Phorbol Ester-induced Transcription of a Fibroblast Growth Factor-binding Protein Is Modulated by a Complex Interplay of Positive and Negative Regulatory Promoter Elements*

Earlier studies from our laboratory showed that a secreted binding protein for fibroblast growth factors (FGF-BP) is expressed at high levels in squamous cell carcinoma (SCC) cell lines. Overexpression studies or conversely reduced expression of FGF-BP by ribozyme targeting have elucidated a direct role of this protein in angiogenesis during tumor development. We have also observed a significant up-regulation of FGF-BP during TPA (12-O-tetradecanoylphorbol-13-acetate) promotion of skin cancer. Here we investigate the mechanism of TPA induction of FGF-BP gene expression in the human ME-180 SCC cell line. We found that TPA increased FGF-BP mRNA levels in a time- and dose-dependent manner mediated via the protein kinase C signal transduction pathway. Results from actinomycin D and cycloheximide experiments as well as nuclear transcription assays revealed that TPA up-regulated the steady-state levels of FGF-BP mRNA by increasing its rate of gene transcription independently ofde novo protein synthesis. We isolated the human FGF-BP promoter and determined by deletion analysis that TPA regulatory elements were all contained in the first 118 base pairs upstream of the transcription start site. Further mutational analysis revealed that full TPA induction required interplay between several regulatory elements with homology to Ets, AP-1, and CAATT/enhancer binding protein C/EBP sites. In addition, deletion or mutation of a 10-base pair region juxtaposed to the AP-1 site dramatically increased TPA induced FGF-BP gene expression. This region represses the extent of the FGF-BP promoter response to TPA and contained sequences recognized by the family of E box helix-loop-helix transcription factors. Gel shift analysis showed specific and TPA-inducible protein binding to the Ets, AP-1, and C/EBP sites. Furthermore, distinct, specific, and TPA-inducible binding to the imperfect E box repressor element was also apparent. Overall, our data indicate that TPA effects on FGF-BP gene transcription are tightly controlled by a complex interplay of positive elements and a novel negative regulatory element.

Earlier studies from our laboratory showed that a secreted binding protein for fibroblast growth factors (FGF-BP) is expressed at high levels in squamous cell carcinoma (SCC) cell lines. Overexpression studies or conversely reduced expression of FGF-BP by ribozyme targeting have elucidated a direct role of this protein in angiogenesis during tumor development. We have also observed a significant up-regulation of FGF-BP during TPA (12-O-tetradecanoylphorbol-13-acetate) promotion of skin cancer. Here we investigate the mechanism of TPA induction of FGF-BP gene expression in the human ME-180 SCC cell line. We found that TPA increased FGF-BP mRNA levels in a time-and dose-dependent manner mediated via the protein kinase C signal transduction pathway. Results from actinomycin D and cycloheximide experiments as well as nuclear transcription assays revealed that TPA up-regulated the steadystate levels of FGF-BP mRNA by increasing its rate of gene transcription independently of de novo protein synthesis. We isolated the human FGF-BP promoter and determined by deletion analysis that TPA regulatory elements were all contained in the first 118 base pairs upstream of the transcription start site. Further mutational analysis revealed that full TPA induction required interplay between several regulatory elements with homology to Ets, AP-1, and CAATT/enhancer binding protein C/EBP sites. In addition, deletion or mutation of a 10-base pair region juxtaposed to the AP-1 site dramatically increased TPA induced FGF-BP gene expression. This region represses the extent of the FGF-BP promoter response to TPA and contained sequences recognized by the family of E box helix-loop-helix transcription factors. Gel shift analysis showed specific and TPA-inducible protein binding to the Ets, AP-1, and C/EBP sites. Furthermore, distinct, specific, and TPAinducible binding to the imperfect E box repressor element was also apparent. Overall, our data indicate that TPA effects on FGF-BP gene transcription are tightly controlled by a complex interplay of positive elements and a novel negative regulatory element.
FGF-BP 1 is a secreted protein that binds to acidic FGF and basic FGF in a non-covalent reversible manner (1). FGF-BP mRNA has been found to be up-regulated in squamous cell carcinoma (SCC) cell lines of different origin, in SCC tumor samples from the head and neck, and in some colon cancer cell lines (1,2). More recently, developmental expression of the mouse FGF-BP gene was found to be prominent in the skin and intestine during the perinatal phase and is down-regulated in adult mice (3). We previously described that expression of FGF-BP in a non-tumorigenic human cell line (SW-13) which expresses bFGF leads to a tumorigenic and angiogenic phenotype (2). Expression of FGF-BP in these cells solubilizes their endogenous bFGF from its extracellular storage and allows it to reach its receptor, suggesting that FGF-BP serves as an extracellular carrier molecule for bFGF (2,4). Expression of FGF-BP under the control of a tetracycline-responsive promoter system in SW-13 cells revealed its role during the early phase of tumor growth (5). To assess the significance of FGF-BP endogenously expressed in tumors, we depleted human SCC (ME-180) and colon carcinoma (LS174T) cell lines of their FGF-BP by targeting with specific ribozymes (6). This study showed that the reduction of FGF-BP reduced the release of biologically active bFGF from cells in culture. In addition, the growth and angiogenesis of xenografted tumors in mice was decreased in parallel with the reduction of FGF-BP, suggesting that some human tumors can utilize FGF-BP as an angiogenic switch molecule.
