The Specificity and Function of the Metal-binding Sites in the Integrin β3 A-domain*

The A-domains within integrin β subunits contain three metal sites termed the metal ion-dependent adhesion site (MIDAS), site adjacent to the metal ion-dependent adhesion site (ADMIDAS), and ligand-induced metal-binding site (LIMBS), and these sites are involved in ligand engagement. The selectivity of these metal sites and their role in ligand binding have been investigated by expressing a fragment corresponding to the β3 A-domain, β3-(109-352), and single point mutants in which each of the cation-binding sites has been disabled. Equilibrium dialysis experiments identified three Mn2+- and two Ca2+-binding sites with the LIMBS being the site that did not bind Ca2+. Although the ADMIDAS could bind Ca2+, it did not bind Mg2+. These results indicate that the Ca2+-specific site that inhibits ligand binding is the ADMIDAS. Two different assay systems, surface plasmon resonance and a microtiter plate assay, demonstrated that the β3 A-domain fragment bound fibrinogen in the presence of 0.1 mm Ca2+ but not in 3 mm Ca2+. This behavior recapitulated the effects of Ca2+ on fibrinogen binding to αvβ3 but not αIIbβ3. Disabling any of the three cation-binding sites abrogated fibrinogen binding. These results indicate that the specificities of the three metal-binding sites for divalent cations are distinct and that each site can regulate the ligand binding potential of the β3 A-domain.

tiple adhesive ligands, and many of the ligands bind to both receptors, including fibrinogen (8 -10). However, there is a fundamental difference in the specificity involved in fibrinogen recognition by the two integrins; ␣ llb ␤ 3 binds fibrinogen through the sequence at the carboxyl terminus of its ␥ chain 406 KQAGDV 411 (11,12), and ␣ v ␤ 3 binds to one of the two 572 RGD 574 sequences in its A␣ chain (13). In addition, sites within the ␥ chain, distinct from the KQAGDV, have been implicated in recognition of fibrinogen by ␣ v ␤ 3 (14). According to recently published crystal structures of the extracellular regions of ␣ v ␤ 3 (15) and ␣ llb ␤ 3 (16) with bound ligand mimetics, the main area of ligand binding lies between the ␤-propeller of the ␣ subunit and the A-domain in the ␤ 3 subunit. This direct demonstration of ␤ 3 A-domain involvement in ligand engagement was preceded by numerous studies using mutational, immunological, and biochemical approaches (17)(18)(19)(20)(21), which indicated that this region played a critical role in ligand binding to the two ␤ 3 integrins.
The binding of most ligands to integrins requires metal ions. This is also true for the ␤ 3 integrins and the cellular responses arising from ligand engagement (e.g. platelet aggregation (22)(23)(24)). The ␤-propeller domain of the ␣ subunit contains four divalent ion-binding sites within the loops of blades 4 -7 near the bottom of the propeller, but these are not directly involved in ligand engagement (25). The ␤ 3 A-domain contains three metal sites for binding ligands, which are more directly involved in ligand binding. These are referred to as the metal ion-dependent adhesion site (MIDAS), 4 a site adjacent to the MIDAS (ADMIDAS), and a ligand-induced metal-binding site (LIMBS). The ␣ v ␤ 3 and ␣ llb ␤ 3 crystal structures establish that cation bound in the MIDAS is involved directly in ligand engagement; the ADMIDAS cation is not directly involved in contacting ligand but may help to regulate ligand binding; and the LIMBS cation may aid in stabilizing the receptor-ligand complex (26). Both Mg 2ϩ and Mn 2ϩ support ligand binding to ␣ v ␤ 3 and ␣ llb ␤ 3 , but Ca 2ϩ has a differential effect on fibrinogen binding to the two integrins. Ca 2ϩ suppresses fibrinogen binding to ␣ v ␤ 3 but supports its interaction with ␣ llb ␤ 3 (23,24,(27)(28)(29). In the ␣ v ␤ 3 structure crystallized in the presence of Ca 2ϩ , the MIDAS and ADMIDAS were partially or completely occupied, respectively. When an Arg-Gly-Asp ligand was cocrystallized with the extracellular portion of ␣ v ␤ 3 in the presence of Mn 2ϩ , the MIDAS and ADMIDAS had Mn 2ϩ bound and the additional site, LIMBS, also had Mn 2ϩ bound. The Mn 2ϩ at the MIDAS was coordinated by the aspartic acid of the ligand, verifying direct involvement of the MIDAS with the ligand. As the LIMBS was not observed in the unliganded structure; it was proposed that this site stabilized ligand binding within the ␤ 3 A-domain. The crystal structures of ␣ llb ␤ 3 with bound ligand mimetics were formed in the presence of Ca 2ϩ and Mg 2ϩ . These structures revealed three metal sites in which the MIDAS had Mg 2ϩ bound that was also coordinated by the ligand, whereas the ADMIDAS and LIMBS had Ca 2ϩ bound.
