A bispecific inhibitor of complement and coagulation blocks activation in complementopathy models via a novel mechanism

Complement is an arm of the immune system that responds to microbial infection by opsonizing microbes and tagging them for phagocytosis as well as lysing them through the action of the membrane attack complex (MAC).^1,2 Disorders including paroxysmal nocturnal hemoglobinuria (PNH), atypical hemolytic uremic syndrome (aHUS), and antiphospholipid syndrome are linked to an abnormal activation of complement because of defects in complement regulatory systems.^3,4 The classical, lectin, and alternative pathways (APs) of the complement converge at the conversion of complement factor C3 to the scaffold protein C3b. Further C3b generation is amplified by a proteolytic complex known as the alternative C3 convertase, C3bBb. The convertase is comprised of C3b and the activated serine protease factor Bb (FBb), which is formed by cleavage of the proenzyme FB into its Ba and Bb fragments by the protease factor D (FD). The complement response is amplified as more C3b is produced, leading to the accumulation of downstream pathway components and the production of the MAC. Importantly, the proinflammatory, chemoattractant anaphylatoxins C3a and C5a are produced as biproducts of proteolytic reactions involving C3bBb and the C5 convertase C3bBbC3b. The activity of convertase complexes is regulated by complement control proteins, such as decay accelerating factor and factors H and I, which disrupt the complex on the membrane surface. The plasma protein properdin is a positive regulator that binds C3bB and C3bBb and acts to stabilize them.^5-7 

It is now appreciated that the interplay between complement and coagulation is involved in the pathology of complement-related diseases.^8,9 Proteases of the 2 pathways can activate each other, and dysregulation of 1 can result in the abnormal activation of the other. Complementopathies are often characterized by tissue damage due to MAC generation, with accompanying microvascular thrombosis.⁹ Like the complement, the coagulation pathway is a proteolytic cascade, with surface-assembled protein complexes playing a central role. One of these is the prothrombinase complex, consisting of coagulation factors Xa (fXa) and Va (fVa), which assembles on the anionic phospholipid surfaces of activated platelets in the presence of calcium.^10-12 The complex catalyzes the cleavage of prothrombin into thrombin, leading to the formation of a fibrin clot.

Lufaxin, a naturally occurring protein found in the saliva of the blood-feeding sand fly, inhibits both the complement and coagulation cascades.^13,14 The protein was first described as an inhibitor of fXa that binds the enzyme and blocks the hydrolysis of small molecule substrates and prothrombin (by the prothrombinase complex).¹³ Later, it was also shown to inhibit activation of the AP of the complement by binding to C3bB and preventing its conversion to C3bBb.¹⁴ Specifically, lufaxin stabilizes the binding of FB to C3b while preventing the cleavage of FB by FD. Here, we elucidate the mechanism of action of lufaxin as both an anticoagulant and complement inhibitor. We use structural analyses to show the binding mode and inhibitory mechanism of this structurally novel inhibitor. We also show that lufaxin binds to fXa at a separate site, allowing it to block coagulation and complement independently and simultaneously, thus making it a potential model for bispecific inhibition of complement-related diseases.

For additional detailed methods, see the supplemental Data, available on the Blood website.

The lufaxin complementary DNA containing a 6-His tag was expressed in HEK-293 cells and purified as described previously.¹³ Human C3b and FB were obtained from CompTech Complement Technologies. Human fXa was obtained from Enzyme Research Laboratories. The anti-C5 monoclonal antibody (mAb) is the antibody incorporated in pharmaceutical grade eculizumab and was provided by Alexion Pharmaceuticals. For examples, see the article by Yuan et al.¹⁵ 

Surface plasmon resonance measurements were performed as described previously.¹⁴ Samples were injected in 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid at pH 7.4 and 0.15 M NaCl (HEPES buffered saline [HBS]) containing either 5 mM MgCl₂ or no divalent cation.

