Germ cells, after fate determination as primordial germ cells (PGCs) in early embryos, undergo various unique changes in epigenetic status during their development, and these changes differ from the epigenetic changes occurring in any other somatic cells. For example, PGCs undergo demethylation of DNA and change histone modification states on a genome-wide scale. Although the full physiological significance of these epigenetic alterations is still unclear, we can now discuss some of their mechanisms due to recent experimental evidence demonstrating the expression of candidate molecules involved in the processes of epigenetic change. On the other hand, DNA demethylation associated with PGC-specific gene expression, reprogramming of imprinted genes and regulation of retrotransposons in PGCs differentially occur from the genome-wide DNA demethylation. The tendency of epigenetic changes to appear on the whole genome, as well as more precise changes in the epigenetic status of particular parts of the genome, may play important roles in establishing the properties of PGCs required for acquiring totipotency.
Among cells consisting of an animal body, germ cells are the only cell type capable of acquiring totipotency after fertilization, thereby guaranteeing the everlasting life cycle from one generation to the next. Mouse germ cells are first derived from a pluripotential epiblast, the source of all the subsequent parts of an embryo, as “fate-determined” primordial germ cells (PGCs). They initially form a cluster of about 30 cells prior to, and distant from, the prospective gonads at the early stage of embryogenesis (at embryonic day [E] 7.25) (Fig. 1) (Ginsburg et al. 1990). Once the fate of PGCs is determined, they are normally monopotent, and differentiate only to eggs or sperm, but not to any other types of cells. However, interestingly, PGCs develop into pluripotential embryonic germ (EG) cells in culture, indicating that they maintain the potential to be converted into pluripotential cells (Matsui et al. 1992). Shortly before fate determination, precursors of PGCs locate to form a cluster inside the proximal region of extra-embryonic mesoderm near the posterior end of the embryo (Fig. 1). Analyses of the gene expression profiles suggest that the PGC precursors are influenced by mesoderm induction signals, but they thereafter escape from the mesodermal fate and are committed into PGCs (E6.0–E7.25) (Okamura et al. 2003; Kurimoto et al. 2008). After fate determination, PGCs start rapidly proliferating and migrating to the genital ridges, i.e., the emerging gonads (Fig. 1) (Tam & Snow 1981; Anderson et al. 2000; Molyneaux et al. 2001; Runyan et al. 2006; Seki et al. 2007). By E8.5, approximately 120 PGCs are localized around the hindgut endoderm, and by E9.0 they are incorporated into the hindgut epithelium. PGCs emigrate from the dorsal aspect of the hindgut between E9.0 (∼150 PGCs) and E9.5 (∼250 PGCs), separate into left and right streams of individual cells, and migrate laterally across the dorsal body wall mesenchyme. At E10.5, approximately 1000 PGCs close to the genital ridges continue to migrate, singly or in clusters, into the genital ridges to form the primary sex cords, while ectopic PGCs remaining in the midline structures are eliminated by apoptosis (Runyan et al. 2006). In the genital ridges, PGCs still proliferate up to about 26,000 cells by E13.5, when their division stops, and they finally undergo male or female gametogenesis. During their development, the expression of a number of PGC-specific genes is induced, and recent studies have revealed that PGCs show various unique changes in epigenetic status, which are quite different from those that occur in any other somatic cells.
In the present review, we summarize the latest information concerning the development and differentiation of PGCs from the point of view of epigenetics, in particular the changes in DNA methylation and histone modifications, and discuss the possible mechanisms and physiological significance of epigenetic changes in PGCs.
