Characteristics of Cation Binding to the I Domains of LFA-1 and MAC-1

The crystal structures of the I domains of integrins MAC-1 (αMβ2; CD11b/CD18) and LFA-1 (αLβ2; CD11a/CD18) show that a single conserved cation-binding site is present in each protein. Purified recombinant I domains have intrinsic ligand binding activity, and in several systems this interaction has been demonstrated to be cation-dependent. It has been proposed that the I domain cation-binding site represents a general metal ion-dependent adhesion motif utilized for binding protein ligands. Here we show that the purified recombinant I domain of LFA-1 (αLI) binds cations, but with significantly different characteristics compared with the I domain of MAC-1 (αMI). Both αLI and αMI bind54Mn2+ in a conformation-dependent manner, and in general, cations with charge and size characteristics similar to Mn2+ most effectively inhibit54Mn2+ binding. Surprisingly, however, physiological levels of Ca2+ (1–2 mm) inhibited 54Mn2+ binding to purified αLI, but not to αMI. Using45Ca2+ and 54Mn2+ in direct binding studies, the dissociation constants (K D ) for the interactions between these cations and αLI were estimated to be 5–6 × 10−5and 1–2 × 10−5 m, respectively. Together with the available structural information, the data suggest differential affinities for Mn2+ and Ca2+binding to the single conserved site within αLI. Antagonism of LFA-1, but not MAC-1, -mediated cell adhesion by Ca2+ may be related to the Ca2+ binding activity of the LFA-1 I domain.

LFA-1 (␣ L ␤ 2 ) and MAC-1 (␣ M ␤ 2 ) are closely related leukocyte integrins that are essential for normal immune system functions (1,2). Both integrins bind to several distinct ligands, but share in the ability to bind ICAM-1, a widely expressed cell surface protein (3)(4)(5). Recent studies suggest that in the ␤ 2 and other non-RGD binding integrins, additional and perhaps multiple subdomains within both the ␣ and ␤ subunits may contribute to form the complete ligand-binding domain. One of these subdomains is the I domain (also known as the A domain), a region of approximately 200 amino acids found in a variety of proteins as well as the ␣ subunit ectodomain of all ␤ 2 integrins and VLA-1 (␣ 1 ␤ 1 ), VLA-2 (␣ 2 ␤ 1 ), and ␣ E ␤ 7 (6,7).
Several lines of evidence suggest that the I domain in the context of the complete integrin may play a significant and direct role in ligand binding. The activities of both integrin-neutralizing and integrin-activating monoclonal antibodies, and NIF, a hookworm-derived MAC-1 inhibitor have been mapped to the I domain region (8 -14). Significantly, purified recombinant forms of the I domains derived from LFA-1, MAC-1, and VLA-2 have intrinsic ligand binding activity (12,(15)(16)(17)(18)(19) and in several systems it has been shown that this protein-protein interaction is cation-dependent (12,19). Determination of the I domain crystal structures has provided a structural basis for conceptualizing the role of cations in ␣ M I and ␣ L I interaction with ligands (19 -21). The structures show that ␣ M I and ␣ L I domains contain a single metal cation-binding site, and that residues involved in coordinating the metal ion in each protein are completely conserved (21). Lee and co-workers (19) proposed that this novel cation-binding site represents a general metal ion-dependent adhesion (MIDAS) 1 motif for binding protein ligands. Interestingly, crystallization of ␣ M I in the presence of different cations has been shown to result in significant changes in metal coordination and protein structure (19), suggesting that differential effects of cation binding on integrin function may be possible.
