Progesterone Stimulation of Human Insulin-like Growth Factor-binding Protein-5 Gene Transcription in Human Osteoblasts Is Mediated by a CACCC Sequence in the Proximal Promoter*

Insulin-like growth factor-binding protein-5 (IGFBP-5) is produced by osteoblasts and potentiates insulin-like growth factor mitogenic stimulation in osteoblast cell cultures. Progesterone (PG) increased IGFBP-5 expression in normal human osteoblasts and increased IGFBP-5 transcription in U2 human osteosarcoma cells. We developed a chloramphenicol acetyltransferase reporter construct containing the human IGFBP-5 proximal promoter sequence, which includes TATA and CAAT boxes, and five putative PG response element half-sites. 10−8 m PG increased promoter activity of this construct in U2 cells co-transfected with a PG receptor isoform A (PRA) expression vector. Analysis of 5′ deletion constructs indicates that PG transactivation of IGFBP-5 promoter activity does not require the PG response element half-sites but does require the region −162 to −124 containing two tandem CACCC box sequences. Mutation of the proximal CACCC box at −139 eliminated PG transactivation. Gel shift assays using a −162 to −124 DNA fragment, U2 cell nuclear extracts, and purified PRA protein indicate that nuclear factors bind to a CACCC sequence at −139 and that PRA alters the pattern of transcription factor interaction with the CACCC sequence. Using a luciferase reporter construct containing base pairs −252 to +24 of the IGFBP-5 promoter, we found that both PRA and PRB isoforms mediated PG stimulation of promoter activity. These results suggest that PG may stimulate IGFBP-5 gene transcription via a novel mechanism involving PR and CACCC-binding factors.

Insulin-like growth factor-binding protein-5 (IGFBP-5) 1 belongs to a family of six structurally related proteins that are unrelated to the IGF receptors and that bind IGF-I and IGF-II with high affinity (1)(2)(3). IGFBPs modulate the mitogenic and metabolic activities of the IGFs in vitro, suggesting an important role in IGF physiology in vivo. Of the various IGFBPs, IGFBP-5 is unique in that it stimulates IGF-induced osteoblastic cell proliferation when added simultaneously with IGF I or II to serumfree osteoblast-like cell cultures (4 -7). Recent evidence suggests that IGFBP-5 may stimulate cell proliferation by an IGF-independent mechanism involving IGFBP-5 specific cell surface binding sites (7,8). In addition, IGFBP-5 binds with high affinity to hydroxyapatite and to various extracellular matrix proteins, properties that are consistent with a role for this binding protein in fixing IGFs to extracellular matrices including mineralized bone (7,9,10). IGFBP-5 may be a more potent enhancer of IGF activities when bound to extracellular matrix than in solution (10). Coding sequences and 5Ј-flanking regions of IGFBP-5 genes are highly conserved in human, rat, and mouse (4,5,(11)(12)(13), and the mouse and human 5Ј-flanking regions have been found to impart basal promoter activity to reporter gene constructs in several cell types, including osteoblasts (12)(13)(14)(15)(16).
Pertinent to this study, we found that progesterone (PG) stimulated human osteoblast cell proliferation (25,26) and increased IGFBP-5 mRNA levels (27). However, the molecular mechanism by which PG increased IGFBP-5 mRNA levels is not known. As a first step toward understanding how bone cell production of IGFBP-5 is regulated by PG, we isolated the 5Ј-flanking region of the human IGFBP-5 gene, linked the IGFBP-5 proximal promoter to a chloramphenicol acetyltransferase (CAT) reporter, and measured basal and progesteroneinducible promoter activity in transient transfection assays with U2 human osteosarcoma cells. We also examined the roles of progesterone receptor isoforms PR A and PR B in mediating PG induction of IGFBP-5 transcription. PRs are expressed in various cells as two different isoforms that are transcribed from separate promoters in the PR gene (28,29). The B-receptor isoform (PR B ) is 933 amino acids in length, and the A-receptor isoform (PR A ) lacks 164 amino acids at the N terminus.
Cell Culture-U2 cells were maintained in DMEM supplemented with 10% CS at 37°C in a humidified atmosphere composed of 95% air and 5% CO 2 . In most experiments, cells were changed to phenol red-free DMEM and CD-FBS (described below). PG was dissolved in ethanol at 10 Ϫ2 M and was further diluted with serum-free DMEM containing 0.1% bovine serum albumin before adding to the cells. Control cultures were treated with ethanol diluted to the same extent as for PG treatment.
