Epitope Mapping of Monoclonal Antibody to Integrin αL β2 Hybrid Domain Suggests Different Requirements of Affinity States for Intercellular Adhesion Molecules (ICAM)-1 and ICAM-3 Binding*

Integrin undergoes different activation states by changing its quaternary conformation. The integrin β hybrid domain acts as a lever for the transmission of activation signal. The displacement of the hybrid domain can serve to report different integrin activation states. The monoclonal antibody (mAb) MEM148 is a reporter antibody that recognizes Mg/EGTA-activated but not resting integrin αL β2. Herein, we mapped its epitope to the critical residue Pro374 located on the inner face of the β2 hybrid domain. Integrin αL β2 binds to its ligands ICAM-1 and ICAM-3 with different affinities. Integrin is proposed to have at least three affinity states, and the position of the hybrid domain differs in each. We made use of the property of mAb MEM148 to analyze and correlate these affinity states in regard to αL β2/intercellular adhesion molecule (ICAM) binding. Our study showed that Mg/EGTA-activated αLβ2 can adopt a different conformation from that activated by activating mAbs KIM185 or MEM48. Unlike ICAM-1 binding, which required only one activating agent, αL β2/ICAM-3 binding required both Mg/EGTA and an activating mAb. This suggests that αLβ2 with intermediate affinity is sufficient to bind ICAM-1 but not ICAM-3, which requires a high affinity state. Furthermore, we showed that the conformation adopted by αLβ2 in the presence of Mg/EGTA, depicting an intermediate activation state, could be reverted to its resting conformation.

Integrins represent a large family of type I heterodimeric (␣ and ␤ subunits) membrane proteins capable of bidirectional signal transduction serving cell growth, differentiation, and apoptosis (1). The ␤ I-like domain is flanked by the hybrid and PSI 1 (for plexins, semaphorins, and integrins) domains ( Fig. 1A) (2,3). The hybrid domain has been shown to be important in the propagation of the activation signal from one end of the ␤ subunit to the other (2,(4)(5)(6). Recently, superimposed structural coordinates of liganded-open ␣ IIb ␤ 3 headpiece with that of unliganded-closed ␣ V ␤ 3 revealed an ϳ10-Å shift and a concomitant rotation of the hybrid domain relative to the last helix of the ␤ I-like domain upon ligand binding (7). mAbs that attenuate this swing-out motion of the integrin ␣ 5 ␤ 1 hybrid domain prevent effective allosteric activation of the ␤ I-like domain (5,6).
Collective observations from electron microscopy images and crystal structures of integrin ␣ IIb ␤ 3 , ␣ V ␤ 3 , and ␣ 5 ␤ 1 suggest that integrin may undergo at least three activation states depicted by different quaternary conformations (3,(7)(8)(9)(10)(11)(12)(13). A bent integrin with the hybrid domain in close proximity of the integrin epidermal growth factor (I-EGF) 3 and 4 domains represents the resting state. The extended integrin with the hybrid domain distally separated from I-EGF 3 and 4 but orientated toward the ␣ subunit ␤-propeller depicts a low affinity state, whereas the extended integrin with a swing-out hybrid domain away from the ␣ subunit ␤-propeller represents a high affinity state.
The conceptualization of different integrin affinity states also derives from earlier functional studies. Observations were made on integrins having different ligand binding properties under different cellular or extracellular conditions. Resting platelet integrin ␣ IIb ␤ 3 binds immobilized fibrinogen, but binding to soluble fibrinogen, fibronectin, or von Willebrand factor requires platelet activation by agonist (14 -16). Divalent cation Mn 2ϩ activates integrin ␣ 4 ␤ 1 to bind VCAM-1, whereas adhesion to fibronectin-derived CS-1 peptide requires additional activating mAb (17). Real time analysis of integrin ␣ 4 ␤ 1 binding to fluorescent-conjugated ligand mimetic via chemokine receptor activation on leukocytes also suggests integrin acquiring multiple affinity states (18). For the integrin ␣ L ␤ 2 , ligand ICAM-1 exhibits higher affinity for purified ␣ L ␤ 2 from T cells than ICAM-3 (19). Prior exposure of ␣ L ␤ 2 to ICAM-1 increased its binding to ICAM-3 (20). Soluble ICAM (sICAM)-3 binds to ␣ L ␤ 2 with 9-fold lower affinity than sICAM-1 (21). We also reported the requirement of two ␣ L ␤ 2 -activating mAbs, KIM185 and KIM127, for their adhesion to ICAM-3 as compared with ICAM-1 (22). Crystal structures of engineered intermediate affinity ␣ L I-domain (L161C/F299C) and high affinity ␣ L I-domain (K287C/K294C) in complex with ICAM-1 showed differences in their interactions (23). Together, it is apparent that distinct ␣ L ␤ 2 affinity states could serve binding to different ICAM ligands. Structural data of intact ␣ L ␤ 2 interacting with ICAM are lacking. Hence, structural data derived from the studies of ␤ 3 and ␤ 1 integrin are useful as a hypothetical activation model to correlate ␣ L ␤ 2 functional states to its conformation.
