Special Reviews

Epigenetic Decisions in Mammalian Germ Cells

See allHide authors and affiliations

Science  20 Apr 2007:
Vol. 316, Issue 5823, pp. 398-399
DOI: 10.1126/science.1137544


Specific sequences are designated for de novo DNA methylation at CpG dinucleotides in mammalian germ cells. The result is the long-term transcriptional silencing of the methylated sequences, most of which are retrotransposons and CpG-rich sequences associated with imprinted genes. There is profound sexual dimorphism in both the nature of the sequences that undergo de novo methylation in germ cells and in the mechanism by which de novo methylation is regulated. The restriction of future gene expression by the imposition of heritable methylation patterns in germ cell genomes is characteristic of mammals but is rare in other taxa.

Genomic methylation patterns are largely erased during the proliferation and migration of primordial germ cells (PGCs) and reestablished in sex-specific patterns during spermatogenesis and oogenesis (1). Most de novo DNA methylation is directed to transposons and their remnants and to clustered repeats (primarily pericentric satellite DNA), with lesser amounts at single-copy sequences and the differentially methylated regions (DMRs) of imprinted loci (2). CpG islands associated with promoter regions of nonimprinted genes are mostly unmethylated (3). Loss of methylation, or a failure to establish methylation patterns, reanimates retrotransposons in germ and somatic cells and causes biallelic expression or repression of imprinted genes in embryonic tissues (46). Even minor disruption of methylation patterns can be lethal, and a broad range of developmental abnormalities is seen when specific genomic regions are abnormally methylated (7).

The DNMT3A/B and DNMT1 families of DNA cytosine-5 methyltransferases are responsible for the establishment and maintenance of methylation patterns, respectively, and are expressed in most dividing cell types (2). DNMT3L (DNMT3-like) is related in sequence to DNMT3A and DNMT3B but lacks enzymatic activity. It is expressed only in germ cells and only at the stages where de novo methylation occurs (5), and acts as a regulator of DNMT3A and DNMT3B (8). Most Dnmt genes contain sex-specific germline promoters that are activated at specific stages of gametogenesis (911). These promoters give rise to germ cell–specific transcripts that lead to the production of truncated forms of the proteins, or to untranslated mRNAs that may have regulatory properties (Fig. 1). The functional consequences of the alternative mRNAs and proteins are largely unknown. The oocyte-specific form of DNMT1 (DNMT1o) has a higher stability compared to the somatic form. It accumulates to very high levels in the cytoplasm of oocytes and persists in preimplantation embryos where it enters nuclei at the eight-cell stage to maintain imprinted methylation patterns (12).

Fig. 1.

Expression of germ cell–specific forms of Dnmt family members. Red indicates germ cell–specific exons and splice variants at left; approximate developmental timing of expression of each form is shown at right. DNMT1s and DNMT1o are active DNA methyltransferases; DNMT1p mRNA contains multiple stop codons in an alternative 5′ exon and is not translated. DNMT3L is expressed in both male and female germ cells but from different promoters, and a third alternative transcript, which cannot encode functional DNMT3L, is present in spermatids. De novo methylation occurs before birth in the entire cohort of prospermatogonia in males, but de novo methylation in females occurs before ovulation over the reproductive life span of the female. Also, de novo methylation occurs long before spermatogonia enter into meiosis in males, but de novo methylation occurs only after the crossing-over stage of meiosis I in females. PGC, primordial germ cell.

DNMT3L-directed de novo methylation occurs in populations of nondividing cells in both male and female germ lines (Fig. 2). However, methylation acquisition is a premeiotic phenomenon in the male germ line, where it occurs in prenatal prospermatogonia, whereas methylation is acquired postmeiotically in the female germ line in growing oocytes that are arrested at the diplotene stage of meiosis I (1, 5). Loss of DNMT3L results in very different phenotypes depending upon the sex examined. Deletion of Dnmt3L in female mice prevents establishment of maternal methylation imprints in oocytes without marked effects on retrotransposon methylation (5). The result is a maternal-effect lethal phenotype in which the heterozygous offspring of homozygous Dnmt3L-null females (which are of normal phenotype) show biallelic expression of genes that are normally maternally methylated and repressed. This leads to abnormal development of extraembryonic structures and death of the embryo before mid-gestation.

Fig. 2.

Expression of Dnmt3L in female and male germ cells. A β-galactosidase–neomycin resistance marker was fused to the promoter of the endogenous Dnmt3L gene. Expression (blue) is seen in growing oocytes in the adult mouse ovary at left and in prospermatogonia (a nondividing precursor to spermatogonial stem cells) in the testis from a newborn mouse at right. Dnmt3L is expressed in testis from ∼7 days before birth to ∼3 days after birth.