Northern Blot Analysis-Total RNA was extracted from cell cultures by a single-step acid guanidinium thiocyanate-phenol-chloroform method (38). 20 g of total RNA was subjected to Northern analysis as described previously (23), using Magnagraph hybridization membranes (MSI, Westboro, MA). The human IGFBP-5 cDNA was a 0.35-kilobase fragment containing the coding region of exon 1 and was labeled with 32 P and hybridized as described (23). Relative abundance of each mRNA species was quantitated by laser densitometry (Biomed Instruments, Fullerton, CA) and corrected as described (39) for the amounts of RNA transferred to the hybridization membrane by dividing the densities of 28 S rRNA bands that were stained with methylene blue before hybridization. Relative mRNA levels are presented as the ratio of treated to control.
Nuclear Run-on Analysis-The effect of PG on the rate of IGFBP-5 gene transcription was determined with a modified nuclear run-on analysis as described (40). Briefly, U2 cells were treated with 10 nM PG or with vehicle for 1 and 4 h. The cells were rinsed and scraped into ice-cold phosphate-buffered saline, nuclei were isolated, and then nuclei were incubated in transcription buffer consisting of 0.6 M KCl, 12.5 mM MgCl 2 , 2.5 mM of ATP, CTP, and GTP, and 100 Ci of [ 32 P]UTP (800 Ci/mmol). RNA was isolated and equal amounts (2 ϫ 10 6 dpm) were hybridized to 10 g of alkali-denatured plasmid DNAs (IGFBP-5 cDNA, ␤-actin cDNA, and pBluescript DNA) that had been immobilized to Hybond-N filters. The filters were hybridized in a solution containing 10 mM TES, pH 7.4, 10 mM EDTA, 0.2% SDS, and 0.6 M NaCl at 68°C for 48 h. After hybridization, the filters were washed twice with 6ϫ SSPE, 0.1% SDS at 42°C for 30 min and once with 1ϫ SSPE, 0.1% SDS at 42°C for 30 min and autoradiographed with two DuPont Cronex III intensifying screens for 3 days. The relative intensity of each band on the autoradiographs was quantitated by laser densitometry (Biomed Instruments, Fullerton, CA).
Isolation of IGFBP-5 Genomic Clones-A mixture of two oligodeoxynucleotides, 5Ј-CCCAGGTAAGAGGAGAGGAA-3Ј and 5Ј-CTGGC-TAGAGGAGGAGACA-3Ј, corresponding to the 3Ј-untranslated region of the human IGFBP-5 gene, was labeled with [␥-32 P] ATP using T4 polynucleotide Kinase and was used to screen 6 ϫ 10 6 clones of a human cosmid (pWE15) library derived from placental DNA (CLONTECH, Palo Alto, CA). Positive clones were rescreened with a human IGFBP-5 full-length coding region cDNA probe. DNA from a selected clone was digested with restriction enzymes EcoRI, PstI, SacI, BamHI, and Hind-III, and analyzed by Southern blot using a 32 P-labeled oligodeoxynucleotide, 5Ј-ACTCTCGCTCTCCTGCCCCA-3Ј, corresponding to the 5Јuntranslated exon 1 region of human IGFBP-5. A 4.6-kilobase EcoRI fragment was gel purified and subcloned into pGEM3Zf(Ϫ) (Promega, Madison, WI). Both DNA strands of the fragment were sequenced with the dideoxy chain termination method (41). It contained 1390 bp of 5Ј-flanking region, all of exon 1, and 2.9 kilobases of intron 1. Primer extension analysis was carried out as described (42).
Reporter Plasmid Construction-Seven IGFBP-5 promoter-reporter constructs containing 753, 461, 345, 325, 252, 162, and 124 bp of IGFBP-5 5Ј-flanking region were prepared by inserting selected restriction enzyme fragments or polymerase chain reaction products into a CAT reporter gene vector pJFCAT1 (43). For pCAT162 CACCC box mutants, double-stranded oligodeoxynucleotides with the sequence corresponding to BP5-PDREm1 flanked by SphI sites were synthesized, annealed and inserted in the sense orientation into the SphI site of pCAT124 upstream of the IGFBP-5 5Ј-flanking region. The IGFBP-5 promoter fragment from Ϫ252 to ϩ24 bp was subcloned to pGEM7 and then to pGL3 (Promega, Madison, WI) to create the luciferase reporter construct pLUC-252. The sequence and orientation of all promoter constructs used were confirmed using the dideoxy method.
