Free Thiols of Platelet Thrombospondin EVIDENCE FOR DISULFIDE ISOMERIZATION*

The free thiols of platelet thrombospondin (TSP) were derivatized with labeled N-ethylmaleimide (NEM) or iodoacetamide (IAM). When Ca2+ was che- lated with EDTA, 2.9 mol of NEM mol of IAM reacted/m01 of native TSP. No additional thiols were found after denaturation with urea. Since TSP has three apparently identical polypeptide chains, sug-gests one free thiol/polypeptide chain. Ca” protected all of the thiols from reaction with IAM. In Ca” about half the thiols reacted normally with NEM and the others were unreactive, indicating that the thiols of TSP are not identical. The number of reactive thiols as a function of [Ca”] revealed a sigmoidal curve with a transition midpoint of 207 PM. The ability of analogs of NEM to compete for derivatization of the thiols with labeled NEM was greater with larger, more hydrophobic agents. separation of labeled TSP that

The free thiols of platelet thrombospondin (TSP) were derivatized with labeled N-ethylmaleimide (NEM) or iodoacetamide (IAM). When Ca2+ was chelated with EDTA, 2.9 mol of NEM or 2.6 mol of IAM reacted/m01 of native TSP. No additional thiols were found after denaturation with urea. Since TSP has three apparently identical polypeptide chains, this suggests one free thiol/polypeptide chain. Ca" protected all of the thiols from reaction with IAM. In Ca" about half the thiols reacted normally with NEM and the others were unreactive, indicating that the thiols of TSP are not identical.
The number of reactive thiols as a function of [Ca"] revealed a sigmoidal curve with a transition midpoint of 207 PM. The ability of analogs of NEM to compete for derivatization of the thiols with labeled NEM was greater with larger, more hydrophobic agents.
Gel electrophoretic separation of labeled TSP that had been partially digested with thrombin and trypsin indicated that some of the label was in the C-terminal tryptic fragment but that most was in the adjacent trypsin-sensitive region.
After cyanogen bromid cleavage of the labeled and reduced protein, four labeled fractions were obtained from a gel filtration column.
With subsequent combinations of tryptic digestion and reversed-phase high performance liquid chromatography, labeled peptides were purified from these four fractions, and the amino acid sequences were determined.
Twelve labeled cysteines were identified, each with a specific radioactivity less than that of the thiol labeling reagent, indicating that only a fraction of that cysteine in a population of TSP molecules was a free thiol at the time of derivatization.
The disulfide bonds most sensitive to reduction by dithioerythritol were also stabilized by Ca2+, implying location in the Ca2+-sensitive part of the molecule.
It is proposed that one equivalent of free thiol/polypeptide chain is distributed among 12 different cysteine residues through an intramolecular thioldisulfide isomerization.
* This work was supported by National Institutes of Health Grant HL-16355.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ To whom correspondence should be addressed.

Thrombospondin
(TSP)l is a large (MI = 420,000) glycoprotein composed of three apparently identical, disulfidelinked polypeptide chains (for a review, see Refs. 1 and 2). Each chain has a globular domain at the C-terminal end connected through a long thin strand to a smaller globular region at the N-terminal end. The three chains are disulfidelinked near the N-terminal globular region. The conformation of TSP is dependent on Ca2+ (3-5). TSP in the presence of Ca" is more compact (3, 4) and is less susceptible to proteolysis (3) than is TSP in the absence of Ca*+. On addition of EDTA to a TSP solution containing Ca*+, there is an increase in the length of the connecting strand in TSP and a concomitant decrease in the size of the C-terminal globular region (4). Thus, it appears that the C-terminal half of the molecule is sensitive to Ca'+. An amino acid sequence deduced from cDNA revealed eight repeat sequences that were proposed to be the calcium-binding region (6); these repeats are approximately at the predicted junction of the thin connecting strand and the C-terminal globular region. TSP is secreted from a-granules by activated platelets (7), and it is released into the medium by various cultured cells, including endothelial cells (8, 9), fibroblasts (lo), aortic smooth muscle cells (ll), monocytes and macrophages (12), and neuroglial cells (13). TSP released from cultured cells is incorporated into their extracellular matrices (10). TSP exhibits binding affinities for a variety of molecules, including heparin (14), fibrinogen (15, 16), fibronectin (16,17), and collagen (17)(18)(19). TSP has been implicated in such cellular activities as cell attachment and spreading (20)(21)(22), cell growth and tissue repair (13, [23][24][25], platelet aggregation (26-28), and tumor metastasis (29). Thus, TSP appears to be an adhesive macromolecule that is involved in cell-cell and cellmatrix interactions.
