Dnmt3L and the Establishment of Maternal Genomic Imprints

See allHide authors and affiliations

Science  21 Dec 2001:
Vol. 294, Issue 5551, pp. 2536-2539
DOI: 10.1126/science.1065848


Complementary sets of genes are epigenetically silenced in male and female gametes in a process termed genomic imprinting. TheDnmt3L gene is expressed during gametogenesis at stages where genomic imprints are established. Targeted disruption ofDnmt3L caused azoospermia in homozygous males, and heterozygous progeny of homozygous females died before midgestation. Bisulfite genomic sequencing of DNA from oocytes and embryos showed that removal of Dnmt3L prevented methylation of sequences that are normally maternally methylated. The defect was specific to imprinted regions, and global genome methylation levels were not affected. Lack of maternal methylation imprints in heterozygous embryos derived from homozygous mutant oocytes caused biallelic expression of genes that are normally expressed only from the allele of paternal origin. The key catalytic motifs characteristic of DNA cytosine methyltransferases have been lost from Dnmt3L, and the protein is more likely to act as a regulator of imprint establishment than as a DNA methyltransferase.

Genomic imprinting imposes a requirement for biparental reproduction. Uniparental mammalian conceptuses are inviable (1–3), and lack of imprinting of specific chromosomes or chromosome segments as a result of uniparental isodisomy or deletions of imprinting centers causes distinct developmental defects according to the chromosomal region involved (4, 5). The poor success rate and unpredictable phenotypic variation seen in mammals produced by cloning procedures are also likely to involve disturbances of genomic imprints (6). However, very little is known about the mechanisms that establish and maintain genomic imprints.

Maternal imprints are established in growing diplotene oocytes and paternal imprints in perinatal prospermatogonia (7–10). In many cases imprinting involves the de novo methylation of regulatory regions of affected genes (4); these methylation marks are maintained throughout development and are only erased and reestablished in the germ line. The DNA methyltransferases and regulatory factors involved in the establishment of imprints in germ cells have not been identified. Dnmt3L (11) became a candidate for such an activity on the basis of sequence similarity to Dnmt3A and Dnmt3B, which have been shown to catalyze de novo methylation (12). However, Dnmt3L lacks the sequence motifs shown to be involved in activation of the target cytosine, binding of the methyl donorS-adenosyl l-methionine, and sequence recognition (13). Similarity of Dnmt3L to Dnmt3A and Dnmt3B is largely restricted to a cysteine-rich region of unknown function and regions between catalytic motifs (Fig. 1A).

Figure 1

Relationship of Dnmt3L to other mammalian DNA methyltransferases and disruption of Dnmt3L gene. (A) Sequence relationships among mammalian DNA methyltransferases. Catalytic motifs are designated with roman numerals. Motifs are absent from Dnmt3L, whereas the cysteine-rich regions and other sequences show strong similarities with Dnmt3A and Dnmt3B. At right is a ClustalW representation of sequence similarities within the region spanning catalytic motifs I to VIII. The corresponding region of Dnmt3L was identified by alignment with Dnmt3A and Dnmt3B. (B) Disruption of the Dnmt3L gene by homologous recombination in ES cells. Methods were as described (25,27), except that CSL3 ES cells derived from blastocysts of strain 129SvEv/Tac were used. The disruption brings the β-geo reporter/resistance gene under the control of the endogenousDnmt3L promoter. The X 3′ of exon 1 indicates the site of three in-frame stop codons; the polyadenylation signal prevents expression of downstream exons. Restriction endonuclease sites: B, Bam HI; H, Hind III; Xh, Xho I; Xb, Xba I. DNA blot hybridization (C) after cleavage of ES cell DNA with Bam HI confirmed that the expected homologous recombination event had taken place. The mutant allele was designated Dnmt3LG .

Dnmt3L was disrupted by homologous recombination in mouse embryonic stem (ES) cells by means of a deletion-replacement mutation that removed four exons and inserted a β-galactosidase–neomycin phosphotransferase (β-geo) fusion gene (14) under the control of the endogenousDnmt3L promoter (Fig. 1, B and C). The disrupted allele was termed Dnmt3LG . HeterozygousDnmt3LG mice were of normal phenotype and showed high-level expression of the β-geo marker exclusively in the cell types in which genomic imprints are established (7–10): growing oocytes in adult females and prospermatogonia in perinatal males (Fig. 2, A and B). Homozygous animals of both sexes were viable and of normal visible phenotype, but both sexes were sterile. Testes of homozygous Dnmt3LG animals contained normal complements of germ cells at birth, but adult testes had severe hypogonadism and Sertoli cell–only phenotype (Fig. 2, C to E).

