The Kaposi’s Sarcoma-associated Herpesvirus Complement Control Protein Mimics Human Molecular Mechanisms for Inhibition of the Complement System*

Surface Plasmon Resonance (BIAcore)— The interaction between C3b/C4b and mutants K64Q/K65Q/K88Q, K131Q/K133Q/H135Q, H158A/H171A/H213A, and F195A/F207A/L209A was studied using surface plasmon resonance (BIAcore 2000). All four flow cells of a CM5 sensor chip were activated with 20 (cid:4) l of 0.2 M 1-ethyl-3-(3-dimethyl- aminopropyl)carbodiimide with 0.05 M N -hydroxy-sulfosuccinimide at a flow rate of 5 (cid:4) l/min. Wild type KCP and two of the mutants (0.02 mg/ml in 10 m M sodium-acetate buffer, pH 4) were injected over flow cell 2, 3, or 4 to reach 10,000 units prior to blocking unreacted groups with 20 (cid:4) l of 1 M ethanolamine, pH 8.5. Flow cell 1 served as a negative control to which no protein was coupled prior to blocking. The associa- tion kinetics were studied over a concentration range of 0.006–0.5 mg/ml for C3b and 0.0015–0.5 mg/ml for C4b using the standard flow buffer (10 m M Hepes-KOH, pH 7.4, supplemented with 75 m M NaCl, 0.005% Tween 20, 2.5 m M CaCl 2 , and 20 (cid:4) M ZnCl 2 ). Protein solutions were injected for 300 s to achieve saturation during the association phase at a constant flow rate of 20 (cid:4) l/min. The sample was first injected over the negative control surface and then over immobilized KCP flow cells and analyzed for a dissociation phase of 200 s at the same flow rate. Signals were normalized by subtracting the nonspecific signal measured by control flow cell 1. Between each different concentration of C3b or C4b tested, the flow cell surfaces were regenerated with a 30- (cid:4) l injection of 2

Kaposi's sarcoma-associated herpesvirus (KSHV 1 /HHV-8) is the most recently discovered human herpesvirus and the likely etiologic agent of the diseases Kaposi's sarcoma (1), multicentric Castleman's disease (2), and primary effusion lymphoma (3). The genome of KSHV contains Ͼ80 open reading frames, of which many have been predicted to protect the virus against the immune system (4). One of those, open reading frame 4, is a secondary lytic gene (5) encoding KCP (KSHV complement control protein), which is expressed at the surface of KSHVinfected cells (6). We and others have shown that KCP is a potent complement inhibitor (6 -8) and, thus, likely to be important for viral pathogenesis by protecting the virus against eradication by the complement system.
The biological importance of viral evasion from the complement system is illustrated by the numerous examples of viral complement inhibitors encoded by many viruses (9 -12) (for a review see Ref. 13), many of which were found to directly contribute to viral pathogenesis in vivo in experimental models (14 -16). Kapadia et al. found that deletion of the KCP homologue from the murine ␥HV68 virus, a rhadinovirus closely related to KSHV, dramatically altered the establishment of acute infection and the ability to maintain persistent infection (16).
KCP is a type 1 membrane bound protein with four Nterminal complement control protein (CCP) domains protruding into the extracellular space. The full-length KCP contains 550 amino acids (accession number P88903 at TrEMBL data base), but KSHV-infected cells also express two shorter isoforms of KCP that arise because of alternative splicing of the mRNA, which retain both the CCP domains and the transmembrane-spanning region (6).
KCP inhibits the complement system at the C3 convertase level. This is a strategic site of regulation, because formation of the C3 convertase and the subsequent activation of C3 are key events in complement activity, eventually leading to release of anaphylatoxins, opsonization of pathogens, and initiation of the terminal lytic pathway. The classical C3 convertase, comprised of C4b and C2a, is formed following complement activation through either the classical or the lectin pathway. C4b localizes the convertase to the membrane surface via a covalent bond, whereas C2a forms the active protease site. The alternative C3 convertase, formed during alternative pathway activation, is comprised of C3b (homologue to C4b) and Bb (C2a homologue). KCP inhibits the C3 convertases through two separate mechanisms, i.e. KCP directly accelerates the decay of classical pathway C3 convertases and also acts as a cofactor for factor I (FI) cleavage of both C3b and C4b (7,8). Therefore, KCP not only accelerates the decay of the C3 convertases, it also inhibits their formation.
