ADAMTS4 (Aggrecanase-1) Activation on the Cell Surface Involves C-terminal Cleavage by Glycosylphosphatidyl Inositol-anchored Membrane Type 4-Matrix Metalloproteinase and Binding of the Activated Proteinase to Chondroitin Sulfate and Heparan Sulfate on Syndecan-1*

C-terminal truncation of ADAMTS-4 from the p68 form to the p53 form is required for activation of its capacity to cleave the Glu 373 -Ala 374 interglobular domain bond of aggrecan. In transfected human chondrosarcoma cells, this process is not autoproteolytic because the same products form with an inactive mutant of ADAMTS4 (a disintegrin and metalloproteinase with thrombospondin-like motif 4) and truncation is completely blocked by tissue inhibitor of metalloproteinase-1. Instead, activation can be mediated by glycosylphosphatidyl inositol- anchored membrane type 4-matrix metalloproteinase (MT4-MMP, MMP-17) because co-transfection with the active form of MT4-MMP markedly enhanced activation, whereas an inactive mutant of MT4-MMP was ineffec- tive. Treatment of co-transfected cells with phosphati-dylinositol-specific phospholipase C liberated the com- plex of MT4-MMP


C-terminal truncation of ADAMTS-4 from the p68 form to the p53 form is required for activation of its capacity to cleave the Glu 373 -Ala 374 interglobular domain bond of aggrecan. In transfected human chondrosarcoma cells, this process is not autoproteolytic because the same products form with an inactive mutant of ADAMTS4 (a disintegrin and metalloproteinase with thrombospondinlike motif 4) and truncation is completely blocked by tissue inhibitor of metalloproteinase-1. Instead, activation can be mediated by glycosylphosphatidyl inositolanchored membrane type 4-matrix metalloproteinase (MT4-MMP, MMP-17) because co-transfection with the active form of MT4-MMP markedly enhanced activation, whereas an inactive mutant of MT4-MMP was ineffective. Treatment of co-transfected cells with phosphati-
dylinositol-specific phospholipase C liberated the complex of MT4-MMP and p68 ADAMTS4 from the cell membrane, but the p53 ADAMTS4 remained associated. Specific glycosaminoglycan lyase digestions, followed by product analyses using fluorescence-assisted carbohydrate electrophoresis and immunoprecipitation experiments, showed that the p53 form is associated with syndecan-1 through both chondroitin sulfate and heparan sulfate. We conclude that ADAMTS-4 activation in this cell system involves the coordinated activity of both glycosylphosphatidyl inositol-anchored MT4-MMP and the proteoglycan form of syndecan-1 on the cell surface.
The proteolytic processing of extracellular matrix (ECM) 1 is widely studied in many tissues and cell types in relation to normal progression of fertilization, embryogenesis, develop-ment, growth, and aging (1). Uncontrolled proteolysis of ECM has also been implicated in many disease states, such as those characterized by tumor invasion (2), chronic inflammation (3), non-healing wounds (4), or excessive tissue destruction (5).
Control of both MMP and ADAMTS activity is exerted at multiple points including transcription, translation, posttranslational processing (including zymogen activation), substrate accessibility, and inhibition by naturally occurring inhibitors such as tissue inhibitors of metalloproteinase (TIMPs). Zymogen activation of both MMPs and ADAMTSs requires, minimally, the removal of the N-terminal prodomain; this can be achieved by intracellular furin-mediated cleavage, such as occurs with the MT-MMPs (12) and the ADAMTSs (7). Prodomain removal can also be achieved by the action of other MMPs, such as for the MT1-MMP-mediated activation of pro-MMP-2 (13), or through activation cascades involving co-activators such as plasmin (14). In this regard, we have been studying the activation of ADAMTS4 in a human chondrosarcoma cell line stably transfected to express the full-length protein and have found (15) that proteolytic activity (evaluated by cleavage at Glu 373 -Ala 374 in aggrecan or cleavage at Glu 441 -Ala 442 in versican V1) requires not only furin-mediated removal of the prodomain but also truncation of the C-terminal region of the proteinase. Moreover, in this cell system, the C-terminal truncation is inhibited by TIMP-1 and a hydroxamate-based MMP inhibitor, suggesting a requirement for a TIMP-1-sensitive MMP in the activation process, an idea first suggested by the finding of TIMP-1 and n-carboxyalkyl peptide inhibition of IL-1-mediated aggrecanolysis in bovine cartilage explants (16).
Moreover, in previous work on the induction of aggrecanasemediated aggrecan degradation in rat chondrosarcoma cells and bovine cartilage explants (17,18), we found that both IL-1 and retinoic acid-induced aggrecanolysis was markedly inhibited by agents (such as mannosamine, 2-deoxyfluoroglucose, and phosphatidylinositol-specific phospholipase C (PIPLC)), which are known to interfere with the synthesis or function of glycosylphosphatidyl inositol (GPI)-anchored proteins. Furthermore, these agents were effective inhibitors of ADAMTS4 truncation and activation in the human chondrosarcoma cell system (15).
When taken together, these results suggest that the control of ADAMTS4 (aggrecanase-1) activity in many cell-dependent systems is controlled by activation through a TIMP-1-inhibitable, GPI-anchored MMP. In the present report, we provide data that indicate that MT4-MMP (MMP-17) (12) is the proteinase responsible for ADAMTS4 activation and that moreover, the activated enzyme form can be bound to the cell surface through the glycosaminoglycan chains of membraneassociated syndecan-1.

