Postnatal Ovarian Transdifferentiation in the Absence of Estrogen Receptor Signaling Is Dependent on Genetic Background

Abstract

Normal ovarian function requires the expression of estrogen receptors α (ESR1) and β (ESR2) in distinct cell types within the ovary. The double estrogen receptor knockout (αβERKO) ovary had the appearance of seminiferous tubule–like structures that expressed SOX9; this phenotype was lost when the animals were repeatedly backcrossed to the C57BL/6J genetic background. A new line of ERKO mice, Ex3αβERKO, was developed for targeted disruption on a mixed genetic background. Histological examination of the ovaries in the Ex3αβERKO showed the appearance of seminiferous tubule–like structures in mice aged 6 to 12 months. These dismorphogenic regions have cells that no longer express granulosa cell–specific FOXL2, while other cells express Sertoli cell–specific SOX9 as examined by immunohistochemistry. Whole ovarian gene expression analysis in Ex3αERKO, Ex3βRKO, and Ex3αβERKO found many genes differentially expressed compared to controls with one Esr1 and Esr2 allele. The genes specific to the Ex3αβERKO ovary were compared to other models of postnatal ovarian transdifferentiation, identifying 21 candidate genes. To examine the genetic background contributions, DNA was isolated from αβERKO mice that did not show ovarian transdifferentiation and compared to DNA from Ex3αβERKO using Mouse Diversity Array. A genomic region putatively associated with transdifferentiation was identified on Chr18 (5-15 M) and genes in this region were compared to the genes differentially expressed in models of ovarian transdifferentiation. This work demonstrates the importance of ESRs in maintaining granulosa cell differentiation within the ovary, identifies several potential gene candidates, and suggests that genetic background can be a confounding factor.

Ovarian development and function require coordinated signaling events that ultimately give rise to functional ovarian follicles consisting of several cell types, including oocytes, granulosa cells, and theca cells. Estrogen receptor (ESR) signaling is essential for folliculogenesis and ovulation, where each ESR specific knockout (KO) mouse has a unique ovarian phenotype (1-7). Esr1 (ERα) is expressed predominately in theca cells, and knockout mice have hemorrhagic and cystic ovarian follicles, altered hormonal production, and no evidence of ovulation (8-12). Esr2 (ERβ) is expressed in granulosa cells and knockout mice are subfertile to infertile due to reduced ovulatory capacity (4, 11, 13-15) and have altered gonadotropin response in granulosa cells (15). The single-ESR mouse models demonstrated that ESR signaling is not necessary for fetal ovarian differentiation or early ovarian development; however, a complete loss of ESR signaling leads to postnatal ovarian transdifferentiation of granulosa cells to Sertoli-like cells typically found in the testis (11, 16, 17). A similar phenotype is observed in the ovaries of mice lacking Cyp19a1/P450 aromatase when animals are fed a soy-free diet (18-22). The loss of granulosa cell differentiation in adult ovaries suggests that proper estrogen signaling is required for ovarian maintenance.

During gonadal sex determination in mammals, the bipotential gonad differentiates into either a testis or ovary depending on signals present in the microenvironment (23-25). In the presence of the sex-determining region of the Y chromosome (SRY), Sox9 (26) expression is induced contributing to increased expression of Fibroblast growth factor 9 (Fgf9) (27), and doublesex and mab-3 related transcription factor 1 (Dmrt1) in Sertoli cells leading to normal testis development (28, 29). Conversely, ovarian development is initiated with wingless-type MMTV integration site family member 4 (Wnt4) (27, 30, 31), along with increased expression of R-spondin 1 (Rspo1) (32-34) and Ctnnb1/CTNNB1 activity (30, 34). Ovarian development also requires forkhead transcription factor L2 (Foxl2) (35-37). Disruption of a number of genes can lead to neonatal gonadal dysgenesis or sex reversal (25, 28, 38-46); however, gonadal transdifferentiation in postnatal models is less common. The loss of estrogen signaling, either through deletion of both ESRs in the αβERKO (11, 16), or the Cyp19a1 KO (18-21), demonstrates the requirement of estrogen to maintain the granulosa cells in the ovary. Postnatal granulosa to Sertoli-like transdifferentiation was shown in adult ovaries in mice lacking Foxl2 (38) and tripartite motif-containing 28 (Trim28) (47) or ectopic expression of testis-specific Dmrt1 in the ovary (48, 49). The loss of both ESRs in the ovary leads to the development of seminiferous tubule–like structures in adult animals, even though at the time of birth, ovarian development appears normal (16). Although this phenotype was first published in 1999 in αβERKO mice (16), the animals were on a mixed C57BL/6J and 129S6 background, and the phenotype was lost when the animals were fully backcrossed onto the C57BL/6J genetic background (Korach laboratory). The creation of a new line of ESR knockout (ERKO) mice, specifically Ex3αERKO (10) and Ex3βERKO (15), generated animals that have a genetic background similar to the original αβERKO mice (16) and provided an opportunity to examine ovarian transdifferentiation in the absence of ESR signaling and the role of genetic background contributions in maintenance of granulosa cells in the ovary. Herein, we describe the appearance of seminiferous tubule–like structures in the ovary of Ex3αβERKO mice, including reduced expression of granulosa cell–specific FOXL2 and increased expression of Sertoli cell–specific SOX9 in certain areas of the ovary. Analysis of gene expression using microarrays demonstrated that individual single and double ERKO ovaries had unique transcriptomes, providing insight into possible genes involved in ovarian transdifferentiation. Further examination of mouse single nucleotide polymorphisms (SNPs) using the Mouse Diversity Array (50) identified a putative genomic region on chromosome 18 that potentially contributes to the maintenance of granulosa cell differentiation in adult ovaries and could prevent ovarian transdifferentiation.

Methods

Animals

Animal procedures were performed under approval from the National Institute of Environmental Health Sciences Institutional Animal Care and Use Committee (Protocol #01-30). Animals were fed NIH-31 chow and maintained on a 12-hour light and dark cycle. The generation of ExαβERKO mice requires at least 4 rounds of breeding to first generate heterozygous Ex3αERKO and Ex3βERKO male and female animals. Each single Ex3αERKO animal is infertile (males and females) while the Ex3βERKO females are subfertile to infertile (10, 15). The double heterozygous ExαβERKO females were then bred with males that were heterozygous Ex3αERKO and either heterozygous or homozygous Ex3βERKO to generate Ex3αβERKO females. One in 32 mice were needed to generate a female homozygous for Ex3αβERKO. Animals were genotyped as previously described (10, 15). Control animals used in this study had at least one copy of each ESR allele present. Animals were aged 6 to 12 months and euthanized via CO₂ followed by cervical dislocation and tissue collection. Genotypes were confirmed after tissue collection and prior to further analysis. Ovaries were collected, and one was fixed in 10% formalin for histological analysis while the other was snap frozen and stored at −80 °C for further analysis.

