Cloning of the Murine b5 Integrin Subunit Promoter IDENTIFICATION OF A NOVEL SEQUENCE MEDIATING GRANULOCYTE-MACROPHAGE COLONY-STIMULATING FACTOR-DEPENDENT REPRESSION OF b5 INTEGRIN GENE TRANSCRIPTION*

We previously noted that the initial receptor by which murine osteoclast precursors bind matrix is the integrin avb5 and that granulocyte-macrophage colony-stimulating factor (GM-CSF) decreases expression of this heterodimer by suppressing transcription of the b5 gene. We herein report cloning of the b5 integrin gene promoter and identification of a GM-CSF-responsive sequence. A 13-kilobase (kb) genomic fragment containing part of the b5 gene was isolated by screening a mouse genomic library with a probe derived from the most 5*-end of a murine b5 cDNA. A combination of primer extension and S1 nuclease studies identifies two transcriptional start sites, with the major one designated 11. A 1-kb subclone containing sequence 2875 to 1 110 is transcriptionally active in a murine myeloid cell line. This 1-kb fragment contains consensus binding sequences for basal (Sp1), lineage-specific (PU.1), and regulatable (signal transducer and activator of transcription) transcription factors. Reflecting our earlier findings, promoter activity is repressed in transfected myeloid cells treated with GM-CSF. Using deletion mutants, we localized a 109-base pair (bp) promoter region responsible for GM-CSF-inhibited b5 transcription. We further identified a 19-bp sequence within the 109-bp region that binds GM-CSF-induced nuclear proteins by gel shift/competition assays. Mutation of the 19-bp sequence not only ablates its capacity to bind nuclear proteins from GM-CSF-treated cells, in vitro, but the same mutation, when introduced in the 1-kb promoter, abolishes its ability to respond to GM-CSF treatment. Northern analysis demonstrates that cycloheximide treatment abrogates the capacity of GM-CSF to decrease b5 mRNA levels. In summary, we have identified a 19-bp cis-element mediating GM-CSF-induced down-regulation of b5 by a mechanism requiring protein synthesis.

The osteoclast is a physiological polykaryon, related to, but distinct from, foreign body giant cells (1). Both multinucleated cells are derived by fusion of macrophages in a process apparently requiring attachment of the mononuclear precursors to matrix (2,3). Cell-matrix interactions are mediated by several classes of receptors, including the family of heterodimers known as integrins (4 -6). The integrin ␣ v ␤ 3 plays a critical role in osteoclastic bone resorption (7)(8)(9)(10). In contrast, the integrin ␣ v ␤ 5 , although closely related to ␣ v ␤ 3 in structure and sharing many of the same target ligands (4), is not expressed on mature bone resorbing cells (11).
We have demonstrated that murine osteoclast precursors initially express and utilize the integrin ␣ v ␤ 5 and not ␣ v ␤ 3 for attachment to matrix (12). In this circumstance, ␣ v ␤ 3 and ␣ v ␤ 5 levels rise and fall, respectively, during formation of boneresorbing polykaryons. The reciprocal changes in the two functional integrins are replicated by treating osteoclast precursors with GM-CSF, 1 a cytokine that is important for osteoclastogenesis (13)(14)(15). Reflecting their surface heterodimers, ␤ 5 and ␤ 3 mRNAs fall and rise, respectively, in response to GM-CSF, whereas those of ␣ v are unaltered. GM-CSF-mediated inhibition of ␤ 5 expression reflects decreased transcriptional activity of the ␤ 5 gene (12).
In order to study the molecular mechanism by which integrin ␤ 5 gene is regulated by the cytokine, we have isolated a ␤ 5 integrin genomic fragment containing the transcriptional start sites and a portion of the regulatory domain of the gene. We demonstrate that the cloned sequence exhibits promoter activity and responds to GM-CSF in the same manner as the intact promoter in osteoclast precursors. Finally, we identify a 19-bp sequence within the ␤ 5 promoter that mediates the GM-CSF responsiveness of this gene. This sequence does not contain consensus sequences for any transcription factor known to be regulated by GM-CSF and, as such, represents a novel GM-CSF response element.

MATERIALS AND METHODS
Cloning of a Full-length Murine Integrin ␤ 5 cDNA-Human ␤ 5 integrin cDNA (provided by Dr. Eric Brown, Washington University School of Medicine) was labeled with digoxigenin to screen a 5Ј-stretch mouse kidney cDNA library (CLONTECH), and three clones were isolated and sequenced by dideoxynucleotide chain termination (16). Because none contained a translational start site, the most 5Ј sequence available was used to screen a mouse brain cDNA library by the GeneTrapper Screening Method (Life Technologies, Inc.). Several new clones were isolated and completely sequenced.
