Ectopic expression of CGG-repeats alters ovarian response to gonadotropins and leads to infertility in a murine FMR1 premutation model

Abstract

Women heterozygous for an expansion of CGG repeats in the 5’UTR of FMR1 risk developing fragile X-associated primary ovarian insufficiency (FXPOI) and/or tremor and ataxia syndrome (FXTAS). We show that expanded CGGs, independent of FMR1, are sufficient to drive ovarian insufficiency and that expression of CGG-containing mRNAs alone or in conjunction with a polyglycine-containing peptide translated from these RNAs contribute to dysfunction. Heterozygous females from two mouse lines expressing either CGG RNA-only (RNA-only) or CGG RNA and the polyglycine product FMRpolyG (FMRpolyG+RNA) were used to assess ovarian function in aging animals. The expression of FMRpolyG+RNA led to early cessation of breeding, ovulation and transcriptomic changes affecting cholesterol and steroid hormone biosynthesis. Females expressing CGG RNA-only did not exhibit decreased progeny during natural breeding, but their ovarian transcriptomes were enriched for alterations in cholesterol and lipid biosynthesis. The enrichment of CGG RNA-only ovaries for differentially expressed genes related to cholesterol processing provided a link to the ovarian cysts observed in both CGG-expressing lines. Early changes in transcriptome profiles led us to measure ovarian function in prepubertal females that revealed deficiencies in ovulatory responses to gonadotropins. These include impairments in cumulus expansion and resumption of oocyte meiosis, as well as reduced ovulated oocyte number. Cumulatively, we demonstrated the sufficiency of ectopically expressed CGG repeats to lead to ovarian insufficiency and that co-expression of CGG-RNA and FMRpolyG lead to premature cessation of breeding. However, the expression of CGG RNA-alone was sufficient to lead to ovarian dysfunction by impairing responses to hormonal stimulation.

Introduction

Spontaneous primary ovarian insufficiency (POI) affects ~1% of chromosomally normal women before the age of 40 (1). The most commonly known genetic etiology identified is heterozygosity for a premutation length (55–200) CGG tract within the 5’ UTR of Fragile X Mental Retardation 1 (FMR1) (2). A clinical diagnosis of fragile X-associated primary ovarian insufficiency (FXPOI), the most severe manifestation of ovarian dysfunction, requires a cessation of menses for 4–6 months and repeated elevation of follicle stimulating hormone (FSH) levels prior to age 40 (2). While it is estimated that between 1:150–1:200 US women are premutation carriers, approximately 20% of carriers, most often ascertained through families with fragile X syndrome, meet the criteria of FXPOI (3–6). However, an additional ~20% of premutation females present with irregular or missing periods or early menopause and another ~13% report difficulty conceiving (7,8). Biochemical indices of ovarian dysfunction are also found, including low anti-Müllerian hormone (AMH), elevated FSH and poor response to hormonal stimulation (9–13). Notably, the average age at menopause for women carrying a premutation is 5 years younger than that of the general population (12). The highest incidence of POI is found in women with a CGG tract size of 70–100 (6,12,14,15). This is different from the neurodegenerative FXTAS disorder, which displays a linear correlation between repeat size and disease pathology in humans and model organisms (16–18).

On a molecular level, FXTAS and FXPOI share common features, including increased FMR1 mRNA in peripheral blood collected from premutation carriers (19,20). FMR1 mRNA is elevated in postmortem brain tissues of FXTAS males as well as in ovarian granulosa cells isolated from female carriers undergoing in vitro fertilization (13,21). Levels of the FMR1 protein product FMRP are reduced by ~60% in peripheral blood collected from FXTAS males with 140–190 CGG repeats (22). Mouse models of the Fmr1 premutation generated at the National Institutes of Health (NIH mouse) and in the Netherlands (Dutch mouse) exhibit variable decreases of FMRP, depending on CGG repeat length, in brain and ovarian tissues (16,23,24). Studies demonstrating degenerative phenotypes in Drosophila and mice from the expression of expanded CGG-containing RNAs rather than modulation of expression of protein-coding FMR1 provided the basis for the two models of pathogenesis (25–27). Importantly, both models have the potential to contribute to the premutation disorders. In the RNA-mediated toxicity model, expanded tracts of CGG-repeat RNA interact with RNA-binding proteins and sequester them, limiting the availability of these proteins for their normal cellular functions (28–32). The second model proposes that initiation of translation from a non-ATG codon (RAN translation) generates a toxic homopolymeric peptide from the expanded repeat RNA (27,33). A polyglycine peptide translated from the 5’ UTR of FMR1, termed FMRpolyG, has been shown to aggregate in postmortem brain tissues of FXTAS patients and in animal models of the disease (27,34,35). FMRpolyG-positive aggregates were also observed in the ovarian stroma of mice expressing expanded CGGs (36,24), suggesting that RAN translation is involved in FXPOI (37–39).

