A novel inhibitor of the alternative pathway of complement reverses inflammation and bone destruction in experimental arthritis

Previously, we have shown that the IgV domain of human CRIg binds to the β chain of C3b, resulting in inhibition of the AP of complement (14). Here, we present the crystal structure of murine CRIg at 1.0 Å resolution. Comparison of the human and mouse CRIg reveals that both proteins are very similar. They share 83% sequence homology within the IgV domain, and they superimpose with a root mean square deviation of 0.37 Å for 110 C_α atoms (Fig. 1 A, left and middle). To determine whether the structural requirements for C3b binding and complement inhibition are similar in human and mouse, we performed alanine substitution of residues M86 and D87 (mutant “A”) or residues H57 and Q59 (mutant “B”; Fig. 1 A, right). These residues form key contact points with the MG3 and MG6 domains of human C3b, respectively (14). Recombinant soluble versions of these mutant proteins, consisting of the extracellular domain of murine CRIg fused to the Fc portion of murine IgG1 (CRIg(A)-Fc and CRIg(B)-Fc), showed a >100-fold loss of binding affinity to C3b when compared with wild-type protein (CRIg-Fc; not depicted). The mutant proteins were folded correctly as shown by similar circular dichroism patterns (not depicted). We next determined if the loss in binding activity translates into a loss of functional activity, i.e., inhibition of the C3 and C5 convertases. Although CRIg-Fc inhibited C3 and C5 convertases of the AP in mouse serum, CRIg(A)-Fc, CRIg(B)-Fc, and the control-Fc proteins lost the ability to inhibit either the C3 or the C5 convertases (Fig. 1 B). The inhibition was selective for the convertases of the AP because CRIg-Fc did not inhibit CP activation in hemolytic assays (Fig. 1 C and Fig. S1). Collectively, these results indicate that the structural requirements for binding to and inhibition of the AP convertases, as well as the selective inhibition of the AP, are conserved in human and mouse CRIg.

We next determined if CRIg-Fc can inhibit inflammation in vivo caused by activation of the AP of complement. We used two models of arthritis in which an intact AP is required for initiation of the disease (12, 15–17). CRIg-Fc treatment after secondary immunization with collagen (day 24) significantly (P < 0.01) reduced clinical scores, inflammation, and bone loss compared with control-Fc–treated mice (Fig. 2, A and D). A recent study has implicated a role for CRIg in Th cell–dependent B cell responses (18). In the CIA studies, systemic muCRIg-Fc did not affect B cell responses (Fig. 2 B), indicating that CRIg-Fc acts on the innate, and not adaptive, arm of the immune response in this disease model. To test whether CRIg-Fc could inhibit already established inflammation, mice were treated with wild-type or control CRIg proteins starting 42 d after primary immunization. CRIg-Fc treatment significantly (P < 0.0001) reduced clinical scores and inhibited inflammation and bone loss compared with treatment with a control fusion protein (Fig. 2, C and D), indicating that the AP of complement is required for progression of joint inflammation and bone destruction in CIA.

To further determine if CRIg inhibits the effector arm of immune complex–mediated inflammation, a cocktail of anti-collagen antibodies were transferred to mice, bypassing B and T cell–mediated immune responses (19). CRIg-Fc reduced clinical scores and inflammation in this model (Fig. 3, A and B), confirming that CRIg acts independently of T and B cell activity and that the AP is required for initiation of disease in AIA (12). The antiinflammatory effect of CRIg-Fc treatment was independent of the presence of endogenously expressed CRIg, as clinical scores in CRIg–wild-type and -knockout mice, injected with the anti-collagen antibodies, were similar (Fig. S2 A). Moreover, CRIg-Fc did not influence the levels of anti-collagen antibodies in the circulation (Fig. S2 B), indicating that CRIg-Fc did not interfere with macrophage-mediated clearance of circulating anti-collagen antibodies or antibody complexes. Finally, the inhibitory effect was not due to Fc receptor–mediated effector functions because treatment of mice with a CRIg-Fc fusion protein containing two mutations within the Fc effector domain (D265A and N297A) was equally effective compared with wild-type Fc protein in inhibiting clinical signs of arthritis (Fig. S2 C). Thus, CRIg-Fc inhibits immune complex–induced inflammation through the interaction of murine CRIg extracellular domain with C3b, a central component of the complement convertases.

