Immunochemical Analysis of the Structure of the Signature Domains of Thrombospondin-1 and Thrombospondin-2 in Low Calcium Concentrations*

Thrombospondins (TSPs) undergo conformational changes upon removal of calcium. The eight C-type and five N-type calcium-binding repeats of TSP-2 form a circuitous wire that, in 2 mm calcium, interacts at its ends with more N-terminal epidermal growth factor (EGF)-like modules, EGF2 and EGF3, and the C-terminal lectin-like module. These components, along with the other EGF-like module(s), form the signature domain of TSPs. Characterization of conformation-sensitive epitopes of monoclonal antibodies to human TSP-2 and its TSP-1 homolog have given insights into the structure of the signature domain in the absence of calcium. The epitope for 4B6.13 anti-TSP-2 was localized to His-722 and Leu-703 in repeat 1C of the wire; recognition only occurred in constructs that included EGF3, the rest of the wire, and the lectin-like module and in the presence of calcium. The epitope for C6.7 anti-TSP-1 was localized to Glu-609 in the EGF2 module. The C6.7 epitope was preferentially recognized when EGF2 was expressed in the context of EGF1, EGF3, the wire, and the lectin-like module. Preferential recognition of the C6.7 epitope did not require calcium. Rotary shadowing electron microscopy of TSP-1 has shown elongation of the stalk and diminution of the C-terminal globule. We propose a model whereby at low calcium concentrations the lectin-like module drops away from EGF3 concomitant with changes in conformation of the wire and loss of the 4B6.13 epitope. A critical feature of the model is interaction of repeat 12N of the wire with EGF2 in both the presence and absence of calcium.

Thrombospondins (TSP) 2 are a family of Ca 2ϩ -binding extracellular matrix glycoproteins. The vertebrate TSPs are divided into two groups based on their oligomerization and modular organization. Subgroup A forms trimers and consists of TSP-1 and TSP-2, and subgroup B forms pentamers and consists of TSP-3, TSP-4, and TSP-5/COMP (cartilage oligomeric matrix protein) (1,2). A monomer of TSP from subgroup A contains an N-terminal module (N), an oligomerization domain (o), a vWF-C module (C), three properdin or type 1 modules (P), three epidermal growth factor (EGF)-like or type 2 modules (E), eight C-type and five N-type Ca 2ϩ -binding repeats to form the wire module (Ca), and a lectin-like module (G) (Fig.  1A). Subgroup B does not have the vWF-C or properdin modules, and contains four EGF-like modules. The hallmark of the TSP family is the highly conserved signature domain or thrombospondin C-terminal cassette (3). The signature domain is stabilized by Ca 2ϩ and composed of interacting elements. These elements are three or more contiguous EGF-like modules, the Ca 2ϩ -binding wire, and the lectin-like module (4).
Atomic absorption spectroscopy estimated that 28 Ca 2ϩ bind to the TSP-2 signature domain (E123CaG-2) in solution (5). This number agrees with the crystal structure, which revealed 26 Ca 2ϩ ions bound to the wire, 3 to the lectin-like module, and 1 to EGF2, for a total of 30 bound Ca 2ϩ ions (4). Biophysical analysis on constructs with various parts of the soluble signature domain of TSP-2 (6) indicated that the elements of the signature domain interact with each other as in the crystal structure (4) and thus comprise a complex structural unit that is influenced by the presence or absence of Ca 2ϩ . Rotary shadowing EM has shown that the removal of Ca 2ϩ dramatically alters the structure of TSP-1, resulting in a decrease in size of a large, apparently globular, C-terminal structure and an increase in the length of adjacent connecting stalk with no change in the size of the N-terminal globule (7)(8)(9). Similar diminution of the C-terminal globule and elongation of the stalk upon depletion of Ca 2ϩ have been demonstrated for TSP-3, TSP-4, and TSP-5 (for review see Ref. 3). How the signature domain rearranges to cause the changes seen in rotary shadowing EM is not known. Here we characterize the epitopes of conformation-sensitive mAbs to TSP-1 and TSP-2 and work from this information to derive a model of the signature domains of TSP-1 and TSP-2 in low Ca 2ϩ .

