REVIEWS

GENOMIC IMPRINTING: PARENTAL INFLUENCE ON THE GENOME

Wolf Reik* and Jrn Walter
Genomic imprinting affects several dozen mammalian genes and results in the expression of those genes from only one of the two parental chromosomes. This is brought about by epigenetic instructions imprints that are laid down in the parental germ cells. Imprinting is a particularly important genetic mechanism in mammals, and is thought to influence the transfer of nutrients to the fetus and the newborn from the mother. Consistent with this view is the fact that imprinted genes tend to affect growth in the womb and behaviour after birth. Aberrant imprinting disturbs development and is the cause of various disease syndromes. The study of imprinting also provides new insights into epigenetic gene modification during development.
EUTHERIANS

Mammals that give birth to live offspring (viviparous) and possess an allantoic placenta.

*Laboratory of Developmental Genetics and Imprinting, Developmental Genetics Programme, The Babraham Institute, Cambridge CB2 4AT, UK. Max-PlanckInstitut fr Molekulare Genetik, Ihnestr. 73, 14195 Berlin and Universitt des Saarlandes, Genetik, 66041 Saarbrcken, Germany. e-mails: wolf.reik@bbsrc.ac.uk; walter@molgen.mpg.de

Genomic imprinting in mammals was discovered in the early 1980s as a result of two types of mouse experiment. Nuclear transplantation was used to make embryos that had only one of the two sets of parental chromosomes (uniparental embryos) and other sophisticated genetic techniques were used to make embryos that inherited specific chromosomes from one parent only (uniparental disomy). In both cases, the surprising finding was that mammalian genes could function differently depending on whether they came from the mother or the father16. The early 1990s then saw the discovery of the first imprinted genes, which were indeed expressed differently on maternal and paternal chromosomes79, and the realization that imprinting had a substantial effect on human genetic disease10,11. It was also found that DNA methylation was a key molecular mechanism of imprinting; methylation marks the imprinted genes differently in egg and sperm, and inheritance of these epigenetic marks leads to differential gene expression1217. Substantial progress has been made in our understanding of imprinting in the past few years: important phenotypic effects of imprinted genes have been discovered, particularly in the control of fetal growth and behaviour after birth; a number of cis-acting sequences are being defined that are important for the control of imprinted gene expression; the evolutionary under-

standing of imprinting and its likely biological purposes is increasing18,19; and the study of imprinting is providing general insights into the importance of epigenetic mechanisms in development. Here we review these recent developments. We begin with a brief summary of imprinted genes, then look at what is known about establishment and maintenance of imprints, and the important role of the germ line. We review the various reading mechanisms that convert the imprint into differential gene expression. We discuss the evolution of imprinting, and its main phenotypic effects, in healthy and diseased states. Finally, we consider the effect of imprinting on important general issues in epigenetics, such as cloning and genome reprogramming.
Imprinted genes

Using several approaches (BOX 1), around 45 imprinted genes have so far been identified in the mouse (see the Harwell imprinting web site for up-to-date statistics on imprinted gene numbers and characteristics). Some of these genes have been tested in other mammals and for many (but not all), the imprinting status is conserved in humans, in some other EUTHERIAN mammals and in a marsupial2022 (but only a few genes have been tested). What are the genetic and epigenetic features that characterize imprinted genes?

NATURE REVIEWS | GENETICS

VOLUME 2 | JANUARY 2001 | 2 1

2001 Macmillan Magazines Ltd

REVIEWS
genes, so these features cannot be used in a systematic search for new imprinted genes. The great majority of imprinted genes examined so far show differences in DNA methylation between the parental alleles (FIG. 2), but the differentially methylated regions (DMRs) can have different properties. For example, the differential DNA methylation in some DMRs is introduced in parental germ cells and maintained in all developmental stages and tissues 2528. Others show considerable changes in methylation during development and acquire tissuespecific methylation patterns29, which can be associated with tissue-specific imprinted expression. Some DMRs are methylated in the inactive gene copy, whereas others are methylated in the active one. Imprinted genes can also differ with respect to bulk chromatin structure, as well as with respect to more specific modifications, such as histone acetylation3036 (R. Feil and R. Gregory, personal communication). Two other epigenetic features have been discovered that might reflect the larger-scale organization of imprinted genes into clusters or domains. First, it has been shown that the DNA in imprinted regions replicates asynchronously in the S phase of the cell cycle; for most imprinted regions, the paternal copy replicates earlier than the maternal one37,38. Because maternally and paternally expressed genes are interspersed in some regions, this is not likely to be a gene-specific property and its molecular basis is not understood. Second, different frequencies of meiotic recombination are found in or near to imprinted clusters, with an elevated recombination rate during male meiosis39,40. How these regional epigenetic features are linked with methylation and chromatin structure is not known. The precise nature of the primary imprint and its fate during development is still a mystery, but it is likely that all the above epigenetic modifications are relevant to imprinting. However, at present there is no direct evidence that histone or other chromatin modifications have roles in imprinting that are independent of DNA methylation. Indeed, the importance of DNA methylation, at least in the maintenance of imprints, has been clearly established genetically17. For the most part, we therefore equate imprints with methylation imprints or differential methylation to simplify the discussion. Imprinted expression is then a result of the reading of the imprint in somatic tissues.
The life cycle of imprints

Box 1 | Finding imprinted genes

Monoparental disomies Expression screen Subtraction

cDNA

Methylation screen Subtraction

DNA fragment

Methylation

Imprinted genes have been identified in various ways: by chance (usually knockouts that then showed parent-specific expression); based on position (next to other imprinted genes or in a chromosome region associated with an imprinting phenotype); or by using two types of systematic screen. In both screens, embryos are used that have a duplication of one of the parental chromosomes or genomes, together with embryos that have the opposite parental chromosome duplicated. This results in gene expression or methylation in one type of embryo but not the other if the gene is imprinted. The first screen is based on subtraction of cDNAs between such uniparental embryos115. The second is based on methylation differences. One approach using restriction landmark genome scanning (by two-dimensional electrophoresis of DNA) has estimated that there are roughly 100 imprinted genes in the mouse genome116. Another methylation screen uses representational difference analysis117(RDA). The estimate of 100 imprinted genes in the genome is likely to be an underestimate but, in any event, imprinted genes constitute a minority of all the genes in the genome.

CPG ISLAND

DNA region of >500 bp that has a high CpG density and is usually unmethylated. CpG islands are found upstream of many mammalian genes.

One remarkable and characteristic feature of imprinted genes is that they are rarely found on their own: around 80% are physically linked in clusters with other imprinted genes (FIG. 1). The clustered organization of imprinted genes is thought to reflect coordinated regulation of the genes in a chromosomal domain. By analogy to X-chromosome inactivation in which an Xinactivation centre controls the inactivation of the entire chromosome, imprinting centres or imprinting control elements (ICs) have been discovered in some clusters. These ICs are needed for the regional control of imprinting or imprinted expression. No common features are recognizable when comparing the protein sequences encoded by imprinted genes, although there are functional relationships between some proteins with roles in fetal growth and development. Furthermore, two general features of the DNA sequence environment of imprinted genes have been noted. First, they are unusually rich in CPG 23 ISLANDS : around 88% of mouse imprinted genes have CpG islands, compared with the average figure of 47%. Second, clustered, direct repeats are common near to or within the CpG islands. The repeats might or might not belong to one of the known repeat families and they have been proposed to be involved in conferring or maintaining differential methylation24. Neither the repeats nor the CpG islands are unique to imprinted

Genomic imprints change in characteristic ways during the life cycle of the organism (FIG. 3). Imprints are established during the development of germ cells into sperm or eggs. After fertilization, they are maintained as chromosomes duplicate and segregate in the developing organism. In the germ cells of the new organism, imprints are erased at an early stage. This is followed by establishment again at a later stage of germ-cell development, thus completing the imprinting cycle. In somatic cells, imprints are maintained and are modified during development. For example,

22

| JANUARY 2001 | VOLUME 2

www.nature.com/reviews/genetics

2001 Macmillan Magazines Ltd

REVIEWS
a
Mouse distal chromosome 7 Human chromosome 11p15.5

b
Mouse central chromosome 7 Human chromosome 15q11q13

TEL Obph1 Tnfrh1 Tnfrh2 Cars Nap1l4 Tssc3/Ipl Tssc5/Impt1 ? Cdkn1c/p57kip2 ? IC2 Kcnq1ot1/Kvlqt1-as Kcnq1/Kvlqt1 Ltrpc5/mtr1 Tssc4 CD81/Tapa Tssc6 Mash2 Th Ins2 Igf2as Igf2 IC1 H19 Rpl23 Nttp1 CEN

