Transcription Factor AP-2 Controls Transcription of the Human Transforming Growth Factor-α Gene*

The epidermal growth factor receptor is vital for normal development and plays a role in oncogenesis. The level of activation of this receptor by transforming growth factor-α (TGF-α) is controlled, in part, by the rate of transcription of the TGF-α gene. In the characterization of the proximal TGF-α promoter by DNase I footprinting, a 43-base pair element (−88 to −130 relative to the transcription start site), designated TαRE I, was found that was specifically protected by nuclear proteins from human mammary carcinoma MDA468 cells. TαRE I was essential for the maximal expression of the TGF-α gene as indicated by deletion and mutagenesis analyses. TαRE I consists of two cis-acting elements, a proximal regulatory element (PRE, −89 to −103) and a distal regulatory element (DRE, −121 to −128). Both elements were able to form specific complexes with protein from MDA468 cell nuclear extracts and are necessary for the full activity of the entire 1.1-kilobase pair TGF-α promoter. Competition and antibody studies determined that the DRE contains a binding site for the transcription factor AP-2, while the protein that binds to the PRE has yet to be identified. When linked upstream to the heterologous herpes simplex thymidine kinase promoter, the TαRE I enhanced transcription up to 11-fold in MDA468 cells. Cotransfection of an AP-2 expression vector was able to activate transcription from the TαREI-TK construct in a DRE-dependent manner. These results further our understanding of how TGF-α transcription is regulated.

Transforming growth factor-␣ (TGF-␣) 1 is a polypeptide mitogen belonging to the epidermal growth factor (EGF) family (1). TGF-␣ is synthesized as a 160-amino acid transmembrane precursor from which the 50-amino acid mature peptide is released (2). Both the membrane-anchored and soluble forms are biologically active through interaction with cell surface EGF receptors.
First discovered in the media of retrovirally transformed fibroblasts (3,4), TGF-␣ appears to play an important role in oncogenesis. It has been found to be associated with many tumors such as mammary, squamous, and renal carcinomas, melanomas, hepatomas, and glioblastomas (5)(6)(7). When trans-fected with a TGF-␣ expression vector, certain cell lines became transformed (8 -10). Also, transgenic mice overexpressing TGF-␣ often develop neoplastic lesions in the tissues to which TGF-␣ overexpression has been directed (11)(12)(13)(14). In many cases, tumor cells coexpress TGF-␣ and the EGF receptor, suggesting an autocrine mechanism of growth stimulation (15). That TGF-␣ is also a potent inducer of angiogenesis in vivo (16) suggests that host support for the tumor might also be induced through the expression of this growth factor by the tumor cells. These models, indicating the involvement of TGF-␣ in tumorigenesis, utilize heterologous promoters; however, TGF-␣ gene transcription is also up-regulated through its own promoter. For example, the TGF-␣ promoter regulates expression of this gene through development (17) and mediates changes induced by DNA methylation (18), hormones (19 -23), and high glucose concentrations (24). To determine how the endogenous TGF-␣ gene is regulated, the promoters of the human and rat TGF-␣ genes have been cloned (23,25,26). Their most characteristic features are the apparent lack of TATA and CCAAT sequences (25) and high GC contents. Elements in the human and rat TGF-␣ promoters have been partially characterized. The human core promoter contains a nonconsensus TATA box that is recognized by TATA-binding protein (27) and an initiator element that accurately orients the transcription complex, giving rise to an unique transcription initiation site (28). In addition, Sp1 (human and rat) and p53 (human) play roles as transcriptional enhancers for the TGF-␣ promoter (18,27,29).
In our characterization of the proximal TGF-␣ promoter, we noted a 43-bp DNase I footprint that contained a GC-rich element that resembled a consensus Sp1 binding site. This paper describes a more detailed analysis of the proteins that bind to this element. These studies revealed that the footprint results from the binding of two sequence-specific transcription factors, one of which we have identified as AP-2. These proteins bind simultaneously to this element and are involved in the regulation of the 1.1-kb TGF-␣ promoter activity. Our studies indicate that AP-2 regulates TGF-␣ transcription in vivo through its site in the TGF-␣ promoter. Our demonstration of the involvement of AP-2 in the regulation of the TGF-␣ promoter may contribute to our understanding of how TGF-␣ expression is regulated during carcinogenesis and during development.

