X-ray Crystal Structure of C3d: A C3 Fragment and Ligand for Complement Receptor 2

See allHide authors and affiliations

Science  22 May 1998:
Vol. 280, Issue 5367, pp. 1277-1281
DOI: 10.1126/science.280.5367.1277


Activation and covalent attachment of complement component C3 to pathogens is the key step in complement-mediated host defense. Additionally, the antigen-bound C3d fragment interacts with complement receptor 2 (CR2; also known as CD21) on B cells and thereby contributes to the initiation of an acquired humoral response. The x-ray crystal structure of human C3d solved at 2.0 angstroms resolution reveals an α-α barrel with the residues responsible for thioester formation and covalent attachment at one end and an acidic pocket at the other. The structure supports a model whereby the transition of native C3 to its functionally active state involves the disruption of a complementary domain interface and provides insight into the basis for the interaction between C3d and CR2.

Serum complement protein C3 is a central component of host defense because its proteolytic activation is the point of convergence of the classical, alternative, and lectin pathways of complement activation. C3 cleavage products mediate many of the effector functions of humoral immunity, including inflammation, opsonization, and cytolysis. Proteolytic cleavage of C3 into C3a and C3b exposes an internal thioester bond that through transacylation mediates covalent attachment of C3b to the surface of foreign pathogens (1). Although surface-bound C3b is itself a ligand for complement receptor 1 (CR1; also known as CD35), it can subsequently be degraded into the successively smaller fragments iC3b and C3dg, tagging the pathogen for recognition by other receptors, including the B cell complement receptor CR2 (CD21) (1). The interaction between B cell CR2 and antigen-bound iC3b or C3dg is an essential component of a normal antibody response (2), making an important link between the innate and adaptive arms of the immune system (3). The C3d fragment (a CR2-binding, 35-kD, protease-resistant fragment of C3dg), when conjugated to a soluble antigen, is a potent molecular adjuvant (4).

A Cys1010 → Ala1010 mutant form of human C3d (C3 residues 996 through 1303; prepro-C3 numbering used throughout) was expressed in Escherichia coli in both its native and selenomethionyl-labeled forms. The alanine mutation was introduced to ensure correct disulfide bonding between Cys1101 and Cys1158 (Cys1010 contributes the thiol moiety of the thioester, which would only be expected to form in native C3). A rosette inhibition assay confirmed that the recombinant C3d fragment was functionally active and bound CR2 to an extent comparable to that of serum-derived C3dg (5). Crystals were grown and the structure was determined by multi-wavelength anomalous diffraction (MAD) phasing from data collected at the selenium absorption edge (6). The structure was refined to anR cryst of 19.3% and anR free of 23.3% for the data in the 5.0 to 1.8 Å resolution range (Table 1). The final model contains residues 1 through 294 (vector residues methionine and leucine plus C3 residues 996 through 1287) of the 310–amino acid construct, as well as 110 solvent molecules.

Table 1

Structure determination and refinement. The structure was determined by the MAD phasing method. Three data sets were collected at the Se absorption edge corresponding to the peak, edge, and a remote wavelength. Atomic positions for the 8 Se atoms were obtained by means of Patterson methods, with the aid of the program HEAVY (24), and difference Fourier maps. The three data sets were then treated as a multiple isomorphous replacement problem, using the remote wavelength as “native” and the peak and edge wavelengths as “derivatives.” The selenium positions were refined and phases were calculated with the program PHASES (25). The resulting experimental map was of very high quality, with an overall figure of merit of 0.8 to 2.0 Å (0.9 after solvent flattening) and was traced with the program O (26). The initialR cryst and R free values (7% of the reflections set aside for the test set) of the model before refinement were 37.1 and 37.4%, respectively, in the 8.0 to 1.8 Å resolution range against the native data set. The model was then refined with the program X-PLOR (27). rsmd, root mean square deviation. After the model was subjected to simulated annealing, successive rounds of manual rebuilding, and positional/B-factor refinement, as well as the application of an overall anisotropic B factor to the data, theR cryst and R free values were reduced to 19.3 and 23.3%, respectively (the 5.0 to 1.8 Å range). The final model contained 294 residues, 104 water molecules, and six glycerol molecules. Electron density for the six NH2-terminal residues (1 through 6), residues 42 through 46, residue 143, and residues 166 through 171 was weak. No electron density was present for residues 295 through 310. Relative residue accessibility was calculated with the program NACCESS (28), secondary structure was assigned with the use of the program PROMOTIF (29), and structural similarity was searched for with the program DALI (30). The multiple sequence alignment was performed with ClustalW (27), and conservation indices were calculated with AMAS (32).

