Instructing Mouse Germ Cells to Adopt a Female Fate

Accurate and efficient production of gametes is critical for the transfer of genetic and epigenetic information to the next generation. Underlying gametogenesis is meiosis, a form of reductive cell division restricted to the germ cell lineage; meiosis is essential for conversion of germ cells from diploid to haploid [for reviews, see Handel and Schimenti, 2010; Kimble, 2011]. This process is necessarily complicated, with numerous steps involved, and is generally well conserved among eukaryotes. The exception to this conservation is the mechanism of meiotic onset which varies substantially between species [Kimble, 2011]. The purpose of this review is to revise progress in recent years regarding the adoption of the female fate in mammalian germ cells, focusing on work in the mouse and highlighting recent advances and remaining gaps in our understanding. We concentrate mainly on the question of how meiotic onset is orchestrated in germ cells of the fetal ovary, though we also touch on issues such as the acquisition of meiotic competence, the suppression of meiosis in germ cells of the fetal testes, and the maintenance of progression through meiotic prophase I.

In mammals, primordial germ cells (PGCs) are specified anew each generation. In the mouse, this begins in the posterior proximal epiblast, at around 6.25 dpc (days post coitum). Under the influence of local inductive factors produced by the surrounding somatic cells, predominantly bone morphogenetic proteins (BMPs) [Lawson et al., 1999; Saitou et al., 2002], a founding population of approximately 40 cells downregulates characteristic somatic genes, re-expresses or maintains expression of genes associated with pluripotency (such as Pou5f1,also known as Oct4, Sox2, and Nanog), and upregulates early germ cell-specific genes such as Prdm1 (also known as Blimp-1) and Dppa3 (also known as Stella) [reviewed by Saitou and Yamaji, 2012] (Fig. 1; for gene and protein names, see Table 1). At this stage, the PGCs are highly methylated. From 8.0 dpc, the PGCs migrate through the embryo, proliferating rapidly and heading towards the location of the future gonads. During their migration, the PGC begin the process of genome-wide demethylation (stage I DNA demethylation) [Seki et al., 2005; Seisenberger et al., 2012] that occurs primarily through replication-coupled passive demethylation, assisted perhaps by the repression of de novo DNA methyltransferases DMRT3a and DMRT3b at this time of specification [Kagiwada et al., 2013]. The genome is not completely demethylated during the stage I period, however, as certain loci including imprinting control regions and the promoters of genes involved in meiosis remain methylated [Seisenberger et al., 2012]. Retention of discrete sites of methylation despite stage I demethylation is orchestrated by the action of DMRT1, the only DNA methyltransferase still expressed at that time [Kagiwada et al., 2013; Hargan-Calvopina et al., 2016]. During migration, PGCs are also subjected to additional reprogramming, such as exchange of histone variants and erasure of histone modifications; these changes likely to contribute to the repression of somatic fate in the PGCs [Seki et al., 2007; Hajkova et al., 2008].

The PGCs finally lodge in the nascent bipotential gonads at approximately 10.5 dpc; early events after their arrival are critical for their commitment to gametogenesis [Nicholls et al., 2019]. At the time of PCG colonisation, the newly formed gonadal tissues are in a state of profound change, because a subset of their somatic cells (the ‘supporting’ cells that will become the Sertoli or granulosa cells) are undergoing commitment and sexual fate determination [reviewed by Stevant and Nef, 2019]. Once in the gonads, germ cells undergo stage II DNA demethylation, which involves TET1 and TET2 action, reaching a hypomethylated ‘epigenetic ground state’ by 12.5–13.5 dpc [Hajkova et al., 2002, 2008; Maatouk et al., 2006; Hajkova, 2011; Seisenberger et al., 2012; Hackett et al., 2013; Kagiwada et al., 2013; Yamaguchi et al., 2013]. As silencing marks are removed, parental imprints are erased, X chromosome reactivation occurs, and certain genes involved in meiotic prophase I either initiate low levels of expression or become vulnerable to induction. In neighbouring somatic cells, these classes of genes remain silenced by methylation [Maatouk et al., 2006]. It seems likely that stage II demethylation is responsible for the baseline expression of germ cell marker GCNA and meiotic protein SYCP3 in germ cells of both sexes upon entry into the gonads [Di Carlo et al., 2000; Chuma and Nakatsuji, 2001; Maatouk et al., 2006].

