CD21)–C3d Interaction Reveals a Putative Charged SCR1 Binding Site for C3d

doi:10.1016/j.jmb.2004.12.007

J. Mol. Biol. (2005) 346, 845–858

Mutational Analysis of the Complement Receptor Type 2 (CR2/CD21)–C3d Interaction Reveals a Putative Charged SCR1 Binding Site for C3d Jonathan P. Hannan1, Kendra A. Young1, Joel M. Guthridge2 Rengasamy Asokan1, Gerda Szakonyi3, Xiaojiang S. Chen3 and V. Michael Holers1* 1

Department of Medicine and Immunology, University of Colorado Health Sciences Center, 4200 East Ninth Ave. Denver, CO 80262, USA 2

Arthritis and Immunology Program, Oklahoma Medical Research Foundation, 825 N.E. 13th Street, Oklahoma City OK 73104, USA 3

Department of Biochemistry and Molecular Genetics University of Colorado Health Sciences Center, 4200 East Ninth Ave., Denver, CO 80262 USA

We have characterized the interaction between the first two short consensus repeats (SCR1-2) of complement receptor type 2 (CR2, CD21) and C3d in solution, by utilising the available crystal structures of free and C3d-bound forms of CR2 to create a series of informative mutations targeting specific areas of the CR2–C3d complex. Wild-type and mutant forms of CR2 were expressed on the surface of K562 erythroleukemia cells and their binding ability assessed using C3dg-biotin tetramers complexed to fluorochrome conjugated streptavidin and measured by flow cytometry. Mutations directed at the SCR2–C3d interface (R83A, R83E, G84Y) were found to strongly disrupt C3dg binding, supporting the conclusion that the SCR2 interface reflected in the crystal structure is correct. Previous epitope and peptide mapping studies have also indicated that the PILN11GR13IS sequence of the first inter-cysteine region of SCR1 is essential for the binding of iC3b. Mutations targeting residues within or in close spatial proximity to this area (N11A, N11E, R13A, R13E, Y16A, S32A, S32E), and a number of other positively charged residues located primarily on a contiguous face of SCR1 (R28A, R28E, R36A, R36E, K41A, K41E, K50A, K50E, K57A, K57E, K67A, K67E), have allowed us to reassess those regions on SCR1 that are essential for CR2–C3d binding. The nature of this interaction and the possibility of a direct SCR1–C3d association are discussed extensively. Finally, a D52N mutant was constructed introducing an N-glycosylation sequence at an area central to the CR2 dimer interface. This mutation was designed to disrupt the CR2–C3d interaction, either directly through steric inhibition, or indirectly through disruption of a physiological dimer. However, no difference in C3dg binding relative to wild-type CR2 could be observed for this mutant, suggesting that the dimer may only be found in the crystal form of CR2. q 2004 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: complement; short consensus repeats; mutagenesis; flow cytometry; tetramers

Introduction

Abbreviations used: CR2, complement receptor type 2; CCP, complement control protein; SCR, short consensus repeat; PBS, phosphate-buffered saline; IPTG, isopropylb-D-thiogalactoside; FITC, fluorescein isothiocyanate; PE, phycoerythrin; MFI, mean fluorescence intensity; SEM, standard error of the mean. E-mail address of the corresponding author: [email protected]

Complement receptor 2 (CR2 or CD21) is a 145 kDa type I transmembrane protein found primarily on the surface of mature B cells, follicular dendritic cells and some T lymphocytes. On B cells, CR2 is found in a CR2/CD19/CD81 complex or in a CR2/CR1 complex and plays an integral role in cell activation and the initiation of normal immune responses.1–4 Interaction of foreign antigen coated with C3d and CR2 results in a cell-signaling event occurring through CD19 in which the signaling

0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.

846 threshold for B cell activation is lowered via a signal transduction cascade.5 CR2 has also been demonstrated to play a critical role in the development of humoral immune responses to T-dependent antigens and thus acts as a bridge between the adaptive and innate immune systems.6–8 CR2 is a member of the structural family of C3/C4 receptor and regulatory proteins known as the regulators of complement activation (RCA family). Members of this family are characterized by the presence of repeating modules known as short consensus repeats (SCRs), also called complement control protein (CCP) or Sushi domains. The molecular structure of CR2 is well known and comprises a 15 or 16 SCR extracellular domain, a 24 amino acid residue transmembrane domain, and a short 34 amino acid residue intracellular tail. Each SCR unit is about 60–70 amino acid residues in length, and variable linker regions of three to eight amino acid residues separate neighboring units. All SCRs contain a number of conserved residues including four cysteine residues and an invariant tryptophan residue. The conserved cysteine residues form a pattern of disulfide bridges connecting Cys1–Cys3 and Cys2–Cys4. The primary ligand for CR2 is the C3d activation fragment of complement component C3, although the larger C3dg and iC3b fragments of C3 also bind CR2 with high affinity.9,10 CR2 also serves as a receptor for CD23,11,12 interferon a,13 and the Epstein–Barr virus (EBV) glycoprotein gp350/ 220.14,15 The two amino-terminal SCRs (SCR1-2) of CR2 are responsible for the binding of these ligands, although an additional glycosylation-dependent interaction with the low-affinity IgE receptor, CD23, involves SCR5-8.12 The X-ray crystal structure of human CR2 SCR1-2 ˚ in its complex with C3d has been elucidated at 2.0 A resolution (Figure 1(a)), the first such ligand– receptor pair in the RCA family for which a cocrystal has been generated.16 Crystal structures of human forms of free CR2 SCR1-2 and C3d have also been determined.17,18 The structure of CR2 SCR1-2 in both the C3d-bound and unbound states reveal that both SCR domains are in the compact b-barrel configuration characteristic of these proteins, and that both domains are packed against each other in a tight V-shape. The interface region between the two SCRs appears to be lined with non-polar and hydrophobic residues (Figure 1(b)). However, our recent analytical ultracentrifugation studies and X-ray scattering experiments, in conjunction with previous electron microscopy data, appear to dispel the ready conclusion that SCR1 and SCR2 exhibit the same side-side packing in solution, and instead indicate that CR2 SCR1-2 in solution is in fact in a much more extended conformation than the two crystal structures suggest.19–21 Prota et al. also suggest that the presence of a large N-glycan moiety attached to N107 of SCR2 would sterically inhibit the close packing of SCR1 and SCR2 (Figure 1(b)); notably both the bound and free forms of CR2 were deglycosylated prior to crystallization.17