View this table:

C3d is an α-α barrel with overall dimensions of approximately 30 Å × 30 Å × 50 Å (Fig. 1, A and B). This α-barrel structure, although uncommon, has been seen in glucoamylase, endoglucanase, and the β subunit of protein farnesyltransferase (7). C3d shows no significant sequence similarity with these proteins. The topology of the molecule is such that consecutive helices alternate from outside to inside, resulting in a core of six parallel helices (α1, α3, α5, α8, α10, and α12), surrounded by a second set of six parallel helices (T1, α2, α4, α7, α9, and α11) running antiparallel to those of the core (Fig. 1A). The most NH2-terminal helix (T1) is one of five (T1 through T5) short segments of 310 helix found in the structure. Although the core of the protein is formed mainly by apolar side chains, there are several polar residues pointing into the core, hydrogen bonded to buried water molecules. The two opposing ends of the barrel are structurally distinct (Fig. 1B): One is a convex surface presenting the amino acid side chains responsible for covalent attachment to antigen, whereas the other is a more concave surface containing an extended acidic pocket.

Figure 1

C3d structure. (A) Ribbon representation of C3d. The single intrachain disulfide bond is shown in green here and in (B). (B) Ribbon representation viewed into the side of the α-α barrel. Shown are the residues responsible for covalent attachment to antigen: Ala17(Cys17 in wild-type C3d), Gln20, and His133 on the convex end of the molecule. Additional residues have been labeled to serve as reference points. Red and blue label oxygen and nitrogen atoms, respectively, here and in (C). (C) The thioester region. Superimposed in green are the modeled Cys17 and Gln20 with an intact thioester bond. The loop containing the thioester residues is stabilized by helix α2, well packed within the core of the protein, as well as stacking interactions with Phe66, making it unlikely that it undergoes much movement relative to the α-α barrel framework. In contrast, His133 is located at the midpoint of a long surface-exposed loop (residues 119 through 145 connecting helices α4 and α5), which caps the barrel. The conformational change required to bring about the open conformation seen in this structure is more likely to occur in this region. Apolar interactions with Ile132, located on the tip of the His133-containing loop, suggest a direct means of influencing the conformation of this loop in native C3. Glu19 forms the base of a pocket (see also Fig. 2B) that may accommodate the His133 side chain in the closed conformation. Although C3 and C4B share a similar reaction mechanism (9), the former reacts primarily with hydroxyl nucleophiles, whereas the latter shows considerable reactivity with amino nucleophiles as well. These differences have been proposed to reflect an increased rate of acyl-imidazole formation in C3 (1). Such a rate increase could be explained by the fact that His133 is hydrogen bonded to negatively charged Glu135 in C3, rendering it a stronger nucleophile; in C4B, the residue equivalent to Glu135 is a serine, which is probably unable to perform a similar function. [The figure was prepared with MOLSCRIPT (33) and RASTER3D (34).]

