Crystal structures of the noncatalytic domains of ADAMTS13 reveal multiple discontinuous exosites for von Willebrand factor

Abstract

ADAMTS13 specifically cleaves plasma von Willebrand factor (VWF) and thereby controls VWF-mediated platelet thrombus formation. Severe deficiencies in ADAMTS13 can cause life-threatening thrombotic thrombocytopenic purpura. Here, we determined 2 crystal structures of ADAMTS13-DTCS (residues 287–685), an exosite-containing human ADAMTS13 fragment, at 2.6-Å and 2.8-Å resolution. The structures revealed folding similarities between the disintegrin-like (D) domain and the N-terminal portion of the cysteine-rich domain (designated the C_A domain). The spacer (S) domain forms a globular functional unit with a 10-stranded β-sandwich fold that has multiple interaction sites with the C_A domain. We expressed 25 structure-based mutants of ADAMTS13-MDTCS (residues 75–685) and measured their enzymatic activity. We identified 3 VWF-binding exosites on the linearly aligned discontinuous surfaces of the D, C_A, and S domains traversing the W-shaped molecule. Since the MDTCS domains are conserved among ADAMTS family proteins, the structural framework of the multiple enzyme-substrate interactions identified in the ADAMTS13-VWF system provides the basis for a common substrate recognition mode in this class of proteinases.

The human ADAMTS (a disintegrin-like and metalloproteinase with thrombospondin type-1 motif) family is composed of 19 genes that encode extracellular multidomain enzymes containing a reprolysin-type metalloproteinase domain and several conserved domains following the metalloproteinase domain (1). In contrast to the phylogenetically related ADAM (a disintegrin and metalloproteinase) family proteins, most of which have a transmembrane and a cytoplasmic domain in the C-terminal region (2), ADAMTSs are secretary proteinases that lack these domains and instead have at least 1 thrombospondin-1 (TSP-1) type-1 repeat (TSR). ADAMTSs have diverse functions including procollagen processing, aggrecan degradation, and organogenesis (1). ADAMTS13 controls platelet thrombus formation through cleavage of the von Willebrand factor (VWF).

VWF is a plasma glycoprotein that plays an essential role in platelet-dependent hemostasis (3, 4). VWF mediates platelet adherence to damaged blood vessels through interactions with glycoprotein Ib on the platelet surface and collagen in the subendothelium and contributes to platelet aggregation through interactions with integrin α_IIbβ₃. VWF, synthesized mainly in vascular endothelial cells, contains 2,050 aa residues and is released into the plasma as disulfide-bonded ultralarge VWF (UL-VWF) multimers having a mass greater than 20,000 kDa. In healthy individuals, UL-VWF multimers undergo limited proteolytic processing (5). ADAMTS13 specifically cleaves the Tyr-1605-Met-1606 peptidyl bond within the A2 domain of VWF (6) in a fluid shear-stress-dependent manner (7). Because VWF multimers have an alternate head-to-head and tail-to-tail disulfide-bonded architecture between neighboring subunits, cleavage by ADAMTS13 gives rise to a series of circulating multimers with molecular masses ranging from 500 to 15,000 kDa. Control of the size distribution of VWF multimers is important for normal hemostasis, as large multimers are hemostatically more active than small multimers (3). Deficiencies in ADAMTS13 activity, caused either by genetic mutations in the ADAMTS13 gene or by acquired inhibitory autoantibodies directed against the ADAMTS13 protein, results in the accumulation of UL-VWF in the plasma (8–11). The UL-VWF accumulation leads to the formation of disseminated platelet-rich microthrombi in the microvasculature, which results in the life-threatening disease, thrombotic thrombocytopenic purpura (TTP).

