TIMP-3 Inhibition of ADAMTS-4 (Aggrecanase-1) Is Modulated by Interactions between Aggrecan and the C-terminal Domain of ADAMTS-4*

ADAMTS-4 (aggrecanase-1) is a glutamyl endopeptidase capable of generating catabolic fragments of aggrecan analogous to those released from articular cartilage during degenerative joint diseases such as osteoarthritis. Efficient aggrecanase activity requires the presence of sulfated glycosaminoglycans attached to the aggrecan core protein, implying the contribution of substrate recognition/binding site(s) to ADAMTS-4 activity. In this study, we developed a sensitive fluorescence resonance energy transfer peptide assay with a Km in the 10 μm range and utilized this assay to demonstrate that inhibition of full-length ADAMTS-4 by full-length TIMP-3 (a physiological inhibitor of metalloproteinases) is enhanced in the presence of aggrecan. Our data indicate that this interaction is mediated largely through the binding of glycosaminoglycans (specifically chondroitin 6-sulfate) of aggrecan to binding sites in the thrombospondin type 1 motif and spacer domains of ADAMTS-4 to form a complex with an improved binding affinity for TIMP-3 over free ADAMTS-4. The results of this study therefore indicate that the cartilage environment can modulate the function of enzyme-inhibitor systems and could have relevance for therapeutic approaches to aggrecanase modulation.

The disintegrin metalloproteinases with thrombospondin motifs (ADAMTS) 3 are a novel family of extracellular proteases forming an integral part of the extracellular matrix itself. Along with serine proteases, matrix metalloproteinases, bone morphogenetic protein-1/tolloid metalloproteinases, and ADAM (a disintegrin and metalloproteinase) proteins, ADAMTS pro-teins play a pivotal role in the proteolytic processing and turnover of the component molecules of the extracellular matrix of a broad range of tissues and have potential roles in the turnover of cell-surface proteins. ADAMTS proteins share the characteristic protease, disintegrin-like, and cysteine-rich domains in common with ADAM proteins, but differ in being soluble rather than membrane-bound and by the presence of thrombospondin type 1 (TSP-1) repeats (1,2). ADAMTS-2, -3, and -14 are potential procollagen N-proteinases, and ADAMTS-13 has been identified as a von Willebrand factor-cleaving protease. ADAMTS-4 is a member of the "angiogenesis/aggrecanase" group of ADAMTS proteases (which also includes ADAMTS-1, -5, -8, -9, and -15) and is unique among the currently known ADAMTS proteases in containing only a single TSP-1-like motif, located between its disintegrin-like and cysteine-rich domains, and lacking any C-terminal TSP-1-like repeats (see Fig. 1). Like the other members of the angiogenesis/ aggrecanase group and ADAMTS-9, ADAMTS-4 has been demonstrated to act as an aggrecanase in vitro. In common with other aggrecanases, ADAMTS-4 is able to cleave aggrecan at multiple sites (five in total) (see Fig. 1) (3); however, it is one of only four that cleave aggrecan at the Glu 373 -Ala 374 "interglobulin domain" cleavage site (the others being ADAMTS-1, -5, and -8, which cleave with varying affinities). Aggrecan is a proteoglycan that is present in articular cartilage and that plays an important role in the ability of cartilage to withstand damage during the mechanical compression of a joint; thus, aggrecanases represent a potential target for the treatment of both osteoarthritis and rheumatoid arthritis (4 -6). The efficient cleavage of aggrecan by ADAMTS-4 has been shown to require the presence of both the sulfated glycosaminoglycans (GAGs; chondroitin sulfate and keratan sulfate) attached to the aggrecan core protein and the TSP-1-like motif of ADAMTS-4, which contains a putative GAG-binding site. The treatment of bovine aggrecan to remove keratan sulfate was shown to result in loss of cleavage at the Glu 373 -Ala 374 cleavage site (7,8). In contrast, the C-terminal domains of ADAMTS-4 (cysteinerich/spacer) have been shown to interact directly with the GAG side chains of aggrecan to inhibit cleavage at Glu 373 -Ala 374 , with C-terminal truncation of these domains resulting in markedly increased activity relative to the full-length protein. Indeed, with the spacer domain intact, full-length ADAMTS-4 demonstrates only a relatively weak activity at the bond between Glu 373 and Ala 374 , instead cleaving preferentially at sites in the chondroitin sulfate-rich region (see Fig. 1) (5,9).
