Previous work has indicated a role for the NH2-terminal segment of the C3 α′-chain in the binding interactions of C3b with a number of its protein ligands. In particular, we have identified two clusters of acidic residues, namely, E736 and E737 and to a lesser extent D730 and E731, as being important in the binding of C3b to factor B and complement receptor 1 and the binding of iC3b to complement receptor 3. Whereas human C3 and C4 have an overall sequence identity of 29%, over a segment near the NH2 termini of their respective α′-chains the sequence identity is 56% (70% chemical similarity). Given the functional similarity between the C4b-C2 and C3b-B interactions in the respective formation of the classical and alternative pathway C3 convertases, as well as the sequence conservation of two acidic clusters, we hypothesized that residues 744EED and 749DEDD within the NH2-terminal segment of the C4 α′-chain would mediate in part the binding of C2 to C4b. We tested this hypothesis using three independent approaches. Site-directed mutagenesis experiments revealed that replacing subsets of the charged residues by their isosteric amides within either acidic cluster resulted in molecules having reduced C2 binding activity. Moreover, a synthetic peptide (C4 residues 740–756) encompassing the two acidic clusters was a specific inhibitor of the binding of C2 to red cell-associated C4b. Finally, Ab raised against the above peptide was able to block the interaction between red cell-associated C4b and fluid phase C2. Taken together, these results strongly suggest that the NH2-terminal acidic residue-rich segment of C4 α′-chain contributes importantly to the interaction of C4b with C2.
In terms of the functional manifestations of complement activation, including its opsonic, inflammatory, and lytic roles in immune clearance, the respective cleavages of C3 and C5 represent the most critical steps in the activation process. In the classical pathway, the cleavage of C3 and C5 is accomplished by convertases having the compositions C4b2a and C4bC3b2a, respectively. Because only in its C4b-bound state is C2a capable of cleaving its physiologic protein substrates C3 and C5 (1, 2), C4b may be thought of as a modulator of the serine protease activity of C2a. The equivalent C3 and C5 convertases of the alternative pathway have the composition C3bBb and C3bC3bBb, respectively. Thus, in these convertases, the serine protease moiety of factor B (Bb) fulfills the proteolytic role and C3b fulfills the enzyme-modulatory role. Therefore, both in terms of sequence and function, C3 and C4 are homologs, as are C2 and factor B.
Human C2 is a 102-kDa serum glycoprotein showing 39% sequence identity with its functional homolog factor B (3). C2 and factor B have similar modular structures consisting from the NH2 to COOH terminus of three short consensus repeats (SCRs)4 of the type found in complement regulatory proteins encoded within the regulators of compliment activation locus, a von Willibrand factor type A (vWFA) domain, and a serine protease domain. The genes encoding C2 and factor B are located within the MHC class III region (4), are <500 bp apart, and have similar intron/exon structures (5, 6, 7). Cumulatively, these data suggest that a gene duplication event has led to the divergence and evolution of factor B and C2 and, indeed, there is an example of a single protein in rainbow trout displaying both classical and alternative pathway C3 convertase subunit activity (8).
When viewed by transmission electron microscopy, human C2 and factor B appear structurally similar (9) with each molecule consisting of three 40-Å-diameter lobes. Presumably, the NH2-terminal-derived C2b and Ba fragments, which are each comprised of three SCR domains, would occupy a single lobe, whereas C2a and Bb, each being comprised of a vWFA domain and a serine protease domain, would occupy two of the lobes. For both C2 and factor B, evidence has been presented that C4b/C3b binding sites are located both within the C2b (2, 10, 11) and Ba activation fragments (12, 13), as well as within their respective vWFA domains (14, 15).
The proteins C3 and C4 contain an intramolecular thioester bond that not only controls their conformational state (16), and thereby their ligand binding properties, but also mediates their covalent attachment to target nucleophiles on pathogen surfaces in a proteolytic activation-dependent manner (17). Whereas mature plasma C3 is a disulfide-linked heterodimer consisting of a 119-kDa α-chain and a 75-kDa β-chain, plasma C4 is a disulfide-linked heterotrimer made up of a 93-kDa α-chain, a 75-kDa β-chain, and a 33-kDa γ-chain. In both cases, proteolytic removal of a 77-residue activation peptide from the NH2-terminal of the respective α-chains, i.e., C4a and C3a, respectively, results in exposure and activation of the thioester. Following thioester transacylation, or the competing hydrolysis reaction, the resulting C3b and C4b molecules acquire ligand-binding properties that were not present in the respective native molecules. The same conformational end state, and therefore the same ligand-binding profile, can also be achieved by direct scission of the thioester in the absence of proteolytic activation (16). Among the new binding sites in C4b or thioester-cleaved forms of C4, e.g., C4(CH3NH2) or C4(H2O), are those for C2, nascent C3b, C5, and the complement regulatory proteins complement receptor 1 (CR1), C4 binding protein (C4BP), membrane cofactor protein, decay accelerating factor, and factor I (18). Likewise, C3b, and its thioester-cleaved analogs such as C3(CH3NH2) and C3(H2O), acquire interaction sites for a similar set of complement family ligands and receptors except that instead of C2 and C4BP, these molecules bind the respective alternative pathway homologs, factors B and H.
There have been quite a large number of studies aimed at identifying ligand interaction sites in C3 (18). There is now general agreement that at least one of the major contacts through which C3b binds the proteins B, H, and CR1 is located within a 42-aa peptide segment at the NH2 terminus of the C3b α′-chain. This conclusion has been reached using a number of distinct, but complementary, experimental approaches including the use of synthetic peptide mimetics (19, 20, 21, 22, 23), the functional site-blocking effect of an anti-peptide Ab (20), the functional profile of engineered chimeric or segment-deleted molecules (24), and, finally, the use of site-directed mutagenesis to identify specific residues in the segment contributing to these various interactions (25, 26). The latter two studies have specifically implicated a number of acidic residues spanning amino acids 730–758 (mature protein numbering of human C3 and C4 used here and throughout), although the subset of the most important acidic residues for each of the three interactions is only partially overlapping. Other studies have suggested the presence of a second contact point for B, H, and CR1 within the C3d region of the molecule (23, 27), although specific C3d contact residues have yet to be identified.
