Mutational Analysis of the Primary Substrate Specificity Pocket of Complement Factor B

Factor B is a serine protease, which despite its trypsin-like specificity has Asn instead of the typical Asp at the bottom of the S1 pocket (position 189, chymotrypsinogen numbering). Asp residues are present at positions 187 and 226 and either one could conceivably provide the negative charge for binding the P1-Arg of the substrate. Determination of the crystal structure of the factor B serine protease domain has revealed that the side chain of Asp226 is within the S1 pocket, whereas Asp187 is located outside the pocket. To investigate the possible role of these atypical structural features in substrate binding and catalysis, we constructed a panel of mutants of these residues. Replacement of Asp187caused moderate (50–60%) decrease in hemolytic activity, compared with wild type factor B, whereas replacement of Asn189resulted in more profound reductions (71–95%). Substitutions at these two positions did not significantly affect assembly of the alternative pathway C3 convertase. In contrast, elimination of the negative charge from Asp226 completely abrogated hemolytic activity and also affected formation of the C3 convertase. Kinetic analyses of the hydrolysis of a P1-Arg containing thioester by selected mutants confirmed that residue Asp226 is a primary structural determinant for P1-Arg binding and catalysis.

Complement is a major effector system of host defense. Activation of complement leads to the generation of protein fragments and protein-protein complexes that mediate acute inflammatory responses, phagocytosis and killing of pathogens, and regulation of adaptive immune responses. Activation-associated production of biologically active protein fragments is catalyzed by a group of eight atypical complement serine proteases (SPs) 1 of the chymotrypsin superfamily (1). Understand-ing the structural basis for the highly restricted proteolytic activity of these SPs is an important first step toward pharmacologic control of complement activation (2).
Members of the chymotrypsin family have very similar three-dimensional structures but distinct substrate specificities. To a great extent specificity is determined by the side chains of the amino acid residues that line up the primary substrate specificity pocket (S 1 site). The pocket has three walls formed by residues 189 -195, 214 -220, and 225-228 (chymotrypsinogen numbering has been used for all SPs or SP domains throughout this paper) (3). The presence at the bottom of the pocket of Asp 189 endows trypsin with preference for positively charged Arg and Lys residues (4,5), whereas in chymotrypsin the specificity for bulky aromatics is largely determined by Ser 189 (6). Residues at position 216 and 226 also contribute to substrate specificity (7). All complement SPs exhibit trypsinlike specificity for positively charged Arg residues and all have an Asp at position 189, except for factor B and C2 (Fig. 1).
Factor B and C2 are structurally similar modular proteins that play a central role in complement activation by providing the catalytic subunits of two key enzymes, namely the C3/C5 convertases of the alternative and the classical pathway, respectively. Complement convertases cleave the same single peptide bonds in C3 and C5. In addition to having Asn and Ser, respectively, instead of Asp at position 189, factor B and C2 also lack the highly conserved free N-terminal sequence of SPs. In typical SPs, the N-terminal sequence constitutes an essential structural element largely responsible for the transition from zymogen to active enzyme (8). Full expression of the proteolytic activities of factor B and C2 only occurs in the context of the complexes, C3bBb(C3b) and C4b2a(C3b), respectively (9). The SP domain resides in the C-terminal half of Bb or C2a and is preceded by a von Willebrand factor type A module (VWFA) which is noncovalently associated with C3b or C4b, respectively, in a Mg 2ϩ -dependent manner. These atypical structural features of factor B and C2 indicate a novel activation mechanism and probably also a distinct substrate binding arrangement at the primary specificity pocket.
In addition to their natural protein substrates C3 and C5, factor B and C2 and their fragments Bb and C2a hydrolyze a small number of C3-and C5-like synthetic substrates (11)(12)(13)(14). Overall, C3-like substrates are considerably more reactive than C5-like substrates. However, even toward their best substrates, the k cat /K m values of factor B, Bb, C2, and C2a are about 3 orders of magnitude lower than the 7.8 ϫ 10 6 s Ϫ1 M Ϫ1 value measured under the same conditions for the hydrolysis of the most reactive thioester by trypsin (14). By comparison, the catalytic efficiency (k cat /K m ) of C3bBb for C3 cleavage was reported to be 3.1 ϫ 10 5 s Ϫ1 M Ϫ1 (10). No natural serine protease inhibitor has been found for factor B or C2 and regulation of the proteolytic activity of C3 convertases is effected largely through control of the assembly and decay of the bimolecular complexes. The structural correlates of the low esterolytic activity and extremely restricted substrate specificity as well as the conformational change(s) associated with zymogen activation are not understood. Determination of the structure of the factor B serine protease domain (B-SP) at 2.1-Å resolution has revealed the expected chymotrypsin fold but also unique features of surface loops and of the oxyanion hole. 2 The backbone conformation of the S 1 pocket is similar to that of trypsin, but there are substitutions of functionally important residues. In this study we used site-directed mutagenesis to analyze possible effects of the factor B-specific residues on the assembly and activity of the C3 convertase. The data indicate that Asp 226 is a primary structural determinant of P 1 -Arg binding and that the native conformation of Asp 226 and Asn 189 are important determinants for C3 cleavage.

