Role of ADMIDAS Cation-binding Site in Ligand Recognition by Integrin α5β1*

Integrin-ligand interactions are regulated in a complex manner by divalent cations, and multiple cation-binding sites are found in both α and β integrin subunits. A key cation-binding site that lies in the β subunit A-domain is known as the metal-ion dependent adhesion site (MIDAS). Recent x-ray crystal structures of integrin αVβ3 have identified a novel cation binding site in this domain, known as the ADMIDAS (adjacent to MIDAS). The role of this novel site in ligand recognition has yet to be elucidated. Using the interaction between α5β1 and fibronectin as a model system, we show that mutation of residues that form the ADMIDAS site inhibits ligand binding but this effect can be partially rescued by the use of activating monoclonal antibodies. The ADMIDAS mutants had decreased expression of activation epitopes recognized by 12G10, 15/7, and HUTS-4, suggesting that the ADMIDAS is important for stabilizing the active conformation of the integrin. Consistent with this suggestion, the ADMIDAS mutations markedly increased the dissociation rate of the integrin-fibronectin complex. Mutation of the ADMIDAS residues also reduced the allosteric inhibition of Mn2+-supported ligand binding by Ca2+, suggesting that the ADMIDAS is a Ca2+-binding site involved in the inhibition of Mn2+-supported ligand binding. Mutations of the ADMIDAS site also perturbed transduction of a conformational change from the MIDAS through the C-terminal helix region of the βA domain to the underlying hybrid domain, implying an important role for this site in receptor signaling.

Integrins constitute a large family of ␣/␤ heterodimeric transmembrane receptors found in all metazoa (1). Cell-matrix and cell-cell interactions mediated by integrins are central to many fundamental biological processes such as embryonic morphogenesis, leukocyte trafficking, and platelet aggregation. Integrins can exist in either active (ligand competent) or inactive conformational states; the equilibrium between these two states is regulated intracellularly by the binding of cytoskeletal and signaling molecules (2,3). Integrin-ligand interactions also require divalent cations and are regulated in a complex manner by changes in the concentrations of these ions. The effects of activation can be mimicked in vitro by cations such as Mn 2ϩ or Mg 2ϩ , whereas Ca 2ϩ typically favors the inactive state. The different effects of these cations are related to their differential abilities to induce the integrin to undergo the shape changes involved in activation (4 -6).
The molecular basis of integrin function has been greatly elucidated by x-ray crystal structures of the extracellular domains of ␣ V ␤ 3 in the unliganded and liganded states (7,8). The overall structure of the heterodimer is that of a "head" on two "legs." The head region (where ligand binding takes place) comprises a seven-bladed ␤-propeller in the ␣ subunit and a von Willebrand factor A-type domain in the ␤ subunit (␤A 1 ; also referred to as "I-like domain"), an ␣,␤-fold, which is inserted by short N-and C-terminal linkers into a "hybrid" domain. The hybrid domain is a ␤-sandwich fold made up of the ϳ60 amino acid residues preceding and the ϳ90 residues following ␤A. Both tertiary and quaternary structural changes are observed upon the binding of a ligand mimetic peptide containing the RGD recognition sequence to the preformed integrin crystal (8), however, a pivotal conformational change appears to be an inwards movement of the ␣1 helix of ␤A. This shift of the ␣1 helix appears to be necessary for activation (6). We have also shown that ␣1 helix motion is linked to a movement in the ␣7 helix region and a swing of the hybrid domain away from the ␣ subunit (9). These latter conformational changes were not observed in the liganded ␣ V ␤ 3 crystal structure (8), probably because of restraints imposed by lattice contacts (10,11).
