AU-Rich Elements in the Collagenase 3 mRNA Mediate Stabilization of the Transcript by Cortisol in Osteoblasts*

Collagenase 3 degrades collagen fibrils and is necessary for bone resorption. Cortisol increases collagenase 3 mRNA in osteoblasts by stabilizing collagenase 3 transcripts. To understand mechanisms involved, we used RNA electrophoretic mobility shift assay and RNA turnover studies. Cortisol increased the binding of Ob cell cytosolic extracts to AU-rich sequences in the collagenase 3 3′-untranslated region (UTR). No cortisol-dependent protein complexes were formed with the coding region or the 5′-UTR. Functional assays, using transient transfections of CMV-driven c-fos collagenase 3′-UTR chimeric constructs, demonstrated that the 3′-UTR of collagenase 3 stabilizes c-fos mRNA in transcriptionally arrested Ob cells, cortisol prolongs the transcript half-life, and mutations of AU-rich sequences destabilize c-fos transcripts precluding the cortisol effect. Purification of osteoblast cytosolic extracts by ultracentrifugation, ion exchange, and RNA affinity chromatography, and polyacrylamide gel electrophoresis followed by mass spectroscopy identified specific proteins. RNA gel mobility supershift assays demonstrated that vinculin and far upstream element (FUSE)-binding protein 2 interacted with collagenase 3 3′-UTR sequences, and RNA interference demonstrated these proteins altered collagenase mRNA stability. In conclusion, AU-rich sequences of the 3′-UTR of collagenase 3 and vinculin and FUSE-binding protein 2 regulate collagenase mRNA stability in osteoblasts.

In skeletal tissue, glucocorticoids primarily target cells of the osteoblastic lineage (1,2). Glucocorticoids decrease the replication and differentiation of precursor cells toward the osteoblastic pathway, thereby decreasing the number of bone-forming cells (3,4). In addition, they impair the functional activity of mature osteoblasts (5). In humans, the main consequence of skeletal tissue exposure to glucocorticoids is decreased bone formation, but glucocorticoids also increase bone resorption by mechanisms involving osteoblastic cells. Osteoblastogenesis and osteoclastogenesis are coordinated events and expression of receptor activator of nuclear factor B ligand (RANK-L) and colony stimulating factor (CSF) 1 by osteoblasts is critical for the genesis of bone-resorbing osteoclasts. Glucocorticoids increase the expression of RANK-L and CSF 1, and decrease the expression of osteoprotegerin, a RANK-L decoy receptor (6,7).
Collagenases are matrix metalloproteases (MMP) 1 that can initiate the cleavage of collagen fibrils at neutral pH (8). Consequently, they are considered central to the process of collagen degradation and matrix breakdown. Three collagenases have been described: collagenase 1 (MMP-1), collagenase 2 (MMP-8), and collagenase 3 (MMP-13) (8 -10). Rat and human osteoblasts and rat osteosarcoma cells express collagenase 3, but only human osteoblasts express the collagenase 1 gene (11,12). Collagenase 1 and 3 degrade type I collagen fibrils with similar efficiency (8). Unstimulated osteoblasts secrete limited amounts of collagenase and changes in the synthesis of collagenase correlate with changes in bone resorption and osteoblast differentiation (13). The importance of collagenase 3 in bone remodeling is documented by studies demonstrating that inhibitors of collagenase activity and collagenase neutralizing antibodies inhibit bone resorption (14). Furthermore, mice with a mutant ␣1 (1) type I collagen gene, whose collagen is resistant to collagenase 3 cleavage, fail to resorb bone following exposure to parathyroid hormone (PTH) (15).
The synthesis of collagenase 3 by osteoblasts is regulated by hormones and by cytokines present in the bone microenvironment acting, for the most part, by transcriptional mechanisms (13). Glucocorticoids are an exception, and increase collagenase 3 mRNA and immunoreactive protease levels in osteoblasts by post-transcriptional mechanisms, leading to the stabilization of collagenase 3 mRNA (16). This could play a role in the actions of these steroids on bone resorption. Regulation of gene expression by post-transcriptional mechanisms occurs in multiple cellular systems, and cytosolic protein-RNA interactions play an important role in the modulation of RNA stability (17). Regulatory elements in the 5Ј-and 3Ј-untranslated regions (UTR) of mRNAs frequently modulate mRNA stability. One well studied family of RNA stability motifs consists of adenosine-uridine regions (AUUUA) or AU-rich elements (ARE), and cytosolic extracts interacting with AREs are capable of stabilizing or destabilizing mRNA (18 -28).
The collagenase 3 RNA contains multiple potential AREs that may regulate RNA stability, but the role of these ARE motifs and the factors that determine basal collagenase 3 RNA stability, or changes in stability following glucocorticoid exposure, are unknown. In the present study, we used RNA electrophoretic mobility shift assay (REMSA) to identify elements present in the coding region, and 5Ј-and 3Ј-UTR of the collagenase 3 RNA forming cortisol-dependent complexes with cytosolic proteins from rat osteoblast-enriched cells (Ob cells).
