Sp1/Sp3 and PU.1 differentially regulate beta(5) integrin gene expression in macrophages and osteoblasts.

Murine osteoclast precursors and osteoblasts express the integrin alpha(v)beta(5), the appearance of which on the cell surface is controlled by the beta(5), and not the alpha(v), subunit. Here, we show that a 173-base pair proximal region of the beta(5) promoter mediates beta(5) basal transcription in macrophage (osteoclast precursor)-like and osteoblast-like cells. DNase I footprinting reveal four regions (FP1-FP4) within the 173-base pair region, protected by macrophage nuclear extracts. In contrast, osteoblast nuclear extracts protect only FP1, FP2, and FP3. FP1, FP2, and FP3 bind Sp1 and Sp3 from both macrophage and osteoblast nuclear extracts. FP4 does not bind osteoblast proteins but binds PU.1 from macrophages. Transfection studies show that FP1 and FP2 Sp1/Sp3 sites act as enhancers in both MC3T3-E1 (osteoblast-like) and J774 (macrophage-like) cell lines, whereas the FP3 Sp1/Sp3 site serves as a silencer. Mutation of the FP2 Sp1/Sp3 site totally abolishes promoter activity in J774 cells, with only partial reduction in MC3T3-E1 cells. Finally, we demonstrate that PU.1 acts as a beta(5) silencer in J774 cells but plays no role in MC3T3-E1 cells. Thus, three Sp1/Sp3 sites regulate beta(5) gene expression in macrophages and osteoblast-like cells, with each element exhibiting cell-type and/or activation-suppression specificity.

Recognition of bone matrix proteins by osteoblasts and osteoclast precursors profoundly affects their differentiation and function. Such interactions are mediated principally by integrins, which are heterodimeric transmembrane glycoproteins consisting of ␣ and ␤ chains (1)(2)(3).
The osteoclast is a physiological polykaryon derived by fusion of macrophages in a process apparently requiring attachment of the mononuclear precursors to bone matrix (4,5). The integrin ␣ v ␤ 3 is expressed predominantly in mature osteoclasts and plays a critical role in osteoclastic bone resorption (6 -10). In contrast, the integrin ␣ v ␤ 5 , although structurally related to ␣ v ␤ 3 and sharing many of the same target ligands (1), is expressed on rodent osteoclast precursors but not on mature bone-resorbing polykaryons (11). Murine osteoclast precursors utilize integrin ␣ v ␤ 5 and not ␣ v ␤ 3 for attachment to matrix (12).
Given these facts, it is not surprising that ␣ v ␤ 3 and ␣ v ␤ 5 levels rise and fall, respectively, during osteoclast differentiation (12).
PU.1, a member of the E twenty-six family of transcription factors, is expressed in macrophages, B cells, mast cells, and neutrophils (27,28). E twenty-six family proteins are characterized by a DNA binding domain that recognizes purine-rich sequences, typically containing a 5Ј-GGAA-3Ј core. PU.1 is necessary for both normal myelopoiesis (29,30) and osteoclast differentiation (31).
Given that the ␤ 5 integrin subunit is expressed in both osteoclast precursors and osteoblasts, we turned to the molecular mechanism by which the gene is regulated in these two important cell types. Taking advantage of our recently cloned murine integrin ␤ 5 gene promoter (32), we identified three Sp1/Sp3 sites in a 173-bp 2 ␤ 5 proximal promoter region and established that they mediate basal transcription in osteoblasts. Each of the Sp1/Sp3 sites plays a different role in this process. The first Sp1/Sp3 site (-53 to -48) acts as weak enhancer, the second site (-26 to -17) serves as a strong enhancer, and the third (ϩ29 to ϩ36) represses transcription. The same three Sp1/Sp3 regulate the ␤ 5 promoter in macrophages in a manner similar to that in osteoblasts. In contrast to its typical enhancer function, a downstream PU.1 site (ϩ73 to ϩ84), which is functional only in osteoclast precursors, serves as a silencer of the ␤ 5 promoter.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-pGL3-basic plasmid, a promoterless luciferase construct, was purchased from Promega (Madison, WI). A 1-kb ␤ 5 promoter-luciferase construct (pGL3-1kb(ϩ)) and its deletion mutants (pGL3(-796), pGL3(-633), pGL3(-483), pGL3(-340), pGL3(-172), and pGL3(-63)) were prepared as in our previous study (32). Briefly, in pGL3-1kb(ϩ), a nearly 1-kb proximal region (from -875 to ϩ110) of the ␤ 5 promoter containing the transcriptional start site (which was designated ϩ1) was placed, in the sense orientation to the luciferase gene, in the pGL3-basic plasmid. All mutants were made by deleting the 5Ј end of the 1-kb promoter fragment in the pGL3-1kb(ϩ) in a progressive fashion (32). Therefore, these constructs contain promoter fragments with the same 3Ј ends (ϩ110) and different 5Ј ends (the locations of these different 5Ј ends are indicated by the numbers in the parenthesis of these mutant constructs). In the current study, we made the point mutation constructs (Mu-1, Mu-2, Mu-3, Mu-12, Mu-13, Mu-23, Mu-123, and Mu-4) as detailed below, under "Site-directed Mutagenesis." Cell Culture and Transfection-The mouse macrophage cell line J774 (J774A.1) and a mouse pre-osteoblast cell line, MC3T3-E1, were cultured in minimum essential medium ␣ modification (Sigma) containing 10% heat-inactivated fetal bovine serum from Life Technologies, Inc. Cells were transfected with LipofectAMINE Plus TM reagent (Life Technologies, Inc.) as follows. One day prior to transfection, J774 cells were scraped off tissue culture dishes with Cell Lifters (Costar, Corning, NY), whereas MC3T3-E1 cells were lifted by trypsinization. The cells were counted and replated in six-well plates (5 ϫ 10 5 cells/well). The next day, every well was treated with a mixture prepared as follows. 1 g of reporter plasmid and 0.1 g of cytomegalovirus-␤-galactosidase (␤-gal) plasmid were diluted with Opti-MEM (Life Technologies, Inc.) and mixed with the Plus reagent to form a pre-complex at room temperature for 15 min. The pre-complex was then mixed and incubated with the LipofectAMINE reagent (previously diluted with Opti-MEM) at room temperature for 15 min to form DNA Plus LipofectAMINE, which was added to each well. The complex was mixed gently into the medium and incubated at 37°C at 5% CO 2 for 4 h. Then, the medium was replaced with minimum essential medium ␣ modification containing 10% fetal bovine serum and incubated at 37°C at 5% CO 2 . 24 h later, cells were washed with phosphate-buffered saline twice, lysates were prepared, and luciferase activity was measured using the luciferase assay kits from Promega (Madison, WI), with normalization to ␤-gal activity measured separately.
