Complement C3 recognition by C3 convertases

Abstract

The complement system plays a fundamental role in immunity, and its dysregulation is implicated in numerous human diseases. Activation of complement occurs through three main pathways: classical, lectin, and alternative; which converge at the central component, component of complement 3 (C3). The classical and lectin pathways use the C4b2a convertase to cleave C3 and initiate complement activation, while the alternative pathway uses the C3bBb convertase, which is further stabilized by properdin. The molecular mechanisms governing C3 recognition by these convertase complexes remain incompletely understood. Here, we present the 3.1-angstrom cryo–electron microscopy structure of the C4b2a-C3 Michaelis complex, alongside 2.9- and 3.1-angstrom structures of the C4b2 zymogen in loading and activation states, elucidating the structural basis for C3 engagement by C4b2a and conformational changes during the classical and lectin pathway convertase maturation. Furthermore, a 2.6-angstrom structure of C3bBb-properdin in complex with C3 uncovers unique substrate-binding features of C3bBb and properdin’s stabilizing role in the alternative pathway. These results offer comprehensive mechanistic insights into complement activation.

Cryo-EM structures reveal how immune complement pathways activate, advancing mechanistic understandings and therapeutic insights.

INTRODUCTION

The complement system is a vital component of innate immunity, comprising a complex network of more than 50 soluble proteins and cell surface receptors that play essential roles in pathogen clearance, immune modulation, and the maintenance of homeostasis (1, 2). Dysregulation of the complement system has been associated with various human disorders, including renal, rheumatic, neurological, and cardiovascular diseases (3–6). More recently, complement dysregulation has also been implicated in severe acute respiratory syndrome coronavirus 2–related pathogenesis, particularly with respect to thrombosis (7, 8).

The complement system operates through three distinct but interconnected activation pathways: the classical pathway, the lectin pathway, and the alternative pathway. Each of these pathways converges at the central component, component of complement 3 (C3), which is cleaved by the C3 convertases to produce the fragments C3a and C3b. C3a functions as an anaphylatoxin (ANA), triggering inflammatory responses through its interaction with the G protein–coupled receptor C3aR, while C3b acts as an opsonin that binds to pathogens or abnormal host cells, enhancing their recognition and clearance by phagocytes. In addition, C3b serves as a critical subunit for the alternative pathway C3 convertase, promoting the amplification of the complement response.

C3 convertases are key enzymatic complexes within the complement system that play a crucial role in activating the complement cascade (Fig. 1A). There are two main forms of C3 convertase: C4b2a (here, the convention where “a” denotes the enzymatically active fragment of C2 is used) in the classical and lectin pathways and C3bBb in the alternative pathway (9, 10). C4b is produced by C1s in the classical pathway or by MASP2 (mannose-binding protein-associated serine protease 2) in the lectin pathway. The C4b2a convertase is then generated when the C4b binds to C2, and C1s or MASP1 or MASP2 then cleaves the bound C2 to form the active C4b2a complex. In contrast, the C3bBb convertase arises from the binding of C3b to factor B (FB), which is activated by factor D (FD). The resulting C3bBb convertase is stabilized by properdin (11). Alternatively, a small fraction of C3 can undergo spontaneous hydrolysis, known as “C3 tick over,” yielding C3(H₂O) that adopts a C3b-like conformation (12), and can thereby lead to the formation of the C3(H₂O)Bb convertase. Recent studies have shown that granzyme K can also activate the complement cascade by cleaving C4 and C2, which leads to the formation of the C4b2a convertase (13), or by directly cleaving and activating C3 (14).

Fig. 1. Overview of the complement C3 convertases.

(A) A simplified diagram of the complement cascade. The molecular complexes resolved in this study are highlighted using red dotted boxes. (B) Schematic representations of the domain organization of human C4, C3, C2, FB, and properdin. All proteins are numbered starting from the first methionine residue, including signal peptides. Domain boundaries of C4: macroglobulin domain 1 (MG1; 20 to 137), MG2 (138 to 240), MG3 (241 to 364), MG4 (365 to 465), MG5 (466 to 563), MG6β (564 to 605), C4a (680 to 756), α′ chain N-terminal (α′-NT; 757 to 778), MG6α (779 to 832), MG7 (833 to 935), CUB (complement C1r-C1s, urchin embryonic growth factor, bone morphogenetic protein 1; 936 to 983 and 1324 to 1388), thioester domain (TED; 984 to 1323), MG8 (1389 to 1427 and 1454 to 1573), anchor (1574 to 1594), and C345C (1595 to 1744). Domain boundaries of C3: MG1 (23 to 126), MG2 (127 to 231), MG3 (232 to 350), MG4 (351 to 451), MG5 (452 to 556), MG6β (557 to 599), C3a (675 to 742), α′-NT (749 to 756), MG6α (757 to 828), MG7 (829 to 934), CUB (935 to 970 and 1324 to 1345), TED (980 to 1323), MG8 (1356 to 1496), anchor (1486 to 1517), and C345C (1518 to 1661). Domain boundaries of C2: complement control protein 1 (CCP1; 21 to 85), CCP2 (86 to 147), CCP3 (148 to 206), vWA (237 to 452), SP (467 to 752). Domain boundaries of FB: CCP1 (35 to 100), CCP2 (101 to 160), CCP3 (163 to 220), von Willebrand A domain (vWA; 270 to 469), and serine protease (SP; 477 to 757). Domain boundaries of properdin: transforming growth factor beta binding protein-like domain (TB; 28 to 76), thrombospondin type I repeat 1 (TSR1; 77 to 134), TSR2 (136 to 191), TSR3 (193 to 255), TSR4 (257 to 313), TSR5 (315 to 377), and TSR6 (379 to 462). The yellow connecting lines indicate interchain disulfide bonds, while the triangular arrows denote cleavage sites on C4 and C3. (C) SDS–polyacrylamide gel electrophoresis analysis of purified proteins, including C4, C4b, C2, C3, C3b, FB, properdin, and FD. Asterisks (*) indicate contaminant proteins.

