A discrete site modulates activation of I domains. Application to integrin alphaMbeta2.

A central characteristic of integrin adhesion receptors is their capacity to become activated, thereby enhancing their affinity for ligands. Here, we report the identification of a discrete site within the I domain of integrin αMβ2, which modulates the adhesive activity of this receptor. Based upon the crystal structure, this region is composed of two short and spatially proximal loops, E162QLKKSKTL and Q190NNPNPRS. Mutations in these loops yield receptors which support spontaneous cell adhesion to fibrinogen, whereas mutation of an adjacent region and wild-type receptors require activation to adhere to this substrate. An activating monoclonal antibody enhanced the adhesive activity of one but not the other loop mutants, suggesting that the activation states of these two mutant receptors were not identical. Given that similar I domains exist in several other integrin α subunits and non-integrin proteins, and possibly in all integrin β subunits, these two loop segments may represent a universal target for controlling integrin activation and the function of other I domain-containing proteins. In support of this hypothesis, several naturally occurring mutations that activate von Willebrand factor map to the same loops of its I(A) domain.

A central characteristic of integrin adhesion receptors is their capacity to become activated, thereby enhancing their affinity for ligands. Here, we report the identification of a discrete site within the I domain of integrin ␣ M ␤ 2 , which modulates the adhesive activity of this receptor. Based upon the crystal structure, this region is composed of two short and spatially proximal loops, E 162 QLKKSKTL and Q 190 NNPNPRS. Mutations in these loops yield receptors which support spontaneous cell adhesion to fibrinogen, whereas mutation of an adjacent region and wild-type receptors require activation to adhere to this substrate. An activating monoclonal antibody enhanced the adhesive activity of one but not the other loop mutants, suggesting that the activation states of these two mutant receptors were not identical. Given that similar I domains exist in several other integrin ␣ subunits and non-integrin proteins, and possibly in all integrin ␤ subunits, these two loop segments may represent a universal target for controlling integrin activation and the function of other I domain-containing proteins. In support of this hypothesis, several naturally occurring mutations that activate von Willebrand factor map to the same loops of its I(A) domain.
I(A) domains, inserted domains of ϳ200 amino acids, are common structural motifs in numerous proteins, including several integrin ␣-subunits, von Willebrand factor (vWF), 1 and the complement components C2 and factor B (1). Interactions of these proteins with other proteins are often mediated by their I domains. As two examples pertinent to this study: 1) the I domain mediates the interaction of vWF with collagen and with its platelet receptor GPIb/IX, thereby allowing vWF to bridge platelets to the subendothelial matrix, and 2) contact sites for several ␣ M ␤ 2 ligands, including NIF (2), C3bi (3), and ICAM-1, and fibrinogen (Fg) (4) map to the I domain of the ␣ M subunit (␣ M I-domain). The crystal structures of the ␣ M I-and ␣ L Idomains recently have been determined (1,5) and are very similar to ␣/␤ structures, composed of seven ␣-helices and five ␤-sheets. A cation binding site is formed at the vertex of the ␤-sheets, and the bound cation is coordinated in a "MIDAS" motif (1). Structural flexibility within the I domains allows them to exist in two distinct conformational states, which differ in the mode of cation coordination (5,6).
Common to many I domain-containing proteins is their capacity to be "activated" such that their affinities for other proteins are enhanced. vWF can be activated by botrocetin (7), a snake venom protein, to acquire the capacity to interact with GPIb/IX. ␣ M ␤ 2 can be activated on resting neutrophils and monocytes to recognize certain ␣ M I-domain ligands, including Fg (8) and C3bi (9), and to execute certain ␣ M ␤ 2 -dependent functions,suchastheadhesiontoendothelium(10),theadhesiondependent respiratory burst (11), and phagocytosis (12). Certain cytokines (11), bacterial products (lipopolysaccharides), and mAbs, such as KIM185 (13), can activate ␣ M ␤ 2 . The inherent flexibility noted in the crystal structure of the ␣ M I-domain (5, 6) may be relevant to such activation.
