Opposite Smad and chicken ovalbumin upstream promoter transcription factor inputs in the regulation of the collagen VII gene promoter by transforming growth factor-beta.

A critical component of the epidermal basement membrane, collagen type VII, is produced by keratinocytes and fibroblasts, and its production is stimulated by the cytokine transforming growth factor-beta (TGF-beta). The gene, COL7A1, is activated by TGF-beta via Smad transcription factors in cooperation with AP1. Here we report a previously unsuspected level of complexity in this regulatory process. We provide evidence that TGF-beta may activate the COL7A1 promoter by two distinct inputs operating through a common region of the promoter. One input is provided by TGF-beta-induced Smad complexes via two Smad binding elements that function redundantly depending on the cell type. The second input is provided by relieving the COL7A1 promoter from chicken ovalbumin upstream promoter transcription factor (COUP-TF)-mediated transcriptional repression. We identified COUP-TFI and -TFII as factors that bind to the TGF-beta-responsive region of the COL7A1 promoter in an expression library screening. COUP-TFs bind to a site between the two Smad binding elements independently of Smad or AP1 and repress the basal and TGF-beta-stimulated activities of this promoter. We provide evidence that endogenous COUP-TF activity represses the COL7A1 promoter. Furthermore, we show that TGF-beta addition causes a rapid and profound down-regulation of COUP-TF expression in keratinocytes and fibroblasts. The results suggest that TGF-beta signaling may exert tight control over COL7A1 by offsetting the balance between opposing Smad and COUP-TFs.

A critical component of the epidermal basement membrane, collagen type VII, is produced by keratinocytes and fibroblasts, and its production is stimulated by the cytokine transforming growth factor-␤ (TGF-␤). The gene, COL7A1, is activated by TGF-␤ via Smad transcription factors in cooperation with AP1. Here we report a previously unsuspected level of complexity in this regulatory process. We provide evidence that TGF-␤ may activate the COL7A1 promoter by two distinct inputs operating through a common region of the promoter. One input is provided by TGF-␤-induced Smad complexes via two Smad binding elements that function redundantly depending on the cell type. The second input is provided by relieving the COL7A1 promoter from chicken ovalbumin upstream promoter transcription factor (COUP-TF)-mediated transcriptional repression. We identified COUP-TFI and -TFII as factors that bind to the TGF-␤-responsive region of the COL7A1 promoter in an expression library screening. COUP-TFs bind to a site between the two Smad binding elements independently of Smad or AP1 and repress the basal and TGF-␤stimulated activities of this promoter. We provide evidence that endogenous COUP-TF activity represses the COL7A1 promoter. Furthermore, we show that TGF-␤ addition causes a rapid and profound down-regulation of COUP-TF expression in keratinocytes and fibroblasts. The results suggest that TGF-␤ signaling may exert tight control over COL7A1 by offsetting the balance between opposing Smad and COUP-TFs.
Type VII collagen belongs to an extensive family of closely related proteins involved in cell anchoring to extracellular matrix and cartilage formation. Although some collagens have a widespread distribution, type VII collagen is found exclusively in the basement membrane of stratified squamous epithelia (1,2). Its subunit, the type VII collagen ␣-chain (COL7A1), is expressed in both dermal fibroblasts and epidermal keratinocytes (3)(4)(5). COL7A1 forms homotrimers that are assembled into fibrils (6). These fibrils are thought to anchor the epider-mal basement membrane to the underlying dermal extracellular matrix (7). Mutations that cause structural alterations or defective expression of COL7A1 lead to dystrophic epidermolysis bullosa, a group of inherited skin disorders in which blisters form between the basement membrane and the papillary dermis (8).
Transforming growth factor-␤ (TGF-␤) 1 in particular is a potent inducer of COL7A1 expression in fibroblast and keratinocytes (9 -11). A multifunctional cytokine, TGF-␤ critically regulates cell adhesion and extracellular matrix production among other cellular functions (12,13). In addition to activating the production of collagen VII and other types of interstitial collagens, TGF-␤ controls the expression of fibronectin, extracellular matrix proteoglycans, integrin cell adhesion receptors, pericellular proteases, and protease inhibitors. The effects of TGF-␤ on genes encoding the cell adhesion apparatus, along with its effects on cell proliferation and differentiation, exert a profound influence on tissue development and homeostasis.
