A Mechanism for Antibody-mediated Outside-in Activation of LFA-1*

MEM83 is an inserted domain (I-domain)-specific antibody that up-regulates the interaction of LFA-1 with ICAM-1 through an outside-in activation mechanism. We demonstrate here that there is no change in the affinity of the MEM83 antibody for the I-domain in either its low (wild-type) or high affinity form and that MEM83 does not enhance the binding of the wild-type I-domain to ICAM-1. Furthermore, we show that the antibody acts as an activating agent to induce LFA-1/ICAM-1-dependent homotypic cell aggregation only as an IgG, but not as a Fab fragment. On the basis of these data, we propose an avidity-based mechanism that requires no direct activation of the LFA-1 I-domain by the binding of the antibody; rather, activation is enhanced when there is an interaction with both arms of the IgG. A molecular model of the antibody interaction with LFA-1 illustrates the symmetry and accessibility of the two MEM83 epitopes across the LFA-1/ICAM-1 heterotetramer. We hypothesize that MEM83 stabilizes adjacent LFA-1 molecules in their active form by the free energy that is gained from the binding of the I-domains to each arm of the IgG. This leads to stabilization of the open state of the integrin and outside-in signaling. Our model supports a mechanism in which both affinity and avidity regulation are required in the activation of LFA-1.

immune responses and may prevent inflammatory and autoimmune disease (8 -10). Similarly, up-regulation of LFA-1 can be used to enhance immune responses (11)(12)(13) and may have roles in restoring T cell function in human immunodeficiency virusimmunocompromised individuals (14) or in overcoming tumor-specific toleration (15).
Activation of the integrin LFA-1 results in a rapid and reversible increase in affinity for its ligands, ICAM-1-5 (16 -19). ICAM-1-3 regulate T cell activation (13) and leukocyte homing (20,21); ICAM-4 is a red blood cell-specific ligand that can interact with a several ␤2/CD18 integrins (22); and ICAM-5 plays a role in leukocyte recruitment to neurons in the central nervous system (19). In T cells, LFA-1 and ICAM-1 act as costimulatory molecules to the peptide/major histocompatibility complex (23,24). The role of LFA-1 and ICAM-1 interaction is critical to the formation of the immunological synapse, and aberrant activation or expression of LFA-1 may lead to selectively disabled T cell receptor clustering (25) or the development of autoimmune diseases (26,27).
Recognition and binding of its ICAM-1 ligand by LFA-1 are mediated through the inserted domain (I-domain) in its ␣-subunit; this forms a heterotetrameric complex through D1/D1domain dimerization within ICAM-1 (28). Activation of this integrin entails both a quaternary rearrangement of the ␣L␤2 heterodimer and a tertiary structural rearrangement within the I-domain (29). As a result, the affinity of the isolated I-domains for ICAMs can be modulated by locking the I-domain in its high affinity open conformation by the introduction of a disulfide bridge (28,30). Yeast display selection also identified two activating I-domain mutations (F265S/F292G) that enhance ICAM affinity by 200,000-fold (31).
Integrins switch from the inactive to active form by either inside-out or outside-in activation (32,33). Inside-out activation of LFA-1 is induced by chemokine activation of G-proteincoupled receptors (34,35) and can be mediated by the use of phorbol esters (phorbol 12-myristate 13-acetate (PMA)) (36). Outside-in activation of ␤2 and other integrins is a method for ligand-induced "cross-talking" to modulate signaling pathways within a cell (37,38). Outside-in activation of LFA-1 that leads to high affinity ICAM binding can be directly triggered by the addition of manganese (39). There are also several monoclonal antibodies known to outside-in activate LFA-1 and specifically bind to the ␣L-subunit (MEM83 (40 -42) and NKI-L16 (43)) or the ␤2-subunit (KIM185 and MEM48 (44 -46)). The activating antibody NKI-L16 was mapped to a region within the ␣L-subunit near the transmembrane segment of the molecule (43).
