Identification of regulatory elements required for Stra8 expression in fetal ovarian germ cells of the mouse

The germline provides the genetic and epigenetic link between generations. In order to produce haploid germ cells (oocytes and sperm, respectively, for females and males), diploid germ cells must undergo a specialized form of cell division known as meiosis. Meiosis initiates at different life stages in the two sexes: in females, it is initiated in germ cells during fetal development, shortly after gonadal sex determination, whereas, in the male, meiosis is first triggered postnatally, just before puberty. Stimulated by retinoic acid gene 8 (Stra8) is considered to be the crucial master regulator of meiosis in mammals because Stra8-null mice of both sexes are infertile due to the inability of germ cells to enter and progress through meiosis (Anderson et al., 2008; Baltus et al., 2006; Mark et al., 2008). Recent studies suggest that STRA8 acts as a transcription factor and regulates the expression of thousands of genes during the initiation of meiosis (Kojima et al., 2019). Despite the importance of timely onset of Stra8 expression in pre-meiotic cells of both sexes, the details of how it is induced in germ cells at the appropriate time are not yet clear.

The expression of Stra8 is highly cell-type specific and is temporally controlled, and there is much evidence that the signalling molecule retinoic acid (RA) is involved in inducing its expression (Griswold et al., 2012). In vivo, Stra8 is expressed exclusively in the germline during normal development. In the fetal ovary, Stra8 transcripts are detected from 12.5 days post coitum (dpc) in pre-meiotic germ cells, and Stra8 expression follows an anterior to posterior ‘wave’ running from 12.5 to 16.5 dpc (Bowles et al., 2006; Menke et al., 2003). This ‘wave’ pattern is also seen in embryonic ovaries of the RA responsive elements (RARE)-lacZ mouse line, which reports RA activity (Bowles et al., 2016, 2006; Rossant et al., 1991). Germ cells of the male fetal testis do not express Stra8 (Menke et al., 2003); in the male germline Stra8 is first expressed in mitotic cells of the postnatal testis and then in pre-meiotic spermatogonia throughout life (Oulad-Abdelghani et al., 1996). In vitro, only various pluripotent cell lines [embryonic stem (ES), embryonal carcinoma (EC), embryonic germ (EG) and cultured spermatogenic stem cells (SSC)] respond to RA by upregulating Stra8 (Oulad-Abdelghani et al., 1996; Wang et al., 2016). In other cell lines, and in all somatic cell types, Stra8 expression appears insensitive to induction by RA, suggesting that the Stra8 locus is tightly held in an epigenetically ‘closed’ configuration and/or that crucial co-regulatory factors are not present in such cells. It makes sense that additional mechanisms would be required to regulate sensitivity to RA, in terms of Stra8 expression, given that RA is widely available during embryonic and adult life. Aberrant expression of Stra8 has been observed in a somatic cancer (Kuang et al., 2019), perhaps indicative of the importance of stringent silencing of Stra8 expression.

For mouse Stra8, multiple silencing mechanisms have been identified, including promoter methylation, the actions of polycomb repressive complex 1 (PRC1) and class I/II histone deacetylase (HDAC) function. Germ cell-specific deletion of the DNA methylase enzyme DNMT1 revealed that the Stra8 promoter is normally methylated and repressed for several days after germ cells colonize the fetal gonad (Hargan-Calvopina et al., 2016). Similarly, silencing by PRC1, and specifically by the repressive mark H3K27me3, is required to prevent precocious Stra8 expression in germ cells of the fetal ovary (Yokobayashi et al., 2013). It has also been demonstrated that interference with class I/II HDAC function, by in utero exposure to trichostatin-A (TSA), leads to precocious Stra8 expression in germ cells of the fetal ovary (Wang and Tilly, 2010). Presumably, germ cells become more RA sensitive than normal, in terms of Stra8 expression, when epigenetic silencing is lost. Importantly, in none of the above-mentioned scenarios did loss of repressive epigenetic marks lead to ectopic Stra8 expression and meiotic onset in germ cells resident in the fetal testis, indicating that these epigenetic mechanisms are permissive, not instructive, in terms of Stra8 expression. On the other hand, one potential instructive regulator of Stra8 expression is the transcription factor doublesex and mab-3 related transcription factor 1 (DMRT1). DMRT1 can bind to the Stra8 promoter and, when Dmrt1 is ablated, reduced Stra8 expression is observed in the fetal ovary (Krentz et al., 2011). Furthermore, recent studies have shown that bone morphogenetic protein (BMP) signalling, using the transcriptional regulator ZGLP1 as its effector, can positively impact on Stra8 expression, among other functions crucial for the oogenic programme, in a mouse primordial germ cell-like (mPGCLC) model (Miyauchi et al., 2017; Nagaoka et al., 2020).

There has been considerable debate over the question of whether RA is required to trigger meiotic onset in the fetal ovary of the mouse. Treatment of gonadal or germ cell cultures with RA, RA receptor (RAR) agonists and RAR antagonists, shows changes in Stra8 expression that are consistent with the hypothesis that RA induces Stra8 expression (Bowles et al., 2010, 2006; Koubova et al., 2006; MacLean et al., 2007; Mu et al., 2013; Ohta et al., 2010; Tedesco et al., 2013; Trautmann et al., 2008). Furthermore, deletion of the RA-degrading enzyme CYP26B1 in embryonic testes allows germ cells to ectopically express Stra8 (Bowles et al., 2006; MacLean et al., 2007). Importantly, in vivo studies carried out in rats demonstrated that RA is not just sufficient but also required for normal meiotic onset because, in the absence of the RA precursor vitamin A, ovarian germ cells remained undifferentiated, as judged by their ongoing expression of the pluripotency-associated marker POU5F1 (also known as OCT4, Li and Clagett-Dame, 2009). On the other hand, concerns about the validity of the hypothesis that RA induces meiosis in the fetal ovary were raised when the deletion of two genes encoding RA-synthesising enzymes Aldh1a2 and Aldh1a3 was not sufficient to ablate all Stra8 expression (Kumar et al., 2011). This led to the suggestion that meiotic initiation in germ cells of the fetal ovary is independent of RA. Although we now know that another RA-synthesising enzyme, encoded by Aldh1a1, is also present and able to produce RA and induce Stra8 in the fetal ovary (Bowles et al., 2016), there have been additional recent studies that demonstrate that depletion of RA-producing enzymes, or RARs, does not completely ablate Stra8 expression, and the controversy continues (Bellutti et al., 2019; Chassot et al., 2020; Kumar et al., 2013; Raverdeau et al., 2012; Vernet et al., 2020).