Nuclear Extract Preparation-Nuclear extracts used for both DNase I footprinting assays and gel shift assays were prepared as follows. Both cell types cultured until they reached confluence were washed three times with cold phosphate-buffered saline and incubated with 20 ml of phosphate-buffered saline containing 5 mM EDTA and 5 mM EGTA for 30 min on ice. Cells from two plates were scraped off the dishes with rubber policemen, pooled, spun down, resuspended in 1.5 ml of cold phosphate-buffered saline, and transferred to 2-ml microcentrifuge tubes. The cells were pelleted in a microcentrifuge for 30 s, media were removed, and the cells were resuspended in 500 l of hypotonic lysis buffer (10 mM Hepes-KOH, pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 , 0.5 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride; DTT and phenylmethylsulfonyl fluoride were added freshly). Cells were lysed for 15 min on ice, at which time 32 l of 10% Nonidet P-40 was added to the suspension, followed by vortexing the tube for 15 s and incubating on ice for 10 min. Nuclei were spun down and resuspended in 100 l of nuclear extraction buffer (20 mM Hepes-KOH, pH 7.9, 420 mM NaCl, 1.2 mM MgCl 2 , 0.2 mM EDTA, 25% glycerol, 0.5 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM 4-(2-aminoethyl) benzenesulfonyl fluoride, 5 g/ml pepstatin, and 5 g/ml leupeptin). DTT, phenylmethylsulfonyl fluoride, 4-(2aminoethyl) benzenesulfonyl fluoride, pepstatin, and leupeptin were added freshly to the buffer. The extract was kept for 20 min on ice and spun down in a microcentrifuge. The supernatant (nuclear extract) was aliquoted, quickly frozen in dry ice/ethanol bath, and stored at Ϫ70°C. Protein concentration of nuclear extracts was determined using the Micro BCA kit (Pierce).
DNase I Footprinting Assays-Two DNA probes, BssHI*-BglII and BssHI-HindIII* (the asterisk indicates the site of 32 P labeling), were prepared for DNase I footprinting assays as follows. Reporter plasmid pGL3-1kb(ϩ), which contains a 1-kb ␤ 5 promoter fragment (-875 to ϩ110), was cut with BssHI and HindIII to generate a proximal promoter fragment BssHI-HindIII (see Fig. 2A for details), which was end-labeled with 32 P using Klenow fragment (Life Technologies, Inc.). To prepare single-end-labeled probes, the double-end-labeled fragment (BssHI*-HindIII*) was extracted with phenol, precipitated with ethanol, resuspended in ddH 2 O, and digested with BglII at 37°C for 2 h. The large and small fragments (BssHI*-BglII and BglII-HindII*I) were separated by applying the digestion mixture to Quick Spin TM Sephadex Column G-50 (Roche Molecular Biochemicals). The purified large fragment BssHI*-BglII has only one end labeled (BssHI). To prepare probe BssHI-HindIII*, pGL3-1kb(ϩ) was cut with PstI and HindIII to obtain the fragment PstI-HindIII (see Fig. 2A). The fragment (PstI-HindIII) was labeled with 32 P, redigested with BssHI, and purified using Quick Spin TM Sephadex Column G-50 as described above. The resulting fragment BssHI-HindIII*, was labeled only at its HindIII end.
DNase I footprinting experiments were performed as described by Kucharczuk and Goldhamer (33). In short, 1 l (around 1 ϫ 10 5 cpm) of the one-end-labeled DNA fragment (BssHI-BglII) or (BssHI-HindIII) was incubated with 10 g of nuclear extract, prepared as described above, in 25 l of binding reaction solution (containing 12 mM Tris-Cl, pH 8.0, 1 mM MgCl, 5 mM NaCl, 1 mM CaCl 2 , 0.1 g/ul bovine serum albumin, 0.1 mM DTT, 5% glycerol, 50 mM KCl; 40 ng/l poly(dI⅐dC), and 2% polyvinyl alcohol) on ice for 60 min. For the bovine serum albumin control, 2.5 g of bovine serum albumin was used instead of 10 g of nuclear extract in binding reaction containing 60 mM instead of 50 mM KCl. Three binding reactions for the experimental sample and three control reactions were set up to allow for three different concentrations. DNase I stock (10 unit/l) was purchased from Roche Molecular Biochemicals. After a 60-min incubation on ice, the experimental binding reactions were treated with 1 l of 1/10, 1/20, or 1/40 dilution of the stock for 2 min, and the bovine serum albumin controls were treated with 1 l of 1/400, 1/800, and 1/1600 dilution of the stock for 2 min. DNase I-treated binding reactions were digested with proteinase K, followed by extraction with phenol. The DNA fragments were precipitated with ethanol and then separated by 6% sequencing gel. An unrelated sequencing reaction was used as a size marker.