The modified Ham (mHam) assay, conventional Ham assay, thrombinoscopy, and thromboelastography were conducted as previously described.^16-21 All patients gave written informed consent. This study was approved by the institutional review board and conducted in accordance with the Declaration of Helsinki.

Lufaxin was crystallized using the hanging drop vapor diffusion method from 100 mM Tris HCl, pH 7.0, 10% polyethylene glycol 6000. Crystals were soaked in 0.5 M potassium bromide. Diffraction data were collected at beamline 22-ID at the Advanced Photon Source. Diffraction data were processed using XDS²² Phenix Autosol and Phenix Autobuild. The structure was rebuilt and refined using Coot²³ and Phenix Refine²⁴ (Table 1).

The C3bB-lufaxin and C3bB-lufaxin-fXa complexes were prepared by mixing 0.57 nmol of C3b, 0.57 nmol of FB, and 1.1 nmol of lufaxin in 1.0 mL of HBS buffer containing 5 mM MgCl_2, followed by concentrating the mixture. These samples were applied to freshly glow-discharged C-flat grids (Protochips, CF1.2/1.3-3Au). Grids were vitrified by plunging into liquid ethane and stored in liquid nitrogen before examination via cryo-EM. Images were recorded on a Glacios TEM at 200 kV equipped with a K3 direct electron detector in superresolution mode.

Movies of the C3bB-lufaxin and C3bB-lufaxin-fXa complexes were processed with MotionCor2,²⁵ and the contrast transfer function was estimated in Ctffind4.²⁶ Particle picking was conducted in Gautomatch (http://www.mrc-lmb.cam.ac.uk/kzhang/Gautomatch/) using references generated with EMAN2.²⁷ The picked particles were subjected to 2D classification using RELION-3.0.8 (Table 2; supplemental Figure 1).²⁸ A previously calculated map was used as an initial reference for 3D classification. The best class was selected for subsequent refinement (Fourier shell correlation = 0.143) in both RELION-3.0.8 and cisTEM²⁹ for C3bB-lufaxin and RELION 3.2 for C3bB-lufaxin-fXa (Table 2; supplemental Figure 1).

Models used to build the structures include the C3bB complex with nickel ion (Protein Data Bank accession code 2XWJ³⁰), the C3bB-FD complex with magnesium ion (2XWB³⁰), wild-type FB (2OK5³¹), and the structure of lufaxin. fXa was added to the model using coordinates from 1HCG³² and 1FAX.³³ Positioning and real space refinement were performed in Phenix²⁴ combined with manual rebuilding using Coot.²³ Model quality was evaluated using MolProbity.³⁴ Structural figures were produced with the PyMOL Molecular Graphics System, version 2.0 Schrodinger, Inc. or UCSF Chimera X (https://www.rbvi.ucsf.edu/chimerax/).³⁵ 

Lufaxin has previously been shown to inhibit the formation of the AP C3 convertase (C3bBb) by binding and stabilizing the proconvertase (C3bB; supplemental Figure 2) in a form that is poorly cleaved by FD.¹⁴ The protein was also shown to inhibit the lysis of rabbit erythrocytes by diluted human serum with an IC₅₀ of ∼60 nM.¹⁴ We evaluated the efficacy of lufaxin in preventing complement activation in ex vivo models of aHUS and PNH using the mHam and conventional Ham tests and compared its efficacy with that of the mAb component of eculizumab. In the mHam test, serum from patients with aHUS was used to activate the AP on a PIGA-mutant TF-1 cell line deficient in the GPI-anchored regulators CD55 and CD59. In addition to healthy human serum, Shiga toxin (Stx) and sialidase (Sia) are also potent activators of the complement in the mHam test and serve as positive controls (Figure 1A). Responses to Stx and Sia are inhibited by anti-C5 mAb at 0.34 μm and lufaxin at 2 μM (Figure 1A). Sia-induced activation is AP-specific (because it removes the receptor for factor H) and is inhibited by both mAb and lufaxin to a greater degree than Stx-induced activation, which involves contributions from the lectin pathway (Figure 1A). In the mHam test, lufaxin also significantly reduced the percentage of cell death in response to aHUS serum in a concentration-dependent manner, from 29.5% in the absence of lufaxin to 17.3% at 2 μM of inhibitor, making it approximately as effective as mAb (15.2% cell death) at a concentration of 0.34 μm (Figure 1B). Both inhibitors reduced cell death below the 20% level accepted as the threshold for a positive test.²⁰ In the conventional Ham test, lufaxin inhibited the lysis of erythrocytes from patients with PNH (relative to that in a healthy control) in an acidified healthy human serum (Figure 1C). Again, at 2 μM lufaxin, the degree of lysis (5.8%) was nearly identical to that observed with mAb at 0.34 μM (6.6%), and both inhibitors significantly reduced erythrocyte lysis from the level observed with serum alone (22.4%; Figure 1C).