EXPERIMENTAL PROCEDURES
Plasmids and Site-directed Mutagenesis-The 220-bp TGF-␣ promoter (Ϫ181 to ϩ38, relative to the transcription initiation site) was cloned upstream of the luciferase gene in the pGL2 basic plasmid (Promega) to make pGL-181/ϩ38 as described previously (28). The 1.1-kb TGF-␣ promoter (Ϫ1069 to ϩ38) was cloned into the SacI and HindIII sites in the same vector to give pGLT1.1.
Mutations were introduced into the TGF-␣ promoter by site-directed mutagenesis. pGL2 basic plasmid contains a bacteriophage replication origin and was used to generate single-stranded template DNA. The sequences of the mutagenic oligonucleotide primers for each mutant * This work was supported by a grant from the National Institutes of Health, DK43652. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
To construct deletion mutant p⌬Ϫ89/Ϫ128, pMϪ89/Ϫ96, and pMϪ121/Ϫ128 were digested with EcoRI. The resulting 5-kb fragment from pMϪ121/Ϫ128 and the 0.7-kb fragment from pMϪ89/Ϫ96 were purified by agarose gel electrophoresis and ligated to each other in proper orientation. To construct pDϪ68, the sequence from Ϫ65 to Ϫ70 in the TGF-␣ promoter was mutated to a ApaI site by oligonucleotidedirected mutagenesis. The plasmid was then cut with ApaI and EcoRI, and the 0.7-kb fragment was isolated and ligated into the same sites in pBluescriptKS(ϩ) (Stratagene). The resulting plasmid was cut with HindIII and KpnI, and the 110-bp fragment was ligated into the same sites in pGLϪ181/ϩ38, thus substituting for the wild type 220-bp TGF-␣ promoter. The same strategy was used to construct pDϪ126 except that the sequence from Ϫ123 to Ϫ128 in the TGF-␣ promoter was mutated to an ApaI site. pDϪ15 was constructed by digesting pGLϪ181/ϩ38 with SacII and religating the resultant ϳ5.8-kb fragment.
Oligonucleotides containing either wild type T␣RE I (Ϫ83 to Ϫ138 of the TGF-␣ promoter) or T␣RE I with a mutation at the distal regulatory element (DRE) were synthesized, with the HindIII and XhoI sites at the 5Ј-and 3Ј-ends, respectively. These oligonucleotides were annealed and ligated into the same sites in pT81Luc (ATCC, plasmid 37584) to make pT␣REI-TK and pDREm-TK. pSX is a eukaryotic expression vector containing the SV-40 promoter. AP-2 cDNA was cloned into multiple cloning sites of pSX to make pSX-AP2.
Cell Culture and Transfection-MDA468 and HepG2 cells were maintained at 37°C in a CO 2 incubator in Dulbecco's modified Eagle's medium supplemented with 10% newborn calf serum or 10% fetal bovine serum, respectively. Medium of HepG2 cells also contains 0.1 mM minimum essential medium nonessential amino acid solution (Life Technologies, Inc.). Cells were grown to confluence, washed with phosphate-buffered saline, trypsinized, and resuspended in full culture medium. For transfection with MDA468 cells, approximately 1.0 ϫ 10 7 cells were transfected by electroporation with a total of 30 g of plasmid DNA consisting of 20 g of the indicated promoter-luciferase construct and 10 g of the pCMV-␤-gal, a plasmid containing the CMV promoterdriving ␤-galactosidase as a control for transfection efficiency (18). For transfection with HepG2 cells, approximately 5.0 ϫ 10 6 cells were electroporated with 5 g of the indicated luciferase reporter plasmid, 5 g of pCMV-␤gal, and 1 g of either pSX or pSX-AP2. Cells were harvested 24 h after transfection, and luciferase and ␤-galactosidase activities were assayed as described previously (18).
