ADAMTS4 (aggrecanase-1) interaction with the C-terminal domain of fibronectin inhibits proteolysis of aggrecan.

ADAMTS4 (aggrecanase-1), a secreted enzyme belonging to the ADAMTS (a disintegrin and metalloproteinase with thrombospondin motifs) gene family, is considered to play a key role in the degradation of cartilage proteoglycan (aggrecan) in osteoarthritis and rheumatoid arthritis. To clone molecules that bind to ADAMTS4, we screened a human chondrocyte cDNA library by the yeast two-hybrid system using the ADAMTS4 spacer domain as bait and obtained cDNA clones derived from fibronectin. Interaction between ADAMTS4 and fibronectin was demonstrated by chemical cross-linking. A yeast two-hybrid assay and solid-phase binding assay using wild-type fibronectin and ADAMTS4 and their mutants demonstrated that the C-terminal domain of fibronectin is capable of binding to the C-terminal spacer domain of ADAMTS4. Wild-type ADAMTS4 was co-localized with fibronectin as determined by confocal microscopy on the cell surface of stable 293T transfectants expressing ADAMTS4, although ADAMTS4 deletion mutants, including Delta Sp (Delta Arg(693)-Lys(837), lacking the spacer domain), showed negligible localization. The aggrecanase activity of wild-type ADAMTS4 was dose-dependently inhibited by fibronectin (IC(50) = 110 nm), whereas no inhibition was observed with Delta Sp. The C-terminal 40-kDa fibronectin fragment also inhibited the activity of wild-type ADAMTS4 (IC(50) = 170 nm). These data demonstrate for the first time that the aggrecanase activity of ADAMTS4 is inhibited by fibronectin through interaction with their C-terminal domains and suggest that this extracellular regulation mechanism of ADAMTS4 activity may be important for the degradation of aggrecan in arthritic cartilage.

Members of the ADAMTS (a disintegrin and metalloproteinase with thrombospondin motifs) gene family are composed of at least 19 molecules (1) and are involved in various biological and biochemical events such as fertilization, proteoglycan degradation, processing of fibrillar collagens, and intravascular coagulation (2)(3)(4)(5)(6)(7). Among them, ADAMTS1, ADAMTS4 (also referred to as aggrecanase-1), ADAMTS5 (aggrecanase-2), and ADAMTS9 cleave specific Glu-X bonds (where X is most often Ala or Gly) of the core protein of aggrecan, a major proteoglycan in articular cartilage (3,6,7). Previous studies suggested that the major aggrecan fragments found in vitro in response to cytokine-stimulated cartilage degradation and in vivo in arthritic joint fluids are generated through cleavage at Glu-X bonds by the glutamyl endopeptidase activity of these AD-AMTS members (8,9). Recent studies further demonstrated that ADAMTS4 is induced by stimulation of chondrocytes and synovial cells with interleukin-1, tumor necrosis factor-␣, or transforming growth factor-␤, although ADAMTS5 is constitutively expressed (6,10). In addition, ADAMTS4 is overexpressed by synovial cells and chondrocytes in osteoarthritis and rheumatoid arthritis (10). Thus, ADAMTS4 is considered to play an important role in the aggrecan degradation of articular cartilage in osteoarthritis and rheumatoid arthritis. The aggrecanase activity of ADAMTS4 is inhibited by TIMP-3 (tissue inhibitor of metalloproteinases-3) among the four TIMP proteins (TIMP-1-4) (11,12), all of which were originally cloned as inhibitors of matrix metalloproteinases (MMPs). 1 However, it is not known whether this is the only regulatory mechanism of ADAMTS4 activity. ADAMTS4 consists of a prodomain, a furin cleavage site, a catalytic domain, a disintegrin-like motif, a thrombospondin-1 (TSP) motif, a cysteine-rich (CR) domain, and a C-terminal spacer domain. As shown with ADAMTS1 (13), ADAMTS4 has an affinity for extracellular matrix (ECM) molecules, being deposited in the ECM after synthesis (14). In fact, the binding activity of C-terminal CR and/or spacer domains of ADAMTS4 for sulfated glycosaminoglycans of aggrecan has been reported (15). Interestingly, the aggrecanase activity of full-length active ADAMTS4 is blocked probably through interaction with ECM molecules, and activity appears after removal of the C-terminal spacer domain (14). Thus, these data suggest that ADAMTS4 may have binding molecules by which the activity is regulated. However, no information is available for proteins interacting with ADAMTS4.
