Transforming Growth Factor-β Up-regulates the β5Integrin Subunit Expression via Sp1 and Smad Signaling*

Integrin-mediated cell-matrix interactions play important roles in regulating cell function. Since transforming growth factor-β (TGF-β) modulates many osteoblast activities, we hypothesized that the growth factor acts in part by modulating integrin expression. TGF-β increased cell adhesion to vitronectin and up-regulated the surface level of αvβ5 via increasing β5 protein synthesis by a transcriptional mechanism. Promoter activity analysis demonstrated that a TGF-β-responsive element resides between nucleotides −63 and −44. Electrophoretic mobility shift assay and immunoprecipitation/Western studies indicated that the nuclear complex formed using the −66/−42 oligonucleotide contained both Sp1/Sp3 and Smad proteins. Since nuclear Sp1/Sp3 levels were not altered, whereas Smad levels were increased by TGF-β, we investigated the roles of Smad proteins in the up-regulation of β5 gene activation. Co-transfection of cells with β5 promoter reporter construct and expression vectors for Smad3, Smad4, and Sp1 increased the stimulatory effect of TGF-β. Furthermore, expression of dominant negative Smad3 or Smad4 in cells decreased or abolished the stimulation of β5 promoter activity by TGF-β. Smad4 mutant also inhibited the up-regulation of surface β5 level by TGF-β. Thus, TGF-β increases expression of the integrin β5 gene by mechanisms involving Sp1/Sp3 and Smad transcription factors.

Expression of integrins can be regulated by cytokines and growth factors (39). TGF-␤ increases the expression of ␣ v ␤ 3 , ␣ v ␤ 5 , and several ␤ 1 containing integrins in a variety of cells (40 -42). Little is known about the underlying signal transduction mechanisms. It has been well established that Smad proteins mediate the signal transduction for TGF-␤ (43)(44)(45)(46). Upon TGF-␤ binding to its receptors, Smad2 and Smad3 are activated via phosphorylation at the C-terminal end. These pathway-restrictive Smads then form complexes with Smad4, and these complexes migrate into the nucleus, where they exert transcriptional activities either directly or indirectly. Since TGF-␤ can induce osteoblast migration and differentiation and ␣ v ␤ 5 , found on osteoblasts, has been reported to mediate cell migration, adhesion, and function in other cell systems (33,47,48), we hypothesized that TGF-␤ modulates osteoblast activities in part via regulating expression of ␣ v ␤ 5 , and this regulation is dependent on Smad signals. We report here that TGF-␤ augments surface expression of ␣ v ␤ 5 by enhancing transcription of the ␤ 5 subunit gene in murine osteoblastic cell line MC3T3-E1, via a mechanism that requires both Sp1/Sp3 and Smad proteins. The higher surface levels of the integrin result in increased adhesion of growth factor-treated osteoblasts to vitronectin.
Cell Culture-The murine osteoblastic cell line MC3T3-E1 was cultured in ␣-minimum Eagle's medium with 10% heat-inactivated fetal bovine serum (HIFBS) until confluence. After changing to medium containing 0.2% HIFBS overnight, cells were treated with either vehicle or TGF-␤, which was dissolved in 4 mM HCl containing 1 mg/ml bovine serum albumin (BSA). Only cells less than passage 22 from our stocks were used.
Cell Adhesion-Adhesion of cells to vitronectin was performed as described previously (34). MC3T3-E1 cells were treated with TGF-␤ (1 ng/ml) or vehicle for 24 h. Single cell suspensions obtained after collagenase and trypsin/EDTA digestion were washed three times with serum-free ␣-minimum Eagle's medium supplemented with 0.1% BSA and allowed to recover for 30 min on a rotating platform at 37°C. 1 ϫ 10 5 cells were seeded to each well of 48-well Costar plates, which were pre-coated with vitronectin (5 g/ml, 0.25 ml/well) or BSA. After incubation at 37°C for 1 h, wells were washed three times with PBS, and the number of adherent cells measured by absorbance at 630 nm after staining the cells with 0.5% toluidine blue in 4% paraformaldehyde and dissolving the blue stain in 1% SDS.
