The α-Subunit Regulates Stability of the Metal Ion at the Ligand-associated Metal Ion-binding Site in β3 Integrins*

Background: Metal ions at LIMBS and MIDAS are essential for integrin-ligand interactions. Results: MD simulations showed strikingly different and functionally relevant conformations of LIMBS in αVβ3 and αIIbβ3. Conclusion: The α-subunit regulates metal ion coordination at LIMBS and hence function of β3 integrins. Significance: The results reveal a new mechanism of integrin regulation by the α-subunit. The aspartate in the prototypical integrin-binding motif Arg-Gly-Asp binds the integrin βA domain of the β-subunit through a divalent cation at the metal ion-dependent adhesion site (MIDAS). An auxiliary metal ion at a ligand-associated metal ion-binding site (LIMBS) stabilizes the metal ion at MIDAS. LIMBS contacts distinct residues in the α-subunits of the two β3 integrins αIIbβ3 and αVβ3, but a potential role of this interaction on stability of the metal ion at LIMBS in β3 integrins has not been explored. Equilibrium molecular dynamics simulations of fully hydrated β3 integrin ectodomains revealed strikingly different conformations of LIMBS in unliganded αIIbβ3 versus αVβ3, the result of stronger interactions of LIMBS with αV, which reduce stability of the LIMBS metal ion in αVβ3. Replacing the αIIb-LIMBS interface residue Phe191 in αIIb (equivalent to Trp179 in αV) with Trp strengthened this interface and destabilized the metal ion at LIMBS in αIIbβ3; a Trp179 to Phe mutation in αV produced the opposite but weaker effect. Consistently, an F191/W substitution in cellular αIIbβ3 and a W179/F substitution in αVβ3 reduced and increased, respectively, the apparent affinity of Mn2+ to the integrin. These findings offer an explanation for the variable occupancy of the metal ion at LIMBS in αVβ3 structures in the absence of ligand and provide new insights into the mechanisms of integrin regulation.

The aspartate in the prototypical integrin-binding motif Arg-Gly-Asp binds the integrin ␤A domain of the ␤-subunit through a divalent cation at the metal ion-dependent adhesion site (MIDAS). An auxiliary metal ion at a ligand-associated metal ion-binding site (LIMBS) stabilizes the metal ion at MIDAS. LIMBS contacts distinct residues in the ␣-subunits of the two ␤ 3 integrins ␣ IIb ␤ 3 and ␣ V ␤ 3 , but a potential role of this interaction on stability of the metal ion at LIMBS in ␤ 3 integrins has not been explored. Equilibrium molecular dynamics simulations of fully hydrated ␤ 3 integrin ectodomains revealed strikingly different conformations of LIMBS in unliganded ␣ IIb ␤ 3 versus ␣ V ␤ 3 , the result of stronger interactions of LIMBS with ␣ V , which reduce stability of the LIMBS metal ion in ␣ V ␤ 3 . Replacing the ␣ IIb -LIMBS interface residue Phe 191 in ␣ IIb (equivalent to Trp 179 in ␣ V ) with Trp strengthened this interface and destabilized the metal ion at LIMBS in ␣ IIb ␤ 3 ; a Trp 179 to Phe mutation in ␣ V produced the opposite but weaker effect. Consistently, an F191/W substitution in cellular ␣ IIb ␤ 3 and a W179/F substitution in ␣ V ␤ 3 reduced and increased, respectively, the apparent affinity of Mn 2؉ to the integrin. These findings offer an explanation for the variable occupancy of the metal ion at LIMBS in ␣ V ␤ 3 structures in the absence of ligand and provide new insights into the mechanisms of integrin regulation.
Integrins are ␣␤ heterodimeric cell adhesion receptors that mediate divalent cation-dependent cell-matrix and cell-cell adhesion during morphogenesis, as well as the maintenance of tissues and organs in adult life. 18 ␣and 8 ␤-subunits assemble into 24 integrin receptors in mammals. Integrins regulate fundamental aspects of cell behavior, including migration, adhe-sion, differentiation, growth, and survival, by communicating bidirectional signals between the extracellular environment and the intracellular cytoskeleton (1).
Integrins are unusual receptors as they do not engage physiologic ligand unless activated. This property allows patrolling blood leukocytes and platelets, for example, to circulate without aggregating or interacting with the vessel walls. Inappropriate activation of integrins contributes to the pathogenesis of common diseases including heart attacks, stroke, and cancer growth and metastasis. Thus understanding how these receptors are regulated is important in promoting health and treating disease (2).