Both C4b2a and C3bBb are inherently labile, dissociating within minutes after formation (15, 16), posing substantial challenges for their study. Previous breakthroughs, in particular the determination of the C3bB zymogen structure in complex with FD (17) and the C3bBb-SCIN (staphylococcal complement inhibitor) complex (18), have provided foundational insights into the alternative pathway convertase assembly. The general location of the C3 substrate-binding site on C3bBb was inferred from the crystallographic dimer interface formed between the two C3b molecules in the C3bBb-SCIN structure and also the structure of C5 in complex with the cobra venom factor (CVF), a homolog of C3b (19). However, the precise molecular mechanism of substrate engagement by the C3 convertases remains elusive. Furthermore, the assembly and activation process of the classical and lectin pathway C4b2a convertase are poorly understood.

In this study, we use cryo–electron microscopy (cryo-EM) to resolve the molecular architecture of these convertases in functional states. We report a 3.1-Å structure of the C4b2a-C3 Michaelis complex, elucidating the structural basis for C3 recognition by the classical and lectin pathway convertase. We also resolve the 2.9-Å and 3.1-Å structures of the C4b2 zymogen (standardized notation for the C4bC2 complex) in the loading and activation states and thereby reveal sequential conformational transitions during the classical and lectin pathway convertase formation and maturation. Furthermore, we capture a 2.6-Å structure of the C3bBb-C3 Michaelis complex in the presence of properdin, which unveils distinct features governing substrate engagement by C3bBb. Detailed analysis also delineates how properdin stabilizes the C3b-Bb interaction, thereby positively regulating the alternative pathway convertase. These high-resolution snapshots address long-standing questions in complement biology.

RESULTS

Cryo-EM structure of the C4b2a-C3 complex

To capture a stable C3 convertase for cryo-EM analysis, we first isolated C3 and C4 from human plasma as previously described (Fig. 1, B and C) (20, 21). C4b was then generated by incubating C4 with C1s. In addition, we produced C2(S679A; Ser⁶⁷⁹→Ala), a catalytically inactive mutant of C2, using human embryonic kidney (HEK) 293F cells, which allowed efficient Michaelis complex formation in solution (fig. S1). Following this, C4b, C2(S679A), and C3 were mixed in an equal molar ratio, with Ni²⁺ included to enhance the C4b-C2 interaction (22). Afterward, this sample was briefly treated with C1s and then rapidly plunged in liquid ethane for cryo-EM analysis. The C1s treatment resulted in partial cleavage of C2, leading to the formation of the C4b2a convertase, which readily engages C3 and enables us to trap the classical pathway C3 convertase-substrate complex and visualize the structure at a resolution of 3.1 Å (Fig. 2A, fig. S2, and table S1). In addition, we captured two functional states of the C4b2 proconvertase containing intact C2 from the same sample: a loading state at 2.9 Å and an activation state at 3.1 Å.

Fig. 2. Cryo-EM structure of the classical and lectin pathway C3 convertase.

(A) Two views of the cryo-EM reconstruction of the C4b2a-C3 complex. Complement proteins C4b, C3, and C2b are depicted in blue, orange, and green, respectively, while the C3a/ANA domain is highlighted in red. (B) A ribbon diagram of the C4b2a-C3 structure accompanied by a schematic representation. (C) The interaction between C4b and C3 involves multiple MG domains from both proteins. (D) Interaction details between C3-MG4 and C3-MG5 and C4b-MG5 and C4b-MG4. Dashed lines indicate polar interactions including hydrogen bonds and salt bridges at distances less than 3.5 Å. (E) Interaction between the C3 scissile loop and the C2a-SP domain, including enlarged views of the density map within the catalytic pocket on the top (red and green represent C3 scissile bond and C2a SP domain, respectively) and detailed molecular interactions on the right. In the right panel, C3a/ANA is shown in red and α′-NT in orange. The cleavage site is indicated by a black arrowhead. The oxyanion hole in C2a-SP is highlighted in blue, and the catalytic triad Ser⁶⁷⁹(A), His⁵⁰⁷, and Asp⁵⁶¹ are highlighted in yellow. The hydroxyl group of Ser⁶⁷⁹ is depicted as hypothetical and is rendered in a lighter color.

In the C4b2a-C3 complex, C4b adopts an active conformation typical of C3b and C4b, resembling a “puppeteer” structure (Fig. 2B) (23). This includes a head (C345C domain), a neck (anchor), shoulders [macroglobulin domain 7 (MG7) to MG8], a body [MG1 to MG6; Linker domain (LNK)], and a downward-extending arm [complement C1r-C1s, urchin embryonic growth factor, bone morphogenetic protein 1 (CUB) domain], with a “puppet” [thioester domain (TED)] dropped to the level of MG1-MG4-MG5. Compared to the structure of C4b alone (24), the C345C domain exhibits a slight rotation to accommodate C2a (fig. S3A). C2a binds to the C345C domain of C4b via its von Willebrand A domain (vWA) domain, while its serine protease (SP) domain swings toward C3. The substrate C3 adopts a characteristic inactive conformation, with the TED securely tucked beneath the CUB and MG8 domains, effectively concealing the Cys¹⁰¹⁰-Gln¹⁰¹³ thioester warhead (fig. S3B) (25). The upper half of C3, including the MG7-MG8-CUB-TED-C345C domains, rotates by 9° relative to the MG1-to-MG6 body compared with the corresponding region in the free human C3 structure, which enhances contact with C4b2a. Similar rotations are observed when compared to the bovine C3 structures (26). The C3a/ANA domain, consisting of four helices, points toward the SP domain of C2a, facilitating the docking of the scissile loop that contains the Arg⁷⁴⁸-Ser⁷⁴⁹ cleavage site into the catalytic center of C2a.

