Main

A subset of the mouse and human genomes is expressed from only one allele in a parent-of-origin–specific manner. This subset includes imprinted X-chromosome inactivation and autosomal imprinted loci. It has been proposed that this epigenetic regulation is accomplished through covalent modifications of both the DNA and the N-terminal tails of core histones in nucleosomes1,2. There are more than 60 identified autosomal imprinted genes, about half of which are paternally repressed and half maternally repressed. Most imprinted genes that have been examined contain at least one DMR located in the 5′ promoter region or in the body of the gene itself3. Recently, several proteins (DNA methyltransferases, CpG methyl binding proteins, chromatin insulators) have been identified as trans-acting factors involved in the epigenetic regulation of these loci4. Many of these factors either possess or associate with proteins that possess DNA methyltransferase activity. Additionally, recent studies have shown correlations between covalent histone modifications and the transcriptional status of imprinted alleles5. In particular, methylation of histone H3 has been associated with the inactive X chromosome6.

PcG protein complexes are thought to maintain long-term gene silencing during development through alterations of local chromatin structure7. In both Drosophila and mammals, recent reports have shown that the Eed/Ezh2 PcG complex contains histone methyltransferase (HMT) activity, methylating histone H3, and that mutations in the SET domain of the Ezh2 fly homolog, E(Z), abolish the enzymatic activity of the complex in vitro8,9,10. The Eed/Ezh2 complex has also been shown to interact with histone deacetylases (HDACs; ref. 11). These results provide biochemical mechanisms by which the Eed/Ezh2 complex may establish heritable epigenetic states leading to long-term gene repression in vivo. We recently showed that loss of function of Eed in the mouse results in reactivation of the imprinted X chromosome in extraembryonic lineages12. Additionally, the Eed and Ezh2 proteins have been reported to co-localize with XIST on the imprinted X chromosome of mouse trophoblast stem cells13. Thus, we investigated whether Eed is also required for epigenetic regulation of autosomal imprinted loci.

We examined the expression patterns of 18 imprinted genes during peri-gastrulation stages (embryonic days (E) 5.5–8.5) using embryos derived either from an intersubspecific cross between Mus musculus domesticus (CD1) and Mus musculus molosinus (JF1) or from an intrasubspecific CD1 × C57BL/6 cross. For each gene examined, we identified a polymorphism in the mRNA to distinguish between maternal and paternal transcripts after RT–PCR. At all stages examined, the paternally repressed genes Cdkn1c, Ascl2, Grb10 and Meg3 were expressed only from the maternal allele in wild-type mice (Fig. 1), showing that these four genes are imprinted during gastrulation. In Eed17Rn5-3354SB null homozygous (herein referred to as Eed−/−) littermates, however, we found mRNA expression from both maternal and paternal alleles of these genes (Fig. 1a–d). Four other paternally repressed (Kcnq1, Slc22a1l, Tssc3 and Igf2r) and six maternally repressed genes (Kcnq1ot1, Snrpn, Peg3, Dlk1, Nnat and Plagl1) showed normal parent-of-origin expression patterns in Eed−/− embryos (Figs. 1e,f and 2). These data indicate that loss of Eed function affects a subset of imprinted loci that are expressed early in development resulting in biallelic expression, herein referred to as loss of imprinting (LOI; ref. 14).

Figure 1: Expression patterns of imprinted genes in wild-type (wt) and Eed−/− embryos.
figure 1

Representative results from individual embryos (each lane) are shown for each assay. m, maternal control; p, paternal control. a, RFLP analysis of Cdkn1c at E7.5 b, Typical results for direct sequencing of Ascl2 RT–PCR products. Arrowhead indicates base position of the expressed polymorphism (maternal = A, paternal = G). c, RFLP analysis of Grb10 at E7.5. d, RFLP analysis of Meg3 at E7.5. e, RFLP analysis of Kcnq1ot1 at E7.5. f, RFLP analysis of Kcnq1 at E7.5.

Full size image
Figure 2: Imprinted expression summary.
figure 2

Schematic representation indicating the chromosomal location of each of the 18 genes examined for imprinted expression in wild-type gastrulation-stage embryos. Ube3a, Igf2 and H19 showed biallelic expression, and U2af1-rs1 was not detected by RT–PCR; therefore, these genes were not informative. Maternally repressed genes and paternally repressed genes are listed in separate columns on the right side of each chromosome (number in centromere denotes chromosome). Genes that showed LOI in Eed−/− embryos are marked with an asterisk.