The integrin heterodimer comprises a head segment that sits on top of two leg segments each spanning the plasma membrane once and ending with a short cytoplasmic tail. The ligand-binding head consists of a seven-bladed ␤-propeller domain from the ␣-subunit that associates noncovalently with a GTPase-like domain, ␤A, from the ␤-subunit (3). Contacts between the cytoplasmic tails and transmembrane segments hold the integrin in an inactive state (unable to bind physiologic ligand). Binding of talin to the ␤-cytoplasmic tail breaks these contacts, switching the ectodomain into the active (ligandcompetent) conformation, a process called "inside-out" signaling (4). Ligand binding then triggers global conformational changes that propagate through the plasma membrane to the cytoplasmic tails, leading to "outside-in" signaling.
Integrin-ligand interactions are regulated in a complex manner by divalent cations (5)(6)(7)(8). Although Mn 2ϩ and, to a lesser extent, Mg 2ϩ stimulate ligand binding, Ca 2ϩ is typically inhibitory. The ligand-binding face of the ␤A domain is decorated by three metal ion-binding sites: a metal ion-dependent adhesion site (MIDAS), 4 flanked on one side (facing the propeller domain of the ␣-subunit) by a ligand-associated metal ion-binding site (LIMBS), and on the opposite side by an Adjacent to MIDAS (ADMIDAS) (9). A ligand aspartate completes the octahedral metal coordination of an Mg 2ϩ (or Mn 2ϩ ) at MIDAS in ligand-bound integrins. Ca 2ϩ but not Mg 2ϩ binds preferentially at LIMBS and ADMIDAS in physiologic buffer conditions, and both sites can coordinate Mn 2ϩ . The metal ion at LIMBS stabilizes the one at MIDAS (9 -11), thus acting as a positive regulator of ligand binding to integrins, whereas the ADMIDAS metal ion can stabilize alternate inactive and active conformations of the integrin (2).
At the ligand-binding face of ␤A domain, the LIMBS loop residues Arg 216 , Asp 217 , and Ala 218 contact residues in the ␣-subunit propeller domain (Fig. 1), but a potential role of the ␣-subunit in regulating metal ion occupancy in the ␤A domain has not been explored. In this study, we carried out computational and functional studies on the two ␤ 3 integrins ␣ V ␤ 3 and ␣ IIb ␤ 3 . Our studies reveal an important role of the ␣-subunit in regulating metal ion stability at LIMBS in ␤ 3 integrins. The significance of this finding is discussed.

EXPERIMENTAL PROCEDURES
Molecular Dynamics Simulations Design-The crystal structures of the ectodomains of ␣ IIb ␤ 3 (Protein Data Bank (PDB) 3fcs) (12) and ␣ V ␤ 3 (PDB 4g1e) (13) were downloaded from the Protein Data Bank. In ␣ IIb ␤ 3 , LIMBS (also known as SymBS, synergistic metal-binding site), MIDAS, and ADMIDAS were occupied by Ca 2ϩ , Mg 2ϩ , and Ca 2ϩ respectively. Because only LIMBS was metal-occupied (by Ca 2ϩ ) in unliganded ␣ V ␤ 3 structure, Mg 2ϩ and Ca 2ϩ were respectively placed at MIDAS and ADMIDAS such that their initial distances (i.e. before minimization) from all the proximal oxygen atoms were between 2.4 and 4 Å. Mutations were made to the native structures using the software Swiss-Pdb Viewer 4.1.0 (14). All non-protein atoms were removed from ␣ IIb ␤ 3 and ␣ V ␤ 3 structures, leaving ␣and ␤-subunits with nearly 24,000 atoms for each, including hydrogens. Proteins were solvated in water boxes of sizes 168 ϫ 169 ϫ 207 Å (for ␣ V ␤ 3 ) and 180 ϫ 144 ϫ 226 Å (for ␣ IIb ␤ 3 ), adding ϳ183,000 and 185,000 water molecules to ␣ V ␤ 3 and ␣ IIb ␤ 3 , respectively, and ionized with 150 mM KCl. To investigate the effects of Phe 191 in ␣ IIb and Trp 179 in ␣ V on coordinating divalent cations at LIMBS, six distinct combinations of cation type and mutation for ␣ IIb ␤ 3 and four for ␣ V ␤ 3 were tested. To investigate the effects of each cation arrangement, the ␣ IIb ␤ 3 and ␣ V ␤ 3 structures were equilibrated with LIMBS-MIDAS-ADMIDAS occupancies of Ca 2ϩ -Mg 2ϩ -Ca 2ϩ and Mn 2ϩ -Mn 2ϩ -Mn 2ϩ .