The fact that FGF-BP has been detected in only a few types of tumors, where it seems to play a crucial role in angiogenesis, led us to investigate the mechanisms responsible for turning its expression on or off. Studying the regulation of FGF-BP in SCC cell lines, we showed that all-trans-retinoic acid, used as a chemotherapeutic agent against SCCs, down-regulates FGF-BP gene expression in vitro by both transcriptional and post-transcriptional mechanisms (7). In vivo all-trans-retinoic acid treatment reduces FGF-BP expression in SCC xenografts and inhibits their tumor growth and angiogenesis (8). On the other hand, FGF-BP mRNA expression in the adult mouse skin was found to be dramatically increased during the early stages of 7,12-dimethylbenz[a]anthracene/TPA-induced mouse skin papilloma formation (3), as well as in 7,12-dimethylbenz[a]anthracene/TPA-treated human skin grafted onto SCID mice. 2 Similarly, FGF-BP expression in vitro was up-regulated in epidermal cell lines carrying an activated ras gene, implicating the ras/PKC pathway in the regulation of FGF-BP (3).
In this context, and given the fact that FGF-BP could play a critical role in the development of human skin cancer, we decided to investigate the effects of the tumor promoter TPA on FGF-BP gene regulation. Our results show that FGF-BP mRNA expression is up-regulated by TPA in the ME-180 SCC cell line and that this induction is mediated by direct transcriptional mechanisms. Analysis of the human FGF-BP promoter reveals that the TPA induction is mediated by cooperation of several inducible regulatory elements. Furthermore, the induction of gene expression by TPA can be modified by a repressor element juxtaposed to the AP-1 site which contains sequences recognized by E box element factors.

MATERIALS AND METHODS
Cell Culture-The ME-180 squamous cell carcinoma cell line was obtained from American Type Culture Collection (ATCC, Rockville, MD). Cells were cultured in improved minimum essential medium (Biofluids Inc., Rockville, MD) with 10% fetal bovine serum (Life Technologies, Inc.).
Northern Analysis-ME-180 cells were grown to 80% confluence on 150-mm tissue culture dishes, washed three times in serum-free IMEM, and then treated 16 h later with 12-O-tetradecanoylphorbol-13-acetate (TPA) (Sigma) in serum-free IMEM. Total RNA was isolated with the RNA STAT-60 method using commercially available reagents and protocols (RNA STAT-60 TM , Tel-Test, Friendswood, TX). 30 g of total RNA were separated by electrophoresis in 1.2% formaldehyde-agarose gel and then blotted onto nylon membranes (MSI, Westboro, MA). The blots were prehybridized in 6ϫ SSC (0.9 M sodium chloride, 0.09 M sodium citrate, pH 7.0), 0.5% (w/v) SDS, 5ϫ Denhardt's solution (0.1% (w/v) Ficoll, 0.1% (w/v) polyvinylpyrrolidone, 0.1% (w/v) bovine serum albumin, 100 g/ml sonicated salmon sperm DNA) (Life Technologies, Inc.) for 4 h at 42°C. Hybridization was carried out overnight at 42°C in the same buffer. After hybridization, blots were washed three times with 2ϫ SSC and 0.1% SDS for 10 min at 42°C and finally once with 1ϫ SSC and 0.1% SDS for 20 min at 65°C. Autoradiography was performed using intensifying screens at Ϫ70°C. Blots were stripped by boiling 2 ϫ for 10 min in 1ϫ SSC and 0.1% SDS. Hybridization probes were prepared by random-primed DNA labeling (Amersham Pharmacia Biotech) of purified insert fragments from human FGF-BP (2) and human GAPDH (CLONTECH). The final concentration of the labeled probes was always greater than 10 6 cpm/ml hybridization solution. Quantitation of mRNA levels was performed using a PhosphorImager (Molecular Dynamics).
In Vitro Transcription on Isolated Nuclei-ME-180 cells were grown to 80% confluence on 150-mm tissue culture dishes. Cells were washed three times in serum-free IMEM and then treated 16 h later with TPA in serum-free IMEM for indicated times. Nuclei from 10 7 cells for each time point were isolated after incubation in lysis buffer containing 0.5% Nonidet P-40 as described (7). Nuclear transcription assays were performed with [␣-32 P]UTP (Amersham Pharmacia Biotech) as described (7). Equal amounts of radioactivity (0.5-1 ϫ 10 7 cpm) were hybridized to nitrocellulose filters containing 3 g of each plasmid. After hybridization for 4 days at 42°C, the filters were washed 4 times with 2X SSPE, 0.1% SDS for 5 min at 25°C and treated for 30 min at 25°C in 2X SSPE containing 20 g/ml RNase A. The filters were then washed 4 times for 30 min in 1X SSPE, 1% SDS at 65°C. The amount of radioactivity present in each slot was determined using a PhosphorImager after overnight exposure, and autoradiograms were exposed for 1-3 days with intensifying screens.