TGF-␤ activates COL7A1 expression at the transcriptional level (9 -11). Although a general signal transduction pathway for transcriptional regulation by TGF-␤ has been established, little is known about the specific mechanisms involved in the COL7A1 gene response. Activated TGF-␤ receptors directly phosphorylate Smad2 and Smad3, inducing their accumulation in the nucleus to regulate the expression of a large set of genes (14). Receptor-phosphorylated Smads associate with Smad4, which in most instances is indispensable for transcriptional regulation. Smad proteins recognize the sequence CAGAC, commonly referred to as the Smad binding element (SBE). To regulate specific target genes, however, activated Smad complexes must additionally interact or functionally cooperate with other transcription factors. In the case of COL7A1 activation by TGF-␤, previous studies have shown the involvement of Smad and AP1 transcription factors (9 -11). The TGF-␤-responsive region in the COL7A1 contains a canonical SBE. Mutation of this element inhibits the activation of the COL7A1 promoter by TGF-␤ in fibroblasts (9). Paradoxically, this SBE is not required for this response in keratinocytes (11) suggesting that the cellular context and other complexities play an important role in the regulation of COL7A1 expression by TGF-␤.
To address these questions, we investigated the role of various elements present in the TGF-␤-responsive region of the COL7A1 promoter in keratinocytes and fibroblasts. We report the existence of a second Smad binding site in the COL7A1 promoter that provides cell type-dependent redundancy, clarifying previous controversies. Our results also suggest that COL7A1 activation in response to TGF-␤ involves not only activation by a Smad complex but also the relief of inhibition by COUP-TF transcriptional repressors.
Plasmids-The human COL7A1 promoter fragment spanning from Ϫ496 to ϩ92 was generated by PCR, using human DNA as template and oligonucleotides primers with flanking SalI/HindIII sites. This fragment was cloned into the XhoI/HindIII sites of the low basal activity luciferase reporter plasmid pGL2-basic (Promega). Mutant forms of the wild type promoter were obtained by site-directed mutagenesis using oligonucleotides carrying the indicated mutations. FLAG-tagged versions of the COUP-TFI and -TFII cDNAs were generated and cloned into the pCMV5 vector. FLAG-tagged COUP-Dim was generated by PCR, amplifying the C-terminal region of COUP-TFII (residues 300 -414) comprising the dimerization domain, and was cloned into the pCMV5 vector. COS1 cells were transiently transfected with FLAGtagged COUP-TFI or -TFII or COUP-Dim using LipofectAMINE (Invitrogen) according to the manufacturer's instructions.
Transcriptional Assays-HaCaT cells were transfected by using the DEAE-dextran method as described previously (15). NIH 3T3 cells were transfected using LipofectAMINE or the calcium-phosphate precipitation method. Briefly, 3 g of plasmid DNA were diluted in 250 l of 2 M CaCl 2 . This solution was mixed with 250 l of 2ϫ HEPES-buffered saline (50 mM HEPES, 280 mM NaCl, 1.5 mM Na 2 HPO 4 , to final pH 7.1). 80-l aliquots of this precipitate were overlaid on wells of a 12-well dish containing 1 ml of freshly added Dulbecco's modified Eagle's medium plus 10% FBS. After transfection cells were incubated in medium containing 10% FBS for 6 -8 h. The medium was changed to 0.2% FBS, and cells were incubated with 100 pM TGF-␤ for 20 -24 h. Cell lysates were then subjected to luciferase assays (Promega) in a Berthold luminometer (Nashua, NH). A cytomegalovirus promoter Renilla luciferase plasmid (Promega) was used as a control to normalize the transfection efficiency and was assayed as described previously (16).
Oligonucleotide Precipitation Assays-Cells were treated with TGF-␤ for 1 h under normal culture conditions and then lysed by sonication in buffer (10 mM Hepes, pH 7.9, 100 mM KCl, 10% glycerol, 1 mM dithiothreitol, 0.5% Nonidet P-40) with phosphatase and protease inhibitors. Cell debris was removed by 5-min centrifugation at 10,000 ϫ g at 4°C. Cell extracts were incubated for 16 h with 1 g of biotinylated doublestrand oligonucleotide corresponding to the wild type or mutant forms of the COL7A1 Ϫ495/Ϫ431 promoter region. DNA-bound proteins were collected with streptavidin-agarose beads for 1 h, washed with lysis buffer, separated on a SDS-polyacrylamide gel, and identified by Western blotting.