EXPERIMENTAL PROCEDURES
Materials-D-Mannosamine, 2-deoxyfluoroglucose, gelatin (porcine skin), anti-MT4-MMP (M3684), heparitinase II (H 6512), and heparan sulfate (HS) disaccharide standards were from Sigma. PIPLC (5 units/ 100 l) was from Boehringer. Goat anti-rabbit IgG and protein A/Gagarose were from Calbiochem. Dulbecco's modified Eagle's medium (product number 12800 for cell growth and 23800 for serum-free culture) powder and culture-tested distilled water were from Invitrogen. Fetal bovine serum was from HyClone (Logan, UT). The ECL detection kit was from Amersham Life Science, and TIMP1 was supplied from Merck. JSCYNH was prepared and described previously as anti-YNHR (15). JSCNIT, JSCVMA, JSCFRK, and JSCALT were raised in rabbits against the ovalbumin-conjugated peptides CGGNITEGE, CVMAHVD-PEEP, CGGSGSFRK, and CGGTGSALT, respectively (by Invitrogen). The antibodies were affinity purified against the relevant immunizing peptide on Sulfolink (Pierce) and used at 0.2-2.0 g of IgG per ml. Chondroitinase ABC (protease-free), keratanase II, and endo-␤-galactosidase were from Seikagaku Corporation (Cape Cod Associates). Purified recombinant ADAMTS4 was a kind gift from Wyeth Inc. (Boston, MA). Anti-MMP2 (rabbit Ab 45) was kindly provided by Dr. William Stetler-Stevenson (National Cancer Institute, Bethesda, MD). Antibodies mS1ED and BB-4 to syndecan-1 were kindly provided by Dr. Alan Rapraeger (University of Wisconsin), and antibodies to biglycan were from Dr. Peter Roughley (19). The expression vectors to murine MT4-MMP (HA) in pcDNA3.1 Zeoϩ and the inactive mutant (human MT4-MMP (E/A) in PSG6) were kindly provided by Dr. Motoharu Seiki (University of Tokyo). The expression vector for mouse furin (pCMVm-Fur) was kindly provided by Dr. Kazuhisa Nakayama of the University of Tsukuba.
Cell Culture, Co-transfection, and Inhibitor Studies-Preparation and culture of the JJ012-TS4 (human chondrosarcoma cell line stably transfected with ADAMTS4) and VA13-TS4 (human fibroblast cell line stably transfected with ADAMTS4) cells was as described previously (15). Briefly, stably transfected cells (about 1 ϫ 10 6 cells) were cultured in 60-mm dishes in 3 ml of standard growth medium containing 10% fetal calf serum to ϳ80% confluence. The day before co-transfection, the cells were replated into fresh growth medium, and the following day, the cells were transferred to serum-free medium and exposed for 3 h to expression plasmids in GenePORTER transfection reagent (Gene Therapy Systems, San Diego, CA) according to the manufacturer's instructions. After removal of the reagent, the cells were maintained for an additional 48 h in growth medium containing 10% FBS. Finally, the cell layer was washed in serum-free medium and maintained for an additional 3 days under serum-free conditions. ECM and cell pellets were prepared as described previously in detail (15). For experiments with 2-DFG, mannosamine, and TIMP-1, the inhibitors were added during the last 3 days of maintenance in serum-free medium. Aggrecanase activity (Glu 373 -Ala 374 cleavage) in culture media was assayed with human aggrecan and antibody JSCNIT as described (15).
Treatment with PIPLC, Autoproteolysis, Immunodepletion Experiments, and Gelatin Zymography-For treatment with PIPLC, after the 3 days of serum-free culture described above, the cells were removed from the dish by gentle suspension in Ca 2ϩ /Mg 2ϩ -free PBS and pelleted by centrifugation. These pellets were gently resuspended in 120 l of serum-free medium, and two 60-l portions were incubated at 37°C for 2 h in the absence or presence of 1.5 international units of PIPLC. After incubation, the cells were re-pelleted at 350 g for 10 min; the supernatant was removed, and cells were lysed in 40 l of 50 mM Tris-HCl (pH 8.0), 5 mM EDTA, 150 mM NaCl, and 0.5% Nonidet P-40. Both supernatant and solubilized cells were taken for Western analysis (15). For autoproteolysis studies, 500-ng portions of purified recombinant AD-AMTS4 (20) were incubated in 50 l of 20 mM Tris, 100 mM NaCl, 10 mM CaCl 2 (pH 7.5) for 2 h at 37°C, and the reactions were terminated by adjustment to 5 mM EDTA. For MMP immunodepletion experiments, 500 l of ECM sample (15) was concentrated to 50 l by lyophilization and added to 8 l of anti-MMP2 or 8 l of anti-MT4-MMP serum. Controls were run in parallel with non-immune rabbit serum. After incubation for 16 h at 4°C, 15 l of protein A/G agarose suspension was added, and incubation was continued for 1 h on a rocker. Protein A/G was pelleted by brief centrifugation, the supernatants were removed, and 50% was used for gelatin zymography under non-reducing conditions on 8% SDS-PAGE gels prepared to include 0.1% (w/v) gelatin. After electrophoresis, gels were washed twice for 15 min with 2.5% Triton X-100 in water, rinsed with water, and added to activating buffer (1% Triton X-100, 10 mM CaCl 2 , 50 mM Tris-HCl, and 100 mM NaCl, pH 7.5) at 37°C for 16 h with rocking. After removal of buffer, the gels were stained for 30 min with Brilliant Blue R-250 (0.1% (w/v) in 50% (v/v) methanol, 1% (v/v) acetic acid) and destained in 40% (v/v) methanol, 10% (v/v) acetic acid.
Co-immunoprecipitation Protocols-For immunoprecipitation experiments, JJ012-TS4 cells were cotransfected with MT4-MMP and after 3 days of serum-free culture were removed from the dish in cold PBS by gentle pipetting and pelleted by centrifugation. Cells (ϳ10 6 ) were treated with PIPLC and/or Chase/heparitinase II or directly lysed, and 200 l of supernatant from enzyme-treated cells or 200 l of cell lysate was pre-cleared with 20 l of a 50% slurry of protein A/G-agarose. Then the supernatant was mixed with 15 l of primary antibody (0.25-1 g) overnight at 4°C. The antigen/antibody complexes were precipitated by incubating with 20 l of protein A/G-agarose for 1 h at 4°C, and after brief centrifugation, the beads were washed four times with lysis buffer, resuspended in gel loading buffer, and heated at 100°C for 5 min. Finally, the beads were pelleted, and the supernatants were taken for Western analysis (15).