Primer Extension-Primer 1 was designed from the coding region of the human FGF-BP cDNA (5Ј-GTGAGGCTACAGATCTTC-3Ј), primer 2 from the FGF-BP 5Ј-UTR (5Ј-GTTCACCTTGTTCTGAGCACACG-GATCCA-3Ј), and a control primer for the 1.2-kb kanamycin RNA (Promega). 10 pmol of each primer was labeled with T4 polynucleotide kinase (Promega) and 30 Ci of [␥-32 P]ATP (Amersham Pharmacia Biotech) for 1 h, and labeled primers were purified over a Chro-maSpin-10 gel filtration column (CLONTECH). Total RNA from ME-180 cells was isolated as described for Northern analysis. ME-180 mRNA was purified from total RNA over an oligo(dT)-cellulose column (Life Technologies, Inc). 100 fmol of each FGF-BP-specific primer was incubated with or without 7 g of ME-180 mRNA or control primer with or without 2 ng of 1.2-kb kanamycin control RNA (Promega) in the presence of avian myeloblastosis virus reverse transcriptase buffer (50 mM Tris-HCl, 8 mM MgCl 2 , 30 mM KCl, 1 mM DTT, pH 8.5, Boehringer Mannheim) and allowed to anneal for 1 h at 50°C. Annealed mixtures were then incubated in the presence of 2 mM dNTPs, 50 units of RNase inhibitor (Boehringer Mannheim), and 40 units of avian myeloblastosis virus reverse transcriptase (Boehringer Mannheim) for 1 h at 42°C. Samples were then run on a 6% polyacrylamide sequencing gel along with radiolabeled HinfI markers and exposed overnight for autoradiography.
Cloning of FGF-BP Gene Promoter-1.8 kb of genomic sequence lying upstream of the human FGF-BP gene was isolated from a human genomic library using the PCR-based PromoterFinder DNA Walking Kit (CLONTECH) according to the manufacturer's recommendations using rTth XL DNA polymerase (Perkin-Elmer). Gene-specific primers derived from the 5Ј-UTR of the human FGF-BP cDNA were 5Ј-ACACG-GATCCAGTGCAATCC-3Ј (ϩ91 to ϩ72) for the primary round of PCR and 5Ј-GGAGTGAATTGCAGGCTGCAGCTGTGTCAG-3Ј (ϩ62 to ϩ33) for secondary PCR. Secondary PCR products from a DraI library (1.1 kb) and PvuII library (1.8 kb) were cloned into a TA Cloning Vector pCR2.1 (Invitrogen), sequenced by automated cycle sequencing (ABI PRISM Dye Terminator Cycle Sequencing, Perkin-Elmer), and confirmed to contain contiguous genomic sequence. This sequence has been submitted to GenBank TM (accession number AF062639).
Internal promoter deletions were generated by PCR-based site-directed mutagenesis. Complementary overlapping oligonucleotides containing specific promoter deletions were generated as follows. Ets site deletion from Ϫ76 to Ϫ67 was incorporated into primers spanning Ϫ58 to Ϫ93 and Ϫ85 to Ϫ47; AP-1 site deletion from Ϫ65 to Ϫ58 was incorporated into primers spanning Ϫ47 to Ϫ83 and Ϫ72 to Ϫ35; Ets/AP-1 site deletion from Ϫ76 to Ϫ58 was incorporated into primers spanning Ϫ49 to Ϫ93 and Ϫ85 to Ϫ36; C/EBP␤ site deletion from Ϫ47 to Ϫ33 was incorporated into primers spanning Ϫ24 to Ϫ66 and Ϫ57 to Ϫ11; and deletion from Ϫ57 to Ϫ47 was incorporated into primers spanning Ϫ35 to Ϫ75 and Ϫ69 to Ϫ27. The Ϫ58 point mutant was made by incorporating a C to T point mutation at position Ϫ58 into primers representing both strands from Ϫ70 to Ϫ51. For each construct, two separate PCR reactions were carried out with either the BamHI-linked Ϫ118 to Ϫ99 upstream primer or the luciferase-specific downstream primer. The products were separated from excess primers and mixed, denatured, and allowed to reanneal. Amplification of the heteroduplex with overlapping 3Ј ends was carried out by 3Ј extension in the absence of primers followed by amplification using outside primers (BamHIlinked Ϫ118 to Ϫ99 primer and Luciferase-specific primer) in a secondary round of PCR. Final PCR products were digested with BamHI, gel-purified, and ligated into the PXP1 BamHI site. Correct sequence and orientation were verified by dideoxynucleotide chain termination sequencing with a Sequenase kit 2.0 (U. S. Biochemical Corp.). To control for TPA effects on vector sequences unrelated to the FGF-BP promoter insert, we tested a number of non-TPA responsive promoters inserted into the PXP 1 vector for their response to TPA under the conditions used in our experiments. Control vectors, as seen in Fig. 6, consisted of fragments of the pro-opiomelanocortin promoter (10), the thymidine kinase minimal promoter (9), or the CMV minimal promoter 2 A. Aigner and A. Wellstein, unpublished data. which were cloned into PXP1 vector. All these vectors demonstrated an approximately 2-fold induction after TPA treatment (see "Results").
Transient Transfections and Reporter Gene Assays-24 h before transfection, ME-180 cells were plated in 6-well plates in IMEM, 10% FBS at a density of 750,000 cells/well. For each transfection, 1.0 g FGF-BP-luciferase construct and 10 l of LipofectAMINE Reagent (Life Technologies, Inc.) were combined in 200 l of IMEM, and liposome-DNA complexes were allowed to form at room temperature for 30 min. Volume was increased to 1 ml with IMEM, added to rinsed cells, and incubated for 3 h at 37°C. Cells were washed and incubated in IMEM for 3 h and then treated for 18 h with vehicle alone (Me 2 SO, final concentration 0.1%) or 10 Ϫ7 M TPA. Transfection efficiency for each construct was determined by co-transfection with 1.0 ng of a CMVdriven Renilla luciferase reporter vector pRL-CMV (Promega) and found to be the same for all BP-PXP1 constructs. However, due to a 2-fold background TPA induction of pRL-CMV (see above), results were normalized for protein content and not for Renilla luciferase activity. Cells were lysed by scraping into 150 l of Passive Lysis buffer (Promega), and cell debris was removed by brief centrifugation. 20 l of extract was assayed for both firefly and Renilla luciferase activity using the Dual-Luciferase TM Reporter assay system (Promega). Light intensity was measured in a Monolight 2010 luminometer. Light units are expressed firefly light units/g of protein. Protein content of cell extracts was determined by Bradford assay (Bio-Rad).