EXPERIMENTAL PROCEDURES
Monoclonal Antibodies (mAb)-The murine mAb 4B6. 13 was found in a panel of mAbs raised against overlapping recombinant human (h) TSP-2 constructs in collaboration with Dr. MaryAnn Accavitti-Loper and the University of Alabama-Birmingham Hybridoma Core. TSP-1 mAbs included HB8432 (10), A6.1 (11), and C6.7 (12). A6.1 (NeoMarkers TSP Ab-4) and * This work was supported by National Institutes of Health Grant R01 HL054462 (to D. F. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Polyclonal Antibodies-Polyclonal rabbit anti-mouse (m) TSP-2 antibodies were those described previously (13). Polyclonal rabbit antibodies directed against hTSP-2 were produced by the same method.
Expression of TSP Modules Using pAcGP67.coco-We used the vector pAcGP67.coco to express the His-tagged TSP-derived constructs as secreted proteins using recombinant bacu-lovirus (14,15). The baculovirus expression system has been shown to produce arrays of TSP modules that are native, glycosylated, and functional (16 -21). The His-tagged recombinant proteins were purified by Ni 2ϩ -chelate chromatography. Some constructs contained permutations of parts of hTSP-1 and hTSP-2 and are described using nomenclature based on the modules present in the construct followed by a species and type designation (Fig. 1B). In addition, we studied E123CaG signature domain hybrids of hTSP-2 and hTSP-1 or of hTSP-2 and mTSP-2. The signature domain hybrids were designed to include three contiguous segments as follows: 1) the three EGFlike modules, 2) wire repeats 1C-8N, and 3) wire repeats 9C-13C and the lectin-like module. The first and the third segments were from the same species or type, and the second segment was from the other species or type. A construct that consists of mTSP-2 in the first and third segments and hTSP-2 in the second is thus named m2h2m2 (Fig. 1C). Finally, mutations of residues that differed between mouse and human TSP-1 or TSP-2 were introduced into E123CaG-1 from human or E123CaG-2 from mouse ( Fig. 1, B and C).
Direct ELISA-Antigen was diluted to 10 g/ml in TBS (10 mM Tris, 150 mM NaCl, pH 7.4) plus either 0.3 mM Ca 2ϩ , 2.0 mM Ca 2ϩ , or 5 mM EDTA and used to coat 96-well microtiter plates (Costar 3590 high binding, Corning Glass) with 50 l per well, for 16 h at 4°C. Calcium was kept at the concentrations indicated in figure legends with the exception of Ca 2ϩ titration experiments, in which the initial wash and blocking were with no Ca 2ϩ , and subsequent washing and antibody incubation was performed with the appropriate Ca 2ϩ concentration. The plates were washed one time with TBS plus 0.05% Tween 20 (TBST) and blocked with 5% BSA in TBST for 1 h. After washing three times with TBST, the mAbs diluted appropriately in TBST containing the stated Ca 2ϩ concentration were incubated with the plate for 2 h. Plates were washed four times with TBST plus the stated Ca 2ϩ concentration. Alkaline phosphatase-conjugated secondary antibody was incubated with the plate for 1 h. After washing four times with TBST, Sigma 104 AP substrate at 1 mg/ml in TBS, pH 9.0, was added to each well. Color development was monitored at 405 nm.
