Steroidogenic factor 1 (SF-1; Nr5a1) regulates the formation of the ovarian reserve

Oocyte cyst breakdown begins as early as ED17.5 in the mouse, and few early follicles are observed around this time (7, 8). Primordial follicle assembly continues through PND1 and is mostly complete by PND4 (Fig. 1A) (7, 8). To validate the use of the Amhr2-Cre-mediated depletion model as a tool to investigate the role of LRH-1 and SF-1 in the processes of oocyte cyst breakdown and primordial follicle assembly, we undertook temporal analysis of AMHR2 expression. The Amhr2 transcript was detectable in the wild-type fetal ovary on ED17.5, PND1, and PND4 by qPCR (Fig. 1B) and on ED17.5 by in situ hybridization (RNAscope) (Fig. 1C). AMHR2 protein was detectable by immunohistochemistry at the same developmental times (Fig. 1D). Next, we evaluated the temporal expression of SF-1 and LRH-1 in the wild-type murine ovary during the period of primordial follicle assembly. There was a substantial increase in Nr5a2 on PND4 compared to ED17.5 and PND1 (Fig. 1E). LRH-1 was not expressed on ED17.5 but was detectable in the pregranulosa cells of a subset of primordial follicles on PND1 (Fig. 1F), consistent with our previous observation of LRH-1 on PND4 (14). In contrast, SF-1 was strongly and specifically expressed in all somatic cells as early as ED17.5, but not oogonia or oocytes, and its abundance did not change appreciably throughout primordial follicle assembly (Fig. 1 G and H).

To explore the evolutionary conservation of SF-1 and LRH-1 expression during the developmental trajectory of the ovary, we evaluated the expression in silico of Nr5a1 and Nr5a2 transcripts in diverse species, including humans, marsupials, and birds, using the Evo-Devo database [https://apps.kaessmannlab.org/evodevoapp/; (26). In all six species analyzed, Mus musculus (mouse), Rattus norvegicus (rat), Homo sapiens (human), Oryctolagus cuniculus (rabbit), Monodelphis domestica (opossum), and Gallus gallus (chicken), Nr5a1 was robustly expressed before and throughout the period of follicular assembly, whereas Nr5a2 expression was nearly undetectable (Fig. 2), confirming our previous observation. The prenatal detection of SF-1 in this wide range of species led us to infer that there is a role for this factor in development of the primordial follicle population that is conserved across vertebrate taxae.

Next, we confirmed the effectiveness of the Amhr2-Cre in depletion of ovarian SF-1 and LRH-1. On ED17.5 (SI Appendix, Fig. S1 A and B), PND1 (SI Appendix, Fig. S1 C and D), and PND4 (SI Appendix, Fig. S1 E and F), both Nr5a1 transcripts and SF-1 protein were substantially depleted as a result of Cre-mediated recombination. Similar patterns of depletion were observed in LRH-1 cKO mice on PND1 (SI Appendix, Fig. S2 A and B). To confirm that Amhr2-Cre-mediated excision was restricted to the ovary, other sites of SF-1 expression, the adrenal gland and spleen, were evaluated. On ED17.5 and PND1, the relative abundance of Nr5a1 was unaffected by genotype in these tissues (SI Appendix, Fig. S1 G, H, J, and K). In addition, the abundance of Cyp11a1, a steroidogenic enzyme and key downstream target of SF-1 (38), was unchanged in SF-1 cKO adrenals (SI Appendix, Fig. S1 I and L) indicating ovarian specificity of Amhr2-Cre depletion. The hypothalamus and pituitary are also sites of SF-1 expression. Nevertheless, previous work has demonstrated normal follicular development to the secondary or early antral stage in mice with hypothalamus- (39) or pituitary gonadotrope-specific (40) depletion of SF-1. Therefore, evaluation of the hypothalamus and pituitary even for nonspecific Cre activity or SF-1 depletion was judged irrelevant to evaluation of an ovary-specific phenotype at this early stage of ovarian development. We then used this model to deplete SF-1 and LRH-1 during ovarian development in the mouse, omitting investigations of ED17.5 in the LRH-1 model, given that the LRH-1 signal is undetectable in the ovary at this time.