Divalent cations have multiple effects on integrin-mediated cell adhesion including enhancement, suppression, and modification of ligand binding activity. Mg 2ϩ and Mn 2ϩ induce conformational alterations of several integrins, including LFA-1 and MAC-1, concomitant with activation of integrinmediated adhesion to ligands (22)(23)(24). In contrast, Ca 2ϩ has been shown to inhibit LFA-1, but not MAC-1, mediated adhesion to ligands (24 -28). We speculated that the differential effects of Ca 2ϩ and Mn 2ϩ /Mg 2ϩ on LFA-1 and MAC-1 function might be related to differences in the divalent cation binding properties of their I domains. In the work described herein, we compared the cation binding properties of purified recombinant ␣ L and ␣ M I domains. Our results indicate that like ␣ M I, ␣ L I preferentially binds Mn 2ϩ over most other cations. However, Mn 2ϩ interaction with ␣ L I was inhibitable to some degree using a variety of cations, and these studies revealed a pattern of binding selectivity that was clearly distinct from ␣ M I. Interestingly, Ca 2ϩ inhibited Mn 2ϩ binding to ␣ L I, but had little effect on Mn 2ϩ binding to ␣ M I, underscoring a fundamental difference in the cation-binding properties of these two I domains. Furthermore, experiments confirmed that ␣ L I binds both Mn 2ϩ and Ca 2ϩ and the results are consistent with the notion that the ␣ L I contains a single mixed-type Mn 2ϩ /Ca 2ϩbinding site. It is possible that the activation state of LFA-1 is regulated in part by the interaction of the I domain with cations present in the extracellular environment. Ca 2ϩ antagonism of LFA-1, but not MAC-1, mediated cell adhesion may be a consequence of the Ca 2ϩ binding activity of the I domain of LFA-1 in that the Mn 2ϩ and Ca 2ϩ complexes of ␣ L I may represent, respectively, high-and low-affinity ligand binding states.

EXPERIMENTAL PROCEDURES
Reagents-Purified maltose-binding protein (MBP2) was obtained from New England Biolabs (Beverly, MA). Purified bovine serum albumin (BSA) was obtained from Pierce (Rockford, IL). Concentrations of proteins were determined using the Bio-Rad Protein Assay Reagent (Bio-Rad) with BSA as standard.
All buffer solutions were prepared in single-use plastic containers. Water was obtained by a Milli-RO/Milli-Q water system (Millipore Corporation, Bedford MA). Reagents used for the preparation of buffers were of the highest quality available. The concentration of Ca 2ϩ and Mn 2ϩ in the water and overlay buffer (described below) was measured by inductively coupled plasma-atomic emission spectrometry (ICP-AES) using a Jarell Ash Atomcomp 975 (Thermo Jarell Ash Corp., Franklin, MA) instrument. The lower limit for detection of Ca 2ϩ and Mn 2ϩ in water and overlay buffer by ICP-AES was 0.50 and 0.55 M, respectively, and 2.5 and 0.91 M, respectively. The measured Ca 2ϩ and Mn 2ϩ concentrations in water and buffer were below the limit of detection by ICP-AES.
Expression and Purification of I Domain Polypeptides-␣ M I was expressed as a fusion protein with glutathione S-transferase (GST), purified from the soluble fraction of the cleared Escherichia coli lysate by affinity chromatography and then cleaved with thrombin to separate the I domain from the GST moiety. The ␣ M I domain was amplified by polymerase chain reaction, using as template plasmid pMON24304 (obtained from B. Harding, Monsanto Co., St. Louis, MO) that contains a 2.45-kilobase cDNA fragment that includes the I domain region of CD11b. The primers used were 5Ј-GCGGATCCAACCTACGGCAG-CAG-3Ј and 5Ј-GCGCGGCCGCGCAAAGATCTTCTCCCGAAG-3Ј. The fragment was digested with BamHI and NotI, cloned into pGEX 4T-1 (Pharmacia LKB, Uppsala, Sweden), and the accuracy of the DNA sequence was verified. The construct joins the glutathione S-transferase gene to the ␣ M I coding sequence and adds five vector-derived amino acids (RPHRD) at the carboxyl terminus. The plasmid was transformed into E. coli strain DH5␣, and expression and purification of the ␣ M I polypeptide was carried out essentially by the procedure of Michishita et al. (29), in which glutathione-Sepharose 4B affinity chromatography is followed by thrombin digestion and gel filtration using a Superose 12 column (Pharmacia LKB, Uppsala, Sweden).
