Abstract
A vital role for complement in adaptive humoral immunity is now beyond dispute. The crucial interaction is that between B cell and follicular dendritic cell-resident complement receptor 2 (CR2, CD21) and its Ag-associated ligands iC3b and C3dg, where the latter have been deposited as a result of classical pathway activation. Despite the obvious importance of this interaction, the location of a CR2 binding site within C3d, a proteolytic limit fragment of C3dg retaining CR2 binding activity, has not been firmly established. The recently determined x-ray structure of human C3d suggested a candidate site that was remote from the site of covalent attachment to Ag and consisted of an acidic residue-lined depression, which accordingly displays a significant electronegative surface potential. These attributes were consistent with the known ionic strength dependence of the CR2-C3d interaction and with the fact that a significant electropositive surface was apparent in a modeled structure of the C3d-binding domains of CR2. Therefore, we have performed an alanine scan of all of the residues within and immediately adjacent to the acidic pocket in C3d. By testing the mutant iC3b molecules for their ability to bind CR2, we have identified two separate clusters of residues on opposite sides of the acidic pocket, specifically E37/E39 and E160/D163/I164/E166, as being important CR2-contacting residues in C3d. Within the second cluster even single mutations cause near total loss of CR2 binding activity. Consistent with the proposed oppositely charged nature of the interface, we have also found that removal of a positive charge immediately adjacent to the acidic pocket (mutant K162A) results in a 2-fold enhancement in CR2 binding activity.
The essential role of C3 in many aspects of the innate immune response (opsonization, phagocytosis, inflammation, and complement-mediated cytolysis) has long been recognized. More recently, there has been an accumulating body of evidence indicating that C3 also participates in the regulation of the adaptive immune response (reviewed in Ref. 1). This regulatory action is mediated by the interaction of C3dg, an activation product of C3, with its complement receptor CR24 (CD21), the latter being primarily present on the membranes of B cells and follicular dendritic cells. Several in vivo studies have pointed out the importance of this interaction in the modulation of the Ab-mediated humoral response. For example, the administration to mice of a blocking anti-mouse CR2 mAb (2), or of a soluble form of CR2 (3), at the time of the primary immunization with suboptimal doses of T-dependent Ags-diminished the primary response and abolished the secondary response and isotype switching. Knockout mice generated by disruption of the Cr2 gene displayed an impaired primary response, virtually no secondary response, and greatly reduced number and size of germinal centers upon immunization with T-dependent Ag (4, 5). Because CR1 and CR2 are differential splice products of the Cr2 gene in the mouse (6), the phenotype of the knockout reflects losing both types of receptors. Nevertheless, in view of the similarity of the immune response profile of the CR1/CR2 knockout mice to the mice in which binding of ligand to CR2 was blocked by a CR2-specific mAb, or by a competing soluble form of CR2, it suggests that minimally an interaction between iC3b/C3dg-coated Ag and CR2 is required for an optimal immune response. The phenotype of the Cr2 gene knockout is in fact very similar to that observed in C3- and C4-deficient mice (7) and to that of CD19-deficient mice (8, 9). The mechanism through which CR2 is thought to participate in a B cell signaling event involves its association with a pre-existing signal transduction complex, namely the CD19/CR2/TAPA-1 complex (10). CD19 is considered to be the key molecule in the signaling process, but a direct ligand for it has not been identified. Thus, a model has been proposed that bridges the B cell receptor (BCR) to CD19 via the dual recognition of a C3dg-coated Ag by CR2 and by the Ab component of the BCR complex (11). In support of this model, it has been found that there was about a 10-fold decrease in the threshold concentration of anti-BCR required for B cell activation in vitro when Abs that simultaneously cross-linked the CD19/CR2 complex were also present (12). A study by Dempsey et al. (13) has provided compelling evidence for the threshold hypothesis in vivo. Specifically, when hen egg lysozyme (HEL) that had been biosynthetically conjugated to two or three tandem copies of C3d (a proteolytic limit fragment of C3dg retaining full ability to bind CR2) was administered to mice in the absence of any other adjuvant, it was found to be, respectively, 1000- and 10,000-fold more immunogenic than similarly administered HEL alone. Moreover, the IgG response to HEL-C3d3 was 100-fold higher than that to HEL administered in CFA. Besides providing in vivo evidence for the threshold hypothesis, this study clearly demonstrated the potential of using C3d as a molecular adjuvant. A subsequent investigation showed that a single molecule of C3b that was conjugated to HEL through the physiologically relevant thioester-mediated transacylation mechanism also greatly enhanced the immunogenicity of HEL (14). Although providing a second example of the use of complement ligand-receptor interactions as a molecular adjuvant, this experiment did not distinguish among the contributions of CR1, CR2, CR3, and CR4 to the adjuvant effect as at least a portion of the Ag-conjugated C3b would have been cleaved to iC3b postimmunization, thus changing its receptor class binding preference from CR1 to the latter three CRs.
