Decay-accelerating factor (DAF), complement receptor 1 (CR1), and factor H dissociate the complement AP C3 convertase (C3bBb) via sites on the type A domain of Bb.

The AP C3 convertase, C3bBb(Mg(2+)), is subject to irreversible dissociation (decay acceleration) by three proteins: DAF, CR1, and factor H. We have begun to map the factor B (fB) sites critical to these interactions. We generated a panel of fB mutations, focusing on the type A domain because it carries divalent cation and C3b-binding elements. C3bBb complexes were assembled with the mutants and subjected to decay acceleration. Two critical fB sites were identified with a structural model. 1) Several mutations centered at adjacent alpha helices 4 and 5 (Gln-335, Tyr-338, Ser-339, Asp-382) caused substantial resistance to DAF and CR1-mediated decay acceleration but not factor H. 2) Several mutations centered at the alpha 1 helix and adjoining loops (especially D254G) caused resistance to decay acceleration mediated by all three regulators and also increased C3b-binding affinity and C3bBb stability. In the simplest interpretation of these results, DAF and CR1 directly interact with C3bBb at alpha 4/5; factor H likely interacts at some other location, possibly on the C3b subunit. Mutations at the C3b.Bb interface interfere with the normal dissociation of C3b from Bb, whether it is spontaneous or promoted by DAF, CR1, or factor H.

The AP C3 convertase, C3bBb(Mg 2؉ ), is subject to irreversible dissociation (decay acceleration) by three proteins: DAF, CR1, and factor H. We have begun to map the factor B (fB) sites critical to these interactions. We generated a panel of fB mutations, focusing on the type A domain because it carries divalent cation and C3bbinding elements. C3bBb complexes were assembled with the mutants and subjected to decay acceleration. Two critical fB sites were identified with a structural model. 1) Several mutations centered at adjacent alpha helices 4 and 5 (Gln-335, Tyr-338, Ser-339, Asp-382) caused substantial resistance to DAF and CR1-mediated decay acceleration but not factor H. 2) Several mutations centered at the ␣ 1 helix and adjoining loops (especially D254G) caused resistance to decay acceleration mediated by all three regulators and also increased C3bbinding affinity and C3bBb stability. In the simplest interpretation of these results, DAF and CR1 directly interact with C3bBb at ␣ 4/5; factor H likely interacts at some other location, possibly on the C3b subunit. Mutations at the C3b⅐Bb interface interfere with the normal dissociation of C3b from Bb, whether it is spontaneous or promoted by DAF, CR1, or factor H.
Complement activation can be initiated by three different pathways: the classical pathway (CP), 1 the alternative pathway (AP), or the lectin pathway (1). Each initiation pathway functions in common to form C3 convertases, active serine proteases that amplify complement activation by cleaving the serum protein C3 into two fragments. One C3 fragment, C3a, is an anaphylactic agent, while the other fragment, C3b, binds covalently to activating targets, marking foreign substances for lysis and/or immune clearance and participating in the selection of and enhancement of the antibody repertoire (2)(3)(4).
The AP C3 convertase is assembled in two steps. First, C3b associates in the presence of divalent cation (Mg 2ϩ ) with factor B, a zymogen that carries the convertase serine protease domain. Second, the C3bB(Mg 2ϩ ) complex is cleaved by the serum protease factor D at a single site in factor B, producing Ba and Bb fragments. The Ba fragment dissociates from the complex, while Bb remains bound to C3b and Mg 2ϩ to form the active C3 convertase, C3bBb(Mg 2ϩ ). A similar process occurs in the assembly of the classical and lectin pathway C3 convertase, C4b2a(Mg 2ϩ ), a complex which is structurally and functionally homologous to the AP C3 convertase.