In a previous study, we demonstrated that recombinant fragments containing portions of the ␤ 3 A-domain could be expressed in bacteria, folded through a denature/renature protocol, and could then support fibrinogen binding in a divalent ion-dependent manner. High Ca 2ϩ (Ͼ1 mM) suppressed fibrinogen binding, where low Ca 2ϩ (0.1 mM) supported binding (30). In this study, we have expressed wild-type ␤ 3 A-domain and mutants in which a single metal ion site has been disabled. These derivatives have been used to clarify the specificity and determine the contribution of each cation-binding site to fibrinogen recognition by the ␤ 3 A-domain.
Expression and Purification of Recombinant ␤ 3 A-domains-The cDNA coding for the ␤ 3 A-domain, ␤ 3 -(109 -352), was inserted into pRSETa vector (Invitrogen) and expressed in Escherichia coli as an amino-terminal His tag protein as described previously (33). Single mutations of aspartic acids 119, 126, and 217 to alanines were created using the QuikChange site-directed mutagenesis kit from Stratagene (La Jolla, CA). The nucleotide sequences of the wild-type ␤ 3 A-domain and each mutant were confirmed. The expressed proteins were purified on a Ni 2ϩ chromatography column under denaturing conditions (8 M urea). The proteins were refolded by dialysis to reduce the concentration of urea to less than 1 mM. The purity of the recombinant proteins was analyzed by SDS-PAGE and Western blotting, and the major band migrated as a single band at 31 kDa, which is of the appropriate molecular weight for the ␤ 3 A-domain. To prepare ␤ 3 A-domain fragment for NMR analyses, 2 liters of cultures were grown in minimal media containing 15 NH 4 Cl as a sole source for nitrogen. The expression and purification conditions were the same as described above. The heteronuclear single quantum coherence spectrum was collected in 100 mM NaCl, 50 mM phosphate buffer, pH 7.4, on a Bruker 600-MHz instrument equipped with a cryoprobe.
Equilibrium Dialysis Cation-binding Assays-Binding of Ca 2ϩ and Mn 2ϩ to the ␤ 3 A-domain fragments was determined by equilibrium dialysis using a 16-mirocell dialyzer (Sialomed Inc., Columbia, MD). Regulation of the free Ca 2ϩ concentration by EGTA was used to obtain the Ca 2ϩ -binding isotherms of ␤ 3 A-domain fragments in 50 mM Tris, 140 mM NaCl, pH 7.3, containing 50 mM octyl glucoside (equilibration buffer). Total Ca 2ϩ concentrations were determined spectrofluorimetrically (PerkinElmer Life Sciences LS50), using Fura-2 and Ca 2ϩ standard solutions obtained from Molecular Probes, Inc. 45 Ca 2ϩ and 54 Mn 2ϩ were measured in a scintillation counter (Beckman LS6000LL) using 45 Ca and 54 Mn programs, respectively. Below 10 M, the required free Ca 2ϩ concentration was adjusted by addition of 7 mM EGTA, as calculated according to Fabiato (34) to account for influence of pH, temperature, and ionic strength on the equilibrium constants used. Demineralization of the analyzed proteins was done by dialysis for 1 h at 22°C against the equilibrium buffer containing enough EGTA to reduce the free Ca 2ϩ to below 1 nM as described by Rivas and Gonzalez-Rodriguez (35). Equilibrium dialysis experiments were then performed as described previously (33). Various amounts of 10 mM CaCl 2 or MnCl 2 were added in the equilibrium buffer containing 45 Ca 2ϩ (3 Ci/ml) or 54 Mn 2ϩ (0.5 Ci/ ml) to achieve free Ca 2ϩ and Mn 2ϩ concentration between 0.01 and 1000 M or 0.01 and 100 M, respectively. Free Ca 2ϩ concentrations were calculated from the contaminating Ca 2ϩ , determined spectrofluorimetrically, plus the amount of added Ca 2ϩ . 