Six cation binding sites were found in the unliganded and eight in the liganded ␣ V ␤ 3 structures (7,8). Four sites are present on the lower face of the ␣ subunit ␤-propeller. Although originally thought to be EF-hand-like, these sites are now known to form ␤-hairpin loops (7,12). All four hairpin loops are linked in a chain-like arrangement and the ␤-hairpin loop in blade 7 is probably important for stabilizing interactions with the underlying "thigh" domain. A fifth cation binding site is observed at the junction between the ␣ subunit "thigh" and "calf " domains. Three sites are present on the top face of ␤A. The first, known as MIDAS (metal ion-dependent adhesion site), plays a central role in ligand recognition. One of the carboxylate oxygens of the aspartic acid side chain of RGD coordinates directly to a metal ion bound at the MIDAS, thus explaining the absolute dependence of ligand binding on divalent cations (8). Occupancy of the MIDAS also induces conformational changes associated with activation of this domain. The second site lies adjacent to the MIDAS and is therefore termed ADMIDAS. The third is termed LIMBS (ligand-associated metal-binding site). The MIDAS and LIMBS sites were occupied only in the liganded structure (8). Residues involved in the cation coordination of the MIDAS and ADMIDAS sites are shown in Table I.
We have previously used the interaction between integrin ␣ 5 ␤ 1 and fibronectin as a model system to identify and characterize cation-binding sites that can support or modulate ligand recognition (13). Our results showed that several classes of cation-binding sites could be identified. Occupancy of the first class by Mg 2ϩ or Mn 2ϩ was required for ligand binding (ligandcompetent sites). The second class was implicated in the allosteric inhibition of Mn 2ϩ -supported ligand binding by Ca 2ϩ (inhibitory sites), whereas the third class was involved in the stimulation of Mg 2ϩ -supported ligand binding by Ca 2ϩ (stimulatory sites). Recently we have shown that the ligand competent site for both Mg 2ϩ and Mn 2ϩ is the MIDAS of the ␤ 1 A domain (4). The identity of the two classes of Ca 2ϩ -binding regulatory sites is currently unclear. In addition, how occupancy of these sites affects conformational movements has not yet been investigated.
Here we have examined the role of the ADMIDAS site in ligand binding by ␣ 5 ␤ 1 . Our findings provide evidence that the ADMIDAS is a member of the class of Ca 2ϩ -binding inhibitory sites and, while not essential for ligand binding, the ADMIDAS may have a role in stabilizing the active conformation through an effect on the ␣1 helix of ␤A. The ADMIDAS is also important for the transduction of cation-induced conformational changes from ␤A to the underlying hybrid domain.
Expression Vector Construction and Mutagenesis-C-terminal truncated human ␣ 5 and ␤ 1 constructs encoding ␣ 5 residues 1-613 and ␤ 1 residues 1-455 fused to the hinge regions and C H 2 and C H 3 domains of human IgG␥1 (␣5-(1-613)-Fc and ␤1-(1-455)-Fc)) were generated as previously described (16). C-terminal truncated constructs containing the full-length extracellular domains (␣5-(1-951)-Fc and ␤1-(1-708)-Fc)) were produced as before (16). To aid the formation of heterodimers, the C H 3 domain of the ␣ 5 construct contained a "hole" mutation, whereas the C H 3 domain of the ␤ 1 constructs carried a "knob" mutation as described (16,17). The mutations in the ␤ 1 subunit were carried out using oligonucleotide-directed PCR mutagenesis, as described (9). Oligonucleotides were purchased from MWG Biotech (Southampton, UK). The presence of the mutations (and the lack of any other changes to the wild-type sequence) was verified by DNA sequencing.
For comparison of purified wild-type heterodimers with heterodimers containing the ADMIDAS or LIMBS mutations in ␤ 1 , 75-cm 2 flasks of subconfluent CHOL761h cells were transfected with 5 g of wild-type or mutant ␤ 1 construct, and 5 g of wild-type ␣ 5 construct as described above. After 4 days, culture supernatants were harvested by centrifugation at 1000 ϫ g for 5 min. Wild-type or mutant heterodimers were purified using Protein A-Sepharose essentially as described before (16). Concentration measurements of wild-type or mutant heterodimers were performed using a purified ␣ 5 ␤ 1 -Fc standard of known concentration (a gift from P. Stephens, M. Robinson, and H. Kirby, Celltech Chiroscience, UK).
