SMAD3 Regulates Follicle-stimulating Hormone Synthesis by Pituitary Gonadotrope Cells in Vivo*

Pituitary follicle-stimulating hormone (FSH) is an essential regulator of fertility in females and of quantitatively normal spermatogenesis in males. Pituitary-derived activins are thought to act as major stimulators of FSH synthesis by gonadotrope cells. In vitro, activins signal via SMAD3, SMAD4, and forkhead box L2 (FOXL2) to regulate transcription of the FSHβ subunit gene (Fshb). Consistent with this model, gonadotrope-specific Smad4 or Foxl2 knock-out mice have greatly reduced FSH and are subfertile. The role of SMAD3 in vivo is unresolved; however, residual FSH production in Smad4 conditional knock-out mice may derive from partial compensation by SMAD3 and its ability to bind DNA in the absence of SMAD4. To test this hypothesis and determine the role of SMAD3 in FSH biosynthesis, we generated mice lacking both the SMAD3 DNA binding domain and SMAD4 specifically in gonadotropes. Conditional knock-out females were hypogonadal, acyclic, and sterile and had thread-like uteri; their ovaries lacked antral follicles and corpora lutea. Knock-out males were fertile but had reduced testis weights and epididymal sperm counts. These phenotypes were consistent with those of Fshb knock-out mice. Indeed, pituitary Fshb mRNA levels were nearly undetectable in both male and female knock-outs. In contrast, gonadotropin-releasing hormone receptor mRNA levels were significantly elevated in knock-outs in both sexes. Interestingly, luteinizing hormone production was altered in a sex-specific fashion. Overall, our analyses demonstrate that SMAD3 is required for FSH synthesis in vivo.

Pituitary follicle-stimulating hormone (FSH) 2 is an essential regulator of gonadal function (1,2). FSH acts on ovarian granulosa cells to regulate follicle development and on testicular Sertoli cells to regulate spermatogenesis (1,3,4). In both humans and rodents, mutations in genes regulating FSH synthesis or action cause primary amenorrhea because of the arrest of follicle development at the pre-antral or early antral stage (5)(6)(7)(8)(9). In contrast, the effects of these mutations in males are species-specific. For example, FSH-deficient mice are oligospermic but fertile (1), whereas FSH-deficient men are azoospermic and infertile (10 -12).
As members of the TGF␤ superfamily, activins signal through complexes of serine/threonine kinase receptors and SMAD signaling proteins (28,29). According to current in vitro models, activins stimulate the phosphorylation and nuclear accumulation of the receptor-regulated SMAD protein SMAD3 in gonadotrope cells. In the nucleus, phospho-SMAD3 partners with the ubiquitous co-factor SMAD4 and the cellspecific forkhead box L2 (FOXL2) to promote transcription of the murine Fshb gene (30,31). Specifically, SMAD4 and FOXL2 bind adjacent cis-elements in the proximal Fshb promoter (Fig.  1, top panel, SBE2/FBE2). The two proteins are then linked through their mutual association with the C-terminal Mad homology 2 (MH2) domain of SMAD3 (30,32). The in vitro model further suggests that SMAD3 can regulate murine Fshb with SMAD4 at a canonical SMAD binding element (SBE1) located 144 bp upstream of SBE2/FBE2 (30) and with FOXL2 via a third response element (FBE1) ϳ85 bp upstream of the SBE (33) (Fig.  1, top panel). Importantly, the actions of SMAD3 and SMAD4 at SBE1 appear to be independent of FOXL2, whereas the actions of SMAD3 and FOXL2 at FBE1 are independent of SMAD4 (30,31). Thus, only at SBE2/FBE2 do all three proteins work together, but SMAD3 represents the common mediator of activin signaling at all three cis-elements (30,31).
Consistent with the in vitro model, conditional deletion of either Smad4 or Foxl2 from pituitary gonadotrope cells (hereafter, Smad4 cKO and Foxl2 cKO, respectively) produces profound FSH deficiency in mice in vivo (34,35). Female cKOs are subfertile, producing smaller litters (both models) at reduced frequency (Foxl2 cKO) compared with wild-type littermates (34,35). This contrasts with the more extreme phenotype of Fshb knock-out females, which are infertile (1). However, it is important to note that in both Smad4 and Foxl2 cKOs, females still produce some FSH, which may be sufficient to drive some ovarian follicle development (34,35). This residual FSH production may reflect the fact that there are cis-elements in the Fshb promoter where FOXL2 and SMAD4 can act independently of one another (see above and Fig. 1). Consistent with this idea, ablation of Smad4 and Foxl2 together (hereafter, Smad4/Foxl2 cKO), which would affect protein complex binding to all three cis-elements, results in a more extreme FSH deficiency, as well as infertility in female mice (34).
In apparent contrast with the in vitro model, however, mice with a conditional deletion in the Smad3 gene in gonadotropes (hereafter Smad3 cKO) produce FSH at normal levels and are fertile (36). This was a surprising result because, a priori, these mice would be predicted to exhibit a phenotype comparable with that of Smad4/Foxl2 cKOs. The absence of a phenotype is not explained by compensation via the related receptor-regulated SMAD2, because mice with conditional deletions in both Smad2 and Smad3 in gonadotropes exhibit normal fertility and quantitatively normal FSH levels (36). Instead, the absence of a phenotype may be explained by the observation that the conditional deletion in the Smad3 gene does not remove all Smad3 expression (and hence function) in gonadotropes of these mice (36).
