A Cluster of Positively Charged Amino Acids in the C4BP α-Chain Is Crucial for C4b Binding and Factor I Cofactor Function*

C4b-binding protein (C4BP) is a regulator of the classical complement pathway, acting as a cofactor to factor I in the degradation of C4b. Computer modeling and structural analysis predicted a cluster of positively charged amino acids at the interface between complement control protein modules 1 and 2 of the C4BP α-chain to be involved in C4b binding. Three C4BP mutants, R39Q, R64Q/R66Q, and R39Q/R64Q/R66Q, were expressed and assayed for their ability to bind C4b and to function as factor I cofactors. The apparent affinities of R39Q, R64Q/R66Q, and R39Q/R64Q/R66Q for immobilized C4b were 15-, 50-, and 140-fold lower, respectively, than that of recombinant wild type C4BP. The C4b binding site demonstrated herein was also found to be a specific heparin binding site. In C4b degradation, the mutants demonstrated decreased ability to serve as factor I cofactors. In particular, the R39Q/R64Q/R66Q mutant was inefficient as cofactor for cleavage of the Arg937-Thr938 peptide bond in C4b. In contrast, the factor I mediated cleavage of Arg1317-Asn1318 bond was less affected by the C4BP mutations. In conclusion, we identify a cluster of amino acids that is part of a C4b binding site involved in the regulation of the complement system.

C4b-binding protein (C4BP) is a regulator of the classical complement pathway, acting as a cofactor to factor I in the degradation of C4b. Computer modeling and structural analysis predicted a cluster of positively charged amino acids at the interface between complement control protein modules 1 and 2 of the C4BP ␣-chain to be involved in C4b binding. Three C4BP mutants, R39Q, R64Q/R66Q, and R39Q/R64Q/R66Q, were expressed and assayed for their ability to bind C4b and to function as factor I cofactors. The apparent affinities of R39Q, R64Q/R66Q, and R39Q/R64Q/R66Q for immobilized C4b were 15-, 50-, and 140-fold lower, respectively, than that of recombinant wild type C4BP. The C4b binding site demonstrated herein was also found to be a specific heparin binding site. In C4b degradation, the mutants demonstrated decreased ability to serve as factor I cofactors. In particular, the R39Q/R64Q/R66Q mutant was inefficient as cofactor for cleavage of the Arg 937 -Thr 938 peptide bond in C4b. In contrast, the factor I mediated cleavage of Arg 1317 -Asn 1318 bond was less affected by the C4BP mutations. In conclusion, we identify a cluster of amino acids that is part of a C4b binding site involved in the regulation of the complement system. The human complement system comprises about 35 known plasma-and membrane-bound proteins involved in efficient activation and tight regulation of the system. Complement proteins form multi-molecular complexes, and limited proteolysis is central for both activation and regulation. Factor I is a serine protease responsible for down-regulation of both classical and alternative pathways of complement. Factor I degrades activated forms of complement factors C3 (C3b) and C4 (C4b) when they are bound to factor H or C4b-binding protein (C4BP), 1 respectively (1, 2).
C4BP is a large plasma protein (molecular mass, ϳ570 kDa) consisting of seven identical ␣-chains and a unique ␤-chain linked together by disufide bridges (3,4). The ␣and ␤-chains contain eight and three complement control protein (CCP) mod-ules, respectively (5). CCP modules consist of approximately 60 amino acids forming a compact hydrophobic core surrounded by five or more ␤-strands organized into ␤-sheets (6). Electron microscopy of C4BP demonstrated a spider-like conformation, with the seven ␣-chains forming extended tentacles (7,8). Synchrotron x-ray scattering and hydrodynamic analysis suggested that C4BP, in solution, is a bundle of seven extended arms held together at their C termini with an average arm-axis angle of 10°rather than being a spider-like molecule (9). C4BP interacts not only with C4b (3) but also with protein S (single binding site on ␤-chain) (10), serum amyloid P component (11), bacterial surface proteins from Streptococcus pyogenes (12), and heparin (13).
C4BP controls C4b-mediated reactions, thereby regulating the classical pathway of complement. Apart from acting as a cofactor to factor I, C4BP also accelerates the natural decay of the C4bC2a complex, which is the classical pathway C3 convertase (14). C4, which is the precursor of C4b, is composed of a 93-kDa ␣-chain, a ␤-chain (75 kDa), and a 32-kDa ␥-chain, which are linked by disulfide bridges (15)(16)(17). During complement activation, C1s cleaves a 9-kDa fragment (C4a) from the N terminus of the ␣-chain (18,19). The remaining part of the molecule (C4b) acquires transient capability to bind amino and hydroxyl groups via a reactive glutamyl residue that is exposed upon cleavage of an internal thiolester bond (20). C4b-like molecules can also be generated from C4 by treatment with amines, chaotropes, or repeated freezing and thawing, which result in the cleavage of the internal thiolester bond without liberation of C4a (21).
