Mitogen-induced expression of the fibroblast growth factor-binding protein is transcriptionally repressed through a non-canonical E-box element.

The fibroblast growth factor-binding protein (FGF-BP) stimulates FGF-2-mediated angiogenesis and is thought to play an important role in the progression of squamous cell, colon, and breast carcinomas. 12-O-Tetradecanoylphorbol-13-acetate (TPA) induction of the FGF-BP gene occurs through transcriptional mechanisms involving Sp1, AP-1, and CCAATT/enhancer-binding protein sites in the proximal FGF-BP gene promoter. The level of TPA induction, however, is limited due to the presence of a repressor element that shows similarity to a non-canonical E-box (AACGTG). Mutation or deletion of the repressor element led to enhanced induction by TPA or epidermal growth factor in cervical squamous cell and breast carcinoma cell lines. Repression was dependent on the adjacent AP-1 site, without discernible alteration in the binding affinity or composition of AP-1. We investigated the following two possible mechanisms for E-box-mediated repression: 1) CpG methylation of the core of the E-box element, and 2) binding of a distinct protein complex to this site. Point mutation of the CpG methylation site in the E-box showed loss of repressor activity. Conversely, in vitro methylation of this site significantly reduced TPA induction. In vitro gel shift analysis revealed distinct and TPA-dependent binding of USF1 and USF2 to the repressor element that required nucleotides within the E-box. Furthermore, chromatin immunoprecipitation assay showed that USF, c-Myc, and Max proteins were associated with the FGF-BP promoter in vivo. Overall, these findings suggested that the balance between trans-activation by AP-1 and repression through the E-box is an important control mechanism for fine-tuning the angiogenic response to growth factor-activated pathways.

blast growth factors (FGF-1 and FGF-2) 1 that are present at high levels in most tissues where they are bound to heparan sulfate proteoglycans and sequestered in the extracellular matrix (1,2). Tumor cells can release FGF-2 activity through the induced expression of an FGF-binding protein (FGF-BP), which is secreted from tumor cells and binds and mobilizes stored FGF-2, leading to the activation of FGF-2-dependent processes such as angiogenesis (3,4).
FGF-BP is found in only a limited number of epithelial tissues where its expression is tightly regulated. During mouse embryonic development, FGF-BP expression is up-regulated in the epithelial layers of the skin, intestine, and lung where it coincides with development of these tissues (5). After peak FGF-BP expression at embryonic day 18, levels drop significantly after birth and remain low in most tissues of the adult mouse (5). In human tissues, FGF-BP expression is low but was found to be significantly up-regulated in certain tumors including squamous cell carcinomas (SCC) derived from skin, cervix, lung, or head and neck region (4). FGF-BP is also highly expressed in some colon cancers (6) and breast adenocarcinomas. 2 A functional role for FGF-BP in these tumors has been shown through the use of ribozyme targeting, where as little as 20% reduction in FGF-BP steady-state mRNA levels led to a decrease in tumor growth and angiogenesis of xenografted cervical SCC and colon tumors (6). Thus it appears that for at least some tumors, FGF-BP expression is rate-limiting for tumor growth and angiogenesis.
A relationship between FGF-BP expression and tumor formation has also been established by the observation that levels of FGF-BP increase during 7,12-dimethylbenz[a]anthracene-and 12-O-tetradecanoylphorbol-13-acetate (TPA)induced mouse skin carcinogenesis (5). We subsequently found that FGF-BP gene transcription is directly induced by TPA or epidermal growth factor (EGF) treatment of SCC cell lines (7,8). Analysis of the FGF-BP promoter showed that TPA and EGF induction is mediated within the first 118 base pairs of the proximal promoter and requires several positive regulatory cis-elements in the FGF-BP promoter including Sp1, AP-1, and C/EBP (7,8). In addition, we identified a region of the promoter between the AP-1 and C/EBP sites that mediated a repressive effect on FGF-BP transcription. Deletion or mutation of the repressor region had no effect on basal activity of the promoter but significantly enhanced the level of TPA induction (7), indicating that this region of the promoter functions to limit the overall response to TPA induction of FGF-BP gene expression. In this study, we investigate more closely the mechanisms by which the FGF-BP repressor region can limit transcriptional induction of this gene in response to either TPA or EGF.
