Rearranging Exosites in Noncatalytic Domains Can Redirect the Substrate Specificity of ADAMTS Proteases*

Background: ADAMTS metalloproteases are multidomain proteins with remarkable substrate specificity. Results: Swapping noncatalytic domains between ADAMTS13 and ADAMTS5 causes reciprocal changes in the cleavage of their natural substrates. Conclusion: ADAMTS exosites in noncatalytic domains are portable modifiers of proteolytic activity. Significance: Shuffling and recombination of ADAMTS ancillary structural domains may be exploited to evolve or engineer new protease functions. ADAMTS proteases typically employ some combination of ancillary C-terminal disintegrin-like, thrombospondin-1, cysteine-rich, and spacer domains to bind substrates and facilitate proteolysis by an N-terminal metalloprotease domain. We constructed chimeric proteases and substrates to examine the role of C-terminal domains of ADAMTS13 and ADAMTS5 in the recognition of their physiological cleavage sites in von Willebrand factor (VWF) and aggrecan, respectively. ADAMTS5 cleaves Glu373–Ala374 and Glu1480–Gly1481 bonds in bovine aggrecan but does not cleave VWF. Conversely, ADAMTS13 cleaves the Tyr1605–Met1606 bond of VWF, which is exposed by fluid shear stress but cannot cleave aggrecan. Replacing the thrombospondin-1/cysteine-rich/spacer domains of ADAMTS5 with those of ADAMTS13 conferred the ability to cleave the Glu1615–Ile1616 bond of VWF domain A2 in peptide substrates or VWF multimers that had been sheared; native (unsheared) VWF multimers were resistant. Thus, by recombining exosites, we engineered ADAMTS5 to cleave a new bond in VWF, preserving physiological regulation by fluid shear stress. The results demonstrate that noncatalytic thrombospondin-1/cysteine-rich/spacer domains are principal modifiers of substrate recognition and cleavage by both ADAMTS5 and ADAMTS13. Noncatalytic domains may perform similar functions in other ADAMTS family members.

The ADAMTS (a disintegrin-like and metalloprotease domain, with thrombospondin type-1 motif) superfamily contains 19 metalloproteases with a modular structure that includes a reprolysin-like metalloprotease domain (M), 3 a dis-integrin-like domain (D), a thrombospondin type 1 repeat (T), a Cys-rich domain (C), and a spacer domain (S), and a variable number of additional thrombospondin type 1 repeat and other domains. The ADAMTS superfamily also contains seven ADAMTS-like (ADAMTSL) proteins that lack M and D domains.
ADAMTS proteases, sometimes with assistance from ADAMTSL proteins, participate in many biological processes including procollagen processing, hemostasis, and extracellular matrix proteolysis relating to morphogenesis, angiogenesis, cancer, and osteoarthritis (1). For example, ADAMTS4 and ADAMTS5 degrade the cartilage proteoglycan aggrecan, which contributes to the development of arthritis. ADAMTS5 appears to be the major aggrecanase because it has higher aggrecanolytic activity than ADAMTS4, and genetic deletion of the ADAMTS5 catalytic domain protects mice from cartilage erosion in experimental models of osteoarthritis (2)(3)(4).
ADAMTS13 cleaves von Willebrand factor (VWF), which is required for normal platelet adhesion at sites of vascular injury. Interestingly, the susceptible peptide bond is buried in the native VWF A2 domain but is exposed when VWF is stretched as occurs in vivo within platelet-rich thrombi in flowing blood. Shear stress-induced VWF cleavage is an essential feedback inhibitory mechanism: congenital or acquired ADAMTS13 deficiency causes thrombotic thrombocytopenic purpura, which is characterized by life-threatening microvascular thrombosis (5)(6)(7).
