A Novel AP-1 Site Is Critical for Maximal Induction of the Follicle-stimulating Hormone β Gene by Gonadotropin-releasing Hormone*

Regulation of follicle-stimulating hormone (FSH) synthesis is a central point of convergence for signals controlling reproduction. The FSHβ subunit is primarily regulated by gonadotropin-releasing hormone (GnRH), gonadal steroids, and activin. Here, we identify elements in the mouse FSHβ promoter responsible for GnRH-mediated induction utilizing the LβT2 cell line that endogenously expresses FSH. The proximal 398 bp of the mouse FSHβ promoter is sufficient for response to GnRH. This response localizes primarily to an AP-1 half-site (–72/–69) juxtaposed to a CCAAT box, which binds nuclear factor-Y. Both elements are required for AP-1 binding, creating a novel AP-1 site. Multimers of this site confer GnRH induction, and mutation or internal deletion of this site reduces GnRH induction by 35%. The same reduction was achieved using a dominant negative Fos protein. This is the only functional AP-1 site identified in the proximal 398 bp, since its mutation eliminates FSHβ induction by c-Fos and c-Jun. GnRH regulation of the FSHβ gene occurs through induction of multiple Fos and Jun isoforms, forming at least four different AP-1 molecules, all of which bind to this site. Mitogen-activated protein kinase activity is required for induction of FSHβ and JunB protein. Finally, AP-1 interacts with nuclear factor-Y, which occupies its overlapping site in vivo.

Follicle-stimulating hormone (FSH) 1 is a key regulator of reproduction, since it is essential for oogenesis in females and regulates spermatogenesis in males (1). Production of FSH is limited to anterior pituitary gonadotrope cells, which also produce luteinizing hormone (LH). FSH is a heterodimeric glyco-protein hormone consisting of two subunits: an ␣-subunit, which is common to LH, thyroid-stimulating hormone, and chorionic gonadotropin (CG), and a unique ␤-subunit that confers specific biological activity (2). Expression of the ␤-subunit gene is the limiting factor in FSH synthesis, and its transcription is regulated primarily by gonadotropin-releasing hormone (GnRH), gonadal steroids, and the activin-inhibin-follistatin system (3)(4)(5).
GnRH is a decapeptide neurohormone, released by a subset of hypothalamic neurons into the hypophyseal portal system, where it binds its receptor on the pituitary gonadotrope membrane. The GnRH receptor belongs to the class of G proteincoupled receptors and, upon ligand binding, activates the protein kinase C and mitogen-activated protein kinase (MAPK) signaling pathways (6). GnRH administration, either to cells in culture or to hypogonadotropic animals, induces transcription of the early response genes, c-fos, c-jun, and egr-1 (7)(8)(9). The transcription factor AP-1, which is composed of Jun/Jun homodimers or Jun/Fos heterodimers, has been implicated in GnRH induction of the FSH␤ gene. Previous studies reported that nuclear proteins from GnRH-treated cells bind an AP-1 consensus sequence (10), that purified c-Jun protein binds putative AP-1 sites in the ovine FSH␤ promoter (11), and that mutation of putative AP-1 sites in this promoter reduces GnRH induction in heterologous HeLa cells (10). However, in mice carrying a transgene of the ovine FSH␤ 5Ј-flanking region linked to luciferase in which the same AP-1 sites were mutated, transgene response to GnRH did not differ from the wild-type ovine FSH␤ promoter (12). Furthermore, one of these AP-1 sites is not conserved in the mouse, rat, or human promoters. Therefore, there is a need to examine FSH␤ regulation of the mouse gene, especially in light of the fact that there is a high degree of conservation between mouse and human FSH␤ genes and that targeted disruption of the FSH␤ gene in mice has a phenotype similar to loss-of-function mutations in humans (1).
Until recently, no FSH␤-producing cell lines were available. Models of GnRH action using reconstitution of GnRH receptor in non-gonadotrope-derived cell lines may lack signaling molecules or transcription factors necessary for appropriate induction of gonadotrope-specific genes, whereas primary pituitary cell cultures contain only about 5% gonadotropes and are difficult to manipulate in vitro. The gonadotrope-derived L␤T2 cell line expresses FSH␤ endogenously (13) and secretes FSH in response to activin (14). These cells also express other markers of pituitary gonadotropes, most notably ␣-subunit, GnRH receptor, LH␤, and all of the components of the activin system autocrine loop: activin, follistatin, and activin receptor (13) as well as inhibin and inhibin receptor (15). Therefore, the L␤T2 cell line is an excellent model in which to directly study regulation of FSH␤ gene expression.
Indeed, these cells have been used to investigate GnRH signal transduction (16) and, more recently, the molecular ba-sis for cell-specific expression of FSH␤, by comparison with the non-FSH-producing gonadotrope-derived cell line, ␣T3-1 (17). In the latter study, we identified specific promoter elements binding steroidogenic factor-1, an orphan nuclear receptor that is specifically expressed in the gonadotrope population and regulates gonadotrope-specific genes within the pituitary. We also identified a conserved binding site in the proximal region of the promoter for nuclear factor-Y (NF-Y), a ubiquitously expressed heterotrimeric transcription factor (18), and showed a role for both NF-Y and steroidogenic factor-1 in gonadotropespecific expression of the mouse FSH␤ gene (17).
The goal of the present study is to gain an understanding of the molecular mechanisms by which GnRH induces FSH␤ gene expression, using the mouse L␤T2 gonadotrope cell model. We demonstrate that regulation by GnRH is mediated in part by induction of multiple AP-1 isoforms. These AP-1 isoforms bind a novel site that overlaps the element that binds the basal transcription factor NF-Y. This novel site consists of a half-site of the AP-1 consensus binding sequence and an adjacent CCAAT box. Furthermore, NF-Y and AP-1 physically interact and, following GnRH stimulation, co-occupy this site in vivo.