In this study, by screening a human chondrocyte cDNA library, we sought binding proteins that may be involved in regulating the activity. Since the spacer domain of ADAMTS4 is not conserved among ADAMTS members, we used the domain as bait in a yeast two-hybrid system and found fibronectin to be a candidate for a regulator of ADAMTS4 activity. The data in this study demonstrate that fibronectin inhibits the aggrecanase activity of ADAMTS4 through the interaction between the C-terminal regions of each molecule.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid System-MATCHMAKER Gal4 two-hybrid system 3 and the MATCHMAKER human chondrocyte cDNA library were * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Chemical Cross-linking-To verify the binding of ADAMTS4 to fibronectin, 125 I-labeled ADAMTS4 (13 nM) was incubated with human plasma fibronectin (0, 9.1, or 91 nM; Chemicon International, Inc., Temecula, CA) or with a 120-kDa fibronectin fragment with the central cell-binding domain (Fn-f120; 100 nM; Invitrogen) in phosphate-buffered saline (PBS) for 16 h at 4°C and cross-linked by treating the mixture with 1 mM disuccinimidyl suberate (DSS) for 2 h on ice. The reaction was stopped by incubation with 100 mM Tris-HCl (pH 7.5) for 15 min at room temperature, and the sample was subjected to SDS-PAGE (12.5% total acrylamide) under reducing conditions. The gel was dried and analyzed by with a Fuji Film BAS 2000 analyzer.
Purification of Recombinant ADAMTS4 Proteins-293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum and transfected with the above-mentioned pcDNA3.1/Zeo(Ϫ) plasmid containing cDNA encoding ADAMTS4 or its mutants using DOTAP liposomal transfection reagent (Roche Diagnostics). Transfected cells were cultured for 2 days and selected with 150 g/ml Zeocin (Invitrogen) for 3 weeks. Stable transfectants expressing full-length ADAMTS4 or its truncated mutants were established by expansion from a single cell, and the conditioned culture media from these transfectants were harvested after a 48-h incubation in serumfree DMEM. For the purification of full-length ADAMTS4 and ⌬T/CR/ Sp, the conditioned media were concentrated using an Amicon Diaflo apparatus fitted with a YM-10 membrane; mixed with 4 volumes of 50 mM Tris-HCl (pH 7.5), 10 mM CaCl 2 , 0.05% Brij 35, and 0.02% NaN 3 (TCB buffer); and applied to an SP-Sepharose fast flow column (2.5 ϫ 10 cm; Amersham Biosciences) equilibrated with TCB buffer. The recombinant proteins were eluted by a linear gradient of 0 -1 M NaCl, and the combined fractions containing ADAMTS4 or ⌬T/CR/Sp were applied to a column of anti-FLAG antibody M2 affinity gel (1 ϫ 6 cm; Sigma) equilibrated with 50 mM Tris-HCl (pH 7.5), 0.15 M NaCl, 10 mM CaCl 2 , 0.05% Brij 35, and 0.02% NaN 3 (TNCB buffer). ADAMTS4 and ⌬T/ CR/Sp were eluted with 6 M urea in TNCB buffer containing 1 M NaCl after washing the column with the buffer containing 1 M NaCl and dialyzed against TNCB buffer to remove urea. For the purification of ⌬Sp and ⌬CR/Sp, which did not show a strong affinity for the anti-FLAG antibody gel, concentrated culture media were applied to a DEAE-Sepharose fast flow column (2.5 ϫ 10 cm; Amersham Biosciences) equilibrated with TCB buffer, and unbound fractions mixed with 9 volumes of TCB buffer were applied to an SP-Sepharose fast flow column. The mutants were eluted by a linear gradient of 0 -1 M NaCl as described above, dialyzed, and then applied to a chelating Sepharose fast flow column (Amersham Biosciences) charged with ZnCl 2 according to our previous method (17). ⌬Sp and ⌬CR/Sp were eluted by a linear gradient of 0 -1 M NaCl and dialyzed against TNCB buffer. The concentrations of these ADAMTS4 species were determined using a BCA protein assay kit (Pierce). The purity of recombinant proteins was evaluated by SDS-PAGE, followed by silver staining of the gels and/or autoradiography of iodinated proteins according to our previous methods (18).