The monoclonal antibody MEM83 binds the ␣L-subunit of LFA-1 (40) and was later identified as an activating antibody for the LFA-1/ICAM-1 interaction (40 -42, 47-49). Interestingly, this antibody was also found to block LFA-1 binding to ICAM-3-4 (47,50). The epitope of MEM83 maps to two regions within the I-domain (Asp 153 -Phe 183 and Thr 217 -Ile 248 ), with Asp 182 being required in the first region and Glu 218 being required combinatorially in the second (51). Here, we investigate potential mechanisms for MEM83 interaction with LFA-1.

EXPERIMENTAL PROCEDURES
LFA-1 I-domain Constructs-Wild-type (WT) and mutant (high affinity (HA) and intermediate (IA) affinity) I-domains (Gly 123 -Tyr 304 ) were expressed in Escherichia coli BL21(DE3) cells using supplemented M9 minimal medium. The recombinant proteins were subsequently refolded and purified as described (52,67). Proteins were concentrated using Centriprep C10 (Millipore Corp.), and sample concentrations were determined using BCA (Pierce) and at A 280 nm . Each LFA-1 I-domain was also 15 N-labeled to allow acquisition of 1 H/ 15 N heteronuclear single quantum coherence nuclear magnetic resonance spectra for quality assurance prior to surface plasmon resonance (SPR) analysis (data not shown).
MEM83 Binding Assays-MEM83 was provided as a gift by Dr. Václav Hořejší (Institute of Molecular Genetics AS CR, Prague, Czech Republic) and was purified by size exclusion chromatography (Superdex TM 75, GE Healthcare). An aliquot of MEM83 was digested by papain, and the Fab fragments were purified on a protein A column as described (41). MEM83 fractions (either IgG or Fab) were concentrated, and sample concentrations were determined by both Bradford (Bio-Rad) and BCA assays. MEM83 IgG was coupled to a CM5 sensor chip (Biacore/GE Healthcare) by amine coupling following the manufacturer's protocols to a response unit of 4630. LFA-1 I-domains (WT, IA, and HA) at concentrations of 20, 50, and 100 nM were subsequently injected over the chip at 10 l/min with Running Buffer A (10 mM HEPES (pH 7.4), 150 mM NaCl, 0.5 mM MgSO 4 , and 0.005% (v/v) Tween 20). For the Fab binding analysis, the HA I-domain was bound to a CM5 chip by amine coupling to a response unit of 743. Buffers and binding conditions were as described for the MEM83 binding experiments.
ICAM-1/Fc Binding Assays-ICAM-1/Fc (R&D Systems) was coupled to a CM5 sensor chip by amine coupling. LFA-1 I-domains (WT, IA, and HA) at concentrations of 100, 200, and 400 nM and LFA-1 I-domains complexed with excess (10-fold) MEM83 were injected over the chip at 10 l/min with Running Buffer A. Negative binding controls with matching MEM83 concentrations were run as described above except that 3 mM EDTA was substituted for MgSO 4 (Running Buffer B).
Biacore Binding Analysis-All SPR (Biacore) sensorgrams were analyzed from triplicate runs using a 1:1 Langmuir model by Scrubber (Version 2.0, Center for Biomolecular Interaction Analysis, University of Utah).
Homotypic Aggregation Assay-The aggregation experiment was performed in flat-bottom 96-well plates at 100 l cells/well as described (35). Human JY cells were washed twice and diluted in RPMI 1640 medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin (1000 units/ml

TABLE 1 SPR binding parameters of MEM83 for the LFA-1 I-domains
Values are the means Ϯ S.D. from three independent measurements at three different concentrations: 20, 50, and 100 nM.