By contrast with C3, relatively little is known about the location of the functionally homologous binding sites in C4 for CR1, C2, and C4BP. One mAb that blocks C4BP binding to C4b also blocks C2 binding to this fragment (28), a finding consistent with C2 and C4BP being antagonistic ligands. Another anti-C4 mAb that blocks the C4b-C4BP interaction has been an epitope mapped to an 89-aa segment located at the NH2 terminus of the C4 α′-chain (29, 30), i.e., the analogous segment of the C4 α-chain to that which has been identified in C3 as containing a binding site for factors B, H, and CR1. Whereas human C3 and human C4 have an overall sequence identity of ∼29%, over a stretch encompassing 27 residues near the NH2 terminus of their respective α′-chains the sequence identity is ∼56% (70% if chemically similar residues are included; see Fig. 1). This region of sequence similarity is then immediately followed by a stretch of sequence showing little or no similarity between C3 and C4. It can also be seen in Fig. 1 that when comparing several representative species of C3 and C4, although the sequence identities are not always absolute, within the two NH2-terminal-most acidic subclusters there tends to be overall compositional conservation of acidic amino acids. Given the functional homology between the C3b-B and C4b-C2 interactions, as well as the sequence conservation noted above, we hypothesized that the acidic residues within the NH2-terminal segment of the C4 α′-chain would mediate, at least in part, the binding of C2 to C4b. Accordingly, in this study we have assessed the C2 binding activities of a series of human C4B isotype mutants in which residues within the negatively charged clusters 744EED and 749DEDD were systematically replaced by their isosteric amides. We have also tested our hypothesis through the use of both the synthetic peptide mimetic and anti-peptide Ab approaches. Our cumulative results clearly demonstrate the involvement of the acidic residue-rich NH2-terminal segment of the C4 α′-chain in mediating the interaction between C4b and C2.
Materials and Methods
The following diethyl barbiturate (veronal)-NaCl buffers were used (31): veronal-buffered saline (VBS), 4 mM veronal, 0.15 M NaCl, 0.15 mM CaCl2, and 0.5 mM MgCl2, pH 7.2 (μ = 0.15); GVB, VBS containing 0.1% gelatin; GVB-Mg2+, GVB containing 5 mM MgCl2; GVBE, VBS containing gelatin and 10 mM EDTA; low ionic strength VBS made isotonic with sucrose (SVB), 4 mM veronal, 0.06 M NaCl, 0.15 mM CaCl2, 0.5 mM MgCl2, and 0.17 M sucrose, pH 7.2 (μ = 0.06); SVB-Mg2+, SVB containing 5 mM MgCl2; SVB containing 0.1% gelatin (SGVB); SGVB-Mg2+, SGVB containing 5 mM MgCl2.
Purified complement components
C1̄s (32), C2 (33), and C4 (34) were purified from fresh frozen human plasma as described previously. A functionally pure human C1 reagent, used in the preparation of complement component cellular intermediates, was prepared from a euglobulin precipitation of whole human serum (35). Guinea pig complement (Sigma, St. Louis, MO) was diluted 50-fold in VBS containing gelatin and EDTA to obtain a C3-C9 reagent. Protein radioiodination was performed by the lactoperoxidase procedure and yielded typical specific activities of 106 cpm/μg (36).
Cell culture media
The high-glucose formulation of DMEM supplemented with 2 mM l-glutamine and 100 U/ml of penicillin/streptomycin was the basal tissue culture medium used in this study. The pH of the medium was maintained by 5% CO2 in a humidified incubator. COS-1 cells were maintained in DMEM supplemented with 10% heat-inactivated FCS (complete DMEM; Life Technologies, Grand Island, NY). The growth medium for transfected COS-1 cells was complete DMEM. FCS-free DMEM consisted of complete DMEM in which the 10% FCS was replaced with 1% Nutridoma-HU (Boehringer Mannheim, Montreal, Quebec, Canada). In the metabolic labeling experiment described below, both the basal DMEM and the Met- and Cys-free DMEM were supplemented with 4% K76-COOH-treated FCS and 1% Nutridoma (DMEM/K76). The procedure for K76-COOH treatment of FCS, which irreversibly inactivates bovine factor I, has been described previously (25).
The synthetic peptide EILQEEDLIDEDDIPVR, corresponding to residues 740–756 of mature C4 (C4740–756), was obtained from two independent sources: Procyon Biopharma (London, Ontario, Canada) and the Alberta Peptide Institute (API; Edmonton, Alberta, Canada). The Procyon peptide had a C-terminal cysteine-amide residue in place of the naturally occurring serine, and the API peptide had at its C terminus the photoactivatable amino acid derivative ornithine-amide-benzoylbenzoate. The API peptide was also supplied as photo-cross-linked conjugates of keyhole limpet hemocyanin (KLH) and BSA, each substituted with on average 17 molecules of peptide per molecule of carrier protein. The KLH- and BSA-peptide conjugates were used, respectively, for rabbit immunizations and for anti-peptide Ab detection and purification (see below). The peptides SIERPDSAPPRVGDT and EDPGKQLYNVEATSY, respectively, corresponding to residues 455–469 of mature C4 β-chain and residues 1198–1212 of human C3, were obtained from Chiron Mimotopes (Raleigh, NC) for previous studies and were used as control peptides. Before use, all peptides were lyophilized from H2O several times to remove traces of volatile compounds and then dissolved in H2O to yield a concentration of about 700 μM. The exact concentration was determined spectrophotometrically at 220 nm using an E220 nm, 1 cm1% of 100.
