Structural insight on the recognition of surface-bound opsonins by the integrin I domain of complement receptor 3

Significance

Fragments of complement component C3 tag surfaces such as those presented by microbial pathogens or dying host cells for recognition by cells from the innate immune system. Complement receptor (CR) 3 enables efficient binding of complement-tagged surfaces by macrophages and dendritic cells, which eventually transport the CR3-bound material into lymph nodes. The study identifies in atomic details the fragments of CR3 and C3 required for such binding. The structural organization permits concomitant recognition by another complement receptor, namely CR2, expressed on cells of the adaptive immune system, suggesting a structural rationale for the exchange of antigens between leukocytes of the innate and adaptive immune systems critical in the formation of humoral immune responses.

Abstract

Complement receptors (CRs), expressed notably on myeloid and lymphoid cells, play an essential function in the elimination of complement-opsonized pathogens and apoptotic/necrotic cells. In addition, these receptors are crucial for the cross-talk between the innate and adaptive branches of the immune system. CR3 (also known as Mac-1, integrin α_Mβ₂, or CD11b/CD18) is expressed on all macrophages and recognizes iC3b on complement-opsonized objects, enabling their phagocytosis. We demonstrate that the C3d moiety of iC3b harbors the binding site for the CR3 αI domain, and our structure of the C3d:αI domain complex rationalizes the CR3 selectivity for iC3b. Based on extensive structural analysis, we suggest that the choice between a ligand glutamate or aspartate for coordination of a receptor metal ion-dependent adhesion site–bound metal ion is governed by the secondary structure of the ligand. Comparison of our structure to the CR2:C3d complex and the in vitro formation of a stable CR3:C3d:CR2 complex suggests a molecular mechanism for the hand-over of CR3-bound immune complexes from macrophages to CR2-presenting cells in lymph nodes.

Activation of complement leads to proteolytic cleavage of the central complement component, C3. Its major fragment, C3b, acts as an opsonin and becomes covalently bound to the activating surface via a reactive thioester located in the thioester (TE) domain of nascent C3b (Fig. S1A). Proteolytic processing by factor I within the CUB domain of C3b leads to the formation of iC3b and C3dg. Finally, C3d—which practically corresponds to the TE domain present in C3, C3b, and iC3b (Fig. S1 B–G)—is formed by other plasma proteases. These activation products are ligands for five complement receptors (1), with iC3b being the primary ligand of complement receptors (CRs) CR3 and CR4 (also known as CD11c/CD18, p150,95, or integrin α_Xβ₂), which is structurally similar to CR3.

Like other integrins, CR3 is a heterodimeric complex of two transmembrane proteins, α_M and β₂. It is abundantly expressed on myeloid leukocytes, including neutrophil granulocytes, dendritic cells, monocytes, and macrophages and also on lymphoid natural killer (NK) cells (2). Most ligands, including iC3b (3), are bound by the Von Willebrand factor A (VWA) domain in the α-chain, also referred to as the αI domain owing to its insertion in the β-propeller domain. I domain residues coordinate a metal ion essential for ligand recognition through a metal ion-dependent adhesion site (MIDAS). Integrins adopt at least three major conformations in the cell membrane. The bent-closed conformation is inactive in ligand binding, the extended-closed conformation has low ligand affinity, and the extended-open conformation binds ligands with high affinity. The transition from the bent-closed to the open-extended conformation is exerted by a cytoplasmic force on the leg of the β-subunit, a process usually referred to as the inside-out signaling (4).

Binding of ligands to CR3 leads to conformational changes in its ectodomain transmitting an outside-in signal through the cell membrane. This may lead to actin remodeling, phagocytosis, degranulation, and changes in leukocyte cytokine production (2, 5–7). CR3, and to a lesser degree CR4, are essential for the phagocytosis of complement-opsonized particles or complexes (6, 8, 9). Complement-opsonized immune complexes are captured in the lymph nodes by CR3-positive subcapsular sinus macrophages (SSMs) and conveyed directly to naïve B cells or through follicular dendritic cells (10) using CR1, CR2, and Fcγ receptors for antigen capture (11, 12). Hence, antigen-presenting cells such as SSMs may act as antigen storage and provide B lymphocytes with antigens (10, 12).