CR2 is a member of the regulators of complement activation (RCA) family (15) and depending on splice site usage, its extracellular region consists of 15 or 16 short consensus repeat (SCR) domains. The binding sites for its ligands, C3dg and iC3b, have been localized to SCR domains 1 and 2 (16) and peptide segments 10-LNGRIS-15 of SCR-1, 84-GSTPYRHGDSVTFA-97 of SCR-2 (17), and 63-EYFNKYS-69 located between SCRs 1 and 2 (18), have been suggested to be important for iC3b/C3dg binding. The three-dimensional structure of CR2 has not been determined but a model of CR2 SCR-1 and SCR-2 has been produced based on the nuclear magnetic resonance structure of SCRs 15–16 of factor H (19). In this model the peptide segments considered important for iC3b/C3dg binding are largely surface exposed and with some rotation about the interdomain segment, the 10–15 and 84–97 peptide segments can be configured to form a contiguous patch (17). An electrostatic surface potential rendition of the modeled CR2 domains indicated that the basic residues within these latter two segments were part of two prominent electropositive patches (20).
The identification of a site(s) within C3d that mediates the interaction with CR2 has been the subject of some controversy, as has been the issue of whether there are supplementary sites in iC3b in addition to the site(s) in C3d. In 1985, Lambris et al. (21), based largely on synthetic peptide mimetic studies, proposed that a C3d segment having the sequence 1199-EDPGKQLYNVEA-1210 (mature C3 numbering) played a major role in the interaction with CR2. Subsequent studies from this group also suggested that in addition to the C3d contacts, there were additional contacts for CR2 in the C3c part of iC3b (22). However, in 1995 our laboratory demonstrated that very extensive mutagenesis of the 1199–1210 segment of human C3 led to minimal effects on the binding of either iC3b or C3dg to CR2, suggesting no more than a minor role for this segment of C3d in mediating binding to CR2 (23). Additionally, our work, like that of an earlier study by Kalli et al. (24), failed to detect a substantial difference between the respective CR2-binding activities of iC3b and C3dg.
In the absence of any further clear candidate sites in C3d to test for CR2 binding, we turned our attention toward obtaining a structure of C3d in the hope that the structure might suggest a candidate site to guide further mutagenesis work. The structure of human C3d has recently been published (20) and reveals the molecule to be a highly helical α-α barrel presenting two distinct surfaces, a convex surface bearing the thioester residues and, at the opposite side, a concave surface. Whereas one might expect that the residues located on the convex end to be partially occluded because of the attachment to the target Ag, the concave surface is fully accessible to the solvent, and thus potentially available for receptor interactions. In addition, the rim of the cavity on the concave surface is surrounded by acidic residues that form a V-shaped entity providing a surface with a prominent negatively charged electrostatic potential. The involvement of such an electronegative patch in forming ionic bonds with a complementary surface on CR2 would be consistent not only with the electropositive surface on CR2 revealed by the modeling, but also with the well established ionic strength dependence of the iC3b/C3dg-CR2 interaction (23, 25). Of interest, mutagenesis studies have suggested that a similar charge complementarity exists between acidic residues in C3b (26, 27) and basic residues in known C3b-binding regions of CR1 (28), the latter protein also consisting of multiple SCR motifs.
Because of the above reasons, we considered the acidic pocket of C3d to be an attractive candidate for the site mediating the interaction with CR2. In the present study we have performed alanine scanning mutagenesis of all of the residues within, and immediately adjacent to, the acidic pocket in C3d and assessed the CR2 binding ability of the recombinant proteins in two independent binding assays. This structure-guided analysis has proved fruitful in allowing us to identify two clusters of C3d residues that are crucial for the interaction with CR2.
Materials and Methods
Buffers, cell media, and purified components
The following buffers were used: Veronal-buffered saline (VBS; 4 mM diethylbarbiturate (Veronal), 0.15 M NaCl, 0.15 mM CaCl2, and 0.5 mM MgCl2; pH 7.2); sucrose-gelatin-Veronal buffer (SGVB; low ionic strength VBS containing 227 mM sucrose, 30 mM NaCl, and 1 g/L of gelatin); SGVB-EDTA buffer (SGVB containing 10 mM EDTA); dextrose-Veronal buffer (DVB; 2.5 mM Veronal (pH 7.2), 140 mM glucose, 71 mM NaCl, 0.5 mM MgCl2, and 0.15 mM CaCl2); and dextrose-gelatin-Veronal-buffer (DGVB; DVB containing 1 g/L gelatin).
COS-1 cells were maintained in high-glucose DMEM containing 10% heat-inactivated FCS, 2 mM glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (complete DMEM). The murine myeloma cell line J558L-pSNRCR2, which secretes the fusion protein (CR2)2-IgG1 (Ref. 3 ; provided by Dr. Henry Marsh, Avant Immunotherapeutics, Needham, MA), was maintained in complete DMEM supplemented with 50 μM 2-ME and 0.6 mg/ml active G418-sulfate (Geneticin, Life Technologies, Burlington, Ontario, Canada). Raji B lymphoblastoid cells were maintained in complete RPMI 1640 medium.
Complement components C1 (29), C2 (30), C4 (31), C3 (31), factor I (32) and factor H (33) were purified from human plasma as previously described, and C3dg was purified from Mg2+-activated human plasma as described (34). Native C3 was converted to iC3b as described in (35).