The human ADAMTS13 gene encodes a precursor protein of 1,427 aa with a modular structure consisting of a signal peptide, a propeptide (P), a metalloproteinase (M) domain, a disintegrin-like (D) domain, a TSR (T1), a cysteine-rich (C) region, a spacer (S), 7 TSRs (T2–T8), and 2 CUB (complement components C1rC1s/urinary epidermal growth factor/bone morphogenetic protein-1) domains (11–13). The M domain of ADAMTS13 alone is not sufficient for recognition and specific cleavage of VWF, but full VWF-cleaving activity is achieved in vitro with an M-D-T1-C-S domain fragment (14–17). In addition, antibodies isolated from idiopathic TTP patients commonly inhibit ADAMTS13 activity by binding to the C and S domains of ADAMTS13 (14, 18, 19). Collectively, these observations indicate that the noncatalytic domains, especially the proximal C-terminal domains including the D, T1, C, and S domains (designated ADAMTS13-DTCS), are essential for recognition of VWF. The crystal structures of the MD domains (ADAMTS-MDs) of ADAMTS1 (20), ADAMTS4 (21), and ADAMTS5 (21) have been determined, but no structural information is currently available for the T1, C, and S domains of ADAMTS proteins. To gain insight into the molecular mechanism of VWF recognition by ADAMTS13, we solved the crystal structures of ADAMTS13-DTCS (residues 287–685) and performed a series of structure-based mutagenesis experiments to identify VWF-binding exosites. The present structure is the first for the TCS domains of any ADAMTS family member and will provide a template for understanding the role of these domains in substrate recognition by ADAMTS proteins.

Results

Structure Determination.

The structure of ADAMTS13-DTCS was solved using the multiple-wavelength anomalous dispersion (MAD) method at 2.9 Å using data sets obtained from a single osmium derivative crystal (Table S1). The structure was further refined against 2 native data sets, form-1 (space group C2, a = 152.7 Å, b = 52. 9 Å, c = 76.2 Å, and β = 111.4°) and form-2 (space group C2, a = 138.6 Å, b = 51.4 Å, c = 76.4 Å, and β = 106.7°) at 2.6-Å (R = 0.243; R_free = 0.289) and 2.8-Å (R = 0.229; R_free = 0.280) resolution, respectively (Table S1). Each crystal contained 1 ADAMTS13-DTCS molecule per asymmetric unit. The final model of the form-1 (form-2) crystal includes ADAMTS13 residues 299–322 (323), 331 (330)-458, and 466–682. Electron densities for carbohydrate moieties attached to 3 of the 4 potential N-linked (Asn-552, Asn-579, and Asn-614) and one O-linked (Ser-399) site were observed (SI Text and Fig. S1). Pro-379, Pro-414, Pro-475, and Pro-618, were in the cis conformation.

Overall Structure.

The N-terminal portion of the C domain (residues 440–531, designated the C_A domain here) in ADAMTS13 has a fold structurally homologous to that of the C domain of ADAMs, despite the lack of sequence similarity. The D domain (residues 306–383) of ADAMTS13 also has a fold similar to the C domain of ADAMs, which is consistent with recent crystallographic studies (20–22). Therefore, ADAMTS13 possesses 2 homologous domains that belong to the ADAM_CR family (Pfam database entry: pfam08516). The remaining C-terminal portion of the C domain (residues 532–555) is highly conserved in amino acid sequence among ADAMTS family proteins (Fig. S2, here called the C_B domain). The domain architecture of ADAMTS13 is schematically represented in Fig. 1A.

Fig. 1.

Structure of ADAMTS13-DTCS. (A) Schematic representation of the domain structures of full-length ADAMTS13 and ADAMTS13-DTCS. (B) Ribbon structure of ADAMTS13-DTCS (form-1) in stereo. Domains are colored as in A. Strands in the S domain are numbered.