Another consequence of stage II demethylation is the onset of DAZL expression in germ cells of both sexes [Cooke et al., 1996; Seligman and Page, 1998; Maatouk et al., 2006; Hargan-Calvopina et al., 2016]. Expression of DAZL, an RNA-binding protein, is associated with a loss of cellular pluripotency and is necessary for germ cell determination (irreversible commitment to the germ cell lineage) [Chen et al., 2014; Nicholls et al., 2019] (Fig. 1). DAZL is also required in both sexes for acquisition of competence (licensing) to respond to developmental cues and undergo either female- or male-specific differentiation and gametogenesis [Ruggiu et al., 1997; Lin et al., 2008; Haston et al., 2009; Gill et al., 2011]. The onset of Dazl expression in post-migratory germ cells appears to depend on signals from the gonadal soma rather than to be a cell autonomous event, because PGCs in Gata4-null embryos that lack gonads never induce Dazl expression [Hu et al., 2015]. Shortly after they migrate into the gonads, germ cells also begin to express another RNA binding protein, the evolutionarily conserved germ cell marker DDX4 (also commonly known as MVH, mouse Vasa homologue). This is significant because DDX4 is a marker that distinguishes germ cells from pluripotent stem cells [Fujiwara et al., 1994; Tanaka et al., 2000].

In both developing ovary and developing testis germ cells stop proliferating at approximately 13.5 dpc. At this stage, mitotic germ cells take on a distinctive pre-meiotic morphology [McLaren and Southee, 1997] and are no longer considered PGCs, but are more correctly referred to as ‘oogonia’ or ‘M-prospermatogonia’ (M for multiplying) in the fetal ovary and testis, respectively (Fig. 2). In the fetal ovary, oogonia forsake the mitotic cell cycle to enter prophase of meiosis I, progressing through the leptotene, zygotene, and pachytene stages, arresting in late diplotene stage (dictyate) just prior to birth. The resulting primary oocytes are maintained in primordial follicles and do not develop further until the female reaches sexual maturity. In the fetal testis, M-prospermatogonia also stop proliferating, but they avoid entering meiosis, arresting in the quiescent G₀/G₁ stage of the mitotic cell cycle [McLaren, 2003]: at this stage they are called T1-prospermatogonia (T1 referring to primary transitional). Shortly after birth they resume mitotic activity (T2-prospermatogonia, T2 referring to secondary transitional) and, after a few rounds of mitotic replication, they give rise to type A spermatogonia. It is only at puberty that cells of the male germline finally initiate meiosis.

There have been conflicting theories as to how meiosis initiates in the fetal ovary and is avoided in the fetal testis. Historically, it was believed that both XX and XY germ cells in the mouse embryo are primed to enter meiosis by default, according to an intrinsic clock and without the need for any particular instructions from the soma. This opinion was based on observations that germ cells that do not safely arrive in the fetal gonad, but instead lodge in sites such as the adrenals or the mesonephros, embark on meiosis at about the same age (dpc) as do germ cells that colonise the fetal ovaries; meiosis also initiates in germ cells co-cultured with fetal lung cells [Zamboni and Upadhyay, 1983; McLaren, 1984; McLaren and Southee, 1997]. The prevailing theory was that a repressor of meiosis was necessarily active in the fetal testis and if any meiotic inducer was required then it must be present in male as well as fetal gonads, as well as in other fetal organs [McLaren and Southee, 1997; McLaren, 2003]. There were, however, some dissenting views: in particular, Byskov and colleagues provided evidence that secretions from the rete ovarii are essential for the initiation of meiosis in oogonia [Byskov, 1974]. In addition, recognition that oogonia enter meiosis in an anterior to posterior wave suggested that an extrinsic inductive signal from cells near or at the anterior end of developing gonad may initiate meiosis [Menke et al., 2003; Yao et al., 2003; Bullejos and Koopman, 2004].