Effect of ADMIDAS Mutations on 3Fn6 -10 Binding-Solid phase ligand binding assays using either Fc captured or directly coated integrin were performed essentially as previously described (6,9,13,16). In these assays 3Fn6 -10 coupled to sulfo-NHS LC biotin was used at 0.1 g/ml, unless stated otherwise. Measurements obtained were the mean Ϯ S.D. of four replicate wells.
Surface Plasmon Resonance-Experiments were performed using the BIAcore 3000 (Biacore AB). Running buffer was 150 mM NaCl, 25 mM Tris-Cl, 1 mM MnCl 2 , pH 7.4. 3Fn6 -10 coupled to biotin maleimide (Sigma) was bound to the surface of a streptavidin-coated chip (Biacore AB). Dilutions of wild-type ␣ 5 ␤ 1 -Fc or ␣ 5 ␤ 1 -Fc with the D137A or D138A mutations in the running buffer were injected at 10 l/min, 25°C, into flow cells containing approximately 300 response units of 3Fn6 -10 fragment. The same buffer containing 5 mM EDTA in place of MnCl 2 was used to regenerate the surface after each injection. No binding was observed when samples were injected in the presence of EDTA. All measurements were baseline-corrected by subtracting the sensorgram obtained with that from a control flow cell coated with streptavidin alone. Kinetic parameters were determined by fitting the data to a 1:1 Langmuir binding model using BIAevaluation software version 3.1.
Effect of Divalent Cations on 15/7 Binding-96-well plates (Costar 1 ⁄2-area EIA/RIA, Corning Science Products, High Wycombe, UK) were coated with goat anti-human ␥1 Fc (Jackson Immunochemicals, Strat- for 30 min at room temperature (50 l/well). Wells were then washed four times with buffer A, and color was developed using 2,2Ј-azinobis(3ethylbenzthiazoline-6-sulfonic acid) substrate (50 l/well). Absorption at 405 nm was measured using a plate reader (Dynex Technologies). Background binding to mAbs to wells incubated with supernatant from mock-transfected cells was subtracted from all measurements. Measurements obtained were the mean Ϯ S.D. of four replicate wells.

Comparison of Epitope Expression by Wild-type and Mutant
Heterodimers-Plates were coated with anti-human Fc and blocked as described above. The blocking solution was removed and cell culture supernatants were added (25 l/well) for 1-2 h. All supernatants were assayed in triplicate, and supernatant from mock-transfected cells was used as a negative control. The plate was washed 3 times with buffer A with 1 mM MnCl 2 (buffer B) (200 l/well), and anti-␣5 or anti-␤1 mAbs (5 g/ml) were added (50 l/well). The plate was incubated for 2 h and then washed 3 times in buffer B. Peroxidase-conjugated anti-rat or anti-mouse secondary antibodies (1:1000 dilution in buffer B; Jackson Immunochemicals) were added (50 l/well) for 30 min, the plate was washed four times in buffer B, and color was developed as described above. All steps were performed at room temperature. Results shown are representative of three separate experiments.
In each assay involving a comparison between different heterodimers the binding of mAb 8E3 (5 g/ml) was used to normalize for any differences between the amounts of the different heterodimers bound to the wells. For example, normalized A 405 for 15/7 binding ϭ (AM 15/7 Ϫ Am 15/7 ) ϫ ((AWT 8E3 Ϫ Am 8E3 )/(AM 8E3 Ϫ Am 8E3 )), where A 15/7 M ϭ mean absorbance of wells coated with mutant integrin, Am 15/7 ϭ mean absorbance of wells coated with mock supernatant, AWT 8E3 ϭ mean absorbance of 8E3 binding to wells coated with wild-type integrin, Am 8E3 ϭ mean absorbance of 8E3 binding to wells coated with mock supernatant, and AM 8E3 ϭ mean absorbance of 8E3 binding to wells coated with mutant integrin. 8E3 recognizes a non-functional epitope in the N-terminal region of the ␤ 1 subunit (16). In experiments using heterodimers captured from cell culture supernatants similar results were obtained using Protein A-purified heterodimers (data not shown). Each experiment shown is representative of at least three separate experiments.

