The I Domain of Integrin Leukocyte Function-associated Antigen-1 Is Involved in a Conformational Change Leading to High Affinity Binding to Ligand Intercellular Adhesion Molecule 1 (ICAM-1)*

On T cells the leukocyte integrin leukocyte function-associated antigen-1 (LFA-1) (CD11a/CD18) can be induced to bind its ligand intercellular adhesion molecule 1 (ICAM-1) (CD54) either by increasing the affinity of the receptor with Mg2+ and EGTA or by receptor clustering following activation with phorbol ester. The existence of these two adhesion-inducing pathways implies that alternative mechanisms might exist by which LFA-1 engages ICAM-1. The LFA-1 α subunit I domain contains a major binding site for ICAM-1. In this study we show that soluble LFA-1 I domain blocks ICAM-1 binding of the high affinity Mg2+-induced form of LFA-1 but not the phorbol ester-induced form. Under conditions of Mg2+-activation, the soluble I domain also prevents expression of an activation dependent epitope on LFA-1, implying that it inhibits a conformational change necessary for conversion to the high affinity form of this integrin. In addition, the binding of Mg2+-activated LFA-1 to ICAM-1 is blocked by peptides covering the α4-β3 loop, the β3-α5 loop, and the α5 helix of the I domain, whereas none of the peptides tested blocks phorbol ester-mediated adhesion. The blocking peptides localize to the same face of the crystal structure of the LFA-1 I domain and define an area that, during activation, may be involved in association of the I domain with another region of LFA-1, potentially the β-propeller domain. This is the first evidence linking a structural domain of an integrin, in this case the I domain, with a particular activation mechanism.

requires activation, with the characteristics of the active LFA-1 depending on the method of stimulation. In vitro, triggering of T cells with phorbol ester or by increasing intracellular Ca 2ϩ concentration does not induce a detectable change in the affinity of the individual LFA-1 molecules. These stimuli, however, induce clustering of LFA-1, thereby increasing the overall strength of binding (4,5), which is described as an increase in avidity. In contrast, activation from the outside of the cell with Mg 2ϩ and EGTA results in the formation of a higher affinity form of LFA-1, as assessed by an increased ability to bind soluble ICAM-1, and in expression of an activation reporter epitope recognized by mAb 24 (1,6,7). In this situation there is no evidence for LFA-1 clustering (5). These findings and those of others (8,9) suggest that physiological stimuli regulate adhesion by two major mechanisms involving alterations either in the affinity of the individual integrin molecules or in the overall avidity of binding. Such different forms of integrin adhesion might dictate the nature of signals transmitted into the cell and have functional consequences. It is not completely understood how these two forms of adhesion relate to the in vivo activation of LFA-1; however, the idea of a two-stage model of activation has been proposed in which integrin clustering is followed by a ligand-induced affinity increase (1).
Whether LFA-1 adhesion activated by different stimuli involves distinct regions within the integrin is unknown. Several ligand binding sites have been identified that colocalize with sites where divalent cation is also thought to bind (reviewed in Refs. 10 and 11). Most attention has focused on the ligand binding activity of the "I" domain, a ϳ200 amino acid region inserted into the amino-terminal region of 7 of the 16 integrin ␣ subunits (reviewed in Refs. 1 and 10). Isolated I domains can bind directly to ligand (12)(13)(14)(15)(16)(17). The crystal structures of the LFA-1 (18,19), Mac-1 (CD11b/CD18) (20,21), and ␣2␤1 (CD49b/CD29) (22) I domains show homology to the classical dinucleotide binding fold (the Rossmann fold), composed of a central sheet of 6 ␤ strands surrounded by 7 helices. A Mg 2ϩ / Mn 2ϩ binding site, termed the metal ion-dependent adhesion site (MIDAS), is located at the "top" of the domain (20) and is conserved between I domains. This motif is critical for ligand binding in LFA-1 (23,24), Mac-1 (13,23,25,26), and ␣2␤1 (15,27). The seven N-terminal repeat sequences of a typical integrin ␣ subunit have been modeled as a ␤-propeller fold (28), and the I domain is predicted to bind to the upper surface of this model, resembling the relationship between the ␣ and ␤ subunit of a heterotrimeric G protein (29).