Surface and Metabolic Labeling and Immunoprecipitation-Cells in p-150 culture dishes were treated with TGF-␤ (1 ng/ml) or vehicle for 24 h, washed with PBS, and surface-labeled by treatment with lactoperoxidase (20 g) and glucose oxidase (0.05 units) in 1 ml of 5 mM ␤-D-glucose in PBS containing 125 I-NaI (250 Ci/plate) as described previously (34). Cell layers were extracted with 1 ml of cell lysis buffer (10 mM Tris-HCl, pH 8.5, 0.15 M NaCl, 1 mM CaCl 2 , 0.02% NaN 3 , 2% Renex 30, and 1 mM AEBSF) and further homogenized by passing through 18-gauge needles. After micro-centrifugation, aliquots (approximately 400 l) of each supernatant containing equal trichloroacetic acid-precipitable radioactivity were precleared with protein A-Sepharose followed by incubation with polyclonal antibody against ␤ 5 (5 l) and 200 l of protein A-Sepharose overnight at 4°C on a Nutator. After centrifugation, pellets were washed twice with 500 l of RIPA buffer (10 mM Tris-HCl, pH 7.4, 0.15 M NaCl, 1% deoxycholic acid, 1% Triton X-100, 0.1% SDS, 0.12 TIU/ml aprotinin, and 0.02% NaN 3 ), followed by PT (0.5% Tween 20 and 0.02% NaN 3 in PBS) containing 1 mg/ml ovalbumin, and finally with PT. Pellets were extracted with sample buffer without reducing reagent (80 l/tube) and subjected to SDS-PAGE. The immunoprecipitated integrins were visualized by autoradiography. The band intensity was quantitated by image analysis using ISS SepraScan 2001 (Integrated Separation Systems, Natick, MA). For metabolic labeling, cells were treated with vehicle or TGF-␤ together with Tran 35 S-label (25 Ci/ml) for 24 h, and the cell layer was harvested. Immunoprecipitation with anti-␤ 5 antibody was performed as described above.
Northern Blot Analysis-Cells in p-150 culture dishes were treated with either TGF-␤ (1 ng/ml) or vehicle for 24 h. Poly(A) ϩ mRNA-enriched RNA was extracted by using the Ultra RiboSep kit according to the protocols provided by the manufacturer, separated by electrophoresis on 1.0% agarose-formaldehyde gels, transferred to nylon membrane, and fixed under UV light. mRNA on the membrane was hybridized with Megaprime-labeled [ 32 P]cDNA for mouse ␤ 5 integrin and reprobed with [ 32 P]cDNA for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) for loading variation. Bands of mRNA were visualized by autoradiography and quantitated by image analysis using ISS SepraScan.
␤ 5 Promoter Luciferase Reporter Constructs-Generation of progressive 5Ј-to 3Ј-deletion constructs of murine ␤ 5 promoters has been described previously (49). Briefly, a 1-kb fragment was isolated from the AccI-digested products of the ␤ 5 genomic 9-kb fragment. Deletion constructs of this 1-kb fragment were obtained by using the Exonuclease III/Mung Bean Nuclease kit from Stratagene. All the fragments were subcloned into pGL3-basic vector containing the luciferase reporter gene. The promoter constructs used in this study spanned from Ϫ875, Ϫ483, Ϫ340, Ϫ310, Ϫ274, Ϫ63, Ϫ43, and Ϫ28 to ϩ110 from the transcription start site.
Transfection and Luciferase Assay-MC3T3-E1 cells were transfected using DEAE-dextran for promoter activity analysis as described previously (34). Briefly, cells were plated at high density (150,000/well) onto 24-well plates in ␣-minimum Eagle's medium containing 10% HIFBS. Eighteen hours later, cells were transfected with 2 g/ml promoter construct to be tested and 0.7 g/ml CMV␤-gal plasmid using DEAE-dextran and a 90-s shock with 10% dimethyl sulfoxide. After 24 h of recovery in growth medium, cells were treated with vehicle or TGF-␤ (1 ng/ml) in medium containing 0.2% HIFBS for an additional 24 h, at which time cells were lysed in reporter lysis buffer followed by measurement of luciferase activity using Optocomp II Luminometer (MGM Instruments, Inc, Hamden, CT). Luciferase activities were normalized with the ␤-galactosidase activities in extracts measured using the ␤-Galactosidase Enzyme Assay System with Reporter Lysis Buffer kit. To test the effects of various expression vectors on promoter activities, cells were transfected with ␤ 5 promoter (0.2 g/well) together with indicated expression vector (0.2 g/well for each vector) or empty vector (pcDNA3) and CMV␤-gal plasmid (0.07 g/well) using LipofectAMINE Plus reagent according to the protocols provided by the manufacturer. After overnight incubation, cells were treated and analyzed as described above.
