Ligand Binding to Integrin αIIbβ3 Is Dependent on a MIDAS-like Domain in the β3 Subunit

Substitution of β3 residue Asp119, Ser121, or Ser123 results in a loss of the ligand binding function of integrin αIIbβ3. Homologous residues in other integrin β subunits are similarly critical for ligand binding function. This DXSXS motif is also present in the I domain of certain integrin α subunits, where it constitutes a portion of the unique metal ion-dependent adhesion site (MIDAS). In this report, we have utilized the crystal structure of the recombinant αM I domain to produce a three-dimensional model of the homologous region in the integrin β3 subunit. We performed mutagenesis of candidate amino acid residues predicted from this model to be involved in cation coordination and ligand binding. We report the identification of Asp217 and Glu220 as residues essential for the ligand binding function of αIIbβ3. Alanine substitution of these residues did not affect receptor expression but abolished the binding of activation-dependent (PAC1) and -independent (OPG2) ligand mimetic antibodies. In our proposed model, β3 Asp217 is analogous to a metal-coordinating residue in the αM MIDAS domain, while Glu220 does not correspond to a functional MIDAS domain residue. Substitution of the highly conserved β3 residue Thr197 corresponding to a critical MIDAS metal-coordinating Thr residue did not affect ligand binding function, suggesting that this region of β3 adopts a structure that is very similar to but not identical to that of the MIDAS domain. These data support a functional linkage between these two sequences and further define a common feature of ligand binding to integrins.

Substitution of ␤ 3 residue Asp 119 , Ser 121 , or Ser 123 results in a loss of the ligand binding function of integrin ␣ IIb ␤ 3 . Homologous residues in other integrin ␤ subunits are similarly critical for ligand binding function. This DXSXS motif is also present in the I domain of certain integrin ␣ subunits, where it constitutes a portion of the unique metal ion-dependent adhesion site (MIDAS). In this report, we have utilized the crystal structure of the recombinant ␣ M I domain to produce a three-dimensional model of the homologous region in the integrin ␤ 3 subunit. We performed mutagenesis of candidate amino acid residues predicted from this model to be involved in cation coordination and ligand binding. We report the identification of Asp 217 and Glu 220 as residues essential for the ligand binding function of ␣ IIb ␤ 3 . Alanine substitution of these residues did not affect receptor expression but abolished the binding of activation-dependent (PAC1) and -independent (OPG2) ligand mimetic antibodies. In our proposed model, ␤ 3 Asp 217 is analogous to a metal-coordinating residue in the ␣ M MIDAS domain, while Glu 220 does not correspond to a functional MIDAS domain residue. Substitution of the highly conserved ␤ 3 residue Thr 197 corresponding to a critical MIDAS metalcoordinating Thr residue did not affect ligand binding function, suggesting that this region of ␤ 3 adopts a structure that is very similar to but not identical to that of the MIDAS domain. These data support a functional linkage between these two sequences and further define a common feature of ligand binding to integrins.
Platelet adherence to components of the subendothelial matrix, to other platelets, and to other cells plays a fundamental role in normal hemostasis. These platelet-adhesive interactions are mediated in large part through the major platelet integrin ␣ IIb ␤ 3 (glycoprotein IIb-IIIa). Elucidation of the mechanism of ligand recognition is central to the understanding of ␣ IIb ␤ 3 receptor function and is the first step toward modulation of platelet function and the development of potential antithrombotic therapies. Moreover, since ␣ IIb ␤ 3 has served as the prototype integrin, determination of the molecular basis of ligand recognition may provide additional insights into integrin-ligand interactions in general.
Significant progress has been made in the identification of potential ligand binding sites in ␣ IIb ␤ 3 . Previous studies indicate that the minimal ligand binding structures are located in the amino-terminal half of ␣ IIb ␤ 3 (1)(2)(3)(4)(5)(6). The identification of more discrete regions within each of the subunits has benefited from the use of small ligand mimetic peptides as probes of ligand binding sites. Peptides derived from the carboxyl terminus of the fibrinogen ␥ chain cross-link to a region of ␣ IIb defined by residues Ala 294 -Met 314 (7). This location is particularly noteworthy, since it encompasses the second divalent cation-binding repeat of ␣ IIb (8). Peptides derived from this sequence inhibit ligand binding and directly bind fibrinogen in a cation-dependent manner (9), supporting a role in ligand binding.
