The DNA methyltransferase DNMT3C protects male germ cells from transposon activity

See allHide authors and affiliations

Science  18 Nov 2016:
Vol. 354, Issue 6314, pp. 909-912
DOI: 10.1126/science.aah5143

Combating parasitic DNA by methylation

DNA methylation plays an important role in repressing the expression of “parasitic” DNAs, such as transposable elements, which have invaded our genomes. Mammals have three DNA methyltransferase enzymes. Barau et al. discovered a fourth DNA methyltransferase enzyme in mice. The enzyme DNMT3C is a duplication of DNMT3B and is found in male germ cells. There it targets evolutionarily young transposons, of which there is a heavy burden in the mouse genome. DNMT3C methylates and silences the young transposons, preserving male fertility.

Science, this issue p. 909


DNA methylation is prevalent in mammalian genomes and plays a central role in the epigenetic control of development. The mammalian DNA methylation machinery is thought to be composed of three DNA methyltransferase enzymes (DNMT1, DNMT3A, and DNMT3B) and one cofactor (DNMT3L). Here, we describe the discovery of Dnmt3C, a de novo DNA methyltransferase gene that evolved via a duplication of Dnmt3B in rodent genomes and was previously annotated as a pseudogene. We show that DNMT3C is the enzyme responsible for methylating the promoters of evolutionarily young retrotransposons in the male germ line and that this specialized activity is required for mouse fertility. DNMT3C reveals the plasticity of the mammalian DNA methylation system and expands the scope of the mechanisms involved in the epigenetic control of retrotransposons.

Genome defense via transcriptional silencing of transposable elements has been proposed to be a driving force for the evolution of DNA methylation (1). Retrotransposons occupy half of mammalian genomic space, and their control is of paramount importance in the germ line: Their activity can damage the hereditary material with an impact on fertility and the fitness of subsequent generations (2). In mammals, after germline epigenetic reprogramming, small RNA-directed DNA methylation establishes life-long epigenetic silencing of retrotransposons during the perinatal period of spermatogenesis (3). Piwi-interacting RNAs (piRNAs) are the cleavage products of retrotransposon transcripts and guide DNA methylation to the promoters of these elements through homology recognition (4, 5). Mammals have specifically evolved a catalytically inactive DNA methyltransferase (DNMT) cofactor, DNMT3L, which acts downstream of the piRNA pathway (6, 7). The inactivation of DNMT3L or PIWI-pathway proteins invariably results in hypomethylation and reactivation of retrotransposons, meiotic failure, azoospermia, and male sterility marked by small testis size (hypogonadism) (8).

To gain insights into the biology of retrotransposon silencing in the germ line, we screened a collection of hypogonadal male mice generated through N-ethyl-N-nitrosourea (ENU) mutagenesis for ectopic retrotransposon activity (fig. S1, A and B). Five independent positive lines were obtained, but all showed linkage to the same genomic interval on chromosome 2 (fig. S1C), suggesting that they shared a spontaneous, ENU-independent mutation. Whole-genome sequencing revealed a de novo insertion of an IAPEz element, a subclass of inracisternal A-particle (IAP) retrotransposon, in the last intron of the Gm14490 gene (Fig. 1A and fig. S1, D to F). Gm14490 maps 9 kilobases (kb) downstream of the Dnmt3B gene and was annotated as a nonfunctional tandem duplication of Dnmt3B, based on lack of transcription and recognizable open reading frames (ORFs) (9).

Fig. 1 Gm14490 encodes a male germ cell–specific de novo DNA methyltransferase, DNMT3C.

