A role for orphan nuclear receptor liver receptor homolog-1 (LRH-1, NR5A2) in primordial follicle activation

Liver receptor homolog-1 (NR5A2) is expressed specifically in granulosa cells of developing ovarian follicles where it regulates the late stages of follicle development and ovulation. To establish its effects earlier in the trajectory of follicular development, NR5A2 was depleted from granulosa cells of murine primordial and primary follicles. Follicle populations were enumerated in neonates at postnatal day 4 (PND4) coinciding with the end of the formation of the primordial follicle pool. The frequency of primordial follicles in PND4 conditional knockout (cKO) ovaries was greater and primary follicles were substantially fewer relative to control (CON) counterparts. Ten-day in vitro culture of PND4 ovaries recapitulated in vivo findings and indicated that CON mice developed primary follicles in the ovarian medulla to a greater extent than did cKO animals. Two subsets of primordial follicles were observed in wildtype ovaries: one that expressed NR5A2 and the second in which the transcript was absent. Neither expressed the mitotic marker. KI-67, indicating their developmental quiescence. RNA sequencing on PND4 demonstrated that loss of NR5A2 induced changes in 432 transcripts, including quiescence markers, inhibitors of follicle activation, and regulators of cellular migration and epithelial-to-mesenchymal transition. These experiments suggest that NR5A2 expression poises primordial follicles for entry into the developing pool.

The development of the mouse ovary begins with sexual differentiation on embryonic day 10.5, characterized by mitotic replication of oogonia 1 . Most of the incipient oocytes initiate meiosis between embryonic days 13.5 and 15.5 2 . Many are lost, but those that persist can be found densely packed in structures known as germ cell cysts 3 . Beginning on approximately day 19.5 of gestation, these cysts undergo a programmed breakdown, when the pre-granulosa cells invade and surround each oocyte, forming the primordial follicles 4 . This process continues through the first four to six days of postnatal life, and oocytes that fail to be encapsulated succumb to apoptosis 5 . In the mouse, approximately 33% of the oocytes survive to form the primordial follicles 4 . The consensus view is that this process of primordial follicle formation establishes the ovarian reserve that will provide oocytes throughout reproductive life 6 .
Although enormous progress has been made in understanding the events and regulation of the trajectory of ovarian follicular development, the early stages of development remain, to a large extent, a mystery 7 . The majority of the follicles in the reserve are dormant. It has been shown that the maintenance of primordial follicles in this quiescent state is not a passive process, but that the resting pool of primordial follicles is under a constant inhibitory influence 8,9 . Nonetheless, once activated, continuation of development is an irreversible process, and follicles that have initiated growth undergo atresia if not selected for subsequent stages of maturation 10 . Activation of primordial follicles occurs via a gonadotropin-independent process, whereby they are gradually selected from the quiescent reserve into the growing follicle pool 11,12 . Extensive bidirectional signaling takes place between oocytes and granulosa cells to ensure follicle development from primordial stage onwards 13,14 . As oocytes begin to increase in size, the granulosa cells undergo a partial epithelial-to-mesenchymal transition under the control of ZEB, SNAIL and TWIST family members 15 . This transition engenders change in the shape of these somatic Results NR5A2 mRNA and proteins are depleted in conditional (cKO) mice at the termination of primordial follicle formation. To confirm the depletion of NR5A2 from the neonatal (PND4) and PND13 ovaries, we first demonstrated by in situ hybridization and immunohistochemistry that AMHR2 is expressed in pre-granulosa cells of the primordial follicle. The Amhr2 transcript localized to the granulosa cells of primordial, primary and secondary follicles in the immature ovary, as well as to the surface epithelium (Fig. 1a). Immunohistochemistry further confirmed the AMHR2 protein to be present in these three classes of follicles (Fig. 1b). Together these findings provide evidence for Cre/lox recombination at least as early as the primordial follicle stage of the follicle development trajectory. We then substantiated the depletion of NR5A2 from perinatal ovaries between PND3 and PND6, the period of termination of primordial follicle formation, by qPCR, in situ hybridization and immunohistochemistry ( Fig. 1c-f). Whole ovary qPCR revealed 80% decrease (p < 0.0001) in Nr5a2 mRNA abundance in the cKO mice (Fig. 1c). In keeping with the reduction seen by qPCR, Nr5a2 mRNA in cKO ovaries at PND6 was undetectable by in situ hybridization (RNAScope; Fig. 1d). Immunohistochemical evaluation of NR5A2 protein demonstrated the expected nuclear localization of NR5A2 in CON ovaries (Fig. 1e). Little signal was observable in primordial and primary follicles in the cKO ovaries, as confirmed by a significant reduction in fluorescence, detectable by CellProfiler software (Fig. 1f).