Two distinct regions within ␤ 3 have been implicated in ligand binding function of the receptor. Arg-Gly-Asp (RGD)containing peptides cross-link to a discrete region of ␤ 3 defined by residues Asp 109 -Glu 171 (10). In addition, monoclonal antibodies (mAbs) 1 specific for this region inhibit ligand binding and platelet aggregation (11,12). Strong evidence for a direct interaction of this region in ligand binding comes from genetic analysis of a variant ␣ IIb ␤ 3 characterized by complete loss of ligand binding function (13). The molecular basis for this defect is substitution of Asp 119 3 Tyr in mature ␤ 3 (14). This Asp residue is absolutely conserved in all the integrin ␤ subunits, suggesting that this residue may play a role in ligand binding to all integrins. This hypothesis has been substantiated by the observation that substitution at residues homologous to ␤ 3 Asp 119 in other integrin ␤ subunits abrogates ligand binding function (15)(16)(17). Moreover, substitution at this residue exerts dominant negative effects blocking the binding of RGD-dependent ligands as well as RGD-independent ligands, further suggesting that this residue may play a role in a common mechanism of ligand binding.
A second potential ligand-interactive site in ␤ 3 is defined by residues Ser 211 -Gly 222 . Peptides corresponding to this sequence bound fibrinogen and blocked its binding to ␣ IIb ␤ 3 (18). Antibodies directed against this peptide also blocked the binding of adhesive proteins to purified receptor (18). In addition, two natural receptor variants characterized by loss of ligand binding function contain substitutions at Arg 214 (19,20), further implicating this region in receptor function.
Despite the identification of these putative ligand contact points, the relationship, if any, between these discrete sites and the precise molecular mechanism by which these sites mediate ligand binding remains to be determined. A paradigm of ligand binding to all integrins is an absolute dependence upon divalent cations. Thus it is significant that the Asp 119 3 Tyr mutation in ␤ 3 also alters the conformation of ␣ IIb ␤ 3 in a manner consistent with loss of bound divalent cation (13). This observation led to the hypothesis that this region may be part of a divalent cation binding site (14). Indeed, a synthetic ␤ 3 peptide corresponding to ␤ 3 residues Met 118 -Ile 131 , directly binds the luminescent calcium analog terbium (21). Substitution of Asp 119 by alanine in this peptide substantially reduces the terbium/peptide interaction, providing support for a role of this region in both ligand recognition and cation binding. Mutational analysis of residues proximal to Asp 119 assigned critical functional roles to Ser 121 and Ser 123 in the ligand binding function of ␣ IIb ␤ 3 (22). Similarly, a role in ligand binding function has been reported for a corresponding Ser residue in ␤ 2 (17). Together Asp 119 , Ser 121 , and Ser 123 of ␤ 3 compose a DX-SXS sequence that is absolutely conserved in all the integrin ␤ subunits. The divalent cation dependence of integrin function together with the high degree of conservation of these functionally significant residues forms the basis of the hypothesis that ligands interact with divalent cations bound to this highly conserved site in the ␤ subunits (14). The presence of critical oxygenated residues in integrin ligands (14,(23)(24)(25)(26)(27) and the displacement of divalent cations following ligand binding (21,28) support this view.
The essential DXSXS motif in the integrin ␤ subunits is also highly conserved in the inserted (I) domain present in six of the integrin ␣ subunits. These ϳ200 amino acid residue I domains are homologous to the A domain of von Willebrand factor and are critical for both cation and ligand binding to I domaincontaining integrins (29). High resolution crystal structures of isolated recombinant ␣ M (30) and ␣ L (31) I domains clearly establish that this DXSXS sequence constitutes a portion of a unique metal coordination site designated the metal ion-dependent adhesion site (MIDAS) (30). The conservation of the metal-coordinating consensus sequence DXSXS and the similarity of hydropathy plots between this region of ␤ 3 and the ␣ M I domain suggested that this conserved region of the integrin ␤ subunits may also adopt an I domain fold (30,32). To test this hypothesis, we have utilized the crystal structure of the recombinant ␣ M I domain to model the homologous region in the integrin ␤ 3 subunit and performed mutagenesis of candidate amino acid residues predicted to be involved in cation coordination and ligand binding. We report the identification of Asp 217 and Glu 220 as residues essential for ligand binding function of ␣ IIb ␤ 3 . Asp 217 is analogous to a metal-coordinating residue in the I domain MIDAS motif. However, Glu 220 does not correspond to a functional ␣ M MIDAS domain residue, indicating that the two regions may adopt a similar but not identical structure. These results establish a functional linkage between these two sequences and further define a novel cationbinding motif essential for integrin receptor function.