KCP is a structural and functional homologue to a group of proteins belonging to the regulators of complement activation (RCA) gene family, whose members inhibit the C3 convertases by decay acceleration and cofactor activity. Regulators accelerate the decay of C3 convertases through binding to C4b and dissociation of C2a or binding to C3b and dissociation of Bb, but the exact mechanism for this event has not been fully elucidated. Most of the RCA family can also act as cofactors for the soluble serine protease FI, enabling it to cleave either C3b or C4b both in convertase-associated or free soluble forms, thereby either inactivating the convertase or preventing its formation. However, cofactor specificity varies among the RCA proteins. Some RCA proteins have a greater avidity for the cleavage of C4b compared with C3b (or vice versa), and complement receptor 1 (CR1; CD35) is unique in its ability to induce a third cleavage of C3b to yield the C3c and C3dg fragments. The exact requirements for FI cofactor function have yet to be elucidated. Direct binding of the cofactor to C3b or C4b is required but is in itself insufficient to induce cleavage by FI; it has been proposed that in order for cofactor activity to occur, the binding RCA protein must be able to induce a conformational change in C3b or C4b necessary for FI binding and cleavage (17).
There are several proteins in the human RCA gene family and these contain between 4 and 35 CCP domains. The three prominent membrane-bound complement regulators are CR1 (30 CCPs), the decay accelerating cofactor (DAF; four CCPs), and the membrane cofactor protein (four CCPs), whereas the two most prominent soluble forms are the oligomeric C4bbinding protein (C4BP; eight CCPs in each of the seven ␣-chains and three CCPs in the ␤-chain) and factor H (20 CCPs). A CCP domain contains ϳ60 amino acids arranged into a relatively compact hydrophobic/aromatic core surrounded by short ␤-strands and internally stabilized by two disulfide bonds (18).
Domain deletion and site-directed mutagenesis studies have revealed that the functional sites of human RCA proteins are comprised of 2-4 CCP domains, and the positively charged patches located at the linker region between the CCP domains are commonly identified as essential for functional activity. The viral RCA homologues have not been studied extensively with respect to structural requirements for function. A recent domain deletion study of the vaccinia virus complement control protein (VCP) suggested that all four CCP domains are needed for C3b binding (19), whereas a neutralizing antibody study suggested that CCP1 is dispensable (20) for C3b/C4b binding and complement inactivation. To date, VCP is the viral RCA homologue that has been the most extensively investigated, including having its three-dimensional structure determined by x-ray crystallography (21).
The aim of the present study was to delineate structural motifs that define the molecular basis of KCP function. Therefore, we constructed a three-dimensional structure of the four CCP domains of KCP by homology modeling. Based on this predicted structure and taking into account the previous work on human RCA proteins, the sites in KCP potentially required for protein-protein interactions and function were mutated, and the functional consequence was determined relative to wild type KCP. The complement component binding, the cofactor activity for both C3b and C4b, and the classical decay acceleration were all independently assessed. Our findings indicate that the binding of C4b, the cofactor activity for C4b cleavage, and the decay acceleration required amino acids in CCP1-3. A patch of positively charged amino acids in CCP1 that stretches into CCP2 (Arg-20, Arg-33, Arg-35, Lys-64, Lys-65, and Lys-88) resembles functionally important sites identified in members of the RCA gene family. We also show that the interaction with C3b (both binding and degradation) required regions in CCP 4 of KCP that were not needed for interaction with C4b. Importantly, this study is the first comprehensive analysis of the structure-function relationship of a viral complement inhibitor at the amino acid level and reveals the spatial requirements for broad complement regulatory activities.

MATERIALS AND METHODS
Homology Modeling and Structure Analysis-Homology modeling was carried out on a Silicon Graphics workstation using the software package Insight II (Accelrys) and the modules Homology, DelPhi, and Discover. The three-dimensional structure of the four CCP domains of VCP has been determined by x-ray crystallography (21) (Protein Data Bank file 1G40). The amino acids of KCP were numbered exclusive of the signal peptide starting with Gln-1, which is the 20th amino acid after the start (Met). To adjust for an insertion in CCP2 (relative to the VCP sequence), coordinates for Ile-79 to Lys-88 were grafted from CCP15 of factor H (amino acids 18 -27; Protein Data Bank file 1HFI), and an insertion in CCP3 was adjusted for by modeling amino acids 175-188 on coordinates of amino acids 105-118 from the ␤-2-glycoprotein (Protein Data Bank file 1C1Z). Other insertions were modeled by using coordinates from a data base of high-resolution structures generated in our group. In some cases, side chains were modified manually to low energy rotamers that also fitted well in the three-dimensional structure (i.e. all rotamers for a given residue were tested while monitoring non-bonded energy values). The structure was then subjected to energy minimization (Discover module) that was initially restricted to deletion and insertion areas followed by energy minimization of the whole structure, first in an uncharged state and then with charged amino acids surrounded by an 8-Å shell of water. The electrostatic potential at the surface was calculated using the DelPhi module with a dielectric constant of 4 inside the molecule surface and 80 outside the protein. A solvent accessibility computation for each amino acid was computed with the method of Lee and Richards (22) to identify buried and exposed residues.
