An AP-1 binding sequence is essential for regulation of the human α2 (I) collagen (COL1A2) promoter activity by transforming growth factor-β

Previous studies have shown that transforming growth factor-β (TGF-β) and tumor necrosis factor-α (TNF-α) modulate type I collagen gene expression in fibroblasts. To fine-map the corresponding response elements in the human α2(I) collagen (COL1A2) promoter, we have generated a series of 5′ deletion promoter/chloramphenicol acetyltransferase (CAT) reporter gene constructs. Transient cell transfection assays using human dermal fibroblasts and stable transfection experiments using NIH 3T3 fibroblasts identified the region located between residues −265 and −241, as critical for TGF-β response. Specifically, we demonstrate that this 25-base pair region mediates the up-regulatory effect of TGF-β on COL1A2 promoter activity and allows antagonistic activity of TNF-α on the TGF-β effect. Gel mobility shift assays indicate that nuclear factor binding to this 25-base pair region of COL1A2 promoter is competed by AP-1, but not NF-1 or NF-κB, oligonucleotides. Transient cell transfection experiments with plasmid constructs in which the potential AP-1-binding site located within this short region of promoter was modified by sitedirected mutagenesis indicated that this element plays a significant role in the basal activity of the promoter. Furthermore, this sequence is essential for TGF-β response and does not require the presence of the three Sp-1-binding sites located further upstream, between nucleotides −273 and −304. In addition, overexpression of c-jun in co-transfection experiments with COL1A2 promoter/CAT constructs blocks the TGF-β response, further implicating AP-1 in the regulation of COL1A2 gene expression. Our results clarify the molecular mechanisms involved in the regulation of type I collagen gene expression and further emphasize the importance of AP-1 in mediating some of the TGF-β effects on gene transcription.

Previous studies have shown that transforming growth factor-␤ (TGF-␤) and tumor necrosis factor-␣ (TNF-␣) modulate type I collagen gene expression in fibroblasts. To fine-map the corresponding response elements in the human ␣2(I) collagen (COL1A2) promoter, we have generated a series of 5 deletion promoter/ chloramphenicol acetyltransferase (CAT) reporter gene constructs. Transient cell transfection assays using human dermal fibroblasts and stable transfection experiments using NIH 3T3 fibroblasts identified the region located between residues ؊265 and ؊241, as critical for TGF-␤ response. Specifically, we demonstrate that this 25-base pair region mediates the up-regulatory effect of TGF-␤ on COL1A2 promoter activity and allows antagonistic activity of TNF-␣ on the TGF-␤ effect. Gel mobility shift assays indicate that nuclear factor binding to this 25-base pair region of COL1A2 promoter is competed by AP-1, but not NF-1 or NF-B, oligonucleotides. Transient cell transfection experiments with plasmid constructs in which the potential AP-1-binding site located within this short region of promoter was modified by sitedirected mutagenesis indicated that this element plays a significant role in the basal activity of the promoter. Furthermore, this sequence is essential for TGF-␤ response and does not require the presence of the three Sp-1-binding sites located further upstream, between nucleotides ؊273 and ؊304. In addition, overexpression of c-jun in co-transfection experiments with COL1A2 promoter/CAT constructs blocks the TGF-␤ response, further implicating AP-1 in the regulation of COL1A2 gene expression. Our results clarify the molecular mechanisms involved in the regulation of type I collagen gene expression and further emphasize the importance of AP-1 in mediating some of the TGF-␤ effects on gene transcription.