Spontaneous binding of vWF to platelets is observed in type IIB vWF disease (14), which is associated with platelet aggregation and thrombocytopenia. This activation of vWF has been ascribed to several mutations which map to its A1 domain. Based upon primary sequence homology and the crystal structures of the ␣ M I-and ␣ L I-domains, these vWF mutations are clustered in a spatially proximal site composed of two loop (L) segments, which connect secondary structural elements: 1) the loop region between helix 1 and ␤-sheet B, termed L(1-B), and 2) the loop between helix 2 and 3, termed L(2-3), of the ␣ M Idomain. The recognition sequence in the A1 domain of vWF for botrocetin also is located in this region (7).
In view of the structural similarity between I domains and the clustering of activating mutations to a local region within this domain of vWF, we hypothesized that mutations within the L(1-B) and L(2-3) region of the ␣ M I-domain might influence the activation state of ␣ M ␤ 2 . To test this hypothesis, these two segments were switched individually to their counterparts in ␣ L to minimize gross perturbations of structure. These mutations are, in fact, shown to differentially influence the activation state of ␣ M ␤ 2 . Thus, a very precise region that may be generally involved in the activation of I domain-containing proteins has been defined. Site-directed Mutagenesis and Expression-All mutations were created by oligonucleotide-directed mutagenesis using uracil-containing single stranded M13mp18 DNA (15). To mutate M 153 KEFVST of helix 1 in ␣ M to its counterpart in ␣ L (ILDFMKD) to create ␣ M [H(1)], a 51-mer primer (5Ј-TAATTGCTCCATCACGTCTTTCATGAAGTCCAGGATCC-GCCGAAAGTCATG-3Ј) (the underlined nucleotide represents the mutated bases) was used, and resultant plaques were screened with BamHI (GGATCC). Primer 5Ј-CTGCATCAAAGAGAACTGGTAGCTG-GTGTTGCTTAATTTCTTCATCACAGTTGAGAC-3Ј (57-mer) was used to create mutant ␣ M [L(1-B)], in which the ␣ M segment of E 162 QLKKS-KTL was changed to the corresponding ␣ L segment of KKLSNTSYQ. Primer 5Ј-TATTGGCTTCACCAGTGCGTCGGGGTCTTTCCATTTAA-CGAACTCTTTGAAGGT-3Ј (54-mer) was used to create mutant ␣ M -[L(2-3)], in which the ␣ M segment of Q 190 NNPNPRS was changed to the corresponding segment of VKWKDPDA in ␣ L . DNA sequencing of the entire I domain was conducted to confirm the desired mutations without alteration of the remaining sequence. The mutant I domains were reinserted into ␣ M using previously introduced ClaI 535 and NheI 1141 sites (16). The reconstructed cDNAs of ␣ M and ␤ 2 were then inserted separately into the expression vector pCIS2M by employing XbaI and XhoI sites. In this vector, ␣ M and ␤ 2 expression is under the control of human cytomegalovirus promoter and enhancer (15). Each expression vector was purified using CsCl gradients and transfected, together with pRSVneo (neomycin resistance gene), into 293 cells according to our established procedure (15). G418 (600 g/ml) resistant colonies were pooled, and ␣ M ␤ 2 -expressing cells were sorted by FACS analysis (FAC-Star, Becton Dickinson, San Jose, CA) using an ␣ M -specific mAb, OKM1, which recognizes an epitope outside of the I domain (17).
Ligand Binding to ␣ M ␤ 2 -expressing Cells-NIF binding assay was performed as described previously (16), using 0.05 to 1 g of 125 I-NIF and 10 6 ␣ M ␤ 2 -expressing cells in a total of 200 l of HBSS containing 2.5 mM Ca 2ϩ . The procedure for C3bi binding was modified from the original method of Bilsland (18) as described previously (16). For adhesion, denatured ovalbumin (dOva) (19) and Fg (20) were used as substrates. The exact procedure has been described elsewhere (16).
