Regulation of the Cell-specific Calcitonin/Calcitonin Gene-related Peptide Enhancer by USF and the Foxa2 Forkhead Protein*

An 18-bp enhancer controls cell-specific expression of the calcitonin/calcitonin gene-related peptide gene. The enhancer is bound by a heterodimer of the bHLH-Zip protein USF-1 and -2 and a cell-specific factor from thyroid C cell lines. In this report we have identified the cell-specific factor as the forkhead protein Foxa2 (previously HNF-3β). Binding of Foxa2 to the 18-bp enhancer was demonstrated using electrophoretic mobility shift assays. The cell-specific DNA-protein complex was selectively competed by a series of Foxa2 DNA binding sites, and the addition of Foxa2 antiserum supershifted the complex. Likewise, a complex similar to that seen with extracts from thyroid C cell lines was generated using an extract from heterologous cells expressing recombinant Foxa2. Interestingly, overexpression of Foxa2 activated the 18-bp enhancer in heterologous cells but only in the presence of the adjacent helix-loop-helix motif. Likewise, coexpression of USF proteins with Foxa2 yielded greater activation than by Foxa2 alone. Unexpectedly, Foxa2 overexpression repressed activity in the CA77 thyroid C cell line, suggesting that Foxa2 may interact with additional cofactors. The stimulatory role of Foxa2 at the calcitonin/calcitonin gene-related peptide gene enhancer was confirmed by short interfering RNA-mediated knockdown of Foxa2. As seen with Foxa2 overexpression, the effect of Foxa2 knockdown also required the adjacent helix-loop-helix motif. These results provide the first evidence for combinatorial control of gene expression by bHLH-Zip and forkhead proteins.

The calcitonin/calcitonin gene-related peptide (CT 1 /CGRP) gene expresses the hormone CT in thyroid C cells and the neuropeptide CGRP in neurons. CT is a therapeutically useful modulator of bone metabolism and calcium homeostasis (1), but it may also play additional roles based on the recent unexpected finding of increased bone mass in CT/CGRP and CT receptor knock-out mice (2)(3)(4). CGRP is an alternative splicing product from the CT/CGRP gene (5) that is now appreciated to be a member of a gene family of multifunctional neuropeptides that act on related receptors (6 -8). Most notably, CGRP is the most potent peptide dilator known (9,10), and abnormal levels of CGRP have been implicated in neurovascular and cardiovascular disorders (7,(11)(12)(13). Importantly, CGRP antagonists have recently been reported to be an effective treatment for migraine (14).
Regulation of CGRP expression in response to extracellular stimuli is controlled exclusively at the transcriptional level. Studies on both the human and rat CT/CGRP gene have identified a cell-specific enhancer containing helix-loop-helix (HLH) motifs ϳ1 kb upstream of the transcription start site (15)(16)(17)(18)(19). We have shown that an 18-bp element with a single HLH site and a flanking octamer-like motif is both sufficient and necessary for cell-specific enhancer activity (18,19). Using nuclear extracts prepared from the neuronal-like CA77 thyroid C cell line, bHLH-Zip proteins USF-1 and -2 were shown to bind the enhancer in vitro (19). Although Oct1 can bind the A/T-rich half of the flanking site, Oct1 cannot transactivate the enhancer (18). Instead, a cell-specific protein referred to as OB2 was shown to bind an adjacent 8-bp site that partially overlaps the HLH site (19). OB2 was identified as an ϳ68-kDa protein by UV-cross-linking and is present in human and rat thyroid C cell lines but not in cell lines that do not express the CT/CGRP gene (19).
In this study have shown that OB2 is the forkhead protein Foxa2, formerly called HNF-3␤. We demonstrate that Foxa2 is able to bind and activate the CT/CGRP enhancer in heterologous cells. Importantly, this transactivation requires the adjacent HLH site and is increased upon co-expression of USF. The combinatorial role of Foxa2 and USF was further demonstrated by short interfering RNA (siRNA)-mediated knockdown of Foxa2 levels, which reduced enhancer activity but only in the presence of the HLH site. These results demonstrate that Foxa2 and USF cooperate to regulate CT/CGRP gene expression in thyroid C cells.
