The Integrin (cid:1) L (cid:2) 2 Hybrid Domain Serves as a Link for the Propagation of Activation Signal from Its Stalk Regions to the I-like Domain*

Integrin activation involves global conformational changes as demonstrated by various functional and structural analyses. The integrin (cid:2) hybrid domain is proposed to be involved in the propagation of this activation signal. Our previous study showed that the integrin (cid:2) 2 -specific monoclonal antibody 7E4 abrogates monoclonal antibody KIM185-activated but not Mg 2 (cid:3) / EGTA-activated leukocyte function-associated anti-gen-1 (LFA-1; (cid:1) L (cid:2) 2 )-mediated adhesion to ICAM-1. Here we investigated the allosteric inhibitory property of 7E4. By using human/mouse chimeras and substitution mutations, the epitope of 7E4 was mapped to Val 407 , located in the mid-region of the (cid:2) 2 hybrid domain. Two sets of constitutively active LFA-1 variants were used to examine the effect of 7E4 on LFA-1/ICAM-1 binding. 7E4 attenuated the binding of variants that have modifica-tions to regions membrane proximal with respect to the (cid:2) 2 hybrid domain. In contrast, the inhibitory activate adhesion. was MOLT-4 assay, ester ng/ml) as

Integrins are key proteins involved in cell-cell and cell-matrix interactions, mediating essential biological processes such as embryogenesis, the immune response, and inflammation (1,2). They are heterodimeric type I membrane glycoproteins formed by noncovalent association of an ␣ and a ␤ subunit. In human, 9 of the 18 ␣ subunits have an I (Inserted)-domain found between blades 2 and 3 of a seven-bladed ␤-propeller structure at its N-terminal end. For this particular subset of integrins, the I-domains are involved in ligand binding via their MIDAS 1 motifs (3)(4)(5)(6). The ␤ subunit is linearly organized into an N-terminal PSI-domain (for Plexins, Semaphorins, and Integrins) (7), a spacer region, an I-domain like structure (also known as the I-like domain or ␤ A-domain), a mid-region, a cysteine-rich region containing four tandem IEGF (Integrin Epidermal Growth Factor)-domains, and a terminal domain followed by the transmembrane and cytoplasmic segment (Fig.  1A). In the integrins without an I-domain in their ␣ subunits, the ␤ I-like domain participates directly in ligand binding (8). However, in the integrins with an I-domain in their ␣ subunits, the ␤ I-like domain may, in addition, serve a regulatory role in the integrin-mediated adhesion (9).
Previously, we constructed integrin ␤ 2 /␤ 7 chimeras in which the N-terminal region (NTR, the combined PSI domain and spacer segment), I-like domain, and the mid-region of the ␤ 2 subunit were replaced with those of the ␤ 7 subunit (10). The epitopes of five ␤ 2 -specific mAbs were mapped by using these chimeras and were found to require both NTR and mid-region for their expression. Coupled with the observation that removal of the I-like domain did not affect the folding of the NTR-midregion complex, we concluded that the NTR and mid-region interact extensively. This conclusion was in agreement with the crystal structure of ␣ V ␤ 3 , in which the I-like domain of the ␤ 3 subunit is connected via its N and C termini to a hybrid domain formed by the spacer and mid-region, assuming an I-set immunoglobulin fold (11). The function of the hybrid domain has been suggested to involve the shape shifting of the I-like domain in ␣ 5 ␤ 1 (12,13). Various mAbs (activating, inhibitory, and reporter) against integrins have provided much information on the regulation of integrin function (1,14,15). Therefore, in our previous study, the properties of the five anti-␤ 2 mAbs to the hybrid domain were investigated. Of note, 7E4 abrogated mAb KIM185-activated LFA-1-mediated adhesion to ICAM-1 but not Mg 2ϩ /EGTA-activated LFA-1/ICAM-1 binding. Although the end effect of both activating agents on LFA-1 is similar, their modes of action differ. Whereas the epitope and site of action of KIM185 lie in the IEGF-4 and terminal domain of the ␤ 2 subunit (16), the effect of Mg 2ϩ / EGTA lies predominantly on the MIDAS and ADMIDAS of the ␤ subunit (1). Taken together, we reasoned that the properties exhibited by 7E4 provide a unique opportunity to address how the activation signal is propagated via the hybrid domain of LFA-1.