Nuclear Extract Preparation and Electrophoretic Mobility Shift Assay-Nuclear extracts from MC3T3-E1 cells, previously treated with 1 ng/ml TGF-␤ or vehicle, were prepared as described (50). 2-4 ϫ 10 7 cells were lysed in ice-cold buffer containing 10 mM Hepes-KOH, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride, and 0.6% Nonidet P-40 for 10 min to allow swelling. After vortexing for 10 s, lysates were micro-centrifuged for 10 min at maximum speed to obtain nuclei pellets which were then extracted with high salt buffer (20 mM Hepes-KOH, pH 7.9, 1.2 mM MgCl 2 , 420 mM NaCl, 25% glycerol, 0.5 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml pepstatin, and 0.12 TIU/ml aprotinin) for 1 h. Protein concentration of extracts was measured using Bio-Rad DC protein assay kit. For EMSA, radioactive double-stranded oligonucleotide Ϫ66/Ϫ42, labeled with T4 polynucleotide kinase and [␥-32 P]ATP, was incubated with nuclear extracts (2 g) or 0.5 footprint unit (1 footprint unit is the amount required to give full DNase I protection under standard conditions) of recombinant Sp1 protein in an ice bath for 20 min in 14 l of binding buffer (50 mM Tris, pH 7.5, 250 mM NaCl, 2.5 mM EDTA, 2.5 mM DTT, 20% glycerol, and 0.25 mg/ml poly(dI-dC)). Assays were terminated by addition of 1 l of 5ϫ TBE buffer (0.445 M Tris borate, pH 8.3, and 10 mM EDTA) containing 0.03% bromphenol blue, 0.03% xylene cyanol, and 30% glycerol and analyzed by electrophoresis at 4°C using 4 -20% polyacrylamide gels in 0.3ϫ TBE at 125 V for 2.5 h. Gels were dried, and autoradiography was performed. For competitive oligonucleotide or supershift assays, 100-fold unlabeled double-stranded oligonucleotides, 1 g of indicated antibodies, or 10 M mithramycin were incubated with nuclear extracts for 30 min before the addition of the radioactive probe.
Western Blot Analysis-Nuclear extracts (25 g) from control or TGF-␤-treated MC3T3-E1 cells were applied to a 10% SDS-PAGE and then blotted to an Immobilon-P membrane. Western analysis was performed by incubating the membrane with rabbit anti-Sp1 or Sp3 antibody (1:2000) or anti-pan-Smad antibodies (an equal mixture of anti-Smad antibody (C-17) and anti-Smad4 antibody (B-8)) (1:150), followed by horseradish peroxidase-conjugated secondary antibodies (1:2000), according to the rapid detection protocol provided by Millipore. Sp1, Sp3, and Smad proteins were visualized by enhanced chemiluminescence using an ECL kit. To detect Smad proteins in complexes with Sp1 and Sp3, RIPA buffer-diluted nuclear extracts (100 g of protein) were first immunoprecipitated with goat anti-Sp1 antibody (5 l) and rabbit anti-Sp3 antibody (1 l), respectively, and extracted with protein A-Sepharose. The extracts were subjected to SDS-PAGE, transferred to membranes, and probed with anti-Smad4 antibody (1:500).
Generation of MC3T3-E1 Mutants Expressing Dominant Negative Smad3 and Smad4 Proteins-MC3T3-E1 cells were transfected with pcDNA3 plasmid carrying FLAG-tagged dominant negative Smad3 or Smad4 cDNA (51) using LipofectAMINE Plus reagent. Stably transfected cell lines were generated by incubation in medium containing G418 (1 mg/ml) and used within five passages. Expression of these dominant negative Smad proteins was verified by Western blot analysis using anti-FLAG antibody.