The lufaxin family of proteins is present in the saliva of new (Lutzomyia)– and old (Phlebotomus)–world sand flies but shows no significant sequence homology to any protein family outside of the sand flies.¹³ We crystallized recombinant lufaxin and determined its structure using single-wavelength anomalous diffraction methods after soaking crystals in a precipitant solution containing potassium bromide (Table 1). The protein consists of 2 β-sandwich domains oriented at approximate right angles to one another (Figure 2A). The N-terminal domain (114 residues) is made up of 2 4-stranded β-sheets and is stabilized by 3 disulfide bonds linking Cys 29 and 37, Cys 55 and 114, and Cys 79 and 89 (Figure 2A). The C-terminal domain (164 residues) contains 6-stranded and 3-stranded β-sheets, with 1 disulfide bond linking Cys 235 with Cys 242. A C-terminal coil region is present, which is disordered after Asp 274. The 2 domains contact one another at an interface running along the lengths of βA and βH, including connecting loops from the N-terminal domain as well as termini and connecting loops of βA∗-E∗ of the C-terminal domain (Figure 2A). Glycan modifications are present at Asn 40 of the N-terminal domain and Asn 239 of the C-terminal domain (Figure 2A). Structural searches using DALI³⁶ did not detect any close structural homologs to lufaxin.

Conformational changes of C3b-bound FB are known to regulate its cleavage by FD, a serine protease, to produce the Ba and Bb fragments.^30,37,38 Two conformers of C3b-bound FB have been observed, and their ratio is affected by the type of divalent cation bound at the metal ion–dependent adhesion site (MIDAS) of the von Willebrand factor type A (VWA) domain of FB.³⁸ The C-terminal (αʹ-chain) carboxylate of the C345C domain of C3b is also inserted into this site and chelates the metal ion.^30,37 Divalent cation binding induces allosteric changes in the complex that enhance the binding of FB with C3b. In the presence of Mg²⁺, C3bB exhibits a high proportion of the closed conformation (35%) of bound FB, with its scissile bond being inaccessible to FD.³⁸ The closed conformation is exemplified by the published crystal structure of the cobra venom factor-FB complex (CVFB), which is FD-activated more slowly than C3bB (Figure 3A).³⁷ Replacement of Mg²⁺ by Ni²⁺ dramatically shifts the conformational equilibrium of FB in C3bB to an almost exclusively open conformation (98%), with the loop containing the scissile bond projecting away from the protein, exposing the P1 arginine residue and making the scissile bond accessible to FD (Figure 3B).^30,38 In the open conformation of C3bB, the serine protease domain of FB is rotated by 84° from its position in CVFB, bringing it into contact with the complement C1r/C1s, Uegf, Bmp1 (CUB) and macroglobulin-2 domains of C3b (Figure 3B), whereas in the closed conformation, the SP domain extends away from the surface and does not contact C3b (Figure 3A).³⁰ Binding of and cleavage by FD require formation of the open conformation, as was established in the crystal structure of C3bB bound with a catalytically inactive mutant of FD in the presence of Mg²⁺.³⁰ 