Recently, significant progress has been made in understanding the expression of the human ␣2(I) collagen (COL1A2) gene and its transcriptional regulation by cytokines and growth factors. In particular, it has been shown that a GC-rich region located between residues Ϫ303 and Ϫ271, containing Sp-1-binding sites, is important for high basal promoter activity (Tamaki et al., 1995). This region is comprised within a larger segment of the COL1A2 promoter which has been shown to confer both TGF-␤ 1 (Inagaki et al., 1994) and TNF-␣ responsiveness (Inagaki et al., 1995). However, despite extensive analyses, these studies did not allow precise characterization of TGF-␤ or TNF-␣-response element(s) within the COL1A2 promoter. It was suggested that the TGF-␤-responsive element (TbRE) is located within a 131-bp region, between nucleotides Ϫ378 and Ϫ255, and consists of at least two cis-elements which act in a concerted manner to mediate the effect of TGF-␤. Once inserted upstream of the thymidine kinase promoter, the TbRE confers TGF-␤ inducibility to this heterologous promoter. Two protein binding sequences within the TbRE, box 3A between residues Ϫ313 and Ϫ286 which contains Sp-1 binding sites, and box B between residues Ϫ271 and Ϫ255, were shown to interact to confer both nuclear protein binding and promoter inducibility, otherwise not observed with either box alone. In addition, these authors suggested that TNF-␣ inhibitory effect requires the contribution of both the Sp-1 binding sequence of the TbRE and an inhibitory element, box 5A, immediately upstream of the TbRE, but excluded, using an antibody interference experiment, participation of both AP-1 and NF-B transcription factors in this phenomenon (Inagaki et al., 1995).
To characterize the TGF-␤ and TNF-␣ response elements within the human COL1A2 promoter in further detail, our experimental approach consisted of (a) development of a repertoire of 5Ј deletion constructs of the COL1A2 promoter and (b) site-directed mutagenesis of specific sequences characterized as essential for growth factor response. This approach allowed us to map the growth factor response elements, in the presence of homologous downstream sequences, reaching the position ϩ58 of the COL1A2 gene. Specifically, we have narrowed the TGF-␤ response element(s) to a 25-bp segment of the promoter, located between residues Ϫ265 and Ϫ241. In addition, we show that this fragment is sufficient to allow inhibition of the promoter activity by TNF-␣. Furthermore, using sitedirected mutagenesis, we have established that the potential AP-1-binding site, CGAGTCA, located within this short region of promoter, is essential for TGF-␤ response.

MATERIALS AND METHODS
Cell Cultures-Human dermal fibroblast cultures, established by explanting tissue specimens obtained from neonatal foreskins, were utilized in passages 3-8. The cell cultures were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum, 2 mM glutamine, 100 units/ml of penicillin, and 50 g/ml of streptomycin.
Cytokines/Growth Factors-Human recombinant TGF-␤ 2 was a generous gift from Dr. David R. Olsen, Celtrix Laboratories, Santa Clara, CA. Human recombinant TNF-␣ was purchased from Boehringer Mannheim.
Plasmid Constructs-To study the transcriptional regulation of COL1A2 gene expression, transient transfection experiments were performed with various 5Ј deletion constructs derived from pMS3.5CAT, kindly provided by Dr. Francesco Ramirez, Mt. Sinai School of Medicine, New York, a plasmid containing ϳ3.5 kilobases of human COL1A2 promoter linked to the CAT reporter gene in the expression vector p8-CAT, a derivative of pEMBL plasmid (Boast et al., 1990). The promoter fragments were generated by polymerase chain reaction (PCR) and cloned as BamHI/SmaI fragments into similarly digested empty expression vector. All deletion constructs were sequenced by automated sequencing (ABI) to verify the accuracy of the PCR. pRSVc-jun (Chiu et al., 1989) was used to overexpress c-Jun in co-transfection experiments with COL1A2 promoter/CAT constructs. Empty pRSVe was used as a control.
Site-directed Mutagenesis of the Putative AP-1-binding Site-Two point mutations were introduced into the putative AP-1-binding site using asymmetric PCR and a single mutant primer (Perrin and Gilliland, 1990). Specifically, the first PCR was prepared using a mutagenic primer containing two point mutations (bold) in the putative AP-1binding site (CGAGTCA 3 CCAGTGA) and a flanking primer closer to the transcription initiation site. The PCR product was gel-purified on a 2% agarose gel and used as a megaprimer for the second PCR with another flanking primer in the opposite direction and the Ϫ342/CAT and Ϫ265/CAT deletion constructs as templates. The promoter sequences were excised from the Ϫ342/CAT and Ϫ265/CAT constructs by digestion with BamHI and XmaI restriction enzymes and the final PCR products, digested with the same restriction enzymes, were ligated together, generating Ϫ342/CAT and Ϫ265/CAT constructs containing a mutated AP-1-binding site. Fidelity of the new plasmid constructs was checked by automated sequencing as described above.
