FMR1 and AKT/mTOR Signaling in Human Granulosa Cells: Functional Interaction and Impact on Ovarian Response

We aimed to determine whether a functional link with impact on female ovarian reserve exists between FMR1 expression and expression ratios of AKT/mTOR signaling genes in human granulosa cells in vivo, as suggested from prior in vitro data. Three hundred and nine women, who were classified as normal (NOR; n = 225) and poor (POR; n = 84) responders based on their ovarian reserve, were recruited during stimulation for assisted reproductive techniques. Expressions of FMR1 and of key genes of the AKT/mTOR and AKT/FOXO1/3 signaling pathways were comparatively analyzed in their granulosa cells. FMR1 expression in granulosa cells of NOR and POR correlated significantly with AKT1, TSC2, mTOR, and S6K expression. No correlation was found between FMR1 and FOXO1 in all, and FOXO3 expression in POR, patients. AKT1 expression was significantly higher and FOXO1 expression lower in POR samples, whereas AKT1 expression was lower and FOXO1 expression was higher in NOR samples. In human native granulosa cells, FMR1 expression significantly correlated with the expression of key genes of the AKT/mTOR signaling pathway, but not with the FOXO1/3 signaling pathway. Our data point to a functional link between FMR1 expression and expression of the AKT/mTOR signaling pathway genes controlling human follicular maturation.