Dynamic changes in epigenetic status during PGC development and differentiation
Mechanisms of genome-wide DNA demethylation in early PGCs
Soon after fate determination, PGCs begin to undergo genome-wide DNA demethylation (Fig. 1) (Seki et al. 2005; Popp et al. 2010). Evidence suggests that changes in the expression of DNA methyltransferases and their cofactors are involved in the genome-wide DNA demethylation in PGCs. In vertebrates, cytosine (C) residues at the CpG (5′-CG-3′) sites in DNA are methylated to generate 5-methylcytosines (C5me) by DNA methyltransferases, as an epigenetic modification. In PGC development, the expression of DNA methyltransferases specifically and dynamically changes. For example, a maintenance DNA methyltransferase, Dnmt1, by which methylation patterns of mother strands are maintained after DNA replication, is transiently downregulated at least in a portion of PGCs after the time of their fate determination, and is subsequently re-expressed by the time of their migration (E8.25) (Fig. 1) (Seki et al. 2005). In addition to that of Dnmt1, the expression of Np95, one of Dnmt1’s cofactors, is also repressed in emerging PGCs (E7.0) (Fig. 1) (Kurimoto et al. 2008). Np95 is known to recognize CpGs methylated only in the mother strand (i.e. hemi-methylated CpGs) and recruits Dnmt1 to these sites (Sharif et al. 2007). Two mechanisms of DNA demethylation, that is, a passive process (dependent on DNA replication) and an active process (independent of DNA replication), are possible; the former may occur in the early PGCs after their fate determination because of loss of both Dnmt1 and Np95, although we cannot exclude the possibility that the active process is also involved. Moreover, the expression of de novo DNA methyltransferases Dnmt3a and 3b, which newly methylate unmethylated DNA, are also downregulated just before fate determination (E7.0) (Fig. 1) (Seki et al. 2005; Yabuta et al. 2006). Some speculate that the low level expression of de novo Dnmt3a/3b may contribute to maintenance of the unmethylated DNA state generated by the preceding DNA demethylation.
From E7.75 to E9.0, when PGCs undergo migration towards the forming genital ridges, most PGCs stop their cell division at the G2 phase of the cell cycle (“G2 arrest”) (Fig. 1) (Seki et al. 2007). Immunohistochemistry has revealed that the levels of 5-methylcytosine (C5mepG) decrease in a portion of PGCs at E8.0 and in most PGCs at E9.5 (Seki et al. 2005). Thus, genome-wide DNA demethylation clearly progresses independently of DNA replication during G2 arrest. However, the molecular mechanisms involved in this process remain unclear. In recent years, the Gadd45 (growth arrest and DNA-damage-inducible protein 45) nuclear protein family, which is involved in regulation of genomic stability, DNA repair and cell proliferation, has been considered as a key player in DNA replication-independent active DNA demethylation by DNA excision repair (Barreto et al. 2007; Rai et al. 2008; Ma et al. 2009; Schmitz et al. 2009; Sen et al. 2010). In this model, it is assumed that a conversion of C5mepG into thymine (TpG) by a DNA deaminase gives rise to a thymine:guanine (T:G) mismatch, and an excision of this thymine (–pG) by DNA glycosylase followed by a replenishment of a cytosine (CpG) by DNA repair cancels the T:G mismatch. In this case, Gadd45 family molecules seem to recruit a DNA deaminase (AID) and a DNA glycosylase (Mbd4) to the C5mepG to demethylate and to mediate their interaction (Rai et al. 2008). Moreover, since Gadd45α is known to be upregulated specifically in PGCs at fate determination (E7.0-E7.25) (Fig. 1) (Kurimoto et al. 2008), the expression of Gadd45α is likely maintained in PGCs during their G2 arrest (E7.75-E9.0). In addition, a very recent study has indicated that AID deaminase is required for DNA demethylation of the pluripotency-related OCT4 and NANOG promoters and for their expression, through binding to these promoters when synchronous reprogramming of somatic cell nuclei is induced in interspecies heterokaryons between mouse embryonic stem (ES) cells and human fibroblasts (Bhutani et al. 2010). This strongly suggests that the upregulation of Oct4 and Nanog during induction of pluripotent stem (iPS) cells from fibroblasts (Takahashi & Yamanaka 2006; Okita et al. 2007; Takahashi et al. 2007; etc.) can also be attributed to active DNA demethylation of their promoters by AID. Furthermore, since AID is specifically and highly expressed in PGCs at least at E11.5 and E12.5 (Fig. 1) (Morgan et al. 2004), this gene is also likely to be expressed in PGCs during their G2 arrest from E7.75 to E9.0. The expression of Gadd45α as well as AID in PGCs implies their possible involvement in active DNA demethylation in PGCs. On the other hand, global DNA demethylation still occurs in AID-deficient PGCs, although it is less efficient than that in wild type PGCs (Popp et al. 2010), suggesting that other molecules and mechanisms may also be involved.
