Parathyroid Hormone Regulates the Rat Collagenase-3 Promoter in Osteoblastic Cells through the Cooperative Interaction of the Activator Protein-1 Site and the runt Domain Binding Sequence*

Parathyroid hormone induces collagenase-3 gene transcription in rat osteoblastic cells. Here, we characterized the basal, parathyroid hormone regulatory regions of the rat collagenase-3 gene and the proteins involved in this regulation. The minimal parathyroid hormone-responsive region was observed to be between base pairs −38 and −148. Deleted and mutated constructs showed that the activator protein-1 and the runt domain binding sites are both required for basal expression and parathyroid hormone activation of this gene. The runt domain site is identical to an osteoblast-specific element-2 or acute myelogenous leukemia binding sequence in the mouse and rat osteocalcin genes, respectively. Overexpression of an acute myelogenous leukemia-1 repressor protein inhibited parathyroid hormone activation of the promoter, indicating a requirement of acute myelogenous leukemia-related factor(s) for this activity. Overexpression of c-Fos, c-Jun, osteoblast-specific factor-2, and core binding factor-β increased the response to parathyroid hormone of the wild type (−148) promoter but not with mutation of either or both the activator protein-1 and runt domain binding sites. In summary, we conclude that there is a cooperative interaction of acute myelogenous leukemia/polyomavirus enhancer-binding protein-2-related factor(s) binding to the runt domain binding site with members of the activator protein-1 transcription factor family binding to the activator protein-1 site in the rat collagenase-3 gene in response to parathyroid hormone in osteoblastic cells.

Parathyroid hormone (PTH) 1 is an essential regulator of calcium homeostasis (1). In addition to kidney, its major target tissue is bone, the body's main calcium store. While PTH increases serum calcium partly by activating osteoclasts, these cells do not display PTH receptors. Instead, PTH exerts a direct effect on osteoblasts, causing them to cease synthesis of type I collagen (2,3), the major organic component of bone. Most relevant to the current study, we and others have demonstrated that, in vitro, PTH can stimulate the osteoblastic synthesis of interstitial collagenase, the enzyme that specifically degrades fibrillar collagens (4,5). Although collagenase synthesis and secretion by osteoblasts has been well documented, the signaling mechanism through which PTH stimulates its expression in this cell type is not fully understood. We have employed the UMR 106-01 (UMR) rat osteosarcoma cell line to investigate PTH regulation of collagenase-3 gene expression in osteoblasts. This cell line displays classical osteoblastic markers including PTH receptors, type I collagen, and high alkaline phosphatase expression. Most importantly to the present study, UMR cells decrease collagen synthesis and begin production of interstitial collagenase in response to PTH treatment.
Previously, we reported that UMR cell collagenase induction by PTH is due to an increase in transcription and is a secondary response since it requires de novo protein synthesis (6). In the present work, we have dissected the minimal PTH-responsive region of the rat collagenase-3 gene. This was achieved by transiently transfecting 5Ј-deleted and internally mutated rat collagenase-3 promoter constructs into UMR cells to assess the effect of PTH on each region within this gene. These constructs revealed that the minimum PTH regulatory region is within 148 base pairs upstream of the transcriptional start site. This region contains several consensus transcription factor recognition sequences including C/EBP, runt domain binding sequence (RD site), p53, PEA-3, activator protein (AP)-2, and AP-1. The following report describes concurrent participation of the AP-1 and RD sites, which are both necessary for PTH induction of the collagenase-3 gene, and shows that the AP-1 site is a basal element. We also identify the families of proteins associated with those sites. By overexpression studies, we confirm a functional requirement of both the native AP-1 and RD sites and the AP-1 and AML-related protein(s) as involved in PTH regulation of the collagenase-3 promoter. It is likely that the signaling mechanism we describe for PTH-induced collagenase transcription is specific to osteoblasts and therefore provides a relevant contribution to the current understanding of bone biology.

EXPERIMENTAL PROCEDURES
Materials-Parathyroid hormone (rat PTH 1-34) was purchased from Sigma. Restriction endonucleases were products of Promega Corp. (Madison, WI), and radionuclides were obtained from NEN Life Science Products. Synthetic oligonucleotides were synthesized by Midland Certified Reagent Company (Midland, TX). Radiolabeled [ 14 C]chloramphenicol was obtained from Amersham Pharmacia Biotech. Tissue culture media and reagents were obtained from the Washington University Tissue Culture Center (St. Louis, MO). Fetal bovine serum was a product of JRH Biosciences (Lenexa, KS) and was also purchased through Washington University. All other chemicals were obtained from Sigma or Fisher.