EXPERIMENTAL PROCEDURES
Cell Culture and Transient Transfection-L␤T2 cells were plated on 6-well plates 1 day prior to transfection. Transfection was performed in Dulbecco's modified Eagle's medium with 10% fetal bovine serum using Fugene 6 reagent (Roche Applied Science) following the manufacturer's instructions. Each well was transfected with 1 g of mFSH␤-luc. Plasmid construction and preparation has been described previously (17). The 398-bp mouse FSH␤ promoter was PCR-amplified from a genomic clone that was kindly provided by Dr. Malcolm Low and ligated into the SmaI restriction site of the pGL3 luciferase reporter plasmid (Promega, Madison, WI) to generate Ϫ398 mFSH␤-luc. The Ϫ304 and Ϫ230 truncations were created by digesting the MluI/BglII fragment of Ϫ398 mFSH␤-luc with XmnI and BbvI restriction enzymes, respectively. The truncated promoter fragments were then blunt-ended with Klenow fragment (New England Biolabs, Beverly, MA) and ligated into the SmaI restriction site of pGL3. The Ϫ129 reporter was created by digesting the MluI/BglII fragment of Ϫ304 mFSH␤-luc with DpnI, bluntended with Klenow, and ligated into the SmaI restriction site of pGL3. The Ϫ194 plasmid was generated by digesting the MluI/BglII fragment of Ϫ230 mFSH␤-luc with the RsaI restriction enzyme and cloning the appropriate promoter fragment into the SmaI and BglII sites of pGL3. The Ϫ95 FSH␤Luc plasmid was created by PCR-amplifying the promoter from Ϫ398 mFSH␤-luc using a forward primer corresponding to the sequence of the mouse FSH␤ promoter from Ϫ95 to Ϫ77 bp and containing a KpnI linker and a reverse primer spanning the HindIII restriction site from the pGL3 vector. The PCR product was digested with KpnI and BglII restriction enzymes and ligated into the corresponding sites in pGL3. The reporter plasmid with the multimerized novel AP-1 site was created using an oligonucleotide, containing four copies of the Ϫ78/Ϫ67 sequence of the mouse FSH␤ promoter between KpnI and NheI linkers and was ligated into corresponding sites in pGL3, upstream of the Ϫ81 bp herpes thymidine kinase promoter, which was ligated between the XhoI and BglII sites. The sequences of all promoter fragments were confirmed by dideoxynucleotide sequencing performed by the DNA Sequencing Shared Resource, University of California San Diego Cancer Center.
An expression plasmid containing ␤-galactosidase driven by the Herpesvirus thymidine kinase promoter was co-transfected with mFSH␤luc and used as an internal control. Sixteen h after transfection, the cells were switched to serum-free Dulbecco's modified Eagle's medium supplemented with 0.1% bovine serum albumin, 5 mg/liter transferrin, and 50 nM sodium selenite. The following day, cells were treated with 10 nM GnRH (Sigma) for 6 h, unless otherwise indicated. The cells were then lysed with 0.1 M potassium phosphate buffer (pH 7.8) with 0.2% Triton X-100. Equal volumes of each lysate were placed in 96-well plates, and luciferase activity was measured on a luminometer (EG&G Berthold Microplate) by injecting 100 l of a buffer containing 100 mM Tris-HCl (pH 7.8), 15 mM MgSO 4 , 10 mM ATP, and 65 M luciferin per well. Galactosidase activity was measured using the Galacto-light assay (Tropix, Bedford, MA) following the manufacturer's instructions. All transfection experiments were performed in triplicate and repeated at least three times. Luciferase values from reporter gene-transfected cells were consistently at least 100 times higher than values from mock-transfected cells. Results represent the mean Ϯ S.E. of all samples analyzed. An asterisk marks a statistically significant difference from the control-treated cells, determined by analysis of variance followed by Tukey-Kramer HSD post hoc multiple range test for individual comparison with p Յ 0.05.
Western Blot-Following overnight starvation and GnRH treatment, L␤T2 cells were rinsed with phosphate-buffered saline and lysed with lysis buffer (20 mM Tris-HCl, pH 7.4, 140 mM NaCl, 0.5% Nonidet P-40, 0.5 mM EDTA, with protease inhibitors: aprotinin, pepstatin, and leupeptin at 10 g/ml each and 1 mM phenylmethylsulfonyl fluoride. Protein concentration was determined with Bradford reagent (Bio-Rad), and an equal amount of protein per sample was loaded on SDS-PAGE gel. After proteins had been resolved by electrophoresis and transferred to a polyvinylidene fluoride membrane, they were probed with specific antibodies for Fos and Jun proteins (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) and NF-YA (Chemicon, Temecula, CA). The bands were detected with secondary antibodies linked to horseradish peroxidase and enhanced chemiluminescence reagent (Amersham Biosciences).
Electrophoretic Mobility Shift Assay (EMSA)-After GnRH treatment, cells were scraped in hypotonic buffer (20 mM Tris-HCl, pH 7.4, 10 mM NaCl, 1 mM MgCl 2 with the same protease inhibitors as mentioned above for the lysis buffer) and allowed to swell on ice. Cells were lysed by passing through a 25-gauge needle, and the nuclei were pelleted by centrifugation. Nuclear proteins were extracted in hypertonic buffer (20 mM Hepes, pH 7.8, 420 mM KCl, 1.5 mM MgCl 2 with protease inhibitors, and 20% glycerol). Two g of nuclear proteins per sample was used in the binding reaction (10 mM Hepes, pH 7.8, 50 mM KCl, 5 mM MgCl 2 , 5 mM dithiothreitol, 0.1% bovine serum albumin, 0.1% Nonidet P-40 with 0.5 g/ml poly(dI-dC) and 2 fmol per reaction of end-labeled probe). Oligonucleotides were labeled with [␥-32 P]ATP using T4 kinase. In the competition experiments, competitor oligonucleotide was added 10 min prior to the addition of the probe, as were the antibodies in the supershift assays. The Fos, Jun, and NF-Y antibodies used are the same as in the Western blot, whereas the nonspecific IgG is from Santa Cruz Biotechnology. The reaction was loaded on a 5% acrylamide gel in 0.25ϫ TBE and electrophoresed at 1-V/cm 2 constant voltage. After drying, gels were exposed to autoradiography.
Mutagenesis-Mutagenesis and deletion of the FSH␤-luc plasmid were performed using the QuikChange TM site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's protocol. The oligonucleotide used to mutate the AP-1 half-site (GTCA) was 5Ј-CAGCAGGCTTTATGTTGGTATTGGTCCCGTTAACACCC-3Ј (top strand; mutated nucleotides are underlined), and the oligonucleotide used to mutate the NF-Y site (ATTGG) was 5Ј-CAGCAGGCTTTATGT-TGGTACCGGTCATGTTAACACCC-3Ј (top strand; again, underlined nucleotides indicate a change from the wild-type sequence). To create an AP-1 consensus, the NF-Y site/AP-1 half-site was mutated to a consensus AP-1 site using the oligonucleotide 5Ј-CAGCAGGCTTTAT-GTTGGTATGAGTCATGTTAACACCC-3Ј. To create a reporter lacking the AP-1 site, a specific internal deletion was made using the oligonucleotide 5Ј-CAGCAGGCTTTATGTTGGTGTTAACACCCAGTAAATCC-3Ј. Mutations were confirmed by dideoxyribonucleotide sequencing, as above.