Due to the limited accessibility of human ovarian tissue for assessment, an understanding of the cellular events leading to POI depends on the use of animal models. Previous studies examined the effects of expanded CGG repeats engineered into the endogenous mouse Fmr1 promoter. This work demonstrated that the number of primordial follicles in aged premutation ovaries was similar to or increased compared to controls (23,24). However, the population of growing follicles exhibited increased atresia suggesting that dysfunctional follicular development may lead to POI rather than a depletion of the primordial follicle reserve (23,24,40–42). Follicular cysts, fluid-filled sacs commonly derived from dysfunctional follicles with degenerated oocytes, were also observed in CGG-expressing ovaries (23). Follicular cysts result from several sources but are frequently associated with hormonal imbalance, androgen excess and the postmenopausal ovary where estrogen production is markedly decreased (37–39). This histological feature could be an indicator of premature aging in CGG-expressing mice or a signal that follicular maturation is perturbed.

A particular limitation of existing models for study of ovarian phenotypes of the premutation is that they alter expression of coding Fmr1, and this impaired the ability to distinguish between contributions of CGG-containing mRNAs and changes in FMRP (23,24). The ectopic expression of CGG repeats in neuronal cells and tissues has demonstrated that the repeats are sufficient to drive cell death and hallmarks of FXTAS (26,27,34), but this sufficiency is not established in a mouse model in the context of ovarian insufficiency. Here we demonstrate the sufficiency of ectopically expressed CGG repeats to lead to ovarian dysfunction and infertility by utilizing two mouse models with the CGG-containing portion of the 5’ UTR of FMR1 expressed from the Rosa26 locus, one capable of expressing FMRpolyG and CGG RNA and the other able to express only the CGG RNA (34). We show clear defects in response to gonadotropin stimulation in both models, but only expression of FMRpolyG and CGG RNA leads to a progressive loss of fertility with age. This loss of fertility is driven by a loss of ovulation and appears to be connected to alterations in steroidogenesis.

Results

Generation and characterization of mice globally expressing CGG repeats

CGG-containing alleles targeted to Rosa26, with a 551 bp portion of the 5’UTR of FMR1 (297 bp of CGG-tract and 127 bp of both 5-and 3-flanking sequence) plus an eGFP coding sequence tag were previously generated (34). We have termed the ‘full 5 UTR FMR1 or FMRpolyG mouse’ engineered by Nicolas Charlet-Berguerand, the FMRpolyG+RNA line, as it is capable of producing expanded CGG RNA and contains the near-cognate ACG codon used to produce a polyglycine peptide via RAN translation. The second line, identified as the 5 UTR FMR1, we referred to as the CGG RNA-only line because it produces the CGG₉₉ RNA but lacks the ACG codon and produces no detectable polyglycine (Fig. 1A). To develop mice with global expression, we crossed FMRpolyG+RNA^Fl/Fl and RNA-only^Fl/Fl animals to a Gdf9-cre line⁴⁶. Females with oocyte-specific expression of CGG repeats were then bred to control males (no Gdf9-cre present). Due to recombination in the preconception oocytes, progeny resulting from a CGG allele passed from the mother express the recombined allele ubiquitously (Fig. 1B). Previous studies found intranuclear aggregates containing FMRpolyG within multiple brain regions and outside of the central nervous system (24,27,34,35), and we confirmed this observation in the hypothalamus of FMRpolyG+RNA-expressing mice at 6 months of age (Supplementary Material, Fig. S1). We examined the other components of the hypothalamic–pituitary-ovarian axis (HPO) for FMRpolyG-positive inclusions. We noted that incidence of FMRpolyG-positive inclusions was markedly different between the pituitary lobes, with the highest and lowest burdens appearing in the intermediate and posterior lobes, respectively (Fig. 2A). This finding was consistent with Ubiquitin staining previously reported in the anterior pituitary lobes of autopsy specimens from two men who died with FXTAS as well as in the pituitary glands of premutation mice (17,43). All regions of the pituitary in FMRpolyG+RNA^rec/+ females exhibited some FMRpolyG-containing aggregates but not RNA-only^rec/+ or control tissues (Fig. 2B). We found the number of FMRpolyG-positive inclusions in the anterior pituitary increased from a median of 4 to 16 inclusions per field from 3 to 6 months of age (Fig. 2C).

Figure 1 .

Mice generated by Sellier et al. (34) with global expression of ectopic CGG-repeats. (A) A 551 bp fragment of human FMR1 5’UTR containing 99 CGG repeats is expressed with an eGFP under the Rosa26 promoter. (i) The FMRpolyG+RNA allele contains a near-cognate ACG codon used to produce a polyglycine peptide when expanded CGGs are downstream. (ii) The RNA-only allele has a 104 bp deletion surrounding the ACG codon and allows for the production of CGG-containing RNA without FMRpolyG. (B) Females with global CGG expression are produced by breeding a female heterozygous for a conditional CGG allele and Gdf9-cre (CGG^Fl/+;Gdf9+) to a male heterozygous for a conditional CGG allele (CGG^Fl/+).

Figure 2 .

FMRpolyG-positive inclusions accumulate in the anterior lobe of the pituitary. (A) FMRpolyG staining is different across the lobes of the pituitary with N-terminal antibody against FMRpolyG, scale bar = 50 □m. (B) FMRpolyG was only detectable in the pituitary lobes of FMRpolyG+RNA brains at 6 months. (C) Inclusion number in the anterior pituitary increased between 3 and 6 months [counted per field of 150□m²], scale bars = 20 □m and error bars represent ±SD. A = anterior, I = intermediate, P = posterior.