In the above described models, the Cre-lox system was used to remove functionally important exons in both the Smad2 and Smad3 genes in gonadotropes (36). Recombination of the Smad2 gene effectively abolished SMAD2 protein expression. In the case of Smad3, however, the floxed region (exons 2 and 3) encodes only one part of the SMAD3 protein: the N-terminal MH1 domain, which mediates SMAD3 binding to DNA (37). In gonadotropes of Smad3 cKO mice, the Smad3 transcript lacked exons 2 and 3, as predicted, but contained the remaining 7 exons (exons 1 and 4 -9). Moreover, there was an increase in the abundance of this novel mRNA. Although translation is predicted to start in exon 1 and for the protein to be translated out of frame in exon 4, the translation start in exon 1 does not conform to a consensus Kozak sequence (38). In contrast, such a sequence exists in exon 4, allowing the C-terminal half of SMAD3 to be translated in frame. This includes the entirety of the MH2 domain, which is phosphorylated by activin type I receptors, and interacts with SMAD4 and FOXL2 (39,40). In vitro, this truncated protein functions identically to full-length SMAD3, in the presence of SMAD4 (36). If this is also the case in vivo, it could explain the absence of an FSH or fertility phenotype in Smad3 cKO mice. This possibility is further supported by the in vitro model (Fig. 1), which indicates that SMAD3 does bind DNA at the FBE1 or SBE2/FBE2 cis-elements. Unfortunately, these data only indirectly suggest a role for SMAD3 in FSH synthesis in vivo.
To definitively rule in or out a role for SMAD3, we need to specifically and completely ablate SMAD3 protein function (including the MH2 domain) in gonadotropes of mice. To our knowledge, there are no other floxed Smad3 strains currently available that would enable this approach. Still, existing models may allow an assessment of SMAD3 function in FSH synthesis. Specifically, residual FSH production in Smad4 cKO mice may be explained by the ability of SMAD3 to partially compensate FIGURE 1. Models of SMAD signaling to the Fshb promoter in murine gonadotrope cells of wild-type and Smad4 knock-out mice. Top panel, in wild-type mice, activins stimulate the formation of complexes of SMAD proteins and FOXL2, which act via at least three cis-elements in the proximal Fshb promoter. SMADs can bind DNA via SBEs, and FOXL2 binds via FBEs. FOXL2 binds the distal FBE1 site and recruits SMAD3 via protein-protein interaction. Complexes of SMAD3 and SMAD4 can bind SBE1, which contains binding sites for both proteins. More proximally, SMAD4 binds SBE2, FOXL2 binds FBE2, and the two proteins are linked through their mutual association with SMAD3. Bottom panel, in mice lacking SMAD4 in their gonadotropes (cKO), SMAD3/SMAD4 binding to SBE1 is lost. However, SMAD3/FOXL2 binding to FBE1 is intact. In the absence of SMAD4, SMAD3 can bind SBE2, enabling activins to stimulate Fshb production, although at reduced levels relative to wild type. MH1, DNA binding domain; MH2, protein-protein interaction domain.
for the absence of SMAD4 (Fig. 1, bottom panel). In vitro, SMAD3 can bind to SBE2 in the Fshb promoter (30) (Fig. 1). Moreover, SMAD3-FOXL2 can stimulate transcription via SBE2/ FBE2, although less potently than SMAD3-SMAD4-FOXL2 (30). If SMAD3 does in fact partially compensate for the absence of SMAD4 at this element and must bind DNA to do so, mice lacking SMAD4 and SMAD3 DNA binding activity should exhibit more extreme FSH deficiency and fertility phenotypes than mice lacking SMAD4 alone. Here, we tested this hypothesis using the Cre-lox system to simultaneously remove SMAD4 and the SMAD3 MH1 domain in gonadotrope cells.

Results
Truncated SMAD3 Stimulates Fshb Transcription in the Presence but Not the Absence of SMAD4 in Vitro-Mice lacking Smad3 exons 2 and 3 express a truncated form of SMAD3, which lacks the DNA binding MH1 domain (hereafter SMAD3-MH2) (36). We proposed that this protein functions equivalently to full-length SMAD3, but only in the presence of SMAD4, enabling transactivation of the Fshb gene. To further test this idea, we compared the activities of full-length SMAD3 and SMAD3-MH2 in a heterologous reporter assay. We cotransfected SW480.7 cells, a SMAD4-deficient human colon adenocarcinoma-derived cell line (41,42), with a murine Fshb promoter-reporter and the indicated combinations of SMAD3 or SMAD3-MH2, SMAD4, FOXL2, and a constitutively active form of the canonical activin type I receptor (T206D or TD) (43). The latter was used as a surrogate for ligand. Co-transfection of SMAD3 and FOXL2 increased reporter activity, which was significantly stimulated by ALK4TD (Fig. 2). In contrast, the combination of SMAD3-MH2 and FOXL2 failed to turn on the reporter, whether or not ALK4TD was co-transfected. SMAD3 and SMAD3-MH2 were expressed at comparable levels (data not shown). Remarkably, when SMAD4 was introduced, SMAD3 and SMAD3-MH2 were comparable in their abilities to potently stimulate the Fshb reporter in the presence of FOXL2. Thus, as predicted, SMAD3-MH2 can activate the Fshb promoter, but only in the presence of SMAD4 (and FOXL2). These results reinforce the idea that SMAD3-MH2 should be unable to promote Fshb expression in the absence of SMAD4 in vivo.
