Distinct recognition of complement iC3b by integrins α_Xβ₂ and α_Mβ₂

Significance

Complement is deposited on foreign antigens and particles and marks them for recognition by immune cells and removal by phagocytes. Immune cells have two homologous integrins, α_Xβ₂ and α_Mβ₂, that recognize the complement fragment iC3b. Although it might have been suspected that these integrins would bind to the same site on iC3b, direct comparisons here show they do not. Using negative stain electron microscopy and purified integrin fragments and iC3b, we show that α_Xβ₂ binds to two distinct sites on iC3b. A single α_Mβ₂ integrin binds to two different sites on iC3b, each of which is distinct from those to which α_Xβ₂ binds. Our findings reveal remarkable diversity among integrins that recognize complement and suggest possible cooperative responses.

Abstract

Recognition by the leukocyte integrins α_Xβ₂ and α_Mβ₂ of complement iC3b-opsonized targets is essential for effector functions including phagocytosis. The integrin-binding sites on iC3b remain incompletely characterized. Here, we describe negative-stain electron microscopy and biochemical studies of α_Xβ₂ and α_Mβ₂ in complex with iC3b. Despite high homology, the two integrins bind iC3b at multiple distinct sites. α_Xβ₂ uses the α_X αI domain to bind iC3b on its C3c moiety at one of two sites: a major site at the interface between macroglobulin (MG) 3 and MG4 domains, and a less frequently used site near the C345C domain. In contrast, α_Mβ₂ uses its αI domain to bind iC3b at the thioester domain and simultaneously interacts through a region near the α_M β-propeller and β₂ βI domain with a region of the C3c moiety near the C345C domain. Remarkably, there is no overlap between the primary binding site of α_Xβ₂ and the binding site of α_Mβ₂ on iC3b. Distinctive binding sites on iC3b by integrins α_Xβ₂ and α_Mβ₂ may be biologically beneficial for leukocytes to more efficiently capture opsonized pathogens and to avoid subversion by pathogen factors.

Activation of mammalian complement is critical for the clearance of pathogens and altered host cells, whereas excessive activation results in tissue damage (1). Complement activation can be initiated by three distinct pathways with proteolytic cleavage of complement component C3 as the pivotal step in each pathway. C3 is cleaved to C3b by C3 convertases. C3b participates in some C3 convertases to amplify C3 cleavage; however, further cleavage of C3b to iC3b inactivates convertase activity. In turn, iC3b can be digested to yield C3c plus C3dg, and C3dg can be further cleaved to C3d (2) (Fig. 1A). Conversion of C3 to C3b exposes the otherwise buried reactive thioester bond in the thioester domain (TED) and enables it to covalently attach to hydroxyl and amino groups on pathogenic, immunogenic, and apoptotic cell surfaces (3). Opsonic fragments of C3 covalently bound through the TED, i.e., C3b, iC3b, and C3dg/C3d serve as ligands for five distinct complement receptors (CRs). Each CR has been characterized for its preference for specific C3 opsonic fragments. CR type 3 (CR3, also known as CD11b/CD18, Mac-1, or integrin α_Mβ₂) and CR type 4 (CR4, also known as CD11c/CD18, p150, 95, or integrin α_Xβ₂) specifically recognize iC3b as shown by selective rosette formation with erythrocytes opsonized with iC3b but not C3b or C3d (4–6), and specific isolation from cells by affinity chromatography on iC3b-Sepharose beads (7). However, at high concentrations on opsonized cells, C3d and C3dg can show reactivity with CR3 (8, 9).

Fig. 1.

Integrins, iC3b, and negative-stain EM of α_X I domain and iC3b complexes. (A) Schematic of iC3b. (B) Schematic of domain organization and conformational states of αI-integrins α_Xβ₂ and α_Mβ₂. (C) Superdex 200 size exclusion chromatography profiles of iC3b in presence or absence of the α_X αI F273S/F300A mutant in presence of 2 mM Mg²⁺. (D) Better resolved class averages of α_X αI and iC3b complexes. (Scale bar: 10 nm.) Schematic interpretations of classes are shown to the right.

