Favoritism in DNA Methylation Free

Perspective on Candiloro and Dobrovic, p. 862

That some diseases may be caused by epigenetic aberrations induced by underlying sequence changes in the genome has been recognized for a long time. The following examples illustrate such changes: (a) classic “imprintor mutations” caused by microdeletions within imprinting control loci; these microdeletions result in a failure to correctly reset the epigenetic status of imprinted genes on the affected allele in offspring and give rise to congenital imprinted disorders (1, 2); (b) triplet repeat expansion resulting in epigenetic dysregulation via CpG methylation or localized histone modifications in the vicinity of the promoter of the associated gene in neuromuscular degenerative diseases (3); (c) α-Thalassemia caused by epigenetic silencing of the HBA2 gene mediated by antisense transcription upstream through the gene; this effect was initiated by the repositioning of the LUC7L promoter by an interstitial deletion (4); and (d) epimutations of the mismatch repair gene MSH2 in Lynch syndrome due to deletion of the transcription termination signal of the upstream EPCAM gene; this deletion results in failure of transcriptional termination and the generation of EPCAM-MSH2 fusion transcripts (5).

There are less frequent examples of a single nucleotide change (rather than a deletion) that results in the altered epigenetic status of a gene. In families with a predisposition to B-cell chronic lymphocytic leukemia, promoter methylation and transcriptional suppression of the affected allele of the DAPK1 gene is attributable to a germline point mutation upstream of the DAPK1 promoter that results in a higher binding affinity for the HOXB7 transcriptional repressor (6). It may be that transcriptional inhibition results in the accrual of repressive epigenetic modifications including methylation to the DAPK1 promoter in these cases. The choice of X-chromosome inactivation is normally random. X-inactivation (or inactivation of one of the two copies of the X chromosome in females) can be skewed in the somatic cells of XX females with a germline point mutation at a nucleotide position 43 bp upstream of the transcription start site of the Xist gene (−43C>A change; refs. 7, 8). This point mutation is located within the CTCF binding site of the Xist gene and, depending on the nature of the change, results in either the abrogation or increased binding affinity of CTCF. The −43C>A change abolishes CTCF binding with the resultant loss of expression of Xist from this homologue and hence escape from X-inactivation. Conversely, the −43C>G change was shown to increase CTCF binding, inducing Xist expression, which in turn, resulted in preferential inactivation of this homologue. The direct epigenetic effect of these single nucleotide changes on the Xist gene itself was not investigated, but the single nucleotide changes induced gross epigenetic changes to an entire homologue by dictating choice of X-inactivation (9). Although such high-penetrance cases seem to be rare, they nevertheless provide evidence for the close interplay between genetic and epigenetic factors in regulating gene activity. In this issue of the journal, Candiloro and Dobrovic (10) move the field forward by demonstrating a relationship between a single nucleotide polymorphism (SNP) and constitutional methylation of the MGMT gene in healthy individuals. All these examples raise the important possibility that genetic-epigenetic interplay may influence the development of common diseases.

Candiloro and Dobrovic (10) provide the first evidence for a “constitutional methylation epigenotype” for the germline C>T SNP (rs16906252) located within the promoter and 5′ untranslated region of the first exon of the cancer-related MGMT gene. MGMT encodes a base-excision repair gene and is frequently somatically methylated in a variety of cancers, including colorectal cancers, gliomas, and lymphomas, wherein promoter methylation correlates closely with loss of protein expression (11). Recent evidence suggests that loss of MGMT expression is associated with good response rates to alkylating drugs, and this finding may influence present patient treatment choices (12). For these reasons, the mechanism underlying the dysregulation of MGMT is of considerable interest to cancer researchers.