Gel Shift Assays-Oligonucleotides (oligos) used for gel shift assays were synthesized by Life Technologies, Inc. and end-labeled with 32 P by T4 polynucleotide kinase (Life Technologies, Inc.). 1 ϫ 10 5 cpm probe was incubated with 2 g of nuclear extracts (prepared as described above) in a 20-l volume of binding reaction (10 mM Tris-Cl, pH 7.5, 100 mM NaCl, 10% glycerol, 50 ng/ml poly(dI⅐dC)) on ice for 20 min. In competition experiments, a 20ϫ or 100ϫ excess amount of unlabeled competitors was premixed with 1 ϫ 10 5 cpm of labeled probe before being added to the binding mixture. The binding reaction was then allowed to proceed for 20 min on ice. In supershift experiments, a 1 ϫ 10 5 cpm probe was incubated with 2 g of nuclear extracts in a 20-l volume of binding reaction for 20 min on ice, at which time 2 l of nonimmune serum or 2 l of specific antibodies (2 g/ul) was added, followed by incubation on ice for an additional 30 min. All binding mixtures were separated, using 0.5ϫ TBE buffer as the running buffer, at 4°C at 100 V for 3.5 h by 4 -20% gradient TBE gels (Novex, San Diego, CA) in a Novex XCell II TM minicell electrophoresis system. The gels were transferred to 3M blotting paper, dried, and exposed to film with an intensifying screen at Ϫ70°C. Antibodies (anti-Sp1, anti-Sp3, and anti-PU.1) and the nonimmune serum were purchased from Santa Cruz Biotechnology (Santa Cruz, CA).
Site-directed Mutagenesis-Point mutations were introduced in the context of pGL3(-63), which contains 173-bp proximal ␤ 5 promoter fragment (from -63 to ϩ110), using a QuickChange TM site-directed mutagenesis kit (Stratagene, La Jolla, CA). Oligos used to mutate the Sp1/Sp3 site within FP1 were 5-GGTCCCGGTGCAGttCGGAGCTG-G-3Ј and 5Ј-CCAGCTCCGaaCTGCACCGGGACC-3Ј. Oligos used to mutate the Sp1/Sp3 site within FP2 were 5Ј-GAGCTCGCCCCGaaCC-GTCCCGCC-3Ј and 5Ј-GGCGGGACGGttCGGGGCGAGCTC-3Ј. Oligos used to mutate the Sp1/Sp3 site within FP3 were 5Ј-GCAGGAGAGG-GttGAGGAGAAAGC-3Ј and 5Ј-GCTTTCTCCTCaaCCCTCTCCTGC-3Ј. Oligos used to mutate the PU.1 site within FP4 were 5Ј-CTGGCGGC-CGAGGAAAAAcccAAGGGTCTCCGAGAGTAG-3Ј and 5Ј-CTACTCTC-GGAGACCCTTgggTTTTTCCTCGGCCGCCAG-3Ј, with the lowercase letters indicating the mutation sites. These oligonucleotides were purchased from Life Technologies, Inc. and purified by polyacrylamide gel electrophoresis. PCRs were performed in a 50-l volume with Pfu polymerase (Stratagene, La Jolla, CA), 10 ng of DNA template, and 125 ng of each oligo using the following conditions: 95°C for 30 s, 1 cycle; 95°C for 30 s, 55°C for 1 min and 68°C for 12 min, 16 cycles; and 4°C. The PCR was treated with DpnI (10 units) for 60 min at 37°C. XL1-Blue supercompetent cells were transformed with the DpnI-treated PCR mixture as described in the instruction manual and plated on ampicillin plates. Plasmids were prepared from individual colonies and sequenced to confirm the correctness of the introduced mutations. All single-site mutants (Mu-1, Mu-2, Mu-3, and Mu-4) were generated using pGL3(-63) as template, the respective pair of oligos, and the PCR condition described above. The double-site mutants (Mu-12, Mu-13, and Mu-23) were made by performing a second round of PCR using the single-site mutants as template and the appropriate pairs of oligos. The generation of the triple-site mutant (Mu-123) was just an extension of the same approach, using a third set of oligos to perform another round of PCR.
Sequence Analysis-Sequence analysis was performed using the Genetic Computer Group (Madison, WI) sequence analysis software.

Identification of Transcription Factor Binding Sites in the
Proximal Region of the ␤ 5 Promoter-Osteoclast precursors and osteoblasts express the integrin ␣ v ␤ 5 (11,12,14,20). We previously cloned the integrin ␤ 5 promoter and observed that a 173-bp proximal ␤ 5 promoter fragment, pGL3(-63) (-63 to ϩ110), is capable of mediating ␤ 5 gene transcription in the myeloid cell line FDCP-Mac11 (32), suggesting that this region contains basal transcription factor binding sites. To pursue this observation, we turned to the macrophage cell line, J774, which is more easily transfected and in which pGL3(-63) also conferred basal promoter activity (Fig. 1A). Similar results were obtained with the murine osteoblast cell line MC3T3-E1 (Fig. 1B).