Here, we establish the C3d fragment as the minimal and high-affinity binding partner for the CR3 I domain. By contrast, the binding site for the CR4 I domain was located in the C3c fragment by electron microscopy (13). We present the crystal structure of the CR3 I domain in complex with C3d. The classic observation of CR3 binding to iC3b, but not to its precursor C3b (14), is consistent with our structure. In addition, our structure and functional data suggest simultaneous binding of CR3 and another complement receptor, CR2, to C3 fragments, which might provide the basis for trafficking of complement-opsonized immune complexes from macrophages to B cells and follicular dendritic cells in lymph nodes.

Results

CR3 and CR4 I Domains Recognize Distinct Binding Sites on iC3b.

To quantitatively compare binding properties of the CR3 and CR4 I domain with regard to binding of C3 proteolytic fragments, C3b, iC3b, C3c, C3dg, and C3d were immobilized in surface plasmon resonance (SPR) flow cells. For both I domains good binding signals were observed with C3b, iC3b, and C3c as ligands (Fig. S2 A–C and F–H). Nevertheless, even high concentrations (100 μM) of the CR3 or CR4 I domain did not lead to saturation. This is consistent with X-ray crystallography (15) and inhibition experiments showing that both the CR3 and CR4 I domain interact weakly (K_d ∼300 μM) with acidic side chains acting as ligand mimetics (16). The C3d and C3dg-coated surfaces produced robust SPR signals of ∼1,200 resonance units (RU) and showed signs of saturation at high CR3 I domain concentrations (Fig. S2 D and E). By contrast, the CR4 I domain only poorly bound these fragments (Fig. S2 I and J). As detailed in other studies (16, 17), the binding kinetics of the CR3 and CR4 I domain ligand binding are not well-described with simple 1:1 Langmuir isotherms. The interactions were quantified by analysis of the sensorgrams with the EVILFIT algorithm, which calculates the minimal distribution in binding kinetics for the heterogeneous interactions with ligands (Fig. 1). In general, the modeled distribution in kinetics efficiently described the experimental data as reflected in the small rmsds. For the CR3 and CR4 I domain, some of the interactions with C3b, iC3b, and C3c were of modest strength with K_d ∼10⁻⁴ to 10⁻³ M (Fig. 1 A and B and D–F), that is, quantitatively equivalent to binding of acidic side chains reported earlier (16). However, the classic CR3 ligand iC3b presented a population of interactions with K_d ∼10⁻⁷ to 10⁻⁶ M (Fig. 1B), not found with either C3b or C3c or for any fragments probed with the CR4 I domain (Fig. S3). This high-affinity type of interaction dominated the binding of the CR3 I domain to C3dg and C3d (Fig. 1 and Fig. S3).

Fig. 1.

SPR analysis of the C3 fragment binding selectivity of the CR3 (A–C) and CR4 I (D–F) domains. The CR3 or CR4 I domain, stabilized by mutagenesis in the open, ligand-binding conformation (50, 51), were injected in concentrations ranging from 250 nM to 100 μM over surfaces coupled with C3b (A and D), iC3b (B and E), or C3d (C and F). The data were analyzed with the EVILFIT algorithm with settings assuming a priori all binding parameters to be equally likely (52, 53). The volume of interactions, indicated with colored contours (in RU as shown by scale bars) was plotted as a function of the dissociation constant (10⁻⁸ M ≤ K_D ≤ 10⁻¹ M) and rate (10⁻³ s⁻¹ ≤ k_d ≤ 10⁰ s⁻¹). Red arrows indicate a population of high-affinity interactions for the CR3 I domain (K_D ∼0.4 μM) shared between iC3b and C3d but not observed for interactions with C3b.

Crystal Structure of CR3 I Domain in Complex with C3d.