Histology and Immunohistochemistry

Ovaries were fixed in 10% formalin, embedded in paraffin, and then sectioned and stained using standard hematoxylin and eosin (H&E) protocols. Immunohistochemical analysis was performed using antibodies specific for SOX9 (RRID: AB_778028) and FOXL2 (RRID: AB_2106188). Briefly, antigen retrieval with a citrate solution (Biocare Medical, Pacheco, CA) followed by blocking with endogenous peroxidase, followed by normal serum, and then Avidin D Blocking Kit (Vector Laboratories, Burlingame, CA). Slides were incubated at room temperature for 1 hour in primary antibody, washed, and then incubated with secondary for 30 minutes. Slides were then incubated in extra avidin peroxidase for 30 minutes, followed by 3 washes, and staining was visualized with Vector® NovaRed® Peroxidase (HRB) Substrate Kit (Vector Laboratories, Burlingame, CA). The slides were dehydrated through graded ethanol, cleared in xylene, and coverslipped using Permount. Images were taken using an EVOS XL Core Cell Imaging System (ThermoFisher Scientific, USA) microscope using coverslip corrected objectives.

Luteinizing Hormone Assay

Blood samples were collected via cardiac puncture after euthanasia, serum was immediately separated and stored at −80 °C until analysis. Serum samples were assayed for luteinizing hormone (LH) using a sensitive low-volume immunoassay previously described (10, 51).

Gene Expression and Microarray Analysis

RNA was isolated from frozen ovaries using TRIzol® (Invitrogen, Carlsbad, CA) following the manufacturers’ protocol. One microgram (1 µg) of RNA was reverse transcribed to cDNA using the SuperScript First-Strand Synthesis System (Invitrogen, Carlsbad, CA). Data are shown as a ratio of gene of interest to Rpl7 as previously described (52). Primer sequences are available upon request.

Gene expression analysis was performed in the National Institute of Environmental Health Sciences (NIEHS) Microarray Core using Agilent Whole Mouse Genome 4 × 44 multiplex format oligo arrays (014868) following the manufacturers’ protocol (Agilent Technologies, Santa Clara, CA). Briefly, 250 to 350 ng of total RNA was Cy3-labeled for each sample control (n = 4), Ex3αERKO (n = 4), Ex3βERKO (n = 4), and Ex3αβERKO (n = 4). For each sample, 1.65 µg of Cy3-labeled cRNAs were fragmented and hybridized for 17 hours in a rotating hybridization oven, washed, and scanned with an Agilent Scanner. Data was obtained using Agilent Feature Extraction software (v9.5) and data was analyzed using Partek® Genomics Suite® (Partek Incorporated, St. Louis, MO). Differentially expressed probes were identified using analysis of variance (ANOVA) and a false discovery rate P value of P < .05 to designate significant differences in transcripts between control and Ex3αERKO, Ex3βERKO, or Ex3αβERKO, respectively. In addition, a fold change of 2 was used to generate the gene lists used for further analysis between each group. Mapped genes were identified, gene lists were compared, and pathway analysis was performed using Ingenuity Pathway Analysis (IPA, Qiagen, Germantown, MD). All microarray data files are available from the GEO database (GSE282858).

To compare transcriptome changes in postnatal ovarian transdifferentiation models, we compared the Ex3αβERKO microarray data with the aromatase (Cyp19A1) knockout (Cyp19a1 KO) (21) and Foxl2 conditional KO (cKO) (38) datasets, using Partek Genomics Suite. Differentially expressed genes in each dataset were identified using ANOVA and a false discovery rate P value of P < .05. Additionally, a fold change of 2 was used for genes in the Ex3αβERKO compared to control (2457 probes) and for the Foxl2 cKO compared to wild-type (845 probes), while a fold change of 1.5 was used for genes in the Cyp19a1 KO compared to wild-type (157 probes). The analysis was done in this manner due to the microarray platforms being different in each study, and the gene list generated for the Cyp19a1 KO was shorter in our analysis than reported in their study (21). These genes were then compared to the RNA-seq dataset from the Trim28 cKO (42) to generate a list of genes differentially expressed in 4 models of postnatal ovarian transdifferentiation (Table 1).

Table 1.

Common genes differentially expressed in mouse models of postnatal ovarian transdifferentiation

Gene	Ex3αβERKO	Cyp19a1 KO (21)	Foxl2 cKO (38)	Trim28 cKO (47)	Official name & function
Cidea	89.41	4.56	15.86	139.09	Cell death-inducing DFFA Like Effector A. Activator of apoptosis.
Rarres1	53.74	2.99	5.86	101.78	Retinoic acid receptor responder 1. Role in cell proliferation, colony formation, cell cycle, and migration.
Defb36	27.03	13.39	14.84	51.56	Defensin beta 36. Potential role in innate immune response.
Cldn11	25.63	10.89	20.56	18.03	Claudin 11. Tight junction protein expressed in CNS and blood-testis barrier.
Cst9	23.53	8.73	40.35	31.93	Cystatin 9. Possible protease inhibitor and may be involved in testis development and hematopoietic differentiation.
Tmc7	22.64	1.78	12.72	8.13	Transmembrane Channel Like 7. Predicted ion channel.
Ttyh1	17.99	4.02	34.25	51.95	Tweety family member 1. Chloride anion channel and plays a role in maintaining neuronal stem cells via Notch signaling.
Itgbl1	11.54	3.03	3.60	7.28	Integrin, beta-like 1. Beta integrin related protein member of the EGF-like family.
Sox9	8.11	4.19	12.61	12.44	SRY (sex-determining region Y)-box 9. DNA binding transcription factor in the testis.
B4galnt1	4.55	1.93	3.71	7.99	Beta-1,4-N-acetyl-galactosaminyl transferase 1. Acts within glycosphingolipid metabolic processes, lipid storage, and spermatogenesis.
Gstm6	4.05	3.29	15.59	11.39	Glutathione S-transferase mu 6. Predicted to be involved in glutathione binding, transferase, and protein homodimerization.
Tbc1d9	3.97	1.67	2.46	3.35	TBC domain family, member 9. May be a GTPase-activating protein.
Map3k21	3.80	2.82	15.94	16.93	Mitogen activated protein kinase kinase kinase 21. Involved in protein phosphorylation and signal transduction.
Cp	3.57	4.66	4.30	7.05	Ceruloplasmin. Copper binding glycoprotein that contributes to iron homeostasis.
Timp2	3.54	3.00	2.64	4.32	Tissue inhibitor of metalloproteinase 2. Role in cell proliferation, regulation of MAPK, and neuron differentiation.
Sort1	3.06	2.15	3.13	2.06	Sortilin 1. Role in protein trafficking.
Jakmip1	2.94	2.67	6.21	11.24	Janus kinase and microtubule interacting protein 1. Role in kinesin binding activity and may act in protein trafficking and microtubule rearrangements.
Sgk3	2.91	2.02	4.97	2.68	Serum/glucocorticoid regulated kinase 3. Role in apoptosis signaling.
B4galt6	2.40	1.55	4.66	2.87	Beta-1,4-galactosyltransferase 6. Involved in ganglioside biosynthetic processes. May be active in Golgi apparatus.
Mns1	2.35	2.93	2.72	2.04	Meiosis-specific nuclear structural protein 1. Role in regulation of cilium assembly and sperm axoneme assembly.
Sdc1	−2.82	−2.18	−2.87	−6.67	Syndecan 1. Integral membrane protein.