Mutational Analysis of the CR2–C3d Interaction

Figure 1. (a) Ribbon representation of the CR2 SCR1-2– ˚ resolution (PDB C3d co-crystal structure at 2.0 A accession no. 1GHQ). The C3d molecule is represented in orange and yellow, and is folded into a distinctive a-helical barrel structure. The two CR2 molecules making up the CR2 homodimer are shown in cyan and red, respectively, and exhibit characteristic SCR b-barrel folds. SCR1 and SCR2 of each molecule are indicated. The numbering system used here for CR2 is based on the reported sequence for mature CR2 and differs by one residue from that reported in the co-crystal structure.38 Q20, which is the site of antigen attachment on C3d, is indicated. This Figure and subsequent molecular illustrations were generated using MOLMOL,39 unless otherwise specified. (b) The structure of free CR2 SCR1-2 (PDB accession no. 1LY2). Aromatic and non-polar residues involved in the hydrophobic packing of SCR1 and SCR2 against each other, and in the ordered linker region between the two SCRs are shown. Also indicated are two N-glycan moieties attached to N101 and N107, respectively. In the crystal structure of free CR2 the N-glycan attached to N107 exhibits a H bond with the carbonyl group of D56.

With regards to the structural elucidation of C3d, in both the CR2-bound and unbound states, little controversy exists. Both forms of C3d show the same dome-shaped a-helical barrel motif

847

Mutational Analysis of the CR2–C3d Interaction

comprising a core of six parallel helices, surrounded by a second set of six parallel helices running antiparallel to the core. An N-terminally truncated form of rat serum C3d has also been determined by X-ray diffraction and this closely resembles the globular structures described above.22 Subsequent to the structural determination of free C3d, mutagenesis screening studies identified a negatively charged region on the concave surface of C3d that appeared to be important in CR2 binding.23 Two patches of acidic residues in particular were seen to play a critical role, E37/E39 and E160/ D163/E166. These data seemed to be in agreement with a number of studies that show that the CR2– C3d association is dependent on pH and ionic strength.19,21,24 However, the determination of the CR2–C3d co-crystal complex revealed a number of contradictory and unexpected features about CR2– C3d binding. Important structural features of this interaction included the presence of a novel C3d binding site on CR2, where only SCR2 makes direct contact with C3d at a site spatially discrete from the negatively charged concave surface. Previous studies have indicated that both SCR1 and SCR2 are necessary for C3d binding to occur,25–28 although it has been proposed that SCR1 may contribute allosterically to the binding interaction.17 Furthermore, in the co-crystal structure, networks of hydrogen bonds between main-chain atoms and well-ordered water molecules, together with hydrophobic and van der Waals interactions, dominate complex formation between the two molecules. A last significant feature of the CR2–C3d structure is the presence of a CR2 homodimer formed by hydrogen bonds between sites within SCR1 of each CR2 molecule. To further study whether the CR2–C3d co-crystal structure is an accurate reflection of complex formation in solution, we have undertaken a series of informative mutagenesis studies on human fulllength 15 SCR forms of CR2 expressed on the surface of K562 erythroleukemia cells. Relative binding affinities of wild-type and mutant forms of CR2 were assessed using multi-color flow cytometry. Fluorescein isothiocyanate (FITC) conjugated monoclonal antibody (mAb) HB5 was used to divide each population of cells according to high, medium or low CR2 expression,26,29 and concurrently, binding ability of the various forms of CR2 was measured using tetramers of recombinant human C3dg-biotin complexed to phycoerythrin (PE) conjugated streptavidin.30 Data were recorded using two differing sets of conditions: firstly, aliquots of CR2-expressing cells were separately incubated with serial dilutions of a stock solution initially containing 2 mg of C3dg-biotin and 0.4 mg of PE streptavidin. To verify that any diminishing effects on CR2 binding were related to the C3dgbiotin concentration, and not simply due to the decreasing concentration of the PE streptavidin we also separately incubated aliquots of CR2-expressing cells in the presence of a fixed concentration of PE streptavidin (0.4 mg), but with various

concentrations of C3dg-biotin (2, 1, 0.5, 0.25, 0.125 and 0.0625 mg). Features of the CR2–C3d co-crystal complex were used to direct mutations on CR2, in particular, around the SCR2–C3d interface (R83A, R83E, G84Y) and at the SCR1–SCR1 dimer interface (D52N). Additional crystal structure driven mutations were selected on SCR1 in close spatial proximity to, and including R13 (N11A, N11E, R13A, R13E, Y16A, S32A, S32E). R13 is part of the PILN11GR13IS sequence of the first inter-cysteine region of SCR1 and has previously been implicated as a contact site for iC3b.28,31 Further analysis of the free and bound forms of CR2 reveals that one face of CR2 is rich in positively charged residues. As previous studies have indicated that the CR2–C3d interaction is likely to be charge-dependent in nature, it was decided to selectively target a number of lysine and arginine residues on SCR1 likely to be involved in the generation of salt-bridges, and/or in the formation of long-range pseudospecific electrostatic interactions.32 Accordingly, CR2 mutants containing R28A, R28E, R36A, R36E, K41A, K41E, K50A, K50E, K57A, K57E, K67A, and K67E were constructed.

Results SCR2–C3d interaction As most of the interactions between SCR2 and C3d that are observed in the co-crystal structure are main-chain in nature (Figure 2) it was difficult to design a substantial number of mutations on SCR2 that adequately map the extent of C3d association in this area. However, structural analysis of the CR2–C3d complex reveals that R83 plays a critical role in the binding interaction with C3d, where the

Figure 2. Detailed structure of the CR2–C3d interface. Shown are the interactions between SCR2 and C3d. Amino acid residues whose side-chain and main-chain oxygen (red) and nitrogen (blue) atoms are involved in hydrogen bond formation (represented by black broken lines) with each other and with ordered water (purple) molecules are indicated.