In 2006, using the technique of cDNA subtractive hybridisation [Bowles et al., 2000; Menke and Page, 2002; Menke et al., 2003], 2 teams showed that meiosis was likely induced by the direct action of retinoic acid (RA), a signalling molecule present in the fetal ovarian environment, on pre-leptotene germ cells [Bowles et al., 2006; Koubova et al., 2006]. The female versus male fetal gonad comparison screens revealed that Stra8 (stimulated by retinoic acid gene 8) was specifically expressed in ovarian germ cells, beginning at 12.5 dpc, whilst Cyp26b1 was initially expressed in ovary and testis, but was strongly upregulated in somatic cells after they committed to a testicular fate and concomitantly downregulated in ovarian somatic cells after they committed to an ovarian fate. Because Stra8 is highly sensitive to induction by RA [Oulad-Abdelghani et al., 1996] and CYP26B1 degrades RA [White et al., 2000], these expression patterns suggested a system whereby endogenous levels of RA would build up once CYP26B1 was lost in the ovary and would be able to act directly on germ cells to induce Stra8 (Fig. 2). Although the level of RA detectable in the fetal ovary is low, there is measurably more RA present in the fetal ovary than the fetal testis, and differing availability of RA in the 2 structures can be correlated with expression levels of CYP26B1 [Bowles et al., 2006, 2016; MacLean et al., 2007; Trautmann et al., 2008]. These findings tied in with earlier studies, showing that RA accelerates entry into meiosis in rat fetal ovarian cultures [Livera et al., 2000].

When Cyp26b1 is genetically deleted, ectopic RA induces testicular germ cells to embark on prophase of meiosis I, rather than follow the normal course of development which is to avoid entering meiosis and, instead, to undergo mitotic arrest and remain quiescent in G₀/G₁ until after birth [McLaren, 2003]. Although XY germ cells certainly initiate ectopic meiosis in this mutant, with timing even slightly earlier than is normal in the wild-type fetal ovary, they do not progress beyond leptotene/zygotene and, ultimately, they undergo apoptosis prior to birth [Bowles et al., 2006; MacLean et al., 2007]. In this model, the XY germ cells initiate expression of Stra8, encoding STRA8, the critical pre-meiotic factor in both sexes, and also upregulate SYCP3 and other markers of meiosis [Baltus et al., 2006]. One known abnormality that may underlie their demise is that they fail to downregulate pluripotency markers, such as Pou5f1 and Sox2 [Bowles et al., 2010]. This is at least partially because of the actions of FGF9 which is produced at high levels in the developing testis [Barrios et al., 2010; Bowles et al., 2010] (Fig. 2).

Meiosis in testicular germ cells is also prevented by the eventual induction of the RNA-binding protein NANOS2, which degrades Dazl mRNA as well as meiosis-associated transcripts [Suzuki et al., 2010; Kato et al., 2016]. An apparent obligate partner of NANOS2, DND1, is also involved in this transcript degradation [Suzuki et al., 2016b]. DND1 is expressed shortly after PGC specification and is lost as germ cells enter meiosis, in both sexes [Ruthig et al., 2019]. The timing of Nanos2 expression is consistent with the idea that the NANOS2/DND1 complex takes over the task of protecting testicular germ cells from a meiotic fate as the RA-degrading actions of CYP26B1 begin to wane [Suzuki and Saga, 2008]. Interestingly, germ cells in both testis and ovary of the Cyp26b1-null embryo enter meiosis earlier than do those in the wild-type fetal ovary, presumably because the complete absence of CYP26B1 enzyme allows endogenous RA levels to reach a necessary threshold earlier [Bowles et al., 2006].

More recent studies have shown that RA also acts on gonadal germ cells to directly upregulate Rec8, a meiosis-specific component of the cohesion complex, and that this occurs independently of STRA8 [Koubova et al., 2014]. Similarly, Meiosin (meiosis initiator, previously hypothetical gene GM4969), is regulated by RA in a STRA8-independent manner [Ishiguro et al., 2020]. Meiosin expression initiates at the same time as Stra8 in germ cells of the fetal ovary. MEIOSIN is a transcription factor and binding partner of STRA8 in the germline of both sexes (see below).