In this article we described the reporter mAb MEM148, which recognizes the free integrin ␤ 2 subunit or Mg 2ϩ /EGTAactivated ␣ L ␤ 2 but not resting ␣ L ␤ 2 (2, 24). We mapped the epitope of MEM148 to the hybrid domain, and the location of the epitope when modeled with the bent structure of ␣ V ␤ 3 faces the ␣ subunit and could, therefore, be masked. Activation as a result of integrin extension could lead to the exposure of the epitope. In addition, we make use of the activation-reporter property of MEM148 to analyze the different ␣ L ␤ 2 affinity states required for ICAM-1 and ICAM-3 binding.
cDNA Expression Constructs-The ␣ L and ␤ 2 cDNAs in the expression vector pcDNA3 (Invitrogen) were described previously (26,27). Previously, we had numbered the N-terminal Met of ␤ 2 as "1" (4) and relevant references therein. For the purpose of clarity and ease of reference with other ␤ 2 and different ␤ integrins functional and structural studies, herein ␤ 2 Met 1 is re-numbered as Met Ϫ22 (28). Construction of ␤ 2 Hu/Mo A in which Met Ϫ22 -Asn 562 of human integrin ␤ 2 was replaced with the corresponding region from mouse ␤ 2 was described previously (4). Amino acid substitutions on the mouse or human ␤ 2 constructs were made using the QuikChange™ site-directed mutagenesis kit (Stratagene). All constructs were verified by sequencing (DNA Sequencing Facility, Department of Biochemistry, University of Oxford, Oxford, UK and Research Biolabs sequencing service, Singapore).
Flow Cytometry Analysis-Flow cytometry analysis of integrin cell surface expression was performed as described previously (29). Briefly, cells were incubated with 20 g/ml primary mAb in RPMI 1640 for 1 h at 4°C. Thereafter, cells were washed and incubated with FITC-conjugated sheep anti-mouse F(abЈ) 2 secondary Ab (1:400 dilution; Sigma) for 45 min at 4°C. Stained cells were washed once and fixed in 1% (v/v) formaldehyde in PBS. Cells were analyzed on a FACSCalibur (BD Biosciences). Data were analyzed using CellQuest software (BD Biosciences). The expression index (EI) was calculated by % cells gated positive ϫ geo-mean fluorescence intensity.
Analysis of mAbs MEM148 and KIM127 Epitope Exposure-MOLT-4 cells were incubated in Mg/EGTA (5 mM MgCl 2 and 1.5 mM EGTA) at 37°C for 0.5 h together with MEM148 or KIM127. Cells were washed once followed by incubation with FITC-conjugated secondary antibody for 45 min at 4°C. Treatment with EDTA was performed by subsequent incubation of Mg/EGTA-treated cells in 5 mM EDTA at 37°C for 15 min followed by staining with the respective mAbs. Fluorescence staining was detected by flow cytometry. Washing experiments were performed by subjecting Mg/EGTA-treated cells to two or three washes in RPMI before incubating with antibodies.
Immobilized Ligand Binding Assay-Adhesion of MOLT-4 cells to ICAM-1 or ICAM-3 coated on Polysorb microtiter wells (Nunc, Rosklide, Denmark) were performed as reported previously (29). Briefly, each microtiter well was coated with 0.5 g of goat anti-human IgG (Fc-specific) in 50 mM bicarbonate buffer (pH 9.2). Nonspecific binding sites were blocked with 0.5% (w/v) bovine serum albumin in PBS for 30 min at 37°C. ICAM-Fc at 1 ng/l or other specified concentrations in PBS containing 0.1% (w/v) bovine serum albumin was added to each well and incubated for 2 h at room temperature. Wells were washed twice in RPMI wash buffer (RPMI media containing 5% (v/v) heat-inactivated fetal bovine serum and 10 mM HEPES (pH 7.4)) before assay. Cells labeled with 3.0 mM 2Ј7Ј-bis-(2-carboxyethyl)-5-(and-6) carboxyfluorescein fluorescent dye (Molecular Probes, Eugene, OR) were incubated in RPMI wash buffer containing Mg/EGTA (5 mM MgCl 2 and 1.5 mM EGTA) and/or activating mAb (10 g/ml) to activate ␣ L ␤ 2 -mediated adhesion. The activating mAbs are MEM48 and KIM185. ␣ L -Specific function-blocking mAb MHM24 (10 g/ml) was included to demonstrate binding specificity. Cell fluorescence, which indicates the number of cells adhering to ligand-coated wells, is measured using FL600 fluorescent plate reader (Bio-Tek Instruments, Winooski, VT).