Different ADAMTS proteases recognize very distinct substrates but employ similar mechanisms to establish strict substrate specificity: the metalloprotease domain determines cleavage site specificity, and C-terminal ancillary domains provide additional binding precision or localization (1). For example, ADAMTS4 and ADAMTS5 bind to the glycosaminoglycan chains of aggrecan and other extracellular matrix proteoglycans through the S domain (8) and C domain (9), respectively, and these interactions can profoundly affect substrate recognition. ADAMTS4 and ADAMTS5 both cleave aggrecan at several sites, including the Glu 373 -Ala 374 bond in the interglobular domain (IGD) and the Glu 1480 -Gly 1481 bond in the chondroitin sulfate-2 (CS-2) domain (bovine aggrecan numbering), and deletion of the ADAMTS4 S domain (10) or the ADAMTS5 CS domains (9) markedly impairs cleavage at these sites. For ADAMTS13, optimal VWF cleavage depends on contacts between successive segments of the VWF A2 domain and corresponding binding sites in the proximal T, C, S, and distal thrombospondin type 1 repeat domains of ADAMTS13 (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23).
The modular structure of ADAMTS active sites and exosites suggests that substrate specificity could be investigated and intentionally modified by reassorting metalloprotease and noncatalytic domains from different family members. The altered properties of chimeric proteases constructed from ADAMTS4 and ADAMTS5 support this concept (10). We have now created chimeric ADAMTS5 and ADAMTS13 proteases and have characterized their activity toward model aggrecan IGD and VWF substrates. The results confirm the modularity and portability of ADAMTS exosites and show definitively that ADAMTS13 TCS domains specify shear stress-dependent cleavage of VWF by conferring that activity onto ADAMTS5 MD domains.
ADAMTS protein concentrations were determined as described (11) by SDS-PAGE with V5-tagged Positope reference protein standards (Invitrogen), Western blotting on PVDF membranes with anti-V5 antibody and peroxidase-conjugated goat anti-mouse IgG (1:10,000 dilution, A3673; Sigma), and chemifluorescence detection (ECL Plus; GE Healthcare). Signals were collected and analyzed with a fluorescence imaging system (Typhoon Trio; GE Healthcare) and ImageQuant TL software (GE Healthcare). The concentration of ADAMTS proteases in conditioned medium typically was 12-36 nM (1-3 g/ml). The media were concentrated 5-10-fold by ultrafiltration on YM30 membranes (Millipore, Inc.) and dialyzed into an appropriate reaction buffer. Proteases were aliquoted and stored at Ϫ30°C until used without further purification. When assayed with substrate FRETS-VWF73 (26), the activity of 0.6 nM MDTCS13 was equivalent to that of 0.6 nM ADAMTS13 in standard pooled normal plasma.
To assess catalytic efficiency (k cat /K m ), reactions containing 3-7 nM substrate, 0.5 or 1 nM protease, and 0.05% Tween 20 were incubated at 37°C for different times (t) and stopped with an equal volume of 50 mM EDTA, and products were analyzed by sandwich ELISA as described (16). Briefly, reactions were adsorbed onto microtiter plates coated with anti-GST antibodies (Pierce), and uncleaved substrates were detected with peroxidase-conjugated anti-His antibodies (Invitrogen). MDTCS13 cleaves GST-gst-VWF73 with K m Ն800 nM (14), and bovine aggrecanase (principally ADAMTS5) cleaves a 40-residue bovine IGD peptide substrate with K m ϳ480 M (27). Therefore, in all reactions the enzyme concentration (E) and initial substrate concentration (S 0 ) are much lower than the K m , and the time courses of product (P) generation were analyzed to obtain k cat /K m (corresponding to the initial rate) by fitting to the following integrated Michaelis-Menten equation.