ADMIDAS Mutants Have Low Constitutive Activity and Show Reduced Expression of Activation Epitopes-
The ADMI-DAS site contains residues Ser 134 , Asp 137 , Asp 138 , and Ala 341 (or Asp 259 ) (see Table I). The side chains of both Asp 137 and Asp 138 contribute two carboxylate oxygens to cation coordination, hence mutation of either residue would be expected to The mutation E140A was used as a control. The non-function perturbing mAb N29 had no effect on 3Fn6 -10 binding (data not shown).

FIG. 2.
Effect of ADMIDAS mutations on binding of 3Fn6 -10 fibronectin fragment to ␣ 5 ␤ 1 -Fc. Binding of 3Fn6 -10 to wild-type integrin or D137A or D138A mutants was measured in the absence of mAbs (open bars) or presence of the activating mAbs TS2/16 (crosshatched bars) or 12G10 (diagonally hatched bars). The non-function perturbing mAb K20 had no effect on 3Fn6 -10 binding (data not shown). Assay was performed using Fc-captured integrin.
abrogate cation binding to this site. The other ADMIDAS residues contribute only a backbone carbonyl (except in the case of Asp 259 , which also participates in cation coordination at the MIDAS site). Hence to selectively test the role of the ADMIDAS site we made mutants D137A and D138A. For these studies we employed a recently described system for expression of recombinant soluble ␣ 5 ␤ 1 (16). We mainly used a truncated version of ␣ 5 ␤ 1 , ␣5-(1-613) ␤1-(1-455), fused to the Fc region of human IgG␥1 (hereafter referred to as tr␣5␤1-Fc). This heterodimer contains the ␣ subunit ␤-propeller and thigh domain with the ␤ subunit A, PSI, and hybrid domains (7), and retains the ligandbinding properties of the full-length receptor (16). The advantages of this system have been described previously (6,9).
The ligand binding activity of the wild-type or mutant integrins was tested after either capture of the integrin onto a 96-well plate coated with anti-human Fc polyclonal antibody, or by directly coating the purified integrin onto the plate (16). Wild-type tr␣5␤1-Fc had low activity when Fc captured but had high activity after direct adsorption (Fig. 1, A and B); the activating mAb 12G10 (15) restored the activity of Fc-captured integrin but did not further increase the activity of directly coated receptor. In contrast, both ADMIDAS mutants had very low activity after either Fc capture or direct adsorption. 12G10 only partially rescued the activity of either Fc-captured or directly adsorbed D137A and D138A mutants (in the case of directly adsorbed integrin to ϳ15 and ϳ70% of wild-type levels, respectively).
The D137A and D138A mutations were also made in a construct containing the full-length extracellular domains of ␣ 5 and ␤ 1 (hereafter referred to as ␣ 5 ␤ 1 -Fc). The mutations had little or no effect on the expression of ␣ and ␤ subunit epitopes, apart from low expression of the 15/7 activation epitope in the D138A mutant (data not shown). After Fc capture, the wildtype ␣ 5 ␤ 1 -Fc had constitutively high ligand binding activity but the mutants displayed little or no activity (Fig. 2). Activating mAbs TS2/16 and 12G10 had little effect on the ligand binding activity of the wild-type receptor but partially (D137A) or com-pletely (D138A) restored the activity of the ADMIDAS mutants (Fig. 2).
