Disruption of the Integrin αLβ2 Transmembrane Domain Interface by β2 Thr-686 Mutation Activates αLβ2 and Promotes Micro-clustering of the αL Subunits*

Integrins are type I heterodimeric cell adhesion molecules that mediate a wide array of biological processes. Integrin bidirectional signaling allows communication between the cell interior with its microenvironment. The integrin transmembrane domains (TMs) are the transducers of activation signal that is relayed from the cytoplasmic domains to the distal ligand binding site located in the ectodomain of the integrin and vice versa. In this study, we showed that the disruption of the αLβ2 TMs by mutation of a key interface residue Thr-686 in the β2 TM promoted αLβ2 activation with ICAMs binding properties that are reminiscent of an intermediate affinity receptor. The activated αLβ2 TM mutants, however, showed minimal reactivity with the reporter mAb KIM127 that recognizes a highly extended αLβ2. Two models of αLβ2 TM interaction were proposed previously. One with GXXXG-type interaction, and another that is based on TM cysteine-scanning analyses. Our data are consistent with a GXXXG-type interaction of the αLβ2 TMs. Finally, we observed by FRET analyses that perturbation of the αLβ2 TMs by β2 Thr-686 mutation facilitated αL micro-cluster formation. This was diminished by linking the αLβ2 TMs with a disulfide bond, which served to clasp the TMs. These data suggest that disruption of the TM interface changes αLβ2 ligand binding affinity, and it may contribute to αL micro-cluster formation.

The metazoan cell adhesion molecules integrins are a large family of type I membrane proteins form by noncovalent and specific pairing of ␣ and ␤ subunits (1). Integrins have a large ectodomain, two transmembrane domains (TMs), 3 and short cytoplasmic tails with the exception of the ␤4 subunit. Devoid of intrinsic enzymatic activities, the integrins trigger intracel-lular signaling circuits by selective recruitment of cytosolic molecules to their cytoplasmic tails (2). The phosphorylations of integrin cytoplasmic tails are also important events in modulating integrin activity and signaling capacity (3,4). Together, these cytosolic interactions and post-translational modifications are required for integrin bi-directional signaling that allows communication between the interior of the cell with its external microenvironment. The TMs of the integrins provide the connection for signal transmission between their cytoplasmic tails and their ectodomains. Electron cryomicroscopy study of human integrin ␣IIb␤3 suggests that the integrin TMs interact with each other in the resting or low affinity state (5). Conceivably, structural changes in the integrin ectodomain elicited by ligand binding or the separation of the integrin cytoplasmic tails by cytosolic molecule such as talin will trigger reorientation or/and separation of the TMs (6), events that are required to relay signal from one end of the integrin molecule to the other. Indeed, in the absence of activation signal impinging on the integrin ectodomain or its cytoplasmic tails, the disruption of the TM interface alone is sufficient to promote ␣IIb␤3 activation and intermolecular homomeric interactions (7)(8)(9). In line with these observations, the introduction of a disulfide bond linking the ␣ and ␤ TMs constrained the movement of the TMs, which blunted both inside-out and outside-in signaling in ␣IIb␤3 (10,11).
To fully characterize integrin bi-directional signaling, the structure of the TMs, and more importantly the orientation of the ␣ and ␤ TMs with respect to each other require investigations. Based on computational studies, it was proposed that a GXXXG-like motif in the ␣ and ␤ TMs, in particular those of ␣IIb␤3, favors interactions similar to that of the homodimer glycophorin A (GpA) (12,13). However, different models for the packing of integrin ␣IIb␤3 TMs have also been suggested (5,10). These models of TM heterodimers may represent distinct stages of ␣ and ␤ TM rearrangements during integrin activation or affinity change (10,12,14,15). Another interesting but debatable concept is the involvement of TM in the process of integrin clustering. Homo-oligomerization of ␣IIb and ␤3 TMcytoplasmic tail fragments in dodecylphosphocholine (DPC) micelles were detected by methods of ultracentrifugation, NMR, and SDS-PAGE (16). The same group also reported dimerization of the ␣IIb TM by TOXCAT assay that measures oligomerization in the inner membrane of Escherichia coli (17). However, using the GALLEX assay, a LexA DNA binding domain/lacZ reporter system in specific E. coli strains, the integrin ␤7 TM but not the ␤1 or ␤3 TM showed strong propensity to form homo-oligomers comparable to that of the GpA TMs; homo-oligomerization of ␣IIb was only weakly detected (18). Recently, NMR spectroscopy analysis of ␤3 TM peptide embedded in bicelles and micelles also showed monomeric TM segments (19). In CHO cell system in which full-length ␣IIb␤3 was investigated, disruption of the ␣IIb␤3 TM interface by the point mutation G708N in ␤3 promoted receptor activation, homo-oligomerization, and facilitated ␣IIb␤3 macro-cluster formation (7). In a separate study, the same mutation increased the ligand-binding affinity of ␣IIb␤3, but failed to form macroclusters (8). What accounts for these different observations is not known.