Plasmids-Human FGF-BP promoter fragments were cloned into the pXP1 promoterless luciferase reporter vector and have been described elsewhere (7). The FGF-BP promoter constructs from Ϫ118 to ϩ62 carrying a mutated AP-1 site (mAP-1 and mAP-1/m-58) or mutated E-box (m-55/Ϫ56) were generated by PCR-based site-directed mutagenesis as described previously (7). Briefly, point mutations were introduced into complementary overlapping PCR primers that convert the AP-1/E-box site from GTGAGTAACGTG (Ϫ66 to Ϫ55) to TGGAG-CAACGTG or TGGAGCAATGTG to generate the mAP-1 Luc and mAP-1/m-58 Luc constructs, respectively. For the m-55/Ϫ56 Luc construct, PCR primers were generated to introduce point mutations at positions Ϫ55 and Ϫ56 of the E-box, converting the site to GTGAGTAACGGT (Ϫ66 to Ϫ55). Primary and secondary PCRs were carried out exactly as described previously (7) and cloned into the pXP1 BamHI site. pRL-CMV Renilla luciferase vector (Promega) was used as transfection efficiency control.
Transient Transfections and Reporter Gene Assays-24 h before transfection, cells were plated in 6-well plates at a density of either 750,000 cells/well (ME-180) or 250,000 cells/well (all other cell lines). For each transfection, 1.0 g of FGF-BP promoter-luciferase construct, 0.2 ng of pRL-CMV (transfection efficiency control), and 8 l of Lipo-fectAMINE (Life Technologies, Inc.) were combined and added to cells for 3 h in serum-free IMEM. The transfected cells were treated for 18 h in serum-free IMEM containing vehicle alone (Me 2 SO, 0.1%), TPA (100 nM), or EGF (5 ng/ml). Cells were lysed and assayed for luciferase activity as described previously (7). Due to a slight background TPA or EGF induction of pRL-CMV, results shown have been normalized for protein content and not for Renilla luciferase activity. Protein content of cell extracts was determined by the Bradford assay (Bio-Rad). Results from transient transfection of all FGF-BP promoter/luciferase constructs were consistent using multiple plasmid preparations. All results were analyzed for statistical significance using t test analysis.
Gel Shift Assays-ME-180 cells were grown to 80% confluency on 150-mm dishes, serum-starved for 16 h, and treated with vehicle alone (Me 2 SO, 0.1%) or with 100 nM TPA for 1 h. Nuclear extract preparation, probe labeling, and binding reactions were carried out exactly as described previously (7), using 5 g of ME-180 nuclear extracts and 200 ng of poly(dI-dC). Reactions were incubated 10 min on ice. 50-Fold molar excess (unless indicated otherwise) of unlabeled competitor oligonucleotides or 2.0-g supershift antibodies were added and incubated for another 10 min before adding 20 fmol of labeled probe. 0.2 g of USF-1 or USF-2 blocking peptides were added concurrently with the antibodies. Reactions were carried out 40 min on ice and analyzed by 6% polyacrylamide gel electrophoresis. All promoter fragments for gel shift were generated by annealing synthetic oligonucleotides. Sequence of the consensus AP-1 element was 5Ј-CTAGTGATGAGTCAGCCGGATC-3Ј. Supershift antibodies were purchased from Santa Cruz Biotechnology and included antibodies for c-Myc In Vitro Methylation-20 g of FGF-BP promoter constructs Ϫ118/ ϩ62 and m-58 were CpG-methylated in vitro with 10 units of SssI methylase (CpG methylase) (New England Biolabs), 160 M S-adenosylmethionine (New England Biolabs), and magnesium-free buffer containing 50 mM NaCl, 10 mM Tris-HCl, and 10 mM EDTA. Methylation reaction was carried out for 5 h at 37°C. Plasmids were purified by phenol/chloroform and ethanol precipitation prior to transfection. Complete CpG methylation was confirmed by digestion with the methylation-sensitive restriction enzyme HpaII (New England Biolabs).
Formaldehyde Cross-linking and Chromatin Immunoprecipitation (ChIP Assay)-Approximately 10 7 ME-180 cells were serum-starved for 16 h followed by treatment for 1 h with 10 Ϫ7 M TPA. Proteins were cross-linked to DNA by adding formaldehyde directly to culture medium to a final concentration of 1% for 15 min at room temperature. Cells were subsequently washed and scraped into 1 ml of 1ϫ phosphatebuffered saline containing 1ϫ protease inhibitor mixture (Roche Molecular Biochemicals). Cell pellets were lysed in 200 l of lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.0, 1ϫ protease inhibitor mixture) for 10 min on ice. Lysates were sonicated on ice to an average DNA length of 100 -500 base pairs and centrifuged to remove cell debris. Supernatant was diluted 5-fold in immunoprecipitation buffer (0.1% SDS, 1% Triton X-100, 0.1% sodium deoxycholate, 140 mM NaCl, 1ϫ protease inhibitors) and pre-cleared with 50 l of GammaBind TM Plus Sepharose TM (Amersham Pharmacia Biotech), 20 g of salmon sperm DNA, and 50 g of bovine serum albumin for 30 min at 4°C. Beads were pelleted, and 10 g of antibody (see previous section) was added to supernatant and incubated overnight at 4°C. Immune complexes were collected with GammaBind TM Plus Sepharose TM and washed (9). DNA was eluted with 1% SDS, 0.1 M NaHCO 3 for 15 min at room temperature. Cross-links were reversed by incubating eluates at 65°C for 4 h in 0.2 M NaCl, followed by digestion with 40 ng/l proteinase K in 10 mM EDTA, 40 mM Tris-HCl, pH 6.5, for 2 h at 45°C. DNA was recovered by phenol/chloroform/isoamyl alcohol extraction and ethanol precipitation.