Autocatalytic processing of ADAMTS-4 (68 kDa) has been demonstrated, resulting in the production of two smaller isoforms of 53 and 40 kDa through cleavage within the spacer and cysteine-rich domains, respectively (10). In addition, membrane type 4 matrix metalloproteinase has been shown to remove the spacer domain of ADAMTS-4 (11). It is therefore possible that proteolytic modulation of these interactions may represent a regulatory mechanism in vivo.
Although there is plenty of evidence for the importance of these C-terminal domains in the interactions between ADAMTS-4 and its physiological substrates, no investigation has yet studied their effects on the inhibitory action of the tissue inhibitors of metalloproteinases (TIMP), a series of well described inhibitors of ADAMTS-4 (12,13). Of the four currently identified species of TIMP, only TIMP-3 has shown potent inhibition of aggrecanase activity by ADAMTS-4 (K i ϭ 7.9 nM), and TIMP-1 and TIMP-2 are inhibitory at higher concentrations (K i Ն 350 nM) (12). At present, the mechanisms underlying the regulation of TIMP inhibition of ADAMTS-4 in vivo remain undefined. In this study, we propose that TIMP interactions are selectively regulated by electrostatic modification of the C-terminal domains of ADAMTS-4 by aggrecan GAGs, invoking a mechanism independent of the conformational state of the catalytic cleft.
Identification of a Human ADAMTS-4 Substrate from a Phage-displayed Peptide Library and Synthesis of Fluorescent Peptide Substrate-An M13 phage-displayed 10-mer random peptide library (fGWmAb179X10) was constructed by inserting a partially degenerated DNA cassette into BbsI-digested phage display vector fGW (16). The DNA cassette encoded the monoclonal antibody (mAb) 179 tag (ACLEPYTACD) immediately downstream of the M13 gene III signal sequence and a 10-residue stretch of random amino acids (ACLEPYTACD-SAXGGGGS, where X represents any of the 20 common amino acids). Following library phage preparation, a small fraction of transformed Escherichia coli MC1061 cells was plated on LB/tetracycline plates to determine the total number of primary transformants, which was estimated to be ϳ8.5 ϫ 10 9 . A total of 1 ϫ 10 12 fGWmAb179X10 phage were digested at 37°C for 1 h by 10 nM human aggrecanase-1 (ADAMTS-4) in a 250-l reaction mixture containing 1ϫ aggrecanase-1 buffer (50 mM HEPES (pH 7.5), 100 mM NaCl, and 10 mM CaCl 2 ). The reaction was terminated by the addition of 0.5 M EDTA to a final concentration of 5 mM. mAb 179 (100 g) was added to the terminated reaction mixture, followed by the addition of bovine serum albumin to 0.1%. After 30 min of incubation on ice, allowing mAb 179 to capture uncleaved phage, Pansorbin cells (100 l) were added, and the reaction mixture was rotated at 4°C for 1 h. The mixture was centrifuged for 2 min at 12,000 ϫ g, and the supernatant containing cleaved phage was recovered. The uncleaved phage-mAb complex bound to Pansorbin cells in the pellet was discarded. Phage in the supernatant was transfected into E. coli K91 cells and amplified overnight at 37°C. Phage in a small aliquot of the supernatant was titered on E. coli K91 cells. The above screening was performed for five rounds as described previously except only mAb 179 was used (17). Single phage clones selected by human ADAMTS-4 were picked randomly and grown in 4 ml of LB medium for isolation of DNA for sequencing. A human ADAMTS-4 substrate with a peptide sequence of AELQGRPISIA was identified. This sequence was then used to construct a fluorescence resonance energy transfer (FRET) peptide substrate, carboxyfluorescein (FAM)-AELQGRPISIAK-N,N,NЈ,NЈ-tetramethyl-6-carboxyrhodamine (TAMRA), using the FRET pairs fluorescein and tetramethylrhodamine and the protocol described by Pope et al. (18) and Alvarez-Iglesias et al. (19).