45 CaCl 2 (0.48 Ci) or 54 MnCl 2 (0.05 Ci) was injected into one half-cell, and the proteins into the other one. After dialysis for 24 h at 21°C, aliquots were removed for 45 Ca 2ϩ and protein determinations, and SDS-PAGE was performed to exclude protein degradation. The recombinant ␤ 3 A-domain mutants were used at the concentration of 30 -40 M. To test the effects of Mg 2ϩ and Mn 2ϩ on Ca 2ϩ binding, the equilibrium dialysis was performed in the presence of 45 CaCl 2 (500 M) sufficient to saturate cation-binding sites in ␤ 3 A-domain fragments, and serial dilutions of either Mg 2ϩ , ranging from 0 to 100 mM, or Mn 2ϩ , ranging from 0 to 1 mM, were added. Similarly, to study the effect of Ca 2ϩ on Mn 2ϩ binding, equilibrium dialysis was performed in the presence of 45 MnCl 2 (50 M) sufficient to saturate cation-binding sites in ␤ 3 A-domain fragments, and serial dilutions of Ca 2ϩ in the range of 0 -10 mM were added. Binding isotherms were analyzed via nonlinear regression analysis using Prism 2.01 (GraphPad Prism Software, San Diego) to determine the maximal number of binding sites (B max ) and the apparent equilibrium dissociation constant (K D ) values. The binding experimental data were fit to two ligand binding models as follows: a single class of independent, noninteracting sites or two classes of independent, noninteracting sites, according to the following equations: for a single class of sites, B ϭ nK[L]/(1 ϩ K[L]); or for two classes of sites, ). In these equations, B is the ratio of Ca 2ϩ or Mn 2ϩ bound (micromoles) to micromoles of fusion protein; n refers to the number of binding sites; K is the equilibrium association constant; and [L] is the free Ca 2ϩ or Mn 2ϩ concentration.
Surface Plasmon Resonance-The interaction of fibrinogen with ␣ IIb ␤ 3 , ␣ v ␤ 3 , the ␤ 3 A-domain, and its mutants were determined by SPR. The association and dissociation rate constants (k on and k off ) and the dissociation constant (K D ) were measured using either the BIAcore X or BIAcore 3000 biosensor (Biacore AB, Uppsala, Sweden). The ␤ 3 A-domain and its mutants or fibrinogen were immobilized onto the CM5 sensor chip via amine coupling (BIAcore AB, Uppsala, Sweden) according to the manufacturer's instructions. Experiments were performed at 37°C in 10 mM Hepes buffer, pH 7.4, with or without calcium, magnesium, or EDTA. Experimental data were analyzed using BIAevaluation 3.2 software supplied by Biacore AB. The results were expressed in resonance units (RU), an arbitrary unit specific for the BIAcore instrument (1000 RU correspond to ϳ1 ng of bound protein/mm 2 ). The association rate constant, k on , and dissociation rate constant, k off , were determined from individual association and dissociation phases, respectively, assuming one-to-one interactions. The overall dissociation constant, K D , was derived from k off /k on .
Microtiter Plate Assays-Microtiter wells were coated with 5 g/ml ␤ 3 A-domain overnight at 4°C in 20 mM sodium bicarbonate buffer, pH 9.5. The plate was washed and coated with 2% bovine serum albumin in Hepes buffer, pH 7.4, for 2 h at 25°C. The plate was incubated with 50 nM 125 I-fibrinogen in varied concentrations of calcium, magnesium, or manganese in Hepes buffer, pH 7.4, for 4 h at 25°C. Binding of fibrinogen was determined by direct ␥-counting of the individual plastic wells. Nonspecific binding was determined by the presence of 2.5 M nonlabeled fibrinogen. Assays were performed in triplicates per experiment; multiple experiments were carried out, and the data were averaged.