Three-dimensional Comparison of Functional Sites-Homology models of KCP CCPs 1 and 2 and C4BP CCPs 1 and 2 (23), the crystal structure of CR1 CCPs 15 and 16 (Protein Data Bank File 1GKN) and DAF CCPs 2 and 3 (Protein Data Bank file 1OK3) were converted to be represented in the internal coordinate system using the program ICM (Molsoft). The CCP domains of KCP and C4BP were positioned to adopt the interdomain angle of CR1 CCPs 15 and 16 by tethering all atoms of the corresponding structural anchors (i.e. residues involved in disulfide bonds). The tethers were then minimized, and only the torsional angle values of the residues belonging to the linker region were allowed to change within the range permitted by the Ramachandran plots. Thus, during this process the individual CCP domains behaved as rigid bodies connected through flexible links. Finally, the interdomain loop and its vicinities were submitted to local energy minimization as implemented in the program ICM to resolve clashes and optimize local interactions.
Expression and Purification of KCP-Isolation, cloning, and functional characterization of the four CCP domains of wild type KCP, used as the template for site-directed mutagenesis here, were described previously (6). The CCP domains and 48 additional C-terminal spacer amino acids were subcloned into the prokaryotic vector pBluescript using the XbaI and NotI restriction enzyme sites, and site-directed mutagenesis was performed using the QuikChange TM site-directed mutagenesis kit (Stratagene) per the manufacturer's instructions. Primers used for mutagenesis are presented in Table I. In two cases two separate primer sets were used, one after the other, to obtain the desired mutations (F207A/L209A primers 1 and 2 and E152Q/D155Q primers 1 and 2). Fidelity and specificity of the mutated cDNA was confirmed by DNA sequencing prior to and after subcloning back into the original eukaryotic expression vector, which yields a fusion protein with the addition of an in-frame C-terminal human IgG1 Fc region (24). Stable transfectants of all mutated KCP Fc cDNA were created in Chinese hamster ovary cells using Lipofectin (Invitrogen) following selection with 400 g/ml hygromycin B (Roche Applied Science). Thereafter, cells were grown in a RPMI 1640 cell medium containing 10% fetal calf serum, 4 mM L-glutamine, 50 units/ml penicillin, and 50 g/ml streptavidin. The cell supernatant containing each mutated KCP-Fc was collected, and fusion proteins were purified as follows. Cell debris was removed by centrifugation, and proteins other than KCP-Fc were removed by precipitation with (NH 4 ) 2 SO 4 (242 g/liter). KCP-Fc was then precipitated from the supernatant with (NH 4 ) 2 SO 4 using a final concentration of 372 g/liter. The pellet was then dissolved in phosphatebuffered saline, and KCP was purified by affinity chromatography using protein A-Sepharose (Amersham Biosciences). The protein concentration was determined with amino acid hydrolysis as well as with absorbance measurement at 280 nm. The apparent molecular weight of all proteins was investigated by separating 3 g of each protein by SDS-PAGE on a 10% gel and visualizing it by staining with Coomassie Blue. The protein integrity was also investigated by Western blot analysis, i.e. 1 g of protein was separated by SDS-PAGE on a 10% gel, electroblotted to a polyvinylidene difluoride membrane, and detected with polyclonal anti-KCP rabbit antiserum.
Polyclonal Antibodies against KCP-The polyclonal antibodies were raised in-house by immunization of rabbits with a recombinant fusion protein containing CCPs 1 and 2 of KCP and a cellulose binding domain tag encoded by the prokaryotic expression vector pET-38b (ϩ) (Novagen). The protein was expressed in Escherichia coli BL21-CodonPlus RP (Stratagene), where it was deposited in inclusion bodies. Recombinant protein was extracted by dissolving the inclusion bodies in 6 M guanidine HCl, 10 mM dithiothreitol, 20 mM Tris-HCl, and 10 mM imidazole and then isolated with a nickel-nitrilotriacetic acid Superflow column (2.6 ϫ 12 cm, Qiagen). The protein was refolded in situ on the column by application of a linear gradient (300 ml) from 6 M guanidine HCl, 10 mM dithiothreitol, 20 mM Tris-HCl, and 10 mM imidazole to 50 mM Tris-HCl and 10 mM imidazole, pH 8. The refolded protein was then eluted by 0.7 M imidazole and dialyzed against phosphate-buffered saline overnight prior to use for the immunization of rabbits.
