Peroxisome Proliferator-activated Receptor-γ Down-regulates Chondrocyte Matrix Metalloproteinase-1 via a Novel Composite Element*

Interleukin-1β (IL-1β) induces degradation via hyperexpression of an array of genes, including metalloproteinases (MMP), in cartilage cells during articular degenerative diseases. In contrast, natural ligands for peroxisome proliferator-activated receptors (PPARs) display protective anti-cytokine effects in these cells. We used the PPAR agonist rosiglitazone (Rtz) to investigate PPAR-γ isotype on IL-1β-target genes. Immunocytochemistry, electrophoretic mobility shift, and transient transfection assays revealed a functional PPAR-γ in chondrocytes in vitro. Rtz displayed significant inhibition of IL-1β effects in chondrocytes. Low Rtz concentrations (close to Kd values for PPAR-γ, 0.1 to 1 μm) inhibited the effects of IL-1β on 35S-sulfated proteoglycan production and gelatinolytic activities and downregulated MMP1 expression at mRNA and protein levels. We have investigated the mechanism of action of Rtz against IL-1β-mediated MMP1 gene hyperexpression. Rtz effect occurs at the transcriptional level of the MMP1 promoter, as observed in transiently transfected cells with pMMP1-luciferase vector. Transient expression of wild type PPAR-γ enhanced Rtz inhibitory effect in chondrocytes, whereas a mutated dominant negative PPAR-γ abolished it, supporting the role of PPAR-γ in this effect. MMP1 gene promoter analysis revealed the involvement of a cis-acting element located at -83 to -77, shown to be a composite PPRE/AP1 site. Gel mobility and supershift assays demonstrated that PPAR-γ and c-Fos/c-Jun proteins bind this cis-acting element in a mutually exclusive way. Our data highlight a new PPAR-γ-dependent inhibitory mechanism on IL-1β-mediated cartilage degradation occurring through DNA binding competition on the composite PPRE/AP1 site in the MMP1 promoter.

PGJ2), exert anti-IL-1␤ effects in cartilage cells in vivo and in vitro. Intra-articular injection of 15d-PGJ2 inhibits adjuvantinduced arthritis in rats (1) and addition of 15d-PGJ2 to IL-1␤treated human or rat chondrocytes in culture counteracts IL-1␤induced effects (2)(3)(4). As IL-1␤ plays a central role in the degradation of cartilage in rheumatoid arthritis and osteoarthritis (5,6), the above data suggest a role for PPAR-␥ in cartilage homeostasis leading to new directions in therapeutic research.
PPAR-␥ is a member of the nuclear receptor superfamily of ligand-dependent transcription factors. PPAR-␥ forms a heterodimeric complex with the retinoid X receptor (7) and binds to specific nucleotide motifs (direct repeats with single spacing, DR1) located in the promoter of target genes. Natural ligands such as 15d-PGJ2 have been shown to activate PPAR-␥, yet the true physiological ligands of this receptor remain elusive. Numerous synthetic PPAR-␥ activators have been identified, including thiazolidinediones (such as rosiglitazone (Rtz)) used in the treatment of type II diabetes (8). PPAR-␥ was first demonstrated to regulate adipocyte differentiation, and it has been recognized recently (9) to be implicated in the modulation of cancer cells growth and inhibition of the inflammatory process in immunocompetent cells (10 -13) and other cell types (14), including cartilage cells (2)(3)(4). Most anti-inflammatory effects have been observed with 15d-PGJ2, a poor agonist of PPAR-␥, whereas synthetic PPAR-␥ ligands such as Rtz were less effective or not active, despite their higher affinity for the receptor (Rtz K d Յ 100 nM). Thus, the question regarding the exact role and the implication of PPAR-␥ in these effects has been largely discussed and documented in cells from the immune system. Overexpression of PPAR-␥ in macrophages maximized the ability of diverse PPAR-␥ agonists to inhibit the expression of inflammatory response genes consistent with a PPAR-␥-mediating anti-inflammatory mechanism (15,16). Other data indicated PPAR-␥-independent anti-inflammatory effects. Treatment of macrophages derived from PPAR-␥-null embryonic stem cells with high concentrations of synthetic PPAR-␥ ligands was reported recently (17) to inhibit the induction of the inducible nitric-oxide synthase and COX2. Similar data were observed in the RAW 264.7 macrophages (18 -20).
In sharp contrast, the role of Rtz-activated PPAR-␥ as an antagonist of IL-1␤-induced cartilage degradation is largely unknown. The aim of this work was to analyze this role through the effects of the hypolipidemic drug rosiglitazone on IL-1␤-induced target genes active in the degradative process of matrix proteins synthesized by culture chondrocytes. We report a PPAR-␥-dependent inhibitory mechanism on IL-1␤-induced MMP1 expression acting through a new composite PPRE/AP-1 site at positions Ϫ83 to Ϫ71 upstream from the MMP1 rabbit gene transcription start site.