Gel Shift Assays-ME-180 cells were grown to 80% confluency on 150-mm dishes, serum-starved for 6 h, and treated with or without 10 Ϫ7 M TPA for 90 min. Nuclear extracts were prepared according to Dignam et al. (11) with the following modifications. Pelleted cells were resuspended in 1 ml of buffer A (15 mM KCl, 10 mM HEPES, pH 7.6, 2 mM MgCl 2 , 0.1 mM EDTA, 1 mM DTT, 0.1% Nonidet P-40, 1 mM sodium orthovanadate) (12) with 1ϫ Complete TM protease inhibitor mixture (Boehringer Mannheim) and incubated on ice for 10 min. Crude nuclei were pelleted at 700 g and resuspended in 50 l of ice-cold buffer C (0.42 M NaCl, 20 mM HEPES, pH 7.9, 25% glycerol, 0.2 mM EDTA, 0.5 mM DTT, 1 mM sodium orthovanadate, 1ϫ Complete TM protease inhibitor mixture) and vortexed at 4°C for 15 min. After centrifugation for 10 min at 1000 ϫ g, supernatant was used directly in binding assays and stored at Ϫ70°C.

RESULTS
TPA Increases FGF-BP mRNA in SCCs-We have previously detected an up-regulation of FGF-BP mRNA following TPA treatment of mouse skin during the development of skin tumors (3) and also in human skin xenografts. 2 These data suggest that the control of this angiogenic switch factor may play an important role in skin carcinogenesis. To examine this further we studied the effect of the tumor promoter TPA on FGF-BP gene expression in ME-180 cells which express high levels of the FGF-BP transcript (7). Cells were treated with 10 Ϫ7 M TPA from 1 to 24 h which resulted in an increase in the steady-state levels of FGF-BP mRNA detectable 1 h after treatment (Fig. 1A). PhosphorImager analysis showed that the induction was maximal after 6 h by 452 Ϯ 44% (Fig. 1A). GAPDH mRNA remained unaffected by TPA treatment, as judged relative to the total amount of RNA loaded and was used to standardize FGF-BP mRNA. The dose dependence of TPA induction of FGF-BP mRNA in ME-180 is shown in Fig. 1B. We estimated the half-maximal effective concentration as 1 nM. The inductive effect of TPA on FGF-BP mRNA was also observed in two other SCC cell lines, FaDu and A431 (data not shown) demonstrating that TPA induction of FGF-BP mRNA is generally preserved in SCC cell lines.
To establish whether TPA induction of FGF-BP mRNA was mediated through a PKC-dependent pathway, ME-180 cells were pretreated or not pretreated for 2 h with 100 nM highly specific PKC inhibitor, calphostin C (13), and then treated with or without 10 Ϫ7 M TPA for 4 h. As can be seen in Fig. 1C, pretreatment of the cells with calphostin C prior to TPA treatment totally blocked the TPA effect, demonstrating that the induction of FGF-BP transcript by TPA is mediated via a PKCdependent mechanism. In addition, it is known that TPA causes an immediate up-regulation of PKC followed by long term down-regulation of PKC activity. In contrast, although long term calphostin C appears to be able to down-regulate protein kinase C, it does not cause early induction of PKC activation but rather blocks the inductive effects of TPA on PKC (14). Thus the fact that Calphostin C causes no induction of FGF-BP mRNA and blocks the TPA effect argues that induction of FGF-BP mRNA is through an up-regulation of PKC activity rather than a consequence of long term down-regulation.
Mechanism of TPA Induction of FGF-BP mRNA-We have previously shown that FGF-BP gene expression can be regulated through both transcriptional and post-transcriptional mechanisms (7). Therefore we next attempted to determine whether the TPA induction of FGF-BP mRNA was at the transcriptional or post-transcriptional level. We first assessed whether TPA treatment affected the stability of the FGF-BP mRNA. Experiments were performed to determine whether addition of inhibitors of transcription (actinomycin D) or translation (cycloheximide) could inhibit the TPA induction of FGF-BP mRNA. Actinomycin D (5 g/ml) or cycloheximide (10 g/ml) were added with or without TPA (10 Ϫ7 M), and FGF-BP mRNA levels were determined 6 h after treatment. As shown in Fig. 2A, simultaneous addition of TPA and actinomycin D completely blocked the TPA induction, whereas simultaneous addition of TPA and cycloheximide had no effect. These data suggest that TPA directly increased the rate of FGF-BP gene transcription independently of de novo protein synthesis and did not affect the stability of the FGF-BP transcript. To verify further that the stability of the FGF-BP transcript was not modified by TPA treatment, ME-180 cells were pretreated for 2 h with TPA and then actinomycin D was added to inhibit transcription. As shown in Fig. 2B, pretreatment of cells with TPA did not increase the half-life of the FGF-BP mRNA indicating that the stability of the FGF-BP transcript is not affected by TPA.