RESULTS AND DISCUSSION
Binding of LFA-1 I-domains by MEM83 as Measured by SPR-A potential mode of MEM83 activation may be due to an enhanced affinity of the antibody for the I-domain in its open HA form relative to its closed low affinity (WT) form. Hence, activation is mediated by direct tertiary interactions within the I-domain. To test this, we expressed recombinant LFA-1 I-domains (Gly 123 -Tyr 307 ) in E. coli and purified and refolded these protein samples from inclusion bodies. We used SPR (Biacore) to measure the affinity of MEM83 IgG and Fab for the WT, disulfide-locked IA (L161C/F299C), and HA (K287C/ K294C) forms (52). No binding preference for the open form was observed (Fig. 1, A-C, and Table 1); thus, activation cannot be mediated by an induced tertiary switch directly upon antibody binding.
To determine whether ICAM-1 and the MEM83 antibody are both required to enhance LFA-1 I-domain activation (47), we bound ICAM-1/Fc to the Biacore CM5 chip surface and measured the affinity of ICAM-1 for the HA and WT LFA-1 I-domains in the presence and absence of the antibody (Fig. 2). Although there was an increased response unit upon premixing the antibody with the HA I-domain, no change in affinity was observed ( Table 2). The presence of a divalent cation was required, as the addition of EDTA to the running buffer prevented the interaction with ICAM-1 (data not shown). Finally, no binding was observed for the WT I-domain in either the presence or absence of the antibody (Fig. 2).
Our data indicate that MEM83 neither directly enhanced nor blocked the binding of the LFA-1 I-domain to ICAM-1. The HA I-domain in its open conformation enhances stable adhesion with ICAM-1 (53, 54) and was not blocked by the addition of MEM83 (Table 2). This is consistent with the mapped MEM83 epitope, which indicates that the antibody binds to a region within the I-domain that does not significantly change its local structure upon activation (Fig. 1D) (51). Published work on glycosylphosphatidylinositol-anchored LFA-1 I-domains showed increased cell rolling upon binding of MEM83 (47). Based on our data, this increase in rolling efficiency in the presence of MEM83 may be due to an increased affinity that is below the threshold of detection for SPR. Alternatively, MEM83 binding may preferentially orient the glycosylphosphatidylinositol-anchored I-domain for ICAM-1 binding or be the result of shear flow conditions (55).
Avidity and Steric Wedge Activation Mechanisms-Rather than a directly induced change in the tertiary structure of the I-domain, our data suggest that the activation mechanism for MEM83 results from induced changes in the overall quaternary structure of LFA-1. We propose two models, avidity and steric wedge, for MEM83-mediated activation of LFA-1 that are based on the accessibility of the antibody to bind the I-domain in the full-length protein.
An avidity model is based on the binding of the LFA-1 I-domains to both arms of the IgG. In this model, MEM83 can bind only one I-domain in the bent/closed conformation because of the close proximity of the I-domain to the membrane, whereas the IgG is able to bind and cross-link I-domains from two separate LFA-1/ICAM-1 heterotetramers (28) in its extended/ open conformation. Thus, an avidity model of outside-in activation of LFA-1 links both arms of an IgG molecule to two LFA-1/ICAM-1 heterotetramers, and vice versa, two I-domains within a single heterotetramer to two separate IgG molecules, to generate a one-dimensional array.
An alternative mode of activation could be that MEM83 acts as a steric wedge between the LFA-1 I-domain and the adjacent ␤-propeller domain or that the accessibility for the MEM83 epitope is blocked by the membrane when the integrin is in its The primary difference between these models is that the avidity model depends on the presence of two binding sites on the intact IgG portion of MEM83. Thus, to explore whether both binding sites are required, we used homotypic cell aggre-   gation assays (41,57) to compare the MEM83 IgG and Fab forms. We obtained identical results in three independent homotypic cell aggregation dose dependence experiments with MEM83 IgG, where the aggregation score dropped to 3 at 0.4 nM, and no aggregation above background was observed below 0.04 nM. Cell aggregation was evident only in the MEM83 IgG samples, with an aggregation profile similar to that of the Mn 2ϩ -and PMA-positive controls (Fig. 3). However, the Fab fragment of the antibody did not induce cell aggregation at the same concentrations. These data indicate that LFA-1 I-domain binding by both arms of MEM83 IgG is necessary for activation and suggest that only one arm is accessible to bind when the I-domain is in its closed conformation.