These results show that under conditions where the wildtype receptor is fully active (either constitutively or after mAbor coating-induced activation), the ADMIDAS mutants have low activity. Because activity can be rescued (at least in part) by activating mAbs, the ADMIDAS mutants are not defective in ligand binding but instead the mutations appear to inhibit ligand binding by inactivating the receptor (i.e. shifting the equilibrium toward the inactive state). Consistent with this interpretation, the D137A and D138A tr␣5␤1-Fc mutants showed decreased expression of activation epitopes, such as that recognized by mAb 12G10 (Table II). Interestingly, Asp 137 and Asp 138 lie at the top of the ␣1 helix of ␤A and we have previously shown that a movement of this helix is important for activation (6). The 12G10 epitope lies partly in the ␣1 helix and changes in expression of this epitope parallel to a shift in ␣1 (6). The D137A and D138A tr␣5␤1-Fc mutants also showed a dramatic reduction in the binding of the 15/7 and HUTS-4 mAbs. The low expression of 12G10, 15/7, and HUTS-4 activation epitopes appears to correlate with the low ligand binding activity of the tr␣5␤1-Fc D137A and D138A mutants. (This correlation was only approximate, however, because the D138A mutant had lower expression of these epitopes but higher ligand binding activity than the D137A mutant.) The ADMIDAS Mutations Increase the Dissociation Rate of Fibronectin from ␣ 5 ␤ 1 -To gain further insight into why the ADMIDAS mutants were defective in ligand binding we used surface plasmon resonance to examine the kinetics of the interaction of 3Fn6 -10 with wild-type and mutant ␣ 5 ␤ 1 -Fc (Fig.  3, A-C). The data showed that the ADMIDAS mutants had similar association rate constants but much faster dissociation rate constants than the wild-type receptor (Table III). These results show that whereas ligand binds to the ADMIDAS mutants at a similar rate to the wild-type integrin, the complex rapidly dissociates after binding takes place. (These rapid offrates may preclude detection of ligand binding to the mutants in solid phase assays, where there is a considerable time delay between removing unbound ligand and detecting bound ligand.) Hence, these findings support the suggestion that the ADMIDAS site is important for the stabilization of the active (ligand-competent) state of the integrin. Furthermore, the rapid dissociation rate observed for the D138A mutant was decreased approximately 10-fold in the presence of a Fab fragment of 12G10 (Fig. 3D), demonstrating that the effect of the mutation could be overcome by stabilizing the active conformation.

ADMIDAS Mutants Have a Similar Divalent Cation Dependence of Ligand Binding to Wild-type Integrin-Fibronectin
binding to ␣ 5 ␤ 1 is supported by Mn 2ϩ and Mg 2ϩ but only very weakly by Ca 2ϩ (13). We compared the effect of these ions on binding of the 3Fn6 -10 fragment to wild-type tr␣5␤1-Fc and the D138A mutant in solid phase assays. In these, and all subsequent ligand-binding assays shown, we used Fc-captured integrin activated by 12G10, to measure ligand binding under similar conditions for wild-type and mutant receptors. The results (Fig. 4, A and B) showed that ligand binding to the D138A mutant was modulated by divalent cations in a very similar manner to the wild-type integrin. The only significant difference between the mutant and wild type was that Mg 2ϩ ions showed a reduced ability to support ligand binding to the mutant relative to Mn 2ϩ . No ligand binding was observed in the absence of cations. Results obtained for the D137A mutant were analogous to those for D138A (data not shown).
The ADMIDAS lies very close to the MIDAS and therefore mutations in the ADMIDAS could influence cation binding to the MIDAS. The affinity of the MIDAS for a particular divalent cation is reflected by the concentration of the ion that supports half-maximal ligand binding (18) (provided that the ligand concentration is low, Ref. 19). This assumption is valid because the MIDAS cation is essential for ligand binding and therefore ligand binding is directly proportional to the fraction of integrin occupied by cation. We therefore tested whether the D138A mutation affected the concentrations of Mn 2ϩ or Mg 2ϩ required for half-maximal ligand binding. The EC 50 values for   Fig. 3) were analyzed using the BIAevaluation 3.1 software (Biacore AB) by separate fitting of data from association and dissociation phases. Results shown are mean Ϯ S.D. from a single experiment but essentially identical data was obtained in two additional experiments (not shown).