Forkhead box protein L2 (FOXL2) is a granulosa cell-specific winged-helix transcription factor that is required for activation of the primordial follicle to the primary state 26 . In a recent publication 27 it was shown that this factor is expressed differentially between subpopulations of pregranulosa cells during primordial follicle formation. We hypothesized that if NR5A2 could elicit its effect on follicular populations by regulating abundance of this key transcription factor, it could also serve as a marker for the efficacy of depletion of the nuclear receptor. Comparison of FOXL2 mRNA and protein abundance in cKO and CON ovaries at PND4 revealed that its expression was markedly lower in the NR5A2 depleted ovaries ( Fig. 1 g-i).
Phenotypic differences between the cKO and CON ovaries at postnatal day 4. Histological evaluation of ovaries of mice at PND4, when the follicle population is limited to primordial and primary follicles 28 confirmed the presence of these two follicle subtypes in both CON and cKO mice (Fig. 2a). At PND4, we observed a larger number of primordial follicles, but fewer primary follicles in the cKO ovary compared to the control mice ( Fig. 2a-c). This increase was also present at the time of appearance of the first antral follicles, on PND13 (Fig. 2d,e) 28 .
The number of primordial follicles in five-month-old cKO mice relative to CON was similar. In contrast, the number of atretic follicles in the cKO ovary, identified by the presence of a degenerate oocyte, was consistently and substantially greater than found in CON mice (Fig. 2f,g).
To further study the role of NR5A2 in follicular development, we established an organ culture model of postnatal day 4 (PND4) ovaries, incubated over 10 days. We identified oocytes by localization of the gamete specific factor, DDX4 (red; Fig. 2h,j). In CON mice we observed primordial follicles in the cortex and numerous developing primary follicles, mainly located in the medulla, consistent with previously published studies 29 . Moreover, the number of developing follicles in the medullary zone of the ovary in cKO mice was significantly lower (Fig. 2i). Further, in the cKO ovaries, the ratio of primordial to primary follicles was substantially greater relative to the CON (Fig. 2k), in keeping with in vivo results.
NR5A2 defines two subsets of primordial follicles. To investigate the role of NR5A2 in primordial to primary follicle transition, we performed RNA in situ hybridization and fluorescent immunocytochemistry in ovaries of wild type mice on PND3 and PND6. Localization of the Nr5a2 transcript revealed the existence of two populations of primordial follicles in the neonatal mouse ovary, one expressing NR5A2 and a second that was devoid of the signal (Fig. 3a,b,d). To further corroborate the observation that the follicles expressing NR5A2 have not yet been activated, we performed a double immunofluorescence to detect NR5A2 and KI-67, a cellular proliferation marker (Fig. 3c,d). As expected, we found that activated primary follicles expressed both NR5A2 Figure 1. Depletion of NR5A2 from knock-out (cKO) ovaries at the termination of primordial follicle formation, via AMRH2-cre. (a) RNA in situ hybridization of Amhr2 in CON ovaries at PND6. (b) Immunolocalization of AMHR2 in the ovaries of PND13 CON mice. (c) Abundance of Nr5a2 transcripts in CON and cKO mouse ovaries at PND4 (n = 5-10 per genotype). (d) RNA in situ hybridization of Nr5a2 in CON and cKO ovaries at PND6. (e) Immunolocalization