IgG fractions of rabbit anti-human C3c (Sigma, St. Louis, MO) and rabbit anti-human C3d (Serotec Canada, Toronto, Ontario, Canada) were radioiodinated by the lactoperoxidase procedure (36).
Site-directed mutagenesis and expression of recombinant C3
All mutants were generated by the overlap-extension PCR method (37) in a modified version (pSV-C3(LC)) of the expression plasmid pSV-C3 containing the full-length cDNA of human C3 (38). pSV-C3 was modified to introduce an AgeI restriction site at nucleotide 3713 and to delete a BglII site at nucleotide 4220 of the C3 cDNA sequence. In neither case did the base substitutions alter the amino acids encoded by the nucleotide sequence. As delimited by the sense and anti-sense flanking primers used for the overlap-extension PCR, the target region for mutagenesis spanned nucleotides coding for amino acid residues 925 through 1371 of C3 (mature C3 numbering), which includes within it the nucleotides encoding the entire C3d fragment (residues 980 to 1281). The overlap-extension PCR fragments of mutants located upstream of the AgeI site were restricted with SalI and AgeI to produce a 670-bp fragment; for mutants located downstream of the AgeI site the overlap-extension fragments were restricted with AgeI and BglII to produce a 507-bp fragment. The restricted fragments were then exchanged with the corresponding segments in wild-type pSV-C3(LC). The mutations were confirmed by strand denaturation dideoxy DNA sequencing using a T7 polymerase sequencing kit (Amersham Pharmacia Biotech, Canada).
Wild-type or mutant plasmid DNAs (5 μg) were transfected into COS-1 cells by our previously described modification of the DEAE-dextran method (39, 40) in 100-mm plates. Culture supernatants (about 8 ml) were harvested after 5 days, dialyzed against VBS and concentrated about 5-fold using Biomax concentrators (Millipore, Bedford, MA). Recombinant C3 protein concentration in the concentrated supernatants was determined by a C3-specific ELISA, using plates coated with 10 μg/ml rabbit anti-C3c Ab (Sigma), a goat anti-C3 sandwiching Ab (Quidel, San Diego, CA), and finally an alkaline phosphatase-conjugated anti-goat IgG for detection. Purified human C3 was used to obtain a standard curve.
Metabolic labeling of transfected COS-1 cells was performed 48 h posttransfection as described previously (23).
Preparation of iC3b-coated erythrocytes
The preparation of sheep erythrocytes bearing varying amounts of wild-type or mutant iC3b was done via minor modifications of our previously described procedure (23). Briefly, 0.25 ml EAC4b2a (109 cells/ml) in SGVB-EDTA buffer were incubated with variable amounts of concentrated transfection supernatant (from 0.175 ml to 1.5 ml at a concentration of 1–1.5 μg/ml) for 2 h at 37°C. The resulting EAC423b were converted to EAC423bi cells by incubating with 4 μg factor H and 0.4 μg factor I for 3 h at 37°C, followed by washing.
The relative number of iC3b molecules/cell was determined by evaluating the binding of 2.5 μg 125I-labeled anti-C3c Ab to 50 μl containing 5 × 107 EAC423bi cells as previously described (26). The nonspecific binding component was determined by incubating the 125I-labeled anti-C3c with an equal number of EAC4b2a cells.
The conversion of red cell-bound recombinant C3b to iC3b was monitored in experiments using metabolically labeled culture supernatants as the source of recombinant C3 for building the EAC423bi cells. As described in detail in an earlier study (26), membranes of the EAC423bi cells were solubilized in 0.75% SDS, treated with 1 M hydroxylamine to break the ester linkages between the 35S-iC3b and red cell surface molecules, and then the released labeled molecules were analyzed by SDS-PAGE autofluorography under reducing conditions.
Preparation of C3dg-coated erythrocytes
Wild-type and mutant C3dg-coated erythrocytes were prepared as described above for the conversion of EAC423b to EAC423bi except that instead of using factor H as the I cofactor, soluble CR1 (sCR1; provided by Dr. Richard Smith while at SmithKline Beecham, Harlow, U.K.) was used for this purpose (5 μg/3 × 108 EAC423b) as sCR1, but neither factor H nor soluble membrane cofactor protein, is an efficient cofactor for the so-called third factor I-mediated cleavage converting iC3b to C3c and C3dg (27). Because the C3dg fragment remains cell-associated, whereas the C3c fragment does not, the success of the conversion was in each case ascertained by first measuring the binding of 125I-anti-C3c at the EAC423b stage and subsequently to the washed cells following the treatment with sCR1 and factor I. In all cases, binding of the 125I-anti-C3c was reduced to background levels following treatment with sCR1 and factor I, whereas specific binding of a 125I-labeled rabbit anti-C3d to the cells was retained.