Isolation of Murine Integrin ␤ 5 Subunit Genomic Clones-A mouse genomic library in FIX II vector (Stratagene) was screened with a 230-bp digoxigenin-labeled fragment derived from the most 5Ј-end of the longest cDNA clone. One positive clone, identified after examining 500,000 plaques, was purified by secondary and tertiary screenings. Phage DNA was prepared from the clone and was digested with NotI, resulting in two DNA fragments of 4 and 9 kb. Both fragments were subcloned into a pBluescript plasmid. Southern blot analysis showed that the 9-kb fragment positively reacted with the 230-bp probe derived from the most 5Ј-end of the longest cDNA clone.
Primer Extension Analysis-Primer extension was performed using total RNA prepared from mouse osteoclast precursors (bone marrow macrophages) and a 30-mer oligonucleotide complementary to a sequence 23 bp downstream of the 5Ј-end of the longest cDNA clone (Fig.  1). Excess oligonucleotide, end-labeled with [ 32 P]ATP, was hybridized for 90 min at 65°C with total RNA (previously denatured for 5 min at 95°C). Annealed primer was extended with SuperScript II reverse transcriptase (Life Technologies, Inc.) at 42°C for 60 min. The reaction product was treated with DNase-free RNase at 37°C for 15 min to remove the RNA templates, and the resulting mixture was separated on a denaturing gel, with an unrelated and previously defined marker to establish the size of specific bands. S1 Nuclease Analysis-S1 nuclease experiments were performed using total RNA prepared from mouse bone marrow macrophages and a single-stranded 70-mer oligonucleotide comprising a sequence complementary to the first 44 bases of the 5Ј-end of the longest cDNA clone, the contiguous 16-base genomic sequence immediately upstream of the 5Ј-end of the longest clone, and a 10-base nonspecific sequence (as a control for nuclease activity; see Fig. 3B). The oligonucleotide, endlabeled with [ 32 P]ATP, was mixed with excess total RNA, and the mixture was denatured at 75°C for 10 min and then incubated overnight at 55°C. The annealed oligonucleotide was treated with S1 nuclease (Boehringer Mannheim) at 37°C for 60 min, and the resulting products were separated on a denaturing sequencing gel, using an unrelated and previously defined marker to establish the size of specific bands.
Construction of a ␤ 5 Promoter-luciferase Reporter Plasmid and Deletion Mutants-Sequence analysis indicates that there is a AccI site in the 5Ј-UTR of the full-length clone. A 30-mer oligo complementary to a sequence upstream of the AccI site was synthesized and end-labeled with digoxigenin. The 9-kb genomic fragment was digested with AccI, and the products were separated in an agarose gel, transferred to nitrocellulose, and hybridized with the 30-mer probe. A 1-kb fragment from the digestion reacted with the probe. Based on our screening strategy, this 1-kb fragment, which should contain the transcriptional start site, was subcloned into pGL3-basic plasmid containing the luciferase reporter gene, to construct a reporter plasmid named pGL3-1kb(ϩ). A control reporter plasmid, pGL3-1kb(Ϫ), was also obtained by placing the 1-kb ␤ 5 promoter fragment in antisense orientation in front of the luciferase gene.
Deletion mutants of the 1-kb fragment were made by using the exonuclease III/mung bean nuclease kit from Stratagene. The presence of two unique restriction sites in the multiple cloning sites region, one generating a 3Ј-overhang (KpnI) and another generating a 5Ј-overhang (Mlu I), made it possible to use this method. Following religation, the sizes of all mutants were determined by sequencing, using a common primer derived from vector sequence. The series of mutants generated in this manner is as follows: pGL3(Ϫ796), with a sequence from Ϫ796 to ϩ110; pGL3(Ϫ726), from Ϫ726 to ϩ110; pGL3(Ϫ633), from Ϫ633 to ϩ110; pGL3(Ϫ483), from Ϫ483 to ϩ110; pGL3(Ϫ340), from Ϫ340 to ϩ110; pGL3(Ϫ172), from Ϫ172 to ϩ110; pGL3(Ϫ63), from Ϫ63 to ϩ110; pGL3(Ϫ28), from Ϫ28 to ϩ110; and pGL3(ϩ10), from ϩ10 to ϩ110.