α_Xβ₂ and α_Mβ₂ are heterodimeric proteins belonging to the β₂ integrin subfamily. They are expressed on myeloid cells including neutrophil granulocytes, monocytes, macrophages, and also on activated lymphocytes and lymphoid natural killer cells (10–12). α_Xβ₂ and α_Mβ₂ are essential for recognition and phagocytosis of pathogens and immune complexes in vivo (13). Deficiency of α_Xβ₂ and α_Mβ₂ in leukocyte adhesion deficiency results in recurring bacterial infections (14) and deficiency of α_Mβ₂ alters susceptibility to injury from immune complexes (15). The β₂-associated α subunits all contain an inserted domain (αI domain), which adopts a Rossmann fold with a metal ion-dependent adhesion site (MIDAS) at the ligand-binding site (Fig. 1B) (16).

The αI domain of α_Xβ₂ and α_Mβ₂ plays a key role in recognition of iC3b by both integrins; however, alternative recognition modes have also been suggested. Mutagenesis and antibody blocking studies on α_Xβ₂ and α_Mβ₂ showed that the αI domain was required for binding to iC3b (6, 17, 18). More recently, electron microscopy (EM) studies with the α_Xβ₂ ectodomain revealed binding solely through its αI domain to a site at the interface between the macroglobulin (MG) 3 and MG4 domains on the C3c moiety of iC3b (Fig. 1A) (19). However, stoichiometry suggests multiple binding sites for the α_X αI domain in iC3b (20). Because of the high similarity between α_Xβ₂ and α_Mβ₂, it was expected that α_Mβ₂ would bind to the same site on iC3b as α_Xβ₂. It was surprising then when a crystal structure of a complex between the α_M αI domain and TED (C3d) revealed a specific complex between them (21). As the authors pointed out, binding of the α_M αI domain alone to TED could not explain the specificity of α_Mβ₂ for binding to iC3b and not to C3d (21). Moreover, the α_M subunit β-propeller domain and the β₂ subunit inserted domain (βI) (Fig. 1B) have been suggested to additionally contribute to iC3b binding (22–24). Both α_Mβ₂ and α_Xβ₂ recognize a variety of unrelated ligands and proteolyzed or denatured proteins (25); given this broad specificity, α_Xβ₂ has been termed a “danger receptor” (26). Furthermore, the isolated α_M αI domain has been crystallized in the open conformation bound to a neighboring αI domain in a ligand-mimetic lattice contact (27). Thus, it was interesting to determine whether the α_M αI domain complex with the TED domain in crystals (21) could be extended to larger integrin and complement fragments and whether additional contacts might be found. These findings also suggested the importance of further comparisons between α_Mβ₂ and α_Xβ₂, particularly with integrin fragments complementary to those previously used in structural studies, i.e., the ectodomain in the case of α_Xβ₂ (19) and the αI domain in the case of α_Mβ₂ (21).

Using negative-stain EM with the α_X αI domain and the α_Mβ₂ headpiece here, we have shown that α_Xβ₂ and α_Mβ₂ bind to different regions of iC3b. Distinct recognition modes toward iC3b by two closely related β₂ integrins reveal surprising diversity that might be important for complement recognition in the context of the quite different surfaces or antigens on which complement can be deposited and enable cooperative rather than competitive functions for α_Xβ₂ and α_Mβ₂.

Results

Interaction of α_Xβ₂ with iC3b.

To stabilize the high-affinity state of the isolated α_X αI domain (α_X residues E130-G319), we introduced mutations F273S and F300A. Homologous mutations F265S and F292A had been found by random mutagenesis followed by directed evolution to increase the affinity of the α_L αI domain and stabilize its open conformation (28). The corresponding residues in α_X have identical structural environments, which suggested that their mutation should similarly stabilize the open, high-affinity conformation. Indeed, the F273S/F300A α_X αI domain, used in all studies below, formed a complex with iC3b that was stable to gel filtration in buffer containing 2 mM Mg²⁺ and eluted earlier than iC3b alone (Fig. 1C). In contrast, the wild-type α_Xβ₂ ectodomain required not only the activating metal ion Mn²⁺, but also Fabs that stabilize the open integrin headpiece conformation (Fig. 1B), to form complexes with iC3b that were stable to gel filtration (19).