EXPERIMENTAL PROCEDURES
Isolation and Culture of Chondrocytes from Rabbit Articular Cartilage-Articular chondrocytes were cultured as described previously (21). Briefly, 5-week-old Fauve de Bourgogne rabbits (CPA, Orleans, France) were killed, and the femoral, humeral, and tibial articular cartilages were removed under sterile conditions. Thin slices of cartilage were sequentially digested with 0.2% hyaluronidase and 0.2% collagenase in Geys' balanced salt solution (Eurobio, Les Ulis France). The resulting cell suspension was plated at 10 5 cells per cm 2 (high density) into 75-cm 2 flasks or 6-well plates and incubated in Ham's F-12 medium (Invitrogen) containing 10% fetal calf serum (FCS), 100 IU/ml penicillin, and 100 g/ml streptomycin. Cells were then cultured at 37°C in an atmosphere of 8% CO 2 in air, and the medium was changed once until the cells were confluent (day 6 of the primary culture). The medium was replaced with 2% FCS/Dulbecco's modified Eagle's medium (DMEM) (Invitrogen) to which was added either IL-1␤ alone (Immugenex Corp., Los Angeles) or combined with Rtz at concentrations ranging between 0.1 and 10 M. Cells were incubated for 18 h and analyzed as below.
Immunological Studies-Chondrocytes were cultured in 4-chamber labtek slides (Merck) and incubated for 18 h in 2% FCS/DMEM with or without 1 M Rtz. After being washed in PBS, cells were fixed with 4% paraformaldehyde in 0.1 M sodium phosphate and washed in blocking buffer (PBS with 5% bovine serum albumin and 0.2% Tween 20). Fixed cells were then exposed to polyclonal PPAR-␥ antibody (Santa Cruz Biotechnology) diluted at 1:100 in blocking buffer for 2 h at room temperature (20°C) in a humid chamber. After a quick rinse in 0.1% PBS/Tween, cells were exposed to the rabbit anti-goat Alexa 546-coupled secondary antibody (Molecular Probes and Invitrogen) diluted at 1:800 for 2 h. Cells were rinsed in PBS/Tween and mounted with Vectashield. The slides were observed with a Nikon Labophot 2 and photographed with a Spot camera driven by the Spot Jr software (Diagnostic Instrument). Controls were made by omission of primary antibody.
Transient Transfection and Luciferase Assay-Rabbit chondrocytes were cultured in Ham's F-12 containing 10% FCS as described above, trypsinized, and plated in 6-well plates. Cells were then transfected at 70% confluence via a calcium phosphate precipitation method. Briefly, cells were incubated for 4 h with the precipitate obtained with 0.02 M CaCl 2 in a Hepes buffer, pH 7.0, and 1.5 g of DNA/well, i.e. either by adding 1 g of the 3xPPRE-TK-Luc construct (a gift from Dr. P. Grimaldi, Nice, France) or the chimeric MMP1 promoter-luciferase construct with 500 ng of ␤-galactosidase expression vector. Each point was made in triplicate. Cells were then lysed in 400 l of lysis buffer (Promega, France). Luciferase assay was performed using 20 l of cell lysis on a Lumat LB 9507 luminometer (Berthold, France). Efficiency of the transfection was shown to be 50 -70% of total cells, as determined by ␤-galactosidase assay (Promega, France).

Measurement of [ 35 S]Sulfate Incorporation into Proteoglycans (PG)-
The incorporation of [ 35 S]sulfate into PG by articular chondrocytes was assessed as described previously (24). At confluency, chondrocytes were incubated in 2% FCS/sulfate-free DMEM plus 1.5 Ci/ml Na 2 [ 35 SO 4 ] (75 MBq/ml, Amersham Biosciences) for 18 h in the presence or absence of effectors. Controls were prepared without any additive compound. Each point was prepared in triplicate. Proteoglycans were extracted from culture media and cell layer at 4°C by incubation with 3 M guanidinium chloride in 0.05 M Tris-HCl (Merck), pH 7.4, in the presence of protease inhibitors (10 mM Na 2 EDTA (Sigma), 5 mM benzamidine (Merck), and 0.1 mM 6-aminohexanoic acid (Prolabo, Fontenay, Bois, France). Aliquots of the guanidinium extract were then spotted on a set of Whatman 3MM paper and allowed to dry. The set was used to quantify total high molecular weight/highly charged sulfated PG subunits, which were precipitated with 5% cetylpyridinium chloride in 0.3 M NaCl (Sigma) according to Larsson and Kuettner (25). Scintillation liquid was added to each strip, and the strips were counted in a Packard tricarb ␤-spectrometer. Each measurement was made in triplicate. For each experiment, the amount of DNA was measured in sister flasks by the fluorometric method described by Kapuscinski and Skoczylas (26). Results were calculated as mean Ϯ S.D. cellular or medium dpm/10 g of DNA of four similarly treated wells. Results are presented as total radiolabeled PG expressed in percent of controls performed with cells from three different animals.
