A Smad-binding Element in Intron 1 Participates in Activin-dependent Regulation of the Follistatin Gene*

Follistatins exert critical autocrine or paracrine control in many tissues by binding and bio-neutralizing activin and several other transforming growth factor-β ligands. In the pituitary, activin acts locally to induce follistatin expression and thus modulate its own actions. This local feedback loop safeguards against excessive activin signaling and maintains the necessary balance of activin and follistatin tone. To better understand the mechanisms underlying the activation of follistatin by activin A, follistatin transcription was evaluated in gonadotrope-derived αT3-1 cells. Transient transfection experiments established that follistatin-luciferase plasmids that incorporate up to 2.86 kb of the upstream region of the rat follistatin gene are not induced by activin A in αT3-1 cells. On the other hand, plasmids that incorporate intron 1 are responsive to activin A and induced by a constitutively active form of ALK4. These experiments ultimately identified a conserved Smad-binding element (SBE1) in intron 1, between +1791 and +1795. In αT3-1 cells treated with activin A, SBE1 preferentially recruits Smad3, but not Smad2, and mediates Smad3-dependent activation of follistatin transcription. shRNA knockdown of endogenous Smad3 in these cells compromises SBE1-mediated transcription in response to activin A and interferes with its ability to positively regulate follistatin mRNA levels. The findings of the current work illustrate the critical role of intron 1 of the follistatin gene in mediating Smad-dependent effects of activin and regulating the expression level of this gene in some cell types, such as pituitary cells of gonadotrope lineage.

Follistatins exert critical autocrine or paracrine control in many tissues by binding and bio-neutralizing activin and several other transforming growth factor-␤ ligands. In the pituitary, activin acts locally to induce follistatin expression and thus modulate its own actions. This local feedback loop safeguards against excessive activin signaling and maintains the necessary balance of activin and follistatin tone. To better understand the mechanisms underlying the activation of follistatin by activin A, follistatin transcription was evaluated in gonadotrope-derived ␣T3-1 cells. Transient transfection experiments established that follistatin-luciferase plasmids that incorporate up to 2.86 kb of the upstream region of the rat follistatin gene are not induced by activin A in ␣T3-1 cells. On the other hand, plasmids that incorporate intron 1 are responsive to activin A and induced by a constitutively active form of ALK4. These experiments ultimately identified a conserved Smad-binding element (SBE1) in intron 1, between ؉1791 and ؉1795. In ␣T3-1 cells treated with activin A, SBE1 preferentially recruits Smad3, but not Smad2, and mediates Smad3-dependent activation of follistatin transcription. shRNA knockdown of endogenous Smad3 in these cells compromises SBE1-mediated transcription in response to activin A and interferes with its ability to positively regulate follistatin mRNA levels. The findings of the current work illustrate the critical role of intron 1 of the follistatin gene in mediating Smad-dependent effects of activin and regulating the expression level of this gene in some cell types, such as pituitary cells of gonadotrope lineage.
Activins are members of the evolutionarily conserved transforming growth factor-␤ (TGF-␤) 3 superfamily of factors implicated in the control of a wide array of cellular processes of embryonic and adult tissues (1)(2)(3). Activin A and activin B are homodimers of inhibin ␤A and ␤B subunits, respectively, with established roles in the control of many cell types including those throughout the reproductive axis (1, 4 -6). The inhibin ␤A and ␤B subunits are both expressed in the pituitary and activin B arising from gonadotropes is a critical component of the network of autocrine or paracrine factors of this tissue (7). By exerting control on FSH␤ expression, activin B is hypothesized to locally provide a positive signal for the differential production of FSH over LH and to thereby facilitate the cyclic fluctuations of these hormones during the estrus cycle (7,8). The actions of activins are strictly controlled by the concerted actions and the preferential usage of several extracellular modulators (9,10). Of these, inhibin and follistatin have established roles in the regulation of pituitary gonadotropes (7).
Follistatins are cysteine-rich glycoproteins that bind and bioneutralize activin with high affinity at a 2:1 molar ratio (11)(12)(13). They also bind and inactivate myostatin and some bone morphogenetic proteins with varying degrees of affinity (14). Structural studies of the follistatin-activin complex have elucidated the basis for this interaction and provided clues about the determinants of follistatin binding to other ligands (15,16). The follistatin gene comprising 6 exons produces two alternatively spliced mRNA transcripts and the corresponding proteins of 315 or 288 amino acids (FS315 or FS288) (17,18). The two isoforms, FS315 and FS288, are differentially expressed in most tissues (19,20). The FS315 form is the predominant circulating form, whereas the shorter FS288 form lacking the C-terminal tail associates tightly with heparan sulfate chains of cell surface proteoglycans and is presumed to be the form that acts locally to modulate activin signaling (2,(21)(22)(23). This local function of follistatin is, in turn, controlled by the actions of activin. In cultured rat anterior pituitary cells, activin increases steadystate follistatin mRNA and secreted protein levels (24 -28). Given that inhibin antagonizes the activin-induced rise in pituitary follistatin, this pinpoints gonadotropes as the primary targets of this action of activin (28,29).
Activin signals by sequentially binding to activin-specific Type II (ActRII or ActRIIB) then Type I (ALK4) serine/threonine kinase receptors to form a multimeric complex (3,30,31). In this complex, the Type II receptors trans-phosphorylate ALK4 and allow it to transiently interact and phosphorylate the downstream signaling proteins, Smad2 and Smad3, at their C-terminal SSXS motifs (31)(32)(33). The phosphorylated forms of Smad2 and Smad3 in turn translocate to the nucleus where they bind to DNA targets and regulate transcription in association with the common Smad4/DPC4 and other partners (34).
The mechanism by which activin induces follistatin expression is not well characterized. A luciferase reporter construct that incorporates the Ϫ757/ϩ136 fragment of the rat follistatin gene is activated by a combination of cAMP and TPA (35,36). In L␤T2 cells, the same fragment is a target of GnRH (37) but not of activin A in ␣T3-1 cells (38). On the other hand, a luciferase reporter construct designated rFS(rin3)-luc that incorporates ϳ2.8 kb of the 5Ј-flanking sequence and extends into the third intron of the rat follistatin gene is induced by activin in ␣T3-1 cells (39). Finally, it has been reported that the human follistatin promoter is activated in HepG2 cells by Smad signaling, downstream of activin or TGF-␤ (40,41). The present study was undertaken to further evaluate the mechanism involved in the transcriptional activation of the follistatin gene by activin and to identify activin-responsive regulatory elements. The mouse ␣T3-1 gonadotrope cell line, which expresses endogenous follistatin, was used as a cellular model. The results indicate that activin induction of the follistatin gene is mediated by a conserved Smad-binding element that localizes to the first intron.

