Binding affinity of metal ions to the CD11b A-domain is regulated by integrin activation and ligand

physiologic (glutamate-based) it is that the nature the ligand binding interface The present findings help explain several of the apparently discrepant effects of Ca 2+ on α A-integrins. The present studies also reveal a potential activation-driven Ca 2+ : Mg 2+ exchange at MIDAS taking place under the physiologic mM concentrations of these cations, which may contribute to dynamic adhesion in vivo.


Introduction
Heterodimeric αβ integrins are a large family of cell surface receptors that mediate cell-cell and cell-matrix adhesion, thus regulating most functions of living cells (1,2).
The divalent cation-dependent binding of physiologic ligands to integrins is triggered allosterically by "inside-out" activation signals that are propagated across the plasma membrane to induce ligand-competency of the ectodomain. Liganded integrins in turn initiate signals that travel from "outside-in" to modify cell behavior. This bidirectional signaling is tightly regulated to ensure the proper titration of cell adhesion to physiologic needs.
Structural studies of integrins have elucidated the basis of metal-dependent ligand binding (3)(4)(5)(6). Integrins contain a ligand-binding von Willebrand Factor A-domain (vWFA) in the α-and/or β-subunits (αA and βA, respectively). αA (or I-domain) assumes a nucleotide-binding (Rossmann) fold, with a mostly parallel β-sheet surrounded by α-helices ( Figure 1). Ligand binding is mediated by a divalent cation at the apex of the β-sheet, which coordinates side chains from three non-contiguous surface loops. A glutamate residue from the ligand completes an octahedral coordination sphere around the metal (Figure 1), with ligand binding specificity arising from additional contacts made with the surrounding surface of the integrin.
The ligand-binding site in αA is named MIDAS (for metal ion-dependent adhesion site). Structural, mutagenesis and biophysical studies revealed an "open" and "closed" states of αA corresponding to active, and inactive states, respectively (7)(8)(9).
The open and closed states in the native integrin are in a dynamic equilibrium that normally favors the closed state in quiescent cells (10). The open state (vs. closed) is characterized by a changed position and packing of the F-α7 loop (between the F strand and the c-terminal α7 helix), an inward movement of the α1 helix, and a 10Å downward slide of the α7 helix (11) (Figure 1). These tertiary changes alter metal ion coordination at MIDAS and the shape and charge of the MIDAS face, enabling it to bind physiologic ligands (5,6). The active state is favored/stabilized by bound ligand (7). It can also be induced allosterically by destabilizing the hydrophobic contact of α7 with the central β-sheet (8,10). This allosteric upregulation can be produced in vitro by a Gly substitution of the invariant α7 residue Ile 316 which by guest on March 24, 2020 http://www.jbc.org/ Downloaded from resides into the hydrophobic SILEN (site-for-isoleucine) pocket in inactive CD11bA (8,9) (Figure 1), or by engineered disulphides resulting in an open but unliganded conformation (12). In integrins with both αA and βA domains, a conserved Cterminal glutamate from open αA acts as an endogenous ligand, by contacting the MIDAS cation in βA (13,14). This allows the propagation of conformational changes from one domain to the other in the whole integrin, providing a