Most of these interactions are presumed to be non-covalent, but the demonstrations that TSP forms disulfide-linked multimers (30,31) and disulfide-linked complexes with thrombin (32) suggest the possibility that intermolecular thiol-disulfide exchange between TSP and other proteins with which it interacts may be a general mechanism. Since a thiolate anion is essential for thiol-disulfide exchange, we have focused our attention on the free thiols of TSP. A thiol was first reported by Danishefsky et al. (32), who estimated that 2.6 mol of NEM reacted per mol of denatured TSP. The thiols were predicted to be Ca*+-sensitive because formation of disulfidebonded complexes with other proteins and with activated thiol-Sepharose was inhibited by Ca2+ (31,32), and because a spin label on thiols of TSP had less mobility in a medium '   We report here that there are 3 equivalents of thiol/molecule of platelet TSP (presumably one/polypeptide chain) but that these 3 equivalents are distributed among at least 12 different cysteine residues in the Ca*'-sensitive part of the molecule. This implies multiple disulfide bond pairings, indicating that secreted TSP is in transition to a new stable conformation with different disulfide bond configurations or that TSP exists with multiple conformations in equilibrium. A secreted disulfide isomerase is suggested.

RESULTS
Free thiols of secreted platelet proteins were derivatized by addition of EDTA followed by either [3H]NEM or ['%]IAM to the supernatant solution of A23187-activated platelets within 30 min of secretion. Derivatized TSP was purified by heparin-agarose chromatography as described under "Materials and Methods." TSP was the major protein in the supernatant solution to react with either of the labeled thiol reagents ( Fig. 1).
Quantification of ThioLs-Th e extent of derivatization with thiol reagents is shown in Table I, and the time courses of derivatization are in Fig. 2. Consider reactions with NEM first. 2.9 mol of NEM reacted with 1 mol of native TSP in EDTA (Table I). Addition of 8 M urea caused the reaction to be faster ( Fig. 2) but to the same extent (Table I), demonstrating the absence of completely buried thiols. Preincubation with 5,5'-dithiobis(2-nitrobenzoic acid), which is absolutely specific for free thiols (34), completely blocked derivatization with NEM in urea, indicating that the reaction of NEM was completely specific for thiols in TSP. These data suggest one ' Portions of this paper (including "Materials and Methods," part of "Results," Figs. 1 and 6-16, and Tables  II and III)   thiol/polypeptide chain for the three-chain protein.
In 2 mM Ca2+ the rate of reaction (Fig. 2) was similar, but there was only about half as much NEM reacted with TSP (Table I). This demonstrates that the thiols were not identical; some reacted quickly and the others were unreactive.
Derivatization with IAM differed slightly. While the number of reactive groups in EDTA (2.6 mol of IAM reacted/m01 of TSP) was similar, inhibition by Ca2+ was nearly complete (Table I, Fig. 2). Urea caused an increase in the extent of derivatization (Table I), but this increase was not blocked by 5,5'-dithiobis(2-nitrobenzoic acid), indicating that when the protein is denatured, IAM reacts with groups other than thiols. The time courses (Fig. 2) are consistent with this conclusion; in urea the initial rate was that expected for reaction of IAM with thiols, but it was followed by a slower, nearly linear increase in reaction to a level well above 3 mol IAM/mol TSP. The time courses for derivatization and the extent of derivatization were similar whether TSP was derivatized in the supernatant solution or after heparin purification (data not shown), suggesting that no other secreted component significantly modifed the derivatization of TSP. We conclude from the data of Table I   The blank was labeled NEM without a competing derivative. The samples were analyzed by SDS-PAGE. out indirect complications due to possible calcium-dependent TSP-TSP interactions, the experiment was repeated with the same TSP preparation diluted 5-fold; the results were identical.