To prove directly that TPA increases the rate of transcription of the FGF-BP gene, we then performed nuclear transcription run-on assays. ME-180 cells were treated with or without 10 Ϫ7 M TPA for various periods. 32 P-Labeled nascent transcripts were prepared from isolated ME-180 cell nuclei and hybridized to a nylon membrane bearing immobilized target DNA sequences. As shown in Fig. 3, TPA increased FGF-BP transcript levels maximally after 1 h of treatment. Quantitation and normalization to ␤-actin showed that FGF-BP transcription was up-regulated by 647 Ϯ 1, 448 Ϯ 16, and 197 Ϯ 31% after 1, 4, and 24 h of treatment, respectively (Fig. 3). ␤-Actin plasmid DNA was used as a control since transcription of this gene remained constant. These findings are consistent with the above results studying steady-state mRNA and the effects of actinomycin D and cycloheximide treatment. Clearly the induction of FGF-BP mRNA by TPA in ME-180 cells is directly due to a rapid up-regulation of transcription.
Isolation and Characterization of the Human FGF-BP Promoter-In order to understand better the transcriptional reg-ulation of the human FGF-BP gene, 1.8 kb of genomic sequence upstream to the known 5Ј-UTR sequence of human FGF-BP cDNA was isolated from a human genomic library and sequenced. The transcription start site of the human gene was determined using primer extension analysis with nested primers derived from known cDNA sequence (Fig. 4). The precise start site compatible with the primer extension results is indicated in Fig. 5. Alignment between the human and mouse FGF-BP promoter which we cloned previously (3) revealed a region of high homology with 70% nucleotide identity within the first 200 nucleotides upstream from the transcription start (Fig. 5). Nucleotide homology dropped significantly in more upstream sequences, suggesting that the proximal conserved 200 nucleotides of the promoter could be important for transcriptional regulation of FGF-BP in both species.
Sequence analysis of the promoter demonstrated the presence of numerous consensus transcription factor binding sites that were conserved between mouse and human FGF-BP promoters and that may have functional importance in FGF-BP regulation. As shown in Fig. 5, a consensus TATA box is located at about Ϫ25 base pairs upstream from the transcription start for both promoters. Between Ϫ48 and Ϫ40 of the human FGF-BP promoter we found a highly conserved consensus binding site for C/EBP␤, a member of the CCAAT/enhancer binding protein (C/EBP) family of leucine zipper transcription factors which play a central role in the acute phase response and in a number of cell differentiation pathways (15)(16)(17). An AP-1 consensus binding site (Ϫ65 to Ϫ59) lies juxtaposed to a sequence with homology to an Ets factor binding motif (Ϫ76 to Ϫ68), suggesting potential functional similarity to the juxtaposed Ets/AP-1 site found in the polyoma virus enhancer and in the collagenase promoter (18,19). In addition, a consensus Sp1 factor binding site (Ϫ90 to Ϫ80), an additional Ets factor binding motif (Ϫ107 to Ϫ100), and a potential NF-B-binding site (Ϫ185 to Ϫ176) are located in the conserved region of the promoter and may play a role in transcriptional regulation of FGF-BP as well.
Functional Analysis of the Human FGF-BP Promoter-To identify the functional promoter elements involved in FGF-BP gene regulation by TPA, progressive 5Ј deletion mutants were constructed based on the location of consensus factor binding sites on the promoter. Deletion constructs were transiently transfected into ME-180 cells, and their relative luciferase activity was assayed in the absence or presence of TPA (Fig. 6). The basal activity of each vector is shown in the left panel of Figs. 6 and 7. The empty PXP1 vector had no detectable luciferase activity either in the absence or presence of TPA (data not shown). However, we did observe a background, approximately 2-fold, TPA induction of the PXP1 vector when several unrelated minimal promoters (i.e. thymidine kinase minimal promoter, the CMV minimal promoter, and the pro-opiomelanocortin minimal promoter) were treated with TPA after transfection into ME-180 cells (Fig. 6, control vector). Background induction by TPA of a variety of vectors has been described previously and is presumably mediated through cryptic sites in the PXP1 plasmid (20). The TPA induction due to the inserted FGF-BP promoter was considered to be that observed above the background control vector induction. The FGF-BP promoter from Ϫ1060 to ϩ62 was induced about 4-fold above control vector in the presence of TPA and showed the same TPA inducibility as the full-length 1.8-kb promoter construct (data not shown). Deletion from Ϫ1060 to Ϫ118, which removed 950 base pairs of promoter sequence including the potential NF-B

FIG. 2. Mechanism of TPA induction of FGF-BP mRNA levels.
The respective upper panels of A and B are representative Northern blot analyses performed as described in the legend to Fig. 1. The respective lower panels of A and B represent quantification of data in upper panels. Signal intensities were quantified by phosphorimaging and normalized to GAPDH. Results represent mean Ϯ S.D. of two independent experiments. A, effect of actinomycin D and cycloheximide on the FGF-BP mRNA induction by TPA. ME-180 cells were treated for 6 h in the absence or presence of 10 Ϫ7 M TPA in combination with 5 g/ml actinomycin D (ActD) or 10 g/ml cycloheximide (CHX). B, analysis of turnover of FGF-BP mRNA in TPA-treated cells. ME-180 cells were treated with vehicle alone or 10 Ϫ7 M TPA for 2 h, and 5 g/ml actinomycin D was then added to control and to TPA-treated cells for 0 -16 h. Total RNA was isolated and hybridized sequentially with FGF-BP and GAPDH probes as described in Fig. 1. site, had no effect on TPA induction and was also induced 5-fold above background (Fig. 6). Similarly, deletion from Ϫ118 to Ϫ93, which removed one of the potential Ets-binding sites, retained full TPA induction. Removal of the consensus Sp1 binding site from Ϫ93 to Ϫ77 had no effect on TPA induction of the FGF-BP promoter. However, the 5Ј deletion to Ϫ77 caused an 80% decrease in basal activity of the promoter (Fig. 6, left panel) suggesting that the Sp1 consensus site is a predominant mediator of basal promoter activity of FGF-BP but is not required for TPA induction.