Soluble Ligand Binding Assay-sICAM-1 or sICAM-3 binding assay was performed as described with slight modifications (30,31). Soluble ICAMs were prepared by incubating 20 g/ml ICAM-Fc with 100 g/ml FITC-conjugated rabbit anti-human IgG Fc antibody (Pierce) in 50 l of RPMI 1640 media containing 5% (v/v) heat-inactivated fetal bovine serum and 10 mM HEPES (pH 7.4) for 30 min at room temperature. Thereafter, purified mouse IgG was added to a final concentration of 100 g/ml mAb for 20 min at room temperature. This was performed to quench cross-reactivity to mouse IgG, which was included in the experiment as activating or function-blocking mAb. Cells (2 ϫ 10 5 ) were resuspended in this sICAM mixture with or without ␣ L ␤ 2 -activating agents for 30 min at 37°C. Cells were washed with RPMI media and fixed in 1% (v/v) formaldehyde in PBS followed by flow cytometry analyses. For activating studies, Mg/EGTA (5 mM MgCl 2 and 1.5 mM EGTA) and/or activating mAbs (MEM48 or KIM185) (10 g/ml) were used. For blocking studies, function-blocking ␣ L -specific mAb MHM24 (10 g/ml) was included in the samples.

Mapping of mAb MEM148
Epitope-The epitope of mAb MEM148 resides in the integrin ␤ 2 hybrid domain and is not expressed in resting integrin ␣ L ␤ 2 but is exposed upon Mg/ EGTA treatment (2). Because it may serve as a useful reporter mAb for subsequent analyses of integrin affinity states, we further characterize its epitope using a panel of integrin ␤ 2 human/mouse "knock-out" mutants generated by site-directed mutagenesis (4). These human integrin ␤ 2 mutants have their residues replaced by corresponding mouse ␤ 2 residues at positions where they differ in the mid-region (Fig. 1B). COS-7 cells, which can express ␤ 2 integrin in the absence of the ␣ subunit (2), were transfected with the ␤ 2 human/mouse knock-out mutant cDNAs followed by immunofluorescence staining with mAb MEM148 and flow cytometry analyses. Expression of MEM148 epitope on transfectants was determined ( Fig. 2A). The mAb MEM48, which maps to residues Leu 534 , Phe 536 , and His 543 of human integrin ␤ 2 I-EGF 3 (32), was included as the receptor expression control. An approximately 3-fold reduction in MEM148 epitope expression was observed for transfectants expressing ␤ 2Hu (H370S/R371I) or ␤ 2Hu (N372G/Q373K) as compared with ␤ 2Hu (wt). The ␤ 2Hu (P374S) variant showed significant (Ͼ90%) abrogation of MEM148 epitope expression.
We next employed an integrin ␤ 2 human/mouse chimera (␤ 2Hu/Mo A) (Fig. 1A) in which Met Ϫ22 -Asn 562 of human integrin ␤ 2 is replaced by the corresponding region from mouse integrin ␤ 2 (4) to generate two "knock-in" mutants. The ␤ 2Hu/ MoA(S374P) has the Ser 374 of the mouse mid-region substituted by the corresponding human Pro residue. The ␤ 2Hu/ MoA(A368V-S374P) has the segment Ala 368 -Ser 374 of the mouse mid-region replaced by the corresponding human segment Val 368 -Pro 374 . In this case, mAb KIM185, which maps to integrin ␤ 2 I-EGF 4 and ␤-tail domain (32), was included as the receptor expression control because the epitope of MEM48 is absent in ␤ 2Hu/Mo A. The expression of MEM148 epitope on ␤ 2Hu/Mo A(S374P) was low as compared with ␤ 2Hu (wt) (Fig. 2B). However, transfectant bearing ␤ 2Hu/Mo A(A368V-S374P) fully restored the epitope of MEM148. Thus, we may conclude that although human integrin ␤ 2 Pro 374 is a critical residue, other residues, His 370 , Arg 371 , Asn 372 , Gln 373 , are also required for the effective presentation of MEM148 epitope.