Proteins-C3b and C4b were purchased from Advanced Research Technologies and iodinated with 125 I using the chloramine T method (25). Human plasma FI (26), C1 (27), C2 (28), C3 (29), and C4 (29) were purified from plasma as described previously. C4 and C3 were treated with methylamine to hydrolyze the internal thioester bond, thus changing the conformation to C4met and C3met, respectively, which, for simplicity, are referred to as C4b and C3b in the text. Proteins were prepared by incubation with 100 mM methylamine, pH 8 -8.5, at 37°C for 1 h followed by dialysis in 50 mM Tris-HCl, pH 7.5, and 150 mM NaCl or 10 mM Hepes-KOH, pH 7.4, supplemented with 75 mM NaCl.
Circular Dichroism-The KCP fusion proteins were dialyzed against 150 mM NaF at a concentration of 0.1-0.15 M, and a Jasco J-720 spectropolarimeter was used to analyze the proteins in the far UV region (250 -195 nm; resolution of 1 nm, speed of 20 nm/min, and sensitivity of 50 millidegrees). Each protein was analyzed a total of five separate times, and the results were averaged following subtraction of the background (buffer alone control values).
C3b/C4b Degradation Assay-Test or control cofactor proteins (0.3 M) were incubated with 375 nM C3b or 125 nM C4b, 60 nM FI, and trace amounts of 125 I-labeled C3b/C4b in 50 l of 50 mM Tris-HCl and 150 mM NaCl, pH 7.4. The samples were incubated at 37°C for 1 h (C4b degradation) or 1.5 h (C3b degradation) prior to separation by SDS-PAGE (10 -15% gradient) under reducing conditions. The radiolabeled C3b or C4b was visualized by autoradiography using phosphorimaging analysis (Amersham Biosciences). Cleavage products were quantified by densitometry.
Decay Acceleration of Classical C3 Convertase-Sheep erythrocytes at a concentration of 10 9 cells/ml were washed three times with DGVB 2ϩ (2.5 mM veronal buffer, pH 7.35, 72 mM NaCl, 140 mM glucose, 0.1% gelatin, 1 mM MgCl 2 and 0.15 mM CaCl 2 ). The erythrocytes were incubated with an amboceptor (Roche Applied Science) and diluted 1:4000 at 37°C for 20 min to make EA cells. The EA cells were subsequently washed twice with ice-cold DGVB 2ϩ . C1 suspended in DGVB 2ϩ and warmed to 30°C was added dropwise to the EA cells (equally prewarmed to 30°C) to yield a final concentration of 5 g/ml C1. This mix was incubated for 20 min at 30°C to obtain EAC1 cells. The EAC1 cells were washed twice with DGVB 2ϩ . EAC14 cells were created by the dropwise addition of a suspension of C4 in DGVB 2ϩ , prewarmed to 30°C, to equally prewarmed erythrocytes to a concentration of 1 g/ml C4 followed by incubation for 20 min at 30°C. The cells were then washed twice with DGVB 2ϩ . All incubations were done with agitation. The EAC14 cells were incubated with C2 (5.7 g/ml) for 5 min at 30°C to allow C3 convertase formation. The cells were then placed on ice for 1 min, centrifuged, and resuspended in DGVB 2ϩ . An equal volume of these EAC142 cells was added to wild type or mutated KCP at various concentrations, diluted in DGVB 2ϩ or only DGVB 2ϩ (no inhibition control), and prewarmed to 30°C. Incubation was allowed for 5 min under constant shaking. 100-l aliquots of each sample were removed and added to 100 l of ice-cold guinea pig serum diluted 1:50 in EDTA-GVB (3 mM veronal buffer, 40 mM EDTA, and 0.1% gelatin). Membrane attack complex formation was allowed for 1 h at 37°C under constant shaking. The resulting lysis was detected by measuring the absorption at 405 nm after the sedimentation of intact erythrocytes. Background lysis was obtained from a control lacking C2. The background was subtracted from all values to calculate convertase efficiency. The lysis yielded without the presence of inhibitor was set as 100%.
Binding to C4b Immobilized on HP Affinity Column-C4b (2.4 mg) was coupled to a HiTrap NHS-activated HP affinity column, Sepharose support (Amersham Biosciences), using the manufacturer's instructions. The column was connected to an Ä KTA explorer and equilibrated with 10 mM Tris-HCl, pH 7.35. ϳ60 g of wild type or mutated KCP was added at a flow rate of 1 ml/min followed by washing with 3 ml of equilibration buffer. Bound protein was then eluted with a 10-ml gradient of 0 -0.5 M NaCl in equilibration buffer. Protein elution was detected by measurement of absorbance at 214 nm, and the NaCl concentration at the elution peak was calculated using the conductivity values measured by the Ä KTA. The percentage of KCP that bound at and above 150 mM NaCl was determined with the Unicorn software used to control the Ä KTA explorer.