Each ␣-chain of C4BP contains a C4b binding site, but, most likely due to sterical hindrance, only up to four C4b molecules can bind to one C4BP molecule (22). Several different regions of the ␣-chains have been suggested to be involved in C4b binding. Initially, an N-terminal 48-kDa ␣-chain fragment, formed by chymotrypsin digestion, was found to bind C4b and to express factor I cofactor activity (23,24). This agreed well with electron microscopy results, which demonstrated C4b binding to the peripheral end of each C4BP tentacle (7). Several reports agree with the concept that the three most N-terminal CCPs are necessary and sufficient for C4b binding (25)(26)(27). The C4b binding site in human C4BP partially overlaps with the binding site for the S. pyogenes surface proteins (12,26), and recently, it was suggested that Arg 66 and/or His 67 is involved in these interactions (26).
The elucidation of the regulatory mechanisms of complement is hindered by the lack of structural information. To overcome these limitations, we have used a combination of molecular modeling and site-directed mutagenesis to study the interaction between C4b and C4BP. In a recent study, we identified a cluster of positively charged amino acids present at the surface of the CCP1-2 modules of the C4BP ␣-chain as a potential binding site for C4b and heparin (28). We now report results obtained with a series of C4BP mutants (R39Q, R64Q/R66Q, and R39Q/R64Q/R66Q), which show this cluster to be crucial for binding of both C4b and heparin.

EXPERIMENTAL PROCEDURES
Materials-The sensor chip CM5 and amine coupling kit were from Biacore AB. Molecular weight markers for electrophoresis were from Amersham Pharmacia Biotech. Lipofectin, Opti-MEM, and Dulbecco's modified Eagle's medium were from Life Technologies, Inc. Chondroitin sulfate and streptavidin were from Sigma, unfractionated heparin was from Lövens, and low molecular weight heparin (Fragmin) and hyaluronan (Healon) were from Amersham Pharmacia Biotech.
Protein Modeling and Geometry Optimization-The reported C4BP model (28) was based on the structure of a CCP pair from factor H. Since then, NMR coordinates of another CCP pair (CCP3-4 of the vaccinia virus complement protein VCP) (29) have become available. However, based on additional modeling work and analysis of our experimental data, we conclude that CCP15 and CCP16 of factor H (30) were the best starting templates to build CCP1 and CCP2 of C4BP, respectively. Our initial model was refined via molecular dynamics simulation using five different simulation protocols (e.g. different set of partial charges, different dielectric or tethering constants). The same procedure was also applied to the VCP and factor H NMR pairs. We performed these calculations in order to select the best set of parameters. The structural refinement of the initial model included energy minimization in vacuo using Discover (Biosym-MSI). All calculations were carried out using the CVFF force field parameters and a 20-Å cut-off distance for nonbonded interactions. Hydrogen atoms were added to the model, and partial charges were assigned to all atoms. In an effort to account for the lack of solvent and ions, potentially charged residues were given appropriate parameters to obtain electrostatic neutrality (31). A total of 10 consensus residues belonging to the central core regions of CCP1 and CCP2 were initially tethered during the minimization and molecular dynamics simulation procedures (the force constant, K, added on all heavy atoms was 5 kcal⅐Å Ϫ2 ). This constant was subsequently relaxed (K ϭ 0) in the final energy minimization. A 100-ps molecular dynamics simulation at 300 K was performed after a 10-ps equilibration. The coordinates saved in the 100-ps history file were collected at 1-ps intervals, and each 100 structures were energy minimized using steepest descent algorithm and conjugate gradients minimizer. Energy minimization was stopped when the maximum Cartesian derivatives of the potential energy function were less than 1.0 kcal/mol-Å (31). The average structure from 55 to 100 ps was generated and energy minimized. A Silicon Graphics Indigo2 R10000 workstation was used.
Proteins-C4BP (8), C4 (32), and factor I (33) were purified from human plasma as described previously. The concentrations were determined by measurement of absorbance at 280 nm, and extinction coefficients (1%, 1 cm) used were 14.1, 8.3, and 14.3 for C4BP, C4 and factor I, respectively. C4b-like molecules (C4met) were prepared by incubation of purified C4 with 100 mM methylamine for 1 h at 37°C and subsequent dialysis against 100 mM Tris-HCl, pH 7.5, and 150 mM NaCl (TBS). Throughout the study, the C4met derivative was used but will in the text be referred to as C4b for reasons of clarity. Proteins were labeled with 125 I using the chloramine T method. The specific activity was 0.4 -0.5 MBq/g of protein.
cDNA Clones for Recombinant Proteins-Full-length cDNA coding for human C4BP ␣-chain (34) was cloned to an eucaryotic expression vector; pcDNA3 (Invitrogen). It was then used as a template, and the following mutations were introduced using the Quick Change sitedirected mutagenesis kit (Stratagene): Arg 39 3 Gln (primers: 5Ј-CCT GGC TAC GTC CAA TCC CAT TCA ACT-3Ј and the corresponding antisense primer), Arg 64 /Arg 66 3 Gln-Gln (primers: 5Ј-TGT ATC TAC AAA CAA TGC CAA CAC CCA GGA GAG-3Ј and the corresponding antisense primer), Arg 39 /Arg 64 /Arg 66 3 Gln-Gln-Gln (above primers were combined). Nucleotides corresponding to changed amino acid residues are underlined. All mutations were confirmed by an automated DNA sequencing (Perkin-Elmer).