Rosetting of iC3b-coated erythrocytes to CR2-bearing Raji cells
Raji cells (5 × 104) were incubated with 4 × 106 EAC423bi in 50 μl of the low ionic strength buffer, SGVB, for 30 min at 37°C with gentle rotation. Cells were fixed with 0.2% glutaraldehyde for 5 min before quenching with 20 mM Tris-HCl (pH 8.2). Rosette formation was evaluated in duplicate by assessing the percentage of Raji cells bearing four or more erythrocytes. Binding specificity was determined by incubating Raji cells with EAC42 and also by preincubating Raji cells with 4 μg/ml of OKB7, an IgG2b functional site blocking anti-human CR2 mAb, (provided by Dr. P. Rao, Ortho Diagnostics Systems, Raritan, NJ) before the addition of iC3b-coated erythrocytes. The IgG2b anti-CR3 mAb OKM1 (provided by Dr. W. Reed, University of North Carolina, Chapel Hill, NC) was used as a nonspecific Ab control. Rosetting of C3dg-coated erythrocytes to Raji cell CR2 was conducted exactly as described for the iC3b-coated erythrocytes.
Rosette inhibition by purified iC3b and C3dg was performed as above for the OKB7 blocking mAb experiments except that instead of OKB7, the Raji cells were preincubated with various concentration of purified iC3b or C3dg before the addition of EAC423bi cells made with sufficient purified C3 to yield ∼80–90% rosette formation in the absence of inhibitor.
Binding of soluble CR2 to iC3b-coated erythrocytes
Culture supernatants from J558L-pSNRCR2 cells were dialyzed against DVB. The amount of secreted (CR2)2-IgG1 was determined by ELISA, using plates coated with a rabbit Ab to mouse IgG (Jackson ImmunoResearch, West Grove, PA) and a sandwiching rabbit Ab to mouse IgG that is conjugated to alkaline phosphatase (Jackson ImmunoResearch). Purified mouse IgG (Sigma) was used to obtain a standard curve.
For the binding assay, 50 μl culture supernatant containing 4 μg/ml (CR2)2-IgG1, were incubated overnight at 4°C with 5 × 106 iC3b-coated erythrocytes. Binding specificity was established by preincubating (CR2)2-IgG1 with 40 μg/ml OKB7 (or OKM1 as an isotype-matched negative control) at 37°C for 1 h. Erythrocytes were then washed with DGVB and binding was detected by using a fluorescein-conjugated donkey Ab to mouse IgG (Jackson ImmunoResearch). After 1 h of incubation at 4°C, cells were washed and resuspended in 1 ml DGVB. Cell-bound fluorescence intensity was determined using a FACScalibur flow cytometer (Becton Dickinson) and analyzed as a function of cell number using the CellQuest software package (Becton Dickinson).
Results and Discussion
As mentioned in the Introduction, we considered the acidic residue-lined pocket on the concave surface of the C3d molecule to be a candidate CR2 binding site because of its accessibility, because of the known ionic strength dependence of the CR2-C3d interaction, and because a molecular model of the binding domains of CR2 indicated the presence of a fairly extensive electropositive surface (20). To test the validity of our candidate binding site, we have performed an alanine mutagenic scan on all of the C3d residues, both charged and neutral, that have surface-exposed side chains lying within or just adjacent to the boundaries of the negatively charged pocket visualized in the electrostatic surface rendition of the molecule (20). These mutations were performed within the context of the intact human C3 molecule, and following transient expression in COS-1 cells of the mutant C3 cDNAs, the culture supernatants served as a source of rC3 for deposition onto C3 convertase-bearing sheep red cells (EAC42). The deposited C3b molecules were then converted to iC3b by treatment of the cells with excess H and I and these EAC423bi cells were then used in rosetting assays with Raji cells that bear CR2, but no other CRs on their surfaces. We have previously confirmed in rosetting assays that CR2 on Raji cells binds iC3b and C3dg with similar avidities (23). Because conversion of the red cell-deposited C3b to C3dg either requires considerably more H and I than is required for conversion of C3b to iC3b, or it requires the use of sCR1 as an I cofactor, for reasons of reagent economy, in most of our experiments we have measured the effect of our mutations on the iC3b-CR2 interaction, but have interpreted the results within the context of the structure of C3d. Further justification of this approach comes from a rosette inhibition assay using as fluid-phase competitors purified iC3b and C3dg. It can be seen in Fig. 1⇓ that on a molar basis these two ligands were equipotent in inhibiting rosette formation between iC3b-coated red cells and CR2-bearing Raji cells.
Comparison of the ability of purified iC3b and C3dg (and C4 as a negative control) to inhibit rosette formation between iC3b-coated red cells and CR2-bearing Raji cells. The solid line at the top indicates the level of rosette formation in the absence of inhibitor. The range of duplicate rosette counts within the assay is indicated by the error bars.
The molecular locations of the 17 residues targeted for mutation in this study are shown in Fig. 2⇓, together with a listing of the 27 mutants that were analyzed. The numbering system used refers to the C3d residue number because this most readily allows interpretation with reference to the C3d structure (add 971 to convert to mature C3 numbering). In addition to the 17 single-residue mutants constructed, acidic residues within each of the 36–39 and 160–167 segments were mutated in several combinations. In two cases, mutants D36N/E37Q/E39Q and D163N, the effects of isosteric amide substitutions were also examined, and R49 was replaced by both alanine and the more isosteric residue methionine. With the exception of the R49A mutant, which was expressed at <100 ng/ml, the expression levels of all of the other mutant C3 molecules fell within the range (250–450 ng/ml) observed for the wild-type molecule in the numerous replicate transfections conducted in the course of this study. Because the more isosteric R49 M mutant was expressed at normal levels, this permitted the evaluation of the role of this residue in mediating CR2 binding. The only other anomaly noted was for the isosteric amide substitution mutant D163N that, although secreted at wild-type levels, was only poorly deposited on the C3 convertase-bearing red cells. In contrast, the D163A mutant was normal with respect to both its secretion level and deposition efficiency.