Statistical Analysis-Statistical analysis was performed using Student's t test. Each experiment was performed at least twice. The data were presented as mean Ϯ S.E.

TGF-␤ Increased Adhesion of MC3T3-E1 Cells to Vitronectin-Treatment of MC3T3-E1 cells with TGF-␤ for 24 h in-
creased cell adhesion to vitronectin to 2.9-fold of the control level (Fig. 1). Since adhesion of osteoblasts (34) and other cell types (47,48,52) to vitronectin is dependent on ␣ v ␤ 5 , these data suggested that TGF-␤ up-regulated the expression or the activities of ␣ v ␤ 5 integrin on the surface of osteoblasts.
TGF-␤ Increased ␣ v ␤ 5 Expression on Murine Osteoblastic Cells-Surface labeling of MC3T3-E1 cells, followed by immunoprecipitation with a polyclonal antibody against the cytoplasmic tail of ␤ 5 and SDS-PAGE indicated that TGF-␤ treatment increased surface levels of ␤ 5 (and hence ␣ v ␤ 5 ) to 2.7-fold of the control level ( Fig. 2A). To determine whether this increase in surface expression was derived via increased synthesis, cells were metabolically labeled with Tran 35 S-label during treatment with TGF-␤ or vehicle, and immunoprecipitation was repeated as described above. As shown in Fig. 2B, TGF-␤ up-regulated the synthesis of ␤ 5 protein (6.6-fold of control level). Consistent with the increased ␤ 5 protein synthesis, normalized Northern blot analysis demonstrated that TGF-␤ increased ␤ 5 steady-state mRNA level to 2.3-fold of the control level (Fig. 2C). Thus, TGF-␤ stimulated expression of ␣ v ␤ 5 by increasing ␤ 5 mRNA level and protein synthesis.
TGF-␤ Up-regulated ␤ 5 mRNA by a Transcriptional Mechanism-To examine whether up-regulation of ␤ 5 mRNA by TGF-␤ resulted from increased gene transcription, MC3T3-E1 cells were transfected with progressive 5Ј end-deleted ␤ 5 promoter constructs carrying luciferase reporter gene. As demonstrated in Fig. 3, TGF-␤ stimulated the activities of all the promoter constructs that included the first 63 nucleotides upstream from the transcription start site. Further deletion of the promoter construct at the 5Ј end to Ϫ43 or Ϫ28 eliminated TGF-␤-mediated up-regulation (Fig. 3). Thus, TGF-␤ stimulated ␤ 5 expression via transcriptional mechanism(s), and a TGF-␤-responsive element resided between nucleotides Ϫ63 and Ϫ44 (Ϫ63/Ϫ44) of the murine ␤ 5 integrin gene.
TGF-␤ Increased Binding of Nuclear Factors to the Oligonucleotide Ϫ66/Ϫ42-To confirm that a TGF-␤-responsive element was indeed present between nucleotides Ϫ63 and Ϫ44 in the ␤ 5 proximal promoter, EMSA was performed after incubating the nuclear extracts, derived from either control or TGF-␤treated cells, with double-stranded radiolabeled oligonucleotide Ϫ66/Ϫ42. As shown in Fig. 4, three bands were observed in all samples tested. Two bands migrated as a closely spaced doublet, and the third was a faint, more rapidly migrating band. Treatment with TGF-␤ for 6 or 24 h increased the band intensities as compared with the corresponding control levels (Fig. 4). The stimulation by TGF-␤ tapered off after 48 h (Fig.  4). These data indicated that TGF-␤ stimulated the binding of nuclear factor(s) to the region Ϫ66/Ϫ42 of the murine ␤ 5 integrin gene.
The TGF-␤-responsive Element Was an Sp1/Sp3 Site Residing between Nucleotides Ϫ53 and Ϫ48 -Sequence analysis of the ␤ 5 proximal promoter (49) revealed the presence of three potential Sp1/Sp3-responsive elements, between nucleotides Ϫ53/Ϫ48, Ϫ24/Ϫ19, and ϩ28/ϩ33 (Fig. 5). The promoter analysis described above (Fig. 3) indicated that the site at Ϫ53/Ϫ48 was the likely TGF-␤-responsive element, whereas the remaining two Sp1/Sp3-like sites were inactive. To confirm this hypothesis, we used site-specific mutagenesis to generate promoter luciferase constructs carrying mutation at each of the three Sp1/Sp3-like sites. As expected, TGF-␤ stimulated the wild type promoter activity (Fig. 6A). When the nucleotides GG at Ϫ52 and Ϫ51 positions of the Ϫ53/Ϫ48 were mutated to TT The membranes were probed with 32 P-labeled mouse ␤ 5 cDNA followed by GAPDH. ␤ 5 mRNA levels were normalized with those of GAPDH.