We also assessed ovarian tissue directly to establish whether FMRpolyG was detectable in the ovaries of FMRpolyG+RNA^rec/+ females. FMRpolyG was reported in ovarian stroma but not the follicular cells of aged mouse and human premutation ovaries (24). Our study included staining of prepubertal ovaries, which had not been analyzed for the presence of FMRpolyG previously. We identified FMRpolyG+ aggregates in follicular cell types, including in the oocytes, at postnatal day 15 (PND 15) but not in RNA-only^rec/+ or control ovaries (Fig. 3A). Inclusions within oocytes were not exclusively intranuclear, in contrast to those reported in neuronal tissues. We saw FMRpolyG puncta in either the cytoplasm or nucleus of oocytes, but only within the nucleus of somatic cells. These intranuclear inclusions found within aged FMRpolyG+RNA^rec/+ ovaries were predominantly in stromal and interstitial cells (Fig. 3B) but could be observed in rare granulosa cells. There was no apparent increase in FMRpolyG-containing aggregates with age and no FMRpolyG-containing aggregates were detected in oocytes of aging animals when analyzed at 3 or 6 months of age. It is unclear why FMRpolyG is only observed in the germ cells of prepubertal ovaries.

Figure 3 .

Polyglycine-positive inclusion bodies were identified in oocytes of FMRpolyG+RNA^rec/+ females at postnatal day15 (A) by immunofluorescence, scale bars = 20 □m. (B) At 6 months, FMRpolyG-aggregates were largely restricted to the ovarian stroma, as shown by immunohistochemistry (scale bar = 20 □m). Nuclear localization is shown in the higher magnification inset (scale bar = 10 □m). No inclusions were seen in oocytes in sexually mature females. O = oocyte, St = stroma, Ly = lymphatic vessel, OSE = ovarian surface epithelium.

Ectopic expression of FMRpolyG+RNA is sufficient to lead to premature reproductive senescence

Previously published data on the neuronal effects of ectopic expanded CGG repeat expression show increased phenotypic severity with age (17,27,34,44). To determine if fertility is similarly affected, females were continuously bred with control males and litter characteristics tracked. FMRpolyG+RNA^rec/+ females exhibit decreased cumulative progeny through 8 months of age (25.6 ± SD 7.83), while females expressing CGG RNA-only^rec/+ show comparable numbers (45.6 ± 13.49) of pups to control females (53.3 ± 10.39) (Fig. 4A), distinguishing the effects of CGG repeat allele products. This decreased fecundity in FMRpolyG+RNA^rec/+ females was due to a cessation of breeding after an average of 3 ± 0.71 litters compared to the 7.17 ± 0.98 or 6.5 ± 0.55 litters per female produced by RNA-only^rec/+ and control females, respectively (Fig. 4B). The number of pups produced per litter was unchanged across genotypes (Fig. 4C). A confounder of the observed infertility was the significant weight gain in FMRpolyG+RNA^rec/+ females compared to females of other genotypes (Fig. 4D). This increase became apparent at 3 months of age and continued through the end of life. This finding corroborates the previous study of mice ubiquitously expressing this Rosa26 allele (34).

Figure 4 .

Ectopic expression of expanded CGGs leads to infertility in aging females. This was apparent from (A) the reduction of cumulative pups born via continuous breeding to FMRpolyG+RNA females and (B) decreased numbers of litters born to FMRpolyG-expressing females. (C) All groups birthed similar numbers of pups per litter. (D) FMRpolyG+RNA but not CGG RNA-only females gained significantly more weight by 3 months of age. Error bars indicate ±SD ^* < 0.05, ^*** < 0.0005, ^**** < 0.00005.

Expression of both CGG RNA and FMRpolyG leads to progressive loss of ovulation and altered steroidogenesis in the ovaries of aging females

To determine the drivers of reproductive senescence in the FMRpolyG+RNA^rec/+ females, we examined ovarian histology at 3 and 6 months. We found that CGG-expressing females exhibited grossly normal histology at 3 months of age but by 6 months of age corpora lutea were completely absent specifically from FMRpolyG+RNA^rec/+ ovaries (Fig. 5). This indicated that loss of fertility resulted from a lack of ovulation and was supported by luteinized and hyperplastic stroma apparent in FMRpolyG+RNA^rec/+ ovaries (Fig. 5B and D). We counted ovarian follicle populations and determined that there was no significant reduction in any individual follicle class at 6 months of age, however, the total number of follicles in FMRpolyG+RNA^rec/+ ovaries was decreased (Fig. 5C). This was consistent with follicle counts from previous studies of aging premutation ovaries, as was our observation of increased atresia within FMRpolyG+ RNA^rec/+ ovaries (23,24). Atresia in the RNA-only^rec/+ ovaries displayed an increase of roughly half the number of dying follicles observed in FMRpolyG+RNA^rec/+ ovaries, it was not statistically significant.

Figure 5 .