Hence, despite its crucial role in fertility, we are still very unclear about how Stra8 is regulated. In addition to the need for Stra8 to be expressed in a female-specific manner during fetal life, the exact timing of the onset and duration of Stra8 expression is likely crucial, as is its complete silencing in somatic cells. To provide further information that might help resolve the issue of whether RA signalling is necessary or dispensable for Stra8 expression, we aimed to identify upstream regulatory elements involved in the endogenous induction of Stra8 expression in fetal ovarian germ cells. We created a transgenic line in which 2.9 kb of the mouse Stra8 promoter drives eGFP expression in the characteristic anterior to posterior wave of germ cell-specific expression. We then identified functional RAREs in the 2.9 kb fragment using F9 EC cells as a model for fetal germ cells. Finally, we subtly mutated two RAREs, singly and in tandem, demonstrating substantially diminished Stra8 expression, in vivo.

RA regulates transcription by acting as a ligand for nuclear RA receptors (RARs), which partner with retinoid X receptors (RXRs) to form heterodimers. Binding of the RA/RAR/RXR complex to RAREs, typically located in regulatory regions upstream of direct target genes, recruits nuclear receptor co-activators or co-repressors, thereby directly activating or repressing transcription of the associated gene (Mark et al., 2006). RAREs are generally direct repeats (DR) of the hexanucleotide sequence [5′-(A/G)G(G/T)TCA-3′] separated by 1 (DR1), 2 (DR2) or 5 (DR5) nucleotides (Rochel and Moras, 2014). Others previously noted two potential RAREs in the 400 bp proximal Stra8 promoter, one designated RARE1 (DR2 type, supported by ChIP experiments, Giuili et al., 2002; Kumar et al., 2011; Raverdeau et al., 2012), and the other designated as RARE2 (inverted repeat, IR5 type, Giuili et al., 2002), and one or both of these were found to be necessary for activation of the promoter by RA in F9 EC cells (Wang and Tilly, 2010). An additional putative RARE (DR4 type) has been noted ∼800 bp upstream of the transcription start site (TSS) (Raverdeau et al., 2012).

Using MatInspector, we analysed 5 kb of genomic DNA sequence upstream of the mouse Stra8 TSS (Fig. S1). In addition to RARE1, we found numerous putative RAREs, but our analysis did not predict the unnamed RARE (DR4) (Raverdeau et al., 2012), nor RARE2 (Giuili et al., 2002). We did identify a putative RARE (DR1 type) partially overlapping with the previously noted IR5 type RARE2 site. We also identified a putative DMRT1-binding site that lies within the DMRT1-binding region previously identified by ChIP (Krentz et al., 2011), as well as others further upstream.

When others made transgenic mouse lines, using random genomic integration, they were unable to detect transgene activation in fetal female germ cells using either 0.4 kb (Giuili et al., 2002) or 1.4 kb Stra8 genomic fragments (Antonangeli et al., 2009; Nayernia et al., 2004; Sadate-Ngatchou et al., 2008), despite strong induction in the adult male testis. To determine whether the lack of induction in germ cells of the fetal ovary could be due to genomic integration effects, or insufficient regulatory sequence, we generated three lines of mice, each harbouring a single targeted copy of a transgene comprising eGFP driven by ∼1.4, ∼1.5 or ∼2.9 kb of Stra8 putative promoter sequence (Fig. 1A). The 1.5 kb version was included because a previous study in F9 cells demonstrated that the 1.5 kb Stra8 upstream genomic fragment has significant promoter activity (Kwon et al., 2014); the 2.9 kb Stra8 promoter was the largest promoter length that we successfully inserted into the Col1a1 locus. Because all three transgenes are inserted single-copy at the same defined open autosomal locus, 3′ to the Col1a1 gene, expression levels in the three independent lines can be compared directly (Beard et al., 2006).

All three lines showed eGFP expression in the adult testes (Fig. S2) in a pattern similar to that of previously published transgene studies using the 0.4 kb and 1.4 kb Stra8 promoter lengths (Giuili et al., 2002; Sadate-Ngatchou et al., 2008). Consistent with published reports, the 1.4 kb promoter fragments did not drive visible eGFP expression in the developing ovary, and this was also the case for the slightly larger 1.5 kb promoter fragment. However, the 2.9 kbStra8-eGFP transgene drove strong eGFP expression in fetal ovaries that could be visualized directly and proceeded in a dynamic anterior to posterior wave (Fig. 1B) similar to that well-established for endogenous Stra8 expression (Bowles et al., 2006; Fu et al., 2017; Menke et al., 2003). Despite the fact that visible eGFP expression could not be detected for the 1.4 kb and 1.5 kb fragments, eGFP transcript detection was higher in fetal ovarian tissue than in testicular tissue at 13.5 dpc, as revealed by qRT-PCR analysis (Fig. 1C). Analysis of tissue sections of 2.9 kb ovaries by immunofluorescence showed eGFP staining only in germ cells (marked by DEAD box polypeptide 4, DDX4, Fig. 1D) and not in cells marked by the ovarian somatic cell marker forkhead box L2 (FOXL2; Fig. 1E). A close correlation was observed between the cytoplasmic eGFP and nuclear staining for endogenous STRA8 (Fig. 1F). When we quantified the co-expression of endogenous STRA8 and eGFP at 13.5 dpc, we found that eGFP was detectable in ∼68% of STRA8⁺ germ cells, and there were no eGFP⁺/STRA8⁻ germ cells (Fig. S3A,B). Observations of 12.5 to 16.5 dpc 2.9 kbStra8-eGFP whole embryos under a fluorescence microscope, before and during dissection, did not reveal any significant eGFP signal outside of the developing ovary.