Rabbit anti-sheep RBC stroma, rabbit polyclonal IgG against human C4, rabbit polyclonal IgG against human C3c, and goat alkaline-phosphatase-conjugated Ab against rabbit IgG were purchased from Sigma. Abs against the synthetic peptide C4740–756 were raised in rabbits by s.c. injections of 500 μg of the KLH-peptide conjugate, initially in CFA, and at two subsequent 2-wk intervals in IFA. The immunizations and the collection of the rabbit antisera were done by technical staff in the Department of Comparative Medicine, Faculty of Medicine, University of Toronto, in accordance with the guidelines of the Canadian Council on Animal Care. The rabbit IgG was fractionated from other serum proteins by ion-exchange chromatography on a QAE Sephadex A-50 as described previously (37) and dialyzed against PBS, pH 7.3. An immunoadsorbant column was prepared by coupling the BSA-C4-peptide conjugate to Affi-Gel 15 (Bio-Rad, Hercules, CA) according to the manufacturer’s instructions. The affinity matrix was equilibrated with PBS, pH 7.3, and the IgG fraction of the rabbit antiserum was applied to the column by gravity flow. Following washing in PBS, the bound Abs were eluted with 100 mM glycine, pH 2.5, and neutralized to pH 8.0 by the addition of 1.0 M phosphate buffer, pH 8.0. The pooled anti-peptide Ab fractions were dialyzed extensively against SVB-Mg2+ containing 0.02% sodium azide.
ELISA for determining the specificity of the C4740–756 anti-peptide Ab
Standard ELISAs in which BSA-C4 peptide conjugate, control BSA, or C4(CH3NH2) were coated on the wells as capture Ag were used to assess the specificity and titer of the preimmune and postimmune rabbit antisera. To assess the relative binding activity of the immunoaffinity-purified anti-C4740–756 Ab toward native C4, C4b (i.e., C1̄s-treated native C4), and C4(CH3NH2), the immunoaffinity-purified Ab was coated as capture Ab on ELISA wells (10 μg/ml). Following blocking, the wells were incubated with a 2-fold dilution series of native C4, C4b, or C4(CH3NH2), where the initial Ag concentration was 50 μg/ml. Using standard ELISA procedures, Ag capture was detected using a goat anti-human-C4 polyclonal Ab followed by alkaline phosphatase-conjugated rabbit anti-goat IgG. Control wells contained no capture Ab, but were otherwise treated the same as experimental wells. Relative binding activity was assessed by the horizontal displacement of the dose-response curves, with the concentration of Ag being plotted on a logarithmic scale.
Plasmid constructions and site-directed mutagenesis
The cDNA expression plasmid pSV-C4B, coding for the wild-type human C4B isotype under the control of an SV40 promoter, has been described previously (38). The plasmid pUC-C4-5′ consists of a 2.8-kb 5′ SalI-EcoRI fragment of the C4B cDNA cloned into pUC19. Site-directed mutants were produced by the overlap extension PCR mutagenesis method (39) using the proofreading enzyme Vent DNA polymerase (New England Biolabs, Beverly, MA) and pSV-C4B as the wild-type template. The resulting 932-bp PCR fragment encompassed unique restriction sites for Eco47III and EcoRI (at bases 2132 and 2788 of the C4B cDNA, respectively) and was digested with these enzymes to generate a 657-bp mutation-containing fragment that could be subcloned into similarly digested pUC-C4-5′, where the sites for Eco47III and EcoRI are also unique (the intermediate vector step is required because the digestion site for Eco47III is not unique in pSV-C4B). Following confirmation of the desired mutation(s), and the absence of any undesired mutations within the 657-bp target sequence by strand-denaturation dideoxy sequencing (T7 polymerase sequencing kit; Amersham Pharmacia Biotech, Piscataway, NJ), pUC-C4-5′ was restricted with SalI and EcoRI to produce a 2780-bp fragment encompassing the mutation(s). The SalI/EcoRI mutation-containing fragment was then exchanged for the corresponding wild-type region in pSV-C4B.
Expression of recombinant C4 (rC4)
COS-1 cells were transiently transfected by the DEAE-dextran method essentially as described in our earlier studies (40) using 15 μg of pSV-C4B plasmid per 100-mm plate seeded 16–20 h earlier with 106 cells. Following the transfection, the cells were allowed to grow in 9 ml complete DMEM for 72 h, after which time the medium was changed to FCS-free DMEM. Supernatants containing rC4 were harvested after a further 48–72 h of incubation.
Metabolic labeling, biosynthetic characterization, and immunoprecipitation
After being allowed to grow for 72 h in complete DMEM, transfected cells were washed in PBS and then incubated for 1 h in Met- and Cys-free DMEM (ICN Biomedicals, Costa Mesa, CA) containing 4% K76-treated FCS and 1% Nutridoma-HU (3 ml/100-mm plate) to deplete internal stores of methionine and cysteine. The medium was then supplemented with 250 μCi of [35S]methionine/[35S]cysteine (ICN Biochemicals TransLabel, ∼1300 Ci/mmol). After 5 h of incubation, an equal volume of Met- and Cys-sufficient DMEM/K76 was added, and the incubation was continued overnight. To assess the biosynthetic processing and C1̄s cleavability of the rC4, metabolically labeled supernatants were immunoprecipitated with rabbit IgG anti-human C4 in conjunction with a Staphylococcus aureus suspension (Sigma), both with and without prior treatment of the supernatants with C1̄s (2 μg/ml, 1 h, 37°C). The buffers and wash procedures for the immunoprecipitations have been described previously (41). All samples were analyzed on 8% SDS-PAGE under reducing conditions, followed by phosphorimage analysis using a Storm 860 scanner (Molecular Dynamics, Sunnyvale, CA). Quantitative measurement of band intensities within a lane was accomplished by peak profile integration using the program IPLabGel 2.0 f (Signal Analytics, Vienna, VA).