Activation of human C3 is regulated by a family of related proteins termed the regulators of complement activation (RCA) (5-7). The RCA proteins control complement activation by virtue of two different activities. 1) RCA proteins can promote the irreversible dissociation of complement convertases. 2) RCA proteins can serve as cofactors in the factor I-mediated cleavage of C3b and C4b. The AP C3 convertase, C3bBb(Mg 2ϩ ), is subject to decay acceleration by three RCA proteins: DAF, factor H, and CR1. These three regulators, like all other RCA proteins, are composed of arrays of tandem globular domains termed CCPs (complement control protein repeats) or SCRs (short consensus repeats). Only a portion of each protein is necessary and sufficient for decay acceleration of C3bBb(Mg 2ϩ ): CCPs 2-4 of DAF (8), CCPs 1-3 of CR1 (9), and CCPs 1-4 of factor H (10). While the mechanism of decay acceleration is largely unknown, previous studies have indicated that DAF may interact with the Bb subunit during this process (11). In addition, it has been postulated that a positively charged surface area in DAF CCP 2 and CCP 3 is the primary C3 convertase recognition area (12).
Here, we sought to elucidate the mechanism of AP C3 convertase decay acceleration by mapping sites on the Bb subunit critical to decay acceleration. We previously developed a simple ELISA method for the investigation of decay acceleration of the AP C3 convertase (13). In this report we describe a panel of Bb mutations, their assembly into mutant C3bBb complexes, and the decay acceleration of these complexes by DAF, CR1, and factor H. Interpretation of the results with a structural model of the factor B type A domain implicates two Bb sites critical to DAF and CR1 mediated decay acceleration, one of which is not critical to factor H-mediated decay acceleration.

EXPERIMENTAL PROCEDURES
Production and Assessment of Mutant Factor B Proteins-The transient expression of mutant and wild type factor B proteins was conducted with a human factor B cDNA in pSG5 vector using human 293T kidney cells in serum-free medium (14). Mutations were introduced into the factor B clone using the QuikChange site-directed mutagenesis method (Stratagene).
Each mutant was assayed for its capacity to participate in C3bBb assembly by three assays. 1) A fluid phase assay was used to examine the C3b-and factor D-dependent cleavage of factor B (15). 2) An ELISA-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
based assay monitored the generation of C3bBb(Ni 2ϩ ) complexes. In this assay microtiter wells were coated with C3b (3 g/ml), blocked, and then incubated for 2 h in the presence of factor D (25 ng/ml), 2 mM NiCl 2 , 25 mM NaCl, and phosphate buffer supplemented with 4% BSA and 0.1% Tween 20 with mutant factor B (usually 400 ng/ml) or wild type recombinant factor B (100, 200, 400, and 800 ng/ml); only 20 ng/ml was used for D254G, which binds C3b with greater affinity than wild type (16). The microtiter wells were washed extensively and C3bBb(Ni 2ϩ ) complexes detected by ELISA methods using polyclonal goat anti-human factor B antibody followed by peroxidase-conjugated mouse anti-goat polyclonal antibody (14). Each reaction was performed in duplicate and the OD values averaged. The relationship between OD and factor B concentration calculated by regression analysis was used to derive the relative efficiency of each mutant factor B protein to form C3bBb(Ni 2ϩ ). 3) A factor B-dependent hemolytic assay was used to assess overall function (14).
Assay of Decay Acceleration-The data shown are the results of 29 different experiments with an average of three mutants per experiment (Fig. 1). In each experiment C3bBb(Ni 2ϩ ) was generated in microtiter wells coated with C3b (3 g/ml), blocked, and then incubated for 2 h with mutant or wild type recombinant factor B (in most cases 400 ng/ml) in the presence of factor D (25 ng/ml), 2 mM NiCl 2 , 25 mM NaCl, and phosphate buffer supplemented with 4% BSA and 0.1% Tween 20, as above. The microtiter wells were washed extensively, and the platebound C3bBb(Ni 2ϩ ) complexes were incubated for an additional 10, 20, and 30 min in the presence of DAF (25 ng/ml), CR1-A (see below) (50 ng/ml), factor H (500 ng/ml), or buffer supplemented with 4% BSA and 0.1% Tween 20 alone. The remaining C3bBb(Ni 2ϩ ) complexes were detected by ELISA as described above. Each reaction was performed in duplicate, and the two values averaged. The percent change in average OD at 30 min, which reflected loss of C3bBb(Ni 2ϩ ) in buffer alone, was determined for each factor B form using the following equation: {[(OD 0Ј buffer) Ϫ (OD 30Ј buffer)]/[(OD 0Ј buffer) Ϫ (OD bgd )]}*100. OD bgd was the OD obtained at t ϭ 0 when no factor B was included in the microtiter well. Analysis of the 10-min time point allowed the most sensitive comparison between the mutant and wild type factor B proteins, although analyses performed with the 20-min and 30-min time points yielded consistent results. The percent change in average OD at 10 min, which reflected loss of C3bBb(Ni 2ϩ ) due to the presence of DAF (CR1-A or factor H), was determined for each factor B form using the following The pairwise t test was used to determine statistically significant differences between mutant and wild type proteins and, in some cases, between different mutant proteins. The activity of the mutants was normalized to the wild type values. By this assay the relative specific activities of the three regulatory protein were DAF, 100%; CR1-A, 88%; and factor H, 25%.