Expression and Purification of Recombinant ADAMTS-4-A total of five isoforms of ADAMTS-4 were used in this study (see Fig. 1). The full-length mature form of ADAMTS-4 with a C-terminally His-tagged sequence (TS4-1) was expressed in an Epstein-Barr virus nuclear antigen-expressing human embryonic kidney 293 cell line and purified from conditioned medium by chromatography on a Radial Q column and heparin-Sepharose. Two C-terminal domain deletion enzymes containing both catalytic and disintegrin domains (TS4-4) and the catalytic domain alone (TS4-5) were expressed in Drosophila S2 cells with a C-terminally His-tagged sequence and purified from diafiltered insect cell culture supernatants by nickel-nitrilotriacetic acid (Ni-NTA) chromatography. In addition, a C-terminal domain deletion enzyme containing the catalytic, disintegrin, and thrombospondin domains (TS4-3) with a C-terminal His tag expressed in a baculovirus system and purified from insect cell culture supernatants was purchased from Calbiochem.
Analysis of Enzyme Activity-Enzyme activity measurements were performed by continuous monitoring of peptide cleavage, i.e. real-time increases in fluorescence intensity, at an excitation wavelength of 485 nm and an emission wavelength of 535 nm using a Bio-Tek Instruments FL600 microplate spectrofluorometer.
Inhibitor samples were diluted in 50 mM Tris-HCl (pH 7.5) and 12.5% glycerol and dispensed into Nunc black polystyrene 384-well microtiter plates (VWR International, Leicestershire, UK) in a volume of 10 l. ADAMTS-4 and FAM-AELQGRPISIAK-TAMRA peptide were diluted 50 mM BisTris propane (pH 8.0), 150 mM NaCl, 10 mM CaCl 2 , and 0.01% Triton X-100, and 20 l was added to the inhibitor plate such that their final assay concentrations were 200 pM and 0.5 M, respectively. Peptide hydrolysis was measured for 4 h at 20°C.
Apparent K i values were determined using the Robosage Version 7.31 add-in (GlaxoSmithKline, Research Triangle Park, NC) for Excel (Microsoft Corp., Redmond, WA) and the four-parameter logis- where x is the concentration of the test sample, y is the percent inhibition, A is the minimum response, D is the maximum response, C is the apparent K i (K i(app) or IC 50 ) for the curve, and B is the Hill slope. Mean values for K i(app) were determined from a fit of the mean percentage inhibition values, at each concentration of inhibitor, to the four-parameter logistic equation. The S.E. of the curve fit (S.E. of the estimate) was determined from the average scatter of the points from the fitted curve using the Robosage Version 7.31 curve-fitting tool. For significance testing, individual paired values of K i(app) were converted to pK i(app) (Ϫlog K i(app) ) and analyzed using the "t-Test: Paired Two Sample for Means" analysis tool in Microsoft Excel.   . This sample also contained a co-purified N-terminally truncated isoform of ADAMTS-4 lacking a functioning catalytic domain (identity confirmed by peptide mass fingerprinting using MALDI mass spectroscopy) (lane 1, lower band). Reaction products were purified through sequential Ni-NTA spin columns to produce the purified TS4-2* isoform (lanes 3 and 4). B, amino acid sequence of the C-terminal cysteine-rich and spacer domains of ADAMTS-4 showing (i) the most C-terminal amino acid residue confirmed in TS4-2* by MALDI mass spectroscopy sequence assignment (arrow) and (ii) the autocatalytic cleavage site described by describes the enzyme concentration. In determining K m values, the effects of internal quench due to inner filter from the FRET peptide substrate on the kinetic rates determined were corrected for by calibration against a fluorescein-labeled peptide control (20).