In this study, we use recombinant I domain and constituent peptides in order to understand how this domain participates in LFA-1 binding to ICAM-1 when different integrin activators are used. We demonstrate that soluble I domain inhibits Mg 2ϩinduced LFA-1-mediated adhesion to ICAM-1 but not phorbol ester-induced binding. We provide evidence to suggest that a conformational change in LFA-1, which leads to the high affinity form, is blocked by the soluble I domain. Using peptides, the inhibiting activity is pinpointed to two loops associated with the ␣4 and ␣5 helices that lie on one face of the I domain.

Preparation of LFA-1 I Domain and ICAM-1Fc Proteins
The LFA-1 I domain residues Leu 111 -Ser 327 , termed (Leu 111 -Ser 327 )I dom, includes the predicted I domain sequence and an amino-terminal extension of ϳ17 amino acids (30). It was made by a polymerase chain reaction from a full-length LFA-1 ␣ subunit cDNA using a standard protocol. The polymerase chain reaction oligos were used to add restriction sites at the 5Ј and 3Ј ends and the digested product was ligated into pGEX2T (31) to encode a glutathione S-transferase fusion protein.
Sequencing was carried out using an automated sequencer (Applied Biosystems). The expressed glutathione S-transferase fusion protein was purified as described previously (17), and was directly cleaved from the glutathione S-transferase moiety using thrombin (Amersham Pharmacia Biotech) while still attached to the glutathione-Sepharose (Amersham Pharmacia Biotech). The electrophoretic mobility of the isolated protein, as determined by SDS-polyacrylamide gel electrophoresis, corresponded to the predicted size of ϳ25 kDa.
A second I domain construct, (Gly 128 -Ser 327 )I dom, the sequence of which conforms to the boundaries of the LFA-1 I domain, was also made as a glutathione S-transferase fusion protein. (Gly 128 -Ser 327 )I dom was isolated in two stages, involving elution of the glutathione S-transferase chimeric protein from glutathione-Sepharose followed by thrombin cleavage. Both (Leu 111 -Ser 327 )I dom and (Gly 128 -Ser 327 )I dom proteins expressed 10 I domain mAb epitopes, assessed as described previously (12). In comparison with (Leu 111 -Ser 327 )I dom, (Gly 128 -Ser 327 )I dom was less soluble, produced in lower yield, and although intact as assessed by SDS-polyacrylamide gel electrophoresis, lost ICAM-1 binding activity after short term storage at 4°C. This inactive protein was used as a negative control in the relevant experiments.
The ICAM-1Fc chimeric protein contained the five extracellular domains of ICAM-1 fused to a human IgG 1 Fc sequence (32) and was isolated from a transfected Chinese hamster ovary cell line after 10 days in serum-free medium (17).

T Cell Adhesion to ICAM-1Fc
Standard T Cell Adhesion Assay-T cell culture was performed as described previously (7), and the cells were labeled in either of two ways. Five ϫ 10 7 human T cells were labeled with 25 Ci of [ 3 H]thymidine overnight at 37°C then washed in RPMI or HEPES buffer (20 mM HEPES, 140 mM NaCl, 2 mg/ml glucose, pH 7.4). Alternatively, the T cells (4 ϫ 10 6 /ml) were labeled with 2.5 M BCECF/AM (Calbiochem) for 30 min at 37°C, then washed and resuspended in HEPES buffer. Immulon 1 96-well plates (Dynatech) were coated with ICAM-1Fc (240 ng/well) overnight at 4°C. Nonspecific sites were then blocked with 2.5% bovine serum albumin in phosphate-buffered saline for 1 h at room temperature, followed by 4 washes in the appropriate buffer. Labeled T cells (2 ϫ 10 5 in 50 l) were added to the ICAM-1Fc-coated wells in the presence or absence of 50 l of proteins, peptides or mAbs in HEPES buffer containing 2 mM MgCl 2 , 2 mM EGTA (2ϫ final concentration) or 0.6 mM MgCl 2 , 2 mM CaCl 2 , 100 nM phorbol 12,13-dibutyrate (PdBu) (2ϫ final concentrations). The plates were spun for 1 min at 30 ϫ g and incubated for 30 min at 37°C, followed by two to four washes with prewarmed RPMI or HEPES containing the appropriate cations. According to labeling procedure, the bound cells were either harvested and radioactivity measured using a beta-counter (LKB) or quantified using a fluorescence plate reader (Fluoroskan II, Labsystems). Function blocking mAbs specific for the LFA-1 ␣ subunit (mAb 38; IgG 2a ) and ICAM-1 (mAb 15.2; IgG 1 ) and control mAbs 4U (IgG 2a ) and 52U (IgG 1 ), all at 10 g/ml, were included as controls in all assays.