RNA Assays-Exponentially growing cells were incubated with 100 pM TGF-␤ for the indicated time. Cells were harvested, and total RNA was extracted by using Qiagen (Chatsworth, CA) RNeasy minikit. 100 mg of total RNA was then used to obtain poly(A) RNA using a Clontech kit. The poly(A) RNA obtained was run on parallel denaturing gel and subjected to Northern analysis. Blots were probed with probes corresponding to human or mouse COL7A1, actin, COUP-TFI, or COUP-TFII.
Yeast One-hybrid Screening-A NIH 3T3 cDNA library in the pGAD10 fusion vector (Clontech), which provides an N-terminal GAL4 fusion transcriptional activation domain, was transformed into a yeast strain bearing four consecutive copies of the Ϫ495/Ϫ431 region of the COL7A1 promoter upstream of both HIS3 and a LacZ reporter gene. cDNA clones that allowed growth in ϪHis plates and gave a strong ␤-galactosidase activity in a colony lift assay were identified as positive.

Role of Two Smad Binding Elements in the COL7A1
Promoter-To investigate the transcriptional activation of COL7A1 by TGF-␤, we generated luciferase reporter constructs driven by the Ϫ496/ϩ92 region of the human COL7A1 (9) (Fig. 1A). Versions of this promoter were generated containing mutations that target the previously described SBE site (9), which we refer to as 5Ј-SBE, and an adjacent AP1 site (10) as well as other sites in this region that are conserved in the human and mouse genes (Fig. 1A). TGF-␤ addition strongly stimulated the expression of the wild type COL7A1 promoter in mouse NIH 3T3 fibroblasts, and this response was diminished by mutations targeting the 5Ј-SBE (Fig. 1B). These results are in full agreement with those reported in human dermal fibroblasts (9). TGF-␤ also stimulated the expression of the wild type COL7A1 promoter in HaCaT human skin keratinocytes. Surprisingly, however, this response was not diminished by mutations targeting the 5Ј-SBE (Fig. 1B).
Several lines of evidence suggested that this tolerance for a mutant 5Ј-SBE in HaCaT keratinocytes reflected a genuine difference between fibroblasts and keratinocytes and not an anomaly of the HaCaT cell line. HaCaT cells are well charac- The data are averages of triplicate assays and are expressed as the -fold induction relative to cells that did not receive TGF-␤. WT, wild type; mt, mutant. C, COL7A1 mRNA levels in HaCaT cells that were incubated with 100 pM TGF-␤ for the indicated times. Poly(A) RNA was isolated and subjected to Northern analysis. Blots were probed with human COL7A1 and actin probes. The COL7A1 signal was normalized against the actin signal, and the values are expressed as -fold increase relative to the controls not treated with TGF-␤.
terized in terms of their responsiveness to TGF-␤ (17). Further, we verified that TGF-␤ stimulates the expression of the endogenous COL7A1 gene at the mRNA level (Fig. 1C). Moreover, the lack of an effect of mutations in this SBE has been reported recently in mouse keratinocytes as well (11).
Searching for a possible basis for this difference between keratinocytes and fibroblasts, we noticed a perfect inverted SBE (3Ј-SBE) ϳ50 base pairs downstream of the 5Ј-SBE (Fig.  1A). This site was not investigated in previous studies (9 -11). We tested the effect of mutating this site alone or in combination with the 5Ј-SBE. Mutation of the 3Ј-SBE affected the response of the COL7A1 promoter to TGF-␤ in a manner similar to mutation of the 5Ј-SBE site, as the 3Ј-SBE mutations diminished the TGF-␤ response in fibroblasts but not in keratinocytes (Fig. 1B). Remarkably, the simultaneous mutation of the 5Ј-SBE and 3Ј-SBE sites completely eliminated the TGF-␤ response not only in fibroblasts but also in keratinocytes (Fig.  1B). These results suggested that the 5Ј-SBE and 3Ј-SBE in the COL7A1 promoter are important for the TGF-␤ response in both cell types, acting redundantly in keratinocytes but not in fibroblasts.