Proteins-Complement proteins factor D, C3b, and factor H were obtained from Advanced Research Technologies, La Jolla, CA. DAF consisted of a soluble derivative containing the 4 CCP active region with a single amino acid substitution (N61Q) that removed the N-linked oligosaccharide situated between CCP 1 and 2 (17,18). CR1-A is a soluble CR1 derivative, consisting of CCP 1-7 as previously described (19).
Protein Modeling-The factor B type A domain model was constructed with the spatial coordinates of the iC3b receptor (CR3) type A domain (20), PDB entry 1IDO using the alignment provided by Hinshelwood et al. (21), the automated program Modeler (22), and the InsightII software platform, MSI.

Construction of a Panel of Factor B Type A Mutants-The
Bb fragment consists essentially of two protein domains: a von Willibrand factor type A domain and a serine protease domain (23). From structural studies of type A domains of other proteins, the factor B type A domain is expected to be a globular structure composed of several parallel ␤ sheets surrounded by seven ␣ helices (20,24). The three-dimensional coordinates of the type A domain of factor B have not been reported although spatial models can be produced based on the three-dimensional coordinates of other type A domains (21,25). Thus, a spatial model was generated using the three-dimensional coordinates of the type A domain of CR3 that had been derived by x-ray crystallography. The CR3 domain was chosen because it binds the C3b derivative iC3b in the presence of Mg 2ϩ (26). The model indicates that a divalent cation-binding site is located in a cleft at the apex of the domain, near several C3b-binding determinants (see "Discussion"), while the base of the domain likely adjoins the serine protease domain ( Fig. 2A).
We constructed a panel of factor B type A mutants. Most mutations consisted of alanine substitutions for one or two charged or polar amino acids that are predicted to be on the protein's exterior. According to our model, the mutations were distributed over much of the type A domain surface (Fig. 2B). Among the recombinants produced were 31 mutant proteins that showed function in a factor D-dependent, divalent cationdependent, C3b-binding assay and could be used to detect decay acceleration. Together, the mutations encompassed changes at 29 amino acid positions, including those of the two N-linked glycans.
Effects of Factor B Mutations on Decay Acceleration-Each factor B mutation was examined for its effects on the decay acceleration activity of DAF; mutant and wild type C3bBb(Ni 2ϩ ) complexes were generated in C3b-coated microtiter wells. The Ni 2ϩ cation was used instead of Mg 2ϩ to promote more stable C3bBb complexes and provide for greater signal. Complexes were then washed and treated for 10, 20, or 30 min with either DAF's functional CCPs 1-4 or buffer alone. C3bBb(Ni 2ϩ ) complexes remaining were detected by ELISA using anti-human factor B goat polyclonal antibody. DAF sensitivity was determined for C3bBb(Ni 2ϩ ) generated with each type A mutant and compared with that obtained with wild type C3bBb(Ni 2ϩ ).
The results are shown in Fig. 3 and Table I. Under the conditions used, most of the amino acid substitutions had no statistically significant effect on DAF-mediated decay acceleration. In contrast D254G, K265A/K266A (replacement of Lys-265 with A and Lys-266 with A), Q335A, Y338A, S339A, and D382A were markedly DAF-resistant, exhibiting 1-36% of the sensitivity of the wild type complex to DAF. Four other cases (E301A, E316A, W343A, and E379A) were slightly resistant to DAF (50 -75% relative DAF sensitivity). Of the two mutations that removed an N-glycan, N*260D was markedly DAF-resistant while N*353A was indistinguishable from the wild type.
