Follicle-stimulating hormone synthesis and fertility depend on SMAD4 and FOXL2

Abstract

Follicle-stimulating hormone (FSH) is an essential regulator of gonadal function and fertility. Loss-of-function mutations in the FSHB/Fshb gene cause hypogonadotropic hypogonadism in humans and mice. Both gonadotropin-releasing hormone (GnRH) and activins, members of the transforming growth factor β (TGFβ) superfamily, stimulate FSH synthesis; yet, their relative roles and mechanisms of action in vivo are unknown. Here, using conditional gene-targeting, we show that the canonical mediator of TGFβ superfamily signaling, SMAD4, is absolutely required for normal FSH synthesis in both male and female mice. Moreover, when the Smad4 gene is ablated in combination with its DNA binding cofactor Foxl2 in gonadotrope cells, mice make essentially no FSH and females are sterile. Indeed, the phenotype of these animals is remarkably similar to that of Fshb-knockout mice. Not only do these results establish SMAD4 and FOXL2 as essential master regulators of Fshb transcription in vivo, they also suggest that activins, or related ligands, could play more important roles in FSH synthesis than GnRH.—Fortin, J., Boehm, U., Deng, C.-X., Treier, M., Bernard, D. J. Follicle-stimulating hormone synthesis and fertility depend on SMAD4 and FOXL2.

Approximately 10% of couples are infertile. Although both male and female factors contribute to the problem, the underlying causes are frequently unknown, stemming from our incomplete understanding of the physiological processes controlling reproduction. Gonadal (testicular and ovarian) function is regulated by hormonal signals from the brain and pituitary gland. Impairments in the synthesis, secretion, or action of these hormones can cause hypogonadotropic hypogonadism (HH), which usually manifests as delayed or absent puberty (1–9). Although there are many causes of HH, impaired gonadal function ultimately results from insufficient stimulation by the gonadotropins, luteinizing hormone (LH), and follicle-stimulating hormone (FSH). Circulating LH and FSH levels are also perturbed in other forms of infertility, including polycystic ovarian syndrome and premature ovarian failure (10, 11).

LH and FSH are dimeric glycoproteins produced by gonadotrope cells of the anterior pituitary. They share a common α subunit [chorionic gonadotropin α (CGA)] noncovalently linked to hormone-specific β subunits (LHβ and FSHβ). Both the hypothalamic decapeptide GnRH (2, 9, 12, 13) and intrapituitary activins (14) have been implicated as the primary stimulators of FSHβ synthesis.

Activins are members of the transforming growth factor β (TGFβ) superfamily and were discovered (and named) for their abilities to stimulate FSH secretion (14–16). Analyses in model cell lines have uncovered candidate mechanisms of activin action in vivo (14, 17). In brief, activins bind to receptor serine/threonine kinases on the plasma membrane of gonadotropes, leading to phosphorylation of the intracellular signaling proteins Sma- and Mad-related protein family member 2 and 3 (SMAD2 and SMAD3). Phosphorylated SMAD2 and SMAD3 dissociate from the receptors and associate with the obligate cofactor SMAD4 (18). SMAD complexes then accumulate in the nucleus and activate transcription of the FSHβ subunit gene (Fshb), which is generally considered the rate-limiting step in FSH synthesis (14, 19, 20). In vitro data indicate that SMADs partner with the cell-restricted transcription factor forkhead box L2 (FOXL2; refs. 21–24), and perhaps other molecules (25–27), to regulate transcription via both conserved and species-specific cis-regulatory elements in the proximal Fshb promoter.

Consistent with this model, gonadotrope-specific ablation of Foxl2 causes selective FSH deficiency and subfertility in mice (28). Conversely, however, mice lacking SMAD2 and the DNA-binding domain of SMAD3 have normal FSH levels and fertility (29). Thus, it is presently unclear whether activins, or other TGFβ superfamily ligands, signal through SMADs or rather a noncanonical pathway to regulate Fshb expression and FSH synthesis in vivo. To discriminate between these possibilities, we selectively ablated Smad4 in gonadotrope cells in vivo using conditional gene targeting in mice. As SMAD4 mediates the actions of all TGFβ superfamily ligands (18), this approach enabled us to assess whether or not SMAD-dependent signaling is required for FSH synthesis in the intact murine pituitary gland. Furthermore, we ablated Smad4 and Foxl2 in combination to determine whether the two proteins function cooperatively and/or independently to regulate FSH in vivo.

MATERIALS AND METHODS

Mice

The Smad4^fl, Foxl2^fl, gonadotropin-releasing hormone receptor (Gnrhr)^{IRES-Cre(GRIC}), and ROSA26^eYFP alleles, as well as corresponding genotyping protocols, were described previously (30–33). To generate mice with the experimental genotypes, the Gnrhr^GRIC allele was always introduced from the females, due to germline Cre activity in males (34). To obtain Smad4 conditional-knockout (S4cKO; Smad4^fl/fl;Gnrhr^GRIC/+) mice, Smad4^fl/fl males were crossed with Smad4^fl/+;Gnrhr^GRIC/+females. Smad4^fl/fl littermates were used as controls. To obtain Smad4/Foxl2 conditional-knockout (S4F2cKO) mice (Smad4^fl/fl;Foxl2^fl/fl;Gnrhr^GRIC/+), Smad4^fl/fl;Foxl2^fl/fl males were crossed with Smad4^fl/+;Foxl2^fl/+;Gnrhr^GRIC/+ females. Smad4^fl/fl;Foxl2^fl/fl littermates were used as controls. To obtain S4cKO-yellow fluorescent protein (YFP) mice (Smad4^fl/fl;ROSA26^eYFP/+;Gnrhr^GRIC/+), Smad4^fl/fl;ROSA26^eYFP/eYFP males were crossed with Smad4^fl/+;Gnrhr^GRIC/+females. To obtain GRIC-YFP mice (Gnrhr^GRIC/+;ROSA26^eYFP/+), ROSA26^eYFP/eYFP males were crossed with Gnrhr^GRIC/GRIC females. All animal experiments were performed in accordance with institutional and federal guidelines and approved by the McGill University Institutional Animal Care and Use Committee (protocol 5204).