Chromatin Immunoprecipitation (ChIP) Assay-L␤T2 cells were starved overnight and treated with GnRH for 2 h. Proteins were crosslinked to DNA by the addition of 1% formaldehyde directly to the cell medium, and, after obtaining the nuclear fraction, chromatin was sonicated to an average length of 1 kb in sonication buffer (50 mM Hepes, pH 7.9, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS). After preclearing with protein A beads, protein-DNA complexes were bound overnight to the same c-Jun, c-Fos, and NF-Y antibodies as were used in the EMSA experiments for supershift assays and precipitated with protein A beads (Amersham Biosciences). After extensive washing (two times each with sonication buffer defined above, high salt sonication buffer (500 mM NaCl with other components as defined above), lithium chloride buffer (20 mM Tris, pH 8, 250 mM LiCl, 1 mM EDTA, 0.1% Nonidet P-40, 0.1% sodium deoxycholate), and TE buffer), cross-linking was reversed by the addition of 300 mM NaCl and incubation at 65°C, and proteins were digested by incubation with Proteinase K. DNA was phenol-chloroformextracted and ethanol-precipitated, and the sequence of interest was amplified by PCR. Primers used in PCR were 5Ј-GGTGTGCTGCCAT-ATCAGATTCGG-3Ј and 5Ј-GCATCAAGTGCTGCTACTCACCTGTG-3Ј and spanned the 280-bp sequence in the mouse FSH␤ gene from Ϫ223 to ϩ57. The specificity of the product was assessed by the presence of a single band of the expected size on an ethidium bromide-stained agarose gel. For quantification, the PCR product was labeled by including [␣-32 P]dATP in the nucleotide mix used in PCR and run on a 5% acrylamide gel in 0.5ϫ TBE. The gels were dried, subjected to autoradiography, and quantified using a PhosphorImager Optical Scanner Storm 860 (Amersham Biosciences) and the ImageQuant program (Amersham Biosciences).
GST Interaction Assay-The glutathione S-transferase (GST)-Jun in the pGEX vector was kindly provided by Dr. Michael Karin (19), whereas GST-NF-YA and GST-NF-YB were kindly provided by Dr. Sankar Maity (20) and Dr. David Gardner (21), respectively. The c-Jun expression vector was obtained from Dr. Michael Birrer (22), and the c-Fos expression vector was obtained from Dr. Eugene Tulchinsky (23). The NF-Y and green florescent protein expression vectors were provided by Dr. Roberto Mantovani (24) and Dr. Douglass Forbes, respectively. 35 S-Labeled proteins were produced using the TNT® T7 coupled reticulocyte lysate system (Promega). Bacteria transformed with the pGEX vectors were grown to an OD of 0.6, upon which protein expression was induced by the addition of 0.25 mM isopropyl-␤-D-thiogalactosidase. Bacterial pellets were sonicated in phosphate-buffered saline with 5 mM EDTA and 0.1% Triton X-100 and centrifuged, and the supernatant was bound to glutathione-Sepharose beads (Amersham Biosciences). Beads were washed four times with sonication buffer, followed by equilibration in the binding buffer (below), and split equally between different samples and the control. 35 S-Labeled proteins were added to the beads and bound for 1 h at 4°C in 20 mM Hepes (pH 7.8), with 50 mM NaCl, 10 mg/ml bovine serum albumin, 0.1% Nonidet P-40, and 5 mM dithiothreitol. After extensive washing, samples were eluted from the beads by boiling in Laemmli sample buffer and subjected to SDS-PAGE. Afterward, the gels were dried and autoradiographed.

GnRH Induces FSH␤ through Proximal Regulatory Sequences-
Since GnRH is a major regulator of FSH␤ synthesis, we sought to determine the molecular mechanisms by which GnRH induces FSH␤ gene expression. A plasmid containing the proximal 398 bp of the mouse FSH␤ 5Ј regulatory region linked to a luciferase reporter gene (mFSH␤-luc) was transiently transfected into L␤T2 cells. We chose to study the mouse FSH␤ regulatory sequence transfected into this FSH-expressing, murine pituitary gonadotrope cell line to provide a homologous model for analysis of gene expression. This region of the mouse gene is highly conserved with ovine, rat, and human FSH␤ genes. We have previously shown that 398 bp of the 5Ј regulatory region of the mouse FSH␤ gene is sufficient to provide gonadotrope-specific expression (17). L␤T2 cells transiently transfected with mFSH␤-luc were treated with increasing concentrations of GnRH to test whether this region of the mouse FSH␤ gene is also sufficient for GnRH responsiveness. Stimulation with GnRH over a range of doses and time periods revealed that this short regulatory region contains elements that allow response to GnRH in a time-and dose-dependent manner. Maximal induction is observed after 6 h of GnRH treatment, and expression returns to basal level within 24 h of GnRH treatment (Fig. 1A). Increasing doses of GnRH result in increasing activity of the mouse FSH␤ promoter. In the following experiments, cells were treated for 6 h with 10 nM GnRH, a concentration closer to the physiological range of GnRH during the estrous cycle.
To identify which promoter elements convey GnRH responsiveness, we mapped regions of the mouse FSH␤ gene promoter that confer GnRH response using truncation deletion analysis. L␤T2 cells were transiently transfected with a series of truncations of the mouse FSH␤ gene 5Ј-flanking region, ranging in length from 398 to 95 bp upstream of the transcription start site, and the ability of GnRH to induce transcription was assayed ( Fig. 1B). Fig. 1B displays the GnRH regulation as -fold induction over vehicle-treated for each of the truncated promoter regions. As we have previously shown, the basal level of expression is not significantly changed by truncation through the region from Ϫ398 to Ϫ95 in L␤T2 cells (17). However, here we show that GnRH induction is significantly reduced by sequential deletion of either of two regions of the mouse FSH␤ promoter. Significant decreases in GnRH responsiveness were found when the promoter was truncated from Ϫ304 to Ϫ230 bp and in the most proximal region between Ϫ95 bp and the start site of transcription. The reduction due to truncation from Ϫ304 to Ϫ230 is minor, and an apparent increase in responsiveness from Ϫ127 to Ϫ95 is not statistically significant, but it leads to the finding that no statistically significant decrease in responsiveness exists between truncations Ϫ304 and Ϫ95. Indeed, a substantial level of induction is retained in the 95-bp most proximal region, and this is the focus of the following investigation.
An AP-1 Half-site Overlapping the CCAAT Box Is Bound by AP-1 following GnRH Stimulation and Is Essential for Maximal Induction of FSH␤ by GnRH-Because the proximal 95-bp region retains 2.6-fold induction by GnRH, whereas the entire 398 bp of the regulatory region is induced 3.4-fold, we focused on this proximal region to determine what transcription factors confer GnRH induction. Using the TransFac® data base (25), the proximal region of the promoter was analyzed for putative transcription factor binding sites. This search revealed a half-

FIG. 1. Localization of the GnRH-responsive region in the mouse FSH␤ promoter.
