The Osteoblast-specific Transcription Factor Cbfa1 Contributes to the Expression of Osteoprotegerin, a Potent Inhibitor of Osteoclast Differentiation and Function*

Bone formation and resorption are tightly coupled under normal conditions, and the interaction of osteoclast precursors with cells of the osteoblast lineage is a prerequisite for osteoclast formation. Cbfa1 is an osteoblast-specific transcription factor that is essential for osteoblast differentiation and bone formation. At present, it is not known whether Cbfa1 regulates any of the osteoblast-derived factors involved in the bone resorption pathway. Osteoprotegerin (OPG) is an osteoblast-secreted glycoprotein that functions as a potent inhibitor of osteoclast differentiation and bone resorption. Cloning and computer analysis of a 5.9-kilobase human OPG promoter sequence revealed the presence of 12 putative Cbfa1 binding elements (osteoblast-specific element 2 (OSE2)), suggesting a possible regulation of OPG by Cbfa1. We cloned the promoter upstream of the β-galactosidase reporter gene (pOPG5.9βgal) and evaluated whether Cbfa1 could regulate its expression in transient transfection assays. The 5.9-kilobase promoter directed increased levels of reporter gene expression, reminiscent of OPG protein levels in osteoblastic cell lines (BALC and U2OS) as compared with the nonosteoblastic cell line COS1. Cotransfection of a Cbfa1 expression construct along with pOPG5.9βgal reporter construct led to 39-, 7-, and 16-fold increases in β-galactosidase activity in COS1, BALC, and U2OS cells, respectively. Removal of all the putative OSE2 elements led to an almost complete loss of transactivation. Mutational analysis demonstrated that the proximal OSE2 element contributes to a majority of the effects of Cbfa1, and Cbfa1 bound to the proximal element in a sequence-specific manner. Further, overexpression of Cbfa1 led to a 54% increase in OPG protein levels in U2OS cells. These results indicate that Cbfa1 regulates the expression of OPG, thereby further contributing to a molecular link between bone formation and resorption.

rived from mesenchymal stem cells, and their formation and function are under the control of an osteoblast-specific transcription factor known as core binding factor a1 (Cbfa1) or osteoblast specific factor 2 (Osf2) 1 (1)(2)(3). Cbfa1 has been shown to regulate the expression of genes that characterize the osteoblast phenotype, including osteocalcin, osteopontin, type I collagen, bone sialoprotein, and collagenase-3, by binding to DNA sequence elements called OSE 2 that are present in the control regions of these genes (1, 4 -6). Cbfa1 knock out animals completely lack bones because of a maturational arrest of osteoblasts and die soon after delivery (2,3). Further, haploinsufficiency of the Cbfa1 gene product in mice and humans heterozygous for the Cbfa1 gene results in a condition called cleidocranial dysplasia that is characterized by delayed ossification (2,3,(7)(8)(9)(10).
Osteoclasts that are primarily involved in bone resorption are multinucleated cells derived from hematopoietic precursors of the monocyte/macrophage series. It has been known for a long time that osteoclast precursors need to interact with cells of the osteoblast lineage to differentiate into mature osteoclasts (11)(12)(13)(14). Recently, proteins involved in this interaction have been identified and are now being extensively characterized (15)(16)(17)(18). They include RANK (receptor activator of NF-B) ligand (15,18), also known as osteoclast differentiation factor or TRANCE (tumor necrosis factor-related activation-induced cytokine) (19). RANK ligand is a membrane-bound protein of the tumor necrosis factor ligand family that is expressed on the osteoblast cell surface and has been shown to play a major role in osteoclast differentiation along with macrophage colonystimulating factor (20). RANK ligand binds to its receptor RANK (16) on hematopoietic cells and initiates a cascade of signaling events that leads to osteoclast differentiation. Furthermore, stromal/osteoblastic cells also secrete a glycoprotein called osteoprotegerin (OPG) (17), a soluble member of the tumor necrosis factor receptor superfamily that acts as a decoy receptor for RANK ligand and prevents its interaction with the cognate receptor RANK (21,22). OPG has been shown to be a potent inhibitor of osteoclast differentiation, survival and function in vitro and bone resorption in vivo (23)(24)(25). OPG knock out mice show severe early onset osteoporosis (24,26), whereas transgenic animals overexpressing OPG are osteopetrotic (17).
While the role of Cbfa1 in osteoblast differentiation and bone formation is fairly well understood, it is not known whether it can modulate the expression of regulators of osteoclast differ-entiation and function, namely RANK ligand and OPG, that are generated by cells of the osteoblast lineage. To search for any potential role of Cbfa1 in regulating osteoclast formation, we directly examined the effect of Cbfa1 on OPG gene transcription by cloning a 5.9-kb region of the human OPG promoter and analyzed its transcriptional activity in cotransfection experiments. We present evidence that Cbfa1 increases OPG gene transcription in osteoblastic and nonosteoblastic cells, via sequence-specific binding and transactivation mediated by the OSE 2 elements and that Cbfa1 overexpression enhances OPG protein levels in osteoblastic cells.

EXPERIMENTAL PROCEDURES
Cloning of the Human OPG Promoter-To clone the human OPG promoter, two approaches were combined to identify and isolate an approximately 5.9-kb genomic fragment located immediately 5Ј to the coding region of the OPG gene. The first approach involved using the Genome Walker kit (CLONTECH, Palo Alto, CA). This is a PCR-based method used to "walk" genomic DNA adjacent to known sequences. Gene-specific primers were designed from the published OPG sequence (GenBank TM accession number U94332) (27) as follows: GSP1 5Ј-gga gat gtc cag aaa cac gag cgc gca g-3Ј and nested GSP2 5Ј-cac agc aac ttg ttc att gtg gtc ccc-3Ј. These primers were used in combination with primers specific to the Genome Walker library adapter to amplify 1874-, 1473-, and 391-bp fragments of DNA upstream of the OPG coding region. Sequence analysis demonstrated that these were overlapping proximal fragments of the 5Ј-flanking region of the OPG gene.