Western Blotting-Proteins were extracted from culture medium by addition of 5 volumes of cold acetone and centrifuged at 12,000 ϫ g for 5 min. The pellet was collected in 10 mM phosphate buffer containing protease inhibitors. Proteins were size-separated by SDS-PAGE in 10% polyacrylamide gel. A Colored Rainbow mix, code RPN 756 (Amersham Biosciences), was used as the molecular weight marker. The gel was electroblotted to nitrocellulose (Bio-Rad) in transfer buffer (25 mM Tris-HCl, 192 mM glycine, pH 8.3) in a Transblot Cell (Bio-Rad). Equal transfer of proteins was confirmed by staining the nitrocellulose with Ponceau Red (0.2% w/v H 2 O diluted 99:1 in 10% trifluoroacetic acid). The MMP1 antibody (polyclonal goat anti-human MMP1, Santa Cruz Biotechnology (Tebu, France)) was diluted at 1:200 and used as primary antibody overnight at 4°C. A peroxidase-conjugated rabbit anti-goat IgG (Dako P0160) was used as secondary antibody at 1:2000 and incubated for 1 h at room temperature. Blots were revealed by enhanced chemiluminescence as instructed by the manufacturer (ECL, Amersham Biosciences). Semi-quantitative scanning densitometry was performed with the NIH image software (results were standardized using Ponceau Red staining for homogeneity).
Production of Rabbit cDNA Probes by RT-PCR-MMP probes were generated by RT-PCR in accordance with the human sequences, selected in a domain conserved across species for MMP1, MMP3, and MMP13 (GenBank TM accession numbers M17821, M25664, and M002427, respectively). Total RNA was prepared from control or IL-1␤ (10 ng/ml)-treated rabbit chondrocytes and synthesized into cDNA using Moloney murine leukemia virus-reverse transcriptase (Promega). Briefly 1 g of total RNA was reverse-transcribed at 42°C for 1 h in a 25-l reaction according to the manufacturer's instructions. The resulting cDNA was amplified for PCR on a DNA thermal cycler (PerkinElmer Life Sciences) using the following program: 96°C for 5 min, followed by 35 cycles of 96°C for 30 s, 55-60°C for 30 s, and 72°C for 1 min. A 15-min 72°C extension step completed the reaction. PCR products were subcloned in pGEM vector.
Mutagenesis-All mutants were generated with the Quick Change site-directed mutagenesis kit from Stratagene (France). The dominant negative PPAR-␥ was constructed following the procedure (27) described by Gurnell et al. (28) in which Leu 468 and Glu 471 were mutated to alanine in the ligand binding domain. A double point mutation (AAAGCA to AGTAGT) was introduced into the proximal PPRE-like site Ϫ83 to Ϫ78 to generate the pGL-MMP1-⌬PPRE, by using the forward sequence primer 5Ј-TCAAGAGGATGTTATAGTAGTTGAGT-CACACAGC-3Ј, corresponding to region Ϫ99 to Ϫ65 of the rabbit MMP1 promoter.
DNA Sequencing-Automated sequencing was performed by Genome Express to confirm the identity of all DNA probes and constructs used in the study.
RNA Isolation and Northern Blot-Total RNA was extracted from cells by a single-step method using Trizol reagent (Invitrogen). Samples of 10 g of total RNA were used for Northern blot analysis. The rabbit MMP1, -3, and -13 cDNA probes were synthesized by RT-PCR. All probes were labeled by random priming with [␣-32 P]dCTP (Amersham Biosciences) to a specific activity of 10 8 to 10 9 cpm/g. Results were normalized by using a human glyceraldehyde-3-phosphate dehydrogenase probe and quantified using Studio Scan II (Agfa, Paris, France) with the NIH software.
Statistical Analysis-Values are expressed as the mean Ϯ S.D. Statistics were performed by analysis of variance. Multiple comparisons were performed with the Tukey test. Differences between groups were considered significant for p Յ 0.05.

Presence of Functional PPAR-␥ in Cultured Articular
Chondrocytes-PPAR-␥ displayed nuclear localization in our in vitro model of cultured rabbit chondrocytes, independently of Rtz treatment (Fig. 1A). EMSA was carried out with nuclear extracts from 1 M Rtz-treated chondrocytes and an oligonucleotide containing the canonical PPRE sequence described under "Experimental Procedures." Fig. 1B showed the involvement of the PPRE-containing oligonucleotide in a robust protein-DNA complex. Binding specificity was confirmed by using an 80-fold molar excess of the unlabeled oligonucleotide. This complex was supershifted in the presence of anti-PPAR-␥ antibody. Finally, Rtz increased luciferase activity in chondrocytes transiently transfected with a PPRE-luciferase gene reporter, in a dose-dependent manner from 0.1 to 1 M. Maximum activity (4-fold induction) was observed at 1 M and stayed in plateau up to 10 M (Fig. 1C).