Full size image

To determine whether LOI contributes to the early lethality of Eed−/− embryos, we used a targeted null mutation of Cdkn1c15 to reduce Cdkn1c expression levels. Maternal inheritance of the Cdkn1c null allele resulted in Eed−/− Cdkn1c+/− embryos that had developed substantially further than had Eed−/− Cdkn1c+/+ embryos, although they were smaller and developmentally delayed when compared with wild-type littermates (data not shown). Morphological analysis at E8.5–9.0 showed that these embryos produced neural folds, a beating heart and somites, tissues that Eed−/− embryos do not develop16. These data indicate that the LOI at Cdkn1c is partially responsible for the Eed−/− embryonic phenotype.

To address the molecular mechanisms leading to LOI at affected loci, we examined the methylation status of CpG islands at Cdkn1c, Grb10, Snrpn and KvDMR1 by bisulfite sequencing of genomic DNA collected from E7.5 wild-type and Eed−/− embryos. Analysis of a region 5′ to the transcriptional start site of Cdkn1c (CpG2; Fig. 3a) showed that the active maternal allele was hypomethylated and the inactive paternal allele was hypermethylated in wild-type embryos (Fig. 3b,d). These data extend previous observations from methyl-sensitive restriction digests that showed parent-of-origin methylation at three sites in the Cdkn1c coding sequence17. In Eed−/− embryos, parent-of-origin methylation was present, but there were specific changes in the methylation pattern when compared with that of their wild-type counterparts (Fig. 3c). Eed−/− maternal Cdkn1c alleles were unmethylated except for two CpGs at the 3′ end of the region analyzed. The Eed−/− paternal Cdkn1c alleles also had a methylation pattern that was distinct from that of the wild-type paternal allele (Fig. 3c). Notably, the Eed−/− paternal Cdkn1c allele was active even though it was hypermethylated. This result was not anticipated because the silent paternal allele is hypermethylated in wild-type embryos. These data indicate that methylation at this DMR does not silence paternal Cdkn1c gene expression in Eed−/− embryos. The specific pattern of CpG methylation, however, could be essential to the epigenetic state (inactive) of the paternal Cdkn1c allele. It is also notable that whereas there is parent-of-origin methylation in wild-type embryos at this DMR, it is not differentially methylated in germ cells, indicating post-fertilization regulation of methylation at this CpG island.

Figure 3: Cdkn1c CpG methylation.
figure 3

a, Schematic representation of a 10-kb genomic fragment of the Cdkn1c locus, indicating the location of exons (boxes 1–4) and CpG islands in the region (CpG1 and CpG2). CpG2 is expanded to indicate all 35 CpG dinucleotides (filled ovals) in the region analyzed after bisulfite treatment. The SmaI restriction site used in d is indicated. b,c, Parent-of-origin methylation patterns in wild-type (wt; b) and Eed−/− (c) E7.5 embryos. m, maternal; p, paternal. Filled ovals indicate methylated CpGs; open ovals indicate unmethylated CpGs. Wild-type includes both Eed+/+ and Eed+/− embryos. Symbols denote changes in the Eed−/− methylation patterns. Arrows, gain of paternal-allele methylation; asterisks, loss of paternal-allele methylation; arrowheads, gain of maternal-allele methylation. d, Genomic DNAs were digested with SmaI before PCR and then analyzed by Tsp509I RFLP to assign parent-of-origin to each of the alleles recovered through bisulfite sequencing, as there was no SNP in the region examined (b,c). In both wild-type and mutant embryos, only the paternal uncut PCR product (arrow) remained after SmaI digestion, indicating that the paternal allele was methylated at this site (arrowhead in a) whereas the maternal allele was not. m, maternal; p, paternal. F and R indicate locations of forward and reverse PCR primers, respectively; SmaI and Tsp509I restriction site locations are indicated in schematic (a and d).

Full size image

For Grb10 on mouse chromosome 11, we identified a potential DMR containing 146 CpG dinucleotides in 1,200 bases located between exons 1 and 2. Bisulfite sequencing of 194 bases (Fig. 4a) containing 22 CpGs confirmed parent-of-origin methylation at this CpG island. The active maternal allele was hypermethylated and the silent paternal allele was hypomethylated in wild-type embryos (Fig. 4b). Three consecutive CpGs were always methylated (Fig. 4c) on the paternal allele in wild-type embryos, whereas all other CpGs were unmethylated. The Eed−/− paternal Grb10 allele showed a variable pattern of hypomethylation, and the three CpGs that were invariantly methylated in wild-type embryos were not methylated in mutant embryos (Fig. 4c). We observed no differences between the wild-type and Eed−/− maternal Grb10 alleles, both of which were methylated at every CpG in the region analyzed. Similar analyses of the Snrpn DMR18, whose imprinted expression is not affected in Eed−/− embryos (Fig. 2), indicated no changes in parent-of-origin methylation patterns (Fig. 4d).