Cell Culture and Transient Transfection-FDC-P1/MAC11 cells were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 10% fetal bovine serum (FBS) with 5% WEHI cell conditioned medium as a source of interleukin-3. Prior to transfection, cells were counted and spun down at 1400 ϫ g for 7 min. The medium was then removed, and cells were resuspended in Opti-MEM (Life Technologies, Inc.) at a concentration of 8.75 ϫ 10 6 cells/ml. Four hundred microliters of this cell suspension were transiently transfected, by electroporation at 325 V and 950 microfarads in Gene Pulser II (Bio-Rad) in a 0.4-cm Gene Pulser cuvette, with 20 g of reporter plasmid and 0.5 g of CMV-␤-gal as a control plasmid for transfection efficiency. Transfected cells were grown in Dulbecco's modified Eagle's medium containing 10% FBS with 5% WEHI cell conditioned medium in 6-well tissue culture dishes for 16 h. Cell lysates were prepared, and luciferase activities were measured using the luciferase assay kits from Promega, with normalization to ␤-gal activities measured separately. In experiments utilizing GM-CSF treatment, FDC-P1/MAC11 cells were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 10% FBS with 5 ng/ml recombinant mouse (R & D Systems). When cell density reached 5 ϫ 10 5 /ml, recombinant mouse GM-CSF (R & D Systems) was added to the cell culture at 10 ng/ml. For controls, PBS containing 0.1% BSA was used. After 24 h, cells were counted and spun down at 1400 ϫ g for 7 min. The culturing medium was then removed and saved, and the cells were resuspended in Opti-MEM (Life Technologies, Inc.) at a concentration of 12.5 ϫ 10 6 cells/ml. 0.4 ml of this cell suspension was transiently cotransfected by electroporation at 325 V and 950 microfarads in Gene Pulser II (Bio-Rad) in a 0.4-cm Gene Pulser cuvette, with 20 g of reporter plasmid and 0.5 g of CMV-␤-gal as a control plasmid for transfection efficiency. The transfected cells were grown in 6-well tissue culture dishes with 3 ml of the saved culturing medium per well for 16 h. Luciferase and ␤-gal activities were measured as described previously.
Electrophoretic Mobility Shift Assays (EMSAs)-Nuclear extracts for EMSAs were prepared as follows. Bone marrow macrophages were isolated and cultured in minimum Eagle's medium containing 10% heat-inactivated FBS with daily supply of recombinant mouse macrophage colony-stimulating factor (R & D Systems; 10 ng/ml) for 2 days in 150-mm Petri dishes (1 ϫ 10 7 cells were plated in one dish). Then, the cells were treated with mouse recombinant GM-CSF (R & D Systems) for 48 h. For controls, PBS containing 0.1% BSA was used. The cells were washed three times with cold PBS and incubated with 20 ml of PBS containing 5 mM EDTA and 5 mM EGTA for 30 min on ice. Cells from two plates were scraped with rubber policemen, pooled, spun down, resuspended in 1.5 ml of cold PBS, 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 dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride; dithiothreitol and phenylmethylsulfonyl fluoride were added freshly). Cells were lysed for 15 min on ice, at which time 32 l of 10% Nonidet P-40 were added to the suspension, followed by vortexing of 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 dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM 4-(2aminoethyl)-benzenesulfonyl fluoride, 5 g/ml pepstatin, and 5 g/ml leupeptin). Dithiothreitol, phenylmethylsulfonyl fluoride, 4-(2-aminoethyl)-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 a dry ice/ethanol bath, and stored at Ϫ70°C. Protein concentration of nuclear extracts was determined using the Micro BCA kit (Pierce).
All oligos used for gel shift assays were synthesized by Life Technologies, Inc. Oligos were end-labeled with 32 P by T4 polynucleotide kinase (Life Technologies, Inc.). 1 ϫ 10 5 cpm of probe were incubated with 2 g of nuclear extracts in a 20-l volume of binding reaction (10 mM Tris-Cl, pH 7.5, 100 mM NaCl, 10% glycerol, 50 ng/ml poly(dI-dC)) on ice for 30 min. In competition experiments, different amounts of cold competitors were premixed with equal counts of hot probe before being added to the binding mixture. The binding mixture was separated at 300 V for 3.5 h by 4% high ionic strength native polyacryamide gel with an acrylamide: bis ratio of 80:1 as described (17), using a Hoefer SE 600 series gel unit. The gels were transferred to 3 M blotting paper, dried, and exposed to film with an intensifying screen at Ϫ70°C.
Site-directed Mutagenesis-The point mutations were introduced in the context of the 1-kb promoter (pGL3-1kbϩ) using a site-directed mutagenesis kit (Stratagene). Oligos used are 5Ј-CCCACAAGTGGtTt-AGCTGTtAtAGCCACCTGGG-3Ј and 5Ј-CCCAGGTGGCTaTaACAGC-TaAaCCACTTGTGGG-3Ј. Lowercase letters indicate the mutation sites. The oligos were purified by polyacrylamide gel electrophoresis. PCR was performed in a 50-l volume with Pfu polymerase (Stratagene), 10 ng of pGL3-1kbϩ (template), and 125 ng of each oligo under the following conditions: 1 cycle at 95°C for 30 s; 16 cycles at 95°C for 30 s, 55°C for 1 min, and 68°C for 12 min; and parking at 4°C. The PCR was treated with Dpn I (10 units) for 60 min at 37°C. XL1-Blue supercompetent cells were transformed with the Dpn I-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.