The rat CT/CGRP promoter and herpes simplex thymidine kinase (TK) promoter luciferase plasmids and the empty expression vector CMV-5 have been described previously (18 -20). The HLH mutation plasmid has the enhancer sequence 5ЈCAGcggatccgCTGTGCAAT3Ј (BamHI linker in lowercase), which contains a disrupted HLH motif with a reconstituted OB2 site. The CMV-Foxa2 (HNF-3␤) expression vector containing the full-length rat cDNA (1.6-kb EcoRI fragment) and two truncated Foxa2 plasmids expressed from the cytomegalovirus promoter and artificial ATGs were kindly provided by R. Costa, University of Illinois. Both include the winged helix DNA binding domain but are transcriptionally inactive. They begin with amino acids 122 and 153. The Foxa2 isoform-3 plasmid was generated by PCR from CMV-Foxa2 plasmid DNA using primers with EcoRI and BamHI restriction sites. The upstream primer was 5Ј-CGGAATTCTACTCTTCCGTGAG-CAAC-3Ј, and the downstream primer was 5Ј-CGCGGATCCGGAC-GAGTTCATAATAG-3Ј. The PCR product was gel-purified using the QIAquick gel extraction kit (Qiagen), cloned into pcDNA 3.1 (Ϫ)/Myc-His vector (Invitrogen), and confirmed by DNA sequencing. The human USF-1 (pN3) and mouse USF-2 (pN4) in the pSG5 expression vector were originally provided by Dr. M. Sawadogo (University of Texas, MD Anderson Cancer Center).
CA77 and COS-7 cells were transfected by electroporation as described previously (19). Approximately 1-2 ϫ 10 6 cells were transfected with 5-10 g of luciferase reporter DNA and 0.5-10 g of expression vectors. The transfected cells from a single cuvette were grown on a 60-mm dish for 16 -24 h. For all experiments, the amount of DNA transfected into the cells was kept constant by the addition of CMV-5 or pcDNA3.1 empty vector DNA. In later experiments, including all of the siRNA experiments, the CA77 cells were transfected by LipofectAMINE 2000 (Invitrogen) as described previously (22). NCI-H460 cells were transfected by electroporation essentially as described (21). The conditions were 0.8 ml of cells (1 ϫ 10 7 cells/ml) in a 0.4-cm cuvette with 10 g of luciferase reporter DNA and 13 g of each expression vector DNA using a Bio-Rad Gene Pulser II at 260V, 1050 F. Immediately after electroporation, 0.4 ml of cells were pipetted into medium in a 10-cm dish and cultured for 20 -24 h prior to harvest. All cells were harvested using 1ϫ reporter lysis buffer (Promega) and assayed for luciferase activity using reagents from Promega. Protein concentrations were determined by Bradford assays (Bio-Rad). Each experimental condition was repeated in at least three independent experiments in duplicate unless otherwise noted. Statistical significance was determined by Student's t test (paired samples).
Electrophoretic Mobility Shift Assay (EMSA)-Electrophoretic mobility shift assays with the CT/CGRP 18-bp enhancer as a probe were performed as described (19). The probe was prepared by annealing 10 pmol of complementary oligonucleotides with overhanging BamHI ends (lowercase) (5Ј-gatccGGCAGCTGTGCAAATCCTg-3Ј, 5Ј-gatccAGGAT-TTGCACAGCTGCCg-3Ј) and labeled with [ 32 P]dATP using Klenow polymerase. All oligonucleotides used as competitors were either bluntended or had BamHI ends that were Klenow-filled. The binding reactions contained 0.02 pmol of labeled probe (50,000 cpm), 3 g of nuclear extract, binding buffer (10 mM Tris, pH 7.5, 5% glycerol, 50 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol), 0.1 g poly(dI⅐dC), and 0.1 pmol of an unrelated double-stranded oligonucleotide (5Ј-GATCCACTATGTCTA-GAG-3Ј) that eliminated a nonspecific complex. Competitor DNAs and peptides were preincubated for 10 min on ice with nuclear extract before the addition of probe. Antibodies were incubated for 10 -15 min after the addition of probe. The rabbit antisera against the N terminus of rat HNF-3␣ (amino acids 7-103) and HNF-3␤ (amino acids 7-86) used for EMSA were kindly provided by R. Costa. All other antibodies were rabbit polyclonal IgG obtained from Santa Cruz Biotechnology. All reactions were incubated for 10 min on ice before the addition of 4 l of 20% Ficoll dyes or 3 l of 50% glycerol dyes, and then resolved by electrophoresis through a 6% nondenaturing polyacrylamide gel (1:29 bis/acrylamide) and exposed to film overnight.