CEN OBPH1 ? ? CARS NAP1L4 TSSC3/IPL TSSC5/IMPT1/ORCT2L ORCTL2s CDKN1C/P57KIP2 BWRT

TEL

CEN

Znf127/Znf127as Ndn Magel2

ZNF127/ZNF127-AS/MKNR3 NDNL1 MAGEL2

C15ORF2

KCNQ1OT1/KVLQT1-AS/LIT1 KCNQ1/KVLQT1 LTRPC5/MTR1 TSSC4 CD81/TAPA TSSC6 ASCL2/HASH2 TH INS IGF2-AS IGF2 H19 RPL23L NTTP1 TEL

IC Snurf-Snrpn Ipw

SNURF-SNRPN IPW

Ube3a

UBE3A/UBE3A-AS

Gabrb3 Gabra5 Gabrg3

GABRB3 GABRA5 GABRG3

CEN

TEL

Figure 1 | Imprinting clusters in human and mouse genomes. Human chromosomes a | 11p15.5 and b | 15q11-q13 and orthologous clusters on a | mouse chromosome distal 7 and b | central 7. The relative location and transcriptional orientation of genes are indicated by arrows. The imprinting status is shown in red (maternally expressed), blue (paternally expressed), black (biallelic expression) and green (imprinted expression not known or not yet precisely defined). Question marks (?) indicate that the orthologues of the mouse or human genes, respectively, are not known. The drawings are not to scale. The BeckwithWiedemann (BWS) cluster (a) comprises about 1 Mb, and the PraderWilli syndrome/Angelman syndrome (PWS/AS) cluster (b) roughly 2 Mb. Imprinting centres (IC) are marked by circles coloured according to the parental origin of the imprint.

methylation may spread from an IC into the promoter. The imprints are eventually read, resulting in parent-specific gene expression. Erasure. The germ line has the role of resetting imprints such that in mature gametes they reflect the sex of that germ line. For most imprints, current evidence indicates that there might be two stages for this resetting process the first one is erasure. This is followed later by establishment. During erasure, there is marked and apparently genome-wide demethylation in germ cells, which is completed by embryonic day 1213 (E1213) in both sexes41,42 (FIG. 4). Indeed, germ cells cultured from these stages (EG cells) have a dominant demethylating activity when fused with somatic cells43; whether this demethylation is active or passive is not

known. The evidence so far indicates that all methylation imprints probably become erased at this stage41,42,44,45. This is important because it implies that imprints inherited from a parent with the same sex as the developing embryo are erased and are unlikely to persist unchanged. There is preliminary evidence that methylation imprints are still present and may be functionally intact before the erasure stage46. After erasure, functional evidence from nuclear transplantation experiments with both male and female germ-cell nuclei indicates that imprints have indeed been substantially altered47,48; expression of imprinted genes in these reconstituted embryos reflects their lack of methylation (for example, H19 is expressed and Igf2 is not expressed). In some instances, this has led to interest-

NATURE REVIEWS | GENETICS

VOLUME 2 | JANUARY 2001 | 2 3

2001 Macmillan Magazines Ltd

REVIEWS
Establishment. After erasure, de novo methylation begins in both germ lines at late fetal stages, and continues after birth41,50 (FIGS 3,4). Oocytes are in meiotic arrest and methylation occurs during their growth47, whereas during spermatogenesis, methylation occurs before meiosis44,45. Nuclear transplantation experiments indicate that this DNA methylation coincides roughly with the acquisition of functional imprints both for autosomal genes and for X chromosome imprinting, at least in oocytes47,51. It is not yet clear which enzymes are responsible for de novo methylation in germ cells (BOX 2). Dnmt1 (DNA methyltransferase 1) and its germ-cellspecific isoforms are candidates52, but it is also possible that Dnmt3a or Dnmt3b, which are required for de novo methylation in postimplantation embryos53, carry out this function in germ cells. It is also unclear how Dnmts specifically target DMRs in either female or male germ cells. DMRs in imprinted genes might be specifically targeted for de novo methylation in one of the germ lines. It is equally possible that there is general de novo methylation in both germ lines and that DMRs are specifically protected from methylation in one germ line but not in the other. In either case, this would require factors that recognize DMRs and that are germline-specific. The existence of such factors is supported by the observation that deficiency of Dnmt1 causes loss of imprints postzygotically, and the imprints cannot be restored by Dnmt1, Dnmt3a or Dnmt3b54. DMRs are generally CpG rich and often fulfil the criteria for CpG islands (see below). However, autosomal CpG islands do not become methylated de novo. So it is likely that imprinted DMRs are genetically or epigenetically modified so that de novo methylation can occur. Genetic modification has been previously postulated to be due to stretches of unique direct repeats that often flank DMRs24. More recent work has shown that the repeats are not necessarily unique to DMRs but that clusters of known repeat families, such

Allele 1

CpG

Allele 2

CpG

Methylation Acetylation Transcription complex

Figure 2 | Characteristics of imprinted genes. The figure shows a schematic pair of imprinted alleles. Hallmarks of imprinted genes such as CpG islands and repeats (arrows) are indicated. The enlarged region below the chromosomes highlights the allele-specific epigenetic changes, such as nucleosomal condensation through deacetylation, and methylation (allele 1) and opening of the chromatin by acetylation and demethylation (allele 2). The transcriptional competence of allele 2 is indicated by the binding of a transcription complex.

ing new insights about the role of methylation imprints (for example, Cdkn1c requires a maternal methylation imprint to be expressed; see below). In addition to methylation imprints, differential replication of DNA is also apparently erased in both germ lines; in the female germ line this coincides with demethylation, but in the male germline it occurs substantially later, after birth49.

Box 2 | DNA methylation and demethylation

De novo methylation Dnmt3b Dnmt3a Dnmt1 Active demethylation Maintenance methylation

Passive demethylation

DNA methylation in mammals occurs in the dinucleotide CpG. Methyl groups can be introduced into unmethylated DNA by the de novo methylation enzymes Dnmt3a and Dnmt3b (and perhaps others). When DNA is replicated, the methyl group on the template strand is recognized and a new one is introduced on the opposite (daughter) strand by the enzyme Dnmt1, which can be associated with the replication machinery. In the presence of Dnmt1, hemi-methylated DNA becomes fully methylated and so DNA methylation patterns tend to be maintained (maintenance methylation). Demethylation can occur in the absence of Dnmt1 with continued rounds of DNA replication (passive demethylation), as well as actively (without DNA replication). The nature of demethylases is unknown.

24

| JANUARY 2001 | VOLUME 2

www.nature.com/reviews/genetics

2001 Macmillan Magazines Ltd

REVIEWS
Zygote Blastocyst Embryo

Maintenance

Maintenance

Reading Mature gametes

IC2 IC1

Establishment

Erasure

Primordial germ cells

Figure 3 | Life cycle of methylation imprints. Erasure, establishment and maintenance of methylation imprints at imprinting centres during germ cell and embryonic development. Imprinting control elements 1 (IC1) and IC2 are shown as examples (see chromosome 11p15.5 in FIG. 1). Grey indicates modification and white indicates no modification at the corresponding alleles. Parental chromosomes are marked according to their sex in blue (male) or red (female). The reading (transcriptional interpretation of the primary imprints) in the developing embryo is indicated by arrows.

LINE-1 ELEMENT

A class of repetitive transposable element interspersed between and within genes throughout the genome. Most have degenerated and lost transposition activity.