The isolated α_X αI complex with iC3b was subjected to negative-stain EM with multivariate K-means classification and multireference alignment of >5,000 particles into 50 class averages (Fig. S1). Class averages were visualized corresponding to both the α_X αI domain bound to iC3b (Fig. 1D, 1–4) and iC3b alone (Fig. 1D, 5–8). The α_X αI domain bound to the key ring moiety formed by MG domains 1–8 of iC3b (Fig. 1 A and D). The key ring moiety can be oriented both (i) by its weaker density on the side with only MG3 and MG4 compared with the thicker density on the side where MG1 and MG2 stack atop MG5 and MG6 and (ii) by the slant of the C345C knob toward the thick side (Fig. 1A). This orientation confirmed binding of the αI domain to MG3 and MG4 near their interface (Fig. 1D, 1–4 and schematics).

Fig. S1.

Complete set of class averages of α_X αI and iC3b complex (6,202 particles). Classes are ranked from most to least populous from left to right and then from top to bottom.

In conversion from C3b to iC3b, factor I cleaves the CUB domain in its C-terminal linkage to TED. Cleavage leaves TED tethered to the C3c moiety through its N-terminal linkage. In agreement with previous results (29), we found that the TED is flexibly tethered to the C3c moiety in iC3b, as shown by its appearance in different orientations and distances relative to the C3c moiety (Fig. 1D, 1, 2, 5, and 6) and its lack of defined orientation (absence of density) in other class averages (Fig. 1D, 3, 4, 7, and 8). The TED (296 residues) was clearly distinguished by its stronger density from the smaller αI domain (190 residues) (Fig. 1D).

Interaction of α_Xβ₂ with C3c.

Further cleavage by factor I of iC3b in the CUB domain separates TED (C3dg) from C3c. We prepared α_X αI domain complexes with C3c in Mg²⁺ by gel filtration (Fig. 2A) and characterized them by EM (Fig. 2B and Fig. S2A) using the same methods as described above for iC3b. Although some particles showed C3c alone (Fig. 2B, 1 and 2), most (30 classes, corresponding to 68% of particles) showed αI bound near the MG3-MG4 interface on the MG key ring (Fig. 2B, 3–8). Strikingly, at least 8 of 50 class averages showed an additional density for αI near the C345C knob (Fig. 2B, 5–8).

Fig. 2.

Negative-stain EM of α_X αI domain complexes with C3c and comparison with α_Xβ₂ ectodomain complexes with C3c and iC3b. (A) Superdex 200 size exclusion chromatography profiles of C3c, the α_X αI F273S/F300A mutant, or their mixture in presence of 2 mM Mg²⁺. (B) Better resolved class averages of α_X αI and C3c complexes. Schematic interpretations of the averages are shown to the right. (C–E) Complexes (2:1) containing two integrin α_Xβ₂ ectodomains bound to C3c (C) or iC3b (D) or for comparison 1:1 α_Xβ₂:C3c class averages (E) (19). The 2:1 complexes were identified among published 1:1 complexes (19) as described in Results. To confirm the position of the C3c or the C3c moiety of iC3b in the class averages (Top), class averages were masked (Middle) and cross-correlated with a C3c crystal structure (Bottom) with cross-correlation score and C3c ribbon cartoon shown below. (Scale bars: 10 nm.) (F) Schematic of 2:1 complexes. Three Fabs bound to α_Xβ₂ to stabilize its extended-open conformation are shown. E reproduced from ref. 19.

Fig. S2.

EM of α_X αI and α_Xβ₂ ectodomain complexes with C3c. (A) Complete set of class averages of α_X αI and C3c complex (5,042 particles) as described in Fig. S1. (B) An EM field containing α_Xβ₂-three Fabs–C3c complexes. Two α_Xβ₂-C3c (2:1) complex-like particles are circled in red.

In five class averages (9.5% of particles), the second αI domain appeared on the thick side of the MG key ring (Fig. 2B, 5 and 6), whereas in others (3.3% of particles), it appeared on the opposite side of the C345C knob nearer the thin side of the MG key ring (Fig. 2B, 7 and 8). The differing orientations might reflect binding to the flexible remnants of the cleaved CUB domain, which localize near the C345C knob, and/or may reflect a 3D orientation of αI out of the C3c plane visualized in Fig. 2B and the tendency of particles to lie flat on EM substrates.