Guided by the quantitative investigations made above, we determined the crystal structure of the CR3 I domain (subunit α_M residues 127–321, mature numbering) in complex with C3d (C3 residues 993–1,288, prepro numbering) at 2.8 Å resolution (Table S1 and Fig. S4A). The structure reveals a 1:1 complex between the CR3 I domain and C3d (Fig. 2A). The integrin I domain folds into an α/β Rossmann fold and adopts its open conformation as shown by comparison with the open-conformation structure of the I domain (Fig. S4B). The I domain α7 helix is shifted toward its C terminus, and this conformation is most likely favored by the I316G mutation introduced for this purpose. The open conformation is likewise adopted in the MIDAS site, where the hexa-coordinated metal ion is coordinated by Ser142, Ser144, Thr209 and two water molecules (Fig. 2B), whereas the last coordination position is occupied by an aspartate from C3d. Because Mg²⁺ was not compatible with the electron density in the MIDAS site, and because Ni²⁺ was present in the crystallization buffer, we used anomalous diffraction data to confirm the presence of a Ni²⁺ ion in the MIDAS site (Fig. 2B and Table S1). The ability of Ni²⁺ to stabilize MIDAS interactions with a ligand is well known from the complement convertases (18).

Fig. 2.

The structure of the C3d:I domain complex. (A) The edge of C3d (brown) interacts with the MIDAS (marked by the Ni²⁺ ion) of the CR3 I domain (purple). (B) The Ni²⁺ ion bound in the MIDAS. The electron density contoured at 6 σ obtained from anomalous differences in diffraction data (Table S1) is shown as a mesh around the Ni²⁺ ion. (C) Details of the intermolecular interface with putative hydrogen bonds and electrostatic interactions indicated by dashed lines. C3 labels are underlined in all panels. Fig. S4D shows a stereo view of the interface.

Within the complex, C3d adopts the well-described compact α-α6 barrel structure (Fig. 2A and Fig. S4C) with only minor conformational differences compared with known structures containing C3d (19–21). The α-helix connecting loop regions of C3d are presented in an alternating fashion at the circumference of two opposite surfaces: a concave, mainly negatively charged surface and a positively charged convex surface. At the rim of the concave surface Asp1247, situated in a loop region connecting helices α10 and α11 (Fig. 2C and Fig. S4 C and D), provides the final coordination bond to the divalent cation in the CR3 MIDAS (Fig. 2B). The quite polar interface between C3d and the I domain is modest, with an area of ∼490 Ų, which is smaller but comparable to similar I domain:ligand complexes (Table S2). Besides Asp1247 on C3d involved in the MIDAS ion interaction, the nearby Asp1245 in C3d engages in a salt bridge with CR3 Arg208 (Fig. 2C and Fig. S4D). C3d Lys1217 also seems to stabilize the interaction, because it is capable of forming salt bridges with CR3 Glu178 and Glu179 and hydrogen bonds with main chain carbonyls of Leu205 and Leu206. Finally, C3d Arg1254 also contributes to the interaction by forming hydrogen bonds with main-chain carbonyls of Gly143 and Ile145 in the I domain via a water molecule.

Conserved Features of MIDAS-Dependent Ligand Interactions.

Besides iC3b, the CR3 I domain is responsible for interacting with a variety of ligands, including fibrinogen, ICAM-1, and RAGE (22), but structures of their I domain complexes are not available. To identify general features among ligand side chains engaging in MIDAS ion coordination that might promote the identification of I domain interacting residues in other CR3 ligands, we identified all unique structures containing a VWA domain or a βI domain engaging in a MIDAS-dependent interaction with a ligand protein. Comparison of these structures revealed obvious trends for the use of either aspartate or the longer glutamate side chains as MIDAS ligands. If the small aspartate side chain is used, as in the C3d complex, it is located in a loop region or at the termini of a peptide (Fig. S5 A–D). This is even true for complexes between antibody Fab fragments and I domains (Fig. S5 E–G), where an aspartate is located peripherally in the heavy chain CDR3 loop. Even smaller than the aspartate and located at the end of a flexible region is the exposed C-terminal carboxylate group in the noncatalytic subunit of complement convertases (C3b/cobra venom factor/C4b) coordinating the MIDAS ion in the VWA domain (Fig. S5H) of the catalytic subunit (factor B/C2). We found three unique examples involving the longer glutamate side chain (Fig. S5 I–K). In complexes between the LFA-1 I domain and ICAM-1/3/5, a glutamate side chain at the end of a β-strand in an ICAM Ig domain is a MIDAS ion ligand. A resembling situation is found for the α₂β₁ I domain:collagen complex. In both cases the glutamate is protruding from a region containing regular secondary structure with little curvature and flexibility that would permit a closer approach to the MIDAS ion and the use of the shorter aspartate. In a third case, a crystal-packing interaction by a glutamate located immediately after an α-helix mimics a ligand of the CR3 I domain (15). Overall, there seems to be a steric selectivity in the MIDAS ion coordination by ligand acidic groups: An aspartate side chain is preferred in loops or flexible termini and the longer glutamate is preferred in regions with secondary structure, whereas both aspartate and glutamate located next to secondary structure are possible MIDAS ion coordinators.

