Site-specific DNA hypermethylation is frequently implicated as a cancer-causing mechanism, but what of genome-wide hypomethylation, which also occurs in many human tumours? Whether this is a cause or consequence of cancer has long been debated, but Rudolf Jaenisch and colleagues, reporting in the 18 April issue of Science, have now developed a mouse model to address this question.

They generated viable, but small, mice that were compound heterozygous for a hypomorphic allele and a null allele of Dnmt1 — the DNA methyltransferase that maintains DNA methylation. These mice expressed just 10% of wild-type levels of the protein, and Southern blot analysis following digestion with a methylation-sensitive restriction enzyme revealed that global methylation levels were decreased. Interestingly, 80% of the mice developed aggressive T-cell lymphomas at 4–8 months of age. These were shown to be monoclonal, which indicates that hypomethylation initiates cancer in a single cell that undergoes other events to become a malignant tumour.

So how might hypomethylation induce this lymphomagenesis? The authors proposed three possible mechanisms: induction of endogenous retroviral elements could insertionally activate proto-oncogenes; proto-oncogenes could be activated by epigenetic effects; or genomic instability might be induced.

The first possibility was ruled out as retroviral element activation was not observed. c-Myc had already been found to be overexpressed in most of the hypomethylation-induced T-cell lymphomas, but this locus was not rearranged in any of the 18 tumours tested. This was in contrast to Moloney murine leukaemia virus (MMLV)-induced tumours, in which 3 of 12 tumours had an insertional rearrangement in c-Myc. It was also thought unlikely that the mechanism could be epigenetic, as hypomethylation existed throughout development and c-Myc was expressed at normal levels in the thymuses of 2–4-week-old mice.

However, there was already evidence that hypomethylation affected genomic stability, and this was confirmed by carrying out array-based comparative genome hybridization. When hypomethylation-induced tumours were compared with MMLV-induced tumours, a significant increase in chomosome gains — particularly of chromosome 15, which contains c-Myc — was observed. Only 2 of 12 tumours did not have this change, and these also had lower levels of c-Myc.

So, it seems that hypomethylation can cause tumorigenesis through genomic instability. A report by the same group, also in Science, provides further support for this, as they show that hypomethylation promotes cancer in tumour-prone mice — due to heterozygosity of the tumour suppressors Nf1 and Trp53 — because it increases the rate of loss of heterozygosity. Perhaps we should reconsider the use of demethylating agents to treat cancer in light of these results.