Having found two binding sites for the isolated α_X αI domain in the C3c moiety, we then asked whether two binding sites could also be found with the intact α_Xβ₂ ectodomain. We reexamined grids from a published study (19) on C3c complexes formed with the α_Xβ₂ ectodomain stabilized in the high-affinity state with three Fab fragments. We found in the same fields from which 4,387 1:1 α_Xβ₂:C3c particles had been picked 121 putative 2:1 particles that were classified into two classes (Fig. 2C and Fig. S2B). Indeed, both class averages revealed 2:1 α_Xβ₂ ectodomain-Fab:C3c complexes. The orientations were similar in the two classes (Fig. 2C, diagrammed in Fig. 2F), although density was clearest in the most populous average (Fig. 2C, 1). Our domain assignments were confirmed by masking all segments except C3c, and cross-correlation with the C3c crystal structure (Fig. 2C, Middle and Bottom). The cross-correlation scores were excellent and were comparable to those using the C3c moiety from 1:1 complexes (Fig. 2E, Middle and Bottom). The orientation of the two integrins in the class averages is readily assigned based on densities for the head containing the αI, β-propeller, and βI domains, and weaker densities for the α- and β-subunit legs. The two integrins orient with their β-subunits facing toward one another as shown by the stronger density of the large β-propeller domain, which marks the α-subunit side of the head, and m24 Fab, which binds to the βI domain and marks the β-subunit side of the head (Fig. 2 C and F). Importantly, these landmarks, and the densities for the two αI domains, confirm that the αI domain in one integrin binds near the MG3-MG4 interface in the MG key ring and that the αI domain in the other integrin binds near the C345C knob, exactly as described above for isolated α_X αI domains.

We then asked whether 2:1 complexes might be present within α_Xβ₂ complexes with iC3b (19). In the same fields from which 5,134 1:1 complexes had been subjected to class averaging (19), we found 161 putative 2:1 complexes that were subjected to classification and averaging into two classes (Fig. 2D). The most populous class average showed the same two αI domain sites on iC3b as seen with C3c, with one binding site near the MG3-MG4 interface and the other near the C345C knob (Fig. 2D, 1). The second class average was less clear, although density for m24 Fab bound to each integrin suggested a similar integrin orientation as in the other class average.

Recognition of iC3b but Not C3c by α_Mβ₂.

A headpiece fragment of α_Mβ₂ in 1 mM Mn²⁺ and 0.2 mM Ca²⁺ formed complexes with iC3b that were stable to gel filtration as shown by elution of the complex earlier than iC3b alone or the α_Mβ₂ headpiece monomer (Fig. 3 A and C). EM class averages showed that the α_Mβ₂ headpiece bound through its αI domain to the iC3b TED (Fig. 3D and Fig. S3A). The position of the TED with respect to the C3c moiety of iC3b varied much less in complexes with α_Mβ₂ than in complexes with the α_X αI domain or in iC3b alone (Fig. 1D) (19, 29). Furthermore, the C3c moiety adopted similar, nonrandom orientations with respect to the integrin, with its C345C knob close to the β-propeller and βI domains in the integrin head (Fig. 3D). In some class averages, the knob approached the head closely (Fig. 3D, 5 and 6). The nonrandom orientation of the C3c moiety and close approach of its knob to the integrin strongly suggest a second three-dimensional contact between the integrin and iC3b that is less stable than the αI-TED contact and is susceptible to disruption when complexes adsorb to the grid and become largely planar.

Fig. 3.

Negative-stain EM of α_Mβ₂ headpiece complexes. (A–C) Superdex 200 size exclusion chromatography profiles of complexes and individual components in Mn²⁺ with iC3b (A), C3c (B), or the α_Mβ₂ headpiece alone in Mg²⁺ or Mn²⁺ (C). (D and E) Better resolved class averages of the α_Mβ₂ headpiece complex with iC3b (D) and α_Mβ₂ headpiece homodimer complexes in Mn²⁺ (E). To figure out whether the α_Mβ₂ headpiece is in open or closed conformation, class averages were masked (row 2) and cross-correlated with a modeled open form α_xβ₂ headpiece (row 3) or a crystal structure of α_Xβ₂ headpiece in closed conformation (PDB ID code 4NEH) (row 4). The cross-correlation scores are labeled. The schematic interpretations of the class averages are shown below. (Scale bar: 10 nm.)