Sensitivity of germ cells to RA, in terms of Stra8 expression, is highly time-restricted, and this likely reflects not only the presence of DAZL [Lin and Page, 2005; Gill et al., 2011] but also the need to achieve a permissive epigenetic state. It makes sense that there would be mechanisms to control the period of RA sensitivity, given the broad availability of this signalling molecule during embryogenesis. Thus far, we know several factors that use epigenetic means to influence the predisposition of fetal germ cells to embark on meiosis. The polycomb repressor proteins, PRC1 and PRC2, both large multiprotein complexes that act as chromatin modifiers, seem to be key epigenetic regulators of the sensitivity of the Stra8 promoter to RA induction. Yokobayashi et al. [2013] found that deletion of Rnf2 (encoding a central component of Polycomb Repressive Complex 1 [PRC1]) led to precocious expression of Stra8 (and other meiosis-associated genes including Rec8) in ovarian fetal germ cells, this occurring about 1 day earlier than normal. In wild-type 11.5 dpc germ cells of both sexes, the PRC1 complex physically binds to the Stra8 promoter which at this time is characterised by both active (H3K4me3) and inactive (H3K27me3) methylation marks, the latter being indicative of PRC2 activity. These results suggest that the Stra8 promoter is ‘poised’ in germ cells at this point of development. Two days later, at 13.5 dpc, both PRC1 binding and repressive H3K27me3 marks are lost from the Stra8 promoter, specifically in ovarian germ cells, concomitant with the onset of normal Stra8 expression. Overall, these findings suggest that PRC1 and PRC2 act to restrain Stra8 expression for several days after germ cells arrive in the gonad. Importantly, the precocious Stra8 expression observed upon genetic removal of the PRC1 component RNF2 is only observed in the presence of endogenous ovarian RA in this model. A recent study reports that germ cell-specific deletion of Snf5, encoding a core component of the SWI/SNF complex, results in impairment of sex-specific development of germ cells in both sexes [Ito et al., 2021]. This is consistent with evidence that the SWI/SNF acts to evict PRCs from H3K27me3-marked chromatin and thereby allows activation of gene expression.

It is known that treatment of embryonic stem (ES) cells with RA can induce them to express Stra8, however, this is not associated with entry into meiosis [Bouillet et al., 1995; Oulad-Abdelghani et al., 1996]. A recent finding may help explain why this is the case. Deletion of the gene encoding MAX (MYC associated factor X), a known partner of MYC family proteins, renders ES cells more sensitive to RA-induced ectopic Stra8 expression, and this was even accompanied by some upregulation of germ cell-associated genes (Dazl and Ddx4), some other meiosis-related genes (e.g., Sycp3), and some evidence of cytological changes reminiscent of very early steps in meiotic prophase I [Suzuki et al., 2016a]. These Max-null ES cells also expressed genes characteristic of 2-cell mouse embryos and showed evidence of erasure of imprints, suggesting MAX-associated repression is a rather general mechanism used in pluripotent cells to impede differentiation [Hishida et al., 2011]. Loss of MAX also increased propensity to express Stra8 and initiate the first steps of prophase I in other pluripotent cell types, such as germline stem cells (GSCs) and iPS cells (iPSCs), but not in somatic cells, such as mouse fibroblasts and NIH3T3 cells. As for the Rnf2-null model, loss of Max in germ cells of the fetal testis does not lead to ectopic Stra8 expression. In vivo, it appears that MAX is transiently lost as germ cells enter meiosis, but the mechanism for this is not known. MAX is a component of an atypical PRC1 form, PRC1.6, and, in the context of antagonising expression of meiotic genes, it seems that its partner protein is not the canonical MYC but MGA [Suzuki et al., 2016a]. These authors suggest that the effects of Rnf2 deletion, outlined above [Yokobayashi et al., 2013], may reflect an in vivo role for the PRC1.6 variant rather than PRC1 [Suzuki et al., 2016a].

As mentioned above, post phase I demethylation, DNMT1 is responsible for maintaining the methylation marks that are eventually removed in phase II demethylation. Accordingly, germ cell-specific loss of Dnmt1 results in precocious Stra8 expression and, thus, early activation of the meiotic program in the fetal ovary [Hargan-Calvopina et al., 2016]. As is the case for Rnf2-null and Max-null models, Stra8 comes to be expressed about 1 day earlier than normal. Testicular germ cells that lack DNMT1 also differentiate precociously, but there is no evidence that they upregulate Stra8 or embark on meiosis [Hargan-Calvopina et al., 2016]. Hence, for each of these examples it appears that lifting of repression is permissive rather than instructive: Stra8 expression is potentiated, but RA and perhaps some other component of the fetal ovarian environment is still required for its induction.