Treatment of Cells and Immunoprecipitates with Glycosaminoglycan Lyases and Analysis of HS and CS-Replicate cultures of JJ012-TS4 cells were co-transfected with MT4-MMP and switched to serum-free medium for 3 days as above. Cells (10 6 per treatment) were recovered from dishes by gentle agitation and pelleted for digestion with PIPLC (see above). Cells were pelleted, washed, and then digested by Chase ABC (proteinase-free) (0.16 unit/ml in DMEM medium, 50 mM sodium acetate, and 10 mM EDTA, pH 7.5) alone, heparitinase II (2.5 units/ml in DMEM medium and 5 mM CaCl 2 , pH 7.5) alone, or the enzymes sequentially, in which case the cells after Chase ABC digestion were pelleted, washed, and then digested by heparitinase II as above. After each treatment (30 min at 37°C), cells and supernatants were separated by brief centrifugation. In experiments where syndecan-1 was immunoprecipitated in the proteoglycan form, the protein A/G agarose beads were treated with Chase ABC and heparitinase II sequentially as described above, except that the enzyme digestions were for 2 h each, and after the Chase digestion, the CaCl 2 concentration was adjusted to 20 mM before addition of the heparitinase II. The total sample was then added to an equal volume of 2ϫ gel loading buffer and heated at 100°C for 5 min. Finally, the beads were pelleted, and the supernatants were taken for Western analysis.
FACE Analysis-Cell layers were digested with proteinase K, and the solubilized glycosaminoglycan (GAG) was desalted by G50 chromatography and concentrated by lyophilization as described (21). These were digested with either chondroitinase ABC or heparitinase II (in 100 l of 0.1 M ammonium acetate, pH 7.3). Buffer salts were removed by Speedvac lyophilization. Chondroitinase products were fluorotagged at 37°C for 18h with 5 l of 0.1 M 2-aminoacridone (dissolved in glacial acetic acid:Me 2 SO, 3:17, v/v) plus 5 l of 1 M sodium cyanoborohydride. Heparitinase products were reacted with 0.1 M 2-aminoacridone (dissolved in glacial acetic acid:Me 2 SO, 3:197, v/v) plus 5 l of 1 M sodium cyanoborohydride (21). The reaction was terminated by the addition of 20 l of 25% (v/v) glycerol; samples were mixed and stored at Ϫ20°C until electrophoretic separation on 20% polyacrylamide gels, which were prepared as follows. Cassettes were assembled in a plastic pouch using two 10 ϫ 10-cm glass plates (Owl Separation Systems) separated by 0.8-mm spacers and then placed into a vertical protein gel caster (Fisher Scientific). Each cassette was filled to a height of ϳ7 cm with degassed separating gel solution (20% acrylamide, N-methyl-bisacrylamide (38.5:1.5), 5% (v/v) glycerol, 0.1 M Tris, 0.09 M boric acid, 5 mM EDTA (pH 8.3), 0.25% (w/v) ammonium persulfate, and 0.5% TEMED) and then overlaid with water. After polymerization (5-10 min at room temperature), the water was discarded, and the cassette was filled with 2 ml of the degassed stacking gel solution (8% acrylamide, N-methylbisacrylamide (38.5:1.5), 5% (v/v) glycerol, 0.1 M Tris borate (pH 8.3), 0.25% (w/v) ammonium persulfate, and 0.5% (v/v) TEMED), and an 8-well comb was inserted during polymerization. Before electrophoresis, gels were removed from the storage pouch, plates were washed with water, and combs were released from the stacking gel. The resulting sample wells were rinsed and then filled with electrophoresis buffer (0.1 M Tris, 0.09 M boric acid, and 5 mM EDTA, pH 8.3), and cassettes were placed in the Glyco electrophoresis gel tank (Prozyme, San Leandro, CA) filled with pre-cooled electrophoresis buffer. Aliquots (5 l) of the fluorotagged samples were loaded, and products were separated by electrophoresis at 500 V (ϳ40 mA/gel) for 80 min at 4°C. For image analyses, glass plates were removed, and the gels were placed directly onto a UV light box. Images were captured and recorded, and fluorescent bands were quantitated using the Kodak EDAS Imaging System and corresponding software.
Generation of the Inactive Mutant (E362Q) of ADAMTS-4 -Fulllength ADAMTS-4 cDNA was obtained by reverse transcription-PCR strategy based upon the published sequence (22). Briefly, total RNA (1 g) was isolated from human primary synoviocytes and used for cDNA synthesis with SMART kit (Clontech Laboratories, Inc., Palo Alto, CA) according to the manufacturer's protocol. The full-length ADAMTS4 cDNA was amplified from the synoviocyte cDNA by PCR using primers designed to engineer a NotI site (5Ј-GCGGCCGC-3Ј) at the 5Ј-end and a XbaI site (5Ј-AGATCT-3Ј) at the 3Ј-end of the gene for the convenience of subcloning it into the pcDNA 3.1(ϩ) mammalian expression vector (Invitrogen). The primers were designed to engineer a NotI site (5Ј-GCGGCCGC-3Ј) at the 5Ј-end and an EcoRV site (5Ј-GATATC-3Ј) at the 3Ј-end of the gene for the convenience of subcloning it into the mammalian expression vector pcDNA 3.1(-) (Invitrogen, San Diego). The pairs of primers for amplifying ADAMTS-4 were 5Ј-ATGCGGCCGCCT-CAATCCTGCAAGCAAGTG-3Ј and 5Ј-CGGATATCGCAAGGTCAC-CACTGTCAC-3Ј). The cycling conditions for PCR reactions were 30 cycles of 94°C for 1 min, 55°C for 1 min, and 72°C for 3 min. The PCR products were digested and ligated into NotI/EcoRV-digested pcDNA3.1(-) vector. The sequence of the construct was verified by DNA sequence analysis and restriction enzyme digestion. The primer (C CAG TCA GCC TTC ACT GCT GCT CAT CAA(Q362) CTG GGT CAT GTC TTC A) used for site-directed mutagenesis was synthesized and purified by Integrated DNA Technologies, Inc. (Coralville, IA) and contained a single point mutation (G3 C), which led to the amino acid change at residue 362 (E3 Q) located within the catalytic domain of ADAMTS4. In addition, the pcDNA3.1(ϩ) expression plasmid containing the wildtype ADAMTS4 was used as a template for site-directed mutagenesis. The mutant construct containing the desired single mutation was generated using the QuikChange Multi Site-Directed Mutagenesis kit (Stratagene) according to the manufacturer's protocol, and clones were sequenced for verification.