Maleimide Derivatives and Thiol Deriuatization-To evaluate steric acid hydrophobic factors in thiol reactivity, various maleimide derivatives were added simultaneously with an equal concentration of [ 'H]NEM and allowed to compete for the thiols in TSP (Fig. 4). As a negative control, ['H]NEM was added with solvent alone, and as a positive control, nonradioactive NEM was allowed to compete with labeled NEM. Inhibition increased with size and hydrophobicity until N-pyrenylmaleimide apparently encountered some steric hindrance to thiol derivatization.
Localization of Thiols-Limited thrombin-and trypsin-catalyzed digestions of IAM-labeled TSP are analyzed in Fig. 5. The mass of each peptide was estimated and compared with those reported by Lawler et al. (4); despite some minor differences in size, the major peptides reported for thrombin diges- Proteins in the supernatant solution from activated platelets were derivatized with ["'CIIAM in EDTA, and TSP was purified on a heparin-agarose column. TSP (400-450 rg/ml) was diluted by a factor of 2.5 with 6 mM EDTA, 100 mM Tris-HCl, pH 8.1, and incubated with 100 nM thrombin at 37 "C. A, the digestion proceeded for 0 h (lane I), 6 h (lane 2), or 12 h (lane 3) before addition of the samples to reducing SDS-PAGE sample buffer. The samples were resolved on duplicate 8816% gels, which included a thrombin control (no TSP). One gel was stained with Coomassie Blue, and the other was prepared for fluorography. The apparent mass of each peptide was calculated from the mobility of the molecular weight standards as indicated. B, the 12 h sample in A was divided into 3 aliquots and chilled to 0 "C. Trypsin (in 0.1 mM HCl) was added at a trypsin/TSP ratio (w/w) of 0 (lane I), l:lO,OOO (lane 2), and l:l,OOO (lane 3). The digestion proceeded at 0 "C for 40 h, at which time 1.5 mM tosyl-lysyl chloromethyl ketone, 1 mM phenylmethylsulfonyl fluoride, and 20 mM HCI were added. SDS-PAGE analysis of the digested samples was similar to that in A. Similar results were obtained with ["HINEM-TSP. Since the 90-kDa fragment is derived from the N-terminal part of the 123-kDa thrombic peptide (4), we conclude that most of the radioactivity was associated with the C-terminal portion of the 123-kDa thrombic peptide, the part of the molecule that is most sensitive to tryptic proteolysis (4). We tentatively conclude that most of the radioactivity was in the C-terminal part of the 123-kDa thrombic peptide and was lost as small peptides after digestion with trypsin. In addition, some label was in the C-terminal end of the intact molecule (the 20-kDa thrombic peptide).
For a more precise location of the thiols, derivatized TSP was cleaved for isolation of labeled peptides in order to determine their sequence for location in the published sequence of TSP (6). We were unsuccessful in our attempts to purify NEM-labeled peptides, a problem also reported by others (35-37). We therefore used IAM-labeled TSP, resulting in more easily purified peptides. While this caused some concern about nonspecific labeling of the protein, radioactivity was recovered only in a cycle of the sequenator where a cysteine residue was identified (see Table II). ['CJIAM-TSP was heparin-purified, reduced, alkylated, and cleaved with cyanogen bromide. The cyanogen bromide peptides were separated into four labeled fractions on a gel filtration column. These fractions were further resolved on a reversed-phase (03) HPLC column either before or after further digestion with trypsin. Additional Cs and Cl8 chromatographic separations were used to further purify some peptides. Fig. 6 is a flow diagram that shows the sequence of the digestions and separations as well as our nomenclature of the peptides, and Figs. 7-16 show the separation on each column. The amino acid sequences of the labeled peptides were determined on a gas-phase sequenator; data are shown in Table II. It was apparent that there were multiple labeled peptides; they are summarized in Table III and indicated in the partial sequence of TSP in Fig. 17.