Substrate recognition mechanism of C4b2a

The structure of the C4b2a-C3 Michaelis complex illustrates the specific recognition mechanism used by the classical and lectin pathway C3 convertase. Both C4b and C2a are involved in C3 recognition, contributing to the stringent substrate specificity. The interface between C4b and C3 buries 2140 Ų from each molecule and involves multiple MG domains from both proteins (Fig. 2C). Specifically, C4b-MG4 interacts with C3-MG5, facilitated by residues Pro³⁸³_C4b, Pro⁴³⁴_C4b, and the Leu⁵¹⁴_C3-Pro⁵¹⁸_C3 strand (Fig. 2D). Between C4b-MG5 and C3-MG5, Pro⁴⁷⁸_C4b and Asp⁴⁷⁹_C4b form interactions with Leu⁴⁶¹_C3 and Thr⁴⁶³_C3, while Arg⁴⁸¹_C4b and Arg⁴⁹⁴_C4b engage with Glu⁴⁶⁹_C3 and Asp⁵¹³_C3, respectively. Additional interactions occur between C4b-MG5 and C3-MG4, where Arg⁵²⁴_C4b is surrounded by Asn⁴¹²_C3-Pro⁴¹⁵_C3, and Thr⁵²⁵_C4b-Leu⁵²⁶_C4b pack against Met³⁶⁸_C3-Pro³⁶⁹_C3.

The C4b-MG6 domain, which consists of two split subdomains (MG6α and MG6β), interacts with three domains in C3 (fig. S3C): Asp⁵⁷⁶_C4b forms a salt bridge with Lys²⁶⁴_C3 in C3-MG3; Glu⁵⁸⁵_C4b bonds with Arg⁸⁵⁵_C3 in C3-MG7; and Arg⁸²⁸_C4b interacts with Asp⁵⁷²_C3 in C3-MG6. In addition, C4b-MG7 and C3-MG7 align closely, with Arg⁸³¹_C4b, Glu⁸³²_C4b, and Leu⁸⁵⁹_C4b accommodating Phe⁹²⁰_C3. Overall, two-dozen hydrogen bonds and salt bridge contacts are formed between C4b and C3 (table S2), ensuring their specific interaction.

The interaction between C2a and C3 buries 1360 Ų from each molecule. The scissile loop of C3 that contains Arg⁷⁴⁸ (P1 residue; the residues on the N-terminal side of the SP scissile bond are numbered P3, P2, P1, etc.; whereas the residues on the C-terminal side are numbered P1′, P2′, P3′, etc.) exhibits clear densities and is positioned within the active site of C2a-SP (Fig. 2E). Arg⁵¹⁰_C2a contacts the C-terminal end of the α₄ helix in the C3a/ANA domain, facilitating the positioning of Leu⁷⁴⁴_C3 (P5 residue). Residues Glu⁵⁵⁶_C2a, Tyr⁵⁵⁸_C2a, and Lys⁶⁷⁶_C2a orient Gly⁷⁴⁵_C3-Leu⁷⁴⁶_C3-Ala⁷⁴⁷_C3 (P4 to P2 residues) through hydrogen bond interactions with their main-chain groups, and Met⁶⁴⁸_C2a further packs on Leu⁷⁴⁶_C3. Gly⁷⁰⁰_C2a-Asn⁷⁰³_C2a, located at the beginning of the long loop 2 (residues 700 to 726), accommodates Arg⁷⁴⁸_C3 (Fig. 2E). In addition, Lys⁶⁷⁶_C2a interacts with the main chain of Ser⁷⁴⁹_C3 (P1′ residue), while Lys⁴⁸⁷_C2a forms a salt bridge with Glu⁷⁵³_C3 (P5′ residue). Notably, in the absence of C3 binding, Lys⁶⁷⁶_C2a is drawn to Ser⁶⁹⁸_C2a, causing its main-chain carbonyl group to form a hydrogen bond with Ser⁶⁷⁹_C2a, which distorts the oxyanion hole (27). Upon C3 binding, Lys⁶⁷⁶_C2a interacts with Ala⁷⁴⁷_C3 and Ser⁷⁴⁹_C3, leading to the establishment of a functional oxyanion hole (fig. S3D). In addition to engaging the scissile loop of C3, C2a also anchors to the C3-MG3 domain: A pocket formed by Lys⁴⁸⁵_C2a, Leu⁵¹⁷_C2a, Arg⁵¹⁹_C2a, and Glu⁵³²_C2a accommodates Arg²⁹⁰_C3, while Trp⁵²⁹_C2a packs with His³³³_C3 (Fig. 2C and fig. S3E). As a result of these interactions with both C2a and C4b, C3-MG3, along with the upper tip of C3-MG4, undergoes a ~25° rotation within the MG1-to-MG6 ring compared to free C3 (fig. S3B). Collectively, these interactions ensure the precise positioning of the Arg⁷⁴⁸_C3-Ser⁷⁴⁹_C3 peptide bond, priming it for cleavage by the Asp⁵⁶¹_C2a-His⁵⁰⁷_C2a-Ser⁶⁷⁹ (Ala)_C2a catalytic triad.

Formation of the classical and lectin pathway C3 convertase

We also captured the structure of the C4b2 proconvertase at the loading and activation states. Together with the C4b2a structure in complex with the C3 substrate, these structures allow us to visualize the formation of the C3 convertase at discrete steps (Fig. 3, A and B).

Fig. 3. Formation of the classical and lectin pathway C3 convertase.