A transcription factor that is endogenously expressed in fetal germ cells, DMRT1 (Dsx- and mab-3 related transcription factor 1), has a role in ensuring efficient Stra8 expression in vivo in fetal ovaries [Krentz et al., 2011]. In Dmrt1-null ovaries, some germ cells express Stra8 at only low levels, though the effect is quite variable. Recently we showed in vitro that DMRT1 can indeed enhance Stra8 expression but that this occurs only when the DMRT1 binding site in the Stra8 proximal promoter has an adjacent intact RARE (retinoic acid-response element) [Feng et al., 2021]. This finding correlates well with evidence that DMRT1 is expressed not only in germ cells of the fetal ovary but also those of the fetal testis and could not, therefore, be expected to independently drive Stra8 expression [Krentz et al., 2011]. Presumably ovarian-specific availability of RA provides the necessary specificity in terms of Stra8 expression. Interestingly, DMRT1 antagonises Stra8 expression and meiotic onset in spermatogonia of the adult testis [Matson et al., 2010].

In recent years, it has proven possible to generate PGC-like cells (PGCLCs) from pluripotent cell types, ES (embryonic stem) and iPS (induced pluripotent stem) cells [Hayashi et al., 2011, 2012; Hikabe et al., 2016; Ohta et al., 2017], providing the opportunity to reconstitute the sex-determination pathway of germ cells in vitro, under defined conditions. Using this in vitro system, Miyauchi and colleagues showed that RA and STRA8 are not sufficient to direct the oogenic fate and that BMP signalling is also required [Miyauchi et al., 2017]. A requirement for BMP signalling in vivo is supported by the fact that genes encoding several BMP molecules are abundantly produced by ovarian fetal somatic cells: these include Bmp2, Bmp4, and Bmp5 [Jameson et al., 2012]. In particular, Bmp2 is strongly expressed by fetal ovarian somatic cells downstream of WNT4 signalling [Yao et al., 2004]. In vitro, BMP2, 4, 5, or 7 were each able to work synergistically with RA to push PGCLCs into a germ cell phenotype, characterised by downregulation of pre-gonadal PGC marker Prdm1 (Blimp-1) and induction of gonadal early germ cell markers Dazl and Ddx4 (Mvh). This requirement for BMP in ensuring some aspects of correct oogenesis ties in with earlier results, where oogenesis was compromised by deletion of the gene encoding a critical BMP and TGFβ transducer, SMAD4 [Wu et al., 2016].

Following the in vitro BMP study of Miyauchi et al. [2017], it was established that the key effector of BMP signalling, in the context of mouse fetal ovarian germ cells, is the GATA-like zinc finger transcription factor ZGLP1 [Nagaoka et al., 2020] (Fig. 2). ZGLP1 protein is detected as early as 12.0 dpc in fetal oogonia, whilst STRA8 protein is not normally detected until 13.0–13.5 dpc. Zglp1-null oocytes expressed DDX4, DAZL, and STRA8 but were not able to progress through meiotic prophase I. Analysis of the role of ZGLP1, using the PGCLC model, suggests that BMP signalling induces ZGLP1 expression, which launches the overall oogenic program and that RA augments this program once initiated [Nagaoka et al., 2020]. Although RA alone was able to efficiently upregulate Stra8 and Rec8, upregulation of other meiosis-associated genes required the presence of ZGLP1 as well. The new model proposes that ZGLP1 drives the oogenic program in general, whilst STRA8 is essential for the initiation of meiotic prophase and RA signalling plays a particular role in repression of the PGC program. In both germ cell-specific Smad4 deletion [Wu et al., 2016] and Zglp1 deletion [Nagaoka et al., 2020], ovarian germ cells that do not progress through meiosis instead undergo apoptosis, as would be expected for abnormal fetal germ cells.