Northern Analysis-Mouse bone marrow macrophages were isolated and cultured in minimum Eagle's medium containing 10% heat-inactivated FBS with daily supply of recombinant mouse macrophage colonystimulating factor (10 ng/ml; R & D Systems) for 2 days. Cells were treated with recombinant mouse GM-CSF (10 ng/ml; R & D Systems) for 6 h (PBS containing 0.1% BSA was used for the control). Total RNA was prepared from these cells using Trizol Reagent (Life Technologies, Inc.). For cycloheximide treatment experiments, macrophages were treated with GM-CSF (10 ng/ml) 30 min after the addition of cycloheximide (5 g/ml). Cells then were cultured for 6 h before total RNA was isolated. 10 g of each sample was separated on 1% agarose gel, and Northern analysis was performed using NorthernMax kit (Ambion, Austin, TX). A 1.4-kb mouse ␤ 5 cDNA fragment was labeled by Bright-Star Psoralen-Biotin labeling kit (Ambion) and used as a probe in the Northern analysis (30 ng/ml). The signal was detected using BrightStar nonisotopic detection system (Ambion). The blot was stripped and reprobed with 1.3-kb rat GAPDH probe (30 ng/ml, labeled in the same way The deduced amino acid sequence is shown below the nucleotide data. The polyadenylation sequence is double underlined. Location of the primer (PE primer) used for the primer extension experiment (Fig. 2) is also shown below its complementary sequence (bp 24 -54).
as the ␤ 5 probe) to normalize loading. To analyze the intensity of each band on the autoradiograms, the x-ray films were subjected to image analysis using ISS SepraScan 2001 (Integrated Separation Systems, Natick, MA). ␤ 5 mRNA levels were normalized to the GAPDH signal.
DNA Sequence Analysis-Sequence analysis was performed using the Genetic Computer Group sequence analysis software.

Molecular
Cloning of a Full-length Murine ␤ 5 cDNA-Because integrin cDNA sequences are highly con-served across species (18,19), we labeled a human ␤ 5 cDNA (20 -22) and used it as a probe to screen more than 1,000,000 plaques derived from a 5Ј-stretch mouse kidney cDNA library (CLONTECH), which yielded three positive clones. Phage DNA was prepared, and the inserts were subcloned into pBluescript plasmid. Sequence analysis revealed that all three clones are incomplete, each terminating at ϩ400, based on the human ␤ 5 sequence. Because we lacked sequence coding for the 5Ј-end of the cDNA, we used the GeneTrapper method (Life Technologies, Inc.) to rapidly screen a large number of clones. This approach, using as a probe the 5Ј sequence derived from our first round of cloning, resulted in isolation of several new clones, the longest containing 294 bp of the 5Ј-untranslated region, the entire coding sequence, and 443 bp of the 3Ј-untranslated region, including a polyadenylation sequence ( Fig.  1) (sequence submitted to GenBank TM ; accession number AF022110). The deduced amino acid sequence is more than 90% identical to the human ␤ 5 protein (not shown).
Isolation of Murine ␤ 5 Integrin Subunit Genomic Clones and Sequence Analysis of the 5Ј-Upstream Region-Primer extension was performed to determine whether the largest ␤ 5 cDNA clone contains the complete 5Ј-untranslated region. The size of oligonucleotide used as primer (30 bp), plus the amount of untranslated sequence 5Ј of the ligation site of this primer (23 bp) would predict formation of a product of at least 53 bp (Fig.  1). A single band corresponding to a 59-mer oligonucleotide was detected (Fig. 2), indicating that the available ␤ 5 cDNA clone lacks 6 bases at its 5Ј-end. A 230-bp fragment derived from the most 5Ј-end of this cDNA was labeled and used to screen a mouse genomic library. After screening 500,000 plaques, one positive clone was identified, from which phage DNA was prepared. Restriction digestion analysis reveals the phage contains a 13-kb insert. A restriction map for the 13-kb was created by restriction digestion and Southern analysis (Fig. 3A).