DNA Transfection-To determine basal promoter activities of the IGFBP-5 promoter CAT constructs, 3 ϫ 10 5 U2 cells were seeded in 60-mm dishes in DMEM supplemented with 10% CS. After 18 h of incubation, cells were rinsed once with DMEM and 1.6 ml of DMEM containing 12 l of LipofectAMINE (Life Technologies, Inc.), and 6 g of Qiagen purified DNA (5 g of CAT construct ϩ 1 g of pCMV␤-gal) was added to the cells. The cells were incubated for 4 h at 37°C, in 95% air, 5% CO 2 . Medium was then replaced with DMEM supplemented with 10% CS and cells were incubated for 48 h before ␤-gal and CAT assay. To evaluate the effect of PG on promoter activity, U2 cells were grown in phenol red-free DMEM supplemented with 5% CD-FBS for 72 h before transfection. U2 cells were plated at 2 ϫ 10 5 /dish and, in most experiments, were transfected with 2 g of phPR-A plasmid (44), 3 g of IGFBP-5 promoter construct, and 1 g of pCMV-␤-gal. After transfection, cells were incubated in phenol red-free DMEM supplemented with 2% CD-FBS for 24 h. Medium was then replaced with phenol red-free DMEM supplemented with 2% CD-FBS, plus 10 nM PG or solvent, for another 24 h. Cells were lysed with reporter lysis buffer (Promega, Madison, WI) as recommended by the manufacturer. CAT and ␤-galactosidase activities were determined as described (45,46), and CAT activity was expressed per unit of ␤-galactosidase activity. To evaluate transactivation of IGFBP-5 promoter activity in a luciferase reporter construct, cells were plated in 6-well plates at 120,000/well. All incubation, transfection, and treatment procedures were as described for the CAT reporter construct studies except that amounts of pLUC-252 and pCMV-␤-gal were 0.375 and 0.125 g/well, respectively, phPR-A (and phPR-B) were tested at 0.0625-0.50 g/well, and total DNA added was kept constant at 1.0 g/well by adding pGEM7 plasmid DNA. Luciferase activities were determined using the Enhanced Luciferase Assay Kit (Pharmingen, San Diegao, CA). Because this assay was less complex to perform than the CAT assay, experiments were done with 6 replicate wells/group. Statistical analysis of differences between groups by analysis of variance was done using the Systat computer program (Systat Inc., Evanstan, IL).
Electrophoretic Mobility Shift Assay-U2 cells were plated at 5 ϫ 10 5 cells per 100 mm dish in 5% CD-FBS for 18 h. The medium was replaced with 2% CD-FBS and incubated for 48 h. Nuclear extracts were prepared as described (47). For the EMSA, the binding reaction contained 4% glycerol, 1 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, 50 mM NaCl, 10 mM Tris-HCl, pH 7.5, 0.05 mg/ml poly(dI-dC), 1.5 l (3 g of protein) of nuclear extract, and 1 l (50,000 cpm) of 32 P-labeled doublestranded oligodeoxynucleotides in a total volume of 20 l. For cold competition experiments, nuclear extract was preincubated with unlabeled double-stranded oligodeoxynucleotides for 10 min at room temperature before the addition of labeled double-stranded oligodeoxynucleotides. The oligodeoxynucleotides used for competition with the labeled IGFBP-5 CACCC box sequence are described under "Materials." The mouse mammary tumor virus PRE oligodeoxynucleotide was a double-stranded 28-mer based on the sequence 189 -162 bp 5Ј of the transcription start site and was described previously (37).
Purified PR A was obtained from a baculovirus expression system and purified as described (31). To evaluate PR interactions with U2 cell nuclear factors, different concentrations of PR proteins were preincubated with 5 g of U2 cell nuclear extract protein before the addition of labeled double-stranded oligodeoxynucleotides. The reaction mixture was incubated at room temperature for 20 min and analyzed by electrophoresis using a 5% polyacrylamide-0.125% bis-acrylamide gel in 0.25 ϫ TBE. For supershift experiments, 1 g of anti-PR antibody PR52 (41) was added to the reaction and incubated on ice for an additional 15 min before loading onto the gel. After electrophoresis, the gel was dried and autoradiographed using Kodak X-OMAT or Biomax MS film and Cronex III screens.