The identification of proteins forming complexes with RNA, and the characterization of RNA sequences responsible for the stabilization of collagenase 3 RNA were pursued.

EXPERIMENTAL PROCEDURES
Cell Culture-The culture method used was described in detail previously (29). Timed pregnant Sprague-Dawley rats were obtained from Charles River Laboratories (Cambridge, MA), and parietal bones were removed from 22-day fetal rats immediately after the mothers were sacrificed by CO 2 asphyxiation. This project was approved by the Institutional Animal Care and Use Committee of Saint Francis Hospital and Medical Center. Cells were obtained by five sequential digestions of parietal bones, using bacterial collagenase (CLS II, Worthington Biochemical, Freehold, NJ). Cell populations harvested from the third to fifth digestions were cultured as a pool at a density of ϳ10,000 cells/cm 2 and have been shown previously to have osteoblastic characteristics (29). Cells were cultured in Dulbecco's modified Eagle's medium supplemented with nonessential amino acids (Invitrogen, Inc.) and 10% fetal bovine serum (Atlanta Biologicals, Inc., Atlanta, GA). At confluence, cells were rinsed and transferred to serum-free medium for 24 h and then exposed to test or control medium in the absence of serum for 2-24 h. Cortisol and actinomycin D (both from Sigma Chemical Co.) were dissolved in absolute ethanol, and at dilutions Ͻ 1:10,000 an equal amount of ethanol was added to control cultures. 5,6-dichloro-1-␤-Dribofuranosylbenzimidazole (DRB) (Biomol, Plymouth Meeting, PA) was dissolved in Me 2 SO at a concentration of 20 mg/ml, and diluted in culture medium 1:200. Except for protein purification studies, primary cultures of Ob cells were used in all experiments. For protein purification, Ob cells were grown to 75% confluence, trypsinized, and split 1:4, and grown to confluence.
Plasmid, Clones, and in Vitro RNA Transcripts-To generate RNA probes for testing in REMSA, 13 plasmids were constructed containing 28 -288 base pair (bp) sequences spanning the entire 2.6-kilobase (kb) rat collagenase 3 gene. Polymerase chain reaction (PCR) amplification of a 2.6-kb collagenase gene template (a gift from C. Quinn, St. Louis, MO) was performed with Pfu polymerase (Stratagene, La Jolla, CA), in the presence of 5Ј-and 3Ј-collagenase specific primers, and the PCR products were subcloned into the cytomegalic virus (CMV) driven expression vector pcDNA3.1(Ϫ) (Invitrogen) (30). A 200-bp murine osteonectin sequence, bp ϩ981 to ϩ1317 (a gift from B. Hogan, Nashville, TN), was similarly constructed (31). A c-fos triple ARE repeat RNA sequence was produced by annealing the single strand oligonucleotides 5Ј-CCGCTCGAGGATTTACGATTTTACGATTTTTACGAAGCTTGG-G-3Ј and 5Ј-CCCAAGCTTCGTAAAAATCGTAAAATCGTAAATCCTC-GAGCGG-3Ј to form double stranded DNA prior to subcloning into pcDNA3.1(Ϫ). Mutations in ARE sequences were generated by PCR using gene splicing by overlap extension method (32). The identity of each construct was verified by DNA sequencing prior to use. RNA templates were linearized with HindIII and transcribed with T7 RNA polymerase and Maxiscript Kit (Ambion, Austin, TX) in the presence of [ 32 P]uridine 5Ј-triphosphate (UTP) (800 Ci/mmol; PerkinElmer, Boston, MA) or unlabeled UTP to produce transcripts with a specific activity of 3.5 ϫ 10 7 disintegrations per minute (dpm)/g of RNA or unlabeled RNA to be used in competition reactions. The DNA template was removed with DNase I and unincorporated ribonucleotides removed by ethanol precipitation with 0.5 M ammonium acetate. Biotinylated RNA, used for RNA affinity purification, was generated by incubating unlabeled RNA transcripts with Photoprobe S-S Biotin (Vector Laboratories, Inc., Burlingame, CA) followed by extraction with 2-butyl alcohol (Malinkrodt, Paris, KY) and ethanol precipitation.