EXPERIMENTAL PROCEDURES
Primary Cells and Cell Lines-Primary rat anterior pituitary (RAP) cells were prepared by collagenase-mediated dispersion of anterior pituitaries obtained from male Sprague-Dawley rats (180 -200 g) as previously described (24). The dispersed RAP cells were seeded on tissue culture plates and maintained at 7.5% CO 2 in a humidified 37°C incubator in a specially formulated medium (designated ␤PJ) containing appropriate growth factors and 2% fetal bovine serum (FBS) (24). The cells were allowed to recover for at least 3 days before initiating experiments. The mouse gonadotrope-derived ␣T3-1 (42) and the human embryonic kidney (HEK) 293T cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% FBS and 2 mM glutamine.
Analysis of Follistatin Transcript Levels by RNase Protection Assays and Real-time PCR-On the 4th day following dispersion, RAP cells (10 7 per 10-cm tissue culture dish) were washed and equilibrated overnight in with 0.2% FBS in ␤PJ medium, then washed again and treated in fresh medium with activin A or vehicle. Total RNA from either the nuclear or cytoplasmic compartment of 10 7 RAP cells was isolated by lysis in ice-cold Nonidet P-40 buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl 2 , 0.5% Nonidet P-40). The nuclei were collected by a 10-min centrifugation at 1000 ϫ g, then subjected to sequential digestion with RNase-free DNase (Promega, Madison, WI) for 10 min at 37°C and proteinase K (EM Science, Gibbstown, NJ) for 30 min at 42°C. The cytoplasmic fractions were treated for 1 h at 42°C with 100 g/ml proteinase K. The RNA recovered from each compartment was used in its entirety for each hybridization reaction to measure follistatin transcripts. In the case of ␣T3-1 cells, total RNA was extracted with the RNeasy kit (Qiagen, Hilden, Germany), and ϳ50 g was used to evaluate follistatin transcript levels. The rat and mouse follistatin cDNA templates were constructed by subcloning into pBluescript IISK (Stratagene, La Jolla, CA) the corresponding genomic fragments spanning the junction of exon 3 and intron 3. Antisense riboprobes corresponding to rat or mouse follistatin were synthesized in the presence of [␣-32 P]UTP (3000 Ci/mmol) using T3 or T7 RNA polymerase from templates linearized with either HindIII or BamHI, respectively. The rat fol-listatin riboprobe protects 463 or 203 nt corresponding to primary (unprocessed) or mature mRNA transcript, respectively. The mouse follistatin riboprobe protects 470 nt of primary transcript or 190 nt of mature mRNA. Rat and mouse glyceraldehyde-3-phosphate dehydrogenase (GAPDH) antisense riboprobes were synthesized using T3 RNA polymerase to yield a protected fragment of 136 nt, as previously described (28). The samples were resolved on 5% polyacrylamide, 8 M urea gels, and band intensity was quantified using the PhosphorImager system (Molecular Dynamics, Sunnyvale, CA) and the Image-Quant 4.0 software package. Follistatin transcript levels were normalized to internal GAPDH levels and are reported as means ϩ S.E.
Plasmid Constructs-Expression plasmids encoding N-terminally Myc-tagged human (h) Smad2, -3, or -4 and the constitutively active ALK4 (caALK4) were generated by PCR and subcloned into the pCS2ϩ expression vector. The numeric designations of the rat genomic fragments used to describe the follistatin-luciferase plasmids are relative to the most downstream transcription initiation site observed in ␣T3-1 cells 4 and previously designated as "␣" (35). The rFS(2.9)-luc reporter was constructed by subcloning the KpnI/EcoNI (Ϫ2864/ϩ136) fragment of the rat follistatin gene into the pGL2 basic luciferase vector (Promega). The rFS(2.9i)-luc plasmid contains the entire first intron (ϩ227/ϩ2097) of the rat follistatin gene just downstream of the Ϫ2864/ϩ136 fragment. To exclude the ATG initiation site from this construct, a 90-bp fragment overlapping exon 1 was deleted by replacing the Aat II/Blp I (Ϫ3/ ϩ758) fragment with two PCR-amplified fragments corresponding to Ϫ3/ϩ136 (forward and reverse primers: 5Ј-TCTAGATTTAAAGC and 5Ј-ACTAGTGGCAGGCGCG-GGGCGAGGA) and ϩ227/ϩ758 (forward and reverse primers: 5Ј-ACTAGTGAGTGGAGGGGATGCGCCCA and 5Ј-ACTTCGGGCTAATAATTTGGTTTG) that were ligated via an engineered SpeI site. The rFS(0.3i)-luc and rFS(0.3iP)-luc plasmids were generated by digesting rFS(2.9i)-luc at the unique Mlu I site to remove the portion upstream of Ϫ312 or at the PmlI site to delete 313 bp from the 3Ј-end of intron 1. The rFS(0.3i)-luc plasmid was further modified using the Erase-A-Base System (Promega) to serially truncate the 3Ј-end of intron 1 and generate rFS(0.3i45)-luc, rFS(0.3i91)luc and rFS(0.3i115)-luc (described in Fig. 4). The rFS(0.3i45)-luc plasmid was further digested with SpeI and Pml I to remove most of intron 1 (ϩ228/ϩ1784) and generate rFS(0.3ex45)-luc. Point mutations within the SBE1 site of the rFS(0.3ex45)-luc plasmid were introduced using a PCR approach with upstream primers containing the indicated substitutions (described in Fig. 4) and a reverse primer within the pGL2 vector. All constructs were subjected to sequence analysis.