Integrin ␣L␤2 (LFA-1, CD11aCD18) is being studied extensively because of its primary function in immune homeostasis involving leukocyte adhesion, migration, and proliferation (20 -22). Many studies investigate the ectodomain conformation of ␣L␤2, and the membrane proximal events impinging on its cytoplasmic tails with regard to its functional regulation (23,24). Although computational and modeling studies provide a guide to ␣L and ␤2 TM interactions that are possible, they remain to be tested in a full-length receptor in a cell-based system (15). Unlike ␣IIb␤3, much less is known about the ␣L␤2 TM interface. Previously, we showed by TM exchange that replacing ␤2 TM with ␤4 TM generated an ␣L␤2 mutant that exhibited wild type activity (25). This was in contrast to ␣L␤2 mutants with ␤2 TM replaced by other ␤ TMs, many of these were activated. In this study, we show that the ␤2 TM residue Thr-686 within the GXXXG-like motif contributes toward the formation of the ␣L␤2 TM interface. Mutation of Thr-686 induced ␣L␤2 activation, and promoted ␣L micro-cluster formation.
Expression Plasmids-The positions of the integrin amino acids are numbered accordingly to Barclay et al. (35). Expression plasmids encoding full-length human integrin ␣L and ␤2 subunits were described previously (36). The integrin TM chi-mera ␤24, in which the ␤2 TM is replaced with the corresponding TM segment from ␤4, was reported previously (25). All point mutations were generated using the QuikChange TM sitedirected mutagenesis (SDM) kit (Stratagene, La Jolla, CA) with relevant primers. For fluorescence resonance energy transfer (FRET) experiments, the expression plasmids containing ␣L monomeric CFP (mCFP) and ␤2 monomeric YFP (mYFP) have been described, and ␣LmYFP was generated by the same procedure (25). Briefly, mCFP was fused to the C terminus of the ␣L cytoplasmic tail with a linker that contains the amino acids GPVAT, and the mYFP was fused to the C terminus of the ␤2 cytoplasmic tail with a linker that contains amino acids GGP-VAT as described (6). All expression plasmids were verified by sequencing (Research Biolabs, Singapore).
Cell Culture and Transfection-The 293T and K562 cells were obtained from ATCC. 293T cells were cultured in Dulbecco's modified Eagle's medium containing 10% (v/v) heat-inactivated fetal bovine serum (HI-FBS), 100 international units/ml of penicillin and 100 g/ml of streptomycin (Hyclone, South Logan, UT). K562 cells were cultured in RMPI 1640 containing 10% (v/v) HI-FBS, 100 international units/ml of penicillin and 100 g/ml of streptomycin. The Polyfect transfection reagent (Qiagen) was used for 293T transient transfection. K562 cells were transiently transfected by the method of electroporation using the Amaxa Nucleofector device, and reagents per the manufacturer's instructions (Amaxa Gmbh, Germany).
Flow Cytometry-The surface expression levels of integrins on transfectants were determined by immunostaining cells with mAb IB4 (20 g/ml) and secondary fluorescein isothiocyanate-conjugated sheep anti-mouse IgG (Sigma) (1:400), followed by analysis on a FACS Calibur with the software CellQuest (Becton Dickinson Biosciences, Mountain View, CA) essentially as described previously (37). Expression index (EI) was calculated by % gated positive ϫ geo-mean fluorescence intensity (37).
Biotin Labeling of Cell Surface Proteins and Immunoprecipitation-In brief, transfectants were washed twice in PBS and incubated in PBS containing 0.5 mg/ml sulfo-NHS-biotin (Pierce) for 15 min at room temperature. The biotin-labeling reaction was terminated by washing cells in PBS containing 10 mM Tris-HCl, pH 8.0 and 0.1% (w/v) bovine serum albumin. For the analyses of ␣L␤2 conformation, cells were maintained in Dulbecco's modified Eagle's medium containing 5% (v/v) HI-FBS, 10 mM HEPES (pH 7.4), and 2 g each of mAb MHM23 or KIM127 at 37°C for 30 min. 5 mM MgCl 2 and 1 mM EGTA were included for the ␣L␤2 activating condition. Unbound mAb was removed by washing cells twice in medium. Cells were lysed in lysis buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% (v/v) Nonidet-P 40) containing protease inhibitors at 4°C for 30 min. Immunoprecipitation was performed using rabbit anti-mouse IgG (Sigma) coupled to protein A-Sepharose beads (Amersham Biosciences) as described previously (37). Bound proteins were resolved on a 7.5% SDS-PAGE gel under reducing conditions (ϩDTT), and electroblotted onto Immobilon P membrane (Millipore, Bedford, MA). Biotinylated protein bands were probed with streptavidin-horseradish peroxidase and detected by enhanced chemiluminescence (ECL) using the ECL-plus kit (Amersham Biosciences).