Cytosolic Extracts-Confluent Ob cells were rinsed with ice-cold phosphate-buffered saline (Sigma), scraped, and resuspended in HEPES-binding buffer, consisting of 10 mM HEPES buffer, pH 7.1, 3 mM MgCl 2 , 14 mM KCl, 5% glycerol (all from Sigma), 1 mM dithiothreitol (Invitrogen), and the freshly added proteinase inhibitors: 0.5 mM phenylmethanesulfonyl fluoride, 10 g/ml leupeptin, and 2 g/ml aprotinin (all from Calbiochem, San Diego, CA), and immediately frozen at Ϫ80°C. Cell lysis was accomplished by 4 cycles of thawing at 37°C for 5 min followed by snap-freezing in acetone on dry ice for 5 min. Cellular debris was cleared by centrifugation, and protein concentrations were determined using a Bio-Rad protein assay kit.
RNA Electrophoretic Mobility Shift Assay-RNA-cytosolic extract binding reactions were performed as described (33,34). 20 g of cytosolic extract were incubated with 5 ϫ 10 5 dpm [ 32 P]RNA sequence at 25°C for 30 min in HEPES binding buffer followed by the addition of 1 unit of RNase T1 (Roche Applied Science, Indianapolis, IN) and heparin (Sigma) at 5 mg/ml for 10 min each. Samples were subjected to electro-phoresis (200 volts) on a 4% native acrylamide gel (acrylamide/bisacrylamide ratio of 29:1; Sigma) at 4°C, and radiolabeled complexes were detected by autoradiography on Kodak X-AR film (Kodak, Rochester, NY), and their relative intensity determined by densitometric analysis. In RNA competition assays, unlabeled, homologous or non-homologous RNA sequences were added in excess (300 -500-fold) to the binding reactions prior to the addition of radiolabeled RNA probes. Supershift assays were performed by adding a 1:100 dilution of antibodies to far upstream element (FUSE)-binding protein-1 (a gift from D. Levens, Bethesda, MD), FUSE-binding protein 2 (a gift from D. Black, Los Angeles, CA), vinculin (Sigma), gelsolin, enolase, annexin I, and microtubule associated protein-2 (all from Santa Cruz Biotechnology, Santa Cruz, CA). Antibodies were added to the binding reactions simultaneously with the cytosolic extract and radiolabeled RNA sequences or immediately following the binding reaction with analogous results.
To estimate the molecular mass (M r ) of cytosolic protein-RNA complexes, RNA-protein binding reactions were carried out using 40 g of cytosolic extract and 10 ϫ 10 5 dpm of [ 32 P]RNA-labeled sequences. Following the binding reaction, samples were irradiated with UV light (2500 Jules) (CL-1000, UV Crosslinker, UVP, Inc., San Gabriel, CA) on ice for 15 min prior to incubation with 1 g RNase A (Qiagen, Valencia, CA) at 37°C for 15 min. Samples were boiled in SDS sample buffer and RNA-protein complexes resolved by SDS-polyacrylamide gel electrophoresis on 12% gels in Tris borate-EDTA buffer (Invitrogen). Gels were dried, exposed to Kodak X-AR film and analyzed by autoradiography. Benchmark molecular weight markers (Invitrogen) were used to estimate the M r of the complex.
Isolation and Identification of RNA-binding Proteins-S100 extracts were prepared from pooled Ob cell cytosolic extracts by centrifugation at 100,000 ϫ g for 1.5 h (Sorvall Ultra 80; Perkin-Elmer) at 4°C, and the supernatant stored at Ϫ80°C. Further purification was carried out by ion-exchange and RNA affinity chromatography. For this purpose, the S100 supernatant was applied onto a bed of Accell Plus QMA ion-exchange resin (Waters Inc., Milford, MA), previously equilibrated with HEPES buffer. The flow-through and two additional HEPES buffer washes were pooled and mixed with 5 mg/ml heparin, 5 mM EDTA and 75 g each of the S-S biotinylated collagenase RNA sequences ϩ1733 to 1933 and ϩ2133 to 2333 at 25°C for 2 h. Vectrex avidin D (Vector Laboratories, Inc.) slurry was then added and mixed for 2 h at 25°C (35). The sample was centrifuged at 2000 ϫ g for 5 min and the pellet washed and re-suspended in HEPES buffer containing 50 mM dithiothreitol and 500 mM NaCl, and incubated at 37°C for 45 min. Samples were centrifuged at 2000 ϫ g for 5 min, and the supernatant dialyzed in 3,500 molecular weight cut-off dialysis tubing (Spectrum Laboratories, Inc., Rancho Dominguez, CA) against 0.05 or 0.005 M acetic acid, at 4°C, prior to sample lyophilization. Aliquots from each purification step were stored at Ϫ80°C for subsequent determination of protein content and REMSA. Following the RNA affinity purification step, samples were fractionated by SDS-polyacrylamide gel electrophoresis in non-denaturing gels followed by Coomassie Blue staining (Bio-Rad). Selected stained gel slices were excised and analyzed by liquid chromatography (LC) (Ultimate LC Packings, Inc., Amsterdam, The Netherlands), electrospray ion-trap mass spectrometry (LCQ Deca from Thermoquest Corp., San Jose, CA), and tandem mass spectrometry (National Center for Biotechnology Information) after in-gel trypsinolysis at the University of Massachusetts Medical School, Worcester Foundation Campus (Shrewsbury, MA).