ADMIDAS Mutants Show Reduced Allosteric Inhibition of Mn 2ϩ -supported Ligand Binding by Ca 2ϩ -
The crystal structure of ␣ V ␤ 3 in the unliganded state showed that the ADMI-DAS site could be occupied by Ca 2ϩ . We have previously shown that one of the effects of Ca 2ϩ ions on ␣ 5 ␤ 1 -fibronectin interactions is an allosteric inhibition of ligand binding supported by Mn 2ϩ (13). We therefore tested whether the D138A mutation affected the ability of Ca 2ϩ ions to inhibit binding of the 3Fn6 -10 fragment in the presence of a constant concentration of Mn 2ϩ (100 M). The results (Fig. 5) showed that the D138A mutation reduced the ability of Ca 2ϩ to inhibit ligand binding. Similar results were obtained with the D137A mutant, whereas the E140A mutation had no effect (data not shown). The inhibition of Mn 2ϩ -supported ligand binding by Ca 2ϩ had a multiphasic pattern suggesting that this inhibition was mediated by Ca 2ϩ binding to at least three separate sites of differing affinities (high, medium, and low). Comparison of the pattern of inhibition for the D138A mutant with that of the wild-type receptor showed that the effect of the medium affinity site was lost. Making the assumption that the concentration of Ca 2ϩ for half-maximal inhibition in this range reflects the affinity of Ca 2ϩ ions for the medium affinity site, mathematical modeling of the inhibition (not shown) indicated that this site has an apparent K D of 160 Ϯ 30 M (n ϭ 4). Taken together, these findings show that the ADMIDAS is a selective, medium affinity Ca 2ϩ -binding site that contributes to the allosteric inhibition of Mn 2ϩ -supported ligand binding.
ADMIDAS Mutants Show No Change in the Ability of Ca 2ϩ to Modulate Mg 2ϩ -supported Ligand Binding-A second effect of Ca 2ϩ ions on ␣ 5 ␤ 1 -fibronectin interactions is modulation of Mg 2ϩ -supported ligand binding. Ca 2ϩ ions can either stimulate or inhibit Mg 2ϩ -supported ligand binding, dependent upon the Ca 2ϩ concentration (13). Low Ca 2ϩ concentrations markedly increase the affinity of Mg 2ϩ for the MIDAS site (and thereby stimulate ligand binding at low Mg 2ϩ concentrations) but at high Ca 2ϩ concentrations Ca 2ϩ ions compete directly with Mg 2ϩ for binding to the MIDAS and thereby inhibit ligand binding. The D138A mutation had little or no effect on the modulation of Mg 2ϩ -supported ligand binding by Ca 2ϩ (Fig. 6). Similar results were obtained for the D137A mutant (not shown). These findings suggest that the ADMIDAS site is not a stimulatory Ca 2ϩ -binding site.
ADMIDAS Mutations Block Transduction of a Conformational Change to the Hybrid Domain-Asp137 and Asp138 lie in the ␣1 helix. We have previously shown that ␣1 helix movement is linked to a movement in the ␣7 helix region, and thereby to an outward swing of the hybrid domain, which exposes the 15/7 and HUTS-4 epitopes (9). Mn 2ϩ or Mg 2ϩ binding to the MIDAS is able to promote this conformational change (9). We therefore tested whether the ADMIDAS mutations affected induction of the 15/7 epitope by these cations (Fig. 7). The results showed that the D137A and D138A mutants had lower levels of 15/7 binding compared with the wild- type integrin in the absence of divalent ions and Mn 2ϩ or Mg 2ϩ failed to increase 15/7 binding. The control mutation E140A had no effect on the induction of 15/7 binding by these ions (data not shown). Similar results were obtained with the HUTS-4 mAb (data not shown). Hence these findings suggest that the ADMIDAS mutations block the pathway of conformational change from the MIDAS to the hybrid domain.