Smad3 and Smad4 Genes Were Efficiently and Specifically Recombined in Gonadotropes-To ablate the SMAD3-MH1 domain and SMAD4 in gonadotropes, we crossed mice harboring floxed alleles for Smad3 and Smad4 (Fig. 3A) with GRIC mice, which express Cre from the GnRH receptor locus. As anticipated, we observed recombination of the floxed alleles in pituitaries of both male and female cKO mice, as well as in the testes and epididymides of cKO males (Fig. 3B). Note that, in addition to gonadotropes, the GRIC allele is active in the male germ line (44). An analysis of gonadotropes (YFPϩ cells) isolated from these animals revealed highly efficient depletion of Smad3 (Fig. 3C) and Smad4 (Fig. 3E) mRNAs, when examined using primers against the targeted exons. As reported previously (36), there was an overall up-regulation of the Smad3 mRNA levels in gonadotropes of cKO mice, although this transcript lacked the targeted exons (Fig. 3D). Note that Smad3 and Smad4 mRNA expression in non-gonadotrope cells of the pituitary (YFPϪ) was equivalent between genotypes ( Fig. 3, C-E, white bars). The purity of YFPϩ cells was 79.3% for the control and 78.3% for cKO mice, respectively. No YFPϩ cells were observed in the YFPϪ population.
cKO Mice Are FSH-deficient-Serum FSH levels were significantly depleted in adult cKO mice relative to littermate controls (Fig. 4, A and B). Similarly, intrapituitary FSH protein content was profoundly reduced in cKO animals as assessed by RIA (Fig. 4, C and D) or immunofluorescence (Fig. 5). Reductions in FSH protein derived from a robust decrease in pituitary Fshb mRNA expression in cKOs (Fig. 6, A and B).
Male cKO mice also showed diminished serum LH relative to controls (Fig. 4E). In contrast, serum LH was significantly elevated in cKO females (Fig. 4F). Consistent with the sex difference in LH secretion, pituitary Lhb and Cga mRNA levels were modestly decreased in cKO males but markedly increased in cKO females (Fig. 6, A and B). Interestingly, these RNA differences were not reflected at the level of intrapituitary LH protein content (Figs. 4, G and H, and 5). In both sexes, pituitary Gnrhr mRNA expression was up-regulated in cKOs relative to controls (Fig. 6, A and B). Expression of Smad3 or Smad4 did not differ between genotypes of either sex, except for Smad4, which was modestly decreased in cKO males (Fig. 6, A and B). Given the relatively small proportion of gonadotropes in the pituitary (ϳ5-10%) and the ubiquitous expression of SMADs, we did not expect to detect changes in Smad3 or Smad4 in the analysis of the whole gland, although, as described above, recombination in gonadotropes was highly efficient (Fig. 3).
Impaired Fertility in cKO Females and Males-8-week-old male and female control and cKO mice were paired with agematched wild-type C57BL/6 mice for a period of 6 (males) or 4 (females) months. Littermate controls and cKOs were mated on the same day. cKO and control males sired equivalent numbers of litters at similar intervals over 6 months (Fig. 7A); however, cKOs fathered smaller litters (Fig. 7B). Note that two of the six cKO males only sired one litter and were excluded from the analysis. Control females were fertile and produced an average of 4.5 litters and 6.25 pups/litter over the course of 4 months. In contrast, cKO females were uniformly sterile; none produced a single pup or showed any overt evidence of pregnancy during the mating trial (Fig. 7C).
Spermatogenesis Is Impaired in cKO Males-Despite having equivalent body mass relative to littermate controls (supplemental Fig. S1A), cKO males exhibited a 50% reduction in testicular mass (Fig. 8, A and B). Cauda epididymal sperm count was similarly reduced by 50% in cKO males relative to controls (Fig. 8C). In contrast, there were no differences between genotypes in progressive sperm motility (Fig. 8D) or seminal vesicle weights (Fig. 8E). Similar results were seen in control and cKO males at the conclusion of the fertility trials (data not shown). Consistent with the normal seminal vesicle weights, but in apparent contrast with the reduced serum LH, cKO males had statistically normal serum testosterone levels (supplemental Fig. S2). Despite impairments in spermatogenesis, cKO males showed overtly normal testicular histology (Fig.  8F).
cKO Females Are Acyclic and Hypogonadal-We next examined the underlying cause of infertility in cKO females. Adult cKO females exhibited normal body mass (supplemental Fig. S1B) but were profoundly hypogonadal with hypoplastic uteri (Fig. 9, A-C). The latter condition was emblematic of estrogen deficiency, although we did not measure levels of 17␤-estradiol in these animals. In our experience, 17␤-estradiol ELISAs lack sufficient sensitivity to accurately measure the hormone on cycle stages other than proestrus in the mouse. Although cKO and control littermates exhibited vaginal opening at roughly the same postnatal age (Fig. 9D), we were unable to reliably measure estrous cyclicity in cKO females (data not shown). Indeed, as indicated in ovarian tissue sections, cKO females showed an arrest in follicle development at the pre-antral or early antral stage and the complete absence of corpora lutea (Fig. 9E). cKO mice also showed impaired responsiveness (ovulation) to exog- C-E, RT-qPCR analysis of mRNA levels of Smad3 and Smad4 in purified gonadotropes (YFPϩ) and non-gonadotropes (YFPϪ) from adult control (YFP/ϩ;GRIC) and cKO (Smad3 fx/fx ;Smad4 fx/fx ;YFP/ϩ; GRIC) mice. In C and E, primers were directed against the deleted region of the genes/transcripts. In D, primers were directed against non-targeted exons in Smad3.