Low Concentrations of Rtz (Յ1 M) Selectively Inhibit IL-1␤induced Proteoglycan Degradation, Gelatinolytic Activities, and MMP1 Production-Chondrocytes were incubated for 18 h in sulfate-free medium in the presence of 35 S-radiolabeled sulfate with or without effectors. 35 S-Sulfate incorporation into cetylpyridinium chloride-precipitable PG was measured in the culture medium. Fig. 2A shows that in the presence of IL-1␤ alone total 35 S-sulfated PG were decreased by 35 Ϯ 10%, from 6.35 Ϯ 0.56 10 4 dpm/10 g DNA in control conditions to 4.12 Ϯ 0.39 10 4 dpm/10 g DNA (p Ͻ 0.01). Addition of Rtz reproducibly induced a dose-dependent stimulation of 35 S-sulfated PG and fully restored control values at doses as low as 1 M, being stable at 10 M. These results show that low concentrations (Յ1 M) of Rtz counteract the detrimental effect of IL-1␤ on proteoglycans production.
We then asked whether this inhibitory effect involved the inhibition of the documented induction of matrix metalloproteinases production induced by IL-1␤ in chondrocytes. Among them MMP1 can be studied by gelatinolytic activity at an apparent molecular mass of 53/58 kDa. In our control conditions, chondrocytes secreted two main gelatinolytic activities at apparent molecular masses of 92 and 72 kDa, presumably corresponding to MMP9 and MMP2, respectively. gelatinolytic activity was decreased by 40 Ϯ 3% and by 50 Ϯ 5% (p Ͻ 0.01) with 1 and 10 M Rtz, respectively. The IL-1␤induced 53/58-kDa gelatinolytic activity was decreased by 35 Ϯ 5% and by 55 Ϯ 7% (p Ͻ 0.01) with 0.5 and 1 or 10 M Rtz, respectively. Similar results were observed at the protein level as shown by Western blot analysis. Fig. 2C is representative of three separate experiments. Scanning densitometry of the bands showed that the mean Ϯ S.D. of IL-1␤ overexpression of the MMP1 signal (apparent molecular mass, 52 kDa) was significantly decreased by Rtz in a dose-dependent manner, i.e. by 41 Ϯ 3.8% (p Ͻ 0.01), 70 Ϯ 11% (p Ͻ 0.001), and 81 Ϯ 6% (p Ͻ 0.001), with 0.1, 0.5, or 1 M Rtz, respectively. These concurring data strongly suggest that the gelatinolytic activity observed at 53/58 kDa corresponds to the MMP1 protein.
The Inhibitory Effect of Rtz on IL-1␤-induced MMP1 mRNA Expression Is a Direct Mechanism-Steady state levels of MMP1 mRNA were analyzed by Northern blotting, as compared with MMP3 and MMP13 mRNAs. As shown in Fig. 3A, MMP1, -3, and -13 mRNA levels were close to detection limits in control conditions, but were strongly expressed in the pres-ence of IL-1␤. Co-treatment of Rtz and IL-1␤ decreased mRNA levels of MMP1, -3, and -13 by 25, 10, and 15%, respectively (0.5 M Rtz), and by 70, 40, and 55% (1 M Rtz) (Fig. 3B). This effect did not appear mediated by protein synthesis, as cycloheximide did not modify the inhibitory effect of Rtz on MMP1, -3, or -13 mRNA levels (Fig. 3A). These data suggest that down-regulation of IL-1␤-induced proteoglycan degradation by Rtz may be mediated by inhibition of MMP production by chondrocytes. We then focused on the Rtz effect on MMP1 gene transcription.
The Inhibitory Effect of Rtz on IL-1␤-induced MMP1 Gene Is Transcriptional and Requires PPAR-␥-Chondrocytes were transiently transfected with a pMMP1-luciferase construct containing 3448 bp of the 5Ј-flanking region of the rabbit MMP1 gene linked to the luciferase gene and supplemented with IL-1␤ and/or Rtz as described under "Experimental Procedures." Each point was done in triplicate, and this experiment was performed three times with different animals for the dose-response curve and twice for the dominant negative experiments. Fig. 4 showed that luciferase activity was not modified by addition of 1 M Rtz alone, whereas it was increased 4-fold by IL-1␤. Co-treatment of chondrocytes with IL-1␤ and increasing concentrations of Rtz led to a dose-dependent reduction of MMP1 promoter activity corresponding to 23 Ϯ 3% and 42 Ϯ 3% (n-1 ϭ 17, p Ͻ 0.001) inhibition with 0.5 and 1 M Rtz, respectively, as compared with IL-1␤-treated cells. In addition, a more pronounced inhibitory effect of Rtz (60 Ϯ 1% inhibition, n-1 ϭ 11 p Ͻ 0.001) was observed in cells overexpressing the wild type PPAR-␥. The involvement of PPAR-␥ in the inhibitory effects of Rtz on MMP1 transcription was further demonstrated by using a mutant dominant negative of PPAR-␥ (DN-PPAR-␥). Preliminary experiments showed that the DN-PPAR-␥ mutant abolishes the stimulatory effect of Rtz on a canonical PPREluciferase construct when co-expressed in chondrocytes (data not shown). As shown on Fig. 4, the inhibitory effect of Rtz on pMMP1-Luc activity was completely abolished by co-expression of DN-PPAR-␥, supporting the implication of PPAR-␥ in this effect.