To identify potential transcription factor binding sites by which the 173-bp proximal promoter region controls ␤ 5 transcription, we performed in vitro DNase I footprinting assays with nuclear extracts of J774 or MC3T3-E1 cells. Four restriction sites (PstI, BssHI, BglII, and HindIII) flanking the 173-bp region were used to prepare two probes ( Fig. 2A). Probe BssHI-HindIII* was labeled at HindIII end, and probe BssHI*-BglII was labeled at BssHI end. Using probe BssHI-HindIII* (Fig.  2B), four regions (FP1-FP4) were protected by J774 nuclear FIG. 1. A 173-bp proximal ␤ 5 promoter region mediates basal transcription activity in both J774 and MC3T3-E1 cells. A, J774 cells were cotransfected with a series of 5Ј-deletion mutants of the murine ␤ 5 integrin promoter linked to luciferase, described previously (32), and cytomegalovirus-␤-gal plasmid. Luciferase activity was normalized to levels of ␤-gal. The experiment was repeated three times, and a representative result is shown. Each bar is the mean of three replicates Ϯ S.D. The smallest deletion mutant, pGL3(-63), which contains a promoter fragment from -63 to ϩ110, continues to mediate basal promoter activity. B, the experiment shown in A was repeated with MC3T3-E1 cells.
FIG. 2. In vitro footprinting assays reveal four regions within the 173-bp ␤ 5 promoter protected by J774 nuclear proteins and three regions protected by MC3T3-E1 nuclear proteins. A, nucleotide sequence of a proximal ␤ 5 promoter (-215 to ϩ110). This sequence is derived from pGL3-1kb(ϩ) construct in which a 1-kb ␤ 5 promoter fragment was subcloned, in the sense orientation to luciferase coding sequence, into pGL3-basic vector (32). The 35-bp pGL3 sequence is detailed in lowercase. The transcription start site (designated ϩ1) is indicated by an arrow. Two restriction sites (PstI and BssHI) located about 200 bp upstream of the start site, and two (BglII and HindIII) in the pGL3 vector were used to generate suitable probes for footprinting assays. B, in vitro footprinting assays, using probe BssHI to HindIII* and J774 or MC3T3-E1 nuclear extracts. The protected regions are shown by bars under the sequence in A. C, a similar experiment using probe BssHI* to BglII and J774 or MC3T3-E1 nuclear extracts. The protected regions from this experiment are indicated by bars above the sequence in A.
proteins, whereas MC3T3 extracts protected only three sites (FP1-FP3). To confirm this result, we repeated the footprinting assays with the probe BssHI*-BglII (Fig. 2C). Similar results were obtained with nuclear extracts prepared from mouse primary bone marrow macrophages or osteoblasts (data not shown).
To determine whether these protected genomic sequences bind nuclear proteins, we synthesized four oligonucleotides (FP1-FP4), each containing one of the protected sequences (Fig. 3A). In gel shift assays (Fig. 3B), oligos FP1, FP2, and FP3 gave rise to slowly (A) and quickly (B) migrating major bands with both J774 and MC3T3-E1 nuclear extracts. As demonstrated by subsequent studies (Figs. 4-6), A comprises two or three bands, depending on which oligo was used for binding. Oligo FP2 also yielded a number of minor, high mobility bands bands (band C) from J774 nuclear extracts and one (band D) from MC3T3 nuclear extracts. Oligo FP3 generated a minor band (band E) with high mobility only from MC3T3 nuclear extracts. 3 Consistent with the footprinting experiments (Fig.  2), oligo FP4 binds a nuclear protein (band F) only from J774 cells. Thus, several cis-elements within the proximal promoter region may mediate basal transcription of the ␤ 5 gene. FP1, FP2, and FP3 bind Sp1/Sp3, and FP4 Binds PU.1 from Macrophage Nuclear Extracts-FP1, FP2, and FP3 each contain potential Sp1 sites. A canonical PU.1 consensus sequence is located partially within FP4, which also includes a noncanonical PU.1 binding region (34) (Fig. 3A). To determine whether the proteins from J774 nuclear extracts binding to FP1-FP3 include Sp1 family members and those binding to FP4 include PU.1, respectively. we performed gel shift/competition assays.
As seen in Fig. 4A, 32 P-labeled FP1 oligos give rise to four bands (a-d, lane 1), each being eliminated by excess unlabeled probe (lanes 2 and 3). A 22-bp oligo containing an Sp1 consensus sequence abolished bands a, b, and c (lanes 4 and 5). When mutated, the consensus Sp1 oligo had no effect on bands a, b, or c ( lanes 6 and 7), suggesting that they are Sp1-related proteins. Given that band d was not affected by excess unlabeled wild type Sp1 oligo, it probably represents a nuclear protein that is not a member of the Sp1 family. Gel shift analysis using 32 Plabeled FP2 oligo yielded three major bands, a, b, and c, as well as several minor bands, all eliminated by excess unlabeled FP2, (Fig. 4B, lanes 1 and 2). Wild type, but not mutated, Sp1 oligo abrogated all bands. As seen in Fig. 4C, FP3 also specifically bound Sp1 related proteins. FP4 probe, incubated with J774 nuclear extract, yielded one band (Fig. 4D, lane 1, e), diminished by unlabeled FP4 (lanes 2 and 3). Wild type (lanes 4 and 5) but not mutated (lanes 6 and 7) PU.1 oligo also abolished the band.