Fig. S3.

EM of high molecular weight, putative complex fractions from gel filtration in Mn²⁺. Complete sets of class averages are shown for α_Mβ₂ headpiece with iC3b (A, 4,883 particles), α_Mβ₂ headpiece with C3c (B, 4,206 particles), and α_Mβ₂ headpiece alone (C, 2,842 particles). Classes are ranked by population as described in Fig. S1.

In contrast to iC3b, C3c failed to complex with the α_Mβ₂ headpiece (Fig. 3B). An early eluting peak was present in samples containing C3c and α_Mβ₂ or α_Mβ₂ alone, but not in C3c alone (Fig. 3B). This peak contained no α_Mβ₂–C3c complex but did contain an α_Mβ₂ headpiece dimer (Fig. S3B) as described in Dimerization of the α_Mβ₂ Headpiece.

Ligand binding by other αI domain-containing integrins has been associated with headpiece opening, as visualized in EM by swing-out of the hybrid domain (Fig. 1B). To determine whether the headpiece was open or closed in α_Mβ₂ bound to iC3b, we cross-correlated it to crystal structures or models of the closed and open α_Xβ₂ headpiece. Better correlation to the open headpiece revealed that the α_Mβ₂ headpiece is open when bound to iC3b (Fig. 3D, Bottom). Moreover, cross-correlation showed that the headpiece oriented with its β-subunit adjacent to the C3c moiety in all class averages. This nonrandom orientation of the α_Mβ₂ headpiece provided further evidence for a specific interaction of its β-propeller and βI domain region with a C345C knob-proximal region of the C3c moiety of iC3b.

Dimerization of the α_Mβ₂ Headpiece.

When incubated alone in buffer with 1 mM Mn²⁺ and 0.2 mM Ca²⁺, the α_Mβ₂ headpiece readily formed a dimer, as shown both by gel filtration (Fig. 3 A–C) and EM (Fig. 3E and Fig. S3 B and C). Dimer formation was not seen in buffer with 1 mM Mg²⁺ and 1 mM Ca²⁺ (Fig. 3C). When the α_Mβ₂ headpiece and iC3b were mixed in buffer with 1 mM Mn²⁺ and 0.2 mM Ca²⁺, α_Mβ₂ complex formation with iC3b was favored over α_Mβ₂ dimer formation, with only 1 of 50 class averages showing the dimer (Fig. S3A). Dimer formation appears to reflect the ability of α_Mβ₂ to promiscuously recognize a wide range of ligands (Introduction). α_Mβ₂ headpiece dimers displayed dyad symmetry, with the αI domain of one monomer bound to the distal end of the β-leg, i.e., to the I-EGF1 or PSI domain, of the other monomer (Fig. 3E, see diagram below). Furthermore, cross-correlation showed that the headpiece was open when one α_Mβ₂ monomer bound to another monomer (Fig. 3E).

Discussion

Recognition of the complement degradation product iC3b by leukocyte integrins α_Xβ₂ and α_Mβ₂ is critical for phagocytosis of opsonized foreign particles in host defense. The α_X and α_M subunits are 60% identical in sequence overall, and their αI domains have 57% sequence identity. Based on sequence homology and recognition of the same ligand, it might be expected that α_Xβ₂ and α_Mβ₂ would bind similarly to iC3b (19). Here, using negative-stain EM, we clearly show that α_Xβ₂ and α_Mβ₂ bind to distinct sites on iC3b.

Structures of the α_Xβ₂ and α_Mβ₂ αI domains, including the α_X αI domain trapped in the open conformation by a lattice contact in bent α_Xβ₂ ectodomain crystals (30), and of the open α_M αI domain bound to the C3 TED (21) show marked differences that rationalize their differences in specificity. Residues in the α_M αI domain that contact the TED, which lie in loops that surround the MIDAS Mg²⁺ ion, differ completely in α_X. Thus, the most important TED-contacting residues in α_M, with the equivalent α_X residues in parenthesis, are Glu-178 (Asn), Glu-179 (Lys), Leu-206 (Gln), Arg-208 (Phe), and Phe-246 (Glu). The differences in side-chain structure, charge, and hydrophobicity are major and readily explain why α_Xβ₂ and α_Mβ₂ bind to different regions of iC3b.