as LINE-1 ELEMENTS, can also be found next to DMRs55. Intriguingly, in various organisms there are epigenetictargeting systems including methylation, that lead to heterochromatization and inactivation of tandemly repeated gene arrays56. Local heterochromatization could therefore lead to methylation of nearby CpGrich DMRs. CpG-rich DMRs could also be modified epigenetically so that they can become methylated. A hallmark of imprinting clusters is their different timing of DNA replication in S phase; recent work indicates that this property might be already acquired in the germ cells and, at least in oocytes, might precede acquisition of the methylation imprint 49; it is thus possible that different timing of replication results in different accessibility of DNA to the de novo or maintenance methylation machinery. Indeed, regulation of methyltransferases can be cell-cycle specific57. It is also possible that some imprints are established not in the germline, but rather by immediate demethylation or de novo methylation, after fertilization, of only one of the alleles. Parent-specific demethylation or de novo methylation immediately after fertilization has indeed been documented (see below). Maintenance. It might be imagined that maintenance of methylation imprints after fertilization is trivial and that, once DMRs are differentially methylated in

the parental germ cells, these patterns are simply maintained after fertilization by Dnmt1. The problem is that there is genome-wide demethylation after fertilization and a wave of de novo methylation after implantation58, both of which have to be resisted by DMRs (FIG. 4). How is demethylation resisted by DMRs? First, demethylation occurs both by active and by passive mechanisms (BOX 2). Whereas the paternal genome is largely demethylated by an active mechanism only hours after fertilization when the parental genomes are still separate in the pronuclei 59,60, the maternal genome is largely demethylated passively by failure to maintain methylation during DNA replication61,62. The maternal genome presumably needs a protection mechanism against active demethylation at fertilization; because demethylation might be triggered by chromatin remodelling of the sperm genome, the maternal genome might be protected by its nucleosomal chromatin structure. Thus it would be interesting to know whether paternally methylated DMRs adopt a specialized chromatin structure in the sperm that does not undergo substantial remodelling after fertilization. Passive demethylation is thought to come about by exclusion of Dnmt1 from the nucleus during cleavage divisions63. So it remains a mystery how methylation in DMRs is maintained during cleavage and which enzymes might be involved. The different requirements that the parental genomes have for their protection from demethylation might explain why there seem to be more maternal germline methylation imprints than paternal ones (W.R. and J.W., unpublished). Resistance to de novo methylation after implantation might be conferred by the specialized chromatin features of unmethylated alleles, or their earlier replication during S phase. Indeed, CpG islands on the inactive X chromosome become methylated de novo owing to the action of Dnmt3a/Dnmt3b after implantation, when the inactive chromosome becomes late-replicating64.
Reading mechanisms

Once the imprints are introduced in the parental germlines, maintained in the early embryo and fully matured during differentiation, they need to be read. Reading means the conversion of methylation or chromatin imprints into differential gene expression. Differential gene expression is thought to be largely at the level of transcription, although there might be the possibility of post-transcriptional mechanisms as well65. A general feature of the reading mechanisms of imprinted genes is that they seem to be complex. The complexity is likely to arise from the fact that many imprinted genes are clustered, and that clustering involves interactions between neighbouring genes and their control sequences. The clustering and interactions between neighbouring genes might be explained by extension of the conflict theory to mechanistic interactions between genes that oppose each others function66. So far, imprinted genes have been shown to be regulated by epigenetic

NATURE REVIEWS | GENETICS

VOLUME 2 | JANUARY 2001 | 2 5

2001 Macmillan Magazines Ltd

REVIEWS
Kcnq1ot1, respectively)7072. Both of these antisense RNAs originate in introns of the sense genes and are colinear with DNA. The Air gene overlaps the promoter of the sense gene70, but this is not the case for Kcnq1ot1 (REF. 55). The promoter regions of the antisense transcripts are CpG-rich and are methylated on the inactive maternal allele; for Air and Kcnq1ot1, these CpG islands or sequences near them carry the primary germline imprint25,55. Deletions of both of these DMRs leads to loss of expression of the antisense transcripts and loss of imprinting of the sense genes73,74 (Igf2r or KCNQ1). It is not known whether expression of the antisense transcript itself interferes with transcription of the sense gene. The antisense transcript could lead to alterations of chromatin structure and DNA methylation, to promoter exclusion or work by RNA-directed mechanisms, such as RNA interference. Alternatively, it is possible that the antisense transcripts have no role by themselves, but simply reflect the activity of other regulatory elements such as silencers or boundaries (see below). Indeed, deletion of the KCNQ1OT1 DMR leads to loss of imprinted repression not only of KCNQ1, but also of CDKN1C, which does not overlap with the antisense transcript74. It is important to note that this deletion was made in somatic cells, so that any effects of germline transmission could not be assessed. Finally, in a marsupial, the opossum, the Igf2r gene is imprinted without having a differentially methylated antisense promoter or an antisense transcript22. This may also indicate that imprinting mechanisms can evolve rapidly and that the primitive (or primordial) imprinting mechanism of Igf2r worked differently from the one now seen in the mouse (W.R. and J.W., unpublished). Boundaries. The observation that endoderm-specific enhancers can be shared between the paternally expressed Igf2 and the maternally expressed H19 (a noncoding RNA) genes suggested the possibility that chromatin boundaries might be involved (FIG. 5c). The region upstream of H19 carries the paternal germline methylation imprint; when this was deleted, the maternal Igf2 gene was expressed, albeit not at its full level in all tissues67. This led to the model that the H19 DMR is a chromatin boundary that is closed when unmethylated, and open when methylated. A specific chromatin structure with several DNaseI hypersensitive sites exists on the unmethylated maternal allele33,34, and the previously characterized repressor factor CTCF (CCCTC-binding factor), which is important for the function of a chick globin boundary element, binds to this maternal allele but not to the paternal methylated one7578. This region has a boundary function in transfection assays and deletion of CTCF-binding motifs abolishes this function7578. It is not known how CTCF and perhaps other factors prevent the H19 endoderm enhancers from activating the Igf2 promoters. However, it is remarkable that the maternal Igf2 promoters are nevertheless DNaseI hypersensitive79, indicating factor binding despite transcriptional silence. Enhancers for lineages other than endoderm have so far not been discovered, although there are candidates for muscle-specific enhancers80,81.

Germ cells Primordial High Mature

Fertilization/ zygote

Embryo E5 E10

Low

Figure 4 | Methylation reprogramming in the germ line and embryo. The figure shows the level of methylation in methylated (black) and non-methylated (grey) imprinted genes and non-imprinted sequences (red, maternal; blue, paternal) during germ-cell and early embryonic development. The horizontal time axis and the vertical axis indicating the relative methylation levels are not to scale. (E, embryonic day.)

Methylation

modifications of promoter sequences, of SILENCERS, of BOUNDARY ELEMENTS, and possibly of overlapping antisense transcripts. So imprinted genes use the normal arsenal of transcriptional control mechanisms, but some of these are controlled by differential epigenetic modifications on parental chromosomes. Promoter methylation. A fairly common way to achieve transcriptional silencing of one allele is by promoter methylation (FIG. 5a). Here, the promoter region, which is often CpG rich, is heavily methylated on one allele. Note that the primary germline imprint region (the region in which epigenetic differences occur between egg and sperm) is often distinct from the promoter, but is necessary for promoter methylation during development67. In the cases tested, DNASEI HYPERSENSITIVE SITES are absent from a paternally methylated promoter30,31 and a maternally methylated promoter32. The details of how methylated promoters are transcriptionally repressed are not known yet, but it is clear that several methylCpG-binding proteins (MBDs), as well as the maintenance methyltransferase itself (Dnmt1), form a complex with histone deacetylases. This presumably leads to a closed chromatin conformation in which transcription factors cannot gain access to the promoter68. Indeed, differences in specific patterns of histone acetylation consistent with this model have been observed between alleles of imprinted genes (R. Feil and R. Gregory, personal communication). Antisense transcripts. A considerable proportion of imprinted genes are associated with antisense transcripts (at present 15%; FIG. 5b). Surprisingly, all antisense transcripts discovered so far in imprinted genes are themselves imprinted and are paternally expressed (with the exception of Tsix, the antisense transcript to Xist69), regardless of whether they occur in sense genes that are paternally or maternally expressed (W.R. and J.W., unpublished). Most antisense transcripts are noncoding and may have regulatory functions. Among the best studied are those overlapping the maternally expressed Igf2r and Kcnq1 genes (called Air and

SILENCER

DNA sequence at which repressor factors bind and mediate silencing of promoters through interaction with the basal transcriptional machinery or the enhancer.
BOUNDARY ELEMENT

DNA sequence that lies between two gene-controlling elements, such as a promoter and enhancer, preventing their communication or interaction. Boundary element function is usually mediated by the binding of specific factors.
DNASEI HYPERSENSITIVE SITE

Chromosomal region highly accessible to cleavage by deoxyribonuclease I (DNAseI). Such sites are associated with open chromatin conformations and transcriptional activity.