Ovarian volume was not reduced in SF-1 cKO mice on ED15.5 (prior to the beginning of assembly; Fig. 3A), despite expression of SF-1 on this day (SI Appendix, Fig. S3). ED15.5 ovaries were also histologically indistinguishable from control counterparts collected on the same day (Fig. 3B). In distinct contrast, on ED17.5, PND1, and PND4, loss of SF-1 substantially reduced ovarian volume and prevented the ovarian growth observed in control ovaries (Fig. 3A and SI Appendix, Fig. S4). In addition, the number of oocytes in nests tended to be reduced in cKO ovaries on ED17.5 and PND1 (Fig. 3 C and D). Thus, the onset of a phenotype resulting from depletion of SF-1 with Amhr2-Cre is approximately concurrent with the onset of oocyte nest breakdown and follicular assembly.

In keeping with the ovarian volume results, the number of primordial follicles was substantially reduced in cKO ovaries during the period of primordial follicle assembly (PND1; Fig. 3E). To test whether the reduced primordial follicle population was a result of a reduced oocyte population, we calculated the number of primordial follicles per oocyte. This population was also reduced in cKO ovaries (Fig. 3E), demonstrating that there was a reduced rate of primordial follicle assembly. The phenotype of reduced ovarian reserve was maintained on PND4 (Fig. 3F) and PND13 (Fig. 3G), showing that conditional depletion of SF-1 has a persistent effect on the size of the ovarian reserve.

Quantification of primordial follicles in LRH-1 cKO mice on PND1 revealed no difference in either follicle or oocyte numbers (SI Appendix, Fig. S5). In addition, there was no phenotype of reduced ovarian volume (SI Appendix, Figs. S4 and S5). Despite the lack of a histological phenotype, qPCR analysis of regulators of cyst breakdown and primordial follicle assembly in LRH-1 cKO mice revealed changes in transcript abundance of anti-Müllerian hormone (Amh) on PND1, while thirteen other genes were unaffected and three [growth differentiation factor 9 (Gdf9), cadherin 1 (Cdh1), and snail family zinc finger 1 (Snai1)] tended to change (P = 0.08). In summary, the change in transcript abundance in response to loss of LRH-1 was minor to undetectable (SI Appendix, Fig. S5), consistent with a lack of histological phenotype.

To further confirm the lack of role of LRH-1 in primordial follicle assembly, we generated a line of mice with depletion of both LRH-1 and SF-1 (double cKO). These mice recapitulated the phenotype of SF-1 cKO mice on PND4, with a substantially reduced pool of ovarian primordial follicles and reduced ovarian size (Fig. 3H). Nevertheless, double cKO follicular populations did not differ from those in the SF-1 cKO model, indicating no additive effect. In summary, depletion of SF-1, but not LRH-1, repressed oocyte nest breakdown and follicular assembly, thus substantially reducing the ovarian reserve.

To delineate the SF-1-regulated mechanisms altering follicle assembly, we compared transcriptional profiles of ovaries collected from SF-1 cKO and control mice on ED17.5, PND1, and PND4 (Datasets S1–S7, NCBI Gene Expression Omnibus, accession number GSE216420). Analysis of the mRNAseq dataset revealed that age and genotype were the primary drivers of variation, as exemplified by principal component analysis (Fig. 4A). On both PND1 and PND4, more genes were up-regulated than down-regulated by the loss of SF-1 (Fig. 4B). Thirty-six mRNAs with known importance in ovarian development were evaluated by qPCR, and there was clear agreement of qPCR with mRNAseq results (SI Appendix, Tables S1 and S2). There was a substantial (~20 to 50%) overlap among transcripts that differed in SF-1 cKO ovaries on each of the three days investigated, suggesting that the disrupted transcriptional program caused by SF-1 depletion is maintained through ovarian development (Fig. 4C). To identify the salient disruptions resulting from loss of SF-1, pathway analysis was conducted on transcripts that differed in SF-1 cKO ovaries on all three days of ovarian development, comprising 216 genes in total (Fig. 4C). Among the pathways we identified were embryonic organ morphogenesis and gonad development (Fig. 4D and Dataset S8), pathways that included the Iroquois homeobox (IRX) transcription factors Irx3 and Irx5.