Nucleotide Sequence Accession Number-The GenBank TM data base accession number of the human IGFBP-5 5Ј-flanking region determined in this study is U20271.

Effect of PG and RU486 on IGFBP-5 Levels in Normal
Human Osteoblasts-Previously we found that PG increased IG-FBP-5 mRNA levels in normal human osteoblasts and in MG63 and U2-OS osteosarcoma cells (27). In the present study we replicated this earlier finding with normal human vertebral osteoblasts and determined whether the progestin antagonist RU486 would affect PG induction of IGFBP-5 expression. Treatment with 10 Ϫ8 M PG for 4 h increased IGFBP-5 mRNA levels to 190% of control (Fig. 1). 10 Ϫ6 M RU486 by itself did not affect IGFBP-5 mRNA levels and prevented the increase in IGFBP-5 mRNA levels induced by PG. These results support the conclusion that PG increases IGFBP-5 expression by a PR-dependent mechanism.
Effect of PG on IGFBP-5 Transcription-U2 cells were selected to investigate the mechanism of PG action in the present study because both IGFBP-5 expression and transfection efficiency were higher in this cell type than in other human osteoblast-like cell lines. PG treatment for 1 h increased IGFBP-5 gene transcription determined by nuclear run-on analysis to 334% of control (p Ͻ 0.002), whereas the rate of ␤-actin gene transcription was not significantly altered (Fig. 2). After 4 h of PG treatment, the IGFBP-5 transcription rates in both the treated and control groups were elevated such that the difference between PG-treated and control groups was less compared with that found after 1 h of PG treatment (data not shown). To determine whether PG effected IGFBP-5 mRNA stability, U2 cells were treated with PG for 4 h, and thereafter RNA polymerase II-mediated transcription was inhibited with 4 ϫ 10 Ϫ5 M 5,6-dichlorobenzimidazole-1-␤-D-ribofuranoside. IGFBP-5 mRNA levels in control and PG-treated cultures were determined at 5-24 h by Northern analysis. In these experiments, IGFBP-5 mRNA half-life was estimated to be approximately 24 h and was not affected by PG (data not shown). The long IGFBP-5 mRNA half-life and absence of a PG effect on half-life are consistent with the results of nuclear run-on analysis, suggesting that PG rapidly increased IGFBP-5 mRNA levels in human osteoblasts primarily by a transcriptional mechanism.
Functional Analysis of the IGFBP-5 Gene-We isolated a pWE15 human genomic cosmid clone containing the IGFBP-5 gene and then isolated a 4.6-kilobase EcoRI fragment containing the 5Ј-flanking region. The nucleotide sequence from position Ϫ461 through the first exon was identical to that reported previously (12). The proximal 5Ј-flanking region contains a TATA box, a CAAT box, and several putative response elements. Primer extension analysis in U2 cells confirmed a single transcription start site 33 nucleotides from the TATA box (data not shown) that was reported previously in a study of human breast cancer cells (12).
To determine whether the putative IGFBP-5 promoter functioned in human osteoblasts, CAT reporter constructs were made containing a PstI fragment from Ϫ753 to ϩ23 (pCAT753) and a 484-bp HindIII/PstI fragment from Ϫ461 to ϩ23 (pCAT461). U2 cells were transiently transfected with sense reporter constructs, antisense constructs, or the pJFCAT1 promoterless control vector and then incubated for 48 h in medium with 10% CS. CAT activity in cells that were transfected with pCAT753 or pCAT461 sense constructs was 620 and 730%, respectively, of CAT activity in cells transfected with pJF-CAT1. In contrast, CAT activity did not increase in U2 cells transfected with antisense promoter constructs.
Identification of IGFBP-5 Promoter Sequences That Confer PG Responsiveness-In experiments for assessing effects of PG treatment, concentrations of steroids and other factors in serum were minimized by maintaining U2 cells for 72 h in phenol red-free DMEM plus 5% CD-FBS and then plating the cells in 5% CD-FBS at two-thirds of the cell density used in the basal promoter expression experiments. Cells were co-transfected with pCAT753 and pCMV-␤-gal, plus PR expression vector phPR-A (44), or a PR control vector from which the PR coding region was removed. Cells were incubated 48 h in phenol redfree DMEM supplemented with 2% CD-FBS in the absence or presence of 10 nM PG. In these conditions basal promoter activity from pCAT753 was 285% of activity from the promoterless control pJFCAT1. The reduction of basal promoter activity in lower density U2 cells maintained in 2% CD-FBS compared with 10% CS suggests that serum factors and cell density-dependent signals may increase promoter activity.