C-terminal Truncation of ADAMTS-4 in Cell Culture Is Not
Autoproteolytic but the Same Products Can Be Generated by Autoproteolysis-Structural details of the molecular forms of ADAMTS-4, the epitope locations, and the proposed cell surface complexes described in this report are provided in a schematic (see Fig. 12). Because the C-terminal truncation and activation of ADAMTS4 in the JJ012 cell system was completely blocked by 125 nM TIMP-1 (15) and because ADAMTS4 is not inhibited by TIMP-1 at these concentrations (23), it appeared to be unlikely that the cell-dependent process was autoproteolytic. However, it was shown recently (20) that a very similar Cterminal truncation process occurs on autoproteolytic digestion of recombinant ADAMTS4, and that this also results in marked increases in the aggrecanase activity of the recombinant enzyme (24). Therefore, to determine the relationship between the truncated forms generated in the JJ012 cell culture and by autoproteolysis, we prepared neo-epitope antibodies (JSCFRK and JSCALT) (see Fig. 12) to the C-terminals of the autoproteolytic forms (20) and compared the products from the two systems by Western analysis (Fig. 1) with the neo-epitope antibodies (Fig. 1, middle and right panels) as well as with combined antibodies to two internal peptide epitopes (JSCVMA and JSCYNH; Fig. 1, left panel). In keeping with previous reports (20,23), the purified recombinant preparation (rTS4) was largely the p68 form, although it did contain some of the p53 form (see left panel), and the p53 form was, as expected, highly reactive with JSCFRK (center panel) but not with JS-CALT (right panel). The autoproteolyzed sample (rTS4 auto) contained all three forms (left panel), and again, as expected, the p53 form was detected with JSCFRK (center panel), and the p40 form was detected with JSCALT only (right panel). The cell culture products (JJ012) migrated in the same positions as the autoproteolytic forms (left panel), and the species migrating at the p53 and p40 positions reacted with the JSCFRK (middle panel) and JSCALT (right panel) antibodies, respectively. It is clear that the neo-epitope antibody to the p53 form (middle panel) cross-reacts with the p40 form but not the p68 form, whereas the antibody to the p40 form (right panel) appears to be highly specific for that species. These data are consistent with the idea that the products generated by JJ012 cell-dependent proteolysis (JJ012, Fig. 1) result largely, if not entirely, from cleavage at the same sites (Lys 694 -Phe 695 for p53 and Thr 581 -Phe 582 for p40), as are cleaved autoproteolytically. We will therefore adopt the p68, p53, and p40 notation (20) for these three species in this paper (they were referred to previously as p75, p60, and p50, respectively (15)), and we will continue to refer to the full-length ADAMTS4 (containing the prodomain) as p100 (Fig. 1, left panel; see also Fig. 12).
The conclusion from the TIMP-1 data (15) that the cellderived p53 and p40 forms are not autoproteolytic was further supported by the finding that the same species were also generated by JJ012 cells transiently expressing an inactive mutant (E362Q) of ADAMTS4 (Fig. 2, lane 2), which was shown previously (20) not to generate the p53 and p40 species by autoproteolysis. In addition, we found (data not shown) that the p53 and p40 species formed by cells expressing the inactive mutant reacted with the appropriate neo-epitope antibodies (JSCFRK and JSCALT, respectively) and also confirmed that the C-terminally truncated active site mutant product was indeed inactive as an aggrecanase (Fig. 2, lane 4), under conditions where the wild-type product (see Fig. 2, lane 1) was active (Fig. 2, lane 3).
Identification by Immunodepletion of MT4-MMP as the Major Gelatinase of ADAMTS4-transfected Human Chondrosarcoma Cells-Because the C-terminal truncation and activation of ADAMTS4 in the JJ012 cells was not autoproteolytic but required a TIMP-1-inhibitable, GPI-anchored MMP, we next examined the possible role of MT4-MMP in this process. This was motivated by the finding (25) that MT4-MMP (MMP-17) is quite widely expressed in cell lines, whereas the only other known GPI-anchored MMP (MT6-MMP) appears to be very restricted in expression to neutrophils and some brain tumors (26). We examined both cell lysates and ECM preparations for MT4-MMP by Western analysis and also for gelatinase activity. This revealed major immunoreactive bands at about 64 and 105 kDa, and somewhat surprisingly, the major gelatinolytic activity in these samples also ran at 64 kDa (data not shown), consistent with its identification as MT4-MMP (27). To confirm this, portions of ECM were immunodepleted with antisera to MT4-MMP and MMP2, and the gelatin zymography clearly indicated that the 64-kDa species represented MT4-MMP (data not shown).  (Fig. 3, lower panel) and a parallel increase in the C-terminally truncated p53 form of ADAMTS4 in the medium (Fig. 3, upper panel, lanes 1-4) (Fig. 4). Because this cell line appears to be furin-deficient (15), transfection with ADAMTS4 alone (Fig. 4, lane 1) generated only the p100 full-length protein, whereas co-transfection with furin (Fig. 4, lane 2) generated mainly the furin-cleaved form (p68), and cotransfection with both furin and MT4-MMP (Fig.  4, lane 4) resulted in increased conversion of the p68 form to the p53 form (lane 4 versus lane 2). Interestingly, co-transfection with MT4-MMP alone (lane 3) did not alter the processing seen in control (lane 1), showing that furin-mediated cleavage to generate the p68 form (lane 2) is a prerequisite for MT4-MMP-mediated C-terminal truncation. Analysis of the media from these cultures for aggrecanase activity confirmed that only the medium containing high levels of the p53 form (Fig. 4,  lane 4) was active (not shown). It is therefore clear that the combination of furin cleavage, followed by MT4-MMP cleavage processing, is a feature of fibroblast-like cells as well as chondrocytic cells.