Purification of Recombinant Proteins-Human kidney 293 cells (ATCC catalog no. 1573-CRL) were transfected with the various C4BP constructs using Lipofectin, according to the manufacturer's instructions (Life Technologies, Inc.). The neomycin analogue G418, at a concentration of 400 g/ml, was used for selection of transfected cells. Colonies of cells showing the highest expression levels, as judged by immunoblotting, were expanded in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 3.4 mM glutamine, 100 units/ml of penicillin and streptomycin, and 400 g/ml G418. Medium from transfected cells was collected in 3-day periods during which the cells were cultured in Opti-MEM Glutamax. It was then stored at Ϫ20°C until about 4 liters of medium was collected. The medium was centrifuged for 30 min at 5000 ϫ g to remove cell debris and applied at a flow of 90 ml/h on an affinity column with mAb 104 (35) coupled to Affi-Gel 10 (2.6 ϫ 12 cm; Bio-Rad) previously equilibrated with TBS. The column was washed with TBS, 1 M NaCl, and the recombinant protein was subsequently eluted with 3 M guanidinium chloride and dialyzed extensively against TBS. All preparative work was done at 4°C. The total amount of each protein obtained was about 8 mg, and the exact concentrations were determined from the amino acid composition analysis after 24 h hydrolysis in 6 M HCl.
Electrophoretic and Blotting Techniques-Unreduced and reduced samples containing purified proteins were subjected to 5 and 10% SDS-PAGE, respectively. The proteins were then detected either by Coomasie Brilliant Blue R-250 staining or transferred to a polyvinylidene difluoride membrane (Millipore). Membranes were quenched with 50 mM Tris-HCl, pH 7.5, containing 150 mM NaCl, 3% fish gelatin, and 0.1% Tween 20. They were then incubated with various antibodies diluted in above buffer. Membranes were washed with the same buffer without fish gelatin and incubated with anti-immunoglobulins conjugated with alkaline phosphatase, washed again, and developed.
C4b Ligand Binding Assay-Microtiter plates (Maxisorp, Nunc) were incubated overnight at 4°C with 50 l of solution containing 10 g/ml C4b in 75 mM sodium carbonate, pH 9.6. The wells were washed three times with 50 mM Tris-HCl, 0.15 M NaCl, 0.1% Tween, pH 7.5 (washing buffer), and then incubated at room temperature with 200 l of quench solution (washing buffer supplemented with 3% fish gelatin). After another three washes, increasing concentrations of plasma purified C4BP or recombinant proteins were added in TBS supplemented with 0.1% Tween 20, 0.1% bovine serum albumin, and the plates were incubated for 3 h at room temperature. The wells were then washed three times and incubated with biotinylated mAb 67 (27) diluted in quench solution. After 1 h of incubation, the plates were washed and incubated for 1 h with streptavidin-conjugated horseradish peroxidase, washed, and developed according to the manufacturer's instructions (Dakopatts).
Competition Assay-Microtiter plates were incubated overnight at 4°C with 50 l of solution containing 10 g/ml C4b in 75 mM sodium carbonate, pH 9.6. The wells were washed three times with washing buffer and then incubated at room temperature with 200 l of quench solution. After another three washes, the 125 I-labeled plasma purified C4BP was added (20 kcpm/well) together with various unlabeled proteins diluted in 25 mM Tris-HCl, 50 mM NaCl, 0.1% Tween 20, 0.1% bovine serum albumin, pH 7.5. When the influence of glycosaminoglycans was tested, samples contained 0.1 nM C4BP and increasing concentrations of heparin, chondroitin sulfate, or hyaluronan (molecular mass, ϳ40 kDa). The samples were incubated for 3 h at room temperature and washed five times, and the amount of radioactivity associated with each well was measured in a ␥-counter.
C4b Degradation Assay-The assay was performed as described previously (36).
Heparin Affinity-About 200 g of C4BP or one of the three mutant proteins were applied at a flow of 0.5 ml/min on 2 ml heparin-Sepharose (Amersham Pharmacia Biotech) equilibrated in 20 mM Tris-HCl, pH 7.4. The column was washed with 5 volumes of the starting buffer, and the proteins were eluted with a linear gradient of NaCl from 0 to 0.5 M NaCl at a flow of 1 ml/min. The amount of C4BP in collected fractions (0.33 ml) was estimated with an enzyme-linked immunosorbent assay. NaCl concentration in the samples was calculated from the measurement of conductivity. The interaction between heparin and C4BP was also studied using a BIAcore biosensor system (Biacore AB). Two flow cells of CM5 sensor chip were activated with 20 l of a mixture of 0.2 M 1-ethyl-3-(3 dimethylaminopropyl) carbodiimide and 0.05 M N-hydroxysulfosuccinimide at a flow rate of 5 l/min, after which 45 l of streptavidin (0.1 mg/ml in 10 mM acetate buffer, pH 4.5) was injected. Unreacted groups were blocked with 20 l of 1 M ethanolamine, pH 8.5. Approximately 4000 resonance units of streptavidin were fixed on the surface of each chip. Biotinylated heparin (kind gift of Dr. Markku Salmivirta; Department of Medical Biochemistry and Microbiology, Uppsala University) was injected over one of the two streptavidin surfaces, the other being a negative control. About 600 resonance units of heparin was immobilized. The association kinetics were studied for recombinant C4BP and the three mutants at various concentrations (0 -500 nM). The flow buffer was 10 mM Hepes-KOH, pH 7.4, supplemented with 75 mM NaCl, 0.005% Tween 20, and 3.4 mM EDTA. Aliquots of C4BP stock solutions (0.6 -2 M in flow buffer) were diluted in the flow buffer, and 45 l was injected during the association phase at constant flow rate of 20 l/min. The dissociation was followed for 6 min at the same flow rate. In all experiments, 10 l of 2 M NaCl was used to remove bound ligands.
