One of the most important controversies in the field of cancer epigenetics is the question of whether aberrant CpG island methylation of tumour suppressor genes can cause cancer or whether such abnormalities of DNA methylation are secondary to the malignant process. We address this question by prospectively causing abnormal methylation events in vivo. We have produced transgenic mice which over-express the de novo methyltransferase Dnmt3b under the widely expressed CAG promoter. Tg(Dnmt3b)+mice develop normally and are fertile, but die at 4-5 months, developing dilated cardiomyopathy. CpG island methylation is globally increased, and abnormal methylation of specific genes, including the tumour suppressor genes Cdkn1a, Cdkna2a and Hic1 is detectable. However, there is no spontaneous cancer incidence. Crossing the mice with the intestinal tumour prone Apc^+/-min mice leads to a modest increase in polyp number and the frequency of dysplastic tumours, although methylation of the Apc gene is not detectable in normal mucosa from Dnmt3b overexpressing mice, and methylation of the Apc promoter occurs with equal frequency in microdissected tumours from Apc^+/- mice regardless of Tg(Dnmt3b)+ genotype. The results show that increases of methylation can be well tolerated, suggesting that although methylation can be increased, active silencing by methylation is more unusual. In particular, the active silencing of tumour suppressor genes by de novo methylation is unlikely to be a primary event in the formation of tumours, although it may play a role in modulating tumour progression.