Expression of Recombinant ␤ 3 A-domain-(109 -352) and
Mutant Forms-To determine the cation selectivity of each metal site, single point mutations were created in a series of fragments corresponding to the ␤ 3 A-domain. A metal ion coordinating aspartic acid was mutated to alanine in the MIDAS at position 119, ADMIDAS at position 126, and LIMBS at position 217. The wild-type ␤ 3 fragment and each mutant were expressed in E. coli as amino-terminal His tag fusion proteins. Purification was performed on a Ni 2ϩ chromatography column in 8 M urea, and the proteins were refolded by dialysis to sequentially reduce the concentration of denaturant to less than 1 mM. The purity of each recombinant protein was analyzed by SDS-PAGE, and a major band at 31 kDa is the appropriate molecular mass of the A-domain fragment (Fig. 1A), and the identity was verified by immunoblotting with antibodies to ␤ 3 (Fig. 1B) or the His tag (not shown). We had previously used circular dichroism to obtain evidence that the ␤ 3 A-domain and various mutant forms were properly folded (33), based on the similarities in their estimated helical content to one another and to that predicted from the crystal structure of ␣ v ␤ 3 (25). We also have analyzed the folding of the ␤ 3 A-domain fragment by NMR. For this purpose, the bacteria expressing the fragment were grown in media containing 15 N and then purified and renatured as described above. The chemical shift dispersion of the heteronuclear single quantum coherence spectrum, collected at 25°C in a 600-MHz instrument, was indicative of a folded protein. Such spectra raise the possibility that NMR ultimately could be used to solve the detailed structures of the various ␤ 3 A-domains.
Equilibrium Dialysis-Equilibrium dialysis experiments were performed to test the specificities of the ion-binding sites in the ␤ 3 A-domain. In preparing the recombinant ␤ 3 A-domains for these analyses, the fragments were treated with EGTA, and the concentration of Ca 2ϩ was reduced to less than 1 nM as described by Rivas and Gonzalez-Rodriguez (35). Ca 2ϩ binding to the ␤ 3 A-domains was measured in the concentration range of 0.1 to 752 M, using a radioactive tracer to monitor binding. Inspection of the Ca 2ϩ -binding isotherms ( Fig.  2A) provides clear evidence for two Ca 2ϩ -binding sites in the wild-type ␤ 3 A-domain and the mutant in which the LIMBS had been mutated (D217A), and one site in the mutants in  which the MIDAS (D119A) and ADMIDAS (D126A) had been inactivated. Similar studies were performed with Mn 2ϩ in the range of 0.1 to 100 M, again using a radioactive tracer to monitor binding (Fig. 2B). With this divalent cation, three binding sites were evident in the wild-type ␤ 3 A-domain and mutation of the LIMBS, MIDAS, and ADMIDAS reduced the number of sites in each variant ␤ 3 A-domain to two.
The data in Fig. 2 were further analyzed. Each isotherm for the two cations fit well to a single site binding model described where B is the ratio of Ca 2ϩ or Mn 2ϩ bound (micromoles) to micromoles of the ␤ 3 A-domain or its mutants; n refers to the number of binding sites; K is the equilibrium association constant, and [L] is the free Ca 2ϩ or Mn 2ϩ concentration. The best fit curves of the data points did not improve statistically when equations for two site models were used. The data obtained from this analysis are summarized in Table 1 and confirm that wild-type ␤ 3 A-domain and mutant D217A bound two Ca 2ϩ ions, although D126A and D119A mutants have single Ca 2ϩ -binding sites. Analysis of the Mn 2ϩ -binding isotherms verified that the wildtype ␤ 3 A-domain has three Mn 2ϩ -binding sites, and each of the three mutant fragments has only two Mn 2ϩ -binding sites (Table 1). Thus, the wild-type ␤ 3 A-domain contains three Mn 2ϩ -and two Ca 2ϩ -binding sites. The Mn 2ϩ -selective site must be the LIMBS because disabling this site had no effect on Ca 2ϩ binding but did reduce the Mn 2ϩ -binding sites from three to two. The affinities of the sites derived by nonlinear regression analysis of the binding isotherms are also summarized in Table  1. The affinities of the wild-type and D217A fragments for Ca 2ϩ are similar, consistent with the inability of the LIMBS to bind Ca 2ϩ although the D119A and D126A had lower affinities. The binding affinities of the wild-type and mutant ␤ 3 A-domains for Mn 2ϩ are similar and higher than for Ca 2ϩ .
Additional equilibrium dialysis experiments were performed to determine the specificity of the cation-binding sites for Mg 2ϩ . The ␤ 3 A-domains were incubated with 500 M 45 Ca 2ϩ in the presence of variable concentrations of Mg 2ϩ (0 -100 mM). As shown in Fig. 3A, of the two Ca 2ϩ -binding sites in the wild-type ␤ 3 A-domain, Mg 2ϩ could only displace Ca 2ϩ from one of these two sites. To determine which of the two Ca 2ϩbinding sites does not bind Mg 2ϩ , similar experiments were performed with the MIDAS and ADMIDAS mutants, which only have a single Ca 2ϩ -binding site. The one Ca 2ϩ bound to the ADMIDAS D126A mutant was displaced by Mg 2ϩ . In contrast, Mg 2ϩ was unable to displace the Ca 2ϩ ion bound to the MIDAS D119A mutant. Hence, the ADMIDAS has specificity for Ca 2ϩ as well as for Mn 2ϩ but not for Mg 2ϩ .