Considering these recent findings, the possible mechanisms of the genome-wide DNA demethylation in early PGCs can be summarized as follows. At fate determination, PGCs repress expression of all Dnmts and Np95, and begin genome-wide DNA demethylation, which may be either dependent or independent on DNA replication. Afterwards, DNA replication-independent active DNA demethylation may mainly progress in PGCs in the period of cell division arrest at the G2 phase from E7.75 to E9.0, even though the expression of the maintenance DNA methyltransferase Dnmt1 is resumed by E8.25.
Changes in histone modifications in early PGCs
Early PGCs concomitantly fluctuate their states of repressive histone modifications with genome-wide DNA demethylation. For example, the level of repressive histone H3 lysine 9 dimethylation (H3K9me2) starts to decrease in PGCs at E7.75 (Fig. 1), while that in the surrounding somatic cells does not significantly change (Seki et al. 2005, 2007). Since PGCs do enter G2 arrest (E7.75–E9.0), demethylation of H3K9me2 is most likely a DNA replication-independent process. The levels of expression of the Jumonji family molecules, including Jmjd1a, a demethylase for H3K9me1 and me2, in PGCs are not significantly different from the surrounding somatic cells (Seki et al. 2007), and whether histone demethylases such as Jmjd1a are involved in the demethylation of H3K9me2 in PGCs requires elucidation. Interestingly, expression of the H3K9me2 methyltransferases, Glp and G9a, is respectively repressed in PGCs at E7.25 (Yabuta et al. 2006) and E9.5 (Seki et al. 2007) (Fig. 1), suggesting that their repression contributes to the maintenance of the unmethylated state of H3K9me2 generated at earlier stages. In contrast, the level of repressive histone H3 lysine 27 trimethylation (H3K27me3) starts to increase in PGCs at E8.25 (Fig. 1) (Seki et al. 2005, 2007), and Ezh2, an H3K27me3 methyltransferase, may be involved in the elevation of H3K27me3, because it is stably expressed in PGCs at least by E8.25 (Fig. 1) (Yabuta et al. 2006).
Possible physiological significance of epigenetic changes in early PGCs
DNA methylation and H3K9 methylation often closely interact with each other. For example, in a human cancer cell line, deficiency of the maintenance DNA methyltransferase DNMT1 results in a genome-wide reduction in H3K9me2/me3, and transfection of murine Dnmt1 rescues this phenotype (Espada et al. 2004). In addition, in ES cells lacking the H3K9me2 methylase G9a a reduction in DNA methylation levels occurs at dozens of genomic loci, which is also rescued by the G9a-transgene (Ikegami et al. 2007). This suggests that in PGCs, the demethylation of both DNA and H3K9 occurs interdependently, and consequently that transcriptional repression is canceled epigenetically. In the migrating PGCs undergoing G2 arrest from E8.0 to E9.0, RNA polymerase II-dependent transcription is repressed (Fig. 1) (Seki et al. 2007); this might compensate for the de-repressive epigenetic state caused by the demethylation of DNA and H3K9. PGCs subsequently cancel the repression of RNA polymerase II-dependent transcription after the acquisition of H3K27me3 as a new repressive histone modification. Although the mechanisms involved in the G2 arrest of migrating PGCs are unknown, they are possibly linked to DNA demethylation, because Gadd45 family molecules, which are involved in DNA demethylation, bind to Cdc2 and inhibit interaction between Cdc2 and cyclin B1, leading to induced G2 arrest in cultured cells (Wang et al. 1999; Zhan et al. 1999; Vairapandi et al. 2002). The functions of the Gadd45 family molecules in G2 arrest and DNA demethylation of migrating PGCs are of great interest and require further investigation.