Of the total 69 cysteines in TSP, 12 were labeled. All of the labeled cysteines are in the C-terminal part of TSP at positions 687, 695, 700, 720, 838, 856, 876, 892, 912, 928, 974, and 1149. All but the final two are in the proposed calcium-binding region (type 3 repeats) identified by Lawler and Hynes (6). The specific radioactivities of the labeled cysteines were estimated (see Miniprint) to be only 3-25% of the specific radioactivity of the IAM used to label them, confirming that only a small fraction of that residue was a free thiol at the time of labeling. That is, TSP existed as a mixture of molecules with the free thiol at different cysteines. Three of the labeled cysteines may be special. CYS'~' and Cysg2* (the last cysteines of the final two type 3 repeats) and Cys974 (the first cysteine after the type 3 repeats) had specific radioactivities  four to six times greater than the others. It is clear, however, that there was no single cysteine that was predominately a free thiol when labeled.
Identification of Labile Disulfide Bonds-The number of labeled cysteines, 12, is clearly greater than the 3 mol of NEM or IAM incorporated/mol of TSP (Table I). This suggested that, rather than a stable thiol(s) in TSP, there is a mechanism of thiol-disulfide exchange among the various labeled cysteines, most of which are in the putative calcium-binding region of the molecule. Is this because the disulfide bonds in this region of the molecule are less stable than those in the remainder of the molecule? To investigate this possibility, the number of free thiols was measured after incubation of TSP with low concentrations of dithioerythritol (Fig. 18). Disulfide bonds were more readily reduced when TSP was in a solution with EDTA compared with a solution with Ca'+. That is, the most labile disulfide bonds were stabilized by Ca*+ and, therefore, are likely to be in the Ca*'-sensitive part of the molecule.

DISCUSSION
There are 3 mol of thiol/mol of TSP (Table I), implying one thiol per polypeptide chain. We identified, however, 12 different labeled cysteine residues, indicating that each of the labeled cysteines was a free thiol in only a small fraction of TSP molecules. That is, the position of the thiol must vary through thiol-disulfide exchange. This is consistent with our observations that the thiols are not identical and that the specific radioactivity of the labeled cysteines was much less than the specific radioactivity of the labeling reagent.
Each of the labeled cysteines (i.e. cysteine with a free thiol) of TSP is in the Ca'+-sensitive C-terminal part of the molecule, and all but 2 (CysgT4 and CYS"~') are in the repeat sequences (Fig. 17) proposed to be the Ca*+-binding part of the molecule (6). Of the eight repeat sequences, we failed to detect a labeled cysteine in only three, encompassing residues 741-822. These repeats contain 17 of the total 69 cysteines in TSP, and 10 of the 17 were labeled. The other 2 labeled cysteines are in the adjacent 120-residue C-terminal sequence, with 1 the fourth residue from the C terminus. Location of the thiols in the Ca*+-sensitive part of the molecule apparently is related to the ability of Ca'+ to modify the reactivity of thiols in TSP and to modify the stability of disulfide bonds.
A basic principle of protein chemistry is that disulfide bond pairing is established by the most stable conformation of the  Why then did we find multiple locations of the thiol and, by inference, multiple disultide pairings? The likely explanation is that we labeled thiols during a transition from one stable conformation (with stable disulfide bonds and a single thiol) to another stable conformation (with different disulfide bonds and a different thiol). There could, for example, be a change from a conformation that exists when TSP is packed in a platelet a-granule, possiblly with little or no calcium, to another conformation after it is secreted into a dilute solution containing Ca*+. Similarly, there could be a change from the Ca*+-bound to the Ca*+-free conformation when EDTA was added prior to the labeling reagent. An example is serum albumin; it contains 17 disulfide bonds and one thiol (38), and in a solution of low ionic strength and alkaline pH, there is a disulfide isomerization (39).