Basal and Activin A-stimulated Fshb Expression Are Reduced in cKO Pituitaries-To more directly assess the effects of the targeted mutations on activin signaling in gonadotropes, we cultured pituitary cells from male and female cKO and control mice. Basal Fshb mRNA expression in murine pituitary cultures depends on activins or related ligands (45). Here, basal Fshb expression was almost completely abolished in cKO cultures relative to controls (Fig. 10, A and B). Activin A induced significant increases in Fshb expression in gonadotropes of cultures from both control males and females (Fig. 10, A and B). In contrast, activin A modestly stimulated Fshb in cultures from male but not female cKOs (Fig. 10, A and B).

SMAD3 Regulates FSH Synthesis in Vivo-
The data presented here provide the first conclusive evidence that SMAD3 regulates FSH synthesis in vivo. Previously, we reported that Smad3 cKO mice are fertile with normal FSH synthesis (36). However, these animals express a truncated and functional form of SMAD3 (i.e. the C-terminal MH2 domain) that lacks DNA binding activity (i.e. the N-terminal MH1 domain) but can still cooperatively regulate Fshb promoter activity with Individual data points are shown (n ϭ 12 or 14 for control and cKO, respectively) as are the group means (horizontal lines). C and D, pituitary FSH content was assessed by RIA (n ϭ 8 or 5 for control and cKO males; n ϭ 11 or 7 for control and cKO females). All samples from cKO females were below the detection limit of the assay. G and H, pituitary LH content was measured by ELISA. The data were analyzed by Student's t test in each panel. SMAD4 and FOXL2 (Fig. 2). Although we will eventually require a model that removes all SMAD3 function in gonadotropes, here we were able to combine two existing mouse strains to probe the role of SMAD3 in FSH synthesis: the Smad3 mice described above (lacking the MH1 domain) and mice lacking Smad4 in gonadotropes (34).
The latter have reduced FSH and are subfertile (34). We argued that the residual FSH in these animals results from partial compensation by full-length SMAD3 and its ability to bind DNA via its MH1 domain (37) (Fig. 1, bottom panel). If this is true, then mice lacking the SMAD3 MH1 domain should be unable to compensate for the absence of SMAD4. Indeed, as we show here, Fshb mRNA and FSH protein are essentially undetectable in these mice. As a result, females are sterile; males are oligospermic and have small testes. These phenotypes are consistent with those of Fshb knock-outs (1). Our data therefore suggest that SMAD3 regulates FSH synthesis in vivo. They also suggest that the MH2 domain of SMAD3 is more critical for Fshb regulation than its MH1 domain and that SMAD complex binding to the Fshb promoter is mediated principally via SMAD4.
We contend that the SMAD3-MH1 domain plays a lesser role than its MH2 domain in Fshb regulation for at least two reasons. First, mice lacking the SMAD3-MH1 domain produce FSH at seemingly normal levels (36). A role for the SMAD3-MH1 domain in FSH synthesis is only observed when SMAD4 is absent (as demonstrated here). Second, the SMAD3-MH2 domain is phosphorylated by the type I receptor (in response to ligand binding and receptor complex activation) and mediates the protein interactions with both SMAD4 and FOXL2 (32,39,40,46,47). Based on these observations, one could argue that SMAD3 may not need to directly bind the Fshb promoter to regulate its transcription under normal conditions. Instead, SMAD complex binding activity is conferred principally by SMAD4 and FOXL2.
If this is true, the 8-bp SMAD binding element (SBE1 in Fig.  1) in the proximal promoter may not play a significant role in FSH synthesis in vivo, because SMAD3-SMAD4 complex binding to this site requires DNA binding activity of both proteins, at least in vitro (31). The role, if any, of this cis-element in vivo should be assessed, perhaps with the aid of newer gene editing technologies (e.g. CRISPR-Cas9). Interestingly, although this was the first activin-responsive cis-element described in the Fshb promoter (25), it has only been observed in rodents thus far (16). The Fshb promoters of other species, such as sheep and pigs, can be robustly stimulated by activins in the absence of this 8-bp SBE (24,48,49). Moreover, adding this element to the human promoter only modestly increases its activin and SMAD sensitivity (31). It is perhaps unsurprising then that it might not play an essential role in mice. Indeed, even point mutations in this element in in vitro promoter-reporter assays only partly attenuate activin responsiveness (17,30,31).