Transient Transfections and CAT Assays-Transient transfections of human foreskin fibroblasts were performed by the calcium phosphate/ DNA co-precipitation method, as described previously . Briefly, the cells were transfected with 10 or 20 g of DNA mixed with 5 g of the pRSV-␤-galactosidase plasmid DNA in order to monitor transfection efficiencies. After glycerol shock, the cells were placed in DMEM containing 1% fetal calf serum 4 h prior to the addition of growth factors and cytokines. In experiments without growth factors and cytokines, the cells were placed in DMEM containing 10% fetal calf serum. After an additional 40 h of incubation, the cells were rinsed twice with phosphate-buffered saline, harvested by scraping, and lysed in reporter lysis buffer (Promega, Madison, WI). The ␤-galactosidase activities were measured according to standard protocols (Sambrook et al., 1989). Aliquots corresponding to identical ␤-galactosidase activity were used for each CAT assay with [ 14 C]chloramphenicol as substrate (Gorman et al., 1982) using thin layer chromatography. Following autoradiography, the plates were cut and counted by liquid scintillation to quantify the acetylated [ 14 C]chloramphenicol.
Stable Transfections-To investigate the expression of promoter constructs with low activity, NIH 3T3 fibroblast cultures were stably transfected with various deletion constructs of the COL1A2 promoter (Ϫ342, Ϫ285, Ϫ-265, and Ϫ241/CAT). For this purpose, each construct was co-transfected with pRC/CMV (InVitrogen, Portland, OR) in a 10 to 1 ratio by the calcium phosphate/DNA co-precipitation method. Four days after transfection, Geneticin® (Life Technologies, Inc.), 0.8 mg/ml, was added to the culture medium to allow selection of transfected cells. Medium was changed every other day and fresh Geneticin® (0.8 mg/ml) was added. After 18 days, all colonies (ϳ100 -150/construct) were pooled so as to eliminate any potential influence of the integration site of the constructs within the cell genome on the promoter activity. The stably transfected cultures generated were maintained in DMEM containing 10% newborn calf serum and 0.4 mg/ml of Geneticin® and utilized at confluence to study the regulation of the COL1A2 promoter. Four hours prior to addition of cytokines and growth factors, the cultures were rinsed once with phosphate-buffered saline and incubated in medium containing 1% newborn calf serum. Fourty hours later, cells were lysed in lysis buffer (see above). The protein concentration of each extract was determined with a commercial assay kit (Bio-Rad), and CAT activity was measured as described above, using identical amounts of protein in each sample (50 g).
Gel Mobility Shift Assays-Nuclear extracts were prepared according to the method of Andrews and Faller (1991). For gel mobility shift assays, a 37-bp double-stranded DNA oligomer corresponding to the region Ϫ271 to Ϫ235 of the human COL1A2 promoter, overlapping the element(s) between residues Ϫ265 and Ϫ241 responsible for TGF-␤ response (see "Results"), was generated: 5Ј-GAGGTATGCAGACAAC-GAGTCAGAGTTTCCCCTTGAA-3Ј. The end-labeled oligomer (ϳ7 ϫ 10 4 cpm) was incubated with 10 g of protein extracts for 30 min on ice in 20 l of binding reaction buffer (12 mM HEPES, pH 7.9, 4 mM Tris, pH 7.9, 60 mM KCl, 1 mM EDTA, 12% glycerol), in the presence of 2 g of poly(dI-dC), as described previously (Dignam et al., 1983). For competition experiments, 20 -60-fold molar excess of DNA was added to the binding reaction. Details of the competition assays are provided in the legend to Fig. 3 and the corresponding text under "Results." DNAprotein complexes were separated from unbound oligomers on 4 or 6% polyacrylamide gels in 0.4 ϫ TBE. The gels were fixed for 3 ϫ 10 min in 30% methanol, 10% acetic acid, vacuum-dried, and exposed to x-ray films at Ϫ70°C. In some experiments, nuclear extracts were preincubated with antisera against c-Jun (Santa Cruz Biotechnology, Santa Cruz, CA) or Jun-B (a kind gift from Dr. R. Bravo, Bristol-Myers Squibb Pharmaceutical Research Institute, Princeton, NJ).