(A) Structure of Gm14490 (Ensembl 2011) and position of the IAPEz insertion (antisense orientation). RACE and RNA-seq analysis of E16.5 and P10 testis identifies a long isoform with coding potential (ATG, green triangle). (B) Gm14490 is detected in testis by reverse transcription–quantitative polymerase chain reaction (RT-qPCR), but not in germ cell–depleted testis from Dnmt3LKO/KO animals. Tissues from 10-week-old mice, unless otherwise specified. (C) RT-qPCR of the two Gm14490 isoforms during testis development. Predominant germ cell populations are represented. Primordial germ cells (PGC), spermatogonial stem cells (SSC), and spermatogonia (Spg). (D) DNMT3C shows characteristics of DNMT3 proteins—conserved methyltransferase (MTase) motifs and an ADD domain, but no PWWP domain. (E) RNA-seq supporting wild-type and mutant Dnmt3C IAP splicing events in E16.5 testis. (F) LUminometric methylation assay (LUMA) of global CpG methylation in Dnmt-tKO ESCs that transiently express Dnmt3C- and Dnmt3B-3XFLAG alleles. Data are mean ± SD from three technical replicates in (B) and (C) and from three biological replicates in (F). nd, not detectable.

We found that Gm14490 was exclusively expressed in male germ cells (Fig. 1B and fig. S2A). Using RNA sequencing (RNA-seq) and rapid amplification of cDNA ends (RACE), we annotated two transcript isoforms, whose expression was tightly regulated during spermatogenesis (Fig. 1, A and C). The short, noncoding isoform was expressed in postnatal testes. However, the long isoform (2.8 kb) possessed a 709-codon ORF, and its expression sharply peaked around embryonic day 16.5 (E16.5), coinciding with male germline de novo DNA methylation (5, 10). In comparison, Dnmt3B is expressed in germ cells, early embryos, and somatic tissues of both sexes (11, 12). Using discriminating primers, we showed independent regulation of Dnmt3B and Gm14490 during spermatogenesis (fig. S2B). The coding potential and specific developmental regulation of Gm14490 led us to reconsider it as a functional paralog of Dnmt3B rather than a pseudogene, and thus we renamed it Dnmt3C.

The long Dnmt3C isoform encodes a protein with an organization characteristic of DNMT3 enzymes: six methyltransferase motifs (I, IV, VI, VIII, IX, and X) in C-terminal position and an N-terminal ATRX-DNMT3L-DNMT3A (ADD) domain, which binds unmethylated lysine 4 residues of histone H3 (H3K4) (Fig. 1D and fig. S2C) (13). However, DNMT3C lacks the Pro-Trp-Trp-Pro (PWWP) domain, which targets DNMT3 proteins to gene bodies through recognition of H3K36 trimethylation (H3K36me3) (14, 15). Overall, DNMT3C exhibits 70% identity with DNMT3B, while DNMT3A and DNMT3B are 46% identical. In hypogonadal mutants, the IAP insertion did not affect Dnmt3C transcript levels but provided an alternative splice acceptor site, which led to the exclusion of Dnmt3C last exon in favor of the retrotransposon sequence in a chimeric Dnmt3C-IAP mRNA (Fig. 1E and fig. S2, D and E). Its predicted translation product lacks motifs IX and X (fig. S2C), which are essential for the AdoMet-dependent methyltransferase fold and for the binding of the methyl donor S-adenosyl methionine, respectively (16).

To demonstrate that DNMT3C is catalytically active, we performed an in vivo DNA methylation assay. Dnmt3C is not expressed in mouse embryonic stem cells (ESCs) (Fig. 1B). By transfecting constructs driving Dnmt3C expression in DNA methylation-free ESCs (Dnmt1, Dnmt3A, and Dnmt3B triple-knockout; Dnmt-tKO) (17), we observed a gain of CpG methylation (10 to 20%), similar to that observed upon Dnmt3B transfection (Fig. 1F and fig. S2, F and G). The mutant Dnmt3CIAP allele failed to raise CpG methylation levels, as did Dnmt3C and Dnmt3B mutant alleles with a missense mutation in the catalytic site (DNMT3C C507A and DNMT3B C658A, in which cysteine at position 507 and 658 is replaced by alanine). An in vitro DNA methylation assay using the DNMT3C methyltransferase domain showed concordant results (fig. S2, H and I). These findings demonstrate that DNMT3C is an enzymatically active member of the DNMT3 family of de novo DNA methyltransferases.