Competition ELISA-Antigen (5.0 nM TSP-1 purified from human platelets or 10.0 nM recombinant hTSP-2) was diluted in TBS plus 2 mM Ca 2ϩ or 5 mM EDTA and used to coat microtiter plates as in direct ELISA. The plates were washed one time with TBST and blocked with 5% BSA in TBST for 1 h. Calcium was kept at the concentrations indicated in the figure legends. The recombinant TSP fragments (competitors) and the mAb being tested were diluted in TBST plus 0.1% BSA so that when combined the final concentrations for each competitor were 3.0, 1.0, 0.3, 0.1, 0.03, or 0.01 M, and the final concentration of the mAb was 0.5 g/ml for C6.7 and HB8432, or a 1:50,000 dilution of ascites for 4B6. 13. The competitors were added to the mAb in , a vWF-C module (C), three properdin or type 1 modules (P), three epidermal growth factor (EGF)-like or type 2 modules (E) 13 repeats to form the Ca 2ϩbinding wire module (Ca), and a lectin-like module (G). B, our studies concentrated on the signature domain of TSP-1 or TSP-2, designated E123CaG. Constructs other than the signature domain hybrids are named based on their modular components and are followed by a species and type designation. Constructs derived from hTSP-1, hTSP-2, mTSP-1, and mTSP-2 are identified with-1, -2, -m1, and -m2, respectively. Residues in the EGF-like modules E1 and E2 that differ between hTSP-1 and mTSP-1 are indicated and were changed individually to the mTSP-1 residue to map the C6.7 epitope. Residue numbering is based on the full TSP-1 sequence with the initiating Met being residue 1. C, signature domain hybrids of hTSP-2 and hTSP-1 and of hTSP-2 and mTSP-2 were expressed. The hybrids were designed to include three contiguous segments. The first and the third segments were from the same species and type, and the second segment was from the other species or type. A construct that consists of mTSP-2 in the first and third segments and hTSP-2 in the second is thus named m2h2m2. Constructs m2h2m2-A, m2h2m2-B, and m2h2m2-C were used to map the 4B6.13 epitope and are the signature domain of mTSP-2 except for the indicated human residues as follows: cluster A, Leu-703 and His-722; cluster B, Asp-742, Thr-750, and Ala-765; and cluster C, Val-841, Thr-847, and Val-849. Residue numbering for the hybrid constructs is based on the full TSP-2 sequence that has two residues more than TSP-1 inserted N-terminal of the signature domain. microcentrifuge tubes and allowed to incubate for 1 h before being added to the wells coated with TSP-1 or TSP-2. This competition mixture was incubated in the wells for 2 h. The plates were then washed four times with TBST plus the stated Ca 2ϩ concentration. Alkaline phosphatase-conjugated secondary antibody was added, and the amount bound was quantified as above and compared with the absence of competitor.
Statistics-All ELISA experiments were repeated, each with triplicate data points, on multiple occasions. Data are expressed as the mean of the different experiments Ϯ S.E.

Studies of 4B6.13
Anti-TSP-2-4B6.13 is a murine mAb raised against hTSP-2 that caught our interest in preliminary experiments because the epitope was present in hTSP-2 only when Ca 2ϩ was present. We therefore mapped the 4B6.13 epitope. Intact hTSP-2, the complete signature domain of hTSP-2 (E123CaG-2), and portions of the signature domain (E3CaG-2, E3Ca-2, E3-2, and CaG-2) were tested with 4B6.13 in a direct ELISA in the presence of 2 mM Ca 2ϩ . Only intact hTSP-2, E123CaG-2, and E3CaG-2 were reactive (not shown). A competition ELISA for adsorbed intact hTSP-2 was performed to evaluate the possibility that the lack of recognition of the smaller constructs in direct ELISA was because of altered conformation of adsorbed protein. Competition for 4B6.13 was found only when EGF3, the wire, and the lectin-like module were all present in the same construct ( Fig. 2A). E123CaG-2 and E3CaG-2 competed equally well for 4B6.13 binding. These results indicate that the interactions among EGF3, the wire, and the lectin-like module control expression of the 4B6.13 epitope.
4B6.13 did not react with E123CaGs from hTSP-1 or mTSP-2 in direct (not shown) or competition ELISAs ( Fig. 2A). Therefore, hybrid constructs were constructed to map the epitope of 4B6.13 within the E3CaG complex. The hybrids contained a complete signature domain, made up of segments of hTSP-2 in combination with segments of either mTSP-2 or hTSP-1 (Fig. 1C). The design of the hybrids was based on the interactions of the wire with EGF2 and EGF3, and the interactions of EGF3 with the lectin-like module as observed in the crystal structure of the Ca 2ϩ -replete signature domain of hTSP-2 (4). By keeping segments 1 and 3 from the same species and type, we sought to preserve these interactions and retain the integrity and structure of the signature domain. In competitive ELISA, E123CaG-2, m2h2m2, and h1h2h1 competed for 4B6.13 binding but h2h1h2 and h2m2h2 did not (Fig. 2B). The three constructs that competed for binding contained the hTSP-2 sequence in segment 2 (wire repeats 1C-8N), indicating that the 4B6.13 epitope is located within this segment. E123CaG-2 and m2h2m2 competed equally well. Human TSP-2 segment 2 in the context of hTSP-1 (h1h2h1) was ϳ10fold less effective (Fig. 2B). Thus, the epitope for 4B6.13 lies within repeats 1C-8N of the wire and requires surrounding segments, optimally from TSP-2.