Inhibitors of GPI Anchor Synthesis Prevent Enhanced Cterminal Truncation by Co-transfection with MT4-MMP-We
had shown previously (15) that inhibitors of GPI anchor synthesis, such as 2-DFG and mannosamine, were capable of preventing ADAMTS4 C-terminal processing in JJ012-TS4 cells. To confirm that the identical pathway is operating in cells with high levels of processing attributable to co-transfection with MT4-MMP, we repeated this experiment at the same concentration of inhibitors (Fig. 5). The expected marked increase in p53 product was observed with MT4-MMP cotransfection (Fig.  5, lane 2 versus lane 1) and either 2-DFG (lane 3) or mannosamine (lane 4) were completely effective in blocking this increase. Indeed, the 2-DFG blocked even the control level of p53 formation. Analysis of portions of these media for aggrecanase activity (Fig. 5, lower panel) showed, as expected, that the Glu 373 -Ala 374 cleaving activity was well correlated with the abundance of the p53 form.
Kinetics of Processing of ADAMTS-4 by Association with GPI-anchored MT4-MMP-We showed previously that 150 nM TIMP-1 or a hydroxamate-based MMP inhibitor could completely block the formation of the p53 and p40 species but, unlike the GPI anchor inhibitors, allowed for appearance of the p68 species in the ECM fraction (see Fig. 5, Ref. 15). This is consistent with a processing pathway in which the transfer of the p68 form to the extracellular compartment is dependent on its association with intracellular MT4-MMP, followed by presentation of the complex on the cell surface for MT4-MMPmediated C-terminal truncation (see the schematic in Fig. 12). To further investigate features of the secretion and activation process, we next examined the effect of increasing concentrations of TIMP-1 on the abundance of the different forms in the cell-associated, ECM, and medium compartments (Fig. 6).
It can be seen that whereas 30 nM TIMP-1 was completely ineffective as an inhibitor of processing (Fig. 6, compare lanes  2, 5, and 8 to control lanes 1, 4, and 7, respectively), 270 nM TIMP-1 completely blocked p53 formation and caused an accumulation of the p68 form in the medium and ECM and also an accumulation of the p100 form in the cells. This is consistent with a model in which the p100 form is processed intracellularly by furin cleavage to the p68 form, which in turn is secreted in association with MT4-MMP for cleavage, in a TIMP-1-sensitive step, on the cell surface.
To directly test the idea that the p68 associates with GPIanchored MT4-MMP on the cell surface, we next examined the capacity of PIPLC treatment of isolated cells to solubilize the two proteins. As expected, cleavage of GPI anchors by PIPLC treatment released a proportion (Fig. 7A, lane 4), but not all, of the 64-kDa MT4-MMP from the cells (panel A, lane 2), whereas there was essentially no release without PIPLC (lane 3); the same pattern was observed in the release of the p68 form of ADAMTS4 (Fig. 7B, lanes 6 and 8), showing that it is released by PIPLC treatment in a manner consistent with its association with GPI-anchored cell surface MT4-MMP. Interestingly, a proportion of the p100 form was released from the cells by incubation in medium alone (Fig. 7B, lane 7), and this process was enhanced by PIPLC treatment, as seen in the disappearance from the cell pellet (lane 6 versus lane 5) and the increase in the medium (lane 8 versus lane 7). This suggests that a proportion (ϳ50%) of the p100 form is loosely associated with Cultures of VA13-TS4 cells were grown to 80% confluence as described previously (15), and the medium was changed to growth medium containing GenePORTER with no addition (lane 1), 1 g of cDNA to furin (lane 2), 1 g of cDNA to MT4-MMP (lane 3), or 1 g of cDNA to furin and 1 g of cDNA to MT4-MMP (lane 4). After 48 h, the medium was changed to serum-free medium, and after an additional 3 days, the medium was analyzed for ADAMTS4 with combined antibodies JSCVMA/JSCYNH.

FIG. 5. MT4-MMP-enhanced C-terminal truncation of AD-AMTS4 is blocked by inhibitors of GPI anchor synthesis.
Cultures of JJ012-TS4 cells were grown to ϳ80% confluence, and the medium was changed to growth medium containing GenePORTER with no addition (lane 1) or 1 g of cDNA for MT4-MMP (lanes 2-4). After 48 h, the medium was changed to serum-free medium containing no addition (lanes 1 and 2), 10 M 2-DFG (lane 3), or 1.5 mM mannosamine (lane 4). After an additional 3 days, the medium was analyzed for ADAMTS4 with antibody JSCVMA. Portions of medium were also assayed for aggrecanase activity (lower panel) with JSCNIT, and the G1-NITEGE product is shown (arrowhead).

FIG. 6. MT4-MMP-enhanced C-terminal truncation of AD-AMTS4 is blocked by TIMP-1.
Triplicate cultures of JJ012-TS4 cells were grown to ϳ80% confluence, and the medium was changed to growth medium containing GenePORTER with 1 g of cDNA for MT4-MMP. After 48 h, the medium was changed to serum-free medium containing no addition (lanes 1, 4, and 7), 30 nM TIMP-1 (lanes 2, 5, and 8), or 270 nM TIMP-1 (lanes 3, 6, and 9). After an additional 3 days, the culture reactions were terminated by separation of the medium, ECM, and cell pellet, and 10% proportions of each were taken for Western analysis for ADAMTS4 with JSCVMA. the cells, whereas the remainder is associated with GPI-anchored MT4-MMP, indicating that in these cells, it can appear on the cell surface without removal of the prodomain. A repeat of this experiment with ADAMTS-4-transfected VA-13 fibroblast-like cells also showed that treatment with PIPLC released all of the p68 form, and again the p100 form was ϳ50% released in buffer alone and 50% released by PIPLC (data not shown).