of NR5A2 in PND4 CON and cKO mouse ovaries. Quantitative analysis of NR5A2 expression in CON and cKO PND4 mouse ovaries (n = 3-6 animals per genotype). (f) Quantification of signal from immunofluorescence in CON and cKO mouse ovaries on PND4. (g) Localization of FOXL2 in CON and cKO mice ovaries at PND4 showing absence of FOXL2 signal in primordial follicles of the cKO mouse. (h) Quantitative analysis of FOXL2 expression between CON and cKO PND4 mice ovaries (n = 3-5 animals per genotype). (i) Abundance of Foxl2 transcripts in CON and cKO mice ovaries at PND4 (n = 5-10 animals per genotype). Quantification or protein expression for NR5A2 and FOXL2 was performed using CellProfiler version 4.06, www.cellp rofil er.org.  Table S1). Validation of candidate genes related to the mechanisms that differed between the cKO and CON in RNAseq analysis was achieved by qPCR (Table 1). Eighteen transcripts were identical in pattern of change and statistical significance in RNAseq and qPCR, and an additional eight transcripts tended to change (p < 0.05, but padj > 0.05) in RNAseq, but were confirmed as significantly changed in qPCR. Changes in transcript abundance of only five genes from RNAseq (padj < 0.05) proved undetectable by qPCR (p < 0.05), and all of these had consistent fold changes in RNAseq and qPCR (Table 1). Pathway analysis revealed that transcripts downregulated by Nr5a2 cKO were primarily associated with cell migration, including pathways for leukocyte migration, cellular extravasation, and chemotaxis (Fig. 4c). The mRNA associated with these pathways included components of cytokine and chemokine signaling (Cxcr2, Cxcl14, A2m, Csf3r), adhesion molecules (Sell, Selp, Selplg), and metalloproteases (Mmp8, Mmp9), among others (Fig. 4d). Pathways associated with vascular development were also disrupted, including transcripts for Angptl7, Fgfbp1 and Mmp9. Other top pathways included those for normal cellular functions, such as ion transport, DNA binding and receptor binding, indicating that downregulated mRNA may be involved in maintaining cellular homeostasis (Fig. 4c, Supplementary Table S2). We hypothesized that, because cellular migration-related functions were negatively affected by loss of Nr5a2, this transcription factor may mediate the epithelial-tomesenchymal transition that characterizes follicular activation 15 . Confirmation by qPCR of the decreases of mesenchymal markers Mmp9 and Snail1 by RNAseq supports this postulate (Fig. 4e). To provide additional evidence for modulation of the mesenchymal transcription factor Snail1, two of its downstream targets, Cdh1 and Gja1, were also quantified by qPCR. Both Cdh1 and Gja1 have been reported to decrease during the escape from the epithelial state [30][31][32] . As expected, both these epithelial markers were significantly upregulated in our cKO mouse model suggesting active maintenance of the epithelial condition (Fig. 4f).

Scientific Reports
Some of the mRNA upregulated in cKO ovaries were oocyte-specific transcripts. Among the primary pathways identified as increasing in cKO were those related to meiosis ( Supplementary Fig. S1, Supplementary  Table S3). Interestingly, among oocyte transcripts that were in greater abundance in cKO ovaries were Gdf9, Kit, Figla, and Bmp15. The former three were evaluated by qPCR and both Kit and Figla were more abundant in cKO ovaries than in control ( Supplementary Fig. S1).

NR5A2 depletion affects quiescence gene expression in primordial follicles.