Cleavage Site Characterization-GST-peptide substrates (5 g) were incubated with ADAMTS protease (10 nM) in reaction buffer at 37°C. After 0.5 h, substrate cleavage was confirmed by SDS-PAGE and Western blotting using peroxidase-conjugated anti-His tag antibody (Tetra-His; Qiagen). After incubation for 5 h, the products were separated by SDS-PAGE on a 10 -20% gradient gel, transferred onto a PVDF membrane (0.2 m; Invitrogen), and stained with SimplyBlue SafeStain (Invitrogen). C-terminal product bands (8.8 -11 kDa) were excised from the membrane and sequenced by automated Edman degradation (W. M. Keck Foundation Biotechnology Resource Laboratory, Yale University).

ADAMTS Proteases and Extracellular Matrix
Binding-Chimeric ADAMTS proteases were designed to assess the ability of noncatalytic domains to determine the cellular localization and substrate specificity of the metalloprotease domain. For ADAMTS13 (Fig. 1A), MDTCS13 was selected as the reference construct because it cleaves model peptide substrates efficiently and is sufficient for shear-dependent cleavage of VWF multimers (11,13,30). In addition, the role of ADAMTS13 D, T, C, and S domains in substrate recognition is well supported (11-17, 19, 20, 22, 23). ADAMTS5 (Fig. 1A) was chosen for domain exchanges because it is at least 20-fold more active than ADAMTS4 toward aggrecan, and ADAMTS5 noncatalytic domains confer increased activity when transferred to ADAMTS4 (10). A truncated ADAMTS5 consisting of only the M domain has little or no proteolytic activity, and sequential addition of DTCS domains progressively increases activity toward many substrates (9). The ADAMTS5 distal thrombospondin type 1 repeat is dispensable for efficient substrate recognition (9), and MDTCS5 was selected as the reference ADAMTS5 construct.
For both ADAMTS13 and ADAMTS5, the smallest known fragment with detectable proteolytic activity consists of MD domains (9,14), and the crystal structure of ADAMTS5 MD domains indicates that M and D domains are structurally integrated (31). Therefore, the more distal TCS domains were exchanged in chimeric constructs MD13/TCS5 and MD5/ TCS13. Additional constructs MDT13/CS5 and MDT5/CS13 were prepared by exchanging the distal CS domains.
Recombinant ADAMTS variants were expressed in stably transfected T-Rex 293 cell lines and analyzed by SDS-PAGE and Western blotting in cell lysates and conditioned medium (Fig. 1B). ADAMTS5 CS domains bind proteoglycans and cause the retention of ADAMTS5 in extracellular matrix unless heparin is included in the cell culture medium (9). The three constructs with distal ADAMT5 domains, MDT13/CS5, MD13/ TCS5 and MDTCS5, were released into conditioned medium only in the presence of heparin (Fig. 1B), demonstrating that CS5 domains are sufficient to specify matrix association.
Two constructs with distal ADAMTS13 domains, MD5/ TCS13 and MDTCS13, were secreted without added heparin. MDT5/CS13 was retained intracellularly, suggesting that it is misfolded. Another MDT5/CS13 construct with a different junction between the T and C domains (ADAMTS5 Met 1 -Pro 622 and ADAMTS13 Lys 440 -Ala 685 ) also was not secreted (data not shown).
ADAMTS5 has a 243-amino acid residue propeptide that is processed by furin, and cleavage of the propeptide is required for ADAMTS5 activity (32). Evidence of this proteolytic processing is apparent for MDTCS5 and MD5/TCS13 (Fig. 1B), which were detected in cell lysates mainly (MDTCS5) or entirely (MD5/TCS13) as larger unprocessed ϳ100-kDa species but were processed completely to ϳ82 kDa species in conditioned medium. ADAMTS13 has a much smaller 41-residue propeptide, and its precursor and processed forms are difficult to resolve by SDS-PAGE (33); size differences were not observed between intracellular and secreted MDT13/CS5, MD13/TCS5, or MDTCS13 (Fig. 1B).
A faint 54-kDa species was detected with anti-V5 antibody in some recombinant proteases that may be a C-terminal proteolytic fragment (Fig. 1B). If so, the size of the fragment is consistent with cleavage in the metalloprotease domain.