Data shown are fold change compared to control or wild-type ovary for each genotype as appropriate for the study. Ex3αβERKO, Cyp19a1 KO, and Foxl2 cKO are microarray data while Trim28 cKO is from RNA-seq experiments. Gene name and descriptions were obtained from NCBI Gene Summary, GeneCards® and The Human Protein Atlas.

Genetic Analysis by Mouse Diversity Array

The original αβERKO mice were generated via insertional disruption (9, 14) while the Ex3αβERKO mice were generated via targeted deletion (10, 15); both strains were initially on a mixed C57BL/6J-129S6 mixed genetic background. DNA was isolated from frozen hypothalamus samples for 3 “original” αβERKO adult mice, tail samples from 9 experimental Ex3αβERKO mice and 3 littermate controls aged 1 year using QIAamp DNA kits (QIAGEN) following the manufacturers’ protocol. DNA was analyzed spectrophotometrically to quantitate and shipped to Jackson Laboratories (Bar Harbor, ME) where JAX Mouse Diversity Array (MDA) (50) was performed. The complete MDA data files are available upon contacting the corresponding author. The R MouseDivGeno Software Package (53) was used to generate allele calls for each SNP on the MDA. To analyze the dataset, all SNPs that were the same between C57BL/6J (B6) and 129S6 (129) were removed leaving approximately 23 000 SNPs. University of North Carolina (UNC) Compgen Tool Suite (54) was then used to visualize the haplotypes between the αβERKO and Ex3αβERKO experimental samples.

Results

Ovarian Morphology Shows Regions of Transdifferentiation in Ex3αβERKO Ovary

Several studies have examined the loss of ESRs in the ovary; these studies have been done primarily in animals with mixed genetic backgrounds (11, 16, 17). Studies in the Korach laboratory in 1999 showed that loss of both ESR1 (ERα) and ESR2 (ERβ) led to increased Sox9 expression and ovarian transdifferentiation (16); however, this phenotype was lost as the animals were fully backcrossed onto a C57BL/6J genetic background (our unpublished data). Development of new global knockout mouse lines for ERα and ERβ (10, 15) were generated in a mixed C57BL/6J and 129S6 genetic background. Because the genetic background of these new ERKOs is similar to that of the original αβERKO animals, single heterozygous mice were bred to obtain Ex3αβERKO, and female mice were aged up to 1 year. Histological analysis from both single and double Ex3ERKO mice show no obvious morphological changes in prepubertal ovaries of either genotype (data not shown). As the animals aged, the ovaries from the single KO mice had distinct phenotypes, where the Ex3αERKO mice had hemorrhagic and cystic ovaries (Fig. 1D) (10) and the Ex3βERKO mice have reduced ovulatory response including fewer large pre-ovulatory follicles (Fig. 1G) (15). Loss of both ERs in the Ex3αβERKO ovaries showed dismorphogenic regions that resembled seminiferous tubules (Fig. 1J) similar to those observed previously (16). The penetrance of these tubule-like structures and the age of appearance varied among the animals studied and was not obvious until histological examination of the ovary. The Ex3αβERKO ovaries also had increased collagen formation and hemosiderin ladened macrophages (stars) in the stroma (Fig. 1J-1O). The appearance of hemorrhagic cystic ovaries was also present in the Ex3αβERKO, presumably due to an increase in serum LH concentrations similar to the Ex3αERKO (Supplementary Fig. S1 (55)) (10) and the luteinizing hormone β c-terminal peptide (LHβCTP) transgenic mouse that has excess LH (56) although the penetrance of these cysts also varied in the Ex3αβERKO animals at all ages examined.

Figure 1.

Altered morphology and expression of SOX9 and FOXL2 in ERKO ovaries. Ovaries from 1-year-old control (A–C), Ex3αERKO (D–F), Ex3βERKO (G–H), and Ex3αβERKO (J–L) animals were formalin fixed and paraffin embedded. H&E staining demonstrate the presence of several follicles and a corpus luteum (CL) in control (A), Ex3αERKO has small follicles and a cyst (D), Ex3βERKO have large follicles and no CLs (G&M), and Ex3αβERKO show an ovarian follicle adjacent to several dismorphogenic tubule-like structures typically found in the testis (outlined in dotted lines). Immunohistochemical staining with granulosa cell–specific FOXL2 antibody has positive staining in granulosa cells within follicles in control (B), Ex3αERKO (E), Ex3βERKO (H), and Ex3αβERKO (K&N) ovaries. Serial sections were used for immunohistochemical analysis using a SOX9 antibody; while few cells stained positive in the control (C), Ex3αERKO (F), and Ex3βERKO (I), the Ex3αβERKO (L&O) showed the most positive cells, specifically in the dismorphogenic tubule-like structures typically found in the testis. Dismorphogenic regions were captured at 400× with the region in the outlined dotted lines visible on the left in (J*-L*) where the differentiation had occurred and cells lacked FOXL2, expressed SOX9, and had a seminiferous tubule–like appearance. The region outlined with dotted lines to the right (J^-L^) and in a different Ex3αβERKO ovary (M–O) show dismorphogenic regions that have cells lacking both FOXL2 and SOX9 (arrowheads). The Ex3βERKO and Ex3αβERKO ovaries have increased hemosiderin laden macrophages (marked with a star) while the control and Ex3αERKO do not. Scale bar, 200 microns, A–L.

Immunohistochemical analysis was performed on serial sections to examine expression of granulosa cell–specific FOXL2 and Sertoli cell–specific SOX9 for potential transdifferentiation in these regions that resemble seminiferous tubules typically seen in the testis. As expected, FOXL2 expression was present in the granulosa cells within different follicles in control, Ex3αERKO, Ex3βERKO, and Ex3αβERKO mice (Fig. 1B, 1E, 1H, and 1K). FOXL2 staining was also present in the corpus luteum (CL) formed in the control ovary (Fig. 1B). Testis-specific SOX9 staining was absent in granulosa cells of the control (Fig. 1C) and Ex3βERKO (Fig. 1I) ovaries, and with increased SOX9-positive cells observed in the Ex3αERKO ovary (Fig. 1F), specifically on the edge of the follicle where theca cells are present. SOX9-positive cells were observed in some ovarian surface epithelial cells and the oviductal epithelium in the sections of the different genotypes observed. The Ex3αβERKO ovary exhibits regions with seminiferous tubule–like structures (Fig. 1J–1M, outlined in dotted lines on left), that do not express FOXL2 (Fig. 1K); however, the same areas show strong expression of SOX9 (Fig. 1L). These regions were captured at 400× (Fig. 1J*-1L*). Interestingly, the Ex3αβERKO ovary still contains some ovarian follicles, which express FOXL2 (Fig. 1J, 1K, 1M, and 1N). The seminiferous tubule–like structures in the Ex3αβERKO ovary specifically express SOX9 (Fig. 1L and 1O), indicating ovarian transdifferentiation is occurring within specific regions of the ovary. The ovary also has regions that appear dismorphogenic (Fig. 1J–1L, outlined in dotted lines on right) that were captured at 400× (Fig. 1J^-L^) and have varied expression of FOXL2 or SOX9 within the cells (Fig. 1J-1L, arrowheads) suggesting these may be intermediate structures for this transdifferentiation process. A second dismorphogenic region from a different Ex3αβERKO ovary is also shown (Fig. 1M-1O) to demonstrate the variability observed within animals and these intermediate structures prior to a seminiferous tubule–like appearance.