Nuclear Extracts and Gel Mobility Shift Assay-Nuclear extracts were prepared from either MDA468 cells or HepG2 cells (30). Doublestranded oligonucleotides for gel mobility shift assays were labeled with either [␣-32 P]dATP or [␣-32 P]dCTP using the Klenow fragment of DNA polymerase I. The binding reactions were carried out at room temperature for 30 min in a total volume of 30 -35 l consisting of 50,000 -80,000 cpm of probes, 6 g of MDA468 cell nuclear extract protein, 20 mM HEPES (pH 7.9), 25 mM KCl, 2 mM spermidine, 0.1 mM EDTA, 2 mM MgCl 2 , 10% glycerol, 1 mM dithiothreitol, 1 g of poly(dI-dC), and 0.1 mg/ml bovine serum albumin, with or without an excessive amount of competitor oligonucleotides. The reactions were then resolved on a 6% nondenaturing acrylamide gel in 0.5 ϫ Tris borate-EDTA buffer. For the antibody supershift assay, 0.5-2 ng of the rabbit polyclonal anti-AP-2 antibody (Santa Cruz Biotechnology, Inc.) was also included in the reaction.
DNase I Footprinting-pGL-181/ϩ38 was cut by HindIII, and the exposed 3Ј-end of the noncoding strand of the TGF-␣ promoter was labeled with [␣-32 P]dATP using the Klenow fragment of DNA polymerase I. The plasmid was then purified, and the 220-bp radiolabeled fragment of the TGF-␣ promoter was excised from the plasmid DNA using KpnI. The resulting labeled fragment was purified by agarose gel electrophoresis. The binding reaction for DNase I footprinting was carried out in a total volume of 45 l containing 30,000 -50,000 cpm of probe DNA, 0 or 5 l of partially purified MDA468 cell nuclear extract, 50 mM KCl, 0.1 mg/ml bovine serum albumin, 25 mM Tris-HCl (pH 8.0), 6 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, 10% glycerol, and 100 ng of poly(dI-dC). After incubation on ice for 15 min, the reaction was subjected to DNase I digestion in the presence of 2.5 mM CaCl 2 for 1 min at room temperature. 100 l of the stop solution (100 mM NaCl, 20 mM EDTA, 1% SDS, 0.2 mg/ml proteinase K) was added to abolish DNase I digestion. The DNA was then separated from the proteins by phenol: chloroform extraction and ethanol precipitation and resolved on a 5% denaturing polyacrylamide gel in a DNA sequencing gel apparatus (Bio-Rad).

Identification of a Potential Target for Transcription Factors in the Human TGF-␣ Promoter-
To identify the sequence(s) that is important for TGF-␣ promoter activity, 5Ј-deletion mutants of the promoter were placed upstream of the luciferase gene and transfected into MDA468 human mammary carcinoma cells. As shown in Fig. 1A, the deletion of the distal 1-kb sequence, from Ϫ1069 to Ϫ181, caused only a minimal reduction (35%) of the reporter activity. Strikingly, further deletion of a 114-bp fragment (from Ϫ181 to Ϫ68) resulted in the almost complete loss of the promoter activity. Similar results were obtained using HepG2 cells (data not shown). These data indicate that this proximal region of the TGF-␣ promoter contains most of the basal regulatory information for TGF-␣ transcription.