For the analyses of disulfide bond formation between the ␣L and ␤2 TMs by the introduction of a cysteine pair, biotin labeling of ␣L␤2, immunoprecipitation, and protein bands detection were performed as described above with the following changes. The mAb IB4 (2 g) was used for immunoprecipitation. Proteins were resolved on 7% SDS-PAGE gels under reducing (ϩDTT) or non-reducing (ϪDTT) conditions.
Cell Adhesion Assays-Adhesion of transfectants to immobilized ICAMs was performed essentially as described (37). Polysorb microtiter well (Nunc, Roskilde, Denmark) was coated with 0.5 g of goat anti-human IgG Fc-specific (Sigma) in 50 mM bicarbonate buffer (pH 9.2), and nonspecific binding sites blocked with 0.5% (w/v) bovine serum albumin in PBS. Thereafter, ICAM-Fc (50 ng) in PBS was added to each well and incubated at room temperature for 2 h. Cells were labeled with 2Ј,7Ј-bis-(2-carboxyethyl)-5-(and-6) carboxyfluorescein (Molecular Probes, Eugene, OR), and ϳ2 ϫ 10 4 cells in medium containing 5% (v/v) HI-FBS and 10 mM HEPES (pH 7.4) were added to each ICAM-coated well and incubated at 37°C for 30 min in a humidified 5% CO 2 incubator. The activating agents mAb KIM185 (10 g/ml) and/or Mg/EGTA (ME)(5 mM MgCl 2 , 1 mM EGTA) were included in the experiments. Non-adherent cells were removed by washing the wells twice with medium. The fluorescence signal that corresponds to the number of adherent cells was measured with a fluorescent plate reader (FL600) (Bio-Tek Instruments, Winooski, VT). In all cases, the level of adhesion to the ICAMs was determined and expressed as relative adhesion to ligand with respect to that of wild-type ␣L␤2 in the absence of activation.
Fluorescence Resonance Energy Transfer (FRET) Analyses-For FRET experiments, K562 transfectants were adhered onto poly-L-lysine coated glass slides by brief spinning. Acceptor photobleaching, FRET (38) was performed on a Zeiss LSM510 confocal microscope (Carl Zeiss, Inc., Thornwood, NY) to detect ␣L micro-clusters using similar procedures as described previously (25). When ␣LmCFP oligomerizes with ␣LmYFP, FRET will be detected. The settings for FRET used were: mCFP, excitation wavelength 458 nm; emission filter BP 470 -500 nm; mYFP, excitation wavelength 514 nm; emission filter LP 530 nm. Cells were visualized with an oil immersion 63ϫ objective. The mYFP of an entire cell was photobleached by scanning the entire cell 20 times using the 514 argon laser line that was set at maximum intensity. The mCFP signal of the region of interest (ROI), which is the cell membrane, was acquired before and after the photobleach. The high density of ␣L␤2 in the center of the cell (Golgi compartment) was excluded from the analyses (6). FRET efficiency (E F ) was calculated as a percentage using the equation E F ϭ (I 6 Ϫ I 5 ) ϫ 100/I 6 , where I n is the mCFP intensity at the n th time point. Bleaching was performed between the 5 th and 6 th time points. The mean noise computed as N F ϭ (I 5 Ϫ I 4 ) x 100/I 5 in which the mCFP signals at the 4t h and 5 th time points before the bleaching process was close to zero in all cases.
Immunofluorescence Co-localization Analyses-K562 cells transfected with ␣LmCFP, ␣LmYFP, and ␤2 or ␤2T686F were adhered onto poly-L-lysine slides by brief spinning, and fixed in 3.7% (w/v) paraformaldehyde in PBS for 10 min at room tem-perature. Cells were washed in HBSS, followed by incubation in HBSS containing 0.3%(v/v) Triton X-100 for 3 min at room temperature. Cells were stained with anti-Golgi matrix protein GM130 mAb (1:50 dilution) for 1 h at room temperature. Cells were washed in HBSS followed by staining with Alexa Fluor 594-conjugated goat anti-mouse antibody (1:500 dilution) (Molecular Probes) for 1 h at room temperature. Cells were mounted in Vector shield and examined on a Zeiss LSM510 confocal microscope for co-localization of integrins with GM130.