Strength of Adhesion Assays-To measure the strength of adhesion, BCECF/AM-labeled T cells were allowed to adhere to ICAM-1Fc for 30 min at 37°C as above. Each well was then filled with HEPES buffer, and the plate was tightly sealed with adhesive sealing film (SealPlate, Anachem). Care was taken to avoid air bubbles. Sealed plates were gently inverted and centrifuged at the relevant force (411-1643 ϫ g) for 8 min at room temperature. Plates subjected to the 1 ϫ g force were sealed, inverted, and allowed to stand for 8 min at room temperature. After centrifugation the plate sealers were removed and the inverted plates were allowed to drain. Cells remaining bound to the ICAM-1Fccoated surface were quantified using the fluorescence plate reader.
Preincubation Assays-Preincubation of T cells or immobilized ICAM-1Fc with I domain was performed prior to the adhesion assay as follows. For plate preincubation, plates that had been ICAM-1Fc-coated and bovine serum albumin-blocked were incubated for 30 min at 37°C with 50 l of proteins or mAbs in HEPES buffer containing 2 mM MgCl 2 , 2 mM EGTA. The solutions were removed, and 50 l of HEPES buffer containing 2 mM MgCl 2 , 2 mM EGTA were added followed by the addition of 50 l of BCECF/AM-labeled cells (2 ϫ 10 5 /well). For cell preincubation, BCECF/AM-labeled cells (2-4 ϫ 10 6 /ml) in HEPES buffer containing 1 mM MgCl 2 , 1 mM EGTA were incubated in the presence or absence of proteins or mAbs for 15 min at 37°C. After centrifugation for 1 min at 30 ϫ g, the solutions were removed, and the cells were resuspended in HEPES buffer and added to ICAM-1Fccoated, bovine serum albumin-blocked, and washed wells containing 50 l of HEPES buffer and 2 mM MgCl 2 , 2 mM EGTA. The assay was carried out as outlined above.

Detection of LFA-1 Epitope Expression by Flow Cytometry
T cells were incubated at 2 ϫ 10 6 /ml on ice for 15 min in HEPES buffer plus 0.5 mM MgCl 2 , 1 mM EGTA, and (Leu 111 -Ser 327 )I dom. Anti-LFA-1 mAbs (mAb 24; G25.2 (Becton Dickinson); S6F1 (a kind gift from Chikao Morimoto, Boston); TS2/4 (ATCC)) were then added at 10 g/ml. After a 30-min incubation at 37°C the cells were washed three times in FACSWASH (phosphate-buffered saline ϩ 0.2% bovine serum albumin ϩ 0.1% sodium azide). The T cells were resuspended at 4 ϫ 10 6 /ml in fluorescein isothiocyanate-conjugated goat anti-mouse IgG (Sigma) diluted in FACSWASH, washed three times as above, fixed in 2% formaldehyde in phosphate-buffered saline, and analyzed on a FACScan flow cytometer (Becton Dickinson). In experiments testing the effect of peptides on mAb 24 expression, peptides were added instead of the soluble I domain, and the same procedure was followed.