FIG. 2. Identification of the ␤ 5 cDNA 5-UTR by primer extension analysis. Total RNA prepared from murine bone marrow macrophages was hybridized with an end-labeled 30-mer oligonucleotide complementary to a sequence 23 bp downstream of the 5Ј-end of the longest ␤ 5 cDNA clone (see Fig. 1). Annealed primer was extended with Superscipt II reverse transcriptase (Life Technologies, Inc.). The reaction products were treated with DNase-free RNase, separated on a sequencing gel, transferred onto filter paper, dried, and exposed to x-ray film. A single band, the size of which corresponded to a 59-mer oligonucleotide, was detected (lane 2). Lane 1 is an unrelated sequencing reaction used as a DNA marker. The 109-bp GM-CSF-responsive region, identified by transfections with the deletion mutants (see Fig. 7), is double underlined. The 19-bp GM-CSF-responsive sequence (identified in Figs. 9 -11) is indicated by asterisks. Location of the oligos used for the primer extension experiment (PE primer) and the S1 nuclease assay (S1 nuclear assay oligo) (see Fig. 4) are also shown. Arrow a indicates the major transcriptional site (designated ϩ1), corresponding to band A in Fig. 4. Arrow b indicates an alternative transcriptional start site, corresponding to band B in Fig.  4. Arrow c shows the transcriptional start site identified by primer extension experiment (Fig. 2). Arrow d marks the end of the sequenced cDNA clone (Fig. 1).
Digestion of the phage DNA with NotI yielded two fragments, 4 and 9 kb. Both fragments were subcloned into pBluescript plasmid. By Southern analysis, a probe containing the translational start site hybridized to the 9-kb clone. Sequencing of a portion of this 9-kb fragment, using primers (initially based on one vector arm and subsequently based on generated sequences), provided approximately 1000 bases representing, in the 3Ј to 5Ј orientation, a 30-bp intron, 70 bp coding for the known 23-amino acid leader sequence, and a portion of the 5Ј-untranslated region. The combined results yielded a region subsequently identified as the proximal regulatory domain of the ␤ 5 integrin gene (see Fig. 3B). Given the contiguous nature of the regulatory domain, 5Ј untranslated region and leader sequences, which end in the canonical AG sequence typical of intron/exon splice boundaries, this genomic fragment contains the first exon, which codes for the entire 5Ј-UTR and leader peptide, and the first base of the following codon. The size of the first intron, of which we have sequenced only 30 bp, is unknown.
Identification of the Transcriptional Start Site-To identify the 5Ј-end of the ␤ 5 cDNA we turned to S1 nuclease protection (Fig. 4). Using this technique, two bands were detected, suggesting that this gene may have two transcriptional start sites, a documented feature of TATA-less promoters (23). The more prominent band, corresponding to a transcriptional start site that is 5 bp more upstream than that predicted by primer extension (Figs. 2 and 3B), identifies a major transcriptional start site, designated ϩ1 (Fig. 3B). The discrepancy between this result and that obtained by primer extension (Fig. 2) probably reflects secondary structure near the GC-rich transcriptional start site inhibiting primer extension.
Sequence Analysis and Activity of Promoter Region-A 1-kb genomic fragment, derived from the original 9-kb genomic cDNA by restriction digestion and shown, by Southern analysis, to include the 5Ј-UTR of the ␤ 5 cDNA, was subcloned into pGL3-basic, which contains the reporter gene luciferase. The resultant 1-kb plasmid, pGL3-1kb (ϩ), contains 875 bp upstream and 110 bp downstream of the transcriptional start site. The putative 875-bp promoter includes no TATA box, but two Sp1 sites are located at Ϫ27 to Ϫ18 and Ϫ192 to Ϫ283. Putative STAT binding sites are found at Ϫ528 to Ϫ520 and Ϫ402 to Ϫ393, whereas PU.1 recognition sequences are present at Ϫ778 to Ϫ773, Ϫ772 to Ϫ767, Ϫ504 to Ϫ499, and ϩ68 to ϩ73 (Fig.  3B). To determine whether the cloned region is an active promoter, we transfected FDC-P1/MAC11 cells with either the 1-kb fragment pGL3-1kb(ϩ) or pGL3-1kb(Ϫ), in which the same genomic fragment was placed, in the same reporter, in the antisense orientation. As controls, we transfected the promoterless pGL3-basic and pGL3-promoter (SV40 promoter). Luciferase activity was normalized to ␤-gal content of cell extracts reflecting transfection with CMV-␤-gal. As seen in Fig. 5, the 1-kb ␤ 5 upstream fragment is transcriptionally active in murine myeloid progenitor cells. To locate the region essential for transcriptional activity, FDC-P1/MAC11 cells were transfected with pGL3-basic, pGL3-1kb(Ϫ), pGL3-1kb(ϩ), or a series of nested deletion constructs of the ␤ 5 promoter, pGL3(Ϫ726), pGL3(Ϫ340), pGL3(Ϫ172), pGL3(Ϫ63), pGL3(Ϫ28), and pGL3(ϩ10). The results indicate that basal promoter activity is regulated by a region located between Ϫ28 and ϩ10 (Fig. 5), a sequence containing a consensus domain for the basal transcription factor Sp1. Additional sequences may act as modest enhancers (e.g. pGL3(Ϫ172)) or silencers (e.g. pGL3(Ϫ340)) of basal ␤ 5 integrin transcription.