Molecular Dynamics Simulations-MD simulations of the integrin ectodomain were performed with NAMD cvs_ 20130828 software package (15). The CHARMM27 force field parameter (16) was used to model the protein. The TIP3P model (17) was used for water molecules. Structures were visualized with Visual Molecular Dynamics (VMD) (18) or Chimera (19). The crystal structures of the ligand-free forms of ␣ IIb ␤ 3 and of ␣ V ␤ 3 ectodomains were used without modification, except for the manual placement of metal ions at LIMBS and ADMIDAS in ␣ V ␤ 3 and the removal of the sugar and water molecules before solvation. All simulations were carried out at the computing facilities of the National Energy Research Scientific Computing Center (NERSC). Periodic boundary conditions were applied in all three directions. Afterward, the entire system was minimized for 20,000 steps followed by 20 or 40 ns of equilibration. A time step of 2 fs was used in all simulations. The temperature and pressure of the systems were held constant at 1 atmosphere and 310 K respectively, using the isothermal-isobaric ensemble with the Langevin piston and Hoover method, as successfully used for modeling integrins (20 -22). We performed 20 or 40 ns of MD simulation for each run, and the trajectories were used for all analyses. The cutoff distance for non-bonded interactions was 1.2 nm, and the particle mesh Ewald method was used for electrostatic force calculations (15). All B-factors were set at zero. The hydrogen atom bond length was constrained using the SHAKE algorithm (23).
Structure Analysis-Two parameters, root mean square deviation (r.m.s.d.) values of the LIMBS metal ion and the energy of interaction of the metal ion with the LIMBS pocket, were used to evaluate stability of the metal ion at LIMBS. The r.m.s.d. of a single atom is a measure of its distance at each time step from the initial position of the ion. Hence, higher r.m.s.d. values represent fluctuations with larger amplitudes and less stable ion pocket bonds. Higher interaction energies reflect higher bond stability between the ion and LIMBS or between the ␣-subunit and the LIMBS loop comprising residues Arg 216 , Asp 217 , and Ala 218 . To let systems equilibrate for a considerable time span before starting to take samples for r.m.s.d. measurements, r.m.s.d. values for the ion were averaged over time steps between t ϭ 16 -20 ns for all simulations (n ϭ 40). Energies of interaction between the divalent ion and the LIMBS pocket were estimated using Langevin dynamics. As this energy of interaction did not show significant fluctuations throughout the simulations, values for energy were averaged over t ϭ 0 -20 ns (n ϭ 200).
Reagents and Site-directed Mutagenesis-Restriction and modification enzymes were obtained from New England Biolabs Inc. (Beverly, MA), Invitrogen Life Technologies, or Fisher Scientific. All cell culture reagents were obtained from Invitrogen Life Technologies. The non-inhibitory monoclonal antibody (mAb) AP3 (American Type Culture Collection, ATCC) detects the ␤ 3 subunit in all conformations. The heterodimer-FIGURE 1. ␣-subunit residues facing LIMBS in ␤ 3 integrins. A ribbon diagram showing the residues from ␣ IIb and ␣ V facing LIMBS loop residues Arg 216 -Ala 218 is presented. The ␤ 3 subunits of unliganded ␣ IIb ␤ 3 (green, 3fcs.pdb) and ␣ V ␤ 3 (yellow, 4g1e.pdb) ectodomains were superposed with Chimera. The metal ion (M 2ϩ ) at LIMBS (sphere) has the color of the respective integrin. ␣ IIb and ␣ V residues are labeled in green and black, respectively, and the LIMBS residues are labeled in red. specific mouse mAbs CD41P2 to ␣ IIb ␤ 3 and LM609 to ␣ V ␤ 3 were from Millipore (Danvers, MA). The function-blocking anti-␤ 1 mAb P5D2 was from R&D Systems, Inc. (Minneapolis, MN). Mouse mAb AP5 detects the N-terminal sequence in the PSI domain only in high affinity/ligand-bound states. The Fab fragment of AP5 was prepared by papain digestion followed by anion exchange and size-exclusion chromatography. The allophycocyanin-labeled goat anti-mouse Fc-specific IgG antibody was from Jackson ImmunoResearch Laboratories (West Grove, PA). Recombinant ␣ V ␤ 3 -specific high affinity fibronectin 10 domain (hFN10) (24) was expressed and purified from Escherichia coli as described (25), and fibronectin-depleted human fibrinogen (FB) was obtained from Enzyme Research Laboratories (South Bend, IN,). Wild-type ligands FN10 and FB and mAbs AP5 (Fab) and AP3 (IgG) were labeled respectively with N-hydroxy succinimidyl esters of Fluor 488 (Alexa Fluor 488) or Alexa Fluor 647 (Invitrogen) according to the manufacturer's instructions. Excess dye was removed using Centri-Spin sizeexclusion microcentrifuge columns (Princeton Separations, Adelphia, NJ). The final hFN10, FB, AP5, and AP3 concentrations and dye-to-protein molar ratios (F/P) were determined spectrophotometrically, giving dye:protein molar ratios of 1-5. F191/W (F/W) and F992F/A992A (FF/AA) substitutions in human ␣ IIb , W179/F (W/F), and F990F/A990A (FF/AA) substitutions in human ␣ V and ␤-genu deletion (⌬-genu) or D158/N (D/N) substitutions in ␤ 3 (26) were introduced using site-directed mutagenesis with the QuikChange kit (Agilent Technologies) and confirmed by DNA sequencing.