To clone larger promoter fragments containing more distal sequences we next screened a conventional genomic library. To do this, the fulllength OPG cDNA (GenBank TM accession number U94332) was sent to Genome Systems (St. Louis, MO) and used to screen a human P1 library. The screen yielded three positive clones, each containing 70 -100 kb of genomic DNA. To determine which P1 clone contained the promoter, each clone was digested with a panel of restriction enzymes and then Southern blotted with the 1874-bp promoter fragment described above. Clone 19401 digested with SstI and Eco47III yielded a 5936-bp band that hybridized to the 1874-bp OPG promoter fragment. This 5936-bp fragment was then gel purified and cloned.
For site-directed mutagenesis of the proximal OSE 2 element, the OSE 2 core element (AACCTCA) (OSE 2 element 1, Ϫ309 to Ϫ303; see Table I and Fig. 1) in the pOPG0.9␤gal construct (in the p␤gal-Basic vector backbone) was substituted with a random sequence of 7 nucleotides (AGATATC; the EcoRV recognition site is underlined) using the two-step PCR strategy (28). The mutant primers used for the two first step PCR reactions were 5Ј-ctc atc aat gta tct tat gg-3Ј (p␤gal-F: vector primer) with 5Ј-gct gtc tcc gcg ggg ctc gat atc ttc ccg gcc cct tcc cgc c-3Ј (OSEmutRev) and 5Ј-ggc ggg aag ggg ccg gga aga tat cga gcc ccg cgg aga cag cag ccg-3Ј (OSEmutFor) with 5Ј-gtc aaa gta aac gac atg-3Ј (p␤gal-R: vector primer). The second step PCR reaction was performed with flanking primers (p␤gal-F and p␤gal-R), using the first step PCR products as template. To generate the OSE 2 mutant construct (pOPG0.9OSEmut␤gal), the second step PCR product was digested with Asp718 and BglII and was ligated to p␤gal-Basic vector containing Asp718 and BglII ends. To generate the OSE 2 mutation in pOPG0.4␤gal, the 0.4-kb region containing the mutation was PCR amplified from pOPG0.9OSEmut␤gal using the primers 5Ј-ata ggt acc gcc cag ccc tcc cac cgc tgg t-3Ј (OPG392AspF) and p␤gal-R. The PCR product was digested with Asp718 and BglII and ligated to p␤gal-Basic vector that was digested with the same enzymes. To create a substitution mutation in the proximal OSE 2 element in pOPG5.9␤gal, a XhoI fragment was removed from pOPG5.9␤gal and replaced with the XhoI fragment from pOPG0.9OSEmut␤gal. Deletion of the proximal OSE 2 element (pOPG 0.3␤gal) was achieved by subcloning the 312-bp SacII-BglII fragment from pOPG0.4␤gal into p␤gal-Basic.
For the addition of OSE 2 element(s) to pOPG0.2␤gal, either one or three copies of the consensus OSE 2 sequence from the osteocalcin promoter was included in the 5Ј (forward) primers (1XocOSEOPG primer: 5Ј-ata tgg tac cgc tgc aat cac caa cca cag cgg atc ctt tcc gcc cca gcc ctg a-3Ј; 3XocOSEOPG primer: 5Ј-ata tgg tac cgc tgc aat cac caa cca cag cgc tgc aat cac caa cca cag cgc tgc aat cac caa cca cag cgg atc ctt tcc gcc cca gcc ctg a-3Ј), and the 0.2-kb OPG promoter fragment was amplified using one of the forward primers and p␤gal-R reverse primer, from pOPG0.2␤gal template. The resultant PCR products were digested with Asp718 and BglII and ligated to p␤gal-Basic vector that was digested with the same enzymes, to generate the 1xOSE-pOPG0.2␤gal and 3xOSE-pOPG0.2␤gal plasmids.
pEF/myc/cyto (control vector) was purchased from Invitrogen, Carlsbad, CA. It contains an SV40 origin of replication, the elongation factor EF1␣ promoter, and a Myc epitope tag coding sequence at the 3Ј end of the multiple cloning site. The region of Cbfa1 that encodes the protein isoform starting with the amino acids MASNS and ending with VWRPY (Osf2Met 69 ) (29) was PCR amplified from pCMV5-Osf2 (obtained from Dr. Gerard Karsenty, Houston, TX) (1). The PCR product was then subcloned into the NcoI and XhoI sites of pEF/myc/cyto to generate pEF-Cbfa1. Cbfa1 is expressed from this plasmid as a fusion protein with a Myc epitope tag at the C-terminal end. The integrity of all plasmid constructs were confirmed by restriction mapping and automated DNA sequencing.
Sequence Analysis-The sequence of the 5.9-kb OPG promoter was analyzed for the presence of consensus transcription factor binding sites, including OSE 2 (Cbfa1-binding site) and OSE 2 -like elements, using the GCG Wisconsin Package, Genetics Computer Group, Inc. (Madison, WI).