A, the 398-bp mouse FSH␤ promoter was linked to a luciferase reporter gene (mFSH␤Ϫluc) and transiently transfected into gonadotrope-derived L␤T2 cells. The herpes thymidine kinase ␤-galactosidase reporter was co-transfected as an internal control. After overnight starvation in serum-free Dulbecco's modified Eagle's medium, the cells were treated with the indicated concentrations of GnRH for different lengths of time (indicated on the abscissa), after which the luciferase activity in the lysates was measured and normalized to ␤-galactosidase. Results represent the mean Ϯ S.E. of at least three independent experiments, each performed in triplicate, and are presented as -fold induction from vehicle control. B, different lengths of the mouse regulatory region were transiently transfected into L␤T2 cells and after overnight starvation, the cells were treated with 10 nM GnRH for 6 h. The results are represented as -fold induction from vehicle-treated cells for each truncation. Significantly different induction in the treated cells versus the control cells for each truncation is marked with an asterisk, whereas a significant drop in induction between the subsequent truncations is marked with a number sign. Results represent the mean Ϯ S.E. of four independent experiments, each performed in triplicate. site for the AP-1 transcription factor adjacent to the CCAAT box, a binding site for the basal transcription factor NF-Y (17). The AP-1 consensus sequence is TGA(C/G)TCA. In the mouse FSH␤ gene promoter, the first half of the AP-1 inverted repeat is replaced by a CCAAT box (ATTGG on the coding strand), and the second half of the repeat (GTCA) is juxtaposed to this CCAAT element. Sequence alignment revealed that the AP-1 half-site and the adjacent NF-Y site are conserved in the mouse, rat, and human FSH␤ promoters but absent from the ovine and bovine promoters ( Fig. 2A).
Next, EMSA was utilized to determine whether NF-Y and/or AP-1 bind to the identified elements and, if so, whether binding is affected by GnRH treatment. For that purpose, a radiolabeled 35-bp sequence from Ϫ99 to Ϫ65 bp, which spans these sites in the mouse gene, was incubated with nuclear extracts from L␤T2 cells treated with GnRH for different lengths of time (Fig. 2B). Several bands were detected that show altered intensity following GnRH treatment. These include complex 1, which shows increased intensity, and complex 2, which is only present following GnRH treatment. To examine whether these complexes contain AP-1 or NF-Y, we used antibodies that recognize all of the Fos and Jun protein isoforms, respectively, and an antibody directed against the NF-YA subunit, which is the most regulated subunit of the NF-Y heterotrimer (18). The Fos antibody induces a complete supershift of the faster mobility complex 2 and a partial supershift of the slower mobility complex 1, reducing the intensity of complex 1 to the untreated (0 GnRH) control level, as does the Jun antibody. These results indicate that both GnRH-regulated bands contain AP-1 complexes, which consist of Fos/Jun heterodimers. The NF-Y antibody completely supershifts complex 1 in the control extracts, as well as some of complex 1 in GnRH-treated nuclear extracts. We postulated that complex 1 is composed of an AP-1 complex co-migrating with an NF-Y complex and that complex 2 is another form of AP-1. Attempts to resolve the bands within complex 1 were unsuccessful. However, the simultaneous in- A, an NF-Y site and an adjacent AP-1 half-site were identified in the mouse FSH␤ promoter using the Transfac® data base. Alignment of the sequence from Ϫ99 to Ϫ65 of the mouse FSH␤ gene regulatory region reveals that the NF-Y site (underline) and the AP-1 half-site (dashed underline) are conserved in human and rodent species but are absent from the ovine and bovine promoters. Consensus binding sites are noted below the alignment. B, EMSA analysis of nuclear extracts from L␤T2 cells treated with vehicle (0 h) or 10 nM GnRH (0.5, 2, or 6 h), using the Ϫ99/Ϫ65 sequence as a probe, are shown. The length of GnRH treatment in hours is indicated above each lane, and the antibodies used in supershift assay are marked above corresponding lanes. IgG represents a nonspecific antibody used as a control. The supershifted bands are indicated with ss, whereas 1 and 2 designate complexes that change following the treatment.
clusion of both NF-Y and AP-1 antibodies supershifts complex 1 completely, indicating that this band is composed of two complexes, one induced by GnRH (Fos/Jun) and one not affected by GnRH (NF-Y). Thus, AP-1 in nuclear extracts from cells treated for either 2 or 6 h with GnRH, but not from untreated cells, bind to this proximal region of the mouse FSH␤ promoter. NF-Y, on the other hand, binds to its site in both control and GnRH-treated cells, and the intensity of NF-Y binding does not change in response to GnRH.
To determine which nucleotides are needed for NF-Y and AP-1 binding, competitions with a 100-fold excess of unlabeled mutant oligonucleotides were utilized (Fig. 3). Two base-pair scanning mutations (mutants A-J) and a one base-pair mutation (mutant K), were introduced into the Ϫ99/Ϫ65 oligonucleotide (Fig. 3A). Nuclear extract from vehicle-treated L␤T2 cells was used in EMSA to confirm that NF-Y binds the CCAAT box (Fig. 3B). Oligonucleotides with mutations in the NF-Y element (mutants E, F, G, and to some extent H) cannot bind NF-Y and do not compete with the labeled wild-type probe. The AP-1 consensus sequence (AP-1 lane) cannot compete for NF-Y binding (Fig. 3B). In a similar competition EMSA, using nuclear proteins from cells treated with GnRH for 6 h (Fig. 3C), mutants E-I cannot compete successfully for AP-1 binding. Inclusion of an NF-Y antibody to induce a supershift of NF-Y allows visualization of the AP-1 band, which is normally obscured by their co-migration (Fig. 3D). Again, binding of NF-Y, which is supershifted (ssNF-Y), requires bases mutated in E, F, G, and partially H, whereas binding of AP-1 requires nucleotides mu-tated in oligonucleotides E-I, which could not compete for AP-1 binding. As expected, the AP-1 consensus sequence competes for AP-1 binding but not for NF-Y binding. These results indicate that the AP-1 binding site overlaps the NF-Y binding site and that both the AP-1 half-site and adjacent CCAAT box are required to bind AP-1. Together, these two regulatory elements create a novel AP-1 site.
To assess the contribution of these sites to FSH␤ induction by GnRH, we introduced selective mutations into mFSH␤-luc. We chose mutations that either inhibit both AP-1 and NF-Y binding in vitro (mutant F in EMSA) or inhibit AP-1 binding only (mutant I in EMSA). These mutations reduce GnRH induction of luciferase activity by 25 and 35%, respectively, compared with the wild-type mFSH␤-luc plasmid response (Fig.  4A). Approximately the same level of reduction in the response to GnRH was achieved when putative AP-1 sites were mutated in the ovine FSH␤ promoter (10). However, mutation of the NF-Y site/AP-1 half-site to an AP-1 consensus sequence (labeled AP-1), dramatically increases GnRH induction of luciferase activity to 15-fold, a 6-fold increase over the level of induction of the wild-type sequence containing the novel AP-1 site, indicating that the site in the wild-type promoter is a low activity site. Such low activity sites have potentially important physiological roles in the regulatory regions of the genes such as FSH␤, which need to be tightly regulated in the course of the estrous cycle, but fluctuate only up to 4-fold in vivo (26).