Model of ␣ L ␤ 2 Illustrating the MEM148 Epitope-Fine mapping of MEM148 epitope allows us to pinpoint its location in the quaternary structure of integrin. A model of integrin ␣ L ␤ 2 was generated by MODELLER using ␣ V ␤ 3 structural coordinates as the template (3) (Fig. 3A). The structure of an intact I-domain-containing integrin is not solved; hence, the ␣ L I-domain was excluded from the model. For clarity, the ␣ L Calf-2, ␤ 2 PSI, I-EGFs, and ␤-tail domains were not included. The critical residue Pro 374 resides on the surface of the hybrid domain facing the ␣ L subunit, in contrast to the epitope of mAb 7E4 (Val 385 ) (4), which is located away from the ␣ L subunit (Fig.  3B). This could explain why mAb MEM148 fails to bind resting ␣ L ␤ 2 ; presumably it assumes a severely bent conformation similar to the resting ␣ V ␤ 3 (3) because its epitope is shielded in this conformation (Fig. 3C). Along the same line of reasoning, the binding of mAb MEM148 to activated ␣ L ␤ 2 adopting an extended conformation would, therefore, be favorable in conjunction with previous observations (2,24).
The Transition of ␣ L ␤ 2 from One Affinity State to Another Is Reversible-The mAb MEM148 does not bind to ␣ L ␤ 2 on MOLT-4 cells unless the cells were treated with Mg/EGTA (2). Divalent cations have a major influence on the ␣ I-domain and ␤ I-like domain, and it is widely accepted that the activated integrin should adopt an opened and/or extended conformation. However, defining the precise mechanism for the transition of integrin from one affinity state to the next in the presence of activating divalent cations remains challenging. Recently, several quaternary integrin conformations were proposed to depict such transitions (7). Of note, the reversion from one conformation to another may be physiologically relevant to maintain a dynamic integrin population responding to different cellular activation milieu.
To this end we determine whether Mg/EGTA-treated ␣ L ␤ 2 can revert to its resting state using MEM148 as the reporter mAb. MOLT-4 cells were incubated in Mg/EGTA-containing RPMI and either MEM148 or KIM127 at 37°C. The epitope of KIM127 resides in ␤ 2 I-EGF 2, and like MEM148, its epitope expression depends on integrin activation (32,33). Significant staining was detected for both mAbs only in the presence of Mg/EGTA (Fig. 4A). The integrin ␣ L -specific mAb MHM24 was included as a control. Next, cells in the presence of Mg/EGTA were subsequently treated with EDTA to deplete existing Mg 2ϩ before incubation with respective mAbs. Epitope expressions of MEM148 and KIM127 were significantly reduced in these samples, whereas control mAb MHM24 staining was not affected.
We next tested whether Mg/EGTA-treated cells after washing in media could still express the epitopes of MEM148 and KIM127 (Fig. 4B). Mg/EGTA-treated cells were subjected to three washes in RPMI without additives followed by staining for either of the two mAbs. Low levels of MEM148 or KIM127 epitope expression were detected as compared with Mg/EGTAtreated cells without washing. Cell binding assay to ICAM-1 was investigated to determine whether Mg/EGTA-treated cells followed by washing could still adhere to ICAM-1 (Fig. 4C). In  (7) is underlined with the Cys 425 involved in the long range disulfide bond indicated by an asterisk (*).

FIG. 2.
Epitope expression of mAb MEM148 on ␤ 2 knock-out and knock-in mutants. A, ␤ 2 knock-out mutants were generated and stained for mAb MEM148 followed by flow cytometry analyses. The epitope expression of MEM148 is calculated by MEM148 EI/MEM48 EI for each mutant. The mAb MEM48 maps to the C-terminal region of human ␤ 2 subunit and was used as a reference antibody in this case. B, ␤ 2 knock-in mutants were analyzed for the restoration of MEM148 epitope expression. The epitope expression of MEM148 is calculated by MEM148 EI/KIM185 EI. The mAb KIM185 maps to ␤ 2 I-EGF 4 and ␤-tail domain (32) and was included as a reference mAb because the epitope of MEM48 is absent in ␤ 2Hu/Mo A. EI was calculated by % percentage cells gated positive ϫ geo-mean fluorescence. the presence of Mg/EGTA, cells adhered significantly to ICAM-1. However, adhesion was minimal when cells were treated with Mg/EGTA followed by two or three washes in RPMI wash buffer before allowing adherence to ICAM-1. Binding specificity was shown by including ␣ L -specific function-blocking mAb MHM24. Together, our data suggest that reversion of Mg/EGTA-activated ␣ L ␤ 2 to its resting conformation is possible when the activating cation is depleted.