To determine the specificity of the cation-binding sites for Mn 2ϩ , additional equilibrium dialysis experiments were performed. The ␤ 3 A-domains were incubated with 500 M 45 Ca 2ϩ in the presence of Mn 2ϩ ranging from 0 to 1 mM. As shown in Fig. 3B, Mn 2ϩ competed with Ca 2ϩ for both Ca 2ϩ -binding sites. Similarly, to study the effect of Ca 2ϩ on Mn 2ϩ binding, equilibrium dialysis was performed in the presence of 45 MnCl 2 (50 M) sufficient to saturate cation-binding sites in the ␤ 3 A-domain  fragments, and serial dilutions of Ca 2ϩ in the range of 0 to 10 mM were added. As depicted in Fig. 3C, Ca 2ϩ was able to replace only two of the three bound Mn 2ϩ . Fibrinogen Recognition by Intact Receptors, ␣ IIb ␤ 3 , ␣ v ␤ 3 , and the ␤ 3 A-domain by Surface Plasmon Resonance (SPR)-SPR was used to examine and quantify the interactions of fibrinogen with the two ␤ 3 integrins and the ␤ 3 A-domain fragments in real time. In the initial experiments, fibrinogen at various concentrations was injected over the ␣ IIb ␤ 3 immobilized onto the biosensor chip by amine coupling, and the binding kinetics were measured. The dissociation constant (K D ), calculated from the measured k on and k off rate constants, was 3.3 ϫ 10 Ϫ9 M in the presence of 1 mM Ca 2ϩ ϩ 1 mM Mg 2ϩ (see Table 2). This value is consistent with estimates of the K D of this interaction by SPR (36) and other methods (27,37). A reduction in Ca 2ϩ concentration to 0.1 mM or an increase to 3.0 mM led to a modest decrease in fibrinogen binding affinity compared with the 1.0 mM Ca 2ϩ condition. When interaction of fibrinogen with ␣ v ␤ 3 was tested by the same approach, it showed lower binding affinity than that for ␣ IIb ␤ 3 at 1.0 mM Ca 2ϩ ( Table 2). As described before (28,38), the binding of fibrinogen to ␣ v ␤ 3 was dependent upon Ca 2ϩ concentration. Its affinity for fibrinogen was maximal at 0.1 mM Ca 2ϩ and was reduced significantly at 3.0 mM Ca 2ϩ ( Table 2). The inhibitory effect of increasing Ca 2ϩ on fibrinogen binding to ␣ v ␤ 3 was due primarily to a decrease in the k on .
Next, we performed SPR experiments with the ␤ 3 A-domain immobilized onto the biosensor chip and various concentrations of fibrinogen as the solute at either low (0.1 mM) or high (3.0 mM) Ca 2ϩ concentrations. For simplicity, data are presented at a single fibrinogen concentration in Fig. 4. As fibrinogen was injected over the immobilized A-domain, the response (RU) decreased when the calcium concentration was increased from 0.1 to 3.0 mM. The dissociation constant (K D ), derived from the rate constants of dissociation and association, was calculated at five different fibrinogen concentrations. The K D was 2.1 ϫ 10 Ϫ7 M at 0.1 mM Ca 2ϩ and decreased by approximately 1 order of magnitude to K D ϭ 2.7 ϫ 10 Ϫ6 M in the presence of 1.0 mM Ca 2ϩ . When the binding affinity was measured at 3.0 mM Ca 2ϩ , the K D was further decreased to 6.9 ϫ 10 Ϫ5 M. The extent of fibrinogen binding at 3.0 mM Ca 2ϩ was similar to the binding observed in the presence of EDTA (data not shown). The change in K D with increasing Ca 2ϩ arose from a decrease in the k on rate constant for fibrinogen binding to the ␤ 3 A-domain as was observed with ␣ v ␤ 3 .