A list of the 27 single and combination mutants analyzed in this study (A) and their location (B) as depicted on a space-filling model of the concave surface of C3d (20 ). The figure was produced using the program RasMol 2.6 and the numbering system is that employed in the C3d structure’s Brookhaven Protein Data Bank file 1C3D. Acidic, basic, and neutral side chain-containing amino acids are respectively colored red, blue, and yellow. The orientation is the same as the figure showing the electrostatic surface potential of this face of the molecule (figure 3⇓ of Ref. 20).
Characterization of the C3-bearing red cells
Because it is known that there is a binding site for factor H within C3d (41, 42), and further that red cells on which the deposited C3b was not converted to iC3b would show very much impaired ability to form CR2-dependent rosettes (23), it was important to establish that none of the mutations that we introduced prevented the factor H- and I-mediated conversion of the cell-associated C3b to iC3b. To examine this issue, culture supernatants from transfected COS-1 cells that had been metabolically labeled with [35S]Met/Cys were used to build the red cells bearing the various mutant C3b molecules. These cells were then treated with factors H and I under our standard conditions to convert the deposited C3b to iC3b. Following hypotonic lysis and recovery of the membrane ghosts bearing ester-linked adducts of 35S-labeled iC3b (or C3b if the conversion had not worked), hydroxylamine was added to break the ester linkages and the chain structure of the released 35S-labeled protein was assessed by SDS-PAGE autofluorography. The results of this experiment are shown in Fig. 3⇓. As a reference point, the hydroxylamine-released chains of the wild-type protein are shown both before and after treatment with factors H and I (right side of upper panel). Whereas the former shows predominantly bands migrating at the positions of α′-chain and β-chain, the latter shows the chains characteristic of iC3b, namely β, α′-67, and a doublet of α′-43/α′-40, where the doublet represents the respective products of the first and second factor I-mediated cleavages. It has long been assumed, and now has been directly demonstrated that it is the first factor I-mediated cleavage that correlates with the functional status of the molecule (43). It can further be seen in Fig. 3⇓ that all of the mutants examined in our study displayed wild-type-like behavior with respect to extent of deposition and conversion to their respective iC3b forms.
Demonstration that the mutant C3b molecules deposited on the C3 convertase-bearing red cells were convertible to iC3b by factors H and I. This experiment was performed using ∼600 ng of [35S]Met/Cys metabolically labeled rC3 to build the EAC423bi cell under our standard conditions (see Materials and Methods). Hydroxylamine treatment (1 M, pH 9) of the detergent-solubilized membranes of these cells released the ester-linked C3 activation fragments, which were then analyzed by SDS-PAGE (reducing conditions) and autoflurography. As a reference to indicate the migration position of intact α′-chain (and β-chain), the far right of the upper panel shows the results from EAC423b cells made with 35S-labeled recombinant wild-type C3, but not treated with H and I so as to preserve the deposited molecule as C3b. The complete conversion to iC3b of all of the mutants tested in this study is indicated by the absence of intact α′-chain and the appearance of the α′-67 and α-43/40 doublet that are the characteristic chains of iC3b.
Rosetting of wild-type and mutant iC3b-coated red cells to CR2-bearing Raji cells
Aliquots of C3 convertase-bearing red cells were treated with various amounts of recombinant C3-containing supernatants to create sets of cells with various amounts of iC3b on their surfaces. Within an experiment, the relative amount of iC3b deposited on each group of red cells was determined by measuring the binding of 125I-labeled anti-C3c to a fixed number of red cells. Fig. 4⇓ presents the rosette formation dose response curves for the various substitution mutants engineered for this study. Each panel represents a separate experiment in which the dose-response curve for the wild-type molecule acts as an internal reference point. These data are representative of minimally two, and often three, repeat experiments for every mutant examined in this study. The data were also plotted using a semilogarithmic scale for the x-axis, which denoted the relative amount of iC3b ligand per red cell (not shown). The horizontal displacement of the psuedo-linear part of a given dose-response curve from that of the wild-type curve allows one to estimate the extent of the defect or enhancement in CR2-binding introduced by any given mutation. This in turn allowed us to summarize in Fig. 5⇓ the results of all of these rosette experiments in a more convenient bar graph format.