Figure 4: Grb10 and Snrpn CpG methylation.
figure 4

a, Schematic representation of the genomic Grb10 locus (roughly 40 kb shown), indicating exons (boxes 1,2) and the location of the CpG island 4 kb 3′ of exon 1. The CpG island is expanded to show all 22 CpG dinucleotides in the region analyzed after bisulfite treatment. Two SNPs in the region allowed for assignment of alleles recovered, and these are indicated by the base-pair changes (paternal (JF1) = A, maternal (CD1) = G at each SNP). b,c, Parent-of-origin methylation patterns in wild-type (wt; b) and Eed−/− (c) E7.5 embryos. Filled ovals indicate methylated CpGs; open ovals indicate unmethylated CpGs. Wild-type includes both Eed+/+ and Eed+/− embryos. Symbols denote changes in the Eed−/− methylation patterns. Arrows, gain of paternal-allele methylation; asterisks, loss of paternal-allele methylation; arrowheads, gain of maternal-allele methylation. m, maternal; p, paternal. d, Methylation analyses at the Snrpn promoter region. Filled ovals indicate methylated CpGs; open ovals indicate unmethylated CpGs. Wild-type includes both Eed+/+ and Eed+/− embryos. m, maternal; p, paternal.

Full size image

The imprinting control region (ICR) KvDMR1 on distal chromosome 7 has been shown to regulate several closely linked, paternally repressed genes including Cdkn1c, Tssc3, Slca221l, Kcnq1 and Ascl2 (Fig. 5a), as paternal inheritance of a deletion of KvDMR1 results in LOI at these imprinted loci19. The observed LOI at these loci may be due to removal of the CpG island at the ICR or to disruption of the Kcnq1ot1 antisense transcript. To test the role of Eed at this ICR, we examined both parent-of-origin methylation at KvDMR1 (Fig. 5) and expression of the antisense Kcnq1ot1 transcript (Fig. 1f) in mutant embryos. Bisulfite sequencing in this ICR showed that parent-of-origin methylation was present in Eed−/− embryos. Maternal KvDMR1 alleles in both wild-type and Eed−/− embryos were hypermethylated, but there were two CpGs that were never methylated in Eed−/− maternal KvDMR1 alleles but always methylated in wild-type maternal alleles (Fig. 5b,c). We recovered both methylated and unmethylated populations of paternal alleles from wild-type and Eed−/− littermates. Preliminary tissue-specific analysis suggests that the paternally methylated populations are predominantly extraembryonic in origin (data not shown). In the methylated paternal KvDMR1 alleles, there was one CpG that was always methylated in wild-type embryos and never methylated in Eed−/− embryos (Fig. 5c).

Figure 5: KvDMR1 CpG methylation.
figure 5

a, Schematic representation of genomic KvDMR1 locus indicating the six paternally repressed genes (boxes) that have been shown to be regulated by this ICR, located in intron 10 of Kcnq1 (ref. 28). KvDMR1 is expanded to indicate all 33 CpG dinucleotides in the region analyzed after bisulfite treatment. One SNP in the region allowed for assignment of alleles recovered, and these are indicated by the base-pair change (paternal (JF1) = G, maternal (CD1) = T). One putative CTCF binding site was identified containing 11 of 16 bases identical to conserved CTCF binding sites29. Two EagI sites previously examined for parent-of-origin methylation are shown30. b,c, Parent-of-origin methylation at the ICR in wild-type (wt; b) and Eed−/− (c) embryos. Note the presence of two populations of paternal alleles in both wild-type and Eed−/− embryos. Filled ovals indicate methylated CpGs; open ovals indicate unmethylated CpGs. Wild-type includes both Eed+/+ and Eed+/− embryos. Symbols denote changes in the Eed−/− methylation patterns. Asterisks, loss of paternal-allele methylation; arrowheads, gain of maternal-allele methylation. m, maternal; p, paternal. Open circles, loss of maternal methylation.