When U2 cells were co-transfected with the expression vector phPR-A, along with pCMV-␤-gal and pCAT753, PG reproduc-

FIG. 2. PG increases IGFBP-5 gene transcription in U2 cells.
U2 cells were treated for 1 h with 10 Ϫ8 M PG or vehicle control (Con), and nuclei were isolated. Transcription in 10 7 nuclei/group was continued in vitro in the presence of [ 32 P]UTP. 32 P-Labeled RNA was hybridized to immobilized ␤-actin cDNA, IGFBP-5 cDNA, and pBluescript vector DNA as described under "Experimental Procedures." Autoradiographs from three preparations are shown. Labeled RNA binding to target DNAs were measured by laser densitometry of the autoradiographs and are expressed as percentages of solvent-treated control Ϯ S.E. ␤-actin, 130 Ϯ 66%; IGFBP-5, 334 Ϯ 32%, p Ͻ 0.01. ibly increased IGFBP-5 promoter activity to 200 -300% of control (p Ͻ 0.001, n ϭ 5). Similar results were obtained after treatment of co-transfected cells with promegestone (R5020), a nonmetabolizable progestin analog (data not shown). Transfection with the PR control vector did not impart PG responsiveness, suggesting that the effects of phPR-A were mediated by PR A and not by the expression vector backbone. PG increased activity of the MMTV promoter pMSG-CAT to 170% of solventtreated control in cells co-transfected with phPR-A (data not shown) (48).
PG responsiveness of each deletion construct was determined in cells precultured in DMEM ϩ 5% CD-FBS, co-transfected with phPR A , pCMV␤-gal and the deletion construct, and then changed to DMEM ϩ 2% CD-FBS with either solvent or 10 nM PG. In multiple experiments, PG reproducibly increased promoter activity 200 -300% in all constructs containing from 753 to 162 bp of proximal promoter (Fig. 3). PG did not increase CAT activity in cells transfected with pCAT124, which contained only the sequence from the CAAT box to the start of transcription, although pCAT124 and pCAT162 had similar levels of basal promoter activity in the same experiments. These data suggest that none of the PRE half-sites were re-quired for PG-dependent transactivation of the IGFBP-5 promoter and that elements between Ϫ162 and Ϫ124 that contain two tandem CACCC box sequences, 5Ј-CCCCACCCCCACCC-3Ј, may impart PG responsiveness to the IGFBP-5 promoter.
The CACCC box is an important enhancer element that binds several different transcription factors (32,33,35,(52)(53)(54)(55)(56)(57)(58). CACCC-binding proteins can either bind directly to the CACCC box or interact with other transcription factors and bind as a complex (58). To determine whether CACCC sequences between positions Ϫ162 and Ϫ124 are required for PG stimulation of IGFBP-5 promoter activity, mutations were made in the distal (pCAT162mDi), proximal (pCAT162mPr), or both (pCAT162mut) CACCC sequences in pCAT162 (changing CCCACCC to aaaACCC). In multiple experiments, PG increased pCAT162 and pCAT162mDi promoter activities but failed to increase pCAT162mut and pCAT162mPr promoter activities (Fig. 4). Thus mutation of the proximal CACCC box at position Ϫ139 abolished PG transactivation of hIGFBP-5 promoter activity. These data provide strong evidence that the proximal CACCC box at position Ϫ139 is required for PG-dependent transactivation of IGFBP-5 promoter activity.