When similar SPR experiments were performed in the presence of either Mg 2ϩ or Mn 2ϩ , no interpretable results were obtained. To examine the effects of divalent cations on fibrinogen binding to the ␤ 3 A-domain, a microtiter plate assay was used, similar in format to that described previously (33). The ␤ 3 A-domain fragment was bound to wells of microtiter plates, and the specific binding of 125 I-fibrinogen (inhibitable by excess nonlabeled fibrinogen) was determined. In this assay, the binding of fibrinogen to the ␤ 3 A-domain was cation-dependent, analogous to both integrins ␣ IIb ␤ 3 and ␣ v ␤ 3 . Similar to ␣ v ␤ 3 and the SPR experiments with the ␤ 3 A-domain fragment, the interaction was inhibited by high levels of Ca 2ϩ (Fig. 5). Fibrinogen binding to the ␤ 3 A-domain was observed in the presence of Mg 2ϩ or Mn 2ϩ at varied concentrations (Fig. 5) and was not suppressed at higher levels of these cations. Collectively, the microtiter plate and SPR experiments show that the ␤ 3 A-domain fragment can bind fibrinogen in a divalent cation-dependent manner and that the interaction is suppressed by higher Ca 2ϩ concentrations.
Fibrinogen Binding to Mutant ␤ 3 A-domains-Representative SPR response curves of mutant and wild-type ␤ 3 A-domain fragments are shown in the presence of 0.1 mM Ca 2ϩ in Fig. 6A. The mutants have negligible fibrinogen binding when compared with wild-type ␤ 3 A-domain fragment. Thus, binding of fibrinogen to each mutant with a disabled metal-binding site was essentially abolished. When fibrinogen binding to the

TABLE 2
Binding constants for the interaction of fibrinogen with ␣ IIb ␤ 3 and ␣ v ␤ 3 measured by SPR ␣ IIb ␤ 3 or ␣ v ␤ 3 were covalently immobilized onto BIAcore CM5 sensor chips. Fibrinogen was injected at 25°C, and the binding was followed over time as the change in RU in the presence of increasing concentrations of calcium ions. The k on and k off values were determined from the association and dissociation phases, respectively. The apparent K D is calculated from the ratio of k off /k on . mutants was tested in the presence of 3.0 mM Ca 2ϩ , the response curves remained flat (maximum RU of 10), although that of the wild-type ␤ 3 A-domain yielded an RU of ϳ100 (Fig.  6B). Thus, binding of fibrinogen to mutant ␤ 3 A-domains was not supported at either low or high Ca 2ϩ . When similar binding experiments were performed using a microtiter plate assay, specific fibrinogen binding to the mutant ␤ 3 A-domains also could not be detected. Therefore, the microtiter plate data corroborate the SPR findings that no fibrinogen binding was detected with the A-domain mutants.

DISCUSSION
In this study, we have sought to understand the cation specificity of the metal-binding sites in the ␤ 3 A-domain and to determine the contribution of each site to ligand binding. Although crystal structures are available and provide detailed insights as to how ligand and cation engage the ␤ 3 A-domain, they do not show the specificity of each cation-binding site and do not clarify how these sites regulate ligand binding. Hence, we have expressed a wild-type ␤ 3 A-domain fragment, as well as fragments in which each of the three cation-binding sites had been inactivated by a single point mutation. Our major findings in analyzing these fragments are as follows: 1) the cation specificity of the MIDAS, ADMIDAS, and LIMBS sites within the ␤ 3 A-domain are not identical; 2) in addition to cation binding, the isolated ␤ 3 A-domain is capable of binding fibrinogen with properties consistent with those of intact ␣ v ␤ 3 ; and 3) mutation of any of the three metal sites has profound effects on the ligand binding capacity of the isolated ␤ 3 A-domain.
The fragments used in this study were designed to correspond to the boundaries of the ␤ 3 A-domain. Evidence that our renaturation protocol led to appropriate folding is supported by our prior CD analyses of related ␤ 3 A-domain fragments, which indicated that their helical content was similar to that predicted from the crystal structure (25), and NMR analyses, which demonstrated that our ␤ 3 A-domain fragment was indeed folded. Also of relevance to this point, the number of divalent cations bound to each ␤ 3 A-domain fragment is consistent with the published crystal structures of this region (15,16,25). Formation of each of the three metal-binding sites requires a complex fold, supporting our contention that the various fragments had a native conformation.