In 2007, Ogino and colleagues (13) advanced our understanding of these mechanisms in finding a close association between methylation of the MGMT promoter and the presence of the relatively common T allele of the germline C>T SNP in a large group of sporadic colorectal cancers from two prospective patient cohorts. Furthermore, MGMT methylation has been found (albeit at lower levels) in the normal colonic mucosa of colorectal cancer patients with corresponding MGMT methylation in tumors, and in the colonic mucosa of individuals without colorectal neoplasia (14_17). It was suggested that this methylation might precede and predispose to neoplastic development resulting from a field defect, or “field cancerization” (originally conceived by Slaughter in 1953), whereby the accrual of genetic alterations in patches of preneoplastic cells underlies the development of cancer (18). Candiloro and Dobrovic (10) extrapolated from these two cogent lines of evidence in hypothesizing that the presence of the T allele of the germline C>T SNP would lead to MGMT promoter methylation in normal somatic tissues. To lend credence to this hypothesis, any positive finding of methylation in normal tissues would need to be distinguishable from that of a “field defect” that predisposes to cancer in that particular tissue.

Therefore, the authors elected to screen for MGMT methylation in peripheral blood mononuclear cells, an easily accessible source of normal cells, from individuals unaffected by cancer. To this end, they used two sensitive, but subtly different, techniques designed to detect low levels of methylation-MethyLight and Sensitive Methylation Analysis after Real-Time Methylation-Specific PCR (SMART-MSP). Their salient findings are that low levels of MGMT methylation were indeed detectable in the peripheral blood of a small proportion of healthy individuals, and this methylation was statistically associated with the T allele of the MGMT SNP. Furthermore, molecular data of these investigators indicated that when MGMT methylation occurred in heterozygous individuals, it occurred on the T allele. Another recent study has shown a similar association between the normal colonic mucosa and tumors of colorectal cancer patients and preferential methylation of the T allele in colorectal tumors (14).

The data of Candiloro and Dobrovic (10) showed minor discrepancies in the frequency and level of MGMT methylation that likely were due to idiosyncrasies of the two detection methods. For example, the methylation-specific probe in MethyLight affords specificity but possibly at the expense of sensitivity. Compared with MethyLight, SMART-MSP has the potential to detect a greater degree of mosaicism of DNA methylation (comprising stretches of methylated DNA interspersed with unmethylated sites) and thus may have a relatively increased sensitivity for detecting methylation at a lesser density. This aspect of SMART-MSP could explain why its detection of methylation in blood was marginally higher than was that of the MethyLight assay. The authors used a further SMART-MSP assay encompassing the germline C>T SNP site within the amplification fragment, followed by sequencing of the amplicons, to show that MGMT methylation was linked to the T allele in heterozygotes. These data may be disputed, however, by invoking the possibility that they merely represent an experimental artifact caused by allelic bias in PCR amplification, although the feasible alternative allelic origins of methylation are limited when methylation levels occur in such a low range (0.1-10%). Still, methylation was detected in only a small number of individuals by either method, including in 12 of 89 people by the more sensitive SMART-MSP, which did not show that methylation occurred exclusively at the T allele. Therefore, further studies in larger cohorts of healthy individuals including additional ethnic groups will be needed before it would be justifiable to conclude that methylation occurs preferentially at the “T” SNP in germline DNA.

Nevertheless, these findings are of significant interest from a mechanistic perspective. The precise mechanism by which the germline C>T SNP renders the promoter susceptible to methylation remains obscure. The germline C>T SNP is located within a cis-acting enhancer element that spans the first exon-intron boundary of MGMT, which is required for efficient promoter activity. Functional studies have shown that deletion of this enhancer element reduces transcriptional activity of the MGMT promoter by 95% (19). The SNP site is located just 24 to 33 bp upstream of a minimal protein binding motif of 9 bp, which binds a transcriptional activator called the MGMT enhancer binding protein, but flanking sequences also were shown to contribute to transcriptional activity (20). The germline C>T SNP plausibly could reduce the binding affinity for the MGMT enhancer binding protein or other transcriptional activator and thus result in down-regulation of transcription. Alternatively, the germline C>T SNP could create a new binding site for a transcriptional repressor, ultimately resulting in the induction of epigenetic modifications of the T allele, including methylation. Regardless of the precise mechanism, it is likely that the epigenetic changes associated with the T allele represent a secondary event. A precedent for this conclusion is provided by interesting new evidence showing that transient down-regulation of gene activity can induce long-term epigenetic silencing (21). Because the germline C>T SNP is exonic, allelic-expression analyses of the MGMT transcript at this site in heterozygous individuals without any detectable methylation would reveal whether the T allele is intrinsically down-regulated and thus whether methylation is potentially the secondary consequence of reduced expression of the T allele per se.