Transfection of siRNA-CA77 cells were grown to 50 -90% confluency and then trypsinized, collected by centrifugation, and resuspended in fresh medium without antibiotics. Approximately 225,000 cells in 1 ml were added to each well of a 12-well plate (Falcon). For each well, a solution containing 0.14 g of siRNA and 0.28 g of reporter plasmid DNA (ratio 1:2) was added to Eppendorf tubes containing 50 l of prewarmed Opti-MEM I reduced serum medium (Invitrogen). In tubes without control or Fox siRNA, the empty plasmid CMV5 was included. LipofectAMINE 2000 (1 l of a 1 mg/ml stock) was added to additional tubes containing 50 l of Opti-MEM I for 5 min. The transfection reagent and siRNA were mixed and incubated 20 min. siRNA/DNA-LipofectAMINE 2000 complexes (100 l) were added to the wells dropwise (before the cells had adhered). For the lower dosages of Foxa2 siRNA, 0.02 g of siRNA was combined with 0.12 g of non-silencing siRNA and 0.28 g of reporter plasmid DNA (ratio 1:2). The control samples always contained the equivalent amount (0.14 g) of the nonsilencing siRNA. Each condition was conducted in duplicate or triplicate. Fresh medium was added 24 -48 h after transfection. Approximately 72 h after transfection, cells were washed in phosphate buffered saline (PBS) and scraped, and lysates were prepared in 50 l of 1ϫ reporter lysis buffer (Promega) with a single round of freeze-thawing to aid lysis. Protein was determined by the Bradford method.
For cell sorting, the transfections were scaled up to 12 wells for each sample. Cell sorting was carried out ϳ72 h after transfection using a BD Biosciences fluorescence-activated cell sorter (FACS) DiVA. Cells were removed by trypsinization, pooled, collected by centrifugation, and resuspended in 1 ml of medium at about 10 6 cells/ml. Each cell suspension was filtered through 70-m mesh filters (Falcon 2350) to remove foreign particles, and transferred to 12 ϫ 75 mm tubes (Falcon 2052) on ice. GFP-positive cells were sorted into tubes containing 5 ml of medium. Cells were collected by centrifugation, washed in PBS, transferred to Eppendorf tubes, and spun down in a microcentrifuge for 5 min. Lysates were prepared by resuspending the cells in 100 l of 1ϫ reporter lysis buffer (Promega) and freezing samples at Ϫ20°C.
Western Blots-Equal amounts of cell lysate (3-8 g in different experiments) were resolved by 10% SDS-PAGE for 1 h at 30mA. The Precision Plus Protein (Bio-Rad) ladder was used as size standards. Proteins were transferred to Immobilon-P membranes (Millipore Corp.) overnight at 4°C at 30V. The membranes were blocked with 5% nonfat dry milk in 0.1% Tween-20 in PBS (PBST) for 1 h at room temperature followed by a 5-min wash in PBST. Primary antibody was a 1:1000 dilution of sheep anti-HNF3␤ IgG (Upstate Biotechnology) in PBST, 0.1% bovine serum albumin, 0.02% sodium azide, which was incubated either for 2 h at room temperature or overnight at 4°C. The membranes were washed in PBST for 15 min and then incubated with a 1:5000 dilution of donkey anti-sheep IgG conjugated to horseradish peroxidase (Santa Cruz Biotechnology) with 0.1% bovine serum albumin for 30 min. After extensive washing with PBST, the immunoreactive bands were visualized using the enhanced chemiluminescence reagents (Amersham Biosciences). Membranes were placed back in 5% milk in PBST and then probed using an antibody that recognizes ␣-tubulin or glyceraldehyde-3-phosphate dehydrogenase (Santa Cruz Biotechnology). Signal intensities were estimated by densitometry of multiple exposed films using NIH Image 1.63 software.