TPA induction was significantly reduced upon deletion of the potential Ets factor binding site from Ϫ77 to Ϫ67 and is reduced even further upon deletion of the AP-1 site from Ϫ67 to Ϫ56 (Fig. 6), indicating that each of these sites contributes to some degree in TPA induction. The basal activity of these constructs was similar at about 10 -20% of the Ϫ118 construct. Finally, deletion from Ϫ56 to Ϫ31, which removes sequences containing homology to a C/EBP␤-binding site, abolished any remaining TPA induction to the background control vector level. The basal activity of this vector was 5% that of the Ϫ118/ϩ62 vector.
Contribution of Ets, AP-1, and C/EBP␤ Sites to TPA Induction-In order to better understand the contribution of each individual consensus binding site to TPA induction, internal promoter deletions were introduced and tested for TPA inducibility within the context of the promoter from Ϫ118 to ϩ62. Deletion of the Ets site alone (Ϫ76 to Ϫ67) or deletion of the AP-1 site alone (Ϫ65 to Ϫ58) reduced TPA induction slightly to the intact promoter (Fig. 7). Deletion of both Ets and AP-1 (Ϫ76 to Ϫ58), however, resulted in a significant decrease in both basal activity and in TPA induction, indicating that both sites act in cooperation for full promoter activity. However, loss of the juxtaposed Ets/AP-1 site does not completely abolish TPA induction, suggesting that additional sites are also involved.
The contribution of the C/EBP␤ binding motif to TPA induction of the FGF-BP promoter was determined by an internal deletion from Ϫ47 to Ϫ33 (Fig. 7, ⌬C/EBP␤). Consistent with the 5Ј deletion construct which contained only the C/EBP␤ site and retained some TPA inducibility (Fig. 6, Ϫ56/ϩ62), an internal deletion of this site showed a significant decrease in TPA induction. Activation of C/EBP␤ has been shown to occur through ras-dependent phosphorylation and is involved in phorbol ester induction of genes such as MDR1 (21)(22)(23). Similarly, C/EBP␤ seems to play a role in TPA induction of the FGF-BP promoter since deletion of this site reduces the overall induction by TPA.
TPA Regulation of the FGF-BP Promoter Involves a Repressor Element Juxtaposed to the AP-1 Site-Between the AP-1 site and the C/EBP␤ site lies a region of low homology between the human and mouse EGF-BP promoter sequences. Because this region was not suspected to have any effect on TPA induction, an internal deletion removing this region (Ϫ57 to Ϫ47) was tested as a control. Surprisingly, in the ⌬57/47 construct, TPA induction of the FGF-BP promoter increased from approximately 5 to 11-fold, suggesting the presence of a possible repressor which may interact with this site. The lack of sequence conservation between the human and mouse in this region may reflect a difference in the regulation of FGF-BP between the two species. The Ϫ57 to Ϫ47 deletion disrupts an AACGTG (Ϫ60 to Ϫ55) which is juxtaposed to the 3Ј end of the AP-1 site and which shows some similarity to the CACGTG E box element recognized by a number of helix-loop-helix factors (24). To test this imperfect E box for repressor activity, a C to T point mutation at position Ϫ58 was introduced into the Ϫ118/ϩ62 BP promoter construct (Fig. 7). This mutant-58 (m-58) construct showed a dramatic increase in TPA induction up to 16-fold above background. Moreover, when the Ϫ58 point mutation is introduced into the ⌬C/EBP␤ construct (Fig. 7, ⌬C/EBP␤/m-58), this promoter mutant also showed increased fold induction by TPA, suggesting that repression mediated by this site is not dependent on the C/EBP␤ site. These data show that the point mutation at position Ϫ58, as well as the internal deletion from Ϫ57 to Ϫ47, disrupts repression of the FGF-BP promoter which normally limits the response to TPA.
Transcription Factor Binding to FGF-BP Promoter Elements-In order to ascertain that TPA induction of FGF-BP was due to direct activation by transcription factors, we performed gel retardation analysis to show transcription factor binding to FGF-BP promoter elements. By using labeled promoter sequence from Ϫ80 to Ϫ63 containing the putative Etsbinding site as a probe (Fig. 8A), the binding of three specific protein complexes in the presence of ME-180 nuclear extracts was detected (Fig. 8B). Protein binding to all three complexes was increased in the presence of TPA (Fig. 8B, lane 3) and was specifically competed away in the presence of excess unlabeled Ϫ80/Ϫ63 oligonucleotides (lanes 4 and 5). Further competition analysis showed that the factors binding to the Ϫ80/Ϫ63 element were only weakly competed by consensus Ets elements from the collagenase promoter (25) and polyoma virus enhancer (26), requiring over 100-fold excess in order to compete for binding (data not shown). It has previously been described that specific residues flanking the GGA trinucleotide motif of the Ets site are required for high affinity sequence-specific binding of individual Ets family members (27)(28)(29). Therefore, our data suggest that the Ϫ80/Ϫ63 element on the FGF-BP promoter could bind an Ets family member other than the Ets-1 or Ets-2 proto-oncogenes (26).