A similar strategy was employed for the expression of the I domain of LFA-1. A cDNA encoding the full-length ␣ L (CD11a) protein was identified using standard hybridization techniques to screen a human inflamed colon library constructed in the ZAP II vector (Stratagene, San Diego, CA). The ␣ L I coding region of the full-length ␣ L (CD11a) template was amplified by polymerase chain reaction using the following oligonucleotides: 5Ј-GCGGATCCAATCTGCAGGGTCCCATGCTG-3Ј and 5Ј-GCGAATTCAGCTCCATGTTGAAGGAAGT-3Ј. The product was digested with BamHI and EcoRI, cloned into pGEX-2T (Pharmacia LKB, Uppsala, Sweden), and sequenced. This construct joins the glutathione S-transferase gene to the ␣ L I-coding sequence and adds six vectorderived amino acids (NSIVTD) at the carboxyl terminus. ␣ L I expressed in E. coli as a fusion protein with GST and grown under the conditions utilized for expression of ␣ M I formed intracellular inclusion bodies and uncleavable fusion protein. However, modification of the conditions for expression of the GST-␣ L I fusion as described below yielded soluble and cleavable fusion protein. Transformed cells were grown in 3 liters of M9 minimal medium containing 50 g/ml proline at room temperature to an optical density (A 600 ) of 0.7. Expression was induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside for 24 h at room temperature. Cells were pelleted, stored at Ϫ70°C overnight, and then lysed by sonication in 120 ml of column buffer (200 mM NaCl, 5 mM MgCl2, 20 mM Tris, pH 7.4). Triton X-100 (0.1%) was added and the lysate incubated for 30 min at 4°C. The sample was then centrifuged, filtered (0.2 M), and incubated with 3 ml of a 50% slurry of glutathione-Sepharose 4B (Pharmacia Biotech, Uppsala, Sweden) for 30 min at room temperature. The resin was washed 4 times with column buffer, and the protein was eluted with 10 mM reduced glutathione, 50 mM Tris-HCl, pH 8.0. The sample was treated in column buffer with 1 unit of restriction grade thrombin (Novagen, Madison, WI) per mg of protein for 16 h at room temperature, and digestion was terminated with 1 mM phenylmethylsulfonyl fluoride. After dialysis of the sample against 2000 volumes of column buffer, free GST was removed by batch adsorption using the glutathione resin. The remaining protein was then charac-terized by protein sequencing, MALDI mass spectroscopy, and SDS-PAGE.
The GST-␣ L I fusion protein expressed by the methods described above yielded relatively small quantities of soluble fusion protein. To generate larger amounts of protein needed for cation-binding studies, a soluble MBP (maltose-binding protein) fusion form of ␣ L I was produced. MBP-␣ L I was generated by polymerase chain reaction-mediated amplification of the I domain-coding region of the cDNA encoding the fulllength ␣ L (CD11a) using the following oligonucleotides as primers: 5Ј-GCGAATTCAATCTGCAGGGTCCCATGCTG-3Ј and 5Ј-GCAAGCTT-TCACAGCTCCATGTTGAAGGAAGT-3Ј. The product was digested with HindIII and EcoRI, cloned into pMAL-c2 (New England Biolabs), and the accuracy of the DNA sequence was verified. The resulting coding sequence joins the MBP to CD11a at amino acid Asn 110 and extends to Leu 324 corresponding ot the same residues contained in the free ␣ L I protein described above. An overnight culture of E. coli strain DH5␣ containing the expression plasmid was subcultured 1:10 into 2 liters of LB broth with ampicillin and grown at 37°C for 1 h. Expression was induced by the presence of 0.3 mM isopropyl-1-thio-␤-galactoside for 2 h. Cells were centrifuged and stored frozen at Ϫ80°C overnight. After thawing, cells were resuspended in 60 ml of MBP column buffer (20 mM Tris pH7.4, 200 mM NaCl, 5 mM MgCl 2 , 1 mM phenylmethylsulfonyl fluoride), frozen in a dry ice/ethanol bath, thawed, and sonicated on ice for 30-s intervals until lysis was achieved. The sample was centrifuged, and the cleared lysate filtered (0.2 M pore size) before applying to an amylose resin (New England Biolabs) column (70 ml bed volume) equilibrated with MBP column buffer. The column was washed extensively with MBP column buffer and the protein eluted into fractions with MBP column buffer containing 10 mM maltose. Protein identity and purity was confirmed by amino-terminal peptide sequencing and SDS-PAGE analysis, respectively.