Transfections and Luciferase Reporter Assays-The ␣T3-1 cells were seeded in poly-L-lysine-coated 12-well tissue culture plates at a density of 3 ϫ 10 5 cells/well in 2 ml of complete medium (Dulbecco's modified Eagle's medium, 10% FBS, and 2 mM glutamine). The cells were transfected the next morning by incubating them for 6 h with a mix of the Superfect Transfection Reagent (Qiagen; Hilden, Germany), 0.6 g/well of luciferase reporter plasmid and 0.2 g/well cytomegalovirus (CMV)-␤-galactosidase (␤-Gal) plasmid as an internal control. Where indicated, varying amounts of expression plasmids encoding Myc-tagged Smads, caALK4, or empty vector were co-transfected along with the reporters. At the end of the 6-h transfection period, the cells were washed and treated with vehicle or activin A in Dulbecco's modified Eagle's medium supplemented with 2% FBS and 2 mM glutamine. The cells were harvested 15 h later in lysis buffer (1% Triton X-100, 25 mM glycylglycine, pH 7.8, 15 mM MgSO 4 , 4 mM EGTA, and 1 mM DTT). Luciferase reporter activity was measured using D-luciferin luciferase substrate (Biosynth, Naperville, IL) with a Lumimark microplate luminometer (Bio-Rad) or a Lumat LB 9507 (EG&G Berthold, Bad Wildbad, Germany) and normalized to that of CMV-␤-Gal. Reported data correspond to luciferase/␤-Gal ratios of each plasmid relative to the activity of the pGL2 basic vector.
Chromatin Immunoprecipitation (ChIP)-The method used for ChIP analysis was essentially as described previously (43). Activin A or vehicle-treated ␣T3-1 cells were cross-linked with 1% formaldehyde for 15 min at room temperature. The cells were lysed by incubating them for 10 -15 min on ice in lysis buffer (25 mM HEPES, pH 7.8, 1.5 mM MgCl 2 , 10 mM KCl, 0.1% Nonidet P40, 1 mM DTT, and protease inhibitors). The nuclear fraction that was recovered by centrifugation (5 min at 5000 ϫ g) was resuspended in ChIP buffer (50 mM HEPES, pH 7.8, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, and protease inhibitors) and sonicated on ice (Misonix XL200 ultrasonic cell disruptor) to achieve an average chromatin length of 500 -1000 bp. The sonicated samples were precleared by incubation with protein A-Sepharose in the presence of 8 g/ml salmon sperm DNA, 0.3% normal rabbit serum, and 0.05% bovine serum albumin followed by centrifugation. The material recovered from the equivalent of ϳ10 7 ␣T3-1 cells was incubated overnight at 4°C with 5 l of either normal rabbit IgG, anti-hSmad2/3, or anti-hSmad2/3 preabsorbed with the peptide antigen along with protein A-Sepharose. The protein A-purified rabbit anti-hSmad2/3 used for these experiments is directed against a peptide within the linker of hSmad2 (amino acids 199 -215), which is conserved in hSmad3. The protein A-Sepharose beads were washed sequentially once with ChIP buffer, twice with ChIP buffer containing 0.5 M NaCl, once with 0.25 M LiCl buffer (20 mM Tris-HCl, pH 8.0, 0.5% Nonidet P-40, 0.5% deoxycholate, 1 mM EDTA), and finally, twice with TE buffer (10 mM Tris-HCl, pH 8, 1 mM EDTA). The specifically bound complexes were eluted from the protein A-Sepharose beads by two 15-min incubations at 65°C with TE elution buffer (10 mM Tris-HCl, pH 8, 1 mM EDTA, 1% SDS). The immunoprecipitated complexes and the starting material (input) were incubated overnight at 65°C to reverse cross-linking, then treated with proteinase K and purified using QIAquick Spin Columns (Qiagen, Hilden, Germany). The DNA samples were recovered in 50 l of 10 mM Tris-HCl, pH 8.5, and analyzed by semiquantitative PCR using primers that amplify a fragment of 185 bp overlapping the SBE1 site within intron 1 of the endogenous mouse follistatin gene (forward: 5Ј-GTCGCTGCAGGTTATGAAATGG and reverse: 5Ј-AAAG-GGGAGAGTGGGGAAGGAC). The amplified fragment was then resolved on a 2% agarose gel and analyzed by ethidium bromide staining. Alternatively, DNA from 1-2 l of each sample was quantified by real-time PCR using the SYBR GREEN PCR Master Mix and the ABI PRISM 7700 Sequence Detector (Perkin-Elmer Applied Biosystems, Foster City, CA). The fragment containing the SBE1 site was amplified using primers flanking the site (forward: 5Ј-AACAGTCTAGTAAAAGTCA-ATGCAAGCT and reverse: 5Ј-TGCGCCCCAGCCATAT). A primer set that amplifies a fragment ϳ5-kb upstream of the transcription start site of the mouse follistatin gene (forward: 5Ј-AGATAGAGATCCCACCACAGAACAA and reverse: 5Ј-GGATGGACTTGGGTGGTATCTGTA) or a fragment of the mouse ␤-actin gene (forward: 5Ј-TTCCCTTCCACAGGGT-GTGA and reverse: 5Ј-ACATAGGAGTCCTTCTGACCC-ATT) were used as internal controls. Primer pairs flanking the upstream putative Smad3 site at Ϫ1604 (forward: 5Ј-CGGCTGTATTTCGGGATCTATT and reverse: 5Ј-ACTG-CAGGAGATAGTGCTAATCTTTTAAT) and Smad4 site at Ϫ895 (forward: 5Ј-GAAAGGGAGAGGGCGAGACT and reverse: 5Ј-CCCTCGGGCTCCACAAGT) were used to amplify the corresponding fragments.