(A) The classical and lectin pathway C3 convertase at three distinct steps: loading, activation, and C3 engaged. Volumes in the cryo-EM map corresponding to C4b, C2b (CCP1 to CCP3), and C2a (vWA-SP) are depicted in purple, salmon, and green, respectively. In the loading state, the C2a-SP domain exhibits flexibility and cannot be modeled and is represented by a dashed circle. (B) Ribbon diagrams of the C4b2 proconvertase and C4b2a convertase structures depicted in the same orientations as in (A). (C) The positions of C4b-Val¹⁷⁴⁴ (the C-terminal end residue of C4b) relative to the Ni²⁺ ion in C2-vWA. The three density maps are shown at the same contour level for direct comparison. Ser²⁶², Ser²⁶⁴, Thr³³⁷, and Asp³⁷⁶ in the metal ion–dependent adhesion site (MIDAS) motif of C2-vWA are highlighted in orange. (D) Structural comparison between the loading and activation states of C4b2. The loading state is presented in gray, while the activation state uses the same color scheme as in (A). (E) Structural models of C1s or MASP2 bound to C2. These models are generated by aligning the activation state of the C4b2 structure with that of C3bBD [Protein Data Bank (PDB): 2XWB] and then aligning the structures of C1s (PDB: 1ELV) and MASP2 (PDB: 5JPM) to FD. (F) Structural comparison between the activation and C3-engaged states. The C3-engaged state is shown in gray, while the activation state is presented in color.

The loading state of the C4b2 structure exhibits some similarities to the FB complexed with CVF (fig. S4A) (28). Although the overall map is at 2.9 Å, the C2-SP domain remains flexible and could not be modeled. A lower-resolution map, reconstructed from a separate dataset acquired on a 200-kV microscope, indicates that the general position of C2-SP is likely near the vWA domain (fig. S4B), similar to the CVF-B complex. In the 2.9-Å structure, the C2–von Willebrand factor (vWF) region [complement control protein 1 to 3 domains (CCP1 to CCP3)] and the C2-vWA domain interact with C4b. In particular, C2-CCP1 and C2-vWA together engage the C4b-C345C domain. Notably, akin to the interaction observed between CVF and FB, the carboxyl group of the terminal Val¹⁷⁴⁴_C4b coordinates with the Ni²⁺ ion in the metal ion–dependent adhesion site (MIDAS) of C2-vWA, completing the metal ion coordination (Fig. 3C). In addition, the CCP2-to-CCP3 domains contact C4b-MG2, C4b-MG6, C4b-MG7, C4b-CUB, and the C4b–α′ chain N-terminal (α′-NT) region (fig. S4C).

The activation state of the C4b2 structure closely resembles the C3bB structure (fig. S4D), with an overall root mean square deviation (RMSD) of 1.4 Å, and provides a complete visualization of C2 (Fig. 3, A and B). In this state, C2-SP adopts a stable conformation and is anchored to the MG2 and CUB domains of C4b. In addition, the vWA-SP linker in C2 is more ordered and is positioned between C2-vWA and C2-SP (fig. S4E). This change pulls C2-vWA closer to C2-SP, which, in turn, induces a rotation of the C4b-C345C domain. As a result, the interaction between C2-CCP1 and C4b-C345C is reduced (Fig. 3D and fig. S4F). This displacement likely contributes to the release of C2b prosegment following C2 cleavage. Models of C1s or MASP2 can be generated on the basis of FD, using the C3bBD complex structure as a template (Fig. 3E). In the resulting models, the scissile loop would be readily accommodated in the active sites of these proteases, facilitating C2 cleavage and C4b2 convertase maturation.

In the C4b2a-C3 complex, C2a adopts a conformation closely resembling that of the C2a alone structure (fig. S4G), with the C2a-SP drawn toward the substrate C3 as described above. Consequently, a tectonic reorientation of C2a is observed when compared to the C2a region in the activation state of C4b2 (Fig. 3F and movie S1). The C2a-vWA domain shifts into the position previously occupied by CCP1 to CCP3, while the C2a-SP undergoes a ~118° swing. Thus, the CCP1-to-CCP3 prosegment must be released for the activated C2a to effectively dock onto C3. The large swing of the C2a-SP is further facilitated by a ~63° rotation of the α₇ helix in the C2a-vWA domain, accompanied by a complete reorientation of the vWA-SP linker. The C2a-vWA domain remains attached to the C4b-C345C domain. Notably, in the C3-bound state, the carboxyl group of Val¹⁷⁴⁴_C4b appears to be positioned further from the Ni²⁺ ion in the MIDAS motif of C2a-vWA, as indicated by the density map (Fig. 3C). This suggests that engagement of C2a to C3 may prompt the concurrent release of C2a from C4b, providing some explanation for the labile nature of the C4b2a convertase.

Cryo-EM structure of the C3bBbP-C3 complex

To investigate C3 recognition by the alternative pathway C3 convertase, we generated C3b from C3 through limited trypsin processing (29). A catalytically inactive FB containing the S699A mutation was prepared using HEK293 cells. Properdin is an oligomeric protein that forms a mixture of dimers, trimers, and tetramers (30–32). We produced a two-chained monomeric properdin by inserting a tobacco etch virus (TEV) protease cleavage site between its TSR3 and TSR4 domains (Fig. 1B) and then processing the purified properdin oligomer with TEV, as previously described (33). Using a similar approach to the generation of the C4b2a-C3 complex, we mixed purified C3b, FB, properdin, and C3 in equal molar ratios with Ni²⁺. The mixture was then briefly treated with FD, converting FB into the active fragment Bb. Subsequently, cryo-EM analysis was conducted, resulting in the structure determination of the C3bBbP-C3 complex at an overall resolution of 2.6 Å (Fig. 4, A and B; fig. S5; and table S1).