To this end, we extended our investigation by fine mapping the epitope of 7E4 using human/mouse chimeras and amino acid substitution analyses. We were able to locate the key residue for the 7E4 mAb. In addition, the allosteric inhibitory property of 7E4 on LFA-1/ICAM-1 binding was examined by using ␣ L and ␤ 2 variants, which confer constitutive activation of the resultant heterodimer.
cDNA Expression Constructs-The ␣ L and ␤ 2 cDNA in the expression vector pcDNA3 (Invitrogen) were described previously (22). ␤ 2 human/ mouse chimeras were constructed using standard molecular biology techniques. The mouse ␤ 2 fragments Mo(Met 1 -Asp 77 ), Mo(Met 1 -Glu 298 ), and Mo(Glu 298 -Asn 584 ) were generated by PCR using a mouse cDNA library as template. An EcoRV site is found in the mouse CD18 cDNA inclusive of the codons for the invariant residues Asp-Ile at positions 77-78 (see Fig. 1B). An EcoRV site was introduced into the equivalent position in the human CD18 cDNA. Mo(Met 1 -Asp 77 ) was inserted into the NheI and EcoRV sites of CD18-pcDNA3.1/zeo(ϩ) that had their corresponding human ␤ 2 fragments removed. Mo(Met 1 -Glu 298 ) was inserted into the NheI and EcoRI sites and Mo(Glu 298 -Asn 584 ) into the EcoRI and SacII sites of CD18-pcDNA3.1/zeo(ϩ) (23) that had their corresponding human ␤ 2 fragments removed. They were named ␤ 2,Hu/Mo B, ␤ 2,Hu/Mo C, and ␤ 2,Hu/Mo D, respectively. ␤ 2,Hu/Mo A was constructed by inserting Mo(Glu 298 -Asn 584 ) into ␤ 2,Hu/Mo C that had the corresponding fragment removed. Amino acid substitutions on the mouse or human ␤ 2 constructs were made using QuikChange TM sitedirected mutagenesis kit (Stratagene, La Jolla, CA). The constructs bearing C36S (24), N351S (25), R593C (20), ␤2C23*, a ␤ 2 truncation mutant in which the codon of Cys 483 (the 23rd of the 56 extracellular cysteines of the ␤ 2 subunit) was converted to a stop codon (26), and ␣ L,C-C , with an engineered disulfide bond at K312C and K319C (27) and human ␤ 2 /␤ 1 chimera ␤ 2 /␤ 1 NV1 (23) were described previously. All constructs were verified by sequencing (DNA Sequencing Facility, Department of Biochemistry, University of Oxford, Oxford, UK). The initiation methionine is assigned number 1 in the protein sequence.
Cell Adhesion Assay-For analysis of LFA-1-mediated adhesion to ICAM-1, the wells of Polysorb microtiter plates (Nunc, Roskilde, Denmark) were coated with 100 l/well of goat anti-human IgG (Fcspecific) at 5 g/ml in 50 mM sodium bicarbonate buffer (pH 9.2) for 16 -20 h at 4°C. Nonspecific binding sites were blocked with 0.5% (w/v) bovine serum albumin (Sigma) in phosphate-buffered saline for 30 min at 37°C. Thereafter, 50 l of 1 g/ml ICAM-1/Fc in phosphate-buffered saline containing 0.1% (w/v) bovine serum albumin was added to each well of the coated plates and incubated for 2 h at room temperature. Subsequent steps were described previously (10,21). MHM24 was used at 10 g/ml to inhibit LFA-1-mediated adhesion. 7E4 was studied for its effect on adhesion at three different concentrations of 20, 10, and 3 g/ml. Mg 2ϩ /EGTA (5 mM MgCl 2 and 1.5 mM EGTA) was used to activate LFA-1-mediated adhesion. Total cell fluorescence was determined using a fluorescence plate reader (Cytofluor 4000; PerSeptive Diagnostics, Framingham, MA). For MOLT-4 (ATCC, Manassas, VA) ICAM-1 binding assay, the phorbol ester PMA (50 ng/ml) was employed for treatment of cells as described previously (10).