Rosette formation between CR2-bearing Raji cells and sheep red cells bearing various amounts of recombinant iC3b. The latter is indicated as cpm 125I-labeled anti-C3c bound per aliquot of 5 × 107 iC3b-bearing red cells. The solid line near the bottom of each panel represents the background rosetting observed in control experiments employing EAC42 cells. The presence in the assay of the CR2-specific, functional site-blocking mAb OKB7 yielded a similar level of background rosetting when EAC423bi cells bearing high levels of wild-type iC3b were employed as the source of CR2 ligand. A recombinant wild-type control dose-response curve was included in all experiments. The error bars indicate the range of duplicate rosette counts (i.e., separate wet mount samplings) within an experiment. Data for the single and multiple alanine-substitution mutants of the 36–39 charged residue cluster are shown in A and B, respectively. Data for the single and multiple alanine-substitution mutants of the charged residues within the 160–167 cluster are shown in C and D, respectively. E–G, The results for the remainder of the mutants tested.
Bar graph summary of the CR2 binding activities of all of the mutants engineered for this study as determined from the CR2-dependent rosette assay. The percentage of wild-type CR2 binding activity of the mutant iC3b molecules was calculated from the horizontal displacement of the pseudo-linear part of a given dose-response curve from that of the wild-type curve on a semilogarithmic plot (relative amount of iC3b per cell is the log axis). Results from at least two (often three) independent experiments involving separate transfections were averaged. The error bars represent the range of activities observed in the replicate measurements.
We initially focused our attention on the two most prominent clusters of acidic residues on the concave end of the molecule because the hypothesis we were testing predicted that either or both of them should serve as a contact for the electropositive surface contributed by the first two SCR domains of CR2. The D36/E37/E39 cluster lines one edge of the negatively charged depression visible in the electrostatic surface potential rendering of the C3d molecule (see figure 3⇑ in Ref. 20), D163 and E166 line the opposite edge, and the carboxylate side chain of E160 extends down into the depression (Fig. 2⇑). Dealing first with the D36/E37/E39 cluster (Fig. 4⇑, A and B, and Fig. 5⇑), it can be seen that whereas mutation of D36 on its own is without effect, both the E37A and E39A mutant molecules show about a 2-fold defect in CR2 binding activity in the rosette assay. The combination E37A/E39A mutant molecule was more than 4-fold compromised in CR2-binding activity and mutation of D36A in combination with E37A somewhat magnified the defect caused by the mutation of E37A alone. Although the combination mutant D36A/E39A showed a similar CR2-binding defect as did the single E39A mutant, the D36A/E37A/E39A triple mutation, which would effectively remove the negatively charged rim from that side of the depression, yields a molecule that displays only about 10% of the wild-type CR2-binding activity. To further test the importance of the negative charges per se, and at the same time address the issue of side chain stereochemistry that becomes more of a factor when analyzing the results of a triple mutation, the isosteric amide version of this mutant was engineered. Like its alanine-substituted counterpart, the D36N/E37Q/E39Q mutant was also found to be essentially inactive in mediating CR2 binding. Finally, alanine substitutions were made at E42, an acidic residue that protrudes up at the extreme edge of this surface of the molecule, and D292, which extends the negative surface at the other edge of this rim of acidic residues (Fig. 2⇑). In each case we found that the resultant mutant molecules retained between 80 and 90% of wild-type CR2-binding activity (Figs. 4⇑, E and F, and 5). Collectively, these data suggest that the side chains of the D36/E37/E39 cluster make important ionic interactions with CR2 and the contacts with E37 and E39 appear to be the most important within this grouping of acidic residues.
We next turned our attention to the acidic residues along the opposite rim of the depression, which encompasses residues 160–167 of the primary sequence (Fig. 2⇑). Also within this surface patch, toward the edge of the cavity rim, is the hydrophobic side chain of I164 and protruding away from the rim, but in between acidic residues D163 and E166, is the basic side chain of K162 (Fig. 2⇑). This region appears to be extremely important for the interaction between C3d and CR2 as even single-alanine substitutions were sufficient to decrease CR2 binding levels to <5% of wild-type for D163 and I164 and to ∼10% of wild-type for E166 (Figs. 4⇑, C and E, and 5). D163 was also mutated to its isosteric amide, and although the deposition efficiency onto the C3 convertase-bearing cells was inexplicably lower for the D163N mutant protein than for any of the other mutants examined, by using more concentrated culture supernatants, sufficient deposition was achieved to ascertain that it was the negative charge per se at the D163 position that was required to mediate binding to CR2 (Fig. 5⇑). The mutation to alanine of D160, which lies at the bottom of the cavity, had a somewhat lesser effect, but still reduced CR2 binding 4-fold relative to wild-type. In contrast, E167, which relative to the cavity is the most distally located acidic residue in this segment, could be mutated to alanine with minimal effect on CR2-binding activity. One of the most interesting results was obtained with the K162A mutant protein as this molecule displayed a 2-fold enhancement in CR2 binding activity. Our interpretation of this result is that the positive side chain of K162 partially inhibits the binding to this region of C3d of the electropositive surface contributed by CR2, so its removal results in increased binding of CR2. In keeping with this interpretation, we have found that in contrast to the almost complete absence of CR2-binding activity displayed by the D163A mutant, the double mutant K162A/D163A retains much of the CR2-binding activity of the wild-type protein (Fig. 4⇑D and 5). It would seem that to overcome the charge repulsion caused by K162, contact with acidic residues D163, E166, and perhaps E160 is necessary. However, in the absence of the K162 positive charge, the ionic contribution of at least D163 to the binding interface with CR2 becomes dispensable.