Changes in Ovarian Gene Expression

Due to the ovarian transdifferentiation demonstrated by the loss of granulosa cell–specific FOXL2, increased expression of SOX9, and reorganization of the cells observed in the Ex3αβERKO ovaries, we examined the expression of genes necessary for embryonic gonadal sex determination in whole ovaries from control, Ex3αERKO, Ex3βERKO, and Ex3αβERKO using quantitative reverse-transcriptase polymerase chain reaction (RT-PCR). These genes include ovarian-specific Foxl2, Wnt4, and Rspo1 and testis-specific Sox9, Dmrt1, and Fgf9. Loss of either single ESR does not alter Foxl2 expression, while Foxl2 expression is reduced in the Ex3αβERKO compared to control or either single ERKO, suggesting reduced numbers of granulosa cells (Fig. 2A). Wnt4/WNT4 is essential for ovarian development during embryogenesis (27, 30, 31); however, the loss of ESR signaling in the ovary does not alter the expression of Wnt4 in adult ovaries (Fig. 2B). The downstream regulator of WNT signaling Rspo1 is also necessary for gonadal ovarian development (32-34) and is increased in both the Ex3αERKO and the Ex3αβERKO (Fig. 2C). Increased expression of Sox9 is observed in the Ex3αβERKO ovary, although the increase is only significant compared to the control and Ex3βERKO, suggesting that the increased testosterone observed in the Ex3αERKO (10) may regulate an increase in Sox9 mRNA expression (Fig. 2D). Increased expression of SOX9 protein was observed in some cells around a follicle in the Ex3αERKO ovary, although similar expression was also observed in the control animals with at least one copy of each ESR (Fig. 1C and 1F). The Ex3αβERKO ovary showed SOX9-positive cells in regions of the ovary, which appear similar to seminiferous tubules typically found in the testis, suggesting that the increased mRNA observed was translated to protein (Fig. 1L). Dmrt1 expression increased specifically in Ex3αβERKO ovaries (Fig. 2E) while Fgf9 expression increased in the Ex3αβERKO compared to controls, but no differences were observed between the Ex3αERKO or Ex3βERKO ovaries (Fig. 2F), suggesting that their increased expression correlates with loss of ESR signaling in general.

Figure 2.

Altered expression of genes involved in embryonic sex determination in Ex3αβERKO ovaries. Whole ovaries were collected from 1-year old mice and snap frozen, RNA was isolated using TRIzol®, reverse transcribed into cDNA, and quantitative RT-PCR was performed using primers specific for ovarian-specific genes Foxl2 (A), Wnt4 (B), and Rspo1 (C), or testis-specific genes Sox9 (D), Dmrt1 (E) or Fgf9 (F). Data shown are a ratio of gene of interest compared to Rpl7 as mean ± SEM. Data was analyzed using one-way ANOVA followed by Tukey's multiple comparison test, different letters represent statistical significance among groups (P < .05).

While we observed significant differences in gene expression of genes important for embryonic gonadal differentiation including a reduction of ovarian Foxl2 and increased expression of testis-specific Dmrt1 and Sox9/SOX9, we hypothesize these changes alone are not sufficient enough to drive the ovarian transdifferentiation into the Sertoli-like cells observed in the Ex3αβERKO. Furthermore, these morphological changes are not observed in the ovary of either single ESR knockout ovaries, suggesting possible compensation supported by the increased expression of Esr1 in granulosa cells of Ex3βERKO animals (15). Whole ovarian microarray analysis was performed to examine the transcriptome of each ESR knockout ovary, the Ex3αβERKO ovary, and littermate controls with at least one functional allele of each ESR. Initial analysis found that the Ex3αERKO ovary had 832 probes differentially expressed compared to control, while the Ex3βERKO ovary had only 117 compared to control. In the Ex3αβERKO ovary, 2961 probes were differentially expressed. Further analysis of these differentially expressed probes by Ingenuity Pathway Analysis (IPA) indicated the Ex3αERKO had 645 analysis ready genes as defined by IPA, with 306 genes downregulated and 339 genes upregulated compared to control (Supplementary Table S1 (55)). The Ex3βERKO ovary had 94 genes differentially expressed, with 19 downregulated and 75 upregulated compared to control (Supplementary Table S2 (55)). The Ex3αβERKO ovary had the most differentially expressed genes, with 2193 genes, of which 770 were downregulated and 1423 were upregulated (Supplementary Table S3 (55)). Interestingly, 43 genes were differentially expressed in all 3 ERKO ovaries (Supplementary Table S4 (55)). While there were 1887 genes specific to the Ex3αβERKO ovary, the Ex3αERKO ovaries had 327 genes and the Ex3βERKO ovaries had 25 genes differentially expressed that were unique to these single KO ovaries (Fig. 3A) when compared to the control ovaries.

Figure 3.

Differentially regulated genes in Ex3ERKO ovaries and several models of postnatal ovarian transdifferentiation. Microarray analysis was done on ERKO ovaries and data was analyzed for each ERKO compared to control. (A) Venn diagram generated in Ingenuity Pathway Analysis using annotated gene lists identified using a false discovery rate of P < .05 and a fold change of 2. The single Ex3ERKO ovaries were each compared to control ovaries isolated from animals at the same age in order to generate each ERKO specific gene list (Supplementary Tables S1-S3 (55)). Ex3αERKO had 645 unique genes, the Ex3βERKO ovary had 94 unique genes, while the Ex3αβERKO had 2193 unique genes. There were 43 genes shared among the 3 models (Supplementary Table S4 (55)). (B) Venn diagram created from comparison of gene lists done in Partek Genomics Suite using false discovery rate of P < .05 and a fold change of 2 compared to appropriate control for the Ex3αβERKO and Foxl2 cKO. A fold change of 1.5 was used for the Cyp19a1 KO dataset. The numbers describe the probes present on each microarray as appropriate. (C) The 23 genes shared in B were compared to RNA-seq dataset from the Trim28 cKO. There are 21 shared genes that are listed in Table 1.