Quantitative measurement of secreted rC4
The concentration of rC4 in dialyzed culture supernatants of transfected cells was determined by a competitive solid-phase RIA (42) using 125I-labeled purified human C4 as the probe and rabbit IgG anti-human C4 as the capture Ab adsorbed to opaque polystyrene microtiter plates (Packard Instruments, Meriden, CT). Purified human C4 was used to obtain a standard curve. Radioactivity was measured by liquid scintillation counting directly in the plates using a TopCount instrument (Packard Instruments).
Classical pathway-dependent hemolytic assay
The hemolytic activity of rC4 in SVB-Mg2+-dialyzed transfection supernatants was determined by using sheep erythrocytes coated with Ab and C1 (EAC1), iodine-oxidized C2 (oxyC2), and C3-C9 reagent as described previously (35, 40). After correcting for background, the degree of specific lysis was converted to Z units where Z = −ln(1 − fractional lysis) and is physically equal to the number of hemolytically effective molecules per erythrocyte. Comparisons of activity were made on the basis of Z units per amount of immunochemically determined rC4.
Fluid-phase C2-dependent hemolytic inhibition assay
Secreted rC4 was converted to rC4(CH3NH2) by treating the culture supernatants from transfected cells with 0.1 M methylamine, pH 8.0, for 6 h at 37°C (43). The methylamine-treated supernatants were then dialyzed extensively against SVB-Mg2+ and concentrated to approximately one-eighth of original volume by using either Centricon-100 concentrators (Amicon, Beverly, MA) or Ultrafree-15 Centrifugal Filter Devices (Millipore, Bedford, MA). Following this procedure, the concentration of the rC4(CH3NH2) was determined by a competitive RIA as described above. oxyC2 (44) in SGVB-Mg2+ in an amount sufficient to generate ∼80% hemolysis in the absence of inhibitor protein in a C2 hemolytic assay to be described below (typically 16 ng of C2) was incubated with variable amounts of rC4(CH3NH2) in a total volume of 200 μl of SGVB-Mg2+ at 0°C for 5 min. EAC4b cells (1.5 × 107), prepared as described previously (45), were added, and the tubes were further incubated at 0°C for 10 min, at which time the cells were washed with cold GVB-Mg2+. The resulting EAC4b2 cells were resuspended in 200 μl of GVB-Mg2+ containing excess C1 reagent and incubated for 10 min at 30°C. Lysis was developed by the addition of C3-C9 reagent (1 ml) at 37°C for 45 min. After spinning down unlysed cells, the degree of lysis was determined by measuring the absorbance of the supernatant at 412 nm. Hemolytic data were used to calculate the relative C2 binding ability of each rC4(CH3NH2) species by comparing the 50% inhibition of hemolytic activity of a given molecule to that of the wild-type molecule.
Synthetic peptide competition assay
Varying amounts of synthetic peptide in 160 μl of H2O were mixed with 40 μl of 5× SGVB-Mg2+, and these solutions were then incubated with 16 ng of oxyC2 for 10 min at 0°C. EAC4b cells (1.5 × 107) were then added and the remaining steps of the assay were done exactly as described above for the fluid-phase C2-dependent hemolytic inhibition by rC4(CH3NH2).
Anti-peptide Ab hemolytic inhibition assays
In one type of assay, various amounts of anti-peptide Ab were incubated with EAC4b cells (1.5 × 107) in 65 μl SGVB-Mg2+ at 0°C for 60 min, at which time 16 ng of oxyC2 were added and the assay volume was increased to 200 μl with SGVB-Mg2+. After 10 min of incubation at 0°C, all samples were centrifuged and the supernatants were discarded. The resulting EAC4b2 cells were treated with C1 and C3-C9 reagents, and the degree of lysis was determined as described above. A second assay using the anti-C4740–756 reagent was designed to restrict the effect of the Ab solely to the C4b-C2 interaction. In this assay, EAC4b cells (1 × 108), in sufficient amount to bind virtually all of a test quantity of C2, were incubated with various amounts of anti-C4740–756 for 60 min at 0°C, followed by washing with cold GVB-Mg2+. The cells were then resuspended in 100 μl of SGVB-Mg2+, and 23 ng oxyC2 were added. After incubation for 10 min at 0°C, the cells were centrifuged, and fixed amounts (95 μl) of the respective supernatants were transferred to another tube. The supernatants were then supplemented with SGVB-Mg2+ to give a total volume of 200 μl and incubated with fresh EAC4b cells (1.5 × 107) for 10 min at 0°C. The hemolytic activity of C2 captured on the second set of cells was then determined by the addition of C1 and C3-C9 reagents as described above.
The site-directed mutagenesis approach
The first approach used to test the hypothesis that the two acidic clusters of residues at the NH2 terminus of the C4b α′-chain contribute to a binding site for C2 involved the systematic replacement of these residues with their isosteric amides. Isosteric amide substitution was chosen in preference to alanine scanning, as this approach did not superimpose a potential steric effect on the loss of charge mutation. The various mutant molecules constructed, and their respective names, are indicated in Table I. The mutations were introduced either into the first cluster having the sequence 744EED or the second cluster having the sequence 749DEDD, but not simultaneously into both clusters. Therefore, for simplicity of annotation, the nomenclature of the mutants that we have used will be of the form 744EEN, 749DEND, 749DQND, etc., where the amide replacement residues are indicated in the context of their neighbors.