To address whether the AP-1 element contributes independently to GnRH induction, we introduced a 9-bp deletion, elim-FIG. 3. Competition EMSA using oligonucleotides with scanning mutations as competitors reveals the base pairs required for NF-Y and AP-1 binding. A, alignment of wild-type sequence (WT) found in the mouse FSH␤ promoter Ϫ99/Ϫ65 and oligonucleotides used as competitors (labeled A-K) is shown. Scanning mutations were introduced into oligonucleotides A-K, and these changes are underlined. These unlabeled oligonucleotides were used in a 100-fold excess in EMSA experiments in B-D, whereas wild-type sequence (WT) was used as a probe. The NF-Y binding site is underlined with a solid line in the wild-type sequence, whereas the AP-1 half-site is underlined with a dashed line. B, nuclear extracts from control cells were subjected to EMSA with radiolabeled wild-type probe. Mutated oligonucleotides were used as competitors in a 100-fold excess in the corresponding lanes. In the lane labeled AP-1, the AP-1 consensus sequence was used as a competitor in the same manner. C, nuclear extracts from cells treated with GnRH for 6 h were subjected to EMSA with the wild-type probe, and the same oligonucleotides with mutations as above were used as competitors. D, NF-Y antibody was added to nuclear extracts from the cells treated with 10 nM GnRH for 6 h, and competition EMSA was performed with the wild-type probe and competitor oligonucleotides indicated above the lanes.
inating the entire novel AP-1 element (⌬AP-1) into the Ϫ398 mFSH␤-luc reporter. This deletion reduces responsiveness to GnRH by 25% compared with the wild-type mFSH␤-luc reporter, a similar level of reduction to that obtained with point mutations in the site (Fig. 4B). Therefore, a reporter plasmid without this novel AP-1 site retains a substantial response to GnRH but cannot reach the maximal induction obtained with the reporter that contains this site. To further characterize this site and determine whether it is sufficient to confer GnRH responsiveness to a heterologous promoter, we created a reporter with four copies of this AP-1 site linked to a minimal Ϫ81 bp herpes thymidine kinase promoter driving luciferase expression. This reporter was induced 4-fold by GnRH, whereas the control reporter containing only the thymidine kinase promoter was minimally induced 1.4-fold (Fig. 4C). This result indicates that, in addition to being essential for maximal induction by GnRH, this site is sufficient for GnRH response.
Multiple AP-1 Isoforms Are Induced by GnRH and Bind to the Novel AP-1 Site-When we identified AP-1 binding to this novel site by EMSA (Fig. 2B), we identified two complexes that changed following GnRH treatment. Both were supershifted with antibodies against Fos and/or Jun proteins. To determine whether these complexes differ in their protein components, we first tested whether one of these bands is AP-1 protein in complex with another factor, using in vitro transcribed/translated c-Jun and c-Fos proteins in EMSA. We found that AP-1 composed only of c-Jun and c-Fos binds DNA directly at the Ϫ99/Ϫ65 region as a single band (lane 4), not present in the reticulocyte lysate control (lane 1, Fig. 5A). This c-Jun/c-Fos complex co-migrates with complex 1 from nuclear extracts treated with GnRH (lane 3), indicating that this upper complex is indeed AP-1 factor alone. Furthermore, this result confirms that AP-1 can bind directly and does not serve only as a co-factor; nor does it need to be in complex with NF-Y to bind DNA in vitro.
Since AP-1 does not need another factor to bind to the Ϫ99/ Ϫ65 region, we postulated that the two different complexes observed following GnRH treatment are composed of different AP-1 isoforms. To determine which AP-1 isoforms are induced in L␤T2 cells following stimulation with GnRH, Western blots of whole cell lysates with and without GnRH treatment were performed. GnRH selectively induces c-Fos, c-Jun, FosB, and JunB but not JunD in L␤T2 cells (Fig. 5B). As expected from the EMSA results in which the NF-Y complex intensity did not change (observable after Fos was supershifted), the amount of NF-YA in the cells does not change with GnRH treatment.
To examine which of these GnRH-induced AP-1 isoforms binds to the FSH␤ promoter, isoform-specific antibodies were used in EMSA (Fig. 5C). Inclusion of an antibody to c-Fos induces a supershift in the upper, slower migrating AP-1 complex (complex 1), whereas an antibody to FosB supershifts the lower, faster migrating complex (complex 2). Non-isoform-specific antibodies, which recognize all of the Fos or all of the Jun isoforms (labeled ns in Fig. 5), supershifted both complexes. Both of the AP-1 bands appear to contain c-Jun and, to a lesser degree, JunB, since both bands were diminished upon inclusion of antibodies specific for c-Jun and JunB. Thus, the upper band (complex 1) contains c-Fos/Jun heterodimers, whereas the lower AP-1 band (complex 2) contains FosB/Jun heterodimers binding the AP-1 half-site in the mouse FSH␤ promoter. (27), was co-transfected with mFSH␤-luc into L␤T2 cells treated with vehicle (control) or GnRH. A-Fos has an acidic extension on the N terminus of the Fos leucine zipper, which physically interacts with the Jun basic region, thus preventing the basic region of the heterodimer from binding DNA (27). Introduction of A-Fos with mFSH␤-luc into L␤T2 cells has no effect on the expression of the reporter in untreated control cells (data not shown). In cells treated with 10 nM GnRH for 6 h, the dominant negative mutant of c-Fos (A-Fos) reduces GnRH induction of mFSH␤-luc by 30% compared with cells transfected with the vector control and treated with GnRH (Fig. 6A). The result demonstrates that functional AP-1 protein is necessary for maximal induction of FSH␤ by GnRH. A similar level of reduction in GnRH responsiveness was observed when the AP-1 site in the promoter was mutated (mutant I, Fig. 4). However, when this mutant I was used in transfection instead of wild-type mFSH␤-luc, A-Fos did not further reduce GnRH response (Fig. 6A). These data suggest that the novel AP-1 site is the only active AP-1 site in the promoter region tested. Fig. 3) were introduced into the mFSH␤-luc vector, and transfections were performed using L␤T2 cells. Cells were treated with 10 nM GnRH for 6 h, after which the luciferase activity was measured and normalized to ␤-galactosidase. The results are represented as -fold induction from the control cells transfected with the same reporter vector. The bar labeled AP-1 represents the induction of the reporter with a consensus AP-1 binding site introduced into the mFSH␤-luc reporter vector in place of the NF-Y/ AP-1 site. Significantly different induction in the treated cells versus the control cells for each reporter is marked with an asterisk, whereas a significant difference in induction of the mutated reporter from induction of the wild-type reporter is marked with a number sign. Results represent the mean Ϯ S.E. of four independent experiments, each performed in triplicate. B, a 9-bp deletion of the NF-Y/AP-1 site was created in mFSH␤-luc, and its induction following GnRH treatment was compared with the wild-type reporter. Significantly different induction in the treated cells versus the control cells for each reporter is marked with an asterisk, whereas a significant difference in induction of the mutated reporter from induction of the wild-type reporter is marked with a number sign. C, a reporter gene with four copies of the NF-Y/AP-1 element upstream of the thymidine kinase promoter was created, and its induction was compared with the induction of the control luciferase reporter driven by the thymidine kinase promoter. The activity from GnRH-treated cells was normalized to the activity from control cells, and results are represented as -fold induction. An asterisk marks that the induction of the reporter with multimerized AP-1 site is significantly different from the induction of the control plasmid.