PCRs contained 4% of input DNA or 20% of immunoprecipitated DNA along with 10 pmol of primers, 1.5 mM MgCl 2 , 0.2 mM dNTPs, 1ϫ PCR buffer (Life Technologies, Inc.), and 5 units of Taq DNA polymerase (Life Technologies, Inc.). Primers used were from Ϫ369 to Ϫ350 and from Ϫ47 to Ϫ73 of the FGF-BP promoter (7). Plasmid containing Ϫ1829 to ϩ62 of the FGF-BP promoter was used as a control template. After 22 cycles of PCR, samples were run on a 1% agarose gel, transferred to nylon membrane, and probed with an FGF-BP-specific primer from Ϫ118 to Ϫ99 end-labeled with T4 kinase. Band intensities were quantitated by PhosphorImager, and the amount immunoprecipitated was expressed as percent of total input.

RESULTS
Identification of a Repressor Element in the FGF-BP Promoter-We previously discovered that deletion of the region between Ϫ57 and Ϫ47, situated between the AP-1 and C/EBP sites, resulted in enhanced TPA induction of transcription (7), indicating the involvement of a negative regulatory element in the TPA regulation of FGF-BP. This region has no homology to any known transcription factor binding sites except for the presence of an imperfect E-box between Ϫ60 and Ϫ55. To characterize further the FGF-BP repressor element, we tested whether mutations in this region would disrupt repressor activity and lead to enhanced TPA or EGF induction of the promoter. The region of the FGF-BP promoter between Ϫ118 and ϩ62 harbors all of the necessary elements for full induction by both TPA (7) and EGF (8). Internal deletion from Ϫ57 to Ϫ47 generated in the context of the Ϫ118 to ϩ62 promoter showed significantly increased induction by both TPA and EGF when transfected into ME-180 cervical squamous carcinoma cells (Fig. 1A). In addition, introduction of a C to T point mutation at position Ϫ58 within the E-box showed a dramatically increased response to TPA, going from a 7-to 18-fold induction (Fig. 1A). EGF induction of the Ϫ58 mutant was also highly increased, going from a 5.5-to 10-fold induction (Fig. 1A). Loss of repression was not reflected at the level of basal promoter activity since the repressor mutant constructs Ϫ57/Ϫ47 and m-58 had the same uninduced promoter activity as the wild-type Ϫ118/ϩ62 (7,8). Therefore, the region between Ϫ58 and Ϫ47 of the FGF-BP promoter appears to function in limiting the overall transcriptional response to both TPA and EGF.
In order to determine whether repressor activity on the FGF-BP promoter existed in other cell types, we tested the TPA and EGF response of the wild-type Ϫ118/ϩ62 or the E-box point mutant (m-58) promoter constructs in other cell lines. The Ϫ58 mutant showed significantly enhanced TPA induction in HeLa (cervical squamous carcinoma) and MCF-7 (breast cancer) cell lines (Fig. 1B). In cell lines where the FGF-BP promoter was EGF-responsive, such as HeLa and BT-549 (breast), the fold EGF induction of the Ϫ58 mutant was also consistently higher (Fig. 1C). These data indicate that repression of the FGF-BP promoter is a general mechanism of FGF-BP transcriptional regulation in response to TPA or EGF stimulation.