848 side-chain of this residue is inserted into a polyanion hole formed by the carbonyl groups of I115, L116, E117, and Q119 of C3d. Accordingly, R83 was selected as an obvious and important candidate for substitution. Mutation of R83 to alanine (R83A), or to glutamic acid (R83E) was found to result in significant decreases in C3dg binding ability relative to wild-type CR2 (Figure 3(a)). Progressively deleterious effects on C3dg binding are seen with the mutation of the positively charged arginine residue to a non-polar alanine residue, and then to a negatively charged glutamic acid, a trend that appears consistently for most of the substitutions reported throughout this study. Further evidence that R83 is central to the SCR2– C3d association in solution, as well as in the crystalline state, is provided by the mutation of

Mutational Analysis of the CR2–C3d Interaction

G84 to tyrosine (G84Y). The introduction of the bulky side-chain at this site appears to sterically interfere with the binding interaction between the two molecules in such a way that only a minimal amount of C3dg binding is observed (Figure 3(b)). The two mutations reported here provide support for the hitherto unanticipated SCR2–C3d interface seen in the co-crystal structure, which is located away from the previously proposed CR2 binding site located on the concave surface of C3d.23 Mutations centered around the PILN11GR13IS sequence on SCR1 Previous epitope and peptide mapping studies have identified the P8-S15 sequence on SCR1 (Figure 4) as being important in the association of

Figure 3. Flow cytometric analysis of mutations targeting the SCR2–C3d interaction. Binding affinities of C3dg-biotin tetramers to (a) R83A, R83E and (b) G84Y mutants relative to wild-type CR2 are given. The maximal mean fluorescence intensity value for wild-type CR2 was set at a relative value of 100 and the other samples normalized to this value. Graphs representing serial dilutions of a stock mixture containing 2 mg of C3dg-biotin and 0.4 mg of PE streptavidin (left), and for varying concentrations of C3dg-biotin (2, 1, 0.5, 0.25, 0.125 and 0.0625 mg) with 0.4 mg of PE streptavidin (right), are shown. The average and SEM of the normalized values for the mean fluorescence intensity (MFI) are demonstrated. At least three separate experiments were performed for each mutation. Data given here are for K562 cells exhibiting medium expression of CR2, but similar relative affinities were observed for cells expressing high and low levels of CR2. This was also the case for mutations described in Figures 5, 7, 8(c) and 9.

Mutational Analysis of the CR2–C3d Interaction

Figure 4. Schematic representation of the PILN11GR13IS sequence of the first inter-cysteine region of SCR1. Mapped epitope positions on SCR1 using inhibitory monoclonal antibodies indicate that residues P8-S15 are important for the association of CR2 with iC3b.28,31 Residues N11 and R13 were selected from within this sequence for mutagenesis screening. Also shown are residues Y16 and S32, which are in close spatial proximity to residues 8–15 and which were also selected as candidates for substitution analysis.

849 CR2 with iC3b.28,31 Introducing alanine and glutamic acid residue mutations at the R13 site (R13A, R13E) resulted in significant reductions in C3dg binding relative to wild-type (Figure 5(a)). Modifications to residues S32 and Y16 (Y16A, S32A, S32E), which are spatially close to the PILN11GR13IS sequence, also demonstrated decreases in C3dg binding; S32E in particular exhibited almost no measurable binding association (Figure 5(b) and (c)). Interestingly, the mutation of N11 to an alanine (N11A) resulted in little or no decrease in C3dg binding affinity, while the introduction of the more disruptive glutamic acid residue (N11E) at this site still resulted in appreciable C3dg binding (Figure 5(d)). Subsequent review of the CR2–C3d co-crystal complex reveals that R13, Y16 and S32 are located on the outward face of SCR1 facing the C3d structure; N11 is located behind these residues, and is oriented away from the plane they form. As such, mutation of N11 to either alanine or glutamic acid is unlikely to significantly interfere with a possible SCR1–C3d interaction. On the other hand, R13, previously implicated in playing a role in iC3b binding, appears to be essential for the CR2–C3d interaction.

Figure 5. C3dg-biotin tetramer binding analysis for mutations selected around and within the PILN11GR13IS sequence of SCR1. Shown are the normalized binding affinities relative to wild-type CR2 of (a) R13A, R13E; (b) Y16A; (c) S32A, S32E; and (d) N11A, N11E mutants. Data are shown for serial dilutions of a C3dg-biotin PE streptavidin stock solution. The data reported here are for K562 cells exhibiting medium expression of CR2.

850 Mutations targeted at the disruption of the positively charged SCR1 face Human CR2 SCR1-2 contains a preponderance of positively charged residues, with 15 arginine and lysine residues, and only eight aspartic acid and glutamic acid residues. Most of this positive charge appears to be concentrated on a single external face of the CR2 molecule (Figure 6). To investigate a possible role for this region within SCR1 in the CR2–C3d interaction, mutations were selected at sites R28, R36, K41, K50 and K57 on SCR1, and at K67, located in the linker region between SCR1 and SCR2. Our data demonstrate how essential this area is to the CR2–C3d interaction (Figure 7(a)–(f)). Alanine screening experiments revealed that R28A, K41A, K50A and K57A mutants exhibited significant decreases in C3dg binding activity, whereas R36A and K67A mutants showed binding associations that were close to, or identical with, that seen for wild-type CR2. Single site glutamic acid substitutions, R28E, R36E, K41E, K50E, K57E and K67E all exhibited a substantial decrease in binding avidity relative to wild-type CR2. However, of these mutations K50E and K67E showed the higher apparent affinities for C3dg. Overall these data, in conjunction with the data provided above for R13, provide unequivocal experimental evidence that SCR1 plays an important role in the binding interaction between CR2 SCR1-2 and C3d, and also significantly expands and modifies the area on SCR1, previously delineated by epitope and peptide mapping that is required for this interaction.

Mutational Analysis of the CR2–C3d Interaction

The dimer interface The existence of a physiological SCR1–SCR1 homodimer similar to that seen in the CR2–C3d co-crystal structure is one of the more intriguing aspects of the CR2–C3d association. In the crystal structure D52 plays an integral part in the symmetrical dimer interface, making a number of hydrogen bonds to residues on the corresponding SCR1 domain (Figure 8(a)). In murine CR2, the equivalent residue to D52 is a histidine residue (H52). However, a number of mouse strains also contain a cytosine to adenosine (1342 c/a) single nucleotide polymorphism at H52, thus introducing an asparagine residue, and concurrently, an N-linked glycan.33 The introduction of a carbohydrate moiety at this site may be associated with the development of autoimmune disease in murine models, and it has been proposed that the structural rationale behind these disease states may be a result of the N-glycan sterically inhibiting either the formation of the SCR1–SCR1 dimer, or directly interfering with the CR2–C3d interaction.33 To test whether or not the presence of an N-linked glycosylation site on human CR2 interferes with C3d binding we mutated D52 to asparagine (D52N) thus creating an N-linked glycosylation sequence (Asn-X-Ser/Thr). Subsequent SDS-PAGE analysis of wild-type CR2 and the D52N form of this protein indicated that the D52N mutant migrated at a higher apparent molecular mass than wild-type CR2, signifying that the asparagine residue had been glycosylated (Figure 8(b)). However, no appreciable decrease in C3dg binding was seen for the D52N mutant relative to wild-type CR2 (Figure 8(c)). These data indicate that the introduction of an N-glycan moiety at the D52 position does not have a significant effect on the CR2–C3d interaction.