(A) Ovaries with expanded CGGs exhibit grossly normal morphology after the onset of puberty, but the FMRpolyG+RNA^rec/+ allele leads to (B) a progressive loss of corpora lutea (CL) in females by 6 months of age. (C) Follicle classes counted in ovaries at 6 months showing increased follicle death and reduction of total follicle numbers. (D) CLs are absent in FMRpolyG+RNA^rec/+ ovaries at 6 months but (E) both CGG-expressing genotypes display ovarian cysts, N = 4/genotype. (F) mRNA expression of the steroidogenesis-related StAR and Vcam1 genes were altered in FMRpolyG+RNA^rec/+ ovaries at 6 months, N = 3/genotype. CLs marked by asterisks, and error bars represent SD. (C-F) ^* < 0.05, ^** < 0.005.

Ovarian cysts were present at increased frequency in both FMRpolyG+RNA^rec/+ and RNA-only^rec/+ ovaries (Fig. 5E). Subsets of the cysts observed in ovaries from both CGG-expressing genotypes were lined with cells suspected to be epithelia due to the presence of some ciliated cells (data not shown), and positive staining for cytokeratin 8 and Pax8 (Supplementary Material, Fig. S2, Supplementary Material, Table S3). These inclusion cysts were not localized to a particular region of the ovary and were noted both adjacent to hilus and opposite the hilus and oviduct. The increased incidence of ovarian cysts has been associated with ovarian aging and perturbed steroidogenesis, which may alter follicle development. We assessed steroidogenesis, the conversion of cholesterol and androgenic precursors into estrogens, by measuring transcript levels of the rate-limiting enzymes. When whole ovary lysates were analyzed it was observed that FMRpolyG+RNA^rec/+ ovaries had decreased expression of steroidogenic acute regulatory protein (StAR) mRNA (Fig. 5F and B). The androgen-responsive gene Vcam1, however, was increased, suggesting that there may be high local androgen concentrations in FMRpolyG+RNA^rec/+ ovaries. These patterns of expression remained similar when assayed in 8-month-old ovaries, but at this later time point ovaries expressing CGG RNA-only^rec/+ revealed a significant increase in StAR transcript (Fig. 5F).

To corroborate the noted histological and molecular findings with clinical markers reported in FXPOI patient cohorts (8–10,15), circulating hormones were assayed from two different cohorts of females at 3- and 6 months of age. Pituitary gonadotropins LH and FSH, as well as granulosa-produced inhibins A and B were measured in each female where serum volume allowed. There were no significant differences in FSH or LH levels across genotypes at 3 months (Fig. 6A). This was also true for pituitary hormones measured at 6 months (Fig. 6B). This result was surprising given the robust histological changes seen in FMRpolyG+RNA^rec/+ ovaries and overt reproductive senescence. It should be noted that there is variability seen, particularly within LH and FSH levels that may result from using a window of late estrus to early diestrus within the cycle. Decreased inhibin A in circulating serum within the same cohort of females illustrated that ovary-derived signals were perturbed (Fig. 6B). This reduction was discordant with cohort of FSH levels. Within individual females, there was no clear inverse relationship with FSH and inhibin levels (individuals not specified in figures for clarity). Given that in a functional hypothalamic–pituitary-ovarian axis the reduction of inhibin would be concomitant with increased FSH (45,46), we take these data as an indication that the axis was disrupted.

Figure 6 .

To support the histological features of premature reproductive senescence in FMRpolyG+RNA^rec/+ ovaries circulating hormones were measured. (A) Serum collected at 3 months of age displayed no difference in luteinizing hormone (ng/mL) or follicle stimulating hormone (ng/mL) across genotypes, N = 3–7/genotype. (B) At 6 months FSH and LH remained comparable across genotypes, but ovary-derived Inhibin A was significantly reduced. Inhibin B displayed a strong trend toward reduction in FMRpolyG+RNA^rec/+ females, N = 4–8/genotype. Error bars denote SD ^* < 0.05.

Expression profiling of aging CGG-expressing ovaries reveals transcriptional changes in cholesterol, steroid, and lipid-associated GO terms

To expand our understanding of changes in aging ovaries, we employed RNA-seq to profile ovarian transcriptomes. We compared the effects of repeat expression with age by analyzing ovaries at postnatal day 21, 3 months and 6 months. There was little overlap in differentially expressed (DE) transcripts between CGG-expressing genotypes or within individual genotypes across ages (Table 1). Gene set enrichment analysis (GSEA) was used to determine if there may be pathways perturbed as a consequence of CGG expression. DE genes identified in ovaries from sexually mature females (3- and 6 months) were enriched for biological processes involving cholesterol and lipid biosynthesis and metabolism, and included Hmcgr, Pmvk, Mvd, Sqle, Idi1 and Fdps, genes central to cholesterol formation (Fig. 7A, Supplementary Material, Fig. S3). These enriched clusters appear to be linked hormone metabolic processes and the regulation of hormone levels. These genes include the previously assessed StAR, Cyp11a1 and Vcam1, as well as Cyp19a1, Insl3, Inhba, Inhbb, Fshr and Lhcgr (Fig. 7A). Perturbations in lipid and cholesterol synthesis in FMRpolyG+RNA^rec/+ were exacerbated by increased body weight, but the enrichment of these GO terms in the ovaries of RNA-only^rec/+ females supported the connection of pathways to CGG-mediated pathology even in the absence of obesity. Comparison of genes from these enriched clusters across ages revealed dynamic changes expression between time points (Fig. 7A). The changes in expression correlated with the histology of the ovaries at each time point as well. Alterations in the expression of Fshr and Lhcgr in FMRpolyG+RNA^rec/+ ovaries also provide insight into the lack of ovulation in these females at 6 months of age. While primordial follicles are not significantly reduced, dysfunctional or ablated response to FSH or LH in growing follicle populations could lead to the loss of ovulatory events. In addition to the most highly enriched clusters, there were other categories linked to recorded phenotypes. One interesting cluster was associated extracellular matrix (ECM) formation and organization (Supplementary Material, Fig. S3). CGG-expressing ovaries were enriched for alterations in ECM-associated genes at both 3- and 6 months of age. The ECM plays a role in follicular architecture during growth but is also implicated in steroidogenic processes. This cluster was of particular interest to us since there were deficits in the expansion of ECM following superovulation (47).