Interaction of IGFBP-5 Promoter CACCC Box Sequences with Nuclear Extract Proteins-To determine whether U2 cell nuclear proteins interact with the CACCC box motif of the IG-FBP-5 promoter, EMSA were performed. 32 P-Labeled doublestranded synthetic oligodeoxynucleotides corresponding to the wild type IGFBP-5 promoter sequence from Ϫ155 to Ϫ128, designated the IGFBP-5 progesterone-dependent responsive element (BP5-PDRE), were incubated with U2 cell nuclear extract proteins. Analysis by EMSA resulted in a reproducible set of shifted bands (Fig. 5A). Unlabeled BP5-PDRE oligodeoxynucleotide or different oligodeoxynucleotides containing CACCC box core motif sequences from enhancers of other promoters were used to compete for binding to labeled BP5-PDRE. These enhancers were RCE sequences from the TGF-␤ and c-fos promoters (32,33) and the MNF-binding site sequence from the myoglobin promoter (34). Other competitors tested were the consensus AP-2-binding site sequence from the metallothionein-IIA promoter (35) and the consensus Sp1-binding site sequence from Simian virus 40 (36) because Sp1 and AP2 have been shown to bind to nonconsensus sequences containing CACCC boxes (52,57,59). As shown in Fig. 5A, 50-fold excess Progesterone inductions of promoter activity in cultures co-transfected with each IGFBP-5 promoter construct, phPR-A and pCMV-␤-gal, and then changed to CD-FBS and treated with solvent or 10 Ϫ8 M PG are shown in the second column, expressed as percentages of the respective solvent control group. All data are based on normalizing CAT to ␤-galactosidase activities, and are means Ϯ S.E. of at least three independent experiments for each construct. There was no significant difference in fold induction of CAT expression by PG among groups pCAT753-pCAT162 as determined by analysis of variance. PG induction of CAT expression in the pCAT124 group was not significant and was significantly lower than in all other groups (p Ͻ .001). Basal promoter activities (minus PG) in pCAT162 and pCAT124 were the same, approximately 3-fold of pJFCAT1 control.
FIG. 4. Mutation of CACCC boxes eliminates PG induction of IGFBP-5 promoter activity. U2 cells were transfected as described under "Experimental Procedures" with phPR-A, with pCMV␤-gal, and with pCAT124, wild type pCAT162, or mutant pCAT162 promoterreporter constructs. In pCAT162mut both CACCC boxes at positions Ϫ147 to Ϫ134 were mutated from CCCACCCCCACCCC to aaaAC-CaaaACCCC. Only the distal CACCC box was mutated to aaaAC-CCCCACCCC in pCAT162mDi, and only the proximal CACCC box was mutated to CCCACCaaaACCCC in pCAT162mPr. Transfected cells were incubated in 2% CD-FBS in the presence or absence of 10 Ϫ8 M PG. CAT activity was normalized to ␤-galactosidase activity and expressed as a percentage of respective control group (transfected cells incubated without PG), mean Ϯ S.E. of three independent experiments. PG significantly increased CAT activity in pCAT162-and pCAT162mDi-transfected cells (p Ͻ 0.001) but did not significantly affect CAT activities in pCAT162mut, pCAT162mPr, or pCAT124-transfected cells.
unlabeled BP5-PDRE specifically competed with labeled BP5-PDRE for binding to U2 cell nuclear proteins and eliminated bands 1-4. At a 50-fold excess molar ratio of unlabeled DNA to labeled BP5-PDRE DNA, consensus AP-2, Sp-1, MNF, and the RCE sequence from the TGF-␤ promoter competed as effectively as unlabeled BP5-PDRE for binding in complexes 1, 3, and 4 (Fig. 5A). AP-2, Sp-1, and c-fos RCE sequences did not compete as effectively as BP5-PDRE, MNF, or TGF-␤ RCE sequences for protein binding in complex 2. Competition with 25-, 50-, and 100-fold molar excess of BP5-PDRE and MNFbinding sequences are compared in Fig. 5B. Competition for binding to band 5 was variable and may be nonspecific. These results suggest that nuclear protein(s) in the band 2 complex are specific for the CCCCACCCC sequence, which is identical in the BP5-PDRE, MNF, and TGF-␤ RCE sequences and different in the other sequences that were tested. Binding of nuclear protein(s) in bands 1, 3, and 4 may be more specific for sequences adjacent to CCCCACCCC.
When CCCACCC sequences at positions Ϫ147 and Ϫ139 were mutated to aaaACCC (BP5-PDREm1), this mutant oligodeoxynucleotide did not compete with labeled BP5-PDRE oligodeoxynucleotides (Fig. 6). However, when the distal CACCC box alone was mutated from CCCACCC to aaaACCC (BP5-PDREm2), the unlabeled BP5-PDREm2 competed very effectively with labeled BP5-PDRE, supporting the conclusion that only the proximal CACCC box at Ϫ139 is required for PG-dependent transactivation of the IGFBP-5 promoter. To further determine which of the tandem CACCC boxes in the IGFBP-5 promoter is required for binding to nuclear factors, single base pair mutations from CCCACCC to CCaACCC were made. Mutation of both distal and the proximal CACCC sequences or mutation of only the proximal CACCC sequence (BP5-PDREm3 and BP5-PDREm5) eliminated the ability of the sequence to compete with wild type labeled BP5-PDRE for nuclear factor binding (Fig. 6). However, when the distal CACCC sequence alone was mutated (BP5-PRDEm4), this oligodeoxynucleotide competed very effectively with labeled BP5-PDRE for nuclear factor binding. These results suggest that the distal CACCC sequence at Ϫ147 may not be important for forming the complexes observed by EMSA. Furthermore, these data provide evidence that the cytosine at position Ϫ139 in the proximal CACCC box is important for the binding of U2 cell nuclear protein(s).