FIG. 4. The AP-1 binding site is required for maximal induction with GnRH. A, mutations I and F (see
Additionally, c-Fos and c-Jun are sufficient to induce mouse FSH␤ promoter activity. c-Jun and c-Fos expression vectors, co-transfected with wild-type mFSH␤-luc into L␤T2 cells, induce mFSH␤-luc greater than 6-fold over the vector controltransfected cells (Fig. 6B). In contrast, c-Jun and c-Fos expression vectors do not significantly increase luciferase expression when the AP-1 site mutant I is used as a reporter. This again confirms that the novel AP-1 site we identified is the only site for AP-1 binding within the 398-bp regulatory sequence. Furthermore, it suggests that the GnRH responsiveness remaining in the Ϫ398 FSH␤ promoter after mutation of the AP-1 site is due to an activity unrelated to induction of AP-1.
MAPK Is Involved in FSH␤ Induction by GnRH through JunB and c-Fos-MAPK is acutely activated following GnRH treatment of L␤T2 cells, and this activation is involved in the induction of the ovine FSH␤ promoter (16). To test whether MAPK plays the same role in the induction of the mouse FSH␤ promoter by GnRH, we treated L␤T2 cells with the MEK inhibitor, UO126, for 30 min prior to and during GnRH treatment. We first established a dose response for MEK inhibition by UO126 in L␤T2 cells, by Western blotting for phospho-MAPK in whole cell lysates following the treatment (data not shown). The minimal concentration of the inhibitor needed to completely inhibit phosphorylation of MAPK by GnRH (1 M) was then used in our experiments. As expected, MAPK signal-ing plays a role in induction of the mouse FSH␤ promoter by GnRH, since this induction is reduced by 46% in the presence of the inhibitor (Fig. 7A).
To test whether levels of AP-1 proteins in L␤T2 cells were altered by inhibition of the MAPK pathway, Western blotting was performed following GnRH treatment. MEK inhibition prevents maximal JunB induction by GnRH (Fig. 7B); however, it does not reduce the induction of the c-Jun or FosB isoforms found to be induced by GnRH in Fig. 5B (data not shown). In accordance with previously published results (6), c-Fos levels are also reduced with MAPK inhibition (date not shown). Therefore, we conclude that the MAPK pathway is involved in GnRH induction of the FSH␤ gene, in part through the induction of JunB and c-Fos proteins.
AP-1 and NF-Y Interact and Co-occupy the Site in Vivo following GnRH Stimulation-Since the AP-1 binding site overlaps the NF-Y binding site, as demonstrated in competition EMSA in Fig. 3, we hypothesized that there are two possible ways that this region can accommodate these transcription factors. One possibility is that AP-1 displaces NF-Y on this site in vivo following Fos and Jun induction by GnRH. This was tested with a ChIP assay, which allows examination of the proteins binding to this gene region in vivo with and without GnRH treatment. After overnight serum starvation, the L␤T2 cells were treated with GnRH for 3 h. Then, after lysis and Equal amounts of protein from whole cell lysates were run on the gel, and after transfer, the membranes were probed with antibodies specific for the indicated proteins. After secondary antibody, enhanced chemiluminescence was performed, and the blots were exposed to film. C, EMSA using the Ϫ99/Ϫ65 sequence from the mouse FSH␤ promoter indicates that at least four different AP-1 isoforms bind the site. The length of GnRH treatment and the isoform specificity of the antibodies used for the supershift assay are indicated above the lanes. Lanes marked ns in the isoform lane included non-isoform-specific antibodies to Fos and Jun proteins, which therefore interacted with all isoforms of Fos and Jun families, respectively. sonication, chromatin was precipitated with antibodies specific to NF-YA, Fos, or Jun. DNA was extracted and subjected to PCR analysis, amplifying the FSH␤ gene regulatory sequence. The ChIP assay shows that Fos and Jun proteins bind the mouse FSH␤ promoter sequence more intensely following GnRH treatment, consistent with their induction by GnRH, whereas NF-Y binding to this region does not change (Fig. 8).
This site is the only CAATT box identified in the region amplified with the primers, and, to our knowledge, there are no other CAATT boxes in the proximity of this site. This assay assesses binding in a cell population, so if NF-Y were dislodged by AP-1 in even a portion of the cells following GnRH treatment, we would expect the NF-Y antibody-precipitated band to be diminished. Therefore, AP-1 does not appear to displace NF-Y on the mouse FSH␤ promoter.
An alternative mechanism by which these two transcription factors could occupy the same site is if AP-1 and NF-Y physically interact. We tested this hypothesis using GST pull-down assays in which in vitro transcribed and translated c-Fos and c-Jun proteins were tested for their ability to interact with NF-YA-GST fusion protein. In this assay, c-Jun interacts with the NF-YA subunit (Fig. 9, top panel). 35 S-Labeled c-Jun, but not c-Fos, precipitates with glutathione beads through an interaction with GST-NF-YA. Additionally, we performed the reverse experiment in which NF-Y proteins were labeled and synthesized in vitro and then tested for interaction with c-Jun-GST fusion protein. Labeled NF-YA is retained in the precipitate by the interaction with GST-c-Jun (Fig. 9, bottom left  panel), whereas NF-YB and NF-YC are not (data not shown). No interactions were observed using GST alone; nor did labeled green florescent protein, which serves as a control, interact with any of the used GST fusion proteins (Fig. 9, bottom right  panel). Thus, NF-Y and AP-1 form heteromeric complexes in vitro, through protein-protein interaction between Jun and NF-YA.

FIG. 6. Co-transfections using dominant negative Fos (A-Fos) indicate that AP-1 is necessary for maximal induction either by GnRH (A) or by overexpression of c-Jun and c-Fos (B).
A, cells transfected with wild-type mFSH␤-luc or mutation I introduced into mFSH␤-luc were treated for 6 h with vehicle (control) or 10 nM GnRH. Cells were co-transfected with an expression vector for dominant negative Fos (A-Fos) or its vector control to assess the necessity for AP-1 in the inductions by GnRH. Results represent the means of four independent experiments, each performed in triplicate. Results were analyzed by analysis of variance and Tukey-Kramer post hoc test, and an asterisk indicates a statistically significant difference from the control vehicletreated cells, whereas a number sign indicates a significant difference from GnRH-treated cells co-transfected with wild-type reporter and empty vector control for A-Fos. B, cells transfected with wild-type mFSH␤-luc or mutation I introduced into mFSH␤-luc were co-transfected with c-Jun and c-Fos and with either dominant negative A-Fos or vector control. Twenty-four h after transfection, luciferase activity was measured and normalized to ␤-galactosidase. None of the vector controls for either c-Jun, c-Fos, or A-Fos had any effect on reporter expression (data not shown). Results represent the means of four independent experiments, each performed in triplicate. An asterisk indicates a statistically significant difference from the control cells, whereas a number sign indicates a significant difference from cells transfected only with c-Jun and c-Fos.