Binding of ADAMTS4 to Immobilized Fibronectin and Its Fragments-Microtiter plates with 96 wells (Immunomodule, Nalge Nunc, Rochester, NY) were coated by incubation with 50 l of intact fibronectin, the C-terminal 40-kDa fibronectin fragment (Fn-f40; 25 nM; Invitrogen), or 25 nM Fn-f120 for 16 h at 4°C. The plates were washed twice with TNCB buffer and subsequently blocked with 1% bovine serum albumin (BSA) in TNCB buffer for 2 h at room temperature. They were then incubated with 125 I-labeled full-length ADAMTS4 or its deletion mutants (ϳ5 ϫ 10 5 cpm, 20 ng/well) for 24 h at 4°C. To confirm the specific binding, 125 I-labeled ADAMTS4 was mixed with a 10-or 50-fold excess amount of unlabeled ADAMTS4 or buffer and then subjected to the binding assay using microtiter plates coated with intact fibronectin. After washing twice with TNCB buffer, the bound proteins were dissociated by treatment with 1 N NaOH, and the radioactivity of the bound fractions was counted using a ␥-counter (ARC-600, Aloka, Tokyo, Japan). Similarly, the possibility of interaction between fibronectin and aggrecan was examined in a binding assay by incubating 125 I-labeled fibronectin in the aggrecan-coated wells and measuring the bound radioactivity.
Laser Scanning Confocal Microscopy of Transfected Cells-Stable transfectants expressing full-length or truncated ADAMTS4 species were established as described above and detached from the dishes by incubation with 0.25% trypsin and 0.02% EDTA for 3 min at 37°C. After blocking the activity of trypsin with 10% fetal bovine serum, the cell suspensions were washed twice with PBS and incubated with 1 g/ml fibronectin in PBS for 10 min at 37°C. The cells were suspended in serum-free DMEM containing 1% insulin/transferrin/selenium/X supplement (Invitrogen) after washing with serum-free DMEM. Fibronectin-treated cells were then cultured on Lab-Tek chamber slides (1 ϫ 10 5 cells/well; Nalge Nunc) for 1 day. The cells were incubated with 3% normal goat serum in PBS to block nonspecific binding and then reacted with anti-FLAG antibody (1:100), anti-fibronectin antibody (1: 20; Santa Cruz Biotechnology Inc., Santa Cruz, CA), or non-immune mouse IgG (Dako Corp., Glostrup, Denmark) for 1 h at room temperature. They were fixed with methanol/acetone/formaldehyde (19:19:2, v/v) and incubated with fluorescein isothiocyanate-and rhodamineconjugated secondary antibodies (1:50; Dako Corp.) after washing with PBS. Transfected cells expressing the pcDNA3.1/Zeo(Ϫ) vector (mock transfectants) and parental 293T cells were also subjected to double immunostaining as a negative control. All preparations were viewed under an Olympus laser scanning confocal microscope at a similar sensitivity (550 V for fluorescein isothiocyanate and 600 V for rhodamine), and differential interference contrast images were also viewed for comparison.
Detection of Aggrecanase Activity and Its Inhibition by Fibronectin-Aggrecan (100 g) prepared from bovine nasal cartilage (12) was incubated with purified full-length ADAMTS4 (8 nM) or truncated AD-AMTS4 species (⌬Sp, ⌬CR/Sp, or ⌬T/CR/Sp; 8 nM each) for 16 h at 37°C and deglycosylated after termination of the reaction with 20 mM EDTA as described previously (12). The digestion products were then subjected to SDS-PAGE (10% total acrylamide) under reducing conditions. Proteins in the gel were transferred onto polyvinylidene fluoride membranes, and aggrecanase activity was evaluated by detecting aggrecan fragments with the neoepitope (NITEGE 373 ) by immunoblotting using the neoepitope-specific antibody I19C (2 g/ml), which was kindly provided by Drs. Kotaro Sugimoto and Kazuhiko Tanzawa (Sankyo Co., Ltd., Tokyo) (19). After reaction with horseradish peroxidase-linked anti-IgG antibody (1:5000; Amersham Biosciences), immunoreactive proteins on the membranes were detected using the ECL TM Western blot detection system (Amersham Biosciences) according to the manufacturer's instructions.