C3d Asp1247 Is Essential for CR3 I Domain Interaction.

Because integrins may promiscuously bind acidic residues exposed on protein surfaces (15, 16, 22) and to exclude crystal-packing artifacts, we mutated C3d residues in the intermolecular interface. We tested iC3b and WT or mutated C3d for their ability to interact with CR3 I domain by isothermal titration calorimetry (ITC) experiments (Fig. 3). All of the binding ligands interacted with the CR3 I domain in a 1:1 stoichiometry, and iC3b generated from C3b with factor I bound the I domain with a K_D of 600 nM. The I domain bound WT C3d with a K_D of 450 nM. These numbers are in good agreement with the high-affinity site identified by SPR (Fig. 1 B and C and Fig. S3B). Mutation of C3d Asp1247 to alanine abolished CR3 I domain binding to C3d, suggesting a crucial role of this aspartate in the interaction with the MIDAS-coordinated cation. Replacing Asp1247 by a glutamate also impaired the interaction, showing that simple conservation of charge is not sufficient (Fig. 3). Most likely, when the longer glutamate side chain in this C3d mutant interacts with the I domain MIDAS site the adjoining interactions in the interface are difficult to form. The R1254A mutant (Fig. 3) displayed a lowered affinity of 2.2 μM, whereas the mutant C3d proteins K1217A and K1217A/ R1254A showed no detectable binding. Together, these data indicate that C3d Asp1247 is essential for the interaction with CR3 I domain but is not sufficient, because other residues of the opposite charge, namely K1217 and R1254, are required to steer the interaction.

Fig. 3.

ITC studies of the CR3 I domain interaction with iC3b, C3d WT, C3d R1254A, and C3d D1247E mutants. Raw titration isotherms and the integrated peak areas are shown. Dissociation constants calculated according to a simple independent binding site model are displayed.

Physiological Significance of the C3d:I Domain Structure.

The in vitro confirmation of our crystal structure by ITC and SPR experiments is also supported by prior data, because the iC3b binding site on the CR3 I domain earlier suggested (23) is in excellent agreement with our structure. Furthermore, the participating residues are strictly or highly conserved in mammalian C3 and CR3 α_M sequences (Fig. S6A). In contrast, none of the C3 residues involved in the I domain interaction is conserved in the structural C3 homolog C4 (Fig. S6B) (24), also being cleaved to C4b and undergoing further degradation to iC4b and C4d. Likewise, α_M residues with side chains interacting with C3d (Arg208, Glu178, and Glu179) are not conserved in CR4. Importantly, our structure also offers an explanation for why the iC3b proteolytic stage must be reached to allow CR3 interaction (25). The CUB domain in C3b connects the TE domain to the C3c moiety of C3b, and proteolytic degradation of the CUB domain probably causes its partial collapse in the resulting iC3b (Fig. S1E). Superposition of known structures of C3b suggests that the CUB domain prevents the I domain from interacting with the TE domain by steric hindrance (Fig. 4). Finally, our structure is in accordance with a favorable geometry where the CR3 binding site of iC3b or C3d(g) opsonizing the activator is accessible for a phagocytotic CR3-presenting cell. The CR3 binding site is separated by 40 Å from the Gln1013 forming the covalent bond to the complement activator surface (Fig. 4A). In conclusion, the in vivo relevance of our structure is strongly supported by (i) SPR and ITC experiments quantitating the interaction of iC3b, WT C3d, and mutated C3d with the I domain, (ii) the conservation of the residues in the molecular interface, (iii) our ability to rationalize the discrimination against CR3 binding to C3b, and (iv) the inferred lack of steric hindrance imposed by the activator upon CR3 binding.