Although the detailed molecular mechanisms and physiological significance of these epigenetic events in developing PGCs remain obscure, genome-wide de-repression of chromatin and transcriptional repression during their growth arrest might reset their chromatin state, which is established during fertilization to the epiblast stage and might be favorable to somatic cell differentiation; such epigenetic reprogramming may conceivably contribute to future acquisition of totipotency in germ cells.
Epigenetic changes in late PGCs
Immunohistochemistry has shown that the levels of C5mepG in PGCs decrease at E12.5 compared with those at E9.5, clearly indicating that after reaching the genital ridges, PGCs undergo further genome-wide DNA demethylation (Fig. 1) (Seki et al. 2005). DNA methylation analysis based on the bisulfite-sequencing method (Popp et al. 2010) confirmed these findings, and further demonstrated that various genomic regions including introns, intergenic regions and transposable elements, are hypomethylated in PGCs at E13.5 compared with those in ES cells and embryonic tissues.
The state of histone modification also dynamically fluctuates in PGCs in the genital ridges. For instance, the levels of permissive histone H3 lysine 9 acetylation (H3K9ac) and H3K4me are elevated in PGCs by E10.5, and subsequently decrease to the same levels found in the surrounding somatic cells by E12.5 (Fig. 1) (Seki et al. 2005). Interestingly, the efficiency of pluripotential EG cell formation from PGCs drops off after E11.5 (Matsui & Tokitake 2009), which may imply that a decline in the levels of permissive histone modifications is a required chromatin status for EG cell formation. This possibility is supported by the fact that the presence of trichostatin A (TSA), an inhibitor of histone deacetylases (HDAC) in culture medium facilitates the formation of EG cells (Durcova-Hills et al. 2008). In contrast, the levels of repressive H3K9me3, H3K27me3 and H3K64me3 transiently decline in PGCs around E11.5–E12.0, and are afterwards re-elevated (Fig. 1) (Hajkova et al. 2008; Daujat et al. 2009). This suggests that in PGCs, the repressive histone modifications may suppress the genome-wide transcriptional activation likely caused by the preceding genome-wide DNA demethylation and elevation of the permissive histone modifications, and may contribute to the appropriate expression of a set of genes required for the proper development and differentiation of PGCs.
We have so far discussed regulation of the genome-wide epigenetic changes and their possible physiological significance during PGC development. Although the exact functions of the epigenetic changes are still largely obscure, it is expected that the roles of epigenetic regulation in PGC development will soon be demonstrated through manipulation of their epigenetic status by suppressing or overexpressing key epigenetic regulators in PGCs. Indeed, a recent study noted that growth of newborn pups and litter size was affected in AID-deficient mice in which the PGC genome had become relatively hypermethylated (Popp et al. 2010), although detailed analysis is required to confirm that the reproductive phenotype is caused by AID deficiency. On the other hand, the significance of the epigenetic changes in particular genomic regions, such as imprinted genes and retrotransposons in PGCs, has already been determined, and the DNA methylation profiles associated with these regions differ from those of genome-wide DNA demethylation as mentioned below.
Demethylation of imprinted genes in PGCs
Genomic imprinting is a phenomenon whereby the parent-of-origin of the particular gene is memorized, and as a result, either the paternal or maternal allele of the gene is expressed; these genes are called imprinted genes. Genomic imprinting is an epigenetic regulatory mechanism of gene expression found only in mammals among higher vertebrates. As a consequence of this mechanism, both paternal and maternal genomes are essential for proper development, and gynogenesis is normally prevented in mammals. Mono-allelic expression of imprinted genes is regulated by DNA methylation of differentially methylated regions (DMRs) located near or within the imprinted genes. Genomic imprinting is erased in PGCs by DNA demethylation of DMRs between E9.5 and E12.5 before or after the PGCs reach the genital ridges to ensure establishment of sex-specific new imprints during gametogenesis (Figs. 1 and 2) (Hajkova et al. 2002; Lee et al. 2002; Sato et al. 2003).