We used cryo-EM to analyze a C3bB-lufaxin complex formed in the presence of Mg²⁺ and refined its structure to 3.2 Å resolution (Table 2; supplemental Figure 1). Overall, the C3bB part of the complex is like that of the published crystal structure of CVFB but with C3b in place of CVF (Figure 3C-E).³⁷ The macroglobulin (MG1-6) ring, C345C, and CUB domains of C3b are well ordered. The thioester-containing domain of C3b is visible but poorly ordered because of its high mobility, as is the complement control protein-1 (CCP1) domain of FB (Figure 3C-F). Any disorder around the MIDAS in the VWA domain of FB makes it difficult to assess its Mg²⁺ occupancy. Weak density corresponding in position to the SP domain of FB in the CVFB crystal structure is also present (Figure 3D). Helix α7 of the VWA domain leading to the VWA-SP linker is well ordered and contains 4 full helical turns, as that in CVFB, rather than the 3 turns in the open conformation of C3bB or in the C3bB-FD complex (supplemental Figure 3A).^30,37 Helix αL of the VWA domain is shortened by 2 turns relative to the open conformation of C3bB and is like that of the CVFB complex with FB in the closed conformation (supplemental Figure 3A).³⁷ The scissile loop itself is disordered between Thr 216 and Lys 233, but its visible part is oriented similarly to that of CVFB, with the side chain of the P1 residue (Arg 234) contacting the surface of the VWA domain where it is inaccessible to FD (supplemental Figure 3B). Together, the orientation of the scissile bond loop, the position of the visible density for the SP domain, and the structures of α7 and αL clearly show that the C3bB-lufaxin complex contains FB in the activation-resistant, closed conformation (Figure 3A-B,F; supplemental Figure 3).

Lufaxin was fit into the unoccupied density projected from the region of the map where the CUB domain of C3b, the FBa region, and the VWA domain intersect. (Figure 3C-D). The lufaxin-binding site overlaps considerably with that of the SP domain in the open conformation of the proconvertase,³⁰ suggesting that the presence of the inhibitor blocks the conformational change of FB that is necessary for its activation (compare Figure 3A-B,F with Figure 4). The face of lufaxin contacting C3bB is exclusively contained in its smaller N-terminal domain, whereas the C-terminal domain of the inhibitor extends away from the C3bB surface between the thioester-containing domain and SP domains and does not contact the complex (Figures 2 and 3C-D).