Fluorescence-activated cell sorting (FACS)

Dissociated pituitary cell suspensions were prepared from adult (> 8-wk-old) S4cKO-YFP and GRIC-YFP mice as described previously (28, 35). Sorting was performed on a FACSAria cell sorter (BD Biosciences, San Jose, CA, USA) at the flow cytometry core facility of the McGill University Life Sciences Complex. RNA was obtained from the YFP⁺ and YFP⁻ cell populations using the Total RNA Mini Kit (Geneaid, New Taipei City, Taiwan) following the manufacturer's instructions.

Reproductive organ analysis, testicular/ovarian histology, and sperm and follicle counting

Reproductive organs were collected from adult mice at the indicated ages and weighed on a precision balance. Formalin-fixed ovaries were paraffin embedded and cut into 5 μm sections. Every fifth section was hematoxylin and eosin (H&E) stained and analyzed by transmitted light microscopy. This allowed tracking of corpora lutea and antral follicles across several sections. Follicle staging and counting were performed following published guidelines (36, 37). The number of preantral follicles in each section was estimated by counting the number of oocytes. For testicular histology, testes were fixed in Bouin's with gentle motion overnight. The testes were then washed with 95 and 70% ethanol and paraffin embedded. Transverse sections (7 μm) were obtained in the middle of the tissue, H&E stained, and examined by light microscopy. For sperm counting, cauda epididymides were dissected and immediately frozen on dry ice. Homogenization-resistant sperm count was performed as described previously (28).

Fertility assessment

To assess fertility, 10-wk-old male or female experimental and control mice were paired with C57BL/6J mice (Charles River, St-Constant, QC, Canada) of the opposite sex. Starting from 20 d after pairing, the cages were inspected daily for the presence of newborn mice. As soon as a litter was present, pups (living or dead) were carefully counted and put back into the cage. Pups were separated from the mother at postnatal day 15 (P15) to avoid interfering with the following pregnancy. The mating trial was performed for 6 mo in the case of S4cKO and control animals and 4 mo in the case of S4F2cKO and control females. S4F2cKO females, which never delivered a litter, were carefully inspected at several time points for signs of pregnancy, which were never observed.

Puberty and estrous cyclicity assessment

Starting from the day of weaning (P21), females were inspected daily for vaginal opening following published guidelines (38). At 7 wk of age, estrous cyclicity was assessed daily in the morning (∼10 AM). A cotton swab wet with sterile saline was introduced ∼5 mm into the vagina, and collected cells were smeared on a glass slide. The smears were stained with 0.1% methyl blue and examined by light microscopy. Staging was assessed according to published guidelines (38). One cycle was defined as the sequential appearance of all estrous cycle stages, regardless of the number of days spent in each stage. For estrus morning experiments, cyclicity was first assessed for ≥2 complete cycles to facilitate stage assignment. Thereafter, blood and pituitaries were collected at 7 AM the day following a clear proestrous smear.

Superovulation

Juvenile (P23–P25) females were injected intraperitoneally with 5 IU pregnant mare serum gonadotropin (PMSG; Sigma, St. Louis, MO, USA) at 11 AM on d 1. On d 3, mice were injected intraperitoneally with 5 IU of human chorionic gonadotropin (hCG; Sigma) at 7 AM. After 14 h, mice were killed, and cumulus-oocyte complexes (COCs) were retrieved by puncturing the oviduct under a dissecting microscope. The COCs were transferred to phosphate-buffered saline (PBS). Oocytes were dissociated from COCs by a 10 min treatment with 2 mg/ml hyaluronidase in PBS at 37°C and counted under a light microscope.

Hormone analyses

Blood was collected by cardiac puncture and left to clot at room temperature for 20 min. Serum was obtained following centrifugation at 3000 g and stored at −20°C until analysis. For measurements of pituitary FSH content, whole pituitaries were homogenized in 0.5 ml PBS. Samples were centrifuged for 5 min at 13,000 g. Supernatants were isolated and diluted at 1:200 for measurement. All hormone assays were performed at the Ligand Assay and Analysis Core of the Center for Research in Reproduction at the University of Virginia (Charlottesville, VA, USA). Serum LH and FSH were measured by multiplex enzyme-linked immunosorbent assay (ELISA). Estradiol was measured by ELISA. Pituitary FSH, testosterone, and progesterone were measured by radioimmunoassay.

Quantitative polymerase chain reaction (qPCR)

Pituitaries were frozen on dry ice immediately on dissection and stored at −80°C. Single pituitaries were homogenized in 500 μl TriZol reagent (Invitrogen, Carlsbad, CA, USA), and RNA was isolated following the manufacturer's protocol. RNA (1.5 μg)was reverse transcribed using Moloney murine leukemia virus (MMLV) reverse transcriptase (Promega, Madison, WI, USA) as described previously (19) in a final volume of 40 μl. For qPCR analysis, 1 μl of complementary DNA was used in triplicate reactions and assayed on a Corbett Rotorgene 6000 instrument (Corbett Life Science, Mortlake, NSW, Australia) using Platinum qPCR Supermix-UDG (Invitrogen). Gene expression was determined using the 2^−ΔΔCt method (39) relative to the expression of the housekeeping gene Rpl19, using primers described previously (28, 29, 35).

