- Email: [email protected]
Imprints on islands
ADRIAN P. BIRD
GENOMIC IMPRINTING
Imprints
on islands
Recent results with parentally imprinted mouse genes suggest that the imprinting mechanism may involve de nova methylation of CpG islands. Diploid organisms carry two copies of every autosomal gene, one from each parent. In the great majority of cases, the two copies are repressed or transcribed identically, but this is not the case for genes that exhibit the phenomenon of parental imprinting. Imprinted genes show markedly different behavi0u.r depending on their parental origin (reviewed in [l]). Fo’r example, the paternally derived insulin-like growth factor gene, I@, of mice is expressed, whereas the maternally derived copy of the gene is repressed [ 21. Conversely, the gene encoding a receptor for Igf2,Z@r, is active when derived from the mother, but inactive when derived from the father [3]. How are apparently identic:al genes in the same nucleus programmed to behave differently? An obvious difference is their history: maternal genes have experienced oogenesis, whereas paternal genes have had the very different experience of spermatogenesis. That, as a result, maternal and paternal diploid embryos are not equivalent is clearly shown by the failure of development of diploid embryos that have two egg genomes or two sperm genomes. The essential questions therefore become: how are genomes marked differentially during gametogenesis and how is this mark maintained on the gene throughout development? This article discusses recent work that promises to lead to answers to these questions. The characteristics of DNA methylation seem tailor-made for imprinting: methylation affects gene expression, usually being associated with transcriptional inactivity; it is potentially heritable, persisting through cycles of cell division; and it is potentially reversible. Although the idea of a link between imprinting and DNA methylation has pervaded the field for some years, decisive experimental evidence has been lacking. Now that authentic imprinted mouse genes are available, several groups have begun to assess critically the involvement of methylation in imprinting. It is still early days, but there are already intriguing results. In particular, a recent paper by Stager and co-workers [4] provid,es the first glimmerings of etidence that methylation may, in some cases at least, be the imprint itself. In order to appreciate the evidence, it is necessary to know that, with respect to methylation, there are two sorts of DNA in the mammalian genome. In most of the genome (98%), the dinucieotide sequence CpG is heavily methylated and occurs only rarely, at about one quarter of its expected frequency; in the remaining 2 % of the genome, however, CpG is non-methylated and 0
Current
Biology
occurs at its statistically expected frequency. Within this 2 %, the sequence is distributed as discrete ‘CpG islands’, which are usually about 1 kb long and are often found at the 5’ ends of genes 151.A crucial property, in the present context, is that the absence of methylation at CpG islands is, in almost all cases, stable. In other words, an island is methylation-free whether the associated gene is transcriptionally active or inactive. A major exception to this generalization is the one inactivated X chromosome of the pair in somatic tissues of female placental mammals. CpG islands on this chromosome become heavily methylated soon after the start of X-inactivation, and the stability of inactivation seems to be considerabiy enhanced as a result [6]. Indeed, there is ample evidence that methylation of promoter-containing CpG islands leads inescapably to silencing of the associated gene [7], The first three imprinted genes to be isolated, Z@r, I&$?, and HlY, all have CpG islands associated with their promoters. The obvious question, therefore, is whether these islands become methylated on the imprinted, inactive allele. The answer is that they do, but not always. Sasaki and co-workers [S] used mice in which both copies of chromosome 7 were maternally-derived to study the methylation status of the imprinted Z@ gene, which is on chromosome 7. In this case, not only is the CpG island at the Imprinted promoter unmethylated, but there is a DNase l-hypersensitive site at the promoter. In these respects, there iS no obvious difference between the active and Inactive alleles. More surprising was the tinding that a low level of transcription from the imprinted Z@ gene could be detected. It appears that the imprinted promoter is set up and ready to go, but something, as yet unknown, prevents efficient transcription. Could that unknown ‘something’ be the ZZl9 gene [9] ? It is only 90 kb away from the Z@ gene and is oppositely imprinted: it is active on the maternal but inactive on the paternal chromosome 7. HZ9 is a curious gene that apparently lacks an open reading frame, in the mouse at least, and is lethal when over-expressed. If expression of HlY were to be incompatible with expression of Z@, then imprinting of both genes could, in theory, be accomplished by regulation of HI9 alone. Ferguson-Smith and co-workers [lo] have analysed the methylation status of HlY. They find that its 5’ CpG island is, in fact, methylated when derived from the sperm, but non-methylated when derived from the egg. Methylated (inactive) and non-methylated (active) forms of the HZ9 CpG island therefore coexist in cells of the embryo. The available precedents strongly 1993, Vol 3 No 5
275
276
Current
Biology 1993, Vol 3 No 5
suggest that high-density methylation of the HI9 island will ensure stable repression of the paternally-derived gene.