A Potential Role for SMAD2 in FSH Synthesis?-Our contention that the SMAD3-MH2 domain is essential for FSH synthe-  sis is challenged by earlier findings. To date, four mouse strains with targeted mutations in Smad3 have been described. As we argued previously (36), we would predict that mice in two (50,51), if not three (52), of these strains likely produce Smad3 transcripts that encode the MH2 domain, akin to what we describe in the present model. A fourth strain (hereafter Smad3 ⌬ex8 ), however, lacks exon 8 (of the 9 exon gene), and as a result, the terminal 89 amino acids of the MH2 domain of the protein (53). This includes the critical C-terminal serines that are phosphorylated by type I receptors. FSH has been measured in these mice, but not in great detail and with seemingly conflicting results. Female Smad3 ⌬ex8 mice appear to have increased circulating FSH, which may result from the loss of estrogen and/or inhibin feedback (i.e. these animals also have ovarian defects (54)). In contrast, pituitary Fshb mRNA levels are reduced by ϳ30% in 8-week-old Smad3 ⌬ex8 males (55). Although the basis of the sex difference is not yet clear, the fact remains that these mice lack a functional SMAD3-MH2 domain and can still produce FSH protein or transcript at slightly reduced to elevated levels.
The prediction of our model is that a complete loss of SMAD3 (and in particular the MH2 domain) should abrogate FSH production, as we see in Smad4/Foxl2 cKO mice (34) and the Smad3/4 cKO animals described here. The Smad3 ⌬ex8 mice challenge this idea and raise the possibility that SMAD2 might compensate for the loss of functional SMAD3 in these animals. This is plausible for a few reasons. First, SMAD2 is abundantly expressed in gonadotropes and is phosphorylated in response to activin stimulation, at least in immortalized gonadotropes (22,25). Second, SMAD2 interacts with both SMAD4 and FOXL2 (32,39). The latter interaction appears to be weaker than the association between SMAD3 and FOXL2 but is still measureable. Moreover, the interaction might be enhanced in the absence of SMAD3. Recall that Smad3 ⌬ex8 mice would not express a functional SMAD3-MH2 domain, which is required for the FOXL2 interaction. Third, SMAD2 was previously discounted because it lacks DNA binding activity (36). However, as we discuss above, SMAD3 does not need to bind to DNA to regulate Fshb. Thus, if SMAD2 is indeed partially compensating for the loss of SMAD3 function in Smad3 ⌬ex8 mice, then cross- FIGURE 8. cKO males are oligospermic. A, testes and seminal vesicles from adult control and cKO males. B, testicular weights of 10-week-old control and cKO males (n ϭ 13/genotype). C, sperm counts from snap frozen epididymides of 10-week-old control and cKO males (n ϭ 9/genotype). D, sperm motility assessment from fresh cauda epididymides in adult control and cKO males (n ϭ 8 or 6 for control and cKO, respectively). E, seminal vesicle weights of 10-week-old control and cKO males (n ϭ 13/genotype). F, H&E staining of testicular sections from representative control and cKO males. Scale bars, 100 m. FEBRUARY 10, 2017 • VOLUME 292 • NUMBER 6

JOURNAL OF BIOLOGICAL CHEMISTRY 2307
ing these animals with those lacking SMAD2 in gonadotropes (36) should lead to severe FSH deficiency.
We would still argue that, under normal circumstances, SMAD2 plays little to no role in Fshb regulation. Instead, SMAD3 is likely the primary transducer of type I receptor activation to the Fshb promoter. This concept is supported by the absence of an FSH phenotype in gonadotrope-specific Smad2 knock-out mice (36). Rather, the ability of SMAD2 to regulate Fshb may only be manifested when SMAD3 function is compromised, as in Smad3 ⌬ex8 mice. If this is true, it begs the question of why SMAD2 does not compensate in the Smad3/4 cKO mice described here. The simple answer is likely because it lacks DNA binding activity. That is, in Smad4 cKO mice, SMAD3 can partially compensate because it can bind DNA via its MH1 domain. In Smad3/4 cKO mice, the inability of SMAD2 to bind DNA and its weaker FOXL2 interaction would likely preclude compensation. There is a naturally occurring splice variant of SMAD2 that can bind DNA, but it is expressed at far lower levels than the full-length isoform (22,25,56,57).
Shared and Unique Phenotypes of Smad3/4 versus Smad4/ Foxl2 Knock-out Mice-The phenotypes of Smad3/4 cKO mice are highly similar but not identical to those of Smad4/Foxl2  cKO mice (34) ( Table 1). In both models, FSH production is severely impaired, leading to oligospermia in males and sterility in females. Females exhibit a block in folliculogenesis at the pre-antral or early antral stage akin to what was described in Fshb knock-out mice (1). In both models, LH secretion and Lhb mRNA expression are increased in females. Cga mRNA expression and LH secretion are reduced in males of both models. Interestingly, however, Lhb mRNA levels are decreased in Smad3/4 cKO males but elevated in Smad4/Foxl2 cKO males. Lhb mRNA is not altered in males lacking Smad4 alone, suggesting that SMAD3 and FOXL2 play distinct roles in Lhb regulation. At present, there is no evidence that FOXL2 regulates Lhb expression (35). However, SMAD3 can potentiate the stimulatory effects of EGR1 (a mediator of GnRH action) on Lhb promoter activity (58) in vitro, and this effect is dependent upon SMAD binding elements (55). Therefore, the loss of SMAD3 binding to the Lhb promoter in Smad3/4 cKO males may contribute to the reduction in Lhb mRNA expression. Unfortunately, we did not measure Lhb mRNA in mice lacking SMAD3-MH1 (36), so it is unclear whether the down-regulation depends on coordinated loss of both SMAD3 and SMAD4 DNA binding activity. Moreover, it is not yet clear why Lhb levels are reduced in Smad3/4 cKO males but elevated in females, although this may reflect sex-specific endocrine effects.