Primer Extension Analysis-Primer extension was carried out using a modified procedure of Boorstein and Craig (7). A 21-mer synthetic oligonucleotide, representing the complement of nucleotides 23-43 of the rat collagenase cDNA (8), was radiolabeled by T4 polynucleotide kinase and [␥-32 P]ATP to a specific activity of 1 ϫ 10 8 cpm/g. Five nanograms of the probe was annealed at 58°C for 45 min with 5 g of poly(A ϩ ) RNA extracted from cells treated with PTH (10 Ϫ8 M) for 4 h. The annealing mixture (250 mM KCl, 10 mM Tris-HCl, pH 8.3, 10 mM EDTA) was cooled slowly; primer extension was initiated by the addition of reverse transcriptase (200 units of SUPERSCRIPT II, Life Technologies, Inc.) in reverse transcriptase buffer (50 mM Tris-HCl, pH 8; 75 mM KCl; 10 mM dithiothreitol; 3 mM MgCl 2 ; 0.5 mM each of dGTP, dATP, dCTP, and dTTP; 0.13 mg/ml actinomycin D; 450 units/ml of RNasin); and incubation was continued at 42°C for 1 h. Following the primer extension, the reaction mixture was precipitated with ethanol, centrifuged, dried, resuspended in 10 l of loading buffer, and electrophoresed on an 8% denaturing sequencing gel beside a sequencing ladder. The sequencing reaction was carried out by the Sequenase version 2.0 DNA sequencing kit (Amersham Pharmacia Biotech) with the same primer and rat collagenase-3 genomic clone 600Ϫ10 DNA.
Preparation of 5Ј-Deleted Promoter Constructs-Previously in this laboratory, Dr. Cheryl Quinn cloned and characterized the rat collagenase genomic clone 600-10, which contains the 5Ј-end of the gene (9). From this, using convenient restriction sites, the longest collagenase promoter constructs were subcloned upstream of a chloramphenicol acetyltransferase (CAT) reporter gene in pBluescript SK (Stratagene).
The shorter collagenase promoter fragments were generated by polymerase chain reaction (PCR) and linked to the bacterial CAT gene in the reporter plasmid pSV0CAT (Promega). All fragments were linked to the CAT gene at a site that was 28 base pairs 3Ј to the transcriptional start site. Two synthetic oligonucleotides were designed as right and left primers for each deletion mutant. Three nucleotides, GAC, of a palindromic six-base SalI restriction site were added to the 5Ј-end of each primer in order to engineer SalI linkers at each end of the PCR product. The conditions for PCR were as follows: left and right primers (50 M each), DNA template (10 ng), dNTPs (2.5 mM each), 10ϫ amplification buffer (10 l) (500 mM KCl, 100 mM Tris-HCl, pH 8.4), MgCl 2 (2.5 mM), Taq polymerase (2.5 units), and H 2 O to a total volume of 100 l in a thin walled PCR tube. Mineral oil (100 l) was added to the top of the solution. The initial temperature for PCR was 94°C (2 min) and then 40 cycles of denaturation (94°C, 1 min), annealing (55°C, 1 min), and elongation (72°C, 1 min). The final extension was at 72°C (5 min), and then refrigeration was at 4°C. The PCR products were purified and precipitated by the QIAquick-spin PCR purification kit (Qiagen). Extra 3Ј-A were filled with Klenow and dNTPs (2.5 mM), concatemerized with T4 DNA ligase, and digested with SalI to obtain the desired fragments. The final product was then ligated into SalI-digested pSV0CAT and transformed into Escherichia coli (DH-5␣). The sequence and orientation of each construct was confirmed using the Sequenase version 2.0 DNA sequencing kit (Amersham Pharmacia Biotech).
Site-directed Mutagenesis-Mutations were introduced into desired constructs using the Chameleon double-stranded site-directed mutagenesis kit (Stratagene). A set of selection and mutagenic oligonucleotide primers was designed and synthesized for each mutant. The AP-1 site TGACTCA was changed by mutating G to A at the second base or TGA to ACT at the first three bases (10). The PEA-3 site consensus sequence AGGAAGT was altered to AAAAAGT (11), and the runt domain binding sequence (RD site) AACCACA was changed to ACTAACA (12). The selection primer was designed to change a BamHI restriction site in pSVOCAT vector into a BanI site. The identity of the mutants was confirmed by the Sequenase version 2.0 DNA sequencing kit (Amersham Pharmacia Biotech).
Transient Transfection-Cells were seeded at 1 ϫ 10 6 cells/100-mm diameter Petri dish and transiently transfected the following day using a calcium phosphate coprecipitation method modified from Rosenthal (14). Briefly, the cells were transferred to 10 ml of fresh maintenance medium 3 h before DNA application. DNA was added to 550 ml of a 0.25 M CaCl 2 solution and mixed dropwise into 550 ml of 2 ϫ HEPESbuffered saline solution (280 mM NaCl, 50 mM HEPES, 3 mM Na 2 HPO 4 , pH 7.1) per dish. When a cloudy precipitate was visible, the solution was added to the culture dishes, which were incubated in 8% CO 2 for 4 h. Following glycerol shock, the cells were returned to maintenance medium overnight and then treated with the appropriate agent(s). A preliminary time course indicated that maximal collagenase promoter activity was detectable after 24 h of PTH treatment. Therefore, all treatments were for this duration. To harvest, the cells were washed four times with phosphate-buffered saline and scraped into 1 ml of TEN (40 mM Tris-HCl, pH 7.5, 1 mM EDTA, 150 mM NaCl). The cells were then pelleted by centrifugation (1000 ϫ g, 10 min, 4°C) and resuspended in 150 l of 0.25 M Tris-HCl, pH 8.0. Cell lysis was achieved by three freeze-thaw cycles. Endogenous acetylases were inactivated by incubating samples at 60°C for 10 min (except samples to be assayed for ␤-galactosidase activity), and cellular debris was removed by centrifugation (12,000 ϫ g, 10 min, 4°C). The supernatant was assayed for CAT activity by a modification of the Seed and Sheen (15) procedure as described below. Separate plates were transfected with a ␤-galactosidase expression plasmid to verify transfection uniformity from experiment to experiment. This was done because cotransfecting this plasmid with other constructs had previously been shown to result in squelching (data not shown). Expression of the pSV-␤-galactosidase control vector was assayed as described below. Background was defined as the activity of the promoterless, enhancerless vector, pSV0CAT. The pSV2CAT plasmid was transfected into separate plates as a positive control in all experiments. All experiments were done 3-7 times with duplicate samples in each. Overexpression of transcription factors into UMR cells was done using LipofectAMINE as recommended by the manufacturer (Life Technologies, Inc.) in six-well plates (seeded at 2 ϫ 10 5 cells/well).