Cell Type-dependent Cooperation with AP1-AP1 sites and the Fos-Jun complexes that recognize these sites have been shown to play a role in certain TGF-␤ gene responses, including the COL7A1 response (10). An AP1 site that is adjacent to the 5Ј-SBE (see Fig. 1A) has been implicated in the TGF-␤ response of the COL7A1 promoter in fibroblasts (10). We found that the requirement of this site for the TGF-␤ response is cell type-dependent. Mutation of the AP1 site had no effect on the TGF-␤ response of the COL7A1 promoter in the keratinocytes, but it inhibited this response in fibroblasts (Fig. 1B).
To investigate the interaction of endogenous Smad and AP1 proteins with these sites in the COL7A1 promoter, we conducted DNA-mediated precipitation assays using biotinylated double-stranded oligonucleotides corresponding to the human COL7A1 Ϫ495/Ϫ431 promoter region. Extracts from control and TGF-␤-treated cells were incubated with wild type or mutant probes, and bound complexes were collected using streptavidin-agarose beads. Proteins of interest were investigated by Western immunoblotting of these precipitates using specific antibodies. Both HaCaT and NIH 3T3 cells showed a TGF-␤dependent formation of a Smad complex capable of binding to the wild type COL7A1 probe, as determined using anti-Smad4 antibody (Fig. 2). Binding of this complex to the probe was not affected by mutations that disrupt the AP1 site. Mutation of the 5Ј-SBE or the 3Ј-SBE diminished the TGF-␤-dependent binding of Smad4, and a combination of these mutations essen-tially eliminated Smad4 binding to the probe (Fig. 2).
To investigate the binding of AP1 complexes to these probes, we used antibodies that cross-react with various members of the Jun family. The wild type COL7A1 probe bound endogenous Jun proteins, and the level of binding was similar in control cells and TGF-␤-treated cells (Fig. 2). Mutation of either or both SBE sites did not affect the binding of Jun proteins to the probe, whereas mutation of the AP1 site completely prevented Jun binding (Fig. 2). Collectively these results suggest that Smad and AP1 complexes may bind to the SBE and AP1 sites in the COL7A1 promoter independently of each other, and this binding is dependent on TGF-␤ stimulation in the case of the Smad site but independent in the case of the AP1 site. Furthermore, this protein binding behavior was the same regardless of whether the SBE sites were functionally redundant in keratinocytes or were non-redundant and cooperative with AP1 in fibroblasts.
COUP-TF Binding to the TGF-␤ Regulatory Region of the COL7A1 Promoter-To determine whether additional factors may control the COL7A1 response to TGF-␤, we tested the role of other segments in this region that are conserved in the human and mouse genes. We generated a reporter construct containing mutations in a conserved segment around position Ϫ450 (Ϫ450 conserved box; Fig. 1A). This construct had much higher basal activity than the wild type construct in transcriptional assays in keratinocytes (Fig. 3A). The activity of this construct was further increased by TGF-␤, suggesting that a repressor factor may bind to this region under basal conditions and may determine the overall responsiveness of the COL7A1 promoter to TGF-␤.
To identify factors that may regulate COL7A1 from the Ϫ495/Ϫ431 promoter region, we screened a cDNA expression library for gene products that bind to this region. To this end, we generated a yeast strain expressing HIS3 and LacZ under the control of four copies of this promoter region and then used these cells to screen a mouse fibroblast cDNA library fused to the GAL4 transcriptional activation domain. In this approach, cDNAs encoding GAL4 fusion proteins that bind to the Ϫ495/ Ϫ431 promoter region would confer HIS3 phenotype and activate LacZ expression. Only four cDNA clones were isolated that fulfilled these criteria. One of these cDNAs encoded the fulllength COUP-TFI (also known as EAR3), and the other three encoded the full-length COUP-TFII (also known as ARP-1) (Fig. 3B).
COUP-TFI and COUP-TFII are closely related orphan members of the nuclear/steroid receptor family (18,19). They consist of an N-terminal DNA binding domain containing two zinc finger motifs and a C-terminal transcription regulatory domain containing a dimerization region (Fig. 3B). COUP-TFI and COUP-TFII bind to DNA as dimers that recognize two direct TGACC(C/T) repeats separated by a single base pair spacer (20). COUP-TFI and -TFII are thought to act primarily as transcriptional repressors (20 -25).