EXPERIMENTAL PROCEDURES
Construction of Mutant Factor B cDNA-The factor B cDNA clone BHL4-1 (15) in the expression vectors pRc/CMV or pcDNA3 (Invitrogen, Carlsbad, CA) was used as wild type (wt) template in site-directed mutagenesis. Factor B mutant cDNA constructs were obtained by the method of Zollar and Smith (16) as modified by Kunkel (17). Alternatively, the QuikChange Site-directed mutagenesis kit (Stratagene, La Jolla, CA) was used according to the manufacturer's protocol. All cDNA constructs of mutant factor B were verified by restriction mapping and dideoxynucleotide sequencing (18) of the region around the mutation. Oligonucleotides were synthesized by the phosphoramidite method (19), using a DNA/RNA synthesizer (Model 394 Applied Biosystems, Foster City, CA).
Expression of wt and Mutant Factor B cDNA-Transient transfection of COS cells with 30 -40 g of cDNA was performed by electroporation as described (20). Cell culture supernatant containing secreted factor B proteins was harvested 72-90 h after transfection. Cell debris was removed by centrifugation and the supernatant was stored frozen at Ϫ80°C in small aliquots. The concentration of recombinant factor B in the medium was measured by enzyme-linked immunosorbent assay (15), using a rabbit anti-human Bb IgG (50 g/ml) as capturing antibody and the mouse anti-Ba monoclonal antibody (mAb) HA4-ID5 (1.5 g/ml) as reporter. The assay was developed with 1:1000 dilution of affinity-purified goat anti-mouse IgG1 alkaline phosphatase conjugate (Southern Biotechnology Associates, Birmingham, AL) and Sigma substrate 104 (Sigma). Color development was measured at 405 nm. The concentration of factor B was calculated from a standard curve constructed using human serum of known factor B concentration. The sensitivity of the assay was approximately 1-2 ng/ml and the concentration of specific protein in the culture medium ranged from 0.3 to 2 g/ml.
To obtain large amounts of recombinant proteins, stable transfection of Chinese hamster ovary cells (CHO-K1, ATCC) was carried out with selected mutants by a modification of a previously described method (21). CHO-K1 cells were maintained in Ham's F-12 (Cellgro, Herndon, VA) supplemented with 10% heat-inactivated fetal bovine serum (Life Technologies, Grand Island, NY), and 2 mM glutamine at 37°C in a humidified, 5% CO 2 incubator. Forty micrograms of each CsCl-purified plasmid DNA was transfected into 4 -6 ϫ 10 6 CHO-K1 cells by electroporation as described (21). Selection of neomycin-resistant cells was started 72 h after transfection with 750 g of G418 (Cellgro) per ml of the above medium. Subcloning of the G418-resistant cells was performed approximately 7 days after initiating selection by limiting dilution of cells at 0.8 cell/well in 96-well tissue culture plates. Clones were allowed to grow in G418-containing medium with 15% heat-inactivated fetal bovine serum for 10 -12 days before screening for factor B production by enzyme-linked immunosorbent assay. The highest producing wt and mutant factor B clones were selected, expanded, and adapted to large-scale production by growing in suspension culture for 2 weeks. Protein purification was facilitated by culturing cells in ExCell 301 serum-free medium (JRH Bioscience, Lenexa, KS) supplemented with 0.5-2% fetal bovine serum, 2 mM glutamine, and 200 g/ml G418.
Purification of Recombinant wt and Mutant Factor B-One to two liters of the stably transfected CHO cell culture medium were harvested, concentrated to approximately 150 ml, and applied to a 30-ml column of CM Sephadex C-50 equilibrated with 0.1 M sodium acetate, 20 mM ⑀-amino-n-caproic acid, 20 mM EDTA, pH 6.5. Factor B was eluted with a gradient of 0 -0.2 M NaCl in the starting buffer. For further purification, factor B-containing pools were dialyzed against 20 mM Tris-HCl, pH 8.0, and subjected to fast protein liquid chromatography, using a Mono-Q column (Amersham Pharmacia Biotech). Factor B was eluted with a gradient of 0 -0.3 M NaCl in the starting buffer. For some mutants Mono-Q chromatography was repeated. Purity of factor B proteins assessed by 10% SDS-PAGE was between 80 and 95%.