Three-dimensional Model of KCP Suggests Putative Protein-Protein Interaction
Sites-A three-dimensional model of the four CCP domains of KCP was created by homology based modeling. The crystal structure of VCP was chosen as the major template, because this protein is the most similar to KCP in sequence among the experimentally determined CCP domain structures. For the most part, VCP served as an adequate framework for the modeling of the KCP structure with the exception of two loop segments that were modeled on equivalent structures found in factor H or ␤-2GPI (see "Materials and Methods"). The electrostatic potential at the surface was determined, and exposed hydrophobic amino acids were located. (Fig. 1, A and B). Based on this model of KCP structure, twelve separate mutants were created as presented in Fig. 1C. The mutants were designed to assess the importance of patches of positive or negative charge as well as hydrophobic patches. We were also guided by the knowledge of functionally important residues in human RCAs.
Expression of KCP Mutants-The 12 altered forms of KCP (containing between one and three altered amino acids) were expressed as four CCP domains with an additional 48-amino acid spacer region fused to a C-terminal Fc tail from human IgG (therefore, the resulting proteins are soluble dimers). We have shown previously that dimeric KCP retains the same function as monomeric KCP (8). The purified proteins were analyzed by Western blot analysis ( Fig. 2A) or visualized by Coomassie staining following separation by SDS-PAGE to demonstrate that the recombinant proteins were of the expected apparent molecular mass (Fig. 2B). The Western blot analysis demonstrated that our polyclonal antibody against KCP recognized all mutants, and correct folding of the KCP proteins was confirmed by circular dichroism analysis (Fig. 3). The spectra for all mutants as well as wild type KCP showed a peak at ϳ230 nm, which is characteristic for CCP domains and indicates that the mutations had not introduced any major structural changes (30).
Decay Acceleration of Classical C3 Convertase Requires Sites in CCP1-3-We have shown previously that KCP is a potent decay accelerator of the classical C3 convertase (8). To investigate whether any of the introduced mutations altered the classical decay acceleration function, we performed a hemolytic assay and measured the ability of mutated KCP to protect sheep erythrocytes coated with classical C3 convertases from lysis as compared with wild type KCP (Fig. 4). For classical C3 convertase decay acceleration, we identified important sites located throughout CCP1-3, but not within CCP4. The mutants R20Q/R33Q/R35Q and K64Q/K65Q/K88Q were both almost completely devoid of decay acceleration function (Fig. 4), The three-dimensional structure of the four CCP domains of KCP was modeled on a Silicon Graphics workstation with Insight II software. The crystallized structure of VCP was used as major template. Panels A-C show a solid surface presentation of the model from two opposite angles, where the model has been rotated 180°around the y-axis. A, the electrostatic potential at the surface was calculated using Delphi software. Negatively charged areas are colored red, and positively charged areas are colored blue. B, hydrophobic amino acids exposed at the surface are shown in yellow. C, positively charged (blue), negatively charged (red), and hydrophobic (yellow) amino acids were mutated to glutamine or alanine. The numbering refers to the processed protein, where the signal peptide is cleaved off. Amino acid number 1 is thus the first amino acid of CCP1. Single letter amino acid abbreviations are used with position numbers in the model.
indicating that this positive patch in CCP1 stretching into CCP2 (especially Lys-65) was crucial for this aspect of KCP function. The hydrophobic amino acids Met-113 and Met-120 were also found to be important for classical C3 decay acceleration (Fig. 4C) as well as a second positively charged patch (Lys-131, Lys-133, and H135) and a negatively charged patch (Glu-99, Glu-152, and Asp-155) at the interface between CCPs 2 and 3 (Fig. 4).
Correlation between Impaired C4b Binding and Decreased Decay Acceleration-The classical C3 convertase is comprised of C4b and C2a. Decay acceleration of this convertase by RCA proteins has been thought to occur through direct RCA binding to C4b, resulting in the dislocation of C2a. We addressed the question of whether weak decay acceleration was due to weak C4b binding. The C4b binding capacity of the mutants was studied in an assay where KCP was injected to a column with immobilized C4b. We found that C4b binding was considerably reduced for the mutants R20Q/R33Q/R35Q, K65Q, K64Q/ K65Q/K88Q, and E99Q/E152Q/D155Q (Fig. 5), consistent with their weak decay acceleration. As expected, R136Q/K138Q, H158A/H171A/H213A, F195A/F207A/L209A, and D21A/D23A/ E28A, which were efficient in the decay of the C3 convertase, FIG. 3. Expressed KCP mutants retain the wild type structure. CD analysis was performed on wild type (wt) and mutated KCP recombinant proteins. The ellipticity obtained for the buffer control was subtracted. Shown here are the spectra for wild type and the three mutants with the most important reduction in function. These spectra are representative for all the mutants described in this study, which all presented a positive peak at ϳ230 nm (characteristic for CCP domains).