Because cytokines play an important role in inflammation associated with arthritis (20, 21), we determined whether CRIg treatment affects cytokine production in the joint. CRIg-Fc did not affect cytokine responses shortly after LPS priming (day 3, 2 h after LPS treatment), but reduced local IL-1β and IL-6 levels that reflected the degree of joint inflammation at later time points (Fig. 3 C).

To test if CRIg-Fc can inhibit established inflammation, we initiated treatment 5 d after antibody transfer. Treatment with CRIg-Fc in established disease significantly (P < 0.001) reduced clinical scores and joint inflammation (Fig. 3 D). In contrast, treatment with CRIg(B)-Fc, that has lost its ability to inhibit the AP of complement, was not efficacious. Similarly, treatment of established disease with TNFRII-Fc fusion protein did not reduce clinical scores or inflammation. Thus, while TNF-α is required for induction of disease in this arthritis model, the AP of complement is required for maintaining inflammation in the joint.

We next determined if CRIg inhibited immune complex– induced inflammation through selective inhibition of the AP of complement. Anti-collagen antibodies in complex with collagen activate both the AP and CP of complement in mouse serum (22). CRIg-Fc inhibited C3 activation generated by the AP of complement (Fig. 4 A), but not C3 activation resulting from combined classical and MBL pathways (not depicted). Next to inhibiting complement activation induced by these immune complexes, CRIg-Fc significantly (P < 0.0001) inhibited complement activation in serum from arthritic mice (Fig. 4 B). As expected, control-Fc or CRIg(A)-Fc did not inhibit C3 deposition (Fig. 4 A) or anaphylatoxin generation (Fig. 4 B) after exposure of immune complexes to mouse serum. The reduced inhibitory activity of CRIg-Fc for C3 compared with C5 convertase activity in whole serum is likely due to the lower binding affinity of CRIg-Fc for the monomeric subunit of the C3 convertase (23) as compared with CRIg-Fc affinity for the dimeric C3b₂ subunit of the C5 convertase (Fig. S3). In addition to its complement inhibitory activity in mouse serum, CRIg inhibited complement activation locally in the joint as shown by the near absence of articular C3 fragments (Fig. 4 C) and a significant reduction in the C3 activation product, C3a Des arg (Fig. 4 D). Consistent with CRIg's selective inhibition of complement activation, anti-collagen antibodies were present in the joint of both CRIg-Fc– and control-Fc–treated mice. CRIg-Fc protein colocalized with C3 fragments in the joint (Fig. 4 E), indicating that systemically delivered protein localizes to the sites of complement C3 activation. Thus, CRIg inhibits antibody-induced AP complement activation in the joint resulting in a significant reduction in inflammation. These results indicate that the inhibitory effect of CRIg-Fc in vivo was fully dependent on its binding to, and inhibition of, the AP convertases.

This study presents CRIg as the first complement receptor that, in its soluble form, selectively inhibits the AP in vitro and in vivo. Its mode of inhibition is different from that of other native complement inhibitors. CRIg does not have decay or cofactor activity, but instead competes with binding of C5 to the convertases of the AP, but not CP (14). Because CRIg does not influence complement activation through the CP and MBL pathway it leaves part of the complement-mediated host defense system intact. Although the extracellular domain of CRIg inhibits complement activation, we have not found evidence that the membrane-bound form of the molecule acts as an intrinsic inhibitor of complement activation. It is however likely that CRIg acts as an AP regulator of convertases bound to the macrophage cell surface during clearance of serum-opsonized particles (13). Further studies, in which the complement inhibitory function of CRIg is uncoupled from its role as a phagocytic receptor, are necessary to clarify the role of macrophage-expressed CRIg in complement regulation.

Collectively, CRIg's ability to inhibit complement activation is well conserved in human and mouse and a soluble version provides a novel tool to study the effect of therapeutic intervention of the AP of complement in preclinical models of disease. Furthermore, the recently presented structure of CRIg in complex with C3b offers the opportunity for rational design of CRIg proteins with enhanced inhibitory efficacy. Collectively, CRIg presents a promising new therapeutic to target diseases in which the AP of complement plays a prominent role, including rheumatoid arthritis, anti-phospholipid syndrome, intestinal and renal ischemia-reperfusion injury, type II membranoproliferative glomerulonephritis, and age-related macular degeneration (8, 24).