Introduction
Controlled ovarian follicular maturation (folliculogenesis) after puberty is essential for successful reproduction. It results in an individual ovarian reserve dependent on age [1]. Related disorders are diminished ovarian reserve (DOR) and premature ovarian insufficiency (POI), which significantly influence spontaneous pregnancy rates and success rates in assisted reproductive techniques (ART) [2]. When women with DOR undergo controlled ovarian stimulation during ART, there is an increased risk of poor ovarian response leading to a reduced number of mature follicles. Normal responders (NOR) and poor responders (POR) are distinguished according to the ESHRE Bologna guidelines [3].
A number of genes and signaling pathways are presumed to play crucial roles in the control of follicular maturation and the ovarian reserve [4]. Of these, FMR1 (Fragile X-Mental Retardation 1, OMIM: *309550), located on the X chromosome (Xq27.3) [5], is the most prominent, because of its high mutation frequency in the female germline. Its protein FMRP is mainly localized in the granulosa cells (GCs) of ovarian follicles at different stages [6]. The gene contains a CGG base triplet in its 5 UTR (untranslated region) with usually 26-34 repeats in distinct human populations [7]. CGG triplet repeats between 54 and 200 are termed as "premutation" (PM), because this longer CGG block becomes subsequently unstable during further heredity and may lead to transmission of a full mutation (FM) allele with >200 CGG repeats in the next generation. In such cases, FMR1 is silenced due to complete CpG methylation of its promoter domain, finally causing fragile X syndrome (FXS; OMIM #300624) [8]. Up to 20% of women with FMR1-PM and so-called "grey zone" alleles (45-55 repeats) are affected by POI [9,10], making it the most common genetic cause of POI or premature ovarian failure (POF), also called fragile X POI (FXPOI, OMIM #311360) [11]. POI/POF is defined as severe or complete ovarian exhaustion before the age of 40 with hypergonadotropic oligo-to amenorrhea lasting longer than four months [12]. Approximately 1% of women are affected by this, and, in 20-25%, a genetic component is suggested [13]. Women with a FMR1-PM allele produce elevated FMR1-mRNA levels, causing reduced FMRP levels due to a negative feedback loop mechanism [14]. Increased FMR1 expression thus appears to be an important ovarian pathomechanism [15,16]. Accordingly, elevated FMR1 expression analyzed in native GCs of women with distinct ovarian responses demonstrated an association with poor ovarian reserve in women with CGG repeats below the norm of 26-34 repeats [17]. Additionally, variable FMR1 genotypes with CGG repeats below or above 26-34 impact ovarian reserve [18][19][20]. Recently, we discovered that the FMR1 transcription rate in human GCs is dependent on epigenetic factors, namely, CpG methylation patterns in three distinct genomic regions, designated as VMR1-3 (variably methylated regions) [21].
The AKT/mTOR (AKT serine/threonine kinase/mechanistic Target of Rapamycin) signaling pathway controls early folliculogenesis by maintaining the primordial follicular pool. This includes primordial follicular activation, GC proliferation, and oocyte-GC and inter-GC communication control [22][23][24]. Moreover, the pathway is active during later gonadotropin-dependent follicular maturation until preovulatory follicle formation. In animal models dependent on FSH ligation to its receptor, follicular activation through activation of this pathway has been demonstrated [25,26]. Three isoforms of AKT that share approximately 80% of their amino acids and are widely expressed in various tissues are known. The role of Akt3 in the ovary is unclear. Akt2-deficient mice show normal fertility and Akt1-deficient female mice are reported to have reduced fertility [27].
AKT1 expression has been identified as a marker in cumulus cells for a positive pregnancy outcome in humans [28]. AKT is involved in mTOR signaling; it induces phosphorylation of tuberous sclerosis protein 2 (TSC2), which forms a dimer with tuberous sclerosis protein 1 (TSC1) and triggers mTOR complex (mTORC) formation. mTORC phosphorylates, and thereby activates, S6K (ribosomal protein S6 kinase), which is the key control element for cellular transcript translation and protein synthesis. Likewise, AKT regulates nuclear transcription factors termed forkhead box O (FoxOs). While three FoxOs (FoxO1, 3, and 4) are broadly expressed, a fourth (FoxO6) is more limited to the brain [29]. In animal studies, FoxO1 and O3 have been shown to be important for gametogenesis in mammals. They are highly expressed in GCs [30] and, in their unphosphorylated form, suppress ovarian follicular activation at the earliest stages of follicular growth, avoiding accelerated follicular initiation. Further GC proliferation is inhibited and a proapoptotic pathway is initiated [31].
Previously, in the COV434 granulosa cell line, we described a putative functional linkage of FMR1/FMRP expression with the expression of key genes of the AKT/mTOR signaling pathway. Moreover, AKT1, mTOR, S6K, and FMR1/FMRP expression levels were altered depending on rFSH (recombinant follicle-stimulating hormone) stimulation [32].
Further support of the functional linkage between FMR1 expression and genes of the AKT/mTOR signaling pathway is indicated in mice by the binding of Tsc2 and mTor with FMRP within the RNA-Interference Silence Complex (RISC) [33]. RISCs locate target RNAs, leading to translational repression and/or epigenetic silencing [34]. Moreover, Fmr1 knockdown causes elevated mTOR signaling [35], and S6K is a major FMRP-phosphorylating enzyme in human neuronal cells [36]. Additionally, in mouse ovary, Fmr1-PM was associated with increased AKT phosphorylation [37].
In this study, we aimed to confirm our earlier COV434 cell culture experiments using native GCs from women undergoing ART at our outpatient clinic. Our analysis of FMR1 expression and of key genes of the AKT/mTOR and AKT/FOXO1/3 signaling pathways in native GCs of 309 women with distinct ovarian responses also indicates a functional linkage between the AKT/mTOR signaling key genes and FMR1 in vivo.

Design and Patients
We performed a prospective study and enrolled 309 women with different ovarian responses at the Heidelberg University Women's Hospital (Germany) from February 2013 to October 2019. They all provided written informed consent and completed a clinical questionnaire for collecting blood and GC samples. The study was approved by the local ethics committee of the Ruprecht Karls University of Heidelberg, Germany (number S-602/2013) and conducted according to the principles expressed in the Declaration of Helsinki.