Gene ontology analysis of transcripts up-regulated in response to SF-1 depletion on ED17.5 revealed increases in transcripts associated with extracellular matrix formation and cell migration. Down-regulated pathways included hormone production and cell signaling (SI Appendix, Fig. S6). Notably, gonad development was among the down-regulated pathways, highlighting the importance of SF-1 in ovarian development. Conditional depletion of SF-1 on PND1 resulted in increases in transcripts in pathways associated with extracellular matrix and cell migration and downregulation of pathways related to metabolism.

More than 84% of the transcripts that differed in SF-1 cKO ovaries relative to control had a predicted Nr5a1/Nr5a2 motif found in the territory of the gene, including proximal sites 5.0 kb upstream and 1.0 kb downstream and distal sites up to 500 kb. This compares to identification of predicted Nr5a1/Nr5a2 motifs in only 30% of genes that did not change in SF-1 cKO ovaries (Fig. 4E and Dataset S9). In addition, transcripts that changed in SF-1 cKO ovaries had more total Nr5a1/Nr5a2 motifs in cis-regulatory regions of corresponding genes than did unaffected genes (SI Appendix, Fig. S7). This illustrates that the observed changes are likely directly regulated by SF-1.

It is known that IRX3 and IRX5 are essential for formation of granulosa cell basement membranes in follicular assembly (41, 42). The transcripts encoding these homeobox factors were down-regulated in SF-1 cKO ovaries (Fig. 5A), and the cis-regulatory regions of the genes had the predicted Nr5a1/Nr5a2 motifs (Dataset S9). Their inhibition in the cKO model was confirmed by in situ hybridization (Fig. 5B). The granulosa cell basement membrane is composed primarily of laminins, and its formation is regulated by IRX3 and IRX5 (42). Thus, the potential exists for IRX genes to be mechanistic regulators downstream of SF-1 action in the perinatal ovary. Indeed, in the SF-1 cKO ovary, there is a switch in the aggregate laminin composition, with laminin alpha (Lama)1 and Lama3 decreasing and Lama2, Lama4, and Lama5 increasing (Fig. 5C). Reduced Lama1 by qPCR confirmed the observed depletion of this transcript (Fig. 5D). The changes in laminin composition in the cKO ovary resulted in an overall increase in total ovarian laminin protein (Fig. 5 E and F) and completely disrupted ovarian laminin deposition (Fig. 5F). In control ovaries, a single, distinct follicular laminin basement membrane was evident, while in cKO ovaries, laminin was distributed throughout the ovarian stroma and between granulosa cells, in contrast to the normal pattern of organization (Fig. 5F and SI Appendix, Fig. S8). IRX3 and IRX5 also regulate formation of oocyte-granulosa cell gap junctions (42). In this context, the deposition of connexin 43 (CX43) protein was disrupted in SF-1 cKO ovaries, similar to that observed in IRX3 and IRX5 mutants (42) (SI Appendix, Fig. S9). Changes to laminins and gap junctions would be expected to result in misassembled follicles. Indeed, in ovaries collected from mature SF-1 cKO mice, inappropriately formed, disorganized follicles with small oocytes were evident (Fig. 5G). Taken together, these results demonstrate an essential role for SF-1 in the formation of the follicular structure and the granulosa–oocyte communication network.

To determine whether loss of SF-1 alters any normal follicular assembly processes, the global RNA analysis of temporally regulated (PND1 vs. ED17.5) transcripts in control mice (SI Appendix, Fig. S10) was compared to that from the cKO ovary. In total, there were 317 transcripts that changed between ED17.5 and PND1 (Padj < 0.05; log2 fold change > |1|) in control ovaries and failed to change in the cKO tissue (Padj > 0.15; log2 fold change < |1|; Fig. 6A). The transcripts for physiological mechanisms that decreased over time only in the control ovaries were immune related, including pathways related to response to pathogens, cell adhesion and chemotaxis, immunodeficiency, and phagocytosis (Fig. 6 A and B and SI Appendix, Fig. S11). These results suggest the presence of an inappropriate inflammatory environment in SF-1 cKO ovaries, which may be either a cause or consequence of increased oocyte death.

