Structure-Function of the Putative I-domain within the Integrin β2 Subunit

The central region (residues 125–385) of the integrin β2 subunit is postulated to adopt an I-domain-like fold (the β2I-domain) and to play a critical role in ligand binding and heterodimer formation. To understand structure-function relationships of this region of β2, a homolog-scanning mutagenesis approach, which entails substitution of nonconserved hydrophilic sequences within the β2I-domain with their homologous counterparts of the β1I-domain, has been deployed. This approach is based on the premise that β1 and β2 are highly homologous, yet recognize different ligands. Altogether, 16 segments were switched to cover the predicted outer surface of the β2I-domain. When these mutant β2 subunits were transfected together with wild-type αM in human 293 cells, all 16 β2 mutants were expressed on the cell surface as heterodimers, suggesting that these 16 sequences within the β2I-domain are not critically involved in heterodimer formation between the αM and β2 subunits. Using these mutant αMβ2 receptors, we have mapped the epitopes of nine β2I-domain specific mAbs, and found that they all recognized at least two noncontiguous segments within this domain. The requisite spatial proximity among these non-linear sequences to form the mAb epitopes supports a model of an I-domain-like fold for this region. In addition, none of the mutations that abolish the epitopes of the nine function-blocking mAbs, including segment Pro192–Glu197, destroyed ligand binding of the αMβ2 receptor, suggesting that these function-blocking mAbs inhibit αMβ2 function allosterically. Given the recent reports implicating the segment equivalent to Pro192–Glu197 in ligand binding by β3 integrins, these data suggest that ligand binding by the β2 integrins occurs via a different mechanism than β3. Finally, both the conformation of the β2I-domain and C3bi binding activity of αMβ2 were dependent on a high affinity Ca2+ binding site (K d = 105 μm), which is most likely located within this region of β2.

Central to the ligand binding function of ␣ M ␤ 2 is its I(A) domain. The ␣ M I-domain is an inserted segment of ϳ200 amino acids and is highly homologous to several I-domains found in integrin ␣ subunits (9). The three-dimensional structures of several I-domains (␣ M , ␣ L , ␣ X , ␣ 2 , etc.) have been solved (10 -13). These I-domains are composed of six or seven ␣-helices and six ␤-sheets arranged in a Rossman-type fold. A cation binding site, termed the MIDAS motif, is located within the I-domain. In the MIDAS motif, cation coordination is provided by a DX-SXS sequence and by other two distant (in terms of primary sequence) oxygenated residues (10). In addition to the ␣ subunits with their I-domains, the ␤ subunits also contribute to ligand binding to integrins. Studies of the ␤ subunits have been focused primarily on their central regions (residues ϳ125-385 in a typical ␤ subunit of Ͼ700 amino acids). This region is predicted to contain a MIDAS motif, and candidate residues for cation coordination have been identified by mutagenesis (14 -17). Protein sequence analysis suggests that this region may also fold into an I-domain-like structure (10,18). However, due to the low homology between the I-domains of the ␣ and ␤ subunits, it is uncertain whether this putative I-domain region does, indeed, fold into an I-domain, or merely contains a MI-DAS motif. What is clear is that this region does play a critical role in mediating ligand binding to integrins. In ␤ 3 , it was reported that bound RGD peptides can be cross-linked to this region (19,20). Substituting this segment within the ␤ 1 I-or ␤ 5 I-domain with its homologous counterpart from ␤ 3 imparts ␤ 3 ligand specificity to the ␤ 1 or ␤ 5 integrin (21,22). A natural mutation of Arg 214 to Gln in ␤ 3 abolishes ligand binding of ␣ IIb ␤ 3 , and a synthetic peptide containing the sequence of ␤ 3 (211-222) blocks Fg binding to purified ␣ IIb ␤ 3 (23). Similar observations implicate the ␤ 1 I-domain in the ligand binding functions of the ␤ 1 integrins. For example, it was shown that both activating and inhibiting mAbs recognize a small stretch of ␤ 1 (residues 124 -160 and 207-218) (24,25). Recently, the D 134 XSXS sequence of the proposed MIDAS motif within ␤ 2 was implicated in the binding of Fg, C3bi, and ICAM-1 to ␣ M ␤ 2 (26,27). These data indicate that this putative I-domain is important to ligand binding functions of the ␤ 2 integrins as well.
Recently, we have deployed homolog-scanning mutagenesis (28) to identify several segments critical to Fg and C3bi binding within the ␣ M I-domain (8,29). This approach entails switching sequences within the ␣ M I-domain to their homologous sequences within the ␣ L I-domain. This approach is feasible because the ␣ M I-and ␣ L I-domains are highly homologous, but ␣ M ␤ 2 and ␣ L ␤ 2 recognize different ligands. In the study reported here, we have applied this same strategy to the putative ␤ 2 I-domain region. Our data are consistent with folding of the region into an I-domain-like structure. However, our results suggest that ligand recognition by the region of the ␤ 2 subunit is achieved in a distinct fashion from that involved in ligand recognition by the ␤ 3 integrins. In addition, we show that the epitopes of several blocking mAbs map to this region but their inhibitory activity is likely to be achieved via an allosteric mechanism. Finally, we show that the conformation and ligand binding functions of the ␤ 2 I-domain are enhanced selectively by Ca 2ϩ , suggesting a unique cation-specific effect on the ␤ 2 Idomain. Taken together, these results provide insight into the structure-function relationships of ␣ M ␤ 2 , which may also extend to other integrins in general.