Reactivity of Factor B Mutants with Module-specific MAbs-Two anti-Ba mAbs, HA4 -1D5 (a subclone of HA4 -1A) and FD3-20, and an anti-Bb mAb, HA4 -15, were described previously (22). The mAb 6B3.3 was raised by using as antigen recombinant factor B VWFA module expressed in Escherichia coli. Reactivity of factor B mutants with these mAbs was examined by enzyme-linked immunosorbent assay similar to that described above. The same rabbit anti-human Bb IgG antibody was used in the solid phase, and each of the four mAbs was used as detectant at a concentration of 1.5 g/ml. The assay was developed with goat anti-mouse IgG ϩ IgM alkaline phosphatase conjugate (Jackson Immunoresearch Laboratory, Inc., West Grove, PA) and phosphatase substrate Sigma 104. Values obtained for each mAb were normalized to those measured for HA4 -1D5 and represent the average of two separate experiments.
Solid-phase Cobra Venom Factor (CoVF) Binding Assay-Binding of wt and mutant factor B to CoVF was determined by enzyme-linked immunosorbent assay as described (23). Culture medium from transfected COS cells containing wt or mutant factor B was dialyzed against half-strength veronal-buffered saline (0.5 ϫ veronal-buffered saline, 2.5 mM sodium 5, 5-diethylbarbiturate, pH 7.4) containing 5 mM MgCl 2 at 4°C overnight. Serial dilutions of factor B in the same buffer were then added to microplates coated with CoVF (Quidel, San Diego, CA). Binding of factor B to CoVF was allowed to occur in the absence or presence of 1.5 g/ml factor D at 37°C for 2 h. Bound factor B or Bb were detected with rabbit anti-Bb IgG (50 g/ml) and goat anti-rabbit IgG alkaline phosphatase conjugate. Results represent the average values of two separate experiments.
CoVF-mediated Factor B Cleavage by Factor D-COS cells (4 -6 ϫ FIG. 1. Alignment of partial amino acid sequences of factor B, C2, chymotrypsin, and trypsin. Residues that form the walls of the primary specificity pocket are shaded. The catalytic triad residue Ser 195 is boxed and marked by an asterisk. Arrows indicate residues targeted for site-directed mutagenesis. Numbers at the top are for residues of the chymotrypsinogen sequence and those at the bottom are for the factor B sequence. CHT, bovine chymotrypsin; TRP, bovine trypsin; HC2, human C2; HFB, human factor B. 10 6 ) were transiently transfected by electroporation with wt or mutant factor B cDNA as described above. The cells were metabolically labeled 72 h later in 1 ml of Dulbecco's modified Eagle's medium without methionine, supplemented with 250 Ci of [ 35 S]Met (specific activity ϳ 1000 Ci/mmol, Amersham Pharmacia Biotech or ICN Radiochemical, Irvine, CA.) for 30 min and chased with cold methionine in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum. After a 3-h chase, 650-l aliquots of the culture supernatants were collected, supplemented with 25 mM Tris-HCl, pH 7.4, 2.5 mM MgCl 2 , and incubated for 2 h at 37°C with factor D (300 and 2 ng) in the absence or presence of 5 g of CoVF. Labeled factor B and Bb were immunoprecipitated by using rabbit anti-Bb IgG antibody and Staphylococcus aureus protein A and analyzed by SDS-PAGE as described (24). To assess factor B cleavage, gel slices corresponding to the autoradiographed bands and blank spaces were cut and digested with 15% H 2 O 2 at 56°C overnight. The blank gel cuts were used to subtract background radioactivity. The released radioactivity was measured with Bio Safe II scintillation fluid (RPI, Mount Prospect, IL) in an LKB liquid scintillation counter (Model 1215 LKB, Gaithersburg, MD) (25).
Factor B Hemolytic Assay-Sheep blood erythrocytes carrying C3b (EC3b) were prepared as described (22), by using freshly purified human factor B (22), factor D (26), and C3 (27). Serial dilutions of culture medium containing wt or mutant factor B were added to 7.5 ϫ 10 6 EC3b, 12.5 ng of factor D, and 125 ng of properdin (Sigma) in a total volume of 150 l in 0.5 ϫ veronal-buffered saline containing 2.5% dextrose, 2.5 mM MgCl 2 , 10 mM EGTA, and 0.1% gelatin. Formation of C3 convertase, C3bBb(P), was carried out at 30°C for 30 min. Then, 0.5 ml of guinea pig serum diluted 1:40 with 10 mM EDTA in veronalbuffered saline was added as source of C3 to C9 and the reaction mixture was incubated for 1 h at 37°C. Percent lysis and hemolytic units/g were calculated as described (28). Values of specific hemolytic activity of each mutant were normalized to that of wt factor B and represent the mean Ϯ S.E. of at least three independent determinations, each performed in duplicate.