AP-1 Dependence of Repressor Activity-The proximity of the AP-1 site to the repressor mutations raised the possibility that the observed increase in the TPA response was due to its impact on the juxtaposed AP-1 site. To test the possible influence that the Ϫ58 mutation may have on the AP-1 site, we generated double mutant constructs carrying the Ϫ58 mutation in conjunction with mutations in either the AP-1 site or the C/EBP site. We have shown previously that EGF or TPA induction of FGF-BP partly depends on the AP-1 and C/EBP sites in the promoter, since mutation of the AP-1 site (mAP-1), or deletion of C/EBP (delta C/EBP) results in a significant de-crease in the amount of induction by EGF and TPA with no effect on basal promoter activity ( Fig. 2 and Refs. 7 and 8). Mutations in the AP-1 site have also been shown to disrupt AP-1 binding (7) (Fig. 5B). As shown in Fig. 2, the Ϫ58 mutation alone showed an enhanced TPA response, whereas mutation of the Ϫ58 and AP-1 sites together showed no loss of repression. Instead, the m-58/mAP-1 double mutant displayed a similar level of TPA or EGF induction as that of the mAP-1 single mutant construct. Conversely, the Ϫ58 mutation in combination with a deleted C/EBP site resulted in loss of repression and increased TPA and EGF induction (Fig. 2). Therefore, the effect of the Ϫ58 mutation is not dependent on the C/EBP site but is dependent on an intact AP-1 site since the mAP-1/m-58 construct does not result in a loss of repression.
The AP-1 dependence of repressor activity prompted us to confirm that the Ϫ58 mutant phenotype was not simply a consequence of altered AP-1 binding affinity to the FGF-BP AP-1 site. We carried out gel shift competition analysis of the AP-1 site in the presence or absence of the Ϫ58 mutation. By using a promoter fragment that spans the AP-1/repressor element (Ϫ70 to Ϫ51) as a probe, we could detect AP-1 binding (upper complex) and binding of additional proteins that are specific to the repressor element (Fig. 3A, upper panel and Refs. 7 and 8). When the Ϫ58 point mutation was introduced into the repressor site, the lower complexes disappeared, and only the AP-1 complex was bound (Fig. 3A, lower panel). Competition with an unlabeled consensus AP-1 element (Fig. 3A, lanes 2-4) or with the FGF-BP AP-1 element (lanes 5-7) could effectively compete for AP-1 binding. Both competitors reduced AP-1 binding to the probe at concentrations between 10-and 20-fold molar excess, as determined by quantitation of the AP-1 band intensity. Similarly, competition for AP-1 bound in the presence of the Ϫ58 mutation also occurred at concentrations between 10-and 20-fold molar excess of unlabeled AP-1 elements (Fig. 3A, lanes 8 -14). Although quantitation of AP-1 binding to the m-58 probe suggests a decreased AP-1 binding affinity, this may be due to the difficulty in obtaining accurate quantitation of band intensity in the presence of multiple bands (Fig. 3A, upper panel). Nevertheless, these results indicate that the enhanced TPA induction caused by the Ϫ58 mutation cannot be explained by increased AP-1 binding affinity.
Alternatively, the Ϫ58 mutation could potentially alter the composition of AP-1 by flanking the AP-1 site with nucleotides that favor binding of different AP-1 family members, leading to increased transcriptional activation. We have published elsewhere that EGF activation of ME-180 cells led to increased binding of c-Fos and JunD proteins to the FGF-BP AP-1 site (8). We asked which AP-1 family members were bound to the FGF-BP promoter after TPA treatment and whether the composition of AP-1 changed in the presence of the Ϫ58 mutation. Gel supershift analysis was carried out using the Ϫ70/Ϫ51 FGF-BP promoter fragment as a probe in the presence of TPAtreated ME-180 extracts and antibodies specific for individual members of the AP-1 family (Fig. 3B, lanes 1-10). Supershift of the AP-1 complex occurred in the presence of cross-reactive Fos and Jun antibodies (lanes 2 and 7, respectively) as well as with specific antibodies for c-Fos (lane 3), Fra2 (lane 6), and JunD (lane 10). The binding of c-Fos, Fra2, and JunD was also prevalent in the presence of the Ϫ58 mutation (Fig. 3B, lanes 13, 16  and 20). Overall, these experiments demonstrate that although repressor activity mediated through the Ϫ58 site is dependent on the adjacent AP-1 site, the Ϫ58 mutation had no obvious impact on either AP-1 binding affinity or the composition of the AP-1 complex.