FIG. 2. Low concentrations of Rtz selectively inhibit IL-1␤induced proteoglycan degradation, gelatinolytic activities, and MMP1 production.
Rabbit articular chondrocytes were cultured at high density and incubated for 18 h in 2% FCS/DMEM with or without 10 ng/ml IL-1␤ alone or combined with increasing concentrations of Rtz from 0.5 to 10 M. A, production of total 35 S-radiolabeled proteoglycans. Chondrocytes were cultured as described above. Total PG production was assayed after 18 h of incubation of the cells in sulfate-free medium containing either IL-1␤ alone or combined with increasing concentrations of Rtz. Each point was done in triplicate, and total PG was expressed as percent of nontreated cells. Data are mean Ϯ S.D. of five experiments performed with cells from different animal donors. **, p Ͻ 0.01 as compared with cells treated with IL-1␤ alone. B, gelatinolytic activities. Culture medium was collected and subjected to gelatin zymography as described under "Experimental Procedures." The levels of gelatinolytic bands were determined by scanning video densitometry. The 72-kDa band is not modified by IL-1␤ or Rtz and was used as internal control. C, Western blot confirmed the expression of MMP1 as a single band identified at ϳ55 kDa. This figure is representative of one of three separate experiments performed with different animals. Scanning densitometry of the MMP1 band observed in culture medium from chondrocytes treated with IL-1␤ alone or combined with 0.1, 0.5, or 1 M Rtz was compared with that observed in nontreated cells. Equal loading of protein was confirmed by Ponceau Red staining. Molecular mass marker sizes are indicated on the right.

FIG. 3. Low concentrations of Rtz counteract IL-1␤-induced mRNA expression of MMP1, -3, and -13 by a direct mechanism.
Cells were incubated as described in Fig. 2

The Inhibitory Effect of Rtz on MMP1 Promoter Transcription Involves a Functional PPRE Situated at
Ϫ83 to Ϫ71-We have identified a degenerated PPRE sequence (AAAGCAT-GAGTCA) in the 3.45-bp region of the rabbit MMP1 promoter, located at Ϫ83 to Ϫ71 (GenBank TM accession number M17820). EMSA analysis showed the formation of a thin retarded complex between the PPRE and nuclear extract from nontreated chondrocytes (Fig. 5A, lane 1). Addition of Rtz strongly enhances this complex (Fig. 5A, lane 2). The specificity of binding activity was confirmed by competition experiments in the presence of an excess of either nonradioactive wild type PPRE or consensus PPRE oligonucleotide (ACOA sequence) (Fig. 5A,  lanes 3 and 4). The signal was reduced and supershifted in the presence of an anti-PPAR-␥ antibody supporting the presence of PPAR-␥ in the complex (Fig. 5A, lane 5). Mutation into the PPRE sequence resulted in oligonucleotides displaying no PPAR-␥ binding activity and, consequently, no supershifted signal (Fig. 5A, lanes 6 -9). Finally, the pMMP1-luciferase construct, bearing the mutated PPRE, was transiently transfected in chondrocytes. This experiment was done twice. Data shown on Fig. 5B showed that this mutation leads to the loss of Rtz inhibitory effect on IL-1␤-mediated MMP1 transactivation. Together, these data demonstrate the following: (i) Rtz-activated PPAR-␥ binds the PPRE sequence situated at Ϫ83 to Ϫ71 in the MMP1 promoter, and (ii) PPAR-␥ binding is fundamental in the inhibition of IL-1␤-induced MMP1 transcription by Rtz.
The PPRE Sequence Situated at Ϫ83 to Ϫ71 in the MMP1 Promoter Is a Composite PPRE/AP1 Element and Promotes a Mutually Exclusive PPAR-␥ or AP1 Binding-This Ϫ83/Ϫ71 PPRE sequence in the MMP1 promoter overlaps the proximal AP1 site at Ϫ77 (TGAGTCA), which is considered a key element in IL-1␤-induced MMP1 expression. EMSA analysis using the composite PPRE/AP1 sequence was performed to highlight the respective involvement of AP1 and PPAR-␥ binding activity. As shown in Fig. 6A, nuclear extracts from chondrocytes treated with IL-1␤ alone elicited a DNA binding complex (lane 2) that was partly supershifted with either the anti-Fos family members (lane 3) or the specific anti-c-Jun antibody (lane 4). In order to characterize the nature of the Fos family member involved, we used specific antibodies against c-Fos, Fra-1, and Fra-2. The association of c-Fos and c-Jun antibodies totally supershifted the signal (Fig. 6A, lane  5). The isolated specific anti-c-Fos antibody substantially supershifted the signal (Fig. 6A, lane 6), whereas anti-Fra-1 or anti-Fra-2 were ineffective (lanes 7 and 8). In addition, the AP1 DNA binding complex was not supershifted upon addition of PPAR-␥ antibody (Fig. 6A, lane 9). These data show that IL-1␤-induced AP1 complex contains c-Fos/c-Jun proteins but no PPAR-␥-binding protein.