We hypothesized that in addition to resulting in inappropriately formed follicles, depletion of SF-1 would result in oocyte death during oocyte cyst breakdown. One potential cause is inappropriate inflammation in cKO ovaries, as described above. Notch signaling, a highly conserved mechanism of intercellular communication, is also known to promote oocyte demise during follicular assembly. Indeed, inhibition of the Notch pathway via granulosa-specific depletion of Notch receptor (NOTCH) two results in increased oocyte survival (43). We found that elements of the Notch pathway were up-regulated in cKO ovaries (Fig. 6C), including Notch receptors (Notch1, Notch2, Notch3, and Notch4), Notch ligands [Delta-like canonical notch ligand (Dll)1, Dll4, and Jagged canonical notch ligand 1(Jag1)], and Notch regulators [LFNG O-fucosylpeptide 3-beta-N-acetylglucosaminyltransferase (Lnfg) and MFNG O-fucosylpeptide 3-beta-N-acetylglucosaminyltransferase (Mfng)]. These changes were confirmed by qPCR (SI Appendix, Fig. S12 and Tables S1 and S2). We next evaluated well-known markers of apoptosis in cKO and CON ovaries and found that these were mostly unchanged (SI Appendix, Fig. S13 A and B). A key mechanism of oocyte elimination is the DNA damage response pathway (44, 45). All genes associated with the gene ontology (GO) term DNA damage response were assessed, and a subset was found to be affected in SF-1 cKO, particularly on ED17.5 (SI Appendix, Fig. S14), suggesting modulation of the DNA damage response may be an early mechanism of SF-1-mediated increase in oocyte death. The complement pathway has been implicated as an effector in oocyte attrition downstream of the DNA damage mechanism (46). Remarkably, among 61 genes associated with the GO term activation of complement, 15 increased in cKO ovaries on PND1, a day of maximal oocyte attrition (Fig. 6D), and eight increased in cKO ovaries on ED17.5 (SI Appendix, Fig. S15). These findings implicate the immune-mediated complement pathway as a factor in the increased oocyte attrition observed in SF-1 cKO ovaries.

Glycogen synthesis kinase-3 beta (Gsk3b) is upstream of the oocyte attrition process mediated through DNA damage checkpoint mechanisms (47) and modulates apoptosis via effects on the translocation of beta-catenin and Trp63 expression (47). Although the specific mechanism involving Gsk3b seems to be largely unaffected by depletion of SF-1 (SI Appendix, Fig. S16A), canonical Wnt signaling was modestly modulated by depletion of SF-1, particularly on PND1 (SI Appendix, Fig. S16B).

Autophagy has also been implicated in oocyte survival and death during primordial follicle assembly, with both excess autophagy (48) and reduced autophagy (49) causing oocyte death. Many genes associated with the GO term regulation of autophagy changed in SF-1 cKO ovaries, with the most distinct changes occurring on PND1 (SI Appendix, Fig. S17 A and B). One outcome of autophagy is an increase in lysosome formation and lysosomal degradation. In cKO ovaries on PND1, transcripts coding for cathepsins, effectors of this pathway, were in substantially greater abundance relative to control counterparts (SI Appendix, Fig. S17C). This increase in lysosomal degradation is a potential driver of increased oocyte death in SF-1 cKO ovaries.

To confirm the increase in oocyte death in SF-1 cKO ovaries, a TUNEL assay was performed to label dead oocytes. On ED17.5 and PND1, when oocyte attrition is rampant, dead oocytes were identified in both SF-1 cKO and CON ovaries, with more oocyte death evident in the cKO ovaries than in control counterparts (Fig. 6E). Few dead oocytes were detectable on PND4 in either control or cKO ovaries (SI Appendix, Fig. S13C). In summary, SF-1 may promote, or perhaps simply permit, the survival of oocytes, as demonstrated by increased oocyte death in SF-1 cKO ovaries.