These experiments suggest, but do not prove, that FP1, FP2,  4. FP1, FP2, and FP3 bind Sp1 family members, and FP4 recognizes PU.1 from J774 nuclear extracts. A, gel shift assays using labeled oligo FP1 as probe were competed with unlabeled oligo FP1 (lanes 2 and 3), an Sp1 consensus sequence (5Ј-ATTCGATCGGG-GCGGGGCGAGC-3Ј) (lanes 4 and 5), and a mutated Sp1 consensus sequence (5Ј-ATTCGATCGGttCGGGGCGAGC-3Ј) (lanes 6 and 7). B, the experiment shown in A was repeated with oligo FP2. C, the experiment shown in A was repeated with oligo FP3. D, competition assays with FP4 as probe. Gel shift assays using labeled oligo FP4 as probe were competed with excess of cold oligo FP4 (lanes 2 and 3), a PU.1 consensus sequence (5Ј-GGGCTGCTTGAGGAAGTATAAGAAT-3Ј) (lanes 4 and 5), and a mutated PU.1 consensus sequence (5Ј-GGGCTGC-TTGAGagaGTATAAGAAT-3Ј) (lanes 6 and 7). and FP3 recognize Sp1-related transcription factors (Fig. 4,  A-C, bands a, b, and c) and that FP4 contains a PU.1 binding site. FP1 also bound a slowly migrating protein not related to the Sp1 family, whereas FP2 yields rapidly migrating proteins. These moieties are either novel members of the Sp1 family or bind to the oligo in a manner requiring concomitant association of Sp1. To confirm the identity of putative Sp1, Sp3, and PU.1 proteins, we performed supershift assays, using appropriate antibodies. Whereas nonimmune serum had no effect on bands generated with probe FP1 (Fig. 5A, lane 2), anti-Sp1 antibody supershifted the lowest mobility species (lane 3), corresponding to band a in Fig. 4A. Anti-Sp3 antibody, in turn, supershifted two bands (lane 4), corresponding to bands b and c in Fig. 4A. The fourth band, d, was not supershifted by either antibody, indicating that it represents a non-Sp1/Sp3 protein. Similar results were obtained using oligo FP2 (Fig. 5B, lanes 1-4), and once again, the minor bands were not supershifted. This latter result, taken with the earlier observation that the two minor bands were displaced by both FP2 and Sp1 oligonucleotides in electrophoretic mobility shift assay studies, suggests they do not interact with Sp1/Sp3 but bind directly to a region of the FP2 sequence sharing homology with the Sp1 oligo. In summary, FP1 and FP2 bind Sp1 and Sp3 as well as other J774 nuclear protein(s). Finally, FP3 bound only Sp1 and Sp3 (Fig. 5C, lanes 1-4), and the FP4-bound protein is PU.1 (Fig. 5D, lanes 1-3).
Identification of Nucleotides in FP1, FP2, FP3, or FP4 Mediating Sp1/Sp3 and PU.1 Binding-We next asked whether FP1-FP4 sites contribute to ␤ 5 basal promoter activity. To this end, we first identified nucleotides in each site critical for nuclear protein binding. We synthesized the mutant FP1, FP2, and FP3 oligos shown in Fig. 6A. These mutant oligos failed to compete for Sp1/Sp3 binding from J774 nuclear extracts (Fig.  6B), indicating that the chosen mutations were sufficient to eliminate the binding capacity of these sites. To exclude the possibility that the chosen mutations, although they abolish Sp1/Sp3 recognition, render the respective oligonucleotides capable of associating with other nuclear proteins, they were used as probes in gel shift assays. Mutant FP1, FP2, and FP3 oligos failed to bind any nuclear proteins (data not shown).
FP4 contains two PU.1-like consensus sequences (Fig. 3A), from which we generated six mutants (M1-M6; Fig. 6C). M1 and M2 were mutated in the canonical putative PU.1 site partially protected by J774 nuclear extracts, and M3-M6 were mutated in the noncanonical sequence present within the fully protected region. M1 and M2 competed with FP4 for PU.1 binding, whereas M3-M6 failed to do so (Fig. 6D). Given that the nucleotides mutated in M1 and M2 retained the ability to compete for PU.1 binding, we conclude that the canonical PU.1 site located partially in the protected region is not functional. In contrast, M3-M6 failed to compete, indicating that their mutated nucelotides are critical for binding. Thus, the atypical site within the protected region of FP4 is a candidate PU.1 response element. Once again, to assess whether the mutant FP4 oligos bind nuclear proteins other than PU.1, we performed gel shift assays with each mutant probe. Whereas M1-M3 each recognized a molecule not seen by the wild type probe, M4 -M6 did not bind J774 nuclear proteins (data not shown).