The overall structure of ADAMTS13-DTCS resembles a distorted W-shape, in which 3 knobs, the D, C_A, and S domains, are connected by 2 elongated structural modules, T1 and C_B (Fig. 1B). The homologous D and C_A domains are separated and related by a pseudo-90° screw rotation with an ≈45-Å translation along T1 (Fig. 1B). T1 has a very similar structure to that of the prototypical TSR, TSR2, in TSP-1 (23) with an rmsd of 1.37 Å for the equivalent Cα atoms (Fig. S1) and an antiparallel 3-stranded fold. Although the C_B domain has no apparent secondary structure, it has a series of turns stabilized by a pair of disulfide bonds (Cys-532-Cys-548 and Cys-545-Cys-555) and forms a rod shape with its N and C termini ≈25 Å apart (Fig. 1B). The C_A and S domains, bridged by the C_B domain, make direct contact through the extended loop of the S domain (Fig. 1B and SI Text). The structures of ADAMTS13 obtained from the 2 crystal forms are essentially the same, with the exception of the relative orientations between the domains (SI Text and Fig. S3). The structural details of the D, C_A, and S domains are described in the following sections and T1 in SI Text.

Comparison of the D and C_A Domains.

The D and C_A domains have only 17% identity in their amino acid sequences (Fig. S2); however, their tertiary structures are quite similar (Fig. 2A, B, and E). They share an N-terminal α-helix, 2 pairs of double-stranded antiparallel β-sheets, and 3 disulfide bonds, constituting the core structure of these domains. The D domain has an additional disulfide bond (Cys-322-Cys-347) that is strictly conserved among the ADAM counterparts (22, 24). The 3 peripheral loops differ markedly in structure between D and C_A in ADAMTS13 (Fig. 2E). The amino acid sequences of these loops are also quite different between D and C_A in ADAMTS13 and in other ADAMTS family members (Fig. S2).

Fig. 2.

Comparison of the D and C_A domain structures. Ribbon representation of the D (A) and C_A (B) domains of ADAMTS13-DTCS, and the C domains of VAP1 (representative of canonical ADAMs, PDB 2ERO) (C) and ADAM10 (PDB 2AO7) (D). The conserved α-helix, β-strands, and disulfide bonds are shown in red, yellow, and orange, respectively. The V-loop and HVR are shown in gray and blue, respectively. Disordered regions in the crystals are shown as dotted lines. The numbers of the terminal amino acid residues are indicated. (E) Superimposition of the D (orange) and C_A (green) domains in stereo. Disulfide bonds are indicated in stick representations.

The loop following the first α-helix of the C_A domain (residues 454–469) is 12 aa residues longer than that of the D domain, protrudes from the main body of C_A, and is disordered along the distal side (Fig. 2 B and E). We designated this C_A-specific loop as the protruding (P) loop. One region in the D domain (residues 323–329) is disordered (Fig. 2A), not only in the current ADAMTS13 structures, but also in other reported ADAMTS1 structures (20), although the corresponding region of C_A is clearly defined in electron density maps. We designated this loop the variable (V) loop because of its variability in both length and amino acid sequence (Fig. S2). Canonical ADAM family members have a helix-loop insertion of 26–30 aa residues in the V-loop (Fig. 2C), whereas the atypical ADAM10 does not, and its C domain is more similar to the D and C_A domains of ADAMTS13, except for the hypervariable region (HVR) (Fig. 2D) (22, 24). ADAMTS13 has an insertion of 6 residues (residues 512–517) just before the C-terminal β-sheet of C_A (Fig. 2 B and E), which is not found in other ADAMTS members (Fig. S2). We designated this loop the ADAMTS13-C_A-unique (U) loop. Each C domain contains a HVR that differs markedly among ADAMs and may play a central role in substrate recognition (22, 24). ADAMTS13 has shorter HVRs in both the D and the C_A domains than those present in the ADAMs (Fig. 2 A–D). These loops and HVRs were targeted for mutations (see below).

There is an Arg-498-Gly-499-Asp-500 integrin recognition sequence in the C_A domain. The side chain of Arg-498 is buried and unavailable for protein–protein interactions, but the Asp-500 side chain is exposed toward the solvent.

Spacer Domain.