It was previously suggested that signalling molecules WNT4 and RSPO1, produced by ovarian somatic cells and acting through β-catenin, promote germ cells to initiate meiosis [Naillat et al., 2010; Chassot et al., 2011], but, because both of these factors are important in ovarian somatic development, there was no clarity as to whether this was a direct effect on germ cells. In a new study, Ctnnb1 (encoding β-catenin) was deleted from the germline, and results suggest that rather than inducing meiosis, intrinsic β-catenin signalling ensures that oogonia maintain their pluripotency, i.e., WNT/β-catenin signalling prevents precocious Stra8 expression and meiotic onset [Le Rolle et al., 2021]. A role in maintenance of pluripotency in this context parallels a known role of WNT/β-catenin signalling in embryonic stem cells [Sato et al., 2004]. It seems that signalling through β-catenin must arrest before efficient Stra8 expression can occur. The mechanism for this arrest is proposed to rely on expression of ZNRF3, a known negative regulator of WNT/β-catenin signalling, but it is not yet clear how Znrf3 is upregulated in oogonia at the appropriate time. The model proposes that when β-catenin signalling is lost, oogonia gain the capacity to differentiate because pluripotency-associated POU5F1 is translocated from the nucleus to the cytoplasm where it forms a complex with β-catenin. No information was given in the recent study [Le Rolle et al., 2021] regarding the effects of deletion of β-catenin in testicular germ cells; this will be required in order to assess whether removal of WNT/β-catenin signalling is instructive or permissive, with respect to the onset of Stra8 expression and meiotic onset.

It is established that STRA8 is a critical master regulator of meiosis in both sexes, responsible for pre-meiotic DNA replication as well as for cohesion, synapsis, and recombination steps within meiotic prophase [Baltus et al., 2006; Anderson et al., 2008; Soh et al., 2015]. It was speculated that STRA8 acts as a bHLH (basic helix-loop-helix) transcription factor [Baltus et al., 2006; Tedesco et al., 2009; Choi et al., 2010; Soh et al., 2015], and this idea has been upheld by the results of recent RNA-Seq and ChIP-Seq analyses, where STRA8 was found to directly upregulate a large cohort of genes, binding primarily at their proximal promotor and recognising a motif of core sequence CNCCTCAG [Kojima et al., 2019]. One caveat of this study, with respect to our understanding of meiotic onset in the fetal ovary, is that it was conducted in preleptotene cells of the postnatal male, for reasons of feasibility. In the Kojima study, 1,351 genes were identified as direct transcriptional targets of STRA8. Interestingly, STRA8 seems to substantially amplify expression of genes that are already ‘on’ rather than initiating their expression de novo. Related to this is the observation that most STRA8 target promoters are characterised by CpG islands, sequences with a large number of CpG dinucleotide repeats that are subject to silencing by methylation. It is possible that these genes are released from strong repression during the second phase of demethylation upon gonadal colonisation (see above). Kojima et al. [2019] suggested that STRA8 may dimerise with another bHLH factor before it translocates to the nucleus to bind gene promoters. Accordingly, such a protein has recently been identified [Ishiguro et al., 2020]. MEIOSIN is induced by RA, and its expression initiates at the same time as STRA8 in germ cells of the fetal ovary. Like STRA8, MEIOSIN initiates expression of a large cohort of direct targets, including many meiotic genes that are involved in establishing the meiosis-specific chromosome structure. The current model holds that STRA8 and MEIOSIN work collaboratively to direct the mitosis-to-meiosis switch in both sexes.

After STRA8 and MEIOSIN are produced, germ cells of the fetal ovary enter and begin to progress through the highly extended prophase of meiosis I that, in the mouse fetal ovary, continues for approximately 1 week. STRA8 is only transiently expressed at the time of meiotic entry [Soh et al., 2015], but certain other germ cell-specific factors have been identified as playing critical roles in maintaining meiotic prophase I. One such protein is MEIOC [Abby et al., 2016; Soh et al., 2017]. Meioc (previously putative gene Gm1564) is a direct target of ZGLP1 [Kojima et al., 2019] though there is also some evidence that it is at least partially dependent on RA and STRA8 for its expression [Soh et al., 2017]. In the absence of MEIOC, germ cells still express STRA8, undergo pre-meiotic DNA replication, and enter meiosis, but progression through prophase I is severely compromised, with failure to progress past the zygotene stage. Remarkably, although some cells achieve the leptotene or zygotene stages reasonably normally, others exit meiotic prophase I and an aberrant metaphase state resembling mitosis is observed [Abby et al., 2016; Soh et al., 2017]. It appears that the Meioc-null cells are simultaneously running the mitotic and meiotic cell cycle programs [Soh et al., 2017]. MEIOC interacts with YTHDC2, an RNA helicase, and the Ythdc2-null phenotype in fetal ovarian germ cells is very similar to that of the Meioc-null [Abby et al., 2016; Bailey et al., 2017; Soh et al., 2017]. The MEIOC/YTHDC2 complex physically interacts with and post-transcriptionally regulates mRNAs: one group proposes that the MEIOC/YTHDC2 complex stabilizes mRNAs involved in the chromosomal program of meiosis, thus extending meiotic prophase I [Abby et al., 2016], while another proposes that this complex destabilizes mRNAs that encode mitotic cell cycle regulators [Soh et al., 2017].