Another major difference between Smad3/4 and Smad4/ Foxl2 cKO models is in pituitary expression of Gnrhr, which is elevated in the former but unchanged in the latter. The Gnrhr is similarly up-regulated in pituitaries of mice lacking Smad4 alone (34) but is unaltered in mice lacking Foxl2 alone (35). These data suggest that the loss of Smad4 likely underlies the increase in Gnrhr in Smad3/4 cKO mice, although we cannot rule out the effects of the loss of SMAD3 DNA binding activity, because we did not measure Gnrhr in mice with the isolated Smad3 deletion (36). Interestingly, Gnrhr up-regulation is not observed in pituitaries cultured from Smad3/4 cKO males or females relative to controls (supplemental Fig. S3, E and F). This suggests that endocrine factors (e.g. GnRH) contribute to the increase in gene expression observed in vivo. As such, neither SMAD4 nor the SMAD3-MH2 domain appears to regulate Gnrhr expression directly.
In conclusion, the data presented here indicate that: 1) SMAD3 regulates FSH synthesis by gonadotrope cells in vivo, 2) FSH production in Smad4-deficient mice is partly rescued by full-length SMAD3, 3) SMAD4 DNA binding activity plays a more essential role in transcriptional regulation of Fshb than does SMAD3 DNA binding activity, 4) SMAD3/SMAD4 bind-ing to the 8-bp SBE at Ϫ266/Ϫ259 of the murine Fshb promoter (SBE1) may be dispensable for transcription regulation of the gene, and 5) SMAD3 and SMAD4 play essential roles in Fshb expression in vivo. The gene deletions do not appear to markedly perturb gonadotrope cell development because expression of other gonadotrope-specific genes such as Gnrhr, Lhb, and Cga is largely intact. Because SMAD3 and SMAD4 are mediators of TGF␤ superfamily signaling and have not previously been implicated in GnRH action, these data indicate that activins play a dominant role over GnRH in the regulation of FSH and/or that GnRH action is somehow dependent upon a functional activin signaling cascade.

Pituitary gene expression profiles in different conditional knockout models
The following abbreviations and symbols are used in the table: ND, not determined; 1, increased in cKO relative to control (more arrows indicate larger effects); 2, decreased in cKO relative to control; ϭ, equivalent between cKO and control; ϩϩϩ, comparable litter sizes and frequency to controls; ?, males reproduce, but their fertility was not systematically characterized; ϩϩ, smaller litters than controls; ϩ, smaller litters and at lower frequency than controls; Ϫ, sterile.

Fshb
Lhb Cga Gnrhr Fertility   Male  Female  Male  Female  Male  Female  Male  Female  Male  Female Fshb

JOURNAL OF BIOLOGICAL CHEMISTRY 2309
LH reference (AFP5306A) and rabbit anti-LH (AFP240580Rb) used in the LH ELISA were provided by Dr. Parlow (National Hormone & Peptide Program). Monoclonal anti-bovine LH mAb (51887) used in the LH ELISA was provided by Dr. Roser (University of California-Los Angeles). Donkey anti-rabbit HRP (P0448) was purchased from Dako Canada ULC (Mississauga, Canada).
Constructs-The full-length and SMAD3-MH2 expression constructs were described previously (36). To circumvent expression differences between full-length and truncated SMAD3 constructs, we introduced a mutation in the start codon in exon 1 of the latter to bias translation initiation toward the AUG in exon 4. We used the QuikChange site-directed mutagenesis protocol and the following primer set on the MH2 expression vector (forward, CTT CGG TGC CAG CCT TGT CGT CCA TCC TG; reverse, CAG GAT GGA CGA CAA GGC TGG CAC CGA AG). The constitutively active ALK4 expression vector (T206D) and Ϫ1990/ϩ1 mFshb-luc reporter were described previously (22,43).
Cell Culture, Transfection, and Promoter-Reporter Assays-SW480.7 cells (human SMAD4-deficient colon carcinoma; gift from Dr. Fang Liu, Rutgers University, NJ) (41,42) were cultured in DMEM with 10% FBS (v/v) at 37°C and 5% CO 2 . For promoter-reporter experiments, the cells were seeded in 48-well plates at a density of 70,000 cells/well. Approximately 24 h after seeding, the cells were transfected with 225 ng/well of the Ϫ1990/ϩ1 murine Fshb-luc reporter plasmid, along with 4.15 ng of FOXL2 and/or 25 ng of the indicated SMAD expression constructs using polyethylenimine in serum-free medium for 2 h. Where indicated, ALK4-TD was co-transfected at 10 ng/well. After transfection, serum-free medium was changed to complete medium. 48 h later, the cells were washed with PBS. The lysates were prepared (28,59), and luciferase assays were performed as described (60) on an Orion II microplate luminometer (Berthold detection systems, Oak Ridge, TN). All conditions were performed in duplicate wells in three independent experiments.