Wild Type, but Not Mutated, FP1, FP2, and FP3 Bind Sp1/ Sp3 from Osteoblast Nuclear Extracts-To determine whether osteoblast nuclear proteins recognizing FP1, FP2, and FP3, like those derived from J774 cells, are also Sp1 and Sp3, we performed gel shift/competition assays with MC3T3-E1 nuclear extracts. As shown in Fig. 7A, the FP1 probe gave rise to three bands (a, b, and c, lane 1) all of which were abolished by unlabeled FP1 and Sp1 (lanes 2-5) but not mutant (m) Sp1 or FP1 oligos (lanes 6 -9). FP2, but not mFP2, also specifically bound Sp1-related nuclear proteins, and once again wild type, but not mutant, Sp1 oligos competed for binding (Fig. 7B). Finally, similar to our studies with J774 cells, five MC3T3-E1 derived bands appeared in gel shift assays with FP3 probe (Fig.  7C, lane 1, bands a-e) each being disrupted by unlabeled oligo FP3 (lanes 2 and 3). Unlabeled Sp1 oligos competed for binding nuclear proteins of bands a, b, and c, but less effectively for the minor bands (lanes 4 and 5). Neither mutant Sp1 nor FP3 oligos affected these bands (lanes 6 -9). We also found that mFP1, mFP2, and mFP3, although they lost their capacity to bind Sp1-related proteins, did not associate with other MC3T3-E1 nuclear proteins (not shown). Supershift assays revealed that the nuclear proteins associating with FP1 and FP2 were indeed Sp1 and Sp3 and that FP3 binds Sp1, Sp3, and other unknown proteins (Fig. 8). Furthermore, these experiments revealed that unlike our findings with J774 nuclear extracts, a third slowly migrating band (Fig. 4A, d) 1-4), FP2 (B, lanes 1-4), FP3 (C, lanes 1-4), and FP4 (D, lanes 1-3). In all cases, 2 l of nonimmune serum (NI) had no effect (lane 2). Sp1 and Sp3 antibodies supershifted the bands from assays with oligos FP1, FP2, and FP3 (lanes 3 and 4). PU.1 antibody supershifted the band from the assays with FP4 oligo (lane 3). In all cases, 2 l of specific antibodies (2 g/ul) were added to each incubation. NE, nuclear extracts. The region marked SS designates supershifted bands. d, a band not supershifted by either Sp1 antibody or Sp3 antibody. macrophages and osteoblasts, we introduced the point mutations, individually and in combination, which block Sp1/Sp3 binding in gel shift assays, into the parental pGL3(-63) reporter construct (Fig. 9A). When transfected into J774 macrophage-like cells, the triple mutant (Mu-123) exhibited only background luciferase activity (Fig. 9B). Thus, one or more of the three Sp1/Sp3 sites are critical for ␤ 5 basal promoter activity. Because the activity of Mu-1 was reduced compared with wild type construct (p Ͻ 0.0001) whereas Mu-23 was more active than Mu-123 (p Ͻ 0.0001), we conclude FP1 contains a weak enhancer element. Additionally, the Sp1 site in FP2 was essential for basal transcription because Mu-2 exhibited only background activity and Mu-13 was more active than Mu-23 (p Ͻ 0.0001). Finally, the FP3 Sp1/Sp3 site represses ␤ 5 transcription, because its sole mutation (Mu-3) prompted increased reporter activity (p Ͻ 0 .0001). Thus, the three Sp1 sites differentially mediate ␤ 5 promoter activity in J774 cells. The regulatory activity of the same three sites in MC3T3-E1 cells mirrors that in J774 cells other than the fact that the Sp1 site in FP2 is not critical in osteoblasts, because its mutation did not result in total loss of promoter activity (Fig. 9C).
PU.1 Represses ␤ 5 Gene Transcription in Macrophages-We finally introduced the point mutation contained in M4 (Fig.  6C), which blocks PU.1 binding (Fig. 6D), into pGL3(-63) reporter construct (Fig. 10A). When transfected into J774, this construct exhibited enhanced basal ␤ 5 transcription (Fig. 10B, p Ͻ 0.002), indicating that in this circumstance, PU.1 acts as a transcription repressor. DISCUSSION The integrin ␣ v ␤ 5 mediates specific activities of a wide range of cells following binding to its various ligands, each of which contains the amino acid sequence RGD. Thus, internalization of vitronectin from extracellular matrix and its subsequent  5 for FP1, lanes 9 and 10 for  FP2, and lanes 14 and 15 for FP3). C, sequences of wild type and mutant FP4 oligonucleotides. In all oligos, the sequence identified by the footprinting assays that contains a putative PU.1 site is underlined. A PU.1 consensus sequence located partially in the protected region is indicated by asterisks below the sequence. Mutated nucleotides are shown in lowercase. D, gel shift competition assays with the six mutant FP4 oligos. Gel shift assays were performed using labeled wild type FP4 oligo and J774 nuclear extracts (lane 1) with or without mutants M1-M6. intracellular degradation by fibroblasts are blocked by an antibody to ␣ v ␤ 5 but not by an antibody to ␣ v ␤ 3 (35). Similarly, although both ␣ v ␤ 3 and ␣ v ␤ 5 are present on mesoendothelial cells, antibodies to the latter integrin, but not the former, inhibit internalization of asbestos fibers in vitro. Coating of the fibers with vitronectin, a ligand for ␣ v ␤ 5 , enhances further the uptake capacity. Attesting to biological significance, ␣ v ␤ 5 , but not ␣ v ␤ 3 , shows immunological colocalization with the fibers in vivo (36). Furthermore, adenovirus-induced permeabilization is mediated by attachment of cells to the RGD-containing penton base domain of the virus, through ␣ v ␤ 5 (37). Whereas attachment of smooth muscle cells is regulated by ␣ v ␤ 3 or ␣ v ␤ 5 , migration is dependent solely on ␣ v ␤ 5 (38).