26

| JANUARY 2001 | VOLUME 2

www.nature.com/reviews/genetics

2001 Macmillan Magazines Ltd

REVIEWS
In the previous section, we suggested that the Kcnq1ot1 DMR could also have boundary functions at least for Cdkn1c, in addition to being a promoter region for the antisense transcript, and this could be true of other DMRs as well. A new imprinted gene cluster on mouse chromosome 12 and human chromosome 14 has been isolated recently that consists of the paternally expressed Dlk gene flanked by a maternally expressed gene, Gtl2, which is a non-coding RNA8284. This
a
Allele 1 CpG

EPIGENOTYPE

Methylation and chromatin pattern of a gene or allele.

arrangement is intriguingly similar to that of Igf2 and H19, including a paternally methylated DMR upstream of Gtl2 and a paternally methylated DMR in Dlk8284. It is puzzling how such a markedly similar arrangement could have evolved in different genes. Silencers. Several imprinted genes have DMRs that are methylated on the active allele. This has led to the proposal that these sequences contain silencers that are inactivated by methylation, perhaps by excluding repressor factors25,29,41,79 (FIG. 5d). This model has now been corroborated in the case of DMR1 in Igf2. Igf2 is paternally expressed in various fetal tissues and DMR1 functions as a maternal silencer in a subset of those tissues. DMR1 is paternally methylated in mesodermal tissues (heart, kidney and lung), in which Igf2 is expressed. In a DMR1 knockout, the maternal allele of Igf2 is derepressed in these tissues85. In addition, Igf2 continues to be expressed postnatally in the same tissues, when it is normally silenced soon after birth85. These experiments reveal that DMR1 is a mesodermal silencer; in vitro transfection experiments confirm this and show that the silencer is methylation sensitive (H. Cedar, personal communication). Another tissue-specific silencer has been detected between Igf2 and H19 in a conserved region that shows DNaseI-hypersensitive sites but no differential methylation. Deletion of this sequence from a yeast artificial chromosome (YAC) transgene results in expression of the maternal Igf2 allele specifically in skeletal muscle and tongue86. So it is possible that for specific sets of enhancers (acting in different tissues) there are specific silencers that might or might not be epigenetically controlled. Indeed, the H19 DMR also contains a silencer (in addition to the boundary) that is endoderm-specific87. The findings on antisense transcripts, boundaries and silencers reinforce the idea that various elements, some of which are under epigenetic control, interact to regulate expression of imprinted genes in clusters.
ICs and epigenetic spreading in clusters

Allele 2

CpG

b
Allele 1 CpG CpG

Allele 2

CpG

CpG

c
Allele 1 CpG

Allele 2

CpG

d
Allele 1 CpG

Allele 2 Activator Repressor Methylation

CpG

Figure 5 | Reading mechanisms in imprinted genes. a | Differential silencing by CpG island or promoter methylation. b | Regulation by antisense transcripts in conjunction with CpG island or promoter methylation; c | Allele-specific regulation of neighbouring genes by differential methylation of boundary elements within a CpG island. Factors such as CCCTC-binding factor (CTCF) (red disc) bind to the unmethylated allele and block the access of upstream promoters to downstream enhancers (green), leading to transcriptional repression of the upstream gene. d | Differential methylation results in differential binding of silencing factors (red, in this case methylation-sensitive), which repress the promoter in cis.

The existence of ICs was first proposed from the molecular- genetic analysis of imprinting disorders, and from knockouts in the mouse8891. In the imprinted region on human chromosome 15 (FIG. 1), small deletions were found in patients with PraderWilli syndrome (PWS) in the promoter region of SNURFSNRPN, and, a few kilobases upstream of this, deletions were found in patients with Angelman syndrome (AS). PWS requires paternal transmission of the deletion, whereas AS requires maternal transmission. The intriguing feature of these deletions is that they lead to altered expression and altered methylation patterns of many of the imprinted genes in the region, even if the genes are separated from the deletions by several megabases88. This is defined as EPIGENOTYPE spreading. In the PWS deletions, when paternally transmitted, otherwise paternally expressed genes are silenced and methylated. In the AS deletions with maternal transmission, genes that are otherwise repressed are now demethylated and

NATURE REVIEWS | GENETICS

VOLUME 2 | JANUARY 2001 | 2 7

2001 Macmillan Magazines Ltd

REVIEWS

WT Early germ cells IC Allele 1 Allele 2

II

III

IV

Late germ cells Allele 1

Allele 2

Soma Allele 1

Allele 2 Erasure defect Establishment defect Maintenance defect

Figure 6 | Control of imprinting clusters by imprinting centres. An imprinting centre (IC) with neighbouring imprinted genes is shown. Imprinting centre deletions have been shown to lead to altered regional imprinting (and altered epigenotype spreading) but the developmental stage at which this occurs is not known. The figure shows the three developmental stages at which defects might occur. In the erasure defect, IC deletion leads to the inability to remove the previous methylation imprints in early germ cells (box I). In the establishment defect, an IC deleted chromosome is aberrantly methylated (box II) or wrongly demethylated (box III) in late germ cells. In the maintenance defect, an IC deletion leads to postzygotic methylation (box IV) or loss of methylation (box V). (WT, wild type.)

MONOTREMES

Non-eutherian, egg-laying mammals.

expressed. The same is true in a mouse knockout model of the PWS deletion91. A model has been proposed whereby deletion of these ICs makes the switching of the regional epigenotype in the germ line impossible88 (FIG. 6). Thus, in the paternal germ line, it is suggested that the incoming deleted chromosome from the grandmother cannot be demethylated, and remains in a maternal epigenotype despite passing through a male germ line. However, no direct proof of this model has been obtained. In fact, recent evidence has shown that a deleted chromosome switched its epigenotype (from unmethylated to methylated) after fertilization, indicating that an IC might be important for maintenance of the epigenotype92. Methylation analysis of the germ cells from parents of patients with PWS and AS, and of the appropriate mouse models, is urgently needed to resolve this important issue. There is functional evidence for ICs in other clusters as well. Deletion of the maternal DMR upstream of H19 leads to expression of the maternal Igf2 and Ins2 genes, and to methylation of DMR1 and DMR2 in Igf2 (REFS 67,90,93); and deletion of the DMR in KCNQ1OT1 alters imprinting of KCNQ1 and CDKN1C74 (see above). It is remarkable that for the imprinting cluster on distal mouse chromosome 7/ human chromosome 11p15.5, there seem to be two ICs71,72,94,95(H19 DMR, Kcnq1ot1 DMR, FIG. 1) that are largely independent from each other. However, the possibility cannot be excluded that there are other mechanisms that coordinate the two ICs, perhaps involving cluster-wide epigenetic modifications, such as altered replication timing of DNA. Indeed, some patients with BeckwithWiedemann syndrome (BWS) have imprinting alterations in both subdomains of the cluster96.

It is still unknown whether epigenotype spreading occurs in the germ line or postzygotically (or perhaps both) (FIG. 6). But more mysterious still is how methylation spreading works mechanistically. It could be imagined that non-transcribed promoters become methylated, or that promoters that are overlapped by antisense transcripts become methylated. Alternatively, methylation spreading may be independent of transcription. It is possible that DMRs in a cluster adopt specific spatial arrangements that bring them in close contact with each other. They could thus protect each other from becoming methylated, or conversely attract methylation to their points of contact. Indeed, a special nuclear arrangement (possibly involving matrixattachment sites) has been described for the maternal copy of the PWS/AS cluster97.
Evolution conflicts and arms races

Imprinting has been found in eutherian mammals, marsupials and flowering plants. However, in MONOTREMES; and all other vertebrates and invertebrates, analysis of uniparental embryos or of genes imprinted in other organisms indicates that imprinting might not be present. But some of these assays are crude and certainly not comprehensive. More subtle effects of imprinting in other organisms cannot therefore be ruled out. This phylogenetic distribution of imprinting, together with the observation that a sizeable proportion of imprinted genes affect fetal growth in a potentially antagonistic manner (paternally expressed genes enhance fetal growth and maternally expressed ones suppress fetal growth) led to the proposal that genetic conflict over maternal resources was the driving force in the evolution of imprinting18. Paternally expressed genes are therefore selected to extract more resources

28

| JANUARY 2001 | VOLUME 2

www.nature.com/reviews/genetics

2001 Macmillan Magazines Ltd

REVIEWS
from the mother to benefit offspring fitness, whereas maternally expressed genes tend to conserve resources, to divide them among more offspring and to maximize reproductive performance of the female18. Accordingly, some imprinted genes could be in arms races with others and this might result in accelerated rates of evolution of imprinted genes18. However, the rate of evolution of proteins encoded by imprinted genes has not been found to be elevated over that of normal genes98. Instead, the imprinting patterns themselves seem more variable. Although the imprinting status of a considerable number of imprinted genes is conserved between mouse and human, there are now several notable exceptions where imprinting status is markedly different. For example, the human IGF2R gene only shows imprinted expression during early development, or imprinted expression is polymorphic, whereas the mouse gene shows stable imprinted expression in most fetal tissues and developmental stages(Harwell imprinting web site). Moreover, disruptions of imprinting of various genes were observed in interspecific hybrids of the deermouse Peromyscus maniculatus99. So, it may be that regulation of imprinted genes rather than the encoded proteins evolves rapidly, a phenomenon that could also contribute to mammalian speciation mechanisms99. Such changes might take place at the level of the multiple regulatory sequences involved in reading imprints (enhancers, silencers or boundary elements), as well as at the level of epigenetic modification. One possible evolutionary territory for imprinting arms races is in the zygote. Here, the observed paternalspecific active demethylation might make it difficult to maintain a paternal germline methylation imprint. In other words, the capacity of the oocyte to demethylate paternal genomes might have evolved as a means to counteract paternal growth-enhancing strategies(W.R. and J.W., unpublished). It is interesting to note in this regard that genome-wide demethylation does not occur in the zebrafish100 (which does not have imprinting), but is present in mammalian species with imprinting.