Our equilibrium dialysis experiments indicated that all three metal-binding sites present in the ␤ 3 A-domain are functional in the absence of the entire ␣ subunit and other domains of the ␤ 3 subunit. The three cation-binding sites also appear to be functionally independent. Disabling any single site did not alter cation binding to a second site, even though the MIDAS and LIMBS share Glu-220, whereas the MIDAS and ADMIDAS share Ser-123 and Asp-251 as part of their cation coordinating constellation. Overall, the affinity of the ␤ 3 A-domain was higher for Mn 2ϩ than for Ca 2ϩ , and this preference was maintained even when the Mn 2ϩ -specific LIMBS was disabled by mutation. The cation specificity of the three metal-binding sites was as follows: the MIDAS could be occupied by all three divalent cations, but its affinity for Mn 2ϩ was greater than for Ca 2ϩ . The LIMBS is a functional Mn 2ϩ -binding site and cannot bind Ca 2ϩ with high affinity. Our data are in agreement with the crystal structures (15,25) in which the LIMBS was occupied by Mn 2ϩ but not by Ca 2ϩ . The MIDAS and ADMIDAS can bind either Ca 2ϩ or Mn 2ϩ . Although the ADMIDAS can bind Ca 2ϩ or Mn 2ϩ , it cannot engage Mg 2ϩ . The latter specificity identifies this site as the "I" site, which was previously defined functionally by Smith and co-workers (28,38) as a metal-binding site that exhibits selectivity for Ca 2ϩ , and its occupancy by this cation suppresses fibrinogen binding to ␣ v ␤ 3 .  Consistent with our previous observations made in a microtiter plate assay system with a related fragment (33), the ␤ 3 A-domain is able to bind fibrinogen. Our SPR experiments in this study verify the ligand binding potential of this fragment and document the cation binding dependence of this interaction. The interaction was optimally supported by low Ca 2ϩ and suppressed by high Ca 2ϩ . This behavior is similar to the effect of Ca 2ϩ on fibrinogen binding to ␣ v ␤ 3 observed in several independent studies (29) and verified by SPR in this study. Based on comparison of the K D values, the contribution of the ␤ 3 A-domain fragment to the affinity for fibrinogen at the same Ca 2ϩ condition was 2 orders of magnitude lower than that exhibited by ␣ v ␤ 3 , indicating a substantial contribution of the domain to the overall affinity of the interaction. The influence of high Ca 2ϩ on fibrinogen binding to ␣ v ␤ 3 was to increase the K D by ϳ2 orders of magnitude, relative to that observed at low Ca 2ϩ primarily by decreasing the association rate, k on . High Ca 2ϩ also increased the K D value of fibrinogen binding to the ␤ 3 A-domain by ϳ2 orders of magnitude through a reduction in the k on rate constant. The suppressive effect of high Ca 2ϩ was also observed in microtiter plate assays, a system that allowed us to demonstrate that this effect was specific for Ca 2ϩ and not observed with increasing Mg 2ϩ or Mn 2ϩ concentrations. Hence, our data and those of others (38 -41) indicate that this phenomenon is dependent on occupancy of the Ca 2ϩ -specific site, the ADMIDAS. The crystal structure of ␣ v ␤ 3 with a bound RGD ligand (15), which is a sequence through which fibrinogen binds to this integrin, suggests that this effect may be due to change in binding surface electrostatics. However, extrapolation of this explanation to the suppressive effect of Ca 2ϩ on fibrinogen binding may not be appropriate because this effect of Ca 2ϩ is not observed with vitronectin, another RGD ligand of ␣ v ␤ 3 (28,29). The differences in the behavior of these two ligands may arise from differences in their affinities for the receptor or from the effects of the ADMIDAS on additional and necessary contacts with fibrinogen and the ␤ 3 A-domain. By equilibrium dialysis, saturation is reached at more than 0.5 mM calcium (Fig. 2). During equilibration dialysis, measurements are done in the presence of excess free cations, and there is an equilibrium between bound and free calcium ions. Our microtiter plate assay involves a washing procedure, which might cause dissociation of calcium bound to the ␤ 3 A-domain, and higher calcium concentrations may be required to keep the binding sites fully occupied. However, we also see effects of high calcium in SPR experiments, which do not involve a washing step. As possible explanations, the metal-binding sites being analyzed may need to be fully saturated to obtain the suppressive effects, or the suppressive effects may be mediated by unrelated low affinity calcium-binding sites, although such sites were not reported in the crystal structures (15,16,25). We also cannot exclude the possibility that the effect of calcium may be on fibrinogen because fibrinogen does bind calcium (42) and Ca 2ϩ does exert unusual effects on fibrinogen (43). However, suppressive effects of Ca 2ϩ are observed with ligands of ␣ v ␤ 3 other than fibrinogen; recognition of osteopontin (44) is also suppressed by high Ca 2ϩ , and it has been suggested that the effects of Ca 2ϩ concentrations on recognition of this and other bone matrix ligands may regulate the contribution of ␣ v ␤ 3 to bone resorption (45). Ca 2ϩ suppression is also observed with other integrin family members (45,46).