Because our analysis showed a large number of genes differentially expressed in the Ex3αβERKO compared to either single-ESR knockout ovary, we compared this dataset with 2 microarray datasets generated from models that demonstrate postnatal ovarian transdifferentiation. This includes the estrogen deficient Cyp19a1 KO mouse maintained on a soy-free diet (21) and the FoxL2 cKO mouse (38). Using Partek Genomics Suite, the raw data from the 3 experiments were analyzed. The Ex3αβERKO ovary had 2457 probes differentially expressed, while the Cyp19a1 KO ovary had 157 probes, and the conditional Foxl2 ovary had 845 probes differentially expressed (Fig. 3B). Of the overlapping probes significantly different in these ovarian models compared to their respective controls, only 23 genes were differentially expressed in all 3 models of ovarian transdifferentiation analyzed by microarray (Supplementary Table S5 (55)). These 23 genes identified were then compared with the RNA-seq dataset from Trim28 cKO mice that also demonstrate postnatal ovarian transdifferentiation (47) and have increased Sox9 expression in the ovary. Of these genes, 21 are differentially expressed in all 4 mouse models of postnatal ovarian transdifferentiation (Fig. 3C). Each of these genes with their primary functions are listed in Table 1.

Several genes were selected from Table 1 to confirm increased expression in Ex3αβERKO ovaries, including genes reported to be necessary for testis development, such as the tight junction protein Claudin 11 (Cldn11) expressed in the blood-testis barrier (57) and Cystatin 9 (Cys9), a protease inhibitor predicted to be involved in testis development and hematopoietic differentiation. Cldn11 and Cst9 are increased in both Ex3αERKO and Ex3αβERKO ovaries (Fig. 4A and 4B). Two genes involved in apoptosis, including Cell death-inducing DFFA Like Effector A (Cidea), an activator of apoptosis (58) and Serum/glucocorticoid regulated kinase 3 (Sgk3), an apoptotic signaling molecule, were studied. Cidea is highly expressed specifically in the Ex3αβERKO ovary (Fig. 4C), while Sgk3 is increased only in the Ex3αERKO ovary (Fig. 4D). Finally, Tweety family member 1 (Ttyh1), a gene involved maintaining neuronal stem cells through interactions with Notch signaling pathway (59, 60), was also shown to have increased expression specifically in the Ex3αβERKO ovary (Fig. 4E) while the retinoic acid receptor responder 1 (Rarres1), a gene implicated in cell proliferation and migration, was found upregulated in all 3 ESR knockout ovaries, with the only specific increases compared to control in the Ex3αβERKO (Fig. 4F). Some genes were specifically altered in the Ex3αβERKO, while other genes had similar expressions in the single-ESR ovaries, suggesting that there could be compensation in the absence of only one ESR.

Figure 4.

Genotypic variation in genes differentially expressed. Whole ovaries were collected from 1-year-old mice, RNA was isolated, reverse transcribed into cDNA, and quantitative RT-PCR was performed using primers specific for genes listed. Testis-specific genes (A–B) Cldn11 and Cst9 have similar expression in the Ex3αERKO and Ex3αβERKO ovaries. Genes implicated in apoptotic signaling including Cidea (C) is specifically increased in the Ex3αβERKO, and Sgk3 (D) was increased in the Ex3αERKO. Cell signaling molecule Ttyh1 (E) was increased in the Ex3αβERKO, while Rarres1 (F) had no significant difference between the ER genotypes. Data shown are a ratio of gene of interest compared to Rpl7 as mean ± SEM from a minimum of 5 samples except for Ex3βERKO for Cldn11 where SD is shown as n = 2. Data were analyzed using one-way ANOVA followed by Tukey's multiple comparison test, different letters represent statistical significance between groups (P < .05).

The initial observation that loss of ESR signaling led to the appearance of Sertoli-like cells in the ovary occurred in 1999 (16). In order to continue our studies and analysis of the unique phenotype, we backcrossed the mice to a pure C57BL/6J background in the colony at NIEHS; however, the phenotype was lost (Korach laboratory). As the Ex3αβEKRO mice again demonstrate similar ovarian transdifferentiation and the appearance of seminiferous tubule–like structures as the αβERKO did, DNA was isolated to examine the contribution of genetic background to the phenotype.

DNA was isolated from frozen tissue samples from adult αβERKO (16) (n = 3), which were repeatedly backcrossed onto the C57BL/6J genetic background and were reported to not have obvious tubule-like structures (personal communication). DNA was also isolated from 1-year-old Ex3αβERKO animals described in this study (n = 9), and an MDA analysis was performed, and then assignment of allele calls of C57BL/6J (B6) or 129S6 to each SNP was performed at The Jackson Laboratories. UNC Compgen Tool Suite (50) was then used to visualize haplotypes as homozygous B6 (yellow, top line), heterozygous B6/129S6 (purple, middle line) or homozygous 129S6 (blue, bottom line). Each individual chromosome was analyzed to compare genotypes for significant differences between the control αβERKO DNA and the experimental Ex3αβERKO DNA. A candidate region was identified on Chr18 (5-15 M) (Fig. 5A) where the experimental Ex3αβERKO DNA was genetically different from the original Ex3αβERKO DNA (samples denoted as C for αβERKO vs E for Ex3αβERKO in Fig. 5). This is shown where the samples C1-C3 show all yellow (single top line), demonstrating they are homozygous for B6 across this genomic region, while the experimental samples E1-E9 have either purple (middle) and/or blue (bottom) across part of this region. A principal component analysis (PCA) was performed using the reference strains B6 and 129S6 on Chr18: 5-15 MB to examine each control αβERKO compared to experimental Ex3αβERKO animals (Fig. 5B and 5C). The control original αβERKO samples (C1-C3) clustered with the B6 reference strain, while the experimental Ex3αβERKO samples showed 2 distinct clustering patterns. Five Ex3αβERKO samples clustered near the reference 129S6 (E1, 3, 4, 7, and 9), while 4 of the samples were heterozygous with one allele B6 and one allele 129S6 (E2, 5, 6, and 8). The genes from this predicted genomic region were exported from Mouse Genome Database (MGD) (61) at the Mouse Genome Informatics (MGI) website, The Jackson Laboratory, Bar Harbor, Maine (accessed August 12, 2023).

Figure 5.

Visualization of SNP allele calls on candidate chromosome 18 identified from MDA analysis. MDA analysis was done using 3 “original” αβERKO DNA samples (C1–3) and 9 Ex3αβERKO DNA samples (E1–9). DNA was isolated from tissues, quantified, and up to 5 ug was sent to The Jackson Labs for analysis. The MDA was run, and allele calls were made on each SNP on the array to either C57BL/6J (B6) or 129S6 the genetic strains used in generation of the knockout mice. Allele calls are demarked as homozygous B6 (yellow, top line), heterozygous B6/129S6 (purple, middle line), or homozygous 129S6 (blue, bottom line). Allele calls were visualized for all chromosomes and regions showing difference between the control and experimental samples. (A) A region of interest was identified on Chromosome 18 where the allele calls in αβERKO control DNA samples (C1-C3) were all B6 while the Ex3αβERKO experimental DNA samples (E1-E9) had predominately 129S6. (B-C) PCA analysis of the MDA genotypes for this 10 Mb Chr18 region for first and second PC (B) or first and third PC (C). Labels for each sample correspond to the Ch18 visualization in A with the reference genes for B6 or 129S6 as designated.