The N-terminal domain of lufaxin interacts with the C3bB surface at 3 points contained in C3b and FB identified using the program PISA,³⁹ suggesting a bridging mechanism for binding that explains its stabilization of the proconvertase complex observed in surface plasmon resonance (SPR) and chromatographic experiments. The inhibitor is positioned where the main chains of the CCP3 domain of the Ba fragment, the VWA domain of FB, and the CUB domain of C3b lie within 9 Å of one another. At the first interaction point, loop B-C in the N-terminal domain of lufaxin (Gly 23-Ile 31) binds the CCP3 domain of FB (Figure 4A-C; supplemental Figures 4 and 5). This interface includes the loops of CCP3 containing Gln 181-Ser 185 and Gly 136-Ser 141, which are traversed by Lys 25-Ile 31 of the lufaxin loop B-C (Figure 4C). This region contains hydrogen bonds between Arg 26 of lufaxin and the carbonyl oxygen atoms of Gly 138 and Tyr 139 of FB as well as between the amide nitrogen of Lys 25 of lufaxin and the carbonyl oxygen of Gly 183 of FB (Figure 4C; supplemental Figure 5). At the second contact point, loop G-H at one end of the N-terminal domain of lufaxin contacts the VWA domain at its end opposite the MIDAS (Figure 4A-C; supplemental Figures 4 and 6). Hydrogen bonds are formed between the Arg 98 side chain of lufaxin and the hydroxyl group of Tyr 397 of FB (Figure 4C) and between the side chain of Asp 27 in loop BC of lufaxin and Asn 423 of FB. Sequence comparisons of FB from different vertebrate species show that these interface sequences are highly conserved, suggesting that lufaxin would interact with C3b-bound FB in mammals and other vertebrate classes (supplemental Figure 6). At the third interaction point, β-strands D and E of the CUB domain of C3b run parallel to β-strands B and H of the N-terminal domain of lufaxin (Figure 4A-B,D). Phe 19 of lufaxin is positioned to form a possible cation-π interaction (4.2 Å) with Arg 1288 from the CUB domain, and Glu 105 of lufaxin also forms a salt bridge with Arg 1288 of the CUB domain (Figure 4D). A second hydrogen bond is formed between the hydroxyl group of Tyr 64 from lufaxin and the amino group of Lys 1284 in the CUB domain. Sequence comparisons of C3 from mammals, reptiles, and amphibians show incomplete conservation of CUB domain–binding interface residues (supplemental Figure 7). Nevertheless, Arg 1288 is conserved across the group, including in CVF from Naja naja kaouthia (supplemental Figure 7). We performed SPR and size exclusion chromatography experiments, showing that lufaxin stabilizes binding of FB with CVF in the presence or absence of divalent cations as it does with C3b, further suggesting that interactions involving Arg 1288 (or equivalents) may play a key role in the binding of lufaxin (supplemental Figure 8).

In addition to inhibiting complement, recombinant lufaxin is a potent inhibitor of coagulation in human plasma. Using the thrombinoscope, thrombin generation was found to be reduced 15-fold at a concentration of 2.5 nM, and little or no thrombin was generated at a lufaxin concentration of 5 nM (Figure 5A). Likewise, the thromboelastogram exhibits large concentration-dependent increases in the R and K values, which become difficult to measure at lufaxin concentrations above 1.5 nM (Figure 5B). In addition, size exclusion chromatography of C3bB-lufaxin-fXa mixtures shows fXa binding to lufaxin in the presence of C3bB. All components are included in a single complex, as indicated by the coelution of the C3bB components, lufaxin, and the chains of fXa in the complex peak (Figure 5C-D, bottom gel). In the absence of lufaxin, fXa does not interact detectably with C3bB (Figure 5C,D, top gel). SPR results show that fXa causes only a small reduction in the rate of complex formation and essentially no change in the dissociation rate constant, suggesting that the complex assembles freely without significant steric hindrance (Figure 5E). In accordance with this, lufaxin (30 nM) strongly inhibits hydrolysis of the substrate S-2222 by fXa in the presence of Mg²⁺ and Ca²⁺ and an excess of C3bB (Figure 5F).

The structure of lufaxin bears no resemblance to that of other fXa inhibitors, and its binding mechanism is unknown. With the goal of understanding this mechanism, we determined the binding mode of lufaxin with fXa by producing a C3bB-lufaxin-fXa complex in the presence of equimolar lufaxin and fXa and determined its structure using cryo-EM (Table 2). In the 3D reconstruction, C3bB-lufaxin is present in the arrangement described earlier but with additional density contiguous with the C-terminal domain of lufaxin (Figure 6A-B). This extra density was assigned as the serine protease domain of fXa, which is well defined, whereas the epidermal growth factor and Gla domains of fXa are less well ordered and project into the solution away from the C3bB complex (Figure 6A-B). In this orientation, fXa does not interfere with lufaxin binding to C3bB because the inhibitor interacts with C3bB via its N-terminal domain and with fXa via its C-terminal domain.