Fig. 4.

The CR3 I domain discriminates against C3b. (A) The structure of C3b superimposed with the C3d:CR3 I domain complex. The βI domain and the α-chain β-propeller within CR3 are shown schematically. (B) Close-up of the region framed in A. Notice the overlap between the CR3 I domain and the C3b CUB domain in the hypothetical CR3:C3b complex. C3 labels are underlined.

C3 Thioester Domain As a Molecular Hub.

Our structure of the C3d:I domain complex together with the structures of C3 (26, 27), the C3d complexes with factor H (20, 28), and CR2 (21) demonstrate that a substantial fraction of the surface of the C3 TE domain is involved in intra- or intermolecular contacts in one of the multiple functional states of C3. Strikingly, surface residues on the TE domain interacting with other domains in native C3 are almost perfectly separated from those surface patches forming contacts with fH, CR3, and CR2 in their C3d complexes. In C3, the convex surface of C3d forms contacts with the MG2, CUB, and MG8 domains (total interface area of ∼2,650 Ų) that bury more than 20% of the TE domain surface (Fig. 5A). On the circumference and the rim of the concave surface of C3d, surface patches interacting with fH (∼600 Ų), CR3 (∼500 Ų), and CR2 (∼1,100 Ų) are likewise almost perfectly nonoverlapping with each other and with the above-mentioned C3 interdomain interface (Fig. 5 B and C). Hence, especially in the iC3b state, the TE domain may bind simultaneously and strongly to the two complement receptors and collateral binding of factor H and CR3 seems possible as well.

Fig. 5.

The multiple interactions of the TE domain in C3 and its proteolytic degradation products. (A) Surface areas (gray) in the C3 TE domain interacting with the CUB, MG8, and the MG2 domains in native C3 are mapped onto C3d from its CR3 complex; the thioester Gln1013 is shown in red. (B) As in A after a 180° rotation. Surface areas of C3d interacting with CR2 (pink), factor H (green), and CR3 (purple) are indicated. (C) Superposition of the complex between C3d (brown surface) with the CR3 I domain (purple cartoon) with that of the C3d:CR2 complex (21) and the C3d:factor H complex (20). (D) Silver-stained gel after denaturing SDS/PAGE analysis of fractions from the CR3 I domain affinity column. Labels + and − indicate, respectively, the presence or absence of the protein in the pull-down assay at that particular step. From left to right, molecular weight (Mw) marker together with purified CR2 CCP 1-2 (insect cell-expressed) and C3d and CR3 I domain affinity column pull-down using C3d WT showing the flow-through and the wash steps and the final EDTA elution. The same fractions are shown for the C3d D1247A and the D1154A control mutants unable to bind CR3 and CR2, respectively. All the volumes in wash and elution steps were the same. Likewise, the volumes loaded for SDS/PAGE analysis were identical. For comparison, the pull-down was performed with the CR2 expressed in either bacteria or insect cells (Fig. S7). (E) Silver-stained gel after denaturing SDS/PAGE analysis of fractions from analytical size-exclusion chromatography (Fig. S8). The same fractions resulting from the C3d:CR2, C3d:CR3 and CR3:C3d:CR2 runs were analyzed.

Simultaneous Binding of CR2 and CR3 to iC3b.