RESULTS AND DISCUSSION
Deletion Analysis of the Human COL1A2 Promoter-We have constructed by a PCR-based methodology a large battery of 5Ј deletions of the human COL1A2 promoter linked to the CAT gene. A schematic diagram of the promoter is shown in Fig. 1A, indicating the 5Ј end position of each of the CAT constructs, relative to the transcription start site ϩ1. The relative activity of the different constructs in transiently transfected human neonatal fibroblast cultures was compared with that of the longest one, the 3.5 kilobases promoter construct, set arbitrarily at 100. Deletion from position Ϫ3500 to Ϫ772 led to a dramatic reduction of promoter activity, which was partially recovered with further deletion to position Ϫ376 (Fig.  1B). These data are in agreement with the data published by Boast et al. (1990). Examination of the region comprised between residues Ϫ376 and Ϫ108 indicated the presence of several cis-elements potentially important for regulation of the basal expression of the COL1A2 promoter. Specifically, three potential Sp-1-binding sites, one potential AP-1 and one potential NF-B-binding site were noted (Fig. 1A). Deletion of promoter sequences between residues Ϫ376 and Ϫ287 led to a ϳ50% reduction of promoter activity, whereas further deletion to position Ϫ265 did not reduce the activity further, as compared with the Ϫ287 fragment (Fig. 1B). Thus, our data indicate that the two upstream Sp-1 sites between nucleotides Ϫ432 and Ϫ287 play an important role in providing high basal activity to the promoter constructs, while the Sp-1 site further downstream (between residues Ϫ287 and Ϫ265) has little, if any, effect. These data contrast recent observations by Tamaki et al. (1995) who showed that deletion of all three Sp-1-binding sites located within this region of promoter is necessary to significantly alter the expression of the promoter. The reasons for this discordance are unknown, but both studies emphasize the importance of Sp-1 sites in maintaining high activity of the COL1A2 promoter.
Additional deletions to position Ϫ241 and to position Ϫ161, which removed two potential regulatory elements, AP-1-and NF-B-binding sites (see Fig. 1A), did not lead to further reduction in promoter activity (Fig. 1B), suggesting that the elements important for the basal expression of the COL1A2 promoter are mostly the Sp-1 sites described above and the sequences upstream of nucleotide Ϫ772. Further deletion to position Ϫ108 reduced the promoter activity to levels about 2-5% of the 3.5-kilobase promoter construct, suggesting that sequences between Ϫ161 and Ϫ108 are a prerequisite for the expression of COL1A2 at a significant level.
Delineation of TGF-␤-responsive Elements within the Human COL1A2 Promoter-The TGF-␤ responsiveness of the different deletion constructs described above was investigated. The stimulatory effect of TGF-␤ (ϳ5-10-fold) was observed with every construct containing at least 265 bp of COL1A2 promoter (Fig.  2B). Together with the analysis of the promoter sequences present upstream of nucleotide Ϫ265, these data suggest that the Sp-1-binding sites located between nucleotides Ϫ330 and Ϫ265 (see Fig. 2A), important for basal expression of the promoter (Tamaki et al., 1995; our results above), are not required for TGF-␤ response. Further deletion of 5Ј sequences to position Ϫ241 of the promoter led to complete loss of TGF-␤ responsiveness, suggesting that element(s) essential for TGF-␤ response are located within a relatively short stretch of DNA, the 25-bp segment between positions Ϫ265 and Ϫ241 of the COL1A2 Confluent fibroblast cultures were co-transfected with the 5Ј deletion constructs of the COL1A2 promoter linked to the CAT gene, together with pRSV-␤-galactosidase, by the calcium phosphate/DNA co-precipitation procedure, as described under "Materials and Methods." After glycerol shock, the cultures were incubated in medium containing 1% fetal calf serum for another 40 h prior to CAT assay. The cell extracts corresponding to the same levels of ␤-galactosidase activity were assayed for CAT activity using [ 14 C]chloramphenicol as a substrate by thin layer chromatography. The results are the mean of fifteen independent experiments utilizing overlapping sets of deletion constructs. The position of the 5Ј end of each construct is indicated below each histogram bar and is aligned with the schematic diagram shown in A.