The residues changed are indicated in bold letters.
When transiently expressed in COS-1 cells and harvested in serum-free medium, all of the rC4 molecules were expressed at levels that were comparable to the wild-type control in terms of total C4 Ag per milliliter of culture supernatant (typically ∼1.5–2 μg/ml at the time of harvesting). Fig. 2 shows the relative hemolytic activities of the mutants in a purely classical pathway-dependent assay. It can be seen that 744QQN, 749DEND, 749DQND, and 749NQNN are severely compromised in this activity with the defect being ≥20-fold. An intermediate level of defect ranging from 8- to 3-fold is seen for mutants 744QQD, 744QED, and 749DQDD, but by contrast mutant 744EEN shows a <2-fold defect in hemolytic activity. However, the hemolytic activity defect need not necessarily arise solely from a defect at the C2 binding stage as other impairments, such as in biosynthetic processing or cleavability by C1̄s, could also give rise to the defect in C4 hemolytic activity. These latter two points were addressed by performing a metabolic labeling experiment on the various transfectants and then carrying out an immunoprecipitation before and after treatment of the labeled supernatants with C1̄s. The results of such an experiment are shown in Fig. 3, A and B. We have previously noted that the posttranslational proteolytic processing of even wild-type human C4 to the mature three-chain form in COS-1 cell transfections is incomplete (40, 46), especially at the junction between β- and α-chains. Thus the β-α band is a major species on the SDS-PAGE phosphorimage analysis of the metabolic labeling experiment shown. Additionally, because the α-γ junction is cleaved more efficiently than is the β-α junction, the γ-chain band is overrepresented relative to β and α. To compare quantitatively the extent of processing to the mature three-chain form of C4, the only form that has hemolytic activity (47), the gels shown in Fig. 3, A and B, were subjected to quantitative peak profile analyses. We then determined for all of the non-C1̄s-treated lanes the ratio of the α-chain pixel intensity to the total pixel intensity of all the chains in the lane. From the bar graph representation of this analysis shown in Fig. 3,C, it can be seen that mutants 744EEN, 744QED, 744QQD, and 749DQDD were fairly comparable to the wild type with respect to their posttranslational proteolytic processing. By contrast, mutants 744QQN, 749DEND, 749DQND, and 749NQNN showed defects of 2- to 3-fold in their respective processing relative to wild-type C4. It can also be seen in Fig. 3 that whereas mutants 744EEN, 744QED, 744QQD, and 749DQDD are cleaved by C1̅s to an extent comparable to that occurring in the wild-type molecule, whatever α-chain is present in the 749DQND, 749NQNN, 749DEND, and 744QQN mutants is relatively resistant to cleavage by C1̄s. It is known from the work of Ogata and colleagues that cleavability by C1̅s is quite sensitive to changes immediately downstream of the cleavage site (48). Even in the absence of sequence changes in this region, thioester-hydrolyzed C4, which spontaneously forms in the culture supernatants, will be resistant to cleavage by C1̄s (49).
The acquisition by C4 of a conformation that is capable of interacting with C2 is achievable either by proteolytic activation with C1̄s, which would yield a thioester-hydrolyzed fluid-phase C4b fragment, or by direct nucleophilic scission of the thioester in native C4 with the small nucleophile methylamine (43). To avoid the problem of C1̄s cleavage resistance and, at the same time, to maximize the concentration of the rC4 molecules in the supernatants capable of interacting with C2, the supernatants were reacted with methylamine to generate the C4(CH3NH2) species having a C4b-like conformational state. Molecules that had already undergone spontaneous thioester hydrolysis (i.e., forming the species C4(H2O)) would also be in a C4b-like conformation. These C4b-like molecules were then assessed for their ability to interact with C2 in the fluid phase and thereby inhibit its detection in a subsequent C2 hemolytic assay. Thus, culture supernatants containing various amounts of methylamine-modified recombinant wild-type or mutant C4 in SGVB-Mg2+ were allowed to react in the fluid phase with a constant amount of C2, the amount of C2 having been predetermined to yield ∼80% hemolysis in a C2 hemolytic assay. Uncomplexed C2 was then captured by EAC4b cells made with an excess of purified C4. Following washing, hemolysis of these EAC4b2 cells was developed by the addition of C1 and subsequently C-EDTA, a C3-C9 reagent, as a source of C3-C9. Therefore, if the 744–752 segment mutants have impaired C2 binding capacity, their fluid-phase C4(CH3NH2) derivatives will not bind C2 in the preincubation step and thus there will be no decrease in the amount of C2-dependent hemolytic activity observed relative to a sham preincubation. In contrast, active C4(CH3NH2) molecules will yield a dose-dependent inhibition curve, and partial defects will yield a horizontal displacement from wild-type inhibition curve when the data are presented in the form of a semilogarithmic plot. The results of one such C2 binding experiment are depicted in Fig. 4, with the data for the 744EED and 749DEDD clusters being grouped in A and B, respectively. Fig. 4,C shows a summary in bar graph form of the relative C2 binding capacities of the various mutant C4 molecules in which data from three independent experiments have been averaged. These data provide strong evidence for the involvement of residues in both acidic clusters in mediating C2 binding and generally indicate that multiple mutations result in a cumulative defect. One notes a striking similarity between the bar graph showing the relative C2 binding activities of the mutants (Fig. 4,C) and the bar graph showing overall hemolytic activity (Fig. 2). For the case of mutants 744EEN, 744QED, 744QQD, and 749DQDD, which show normal processing and C1̅s cleavability, the correlation between these two assays suggests that the hemolytic defect can indeed be accounted for at the C2 binding stage. For the case of mutants 744QQN, 749DEND, 749DQND, and 749NQNN, which have, respectively, ∼3, ∼6, ∼1, and <1% of wild-type C2 binding activity, interpretation of the data is clouded by the defects in posttranslational proteolytic processing. Nevertheless, quantitative analysis of the metabolic labeling experiment (Fig. 3) suggests that the extent of the C2 binding defect is much greater than the processing defect in all cases. For example, whereas the processing defect in 749DEND is about 2-fold, the C2 binding defect is about 16-fold, thereby suggesting an important contribution by C4 residue D751 to the interaction with C2.