To determine transcription factor binding to the C/EBP␤ site, gel shift analysis was carried out using labeled promoter sequence from Ϫ55 to Ϫ30 (Fig. 8A) as a probe. In the presence of ME-180 extracts, the Ϫ55/Ϫ30 element bound one predominant complex (Fig. 8C), which demonstrated increased binding in the presence of TPA (lane 3). The majority of the complex was competed away in the presence of excess unlabeled Ϫ55/ Ϫ30 oligonucleotide (lanes 4 and 5), indicating that binding was specific. Competition by the C/EBP␤ site from the p21 WAF1/CIP1 gene promoter (30) for the specific complex was effective only at high molar excess (data not shown) indicating that the FGF-BP Ϫ55/Ϫ30 element may bind a different C/EBP␤-related factor.
To investigate further the transcriptional activation of the FGF-BP promoter by AP-1 and the involvement of the variant E box repressor element, gel shift experiments were carried out using the labeled promoter sequence spanning the juxtaposed AP-1/repressor element as a probe (Ϫ70/Ϫ51, Fig. 8A). In the presence of ME-180 nuclear extracts, the Ϫ70/Ϫ51 element bound an upper complex (Fig. 8D, arrow) and a lower doublet (bracket). The binding of all three complexes was highly induced by TPA (Fig. 8D, lane 3) and was effectively competed by molar excess of the unlabeled Ϫ70/Ϫ51 oligonucleotide (lanes 4 and 5). To understand better the specific composition of these complexes, point mutations were introduced in either the AP-1 site (mut AP-1) or the repressor site (mut Ϫ58) and tested for their abilities to compete for binding. Competition with the mutant AP-1 site resulted in a decrease of only the bottom doublet and no competition for the upper band (lanes 6 -8), suggesting that the upper band corresponds to factors bound specifically to the AP-1 site. Conversely, when competition was carried out with the repressor mutant (mut Ϫ58), binding of the doublet on the probe remains intact, whereas binding to the AP-1 site is reduced (lanes 9 -11), indicating that the lower two bands represent distinct protein binding to the repressor element. Furthermore, when competition was carried out with an AP-1 consensus, which contains an AP-1 site flanked by sequences which are not homologous to the FGF-BP promoter, competition for only the upper AP-1 complex was observed (lanes 12 and 13). These results show that the AP-1 site and the repressor site of the FGF-BP promoter bind distinct and specific transcription factor complexes that are induced in the presence of TPA. Taken together, our data show that sequences between Ϫ77 and Ϫ33 of the FGF-BP promoter form a novel TPA regulatory cassette consisting of interacting positive and negative control elements. DISCUSSION In this report we demonstrate that TPA induction of FGF-BP mRNA levels is primarily through stimulation of gene transcription. This is in contrast to the retinoid repression of FGF-BP gene expression which we have previously shown is mediated through post-transcriptional and transcriptional mechanisms (7). In fact, at least at early time points after retinoid administration, the post-transcriptional mechanism which is dependent on new protein synthesis predominates since the half-life of the FGF-BP mRNA is greater than 16 h (7). Our studies show that the TPA induction of FGF-BP mRNA is rapid, requiring no new protein synthesis and involves direct activation by transcription factors whose site of action is clustered in the first 118 base pairs upstream of the transcription start site. Within this region the majority of the TPA stimulation of the FGF-BP promoter can be explained by the additive effects of two sites positioned between Ϫ76 to Ϫ58 and from Ϫ47 to Ϫ33.
The Ϫ76 to Ϫ58 site harbors a perfect consensus to the AP-1 transcription factor binding site NTGAGTCA (31). The AP-1 transcription factor complex comprises the c-fos and c-jun proto-oncogenes which are known to be activated as a result of TPA stimulation of PKC-dependent pathways (32). However, deletion of the AP-1 site alone in the FGF-BP promoter caused only a slight reduction in TPA effects on the FGF-BP promoter. This result is consistent with the emerging picture that AP-1 acts synergistically with other transcription factors, such as the Ets family of transcription factors, to mediate gene expression in response to TPA and other stimuli (28,29). In the FGF-BP promoter deletion of sequences 5Ј to the AP-1 consensus significantly decreases the TPA stimulation in comparison with deletion of the AP-1 site alone. These 5Ј sequences contain the core GGA found in the center of the Ets family DNA consensus recognition site (29). Considering the body of evidence that suggests that Ets/AP-1 cooperate for full transcriptional activation, it seems likely that this may be the function of the Ϫ76/Ϫ58 element. For instance similar cooperation between Ets and AP-1 occurs through a juxtaposed Ets/AP-1binding site in the polyoma virus enhancer (18) and has subsequently been implicated in the regulation of genes involved in invasion and metastasis, including collagenase and urokinase plasminogen activator (19,(33)(34)(35)(36)(37). Although we found that the collagenase Ets element or the polyoma virus Ets element did not effectively compete for binding to the FGF-BP Ets element, this may reflect the binding of another Ets family member to the FGF-BP promoter whose recognition site could be determined by sequences flanking the GGA core (27).