Intercellular signaling regulated by KITL and its interaction with KIT proto-oncogene receptor, tyrosine kinase (KIT), promote cell survival and are among the best-known regulators of follicular activation. This pathway has also been identified as a regulator of germ cell cyst breakdown and follicular assembly (50). Herein, we show that the oocyte-specific receptor Kit was unaffected by loss of SF-1 in ED17.5 and PND1 ovaries, but Kitl was less abundant on both days (Fig. 7A). Ovarian culture experiments tested whether recombinant KITL rescued the phenotype of reduced ovarian reserve. Cultures were initiated at ED17.5, and, after 6 d of treatment, cKO ovaries cultured with KITL had increased ovarian reserve (Fig. 7 B and C) resulting in a follicular population comparable to untreated control ovaries. Interestingly, in cKO ovaries that were treated with KITL, unusual follicular phenotypes were observed, including double oocyte follicles (Fig. 7D) and follicles with large-volume oocytes but flattened, primordial granulosa cells (Fig. 7E). This suggests that although KITL treatment partially rescued the phenotype of reduced ovarian reserve, follicular dysfunction remains, perhaps due to the malformation of the laminin follicular basement membrane.

All mouse experiments were conducted according to the guidelines of the Canadian Council of Animal Care and approved by the Université de Montréal (UdeM) Comité d’éthique de l’utilisation des animaux. The mice used were of the C57BL/6 genetic background, were housed in the laboratory animal facility at the UdeM Faculté de médecine vétérinaire, maintained on a 14 h light, 10 h dark cycle, and provided with pelleted feed and water ad libitum.

The Nr5a1 floxed mice were obtained from The Jackson Laboratory (Stock No. 007041; B6Ei.129P2-Nr5a1tm2Klp/EiJ; referred to as Nr5a1 f/f) and have been described previously (24, 25, 73), and the Nr5a2 floxed mice (referred to as Nr5a2 f/f) have been described in detail by our research group and others (14 –16, 74). These mice were bred to mice with Cre-recombinase expression driven by the Amhr2 promoter (21). Detailed information on genotyping and sample collection is in SI Appendix, Methods.

Ovarian culture was conducted as reported previously (14, 65, 75), with complete description in SI Appendix, Methods.

For follicle quantification, fixed ovaries were sectioned at 4 μm, and every 8th section was saved for counting, as previously described (76, 77). Thus, sections were separated by approximately 28 μm, ensuring that no follicles were counted twice and all regions of the ovary were represented. Further detail is in SI Appendix, Methods.

Details of RNA isolation, library construction, and qPCR and RNAseq data analysis may be found in SI Appendix, Methods. Sequencing was performed at the RNomics platform at Université de Sherbrooke using Illumina NextSeq, with NextSeq® 500/550 High Output Kit v2 (75 cycles). Approximately 15 million reads per sample were generated. Quality was assessed using FASTQC (78), and data were of high quality. Data were analyzed using the GenPipes RNAseq pipeline (Python version 3.9.1 and GenPipes version 4.3.0), with default settings (79) and genome version GRCm38 (mm10). The raw count matrix obtained from GenPipes was normalized and statistically analyzed using DEseq2 (80). mRNAseq data are available in the NCBI Gene Expression Omnibus, accession number GSE216420.

The RNAscope platform (ACDBio) was used for in situ hybridization, according to the manufacturer’s protocol and as previously described (81). All immunohistochemistry experiments were carried out according to the same general protocol, with minor modifications for some antibodies (SI Appendix, Table S4) according to Meinsohn et al (14). To confirm specificity of binding, a secondary antibody only control was run with each experiment (SI Appendix, Fig. S18). Additional detail may be found in SI Appendix, Methods.

Data were analyzed by the t test or ANOVA, as appropriate. A Welch’s correction was used in cases of unequal variances, and outliers were identified using a Grubbs outlier test and removed. Ovarian culture data were analyzed by a paired t test, with mouse as pairing factor. P < 0.05 was considered statistically significant in all cases. Graphpad prism was used for graphing and statistical analysis. For mRNAseq data, data were analyzed in DEseq2 as described above, and transcripts were considered statistically different if Padj < 0.05, with an additional requirement for Log2fc > |1| for inclusion in pathway analyses.

C.H.K.H., O.E.S., N.G., and B.D.M. designed research; C.H.K.H., O.E.S., M.-C.M., and M.B. performed research; C.H.K.H., M.B., N.G., and B.D.M. analyzed data; and C.H.K.H., O.E.S., M.-C.M., M.B., N.G., and B.D.M. wrote the paper.

The authors declare no competing interest.