To confirm that the Ϫ53/Ϫ48 site binds proteins of the Sp1/ Sp3 family, we carried out a series of EMSA-based studies. As expected, TGF-␤ increased the binding of all three bands (A-C) to radioactive oligonucleotide Ϫ66/Ϫ42 in EMSA (lanes 1 and 2,  Fig. 7). Preincubation of nuclear extracts with 100-fold unlabeled Ϫ66/Ϫ42 oligonucleotide inhibited binding of nuclear extracts to labeled probe (lanes 3 and 4). Consistent with the region being an Sp1/Sp3 site, addition of 100-fold excess con-sensus Sp1 oligonucleotide abolished all three bands in both control and TGF-␤-treated cells (lanes 5 and 6), whereas preincubation of the nuclear extract with 100-fold consensus AP-1 oligonucleotide failed to prevent complex formation (lanes 7 and 8). These combined data indicated that A-C bands in EMSA were specific complexes formed between oligonucleotide Ϫ66/Ϫ42 and the members of Sp1 family present in the nuclear extracts.
Since both Sp1 and Sp3 bind to the same DNA consensus sequence, we employed supershift EMSA to identify the proteins bound to Ϫ66/Ϫ42. Nuclear extracts were preincubated with either anti-Sp1 antibody or anti-Sp3 antibody before ad- Nuclear extracts from either control (C) or TGF-␤ (T)-treated cells were preincubated with the indicated competitive oligonucleotide or antibody followed by incubation with radiolabeled Ϫ66/Ϫ42 oligonucleotide. The complexes were separated on a 4 -20% gradient polyacrylamide gel. Lanes 1 and 2, no competitor added. Lanes 3-8, nuclear extracts were preincubated with indicated oligonucleotide before addition of radiolabeled Ϫ66/Ϫ42. Lanes 9 -12, nuclear extracts were preincubated with the indicated antibody. Lanes 13 and 14, nuclear extracts were preincubated with mithramycin, which interfered with Sp1/Sp3 binding. Lanes 15 and 16, nuclear extracts were incubated with radiolabeled double-stranded oligonucleotide Ϫ66/Ϫ42 m in which the GG residues at Ϫ52 and Ϫ51 positions in the Sp1/Sp3 site were replaced with TT. NC, no nuclear extract added. The bands right below band B and band C were not identified. dition of radioactive Ϫ66/Ϫ42. As shown in Fig. 7, lanes 9 and 10, band A was supershifted by the presence of anti-Sp1 antibody. In contrast, bands B and C were supershifted in the presence of anti-Sp3 antibody (lanes 11 and 12). Thus, both Sp1 and Sp3 mediate TGF-␤ up-regulation of the ␤ 5 integrin gene. Further evidence that the responsive element binds Sp1 family members was provided by the study in which preincubation of nuclear extracts with mithramycin, which binds to GC pairs of the DNA and interferes with Sp1/Sp3 binding, abolished all three bands in EMSA (lanes 13 and 14). Finally, radioactively labeled Ϫ66/Ϫ42 m , in which GG at Ϫ52 and Ϫ51 positions in the Sp1/Sp3 site were replaced with TT, failed to form complexes with nuclear extracts from both control and TGF-␤treated cells (lanes 15 and 16). Taken with the findings in Fig.  6, these observations provide strong support for the concept that an Sp1/Sp3 site residing in Ϫ53/Ϫ48 is the responsive element for TGF-␤ in the up-regulation of ␤ 5 integrin subunit.