Assay of CAT Activity-CAT activity was measured by reacting 25 or 50 (six-well plate) l of cell lysate in duplicate in a 100-l reaction volume consisting of final concentrations of 250 M n-butyryl-coenzyme A and 23 M [ 14 C]chloramphenicol (0.125 Ci/assay). The reaction volume was adjusted to 100 l with 0.25 M Tris-HCl, pH 8.0, and reacted for 2 h at 37°C. Butylated chloramphenicol was removed by pre-extraction with 200 l of mixed xylenes (Aldrich). The xylene phase was back extracted with 100 l of 0.25 M Tris-HCl, pH 8.0. Butylated chloramphenicol retained in the final organic layer was determined by scintillation counting. The values were normalized to protein as determined by the Bradford (16) dye binding (Bio-Rad reagent) method. A standard curve using purified CAT was conducted every experiment to determine the linear range of the enzyme assay. Experimental CAT activity was always within the linear range.
␤-Galactosidase Assay-Cell lysate (150 l) was incubated with 150 l of 2ϫ assay buffer (120 mM Na 2 HPO 4 , 80 mM NaH 2 PO 4 , 2 mM MgCl 2 , 100 mM ␤-mercaptoethanol, 1.33 mg/ml o-nitrophenyl ␤-D-glucopyranoside) at 37°C until a yellow color developed. The reaction was terminated with 500 l of 1 M sodium carbonate, and the absorbance was read at 420 nm. Readings were compared with a standard curve made from the reaction product, o-nitrophenol.
Gel Mobility Shift Assay-Nuclear extracts were prepared from control and PTH-treated UMR 106-01 cells as described (17). Approximately 5 g of nuclear extract was incubated in a volume of 20 l containing binding buffer (final concentrations: 4% glycerol, 1 mM MgCl 2 , 0.5 mM dithiothreitol, 50 mM KCl, 10 mM Tris-HCl, pH 7.5), 100 ng/l poly(dI-dC), and antisera or competitor DNA at room temperature for 15 min. 32 P-labeled double-stranded oligonucleotide was added to the reaction immediately following the above reagents. The incubation was carried out for 15 min at room temperature. The reaction was stopped by the addition of 2 l of 10ϫ gel loading dye. Electrophoresis was performed at 4°C on a 6% nondenaturing polyacrylamide gel in TGE buffer (25 mM Tris, 190 mM glycine, and 1.1 mM EDTA, pH 8.5). The protein-DNA complexes were visualized by autoradiography. The sequences of the oligonucleotide probes were as follows (RD, AML-1, and AP-1 sites are underlined, and mutations are in boldface type).
The Primer Designer program was used to choose flanking sequences from the rat collagenase-3 promoter for the double-stranded oligonucleotides. This program predicts secondary structure formation and T m allowing stable hybridization. The AML-1 consensus binding site has been previously reported (12,18).

Identification of the Rat Collagenase-3 Gene Transcription
Start Site-Since the cDNA we obtained for rat collagenase was not full-length (8), we needed to identify the transcription start site of the gene. It should be noted that there are three human collagenase genes, known as human collagenase-1, -2, and -3 (or MMP-1, -8, and -13). In rat and mouse, only one interstitial collagenase has been identified (8,19), which is homologous to human collagenase-3. Thus, we have adopted this nomenclature to describe the rat collagenase promoter as that for rat collagenase-3. Using primer extension, we detected three major primer-extended products of 86, 87, and 88 nucleotides, corresponding to A, G, and G, respectively (Fig. 1). The nucleotide sequence of the missing preproenzyme not described in our previous work is included and is 17 nucleotides (8,9). By comparing the sizes of the cDNA, the genomic DNA, and the primer-extended products, we concluded that the transcription start site is 26 -28 nucleotides upstream from the translation start site. Since the major transcription start site appears to be 28 nucleotides upstream of the translational start site, we have used this to number all of the regulatory region constructs.