Beginning in 2011, the requirement for RA to induce meiosis, in vivo, has been challenged. In the first such study, genes encoding 2 known RA-synthesising enzymes, aldehyde dehydrogenases ALDH1A2 and ALDH1A3, were genetically deleted [Kumar et al., 2011]. The rationale for focusing on these 2 enzymes was that they are both expressed in the mesonephros, which is adjacent to the ovary, and the postulated source of meiosis-inducing RA [Bowles et al., 2006; Bowles and Koopman, 2007]. As some germ cells in the XX gonad still began to express STRA8 and entered meiosis, it was claimed that RA was not required in vivo for meiotic onset. Nonetheless, an anti-meiotic function for CYP26B1 was still supported by these authors, and it was postulated that the enzyme plays a critical role that does not involve the degradation of RA [Kumar et al., 2011]. This work did not include genetic removal of Aldh1a1, encoding ALDH1A1, a less efficient RA synthesising enzyme, a point elaborated on in a later study [Bowles et al., 2016] and supported by evidence that ovarian RA also impacts on meiotic onset in the mouse [Mu et al., 2013] as well as the human [Le Bouffant et al., 2010; Childs et al., 2011]. In 2020, 2 more studies addressed the question of whether RA is required in vivo for meiotic onset in the fetal ovary of mice. In one, expression of all 3 Aldh1a genes was diminished using a CRE recombinase-based conditional approach [Chassot et al., 2020], whilst in the other, genes encoding retinoic acid receptors (RARs, which mediate the RA response) were deleted, completely in the case of RARβ and by use of tamoxifen-induced CRE recombinase for RARα and RARγ [Vernet et al., 2020]. These authors concluded that RA is dispensable for meiotic entry, though it does contribute to ensuring timely expression of Stra8.

Although these studies are certainly thought-provoking, there are several caveats in terms of interpretation. For example, attempts to delete genes using CRE recombinase systems may result in inconsistent, incomplete, and mosaic patterns of deletion, when compared to straight genetic deletions [Abram et al., 2014; Tian and Zhou, 2021]. Relevant to this is the knowledge that gonadal germ cells, when at the appropriate stage of development, are highly sensitive to low levels of RA in terms of their preparedness to express Stra8 [Bowles et al., 2010]. In terms of a requirement for RA in vivo, we note the 2009 study of Li and Claggett-Dame where it was shown that diminishing RA availability, by making pregnant rat dams vitamin A deficient (VAD), resulted in an oogonial population that failed to enter meiosis and remained undifferentiated [Li and Clagett-Dame, 2009]. Recently, we showed that subtle mutation of 2 RA response elements (RAREs) upstream of mouse Stra8, tested in the context of a 2.9-kb promoter-luciferase construct, is sufficient to completely ablate Stra8 expression in F9 (mouse embryonal carcinoma) cells [Feng et al., 2021]. Mutation of the same 2 sites in vivo resulted in loss of about 50% of Stra8 expression. It is not clear whether the residual Stra8 expression in the in vivo model reflects the presence of as yet unidentified but functional RAREs, perhaps further upstream on in the first intron of Stra8 [Feng et al., 2021]. When these results are considered together with the VAD studies we conclude that RA acts directly to drive the onset of Stra8 expression in vivo, whilst acknowledging that other extrinsic factors may also contribute to or influence this.