Mutagenesis-Expression constructs encoding wild type ␣ IIb , CD2b, and wild type ␤ 3 , CD3a, have been previously described (36). A 2.6kilobase pair fragment of ␤ 3 encoding the entire coding sequence and a portion of the 3Ј-untranslated sequence was removed from CD3a by digestion with HindIII and DraI and ligated into the expression vector pcDNA3 (Invitrogen, La Jolla, CA) that had been digested with XhoI, blunt-ended with T4 polymerase, and subsequently digested with Hin-dIII. The resulting construct was designated pc3A. Site-directed mutagenesis of selected ␤ 3 residues was performed by splice overlap extension as described (38). Polymerase chain reaction-generated fragments containing mutations were digested with Kpn and BstXI, and the resulting 448-base pair fragment was gel-purified and ligated to Kpn/BstXI-digested pc3A. All ligated polymerase chain reaction fragments were sequenced in their entirety to verify the introduction of the mutation and the absence of any other substitutions. DNA constructs were purified on Qiagen columns (Chatsworth, CA) prior to transfection.
Cell Culture and Transfection-Chinese hamster ovary cells obtained from the American Type Culture Collection (Rockville, MD) were maintained in Dulbecco's modified Eagle's medium (BioWhittaker, Walkersville, MD) containing 10% fetal bovine serum, 1% nonessential amino acids, 2 mM glutamine, 100 units/ml penicillin, and 100 mg/ml streptomycin. Cells were transiently transfected using Lipofectamine ® (Life Technologies Inc.) as described previously (39). Functional analysis of cells was routinely performed 48 h after transfection.
Flow Cytometry-Surface expression of transfected integrins was analyzed by flow cytometry as described (40). PAC1 binding was analyzed by two-color flow cytometry as described in detail (39). PAC1 binding was analyzed only on the subset of cells positive for ␣ IIb ␤ 3 expression that were gated utilizing biotinylated D57. The binding of OPG2 was performed essentially as described for PAC1 binding, except biotinylated OPG2 was used and the activating mAb anti-LIBS6 was omitted. OPG2 binding was analyzed on the subset of cells positive for ␣ IIb ␤ 3 expression that were gated utilizing fluorescein isothiocyanateconjugated D57. Analysis was performed on a FACScan (Becton Dickinson, Mountain View, CA).
Molecular Modeling-A set of models for the potential MIDAS domain of ␤ 3 were produced using homology modeling based on the suggested topological similarity between the integrin ␣ M I domain and the integrin ␤ subunits (30). Secondary structure prediction was performed on the amino acid sequence of ␤ 3 Tyr 110 -Leu 294 using the program PHD (41). The sequence alignment used for homology modeling was derived in two stages. First, the amino acid sequences of those proteins with similar sequence to the ␣ M I domain found using BLAST (42) (26 in total) were multiply aligned against each other using Clustal W (43). Similarly, proteins with sequence similar to ␤ 3 (30 in total) were multiply aligned against each other. Second, these two sequence alignments were aligned against each other as two "profiles" within Clustal W. Thus, the integrity of each of the two original alignments was maintained in combining them to obtain the final alignment. This alignment was then manually adjusted so that 1) ␣ M Thr 209 , which has a hydroxyl group coordinating with Mg 2ϩ , corresponded with ␤ 3 Thr 197 that is completely conserved among the integrin ␤ subunits; 2) ␣ M Asp 242 , which coordinates the Mg 2ϩ ion via a water molecule was aligned with the totally conserved ␤ 3 Asp 217 ; 3) wherever possible, no insertion or deletion occurred within the crystallographically determined secondary structural elements of ␣ M ; and 4) there was reasonable agreement between the predicted secondary structure of ␤ 3 and the determined structure of ␣ M .