On Chr 18 (5 -15 M) MGI reported there are 66 protein coding genes, 40 pseudogenes, 3 QTL, 6 miRNA genes, 12 lncRNA genes, 1 rRNA gene, and 1 unclassified non-coding RNA gene (Supplementary Table S6 (55)) according to MGI feature classifications (61). These 66 protein coding genes were compared to the gene lists generated from the ERKO microarrays (Supplementary Tables S1-S3 (55)). In the in silico analysis, a glycoprotein Desmocollin 2 (Dsc2) is downregulated in Ex3αERKO (−4.29), Ex3βERKO (−7.47) and Ex3αβERKO (−4.61) ovaries. Dystrobrevin alpha (Dtna) is increased in Ex3αERKO (+3.9) and Ex3αβERKO (+3.9). Several genes in this Chr18 region of interest were found to be specifically expressed only in the Ex3αβERKO ovary. These genes included NPC intracellular cholesterol transporter 1 (Npc1) (+1.84), Impact, RWD domain protein (Impact) (+1.5), Desmoglein 2 (Dsg2) (+1.51), Molybdenum cofactor sulfurase (Mocos) (+1.15), Formin homology 2 containing domain 3 (Fhod3) (−1.29), Potassium channel tetramerization domain-containing 1 (Kctd1) (−1.25), and Beta-1–4-galactosyltransferase 6 (B4galt6) (+1.89). As these values were below the fold change of 2 that was used for the comparison with Cyp19a1 KO and Foxl2 cKO ovaries (Fig. 3), the published datasets from the Cyp19a1 KO and Foxl2 cKO were examined (21, 38). Dtna and Kctd1 and B4galt6 were differentially expressed in Foxl2 cKO ovaries (38) and B4galt6 was the only gene in this region reported in Cyp19a1 KO gene list (21) (Table 2). The candidate genes found in this region of interest on Chr18 that were differentially expressed in the Ex3αβERKO were also compared to genes from the RNA-seq dataset from Trim28 cKO ovaries (47), and 4 genes were also reported to be significantly different in Trim28 cKO compared to control ovaries, including Dsc2, Dtna, Kctd1, and B4galt6. Expression of Kctd1 was downregulated in the Ex3αβERKO and upregulated in the TRIM28 cKO (47), while the other genes showed similar direction of altered expression (Table 2) in the models of postnatal ovarian transdifferentiation examined.

Table 2.

Genes located on Ch18 5-15 M and differentially expressed in ovarian transdifferentiation

Gene symbol	Ex3αβERKO	Cyp19a1 KO (21)	FOXL2 cKO (38)	TRIM28 cKO (47)	Official name & function
B4galt6	1.89	1.55	3.6	2.87	Beta-1,4-galactosyltransferase 6. Involved in ganglioside biosynthetic processes. May be active in Golgi apparatus.
Dsc2	−4.61	NL	NL	−5.63	Desmocollin 2. Cadherin-like transmembrane glycoproteins that mediate cell-cell junctions.
Dtna	3.9	NL	4.4	3.16	Dystrobrevin alpha. Predicted to enable zinc ion binding and may be involved the formation and stability of synapses.
Kctd1	−1.25	NL	2.0	2.89	Potassium channel tetramerization domain-containing 1. Negatively regulates AP-2 transcription factors and the Wnt signaling pathway.

Data shown are fold change compared to control or wild-type ovary for each genotype as appropriate for the study. NL indicates not listed in the dataset. Gene name and descriptions were obtained from NCBI Gene Summary, GeneCards® and The Human Protein Atlas.

Discussion

ESR signaling is necessary for the maintenance of granulosa cell identity in the ovary throughout adulthood, as demonstrated in the Ex3αβERKO showing ovarian transdifferentiation of granulosa cells to Sertoli-like cells (Fig. 1). Herein, we examined the ovarian morphology and gene expression differences in 3 different ESR knockout models, the Ex3αERKO, Ex3βERKO, and the double Ex3αβERKO. We also demonstrate that differences in genetic background contribute to the variability in the transdifferentiation phenotype in the Ex3αβERKO ovary, and we identified a putative genomic region on Chr 18 5-15 M (Fig. 5) that could be involved in the loss of granulosa cell identity in the Ex3αβERKO compared to the “original” 3αβERKO (16). Early ovarian development appears normal in all prepubertal mice examined at postnatal day 12; however, as the animals aged, we observed this ovarian transdifferentiation in Ex3αβERKO mice aged 6 to 12 months that varied by size and number of tubule-like structures present in each ovary. We attempted to characterize the timing of the transdifferentiation by examining ovaries from mice aged 3 to 12 months; however, the variation observed between animals, the variability of influences in genetic strain, and requirement to section entire ovaries and stain for SOX9, we were unable to precisely determine time of transdifferentiation in this model system.

Hemosiderin laden macrophages are abundant in the Ex3αβERKO ovaries (Fig. 1J-1O, stars) and some of the Ex3αERKO and Ex3αβERKO have hemorrhagic cysts present while others had less severe cystic phenotypes. The presence of these cysts is presumably due to increased LH in the Ex3αERKO and the Ex3αβERKO (8, 10) as compared to the Ex3βERKO animals, which had LH concentrations similar to control mice (Supplementary Fig. S1 (55)). Previous work has shown that Ex3αERKO ovaries have hemorrhagic cysts (10) as do the LHβCTP mouse model with excess LH (56); however, when this LHβCTP mouse was crossed with an βERKO, these cysts did not form (62). The presence of these hemorrhagic cysts in the Ex3αβERKO suggests that a genetic component may contribute to this phenotype as well, since the loss of Esr2 was not able to protect from the formation of hemorrhagic cysts in some animals. Similar findings were observed with hemosiderin laden macrophage infiltration into the stroma. These macrophages emit autofluorescence in both green and red channels; therefore, immunofluorescent staining was not possible to perform co-localization studies. It is unclear if the increased immune cells are due to aging of the ovaries, loss of paracrine signaling, or related to the hemorrhagic cyst variability. The morphological differences observed in the ESR knockout mice are most evident in the Ex3αβERKO, with the appearance of dismorphogenic regions that have lost the expression of FOXL2 and begin to express SOX9, demonstrating transdifferentiation into Sertoli-like cells in the ovary (Fig. 1, green arrows).

Estrogen Signaling is Necessary for Maintenance of Granulosa Cell Differentiation

Complete loss of ESR signaling leads to postnatal ovarian transdifferentiation of granulosa cells to Sertoli-like cells in the Ex3αβERKO (Fig. 1), similar to that reported in Cyp19a1 KO mice on soy-free diets (19-22). We first examined the expression of several genes implicated in gonadal sex determination and formation of either the ovary or testis from the bipotential gonad (Fig. 6A–6B). We observed a decrease of granulosa cell–specific Foxl2 in the Ex3αβERKO, while Wnt4 showed no difference in expression, and Rspo1 was increased in both Ex3βERKO and Ex3αβERKO ovaries. Sertoli cell–specific genes Sox9 and Fgf9 were increased in Ex3αERKO and Ex3αβERKO ovaries, presumably due to the excess LH and testosterone in absence of Esr1 (10). The Ex3αβERKO ovary had increased expression of Dmrt1, a hallmark of Sertoli cell differentiation (48, 63).