Discussion

Figure 6. Electrostatic surface representation of CR2 SCR1-2. A number of positively charged residues (R28, R36, K41, K50, K57, K67) located in SCR1 and in the linker region between the two SCR domains were chosen for mutagenesis and subsequent C3dg-biotin tetramer binding analysis. R83 (integral to the SCR2–C3d interaction) and R13 are also indicated.

The construction of a large number of instructive CR2 mutants expressed on the surface of K562 erythroleukemia cells, and our analysis of their ability to bind C3 proteolytic fragments, in the form of recombinant C3dg-biotin tetramers, has allowed us to experimentally address a number of questions that have arisen subsequent to the elucidation of the CR2 SCR1-2–C3d co-crystal structure. Mapping the relative C3dg binding affinities of our alanine substitutions (Figure 9(a)) and glutamic acid and tyrosine substitutions (Figure 9(b)) onto the surface of CR2 has allowed us to identify two specific areas on SCR1 and SCR2 that are essential for the CR2– C3d association; on SCR2, the area including and around R83 and G84; and on SCR1 a number of arginine and lysine residues (R13, R28, K41, K57) which form a contiguous region of positive charge and which is conserved in both the human and the murine forms of CR2. We have also attempted to ascertain the nature of a CR2 SCR1–SCR1 dimer

Mutational Analysis of the CR2–C3d Interaction

851

Figure 7. Mutagenesis screening of residues located within the positively charged face of CR2. Shown are the normalized binding affinities relative to wild-type CR2 of (a) R28A, R28E; (b) R36A, R36E; (c) K41A, K41E; (d) K50A, K50E; (e) K57A, K57E; and (f) K67A, K67E mutants. Data are shown for serial dilutions of a C3dg-biotin PE streptavidin stock solution. The data reported here are for K562 cells exhibiting medium expression of CR2.

structure by introducing an N-glycan moiety at residue 52. Our data directly targeting the SCR2–C3d interaction, as identified in the CR2–C3d co-crystal structure, indicate that introducing single site

alanine and glutamic acid substitutions at R83 and a tyrosine substitution at G84 inhibit C3dg-biotin tetramer binding to a significant extent. In conjunction with these data, previous peptide and mAb mapping studies utilizing human–mouse chimeras

852

Figure 8. (a) Detailed view of the SCR1–SCR1 interaction: specifically the integral role D52 plays in the creation of a network of hydrogen bonds securing the CR2–CR2 homodimer. Mutation of D52 to asparagine creates an N-glycosylation consensus sequence. Presence of an N-glycan moiety would be expected to impair the formation of a physiologic dimer. The two CR2 molecules are represented in cyan and red, respectively. (b) Immunoprecipitates of the wild-type and D52N forms of CR2 from surface biotinylated K562 erythroleukemia cells. The D52N mutant migrates at a higher apparent molecular mass than wild-type CR2 due to the presence of an additional N-glycan at position 52. (c) C3dg-biotin binding analysis of the D52N mutation. Data are given for cells exhibiting medium CR2 expression.

have implicated the first inter-cysteine region of SCR2 as playing an important role in iC3b binding.28 Both of these studies are in full agreement with those areas of CR2 SCR2 that make up the interface region with C3d in the co-crystal complex. It should also be considered that the surfaces of

Mutational Analysis of the CR2–C3d Interaction

SCR2 and C3d appear to be complementary in shape. Accordingly, it is almost assured that the SCR2–C3d interface, as reported in the CR2–C3d structure, is an accurate reflection of the binding association in solution. The crystal structures of both bound and free forms of CR2 SCR1-2 indicate that the two SCR domains are tightly packed against each other in an association seemingly dominated by hydrophobic interactions involving I38, W111, P120, but also through a salt-bridge between E64 and H91.17 As such, SCR1 appears to be unable to interact directly with C3d due to constraints on domain mobility provided by the side-side packing. However, our alanine and glutamic acid substitution data have allowed us to identify a number of charged residues on an external edge of SCR1 that are essential for the CR2–C3d interaction to occur. These data not only enable us to confirm that both SCR1 and SCR2 are necessary for CR2 to function as a C3d/C3dg receptor, but also allow us to significantly expand the region of SCR1 that is critical for the association of CR2 and C3d; previous epitope and peptide mapping experiments have determined that residues P 8–S 15 are critical for the C3d–CR2 association.28,31 However, our data not only allow us to isolate R13 from the PILN11GR13IS sequence of the first inter-cysteine region of SCR1 as playing a prominent role in C3d binding, but also to identify an additional essential participation centered around residues R28, K41 and K57. These data unequivocally demonstrate that the CR2–C3d association is facilitated by electrostatic interactions, although the exact nature of this charge dependence is currently undetermined. A recent study, utilizing theoretical electrostatic potential and apparent pKa calculations, has proposed that the pH and ionic strength dependencies exhibited by the CR2–C3d interaction are driven by the overall respective net charges of the CR2 and C3d molecules.32 Specifically, the formation of a CR2–C3d complex is influenced by long-range pseudospecific interactions leading to the establishment of an encounter complex in which the interface region between the two molecules is dominated by hydrogen bonds, hydrophobic interactions and van der Waals contacts. The existence of a long-range electrostatic interaction between CR2 and C3d, rather than a specific requirement for ion pair formation, allows a number of the disparities that are seen between the solution phase studies of the CR2–C3d interaction and the CR2–C3d cocrystal structure to be reconciled; in particular, an alanine screening study utilizing mutant forms of iC3b that indicated that single, double and triple mutations directed towards an acidic pocket of residues on the concave surface of C3d, remote from the SCR2–C3d interaction site, severely inhibited binding to CR2-bearing Raji cells.23 According to the theoretical binding model, mutations disrupting the negative electrostatic potential associated with this acidic pocket would also affect the overall negative character of the C3d molecule and

Mutational Analysis of the CR2–C3d Interaction

853

Figure 9. (a) Alanine substitutions and (b) glutamic acid and tyrosine substitutions mapped onto the surface of free CR2. The scheme used to color mutations represents the percentage of C3dg-biotin tetramer binding relative to wildtype CR2 (at a concentration of 2 mg of C3dg-biotin/0.4 mg of PE streptavidin). All values given are for K562 cells exhibiting medium CR2 expression. This Figure was generated using the PYMOL Molecular Graphics System (Delano Scientific, San Carlos, CA, USA).