Analytical Procedures-For FACS analyses, our previous method (16) was used. For immunoprecipitation, ␣ M ␤ 2 -expressing 293 cells were 125 I-surface-labeled with lactoperoxidase according to a published method (21) and immunoprecipitated with 10 g mAb as described (16). For protein modeling, the coordinates for ␣ M I-domain were assigned according to the crystal structure of ␣ L I-domain (5) using the HOMOL-OGY program and then energy minimized using the DISCOVER program (Biosym Technologies, San Diego, CA).

␣ M ␤ 2 Mutants Are Cell-surface Expressed as Heterodimers-
Based on the location of the vWF-activating mutations (14) 5). Upon introduction of these mutant and wild-type ␣ M ␤ 2 into 293 cells, all were stably expressed. Immunoprecipitation of surface labeled cells with OKM1, a mAb to the ␣ M subunit, or TS1/18, a mAb to the ␤ 2 subunit, showed two bands by polyacrylamide gel electrophoresis in sodium dodecylsulfate (nonreducing) of the same mobility as wild-type ␣ M ␤ 2 with estimated molecular weights of 165 and 95 kDa, indicating that all three mutants were correctly assembled into heterodimers on the cell surface. FACS analyses showed that the mean fluorescence intensities of the three mutants with any one of a panel of mAbs against ␣ M (OKM1, 44, LM2/1, M1/70) or ␤ 2 (MHM23) differed by less than 15%, suggesting that the three mutants were expressed at similar levels as the wild-type receptor.
Interaction of Mutant ␣ M ␤ 2 -bearing Cells with NIF, C3bi, and dOva Does Not Require Activation-The interactions of NIF, C3bi, and dOva with the ␣ M ␤ 2 mutants were examined. Varying quantities of 125 I-NIF were added to the transfectants. The binding isotherms obtained with all the ␣ M ␤ 2 -bearing cells were very similar (Fig. 1A), providing evidence for saturable binding. Specificity was supported by the failure of mock-transfected 293 cells to bind NIF and the abolition of 125 I-NIF binding (Ͼ98% inhibition) by excess unlabeled NIF or EDTA. From the binding curves, the following K d values were calcu- The similarity in binding parameters for the three mutant and wild-type ␣ M ␤ 2 indicates that the mutated sites are not directly required for NIF recognition (␣ L ␤ 2 does not bind NIF (2)) and that overall conformation and function of the I domain in the ␣ M ␤ 2 receptors had been preserved.
EC3bi binding (Fig. 1B) and adhesion to dOva (Fig. 1C) also were not altered by the mutations. Verifying the specificity of EC3bi binding, uncoated or IgM-coated erythrocytes did not interact with the three mutant and wild-type ␣ M ␤ 2 -transfectants; mock-transfected 293 cells did not rosette with EC3bi; and EDTA abolished binding (Fig. 1B). Thus, EC3bi bound to the mutants without a requirement for activation, consistent with the reports in the literature (16,22). The interactions of the mutants with dOva were completely inhibited by 0.5 mM EDTA, indicating divalent cation dependence, or 10 g/ml NIF, indicating ␣ M ␤ 2 dependence. Mutation of the H(1) segment of ␣ M had no effect on adhesion, whereas the L(1-B) and L(2-3) mutants exhibited a 2-fold increase in adhesion. (23), blocks many ␣ M ␤ 2 -mediated functions including ICAM-1, Fg, and C3bi binding and neutrophil aggregation (17). Consistent with these reports, 2LPM19c inhibited the adhesion of wild-type ␣ M ␤ 2 to dOva ( Fig. 2A). However, the mAb had no effect on the adhesion of Upon introduction of the cDNAs into 293 cells, all three mutant and wild-type ␣ M ␤ 2 were expressed as functional heterodimers on the cell surface as evidenced by their ability to: 1) react with mAbs to both the ␣ M and ␤ 2 subunits, 2) bind NIF with similar affinities, 3) rosette with C3bi, and 4) adhere to dOva. As the latter three properties are attributes of ␣ M ␤ 2 which are not shared with ␣ L ␤ 2 (13), the region composed of H(1), L(1-B), and L(2-3) is not required for binding of these representative ligands.