Preparation and Characterization of Synthetic Peptides
Peptides spanning the LFA-1 I domain were synthesized on a model 430A Applied Biosystems solid phase synthesizer using 9-fluorenylmethoxycarbonyl chemistry (33) and characterized by amino acid analysis and plasma desorption mass spectrometry. A peak corresponding to the expected molecular weight was found for each peptide, and in all cases peptides contained the correct sequence. Synthesized peptides were desalted using Sephadex G10 (Amersham Pharmacia Biotech) to remove potentially cytotoxic compounds and stored as a lyophilized powder under desiccating conditions at room temperature. All peptides used in these experiments were tested for cytotoxic activity using an 3-(4,5-dimethyl thiazol-2-yl)-2,5-diphenyl tetrazolium bromide assay with modification as described previously (34) and were tested at concentrations at which they were fully soluble.

LFA-1 I Domain Blocks Mg 2ϩ but Not Phorbol Ester-induced
Adhesion-The role of the I domain in different forms of LFA-1-mediated adhesion to ICAM-1 was investigated by testing the ability of soluble recombinant (Leu 111 -Ser 327 )I dom to interfere with Mg 2ϩ /EGTA or PdBu-stimulated T cell binding to immobilized ICAM-1. Experimental conditions were chosen such that similar percentages of T cells added to the reaction adhered with both stimuli. The I domain inhibited Mg 2ϩ /EGTAinduced adhesion in a dose-dependent manner with half-maximal blocking at 2 M, but did not interfere with PdBu-induced LFA-1 adhesion (Fig. 1A). Both forms of adhesion were completely blocked by mAbs against ICAM-1 or LFA-1 ( Fig. 1B and data not shown). Inactive I domain (control protein, see "Experimental Procedures") did not interfere with Mg 2ϩ -induced adhesion (Fig. 1B) indicating that the blocking effect of the active I domain was not due to nonspecific protein interactions.
One possible reason for the failure of the (Leu 111 -Ser 327 )I dom to block PdBu-mediated adhesion is that PdBu might induce a higher overall binding strength than is achieved with Mg 2ϩ /EGTA and therefore the binding of Mg 2ϩ /EGTA-treated cells would be more easily inhibited than the binding of cells treated with phorbol ester. To investigate this we titrated the amount of anti-ICAM-1 mAb 15.2 required to prevent adhesion ( Fig. 2A). The titration curves overlap, showing that both forms of adhesion are inhibited equally by a blocking antibody. To substantiate these data we compared the overall adhesive strength of Mg 2ϩ /EGTA-and of PdBu-activated cells by a second method. T cells were allowed to adhere to ICAM-1 and then the strength of adhesion was evaluated by a quantitative centrifugal removal assay (35), in which the cells were subjected to increasing relative centrifugal forces (Fig. 2B). The number of cells that remained bound to ICAM-1 at each relative centrifugal force was the same for both forms of adhesion indicating that treatment of cells with either 1 mM Mg 2ϩ and 1 mM EGTA or with 50 nM PdBu results in a very similar overall strength of adhesion.
We have previously shown that (Leu 111 -Ser 327 )I dom binds to ICAM-1 under conditions similar to the Mg 2ϩ /EGTA activation procedure (17) raising the possibility that the lack of inhibition of PdBu-mediated T cell adhesion is due to the inability of soluble I domain to bind ICAM-1 under the PdBu assay conditions. When investigated by solid phase enzyme-linked immunosorbent assay, however, the soluble I domain was found to bind ICAM-1 to a similar extent when either PdBu or Mg 2ϩ /EGTA assay buffers were used (data not shown).
One explanation for the blocking effect of (Leu 111 -Ser 327 )I dom might be that it competes for ICAM-1 binding with Mg 2ϩ / EGTA activated LFA-1 but not with PdBu activated LFA-1. Another explanation is that soluble I domain inhibits by interacting with LFA-1 on T cells thus preventing a conformational change essential for high affinity binding induced by Mg 2ϩ / EGTA. We attempted to discriminate between these two possibilities by preincubating either T cells or immobilized ICAM-1 with (Leu 111 -Ser 327 )I dom, then removing unbound protein prior to the adhesion assay (Fig. 3). Neither preincubation of T cells nor of immobilized ICAM-1 with soluble I domain caused inhibition of T cell adhesion, therefore preventing assessment, using this approach, of which of the two mechanisms of inhibition applies. The need for the constant presence of the I domain in the assay, however, implies that the interaction of the I domain with ICAM-1 or T cells is of low affinity.