TGF-␤ Up-regulation of ␤ 5 Involved Smad Protein Signaling-To determine if the mechanism by which TGF-␤ up-regulated ␤ 5 gene expression involved increasing nuclear levels of Sp1 and Sp3 proteins, we performed Western analysis on nuclear extracts from cells treated with vehicle or TGF-␤. As shown in Fig. 8A, TGF-␤ failed to increase the levels of nuclear Sp1 and Sp3. Furthermore, the complex formed between recombinant Sp1 protein and the Ϫ66/Ϫ42 probe migrated faster than the bands formed when nuclear extracts were the source of protein (Fig.  8B). This latter finding, taken with the unaltered level of Sp1 and Sp3, suggested that nuclear extracts contained additional protein(s) forming a complex with Sp1/Sp3 at its DNA-binding site. It has been well documented that Smad proteins translocate to the nucleus following initiation of TGF-␤ signaling (43)(44)(45)(46). Consistent with this finding, TGF-␤ increased the concentration of Smad proteins in nuclear extracts of MC3T3-E1 cells (Fig. 8A), suggesting that these molecules may play a role in the enhanced binding of Sp1 and Sp3 to the TGF-␤-responsive element in murine ␤ 5 integrin promoter.
To test this hypothesis, a supershift EMSA using anti-pan-Smad antibodies was performed. Preincubation of nuclear extracts with this mixture of antibodies (anti-Smad (C-17) ϩ anti-Smad4 (B-8)) drastically reduced the intensity of all bands in both control and TGF-␤-treated cells (Fig. 9, left). The reduction of the band intensity by anti-pan-Smad antibody in the control lane may derive from the basal secretion of active TGF-␤ by MC3T3-E1 cells (53,54), which would provide some activated Smad proteins available for interaction with Sp1/ Sp3, which ultimately bind to the responsive element. Since no supershifted bands were detected using anti-pan-Smad anti-bodies, we performed the same experiment using anti-Egr-1 antibody to support the specificity of the anti-pan-Smad antibodies. Egr-1 has been shown to interact with Sp1 and can prevent Sp1 from binding to its responsive element in mCSF promoter (55). Alternatively, the sequence of the responsive element for Sp1 overlaps with that of Egr-1 (GGCGGG versus GCGGGGGCG), and Egr-1 can compete with Sp1 for DNAbinding site in platelet-derived growth factor A chain promoter (56). As shown in Fig. 9, right, anti-Egr-1 antibody did not affect the band intensity in either control or TGF-␤-treated samples. Thus, the reduction of the band intensities by pan-Smad antibodies in both control and TGF-␤ lanes appeared to be specific. Moreover, Egr-1 did not appear to be involved in TGF-␤-mediated up-regulation of ␤ 5 integrin.
To confirm further the interactions between Sp1/Sp3 and Smad proteins, nuclear extracts were immunoprecipitated with anti-Sp1 or anti-Sp3 antibody followed by Western blot analysis for Smad4 protein. As demonstrated in Fig. 10, Smad4 binds to Sp1 and Sp3 in the nuclear extracts of both control and TGF-␤-treated cells, whereas nonspecific anti-rabbit and antigoat IgGs did not yield any Smad4 bands. Thus, Smads by translocating to the nucleus, where they associate with Sp1 and Sp3, are likely involved in regulating the TGF-␤-dependent increase in transcription of the murine ␤ 5 integrin gene. To test this possibility, we co-transfected expression vectors for Sp1 (kindly provided by Dr. Sunil Srivastava, University of Cincinnati, Cincinnati, OH), Smad3, and Smad4 (51) with ␤ 5 promoter reporter construct and determined the outcome by luciferase assay. As expected, TGF-␤ stimulated ␤ 5 promoter activity in cells co-transfected with the empty vector pcDNA3 (basal group) (Fig. 11). Sp1 expression vector alone had no effect on the promoter activity in the Control cells but significantly, although only slightly, increased up-regulation by TGF-␤ as compared with the basal group (26.12 Ϯ 1.41 versus 21.01 Ϯ 0.42, p Ͻ 0.05). Likewise, co-transfection with expression vectors of Smad3 and Smad4 had no effect on the ␤ 5 promoter activity in the Control cells but increased slightly the up-regulation by TGF-␤ as compared with the basal group (30.38 Ϯ 2.12 versus 21.01 Ϯ 0.42, p Ͻ 0.05). When cells were co-transfected with all three expression vectors (Sp1 ϩ Smad3 ϩ Smad4), Control and TGF-␤-stimulated ␤ 5 promoter activities were increased to 1.9-and 3.3-fold of their respective basal activity. These data confirmed that all three transcription factors (Sp1, Smad3, and Smad4) were involved in the up-regulation of ␤ 5 by TGF-␤.