Fig. 4. Cryo-EM structure of the alternative pathway C3 convertase.

(A) Two views of the C3bBbP-C3 complex cryo-EM reconstruction. C3b, Bb, and C3 are depicted in khaki, light blue, and orange, respectively, while properdin is shown in aquamarine. C3a/ANA is highlighted in red. (B) Ribbon rendering of the C3bBbP-C3 structure accompanied by a schematic representation. (C) Interaction between Bb-SP and C3, with Bb-SP, C3a/ANA, α′-NT, C3-MG3, and C3-MG8 shown in light blue, red, orange, cyan, and dark green, respectively. The β-hairpin in Bb-SP is highlighted in pink. (D) The β-hairpin in Bb-SP is longer than that of C2a-SP, inducing the formation of a short strand in the C3 α′-NT region. The bottom panel shows the density map of the strand region in Bb-SP. (E) Interaction between the C3 scissile loop and Bb-SP, including enlarged views of the density map within the catalytic pocket on the top and detailed molecular interactions on the right. C3a/ANA is shown in red and α′-NT in orange. The cleavage site is indicated by a green arrowhead. The oxyanion hole in Bb-SP is highlighted in green, and the catalytic triad Ser⁶⁹⁹(A), His⁵²⁶, and Asp⁵⁷⁶ are highlighted in yellow. The hydroxyl group of Ser⁶⁹⁹ is depicted as hypothetical and shown in a lighter color.

In this complex, C3b and Bb individually closely resemble their respective structures observed in the C3bBb-SCIN complex; however, it is apparent that Bb can now properly latch onto the substrate C3 without the steric hindrance imposed by SCIN (fig. S6A). The C3 molecule adopts a conformation nearly identical to that seen in the C4b2a-C3 complex (fig. S6B), with an RMSD of 0.9 Å. Similar to the C4b2a-C3 complex, both C3b and Bb engage the C3 molecule. C3b forms a quasidimer with C3, involving the MG3-to-MG7 domains of both proteins (fig. S6C). Compared to the C4b-C3 interface in the C4b2a-C3 structure, the C3b-C3 interaction involves fewer hydrogen bonds and salt bridge interactions (table S2), and the interface area is slightly smaller (2040 Ų versus 2140 Ų), suggesting that C4b forms a somewhat more optimized interaction with C3 than C3b.

In contrast, the interaction between Bb and C3 exhibits unique features and appears enhanced compared to the C2a-C3 interaction. Compared to the C3bB structure, a large swing of Bb-SP is observed (fig. S6D), similar to that seen with C2a-SP describe above. The β-hairpin comprising the first two β strands in the Bb-SP domain (Gln⁴⁹⁶_Bb-Val⁵¹⁵_Bb) is longer than the corresponding region in C2a-SP (His⁴⁸¹_C2a-Leu⁴⁹⁶_C2a) and effectively packs onto Asn⁷⁵⁰_C3-Asp⁷⁵²_C3 (P2′ to P4′ residues) by inducing the formation of a short strand (Fig. 4, C and D). A hydrophobic pocket formed by Pro⁵⁷⁰_Bb and Ala⁶⁶³_Bb-Tyr⁶⁶⁶_Bb accommodates Leu⁷⁴⁴_C3 and Leu⁷⁴⁶_C3 (Fig. 4E). Gly⁷²⁰_Bb-Asp⁷²³_Bb at the beginning of loop 2 (residues 720 to 739) holds Arg⁷⁴⁸_C3 in a similar manner as Gly⁷⁰⁰_C2a-Asn⁷⁰³_C2a (Fig. 2E), with the negatively charged Asp⁷²³_Bb better coordinating Arg⁷⁴⁸_C3 compared to Asn⁷⁰³_C2a. The Asp⁵⁷⁶_Bb-His⁵²⁶_Bb-Ser⁶⁹⁹ (Ala)_Bb triad adopts nearly identical positions to Asp⁵⁶¹_C2a-His⁵⁰⁷_C2a-Ser⁶⁷⁹(Ala)_C2a, ready to attack the Arg⁷⁴⁸_C3-Ser⁷⁴⁹_C3 peptide bond.

Besides targeting the α′-NT segment of C3 that contains the cleavage site, Bb also contacts C3-MG3 and C3-MG8. For example, Ile⁵⁰²_Bb forms hydrophobic interactions with Phe²⁷⁰_C3 and Leu²⁸⁸_C3 in C3-MG3 (Fig. 4C and fig. S6E). Pro⁵⁰⁴_Bb-Ser⁵⁰⁵_Bb at the tip of the β1-and-β2 hairpin forms van der Waals interactions with Pro²⁸⁵_C3-Glu²⁸⁶_C3. Asp⁵⁴⁸_Bb coordinates Arg²⁹⁰_C3. Several polar residues in loop 2 of Bb, including Lys⁷²⁶_Bb and Asn⁷²⁷_Bb, form polar interactions with residues in C3-MG8, such as Glu¹³⁷²_C3 and Arg¹³⁷⁶_C3. These exosite interactions further ensure the specific recognition of C3 by Bb.