Flow Cytometry-Cells were incubated with primary mAb, at 10 g/ml unless otherwise stated, in RPMI 1640 for 1 h at 4°C. They were then washed twice and incubated with fluorescein isothiocyanate-conjugated sheep anti-mouse F(abЈ) 2 secondary antibody (1:400 dilution; Sigma) for 45 min at 4°C. Stained cells were washed once and fixed in 1% (v/v) formaldehyde in phosphate-buffered saline. Cells were analyzed on a FACScan flow cytometer (BD Biosciences). Data were analyzed using the CellQuest software (BD Biosciences). Expression index was calculated by EI ϭ %GP ϫ MFI/100, where EI is expression index, %GP is % cells gated positive, and MFI is mean fluorescence intensity.
Positive gates were set to include 5% of the brightest cells stained with the irrelevant antibody KB43.

RESULTS
Epitope Mapping-In a previous study (10), we reported the mapping of several anti-␤ 2 mAbs using ␤ 2 /␤ 7 hybrid chimeras. The epitope of mAb 7E4 was further studied by using human/ mouse ␤ 2 chimeras (␤ 2,Hu/Mo ). Four chimeras were generated: ␤ 2,Hu/Mo A contains the Met 1 to Asn 584 residues of the mouse ␤ 2 sequence; ␤ 2,Hu/Mo B contains the Met 1 to Asp 77 residues; ␤ 2,Hu/Mo C contains the Met 1 to Glu 298 residues; and ␤ 2,Hu/Mo D contains the Glu 298 to Asn 584 residues (Fig. 1A). These cDNAs were co-transfected with ␣ L cDNA into COS-7 cells and analyzed by flow cytometry. The mAb KIM185 has its epitope located at IEGF-4/TD (Tail Domain) of the ␤ 2 subunit (16); hence it was included to monitor surface expression of the ␤ 2 wild-type and chimeric subunits. The expression of the 7E4 epitope was greatly reduced on the ␣ L ␤ 2,Hu/Mo A transfectants (Table I)  and mid-region (Leu 363 -Cys 445 ) of the human and mouse ␤ 2 subunits. Amino acids that differ are shown by *. The I-like domain is omitted in the alignment but has its location indicated between the spacer and mid-region. In all cases, the initiation methionine is assigned the number 1.
was detected on ␣ L ␤ 2,Hu/Mo D transfectants. Our previous study using ␤ 2 /␤ 7 chimeras showed that the expression of the 7E4 epitope requires the ␤ 2 residues both in the spacer and the mid-region, and expression is abolished if either region was substituted with the corresponding one from ␤ 7 (10). In the data presented here, its expression is only affected if the midregion is from the mouse ␤ 2 subunit. Thus, we may conclude that the species-specific residues for the epitope reside in the mid-region, but residues in the spacer common to the human and mouse ␤ 2 subunits are also required for the expression of the 7E4 epitope.
The number of residues that are different between the human and mouse ␤ 2 subunits in the mid-region are limited (Fig.  1B); therefore, we introduced the mouse residues into a human ␤ 2 subunit by site-directed mutagenesis. As before, the cDNA constructs were transfected into COS-7 cells, and surface expression of the epitope was monitored by flow cytometry. Of the mutants generated and tested, the 7E4 epitope was abolished with ␤ 2,Hu (V407N) but not with all other substitutions (Table  II).
"Knock-in" mutants were constructed on the ␤ 2,Hu/Mo A background, in which the PSI, I-like domain, hybrid domain (made up of the spacer and mid-region), IEGF-1, -2, and -3, and part of IEGF-4 (Fig. 1A) were those of the mouse ␤ 2 subunit. The 7E4 epitope expression was restored in the ␤ 2,Hu/Mo A(N407V) variant which had the mouse Asn changed to a human Val residue (Table III). Taken together, the above experiments suggest that the key residue for the 7E4 epitope is the valine residue at position 407 of the human ␤ 2 subunit.