Murine CRIg-Fc was generated as described elsewhere (13). Mutants in the extracellular domain of CRIg and mutations in the Fc effector domain of the molecule were introduced with the Quick-Change Mutagenesis kit (Stratagene) as described elsewhere (14). Circular dichroism spectra from the mutant proteins (CRIg(A)-Fc and CRIg(B)-Fc) were obtained as described previously (14). TNFRII-Fc was generated by fusing the extracellular domain of TNFRII with the Fc portion of murine IgG1. Surface plasmon resonance analysis was performed as described previously (13).

Fc-fusion proteins were diluted in GVB⁺/EGTA (0.1% gelatin/veronal buffer [BioWhittaker]/15 mM EGTA/24 mM MgCl₂) and mixed with 4% fresh BALB/c serum and 0.1 mg/ml activated zymosan in a final volume of 100 μl for 45 min at 37°C with shaking, and the reaction stopped with 900 μl GVB containing 50 mM EDTA. Zymosan was pelleted, and the supernatants were used for mouse C3a and C5a ELISAs.

Hemolysis assays for the AP and CP were performed as described previously (14). AP and CP hemolytic assays using assembled components were performed essentially as described previously (25). IgM-coated sheep erythrocytes were obtained from Comptech.

All animals were held under sterile pathogen-free conditions, and animal experiments were approved by the Institutional Animal Care and Use Committee of Genentech. CIA was achieved by immunization of 8–11-wk-old DBA/1J mice (The Jackson Laboratory) with bovine type II collagen as described previously (26). Mice were treated three times per week with 4 mg/kg protein when dosing was initiated on day 24, or three times per week with 12 mg/kg when dosing was initiated on day 42 after primary immunization. Antibody-induced, or passive, arthritis was performed by injection with an arthritogenic monoclonal antibody mixture (Chemicon) in 6-wk-old BALB/c mice (The Jackson Laboratory) as described previously (27). In brief, 2 mg anti-collagen type II antibodies was injected intravenously in mice, followed 3 d later by intraperitoneal injection with 25 μg LPS. Mice were treated daily with 4 mg/kg CRIg-Fc, TNFRII-Fc, or control-Fc (a monoclonal antibody to gp-120, IgG1 isotype). All dosing regiments (CIA and AIA) maintained the levels of CRIg-Fc and TNFRII-Fc between 50 and 100 μg/ml of serum. Clinical scoring was performed by trained personnel blinded to the nature of the treatment as described previously (26). For treatment of mice with already established disease, mice from the initial immunized cohort with an average clinical score of 6 were randomly assigned to the various treatment groups. Histologic evaluation of arthritis severity was performed on formalin-fixed, decalcified forelimb and hindlimb joints. Joints were scored for severity of inflammation by a pathologist blinded to the treatment using a scale of 0–5 for each paw according to Joosten et al. (28), except that synovial, bone, and cartilage changes were considered in aggregate when determining scores. X-ray micro-computed tomography (micro-ct) was performed ex vivo as described previously (26). Micro-ct estimates of joint cortical bone volume were obtained from the metatarsal-phalangeal and metacarpophalangeal joints.

Frozen sections from metatarsal-phalangeal joints were prepared as described elsewhere (29). Immunohistochemistry of joint sections was performed using an FITC-conjugated polyclonal Fab fragment to mouse C3 (Cappel) and an Alexa Fluor 594–conjugated goat anti–mouse monoclonal antibody to IgG2a (Jackson ImmunoResearch Laboratories). CRIg was detected using a primary rat anti–mouse monoclonal antibody (clone 14G6; reference 13), followed by an Alexa Fluor 594–conjugated goat anti–rat polyclonal antibody (Jackson ImmunoResearch Laboratories).