To further examine the association of p68 and MT4-MMP, we next took a PIPLC-released supernatant containing both p68 and MT4-MMP and carried out bidirectional immunoprecipitation. Precipitates were prepared with (and without) anti-MT4-MMP and anti-TS4 (JSCYNH), and the resulting products were analyzed with JSCVMA/JSCYNH and anti-MT4-MMP, respectively. The results (Fig. 8) showed that the p68 and MT4-MMP were indeed associated in the PIPLC-released material and that this association was sufficiently robust to allow for immunoprecipitation with antibodies to either component. In the absence of precipitating antibodies, neither the p68 nor MT4-MMP was detected on blots (not shown), and neither species was detected in immunoprecipitates with antibodies to an irrelevant protein, human biglycan (Fig. 8, lanes 3 and 6).
Most interestingly, in contrast to the p100 and p68 species, the PIPLC treatments released little if any of the cell-associated p53 form (Fig. 7B, lane 6), suggesting that p53 is associated with a cell surface component other than GPIanchored MT4-MMP.
The p53 Form Is Associated with CS and HS on Cells-The studies with PIPLC suggested that during or after formation of p53 by MT4-MMP cleavage, it is transferred to a different cell surface "receptor." Because p53 has been shown to require 0.4 M NaCl for elution from heparin-Sepharose (20), we examined the possibility that it is bound to the cell surface by association with sulfated glycosaminoglycans. For this purpose, we prepared PIPLC-treated cells (as in Fig. 7, lane 6) and digested them with chondroitinase ABC alone, heparitinase II alone, or both enzymes sequentially. Western analysis (Fig. 9) showed that ϳ50% of p53 was released from the cells and recovered in the supernatant after treatment with chondroitinase ABC alone (Fig. 9, lanes 2 and 6) or heparitinase II alone (Fig. 9,  lanes 3 and 7), and it was completely released after sequential treatment with the two enzymes (Fig. 9, lanes 4 and 8). This suggests that the p53 is associated with these cells through an interaction with both CS and HS (see Fig. 12).
To examine whether these GAG-lyase treatments digested all of the GAG on these cells (prepared as in Fig. 9, lane 1), we incubated them with chondroitinase ABC or heparatinase II or buffer alone (see "Experimental Procedures" for details) and subsequently assayed for CS or HS contents by fluorescenceassisted carbohydrate analysis. The data (Fig. 10) showed that JJ012 cells, enriched in the p53 form, contained both cellassociated HS (unsulfated, mono-and disulfated disaccharides in lane 1) and CS (predominantly 4-sulfated disaccharides in lane 3). Moreover, these cell-associated GAGs were each essentially completely removed by digestion with heparitinase II (lane 2) or chondroitinase ABC (lane 4).
The p53 Form Is Associated Only with GAGs on Syndecan-1-The only transmembrane proteoglycan shown clearly to be substituted with both CS and HS is syndecan-1 on normal mouse mammary epithelial cells (28 -30). Because JJ012 cells were found to contain mRNA transcripts for this syndecan-1 FIG. 7. Cleavage of GPI anchors with PIPLC releases both MT4-MMP and p68 ADAMTS4. Cultures of JJ012-TS4 cells were grown to ϳ80% confluence, and the medium was changed to growth medium containing GenePORTER and 1 g of cDNA for MT4-MMP. After 48 h, the medium was changed to serum-free medium, and after an additional 3 days, the culture reactions were terminated by separation of the medium, ECM, and cell pellet fractions. Cell pellets were treated in serum-free medium without and with PIPLC (see "Experimental Procedures" for detail), and the released products were sepa-  Fig. 7) were taken for immunoprecipitation (see "Experimental Procedures") with anti-MT4-MMP, an-tiTS-4 (JSCYNH), or anti-biglycan as a control. The starting material (lanes 1 and 4) and the precipitates (lanes 2, 3, 5, and 6) were taken for Western analysis for ADAMTS-4 with combined antibodies JSCVMA/ JSCYNH (lanes 1-3) or for MT4-MMP (lanes 4 -6). *, IgG heavy chain. Fig. 7, lane 6) were treated with buffer alone, chondroitinase ABC, and/or heparitinase II as described in "Experimental Procedures." Equal proportions of cell-associated products (lanes 1-4) and supernatants (lanes 5-8) were taken for Western analysis with combined antibodies JSCVMA/JSCYNH. The treatment combinations (ϩ and Ϫ) are shown, and both a high exposure (A) and a low exposure (B) are shown to more clearly illustrate the relative abundance of p53 product.