FIG. 4. Decay acceleration of the classical C3 convertase by wild type (wt) and mutated KCPs.
The decay acceleration was investigated with a hemolytic assay. Sheep erythrocytes coated with the classical C3 convertase were incubated for 5 min in DGVB 2ϩ alone or with wild type or mutated KCP recombinant proteins in concentrations ranging from 0.4 to 0.004 M. The remaining active C3 convertases were detected by the addition of guinea pig serum diluted in EDTAcontaining buffer to allow the formation of membrane attack complexes at 37°C for 1 h. The lysis was determined by measuring the absorbance at 405 nm. Data are given as the average of at least three duplicates. The bars represent the S.D.

FIG. 5.
Binding of KCP to C4b. KCP was loaded onto a Sepharose column coupled with C4b and eluted with a linear salt gradient. Eluted protein was detected by absorption at 214 nm. A, elution profile for wild type and mutated KCP proteins (significant binding was considered to occur at a salt concentration of Ն150 mM NaCl in the elution buffer and is depicted in gray). The experiment was repeated twice, of which one set of data is presented. B, the fraction of injected KCP that bound to C4b at 150 mM NaCl was calculated.
were also found to bind C4b with approximately the same apparent affinity as wild type KCP (Fig. 5). However, we found that M113A/M120A bound C4b relatively well (Fig. 5) despite its poor decay acceleration activity (Fig. 4C).
Sites Involved in FI Cofactor Function-We also reported previously that KCP regulates classical and alternative C3 convertases by acting as a cofactor for FI-mediated cleavage of C4b and C3b, respectively (8). RCA cofactor proteins have previously been found to require both direct C3b-or C4b-binding sites and additional sites not involved in C3b or C4b binding. These latter sites have been hypothesized to be required either for direct interaction with FI or to induce a conformational change in C3b or C4b to allow FI cleavage (17). We assessed the cofactor activity of the mutants in a degradation assay by measuring the production of FI cleavage products. Sites found to be essential to FI-mediated C4b cleavage were the positive patches made up by the amino acids Lys-64, Lys-65, and Lys-88 between CCP1 and CCP2 and Lys-131, Lys-133, and His-135 between CCP2 and CCP3 as well as the negative patch Glu-99, Glu-152, and Asp-155 in the border region between CCP2 and CCP3 (Fig. 6). The cofactor activity for FI-mediated cleavage of C3b was also studied by a degradation assay, the results of which are shown in Fig. 7. Mutants K64Q/ K65Q/K88Q and K131Q/K133Q/H135Q were the least efficient cofactors. Also, H158A/H171A/H213A and F195A/F207A/ L209A were severely defective in function. R136Q/K138Q and E99Q/E152Q/D155Q displayed ϳ50% cofactor activity as compared with wild type KCP.
Affinity (K D ) Determined by Surface Plasmon Resonance-To further characterize the mutants that were the weakest cofactors for C3b cleavage, i.e. K64Q/K65Q/K88Q, K131Q/K133Q/ FIG. 6. C4b degradation in the presence of KCP mutants. The capacity of the mutants to act as cofactors for FI in the degradation of C4b was assessed with a degradation assay. FI (60 nM), KCP (300 nM), C4b (125 nM), and trace amounts of 125 I-labeled C4b were incubated under physiological saline conditions at 37°C for 1 h. The proteins were separated by 10 -15% gradient SDS/PAGE under reducing conditions. A, the C4b cleavage products as visualized by autoradiography. Cofactor activity is shown through the FI cleavage resulting in degradation of the ␣-chain and formation of the C4d band. B, the intensity of the bands corresponding to C4d was determined by densitometry, and the background was subtracted. The value obtained for cofactor activity of wild type KCP (wt) was set as 100%. The average of three separate experiments is presented, and the error bars represent S.D.

FIG. 7. C3b degradation in the presence of KCP mutants
The capacity of the KCP mutants to act as cofactors for FI in the cleavage of C3b was assessed with a degradation assay similar to that used for C4b ( Fig. 6) but with 375 nM C3b and 125 I-labeled C3b instead of C4b. FI cleavage results in degradation of the ␣-chain to 64-, 46-, and 43-kDa products. The 64-kDa band was quantified and used to assess the relative cofactor activities (value for wild type KCP (wt) was set as 100%). The results of three separate experiments are presented, and the error bars represent S.D.