FIG. 2. Effects of TGF-␤ and TNF-␣ on the activity of 5 deletions of the COL1A2 promoter in transient cell transfection experiments.
Confluent fibroblast cultures were transfected with various 5Ј deletion constructs of the human COL1A2 promoter linked to the CAT gene by the calcium phosphate/DNA co-precipitation procedure, as described under "Materials and Methods." After glycerol shock, the cultures were first incubated for 3 h in medium containing 1% fetal calf serum and then for another 40 h with either TGF-␤ (10 ng/ml, B) or TNF-␣ (25 ng/ml, C). In every experiment, duplicate culture plates were used for each case. The cell extracts were assayed for CAT activity with [ 14 C]chloramphenicol as a substrate using identical amounts of protein. Results are the mean Ϯ S.D. of 13 (TGF-␤) and 10 (TNF-␣) experiments utilizing overlapping sets of 5Ј deletion constructs. Panel A allows visual identification of the potential cis-elements involved in growth factor response with regard to the 5Ј deletions of the COL1A2 promoter.

promoter.
To verify the TGF-␤ responsiveness in a stable expression system, NIH 3T3 fibroblast cultures stably transfected with constructs Ϫ342, Ϫ287, Ϫ265, and Ϫ241/CAT were generated. Incubation of these transfectants with TGF-␤ confirmed the data obtained in transient transfections of human dermal fibroblasts, indicating that the segment of promoter located between residues Ϫ265 and Ϫ241 is indeed essential for TGF-␤ response. Specifically, all stably transfected constructs, except Ϫ241/CAT, responded to TGF-␤ by a 3-5-fold elevation of promoter activity (not shown). These data contrast those by Ina-gaki et al. (1994) which suggested that sequences extended to span nucleotides Ϫ330 to Ϫ286 had to be present, together with the sequences located between residues Ϫ271 and Ϫ255, to confer TGF-␤ responsiveness to a heterologous promoter, the thymidine kinase promoter, normally unresponsive to TGF-␤. The experimental approach taken by these authors consisted in analyzing the region of the COL1A2 promoter located between residues Ϫ378 and Ϫ183 by DNase footprinting. Two distinct areas protected from nuclease digestion were characterized, one between nucleotides Ϫ271 and Ϫ255, the other between nucleotides Ϫ330 and Ϫ286. Deletion of either one of the protected fragments led to significant reduction in the promoter activity, suggesting that these two regions of COL1A2 promoter are necessary for TGF-␤ effect. In our experiments, upstream Sp-1 sequences were found to be fundamental for high expression of the promoter, as demonstrated by Tamaki et al. (1995), but are not required for TGF-␤ response. Further evidence for the lack of involvement of Sp-1 in TGF-␤ response is provided below (see Figs. 3 and 5 and the corresponding text).
Additional 5Ј deletion to position Ϫ161 of the promoter restored some of TGF-␤ responsiveness (ϳ2.5-fold stimulation) (Fig. 2B). This elevation of promoter activity was lost when the 5Ј end of our construct was decreased to position Ϫ108 of the promoter. Since the activity of the shortest construct was extremely low (see Fig. 1 and related text under "Results"), we established stably transfected NIH 3T3 fibroblast cultures with the Ϫ108/CAT construct. No significant effect of TGF-␤ was detected on the promoter activity using this experimental approach (not shown). It appears, therefore, that an essential TGF-␤-response element is located between residues Ϫ265 and Ϫ241 of the COL1A2 promoter, although additional, yet somewhat secondary, sequences allowing some TGF-␤ responsiveness may exist downstream from nucleotide Ϫ161.