We furthered our analyses by changing the order of MOLT-4 treatment with different agents to test whether MEM148 re- /EGTA were subjected to three washes in RPMI media followed by staining with mAb KIM127 or MEM148. Other conditions are the same as before. C, MOLT-4 cells were allowed to adhere to ICAM-1 in the presence of activating agents Mg/EGTA (ME) as described under "Experimental Procedures." In separate samples, Mg/EGTA-treated cells were subjected to two or three washes in RPMI wash buffer before dispensing into ICAM-1-coated wells. ␣ L -specific function-blocking mAb MHM24 was included to demonstrate binding specificity. D, the experiment was performed as in A and B except that cells were first treated with Mg/EGTA in the presence of MEM148 and washed twice in media followed by EDTA treatment or further washes before secondary antibody staining and flow cytometry analyses. E, MOLT-4 cells were surface-biotinylated. Labeled cells were stimulated with Mg/EGTA or PMA (100 ng/ml) in the presence of irrelevant mAb, MHM23, or MEM148. Cells were washed and lysed in lysis buffer followed by immunoprecipitation and ECL detection ("Experimental Procedures"). mains bound to Mg/EGTA-treated cells even in the presence of EDTA. Cells were first treated with Mg/EGTA in the presence of MEM148 with subsequent treatment with EDTA or additional washes in media (Fig. 4D). MEM148 staining was detected in both cases. This suggests that the ␣ L ␤ 2 ⅐MEM148 complex is stable even when divalent cations were depleted or by extensive washing in media.
PMA is known to promote ␣ L ␤ 2 ligand binding (4,34). PMA also induced the expression of MEM148 epitope on myeloid cells, and this was contributed by a proteolytically cleaved fragment of ␤ 2 unassociated with the ␣ L , ␣ M , or ␣ X subunit rather than the respective heterodimers (35). To test whether PMA induced MEM148 epitope expression on MOLT-4, cells were surface-labeled with biotin followed by treatment with Mg/EGTA or PMA in the presence of MEM148 or other mAbs (Fig. 4E). Cells were lysed, and immunoprecipitation was performed followed by ECL detection. As compared with flow cytometry analyses, cell surface-labeling followed by immunoprecipitation can identify whether there are free forms of ␤2, which could be detected by MEM148, on the cell surface of MOLT-4 as a result of Mg/EGTA or PMA treatment. MEM148 only immunoprecipitated ␣ L ␤ 2 when cells were treated with Mg/EGTA but not with PMA. No corresponding protein bands were detected using irrelevant mAb, but ␣ L ␤ 2 bands were detected under all conditions using the ␤2-specific mAb MHM23. Truncated ␤ 2 (65-70 kDa) (35) was not detected in MEM148 immunoprecipitation samples of PMA-treated MOLT-4 cells. It is possible that the truncated ␤ 2 is mainly expressed and is abundantly up-regulated on the surface of activated myeloid cells instead of lymphocytes (35).
It is also interesting to note that PMA did not induce detectable expression of MEM148 epitope on ␣ L ␤ 2 , although it is reported to promote ␣ L ␤ 2 ligand binding (4,34). It is possible that PMA promotes an active ␣ L ␤ 2 conformation that is different from that triggered by Mg/EGTA for example. Alternatively, PMA increases the lateral mobility of ␣ L ␤ 2 on plasma membrane (36), allowing receptor clustering, which effects ligand binding.
Binding of ␣ L ␤ 2 to Immobilized ICAM-1 and -3 Requires Different Affinity States-To understand the molecular basis of ␣ L ␤ 2 activation with respect to ICAM-1 and ICAM-3 binding, we performed immobilized ligand binding assays using MOLT-4 cells. Cells were allowed to adhere to ICAMs in the presence of Mg/EGTA, mAb, or a combination of both (Fig. 5). MOLT-4 binding to ICAM-1 was minimal in the absence of activating agents. In the presence of Mg/EGTA or any of the activating mAbs KIM185 or MEM48, binding to ICAM-1 was significantly augmented (Fig. 5A). In contrast, in the presence of MEM148 alone, binding to ICAM-1 was minimal. Under the conditions of Mg/EGTA and any of the three mAbs under study, the binding of MOLT-4 to ICAM-1 was further enhanced. ICAM-1 binding was mediated by ␣ L ␤ 2 on MOLT-4 cells because binding was effectively abrogated in the presence of function-blocking ␣ L -specific mAb MHM24. It is, therefore, possible that activation of ␣ L ␤ 2 by either Mg/EGTA or one of the activating mAbs may convert ␣ L ␤ 2 into an intermediate affinity state. Promotion of this population of ␣ L ␤ 2 into a higher affinity state can still be achieved by an additional activating agent. Similar profiles were observed in ␣ L ␤ 2 COS-7 and 293T transfectants (data not shown).