Oligonucleotide Precipitation Assays-For these experiments, a lentiviral delivery system was used to facilitate the expression of Myc-tagged Smads in ␣T3-1. The N-terminally Myc-tagged hSmad2, -3, or -4 cDNAs were subcloned upstream of an IRES GFP marker in the pCSC-SP-PW-IRES/ GFP lentiviral transfer vector (generously provided by Dr. Inder Verma, Salk Institute, La Jolla, CA). The GFP-expressing pCSC-SP-PW-IRES/GFP empty vector was used as control. Recombinant lentiviruses were produced by co-transfecting HEK293T cells with the Smad expression or control transfer plasmid and three additional plasmids required for packaging (pMDL, pRev, and pVSVG) using polyethylenimine as the transfection reagent, as described (44). The supernatants containing the viral particles were collected 48 h after transfection, filtered through a 0.45-m filter, and concentrated by ultracentrifugation for 2.5 h at 50,000 ϫ g. Relative titers were assessed by monitoring the percentage of GFP-positive HEK293T cells infected with serial dilutions of the viral preparations. For oligonucleotide precipitation experiments, ␣T3-1 cells (5 ϫ 10 6 /10 cm dish) were infected in the presence of 8 g/ml polybrene (Sigma) with lentiviral vectors to express only GFP as a control or combinations of Myc-Smad2 and -4 or Myc-Smad3 and -4. The amount of virus necessary to achieve Ͼ90% GFPpositive ␣T3-1 cells was predetermined by performing serial dilutions of equivalent titers of the viral preparations. Protein expression was allowed to progress for 5 days at which time the cells were replated into four 10-cm dishes and allowed to grow for two more days. The cells were supplemented with fresh medium and treated for 30 min with 1 nM activin A or vehicle in duplicate. Lysates of ␣T3-1 cells were prepared by brief sonication in lysis buffer (25 mM Tris-HCl, pH 7.5, 0.1% Triton X-100, 10% glycerol, 1 mM MgCl 2 , 0.5 mM EDTA, 100 mM NaCl, 5 mM NaF, 1 mM Na 4 P 2 O 7 , 1 mM DTT, and protease inhibitors) followed by a 10-min centrifugation at 12,000 ϫ g at 4°C. The supernatant obtained from each 10-cm dish (850 g) was incu-bated for 2 h at 4°C with 1 g of biotinylated double-stranded oligonucleotides, precoupled to streptavidin-agarose beads (Pierce), in the presence of 8 g of poly(dI-dC) (Sigma). The agarose beads were washed three times by centrifugation, and specifically bound proteins were recovered and subjected to Western analysis. The samples were resolved under reducing conditions using 10% NuPAGE SDS gels (Invitrogen) and MOPS as the running buffer then transferred to nitrocellulose membranes. After blocking the membranes with 5% BLOTTO (Pierce), Myc-Smads were detected using an anti-Myc monoclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA) and a horseradish peroxidase-conjugated sheep anti-mouse IgG (Amersham Biosciences, Piscataway, NJ). Immune complexes were then visualized with SuperSignal West Pico chemiluminescence substrate (Pierce). The experiments were performed using biotinylated wild type (forward: 5Ј-CAAGCTG-CACGTGTTGTGTCTGGGTCACTGGTAACTGACATTG-ATATGGCTAGGCGCAGCGGCTGCTGCTC; reverse: 5Јbiotin-GAGCAGCAGCCGCTGCGCCTAGCCATATCAAT-GTCAGTTACCAGTGACCCAGACACAACACGTGCAGC-TTG) and SBE1 mutant (forward 5Ј-CAAGCTGCACGTGTT-GTaatTGGGTCACTGGTA ACTGACATTGATATGGCTA-GGCGCAGCGGCTGCTGCTC; antisense 5Ј-biotin-GAGC-AGCAGCCGCTGCGCCTAGCCATATCAATGTCAGTTA-CCAGTGACCCAattACAACACGTGCAGCTTG) probes corresponding to rat follistatin.
The shRNAs that produced greatest knockdown of the corresponding target (S2#4 and S3#4) were subsequently evaluated for their ability to interfere with or attenuate activin-dependent activation of the rFS(0.3ex45)-luc reporter in ␣T3-1 cells or the induction of endogenous follistatin mRNA. To achieve sustained silencing of Smad2/3 in the majority of cells, ␣T3-1 cells were seeded in 12-or 24-well poly-L-lysine-coated plates (10 5 cells/well), and triplicate wells were transduced with each lentiviral vector under conditions determined to yield Ͼ80% GFPϩ cells, in the presence of 8 g/ml polybrene as described above. Fresh medium was introduced 24 h later, and the cells were cultured for 3 additional days. The cells were subsequently processed for transient transfection experiments, as described above, with rFS(0.3ex45)-luc as the reporter. Alternatively, the cells were supplemented with fresh medium, then treated with either vehicle or activin A for 2 h. These cells were then processed for follistatin mRNA determination. Total RNA was extracted using the RNeasy Micro kit (Qiagen) and reversetranscribed using SuperScript II (Invitrogen). To quantify endogenous mouse mRNA levels, real time PCR was performed on the ABI 7900HT Fast Real-Time PCR System using Power SYBR Green Master Mix (Perkin-Elmer Applied Biosystems). Follistatin transcript levels were quantified using the ⌬⌬C t relative quantification method (48) (forward primer: 5Ј-CCCCA-ACTGCATCCCTTGTA and reverse primer: 5Ј-GGTCCGC-AGTCCACGTTCT) with GAPDH as control (forward primer: 5Ј-GGAAGGGCTCATGACCACAGT and reverse primer: 5Ј-CACAGTCTTCTGAGTGGCAGTGAT).
Reagents-Recombinant human activin A was purified from the conditioned medium of stably transfected CHO cells (generously provided by Dr. Wolfgang Fischer, Salk Institute, La Jolla, CA). The human Smad2 and Smad3 cDNAs (provided by Dr. Rik Derynck, University of California at San Francisco) and the human Smad4/DPC4 cDNA (provided by Dr. Scott Kern, Johns Hopkins University School of Medicine, Baltimore, MD) were modified by adding a Myc tag at the N terminus and subcloned into the CS2ϩ expression vector. The expression plasmid for the constitutively active Type I activin receptor (ALK4(T206D) or caALK4) has been described previously (49).

Activin Increases Follistatin Transcript Levels of Primary Rat
Anterior Pituitary and ␣T3-1 Cells-Activin A causes a dramatic increase in follistatin mRNA levels and follistatin secretion from primary cultures of rat anterior pituitary (RAP) cells, as previously reported (24,28). Several lines of evidence suggest that this reflects the transcriptional activation of the follistatin gene in pituitary gonadotropes (39). To better understand this mechanism and identify a suitable cell line for further studies, activin effects on follistatin mRNA expression were evaluated in greater detail in RAP cells and compared with those of gonadotrope-derived ␣T3-1 cells. In RAP cells, 1 nM activin A caused a rapid and transient increase in primary follistatin transcript levels accompanied by a delayed slower rise of the mature mRNA form (Fig. 1, panel a). Co-treatment with inhibin A antagonized the activin-induced rise in follistatin mRNA levels suggesting that the observed changes in follistatin transcript levels reflect primarily those occurring in inhibin-responsive gonadotropes (Fig. 1, panel b). As in RAP cells, gonadotropederived ␣T3-1 cells responded to activin A with a similar rise in primary and mature follistatin transcript levels that was also antagonized by co-treatment with inhibin A (Fig. 1, panels d  and e). The effect of activin A on follistatin transcript levels of both ␣T3-1 and RAP cells was concentration-dependent over the range tested (Fig. 1, panels c and f). Based upon these observations, the ␣T3-1 cells were used for further evaluation of the follistatin gene.