The Minimal PTH-responsive Region Is within 148 Base Pairs of the Transcriptional Start Site-The rat collagenase-3 promoter region includes consensus binding sites for several DNA-binding proteins, C/EBP, RD, p53, PEA-3, AP-2, and AP-1 ( Fig. 2A) and is shown in comparison with the human and mouse collagenase-3 promoters (Fig. 2B) (20,21). There are four consensus sites, namely an AML/PEBP2/runt (or OSE2 or RD) site, a p53 site, a PEA-3 site, and an AP-1 site, which are highly conserved both in sequence and location in all of these collagenase-3 promoters. In order to determine which region is necessary for PTH to activate transcription, we deleted regions of the rat collagenase-3 promoter from the 5Ј-end and placed the resulting promoter sequences 5Ј of the CAT gene. The deletion constructs were transiently transfected into UMR 106-01 cells and then treated with or without PTH and assayed for CAT activity. The results indicate that CAT activity increased as the constructs decreased until only 148 base pairs remained, and all constructs retained PTH-responsiveness (data not shown). Our major interest was the PTH-responsive elements, which are all contained within the first 148 base pairs upstream of the transcriptional start site.
In order to more precisely determine the minimal PTHresponsive region in the rat collagenase-3 promoter, we further deleted the WT(Ϫ148) construct. For instance, WT(Ϫ54) contains the AP-1 and C/EBP sites, while WT(Ϫ38) retains only a C/EBP site and TATA box. The ability of PTH to activate these shortened promoter regions was then assessed. In comparison with the WT(Ϫ148) construct, both WT(Ϫ54) and WT(Ϫ38) had very little basal expression and almost no PTH response (Fig.  3). These results suggest that the minimal PTH-responsive region lies within 148 base pairs 5Ј of the transcription start site of the collagenase-3 gene, and major basal elements reside between Ϫ148 and Ϫ54.
During the preceding work we considered the possibility that our results could be affected by the transient transfection procedure. We addressed this possibility by comparing the effect of PTH on a stably versus transiently transfected WT(Ϫ6500). We observed no difference in promoter activity (basal or PTHinduced) due to the type of transfection (data not shown). It should be noted that basal transcription is considerably higher with both the transient and stably transfected cells compared with the endogenous gene (6). It suggests that either a "silencer" exists elsewhere in the gene or the transfection process up-regulates basal expression. The latter has been observed for all the matrix metalloproteinase promoters (11). Alternatively, an activator region in the endogenous gene may be masked by nuclear matrix resulting in reduced basal expression.
Mutation of the AP-1 Site Decreases Basal Expression but Does Not Abolish PTH Stimulation-Since deletion to Ϫ54 decreased basal expression so substantially, it was difficult to assess the role of the AP-1 site. In order to understand the role of the AP-1 motif in PTH regulation of the collagenase-3 gene, we mutated this element in several constructs and measured the ability of the mutants to drive CAT activity. One or three nucleotides of the consensus AP-1 sequence were mutated (10). We have altered the second nucleotide in the AP-1 sequence in the WT(Ϫ148) construct from a G to an A (position Ϫ47) and referred to this construct as M(Ϫ148A1). The first three nucleotides in the consensus AP-1 site were also mutated from TGA to ACT in the WT(Ϫ148) and WT(Ϫ54) constructs. The mutations destroyed the first half-site of the AP-1 palindrome sequence and were named M(Ϫ148A3) and M(Ϫ54A3), respectively. The CAT activity obtained when UMR cells were transfected with these constructs in the presence or absence of PTH is shown in Fig. 4. The PTH response or stimulation is represented as -fold stimulation over the control. In compari- son with WT(Ϫ148), M(Ϫ148A1) had reduced basal CAT activity. Previous work has shown that the single base mutation is analogous to a mutation in the human collagenase-1 AP-1 sequence that prevents binding of the Fos-Jun heterodimer (10). Significantly, this mutation did not abolish the PTH response in rat UMR 106-01 osteoblastic cells, indicating that the AP-1 sequence was not absolutely required for the response to PTH. A triple mutation M(Ϫ148A3) showed a lesser basal CAT activity compared with M(Ϫ148A1) but still did not abolish the PTH response; however, mutation of the AP-1 motif greatly affected the basal expression of the rat collagenase-3 promoter constructs.
The PEA-3 Site Does Not Contribute to PTH Stimulation-Since the PEA-3 (polyomavirus enhancer activator-3) site at Ϫ77 to Ϫ71 had been shown to be a contributory element in other metalloproteinase promoters (11), we mutated this site in the WT(Ϫ148) construct. The consensus PEA-3 sequence, AG-GAAGT, was altered to AAAAAGT and named M(Ϫ148P2). This construct had increased PTH response and basal expression compared with WT(Ϫ148), and it suggested that this may be a silencer element. Combined mutation of AP-1 and PEA-3 sites gave essentially the same result as with AP-1 mutation alone; basal expression was decreased, but PTH response was not abolished (Fig. 5).