Glutathione S-Transferase (GST) Pulldown-Full-length Foxa2 was subcloned into the EcoRI site of pGEX-6P2 (Amersham Biosciences). The TNT T7 coupled reticulocyte lysate system (Promega) was used to make [ 35 S]Met-labeled USF-1, USF-2, and PITX2a proteins. The labeled proteins were made in a total volume of 50 l with 3 l of [ 35 S]methionine for 30°C for 90 min as recommended by the manufacturer. 20 l of labeled proteins were incubated with 50 l of GST-coated or GST-Foxa2-coated beads in PBS, 1% bovine serum albumin, 0.3% Nonidet P-40, 1 M dithiothreitol, 300 ng/ml ethidium bromide for 1.5 h at 4°C. The beads were washed with 750 l of binding buffer (minus the ethidium bromide) six times on ice; this was followed by two washes with 750 l of 10 mM Tris-Cl, 50 mM NaCl, pH 7.0. The beads were heated at 95°C, and the bound proteins were resolved by electrophoresis on a 10% SDS-polyacrylamide gel. The gel was dried and exposed to autoradiographic film.

RESULTS
Identification of Foxa2 as OB2-The cell-specific CT/CGRP 18-bp enhancer contains overlapping binding sites for a USF-1 and -2 heterodimer and a cell-specific protein that we have referred to previously as OB2 (Fig. 1A). The OB2 binding site has been defined previously as an 8-bp element that overlaps the HLH motif by 2-bp (19). It should be noted that upon competition of OB2, unidentified complexes near USF and Oct-1 were often (but not always) observed. The identification of OB2 came about by serendipity when we used a liver nuclear extract as a "negative control" for EMSAs. Unexpectedly, a complex was observed that had the same binding properties in EMSAs as the OB2 complex from the CA77 thyroid C cells (data not shown). A survey of known liver transcription factors and their binding sites revealed the forkhead protein Foxa2 as a likely candidate. Foxa2 has approximately the same size (47 kDa) as predicted from the 68-kDa OB2 protein that we had identified previously by UV-cross-linking from CA77 cells (19). The difference in size could be caused by the cross-linked DNA. We tested several known Foxa2 sites, and all were found to compete for OB2 in the EMSA (Fig. 1B). In contrast, oligonucleotides containing similar sites recognized by C/EBP and other factors were not able to compete for OB2. Likewise, mutation of the OB2 site to create palindromic versions of the site did not compete for binding. The most effective competitor was HNF3 #4 (24) (Fig. 1B). As little as a 5-fold molar excess of this Foxa2 binding element was able to specifically compete with the 18-bp enhancer for binding to the OB2 complex (Fig. 2).
To prove that Foxa2 bound the CT/CGRP enhancer we used a Foxa2 antiserum. Addition of Foxa2 antiserum eliminated HLH and OB2 proteins bind independently to overlapping motifs (boxed). Substitutions of the GC dinucleotide (lightly shaded region to the left of the OB2 site) reduce but do not eliminate OB2 binding. The cAMP-and ras-responsive elements at Ϫ250-bp and non-cell-specific enhancer sites (hatched boxes) that flank the 18-bp enhancer are also shown. Sequences of oligonucleotides used in this paper are listed and categorized by their ability to bind OB2 based on EMSA competition assays. The HNF and HFH oligos all bind Foxa2. All oligonucleotides have flanking BamHI ends (not shown). The indicated consensus OB2 site was determined from 32 oligonucleotides used as competitors in DNA binding assays. B, EMSA of HLH-OB2 complexes from CA77 nuclear extracts using the 18-bp enhancer as a radioactive probe and the indicated oligonucleotides as nonradioactive competitors (50-fold molar excess). The Oct1, USF1/2 heterodimer, and OB2 complexes are indicated. the OB2 complex and yielded a supershifted complex (Fig. 2). To reduce the chance that another complex might be missed that comigrated with Foxa2, the supershift was done under both low (Fig. 2, left panel) and high (Fig. 2, right panel) stringency conditions. Under low stringency conditions, a nonspecific complex was seen below OB2 and was removed by competition with all oligos, including a random linker DNA. As controls, OB2 was not shifted by antisera against HNF-3␣ (Fig.  2) or eight other factors (C/EBP-␣, C/EBP-␤, Ets-1, Ets-2, c-Rel, SRF, Elk1, Hmx3) (data not shown).