Transfections and mAb Binding-HEK293T (ATCC) cells were transiently co-transfected with pcDNA3 plasmids encoding different combinations of wild type and mutant ␤ 3 integrins using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's protocol. Plasmids used encoded full-length wild-type ␣ IIb ␤ 3 or ␣ V ␤ 3 , the respective integrin carrying substitutions or encoding constitutively active forms of these integrins by substituting the conserved transmembrane motif F992F (in ␣ IIb ) or F990F (in ␣ V ) to AA (27) ). In addition, two mutations were created in constitutively active ␣ IIb ␤ 3 , one known to cause loss of binding to FB ( , and the second by substituting the LIMBS metal-coordinating residue Asp 158 with asparagine (␣ IIb FF/AA ␤ 3 D/N ). 48 h after transfection, cells were detached (10 mM EDTA/PBS) and washed twice in Hanks' balanced saline solution (HBSS) and once in HBSS containing 1 mM CaCl 2 /1 mM MgCl 2 (HBSS 2ϩ ). 6 ϫ 10 5 ␣ IIb ␤ 3expressing cells in 100 l of HBSSϩ0.5% BSA were stained with the Alexa Fluor 647-labeled Fab fragment of AP5 (10 g/ml; 30 min; room temperature) followed by one wash and then fixed (1% buffered paraformaldehyde). To assess ␤ 3 integrin expression levels, 6 ϫ 10 5 cells in 100 l of HBSSϩ0.5% BSA were incubated with CD41-P2 (anti-␣ IIb ␤ 3 ) mAb or LM609 (anti-␣ V ␤ 3 ) mAb at 10 g/ml for 30 min at 4°C followed by one wash and then addition of allophycocyanin-labeled anti-mouse Fcspecific IgG (10 g/ml) for 20 min on ice. Samples were again washed once in HBSS and then fixed. Other cells were labeled with Alexa Fluor 647-labeled AP3 (anti-␤ 3 ) (at 10 g/ml; 30 min; 4°C), washed once in HBSS, and then fixed. 20,000 cells were analyzed for each sample using a FACSCalibur or LSR-Fortessa flow cytometer (BD Biosciences). Binding of CD41-P2, LM609, AP3, and AP5 to ␤ 3 integrin-expressing HEK293T was expressed as mean fluorescence intensity, as determined using the FlowJo software (BD Biosciences). Cell binding of AP5 was normalized by dividing its mean fluorescence intensity by the mean fluorescence intensity for CD41-P2 and multiplying by 100.