The S domain is a long cysteineless segment and its primary structure shows no apparent homology to known structural motifs. The present study revealed that this region folds into a single globular domain with 10 β-strands in a jelly-roll topology, forming 2 antiparallel β-sheets that lie almost parallel to each other (Fig. 1B and Fig. S4A). The hydrophobic residues forming the core of the β-sandwich (Fig. S4B), a cluster of aromatic residues located on the concave outer surface of the smaller 4-stranded sheet (Fig. S4C), and proline and glycine residues in the loops, are highly conserved among ADAMTS proteins (Fig. S4D). Collectively, these findings suggest that ADAMTS proteins share the S domain architecture observed in ADAMTS13. In contrast, loops located at the distal side of the molecule are highly variable in both length and amino acid sequence among ADAMTS family members (Fig. S4D). The N and C termini of the S domain are in close proximity and thus the T2 following the S domain should be in close proximity to the C_A/S-domain junction but not the distal side of the S domain.

MDTCS Model.

The reported crystal structures of the ADAMTS-MDs and our current ADAMTS13-DTCS structure enabled us to build an ADAMTS13-MDTCS model (Fig. 3A). The currently available ADAMTS-MD structures (20, 21) superimpose well on each other, except for subtle differences in the relative orientations of the M and D domains. We performed a functional assay using the ADAMTS13-MDTCS mutants F216E and A258C/K368C, which have modified interactions between the M and D domains. The results suggest that a stable association between the M and D domains is necessary for ADAMTS13 function (Fig. 3 B and D and SI Text).

Fig. 3.

A model of ADAMTS13-MDTCS and the residues affecting enzymatic activity. (A) Surface representation of the ADAMTS-MDTCS model. The M domain is shown in gray and the other domains are colored as in Fig. 1. The zinc ion is shown in yellow. (B) Close-up view of the interface between the M and D domains in the ADAMTS13-MDTCS model. Potential S1′ and S3′ substrate-binding sites, disordered V-loop (represented by a red dotted line), and the residues substituted by cysteine to form an interdomain disulfide bond (represented by a green dotted line) are indicated schematically. (C) The residues that affect ADAMTS13-MDTCS activity are indicated on the molecular surface, using a red-through-blue color-coding according to the results of the mutational assay shown in D. The molecule is viewed from 2 orthogonal directions. The V-loop of the D domain was disordered in the structures determined in this study and is represented by a red ellipsoid. (D) Summary of the mutational analysis, presenting the effects on secretion and enzymatic activity of the ADAMTS13-MDTCS mutants. Signs denote relative secretion level as follows: −, no detectable secretion; +, less than 30%; ++, 30≈70%; +++, 70∼100% of the secretion level of the wild type; ND, not determined. Relative enzymatic activities of the mutants are shown in the bar graph. The error bars indicate the range. (E) Close-up view of the hydrophobic cluster surrounded by arginine residues (exosite-3) in the S domain.

VWF-Binding Exosites.

We introduced mutations into ADAMTS13-MDTCS and measured the enzymatic activities of the mutants using the synthetic fluorogenic substrate FRETS-VWF73 (25). The results are summarized in Fig. 3 C and D.

In the current model, the D domain abuts the M domain catalytic site (Fig. 3B), suggesting that the surface of the D domain leading to the catalytic site functions as a VWF-binding exosite. Two mutants, 1 with a substitution in the HVR (D), R349D, and the other with 7 residues in the V-loop (D) replaced by a 4-residue linker, the ΔV-loop (D), exhibited diminished enzymatic activity (Fig. 3 C and D). The disordered V-loop (D) contains 4 charged residues, Arg-326, Glu-327, His-328, and Asp-330, which are suggested to lie in the vicinity of Arg-349 in the HVR (Fig. 3B). The cluster of these charged residues may collaboratively create an exosite (exosite-1). Charged amino acid-to-alanine substitutions revealed that Asp-1614, Glu-1615, and Lys-1617 in the VWF A2 domain act synergistically in ADAMTS13-mediated cleavage (26), suggesting that these charged residues in VWF are targets for exosite-1. Recently, Arg-349 was suggested to interact directly with VWF, most probably with Asp-1614 (27). Leu-350 and Val-352, which form a cluster of hydrophobic residues adjacent to the end of the catalytic cleft (Fig. 3B), also interact with VWF (27). This observation suggests that the hydrophobic cluster functions as a part of exosite-1.