Ovarian Stimulation
The appropriate protocol was selected by physicians independently of this study. Either the long protocol of gonadotropin-releasing hormone (GnRH) agonist administration (long GnRH agonist protocol) or the GnRH antagonist protocol was used for ovarian stimulation. For the long GnRH agonist protocol, an initial downregulation, using a GnRH agonist at day 20 + 1 of the menstrual cycle, was used. On day two of the following cycle, gonadotropins (rFSH or HMG: human menopausal gonadotropin) were injected daily to induce proper follicular maturation. When the follicles attained a diameter of 17 mm, ovulation was induced by human chorionic gonadotropin injection and the oocytes were retrieved after 36 hours, using ultrasound-guided follicular puncture. Aspirates then were placed in 14 mL round-bottom tubes containing phosphate-buffered saline (ART-4012, SAGE IVF, Cooper Surgical, Trumbull, CT, USA) and heparin (2.5 IU/mL) or Flush medium (4 GM 501F-100, Gynemed, Lensahn, Germany). For the GnRH antagonist protocol, gonadotropins (mainly recFSH or HMG) were injected daily to induce proper follicular maturation, beginning at day two of the menstrual cycle. When the leading follicle reached an average diameter of 14 mm, a GnRH antagonist was used to prevent preterm spontaneous ovulation. When the follicles reached 18 mm in diameter, the oocytes were retrieved after induction of ovulation as described above. The cumulative dose of gonadotropins was determined based on the patients' responses.

Retrieval of Granulosa Cells
GCs were recovered from the follicular fluid after transvaginal ultrasound-guided follicle puncture for in vitro fertilization (IVF), as described previously [17]. The follicles were aspirated with a specific needle (Premium Fas Single Lumen, #4551 NS-AS1; Gynétics Medical Products N.V., Lommel, Belgium) that was connected to a vacuum pump (Cook Medical, K-MAR-5200, Bloomington, IN, USA). The aspirated follicular fluid was collected in 14 mL round-bottom tubes (Falcon, NY 352001, USA) and maintained at 37 • C. The follicular fluid was then transferred to a cell culture dish (150350 or 150360, Thermo Fisher Scientific, Nunc, Waltham, MA, USA) on a table heated to 37 • C (Workstation L126 Dual, K-Systems, Birkerød, Denmark). Mural GCs were identified morphologically within the follicular fluid using a Nikon SMZ1500 zoom-stereomicroscope (Nikon Instruments Europe B.V., Amsterdam, The Netherlands). In most cases, GCs were directly taken from the follicular fluid without additional washing. A brief washing step in a culture medium

RNA Extraction
GCs in RNAlater were centrifuged at 5000× g for 5 min, and the supernatants were removed. Total RNA was isolated from these GCs using TRIzol (Life Technologies by Thermo Fisher, Carlsbad, CA, USA) according to the manufacturer's instructions [38,39] with PEQGOLD PHA-SETRAP A 1.5 mL tubes (VWR International GmbH, Darmstadt, Germany) or MaXtract TM High Density 1.5 mL tubes (Qiage Germantown, MD, USA), as described previously [17]. RNA was dissolved in RNase-free water, and concentration and purity were determined using a NanoDrop 2000c UV-spectrometer (NanoDrop Products, Wilmington, DE, USA). cDNA of the mRNA fraction was synthesized after oligo-dT priming with the SuperScript III First-Strand Synthesis System (Invitrogen by Life Technologies, Carlsbad, CA, USA) and M-MLV Reverse Transcriptase, RNase H Minus, Point Mutant (Promega, Madison, WI, USA).

Statistical Analysis
Data distribution was first determined using the Shapiro-Wilk test. For a simple comparison between ovarian reserves (NOR and POR), Student's t-test or the Mann-Whitney test was used. Correlation was calculated using the Shearman-Rho correlation coefficient, as not all analyzed genes were normally distributed. Regression curves were tested for variable models for the AKT/mTOR signaling and AKT/FOXO signaling genes' expression values with respect to FMR1 expression values, patient age, and total rFSH or total HMG dose. The CGG subgroups and analysis of cohort demographic data were compared using a one-way analysis of variance or Kruskal-Wallis test as required. The adjusted Chi-square test was used for group comparisons (clinical ovarian response and IVF/intracytoplasmic sperm injection treatment). Results are presented as means ± standard deviations or median and interquartile range (25th-75th percentile; 1st-3rd quartile). For n < 3, data are presented as median and minimum and maximum values (minimum-maximum). Statistical analyses were performed using the Statistical Package for the Social Sciences V. 22.0 and 27 (IBM Corporation, Armonk, NY, USA), and statistical significance was set at p < 0.05.