To identify potential target(s) for transcription factors in this proximal TGF-␣ promoter, we performed DNase I footprinting experiments using the 220-bp promoter fragment (Ϫ181 to ϩ38) as probe. A 43-bp element was footprinted by MDA468 nuclear extract and designated T␣RE I (Fig. 1B, lanes 2 and 5). DNA sequences between Ϫ88 and Ϫ130 on the upper strand and between Ϫ95 and Ϫ114 on the lower strand of the TGF-␣ promoter were protected from DNase I digestion (Fig. 1C). TGF-␣ expression can be up-regulated by EGF (20 -23). However, responsiveness of the TGF-␣ promoter to EGF was not affected by the deletions shown in Fig. 1A (data not shown). Also, protein binding to the promoter remained unchanged upon EGF stimulation (compare lanes 2 and 3 and lanes 5 and 6, Fig. 1B). These data suggest that transcription factors acting through T␣RE I regulate TGF-␣ basal transcription, whereas the EGF-responsive element resides elsewhere in the promoter.
Mutagenesis Analysis of T␣RE I-To further define T␣RE I, a series of site-directed mutations were introduced into this proximal sequence (Fig. 2). The mutated TGF-␣ promoters were cloned upstream of the firefly luciferase reporter gene and transfected into MDA468 cells. Transcription from three of the mutant promoters, MϪ89/Ϫ96, MϪ97/Ϫ103 and MϪ121/Ϫ128, was decreased to 38, 51, and 48% of the wild type promoter activity, respectively, whereas that of MϪ104/Ϫ111 and MϪ112/Ϫ118, retained the wild type activity, at 120 and 97%, respectively. Again, T␣RE I was found to be nonessential for EGF stimulation because each mutant promoter responded equally well to EGF with a 4-fold increase in reporter activity (data not shown).
This mutational analysis suggests that T␣RE I contains two cis-acting elements, Ϫ89 to Ϫ103 and Ϫ121 to Ϫ128, designated the PRE and the DRE, respectively. Furthermore, in MDA468 cells, the T␣RE I was able to enhance transcription from the heterologous herpes simplex TK promoter, when placed upstream of this promoter (Fig. 2). While mutations in either the PRE or DRE resulted in decreased promoter activity, simultaneous deletion of both the PRE and DRE (⌬Ϫ89/Ϫ128) resulted in the same loss of promoter activity that was seen with mutation in each of the elements alone (Fig. 2). Whether these elements act in concert in the regulation of the TGF-␣ promoter activity such that both elements are necessary for full activity of the T␣RE I will be under further investigation. A similar transcription requirement for T␣RE I, especially the DRE, has also been observed in HeLa cells (data not shown).
Decreased Transcriptional Activity Is Related to Altered Protein Binding-We determined whether protein binding to the T␣RE I was affected by mutations at the PRE or DRE. We performed a DNase I footprinting experiment on the 220-bp promoter fragments containing the mutations at either the DRE or PRE (Fig. 3). We found that the three mutant promoters, MϪ89/Ϫ96, MϪ97/Ϫ103, and MϪ121/Ϫ128, which showed decreased promotor activities in the transient transfection assays, also displayed altered patterns of protein binding. Interestingly, mutations directed to either the PRE or DRE alone only abolished protein binding to that same site but did not affect protein binding to the other site. This result also indicates that there are two protein binding sites in T␣RE I that correspond to the functionally defined DRE and PRE. The protein binding to either site is independent on the protein binding to the other site.
T␣RE I Consists of Two Distinct Protein Binding Sites-To further characterize the nuclear proteins that bind to the T␣RE I, we used oligonucleotides containing either the PRE or DRE sequence as probes in gel mobility shift assays (Fig. 4). Both PRE and DRE oligonucleotides formed complexes with MDA468 nuclear proteins (lanes 2 and 6). The complex formed on the DRE and PRE oligonucleotides had similar gel mobility, but the morphology of the bands were different. The DRE complex consisted of a relatively sharply defined band, whereas the PRE complex appeared more diffuse and is of slightly higher mobility. The binding of these proteins to the oligonucleotides was sequence-specific. The formation of these complexes was inhibited by the presence of a 100-fold excess of the corresponding unlabeled oligonucleotides (lanes 3 and 8) but the PRE and DRE oligonucleotides could not compete with each other for protein binding (lanes 4 and 7). This observation suggests that different proteins bind to these elements. We attempted to determine whether the PRE-and DRE-binding proteins interact with each other with sufficient affinity to allow the detection of a complex using the gel mobility shift assay. Assuming that the PRE and DRE oligonucleotides are each bound by a single protein (see below), interaction between these proteins could be detected by the formation of a higher order and lower mobility complex in a binding reaction that included both oligonucleotides. However, in the gel mobility shift assay using both oligonucleotides, we observed only a simple mixture of the PRE and DRE complexes and no higher order complexes (lane 9). As before, the competition reactions confirmed the specificity of the binding proteins for their respective elements (lanes 10 and 11). This result indicates that even if an interaction between these DNA-binding proteins can occur, it is not sufficiently stable to allow detection by the gel mobility shift assay.