C3 attaches to pathogens covalently (8), predominately through ester linkages. To ensure covalent attachment to only nonself elements, an activated short-lived form of C3 is generated in close proximity to the cell surface of pathogenic microorganisms (1). Mechanistically, activation is triggered by proteolytic cleavage of C3, which initiates a series of steps (9) that closely parallel those elucidated for the B isotype of human C4 (10). Key among these steps is nucleophilic attack on the buried thioester linkage (formed by the side chains of Cys1010 and Gln1013) by His1126. The acyl-imidazole-activated Gln1013 side chain is then exposed to solvent, where it can be attacked by cell surface nucleophiles (for example, hydroxyl groups from glycolipids or glycoproteins of pathogens), leading to covalent attachment of C3b to antigen. In a competing reaction, the acyl-imidazole intermediate is rapidly hydrolyzed, thereby limiting the lifetime of the transacylation-competent state. The residues involved in forming the thioester bond [Cys17 (1010 in C3; here mutated to Ala17) and Gln20 (1013 in C3)] and the acyl-imidazole intermediate [Gln20 and His133(1126 in C3)] are surface-exposed in C3d and located on the convex surface of the molecule (Fig. 1, B and C). The Ala17residue is situated on the βturn preceding α1, and Gln20 is the NH2-terminal residue of α1, positioning them in close proximity. Although the side chain of Gln20 is somewhat disordered and Cys17 has been mutated to alanine in this structure, modeling shows that the cysteine sulfur and the carbonyl carbon of the glutamine side chain can be brought to within 2.0 Å of each other (close to a typical S–C bond length) with little movement of the peptide backbone (Fig. 1C). It would appear, then, that the relative disposition of these residues is probably close to that found in native C3. In contrast, His133 is misoriented and too far away from the modeled thioester linkage (4.1 Å) to form the acyl-imidazole intermediate with Gln20 (Fig. 1C). Small changes in the relative orientation of the reactive groups in the active sites of enzymes are crucial in governing the rate of catalysis (11). In this case, a conformational change in the main chain would be required to bring the thioester linkage and histidine side chain into position for nucleophilic attack and covalent bond formation (see Fig. 1C). We see no evidence of disorder or alternate conformations in the polypeptide segments containing these residues, which suggests that in native C3 this region must be held in a strained conformation capable of forming the acyl-imidazole intermediate, termed here the “closed” conformation. This structure showing the large separation between the histidine and glutamine residues (see Fig. 1C) would then represent the relaxed or “open” conformation.

A number of surface-exposed apolar residues, including Ile23, Phe66, Phe76, Ile132, and Tyr273 (Fig. 1C), flank the thioester region. Sequence comparisons of the C3d region from diverse species show that these residues are also highly conserved (Fig.2A). In fact, when the degree of similarity for each position in the sequence (Fig. 2) is mapped onto the structure, we find a large, contiguous, surface-exposed patch, including the apolar residues, whose degree of conservation is as high as that of residues in the core (Fig. 2B). We interpret the conservation of this patch to be a reflection of an important structural role, namely, that of a domain interface in native C3. We further suggest that the domain interface not only serves to protect the thioester from solvent hydrolysis but, in addition, provides some, if not all, of the strain energy required to hold this region in the closed conformation. At the same time, interactions with His133 must prevent nucleophilic attack on the thioester in native C3. Upon C3 cleavage and subsequent disruption of this interface, the strain energy would be transferred to the covalent bond formed between His133 and Gln20, enhancing the reactivity of the acyl-imidazole intermediate while holding the structure in the closed conformation. After nucleophilic attack and relief of the strain, conversion to the open conformation would occur, precluding the possibility of reforming the acyl-imidazole linkage. It has been shown (10) that the Cys17 thiolate anion released after acyl-imidazole formation serves to base catalyze the subsequent nucleophilic attack, an observation supported by the relative dispositions of Ala17, Gln20, and His133 (Fig. 1C).