Cation Binding Assays-54 Mn 2ϩ and 45 Ca 2ϩ binding to purified proteins was measured by procedures based on those previously described by Michishita et al. (29). Equimolar amounts of purified recombinant proteins were bound to nitrocellulose (PH79, 0.1 M pore size, Schleicher and Schuell, Keene, NH) using the Mini-Fold II Slot Blot System (Schleicher and Schuell). Filters were washed twice for 10 min at room temperature with overlay buffer (10 mM imidazole, pH 6.8, 60 mM KCl) containing 10 mM EDTA, followed by four washes for 30 s with overlay buffer lacking EDTA. The filter was then incubated for 10 min in overlay buffer containing 0.5 Ci/ml 54 MnCl 2 (Ͼ40 Ci/g, NEN Life Science Products, Boston, MA) or 45 CaCl 2 (10 -75 Ci/g, NEN Life Science Products) in the presence or absence of cold competitor cations diluted from freshly prepared stock solutions. Filters were washed twice in 50% ethanol and dried. Bound radioactive material was measured either by autoradiography or by PhosphorImage analysis using a Molecular Dynamics (Sunnyvale, CA) PhosphorImager system. The efficiency of protein binding to nitrocellulose was assessed by staining the filter with naphthol blue black as described previously (30), and scanning using an LKB Ultrascan XL densitometer (LKB, Uppsala, Sweden). Binding to filters was determined to be dose-dependent and was highly consistent (i.e. less than 10% variation) between identically loaded wells. In some experiments using cold competitor cations, IC 50 values were derived from a two parameter logistic model using nonlinear regression analysis.
Estimation of Ca 2ϩ and Mn 2ϩ Equilibrium Dissociation Constants-The Ca 2ϩ and Mn 2ϩ binding parameters for MBP-␣ L I were examined using the methods described above with the following modifications. MBP-␣ L I was dialyzed exhaustively against 200 mM NaCl, 20 mM Tris, pH 7.5, and 5 mM MgCl 2 and passed through a 0.2-m filter prior to use. 45 Ca 2ϩ and 54 Mn 2ϩ were diluted with non-radioactive Ca 2ϩ and Mn 2ϩ , respectively, to obtain total concentrations of these cations that ranged from 0.01 to 10 mM. MBP-␣ L I was bound to filters, washed with overlay buffer containing EDTA to remove bound metal, and then rinsed exhaustively with overlay buffer to remove residual EDTA. Filters were incubated with various concentrations of 45 Ca 2ϩ and 54 Mn 2ϩ for 10 min at room temperature, removed, and washed with 50% ethanol as described above. The filters were air dried and the amount of Ca 2ϩ or Mn 2ϩ bound determined by PhosphorImager analysis. A 45 Ca 2ϩ or 54Mn 2ϩ standard curve was utilized to determine the specific activity of each labeled cation expressed as arbitrary units/pmole of cation.

Expression of Recombinant Proteins-
The I domain coding regions were derived by polymerase chain reaction amplification using CD11a and CD11b cDNA clones as templates as described under "Experimental Procedures." The sequence of the MAC-1 I domain (␣ M I) was identical to that reported previously (31). However, the sequence of the LFA-1 I domain (␣ L I), as determined from two independent clones from the cDNA library, varied from that reported by Larson et al. (32) at a single nucleotide that results in substitution of tryptophan 189 by arginine. This substitution has also been identified by other researchers (21) and may represent a natural allelic variation. The position of this residue lies on the surface of the protein, far from the MIDAS motif, and thus is unlikely to affect the metal binding characteristics.
␣ M I was expressed as a fusion protein with GST, purified from the soluble fraction of the cleared E. coli lysate by affinity chromatography, and then cleaved with thrombin to separate the I domain from the GST moiety. In contrast, ␣ L I expressed in E. coli as a fusion protein with GST, using conditions similar to those utilized for the expression of ␣ M I, formed intracellular inclusion bodies. Modification of the expression conditions, as described under "Experimental Procedures," yielded small amounts of soluble GST-␣ L I that was cleavable with thrombin.
In order to generate larger amounts of soluble protein, ␣ L I was expressed as a fusion protein with MBP. In this expression context, the protein was found almost exclusively in the soluble fraction and at high concentration. Unfortunately, cleavage of the MBP-␣ L I fusion protein with thrombin resulted in I domain aggregation and precipitation. Due to the apparent ability of the MBP moiety to enhance the solubility of the LFA-1 I domain and the high yields of recovered protein (e.g. up to 30 mg/liter of cell culture) the majority of experiments utilized ␣ L I expressed as a fusion protein with MBP. However, to verify that structural differences imposed by the nature of the MBP-␣ L I fusion protein construct did not contribute to ␣ L I cation binding properties, some experiments were conducted with ␣ L I cleaved and purified from the GST-␣ L I fusion protein. For the sake of clarity, we subsequently use the term ␣ L I f to refer to MBP-␣ L I fusion protein and the term ␣ L I i to refer to the purified I domain of LFA-1 isolated from the thrombin cleavage of the GST-␣ L I fusion protein. The accuracy of protease cleavage of the fusion proteins for the release of ␣ M I and ␣ L I i was confirmed by identification of the correct amino terminus by amino acid peptide sequencing. The molecular masses of ␣ M I and ␣ L I i , determined by MALDI mass spectroscopy, were within 0.05 and 0.01%, respectively, of their expected values. Purity of each protein was greater than 95% as determined by SDS-PAGE analysis (data not shown).