FIG. 7. Inhibition of the MAPK pathway during GnRH treatment prevents maximal induction of FSH␤ and JunB.
A, cells transfected with wild-type mFSH␤-luc were treated for 6 h with vehicle (control) or 10 nM GnRH. Cells were co-treated with the MEK inhibitor, UO126, at 1 M to assess the necessity of the MAPK pathway in induction by GnRH. Results represent the mean of five independent experiments, each performed in triplicate. Results were analyzed by analysis of variance and Tukey-Kramer post hoc test, and an asterisk indicates a statistically significant difference from the control vehicletreated cells, whereas a number sign indicates a significant difference from GnRH-treated cells without UO126. B, Western blot of JunB in whole cell lysates of L␤T2 cells following 0, 1, or 3 h of GnRH treatment with or without the MEK inhibitor, UO126. The experiment was repeated three times, and a representative gel is shown.

DISCUSSION
GnRH is a key regulator of FSH␤ gene expression and therefore FSH synthesis. Previous studies were limited by a lack of available cell lines that express both FSH subunit genes and respond to GnRH. The scarce number of gonadotropes in the pituitary precludes performing these studies in primary cells. The genesis of the L␤T2 cell line that endogenously expresses FSH␤, allows for the dissection of molecular pathways governing its expression (13,28). In the current study, we have focused on delineating the mechanisms of GnRH induction of the mouse FSH␤ gene.
FSH regulates gonadal development in mammals and is required for folliculogenesis (1). Tight regulation of FSH levels is crucial for the menstrual or estrous cycle. Both FSH protein in circulation and ␤-subunit mRNA in the pituitary normally fluctuate 4-fold during the cycle (26,29). In mice lacking GnRH, serum FSH levels are 60 -90% lower (30). One pulse of GnRH administered to castrated, testosterone-replaced rats (with low endogenous GnRH) increased FSH␤ transcription 4-fold (31). This level is comparable with the induction observed in our studies using the L␤T2 cell line.
We report here that the AP-1 transcription factor, induced by GnRH, can bind a half-site of its consensus sequence, GTCA, when this half-site is juxtaposed to a site involved in basal expression, in this case a CCAAT box binding NF-Y. Further, we show that this AP-1/NF-Y site is necessary for maximal GnRH induction of the mouse FSH␤ gene. As shown in Fig. 4, the half-site is, as expected, a low affinity site for AP-1. Specifically, when we mutate the NF-Y site/AP-1 half-site in the FSH␤ promoter to create a full AP-1 consensus site in this position, luciferase expression after GnRH treatment is greater than 15-fold higher than the untreated control. This is 6-fold higher than GnRH induction of the wild-type promoter. Further, in gel shift assays, an AP-1 consensus competed more effectively for AP-1 bands than the nonlabeled wild-type sequence, and, when used as a probe, the AP-1 consensus requires one-tenth of the protein as wild-type probe to observe AP-1 binding (data not shown). Half-sites may have important physiological roles despite, or perhaps because of, their low affinity. Genes such as FSH␤ fluctuate only 4-fold, but this relatively modest change in expression during the estrous cycle is crucial for normal egg development and selection. A full consensus sequence might bring forth an unnecessarily high induction.
In a recent report describing the regulation of GnRH receptor expression by GnRH and activin in L␤T2 cells, the promoter element studied, GTCTAGTCAC, was of special interest (32). The authors conclude that AP-1 binds a novel 6-bp site, AGT-CAC, instead of its usual 7-bp site, whereas activin-regulated Smad 4 binds a 2-bp site. However, the competition EMSA experiments shown in that report suggest the alternative explanation that Smad 4 binds a Smad half-site GTCT, and AP-1 binds its half-site GTCA, which is separated from the Smad half-site by only one nucleotide. Thus, in the GnRH receptor gene, in light of our findings, it is possible that AP-1 and Smad 4 are stabilized on their respective half-sites by mutual interaction, and this is the reason both sites are needed for response to either activin or GnRH. It would be of interest to examine whether half-sites or low affinity sites are commonly involved FIG. 8. ChIP reveals that NF-Y binds DNA in the proximal mouse FSH␤ promoter in both control and cells treated with GnRH for 3 h, whereas AP-1 binds DNA following GnRH treatment. A, chromatin was isolated from L␤T2 cells treated with vehicle or 10 nM GnRH for 3 h and cross-linked with formaldehyde. After sonication, sheared chromatin was precipitated with the antibodies indicated above the lanes. The precipitated and purified DNA is then amplified in the PCR. The antibody specific for NF-Y precipitates the DNA specific for the sequence in the proximal mouse FSH␤ promoter, in both control and GnRH-treated cells. Fos and Jun, on the other hand, bind DNA in vivo only following the GnRH treatment. In the first two lanes, chromatin was precipitated with protein A beads only serving as controls. B, chromatin prior to precipitation serves as the control for the amount of chromatin used for precipitation in the untreated and GnRH-treated samples. A serial dilution of the chromatin was performed and then used in PCR together with precipitated samples. C, four independent experiments were quantified using a PhosphorImager and then normalized to intensity in the control sample precipitated with protein A beads only to normalize for any difference in the activity of the [␣- 32 9. GST pull-down assays demonstrate that NF-YA can interact directly with c-Jun. 35 S-Labeled proteins, indicated above each panel, were used in the binding assay with GST fusion proteins, labeled above each lane. GST fusion proteins were induced with isopropyl-␤-D-thiogalactosidase overnight, and the bacterial pellets were sonicated. These proteins were bound to glutathione-Sepharose beads, and in vitro transcribed and translated labeled proteins were added. After extensive washing, the precipitates were run on a gel and subjected to autoradiography.
in the induction of gonadotrope-specific genes that usually have a low level of induction but are very tightly regulated throughout the menstrual/estrous cycle.
Another interesting characteristic of the FSH␤ promoter element is that the AP-1 half-site overlaps an NF-Y site involved in basal expression. Recently, overlapping NF-Y and YY1 sites have been identified in the promoter of the Hox4b gene (33); however, this element was in a specialized intronic site able to bind either factor in a mutually exclusive manner. In the FSH␤ promoter, AP-1 and NF-Y occupy this element concurrently. In EMSA experiments with control extracts, NF-Y binds this site, and the complex is completely supershifted with antibodies to NF-YA. Competition EMSA in Fig. 3B also shows that a CCAAT box is needed for NF-Y to bind. The ChIP assay indicates that NF-Y is present in the complex at the same level before and after GnRH treatment in live cells. AP-1 can also bind this site in vitro, without other proteins present. In vitro transcribed and translated AP-1 binds this site, and it comigrates with the AP-1 complex from GnRH-treated cells. Thus, the FSH␤ promoter element can be bound by both NF-Y and AP-1.