C3 Cleavage Assay-C3 was freshly isolated from plasma of a normal individual as described (27) except that a final chromatographic step using hydroxyapatite fast protein liquid chromatography (Amersham Pharmacia Biotech) was added. Purified wt or mutant factor B (50 ng) was mixed with C3 (75 ng) with or without 150 ng of CoVF and 12.5 ng of factor D in a total volume of 25 l of 25 mM Tris-HCl, pH 7.4, containing 75 mM NaCl and 5 mM MgCl 2 . After incubating at 37°C for 1 h, 10 l of each reaction mixture was analyzed on 7.5% SDS-PAGE. C3 and C3 fragments were detected on Western blots by using goat anti-human C3 IgG (Cappel, Durham, NC) and affinity-purified rabbit anti-goat IgG F(ab)Ј 2 horseradish peroxidase conjugate (ICN). The ECL luminescent detection system (Amersham Pharmacia Biotech) was utilized to visualize C3 polypeptide chains following the manufacturer's protocol. The amount of C3 conversion was determined by scanning ␣ and ␣' chain using ScanMaker 5 scanner (MicroTek Lab, Inc., Redondo Beach, CA) and band intensity was quantified using software NIHimage1.58.
Esterolytic Assays-The rate of hydrolysis of Z-Lys-Arg-SBzl (Peninsula Laboratories Belmont, CA) was measured by a modification of the method of Kam et al. (14). Assays were carried out in microplate wells. The B-SP was expressed by Sf9 insect cells infected by recombinant baculovirus and isolated from the serum-free Excell 401 media using Bio-Rex 70 and Mono S ion exchange chromatography. 2 The recombinant B-SP consists of a vector-derived tripeptide Ala-Asp-Pro at the N terminus and the C-terminal 295 amino acid residues of factor B. Purified factor B or B-SP (0.11-0.2 M) was added to 0.08 to 0.8 mM Z-Lys-Arg-SBzl and 1.6 mM Ellman's reagent 5,5-dithiobis-(2-nitrobenzonic acid) (Sigma) in 250 l of 0.1 M HEPES, pH 7.5, containing 0.5 M NaCl and 16% Me 2 SO. Factor B was omitted from control wells used for measuring background hydrolysis of the substrate. Esterolytic rates were measured kinetically for 15 min by using a V max kinetic microplate reader (Molecular Devices, Menlo Park, CA). Kinetic constants were determined by the Lineweaver-Burk method based on at least five substrate concentrations. Correlation coefficients in all cases were greater than 0.98.

RESULTS
To understand the structural implications of the unique factor B residues in and around the primary specificity pocket, the serine protease domain (B-SP) was expressed using a baculovirus system and its crystal structure determined at 2.1-Å resolution by multiple isomorphous and molecular replacement methods. 2 As expected, B-SP was found to display a chymo-trypsin-like, two ␤-barrel structural fold. In the active center, the catalytic triad residues, Asp 102 , His 57 , and Ser 195 , and the nonspecific substrate-binding site (Ser-Trp-Gly 214 -216 ) have typical serine protease configurations (Fig. 2). However, the oxyanion hole displays a zymogen-like conformation due to the inward orientation of the carbonyl oxygen atom of Arg 192 , the backbone of which together with those of Cys 191 , Gly 193 , and Asp 194 form a single-turn 3 10 helix. The three walls of the primary specificity pocket are formed by residues 189 -195, 214 -220, and 225-228. The backbones of these residues, except for the single-turn helix, can be superposed on those of the corresponding residues of trypsin. Asn 189 is located at the bottom of the pocket, replacing the highly conserved Asp of other SPs with trypsin-like substrate specificity. However, the side chain of Asp 226 , which replaces Gly 226 of trypsin, extends toward the bottom of the pocket which suggests that it may be directly involved in binding the P 1 -Arg of the substrate substituting for Asp 189 of other trypsin-like SPs. An Asp residue also replaces a conserved Gly of other SPs at position 187. Asp 187 of factor B is located directly beneath the pocket and forms a salt bridge with Lys 163 . To investigate the possible participation of the three residues, Asp 187 , Asn 189 , and Asp 226 , in substrate binding and catalysis, factor B mutants at these positions were constructed and assayed. In addition, the functional role of Pro 188 , not found at this position in other SPs, was also assessed. In most cases, two independent clones for each mutant were expressed and analyzed to avoid artifactual results. In all cases, results of functional analysis of the two clones of each mutant were consistent. This suggested that functional differences from the wt resulted from the amino acid substitution at the mutation sites.
Reactivity of Factor B Mutants with Module-specific MAbs-To probe for possible effects of the mutations on the overall structure of the molecule, we tested the reactivity of the mutants with a panel of module-specific mAbs. The anti-Bb mAb HA4 -15 (22) has been shown to recognize an epitope on the SP domain (data not shown). MAbs FD3-20 (anti-CCP1-3) and HA4 -1D5 (anti-CCP2) bind to distinct epitopes on the Ba fragment (29), while 6B3.3 (␥1,) recognizes an epitope on the VWFA module at or near the C3b-binding site (data not shown). We did not observe substantial differences in the reactivity of the mutants with the four mAbs (data not shown), suggesting that all epitopes tested are retained in their native conformation.