Cell Culture and DNA Transfection-The monkey kidney cell line COS-1 was obtained from American Type Culture Collection (ATCC CRL 1650) and was grown in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum (FBS) (Hyclone Laboratories, Logan, UT) and 2 mM L-glutamine (Life Technologies, Inc.). COS1 cells were chosen for use in transient transfection studies owing to their ability to initiate replication and generate multiple copies of transfected plasmids that contain the SV40 origin of replication, eventually resulting in the synthesis of substantial amounts of the protein of interest (30). BALC, a mouse calvaria-derived stromal/osteoblastic cell line (31) was grown in RPMI 1640, supplemented with 5% FBS and 2 mM L-glutamine; the human osteoblast-like osteosarcoma cell line U2OS was grown in McCoy's 5A medium supplemented with 10% FBS and 2 mM L-glutamine; and ROS 17/2.8 osteoblast-like osteosarcoma cells were grown in Ham's F-12 medium supplemented with 10% FBS. All cultures were maintained at 37°C in a humidified atmosphere of 95% air and 5% CO 2 . For Cbfa1 transactivation studies, 1 ϫ 10 5 cells were plated per well in 6-well plates and incubated for 24 h. Cells were then transfected with 1 g each of the reporter (OPG promoter constructs linked to ␤gal or the control reporter vector p␤gal-Basic) and the effector plasmid (pEF-Cbfa1 or the control expression vector pEF/myc/cyto), using Fugene transfection reagent (Roche Molecular Biochemicals). The constructs (1 g each in a total volume of 20 l in Tris-EDTA buffer, pH 8.0) were mixed with diluted Fugene reagent (194 l of serum-free medium ϩ 6 l of Fugene) and incubated for 15 min at room temperature. The DNA-Fugene mix was then added dropwise to the plates, and the cells were incubated for an additional 48 h. Following transfection, the plates were washed twice with PBS (Life Technologies, Inc.) and then lysed with 100 l of lysis buffer provided with the ␤gal reporter gene assay kit (Roche Molecular Biochemicals). The cell extracts were spun down for 2 min at 14,000 rpm in a microcentrifuge to precipitate cellular debris. 20 l of the supernatant was transferred to white opaque microtiter plates, and the ␤gal activity was measured as per the manufacturer's instructions, using an automated injection MLX Luminometer (Dynex Corporation, Chantilly, VA). As a positive control and to verify transfection efficiency, separate plates were transfected with a ␤gal expression plasmid (p␤gal-Promoter, CLONTECH, Palo Alto, CA) that has the ␤gal reporter gene coding region under the control of the SV40 early promoter (referred to as SV40-␤gal in the text and in Fig. 4). This was done to avoid possible squelching of factors that could arise when cotransfecting multiple plasmids (32,33). The luciferase reporter vector was not used because of our recent finding that cryptic enhancer elements in the luciferase reporter vector pGL3-Basic could mediate transactivation by Cbfa1, leading to spurious background luciferase expression (34). All the experiments were repeated at least three times in triplicate wells, using different plasmid preparations. The ␤gal activity values represent the integral value of light emitted over a period of two seconds and are expressed as fold induction over basal (control vector-transfected) levels. Results were analyzed using Student's t test, and probability (p) values of less than 0.05 were considered statistically significant.
To analyze the effect of Cbfa1 transfection on OPG secretion, U2OS cells were plated at 1 ϫ 10 5 cells/well in 6-well plates and allowed to grow for 24 h. The culture medium was then removed, cells were washed with 1ϫ PBS (Life Technologies, Inc.) and replaced with medium containing 0.1% FBS and incubated overnight. The next morning, cells were transfected with 1 g of either empty vector (pEF/myc/cyto) or pEF-Cbfa1 using Fugene reagent as described above. 5.5 h after transfection, the medium was removed, plates were washed with 1ϫ PBS, and complete medium was added and incubated for 4 h. The cells were then serum-starved (in medium with 0.1% FBS) and incubated further. 48 h after transfection, the supernatant was collected, and the amount of secreted OPG was quantified in an ELISA as described below.
Nuclear Extract Preparation and DNA Binding Assays-Nuclear extracts were prepared from ROS 17/2.8 cells that were grown in 10-cm tissue culture plates, following the published protocol (35). Using the same protocol, nuclear extracts were also prepared from COS1 cells that were transfected with 10 g of empty vector (pEF/myc/cyto) or the Cbfa1 expression construct (pEF-Cbfa1) using Fugene transfection reagent and incubated for 48 h. DNA binding reactions were performed in a buffer containing 5% glycerol, 100 mM NaCl, 50 mM Tris-HCl (pH 7.5), 0.1% Nonidet P-40, 2 mM EDTA, 1 mM dithiothreitol, and 2 g of poly(dI⅐dC)⅐poly(dI⅐dC). 5 fmol of 32 P-labeled double stranded OSE 2 oligonucleotides from either the osteocalcin promoter (5Ј-GAT CCG CTG CAA TCA CCA ACC ACA GCA-3Ј, OCwt) (36) or the OPG promoter (proximal OSE 2 : 5Ј-GAT CCG CCG GGA AAC CTC AGA GCC CCA-3Ј, OPGwt; or proximal OSEmut: 5Ј-GAT CCG CCG GGA AGA TAT CGA GCC CCA-3Ј, OPGmut) were incubated with the nuclear extracts for 10 min at room temperature to enable protein-DNA complex formation and then electrophoresed on a 5% polyacrylamide gel as described (36).