CR2 binds with similar affinity to iC3b and C3d(g) and with C3d as the minimal ligand (29). To verify whether the formation of the CR2:C3d:CR3 complex suggested above is possible, C3d and CR2 complement control protein (CCP) domains 1-2 were subjected to pull-down experiments with immobilized CR3 I domain. Bound proteins were eluted with EDTA to disrupt MIDAS–ligand interactions. In the presence of WT C3d, both CR2 and C3d were eluted from the CR3 affinity column, demonstrating that the CR3 I domain and CR2 CCP 1-2 are able to simultaneously bind to C3d (Fig. 5D and Fig. S7). Using instead the C3d D1247A mutant no protein was eluted, showing that CR2 is not interacting nonspecifically with the resin and that this C3d aspartate is essential for interaction with the CR3 MIDAS. With the C3d mutant D1154A unable to bind CR2 (30), CR2 was virtually absent from the EDTA eluate (Fig. 5D and Fig. S7). To further corroborate the existence of the ternary CR2:C3d:CR3 complex, size-exclusion chromatography experiments were conducted (Fig. S8). The chromatograms and SDS/PAGE analysis of fractions from these experiments showed that the complexes elute in the expected order CR3:C3d:CR2, CR3:C3d, and CR2:C3d (Fig. 5E and Fig. S8). Importantly, the presence of CR2 in the early fractions from the CR3:C3d:CR2 experiment can only be explained by the fact that CR2 was engaged in the ternary complex together with C3d and CR3. In conclusion, our pull-down and size-exclusion chromatography experiments confirmed the existence of a ternary complex in which the CR3 I domain and the CR2 CCP 1-2 fragment bind simultaneously to C3d.

Discussion

The engulfment by immune cells of complement-opsonized objects is a fundamental property of the human immune defense where CR3 plays an essential role (8). CR3’s specificity for iC3b as ligand has been established for 30 years. Now, our results define where and how CR3 interacts with iC3b and further emphasize the astonishing central role of the C3d moiety in intermolecular contacts. The CR3 I domain is well established as the primary binding site for iC3b (3, 31). The quantitative findings in our study clearly suggest that the binding of the CR3 I domain to C3d is 100- to 500-fold stronger than its weaker binding of simple acidic groups. Hence, the structure of the C3d:CR3 I domain complex shows a type of interaction distinct from the earlier report showing the binding to a ligand mimetic, that is, glutamate side chain contributed by a crystal lattice contact (15). Nevertheless, although the I domain is an important determinant in the binding of iC3b and C3d(g), other regions in CR3 must contribute, because deletion of its I domain results in residual iC3b affinity (32). Both the α_M β-propeller and the β₂ I-like domain (Fig. 4A) have been suggested to be implicated in the interaction with iC3b (33–35). On the iC3b side, mutations in the Nt-α′ region weaken the iC3b–CR3 interaction (36). The degradation product C3d(g) is not normally considered as a CR3 ligand, and soluble C3d cannot outcompete erythrocyte-bound iC3b in a rosette assay (37, 38). However, coating of erythrocytes with C3d facilitates their phagocytosis by monocytes in a metal ion- and CR3-dependent manner, although C3d does this much less efficiently than iC3b (39). This is likely to directly mirror the interaction we observe in the crystal structure and have quantitated by SPR and ITC. In summary, there seem to be at least two contact points between iC3b and CR3, a crucial one involving the C3d:I domain interaction established by us. If the iC3b Nt-α′ region in the C3c moiety of iC3b and the α_M β-propeller and the β₂ I-like domain in CR3 indeed are important for the receptor–ligand interactions, contacts involving these are likely to be spatially well separated from the C3d:I domain interface (Fig. 4A). Our model of iC3b binding to CR3 is clearly distinct from that proposed for CR4 by prior EM studies (13). In side-by-side SPR experiments of the I domain binding of C3 fragments, the iC3b and C3d(g) fragments presented high-affinity binding sites for the CR3 I domain, whereas C3b and C3c had no such interactions. These findings corroborate earlier studies on the intact receptor, supporting the relevance of studying the I domain selectivity towards C3 fragments. The CR4 I domain bound with almost indiscriminate kinetics to C3b, iC3b, and C3c, whereas the C3d(g) fragments were much poorer ligands. This is also in close agreement with the work by Chen et al. (13) on the intact ectodomain of CR4, which identified a major binding site for this receptor involving the MG3 and MG4 domains of C3c. The orientation of these domains, and hence the binding interface for CR4, are conserved between C3b and C3c (40), likely also making this the case for iC3b. This is quantitatively supported by the nearly identical kinetics for the CR4 I domain binding of these fragments observed in our study.