Of greater relevance to the present studies, ␣ v ␤ 5 is important for the function of both macrophages and osteoblasts. We demonstrate that murine osteoclast precursors, in the form of immature bone marrow macrophages, express ␣ v ␤ 5 and use this receptor to attach to matrix proteins, an early and obligate step in osteoclastogenesis (12). Furthermore, phagocytosis of apoptotic cells and cross-presentation of antigen to cytotoxic T lymphocytes by macrophage-derived dendritic cells involves ␣ v ␤ 5 , working in conjunction with CD36 (39). Finally, an antibody to ␣ v ␤ 5 inhibits attachment of human osteoblasts to vitronectin, whereas a function-blocking antibody to ␣ v ␤ 3 is without effect. 1 Given that ␣ v ␤ 5 plays a central role in the activity of osteoblasts and osteoclast precursors, we turned to the mechanisms regulating expression of this integrin. Our earlier work (12,40,41), and that of others (42) reveals that appearance of ␣ vassociated integrin on the surface of a cell is controlled by expression of the appropriate ␤, and not the ␣ V subunit. When first isolated, both osteoblasts and osteoclast precursors express ␣ v ␤ 5 , whose levels are regulated by TGF ␤ (osteoblasts), 1 granulocyte-macrophage CSF (osteoclast precursors (12), or the osteoclastogenic cytokine OPGL (osteoclast precursors). 4 In this regard, we isolated the murine ␤ 5 integrin promoter and identified a novel response element that mediates a granulocyte-macrophage CSF-dependent decrease in transcription (32). We have also delineated the TGF ␤-responsive region in osteoblasts. 5 In the course of the experiments defining cytokine-responsive components of the ␤ 5 promoter, we noted that basal expression is supported by a region included in -63 to ϩ110. Using footprinting analysis, we identified regions of this sequence protected by nuclear extracts from both osteoblasts and osteoclast precursors. Our data reveal the presence of four protected sites, three of which represent potential binding domains for SP family proteins, whereas the fourth region has two potential binding sites for the lineage-specific transcription factor PU.1. In competition assays using both consensus oligonucleotides and antibodies to the three transcription factors, we confirmed that the three SP sites bind a combination of Sp1 and Sp3, whereas only one of the two PU.1 sites is active. The functional PU.1 site we identified in the ␤ 5 promoter is 5Ј-AAAAGGGAAGG-3Ј, which, although it differs significantly from the core PU.1 sequence, 5Ј-GAGGAA-3Ј, is similar to a previously reported noncanonical PU.1 binding domain (34). Our finding is not without precedent (several PU.1 responsive genes lack a consensus sequence (43)(44)(45)) and supports the hypothesis that sequence alone is not sufficient to determine function of a putative transcription factor binding site.
the -63 to ϩ110 genomic fragment, which supports basal transcription in both osteoblasts and macrophage-like cells. Transfection of the individual mutants, alone or in combination, into both cell types, reveals each binding site plays a separate role in controlling ␤ 5 gene expression.
As expected, the region of the ␤ 5 promoter that binds PU.1 is active in myeloid-derived, osteoclast-related cells but is without effect in osteoblastic cells, which are mesenchymal in origin. PU.1 is found primarily in pre-B and B cells, macrophages, and osteoclasts, which are produced as a result of macrophage differentiation and fusion (46). Consistent with the critical role of PU.1 in macrophage formation, our group has demonstrated that PU.1 is essential for osteoclast differentiation and that levels of this transcription factor are increased during this event (31). In most circumstances, PU.1 acts as a transcriptional enhancer (47). Because PU.1 binds TBP in vitro (48), an attractive hypothesis holds that PU.1 assembles the basal transcription machinery on myeloid promoters, nearly all of which lack a TATA box. A recent report indicates that PU.1 regulates lineage-specific commitment of pluripotent hematopoietic precursors by decreasing expression of a master gene for nonmyeloid commitment, such as GATA-1, while at the same time increasing expression of the macrophage CSF, granulocytemacrophage CSF, and granulocyte CSF receptors (49), each of which plays a role in expansion of a specific myeloid sublineage.
The experiment was repeated three times and a representative result is shown. Each bar is the mean of three replicates Ϯ S.D. C, the transfection experiment shown in B was repeated with MC3T3-E1 cells.  A, schematic presentation of a pGL3(-63) reporter containing a mutation in the functional PU.1 binding site. The mutant (Mu4) was generated by introducing the point mutations shown to abrogate the binding capacity in gel shift assays above (Fig. 6) into pGL3(-63). B, basal ␤ 5 promoter activity in J774 and MC3T3-E1 transfected with parental pGL3(-63) or Mu4 mutant. J774 and MC3T3-E1 cells were transfected and assayed as described in Fig. 9, B and C. The experiment was repeated three times, and a representative result is shown. Each bar is the mean of three replicates Ϯ S.D.