Another set of experiments was performed to study the fate of the AP1 complex when Rtz was added together with IL-1␤. As seen on Fig. 6B, the AP1 complex observed with IL-1␤ alone (lane 1) was efficiently competed with a 50-and 100-fold molar excess of unlabeled Ϫ83/Ϫ71 PPRE sequence (lanes 2 and 3). By comparison, this complex was not elicited by nuclear extracts from chondrocytes co-treated with IL-1␤ and Rtz, whereas a new retarded species was observed (Fig. 6B, lane 4). This new band was competed with unlabeled Ϫ83/Ϫ71 PPRE sequence (Fig. 6B, lane 5) and was supershifted after addition of PPAR-␥ antibody (lane 6). No supershift of this complex was observed with PPAR-␣ (Fig. 6C, lane 3) nor PPAR-␤ (lane 4) antibodies, indicating that PPAR-␥ was present in this new complex without other PPAR isotypes. Finally, neither c-Fos nor c-Jun antibodies (Fig. 6C, lanes 6 and 7) modified this signal, confirming the absence of AP1 proteins in this complex. Taken together, these data demonstrate a mutually exclusive binding of PPAR-␥ and AP1 to the composite element Ϫ83 to Ϫ71 of MMP1 promoter.
We then investigated whether the PPAR␥ inhibitory effect on IL-1␤-induced MMP1 transcription might involve NF-B transcription factors. EMSA analysis (Fig. 7) using an oligonucleotide bearing the Ϫ3029 NFB-RE motif from the MMP1 promoter showed that no complex was observed using nuclear extracts from nontreated or Rtz-treated cells (lanes 1 and 2). IL-1␤ treatment of chondrocytes elicited the formation of two DNA binding complexes (Fig. 7, lane 3), which were mostly supershifted with anti-P50 antibodies (lane 4). The lower mobility band was partly supershifted with anti-P65, whereas the higher mobility band was not modified (Fig. 7, lane 5). These data indicate that the lower mobility band mainly consists of the p65-p50 heterodimer, whereas a p50 homodimer is present in the higher mobility band. Rtz treatment had no significant effect on any complex (lanes 6 -8). In conclusion, no detectable cross-talk occurs between IL-1␤-induced NF-B binding activity and Rtz-activated PPAR-␥ at the Ϫ3029 NFB-RE. DISCUSSION We report here on the anti-cytokine effects of PPAR-␥ activation in rabbit cartilage cells. Low concentrations (Յ1 M) of the classical pharmacological ligand Rtz selectively inhibit IL-1␤-induced proteoglycan degradation and MMP1 expression. The latter Rtz inhibitory effect targets the rabbit MMP1 promoter via a novel composite element. PPAR-␥ is constitutively expressed in cultured rabbit articular chondrocytes as shown by immunocytochemistry in Fig. 1A. EMSA experiments and chondrocyte transient transfections with a 3X-PPRE-Luciferase construct demonstrated that endogenous PPAR-␥ is functionally activated by low doses of Rtz (0.1-1 M) with a maximum activity at 1 M (Fig. 1B). This is at variance with the results of Fahmi et al. (3) who found a maximum stimulating effect of such a reporter construct in cultured human chondrocytes by using a 100-fold higher concentration of Rtz. This discrepancy may relate to a lower PPAR-␥ content in cultured human chondrocytes compared with rabbit cells.
We studied the effects of Rtz on a series of IL-1␤-induced genes. Low concentrations of Rtz (0.1 to 1 M) negatively regulated specific subsets of IL-1␤-induced effects in chondrocytes, whereas Rtz alone displayed no effect. A significant decrease of proteoglycan degradation was associated with down-regulation of 53/58-kDa gelatinolytic activity, MMP1 mRNA expression, and MMP1 protein production (Fig. 2). Most interesting, low concentrations of Rtz did not affect IL-1␤-induced COX 2 mRNA expression nor nitrite production, whereas both inflammatory markers, as well as MMP1 expression, were reduced with high doses (10 M) of Rtz (data not shown). This pharmacological variance supports the notion that Rtz differently modulates IL-1␤-induced inflammation and degradation in cartilage cells. Inhibition of MMPs by low concentrations of natural as well as synthetic PPAR-␥ ligands has also been reported in other models (14, 29 -32). The PPAR-␥ mediation of MMPs inhibition has been described but not fully demonstrated. We have investigated the mechanism of the inhibitory effect of Rtz with a focus on IL-1␤-induced MMP1 in chondrocytes.