Examination of the COL7A1 Ϫ495/Ϫ431 region revealed that the sequence of the Ϫ450 conserved box is similar to the COUP-TF consensus binding sequence (Fig. 3C). As mutations of the Ϫ450 conserved box augmented the activity of the COL7A1 promoter, we hypothesized that this effect might be due to a loss of endogenous COUP-TF binding to this region. To determine whether this region can bind COUP-TF proteins, we carried out DNA precipitation assays using wild type and mutant biotinylated COL7A1 Ϫ495/Ϫ431 oligonucleotide probes (Fig. 3C). A FLAG-tagged COUP-TFII construct expressed in COS cells bound to the wild type probe but not to two different probes that contain mutations in the Ϫ450 conserved box (Fig.  3D). COUP-TFII binding to the wild type promoter was not

COL7A1 Regulation by TGF-␤
affected by cell treatment with TGF-␤ and was not disrupted by SBE or AP1 site mutations that block the binding of Smad or Jun, respectively (Fig. 3E). We looked for, but could not find, evidence of Smad3 or Smad4 binding to COUP-TFI or COUP-TFII. In extracts from cells overexpressing Smads and COUP-TFs, these proteins neither enhanced nor interfered with the binding of each other to the COL7A1 Ϫ495/Ϫ431 probe (data not shown). These results suggest that the Ϫ450 conserved box is a COUP-TF site that binds COUP-TF proteins independently of Smad and AP1. (17). We verified by Northern analysis that TGF-␤ addition causes a rapid and profound decrease in the level of COUP-TFII mRNA in HaCaT and NIH 3T3 cells (Fig. 4). COUP-TFI expression, which was present in NIH 3T3 cells but barely detectable in HaCaT cells, was also inhibited by TGF-␤ (Fig. 4). The down-regulation of COUP-TFII by TGF-␤ in HaCaT cells was rapid (t1 ⁄2 Յ 2 h; Fig. 4) and preceded the up-regulation of COL7A1 expression (t1 ⁄2 ϭ 4 h; refer to Fig. 1C). Thus, TGF-␤ action down-regulates the expression of a putative COL7A1 repressor.

Down-regulation of COUP-TF Expression by TGF-␤-Interestingly, GeneChip transcriptomic profiling data in HaCaT cells revealed that TGF-␤ treatment decreases the level of COUP-TFII transcripts
Repression of COL7A1 Promoter by COUP-TF-We investigated whether COUP-TFs can act as repressors of the COL7A1 promoter in transfected HaCaT cells. Indeed, both COUP-TFI and COUP-TFII markedly inhibited the basal activity of the wild type COL7A1 promoter as well as its activation by TGF-␤ (Fig. 5A). The promoter construct containing mutations in the Ϫ450 conserved box (the COUP-TF binding element) was not only hyperactive under basal conditions but also completely resistant to inhibition by exogenous COUP-TFI or COUP-TFII (Fig. 5A).
As no anti-COUP-TF antibodies could be obtained for these studies, we resorted to alternative approaches to determine whether the COL7A1 promoter is sensitive to endogenous COUP-TFs. First, we mutated the COUP-TF binding region to bring this sequence closer to the optimal COUP-TF binding sequence (COUPϩ mutant in Fig. 3C) (20). When tested in oligonucleotide precipitation assays, a COL7A1 probe containing this mutant sequence bound COUP-TFII with higher affinity than did the wild type probe (Fig. 5B). A COL7A1 promoter construct bearing this mutation showed a significantly decreased activity in HaCaT cells, which was suggestive of a response to endogenous COUP-TFs (Fig. 5C). Thus, the COUPϩ mutation hypersensitizes the COL7A1 promoter to COUP-TF repression.