therefore its ability to form a complex with CR2. The proposal of a long-range electrostatic interaction playing a critical role in the orientation of the CR2 and the C3d molecules relative to each other also allows the physiologic existence of a tightly packed V-shaped SCR1–SCR2 conformation similar to that seen for the crystallized forms of CR2. According to this scenario SCR1 would not have to directly interact with C3d, but mutations affecting the overall positive character of the CR2 molecule would disrupt the attraction between the two molecules nonetheless. An alternative explanation for the pH and ionic strength dependencies exhibited by the CR2–C3d binding association may be provided by the fact that some of our alanine and glutamic acid substitutions directed towards conserved positively charged residues on CR2 do not appear to have a dramatic effect on C3dg binding. Residues R36 and K67 appear to be integral to the overall positive charge distribution on the surface of the CR2 molecule (Figure 6), and as such the mutation of these residues can be expected to alter the overall positive character of the CR2 molecule. However, R36A, K67A and K67E exhibit an increased capacity for C3dg binding relative to the R13A, R13E, R28A, R28E, R36E, K41A, K41E, K57A and K57E mutations on SCR1 (Table 1). These data indicate that a specific region of positive charge on CR2 is necessary for the association of this receptor with C3d, and may therefore signify a much more localized electrostatic interaction than that proposed by the theoretical long-range attraction binding model. Namely, CR2–C3d binding would be facilitated or stabilized by the formation of saltbridges between charged residues on CR2 and on C3d. However, for SCR1 to directly interact with the surface of C3d in a charge-dependent manner, two major criteria have to be met: firstly, the identification of a more flexible CR2 structure in solution, in which the two SCR domains are not constrained by side-side packing, and in which the long eight amino acid residue linker region is in a much more

extended conformation than that seen for the crystal structures of bound and free CR2; secondly, assuming the SCR2–C3d interface as reported in the CR2–C3d co-crystal structure is a correct reflection of the binding interaction in solution, and therefore SCR2 is anchored to the top of the C3d molecule, then we must to be able to identify a corresponding negatively charged region on C3d which would be able to act as a SCR1 binding site. Concurrent with this mutagenesis screening study, a separate X-ray scattering and sedimentation modeling analysis of the CR2 SCR1-2–C3d interaction has independently been carried out and these results are presented in the accompanying paper.34 Briefly, however, these data indicate that CR2 in both the C3d-bound and free states adopts an open V-shaped conformation in which the two SCR domains make little or no contact with each other, and in which the extended linker region appears to facilitate domain flexibility.20,34 In addition, these solution studies of the CR2–C3d complex and when combined with a computer modeling approach, indicate that SCR1 is likely to lie close to the surface of the C3d molecule, allowing for the likelihood that both SCR1 and SCR2 directly interact with C3d. Subsequent examination of the CR2–C3d co-crystal complex has allowed us to identify a negatively charged channel on the C3d molecule that may act as a direct interaction site for SCR1 (Figure 10). This channel is directly contiguous to the SCR2 binding site described in the co-crystal structure, and a number of aspartic acid and glutamic acid residues are located within this groove region: E117, D122, D128, and D147. Further computer modeling shows that a solution structure for the complex in which SCR1 is close to or located within this groove would be consistent with the X-ray scattering data (H.E. Gilbert and S.J. Perkins, personal communication). Spatially, this channel is remote from the acidic pocket located within the concave surface of C3d, and this alternative shortrange electrostatic model is unable to explain previous mutagenesis screening studies targeting

Table 1. Summary of CR2–C3dg-biotin binding affinity for high, medium and low CR2 expressing K562 cell populations High

Medium

Low

Mutation

MFI

SEM

Weighting

Mutation

MFI

SEM

Weighting

Mutation

MFI

SEM

Weighting

N11A N11E R13A R13E Y16A R28Aa R28E S32A S32E R36A R36E K41A K41E K50A K50E D52N K57A K57E K67A K67E R83A R83E G84Y

97.1 69.2 62.5 32.1 76.6 77.4 12.1 80.5 7.5 84.5 22.5 40.2 17.4 54.5 49.6 95.0 26.3 26.4 117.1 58.6 65.8 30.0 13.2

6.0 7.6 5.9 5.4 5.3 11.4 2.7 8.7 4.0 8.3 1.7 10.1 3.7 8.8 8.5 10.2 3.8 4.4 11.2 2.0 4.5 10.5 1.3

CCCC CC CC C CCC CCC K CCC K CCC C CC K CC CC CCCC C C CCCC CC CC C K

N11A N11E R13A R13E Y16A R28A R28E S32A S32E R36A R36E K41A K41E K50A K50E D52N K57A K57E K67A K67E R83A R83E G84Y

96.0 71.9 65.5 34.3 76.3 53.9 12.6 78.1 9.2 79.6 22.5 35.7 16.7 49.2 51.6 91.2 23.8 22.4 94.8 54.5 69.6 31.2 13.6

8.4 11.7 4.8 4.8 4.0 6.3 2.6 8.2 4.2 8.2 2.3 4.8 2.8 6.8 9.1 11.5 3.0 3.9 9.0 1.6 3.8 10.5 1.3

CCCC CCC CC C C CC K CCC K CCC C C K CC CC CCCC C C CCCC CC CC C K

N11A N11E R13A R13E Y16A R28A R28E S32A S32E R36A R36E K41A K41E K50A K50E D52N K57A K57E K67A K67E R83A R83E G84Y

94.7 69.2 62.5 35.2 75.4 53.2 13.1 79.2 9.1 72.8 20.5 36.4 16.6 44.8 52.5 99.8 24.3 21.3 85.8 50.5 72.5 31.9 14.2

4.2 7.3 5.9 4.6 4.2 5.0 2.9 5.2 4.1 6.8 1.7 3.5 2.6 3.3 9.4 12.0 3.4 3.8 7.2 1.6 5.0 10.5 1.5

CCCC CC CC C CCC CC K CCC K CCC C C K CC CC CCCC C C CCC CC CCC C K

Percentage binding affinities are reported here for cells incubated in the presence of 2 mg C3dg-biotin/0.4 mg PE streptavidin. Key to binding data: CCCC, 90–120%; CCC, 70–89.9%; CC, 40–69.9%; C, 20–39.9%; K, 0–19.9%. a The high expressing R28A K562 cell population exhibited an inconsistent C3dg tetramer binding affinity with the medium and low expressing populations.