Homology modeling was performed using the program MODELLER (44). Two disulfide bonds are predicted in ␤ 3 Tyr 110 -Leu 294 (45). These two bonds were restrained to be present in the model and are indicated in Fig. 1. A set of 10 models was generated. Discrepancies between the sequence alignment and the resulting models were identified by consistent bond length and bond angle violations in the corresponding regions of the models, and the sequence alignment was adjusted to remove them. The alignment was modified in an iterative manner until no consistent bond length and bond angle violations were observed across the set of 10 models. Further adjustment of the alignment was performed to minimize the number of buried charge residues not involved in salt bridges. The final alignment is shown in Fig. 1.

The Conserved Region of ␤ 3 Adopts an I Domain-like Fold-
Ligand binding to all integrins is dependent upon divalent cations. Substitution of Asp 119 in ␤ 3 abolishes ligand binding function of ␤ 3 integrin and alters a divalent cation conformation of ␣ IIb ␤ 3 (14). The similarity of hydropathy plots between this region of ␤ 3 and the region of the I domain containing the MIDAS motif suggested that this region of ␤ 3 may adopt a similar fold. To test the prediction that this region has a structure functionally equivalent to a MIDAS domain, we built a three-dimensional model of the region of ␤ 3 defined by Tyr 110 -Leu 294 . The proposed ␤ 3 structure has secondary structural elements very similar to those in the I domain with a central ␤-sheet composed of five parallel ␤ strands and one antiparallel ␤ strand surrounded by seven ␣-helices (Fig. 2). The proposed model brings two pairs of cysteine residues into close proximity, where they can form the predicted disulfide bonds (45). In general, the model places hydrophilic residues on the outside and hydrophobic residues on the inside of the structure. One exception is in the ␤D strand, immediately preceding Asp 217 . Here two arginine residues, Arg 214 and Arg 216 (that are not conserved among the integrin ␤ subunits) appear to be buried, although the model provides salt bridge partners (Glu 234 and Asp 259 , respectively) to neutralize them. The buried position of Arg 214 might help explain the sensitivity of this site to substitutions (19,20). The cluster of residues Asp 119 , Ser 121 , and Ser 123 that have been previously implicated in ligand binding are located at the top of the structure in a loop between ␤A and ␣1 and form part of the potential MIDAS-like motif (Fig. 2).
Asp 217 and Glu 220 Are Required for Ligand Binding Function-Metal coordination in the ␣ M I domain is provided in part by the DXSXS sequence and two noncontiguous amino acid residues (Thr 209 and Asp 242 ) located carboxyl-terminal to the DXSXS motif (30). It is likely that oxygenated residues critical for metal coordination would be conserved in all the integrin ␤ subunits. Alignment of the corresponding regions of the known integrin ␤ subunits served to identify a number of conserved candidate residues (Fig. 3). To identify residues essential for ligand binding, we performed alanine mutagenesis and examined the effect of these substitutions on ligand binding function following transient co-transfection of Chinese hamster ovary cells with a mutant ␤ subunit and wild type ␣ IIb .
To determine the effect of these substitutions on the ligand binding function of the expressed mutant receptors, we examined the capacity of the transfectants to bind the mAb PAC1 by flow cytometry. PAC1 is a murine IgM k that binds specifically to the activated conformation of ␣ IIb ␤ 3 , and its binding is blocked by macromolecular ligands as well as ligand mimetic peptides (33,46,47). PAC1 also fails to bind to ligand bindingdefective mutants (22,48). The binding of PAC1 to transfected cells was assayed by flow cytometry in the absence or presence of mAb anti-LIBS6 that acts directly upon ␣ IIb ␤ 3 , provoking high affinity ligand binding function (49). Cells expressing wild type ␣ IIb ␤ 3 bound PAC1 in the presence of the activating mAb anti-LIBS6 (Fig. 4). The binding of mAb PAC1 was specific, since it was completely blocked by the ␣ IIb ␤ 3 -specific peptidomimetic Ro 43-5054. Alanine substitution of a number of highly conserved oxygenated residues in ␤ 3 did not affect heterodimer formation or receptor processing, since all mutants were expressed on the cell surface as assayed with the anti-␣ IIb ␤ 3 -specific mAb D57 (Table I). However, alanine substitution of Asp 217 or Glu 220 resulted in a loss of PAC1 binding (Fig.  4). These two mutants readily bound anti-LIBS6 as assayed by flow cytometry (data not shown). Therefore, the lack of PAC1 binding was not due to failure of anti-LIBS6 to bind to these mutants. Furthermore, the loss of ligand binding to these two mutants was not a generalized effect, since alanine substitution of a number of other highly conserved residues with oxygenated side chains had no effect on ligand binding function (Table I).