The repression of IL-1␤-induced MMP1 overexpression by Rtz is dose-dependent, direct, and occurs at the transcriptional level, as observed by transient expression experiments using the pMMP1-Luc vector (Fig. 4). In addition, the Rtz inhibitory effect was enhanced in chondrocytes transiently transfected with the wild type PPAR-␥, whereas it was abolished in cells transfected with a dominant negative mutant PPAR-␥. This further supports the implication of PPAR-␥ in the inhibitory effect of Rtz. PPAR-␥ has been shown to modulate IL-1␤-induced genes by several mechanisms including protein-protein (NF-B, AP1, STAT) or DNA-protein interactions. Chung et al. (33) have shown that PPAR-␥ inhibits NF-B-driven transcription by physically interacting with p50 and p65 proteins as reported previously by Delerive et al. (34) for PPAR-␣. An NF-B-responsive element has been characterized at position Ϫ3029 to Ϫ2994 in the MMP1 promoter, which is required for maximal IL-1␤ transcriptional induction in synovial fibroblasts (23). In the present study, the Ϫ3029 to Ϫ2994 NFB-RE is shown to form DNA complexes involving p50/p50 homodimer and p50/p65 heterodimer proteins. The binding activity of these complexes was not modified in the presence of Rtz (Fig. 7), suggesting that Rtz-activated PPAR-␥ does not interfere with IL-1␤-induced NF-B binding activity in our model. Activated PPAR-␥ also acts through DNA binding interaction with its cognate PPRE sequences (direct repeats with single spacing nucleotide, DR1) located in the promoter of target genes. For example, the activated PPAR-␥ inhibits regulated on activation normal T cell expressed and secreted (RANTES) transcription and protein production in endometriotic stromal cells through a specific PPRE at Ϫ334 to Ϫ322 upstream from the RNA polymerase start site (35). We have identified a degenerated DR1 sequence (AAAGCATGAGTCA) located at Ϫ83 to Ϫ71 in the rabbit MMP1 promoter. Gel mobility shift and supershift assays showed that this cis-acting element binds PPAR-␥, whereas no binding activity was observed with a mutated PPRE (Fig. 5A). In addition, Rtz failed to inhibit MMP1 promoter activity when chondrocytes were transfected with a pMMP1-Luc vector bearing this mutated PPRE (Fig. 5B). Altogether these data demonstrate that the inhibitory effect of Rtz on IL-1␤-induced MMP1 gene transcription implicates PPAR-␥ binding on the Ϫ83 to Ϫ71 cis-acting sequence of the MMP1 promoter.
The next question was to study whether the reduced IL-1␤induced MMP1 expression by Rtz was because of an interaction between PPAR-␥ and AP1 binding activities on the Ϫ83 Ϫ71 sequence of the MMP1 promoter. Fahmi et al. (36) have shown previously that 15d-PGJ2, a lower affinity PPAR-␥ ligand, reduces IL-1␤-induced MMP1 expression in human synovial fibroblasts and inhibits AP1 DNA binding activity. They concluded that 15d-PGJ2 negatively regulated MMP1 expression by interfering with AP1 activation, although neither the PPAR-␥ dependence of the 15d-PGJ2 inhibitory effect nor the PPAR-␥/AP1 interactions were analyzed. Their EMSA experiment describing the decreased AP1 binding activity was performed with an oligonucleotide corresponding to the restricted AP1 sequence Ϫ77-71 of the MMP1 promoter, which does not comprise the full overlapping PPRE Ϫ83-71 sequence. In addition, BRL49653 (Rtz) was much less effective than 15d-PGJ2, with an inhibitory effect only observed at very high concentrations (10 -100 M) far from its K d value for PPAR-␥, and no information was provided about the effect of Rtz on the AP1 binding activity. Altogether these data supported a PPAR-␥independent pathway for the anti-MMP1 effect of the 15d-PGJ2. Indeed, 15d-PGJ2 has been shown to reduce AP1 activation in several anti-inflammatory responses by interacting at multiple levels along the AP1 pathway (37). More recently, Mix et al. (38) showed that 15d-PGJ2 decreased IL-1␤-induced MMP1 expression in the human SW-1353 chondrosarcoma cell line, through a novel PPAR-␥-independent mechanism that requires Smad signaling.