Figure 6.

Model of postnatal ovarian transdifferentiation. (A) Granulosa cells in the ovary express FOXL2 which actively represses testis-specific genes while TRIM28 regulates Foxl2 and Cyp19a1 and potential sumoylation of FOXL and ESR2 in these cells. (B) In the absence of ESR signaling (loss of both receptors or any ligand), the loss of TRIM28 or FOXL2 granulosa cells transdifferentiate into intermediate cells distinct from other cell types and that of cells in the bipotential gonad during embryo sex determination. (C) Transition from this intermediate cell type into SOX9 and DMRT1 expressing cells that resemble Sertoli cells in the testis could be regulated by Ttyh1, Kctd1, Cidea, or B4galt6 genes that were identified in the 4 models of ovarian transdifferentiation.

To identify possible mechanisms that contribute to loss of granulosa cell differentiation in the ESR ovaries we performed whole ovary microarray analysis. The expression of genes in the Ex3αERKO or the Ex3βERKO ovary is similar to control ovaries than the Ex3αβERKO (Fig. 3). These differences suggest that a compensatory mechanism may be present in each single ERKO ovary that provides enough ESR signaling to prevent loss of granulosa cell differentiation and subsequent ovarian transdifferentiation. The Ex3βERKO granulosa cells that have increased expression of Esr1 compared to granulosa cells from wild-type follicles (15). The Cyp19a1 KO mouse also only showed ovarian transdifferentiation on a soy-free diet, suggesting that the genistein in the soy was sufficient to maintain granulosa cell differentiation, further supporting a compensation for maintenance of granulosa cell differentiation in the single ERKOs.

The microarray analysis was made more complex by the presence of heterozygous ESR alleles in the control animals. To be considered a control, the animal needed to have one copy of each ESR allele. These controls were used due to the infertile phenotype of each single-ESR female (10, 15), and the large number of breeders necessary to obtain Ex3αβERKO animals. The single Ex3αERKO and Ex3βERKO animals were wild-type for the other ESR gene, eliminating the potential for heterozygous interference in the analysis. The compensatory relationship of the 2 ESRs may also complicate the genomic results, as there were a large number of genes that were differentially expressed in the Ex3αβERKO compared to each single ERKO. Furthermore, the number of genes differentially expressed in the Ex3αβERKO mice were much larger than that reported for the Cyp19a1 KO mouse (21), suggesting that loss of the ESRs themselves causes a larger change in the transcriptome than the loss of their hormone ligand. In similarity, ESR1 actively represses a number of genes in the uterus (64), which may also be true in the ovary.

Differentially Expressed Genes in Models of Postnatal Ovarian Transdifferentiation

The genes identified in the microarray analysis and on Chr18 as differentially expressed in Ex3αβERKO ovaries were compared with genes reported to be differentially expressed in other models of ovarian postnatal transdifferentiation, including the Cyp19a1 KO (21), Foxl2 cKO (38), and Trim28 cKO (47). From these comparisons, 22 genes were differentially regulated in all 4 models (Tables 1 and 2). There were several genes from this list that were also upregulated in the Ex3αERKO ovary (Fig. 4), including Cldn11, Cst9, and Dtna, which are expressed during embryonic testis differentiation (57, 65-67). The expression of testis-specific Sox9, Cldn11, Cst9, and Dtna in the Ex3αERKO could be due to increased androgens (10); however, increased androgens alone are not sufficient to induce transdifferentiation as this has not been reported in models of excess androgens, such as polycystic ovarian syndrome (68-72), without loss of other genes necessary for the maintenance of the granulosa cells, such as both ESRs, Cyp19a1 (19, 21, 22, 73), Foxl2 (38), or Trim28 (47) (Fig. 6).

TRIM28 is a scaffold protein expressed in many tissues (74, 75) that modulates progesterone receptor and Esr1 receptor signaling in the uterus (76). Overexpression of TRIM28 in mouse models of premature ovarian insufficiency was shown to increase expression levels of Foxl2 and Cyp19a1 genes, and to restore hormone levels and ovarian function (77). TRIM28 binds directly to SOX9 in Sertoli cells of the fetal testis (78), while in the ovary TRIM28 is essential for the maintenance of granulosa cell differentiation (47). Five genes listed in Table 1 were shown to be bound by TRIM28 in the ovary, including Defb36, Sox9, B4galnt1, Timp2, and B4galt6 (also present on Chr18, Table 2). Of the genes differentially regulated in all 4 models of ovarian transdifferentiation, 10 were found to be bound by both TRIM28 and FOXL2 including Cldn11, Ttyh1, Itgbl1, Gstm6, Tbc1d9, Map3k21, Cp, Sort1, Sgk3, and Sdc1. Three genes located on Chr18 were also differentially regulated in 2 to 3 models of ovarian transdifferentiation (Table 2) and reported to be bound by both TRIM28 and FOXL2 including Dsc2, Dtna, and Kctd1. Chromatin near Esr2 was also reported to be bound by both TRIM28 and FOXL2 (47). TRIM28 also functions as a SUMO-E3-ligase, and sumoylation impacts a number of transcription factors including FOXL2 (79), ESR1 (80), and ESR2 (81), which could be one mechanism of regulation of increased expression of granulosa cell–specific gene expression.

The Trim28 cKO granulosa cells lost expression of FOXL2, transdifferentiated into an intermediate cell type distinct from cells present in the bipotential gonad prior to embryonic gonadal differentiation (47). Loss of estrogen signaling (via loss of both receptors or in absence of the ligand), loss of FOXL2 and/or reduction of TRIM28 expression or sumoylation activity appears to prevent the active repression of genes in the testis-specific pathway, causing the cells to become an intermediate cell type. This is modeled in Fig. 6, where the granulosa cells transition into an intermediate cell type before obtaining Sertoli-like cell markers including SOX9 and DMRT1. In the Ex3αβERKO, FOXL2 expression was lost in cells in the dismorphogenic regions which was followed by expression of SOX9 and structures that appear similar to seminiferous tubules in the testis lacking sperm (Fig. 1J-1L). The plasticity of granulosa cells in the ovary is similar to findings observed in Sertoli cells in the testis, where loss of Dmrt1 (63), Sox8, and Sox9 (40, 82) contribute to postnatal transdifferentiation, demonstrating that continued repression is necessary for maintenance of both the ovary and the testis even after embryonic gonadal development (83).