CGG Repeat Analysis
Briefly, DNA samples were obtained from 10 mL of blood samples with EDTA, as described previously [17]. CGG repeat length in the 5 -UTR of . When the presence of PM was suspected, a Southern blot was performed using a 32P-dCTP radioactively-labeled p2 probe containing FMR1 exon 1 with CGG repeats, as described previously [42].

Non-PM Allele Length Classification
CGG repeat lengths were categorized according to previous studies [19,43]. Patients were classified according to the repeat lengths at both alleles (low: < 26 repeats; normal: 26-34 repeats; high: 35-55 repeats) into six different genotypes: high/high, high/low, normal/high, normal/normal, normal/low, and low/low. The genotype high/high was not present in any patient, nor was PM with alleles over 54 repeats.

General Study Population
A total of 84 POR and 225 NOR samples were included. The demographics are summarized in Table 1. The two groups differed significantly in AFC, AMH, age, total number of oocytes received, and number of mature MII oocytes, but not in BMI, nor in estradiol, FSH, or LH levels.

Gene Expression Analyses, Correlation and Regression Curve Models
In all patients, mRNA expression levels of AKT1, TSC2, mTOR, S6K and FOXO3 had a positive and statistically significant correlation with FMR1 mRNA expression level (p < 0.01 for all parameters) ( Table 2). Table 2. Correlations of FMR1 gene expression and AKT1 (AKT serine/threonine kinase), FOXO3 (forkhead box O3), FOXO1 (forkhead box O1), mTOR (mechanistic Target of Rapamycin), TSC2 (tuberous sclerosis protein 2), and S6K (ribosomal protein S6 kinase) gene expression in all patients. The relative gene expression in granulosa cells of patients was normalized by two housekeeping genes and a granulosa cell calibrator using the ∆∆Ct method. Correlation was calculated using the Spearman-Rho (r) correlation coefficient.

Spearman-Rho Correlation
Co-Efficient FMR1 When analyzed based on ovarian response, correlations within the groups of NOR and POR remained significant (Tables 3 and 4), demonstrating significant correlation coefficients in both groups as well, particularly for AKT1, mTOR, TSC2, and S6K.
However, for FOXO1, no correlation was detected in their expression levels, nor with FMR1 expression levels. This was consistent for all patients, as was confirmed after subgroup analysis considering the ovarian response (Tables 2-4). Moreover, for FOXO3 in POR, significance was lost.
In addition, generalized regression curves for variable regression models were produced and are included as supplemental figures (Supplemental Figure S1-S6). They demonstrate, for AKT, mTOR, TSC2, S6K, and FOXO3 expression values, highly significant regression curves (ANOVA p < 0.000) with FMR1 expression, with the linear model being the best fitting one compared to other models. FOXO1 showed no significant regression for any of the analyzed models with FMR1. Moreover, when analyzing the expression of the genes with respect to age and total rFSH or HMG dose, none of the regression models led to significant results (ANOVA p values n.s. for all). Table 3. Correlations of FMR1 gene expression with AKT1, FOXO3, FOXO1, mTOR, TSC2, and S6K expressions according to ovarian response group (NOR). The relative gene expression in granulosa cells of patients was normalized by two housekeeping genes and a granulosa cell calibrator using the ∆∆Ct method. Correlation was calculated using the Spearman-Rho correlation coefficient.

Spearman-Rho Correlation
Co-Efficient FMR1  Figure 1). Table 4. Correlations of FMR1 gene expression with AKT1, FOXO3, FOXO1, mTOR, TSC2, and S6K expressions according to ovarian response group (POR). The relative gene expression in granulosa cells of patients was normalized by two housekeeping genes and a granulosa cell calibrator using the ∆∆Ct method. Correlation was calculated using the Spearman-Rho correlation coefficient.