Mutation of the AP-1 Site and RD Site Substantially Reduces both Basal Expression and PTH Stimulation-
The sequence AACCACA has recently been identified as an osteoblast-specific element (OSE2) in the mouse osteocalcin gene promoter (22). Merriman et al. (18) have also shown the same consensus element, ACC(A/G)CA as an AML-1 consensus binding sequence in the rat osteocalcin gene promoter and shown that it binds a member of the AML-1/PEBP2/runt domain transcription factor family. The same AACCACA sequence, the RD site, lies between Ϫ132 and Ϫ126 in the rat collagenase-3 promoter.
Other studies have shown that mutation of ACC/TGG into CTA/GAT in this consensus sequence results in a failure of these proteins to bind and pinpoints the importance of the ACC/TGG motif (12,18). Based on these studies, we mutated the same three nucleotides in the RD site of the rat collagenase-3 promoter. To determine the importance of cooperativity with the AP-1 site, we made several constructs: WT(Ϫ125), which lacks the RD site; M(Ϫ125A3), which lacks the RD site and has a mutation in the AP-1 site; M(Ϫ148R3), which has a mutation in the RD site; and M(Ϫ148A3R3), which has mutations in both the RD site and the AP-1 site. All of the above constructs were transfected into UMR cells and assayed for CAT activity (Fig. 6). The absence of the RD site, WT(Ϫ125), resulted in reduced basal activity as well as a slight decrease in PTH response; but deletion of the RD site together with mutation in the AP-1 site, M(Ϫ125A3), essentially reduced both basal expression and PTH responsiveness to background activities. Transfections using M(Ϫ148R3) demonstrated a small reduction in basal expression in comparison with WT(Ϫ148). Interestingly, the M(Ϫ148A3R3) construct, which had mutations in both the RD and AP-1 sites, was similar to the M(Ϫ125A3) construct, and activity was reduced to background for both control and PTH-treated cells. The M(Ϫ148A3R3) construct should be compared with the M(Ϫ148A3) construct shown in Fig. 4; this latter construct still retained a small amount of basal activity and PTH responsiveness. It required mutation or deletion of the RD site to essentially abolish both basal expression and PTH response.
Identification of Proteins Binding to the AP-1 Site-The loss of promoter activity when the AP-1 site and RD site are mutated prompted us to identify the proteins associated with those sites and how they are regulated in response to PTH. Earlier, work in our laboratory revealed that Fos protein accumulation was maximal after 1 h of PTH treatment in UMR cells (24). Fig. 7 demonstrates that proteins in both control and PTH-treated (1 h) nuclear extracts were able to bind to the rat collagenase-3 AP-1 site, but significantly more protein-DNA complex was produced when extract from PTH-treated cells was used, as evidenced by the greater intensity of the shifted bands. To identify the proteins bound to the AP-1 site, antibodies to c-Fos, c-Jun, CREB, and phospho-CREB were incubated with the nuclear extracts of both control and PTH-treated cells, followed by the addition of labeled AP-1 probe. The antibodies to c-Fos and c-Jun, especially c-Fos, caused a significant reduction of protein binding when PTH-treated nuclear extract was used. The antibody to total CREB, but not the antibody to phosphorylated CREB, produced a supershifted band in both control and PTH-treated nuclear extracts. The anti-c-Fos and anti-c-Jun antibodies are known to interact with the DNA binding domain of Fos and Jun, respectively, and would therefore not allow for a supershifted band to be detected but, instead, would abolish binding to the rat collagenase-3 AP-1 site. The specificity of proteins binding to the AP-1 site was also determined by competition with unlabeled homologous AP-1 site and heterologous Sp1 site probes. When we used labeled mutant AP-1 site as the probe, the shifted bands seen with labeled wild AP-1 site probe were not found in both control and PTH-treated nuclear extracts. Further, cold competition with oligonucleotides of the mutant AP-1 site, wild type AP-1, or Sp1 consensus sequence and also incubation with c-Fos and c-Jun antibodies have indicated that the binding is relatively nonspecific and that the proteins bound to the labeled mutant AP-1 are not Fos and Jun (data not shown).
Identification of Proteins Binding to the RD Site-The proteins associated with the RD site were also identified. When the rat collagenase-3 RD site was used as a probe, nuclear extracts of both control and PTH-treated UMR cells were able to alter the mobility of the probe, producing identical shift patterns with no change in the abundance of binding (Fig. 8A). The binding specificity of the protein(s) to this site was examined by competing this protein-DNA complex with excesses of unlabeled RD site probe (Fig. 8A). The protein-DNA complex was similarly competed in both control and PTH-treated samples. At the same time, unlabeled mutant RD site oligonucleotide could not compete (Fig. 8B), and thus both experiments indicate the specificity of proteins binding to the native RD site. Since the AML/PEBP2 transcription factor family has a conserved runt domain (12,23), which binds to a consensus DNA sequence (12), cross-competition studies using the AML-1 consensus sequence were also performed. Fig. 8C shows that the protein-DNA complex was significantly reduced when unlabeled AML-1 consensus binding site oligonucleotide was added at increasing concentrations. This result suggests that the proteins binding to the RD site in the rat collagenase-3 promoter are members of the AML/PEBP2/runt domain transcription factor family.