Together, these results demonstrate that the 2.9 kb genomic region immediately upstream of the Stra8 TSS is sufficient to drive robust germ cell-specific eGFP expression in the mouse fetal ovary. Furthermore, we can conclude that regulatory elements included in this 2.9 kb fragment are sufficient to recapitulate the normal anterior to posterior ‘wave’ pattern of endogenous Stra8 expression, with only a slight temporal lag in eGFP production compared with endogenous STRA8. Element(s) that respond in a sex-specific manner are present in the 1.4 kb and 1.5 kb fragments, as they drive higher levels of expression in the fetal ovary than fetal testis (Fig. 1C). In addition, our results indicate that element(s) crucial for achieving high levels of expression in germ cells of the fetal ovary lie in the region between 1.5 kb and 2.9 kb upstream of the Stra8 TSS.

Exploiting our finding that the 2.9 kb promoter fragment is capable of driving expression in vivo, we aimed to further characterise the functional elements of the Stra8 promoter by in vitro analysis. To this end, we used the same promoter fragments (1.4, 1.5 and 2.9 kb) to drive firefly luciferase in the F9 EC cell line: we wanted to establish whether this line could be used to model ovarian germ cells with respect to Stra8 expression. Transiently transfected F9 cells were cultured in medium containing either 50 nM RA or carrier (DMSO) for 18 h before assaying for luciferase activity. No significant differences were observed between the 1.4, 1.5 and 2.9 kb constructs treated with DMSO; however, in the presence of RA, all three promoters displayed substantially more luciferase activity than their respective controls (Fig. 2A, red hatched bars compared with dark grey bars).

We next investigated the possible regulation of Stra8 by DMRT1 in this F9 model system. First, we tested whether RA induces endogenous Dmrt1 expression in the F9 cell line: Dmrt1 expression was basal in F9 cells and not substantially increased by RA treatment, and levels of Dmrt1 expression were less than 10% of that detected in mouse fetal ovary tissue at 12.5 dpc (Fig. S4; Laursen et al., 2012). For all three Stra8 promoter constructs, co-transfection with a Dmrt1 expression construct alone did not significantly increase luciferase activity (Fig. 2A, light grey bars). However, in the presence of both RA and the Dmrt1 expression construct, we observed an augmentation of luciferase activity and, as our in vivo studies would predict, the 2.9 kb fragment directed significantly more luciferase activity than its 1.4 kb and 1.5 kb counterparts (Fig. 2A, solid red bars). This apparent synergistic effect of DMRT1 and RA on the Stra8 promoter constructs is consistent with evidence that DMRT1 can enhance Stra8 expression in the fetal ovary (Krentz et al., 2011). As DMRT1 has a repressive effect on Stra8 in the postnatal testis (Matson et al., 2010), the fetal ovarian-like response of our Stra8 promoters in the presence of DMRT1 indicates that our approach of testing Stra8 promoter constructs in F9 EC cells is suitable for modelling the regulation of Stra8 in fetal ovarian germ cells (summarised in Fig. 2B).

We systematically decreased the length of the 2.9 kb Stra8 promoter construct to define the possible regulatory regions contributing to its greater sensitivity to RA, and continued to use F9 EC cells as an in vitro model system. No significant reduction in promoter activity was observed until the length was reduced from 1653 to 1550 bp (Fig. 2C). Within the intervening 103 bp region we identified a single putative RARE that we designated RARE3 (as RARE1 and RARE2 had previously been assigned for this promoter, Giuili et al., 2002).

To confirm RARE3 functionality in our in vitro F9 system, we mutated the site in the context of the full 2.9 kb Stra8 luciferase construct and observed more than a 50% reduction in promoter activity (Fig. 2D). In a parallel experiment we also mutated RARE1, an element reported to bind RA receptors in ChIP experiments (Kumar et al., 2011; Raverdeau et al., 2012), and found a similar reduction in promoter activity. When both RAREs were simultaneously mutated the promoter no longer responded to RA in this system (Fig. 2D). These results show that both RARE1 and RARE3 are involved in regulating Stra8 expression and, at least in our in vitro system, subtle mutation of both completely abolishes promoter activity. It also suggests that no other putative RAREs identified within the 2.9 kb region can direct expression in this system.

We next sought to address whether the synergistic effect of DMRT1 and RA (depicted in Fig. 2A) is specific to RARE1 (which is proximal to a ChIP-verified DMRT1 binding site; Fig. 1A; Murphy et al., 2010) or the more distal RARE3. We compared the unaltered and the RARE-mutated 2.9 kb Stra8 promoter luciferase constructs in the presence or absence of the Dmrt1 expression construct (Fig. 3A). We found that DMRT1 was still able to enhance the effects of RA on luciferase activity when RARE3 was mutated, but not when RARE1 was mutated (Fig. 3A, red hatched compared with solid red bars for RARE1 and RARE3 constructs). These results suggest that the synergism between DMRT1 and RA signalling is specific to the RARE1 site in this system.

We also mutated the DMRT1 binding site to test whether its loss has any effect on the ability of RA to promote transcription at the RARE1 site. DMRT1 was unable to promote transcription when either the RARE1 or the DMRT1 site were mutated (Fig. 3B). The ability of RA to elicit expression was retained when the DMRT1 site was mutated, and the level of activation was similar to that seen when RARE1 was mutated. This may mean that destruction of the DMRT1 site prevents signalling through the nearby RARE1 and that the observed RA response is due only to signalling through the intact RARE3. The lack of complete ablation of promoter activity when the DMRT1 site is mutated is consistent with reports showing that DMRT1-null females retain substantial Stra8 expression and are fertile (Krentz et al., 2011). Overall, these results support the concept that RA plays a dominant instructive role and DMRT1 plays a supportive or enhancing role in regulating Stra8 expression in the F9 cell culture model, acting together with RA at the RARE1 site (Fig. 3C).

Results from our in vitro F9 cell studies suggest that RA acts directly to activate the 2.9 kb Stra8 promoter, utilising both RARE1 and RARE3 sites, and that enhanced activity conferred by DMRT1 relies on an intact RARE1. To test whether these findings hold in vivo, we designed CRISPR/Cas9 reagents to specifically and subtly mutate the RAREs in the genomic Stra8 promoter (Fig. S5A,B). Three mouse lines were established in which RARE1, RARE3 or both RAREs (RARE1/3) were mutated to reproduce the sequence changes employed in our in vitro constructs.

qRT-PCR analysis of 14.5 dpc ovaries from embryos with homozygous mutations of RARE1 revealed a significantly compromised Stra8 transcription when compared with wild-type littermates (Fig. 4A; Fig. S6). This result correlates well with our observations regarding the importance of RARE1 for Stra8 expression in vitro. Expression of Rec8, a meiotic marker induced by RA independently of STRA8 (Koubova et al., 2014), was not significantly altered, as expected. The diminished Stra8 expression did not affect expression of meiotic markers Sycp3 and Dmc1 but Pou5f1, encoding the pluripotency marker POU5F1, was expressed at elevated levels, suggesting that differentiation of germ cells was compromised.