Additional Possible Contributing Mechanisms to Ovarian Transdifferentiation

In addition to changes in gene expression, genetic background is important for sex determination, and different strains of mice have dosage differences in expression of several sex determination markers (84). Chromatin-specific regions have been identified that correlate with either Sertoli cell or granulosa cell–specific expression during embryonic sex determination (85, 86) and these regions were shown to be altered when Dmrt1 was lost in the testis, where they resembled chromatin typically associated with granulosa cells (87). Loss of Cbx2, a member of the Polycomb repressive complex that methylates H3K27, is required during embryonic sex determination where it represses ovarian-specific genes that oppose testis development (86). While these genomic studies looked at embryonic sex determination, similar loss of active repression of the opposing gonadal pathway is similar to what is observed in the models of postnatal transdifferentiation. The potassium channel tetramerization domain-containing 1 (Kctd1) gene is a transcriptional repressor of AP-2 transcription factors in vitro and has slight expression in the human ovary (88). The Kctd1 gene is located on Chr18 and its chromatin region is bound by both TRIM28 and FOXL2 in the ovary (47). Kctd1 was upregulated in the Trim28 cKO ovary and slightly downregulated in the Ex3αβERKO but not listed in the Cyp19a1 cKO or Foxl2 cKO datasets (Table 2). The difference in gene expression in these models of ovarian transdifferentiation suggests that further investigation is necessary into the role that Kctd1 may play in ovarian maintenance of granulosa cells through interactions with TRIM28, FOXL2, and/or the ESRs.

Cellular Signaling and Protein Trafficking

Cellular signaling mechanisms may also impact ovarian transdifferentiation, and increased expression of several genes implicated in signaling pathways were found to be differentially expressed in all 4 models of ovarian transdifferentiation and chromatin near the genes that were bound by both FOXL2 and TRIM28 (47). These include upregulated genes Ttyh1, Map3k21, and Tbc1d9, while Sdc1 was downregulated (Table 1). Ttyh1 is a predicted integral membrane protein that regulates neuronal cells through interactions with the Notch signaling pathway (59, 60). Ttyh1 is significantly upregulated in the Ex3αβERKO ovary and other models of postnatal ovarian transdifferentiation, and its expression increases during the time of embryonic sex determination in the pre-Sertoli cells (67). Interestingly, crosstalk between the Notch and Hippo pathways is important for embryonic development. Ttyh1 requires Notch signaling (59), and Notch has been implicated to interact with Hippo targets yes-associated protein 1 (YAP1) to regulate gene expression. YAP1 is targeted by the large tumor suppressors 1 and 2 (Lats1/Lats2) to regulate gene expression (89). LATS1/LATS2 are expressed in the ovary and knockout of both genes caused the appearance of Sertoli-like structures in the ovary of these mice. Furthermore, cultured granulosa cells from the Lats1/Lats2^(flox/flox) mice treated with Cyp19-Cre recombinase lentivirus differentiated into several cell lineages, including osteoblasts in vitro (90), demonstrating the plasticity of these cells. Integral membrane protein Sdc1 is expressed in adult ovary and testis (91) and was downregulated in models of ovarian transdifferentiation. Conversely, the GTPase activating protein Tbc1d9, expressed in adult testis (91), and the mitogen activated kinase Map3k21, expressed in developing gonads at the time of embryonic sex determination (92, 93), are both upregulated in the models of ovarian transdifferentiation. The Hippo signaling pathway is predicted to interact with several different signaling pathways (89), and the differentially regulated signal molecules may crosstalk through common signaling mechanisms to contribute to the plasticity of granulosa cells in adult ovaries after loss of key genes in granulosa cells.

Apoptosis and Oocyte Loss

The Ex3αβERKO has increased expression of Cidea, a gene implicated in regulation of apoptosis and thermogenesis was increased in all 4 models of ovarian transdifferentiation (Table 1). Cidea is also implicated in thermogenesis and the transition of white adipose tissue into beige or bright adipose tissue (94, 95), including in some mouse strains after exposure to excess androgens (68). The expression of Cidea was confirmed to be specific in the Ex3αβERKO ovary (Fig. 4C), suggesting that it could be regulating apoptosis or contributing to the cellular transition of granulosa cells. Another apoptotic signaling molecule, Sgk3 was only found to be upregulated in the Ex3αERKO ovary (Fig. 4D). Oocyte loss has preceded development of Sertoli-like structures in the rat ovary (96), while in the mouse, the loss of germ cells did not block fetal ovarian development or induce expression of SOX9 (97). The transdifferentiated regions in the Ex3αβERKO did not have oocytes present, while adjacent healthy follicles still contained oocytes (Fig. 1). In the rat irradiated ovary, FOXL2 expression is reduced (96) but not lost as seen in the Ex3αβERKO prior to increased expression of SOX9 in the dismorphogenic regions. The Trim28 cKO ovary showed regions that did not have oocytes and lacked FOXL2 expression before SOX9 expression (47). While the loss of the oocyte may be necessary for the development of Sertoli-like regions, preliminary work in our laboratory suggests that loss of the oocyte is not sufficient as irradiation of Ex3αβERKO mice at PND3 did not develop dismorphogenic regions (data not shown). The timing of the loss of oocytes has also been suggested to explain discrepancies in the literature and in different animal models (45, 98), although the loss of oocytes may not be sufficient unless coupled with loss of other regulatory mechanisms that maintain granulosa cell identity throughout the postnatal period.

Conclusion

Herein, we demonstrate altered ovarian gene expression in both single ERKO ovaries, and highlight common genes differentially expressed in several models of postnatal ovarian transdifferentiation. We also highlight a putative genomic region on Chr18 that might contribute to the loss of granulosa cell maintenance and inhibition of testis-specific gene expression in the Ex3αβERKO ovary. Several genes identified as possibly contributing to the differentiation from the intermediate cell type to the Sertoli-like cell observed include cellular signaling molecule Ttyh1, an activator of apoptosis and regulator of thermogenesis Cidea, and the transcriptional repressor Kctd1. Furthermore, B4galt6, a gene that has a potential role in protein sorting in the Golgi apparatus is located on the Chr18 region identified and upregulated in all 4 models of ovarian transdifferentiation. The postnatal ovarian transdifferentiation observed in our study occurs due to changes in a number of cellular processes, seemingly controlled by ESR signaling. While we may not be able to define the precise mechanism, we demonstrate that ESR signaling is necessary for granulosa cell maintenance and that transdifferentiation occurs in ovarian regions adjacent to normal-appearing follicles with oocytes present.

Abbreviations

ANOVA

analysis of variance

cKO

conditional knockout

ERKO

estrogen receptor knockout

ERα

estrogen receptor 1 (ESR1)

ERβ

estrogen receptor 2 (ESR2)

ESR

estrogen receptor

IPA

Ingenuity Pathway Analysis

KO

knockout

LH

luteinizing hormone

MDA

Mouse Diversity Array

RT-PCR

reverse-transcriptase polymerase chain reaction

SNP

single nucleotide polymorphism

SRY

sex-determining region of the Y chromosome

Funding

This research was supported (in part) by the Intramural Research Program of the NIH. Funding was provided by the Division of Intramural Research at the National Institute of Environmental Health Sciences to K.S.K. #1ZIAES0700065.

Disclosures

The authors have nothing to disclose.

Data Availability

Some of all of the datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.