855

Mutational Analysis of the CR2–C3d Interaction

Figure 10. Electrostatic representation of the CR2–C3d complex. Indicated on C3d are residues E117, D122, D128 and D147, which form part of a conserved negatively charged groove located on C3d, immediately proximal to the SCR2 binding site, and may comprise an additional CR2–C3d interaction site.

this region on the iC3b fragment.23 It should be noted, however, that some studies have indicated that iC3b may bind to CR2 at different sites to C3d and C3dg.24,35 Consistent with the idea of a localized charge-dependent association occurring between SCR1 and C3d are surface plasmon resonance data we have acquired which indicate that all phases of the CR2 SCR1-2–C3dg binding association are effected by ionic strength. In 50 mM NaCl the KD of the CR2 SCR1-2–C3d binding interaction was calculated at 67.2 nM, with a Kon of 2.39!10K3 nMK1 sK1 and a Koff of 0.189 sK1. In 125 mM NaCl the KD was calculated at 658 nM, with a Kon of 0.663!10K3 nMK1 sK1 and a Koff of 0.43 sK1. Protein complexes that are dominated by localized charge–charge interactions exhibit binding associations in which both the Kon and the Koff rates are sensitive to ionic strength and pH. However, when long-range electrostatic interactions are dominant in complex formation, only the Kon rate is sensitive to ionic strength and pH.36,37 As such, our surface plasmon resonance data are indicative of a localized charge-dependent interaction occurring between SCR1 and C3d, although the existence of a long-range electrostatic interaction between CR2 and C3d cannot be ruled out. To resolve the exact nature of the CR2–C3d electrostatic interaction it is therefore going to be

necessary to perform additional solution phase studies on the CR2–C3d complex. This will likely involve the use of high field multi-dimensional NMR techniques that allow the determination of primary structure, as well as the identification of those amino acids that contact ligand, by measuring chemical shifts, coupling constants and linewidths. One of the unanticipated interactions revealed by the elucidation of the CR2–C3d co-crystal structure was the presence of an SCR1–SCR1 homodimer. Substantial initial support for the physiologic relevance of such a dimer was provided with the identification in murine CR2 of a single nucleotide polymorphism occurring at residue 52, thus changing the normally found histidine residue to asparagine. This polymorphic CR2 variant has been found to bind C3d with a lower affinity than wildtype CR2 and is predicted to have a disrupted dimer interface due to perturbation caused by the introduction of an N-linked glycosylation site.33 This allele of the murine Cr2 gene appears to be an excellent candidate for the disease associated gene within the Sle1c locus of the NZM2410 model of systemic lupus erythematosus. However, our D52N mutant constructed with the intention of disrupting the formation of a possible SCR1–SCR1 dimer on full-length human CR2, by the introduction of an N-linked glycosylation sequence, failed to show a substantial decrease in C3dg binding ability. This indicates that even under the constraining conditions in which a transmembrane protein is anchored to a cell surface, it is unlikely that dimer formation has a major physiologic effect on the binding of C3d/C3dg. Solution phase studies of the free and C3d-bound forms of CR2 have also determined that the CR2 SCR1-2 molecule is monomeric in physiological buffers,19,20 and it is therefore likely that the presence of a CR2–CR2 dimer may only be a feature of the static ligandbound structure observed in the CR2–C3d co-crystal structure.

Conclusions This mutagenesis screening study of the CR2– C3d interaction targeting specific areas on the SCR1 and SCR2 domains of human CR2 has provided a number of revealing insights into the nature of this binding association. Mutations (R83A, R83E, G84Y) constructed with the aim of disrupting an SCR2– C3d interface, as represented in the CR2–C3d co-crystal structure, exhibited reduced ability to bind C3dg. From these data we have been able to conclude that the unanticipated SCR2–C3d binding site seen in the CR2–C3d co-crystal structure is likely to be a true reflection of this binding interaction under solution conditions. Previous epitope and peptide mapping strategies have identified the P8–S15 sequence on the first inter-cysteine region of SCR1 as playing a salient role in CR2–C3d binding.27,28,31 Our mutational screening of residues likely to be involved in a

856 SCR1–C3d interaction has indicated that a number of conserved, positively charged residues on a single face of SCR1 are critical for CR2–C3d binding to occur (R13, R28, K41, K57). The fact that only R13 is located within this PILN11GR13IS sequence has required us to redefine those regions of SCR1 that are necessary for C3dg binding. Previous surface plasmon resonance strategies screening the CR2– C3d binding association have indicated that this interaction is electrostatic in nature.19,24 As such, these studies are wholly supportive of our data implicating the participation of positively charged residues on SCR1 in a binding interaction with C3d. The precise nature of this charge–charge interaction is currently unresolved, and both long-range,32 and localized (this work) electrostatic contributions to the CR2–C3d binding interaction have been proposed. It is hoped that our mutagenesis data defining the critical roles that residues R13, R28, K41, K57 and R83 play in the CR2–C3d interaction will assist in the future design of therapeutic peptides or molecules that could alter the CR2– C3d binding interaction. Finally, we have utilized the CR2–C3d co-crystal structure to construct a single site mutation on human CR2 that was designed to disrupt a physiologic SCR1–SCR1 homodimer, thereby mimicking a single nucleotide polymorphism found in murine CR2. The creation of a D52N mutant introduced an N-glycosylation consensus sequence at an area on the SCR1 domain integral to the dimer interface. However, no decrease in C3dg binding ability could be seen for this mutant. This result is supportive of concurrent solution studies of the CR2–C3d interaction that indicate that human CR2 is monomeric both in free and C3d-bound forms, 34 but does not preclude the possibility that this mutation in murine CR2 plays a role in the binding of C3d or other ligands.

Materials and Methods Production of CR2 and C3dg-biotin recombinant proteins Wild-type full-length rCR2 was expressed on human K562 erythroleukemia cells with the eukaryotic expression vector pSFFV-neo, as described.26 CR2-expressing transfectants were incubated in the presence of biotinylated anti-CR2 mAb HB-5 and then sorted using streptavidin-coated magnetic beads (Dynabeads, Dynal, Great Neck, NY) to establish stable populations of cells with CR2 protein expression. Mutations of CR2 within the first two SCR domains were carried out utilizing a Quickchange site-directed mutagenesis kit (Stratagene, La Jolla, CA). Mutated full-length CR2 cDNA was sequenced by the University of Colorado Cancer Center Sequencing Core (Denver, CO), and transfected into K562 human erythroleukemia cells for binding analysis. Human C3dg-biotin was produced using established protocols.19,30 Briefly, C3dg-biotin was produced in Escherichia coli BL21 pLysS Codon Plus strain (Stratagene) transformed with the pET11b vector (Novagen Inc.).

Mutational Analysis of the CR2–C3d Interaction

Chloramphenicol and ampicillin-resistant colonies were used to produce starter cultures, which were then expanded to five liters and grown at 37 8C in the presence of 50 mM D-biotin until an A600 nm of w0.3 was attained. Samples were then induced with 0.025 M IPTG at 28 8C and shaken overnight. The cultures were harvested and pellets were resuspended in PBS (pH 8.0) in the presence of complete EDTA-free protease inhibitor tablets (RocheBoehringer Mannheim) and lysed by successive freezethaw cycles. The lysate was then clarified and purified by successive immobilized metal affinity chromatography (IMAC) and size-exclusion chromatography stages. Confirmation that samples were biotinylated was obtained by Western blots utilizing streptavidin-horseradish peroxidase.