Of the 166 residues in segment 2, 8 differ between hTSP-2 and mTSP-2 and are found in three clusters (Fig. 1C). To determine where the 4B6.13 epitope is located, we made variations of m2h2m2 that contained each of these clusters, A, B, or C. In direct ELISA, 4B6.13 reacted with m2h2m2-A but not with m2h2m2-B or m2h2m2-C (Fig. 3), thus localizing the epitope to cluster A, i.e. to Leu-703 and His-722. Further mutations to the residues in m2h2m2-A were generated so that only one residue in each is unique to hTSP-2. These constructs, m2-Leu703 and m2-His722, were tested in direct ELISA (Fig. 3). 4B6.13 bound to m2-His722 and not to m2-Leu703, but the reaction with m2-His722 was only 60% of control compared when both residues are present as in m2h2m2-A and m2h2m2. Rabbit polyclonal antibodies directed against mTSP-2 (13) and hTSP-2 were mixed and demonstrated in a direct ELISA that equivalent amounts of the recombinant constructs used in the 4B6.13 ELISA were coated in the microtiter wells (data not shown). These results indicate that His-722 is the key residue in the 4B6.13 epitope but that Leu-703 also contributes to binding.
EDTA were the same as when the antigen was adsorbed in 2.0 mM Ca 2ϩ before addition of antibody and the test concentration of Ca 2ϩ (data not shown). Thus, the adsorbed protein cycles reversibly between Ca 2ϩ -replete and Ca 2ϩ -depleted conformations that are associated with the expression or not of the 4B6.13 epitope, and the concentration of Ca 2ϩ when the antibody is reacting with TSP-2 is the determining factor for binding.
The EGF1 and EGF2 modules were mutated to determine which residue or residues are responsible for binding of C6.7 and HB8432. Neither mAb reacts with mTSP-1 (15), so the amino acids that differ in EGF1 and -2 (Fig. 1B) were switched individually from the human to the mouse. These mutants were expressed in the context of the hTSP-1 signature domain construct, E123CaG-1. The mutants, E123CaG-1(wt) and mTSP-1, There are eight residues that differ between hTSP-2 and mTSP-2 in segment 2 (wire repeats 1C-8N) of m2h2m2. These variations of m2h2m2 were divided into three clusters (m2h2m2-A, m2h2m2-B, and m2h2m2-C) and were made to determine where the 4B6.13 epitope is located. Cluster A was further mutated to have only one hTSP-2 residue, Leu-703 or His-722. Data are plotted as percent of E123CaG-2 binding. E123CaG-2, m2h2m2, m2h2m2-A, which contain two differences, and m2-His722 were detected by 4B6.13 in direct ELISA. The constructs m2h2m2-B, m2h2m2-C, E123CaG-m2, m2-Leu703, and the BSA control were not recognized by 4B6. 13. Values are expressed as the mean Ϯ S.E. of 4 -5 experiments.

The Signature Domains of TSP-1 and TSP-2 in Low Ca 2؉
were compared in direct ELISA (Fig. 6) with C6.7, HB8432, or as a control A6.1, a mAb that reacts with wire repeat 1C of both human and mouse TSP-1 (15). A6.1 reacted with all five mutants, E123CaG-1, and mTSP-1, whereas C6.7 and HB8432 reacted with E123CaG-1 and all of the mutants except for the E609K construct. Thus, Glu-609 in EGF2 is required for recognition by both C6.7 and HB8432, but only C6.7 requires the presence of the wire and lectin-like module for optimal recognition.
The original characterization of C6.7 showed the ability of immobilized C6.7 to bind 125 I-TSP was maintained at extremely low Ca 2ϩ concentrations (11). We therefore tested the binding of C6.7 in the presence and absence of Ca 2ϩ to the h1h2h1 construct in which epitopes for both 4B6.13 and C6.7 are present. When h1h2h1, E123CaG-1, and E123Ca-1 were tested in competition ELISA with C6.7, the C6.7 competition curves were similar regardless of Ca 2ϩ concentration (Fig. 7A). The h1h2h1 hybrid competed much like E123CaG-1, exhibiting 10-fold better recognition than E123Ca-1, a construct lacking the lectin-like module. HB8342 antibody binding also was not affected by Ca 2ϩ concentration (Fig. 7B).