Association and dissociation rate constants were determined by continuous monitoring of peptide cleavage by 400 pM ADAMTS-4 using GraFit Version 5.0 software. Initial velocities (V 0 ) and steady-state velocities (V s ) of substrate hydrolysis were determined at various concentrations of TIMP-3 (0 -40 nM) and used to estimate the association (k on ) and dissociation (k off ) rate constants by applying the equations F ϭ (V s t ϩ (V 0 Ϫ V s ))/((1 Ϫ e (Ϫkobst) )/k obs ) ϩ F 0 and k obs ϭ k on [I] ϩ k off , where F is the observed fluorescence, F 0 is the initial fluorescence, k obs is the observed rate constant, t is time, and [I] is the inhibitor concentration.  Table  1). The assay was highly sensitive, with peptide hydrolysis by as little as 50 pM ADAMTS-4 producing excellent zЈ values (averaging 0.85), confirming that the assay was statistically highly reproducible (21). TS4-5 was catalytically inactive, in agreement with previously published findings (9). However, the isoforms TS4-2*, -3, and -4 all demonstrated measurable rates of peptide hydrolysis at subnanomolar concentrations, and K m and k cat /K m values similar to those observed using the fulllength ADAMTS-4 construct were observed (Table 1). In each case, complete inhibition of the cleavage was achieved by the addition of 100 M CGS27023 (22).

Cleavage of the FAM-TAMRA Peptide by Recombinant ADAMTS-4 and Truncated
TIMP Inhibition of ADAMTS-4 Using the FAM-TAMRA Peptide Assay-Using the conditions described above, we investigated the effects of the addition of TIMP-3, TIMP-1, and the broad-spectrum metalloproteinase inhibitor CGS27023 to TS4-1 in the FAM-TAMRA peptide assay. A K i of 6.38 Ϯ 1.51 nM was determined for TIMP-3. This compares with a published value of 7.9 nM using aggrecan as the substrate (12). Little or no inhibition of TS4-1 was observed with TIMP-1 (K i Ͼ 350 nM), and a K i of 0.87 Ϯ 0.16 M was determined for CGS27023.
Aggrecan Enhances TIMP-3 Potency-We investigated the effect of the potential physiological substrate aggrecan on the interaction between ADAMTS-4 and TIMP-3 using the FAM-TAMRA peptide assay. The addition of 10 g/ml aggrecan to the assay in the absence of TIMP-3 resulted in an ϳ50% inhibition of peptide hydrolysis due to competitive inhibition by the presence of an alternative substrate at a concentration close to its Michaelis constant (23). The K m and k cat values for ADAMTS-4 determined in the presence of 10 g/ml aggrecan were also consistent with expectations for a purely competitive inhibition of the alternative substrate aggrecan with respect to peptide cleavage (Table 2). Interestingly, the presence of aggrecan resulted in an increase in the potency of TIMP-3, with an order of magnitude increase in potency observed at aggrecan concentrations Ͼ1 g/ml (Fig. 3).
In contrast, CGS27023 inhibition was unaffected by the presence of 10 g/ml aggrecan, with a K i(app) of 1.23 Ϯ 0.24 M being measured. As before, no activity was measured for TIMP-1 under these assay conditions (K i(app) Ͼ 350 nM). Association and disso-

TABLE 1 K m and derived k cat /K m values for ADAMTS-4 constructs TS4-1 to TS4-4
Data are the means (n ϭ 3) calculated from the total protein concentration.  ciation constants for TIMP-3 inhibition of ADAMTS-4 were also determined in the presence and absence of aggrecan and demonstrated that the addition of aggrecan resulted in an increase in the observed k on , but had little effect on k off ( Table 3).