basis for bidirectional signaling. The MIDAS face of liganded βA is decorated by two additional metal ions at ADMIDAS (adjacent to MIDAS) and at LIMBS (ligandassociated metal binding site), on either side of MIDAS (6). In unliganded βA, only ADMIDAS is occupied by a metal that links the α1 helix to the F-α7 loop, maintaining the integrin in a low affinity state. In the liganded state, a metal ion is bound at MIDAS with a ligand Asp completing its coordination sphere, in a manner strikingly similar to that in αA. In the liganded-state, the ADMIDAS metal ion shifts inwards towards MIDAS, helping to stabilize the MIDAS cation, with the α1 helix moving in unison. The F-α7 loop undergoes a major conformational change and its link with ADMIDAS is severed. A third metal ion occupies LIMBS, serving to also stabilize the MIDAS ion and therefore the stable binding of ligand (6). As in the case of αA, the changes in βA can be ligand-induced or may be triggered allosterically through a conformational switch of the F-α7 loop driven directly by an inside-out signal as proposed in the deadbolt model (2).
It is established that the nature of the metal ion plays a critical role in regulating ligand binding affinity to αA (reviewed in (15)). In many αA-integrins, µM Mg 2+ or Mn 2+ (but not µM Ca 2+ ) supports ligand binding and mM Ca 2+ blocks Mg 2+ -mediated adhesion (16)(17)(18)(19)(20)(21). On the other hand, µM Ca 2+ was found to support ligand binding to some αA domains (21) and to the respective native integrins (22)(23)(24), an effect that can be mediated through a direct coordination of Ca 2+ at MIDAS (21,25). The reason(s) for the opposing effects of Ca 2+ on ligand binding in αA integrins is not well understood. Further, since human plasma contains mM concentrations of Mg 2+ and Ca 2+ , the physiologic relevance of the above observations is currently unclear.
Despite extensive studies of the role of metal ions in integrin function, the possibility by guest on March 24, 2020 http://www.jbc.org/ Downloaded from that metal ion binding to integrins is itself regulated by the activation state of the integrin has not been explored. One reason is that integrins contain multiple metal binding sites, making such a determination in the native integrin difficult (6).
Second, some physiologic ligands contain metal-binding sites of their own, complicating such an analysis (26,27). Third, physiologic ligands bind integrins in an activation-dependent manner (1) making it difficult to evaluate an independent role of integrin activation on metal ion binding affinity per se. Forth, physiologic ligands form multivalent interactions with integrins, sometimes with distinct metal ion requirements; for example, binding of integrin CD11a/CD18 to CD54 is Mg 2+dependent, yet clustering of this integrin is Ca 2+ -dependent (28). The feasibility of producing functionally-and structurally-defined water-soluble forms of αA (8,9) and the availability of an activation-independent ligand-mimetic mAb 107 (29) allowed us to explore if metal ion affinity is regulated by the activation state of αA and/or by ligand. We found that while both Mg 2+ and Ca 2+ display low (mM) affinity to MIDAS in inactive αA from integrin CD11b/CD18 (CD11bA), Mg 2+ selectively binds to MIDAS with µM affinity in the active state. The single-chain mAb 107 (scFv107) induced a dramatic increase in affinity of MIDAS to Ca 2+ , favoring it over Mg 2+ . The significance of these findings is discussed.