DISCUSSION
Monoclonal antibodies can be extremely useful probes to study structure-function relationships in conformationally liable proteins, e.g. integrins (24). Herein we describe two mAbs that recognize epitopes that are optimally present when the three elements of the signature domains of TSP-1 and TSP-2, the EGF-like modules, wire, and the lectin-like module, are in the same protein. Reactivities of the two mAbs to the various constructs are summarized in supplemental Table 1. One of these antibodies, 4B6.13 to hTSP-2, requires that the protein bind Ca 2ϩ . The other, C6.7 to hTSP-1, is indifferent to the absence or presence of Ca 2ϩ . Although the mAbs bind to separate members of the TSP family, both react with the h1h2h1 hybrid protein, with each mAb retaining the specific Ca 2ϩ and domain requirements determined for E123CaG-1 and E123CaG-2.
Calorimetric and spectroscopic studies, which indicate that the signature domain of TSP-2 functions as a complex structural unit that is influenced by the presence or absence of Ca 2ϩ (6), are compatible with the requirements for exposure of the 4B6.13 epitope. The antibody only binds when EGF3, the wire, and the lectin-like module are present, and the complex is stabilized with Ca 2ϩ . The epitope for 4B6.13 is in repeat 1C of the wire and requires His-722 with a contribution from Leu-703. The crystal structure of the hTSP-2 signature domain (4) shows these two wire residues in close proximity not only to each other but also to residues that coordinate Ca 2ϩ and interact with EGF3 and the lectin-like module (Fig. 8, A and B). The close proximity of the epitope to residues that coordinate Ca 2ϩ is consistent with the sensitivity of the epitope to Ca 2ϩ concentrations. The interactions needed for exposure of the epitope are maintained in the hybrids, demonstrating that intra-signature domain contacts critical for preserving the structure of the signature domain of hTSP-2 are functional in mTSP-2 and hTSP-1 as  well. However, in competition ELISAs, h1h2h1 did not compete as well as E123CaG-2 or m2h2m2, which suggests a difference in the contributions of segments 1 and 3 of TSP-1 and TSP-2 to the exposure of the 4B6.13 epitope.
Biophysical studies of E123CaG-2 are compatible with the preferred exposure of C6.7 in E123CaG-1 as compared with E123-1. The crystal structure of E123CaG-2 shows interactions of repeat 12N of the wire with EGF2 (4). Differential scanning calorimetry experiments revealed a difference in thermal stability between E3CaG-2 and E123CaG-2; the increase from 50.5 to 53.5°C in the first melting transition is attributed to the interaction between the EGF-like modules 2 or 1 and the wire (5). The hTSP-1-specific antibodies C6.7 and HB8432 each recognized epitopes that were lost when Glu-609 in EGF2 was changed to Lys, as occurs in mTSP-1. Both antibodies bound to the h1h2h1 hybrid construct that contains EGF2 from hTSP-1. Recognition by C6.7 was ϳ10fold better when EGF2 is in the framework of the complete signature domain than when EGF2 is in a construct of tandem EGF-like modules or EGF-like modules and wire. Thus, the interactions within the signature domain, presumably EGF2 with the wire and the lectin-like module with the wire and EGF3, affect the conformation and/or exposure of the C6.7 epitope. The structure of the complete hTSP-1 signature domain has not been solved, and must be modeled on the hTSP-2 structure (4,25). In sequence alignments of hTSP-1 and hTSP-2, Glu-609 in hTSP-1 corresponds to Val-611 from hTSP-2. In the crystal structure Val-611 is positioned in an area near to where repeat 12N of the wire contacts EGF2 (Fig. 8, A and C).