To further localize the region responsible for GM-CSF-mediated inhibition of ␤ 5 integrin gene expression, we repeated our transfection assays, using a series of ␤ 5 promoter deletion mutants. Once again cells were treated with vehicle or GM- FIG. 4. Confirmation of the position of ␤ 5 cDNA 5-UTR by S1 nuclease assay. Total RNA prepared from BMMs was subjected to S1 nuclease protection assay using an end-labeled 70-mer oligonucleotide containing a sequence complementary to the first 44 bases of the 5Ј-end of the longest ␤ 5 cDNA clone, a sequence complementary to the 16-base genomic sequence immediately upstream of the 5Ј-end of the longest clone, and a 10-base nonspecific sequence (see Fig. 3B). The hybridized and digested mixture was separated on a denaturing sequencing gel, transferred to a membrane, dried, and autoradiographed. Lane 1, unrelated sequencing reaction used as DNA marker; lane 2, labeled oligonucleotide hybridized with the total RNA, without subsequent S1 nuclease treatment. The arrow indicates the location of the intact oligonucleotide. Lane 3, bands A and B represent a 55-and a 54-bp oligo, respectively. This result suggests that two transcriptional start sites may exist. CSF prior to transfection. Whereas the activity of constructs pGL3(Ϫ796), pGL3(Ϫ726), pGL3(Ϫ633), pGL3(Ϫ483), pGL3(Ϫ340), and pGL3(Ϫ172) is inhibited by GM-CSF, that of the genomic fragment lacking Ϫ172 to Ϫ63 is not (Fig. 7). Thus, we have identified a 109-bp sequence mediating GM-CSF-dependent repression of the ␤ 5 integrin gene. This promoter region (Ϫ172 to Ϫ63) does not contain the two STAT-like sequences located at Ϫ528 to Ϫ520 and Ϫ402 to Ϫ393 (Fig. 3B).

Identification of a 19-bp Sequence within the 109-bp Region Mediating the GM-CSF-dependent Inhibition of ␤ 5 Gene
Transcription-To identify a smaller sequence mediating GM-CSF responsiveness, we synthesized four overlapping oligos (oligo I, II, III, and IV) spanning the previously defined 109-bp region (Fig. 8A). These short DNA fragments were end-labeled with 32 P and used for EMSA studies. Nuclear extracts used were prepared from bone marrow macrophages treated for 48 h with GM-CSF (10 ng/ml) or PBS containing 0.1% BSA (as control). GM-CSF treatment did not alter the pattern of proteins binding to oligos I, II and IV (Fig. 8B). In contrast, oligo III bound different nuclear proteins, derived from cytokine-treated, as compared with control, cells (Fig. 8B). Thus, untreated extracts, incubated with oligo III, generate a major band (Fig. 8B, labeled C) and a minor band (labeled B). When proteins from GM-CSF-treated cells were used, band C disappeared, band B increased in intensity, and a prominent new band (labeled A) appeared. This result suggests, but does not prove, that alterations in protein binding by bases within oligo III may mediate the inhibitory effect of GM-CSF on transcription of the ␤ 5 gene. To obtain additional information as to the identity of the bases responsible for the specific EMSA patterns seen in Fig. 8, we synthesized 7 deletion mutants of oligo III (Fig. 9A) and assessed their capacity to compete with labeled oligo III in EMSA (Fig. 9, B and C). When untreated nuclear extracts were used, mutants a, b, and g, but not mutants c-f, were efficient competitors (Fig. 9B), a finding that suggests that a 16-bp sequence (bp 8 -23, Fig. 9A) can bind the protein complexes B and C. When GM-CSF-treated nuclear extracts were used as source of protein, mutants a and g again competed efficiently, whereas b-f did not (Fig. 9C). This result indicates that a 19-bp sequence (bp 8 -26, Fig. 9A) binds the proteins present in bands A and B. Taken together, the studies using the various deletion mutations of oligo III, itself derived from the regulatory region of the murine ␤ 5 gene, allow us to conclude that a 19-bp sequence (bp 8 -26, Fig. 9A) is responsible for binding of both basal and GM-CSF-induced nuclear proteins.