For the study of fibronectin inhibition, wild-type ADAMTS4 and ⌬Sp (8 nM each), which showed sufficient aggrecanase activity, were incubated with increasing concentrations of fibronectin (0, 10, 20, 40, 100, 200, 500, and 1000 nM final concentrations) for 2 h at 37°C prior to the reaction. Aggrecan was then digested by incubation with the mixtures for 16 h at 37°C, and aggrecanase activity was monitored as described above. Since the aggrecanase activity of wild-type ADAMTS4 was inhibited by intact fibronectin, the inhibitory effects of the fibronectin fragments (Fn-f40 and Fn-f120) on the activity were also examined using ADAMTS4 preincubated with the fragments at final concentrations of 0, 10, 20, 40, 100, 200, 500, 750, and 1000 nM in the presence of 1 mM phenylmethanesulfonyl fluoride, which was added to completely block a trace amount of serine proteinase(s) contaminating the preparations of fibronectin fragments. The densities of immunoreactive bands were measured by scanning densitometry using NIH Image Version 1.62 according to our previous method (12).
Statistical Analysis-Measured values are expressed as the mean Ϯ S.D. In the solid-phase binding assay, the difference in radioactivity was analyzed by the Bonferroni/Dunn test. Tests were performed using StatView Version 5.0. p Ͻ 0.05 was considered significant.

Screening of Proteins That Interact with ADAMTS4 -To
identify the proteins that interact with ADAMTS4, 3 ϫ 10 6 clones from the human chondrocyte cDNA library were screened by the yeast two-hybrid system using a cDNA fragment encoding the C-terminal spacer domain (Lys 687 -Lys 837 , 151 amino acids) of ADAMTS4 as bait. Among the 156 clones grown on medium lacking tryptophan, leucine, histidine, and adenine, 31 clones (20%) were identified as human fibronectin. These 31 clones had different 5Ј-ends, but all of them except for the shortest clone covered the whole region of the C-terminal heparin-binding domain/fibrin-binding domain of fibronectin (Fig. 1A).
Interaction of ADAMTS4 with Fibronectin Fragments in Yeast-The region of binding of ADAMTS4 to fibronectin was examined by yeast two-hybrid assays. Yeast strain AH109 was cotransformed with the pGBKT7/TS4sp and pACT2/FnI-VI plasmids, containing cDNA encoding fragment I, II, III, IV, V, or VI of fibronectin (Fig. 1A). As shown in Fig. 1B, clones cotransformed with pGBKT7/TS4sp and pACT2/FnVI (referred to as Sp/VI) as well as positive clones expressing p53 and SV40 (referred to as p53/SV40) could grow to form blue-stained colonies on medium lacking the three amino acids and adenine. In contrast to these transformants, other clones expressing ADAMTS4 and fibronectin fragments I-V showed only negligible background growth (Fig. 1B). Although positive control p53/SV40 transformants grew faster than Sp/VI transformants, the intensity of blue color for X-␣-gal staining was much higher in the Sp/VI transformants than in the control (Fig. 1B). When the ␣-galactosidase activity of each transformant was evaluated by measuring the absorbance at 410 nm, Sp/VI transformants showed remarkably higher activity compared with other transformants and control p53/SV40 transformants (Fig. 1C).