Transient Transfections and mRNA Turnover Studies-Transient transfection experiments were carried out with wild type c-fos and wild type and mutant c-fos collagenase chimeric constructs (Fig. 1). For this purpose, the 2.1-kb rat c-fos gene (a gift from T. Curran, Memphis, TN) was EcoRI-digested and subcloned into pcDNA 3.1(Ϫ) to create pCMVcfos (36). The 1.27-kb 5Ј-UTR and coding region of c-fos was isolated from pCMVc-fos by restriction digestion and cloned into pcDNA3.1(Ϫ), creating pCMV5ЈϩCRc-fos. The 1.2 kb collagenase 3Ј-UTR sequence was amplified by PCR using the rat collagenase 3 gene template and specific 5Ј-and 3Ј-primers; the PCR fragment was ligated to pCMV5ЈϩCRc-fos, creating pCMVc-fos/collagenase 3Ј-UTR (pCMVc-fos/case 3Ј-UTR). Mutations in ARE sequences were generated by PCR using gene splicing by overlap extension method, to create pCMVc-fos/case ⌬ 3Ј-UTR (32). The identity of the constructs was verified by DNA sequencing prior to use.
Cells were grown to ϳ75% confluence and transiently transfected with pCMVc-fos, and pCMVc-fos/case 3Ј-UTR, wild type and mutants, by calcium phosphate/DNA co-precipitation for 24 h, followed by glycerol shock as described, and allowed to recover in 10% fetal bovine serum for 48 h (16). Cells were washed, serum-deprived, and exposed to Dulbecco's modified Eagle's medium or 1 M cortisol for 4 h prior to the addition of 10 g/ml actinomycin D (Sigma) to arrest transcription. In experiments where mRNA decay was followed longer than 3 h, actinomycin D was pulsed immediately after the 3 h time point to obtain an additional 10 g/ml concentration. At selected times, cells were removed and total RNA was extracted with an RNeasy Kit (Qiagen) according to the manufacturer's instructions, and subjected to reverse transcription (RT)-PCR analysis for the detection of c-fos transcripts. For this purpose, 1 g of RNA was reverse transcribed in the presence of bovine growth hormone reverse sequence primer (5Ј-TAGAAGGCA-CAGTCGAGG-3Ј) corresponding to a polyadenylation sequence present in the pcDNA3.1(Ϫ), and 1 unit of RNasin (Promega, Madison, WI), with 1.0 l of murine Moloney leukemia virus-reverse transcriptase (Invitrogen). An aliquot was PCR amplified in the presence of forward primers 5Ј-CAATTTATTTATTAAGATGGATTC-3Ј of c-fos sequence, for pCMVc-fos, or 5Ј-CTTAGTGACTGACACTTGG-3Ј of collagenase sequence, for pCMVc-fos/case, and 2.5 Ci of [ 32 P]deoxycytidine 5Јtriphosphate (dCTP) (PerkinElmer), with 1 unit of Taq polymerase (Invitrogen) at 94°C denaturing, 58°C annealing, and 72°C extension temperatures for 26 or 28 cycles respectively. PCR products, 305 bp for c-fos, and 139 bp for c-fos/case 3Ј-UTR, wild type and mutant, were separated on denaturing 6% polyacrylamide gels (Invitrogen), dried, exposed to Kodak X-AR film, and visualized by autoradiography.
RNA Interference (RNAi) and mRNA Turnover-To downregulate vinculin and FUSE-binding protein 2, 21-mer double-stranded RNA (dsRNA) homologous in sequence to FUSE-binding protein 2 bp 289 -310 or vinculin bp 1766 -1787 or scrambled 21 bp non-mammalian gene sequence were transcribed in vitro using a silencer siRNA construction kit according to manufacturer's instructions (Ambion) (37,38). Briefly, sense and antisense siRNA oligonucleotide templates containing T7 promoter primer sequence at their 3Ј-ends, were individually hybridized to a T7 promoter primer and extended using Exo-Klenow DNA polymerase. The sense and antisense dsDNA templates formed were combined, and RNA was transcribed using T7 RNA polymerase at 37°C for 24 h. The 5Ј overhanging leader sequences were removed from the dsRNA product by a single-strand specific ribonuclease, and the DNA template was destroyed by DNase digestion. The resulting siRNAs were purified, quantitated by measuring their absorbance at 260 nm and stored at Ϫ20°C. The ability of dsRNA to downregulate FUSE-binding protein 2 and vinculin transcripts was tested by culturing Ob cells to ϳ75% confluence in the presence of 10% FBS. Cells were rinsed with phosphate buffered saline and transfected in Optimem (Invitrogen) with FUSE-binding protein 2, vinculin or scrambled dsRNA sequences, all at 75 nM, using gene silencer siRNA transfection reagent (Gene Therapy Systems, Inc., San Diego, CA), according to the manufacturer's instructions, and cultured for an additional 48 h period. Total RNA was extracted and processed for Northern blot analysis using a 922-bp human FUSE-binding protein 2 cDNA (a gift from D. Black) or a 1389 bp mouse vinculin cDNA (a gift from E. Adamson, La Jolla, CA). To determine the impact of FUSE-binding protein 2 and vinculin gene silencing on collagenase mRNA stability, cells were grown to ϳ75% confluence and transiently transfected for 48 h with 75 nM FUSEbinding protein 2, and 75 nM vinculin, or 150 nM scrambled dsRNA sequences, as described. Cells were washed, serum-deprived, exposed to DRB to arrest transcription, and at selected times, total RNA was extracted and subjected to Northern blot analysis using a 2.6-kb rat collagenase 3 cDNA.