Although the concept of ‘adoption of the female fate’ for germ cells has historically been associated with observation of the onset of meiotic changes that begin at approximately 13.5 dpc in mouse ovarian germ cells, we now appreciate that meiotic onset and some other aspects of oogenesis are separable. Surprisingly, in Stra8-null female mouse ovaries, some oocyte-like cells were found to have developed, even though these cells had never entered meiosis (i.e., they developed zonae pellucidae, organized follicle structures, and could be induced to ovulate upon hormonal stimulation) [Dokshin et al., 2013]. Similarly, analysis of the Meiosin-null ovaries revealed oocyte-like cells that had apparently skipped meiotic prophase [Ishiguro et al., 2020]. Although detailed consideration of oocyte regulation and development, downstream of meiotic onset, are beyond the scope of this review, it is notable that recent studies have identified a core set of transcription factors that orchestrate oocyte growth at the transition from primordial to primary follicle stage [Hamazaki et al., 2021]. When expression of these transcription factors is forced in pluripotent stem cells (ES cells), oocyte-like cells that are competent for fertilization and cleavage are produced. Amazingly, specification of PGCs, epigenetic reprogramming, and meiosis can all be bypassed in this system. These unexpected findings have led to a new model of oogenesis in which the 2 key components, the chromosomal events of meiosis and cellular growth and differentiation, are recognised as dissociable processes.

Numerous mechanisms have been shown, over recent years, to have some impact on the onset and progression of meiosis. Ultimately, however, it is sex-specific mechanisms that are of greatest interest, as they are the ones critical for driving the adoption of female germ cell fate, working on a background of promoter accessibility for key early meiosis genes. Recent findings have extended our understanding of such mechanisms.

One of the most significant findings in recent years, with respect to the adoption of female germline fate, relates to the revelation that RA, alone, is not sufficient to direct efficient oogenesis in an in vitro model [Miyauchi et al., 2017; Nagaoka et al., 2020]. When mouse ES or mouse iPS cells are pushed through an epiblast-like cell state to PGC-like cells (PGCLCs) they do respond to RA, in terms of Stra8 expression and further meiotic progression, but this does not occur very efficiently, and addition of BMP to the culture greatly improves success. Fine dissection of the roles BMP signalling plays, at various steps in the process (i.e., germ cell commitment, meiotic onset, oogenetic progression, etc.) remains to be done, especially in the in vivo context. Whether ZGLP1 is the only effector of BMP signalling, during oogenetic onset and progression, also remains to be clarified. ZGLP1 may be considered a pioneer factor, because ChIP studies revealed that many of its target genes are in a poised or repressed state prior to their upregulation [Nagaoka et al., 2020]. This contrasts with STRA8 which seems to act by greatly upregulating many genes that are already expressed at low levels, at least in the context of male meiotic onset [Kojima et al., 2019].

A key new piece of information relates to the discovery that, in addition to STRA8 and REC8, RA also induces MEIOSIN, a transcription factor that acts together with STRA8 to ensure initiation of a robust meiotic programme. Mechanistic details underlying this apparent collaboration of 2 HLH-type transcription factors remain to be established.

Another interesting development is evidence that a third ovarian signalling molecule is also important. WNT4 is highly expressed by pre-granulosa cells of the developing ovary, and recent evidence suggests that it, or other members of the WNT family, acts through β-catenin in germ cells to ensure loss of POU5F1 function. This represents a novel method of ensuring a loss of pluripotency at the onset of meiosis; we already knew that Pou5f1 ceases by being expressed upon entry into meiotic prophase and that this occurs in an anterior to posterior wave [Pesce et al., 1998; Bullejos and Koopman, 2004], suggesting perhaps an association with RA availability. Intriguingly, the method of loss of POU5F1 function in germ cells of the fetal testis is distinct again, involving no attenuation of Pouf51 expression but, instead, post-translational repression as testicular germ cells arrest [Western et al., 2010].

Results over recent years have extended and supported the idea that regulatory regions of meiotic and oogenic genes in germ cells need to be in a permissive state before they can respond to signals from the fetal ovarian environment. As far as we know, phenomenon such as global demethylation, maturation, and licencing of germ cells, and the onset of low levels of meiotic gene expression are not sex specific and are therefore, presumably, merely permissive for meiotic onset. This means that sex-specific processes are required to ensure that ovarian germ cells enter meiosis in a timely fashion and that testicular germ cells do not. Presently, the best candidates for an instructive role in the fetal ovary are the signalling molecules RA and BMP2; details of their complementary roles in this context remain to be clarified.

The authors have no conflicts of interest to declare.

This work was funded by a Discovery Project grant from the Australian Research Council (DP200102896 to J.B. and C.S.). C.S. also acknowledges project funding support from Rebecca L Cooper Medical Research Foundation and the Kid’s Cancer Project.

C.S. and J.B. contributed to writing this review.