Soluble Ligand Binding Assays-For ligand binding experiments, 1 mM CaCl 2 /1 mM MgCl 2 or 1 mM MnCl 2 was added to 6 ϫ 10 5 ␤ 3 integrin-expressing HEK293T cells in 100 l of HBSS buffer containing 0.5% BSA and incubated in the presence of a saturating amount of Alexa Fluor 488-labeled FB (160 g/ml) or Alexa Fluor 488-labeled wild-type FN10 (12.6 g/ml) for 30 min at 25°C. To block any potential interaction of FN10 with endogenous ␤ 1 integrins in HEK293T, ␣ IIb ␤ 3 -expressing HEK293T cells were preincubated with the function-blocking anti-␤ 1 mAb P5D2 before adding Alexa Fluor 488-labeled FN10. Saturating amounts of each ligand were derived from dose-response curves, where labeled ligand was added in increasing concentrations to HEK293T cells expressing constitutively active ␤ 3 integrins in the presence of 1 mM MnCl 2 . Integrin-ligand interactions in the presence of varying concentrations of Mn 2ϩ were measured by adding increasing amounts of MnCl 2 to a mixture of ␤ 3 integrin-expressing cells and saturating amounts of Alexa Fluor 488-labeled ligand. Treated cells were then incubated with Alexa Fluor 647-labeled AP3 (10 g/ml; 30 min; 4°C) followed by washing once in HBSS containing the corresponding concentration of Mn 2ϩ . Cells were then fixed with 1% paraformaldehyde and analyzed by flow cytometry. Binding of soluble ligand to ␤ 3 integrin-expressing cells was normalized by dividing mean fluorescence intensity by that for Alexa Fluor 647-labeled AP3 and multiplying by 100. Mean and S.D. values from three independent experiments were calculated and compared using Student's t test. Non-linear curve fittings of the dose-response curves were performed using GraphPad Prism (GraphPad Software Inc., La Jolla, CA).

LIMBS Structure and Stability of the Metal Ion in ␤ 3
Integrins-Measurements of the energy of interaction of ␣ V and ␣ IIb subunits with the ␣-subunit-facing LIMBS loop residues Arg 216 -Asp 217 -Ala 218 revealed a stronger (2-3-fold) interaction of ␣ V with the LIMBS loop when compared with ␣ IIb . This difference was first detected at 8 ns of simulations and maintained through 40 ns ( Fig. 2A).
MD simulations after a few nanoseconds of equilibration showed that Ca 2ϩ coordination at LIMBS in ␣ IIb ␤ 3 comprised six "primary" oxygens (i.e. the two carboxyl oxygens of Asp 158 and Asp 217 , one carboxyl oxygen of Glu 220 , and the carbonyl oxygen of Pro 219 ) (Fig. 2B), which hold a mean distance of 2.2-2.3 Å from the encapsulated cation for the entire simulation, along with two "secondary" oxygens (i.e. the carbonyl oxygens of Asp 217 and Ala 218 ), whose mean distance from the cation was no more than 4.5 Å over the whole trajectory. The primary oxygens remained in close contact with the Ca 2ϩ at LIMBS, forming highly stable bonds with a mean length of 2.2-2.3 Å and fluctuation amplitude below 0.5 Å. The primary coordinating oxygens formed an octahedral arrangement, with a planar surface formed by OD2 of Asp 158 , OE1 of Glu 220 , OD1 of Asp 217 , and the carbonyl oxygen of Pro 219 , with OD1 of Asp 158 and OD2 of Asp 217 at the top and bottom of the plane, respectively (Fig. 2B), restricting cation fluctuations. Replacing Ca 2ϩ with Mn 2ϩ in LIMBS, MIDAS, and ADMIDAS of ␣ IIb ␤ 3 increased the energy of interaction of Mn 2ϩ with LIMBS by 5-8% and had a small but significant effect on r.m.s.d. (Table 1).
In contrast, the LIMBS pocket is distorted toward a planar shape in ␣ V ␤ 3 (Fig. 2C), the result of the stronger interaction that pulled the LIMBS loop toward ␣ V , changing metal ion coordination at this site. MD simulations showed that Ca 2ϩ at LIMBS is coordinated by six primary oxygens (the two carboxyl oxygens of Asp 158 and Asp 217 , the carbonyl oxygen of Pro 219 , and the carboxyl oxygen of Asn 215 ), but only one secondary coordinating oxygen, the carbonyl oxygen from Asp 217 (Fig.  2C). The carboxyl oxygens of Asp 158 and Asp 217 as well as the carbonyl oxygen of Pro 219 all form one planar surface that surrounds the cation, with only the side chain oxygen of Asn 215 interacting with the cation at the bottom of the plane (Fig. 2C). These changes made Ca 2ϩ at LIMBS significantly less stable in ␣ V ␤ 3 when compared with ␣ IIb ␤ 3 , as reflected by the significantly higher r.m.s.d. and lower energy of interaction of the metal with LIMBS (Table 1). Replacing Ca 2ϩ with Mn 2ϩ in LIMBS, MIDAS, and ADMIDAS of ␣ V ␤ 3 had minimal effects on r.m.s.d. or on the energy of interaction of Mn 2ϩ with LIMBS (Table 1).