Statistical Analysis-Slopes of RNA decay experiments were analyzed by the method of Sokal and Rohlf (39).

RESULTS
Examination of the collagenase 3 gene reveals a 28-nucleotide 5Ј-UTR, a 1.42-kb protein coding region, and a 1.2-kb 3Ј-UTR (30) (Fig. 1). Its 3Ј-UTR contains seven ATTT(T,T)A and the coding region contains two ATTT(T)A sequences, AREs in the RNA that may bind cytosolic proteins and modulate mRNA stability (17,24,40). To determine which segments of the collagenase 3 RNA could bind Ob cytosolic proteins, 28 -288-bp radiolabeled RNA sequences spanning the entire collagenase 3 RNA were incubated with cytosolic lysates prepared from Ob cells cultured in the presence or absence of 1 M cortisol, and resolved on 4% native polyacrylamide gels. RNA sequences from the 5Ј-UTR failed to form RNA-protein complexes, and sequences from the coding region formed complexes, but their intensity was not different in cell extracts from control or cortisol-treated cultures (data not shown). Cytosolic extracts from Ob cells formed complexes with three RNA segments present in the collagenase 3Ј-UTR and containing 2 AREs each (Fig. 2). There was a 2-4-fold increase in the intensity of the complex when extracts from cortisol-treated cells were incubated with ϩ1733 to 1933 and ϩ2133 to 2333 radiolabeled 3Ј-UTR RNA sequences, and a slight increase when extracts were incubated with ϩ1933 to 2133 3Ј-UTR RNA sequences. There was no cortisol-dependent complex formation when extracts were incubated with a ϩ1445-1733 segment containing 1 ARE, or with a ϩ2333-2595 segment containing no AREs. Mutations of ARE sequences present in the ϩ1733-1933, ϩ1933-2133, and ϩ2133-2333 3Ј-UTR regions resulted in a decrease in the intensity of the cytosolic protein-RNA complex from control and cortisol-treated cells, indicating that these elements are important in the formation of the complex (Fig. 2).
To confirm the specificity of the binding reaction, competition assays were performed. Incubation of the cytosolic extract with 500-fold excess unlabeled homologous sequence completely prevented the formation of the radiolabeled RNA-protein complex (Fig. 3, left and middle panels). To confirm the protein nature of the cytosolic component binding to the collagenase 3 RNA, cytosolic extracts were incubated with 32 P-  FIG. 3. RNA gel mobility shift assay demonstrating specificity of binding of cytosolic proteins from cortisol-treated Ob cells to 32 P-labeled ؉1733-1933 RNA sequences of the 3-UTR of collagenase 3, and to a 32 P-labeled triple repeat ARE sequence (3؋ AREs). The left panel shows binding of cytosolic extracts from Ob cells cultured in the absence (Ϫ) or presence (ϩ) of 1 M cortisol to 32 Plabeled ϩ1733-1933 RNA sequences, binding following preincubation with 500-fold excess of unlabeled, or cold, homologous RNA sequence (C), and binding followed by treatment of the reaction mixture with proteinase K prior to electrophoresis (PK). The middle panel shows binding of cytosolic extracts from Ob cells cultured in the presence (ϩ) of 1 M cortisol to 32 P-labeled ϩ1733-1933 RNA sequences, and binding following preincubation with 500-fold excess of unlabeled homologous RNA sequences (C), osteonectin RNA sequences (nonspecific, NS) or unlabeled triple ARE repeat sequences (ARE). The right panel shows binding of cytosolic extracts from Ob cells cultured in the presence (ϩ) of 1 M cortisol to 32 P-labeled triple ARE repeat sequences, and binding following preincubation with 500-fold excess of unlabeled homologous ARE repeats, or ϩ1733-1933 (1733) or ϩ2133-2333 (2133) unlabeled collagenase 3 RNA sequences. Arrows indicate location of specific RNAprotein complexes. labeled ϩ1733-1933 RNA sequences and treated with proteinase K prior to electrophoresis. This prevented the visualization of the RNA-cytosolic extract complex, confirming its protein nature. Incubation of cytosolic extracts with 32 P-labeled nonspecific osteonectin sequences did not result in the formation of a complex (not shown), and unlabeled osteonectin RNA sequences did not displace the specific radiolabeled collagenase 3 RNA protein complex (Fig. 3, middle panel). In contrast, preincubation with an unlabeled c-fos triple ARE repeat sequence decreased the radiolabeled RNA-protein complex, indicating that the complex formation involved AREs. This was confirmed by demonstrating a cytosolic extract complex formation with the 32 P-labeled fos triple ARE repeat sequence that could be displaced by homologous ARE sequences and by ϩ1733-1933 and ϩ2133-2333 collagenase 3 3Ј-UTR sequences (Fig. 3, right  panel). The effect of cortisol on cytosolic proteins complexing with collagenase 3 RNA sequences was time-dependent, and the effect was maximal with cytosolic extracts from Ob cells treated with cortisol for 2-6 h (Fig. 4).