Immunoblot Analysis-COS1 cells were transfected with 1 g of either empty vector (pEF/myc/cyto) or pEF-Cbfa1 using Fugene transfection reagent. 48 h after transfection, the plates were washed twice with PBS, and the cells were lysed with 100 ml of lysis buffer (Roche Molecular Biochemicals). 20 l of the lysate was electrophoresed on a 5-12% NuPAGE Bis-Tris gel (NOVEX, San Diego, CA), transferred onto a Nitrocellulose membrane, and then probed with a 1:5000 dilution of mouse monoclonal ␣-Myc antibody (directed against the Myc epitope tag at the C-terminal end of Cbfa1-Myc fusion protein that is encoded by pEF-Cbfa1). Horseradish peroxidase-conjugated anti-mouse IgG (1: 1500 dilution) was used as a secondary antibody, followed by ECL-Plus detection (Amersham Pharmacia Biotech).
OPG ELISA-The amount of OPG secreted into the cell culture medium was determined using an ELISA. Cells (COS1, BALC, and U2OS cells) were plated at a density of 1 ϫ 10 5 cells/well in 6-well plates, in complete medium containing serum, and the culture media were collected after 48 h. Sandwich ELISAs were performed in 96-well Immulon4 plates (Dynex Technologies, Inc., Chantilly, VA) using standard procedures. In brief, wells were coated with 100 l of rabbit polyclonal IgG (5 g/ml) directed against recombinant human OPG (37) and then blocked with 200 l of 1% casein in PBS/Tween 20 for 1 h at room temperature. The plates were then incubated with 100 l of the cell supernatant for 1 h at room temperature, followed by three washes with PBS/Tween 20. Then the plates were incubated with 100 l of biotinylated ␣-OPG IgG (1:1000 to 1:2000 dilution) for 1 h at room temperature. After washing, the plates were incubated with streptavidin-horseradish peroxidase conjugate (1:10,000 dilution). The plates were then washed with PBS/Tween 20, and incubated with 3,3Ј,5,5Ј-tetramethyl benzidine substrate solution for 15 min at room temperature. The reaction was stopped with 1 M phosphoric acid, and the absorbance was determined at 450 nm using a SpectraMAX250 microplate spectrophotometer (Molecular Devices, Sunnyvale, CA). The amount of OPG in the medium was extrapolated from a standard curve generated using recombinant human OPG. The ELISA was sensitive enough to detect picogram amounts of OPG in the culture medium (linear range of the assay: 1-1000 pg/ml). The polyclonal ␣-OPG IgG that we used was able to detect OPG protein from human, rat, and mouse sources.

RESULTS
Cloning of the 5.9-kb OPG Promoter and Analysis of Putative Transcription Factor Binding Sites in the Promoter-We cloned and sequenced a 5.9-kb fragment of the 5Ј-flanking region of the human OPG gene (Ϫ5917 to ϩ19) as outlined under "Experimental Procedures." Computer analysis of the 5.9-kb sequence schematically summarized in Fig. 1 revealed the presence of a putative TATA box (at position Ϫ27), three CCAAT box sequences (at positions Ϫ669, Ϫ752, and Ϫ826), and consensus binding sites for a variety of transcription factors, including AP1, AP2, and SP1, as described in the previously published 1.1-kb promoter sequence (27). In addition, 12 putative binding sites for the osteoblast-specific transcription factor Cbfa1 (OSE 2 elements) were also identified ( Fig. 1 and Table I) (36,38). These putative OSE 2 elements were identical or similar to the ones found in the regulatory regions of several known targets of Cbfa1 that are expressed abundantly in osteoblasts, namely osteocalcin, osteopontin, type I collagen, and bone sialoprotein (1,6).
The basal transcriptional activity directed by the 5.9-kb region was analyzed by its ability to drive the expression of a ␤gal reporter gene after transient transfection into the following cell lines: U2OS, a human osteoblast-like osteosarcoma cell line; BALC, a mouse calvaria-derived stromal/osteoblastic cell line that is known to efficiently support osteoclast differentiation (31,39); and the nonbone cell line COS1, a SV40 immortalized monkey kidney cell line (30). As shown in Fig. 2A, the 5.9-kb promoter fragment directed a high level of expression of the ␤gal reporter gene in U2OS cells. The expression level was the highest (16-fold higher than that of the promoter-less reporter vector p␤gal-Basic) in U2OS cells and was very low in COS1 and BALC cells. To further ensure that promoter expression reflects endogenous OPG expression, we evaluated the amount of OPG protein produced (and secreted into the medium) in these cell lines using an ␣-OPG antibody-based ELISA. As shown in Fig. 2B, high levels of OPG were detected in the culture medium in U2OS cells, and only very low levels were present in BALC and COS1 cell media.
Transactivation of the 5.9-kb OPG Promoter by Cbfa1-The presence of several putative OSE 2 elements in the 5.9-kb OPG promoter, combined with the essential role of Cbfa1 in osteoblast development and the physiologically observed link between cells of the osteoblast lineage and osteoclastogenesis, prompted us to analyze whether OPG is one of the targets of Cbfa1. To test the ability of Cbfa1 to transactivate the OPG promoter, COS1, BALC, and U2OS cells were cotransfected with pOPG5.9-␤gal construct along with either the Cbfa1 expression construct (pEF-Cbfa1) or the empty vector (pEF/myc/ cyto). At first, we determined whether the Cbfa1 protein was synthesized from pEF-Cbfa1, by performing immunoblot analysis of transfected COS1 cell extracts using a monoclonal ␣-Myc antibody. As shown in Fig. 3A, substantial amounts of Cbfa1 expression was detected in cells transfected with pEF-Cbfa1, and the protein was capable of binding to the OSE 2 element (see Fig. 9, lane 8). Cotransfection of pEF-Cbfa1 and pOPG5.9-␤gal led to a 39-fold increase in OPG promoter activity in COS1 cells, a 7-fold increase in BALC cells, and a 16-fold increase in U2OS cells compared with pEF/myc/cyto control (Fig. 3B). This transactivation of the OPG promoter by Cbfa1 is consistent with the presence of functional consensus OSE 2 sites in the OPG promoter, or it could reflect an indirect effect of Cbfa1 via other sites in the promoter.