The I Domains of LFA-1 and MAC-1 Bind Metal Cations-To demonstrate that the LFA-1 and MAC-1 I domains contain qualitatively similar cation-binding sites, ␣ M I and ␣ L I f were immobilized on nitrocellulose filters and incubated with 54 Mn 2ϩ as described under "Experimental Procedures." The autoradiogram shown in Fig. 1A shows that both ␣ L I f and ␣ M I, but not BSA, bind 54 Mn 2ϩ and that unlabeled Mn 2ϩ competitively inhibits binding in a dose-dependent fashion. Quantitative PhosphorImager analysis of the nitrocellulose filters confirmed that 54 Mn 2ϩ binding to control proteins BSA or MBP was less than 10% of 54 Mn 2ϩ binding to either of the purified I domains (Fig. 1, B and C). Moreover, boiling of ␣ L I f prior to immobilization on nitrocellulose reduced 54 Mn 2ϩ binding to background levels. These results showed that the purified I domains of each integrin contain a conformationally sensitive Mn 2ϩ -binding site.
To assess possible qualitative variation in the cation binding activity of the I domain of each integrin, we screened various divalent and trivalent metal ions for the capacity to inhibit 54 Mn 2ϩ binding to ␣ M I and ␣ L I f . The results in Fig. 2 and Fig.  3A demonstrate that 500 M unlabeled Mn 2ϩ reduced 54 Mn 2ϩ binding to ␣ M I and ␣ L I f by approximately 90%. In these exper-iments, the concentration of 54 Mn 2ϩ was approximately 0.2 M, and hence the observed residual binding is likely to be nonspecific. In agreement with the earlier data of Michishita et al. (29), Mg 2ϩ , Ni 2ϩ , Co 2ϩ , Zn 2ϩ , and Cd 2ϩ all were excellent competitors of Mn 2ϩ binding to ␣ M I, as were two previously untested divalent cations, Fe 2ϩ and Cu 2ϩ (Fig. 2). Relative to The results shown in A and B were obtained from the same experiment. Boiling of ␣ L I f for 5 min prior to immobilization inhibited 54 Mn 2ϩ binding (C). 54 Mn 2ϩ binding to ␣ L I f , OE; to ␣ M I, q; and BSA, f. these divalent cations, those with higher states of oxidation (Cr 3ϩ , Fe 3ϩ , and Au(III)) were less effective and reduced the binding of 54 Mn 2ϩ by only approximately 50%. The cations with the largest ionic radii of those tested, Ba 2ϩ , Sr 2ϩ , and Ca 2ϩ , although divalent, had no effect on Mn 2ϩ binding to ␣ M I. In contrast to the results obtained with ␣ M I, all of the tested cations inhibited to some extent 54 Mn 2ϩ binding to ␣ L I f (Fig.  2B). However, the small trivalent cations (Fe 3ϩ and Cr 3ϩ ) and large divalent cations (Sr 2ϩ , Ba 2ϩ , and Ca 2ϩ ) were less effective inhibitors than the divalent cations with ionic radii more similar to Mn 2ϩ (e.g. Cd 2ϩ , Co 2ϩ , Fe 2ϩ , Zn 2ϩ , and Cu 2ϩ ). The smallest of the divalent cations tested, Ni 2ϩ and Mg 2ϩ , exhibited intermediate inhibitory activity. Notably, the Au(III) ion, which has an ionic radius very close to that of Mn 2ϩ , was among the best competitors. Of particular interest was the observation that the physiologically relevant Ca 2ϩ ion had no effect on 54 Mn 2ϩ binding to ␣ M I, but did inhibit 54 Mn 2ϩ binding to ␣ L I f by greater than 70%. Together the data suggest that ionic size may be a more important attribute than charge in determining cation binding specificity to ␣ L I, although the specific coordination geometry preferred by the cations may also play a roll. The data further suggest that the LFA-1 I domain may be complexed with Ca 2ϩ at physiological concentrations.