Cross-linking Study of ADAMTS4 and Fibronectin-To further study the interaction of ADAMTS4 with fibronectin, a cross-linking experiment was carried out by incubating 125 Ilabeled ADAMTS4 with increased amounts of intact fibronectin. As shown in Fig. 1D, 125 I-labeled ADAMTS4, which migrated at 73 kDa, shifted to the site of ϳ400 kDa when reacted with intact fibronectin and cross-linked with DSS. On the other hand, reaction of an ADAMTS4 and Fn-f120 mixture with DSS showed negligible cross-linked products. Although ADAMTS4 became a broad band ranging from 70 to 80 kDa in the presence of DSS, dimerization of the proteinase was not detected. Thus, the molecular shift was considered to be caused by a crosslinked complex of ADAMTS4 and intact fibronectin.
Binding of ADAMTS4 to the C-terminal Region of Fibronectin-To identify the region of fibronectin that interacts with ADAMTS4, a solid-phase binding assay was performed using immobilized intact fibronectin, Fn-f120, and Fn-f40. A large amount of 125 I-labeled ADAMTS4 could bind to intact fibronectin-coated wells, whereas BSA-coated wells showed only background binding. The radioactivity bound to Fn-f40 (but not Fn-f120) was significantly higher than that to BSA (p Ͻ 0.01 versus BSA) (Fig. 1E).
Purification and Aggrecanase Activity of ADAMTS4 and Its Mutants-Full-length ADAMTS4 and its C-terminally truncated mutants (⌬Sp, ⌬CR/Sp, and ⌬T/CR/Sp, which lack the C-terminal spacer domain; most of the CR and whole spacer domain; and the TSP, CR, and spacer domains, respectively) ( Fig. 2A) were expressed in stably transfected 293T cells. These ADAMTS4 species were purified from the conditioned media. As shown in Fig. 2B, purified recombinant ADAMTS4, ⌬Sp, ⌬CR/Sp, and ⌬T/CT/Sp showed major protein bands of 73, 58, 48, and 39 kDa, respectively, on silver-stained gels after SDS-PAGE. All bands were recognized by immunoblotting with anti-FLAG antibody (Fig. 2C). When the aggrecanase activity of each recombinant ADAMTS4 species was examined by immunoblotting of aggrecan digestion products using the neoepitope-specific antibody, an immunoreactive aggrecan fragment of ϳ80 kDa appeared after digestion with wild-type ADAMTS4 and ⌬Sp (Fig. 2D), indicating that these two recombinant ADAMTS4 species have potent aggrecanase activity. On the other hand, ⌬CR/Sp showed only weak activity, and ⌬T/ CR/Sp had no activity (Fig. 2D).
Binding of the C-terminal Spacer Domain of ADAMTS4 to Fibronectin-To determine which domain of ADAMTS4 is involved in binding to fibronectin, a solid-phase binding assay was performed by incubating 125 I-labeled ADAMTS4, ⌬Sp, ⌬CR/Sp, or ⌬T/CR/Sp in fibronectin-coated wells. As shown in Fig. 3A, the binding activity of these ADAMTS4 species was higher than that of BSA, which had only background signals. However, the binding activity of wild-type ADAMTS4 was remarkably ϳ3-fold higher than that of C-terminally truncated ADAMTS4 mutants (Fig. 3A). The specific binding between 125 I-labeled ADAMTS4 and fibronectin was confirmed since the binding was competitively inhibited by unlabeled ADAMTS4 in a dose-dependent manner (Fig. 3B). Thus, these results suggest that the C-terminal spacer domain (Arg 693 -Lys 837 ) of ADAMTS4 is involved in binding to intact fibronectin.
Docking of ADAMTS4 on the Cell Membrane of Fibronectincoated Cells-The pericellular interaction of ADAMTS4 species with fibronectin was examined by double immunostaining of stable transfectants expressing wild-type ADAMTS4, ⌬Sp, ⌬CR/Sp, ⌬T/CR/Sp, or vector alone. As shown in Fig. 4, fibronectin was immunolocalized on the cell surface of all the FIG. 1. Binding of fibronectin and its fragments to ADAMTS4. A, schematic diagram of human fibronectin and its fragments. cDNA fragments I-VI encode the six regions of fibronectin used in yeast two-hybrid assays. Fn-f120 and Fn-f40 denote fibronectin fragments of 120 and 40 kDa, respectively, which were utilized for binding and inhibition experiments. The range of the regions covered by the longest and shortest transfectants. Although wild-type ADAMTS4 was strongly colocalized with fibronectin on the cell surface, negligible or no immunoreaction was observed with ⌬Sp, ⌬CR/Sp, ⌬T/CR/Sp, or mock transfectants (Fig. 4).