Models of ␣L␤2 TMs-The two backbone models for ␣L␤2 TMs interaction were obtained from a previous conformational search that used the TMs of all integrin types (15), where only these two models were found to be evolutionarily conserved, and consistent with previous studies (10,18).

RESULTS
Thr-686 of the ␤2 Subunit Participates in the Interaction of the ␣L␤2 TMs-We reported previously that the specific pairing of the integrin TMs is required for the proper expression of the receptor, and the maintenance of the ␤2 integrins in a low affinity state (25). Conceivably, non-permissive TM pairing leads to disrupted TM packing of the ␣ and ␤ subunits. An interesting observation was made with regard to the TM mutant ␤24 in which the ␤2 TM was replaced with the ␤4 TM. Unlike other ␤2 mutants with TM of ␤1, ␤3, ␤5, or ␤6 that generated constitutively activated receptors when combined with ␣L, the ␣L␤24 mutant showed wild-type ligand binding activity (25). The ␤2 having its TM replaced by ␤7 or ␤8 TM were excluded from the analyses because of the aberrant glycosylation of ␤27 and defective expression of ␤28. Inspection of the TM sequences of ␤1, ␤3, ␤5, and ␤6 shows a conserved Val within the GXXXG-like motif of these subunits (Fig. 1A). However, in ␤2 and ␤4 the corresponding residues are Thr and Leu, respectively. Previously, we have reported that two backbone models of interaction of ␣ and ␤ integrin TMs are evolutionarily conserved (15); a GXXXG-type (GpA-like) interaction (model I, cyan) and a model consistent with previous TM cysteine-scanning analyses (model II, green) ( Fig. 1B) (10). The ␤2 Thr-686 is positioned at the interface of the ␣L and ␤2 TMs in both models.
We examined whether ␤2 Thr-686 is a key residue in the interface of the ␣L␤2 TMs. 293T cells were transiently transfected with either wild-type ␣L␤2 or ␣L␤2T686V (Fig. 2A). The expression level of ␣L␤2T686V was comparable to wild-type ␣L␤2 as determined by flow cytometry using the mAb IB4 that recognizes ␤2 integrin heterodimers (Fig. 2B) (33). To test the activity of ␣L␤2T686V, cell adhesion assays on immobilized ␣L␤2 ligands ICAM-1 (39) and ICAM-3 (40) were performed. Cells expressing wild-type ␣L␤2 showed minimal adhesion to ICAM-1, but adhesion was markedly enhanced by the addition of the ␤2 integrin activating mAb KIM185 (Fig. 2C) (29). Importantly, cells expressing ␣L␤2T686V showed significant constitutive adhesion to ICAM-1, albeit at a lower level when compared with that of KIM185 activation. Adhesion was mediated by ␣L␤2 because it was abrogated in the presence of the mAb IB4 that is also function-blocking (25).
The investigation was extended to examine the adhesion profiles of these transfectants to ICAM-3 (Fig. 2D). It was proposed that an intermediate affinity ␣L␤2 binds ICAM-1 but not ICAM-3, and effective binding to ICAM-3 requires a high affinity ␣L␤2 (41). Binding studies with isolated recombinant ␣L inserted (I) domain, and the ICAMs showed that the binding affinity is in the order of ICAM-1ϾICAM-2ϾICAM-3 (42). Further, we showed that the conformational changes in ␣L␤2 that are required to bind ICAM-1 and ICAM-3 are different (43). In our previous studies, the adhesion of ␣L␤2 expressing cells to ICAM-3 can be induced by a combination of KIM185 and Mg/EGTA (25,41,44,45). In line with these observations, transfectant expressing wild-type ␣L␤2 only adhered significantly to ICAM-3 when both activating KIM185 and Mg/EGTA were included (Fig. 2D). Interestingly, cells expressing ␣L␤2T686V showed significant adhesion to ICAM-3 in the presence of KIM185 alone, and the level of adhesion was augmented when KIM185 and Mg/EGTA were both included. These data suggest that substitution of Thr-686 with Val in ␤2 TM promotes an activated ␣L␤2 with properties of an intermediate affinity receptor.