To confirm that Foxa2 can bind the 18-bp enhancer, we used the heterologous COS-7 cells. These cells do not express Foxa2 protein based on Western blots (data not shown). An OB2 complex was detected using nuclear extracts from COS-7 cells transfected with a Foxa2 expression vector (Fig. 3). As a control, EMSAs with nuclear extracts prepared from untransfected COS-7 cells did not yield an OB2 complex. The complex seen with extracts from the Foxa2 complexes comigrated with the OB2 complex from CA77 nuclear extracts and was specifically competed by an oligonucleotide containing a Foxa2 binding site (Fig. 3). Thus, by competition, supershift, and expression binding studies, OB2 is the forkhead protein Foxa2.
Foxa2 Overexpression Activates the 18-bp Enhancer in Heterologous Cells-Having demonstrated that Foxa2 binds the enhancer, we then asked if expression of Foxa2, along with USF-1 and -2, was sufficient for enhancer activity. We used the heterologous COS-7 cell line, which as mentioned above does not contain endogenous Foxa2 but does contain USF-1 and -2. The COS-7 cells do not have 18-bp enhancer activity as seen by the comparable activity of the reporter genes that contain the TK promoter and the TK promoter with the multimerized 18-bp enhancer (Fig. 4). Likewise, mutation of the HLH site of the 18-bp enhancer within the context of the 1.25-kb CT/CGRP promoter (1.25-CT/CGRP Bam mutant) had no effect on reporter activity in the COS-7 cells. Co-transfection of a Foxa2 expression vector increased activity of the 18-bp enhancer 2-fold (Fig. 4). Although small, this increase was specific for the 18-bp enhancer because Foxa2 did not activate the parental TK promoter reporter. This activation is consistent with the detection of Foxa2 binding activity following transfection of COS-7 with a Foxa2 expression vector in Fig. 3.
Interestingly, Foxa2 was able to activate the 1.25-kb CT/ CGRP promoter but not the 1.25-kb CT/CGRP Bam mutant  promoter with the mutated HLH site in the 18-bp enhancer (Fig. 4). This implicates coordinate action of Foxa2 with HLH proteins at the enhancer because the Foxa2 binding site is not directly affected by this mutation. Similar results were seen in HeLa cells (data not shown). This is consistent with our previous findings that both the HLH and OB2 motifs are required for enhancer activity (18) and indicates that Foxa2 activity requires the adjacent HLH motif.
To further explore the dependence of Foxa2 upon the HLH site, we cotransfected both Foxa2 and USF expression vectors into NCI-H460 cells. This non-small cell lung cancer line was chosen because it expresses relatively low levels of USF-1 and -2 proteins, and overexpression of USF in this cell type has been shown previously to strongly activate the arginine vasopressin promoter (21). The need for a cell line with low USF levels was driven by our finding that overexpression of USF-1 and -2, with or without Foxa2, had little or no effect on enhancer activity in COS-7 cells (data not shown). One possible explanation for the lack of activation by overexpressed USF proteins is that these proteins are already present in COS-7 cells. NCI-H460 cells do not have Foxa2 protein that is detectable on Western blots, and reporter genes have similar activities with or without the 18-bp enhancer (data not shown). Thus, the NCI-H460 cells have low levels of USF and do not have intrinsic 18-bp enhancer activity.
Transfection of USF expression vectors activated the 1.25-kb CT/CGRP promoter in NCI-H460 cells (Fig. 5A). Overexpression of USF-2 consistently yielded greater activation than USF-1 alone, but the greatest activation was seen with the combination of USF-1 and USF-2. This agrees with our previous report that the enhancer is predominantly bound by a USF-1 and -2 heterodimer (19). The cells were then transfected with the Foxa2 expression vector. Expression of Foxa2 yielded 3-4-fold activation of CT/CGRP promoter activity (Fig. 5B). When we coexpressed USF-1 and USF-2 with Foxa2, there was a significantly greater (7-fold) activation of the CT/CGRP promoter (Fig. 5C). As a control, the cells were also transfected with a Foxa2 isoform that lacks the first 30 amino acids (isoform 3) (Fig. 5, B and C). Isoform 3 was previously reported to be unable to activate the glucagon promoter (25). The truncated Foxa2 did not activate the CT/CGRP promoter, and the observed decrease in promoter activity suggests that isoform 3 may be acting as a dominant negative protein. In addition, coexpression of USF-1 and USF-2 did not restore transcriptional activity to the Foxa2 isoform 3 mutant. Taken together with the finding that Foxa2 activation requires the adjacent HLH site, these observations support the conclusion that the CT/CGRP enhancer is activated by the coordinate action of Foxa2 and USF proteins.