To substantiate the results obtained with mAb PAC1, the effect of these substitutions on the binding of another ligand mimetic mAb, OPG2 was also examined by flow cytometry. OPG2 is a murine IgG 1k that blocks the binding of adhesive protein ligands to ␣ IIb ␤ 3 , and its binding is blocked by RGD peptides (34). However, in contrast to PAC1, the binding of mAb OPG2 is activation-independent. Consistent with the results obtained with PAC1, OPG2 failed to bind to cells expressing the ␤ 3 mutants Asp 217 3 Ala or Glu 220 3 Ala (Fig. 5). Alanine substitution of other conserved residues had no effect on the binding of mAb OPG2 (Table I).
Asp 217 in ␤ 3 is found at the end of ␤D strand in the proposed model and is predicted to be the functional equivalent of Asp 242 , which occupies a similar position in the ␣ M I domain (30). In contrast, ␤ 3 Glu 220 does not appear to have a functional equivalent in the I domain MIDAS motif. The loss of ligand binding function of ␣ IIb ␤ 3 after substitution of ␤ 3 Glu 220 suggested that Glu 220 might be involved in metal coordination. To Sec str x-ray, the crystallographically determined secondary structure of CD11b (30). CD61 Sec str model, the secondary structural elements observed in the lowest energy model of CD61 as determined by PROCHECK (67). CD61 Sec str PHD, the secondary structural elements predicted for CD61 on the basis of its amino acid sequence by the program PHD (41). The names of the secondary structural elements and the positions of the disulfide bonds are shown by Sec str/disulfides. Amino acid sequences shown in lowercase for CD11b were not used to define the positions of the corresponding residues in CD61, since the models showed them not to be topologically equivalent.
investigate this hypothesis, two additional sets of models were generated. In the first set, the OE1 atom of Glu 220 was taken to form a hydrogen bond with a coordinating water molecule; a distance restraint of 2.5-3.3 Å was therefore imposed between these two atoms throughout generation of this model. In the second set, the OE1 atom of Glu 220 was taken to coordinate directly with the metal ion; a distance restraint of 1.8 -2.2 Å was therefore imposed between these two atoms throughout generation of this model. Good quality models could be obtained with Glu 220 interacting with the metal ion either directly or indirectly via a water molecule. This interaction appears to be facilitated by Pro 219 becoming cis (it is trans in the model depicted in Fig. 2 generated without Glu 220 being restrained) and is aided by the conformational freedom of Gly 221 and Gly 222 . These results support a hypothesis that Glu 220 could participate in metal coordination. DISCUSSION We report that alanine mutagenesis of selected amino acids in ␤ 3 that are highly conserved among the integrin ␤ subunits has identified Asp 217 and Glu 220 as residues essential for the ligand binding function of ␣ IIb ␤ 3 . A model of this region of ␤ 3 based on the crystal structure of the recombinant ␣ M I domain predicts that this region may share many of the structural elements of the I domain. This proposed model is consistent with the mutagenesis data and suggests that Asp 217 and Glu 220 are located in the ␤D-␣5 loop, where they can participate in the coordination of a metal ion together with the previously identified Asp 119 , Ser 121 , and Ser 123 . Asp 217 but not Glu 220 is analogous to a metal-coordinating residue in the MIDAS domain, indicating that region adopts a similar but not identical fold to that of the MIDAS motif.