Figure 2

Expression of Dnmt3L gene in male and female germ cells and sterility of homozygousDnmt3LG males. (A) Accumulation of β-geo reporter under control of the Dnmt3L promoter specifically in growing oocytes as assessed by staining adult ovaries with X-Gal. Growing oocytes at all stages are stained, but primary oocytes and somatic cells are unstained. The oviduct is at bottom. (B) Transcription from Dnmt3L promoter in seminiferous tubules of fetal testis. Staining for β-geo with X-Gal was as in (A). (Left) Wild-type testis from 17.5 dpc mouse fetus; (right) testis from heterozygous littermate. Staining within seminiferous tubules was present in prospermatogonia (29). Postpartum and adult testes showed much less β-Gal activity (29). (C) Hypogonadism in homozygousDnmt3LG testes. Testis of wild-type littermate is at left. (D) Sertoli cell–only phenotype in seminiferous tubules of homozygous Dnmt3LG adult males. The tubule lumen is occupied only by cytoplasmic processes of Sertoli cells. (E) Section of seminiferous tubule from wild-type littermate.

Oogenesis was normal in homozygous Dnmt3LG females, but the mutation behaved as a maternal-effect lethal in that heterozygous progeny of homozygous females failed to develop past 9.5 days postcoitum (dpc). The most notable anatomical abnormalities within the embryo proper were pericardial edema with exencephaly and other neural tube defects (Fig. 3A). These defects, together with death at midgestation, are common consequences of abnormalities of extraembryonic tissues (15). In the case of heterozygous progeny of homozygous Dnmt3LG females these included a failure of chorio-allantoic fusion (Fig. 3, C and D), hyperproliferation of secondary trophoblastic giant cells and overgrowth of the chorion, hyperproliferation of yolk sac endoderm, and excess maternal blood in the vicinity of the ectoplacental cone (Fig. 3D). The defects were not due to uterine environment effects, because a similar phenotype was seen when heterozygous progeny of homozygous females were transferred to oviducts of wild-type foster females (Fig. 3B).

Figure 3

Maternal-effect lethal phenotype in heterozygous embryos derived from homozygous Dnmt3LG females. (A) Exencephalic embryo at 9.5 dpc. (B) Developmental failure of heterozygous progeny of homozygous Dnmt3LG females when transferred to uteri of wild-type females. The left uterine horn received 15 mutant embryos; the right horn received 15 wild-type control embryos. At 13.5 dpc mutant conceptuses are represented only by necrotic implantation sites (blue spots at left), whereas development of wild-type conceptuses (white spots) in the right horn is normal. (C) Normal chorio-allantoic fusion in a wild-type conceptus at 8.5 dpc. (D) Abnormalities of extraembryonic structures and failure of chorio-allantoic fusion in a heterozygousDnmt3LG 9.5 dpc embryo derived from a homozygous oocyte. Note the abnormal separation of allantois (All) and chorion (Chor) by comparison with (C), thickened chorion, and hyperproliferation of secondary trophoblastic giant cells in ectoplacental cone (EPC) and of yolk sac endoderm (lateral to allantois). Excess maternal blood apposed to ectoplacental cone is also apparent.

A role for Dnmt3L in the establishment of genomic imprints was suggested by the specific expression in germ cells at stages where imprints are established, the sequence affinities with known DNA methyltransferases, and the maternal-effect phenotype. Bisulfite genomic sequencing (16) of the differentially methylated region (DMR) of the imprinted and maternally repressedSnrpn gene (17) revealed that the DMR was heavily methylated at all tested sites in DNA of control oocytes but was markedly undermethylated in oocytes of homozygousDnmt3LG females (Fig. 4A). DNA blot hybridization after cleavage with the methylation-sensitive restriction endonuclease Hha I was used to determine whether the methylation deficiency present in the oocyte persisted on the maternal allele in progeny derived from crosses to wild-type males. Neither allele of Snrpn was detectably methylated in such heterozygous embryos (Fig. 4B). Bisulfite genomic sequencing showed that one-half of the alleles of the imprinted genesH19, Snrpn, and Peg1 were methylated in DNA of control embryos. H19, one of the rare genes whose maternal expression is enforced by methylation of the paternal allele (18, 19), showed normal allele-specific methylation in heterozygous progeny of homozygous females, whereas the maternally imprinted Snrpn and Peg1 genes were unmethylated on both alleles (Fig. 4C). These results showed that Dnmt3L is required for the establishment of maternal methylation imprints during oogenesis, and that a maternal store of Dnmt3L is not required for the maintenance of paternal methylation imprints. Global genome methylation (the large majority of which resides in repeated sequences) was not notably reduced in DNA of heterozygous progeny of homozygous Dnmt3LG mutant females (Fig. 4B), and demethylation imposed by Dnmt3L deficiency during oogenesis was largely restricted to the DMRs of maternally imprinted genes.

Figure 4

Lack of maternal methylation imprints in homozygousDnmt3LG oocytes and in heterozygous embryos derived from them. (A) Methylation of all CpG dinucleotides in the DMR of Snrpn in control embryos and light and variable methylation of the same sequence in homozygousDnmt3LG oocytes. (B) Lanes 1 and 2: lack of methylation at Hha I sites in the DMR of the maternal allele ofSnrpn in heterozygous embryos derived from homozygousDnmt3LG oocytes. DNA was digested with Hha I and Nde I. Lanes 3 to 6: global genome methylation was not decreased in heterozygous embryos derived from homozygousDnmt3LG oocytes, as assessed by sensitivity to Hpa II (lanes 4 and 6). Lanes 3 and 5 contained Msp I digests. Lanes 7 to 10, IAP retroposons (30) and pericentric satellite DNA (29) are not demethylated in heterozygous embryos derived from homozygous Dnmt3LG oocytes. In all cases DNA was purified from embryos at 8.5 dpc. Samples marked G/+ were from heterozygous progeny of homozygous Dnmt3LG females. (C) Loss of allele-specific methylation at the maternally imprinted genes Snrpn and Peg1 with normal monoallelic methylation at the paternally imprintedH19 gene. Controls show monoallelic methylation for all tested sequences [upper portion of (C)].