These EMSA-derived data provide information only on the capacity of the sequence to bind nuclear proteins in vitro. In order to determine whether the 19-bp fragment defined by our EMSA studies is indeed functional in mediating GM-CSF inhibition of ␤ 5 gene transcription, we first generated a set of point mutations (Fig. 10A) in oligo III that abolished its capacity to compete in EMSA for binding of nuclear proteins derived from GM-CSF-treated or untreated BMMs (Fig. 10B). The mutant oligo III (Fig. 10B, M) failed to compete for binding of the nuclear proteins in bands A, B, and C (lanes 4, 5, 9, and 10), whereas wild-type oligo III (III) did so efficiently (lanes 2 3, 7,  and 8). These findings show that the 19-bp sequence containing the indicated point mutations did not bind the nuclear proteins FIG. 6. The 1000-bp proximal ␤ 5 promoter confers GM-CSFmediated repression of integrin ␤ 5 gene expression. FDC-P1/ MAC11 cells were transfected with pGL3-promoter (SV40), pGL3-1kb(ϩ), or CMV ␤-gal, with luciferase activity normalized to that for ␤-gal. The experiment was repeated three times, and a representative result is shown. Each bar is the mean of three replicates Ϯ S.D. FIG. 7. Localization of a 109-bp region mediating GM-CSF-dependent repression of the integrin ␤ 5 gene transcription. FDC-P1/MAC11 cells were cultured, transfected with a series of deletion constructs of the ␤ 5 promoter, and assayed as described in Fig. 6. The experiment was independently repeated three times, and a representative result is shown. Each bar is the mean of three replicates Ϯ SD. in bands A, B, and C. We then introduced the same point mutations in context of the 1kb promoter (pGL3-1kb(ϩ)) and named the resulting plasmid mpGL3-1kb(ϩ). When mpGL3-1kb(ϩ) was transfected into FDC-P1/MAC11 cells, its activity was no longer inhibited by GM-CSF treatment, whereas the wild-type reporter continued to respond (Fig. 11). Thus, we have identified a cis-element in the ␤ 5 promoter that mediates down-regulation of transcription of this gene by GM-CSF.  Fig. 9 is boxed; the mutated nucleotides are in lowercase letters, with their wild-type counterparts above or below. B, EMSA/competition assays were repeated as described in Fig. 9, B and C, with wild-type oligo III (III) or the point mutated oligo III (M).
FIG. 11. Mutation of the 19-bp sequence in the 1-kb ␤ 5 promoter abrogates its capacity to mediate GM-CSF responsiveness. The same four point mutations described in Fig. 10A were introduced in the context of the 1-kb ␤ 5 promoter, generating the construct mpGL3-1kb(ϩ). GM-CSF-treated or untreated FDC-P1/MAC11 cells were transfected with wild-type promoter (pGL3-1kb(ϩ)) or the mutant form, mpGL3-1kb(ϩ), with CMV-␤-gal plasmid as control and assayed as described in Fig. 6. These studies were performed four times, with a representative result shown. Each bar is the mean of three replicates Ϯ SD.  -g). B, EMSA/ competition assays with the deletion mutants shown in A. A 20ϫ or 100ϫ excess of each unlabeled mutant oligo was premixed with 1 ϫ 10 5 cpm of labeled oligo III, prior to EMSA, performed as described in Fig.  8B. C, EMSA/competition assays, as performed in B, but using nuclear extracts from GM-CST-treated cells.
We compared our 19-bp fragment with consensus sequences for transcription factors in a data base (Sitesdata.gcg) obtained from the National Center for Biotechnology Information (anonymous FTP site: ftp://ncbi.nlm.nih.gov/repository/TFD/datasets/), using the Findpattern program of the Genetic Computer Group sequence analysis package. We found that this 19-bp sequence contains putative binding sites for the transcription factors E2A (25), estrogen receptor (26), and GT-IIBa-SV40 (27).
Inhibition of the ␤ 5 Gene by GM-CSF Requires Protein Synthesis-GM-CSF treatment of BMMs for at least 6 h dampens ␤ 5 gene transcription, thus decreasing ␤ 5 mRNA levels (12). To investigate whether inhibition of ␤ 5 transcription by GM-CSF is dependent on protein synthesis, we exposed bone marrow macrophages to cycloheximide (5 g/ml) for 30 min prior to treating the cells with GM-CSF for 6 h. Total RNA was isolated, and Northern analysis demonstrated that cycloheximide treatment abrogates the inhibitory effect of GM-CSF on the ␤ 5 gene (Fig. 12). Thus, GM-CSF down-regulation of ␤ 5 gene transcription requires protein synthesis. DISCUSSION During authentic osteoclastogenesis, expression of the integrin ␣ v ␤ 5 falls in parallel with appearance of ␣ v ␤ 3 , events mimicked by treatment of osteoclast precursors with GM-CSF (12). Blunting of ␣ v ␤ 5 expression in these cells involves transcription suppression of the ␤ 5 , but not the ␣ v , gene. Because GM-CSF, in other circumstances, invariably stimulates gene expression, our observations represent a novel role for the cytokine, namely transcriptional inhibition.