Site-directed Mutagenesis and Development of Stable Cell Lines-The detailed procedures used for homolog-scanning mutagenesis and to establish stable cell lines expressing wild-type and mutant ␣ M ␤ 2 receptors in human kidney 293 cells have been published (30). Similar methods were used to express the ␣ M ␤ 2 heterodimer and the single ␤ 2 subunit on the surface of the Chinese hamster ovary cells. To obtain cell lines with similar expressions, each mutant cell line was subcloned by cell sorting using an ␣ M -specific mAb (2LPM19c). Up to 20 colonies were selected and analyzed for integrin expression by FACS analysis. Cells with receptor expression levels similar to wild-type ␣ M ␤ 2 were chosen, and five different subclones were used for the subsequent studies reported in this work. To exclude the possibility of subcloning artifacts, all studies were repeated using the original pool of each mutant receptor.
C3bi Binding and Adhesion to Fg-The ligand binding activity of the ␤ 2 mutants was assessed using two classic ␣ M ␤ 2 ligands, C3bi and Fg, according to our published methods (27). For adhesion of ␣ M ␤ 2 -expressing cells to Fg, the recombinant ␥-module (10 g/ml) was deposited at the center of each well in a 24-well non-tissue culture polystyrene plate. After blocking with 400 l of 0.05% polyvinylpyrrolidone in DPBS, a total of 2 ϫ 10 6 cells in Hank's balanced salt solution containing 1 mM Ca 2ϩ and 1 mM Mg 2ϩ was added to each well and incubated at 37°C for 20 min. The unbound cells were removed by three washes with DPBS, and the adherent cells were quantified by cell-associated acid phosphatase as described previously (27).
FACS Analysis-A total of 10 6 cells expressing wild-type or mutant

RESULTS
Homolog-scanning Mutagenesis of the ␤ 2 I-domain-As shown in Fig. 1, the purported I-domain within integrin ␤ 2 shares considerable sequence homology with the corresponding region of the ␤ 1 subunit. The major sequence differences are confined to regions that are predicted to be hydrophilic and surface-oriented based on hydropathy plots and molecular modeling, and, thereby, are the segments that are likely to contribute to the unique functions of the ␤ 2 integrins. For example, the ␤ 2 subunit partners with an entirely separate set of ␣ subunits from ␤ 1 , and the ␤ 2 integrins recognize a set of ligands very distinct from the ␤ 1 integrins (there is no known peptide sequence recognized by both ␤ 1 and ␤ 2 integrins). Based on the sequence homology between the ␤ 1 I-and ␤ 2 Idomains, we sought to systematically probe the function of the hydrophilic and unique segments of this region (residues 125-385) using homolog-scanning mutagenesis. Accordingly, we replaced 16 non-conserved segments of three to nine residues within the ␤ 2 I-domain with the corresponding segments from the ␤ 1 subunit (Fig. 1). These 16 segments covered the entire hydrophilic region of the ␤ 2 I-domain predicted from hydropathy plots and molecular modeling. The primers used for mutagenesis are listed in Table I. The DNA sequence of the entire I-domain was confirmed for each mutant before and after transfer back into the pCIS2M expression vector containing the cDNA of ␤ 2 .
Surface Expression and Heterodimer Formation-A large number of natural mutations occur within the ␤ 2 I-domain, which abolish surface expression and/or heterodimer formation (31)(32)(33)(34)(35)(36)(37). Nevertheless, when the ␤ 2 mutants were co-transfected with wild-type ␣ M in human kidney 293 cells, all 16 mutants were expressed on the cell surface as heterodimers and the subunits had appropriate molecular weights. As shown in Fig. 2, immunoprecipitation of surface-labeled cells with 44a, a mAb specific for the ␣ M subunit, yielded two bands of ϳ165 kDa (␣ M ) and 95 kDa (␤ 2 ) on SDS-PAGE. The patterns were similar to those obtained for wild-type ␣ M ␤ 2 (27). In addition, FACS analyses were conducted on these 16 mutants FIG. 1. Sequence alignment between the putative ␤ 1 I-and ␤ 2 Idomains. The amino acid residues are from 141 to 395 for the ␤ 1 Idomain and from 125 to 380 for the ␤ 2 I-domain (the numbering is based on the entire protein sequence including the signal peptide). The conserved residues are underlined, and the mutated segments are shown in brackets.
using a panel of ␤ 2 -specific mAbs (Table II). All 16 ␤ 2 mutants were recognized by three different mAbs to the ␤ 2 subunit MEM48, 7E4, and 6.7, as well as by the ␣ M -specific mAb 44. To exclude selection artifacts, we established at least five independent stable cell lines for each mutant ␤ 2 integrin that expressed similar levels of receptors on their cell surfaces, as judged by FACS analysis using mAb 44. Heterodimer formation, as well as other results described below, was similar for all five clones.