Primary pituitary culture

Primary pituitary cultures were prepared as described previously by our laboratory (28, 40). Cells were seeded at a density of 4 × 10⁵ cells/well in 48-well plates. For in vivo recombination experiments, cells from S4cKO or control littermates were cultured in 10% serum-containing medium for 24 h and then in 2% serum-containing medium supplemented with 1 nM activin A, where indicated, for an additional 24 h. For ex vivo recombination experiments, pituitary cells from Smad4^fl/fl or Smad4^fl/fl;Foxl2^fl/fl mice were cultured for 24 h after plating and then infected with adenoviruses expressing green fluorescent protein (Ad-GFP) or Cre-internal ribosome entry site (IRES)-GFP (Ad-Cre) (Baylor College of Medicine Vector Development Laboratory, Houston, TX, USA) at a multiplicity of infection (MOI) of 60 in 10% serum-containing medium for 24 h. Cells were further cultured for 24 h in 2% serum-containing medium in the presence or absence of 1 nM activin A. Cells were harvested with 0.25% trypsin, and RNA was extracted using the Total RNA Mini Kit (Geneaid) following the manufacturer's instructions.

Immunofluorescence and gonadotrope counting

Deeply anesthetized adult mice were perfused transcardially with PBS, followed by 4% paraformaldehyde (PFA) buffered in PBS. The pituitaries were postfixed in 4% PFA overnight at 4°C, transferred to PBS, and embedded in paraffin. Pituitary sections (5 μm) were progressively rehydrated using a graded series of ethanol solutions and subjected to antigen retrieval in boiling 10 mM sodium citrate (pH 6.0) for 35 min. Sections were blocked in 5% bovine serum albumin (BSA)-containing PBS with 0.2% Tween-20 (PBST) for 1 h at room temperature and incubated with primary antibodies overnight at 4°C in 5% BSA-PBST. After a series of washes in PBST, fluorophore-conjugated secondary antibodies were added in 5% BSA-PBST at a concentration of 1:500 and incubated at room temperature for 1 h. For the Cre antibody, sections were incubated with 1:150 biotinylated anti-rabbit IgG, washed in PBST, and further incubated in fluorophore-conjugated streptavidin at a 1:400 dilution. After another series of PBST washes, sections were mounted with 4′,6-diamidino-2-phenylindole (DAPI)-containing Prolong Gold reagent (Invitrogen) and imaged by epifluorescence on a Zeiss Axioplan 2 microscope (Carl Zeiss, Oberkochen, Germany). Primary antibodies used were anti-FSHβ (NIDDK AFP7798-1289, 1:500, raised in rabbit), anti-LHβ (Santa Cruz Biotechnology, Santa Cruz, CA, USA, sc-7824, 1:500, raised in goat), and anti-Cre (a kind gift from Dr. Jacques Drouin; Novagen 69050, 1:500, raised in rabbit). Secondary antibodies used were: Alexa fluor 594-conjugated anti-rabbit (Invitrogen, A21-207, raised in donkey), Alexa fluor 488-conjugated anti-goat (Invitrogen, A-11055, raised in donkey), Alexa Fluor 594-conjugated anti-goat (Invitrogen, A-11058, raised in donkey), biotinylated anti-rabbit (Vector, BA-1100, raised in horse), and Alexa fluor 488-conjugated streptavidin (Invitrogen, S-11223).

For gonadotrope cell counting, 2 nonadjacent pituitary sections per mouse were immunostained as above with FSHβ and LHβ antibodies and visualized by fluorescence microscopy. For each section, four 0.25 mm² squares were drawn within the anterior lobe. LHβ- and FSHβ-immunoreactive cells were counted within each square, considering only the cells that were fully enclosed within the limits of the square. The total number of cells counted within all 4 squares (1 mm² of tissue) was averaged between the 2 sections.

Statistical analyses

Reproductive organ weights, sperm counts, ovarian follicle counts, ovulated oocyte counts, onset of puberty, estrous cyclicity, pituitary gene expression, litter size, serum hormones, and gonadotrope cell counts were analyzed using unpaired t tests. In the case of pituitary Fshb transcripts and serum FSH levels in S4F2cKO females, Mann-Whitney U tests were used due to high variability in the control group. Primary culture experiments were analyzed using 2-way analysis of variance (ANOVA), followed by Tukey post hoc tests. Data were log transformed when variances were unequal between groups. Statistical analyses were performed using Systat 10.2 (Systat Software, San Jose, CA, USA) or GraphPad Prism 5 (GraphPad, San Diego, CA, USA). Values of P < 0.05 were considered statistically significant.

RESULTS

Generation of gonadotrope-specific S4cKO mice

To generate S4cKO mice lacking Smad4 specifically in gonadotropes, we crossed mice carrying conditional alleles of Smad4 (Smad4^fl/fl; ref. 33) with GnRHR-IRES-Cre (GRIC) mice, which express Cre recombinase in gonadotropes (32). To quantitatively assess the extent of Smad4 deletion in the gonadotropes of S4cKO animals, we crossed in the Cre-dependent ROSA26^eYFP reporter allele (30) on the S4cKO background, generating Smad4^fl/fl;Gnrhr^GRIC/+;ROSA26^eYFP/+ (S4cKO-YFP) mice, thereby enabling purification of gonadotropes by FACS (28, 29, 35). In control Smad4^+/+;Gnrhr^GRIC/+;ROSA26^eYFP/+ (GRIC-YFP) mice, Smad4 expression appeared higher in gonadotropes (YFP⁺) compared with other pituitary cell lineages (YFP⁻) (Fig. 1A). Similar to what we reported in other models using the GRIC allele (28, 29), there was a robust (∼90%) loss of Smad4 mRNA in S4cKO-YFP gonadotropes (compare green bars in Fig. 1A). Smad4 expression was comparable in nongonadotropes (YFP⁻) of the two genotypes.

Figure 1.