When mutations to the acidic residues in the 160–167 segment were conducted in combination, the molecules were generally defective in CR2 binding (Fig. 4⇑D and 5). For example, the residual binding of the E160A mutant is abrogated in the double mutant E160A/D163A and the combined mutation to alanine of D163 and E166, both of which were largely inactive as single mutants, does not result in the acquisition of any substantial new CR2 binding activity. The one exception to this general rule is that the combination of E166A, which was largely inactive as a single mutant, with E167A, which as a single mutant retained a near wild-type level of activity, yielded a molecule retaining ∼30% of wild-type CR2-binding activity. This result is difficult to explain, but it is possible that in the absence of the mutual charge repulsion between the carboxylates of D163, E166, and E167, the side chain of D163 may reorient so as to partially accommodate the loss of the ionic contribution of E166 to the binding energy.
Fig. 4⇑, E and F, shows the rosette assay results for six additional single-residue mutants examined in this study (V97A, N98A, R49 M, K251A, K291A, and Y201A). It can be seen that these mutations were found to have relatively minor effects on CR2 binding activity, the largest effect being seen for the K251A mutant, which retains about 60% of wild-type activity. From these data we conclude that CR2 appears to primarily bridge between contacts in the D36/D37/D39 cluster and the E160/D163/I164/E166 cluster. Because V97 and N98 essentially form a surface bridge between these two contact clusters, it was somewhat surprising that their mutation had virtually no effect on CR2-dependent rosette formation.
A subset of the mutant molecules was also examined in the context of C3dg for their binding activity to CR2 as assessed using the rosette assay. Our purpose was to ensure that the major defects in CR2-binding activity that we had observed in the context of iC3b were not due to a secondary effect of a particular mutation promoting an unfavorable interaction between the C3d and C3c regions of the molecule such that the CR2 binding site of C3d became sterically masked. The results of this experiment are shown in Fig. 6⇓. It can be seen that the relative CR2-binding activities observed in the case of the ligand being C3dg largely mirrored those obtained using iC3b as the CR2 ligand. The only substantial exception noted was that the extent of the defect for the E166A mutant was less than what was observed in the context of iC3b (i.e., 3- to 4-fold as C3dg vs 10-fold as iC3b) and was now similar to the defect observed for the E160A mutant. Thus, although for the E166A mutant there appears to be some superimposed secondary effect, this is not the case for any of the other mutants showing a CR2-binding defect. Therefore, these results lend further support to the concept that the CR2-contacting residues putatively identified in this study are essentially equally accessible for CR2 binding regardless of whether the ligand is iC3b or C3dg.
Rosette formation between CR2-bearing Raji cells and sheep red cells bearing various amounts of recombinant C3dg. The latter is indicated as cpm 125I-labeled anti-C3d bound per aliquot of 5 × 107 C3dg-bearing red cells. The background level of rosette formation in each experiment was determined using EAC42 cells and is indicated by the solid line at the bottom of the figure. The average of duplicate measurements is plotted and the range is indicated by the error bars.
Binding of soluble CR2 to wild-type and mutant iC3b-coated red cells
Because the rosette assay is highly dependent on avidity effects, i.e., there are many potential points of contact between the ligand-bearing red cell and the receptor-bearing Raji cell, it is possible that this assay would miss the effects of mutations that bring about relatively minor changes to the intrinsic affinity governing the interaction between individual iC3b and CR2 molecules. This may be especially true for some of the single-residue mutations examined in the previous section. Although not completely devoid of avidity effects, an assay measuring the binding of a bivalent soluble form of CR2 to iC3b-coated red cells should be less prone to such avidity “masking” effects because the ligand-receptor contacts are maximally bivalent. Accordingly, red cells bearing various amounts of wild-type or mutant iC3b were prepared and incubated with a fixed concentration of soluble CR2 in the form of a (CR2)2-IgG fusion protein (3, 23). Binding of the soluble CR2 to the cells was revealed by an FITC-conjugated anti-mouse IgG in conjunction with flow cytometric analysis. The relative amount of iC3b ligand per cell was determined as before using 125I-labeled anti-C3c and background fluorescence levels were determined using EAC42 cells. The results of such assays employing all of the single mutants engineered for this study are shown in Fig. 7⇓. In general, the results were concordant with the rosette assay. In no case did mutants that showed severe defects in the rosette assay not also show severe defects in the soluble CR2 binding assay; however, as expected, the effects of some mutations were more severe in the soluble CR2 binding assay. For example, whereas the D36A and E42A mutations were without effect in both assays, near complete CR2 binding impairment was seen in mutants E37A and E39A (Fig. 7⇓A), this contrasting with the 2-fold defect seen in the rosette assay for these latter two mutants. The data in Fig. 7⇓B further confirm the enhancing effect of the K162A mutation noted in the rosette assay. Additionally, the E160A mutation on its own was sufficient to completely abrogate CR2-binding activity measured by this assay, whereas this mutant showed about 25% residual activity in the rosette assay. The small decreases in CR2 binding detected by the rosette assay for mutants D292A and K251A were confirmed and mutation of V97, a nonpolar residue lying immediately adjacent to the critical I164, gave rise to an ∼2-fold defect in the soluble CR2 binding assay (Fig. 7⇓D).