Synthetic Peptide-An 18-mer peptide, EILQEEDLIDEDDIPVRS, corresponding to residues 740 -757 of human C4, was purified by high pressure liquid chromatography after synthesis and lyophilized. It was dissolved in phosphate-buffered saline at a 100 M concentration. To test the effect of the peptide on the C4b-C4BP interaction, 20 nM C4BP in 20 mM Tris-HCl, pH 7.5, 0.1% Tween 20, 0.1% bovine serum albumin was mixed with increasing amounts of the peptide (0.1 nM to 50 M). The samples were then added to microtiter plates with immobilized C4b and the binding of C4BP estimated as described for the C4b ligand binding assay. Alternatively, the microtiter plates were incubated overnight at 4°C with 50 l of solution containing 20 g/ml of the peptide in 75 mM sodium carbonate, pH 9.6, and the binding of 125 I-C4BP was assayed as described for the competition assay.

Refined Molecular Model of CCP1-2 and Strategy for Sitedirected Mutagenesis-
The three-dimensional structures of two pairs of CCP modules have been determined with NMR so far: CCP15-16 of factor H and CCP3-4 of VCP. The CCP pairs of these two structures present different intermodular angles (29,30). Our initial model for the C4BP ␣-chain (28) was based on the structure of the factor H pair. The NMR structure of factor H CCP15 was found to be the best starting conformation to build C4BP CCP1 because the number of insertions and deletions between the sequence of the model to build and the CCP15 template was small as compared with other CCPs of known structure. Also, when using CCP15 of factor H as a template, the amino acid side chains in the resulting C4BP CCP1 model run according to rules deduced from analysis of experimentally determined structures. For instance, the positively charged side chain of Arg 39 is oriented toward the solvent, mimicking Arg 41 of factor H (numbering according to the Protein Data Bank file, entry 1hfh). In contrast, when VCP CCP2 was used as a template, Arg 39 of C4BP pointed directly into the hydrophobic core of the module. Even though some other rotamers for Arg 39 could be used to avoid this problem (37), the guanidinium group of Arg 39 would be 15 Å away from R64 and on a different face of the molecule. However, Arg 39 and Arg 64 are most likely topological neighbors (see below) because heparin binds to this region of C4BP. Similar observations were made for the second CCP of the C4BP ␣-chain, suggesting that factor H CCP16 is an appropriate template to build this C4BP module. Moreover, when the VCP intermodular angle was used to build the C4BP CCP1-2 model, Lys 63 of C4BP was "locked" into an aromatic/hydrophobic pocket formed at the interface between the modules. Such problems did not occur when the factor H pair was used as the initial template. The angle between C4BP CCP1 and CCP2 based on the one present in factor H was consistent with appropriate amino acid distribution of C4BP (e.g. charged residues were exposed) and with the presence of a positively charged cluster on one face of the molecule. However, because the ␣-chain of C4BP has an extended shape as shown by electron microscopy and x-ray scattering data, we suggest that the intermodule angle of factor H has to be stretched to represent more closely a native C4BP tentacle in a solution.
The geometry of the initial CCP1-2 model was refined via molecular dynamics simulation and the resulting model is shown in a simplified representation in Fig. 1A. The stereo-FIG. 1. Model for the CCP1 and CCP2 modules of the C4BP ␣-chain. A, solid ribbon representation of the final model for the CCP1 (yellow) and CCP2 (white) modules of C4BP. The side chains of cysteine residues are shown (blue), and the linker region between the two CCPs is marked in red (residues 61-64). The side chains of a key positive cluster for C4b and heparin binding are shown. In the present study, Arg 39 , Arg 64 , and Arg 66 were mutated to glutamine residues. The modeling data also suggest that Lys 63 is part of this electropositive cluster. B, Ramachandran plot for the final C4BP model. This plot shows that most of the backbone dihedral angles are in the energetically favored regions. A few residues were found outside generously allowed regions. These amino acids were mainly glycines that can be located in almost all regions of the plot. This is illustrated by arrows pointing at Gly 36 : Phi ϭ 160, Psi ϭ Ϫ66.8; Gly 52 : Phi ϭ Ϫ143, Psi ϭ 37.7; Gly 112 : Phi ϭ 168.8, Psi ϭ Ϫ76.6; Gly 26 : Phi ϭ 91.9, Psi ϭ 6.1.
chemistry of the final model was analyzed using ProStat (Biosym-MSI). We investigated bond lengths, backbone ⍀-angles, side chain 1 and 2 angles, and chirality. All were found to be in agreement with values obtained for experimentally determined structures. A Ramachandran plot is presented in Fig. 1B and shows that mostangles are within the energetically favored regions supporting the quality of the model. The few peptide bonds of the model that are found in unfavored regions were found to involve mostly glycines (indicated by arrows in Fig. 1B), which is a well known feature of this amino acid. Moreover, Ramachandran plots using NMR structures of other CCP modules yielded a similar percentage of peptide bonds outside the allowed regions, again mainly due to Gly residues (not shown).