A strong argument can be made that the changes in the connecting strand and C-terminal globule of the TSPs upon removal of Ca 2ϩ , visualized by rotary shadowing EM, involve rearrangement of interactions among the stalk, wire, and lectin-like module that interact in the Ca 2ϩ -replete crystal structure (4) of the signature domain of TSP-2 (Fig. 8A, left). The stalk consists of EGF-like modules 1 and 2 and projects out from the remainder of the signature domain. The length for the connecting strand of Ca 2ϩ -replete TSP-1 (8,9,11,26) or TSP-2 (26), as measured from the intersection of the three chains in the oligomerization domain to the closest edge of the C-terminal globule, is ϳ260 to 310 Å. This length matches the axial lengths of the modules of TSP between the N-terminal module and E3 as follows. The length of the oli-gomerization domain, as deduced from the structure of the coiled-coil region of cartilage matrix protein (27), is likely about 50 Å. The vWF-C module has a length of ϳ50 Å based on the NMR structure of the vWF-C module of type II collagen (28); the properdin modules or TSRs of TSP-1 are each about 50 Å long (29), and the EGF-like modules 1 and 2 of TSP-2 are each about 30 Å long (4). These measurements add up to a predicted length for the Ca 2ϩ -replete connecting strand of 250 -310 Å, depending on how much of the oligomerization domain is included in the calculation. Therefore, the C-terminal globule visualized by rotary shadowing EM presumably includes E3, the wire, and the lectin-like module. Fig. 8A attempts to reconcile the immunochemical findings with rotary shadowing EM images demonstrating that removal of Ca 2ϩ results in a decrease in size of the globular C region and an increase in length of the adjacent connecting strand (7,8,11). In a comparison of Ca 2ϩ -replete versus EDTA-treated TSP-1, Lawler et al. (8) found a decrease in size of the shadowed C-terminal globule, from 120 Ϯ 19 to 80 Ϯ 7 Å, and an increase in length of the thin connecting strand, from 290 Ϯ 46 to 380 Ϯ 30 Å. Frazier and co-workers (9,11) also directly compared Ca 2ϩ -replete to EGTA-treated TSP-1 and found a decrease in size of the shadowed C-terminal globule from 170 Ϯ 20 to 100 Ϯ 20 Å and an increase in length of the thin connecting strand from 260 Ϯ 40 to 430 Ϯ 40 Å. Thus, when Ca 2ϩ is removed from TSP, the two-dimensional size of the C-terminal globule decreases 1.5-1.7-fold, and the length of the connecting strand increases by 1.3-1.7-fold.
In the models of Ca 2ϩ -depleted protein shown in Fig. 8A, EGF3 and the wire become part of the elongated stalk, and the lectin-like module comprises the bulk of the diminished C-terminal globule. The wire is made up of eight C and five N repeats, is stabilized by eight disulfides, and has a contour length of ϳ170 Å (4). In the presence of Ca 2ϩ , the wire extends from EGF3 to form a convoluted circle that loops back to interact with EGF2 and EGF3 and then courses toward the center of the circle where the lectin-like module is suspended (4). The EGF3 module is a major organizer of the signature domain and acts as a "clasp" to close the circle formed by the wire (4). Repeats 1C, 8N, and 9C interact with the lectin-like module, whereas repeats 2N to 7C course away from the lectin-like module to form a prominent opening of 14 ϫ 40 Å (4). The models of Ca 2ϩ -depleted protein FIGURE 8. Localization of epitopes in the crystal structure of Ca 2؉ -replete TSP-2 signature domain and models of TSP in low Ca 2؉ . In the crystal structure of the Ca 2ϩ -replete TSP-2 signature domain (4), the Ca 2ϩ -binding repeats form a circuitous wire that interacts at its ends with EGF-like modules 2 and 3 (yellow) and the lectin-like module (wheat). The wire is depicted in blue (sequences outside of disulfide-bonded loops), green (sequences within disulfide-bonded loops), and orange (portion of repeats 11C and 12N, which is constrained by a disulfide and interacts with EGF2). The Ca 2ϩ -binding repeats are numbered and do not correspond to the color-coded constrained and unconstrained loops, i.e. each Ca 2ϩ -binding repeat contains sequences both inside and outside of disulfide-bonded loops (4). Calcium ions are slate blue spheres. The sugars are gray, and water molecules in the surface models are violet. Disulfide bonds are in red except in the insets, where the cysteines are the same color as the module in which they are located. A, calcium-replete (left) and low Ca 2ϩ (center and right) models of the TSP signature domain. In the low Ca 2ϩ models, in which the wire repeats have been manipulated using Adobe Illustrator, the lectin-like module drops away from EGF3 to cause the elongation and the wire changes conformation with loss of the 4B6.13 epitope, but repeat 12N of the wire (orange) continues to interact with EGF2 to maintain the C6.7 epitope. The extent of elongation of the connecting strand (ϳ90 Å) is based on measurements from rotary shadowing EM (8). In the first low Ca 2ϩ model (center), the 15 residues between the cysteines in repeats 12N and 13C extend. In the second low Ca 2ϩ model (right), the 15 residues between the cysteines in repeats 10N and 11C and 12N and 13C extend, and the interactions of repeats 8N and 9C with the lectin-like module are maintained. B, location of residues important in the 4B6.13 epitope, His-722 and Leu-703, both hot pink, in the structure of the hTSP-2 signature domain. C, location of Val-611, hot pink, in a surface representation of the hTSP-2 signature domain. In sequence alignments, Val-611 of hTSP-2 corresponds to Glu-609 from hTSP-1. Glu-609 is a key residue in the epitopes of C6.7 and HB8432. Val-611 is in close proximity to repeat 12N (orange) of the wire that interacts with EGF2.