FIG. 9. Demonstration that ADAMTS-4 (p53) is associated with both CS and HS on the cell surface of JJ012 cells. PIPLC-treated cell pellets from replicate cultures (see
(data not shown), we examined the possibility that the p53 was bound to the GAG chains of this proteoglycan. We approached this in two ways. We took PIPLC-treated cells as starting material and carried out immunoprecipitation (of the lysates of untreated cells and cells after sequential Chase ABC/heparitinase II treatment) with anti-ADAMTS4 (JSCYNH), anti-syndecan-1 (B-B4), and anti-biglycan as a control. The immunoprecipitates containing syndecan-1 as a proteoglycan were deglycosylated (see "Experimental Procedures"), and samples were run for Western analysis with anti-ADAMTS4 (Fig. 11,  lanes 1-3) and anti-syndecan-1 (lanes 4 -6). The results showed that the anti-syndecan-1 immunoprecipitate of untreated cells contained the p53 (lane 1), whereas the immunoprecipitate of deglycosylated cells did not (lane 2). In addition, the anti-TS4 immunoprecipitate of untreated cells contained abundant syndecan-1 (lane 4), whereas the immunoprecipitate of deglycosylated cells contained a much reduced amount (lane 5). The presence of some syndecan-1 in the immunoprecipitate of deglycosylated cells is consistent with the finding (not shown) that in this sample (and in some other experiments), the deglycosylation was not complete, and some p53 remained associated with the cell pellet. The immunoprecipitation of p53 by antisyndecan-1 (Fig. 11, lane 1) and of syndecan-1 by anti-AD-AMTS4 (lane 4) was specific in both cases because no products were obtained with the control antibody (lanes 3 and 6). To confirm that the association of the p53 to syndecan is through the GAG chains alone, we next took PIPLC-treated cells and prepared equal amounts of p53 on untreated cells and in the supernatant of treated cells. These samples were then immunoprecipitated with anti-syndecan-1 (B-B4) and analyzed on Western blot with anti-ADAMTS-4 and anti-syndecan-1 (mS1ED) The results (not shown) confirmed that the B-B4 immunoprecipitate of the cell lysate contained abundant p53, whereas the immunoprecipitate of the supernatant p53 did not. Furthermore, the B-B4 immunoprecipitate of the cell lysate contained syndecan-1, whereas the immunoprecipitate of the supernatant did not. In summary, we conclude that the p53 is bound to syndecan-1 on these cells through interaction with substituent CS and HS and not through interaction with the syndecan-1 core protein. DISCUSSION A Model for ADAMTS4 Processing and Activation-The data presented (summarized in Fig. 12) suggest that ADAMTS4, and therefore perhaps other aggrecanases (ADAMTS-1, -4, -5, -9), are subject to a complex but discrete set of interactions with other cellular products during the process of secretion and cell surface activation. After furin-mediated prodomain removal, it appears that the p68 intermediate form associates with GPIanchored MT4-MMP. Whether this association occurs intracellularly or only on the cell surface is not known, although in most experiments, PIPLC treatment eliminated Ͼ80% of the p68 from the cells (see Fig. 7, lanes 5 and 6), suggesting that there is no large pool of intracellular complex. The association appears to be quite stable, and it does not mask antibody binding sites because it can be trapped by immunoprecipitation with either anti-TS4 or anti-MT4-MMP. In addition, similar experiments with a fibroblast cell line, VA13, and primary bovine synovial cells (unpublished data) confirmed that, in these cells, the p68 form is also generated and associates with a PIPLC-releasable component on the cell surface.
During maintenance of the JJ012 cells in serum-free medium, there follows an MT4-MMP-mediated proteolysis of the p68 to generate the p53 form, which can be detected both on the cell surface (in association with the GAG chains of syndecan-1) and in the medium. By analysis with C-terminal specific neoepitope antibodies (Figs. 1 and 12), we have found that this proteolysis occurs at the Lys 694 -Phe 695 site, identified previously as an autocatalytic cleavage site (20), suggesting that this bond is particularly susceptible to proteolysis by both MT4-MMP and ADAMTS4 and perhaps other metalloproteinases. Such MT4-MMP-mediated proteolysis at Thr 581 -Phe 582 also generates the p40 form, which is generated in variable yield in this system and is not found on the cell surface but appears to be released directly to the medium, consistent with the reduction in GAG-binding affinity observed with the identical autocatalytic product (20). Proteolysis of the p68 to the p53 form was totally blocked by TIMP-1 at 270 nM (Fig. 6), and this was accompanied by accumulation of p68 in the medium and ECM but not in the cell-associated fraction, suggesting that the TIMP-1 was interacting with the MT4-MMP and resulting in a dissociation of the ADAMTS4 p68 from the complex without proteolysis. This mode of inhibition was apparently different from that provided by the GPI anchor synthesis inhibitors (Fig.  5), which resulted instead in the non-appearance of the p68, as would be expected if furin-mediated cleavage of the p100 to p68 and/or p68 translocation to the cell surface were dependent on the continued presence of GPI-anchored MT4-MMP. This conclusion provides a simple explanation for the findings (17,18,31) that inhibitors of GPI-anchor synthesis can block IL-1induced aggrecanolysis in cartilage explants.
In cultures where the cell surface p68 had been depleted by proteolysis or release over 3 days in serum-free medium, the p53 form was recovered largely in the medium (see Fig. 6, lanes  1, 4, and 7); however, in some cultures, it was found that abundant p68 was still present on the cell surface, and in this case there was a similarly abundant cell surface pool of p53 (Fig. 7, lane 5), suggesting that processing was still occurring. In these cultures, the p53 was found to be associated with both CS and HS on syndecan-1 (Figs. 9 -11). Although we have not shown that the CS and HS chains are present on the same core protein (as suggested in Fig. 12), the composition of the GAGs (Fig. 10) is consistent with that described for CS and HS on a single syndecan-1 core protein produced by normal epithelial cells (28,29). It is intriguing that the p53 is found associated with both types of GAG, because this would imply that the binding of the proteinase to the GAG chains is not attributable  1 and 2) and Chase ABC digestion (lanes 3 and 4) of the released products (lanes 1 and 3) and the cell-retained products (lanes 2 and 4). The disaccharides were identified by co-migration with commercially available standards. Unlabeled bands have not been identified. *R, reagent product. to a unique structural motif found in either CS or HS but rather a more general binding property of all GAG chains on syndecan-1 for ADAMTS4. Indeed, it has been postulated (29) that the more distal HS chains on syndecan-1 may transfer bound protein to the membrane-proximal CS chains for further processing. Significantly, the finding that the p53 alone binds to syndecan-1, despite the presence of both p68 and p40 in the system, suggests that the syndecan GAG-binding activity on ADAMTS4 is masked by the spacer domain (on p68) and that both the thrombospondin-1-like and cysteine-rich regions are required for binding of p53 to GAG (see Fig. 12 for domain compositions of p68, p53, and p40). Indeed, a requirement of the thrombospondin-1 motif for binding of ADAMTS-4 to aggrecan has been described previously (32).