In the cKO relative to the CON ovary, we observed a greater expression of PTEN, which is known to be a principal factor constraining primordial follicle activation 33 (Fig. 5a,b). As PTEN is a tumor suppressor, we hypothesized that Nr5a2 depletion would further be manifest in the expression of genes associated with proliferative quiescence. This proved true, as transcriptome analysis revealed modification in a number of these genes (Fig. 5c). Subsequent validation by qPCR confirmed the mRNA abundance of these genes was elevated in the PND4 ovaries from cKO animals ( Table 1; Fig. 5d-g). These included Cxcl1 and Cxcl10, two chemokine ligands, as well as Junb, Egr1, and Lin28a. All of these transcripts have been reported as markers of quiescence [34][35][36][37][38] . Moreover, the key inhibitors of follicular emergence from a quiescent state Smad 2 and Smad3 39,40 were upregulated in NR5A2 cKO ovaries (Fig. 5f). To assess the downstream effect of this maintenance of quiescence, the cyclin-dependent kinase inhibitor, Cdkn1a was evaluated by qPCR. This transcript was substantially elevated in the cKO PND4 ovary (Fig. 5g). Together the evidence supports the view that NR5A2 influences the activation of primordial follicles, both by inducing a partial epithelial-to-mesenchymal transition and by regulating a decrease in quiescence regulators and an increase in regulators of cellular proliferation. Hematoxylin/Eosin staining of postnatal day 4 (PND4) control and cKO mouse ovaries showed an increased number of primordial follicles and a decreased number of primary follicles in cKO compared to control. The populations of primordial (b) and primary (c) follicles on PND4 were quantified (n = 5 animals per genotype). (d) Hematoxylin/Eosin staining analysis showed that the number of primordial follicles was also increased in cKO mouse ovaries at PND13 compared to the CON. The primordial follicles are located principally in the cortices of the ovaries. (e) The population of primordial follicles on PND13 was enumerated (n = 5-9 animals per genotype). (f) Histological analysis of mature (5-month-old) ovaries in cKO and CON animals. (g) The populations of primordial, primary/secondary, antral and atretic follicles were quantified (n = 3-5 animals per genotype). In a, b, d, and e, example primordial follicles groups are circled with a dashed line. In d, one secondary follicle is marked with a black triangle. In (a,b), example primary follicles are marked by a white arrow. (h,j) CON and cKO mouse ovaries cultured for 10 days in vitro, at low (h) and high (j) magnification. Oocytes are stained with antibodies to DDX4 (red). (i) Quantification of the follicles in the medulla of cultured ovaries (n = 8-13 animals per group). (k) Quantification of the primordial follicle population in cultured ovaries. www.nature.com/scientificreports/

Discussion
In the mammalian ovary, the progressive activation of the follicular reserve serves as the source of fertilizable ova during reproductive life. While progress has been made in recognition of the events in primordial follicle formation 41 , factors regulating their activation under physiological conditions remain less well understood. NR5A2, an orphan nuclear receptor, is well known to play essential roles in the ovary in: expansion of the cumulus oophorus 22 , ovulation 23 , granulosa cell proliferation 18 and luteinization 24 . In the present study we show that NR5A2 is expressed in granulosa cells as early as the primordial follicle stage of follicle development. Based on this observation, we investigated the role of NR5A2 in early folliculogenesis in mouse ovaries between PND3 and PND6, i.e. during the window of time of termination of the formation of the primordial follicle pool 42 . Herein, we present evidence to indicate that there are two populations of primordial follicles in the mouse ovary, one quiescent where NR5A2 is absent, and a second, with NR5A2 expressed that is poised to activate and join the growing pool. Once follicles become activated, granulosa-specific expression of NR5A2 persists through follicular growth and ovulation. In aid of confirming the follicle population in which NR5A2 is depleted, we demonstrated, by immunofluorescence and RNA in situ hybridization, that AMHR2 is expressed in all primordial follicles. This concurs www.nature.com/scientificreports/ with earlier studies showing AMHR2 in granulosa cells in the mouse ovarian reserve 43 . Indeed, it has been reported that the Amhr2-Cre allele is expressed even earlier in the mouse female gonad, i.e. by embryonic day E11.5-12.5 44,45 . The Amhr2-driven Cre recombinase has been employed in a prenatal knock-out mouse model to deplete NOTCH2 from germ cell nests 46,47 , and we had previously shown effective recombination and depletion in the ovary 23 . Therefore, we chose this model to explore the effects of NR5A2 on the early follicle population (Nr5a2f. /f ; Amhr2 Cre+ ). Herein we have demonstrated substantial and consistent (80-90%) depletion of Nr5a2 in the PND4 mouse ovary.