Dnmt3CIAP/IAP animals were somatically normal, and only males were sterile (fig. S3A). Hypogonadism was linked to azoospermia with interruption of spermatogenesis at the pachytene stage of meiosis I, in the context of impaired chromosome synapsis (Fig. 2, A to C). The developmental phenotype of Dnmt3C mutant mice was similar to that observed in Dnmt3LKO/KO males (7, 18), suggesting that DNMT3C could be involved in transposon silencing during spermatogenesis. Indeed, the same set of retrotransposons were up-regulated in Dnmt3CIAP/IAP and Dnmt3LKO/KO testes at P20 (postnatal day 20) (Fig. 2D and fig. S3B) (18). Long interspersed nuclear elements (LINEs or L1s) showed the strongest reactivation, and more specifically evolutionarily young subfamilies: type A, T, and Gf transcripts were increased by 10-fold in Dnmt3CIAP/IAP testes, in association with accumulation of L1-encoded ORF1 proteins (Fig. 2E). Among endogenous retroviruses (ERVs), reactivation was specific to some ERVK families (MMERVK10C, IAPEz, and IAPEy). As in the case of the Dnmt3L mutation, L1 and IAPEz derepression was linked to a DNA methylation defect in Dnmt3CIAP/IAP testes (Fig. 2F), despite normal expression of piRNA/DNA methylation genes and piRNA production during fetal spermatogenesis (fig. S3, C to F). Finally, we confirmed DNMT3C function by generating a Dnmt3C knockout mouse through CRISPR-Cas9–mediated deletion (fig. S4A). The Dnmt3CKO allele recapitulated the Dnmt3CIAP/IAP developmental and molecular phenotypes in homozygous Dnmt3CKO/KO males and failed to complement the Dnmt3CIAP allele in Dnmt3CIAP/KO compound heterozygous males (fig. S4, B to E).

Fig. 2 Phenotype of Dnmt3CIAP/IAP males.

(A) Hypogonadism (6-week-old mice). Scale bars, 5 mm. (B) Severe germ cell loss in testis sections (11-week-old mice). Scale bars, 100 μm. (C) Impaired chromosome synapsis at meiosis as detected by immunofluorescence against synaptonemal complex proteins (SYCP1 and SYCP3). Scale bars, 4 μm. (D) RNA-seq heatmap shows overexpression of young L1 and specific ERVK types in Dnmt3CIAP/IAP compared to Dnmt3CIAP/WT testis at P20. Annotations from RepeatMasker. (E) Aberrant expression of L1-ORF1 proteins in Dnmt3CIAP/IAP germ cells (TRA98-positive) at P20. Scale bars, 50 μm. (F) L1A-5′UTR and IAPEz-5′LTR are hypomethylated in Dnmt3CIAP/IAP testis DNA at P20, as in Dnmt3LKO/KO testis. Southern blot analysis after methyl-sensitive Hpa II digestion. Msp I is used as a digestion control.

To assess the contribution of DNMT3C to male germline methylation, we performed whole-genome bisulfite sequencing (WGBS) in sorted germ cells from testes at P10, when de novo DNA methylation is completed. Overall CpG methylation levels of Dnmt3CIAP/IAP mutant cells were not markedly different from the Dnmt3CIAP/WT control (77.7 versus 78.5%). A slight decrease was only apparent when focusing on transposons (81.5 versus 84.2%), and more specifically on LINEs, ERVK, and ERV1 (Fig. 3, A and B). Accordingly, there was only a limited number (264) of differentially methylated regions (DMRs) in Dnmt3CIAP/IAP versus Dnmt3CIAP/WT germ cells (fig. S5A); all reflected hypomethylation in the mutant, and most overlapped with LINEs (34%) and ERVs (48%). RepeatMasker annotations further highlighted that the same families that were transcriptionally derepressed were hypomethylated in Dnmt3CIAP/IAP testes; namely, young L1s and specific ERVKs (Fig. 2D and Fig. 3C). By comparison, deletion of DNMT3L had a stronger and broader impact: Dnmt3LKO/KO germ cells exhibited only 39% of global CpG methylation, and all genomic compartments and retrotransposon classes were affected (Fig. 3, D and E, and fig. S5B).