SPONGIOTROPHOBLAST

Phenotypic consequences

Junctional zone between the labyrinth and the maternal side of the placenta.
LABYRINTHINE TROPHOBLAST

Placental zone where fetal and maternal exchange takes place.

HYPERKINETIC, HYPOKINETIC

Exceeding (hyper) or reduced (hypo) movement of the body or extremeties.

Box 3 | Outstanding questions How and precisely when during germline develoment are old imprints removed and new ones introduced? Which Dnmts, demethylating activities and chromatin factors are involved? How does the spreading of epigenetic information in clusters work, and is this a germline-specific phenomenon, a postzygotic phenomenon or both? How are imprints maintained when there is genome-wide active and passive demethylation in the early embryo? How many fundamentally different arrangements of imprinted genes and imprinting control elements are there in the genome? How conserved is imprinting between mammalian species? How precisely do imprinted genes affect extraembryonic and embryonic development, and the nutritional exchange with the mother? Are there interactions of imprinted genes (particularly antagonistic ones) in known, or in novel, physiological pathways? In addition to growth and behaviour, are there other developmental processes and mechanisms in which imprinted genes have a decisive role, and how will these fit with evolutionary theories?

A substantial proportion of imprinted genes are implicated in the control of fetal growth by transgenic studies, by their location in chromosome regions that (in uniparental disomies) affect growth, or because the gene product indicates a role in growth or cell proliferation (Harwell imprinting web site). Paternally expressed genes generally enhance fetal growth and maternally expressed genes suppress fetal growth. Although the numbers are still small, this pattern is consistent with the conflict theory 18. Regardless of imprinting, only one system of genes has been discovered so far that has an important role in growth of the fetus as a whole. This is the insulin and insulin-like growth factor system together with its receptors, binding proteins and associated signal-transduction pathways101. So far, three members of this system (Igf2, Igf2r, Ins) are known to be imprinted. Imprinted genes that affect growth by an unknown mechanism could either encode new members of the Igf system (Grb10 might well be involved in transducing an Igf/Ins signal) or belong to an as yet unknown system of growth control. Fetal growth depends on the availability of nutrients provided by the mother. An indirect way of regulating fetal growth is therefore to restrict nutrient transfer through altered placental growth or function. Remarkably, most imprinted genes are expressed in the placenta (Harwell imprinting web site). Furthermore, placental growth is generally affected in those transgenic studies that also showed an effect on fetal growth. Two imprinted genes have so far been discovered that might have a role specifically in placental development or growth. Mash2 regulates the development of the SPONGIOTROPHOBLAST102, whereas an Igf2 transcript is expressed specifically in the LABYRINTHINE 85 TROPHOBLAST . Altered growth and development of these important tissues could conceivably lead to effects on nutrient transfer to the fetus, but functional studies are needed to clarify this. Effects of imprinting might also be expected in immediate postnatal growth (from birth to weaning) because resources (milk) continue to be provided by the mother18. The first indication that postnatal behaviour might be influenced by imprinted genes was obtained from studies on mouse pups disomic for distal chromosome 2. Just after birth, the mice with paternal disomy showed HYPERKINETIC behaviours whereas those with maternal disomy were HYPOKINETIC6. Other evidence that imprinted genes have a role in brain development or function is provided by the surprisingly large number of neurological and psychiatric disorders in which parent-of-origin effects are observed103. In PWS and AS (classical imprinting diseases), as well as in autism, bipolar affective disorder, epilepsy, schizophrenia, Tourette syndrome and Turner syndrome, the occurrence or severity of symptoms depends on which parent transmits the disease susceptibility. With the exception of PWS and AS, however, no specific imprinted genes have yet been found that have a role in these diseases. In AS, the only maternally expressed gene found

NATURE REVIEWS | GENETICS

VOLUME 2 | JANUARY 2001 | 2 9

2001 Macmillan Magazines Ltd

REVIEWS
in the cluster so far, UBE3A, is probably solely responsible for the disease symptoms, whereas in PWS, several otherwise paternally expressed genes (FIG. 1) are deleted or silent. The individual contributions that these deficiencies make to the disease are not yet clear. Some more specific insights have been gained from knockout experiments in the mouse. Knockouts of Grf1 (REF. 104) and Ube3a105 (the mouse homologue of the human AS gene) have defects in contextual learning and memory (among others). Peg1 and Peg3, by contrast, have a role in maternal behaviour such that mothers that lack these molecules neglect and do not feed their offspring106,107. How these phenotypes could be interpreted in terms of the genetic conflict theory or other theories is not clear. Because several imprinted genes that affect behaviour also have a role in fetal growth, these effects need to be genetically separated (for example by conditional knockouts), particularly because intra-uterine growth retardation can have long-term effects on cognitive functions.
Epimutations, reprogramming and cloning

death)113 that are typical of deregulation of imprinted genes, perhaps indicating that somatic donor cells might have had aberrant imprint patterns, or that reprogramming might interfere with proper imprint maintenance. Finally, the striking observation that ageing might be reversible by cloning114 prompts the speculation that an important component of the ageing process, as well as of diseases, could be the somatic acquisition of epigenetic modifications.
Perspectives

Just as mutations alter DNA, epimutations alter DNA methylation or chromatin patterns. Epimutations in imprinted genes can lead either to biallelic expression (loss of imprinting) or to biallelic silencing. How frequent these alterations are either in the germ line or during somatic development is not known. Epimutations that are not likely to have been caused by underlying DNA mutations have been observed in several disease situations, including Wilms tumour108,109 (H19 methylation), BWS71,72,89 (H19 methylation, KvDMR1 demethylation), and PWS/AS110 (SNURFSNRPN methylation/demethylation). BWS and PWS epimutations are likely to occur in the germ line presumably by failing to erase the grandparental imprint (or by establishing the wrong imprint), or in the early embryo. By contrast, aberrant H19 methylation in Wilms tumour arises somatically108,109. The possibility that epimutations can arise during development has been explored in embryonic stem cells. Indeed, with prolonged culture of embryonic stem cells, a high frequency of epimutations arises in H19, Igf2, Igf2r and U2af-rs1, which persist during fetal development111 (and are associated with developmental abnormalities). Cloning of various mammalian organisms has been achieved recently using donor nuclei from differentiated cells112. Gene expression and, presumably, epigenetic modifications need to be reprogrammed when the somatic nuclei are introduced into the enucleated oocyte. Whereas the introduced genome might undergo passive or active demethylation, methylation in imprinted DMRs again needs to be protected from this reprogramming so that imprints are maintained intact in the cloned organism. The fact that cloning is still very inefficient and the large majority of clones die during development might indicate that the reprogramming process is inefficient. In addition, cloned animals frequently show abnormalities (placental and fetal overgrowth, and perinatal

It is instructive to view mammalian genomic imprinting from many viewpoints, including mechanisms, phenotypic consequences and evolutionary significance. In the next few years, exciting developments will occur in all these areas, making this synthetic view ever more productive and enjoyable. Particular advances will come from the comparison of the sequences of imprinted regions in different mammalian species, including perhaps marsupials and monotremes, and with other vertebrates that do not have imprinting. This will provide important insights into the evolution of arrangements and clusters of imprinted genes, and will pinpoint conserved regions with possible regulatory roles. Their function can then be examined in precisely timed, and tissue-specific, gene-targeting experiments. The functions of regulatory sequences will depend on complexes that involve the chromatin factors and methyltransferases, and such complexes can be analysed using proteomic approaches. Once all the imprinted genes have been isolated, targeting experiments will provide crucial insights, particularly into interactions between their products in physiological pathways. Many questions about the biology of imprinting remain (BOX 3), but using the array of approaches summarized above, some fascinating answers will surely follow.
Update added in proof

A recent mouse transgenic study118 reports that the Angelman syndrome deletion region contains a sequence that can protect the maternal Snrpn promoter from becoming demethylated after fertilization, and the paternal Snrpn promoter from becoming de novo methylated in sperm and in the postimplantation embryo. This is the first report of a sequence that can protect from de novo methylation as well as maintain a methylation imprint.