ICAM-3 binding was carried out under similar conditions (Fig. 5B). Binding to ICAM-3 was minimal in the presence of Mg/EGTA or any of the mAbs. When Mg/EGTA and one activating mAb were included, binding was significantly augmented. This was in concordance with our previous findings that mAbs KIM185 and KIM127 were required for ␣ L ␤ 2 COS-7 transfectants binding to ICAM-3 (22). The addition of two different mAbs recognizing the same receptor may result in receptor clustering. In this study enhanced adhesion to ICAM-3 cannot be due to receptor clustering because only one activating mAb was employed, whereas the other activating agent was Mg/EGTA. Abrogation of binding by MHM24 demonstrated interaction specificity between ␣ L ␤ 2 and ICAM-3. Noteworthy, the combination of Mg/EGTA with mAb MEM148 only had a marginal effect on ␣ L ␤ 2 /ICAM-3 binding as compared with other mAbs. This could not be due to insufficient ICAM-3 ligand coated on the well (1 ng/l) because increasing ICAM-3 concentration 3-fold had only a marginal effect on cell binding under all conditions (Fig. 5C).
Binding of MEM148 to ␤ 2 hybrid domain is permissible on Mg/EGTA-activated but not resting ␣ L ␤ 2 . The ICAM-3 binding data suggest that Mg/EGTA-activated ␣ L ␤ 2 adopts a conformation in which the epitope of MEM148 becomes exposed. The quaternary conformation adopted by Mg/EGTA-activated ␣ L ␤ 2 may be similar to one of the intermediate affinity states proposed for ␤ 3 integrins.
Binding of ␣ L ␤ 2 to Soluble ICAM-1 and -3-If resting ␣ L ␤ 2 assumes a severely bent conformation like ␣ V ␤ 3 , the accessibility of ligands to the ␣ I-domain and ␤ I-like domain on the cell surface is unfavorable because they are oriented toward the plasma membrane. We further our analyses using sICAMs binding assays (Fig. 6) (30,31). This assay allows ICAMs to be "free" in solution rather than being immobilized on a solid phase, which may provide better accessibility to the head of the bent integrin on the cell surface. MOLT-4 cells bind to sICAM-1 when treated with Mg/EGTA, KIM185, or MEM48 but not MEM148. Binding was increased in the presence of Mg/EGTA with KIM185 or MEM48. In the presence of Mg/EGTA and MEM148 binding was augmented as compared with Mg/EGTA alone, which is similar to that observed under immobilized ICAM-1 binding assay (Fig. 5A).
sICAM-3 binding to MOLT-4 cells was detected only when cells were treated with Mg/EGTA together with KIM185 or MEM48. Similar to immobilized ICAM-3 binding assay, Mg/ EGTA with MEM148 could promote MOLT-4/sICAM-3 binding albeit at a lower level as compared with Mg/EGTA with KIM185 or MEM48. Bindings were ␣ L ␤ 2 -specific because they were effectively abrogated by MHM24. It was noted that there was a difference between immobilized and soluble ICAM assays. The percentage cell binding of Mg/EGTA-and MEM48treated MOLT-4 on ICAM-1, for example, was similar with that on ICAM-3 (Fig. 5, A and B). However, binding to sICAM-3 was much lower than sICAM-1 with Mg/EGTA and MEM48 (Fig. 6,  A and B). Such a difference was consistently detected in repeat experiments performed. Woska et al. (21) also reported weaker sICAM-3 binding to purified ␣ L ␤ 2 as compared with sICAM-1. At present the reason for these observations is unclear. DISCUSSION The headpiece of the integrin ␣ L ␤ 2 may be analogous to the integrin ␣ IIb ␤ 3 consisting of the ␣ I-domain (in I-domain-containing integrins), ␤-propeller, thigh domain, ␤ I-like domain, hybrid, PSI, and I-EGF 1 (7). The ␤ 2 I-like domain of integrin ␣ L ␤ 2 allosterically regulates ligand binding of the ␣ L I-domain through binding of the ␤ 2 metal ion-dependent adhesion site (MIDAS) to Glu 310 , found in the last helix of the ␣ L I-domain; this was shown by replacing ␣ L Glu 310 with Cys and either ␤ 2 MIDAS Ala 210 or Tyr 115 to Cys, allowing a disulfide connection between the two domains (31). Downward displacement of this helix by ␤ 2 I-like domain with a "pull spring" motion triggers I-domain conversion from a closed to an open conformation ready for ligand binding (31). The hybrid domain is linked directly to the ␤ I-like domain (3). Our previous study on ␣ L ␤ 2 hybrid domain maps the epitope of mAb 7E4 to Val 385 and demonstrates the importance of the hybrid domain for the activation signal transfer from ␤ 2 membrane proximal region to the ␣ L I-domain (4). Preventing the movement of integrin ␣ 5 ␤ 1 hybrid domain away from its ␤ 1 I-like domain by mAb SG/19 maintains ␣ 5 ␤ 1 in low affinity for fibronectin, as determined by surface plasmon resonance (6).