The C_A domain has 3 surface loops. The ΔV-loop (C_A) mutant resulted in very low enzymatic activity (Fig. 3 C and D), suggesting that the V-loop (C_A) creates another exosite (exosite-2). A triple alanine substitution in the V-loop (C_A), H476A/S477A/Q478A, and a mutant at the N terminus of the HVR (C_A) adjacent to the V-loop (C_A), R488E, had significantly reduced activity (≈21%), suggesting that these hydrophilic or charged residues play a pivotal role in VWF recognition at exosite-2. The ΔU-loop and the F494Q/M496Q mutants showed reduced activity (≈40%) compared to the ΔP-loop mutant (≈53%). The U-loop (C_A) and residues 494–496 flank the V-loop (C_A) and may contribute to exosite-2. In contrast, the P-loop is distant from the V-loop (C_A) and may contribute less to VWF binding. A mutation in the middle of the HVR (C_A), K497E, maintained enzymatic activity comparable to wild type, even though Lys-497 is the equivalent of Arg-349, the pivotal residue in exosite-1 in the homologous D domain.

The 3 distal loops in the S domain were replaced by short linkers and enzymatic activity in the mutants was assayed. The Δβ7-β8-loop and Δβ9-β10-loop mutants showed greatly reduced activity compared to the Δβ5-β6-loop mutant (Fig. 3 C and D). The β9-β10 loop contains 2 tyrosine residues, Tyr-661 and Tyr-665, which face the solvent. The Y661Q/Y665Q mutant was significantly less active (≈18%) than the wild type. These tyrosine residues and Leu-668 in the β9-β10 loop form a hydrophobic cluster together with residues in the neighboring β3-β4 and β7-β8 loops (Pro-590, Leu-591, Phe-592, Leu-637, and Pro-638) (Fig. 3E). Leu-591 and Phe-592 are located at the center of this hydrophobic cluster. The L591Q/F592Q mutant showed reduced activity (Fig. 3D). Four arginine residues surround the hydrophobic cluster (Fig. 3E). The R568Q/R660Q mutant was significantly less active than the wild type. Collectively, the hydrophobic cluster rimmed with arginine residues in the S domain may to be another VWF-binding exosite (exosite-3).

Mutations of residues (R386S/R421S and L443Q) located apart from the exosites and the O-linked fucosylation site (S339A) did not affect activity. The R393A/R407A mutation in T1 showed reduced activity (≈33%). Arg-407 is located at the bottom of a cleft formed between exosite-1 and exosite-2. This residue may also contribute to VWF binding.

Discussion

This study presents a structural determination of the ADAMTS13 DTCS domains, which constitute important functional part of the proteinase. The structure revealed that the residues important for stabilizing the DTCS core architecture are strictly conserved in all ADAMTS proteins. In contrast, peripheral loops within the D, C_A, and S domains were substantially different in both length and amino acid sequence among ADAMTSs, suggesting that these regions have specific functions that distinguish each ADAMTS member. By systematic mutagenesis, we identified 3 VWF-binding exosites in these loops (Fig. 3C). The exosites are highly conserved among ADAMTS13s from different species (Fig. S5). The 3 exosites were linearly aligned in the 3D structure, traversing the W-shaped ADAMTS13-DTCS molecule (Fig. 3C). This arrangement suggests that these exosites bind collaboratively to multiple discontinuous regions of VWF.