Homology modeling suggests that ␤ 3 Asp 217 is the functional equivalent of ␣ M Asp 242 . In the Mg 2ϩ form of the ␣ M I domain, the carboxylate group of Asp 242 contributes to metal coordination via a hydrogen bond to a coordinating water molecule and through a second hydrogen bond to the hydroxyl side chain of the directly coordinating residue Ser 144 . In the Mn 2ϩ form of the ␣ M I domain, the metal is coordinated slightly differently, since Asp 242 directly coordinates the bound metal (32). Similarly, the Asp 242 homologue in the ␣ L Mn 2ϩ I domain structure, Asp 239 , directly coordinates the bound metal (31). Substitution of ␣ M Asp 242 in the isolated I domain results in loss of both cation binding and ligand binding (29). Furthermore, substitution of Asp 242 in ␣ M and substitutions of the equivalent residues in the I domains of ␣ L (Asp 239 ), ␣ 1 (Asp 253 ), and ␣ 2 (Asp 254 ) all exert dominant negative effects, since each substitution results in the loss of binding of structurally distinct ligands to the intact receptor. Together these results indicate that the participation of this Asp residue in metal coordination is critical for ligand binding. The observed loss of function of ␣ IIb ␤ 3 following substitution of ␤ 3 Asp 217 is consistent with this residue being the functional equivalent of these I domain Asp  residues and strongly suggests that its role in the ligand binding function of ␣ IIb ␤ 3 is due to its participation in cation coordination.
The basis for the requirement of ␤ 3 Glu 220 in the ligand binding function of ␣ IIb ␤ 3 is less obvious. A Glu residue is absolutely conserved at this position in all the integrin ␤ subunits. A Glu residue, separated from the coordinating ␣ M Asp 242 residue by a Gly, is conserved in ␣ L (Glu 241 ) but is less well conserved among other proteins in the A/I domain superfamily (30). The effect of substitution of this Glu residue in the I domain on intact ␣ M ␤ 2 receptor function has not been examined directly by mutagenesis. In the Mg 2ϩ form of the ␣ M I domain, Glu 244 is hydrogen-bonded through its main chain carbonyl to a coordinating water molecule (30). However, one would expect the potential for this indirect main chain interaction with a coordinating water molecule to be the same for Ala as it is for Glu if the structure of these two domains is similar. In this regard, we note that the register at the top of ␤D cannot be assigned with certainty, since ␤ 3 (as well as other integrin ␤ subunits) does not contain the Gly residue that is conserved in each of the integrin I domains. An alternative register makes this loop between ␤D and ␣5 more similar to that in the von Willebrand factor A3 domain 2 that lacks this conserved Gly. This would have the effect of placing Asp 217 one residue further up the loop, and Glu 220 would then be placed in a position where it too could participate in metal coordination either directly or indirectly via a water molecule. Given the inherent limitations of the proposed model, it is also possible that Glu 220 may participate directly in ligand binding in a manner distinct from a potential role in cation coordination. Precise functional assignment of this residue will ultimately require a high resolution crystal structure.
Ligand binding to ␣ L ␤ 2 , ␣ M ␤ 2 , and ␣ 2 ␤ 1 is critically dependent upon a conserved Thr residue within the I domain of the ␣ subunits (50 -53). In the crystal structure of the Mg 2ϩ form of ␣ M , Thr 209 makes a strong bond to the metal via its side chain hydroxyl oxygen (30). In the Mn 2ϩ form of the ␣ M I domain, Thr 209 does not coordinate the metal directly but rather hydrogen-bonds to a coordinating water molecule (32). Similarly, in the Mn 2ϩ form of the ␣ L I domain, the corresponding residue in ␣ L , Thr 206 , is also hydrogen-bonded to a coordinating water molecule (31). In both of these I domains, this critical Thr residue is present on an extended loop between ␣3 and ␣4. Inspection of the corresponding loop region in the ␤ 3 structural model identified two potential residues, Thr 195 and Thr 197 , to be spatially adjacent to the DXSXS sequence. While Thr 195 is not well conserved, Thr 197 is absolutely conserved among all the integrin ␤ subunits. In contrast to the critical role that the conserved Thr residue located at this position plays in ligand binding to I domain-containing integrins, alanine substitution of ␤ 3 Thr 195 or Thr 197 had no effect on ligand binding, since both mutants readily bound the ligand mimetic mAb PAC1, suggesting that metal coordination may be different in these two structures.