Herein, we report the binding determinant of mAb MEM148, consisting of the critical residue Pro 374 and residues His 370 -Gln 373 of the integrin ␤ 2 hybrid domain. Distinct from the Only MHM24 blocking of ME/MEM48 binding at highest concentration of ICAM-3 was included for clarity. exposed epitope Val 385 of mAb 7E4, these residues are hidden in the ␣ L ␤ 2 model using the bent ␣ V ␤ 3 as template (Fig. 3). This shares similarity to mAbs 15/7 and HUTS-4 epitopes located on integrin ␤ 1 hybrid domain that are partly masked by integrin ␣ 5 ␤-propeller (5). Accessibility of mAb MEM148 to its epitope(s) would, therefore, be sterically unfavorable when ␣ L ␤ 2 adopts a bent and closed headpiece conformation. It accounts for the lack of mAb MEM148 reactivity to unactivated ␣ L ␤ 2 observed previously (2,24). However, when exposed to Mg/EGTA, ␣ L ␤ 2 can undergo a conformational change possibly via unbending, thereby exposing the epitope of MEM148. Crystal structures of the integrins ␣ IIb ␤ 3 and ␣ V ␤ 3 with ligand mimetics, electron microscopy images of soluble integrin ␣ V ␤ 3 with cyclo-RGDfV peptide in the presence of Mn 2ϩ or Ca 2ϩ , and that of integrin ␣ 5 ␤ 1 with physiological ligand fibronectin provide insights into the possible conformations that can be adopted by the integrin during its transition from resting to a high affinity state (3, 7, 10 -12). Based on these observations, it was proposed that integrin activation involves at least three affinity states (7). Each conformer can be distinguished not only by the degree of bending but also the conformation of its headpiece as determined by the relative orientation of the hybrid with respect to the ␤ I-like domain. Previously, it was reported that ␣ L ␤ 2 binding to ICAM-1 and -3 exhibit different affinities (19 -22). In this study we found that mAb MEM148 in combination with activating agents Mg/ EGTA further enhanced the affinity of ␣ L ␤ 2 /ICAM-1 binding as compared with Mg/EGTA alone. The observation could be attributed to the disruption of the extensive contacts between the ␣ L ␤ 2 headpiece-tailpiece interface when in the presence of Mg 2ϩ and with Ca 2ϩ depleted. This would facilitate the unbending of ␣ L ␤ 2 , hence allowing MEM148 to access its epitope masked previously.
From our ␣ L ␤ 2 /ICAM-3 binding data, two activating agents are required to promote binding. Collectively, this implies that the conformation adopted by ␣ L ␤ 2 in the presence of Mg/EGTA, mAb KIM185, or MEM48 represents an affinity state(s) capable of ICAM-1 but not ICAM-3 binding. The ␣ L ␤ 2 conformer in the presence of Mg/EGTA could be different from those activated by mAb KIM185 or MEM48 because if they are identical the combination of both agents would not further enhance ␣ L ␤ 2 /ICAM-1 and promote ␣ L ␤ 2 /ICAM-3 binding as compared with using one agent. Indeed, our previous study showed that the mAb 7E4 binds to ␣ L ␤ 2 hybrid domain and prevents ␣ L ␤ 2 activation by mAb KIM185 but not Mg/EGTA (4). The Mg/ EGTA-activated ␣ L ␤ 2 conformer has its hybrid domain re-oriented, as evidenced by its reactivity with MEM148 described herein. However, such re-positioning as a result of Mg/EGTA activation still did not favor ␣ L ␤ 2 binding of ICAM-3. We also found that the combination of Mg/EGTA with mAb MEM148 did not promote effective ␣ L ␤ 2 /ICAM-3 binding as opposed to Mg/EGTA with mAb KIM185 or MEM48. It differs from that of ICAM-1 binding in which Mg/EGTA and MEM148 had a cumulative effect. If mAb MEM148 further displaces the hybrid domain in Mg/EGTA-activated ␣ L ␤ 2 , resulting in a higher affinity state comparable with Mg/EGTA/KIM185 or MEM48 condition, effective ICAM-3 binding should also be established. However, the lack of binding detected in this case would suggest that mAb MEM148 may only stabilize the Mg/EGTAactivated conformation of ␣ L ␤ 2 by binding to its re-positioned hybrid domain.