Stabilization of C3bBb by properdin

Focused refinement enabled clear visualization of the TSR5 and TSR6 domains of properdin that interacts with C3bBb (fig. S5). Previous structures have been determined for properdin in complex with the C-terminal C345C domain of C3b (34) and in complex with the C3bB proconvertase (35, 36). In our structure, properdin binds to C3bBb in a manner largely similar to its interaction with C3bB, with contacts predominantly to C3b. The properdin-C3b interaction buries a surface area of 950 Ų, much larger than the 210-Ų properdin-Bb interface. Notably, two critical loops in properdin, TSR5-stirrup (residues 328 to 333) and TSR6-stirrup (residues 419 to 426), insert into the C3b-Bb interface to enhance convertase stability (Fig. 5A). Key bridging residues include Arg³²⁹_P in the TSR5-stirrup, which stacks against Phe¹⁶⁵⁹_C3b and also forms a hydrogen bond with the main-chain oxygen of Leu³⁴⁹_Bb (Fig. 5B and fig. S7A). In addition, Ile³⁴⁰_P and Pro³⁴¹_P engage Leu³⁴⁹_Bb via hydrophobic interactions. In the TSR6-stirrup, Ser⁴¹⁹_P and Met⁴²⁰_P interact with Tyr³¹⁷_Bb and Lys³⁵⁰_Bb through main-chain groups, while Val⁴²¹_P forms a hydrophobic contact with Met³⁹⁴_Bb. Together, these interactions from properdin effectively secure Bb to C3b, stabilizing the alternative pathway C3 convertase.

Fig. 5. Stabilization of C3bBb by properdin.

(A) The TSR5-stirrup and TSR6-stirrup loops of properdin promote the interaction between C3b and Bb. Bb-vWA, C3b-C345C, and properdin are shown in light blue, khaki, and aquamarine, respectively. (B) Key interactions between properdin and C3bBb. (C) Conformational change of Arg³³⁰_P. Relative to the C3bBP proconvertase complex (gray; PDB: 7NOZ), Arg³³⁰_P adopts a markedly different conformation to facilitate an interaction between C3b and the metal ion in the MIDAS motif of Bb-vWA.

Compared to the C3bBP proconvertase complex (36), a notable difference is seen with properdin Arg³³⁰_P in our C3bBbP structure (Fig. 5C). In both complexes, Arg³³⁰_P is attached to Asp¹⁵³⁴_C3b in the C3b C345C domain. In C3bBP, it also approaches Ser⁷⁸_Ba in the CCP1 domain, likely contributing to stabilize the proconvertase before cleavage and release of the vWF/Ba prosegment. In our mature C3bBbP convertase structure that lacks the vWF/Ba prosegment, Arg³³⁰_P undergoes a marked rotation and packs against Pro¹⁶⁶²_C3b, the penultimate residue in C3b (Fig. 5C and fig. S7B). This arrangement helps anchor the Pro¹⁶⁶²_C3b-Asn¹⁶⁶3_C3b segment, thereby sustaining the interaction of terminal Asn¹⁶⁶³_C3b with the metal ion (Ni²⁺ here) in the MIDAS motif of Bb-vWA. In this way, properdin further stabilizes the mature C3bBb convertase.

DISCUSSION

The complement system is a central arm of immunity and is implicated in many diseases. Central to its activation are the C3 convertases: C4b2a of the classical and lectin pathways and C3bBb of the alternative pathway, which cleave C3 to amplify the complement cascade. In this study, we elucidated the molecular mechanisms governing these enzyme assemblies. In particular, we determined high-resolution structures of two Michaelis complexes: C4b2a bound to C3 and C3bBbP bound to C3. These structures reveal how C3 convertases recognize their substrate, offering key insights with broad implications for complement biology and therapeutics.

First, our findings provide a framework for interpreting complement-related data from animal models. Sequence alignments demonstrate high conservation of C3, C4, C2, and FB across humans, mice, and rats (fig. S8), with most interface residues in the C4b-C3, C2a-C3, C3b-C3, and Bb-C3 contacts preserved. However, C4 exhibits greater similarity between humans and rats than between humans and mice. Notably, several residues at the C4 MG4-to-MG5 domain interface with C3 in the C4b2a-C3 complex, such as Pro³⁸³_C4, Arg⁴⁹⁴_C4, Arg⁵²⁴_C4, and Leu⁵²⁶_C4, are more conserved in rats. This suggests that rat models may better recapitulate human classical and lectin pathway convertase function compared to mouse models, potentially informing species selection in preclinical studies.

These structures also shed light on the mechanisms of C5 cleavage, a critical step in forming the membrane attack complex. Once C3b deposition reaches a threshold density through C3 convertase activity, C5 convertases assemble via association of additional C3b molecule(s) with C4b2a or C3bBb. Although the precise architecture of C5 convertases remains elusive, the structural homology between C5 and C3, as well as between CVF and C3b/C4b, enabled us to model C5-bound Michaelis complexes using the CVF-C5 structure (19) alongside our C4b2a-C3 and C3bBbP-C3 complexes (fig. S9). These models are consistent with affinity measurements: C3b and C4b bind C5 with dissociation constant values around 1 μM, similar to their affinities for C3 (37). However, the models also indicate that the C5a/ANA domain does not align perfectly with the catalytic pockets of C2a or Bb, implying that conformational rearrangements in C5, the convertase, or both are required for cleavage. Additional unresolved questions include the exact positioning of additional C3b molecule(s) in C5 convertase formation and the potential involvement of properdin. Furthermore, the monomeric properdin variant used in this study represents its minimal functional unit for stabilization. In vivo, properdin oligomers provide multivalent binding sites primarily for C3b, localizing C3bBb and free C3b to increase their local density. This organization could facilitate C5 convertase assembly, warranting further investigation into properdin’s role in terminal pathway activation.

Last, our structures lay a foundation for therapeutic strategies targeting C3 convertases in complement-driven diseases. For instance, pegcetacoplan, a compstatin-family cyclic peptide approved for paroxysmal nocturnal hemoglobinuria (PNH) and geographic atrophy (38, 39), binds at the MG4-MG5 interface in C3 or C3b (40, 41). Overlaying this interaction onto our C4b2a-C3 and C3bBbP-C3 complexes shows that compstatin sterically blocks convertase engagement, inhibiting activation across all pathways (Fig. 6A). Iptacopan, an FB inhibitor approved for PNH, C3 glomerulopathy, and immunoglobulin A nephropathy, directly occludes the Bb active site by blocking the substrate’s α′-NT segment, as revealed by comparing our C3bBbP-C3 structure with the Bb-SP-iptacopan complex (Fig. 6B) (42). These insights could guide the refinement of next-generation inhibitors. Inhibiting C2 could also be beneficial for treating diseases driven by classical or lectin pathway activation (43). Given iptacopan’s success, small-molecule C2 inhibitors could be developed, and our C4b2a-C3 structure provides a blueprint for their rational design.