Given the functional relationship between Foxa2 and the USF proteins, we asked if these proteins bind each other in vitro. There was no detectable binding of USF-1 or USF-2 to agarose beads containing GST-Foxa2 protein (data not shown). As controls, we observed binding of the homeodomain protein PITX2a to GST-Foxa2 and there was no detectable binding of either PITX2a or USF proteins to GST agarose. The binding between Foxa2 and PITX2a, although not previously reported, is consistent with the prediction by Foucher et al. (26) that Foxa2 can bind many, if not all, homeodomain proteins.
Foxa2 Is Required for 18-bp Enhancer Activity in CA77 Cells-Then we tested the effect of overexpressing Foxa2 in the CA77 cells, which contain endogenous Foxa2. Instead of activation as seen in the heterologous cells, there was an unexpected 2-4-fold decreased enhancer activity in the CA77 cells (Fig. 6A). This observation was unexpected because mutation of the Foxa2 binding site clearly decreased enhancer activity (18,19). The repression by Foxa2 overexpression was specific for the 18-bp enhancer activity because the minimal TK promoter reporter was not repressed. Repression was seen in the context of 1.25-kb of flanking DNA and required the HLH site because the mutant HLH reporter was not repressed (Fig. 6A). Using less Foxa2 expression vector yielded less repression of the enhancer, but activation was still not seen (data not shown). The repression appears to require the full-length Foxa2 protein because transfection of expression vectors encoding only the DNA binding and C-terminal domains (Fig. 5B) did not affect 18-bp enhancer activity in CA77 cells (Fig. 6B). A caveat of this finding is that we could not confirm expression of the DNA binding domain fragments because they would not be recognized by the antisera. One interpretation of these results is that Foxa2 is not limiting in the CA77 cells and that the transactivation domain of the overexpressed protein "squelches" promoter activity by binding other factors (see "Discussion").
To resolve the unexpected observation that overexpression of Foxa2 inhibited enhancer activity in CA77 cells, we turned to a knockdown strategy. An RNA interference approach was used to reduce the endogenous Foxa2 protein level in CA77 cells. A series of double-stranded siRNAs were generated that are complementary to specific regions of the Foxa2 mRNA sequence. These siRNAs were co-transfected into CA77 cells with the 1.25-kb CT/CGRP promoter reporter gene. One of the siRNA duplexes tested, Fox-si2, repressed CT/CGRP enhancer activity 30-fold (Fig. 7A). Transfection of two other siRNAs directed against Foxa2 (Fox-si1 and -si4) had less of an effect on promoter activity (Fig 7A). The incubation time following transfection was an important parameter. There was little or no inhibition of promoter activity or knockdown of Foxa2 levels at 24 h following transfection, so all experiments were extended to 72-h incubations. As a control, the cells were transfected with a non-silencing siRNA that is predicted to not target any gene. The control siRNA did not affect promoter activity (Fig. 7A).
As a control for target specificity and to further test the role of the adjacent HLH site, we used the 1.25-kb CT/CGRP promoter containing the mutant HLH motif. The effect of two concentrations of Fox-si2 siRNA was compared between the wild-type and mutant promoters. Transfection with 0.14 g of Fox-si2 siRNA repressed activity to 3% of control activity, and 0.02 g of siRNA yielded 40% of control (Fig. 7B). In contrast, the mutant HLH reporter was not repressed by low concentrations of Foxa2 siRNA. At the higher siRNA concentration, the HLH mutant reporter was repressed, but only to 65% of control values in contrast to the 3% level seen with the wild-type promoter (Fig. 7B). These results strongly support the specificity of siRNA action on the 18-bp enhancer and the involvement of the HLH site in Foxa2 action.