Epitope Mapping of Function-blocking mAbs-To help locate the functional sites within the ␤ 2 I-domain, we sought to map the epitopes of several ␤ 2 -specific function-blocking mAbs: MHM23, IB4, 6.5E, TS1/18, CLB54, YFC118.3, R3.3, H20A, 685A5, and 7E4. The ability of these mAbs to block ␤ 2 integrin functions, such as ␣ M ␤ 2 -mediated adhesion and C3bi binding and ␣ L ␤ 2 -mediated binding to ICAM-1, has been well documented (38 -42). Representative FACS analyses using mAb IB4 with five of the ␣ M ␤ 2 mutants are shown in Fig. 3A and a summary of the FACS analyses for all 16 mutants and 12 ␤ 2 -specific mAbs is shown in Table II. Among these 12 mAbs, 3 (6.7, MEM-48, and 7E4) reacted well with all 16 mutants, but not the mock-transfected 293 cells. The other nine mAbs recognized the ␤ 2 I-domain, and their epitopes consisted of at least two noncontiguous sequences. For example, mAb IB4 reacted well with wild-type ␣ M ␤ 2 , and mutants ␣ M ␤ 2 (Leu 154 -Glu 159 ), ␣ M ␤ 2 (Asn 213 -Glu 220 ) and ␣ M ␤ 2 (His 354 -Asn 358 ), but its binding to the two mutants ␣ M ␤ 2 (Arg 144 -Lys 148 ) and ␣ M ␤ 2 (Pro 192 -Glu 197 ) was ablated (Fig. 3A), suggesting that these two segments (Arg 144 -Lys 148 and Pro 192 -Glu 197 ) contribute to the epitope of IB4. As shown in Table II, in addition to IB4, mAbs MHM23, H20A, R3.3, and perhaps 6.5E also depended on segments Arg 144 -Lys 148 and Pro 192 -Glu 197 for their interactions with ␣ M ␤ 2 . mAb 685A5 required segments Arg 144 -Lys 148 , Pro 192 -Glu 197 , and Asn 213 -Glu 220 ; mAb TS1/18 required segments Leu 154 -Glu 159 and Glu 344 -Asp 348 ; mAb CLB54 required segments Leu 154 -Glu 159 and His 354 -Asn 358 ; and finally mAb YFC118.3 required segments Arg 144 -Lys 148 , Leu 154 -Glu 159 , and His 354 -Asn 358 . These epitopes can be roughly divided into two different groups (see Fig. 8). The first contains segments Leu 154 -Glu 159 , Glu 344 -Asp 348 , and His 354 -Asn 358 , and is important for ␣ M ␤ 2 interaction with mAbs TS1/18, CLB54, and YFC118.3, and the second contains segments Arg 144 -Lys 148 , Pro 192 -Glu 197 , and Asn 213 -Glu 220 , and is important for ␣ M ␤ 2 binding of mAbs MHM23, H20A, IB4, R3.3, and 685A5. To further support our epitope mapping results and this grouping of the mAbs, we performed two additional experiments. First, competition was performed between mAbs MHM23, IB4, and R3.3 from group 2, TS1/18 from group 1, and 7E4, which recognizes an epitope that is likely located outside of the ␤ 2 Idomain. In these experiments, ␣ M ␤ 2 -expressing cells were incubated first with the competitor mAb, IB4, R3.3, TS1/18, or 7E4, and then the reporter mAb MHM23 was added. Binding of MHM23 was measured by FACS analysis, and the results are shown in Fig. 3B. As predicated, mAbs IB4 and R3.3, which belong to the same group as MHM23 (group 2), blocked more than 95% of the binding of mAb MHM23 to ␣ M ␤ 2 . In contrast, mAb TS1/18 (group 1) and mAb 7E4 had little effect on MHM23 binding. The specificity of these assays was confirmed by the ability of unlabeled MHM23 but not a control IgG to block the binding of the fluorescence-labeled MHM23 to the cells. Sec-  ond, the ability of mAb IB4 to block adhesion of ␣ M ␤ 2 -expressing cells to a representative ligand, the ␥-module of fibrinogen, was assessed using wild-type and two different ␣ M ␤ 2 mutants. As shown in Fig. 3C, cells expressing these three different ␣ M ␤ 2 receptors all adhered well to the ␥-module in the presence of a control IgG. Addition of mAb IB4 completely inhibited adhesion of cells expressing the wild-type and one of the mutant receptors ␣ M ␤ 2 (Leu 154 -Glu 159 ). However, mAb IB4 had no effect on adhesion by the second mutant ␣ M ␤ 2 (Arg 144 -Lys 148 ). These results are consistent with the FACS data presented in Fig. 3A, which show that the epitope of mAb IB4 was destroyed in mutant ␣ M ␤ 2 (Arg 144 -Lys 148 ) but not in mutant ␣ M ␤ 2 (Leu 154 -Glu 159 ).