Table 1.

RNA-sequencing of ovaries collected from prepubertal and aging females (A) demonstrated increasing transcriptional alterations with age. (B) RNA-only^rec/+ females showed very little overlap in differentially expressed genes between ages while (C) ovaries from FMRpolyG+RNA^rec/+ females exhibited more common alterations

A	Summary	# of genes with FDR <0.2
		RNA-only vs. control	FMR poly G + RNA vs. control
	Postnatal day 21	54	105
	3 month	26	517
	6 month	122	735
B	Overlaps for RNA-only vs. control	PND 21	3 month	6 month
	3 month	54	0	1
	6 month			122
C	Overlaps for FMR poly G + RNA vs. control	PND 21	3 month	6 month
	Postnatal day 21	105	6	20
	3 month		517	50
	6 month			735

Figure 7 .

GSEA pathways identified in aging ovaries and dynamic expression of cholesterol-related processes. (A) Differential expression for ~100 DE genes related to cholesterol, steroid, lipid processes were plotted to show dynamic expression in aging ovaries. Relatively few changes in these genes distinguish CGG-expressing ovaries from controls prior to the onset of puberty. Between 3 and 6 months of age the profiles of CGG-expressing ovaries had robust changes in gene expression. FMRpolyG+RNA^rec/+ females, in particular, exhibited dynamic expression of steroidogenesis-related genes with age. (B) Distributions of cell proportions were found to be different across genotypes in prepubertal ovaries (PND 21) and in ovaries collected at 3 months. This difference in cell proportions was no longer significant across genotypes by 6 months of age.

We observed that several of the genes highlighted by GSEA were associated with specific ovarian cell types. To determine whether CGG expression altered the representation of ovarian cell types, we performed deconvolution analysis. Four populations were delineated by cell type-specific marker genes (Supplementary Material, Table S2), and reference-free deconvolution was used to estimate ovarian cell proportions at each age (Fig. 7B). Surprisingly, we detected significantly different proportions within the different genotypes at postnatal days 21 and 3 months of age. This early indicator of changes in ovarian composition suggested that ovarian function may also be altered at this age. When just the cell proportions of the CGG RNA-only or FMRpolyG+RNA groups were compared to one another we did not observe significant differences at any age (P = 0.356, 0.0527, and 0.680 for ages PND21, 3 months and 6 months, respectively).

Ovarian response to gonadotropin stimulation is perturbed prior to sexual maturity

To interrogate intrinsic ovarian functions that may be perturbed in young CGG-expressing females, we stimulated them with defined doses of gonadotropins (48). We assessed ovarian function by quantification of ovulated oocytes and determined that expression of either CGG RNA-only or FMRpolyG+RNA was sufficient to reduce the number of oocytes collected (Fig. 8A). COCs collected from CGG-expressing females were also smaller in diameter than controls (Fig. 8B). To determine if the process of ovulation was perturbed, we collected ovaries 8 h after the induction of ovulation. We found no statistical difference in the number of periovulatory follicles counted (Fig. 8D) but did observe a robust reduction in the expansion of the cumulus granulosa in periovulatory follicles (Fig. 8C and E). The oocytes contained within these follicles also exhibited a reduction in meiotic resumption (Fig. 8F). Deficiencies in cumulus expansion have been contributed to decreases of cumulus expansion-enhancing factor mRNAs in previous studies (49–51). However, when we assayed whole ovaries at mid-ovulation, we did not observe robust changes in expression across these factors (Fig. 8G). In order to determine whether these ovulatory processes were deficient, rather than delayed, we collected ovaries 11.5 h post-hCG injection and observed persistent unexpanded cumulus in CGG-expressing ovaries (Supplementary Material, Fig. S4).

Figure 8 .

Expanded CGG expression impaired response to hormone stimulation. (A) Decreased numbers of COCs were ovulated after superovulation and (B) COCs had smaller diameters than those of controls, N = 28–32/genotype. (C) CGG-expressing ovaries examined mid-ovulation displayed dense cumulus layers surrounding the oocyte while control follicles had undergone expansion of the cumulus into a dispersed matrix, scale bars = 100 □m. (D) The number of periovulatory follicles within the ovaries of stimulated females revealed no statistical difference but (E) scoring of cumulus expansion on a scale from 1 to 4, completely unexpanded to fully expanded, demonstrated that CGG-expressing follicles were predominantly scored as grade 1, N = 5/genotype. (F) This correlated with a lack or meiotic resumption in oocytes, but (G) transcriptional reductions of canonical cumulus-enhancing factors did not fully explain this phenotype. Error bars on all graphs represent SD.