Hypogonadism in S4cKO male mice. A) YFP⁻ and YFP⁺ cells were sorted from ROSA26^eYFP/+;Gnrhr^GRIC/+ (GRIC-YFP; controls) or Smad4^fl/fl;ROSA26^eYFP/+;Gnrhr^GRIC/+ (S4cKO-YFP) male and female mice by FACS. Smad4 expression was assessed by qPCR and measured relative to Rpl19 expression; n = 3 independent sorting experiments. B) Representative testes from 10-wk-old control and S4cKO mice. C) Testicular weights (expressed as percentage body weight) in 10-wk-old control and S4cKO mice. Body weights did not differ. D) Cauda epididymal sperm counts in adult control and S4cKO males (one epididymis analyzed per mouse). E) Seminal vesicle weights, expressed as percentage of body weight (which did not differ), in 10-wk-old control and S4cKO males. Bars represent means ± sem. Groups/bars with different symbols differ significantly. In some panels, individual data points are plotted as circles or squares; means are shown by horizontal lines. n.s., no significant difference. *P < 0.05.

S4cKO mice are hypogonadal and subfertile

S4cKO mice developed normally and appeared healthy. However, examination of the reproductive organs in adult (10-wk-old) animals revealed hypogonadism. Male S4cKO mice had markedly decreased testicular weights (∼40% less than controls), which was accompanied by a 50% reduction in epididymal sperm counts (Fig. 1B–D). Histology revealed grossly normal testicular morphology, although the seminiferous tubules were generally smaller and many had narrow lumina (Supplemental Fig. S1A). In contrast, seminal vesicle weights and circulating testosterone levels did not differ between S4cKO and control mice (Fig. 1E and Supplemental Fig. S1B). S4cKO males also exhibited normal fertility (Supplemental Fig. S1C).

In contrast, S4cKO females were subfertile (Fig. 2A), producing significantly fewer pups compared with control mice over the course of a 6 mo breeding trial (Supplemental Fig. S1D). This was due to a smaller number of pups produced per litter (Fig. 2A); litter frequency was comparable between genotypes (Supplemental Fig. S1E). Comprehensive assessment of reproductive maturation in S4cKO females revealed normal puberty onset and robust estrous cyclicity (Supplemental Fig. S2A–C), indicating that the subfertility was unlikely caused by abnormal activation of the reproductive axis.

Figure 2.

Subfertility and abnormal ovarian follicular maturation in S4cKO female mice. A) Average litter size delivered by control and S4cKO females during a 6 mo breeding trial. B) Representative H&E-stained ovarian histological sections from 10-wk-old control and S4cKO females. Scale bars = 0.5 mm. C) Diestrous ovarian weights, expressed as percentage of body weight (which did not differ), in 10-wk-old control and S4cKO female mice. D) Corpora lutea (i), late antral/preovulatory (ii), early antral (iii), and preantral (iv) follicle counts in ovaries from 10-wk-old control and S4cKO females (1 ovary examined per mouse). E) Number of oocytes recovered from both oviducts in response to PMSG/hCG stimulation in juvenile (P23–P25) control and S4cKO females. Bars represent means ± sem. n.s., no significant difference. *P < 0.05.

To identify the basis of the subfertility in S4cKO females, we first examined their reproductive organs. Adult S4cKO females had smaller ovaries but normal uterine weights and serum estradiol on metestrus/diestrus compared with controls (Fig. 2B, C and Supplemental Fig. S2D, E). Progesterone levels, which were quite variable, did not differ significantly between S4cKO and control females (Supplemental Fig. S2F). Histological analysis indicated that, while follicles at all stages of development were present, there was a significant decrease in the number of corpora lutea in the ovaries of S4cKO animals (Fig. 2Di). To gain further insight into this phenotype, we counted follicles at different stages of development in S4cKO and control ovaries. We noted a progressive decline in the number of follicles beyond the preantral stage in S4cKO mice, whereas there were no measurable differences in follicle numbers between genotypes at earlier stages (Fig. 2Dii–iv). To rule out intrinsic ovarian defects in S4cKO mice, juvenile females (postnatal d 23–25) were stimulated with PMSG to induce follicle growth and, 48 h later, with hCG to trigger ovulation. S4cKO and control females ovulated a comparable number of oocytes, demonstrating normal ovarian responsiveness to exogenous gonadotropins (Fig. 2E).

S4cKO mice are FSH deficient

Next, we asked whether gonadotropin synthesis was impaired in mice. Indeed, there was a profound (∼90%) reduction in circulating FSH levels and pituitary FSH content in S4cKO males compared with controls, along with a 50% reduction in LH (Fig. 3Ai, iii; B). We also observed FSH deficiency in metestrous/diestrous females, although the decrease was more variable and of a lower magnitude (∼50%) than in males; LH was unaffected (Supplemental Fig. S2G). We therefore examined serum gonadotropins on the early morning of estrus, at the time of the purported activin-dependent (secondary) FSH surge (41). Mean serum FSH levels were reduced ∼4-fold in S4cKO females and robust secondary surges were absent (Fig. 3Aii). LH levels did not differ between genotypes (Fig. 3Aiv). Next, we examined whether the reduction in circulating gonadotropins was due to impaired gonadotropin subunit expression in gonadotropes. First, we analyzed pituitary sections doubly stained with antibodies directed against the LHβ and FSHβ subunits. There was a notable decrease in the intensity of FSHβ staining in both male and female S4cKO pituitaries compared with control littermates (Fig. 3C). LHβ staining appeared normal, suggesting that LH deficiency in S4cKO males does not result from impaired LHβ synthesis (see below) or gonadotrope specification (Fig. 3C). Next, we measured gonadotropin subunit mRNA levels by RT-qPCR. Fshb expression was significantly lower in S4cKO males and females (morning of estrus) compared with control littermates (Fig. 4A). In contrast, LHβ subunit (Lhb) expression was normal in S4cKO males and estrous morning females but modestly increased in metestrous/diestrous females (Fig. 4A and Supplemental Fig. S2H). Expression of Cga, which encodes the common gonadotropin α subunit, was unchanged in females but reduced in S4cKO males, perhaps contributing to their lower LH levels (Fig. 4A and Supplemental Fig. S2H). We also analyzed the pituitary expression of additional genes encoding important regulators of gonadotrope function. Interestingly, Gnrhr transcripts were up-regulated in both male and female S4cKO mice regardless of estrous cycle stage (Fig. 4A and Supplemental Fig. S2H). The activin antagonist follistatin (Fst) was modestly down-regulated but only in estrous morning S4cKO females (Fig. 4A and Supplemental Fig. S2H). Expression of the canonical activin receptor type IB and IIA (Acvr1b and Acvr2a) or the SMAD cofactor Foxl2 did not differ between genotypes (Fig. 4A and Supplemental Fig. S2H). Collectively, these results strongly suggest that FSH deficiency in S4cKO mice derives primarily from impaired Fshb mRNA expression.