DISCUSSION
Three main classes of cation-binding sites that affect integrin function have been deduced from studies on ␣ 5 ␤ 1 , ␣ L ␤ 2 , and ␣ V ␤ 3 (13,18,20,21). These are ligand-competent, stimulatory (or effector), and inhibitory sites. The major ligandcompetent site has recently been shown to be the MIDAS of ␤A (6,8). Using the prototypical interaction between ␣ 5 ␤ 1 and fibronectin, we now report (a) that the ADMIDAS site does not contribute directly to ligand binding but is important for stabilizing the active conformation of the integrin, (b) that the ADMIDAS is a Ca 2ϩ -binding site involved in the allosteric inhibition of Mn 2ϩ -supported ligand binding, and (c) the AD-MIDAS is involved in transduction of a conformational change from the MIDAS through the C-terminal helix region of the ␤A domain to the underlying hybrid domain, implying an important role in receptor signaling.
In allocating functions to the various cation-binding sites in integrins, a central issue is which sites need to be occupied for ligand binding to occur; i.e. the identity of the ligand-competent sites. Both the MIDAS and LIMBS appear to fall into this category (8). The ADMIDAS, however, belongs to the category of Ca 2ϩ -binding inhibitory sites. A surprising feature of the inhibition of ligand binding to ␣ 5 ␤ 1 by Ca 2ϩ is that Ca 2ϩ acts as a competitive inhibitor of ligand binding supported by Mg 2ϩ but not of ligand binding supported by Mn 2ϩ . This previously led us to suggest that there are separate ligand-competent sites for Mg 2ϩ and Mn 2ϩ (13). However, it is now clear that the major ligand-competent site for both Mg 2ϩ and Mn 2ϩ is the MIDAS (6). Why Ca 2ϩ is able to displace Mg 2ϩ but not Mn 2ϩ from the MIDAS will be the subject of future investigations.
Whereas the ligand-competent sites typically bind Mn 2ϩ or Mg 2ϩ , both the stimulatory and inhibitory sites appear to bind Ca 2ϩ selectively (13,20,21). The ␣ V ␤ 3 crystal structures showed that the ADMIDAS can be occupied by either Ca 2ϩ or Mn 2ϩ . However, no insight into the specificity of this site should be inferred from these studies because the crystals were prepared in the presence of a high concentration of a single cation (5 mM Ca 2ϩ or Mn 2ϩ ). The studies presented here show that the ADMIDAS preferentially binds Ca 2ϩ ; this feature may be because of its high content of coordinating groups from negatively charged residues (22). In addition to the ADMIDAS, other sites appear to contribute to the inhibition of Mn 2ϩsupported ligand binding by Ca 2ϩ ; these could include sites on ␣ subunit ␤-propeller. 2 The LIMBS would appear to be a good candidate for a stimulatory Ca 2ϩ -binding site because of its close proximity to the MIDAS. However, it is difficult to test this suggestion experimentally because LIMBS mutations block ligand binding to ␣ 5 ␤ 1 , and therefore it is not possible to check if the synergistic effects of low concentrations of Ca 2ϩ and Mg 2ϩ are abrogated by LIMBS mutations. 3 A second major issue concerns the contribution of cation- binding sites to signal propagation mechanisms through the receptor molecule. We have previously shown that activation of the head region of ␣ 5 ␤ 1 involves a linked movement of the ␣1 and ␣7 helices in the ␤A domain and an outward swing of the hybrid domain (9). The MIDAS is involved in this pathway of conformational changes because MIDAS mutations block both ␣1 and ␣7 movements (6,9). We report here that the ADMIDAS mutations also affect the activation state of the integrin. These mutations could affect activation directly through loss of AD-MIDAS cation, or indirectly through an effect on the ␣1 helix of ␤A, which coordinates the ADMIDAS cation at its N-terminal end. In practice it may be difficult to distinguish between these alternatives. However, in support of at least a partial role for an indirect effect, we have recently found that other mutations in the ␣1 helix can either inactivate or activate the receptor. 