In the cKO animals we observed a significantly smaller population of activated primary follicles at PND4, as indicated by the absence of expression of proliferation markers in the somatic cell component of the follicles. We hypothesized that the consequent increase in the number of primordial follicles observed in NR5A2 cKO mice ovaries results from a decreased rate of recruitment from a constant follicular pool. In support of this hypothesis, we showed that PND4 ovaries in organ culture displayed a larger population of primordial follicles and a reduced number of primary follicles.
PTEN is a well-established regulator of primordial follicle recruitment (reviewed in 48 ). Not unexpectedly, this protein was substantially upregulated in the cKO follicle population at PND4. This finding is consistent with the role of PTEN and its downstream target FOXO3, the transcript of which was also upregulated in the cKO ovary, as the principal inhibitors of primordial follicle activation. Indeed, deletion of PTEN from oocytes of primordial follicles resulted in premature activation of the entire primordial follicle pool 49 . Similarly, studies have shown that Foxo3 negative ovaries were enlarged by PND14 with greater numbers of early-growing follicles, and the primordial follicle reserve was totally depleted by that time 50,51 . Thus, one consequence of depletion of NR5A2 in primordial follicles is disruption of regulation of follicle activation normally exercised by the selective depletion of PTEN.
Molecular analysis has shown that in ovarian somatic cells, the transcription factor FOXL2 plays an important role in controlling both the formation and activation of primordial follicles. It dictates the differentiation of pre-granulosa into granulosa cells during primordial follicle recruitment 26,52 . In the present study, we show substantially reduced FOXL2 transcript and protein abundance in the NR5A2 cKO mouse, associated with the increased number of primordial follicles. This finding is consistent with the results of the germline depleting mutation of the Foxl2 gene where the abrogation of the squamous to cuboidal transition of the primordial follicle granulosa cells following follicular activation was observed 26 . Further, our ChIPseq data in ovulatory adult mouse granulosa cells demonstrate peaks of NR5A2 binding associated with the Foxl2 gene 53 . Although FOXL2 is expressed early in ovarian prenatal development 26 , the present findings indicate that it is likely downstream of NR5A2 in follicular activation in the postnatal primordial follicle.
To further identify NR5A2-regulated factors that may contribute to the transition from primordial to primary follicles, a comparison was made of our RNA sequencing results with a study of these two developmental strata from human ovarian tissue, in which granulosa cell-specific transcripts were identified and analyzed 54 . Fourteen transcripts that were greater in or unique to primordial follicles in that study 54 were found to vary between the NR5A2 cKO and CON ovary in the present study. Among these, eleven were in greater abundance in cKO than CON ( Supplementary Fig. S2). Several were regulators of ion transport or ion balance, including Atp8a2, Gabre, and Slc30a3, while several others were regulators of lipid binding or metabolism, including Pla2g4c, Osbp2, and Atp8a2. Wnt signaling is a positive regulator of both granulosa cell proliferation 55 and epithelial-to-mesenchymal transition (reviewed in 56 ). The suppressor of Wnt signaling, Tle2 57 was more abundant in primordial follicles and in cKO ovaries. Functional annotations of all mRNA that were common to the present dataset and the study of Ernst et al. 54 are in Supplementary Table S4. A potentially counterintuitive finding in the RNA sequencing study was the observation that a number of meiosis-related, oocyte-specific transcripts were upregulated in cKO ovaries. It is unlikely that oocytes were meiotically active on PND4; thus, it is more likely that these differences reflect the abundance of oocytes rather than a difference in meiotic activity. Among oocyte-specific transcripts that were upregulated were well-established oocyte activation signals, including Gdf9, Kit, and Figla. This indicates that the ability of the oocyte to provide a signal for activation remains, or is perhaps increased in the absence of NR5A2, despite an impaired ability of granulosa cells to respond to the signal.