Figure 3.

Impaired FSH synthesis in S4cKO mice. A) Serum FSH (i, ii) and LH (iii, iv) in 10-wk-old male (i, iii) and > 10-wk-old female (ii, iv; assessed on estrous morning) control and S4cKO mice. B) Pituitary FSH content in adult (> 10-wk-old) male S4cKO and control mice. For S4cKO animals, 2 out of 3 samples were below the detection limit of the assay (240 ng/ml). C) Immunofluorescence for LHβ (green) and FSHβ (red) in the pituitaries of 10-wk-old control and S4cKO male (left) and female (right) mice. Scale bars = 25 μm. Individual data points are plotted as circles or squares; means are shown by horizontal lines. n.s., no significant difference. *P < 0.05.

Figure 4.

Fshb deficiency and impaired activin-stimulated Fshb expression in S4cKO pituitaries. A) Expression of selected genes in the pituitaries of 10-wk-old male (top) and > 10-wk-old estrous morning female (bottom) control and S4cKO mice, assessed relative to the expression of the housekeeping gene Rpl19. The animals are the same as in Fig. 3A. Lhb, luteinizing hormone β subunit; Cga, gonadotropin hormones, α subunit; Gnrhr, gonadotropin-releasing hormone receptor; Fst, follistatin; Acvr1b, activin receptor type 1B; Acvr2a, activin receptor type 2A; Foxl2, forkhead box L2. n = 14/group in males; n = 9/group in females. B) Primary pituitary cells prepared from control or S4cKO male (left) or female (right) mice and treated for 24 h with 1 nM activin A or left untreated. Fshb transcripts were assessed by qPCR, relative to the expression of Rpl19. Males: n = 4; females: n = 3 independent experiments. Bars represent means ± sem. Groups/bars with different symbols differ significantly. and n.s., no significant difference. *P < 0.05.

Activin regulation of Fshb expression is impaired in S4cKO mouse pituitaries

The marked reduction of circulating FSH and pituitary Fshb mRNA levels in S4cKO females on the estrous morning suggested that gonadotropes lacking SMAD4 may be impaired in their ability to up-regulate Fshb expression in response to activins or related ligands. To directly address this possibility, we measured activin-stimulated Fshb expression in primary pituitary cultures from control and S4cKO mice. Basal Fshb mRNA levels were dramatically depleted (>98%) in cultures from either male or female S4cKO mice (Fig. 4B). Basal Fshb expression in such cultures is heavily dependent on autocrine/paracrine activin signaling (40, 42). Note that the more profound reduction in Fshb transcripts observed in S4cKO pituitary cultures compared with S4cKO pituitaries in vivo (Fig. 4A) may reflect the absence of GnRH (or additional endocrine signaling) signaling in the former. Interestingly, exogenous activin A was able to stimulate Fshb expression in both control and S4cKO cells, but the absolute magnitude of Fshb induction was much lower in the latter (Fig. 4B). Therefore, the Fshb gene in S4cKO gonadotropes retains some ability to respond to activins, but this is insufficient to produce wild-type levels of Fshb mRNA.

S4F2cKO females are hypogonadal and sterile

The reproductive phenotypes of S4cKO mice are remarkably similar, although not identical, to those of gonadotrope-specific Foxl2 conditional-knockout (F2cKO) animals (28). In both models, females are FSH deficient and subfertile. This contrasts with the phenotype of female Fshb-knockout mice, which completely lack the dimeric hormone and are sterile (43). In both the S4cKO and F2cKO models, incomplete recombination of the floxed alleles could explain their residual FSH production. That said, both models show >90% reductions in the targeted mRNAs. It therefore seems likely that another mechanism accounts for the ability of these mice to produce sufficient FSH to stimulate modest follicle growth. We turned our attention back to the current model of activin-regulated Fshb expression (44), where SMAD4 and FOXL2 form part of the same transcriptional complex binding to a composite SMAD/forkhead cis-element at −115/−107 of the murine Fshb promoter (24). In addition, SMAD4 binds an 8-bp SMAD binding element at −266/−259 independently of FOXL2 (45), whereas FOXL2 binds independently of SMAD4 at a noncanonical forkhead binding element at −350 (21). Therefore, residual Fshb production in the individual Smad4- and Foxl2-knockout models might reflect the fact that the two proteins have both interdependent and independent functions.