The three-dimensional model of CCP1-2 was used as basis for selection of mutations. In order to probe the functional importance of interface between CCP1 and CCP2, three mutants were generated, R39Q, R64Q/R66Q, and R39Q/R64Q/ R66Q. The C4BP mutants were expressed in eucaryotic cells and purified from cell culture media using mAb 104 affinity chromatography (Fig. 2). The epitope for mAb 104 is located in CCP1 or at the interface between CCP1 and CCP2 (27). The expression levels of wild type and mutant C4BP were similar and approximately 2 mg of pure protein was obtained from 1 liter of culture medium. Without reduction (Fig. 2, B and D), recombinant wild type and mutant C4BP demonstrated molecular weights slightly lower than that of plasma derived C4BP, which was due to the lack of the ␤-chain in recombinant C4BP (35). On immunoblotting, the reduced proteins were visualized with a rabbit antibody against human C4BP (Fig. 2C), whereas unreduced mutants were recognized by mAb 67 (Fig. 2D), the epitope of which is located in the middle part of the ␣-chain (27). The surface exposed mutations that were introduced in C4BP did not cause folding problems affecting expression levels, reactivities with various antibodies, and electrophoretic mobilities. The three arginine residues that were the targets for the mutagenesis were solvent exposed and not involved in any clear stabilizing ionic interactions (i.e. salt bridges) and the amino acid substitutions were therefore expected to be well tolerated. The replacement of positively charged Arg by polar Gln is conservative in term of overall size and ability to form hydrogen bonds with the solvent. Collectively, these data strongly suggest that the proteins were well folded.
C4BP Mutagenesis Resulting in Decreased Binding of C4b to C4BP-The interaction between C4b and recombinant C4BP mutants was analyzed with a direct binding assay as well as in a competition assay. Each C4BP molecule contains up to seven C4b binding sites, each of which binds C4b with relatively low affinity (22). The fast dissociation of C4b from C4BP prevented the use of microtiter plate based assays with immobilized C4BP. As an alternative approach to investigate the effects of the mutations, binding of C4BP mutants to immobilized C4b was tested. Due to the multimeric nature of C4BP and the multiple C4b binding sites on each C4BP, this approach only allowed a qualitative analysis of the effect of the mutations on the C4b-C4BP interaction but did not permit calculation of affinity constants. We chose to estimate the apparent affinity from the midpoint of the binding curve in the direct binding assay using purified C4BP and immobilized C4b. The binding was estimated both at physiological NaCl concentration (150 mM) and at 50 mM NaCl in order to elucidate the influence of ionic strength on the interaction. Plasma purified and recombinant wild type C4BP bound to immobilized C4b with similar apparent affinities, as shown in Fig. 3. At physiological NaCl concentrations (Fig. 3B), R39Q and recombinant wild type C4BP reached 50% of maximal binding at 37 and 3 nM (12.3fold difference), respectively. R64Q/R66Q and R39Q/R64Q/ R66Q reached 50% binding at 168 and 400 nM, which is 56 and 133 times higher, respectively, than the 3 nM found for recombinant wild type C4BP (Fig. 3B). At the lower ionic strength, 0.5 nM of recombinant wild type C4BP was required to give 50% maximum binding. These results suggest the C4b-C4BP interaction to be highly dependent on ionic interactions. This was further demonstrated by the complete inhibition of the interaction at 500 mM NaCl when the recombinant wild type C4BP, at a concentration of 5 nM, was incubated with immobilized C4b. Already at 200 mM salt, the interaction was decreased by 50% as compared with physiological conditions (data not shown). At low ionic strength, R39Q, R64Q/R66Q, and R39Q/ R64Q/R66Q reached 50% binding at concentrations that were 19 (9.5 nM), 46 (23 nM), and 200 (100 nM) times higher, respectively, than that of recombinant wild type C4BP (see Fig. 3A).
In the competition assay, the recombinant proteins were allowed to compete with 125 I-labeled C4BP tracer for binding to immobilized C4b (Fig. 3C). In the absence of competitor, 18 -24% of the added 125 I-labeled C4BP tracer bound to the immobilized C4b, and the binding of radiolabeled tracer could be competed out by unlabeled C4BP. Recombinant wild type and plasma C4BP were equally efficient in displacing the 125 I-C4BP tracer from the immobilized C4b, the half maximal competition being reached at 1 nM concentration of C4BP (experiments done at 50 mM NaCl). The concentrations of R39Q, R64Q/R66Q, and R39Q/R64Q/R66Q required to obtain 50% inhibition were 15-, 67-, and 94-fold higher, respectively, than that of recombinant wild type C4BP.