To test further this hypothesis, we generated a FLAG-tagged mutant COUP-TF construct designed to inhibit the dimerization and thus the function of endogenous COUP-TFs. This construct, COUP-Dim, encodes the dimerization domain of COUP-TFII (Fig. 6A). The COUP-Dim product was able to completely inhibit the binding of COUP-TFII to the COL7A1  1A). Cells were untreated or treated with TGF-␤ for 20 h before luciferase activities were determined. B, the four positive clones obtained by screening a mouse cDNA expression library fused to a transcriptional activation domain for clones encoding COL7A1 Ϫ495/Ϫ431 binding factors. All four clones encoded the full-length COUP-TFI or COUP-TFII and were not fused to the GAL4-activating domain, thus acting as activators on their own. Protein domains of interest and amino acid sequence length are schematically indicated. nt, nucleotide. C, the Ϫ450 conserved box in the human and mouse COL7A1 promoters contains an imperfect COUP-TF binding site. Shown are also the mutations (mt) introduced in the human COL7A1 promoter that either disrupt the COUP-TF site (mt-COUP1 and -2) or bring the sequence of this site closer to the consensus COUP-TFII binding sequence (COUPϩ). The mouse sequence as well as the consensus COUP-TFII binding site is shown. D, COS1 cells were transfected with a vector encoding FLAG-tagged COUP-TFII. Cell lysates were precipitated with biotinylated probes corresponding to the wild type or the indicated mutant versions of the COL7A1 Ϫ495/Ϫ431 promoter region. Protein-DNA complexes were subjected to Western immunoblotting with anti-FLAG antibodies. E, cells were transfected with mouse FLAG-tagged COUP-TFII and then treated with 100 pM TGF-␤ for 1 h or not treated. Cell lysates were precipitated with biotinylated probes corresponding to the wild type or the indicated mutant versions of the COL7A1 Ϫ495/Ϫ431 promoter region. Protein-DNA complexes were subjected to Western immunoblotting using anti-FLAG, anti-Smad4, or anti-pan-Jun antibodies. promoter (Fig. 6A), demonstrating that it can act as a dominant-negative construct. When tested in COL7A1 promoter activity assays, COUP-Dim increased the activity of this promoter while still allowing a further activation by TGF-␤ (Fig.  6C). Taken together, these results suggest that the COL7A1 promoter is repressed by COUP-TF or a closely related factor in HaCaT cells. The ability of TGF-␤ to sharply decrease the expression of endogenous COUP-TFs in these cells while simultaneously inducing the formation of a COL7A1-activating Smad complex suggests that COL7A1 induction by TGF-␤ involves a combination of promoter activation and deinhibition inputs.

DISCUSSION
The results of this work suggest that the activation of COL7A1 expression by TGF-␤ is a complex process involving two inputs (schematically summarized in Fig. 7). One input is provided by TGF-␤-dependent Smad-transactivating factors and the other by the relief of COUP-TF-mediated transcriptional repression of the COL7A1 promoter. This complex regulatory process has several features that distinguish it from previously characterized TGF-␤-regulated promoters.
We provide evidence that regulation of the COL7A1 promoter by TGF-␤-activated Smad factors occurs via two SBEs in the TGF-␤-responsive region of this promoter. The SBE, or CAGAC sequence, is the best characterized of the various sequences implicated in DNA binding by signal-activated Smad factors (14,26,27). In many TGF-␤ target promoters characterized to date, the responsive region contains only one SBE. The COL7A1 promoter was previously thought to be in this class, with only one identified functional SBE (referred to here as the 5Ј-SBE) (9 -11). Our evidence, however, indicates that a second SBE, located ϳ50 bp downstream of the 5Ј-SBE in the human and mouse promoters, is also involved. In promoter construct assays at least, these two SBEs are functionally equivalent and act redundantly in keratinocytes but not in fibroblasts. This partial redundancy explains the previously encountered paradox that COL7A1 promoter constructs devoid of the only identified SBE lack TGF-␤ responsiveness in fibroblasts (9) but remain responsive in keratinocytes (11).