Soluble I Domain Inhibits the Expression of the Activation Epitope 24 on T Cell LFA-1-One of the characteristics of Mg 2ϩ -induced LFA-1 adhesion is the expression of an activation reporter epitope recognized by mAb 24 (1, 6, 7). Exposure of this epitope is thought to reflect a conformational change in LFA-1 that leads to a high affinity receptor. PdBu-stimulated adhesion, in contrast, is not associated with any significant increase in mAb 24 expression (4). In order to assess whether soluble I domain interacts with T cells under the condition of Mg 2ϩ -activation, we investigated whether it affects the expression of the mAb 24 epitope. T cells were incubated with (Leu 111 -Ser 327 )I dom and Mg 2ϩ /EGTA in the same manner as for an adhesion assay (except that they were not exposed to immobilized ICAM-1) and the expression of mAb 24 was measured by flow cytometry. Soluble I domain inhibited mAb 24 expression in a dose-dependent manner (Fig. 4, A and B). However, soluble I domain did not affect the expression of three mAbs, S6F1, G25.2, and TS2/4, which, like mAb 24, are not expressed on the I domain (12), 2 but are directed against epitopes on the ␤-propeller domain of LFA-1 (36) (Fig. 4, C and D, and data not shown). These data show that the soluble I domain interferes specifically with expression of a conforma-2 B. Leitinger, unpublished results. tion-sensitive epitope on LFA-1, indicating that there is a direct interaction of the soluble I domain with T cells. This implies that the I domain inhibits T cell binding to ICAM-1 by interacting with LFA-1 and preventing a conformational change necessary for conversion to the high affinity form of this integrin.
The Effect of I Domain Peptides on Mg 2ϩ /EGTA and PdBuinduced T Cell Binding to ICAM-1-To characterize the regions within the I domain required specifically for Mg 2ϩ -stimulation, constituent peptides (Table I) were tested for their ability to interfere with ICAM-1 binding of T cells following the two protocols already mentioned. When T cell LFA-1 was activated by treatment with Mg 2ϩ /EGTA, only peptides I(217-233) and I(238 -254) consistently blocked LFA-1 binding to ICAM-1 (Fig.  5A). The average level of inhibition for I(238 -254) was 72% and for I(217-233) was 50% at 2 mM (n ϭ 13). It is of interest that I(130 -159), which covers the major part of the MIDAS motif (D 137 XSXS), was not inhibitory at 2 mM (data not shown). As anticipated, when LFA-1-mediated adhesion was induced by PdBu, I(217-233), I(238 -254), and the other peptides tested had no blocking activity (Fig. 5B). This result indicates that I domain regions I(217-233) and I(238 -254) are specifically involved in Mg 2ϩ -induced recognition of ICAM-1.
In order to further pinpoint the critical residues, we made a set of overlapping peptides concentrating on the sequence Thr 217 -Ile 254 (Table I) peptide I(238 -254), whereas control peptide I(256 -275) did not interfere at the same concentrations (Fig. 6A). We considered peptide I(242-254) to represent the minimally active sequence. Similarly, peptide I(223-233) corresponded in blocking activity to the parent peptide I(217-233) (data not shown). Unfortunately, peptide I(210 -223) had limited solubility and could not be tested. The dose-dependent inhibition of T cell binding to ICAM-1 by I(223-233) and I(238 -254) was compared with scrambled versions of both these peptides (see Table I), which did not significantly affect binding (Fig. 6B).
To test whether the peptide inhibition occurs by the same mechanism as soluble I domain inhibition of T cell binding to ICAM-1, we analyzed the effect of peptide I(238 -254) on mAb 24 expression. Fig. 7 shows that I(238 -254) at 2 mM, but not its scrambled control, inhibited mAb 24 expression. The average level of inhibition of 24 epitope expression was 55% (n ϭ 4). These results indicate that I(238 -254) and soluble I domain interact with T cells in a similar manner.