Delineation of TNF-␣-responsive Elements within the Human COL1A2 Promoter-Using the same set of 5Ј deletion constructs of the human COL1A2 promoter described above, we analyzed the potential regions of the promoter responsible for transcriptional inhibition by TNF-␣. TNF-␣ had a strong inhibitory effect on the constructs containing at least 265 bp of COL1A2 promoter sequences (between 45 and 74% inhibition as compared with the basal activity of each individual CAT construct, Fig. 2C). Further deletion of the 5Ј end of the promoter to either position Ϫ241 not only abolished the inhibitory effect but actually reversed the response to TNF-␣ (Fig. 2C). The extent of stimulation of the activity of the Ϫ241/CAT construct by TNF-␣ reached ϳ3-fold above control. Similar results were obtained when using NIH 3T3 fibroblast cultures stably transfected with the various COL1A2 promoter/CAT constructs were examined, excluding any misleading result due to the transient transfections (not shown).
Sequences within the Proximal 265-bp Segment of COL1A2 Promoter Are Sufficient to Mediate Antagonism between TGF-␤ and TNF-␣-Since our data indicate that the region located between nucleotides Ϫ265 and Ϫ241 is essential to mediate both TGF-␤ up-regulation and TNF-␣ down-regulation of COL1A2 promoter activity, we proceeded to fine-map the region of the promoter allowing the antagonist effect of these two cytokines. For this purpose, fibroblast cultures were transiently transfected with various deletion constructs and incubated with TGF-␤, in the absence or presence of TNF-␣. The results, presented in Table I, clearly establish that 265 bp of promoter sequences are sufficient to mediate the antagonism between TGF-␤ and TNF-␣. Specifically, when TNF-␣ was added concomitantly with TGF-␤, the promoter activity was reduced to 68 -17% of the activity observed in the presence of FIG. 3. Binding of nuclear proteins to a DNA fragment spanning the region from ؊271 to ؊235 of the COL1A2 promoter. A 37-bp oligonucleotide including the sequence found between residues Ϫ271 and Ϫ235 of the human COL1A2 promoter was end labeled with [␥-32 P]ATP and used in gel electrophoresis mobility shift assays to study its ability to bind transcription factors. A, the nuclear extracts used were as follows: lane 1, none; lanes 2 and 4 -7, untreated confluent fibroblast control cultures; lane 3, TGF-␤-treated confluent fibroblast cultures. Competition binding studies were performed as follows: lanes 1-3, no competitor DNA; lanes 4 and 5, 20-and 60-fold molar excess of homologous unlabeled oligomer; lane 6, 20-fold molar excess of a 15-bp oligonucleotide corresponding to the region Ϫ258 to Ϫ244 of the COL1A2 promoter; lane 7, 20-fold molar excess of a 22-bp oligonucleotide containing the collagenase AP-1-binding site (underlined), 5Ј-CTAGTGATGAGTCAGCCGGATC-3Ј. B, the nuclear extracts used were as follows: lane 1, none; lanes 2 and 4 -9, untreated confluent fibroblast control cultures; lane 3, TGF-␤-treated confluent fibroblast cultures. Competition binding studies were performed as follows: lanes 1-3, no competitor DNA; lanes 4 and 5, 20-and 60-fold molar excess of homologous unlabeled oligonucleotide; lanes 6 and 7, 20-and 60-fold molar excess of a 26-bp oligonucleotide containing the mouse COL1A2 NF-1 binding sequence (underlined), 5Ј-CTTCCAAACTTG-GCAAGGGCGAGAGA-3Ј (Rossi et al., 1988); lanes 8 and 9, 20-and 60-fold molar excess of a 22-bp oligonucleotide containing a consensus NF-B binding sequence (underlined) found in the immunoglobulin light chain gene, 5Ј-GATCGAGGGGACTTTCCCTAGC-3Ј (Lenardo and Baltimore, 1989).
TGF-␤ alone. When Ϫ241/CAT was used in transient transfection experiments, the promoter activity was not enhanced by TGF-␤ but TNF-␣ significantly elevated its activity, i.e. 1.7-5fold above the promoter activity noticed in presence of TGF-␤ alone.