Methylation of the FGF-BP Promoter Represses TPA Induction-We hypothesized that one possible effect of the Ϫ58 mu-tation could be that the C to T point mutation leads to loss of transcriptional repression through disruption of a CpG methylation site. The Ϫ57/Ϫ47 deletion is consistent with this hypothesis since this mutation also destroys the Ϫ58 CpG methylation site. In vivo cytosine methylation occurs preferably at CpG dinucleotides and is closely associated with transcriptional repression of genes with CpG sites in their promoter region (10). In order to determine whether methylation of the Ϫ58 CpG site could mediate repression of FGF-BP promoter induction, we tested the effect of in vitro methylation of the FGF-BP promoter constructs on their transcriptional response to TPA. The wild-type Ϫ118/ϩ62 and the m-58 promoter constructs were methylated in vitro with SssI methylase so that each plasmid would differ in its CpG methylation pattern only at the Ϫ58 site (Fig. 4). Complete methylation of each plasmid construct was confirmed by digestion with HpaII (data not shown). Transfection of the methylated Ϫ118/ϩ62 plasmid into ME-180 cells resulted in a significant 50% decrease in the level of TPA induction compared with the unmethylated Ϫ118/ϩ62 plasmid (Fig. 4). On the other hand, methylation of the m-58 promoter construct, which is unmethylated at position Ϫ58, demonstrated a similar level of TPA induction compared with the unmethylated m-58 construct (Fig. 4). Both methylated constructs displayed an equivalent decrease in basal promoter activity which was 20% lower than the unmethylated plasmids (data not shown) and was independent of the Ϫ58 mutation. These results demonstrate that methylation of the Ϫ58 CpG site significantly represses TPA induction and indicate that methylation of this site is a potential mechanism for limiting the transcriptional response to growth factor or TPA regulation in vivo.
Transcription Factor Binding to the FGF-BP Repressor Element-Although methylation causes repression through the Ϫ58 site, there was still the question of how methylation affects protein binding to the promoter. Furthermore, we also wondered how the C to T point mutation at Ϫ58 might affect protein binding to the unmethylated FGF-BP promoter. We examined protein binding to the FGF-BP repressor element by electrophoretic gel mobility shift assays using a fragment of the promoter from Ϫ70 to Ϫ51 spanning the AP-1/repressor element. The various promoter fragments used are depicted in Fig. 5A, with their sequences shown in Fig. 6A. In the presence of ME-180 nuclear extracts, gel shift with the Ϫ70/Ϫ51 probe resulted in four distinct protein complexes (Fig. 5B, lane 1 see Ref. 7). Protein binding of all four complexes increased in the presence of nuclear extracts from TPA-treated ME-180 cells (Fig. 5B, lane 2). The largest increase was in the uppermost band (complex 1) which represents AP-1 as shown by supershift analysis (Fig. 3B (8)) and by lack of competition with unlabeled Ϫ70/Ϫ51 fragment harboring point mutations in the AP-1 site (Fig. 5B, lane 3). To determine whether the Ϫ58 repressor mutation could affect protein binding to the promoter, competition was carried out with the FGF-BP promoter fragment harboring the C to T point mutation at Ϫ58. The m-58 fragment could effectively compete for binding to the AP-1 band (complex 1) since this fragment retains an intact AP-1 element (Fig. 5B, lane 4). In contrast, mutation of position Ϫ58 resulted in loss of binding to complexes 2-4 (lane 4) suggesting that these bands represent factors that interact with the FGF-BP repressor element.
Due to the observation that methylation of the Ϫ58 site caused repression of TPA induction, we wondered whether methylation itself alters protein binding to the FGF-BP promoter. As seen in Fig. 5C, an unlabeled promoter fragment that was methylated at the Ϫ58 CpG site on both strands (meth-58) was able to compete for AP-1 binding but was a less efficient competitor for the other complexes (lane 2). Similarly, when the methylated fragment was used as a probe, only AP-1 binding was detected (lane 3). This result demonstrates that methylation of the Ϫ58 site does not interfere with AP-1 binding to the promoter in vitro, and thus decreased AP-1 binding does not account for methylation-mediated repression. In contrast, methylation does interfere with binding of complexes 2-4, suggesting that the bindings of these proteins are methylationsensitive. Furthermore, under in vitro gel shift conditions, methylation does not induce binding of additional protein complexes that also might account for transcriptional repression. We next examined the binding of distinct factors to the unmethylated repressor element in order to investigate whether any of these factors might play a role in repression through the Ϫ58 site. In Fig. 5D, the mAP-1 promoter fragment was used as a probe in order to eliminate binding of the AP-1 band (complex 1) and to better resolve the lower bands (complexes [2][3][4]. Surprisingly, only one distinct complex was found to bind to the mAP-1 fragment that co-migrated with complex 2 in the gel shift with the Ϫ70/Ϫ51 promoter fragment (Fig. 5D). The amount of complex 2 binding to the repressor element was significantly increased in the presence of extracts from TPA-treated cells (lane 3). Competition analysis demonstrated that complex 2 was competed by the wild-type Ϫ70/Ϫ51 (lane 4) or mAP-1 (lane 5) promoter fragments such that the resulting band intensities were 22 and 26% of uncompeted binding, respectively. In contrast, complex 2 binding was not efficiently competed by the m-58 fragment (lane 6), which was 63% of uncompeted binding. These results show that binding of complex 2 is independent of AP-1 and is dependent on the cytosine at position Ϫ58, demonstrating an interaction between a distinct complex with the repressor element of the FGF-BP promoter.