The effect of abnormal methylation imprints on imprinted gene expression was tested in embryos derived from crosses of Mus musculus females homozygous for Dnmt3LG to wild-type M. m. castaneus (CAST) males or to a strain in which CAST chromosome 7 had been introgressed into a C57BL/6J strain background to improve breeding efficiency (20). Expressed polymorphisms allowed assignment of the parental origin of transcripts (20). As shown in Fig. 5,Snrpn, Necdin, Zfp127,Kcnq1ot1, and Peg3 were transcribed from both alleles in heterozygous Dnmt3LG progeny of homozygous females but only from the paternal allele in control M. musculus × CAST embryos. The paternally methylated and imprinted H19 gene remained maternally expressed, in agreement with the retention of paternal H19 methylation seen in Fig. 4C. Igf2, which is normally maternally repressed (3), retained paternal-specific expression as predicted by unperturbed H19 imprinting (4,21). Cdkn1 and Ipl were not expressed from either allele, as assessed by reverse transcription–polymerase chain reaction (PCR) with specific primers and an internal Necdin control (Fig. 5B). This is likely due to repression of the maternal allele of Cdkn1 and Ipl as a result of reactivation of the maternal allele of the nearby Kcnq1ot1gene, which is predicted to repress in cis the active allele of other imprinted genes in the cluster (22–24). Development of heterozygous Dnmt3LG embryos is most similar to that of embryos reconstituted from a normal sperm nucleus and a haploid nucleus derived from a nongrowing oocyte, which lacks both maternal and paternal imprints (7).

Figure 5

Biallelic expression of maternally imprinted genes in heterozygous progeny of homozygousDnmt3LG females. Expressed polymorphisms were introduced by crossing homozygous Dnmt3LG females to M. m. castaneus males or to males containingM. m. castaneus chromosome 7 on a C57BL/6J strain background (20). Regions containing the polymorphisms were recovered by RT-PCR and the allele-specific expression determined by direct sequencing of the products in (A) or by agarose gel electrophoresis in (B). (A) Snrpn,Necdin, Zfp127, Kcnq1ot1, andPeg3 were expressed only from the paternal allele in control embryos (top row) but were expressed from both alleles in heterozygous progeny of homozygous Dnmt3LG females (bottom row). H19 remained imprinted and was expressed only from the maternal allele, as predicted from the methylation data of Fig. 4. (B) Expression of Igf2 remained monoallelic and paternal, in agreement with the retention of imprinted expression ofH19 and the reciprocal imprinting of Igf2 andH19 (4). The Cdkn1 and Iplgenes were silenced biallelically as a result of reactivation of maternal Kcnq1ot1 transcription (22–24). Polymorphisms are described in (31). Samples marked G/+ were from heterozygous progeny of homozygous Dnmt3LG females.

Mutations in Dnmt3L and each of the known DNA methyltransferases produce distinct phenotypes. Deletion of the somatic form of Dnmt1 causes global genome demethylation with dysregulation of imprinted genes and ectopic X-chromosome inactivation (25, 26), whereas deletion of the oocyte-specific form of Dnmt1 causes a pure maternal-effect phenotype in which one-half of the normally silent alleles of certain imprinted genes are demethylated and reactivated in heterozygous progeny of homozygous females (27). Mutations inDnmt3B prevent the methylation of specific types of pericentric satellite DNA and cause the human immunodeficiency and chromosome instability disease known as ICF syndrome (28). Although demethylation has not been reported to occur in Dnmt3A-deficient cells, Dnmt3A- Dnmt3Bdouble-mutant ES cells have been reported to undergo global genome demethylation and a loss of the ability to methylate newly integrated retroviral DNA (12). Dnmt3L is required specifically for the establishment of genomic imprints but is dispensable for their propagation, and Dnmt3L is the only gene known to be essential for the de novo methylation of single-copy DNA sequences. The results of this and prior studies (27) confirm that the methylation of single-copy sequences and repeated sequences are independently regulated. The sequence of Dnmt3L suggests that the protein is likely to function not directly as a DNA methyltransferase but as a regulator of methylation at imprinted loci, and identification and characterization of germ cell factors that interact with Dnmt3L should lead to a better understanding of the mechanisms that establish genomic imprints.

  • * Present address: Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences, Shanghai 200031, People's Republic of China.

  • Present address: Department of Biological Sciences, Dartmouth College, Hanover, NH, USA.

  • To whom correspondence should be addressed. E-mail: thb12{at}


View Abstract

Navigate This Article