Deletion of the Ϫ47 to Ϫ33 FGF-BP promoter region also FIG. 7. Effect of internal deletions on basal activity and TPA-induced activity of the FGF-BP promoter. The left histogram shows the impact of each mutation on basal activity of the construct (center). The control vector shown is the thymidine kinase minimal promoter in PXP1. The right histogram shows the transcriptional activity in the presence of TPA 10 Ϫ7 M for each internal deletion of FGF-BP promoter. Each promoter construct was transiently transfected into ME-180 cells, and luciferase activity is expressed as fold induction of TPA-treated over untreated for each construct. Values represent the mean Ϯ S.E. from at least three separate experiments, each done in triplicate wells. Asterisk indicates significant difference (p Ͻ 0.05) from the Ϫ118/ϩ62 promoter construct.
substantially reduces the TPA effects on the FGF-BP promoter. Sequence analysis revealed that a site homologous to the C/EBP␤-binding site is centered in this region of the promoter. The factors binding to FGF-BP C/EBP␤ element, however, are not effectively competed by the C/EBP␤ site from the p21 WAF1/ CIP1 gene promoter, suggesting that transcription of FGF-BP may be mediated by a different C/EBP␤-related factor. The published consensus for C/EBP␤ is T(T/G)NNGNAA(T/G) (38) which is identical in eight positions (underlined) to the site between Ϫ48 to Ϫ41 differing only in the most 3Ј-nucleotide of the consensus. In addition, the involvement of C/EBP␤ in TPAmediated responses has been shown previously. For instance, induction by phorbol esters has been shown to cause increased C/EBP␤ synthesis, phosphorylation, and DNA binding to promoters of a number of genes including MDR1 and collagenase 1 (21-23, 39). Thus, C/EBP␤ or a family member is involved in the activity of the Ϫ47 to Ϫ33 element. Like other leucine zipper family members, C/EBP␤ acts cooperatively with other transcription factors to modulate the level of gene expression in response to extracellular stimuli. For example, C/EBP␤ has been shown to associate with Fos/Jun in vitro (40) and can cooperate in vivo to induce expression of the TSG-6 gene in response to interleukin-1 and tumor necrosis factor-␣ which is mediated through distinct AP-1 and C/EBP␤-binding sites in the TSG-6 promoter (41). Similarly, our data show that the C/EBP␤ consensus element is a major mediator of TPA-induced gene expression of FGF-BP. However, because removal of the C/EBP␤ site alone does not completely abolish TPA induction, this suggests that like other TPA-induced genes, the C/EBP␤ site acts in cooperation with other promoter elements.
A novel aspect of TPA regulation of the FGF-BP promoter is the role of the region Ϫ57 to Ϫ47 between the AP-1 site and the C/EBP␤ site. Deletion of this region substantially increases the TPA response, implying that this region normally represses the extent of the response to TPA. A point mutation in this region also abrogates repression thus making it unlikely that the effect of the deletion is simply to bring the AP-1 and C/EBP␤ sites in closer proximity leading to their increased responsiveness to TPA. In fact, the relief of repression obtained with the Ϫ58 point mutant is observed in the presence of the C/EBP␤ deletion suggesting that the repression impacts on the AP-1 element rather than the C/EBP␤ site. An alternate possibility is that the factor bound to the Ϫ57 to Ϫ47 site interacts with the general transcription machinery in a manner similar to the NC2 repressor (42). However, this seems less likely because we observe no increase in basal activity of the promoter after deletion or mutation of the repressor site in comparison to the Ϫ118 construct (Fig. 7). The Ϫ57 to Ϫ47 deletion destroys an AACGTG (Ϫ60 to Ϫ55) which is a variant of the CACGTG E box element recognized by a number of helix-loop-helix factors (24). The Ϫ58 mutant changes the AACGTG to AATGTG and would perturb the 5Ј part of the dimer recognition sequence (24). However, the wild type sequence alone does not predict which member of the helix-loop-helix family would interact with this site. Interestingly, binding to an AACGTG recognition element has been described in vitro to a homodimer of the aryl hydrocarbon receptor nuclear translocator helix-loop-helix factor (43), and aryl hydrocarbon receptor nuclear translocatordeficient embryonic stem cells have a defective angiogenesis process (44). However, it is unclear whether aryl hydrocarbon receptor nuclear translocator homodimers interact with promoters in vivo. Alternatively, other helix-loop-helix factors are known to function as transcriptional repressors, such as the Mad family of proteins that bind related E box sequences during TPA-induced macrophage differentiation (45,46) and recruit the mSin3-histone deacetylase corepressor complex, leading to a more closed chromatin structure and transcriptional repression (47).
Through gel retardation analysis, we show distinct factor binding to the AP-1 site and to the E box repressor site. Interestingly, factor binding to both of these sites is increased upon stimulation with TPA. TPA-induced transcription factor binding to E box elements has been described for a number of different promoters including c-fos (48 -50). The observation that TPA induces factors which both stimulate and limit induction of FGF-BP suggests a mechanism by which transcription of the FGF-BP gene could be tightly regulated and may reflect a level of tissue-specific expression of this gene.
Overall, our data suggest that the TPA induction of the FGF-BP promoter is induced through both Ets/AP-1 site and a C/EBP␤ site and that the extent of induction is moderated by factors that bind to an E box repressor element which lies adjacent to the AP-1 site. It is known that TPA also induces the expression of genes involved in proteolytic degradation of the extracellular matrix such as stromelysin, collagenase, and urokinase plasminogen activator (33,51,52). Interestingly, these promoters are regulated by similar transcription factors as those which we show are involved in FGF-BP promoter induction, e.g. Ets/AP-1 and C/EBP␤. Thus, our data would support the argument that a specific subset of transcription factors may be induced (or derepressed) to specifically stimulate a panel of genes involved in invasion, angiogenesis, and metastasis during skin tumor development.