Our studies on α_Xβ₂ here extended a previous EM study on the α_Xβ₂ ectodomain (19) by using the isolated αI domain and by identifying the location of a secondary binding site in iC3b. The isolated αI domain bound near the interface of the MG3 and MG4 domains in iC3b and C3c and less frequently to an additional, secondary site near the C345C knob in C3c. We were able to confirm binding to the secondary site by using the intact α_Xβ₂ ectodomain in both iC3b and C3c. Although we did not observe the secondary site using the αI domain with iC3b, the technical limitations of multivariate K-means classification may have been responsible for a failure to resolve both variation among particles in position of the TED and variation in whether a second αI domain was bound in class averages. The TED is larger than the αI domain and, thus, more dominant in K-means classification. We previously experienced similar dominance of the α_Xβ₂ ectodomain over the TED: In α_Xβ₂ complexes with iC3b, the TED was not resolved, whereas averaging iC3b particles in the same fields that were not bound to α_Xβ₂ resolved the TED (19).

The secondary binding site near the C345C knob resolved here both with binding of the α_X αI domain to C3c and binding of the α_Xβ₂ ectodomain to iC3b and C3c is adjacent to the CUB domain linkage to the MG key ring. This region becomes exposed when the CUB domain is cleaved in conversion of C3b to iC3b. Thus, the secondary binding site may be relevant in physiologic recognition of iC3b. Because α_Xβ₂ recognizes negatively charged residues in proteolyzed or denatured proteins as a danger receptor (26), it is reasonable to ask whether negatively charged regions are near the secondary recognition site. One candidate is the cleaved terminus of the CUB domain that precedes the MG8 domain, which is present in both iC3b and C3c. The terminal sequence is SEETKENEG and, thus, rich in acidic glutamic residues. Furthermore, this segment is disordered in the C3c crystal structure (31), consistent with recognition by α_Xβ₂ of denatured proteins and the appearance of the α_X αI domain on either side of the C345C knob in class averages. We found that two α_Xβ₂ ectodomain molecules could simultaneously bind to the primary site near MG3 and MG4 and the secondary site near the C345C knob on a single iC3b molecule. The use of both sites may contribute to more avid recognition by α_Xβ₂ (CR4) of iC3b.

In addition to our main focus here on iC3b recognition, our studies also provide insights into promiscuous recognition and headpiece opening by α_Mβ₂. Ligand binding through the αI domain was associated with swing-out of the β-subunit hybrid domain as visualized by the open headpiece conformation in EM. Studies with Fabs to the β₂ subunit that stabilize the open conformation show that they greatly increase affinity of α_Lβ₂ for its ligand ICAM-1 and of α_Xβ₂ for iC3b (32, 33). Here, the principle that headpiece opening is associated with the high-affinity state of the αI domain (Fig. 1B) has been extended to integrin α_Mβ₂, and was demonstrated for both binding of α_Mβ₂ to a specific ligand (iC3b) and a promiscuous ligand (α_Mβ₂).

Specific and promiscuous recognition by α_Mβ₂ are brought into focus here by its ability to both bind iC3b and self-associate into dimers. Integrin α_Mβ₂ has been found to bind a range of ∼40 different proteins, glycans including heparin, and denatured proteins (reviewed in ref. 25). Nonetheless, α_Mβ₂ (CR3) is highly selective for iC3b. α_Mβ₂ on the surface of myelomonocytic cells selectively rosettes with cells sensitized with iC3b but not with cells bearing IgM, C3b, or C3d (4, 5). The latter is equivalent to the TED.

Our studies on α_Mβ₂ show that it not only binds through its αI domain to the TED in iC3b, but also binds through its head region to the C345C-proximal region of the C3c moiety. The recent crystal structure of the α_M αI domain bound to TED was validated with measurement of a micromolar K_D value and elimination of binding with mutation of a key recognition residue in TED (21). However, because binding to TED (C3d) did not match the specificity of α_Mβ₂ for iC3b, this study raised two possibilities: (i) Binding to TED might reflect the promiscuity of α_Mβ₂ and not be related to iC3b recognition. (ii) As suggested by the authors, binding of the αI domain to TED might represent only one of two interactions, with the second interaction providing selectivity for iC3b over C3d. Although previous structural work on recognition by α_Mβ₂ of iC3b was with the isolated α_M αI domain, our current work is with the α_Mβ₂ headpiece and provides evidence for such a second interaction.