Based on these observations, it may be inferred that Thr-686 of the ␤2 subunit contributes to the packing of the ␣L and ␤2 TMs. To further verify these findings, we asked whether substitution of Leu-686 to Val in the mutant ␤24, which retains a wild type activity when paired with ␣L (25), could also induce an intermediate affinity receptor. The expression level of ␣L␤24 was comparable to that of wild-type ␣L␤2 and ␣L␤24L686V as determined by flow cytometry using mAb IB4 (Fig. 2E). Transfectants expressing wild-type ␣L␤2 and ␣L␤24 showed similar adhesion profiles to ICAM-1 for minimal adhesion was detected in the absence of activating KIM185 (Fig. 2F). This was consistent with our previous observations (25). By contrast, cells expressing ␣L␤24L686V showed constitutive ICAM-1 adhesion, albeit at a lower level when compared with cells treated with KIM185 (Fig. 2F). When ICAM-3 adhesion assay was performed, cells bearing wild-type ␣L␤2 and ␣L␤24 showed similar adhesion profiles. Both KIM185 and Mg/EGTA were required to induce significant cell adhesion to ICAM-3. However, cells expressing ␣L␤24L686V showed significant adhesion in the presence of KIM185 alone, which could be further enhanced by the addition of Mg/EGTA (Fig. 2G). Taken together, these data suggest that the ligand-binding properties of ␣L␤2T686V and ␣L␤24L686V are similar, and that Thr-686 in the GXXXG-like motif of ␤2 has a role in maintaining the ␣L␤2 TM interface.
It was also apparent that the precocious activity of ␣L␤2T686V was not specific to the substituted Val because substitution of Thr-686 with a conserved residue Ser or a hydrophobic and aromatic residue Phe generated receptors that were expressed at comparable levels ( Fig. 3A) and had similar ICAM binding properties as the ␣L␤2T686V mutant (Fig. 3,  B and C), although the activity of ␣L␤2T686S was comparatively less than ␣L␤2T686V and ␣L␤2T686F. We went further to assess the conformation of these receptors by performing immunoprecipitation analyses of transfectants that were surface labeled with biotin (Fig. 4). The mAb KIM127 is a conformational-sensitive mAb that recognizes the IEGF-2 fold, and reports a highly extended ␣L␤2 (31,32). A half-bent ␣L␤2 with an engineered disulfide lock in the ␤2 flexion failed to react with KIM127 (46). In the absence of activation by Mg/EGTA, wildtype ␣L␤2 and mutants reacted minimally with KIM127 with traces of unassociated ␤2 detectable that were attributed to free ␤2 subunits as reported previously (44). Addition of Mg/EGTA markedly increased the level of receptors immunoprecipitated with KIM127. The mAb MHM23 that is ␤2 heterodimer-specific reacted with wild-type ␣L␤2 and mutants even in the absence of receptor activation, and was included as a control The GXXXG-like motif is in the green box. Thr-686 (pink) of the ␤2 subunit and the corresponding residue Val (yellow) in ␤1, ␤3, ␤5, and ␤6 are shown. The amino acid number of the putative first residue in each TM sequence is shown on the left. It should be noted that the membrane boundaries of the TM segments remain to be fully defined for each subunit (19,60,61). Val-683 of the ␤2 subunit is indicated and is discussed in the latter section. B, two models of ␣L␤2 TM interactions described previously are shown (15). Model I (cyan) is consistent with a GXXXG-type (GpA-like) interaction (18). Model II (green) is consistent with the proposed TM interface of ␣IIb␤3 based on cysteine-scanning analyses (10). The side chain of Thr-686 in these models is shown as a stick. (26,27). These data suggest that the mutation of ␤2 Thr-686 promotes activation of the ␣L␤2, but does not induce an extended conformation detectable by KIM127.
A GXXXG-type Packing of the ␣L␤2 TMs-We note that the orientation of the ␣L␤2 TMs in both models are different. In these models, Thr-686 is located at the interface of the TMs but with different degrees of rotation relative to the ␣L TM (Fig. 5,  A and B). This is more pronounced when Val-683 is taken into consideration. Val-683 does not reside in the GXXXG-like motif of the ␤2 TM (Fig. 1A). In the model of a GXXXG-type interaction (cyan), Val-683 is positioned away from the interface, while in the other model (green), Val-683 is located closer to the ␣L TM (Fig. 5B). Thus we have also generated ␣L␤2V683F and examined its ligand-binding properties compared with ␣L␤2T686F and wild-type ␣L␤2. All transfectants expressed comparable levels of integrins as determined by flow cytometry analyses using mAb IB4 (Fig. 5C). When cell adhesion assays to the ICAMs were performed, cells bearing ␣L␤2V683F did not show constitutive adhesion to ICAM-1, and did not adhere significantly to ICAM-3 in the presence of a single activating agent KIM185 (Fig. 5, D  and E). It was apparent that the adhesion profiles of cells expressing ␣L␤2V683F were similar to that of cells expressing wild-type ␣L␤2. By contrast, cells expressing ␣L␤2T686F showed intermediate affinity ICAMs binding properties as discussed in the preceding sections. These data are in line with the model of a GXXXG-type interaction of the ␣L␤2 TMs.