ADAMTS Substrates-Chimeric GST-peptide substrates based on the sequences of human VWF and aggrecan (Fig. 1C) were expressed in E. coli and purified to homogeneity (Fig. 1D). Substrate gst-VWF73 corresponds to the 73-residue Asp 1596 -Arg 1668 segment of VWF, which is the smallest VWF fragment that is cleaved efficiently by ADAMTS13 (12). ADAMTS13 D, T, C, and S domains interact with successive segments of this substrate between Asp 1614 and Arg 1668 , and these cooperative interactions accelerate the cleavage of the Tyr 1605 -Met 1606 bond by several orders of magnitude (16,17,19,20).
Substrate gst-IGD contains the Thr 331 -Gly 458 IGD segment of human aggrecan. A similar GST-peptide substrate containing aggrecan Tyr 330 -Gly 457 and a C-terminal FLAG tag (34) is cleaved rapidly at the Glu 373 -Ala 374 bond by ADAMTS5 (10). The gst-VWF/IGD and gst-IGD/VWF substrates exchange N-terminal cleavage sites and C-terminal potential exosite binding sites. gst-VWF/IGD consists of VWF Asp 1596 -Ile 1616 followed by aggrecan Glu 394 -Gly 458 and therefore retains the Tyr 1605 -Met 1606 bond cleaved by ADAMTS13, as well as a Glu 1615 -Ile 1616 bond that might be cleaved by ADAMTS5. Sub-strate gst-IGD/VWF is the reciprocal construct, with aggrecan Thr 331 -Leu 393 followed by VWF Lys 1617 -Arg 1668 . The retained aggrecan sequence includes the entire Gln 354 -Leu 393 peptide previously reported to be cleaved efficiently by bovine aggrecanase (27).
Substrate Specificity of ADAMTS Variants-Recombinant proteases in conditioned medium were concentrated by ultrafiltration, dialyzed against reaction buffer, and used without further purification. The activity of ADAMTS proteases toward chimeric substrates was assessed by a semiquantitative Western blot assay method. As expected (16), proteases containing the MD domains of ADAMTS13, MDTCS13, MDT13/CS5, and MD13/TCS5, cleaved gst-VWF73 and gst-VWF/IGD to yield a 28-kDa N-terminal product detected with anti-GST antibody (Fig. 2A). These enzymes cleaved gst-VWF/IGD less completely than gst-VWF73, which is consistent with direct interactions between the Lys 1617 -Arg 1668 segment of VWF and the TCS domains of ADAMTS13 (16). The aggrecan IGD has no obvious potential cleavage sites for ADAMTS13, and MDTCS13, MDT13/CS5, and MD13/TCS5 did not cleave gst-IGD/VWF or gst-IGD to a detectable extent ( Fig. 2A), even after increasing the time of digestion from 2 to 24 h or increasing the enzyme concentration 10-fold (Fig. 2B).
Proteases containing the MD domains of ADAMTS5, MDTCS5 and MD5/TCS13, cleaved gst-IGD and gst-IGD/ VWF to generate a 32-kDa product ( Fig. 2A). To establish the site of cleavage, MD5/TCS13 was incubated with gst-IGD and gst-IGD/VWF, and C-terminal cleavage products were isolated and sequenced. The N-terminal sequence obtained was ARGS-VILTVKP in each case, indicating that MD5/TCS13 cleaves both substrates at the Glu 373 -Ala 374 bond (Fig. 3).