NR5A2 depletion has been shown to decrease granulosa cell proliferation by downregulating genes such as cyclins D and E, whose expression has been demonstrated to be depleted in quiescent cells 18,37 . Additional support for a reduced rate of activation of primordial follicles comes from evaluation of quiescence markers. For example, the abundance of the transcript for Cdkn1a, the cyclin dependent kinase inhibitor, is increased in PND4 cKO mice in the present study, consistent with its regulation by NR5A2 in breast cancer cells 58 . This protein is known to induce and maintain cell cycle arrest under conditions such a senescence or quiescence 59,60 . Further, it is a direct transcriptional target of NR5A2 in other tissues 18 . The chemokine ligands CXCL1, CXCL10 as well as EGR1, JUNB and LIN28A are associated with mitotic quiescence [34][35][36][37] . All five of these factors were significantly upregulated at the mRNA level in NR5A2 cKO ovaries at PND4. SMAD2 and 3 are key signal transducers in the TGF-beta signaling pathway. Their expression is associated with follicular quiescence while their loss was associated with activation 40 . SMAD3 directly maintains quiescence of granulosa cells by suppressing expression of the proto-oncogene MYC 39,61 . Both Smad2 and Smad3 were greater in cKO than CON ovaries, while Myc declined. Together, the gene expression patterns support the hypothesis that NR5A2 depletion from primordial follicles diminishes the rate of recruitment of primordial follicles into the growing follicle pool.
A key finding in the present study is that the presence or absence of NR5A2, defines two subsets of follicles in the primordial follicle population. We explored these two subpopulations, those in which NR5A2 was expressed in the pre-granulosa cells and those where it was not present. Neither expressed the mitotic marker KI-67, confirming their quiescence. The concept of more than one subset of primordial follicles is not new, as previous studies of the developmental chronology of follicle development identified two distinct populations of primordial www.nature.com/scientificreports/ follicles 47,62 and lineage tracing has defined an initial small wave of follicles that populates the medulla of the ovaries 27 . These arise from bipotential cells in the stroma of the ovary, and begin to express Foxl2 in late embryonic life. The second population is of surface epithelial provenance and in these cells, robust Foxl2 expression only begins after birth. The follicle of the first wave exists in the ovaries for around three months and contribute to the onset of puberty. The second wave of primordial follicles, from the ovarian surface, replaces the first wave of follicles, providing fertility until the end of reproductive life. In the present study, the ovaries of CON mice collected at PND4 and cultured for ten days showed extensive activation of the medullary follicle population, and this activation was nearly absent in the cKO mice. Thus, the differential expression of NR5A2 in primordial follicles may be related in part to their location in the postnatal ovary, which appears to be a consequence of the respective provenance of the follicles 27 . Nonetheless, our in situ analyses also identified populations of primordial follicles in the cortex of more mature ovaries that either express or are devoid of NR5A2. Given the multiple roles of NR5A2, including induction of cell proliferation, we speculate that the follicles in which NR5A2 is absent are quiescent, while those expressing NR5A2 are in line for activation. Finally, primordial follicle activation is characterized by a partial transition of granulosa cells from epithelial to mesenchymal state 63 . Studies show that NR5A2 promotes migration and invasion in breast cancer through up-regulating MMP9 and inactivating E-cadherin 30,58 . Together these findings suggest that NR5A2 is involved in epithelial mesenchymal transition 64 . Here we showed that the depletion of NR5A2 decreases pathways related to migration, increases the quantity of transcript of epithelial markers, Cdh1 and Gja1, and decreases mesenchymal state related genes, Snail1 and Mmp9. These findings provide another mechanism by which NR5A2 may regulate primordial follicle activation.
The downregulation of transcripts associated with migration in cKO ovaries was not limited to markers of epithelial to mesenchymal transition. As discussed previously, there is evidence for localization of growing follicles to the ovarian medulla, with primordial follicles remaining in the more rigid cortex. This process is not fully understood, but seems to be regulated in part by reorganization of the extracellular matrix 65 . In particular, actin polymerization, via the HIPPO pathway 66 , is a regulator of follicular activation. In the present study, Coro1a and Synpo2, both accessory to actin reorganization, were in reduced abundance in NR5A2-depleted ovaries.