the active, maternal island is not. Once again, methylation is acquired at the post-implantation stage, and is absent in the sperm.
If proven, this may solve part of the imprinting conundrum, namely how Lmprinting is rendered stable. It does not, however, prove that methylation is the original im print. To get at this question, it is necessary to trace back through development the methylation status of the paternal CpG island, to its origin in the sperm. When FergusonSmith et al. [lo] did this, they found that, in common with all other known CpG islands, the HI9 island is not methylated in sperm. In other words, methylation of this island is not itself the imprint. The result is reminiscent of X-chromosome inactivation, in which CpG island methylation occurs in the early embryo to stabilize inactivation of X-linked genes, but is not itself the primary inactivation mechanism. It is as though the message “this CpG island contains an inactive promoter” triggers de nova methylation of the island at this stage of embryogenesis, leading to stable silencing of the associated gene.
Thus, two out of two paternally imprinted genes show de no210embryonic methylation of their promoter CpG islands. In both, methylation is early, but not early enough to be the molecular basis of the imprint itself. It is important to stress the novelty of these findings. Until recently, the only examples of CpG islands that were known to become methylated de nova during development were those on the inactive X chromosome. These two genes therefore join an exclusive group, in which long-term repression seems to be ensured by high-density promoter methylation.
Meanwhile, on chromosome 17, analysis on the I@rgene yielded further evidence relating CpG island methylation to imprinting. Some of these results reinforce those obtained for Hl9, but others are distinctly new. Stiiger and colleagues [4] have located two clear CpG islands in the Ig$?r gene, one covering the promoter and one located 21 kb downstream within intron 1 (Fig. 1). Methylation of the promoter island exactly parallels that of the Hl9 island; that is, the silent, paternal island is methylated, but
!
;
“’
” _‘ ,z
:
\’
The downstream CpG island in the &$?r gene tells a story that is different in two respects [4]. Firstly, this island is methylated on the active maternal allele, but not on the inactive paternal allele. The island is some distance downstream of the promoter, but it is nevertheless surprising that its methylation correlates inversely with expression. The most striking result, however, is that methylation of the island can be traced back, uninterrupted, through morula and blast& stages of development, to the oocyte (Fig. 1). This patch of localized methylation might therefore constitute the oocyte-specific imprint. Tantalizingly, there is as yet no obvious link between this island and the functioning of the I@r gene. Unlike the promoter islands discussed above, the downstream island does not itself lie within the promoter for any obvious transcript.
V Morula
Fig. 1. Schematic summary of changes in-methylation at the paternally irnprinted IgfZrgene in early embryogenesis of mouse, as established by StGger and co-tiorkers W. The darker barrels represent three notional exons; the precise number and position of exons in this gene are not yet known. Ellipsoids RI and R2 represent the promoter and downstream CpG islands, respectively. R2 is known to be about 21 kb downstream of RI in the first intron of the gene. Both islands are approximately 1 kb in length 141. Green CpC islands are non-methylated, and purple are methylated. The curly arrow indicates transcription from the maternallyderived Igf2r gene in post-implantation embryos.