Formation of the CoVFB and CoVFBb Complexes-Expression of proteolytic activity by the factor B SP domain requires binding of factor B to C3b and its proteolytic cleavage by factor D. Introducing mutations in the SP domain could alter C3b binding and/or susceptibility to factor D cleavage, although these functions have been assigned to distal parts of the molecule, namely, the CCP and the VWFA modules (1). We examined the ability of factor B mutants to form the CoVFB and CoVFBb complexes. Choice of CoVF over C3b was dictated by the much longer half-life of the complexes, which facilitates detection. All mutants showed dose-dependent binding to CoVF in the absence (data not shown) and presence (Fig. 3) of factor D. Enhancement of binding to CoVF was observed in the presence of factor D for all mutants. Factor B carrying single mutations at positions 187 or 189 had essentially the same binding activity as wt factor B, except for the D187Y mutant, which only formed about half as much CoVFBb as wt factor B. In the D226 panel of mutants, surprisingly only D226N had wt binding activity. The same substitution combined with N189D resulted in 50% reduction of binding to CoVF compared with either the D226N or N189D mutant. The trypsin-like mutation D226G alone or in combination with the N189D mutation caused 60 and 87% reduction, respectively, in CoVFBb complex formation. Similar reductions in CoVF binding ability of the mutants was also observed without factor D cleavage (data not shown). The results suggested that, with the exception of the D226N mutation, substitutions at position 226 affect initial binding of factor B to CoVF thus sensitivity to factor D proteolysis, since binding is a prerequisite for factor B cleavage. In a more direct factor B cleavage assay, conversion of biosynthetically labeled factor B to Bb by factor D in the presence of CoVF was analyzed by SDS-PAGE and autoradiography (Fig. 4). The results correlated well with the binding data. Mutant D226N was as sensitive to factor D cleavage as wt factor B. Mutants D226N/N189D, D226G, and D226G/N189D were less susceptible to factor D with conversion to Bb estimated at 53, 27, and 16%, respectively, of that of wt factor B at the high concentration of factor D. The combined results suggest that although the overall structural integrity of the mutants was preserved, as indicated by equivalent reactivity with the module-specific mAbs, amino acid substitutions in the SP domain apparently affected CoVF/C3b binding, which is mediated by sites on the other two domains of the molecule.
Hemolytic Activity of Factor B Mutants-The effects of the mutations on the ability of factor B to cleave/activate C3 and C5 were assessed by a hemolytic assay. The hemolytic activity of the mutants relative to that of wt factor B is illustrated in Fig. 5. Elimination of the negative charge of Asp 187 in mutants D187A, D187N, and D187S resulted in 50 -60% loss of hemolytic activity. Substitution of Tyr at the same position caused a more pronounced decrease in hemolytic activity, approximately 80%. The data suggest that the bulky hydrophobic side chain of Tyr is not favored and that full expression of factor B hemolytic activity requires the salt-bridging conformation of Asp 187 . Ala mutation at position 188 in the mutant P188A did not have significant effect on the hemolytic activity.
As revealed in the crystal structure, Asn 189 and the side chain of Asp 226 are located at the bottom of the primary specificity pocket and appear to be accessible to the P 1 -Arg of the substrate (Fig. 2). Replacement of Asn 189 with charged residues, either Asp or Lys, reduced hemolytic activity by 95%, while the Ala mutant retained approximately 30% of wt activity. Although eliminating the negative charge from Asp 226 in the D226N mutant did not affect the assembly of the CoVFBb complex (Fig. 3), it completely abrogated the C3/C5 convertase activity. Replacement of the same residue with Gly present in trypsin also resulted in complete loss of hemolytic activity. Again the loss of hemolytic activity was out of proportion to the only moderately reduced ability to form the CoVFBb complex (Fig. 3). Attempts to construct a trypsin-like pocket by reassigning the negative charge to position 189 in the double mutants D226N/N189D and D226G/N189D failed to restore factor B hemolytic activity, despite the residual CoVF binding activity (Figs. 3 and 5). The hemolytic data strongly indicate that Asp 226 plays a critical and highly specialized role in the expression of C3/C5 convertase activity by factor B. Residue Asn 189 and Asp 187 are also of importance for expression of factor B-dependent proteolytic activity. In contrast, the Pro residue at position 188 has no apparent functional role and likely serves as spacer between structurally crucial residues.
C3 Cleavage Assay-Decrease of the factor B hemolytic activity could reflect a defect of C3 and/or C5 cleavage. The effects of the mutations on C3 proteolytic activity were assessed by a direct cleavage assay. Wt factor B and selected mutants were permanently expressed in CHO cells and purified. Fluid-phase C3 convertases were formed with CoVF in the presence of factor D. Conversion of C3 to C3a and C3b was assessed by the appearance of the ␣' chain of C3b on SDS-PAGE (Fig. 6). As shown, under the experimental conditions used, wt factor B converted 45% of ␣ to ␣' chain, while there was no conversion observed in controls not containing CoVF and factor D. The N189A mutant demonstrated 37% of wt proteolytic activity. This is consistent with the expression of 29% of wt hemolytic activity by this mutant (Fig. 5). As expected from the lack of hemolytic activity, there was no detectable C3 cleavage by the D226N and D226N/N189D mutants even after prolonged exposure of the film. However, there was trace amount of ␣ chain cleavage by the N189D mutant, seen more clearly after long exposure of the film. The C3 cleavage study demonstrated that at least for the factor B mutants tested loss of hemolytic activity could be attributed to loss of proteolytic activity for C3.