Flow cytometry Flow cytometric experiments were carried out using K562 erythroleukemia cells transfected with full-length wild-type or mutant human CR2. For each condition, 5!105 human CR2-transfected K562 cells were first incubated with FITC-conjugated anti-CR2 mAb HB-5 at 1 mg/ml on ice for one hour.26,29 During this incubation, 100 ml of tetramers in PBS, 0.1% (w/v) BSA, 0.01% (w/v) sodium azide were prepared for each condition by adding the appropriate amount of recombinant C3dg-biotin and 0.4 mg of PE conjugated streptavidin (Southern Biotech, Birmingham, AL) and incubating at room temperature for 30 minutes. C3dg-biotin concentrations used were 2, 1, 0.5, 0.25, 0.125, and 0.0625 mg/ml. Serial dilutions up to 1:16 of a stock solution initially containing 2 mg of C3dgbiotin/0.4 mg of PE streptavidin were also prepared. Following washing of the FITC-stained K562 cells, 100 ml of tetramers was added to each sample of cells and incubated for 30 minutes on ice. After washing, the cells were fixed and analyzed by multi-color flow cytometry in the University of Colorado Cancer Center Flow Cytometry Core Facility (Denver, CO). Cells were divided into high, medium and low CR2 expression using 101 as the lower limit and gating on the lower 18%, middle 18% and upper 18% of CR2-expressing cells (FITC positive). Tetramer binding was determined by PE mean channel fluorescence. A minimum of three separate experiments was carried out for each mutation. Pilot experiments revealed no difference in tetramer binding in the presence or absence of mAb HB5.

Immunoprecipitation and SDS-PAGE analysis A total of 1!106 K562 human erythroleukemia cells expressing either wild-type full-length CR2 or the D52N mutant were surface biotinylated with EZ-Link SulfoNHS-LC-Biotin (Pierce). Cells were then lysed with RIPA buffer supplemented with Complete protease inhibitor cocktail tablets (Roche Boehringer Mannheim), 10 mg/ml of pepstatin, and phosphatase inhibitors (0.05 mM sodium orthovanadate, 50 mM sodium fluoride, and 20 mM b-glycerophosphate). Lysates were incubated with mAb HB-5 for 30 minutes on ice, followed by the addition of protein G-Sepharose (Amersham Pharmacia) and a subsequent one hour incubation at 4 8C. Samples were run in 7.5% (w/v) polyacrylamide gels, transferred to a nitrocellulose membrane, incubated with streptavidin-horseradish peroxidase (Zymed) and analyzed by Western blot.

Mutational Analysis of the CR2–C3d Interaction

Acknowledgements These studies were supported by the National Institutes of Health (grant R0-1 CA53615) and assisted by the UCHSC Cancer Center Flow Cytometry Core. We also thank Professor Stephen Perkins and Hannah Gilbert at University College London for extremely helpful discussions.

References 1. Pepys, M. B. (1974). Role of complement in induction of antibody-production in vivo–effect of cobra factor and other C3 reactive agents on thymus-dependent and thymus-independent antibody-responses. J. Expt. Med. 140, 126–145. 2. Ahearn, J. M. & Fearon, D. T. (1989). Structure and function of the Complement Receptors, CR-1 (CD35) and CR-2 (CD21). Advan. Immunol. 46, 183–219. 3. Dempsey, P. W., Allison, M., Akkaraju, S., Goodnow, C. & Fearon, D. T. (1996). C3d of complement as a molecular adjuvant: bridging innate and acquired immunity. Science, 271, 348–350. 4. Carroll, M. C. & Fischer, M. B. (1997). Complement and the immune response. Curr. Opin. Immunol. 9, 64–69. 5. Fearon, D. T. (1998). The complement system and adaptive immunity. Semin. Immunol. 10, 355–361. 6. Ahearn, J. M., Fischer, M. B., Croix, D. A., Georg, S., Ma, M., Xia, J. et al. (1996). Disruption of the Cr2 locus results in a reduction in B-1a cells and in an impaired B cell response to T-dependent antigen. Immunity, 4, 251–262. 7. Croix, D. A., Ahearn, J. M., Rosengard, A. M., Han, S., Kelsoe, G., Ma, M. & Carroll, M. C. (1996). Antibody response to a T-dependent antigen requires B cell expression of complement receptors. J. Expt. Med. 183, 1857–1864. 8. Molina, H., Holers, V. M., Li, B., Fang, Y.-F., Mariathasan, S., Goellner, F. et al. (1996). Markedly impaired humoral immune response in mice deficient in complement receptors 1 and 2. Proc. Natl Acad. Sci. USA, 93, 3357–3361. 9. Weis, J. J., Tedder, T. F. & Fearon, D. T. (1984). Identification of a 145,000 Mr membrane-protein as the C3d receptor of (CR-2) of human lymphocytes-B. Proc. Natl Acad. Sci. USA, 81, 881–885. 10. Kalli, K. R., Ahearn, J. M. & Fearon, D. T. (1991). Interactions of iC3b with recombinant isotypic and chimeric forms of CR2. J. Immunol. 147, 590–594. 11. Aubry, J.-P., Pochon, S., Graber, P., Jansen, K. U. & Bonnefoy, J.-Y. (1992). CD21 is a ligand for CD23 and regulates IgE production. Nature, 358, 505–507. 12. Aubry, J.-P., Pochon, S., Gauchat, J. F., Nueda-Marin, A., Holers, V. M., Graber, P. et al. (1994). CD23 interacts with a new functional extracytoplasmic domain involving N-linked oligosaccharides. J. Immunol. 152, 5806–5813. 13. Delcayre, A. X., Salas, F., Mathur, S., Kovats, K., Lotz, M. & Lernhardt, W. (1991). Epstein-Barr Virus complement C3d receptor is an interferon-alpha receptor. EMBO J. 10, 919–926. 14. Fingeroth, J. D., Weis, J. J., Tedder, T. F., Strominger, J. L., Biro, P. A. & Fearon, D. T. (1984). Epstein-Barr Virus receptor of human lymphocytes-B is the C3d receptor CR-2. Proc. Natl Acad. Sci. USA, 81, 4510–4514.