H135Q, F195A/F207A/L209A, and H158A/H171A/H213A, their affinity for C3b as well as for C4b was measured by surface plasmon resonance analysis. The affinity for C3b binding of K64Q/K65Q/K88Q and H158A/H171A/H213A were 4-and 20fold reduced, respectively (compared with wild type binding; Table II). However, K131Q/K133Q/H135Q and F195A/F207A/ L209A were found to bind C3b with almost the same affinity as wild type KCP. To facilitate the comparison of the role of each mutant in C4b binding, classical C3 convertase decay and FI-mediated cleavage of C3b or C4b, we have summarized our experimental results in Table III. DISCUSSION KCP is a viral RCA homologue encoded by KSHV. It inhibits the C3 convertases by decay acceleration, most important for the classical C3 convertase, and by acting as a cofactor for FI mediated cleavage of both C3b and C4b. In this study, sites in KCP that are important for various aspects of its function were located. For this purpose, the three-dimensional structure of the four CCP domains of KCP was predicted by homology modeling using the crystal structure of VCP as the template. This structure was chosen because it is the closest viral homologue for which the structure has been solved. Moreover, it is one of the few models deposited at the Protein Data Bank that contains the structure of a protein constituted by four CCP domains, as is the case for KCP. Amino acid composition within the interdomain linker amino acids and their lengths suggested that KCP could adopt interdomain angles similar to those observed in VCP and were therefore modeled accordingly. Nevertheless, it should be noted that these VCP domain angles were probably restrained by crystallographic packing, and caution should be used when analyzing their structural significance.
The three-dimensional structure of KCP, in combination with knowledge of functional sites in human RCAs, served as a basis for detecting putative functional sites in this viral RCA. To assess the predicted sites experimentally, mutants were created in which 1-3 amino acids were altered. In the case where several amino acids were mutated at the time, the amino acids were in close structural proximity, forming patches that were either hydrophobic or positively or negatively charged.
The decay of the classical C3 convertase was found to require several sites in CCP1-3. Mutants R20Q/R33Q/R35Q and K64Q/K65Q/K88Q were almost completely devoid of decayaccelerating capacity (Fig. 4), indicating that this patch of positively charged amino acids in CCP1, stretching into CCP2, is a key site for interaction with the classical C3 convertase.
Lys-65 seemed to be the most important amino acid for this interaction. Also Met-113 and Met-120, as well as the positive patch comprised of amino acids Lys-131, Lys-133, His-135 and Glu-99, Glu-152, Asp-155 were incapable of decay acceleration. The finding that negative amino acids in KCP are needed for its function might seem contradictory to the common view that positively charged amino acids make up the interaction sites. However, negatively charged amino acids have previously been shown to be important for function. In DAF, mutations E134A (CCP3) and E239A (CCP4) were shown to reduce classical decay accelerating activity (31,32). Glu-239 is situated in CCP 4, close to the border with CCP3, and could be equivalent in position to the negative patch Glu-99, Glu-152, and Asp-155, (considering that CCP1-3 of KCP are more similar to CCP2-4 than CCP1-3 of DAF). It has also been proposed that a negative area in DAF is required for C3 convertase interaction (33) A conceivable view of the mechanism behind the decay of the classical C3 convertase is that the RCA binds C4b and thereby dislocates C2a. Our C4b binding assays show that the mutants R20Q/R33Q/R35Q, K64Q/K65Q/K88Q, and E99Q/E152Q/ D155Q had a much lower apparent affinity for C4b than did wild type KCP ( Fig. 5 and Table II). This probably explains their weak decay acceleration. However, the mutants M113A/ M120A and, to some extent, K131Q/K133Q/H135Q were not so defective in C4b binding that it is likely to account for the weak decay acceleration. These data suggest that Met-113 and Met-120 contributed to decay acceleration by means other than binding to C4b, perhaps through direct C2a dislocation. This latter hypothesis has also been proposed previously for human RCA proteins, which was supported by an investigation of C2a mutants that were not sensitive to RCA-mediated decay (34). Also, the docking of DAF onto factor B (equivalent to C2 in the alternative convertase) suggested hydrophobic interaction points (33).