Structural Basis for the Stronger Interaction of ␣ V with the LIMBS Loop-In ␣ V ␤ 3 , all three LIMBS loop residues were involved in more extensive interactions with ␣ V : Arg 216 side chain was engaged in strong ionic bonds with Glu 121 and Glu 123 , and its carbonyl oxygen occasionally H-bonded the side chain of Tyr 178 (H-bond probability 0.5%). Arg 216 also formed van der Waals contacts with Phe 154 and with the indole group of Trp 179 (Fig. 3, A and B). In addition, the main and side chains of Asp 217 formed van der Waals contacts with the indole group of Trp 179 , and the side chain of Ala 218 contacted the carboxyl oxygen of Asp 219 in ␣ V . These interactions stretched the LIMBS loop toward ␣ V , distorting LIMBS.
In contrast, interaction of ␣ IIb with LIMBS loop residues Arg 216 -Asp 217 -Ala 218 was primarily limited to Arg 216 . The side chain of Arg 216 formed intermittent H-bonds with the hydroxyl group of Tyr 190 (H-bond probability 5.0%), and its carbonyl oxygen contacted the side chain of Tyr 190 (Fig. 3C). Arg 216 side chain also formed occasional ionic interactions with Glu 123 (Fig. 3D), but made no contacts with Phe 191 . Additionally, Asp 232 made intermittent van der Waals contacts with the side chain of Ala 218 .

FIGURE 2. ␣-subunit/LIMBS interaction energies and LIMBS conformations in ␤ 3 integrins.
A, computed energy of interaction between ␣ IIb and ␣ V subunits and LIMBS loop residues Arg 216 -Ala 218 . The two trajectories appear to equilibrate after around 8 ns of simulation. The transient peak seen afterward (at ϳ16 ns) most likely represents a high energy local minimum state that the system occasionally takes, rather than simply a random fluctuation. B and C, snapshots of molecular dynamics simulations at t ϭ 20 ns, showing structures of LIMBS in ␣ IIb ␤ 3 (B) and ␣ V ␤ 3 (C), in the same orientation, with the LIMBS Ca 2ϩ shown as a large sphere in each case. In ␣ IIb ␤ 3 , the LIMBS metal ion contacts six primary oxygens (magnified red spheres) and two secondary oxygens (small red spheres) (B). In ␣ V ␤ 3 , the LIMBS metal ion also contacts six primary oxygens but only one secondary oxygen (C). Note that the side chain of Arg 216 is stretched out in C versus in B and that five primary oxygens in ␣ V ␤ 3 (C) lie in one plane, in contrast to the octahedral arrangement of the primary oxygens in ␣ IIb ␤ 3 (B).

JOURNAL OF BIOLOGICAL CHEMISTRY 23259
Effects of Modifying ␣ IIb and ␣ v on Stability of the Metal Ion at LIMBS-Of the ␣-subunit residues contacting the LIMBS loop residues Arg 216 -Asp 217 -Ala 218 , the side chains of Phe 191 in ␣ IIb and Trp 179 in ␣ V are superimposable (Fig. 1). We evaluated the impact of interchanging these two residues on stability of the metal ion at LIMBS in the two ␤ 3 integrins. As an internal control, we measured the effects of destabilizing the metal ion at LIMBS through the D158/N substitution (which removes one of the main coordinating oxygens from the metal ion). MD simulations showed that implementing the D158/N mutation in ␣ IIb ␤ 3 yielded a significant increase in r.m.s.d. (ϳ2-fold) and a reduction in the energy of interaction (Table 1), both reflecting destabilization of the metal ion at LIMBS.
Implementing the F191/W substitution in ␣ IIb significantly increased the energy of interaction of the LIMBS loop with the bulkier Trp 191 in ␣ IIb F/W when compared with Phe 191 in wildtype ␣ IIb (Fig. 4A). Snapshot of the structure at t ϭ 20 ns showed that the side chain of Arg 216 of ␤ 3 releases its interaction with Tyr 190 and forms van der Waals contacts with the indole group of Trp 191 in ␣ IIb F/W (compare Fig. 4B with Fig. 3C), pulling the LIMBS loop region toward the ␣ IIb F/W propeller domain. With Trp 191 and Tyr 190 pulling the LIMBS loop in the same direction, Ala 218 is brought closer to Asp 232 to form more contacts, increasing the energy of interaction of ␣ IIb F/W with the LIMBS loop and deforming the octahedral shape of the pocket. These movements displaced the oxygens forming the LIMBS pocket from their native pattern toward a more planar configuration. Although the energy of interaction of Ca 2ϩ or Mn 2ϩ with the LIMBS pocket in the ␣ IIb F/W ␤ 3 structure did not change, r.m.s.d. increased by ϳ4-fold (Table 1). Hence, it appears that the energetic component of the free energy remains unchanged upon applying F191/W, whereas the entropic component is highly changed upon deformation of the pocket as represented by a 4-fold increase in r.m.s.d.  Implementing the W179/F substitution in ␣ v W/F did not change the energy of interaction of the LIMBS loop with ␣ V W/F (Fig. 5A). However, it reduced fluctuations of the Ca 2ϩ or Mn 2ϩ at LIMBS (r.m.s.d. reduced by ϳ2-fold, Table 1), the result of removing the contacts between Trp 179 and the LIMBS loop (Fig. 5B). The W179/F substitution did not significantly change the interaction energy of the metal ion with LIMBS (Table 1) because it did not affect the other interactions of ␣ V with the LIMBS loop, particularly with Arg 216 (Fig. 5B).