FACS-Cell sorting was performed at the Cell Vision Core Facility for Flow Cytometry and Single Cell Analysis of the Life Science Complex at the Rosalind & Morris Goodman Cancer Research Centre. Pituitary cell suspensions were prepared from 14 GRIC/ϩ;Smad3 fx/fx ;Smad4 fx/fx ;ROSA26 YFP/ϩ male and female animals as described previously (35), and single cell suspensions were prepared in PBS. Fourteen GRIC/ϩ; ROSA26 YFP/ϩ male and female mice were used as controls. The cells were then passed through a 70-m nozzle at 70 p.s.i. into a Becton Dickinson FACSAria (software version 8.0.1). The pituitary cells were then run and gated to sort YFPϩ (gonadotropes) and YFPϪ (non-gonadotropes) cells from both control and experimental pituitaries. We obtained 1.7 ϫ 10 4 YFPϩ and 7.1 ϫ 10 5 YFPϪ cells for the controls; 1.4 ϫ 10 4 YFPϩ and 6.8 ϫ 10 5 YFPϪ cells for the experimental animals. Sorting purity was assessed by culturing 5 l of both YFPϩ and YFPϪ cells from control and cKO pituitaries in one well of a 96-well plate with M199 medium. 12 h later, the YFPϩ/Ϫ ratio was calculated in each group. Total RNA was extracted from YFPϩ and YFPϪ cells using a total RNA mini kit according to the manufacturer's

Genotyping
Smad3 instructions. Pituitary selective gene expression was assessed by RT-qPCR. Hormone Analyses-Blood was collected by cardiac puncture and allowed to clot at room temperature for 20 min. Serum was obtained following centrifugation at 3000 ϫ g and stored at Ϫ20°C until analysis. To measure pituitary FSH and LH content, pituitaries were isolated and homogenized in 300 l of PBS, sonicated, and centrifuged for 10 min at 16,000 ϫ g at 4°C. Supernatant was removed and kept at Ϫ20°C until assayed. 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), except for pituitary LH content (below). Serum FSH and LH levels were measured by multiplex ELISA. The reportable ranges were 2.40 -300 and 0.24 -30 ng/ml for FSH and LH, respectively. Intra-assay CVs were Ͻ5.0%. Testosterone was measured by ELISA. The reportable range was 10 -1600 ng/ml, and the intra-assay CV was Ͼ20%. For pituitary FSH content, all samples were diluted 50 times and then measured by radioimmunoassay. The reportable range was 2.9 -40 ng/ml. For pituitary LH content, all homogenized samples were diluted 10 6 times in 0.05% PBST supplemented with 0.2% BSA and measured with an in-house sandwich ELISA as described previously (65).
Immunofluorescence-10 -12-week-old male and female animals were anesthetized with 200 mg/kg avertin by i.p. injection followed by an overdose of isoflurane through a nose cone. The animals were transcardially perfused with PBS followed by freshly made 4% PFA in PBS. Pituitaries were isolated and postfixed in 4% PFA for 3 h at room temperature. After a brief rinse with PBS, pituitaries were immersed in 30% sucrose overnight at 4°C and then embedded in OCT on dry ice. Pituitaries were sectioned (5 m) on a cryostat and mounted on prechilled Micro slides. The slides were then warmed up to room temperature, and sections were blocked in 1.5% donkey serum in PBS for 1 h. The sections were washed three times in 0.2% PBST and then incubated with rabbit anti-rat FSH␤ (1:500) and goat anti-LH␤ (1:500) overnight at room temperature. After three washes with PBST, the sections were incubated with Alexa Fluor 594conjugated donkey anti-rabbit (1:600) and Alexa Fluor 488conjugated donkey anti-goat (1:600) secondary antibodies for 1 h at room temperature. After another three washes with PBST, coverslips were mounted with ProLong Gold antifade reagent with DAPI and dried at 37°C for 15 min in the dark. Images were acquired with an Axiocam 506 color camera using Zeiss blue edition (Zen 2 lite) software on a Zeiss Axio Imager M2 microscope.
RT-qPCR-Pituitaries from 10-week-old mice were extracted and immediately frozen in liquid nitrogen and stored at Ϫ80°C. Control females were euthanized at 7 a.m. on the morning of estrus (as assessed by vaginal cytology). cKO females, which did not cycle, were euthanized the same day as control littermates. Samples were homogenized in 500 l of TRIzol reagent, and the total RNA was extracted according to the manufacturer's protocol. RNA concentration was determined by NanoDrop. Two g of RNA per sample were then reverse transcribed into cDNA using Moloney murine leukemia virus reverse transcriptase and random hexamer primers at a final volume of 40 l. Two l of cDNA were then used as template for qPCR analysis as described previously (22) on a Corbett Rotorgene 600 instrument (Corbett Life Science) using EvaGreen reagent and primers listed in Table 2. mRNA levels of target genes were determined using the 2 Ϫ⌬⌬Ct method. Ribosomal protein L19 (Rpl19) was used as a control for normalization.