MDTCS5 could not cleave substrates gst-VWF73 and gst-VWF/IGD, but MD5/TCS13 was able to cleave them (Fig. 2), and the mobility of the N-terminal product on SDS-PAGE was similar to that generated by MDTCS13. Automated Edman degradation of the C-terminal product gave the sequence IKRLPGDI (Fig. 3B), indicating that MD5/TCS13 cleaves the Glu 1615 -Ile 1616 bond of gst-VWF73 or gst-VWF/IGD, 10 residues C-terminal of the Tyr 1605 -Met 1606 bond cleaved by MDTCS13 (16). Prolonged incubation for 24 h or with 10-fold more MD5/TCS13 resulted in complete cleavage of gst-VWF73 but partial cleavage of gst-VWF/IGD (Fig. 2B). Thus, TCS13 domains confer on MD5 the ability to cleave substrates based on the VWF A2 domain, and the chimeric MD5/TCS13 protease prefers gst-VWF73, which contains sequences that interact with TCS13 domains (12,14,16,17).
Catalytic Efficiency of Substrate Cleavage by ADAMTS Variants-The time course of substrate cleavage (Fig. 4) was analyzed under conditions such that E Ͻ Ͻ K m and S 0 Ͻ Ͻ K m . Fitting to the integrated Michaelis-Menten equation yields values for k cat /K m , a measure of catalytic efficiency (Table 1).
Substrate gst-VWF/IGD lacks the exosite-binding sequences of gst-VWF73 that are important for VWF recognition (12,16,17,19,20), and MDTCS13 cleaved gst-VWF73 ϳ5-fold more rapidly than gst-VWF/IGD as expected (Table 1). MDT13/CS5 cleaved both substrates more slowly than MDTCS13 but also preferred gst-VWF73 over gst-VWF/IGD, which is consistent with a role for the ADAMTS13 T domain in binding the Gln 1624 -Val 1630 segment that is present in gst-VWF73 but missing from gst-VWF/IGD (16). MD13/TCS5 was less active than MDT13/CS5 and also discriminated less effectively between gst-VWF73 and gst-VWF/IGD; the difference in cleavage rate was ϳ4-fold for MD13/TCS5 compared with   (Table 1). Thus, replacement of the ADAMTS13 T domain made substrate recognition less dependent on substrate sequences after VWF Ile 1616 .

10-fold for MDT13/CS5
Replacing the distal domains of MDTCS5 gave complementary results but with interesting differences. MDTCS5 cleaved substrate gst-IGD with efficiency similar to that with which MDTCS13 cleaved gst-VWF73. MDTCS5 also cleaved gst-IGD and gst-IGD/VWF with similar efficiency, indicating that the C-terminal 65 residues of the IGD domain, Glu 394 -Gly 458 , contribute little to substrate recognition by distal domains of MDTCS5. MD5/TCS13 cleaved gst-IGD/VWF ϳ2.2-fold more rapidly than gst-IGD (Table 1), indicating that TCS13 domains confer a modest preference for substrates containing their natural binding partners in the Lys 1617 -Arg 1668 segment of VWF. More strikingly, MD5/TCS13 acquired the ability to cleave gst-VWF73 with surprising efficiency. MDTCS5 was inactive toward gst-VWF73, but MD5/TCS13 cleaved gst-VWF73 at nearly one-third the rate of MDTCS13 (Table 1). In addition, MD5/TCS13 cleaved gst-VWF73 ϳ10-fold faster than gst-VWF/IGD, indicating that efficient gst-VWF73 recognition was mediated by interactions between distal segments of gst-VWF73 and exosites in TCS13 domains.
Aggrecanase Activity of Recombinant ADAMTS Enzymes-The glycosaminoglycan chains of aggrecan bind ADAMTS5 and influence cleavage by this protease. However, gst-IGD and gst-IGD/VWF substrates (Fig. 1) are not glycosylated and cannot be used to evaluate this relationship. To assess the contribution of glycosylation to substrate recognition, ADAMTS variants were incubated with bovine aggrecan, and specific neoepitopes produced by cleavage were detected by Western blotting (Fig. 5).