CRIg-Fc titers in the serum were determined with an ELISA using capture and detection monoclonal antibodies raised against the extracellular domain of murine CRIg (13). Anti-collagen antibody titers in the serum were determined in microtitre plates using bovine type II collagen (Chondrex) for antibody capture and biotinylated isotype-specific antibodies (Jackson ImmunoResearch Laboratories) SA-HRPO and TMB substrate (Kirkegaard and Perry Laboratories, Inc.) for detection. Complement activation induced by arthrogen antibodies was measured by adding 10 μg/ml of mouse anti-collagen antibodies to collagen-coated microtitre plates for 1 h, followed by washing three times with GVB (0.1% gelatin/veronal buffer; BioWhittaker). Fresh BALB/c serum (20% in either GVB⁺⁺ [GVB/2 mM CaCl₂/2 mM MgCl₂] or GVB/EGTA [GVB/30 mM EGTA/48 mM MgCl₂]) was added (10% final serum concentration), mixed for 1 min, and incubated at 37°C for 45 min. Complement C3 binding was detected with a goat anti–mouse horseradish peroxidase–conjugated C3 antibody (Cappel) using TMB substrate. The reaction was stopped by adding H₂SO₄, and optical density was read at 450 nm wavelength. Hemolysis assays specific for alternative and classical complement pathways were performed as described elsewhere (13, 14).

Mouse cytokine analysis was performed as described elsewhere (21). Hind footpads were cut at the borderline of fur growth and frozen in liquid nitrogen. The volume of PBS used for homogenization was adjusted to 75 mg of tissue per ml of PBS. Supernatants were subjected to ELISA for murine IL-1α, TNF-α, and IL-6 (DuoSet; R&D Systems). C3a des Arg levels were quantified with capture and detection antibodies from BD Biosciences (Fig. 4 B) or with a mouse C3a des Arg ELISA kit (Bachem; Fig. 4 D). In Fig. 4 B, OD measurements were converted to arbitrary units using a standard curve generated with a stock solution of zymosan-activated mouse serum (10 mg/ml incubated for 1 h at 37°C). C5a des Arg ELISAs were established using capture and detection antibodies purchased from BD Biosciences.

A DNA fragment encoding residues 20–137 of murine CRIg was cloned into the NdeI/BamHI sites of the pET28b expression vector (Novagen). Expression and purification were performed in the same manner as described for human CRIg (14). Crystals were grown at 19°C using the hanging drop vapor–diffusion method. 20 mg/ml of purified mCRIg in 25 mM Tris, pH 7.5, and 100 mM NaCl was mixed with equal volumes of a reservoir solution containing 0.2 M potassium tartrate and 20% PEG 3350. Crystals formed after 3 d. For data collection, crystals were dipped briefly into a solution-containing reservoir with an addition of 20% glycerol, and then flash-frozen in liquid nitrogen. Data were collected at Advanced Lightg Source beamline 5.0.2 and processed using HKL2000. Crystals of murine CRIg diffracted to 1.0 Å resolution belong to space group P2₁2₁2₁, with cell parameters of a = 31.3, b = 50.7, and c = 62.8 Å. The structure was solved with crystals containing seleno-methionine–labeled protein using multiple anomalous dispersion and program auto-SHARP (30). Refinement using Refmac and manual adjustments with program O resulted in a model with an R_cryst of 13.1% and an R_free of 15.1%.

All p-values were calculated with an unpaired, two-tailed Student's t test assuming equal variance.

Fig. S1 shows that CRIg Fc inhibits AP C5 convertase initiated by assembly of a CP C5 convertase on IgM-coated sheep red blood cells. Fig. S2 shows clinical scores in CRIg wild-type and knockout mice after AIA, the influence of CRIg-Fc on antibody titers in serum, and the efficacy of CRIg-Fc with a mutation in the Fc domain. Fig. S3 illustrates the kinetics of soluble C3b dimer (C3b₂) and monomer (C3b) binding to CRIg-Fc.

We thank Hergen Spits and Paul Godowski for their insightful comments; and Julio Ramirez, Pamela Chan, Jean-Philippe Stephan, Fojan Zamanian, Jason Chinn, Lizette Embuscado, and Laura DeForge for their valuable contributions.

We acknowledge the support staff at beamline 5.0.2 at the Advanced Light Source. The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, Materials Sciences Division of the U.S. Department of Energy under contract number DE-AC03-76SF00098 at Lawrence Berkeley National Laboratory. Coordinates and structure factors have been deposited at the Brookhaven Protein Database (2PND).

The authors have no conflicting financial interests.