The fact that the critical I domain Thr residue coordinates the metal ion differently in the Mg 2ϩ and Mn 2ϩ forms suggests that these two forms may represent two different activation states of the receptor (31,32). Indeed, Mn 2ϩ can stimulate the binding of ligand to ␣ IIb ␤ 3 and a number of other integrins in the absence of other agonists (54 -59). In the present study, PAC1 binding to the expressed mutants was investigated in the presence of the activating mAb anti-LIBS6. Thus, one potential explanation for the observed lack of effect on PAC1 binding to the Thr 197 3 Ala mutant is that this residue may play an essential role in metal coordination in the unactivated conformation of the receptor but is not essential for coordination in the active conformation. However, substitution of Thr 197 or Thr 195 had no effect on the binding of OPG2, which does not require receptor activation. Thus it is unlikely that Thr 197 participates in metal coordination in this structure. The present results cannot rule out that substitution of this Thr residue alone does not destabilize the coordination of bound cation sufficiently to inhibit ligand binding.
The identification of Asp 217 and Glu 220 as putative metalcoordinating residues provides an explanation for previous results implicating this region in ligand binding function (18) and the loss of ligand binding associated with mutations at Arg 214 (19,20). The current model places Arg 214 in the middle of ␤D, which would bury that residue unless the carboxyl-terminal helix (␣7) swings out from the position it occupies in the integrin I domain fold. An altered location of this helix is reasonable given its variable conformation in the integrin I domains. Therefore, while the proposed model does not support a direct role for this residue in metal coordination, the previously described Arg 214 3 Gln or Arg 214 3 Trp substitutions might both be predicted to produce alterations in the local structure, potentially destabilizing metal coordination mediated by Asp 217 . Such a structural alteration may also be the root cause of the impaired stability of the ␣ IIb ␤ 3 heterodimer observed in these mutants (19,20,60).
The current results support a role for this region of ␤ 3 in ligand binding function as a consequence of its adopting a structure very similar to the MIDAS structure present in the I domains. In addition to the lack of a critical threonine residue for ligand binding located carboxyl-terminal to the DXSXS sequence in the ␤ 3 subunit, other significant differences exists in the ligand binding characteristics of I domain-containing integrins compared with ␣ IIb ␤ 3 . First, ligand binding to ␣ IIb ␤ 3 is fully supported by Ca 2ϩ . In contrast, binding of ligands to recombinant I domains or intact I domain-containing integrins requires Mg 2ϩ or Mn 2ϩ and can be inhibited by Ca 2ϩ (29,55,61,62). The functional requirement for the negatively charged Glu 220 but not for the uncharged Thr 197 could explain why Ca 2ϩ , a larger ion with a greater tendency to form ionic bonds, can substitute for Mg 2ϩ in the ligand binding function of ␣ IIb ␤ 3 but not ␣ M ␤ 2 . Second, the results presented here would indicate that I domain integrins potentially possess two MIDAS domains. The significance of two MIDAS domains and their interrelationship in ligand binding has not been addressed, yet mutations of critical residues in either the ␣ or ␤ subunit domains effectively abolish ligand binding to ␤ 2 integrins. Third, it is particularly noteworthy that isolated recombinant I domains can bind ligand directly (29,51,(63)(64)(65)(66). In contrast, high affinity ligand binding to ␣ IIb ␤ 3 requires an intact heterodimer, since individual ␣ IIb or ␤ 3 subunits have not demonstrated ligand binding capacity.
In summary, we conclude that a highly conserved region in the integrin ␤ subunits functions as a MIDAS-like domain and that ␤ 3 residues Asp 217 and Glu 220 play a role in ligand binding function via a likely contribution to a metal coordination sphere. In view of the ubiquitous divalent cation requirement for all integrin receptor function, it is likely that residues homologous to ␤ 3 Asp 217 and Glu 220 in other integrin ␤ subunits will be assigned similar functional roles. The present results and the proposed model do not address the role of ␣ IIb to this structure; however, the cross-linking of ligand mimetic peptides to ␣ IIb substantiates the close proximity of regions of ␣ IIb to bound ligand. Accordingly, this presents the intriguing possibility that a residue located on ␣ IIb may supply a metalcoordinating ligand or interact directly with ligand. Such a possibility might further explain the requirement of the ␣ subunit for ligand binding and ligand recognition specificity in integrins that lack an I domain.