Role of the ␤ 2 I-domain in Ligand Binding-A short disulfide loop of 7-8 amino acids has been implicated in the ligand binding functions of the ␤ 3 integrins (21,22,43). This disulfide loop is conserved in the ␤ 2 subunit, corresponding to Pro 192 -Glu 197 within the putative ␤ 2 I-domain. Given the high degree of homology between the ␤ 2 and ␤ 3 subunits, we tested the hypothesis that segment Pro 192 -Glu 197 is also important to the ligand binding function of ␣ M ␤ 2 . The ␥-module of Fg and C3bi were used as model ␣ M ␤ 2 ligands, and we assessed their interactions with the mutant ␣ M ␤ 2 (Pro 192 -Glu 197 ), in which this segment was replaced with its homologous counterpart of the ␤ 1 subunit. We expected that this mutant would be defective in Fg and C3bi binding, should Pro 192 -Glu 197 constitute a part of the ligand binding site within ␣ M ␤ 2 . As shown in Fig. 4 (A and  B), this mutant bound C3bi and interacted with the ␥-module similarly to wild-type ␣ M ␤ 2 , suggesting that this sequence is not directly involved in ligand binding by ␣ M ␤ 2 . The specificity of the C3bi binding assay was confirmed using mock-transfected 293 cells and by inhibition experiments with EDTA. In addition, the specificity was further verified by blocking experiments using the ␣ M -specific mAb 44a; addition of 44a blocked more than 90% C3bi binding to both the wild-type and the mutant receptors. Similarly, the specificity of the adhesion to the ␥-module was confirmed by blocking experiments with EDTA (data not shown) and mAb 44a. Thus, the contribution of the ␤ 2 subunit to ligand binding is different from that of ␤ 3 , suggesting that ligand binding to the ␤ 2 integrins has different requirements.
The Influence of Ca 2ϩ on the Conformation of the ␤ 2 I-domain-Ligand binding to integrins depends upon divalent cations, and specific cations can influence ligand binding specificity. For example, the ␣ M I-domain adopts different conformations in the presence of Ca 2ϩ versus Mn 2ϩ (44,45), and conformational changes are induced in the ␤ 1 I-domain by Mg 2ϩ and Ca 2ϩ (46,47). In the course of our studies, we observed that binding of mAbs YFC118.3 and TS1/18 to ␣ M ␤ 2 was supported by Ca 2ϩ but not by Mg 2ϩ and that addition of EGTA/Mg 2ϩ or EDTA reduced the binding of these mAbs by 4-fold (for YFC118.3) or 5-fold (for TS1/18) (Fig. 5A). As these two mAbs recognize different and non-contiguous regions within the ␤ 2 I-domain (see Table II; TS1/18 recognizes Leu 154 -Glu 159 and Glu 344 -Asp 348 , whereas YFC118.3 recognizes Arg 144 -Lys 148 , Leu 154 -Glu 159 , and His 354 -Asn 358 ), these results suggested that the overall conformation of the ␤ 2 I-domain is differentially influenced by cations and that the conformation induced by Ca 2ϩ is required for optimal reactivity with these mAbs. To further characterize these observations, we tested the effect of Ca 2ϩ concentrations on the binding of these two mAbs. A constant concentration of each mAb of 20 nM was selected for these analyses, which is below the concentrations of each required for 50% of its maximal binding to ␣ M ␤ 2 . The Ca 2ϩ titration curve for mAb YFC118.3 is shown in Fig. 5B. Binding of the mAb increased with increasing Ca 2ϩ and saturated above 500 M added Ca 2ϩ . These data could be fitted to a single binding site model. The estimated K d of this Ca 2ϩ bind-TABLE II Reactivity of function-blocking monoclonal antibodies with the ␤ 2 I-domain mutants FACS analysis was performed using 1 g of each mAb and 10 6 ␣ M ␤ 2 -expressing cells. A "ϩ" indicates that the mean fluorescence intensity of the mAb is at least 10 times that of the IgG control. A "Ϫ" indicates that the mean fluorescence intensity of the mAb is no more than that of the IgG control. To explore the possibility that the Ca 2ϩ binding site reported by these two mAbs is located within the MIDAS motif (D 134 XSXS) of the ␤ 2 I-domain, we tested YFC118.3 binding to a mutant ␤ 2 , in which Ser 136 , a putative cation coordination site, was replaced by Ala. As shown in Fig. 5C, Ca 2ϩ bound to this mutant ␣ M ␤ 2 with a significantly (p Ͻ 0.03) reduced affinity (K d ϭ 151 Ϯ 10 M), compared with wild-type ␣ M ␤ 2 , indicating that the cation binding site reported by YFC118.3 is likely located within the ␤ 2 I-domain. To exclude the possibility that Ca 2ϩ binding to the ␣ M subunit may allosterically affect YFC118.3 and TS1/18 binding to ␤ 2 , we expressed the ␤ 2 subunit alone on the surface of the Chinese hamster ovary cells. The presence of ␤ 2 and absence of ␣ M on the cell surface was confirmed by FACS analyses using an ␣ M -specific mAb 44 and ␤ 2 -specific mAbs 6.7 (Fig. 6A), 7E4, and MEM-48 (data not shown). That the ␤ 2 subunit is expressed alone on the cell

FIG. 4. Ligand binding to ␤ 2 I-domain mutants containing the epitopes of function-blocking mAbs.