Effect of 7E4 on the Adhesion of MOLT-4 and Transfectants Expressing Wild-type and Constitutively Active LFA-1 to ICAM-1-Previously we have shown that the mAb 7E4 blocks KIM185-activated LFA-1-mediated adhesion to ICAM-1 but not adhesion activated by Mg 2ϩ /EGTA (10). In this work, we determined whether 7E4 can block the adhesion of several "constitutively active" LFA-1 to ICAM-1. ␤ 2 NV1 is a ␤ 2 /␤ 1 chimera in which the cysteine-rich region of the ␤ 2 subunit was replaced that that of ␤ 1 (23). ␤ 2 R593C is a mutant found in an LAD-1 patient that supported expression of a constitutively active LFA-1 in an in vitro transfection system (20). Since then, two other mutations have been investigated, ␤ 2 C36S described for the Irish Setter (24) and ␤ 2 N351S for another LAD-1 patient (25), both were found to support the expression of constitutively active LFA-1. 2 7E4 was found to abolish the adhesion of 293T transfectants expressing the LFA-1 variants 2 S. K. A. Law, unpublished observations. a Epitope expression is measured using EI ‫؍‬ %GP ϫ MFI, where EI is expression index, %GP is % cells gated positive, and MFI is mean fluorescence intensity. Values are calculated using CellQuest software and presented with respect to the expression of the same mAbs on the ␣ L ␤ 2 wild type.
b The mAb KIM185 maps to the C-terminal of the human ␤ 2 subunit and is used as a positive control (16).
c The mAb MEM48 maps to the C-terminal of the human ␤ 2 subunit and is used as a positive control in place of KIM185 (16,37).

FIG. 2. The effect of 7E4 on the adhesion of LFA-1 variants to ICAM-1 in the absence of activating agent (A) or presence of activating agents Mg 2؉ /EGTA (B).
␣ L ␤ 2 C36S, ␣ L ␤ 2 R593C, and ␣ L ␤ 2 NV1 to ICAM-1 with higher efficiencies at high 7E4 concentrations. In contrast, 7E4, even at 20 g/ml, has no significant effect on the adhesion of the ␣ L ␤ 2 N351S variant to ICAM-1 ( Fig. 2A). By using Mg 2ϩ /EGTA as the activating reagent, the blocking effect of 7E4 to ICAM-1 was minimal for all constructs (Fig. 2B). All adhesion was specific as they can be blocked with anti-␣ L antibody MHM24. In addition, similar results were obtained when LFA-1 was expressed on COS-7 cells (data not shown).
In a separate experiment, we investigated the effect of 7E4 on the adhesive ability of transfectants expressing constitutively active LFA-1. ␤ 2 C23* is a ␤ 2 truncation mutant in which the codon of Cys 483 was converted to a stop codon. When combined with wild-type ␣ L , the resultant heterodimer was found to be constitutively active (26). The ␣ L,C-C mutant has the I-domain locked in the active conformation and the LFA-1 that it forms with wild-type ␤ 2 is constitutively active (27,30). 7E4 abrogated adhesion to ICAM-1 of ␣ L ␤ 2 C23* but not ␣ L,C-C ␤ 2,wt (Fig. 3).
7E4 was also tested for its effect on the phorbol ester PMAtreated MOLT-4 adhesion to ICAM-1 (Fig. 4). PMA activates protein kinase C, which is reported to activate LFA-1 via phosphorylation of its ␤ 2 cytoplasmic tail (31,32). PMA significantly augmented the binding of MOLT-4 to ICAM-1. Noteworthy, 7E4 effectively abrogated this binding similar to that observed in the presence of MHM24, an ␣ L -specific function-blocking mAb.
It is possible that 7E4 fails to block the adhesion of the LFA-1 variants ␣ L ␤ 2 N351S and ␣ L,C-C ␤ 2,wt to ICAM-1 because the epitope is not expressed on these variants. This is not the case as shown in Table IV; the epitope is expressed on wild-type and all the variants described in this article.