Our general conclusion drawn from this study was that functional sites in KCP were similar to those described for human RCA proteins. One proof of such similarity is that the site of interaction with C4b (direct binding, cofactor activity, and decay acceleration) is dispersed through three CCP domains. This observation is consistent with previous studies on human RCA proteins; binding to C4b and classical C3 convertase decay has been shown to require 2-3 CCP domains for RCA proteins, i.e. CCPs 2 and 3 of DAF (37), CCP1-3 of C4BP (38), CCPs 2 and 3 (39) or CCP1-4 (40) of the membrane cofactor protein, and CCP1-3 of CR1 (41). Also, the finding that the interaction with C3b required additional sites in KCP CCP4 is similar to previous findings for host complement regulators. RCA binding to and cleavage of C3b has, in some cases, been shown to require an additional CCP domain compared with C4b-binding, i.e. CCP2-4 for DAF (37), CCP1-4 for C4BP (42), and CCP2-4 of the membrane cofactor protein (40). A second apparent similarity in functional requirements between KCP and human RCAs is the involvement of positive amino acids, consistently identified in human RCAs, which are usually located at the junction between two CCP domains. In KCP we identify here a key regulatory site composed of the positively charged residues Arg-20, Arg-33, Arg-35, Lys-64, Lys-65, and Lys-88 that is pivotal for C4b binding (as reflected in the decay of classical C3 convertase and C4b degradation activities).
We compared the structure of sites with the corresponding amino acids (i.e. in the first and second CCP domains of the minimum CCPs required for function) of CR1 (CCP15-16, analyzed by NMR) (43), DAF (CCP2-3; determined by x-ray crystallography) (33), and a previously reported homology model of C4BP (CCP1-2) (23, 44). Direct superimposition of the structures was not possible because of different intermodular angles inherent to each protein. Therefore, we have adjusted the an-gles between CCPs 1 and 2 of KCP and C4BP to be the same as for CR1 (see "Materials and Methods"). We believe that this approach is valid because, although the intermodular angles All of the domains were superimposed using the disulfide bridges (red sticks) as reference points as described under "Material and Methods." Basic residues are shown as blue sticks, and orange shaded areas depict the interdomain linker, thus highlighting its conserved basic character that is present in all four models. E, magnified view of the spatial conservation of basic residues (for the sake of clarity, only KCP and CR1 are shown). The models were rotated ϳ90°in the y-axis compared with the position shown in panels A and B, and the corresponding area is highlighted with a gray shaded box in panel A. Note that the KCP Arg-33 (KCP R33) and the CR1 Lys-912 (CR1 K912) belong to different secondary structure elements. Single letter amino acid abbreviations are used with position numbers throughout the figure. between CCP domains appear to be flexible in solution, they must adopt similar conformations upon the binding of a common ligand (i.e. C4b). Also, the different intermodular angles present in the homology models of KCP and C4BP are inherited from their templates in which the molecules are clearly involved in crystallographic packing and, thus, present a restrained angle probably not reflecting the natural conformation. DAF CCP2-3 superimposes well on CR1 CCP15 and 16 with a main chain root mean square deviation of only 3.57 Å and was therefore not further adjusted. Comparison of KCP, DAF, CR1, and C4BP are presented in Fig. 8. The conservation of the positively charged Lys-64, Lys-65, and Lys-88 functional site of KCP relative to similar residues in CR1, C4BP, and DAF exceeded our expectations. In C4BP, residues Arg-39, Lys-63, and Arg-64 were found to be important for classical decay acceleration and C4b binding (45). For DAF, Lys-125 and Lys-127 but not Lys-126 were required for decay acceleration of the classical C3 convertase (32); however, the corresponding sites (Lys-959, Lys-960, and Lys-961) in CR1 have not yet been assessed by mutagenesis. Nevertheless, Arg-933, Lys-912, and Lys-914 in CCP15 of CR1 were found to be crucial for C4b and C3b binding, and these residues correspond spatially as an exact match to Arg-20, Arg-33, and Arg-35 in KCP. The positive patches between the first and second CCP domain in a regulatory site of the various RCA proteins are not the only sites that have been found to be important for function for this group of proteins. However, the similarity of the CCP1-2 linker patch between the compared proteins indicates that human complement system regulators and their viral homologues such as KCP use similar mechanisms for C3 convertase inhibition based on related structural features. This indication likely reflects required points of contact for docking of regulators to C3b and C4b. It is likely that additional regions are spatially conserved to induce cooperative binding to FI and conformational changes to induce cleavage or dislocation of C2a or Bb. This is the first study where the functional sites for a viral RCA homologue have been described. We consider that the delineation of functional sites of a distant species (virus) is a good means to establish conserved interaction points. This finding will advance the knowledge of RCA function, potentially establishing the minimal requirements needed for C3 convertase inhibition, which could serve as the basis for the development of synthetic C3 convertase inhibitor in the future.