A recent crystallographic study revealed that the Tyr-1605-Met-1606 scissile bond of VWF is buried within the core of the globular A2 domain under static conditions (Fig. 4A, A2 folded) (28). When VWF is subjected to fluid shear stress in circulation or denaturants in vitro, the A2 domain unfolds and adopts a partially extended conformation that makes its scissile peptide bond accessible for cleavage by ADAMTS13 (7, 29, 30) (Fig. 4A). We previously identified VWF73 (residues 1,596–1,668) as a minimum specific substrate for ADAMTS13 and suggested that a segment (residues 1,660–1,668) of VWF73 contains essential residues for recognition by ADAMTS13 (31). VWF73 is more than 200 Å long at its maximum extension, which is almost twice the distance between the catalytic site and the distal exosite-3 in the current ADAMTS13-MDTCS model. NMR spectroscopy has indicated that VWF73 adopts an unfolded structure (5). Therefore, the ADAMTS-MDTCS appears to be able to accommodate, by an induced-fit mechanism, a partially unfolded VWF73 segment along the extended molecular surface encompassing at least 3 critical exosites (Fig. 4B). Exosite-3 forms a cluster of hydrophobic residues rimmed by basic residues (Fig. 3E). Both the surface properties and the size of exosite-3 imply that exosite-3 binds to VWF, such that the VWF segment (residues 1,653–1,668) forms an amphipathic α-helix (α6 as in the crystal structure, Fig. 4C) and makes contact with ADAMTS13 by facing its hydrophobic residues toward exosite-3. Autoantibodies that inactivate ADAMTS13 are the most frequent cause of acquired TTP. These TTP patients possess antibodies directed against ADAMTS13 residues 657–666 (32) that exactly coincide with the β9-β10 loop, a part of exosite-3.

Fig. 4.

ADAMTS13-VWF interactions. (A) Folded and unfolded structures of the VWF A2 domain. The VWF A2 domain adopts a Rossman fold with a central 6-stranded β-sheet surrounded by 5 α-helices (shown as “A2 folded”) (28). The scissile peptide bond (Tyr-1605-Met-1606) is buried within the protein core under static conditions. The C-terminal region (residues 1,596–1,668, corresponding to VWF73) (31) of the A2 domain must be unfolded to expose the scissile bond and the exosite-binding regions under shear-stress conditions (shown as A2 unfolded). (B) ADAMTS13-MDTCS-VWF binding model. The molecular surface of the ADAMTS13-MDTCS model is shown in gray and the bound zinc ion is shown in yellow. Residues that mediate VWF binding are depicted as in Fig. 3C, and the exosites and the catalytic cleft are indicated by red and yellow dotted ellipsoids, respectively. The dotted green line represents a VWF molecule (residues 1,596–1,668) bound to ADAMTS-MDTCS. (C) Close-up view of the α6 helix and surrounding residues in the VWF A2 domain. Hydrophobic residues are indicated with red letters. Systematic charge-to-alanine substitutions revealed that the D1653A and D1663A mutations (cyan) reduced the substrate cleavage, the E1655A mutation (orange) slightly increased cleavage, and the R1659A, E1660A, and R1668A mutations (gray) had no significant effect (34).

The present structure suggests a linear correspondence between the ADAMTS13 domains and their interaction sites in the A2 domain of VWF, consistent with previous systematic mutagenesis studies and kinetic analysis by Gao et al. (33). These authors suggested that the S domain contains an exosite that primarily determines catalytic efficiency by interacting with α6 of the VWF A2 domain (33). They identified 3 other VWF segments that interact with the MD, T1, and C domains of ADAMTS13 (17). Our structural and functional data are in good agreement with these observations, suggesting that the catalytic cleft plus exosite-1, exosite-2, and exosite-3 make cooperative, modular contacts with 3 discrete segments of the VWF A2 domain, the residues flanking the cleavage site (P9-P18′, residues 1,596–1,623), residues 1,642–1,652 and the α6 (residues 1,653–1,668) of the A2 domain, respectively (Fig. 4B). The model is also consistent with the previous observation that decreasing the length of peptides derived from the C terminus of the VWF A2 domain caused a progressive decrease in their potency as ADAMTS13 inhibitors (34). The elongated structure of the stiff, rod-like T1 module and its nonessential interactions with VWF (17) suggest that its primary role is to position the exosites spatially. The mobility of the domains (Fig. S3 and SI Text) suggests that a spectrum of ADAMTS13 conformations exist, with different spatial alignments of the exosites, increasing the possibility of ADAMTS13 interacting with partially unfolded VWF molecules, which also present a wide spectrum of conformations under shear-stress conditions in the circulation. The M domains of ADAMTS4 and ADAMTS5 do not retain specific catalytic activity. The inclusion of the proximal C-terminal domains enhances their aggrecanase activity, suggesting that these ADAMTSs function through multiple exosites (35–39), as observed in the ADAMTS13-VWF system.