Micro-clustering of ␣L␤2 with Disrupted TM-TM Interaction-Previous study showed that the mutation G708N of the ␤3 TM generated an activated ␣IIb␤3, and it induced receptor macro-cluster formation (7). A separate study showed affinity modulation of ␣IIb␤3 when the interface of its TMs was disrupted, but ␣IIb␤3 macro-cluster formation was not detectable (8). Our present data are in line with the concept that a disrupted TM-TM interaction changes the ligand binding affinity of the integrin. Next, we sought to determine whether the disruption of ␣L␤2 TMs interaction could induce subunit oligomerization. To detect micro-cluster formation in which the individual molecules are ϳ10 nm in the proximity of one another (47) instead of macro-cluster with a length scale of Ͼ200 nm, we made use of photobleach FRET analysis. The C terminus of the ␣L cytoplasmic tail was fused with either monomeric cyan fluorescent protein (mCFP) or monomeric yellow fluorescent protein (mYFP) generating ␣LmCFP or ␣LmYFP, respectively. These will be referred to as Irrelevant mAb (open histogram). C and D, adhesion of transfectants to immobilized ICAM-1 or ICAM-3 was performed with or without activating agents. The activating agents are mAb KIM185 and Mg/EGTA (ME). Adhesion specificity was verified using mAb IB4, which is also a function-blocking mAb. The level of adhesion to the ICAMs was determined and expressed as relative adhesion to ligand with respect to that of wild-type ␣L␤2 in the absence of activation. Data points represent means Ϯ S.D. of triplicates. E, expression levels of ␣L␤2, ␣L␤24, ␣L␤24L686V on 293T transfectants were determined by flow cytometry using mAb IB4 (shaded histogram) . Irrelevant mAb (open histogram). F and G, adhesion of transfectants expressing wild-type ␣L␤2 and mutants to the ICAMs is as described before. Data points represent means Ϯ S.D. of triplicates. EI, expression index.
␣LCFP and ␣LYFP henceforth. K562 cells were used as the surrogate cells for the transfection of ␣LCFP, ␣LYFP, and ␤2 as described previously (25), and wild-type ␣L␤2 in K562 transfectants has minimal propensity to oligomerize (6). When ␣L micro-cluster forms, the proximity of ␣LCFP with ␣LYFP (Ͻ10 nm apart) should allow FRET to be detected although there will also be populations of ␣LCFP with ␣LCFP, and ␣LYFP with ␣LYFP that will not lead to FRET (Fig. 6A). In the absence of YFP photobleach (referred to as unbleach), no significant changes in the fluorescence intensity of CFP was detected in transfectants bearing ␣LCFP, ␣LYFP, and ␤2 (Fig. 6, B and C). When photobleach was performed, a marked decreased in YFP signal was detected, but there was minimal increase in CFP signal in these cells. By contrast, cells expressing ␣LCFP, ␣LYFP, and ␤2T686F showed an increase in CFP signal after YFP photobleach. Consistent with the lack of constitutive activity in ␣L␤2V683F shown in the previous section, cells bearing ␣LCFP, ␣LYFP, and ␤2V683F, showed minimal change in CFP signal after YFP photobleach. The % FRET efficiency for each condition was also calculated as described under "Experimental Procedures" and plotted (Fig. 6D). We also examined FRET in cells transfected with ␣LCFP and ␣LYFP (Fig. 6E). FRET signal detected in these cells was low and comparable to cells expressing ␣LCFP, ␣LYFP, and ␤2. Cells expressing ␣LCFP, ␣LYFP, and ␤2T686F consistently showed an increase in FRET signal. It should be noted that the ␣L subunit may not fold properly in its ␤-propeller domain when expressed in the absence of the ␤2 subunit (48). Thus, we made use of cells transfected with ␣LCFP, ␣LYFP, and ␤2 as the control in the FRET analyses. As reported previously (6), the fluorescence detected in the central region of the cells was attributed to integrins in the Golgi because immunofluorescence staining of the Golgi matrix protein GM130 (28) showed co-localization of the integrins with GM130 (Fig. 6F). In this study, only the rim of each cell was taken as the region of interest (ROI) in all the FRET analyses. These observations suggest that mutation of Thr-686 in the ␤2 TM not only disrupts the packing of the ␣L␤2 TMs resulting in receptor activation, micro-clustering of the ␣L subunits was detected.
