Heredity and DNA methylation in colorectal cancer

In their interesting review of colorectal cancer (CRC) as a model for epigenetic tumourigenesis, Wong, et al (in this issue, p 139) discuss the role of hereditary factors in explaining the aetiology of CRC with DNA methylation. Firstly, they assert that larger studies do not support a hereditary aetiology for the CpG island methylator phenotype (CIMP). Secondly, they suggest that MLH1 epimutation (or germline hemi‐allelic methylation) may be heritable. Both of these propositions may be challenged.

In the case of an inherited predisposition to CRCs with acquired DNA methylation or CIMP, a family cancer clinic‐based study that excluded families with the Lynch syndrome, found that members with CRCs showing the CIMP had a 13‐fold increased risk of having a first‐degree relative with cancer (not necessarily CRC) as compared with those without CIMP‐positive CRC.¹ A hospital‐based study by Ward et al² could not confirm this finding, but it is pertinent that they excluded families considered to have hereditary nonpolyposis colorectal cancer (HNPCC). Ward et al regularly use clinical definitions for HNPCC.³ The exclusion of “HNPCC” families is likely to have introduced a major bias, as not all families meeting a clinical definition for HNPCC (eg, the Amsterdam criteria) in fact show evidence of DNA mismatch repair deficiency that would be indicative of a germline defect in a DNA mismatch repair gene.⁴

The second large study cited by Wong et al⁵ was a population‐based series that assessed CIMP in 864 CRCs and defined CIMP‐high as the presence of methylation in at least two of five markers.⁵ This is not a stringent definition of CIMP‐high as indicated by the low frequency of BRAF mutations (32/182, 17.6%) among CIMP‐high/DNA microsatellite stable (MSS) CRCs. As Wong et al point out, an inherent difficulty in establishing whether genetic factors may explain CIMP is the lack of an agreed definition of CIMP. It is clear, however, that mutation of BRAF cosegregates with extensive CIMP⁶ and may therefore be used as a surrogate for high‐level CIMP. In the same large population‐based study group, but now stratified on the basis of BRAF mutation,⁷ the findings were different. In the subset of MSS CRCs, the odds ratio for having a positive family for subjects with BRAF mutation‐positive CRCs was 4.23 (95% confidence interval 1.65 to 10.84) (as compared with subjects with BRAF mutation‐negative CRCs). However, among subjects with high microsatellite instability (MSI‐H) CRCs, a stronger family history of CRC was observed when cancers did not have the BRAF mutation.⁷ This is only to be expected, as subjects with MSI‐H CRCs that lacked BRAF mutations were relatively young and a proportion would have Lynch syndrome.

In 2005, a further paper described a series of 11 Lynch syndrome‐like families, in which some CRCs were MSI‐H but others had low‐level MSI or were MSS.⁸ Both CRCs and polyps in these “MSI‐variable” families showed frequent mutation of the oncogene BRAF or methylation of the CIMP marker MINT31.⁸ Affected subjects also had serrated polyps and two had hyperplastic polyposis. On the basis of these observations, it was suggested that an inherited predisposition to acquired DNA methylation in somatic tissues could give rise to a “serrated pathway syndrome”.⁸ Should MLH1 be implicated, then one might observe the development of CRCs that were MSI‐H. This would only apply to a subset of CRCs, but could by chance, as in the case of a similar serrated pathway syndrome family described in 1997,⁹ affect all CRCs tested in a single family.

Loss of imprinting (LOI) of IGF2 in normal colonic mucosa and normal leucocytes has been associated with a personal and family history of CRC.¹⁰ A study from Japan has shown that IGF2 LOI was more frequent in CRCs with the features of CIMP.¹¹ The link between CIMP and LOI may be explained by methylation of the H19 differential methylated region.¹² These observations provide further evidence for the heritability of CIMP.

Turning to germline hemi‐allelic methylation of MLH1 (or MLH1 epimutation), Gazzoli et al¹³ were the first to identify an early‐onset CRC, in which one MLH1 allele showed methylation. Interestingly, the same allele was also found to be methylated in the subject's lymphocytes. By elegantly exploiting the existence of a common polymorphism in the promoter region of MLH1, Gazzoli et al succeeded in showing that the wild‐type MLH1 allele had been lost in the CRC. On this basis, they introduced the concept of germline hemi‐allelic methylation of MLH1 as a cause of “HNPCC”.¹³ However, they were unconvinced that a methylated allele could be transmitted vertically from parent to child, and concluded that the finding was likely to be both rare and sporadic. Miyakura et al¹⁴ reported four more examples of early‐onset MSI‐H CRC associated with germline hemi‐allelic methylation of MLH1. Although the patients were ascertained through cancer family clinics, they did not have family histories suggestive of Lynch syndrome. They were merely young and some had multiple tumours consistent with Lynch syndrome. Again, Miyakura et al¹⁴ did not imply that germline hemi‐allelic methylation of MLH1 could be transmitted vertically.

A paper from Ward's group arrived at a different conclusion with respect to germline hemi‐allelic methylation or epimutation of MLH1.³ They documented two additional subjects carrying an MLH1 epimutation, who also met clinical criteria indicative of a diagnosis of “HNPCC”. Additionally, the epimutation was present in spermatozoa of one of the affected subjects. The second finding not only fitted with a germline defect but also provided evidence for vertical transmission of the defect. The authors therefore advanced the concept of MLH1 epimutation as a new cause of HNPCC.³ Nevertheless, it may be questioned if epimutations can in fact be inherited. Although germline hemi‐allelic methylation was indeed shown in single members of two families that met certain clinical criteria for HNPCC, this is hardly surprising as the search for the epimutation was conducted exclusively in members of families registered in cancer family clinics. This ascertainment bias aside, it is now clear (as stated above) that when a family happens to meet a particular clinical definition of “HNPCC” this does not automatically prove the existence of an underlying altered DNA mismatch repair gene (the basis for Lynch syndrome).⁴ Although one of the affected subjects indeed showed methylation of MLH1 in spermatozoa, this was in <1% of spermatozoa.³ Should such an affected sperm succeed in fertilising an ovum, subsequent clearance of methylation during early embryogenesis would negate the effects of vertical transmission of the affected allele.

Ward et al subsequently showed the de novo origin of germline hemi‐allelic methylation of MLH1 in a male subject who was shown to have inherited the methylated allele from his mother in whom the same allele was not methylated.¹⁵ These authors nevertheless continued to claim that MLH1 epimutation was “weakly” heritable, although they also contradicted themselves in the same paper by asserting that there was no evidence that MLH1 epimutation could be inherited. Wong et al now cite the four preceding reports on this topic as providing evidence for the heritability of germline epigenetic change.

In summary, the balance of evidence suggests that genetic mechanisms will be found to at least partially explain the evolution of CIMP‐positive CRCs and will account for a subset of families that may mimic Lynch syndrome. On the other hand, there is no evidence to support the inheritance of MLH1 epimutation.