Regulatory Elements in the First Intron of the Follistatin Gene Are Responsive to Activin-To define the mechanism of transcriptional activation of the follistatin gene in pituitary gonadotropes and localize activin-responsive elements, fragments of the rat follistatin gene were cloned upstream of the firefly luciferase reporter in pGL2 and evaluated for inducibility by activin A in ␣T3-1. First, the region from Ϫ2864 to ϩ136 corresponding to the proximal promoter of the rat follistatin gene was cloned into pGL2 (rFS(2.9)-luc) and evaluated. Although this upstream fragment displayed basal activity in ␣T3-1 cells, it failed to mediate activin A effects (Fig. 2a). Equivalent plasmids that were truncated at the 5Ј-end to either Ϫ752 (data not shown) or Ϫ312 (rFS(0.3)-luc) behaved similarly and remained unresponsive to activin A (Fig. 2a). These observations suggested that activin-responsive elements must reside further upstream or downstream. To evaluate the latter possibility, intron 1 (1870 bp) of the rat follistatin gene was cloned at the 3Ј-end of the Ϫ2864/ϩ136 fragment, and the resulting plasmid (rFS(2.9i)-luc) was evaluated in ␣T3-1 cells. Activin A induced the reporter activity of the rFS(2.9i)-luc plasmid 8-fold indicating that regulatory elements located in intron 1 mediate this effect (Fig. 2a). The removal of the region upstream of Ϫ312 somewhat compromised the magnitude of activin inducibility of rFS(0.3i)-luc relative to (rFS(2.9i)-luc) (Fig. 2a). Nevertheless, both plasmids were activated in ␣T3-1 over the same concentration range of activin A (Fig. 2b). These results suggest that regulatory elements between Ϫ2864 and Ϫ312 do not contribute significantly to establishing sensitivity to activin A. Hence, further analyses were confined to plasmids that contained only 312 bp of the proximal promoter of the rat follistatin gene. Responsiveness to activin A was lost by the removal of a 129-bp fragment (ϩ1784/ϩ1912) toward the 3Ј-end of intron 1, indicating that this fragment participates in mediating activin A effects (Fig. 2a). Indeed this ϩ1784/ϩ1912 fragment corresponding to the 3Ј-end of intron 1 conferred activin A responsiveness to the otherwise unresponsive rFS(0.3)-luc plasmid (compare the activity of rFS(0.3)-luc to rFS(0.3ex45)-luc (Fig.  2a). The pattern of activin-responsiveness of the various plasmids described thus far was recapitulated by co-transfected caALK4 (Fig. 2a). This intronic fragment of the rat follistatin gene, herein referred to as the activin-responsive fragment (ARF), has enhancer-like activity as it mediated activin A effects when two or more copies were placed upstream of the unresponsive rFS(0.3)-luc or the heterologous pGL2-(SV40)-promoter plasmid (Promega) (Fig. 2c). A series of 3Ј truncations within ARF using exonuclease III defined a minimal fragment of intron 1 (from ϩ1784 to ϩ1844) capable of activating transcription in response to 1 nM activin A (data not shown) or co-transfected caALK4 (Fig. 2d). Altogether, these results suggest that ALK4-mediated activin signaling in gonadotropes induces follistatin gene transcription via regulatory elements located within the ARF of intron 1.
Smad Signaling Activates Follistatin Gene Transcription-To determine whether the Smad signaling pathway participates in mediating the transcriptional effects of activin on the rat follistatin gene, co-transfection experiments with hSmad2, -3, and -4 expression plasmids were performed in ␣T3-1 cells. Consistent with the data of Fig. 2, whereas the rFS(2.9i)-luc plasmid that incorporates intron 1 was activated in cells treated with activin A, rFS(2.9)-luc lacking intron 1 was not (Fig. 3a). Neither hSmad2 nor hSmad4 alone or together had an appreciable effect on either reporter plasmid (Fig. 3a). Co-transfected hSmad3, on the other hand, induced the basal activity of rFS(2.9i)-luc, but not rFS(2.9)-luc lacking intron 1 (Fig. 3a). The combined actions of hSmad3/4 enhanced the basal and activindependent reporter activities of both rFS(2.9)-luc and rFS(2.9i)luc plasmids without and with intron 1, respectively (Fig. 3a). a, b, and c) and ␣T3-1 (panels d, e,  and f)  These results illustrate that full activin/Smad-dependent activation of the rat follistatin gene in ␣T3-1 cells requires elements in intron 1 but that upstream elements may also contribute to this process.

FIGURE 1. Activin A and inhibin A regulate follistatin transcript levels of primary cultures of RAP cells (panels
A Smad-binding Element in Intron 1 Mediates Activin Effects on Follistatin Gene Transcription-Examination of the rat follistatin gene with the MatInspector sequence analysis tool (Genomatix Software GmbH, Munich, Germany) (50) identified a putative Smad4 (at Ϫ895) and a Smad3 (at Ϫ1604) binding site in the 5Ј-flanking region (Fig. 4a). More importantly, further analysis suggested the presence of a Smad3 binding motif (5Ј-GTCTGggtca-3Ј) located at ϩ1791/ϩ1800 of intron 1 (Fig. 4a). Because the data of Fig. 3a supported the potential importance of a downstream Smadbinding element, this intronic Smad3 binding site, designated SBE1, was subjected to further evaluation. To test its contribution in mediating Smad-dependent transcription of rat follistatin, the SBE1 site of the rFS(0.3ex45)-luc reporter was mutated by introducing 3-nucleotide substitutions within the putative core Smad binding site (Fig. 4a), and this mutant plasmid was then tested in ␣T3-1 cells. By contrast to the wild-type rFS(0.3ex45)-luc plasmid, the corresponding SBE1 mutant plasmid was not induced by activin A nor was it activated by co-transfected hSmad3/4 (Fig. 3b) or caALK4 (Fig.  3c). Collectively, these results suggest that SBE1 has an important role in mediating Smad-dependent transcriptional effects of activin A in ␣T3-1 cells.