Enhancement of TIMP-3 Potency by Aggrecan Is
Mediated through GAGs-Having ascertained that chondroitin sulfate did not significantly modify the K m and k cat values for ADAMTS-4 turnover of the FRET peptide (Table 2), we monitored its effect on TIMP-3 interactions with ADAMTS-4 using this assay. Whereas chondroitin 4-sulfate had little effect on K i values, chondroitin 6-sulfate enhanced TIMP-3 binding to ADAMTS-4 at low doses (Fig. 4). Chondroitin 4-sulfate also had the effect of interfering significantly with the aggrecan potentiation of TIMP-3 inhibition, whereas chondroitin 6-sulfate had less effect. Keratan sulfate demonstrated no determinable effect on TIMP-3 inhibition of ADAMTS-4 in either the presence or absence of aggrecan (Fig. 5). The effects of deglycosylating aggrecan on its enhancement of TIMP-3 potency were also studied, and in line with the results obtained in studies of chondroitin sulfates, deglycosylated aggrecan was markedly less efficient in the modulation of TIMP-3 potency (Fig. 6).
Aggrecan Modulation of TIMP-3 Potency Is Dependent on the ADAMTS-4 C-terminal Domain-TIMP-3 inhibition of the four C-terminally truncated isoforms of ADAMTS-4 was studied in the absence and presence of aggrecan (Fig. 1). Under normal assay conditions, all the forms gave similar binding constants with TIMP-3, although the TS4-4 construct containing the catalytic and disintegrin-like domains showed a small enhancement of affinity for TIMP-3 (Table 4). In the presence of aggrecan, only TS4-1 showed a substantial (20-fold) increase in TIMP-3 affinity. The data indicate that the His tag present in all the TS4 forms except TS4-2* had no effect on TIMP-3 or aggrecan interactions.
The Ionic Environment Modulates TIMP-3 Potency-To investigate whether the interactions observed between ADAMTS-4, aggrecan, chondroitin sulfate, and TIMP-3 could be electrostatic in nature, the experiments described above were repeated using assay buffer lacking NaCl. At low ionic strength, neither hydrolysis of the FAM-TAMRA peptide substrate nor the potency of CGS27023 inhibition was significantly

TABLE 3 Effects of aggrecan on TIMP-3 association and dissociation rates
TS4-1 activity was measured by in vitro FRET peptide substrate assay. Mean rates of hydrolysis over 2.5 h were used in the determination of k on and k off values. Data are means (n ϭ 3). The error is the standard error of the linear fit.

Aggrecan Regulation of ADAMTS-4/TIMP Interactions
changed (data not shown). In contrast, studies using TIMP-3 showed an 8-fold increase in potency. With the further addition of 10 g/ml aggrecan, an overall 30-fold increase in TIMP-3 potency was observed (Table 4). TS4-2 interaction with TIMP-3 was dramatically enhanced under low ionic strength conditions: ϳ10-fold in the absence of aggrecan and a further 10-fold in its presence. On the other hand, TIMP-3 interactions with ADAMTS-4 forms with further C-terminal truncations (TS4-3 and TS4-4) were not modified substantially by low ionic conditions. The addition of aggrecan resulted in a 20-fold increase in TS4-3 binding and had little effect on TS4-4 binding.