Links
DATABASE LINKS H19 | Igf2 | Cdkn1c | Dnmt1 | Dnmt3a |

Dnmt3b | Tsix | Xist | Igf2r | Kcnq1 | Air | Kcnq1ot1 | CTCF | Dlk | Gtl2 | PraderWilli syndrome | SNURFSNRPN | Angelmann syndrome | Ins2 | BeckwithWiedemann syndrome | Ins | Grb10 | autism | bipolar affective disorder | schizophrenia | Tourette syndrome | Turner syndrome | UBE3A | Grf1 | Ube3a | Peg1 | Peg3 | Wilms tumour FURTHER INFORMATION Harwell imprinting web site

30

| JANUARY 2001 | VOLUME 2

www.nature.com/reviews/genetics

2001 Macmillan Magazines Ltd

REVIEWS
1. Surani, M. A. H., Barton, S. C. & Norris, M. L. Development of reconstituted mouse eggs suggests imprinting of the genome during gametogenesis. Nature 308, 548550 (1984). McGrath, J. & Solter, D. Completion of mouse embryogenesis requires both the maternal and paternal genomes. Cell 37, 179183 (1984). References 1 and 2 are key papers that uncovered the existence of genomic imprinting. Nuclear transplantation in mouse embryos showed that embryos with only maternal or paternal chromosomes could not develop normally, despite being diploid. Mann, J. R. & Lovell-Badge, R. H. Inviability of parthenogenones is determined by pronuclei, not egg cytoplasm. Nature 310, 6667 (1984). Lyon, M. F. & Glenister, P. H. Factors affecting the observed number of young resulting from adjacent-2 disjunction in mice carrying a translocation. Genet. Res. 29, 8392 (1977). Searle, A. G. & Beechey, C. V. Complementation studies with mouse translocations. Cytogenet. Cell Genet. 20, 282303 (1978). Cattanach, B. M. & Kirk, M. Differential activity of maternally and paternally derived regions in mice. Nature 315, 496498 (1985). This paper showed that specific chromosome regions could function differently depending on their inheritance from mother or father. It established marked opposing effects of the parental genomes on fetal growth and postnatal behaviour. DeChiara, T. M., Robertson, E. J. & Efstratidiadis, A. Parental imprinting of the mouse insulin-like growth factor II gene. Cell 64, 849859 (1991). Barlow, D. P., Stoeger, R., Herrmann, B. G., Saito, K. & Schweifer, N. The mouse insulin-like growth factor type-2 receptor is imprinted and closely linked to the Tme locus. Nature 349, 8487 (1991). Bartolomei, M. S., Zemel, S. & Tilghman, S. M. Parental imprinting of the H19 gene. Nature 351, 153155 (1991). References 79 describe the first paternally expressed (Igf2) and maternally expressed (Igf2r, H19) imprinted genes. All three genes are involved in fetal growth control. Nicholls, R. D., Knoll, J. H., Butler, M. G., Karam, S. & Lalande, M. Genetic imprinting suggested by maternal heterodisomy in nondeletion PraderWilli syndrome. Nature 342, 281285 (1989). Henry, I. et al. Uniparental parental disomy in a genetic cancer-predisposing syndrome. Nature 351, 665667 (1991). Reik, W., Collick, A., Norris, M. L., Barton, S. C. & Surani, M. A. Genomic imprinting determines methylation of parental alleles in transgenic mice. Nature 328, 248251 (1987). Sapienza, C., Peterson, A. C., Rossant, J. & Balling, R. Degree of methylation of transgenes is dependent on gamete of origin. Nature 328, 251254 (1987). Swain, J. L., Stewart, T. A. & Leder, P. Parental legacy determines methylation and expression of an autosomal transgene: A molecular mechanism for parental imprinting. Cell 50, 719727 (1987). Chaillet, J. R., Vogt, T. F., Beier, D. R. & Leder, P. Parentalspecific methylation of an imprinted transgene is established during gametogenesis and progressively changes during gametogenesis. Cell 66, 7783 (1991). Sasaki, H. et al. Inherited type of allelic methylation variations in a mouse chromosome region where an integrated transgene shows methylation imprinting. Development 111, 573581 (1991). Li, E., Beard, C. & Jaenisch, R. Role for DNA methylation in genomic imprinting. Nature 366, 362365 (1993). This paper showed that DNA methylation has a key role in maintaining imprinting. Imprinted expression of several genes was lost in a knockout of Dnmt1, the maintenance methyltransferase gene. Moore, T. & Haig, D. Genomic imprinting in mammalian development: a parental tug-of-war. Trends Genet. 7, 4549 (1991). Hurst, L. D. & McVean, G. T. Growth effects of uniparental disomies and the conflict theory of genomic imprinting. Trends Genet. 13, 436443 (1997). Feil, R., Khosla, S., Cappai, P. & Loi, P. Genomic imprinting in ruminants: allele-specific gene expression in parthenogenetic sheep. Mamm. Genome 9, 831834 (1998). McLaren, R. J. & Montgomery, G. W. Genomic imprinting of the insulin-like growth factor 2 gene in sheep. Mamm. Genome 10, 588591 (1999). Killian, J. K. et al. M6P/IGF2R imprinting evolution in mammals. Mol. Cell 5, 707716 (2000). 23. Paulsen, M. et al. Sequence conservation and variability of imprinting in the BeckwithWiedemann syndrome gene cluster in human and mouse. Hum. Mol. Genet. 9, 18291841 (2000). 24. Neumann, B., Kubicka, P. & Barlow, D. P. Characteristics of imprinted genes. Nature Genet. 9, 1213 (1995). 25. Stger, R. et al. Maternal-specific methylation of the imprinted mouse Igf2r locus identifies the expressed locus as carrying the imprinting signal. Cell 73, 6171 (1993). 26. Olek, A. & Walter, J. The pre-implantation ontogeny of the H19 methylation imprint. Nature Genet. 17, 275276 (1997). 27. Tremblay, K. D., Duran, K. L. & Bartolomei, M. S. A 5 2kilobase-pair region of the imprinted mouse H19 gene exhibits exclusive paternal methylation throughout development. Mol. Cell. Biol. 17, 43224329 (1997). 28. Shemer, R., Birger, Y., Riggs, A. D. & Razin, A. Structure of the imprinted mouse snrpn gene and establishment of its parental-specific methylation pattern. Proc. Natl Acad. Sci. USA 94, 1026710272 (1997). 29. Feil, R., Walter, J., Allen, N. D. & Reik, W. Developmental control of allelic methylation in the imprinted mouse Igf2 and H19 genes. Development 120, 29332943 (1994). 30. Ferguson, S. A., Sasaki, H., Cattanach, B. M. & Surani, M. A. Parental-origin-specific epigenetic modification of the mouse H19 gene. Nature 362, 751755 (1993). 31. Bartolomei, M. S., Webber, A. L., Brunkow, M. E. & Tilghman, S. M. Epigenetic mechanisms underlying the imprinting of the mouse H19 gene. Genes Dev. 7, 16631673 (1993). 32. Feil, R., Boyano, M. D., Allen, N. D. & Kelsey, G. Parental chromosome-specific chromatin conformation in the imprinted U2af1-rs1 gene in the mouse. J. Biol. Chem. 272, 2089320900 (1997). 33. Khosla, S., Aitchison, A., Gregory, R., Allen, N. D. & Feil, R. Parental allele-specific chromatin configuration in a boundary-imprinting-control element upstream of the mouse H19 gene. Mol. Cell. Biol. 19, 25562566 (1999). 34. Hark, A. T. & Tilghman, S. M. Chromatin conformation of the H19 epigenetic mark. Hum. Mol. Genet. 7, 19791985 (1998). 35. Hu, J. F., Oruganti, H., Vu, T. H. & Hoffman, A. R. The role of histone acetylation in the allelic expression of the imprinted human insulin-like growth factor II gene. Biochem. Biophys. Res. Commun. 251, 403408 (1998). 36. Pedone, P. V. et al. Role of histone acetylation and DNA methylation in the maintenance of the imprinted expression of the H19 and Igf2 genes. FEBS Lett. 458, 4550 (1999). 37. Kitsberg, D. et al. Allele-specific replication timing of imprinted gene regions. Nature 364, 459463 (1993). 38. Knoll, J. H., Cheng, S. D. & Lalande, M. Allele specificity of DNA replication timing in the Angelman/PraderWilli syndrome imprinted chromosomal region. Nature Genet. 6, 4146 (1994). 39. Paldi, A., Gyapay, G. & Jami, J. Imprinted chromosomal regions of the human genome display sex-specific meiotic recombination frequencies. Curr. Biol. 5, 10301035 (1995). 40. Robinson, W. P. & Lalande, M. Sex-specific meiotic recombination in the PraderWilli/Angelman syndrome imprinted region. Hum. Mol. Genet. 4, 801806 (1995). 41. Brandeis, M. et al. The ontogeny of allele-specific methylation associated with imprinted genes in the mouse. EMBO J. 12, 36693677 (1993). 42. Tada, T. et al. Epigenotype switching of imprintable loci in embryonic germ cells. Dev. Genes Evol. 207, 551561 (1998). 43. Tada, M., Tada, T., Lefebvre, L., Barton, S. C. & Surani, M. A. Embryonic germ cells induce epigenetic reprogramming of somatic nucleus in hybrid cells. EMBO J. 16, 65106520 (1997). 44. Ueda, T. et al. The paternal methylation imprint of the mouse H19 locus is acquired in the gonocyte stage during foetal testis development. Genes Cells 5, 649659 (2000). 45. Davis, T. L., Yang, G. J., McCarrey, J. R. & Bartolomei, M. S. The H19 methylation imprint is erased and reestablished differentially on the parental alleles during male germ cell development. Hum. Mol. Genet. 9, 28852894 (2000). 46. Labosky, P. A., Barlow, D. P. & Hogan, B. L. Mouse embryonic germ (EG) cell lines: transmission through the germline and differences in the methylation imprint of insulin-like growth factor 2 receptor (Igf2r) gene compared with embryonic stem (ES) cell lines. Development 120, 31973204 (1994) 47. Obata,Y. et al. Disruption of primary imprinting during oocyte growth leads to the modified expression of imprinted genes during embryogenesis. Development 125,15531560 (1998). 48. Kato, Y. et al. Developmental potential of mouse primordial germ cells. Development 126, 18231832 (1999). References 47 and 48 showed that germ cell nuclei, when transplanted to zygotes, had a characteristically restricted potential for development. This is explained, at least in part, by the erasure of imprints in primordial germ cells. Simon, I. et al. Asynchronous replication of imprinted genes is established in the gametes and maintained during development. Nature 401, 929932 (1999). This work showed that asynchrony in DNA replication of imprinted regions is erased in germ cells. Kafri, T. et al. Developmental pattern of gene-specific DNA methylation in the mouse embryo and germ line. Genes Dev. 6, 705714 (1992). Tada, T. et al. Imprint switching for non-random Xchromosome inactivation during mouse oocyte growth. Development 127, 31013105 (2000). Mertineit, C. et al. Sex-specific exons control DNA methyltransferase in mammalian germ cells. Development 125, 889897 (1998). Okano, M., Bell, D. W., Haber, D. A. & Li, E. DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell 99, 247257 (1999). Tucker, K. L. et al. Germ-line passage is required for establishment of methylation and expression patterns of imprinted but not of nonimprinted genes. Genes Dev. 10, 10081020 (1996). Engemann, S. et al. Sequence and functional comparison in the BeckwithWiedeman region: implications for a novel imprinting centre and extended imprinting. Hum. Mol. Genet. 9, 26912706 (2000). Dorer, D. R. & Henikoff, S. Expansions of transgene repeats cause heterochromatin formation and gene silencing in Drosophila. Cell 77, 9931002 (1994). Robertson, K. D. et al. The human DNA methyltransferases (DNMTs) 1, 3a and 3b: coordinate mRNA expression in normal tissues and overexpression in tumors. Nucleic Acids Res. 27, 22912298 (1999). Monk, M., Boubelik, M. & Lehnert, S. Temporal and regional changes in DNA methylation in the embryonic, extraembryonic and germ cell lineages during mouse embryo development. Development 99, 371382 (1987). Oswald, J. et al. Active demethylation of the paternal genome in the mouse zygote. Curr. Biol. 10, 475478 (2000). Mayer, W., Niveleau, A., Walter, J., Fundele, R. & Haaf, T. Demethylation of the zygotic paternal genome. Nature 403, 501502 (2000). Howlett, S. K. & Reik, W. Methylation levels of maternal and paternal genomes during preimplantation development. Development 113, 119127 (1991). Rougier, N. et al. Chromosome methylation patterns during mammalian preimplantation development. Genes Dev. 12, 21082113 (1998). References 5862 showed that there is genomewide demethylation in mouse preimplantation embryos followed by remethylation after implantation. The paternal genome is rapidly demethylated in the zygote, presumably by an active mechanism. The maternal genome is largely demethylated passively. Carlson, L. L., Page, A. W. & Bestor, T. H. Properties and localization of DNA methyltransferase in preimplantation mouse embryos: implications for genomic imprinting. Genes Dev. 6, 25362541 (1992). Bourchis, D. et al. Abnormal methylation does not prevent X inactivation in ICF patients. Cytogenet. Cell Genet. 84, 245252 (1999). Jouvenot, Y., Poirier, F., Jami, J. & Paldi, A. Biallelic transcription of Igf2 and H19 in individual cells suggests a post-transcriptional contribution to genomic imprinting. Curr. Biol. 9, 11991202 (1999). Reik, W. & Maher, E. R. Imprinting in clusters: lessons from BeckwithWiedemann syndrome. Trends Genet. 13, 330334 (1997). Thorvaldsen, J. L., Duran, K. L. & Bartolomei, M. S. Deletion of the H19 differentially methylated domain results in loss of imprinted expression of H19 and Igf2. Genes Dev. 12, 36933702 (1998). Deletion of the differentially methylated region upstream of H19 results in loss of repression of the linked Igf2 on the maternal allele, and loss of H19 methylation and silencing on the paternal allele. Bird, A. P. & Wolffe, A. P. Methylation-induced repression belts, braces, and chromatin. Cell 99, 451454 (1999). Lee, J. T. Disruption of imprinted X inactivation by parentof-origin effects at Tsix. Cell 103, 1727 (2000). Lyle, R. et al. The imprinted antisense RNA at the Igf2r locus overlaps but does not imprint Mas1. Nature Genet.