PR A Interation with Nuclear BP5-PDRE-binding Protein(s)-Deletion of the region Ϫ162 to Ϫ124 and mutation of the CACCC sequences within this region abolished PG responsiveness of the IGFBP-5 promoter, and the BP5-PDRE oligodeoxynucleotide corresponding to this region of the promoter bound to U2 cell nuclear factors. These results suggested that PR A could either directly bind to the BP5-PDRE or interact with nuclear protein(s) that bind to the BP5-PDRE. To determine whether PR A could bind to the IGFBP-5-PDRE sequence at Ϫ139 we used EMSA to examine binding of baculovirusexpressed, purified human PR A to labeled BP5-PDRE oligodeoxynucleotide. As a positive control the PR A preparation was first tested with a consensus progesterone response element oligodeoxynucleotide (MMTV-PRE) from the mouse mammary tumor virus promoter (30) and the anti-PR monoclonal antibody, AB52 (31). Purified PR A binding to labeled MMTV-PRE was dependent on the presence of nuclear extract proteins, as observed previously (31), and produced a strong mobility shift band, which was supershifted by incubating the protein-DNA complexes with the PR antibody (Fig. 7A). Purified PR A protein alone did not bind the BP5-PDRE oligodeoxynucleotide, which is similar to the result with the MMTV-PRE. Incubation of labeled BP5-PDRE with 5 g of U2 cell nuclear protein alone produced multiple shifted bands (Fig. 7B), as in Fig. 4. When increasing amounts of PR A were preincubated with nuclear extract proteins and then mixed with labeled BP5-PDRE, the intensity of complexes 1 and 3 progressively decreased, whereas the intensity of complex 4 progressively increased (Fig. 7B). Addition of anti-PR antibody together with PR A and nuclear extract proteins did not alter the gel shift pattern of labeled BP5-PDRE, suggesting that either PR A did not participate in direct binding to BP5-PDRE or that the PR A epitope to which the monoclonal antibody binds was masked by the nuclear proteins. As a negative control for the effects of PR A on the BP5-PDRE gel shift pattern, PR A was added with nuclear extract to labeled Sp1 and TGF-␤ RCE oligodeoxynucleotides, which have similar CACCC box sequences. PR A had no effect on the gel shift patterns with these sequences (data not shown).
PR A and PR B Mediate PG Induction of IGFBP-5 Promoter Activity-Initial experiments comparing abilities of PR A and PR B isoforms to transactivate the IGFBP-5 promoter in CAT reporter constructs suggested that PR B was inactive (data not shown), so experiments to define the regions of the promoter that mediate transactivation were limited to PR A . This apparent specificity for PR A could not be confirmed, however, using different phPR-B plasmid preparations and a luciferase reporter construct containing 252 bp of IGFBP-5 5Ј-flanking region. U2-OS cells were co-transfected with 0.0625-0.50 g/well of PR expression plasmids phPR-A and phPR-B together with pCMV-␤-gal and pLUC-252. With this luciferase reporter construct, both PR A and PR B transactivated IGFBP-5 promoter activity in the presence of PG, and PR B was more effective at higher transfected DNA levels (Fig. 8). PG-dependent PR B transactivation of the MMTV promoter was much higher than PR A transactivation of this promoter (data not shown), as described previously (48). We also found that RU486 inhibited PR A and PR B transactivation of the IGFBP-5 promoter, which correlates with the effects of RU486 to decrease IGFBP-5 mRNA levels in normal human osteoblast cultures. DISCUSSION The human IGFBP-5 promoter does not contain classic consensus palindromic progesterone or glucocorticoid response elements (GRE/PRE), defined as two hexameric half-sites with a 3-bp spacer (60). However the IGFBP-5 promoter contains five putative GRE/PRE half-sites that have strong sequence similarity to the hexanucleotide motifs 5Ј-TGTTCT-3Ј of the MMTV promoter and 5Ј-TGTTCA-3Ј of the uteroglobin gene promoter (50,60). Although GR and PR do not bind to GRE/PRE halfsites as effectively as to the palindromic GRE/PRE consensus sequence, clusters of GRE/PRE half-site sequences in MMTV, uteroglobin, and other promoters have been shown to bind the receptors and effectively mediate ligand-dependent transactivation through synergistic interactions between the multiple half-sites (50,60). Furthermore, PR as well as other steroid hormones receptors affect transcription of many genes without directly binding to consensus hormone response elements (61)(62)(63)(64)(65)(66). Many of these actions of nuclear receptors involve interactions with co-activators and co-repressors (67,68). Although the presence of PRE half-sites in the IGFBP-5 promoter suggested the possibility that they might mediate PG transactivation, our results demonstrate that this is not the case. Rather, results of the series of reporter construct IGFBP-5 promoter deletions suggest that a repeated CACCC box motif (BP5-PDRE) proximal to the PRE half-sites mediates PG transactivation.