Discussion

In this study, we sought to understand whether ovarian pathology in FMR1 premutation carriers stemmed directly from expression of expanded CGG repeats hypothesizing that expression of expanded CGG repeats independent of Fmr1 coding sequence in mice would be sufficient to confer ovarian dysfunction. Based on previous studies of premutation models with expanded CGGs introduced into the mouse Fmr1 locus that showed comparable or increased numbers of primordial follicles in the ovaries of aging female mice, we suspected that altered reproductive function resulted from perturbed follicular development. (23,24,41). Here we provide evidence that expanded CGGs expressed independently from Fmr1 can drive ovarian pathologies, ectopically expressed CGG repeats from the autosomal Rosa26 locus are sufficient to confer reproductive senescence in aging females. We determined that advanced reproductive senescence in females with aging required expression of both CGG-containing transcript as well as its translated peptide product. We also observed aggregation of the FMRpolyG peptide within the ovaries of FMRpolyG+RNA^rec/+ females. Inclusions were present in the cytoplasm (Fig. 3) and nucleus of oocytes in prepubertal ovaries but were not observed in aging ovaries. The cyclical proliferation and attrition of follicles in each estrous cycle could underlie this finding. Prepubertal ovaries have initially recruited follicles, but these are not yet dependent on gonadotropin signals. Once females are sexually mature, follicles are recruited each cycle and must respond to FSH and LH cues within a specific time window for successful maturation and ovulation. Cellular stress induced by CGG repeat-RNA and FMRpolyG expression may lead to atresia in these follicles before aggregates are visible via immunostaining. Future in vitro experiments on ovarian cell lines or isolated ovarian follicles with the use of a reporter for FMRpolyG (52) may also reveal differences in polyglycine peptide expression between ovarian and neuronal cells. Lower FMRpolyG expression could lead to fewer FMRpolyG inclusions in the ovary and follicles that do produce polyglycine peptides become rapidly atretic, leading to their absence in aging ovaries.

The breeding, histological and molecular phenotypes described here are robust, but limitations to these models were readily apparent. A primary concern with the FMRpolyG+RNA^rec/+ females was the increased weight observed with age. A previous report described the obesity phenotype in these animals and noted that Lepr expression in the hypothalamus was decreased at 6 months of age (34). Both obesity and reduced Leptin signaling are well-documented modulators of female fertility and hormonal changes, such as the suppression of FSH and LH levels. Human FMR1 premutation carriers with FXPOI do not show increased BMI compared to non-carrier females (53). We must therefore be cautious about ascribing all of the phenotypes observed in the FMRpolyG+RNA^rec/+ mice to ovarian dysfunction caused by repeat expression because it is likely that increased weight can drive or exacerbate these. In future studies, it should be possible to direct expression of the repeats in specific ovarian cell types to determine effects on follicle development while potentially mitigating weight increases. While not faithfully recreating expression patterns of the human condition, this approach serves to dissect the contributions of individual ovarian cell types to the identified phenotypes in an existing model. Lack of primary human tissue for assessment highlights the need to study molecular phenotypes in the ovarian tissue of mouse models. It is likely that leveraging data from mutants with cell-specific expression with that from global mutants will point to biomarkers or cellular functions that have a measurable human correlate. It also allows us to minimize the artifacts of this system, particularly the disrupted gene expression of Leptin signaling, which will influence fertility regardless of weight. To uncover human-specific features of the disorder, cell-based models, such as somatic ovarian cells derived from FXPOI patient iPSCs, will be critical to establish the effects of expanded CGG expression since these cells carry the particular genetic landscape of a human with the disease. Direct assessment of these human cells with standardized biochemical treatments may reveal if follicular response is changed in the presence of expanded CGG repeats. Despite the limitations, we can be confident that many of the observations made in the ovaries of aging females can be attributed to the expression of expanded CGG alleles. Importantly, the fact that mice globally expressing CGG RNA-only^rec/+ alleles exhibit altered gene expression and corroborating morphological changes in similar signaling pathways without any increase in weight led us to conclude that CGG-repeat expression does lead to dysfunction.

One consistent finding was the increased incidence of ovarian cysts. Both lines of CGG-expressing females (with or without the FMRpolyG RAN product) displayed cystic ovaries by 6 months of age, strongly suggesting that the formation of cysts is linked to CGG expression, likely through altered steroidogenic processes. The current interpretation of these findings is that they indicate premature ovarian aging rather than a polycystic ovary phenotype, but there are some similarities to androgen-treated ovaries in gene expression changes Hsd17b1, Wnt4, Vcam1 and Acta2 that regulate and respond to hormone level (54). We did not measure estradiol in females at 6 months so we cannot conclude that estrogen deficiency is a contributor, however the reduction in inhibin A production suggests that this is likely. The presence of follicular and inclusion cysts was also noted in the NIH premutation mouse model, further supporting a connection to expanded CGG expression (23).