2.

49.

50.

3.

51.

4.

52.

5.

53.

6.

54.

55.

7.

56.

8.

57.

9.

58.

10.

59.

60.

11.

61.

12.

62.

13.

14.

15.

63.

16.

64.

17.

65.

66.

18.

67.

19.

20.

68. 69. 70.

21.

22.

NATURE REVIEWS | GENETICS

VOLUME 2 | JANUARY 2001 | 3 1

2001 Macmillan Magazines Ltd

REVIEWS

25, 1921 (2000). 71. Smilinich, N. J. et al. A maternally methylated CpG island in KvLQT1 is associated with an antisense paternal transcript and loss of imprinting in BeckwithWiedemann syndrome. Proc. Natl Acad. Sci. USA 96, 80648069 (1999). 72. Lee, M. P. et al. Loss of imprinting of a paternally expressed transcript, with antisense orientation to KVLQT1, occurs frequently in BeckwithWiedemann syndrome and is independent of insulin-like growth factor II imprinting. Proc. Natl Acad. Sci. USA 96, 52035208 (1999). 73. Wutz, A. et al. Imprinted expression of the Igf2r gene depends on an intronic CpG island. Nature 389, 745749 (1997). Deletion of the intronic DMR in Igf2r leads to loss of imprinted repression on the paternal allele. The DMR contains a promoter for the antisense transcript Air that is not transcribed from the methylated maternal allele. 74. Horike, S. et al. Targeted disruption of the human LIT1 locus defines a putative imprinting control element playing an essential role in BeckwithWiedemann syndrome. Hum. Mol. Genet. 9, 20752083 (2000). 75. Bell, A. C. & Felsenfeld, G. Methylation of a CTCFdependent boundary controls imprinted expression of the Igf2 gene. Nature 405, 482485 (2000). 76. Hark, A. T. et al. CTCF mediates methylation-sensitive enhancer-blocking activity at the H19/Igf2 locus. Nature 405, 486489 (2000). 77. Kanduri, C. et al. Functional association of CTCF with the insulator upstream of the H19 gene is parent of origin-specific and methylation-sensitive. Curr. Biol. 10, 853856 (2000). 78. Szabo, P., Tang, S. H., Rentsendorj, A., Pfeifer, G. P. & Mann, J. R. Maternal-specific footprints at putative CTCF sites in the H19 imprinting control region give evidence for insulator function. Curr. Biol. 10, 607610 (2000). Refs 7578 show that the DMR upstream of H19 contains a chromatin boundary that presumably blocks access of enhancers located downstream of H19 to the Igf2 promoters on the maternal chromosome. The unmethylated but not the methylated DMR binds the known boundary protein CTCF. 79. Sasaki, H. et al. Parental imprinting: potentially active chromatin of the repressed maternal allele of the mouse insulin-like growth factor II (Igf2) gene. Genes Dev. 6, 18431856 (1992). The first description of a differentially methylated region (DMR1), located upstream of the Igf2 gene. This work also showed that the maternal Igf2 promoters, although repressed, were neither methylated nor in a closed chromatin conformation. 80. Ishihara, K. et al. Comparative genomic sequencing identifies novel tissue-specific enhancers and sequence elements for methylation-sensitive factors implicated in Igf2/H19 imprinting. Genome Res. 10, 664671 (2000). 81. Kaffer, C. R. et al. A transcriptional insulator at the imprinted H19/Igf2 locus. Genes Dev. 14, 19081919 (2000). 82. Schmidt, J. V., Matteson, P. G., Jones, B. K., Guan, X. J. & Tilghman, S. M. The Dlk1 and Gtl2 genes are linked and reciprocally imprinted. Genes Dev. 14, 19972002 (2000). 83. Takada, S. et al. Delta-like and Gtl2 are reciprocally expressed, differentially methylated linked imprinted genes on mouse chromosome 12. Curr. Biol. 10, 11351138 (2000).