Mapping of Blocking Peptides on the LFA-1 I Domain Structure-When the two function blocking peptides are superimposed on the crystal structure of the LFA-1 I domain (18), they span two adjacent areas. The sequence Glu 223 -Val 233 corresponds to the loop following the ␣4 helix which terminates in the ␤3 strand and the sequence Ala 242 -Ile 254 consists of the ␤3-␣5 loop and follows the loop to the end of the small ␣5 helix (see Fig. 8A).
The seven N-terminal repeated domains of an integrin ␣ subunit have recently been proposed to form a seven-bladed ␤-propeller structure (28) homologous to the heterotrimeric G protein ␤ subunit (37). Springer (28) has modeled the Mac-1 I domain, which is structurally homologous to the G protein ␣ subunit, onto the integrin ␤-propeller domain, mimicking the G protein ␣␤ interaction (29). We have similarly superimposed 32 C ␣ atoms corresponding to the equivalent ␤-strands of the central ␤-sheet of the LFA-1 I domain (18) onto the G protein ␣ subunit coordinates (root mean squared, 2.6 Å) of the G protein ␣␤ complex (29) (Fig. 8B). Orientation of the I domain relative to the ␤-propeller domain is therefore analogous to the binding of G␣ to G␤. In this model the I(242-254) peptide lies over the central axis of the ␤-propeller domain and the other peptide, I(223-233), forms the only point of close contact between the I domain and the ␤-propeller fold. The validity of the latter assignment will depend on the length and orientation of the linker regions between the I domain and the ␤-propeller.

DISCUSSION
The main findings of this study are: (i) that soluble LFA-1 ␣ subunit I domain interferes with Mg 2ϩ -activated but not phorbol ester-induced LFA-1 adhesion to ICAM-1, indicating that it has a distinct role in adhesion of high affinity LFA-1; (ii) that a key region required for Mg 2ϩ -activation is localized to one face of the I domain and includes the ␣4-␤3 loop, the ␤3-␣5 loop, and the ␣5 helix; and (iii) that this region in the I domain is involved in a conformational change necessary for conversion to the high affinity form of LFA-1. This is the first evidence linking a conformational alteration involving a structural element, in this case the I domain, with a particular activation mechanism for an integrin.
There are several possible explanations for why the soluble I domain blocks binding to immobilized ICAM-1 of the high affinity Mg 2ϩ -induced form but not of the low affinity PdBu form of LFA-1. Although both of these methods of activation of LFA-1 have distinct characteristics, they result in a similar overall strength of cell adhesion, thus discounting this as a possible reason for the difference in behavior of the soluble I domain. Another possibility is that the two forms of LFA-1 bind ICAM-1 in distinctive ways; however, there is no evidence so far to suggest more than one type of binding site on ICAM-1. Both forms of adhesion are blocked in a similar manner by anti-ICAM-1 mAb 15.2, as seen in this study, and by mAbs to

DIDTKAEGAD.TAINSDG
other epitopes on ICAM-1 (32). 3 In addition, both PdBu-and Mg 2ϩ -induced LFA-1 binding are equally sensitive to mutation of Glu 34 or Gln 73 in ICAM-1. 4 These mutations have been shown to severely disrupt LFA-1 binding (17,38). Together, these results indicate that the two forms of LFA-1 recognize the same critical features of the ligand-binding motif on ICAM-1.
The I domain contains the major binding site for ICAM-1 (12,39), with the region implicated in ligand binding surrounding the metal binding MIDAS motif (18 -20, 22). It is speculated that this motif may directly contact ligand and that a metal ion may provide a bridge between the ligand and I domain (17, 40 -42). Mutagenesis and switching ␣M for ␣L segments or mouse for human sequences around the MIDAS motif have further confirmed the ligand binding status of this area (36,43,44). The distinctive role for the I domain in Mg 2ϩ -induced adhesion could be explained if the I domain operates as a ligand binding site only for this form of adhesion. However, the same anti-I domain mAbs interfere with both PdBu-induced as well as Mg 2ϩ -induced LFA-1 adhesion (4,39,45), which is consistent with the idea that the I domain serves as the ligand binding site for both forms of LFA-1.