DISPATCH
The behaviour of the downstream island of Z@r is reminiscent of that of the arl&ial TGA transgene studied by Chaillet and co-workers [ 111. Here again, methylation was acquired by the maternal copy of the gene in the egg, and wzs t~smitted faithfully through embryonic stages of development. The sperm copy of the transgene remained hypomethylated.
expect that we will get this, and with it perhaps, a deeper understanding of DNA methylation.
References 1. 2.
Although it is still early for generalizations, there is already a case to be made that methylation of CpG islands is a component of the imprinting process. There seem to be two stages at which methylation occurs de nova late oogenesis, as in the case of the TGA transgene and the downstream island of Z@c or during post-implantation embryonic development, as in the case of the islands on the inactive X chromosome, paternal HZ9 gene and Z&?r promoter. In the Iatter case, methylation may be important for maintenance of the imprint in somatic cells, but it cannot itself be the original imprint. Cases of methylation de novo during oogenesis, on the other hand, may represent the molecular imprint itself. Indeed, it may be that methylation can occur & nova during oogenesis but not during spermatogenesis This might explain why the vast majority of imprinted transgenes acquire methylation only in the maternal germline. Silencing of a patemallyinherited gene might, in this case, be accomplished indirectly, by inactivating a maternal gene, which in turn represses the paternally-derived gene on the same chromosome (that is, in cis>.The downstream CpG island of Zg$& may be such a cis metier. Now that the journey from genetics to genes is complete, the process of paternal imprinting is ready for functional dissection in the mouse. Over the next few years, we can expect more imprinted genes to be found, imprinted loci to be more extensively sequenced, and detailed methylation maps to be generated. The ultimate aim, of course, is a complete explanation of the molecular biology of parental imprinting. It is not too much to
MONK M, S~JRANI A (ELX): Genomic hnprinting. Dev&pment Suppkment Cambridge: The Company of Biologists Ltd; 1990. DECHIARA TM, ROBERSON FI, EFSTRUUDIS A: Parental imprinting
of the mouse insulin-like growth factor type 2 gene. Cell 1991, 3.
64:84%359. BARLOW DP, STOGER R, HERRMANN BG, &TO’ K, SCHWELFERN: The
mouse insulin-like growth factor type 2 receptor is imprinted and closely linked to the Tme locus. Nature 1991, 34984-87. 4.
STOGER R, KUBICKA P, LIU C-G, m T, RAZIN A, CEDER H, BARLOW DP: Maternal specific methylation of the imprinted
mouse ZgtZr locus identities the expressed locus as carrying the imprinting signal. Cell 1993, in press. BIRD AP: CpG islands and the function of DNA methylation. Nature 1986, 321:209213. RIGGS AD, PFEIFFER GD; X-chromosome
inactivation and cell memory. Trh Gemt 1992, 8:169-173. BUD AP: The essentials of DNA methylation. Cell 1992, 70~5-8. &AKI
H, JONES PA, CHAULET JR, FERGUSON-SIMITH AC, BARTON
SC, REM W, SuRANI MA Parental imprinting: potentially active chromatin of the repressed maternal allele of the mouse insulin-like growth factor II (I$$?) gene. Genes Develop 1992, 9.
6:184$1856. BARTOIOMEI MS, ZEMEL s, TILGHMAN SM: Parental imprinting
of
the mouse HI9 gene. Nature 1991, 351:153-155. 10.
11.
FERGUSON-Sm
AC,
&XI
H,
CATTANACH BM,
SURANI MA:
Parental origin specitic epigenetic modification of the mouse I319 gene. Nature 1993, in press. CHAILLE’~ JR, VOGT TF, BEIER DR, LEDER P:Parental-specific methylation of an imprinted transgene is established during gametogcnesis and progressively changes during embryogenesis. Cell 1991, 66:77-83.
Adrian P Bird, Institute of Cell and Molecular Biology, University of Edinburgh, Kings Buildings, Edinburgh EH93JK UK.