Effect of 5Ј Deletions in the OPG Promoter on Base-line Expression-To assess the relative contribution of the different OSE 2 elements, we made sequential 5Ј deletions of the promoter and obtained six different deletion constructs that pres-

FIG. 2. Correlation between the transcriptional activity of the 5.9-kb OPG promoter and the expression of endogenous OPG in COS1, BALC, and U2OS cells.
A, basal levels of expression directed by the 5.9-kb OPG promoter. The 5.9-kb OPG-␤gal construct was transfected into COS1, BALC, and U2OS cell lines, and the ␤gal activity in cell extracts was measured 48 h after transfection. Three independent transfection experiments were done in triplicate, and the means Ϯ S.E. of ␤gal activity from a representative experiment are shown. B, levels of OPG secreted into the culture medium in COS1, BALC, and U2OS cultures. 1 ϫ 10 5 cells were plated per well in 6-well plates and incubated for 48 h. The amount of OPG secreted into the medium was quantified in a sandwich ELISA using biotinylated ␣-OPG polyclonal antibody as described under "Experimental Procedures."

FIG. 3. Transactivation of the 5.9-kb OPG promoter by Cbfa1 in osteoblastic and nonosteoblastic cells.
A, immunoblot analysis of Cbfa1 expression in transfected COS1 cells. COS1 cells were transfected with either empty vector (pEF/myc/cyto) or pEF-Cbfa1, and the expression of the Cbfa1-Myc fusion protein was detected by immunoblot analysis using a monoclonal ␣-Myc antibody. B, the 5.9-kb OPG-␤gal construct was cotransfected into COS1, BALC, and U2OS cells, along with either the Cbfa1 expression construct (pEF-Cbfa1) or the empty vector (pEF/myc/cyto) using Fugene6 TM transfection reagent. Values represent the fold-induction of ␤-gal activity (means Ϯ S.E.) in Cbfa1transfected cell extracts compared with that in empty vector-transfected cell extracts. The fold induction value observed in each cell line is indicated on top of the bars. ent a sequential decrease in the number of consensus OSE 2 elements they contain. The deletion constructs (3.6-, 1.9-, 1.5-, 0.9-, 0.4-, and 0.2-kb OPG ␤gal) were analyzed for basal expression in BALC and U2OS cells that express low and high amounts of endogenous OPG, respectively (Fig. 2B). As shown in Fig. 4, in both the cell lines, in comparison with the 5Ј deletion constructs, the 5.9-kb promoter directed relatively low levels of expression. Sequential deletions led to a progressive increase in base-line expression, suggesting the presence of negative regulatory elements in the distal region of the promoter. In U2OS, highest expression (2-3-fold) was observed with constructs containing 0.9-and 0.4-kb promoter sequences that include just the proximal OSE 2 element. Similarly, in BALC, the 0.9-kb promoter directed the highest level of expression, but in contrast to U2OS, additional 5Ј deletions (0.4 and 0.2 kb) resulted in decreased basal expression (Fig. 4). As a control, the SV40 promoter-driven ␤gal construct (SV40-␤gal) was used, and it directed very high levels of ␤gal expression compared with any of the OPG promoter-␤gal constructs.

Sequential 5Ј Deletions of the OPG Promoter Result in a Decrease in Cbfa1
Transactivation-To obtain further evidence of a role for the OSE 2 elements in OPG promoter expression, we performed cotransfection experiments in COS1 cells, which are nonosteoblastic and do not express endogenous Cbfa1 (40) or OPG (Fig. 2B). Analysis of the promoter deletions demonstrate that sequential removal of the distal OSE 2 elements (elements 12 through 2; Table I and Fig. 1) led to a decrease in Cbfa1 inducibility of the OPG promoter (Fig. 5). The removal of all the putative OSE 2 elements led to a near complete loss of transactivation, from 25-fold in the 0.4-kb OPG␤gal construct to only 4-fold in the 0.2-kb OPG␤gal construct (Fig. 5), suggesting that the OSE 2 elements may indeed contribute to the observed effects of Cbfa1. To evaluate whether the effects of Cbfa1 observed in COS1 would have any relevance in the stromal/ osteoblastic cell line BALC, and in the human osteosarcoma cell line U2OS, identical cotransfections were performed in BALC and U2OS cell lines using the 5Ј deletion constructs. As shown in Fig. 5, compared with the 7-fold induction of the FIG. 4. Effect of 5 deletions in the OPG promoter on base-line expression. Sequential 5Ј deletions of the OPG promoter were generated in p␤gal-Basic vector using standard cloning procedures, and the constructs were transfected into BALC and U2OS cells. ␤gal activity was measured in cell extracts (from three independent transfection experiments performed in triplicate) and is indicated as the percentage of change compared with that of the 5.9-kb OPG␤gal construct. The SV40-␤gal construct was used as a positive control and for comparison of the relative strength of the OPG promoter.

FIG. 5. Sequential 5 deletions in the OPG promoter result in a progressive decrease in Cbfa1 transactivation.