Fig. 3 DNMT3C methylates evolutionarily young retrotransposon promoters.

(A and B) Tukey box-plot representation of CpG methylation content as determined by WGBS over different (A) genomic compartments and (B) retrotransposon classes in control Dnmt3C IAP/WT (purple) and mutant Dnmt3C IAP/IAP (yellow) germ cells at P10. (C) Percentage of DNA methylation loss within individual retrotransposon families in Dnmt3CIAP/IAP samples. (D and E) As in (A) and (B) but for Dnmt3LKO/WT (green) and Dnmt3LKO/KO (red) germ cells. (F) Plotting of DNA methylation loss over individual copies of L1 families according to genetic distance from consensus sequences. (G and H) Metaplots of DNA methylation over full-length L1s (>5 kb) comparing (G) Dnmt3CIAP/IAP and (H) Dnmt3LKO/KO versus control samples. (I) Left: Heatmap of Dnmt3CIAP/IAP and control DNA methylation levels across paternally imprinted loci. Right: Methylation maps at the Rasgrf1 locus.

The DNA methylation defect in Dnmt3C mutants only reached 30% at the most for young L1s (Fig. 3C). We reasoned that DNMT3C selectively affects transcriptionally active retrotransposon copies, as these are targets of piRNA-dependent DNA methylation during fetal spermatogenesis (4, 5). Indeed, individual L1-A and -T elements with a 5′ promoter (length >5 kb) and the highest similarity toward the consensus sequence showed the greatest DNA methylation loss in Dnmt3CIAP/IAP cells (Fig. 3F and fig. S4C). Additionally, DNMT3C-dependent DNA methylation was not evenly distributed across the length of these transcriptionally competent elements, but rather focalized to their promoters [5′ untranslated region (UTR)] (Fig. 3G), in a pattern previously observed in piRNA-deficient MiliKO/KO males (fig. S5D) (4). Older L1-F elements did not show such trends (Fig. 3F and fig. S5C). By contrast, Dnmt3L deficiency caused demethylation of L1s independently of their age and throughout their sequence (Fig. 3H).

ERVKs exhibited milder methylation loss (10%) than LINEs in Dnmt3CIAP/IAP cells (Fig. 3C), and this was not related to the size or the sequence conservation of individual copies (fig. S5, C and E). ERVKs partially resist the genome-wide erasure of methylation that occurs in the fetal germ line (19, 20); this likely explains their limited dependency toward DNMT3C remethylation activity. Nevertheless, DNMT3C dependency was still greatest over regulatory long-terminal repeat (LTR) sequences of MMERVK10C (fig. S5F) and IAPEz elements when highly conserved copies were analyzed by bisulfite pyrosequencing (fig. S4E). Finally, among paternally methylated imprinted loci, only the Rasgfr1 imprinting control region (ICR) was hypomethylated in Dnmt3CIAP/IAP germ cells (Fig. 3I). This ICR includes an ERVK LTR fragment, which acquires DNA methylation in a piRNA-dependent manner during spermatogenesis (21). This highlights again the genomic convergence between piRNA and DNMT3C targeting in fetal male germ cells.