277
GENOMIC IMPRINTING
Imprints
on islands
Recent results with parentally imprinted mouse genes suggest that the imprinting mechanism may involve de nova methylation of CpG islands. Diploid organisms carry two copies of every autosomal gene, one from each parent. In the great majority of cases, the two copies are repressed or transcribed identically, but this is not the case for genes that exhibit the phenomenon of parental imprinting. Imprinted genes show markedly different behavi0u.r depending on their parental origin (reviewed in [l]). Fo’r example, the paternally derived insulin-like growth factor gene, I@, of mice is expressed, whereas the maternally derived copy of the gene is repressed [ 21. Conversely, the gene encoding a receptor for Igf2,Z@r, is active when derived from the mother, but inactive when derived from the father [3]. How are apparently identic:al genes in the same nucleus programmed to behave differently? An obvious difference is their history: maternal genes have experienced oogenesis, whereas paternal genes have had the very different experience of spermatogenesis. That, as a result, maternal and paternal diploid embryos are not equivalent is clearly shown by the failure of development of diploid embryos that have two egg genomes or two sperm genomes. The essential questions therefore become: how are genomes marked differentially during gametogenesis and how is this mark maintained on the gene throughout development? This article discusses recent work that promises to lead to answers to these questions. The characteristics of DNA methylation seem tailor-made for imprinting: methylation affects gene expression, usually being associated with transcriptional inactivity; it is potentially heritable, persisting through cycles of cell division; and it is potentially reversible. Although the idea of a link between imprinting and DNA methylation has pervaded the field for some years, decisive experimental evidence has been lacking. Now that authentic imprinted mouse genes are available, several groups have begun to assess critically the involvement of methylation in imprinting. It is still early days, but there are already intriguing results. In particular, a recent paper by Stager and co-workers [4] provid,es the first glimmerings of etidence that methylation may, in some cases at least, be the imprint itself. In order to appreciate the evidence, it is necessary to know that, with respect to methylation, there are two sorts of DNA in the mammalian genome. In most of the genome (98%), the dinucieotide sequence CpG is heavily methylated and occurs only rarely, at about one quarter of its expected frequency; in the remaining 2 % of the genome, however, CpG is non-methylated and 0
Current
Biology
occurs at its statistically expected frequency. Within this 2 %, the sequence is distributed as discrete ‘CpG islands’, which are usually about 1 kb long and are often found at the 5’ ends of genes 151.A crucial property, in the present context, is that the absence of methylation at CpG islands is, in almost all cases, stable. In other words, an island is methylation-free whether the associated gene is transcriptionally active or inactive. A major exception to this generalization is the one inactivated X chromosome of the pair in somatic tissues of female placental mammals. CpG islands on this chromosome become heavily methylated soon after the start of X-inactivation, and the stability of inactivation seems to be considerabiy enhanced as a result [6]. Indeed, there is ample evidence that methylation of promoter-containing CpG islands leads inescapably to silencing of the associated gene [7], The first three imprinted genes to be isolated, Z@r, I&$?, and HlY, all have CpG islands associated with their promoters. The obvious question, therefore, is whether these islands become methylated on the imprinted, inactive allele. The answer is that they do, but not always. Sasaki and co-workers [S] used mice in which both copies of chromosome 7 were maternally-derived to study the methylation status of the imprinted Z@ gene, which is on chromosome 7. In this case, not only is the CpG island at the Imprinted promoter unmethylated, but there is a DNase l-hypersensitive site at the promoter. In these respects, there iS no obvious difference between the active and Inactive alleles. More surprising was the tinding that a low level of transcription from the imprinted Z@ gene could be detected. It appears that the imprinted promoter is set up and ready to go, but something, as yet unknown, prevents efficient transcription. Could that unknown ‘something’ be the ZZl9 gene [9] ? It is only 90 kb away from the Z@ gene and is oppositely imprinted: it is active on the maternal but inactive on the paternal chromosome 7. HZ9 is a curious gene that apparently lacks an open reading frame, in the mouse at least, and is lethal when over-expressed. If expression of HlY were to be incompatible with expression of Z@, then imprinting of both genes could, in theory, be accomplished by regulation of HI9 alone. Ferguson-Smith and co-workers [lo] have analysed the methylation status of HlY. They find that its 5’ CpG island is, in fact, methylated when derived from the sperm, but non-methylated when derived from the egg. Methylated (inactive) and non-methylated (active) forms of the HZ9 CpG island therefore coexist in cells of the embryo. The available precedents strongly 1993, Vol 3 No 5
275
276
Current
Biology 1993, Vol 3 No 5
suggest that high-density methylation of the HI9 island will ensure stable repression of the paternally-derived gene.