A, C3bi binding. Biotinylated EC3bi (2 ϫ 10 7 ) were added to 2 ϫ 10 5 cells expressing ␣ M ␤ 2 , which had been pre-seeded onto polylysine-coated 24-well plates. After 60 min at 37°C, the amount of bound EC3bi was determined using avidin-alkaline phosphatase and p-nitrophenylphosphate, measuring the absorbance at 405 nm. The value for wild-type ␣ M ␤ 2 was taken as 100%. Specificity was demonstrated by addition of 1 mM EDTA (shown with asterisks) and further verified by blocking experiments with an ␣ Mspecific mAb 44a; addition of mAb 44a blocked more than 90% C3bi binding to wild-type and three representative mutants: ␣ M ␤ 2 (Arg 144 -Lys 148 ), ␣ M ␤ 2 (Pro 192 -Glu 197 ), and ␣ M ␤ 2 (Leu 154 -Glu 159 ). Data are the means Ϯ S.D. of three to six independent experiments. B, Fg adhesion. Adhesion of ␣ M ␤ 2 -expressing cells to the ␥-module of Fg was performed as described in Fig. 3C except that the number of adherent cells expressing wild-type ␣ M ␤ 2 was taken as 100%. Specificity was verified using ␣ M -specific function-blocking mAb 44a (filled bar). Data are the means Ϯ S.D. of three to six independent experiments. surface is further supported by surface labeling and immunoprecipitation experiments; for the ␣ M ␤ 2 -expressing cells, both mAbs 44a (against ␣ M ) and 6.7 (against ␤ 2 ) yielded two bands of ϳ95 and 165 kDa on SDS-PAGE, whereas for the ␤ 2 -expressing cells, mAb 44a did not produce any detectable band and mAb 6.7 yielded only a single band of 95 kDa (␤ 2 ) (Fig. 6B). These data demonstrate that the ␤ 2 subunit is present alone on the cell surface and not complexed with ␣ M or any other integrin ␣ subunits. To see whether the single ␤ 2 subunit still contains a high affinity Ca 2ϩ binding site, we repeated the above Ca 2ϩ titration experiments with mAbs YFC118.3 and TS1/18. Fig. 6C shows that, similar to the ␣ M ␤ 2 heterodimer, mAb YFC118.3 bound to single ␤ 2 in a cation-dependent manner, and the Ca 2ϩ titration curve can be fitted to a single binding site model. The estimated K d of this Ca 2ϩ binding site is 83 Ϯ 2 M, which is very close to the K d of 105 M for the heterodimeric receptor. A similar K d for Ca 2ϩ binding to ␤ 2 was obtained with TS1/18. Taking these data together, we conclude that the proper conformation for mAb binding to the ␤ 2 I-domain depends upon a Ca 2ϩ binding site within ␤ 2 , possibly composed of Ser 136 within the proposed MIDAS motif of the ␤ 2 I-domain.
Several studies have reported that ligand binding by the ␣ M I-, ␣ L I-, ␣ 1 I-, ␣ 2 I-, ␤ 1 I-, and ␤ 3 I-domains is supported by Mg 2ϩ but not Ca 2ϩ (47)(48)(49)(50). In fact, Ca 2ϩ can inhibit ligand binding to several of these integrins. As our data suggest that the ␤ 2 I-domain contains a unique high affinity Ca 2ϩ binding site, we next tested the effects of Mg 2ϩ and Ca 2ϩ on C3bi binding by ␣ M ␤ 2 . As shown in Fig. 7, C3bi binding is supported by 1 mM Ca 2ϩ . This interaction can be further increased by addition of Mg 2ϩ . However, ␣ M ␤ 2 only exhibited minimal ligand binding in the presence of Mg 2ϩ alone (EGTA was added to exclude possible contributions by Ca 2ϩ ). As expected, addi-tion of 1 mM EDTA completely abolished C3bi binding to ␣ M ␤ 2 , confirming the cation dependence of the C3bi/␣ M ␤ 2 interaction. Thus, in the case of ␣ M ␤ 2 , Ca 2ϩ is not inhibitory but is required for ligand binding. DISCUSSION In this work, we have probed the function of the hydrophilic surface of the putative ␤ 2 I-domain (residues 125-385), using a homolog-scanning mutagenesis approach. Our major findings are as follows. 1) The majority of the hydrophilic surface of the ␤ 2 I-domain is not critically involved in heterodimer formation between the ␣ M and ␤ 2 subunits. 2) Although the epitopes of several function-blocking mAbs map to the putative ␤ 2 I-domain, these epitopes are not involved directly in ligand binding to ␣ M ␤ 2 .