AP-2 Binds to the DRE-Sequence analysis of the DRE indicated the presence of a potential binding site for either transcription factor Sp1 or AP-2 (Fig. 5A). In a gel mobility shift assay using the DRE as a probe, a 100-fold excess of an Sp1 consensus oligonucleotide failed to compete for protein binding (Fig. 5B, left panel, lane 4). However, a 50-fold excess of an AP-2 consensus oligonucleotides completely inhibited the formation of the DRE-protein complex (lane 5). Moreover, the mobility of the DRE-protein complex was further retarded by the addition of rabbit polyclonal IgG against AP-2 (lane 6), but not by the addition of anti-retinoblastoma protein antibody (lane 7). In a parallel experiment, a 25-fold excess of the DRE probe completely blocked protein binding to the consensus AP-2 sequence (Fig. 5C, lane 4). Taken together, these results

FIG. 1. Identification of a potential target for transcription factors in the human TGF-␣ promoter.
A, 5Ј-deletion analysis of the 220-bp TGF-␣ promoter. Various portions of the human TGF-␣ promoter were cloned upstream of the luciferase gene to make pGLT1.1, pGLϪ181/ϩ38, pDϪ126, pDϪ68, and pDϪ15, respectively. 20 g of each construct was transfected into MDA468 cells together with 10 g of the ␤-galactosidase expression vector driven by the CMV promoter. Basal transcription represents luciferase/␤-galactosidase activity, with that of the wild type standardized to 1. Luciferase activity from the wild type promoter ranged between 10,000 and 20,000 light units, while the background activity from promoterless constructs was between 100 and 200 light units. B, DNase I footprinting showing regions within the wild type 220-bp promoter (Ϫ181 to ϩ38) that were protected by MDA468 cell nuclear extracts (NE). Next to the gel, the shaded box indicates protected region on the lower strand, whereas the lined box indicates protected region on the upper strand. C, diagram highlighting protected regions on the TGF-␣ promoter.
indicate that AP-2 binds to the DRE. The complex formed on the PRE could not be supershifted or inhibited by an AP-2 antibody (Fig. 5B, right panel, lanes 3 and 4), indicating that AP-2 neither interacts with the PRE nor is complexed by protein-protein interaction with the transcription factor that binds the PRE.
In Fig. 2, we have shown that a mutation at DRE decreased transcription from the TGF-␣ promoter to 48% of wild type (MϪ121/Ϫ128). An oligonucleotide containing the same mutation failed to compete with either the wild type DRE sequence or the consensus AP-2 element for protein binding (Fig. 5C,  lanes 7 and 11). Thus, the DRE is recognized in a sequencespecific manner both in the intact cell and in vitro.
AP-2 Site in the DRE Is Functional-It has been previously shown that the HepG2 human hepatoma cell line lacks endogenous AP-2 activity. Cotransfection experiments were done to test the functional importance of the AP-2 site in the TGF-␣ promoter (Fig. 6A). When the wild type T␣RE I was placed upstream of a heterologous TK promoter (T␣REI-TK), cotransfection of a eukaryotic expression vector for AP-2 (pSX-AP2) increased transcription 2.2-fold compared with pSX alone. AP-2 cotransfection had no significant effect on expression from the TK promoter alone. This AP-2-dependent increase in transcription requires an intact DRE sequence, as indicated by the fact that AP-2 cotransfection failed to activate when the T␣RE I element contained the MϪ121/Ϫ128 mutation at the DRE site (DREm-TK).