To verify further the roles played by Smad3 and Smad4 in TGF-␤ up-regulation of ␤ 5 , MC3T3-E1 cell lines stably expressing dominant negative FLAG-tagged Smad3 (Smad3 m) or Smad4 (Smad4 m) were generated. The inset in Fig. 12 demonstrated expression of the dominant negative forms of Smad3 and Smad4, as evidenced by the positive Western signals using an antibody against the FLAG epitope. Whereas cells carrying FIG. 8. A, TGF-␤ has no effect on the nuclear Sp1 and Sp3 levels but stimulates those of Smad proteins. Cells were treated with either vehicle (C) or TGF-␤ (T) for 6 h. Nuclear extracts were analyzed for Sp1, Sp3, and Smad protein levels by Western blot analysis. B, the complex formed between oligonucleotide Ϫ66/Ϫ42 and nuclear extracts (NE) migrated slower than that formed with recombinant Sp1 protein in EMSA.
FIG. 9. Anti-Smad antibodies inhibit complex formation between nuclear extracts and the ؊66/؊42 oligonucleotide. Nuclear extracts (NE) of control (C) or TGF-␤ (T)-treated cells were preincubated with or without anti-pan-Smad antibodies (left) or anti-Egr-1 antibody (right) before addition of radiolabeled Ϫ66/Ϫ42. The complexes formed were separated on a polyacrylamide gel and visualized by autoradiography.
control vector (pcDNA3) maintained the up-regulation of ␤ 5 gene expression by TGF-␤, those expressing mutant Smad3 and Smad4 exhibited reduced and abolished stimulation of promoter activity, respectively (Fig. 12). Surface labeling showed that the induction of ␤ 5 by TGF-␤ was greatly diminished in the Smad4m cells, whereas the up-regulation of surface ␤ 5 level by TGF-␤ was maintained in MC3T3 cells carrying pcDNA3 empty vector (Fig. 13). Thus, Smad proteins play an essential role in TGF-␤ up-regulation of the ␤ 5 gene. DISCUSSION We have demonstrated that TGF-␤ up-regulates adhesion of murine osteoblasts to vitronectin and enhances surface expression of the integrin ␣ v ␤ 5 by a transcriptional mechanism targeting the rate-limiting ␤ 5 integrin subunit. We identify a TGF-␤-responsive element to a specific Sp1/Sp3 site between nucleotides Ϫ53 and Ϫ48 in the proximal ␤ 5 promoter. Increased ␤ 5 transcription, however, is not derived from altered Sp1 and Sp3 levels but rather from increased Sp1/Sp3 binding to the TGF-␤-responsive element as a result of TGF-␤-mediated nuclear translocation of Smad proteins, known to be involved in transducing signals initiated by the growth factor. Sp1/Sp3 are ubiquitous nuclear proteins that recognize homologous GC-rich sequences present in many promoters, including that of the murine ␤ 5 integrin subunit. Although Sp1/ Sp3 are abundant housekeeping transcription factors governing the basal activities of many TATAA-less promoters, they have also been implicated in the inducible regulation of several genes (57)(58)(59)(60). In line with the latter findings, our data indicate that Sp1/Sp3 mediates TGF-␤ up-regulation of integrin ␤ 5 subunit. Furthermore, both the pathway-specific Smad3 and the common Smad signaling protein Smad4 are involved in mediating TGF-␤ regulation of ␤ 5 . Although the role of Smad2, another pathway-specific TGF-␤ signaling molecule, was not examined, the inability of dominant negative Smad3 to abolish completely TGF-␤-mediated up-regulation as compared with the total inhibition obtained using dominant negative Smad4 (Fig. 12) suggests that Smad2 also participates in TGF-␤ up-regulation of ␤ 5 transcription. A similar interaction between Sp1 and Smad proteins leading to stimulation of transcription has also been reported in the up-regulation of cell cycle inhibitor p21 induced by TGF-␤ (57). While this manuscript was in revision, Pardali et al. (61) reported that Smad2, -3, and -4 can physically interact with Sp1 protein. They further demonstrated that the Mad homology 1 (MH1) domain of the Smad proteins interacts with Sp1, and the glutamine-rich sequence in the transactivation domain of Sp1 was involved in interaction with Smad proteins. Therefore, the regulation of Sp1 activity via Smad signaling molecules may represent one of the common mechanisms by which TGF-␤ exerts some of its effects on the cell.