Molecular Modeling of the ␤ 2 Subunit Expressing 7E4 Epitopes-By using the coordinates of the integrin ␣ V ␤ 3 as a template, a model of the hybrid and I-like domains of the ␤ 2 subunit was generated using the molecular modeling program MODELLER (Fig. 5). The 7E4 epitope Val 407 is positioned at the N-terminal end of the E strand of the hybrid domain (11) and can be visualized to be close to the I-like domain of the ␤ 2 subunit. DISCUSSION Based on electron micrographs of the representative integrins ␣ IIb ␤ 3 and ␣ 5 ␤ 1 , the general structure of the integrin molecule had been taken as a two-pronged plug having a glob- The absolute level of LFA-1 expression, as indicated by the heterodimer-specific mAb IB4, is quite variable among the different transfectants. The expression levels of LFA-1 with the ␤ 2 C36S and ␤ 2 NV1 are consistently lower than the wild type and other LFA-1 variants. Staining of IB4 was done at 10 g/ml.
c Cell staining of 7E4 were carried out in three different concentrations of mAb at 20, 10, and 3 g/ml, respectively. d Staining of KB43, an irrelevant mAb, was carried out at 10 g/ml. ular head and two legs anchoring them into the membrane (33)(34)(35). From numerous functional studies and other studies, it was also generally accepted that conformational change is an integral component in regulating integrin adhesiveness (36). The conformational change had not been defined, but much experimental data pointed to the model that a resting integrin is highly constrained and a release of these constraints results in activation. The x-ray solution of ␣ V ␤ 3 showed a structure in which the two legs were bent, folding the integrin heterodimer into a compact conformation (11). Extensive analyses of the electron micrograph images of the recombinant extracellular fragment of the ␣ V ␤ 3 integrin in different activation states were performed (33). It was found that the resting integrin in Ca 2ϩ /Mg 2ϩ corresponds to the compact bent conformer, whereas the active integrin induced by Mn 2ϩ or the cyclic RGD ligand corresponds to the extended conformer. These observations were in line with the "switchblade" model of integrin activation proposed based on the NMR analysis of the structure of the IEGF-2 and -3 of the ␤ 2 integrin subunit (37).
Previously, we have reported the effect of the mAb 7E4 on LFA-1-mediated adhesion of MOLT-4 to immobilized ICAM-1 (10). If LFA-1 was activated by the mAb KIM185, 7E4 can block the adhesion. However, if Mg 2ϩ /EGTA was used to activate LFA-1, 7E4 has no effect. The switchblade model suggests that resting LFA-1 is presumably bent, and activation involves the transition of the molecule to the extended conformation. Support for this model for both KIM185 and Mg 2ϩ /EGTA-mediated activation of LFA-1 can be found with the study of the mAb KIM127. The epitope of the mAb KIM127 has been mapped to the IEGF-2 and IEGF-3 of the integrin ␤ 2 subunit (16,38). Superposition of the NMR structure of these two domains on the ␣ V ␤ 3 template (11) showed that the residues forming the epitope are buried (37). It has been shown that the expression of the KIM127 epitope is associated with the activation of LFA-1, which would require the binding of the mAb KIM185 (39) or the incubation of the cells in Mg 2ϩ /EGTA. 2 Taken together, both KIM185 and Mg 2ϩ /EGTA activate LFA-1 by exposing the KIM127 epitope, which may be interpreted as the conversion of the bent conformer into the extended conformer. KIM185-activated adhesion must start with the binding of the antibody to LFA-1 at its epitope site in the IEGF-4 and ␤TD domain of the integrin ␤ 2 subunit (16). This binding induces a conformational change rendering LFA-1 the capacity to bind ICAM-1 at its ligand-binding site that is located at the I-domain of the ␣ L subunit. Binding of 7E4 to the hybrid domain may thus be viewed as a blockade in communication between the activation site, i.e. KIM185-binding site, and the ligandbinding site. Mg 2ϩ /EGTA, on the other hand, is likely to induce the conformational change of LFA-1 at the MIDAS and