To determine whether sequences present in the 3Ј-UTR of the collagenase 3 RNA were responsible for transcript stabilization, the rate of transcript decay was determined in Ob cells transiently transfected with a CMV-driven c-fos expression plasmid (pCMV-fos) or a chimeric construct in which the CMV promoter drives expression of the c-fos coding region linked to the 3Ј-UTR of collagenase 3 (pCMVc-fos/case3ЈUTR). The c-fos gene was chosen because of the short half-life of its mRNA (41). Ob cells were transiently transfected with CMV-driven native c-fos construct (pCMVc-fos) or with the collagenase 3Ј-UTR substituted construct (pCMVc-fos/case3ЈUTR), grown to confluence and treated with actinomycin D to arrest transcription. RNA derived from the transfected plasmids was detected by RT-PCR using construct-specific primer pairs. The half-life of native c-fos transcripts following transcriptional arrest was ϳ5 min, whereas transcripts generated from pCMVc-fos/case 3Ј-UTR did not decay throughout a 30 min period, confirming that the collagenase 3Ј-UTR contained transcript stabilizing sequences (Fig. 5, left panel). In a subsequent experiment, the half-life of pCMVc-fos/case 3Ј-UTR was determined to be 6 h (Fig. 5, middle panel), comparable to the known half-life of native collagenase 3 in Ob cells (16). As reported with native collagenase, cortisol stabi-lized the transcript, increasing its half-life to 16 h. In the presence of cortisol, mutations of AREs present in the collagenase 3Ј-UTR dramatically shortened the life of the transcript so that the half-life of pCMVc-fos/⌬ case 3Ј-UTR was 6 h, as compared with 16 h for the wild-type construct, p Ͻ 0.01. In fact, the half-life of the mutant pCMVc-fos/⌬ case 3Ј-UTR transcript was not significantly prolonged by cortisol (p Ͼ 0.05), confirming that ARE sequences present in the collagenase 3Ј-UTR were required for the cortisol stabilizing effect (Fig. 5, right panel). ARE sequences in the ϩ1733-1933 and ϩ2133-2333 regions of the pCMVcfos/case 3Ј-UTR were mutated and tested individually in transiently transfected Ob cells. The half-life of c-fos transcripts containing a 3Ј-UTR with single ARE mutations was not different from that of the non-mutated pCMVcfos/case construct after transcriptional arrest (data not shown).
To identify cortisol-regulated proteins binding to ARE sequences present in the collagenase 3 RNA, cytosolic extracts from Ob cells treated with cortisol were subjected to S100 ultra-centrifugation and Accell QMA ion-exchange (Fig. 6). The non-retained protein fraction, which contained RNA binding activity, was purified further by RNA affinity chromatography using the S-S-biotinylated collagenase 3 RNA sequences ϩ1733-1933 and ϩ2133-2333. Following dissociation of the RNA-protein complex, the RNA-binding proteins were fractionated by SDS-polyacrylamide gel electrophoresis. Crude (not shown) and S100 cytosolic extracts, the flow through of the QMA ion-exchange chromatography and the proteins recovered after RNA affinity purification formed complexes migrating at ϳ65, 105, and 130 kDa when incubated with radiolabeled ϩ2133-2333 RNA sequences and subjected to UV light crosslinking (Fig. 6). To identify the isolated proteins, RNA affinity purified material was fractionated by SDS-polyacrylamide gel electrophoresis and stained with Coomassie Blue, which revealed five bands migrating at 65, 90, 105, 115, and 150 kDa (Fig. 6). These were excised and subjected to mass spectroscopy analysis. The proteins identified were vinculin (M r 117), FUSEbinding proteins 1 and 2 (M r 70 and 75), gelsolin (M r 80), enolase (M r 45), annexin I (M r 37), and microtubule associated protein-2 (M r 300).