The OPG promoter deletion constructs were cotransfected with either pEF-Cbfa1 or pEF/myc/cyto into COS1, BALC, and U2OS cells, and the ␤gal activity was measured in cell extracts 48 h post-transfection. The average fold changes in promoter activity in Cbfa1transfected cell extracts compared with that in empty vector-transfected extracts are shown. Extracts from cells transfected with the promoter-less reporter vector p␤gal-Basic served as a control. 5.9-kb promoter in BALC cells, sequential 5Ј deletions led to a decrease and eventually resulted in a complete loss of transactivation. Similarly, in U2OS cells, the 16-fold induction in ␤gal activity directed by the 5.9-kb OPG promoter was nearly lost with the removal of regions harboring the OSE 2 elements. Interestingly, deletion of a 183-bp region (Ϫ372 to Ϫ190) containing the most proximal OSE 2 resulted in either a complete loss of transactivation or marginal responsiveness in all three cell lines (Fig. 5; compare 0.4-and 0.2-kb fragments), suggesting that it plays an important role in mediating Cbfa1 transactivation.
Substitution or Deletion of the Proximal OSE 2 Element Results in a Decrease in Cbfa1-mediated Transactivation-To further validate the role of the proximal OSE 2 element, we next asked whether mutations in this element affect promoter activity. A substitution mutation within this element in the 0.4-kb OPG ␤gal construct and an additional deletion mutant that lacks the proximal OSE 2 element (0.3-kb OPG ␤gal) were evaluated for Cbfa1 responsiveness. There was no significant change in the basal expression levels directed by these mutant constructs in COS1 cells, but a 45-65% decrease in basal promoter activity was observed in U2OS osteoblastic cells (Fig.  6). Because the endogenous OPG levels are high in U2OS cells (Fig. 2B), these results suggest the relative importance of the proximal element for basal expression as directed by the 0.4-kb fragment in U2OS cells. To further assess the role of the proximal OSE 2 element, the induction of OPG promoter expression by Cbfa1 was examined in cotransfection experiments in COS1 and U2OS cells. The wild type 0.4-kb fragment showed an approximately 25-fold increase in ␤gal activity in the presence of Cbfa1 in COS1 cells. However, mutations in the proximal OSE 2 element led to a 50% decrease in ␤gal activity, and the deletion of the OSE 2 element along with some of the flanking sequences (from Ϫ372 to Ϫ294; 0.3-kb OPG ␤gal), led to a 70% decrease in Cbfa1 transactivation of the promoter in COS1 cells (Fig. 6). Similarly, in U2OS cells, the ϳ14-fold increase in ␤gal activity directed by the 0.4-kb promoter fragment was decreased by 40% in the substitution mutant (0.4-kb OSE mutant) and by 55% in the deletion mutant (0.3 kb). Having demonstrated a requirement for the proximal element using deletion and substitution mutations in constructs containing only one OSE 2 element, we then directly investigated the con-tribution of the proximal element in the context of the 5.9-kb promoter that contains 11 additional OSE 2 elements. In brief, we assessed whether the 11 distal elements (elements 12 through 2) can rescue or maintain full Cbfa1 inducibility in the absence of a functional proximal element. For these studies, we mutated the proximal element in the 5.9-kb promoter (Fig. 7A) and evaluated its responsiveness to Cbfa1 in cotransfection experiments. There was no change in base-line expression directed by the wild type and mutant promoter fragments. However, mutation of the proximal element in the 5.9-kb promoter resulted in a 60% decrease in Cbfa1-mediated transactivation (Fig. 7B). Collectively, these results substantiate the relative importance of the proximal OSE 2 element in mediating Cbfa1 effects on OPG promoter expression in an osteoblastic cell line.
Addition of OSE 2 Sequences to the 0.2-kb OPG Minimal Promoter Restores Cbfa1 Responsiveness-We then tested whether the addition of OSE 2 elements could restore Cbfa1 inducibility to the 0.2-kb OPG minimal promoter. For this, either one (1xOSE-0.2kbOPG-␤gal) or three copies (3xOSE-0.2kbOPG-␤gal) of a canonical OSE 2 sequence were linked to the 0.2-kb minimal promoter, and the constructs were cotransfected into COS1 cells, along with either pEF-Cbfa1 or the empty vector pEF/myc/cyto. Addition of one copy of the OSE 2 element resulted in a 2-fold increase in activity of the minimal promoter (Fig. 8), consistent with a 2-fold (50%) decrease in activity of the 0.4-kb promoter fragment that had a substitution mutation in the proximal OSE 2 element (0.4-kb OSEmut-␤gal) or when the element was deleted (0.3 kb), as shown in Fig. 6. Interestingly, addition of three copies of the OSE 2 element resulted in a 17-fold increase in Cbfa1 transactivation as compared with the 3.6-fold induction of the 0.2-kb minimal promoter fragment.