Inhibition of the Aggrecanase Activity of ADAMTS4 by Fibronectin and Its Fragments-To study the effect of fibronectin on wild-type ADAMTS4 and ⌬Sp, both of which have definite aggrecanase activity, these recombinant proteinases were incubated with intact fibronectin (0, 10, 20, 40, 100, 200, 500, and 1000 nM final concentrations), and aggrecanase activity was determined by immunoblotting using the neoepitope-specific antibody. As shown in Fig. 5A, fibronectin completely inhibited fibronectin clones is indicated by lines above the fibronectin domains. FHBD, fibrin-and heparin-binding domain; GCBD, gelatin-and collagenbinding domain; CBD, cell-binding domain; HBD, heparin-binding domain; FBD, fibrin-binding domain; RGD, portion containing the RGD sequence. B, yeast strains cotransformed with the pGBKT7/TS4sp plasmid with the ADAMTS4 spacer domain (Sp) and the pACT2 vector containing each fibronectin cDNA fragment (I, II, III, IV, V, or VI). The yeast strains were transformed with the pGBKT7/TS4sp plasmid and the pACT2 vector alone (negative control) and with plasmid p53 and SV40 (positive control). Yeast transformants were cultured for 2 days on X-␣-gal-containing medium lacking leucine, tryptophan, histidine, and adenine. C, ␣-galactosidase activity of each transformant. The activity was measured using p-nitrophenyl-␣-D-galactoside as described under "Experimental Procedures." Errors bars indicate S.D. (n ϭ 5). D, chemical cross-linking of ADAMTS4 and fibronectin. Cross-linking was performed with 125 I-labeled ADAMTS4 and intact fibronectin (Fn) or Fn-f120 in the presence of DSS, and the products were subjected to SDS-PAGE (12% total acrylamide) under reducing conditions as described under "Experimental Procedures." E, binding of 125 I-labeled ADAMTS4 to fibronectin and its fragments immobilized on microtiter plates. The binding activity of 125 I-labeled ADAMTS4 was assayed in triplicate using intact fibronectin, Fn-f120, Fn-f40, or BSA as described under "Experimental Procedures." Errors bars indicate S.D. **, p Ͻ 0.01.  4). Aliquots of aggrecan (100 g) were incubated with each ADAMTS4 species, and the digestion products were analyzed by immunoblotting using the neoepitope-specific antibody as described under "Experimental Procedures." the aggrecanase activity of wild-type ADAMTS4 at 500 nM, although little or no inhibition of ⌬Sp was observed. When data measured by the densitometric analysis were plotted, an Sshaped inhibition curve was obtained for wild-type ADAMTS4 (Fig. 5B), and the IC 50 value (concentration at 50% inhibition) of fibronectin was determined to be 110 nM. We also examined the inhibition of wild-type ADAMTS4 and ⌬Sp with fibronectin fragments. As shown in Fig. 6 (A and B), Fn-f40 completely blocked the activity of wild-type ADAMTS4 at 750 nM, whereas Fn-f120 inhibited only 30% of the activity. Based on the inhibition curve, the IC 50 value of Fn-f40 was 170 nM (Fig. 6B). In contrast, these fibronectin fragments did not inhibit the aggrecanase activity of ⌬Sp (data not shown). In addition, when 125 I-labeled fibronectin was incubated in the aggrecan-coated wells, no binding was observed (data not shown), indicating that inhibition of ADAMTS4 aggrecanase activity by fibronectin is not due to interaction between fibronectin and aggrecan.