Analysis of primary ovarian tissue obtained from FMR1 premutation carriers reported that four of five ovaries exhibited cortical inclusion cysts compared to 2 of 4 controls. The control ovaries displaying inclusion cysts were from either peri-or post-menopausal women (55), while the premutation ovaries were from a mix of actively cycling and post-menopausal women. This link to the human condition is an important one to judge the suitability of mouse models with expanded CGG-repeat expression. Along with the findings in the Rosa26 mouse models, we must consider the identification of inclusion cysts within human controls indicated that these may be the product of ovarian aging rather than a distinct pathology. Further evaluation of primary human tissues and mechanistic studies in the Rosa26 mouse is necessary to fully understand the genesis of this phenotype.

One of our most striking observations was that ovarian dysfunction is apparent in prepubertal females. This phenotype is apparent before evidence of premutation pathology in the brain or alterations in body weight (34). Through RNA-sequencing analysis of ovaries from young mice, we detected enrichment in processes related to ovarian cell biogenesis (Supplementary Material, Figure S3). Using a superovulation paradigm, we were able to show that expression of expanded CGG repeats is sufficient to impair key ovulatory processes in response to exogenous hormones. These experiments also enabled us to examine intrinsic ovarian activities since endogenous signals from the hypothalamus and pituitary were overridden. These studies proved to be particularly instructive, as they revealed a new maturation defect in CGG-expressing ovaries. RNA-seq data collected from aging ovaries indicated several pathways that are associated with the final maturation of follicles and ovulation. Oocytes rely on the granulosa cells for the production and transport of certain metabolites during development. The expression of genes related to cholesterol synthesis is upregulated in the differentiated cumulus cells prior to the LH surge (56). We identified several of the genes necessary for synthesis (Hmcgr, Mvd, Fdps, Idi1, Sqle and Cyp51) or transport (Slc38a3 and Slc38a5) of metabolites for ultimate oocyte maturation to be perturbed in our RNA-seq analyses of whole ovaries at both 3- and 6 months of age. The transcripts mentioned have high expression specifically within the cumulus cells in the later stages of follicle development, and previous studies with similar cumulus expansion defects display the same perturbed transcripts (56). Changes in the expression of these genes and the reduction of Lhcgr might suggest that the follicles within FMRpolyG+RNA^rec/+ ovaries do not mature properly or with appropriate timing, and can therefore, not respond to the LH surge. That could explain the loss of ovulatory events and corpora lutea in the FMRpolyG+RNA^rec/+, despite having comparable numbers of primordial follicles in ovaries at 6 months of age. Impaired or altered response to the LH cue by follicles expressing expanded CGGs could also explain why cumulus expansion is impaired when a standardized dose of luteinizing hCG is provided during superovulation.

We found that ovulated COCs had a measurable reduction in total diameter. This finding could be of significant value in human premutation women as it could be measured in the context of patients undergoing clinical IVF cycles without altering the course of treatment in these individuals. If a similar reduction in COC diameter is observed, it may suggest that follicle maturation in premutation carriers is altered and support the use of this assay in mouse models to address mechanisms and test therapeutic interventions. There is limited information directly measuring ovarian reserve or follicular growth in humans, and measurement of COC diameter from human premutation carriers could provide validation of the model system we are using. As with other phenotypes, refined understanding of the cellular mechanism(s) leading to altered COCs might be possible through use of cell-type specific Cre-recombinase strains of mice to exploit the conditional nature of the expression of the Rosa26 alleles. Restricting expanded CGG-repeat expression to either oocytes or granulosa cells would determine whether expression in a single cell-type is sufficient to impair cumulus expansion or meiotic resumption. Detailed interrogation of the gonadotropin-dependent stages of folliculogenesis are necessary to dissect these phenotypes, and this testing paradigm lends itself to evaluating genetic and/or biochemical modifiers of COC maturation either in vitro or in vivo. While the cellular mechanism(s) are under investigation, it will be valuable to leverage the observations made here to measure human correlates, such as reduced collection of ovulated oocytes and diminished COC diameters. This will enable us to understand how best to model the human condition of FXPOI in mice (13,57).

Materials and Methods

Animals

Transgenic mice with conditional expression of mRNAs carrying premutation length CGGs (99 repeats) from the Rosa26 locus were a kind gift from Nicolas Charlet-Berguerand (IGBMC, France) (34). Two lines were used, one that allows production of the RAN protein FMRpolyG (FMRpolyG+RNA), and one lacking FMRpolyG due to the deletion of the region containing the ACG non-canonical start codon (RNA only). Production of mice with global expression of CGG RNA with or without expression of RAN products was achieved by mating the females from each line with intact loxP sites (FMRpolyG+RNA^Fl/Fl and RNA-only^Fl/Fl) animals to Gdf9-cre mice, then mating FMRpolyG+RNA^Fl/+ and RNA-only^Fl/+;Gdf9-cre females to control males (Fig. 1). Animals were housed in a 12 h light/dark cycle and fed ad libitum. All experiments were conducted in accordance with the NIH Guide for the Use and Care of Laboratory Animals and approved by the Institutional Animal Care and Use Committee of Baylor College of Medicine.