Cbfa1 Binds in a Sequence-specific Manner to the Proximal OSE 2 Element in Vitro-The core sequence of the proximal OSE 2 element is identical to the OSE 2 element in the osteopontin promoter that has been shown to specifically bind Cbfa1 in vitro (1). To assess the ability of Cbfa1 to interact with the proximal OSE 2 element, electrophoretic mobility shift assays were performed with nuclear extracts from ROS 17/2.8 osteoblast-like osteosarcoma cells that express large amounts of Cbfa1 (36,(41)(42)(43). The native proximal OSE 2 element, the mutated proximal OSE 2 element (containing the substitution FIG. 6. Effect of substitution or deletion of proximal OSE 2 element on base-line expression and Cbfa1 transactivation. To assess basal levels of expression, OPG promoter-␤gal constructs containing a substitution or deletion of the proximal OSE 2 element were transfected into COS1 and U2OS cells. The average ␤gal activity measured in cell extracts from three independent transfection experiments is shown as the percentage of change compared with that of the 0.4-kb OPG␤gal construct. To assess the effect of substitution mutation/deletion of the proximal OSE 2 element on Cbfa1 transactivation, wild type as well as mutated OPG promoter ␤gal constructs were cotransfected into COS1 and U2OS cells along with either pEF-Cbfa1 or pEF/myc/ cyto. The average fold change in ␤gal activity in Cbfa1-transfected cell extracts compared with that in empty vectortransfected extracts is shown on the right. mutation that resulted in a 2-fold decrease in Cbfa1 transactivation; Fig. 6), and the OSE 2 element from the mouse osteocalcin promoter (positive control) were radiolabeled and used as probes. As shown in Fig. 9, Cbfa1 protein from ROS 17/2.8 cells did form a specific complex with the proximal OSE 2 element as it did with the positive control probe from the osteocalcin promoter (lanes 2 and 4). The specific complex was absent when a mutated OSE 2 element was used (lanes 3 and 5). This complex can be competed by unlabeled mouse osteocalcin OSE 2 (data not shown) (1). Also, the specific complex was observed with nuclear extracts from Cbfa1-transfected COS1 cells and not from empty vector transfected cells (compare lanes 7 and 8) or with a mutant proximal OSE 2 element (lane 11). The complexes formed with nuclear extracts from Cbfa1transfected COS1 cells that has no endogenous Cbfa1 expression (40), and those formed with nuclear extracts from ROS 17/2.8 cells had identical gel migration patterns suggesting that they are Cbfa1-OSE 2 complexes. Also, a similar Cbfa1-OSE2 complex was seen in nuclear extracts from U2OS cells, but the intensity was much lower (Ref. 4

and data not shown).
Overexpression of Cbfa1 Increases Endogenous OPG Protein Levels in U2OS Cells-Because overexpression of Cbfa1 results in a very strong stimulation of OPG promoter activity, we analyzed whether Cbfa1 could increase OPG protein levels in U2OS cells, using the OPG ELISA. As shown in Fig. 10, cells transfected with the Cbfa1 expression construct had a 54% increase in the levels of OPG protein secreted into the culture medium, compared with empty vector transfected cells (20 pg/10 3 cells versus 13 pg/10 3 cells).

DISCUSSION
It has been speculated that Cbfa1 through its effects on osteoblast lineage commitment and function could also, in part, directly or indirectly regulate the bone resorption process. In the present work, we provide evidence that OPG, a negative regulator of osteoclast formation and function that is secreted by stromal/osteoblastic cells, is a target of Cbfa1. We have cloned and characterized a 5.9-kb human OPG promoter and shown that it is robustly transactivated by Cbfa1 in a sequencespecific manner in both osteoblastic (BALC, U2OS) and nonosteoblastic (COS1) cells. Additionally, we demonstrate that this promoter is sufficient to direct osteoblast-specific expression, reminiscent of endogenous OPG protein expression.
The initial indication that Cbfa1 could regulate the expression of OPG came from the observation that there are 12 putative OSE 2 elements in the OPG promoter. Subsequently, ectopic expression of Cbfa1 in a nonosteoblastic cell line (COS1), which expresses no endogenous Cbfa1 (40) or OPG, resulted in a 39-fold increase in OPG promoter activity. In a stromal/osteoblastic cell (BALC) or osteoblastic (U2OS) back- FIG. 7. Substitution mutation in proximal OSE 2 element in the 5.9-kb OPG promoter significantly decreases Cbfa1-mediated transactivation. A, schematic representation of the 5.9kb OPG promoter region containing either wild type or mutated proximal OSE 2 element. B, effect of substitution mutations in the proximal OSE 2 element on Cbfa1 transactivation of the 5.9kb OPG promoter. U2OS cells were transfected with pOPG5.9-␤gal or pOPG5.9prox. OSEmut-␤gal construct, along with either pEF/myc/cyto or pEF-Cbfa1, in triplicate. The ␤gal activity in cell extracts was measured 48 h after transfection and is expressed as the fold increase in Cbfa1 transfected cells compared with empty vector transfected cells.
FIG. 8. Addition of OSE 2 sequences to the 0.2-kb OPG minimal promoter restores Cbfa1 responsiveness. Constructs containing either one or three copies of the consensus OSE 2 element linked to the 0.2-kb OPG minimal promoter and the wild type 0.2kb OPG␤gal construct were cotransfected into COS1 cells, along with pEF-Cbfa1 or pEF/myc/cyto. The fold induction in ␤gal activity directed by each promoter construct in the presence of Cbfa1, compared with its own control (in the presence of pEF/myc/cyto) is shown. Extracts from cells transfected with the promoter-less reporter vector p␤gal-Basic served as a control. ground, Cbfa1 could robustly stimulate OPG promoter activity (7-and 16-fold, respectively) ( Fig. 3B) suggesting that this effect is functionally relevant in a proper promoter context and cellular environment. The effects of Cbfa1 could be mediated either via direct binding of the protein to one or more OSE 2 elements and/or indirectly via the activation of other factors that in turn bind to and activate the promoter. Functional analysis using a combination of promoter deletion and site-directed mutagenesis approaches, demonstrated that the proximal OSE 2 element is necessary and sufficient to mediate Cbfa1 transactivation of the OPG promoter and also contributes to basal OPG promoter expression in osteoblastic cells (U2OS).
The role of the upstream promoter regions harboring the 11 distal OSE 2 elements (elements 12 through 2; Table I and Fig.  1) is underscored by the effect of sequential deletions of these elements that result in a moderate decline in Cbfa1 transactivation in COS1, BALC, and U2OS cells. Further, substitution mutation in the proximal OSE 2 element in the 5.9kb OPG promoter did not completely abolish Cbfa1 transactivation (Fig.  7B), suggesting that other elements (possibly the distal OSE 2 elements) play a role in mediating Cbfa1 effects on the promoter. Also, the increase in base-line expression seen upon sequential 5Ј deletions (Fig. 4) suggests the possible presence of negative regulatory elements in the distal region of the promoter, the removal of which leads to an increase in base-line expression.