We examined Stra8 expression in 14.5 dpc ovaries of the RARE3 mutant line and found that homozygous RARE3 mutants also had diminished levels of Stra8 expression (Fig. 4B; Fig. S6), but the magnitude of the change was less than we observed for the RARE1 mutant, and less than predicted by our in vitro studies (Fig. 2D). As was the case for the RARE1 mutants, no significant changes were observed in Rec8, Sycp3 and Dmc1 expression, but Pou5f1 expression was elevated.

We next analysed mice with mutations in both RARE1 and RARE3. Ovaries of 14.5 dpc embryos with the double mutation showed a more pronounced loss of Stra8 expression (Fig. 4C; Fig. S6) but, unlike the case in vitro, we did not see a complete loss of Stra8 expression. The retention of ∼45% of Stra8 expression may indicate that active positive regulatory elements lie beyond the 2.9 kb promoter fragment that we used in our in vitro studies, and that these contribute to drive some Stra8 expression in vivo. Whole-mount immunofluorescence showed that the residual STRA8 expression was predominantly at the anterior end of the gonad at 13.5 dpc, suggesting that the anterior to posterior wave is retained (Fig. S7). This result suggests that any remaining positive regulatory elements respond to signal(s) that originate at the anterior end of the gonad, or adjacent mesonephros, in vivo. It is possible that modification of RARE1 and RARE3 makes the locus less sensitive to RA and diminished Stra8/STRA8 expression in the double RARE mutant therefore reflects delayed onset of Stra8 expression. We found that homozygous female mice harbouring RARE1, RARE3 or RARE1/3 double mutations were fertile. This is not unexpected given the largely unperturbed Sycp3 and Dmc1 expression observed at 14.5 dpc.

The CRISPR/Cas9 genome editing experiments leveraged the homology directed repair (HDR) pathway to introduce precise edits to our RARE sites of interest. In the absence of successful HDR, the non-homologous end joining (NHEJ) pathways assumes the DNA repair role; however, erroneous deletions of one to several hundred bases of DNA are often introduced to the target genomic region. During the generation of our RARE1 mutant, we also serendipitously generated a mouse line with a 173 bp deletion due to NHEJ (Δ173, −228 to −55 relative to the TSS; Fig. S5C). Intriguingly, the deleted region includes both RARE1 and the DMRT1 binding site (Fig. 5A). qRT-PCR analysis at 14.5 dpc showed that fetal ovaries homozygous for this mutation did not express Stra8 at all, that expression of both Dmc1 and Sycp3 was substantially diminished, Rec8 was unchanged and Pou5f1 elevated (Fig. 5B). This result was corroborated by the lack of STRA8 and SYCP3 immunofluorescence staining in ovaries of the Δ173 mutant at 14.5 dpc compared with those of the wild type and RARE1/3 double mutant (Fig. 5C), and the retention of POU5F1⁺ germ cells along the entire length of Δ173 mutant ovaries (Fig. S8). Breeding of homozygous Δ173 (n=4 each sex) animals with wild-type breeding partners for ∼3 months established that both sexes are infertile, whereas heterozygous breeders show no apparent abnormalities (Table S4). We replicated this 173 bp deletion in the context of the 2.9 kb Stra8 promoter luciferase construct and saw no significant induction of promoter activity regardless of the presence of RA, DMRT1 or both (Fig. 5D). This result highlights the crucial importance of this small region of upstream sequence and confirms the relevance of our in vitro F9 EC cell model for examining early meiotic germline events.

STRA8 expression in germ cells is a necessary prelude to meiosis in both sexes, but it is not clear how the highly cell- and stage-specific control of Stra8 expression is achieved. One major question relates to how the gene is so comprehensively silenced in somatic cells at all stages of development and adult life, and there is a growing appreciation of the various epigenetic mechanisms that appear to be involved in this silencing (Hargan-Calvopina et al., 2016; Wang and Tilly, 2008; Yokobayashi et al., 2013). In this study, however, we focused on the question of how Stra8 expression is specifically activated in germ cells of the mouse fetal ovary, and in resolving the contentious question of whether RA plays a role in this crucial developmental step.

Previously, transgenic studies employing either 0.4 kb or 1.4 kb of Stra8 upstream sequence reported high levels of germ cell-specific expression in the adult testis, but no expression was detected in the fetal ovary (Antonangeli et al., 2009; Giuili et al., 2002; Nayernia et al., 2004; Sadate-Ngatchou et al., 2008). Here, we generated a transgenic mouse line in which the 2.9 kb fragment immediately upstream of the mouse Stra8 TSS was able to direct eGFP expression in a spatial and temporal pattern that recapitulates the expression of endogenous Stra8 in germ cells of the mouse fetal ovary, with no eGFP induction observed in the fetal testis. We detected what appears to be a slight delay in the production of eGFP compared with endogenous STRA8: for STRA8⁺ germ cells only 68% were also GFP⁺, whereas no GFP⁺ germ cells were STRA8⁻. This apparent lag might result from differences in transcriptional accessibility at the two distinct genomic loci (downstream of Col1a1 and endogenous Stra8, respectively), in the timing required for RNA processing for the two transcripts, or in translation of the two proteins. It is also possible that the 2.9 kb promoter fragment lacks RAREs that impact on sensitivity to RA, or lacks binding sites for some other RA-independent inducing factor, compared with the endogenous Stra8 locus. This line will be a valuable tool for dissecting the pre- and early-meiotic events that occur in female germ cells.