A second possibility for multiple disulfide pairing is that TSP exists in multiple, equally stable conformations, each with its own disulfide pairings. Thus, rather than a stable thiol(s) in TSP, there may be an equilibrium thiol-disulfide exchange among the various residues labeled, with the fraction of a cysteine existing as a free thiol determined by the relative stability of that conformation. Such multiple conformations could help explain the ability of TSP to bind to so many proteins; TSP could alter its disulfide bond pairings to maximize its interactions with other proteins. An example of an equilibrium thiol-disulfide exchange is p-lactoglobulin (40), which exists as an equilibrium of two forms, each with its own disulfide bond pattern and own thiol.
It is clear from the studies described here that TSP has thiols that react with disulfide bonds within the molecule, and it has been reported in other studies that thiols on TSP react with disulfide bonds on other protein molecules (30)(31)(32). What is the physiological significance of these reactions? The greatest reactivity with other molecules occurs in a Ca'+-free medium, but after secretion TSP presumably is in a Ca'+containing medium. It is possible that the Ca*+-sensitive region of the TSP molecule is also sensitive to other environmental factors. Non-covalent interactions with other proteins or adhesion to cell surfaces might induce conformational changes that permit intermolecular thiol-disulfide exchange. For example, Narasimhan et al. (41) reported that one of the two buried thiols of fibronectin became accessible to 5,5'dithiobis(2-nitrobenzoic acid) when the protein was adsorbed to a solid surface. A similar situation may exist with TSP, and Ca*+ may preserve reactive thiols and inhibit intermolecular thiol-disulfide exchange until a need is signaled by environmental factors.
Thiol-disulfide exchange between a thiol on TSP and a disulfide bond on other proteins, such as extracellular matrix proteins or proteins on cell surfaces, obviously raises many intriguing possibilities. It is especially interesting that one of the labeled cysteines is in an Arg-Gly-Asp-Ala-Cysgl* sequence; Arg-Gly-Asp has been shown to mediate the attachment of several different adhesive proteins to cells (42), and Lawler et al. (43) demonstrated that it mediates attachment of TSP in a Ca'+-dependent manner to a receptor on a number of different cell lines. The covalent bonding of TSP to cells and to other matrix proteins through thiol-disulfide exchange would open nearly unlimited possibilities for physiologically significant reactions.
Thiol-disulfide exchange, or disulfide isomerization, is normally a very slow reaction unless catalyzed. It is possible that platelets secrete something that can catalyze disulfide isomerization. There are several proteins known to have disulfide isomerase activity. Protein disulfide isomerase is an enzyme of the endoplasmic reticulum that presumably catalyzes isomerization of disulfide bonds as a protein is synthesized (44). Thioredoxin is a protein with multiple functions involving a very reactive pair of cysteines that readily undergo oxidation to a disulfide and reduction to dithiols (45). It can reduce protein disulfide bonds, and it can catalyze disulfide bond isomerization (45). Protein disulfide isomerase and thioredoxin share an active center sequence (Table IV) (46), and it has been reported (47) that the P-chains of lutropin and follitropin contain a similar sequence and exhibit 300 and 60 times more protein disulfide isomerase activity than thioredoxin. The consensus sequence of these four proteins, -Cys-Gly-X-Cys-, is in TSP at position 554-557 (there are four additional -Cys-X-X-Cys-), but we did not identify it as reactive with labeled IAM. The same sequence is in von Willebrand factor, a protein that also is secreted by activated platelets and that also undergoes thiol-dependent multimerization (48), and TSP and von Willebrand factor share the property of forming multimers optimally at pH 6 (31,48). Each of the sequences in Table IV also has a positive charge adjacent to a cysteine (it has been proposed that this would tend to stabilize the thiolate anion by forming a base pair, Ref. 45), and other residues may be close in the secondary structure and modify reactivity. It is possible that thioldisulfide exchange in TSP is catalyzed by TSP itself or by some other factor secreted by activated platelets.