are based on the premise that repeat 12N of the wire interacts with EGF2 in both the presence and absence of Ca 2ϩ , thus accounting for the higher affinity binding to the C6.7 epitope when the other parts of the signature domain are present. The wire is separated into sequences that are inside disulfide-bonded loops or outside the loops (Fig. 8A). Assuming that only sequences not constrained by disulfides can extend, we propose two models in which there is extension of one or two of the unconstrained sequences (Fig. 8A). In the first model (Fig. 8A, center), the residues between the cysteines in repeats 12N and 13C (1 residue from 12N and 14 from 13C) extend, and interactions of 8N and 9C with the lectin-like module are not maintained. In the alternate model (Fig. 8A, right), the interactions of repeats 8N and 9C with the lectin-like module are maintained, and in addition to the 15 residues between the cysteines in repeats 12N and 13C, the 15 residues between the cysteines in repeats 10N and 11C (1 residue from 10N and 14 from 11C) extend.
In the models, lengthening of the tether to the lectin-like module is because of the extension of repeat 13C or, in the model in which 8N and 9C continue to interact with the lectin-like module, extension of repeats 11C and 13C. Full extension of the 15 residues between the cysteines in repeats 12N and 13C, or between 10N and 11C, would be ϳ54 Å based on a length of 3.6 Å per residue. This distance added to the 30-Å length of EGF3 approximates the 90-Å elongation of the connecting strand found by Lawler et al. (8) (Fig. 8A,  center and right). The models are consistent with the site at which mAb A6.1, which binds to wire repeat 1C (15), decorated the stalk of EGTA-treated TSP-1 in rotary shadowing EM (9). The A6.1-binding site was in a region of the connecting strand close to the remnant of the large globular region (9). If the wire were to simply unroll, losing the interaction between 12N and EGF2, it could extend to the full contour length of ϳ170 Å (4). This length plus the 30 Å from EGF3 would increase the length of the connecting strand by ϳ200 Å, which is greater than the 90 Å lengthening described by Lawler et al. (8), or the 170 Å lengthening determined by Frazier and co-workers (9,11). Furthermore, the A6.1 epitope would be more equidistant between the N-terminal and C-terminal globules.
In these models, the individual wire repeats, other than the one or two with extended sequences, keep their core structure. Persistence of the structure of 11 of the 13 repeats is consistent with far UV circular dichroism that demonstrates only a 10 -20% loss in the absence of Ca 2ϩ of the strong characteristic band of negative ellipticity at 204 nm that has been shown to arise from the wire (6). However, interactions between adjacent repeats are hypothesized to change upon depletion of Ca 2ϩ , allowing the wire to lose its Ca 2ϩ -replete contour and become part of the stalk.
The requirements for optimal expression of the C6.7 epitope and the fact that the higher affinity of C6.7 for constructs that include the complete signature domain is maintained in the absence of Ca 2ϩ have interesting implications. The interactions between EGF2 and the wire that stabilizes the epitope of C6.7 must be independent of Ca 2ϩ binding in repeat 12N. However, the E123Ca-1 construct lacking the lectin-like module does not bind C6.7 with higher affinity even though EGF2 and repeat 12N are present. These observations suggest that during synthesis the lectin-like module is needed for proper folding of the signature domain, in particular for formation of the interaction between repeat 12N of the wire and the EGF2 module. Once the interaction is established, it is apparently maintained despite the dramatic structural changes caused by the removal of Ca 2ϩ . Preservation of the repeat 12N-EGF2 interaction and possibly of the repeat 8N/9C-lectin interaction would allow the signature domain to convert readily between its Ca 2ϩ -replete and Ca 2ϩ -depleted conformations.