Targeting a single transgene copy into an open genomic locus, downstream of Col1a1, allowed us to directly compare the regulatory capability of the 2.9 kb fragment with that of the shorter 1.4 kb and 1.5 kb fragments. Although eGFP was not induced to a visible level with the shorter constructs, as previous studies had also found (Giuili et al., 2002; Sadate-Ngatchou et al., 2008), our in vivo studies revealed that the shorter transgenes were in fact transcriptionally activated in the fetal ovary (but not the fetal testis) and must, therefore, include some elements that respond in a sex-specific manner. We also found that, in vivo, regulatory element(s) that are important for high levels of expression must lie in the region between 2.9 kb and 1.5 kb upstream of the Stra8 TSS.

We characterised the promoter elements required for expression of the transgene, working initially in F9 EC cells as a model for fetal germ cells. F9 cells are known to upregulate Stra8 in response to RA and express modest levels of RARγ but low levels of RARα and RARβ (Laursen et al., 2012), which is a similar RAR profile to pre-meiotic ovarian germ cells at 10.5 dpc and 11.5 dpc (Vernet et al., 2020). Using a low dose of RA (50 nM), we identified a short distal regulatory region of ∼100 bp that was crucial for strong transcriptional activation by RA. By way of site-directed mutagenesis, we demonstrated that a RARE sequence of the DR1 type, designated here as ‘RARE3’, lies within the region and is functional in our in vitro system. We also confirmed that RARE1 is active in vitro, as has been reported previously (Giuili et al., 2002). In light of these results, it seems likely that the additional upstream sequence between 1.5 kb and 2.9 kb, which includes RARE3, allows for greater sensitivity to the limited amount of RA available in the fetal ovary. The fact that shorter Stra8 constructs direct detectable eGFP in the postnatal testis, but not the fetal ovary, is in line with evidence that higher levels of RA are present in the former than in the latter tissue (Bowles et al., 2016, 2006; Snyder et al., 2010).

There is substantial evidence that DMRT1 is an important regulator of Stra8 but, paradoxically, it seems to enhance Stra8 expression in germ cells of the fetal ovary (Krentz et al., 2011) and repress Stra8 expression in adult spermatogonia (Matson et al., 2010). The effect in the fetal ovary seems to be somewhat stochastic, because some germ cells in the Dmrt1-null expressed STRA8 and entered meiosis normally, whereas others failed to do so (Krentz et al., 2011). In our F9 EC in vitro system, we found that DMRT1 alone had no effect on luciferase expression, but that its presence substantially enhanced the activation level achieved by RA. We further found that the ability of DMRT1 to positively influence Stra8 expression, in vitro, required the more proximal RARE1 sequence to be intact, but not the distal RARE3 sequence. Moreover, mutation of the DMRT1 site previously defined by ChIP and located just 38 bp downstream of RARE1 (Krentz et al., 2011) ablated the DMRT1/RA synergistic effect. Our findings suggest a model whereby DMRT1 acts together with RA/RAR/RXR at the RARE1, whereas RA/RAR/RXR action at RARE3 is independent of DMRT1 and ensures, at least, an amplification of transcription (Fig. 3C). This conclusion is consistent with the observation that DMRT1 is present in germ cells of the fetal testis but that it does not induce Stra8 expression and does not bind upstream of Stra8 in that context (Krentz et al., 2011), presumably because RA is not present. Intriguingly, DMRT1 appears to inhibit, rather than enhance, RA-directed Stra8 expression in adult spermatogonia (Matson et al., 2010); the molecular basis for this difference remains to be determined. For this reason, our F9 EC in vitro model is probably not suitable for modelling meiotic onset in postnatal spermatogonia, but our results suggest the F9 cell line is suited for investigating meiotic onset in fetal oocytes.

We aimed to validate our findings in vivo, and used CRISPR/Cas9 methodology to directly and subtly mutate RARE1 and RARE3 elements. Mutation of RARE1 had a greater effect than mutation of RARE3: Stra8 expression fell to ∼65% and ∼85% of normal levels, respectively. When both RAREs were modified, Stra8 expression was reduced further to ∼45% of normal. This is strong evidence that intact RARE1 and RARE3 binding sites are required for optimal Stra8 expression. Nonetheless, ovarian fetal germ cells in the RARE1/3 mutant still entered meiosis and the homozygous XX adults were fertile; this is not surprising, as we found that downstream meiotic factors Sycp3 and Dmc1 remained relatively normal despite the reduced Stra8 levels (Fig. 4C; Fig. S6). Interestingly, the residual expression of STRA8 in the RARE1/RARE3 double mutant ovaries is predominant at the anterior end of the gonad at 13.5 dpc, suggesting that any remaining intact positive regulatory elements respond to an anterior signal source; we consider it likely that this is RA, or some RA-induced factor, as levels are highest at the junction between the anterior gonad and mesonephros (Bowles et al., 2006). Although RARE2 (IR5) (Giuili et al., 2002), the RARE (DR4) found ∼800 bp upstream of the TSS, (Raverdeau et al., 2012) and the putative DR1-type RARE we identified proximal to the TSS, did not seem to function in vitro, it is possible that they do function in vivo if appropriate co-factors are present in that context. It is also possible that there are other important RA-sensitive regulatory elements, possibly including RAREs, beyond the 2.9 kb region or within the first intron of Stra8, that act to maintain Stra8 expression in the absence of RARE1 and RARE3. One other possibility is that RA drives production of WNT4 predominantly at the anterior end of the gonad (Bowles et al., 2018), and that this is influencing Stra8 expression in the germ cells (Chassot et al., 2011; Naillat et al., 2010).

Based on the recent findings of others, it is also possible that residual Stra8 expression in fetal ovaries of the RARE1/3 mutant is due to BMP signalling (Miyauchi et al., 2017). However, unlike RA, there is no reason to anticipate that BMP is especially prevalent at the anterior end of the gonad, as Bmp2 expression is observed in a rather homogeneous pattern along the length of the developing ovary (Yao et al., 2004). A downstream effector of BMP signalling, ZGLP1, has been shown, in mPGCLCs, to bind within the first intron of Stra8, but this occurs only in the presence of RA (Nagaoka et al., 2020). It is possible that, like DMRT1, ZGLP1 enhances Stra8 expression only in the presence of RA (but not necessarily via RARE1 or RARE3). If this is the case, then it would make sense that that anterior to posterior wave would be retained in our RARE1/3 mutant, because RA levels are initially highest at the anterior end. The lack of Zglp1 expression in F9 cells (Fig. S8; Laursen et al., 2012) suggests that the BMP signalling pathway is not active in our in vitro model; this may account for the discrepancy seen between our in vitro and in vivo results in terms of complete ablation or partially retained Stra8 expression, respectively.