Substrate entry to the active site of fXa is prevented by lufaxin, which occludes the prothrombin-binding face of the protease (Figure 6A,C-G).¹² Residues forming this interface were identified using PISA and are shown in the amino acid alignments of supplemental Figure 9. The C-terminal coil of lufaxin runs toward the active site of the protease, and the side chain of Arg 277 fills the S1 subsite at the active site of fXa, participating in several hydrogen bond and salt bridge interactions with Asp 189, Ala 190, and Gly 216 of fXa (Figure 6C; Figure E-F; supplemental Figure 5). The carbonyl group (or C-terminal carboxyl group) of Arg 277 is situated adjacent to the catalytic serine residue (Ser 195; Figure 6D-F) of fXa, and Gly 278 and the 6-histidine tag immediately after it is removed by cleavage, as verified by the absence of the His-tag epitope in western blots after incubation of lufaxin with excess fXa (supplemental Figure 10). Loop A∗-B∗ (Arg 155-Glu 159), residues along β-strand G∗ and residues in the H∗-I∗ loop of lufaxin bind the autolysis loop of fXa. Tyr 228 and His 230 in β-strand G∗ pack against the apex of the loop (Figure 6C; supplemental Figure 5),⁴⁰ whereas loop A∗-B∗ of the inhibitor lies along 1 face of the autolysis loop with the side chain of Gln 156 of lufaxin contacting the side chain of Phe 156 from fXa. Arg 150 in the autolysis loop forms salt bridges with Asp 247 and Asp 249 in the H∗-I∗ loop of lufaxin (Figure 6C; supplemental Figure 5). Arg 222 of fXa, which lies in a loop (220-loop) forming 1 wall of the S1 subsite that has been identified as a sodium ion–binding site,⁴¹ forms a salt bridge with Glu 234 in the G∗-H∗ loop of the inhibitor. Structure prediction of the fXa-lufaxin complex was performed after the experimental solution was determined using Alphafold 2.3.1,⁴² which gives improved predictions of multimeric complexes. The structure is consistent with that determined using cryo-EM (supplemental Figure 11).

The target of lufaxin in vivo is fXa contained in the prothrombinase complex rather than free fXa. To examine the possibility of C3bB-bound lufaxin interacting with this target, we produced a model of C3bB-lufaxin in complex with prothrombinase by superposition of the fXa protease domains of C3bB-lufaxin-fXa and the recently determined structure (Protein Data Bank accession 7TPP) of prothrombin-prothrombinase,¹² with the prothrombin coordinates removed (fVa-fXa; Figure 7). The model suggests that lufaxin can join the 2 complexes into a larger C3bB-lufaxin-prothrombinase complex without introducing significant molecular clashes, making it possible that the inhibitor binds C3bB and prothrombinase simultaneously (Figure 7). Membrane-binding sites in this modeled complex are not perfectly coplanar, making it uncertain whether C3bB and prothrombinase can interact on cell surfaces.

Complementopathies such as aHUS and PNH are associated with the activation of the AP and the coagulation cascade.^9,43-46 C3bB is an attractive target for therapeutic inhibition because it is the precursor of C3bBb, the central convertase of the AP, and a point of amplification for the overall complement response. Blockade at this step prevents the accumulation of C3b and anaphylatoxin C3a as well as the production of downstream terminal pathway intermediates, including anaphylatoxin C5a and components of the MAC. Small molecule FD and FB inhibitors are now available that specifically block the conversion of C3bB to C3bBb and have shown promise in treating complement-related diseases.^21,47,48 Here, we describe the structure and mechanism of lufaxin, a macromolecular inhibitor of C3bB activation that has the added feature of being a potent inhibitor of coagulation through the binding of fXa.