Additional mutations of positions Tyr-338 and Asp-382 were made to identify the chemical properties of the side chains of these residues most critical to decay acceleration; while loss of the tyrosine 4-hydroxyl group from position 338 in the Y338F mutation reduced DAF sensitivity to 18% of the wild type level, further loss of the phenyl group in the Y338A and Y338S mutations reduced DAF effects to even greater degrees (3-5%) (p Ͻ 0.05 for both Y338A and Y338S when each was compared with Y338F). These observations imply that both the 4-hydroxyl and the phenyl group at position 338 make significant contributions to DAF-mediated decay acceleration. In contrast, while both D382N and D382A were DAF-resistant, there was no statistical difference between the affects of the two muta-tions (p ϭ 0.48). That could imply that only the negative charge is of importance at position 382. CR1, like DAF, is capable of decay acceleration activity of the CP and AP convertases. Previously, the ELISA-based assay used in this study had been used to show that the aminoterminal region of CR1 is necessary and sufficient for the decay of C3bBb (27). In this study we used a CR1 amino-terminal construct, CR1-A (19), to test mutant C3bBb(Ni 2ϩ ) complexes for CR1 sensitivity. In general, CR1-A sensitivity correlated well to DAF sensitivity (Table I and Fig. 4). All complexes that were found to be less than 50% DAF-sensitive were also found to be significantly insensitive to CR1. As with DAF, major effects in CR1 sensitivity were seen with D254G, N*260D, K265A/K266A, Q335A, Y338A, Y338F, and Y338S.
We also examined the effects of the factor B mutations on factor H-mediated C3bBb(Ni 2ϩ ) decay acceleration. In comparison to DAF and CR1, fewer mutants affected factor H sensitivity (Table I and Fig. 4). As with DAF and CR1, major effects were seen with N*260D and K265A/K266A (19 -38% the factor H sensitivity of wild type). Of the other substitutions that

FIG. 3. Effects of factor B mutations on DAF-mediated decay acceleration.
C3bBb(Ni 2ϩ ) assembled with wild type or mutant factor B protein were compared with respect to their sensitivity to DAF. The decay acceleration rate of wild type C3bBb(Ni 2ϩ ) was defined as 100%.
Decay Acceleration Sites of the AP C3 Convertase caused markedly increased resistance to DAF, Y338A and Y338F were statistically significant, but relatively minor compared with their effects on DAF and CR1 decay acceleration (relative to the wild type, Y338A conferred 66% factor H sensitivity but only 3% DAF sensitivity and 0% CR1 sensitivity). Moreover, C3bBb assembled with the D254G protein retained 48% factor H sensitivity, but only 9% DAF sensitivity and 4% CR1 sensitivity. No significant factor H effects were seen with Q335A, S339A, or D382A. Minor reductions in factor H sensitivity were seen with M341A, a substitution where no changes were detected in decay acceleration mediated by DAF or CR1.
Identification of Bb Structural Elements Involved in Decay Acceleration-As indicated, according to the model, the type A mutations made in this study are distributed over much of the domain surface (Fig. 2). In contrast, 12 of the 13 positions implicated in DAF/CR1 function were found in two discrete sites (Fig. 5); one site involves six amino acid positions and is defined by nine mutations (Q335A, Y338A, Y338F, Y338S, S339A, W343, E379A, D382A, and D382N). That site is centered on two adjacent ␣ helices (␣ 4 and ␣ 5). The four most critical positions in this region (Glu-335, Tyr-338, Ser-339, and Asp-382) constitute a discrete functional epitope (Fig. 6). The second site involves five amino acid positions and is defined by four mutations (D254G, N260D, K265A/K266A, E316A). That site consists of the ␣ helix 1 and two adjacent loops and closely follows a portion of the previously reported C3b-binding region (see "Discussion"). Some mutations in the ␣ 1 site dramatically affected factor H-mediated decay acceleration, while those in the ␣ 4/5 site had, at most, very minor effects on factor H activity.