Protein Purification
The inactive form of CD11bA was generated by expressing a protein fragment spanning residues Gly 123 -Gly 321 (CD11bA 123-321 ) of human CD11b (8). The active form of CD11bA was made by replacing Ile 316 with Gly (CD11bA I316G ) (8). The MIDASdefective mutant CD11bA D242A was described (3). All proteins were expressed as GST-fusion proteins in Escherichia coli, purified by affinity chromatography, cleaved with thrombin to release recombinant CD11bA, and further purified by ion exchange chromatography on a HiTrap SP Sepharose HP column (Pharmacia, Uppsala, Sweden) using FPLC (Pharmacia). Purified preparations of CD11bA were dialyzed against 20mM Tris-HCl, pH 7.5, 150mM NaCl (TBS).

Expression and purification of single chain 107
cDNAs encoding the variable heavy (VH) and kappa light chains (VL) of monoclonal antibody 107 were isolated by reverse transcription of mRNA derived from the 107 hybridoma cell line (29), using standard primers (30). A 93-bp DNA fragment encoding the flexible linker peptide (Gly4Ser)3 (Pharmacia Biotech) was used to join the VH and VL cDNAs, generating the cDNA encoding scFv107. The latter was cloned into pPICZαB (Invitrogen) with a thrombin cleavage site and a polyhistidine tag introduced C-terminally. The construct was then fully sequenced.
The Bstx I-linearized vector was electroporated into yeast Pichia Pastoris KM71H strain (Invitrogen) and zeocin-resistant clones were selected. Detection of secreted scFv107 was done by western blotting using anti-mouse Fab (Sigma). Large-scale protein expression was done as follows: Yeast was grown at 30°C in buffered glycerol complex medium (BMGY) with maximum aeration (250rpm) until the culture reached an A 600 between 2-6. To induce protein expression, cells were centrifuged (1,500-3,000g) for 10 min at room temperature and resuspended at 1/10th of the original culture volume in buffered methanol complex medium (BMMY) containing 5% casamino acids. To maintain induction, methanol was added to a final concentration of 1% every 24hrs. The time course of protein expression was followed until 120 hours. The culture was then centrifuged and the supernatant concentrated using Amicon concentrator (Millipore) and dialyzed against 50mM sodium acetate, pH=5. Dialyzed scFv107 (lacking the His tag) was passed through a cation-exchange SP column using FPLC (Pharmacia) and eluted with 160-300mM sodium chloride, fractions were pooled, immediately dialyzed against 20mM Tris-HCl, pH=8.2 and concentrated using Centricon (Millipore) to a final concentration of 7 (31) or scFv107 was covalently coupled via primary amine groups to the dextran matrix of a CM5 sensor chip (BIAcore). Chips treated in the same way using bovine serum albumin or no protein were used as a negative controls. CD11bA (0.5µM) was flowed over the iC3b-, scFv-, BSA-coupled or bare CM5 sensor chips at 5 µl/min in TBS running buffer, containing 0.005% P20, MgCl 2 +CaCl 2 (at 1 mM each) or various concentrations of CaCl 2 (0-1mM). 1M NaCl in 20 mM Tris-HCl, pH 8.0 or 10 mM HCl followed by 10mM EDTA was used to remove the bound proteins on the chip and regenerate the surface. The binding data (after subtracting the background binding to BSA-coupled surface) were analyzed by the linear transformation method to obtain the kinetic constants (32).

Titration Calorimetry
The heat flow resulting from the binding of metal ions to CD11bA, with or without

Characterization of CD11bA and scFv107 recombinant proteins
Gel electrophoresis of the purified inactive and active CD11bA, CD11bA D242A , and scFv107 on SDS-containing reducing 12% polyacrylamide gels revealed single Coomassie Blue-stained bands of the expected molecular mass (Figure 2A). Active but not inactive CD11bA bound to the physiologic ligand iC3b in Ca 2+ -Mg 2+containing buffer, as previously shown (8). As in the case of the native 107 mAb (29), Ca 2+ alone mediated optimal binding of the activation-insensitive ligand-mimetic scFv107 to CD11bA ( Figure 2C). This binding required an intact MIDAS, as the

Interaction of Mn 2+ with active and inactive CD11bA
Use of titration calorimetry to measure affinity of metal ions to CD11bA was first established using Mn 2+ , a metal ion that binds to the integrin with high affinity (3). A solution of inactive CD11bA (0.1mM) in the calorimeter cell was titrated with a 0.65 mM solution of MnCl 2 at 25 o C. As the metal ion is added, heat is released and as titration continued, the free protein concentration in the cell decreased. The heats of the reaction after integration of the titration peaks are shown in Figure 3A; injection of the same metal ion solution into pure buffer had negligible heats of reaction. The ion-into-protein titration led to complete binding of CD11bA, with half maximal 9 heat amplitude obtained at Mn 2+ /inactive CD11bA of ~1.0 ( Figure 3A), consistent with the presence of CD11bA in a monodisperse form and in agreement with the existing crystallographic data (11). The binding isotherm for the Mn 2+ -inactive CD11bA interaction is characteristic of an exothermic single binding site interaction with µM binding affinity (37+6, n=4, Table 1) and an observed enthalpy change (∆H o ) of -5.87+0.20 kcal/mol of Mn 2+ (mean+SD, n=4)( Table 1) Table 1).