It is also possible that additional non-RA factor(s) are involved. A number of recent studies have supported the contention that induction of Stra8 expression (and thus meiotic initiation) is independent of RA signalling in the fetal ovary, or at least that RA signalling is dispensable. It is reported that deletion of some or all of the known RA-synthesising enzymes diminishes Stra8 expression but does not ablate it, and meiosis is still able to occur (Chassot et al., 2020; Kumar et al., 2011). Furthermore, deletion of RARs leads to delayed Stra8 expression and fewer STRA8⁺ germ cells, but meiosis can still progress (Vernet et al., 2020). Our findings are not necessarily contradictory to these published works. We used an alternative and complementary strategy to investigate the role of RA in Stra8 regulation and we provide direct evidence that, to achieve normal levels of Stra8 expression in mouse fetal oocytes, it is necessary that two identified RAREs are intact: subtle mutation of both RARE1 and RARE3 diminishes Stra8 expression by 55%, but does not deplete it entirely. Although residual Stra8 expression may reflect RA acting at different RAREs or the actions of WNT(s) or BMP(s), as discussed above, it also remains possible that the remaining Stra8 expression that we observed in this study is induced independently of RARs or RA-synthesising enzymes, as has been suggested (Chassot et al., 2020; Vernet et al., 2020), and that some additional as yet unidentified factor(s) are involved in ensuring Stra8 expression in the context of the fetal ovarian germ cells.

We inadvertently made a 173 bp deletion of the Stra8 proximal promoter that encompassed RARE1 and the DMRT1 binding site, but left the upstream RARE3 (identified here), RARE2 (Giuili et al., 2002) and a TATA box-like sequence (Giuili et al., 2002) intact (Fig. S5). We found that Stra8 expression was ablated completely in fetal ovarian germ cells, that they did not enter meiosis, and that XX and XY individuals were infertile, suggesting that this line is effectively a Stra8-null. When this mutation was modelled in vitro, we similarly found a complete ablation of promoter activity. The severity of the effect, when compared with tandem in vivo mutation of RARE1 and RARE3, suggests that the deleted region includes at least some essential element of the ‘core promoter’ (though as noted, a TATA box-like sequence remains) and that its removal completely destroys the ability of transcription initiation complexes to bind and function. Future studies will focus on this short region of Stra8 upstream sequence.

We conclude that Stra8 is a direct target of RA in germ cells of the fetal ovary, and that sex-specific Stra8 expression in this context is driven in part by canonical RA signalling, with RA/RAR complexes acting through RARE1 and newly identified RARE3. The inducing factor(s) that are responsible for the residual Stra8 expression remain to be determined. We further conclude that the role of DMRT1 in driving Stra8 expression is to enhance, when both RA and an intact RARE1 are present, rather than to instruct: this is in keeping with evidence that DMRT1 is more highly expressed in germ cells of the fetal testis than the fetal ovary and that Dmrt1-null females are fertile. Furthermore, we have demonstrated that 2.9 kb of Stra8 upstream sequence is sufficient to recapitulate the curious anterior to posterior pattern of Stra8 onset, and that the F9 EC cell culture system is suitable for modelling fetal germ cell meiotic onset. This information will underpin future studies of meiotic onset in the mouse model.

All procedures involving animals and their care conformed to institutional, state and national guidelines. This study was approved by the University of Queensland Animal Ethics Committee.

All transgenic mice were generated in house using stock C57BL/6J, CD1 and C57Bl/6JxCBA F1 hybrids purchased from Animal Resource Centre (Western Australia).

A fragment containing the 2.9 kb promoter region of Stra8 was amplified from C57BL/6J genomic DNA using the 2.9 kbStra8_F and Stra8_R primers (Table S1). The PCR fragment was subcloned into pGEM-T-easy (Promega) and inserted into the XhoI/SacII site of pEGFP-N1 (Clontech) to generate the plasmid 2.9 kbStra8-eGFP. A similar plasmid, 1.4 kbStra8-eGFP, based on the work reported in Nayernia et al. (2004), was a kind gift from Julia Young and Kate Loveland (Monash University, Melbourne, Victoria, Australia). To assess the activity of the promoters in vivo the Stra8 promoter, eGFP and polyA sequence from these constructs were cloned into the PGK/ATGfrt vector (MES4490, Thermo Scientific Open Biosystems) to allow targeted integration into ES cells (ESCs) and subsequent mouse line generation (further described in next section). Constructs for in vitro transfection assays used only the Stra8 promoters from 2.9 kbStra8-eGFP and 1.4 kbStra8-eGFP, which were cloned into pGL4.10[luc2] (Promega) to generate Stra8 promoter luciferase reporter constructs (see ensuing sections). The ATG start site of firefly luciferase precisely replaced the eGFP ATG start site of the constructs used for reporter mice generation.

The constructs 2.9 kbStra8-eGFP and 1.4 kbStra8-eGFP were excised with XhoI/SspI and SpeI/SspI, respectively. The 1.5 kbStra8-eGFP was prepared by digesting 2.9 kbStra8-eGFP with BspHI/SspI. All three Stra8-eGFP fragments were blunt-ended with Mung Bean Nuclease (New England Biolabs) before being ligated into the EcoRV site of the PGK/ATGfrt vector (MES4490, Thermo Scientific Open Biosystems). Orientation of the Stra8-eGFP inserts were determined by restriction enzyme digest and clones with the correct insert orientation for each promoter length were further validated by Sanger sequencing. The integration of Stra8-eGFP-PGK/ATGfrt constructs into KH2 ESCs (Beard et al., 2006) was carried out as described previously (Quinn et al., 2014). In brief, each Stra8-eGFP-PGK/ATGfrt construct was electroporated into KH2 ESCs along with FLPe recombinase-expressing construct (pCAGGS-FLPe-puro, MES4488, Thermo Scientific Open Biosystems). Positive colonies were selected using hygromycin before confirmation for correctly targeted integration (Col1a1-Geno7 and Col1a1-Geno8), as well as the presence of the Stra8-eGFP transgene (Stra8_Fw2 and eGFP_Rv1; Table S1). Correctly targeted cells were injected into C57BL/6J blastocysts and transferred into 0.5 dpc or 2.5 dpc pseudopregnant CD1 females. Chimeric founders were identified by coat colour and bred to C57BL/6J females to generate F1 progeny. Heterozygous pups identified by genotyping using the primers above were used to establish breeding colonies to supply mice for timed matings and subsequent analysis.