The complement inhibitory mechanism of lufaxin is novel. The inhibitor binds at a site that includes elements of C3b and FB and prevents activation of FB by blocking the conformational changes necessary for interaction with FD. The complex is stabilized even in the absence of normally required divalent cations through bridging interactions between lufaxin, C3b, and FB. Lufaxin itself shows little interaction with the individual components of complex C3b and FB. The binding of lufaxin¹⁴ in the absence of Mg²⁺ suggests the formation of a low-affinity encounter complex between C3b and FB in its closed conformation that allows less hindered access by lufaxin than in the presence of Mg²⁺. When bound to C3bB, lufaxin directs its C-terminal domain, using its fXa-binding site, away from the surface of the complex, where it is free to interact with fXa, allowing the inhibitor to bind to both targets simultaneously. The structure of lufaxin is not like that of other naturally occurring proteinaceous fXa inhibitors, including various serpins and the tick anticoagulant peptide. The tick anticoagulant peptide is a Kunitz-type inhibitor that interacts with fXa in a noncanonical manner via an N-terminal tyrosine residue whose side chain inserts into the S1 subsite of the protease,⁴⁹ whereas the mechanism of lufaxin is centered on a C-terminal arginine residue that inserts at the active site of fXa.

The co-occurrence of complement and fibrin deposition in many complement-related disorders suggests that concomitant inhibition of complement and coagulation may be an effective avenue for treatment.⁴³ The 2 pathways are activated reciprocally in that tissue damage caused by the MAC results in the generation of procoagulant surfaces and subsequently activated coagulation proteases are capable of activating complement in a noncanonical manner.⁵⁰ In addition, MAC formation on platelet surfaces or microparticles and the activation of platelets by anaphylatoxin C3a lead to the exposure of anionic phospholipids and binding of coagulation complexes.^51,52 Because of the direct interplay between the 2 pathways, complement deposition and thrombosis are colocalized. Our ex vivo assays clearly show that lufaxin limits complement activation in sera from patients with aHUS and red blood cells from patients with PNH to a similar degree as the anti-C5 mAb. Assays of thrombin generation and fXa activity in human plasma and reconstituted enzyme systems demonstrate that lufaxin also potently inhibits coagulation. With separate binding sites for C3bB and fXa allowing simultaneous binding of the 2 targets, lufaxin may serve as a model for the development of bispecific therapeutics. Successes with bispecific antibodies developed for tumor immunotherapy and the treatment of A-type hemophilia suggest that other systems involving multiple extracellular targets can be effectively treated in this manner.^53,54 

The authors thank the staff of the Southeast Regional Collaborative Access Team at the Advanced Photon Source for assistance with diffraction data collection. Elizabeth Fischer and her team in the National Institute of Allergy and Infectious Diseases Research Technologies Branch assisted with cryo-EM data collection. This work required the use of the computational resources of the NIH High-Performance Computing Biowulf cluster (http://hpc.nih.gov).

This work was supported by the intramural research program of the National Institute of Allergy and Infectious Diseases, National Institutes of Health; National Institutes of Health, National Heart, Lung, and Blood Institute, grant R56HL133113 (R.A.B.); Johns Hopkins Hospital grant 80053630 (I.M.B.F.); and an American Society of Hematology Research Training Award for Fellows (G.F.G).

Contribution: J.F.A., I.M.B.F., and R.A.B. designed the study; J.F.A., H.L., E.C.S., I.M.B.F., T.K., G.F.G., X-Z.P., O.A.A., and P.H.A. performed experiments and analyzed data; V.P. provided technical assistance; J.F.A. drafted the manuscript; and J.G.V. and J.M.C.R. contributed to subsequent drafts.

Conflict-of-interest disclosure: R.A.B. has served on an advisory board for Alexion Pharmaceutical Inc. G.F.G serves on a medical advisory board and has received an honorarium from Apellis Pharmaceuticals. The remaining authors declare no competing financial interests.

Correspondence: John F. Andersen, Laboratory of Malaria and Vector Research, National Institutes of Health-National Institute of Allergy and Infectious Diseases, 12735 Twinbrook Parkway, Rockville, MD 20852; e-mail: john.andersen@nih.gov.