Assembly and Function of DAF-resistant C3bBb Complexes-The mutants that were most dramatically DAF-and CR1resistant, seven carrying mutations in the ␣ 4/5 site and three carrying mutations in the ␣ 1 site, were compared with the wild type factor B protein for other biochemical differences. All ten mutants were found hemolytically active (Table II), with many more active than wild type. It is not clear how to account for higher activity since, in addition to complex stability, hemolytic function could be influenced by many factors including assembly rate, the efficiency of protease activation, and possibly interactions with native red blood cell complement regulator molecules. With the exception of D254G, all of the mutants formed C3bBb(Ni 2ϩ ) complexes 1-to 2-fold as readily as the wild type protein (Table II). D254G, previously shown to accentuate C3b binding and convertase stability (16), formed C3bBb(Ni 2ϩ ) complexes about 50-fold more readily than wild type, and much less protein was required to produce equivalent detectable complex.
We also examined the rate of spontaneous dissociation of C3bBb(Ni 2ϩ ) for all the DAF/CR1-resistant mutant C3bBb(Ni 2ϩ ) complexes (Table II). In the ␣ 1 region, D254G increased C3bBb stability 500%, and N*260D increased C3bBb stability 200%. In the ␣ 4/5 region, Y388A increased stability about 50%. Mutations of the other ␣ 4/5 positions (Glu-335, Ser-339, and Asp-382) had lesser effects on the stability of the C3bBb(Ni 2ϩ ) complex. DISCUSSION The C3 convertases are the central amplification enzymes of the complement cascade, and their proper regulation is essential to promote and facilitate elimination of foreign agents by effector cells while at the same time protecting self-tissues from complement-mediated destruction. A family of regulators, the RCA proteins, has evolved to serve this purpose, and these proteins are found both on cell surfaces and in the serum. In this regard, multiple RCA proteins can promote the dissociation (decay acceleration) of the C3 convertases. In this work we sought to map within Bb the biochemical interactions involved in this process.
Four of the human RCA proteins are capable of decay-acceleration: factor H can dissociate AP convertases, C4-binding protein (C4bp) can dissociate CP convertases, and CR1 and DAF can dissociate both AP and CP convertases. Of these four proteins, three can serve as a cofactor for the factor I-mediated cleavage of C3b and/or C4b: factor H (C3b), C4bp (C4b), and CR1 (C3b and C4b). In the case of DAF the analysis of decay acceleration is simpler because it has no potentially confounding interactions providing for stable C3b or C4b binding or serving as a cofactor for factor I-mediated cleavage.
We began our search for DAF⅐Bb interaction sites by constructing a panel of factor B mutants, focusing on the type A domain because search for of the presence of C3b-binding elements in this region. Previous studies showed that when the isolated factor B type A domain is bound to C3b and the complex treated with trypsin, the type A peptide fragments that remain bound to C3b are from the apex of the domain, near the divalent cation-binding site (21) (Fig. 2A). These assignments were consistent with prior mutagenesis studies directed at the C3b-binding elements of factor B (16,25).
Most of the mutants we prepared were single or double alanine substitutions that retained the capacity for C3bBb assembly and the function required for hemolytic activity. C3bBb complexes then were made with mutant or wild type factor B protein, and the DAF sensitivity of mutant complexes were compared with that of the wild type complexes. Several observations were made. 1) Mutations in 12 of 29 type A amino acid positions interfered with DAF-mediated regulation ( Fig. 3 and Table I). 2) Two type A sites were involved in decay acceleration (Fig. 5). One site was centered around two adjacent ␣ helices (␣ 4 and ␣ 5) defined by six amino acid positions; a second site was located on the ␣ 1 helix and two nearby loops defined by five amino acid positions in a region that includes C3b-binding elements. Two DAF-resistant ␣ 1 mutations also markedly increased C3bBb stability (Table II). 3) The above two sites implicated in DAF-mediated decay acceleration are located on opposite sides of the type A domain (Fig. 5).