A sequence in C4 rich in negatively charged amino acids residues, 740 EILQEEDLIDEDDIPVRS 757 , has been proposed FIG. 2. Analysis of purified C4BP and its mutants by SDS-PAGE. Plasma purified C4BP and recombinant proteins (approximately 2 g/well for Coomasie staining and 0.04 g/well for immunoblotting) were separated by SDS-PAGE. The polyacrylamide gel concentrations were 5 and 10% for unreduced and reduced samples, respectively. Proteins were then subjected to Coomasie staining (A and B) or transferred to a membrane (C and D) and allowed to react with rabbit antibody against C4BP (reduced samples, C) or mAb 67 (unreduced samples, D).
to be a binding site for C4BP (38). As the binding site for C4b in C4BP is rich in positively charged amino acid residues, we decided to test whether a synthetic peptide corresponding to this C4b sequence was able to inhibit the C4b-C4BP interaction. However, no inhibition of the binding between C4b and C4BP was observed, even in the presence of 50 M peptide (data not shown). Furthermore, no binding of 125 I-labeled C4BP to immobilized peptide (in microtiter plates) was observed (data not shown). Even though the results obtained with the peptide were negative, they do not exclude the possibility that the proposed region in C4b is part of the C4BP binding site because the peptide in solution may not have adopted the proper conformation.
Mutagenesis of C4BP and Loss of Factor I Cofactor Function-To elucidate whether the impaired C4b binding was matched by a decrease in factor I cofactor activity, the C4BP variants were incubated with C4b, factor I and trace amounts of 125 I-labeled C4b. Proteins were then separated by SDS-PAGE and C4b visualized by autoradiography (Fig. 4A). In the presence of both C4BP and factor I, C4b was degraded, and C4d and ␣4 molecules appeared as described (21). The ␣3 ϩ C4a fragment had the same mobility as the ␤-chain, which is why the two polypeptides cannot be distinguished. Recombinant wild type C4BP was equally efficient as factor I cofactor as plasma purified C4BP (Fig. 4, lanes 3, 4, 9, and 10). The three C4BP mutants as compared with recombinant wild type C4BP demonstrated decreased cofactor activity. To quantify the loss in cofactor activity, the intensities of bands corresponding to C4d and ␣4 were estimated by densitometry (Molecular Dynamics). As judged from the amount of C4d released, the R39Q/ R64Q/R66Q mutant functioned poorly as factor I cofactor (Table I). In contrast, generation of ␣4 was less affected by the introduced mutations, and the level of ␣4 generated in the presence of the R39Q/R64Q/R66Q mutant was only 30% lower than that formed in presence of recombinant wild type C4BP. The time course of the two cleavages in the presence of recombinant wild type C4BP or the R39Q/R64Q/R66Q mutant is shown in Fig. 5. During its inactivation, C4b is cleaved at two positions by factor I, at Arg 937 -Thr 938 and Arg 1317 -Asn 1318 (Fig.  4A). In the presence of recombinant wild type C4BP (Fig. 5,  open circles), the Arg 1317 -Asn 1318 cleavage resulting in the ap- gives rise to a C4d fragment (45 kDa) that is released from the remaining part of the molecule (C4c). C4c consists of two ␣-chain fragments, ␣4 (13 kDa) and a polypeptide composed of ␣3 (25 kDa) and C4a (13 kDa), linked to the ␤-chain with disulfide bridges. The arrow marks the position of the negatively charged 740 -757 peptide (see text). B, C4BP (200 nM) was incubated with 250 nM C4b, 60 nM factor I and trace amounts of 125 I-labeled C4b for 1.5 h at 37°C. Immediately afterward, a sample buffer (with or without a reducing agent, dithiothreitol (DTT)) was added, samples were heated at 95°C, and the proteins were separated by SDS-PAGE electrophoresis (7.5-15% gradient gel). The gel was dried and subjected to autoradiography. As controls, either C4BP (lane 1) or factor I (lanes 2 and 8) was omitted in the incubation mixtures.
pearance of the ␣4 fragment is very rapid. The subsequent cleavage at position Arg 937 -Thr 938 results in the release of the C4d fragment. At the concentrations of reagents used in the degradation assay, the first and the second cleavage were completed within 30 min and 2 h, respectively. When the R39Q/ R64Q/R66Q mutant was analyzed in a similar manner, we found that Arg 1317 -Asn 1318 cleavage was less affected than the Arg 937 -Thr 938 cleavage. As a result, after 105 min of incubation, 70% of the ␣4 was released, but only 15% of the C4d fragment was formed (Fig. 5, closed circles). The same amounts of ␣4 and C4d were released even after 6 h of incubation (results not shown).

Mutagenesis of C4BP and the Loss of Heparin Binding Site-
C4BP is known to bind to heparin (13,39), but the binding site on C4BP has not been localized. Inhibition of the interaction between C4BP and C4b by heparin (40) suggests that the C4b and heparin binding sites overlap. Unfractionated heparin, which contains molecules varying in size from 4 to 30 kDa (mostly 13 kDa), and low molecular weight heparin, prepared by deaminative cleavage, consisting of molecules in a range of 2 to 6 kDa (mostly 4 kDa), were found to be equally efficient in inhibiting the C4b-C4BP interaction (Fig. 6). Under the same conditions, chondroitin sulfate or hyaluronan had no effect.
Affinity chromatography was used to assess the capacity of the mutant C4BP molecules to bind to heparin. Recombinant wild type C4BP eluted as a broad peak with maximum at 273 Ϯ 7 mM NaCl (Fig. 7). Similar results were obtained with plasma purified C4BP eluting at 266 Ϯ 9 mM NaCl (not shown). The heparin binding abilities of C4BP were compromised by the mutations R39Q and R64Q/R66Q, eluting at 216 Ϯ 5 and 181 Ϯ 4 mM, respectively. The R39Q/R64Q/R66Q mutant eluted from the heparin-Sepharose already at 129 Ϯ 3 mM NaCl, indicating that it would not bind to heparin under physiological ionic strength.