One feature that distinguishes the Mg 2ϩ -induced form of LFA-1 from the PdBu-induced form is the expression of an activation reporter epitope recognized by mAb 24 (6,7), the exposure of which is thought to result from a conformational change. The inhibition of expression of this epitope by soluble I domain suggests that the soluble I domain prevents a conformational change within LFA-1 required for transition to the high affinity form of this integrin. Furthermore, the reduction in 24 epitope expression implies that the inhibitory effect of the soluble I domain occurs via interaction with T cell LFA-1 rather than immobilized ICAM-1. Although the 24 epitope has not been precisely mapped on the LFA-1 ␣ subunit (6), it is certain that it does not reside within the I domain (12). 5 This suggests that an interdomain movement involving the I domain is associated with exposure of the cryptic 24 epitope located elsewhere in LFA-1 (46) and that we are not measuring a conformational change in the I domain itself.
The inhibitory I domain peptides used in this study mimicked the behavior of the soluble I domain in preventing Mg 2ϩinduced adhesion, and for the I(238 -254) peptide in preventing mAb 24 epitope expression. The LFA-1 I domain crystal structure shows that these inhibitory sequences cover the ␣4-␤3 loop (Glu 223 -Val 233 ) and the ␤3-␣5 loop to the ␣5 helix (Ala 242 -Ile 254 ) (18). These regions are spatially close, lying on the same face of the domain. For other I domain-containing integrins, there is evidence that adjacent loops distal to the MIDAS motif can influence the ability of the integrin to function. For example, the mutation of the ␣2-␣3 loop and ␣1-␤D loop in Mac-1 I domain resulted in an active integrin able to spontaneously bind fibrinogen (47). Moreover, mutation of Pro 192 or Pro 195 within the ␣2-␣3 loop of LFA-1 and Mac-1, respectively, interfered with adhesion to their ligands ICAM-1 (24) and iC3b (25).
Recently the seven N-terminal repeat sequences of a typical integin ␣ subunit have been modeled as a ␤-propeller fold (28) homologous to the heterotrimeric G protein ␤ subunit (37). The I domain, which is structurally homologous to the G protein ␣ subunit, can be positioned on the upper surface of the ␤-pro- peller, to one side of the central axis, mimicking the relationship between G protein ␣ and ␤ subunits (28,29). Looking down on the central axis of the ␤-propeller (see Fig. 8B), the Ala 242 -Ile 254 sequence protrudes out over the central cavity, and the Glu 223 -Val 233 sequence potentially makes contact with the ␤-propeller domain. The positioning of these ␣4 and ␣5 loop regions suggests that they represent possible contact points between the I domain and the ␤-propeller. The speculation is that the conformational change in LFA-1 that occurs upon Mg 2ϩ -activation involves movement between the I domain and the ␤-propeller domain. The interpretation of our results is that soluble I domain and peptides representing the ␣4 and ␣5 loop regions compete with cellular I domain in intact LFA-1 for binding to the ␤-propeller domain, thereby physically blocking activation. The fact that the soluble I domain does not block LFA-1 activation by PdBu implies that PdBu stimulation does not involve conformational changes involving the I domain, in keeping with the fact that PdBu does not cause an affinity alteration or a detectable conformational change in LFA-1. This does not preclude the involvement of I domain of intact LFA-1 in ligand binding following PdBu stimulation.
Although crystallographic studies show that changes in the tertiary structure of the I domain are possible (21) these would not have been detected in our study. From our data we conclude that the addition of the soluble I domain or peptides interferes with alterations in interdomain or intersubunit associations and thereby prevents changes in the quarternary structure of LFA-1. Our results give rise to the idea that activation of LFA-1 to a high affinity form involves interdomain movement, which alters the association of the I domain with the ␤-propeller domain, and in this way leads to structural changes that increase the affinity for ligand.