The coding region of the human integrin ␤ 3 gene contains 14 exons within 46 kb of genomic DNA (28). The 5Ј-UTR resides in a separate exon, separated by an uncloned intron (29). Because the ␤ 3 and ␤ 5 integrin cDNAs are highly homologous (20), the ␤ 5 integrin gene, like that of ␤ 3 (28), may bear a large first intron. If this is so, it would be difficult to isolate a genomic clone containing the promoter region of the murine ␤ 5 integrin gene without identifying the 5Ј-untranslated end of the cDNA. With this in mind, our first exercise was to clone a full-length ␤ 5 murine cDNA. Using the most 5Ј-untranslated sequence as a probe, we screened a mouse genomic library and isolated a 13-kb fragment containing part of the coding region. A combination of primer extension and S1 nuclease analysis identified two potential transcriptional start sites. Similar to other integrin promoters (30 -33), the ␤ 5 gene does not contain a classical TATA element (25) in the vicinity of the transcriptional start site. The first kb of the ␤ 5 promoter, in contrast, accommodates four consensus sites for the B and myeloid cell-specific transcription factor, PU.1, which plays an essential role in osteoclastogenesis (34). Most importantly, the 1-kb fragment bearing the putative transcriptional start site contains a functional promoter as it enhances transcription at least 15-fold in transfected cells. Deletion analysis of the initial 1-kb promoter shows that basal promoter activity resides between Ϫ28 and ϩ10, a region that contains a consensus recognition site for Sp1. Because SP1 can act as an initiator of transcription in genes lacking a TATA box (35), the finding of a consensus SP1 binding site suggests that it may function in this role in the ␤ 5 promoter. Most importantly, by a combination of transient transfection studies and EMSAs, using both deletion and mutation of the regulatory region of the integrin ␤ 5 gene, we have identified a potentially novel 19-bp sequence responsible for GM-CSF-dependent transcriptional inhibition.
We were concerned because our EMSA experiments had been performed using BMMs as a source of protein, whereas our transfections used exclusively the FDC-P1 cell line. To address this issue, we repeated the EMSA studies using oligos I-IV and nuclear proteins from FDC-P1 cells treated with either GM-CSF or PBS-containing medium as control. In all instances, the pattern of bands was identical to that obtained using BMMs (data not shown), results that buttress the approach of using primary cells and transformed lines to dissect the mechanism of GM-CSF regulation of the ␤ 5 gene.
We initially attempted to determine whether the 19-bp GM-CSF-responsive element in the ␤ 5 promoter can mediate GM-CSF-dependent inhibitory effects on heterologous promoters, subcloning three tandem repeats of the 19-bp sequence upstream of the thymidne kinase minimal promoter. However, when this reporter plasmid was transfected into FDC-P1/ MAC11 cells not treated with GM-CSF, only background luciferase activity was detected (data not shown). When the same three copies of the 19 bp sequence were subcloned in front of the SV40 promoter (pGL3-promoter from Promega), high promoter activity was detected in transfected FDC-P1/MAC11 cells. However, GM-CSF treatment does not decrease the activity of this construct (data not shown). SV40 promoter is a strong TATA-box based viral promoter, whereas integrin ␤ 5 promoter is TATA-less. Such differences may contribute to the failure of the 19-bp sequence to mediate the inhibitory effect of the cytokine on the SV40 promoter. Thus, it is likely that the characterized GM-CSF-responsive sequence functions in a promoter-specific manner.
GM-CSF is a member of a cytokine superfamily that utilizes  2 and 4). Six hours later, total RNA was prepared, and Northern analysis was performed using ␤ 5 and GAPDH cDNAs probes. B, the intensity of ␤ 5 and GAPDH mRNA signals was quantitated using ISS SepraScan 2001 (Integrated Separation Systems, Natick, MA), and ␤ 5 mRNA levels were normalized to GAPDH. The results are displayed graphically. a number of major pathways, including those involving c-Myc (36), STAT5 (36), IRS-2 (37), and Ras (36), to transduce receptor-mediated signaling from the cell surface to the nucleus. STAT proteins are transcription factors that, in their inactive form, reside in the cytoplasm (38 -40). Binding of a wide range of cytokines to their cognate plasma membrane receptor activate members of the Janus kinase family, the major substrates of which are STAT proteins. Phosphorylated STATs dimerize and translocate to the nucleus, where they almost always activate transcription (40). STAT5, which exists in both mice and humans as two isoforms, STAT5A and STAT5B (41,42), in other circumstances mediates GM-CSF-dependent transcription (40). Although the ␤ 5 promoter contains two STAT-like sequences, they do not reside in the GM-CSF-responsive sequence and thus play no role in the mechanism by which the cytokine modulates expression of the gene. To show that the nuclear proteins binding to the 19-bp sequence are indeed not STAT5 proteins, we performed EMSA experiments with anti-STAT5A and anti-STAT5B antibodies and found, as expected, that these reagents fail to supershift the bands (data not shown). Thus, we conclude we have identified a GM-CSF response element binding transcription factor(s) other than STATs. The fact that this response element does not bind proteins known to be involved in transcriptional control by GM-CSF suggests that we may have identified a novel response element.