Characterization of Proteins Binding to the RD Site-To further address the relationship between proteins binding to the RD site and the AML transcription factor family, we examined the immunoreactivity of proteins binding to the RD site using polyclonal antisera raised against human AML-1B and CBF-␤. A larger isoform of AML-1, AML-1B has an N-terminal extension (26); the AML-1B antiserum is specific to AML-1 members, and it has only weak binding to human AML-2 and AML-3 (27), which are other members of the AML family (28). But the human AML-1B antiserum has a cross-reactivity to all three forms of mouse AML (PEBP2␣A, -B, and -C) (29). The antiserum to CBF-␤ was used to determine whether the nuclear extracts also contain CBF-␤, a non-DNA binding partner protein of AML members. This family requires CBF-␤ for enhanced DNA binding activity by heterodimerization (23,27). Fig. 9 shows that the addition of AML-1B antiserum resulted in a supershifted band and a decrease in the faster migrating complex in both control and PTH-treated UMR cells. A similar result was obtained with CBF-␤ antiserum. It clearly indicates that both AML-related protein(s) and CBF-␤ are present in the nuclear extracts of control and PTH-treated UMR cells. Since the antiserum to human AML-1B was used for our gel shift study, the proteins in the nuclear extracts of rat osteoblastic osteosarcoma cell line (UMR) binding to the RD site are immunologically related to AML/PEBP2␣ proteins.
Functional Requirement of the AP-1 and RD Sites and the Proteins Binding to Them-To further elucidate the role of AML-related factor(s) for collagenase-3 promoter activity, we utilized AML-1/ETO, a repressor of AML proteins (12). AML-1/ETO is formed by the t(8:21) chromosomal translocation in which ETO protein is fused in frame at the end of the runt domain protein. This chimeric product recognizes the AML consensus sequence (12,26). The collagenase-3 promoter construct, WT(Ϫ148) was transiently cotransfected into UMR cells with increasing amounts of an AML-1/ETO expression plasmid. The results demonstrate that the PTH response was greatly reduced by overexpression of AML-1/ETO (Fig. 10), suggesting that this effect of the hormone requires endogenous AML-related factor(s) binding to the RD site in UMR cells.
To examine the importance of the AP-1 and AML-related transcription factor(s) and their binding sites in PTH-induced collagenase-3 promoter activity, we transfected wild type and FIG. 7. Gel mobility shift analysis of the rat collagenase-3 AP-1 site. UMR 106-01 cells were treated with control or PTH (10 Ϫ8 M)containing media for 1 h, and nuclear extracts were prepared. Proteins binding to the AP-1 site were assessed. An antibody to c-Fos (2 l), c-Jun (2 l), CREB (2 l), or phospho-CREB (2 l) or the same concentration of IgG (2 l) was included in the incubation before adding labeled wild type collagenase-3 AP-1 site probe. Competition was performed with a 50-fold molar excess of unlabeled AP-1 site oligonucleotide or Sp1 consensus oligonucleotide. Labeled mutant collagenase-3 AP-1 site probe was also used for gel shift analysis with control and PTH-treated nuclear extracts. Three experiments were carried out with similar results.
FIG. 8. Gel mobility shift analysis of the rat collagenase-3 RD site. UMR 106-01 cells were treated with control or PTH (10 Ϫ8 M)-containing media for 2 h, and nuclear extracts were prepared. Proteins binding to the RD site were assessed by gel mobility shift analysis. A, competition was performed with 10-, 50-, and 100fold molar excess of unlabeled native RD site probe. B, competition was performed with 50-and 100-fold molar excess of unlabeled mutant RD site oligonucleotide with PTH-treated nuclear extract. C, cross-competition was performed with 10-, 50-, and 100-fold molar excess of unlabeled AML-1 consensus binding site oligonucleotide. Three separate experiments were carried out with similar results. mutant rat collagenase-3 promoter constructs into UMR cells along with eukaryotic expression constructs coding for c-Fos, c-Jun, OSF2/CBFA1, and CBF-␤. Recently, Ducy et al. (30) have identified and cloned the mouse osteoblast-specific factor and named it OSF2/CBFA1 (the homolog of human AML-3). Since OSF2/CBFA1 has been shown as a regulator of osteoblast differentiation, we have used this construct for our overexpression studies. We initially titrated the optimal DNA concentration required for each expression construct for maximal collagenase-3 promoter activity. Approximately 5 ng of DNA of each was enough to maximally enhance the PTH response. Overexpression of c-Fos, c-Jun, OSF2/CBFA1, and CBF-␤ together with the WT(Ϫ148) construct increased the response to PTH to a greater extent than when each was individually transfected (Fig. 11). To determine whether the AP-1 and RD sites were required for this increased PTH response, we cotransfected all four expression constructs with mutant collagenase-3 promoter constructs (Fig. 12). We obtained the same results as shown in Fig. 6. In other words, mutation of either site prevented the other transcription factor from being effective; both sites and both types of transcription factor were required for stimulation of the collagenase-3 promoter by PTH. It should also be noted that overexpression of the proteins in control cells did not enhance promoter activity; this required stimulation by the hormone. DISCUSSION We have previously shown that the induction of interstitial collagenase by parathyroid hormone in rat osteoblastic osteosarcoma cells, UMR 106-01, is mostly at the transcriptional level (6), and it requires protein synthesis for the first 2 h of PTH treatment. In the present paper, we describe the minimal PTH-responsive region of the rat collagenase gene promoter (collagenase-3) and the proteins associated with the PTH-responsive region. By primer extension analysis, it is evident that the transcription start site is 28 nucleotides upstream of the ATG codon. The fact that more than one product appeared in primer extension analysis for the collagenase transcription start site suggests the possibility of heterogeneity at the 5Ј-end, or this appearance could result from incomplete primer extended products caused by difficulties encountered by reverse transcriptase due to the methylated nucleotides of the 5Ј-cap of eukaryotic mRNA. It does not appear to be artifactual, because it was seen in different RNA preparations. Recently, the transcription start site in the rat collagenase-3 gene was described as being 25 nucleotides upstream of the ATG codon (9). Here we report the addition of three nucleotides determined by our primer extension analysis and deduce that the major transcription start site in the rat collagenase-3 gene is 28 nucleotides upstream of the translation start site. Similarly, it has been shown that the transcription start site is 28 and 29 nucleotides upstream of the translation site in the human (21) and mouse collagenase-3 genes (20), respectively. In most of the genes belonging to the matrix metalloproteinase family, the transcription start site occurs within 30 nucleotides of the ATG codon.