Esterolytic Activity-Because C3 is a large protein substrate, extensive molecular contacts with C3b-bound Bb are probably required for its proteolysis. Hydrolysis of small synthetic thioester substrates containing Arg at the P 1 site could provide further insights into substrate recognition. In the present study we chose Z-Lys-Arg-SBzl as substrate because it was shown to be the most reactive among the P 1 Arg-containing C3 or C5-like substrates tested by Kam et al. (14). The catalytic efficiency (k cat /K m ) of recombinant wt factor B was 1135 M Ϫ1 s Ϫ1 (Fig. 7) which is similar to the 1370 M Ϫ1 s Ϫ1 value reported previously for native factor B (14). The recombinant B-SP had k cat /K m of 198 M Ϫ1 s Ϫ1 , which is 5.7 times lower than that of intact factor B. Measurement of individual kinetic parameters showed that the decreased k cat /K m of B-SP was mainly due to a 4-fold increase in K mi Of the mutants tested, D226N showed 50-fold slower catalytic rate than wt factor B. However, placement of a negative charge at position 189 on the D226N background partially restored esterolytic activity. As shown, the k cat /K m of the double mutant D226N/N189D was about 10-fold higher than that of D226N. As indicated by the lower than wt factor B k cat and unaltered K m , decreased catalytic efficiency of these two mutants could be directly attributed to the decreased catalytic rate. These results strongly suggest that the negatively charged Asp 226 determines binding specificity and catalytic efficiency for the substrate Z-Lys-Arg-SBzl. Substitutions of Asp or Ala for Asn 189 in N189D and N189A caused 2.7-and 6.6-fold lower activity, respectively. Although N189A factor B had slightly lower esterolytic activity than N189D factor B, it had substantially higher proteolytic activity for C3 (Fig. 6). Our findings demonstrated that in addition to Asp 226 , Asn 189 also participates in substrate recognition and in determining specificity for C3. Apparently, the structural configuration of residues Asp 226 and Asn 189 of factor B is critical for recognition and cleavage of C3 and C5. DISCUSSION Determination of the structure of the SP domain of factor B revealed a number of novel insertions and deletions compared with typical SPs and also certain unique structural features of the catalytic apparatus, especially in the primary specificity pocket (data not shown). In the present study, mutational analysis of factor B residues in and around the primary specificity pocket was performed to investigate structural correlates of substrate recognition at the S 1 site. The results are discussed in light of the large amount of available information on SP specificity.
Our results clearly demonstrate that Asp 226 of factor B is a critical structural determinant for substrate binding and catalysis, substituting for Asp 189 of other SPs with trypsin-like specificity. Functional analysis of the D226N mutant provided the most clear-cut results. The observed loss of esterolytic and proteolytic activity of this mutant could be attributed solely to a catalytic defect resulting from inappropriate engagement of the P 1 -Arg in the S 1 site, while other functional sites necessary for the proteolytic activation and substrate binding appeared to be well preserved. A sharp 50-fold decrease in catalytic rate (k cat ) indicates that a negative charge at the bottom of the primary pocket is essential for efficient catalysis, but not for overall substrate binding affinity, because the K m is not altered by the Asn substitution (Fig. 7). Apparently, hydrogen bond formation of the P 1 -P 3 residues to the nonspecific substratebinding site, Ser-Try-Gly 214 -216 , and hydrophobic anchoring of the P 2 and P 3 side chains to S 2 and S 3 pockets, respectively, provide sufficient binding force. Also it seems likely that Asn 226 provides additional binding energy, probably by hydrogen bonding with P 1 -Arg. However, positioning of the scissile bond relative to Ser 195 and the oxyanion hole through the putative hydrogen bonds may differ from that effected by the direct ionic contact made by Asp 226 in wt factor B. Replacing Asp 226 with Asn affected equally esterolytic and C3 proteolytic activity, although D226N factor B could form a CoVFBb complex. In a recent report Hourcade et al. (30) also found that substitution of various residues (Asn, Ala, Ser, and Tyr) for Asp 226 caused severe reduction in proteolytic activity despite normally assembled C3bBb complex. It is of special interest that the conservative substitution of Glu for Asp 226 also abrogated C3 proteolytic activity. This observation suggests that accurate positioning of the carbonyl group of P 1 -Arg of C3 relative to the nucleophilic Ser 195 O-␥ and oxyanion hole can only be achieved by the native residue Asp 226 . A corresponding trypsin mutant, D189E, displayed 2-3 orders of magnitude decrease in catalytic efficiency (k cat /K m ), associated with a 40-fold shift in the preference from Arg to Lys substrates relative to wt trypsin (31). Apparently, the additional methylene group distancing the carboxylate of trypsin D189E from the peptide backbone within the narrow S 1 pocket impeded the proper positioning of the side chain of Arg, which is longer and larger than that of Lys. The loss of C3 catalytic activity by D226E factor B (30) can probably be attributed to a similar spatial effect.