3) The positioning of these epitopes is consistent with an I-domain-like fold for this region of the ␤ 2 subunit, as proposed by several investigators (10,(15)(16)(17)(18)42). 4) Of particular note, segment Pro 192 -Glu 197 , which has been implicated in direct ligand contact by the ␤ 3 integrins (21,22,43), is not critical to ligand binding by ␣ M ␤ 2 , suggesting a fundamental difference between the ligand binding mechanism by ␤ 2 versus ␤ 3 . 5) The optimal conformation of the ␤ 2 I-domain for C3bi binding depends on a functional Ca 2ϩ binding site within the ␤ 2 subunit.
V 275 GSDNH between human and avian ␤ 3 was found to change the specificity of ␣ IIb /␤ 3 association (51). Taken together, these data strongly implicate this central region of the ␤ subunits in either heterodimer formation or in controlling the pairing specificity between the ␣ and ␤ subunits. None of the known ␣ subunits complex with both ␤ 1 and ␤ 2 , and thus we expected heterodimer formation would be perturbed in some of our homolog-scanning mutants, particularly the one involving seg-ment Asp 290 -Glu 298 , which is homologous to V 275 GSDNH of ␤ 3 (51). Nevertheless, when all 16 nonconserved segments within the ␤ 2 I-domain, including segment Asp 290 -Glu 298 , were replaced with their counterpart sequences within the ␤ 1 I-domain, surface expression and heterodimer formation were not affected, as assessed by surface labeling and immunoprecipitation experiments. Therefore, we conclude that the majority of the hydrophilic residues of the ␤ 2 I-domain do not make a significant contribution to the heterodimer formation and specificity pairing of the ␣ M and ␤ 2 subunits. As most of the hydrophobic residues are identical between ␤ 1 I-and ␤ 2 I-domains, it is possible that the remaining few non-identical hydrophobic residues within the ␤ 2 I-domain, most of which have been mutated and found not critical in this study (see Fig. 1), are responsible for the specific paring between ␣ M and ␤ 2 . Alternatively, as the C-terminal cysteine-rich region is not involved in heterodimer formation (52,53), the N-terminal plexin-homologous region is a likely candidate for determining the specificity of the heterodimer formation. It should be noted that most of the naturally occurring point mutations that prevent cell surface expression occur at conserved sites in the ␤ subunit, and these could affect the overall fold of the ␤I-domains, leading to intracellular degradation (the exception to this is Lys 196 , which is not conserved, but when we substituted a Glu at this position, surface expression also was not affected).
It was proposed recently that the central region within the ␤ subunits (residues 125-385 for ␤ 2 ) folds into an I-domain-like structure, similar to that present in several integrin ␣ subunits (10,18). However, homology between the I-domains of the ␣ and ␤ subunits is very low, particularly in the C-terminal portions, and conflicting views exist in the literature as to whether this region assumes an I-domain fold or merely contains a metal binding MIDAS motif (DXSXS), such as that found in I-domains (14 -17). Using the 16 homolog-scanning mutants of the ␤ 2 I-domain, we have mapped the epitopes of nine mAbs (MHM23, IB4, 6.5E, TS1/18, CLB54, YFC118.3, R3.3, H20A, and 685A5). All of these mAbs reactive with the putative ␤ 2 I-domain recognized epitopes that are composed of at least two non-contiguous sequences. For discussion purposes, these epitopes can be divided into two groups (Fig. 8 . The spatial relationship of these segments is consistent with the I-domain fold such that protein folding will bring the distal segments Leu 154 -Glu 159 , Glu 344 -Asp 348 , and His 354 -Asn 358 (group 1), or Arg 144 -Lys 148 , Pro 192 -Glu 197 , and Asn 213 -Glu 220 (group 2) together into spatial proximity to form the overlapping mAb epitopes. Our mapping results are consistent with a very recent study, in which the epitopes of some mAbs from group 1 were mapped and used to construct a threedimensional model for the ␤ 2 I-domain (42). Although our results support this model, there is one major difference. The identification of mAbs (MHM23, H20A, IB4, R3.3, and 685A5) belonging to group 2 allows us to define more accurately the position of the disulfide loop, C 191 PNKEKEC 198 , within the three-dimensional framework. The positioning of this loop is particularly important, given the recent report implicating the segments equivalent to Pro 192 -Glu 197 within this loop in ligand binding to ␤ 3 and ␤ 5 (22,43). As shown in Table II Recent studies from several laboratories have implicated segment 179 -183 of ␤ 3 , which is homologous to Pro 192 -Glu 197 of ␤ 2 , in ligand binding (21,22,43). However, the ␣ M ␤ 2 (Pro 192 -Glu 197 ) mutant interacted well with C3bi and the ␥-module of Fg, similar to the wild-type receptor, suggesting that this segment is not involved directly in ligand contact within the ␤ 2 integrins. Thus, there appears to be a fundamental difference between the ligand binding requirements of ␤ 3 and that of ␤ 2 . Of note, the integrins ␣ IIb ␤ 3 and ␣ V ␤ 3 that utilize this sequence in ligand binding lack I-domains within their ␣ subunits. Therefore, integrins with or without I-domains in their ␣ subunits may employ different mechanisms for ligand binding. Support for this hypothesis also can be derived from recent findings showing that the W2 and W3 repeats within the ␤-propeller of the ␣ subunits are located in close proximity to the sequence corresponding to Pro 192 -Glu 197 of ␤ 2 within their ␤ subunits, and together contribute to formation of the ligand binding site (43). Since the I-domains within the ␣ subunits are predicted to insert between the W2 and W3 repeats (54), this geometry would be altered and not be available for ligand binding to the ␤ 2 integrins.