DISCUSSION
In this study, we have demonstrated that ADAMTS4 binds to fibronectin, which is an adhesive glycoprotein present in many tissues, including articular cartilage (20). This interaction was first discovered in the yeast two-hybrid system by screening a human chondrocyte cDNA library and further demonstrated by the findings of a cross-linking experiment, solid-phase binding assay, and co-immunolocalization. The TSP motifs and spacer domain of ADAMTS1 have an affinity for ECM molecules, although the interacting molecules and mechanisms are unknown (13). Since ADAMTS4 also possesses a similar TSP motif and spacer domain, this metalloproteinase was expected to interact with the ECM. In fact, ADAMTS4 requires the interaction with sulfated glycosaminoglycans attached to aggrecan core protein for efficient aggrecan degradation (21), and binding sites to aggrecan glycosaminoglycans have been identified in C-terminal CR and/or spacer domains (15). However, this study is the first to demonstrate the protein-protein interaction of ADAMTS4 with fibronectin.
Fibronectin is composed of several structural domains with different binding abilities. Thus, we carried out yeast twohybrid assays by coexpressing cDNA fragments encoding the ADAMTS4 spacer domain or six different parts of fibronectin to examine the binding region of fibronectin. Data showing that only the clone cotransformed with pGBKT7/TS4sp and pACT2/ FnVI could grow on the media suggest the possible interaction of the ADAMTS4 spacer domain with the fibronectin C-terminal domain. This was confirmed by the solid-phase binding assay using the C-terminal fibronectin fragment, Fn-f40. On the other hand, since our yeast two-hybrid screening and assays utilized the ADAMTS4 spacer domain as bait, the spacer domain was assumed to be responsible for the interaction with fibronectin. Actually, our binding assay data using recombinant proteins of ADAMTS4 and its deletion mutants indicate that the C-terminal spacer domain (Arg 693 -Lys 837 ) is essential for the binding. Thus, these data demonstrate that the Cterminal domains of ADAMTS4 and fibronectin are involved in their binding. Besides ADAMTS4, fibrin and insulin-like growth factor-binding protein-3 (22) and CD44 (23) also bind to the C-terminal region of fibronectin. However, since a homology search for amino acid sequences showed no consensus sequences among these fibronectin-binding proteins, molecular mechanisms and elements responsible for the interaction between the C-terminal regions of ADAMTS4 and fibronectin remain to be clarified by further study.
The double immunostaining performed in the present study demonstrated that ADAMTS4 and fibronectin are colocalized on the cell surface. Importantly, the co-localization was positive only with fibronectin and wild-type ADAMTS4, but not with C-terminally truncated mutants of ADAMTS4. These observations confirm the complex formation of ADAMTS4 and fibronectin through interaction between their C-terminal domains at a cellular level. It has been shown that full-length active ADAMTS4 is trapped within the ECM after synthesis from cells and that C-terminally truncated species of active ADAMTS4 are released from the ECM by metalloproteinase-mediated or autocatalytic truncation (14). Recent studies also demonstrated the pericellular immunolocalization of ADAMTS4 in cultured chondrocytes (24) and ADAMTS4 transfectants (7,25). Although sulfated glycosaminoglycans of aggrecan were assumed to be binding partners for ADAMTS4 (24), the data in the present study suggest that fibronectin is a preferable candidate molecule for the pericellular docking of ADAMTS4. Fibronectin can directly localize on the cell membrane by binding to integrins such as ␣ 5 ␤ 1 (20), whereas aggrecan may be deposited around chondrocytes with some distance from the cells. In fact, fibronectin is immunolocalized in the pericellular region of chondrocytes in osteoarthritis cartilage (26) and synovial fibroblasts in rheumatoid arthritis synovium (27). Since fibronectin utilizes different sites for binding to cells (central cell-binding domain) and ADAMTS4 (C-terminal heparinand fibrin-binding domain), interaction between ADAMTS4 and fibronectin cannot be disturbed on the cell membrane. Thus, these observations suggest the possibility that fibronectin plays a role in anchoring ADAMTS4 onto the cell membrane of chondrocytes and/or synovial fibroblasts in pathological conditions such as osteoarthritis and rheumatoid arthritis.