Interaction of Mg 2+ with inactive and active CD11bA
In contrast to Mn 2+ , the binding affinity of Mg 2+ to the inactive and active forms of CD11bA were dramatically different. Binding to the former was of low affinity (0.937+0.088mM)( Figure 3B and Table 1

Interaction of Ca 2+ with active and inactive CD11bA
We next assessed the binding affinity of Ca 2+ to the two conformations of CD11bA.
In both cases the binding affinity was very low (~2mM) ( Figure 3C,

Effect of ligand on the thermodynamics of metal ion-CD11bA interactions
The binding affinity of the three metal ions to the inactive and active forms of CD11bA was next measured in the presence of scFv107. None of these metal ions bound to scFv107 directly in the absence of CD11bA ( Figure 4A-C). Whereas the 10 binding affinity of Mn 2+ or Mg 2+ to CD11bA did not change significantly, a 50-and 80-fold increase in Ca 2+ binding to inactive and active CD11bA, respectively were observed ( Figure 4C and Table 1). The binding energy for the Ca 2+ -inactive CD11bA interaction is driven largely by favorable enthalpy; both favorable enthalpy and entropy contributed to Ca 2+ binding to active CD11bA (Table 1). In both cases, halfmaximal heat amplitude is obtained at a Ca 2+ :CD11bA molar ratio of ~1.0, consistent with the presence of a single Ca 2+ binding site in CD11bA. This is most likely at MIDAS, as no binding of scFV107 to the MIDAS-defective mutant CD11bA D242A occurs ( Figure 2C).

Discussion
The major finding in this report is that the activation state as well as ligand regulate metal ion binding affinity to CD11bA. Activation induced a 10-fold increase in the binding affinity of Mg 2+ but not Ca 2+ to CD11bA. On the other hand, the ligandmimetic scFV107 induced a dramatic increase in the binding affinity of Ca 2+ but not Mg 2+ to CD11bA. configuration. This appears to be largely due to formation of a metal:protein bond (via the hydroxyl oxygen of Thr209) which may account for the fact that the observed increase in affinity is largely enthalpy-driven (Table 1).
In contrast to Mg 2+ , the low (mM) affinity of Ca 2+ to CD11bA was dramatically increased by ligand binding, with activation adding only slightly to the binding affinity. This increase was largely driven by a favorable change in enthalpy, which results from formation of a bond between the metal ion and the ligand Asp (Table  12 downward movement of the c-terminal α7 helix (38). Modeling of the antigenbinding site of mAb107 on that of the AQC2-integrin complex predicts that Asp107 from mAb107 may act as MIDAS-coordinating ligand (not shown). Thus Asp107 from scFv107 may create a highly favored Ca 2+ coordination site at MIDAS, accounting for the dramatic increase in binding affinity of Ca 2+ to closed CD11bA/scFV107 complex. It is also interesting to note that the critical influence of ligand on Ca 2+ binding to MIDAS has also been observed in the βA domain. In protein crystals of the αA-lacking integrin αVβ3, the βA MIDAS is metal-bound in the presence but not absence of the prototypical RGD ligand (5,6). Also, in native α4β1 (39) and α9β1 (40), a rapidly exchangeable Ca 2+ -binding site, presumably at MIDAS, and of a similar affinity to that observed here for αA, is only detected in the presence of the ligand. Taken together, these data support the existence of a coupled equilibrium between ligand binding and MIDAS ion binding that is applicable not only to βA (39) but also to αA domains in integrins.
Previous studies have shown that Mn 2+ or Mg 2+ but not Ca 2+ support binding of physiologic ligands to active CD11bA. In these cases, a ligand glutamate (rather than an aspartate) engages the MIDAS ion directly (7,12). In view of the preferred Ca 2+donor atom distances derived recently (41), we suggest that the longer glutamate side chain of physiologic ligands like iC3b and CD54 may prevent Ca 2+ from forming a ternary ligand/metal/integrin complex. In the structure of CD11aA-CD54-Mg 2+ complex for example, the extra 1.52Å methyl group of the ligand glutamate may create a steric conflict for Ca 2+ if it were to replace the MIDAS Mg 2+ in this structure. In other instances where Ca 2+ has been shown to support physiologic (glutamate-based) ligand binding (21), it is conceivable that the nature of the ligand binding interface allows an optimal coordination of Ca 2+ at MIDAS.
The present findings help explain several of the apparently discrepant effects of Ca 2+ on α A-integrins. The present studies also reveal a potential activation-driven Ca 2+ :Mg 2+ exchange at MIDAS taking place under the physiologic mM concentrations of these cations, which may contribute to dynamic adhesion in vivo.