Analysis of the 5Ј upstream region shows the presence of a typical TATA box (Ϫ29 to Ϫ23) in addition to AP-1 (Ϫ48 to Ϫ42) and PEA-3 (Ϫ77 to Ϫ71) sites, similar to several other genes in the matrix metalloproteinase family thus far analyzed (20,31,32). The presence of AML/PEBP2/runt or RD (OSE2), p53, PEA-3, and AP-1 consensus sites in rat, mouse, and human collagenase-3 promoters (20,21), which are conserved in sequence and location, suggests that these sites may be evolutionarily conserved to confer common routes of transcriptional regulation of the collagenase-3 genes. The AP-1 site (TGA(C/ G)TCA) in the human, mouse, and rat collagenase-3 promoters, as in other matrix metalloproteinase promoters, is a target for AP-1-mediated activation of transcription (21,31,34). It has also been shown that c-fos is a primary gene for transcriptional activation of matrix metalloproteinases (31). Earlier studies (25) of ours indicated that PTH treatment causes increased levels of c-fos and c-jun mRNAs in UMR 106-01 cells. Fos protein is induced by PTH from nearly undetectable levels, but Jun is measurable even under control conditions (24,35). The ratio of various Jun and Fos protein family members expressed in cells may be one determinant of their efficiency of transcriptional activation of target genes. It has been shown that Fos may interact with Jun or other nuclear proteins to regulate gene transcription (36,37). In this paper, we show the AP-1 site is necessary for basal expression of the rat collagenase-3 gene by transfection experiments and also report the increased binding by gel shift analysis of AP-1 family member proteins to this site in response to PTH. By gel shift analysis, we can also observe binding of a CREB-related protein to the AP-1 site of the rat collagenase-3 gene. This could be due to the similarity in sequence between AP-1 and cyclic AMP-response element sites in a gel shift assay. The possibility of heterodimer formation of Fos and CREB proteins, at least in a gel shift assay, also cannot be ruled out. Alternatively, the CREB family member could be ATF-1 because the CREB antibody used is able to cross-react with ATF-1. Nevertheless, it should be noted that this protein was not phosphorylated, since the anti-phospho-CREB antibody did not cause a supershift. Although CREB or ATF-1 interaction with the collagenase-3 AP-1 site was seen by gel shift analysis, it is unlikely that this is the mediator and far more likely that the AP-1 proteins are involved in collagenase-3 promoter activity for the following reasons. (i) Previously, we demonstrated that PTH induction of endogenous collagenase-3 gene expression was completely abolished by treating UMR cells with cycloheximide, showing that the PTH effect is a secondary response and requires de novo protein synthesis between 0 and 2 h of PTH treatment (6). A member of the CREB/ATF-1 family would not fulfill this requirement, since these proteins are constitutively synthesized. (ii) In c-fos Ϫ/Ϫ mouse osteoblasts, basal and PTH induction of collagenase-3 transcripts was significantly reduced. 2 (iii) Whether CREB or ATF-1 binds to the AP-1 site in a gel shift, this protein was not phosphorylated in response to PTH, since anti-phospho-CREB antibody did not cause a supershift (Fig. 7), whereas we know this antibody will supershift PTH-induced phosphorylated CREB bound to a cyclic AMP-response element in the c-fos gene (24). (iv) Overexpression of Jun B along with the collagenase-3 promoter construct in transient transfection experiments inhibited the PTH-induced collagenase-3 activity (38). Jun-B is known to be a negative regulator of Fos and Jun in other target genes (39); thus, Fos and Jun appear to be one of the functional induced regulators of the rat collagenase-3 promoter in osteoblastic cells. (v) Last, overexpression of CREB with or without OSF2 in transient transfection experiments did not stimulate collagenase-3 promoter activity in response to PTH (data not shown).