To confirm that the siRNA transfection reduced Foxa2 levels, it was necessary to enrich for the transfected cells by FACS prior to analysis by Western blots. Analysis of total cell lysates did not reveal any knockdown of Foxa2, presumably because only ϳ10 -20% of the cells were transfected (data not shown). We repeated the siRNA transfection conditions with inclusion of the CMV-eGFP reporter to allow detection of transfected cells. Following FACS selection, a specific knockdown of Foxa2 protein levels was observed (Fig. 7C). As a control, the western membrane was re-probed with antibodies directed against either ␣-tubulin or glyceraldehyde-3-phosphate dehydrogenase proteins. These proteins were not reduced by the Foxa2 siRNA. Following normalization of the Foxa2 signals to the ␣-tubulin and glyceraldehyde-3-phosphate dehydrogenase signals, we estimate that there was about a 2-fold knockdown of Foxa2 protein levels in the selected cells. These results support the conclusion that Foxa2 is required for CT/CGRP enhancer activity in CA77 cells. DISCUSSION A single 18-bp enhancer controls both cell-specific and regulated transcription of the CT/CGRP gene. This multifunctional enhancer is synergistically activated by factors that bind an HLH motif and an adjacent octamer-like element (18). We have shown previously that the ubiquitous USF-1 and -2 heterodimer binds the HLH motif and that a cell-specific protein binds the adjacent motif (19). We have now identified the cell-specific protein as the forkhead protein Foxa2 (formerly HNF-3␤). Foxa2 is a member of the forkhead family of winged helix DNA binding proteins that play roles in embryogenesis, tumorogenesis, and maintenance of cell differentiation (27,28). Foxa2 has been proposed to act by remodeling chromatin and to functionally interact with multiple factors (26, 29 -32). Indeed, interactions between cell-specific factors and other regulators are believed to underlie tissue-specific control by the ubiquitous USF proteins (33)(34)(35)(36)(37)(38). Our overexpression and knockdown data indicate that Foxa2 activation of the CT/CGRP 18-bp enhancer requires the adjacent HLH site that is bound by a USF-1 and -2 heterodimer. Combinatorial control by a ubiquitous and a cell-specific factor fits the emerging pattern for The combination of Foxa2 and USF proteins yielded a significantly greater activation of the reporter than Foxa2 or USF-1 or -2 alone (p Ͻ 0.05). The mean and S.E. relative to the reporter alone is shown from n ϭ 19 -20, except for the Foxa2 isoform 3 with and without cotransfection of USF-1 or -2, which is the mean and range from n ϭ 2. many genes, but to our knowledge this is the first report of an enhancer controlled by USF and a forkhead protein. Based on our in vitro data, this communication does not appear to involve direct binding between the two proteins, although further experiments will be needed to address the possibility of interactions that might occur in the context of the nucleus.
Foxa2 was originally believed to be primarily involved in hepatic gene functions, and indeed low levels of CT mRNA have been reported in the liver (39). However, it is now clear that Foxa2 plays important roles outside the liver, including the control of neuronal and neuroendocrine genes. Foxa2 is expressed in pancreatic endocrine cells where it is involved in the regulation of the glucagon, insulin, and pdx-1 genes (40). Another neuronal and neuroendocrine Foxa2 target gene is the neuronal aromatic amino acid decarboxylase gene (41). Interestingly, this promoter is very active in the CA77 thyroid C cell line (42). Foxa2 is also induced during neuronal differentiation of P19 cells (43), and is expressed in the embryonic neural floor plate and ventral midline cells of the central nervous system (44). In addition to Foxa2, FIG. 6. Effect of Foxa2 overexpression on the 18-bp enhancer in CA77 cells. A, the indicated luciferase reporter genes were cotransfected with either the CMV-Foxa2 (ϩ) or an empty expression vector (Ϫ). Promoter activity was measured from the TK promoter with and without the multimerized 18-bp enhancer and the 1.25-kb CT/CGRP promoter, with and without a BamHI linker mutation within the HLH motif. The mean and S.E. relative to the TK-luciferase reporter with empty expression vector is shown. B, cytomegalovirus expression vectors encoding truncated Foxa2 proteins do not repress 18-bp enhancer activity. The truncated proteins contain the indicated amino acids. The luciferase activity from a single experiment with the TK and 18-bp-TK promoters is shown, although similar results were seen in an independent experiment using reporter genes containing the SV40 promoter (pGL3 promoter) and the pGL3-SV40 promoter with the multimerized 18-bp enhancer.