Puberty Onset, Estrous Cyclicity, and Fertility Assessment-To assess puberty onset, the females were monitored daily for vaginal opening starting on postnatal day 24 (66). Starting at 8 weeks of age, estrous cyclicity was assessed daily in the afternoon (ϳ3 p.m.) for 2-3 weeks by collecting vaginal cells with a cotton swab wet with sterile saline. The cells were smeared on a glass slide, stained with 0.1% methyl blue, and examined by light microscopy. Staging was assessed according to published guidelines (66). One cycle was defined as the sequential appearance of all estrous cycle stages, regardless of the number of days spent in each stage. To assess fertility, 8-week-old males (n ϭ 6/genotype) and females (n ϭ 4/genotype) were paired with age-matched, wild-type C57BL/6 mice (Charles River) for 6 or 4 months, respectively. Breeding cages were monitored daily, and the date of birth and number of newborn pups were recorded. The pups were removed from the breeding couple at postnatal day 15. At the end of the mating trials, control and cKO animals were dissected; blood samples were collected as described above. Pituitaries were isolated, snap frozen, and kept at Ϫ80°C until processing for RNA extraction using TRIzol. One ovary or testis was collected, weighed, and fixed in 10% formalin (ovaries) or Bouin's buffer (testes).
Sperm Counting-Epididymal sperm heads were counted using a hemacytometer as described previously (35). Briefly, the left epididymis of the animal (n ϭ 9/genotype) was immediately snap frozen in liquid nitrogen, weighed, and then homogenized (2 ϫ 15 s, separated by 30 s) on ice in 2 ml 0.9% NaCl, 0.1% Thimerosal, and 0.5% Triton X-100 with a Polytron on power setting 4. The samples were then diluted 20 times in PBS. Ten l of each sample were loaded and counted on each side of a hemacytometer.
Sperm Motility Analysis-Sperm motility of 10 -12-week-old male control (n ϭ 8) and cKO (n ϭ 6) mice was analyzed using computer-assisted sperm analysis, as described previously (67). All materials and tools were prewarmed and kept at 37°C before and during all steps. The left side cauda epididymis of each animal was clamped off and briefly rinsed in prewarmed Hanks' salt M199 medium supplemented with 0.5% (w/v) BSA and adjusted to pH 7.4. The tissue was cut 10 times with finepoint scissors, and sperm were allowed to disperse in fresh M199 medium with BSA for 5 min in a 35-mm Petri dish. The sperm suspension was then diluted four times before loading into a 2X-CEL 80-m deep sperm analysis chamber. Subsequently, 20 l of diluted sperm sample was analyzed using a Hamilton-Thorne IVOS automated semen analyzer (Hamilton-Thorne Biosciences, Beverly, MA). IVOS parameters were set according to the standards determined by The Jackson Laboratory. Two slides were analyzed per sample, and five views from each slide were analyzed.
Reproductive Organ Histology-Ovaries and testes were isolated from 10-week-old control and cKO females and males, respectively, weighed on a precision balance and fixed in 10% SMAD3 Regulates FSH Production in Vivo FEBRUARY 10, 2017 • VOLUME 292 • NUMBER 6 formalin or Bouin's buffer at room temperature. Tissues were then paraffin-embedded and cut into 5-m sections. The sections were stained with H&E. The images were acquired with Leica DFC310 FX 1.4-megapixel digital color camera with a high quality transmitted light microscope (Leica DM1000 LED) using Leica Application Suite version 4.0.0 software.
Superovulation-Superovulation was performed in juvenile (postnatal days 23-28) females as described previously (35). Control and cKO females were injected i.p. with 5 IU of pregnant mare's serum gonadotropin at 5 pm. 48 h later, the mice were treated i.p. with 5 IU of human chorionic gonadotropin. After 14 h, mice were killed, and cumulus-oocyte complexes (COCs) were harvested in PBS from ampullae of the oviduct. COCs were enzymatically dissociated using 0.5 mg/ml hyaluronidase for 10 min at 37°C. Oocyte number was then counted under an inverted microscope.
Primary Pituitary Cultures-Primary pituitary cultures were prepared as previously described (68). Briefly, pituitaries were isolated from adult control and cKO animals, single suspended cells were seeded at a density of 300,000 cells/well in 48-well plates. The cells were allowed to settle for 16 h in M199 culture medium supplement with 10% FBS. The media were then replaced with 2% FBS-containing media in the presence or absence of 1 nM activin A. Note that cultures from males and females were prepared and processed separately. Culture media were removed, the cells were harvested, and RNA was extracted using a total RNA mini kit following the manufacturer's instructions.
Statistical Analysis-All of the data were analyzed by unpaired Student's t tests except pituitary gene expression in the primary culture experiments, which was analyzed by multiple t tests. For the reporter assay, two-way analyses of variance were conducted, followed by Bonferroni-corrected multiple comparison tests. Statistical analyses were performed using GraphPad Prism 6 (GraphPad). Values of p Ͻ 0.05 were considered statistically significant.
Author Contributions-Y. L. and D. J. B. were responsible for the experimental design, data analyses, and manuscript preparation. Y. L. conducted the majority of the experiments. G. S. conducted the promoter-reporter assays and some qPCR experiments and edited the manuscript. U. B., C.-X. D., and J. G. generated the Cre and floxed Smad mouse strains and edited the manuscript. All authors reviewed the results and approved the final version of the manuscript.