The alterations in transcripts related to migration could also be indicative of changes in vascularization regulated by NR5A2 as regulators of vascular development, including Mmp9, Fgfbp1, and Angptl7 were less in NR5A2 cKO ovaries. These results are consistent with some recent evidence of a role for vasculature in follicular activation 61 .
As noted above, the consequence of this NR5A2 depletion was a significantly greater number of primordial follicles in PND4 cKO mice, an increase that persists at least until PND13, when the follicle population is composed of primordial, primary and secondary but no antral follicles 28 . The disappearance of the increased population of primordial follicles in the cKO mice in adulthood may be related to the existence of census mechanism in the ovary by which an excess number of primordial follicles around birth is detected and removed from the ovary by adulthood. This concept is consistent with the resultant phenotype of BCL2 cKO mice where depletion was effected by the cKIT promoter 67 . These mice have more primordial follicles at birth but this increase is not maintained in postnatal life, and by postnatal day 30-60 there is no longer a difference 67 . Herein we show a substantial increase in the number of atretic follicles, frequently recognized by fragments of zonae pellucidae 50 . It is therefore possible that long-term blockade of primordial follicle growth in the cKO ovary results in clearance of the non-growing follicles and depletion of the primordial follicle pool by early adulthood, as has been shown by abrogating AKT activation pathway through conditional deletion of PDK1 in the oocytes 13 .
In our transgenic mouse model, NR5A2 is depleted by as much as 90%. Its residual expression and a potential compensation from the other member of the NR5A nuclear receptor family, NR5A1, also expressed in granulosa cells 68 could explain the absence of phenotype in post-pubertal animals relative to the number of primordial follicles. As mature NR5A2 cKO mice present growing follicles and no sign of primary ovarian insufficiency, it may be that the excess number of primordial follicles is limited to a certain window of the developmental trajectory.
In summary, we have demonstrated that depletion of NR5A2 from primordial follicles results in a larger ovarian reserve. We report a decreased rate of recruitment into the growing follicle pool, related to an increase of quiescence and a dysregulation of epithelial to mesenchymal transition (summarized in Fig. 6). We have further shown that NR5A2 is differentially expressed in populations of primordial follicles, and we conclude that this orphan nuclear receptor is an elemental generator of primordial follicle activation.

Materials and methods
Animals and colony maintenance. Animal experiments were approved by the University of Montreal Animal Care Committee and were conducted according to the guidelines of the Canadian Council on Animal Care. All mutant and control (CON) mice were maintained on the C57BL/6 background, under a 14-h light, 10-h dark cycle and provided food and water ad libitum. The Nr5a2 floxed (Nr5a2 fl/fl ) mice have been described previously 21,23,69 . The LoxP sites in this line result in excision of exons 4 and 5 from the NR5A2 transcript 69 . Granulosa-specific depletion of NR5A2 was induced by crossing these animals with mice expressing Cre-recombinase driven by the anti-Mullerian type II receptor (Amhr2 Cre/+ ) 21,23 to produce conditional knockout (cKO) mice (genotype Nr5a2 fl/fl ; Amhr2 Cre/+ ). Following DNA extraction from tails, littermates were genotyped. The genotyping primers bind to regions that are separated by one LoxP site in the region of intronic DNA between exons 3 and 4 of the Nr5a2 gene. Therefore, the genotyping PCR resulted in a single, larger PCR product in mice homozygous for the LoxP site, a single, smaller PCR product in wildtype mice, and two products in mice heterozygous for the LoxP site. Control mice (CON) mice in these trials were nonmutant, Nr5a2 fl/fl ; Amhr2 Cre/− females 70 . www.nature.com/scientificreports/ Histology and follicle counting. Ovaries from postnatal day 4 (PND4), PND13 and mature (five-month old) animals were serially sectioned at 5 µM and stained with hematoxylin and eosin. For follicle counting, 3-9 ovaries were assessed per genotype and follicles were classified according to the following criteria: primordial follicles