The Dnmt3L-null phenotype in male mice is markedly different. Male germ cells that lack DNMT3L show fulminating expression of retrotransposons of the LINE-1 (long interspersed elements) and IAP (intracisternal A particles) classes, severe asynapsis and nonhomologous synapsis at meiotic prophase, and eventual apoptosis of all germ cells before pachytene (6). Methylation patterns at the small number of paternally methylated DMRs are almost normal, but there is a failure to methylate retrotransposons. The DNMT3L protein appears to be essentially identical in both germ lines, but the timing of expression is markedly different in the two sexes, and the methylation phenotypes of Dnmt3L-null male and female germ cells are nearly mirror images. This implies that DNMT3L potentially interprets a preexisting mark that is established by other factors and located at different genomic regions in male premeiotic and female postmeiotic germ cells; the nature of the mark may be a particular posttranslational histone modification or set of modifications. This could explain how histone modifications (which have not been shown to be subject to long-term passive inheritance) might be converted into heritable patterns of DNA methylation that can impose long-term transcriptional silencing on the affected sequences.

The profound sexual dimorphism in the ontogeny of methylation patterns in male and female germ lines has a direct consequence on the genomic characteristics of the DMRs associated with paternally and maternally imprinted genes (13). De novo methylation occurs shortly before ovulation in growing oocytes, and methylation patterns are soon erased in the primordial germ cells of the next generation. CpG dinucleotides that are methylated in the germ line are rapidly mutated to TpG and CpA dinucleotides (14), and because female germline genomes are methylated for only a small number of cell divisions, the incidence of C→T mutations at methylated sites will be low. In the male germ line, the entire cohort of prospermatogonia undergo de novo methylation in the perinatal period, and spermatogonia must maintain methylation patterns for the large and variable number of mitotic divisions that precede entry into meiosis. This increases the load of C→T mutations, especially for the offspring of older males, and the mutational pressures have led to a low density of CpG dinucleotides in paternally methylated DMRs relative to maternal DMRs and to an underrepresentation of paternally methylated DMRs relative to maternally methylated DMRs. For unknown reasons, maternally methylated DMRs usually span the promoters of imprinted genes, whereas paternally methylated DMRs can be located many kilobases away from the affected genes.

Diversity in the nature of the sequences designated for de novo methylation (single-copy versus repeated sequences, clustered versus interspersed organization) makes it likely that multiple signals direct de novo methylation in germ cells, but the nature of those signals is unclear. RNA-directed DNA methylation (RdDM) is a candidate, and such a pathway does exist in plants, where it is involved in transcriptional repression of transposons. However, an adaptive RNA interference (RNAi) response seems unlikely to exist in somatic mammalian cells where double-stranded RNA (dsRNA) provokes a general inhibition of translation via protein kinase RNA-activated (PKR) and the interferon response (15). Most organisms that use RNAi for both transcriptional and posttranscriptional silencing have a Dicer-like protein specific to each process [reviewed in (16)], which cleaves dsRNA to short 21- to 22-nucleotide dsRNAs that inhibit translation or induce degradation of homologous RNAs. Mammalian genomes encode only one Dicer protein, which localizes to the cytoplasm. Mammalian Dicer is most closely related to the Drosophila melanogaster Dicer-1, which is required for microRNA (miRNA) processing and not production of small interfering RNAs (siRNAs, which mediate the adaptive RNAi response). Loss of Dicer activity in mouse embryonic cell lines has a minimal effect on transposon methylation and activity, and siRNAs directed to transposable elements have not been identified (17). It appears that RNAi components in mammalian cells may be used exclusively for miRNA production under normal conditions and that DNA methylation in mammals may not require input from an RNA component. However, if an adaptive RNAi response does operate in mammals, it could be active only in germ cells, where the absence of an interferon response would allow for the use of a dsRNA-responsive mechanism (18). A germ cell–specific class of small RNAs has recently been identified, and their potential functions in the germ line are discussed elsewhere [reviewed in (16)]. They could provide evidence that RNA is implicated in sequence-specific de novo methylation in mammalian germ cells (19).

De novo methylation can be RNA independent. The ascomycete fungus Neurospora crassa genome senses and methylates repeated sequences just before meiosis in a process that is not affected by mutations in components of the RNAi pathway (20). The large majority of 5-methylcytosines in the mammalian genome is in dispersed and tandem repeats (21), and a repeat-sensing mechanism could operate in mammalian germ cells as well.

Errors in establishment or maintenance of germ cell methylation patterns can cause human diseases, as in some cases of the imprinting disorders Beckwith-Wiedeman, Angelmann, and Prader-Willi syndromes (7). High rates of discordance in monozygotic twins reported for psychiatric disorders may involve methylation abnormalities that arise in early development (germline mutations cannot explain the discordance), and the paternal-age effect in schizophrenia may involve errors in the maintenance of genomic methylation patterns during the many cell divisions that male germ cells must undergo as spermatogonia. Epigenetic errors during germ cell development are likely to contribute to many other disorders in which neither genetic nor environmental factors can account for all risk.

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article