To further characterize the proteins binding to the collagenase 3 RNA, cytosolic extracts from control and cortisol treated cultures were incubated with radiolabeled collagenase 3 RNA sequences in the presence of specific antibodies to the proteins identified by mass spectroscopy analysis. Antibodies to vinculin and FUSE-binding protein 2 shifted or nearly abolished the protein complex formed with 32 P-labeled collagenase RNA ϩ2133-2333 sequences (Fig. 7). Antibodies to FUSE-binding protein 1, gelsolin, enolase, annexin I, and microtubule-associated protein-2 (not shown) did not modify the RNA-cytosolic protein complex. These findings confirm that FUSE-binding protein 2 and vinculin interact with RNA sequences of the collagenase 3 RNA. To determine whether these proteins contributed to the stabilization of collagenase 3 transcripts, RNAi experiments were conducted. Transfections of double-stranded RNAs coding for FUSE-binding protein 2 or vinculin into Ob cells downregulated the expression of their respective mRNAs (Fig. 8). Unexpectedly, RNAi of FUSE-binding protein 2 and vinculin stabilized collagenase 3 mRNA (Fig. 8), and cortisol did not cause further collagenase 3 mRNA stabilization under conditions of RNAi (not shown).

DISCUSSION
In the present studies, we examined the role of the collagenase 3 3Ј-UTR on RNA stability and the binding of cortisolregulated proteins to AREs within the collagenase 3Ј-UTR. Cytosolic proteins from Ob cells did not bind to the 5Ј-UTR, and no cortisol-dependent binding to the coding region was detected. Cortisol-dependent complexes were formed with the three regions of collagenase 3 3Ј-UTR RNA sequences ϩ1733-1933, ϩ1933-2133, and ϩ2133-2333. Displacement experiments and mutational analysis identified AREs as the elements responsible for the protein-RNA complexes. This is in accordance with findings by other investigators, since AREs frequently bind cytosolic proteins and determine RNA stability (18,42,43). Functional assays confirmed that AREs present in the collagenase 3 3Ј-UTR play a role in RNA stabilization and are responsible for the effect of cortisol on collagenase mRNA stability since their mutation results in loss of the stabilizing effect of cortisol. Protein purification of cytosolic extracts by S100 centrifugation, ion-exchange and RNA affinity chroma-tography identified various cytosolic proteins. Gel supershift assays revealed that FUSE-binding protein 2 and vinculin interacted with the collagenase 3 3Ј-UTR. Surprisingly, RNAi gene knockdown experiments revealed that down-regulation of vinculin and FUSE-binding protein 2 resulted in stabilization of collagenase 3 mRNA, possibly indicating an independent effect of these two proteins in the destabilization of collagenase FIG. 6. Purification scheme utilized for the isolation of RNAbinding proteins from cytosolic extracts of Ob cell cultures, and RNA gel mobility shift assay characterization and protein profile of cytosolic extracts following purification. In the middle panel, cytosolic extracts were purified sequentially by S100 centrifugation (S100), QMA ion-exchange (QMA), and RNA affinity (RA) chromatography and were incubated with 32 P-labeled ϩ2133-2333 collagenase 3 RNA sequence, UV light cross-linked, and fractionated by electrophoresis on denaturing gels, and visualized by autoradiography. Arrows indicate the location of specific RNA-protein complexes displaced by preincubation with homologous RNA sequences in 500-fold excess. Molecular size in kDa is indicated on the left. In the right panel, 350 g of RNA affinity-purified protein were fractionated by SDS-polyacrylamide gel electrophoresis and stained with Coomassie Blue. Arrows indicate protein bands excised for analysis by mass spectroscopy. Molecular size estimates in kDa are shown to the left of the gel. with double-stranded (ds) scrambled RNAs (control) or dsRNAs to silence vinculin and FUSE-binding protein 2 (RNAi), recovered in serum-containing medium for 48 h, and grown to confluence. On the left panel, RNA was subjected to Northern blot analysis and hybridized with ␣-32 Plabeled FUSE-binding protein 2 (KSRP) or vinculin and visualized by autoradiography. On the right panel, cells were treated with 5,6-dichlorobenzimidazole riboside (DRB), and RNA from control (black circles) or silenced genes (RNAi, white circles) was subjected to Northern blot analysis and hybridized with ␣-32 P-labeled collagenase 3 cDNA, visualized by autoradiography, and quantitated by densitometry. Values are means Ϯ S.E. for four cultures, and are presented as percent collagenase mRNA levels relative to the time of DRB addition. Slopes were statistically different. The inset shows a representative Northern blot analysis demonstrating collagenase mRNA decay after DRB. 3 transcripts. Whereas the results confirm the relevance of FUSE-binding protein 2 and vinculin to collagenase 3 mRNA stability, they did not permit testing whether the two proteins played an intermediary role in the effect of cortisol on collagenase 3 mRNA stabilization.