Binding of Wild-type and F191/W Mutant ␣ IIb ␤ 3 to Soluble Ligand-We next sought experimental validation of the computational studies. We expressed recombinant ␣ IIb F/W ␤ 3 in its resting and mutationally activated (␣ IIb F/WϩFF/AA ␤ 3 , ␣ IIb F/W ␤ 3 ⌬-genu ) states in HEK293T cells and compared its surface expression, structure, and ligand binding capacity with that of constitutively active wild type ␣ IIb FF/AA ␤ 3 . As shown in Fig. 6A, surface expression of the mutant resting or constitutively active heterodimeric receptor was comparable with that of resting or constitutively active wild-type ␣ IIb ␤ 3 , and the F191/W mutation did not change the recognition of the constitutively active integrin by the ligand-induced binding site (LIBS) mAb AP5 (Fig. 6B). Binding of WT ␣ IIb ␤ 3 to soluble Alexa Fluor 488-labeled FB (Alexa Fluor 488-FB) was minimal in the presence of 1 mM each of the physiologic divalent cations Ca 2ϩ and Mg 2ϩ (Ca 2ϩ /Mg 2ϩ ), but increased only slightly in 1 mM Mn 2ϩ (Fig. 6C).
Cellular ␣ IIb F/W ␤ 3 showed minimal binding to soluble Alexa Fluor 488-FB in Ca 2ϩ /Mg 2 buffer, with 1 mM Mn 2ϩ increasing binding by ϳ2-fold (Fig. 6C). Introduction of the F191/W muta-  integrins abolished ligand binding in the presence of Ca 2ϩ /Mg 2ϩ , but did not reduce binding in the presence of 1 mM Mn 2ϩ (Fig. 6C). As a negative control, replacing the ligand contact residue Tyr 189 in ␣ IIb with alanine abolished binding of cellular ␣ IIb YA ␤ 3 ⌬-genu to soluble Alexa Fluor 488-FB in 1 mM Mn 2ϩ (Fig. 6C), in agreement with a published study (28).
Effects of F191/W in ␣ IIb and W179/F in ␣ V on Apparent Affinity of Mn 2ϩ to the Respective Integrins-We measured the binding of constitutively active wild-type and mutant ␤ 3 integrins to saturating amount of soluble ligand across a range of Mn 2ϩ concentrations. Half-maximal binding of ␣ IIb FF/AA ␤ 3 to Alexa Fluor 488-FB was achieved at an Mn 2ϩ concentration of 5.9 M Ϯ 2.1 (mean Ϯ S.D., n ϭ 3) (Fig. 7A). Removing one of the LIMBS metal-coordinating oxygens (through the D158/N mutation) increased the Mn 2ϩ concentration required for halfmaximal binding of ligand to ␣ IIb FF/AA ␤ 3 D/N by ϳ6-fold to 36.7 Ϯ 8.6 M, an indirect measure of the reduction in apparent affinity (K d(app) ) of Mn 2ϩ to ␣ IIb FF/AA ␤ 3 D/N (Fig. 7B). The F191/W substitution in ␣ IIb FF/AA ␤ 3 yielded an almost identical value (34.6 Ϯ 6.2 M) (Fig. 7B). In contrast, the W179/F mutation in ␣ V significantly increased the apparent K d(app) of Mn 2ϩ to the ␣ V W/FϩFF/AA ␤ 3 heterodimer (as judged by mAb LM609 binding, not shown) by ϳ2-fold (mean Ϯ S.D., 5.0 Ϯ 1.9 M from 10.2 Ϯ 3.5 M to ␣ V FF/AA ␤ 3 , p ϭ 0.018) (Fig. 7, C and D).