Shear-dependent Cleavage of VWF-MD5/TCS13 acquired the ability to cleave the Glu 1615 -Ile 1616 bond of gst-VWF73, and therefore multimeric VWF was also evaluated as a substrate (Fig. 6). Plasma VWF and ADAMTS proteases were incubated under static conditions or sheared by vortexing for 200 s. The reactions were analyzed by SDS-PAGE and Western blotting to detect VWF cleavage products (21,29). As a negative control, the reactions were supplemented with EDTA before vortexing. As reported previously (21,29), plasma VWF was resistant to cleavage by MDTCS13 under static conditions but was cleaved rapidly when the reaction was subjected to fluid shear stress. VWF was resistant to cleavage by MDTCS5, with or without shear stress. However, the results for MD5/TCS13 were similar to those for MDTCS13. VWF was not cleaved by MD5/TCS13 under static conditions but was cleaved under fluid shear stress. Thus, recombining exosites enabled MD5/ TCS13 to perform shear-dependent cleavage of VWF multimers.

DISCUSSION
The noncatalytic domains that characterize the ADAMTS protease family mediate several distinct functions in different family members, including substrate recognition, tissue localization, and angiogenesis (1). We have used a domain substitution approach to investigate the role of noncatalytic domains in ADAMTS5 and ADAMTS13. In addition we prepared chimeric model substrates in which cleavage sites and ancillary binding sites were exchanged between VWF domain A2 and aggrecan IGD. The results show that the catalytic center and noncatalytic domains can cooperate to determine substrate specificity, but the importance of each interaction varies considerably for different enzyme-substrate combinations.
MDTCS5 cleaved gst-IGD efficiently as expected but also cleaved gst-IGD/VWF at essentially the same rate. Furthermore, MD5/TCS13 cleaved gst-IGD rapidly despite the absence of distal ADAMTS5 domains. These results suggest that the C-terminal segment of the nonglycyosylated IGD polypeptide does not interact functionally with TCS5 exosites. In addition, MDTCS5 and MD5/TCS13 both cleaved the Glu 373 -Ala 374 bond of native bovine aggrecan with similar efficiency, indicating that TCS5 exosites that interact with glycosaminoglycans (9) may not contribute to cleavage of the aggrecan IGD.
Although MD5/TCS13 cleaved the aggrecan IGD efficiently, it could not cleave the Glu 1480 -Gly 1481 bond in the aggrecan CS-2 domain. However, studies of ADAMTS5/ADAMTS4 chimeric proteases showed that MD5/TCS4 cleaved both IGD and CS-2 sites normally (10). Thus, when combined with MD5, TCS4 can functionally replace TCS5 for both IGD and CS2 cleavage, whereas MD5/TCS13 only preserves IGD cleavage.  Table 1). The data points represent the mean values for at least two independent experiments.
These results are consistent with a model proposed by Fushimi et al. (10) in which cleavage of the Glu 373 -Ala 374 site depends mainly on interactions between protease MD domains and the IGD polypeptide, whereas cleavage of the Glu 1480 -Gly 1481 bond requires interactions between the TCS domains of ADAMTS4 or ADAMTS5 and CS-2 glycosaminoglycan chains.
The ADAMTS13 construct MDTCS13 predictably cleaved gst-VWF73 ϳ5-fold more rapidly than gst-VWF/IGD, which lacks the exosite binding sequences that promote the efficient recognition of gst-VWF73 (12,14,16,17). Unexpectedly, MD5/ TCS13 also cleaved gst-VWF73 very efficiently, ϳ10-fold more rapidly than gst-VWF/IGD, although MDTCS5 was completely inactive toward these substrates. Therefore, appending TCS13 dramatically altered the specificity of MD5, enabling it to cleave an otherwise resistant substrate, and cleavage depended on TCS13 exosite interactions with gst-VWF73 that cannot occur for gst-VWF/IGD. These exosite-substrate interactions are consistent with loss of function phenotypes induced by engineered deletion and missense mutations in ADAMTS13 and VWF (11-17, 19, 22, 23). The gain of function produced by transferring functional ADAMTS13 exosites onto ADAMTS5, in construct MD5/TCS13, shows definitively that TCS13 domains are critical for recognizing and cleaving VWF because they bind to a C-terminal segment of the A2 domain.