ADMI-DAS sites of the I-like domain of the ␤ 2 subunit (1). Because the divalent cation-binding sites and the ligand-binding sites are both distal to the hybrid domain, 7E4 does not affect ICAM-1 binding induced by Mg 2ϩ /EGTA. If 7E4 blocks LFA-1 mediated adhesion by preventing signals originating from regions proximal to the membrane to the ligand-binding site, we should be able to test this hypothesis by using a number of LFA-1 variants that are constitutively active. Indeed, our results presented here showed segregation of these variants. The constitutive activity of LFA-1 due to modification on the membrane side of the hybrid domain is abolished by 7E4. These variants are ␣ L ␤ 2 R593C (20), ␣ L ␤ 2 NV1 (23), ␣ L ␤ 2 C23* (26), and LFA-1 activated by KIM185 that binds to the "foot" of the ␤ 2 subunit. Most interestingly, also in this list is the LFA-1 variant with the C36S mutation (24) in the PSI domain. Because the PSI domain is folded back under the hybrid domain, the mutation C36S can therefore be considered to be located on the membrane proximal side of the hybrid domain. Furthermore, activation of LFA-1 by PMA, originating from the inside the cell, was attenuated by 7E4. The corollary is that LFA-1 with modification distal to the hybrid domain should not be affected by 7E4 because they exert their effects to the ligand-binding site beyond the hybrid domain. They include the activation by Mg 2ϩ /EGTA (10), the LAD mutation N351S in the I-like domain of ␤ 2 (25), and the engineered disulfide lock in the I-domain of the ␣ L subunit (27,30).
The ␤ I-like domain is proposed to regulate the activity of the ␣ I-domain. It has been known that mutations in the MIDAS residues in the I-like domain of the ␤ 2 integrin subunit would not affect LFA-1 expression, but the LFA-1 variants were incapable of ligand binding, irrespective of being treated with activating mAbs or Mg 2ϩ /EGTA (40 -42). Subsequently, an invariant glutamate residue, located in the linker between the I-domain and the ␤-propeller in all I-domain containing ␣ subunits had been identified. Mutation of this glutamate residue, Glu 335 in the ␣ L subunit, rendered the resultant LFA-1 inactive (43). Taken together, these results suggest that an integral part of LFA-1 activation may lie in the binding of the ␤ 2 MIDAS motif with Glu 335 in ␣ L . This hypothesis received strong support from experiments in a recent publication in which the Glu 335 of the ␣ L subunit, and either one of the two residues on the ␤ 2 subunit, Tyr 137 or Ala 232 , were converted to cysteines (44). The two resultant LFA-1 variants were found to have the ␣ L and ␤ 2 subunits disulfide-linked and are constitutively active with respect to ICAM-1 binding. Because both Tyr 137 and Ala 232 are in the proximity of the MIDAS motif of the ␤ 2 I-like domain, the LFA-1 variants may be considered as being "locked" into an active conformation similar to when the ␤ 2 MIDAS binds to the invariant Glu in the ␣ L subunit.
The epitope of 7E4 is complex and requires the split regions of the hybrid domain for expression (10). By using single residue substitutions described in this article, the species-specific residue was identified to be Val 407 . Other residues, common to both human and mouse, must also contribute to the conformation of the epitope. Modeling on the ␣ V ␤ 3 template showed that it is close to the C-terminal helix of the I-like domain (Fig. 5). Thus, binding of 7E4 to the hybrid domain may alter its capacity to interact with the I-like domain, preventing LFA-1 from being activated perhaps by disabling the binding of the ␤ 2 MIDAS motif to the invariant glutamate in ␣ L . In recent studies on the ␣ 5 ␤ 1 integrin (12,13), it was suggested that the "swinging" of the hybrid domain away from the I-like domain is associated with the integrin being activated. In this article, our data do not provide support nor contradict this particular relative motion of the hybrid domain in integrin activation. However, it does concur that the hybrid domain plays a role in the regulation of LFA-1 activity.