the active, maternal island is not. Once again, methylation is acquired at the post-implantation stage, and is absent in the sperm.
If proven, this may solve part of the imprinting conundrum, namely how Lmprinting is rendered stable. It does not, however, prove that methylation is the original im print. To get at this question, it is necessary to trace back through development the methylation status of the paternal CpG island, to its origin in the sperm. When FergusonSmith et al. [lo] did this, they found that, in common with all other known CpG islands, the HI9 island is not methylated in sperm. In other words, methylation of this island is not itself the imprint. The result is reminiscent of X-chromosome inactivation, in which CpG island methylation occurs in the early embryo to stabilize inactivation of X-linked genes, but is not itself the primary inactivation mechanism. It is as though the message “this CpG island contains an inactive promoter” triggers de nova methylation of the island at this stage of embryogenesis, leading to stable silencing of the associated gene.
Thus, two out of two paternally imprinted genes show de no210embryonic methylation of their promoter CpG islands. In both, methylation is early, but not early enough to be the molecular basis of the imprint itself. It is important to stress the novelty of these findings. Until recently, the only examples of CpG islands that were known to become methylated de nova during development were those on the inactive X chromosome. These two genes therefore join an exclusive group, in which long-term repression seems to be ensured by high-density promoter methylation.
Meanwhile, on chromosome 17, analysis on the I@rgene yielded further evidence relating CpG island methylation to imprinting. Some of these results reinforce those obtained for Hl9, but others are distinctly new. Stiiger and colleagues [4] have located two clear CpG islands in the Ig$?r gene, one covering the promoter and one located 21 kb downstream within intron 1 (Fig. 1). Methylation of the promoter island exactly parallels that of the Hl9 island; that is, the silent, paternal island is methylated, but
!
;
“’
” _‘ ,z
:
\’
The downstream CpG island in the &$?r gene tells a story that is different in two respects [4]. Firstly, this island is methylated on the active maternal allele, but not on the inactive paternal allele. The island is some distance downstream of the promoter, but it is nevertheless surprising that its methylation correlates inversely with expression. The most striking result, however, is that methylation of the island can be traced back, uninterrupted, through morula and blast& stages of development, to the oocyte (Fig. 1). This patch of localized methylation might therefore constitute the oocyte-specific imprint. Tantalizingly, there is as yet no obvious link between this island and the functioning of the I@r gene. Unlike the promoter islands discussed above, the downstream island does not itself lie within the promoter for any obvious transcript.
V Morula
Fig. 1. Schematic summary of changes in-methylation at the paternally irnprinted IgfZrgene in early embryogenesis of mouse, as established by StGger and co-tiorkers W. The darker barrels represent three notional exons; the precise number and position of exons in this gene are not yet known. Ellipsoids RI and R2 represent the promoter and downstream CpG islands, respectively. R2 is known to be about 21 kb downstream of RI in the first intron of the gene. Both islands are approximately 1 kb in length 141. Green CpC islands are non-methylated, and purple are methylated. The curly arrow indicates transcription from the maternallyderived Igf2r gene in post-implantation embryos.