Although the ADAMTS5 metalloprotease domain could be induced to cleave VWF in construct MD5/TCS13, none of the tested proteases that contained an ADAMTS13 active site was able to cleave aggrecan IGD sequences or native aggrecan core protein, at least at the sites recognized by aggrecanase neoepitope antibodies. This result probably reflects the extraordinarily strict specificity of ADAMTS13 for the sequence Leu-Xaa-Tyr-Met (35,36), which occurs once in VWF and not at all in aggrecan.
The physiological cleavage of VWF by ADAMTS13 is regulated by fluid shear stress, which unfolds the A2 domain and exposes the buried scissile bond. MDTCS5 did not cleave plasma VWF detectably, but chimeric MD5/TCS13 acquired the ability to cleave VWF in a shear-dependent manner, most likely at the Glu 1615 -Ile 1616 bond that is cleaved in gst-VWF73. In the A2 domain crystal structure, Glu 1615 -Ile 1616 is located on the surface in an extended segment referred to as the ␣4-less loop (37). The side chain of Glu 1615 projects into solvent, but Ile 1616 is buried. Exposure of this bond by shear-induced unfolding of VWF was not sufficient to render it sensitive to MDTCS5. However, exosites conferred by TCS13 domains  (Fig. 4). Substrates not cleaved as determined by Western blotting are indicated by a minus sign (Ϫ). Cleavage of bovine aggrecan at Glu 373 -Ala 374 and Glu 1480 -Gly 1481 bonds was analyzed by Western blotting with site-specific antibodies (Fig. 5); a plus sign (ϩ) indicates cleavage, and a minus sign (Ϫ) indicates no cleavage.   enabled MD5/TCS13 to recognize and cleave VWF with remarkable efficiency and with regulation by fluid shear stress. The Glu 1615 -Ile 1616 selected by MD5/TCS13 is one of six Glu-Xaa bonds in gst-VWF73, and the predilection for this bond reflects the specificity of the ADAMTS5 metalloprotease domain, as well as geometric constraints on positioning a substrate between the active site and exosites. ADAMTS5 prefers certain Glu-Gly, Glu-Leu, and Glu-Ala bonds in aggrecan (4), but little is known about its specificity for extended sequences adjacent to these sites. It seems likely that modifying the scissile bond environment to better match the specificity of ADAMTS5 would further enhance the shear-dependent cleavage of VWF by MD5/TCS13. The evolution of ADAMTS proteases has involved many instances of domain duplication and shuffling (38). These proteins have diverged considerably-ADAMTS5 and ADAMTS13 are just 29% identical over the domains they share-and the reshuffling of domains in vitro is a powerful strategy to investigate structure-function relationships. Our studies and others (10,39,40) indicate that ADAMTS noncatalytic domains often are portable, easily transferred between family members with retention of structural integrity. In addition, their exosite functions tend to be modular, corresponding to one or a few domains. Finally, the lack of stringent distance constraints allows new combinations of exosites and active sites to manifest new proteolytic activities, which could have practical applications. For example, MD5/TCS13 might be able to restore regulated proteolysis of VWF in patients with thrombotic thrombocytopenic purpura caused by autoantibodies against the MD domains of ADAMTS13.
Noncatalytic domains strongly determine the specificity of ADAMTS5 and ADAMTS13, supporting the principle that ADAMTS proteases require their noncatalytic domains to recognize substrates and localize their proteolytic activity (1). A corollary might be that targeting of noncatalytic domains should inactivate ADAMTS proteases in vivo. In a way, this concept is validated by the observation that autoantibodies against the ADAMTS13 spacer domain can inhibit protease activity enough to cause thrombotic thrombocytopenic purpura (5,6). This proof of principle experiment of nature suggests that intentionally targeting the exosites of other ADAMTS proteases may be therapeutically useful.