Activin A Induces the Recruitment of Smad3 to the SBE1 Site of Intron 1 of the Follistatin Gene in ␣T3-1 Cells-Sequence alignment revealed that intron 1 of the rat follistatin gene is 93% identical to the mouse gene and, more importantly, that the SBE1 site is fully conserved in both genes. ChIP assays were performed to evaluate the participation of endogenous Smad signaling and the importance of SBE1 in mediating the transcriptional effects of activin A on the mouse follistatin gene of ␣T3-1 cells. A purified rabbit polyclonal antibody (anti-Smad2/3) used for these experiments was first validated for its specificity (Fig. 5a). Semiquantitative analysis of samples immunoprecipitated with anti-Smad2/3 but not normal rabbit IgG, yielded an activin A-inducible PCR-amplified fragment of the expected size corresponding to mouse follistatin (Fig. 5b). The primer set used for these analyses amplifies a 185-bp follistatin fragment that spans the entire ARF of intron 1. Further analysis using real-time PCR confirmed that activin A induced a timedependent (Fig. 5c) and a concentration-dependent (Fig. 5d) recruitment of Smad2/3 to the SBE1 site of the endogenous mouse follistatin gene in ␣T3-1 cells. Co-treatment with recombinant FS288 bio-neutralized and prevented this action of activin A (Fig. 5e). By contrast to these results, flanking regions corresponding to the putative Smad3 and Smad4 binding sites at positions Ϫ1604 and Ϫ895, respectively, were not enriched by anti-Smad2/3 (Fig. 5f). The numbers shown on the schematics of the various reporter plasmids reflect nucleotide positions relative to the transcription start site (shown with an arrow) of the rat follistatin gene. a, deletion analysis of the upstream promoter and intron 1 identify a 129-bp region of intron 1, designated ARF, that mediates the effects of activin A (1 nM) or a co-transfected constitutively active ALK4 (caALK4). b, two luciferase reporter plasmids that include all of intron 1 but incorporate the upstream portion from either Ϫ2864 (rFS(2.9i)-luc) or Ϫ312 (rFS(0.3i)-luc) are activated over the same concentration range of activin A. c, the ARF of intron 1 has enhancer activity and mediates the effects of activin A (1 nM) when placed upstream of the Ϫ312/ϩ136 fragment of rat follistatin or the SV40-pGL2 reporter. d, serial 3Ј-truncations of ARF within the context of the rFS(0.3ex45)-luc reporter plasmid define a minimal fragment (ϩ1784/ ϩ1844) that activates transcription of this plasmid by co-transfected caALK4 and activin A (not shown). The ␣T3-1 cells were transiently co-transfected with the indicated luciferase reporter plasmids or the basic pGL2 vector and CMV-␤-Gal, as an internal control for transfection efficiency, using the Superfect reagent. The cells were incubated with the DNA/Superfect mix for 6 h, then supplemented with fresh medium, and treated for 15 h with vehicle or activin A. Where indicated, a caALK4 expression plasmid or the empty vector was included. The cells were harvested, and the luciferase activity (arbitrary light units) of each sample was internally normalized to that of ␤-Gal. The data are reported as the ratio of these normalized luciferase values relative to that of the empty pGL2 vector under identical conditions. The experiments were performed in triplicate, and the results shown in all panels are the mean Ϯ S.E. of triplicate determinations from representative experiments.
Because the antibody used in ChIP analyses does not distinguish between Smad2 and Smad3, those experiments could not determine the relative recruitment or the importance of the endogenous proteins in mediating the transcriptional effects of activin A in ␣T3-1 cells. Moreover, both Smad2 and Smad3 are downstream targets of ALK4 and phosphorylated in response to activin signaling in most cells, including ␣T3-1 (data not shown). To resolve this question, oligonucleotide precipitation experiments were performed using extracts of vehicle or activin A treated ␣T3-1 cells in which expression of Myc-tagged hSmad2 and -4 or hSmad3 and -4 were achieved by lentivirusmediated delivery. Cellular extracts from these infected cells were subsequently analyzed by oligonucleotide precipitation experiments using biotinylated probes corresponding to ϩ1774/ϩ1844 within intron 1 and harboring either a wild type or a mutant SBE1 site (see Fig. 4). Western analysis of these samples revealed that hSmad3 forms a complex with the wild type but not the mutant probe and that this association is induced by activin A (Fig. 6, middle panel). By contrast, hSmad2 did not interact with either the wild type or the mutant probe at detectable levels (Fig. 6, bottom panel). In both cases, hSmad4 interacted with the wild-type probe, but this interaction was not induced by activin A (Fig. 6).
shRNA Knockdown of Smad3 Attenuates Activin-dependent Induction of Endogenous Follistatin mRNA-The experiments described thus far suggest that Smad3 is the preferential mediator of activin A effects on follistatin expression in gonadotropes. To determine if this is indeed the case, an shRNA knockdown strategy was used to evaluate the relative importance of endogenous Smad2 and Smad3 in mediating activin A effects. Four shRNAs targeting each of Smad2 or Smad3 were designed and evaluated in HEK293T cells for their ability to knockdown the expression of the corresponding Smad as compared with the empty vector control. Of the Smad2 targets tested, S2#1 and S2#2 were only marginally effective and reduced Smad2 to ϳ50% of control levels (Fig. 7a). By contrast, Smad2 shRNAs S2#3 and S2#4 reduced Smad2 to levels that were undetectable (Fig. 7a). Of the four Smad3 shRNAs, S3#3 was totally ineffective whereas S3#1, S3#2, and S3#4 significantly reduced the level of Smad3 protein detectable by Western (Fig. 7a). These effects were specific in that shRNAs targeting Smad2 did not produce detectable changes in Smad3 or Smad4 expression and, conversely, shRNAs targeting Smad3 did not influence Smad2 or Smad4 levels (data not shown). Substantial knockdown was measurable within 48 h of transduction. The shRNA plasmids that were validated to be effective by Western analysis (S2#4 and S3#4) were then tested for their ability to compromise activin-induced rFS(0.3ex45)-luc activity in ␣T3-1 cells infected with lentiviral vectors. The basal FIGURE 3. A Smad-binding element (SBE1) located within the ARF of intron 1 mediates the effects of activin A or caALK4. a, co-transfected hSmad3 or hSmad3/4, but not hSmad2 or 4 alone or together, activate the rFS(2.9i)-luc plasmid that incorporates intron 1 and enhance the effect of activin A while the plasmid lacking intron 1 (rFS(2.9)-luc) is marginally activated by co-transfected hSmad3/4. (Inset shows a representative Western analysis of co-transfected Smad levels.) b, mutagenesis of the SBE1 site abolishes transcriptional activation of rFS(0.3ex45)-luc in response to 1 nM activin A or co-transfected hSmad3/4. c, mutagenesis of the SBE1 also abolishes responses to caALK4. The ␣T3-1 cells were co-transfected with the indicated luciferase reporter plasmids, CMV-␤-Gal, and Myc-tagged hSmad2, -3, or -4 expression vectors as shown. Details of the substitutions within SBE1 to generate mSBE1 are shown in Fig. 4. For these experiments, transiently transfected ␣T3-1 cells were treated with either vehicle or 1 nM activin and processed as described above. Data represent arbitrary light units (L.U.) internally normalized to ␤-Gal activity and reported relative to pGL2. The results of triplicate determinations of representative experiments are reported as mean Ϯ S.E. FIGURE 4. A schematic representation of the rat follistatin gene. a, a schematic of the region of the rat follistatin gene from Ϫ2864 to ϩ2289, relative to the transcription start site, showing the position of predicted binding sites for several classes of transcription factors, the three major transcription initiation sites (␣, ␤, and ␥) and the activin-responsive region (ARF) of intron 1. b, the nucleotide sequence of rat follistatin corresponding to ARF (from ϩ1764 to ϩ1912) of intron 1. Shown are the positions of the 3Ј truncations generated by digestion with exonuclease III (exo45, exo91, exo115) and the nucleotide substitutions within SBE1 (boxed) to generate the mutant SBE1 form (mSBE1). The forward and reverse arrows correspond to the equivalent position of primers derived from the mouse follistatin gene and used for real-time PCR analysis of ChIP samples described in Fig. 5. The solid underline corresponds to the biotinylated probe used for the oligonucleotide pull-down experiments presented in Fig. 6. or activin A-dependent activation of the rFS(0.3ex45)-luc reporter in cells transduced with Smad2 shRNA S2#4 was indistinguishable from those infected with the empty vector control (Fig. 7b). By contrast, Smad3 shRNA S3#4 significantly attenuated rFS(0.3ex45)-luc reporter activation at all concentrations of activin A tested, compared with control (Fig. 7b). This partial attenuation of reporter activity could have reflected the actions of residual Smad3 protein remaining in cells transduced with Smad3 shRNA S3#4 or the participation of other signaling pathways. To further substantiate the relevance of these observations in the context of the endogenous gene, parallel experi-ments were also performed to evaluate activin A effects on endogenous follistatin mRNA accumulation in ␣T3-1 cells transduced similarly with the empty, S2#4, or S3#4 lentiviral vectors. Cells that were infected with the empty lentiviral vector responded to 1 nM activin A with a 3-fold increase in follistatin mRNA abundance as determined by real-time PCR (Fig. 7c). The Smad2 shRNA S2#4 only slightly blunted the magnitude of this response despite the significant knockdown of Smad2 protein (Fig. 7c). By contrast, S3#4-mediated knockdown of Smad3 blunted the response to activin A to ϳ50% of control, to levels that did not reach statistical significance (Fig. 7c).

DISCUSSION
Activins are key modulators of follistatin expression and many tissues that express and respond to activin often also produce follistatin (2,7). This close anatomic and functional link establishes a local feedback loop that safeguards against excessive activin signaling and insures that the necessary balance of activin and follistatin tone within a given tissue is maintained. Accordingly, it is presumed that many of the demonstrated actions of follistatins reflect their influence as local "buffers" of the bioactivity of activins and other members of the TGF-␤ family of ligands (2,14). In the anterior pituitary, the self-modulating mechanism through which activin induces follistatin production exerts control over activin bioactivity and plays a pivotal role in maintaining FSH␤ expression and FSH production from gonadotropes at levels that are physiologically relevant and necessary for normal reproductive function (7,8). Several lines of evidence also suggest that a balanced activin and follistatin tone is a critical checkpoint for the control of pathogenic mechanisms underlying tumor formation within the pituitary (51,52). Genetic models support these far-reaching actions of follistatin. The overexpression of follistatin in mice is associated with reduced FSH levels, subfertility arising from gonadal defects as well as abnormalities of the skin and hair formation (53). Mice with a targeted deletion of the follistatin gene, on the other hand, die within hours of birth because of numerous developmental abnormalities of musculoskeletal tissue and the skin (54). Despite the established importance of  follistatin, the mechanisms or signaling pathways that regulate its expression and mediate the effects of activin have not been extensively explored.
To start unraveling the mechanism of activin-induced follistatin expression, the mouse ␣T3-1 cell line was used as a model of pituitary gonadotropes to evaluate the transcriptional regulation of this gene. The findings of the current study confirm that follistatin is a transcriptional target of activin A in ␣T3-1 cells. The results further indicate that activin inducibility of follistatin in this cell type is controlled by a Smad-binding element (SBE1) located within a regulatory region of intron 1 (ARF). This intronic ARF displays activin-responsive enhancer activity and mediates activin A effects from either upstream or downstream sites, albeit more weakly than in its native context. The latter is consistent with the possibility that other regulatory regions cooperate with SBE1 and contribute to full transcriptional activation of the follistatin gene.
The results presented here suggest that the SBE1 element identified in the current study mediates Smad3, and to a much lesser extent Smad2 effects and that Smad3 is the preferential downstream mediator of activin in ␣T3-1 cells. This conclusion is supported both by shRNA knockdown and oligonucleotide pull-down experiments. shRNA-mediated knockdown of Smad3 significantly compromised the ability of activin A to regulate SBE1-dependent transcription and, more importantly, to induce endogenous follistatin mRNA accumulation in ␣T3-1 cells. Furthermore, oligonucleotide precipitation experiments demonstrated that activin A induces the association of Smad3, but not Smad2, to a biotinylated probe that integrates the wildtype SBE1. This association was probably not facilitated by Smad4 because activin-dependent Smad3 association with the wild-type SBE1 probe was observed regardless of whether Smad4 was co-expressed ( Fig. 6) or not (data not shown). Moreover, the observed Smad4 association with the wild-type probe was independent of activin A. These results suggesting that Smad4 does not contribute to the inducible assembly of Smad3 complexes to SBE1 are consistent with reporter assays that also failed to reveal Smad4-dependent activation via SBE1. With regards to Smad2, the failure of the wild-type probe to recruit it could have resulted from the absence of FoxH1 in ␣T3-1 cells, given that FoxH1 is an obligatory cofactor of Smad2 at some downstream targets of TGF-␤ ligands (55)(56)(57). This was not the case, however, because concurrent expression of FoxH1 and encoding Myc-tagged hSmad2 or -3. Immunoblots that were sequentially probed with a mAb anti-Myc then a mAb anti-actin as an internal loading control were subjected to ECL for visualization of immune complexes. b, concentration-dependent effects of activin A on the activity of the rFS(0.3ex45)-luc reporter plasmid in ␣T3-1 cells transduced with the empty lentiviral vector as control or vectors encoding Smad2 (S2#4) or Smad3 (S3#4) targeting shRNAs. The results of a representative experiment, performed in triplicate, are shown as the ratio of measured luciferase activity to ␤-Gal. c, activin A effects on endogenous follistatin mRNA levels of ␣T3-1 cells following transduction with the empty lentiviral vector as control or vectors encoding Smad2 (S2#4) or Smad3 (S3#4) targeting shRNAs. On the fourth day postinfection, the cells were treated with either vehicle or 1 nM activin A for 2 h, and total RNA extracted from them was subjected to real-time PCR analysis using primers corresponding to mouse follistatin (forward primer: 5Ј-CCCCAACTGCATCCCTTGTA and reverse primer: 5Ј-GGTCCGCAGTC-CACGTTCT) with GAPDH as the internal control. The mean Ϯ S.E. of triplicate determinations of a representative experiment are shown relative to the baseline of vehicle-treated control samples. Smad2, with or without Smad4, did not activate activin-responsive plasmids such as rFS(2.9i)-luc, rFS(0.3i)-luc or rFS(0.3ex45)-luc plasmids above baseline (data not shown). Therefore, unlike the intronic enhancers of lim-1, Xnr1 and the intronic left side-specific enhancer of mouse Nodal (58 -60), the Smad-binding element in intron 1 of the rat follistatin gene does not assemble FoxH1-Smad2 complexes.