The H(1), L(1-B), and L(2-3) Mutants Map the Epitope for the Function Blocking mAb 2LPM19c-mAb 2LPM19c, which recognizes a conformational epitope in the I domain
The ability of the L(1-B) and L(2-3) region to modulate the activation state of ␣ M ␤ 2 was demonstrated by the following observations: 1) segment switching at either site converted ␣ M ␤ 2 into a constitutively active receptor; i.e. compared to wild-type ␣ M ␤ 2 , the adhesive of the two mutants to Fg increased by about 10-fold. A similar increment in adhesion of neutrophils to Fg is observed upon activation of the cells (24). 2) The two mutant cell lines acquired a distinct morphology. They adhered and spread upon coverslips, formed lamellipodia extensions, and displayed a very flattened morphology. These changes indicate that the activating mutations could transmit outside-in signals, such that the cytoskeletal network within the cell reorganized.
3) The 2LPM19c mAb to ␣ M blocked functions of wild-type but not the two mutant receptors which lacked the 2LPM19c epitope as assessed by FACS analysis. Thus, the inhibitory effects of the mAb is likely to arise from its allosteric effect on the receptor conformation. This explanation suggests that the mAb has similar recognition requirements but the opposite effect of botrocetin on vWF activation. 2LPM19c recognizes H(1), L(1-B), and L(2-3), and these segments are homologous to two critical regions in vWF for botrocetin recognition (residues 539 -553 and 569 -583) (7) (Fig. 5). Thus, the binding sites for 2LPM19c and botrocetin correspond to overlapping regions although the sites are not precisely identical. The alternative possibility that 2LPM19c sterically blocks ligand binding cannot be excluded but seem unlikely in view of the distance of the epitope from the "MIDAS" motif of the receptor (1).
In addition to the two loop regions of the ␣ M I-domain, it is likely that other spatially adjacent segments also might influence receptor activation. In particular, the region composed of amino acids 117-130 of ␣ M is a reasonable candidate. In the crystal structure of the ␣ M I-domain, this segment, which lies outside the I domain and in front of the ␤-sheet A, is in close spatial proximity to the L(1-B) and L(2-3) segments (Fig. 5B). Mutations in the homologous region of vWF induce its activation, including two vWF type IIB mutations, Pro 503 3 Leu (25) and His 505 3 Asp (14).
The mechanism by which the loop mutations in ␣ M ␤ 2 or in vWF induce activation remains uncertain. Qu and Leahy (5) proposed that the mutations within this region of vWF might interfere with interactions between A domains which normally suppress GPIb binding. Our data are consistent with this explanation but also support an alternative; namely, mutations in this region changed the relative conformations of the loop between ␤-sheet A and helix 1, and the loop between helix 3 and 4, thus affecting the receptor activity. The latter possibility was supported by the observation of Lee et al. (6) that the ␣ M I-domain is capable of adopting two different conformations and undergoing large structural changes. Further experiments are required to distinguish these two mechanisms.
One of the two loop mutants, ␣ M [L(2-3)], could be further activated by the ␤ 2 -activating mAb, KIM185, thus producing a "superadhesion" receptor. Mutation of a residue in the ␤ 3 subunit of ␣ IIb ␤ 3 also has been shown to produce a "superintegrin" (26). The distinction between the two loop mutants in response to KIM185 suggests the existent of different activation states of the ␣ M ␤ 2 . In our study, short amino acid segments were altered to generate these different activation states. By analogy to vWF, point mutations or polymorphisms also may lead to ␣ M ␤ 2 receptors with different functional properties. Recently, it has been shown that a polymorphism in the ␤ 3 subunit is associated with a thrombotic phenotype (27). Mutations which alter the activation states of ␣ M ␤ 2 could have broad ranging consequences including altered susceptibility to infection. We are currently searching for such naturally occurring activating mutations.