Our cell assay data are in contrast to those from a previous study in which both MEM83 IgG and Fab resulted in homotypic aggregation (41). A possible explanation for these differences is that the closed form of LFA-1, with its I-domain close to the cell membrane (56), may become sensitive to the accessibility of MEM83 Fab under variable glycosylation patterns of the integrin in the different cell lines and culture media that were used. We note, however, that the literature for MEM83 abounds with seemingly conflicting reports of its role as both an activating and a blocking antibody as well as its being activation-insensitive (40 -42, 47-50, 58).
A Molecular Model of MEM83-mediated LFA-1 Activation-On the basis of our data, we generated a molecular model of the MEM83 interaction with LFA-1 and ICAM-1 (Fig. 4). Our model makes use of the LFA-1 IA I-domain/ICAM-1 D1/D2domain heterotetrameric crystal structure (28) and a template human IgG structure (59). The hypervariable region of this template IgG was oriented adjacent to MEM83 epitope residues Asp 182 and Glu 218 within the first I-domain, creating one repeating unit of our avidity model. In this orientation, positioning of Repeating Unit 2 required simply a translation along the x axis (Fig. 4A). We used domain models of ICAM-1 and LFA-1 in its open conformation to illustrate the overall arrangement of the complex across two cell membranes (Fig.  4B). The relative orientation of the I-domain was dictated by the requirement to place the C-terminal ␣7-helix Glu 310 residue proximal to the ␤2 I-domain (60,61).
Our model demonstrates how cross-linking through both the IgG and two accessible MEM83 epitopes across the heterotetramer would lead to the formation of a one-dimensional array. This accessibility of the MEM83 epitope is supported by modeling and electron microscopy studies indicating that the ␣ I-domain within the integrin headpiece is largely exposed (1,56,60,62,63). Considerable rotational freedom is envisaged within our model for an IgG molecule relative to the MEM83 epitope. Because of steric considerations of the C 2 symmetry, a single IgG molecule binding both LFA-1 I-domains within the heterotetramer is unlikely.
MEM83 acts as a blocking antibody for the interaction of LFA-1 with ICAM-3 (42,47), which is likely due to overlap of the MEM83 epitope residues and the I-domain/ICAM-3 epitope D (Asp 182 -Ser 184 ) (64). MEM83 was also reported to completely block the interaction of LFA-1 with ICAM-4 (50). We note that, unlike ICAM-1 and ICAM-3, ICAM-2 and ICAM-4 have only two extracellular Ig-like domains (65). Ori-entation of the IgGs within our model exhibits how the MEM83 antibody could block an interaction of ICAM-4 with LFA-1 (Fig. 4B).
Our model can readily incorporate ICAM-1 D4/D4-domain dimerization to form a one-dimensional array of adjacent heterotetramers (66), together with MEM83-linked I-domain dimerization. This would link an increase in avidity to an increase in affinity of LFA-1. In this manner, clustering induced by the MEM83 antibody may be analogous to that induced by the T cell receptor in the immunological synapse of the T cell (23).
In light of our data indicating that the binding of MEM83 to isolated I-domains is activation-insensitive, we hypothesize that MEM83 can stabilize adjacent LFA-1 molecules only while they are in their active form. This stabilization occurs via I-domain binding to each arm of the IgG hypervariable regions. The resting state of LFA-1 is the closed/low affinity form, and crosslinking this form by a non-blocking antibody would not enhance integrin activation. Alternately, a non-blocking antibody such as MEM83, which can recognize two I-domain epitopes across two LFA-1/ICAM-1 heterotetramers, would provide the additional free energy necessary to stabilize the active form of LFA-1. Hence, this array-mediated stabilization of the open/high affinity conformation of the integrin leads to outside-in signaling.