AP-1 Binds to the Ϫ265/Ϫ241 Region of the COL1A2 Promoter-To determine whether the region of the COL1A2 promoter located between nucleotides Ϫ265 and Ϫ241 was a site for binding of transcription factors, we performed gel mobility shift assays first with a radiolabeled 37-bp oligonucleotide containing the sequence of the Ϫ271/Ϫ235 region of the human COL1A2 promoter. Our results indicate that this region binds nuclear proteins isolated from both control and TGF-␤-treated fibroblasts (Fig. 3A, lanes 2 and 3). As predicted, the specific binding activity of nuclear extracts from control confluent fibroblast cultures (Fig. 3A, lane 2) could be competed in a dose-dependent manner by addition of 20-and 60-fold excess of homologous DNA (Fig. 3A, lanes 4 and 5). In addition, a 15-bp oligonucleotide containing the potential AP-1-binding site identified at position Ϫ250 (underlined), 5Ј-AACGAGTCA-GAGTTT-3Ј, also competed with the binding when used in a 20-fold molar excess (Fig. 3A, lane 6). Similarly, a 23-bp oligonucleotide containing the collagenase AP-1-binding site (underlined), 5Ј-CTAGTGATGAGTCAGCCGGATC-3Ј, competed with the binding when added in a 20-fold molar excess to the reaction mixture (Fig. 3A, lane 7). To investigate further the specificity of the protein binding to this DNA sequence, additional competition experiments were performed, utilizing unlabeled oligonucleotides containing consensus sequences for either NF-1 or NF-B (Fig. 3B). The binding activity of control nuclear extracts (Fig. 3B, lane 2) was competed by the addition of cold homologous competitor (Fig. 3B, lanes 4 and 5) but was not altered by the addition of either NF-1 or NF-B oligonucleotides (Fig. 3B, lanes 6, 7, and 8, 9, respectively). Collectively, these data indicate that AP-1 binds to the segment of COL1A2 promoter 5Ј-CGAGTCA-3Ј within the TGF-␤-responsive region identified above.
It should be noted that the binding activity, as determined by the intensity of the DNA/protein band, of extracts from TGF-␤-treated cells was not significantly altered as compared with that of control extracts. This quantitative similarity may, however, be masking changes in both the content and the transcriptional activity of the bound complexes. Specifically, the AP-1 complex is a dimer of gene products of the Fos and Jun families of oncogenes, which have closely related recognition sites but different transcriptional activities and DNA binding affinities (reviewed in Vogt and Bos (1990)). For example, whereas c-Jun is a potent activator of the c-jun and collagenase promoters, Jun-B is not and inhibits their trans-activation by c-Jun (reviewed in Mauviel (1993)). To characterize further the protein complex binding to the Ϫ265/Ϫ241 region of human COL1A2, nuclear extracts from TGF-␤-treated fibroblast cultures were incubated with antibodies against c-Jun or Jun-B, prior to detection of DNA/protein interactions by gel mobility shift assay. As shown in Fig. 4  pates in the formation of the complex that binds to the Ϫ265/ Ϫ241 region of COL1A2. This result is in agreement with our previous observations that TGF-␤ is a potent activator of jun-B expression in fibroblasts  and further indicates that certain components of AP-1 are part of the transcription factor binding to the COL1A2 promoter.
Mutation of the AP-1-binding Site within the Proximal COL1A2 Promoter Region Abolishes TGF-␤ Responsiveness-We have previously shown that TGF-␤ is capable of modifying AP-1-driven but not NF-B-driven transcription of the genes encoding collagenase and IL-8, respectively . Together with the data presented above, these observations led us to investigate the role of the putative AP-1-binding site located between residues Ϫ265 and Ϫ241 in mediating TGF-␤ effect on COL1A2 promoter. To investigate this possibility, we employed site-directed mutagenesis to modify the AP-1 site from its original sequence 5Ј-CGAGTCA-3Ј to 5Ј-CCAGTGA-3Ј in two constructs, the Ϫ342/CAT and the Ϫ265/CAT. The first one contains three Sp-1 sites upstream of the AP-1 sequence , whereas the latter construct has the AP-1 binding site located right at its 5Ј end. The mutated constructs were then used in transient transfection experiments in parallel with their wild-type counterparts. Mutation of the AP-1 site in either construct led to a dramatic drop, ϳ50%, of basal activity as compared with their unmutated counterparts and resulted in almost complete loss of TGF-␤ responsiveness (Fig.   