We find that mutation of the active PU.1 site in the promoter of the ␤ 5 integrin gene leads to increased basal transcription, revealing that PU.1 is a suppressor of constitutive ␤ 5 promoter activity. Although, as discussed above, PU.1 is usually a transcriptional activator, it can also inhibit gene expression. For example, the 3Ј enhancer region of the murine Ig locus contains a PU.1 binding region required for tissue B cells/T cellsspecific V -J joining (50), which also acts as a negative element in pre-B cells. Attempts to detect PU.1-binding proteins, which might alter its transcriptional function, were unsuccessful (51). Additionally, transcription of the murine MHC-II gene I-A␤ falls when PU.1 is transiently co-expressed, leading to the proposal that PU.1 inhibits gene expression, by displacing NF-Y, a constitutive transcriptional activator of this gene (52). Finally, binding of PU.1 to the transcriptional start site suppresses expression of the CD11c gene (53). In summary, the proposed mechanisms by which PU.1 inhibits gene expression require either (a) binding close to the transcriptional start site or (b) interaction with other proteins acting as transcriptional co-regulators. Because the PU.1 binding site of the ␤ 5 promoter is not close to the transcriptional start site of the gene and we detected only a single band binding to an oligonucleotide containing the PU.1 sequence, it is unlikely that either displacement of the basal transcriptional complex or recruitment of a transcriptional co-repressor represents the mechanism by which PU.1 suppresses ␤ 5 gene expression. Thus, our findings suggest a potentially novel mechanism by which PU.1 can regulate gene expression.
In addition to the PU.1 binding site, the ␤ 5 proximal promoter region contains three domains recognizing Sp1 and Sp3, each differentially regulating basal transcription in J774 and MC3T3 cells. In both lines, a mutation approximately 30 bp downstream of the ϩ1 residue results in a 4-and 2-fold transcriptional enhancement, respectively, suggesting that binding of the Sp1/Sp3 complex to this genomic region is suppressive. This hypothesis is confirmed by the fact that ␤ 5 gene expression is blocked in the double mutant Mu-12, in which only the putatively inhibitory upstream site is active.
With a single exception (see below) the available data demonstrate that Sp1 invariably activates transcription (22). Sp3, on the other hand can either enhance or suppress transcription, with the outcome dependent on the context of the element within the intact promoter and the number of Sp binding sites (22,25,54). Similar to our findings, basal expression of the human glucagon-like peptide-1 receptor gene is regulated by three Sp1/Sp3 sites in the proximal promoter (55). Although the two more proximal sites act to enhance expression, the third, located at -344 to -339, is repressive. This inhibitory activity is dependent on Sp3, confirming the capacity of this protein to dampen promoters containing multiple Sp1 binding sites. Similarly, the human adenine nucleotide translocase gene is either activated or suppressed, depending on which of the three Sp1 sites is mutated (56).
Our electrophoretic mobility shift assay and supershift data demonstrate that in addition to Sp1 and Sp3, the FP3 region binds a rapidly migrating protein(s) in MC3T3 but not J774 nuclear extracts. We propose these proteins may play a role in ␤ 5 gene suppression in osteoblasts. Although these molecules are presently uncharacterized, one candidate is p74, a protein identified only in metabolic labeling experiments and documented to inhibit Sp1 gene transcription in vivo (57). Because Sp1 and Sp3 are the only proteins in macrophage nuclear extracts that bind to the FP3 oligonucleotide, the mechanism by which this region of the promoter decreases ␤ 5 transcription is speculative. Because Sp proteins are known to bend DNA (58), an event that can modulate gene activation (59), it is possible that upon binding, the Sp1/Sp3 heterodimer causes chromatin bending, thereby limiting access of the basal transcriptional machinery to the initiation site.
Although the downstream Sp binding site in the ␤ 5 gene is inhibitory, the two distal GC-rich elements act as transcriptional enhancers in both MC3T3 and J774 cells. On the other hand, although mutation of FP2 completely arrests transcription in macrophage-like cells, the event is only partially blunted in osteoblasts. Thus, as is the case with the TATA-less carbamoyl-phosphate synthase gene promoter (60), the central Sp1/Sp3 binding site acts as a basal promoter in macrophages. Of interest, nuclear extracts from both cell types, probed with the FP2 oligonucleotide, yield a rapidly migrating band. Attesting to specificity the band is competed by wild type, but not mutated, FP2. Although the identity of this protein remains unknown, it is a candidate for mediating that component of basal transcription that is Sp1/Sp3-independent. The fact that only FP2, and not FP1 or FP3, supports the binding of this molecule is not surprising, because the flanking sequences of all three Sp sites are distinct.
Sp1 and/or Sp3, in many instances, interact with a wide range of transcription factors, each binding to its own response element (61)(62)(63)(64). We found that transcriptional activity of constructs encompassing an 800-bp fragment of the ␤ 5 promoter immediately 5Ј of the region mediating basal activity is greater than that of the basal promoter alone (Fig. 1). Because this larger genomic region contains putative binding sites for a number of transcription factors (32), Sp1 and Sp3, in addition to regulating basal transcription, may interact with one or proteins bound to the upstream regions and thus regulate overall expression of the ␤ 5 gene in the bone cells.
Sp1 and its related protein Sp3 are implicated in the transcriptional regulation of many genes during different stages of development (22,23). Of particular relevance, Sp1 regulates a number of integrin subunits, including CD18, the ␤ chain of the leukocyte integrins (65); CD11c which forms a heterodimeric complex with CD18 (66, 67); and the ␣ 2 (68, 69), ␣ 6 (70), ␣ 7 (71), and ␣ v chains (72). Moreover, Sp3 activates the leukocyte integrin genes CD11c and CD11b in myelomonocytic cells (73). Our findings that Sp1 and Sp3 modulate expression of the rate-limiting ␤ 5 subunit during cytokine-induced differentiation of both major bone cell types reinforces the role of the two transcription factors in integrin biology.