FIG. 7. siRNA mediated knockdown of endogenous Foxa2 protein and enhancer activity. A, cotransfection of the 1.25-kb CT/ CGRP promoter reporter with siRNA duplexes into CA77 cells. None indicates pCMV5 plasmid in place of siRNA, Con-nsi is the control non-silencing siRNA and Fox-si1, Fox-si2, and Fox-si4 are the Foxa2 siRNA duplexes. A fourth siRNA (Fox-si3) had irreproducible effects (data not shown). The mean and S.E. of three independent experiments in duplicate is shown. B, cultures were transfected with the 1.25-kb CT/CGRP promoter or 1.25-kb CT/CGRP promoter with the BamHI mutation in the HLH motif of the 18-bp enhancer to test the specificity of siRNA action. Data were normalized to the each promoter reporter with control non-silencing siRNA. Cells were treated with either 0.14 g of Fox-si2 siRNA or 0.02 g of Fox-si2 siRNA plus 0.12 g of control non-silencing siRNA (total 0.14 g) or 0.14 g of control non-silencing siRNA, as indicated. The mean and S.E. (n ϭ 3-9) relative to each reporter with 0.14 g of control non-silencing siRNA is shown. C, Western blots of lysates prepared from FACS-selected GFP-positive CA77 cells. Cells were transfected with CMV-eGFP with pCMV5 (None), Fox-si2 siRNA, or Con-nsi siRNA. The membrane was first incubated with Foxa2 antiserum and then re-incubated with antiserum against ␣-tubulin as a control. The Foxa2 (47-kDa) and ␣-tubulin (51-54-kDa) proteins are indicated (two bands were observed with the ␣-tubulin blot). As an additional control, the same lysates were analyzed by Western blots using first Foxa2 antiserum, then glyceraldehyde-3-phosphate dehydrogenase (37-kDa) antiserum as indicated.
there are related family members in the brain (45,46). However, although Foxa2 is active in the nervous system, we could not detect Foxa2 mRNA in the adult rat trigeminal ganglia by reverse transcription-PCR, 2 and a mutation that eliminates Foxa2 binding and activity in thyroid C cell lines did not affect activity in cultured trigeminal neurons (47). Interestingly, a Foxa2-immunoreactive band was detected by Western blots of rat trigeminal ganglia. 2 This suggests the possibility Foxa2 contributes to CT/CGRP expression in the thyroid and possibly liver, whereas a related Fox protein may possibly contribute to expression in the trigeminal ganglion.
Are USF and Foxa2 all that is required for the 18-bp CT/CGRP enhancer? The answer is most likely no. We expect that there will be an additional cofactor(s) because Foxa2 expression does not fully match the CT/CGRP expression pattern. For example, the relatively high expression of CT/CGRP in the thyroid compared with the liver suggests the existence of neuronal/neuroendocrine factors. In this regard, we know that there are additional weak but unidentified complexes on the EMSAs. Furthermore, the repression of enhancer activity upon Foxa2 overexpression in the CA77 cells also suggests the possible involvement of a co-activator because one explanation of the repression is that overexpressed Foxa2 squelched activity by forming nonproductive complexes with a putative co-activator.
Identification of the combination of USF and Foxa2 proteins as players at the 18-bp enhancer provides molecular insight into the regulation of the 18-bp enhancer by depolarization, mitogen-activated protein kinases, and anti-migraine drugs in a thyroid C cell line and trigeminal ganglia neurons (20,47,48). Although there are no reports on Foxa2 phosphorylation, USF can be directly activated by p38 mitogen-activated protein kinases and neuronal depolarization in other systems (49,50) and hence is likely to be the target of mitogen-activated protein kinase control of the 18-bp enhancer. It is tempting to speculate that mitogen-activated protein kinase activation may underlie the elevation of CT/CGRP levels in neurovascular headaches and sepsis. The importance of CGRP in migraine is highlighted by the reduction of CGRP levels by the widely used triptan anti-migraine drugs (51) and the efficacy of a CGRP receptor antagonist in recent clinical trials (14). Hence, the factors that bind the 18-bp enhancer may be significant in both normal and pathological regulation of the CT/CGRP gene.