To test this hypothesis, we generated S4F2cKO (Smad4^fl/fl;Foxl2^fl/fl;Gnrhr^GRIC/+) mice lacking both Smad4 and Foxl2 selectively in gonadotropes. S4F2cKO females exhibited pale and barely patent vaginas, and smears were usually devoid of cells. Therefore, we could not reliably assess the onset of puberty or estrous cyclicity in these animals. In a 4 mo breeding trial, S4F2cKO females did not produce any pups, and none showed evidence of pregnancy. In contrast, control (Smad4^fl/fl;Foxl2^fl/fl) females showed normal fertility (Fig. 5A). Consistent with their sterility, S4F2cKO females had dramatically reduced ovarian weights and hypoplastic (thread-like) uteri, suggesting lower estrogen tone (Fig. 5B–D). This could not be directly assessed, however, as serum estradiol levels were below the detection limit of the assay in all of the S4F2cKO mice and in most of the controls (measured at metestrus/diestrus; Supplemental Fig. S3A). Ovarian histology revealed the complete absence of corpora lutea and preovulatory follicles and the presence of only a few early antral follicles in young adult S4F2cKO females (10 wk old; Fig. 5E). In 6-mo-old females, which we analyzed after the conclusion of the breeding trial, ovarian tissue was abnormal in S4F2cKO mice, with few immature follicles and evidence of tubulostromal hyperplasia (Supplemental Fig. S3B). Juvenile S4F2cKO females ovulated in response to exogenous gonadotropins, ruling out an ovarian defect (Supplemental Fig. S3C).

Figure 5.

Profound hypogonadism and female sterility in S4F2cKO mice. A) Average litter size delivered by control and S4F2cKO females during a 4 mo breeding trial. S4F2cKO females delivered no litters. B) Representative reproductive tracts from 10-wk-old control and S4F2cKO mice. C, D) Ovarian (C) and uterine (D) weights, expressed as percentage of body weight (which did not differ), in >10-wk-old female control and S4F2cKO mice. E) Representative H&E-stained histological sections from 10-wk-old control and S4F2cKO female ovaries. Scale bars = 0.5 mm. F) Representative testes and seminal vesicles from 10-wk-old control and S4F2cKO males. G, H) Testicular (G) and seminal vesicle (H) weights, expressed as percentage of body weight (which did not differ) in 10-wk-old male control and S4F2cKO mice. Bars represent means ± sem. In some panels, individual data points are plotted as circles or squares; means are shown by horizontal lines. n.s., no significant difference. *P < 0.05.

Male S4F2cKO mice were similarly hypogonadal, with a 50% reduction in testicular mass relative to controls (Fig. 5F, G). This was a greater decrease than observed in S4cKO (Fig. 1B, C) or F2cKO mice (28). Nevertheless, S4F2cKO males were fertile, siring litters of similar size as controls (data not shown). Testes of S4F2cKO mice showed normal histology, although their seminiferous tubules were smaller in diameter compared with controls (Supplemental Fig. S3D). In contrast, their seminal vesicle weights and serum testosterone levels were normal (Fig. 5H and Supplemental Fig. S3E).

S4F2cKO mice are FSH deficient

Consistent with their reproductive phenotypes, S4F2cKO mice had dramatically impaired serum FSH levels, even more pronounced than those observed in S4cKO or F2cKO mice. Values in males and in most females were near the detection limit of the assay (Fig. 6Ai, Aii). Pituitary FSH content, FSHβ immunoreactivity, and Fshb mRNA were barely detectable in S4F2cKO mice (Figs. 6B, C and 7A). Nonetheless, pituitary glands of S4F2cKO and control animals had similar numbers of gonadotropes (Fig. 6D). In control males and females, LHβ⁺ cells were all colabeled with FSHβ (FSHβ⁺), whereas some cells were uniquely FSHβ⁺ (Fig. 6B, D). Strikingly, in both male and female S4F2cKO mice, LHβ⁺ cells showed a very faint FSHβ signal, and many had undetectable FSHβ staining (Fig. 6B, D). By contrast, the number of FSHβ-only cells scattered throughout S4F2cKO pituitaries was normal (Fig. 6B, D). We verified that all the gonadotropes expressed the Cre recombinase enzyme in S4F2cKO animals. Indeed, both the LHβ⁺ and FSHβ⁺ cells were Cre⁺ (Supplemental Fig. S3F, G). Overall, our results indicate that S4F2cKO mice express low levels of Fshb in most gonadotropes, but this is insufficient to maintain reproductive axis activity.

Figure 6.

Profoundly impaired FSH synthesis in S4F2cKO mice. A) Serum FSH (i, ii) and LH (iii, iv) in 10-wk-old male (i, iii) and adult >10-wk-old female (ii, iv) control and S4F2cKO mice. B) Immunofluorescence for LHβ (green) and FSHβ (red) in pituitaries of 10-wk-old control and S4F2cKO male (left) and female (right) mice. Scale bars = 25 μm. C) Pituitary FSH content in adult (>10-wk-old) male S4F2cKO and control mice. For S4F2cKO animals, all the samples were below the detection limit of the assay (240 ng/ml). D) Quantification of the number of cells immunoreactive for both LHβ and FSHβ (LHβ⁺/FSHβ⁺), only LHβ (LHβ⁺/FSHβ⁻) or only FSHβ (LHβ⁻/FSHβ⁺) in control and S4F2cKO male pituitaries (n=3 mice/genotype). Bars represent means ± sem. In some panels, individual data points are plotted as circles or squares; means are shown by horizontal lines. n.s., no significant difference. *P < 0.05.

Figure 7.

Fshb deficiency in S4F2cKO pituitaries and loss of activin-stimulated Fshb expression on acute deletion of Smad4 and Foxl2. A) Expression of selected genes in the pituitaries of 10-wk-old male (top) and >10-wk-old female (bottom) control and S4F2cKO mice assessed relative to Rpl19; n = 6/group in males; n = 9 and 7 for control and S4F2cKO female mice, respectively. B) Primary pituitary cells prepared from Smad4^fl/fl;Foxl2^fl/fl male (left) or female (right) mice, infected with Cre-expressing (Ad-Cre) or control (Ad-GFP) adenoviruses, and treated for 24 h with 1 nM activin A or left untreated. Fshb transcripts were assessed by qPCR, relative to the expression of the housekeeping gene Rpl19; n = 3 independent experiments. Bars represent means ± sem. Groups/bars with different symbols differ significantly. n.s., no significant difference. *P < 0.05.