Initially, DNA fragments corresponding to the upstream promoter region of the rat follistatin gene were evaluated for their ability to mediate activin A effects in transiently transfected ␣T3-1 cells. Primer extension analyses confirmed that both ␣T3-1 and primary rat anterior pituitary cells preferentially utilize the "␣" transcription start site and to a lesser extent the "␤" and "␥" sites 4 previously identified from studies of granulosa, P19 and F9 embryonic carcinoma cells (35,36). An upstream fragment equivalent to Ϫ2864 to ϩ136, relative to the transcription start site of the rat follistatin gene, failed to confer activin inducibility onto the basic pGL2 luciferase reporter. The same construct was only weakly (1.5-fold) activated in ␣T3-1 cells when the Smad signaling pathway was activated directly via co-transfection of a constitutively active form of ALK4 (caALK4). Truncating the Ϫ2864/ϩ136 fragment at the 5Ј-end to Ϫ752 (data not shown) or Ϫ312 reduced basal reporter activity but did not influence activin responsiveness confirming that elements upstream of the Ϫ312 position of the rat follistatin gene do not contain inhibitory sites nor are they sufficient for mediating the transcriptional effects of activin A in ␣T3-1 cells. Further analysis localized the activin-responsive region of the rat follistatin gene to the 3Ј-end of the first intron, between nucleotides ϩ1784 and ϩ1844 relative to the transcription start site.
Interestingly and in contrast to the results of the current study, it has been reported that regulatory elements located between Ϫ2165 and ϩ2 are sufficient for Smad-dependent transcription of the human follistatin gene in HepG2 cells (40,41). These apparently conflicting results could arise from differences in the human and rat follistatin genes or differences in the mechanisms of activation in the two cell types. The first of these possibilities is unlikely given the high degree of conservation of the follistatin gene in vertebrates at the nucleotide level. Alignment of up to 3 kb of the upstream promoter shows this region to be 95 or 79% identical between rat and mouse or human, respectively. Within this upstream region, putative Smad3 (at Ϫ1604) and Smad4 (at Ϫ895) sites identified in the rat follistatin gene are conserved in mouse and human. One or both of these could have mediated the effect of activin in HepG2 cells but neither was functional in aT3-1 cells. Remarkably, intron 1 is also highly conserved at the level of 93 or 77% sequence identity between rat and mouse or human, respectively. Moreover, the SBE1 site of intron 1 is completely conserved in all three. These observations provide a compelling argument for cell-specific mechanisms or cofactors that dictate the differential utilization of upstream or downstream Smadbinding elements and ultimately control the local concentrations of follistatin within a given microenvironment. Whether intron 1 of the rat follistatin gene is dispensable for the activation of this gene in cell types other than gonadotropes and whether the first intron of the human follistatin gene contributes to inducible expression in HepG2 or other cell types will have to be a subject of future experiments. Initial experiments, however, suggest that SBE1 of intron 1 does not significantly contribute to transcriptional activation in HepG2 and HEK293 cells treated with activin A or TGF-␤ whereas an intact SBE1 is necessary for the induction of both rat and mouse follistatin in gonadotrope-derived ␣T3-1 as well as the more differentiated L␤T2 cell line (data not shown).
Activin-mediated Smad2/3 signaling has a critical role in pituitary gonadotropes to promote differential expression of FSH over LH at appropriate stages of the reproductive cycle and maintain sensitivity to GnRH by inducing FSH␤ (61-63) and regulating GnRH-R expression (64,65). Coordinately and in conjunction with feedback control mechanisms mediated by gonadal inhibin and steroids, activin induces the local production of pituitary follistatin to modulate further activin signaling and promote timely fluctuations of gonadotropins necessary for normal cycling (6,8). In rodents, pituitary follistatin levels peak in late proestrus but are low in the morning of estrus when the activin-dependent secondary FSH surge occurs (66). Studies of the FSH␤ and GnRH-R genes have highlighted the complexity of this system and begun to unravel the mechanism through which activin, GnRH, and gonadal steroids promote the hierarchical recruitment of cofactors and coordinately regulate their expression in gonadotropes. Whether some of the same factors that have been shown to cooperate with Smad2/3 at the GnRH-R promoter, such as Lhx3, Pitx-1, or AP-1 (64,67,68) or at the FSH␤ promoter, such as Pitx1, Pitx2, Lhx3, or TALE homeodomain protein, Pbx1 and Prep1, (63,69,70), also exert control over follistatin expression in gonadotropes remains an open question for future studies.
Altogether, the studies described here demonstrate that a Smad-binding element in intron 1 of the rat follistatin gene mediates the transcriptional effects of activin in mouse gonadotrope-derived ␣T3-1 cells. The results suggest that this Smad-binding element, designated SBE1, preferentially recruits Smad3 to facilitate cell-specific expression of follistatin in response to activin A. By revealing the potential complexity of the mechanisms that may be involved in Smad-dependent regulation of the follistatin gene in different cell types, the present findings provide a framework for future studies for assessing if the mechanism defined in the current work is unique to gonadotropes. Preliminary experiments suggest that SBE1-mediated activation of the follistatin gene is dependent on the cooperation of Smad3 with cell-type restricted factors present in ␣T3-1 but not in HepG2 or HEK293T cells. Ultimately, the identification of these cell-specific factors and a better understanding of the underlying mechanisms could potentially identify tools for cell-specific manipulation of the levels of this important modulator of activin signaling.