Vinculin, a 117-kDa cytoskeletal protein present in focal cell contacts, regulates cell shape and adhesion, and as such, it is considered to play a role in osteoclastic adhesion and function (44 -46). The association of vinculin with collagenase 3 3Ј-UTR is in accordance with a role for cytoskeletal proteins in the targeting of distinct mRNAs to specific subcellular compartments, an event central to the post-transcriptional regulation of protein expression. The localization to specific cellular compartments may involve the targeting of an mRNA to a specific pool of polysomes (47,48). The 3Ј-UTR plays a key role in the targeting of transcripts to cytoskeletal bound polysomes, where transcripts can associate with cytoskeletal proteins, such as actin and possibly vinculin, via their 3Ј-UTR (47). Consequently, the 3Ј-UTR not only participates in transcript stabilization, but also in the targeting of the transcript, and the two processes may involve vinculin, a cytoskeletal protein.
The FUSE-binding proteins (FBP) are a family of three regulatory proteins, termed FBP 1, FBP 2, or KH-type splicing regulatory protein (KSRP), and FBP 3 (49). FUSE-binding protein-1 was initially characterized as a protein targeting the far upstream element, a positive cis-element of the human c-myc gene (50). FUSE-binding proteins preferentially bind to single-stranded DNA and to RNA sequences, and are known to act as transcription factors, but have been postulated to regulate transcript stability (51). The nucleic acid recognition domain of FUSE-binding proteins displays an array of KH motifs, which are RNA and single strand nucleic acid recognizing units (49,50,52). FUSE-binding proteins are expressed by a variety of tissues, although the level of expression of FUSE-binding protein-1, -2, and -3 varies among tissues, suggesting different regulatory mechanisms for each protein (49). FUSE-binding protein-1 is necessary for Myc expression and cell proliferation, and its expression is developmentally regulated, and is dramatically reduced in terminally differentiated cells (50,(53)(54)(55). Since glucocorticoids prevent terminal differentiation of Ob cells, it is possible that the expression of FUSE-binding proteins is regulated, directly or indirectly, by these steroids (4). Interestingly, basal collagenase 3 expression also is dependent on the stage of osteoblastic cell differentiation, and its expression is maximal in differentiated cells (13). This would be opposite to the predicted expression of FUSE-binding protein-1, but less is known about the expression of FUSE-binding protein 2, which associates with the collagenase 3Ј-UTR.
FUSE-binding protein 2 or KSRP was identified as a component of a protein-RNA complex assembled on the downstream control sequence (DCS) complex of the c-src gene, and found to have splicing regulatory properties (56,57). Although other cytosolic proteins bind to a 33 nucleotide core of DCS, FUSEbinding protein 2 or KSRP appears to be the only one to stabilize the protein-RNA complex. Recent work in isolated exosomes reveals that FUSE-binding protein 2 or KSRP not only binds to ARE-containing mRNAs, but it serves to recruit the exosome and to destabilize this class of mRNAs (58). Therefore, the destabilizing effect of FUSE-binding protein 2 or KSRP and vinculin on collagenase 3 RNA in osteoblasts is in accord with these findings. In preliminary experiments, cortisol did not change FUSE-binding protein 2 or vinculin mRNA levels in Ob cell cultures (data not shown), but this does not preclude an effect on activation, binding or degradation of these proteins by cortisol. The fact that the same factor could bind to DNA and RNA and regulate transcriptional and post-transcriptional events is not surprising, since often these proteins shuttle between the nucleus and the cytoplasm (59). Glucocorticoids induce apoptosis of mature osteoblasts, and collagenase 3 may play a role in regulating osteoblastic apoptosis (60,61). Patterns of protein expression during osteoblastic death have not been reported, but Jurkat T cells and the Burkitt lymphoma cell line BL60 overexpress RNA-binding proteins during apoptosis, including FUSE-binding protein 2 and heterogeneous nuclear ribonucleoproteins (62,63). It is possible that these proteins play a role in collagenase 3 mRNA stability and protease expression during apoptosis, an event that may have a significant function in tissue breakdown as cells die.
In conclusion, cortisol induces the formation of cytosolic protein-collagenase 3 3Ј-UTR complexes that probably play a role in mRNA stabilization. Cortisol stabilizes collagenase 3 mRNA in osteoblasts, and AREs in the collagenase 3 3Ј-UTR are required for the effect of cortisol on collagenase 3 transcript stability. FUSE-binding protein 2 and vinculin may play a role in the effects of cortisol, but have independent destabilizing activity on collagenase 3 transcripts.