857 15. Nemerow, G. R., Wolfert, R., McNaughton, M. E. & Cooper, N. R. (1985). Identification and characterization of the Epstein-Barr Virus receptor on human lymphocytes-B and its relationship to the C3d complement receptor (CR-2). J. Virol. 55, 347–351. 16. Szakonyi, G., Guthridge, J. M., Li, D., Young, K., Holers, V. M. & Chen, X. S. (2001). Structure of complement receptor 2 in complex with its C3d ligand. Science, 292, 1725–1728. 17. Prota, A. E., Sage, D. R., Stehle, T. & Fingeroth, J. D. (2002). The crystal structure of human CD21: implications for Epstein-Barr virus and C3d binding. Proc. Natl Acad. Sci. USA, 99, 10641–10646. 18. Nagar, B., Jones, R. G., Diefenbach, R. J., Isenman, D. E. & Rini, J. M. (1998). X-ray crystal structure of C3d: a C3 fragment and ligand for complement receptor 2. Science, 280, 1277–1281. 19. Guthridge, J. M., Rakstang, J. K., Young, K. A., Hinshelwood, J., Aslam, M., Robertson, A. et al. (2001). Structural studies in solution of the recombinant N-terminal pair of short consensus/complement repeat domains of complement receptor type 2 (CR2/CD21) and interactions with its ligand C3dg. Biochemistry, 40, 5931–5941. 20. Gilbert, H. E., Hannan, J., Holers, V. M. & Perkins, S. J. (2004). Synchroton X-ray scattering and ultracentrifugation of unbound CR2 SCR1-2 shows that SCR-1 and SCR-2 form a family of extended V-shapes in solution. Mol. Immunol. 41, 235. 21. Moore, M. D., DiScipio, R. G., Cooper, R. N. & Nemerow, G. R. (1989). Hydrodynamic, electron microscopic and ligand-binding analysis of the Epstein-Barr virus/C3dg receptor (CR2). J. Biol. Chem. 264, 20576–20582. 22. Zanotti, G., Bassetto, A., Battistutta, R., Folli, C., Arcidiaco, P., Stoppini, M. & Berni, R. (2000). ˚ resolution of an N-terminally Structure at 1.44 A truncated form of the rat serum complement C3d fragment. Biochim. Biophys. Acta, 1478, 232–238. 23. Clemenza, L. & Isenman, D. (2000). Structure-guided identification of C3d residues essential for its binding to Complement Receptor 2 (CD21). J. Immunol. 165, 3839–3848. 24. Sarrias, M. R., Franchini, S., Canziani, G., Argyropoulos, E., Moore, W. T., Sahu, A. & Lambris, J. D. (2001). Kinetic analysis of the interactions of complement receptor 2 (CR2, CD21) with its ligand C3d, iC3b, and the EBV glycoprotein gp350/220. J. Immunol. 167, 1490–1499. 25. Lowell, C. A., Klickstein, L. B., Carter, R. H., Mitchell, J. A., Fearon, D. T. & Ahearn, J. M. (1989). Mapping of the Epstein-Barr virus and C3dg binding sites to a common domain on complement receptor 2. J. Expt. Med. 170, 1931–1946. 26. Carel, J.-C., Myones, B. L., Frazier, B. & Holers, V. M. (1990). Structural requirements for C3dg/EpsteinBarr virus receptor (CR2/CD21) ligand binding, internalization, and viral infection. J. Biol. Chem. 265, 12293–12299. 27. Molina, H., Brenner, C., Jacobi, S., Gorka, J., Carel, J.-C., Kinsohita, T. & Holers, V. M. (1991). Analysis of Epstein-Barr virus-binding sites on complement receptor 2 (CR2/CD21) using human-mouse chimeras and peptides: at least two distinct sites are necessary for ligand-receptor interaction. J. Biol. Chem. 266, 12173–12179. 28. Molina, H., Perkins, S. J., Guthridge, J., Gorka, J.,

858

29.

30.

31.

32. 33.

34.

Mutational Analysis of the CR2–C3d Interaction

Kinoshita, T. & Holers, V. M. (1995). Characterization of a Complement Receptor 2 (CR2/CD21) ligand binding site for C3. J. Immunol. 154, 5426–5435. Tedder, T. F., Clement, L. T. & Cooper, M. D. (1984). Expression of C3d receptors during human B cell differentiation: immunofluorescence analysis with the HB-5 monoclonal antibody. J. Immunol. 133, 678–683. Henson, S. E., Smith, D., Boackle, S. A., Holers, V. M. & Karp, D. R. (2001). Generation of recombinant C3dg tetramers for the analysis of CD21 binding and function. J. Immunol. Methods, 258, 97–109. Martin, D. R., Yuryev, A., Kalli, K. R., Fearon, D. T. & Ahearn, J. M. (1991). Determination of the structural basis for selective binding of Epstein-Barr virus to human complement receptor type 2. J. Expt. Med. 174, 1299–1311. Morikis, D. & Lambris, J. D. (2004). The electrostatic nature of C3d-complement receptor 2 association. J. Immunol. 172, 7537–7547. Boackle, S. A., Holers, V. M., Chen, X., Szakonyi, G., Karp, D. R., Wakeland, E. K. & Morel, L. (2001). Cr2, a candidate gene in the murine Sle1c lupus susceptibility locus, encodes a dysfunctional protein. Immunity, 15, 775–785. Gilbert, H. E., Eaton, J. T., Hannan, J. P., Holers, V. M.

35.

36. 37.

38. 39.

& Perkins, S. J. (2005). Solution structure of the complex between CR2 SCR1-2 and C3d human complement: an X-ray scattering and sedimentation modeling study. J. Mol. Biol. 346, 859–873. Esparza, I., Becherer, J. D., Alsenz, J., De la Hera, A., Lao, Z., Tsoukas, D. C. & Lambris, J. D. (1991). Evidence for multiple sites of interaction in C3 for complement receptor 2 (C3d/EBV receptor, CD21). Eur. J. Immunol. 21, 2829–2838. Selzer, T., Albeck, S. & Schreiber, G. (2000). Rational design of faster associating and tighter binding protein complexes. Nature Struct. Biol. 7, 537–541. Kiel, C., Selzer, T., Shaul, Y., Schreiber, G. & Herrmann, C. (2004). Electrostatically optimized Ras-binding Ral guanine dissociation stimulator mutants increase the rate of association by stabilizing the encounter complex. Proc. Natl Acad. Sci. USA, 101, 9223–9228. Fingeroth, J. D., Clabby, M. L. & Strominger, J. D. (1988). Characterization of a lymphocyte-T EpsteinBarr virus C3d receptor (CD21). J. Virol. 62, 1442–1447. Koradi, R., Billeter, M. & Wu¨thrich, K. (1996). MOLMOL: a program for display and analysis of macromolecular structures. J. Mol. Graph. 14, 51–55.

Edited by R. Huber (Received 18 August 2004; received in revised form 2 December 2004; accepted 3 December 2004)