If MGMT methylation in association with the T allele is merely an epigenetic bystander to the constitutional loss of gene activity, its reliability as a potential marker for cancer risk or as a treatment target would be highly questionable. In the Candiloro and Dobrovic study (10), MGMT methylation levels in peripheral blood were low or absent. Even if found to be directly related to the level of gene activity, the methylation levels detected in blood may not represent the levels in other somatic tissues. The epigenetic manifestations associated with this SNP are likely to be more profound in tissue types susceptible to the development of methylated MGMT–related cancers. Somatic mosaicism of methylation is clearly exemplified by MSH2 epimutations induced by EPCAM deletions; MSH2 promoter methylation is either absent or low in blood but abundant in epithelial tissues where the EPCAM gene is highly expressed and the majority of Lynch syndrome cancers arise (5). More concrete information about the functionality of MGMT would come from the genotype of the SNP itself than from the epigenetic manifestations that may be associated with it.

Whether the T allele of this germline SNP is associated with an increased risk of developing methylated MGMT–related cancers and modifies susceptibility to particular therapeutic drugs are questions of significant clinical importance. Candiloro and Dobrovic (10) conclude aptly that it remains to be determined whether the presence of the T allele in germline DNA confers an increased risk of cancer development. This risk association will need to be ascertained in prospective population-based studies. If future studies produce evidence that the T allele at this SNP site does indeed modify the risk of cancer, genotyping at this allele would be a more robust marker of cancer risk than would be the presence or absence of associated methylation. Candiloro and Dobrovic (10) have provided key further evidence of the biological significance of interactions between genetic and epigenetic factors. They also provide a cautionary note about the properties of SNPs themselves, which generally have been assumed not to be associated with epigenetic-related harmful effects. This germline C>T SNP (occurring in about 13% of individuals) may not be benign after all. Rather, in due course, it may prove to be a disease-susceptibility variant of significant clinical importance.

Candiloro and Dobrovic (10) have highlighted the opportunities for using epigenetic markers to identify individuals at risk of cancer and who may be good candidates for epigenetic or other cancer chemoprevention, as well as for screening to reduce cancer mortality. It is well known that cancers are associated with widespread epigenetic changes and that new classes of drugs are now able to induce reexpression of epigenetically inactivated tumor suppressor genes. Epigenetic therapies also can sensitize cancer cells to other treatments. For example, epigenetic therapy in breast cancer cells induces reexpression of the estrogen receptor and sensitizes the cells to antiestrogen therapy (22). Genetic heterogeneity is greater in cancer than in premalignancy, and cancers thus have multiple hits (epigenetic or otherwise) that make them less sensitive and more prone to developing resistance to an intervention; therefore, it is likely that epigenetic therapy will be more successful in treating early (such as prostatic intraepithelial neoplasia or in situ breast cancer) than advanced neoplasia. Further evidence supporting the relevance of epigenetic changes to prevention includes their usual presence before, and effects on, tumor formation (such as in genetic mouse models) and their frequent occurrence in earliest tumor progression (21). Epigenetic therapy also potentially could prevent cancer in individuals with constitutional aberrant DNA methylation and no evidence of neoplasia. The study of epigenetic cancer prevention also is warranted in intraepithelial neoplasias resulting from a clear relationship between environmental factors and epigenetic dysregulation (e.g., Helicobacter pylori infection–related gastric dysplasia).

One caveat to remember is that even when epigenetic treatment successfully reverses repressive epigenetic modifications, the target gene may not reexpress. If promoter methylation associated with the T allele of the MGMT gene has merely occurred as a secondary consequence of prior transcriptional repression (through lack of binding of the transcription factor for example), then treatment with epigenetic drugs may not ultimately result in reactivation of the allele, even if the methylation itself is successfully removed. Additional studies are required to determine the utility of epigenetic therapies for this particular gene. Emerging technology likely will allow us to define study populations in whom epigenetic approaches may be beneficial. A key challenge for these studies will be to develop the means to monitor epigenetic drug effects in subpopulations of cells within the target tissues of at-risk individuals.

No potential conflicts of interest were disclosed.