In the present work, a series of EMSA analyses were performed with the composite PPRE/AP1 element in order to highlight the respective involvement of PPAR-␥ and AP1 binding activities, in cells co-treated with IL-1␤ and Rtz (Fig. 6). The IL-1␤-induced AP1 DNA binding activity observed in chondrocytes was shown to contain immunoreactive c-Fos and c-Jun proteins (Fig. 6A). This is at variance with the AP1 complex induced in fibroblast growth factor-stimulated MMP1 in MC3T3-E1 osteoblasts and in PMA-induced MMP1 in rabbit fibroblasts (39,40). In these cases, the AP1 complex was shown to contain Fra-1-or Fra-2-binding proteins at the Ϫ77 site of the MMP1 promoter. However, our data are in accordance with those from Chamberlain et al. (41) and Hamid et al. (42) who showed an up-regulation of MMP1 by PMA or IL-1␤ involving a c-Fos-c-Jun AP1 complex at Ϫ77 to Ϫ69 of the promoter. Finally, the IL-1␤-induced complex bound to the AP1 element, characterized under the present experimental conditions, was not recognized by anti-PPAR-␥ antibodies indicating the absence of PPAR-␥ proteins in this complex.
When nuclear extracts were prepared from IL-1␤ and Rtz co-treated cells, the AP1 complex totally disappeared, whereas a differently shifted complex was observed (Fig. 6B). This new complex is further supershifted in the presence of an anti-PPAR-␥ antibody but not with c-Fos and c-Jun antibodies and not with PPAR-␣ or PPAR-␤ antibodies, indicating that this complex only contain PPAR-␥ (Fig. 6C). Taken together, our data demonstrate that AP1 and PPAR␥ bindings on the composite PPRE/AP1 sequence are mutually exclusive.
Our data are consistent with a new PPAR-␥-dependent inhibitory mechanism of Rtz on IL-1␤-induced MMP1 expression in chondrocytes. This inhibition appears to be mediated through binding competition between PPAR␥ and the AP1 complex for the composite PPRE/AP1 at position Ϫ83 to Ϫ71. As direct repeat (DR1) cis-acting elements can be recognized by a variety of transcription factors including the chicken ovalbumin upstream promoter-transcription factor or RAR-related orphan receptor (43), we cannot exclude the possibility that in our model other transcription factors may also bind this composite element and take part into the transcriptional downregulation of the MMP1 gene.
Whether PPAR-␥-dependent inhibition is restricted to MMP1 or occurs with other MMPs remains to be investigated. In the present work Rtz was shown to inhibit IL-1␤-induced MMP9 gelatinolytic activities and MMP3 and MMP13 mRNA levels. PPAR␥ agonists were also found to inhibit MMP3 expression in human rheumatoid fibroblast-like synoviocytes (44). Down-regulation of MMP9 by natural (15d-PGJ2) and synthetic (GW7845) PPAR-␥ ligands in PMA-stimulated THP-1 monocytes has been consistently observed (31). An inhibition of MMP13 expression by Rtz was also reported by Fahmi et al. (3) in human chondrocytes. Similarly to the MMP1 gene situation, AP1 has been shown to be involved in the up-regulation of MMP3 in meshwork cells (45), MMP9 in human vascular smooth muscle cells (46), and MMP13 in osteoblasts (27). We searched for the presence of a DR1-PPRE overlapping an AP1 element in the promoter of MMP3, MMP9, and MMP13. An AP-1 site at position Ϫ1929 and a PPRE at position Ϫ1950 have been recently identified in the murine MMP3 (47). Computer analysis of the human MMP3 (GenBank TM accession number AF405705) allowed us to identify a degenerated DR1-PPRE overlapping an AP1 element at position Ϫ73 to Ϫ63 (CAAGGATGAGTCA). Two PPREs were recently identified in the rat MMP9 promoter (48). Most interesting, we found a degenerated PPRE sequence overlapping an AP1 element (TGAGTCAGAAGTTCG) at position Ϫ1661 to Ϫ1646 in the human MMP9 promoter (GenBank TM accession number AF538 844). Finally, we found a DR1-PPRE/AP1 sequence (AAGT-GATGACTCA) at position Ϫ80 to Ϫ67 in the human (Gen-Bank TM accession number X81640), rat (GenBank TM accession number AY135636), and dog (GenBank TM accession number AF384859) promoters of MMP13. We thus speculated that the Rtz inhibitory effect on IL-1␤-induced MMP3, -9, and -13 may occur through a mechanism similar to the Rtz effect on MMP1.
In conclusion, we have determined that Rtz-activated PPAR-␥ represses MMP1 transcription in chondrocytes, by interfering with the binding of c-Fos/c-Jun proteins to the promoter sequences situated at Ϫ83-71. Further studies will be performed to confirm that this interaction occurs in vivo in the chromatin environment and to determine whether other co-factors participate in this effect. Nevertheless, PPAR-␥ appears able to counteract the IL-1␤-induced proteoglycan degradation and MMP1 transcription acting as a protective mechanism in cartilage. This beneficial function can be elicited by low concentrations of Rtz (and probably other thiazolidinediones), in contrast with the much less manageable amounts required with most other PPAR-␥ ligands. Thiazolidinediones thus deserve consideration for clinical testing in degenerative cartilage diseases.