In this study, we mapped the epitopes of nine ␤ 2 -blocking mAbs to specific regions within the ␤ 2 I-domain. The epitopes were restricted to six segments (Arg 144 -Lys 148 , Leu 154 -Glu 159 , Pro 192 -Glu 197 , Glu 344 -Asp 348 , Asn 213 -Glu 220 , and His 354 -Asn 358 ). To determine whether these segments also are involved in ligand binding, we examine their binding of C3bi and Fg, two classic ligands of ␣ M ␤ 2 . All six ␤ 2 mutants interacted well with C3bi and the ␥-module of Fg, in a manner similarly to wild-type ␣ M ␤ 2 , except mutant ␣ M ␤ 2 (Arg 144 -Lys 148 ), which exhibited 50% adhesive activity of the wild-type receptor. These data suggest that none of these segments is critically involved in ligand binding of ␣ M ␤ 2 . Therefore, it is very likely that these ␤ 2 I-domain specific mAbs, like the ␤ 1 -specific function-blocking mAb described by Mold et al. (55), inhibit receptor functions allosterically. We cannot exclude the possibility that some of these mAbs sterically hinder ligand binding. However, several activating mAbs of the ␤ 1 integrins map to the homologous region within the ␤ 1 I-domain (25), suggesting that this region is conformationally flexible, consistent with an allosteric mechanism. In further support of this model, we found that mutant ␣ M ␤ 2 (Asn 213 -Glu 220 ), which interacted more avidly with both C3bi and the ␥-module (Fig. 4, A and B), exhibited an active conformation, judged by its reactivity toward an activation-dependent mAb 24. 2 This mAb has been used in a number of studies to probe the activated state of several ␤ 2 integrin receptors (49,56,57). Investigation of the underlying mechanism of activation is currently under way.
It has been well established that integrin-ligand interactions are cation-dependent, but the nature and location of these cation-binding sites are currently unclear. Recently, a novel cation binding site, termed the MIDAS motif, was identified in the crystal structures within the I-domains of several ␣ subunits and was found to be central to ligand binding functions of these I-domains (10, 58 -61). Evidence for the existence of MIDAS motifs in the I-domains of ␤ 1 , ␤ 2 , and ␤ 3 has also been developed (14,16,17,47,51,62,64). Although the I-domains of the ␣ and ␤ subunits are predicted to have similar MIDAS folds, their cation binding properties appear to differ significantly. Using several different approaches including x-ray crystallography, circular dichroism, and fluorescence, it appears that cation binding to the I-domains of the ␣ subunits and the ␤ 1 , ␤ 3 , and ␤ 5 subunits can lead to changes in conformation and ligand binding activity (14,44,45,47,50,63). Compared with the ␤ 1 and ␤ 3 subunits, the role of cation binding in controlling the conformation of the ␤ 2 subunit is not well understood. In this study, we report that the binding of two mAbs (TS1/18 and YFC118.3) recognizing non-contiguous regions within the ␤ 2 Idomain depend on Ca 2ϩ for optimal recognition of ␣ M ␤ 2 (Fig. 5). A single Ca 2ϩ binding site with a K d value of ϳ105 M was estimated for both mAbs. This K d value is very similar to that determined for Ca 2ϩ binding to the ␣ L I-domain (50 M) (45) and those obtained for Mg 2ϩ binding to the I-domains of ␣ 1 , ␣ 2 , ␤ 1 , and ␤ 5 (80 -100 M) (14,47,50), suggesting that the cation binding site that controls the conformation of the ␤ 2 I-domain is most likely located within the ␤ 2 I-domain itself. To test this hypothesis, we evaluated Ca 2ϩ binding activity of ␣ M ␤ 2 (S136A), in which the predicted coordinating residue within the MIDAS motif was changed. Using mAb YFC118.3 and FACS analysis, we found that the Ca 2ϩ binding affinity obtained for this mutant ␤ 2 was significantly lower than that of wild-type ␤ 2 (151 Ϯ 10 M for the mutant versus 105 Ϯ 9 M for wild-type ␤ 2 , p Ͻ 0.03) (Fig. 5C). Our results are in agreement with the studies of Lin et al. (14), showing that mutations of the 2 Y. Xiong and L. Zhang, unpublished observation.  (42). The model is further modified based on the epitope mapping data in Table II using  residues within the MIDAS motif of the ␤ 5 I-domain changed the apparent affinity of Mg 2ϩ for ␣ v ␤ 5 from 80 -180 M to 125-300 M. To exclude the possibility that Ca 2ϩ could affect YFC118.3 and TS1/18 binding to the ␤ 2 I-domain allosterically by binding to ␣ M (via the Ca 2ϩ binding site within either the ␣ M I-domain or the ␤-propeller), we expressed single ␤ 2 on the cell surface. We found that the ␤ 2 subunit alone, in the absence of ␣ M or any other ␣ subunits, still possessed a high affinity Ca 2ϩ binding site, which is required for optimal binding of mAbs YFC118 and TS1/18 to the ␤ 2 I-domain (Fig. 6). The calculated K d is 83 M, which is very close to that of the ␣ M ␤ 2 heterodimer (105 M). These data strongly suggest that the Ca 2ϩ binding site that promotes YFC118.3 and TS1/18 binding to the ␤ 2 I-domain is located within the ␤ 2 subunit, possibly composed of Ser 136 of the MIDAS motif. However, since mutation of Ser 136 did not completely abolish Ca 2ϩ binding, residues outside the MIDAS motif may also be involved in Ca 2ϩ coordination.
Given the specificity of the ␤ 2 I-domain for Ca 2ϩ , we next tested whether this Ca 2ϩ binding site plays a role in ligand binding by ␣ M ␤ 2 , and found that C3bi binding to ␣ M ␤ 2 was supported more effectively by Ca 2ϩ than Mg 2ϩ (Fig. 7). In light of the report that Ca 2ϩ does not support C3bi binding to the recombinant ␣ M I-domain (48), the Ca 2ϩ binding site that supports C3bi binding of ␣ M ␤ 2 is likely located within the ␤ subunit, most probably in the ␤ 2 I-domain. A similar cation-binding site was reported in the ␤ 1 subunit that modulates both ligand binding and mAb 12G10 recognition by integrin ␣ 5 ␤ 1 (46,65). This mAb (12G10) recognizes an epitope (Val 211 -Met 287 ) within the ␤ 1 I-domain similar to that of TS1/18 and YFC118.3, and its binding depends on a single high affinity cation binding site with a K d of 70 M for Ca 2ϩ (46). Mold et al. (46) proposed that divalent cations induced conformational changes within the ␤ 1 I-domain, leading to an unmasking of the ligand binding site within ␣ 5 ␤ 1 . Given the similarity between the Ca 2ϩ binding sites within the ␤ 1 I-and ␤ 2 I-domains, it is very possible that the same mechanism is involved in the modulation of ␣ M ␤ 2 function by Ca 2ϩ .
In summary, using homolog-scanning mutagenesis, we have systematically probed the hydrophilic surface of the ␤ 2 I-domain. Our data suggest that the majority of the hydrophilic regions of the ␤ 2 I-domain are not critically involved in the specific association of ␤ 2 with ␣ M . Additionally, we have mapped the epitopes of nine ␤ 2 -specific mAbs into two separate groups within the ␤ 2 I-domain and showed that the spatial arrangement of the residues that constitute these mAb epitopes is consistent with an I-domain-like fold in this region. Most importantly, our data strongly demonstrate that the ligand binding site within ␤ 2 is distinct when compared with that of ␤ 3 . This fact leads us to hypothesize that integrins containing I-domains in their ␣ subunits may utilize different regions of the ␤I-domains for ligand recognition than the integrins lacking I-domains in their ␣ subunits. In addition, our C3bi binding and Fg adhesion data showed that the epitopes of the nine ␤ 2 I-domain specific function-blocking mAbs are not critically involved in ligand binding, implying that they block ␣ M ␤ 2 functions by allosteric mechanisms. Finally, we have demonstrated that both the conformation of the ␤ 2 I-domain and C3bi binding to ␣ M ␤ 2 depend on a functional Ca 2ϩ binding site, which is located within the ␤ 2 subunit and probably in the ␤ 2 I-domain. As C3bi binding to the ␣ M I-domain is supported by Mg 2ϩ , but not Ca 2ϩ (48), our data suggest a role for the Ca 2ϩ binding site within the ␤ 2 I-domain in C3bi-␣ M ␤ 2 interactions. Given the high degree of homology between all integrin ␤ subunits, these conclusions should extend to other integrins as well.