S4F2cKO mice show gender-specific alterations in LH secretion

In contrast to FSH, serum LH levels were increased >5-fold in female S4F2cKO mice compared with controls (Fig. 6Aiv). However, serum LH was decreased in S4F2cKO males (Fig. 6Aiii). In both sexes, pituitary Lhb mRNA levels were increased by 3- to 5-fold (Fig. 7A). LHβ immunoreactivity appeared normal, if not elevated, in knockout pituitaries (Fig. 6B). Pituitary Cga mRNA levels were robustly depleted in S4F2cKO males, but not females, probably contributing to the sex difference in serum LH levels (Fig. 7A). Interestingly, Gnrhr expression, which was elevated in S4cKO mice, was normal in S4F2cKO animals (both genders). Fst, Acvr1b, and Acvr2a mRNA levels did not differ between genotypes (Fig. 7A). Foxl2, as expected, was significantly reduced (as in ref. 28).

Basal and activin regulated Fshb expression is abolished in Smad4/Foxl2-depleted pituitaries

Above, we showed that Smad4 deletion impaired basal and activin-stimulated Fshb expression in primary pituitary cultures (Fig. 4B). We also observed a similar, albeit milder, effect on acute deletion of Smad4 (Supplemental Fig. S4A, B) or Foxl2 (28) in primary pituitary cultures from Smad4^fl/f or Foxl2^fl/fl mice infected with Ad-Cre. Therefore, we tested the hypothesis that the residual activin response was FOXL2 or SMAD4 dependent, respectively, by examining the effects of acute recombination of both Smad4 and Foxl2 conditional alleles in pituitary cultures from Smad4^fl/fl;Foxl2^fl/fl mice. In Ad-Cre-infected cultures, we observed a dramatic decrease in basal Fshb transcription and the complete loss of activin responsiveness (Fig. 7B and Supplemental Fig. S4C, D). These results indicate that SMAD4 and FOXL2 are required for basal and activin-stimulated Fshb expression in adult gonadotropes. In Ad-Cre-infected cultures, Lhb expression was normal, whereas Cga was mildly down-regulated in males but not females (Supplemental Fig. S4C, D). This suggests that changes in Cga but not Lhb expression in S4F2cKO pituitaries (Fig. 7A) may be partially explained by cell-autonomous mechanisms. In contrast, the increases in Lhb may represent endocrine (indirect) effects.

DISCUSSION

Loss-of-function mutations in the FSHB/Fshb gene cause hypogonadotropic hypogonadism in humans and mice, demonstrating an essential role for FSH in reproductive development and fertility (6, 7, 43). GnRH and activins are regarded as the primary stimulators of FSH synthesis in mammals; yet, their relative roles and mechanisms of action are poorly understood, particularly in vivo. Here, we show that the canonical mediator of TGFβ superfamily signaling, SMAD4, is absolutely required for normal FSH synthesis in both male and female mice. Moreover, when the Smad4 gene is ablated in combination with the gene encoding its most thoroughly characterized DNA binding cofactor in gonadotrope cells, Foxl2, mice make essentially no FSH and females are sterile. The phenotype of these animals is remarkably similar to that of Fshb-knockout mice. Not only do these results establish SMAD4 and FOXL2 as essential regulators of Fshb transcription, they suggest that activins (or related TGFβ ligands) are as important as GnRH, if not more so, for FSH synthesis in mice.

Regulation of FSH by GnRH vs. activins or related TGFβ ligands

The relative contributions of GnRH and TGFβ superfamily signaling to the regulation of FSH remain to be precisely defined; however, our results suggest a dominant role for the latter. First, SMAD4 and FOXL2 mediate the effects of activins, but not GnRH, on Fshb transcription in vitro (20, 22). Second, although activins stimulate Gnrhr promoter activity in vitro (46–48), Gnrhr transcripts were elevated or normal in S4cKO or S4F2cKO mice, respectively. Gnrhr expression is similarly normal in F2cKO mice (28). Third, Lhb expression is normal or elevated in S4cKO or S4F2cKO mice, indicating that GnRH signaling via the GnRHR is intact. Collectively, these observations suggest that GnRH cannot compensate for the loss of activin signaling via SMAD4 and FOXL2 to the Fshb promoter.

This said, one cannot discount the critical role for GnRH in normal Fshb transcription and FSH synthesis. This is perhaps best demonstrated in GnRH-deficient hpg mice, which have low circulating FSH levels (13). However, the mechanisms by which GnRH stimulate Fshb transcription remain obscure (14). In fact, GnRH regulation of FSH may be mediated indirectly through activins or related ligands. That is, GnRH pulses modulate the local synthesis of follistatin and activin subunits in the pituitary (49, 50). In addition, GnRH might function principally to regulate Fshb in synergy with activins (51). If GnRH requires underlying activin signaling to stimulate FSH, then the loss of SMAD4 and FOXL2 could impair both activin (directly) and GnRH (indirectly) action. This possibility will only be resolved once mechanisms of GnRH action in vivo are more thoroughly described.

Sex-specific changes in LH synthesis in S4cKO and S4F2cKO mice

In addition to the striking FSH-deficiency phenotype, we observed dysregulation of LH synthesis in S4F2cKO males and females and in S4cKO males. Lhb expression was increased in S4F2cKO mice, probably as a result of reduced sex steroid negative feedback at the pituitary and/or hypothalamic levels (52–54). Indeed, S4F2cKO females had thin uteri and were acyclic, suggestive of low estrogen levels (which were indeed below the detection limit of the assay). In S4F2cKO males, testosterone levels were normal at the single time point measured, but the pulsatile nature of testosterone release may have obscured our ability to detect chronic changes. In S4F2cKO females, elevated Lhb expression was associated with increased serum LH. In contrast, serum LH levels were decreased in S4cKO and S4F2cKO males, despite their normal and elevated Lhb expression, respectively. In both cases, these reduced LH levels may have resulted from a dominant effect of impaired Cga expression, which was only observed in males. Contrary to prevailing dogma, this implies that Cga expression may be rate limiting for LH synthesis, at least in male mice.