The I Domain of LFA-1, but Not MAC-1, Contains a Binding Site for Calcium-Ca 2ϩ inhibits LFA-1, but not MAC-1, mediated cell adhesion (24 -28, 33), an effect that conceivably may be related to differences in the divalent cation binding properties of the I domains of these integrins. Since the results presented above suggested that Ca 2ϩ binds to ␣ L I but not ␣ M I, we examined in further detail the binding of this cation to these purified polypeptides. The effect of increasing concentrations of unlabeled Mn 2ϩ and Ca 2ϩ on 54 Mn 2ϩ binding to ␣ M I and ␣ L I f is presented in Fig. 3A. In agreement with the results of Michishita et al. (29), unlabeled Mn 2ϩ and Ca 2ϩ inhibited 54 Mn 2ϩ binding to ␣ M I with estimated IC 50 values (inhibitory concentration resulting in 50% control binding) of 1-2 M and greater than 10,000 M, respectively. In contrast to these results, unlabeled Mn 2ϩ and Ca 2ϩ were both potent inhibitors of 54 Mn 2ϩ binding to ␣ L I f inhibiting 54 Mn 2ϩ binding with IC 50 values of 10 -20 and 50 -100 M, respectively. Furthermore, 1 mM Ca 2ϩ reduced 54 Mn 2ϩ binding to ␣ L I f to a level equivalent to that observed with the control proteins BSA and MBP (data not shown). These results showed clearly that Ca 2ϩ inhibited 54 Mn 2ϩ binding to ␣ L I f but not ␣ M I, and indicate that the I domain of LFA-1, but not MAC-1, has a Ca 2ϩ -binding site.
In the experiments described above, the cation binding comparison was made between ␣ L I f (i.e. the MBP-␣ L I fusion protein) and ␣ M I, the I domain of ␣ M prepared by cleavage of the GST-␣ M I fusion protein. To rule out the possibility that the differences in the cation binding activities of the two I domains was related to structural differences imposed by the nature of the protein constructs, 54 Mn 2ϩ binding to ␣ L I i , the I domain of LFA-1 prepared by thrombin cleavage of the GST-␣ L I fusion protein, was determined in the presence of increasing concentrations of unlabeled Mn 2ϩ and Ca 2ϩ . In agreement with the results shown in the previous experiment, Fig. 3B shows that 54 Mn 2ϩ binds to ␣ L I i , and that both unlabeled Mn 2ϩ and Ca 2ϩ inhibit 54 Mn 2ϩ binding (IC 50 ϭ 39 Ϯ 19 and 278 Ϯ 108 M, respectively). Furthermore, the results suggest that the cation binding activities of ␣ L I i and ␣ L I f are similar, and that the Ca 2ϩ -binding site is associated with the I domain peptide.
The hypothesis that the LFA-1, but not the MAC-1 I domain, has a Ca 2ϩ -binding site was further confirmed by direct binding studies utilizing 45 Ca 2ϩ . In the first experiment, ␣ L I f , MBP, and ␣ M I were immobilized on nitrocellulose paper and then incubated with 0.2 M 45 Ca 2ϩ . Fig. 4A shows that 45 Ca 2ϩ binds directly to ␣ L I f , but not to MPB or ␣ M I, results consistent with those in Fig. 3, and that together confirm that the 45 Ca 2ϩ binding activity is contained within the LFA-1 I domain rather than the associated MBP moiety. To further confirm these results, the experiment was repeated utilizing ␣ L I i , which lacks all MBP sequence. ␣ L I i was immobilized on nitrocellulose and incubated with 45 Ca 2ϩ in the absence or presence (Ͼ1000-fold excess) of unlabeled Ca 2ϩ . The results shown in Fig. 4B demonstrate that 45 Ca 2ϩ binds to ␣ L I i , but not to ␣ M I, and that excess unlabeled Ca 2ϩ completely inhibited 45 Ca 2ϩ binding to ␣ L I i . Together, the results of these experiments strongly suggest that the LFA-1 I domain contains a Ca 2ϩ -binding site. Furthermore, these results suggest that Ca 2ϩ binding to the I domain of LFA-1, but not MAC-1, may occur at cation concentrations found in plasma (34). and Ca 2ϩ Interaction with ␣ L I f -The affinity constant for Ca 2ϩ and Mn 2ϩ binding to ␣ L I f were determined by hot saturation binding studies and analysis of the equilibrium binding data as described below. Initial experiments showed that Ca 2ϩ and Mn 2ϩ binding to ␣ L I f immobilized on nitrocellulose was reversible, reached equilibrium in less that 1 min, and that bound 45 Ca 2ϩ and 54 Mn 2ϩ dissociated rapidly when filters were incubated in overlay buffer containing cold competitor cation (data not shown). Cation dissociation