DISCUSSION
The C-terminal extracatalytic domains of ADAMTS-4 are required for both binding to the extracellular matrix and substrate recognition, and in each case, this is mediated through GAG-binding interactions. The aggrecanase activity of ADAMTS-4 has been shown to be dependent on the GAGbinding interactions of its TSP-1 motif, with activity also enhanced by GAG binding to its cysteine-rich domain. In contrast, interactions between the GAG side chains of aggrecan and the ADAMTS-4 spacer domain serve to suppress cleavage of aggrecan at Glu 373 -Ala 374 , the interglobular domain cleavage site (one of five cleavage sites contained within the aggrecan protein core) associated with aggrecanase activity in vivo and thought to be the most detrimental for loss of cartilage function due to the release the entire highly negatively charged C-terminal region of the aggrecan molecule (3,7,9,10). Autocatalytic cleavage of ADAMTS-4 resulting in loss of the spacer domain has been reported in vitro, suggesting that such interactions may serve a regulatory function in vivo (10). In this study, we have further investigated the biological roles of the ADAMTS-4 C-terminal domains. We have identified a novel function for the interactions between aggrecan and the C-terminal region of ADAMTS-4 that modulate TIMP-3 inhibition of ADAMTS-4.
Inhibition of ADAMTS-4 cleavage of aggrecan by TIMP proteins has been described previously (12,13). However, by utilizing aggrecan as a substrate for the measurement of ADAMTS-4 activity within such studies, it was difficult to identify its regulatory role in TIMP inhibition. In this study, we adopted an alternative approach of measuring the impact of aggrecan on hydrolysis of a second fluorescent substrate, allowing the binding interactions of ADAMTS-4 with aggrecan to be studied in isolation. The FRET peptide substrate that we have identified is extremely sensitive to ADAMTS-4 proteolysis, with a K m of ϳ10 M, and is suitable for detailed kinetic studies. We have shown that the assay reports dissociation constants or IC 50 values for TIMP-3 and low molecular mass peptide inhibitors similar to those obtained with either aggrecan or other protein substrates.
We observed that TIMP-3 inhibition of ADAMTS-4 was enhanced typically by 20-fold in the presence of 10 g/ml aggrecan. This effect was reversed (albeit not completely) either by deglycosylation of aggrecan or by the addition of a 10-fold excess of purified chondroitin 4-sulfate. At the concentration tested, chondroitin sulfate should saturate out any GAG-binding interactions between aggrecan and ADAMTS-4 or TIMP-3, although chondroitin 4-sulfate alone demonstrated only a weak effect on TIMP-3. In contrast, chondroitin 6-sulfate (the major GAG component of human aggrecan) (24) was able to enhance TIMP-3 potency to a level much more similar to that of aggrecan when added at equivalent concentrations of GAGs. Interestingly, keratan sulfate, which has been shown to potentiate aggrecanase activity and which is the predominant GAG found near the Glu 373 -Ala 374 cleavage site (25), had no effect at all on TIMP-3 interactions. These results therefore suggest that the enhanced potency observed with aggrecan is mediated primarily through GAG-binding interactions, potentially through the

TABLE 4 Effects of ADAMTS-4 C-terminal domains on the interaction between ADAMTS-4, TIMP-3, and aggrecan
The mean rates of hydrolysis over 4 h were used in the determination of K i (app)  chondroitin 6-sulfate side chains of aggrecan, but may also involve a contribution from the aggrecan core protein.
The differences between the effects of chondroitin 6-sulfate and keratan sulfate may also suggest a possible partial explanation as to why the interglobular domain is so susceptible to proteolytic cleavage by aggrecanases. However, it should also be noted that, in this study, we investigated only a single form of keratan sulfate, which has been reported previously to undergo both age-related changes in chain length, sulfation, fucosylation, and sialic acid capping (26) and to vary in structure between the keratan sulfate-rich region and interglobular domain of aggrecan (25). Therefore, it remains a possibility that alternative forms of keratan sulfate may demonstrate properties different from those observed in this study.