One of the most interesting findings in this study is that the aggrecanase activity of ADAMTS4 was inhibited by interaction with fibronectin. Concerning regulators of the aggrecanase activity of ADAMTS4, TIMP-3 was reported as an efficient inhibitor (11,12). Since ADAMTS4 shares a common catalytic site with MMPs, it is predictable that TIMP-3, an original inhibitor of MMPs, inhibits the activity of ADAMTS4 probably by binding to the catalytic site of ADAMTS4 (11,12). However, the inhibition of ADAMTS4 by fibronectin is novel in that an ECM molecule acts as an inhibitor of a member of the ADAM family. A recent study showed that proteoglycans in the brain, i.e. testican-3 and its spliced variant (N-Tes), can inhibit the pro-MMP-2 activation activity of membrane-type MMPs (MT1-MMP and MT3-MMP) (28). This suggests that ECM molecules are involved in regulating the activities of MMP and ADAMTS members through molecular interactions. Since we could not carry out kinetic studies because of the long incubation time (at least 12 h) required for the aggrecanase assay, the inhibition mechanism of ADAMTS4 activity by fibronectin is not clear. However, the data showing that aggrecanase activity is inhibited only by complex formation between the non-catalytic domain of ADAMTS4 and fibronectin and that fibronectin itself is not a substrate of wild-type ADAMTS4 (3) suggest that fibronectin may be a noncompetitive inhibitor and not a competitive inhibitor. On the other hand, since binding of the C-terminal CR and/or spacer domain of ADAMTS4 to aggrecan glycosaminoglycans is required for aggrecan degradation by ADAMTS4 (15), the inhibition of ADAMTS4 by fibronectin may be ascribed to the hindrance of access of the proteinase to aggrecan. The IC 50 values of intact fibronectin and Fn-f40 against ADAMTS4 were Ͻ200 nM, i.e. 92 g/ml. Since concentrations of fibronectin in normal serum (280 -375 g/ml) and synovial fluids from healthy subjects (172 g/ml) and patients with rheumatoid arthritis (721 g/ml) or osteoarthritis (556 g/ml) are much higher than the inhibition levels against ADAMTS4 activity (29,30), aggrecanase activity in serum and synovial fluid may be inhibited.
ADAMTS4 is synthesized in a zymogen form of ϳ100 kDa and converted to an active form of ϳ75 kDa after intracellular processing of the prodomain by furin-like proteinases (14). However, Gao et al. (14) demonstrated that the 75-kDa ADAMTS4 species is associated with the ECM in an inactive form after synthesis from cells and processed in the ECM to C-terminally truncated forms of ϳ60 and ϳ50 kDa, which are then released from the ECM into the culture medium. In addition, truncation of the spacer domain is ascribed to the action of MMPs or autocatalysis (14,15). Although these studies gave no information about ECM molecule(s) associated with the 75-kDa full-length active species, the data in our study suggest that inactivation of the ADAMTS4 species observed in such studies may be due to the interaction of fibronectin with the C-terminal spacer domain of ADAMTS4. MMPs such as MMP-2 and MT2-MMP cleave ADAMTS1 at the middle portion of the spacer domain (31). In addition, a recent study has shown that ADAMTS4 is processed by MT4-MMP to the Cterminally truncated active form, which can interact with glycosaminoglycan chains of syndecan-1 (32). Many cells of mesenchymal origin, including chondrocytes and synovial fibroblasts, produce various active MMPs such as MMP-2 and MT1-MMP in the pericellular region within tissues (33,34), and C-terminally truncated ADAMTS4 species have been detected in human cartilage and synovial tissue (14). Thus, the activities of MMPs such as MMP-2 might be involved in the cleavage of ADAMTS4 at sites between the CR and spacer domains to activate ADAMTS4, which is inhibited and anchored to the cell membrane through interaction with fibronectin. Unlike MMPs, the activity of ADAMTS4 is weakly or negligibly inhibited by TIMP-1, -2, and -4 (11,12). Although TIMP-3 efficiently inhibits the activity (11,12), the inhibitor may not always be expressed in ADAMTS4expressing tissues such as the brain, heart, and lung (35). Thus, the tissues need another defense mechanism against the activity to avoid rapid and massive degradation of proteoglycans such as aggrecan. Inhibition of ADAMTS4 activity through interaction with fibronectin, a ubiquitous ECM component, may be an attractive and novel supplementary protection mechanism against aggrecan degradation. Further study is needed to demonstrate this hypothesis at the cellular and tissue levels in pathological tissues such as osteoarthritis cartilage.