84. Wylie, A. A., Murphy, S. K., Orton, T. C. & Jirtle, R. L. Novel imprinted DLK1/GTL2 domain on human chromosome 14 contains motifs that mimic those implicated in IGF2/H19 regulation. Genome Res. 10, 17111718 (2000) 85. Constancia, M. et al. Deletion of a silencer element in Igf2 results in loss of imprinting independent of H19. Nature Genet. 26, 203206 (2000). 86. Ainscough, J. F., John, R. M., Barton, S. C. & Surani, M. A. A skeletal muscle-specific mouse Igf2 repressor lies 40 kb downstream of the gene. Development 127, 39233930 (2000). 87. Drewell, R. A. et al. Deletion of a silencer element disrupts H19 imprinting independently of a DNA methylation epigenetic switch. Development 127, 34193428 (2000). Refs 8587 describe silencers in the Igf2 and H19 genes that are important for the regulation of imprinting and expression. These silencers are tissue-specific and may interact with different sets of enhancers. Silencer function may depend on methylation or chromatin modification. 88. Buiting, K. et al. Inherited microdeletions in the Angelman and PraderWilli syndromes define an imprinting center on human chromosome 15. Nature Genet. 9, 395400 (1995). In patients with the imprinting disorders PWS or AS, an imprinting centre (IC) in the chromosome 15 imprinting cluster is deleted. Deletion of the IC alters imprinting of several genes in the whole domain. 89. Reik, W. et al. Imprinting mutations in the BeckwithWiedemann syndrome suggested by altered imprinting pattern in the IGF2-H19 domain. Hum. Mol. Genet. 4, 23792385 (1995). 90. Leighton, P. A., Ingram, R. S., Eggenschwiler, J., Efstratiadis, A. & Tilghman, S. M. Disruption of imprinting caused by deletion of the H19 gene region in mice. Nature 375, 3439 (1995). A knockout of a region containing an imprinting centre (the H19 DMR) led to altered imprinting of the linked Igf2 and Ins2 genes. 91. Yang, T. et al. A mouse model for PraderWilli syndrome imprinting-centre mutations. Nature Genet. 19, 2531 (1998). 92. Bielinska, B. et al. De novo deletions of SNRPN exon 1 in early human and mouse embryos result in a paternal to maternal imprint switch. Nature Genet. 25, 7478 (2000). 93. Forne, T. et al. Loss of the maternal H19 gene induces changes in Igf2 methylation in both cis and trans. Proc. Natl Acad. Sci. USA 94, 1024310248 (1997). 94. Paulsen, M. et al. Syntenic organization of the mouse distal chromosome-7 imprinting cluster and the BeckwithWiedemann-syndrome region in chromosome 11p15.5. Hum. Mol. Genet. 7, 11491159 (1998). 95. Caspary, T., Cleary, M. A., Baker, C. C., Guan, X. J. & Tilghman, S. M. Multiple mechanisms regulate imprinting of the mouse distal chromosome 7 gene cluster. Mol. Cell. Biol. 18, 34663474 (1998). 96. Maher, E. R. & Reik, W. BeckwithWiedemann syndrome: imprinting in clusters revisited. J. Clin. Invest. 105, 247252 (2000). 97. Greally, J. M. et al. Conserved characteristics of heterochromatin-forming DNA at the 15q11q13 imprinting center. Proc. Natl Acad. Sci. USA 96, 1443014435 (1999). 98. McVean, G. T. & Hurst, L. D. Molecular evolution of imprinted genes: no evidence for antagonistic coevolution. Proc. R. Soc. Lond. B 264, 739746 (1997). 99. Vrana, P. B., Guan, X. J., Ingram, R. S. & Tilghman, S. M. Genomic imprinting is disrupted in interspecific

Peromyscus hybrids. Nature Genet. 20, 362365 (1998). 100. Macleod, D., Clark, V. H. & Bird, A. Absence of genomewide changes in DNA methylation during development of the zebrafish. Nature Genet. 23, 139140 (1999). 101. Efstratiadis, A. Genetics of mouse growth. Int. J. Dev. Biol. 42, 955976 (1998). 102. Guillemot, F. et al. Genomic imprinting of Mash2, a mouse gene required for trophoblast development. Nature Genet. 9, 235242 (1995). 103. Isles, A. R. & Wilkinson, L. S. Imprinted genes, cognition and behaviour. Trends Cog. Sci. 4, 309318 (2000). 104. Brambilla, R. et al. A role for the Ras signalling pathway in synaptic transmission and long-term memory. Nature 390, 281286 (1997). 105. Jiang, Y. H. et al. Mutation of the Angelman ubiquitin ligase in mice causes increased cytoplasmic p53 and deficits of contextual learning and long-term potentiation. Neuron 21, 799811 (1998). 106. Lefebvre, L. et al. Abnormal maternal behaviour and growth retardation associated with loss of the imprinted gene Mest. Nature Genet. 20, 163169 (1998). 107. Li, L. et al. Regulation of maternal behavior and offspring growth by paternally expressed Peg3. Science 284, 330333 (1999). 108. Steenman, M. J. et al. Loss of imprinting of IGF2 is linked to reduced expression and abnormal methylation of H19 in Wilms tumour. Nature Genet. 7, 433439 (1994); erratum in 8, 203 (1994). 109. Moulton, T. et al. Epigenetic lesions at the H19 locus in Wilms tumour patients. Nature Genet. 7, 440447 (1994). 110. Buiting, K. et al. Sporadic imprinting defects in PraderWilli syndrome and Angelman syndrome: implications for imprint-switch models, genetic counseling, and prenatal diagnosis. Am. J. Hum. Genet. 63,170180 (1998). 111. Dean, W. et al. Altered imprinted gene methylation and expression in completely ES cell-derived mouse fetuses: association with aberrant phenotypes. Development 125, 22732282 (1998). 112. Gurdon, J. B. & Colman, A. The future of cloning. Nature 402, 743746 (1999). 113. Young, L. E., Sinclair, K. D. & Wilmut, I. Large offspring syndrome in cattle and sheep. Rev. Reprod. 3, 155163 (1998). 114. Lanza, R. P. et al. Extension of cell life-span and telomere length in animals cloned from senescent somatic cells. Science 288, 665669 (2000). 115. Kaneko-Ishino,T. et al. Peg1/Mest imprinted gene on chromosome 6 identified by cDNA subtraction hybridization. Nature Genet. 11, 5259 (1995). 116. Hayashizaki, Y. et al. Identification of an imprinted U2af binding protein related sequence on mouse chromosome 11 using the RLGS method. Nature Genet. 6, 3340 (1994). 117. Peters, J. et al. A cluster of oppositely imprinted transcripts at the Gnas locus in the distal imprinting region of mouse chromosome 2. Proc. Natl Acad. Sci. USA 96, 38303835 (1999). 118. Shemer, R. et al. The imprinting box of the PraderWilli/Angelman syndrome domain. Nature Genet. 26, 440443 (2000).

Acknowledgements
We thank the many colleagues in our labs who are involved in imprinting work for discussions and comments on the manuscript, in particular W. Dean, M. Constancia, A. Murrell, R. Feil and G. Kelsey. We also thank three anonymous referees for excellent and constructive suggestions. Work in the authors labs is supported by BBSRC, MRC, CRC, HFSP and DFG.

32

| JANUARY 2001 | VOLUME 2

www.nature.com/reviews/genetics

2001 Macmillan Magazines Ltd