The peptide mimetic approach
Although the mutagenesis results strongly suggested that the NH2-terminal α′-chain segment of C4, and, in particular, some of the acidic residues, contributed to the binding interaction with C2, these results on their own were not definitive because, at least for some of the mutants analyzed, there was evidence for some undesirable effects on global conformation. To further test the involvement of the NH2-terminal α′-chain segment in C2 binding from an independent direction, a synthetic peptide corresponding to C4 residues 740–756 and having the sequence EILQEEDLIDEDDIPVR was assessed for its ability to inhibit the binding of C2 to EAC4b cells. It can be seen in Fig. 5 that this peptide was able to compete with cell-associated C4b for the binding of C2, with the 50% inhibition point requiring ∼0.2 mM of peptide. In contrast, a control peptide derived from a segment of human C3d failed to show any interaction with C2. Similar inhibitory effects on C2 binding were obtained when this experiment was repeated with an independently synthesized peptide of the same C4 sequence, whereas a second control peptide, this time corresponding to human C4 β-chain segment 455–469, was still without effect in the assay (data not shown).
The blocking Ab approach
To further corroborate the involvement of the NH2-terminal C4 α′-chain segment in the interaction with C2, a polyclonal rabbit Ab was raised against the peptide corresponding to C4 residues 740–756. This Ab was immunoaffinity-purified and was then assessed for its ability to block the interaction between C4b and C2. In preliminary ELISA experiments, we confirmed that the anti-peptide Ab, but not preimmune rabbit IgG, would not only recognize the peptide as a conjugate of BSA (peptide-KLH was the immunogen) but also would react, albeit with much lower titer, with methylamine-treated C4 that had been coated on the ELISA plate. Moreover, when the affinity-purified anti-peptide Ab was coated on ELISA plates it had the ability to capture C4 Ag from solution with its reactivity against C4b being about 3-fold higher than against either native C4 or C4(CH3NH2) (data not shown). These latter observations suggest that a subset of the epitopes recognized by the anti-peptide polyclonal Ab are available even in the native molecule, but removal of the C4a activation peptide either results in their increased exposure or the liberated free amino group forms part of an epitope.
In the experiment shown in Fig. 6, EAC4b cells, made with excess purified C4, were preincubated with various amounts of specific anti-peptide Ab or with control nonspecific rabbit IgG. The cells were then incubated with a fixed quantity of C2, and then lysis was developed upon addition of excess C1 and C-EDTA. It can be seen that the anti-C4-peptide IgG, but not the control IgG, inhibited C2-dependent hemolysis in a dose-dependent manner. Although the results were strongly suggestive of a blockage of C2 binding to C4b by the anti-peptide Ig, we could not exclude the possibility that this Ab might also interfere with the C5 convertase subunit functionality of C4b, either by sterically masking the transacylation target residues of nascent C3b in C4b α′-chain or by blocking the C5 binding site in C4b. To circumvent this ambiguity, a second hemolytic assay was used in which the effect of the anti-peptide Ab would be limited to the C4b-C2 interaction step. Specifically, EAC4b cells, made with excess human C4, were preincubated with the peptide-specific or control IgG as before. Following washing, these cells were incubated with a limiting amount of C2, an amount that in the absence of any pretreatments would become almost completely bound to the EAC4b cells. The cells were pelleted and the residual C2 activity in the supernatant was assessed using a fresh aliquot of EAC4b cells. The experiment depicted in Fig. 7 shows that as the amount of anti-peptide Ab added in the preincubation step with the EAC4b cells is increased, the amount of C2 that remains uncomplexed, and thus available for binding and inducing hemolysis in the second aliquot of EAC4b cells, also increases. The control IgG added in an amount equivalent to the highest amount of the specific anti-peptide IgG used showed no ability to block the interaction between C2 and the C4b present on the initial sample of EAC4b cells. Cumulatively, these experiments show that an Ab recognizing an epitope within the 740–756 segment of target-bound C4 α′-chain can block the interaction between C4b and C2.
Whereas there is a considerable body of knowledge regarding the localization of complement ligand-binding sites in C3 (reviewed in Ref. 18), comparable studies on C4 are much more limited. At the outset of this investigation, the only binding sites in C4 that were identified at the specific amino-acid level were a C5 binding site surrounding β-chain residue R462 (46) and a noncovalent interaction site for nascent C3b in the vicinity of α-chain residue S1217 (50). The sole other relevant data in the literature pertained to an anti-C4 mAb that could block the interaction between C4b and C4BP and that was shown to bind to an 89-aa segment spanning residues 738–826 within the NH2-terminal-most α-chain fragment of C4c (30). This investigation has identified two acidic clusters of residues that are a subset of the above segment, specifically located near its NH2 terminus, as contributing to the binding interaction with C2. This is consistent with C4BP and C2 being antagonistic ligands, and therefore binding to the same general area of the molecule, although not necessarily to the same key contact residues.