Mapping of Repressor Binding to the E-box-In order to determine precisely which nucleotides were necessary for complex 2 binding to the FGF-BP repressor element, we carried out a complete mutational analysis of the Ϫ70/Ϫ51 promoter fragment. Single point mutations were introduced into the repressor element from Ϫ60 to Ϫ51 (Fig. 6A) and tested for their ability to compete for binding to complex 2. By using the mAP-1 promoter fragment as a probe, gel shift competition showed that point mutations introduced between Ϫ60 and Ϫ55, includ- ing the Ϫ58 mutation shown previously, could no longer compete for binding to complex 2 (Fig. 6B, lanes 3-8), whereas point mutations at the 3Ј end of the fragment from Ϫ54 to Ϫ51 could effectively compete for complex 2 binding (lanes 9 -12). It is noteworthy that only mutations within the non-canonical E-box AACGTG had an effect on complex 2 binding.
We next tested whether disruption of complex 2 binding to the FGF-BP E-box correlated with the loss of repressor activity on the FGF-BP promoter. Point mutations were introduced into the promoter at positions Ϫ55 and Ϫ56 in the 3Ј end of the FGF-BP E-box which convert the E-box from AACGTG to AACGGT (Fig. 6A). Like the Ϫ58 mutation, the Ϫ55/Ϫ56 E-box mutation was unable to compete for complex 2 binding in a gel shift assay (Fig. 7A). The functional impact of the Ϫ55/Ϫ56 E-box mutation was then tested in the context of the Ϫ118 to ϩ62 promoter construct by transient transfection. Like the m-58 E-box mutant, the Ϫ55/Ϫ56 promoter mutant displayed significant loss of repression compared with the wild-type promoter in response to TPA (Fig. 7B) and to EGF (Fig. 7C). These results are consistent with the correlation between protein binding to the E-box and transcriptional repression. It should be noted that the m-55/Ϫ56 mutation does not disrupt the CpG core of the E-box, indicating that loss of repression through E-box mutation can also occur independently of methylation. Thus, two separate mutations in the FGF-BP E-box disrupt complex 2 binding and display enhanced TPA induction of the FGF-BP promoter.
Identification of USF and c-Myc Binding to FGF-BP Promoter-E-box promoter elements are recognized by a number of basic helix-loop-helix leucine zipper (bHLHZip) transcription factors (11)(12)(13). To determine whether any known members of this transcription factor family were present in complex 2, supershift analysis was carried out using antibodies specific for certain bHLHZip factors. The presence of antibodies for c-Myc or Max had no effect on complex 2 binding (Fig. 8, lanes 3 and  4). Similarly, antibodies recognizing the Mad proteins Mad-1 (lane 5) or Mad-2 (Mxi1), Mad-3, and Mad-4 (data not shown) had no effect on complex 2 binding. Although the bHLH transcription factor aryl hydrocarbon receptor nuclear translocator (Arnt) has been shown to bind as homodimers in vitro to an AACGTG variant E-box (14), antibodies specific for Arnt had no effect on complex 2 binding to the FGF-BP promoter (Fig. 8,  lane 6). On the other hand, antibodies specific for the bHLHZip factors USF-1 and USF-2 completely blocked complex 2 binding to the FGF-BP promoter (Fig. 8, lanes 7 and 9). The ability of the USF-1 or USF-2 antibodies to block complex 2 binding was reversed in the presence of competing USF-1 or USF-2 peptides, respectively (lanes 8 and 10), demonstrating antibody specificity. The TPA-induced binding of USF-1 and USF-2 to a non-canonical E-box AACGTG in the FGF-BP promoter indicated that USF may play a distinct role in the transcriptional regulation of this gene.