DISPATCH
The behaviour of the downstream island of Z@r is reminiscent of that of the arl&ial TGA transgene studied by Chaillet and co-workers [ 111. Here again, methylation was acquired by the maternal copy of the gene in the egg, and wzs t~smitted faithfully through embryonic stages of development. The sperm copy of the transgene remained hypomethylated.
expect that we will get this, and with it perhaps, a deeper understanding of DNA methylation.
References 1. 2.
Although it is still early for generalizations, there is already a case to be made that methylation of CpG islands is a component of the imprinting process. There seem to be two stages at which methylation occurs de nova late oogenesis, as in the case of the TGA transgene and the downstream island of Z@c or during post-implantation embryonic development, as in the case of the islands on the inactive X chromosome, paternal HZ9 gene and Z&?r promoter. In the Iatter case, methylation may be important for maintenance of the imprint in somatic cells, but it cannot itself be the original imprint. Cases of methylation de novo during oogenesis, on the other hand, may represent the molecular imprint itself. Indeed, it may be that methylation can occur & nova during oogenesis but not during spermatogenesis This might explain why the vast majority of imprinted transgenes acquire methylation only in the maternal germline. Silencing of a patemallyinherited gene might, in this case, be accomplished indirectly, by inactivating a maternal gene, which in turn represses the paternally-derived gene on the same chromosome (that is, in cis>.The downstream CpG island of Zg$& may be such a cis metier. Now that the journey from genetics to genes is complete, the process of paternal imprinting is ready for functional dissection in the mouse. Over the next few years, we can expect more imprinted genes to be found, imprinted loci to be more extensively sequenced, and detailed methylation maps to be generated. The ultimate aim, of course, is a complete explanation of the molecular biology of parental imprinting. It is not too much to
MONK M, S~JRANI A (ELX): Genomic hnprinting. Dev&pment Suppkment Cambridge: The Company of Biologists Ltd; 1990. DECHIARA TM, ROBERSON FI, EFSTRUUDIS A: Parental imprinting
of the mouse insulin-like growth factor type 2 gene. Cell 1991, 3.
64:84%359. BARLOW DP, STOGER R, HERRMANN BG, &TO’ K, SCHWELFERN: The
mouse insulin-like growth factor type 2 receptor is imprinted and closely linked to the Tme locus. Nature 1991, 34984-87. 4.
STOGER R, KUBICKA P, LIU C-G, m T, RAZIN A, CEDER H, BARLOW DP: Maternal specific methylation of the imprinted
mouse ZgtZr locus identities the expressed locus as carrying the imprinting signal. Cell 1993, in press. BIRD AP: CpG islands and the function of DNA methylation. Nature 1986, 321:209213. RIGGS AD, PFEIFFER GD; X-chromosome
inactivation and cell memory. Trh Gemt 1992, 8:169-173. BUD AP: The essentials of DNA methylation. Cell 1992, 70~5-8. &AKI
H, JONES PA, CHAULET JR, FERGUSON-SIMITH AC, BARTON
SC, REM W, SuRANI MA Parental imprinting: potentially active chromatin of the repressed maternal allele of the mouse insulin-like growth factor II (I$$?) gene. Genes Develop 1992, 9.
6:184$1856. BARTOIOMEI MS, ZEMEL s, TILGHMAN SM: Parental imprinting
of
the mouse HI9 gene. Nature 1991, 351:153-155. 10.
11.
FERGUSON-Sm
AC,
&XI
H,
CATTANACH BM,
SURANI MA:
Parental origin specitic epigenetic modification of the mouse I319 gene. Nature 1993, in press. CHAILLE’~ JR, VOGT TF, BEIER DR, LEDER P:Parental-specific methylation of an imprinted transgene is established during gametogcnesis and progressively changes during embryogenesis. Cell 1991, 66:77-83.
Adrian P Bird, Institute of Cell and Molecular Biology, University of Edinburgh, Kings Buildings, Edinburgh EH93JK UK.
277