Three polar residues, Glu-335, Tyr-338, and Ser-339, and one acidic residue, Asp-382, appear to be most critical in the DAF/ type A interaction. Together they form a contiguous patch on the surface of the type A domain model (Fig. 6). The tyrosine residue at position 338 is particularly relevant; alanine or serine substitution at this single position reduced decay acceleration to a few percent of wild type decay acceleration. Indeed, loss of the 4-hydroxyl group of the Tyr-338 side chain by phenylalanine substitution (Y338F) resulted in substantial loss in DAF sensitivity, suggesting an important role for hydrogen bonding at this point. Given the difference between Y338A and Y338F (Y338F is more active than Y338A), the Tyr-338 phenyl group also appears instrumental in the process. The side chain hydroxyl of Ser-339 is also of interest, since its removal (in S339A) also reduces DAF sensitivity. While it is impossible to determine whether the Tyr, Glu, and/or Ser side chains directly contact DAF, or instead are important structural elements of the ␣ 4/5 contact region, tyrosine has been found to be a frequent "hot spot" of other protein interfaces (28,29).
Each C3bBb complex was also examined for its behavior in the presence of CR1 or factor H. In general, factor B mutations that affected DAF-mediated decay acceleration similarly affected CR1-mediated decay acceleration ( Fig. 4 and Table I). In contrast, fewer factor B mutations altered factor H decay acceleration activity. While the two DAF and CR1-resistant ␣ 1 mutants also rendered C3bBb partially resistant to factor H, the most dramatic DAF and CR1-resistant ␣ 4/5 mutants con- ferred, at most, only subtle changes in factor H sensitivity ( Fig.  4 and Table I).
The ␣ 1 helix partially overlaps the C3b-binding region of the type A domain (Fig. 5). In contrast, the ␣ 4/5 site is somewhat removed from the C3b-binding region (Fig. 5). One simple interpretation of these relationships would be that DAF and CR1 interact with C3bBb at the ␣ 4/5 region and that mutations in the ␣ 4/5 region directly disrupt these interactions. In contrast, the ␣ 1 mutations would exert their effects directly on the C3b/Bb interface, obstructing changes in the interface that are at play during the dissociation of C3b from Bb, whether dissociation is spontaneous or accelerated by DAF, CR1, or factor H.
Although all three regulators were affected by some of the type A mutations, it appears that the mutation/response "spectra" of CR1 and DAF are quite similar, while that of factor H differs ( Fig. 4 and Table I). This could imply that CR1 and DAF utilize similar biochemical mechanisms to promote decay acceleration. Both of these membrane-bound proteins require three CCPs for decay acceleration activity, CCP 2-4 of DAF and CCP 1-3 of CR1, with the two respective regions homologous by amino acid sequence and by intron/exon structure. The factor H protein is also encoded at the RCA cluster and composed of homologous CCPs, but it evolved as a serum protein, which could account for possible mechanistic differences.
Previous studies have shown that DAF CCPs 2-4 are involved in the decay acceleration of AP C3 convertase (8) and a positively charged region between CCP 2 and CCP 3 has been implicated in the process (12,18). Given the importance of the factor B polar residues at the ␣ 4/5 region, the DAF⅐C3bBb interaction may be determined in large part by extensive hydrogen interactions between this region and the positively charged residues of DAF. Given that three DAF CCPs (or three CR1 CCPs) are required for decay acceleration of the AP C3 convertase, it is also possible that DAF and/or CR1 interactions with convertase are not limited to the type A region. Additional DAF⅐C3bBb interactions might extend to C3b or the factor B serine protease domain.
If DAF and CR1 interact with Bb distal to the C3b-binding site, how does this affect the C3bBb association? Allosteric signaling has been observed in other type A domain-containing proteins: In the case of von Willibrand factor, a heparin interaction at the type A1 base down-regulates a glycoprotein Ib (GpIb) binding site located at the domain cleft (21,30,31). Indeed, mutations that are clustered near the base of the domain and disrupt heparin binding also increase GpIb-binding affinity. A similar allosteric mechanism could account for the effects of DAF (CR1) on C3bBb decay and the negative impact of the ␣ 4/5 mutations. It has been previously shown that the isolated factor B type A domain can occur in more than one physical conformation (32). Thus, one possibility is that interaction of C3bBb with DAF or CR1 at the ␣ 4/5 site participates in promoting a type A conformation with relatively low affinity for C3b.