More than 80 causative mutations for congenital TTP have been identified in the ADAMTS13 gene (11, 40, 41), including 16 missense mutations within the DTCS region. These mutations are not restricted to a specific region but are located throughout the molecule, suggesting that most of the mutations cause some structural defect that affects proper folding and secretion (Table S2). The R349C and P353L mutants, however, are likely to affect enzymatic activity: Arg-349 is in exosite-1 and Pro-353 forms part of the potential substrate-binding S3′ pocket (Fig. 3B). Five polymorphisms have been identified within the DTCS region (Table S2). Approximately 10% of the Japanese population are heterozygous for P475S substitution, located in the V-loop (C_A), which reduces VWF-cleaving activity (40, 42). The P618A substitution reduces secretion efficiency in cultured cells (43). Both proline residues adopt the cis conformation and, therefore, substitution by nonproline residues would cause structural distortions.

Shear stress in the blood circulation controls the exposure of the cryptic scissile bond and exosite-binding regions in VWF to ADAMTS13. The M domain of ADAMTS13 is catalytically active, whereas the noncatalytic domains display surface features that are optimized for recognizing an unfolded VWF A2 domain. Therefore, cleavage by ADAMTS13 is primarily dependent on shear-force-induced unfolding of the VWF molecule. The force-induced proteolysis observed for ADAMTS13-VWF represents a model for probing the molecular mechanisms underlying the translation of a mechanical stimulus into a chemical response in a biological system.

Materials and Methods

Preparation, Crystallization and Structural Analysis of ADAMTS13-DTCS.

Production and crystallization of ADAMTS-DTCS has been described previously (44). Briefly, ADAMTS13-DTCS (residues 287–685), with a C-terminal tobacco etch virus proteinase cleavage site followed by tandem His-tag sequences, was expressed in CHO Lec 3.2.8.1 cells. After purification on a Ni-NTA column, ADAMTS13-DTCS was subjected to proteolysis with the tobacco etch virus proteinase and was further purified using HiTrap SP (GE Healthcare). ADAMTS13-DTCS crystals were obtained by the sitting drop vapor diffusion method, with drops containing 0.5 μL protein solution and 0.5 μL reservoir solution (26% (wt/vol) PEG1500, 100 mM Mes, pH 6.0) supplemented with 0.2 μL of 40% (wt/wt) pentaerythritol ethoxylate (3/4 EO/OH) (Hampton Research) equilibrated for several days at 293 K. Os-derivative crystals were obtained by soaking native crystals in reservoir solution supplemented with 1 mM OsCl₃ and 20% glycerol for several hours. Crystals were cryoprotected in reservoir solution supplemented with 20% glycerol and flash cooled under a stream of nitrogen gas at 100 K. All diffraction data were collected at the SPring-8 beamline BL41XU (Table S1). Details of structural analysis are described in SI Text.

Functional Analysis.

Recombinant wild-type and 25 mutants of ADAMTS13-MDTCS (residues 75–685) with a C-terminal His-tag were prepared by transient expression using a cytomegalovirus promoter-driven expression vector and HeLa cells. The culture medium and cell lysates were collected 72 h posttransfection, and the expression levels were quantified by Western blotting using anti-His-tag (Fig. S6). For enzyme assays, culture medium (5 μL) containing equivalent amounts of ADAMTS13-MDTCSs was mixed with reaction mixture (95 μL) containing 2 μM fluorogenic substrate (FRETS-VWF73) (25), 10 mM Hepes (pH 7.4), 150 mM NaCl, 5 mM CaCl₂, and 0.005% Tween-20. Initial velocities of the increase in fluorescence were determined for the enzymatic activity, and the relative activities of the mutants were calculated from a calibration curve for serially diluted wild-type ADAMTS13-MDTCS. The activity for each mutant was determined in duplicate or triplicate experiments.