Full size image

An insulator model for imprinted gene expression at this cluster has been proposed, on the basis of in vitro reporter assays that show CTCF, a zinc-finger protein, binds to unmethylated sites in the KvDMR1 ICR20. Alternatively, imprinted expression at this cluster could be regulated by the Kcnq1ot1 antisense transcript, similar to the model that has been proposed at the Igf2r/Air locus on chromosome 17 (ref. 21). Neither Kcnq1ot1 expression nor the methylation status of the ICR were altered in Eed−/− embryos. These data indicate that Eed represents an additional mechanism of epigenetic regulation that is required to maintain parent-of-origin expression at specific loci in this cluster. In Eed−/− embryos, Cdkn1c and Ascl2 had LOI but Slc22a1l, Tssc3 and Kcnq1 were properly imprinted, whereas all imprinted genes in the cluster were affected in the KvDMR1/Kcnq1ot1 deletion19. Gene-specific LOI was also observed in the cluster of imprinted genes on chromosome 12 (ref. 22), where Meg3 expression was subject to LOI but Dlk1 expression was properly imprinted in Eed−/− embryos.

Our expression and methylation data identify Eed as a trans-acting factor that regulates imprinted gene silencing at a specific subset of autosomal loci during early development. Of the genes affected by mutations in Eed that have been identified thus far, no correlation with chromosomal location (Fig. 2) or methylation patterns was observed in wild-type embryos at ICRs or DMRs. For example, in normal littermates, the DMR of the active maternal Cdkn1c allele was unmethylated whereas at the DMR associated with Grb10, the active maternal allele was entirely methylated. The only consistent finding, including imprinted X-chromosome inactivation12, is that in each case we observed biallelic expression of paternally repressed genes in mutant embryos (Fig. 2). Though these results are suggestive, more loci need to be examined to determine whether regulation by Eed is in fact restricted to paternally repressed genes.

DNA methyltransferase activity has not been associated with any PcG protein complexes. But it is possible that the alterations in DNA methylation patterns that we observed are intimately linked with altered covalent histone modifications. Recent studies in both Neurospora and Arabidopsis have shown that cytosine methylation is dependent on histone methylation23,24. Given the HMT activity of the Eed/Ezh2 complex and its potential for HDAC interactions, we propose that alterations to covalent histone modifications (methylation and possibly acetylation) are the primary defect resulting in LOI when the Eed/Ezh2 complex is disrupted and that the accompanying changes in DNA methylation patterns at Cdkn1c, Grb10 and KvDMR1 in Eed−/− embryos are a secondary consequence.

The Eed/Ezh2 complex interacts through Eed with another PcG protein, Yy1, which has DNA-binding ability25. One possible explanation for the specificity of genes that show LOI in Eed−/− embryos could be that Yy1 targets the Eed/Ezh2 complex to specific recognition sites of imprinted genes, resulting in local histone modifications. In the absence of functional Eed, Yy1 may not be able to recruit the HMT activity of Ezh2, resulting in hypomethylation of histone H3 and a transcriptionally permissive environment leading to derepression of normally silent alleles. This model also explains the observed biallelic expression (as opposed to loss of expression of the active allele at affected loci) and is consistent with the role of Eed in a PcG group complex for maintaining long-term repression11.

At the affected loci examined, LOI cannot be the direct consequence of an inability of Eed−/− cells to distinguish between the parental chromosomes, as parent-of-origin methylation patterns were present in Eed−/− embryos. This observation is in contrast with LOI caused by loss of parental imprints in uniparental disomies, deletion of ICRs or ablation of DNA methyltransferases. Our findings suggest that the presence of parent-of-origin methylation is not sufficient to direct imprinted gene expression at specific loci. It is possible, however, that particular patterns of CpG methylation may be essential for silencing of the paternal allele at affected loci and that the changes observed in Eed−/− embryos lead to LOI.

Incorrect gene expression at imprinted loci has been reported in many human disorders, including Beckwith–Wiedemann syndrome, Prader–Willi syndrome, Angelman syndrome and several types of cancer26. In a substantial fraction of individuals with these diseases, imprinting defects are observed in a subset of imprinted genes in the absence of any detectable mutations in either the affected genes or known cis-acting regulatory elements14. Because Eed is required to maintain the epigenetic state leading to imprinted gene expression, we propose that mutations in EED or other trans-acting regulatory factors may account for the imprinting defects in these diseases. Here we showed that Eed is one such trans-acting factor that is necessary for parent-of-origin silencing of a subset of imprinted genes, providing the first link between a mammalian PcG gene and genome imprinting.

Methods

Expression analyses.