FIG. 2. Mutagenesis analysis of T␣RE I.
The top line shows the wild type T␣RE I sequence. Two cis-elements defined within the T␣RE I, PRE and DRE, are underlined and indicated. Mutant promoters are designated by the 5Ј and 3Ј positions of the mutated or deleted sequences. Each mutated sequence is also shown. The promoter constructs were placed upstream of the luciferase gene in the pGL2 or pT81Luc plasmid, and 20 g of each plasmid was transfected into MDA468 cells. 10 g of the ␤-galactosidase expression vector was cotransfected to normalize for transcription efficiency. Relative promoter activities are diagrammed on the right, with that of the wild type TGF-␣ and the TK standardized to 1.  1-4) or PRE sequence (lanes 5-8) were radiolabeled and incubated with nuclear extracts from MDA468 cells. In lanes 9 -11, both PRE and DRE probes were used. Where indicated, a 100-fold excess of the competitors was also added. The reaction mix was run on a nondenaturing polyacrylamide gel to separate the protein-bound from the free DNA probe. An autoradiogram of the gel is shown.
That the transactivation by AP-2 is a result of AP-2 binding to the DRE is further evidenced by a gel mobility shift assay using nuclear extracts from HepG2 cells (Fig. 6B). No specific DNA/protein interaction was detected on the DRE probe as indicated by competition experiments using a 100-fold excess of the unlabeled probe (compare lanes 2 and 3), consistent with the previous observation that HepG2 cells have no endogenous AP-2 activity. However, the addition of exogenous AP-2 protein resulted in three additional shifted bands, I, II, and III (lane 4). Complexes II and III appear as strong shifts, whereas complex I is much weaker. A 100-fold excess of either unlabeled DRE or AP-2 DNA competed protein binding to all three bands, indicating the specificity of these shifts (lanes 5 and 6, respectively). Complexes I, II, and III were further confirmed to contain AP-2 by the quantitative supershifting of these bands with ␣-AP-2 antibody (lane 7).
The PRE and DRE Are Important in the Context of the 1.1-kb TGF-␣ Promoter-The mutagenesis study in Fig. 2 used the proximal 220-bp promoter. However, in vivo, TGF-␣ transcription is regulated in the context of a larger promoter. We therefore determined the role of the DRE or PRE in the context of the entire promoter as it has thus far been defined. We constructed was also used to "supershift" those complexes containing the epitope recognized by the antibody. The major band in the right part is the PRE-protein complex. The AP-2 antibody failed to supershift this complex. C, gel mobility shift assay using AP-2 (lanes 1-7) or DRE (lanes 8 -11) probe and MDA468 cell nuclear extracts. The arrows to the left and right of the panel indicate specific DNA/protein complexes detected by AP-2 or DRE probe, respectively. Lanes 3, 7, 10, and 11 contain a 100-fold excess of the indicated competitor oligonucleotides. Lanes 4 -6 contain a 25-, 50-, and 100-fold excess of the DRE oligonucleotide as competitor, respectively.

FIG. 6. Functional analysis of the DRE in AP-2-deficient cells (HepG2).
A, cotransfection of an AP-2 expression vector activates transcription directed by T␣RE I. HepG2 cells, which are known to be AP-2-deficient, were transfected with reporter plasmid containing the Herpes simplex TK basal promoter or the TK promoter coupled to wild type (T␣REI-TK) or mutant (DREm-TK) TGF-␣ elements together with either an AP-2 expression vector, pSX-AP2, or the vector without the AP-2 cDNA, pSX. Luciferase activities were normalized by protein concentration in the cell lysates and are displayed as the ratio between transfections with pSX-AP2 and transfections with pSX. B, recombinant AP-2 binds to the DRE. Gel mobility shift assays using the DRE probe and HepG2 nuclear extracts were performed with or without the addition of recombinant AP-2. Where indicated, a 100-fold excess of the competitor DNA was added. Lanes 4 -7 show the shift in an assay that contains ϳ10 ng of the recombinant AP-2 protein (Promega). Lane 7 contains 2 ng of the anti-AP-2 antibody. Arrows indicate DNA-protein complexes resulting from the addition of the AP-2 protein.