Stra8-eGFP and RARE mutant studs from each line were mated with females, with noon of the day that the copulatory plug was observed designated as 0.5 dpc. Embryos were collected from timed matings, between 12.5 dpc and 16.5 dpc. The sex of the embryos was either identified by visual inspection of dissected gonads or by genotyping of tail tissue for ubiquitin-like modifier activating enzyme 1 (Uba1, also known as Ube1) (Chuma and Nakatsuji, 2001). For qRT-PCR, individual gonads were dissected with the mesonephros removed, whereas for direct fluorescence imaging and whole-mount immunofluorescences the mesonephros was left on for orientation and referencing purposes. Whole embryos (minus tail for genotyping) were collected, fixed in 4% paraformaldehyde (PFA)/PBS and processed into paraffin for sectioning and immunofluorescence imaging.

Individual gonad pairs collected from each Stra8-eGFP line and RARE mutant lines were stored in RNAlater (76106, Qiagen) post-dissection. After genotyping, total RNA was extracted using an RNeasy Micro Kit (74106, Qiagen). Total RNA-containing eluate was immediately used for cDNA synthesis by reverse transcription using a High Capacity cDNA Reverse Transcription Kit (4368813, Life Technologies). All qRT-PCRs were performed on a Viia7 Real-Time PCR System (Life Technologies). eGFP transcripts were quantified using SYBR Green PCR master mix (Life Technologies) with primers GFP_F and GFP_R (Table S1). Relative expression levels were determined using the 2^delta-Ct method using germ cell-specific gene Ddx4 (also known as Mvh and Vasa) with primers DDX4_F and DDX4_R (Table S1). For RARE1, RARE3, RARE1/3 and Δ173 mutant ovaries, expression levels of Dmc1 (Mm00494485_m1), Pou5f1 (Mm03053917_g1), Rec8 (Mm00490939_m1), Stra8 (Mm00486473_m1) and Sycp3 (Mm00488519_m1) relative to Ddx4 (Mm00802445t_m1) or Tbp (Mm01277045_m1) were quantified using Taqman Gene Expression Assays (Life Technologies). All kits and assays were performed according to instructions supplied by the manufacturer.

For direct fluorescence imaging of Stra8-eGFP gonads, freshly dissected gonads with mesonephros attached were placed in concave slides suspended in PBS with a glass coverslip. A BX-51 upright fluorescence microscope (Olympus) was used to visualise eGFP expression and images were captured using a DP70 colour camera (Olympus) attached to the microscope. Images were taken at multiple focal planes for each gonad and Adobe Photoshop CS5 version 12.0×64 was used for focus stacking using auto-stack, auto-align and auto-blend settings.

Whole embryos embedded in paraffin were sectioned at 7 µm and immunofluorescence was performed as described previously (Bowles et al., 2010). Primary antibodies used were anti-DDX4 (1:1000, ab13840, rabbit polyclonal, Abcam), anti-FOXL2 (1:2000, rabbit polyclonal, Kashimada et al., 2011), anti-STRA8 (1:300, ab49405, rabbit polyclonal, Abcam), anti-GFP (1:100, ab5450, goat polyclonal), anti-SYCP3 (1:200, ab97672, Abcam) and anti-POU5F1 (1:200, sc-5279, Santa Cruz Biotechnology). Secondary antibodies used (at 1:200) were donkey anti–goat IgG Alexa Fluor 488 (A11055, Molecular Probes), donkey anti-rabbit IgG Alexa Fluor 568 (A10042, Molecular Probes) and donkey anti-mouse IgG Alexa Fluor 647 (A31571, Molecular Probes). Nuclei were stained with DAPI (20 µg/ml, D9542, Sigma Aldrich) and imaged with a Zeiss LSM710 attached to a Zeiss AxioExaminer Z1 Upright microscope.

Whole gonads and mesonephros were dissected out of the embryos and whole-mount immunofluorescence was performed as described previously with minor modifications (Hobbs et al., 2015). Briefly, dissected gonads were fixed in 4% PFA for 30 min at room temperature, washed in PBS before being blocked in PBS supplemented with 2% bovine serum albumin and 10% fetal bovine serum (FBS) for 4 h at 4°C. Gonads were then incubated overnight with primaries antibodies [anti-STRA8, 1:2000, ab49602 (Abcam); anti-GFP, 1:100, 338008 (Biolegend) and anti-TRA98, 1:500, ab82527 (Abcam)], followed by three washes of 30 min with PBS with 0.3% Triton X-100 before incubating for 4 h at 4°C with secondary antibodies [goat anti-rabbit IgG Alexa Fluor 594, A11037 (Invitrogen) and donkey anti-rat IgG Alexa Fluor 488, A21208 (Invitrogen)]. Nuclei were stained with DAPI (20 µg/ml, D9542, Sigma Aldrich) and imaged with a Leica DMi8 SP8 laser point scanning confocal microscope. STRA8⁺ and GFP⁺ cells were manually counted using the cell counter plug-in of ImageJ. Briefly, cells were counted over a 30 µm z-stack at 4 z-planes (0, +10, +20 and +30 µm) for each biological replicate. At each plane, GFP⁺ and STRA8⁺ were manually annotated in ImageJ and counts were generated by the cell counter plug-in.