We examined imprinted expression patterns by intron-spanning RT–PCR from freshly dissected gastrulation-stage embryos. We crossed Eed+/− CD1 dams with Eed+/− JF1 sires to produce embryos that were polymorphic at all loci examined except for Meg3. We used CD1 × C57BL/6 embryos to examine Meg3. For all expression analyses, we extracted mRNA from individual gastrulation-stage embryos using Invitrogen Tri-Reagent. We synthesized cDNA in 20-μl reactions at 37 °C for 1 h using Invitrogen SuperscriptII RT and then amplified it by 32 rounds of PCR with 1 mM MgCl2 at an annealing temperature of 60 °C. Polymorphism position (bp), SNP (CD1/JF1) and type of analysis used (enzyme) for each gene were as follows: Cdkn1c, bp 35,253 A/G, RFLP (Tsp509I); Dlk1, undetermined, bp 7,670–8,062, SSCP; Grb10, bp 897 G/A, RFLP (Cac8I); H19, bp 846 G/A, RFLP (BclI); Igf2, undetermined, bp 2,202–2,471, SSCP. Igf2r, 13-bp length polymorphism, bp 4,486–4,725; Kcnq1, bp 1,018 T/C, RFLP (AluI); Kcnq1ot1, bp 2,883 T/G, RFLP (HpyCH4IV); Ascl2, bp 1,315 T/C, direct sequencing; Meg3, bp 1,835 CD1 = G, B6 = A, RFLp (BsrDI); Nnat, undetermined bp 442–666, SSCP; Peg3, bp 5,329 A/C, RFLP (TaqI); Slc22a1l, bp 680 C/T, RFLP (NdeI); Snrpn, bp 391 (A/C), RFLP (NlaIII); Tssc3, bp 122 A/C, SSCP; Ube3a, bp 2,366 T/C, RFLP (Tsp509I); U2af1-rs1, undetermined, bp 4,966–5,381, SSCP; Plagl1, bp 941 A/G, RFLP (TaqI). Primer sequences are available on request. All mice used were treated according to the recommendations of the Institutional Animal Care and Use Committee of the University of North Carolina at Chapel Hill.

Bisulfite sequencing.

After bisulfite treatment (as described in ref. 27), we carried out nested PCR reactions to amplify the following regions for methylation analysis: Cdkn1c, bp 34,033–34,328; Grb10, bp 65,194–65,398, SNPs at bp 65,297 and bp 65,230 (at both SNPs CD1 = C, JF1 = T); Snrpn, bp 67,729–68,105, SNP at bp 67,875 (CD1 = G, JF1 = A); KvDMR1, bp 2,542–2,934, SNP at bp 2,883 (CD1 = T, JF1 = G). Primer sequences are available on request. The conditions for first-round PCR were 3 min at 95 °C and then 36 cycles of 45 s at 95 °C, 56 °C and 72 °C. The conditions for second-round PCR were 3 min at 95 °C and then 36 cycles of 45 s at 95 °C, 60 °C and 72 °C. All reactions included 10% dimethylsulfoxide and 1.5 mM MgCl2. We cloned PCR products from at least two independent reactions and sequenced individual clones from each sample. Bisulfite conversion efficiency ranged from 94.6% to 99.1% for independently treated samples. Between 59% and 81% of individual clones sequenced (range represents independent PCR reactions) showed unique patterns of bisulfite conversion (considering all cytosine residues).

Methylation-sensitive restriction digest at Cdkn1c.

We first digested genomic DNAs from both mutant and heterozygous E7.5 embryos with the methyl-sensitive restriction enzyme SmaI and then carried out PCR analysis of a 1.4-kb segment (containing no additional SmaI sites; primer sequences available on request). We then digested the PCR products with Tsp509I for RFLP analysis and resolved them on a 0.5% agarose gel (Tsp509I polymorphism at position 35,253; CD1 = A, JF1 = G).

Accession numbers.

Dlk1, Ab047760; Grb10, G754063; H19, NM013123; NM010514;Igf2, NM010514; Igf2r, 20897376; Kcnq1, U70068; Kcnq1ot1, AF119385; Ascl2, BC019520; Meg3, Y13832; Nnat, AK003004; Peg3, AB003040; Slc22a1l, 6679179; Snrpn, 7705508; Tssc3, NM_009434; Ube3a, 27804320; U2af1-rs1, AF309654; Plagl1, AF147785; Cdkn1c, AP001293; Grb10, AL663087; Snrpn, AF332579; KvDMR1, AF119385.