In α_Mβ₂ complexes with iC3b, despite the flexible linkage connecting the αI domain-bound TED to the C3c moiety through the unfolded, partially cleaved CUB domain, almost all class averages showed a similar orientation of the C3c moiety, with the thick side of the MG key ring facing the integrin and the C345C knob-proximal region of the C3c moiety touching or near to the β-propeller/βI domain portion of the α_Mβ₂ head. An interaction in this region of iC3b with the β-propeller/βI domain region of α_Mβ₂ is strongly suggested by the preponderance of class averages with this orientation and the contrast with the random orientation of the TED relative to C3c moiety in class averages of iC3b alone. Evidence for a specific interaction in this region is further strengthened by the uniform orientation of the α_Mβ₂ headpiece with its β-subunit side proximal to the C3c moiety of iC3b. Because in iC3b the TED is flexibly tethered through the unfolded remnant of the CUB domain to the C3c moiety, TED flexibility should generate random orientations of the C3c moiety relative to the α_Mβ₂ headpiece if there were no interactions other than those between the αI domain and TED. In summary, both the similar positions of the C3c moiety relative to the α_Mβ₂ headpiece in all well-resolved class averages of the iC3b–α_Mβ₂ headpiece complex, and the orientation with the thick side of the MG domain ring in C3c facing the headpiece and the β-subunit side of the headpiece facing the C3c moiety, support an additional stabilizing interaction between the C345C-proximal region of the C3c moiety and the β-propeller/βI domain portion of the α_Mβ₂ head.

An interaction of the knob-proximal region of the MG domain ring of iC3b with the β-propeller/βI domain region of α_Mβ₂ is consistent with previous evidence that these regions contribute to CR3 function. When the αI domain is deleted from α_Mβ₂, it retains partial ability to bind iC3b but not several other ligands, and residual CR3 function is completely inhibited by an antibody to the β-propeller domain (22). The antibody binds an epitope that includes a βI domain-proximal loop in blade (β-sheet) 6 of the β-propeller domain (34, 35). Mutagenesis has suggested a role for nearby blade 4 of the β-propeller domain and the βI domain in iC3b recognition (23, 24). Moreover, the recognition site near the C345C knob on the MG key ring of the C3c moiety is consistent with previous mutational data on CR3 recognition of iC3b. Mutation of iC3b residues 736 and 737, which localize to the α′NT segment, markedly diminishes recognition of iC3b by α_Mβ₂ (36). Notably, α′NT localizes to the thick side of the C3c key ring adjacent to the C345C knob (Fig. 1A), i.e., the same region of iC3b that we find interacts with the β-propeller/βI domain region of α_Mβ₂. Taken together, the mutational and antibody inhibition data on integrin α_Mβ₂, the mutational data on iC3b, and our EM data on α_Mβ₂ complexes with iC3b support a model in which the specificity of α_Mβ₂ for iC3b derives from bipartite binding of the αI domain and β-propeller/βI domain of α_Mβ₂ to the TED and C3c moieties, respectively, of iC3b.

Interestingly, we were also able to compare here the selective interaction of α_Mβ₂ with iC3b to a promiscuous self-association with α_Mβ₂. Notably, α_Mβ₂ self-associates in a reciprocal, symmetric interaction. Thus, the particular promiscuous interaction visualized here was favored by the greater avidity of a dimeric than a monomeric interaction. Despite the difference in avidity, when α_Mβ₂ and iC3b were mixed together, recognition of iC3b predominated over self-association, consistent with the selectivity of α_Mβ₂ for iC3b compared with other C3 fragments.