Whereas USF is ubiquitously expressed and is easily detected by gel shift assay, the binding of Myc/Max/Mad proteins from cellular extracts is practically undetectable under these conditions (15). Therefore, the in vitro binding assays are biased toward detecting USF. To get a more realistic picture of E-box factor binding to the FGF-BP promoter, we decided to analyze the in vivo binding of USF or Myc/Max/Mad proteins by formaldehyde cross-linking followed by chromatin immunoprecipitation and PCR amplification of the FGF-BP promoter. This type of analysis (also known as the ChIP assay) has proved successful in determining the binding characteristics of c-Myc and USF to the endogenous cad promoter (16 -18). We therefore used a similar approach to determine whether the binding of c-Myc family members could be detected on the endogenous FGF-BP promoter in ME-180 cells. After immunoprecipitation, a fragment of the promoter from Ϫ369 to Ϫ47 was amplified by PCR and detected by Southern analysis. As seen in Fig. 9, antibodies against c-Myc, Max, or USF-1 could effectively immunoprecipitate the FGF-BP promoter. After correction for the amount of input DNA, we determined the amount of FGF-BP promoter immunoprecipitated was 3.0 -3.8fold higher than the no antibody control. This is in contrast to immunoprecipitation with the Mad-2 ( Fig. 9) or Mad-1 (data not shown) antibodies that were similar to the control. Immunoprecipitation of the FGF-BP promoter with c-Myc and USF-1 was reproducible over multiple experiments and was also detected after amplification of a shorter FGF-BP promoter fragment (Ϫ118 to ϩ62) to the same degree over the control (data not shown). These results suggest that both USF as well as c-Myc-Max complexes can associate with the endogenous FGF-BP promoter in ME-180 cells and could potentially play a functional role in modulating TPA or EGF induction of FGF-BP gene expression.

DISCUSSION
In this study, we show that repression of FGF-BP transcription occurs through a promoter element situated between the AP-1 and C/EBP sites and containing homology to an E-box element. Mutation or deletion of this site substantially increased the level of both TPA and EGF induction of FGF-BP promoter activity in multiple cell lines, implying that this region normally limits the extent of the response to growth factor stimulation. In addition, FGF-BP promoter activity was repressed even further upon CpG methylation of the E-box. Overall our data suggest that the transcriptional response to TPA can be limited by an E-box site that operates via mechanisms involving methylation and/or interaction with USF and c-Myc transcription factors.
Promoter mutations introduced between the AP-1 and C/EBP sites that exhibited enhanced TPA or EGF induction were consistent with the disruption of the E-box motif from Ϫ60 to Ϫ55. E-box elements are generally recognized by the family of bHLHZip transcription factors that bind to the consensus E-box element CACGTG. However, certain bHLHZip factors also recognize non-canonical E-box elements similar to the AACGTG E-box found in the FGF-BP promoter. For example, studies using in vitro binding site selection (11) or in vivo chromatin immunoprecipitation (12) found that c-Myc/Max heterodimers are able to recognize the sequence CAACGTG. USF was also found to bind the CAACGTG heptamer (11) as well as a AACGTG hexamer (13), albeit with lowered affinity compared with the consensus CACGTG E-box under in vitro binding conditions (13). Additionally, the bHLH transcription factor Arnt has been shown to bind as homodimers in vitro to an AACGTG variant E-box (14). However, it is unclear whether Arnt homodimers occur in vivo, and we were unable to detect Arnt binding to the FGF-BP E-box in an in vitro gel shift assay.
In our experiments we found that the non-canonical FGF-BP E-box is recognized by USF-1 and USF-2 in vitro, suggesting that these factors may play a role in the regulation of this gene. The lack of binding by c-Myc family members in vitro is not surprising since c-Myc binding activity is undetectable in most cells (15). However, using in vivo chromatin immunoprecipitation we were able to detect significant binding of USF, c-Myc, and Max to the region of the promoter (between Ϫ118 and ϩ62) containing the AACGTG E-box. Although this region contains no additional E-box or Inr elements, we cannot rule out the possibility that c-Myc-Max complexes bind to an as yet undetermined sequence within this region or that they interact with the promoter through other proteins, such as C/EBP (19). In an attempt to determine whether FGF-BP is indeed a target gene for USF and/or c-Myc we carried out preliminary experiments testing the effect of transient overexpression of USF or c-Myc on FGF-BP promoter activity. Due to indirect effects on cell growth, however, we were unable to detect selective c-Myc-or USF-dependent regulation of FGF-BP, which was consistent with other studies examining c-Myc target genes (15).