Flow cytometric measurement of the binding of soluble CR2 (i.e., (CR2)2-IgG fusion protein) to sheep red cells bearing various amounts of recombinant iC3b. The mean fluorescence was plotted as a function of the relative number of iC3b molecules per red cell, as determined by measuring the binding of 125I-labeled anti-C3c to a fixed number of iC3b-coated red cells. Background fluorescence (solid line at bottom of each panel) was determined by measuring the binding of soluble CR2 to EAC42 cells. Only the single substitution mutants were tested in this assay. A, Results for charged residues located in the 36–42 segment. B, Results for charged residues in the 160–167 segment. C and D, Results for the remainder of the mutants tested. The average of duplicate measurements is plotted and the range is indicated by the error bars.
Concluding remarks
We believe that the results of our experiments conclusively show that two predominantly acidic surface patches provide crucial contacts for the interaction with the C3d-binding domains of CR2 (i.e., CR2 SCR domains 1 and 2). Of the two acidic clusters, E37/E39 and E160/D163/E166, which are located on opposite sides of the depression on the concave surface of C3d, the latter appears to be the more important based on the fact that single mutations within it were sufficient to almost completely abrogate CR2 binding. The contribution of the hydrophobic side chain of the adjacent residue I164 also appears to be vital. The modeled structure of the C3d-binding domains of CR2 (17, 20) indicates that the adjacent residues R89 and H90 are part of one electropositive surface and that R13 is part of another. The significance of these residues is that they are contained within the CR2-derived synthetic peptide segments that were shown to be inhibitory to the CR2-C3d interaction (17). Confirmation that ionic contacts are actually formed between the two acidic patches of C3d identified in this work and the two basic patches suggested by the model of the binding domains of CR2 of course awaits the achievement of the structure of a cocrystal of C3d complexed to SCRs 1 and 2 of CR2.
It is noteworthy that the negative charges of two of the residues that we have identified as being key contributors to the binding interface, specifically E39 and D163, are not conserved in cobra or trout C3, but they are conserved in mouse C3. Whereas cobra and trout iC3b/C3d are known not to be able to interact with human CR2 (44), mouse C3d binds well to human CR2 (45). Additionally I164, which is conserved in human and mouse C3, is replaced by valine in cobra and trout C3 and E166 is replaced by lysine in cobra C3. Based on our current work, these substitutions on their own would account for the lack of CR2 binding by trout and cobra C3. However, this does not preclude the possibility of there being contacts to other regions of the C3d molecule in which there is further nonconservation of key contact residues among human CR2-binding and nonbinding species of C3.
As mentioned in the Introduction, at one point a peptide segment corresponding to C3d residues 228–239 (1199–1210 mature C3 numbering) had been thought to be a major contact region for the interaction with CR2. However, we have previously reported (23) that quite extensive substitution of this segment led to a fairly minimal defect in CR2 binding activity and thus at best, it plays a fairly minor role in mediating the C3d-CR2 interaction. Nevertheless, the coupling of a single copy of a peptide encompassing this segment to the variable region of an Ab was recently reported to have significant molecular adjuvant properties in the production of anti-Id Abs against the immunizing Ig (46). Given this result, one might predict that a similar approach focusing on the more crucial peptide segments identified in the present work would produce in an even greater molecular adjuvant effect. We are currently pursuing such investigations.
Acknowledgments
We thank Russell G. Jones, currently of the Department of Medical Biophysics, University of Toronto, for his contributions to this study. We also thank Cheryl Smith, of the Department of Immunology, University of Toronto, Flow Cytometer Facility for her assistance in carrying out the flow cytometric measurements that we have reported.
Footnotes
-
↵1 This work was supported by funding from Aventis-Pasteur in the form of a Canadian Universities Research Program Grant and by the Medical Research Council of Canada (MT-7081).
-
↵2 Current address: Department of Biopathology and Bio-medical Technologies, University of Palermo, Corso Tukory 211, Palermo 90134, Italy
-
↵3 Address correspondence and reprint requests to Dr. David E. Isenman, University of Toronto, Department of Biochemistry, Medical Sciences Building, Room 5306, Toronto, Ontario M5S 1A8, Canada. E-mail address: d.isenman{at}utoronto.ca
-
↵4 Abbreviations used in this paper: CR, complement receptor; DVB, dextrose-Veronal buffer; DGVB, dextrose-gelatin-Veronal buffer; EAC4b2a, EAC423b, EAC423bi, hemolysin-sensitized sheep erythrocytes coated respectively with complement components C4b and C2a, C4b, C2a and C3b, C4b, C2a and iC3b; HEL, hen egg lysozyme; SCR, short consensus repeat; sCR1, soluble CR 1; SGVB, sucrose-gelatin-Veronal buffer; VBS, Veronal-buffered saline; BCR, B cell receptor.
- Received May 15, 2000.
- Accepted July 18, 2000.
- Copyright © 2000 by The American Association of Immunologists