The two clusters of acidic residues spanning residues 744–752 were targeted for investigation because of their general conservation with the analogous residues in C3 α′-chain and because the C3 residues had been shown to contribute to the binding interaction between C3b and factor B (25), this interaction being the alternative pathway analog of the C4b-C2 interaction. We believe that the congruency of results obtained using three independent binding site-mapping approaches, namely, site-directed mutagenesis, synthetic peptide competition, and anti-peptide Abs, strongly supports a role for this segment of C4b in mediating the interaction with C2. Interpretation of the site-directed mutagenesis data for a subset of the mutants was complicated by the fact that in addition to the loss in C2 binding, there was clear evidence for more general conformational perturbations that affected the extent of posttranslational proteolytic processing and cleavability by C1̄s. Nevertheless, half of the mutants analyzed displayed wild-type-like processing and cleavability by C1̄s and yet also showed 2- to 8-fold defects in C2 binding (i.e., 744EEN, 744QED, 744QQD, and 749DQDD), thereby clearly indicating a role for E744, E745, D746, and E750 in mediating the binding interaction with C2. Furthermore, even for a mutant such as 749DEND that showed a 2- to 3-fold defect in processing, the defect in C2 binding was much greater (∼16-fold) than the processing defect. Thus the C2 binding defect in this case could not simply be accounted for by the lower concentration of the fully processed three-chain form of the molecule. Although clearly not definitive for this amino acid, these data nevertheless point toward D751 also being a participant in the binding interaction with C2. C4 residues E750 and D751 correspond to C3 residues E736 and E737, whose mutation we have previously shown decreases the strength of the C3b-B interaction ∼3-fold (25). One notable difference between the charged residue mutational analysis of the corresponding C4 and C3 segments is the apparently greater importance of the NH2-terminal-most 744EED acidic cluster of C4 for the C2-C4b interaction than is the case for the corresponding C3 730DED cluster in mediating the C3b-B interaction. Although even single mutations within the 744EED acidic cluster of C4 have fairly pronounced effects on the C2-binding interaction, even the double mutation of C3 residues D730 and E731 led to only about a 20–30% decrease in factor B binding activity (25).
The general conclusions from the site-directed mutagenesis data regarding the participation of the two acidic clusters in forming a C2 binding site were further supported by experiments showing the ability of a synthetic peptide encompassing residues 740–756 to inhibit the authentic C4b-C2 interaction in a specific and dose-dependent manner and by the ability of an Ab to this peptide segment to also block the C4b-C2 interaction. Negative results using peptide mimetics are inconclusive because the conformation of the peptide may be very different as a free peptide than within the context of the rest of the protein. However, a positive result such as we have in this case can only be artifactual if the array of side chains available in the peptide are ones that would not normally be accessible in the whole protein. This is clearly not the present case as the anti-peptide Ab was not only able to recognize the epitope in the context of C4b, but also its binding to C4b blocked the interaction with C2. Furthermore, although the steric “footprint” of the Ab on the C4b molecule would almost certainly extend beyond the acidic residue-rich peptide epitope, the fact that both the site-directed mutagenesis and the synthetic peptide competition experiments did not extend much beyond the cluster of acidic residues, and yet in each case had an effect on the C4b-C2 interaction, is consistent with the functional site-blocking effect of the anti-peptide Ab being a direct rather than an indirect one.
An anti-peptide Ab against the NH2-terminal α′-chain segment of C3 corresponding to C3 residues 727–768 was shown to block the C3b-B interaction (20), analogous to the effect that we observe on the C4b-C2 interaction by the anti-C4740–756 anti-peptide Ab. In the case of the anti-C3727–768 Ab, the antigenic determinant recognized only became available upon proteolytic conversion of native C3 to C3b, iC3b, or C3c. Thus the immunodominant epitope behaved as a classic neoepitope. In our case, there is only a 3-fold enhancement in Ab reactivity toward C4b vs native C4. It is possible that in the case of C4, the two acidic clusters are largely available to C2 even in the native molecule, and that it is the formation/exposure of an independent binding site in another part of the molecule that determines the ability of C4b and C4(CH3NH2), but not native C4, to bind C2. Alternatively, there may actually be no difference in the relative exposures of the NH2-terminal acidic clusters in native C3 and native C4. Rather, it may be that a more C-terminal epitope becomes exposed upon proteolytic activation of C3, an immunodominant epitope that was present in the C3727–768 peptide immunogen, but that was not present in the C4740–756 peptide corresponding to only the NH2-terminal-most 40% of the immunogen used in the C3 study.