One objective of this study was to determine how mutations in each of the metal sites affect ligand binding to the isolated ␤ 3 A-domain fragment. When the aspartic acid at position 119 in the MIDAS was mutated to an alanine, ligand binding was abolished. This is not surprising because it is known that the MIDAS site is directly involved in ligand contact. A naturally occurring mutation at this position of aspartic acid to tyrosine disables the ligand binding function of ␣ IIb ␤ 3 and leads to a platelet bleeding disorder, Glanzmann thrombasthenia (47). When the ADMIDAS Asp-126 was mutated to alanine, ligand binding was again lost. In studies of ␤ 1 and ␤ 7 integrins, mutations of the ADMIDAS have been shown to either inhibit ligand binding (41) or reduce the calcium inhibition of ligand binding (39,40). When the aspartic acid in the LIMBS was mutated, ligand binding was also diminished, and this observation is also consistent with the effect of a naturally occurring mutation (D217V), which leads to a nonfunctional ␣ IIb ␤ 3 and Glanzmann thrombasthenia (48). How the ADMIDAS and LIMBS influence ligand binding is not clear. From the study of ␣ 5 ␤ 1 , Mould et al. (41) suggested that the ADMIDAS was not involved directly in ligand contact but rather stabilized the active conformation of the integrin needed for productive ligand binding. Studies by Chen et al. (39,40) on ␣ 4 ␤ 7 suggested counterbalancing roles of the ADMIDAS and LIMBS; the ADMIDAS was a negative regulator of calcium required for leukocyte rolling, although the LIMBS was a positive regulator of calcium required for firm adhesion. It was postulated that these effects arose by the influence of these sites on the conformation of the MIDAS because the LIMBS and the ADMIDAS share residues in common with the MIDAS, one residue with the LIMBS and two with the ADMIDAS. The metal sites may be independent when binding cations; however, when binding fibrinogen there may be some requisite communication between the metalbinding sites. Hence, when the LIMBS and MIDAS are mutated, the loss of fibrinogen binding activity may not be due only to the lack of metal but rather to the repulsive charges no longer neutralized by cation in the unoccupied metal sites. A controversy remains as to whether the LIMBS binds Ca 2ϩ . Our data suggest that LIMBS does not bind Ca 2ϩ , which was also reported by Xiong et al. (25) in the crystal structure of ␣ v ␤ 3 . The LIMBS was not occupied in the presence of Ca 2ϩ and was only occupied by Mn 2ϩ when RGD ligand was present (15). However, the crystal structures of ␣ IIb ␤ 3 with bound ligands show a Ca 2ϩ bound at the LIMBS (16). It remains a formal possibility that the LIMBS might be able to bind Ca 2ϩ but only in the presence of ligand.
When the MIDAS or ADMIDAS was inactivated by mutation, the affinity of the remaining site for Ca 2ϩ was reduced. This is consistent with some communication between these sites and regulation of the MIDAS function by the ADMIDAS. Even though the LIMBS binds Mn 2ϩ and not Ca 2ϩ , this site may still control the function of the MIDAS cation. Xiong et al. (15) speculated the role of the LIMBS was to stabilize and orientate Glu-220 for structure stability of the ligand binding area based on the ␣ v ␤ 3 crystal structures. Even though our data indicate that the cation binding function of the LIMBS does not require ligand, such stabilization of a ligand-permissive conformation is still a viable explanation for its role.
In summary, we have determined the cation specificity of the metal-binding sites in the ␤ 3 A-domain and their contribution to ligand binding. Although much progress has been made in solving high resolution structures of integrins ␣ v ␤ 3 and ␣ IIb ␤ 3 , these structures have provided snapshots of occupied cationbinding sites in the presence or absence of ligands and have not established the precise specificities of these sites. Our studies show that these metal-binding sites have distinct cation preferences and that occupancy of each of these sites regulates ligand binding. Unresolved is the basis for the differential effects of cations on the ligand binding functions of the two ␤ 3 integrins. With the insights into the specificities of the cation-binding sites gained from this study, mutational analyses within the context of the intact receptors will become more interpretable.