FIG. 5. Effect of a mutation in the AP-1-binding site on COL1A2 promoter activity and its regulation by growth factors. Two 5Ј deletion constructs (Ϫ342 and Ϫ265) were subjected to substitution mutations in the potential AP-1-binding site located at position Ϫ250, as described under "Materials and Methods." The mutated and the parent constructs were used in parallel transfections of fibroblasts, as described in the legend to Fig. 2. After glycerol shock, the cultures were incubated for 3 h in medium containing 1% fetal calf serum and incubated for another 40 h without or with TGF-␤ (10 ng/ml), in the absence or presence of TNF-␣ (25 ng/ml). Processing of samples was carried out as described in the legend to Fig. 2. The various promoter constructs used were as follows: panel 1, Ϫ342COL1A2 promoter/CAT; panel 2, mutated Ϫ342COL1A2 promoter/CAT; panel 3, -265COL1A2 promoter/CAT; panel 4, mutated Ϫ265COL1A2 promoter/CAT. A, autoradiograms of a representative experiment. B, the results, presented as relative promoter activity, are the mean Ϯ S.D. of three independent experiments, each performed using duplicate samples, and are expressed as percent of acetylation of the [ 14 C]chloramphenicol substrate. A and B, panels 1 and 3 versus panels 2 and 4, respectively). Specifically, as observed in three separate experiments (Fig.  5B), the level of induction of the parent constructs by TGF-␤, ϳ10 -13-fold induction above control levels for both Ϫ342/CAT and Ϫ265/CAT constructs, was reduced to 1.6 -2.2-fold above control for the mutated constructs. These results clearly demonstrate that the AP-1 binding site within this 25-bp promoter is fundamental for TGF-␤ response, independently of the upstream Sp-1 sites.

5,
Overexpression of c-jun Blocks the Effect of TGF-␤ on COL1A2 Promoter Activity-We have previously established that TGF-␤ and TNF-␣ exert antagonistic activities on AP-1 driven promoters . This antagonism is due to differential induction of oncogenes of the Jun family, namely jun-B and c-jun; their products have been shown to exert opposite transcriptional activities (Schü tte et al., 1989;Deng and Karin, 1993;Mauviel et al., 1993). Specifically, TNF-␣ is a potent inducer of c-jun, whereas TGF-␤ induces jun-B in dermal fibroblasts, leading to opposite effects on collagenase gene expression through AP-1 controlled trans-activation of the promoter (reviewed in Mauviel (1993)). Since TNF-␣ is a potent inducer of c-jun (Brenner et al., 1989;Mauviel et al., 1993) and since mutations in the AP-1 site of the COL1A2 promoter abolish TGF-␤ response (see above), we investigated the effect of c-jun overexpression on TGF-␤-induced activation of the COL1A2 promoter. For this purpose, confluent fibroblast cul-tures were co-transfected with pRSVc-jun and Ϫ342/CAT and treated with TGF-␤. As shown in Fig. 6, overexpression of c-jun led to a dramatic reduction of Ϫ342/CAT activity and abolished the effect of TGF-␤. Specifically, c-jun overexpression reduced the basal COL1A2 promoter activity by ϳ60% and totally blocked the response to TGF-␤. These data suggest that (a) AP-1/c-Jun plays a significant role in the basal activity of the COL1A2 promoter, and (b) the antagonistic effect of TNF-␣ on TGF-␤ induced up-regulation of collagen gene expression may be mediated, at least in part, by c-jun overexpression upon TNF-␣ stimulation.
Conclusions-Using a series of 5Ј deletion constructs of the COL1A2 promoter, we have excluded Sp-1-binding sites as being necessary for TGF-␤ response. In addition, we have characterized a short fragment of the COL1A2 promoter, spanning from nucleotides Ϫ265 to Ϫ241, as an AP-1-binding site. Using site-directed mutagenesis, we have demonstrated that this region plays a regulatory role in the basal activity of the promoter and is fundamental for TGF-␤ response. Furthermore, our data suggest that this AP-1 site may be sufficient to mediate TNF-␣ inhibitory effect on TGF-␤ driven up-regulation of COL1A2 gene expression. This study extends our understanding of the molecular mechanisms involved in the transcription of the COL1A2 gene and provides novel mechanistic information on growth factor and cytokine regulation.