AP-1 binding to this low affinity site may be stabilized by protein-protein interactions with the basal transcription factor NF-Y. Our attempts to observe a higher order complex in EMSA were unsuccessful; this is probably due to its expected size of 220 kDa, which would be too large to migrate into the gel. Further, the higher order complex would consist of five different proteins (NF-YA, NF-YB, and NF-YC, which are all necessary for NF-Y to bind, and Jun and Fos, which form AP-1) and therefore is difficult to reconstitute from recombinant proteins. However, we have established that this is a low affinity site, compared with the AP-1 consensus sequence, and that it is bound by NF-Y prior to and during GnRH treatment, which leads us to speculate that NF-Y may stabilize AP-1 binding. At least three such examples have been reported: SP1, SREBP1, and RF-X. In such cases, NF-Y considerably increases the affinity of the neighboring factor for DNA, making these complexes more stable on the DNA (18). In addition to physical interaction with many transcription factors (21,34,35), NF-Y interacts with several components of the basal transcriptional machinery (36 -38). Therefore, through direct contact with induced and/or activated transcription factors and the basal machinery, NF-Y may serve as a transcriptional coordinator or integrator.
We determined that AP-1 interacts with NF-Y through direct protein-protein contact between c-Jun and NF-YA. This is not surprising, since there is mounting evidence that NF-YA is the regulatory subunit of the trimeric NF-Y complex (18). c-Jun protein has also been reported to physically interact with other transcription factors, such as ER␣ (39), Cbfa1 (40), and Smads (41). How AP-1 is able to circumvent the spatial constraints of binding to a half-site directly adjacent to a NF-Y-occupied CAATT box, without a single nucleotide space, remains to be determined. AP-1 has been shown to bind a weak binding site when it cooperatively associates with transcription factors on juxtaposed sites (42); however, in that case, there were two base pairs between the binding sites. Fos and Jun heterodimers form a flexible fork, which might permit binding of other transcription factors at adjacent sites on the DNA (43). Since the mouse FSH␤ promoter contains an AP-1 half-site, it is possible that only one member of the heterodimer, in this case Fos, directly binds the DNA and that Jun binds to NF-Y, as we have demonstrated in vitro, as well as to Fos, but not directly to the DNA. Alternatively, it is possible that both Fos and Jun proteins contact the DNA. We determined that residues TATTG-GTCAT are needed for AP-1 to bind. Conserved residues in the AP-1 consensus TGAGTCA, are printed in boldface type and/or underlined for easier observation. The CCAAT element (opposite strand: ATTGG) is bound by NF-Y. However, one member of the AP-1 heterodimer can bind an underlined T, which is conserved in our novel site and the AP-1 consensus, and G (in boldface type), which both Fos and Jun bind according to the crystal structure (43). The other partner can bind the GTCA half of the consensus (in boldface type). Fos and Jun bind their site in the major groove, and only four amino acid residues contact the DNA (43). Vast portions of either protein are found perpendicular to the DNA. That conformation and the twist of the DNA helix may allow enough space for NF-Y to bind. NF-Y, on the other hand, contacts the DNA in the minor groove (44). From our studies, it appears that the NF-Y site has to be present for AP-1 to bind, and since NF-Y occupies the site in vivo, NF-Y may stabilize low affinity AP-1 DNA interactions.
We used 398 bp of the mouse FSH␤ regulatory sequence in these experiments, and this relatively short region has both higher expression and greater response to GnRH than the much longer ovine regulatory sequence used in previous studies (13,16). Thus, important species-specific differences in FSH regulation exist. The NF-Y/AP-1 site identified comprises sequences from Ϫ76 to Ϫ69 in the mouse FSH␤ promoter and is not conserved in the ovine or bovine promoters, although it is conserved in the human and rat. Two potential AP-1 sites previously identified in the ovine promoter correspond to Ϫ69/ Ϫ63 and Ϫ106/Ϫ100 sequences in the mouse FSH␤ promoter (10). When we used oligonucleotides spanning those sites as probes in EMSA, we did not detect AP-1 binding or any change in binding complexes following GnRH treatment (data not shown). The reports describing those AP-1 sites used purified proteins to detect binding to these sites (11) or detected AP-1 from extracts of GnRH-treated HeLa cells binding to the AP-1 consensus sequence (10). However, this is not surprising, since GnRH induces Fos and Jun isoforms. Further, mutations in these sites do not affect appropriate regulation of FSH by gonadectomy or GnRH antagonist in transgenic animals (12).
Two AP-1 consensus sites exist within the 398 bp of the mouse regulatory region, at Ϫ10/Ϫ4 and Ϫ181/Ϫ175. As expected, AP-1 from GnRH-treated nuclear extracts can bind the AP-1 consensus sequence. However, based on the result of our transfection experiments, we conclude that those sites are not functional in the context of the promoter. Namely, when we used a reporter containing a mutation in the AP-1 half-site, mutant I, the induction by GnRH decreased by about the same amount as when we co-transfected the dominant negative A-Fos in Fig. 6. Notably, dominant negative Fos cannot reduce the induction by either GnRH or overexpression of c-Jun and c-Fos when mutant I is used instead of the wild-type promoter. This finding strongly suggests that this site is required for AP-1 action in this regulatory region. From the truncation analysis, we determined that there is another region of the promoter between Ϫ230 and Ϫ304 responsive to GnRH to a lesser degree, since there was a statistically significant drop in the induction level upon truncation of that region. Furthermore, it is possible that another GnRH-responsive site exists in the region proximal to the start site, since the Ϫ95 bp truncation maintains two-thirds of the response, whereas the AP-1 site we identified is responsible for one-third of the induction. We are currently investigating the elements in those regions for their role in GnRH response.
The GnRH receptor belongs to the class of G protein-coupled receptors and, upon ligand binding, activates protein kinase C and downstream MAPK signaling pathways (6). Our results showing the role of MAPK in mouse FSH␤ induction are in agreement with previously published reports that MAPK is involved in ovine FSH␤ up-regulation and c-Fos induction by GnRH. Our data extend those findings and demonstrate that MAPK is involved in JunB induction by GnRH but not in c-Jun or FosB induction. However, mutation of the AP-1 site or introduction of the dominant negative Fos reduces induction of FSH␤ by 35 and 30%, respectively, whereas the inhibition of the MAPK pathway causes 46% reduction in the FSH␤ response to GnRH. Thus, the MAPK pathway probably regulates other transcription factor(s) in addition to AP-1, either through activation by phosphorylation or by induction of their gene expression.
In this report, we demonstrate that AP-1 binds a novel site composed of an AP-1 half-site and a CCAAT box. This site is involved in GnRH induction of the mouse FSH␤ gene. The AP-1 half-site overlaps a site involved in basal expression, which is occupied by NF-Y in vivo. Direct physical contact between AP-1 and NF-Y may stabilize AP-1 on this low affinity site, and NF-Y may serve as an integrator between hormonal signals and the transcriptional machinery.