luciferase reporter constructs in which the DRE and PRE mutations were placed in the entire 1.1-kb (Ϫ1069 to ϩ38) TGF-␣ promoter. Mutations in the DRE (T1.1DREm) or PRE (T1.1PREm) that affected promoter function in the context of the 220-bp proximal promoter had essentially the same effect on the 1.1-kb promoter (Fig. 7), whereas a mutation between the PRE and DRE (T1.1(Ϫ104 m)) had no effect on transcription. This result further confirms the importance of these elements in regulating the TGF-␣ gene transcription.

DISCUSSION
The transcription of the TGF-␣ gene is regulated under a variety of physiological and pathological circumstances. TGF-␣ is regulated both temporally and spatially during development, resulting ultimately in the expression of this growth factor in multiple tissues in the adult organism. In most of the tissues examined, TGF-␣ expression appears to be confined to specific cell types. For example, in the bovine pituitary, maximal expression is seen in the lactotrophs (31); in the ovary, the theca cells express TGF-␣ maximally just prior to ovulation (32); in blood vessels, the arterial vascular smooth muscle cells are the predominant site of expression (33); in the brain, neurons in certain brain regions express TGF-␣ (34,35), and in the skin, the basal keratinocytes express this growth factor (21,36). Furthermore, TGF-␣ expression can be regulated in response to metabolic and hormonal signals. In the mammary cells, estrogen (19,37), phorbol esters (38), and EGF (or TGF-␣ itself) (20 -22) can regulate transcription of this gene. In vascular smooth muscle cells, exposure to glucosamine or to superphysiological concentrations of glucose (24) can up-regulate the transcription of this gene. Interestingly, TGF-␣ expression was first observed in tumor cells that had been transformed by retroviruses or spontaneously. The expression in tumors was first detected because these transformed cells express this growth factor at higher levels than the parental cell lines from which they were derived. This observation of TGF-␣ overexpression in tumor cells gave rise to the autocrine hypothesis, which was later confirmed by demonstrating that TGF-␣ overexpression driven by heterologous promoters in cell lines (8 -10) or in transgenic animals (11)(12)(13)(14) results in the development of transformed cells or tumors. That transcriptional activation of the TGF-␣ gene can lead to marked phenotypic changes prompted us to investigate the endogenous TGF-␣ promoter.
The TGF-␣ promoter has many features in common with the promoters of housekeeping genes in that it is GC-rich and does not contain a consensus TATA-box. Like many housekeeping genes, transcription from both the human (18) and rat (29) TGF-␣ promoters is highly dependent upon the transcription factor, Sp1. The human TGF-␣ promoter also contains an initiator element, which is required for optimal transcription and appears to direct transcriptional initiation to a unique site in the TGF-␣ gene (28). More recently, we have defined a p53response element in the human gene that may play a role in the repair of epithelia suffering DNA damage from physical or chemical agents (27). A nonconsensus TATA-box has also been defined in the human TGF-␣ gene whose position and sequence is conserved in the rat gene and that contributes to the transcriptional activity of the proximal promoter (27).