from ␣ L I f was stopped completely by immersing and washing nitrocellulose filters in 50% ethanol as described above (data not shown). Quantitation of Ca 2ϩ or Mn 2ϩ bound to the filters was performed by autoradiography and PhosphorImaging, and the data was analyzed using the EBDA/LIGAND software (Biosoft, Milltown NJ). Ca 2ϩ and Mn 2ϩ binding to ␣ L I f is shown in Fig. 5. Binding to ␣ L I was dose dependent and saturable (Fig. 5, A and B, insets), and the Scatchard plot of the binding data was curvilinear. LIGAND resolved the binding isotherm into two components corresponding to high and low affinity binding sites. The estimate for the K D of the high affinity Ca 2ϩ -and Mn 2ϩ -binding sites was 5.6 Ϯ 0.7 ϫ 10 Ϫ5 and 1.4 Ϯ 0.2 ϫ 10 Ϫ5 M, respectively. The estimate of the K D values for the low affinity Ca 2ϩ -and Mn 2ϩ -binding sites was 1-5 ϫ 10 Ϫ3 M. The maximal binding capacity of the high affinity site for Mn 2ϩ and Ca 2ϩ was 0.4 and 0.8 mol of cation bound/mol of ␣ L I f , respectively. The slope of the Hill plot of the binding data for both Ca 2ϩ and Mn 2ϩ was close to unity, indicating the apparent absence of cooperativity between the high and low affinity sites (data not shown).
We questioned whether the low affinity metal-binding site might represent Ca 2ϩ and Mn 2ϩ binding to the MBP component of the fusion protein. Therefore, the affinity of each of these cations for purified MBP, lacking any I domain se-quences, was determined as described above. Scatchard analysis of the binding data showed a single low affinity site for Ca 2ϩ and Mn 2ϩ in MBP with a K D corresponding to the low affinity site identified in ␣ L I f (data not shown). We conclude that the low affinity metal-binding site in ␣ L I f is due to Ca 2ϩ and Mn 2ϩ binding to the MBP component, whereas the higher affinity metal-binding site is associated with the I domain of the LFA-1 integrin.

DISCUSSION
The results of this study clearly show that the purified I domains of LFA-1 and MAC-1 bind cations with distinct selectivity. Binding of 54 Mn 2ϩ to ␣ L I was inhibitable by a variety of divalent and trivalent cations with a range of ionic radii. In contrast, binding to ␣ M I appeared to be more specific for smaller divalent cations. Notably, Ca 2ϩ inhibited 54 Mn 2ϩ binding to ␣ L I at concentrations well below those found in physiological environments (34), whereas Ca 2ϩ failed to inhibit 54 Mn 2ϩ binding to ␣ M I even when present at approximately 10 5 M excess. These results demonstrate a fundamental biochemical difference between these two closely related integrins, a finding with potential functional significance in the modulation of cellligand interactions.
In The binding data suggest that Ca 2ϩ and Mn 2ϩ bind to a single site within the I domain of LFA-1. 45 Ca 2ϩ and 54 Mn 2ϩ binding was inhibited by either unlabeled cation, demonstrating that these two cations bind to the LFA-1 I domain in a mutually competitive manner. Ca 2ϩ and Mn 2ϩ bind to the I domain of LFA-1 with similar affinity and stoichiometry with 1 mol of cation bound/mol of I domain indicative of a Ca 2ϩ /Mn 2ϩ mixed type binding site. We speculate that the Ca 2ϩ /Mn 2ϩbinding site in ␣ L I may represent an equivalent Ca 2ϩ /Mg 2ϩbinding site since, as we show here, Mg 2ϩ is an effective inhibitor of Mn 2ϩ binding to ␣ L I. Collectively, these results are consistent with the available structural information showing that the I domains of LFA-1 and MAC-1 contain a single Mn 2ϩ / Mg 2ϩ -binding site (19,21,29). Comparable structural data describing an I domain Ca 2ϩ -binding site has not been presented; the information herein is the first demonstration that the I domain of LFA-1 binds Ca 2ϩ as well as Mn 2ϩ and that both cations bind to a single site in a competitive manner. It should be noted that these metal binding studies were conducted using solid-phase binding techniques and it is possible that the physical interaction of the proteins with the nitrocellulose membrane could affect the metal binding characteristics. Consequently, confirmation of the reported binding constants using solution phase equilibrium binding dialyisis would be useful.