Another structural characteristic of the S 1 pocket of factor B is a hydrogen bonding network formed by the carboxyl oxygens of Asp 226 and pocket residues Asn 189 , Thr 190 , and Arg 225 (Fig.  2). This effectively reduces ionic bonding potential available for making contacts with P 1 -Arg of the substrate. On one hand, this distinct feature could possibly explain the overall low esterolytic activity of factor B, Bb (12)(13)(14), and B-SP (Fig. 7). On the other hand, it implies the need for additional bonding between P 1 -Arg and other pocket residues. The side chain of Asn 189 faces the carboxyl of Asp 226 from the opposite wall and occupies a central position at the bottom of the specificity pocket. Although the position of the Asn 189 side chain is about 0.5-1.0 Å lower than that of Asp 226 , it appears accessible to the substrate. Our results indicate a supporting role for Asn 189 in substrate recognition and catalysis. Substitution of Ala, Asp, or Lys at this position caused substantial reduction or abrogation of hemolytic activity, which paralleled a similar reduction in C3 proteolytic activity (Figs. 5 and 6). The Ala substitution caused a decline in synthetic substrate binding affinity (K m ) and catalytic efficiency (k cat /K m ), which strongly indicates participation of Asn 189 in substrate recognition. The amine group of the Asn 189 side chain may mediate P 1 -Arg binding through a hydrogen bond. Absence of this potential binding force may compromise accurate register of P 1 -Arg of C3 for catalysis. Substitution of a charged residue, Asp or Lys for Asn 189 in N189D and N189K, respectively, abrogates C3 proteolytic activity of the C3-or CoVF-bound Bb. Interestingly, the N189D mutant retains substantial esterolytic activity toward the synthetic substrate. These results suggest that the reconstructed S 1 pocket, with free carboxyls at positions 226 and 189, despite its altered geometry could register to the His 57 -Ser 195 dyad, the Arg bond of the synthetic substrate but not that of C3. The free leading or leaving group of the synthetic substrate may account for the observed binding flexibility.
C2 and factor B have identical proteolytic specificity for single Arg peptide bonds of C3 and C5 so that their substratebinding sites can be presumed to be very similar in geometry and chemical nature. Thus, it is not surprising that C2 has Asp and Ser at positions 226 and 189, respectively (Fig. 1). Besides factor B and C2, an acidic residue is also present at position 226 in a few additional members of the chymotrypsin family, namely fiddler crab collagenase (cCOLL) (32), human cathepsin G (CATG) (33), protease 3 (hPRO3) (34), and neutrophil elastase (hnELA) (35). In contrast to C2 and factor B these serine proteases display relatively broad substrate specificity. cCOLL and CATG recognize not only basic but also large hydrophobic side chains (32,36). The Arg/Lys substrate preference is mainly attributed to the presence of Asp 226 /Gly 189 in cCOLL and of Glu 226 /Ala 189 in CATG within the S 1 pocket. The large and flexible S 1 pocket in cCOLL allows this enzyme to adjust to different shapes of the P 1 side chain. Removal of the negative charge from the cCOLL S 1 pocket in the D226G mutant resulted in a significant decrease of catalytic efficiency toward Arg/Lys substrates (37). Similarly to Asp 226 in factor B and cCOLL, the corresponding Glu 226 in human CATG has only one carboxyl oxygen available for substrate binding (33). This may be responsible for the relatively slow catalysis of substrates with P 1 -Lys or Arg. However, the presence of a negatively charged residue at position 226 is not a sufficient condition for specificity for basic residues. Neither hPRO3 nor hnELA, both of which have an Asp 226 , recognizes a Lys or Arg-P 1 residue. The two enzymes display close similarity of their S 1 sites and cleave after small mostly hydrophobic residues, such as Leu/Ile (hnELA), Ala/Ser (hPRO3), and Val/Met (hnELA and hPRO3) (38). The presence of Ile and Val at position 190 of hPRO3 and hnELA, respectively, seems partially responsible for their substrate specificities. In hnELA, loss of specificity for basic residues has been attributed to inaccessibility of Asp 226 that is shielded by Val 190 and Val 216 . Similarly, Asp 226 of hPRO3 is also shielded by Ile 190 and Val 216 . Taken together, the data indicate that Arg/Lys substrate specificity is structurally determined not only by the presence but also by the accessibility of an acidic side chain at the base of the specificity pocket, positioned either at 189 or 226. The carboxyl oxygens of Asp 226 or Glu 226 seem less available to substrate than those of Asp 189 because of participation in hydrogenbonding networks with residues on the wall of the pocket. This appears to be a distinct feature observed in factor B, the neutrophil elastases, and cCOLL.