From our observations, the binding of any of these proteins to the AP-1 site appears not to be sufficient for the PTH response, since mutation of the AP-1 site does not eliminate the response to PTH and suggests the possibility that the AP-1 site may be more involved in basal expression. In certain cases, AP-1 binding has been shown to require contributory effects from other factors binding to the PEA-3 site. Human (40) and rabbit (41) interstitial collagenase genes (collagenase-1) have been shown to require the PEA-3 element in combination with the AP-1 site for full induction to occur in response to phorbol myristate acetate. It has been shown that basal as well as induced transcription from the human urokinase-type plasminogen activator gene requires an enhancer containing two elements, a combined PEA-3/AP-1 and a consensus AP-1 site (42). In the tissue inhibitor of metalloproteinases-1 promoter, c-Ets-1 enhances transcription synergistically with an AP-1 site (37). Our results indicate an absence of any cooperativity of the PEA-3 element with the AP-1 site for PTH responsiveness of the rat collagenase-3 promoter in UMR cells. Similarly, no significant synergistic effect has been found between the AP-1 site and the PEA-3 element in human collagenase-3 promoter activity (21).
The modulation of gene promoter activity by the cooperation of an AP-1 site with other transcriptional regulatory sites has been well documented. The myeloid specific expression of the leukocyte integrin gene, CD11c, is facilitated by cooperative interaction between the Sp1 and AP-1 binding sites (43). It has also been shown that the T cell-specific expression of interleukin-3 is achieved in part through the positive activities of the AP-1 and Elf-1 sites in the interleukin-3 promoter (44). Recently, it was shown that the activation of human collagenase-1 requires a cooperative interaction of the AP-1 site and the signal transducer and activator of transcription-binding element in response to oncostatin M (45). Thus, it is possible that, with the rat collagenase-3 promoter in osteoblastic cells, the AP-1 site acts together with another site in the 148 base pairs upstream of the transcriptional start site.
Banerjee et al. (46) have shown the importance of the AML consensus binding site in the rat osteocalcin gene for osteoblast-specific transcriptional activation. It is very striking that all three collagenase-3 promoters (human, mouse, and rat) have this conserved sequence (AML/PEBP2/runt or RD site or OSE2) in the same position. Since the proteins bound to the RD site of the rat collagenase-3 promoter can be competed by AML consensus sequence oligonucleotide, this result indicates that proteins binding to the RD site belong to the AML/PEBP2␣ transcription factor family. The identical gel shift pattern obtained in both control and PTH-treated conditions suggests that there is no change in abundance of these proteins. There could perhaps be an association of different isomeric forms of AML or AML-related proteins. The AML-1 protein exists in different isomeric forms, namely AML-1A, AML-1B, and AML-1C (47), in which AML-1A and AML-1B act antagonistically for transactivation (48). The requirement of AML-related factor(s) for PTH-induced collagenase-3 promoter activity is clearly demonstrated by overexpression of a chimeric protein, AML-1/ETO, which represses trans-activation functions of AML proteins. Recently it has been reported that the DNAbinding of AML-1 is regulated by changes in the reductionoxidation state of a conserved cysteine residue in its runt domain (49). The AML-1B protein can also be phosphorylated by activation of extracellular signal-regulated kinase (50). It is likely that the post-translational modification of the proteins plays a major role in collagenase-3 gene expression. The increased collagenase-3 promoter activity in response to PTH by overexpression of all four expression constructs with WT(Ϫ148) promoter construct suggests the possibility that PTH might be involved in post-translational modifications of these factor(s).
It is possible that nuclear extracts of both control and PTHtreated UMR cells may contain a unique osteoblast-specific protein(s) related to the AML/PEBP2␣ transcription factor family as suggested by Merriman et al. (18) and Geoffroy et al. (22). Recently, Ducy et al. (30) have identified the murine osteoblast-specific factor, OSF2/CBFA1, which is similar to AML-3/PEBP2␣A. It has also been shown that AML-3/ PEBP2␣A/CBFA1 is mainly involved in osteoblast maturation (51). If the family of AML/PEBP2␣/CBFA transcription factors were solely responsible for PTH induction of the rat collagenase-3 gene, the constructs containing mutations or deletions of the RD site should have totally abolished the CAT activity in response to PTH. This was not the case. Since neither independent mutations in the RD site nor in the AP-1 site could totally abolish the PTH response, it appears that these two sites interact cooperatively to activate collagenase-3 gene expression in response to PTH. Overexpression of the AP-1 factors and OSF2/CBFA1 with its partner protein CBF-␤ using the M(Ϫ148A3R3) construct in co-transfections also confirms a functional role of the AP-1 and RD sites for collagenase-3 promoter activity in response to PTH. Previous reports have shown that AML-1 can interact with Ets-1, Myb, and C/EBP transcription factors (52)(53)(54). At this stage, we are able to formally include the possibility that proteins from the family of AML/PEBP2␣/CBFA-related transcription factor(s) and AP-1 factors can interact together or independently with their corresponding binding sites and hence respond to PTH. In summary, our studies on the rat collagenase-3 gene promoter suggest that while the AP-1 site is required for basal expression of the rat collagenase gene in osteoblasts, concurrent participation of the RD site is required for PTH-mediated collagenase induction.