Investigation of the mechanisms underlying the interactions observed demonstrated that aggrecan behaved as a purely competitive inhibitor of ADAMTS-4, whereas chondroitin sulfate had no effect. Little or no effect on the rate of dissociation of TIMP-3 from ADAMTS-4 was seen; however, a large increase in the TIMP-3 association rate constant was observed in the presence of aggrecan. The lack of change in apparent affinity of the low molecular mass inhibitor CGS27023 in the presence of aggrecan suggests that this effect is not mediated through a conformational change in the catalytic cleft, but results in enhanced TIMP-3 potency possibly by facilitating its binding to ADAMTS-4. Furthermore, the resulting enzyme-inhibitor complex is not affected by aggrecan because the first-order dissociation constant remains unchanged, thus ruling out any significant mechanism preventing the release of TIMP-3 from ADAMTS-4, once bound.
GAG/protein-binding interactions are electrostatic in nature (27), and we were able to demonstrate that the potency of TIMP-3 was enhanced by reducing the ionic strength of our assay buffer. GAG-binding sites on TIMP-3 have been postulated previously since it was shown to bind heparan sulfate and chondroitin sulfate (28). Similarly, the cysteine-rich, TSP-1, and spacer domains of ADAMTS-4 also contain GAG-binding motifs, which have been shown to interact with the GAG chains of aggrecan (5,10). In contrast to full-length ADAMTS-4 (TS4-1), studies of a selection of truncated forms of ADAMTS-4 (TS4-2*, -3, and -4) demonstrated that, at physiological ionic strength, TIMP-3 inhibition of ADAMTS-4 constructs lacking the C-terminal spacer domain was largely unaffected by the presence of aggrecan. The aggrecan-associated enhancement of TIMP-3 potency was restored in constructs containing an intact TSP-1 domain (TS4-2* and TS4-3) when NaCl was omitted from assay buffers; however, TIMP-3 inhibition of TS4-4 remained constant irrespective of NaCl or aggrecan concentrations. Significantly, the equivalent responses of the His-tagged TS4-3 and tag-free TS4-2* truncates to the addition of aggrecan under these conditions serve to rule out the possibility that the presence of a His tag is responsible for these responses. Taken together, these results suggest that the observed aggrecan-dependent enhancement of TIMP-3 potency is mediated through interactions at the TSP-1-like motif of ADAMTS-4, but that, under normal physiological conditions, additional binding interactions between the GAG side chains of aggrecan and the spacer domain of ADAMTS-4 are also necessary to facilitate binding at the TSP-1-like motif.
From the results of this study, we conclude that TIMP-3 interaction with ADAMTS-4 is mediated through the binding of GAGs (specifically chondroitin 6-sulfate) of aggrecan to the TSP-1 motif and spacer domains of ADAMTS-4 to form a complex with an improved binding affinity for TIMP-3 over free ADAMTS-4 either through the allosteric modulation of a TIMP-binding exosite of ADAMTS-4 or through the formation of a TIMP-enzyme-chondroitin 6-sulfate complex with chondroitin 6-sulfate serving as a scaffold to facilitate TIMP-3 presentation at the catalytic cleft. It has been shown that the accelerating effect of heparin on the binding of antithrombin to factor Xa is at least partly due to heparin promoting the ordered assembly of antithrombin and factor Xa in an intermediate ternary complex (29). Interestingly, although a previous study has described potential GAG-binding interactions between TIMP-3 and GAGs within the extracellular matrix (28), our own study investigating TIMP-3-binding interactions with aggrecan found no evidence for direct binding interactions, 4 suggesting that allosteric modulation may present a more probable model.
Within physiological contexts, the unique environment created by the concentration of sulfated sugars associated with different proteoglycans such as aggrecan could act as a significant regulator of both enzyme and inhibitor activity. The efficacy of TIMP-3 interaction with ADAMTS-4 in cartilage could be modulated by changes in the ratio of chondroitin 4-sulfate to chondroitin 6-sulfate (24). In other tissues with less GAG content, TIMP-3 would be a considerably less efficient inhibitor of ADAMTS-4. Knowledge of the precise binding sites of both chondroitin 6-sulfate and TIMP-3 on the exodomains of ADAMTS-4 will provide useful information for the design of highly specific inhibitors for this enzyme.