A Disulfide Bond Linking the ␣L␤2 TMs Attenuates ␣L Micro-clustering Induced by the Disruption of the ␣L␤2 TM Interface-To further verify that the disruption of the TMs in ␣L␤2 facilitates ␣L micro-cluster formation, a disulfide clasp in the ␣L␤2 TMs was generated to prevent TM separation, and its effect on ␣L micro-clustering examined. The ␣LLeu1067 and ␤2Ile679 were selected for mutation to Cys based on model I (GXXXG-type interaction) (Fig. 7A). First, we examined formation of the intersubunit disulfide bond within ␣L␤2. Transfectants expressing wild-type ␣L␤2, ␣L␤2T686F, and ␣LL1067C␤2T686F/I679C were surface labeled with biotin, and integrins precipitated with mAb IB4  Transfectants expressing the indicated integrins were surface-labeled with biotin. Cells were incubated in media containing the indicated mAbs, including the mAb KIM127 that reports a highly extended ␣L␤2 (32). The mAb MHM23, which reacts with ␤2 integrin heterodimers and is not activation-dependent, was included as a control (26,27). Mg/EGTA (ME) was included to induce ␣L␤2 activation. Cells were lysed and immunocomplexes precipitated as described under "Experimental Procedures." Proteins were resolved on a 7.5% SDS-PAGE under reducing conditions. The ␣L and ␤2 bands are indicated. (Fig. 7B). Proteins were resolved on reducing or non-reducing SDS-PAGE gels. Under reducing conditions, the ␣L and the ␤2 protein bands were detected. Under non-reducing conditions, the ␣L and ␤2 subunits were detected for ␣L␤2 and ␣L␤2T686F. However, a single high MW protein band that is equivalent to the MW of ␣L␤2 combined was detected in ␣LL1067C␤2T686F/I679C. Thus, a disulfide bond linking the ␣L␤2T686F TMs was generated. Next, FRET analyses were performed. As observed previously, effective FRET was detected in cells expressing ␣LCFP, ␣LYFP, and ␤2T686F after YFP photobleach as compared with cells expressing ␣LCFP, ␣LYFP, and ␤2 (Fig. 8, A and B). By introducing a disulfide clasp in ␣L␤2T686F (cells expressing ␣LL1067CCFP, ␣LL1067CYFP, and ␤2T686F/I679C), a reduction in FRET signal was detected (Fig.  8A). The reduction in FRET signal was not due to an aberrant effect of the introduced Cys because the introduction of Cys into the ␣L TM or the ␤2T686F TM alone (cells expressing ␣LL1067CCFP, ␣LL1067CYFP, ␤2T686F or cells expressing ␣LCFP, ␣LYFP, ␤2T686F/I679C, respectively) did not reduced significantly the FRET signal (Fig. 8B). FRET in cells expressing ␣LL1067CCFP, ␣LL1067CYFP, ␤2I679C was also comparable to cells expressing ␣LCFP, ␣LYFP, and ␤2. These data suggest that the disulfide clasp could effectively reduce the disrupting effect of ␤2T686F mutation on ␣L␤2 that facilitates ␣L micro-cluster formation.

DISCUSSION
A functional integrin is formed by noncovalent and specific association of an ␣ and a ␤ subunit (1). The ligand binding properties and the cytoplasmic signaling capacities of integrins require conformational changes that are transmitted from their headpieces to their cytoplasmic tails and vice versa (23,24). The integrin TMs serve as the signal transducers for the relay of activation signals between the extracellular and intracellular regions of the molecule. Unlike the well known mechanisms of receptor tyrosine kinases activation that involves receptor dimerization (49), the separation of the integrin TMs is required for integrin bi-directional signaling (6,7,11). It appears that the specific pairing and packing of integrin TMs via parallel evolution is required to maintain integrins in a low affinity conformation (25). Indeed, mutations that disrupt the interface between integrin ␣IIb␤3 TMs have a propensity to generate activated receptors (7,8). Further, an engineered disulfide bond that clasps the integrin TMs and prevents their separation attenuated ␣IIb␤3 outside-in signaling (11).
Here, we showed that the TM residue Thr-686 of the integrin ␤2 subunit participates in the interaction of the ␣L␤2 TMs because mutations of this residue induced ␣L␤2 activation with ligand binding properties of an intermediate affinity receptor. This is based on the capacity of TM disrupted ␣L␤2 to bind constitutively to ICAM-1 and not ICAM-3, which requires an  . The mutation T686F in the ␤2 TM promoted ␣L micro-cluster formation. A, an illustration of FRET in ␣LCFP␤2 and ␣LYFP␤2 micro-cluster. When ␣LCFP␤2 and ␣LYFP␤2 do not form micro-clusters, and the distance between the CFP and YFP is Ͼ10 nm, none or minimal FRET will be detected. The disruption of the TM interface by T686F leads to ␣LCFP␤2 and ␣LYFP␤2 micro-cluster formation (case I) in which ␣LCFP and ␣LYFP are Ͻ10 nm apart, and FRET will be detected. There will also be cases II and III in which no FRET will be detected. additional activation signal as reported previously (41,43). However, these ␣L␤2 mutants did not adopt an extended conformation that was detectable by reporter mAb KIM127 (32). Based on negative stain electron microscopy of ␣L␤2 and ␣X␤2, it is evident that an extended conformation is one of the hallmarks of integrin activation (50). Nonetheless, conformational breathing of a bent ␣L␤2 may also provide certain degree of unbending that will not be poised to react with KIM127, but could allow ligand-engagement. Interestingly, the ligand binding capacity of bent integrin ␣V␤3 has also been described and discussed (24,51). Using the method of FRET-based detection, it was shown that integrin ␣4␤1 on U937 cells did not undergo full extension when activated by Mn 2ϩ or by formyl peptide receptor activation (52). Recently, it was also shown that an engineered disulfide bond linking the plexin-semaphorin-integrin domain and IEGF-2 of the ␤2 subunit generated a mutant ␣L␤2 that was KIM127 negative, but showed a propensity to adhere to ICAM-1, suggesting that full extension need not be an absolute requirement for ␣L␤2 ligand-binding (46). Taken together, it may be inferred that the disrupted packing of the ␣L␤2 TMs by mutations employed in this study induces conformational changes that do not culminate to full extension of the receptor, but these changes are sufficient to activate ␣L␤2.