Methylation patterns in retrotransposons in PGCs
Retrotransposons are transposable elements whose translocation within genomes is mediated by reverse transcriptions. Intracisternal A-particle (IAP) is one of the retrotransposons and is found in numerous positions in the mouse genome. The CpGs of IAPs are only partially demethylated in PGCs (Figs 1 and 2), which seems to be contrary to the genome-wide demethylation pattern (Hajkova et al. 2002; Lane et al. 2003). The partial demethylation of IAPs may be involved in prevention of insertion mutations in the PGC genome, which is caused by the transposition of activated IAPs.
In addition to the erasure of genomic imprinting and the repression of retrotransposons, the regulation of PGC-specific gene expression is also associated with epigenetic alterations as discussed below.
Epigenetic regulation of gene expression in PGCs
Epigenetic requirements for PGC-specific gene expression at their fate determination
The epigenetic regulation of gene expression during PGC determination from epiblasts in vivo is rarely studied due to limited cell numbers. On the other hand, important information has been accumulated from experiments using PGC-like cells induced in vitro from pluripotent stem cells. Such PGC-like cells are similar to in vivo PGCs with respect to upregulation of PGC-marker genes, fluctuation of repressive histone modifications such as H3K9me2 to H3K27me3, and erasure of genomic imprinting, although their ability to form functional gametes has not been shown to date.
Recently, epiblast stem cells (EpiSCs), a type of pluripotent stem cell line, were established from epiblasts (Brons et al. 2007; Tesar et al. 2007). EpiSCs are a useful in vitro model of epiblasts because their gene expression profile is closely correlated with that of epiblasts, while it is quite different from that of ES cells derived from the inner cell mass (ICM) of blastocysts (Brons et al. 2007; Tesar et al. 2007). In EpiSCs as well as epiblasts, the expression of germline-specific Stella and pluripotency-related Rex1 and Fbx15, which are expressed in ICM and ES cells, is not observed, and CpGs of the flanking regions of these genes are hypermethylated, while they become hypomethylated after PGC-like cells are induced (Hayashi & Surani 2009). The flanking regions of these genes are also hypomethylated in ES cells as in the EpiSC-derived PGC-like cells (Imamura et al. 2006). Moreover, the flanking region of Stella has permissive histone modifications such as higher levels of H3K9ac/H3K4me3 and lower levels of H3K27me3 in ES cells (Hayashi et al. 2008). These in vitro observations suggest that PGC-specific genes such as Stella, Rex1 and Fbx15 may also become hypomethylated in their flanking regions in PGCs in vivo at their fate determination (Fig. 2).
In addition to Rex1 and Fbx15, many pluripotency-related genes are expressed specifically in PGCs at fate determination (Yabuta et al. 2006; Kurimoto et al. 2008). For example, after Oct4 is continuously expressed from zygotes to epiblasts, its expression is restricted in PGCs, whereas it is downregulated in the somatic cells after fate determination of PGCs at E7.5. The expression of Nanog is also maintained specifically in fate-restricted PGCs around E7.0. On the other hand, Sox2 is transiently downregulated in epiblasts at E6.25, and is subsequently upregulated in PGCs around E7.0. Consistent with these expression patterns, Oct4 plays important roles in PGC determination (Okamura et al. 2008) as well as in the survival of migrating PGCs (Kehler et al. 2004), and Nanog is involved in PGC survival and/or proliferation (Chambers et al. 2007; Yamaguchi et al. 2009).
Correlation of the expression of these genes, and methylation of the CpGs of their flanking regions, have been shown using cultured cell lines (Imamura et al. 2006), and in case of Oct4 and Nanog expression, suppression by CpG-hypermethylation of their flanking regions has been demonstrated (Hattori et al. 2004, 2007). In addition, suppression of Oct4 expression by GCNF-mediated CpG-hypermethylation of its flanking region in differentiated ES cells has been reported (Gu et al. 2006; Sato et al. 2006). These data suggest that CpG-demethylation of the flanking regions may be involved in maintenance of the expression of Oct4 and Nanog, or upregulation of Sox2, in developing PGCs (Fig. 2). A recent report showing continuous CpG-hypomethylation of the flanking region of Oct4 during PGC-like cell induction from EpiSCs supports this idea (Hayashi & Surani 2009).