In this paper, we describe another element in the proximal TGF-␣ promoter. This element, termed T␣RE I, was first defined by DNase I footprint analysis of the proximal promoter using nuclear proteins from the human breast cancer cell line, MDA468. The T␣RE I spans about 40 base pairs at a position centered around Ϫ110 relative to the transcription initiation site. Deletion of the T␣RE I reduces basal TGF-␣ promoter activity, while this element activates transcription from a heterologous TK promoter to 11-fold. Thus, T␣RE I behaves like a transcriptional enhancer in the TGF-␣ promoter. Further characterization of the T␣RE I indicated that it actually consists of two elements that are recognized by distinct sequence-specific transcription factors. We have termed these elements the DRE and PRE. The evidence for the two elements consists of the following: (i) point mutagenesis of the T␣RE I defined two regions of this element that contributed to the transcriptional enhancer activity both in the context of the proximal promoter and the 1.1-kb promoter; (ii) the point mutations that reduced enhancer activity reduced the span of the DNase I footprint; (iii) gel mobility shift assays indicated that the PRE and DRE were recognized by proteins having distinct DNA sequence specificities; (iv) the DRE is recognized by AP-2. Although the DRE contained a sequence resembling a binding site for Sp1, an AP-2 but not an Sp1 consensus oligonucleotide competed for binding to this segment of the T␣RE I, and the DRE is recognized by recombinant AP-2 when added to the nuclear extract of HepG2 cells that are deficient in AP-2 activity. Also, the DNA-protein complex formed with the DRE was recognized by an AP-2 antibody by not by a control antibody. Furthermore, the DRE conferred sequence-specific AP-2 responsiveness to a heterologous promoter when cotransfected into HepG2 cells with an AP-2 expression vector. Although we have not identified the protein that binds to the PRE, we do know that each protein can bind to its portion of the T␣RE I in the absence of the other as shown both in the footprinting analysis and gel mobility shift assays. Furthermore, AP-2 and the PRE-binding protein do not interact with each other with sufficient affinity to allow detection by gel mobility shift assays. Nevertheless, the DRE and PRE may cooperate functionally. Point mutations in either the DRE or the PRE reduced the activity of the T␣RE I element as much as deletion of the entire T␣RE I from the promoter. The mechanism by which DRE and PRE may cooperate in regulating the activity of T␣RE I requires further investigation.
Although the AP-2 cDNA was cloned several years ago (39), the role of this transcription factor in the control of gene expression remains unclear. AP-2 is a 52-kDa protein that functions as a dimer, recognizing a GC-rich DNA sequence 5Ј-GCCNNNGGC-3Ј (39). Other similar consensus binding sequences for AP-2 have also been reported. AP-2 binding sites are found in genes of SV40, human metallothionein IIa, murine major histocompatibility complex H-2Kb, collagenase, human growth hormone, human proenkephalin, human keratin K14, FIG. 7. PRE and DRE are important in the context of the entire 1.1-kb promoter. Various 1.1-kb promoter constructs were cloned 5Ј to the luciferase gene in the pGL2 plasmid. 20 g of each plasmid was transfected into MDA468 cells together with 10 g of the ␤-galactosidase expression vector. Luciferase activities were determined 48 h posttransfection and were normalized to the ␤-galactosidase levels in the same cell lysate. Each transfection was repeated three times, and variations of the luciferase activities in percentage between each experiment were Ͻ8%. and mouse mammary tumor virus (39 -45). AP-2 appears to cooperate with other transcription factors to mediate transcriptional activation in response to signals induced by the developmental morphogen retinoic acid (39,46) and to the signals mediated through the activation of cAMP-dependent protein kinase and protein kinase C (41,47,48). In the case of the keratin K14 promoter, AP-2 appears to play a role in the tissue-specific expression of K14 in keratinocytes, but it does so in cooperation with a neighboring element (42). This cooperative behavior might also exist between DRE and PRE of the TGF-␣ promoter. AP-2 expression has been detected in the basal layers of the skin (42,49), and this correlation of AP-2, K14, and TGF-␣ localization in the skin is compatible with a role for AP-2 in the control of the tissue-specific expression of these genes. AP-2 has also been shown to play a role in the function of the oncoprotein, Ras. The expression of AP-2 is stimulated by Ras transformation (50) in PA-1 cells. Interestingly, TGF-␣ expression is also increased in cells transformed by Ras, suggesting a possible role for AP-2 in the Ras induction of TGF-␣ (51). The finding of this element in the TGF-␣ promoter will contribute to our understanding of how the transcription of this growth factor is regulated.