Structural and functional consequences of altering the Asp 189 of trypsin have been examined by site-directed mutagenesis, kinetic, and crystallographic analysis (39). The negative charge was relocated to the opposite wall of the binding pocket in rat trypsin mutant D189G/G226D. Kinetic analysis showed that, compared with wt trypsin, this relocation of the negative charge caused 10 4 -and 4.5 ϫ 10 2 -fold decrease in catalytic efficiency (k cat /K m ) toward P 1 -Arg and -Lys containing substrates, respectively. The decrease resulted from a much sharper decline in k cat for the Arg than the Lys substrates, whereas the binding affinity (K m ) for both substrates was equally reduced. The crystal structure of D189G/G226D trypsin in complex with inhibitors showed that in its new position, Asp interacts extensively with other residues in the pocket through hydrogen bonds, which greatly reduce its negative charge potential. Similarly to trypsin D189G/G226D, the native Asp 226 of factor B forms hydrogen bonds and this correlates with the low binding affinity and overall low catalytic efficiency toward P 1 -Arg/Lys peptide substrates (12)(13)(14). Re-constructing the pocket of factor B in the D226N/N189D mutant caused complete loss of hemolytic and C3 proteolytic activity (Figs. 5 and 6), although esterolytic activity toward the P 1 -Arg thioester substrate was partially retained (Fig. 7). The kinetic analysis showed that the 80% reduction in esterolytic activity (k cat /K m ) was almost entirely due to reduction in k cat , whereas the K m was not affected. Thus, the exact location of the negative charge at base of the S 1 site and particularly its spatial relationship to the His 57 -Ser 195 dyad and the oxyanion hole, which is altered in trypsin D189G/G226D and factor B D226N/ N189D, are especially critical for efficient catalysis.
In an effort to directly compare factor B to trypsin, a Gly residue was substituted at position 226 either alone (D226G) or in combination with the N189D mutation (D226G/N189D). Neither mutant had hemolytic activity. However, loss of hemolytic activity could not be attributed exclusively to defective substrate recognition at the S 1 site because the ability of these mutants to participate in the assembly of the C3 convertase was also affected (Figs. 3 and 5). Binding of the mutants to CoVF and their sensitivity to factor D cleavage was substantially decreased indicating conformational changes near or at the C3b/CoVF-binding sites, which are presumed to be distal to the mutation sites. Because overall folding of the polypeptide chain and the conformation of antigenic epitopes appeared unaffected, the conformational alteration of the C3b-binding site must be subtle, albeit functionally significant. At present it is not clear how the catalytic center relates spatially to the C3b/CoVF-binding sites. Hourcade et al. (30) also described a conformational change at a site distal from the mutation in the F227A mutant (30). The mutant was cleavable by factor D, but cleavage did not promote the conformational change to a high affinity C3b-binding proteolytically active state, which characterizes wt factor B. The Bb fragment of this mutant was recognized by a Bb-specific mAb at much lower efficiency than the wt counterpart. As viewed in the structure of B-SP, the RDFHIN 225-230 segment forms an extended internal ␤-strand, which is buried within the protein core. Substituting Ala for Phe at position 227 might destabilize the core, affecting the conformation of the surface epitope recognized by the Bb-specific mAb (30). This epitope is probably located near the RDFHIN 225-230 segment and is only reactive in Bb perhaps because it is sterically hindered by the Ba region of intact factor B or because it undergoes a conformational change upon cleavage/removal of Ba. Our D226G mutants might have conformational change(s) within the same region. However, the relationship between the possible conformational change of the antigenic epitope and that of the C3b-binding site is still unclear.
It is of interest that the RDFHIN 225-230 motif is found in factor B and C2 of most animal species, but is absent from all other complement enzymes (1) as well as from other SPs of the large chymotrypsin family (40,38). This underlines the fundamental role of Asp 226 in the function of factor B and C2 in complement activation. Therefore, the native conformation of Asp 226 and Asn 189 or Ser 189 within the S 1 pocket of factor B and C2, respectively, constitutes one of the structural determinants, which have evolved to optimize the highly specific C3/C5 cleavage. However, C3/C5 recognition and hydrolysis require more extensive enzyme-substrate contacts than interaction of the side chain of P 1 -Arg with residues of the S 1 site. The disparity in catalytic activity toward C3 and dipeptide substrates of N189D and D226N/N189D factor B (Figs. 6 and 7) probably reflects the complexity of the interaction between C3b-bound Bb and its natural substrates, C3 and C5.
In the present study, we correlated the crystal structure of B-SP to the detailed mutational analysis of the factor B S 1 pocket. The resulting information contributes to current understanding of the structural basis for factor B and C2 substrate specificity and catalysis. Such knowledge is crucial for designing highly specific inhibitors that could have therapeutic potential for complement-mediated human diseases.