Whereas the synthesis of GPI-anchored MT4-MMP appears to be required for ADAMTS4 secretion and activation in this cell system, the influence of syndecan-1 on the process is not clear. Although the highly active p40 form is apparently released to the medium without association with syndecan-1, the finding of p53 bound to the CS and HS chains of syndecan-1 suggests that the MT4-MMP and syndecan-1 (both with membrane associations) may themselves be closely associated in an activation complex. Thus, it seems unlikely that the MT4-MMP-generated p53 would become freely diffusible before binding to the GAG components of syndecan-1. Whether the syndecan-1 is required as a cofactor for MT4-MMP activity in this complex is not known; however, such a requirement might explain why we have not been able to reproducibly generate the p53 and p40 forms of ADAMTS4 by incubation of the p68 form with a commercially available recombinant catalytic domain of MT4-MMP. Moreover, whether the p68 and p53 forms are active when present in these associations is unknown. Further work will establish whether this or a related activation complex is also required for processing of the other aggrecanases, ADAMTS-1, -4, -5, and -9.
Control of Aggrecanase Activity by the Activation Complex-The discovery in cartilage explants (33,34) and human synovial fluids (35,36) of products of the "aggrecanase"-dependent cleavage of aggrecan at the Glu 373 -Ala 374 bond has been followed by the description of at least seven proteinases (atrolysin C (37), MMP-8 (38), MT1-MMP (39)), ADAMTS-1 (40,41), ADAMTS-4 (42), ADAMTS-5 (43), and ADAMTS-9 (6) with the capacity to cleave at this "aggrecanase" site. Although some studies have indicated a predominant role for ADAMTS4 and/or ADAMTS5 in human osteoarthritic cartilage (44), the lack of highly specific inhibitors or assay methods and a poor understanding of the activation states of these proteinases means that the relative importance of these enzymes in aggrecanolysis in different normal tissues and pathologies remains unclear. Treatment of most cartilages with IL-1, tumor necrosis factor-␣, retinoic acid, fibronectin fragments, or neprilysin (16,18,(45)(46)(47)(48)(49)(50)(51)(52) markedly enhances aggrecanolysis in the tissue; however, there is a lack of data on the effect of these treatments on mRNA, protein, or activity levels for the different aggrecanases. Indeed, earlier work, which showed that TIMP-1 and MMP inhibitors (16,53) or inhibitors of GPI anchor synthesis  11. Demonstration of the association of ADAMTS-4 (p53) and the GAG chains of syndecan-1 by immunoprecipitation of JJ012 cell lysates. JJ012 cells were prepared (as in Fig. 7, lane 6), and a portion was digested sequentially with Chase ABC and Hepase II. Undigested and digested cell lysates were taken for immunoprecipitation with anti-syndecan-1 (B-B4), anti-ADAMTS4 (JSCYNH), and anti-biglycan. Western analysis of immunoprecipitates for ADAMTS-4 with JSCVMA (lanes 1-3) and for syndecan-1 with mS1ED (lanes 4 -6) is shown. See text for details. ‫,ء‬ IgG heavy chain. (17) could block the induction of aggrecanase activity in cartilage explants, were interpreted in terms of these agents inhibiting an aggrecanase-activator proteinase. Most recently (54), evidence has been presented for the "activation" of constitutively expressed cell surface-bound ADAMTS4 by IL-1 treatment of cartilage explants. Evidence for the existence of an activation cascade in the aggrecanase response in IL-1-stimulated cartilage explants was also suggested by work with inhibitors of both cysteine endopeptidases (55) and serine endopeptidases (56). Maybe these earlier results can be explained by the fact that a cysteine endopeptidase, Gpi8p, is required for the transamidation step in GPI anchor formation (57) and also that both ADAMTS4 and MT4-MMP require "furin-like" serine proteinase activity to remove their prodomains during zymogen activation.
Whether the proposed activation complex is required for activation of ADAMTS4 in all cell types is unknown, although it is clear from analysis of a range of tissues including cartilage (15), synovial membrane (15), aorta (10), and brain (58) that the C-terminally truncated forms appear to predominate in these tissues. In addition, it also seems very likely that MT4-MMP is responsible for ADAMTS4 activation in normal articular cartilages, because in studies with bovine cartilage explants 2 we have shown that IL-1 treatment is accompanied by conversion of p68 to p53 and an increased abundance of extractable 64-kDa MT4-MMP. Furthermore, the inclusion of the GPI anchor synthesis inhibitor, mannosamine, in these cultures blocked the appearance of MT4-MMP and the activation of ADAMTS4 by C-terminal truncation.
If the activity of ADAMTS4 in cartilages, and perhaps in other tissues such as aorta (10) and brain (59), is primarily controlled through cell surface processing on an MT4-MMP/ syndecan-1 complex, this has significant implications for the design of therapeutics aimed at controlling ADAMTS4-mediated degradation of aggrecan, versican, and brevican. Agents that could modify the synthesis, furin-mediated activation, and/or GPI anchoring of MT4-MMP might secondarily result in specific control of ADAMTS4 activation and so prevent uncontrolled aggrecanolysis in arthritic cartilage (60) or promote repair in the injured central nervous system, where intact versican and/or brevican may prevent neurite extension and synapse formation (61). Indeed, agents that could specifically interfere with the association of ADAMTS4, MT4-MMP, and syndecan-1 on the cell surface might be particularly useful.
In summary, the present work suggests that GPI-anchored MT-MMPs and syndecans may collaborate on the surface of cells to maintain normal ECM homeostasis and also to respond to pro-inflammatory signals by increased activation of AD-AMTS proteinases, which cleave aggregating proteoglycans. In this regard, it will be interesting to determine whether the inhibitory effects of sulfated GAGs on cell-dependent aggrecanolysis in cartilage explants (62) can be explained by this model. Finally, in a recent report, recombinantly expressed equivalents of the p68, p53, and p40 forms of ADAMTS4 have been analyzed for activity against the aggrecanase sites in the GAG attachment region and the interglobular domain of bovine aggrecan (63). This study confirmed the reported inactivity of the p68 form and the high activity of the p53 form against the Glu 373 -Ala 374 bond (15) but also showed that the p68 form readily cleaves aggrecan in the CS-attachment region only. It therefore appears that the MT4-MMP-mediated activation process described in the present work converts ADAMTS4 from an enzyme that may promote cartilage matrix assembly and organization, by C-terminal processing of aggrecan, into a form (p53 or p40) that is uniquely capable of destructive aggrecanolysis of the kind seen in human joint diseases (35,64).