displayed a single layer of squamous pre-granulosa cells around the oocyte, primary follicles were defined as those with one layer of cuboidal granulosa cells, secondary follicles were those with more than one layer of cuboidal granulosa cells surrounding the oocyte, and antral follicles were the tertiary follicles presenting an antrum. Follicles with multiple layers of disorganized granulosa cells, non-homogeneous theca layer, damaged oocytes or fragments of zonae pellucidae 50 were classified as atretic. For follicle enumeration in ovaries from PND4, PND13, and 5-month-old mice, multiple slides were counted for each ovary and data were expressed as an average number of unique follicles/slide. For cultured PND4 ovaries, only 1-2 slides/ovary were counted. Therefore, to normalize for differing section sizes, data were expressed as ratio of primordial to primary follicles and as percent of total follicles in the medulla. The Grubbs Outlier Test was performed and a single outlier was removed.  were collected and RNA purified as above. Quality of RNA was assessed using an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA) and only samples of high quality were used. The library preparation was performed with a New England Biolabs mRNA stranded library preparation kit (Rowley, MA) and library QC followed. Pair-ended sequencing was performed at Génome Québec (Montreal, QC), according to their standard procedures with an Illumina HiSeq4000, generating 20,000,000 reads per sample of 100 base pairs per read.

RNA extraction and real
Reads with a quality score less than 20 were removed and alignment was performed using the STAR open source alignment tool 72 . The reference genome employed was the Genome Reference Consortium Mus musculus GRCm38 (mm10). To evaluate RNA abundance at the gene level, reads mapped to each gene were quantified using featureCounts. Normalization and analysis of differential abundance were performed using the package DEseq2 in R, with a Benjamini Hochberg False Discovery Rate (FDR) correction (FDR = 5%). Individual mRNA were considered significantly differentially abundant at Padj < 0.05 and fold change of at least 1.5 up or down (equivalent to log2 fold change of |0.585|). Pathway analysis was performed using Metascape 73 . For integration of these data with the dataset of Ernst et al., (2018), the data of Ernst et al., were accessed directly as reported and Microsoft Excel was used to find commonalities between the datasets. DAVID 74,75 was used for functional annotation. Heatmaps were drawn in the program Morpheus, which is freely available online at https ://softw are. broad insti tute.org/morph eus/.
Fluorescent immunocytochemistry. Slides of formalin or paraformaldehyde-fixed paraffin-embedded PND4 ovaries from control and cKO mice were rehydrated as previously described 21  CY3 conjugated goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories) was used as second antibody for PTEN, FOXL2, cleaved caspase 3 and KI-67 immunolocalization, at the same concentration. Finally, slides were counterstained with DAPI, 1:1000 in PBS for 5 min before being washed three times in PBS and mounted with Permafluor. CellProfiler version 4.06 Software 76 , which is freely available at www.cellp rofil er.org, was used to quantify signal intensity, by the following method: first, the color images were converted to grayscale by splitting, to conserve only the main channel (blue for DAPI, red for NR5A2 and PTEN immunofluorescence, and green for FOXL2). Based on DAPI images, the follicles or areas of interest were then manually delimitated and the intensity of the staining quantified. Finally, the controls were normalized to 1 and cKO intensity was calculated relative to 1.
Culture of neonatal ovaries. Postnatal day 4 (PND4) mice ovaries from CON and cKO animals were dissected and placed in 50 μl drops of medium and cultured for 10 days on 0.4 μm floating filters at 37 °C in a chamber containing 5% CO 2 . Filters were placed in four-well culture plate on 0.5 ml DMEM-F12 medium supplemented with 0.1% Albumax, penicillin-streptomycin, 0.1% fetal bovine serum, 27.5 μg/ml transferrin, and 0.05 mg/ml L-ascorbic acid 77,78 . Culture medium was changed every second day. After 10 days of culture, ovaries were fixed for 4 h in 4% PAF, washed with PBS and mounted in blocks before being immunostained with DDX4 antibody to facilitate follicle counting. www.nature.com/scientificreports/