The results of our oligonucleotide binding experiments show that an endogenous, TGF-␤-dependent Smad complex can specifically bind to the TGF-␤-responsive region of the COL7A1 promoter. This binding requires the integrity of both SBEs but not the integrity of other conserved elements in this region, including AP1 and COUP-TF sites, whose mutation prevents the binding of their cognate factors. While we cannot exclude the possibility that the mutations in the AP1 and COUP-TF sites may still allow other, as yet unknown factors to cooperate with Smads in binding to this promoter, this possibility seems unlikely given the density of our mutational analysis. This is of interest because the affinity of an isolated Smad-SBE interaction is too low to support effective binding and gene regulation in vivo on its own (27). In promoters that contain one single SBE, receptor-phosphorylated Smads must associate with Smad4 and at least one other DNA binding factor to achieve high affinity binding. Examples include the activation of Xenopus Mix2 by a Smad2-FoxH1 complex (28) and goosecoid by a Smad2-Mixer complex (29) in response to nodal/activin signals and the activation of an IgA promoter by a Smad3-Runx2 complex in response to TGF-␤ (30 -32). This also applies to promoters regulated by bone morphogenetic proteins via Smad1, as in the activation of Xenopus Vent2 by a Smad1-OAZ complex (34,35) and mouse Tlx2 by a Smad1-Ecsit complex (36) and the repression of c-MYC by a Smad3-E2F4/5 complex in response to TGF-␤ in human epithelial cells (33), with Smad4 included in these complexes in all cases.
Although the requirement of a DNA binding cofactor for Smad binding to target promoters containing one single SBE is well established, recent work has suggested that another class of Smad target promoters may achieve high affinity binding by cooperative interactions of multiple SBEs with the multiple DNA binding domains present in the phosphoSmad-Smad4 heteromer. Previous examples of this paradigm include Smad7 (37,38) and Id1 (39,40), which are common targets of TGF-␤ and BMP, and CDKN1A (p21Cip1), which is a target of TGF-␤ (41). A ligand-activated Smad complex may bind to these promoters independently of other cofactors, although it may have to recruit additional transcription factors for regulation of the gene. Thus, a TGF-␤-activated Smad3-Smad4 complex recruits FoxO to the CDKN1A promoter for activation (41) and recruits ATF3 to the Id1 promoter for repression (17).
The present results suggest that COL7A1 belongs to this class of TGF-␤ target genes. COL7A1 contains two functional FIG. 5. COUP-TFI and -TFII repress the COL7A1 promoter. A, NIH 3T3 cells were transfected with increasing amounts of COUP-TFI or COUP-TFII expression plasmids together with COL7A1 Ϫ490/ϩ92 reporter constructs containing the wild type (WT) sequence or COUP-TFII binding site mutations (mtCOUP1) (refer to Fig. 3C). Cells were untreated or treated with TGF-␤ for 20 h before luciferase activity was determined. B, lysates from COS1 cells expressing FLAG-tagged COUP-TFII were precipitated with biotinylated probes corresponding to the wild type or the COUPϩ mutant versions of the COL7A1 Ϫ495/ Ϫ431 promoter region. Protein-DNA complexes were subjected to Western immunoblotting with anti-FLAG antibodies. C, NIH 3T3 cells were transfected with increasing amounts of COUP-TFII expression plasmid together with COL7A1 Ϫ490/ϩ92 reporter constructs containing the wild type sequence or the COUPϩ mutations. Cells were untreated or treated with TGF-␤ for 20 h, and luciferase activity was determined. SBEs that are recognized by a Smad complex apparently without a strict dependence on additional factors, and it contains an AP1 site that in fibroblasts at least is required for strong TGF-␤-dependent transactivation (10). COL1A2 (42), PAI1 (43,44), and c-Jun (45) have also been found to respond to TGF-␤ through a cooperation between SBE and AP1 sites. Although Smad and AP1 factors may interact in the cell and have been proposed to bind to the PAI1 promoter as a complex (43), our results agree with reports that Smad and AP1 bind to their respective elements independently of each other (10,45,46).
A second set of findings in the present studies led to evidence that TGF-␤ down-regulates the expression of COUP-TF, and this effect may contribute to relieving COL7A1 from COUP-TF-mediated repression in keratinocytes and fibroblasts. The finding that COUP-TFs can negatively regulate the COL7A1 promoter was unexpected. Interestingly, the only four positive clones identified in our cDNA expression library screening in yeast using the TGF-␤-responsive region of COL7A1 correspond to the full-length COUP-TFI or -TFII. This is despite the fact that the COUP-TF site on the COL7A1 promoter is imperfect and has lower affinity for COUP-TF factors than does the perfect consensus site. COUP-TF sites consist of two direct repeats and bind COUP-TF proteins as homodimers (47). Thus, unlike Smad and AP1 proteins, which normally bind DNA as heteromers, COUP-TF can bind to the COL7A1 promoter region independently of other factors. This may explain the isolation of COUP-TF but not Smads, Jun, or Fos in our expression cloning screening.