The mechanistic basis for the striking sex difference in Cga regulation is at present unclear. In vitro, FOXL2 dose-dependently activates or represses Cga promoter activity depending on the cellular context (55); although pituitary Cga expression is normal in F2cKO mice (28). Direct regulation of Cga by SMADs has not been reported to our knowledge. Interestingly, the Cga promoter is negatively regulated by androgens and directly bound by the androgen, but not estrogen, receptor (56, 57). If the repressive effect of androgens is modulated by FOXL2 and/or SMADs, this could explain why loss of SMAD4 (alone or with FOXL2) affects Cga expression specifically in males, regardless of whether testosterone levels are altered.

Signaling molecules upstream of SMAD4 and FOXL2

While both SMAD4 and FOXL2 can bind the Fshb promoter directly (22–24, 45, 58, 59), neither is a direct target of activin or other TGFβ receptors. In addition, the 2 proteins do not directly interact (23, 60). Therefore, another protein (or proteins) must provide the link between receptor activation and SMAD4/FOXL2 at the level of the Fshb promoter. Although SMAD3, and to a lesser extent SMAD2, are the obvious candidates, we recently showed that SMAD2 and the DNA-binding activity of SMAD3 are dispensable for FSH synthesis and fertility in Smad2/3 conditional-knockout mice (29). We postulated that the C-terminal MH2 domain of SMAD3, which is retained in these mice, can provide the molecular link between activin receptor activation and Fshb expression in the nucleus (Fig. 8i) because it is both phosphorylated by the activin type I receptor and interacts with SMAD4 and FOXL2 (22–24, 60). Indeed, in vitro data show that SMAD4 DNA binding activity is sufficient to confer synergistic activation of Fshb transcription by FOXL2 and DNA-binding-deficient SMAD3 (24). Nonetheless, this and other possibilities, such as TGFβ superfamily signaling through SMADs 1, 5, or 8 (61), require investigation using appropriate knockout models.

Figure 8.

Model of hormonal regulation of Fshb expression in gonadotropes in vivo. Left column: proposed mechanisms of Fshb transcriptional regulation in wild-type mice (i) or mice lacking Smad4 (S4cKO; ii), Foxl2 (F2cKO; iii), or both Smad4 and Foxl2 (S4F2cKO; iv) in gonadotropes. Middle column: Resulting circulating FSH levels across the estrous cycle (1°: GnRH-induced primary FSH surge; 2° activin-induced secondary FSH surge). Right column: Observed fertility outcomes in females of the indicated knockout strains. i) In wild-type mice, activin signaling through SMAD3 (S3) and its obligatory partner SMAD4 (S4) activate Fshb transcription in cooperation with FOXL2 (F2) via a composite SMAD/FOXL2 proximal binding site, an 8 bp SMAD-responsive element, and a distal FOXL2 binding site. Mechanisms of GnRH signaling to the murine Fshb promoter are poorly described, as denoted by “?”. ii) In S4cKO animals, both FOXL2 binding sites can still confer some activin responsiveness, presumably via SMAD3. iii) In F2cKO mice, the activity of the 8 bp SMAD binding element should be preserved. In both ii and iii, this results in a decrease in activin responsiveness of the Fshb promoter, a shallower secondary FSH surge, and reduced fertility. iv) In S4F2cKO mice, activin responsiveness is completely lost. As a result, these mice cannot generate a secondary FSH surge and are sterile. A primary FSH surge should also be absent due to a block of antral follicle growth, resulting in the absence of an estrogen-stimulated GnRH surge.

Independent and cooperative regulation of Fshb by SMAD4 and FOXL2

Because mice lacking both Smad4 and Foxl2 exhibit a more robust phenotype than mice lacking either alone (this study and ref. 28), the data suggest that the two proteins regulate Fshb expression both cooperatively and independently. Indeed, based on in vitro studies, our current model (Fig. 8) predicts that, in the absence of either SMAD4 or FOXL2, some activin signaling to the Fshb promoter can be maintained. That is, without SMAD4, activin-regulated SMAD3 can still activate Fshb transcription in synergy with FOXL2 through at least 2 cis-elements in the proximal murine Fshb promoter (refs. 21–24 and Fig. 8ii). In the absence of FOXL2, in contrast, SMAD complexes can still stimulate Fshb via an 8-bp SMAD binding element (Fig. 8iii and ref. 45). In both scenarios, the residual signaling can maintain FSH levels above a threshold required to drive maturation of some ovarian follicles to the preovulatory stage (Fig. 8ii–iii). In the absence of both SMAD4 and FOXL2, however, all activin responsiveness is lost, rendering animals incapable of generating threshold levels of FSH (Fig. 8iv). Consistent with this idea, the double knockout mice do not cycle and exhibit an arrest in follicle development at the preantral stage.

Although our results clearly demonstrate necessary roles for SMAD4 and FOXL2 in FSH synthesis in mice, it is presently unclear whether the human FSHB gene is similarly regulated. Nonetheless, activins are indirectly implicated in human FSH secretion, as circulating inhibins negatively correlate with FSH levels (62) and a soluble ectodomain of ACVR2A suppresses FSH in women (63). Notably, FOXL2 is expressed in human gonadotropes (64) and the composite cis-element through which SMAD4 and FOXL2 regulate murine Fshb is present in the human FSHB promoter (22). Therefore, it is possible that we have identified a fundamental and conserved mechanism underlying FSH synthesis in all mammals.