We have previously generated two models for integrin ␣/␤ TM interface independently from any experimental data (15) that are consistent with both a GXXXG-type (cyan) interaction, suggested from experiments using the GALLEX assay (18), and that based on cysteine-scanning analyses (green) (10) (Fig. 1B). The orientation of the ␤2 TM residues Val-683 and Thr-686 are different in these models. Whereas both residues are located in the interface of the TMs based on the cysteine-scanning ␣IIb␤3 model (green), Val-683 is projected away from the interface in the GXXXG-type model (cyan) (Fig. 5B). Indeed, mutation T686F but not V683F induced ␣L␤2 activation, which corroborates well with a GXXXG-type interaction. This was further exemplified in the FRET-based analyses of ␣L oligomerization. Other than the heterodimeric TM-TM interaction, homomeric TM-TM oligomerization of integrin subunits have been reported using protein fragments containing integrin TMs (16 -18). In a CHO cell-based system, the mutation G708N in the ␤3 TM increased the ligand binding affinity of full-length ␣IIb␤3 and induced formation of macro-clusters (7). Further, in the same study, oligomerization of purified full-length ␣IIb␤3 under activating conditions was detected by transmission electron microscopy (7). However, macro-cluster formation was not detected using the same system by another group   (8). What accounts for the difference in observations is not known. Nevertheless, in both studies, the ability of ␣IIb␤3 to form micro-clusters rather than macro-clusters in a cell-based system was not addressed. Integrin macro-cluster contains an area with a diameter of Ͼ200 nm that is detectable by microscopy, and it can only provide information on integrin localization but not homo-oligomerization (8). However, micro-cluster of integrins is defined as integrins that are within ϳ10 nm of each other, which is a more accurate assessment of integrin homo-oligomerization (47,53).
FRET-based studies are suitable to detect integrin microcluster because effective FRET requires the CFP and YFP that are fused to the integrin subunits to be Ͻ10-nm apart. Using FRET by the method of photobleach, we detected ␣L microcluster formation as a result of direct disruption of the ␣L␤2 TM interface. In line with the up-regulated ligand binding activity of ␣L␤2T686F and not ␣L␤2V683F, ␣L micro-cluster was detected only in ␣L␤2T686F. Thus, we conjectured that direct disruption of the ␣L␤2 TM interface not only induces receptor affinity change but also facilitates ␣L micro-cluster formation. Because the study was conducted with a protocol of using ␣LCFP, ␣LYFP, and ␤2 in the same cell, the detection of FRET could only suggest oligomerization of the ␣L subunits. Whether the ␤2 subunit homo-oligomerizes as a result of the disrupted ␣L␤2 TMs is not known at present, but would be interesting future studies to pursue.
The micro-cluster formation of ␣L in cells expressing ␣L␤2T686F may be attributed to the separation of ␣L␤2 TMs, but it is also possible that ␣L␤2T686F has greater lateral mobility in the membrane as compared with wild-type ␣L␤2, hence a higher propensity for ␣L to micro-cluster in ␣L␤2T686F cells. This requires further investigations. Indeed, an interesting study reports that different conformational states of ␣L␤2 differ in their lateral mobility in T cells (54). Recently, we have reported a high propensity of ␣M and ␤2 synthetic TM homomeric interactions (55). Also homomeric associations of these subunits could be detected in CHO cells by FRET (56). Additional studies addressing TM homomeric interactions of the ␤2 integrins, and the contribution of TM-TM interactions to receptor lateral mobility will provide useful insights into immune cell homeostasis, as in antigen-presentation and tolerance, in which cytoplasmic signaling emanating from clusters of ␤2 integrins modulates the immune response (57)(58)(59).