306
Distribution of different FMR1 genotypes between NOR and POR groups was com-307 parable ( To exclude the impact of CGG repeat length on our results, relative gene expression 311 levels of AKT1, TSC2, mTOR, S6K, FOXO1, and FOXO3 were subdivided into different 312 FMR1 genotypes (high/high, high/low, normal/high, normal/normal, normal/low, and 313 low/low) and ovarian response groups (NOR and POR). Expression of the key genes of 314 the AKT/mTOR signaling pathway, i.e., AKT and FOXO1/3 in the different FMR1 geno-315 type subgroups was not significantly different in our analysis, in none of the response 316 groups (Supplementary Table S1).

318
Our study data indicate a functional interplay between FMR1 expression rate and the 319 expression of key genes of the AKT/mTOR signaling pathway in native human GCs, with 320 impact on ovarian response. We confirmed comparable in vitro data based on the COV434 321 cell line [32] using ex vivo GC samples. Clinically, NOR and POR samples are distin-322 guished by differences in AFC, AMH, number of total oocytes received, and number of 323 mature MII oocytes. Both patient groups displayed similar BMI and basal estradiol, FSH, 324 and LH levels. Variable FMR1 genotypes also did not impact the expression levels of the 325 analyzed AKT/mTOR or AKT/FOXO signaling pathway genes. 326 During the highly orchestrated molecular processes of human folliculogenesis and 327 oocyte maturation, GCs continuously interact with oocytes and directly control the fate of 328 follicles. They are, therefore, of utmost importance in follicular maintenance and growth, 329 as well as for follicular atresia controlled by apoptotic processes [44]. For TSC2, mTOR, S6K, and FOXO3 expression levels, no significant differences between NOR and POR groups were detected (Table 4).

CGG Repeat Analysis and Gene Expression Values
Distribution of different FMR1 genotypes between NOR and POR groups was comparable (Table 6). No PM was detected. To exclude the impact of CGG repeat length on our results, relative gene expression levels of AKT1, TSC2, mTOR, S6K, FOXO1, and FOXO3 were subdivided into different FMR1 genotypes (high/high, high/low, normal/high, normal/normal, normal/low, and low/low) and ovarian response groups (NOR and POR). Expression of the key genes of the AKT/mTOR signaling pathway, i.e., AKT and FOXO1/3 in the different FMR1 genotype subgroups was not significantly different in our analysis, in none of the response groups (Supplementary Table S1).