Fig. 6. Insights into therapeutic drugs targeting the C3 convertase.

(A) Mechanism of CP40 inhibition on the engagement of C3 with C4b2a and C3bBb. The C4b2a-C3 inhibition model is generated by superimposing C3b-CP40 (PDB: 7BAG) onto the C3 molecule in the C4b2a-C3 complex (left panel). The C3bBbP-C3 inhibition model is generated by superimposing C3b-CP40 onto both C3b and C3 (right panel). The binding of CP40 would introduce steric clashes that hinder C3 binding to either C4b2a or C3bBb. (B) Mechanism of Iptacopan inhibition on the engagement of C3 with C3bBbP. The inhibition model is generated by superimposing Bb-Iptacopan (PDB: 6RAV) onto the structure of Bb-C3a/ANA-α′-NT, which is separated from the C3bBbP-C3 complex. Iptacopan is represented in space-filling format.

In summary, we delineated the mechanisms by which C4b2a and C3bBb recognize C3. Our results highlight distinct substrate-binding characteristics of the classical and lectin and alternative convertases and elucidate the unique role of properdin in the alternative pathway. These findings provide a fundamental understanding of these essential molecular machineries and will guide the design and optimization of future therapeutic molecules aimed at modulating complement activation.

MATERIALS AND METHODS

Cell line

HEK293F cells (Thermo Fisher Scientific) were cultured in FreeStyle 293 Expression Medium (Thermo Fisher Scientific) at 37°C with 5% CO₂ and 55% humidity using a humidified shaker for protein purification.

Human C3 and C4 purification

Human C3 and C4 were purified from plasma provided by Peking University First Hospital Renal Department, with prior ethical approval (protocol no. 2017[1280]). For C3 purification, Na₂SO₄ powdered salt crystals (Sigma-Aldrich) were added to the plasma to a final concentration of 10% (w/v) to precipitate high–molecular-weight (M_w) components. Following centrifugation and dialysis of the supernatant against the DEAE buffer, the protein solution was loaded onto a DEAE ion exchange column for the initial separation step. The C3-containing fraction was then dialyzed against Mono S column buffer and applied to a Mono S column to remove C3 (H₂O). The eluted protein was concentrated using a 10,000–molecular-weight cutoff concentrator (Millipore) and further purified by size exclusion chromatography using a Superdex 200 Increase column. The C3 fraction was collected, centrifuged, and lastly stored in 25 mM Hepes-NaOH buffer (pH 7.4) with 100 mM NaCl.

For C4 purification, trisodium citrate was added to the plasma to a final concentration of 25 mM, followed by the slow addition of BaCl₂ solution to reach a concentration of 60 mM. The resulting precipitate was removed by centrifugation, and the supernatant was applied to a Q FF ion exchange column. The pooled eluent was treated with a 50% polyethylene glycol, molecular weight 6000 solution to induce protein precipitation. After redissolving the protein, the solution was loaded onto a Q HP column. The fractions containing C4 were collected and further purified using a Mono Q column. Last, C4 was subjected to gel filtration using a Superdex 200 Increase column for the final purification step and stored in 25 mM Hepes-NaOH buffer (pH 7.4) containing 150 mM NaCl.

Generation of C3b and C4b

C3b was generated by limited trypsin processing of C3 as previously described (29). First, trypsin was added in a trypsin:C3 mass ratio of 1:40 to cleave C3 at 37°C for 6 min. Subsequently, sample was transferred on ice immediately, and soybean trypsin inhibitor was added to stop the reaction. C3b sample was subjected to gel filtration using a Superdex 200 Increase column for the final purification step and stored in 25 mM Hepes-NaOH buffer (pH 7.4) containing 100 mM NaCl. C4b was produced by adding active C1s (Complement Technology, catalog no. A104) to C4 solution in mass ratio of 1:100, incubated first at 37°C for 3 to 4 hours and then overnight at 4°C.

Recombinant protein expression and purification

For recombinant expression of C2, FB, and FD in HEK293F cells, the genes encoding full-length C2 (UniProt, P06681), FB (UniProt, P00751), and FD (UniProt, P00746) were cloned into the pcDNA3.1 vector with an 8× His-tag. For obtaining inactivated FB and C2 variants, site-directed mutagenesis was performed to introduce the S699A mutation in FB and the S679A mutation in C2.

HEK293F cells were cultured to a density of 1 × 10⁶ cells/ml and transfected with plasmids encoding the target proteins. For a 1-liter cell culture, 1 mg of plasmid DNA was mixed with 2 mg of 40-kDa linear polyethyleneimine (Polysciences) in 50 ml of fresh medium and incubated for 20 min before transfection. The transfected cells were cultured for 4 days, after which the supernatant was collected by centrifugation. The medium was then exchanged with purification buffer [25 mM tris-HCl (pH 7.4) and 150 mM NaCl] using a Hydrosart ultrafilter (Sartorius). Recombinant proteins were isolated by Ni–nitrilotriacetic acid affinity chromatography and eluted with purification buffer supplemented with 300 mM imidazole. The eluted proteins were further purified by size exclusion chromatography using a Superdex 200 Increase column and stored in 25 mM Hepes-NaOH buffer (pH 7.4) with 150 mM NaCl.