To further demonstrate the decrease in heparin affinity as a result of the introduced mutations, surface plasmon resonance technique (BIAcore) was used to monitor the formation and dissociation of surface-bound complexes between the recombinant C4BP variants and immobilized heparin (Fig. 8). Each C4BP protein sample was injected over the surface carrying immobilized streptavidin-biotinylated heparin complexes and also over a chip containing streptavidin alone as control. The sensograms presented were obtained after subtraction of the unspecific binding of C4BP to the streptavidin containing chip. Wild type recombinant C4BP demonstrated specific, dose-dependent binding to the immobilized heparin. The rate of association was high and the dissociation rate was low, suggesting high affinity interaction. However, as the interaction between the fluid phase multimeric C4BP and the immobilized heterogeneous heparin is very complex, it was not possible to calculate meaningful affinity constants for the interaction. The three mutant C4BP molecules yielded weak binding signals, suggesting that most of the specific heparin binding site was lost as a result of the mutagenesis (Fig. 8B).

DISCUSSION
The factor I cofactor function of C4BP resulting in the degradation of C4b is a physiologically important regulatory mechanism of the complement system. Due to the complexity and heterogeneity of the interacting proteins, C4BP, C4b, and factor I, three-dimensional structures of the individual proteins and of protein complexes have not been determined experimen-tally and are unlikely to become available in the near future. However, molecular modeling and computer-guided site-directed mutagenesis are efficient tools for the identification of binding sites and elucidation of structure-function relationships. These techniques have proven successful in other studies, e.g. in the investigation of two CCP modules from CD46 (41).
Based on the sequence similarities between CCP modules in C4BP and various CCPs with known three-dimensional structures, we constructed a three-dimensional model for the ␣-chain of human C4BP (28). Theoretical and interactive analysis of potential binding sites suggested the interface between C4BP CCP1 and CCP2, displaying a cluster of positively charged amino acids, to be of putative functional importance. Sequences of five cDNAs coding for C4BP ␣-chains have been reported: human (42), bovine (43), rat (44), mouse (45), and rabbit (46). Analysis of interspecies similarities of amino acid sequences showed the N-terminal region of the ␣-chain containing the positively charged cluster to be highly conserved suggesting physiological importance (46). Taken together, these observations prompted us to construct several ␣-chain mutants in which arginines in CCP1 and CCP2 were replaced by glutamine residues. Three mutants, R39Q, R64Q/R66Q, and R39Q/R64Q/R66Q, were expressed in the eucaryotic system and purified. From the results of direct binding studies and a competition assay, it is clear that the suggested electropositive cluster indeed represents a key interaction site for C4b. Our data agree with a previous report focusing on the S. pyogenes-C4BP interaction by Accardo et al. (26), which suggested the binding site for C4b to span CCP1-3 and pointed out the importance of the Arg 66 -His 67 pair. A mutant carrying the combined change of Arg 66 to Glu and His 67 to Thr demonstrated decreased binding to C4b-Sepharose at physiological ionic strength. However, at 40 mM NaCl, no difference in binding (estimated as retention on a C4b-Speharose) between mutant and recombinant wild type C4BP could be observed (26).
The C4b-C4BP interaction is highly sensitive to salt concentration, and already at 200 mM NaCl, the binding decreased by about 50%. This stands in sharp contrast to the C4BP-protein S interaction, which is essentially unaffected by high salt concentrations (47). Protein S binds to the most N-terminal CCP of the ␤-chain (48). Structural analysis of a ␤-chain model revealed a cluster of solvent exposed hydrophobic residues, which are potentially involved in the binding of protein S (47). Such clusters of residues with hydrophobic characteristics favor tight and stable interactions and may therefore have been maintained during evolution.
Even though most reports suggest the C4b binding site to be located within CCPs 1-3, other regions of the ␣-chain have been proposed to be involved as well. Thus, proteolytic fragments of C4BP were used to map the C4b binding and factor I cofactor activities to distinct regions spanning CCP6 -7 and CCP3-6, respectively (49). In addition, a monoclonal antibody directed to CCP6 was found to block C4b binding (50), which is difficult to reconcile with the observation that mouse C4BP lacking CCP5-6 binds human C4b (45). Recently, it was shown that a cryptic binding site for C3b is located near the C terminus of the ␣-chain (51). This site was only exposed in recombinant, monomeric cell-bound C4BP, which makes its physiological significance unclear. However, such a cryptic C3b binding site may also have the ability to interact with C4b, which could explain some of the disagreement in published reports. When all available data are taken into consideration, it appears clear that the interface between CCP1 and CCP2 forms a crucial binding site for C4b. However, other regions within CCP1-3 are presumably required for expression of the full binding FIG. 8. Surface plasmon resonance analysis of C4BP-heparin interaction. A, increasing concentrations of recombinant wild type C4BP (0 -500 nM) were injected over chips containing biotinylated heparin, which was bound to immobilized streptavidin. The amount of protein associating with the heparin was measured in resonance units. Identical samples were also injected over a control chip containing streptavidin but no heparin (unspecific binding), and the demonstrated sensograms were obtained by subtraction of the unspecific binding. B, sensograms for the three recombinant C4BP mutants obtained after injection of a solution containing 500 nM protein over heparinized surface and corrected for unspecific binding. capacity between C4BP and C4b and for expression of factor I cofactor activity.