The exact mechanism of E-box-mediated repression is complex. Although Mad-Max complexes have been shown to actively repress transcriptional through E-box binding and recruitment of mSin3 (20), this mechanism is unlikely to be responsible for FGF-BP repression since we were unable to detect Mad-1 or Mad-2 binding to the promoter in vivo. Repression of FGF-BP was dependent on the juxtaposed AP-1 site since the phenotype of the m-58 repressor mutation (i.e. increased TPA induction) was lost in the presence of an AP-1 site mutation. This finding implies a mechanism by which action through the E-box element (such as transcription factor binding or methylation) impinges on the trans-activation ability of AP-1. This interaction occurs without affecting the binding affinity of AP-1 or the composition of the AP-1 complex. USF and AP-1 family members have been shown to functionally interact through their leucine zipper domains, resulting in either AP-1 stimulation, as in the case of USF-2-c-Fos complexes (21), or AP-1 inhibition, as in the case of USF-1-Fra1 complexes (22). Whether or not these complexes require DNA binding, however, was unclear. The close proximity between AP-1 and E-box elements on the FGF-BP promoter suggests that E-box factor binding could functionally inhibit AP-1 transactivation through a specific protein-protein interaction with AP-1. Similarly, the ␣A-crystallin promoter is repressed in a tissue-specific manner through a composite USF/AP-1 site involving JunD and Fra2 family members (23). Thus, transcriptional repression via interactions between USF and AP-1 may be a common mechanism by which AP-1 and/or USF activity is fine-tuned in response to growth factor stimulation or tissuespecific expression.
The presence of both USF and c-Myc binding to the FGF-BP promoter suggest that the relative binding levels of each of these factors may play a role in FGF-BP regulation. Such a model has been proposed for the cad promoter, where both c-Myc and USF are bound to the promoter E-box in vivo (16,17). Induction of cad gene expression, however, is dependent only on the presence of c-Myc and not USF (16,17). Therefore, discrimination at the FGF-BP E-box may depend on the relative protein levels through a mechanism involving interplay between c-Myc, USF, and possibly other bHLHZip family members. Alternatively, binding discrimination could occur posttranslationally. In this regard, we found that USF binding to the FGF-BP E-box increases significantly after TPA treatment, suggesting that its DNA binding (and perhaps repressor activity) is regulated through a phosphorylation event.
Another aspect of this study is the possibility that methylation of the E-box effectively represses TPA induction of FGF-BP. Aberrant methylation is often associated with tumor progression. Hypermethylation of CpG islands is correlated with the silencing of tumor suppressor genes such as p16 (24). On the other hand, global hypomethylation has been associated with formation of colon cancer (25), and the hypomethylation of individual CpG sites can lead to expression of oncogenes such as ras (26). Methylation is generally associated with transcriptional repression that can occur via indirect mechanisms through the binding of factors, such as MeCP1 and MeCP2, which recognize methylated CpG dinucleotides and recruit histone deacetylases, thus leading to a more compact chromatin structure and transcriptional repression (10). Methylation can also lead to site-specific repression by preventing binding of transcription factors containing CpG dinucleotides in their recognition sites. For example, an inverse relationship between methylation and bHLHZip factor binding to E-box elements has been demonstrated for c-Myc (27) as well as USF (28). Our experiments also show that binding of USF to the FGF-BP E-box in an in vitro assay is sensitive to methylation, whereas methylation had no discernible effect on AP-1 binding. This suggests that the mechanism by which methylation inhibits FGF-BP transcription is distinct from alterations in protein (i.e. USF) binding. This is further supported by the finding that the m-55/Ϫ56 mutant disrupts repression without affecting the methylation site, suggesting multiple mechanisms by which the E-box mediates transcription. Our finding suggests a possible correlation between hypomethylation and an increase in FGF-BP gene expression, an event that could significantly contribute to tumor growth and angiogenesis in certain cancer types.
Overall, this study further defines the transcriptional mechanisms by which the angiogenic modulator FGF-BP is regulated in response to stimulation by EGF or TPA. An important aspect of FGF-BP regulation is the presence of an E-box that mediates AP-1-dependent transcriptional repression. Differences in USF and/or c-Myc binding in conjunction with changes in the methylation status of the promoter may be important mechanisms by which the extent of repression exerted on the FGF-BP promoter is de-regulated, thus leading to increased FIG. 9. Binding of c-Myc/Max and USF to the FGF-BP promoter in vivo by ChIP assay. Formaldehyde cross-linked chromatin from TPA-treated ME-180 cells was immunoprecipitated with antibodies to c-Myc, Max, Mad-2, USF1, or in the absence of antibody (no Ab). Input and immunoprecipitated (I.P.) DNA were analyzed by 22 cycles of PCR using primers specific for the FGF-BP promoter, followed by Southern blot with an internal primer. The amount of immunoprecipitated promoter is determined as percent of total input DNA and is expressed to the right of the panel as fold over the no antibody control. The data are representative of three independent experiments. FGF-BP gene expression and activation of an angiogenic phenotype.