A TEV site was introduced between residue P255 and V256 in properdin (UniProt, P27918). Subsequently, this properdin-coding DNA was cloned into pcDNA3.1 vector with an 8× His-tag. After expression and purification from HEK293F cells, TEV protease was added in purified properdin in a TEV:protein mass ratio of 1:20 and incubated for 24 hours at 4°C. The cleaved protein was further purified by size exclusion chromatography using a Superdex 200 Increase column and stored in 25 mM Hepes-NaOH buffer (pH 7.4) with 150 mM NaCl.

Cryo-EM sample preparation and data collection

To prepare the classical pathway C3 convertase-substrate complex, C4b, C2(S679A) were mixed at a 1:1 molar ratio in binding buffer [25 mM Hepes-NaOH (pH 7.4), 100 mM NaCl, and 2 mM NiCl₂], and C1s was then added in a C2:C1s mass ratio of 1:20. After incubation for 5 min at room temperature, the mixture was cooled to 4°C, and an equimolar amount of C3 was added. After a 30-min incubation on ice, the solution was concentrated to a final protein concentration of 1 mg/ml using a 10,000–molecular-weight cutoff concentrator (Millipore).

To prepare the alternative pathway C3 convertase-substrate complex, C3b, CFB(S699A) and properdin-TEV were mixed at an equal molar ratio in binding buffer [25 mM Hepes-NaOH (pH 8.0), 100 mM NaCl, and 2 mM NiCl₂]. Subsequently, FD was added in an FB:FD mass ratio of 1:20. After incubation for 5 min at room temperature, the mixture was then cooled to 4°C, and an equimolar amount of C3 was added. After a 30-min incubation on ice, the solution was concentrated to a final protein concentration of 0.7 mg/ml using a 10,000–molecular-weight cutoff concentrator (Millipore).

The sample preparation was carried out using a Vitrobot Mark IV (FEI). To prepare the C4b2b-C3 cryo-EM sample, 4 μl of the sample was deposited onto glow-discharged holey-carbon gold grids (Quantifoil Au 300 mesh R 0.6/1), which had been treated for 40 s (H₂ and O₂) using a Solarus Model 950 Plasma Cleaner (Gatan). The grids were blotted for 2.5 s with a blotting force of −1 at 4°C and 100% humidity and then plunged into liquid ethane cooled by liquid nitrogen. For the C3bBbP-C3 sample, 4 μl of the sample was deposited onto glow-discharged holey-carbon grids with continuous carbon film (Quantifoil Cu 300 mesh R 1.2/1.3 with 2 nm C), which had been treated for 40 s (H₂ and O₂) using a Solarus Model 950 Plasma Cleaner (Gatan). The grids were blotted for 2 s with a blotting force of −1 at 4°C and 100% humidity and then plunged into liquid ethane cooled by liquid nitrogen.

All grids were initially screened using a 200-kV Talos Arctica microscope equipped with a Falcon 4 camera. Data collection was performed using EPU software (Thermo Fisher Scientific) on a 300-kV Titan Krios G4 microscope equipped with a Falcon 4 camera.

Cryo-EM data processing

Data for determining the structures of C4b2a-C3, C4b2 (activation), and C4b2 (loading) were from the same dataset. A total of 14,681 movies was collected and processed by CryoSPARC (version 4.4.1) (44). After Patch Motion Correction and Patch CTF Estimation, we manually curated the micrographs by exposure curation to remove low-quality images. We then performed template-free particle picking with the Blob Picker and generated templates by subsequent two-dimensional (2D) classification. Using these templates, we conducted template-based picking followed by multiple rounds of 2D classification to discard low-quality particles. Particles from 2D classes corresponding to the C4b2a-C3 complex were selected to train a Topaz model, which we then applied to pick particles with higher accuracy. Several rounds of heterogeneous refinement enriched for C4b2a-C3 particles while removing poor-quality and C4b2-only particles. The resulting C4b2b-C3 particle set was subjected to homogeneous refinement and local refinement to produce the final 3D reconstruction. The C4b2 (loading state) and C4b2 (activation state) particles were processed in a similar manner. To obtain higher-quality local density maps, mask-based local refinements were further performed using CryoSPARC.

For the C3bBbP-C3 complex, 8649 movies were collected and processed in CryoSPARC following the same workflow used for C4b2b-C3. After Patch Motion Correction, Patch CTF Estimation, and exposure curation, particles were initially picked and classified; selected classes were then used to train a Topaz model, which was applied to refine particle picking. Several rounds of heterogeneous refinement enriched for high-quality C3bBbP-C3 particles, which were subsequently subjected to homogeneous and local refinements to yield the final 3D reconstruction at 2.57-Å global resolution. To improve local features, we performed mask-based 3D classification focused on C3b-TED and on properdin, producing higher-quality local maps at 2.84 and 2.73 Å, respectively.

Model building and structure refinement

Initial models of C4b, C3, and C2 were generated using AlphaFold2 (45). Models of C2a, C3b, Bb, and properdin were obtained from previous structure [Protein Data Bank (PDB) ID: 2ODP, 2WIN, and 6S0B]. These models were docked into the cryo-EM density map using UCSF Chimera (46). Further structure model building was performed using Coot (47). No water molecules were built. Structural refinements were carried out using the real-space refinement in PHENIX (48) with secondary structure and geometry restraints. The maximum distance for metal ion coordination is set at 3.5 Å (default in PHENIX). The final model was evaluated and validated in PHENIX. Structural analyses were carried out using Coot and PISA (49). Figures were prepared with UCSF ChimeraX (50).

Supplementary Materials

The PDF file includes:

Figs. S1 to S9

Tables S1 and S2

Legend for movie S1

Other Supplementary Material for this manuscript includes the following:

Movie S1