The CACCC box is an important promoter element required for efficient and accurate gene expression in the TGF-␤, c-fos, c-jun, and IGF I genes (32,33,52,59). In addition, a number of transcription factors including the RCE-binding protein, MNF, and AP2 bind to CACCC box sequences and effect transcription (35,54). CACCC-binding proteins can either bind directly to the CACCC box sequence or interact with other transcription factors and bind as a complex (58). Duan and Clemmons (14) reported that human IGFBP-5 promoter activity is increased by AP-2. They found that the two overlapping AP-2-binding sites in the CACCC box motif at positions Ϫ162 to Ϫ124 were not required for AP-2 stimulation of promoter activity in human hepatoma cells and skin fibroblasts. Rather, an alternate consensus AP-2-binding sequence 5Ј-GCCNNNGGC-3Ј at positions Ϫ52 to Ϫ35, adjacent to the TATA box, mediated AP-2 transactivation. Both of the sequences bound purified AP-2 protein in EMSA experiments even though only the sequence adjacent to the TATA box was involved in AP-2 transactivation of promoter activity. In contrast, we found that PR A transactivation of the IGFBP-5 promoter in osteoblastic cells required the AP-2-binding sequence at position Ϫ139. This suggests that AP-2 and PR A transactivate the IGFBP-5 promoter by different mechanisms.
Although EMSA analysis suggested that purified PR A did not bind to BP5-PDRE directly, addition of PR A altered the gel shift band pattern produced by binding of nuclear extract pro- teins to the BP5-PDRE DNA sequence. This result suggests that PR A interacted with BP5-PDRE-binding proteins, although the identities of the binding proteins are not known. Of the CACCC box sequences tested in competition EMSA experiments, the RCE and MNF-binding sites of the TGF-␤ and myoglobin promoters (33,34) are most similar to the BP5-PDRE sequence, and the RCE and MNF sequences most effectively competed with the BP5-PDRE in binding nuclear extract proteins. These results, although not definitive, are consistent with the possibility that RCE, MNF, or related proteins may be involved in binding BP5-PDRE.
Our results indicate that both PR A and PR B can mediate PG induction of hIGFBP-5 promoter activity. For many genes that are induced by progestins, human PR B is more active than human PR A (44,69,70), although PR A has been reported to be a stronger transactivator than PRB for chicken ovalbumin and human tyrosine aminotransferase genes (44,48). In many PGinducible genes that have been examined, activated PR B increases transcription by binding to a PRE within the gene promoter, and PR A functions as a negative regulator without directly binding to a PRE (44,69). It remains to be established whether PR B mediates transactivation of the IGFBP-5 promoter by a mechanism that is similar to or different from that of PR A (70).
PR A and PR B are differentially expressed in a cell type-and species-specific manner (71,72). Depending on the cell type, either both isoforms are induced by estrogen or PR B is differentially induced (73)(74)(75). In human osteoblasts both isoforms of PR are expressed, and estrogen induced both A and B promoters (74). Progestins have been found to stimulate osteoblast proliferation, differentiation, and growth factor expression (25, 26, 76 -80) and to increase bone formation (81,82). In ovariectomized dogs, PG increased bone formation (83) and in postmenopausal women progestins combined with estrogen increased bone density to a greater extent than estrogen alone (84,85). Because IGFBP-5 increases the mitogenic and anabolic activities of IGFs and IGF activities are important for bone formation, one could speculate that PG induction of IGFBP-5 expression in osteoblasts contributes to the positive effects of PG on bone formation.