Breeding, hormone and superovulation studies

To measure breeding fecundity, females were housed from ~P42 with a control male and allowed to breed continuously until 8 months of age. Litter numbers and pups per litter were recorded for each female. Blood serum samples were taken from one cohort of females at 3 months of age and a second cohort at 6 months of age. Estrous cycles were not monitored continuously but vaginal cytology analyses were performed at the time of blood collection. Females determined to fall within late estrus and early diestrus stages had blood collected. Serum was separated by centrifugation and stored at −80°C until the samples were sent to the UVA ligand core. Samples from mice at 3 months of age were tested for FSH and LH, and the 6 months of cohort were tested for FSH, LH and inhibins A and B. Complete data sets for all mice were not possible in some cases due to limited sample volume or test values outside of the sensitivity range for the assay. Superovulation studies were conducted using females between the ages of P21-P28. Intraperitoneal injection of 2.5 IU PMSG (EMD Millipore) was followed 44–48 h later by injection of 2.5 IU hCG (Sigma-Aldrich). Oviducts were collected 16–20 h post-hCG and oocytes denuded using 0.5 mg/mL hyaluronidase (Sigma).

Histology and immunohistochemistry

Ovaries were collected and fixed in 4% paraformaldehyde or Bouin’s solution, processed through a series of alcohol and xylenes before embedding in paraffin wax (McCormick Scientific Paraplast Plus). Tissues were sectioned at a thickness of 8 □m, and follicles were estimated by counting oocytes with a visible nucleus in every fifth section. The raw totals were multiplied by 5, the section frequency, to determine the follicle number per ovary. Follicles were deemed to be atretic if they exhibited at least 2 of the following: 3 or more pyknotic granulosa, vacuolated or degenerating oocyte, or immune cell infiltration. Sections were prepared for immunohistochemistry and immunofluorescence as previously described (34), and both heat- and enzyme-mediated antigen retrieval was performed utilizing sodium citrate buffer (pH 6.0) and proteinase K (GenDEPOT), respectively. Immunohistochemistry included a step for treatment of endogenous peroxidases. Antibodies were diluted in 0.5% Protifar, 0.15% Glycine in 1X PBS. The anti-FMRpolyG antibody recognized an N-terminal epitope (NTF) (58) and was kindly provided by Peter Todd at the University of Michigan. FMRpolyG NTF and cytokeratin 8 (DSHB) were diluted 1:100. Images were produced with a Zeiss camera and processed using Zeiss ZEN software (ZENBlue 2.3.1). FMRpolyG-positive inclusions were counted using standardized field of 150□m².

qRT-PCR

RNA was isolated from whole ovaries using the Zymo mini RNA isolation kit and treated with DNase I (Zymo Research). cDNA was reverse transcribed using qScript (Quanta BioSciences), and equal concentrations of cDNA loaded per sample for semi-quantitative PCR run. PerfeCTa SYBR green Supermix (Quanta BioSciences) was used to detect products and primer sequences were derived from previous publications (Supplementary Material, Table S1).

RNA-sequencing analyses

RNA-seq libraries were generated from three biological samples per condition using the NEBNext® Ultra RNA Library Prep Kit for Illumina® (New England Biolabs, cat# E7420S). Qualified libraries were sequenced on an Illumina Novaseq Platform using a paired-end 150 run (2 × 150 bases) with 20 million raw reads generated from each library. The FASTQ sequence files were aligned to the mouse reference genome (mm 9) using Bowtie 2.2.6 in pair-end alignment mode. All data analyses were performed using R software (v3.6.0) and Bioconductor packages. Gene counts with a mean read count >2 across all samples were used. These counts were compared between CGG-expressing samples and controls using the Bioconductor package DESeq2 (59). Differentially expressed genes (DE) were defined as those with FDR < 0.2 to include top-ranked DE genes while controlling for the multiple testing issue with the small sample size, and this follows an approach from existing publications (60,61). Gene set enrichment analysis (GSEA) was performed on DE genes at each age using Metascape (62), and gene ontology terms used to cluster DE genes.

Deconvolution data from bulk ovary tissue was sought to extrapolate connections between changes in transcriptomes and ovarian cell types. Since purified cell-type expression profiles for isolated ovarian cells at all ages were not available, the deconvolution was considered to be reference-free. Marker genes identified from existing literature were used to distinguish four major cell populations within the mouse ovary, oocyte (63–65), granulosa cells (64–67), theca cells (67–69) and stroma (70–72) (Supplementary Material, Table S2). Gene counts were first normalized into fragments per kilobase of transcript per Million (FPKM), and the FPKM plus marker genes were used as inputs into R/Bioconductor package TOAST for an in silico deconvolution analysis using a quadratic programming algorithm (73).

Statistical analyses

Results are reported as mean ± standard deviation (SD). Statistical analyses were performed using PRISM Graphpad Version 6. One-way and two-way ANOVA were used to analyze continuous variables and normally distributed data. Kruskal–Walls test was used for nonparametric data. To determine significance between cell proportions in deconvolution data the regression-based MiRKAT test was employed (74).

Funding

The Eunice Kennedy Shriver National Institute of Child Health & Human Development (R21HD084802); National Institute of Neurological Disorders and Stroke (R01NS051630) to D.L.N., the National Institute of General Medical Science’s (T32GM08307 training grant) to BCM’s program in Human and Molecular Genetics. The GEM core is partially supported by the National Institutes of Health (grant P30CA125123).