COUP-TF proteins are orphan members of the nuclear hor-mone receptor family that negatively regulate the activation function of vitamin D, thyroid hormone, retinoic acid, the retinoid X, and the peroxisome proliferator-activated receptors (19). COUP-TF genes are highly conserved from Drosophila through human and are essential for development and differentiation during embryogenesis (19). COUP-TFs are known to act as transcriptional repressors through interactions with corepressors NCoR (nuclear receptor corepressor) and SMRT (silencing mediator of retinoid and thyroid hormone receptors) (21, 23) as well as CTIP1 and CTIP2 (COUP-TF-interacting protein 1 and 2) (48,49). However, the absence of known COUP-TF ligands and regulators has impaired the ability to delineate fully the function of these factors. We provide several lines of evidence suggesting that COUP-TFs may function as repressors of the COL7A1 promoter. First, COUP-TFI and -TFII strongly decrease the activity of the COL7A1 promoter in transcriptional assays. Second, mutations that disrupt the COUP-TF site increase the basal activity of the COL7A1 promoter, whereas mutations that increase the affinity of this site decrease the activity of the promoter. Third, expression of a dominant-negative construct that inhibits COUP-TF binding to the COL7A1 promoter strongly increased the basal activity of this promoter and its stimulation by TGF-␤ in keratinocytes.
We found that TGF-␤ addition to keratinocytes or fibroblasts causes a rapid and profound decrease in the expression of COUP-TFI and -TFII. This decrease preceded the increase in COL7A1 mRNA levels. These observations suggest that in addition to inducing the formation of a COL7A1-activating Smad complex, TGF-␤ action may facilitate the activation of this gene by relieving it from COUP-TF-mediated repression. This dual input is similar to the recently delineated mechanism of activation of CDKN1A (p21Cip1) and CDKN2B (p15Ink4a) by TGF-␤ in epithelial cells. TGF-␤ elevates the expression of these two cyclin-dependent kinase inhibitors through the induction of activator Smad complexes and the relief of c-MYCmediated repression (41,50,51). As in the case of COUP-TF, c-MYC expression is rapidly down-regulated by TGF-␤, allowing the depletion of c-MYC from the CDKN1A and CDKN2B promoters. A dual activation switch on the COL7A1 promoter would likewise afford a tight control over its expression and control by TGF-␤, perhaps reflecting the critical importance of COL7A1 regulation in epidermal homeostasis.
Acknowledgments-We thank Dr. M. J. Tsai for COUP-TF vectors and Mirlene Lindor and Laurie-ann Swaby for technical assistance.  6. A dominant-negative COUP-TF construct activates the COL7A1 promoter. A, a FLAG-tagged construct spanning the C-terminal residues (300 -414) of COUP-TFII including its dimerization domain (FLAG-COUP-Dim) was generated. COS1 cells were transfected with FLAG-tagged COUP-TFII and increasing amounts of FLAG-COUP-Dim. Cell lysates were precipitated with biotinylated COL7A1 wild type, and protein-DNA complexes were subjected to Western blotting and probed with anti-FLAG antibody. Western blotting was performed with whole extracts to verify the expression of the two constructs. B, NIH 3T3 cells were transfected with increasing amounts of FLAG-COUP-Dim expression plasmid together with a COL7A1 Ϫ490/ϩ92 promoter reporter construct. Cells were untreated or treated with TGF-␤ for 20 h, and luciferase activity was determined.
FIG. 7. TGF-␤ inputs into the COL7A1 promoter. TGF-␤ is proposed to activate the COL7A1 promoter via two inputs. TGF-␤ activates a Smad complex that binds to the COL7A1 promoter and activates transcription cooperatively with a TGF-␤-independent AP1 complex. In parallel, TGF-␤ action facilitates the activation of this promoter by relieving it from COUP-TF-mediated repression. TGF-␤ rapidly and profoundly down-regulates the expression of COUP-TFI and -TFII, which act as COL7A1 promoter repressors. The involvement of Smads in the COUP-TF down-regulation response remains to be determined.