Our finding that CR2 and the CR3 I domain can bind to the same molecule of C3d is intriguing, and because CR2 does not seem to recognize elements outside the C3d moiety of iC3b, our results imply that one iC3b or C3d(g) molecule on soluble immune complexes can bridge a CR3-presenting cell with a CR2-presenting cell. The location of the ligand-binding region (CR2 CCP 1-2 and CR3 I domain) at the distant end of these large receptors far from the cell membrane makes immune complex bridging between two cells a realistic scenario. One important example is the suggested hand-over of complement-opsonized immune complexes from CR3-bearing subcapsular sinus macrophages to CR2 on naïve B cells or follicular dendritic cells in lymph nodes (10, 12). Concerning the remaining complement receptors, the binding site for CR4 at the MG3–MG4 domains mapped by EM (13) seems not to overlap with regions implicated in binding CR3. Both receptors are expressed together on many cell types including macrophages, monocytes, neutrophils, and NK cells (2), suggesting that simultaneous CR3 and CR4 binding to the same iC3b molecule is possible. The CRIg receptor expressed on a subset of tissue macrophages recognizes C3b, iC3b, and C3c, and its binding site has been mapped to the MG domains 3, 4, 5, and 6 and the LNK regions of C3b (41). Because the structure of iC3b is only known to low resolution and is controversial (13), the exact spatial relation between the iC3b binding sites for CRIg and the CR3 I domain found by us is unknown but likely to be variable. However, these sites are nonoverlapping, suggesting that both receptors could bind simultaneously to iC3b unless other parts of CR3 compete with CRIg. With respect to complement receptor 1, both CR1 and CR3 binding involve the same part of the Nt-α′ region (36), suggesting mutually exclusive binding of the two receptors to the same iC3b molecule.

Therapeutic intervention through blockade of CR3 has been suggested for treatment of cerebral stroke (42). Furthermore, macrophage migration to the brain is likely to involve CR3 (43). However, owing to the diversity of CR3 ligands, their central role in immune clearance of pathogens and apoptotic cells by phagocytosis (8, 44), adhesion (45, 46), and transmigration (47) overall or even function-specific CR3 blockade is not without risks. As an example, systemic lupus erythematosus is tightly linked with a single mutation in the α_M chain of CR3 that mainly affects its role in phagocytosis (48). In the same manner, C3b and its proteolytic fragments also engage in interactions with a large number of other proteins, thereby severely complicating the development of complement inhibitors. Our structure of the C3d:I domain complex provides an important contribution to the development of more selective complement inhibitors by identifying the surface areas on the CR3 I domain and on the C3d moiety of importance for their mutual interaction. Owing to the proximity of the C3d–CR2 interaction it may also contribute to improve the CR2-based targeting strategies under development for site-specific delivery of complement inhibitors (49).

Methods

WT or mutated C3d and CR3 I domain were prepared as recombinant proteins in Escherichia coli. The CR2 CCP 1-2 fragment was prepared as recombinant protein in insect cells or E. coli. C3b was prepared by trypsin digestion of C3, and iC3b, C3d(g), and C3c were prepared from C3b by factor I digestion in the presence of fH. In the SPR experiments proteolytic C3 derivatives were coupled through primary amine groups to the CM4 sensor chip and the CR3 I or CR4 I domain injected over the chip. ITC measurements were performed at 25 °C by titration of CR3 I domain with iC3b or WT/mutated C3d. For the C3d pull-down experiments, the CR3 I domain was coupled to CNBr-activated Sepharose, and C3d and CR2 were bound in the presence of Mg²⁺ and eluted with EDTA. The C3d:I domain complex was crystallized by vapor diffusion. Diffraction data from frozen crystals were collected with synchrotron radiation at the Swiss Light Source or European Synchrotron Radiation Facility, and the structure of the complex was determined by molecular replacement using Protein Data Bank ID codes 1IDO and 1C3D as search models. Coordinates and structure factors have been deposited at the Protein Data Bank with ID code 4M76. Detailed methods and the associated references can be found in SI Methods.