The possible involvement of histone modifications in PGC-specific gene expression has also been reported. In PGC-like cells induced from ES cells, the flanking regions of VASA, SCP3, OCT4 and NANOG come to have permissive histone modifications such as higher levels of H3K4me2 and lower levels of H3K9me2/H3K27me3 (Fig. 2) (Tilgner et al. 2008).
Taking together this experimental evidence strongly suggests that the epigenetic changes in PGC-specific genes and pluripotency-related genes play important roles in the expression of said genes during PGC determination. In addition, such epigenetic changes in PGCs are also likely involved in PGC determination itself, via regulation of specific gene expression as well as establishment of a unique chromatin status pos-sibly required for reprogramming the PGC genome. Future studies will shed light on these very interesting possibilities.
Epigenetic requirements for PGC-specific gene expression during migration towards, and colonization in, the genital ridges
Although developing PGCs undergo genome-wide DNA demethylation, DNA demethylation associated with the expression of some germline-specific genes also occurs at different times. Mvh, Dazl and Scp3 start their expression in PGCs between E10.5 and E11.5, when the PGCs reach the genital ridges, and DNA demethylation of the flanking regions of these genes concomitantly occurs (Maatouk et al. 2006). In the case of Dazl, its expression is regulated by DNA demethylation of an Sp1-binding site in its flanking region in porcine PGCs (Linher et al. 2009). The Sp1-binding site in the flanking regions of this gene is also found in mice and humans (Linher et al. 2009), suggesting that mammalian species share similar epigenetic regulation of this gene via the Sp1-binding site.
The possible importance of DNA demethylation for PGC-specific gene expression has also been suggested in other genes. For example, Dazl, Mvh, Mage-b4 as well as GCNA1 normally start their expression at E10.5–E11.5 in PGCs, while they are prematurely expressed at E9.5 in Dnmt1-deficient embryos (Maatouk et al. 2006). In addition, Ant4 is specifically expressed in germline cells after E11.5, and CpGs of its flanking region are hypomethylated in spermatogonia (Suzuki et al. 2007). In vitro analyses have demonstrated that DNA demethylation of the flanking region of Ant4 activates its expression (Rodic et al. 2005; Suzuki et al. 2007). Taken together, these results strongly suggest that the expression of a number of germline-specific genes is regulated by DNA demethylation (Fig. 2).
Histone modification is also known to play a role in the regulation of a germline-specific gene Dhx38. This gene is a putative binding target of the Blimp1-Prmt5 complex, which mediates dimethylation of histone H2A arginine 3 (H2AR3) and histone H4 arginine 3 (H4R3), and is expressed in PGCs after E12.5 (Ancelin et al. 2006). The level of H2A/H4R3me2 is relatively high in PGCs by E8.5, but subsequently decreases due to the translocation of Blimp1-Prmt5 from the nucleus to the cytoplasm around E11.5 (Fig. 1) (Ancelin et al. 2006), suggesting that the methylation of H2A/H4R3 by Blimp1-Prmt5 represses the expression of Dhx38 in PGCs by E12.5 (Fig. 2).Therefore PGCs likely use DNA methylation as well as histone modifications for the timely expression of PGC-specific genes during their development (Fig. 2).
Conclusions and outlook
Primordial germ cells undergo unique epigenetic changes during their development. These epigenetic changes play important roles in PGC-specific gene expression, reprogramming of imprinted genes and repression of retrotransposons in PGCs, and are also likely involved in the establishment of the germ cell-specific chromatin context, which may be necessary for germ cells to acquire their totipotency. However, the experimental evidence demonstrating the roles of epigenetic alteration in PGCs is still fragmentary and/or descriptive in nature, and further comprehensive studies are required to elucidate the functions of the unique epigenome in PGCs. To that end, technical breakthroughs such as epigenetic analyses adaptable to a very small number of cells, such as in vivo PGCs, are essential. After clarifying the epigenetic mechanisms that regulate germ cell development, we may realize the conversion of somatic cells into germ cells by epigenetic manipulation in the future.
K. Mochizuki was supported by Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists. Y. Matsui was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan.