Several kilobases upstream of H19 is an element called the differentially methylated domain (DMD), which has multiple functions and is essential for the imprinting of both H19 and Igf2 [4]. The DMD is methylated on the paternal chromosome, where it is needed for switching off H19, but it is unmethylated on the maternal chromosome, where it is needed for switching off Igf2 (Figure 1).

Transfection of the 3' region of the mouse DMD into Drosophila has shown that it can act as a silencer [5]. Now, work in the mouse [1,6] has shown that this 3' region (see Figure 1a) indeed silences H19 on the methylated paternal chromosome. Thus, a targeted deletion of this silencer region on a chromosome inherited from the father allows H19 to be active, whereas when the same chromosome is inherited from the mother, there is no detectable effect on H19. Surprisingly, the deletion did not affect the parent-of-origin-driven differential methylation of H19 [1], suggesting that the sequences controlling differential methylation are upstream of the deletion.

Another role for the DMD has been found when it is maternally transmitted, and therefore unmethylated. Earlier this year, it was shown that in its unmethylated state the DMD acts as a boundary element - a cis-acting element that blocks the effects of distal enhancers (Figure 1b) [7,8,9,10]. The DMD lies between Igf2 and enhancers downstream of H19 (Figure 1). According to the current model, the unmethylated DMD blocks the interaction of enhancers with the maternal Igf2 promoter to prevent Igf2 expression from the maternal chromosome. By contrast, the methylated DMD does not appear to have enhancer-blocking activity, so the enhancers can access Igf2 on the paternal chromosome, allowing expression of the gene to occur.

The identification of this DMD boundary element appears to provide a satisfactory explanation for the repression of maternal Igf2, but it is clearly not the whole story. New work [2,3] has shown that two further elements can prevent Igf2 expression, and both are silencers. These elements act specifically on Igf2 in mesodermal tissues, independent of any effect on H19. Interestingly, the results of the new work indicate that in the mesoderm the boundary does not form a complete block between enhancer and promoter; it can be crossed in some circumstances.

One of the newly identified elements is another differentially methylated region, DMR1, which is about 1.5 kilobases in size and lies at the 5' end of Igf2. This region is hypermethylated on the paternal chromosome but is unmethylated on the maternal chromosome. It had been hypothesized that when it is unmethylated DMR1 has a silencer function, thus preventing transcription of maternal Igf2 [11,12,13,14]; when methylated, the silencer is envisaged to be no longer functional, allowing transcription of paternal Igf2. Confirmation of this idea has come from Constância and colleagues [3], in functional studies with a mouse lacking DMR1; the 'knockout' region also encompasses a block of direct repeats just upstream of DMR1. When the knockout is maternally inherited (Figure 1c), Igf2 is no longer silent: it is transcribed from the maternal chromosome - there is loss of imprinting resulting in biallelic expression. On paternal transmission, transcription in the embryo was unaffected. An additional finding was that the DMR1 silencer appeared to be specific for Igf2 within mesodermal tissues only, with the exception of muscle; it did not silence Igf2 in endoderm.

The other new Igf2 silencer is specific for muscle [2]. This silencer is a 1 kilobase element located midway between Igf2 and H19 that contains nuclease-hypersensitive sites characteristic of control regions and shows homology with the corresponding human sequence but is not differentially methylated [15], so the nature of the epigenetic mark is unclear. Deletion of this element resulted in reactivation of Igf2 in muscle on maternal transmission, particularly in the tongue [2]. This is of interest in relation to the loss of imprinting of IGF2 that occurs in the human genetic disorder Beckwith Wiedemann syndrome (BWS), which is associated with fetal overgrowth and predisposition to childhood tumors. Enlargement of the tongue is the most consistent feature of BWS, a feature that might correspond to the strong reactivation of Igf2, a potent growth factor, in the mouse tongue following deletion of the muscle-specific silencer.

The model discussed here (Figure 1) for switching off maternally derived Igf2 predicts that the boundary element located between the promoter and enhancers prevents the downstream enhancers from accessing the Igf2 promoters, thus stopping Igf2 transcription. The model is based on the well-known endoderm enhancers but is expected to apply to mesodermal tissues as well, because the mesodermal enhancers, like the endoderm enhancers, are distal to the boundary element (see Figure 1) [3,16,17]. According to the boundary model, maternal Igf2 should be switched off in the DMR1 knockout - but contrary to expectation, it is in fact switched on. So how is maternal Igf2 derepressed in the mesoderm of the knockout? Constância et al. [3] proposed two alternative mechanisms. One is that normally the mesoderm enhancers interact with the unmethylated DMR1 silencer on the maternal chromosome (Figure 1b), so that transcription of Igf2 is prevented. To achieve this, the boundary element would have to be circumvented, at least partially. When DMR1 is knocked out, it is postulated that the mesoderm enhancers activate the Igf2 promoters; there is no DMR1 silencing activity (Figure 1c). The second mechanism proposed by Constância et al. [3] is that DMR1 normally interacts with the boundary element to prevent the downstream enhancers from accessing the promoter (Figure 1b). This proposal implies that in mesoderm the boundary element needs to interact with the silencer to be fully functional; in the DMR1 knockout the interaction is lost, and leads to lifting of the boundary (Figure 1c). Interactions between the muscle enhancer and silencer can also be proposed. Whatever the mechanism, the experiments show that control of expression of imprinted genes can involve interaction between enhancers and epigenetically regulated silencers.

The phenotypic effects that result from deletion of DMR1 are intriguing. An unexpected result was that there was no increase in growth in mice carrying the maternally derived DMR1 knockout, despite the loss of imprinting of Igf2. This is surprising, because in other knockouts where biallelic expression of Igf2 occurred, an increase in size was found; but in these mutants H19 expression was also disrupted [4,18,19]. Future studies of the levels of Igf2 peptide in the mesoderm of DMR1 knockout mice may provide an explanation of these apparently disparate findings. A second phenotypic observation is also worthy of note. On paternal transmission, the DMR1 knockout mice underwent intrauterine growth retardation, so they were small at birth. This was expected because of the design of the knockout construct: a paternal placenta-specific transcript is absent. But both maternal and paternal transmission of the knockout allele led to extended postnatal expression of Igf2 (suggesting that a non-imprinted element controlling the duration of expression lies in the sequence covered by the knockout). The extended postnatal expression of Igf2 enabled the mice that had a paternally derived DMR1 knockout allele and were small at birth to catch up in size with their normal sibs by the time of weaning. The situation in mice may be pertinent to that in humans, where loss of imprinting of IGF2 can occur without altering the imprinting of H19 in hepatoblastoma and in many patients with BWS [20,21,22]. The recent findings now show that loss of imprinting of Igf2 independent of H19 occurs in mice as well. The Igf2 silencers identified in mice add to the ever-increasing number of elements controlling the imprinting of Igf2; they provide additional targets for mutations that can lead to disruption of imprinting, and to diseases including cancer.

Imprinted genes tend to be clustered. Igf2 and H19 are at one end of a much larger cluster of at least 12 genes, covering 1 megabase on mouse distal chromosome 7 and the Beckwith-Wiedemann region on human chromosome 11p15.5. Could there be coordinate control of imprinting within the cluster, alongside control at a local level? There is not a clear answer as yet. Genes at either end of the cluster are imprinted in both species, but those in the central region are not, implying that there is not coordinate control. X-chromosome inactivation provides a precedent for a central mechanism for inactivation, but it is known that some genes can escape X inactivation. Comparative genomic sequence analysis may provide some mechanistic answers, and the first such analysis has recently been carried out: 250 kilobases of the central part of the cluster from the mouse has been compared with the orthologous Beckwith-Wiedemann region in humans [23]. Of six orthologous genes identified which showed extensive structural and functional conservation, three were similarly imprinted in both species. Two of the three remaining gene pairs showed differences in imprinting: Tssc4 is imprinted in mice but not in humans, and LTRPC5 is imprinted in humans but not in mice. A third gene pair Tssc6/TSSC6 is not imprinted in either species. Thus, cross-species conservation of domains will not provide us with a simple means of identifying domains important for imprinting. Nevertheless, imprinted genes within the cluster all contained CpG islands, regions that can be differentially methylated, but the non-imprinted genes did not. CpG islands are much more frequent around imprinted genes than elsewhere in the genome; they may be an essential feature of imprinted domains and indicate the presence of controlling elements. With the recently identified silencers and boundary elements, it is increasingly evident that a wide variety of controlling elements have been drawn into the process of imprinting.