DNMT3C is highly specialized at methylating young retrotransposon promoters in the male germ line. In comparison, DNMT3B is not involved in germ cell development but establishes somatic methylation patterns genome-wide in the early embryo (11, 22, 23). DNMT3C has therefore evolved its own regulatory pattern and genomic targets, which have diverged from the ancestral DNMT3B copy. Despite the strict requirement of DNMT3C in the mouse, the Dnmt3C duplication is not universal among mammals but occurred ~46 million years ago in the last common ancestor of the muroid rodents (Fig. 4, A and B, and fig. S6, A and B). These represent the largest mammalian superfamily, with 1518 species accounting for 28% of all extant mammals (fig. S5C) (24), and include the two primary models for biomedical research, particularly in reproduction and endocrinology: the mouse and the rat. The genomes of Muroidea carry a heavy burden in young transposons: 25% have integrated in the last 25 million years with currently thousands of active copies (25). In comparison, most human transposons became extinct during that time (26).

Fig. 4 Phylogenetic distribution of Dnmt3C.

(A) Locus organization and synteny around the Dnmt3B (blue) and Dnmt3C (red) genes in human, rabbit, mouse, and rat genomes. (B) Dnmt3C distribution in the mammalian phylogenetic tree with a focus on the Rodentia order. Species in detail have available sequenced genomes. Red branches: Muroidea families in which Dnmt3C was identified; dashed branches: Muroidea families without available genome sequence.

The lack of transposon methylation defects in the germ cells of Dnmt3A or Dnmt3B mutant mice had been previously interpreted as a sign of functional redundancy between these two enzymes (23, 27). We show here that this function relies instead on an additional de novo DNMT. The mammalian DNA methylation members should now be considered as a quintet: DNMT1, 3A, 3B, 3C, and 3L. The two most recent evolutionary additions are linked to reproduction: The eutherian DNMT3L cofactor stimulates germline methylation genome-wide, and the muroid DNMT3C enzyme methylates young retrotransposons during spermatogenesis. Future research should resolve the biochemical mechanism that designates specific sequences for DNMT3C-dependent DNA methylation. Its PWWP-free status may prevent it from targeting H3K36me3-enriched gene bodies and redirect it to retrotransposon promoters. In conclusion, our discovery of DNMT3C raises a new set of challenges to the current views of the remarkable evolution of DNA methyltransferases, the regulation of the de novo methylation process, and its tight links with the selective pressure to epigenetically control transposons in the germ line.

Supplementary Materials

Materials and Methods

Figs. S1 to S6

Tables S1 to S7

References (2854)

References and Notes

  1. Acknowledgments: We thank members of D.B.’s laboratory, especially M. Greenberg and M. Walter. We are grateful to T. H. Bestor, V. Colot, E. Heard, and E. Mulugeta for advice; S. Fouchécourt, B. Fumel, L. Mourao, and B. Piegu for technical help; A. Bortvin and S. Tajima for antibodies; C. Jouhanneau for excellent mouse husbandry; E. Desale and PHENOMIN-TAAM for the mutagenesis program; and PICT-IBiSA (France–Bioimaging, ANR-10-INSB-04), the Flow Cytometry and Animal Platforms of the Institut Curie. This research was supported by grants from European Research Council (ERC-Cog EpiRepro), Agence Nationale pour la Recherche (ANR TranspoFertil, ANR-10-LABX-0030-INRT, and ANR-10-INBS-07 PHENOMIN). D.B.’s laboratory is part of the Laboratoire d’Excellence (LABEX) DEEP (11-LBX0044). J.B. was a recipient of a postdoctoral fellowship from Agence Nationale de Sécurité Sanitaire (Anses). The Institut Curie ICGex NGS platform was supported by ANR-10-EQPX-03 (Equipex) and ANR-10-INBS-09-08 (France Génomique Consortium) grants and by the Canceropôle Ile-de-France. Mouse line distribution under uniform biological material transfer agreement. All sequencing data sets are available at the National Center for Biotechnology Information Gene Expression Omnibus database under GSE84141.
View Abstract

Stay Connected to Science

Navigate This Article