In summary, the direct comparisons here between α_Xβ₂ (CR4) and α_Mβ₂ (CR3) show that although these integrins are close relatives of one another, they bind to different sites on iC3b and, therefore, appear to have independently evolved their specificity for iC3b. There is extensive overlap in expression of these leukocyte integrins on cell types, with both expressed on macrophages and neutrophils, whereas α_Xβ₂ is selectively expressed on dendritic cells (10, 11). Remarkably, the primary binding site of α_Xβ₂ on the thin side of the C3c key ring lies distal from the α_Mβ₂ binding site on the TED domain and thick side of the C3c key ring and comparison of class averages of α_Xβ₂ and α_Mβ₂ complexes with iC3b (Figs. 2 and 3) suggest that no overlap between the integrins would occur upon simultaneous binding to the same iC3b molecule. The recognition of independent sites by CR3 and CR4 on iC3b provides greater diversity in recognition of opsonized pathogens and immune complexes, and may also allow cooperative interactions between the two integrins.

Materials and Methods

Integrin α_X αI Domain.

The α_X αI domain (residues E130 to G319 with F273S and F300A mutations) was expressed in pET28a with a 6 His tag at the N terminus in Escherichia coli BL 21(DE3) cells. Expression was induced with 1 mM isopropyl β-d-thiogalactopyranoside for 24 h at 16 °C. Washed bacteria were sonicated in 50 mM Tris pH 8.0, 300 mM NaCl, 10% (vol/vol) glycerol, and 1 mM phenylmethanesulfonyl fluoride. Protein was purified with Ni-NTA resin (Qiagen) and then by gel filtration on a Superdex 200 10/300 column (GE Healthcare) in 20 mM Tris pH 8.0, 300 mM NaCl, 10% (vol/vol) glycerol.

Integrin α_Mβ₂ Headpiece.

The cDNA encoding the mature residues 1–752 of α_M was cloned into ET11 vector, which was derived from ET1 vector (34). The cDNA encoding the signal sequence and mature residues 1–460 of β₂ subunit was cloned into pEF1-puro vector as described (37). HEK293 cells were cotransfected with expression constructs, and protein was expressed and purified as described (38).

Preparation of Complement iC3b and C3c.

Human C3 was purified as described (39). iC3b was prepared according to the literature (40) with slight modification. Generally, the C3 was digested by trypsin to generate C3b and then treated with Factor H/I to generate iC3b. Outdated human plasma (500 mL, stored for several weeks at 4 °C) was treated with 25 mg of zymosan (Complement Technology) for 5 d at 37 °C. C3c was purified by polyethylene glycol precipitation, anion exchange chromatography (DEAE Sephacel, GE Healthcare), cation exchange chromatography (SP fast flow Sepharose, GE Healthcare) and size-exclusion chromatography (Superdex 200 16/60, GE Healthcare) as described (31). C3c was finally stored in buffer containing 20 mM Tris pH 7.4, 150 mM NaCl, 2 mM MgCl₂.

Negative-Stain EM.

Purified α_X αI domain was mixed with iC3b or C3c at a molar ratio of 10:1 with MgCl₂ brought to a final concentration of 2 mM. Complexes were purified by using a Superdex 200 10/300 column (GE Healthcare) in 20 mM Hepes pH 7.4, 150 mM NaCl, and 2 mM MgCl₂. The α_Mβ₂ headpiece was mixed with iC3b or C3c at a molar ratio of 1:4 in presence of 1 mM MnCl₂ and 0.2 mM CaCl₂. Complexes were purified with Superdex 200 chromatography, 20 mM Hepes pH 7.4, 150 mM NaCl, 1 mM MnCl₂, and 0.2 mM CaCl₂. Peak fractions were adsorbed to glow-discharged carbon-coated copper grids, washed with deionized water, and stained with freshly prepared 0.75% uranyl formate. Low-dose images were acquired with an FEI Tecnai-12 transmission electron microscope at 120 kV and a nominal magnification of 67,000× or a Tecnai G² Spirit BioTWIN transmission electron microscope at 80 kV and a nominal magnification of 68,000×. Image processing was performed with SPIDER and EMAN as described (19). Five thousand to seven thousand particles were picked interactively and subjected to multireference alignment and K-means classification specifying 50 classes. Cross-correlation was as described (19). The projection of open α_Mβ₂ headpiece was generated by using a modeled open form α_xβ₂ headpiece, which contains α_x from ref. 30 and open β from ref. 41.