It was hypothesized that other intermediates may exist during ␣ V ␤ 3 activation (10). These intermediates may be depicted by half-bent structures with separation of their ␣ and ␤ subunits but still connected at their heads. Solution structure of intact integrin in Mg 2ϩ only is lacking; nonetheless, it is possible that Mg/EGTA-activated ␣ L ␤ 2 adopts a conformation akin to one of these intermediates. This may explain the reactivity of mAb MEM148 to Mg/EGTA-activated ␣ L ␤ 2 . The Mg/EGTAactivated ␣ L ␤ 2 "intermediate" is also a dynamic structure that can be reverted to its resting bent conformation. We found that the reactivity of mAbs MEM148 and KIM127 was diminished in EDTA-treated ␣ L ␤ 2 , which was beforehand activated by Mg/EGTA. EDTA does not directly affect these mAbs because staining could be detected for single integrin ␤ 2 expressed on COS-7 cells in the presence of 5 mM EDTA. 2  MEM148 (2) and KIM127 (32) report ␤ 2 hybrid domain and I-EGF 2 movement during ␣ L ␤ 2 activation, respectively, our data favor a global conformation reversion to its resting state rather than a localized effect.
Structural data revealed an intermediate affinity ␣ L I-domain (Leu 161 -Cys/Phe 299 -Cys lock) capable of ICAM-1 binding in the presence of Mg 2ϩ (23). The transition between low, intermediate, and high affinity conformations of ␣ L I-domain revealed a ratchet-like movement of its ␤6-␣7 loop (23,37). Recently, molecular dynamics simulation reveals that intermediate affinity ␣ L I-domain can be generated by applying a pulling-force on the ␣7 helix (38), which is allosterically regulated by the ␤ 2 I-like domain. We found that Mg/EGTA-treated ␣ L ␤ 2 could be further activated to bind ICAM-1 by the addition of activating mAb KIM185 or MEM48. Mg/EGTA treatment would have converted ␣ L ␤ 2 into an intermediate affinity state. The addition of these mAbs would disrupt the interface between the ␣ L and ␤ 2 leg pieces with eventual propagation of conformational change to the hybrid domain and ␤ 2 I-like domain. This allows further displacement of the I-domain ␣7 helix transforming the ␣ L ␤ 2 from an intermediate to a high affinity state for ICAM-1 binding.
Why does binding to ICAM-1 and ICAM-3 require different ␣ L ␤ 2 conformations? The difference may be affected by dissimilar dissociation rates of the interactions and the conformations of the binding pockets. With respect to the ICAMs, the height of ICAM-1 and -3 on the cell surface is similar, both having five C2-set IgSF domains. The binding sites in the ICAMs for ␣ L ␤ 2 also shares similarity; Glu 34 and Gln 73 in IgSF domain 1 (D1) of ICAM-1 (39) and the corresponding residues Glu 37 and Gln 75 in D1 of ICAM-3 (40 -42). However, Lys 39 , Met 64 , Tyr 66 , and Asn 68 of ICAM-1 are important for binding to ␣ L I-domain (23), but Asn 23 , Ser 25 , and Phe 54 of ICAM-3 are required for effective interaction (40). In addition, ICAM-3 IgSF domain 1 is extensively glycosylated as compared with ICAM-1 IgSF domain 1 (19). These may contribute to the dissimilarity between ␣ L ␤ 2 / ICAM-1 and -3 binding (43). The binding sites of ICAM-1 and -3 on the ␣ L I-domain have also been determined. Residues surrounding the ␣ L I-domain metal ion-dependent adhesion site and those in proximity are found to be critical for ICAM-1 and -3 binding (23,(43)(44)(45). In another study, the Ile-Lys-Gly-Asn motif, located in the N terminus of the ␣ L I-domain, was shown to be important for ICAM-3 but not ICAM-1 binding (46). This is intriguing because the metal ion-dependent adhesion site is positioned at the top of the ␣ L I-domain, whereas its N terminus is located at its bottom. Recent crystal data of an engineered high affinity ␣ L I-domain in complex with ICAM-3 D1 suggest that both ICAM-1 and ICAM-3 share a common docking mode with ␣ L I-domain (43). The exact nature and differences of the interactions between intact ICAMs and ␣ L ␤ 2 require further investigations.
In conclusion, our study using ␤ 2 hybrid domain-specific reporter mAb MEM148 suggests the requirement of different affinity states for integrin ␣ L ␤ 2 binding to ICAM-1 and -3. A proposed model explaining these observations would be the displacement of the ␤ 2 hybrid domain in Mg/EGTA-treated ␣ L ␤ 2 , which renders ␣ L ␤ 2 capable of ICAM-1 binding. However, this movement did not generate an ␣ L ␤ 2 conformer that could bind to ICAM-3 effectively, which requires additional activation signal. It will be interesting in future studies to test the proposed model of ␣ L ␤ 2 activation by structural analyses of intact ␣ L ␤ 2 with ICAMs under different conditions.