DISCUSSION
In this study, we provide computational and functional evidence that the ␣-subunit plays an essential role in stability of the metal ion coordination at LIMBS in ␤ 3 integrins. By combining these two approaches, we demonstrated the following. 1) interaction of the LIMBS loop with ␣ V is more extensive than with ␣ IIb ; 2) metal ion coordination at LIMBS after 20 ns of equilibration becomes planar in ␣ V ␤ 3 ; 3) changing the ␣ IIb -LIMBS loop interface residue Phe 191 to Trp destabilized the metal ion at LIMBS, whereas a Trp 179 to Phe mutation in ␣ V produced opposite but weaker effects; and 4) introducing F191/W in cellular ␣ IIb ␤ 3 reduced the apparent affinity of Mn 2ϩ to this integrin; the reverse was observed upon introducing the W179/F mutation in cellular ␣ V ␤ 3 .
The higher energy of interaction of ␣ V with the LIMBS loop residues Arg 216 -Ala 218 was directly related to the more extensive contacts ␣ V made with this loop when compared with ␣ IIb . The stronger contacts increased fluctuations of the metal ion at LIMBS in ␣ V ␤ 3, as reflected by the ϳ4-fold increase in r.m.s.d., and also increased the mean distances between the coordinating oxygens and the metal ion at LIMBS, as reflected by the reduction in total energy of interaction of the metal ion with the pocket. These observations may offer an explanation for the variable occupancy of LIMBS by metal ion in crystal structures of unliganded ␣ V ␤ 3 ectodomains, where LIMBS was metal-occupied in one (4g1e.pdb, used in this study) (13) but not in four other unliganded ␣ V ␤ 3 ectodomain structures (9,26,29,30). The lack of metal occupancy at LIMBS was attributed to unfavorable crystallization conditions (13). However, LIMBS is metal ion occupied in ␣ V ␤ 3 under the same crystallization conditions when ␣ V ␤ 3 is ligand-bound (9,25). The data produced in this study suggest that variability in LIMBS occupancy by metal is the result of the different contacts the LIMBS loop makes with ␣ V in contrast to ␣ IIb . In the presence of ligand, Glu 220 of the ␤A domain provides an extra primary oxygen, stabilizing the metal ion at LIMBS. In the absence of ligand, this stabilizing influence is lost, and the metal ion is freer to escape LIMBS. This scenario is reflected in the higher r.m.s.d. and lower energy of interaction of the metal ion with LIMBS in unliganded ␣ V ␤ 3 when compared with ␣ IIb ␤ 3 (Table 1). ␣ V ␤ 3 is widely expressed in tissues including bone, where it mediates dynamic cell adhesion (31,32). The majority of Mn 2ϩ in the body is sequestered in bone (33) to levels that approach the K d of ␣ V ␤ 3 for Mn 2ϩ (7), suggesting that Mn 2ϩ plays an important role in regulating ␣ V ␤ 3 function in this tissue. In contrast, ␣ IIb ␤ 3 is solely expressed on circulating platelets in blood containing high levels of its physiologic ligands, mainly fibrinogen, and mM concentrations of the divalent cations Ca 2ϩ and Mg 2ϩ . Maintaining ␣ IIb ␤ 3 in a dormant inactive state in this environment is therefore essential to prevent pathologic thrombosis. The present data provide insights into how regu- lation can be tailored to the particular environment where an integrin is expressed. Occupancy of LIMBS, MIDAS, and ADMIDAS by metal ions in unliganded ␣ IIb ␤ 3 may explain the rapid ligand association rates to activated ␣ IIb ␤ 3 (7). This potential proactivation tendency at the ligand-binding site must be counteracted by energy barrier(s) to activation elsewhere to effectively keep ␣ IIb ␤ 3 in an inactive state on circulating resting platelets. One such barrier may exist in the integrin leg segments, between the ␣-subunit Calf-2 domain and the ␤ 3 subunit EGF-like 4 (IE4) and ␤TD domains (34). Disruption of this interface rendered ␣ IIb ␤ 3 as susceptible to Mn 2ϩ -induced ligand binding as ␣ V ␤ 3 (34). In ␣ V ␤ 3 , where this Calf-2/IE4-␤TD barrier is weak or absent (34), a relatively stronger ␣ V -LIMBS interface may help favor the inactive ␣ V ␤ 3 conformation, thus limiting stable occupancy of the metal ion in this integrin to conditions when ligand is also accessible.