4 Also, if the effect of the D137A and D138A mutations was simply because of loss of the ADMIDAS cation then the two mutations should have an equal effect on ligand binding. Instead, our observation that the D137A mutation has a more severe effect on ligand binding than the D138A mutation suggests that the mutations do not only affect cation binding. The ability of 12G10 to override the effect of the ADMIDAS mutations could be explained by the ability of this mAb to stabilize the inward shift of the ␣1 helix observed in the active state of ␤A (6,8). Thus the ADMIDAS appears to be required for maintaining the active conformation of the integrin head. AD-MIDAS mutants also showed decreased expression of 15/7 and HUTS-4 epitopes in hybrid domain. A plausible explanation for this is that ␣7 helix movement and the outward pivoting of the hybrid domain are linked to ␣1 helix movement (9), and therefore impeding the shift of the ␣1 helix causes a perturbation of this signaling pathway.
Mutagenesis of ADMIDAS residues in cell surface-expressed ␤ 2 and ␤ 3 integrins has produced conflicting results. In ␣ IIb ␤ 3 ADMIDAS mutations had no effect (23), whereas in ␣ L ␤ 2 the mutation D120A (equivalent to D138A) blocked function (24). In ␣ L ␤ 2 this mutation also strongly perturbed binding of mAb 24 (an antibody that recognizes an activation epitope on the ␤2 A domain). Hence, these results for ␣ L ␤ 2 are in overall agreement with our findings for ␣ 5 ␤ 1 . The ADMIDAS could represent the site from which Ca 2ϩ is removed by EGTA/Mg 2ϩ activation of ␤ 2 integrins (5,24). In support of this suggestion, this site has an apparent affinity for Ca 2ϩ in the same range as that described here for the ADMIDAS site (24). The ADMIDAS probably also represents the Ca 2ϩ -binding site on the ␤2 A domain reported by Xiong and Zhang (affinity ϳ 100 M) (25). It is less clear whether the ADMIDAS is equivalent to the inhibitory Ca 2ϩ site described by Hu et al. (21) for ␣ V ␤ 3 , which had an affinity in the millimolar range.
The activation mechanism of integrins has recently been proposed to involve a switchblade-like straightening from a highly bent to an extended form (26,27). Mn 2ϩ but not Ca 2ϩ is able to mediate this conformational rearrangement (27). However, here we have recapitulated the effects of divalent cations on the native receptor (with full-length extracellular domains) (13) using the truncated integrin (which lacks most of the leg regions). Therefore, although cation occupancy and receptor bending/straightening may be linked, the effects of the different cations on activation and ligand binding are not dependent on unbending of the receptor. Instead, these data suggest that the cation-binding sites regulate ligand binding mainly through effects on conformation of the head region. Hence, for example, the activating property of Mn 2ϩ on integrin function appears to be primarily through its effect on the ␤A domain rather than on straightening. This activating effect of Mn 2ϩ , and of mAbs such as TS2/16, appears to be due mainly to a decrease in the dissociation rate (28 -30).
In summary, we have established previously unidentified roles for the ADMIDAS site in regulation of ligand binding to ␣ 5 ␤ 1 . These findings also support our recently proposed model of activation of the head region involving movements of the ␣1 and ␣7 helices of ␤A (9). Control of this shape-shifting pathway may explain, in part, why the ADMIDAS site is absolutely conserved throughout all integrin ␤ subunits sequenced to date (31,32).