Figure 2

Sequence conservation of C3d. (A) Multiple sequence alignment of selected species of C3d and human C4d (B isotype) (21). Residues shaded in yellow are at least 90% buried in the C3d structure, and those shaded in red are residues composing the contiguous surface patch labeled in (B). Numbers correspond to the degree of conservation in C3d sequences only: 0 (not conserved) to A (highly conserved), as determined by the program AMAS (32). In human C4d, approximately 75% of the core residues, as well as the putative domain interface residues, are highly conserved [a conservation index (cons. index) of 7 or higher when included in the AMAS calculation], which suggests that it will adopt a similar fold and possess the analogous domain interface. The helical segments in human C3d are indicated by blue cylinders. [The figure was prepared with ALSCRIPT (35).] (B) Mapping of residue conservation as determined in (A) onto the surface of C3d; white (not conserved) to progressively darker red (highly conserved). [The figure was prepared with GRASP (36).] The conserved patch includes most of the surface apolar residues shown in Fig. 1C.

The presence of a domain interface burying the thioester region and providing the strain energy to hold native C3 in the closed conformation is consistent with a number of chemical inactivation experiments. Even in the absence of proteolytic activation, small nucleophiles such as methylamine and ammonia are known to cleave the thioester linkage, resulting in a slow (∼10-hour) conformational change to the so-called C3b-like state (12). The associated increase in volume and polarity of the cysteine and modified glutamine side chains (spontaneous hydrolysis by water would lead to glutamic acid) would be expected to disrupt the steric and chemical complementary of the domain interface, thermodynamically destabilizing the molecule. Furthermore, mildly denaturing agents are known to promote covalent attachment of C3 to hydroxyl-bearing nucleophiles or the competing thioester hydrolysis reaction (13), again presumably by disrupting the domain interface. Mutations of Cys17, Glu19, Gln20, and His133 show, however, that variations in the interface, including those leading to the loss of the thioester linkage (Cys17 → Ala17, Glu19 → Gln19, and Gln20 → Asn20), do not necessarily preclude a C3-like conformation (14). These results are consistent with the ability of proteins to accommodate changes in core packing or protein-protein interfaces while still retaining function (15). In any case, steric clashes associated with increased bulk, as with methylamine addition, would be expected to be the most destabilizing.

Efforts have been made to identify C3d residues involved in binding the CR2 receptor, and a peptide segment has been proposed (residues 228 through 239 in our structure) (16). However, no consensus has yet been reached, because extensive mutagenesis of this segment (in part located on the convex surface) did not substantially affect binding (17). The CR2 receptor itself consists of 15 to 16 short consensus repeats of 60 to 70 amino acids, only the two NH2-terminal repeats of which are involved in C3d binding (18). Because C3d interacts with pathogen surfaces through a covalent link at Gln20, it is unlikely that the thioester-containing end of the α-α barrel would be involved in CR2 interactions. The structure shows that the opposite end of the barrel (the concave surface in Fig. 1B), which would be completely accessible for receptor interactions even when bound to cell surfaces, presents a potential site of interaction—that of an extended pocket formed primarily by acidic residues (Fig. 3). A number of the residues found in the pocket are highly conserved among diverse C3d sequences (Asp36, Glu160, and Tyr201). This pocket corresponds in location to the substrate binding sites in the α-α barrels of glucoamylase, endoglucanase, and farnesyltranferase. With respect to the receptor, inhibition studies have identified CR2-derived peptides, containing basic residues, that are important in C3d binding (19). Homology modeling [based on the structure of a two-domain fragment of human factor H (20)] shows that these basic residues, among others, form an extensive positively charged surface on the two NH2-terminal repeats of human CR2. These observations make the negatively charged pocket, on the concave end of the barrel, an attractive candidate for the site of CR2 interaction.

Figure 3

Stereo view of an electrostatic surface rendition of C3d, showing the acidic pocket on the concave end of the molecule. Acidic and basic residues are colored red and blue, respectively. Labeled are the surface-exposed residues that form the pocket. The contour level is at ±10 kT. [The figure was prepared with GRASP (36).]

  • * These authors contributed equally to this work.

  • Present address: Centre for Virus Research, Westmead Hospital, University of Sydney, Westmead, NSW 2145, Australia.

  • To whom correspondence should be addressed at the Department of Molecular and Medical Genetics, University of Toronto, Toronto, Ontario, Canada, M5S 1A8. E-mail: james.rini{at}


Stay Connected to Science

Navigate This Article