As its name suggests, the HIC1 (hypermethylated in cancer 1) gene is hypermethylated and hence transcriptionally silent in several types of cancer, indicating that it might act as a tumour suppressor. Now, Stephen Baylin and colleagues have developed a mouse model to further investigate this possibility, and show that complete loss of Hic1 prediposes mice to a gender-dependent array of tumour types.

The authors generated Hic1+/− mice and investigated the incidence of tumour formation. The heterozygous mice began to develop tumours at 70 weeks, and by 100 weeks 34.2% had developed malignant tumours, compared with 14.2% of the wild-type mice. Surprisingly, the tumour types that were identified differed according to the sex of the mice — male heterozygotes developed carcinomas at an increased frequency, whereas female heterozygotes mostly developed lymphomas and sarcomas. The tumours that developed in Hic1 heterozygotes were also more aggressive than those that developed in wild-type mice.

So, was the remaining wild-type allele lost in these cancers? Gross chromosomal deletions were not observed, so the authors investigated the methylation status of the Hic1 promoter. As in humans, Hic1 has two alternative promoters, 1a and 1b; it is 1b that seems to be hypermethylated in most human cancers. In mice, however, methylation-specific polymerase chain reaction revealed that the 1a promoter is predominantly methylated in most lymphomas and sarcomas. All carcinomas from heterozygous mice that do not have a methylated 1a promoter were found to contain dense methylation of the 1b promoter, and, in total, 87% of malignant tumours from the heterozygous mice had dense methylation of one of the alternative promoters. This promoter methylation corresponded with a lack of Hic1 protein, as it does in human cells.

HIC1 is therefore the first candidate tumour suppressor that is transcriptionally silenced, rather than mutated, and that acts as a tumour suppressor in mice. The identification of other such genes, and their confirmation using mouse models, will no doubt follow.