The removal of the proximal OSE 2 element almost completely abolished Cbfa1 effects in all three cell lines. This loss of transactivation was restored by addition of either one or three copies of OSE 2 element to a minimal 0.2-kb OPG promoter (Fig.  8). Electrophoretic mobility shift assays performed with nuclear extracts from ROS 17/2.8 cells and Cbfa1-transfected COS1 cells, showed that Cbfa1 binds to the proximal OSE 2 element in a sequence-specific manner. Interestingly, in transactivation studies, residual effects of Cbfa1 were still observed after deletion of the most proximal OSE 2 element, suggesting that a full complement of Cbfa1 effects on OPG promoter may be mediated in part via noncanonical OSE 2 element(s) present in this region. Also, a substitution or deletion mutation in the proximal OSE 2 element led to a 45-65% decrease in basal promoter activity in U2OS osteoblastic cells and a 40 -70% decrease in Cbfa1-transfected COS1 and U2OS cells, suggesting that additional factors besides Cbfa1 are involved in regulating expression of the OPG promoter. It is conceivable that Cbfa1 induces the expression of another factor(s) that in turn binds to an element(s) in the promoter and induces expression. Cbfa1 could also interact with other promoter binding factors that regulate OPG promoter activity. Precedence for interaction between Cbfa1 and AP-1 has been suggested in the mediation of parathyroid hormone regulation of the mouse (5) and rat collagenase-3 promoter (33), and Cbfa1 and ETS1 have been shown to enhance osteopontin promoter activity in a synergistic manner (6). Collectively, these results suggest that Cbfa1 regulates the OPG promoter in a sequence-specific manner via contributions from one or more of the OSE 2 elements.
This paper provides the first direct demonstration that Cbfa1 regulates the expression of OPG, as reflected by changes in transcription and protein levels. The fact that mutation of the proximal OSE 2 element decreases OPG promoter activity in U2OS cells suggests that Cbfa1 plays a major role in OPG expression in osteoblastic cells. We have shown that Cbfa1 overexpression results in a 54% increase in the level of OPG protein that is secreted into the culture medium in U2OS cells. Because the only known functions of OPG in bone are inhibition of osteoclast formation and function, these results, at least in part, favor a role for Cbfa1 in inhibiting bone resorption. However, it remains to be determined whether Cbfa1 overexpression results in an increase in OPG protein levels in vivo, and the consequence, if any, on osteoclast formation/function and bone resorption is still undetermined. FIG. 9. Sequence-specific binding of Cbfa1 to the proximal OSE 2 element in vitro. Electrophoretic mobility shift assays were performed with nuclear extracts from ROS 17/2.8 cells or transfected COS1 cells. In lanes 1-5, complexes formed using ROS 17/2.8 nuclear extracts and various labeled OSE 2 probes are shown. Lane 1, free probe; lane 2, wild type osteocalcin OSE 2 (OC wt); lane 3, mutant osteocalcin OSE 2 (OC mut); lane 4, wild type OPG proximal OSE 2 (OPG wt); lane 5, mutant OPG proximal OSE 2 (OPG mut). The arrow represents the Cbfa1-specific complex. The consensus OSE 2 element from the osteocalcin promoter served as a positive control. The exposure time for the middle panel (lanes 4 and 5) was longer than that for the left panel (lanes 1-3). Complexes formed using nuclear extracts from transfected COS1 cells are shown in lanes 6 -11. The wild type OPG proximal OSE 2 element (OPG wt, lanes 6 -8) and the mutant OPG proximal OSE 2 element (OPG mut, lanes 9 -11) were used as probes. In a recent study, it has been shown that OPG mRNA is expressed in the fibrous connective tissues that are present in the calvarial region of Cbfa1Ϫ/Ϫ mice embryos (44). However, based on the fact that Cbfa1Ϫ/Ϫ animals lack functional osteoblasts (2,3) and that OPG is expressed abundantly in fibroblasts (45), it would be difficult to unequivocally determine whether Cbfa1 regulates the expression of OPG in osteoblasts in vivo using this system. A previous report in abstract form has also indicated the presence of potential OSE 2 sites in the promoter for RANK ligand, the cognate ligand for OPG (46), but there is as yet no evidence that Cbfa1 overexpression induces the activity of this promoter. It would be interesting to look directly at the role of Cbfa1 in the regulation of RANK ligand gene expression, because the relative levels of OPG and RANK ligand appear to be crucial in determining the extent/ rate of osteoclast differentiation (47).
In summary, in addition to the known role of Cbfa1 in promoting the differentiation of mesenchymal stem cells to form mature osteoblasts, expression of Cbfa1 appears capable of specifically increasing the production of OPG, which in turn could interfere with the interaction of RANK ligand with its receptor, RANK, on osteoclast precursors (15)(16)(17)(18)25), thereby leading to an inhibition of osteoclast differentiation (Fig. 11). Our data suggest the exciting possibility that the net effects of Cbfa1 on bone could result from an increase in osteoblast formation/ activity and inhibition of osteoclast formation/activity. FIG. 11. Model illustrating the mechanism by which Cbfa1 could regulate osteoclastogenesis. Shown is a schematic representation of the cellular and molecular interactions involved in osteoblast and osteoclast formation, showing that Cbfa1, in addition to its role in osteoblast differentiation and osteoblast maintenance, could also inhibit osteoclast formation and activity by stimulating OPG gene expression.