F9 EC cells were obtained from the European Collection of Cell Cultures (ECACC, 85061803, Lot 00C011) via Cellbank Australia. Truncated transfection constructs were generated by restriction enzyme digest of the 2.9 kbStra8luc2 construct. Site-directed mutagenesis was used to mutate the putative transcription factor binding motifs RARE1 and RARE3 on the 2.9 kbStra8luc2 construct with the mutation design, including a SalI recognition site for identification purposes (mutations matched to those in RARE mutant mice, Table S2). Genotyping primers for RARE1 mutant mice were used to amplify the 173 bp deletion from mutant mice, which was used to replace the corresponding region on the 2.9 kbStra8luc2 construct for investigating promoter activity harbouring the 173 bp deletion in vitro. DMRT1 binding site mutation on the 2.9 kbStra8luc2 promoter was also performed using site-directed mutagenesis with the design including an AscI site for identification purposes. cDNA derived from F9 EC cells was used to amplify PCR products consisting of Dmrt1 coding sequences that were subcloned into pGEM-T-easy (Promega) and then into pcDNA3.1 (Life Technologies) to create expression construct pDMRT1. F9 EC cells were cultured in 0.1% gelatin-coated flasks with high glucose Dulbecco's Modified Eagle Medium (DMEM, Gibco, 11995-073) supplemented with 10% v/v FBS (Bovogen, French Origin) and subcultured according to ECACC guidelines. For transfection experiments subconfluent (70-80%) F9 EC cells were treated with 0.25% trypsin (Gibco, 25200-072) before being seeded onto 24-well tissue culture plates at 2×10⁵ cell/cm² and allowed to adhere overnight. Culture medium was replaced ∼1 h before transfection with Lipofectamine 2000 according to the manufacturer's instructions. Each well was transfected with 100 ng of one of the Stra8 promoter-driven firefly luciferase constructs with or without 50 ng of pDMRT1. As a transfection control, 35 ng of Renilla luciferase construct pE1b-Rluc (a gift from Dr Liang Zhao (The University of Queensland, Brisbane, Queensland, Australia)) was also added to the transfection of each well, and pcDNA3.1 empty vector was used to make up for the DNA quantity required to maintain optimal DNA to lipfectamine ratios. Transfected cells were treated for 18 h the following day using medium supplemented with 1:2000 of either 100 µM all-trans-RA (R2625, Sigma Aldrich) or DMSO carrier control (final concentration of RA was 50 nM). Luciferase assays were then carried out using Dual-luciferase Reporter Assay System (Promega) and read with a POLARstar Omega plate reader (BMG Labtech) according to the manufacturer's instructions. The total culture duration of the F9 cell transfection experiments from seeding of the cells to harvesting for luciferase assay was controlled at 48 h.

Guide RNA was designed to target the sequence 5′-GGCGCTAGCCGCCTGGAT-3′ adjacent to RARE1. Oligonucleotides RARE1-top and RARE1-bottom were assembled and sgRNA was produced as described previously (Yang et al., 2013). For RARE3, custom DNA oligonucleotides RARE3_sgRNA_F, targeting the sequence 5′-CAAACCTCCACATTCCTCAC-3′, and a reusable chimeric_sgRNA_R (Table S1) were annealed and amplified using the Expand high fidelity PCR system (Roche). The PCR product consisted of a T7 promoter to drive transcription of sgRNA for RARE3 target sequence followed by chimeric guide RNA scaffold (Cong et al., 2013), and produces a 3′ DraI recognition site when T/A ligated. The PCR product was gel purified and T/A ligated into pGEM-T-easy (Promega) for Sanger sequencing using M13 primers. Correctly oriented clones with 100% sequence identity were digested with DraI, and the fragment containing the original PCR product, but not the T7 promoter from pGEM-T-easy, was gel purified for in vitro transcription (IVT). All sgRNAs were transcribed using a MEGAshortscript T7 Kit (Ambion), and Cas9 mRNA IVT was performed using an mMESSAGE mMACHINE T7 Transcription Kit (Ambion) from linearised pBS-Cas9 (gift from Dr Liang Zhao). IVT sgRNA and Cas9 mRNA were purified using a MEGAclear Transcription Clean-Up Kit (Ambion), with all procedures performed according to the manufacturer's instructions. Repair ssODN for RARE1 and RARE3 (RARE1_ssODN and RARE3_ssODN, respectively; Table S3) was designed with mutations to putative RAR/RXR binding sites, as well as an sgRNA protospacer adjacent motif recognition site to prevent retargeting and a SalI recognition site to facilitate genotype identification (identical mutations to F9 site-directed mutagenesis constructs). An injection cocktail consisting of 20 ng/μl Cas9 mRNA, 10 ng/μl sgRNA and 10 ng/μl ssODN was microinjected into the male pronucleus of one cell embryos derived from superovulated C56BL/6JxCBA F1 females. Injected embryos were cultured overnight in KSOM medium (Merck Millipore, MR-121-D) and embryos that reached the two-cell stage the next day were transferred into pseudopregnant CD1 females using standard methods. Mice harbouring correctly mutated RARE1 and/or RARE3 sites were identified by PCR using their respective genotyping primer pairs (Table S1) and then digested with SalI restriction enzyme. For the RARE1 and RARE3 double mutant, RARE1 mice were used to generate one-cell embryos for injection with RARE3 CRISPR/Cas9 injection cocktail. The Δ173 bp mutation was the result of NHEJ when generating the RARE1 mutant. The 2.9 kb Stra8 promoter and first exon were PCR amplified, subcloned and sequenced from all mutant alleles (Fig. S9). No other mutations other than the intended RARE sites were introduced, but we did note the presence of known SNP rs40012028 in the RARE3 and Δ173 bp mutant alleles. We also amplified, cloned and sequenced exon 2 of the Δ173 bp mutant allele to confirm that there were no mutations in the ATG start codon and surrounding sequence (Fig. S9).

Genomic DNA sequence (GRCm38.mm10) −5000 bp to +5000 bp relative to the Stra8 TSS (start of RefSeq NM_009292.1) was used as input for transcription factor motif prediction in MatInspector (Cartharius et al., 2005; Quandt et al., 1995). GT-Scan (O'Brien and Bailey, 2014) and Optimized CRISPR Design (http://crispr.mit.edu) were used to identify putative CRISPR/Cas9 guide RNA targets. All statistical analysis was performed using GraphPad Prism 7 software.

We thank Dr Rudolf Jaenisch (Massachusetts Institute of Technology, Cambridge, MA, USA) and Dr Konrad Hochedlinger (Harvard University, Cambridge, MA, USA) for the gift of KH2 cells and Tara Davidson for technical assistance. We also acknowledge the School of Biomedical Sciences Core Facilities at the University of Queensland for use of the Leica DMi8 SP8 confocal microscope.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.194977.reviewer-comments.pdf