Although there are three Sp1/Sp3 sites in the ␤ 5 proximal promoter region, only that at Ϫ53/Ϫ48 is responsible for the FIG. 10. Smad4 forms complexes with Sp1 and Sp3 in the nuclear extracts. Nuclear extracts (NE) from control (C) or TGF-␤ (T)treated cells were immunoprecipitated (IP) with either goat anti-Sp1 antibody or rabbit anti-Sp3 antibody and pulled down with protein A-Sepharose. The immunoprecipitated proteins were resolved on SDS-PAGE and transferred to Immobilon-P membranes. The membranes were probed with mouse anti-Smad4 antibody. For negative controls, samples were immunoprecipitated with nonspecific anti-rabbit IgG or anti-goat IgG before Western blot analysis.
FIG. 11. Co-transfection of ␤ 5 promoter luciferase reporter construct with expression vectors for Sp1, Smad3, and Smad4 increases basal and TGF-␤-stimulated promoter activities. MC3T3-E1 cells were transfected with the ␤ 5 promoter-luciferase reporter construct (Ϫ63 to ϩ110) together with the indicated expression vectors or empty vector (pcDNA3, basal group). After overnight recovery, cells were treated with either vehicle (Control) or TGF-␤ for 24 h and luciferase activities measured, which were normalized with cotransfected ␤-galactosidase activities. a, p Ͻ 0.01; b, p Ͻ 0.05 when compared with the corresponding value in the pcDNA3 transfected basal group; c, p Ͻ 0.001 when compared with the control value in the same group. stimulation by TGF-␤. The Sp1/Sp3-responsive element at Ϫ24/Ϫ19 is indispensable for basal expression, since mutation at this site results in a very low (5% of wild type) promoter activity (Fig. 6). In contrast, the Sp1/Sp3 site at ϩ28/ϩ31 is not involved in regulating ␤ 5 expression. The differential responsiveness of the three Sp1/Sp3 sites in the ␤ 5 promoter toward TGF-␤ stimulation may derive from the fact that variations in sequence of the regions flanking the three Sp1/Sp3 binding domains may lead to altered DNA conformation and hence function. Similar findings have been reported for several other promoters, including those of p21, p15 INK4B , and ␣1(I) procollagen (58 -60).
Although we have reported previously that ␣ v ␤ 5 mediates attachment of osteoblasts to vitronectin (34), whether the integrin plays other roles in osteoblast function is unknown. ␣ v ␤ 5 may promote osteoblast migration, since interaction between ␣ v ␤ 5 and vitronectin induces locomotion in several cell systems (48,62). In bone formation, osteoprogenitor cells in the bone marrow must migrate to remodeling sites on the bone surface. The stimulation of ␣ v ␤ 5 by TGF-␤ is consistent with the chemotactic and osteogenic activities of TGF-␤ (3, 14 -17, 19). ␣ v ␤ 5 can also function as an endocytic receptor for vitronectin and is involved in phagocytosis of apoptotic cells and rod outer segment (63)(64)(65)(66). Thus, ␣ v ␤ 5 may also mediate the reported ability of osteoblasts to clear the frayed organic matrix via endocytosis and/or phagocytosis in the resorption lacuna left behind by osteoclasts during the remodeling cycle (67)(68)(69). Recently, the TGF-␤1, -2, and -3 gene knockout mice have been generated (70 -72). Of note is that the TGF-␤2 knockout mice show significant defects in both endochondrial and membranous ossification in vivo (71). Since the formation of these tissues requires migration, proliferation, and differentiation of osteoblast precursor cells and integrins-mediated cell-matrix interactions govern these processes, there exists a possibility that the expressions of ␤ 5 and/or other integrins are compromised in TGF-␤2 null osteoblast precursor cells. Future experiments analyzing the expression patterns of integrins using TGF-␤2deficient osteoblasts may shed some light on this matter.