Discussion
Our study data indicate a functional interplay between FMR1 expression rate and the expression of key genes of the AKT/mTOR signaling pathway in native human GCs, with impact on ovarian response. We confirmed comparable in vitro data based on the COV434 cell line [32] using ex vivo GC samples. Clinically, NOR and POR samples are distinguished by differences in AFC, AMH, number of total oocytes received, and number of mature MII oocytes. Both patient groups displayed similar BMI and basal estradiol, FSH, and LH levels. Variable FMR1 genotypes also did not impact the expression levels of the analyzed AKT/mTOR or AKT/FOXO signaling pathway genes.
During the highly orchestrated molecular processes of human folliculogenesis and oocyte maturation, GCs continuously interact with oocytes and directly control the fate of follicles. They are, therefore, of utmost importance in follicular maintenance and growth, as well as for follicular atresia controlled by apoptotic processes [44].
The AKT/mTOR signaling pathway plays a role in diverse cellular functions and initiates the transcription-and regulates the translation-of multiple genes in many cell types. It also plays roles in autophagy and the biosynthesis of lysosomes and lipids [45]. In this pathway, AKT acts as a prominent protein phosphorylation kinase. In human ovaries, it is expressed in oocytes and GCs. Accordingly, in the mouse AKT isoform, AKT1 is involved in fertility control [27,28]. AKT phosphorylates TSC2, which then forms a dimer with TSC1 to induce mTORC formation. Subsequently, mTORC phosphorylates S6K, a major kinase in a further cell-specific translational control mechanism. In human neuronal cells, elevated S6K expression induces FMRP phosphorylation [36]. The AKT/mTOR signaling pathway also regulates GC differentiation and proliferation and is thus functional for ovarian activation [46].
In this study, we demonstrated that FMR1 expression in native human GCs in women with different ovarian responses is significantly and positively correlated with the expression levels of AKT1, TSC2, mTOR, and S6K. Correlations between FMR1 expression and the AKT-FOXO1/3 signaling pathway genes were not apparent, especially in the POR patient group. Results were consistent after regression curve analysis for all genes. These results thus support a specific functional linkage between FMR1/FMRP expression and AKT/mTOR signaling in human GCs, supporting the results of our previous study on the COV434 granulosa cell line [32]. Putative confounders, i.e., age and total rFSH or HMG dose, were excluded, as no significant regression curve was apparent for any of the analyzed genes values with respect to those factors.
Elevated FMR1 expression levels cause reduced FMRP levels by a negative feedback loop mechanism [14]. Because FMRP binds Tsc2 and mTor within the RISC [33], decreased FMRP levels will suppress the binding of GCs in the RISC. This results in the increased expression of unbound AKT/mTOR gene products, activating GC differentiation and proliferation, and further follicular development. This hypothesis was further supported by the results of the POR patient group, where S6K correlation appeared to be more pronounced. This may reflect a compensational reaction to the lower ovarian reserve, although proof of such a causality needs to be evaluated in future studies to discriminate between an FMR1 expression only or an FMR1 and ovarian response-dependent mechanism. Nevertheless, our results are also in line with those seen in elevated mTOR and S6K levels in Fmr-KO mice, which led to the premature recruitment of the oocyte pool [47], and are comparable with the results for lower AMH level (POR), where increased follicular activation as a compensatory mechanism is suggested [48].
AKT expression is the bridge between mTOR and FOXO signaling. In the latter, it induces FOXO phosphorylation, leading to their exclusion from the nucleus into the cytoplasm and, thereby, to their inactivation [49]. Decreased AKT expression thus results in dephosphorylation of FOXO1/3, which then shuttles back into the nucleus. There, it induces a proapoptotic pathway with DNA fragmentation, granulosa cell death, and, finally, follicular atresia, as well as oocyte silencing [44]. Consequently, women with POI were offered AKT activators, which lead to follicular activation [50].
In this study, FOXO1/3 expression was analyzed in fresh GCs of patients with different ovarian responses. Of these, FOXO1 is highly expressed in human GCs [51]. Although we found no correlative effect of FMR1 on the expression levels of these FOXOs, at least in the poor responding group, an inverse gene expression activation of AKT1 and FOXO1 depending on ovarian response has been identified; higher AKT1 and lower FOXO1 expression were apparent in POR, and the reverse was true in NOR. Comparable results were shown in recent animal studies [52,53]. These studies suggest an opposing AKT1-triggered and FMR1-independent regulating mechanism that controls the ovarian reserve. Downregulated FOXO1 in POR may thus prevent increased proapoptotic events to maintain the already limited ovarian reserve.

Limitations
Our results are limited by the sample size. The number of GCs obtained per puncture is limited as well. In addition, we used GCs only from preovulatory follicles, which may differ from GCs of other, earlier stages of development. Patient ages were slightly different in both groups, with POR patients older than NOR patients by an average of three years. However, both patient groups had a mean age below 40 years. Nevertheless, an age-related effect cannot be excluded. As granulosa cell numbers per patient and per puncture are limited, expression analysis was not possible for all genes in every patient, most notably for TSC2 and FOXO1.
Therefore, further studies including more clinically defined NOR and POR patients, subdivided into different age groups, are needed to support the suggested functional FMR1/AKT/mTOR linkage and its impact on female fertility.

Conclusions
We confirmed a functional linkage between FMR1 expression and the AKT/mTOR signaling pathway genes in native human GCs, which was first revealed in the COV434 granulosa cell line. This linkage is only found with the key genes of the AKT/mTOR signaling pathway. AKT/FOXO1/3 signaling was not associated with FMR1 expression in poor responding patients. This FMR1/AKT/mTOR linkage is, therefore, probably specific to follicular maturation control and indicates a putative, ovarian reserve-dependent balancing mechanism in the female germline.