Considerable evidence supports the idea that cation-binding to integrin I domain MIDAS motifs plays a role in modulating the interaction of intact integrin with ligand. Mg 2ϩ and Mn 2ϩ induce conformational changes in integrins and both cations stimulate LFA-1-mediated cell adhesion to ICAM-1 (22)(23)(24)26,27). In contrast, Ca 2ϩ inhibits Mn 2ϩ -induced activa-tion of LFA-1, and Mg 2ϩ stimulation of LFA-1 mediated cell adhesion to ICAM-1 requires prior chelation of Ca 2ϩ by EGTA treatment (24,26,27). That the effects of divalent cations on integrin-mediated cell adhesion may be related to cation binding and cation-induced changes in I domain conformation is suggested by several observations. First, the recombinant forms of the I domains derived from a variety of integrins have intrinsic ligand binding activity. Second, ligand binding activity is cation-dependent and supported by Mn 2ϩ and Mg 2ϩ , results that reflect the activity of the these cations on the behavior of the intact integrin (12, 16 -19). Third, activationspecific conformational changes (neoepitopes) have been mapped to the I domain of MAC-1 (17,35). Finally the structural data reported by Lee et al. (20) showed ␣ M I crystals grown in Mg 2ϩ display large differences in conformation and dramatic alteration of the surface of the protein implicated in ligand binding compared with crystals grown in Mn 2ϩ . These investigators proposed that the Mg 2ϩ and Mn 2ϩ structures represent conformations of the I domain that exist in the active and inactive states of the integrin, respectively (20). In addition, it was speculated that Ca 2ϩ binding to the integrin I domain might stabilize the inactive form of the integrin (20). Together, these observations suggest that cation binding to the I domain modifies integrin interaction with ligand and raises the question whether cation-I domain complexes might exert activating or inactivating ligand-binding effects that are dependent upon the cation type.
Ca 2ϩ , Mg 2ϩ , and Mn 2ϩ are present in the extracellular environment and all are available to compete for binding to the I domain. The normal concentration of Ca 2ϩ and Mg 2ϩ in serum is approximately 1 mM (34), and it has been estimated that the concentration of Mn 2ϩ may range from 1 to 50 M depending upon the particular tissue environment (36). Assuming that the cation binding properties of the isolated recombinant I domain and the native I domain within the context of the holoprotein are similar, the high concentration of cation relative to the K d for cation binding to the I domain predicts that the metal-binding site in the I domain of LFA-1 will be occupied by cation. Furthermore, based on the evidence discussed above supporting the idea that cation-binding to integrin I domain MIDAS motifs play a role in modifying the integrin interaction with ligand, it is conceivable that I domain of low affinity LFA-1 on circulating leukocytes may be complexed with calcium, and that activation to the high affinity form may be accompanied by the replacement of Ca 2ϩ with Mg 2ϩ or Mn 2ϩ .
It seems possible that integrin activation could be driven in FIG. 5. Determination of the dissociation constant (K D ) for Mn 2؉ and Ca 2؉ interaction with ␣ L I f . 54 Mn 2ϩ (A) or 45 Ca 2ϩ (B) binding to ␣ L I f was determined by PhosphorImager analysis. The binding data was replotted (inset) by the method of Scatchard (43) and analyzed using the program LIGAND as described under "Experimental Procedures." The values reported are the results of a single experiment; this experiment was repeated three times with similar results. some environments by dynamic alteration in the ratio of cation concentration, for example, at sites of vascular or tissue injury or bone resorption (22, 36 -39). However, a more likely scenario is that cation binding to the I domain of LFA-1 is only one of multiple factors that together coordinate and regulate LFA-1 activation. Additional components of the activation process likely include conformational changes conferred by inside-out signaling mechanisms and ligand binding (40 -42). Consequently, structural changes in the integrin ectodomain mediated by inside-out signaling processes, ligand binding, and the association of Mg 2ϩ or Mn 2ϩ cation binding with the I domain, may combine to produce the activated form of LFA-1.