As suggested above, the interaction between C2 and C4b involves multiple points of attachment. In terms of the currently known C4b interaction sites within C2, the NH2-terminal α′-chain segment of C4b could potentially interact with the SCR domains of C2b or with the vWFA domain portion of C2a. Like many complement protein-protein interactions, that between C4b and C2 is known to be ionic strength-dependent (2). Furthermore, the C2b-C4b interaction, which has been hypothesized as being important in the assembly of the classical pathway C3 convertase (10), is not only exquisitely sensitive to ionic strength, but also to the presence of Mg2+ (2). In this latter respect, the interaction between C2b and C4b differs from that of the equivalent alternative pathway interaction between Ba and C3b as the latter was found to be metal ion-independent (13). The ionic strength dependence as well as the Mg2+ dependence of the C2b-C4b interaction are both consistent with the involvement on the C4 side of the interface of the acidic residue-rich segment near the NH2 terminus of the C4b α′-chain. One or more of the carboxylate groups might be a coordinate ligand for the metal ion, which in turn facilitates bridge formation with the partner protein C2. Besides the putative metal coordination site within the C2b region, both C2a and Bb have a known metal binding motif, the MIDAS (metal-ion-dependent adhesion site) motif, which is part of their respective vWFA domains. The three-dimensional structure of examples of vWFA domains from the adhesion molecules αMβ2 (CR3) and αLβ2 (LFA-1) have been determined, and five metal ion-coordinating residues have been identified (51, 52). It has been suggested that the sixth metal ion-coordinating residue would be contributed by the ligand protein (51). Our previous finding that mutation of the C3 α′-chain NH2-terminal acidic residues E736 and E737 resulted not only in a loss in factor B binding by C3b, but also in a loss in CR3 binding by iC3b, led us to suggest that because of the commonality of Mg2+ dependence and the presence of a vWFA domain in both factor B and CR3 that the contact of E736 and E737 might be with the respective vWFA domains (25). Therefore, one might speculate that the C4b NH2-terminal α′-chain acidic cluster interacts with the vWFA domain of C2. Through homology modeling, the vWFA domain of factor B has recently been visualized (53, 54). Furthermore, a study using the technique of surface-enhanced laser desorption-ionization affinity mass spectrometry, in conjunction with homology modeling, has suggested that peptide segments within the MIDAS cleft were directly involved in C3b binding (54). Therefore, by extension, the vWFA domain of C2 should also be involved in contacting C4b. However, electrostatic surface potential renditions of the MIDAS surface for either factor B or modeled factor B-C2 chimeras did not show an obvious electropositive surface that might interact with the highly negatively charged 730–736 segment of C3 α′-chain or, by extension, the corresponding 744–752 segment of C4 α′-chain (53, 54). Indeed, the homology modeling revealed that the floor of the MIDAS cleft is highly negatively charged, and the only significant electropositive patch visualized on the surface of the factor B vWFA is remote from the C3b-binding peptide segments identified by the affinity-based mass spectrometry (54). Although it is still possible that the acidic residues at the amino terminus of the C3b and C4b α′-chain might directly contribute to the coordination of the Mg2+ ion within the MIDAS, it is hard to explain why in the case of C4b so many of the isosteric amide mutations would affect the single coordinating residue that one would expect to come from that side of the interface. Thus, based on the modeled data available at this time, it does not seem likely that the MIDAS surfaces of factor B or C2 would mediate multiple ionic interactions with the acidic residues located near the NH2 termini of, respectively, the C3b and C4b α′-chains.
By contrast to the above, there are several lines of circumstantial evidence suggestive of an ionic interaction between SCR motif-containing proteins and the negatively charged residues located at the NH2 termini of both C3b and C4b α′-chains. First, mutagenic analyses of the charged residues within the 730–768 segment of C3b α′-chain showed that binding interactions with factor H and CR1, both of which are composed entirely of SCR motifs, are compromised only by the loss of negative charges, despite there being four basic residues that were mutated within the target segment (26). Loss of charge mutations at C3 residues E736 and E737 not only cause loss in CR1 binding activity, but also factor B binding activity. Owing to the commonality of SCR motifs to CR1 and factor B, it is tempting to speculate that the defect in factor B binding may reflect the loss of an interaction between the acidic residues of the NH2-terminal α′-chain segment of C3b and the SCR motif-containing Ba region of factor B. Furthermore, it has been observed that in a series of substitution mutants in the SCR domains of CR1 that gain of function mutants for both C3b and C4b binding often correlated with an increase of positive charge, or decrease of negative charge, suggesting an electropositive interface on the CR1 molecule and an electronegative interface on the C3b or C4b molecule (55). Further evidence for an electropositive interface on an SCR-containing protein binding to C4b comes from a recent modeling study on SCR domains of C4BP that predicted an electropositive surface at the interface between SCR modules 1 and 2 (56). Furthermore, this work showed that mutation of these basic residues in C4BP to neutral ones led to a loss in C4b binding activity. Given the evidence in the literature suggesting a C4BP binding site within the NH2-terminal segment of C4b α′-chain, Blom et al. (56) hypothesized that the highly acidic segment of C4b α′-chain that has been the focus of our present study may also be responsible for the interaction with the electropositive interface that they had identified on C4BP. However, they found that a peptide corresponding to the C4b740–757 segment did not inhibit the interaction between C4b and C4BP. This peptide was virtually identical with the one that we used and that was found to inhibit the interaction between C4b and C2. Although C2 and C4BP are antagonistic ligands for C4b, they need not use exactly the same contact residues to sterically block one another. Indeed, for the analogous pair of antagonistic alternative pathway proteins, i.e., factors B and H, data from both site-directed mutagenesis and peptide mimetic studies have shown that these binding sites are separable within the NH2-terminal α′-chain segment of C3b (22, 26) with the acidic residues essential for H binding being ∼7 aa COOH-terminal of those mediating factor B binding.
In summary, we have demonstrated using three independent experimental approaches that the acidic residues 744EED and 749DEDD within the NH2-terminal segment of the C4b α′-chain contribute importantly to the binding interaction between C4b and C2. On balance, we believe that the arguments in favor of these acidic residues interacting with the SCR domains of C2b appear stronger than those that favor the interaction of this region with the vWFA domain of C2a. Clearly, however, future studies using recombinant fragments corresponding to C2b and to the vWFA domain of C2, in conjunction with the mutant C4 molecules engineered in this study, will be necessary to resolve this issue.
This work was supported by funding from the Medical Research Council of Canada (MT-7081).
Abbreviations used in this paper: SCR, short consensus repeat; vWFA, von Willebrand factor type A; CR, complement receptor; C4BP, C4 binding protein; C-EDTA, a C3-C9 reagent; oxyC2, iodine-oxidized C2; EAC, sheep erythrocytes coated with Ab and the indicated complement components or fragments thereof; MIDAS, metal-ion-dependent adhesion site; rC4, recombinant C4; VBS, veronal-buffered saline; SVB, low ionic strength VBS made isotonic with sucrose; SGVB, SVB with gelatin; KLH, keyhole limpet hemocyanin; API, Alberta Peptide Institute.