The C4b-C4BP complex is not the first example of an interaction in which the interface area between two CCP modules plays a pivotal role in forming a binding site. Thus, the binding sites for measles virus on CD46 involves the interface between CCP1 and CCP2 together with parts of CCP2 (41). Moreover, insertion of two and four amino acids between CCP1 and CCP2 and between CCP3 and CCP4 of factor H, respectively, entirely abolished its factor I cofactor function (52). Recently, it has been shown that an antibody blocking the interaction between CR2 and C3dg is directed against an epitope on a recess formed between CCP1 and CCP2 (53).
The decreased binding of mutated C4BP to C4b was matched by loss of its factor I cofactor activity. However, the mutations introduced in C4BP had differential effects on the two cleavage sites in C4b, such as the cleavage of the Arg 937 -Thr 938 peptide bond was more severely affected as compared with the Arg 1317 -Asn 1318 bond. The mechanism by which C4BP and factor H function as factor I cofactors in the degradation of C3b and C4b is unknown. Recently, it was suggested that factor H causes conformational changes in C3b, facilitating its interaction with factor I (54). Thus, it was proposed that factor I binds both factor H and C3b and that the binding sites do not overlap. Possibly, a similar mechanism is involved in the C4BP/factor I mediated C4b degradation. This would be consistent with the presence of a secondary C4b binding site located outside the electropositive cluster in CCP1-2, which possibly is more involved in the cleavage of the Arg 1317 -Asn 1318 peptide bond than the Arg 937 -Thr 938 bond. This could explain the selective impairment of one of the cleavage reactions.
Heparin interacts with several complement proteins, but the affinity of the binding of heparin to C4BP is the highest among the complement proteins, most likely due to the multimeric nature of C4BP (39). We now show that the C4b binding site also interacts with the negatively charged heparin but not with chondroitin sulfate or hyaluronan that is negatively charged due to a presence of carboxyl groups. The pattern, composition, and spacing of basic amino acids in heparin binding peptides and proteins have been investigated in detail (55,56). From these studies, it is known that heparin (characterized by a high charge density) interacts tightly with peptides displaying a series of at least three to five positively charged residues and that the binding is stronger when the cluster contains arginine residues rather than lysines. Our results are fully consistent with these observations as we show that Arg 39 , Arg 64 , and Arg 66 play a key role in heparin binding. In addition, because of its location, we suggest that Lys 63 is also involved in the heparin binding site. Indeed, after numerous molecular modeling trials (i.e. using different intermodule angles), the most rational location for the Lys 63 side chain is in direct vicinity to the three guanidinium groups of Arg 39 , Arg 64 , and Arg 66 . The mutations were found to abrogate the binding of C4BP to heparin almost entirely, which suggested that the electropositive cluster in CCP1-2 constitutes the major heparin binding site in C4BP. Analysis of the three-dimensional model of the whole ␣-chain of C4BP showed that the electropositive cluster between CCP1 and CCP2 was the only one presenting the key characteristics of a heparin binding site. It is noteworthy that heparin binding sites have also been found at the surface of factor H CCP7, CCP13 and CCP20 (57)(58)(59). However, the mechanism of action of heparin in the case of factor H seems to be different than the one for C4BP, because heparin enhances binding of factor H to surface-bound C3b, thereby helping in the inactivation of C3b by factor I (60). Furthermore, low molecular weight heparin (5 kDa) was ineffective in the factor H-C3b system, whereas it still inhibits C4b-C4BP interaction. In a C4b degradation assay, similar concentrations of heparin were required to inhibit factor I cofactor activity as the ones needed for blocking C4BP binding to immobilized C4b (not shown). The multimeric nature of C4BP makes the heparin interaction potentially interesting from a physiological perspective. Binding of C4BP to the host cells via for example three heparin binding sites present on three different ␣-chains would favor degradation of C4b close to the cell surface as the other four ␣-chains would be free to interact with C4b and factor I. Thus, interaction of C4BP with heparin-like molecules, present at the surface of some cells, could protect them from the destructive and inflammatory consequences of complement activation. The localization of the heparin binding site at the tip of the C4BP tentacles is consistent with this hypothesis because if such a site was located closer to the central core region of C4BP, it would not be as easily accessible for binding.
In recent years, several complement components have gained recognition as potential therapeutics, e.g. complement components may have cardioprotective roles, have anti-inflammatory actions, or help to overcome hyper acute rejection during xenotransplantation (reviewed in Ref. 61). For instance, a surface-bound form of human C4BP consisting of only one ␣-chain (CCP1 to CCP8) has been constructed (51). This engineered C4BP molecule efficiently blocked complement-mediated endothelial cell lysis, suggesting that C4BP could be useful for clinical xenotransplantation. However, although complement regulators have therapeutic potential, our knowledge of the intermolecular interactions involving complement regulatory proteins is still rudimentary. The now identified binding site for C4b on the C4BP molecule is part of an effort not only to describe sites for protein-protein interactions but also to enhance our understanding of molecular mechanisms involved in regulation of the complement system by factor I and its cofactor proteins.