ABSTRACT
Meningococcal factor H binding protein (fHbp) is a human species-specific ligand for the complement regulator, factor H (fH). In recent studies, fHbp vaccines in which arginine at position 41 was replaced by serine (R41S) had impaired fH binding. The mutant vaccines elicited bactericidal responses in human fH transgenic mice superior to those elicited by control fHbp vaccines that bound human fH. Based on sequence similarity, fHbp has been classified into three variant groups. Here we report that although R41 is present in fHbp from variant groups 1 and 2, the R41S substitution eliminated fH binding only in variant group 1 proteins. To identify mutants in variant group 2 with impaired fH binding, we generated fHbp structural models and predicted 63 residues influencing fH binding. From these, we created 11 mutants with one or two amino acid substitutions in a variant group 2 protein and identified six that decreased fH binding. Three of these six mutants retained conformational epitopes recognized by all six anti-fHbp monoclonal antibodies (MAbs) tested and elicited serum complement-mediated bactericidal antibody titers in wild-type mice that were not significantly different from those obtained with the control vaccine. Thus, fHbp amino acid residues that affect human fH binding differ across variant groups. This result suggests that fHbp sequence variation induced by immune selection also affects fH binding motifs via coevolution. The three new fHbp mutants from variant group 2, which do not bind human fH, retained important epitopes for eliciting bactericidal antibodies and may be promising vaccine candidates.
INTRODUCTION
Neisseria meningitidis is a major cause of bacterial meningitis and sepsis worldwide (26). The pathogen is exquisitely adapted to survive in the human host (reviewed in reference 48). When the organism invades the bloodstream, activation of the complement cascade is downregulated by the ability of the organism to bind human complement regulators such as complement factor H (fH), which enables the organism to survive in human serum (35, 47, 48, 58). When fH molecules and active complement components are in close proximity on the bacterial outer membrane, fH can inhibit assembly of the active C3 convertase (C3bBb in the alternative pathway) by competing with factor B for C3b binding and by accelerating decay of the convertase (18, 44, 51, 57). Bound fH also acts as a cofactor for cleavage of C3b into inactive C3b (iC3b), which is mediated by factor I. Similar fH binding mechanisms are employed by a number of pathogens, for example, Streptococcus pyogenes (20, 28), Borrelia burgdorferi (27), and Streptococcus pneumoniae (30).
There are two known meningococcal fH ligands, factor H binding protein (fHbp) (36) and neisserial surface protein A (NspA) (33). Both bind specifically to human fH (24, 33). When N. meningitidis was incubated in rat or rabbit serum, the addition of human fH decreased C3b deposition on the bacteria and enhanced survival of the organism (24, 54). Challenge experiments with N. meningitidis in human fH transgenic rats also supported human fH-dependent mechanisms of evasion of innate immunity (54).
In mice and humans, fHbp vaccines elicited serum complement-mediated bactericidal antibody responses (10, 21, 31, 38, 55, 59, 60), which are the serologic hallmark of protection from meningococcal disease (14). Recombinant fHbp is a component of two vaccines in late-stage clinical development that target capsular group B strains (19, 23, 31, 52). The vaccine potential of fHbp also is being investigated for prevention of capsular group A, W-135, and X strains causing epidemics in sub-Saharan Africa (2, 40, 43).
As described above, fHbp binds specifically to human fH (24). Recent data from human fH transgenic mice indicated that the presence of human fH decreased protective antibody responses to recombinant fHbp vaccines that bound human fH (7). Mutant fHbp vaccines in which arginine (R) 41 was replaced by serine (S) had decreased fH binding and elicited bactericidal antibody responses superior to those elicited by the control fHbp vaccines that bound fH (8). Collectively, the data indicated that binding of human fH to fHbp impairs immunogenicity of the vaccine and that mutant fHbp vaccines designed not to bind human fH can elicit greater protective antibody responses.
Currently, there are more than 500 unique meningococcal fHbp amino acid sequences reported in a central repository (http://pubmlst.org/neisseria/fHbp/). Each fHbp amino acid sequence variant is assigned a unique identification (ID) number (“peptide ID”) such as ID 1 or ID 352. Based on analysis of amino acid sequence similarity, fHbp variants have been subdivided into two subfamilies (A and B) (21, 41), three variant groups (v.1, v.2, and v.3) (38), or 10 modular groups (I through X) (53). Modular groups I to VI accounted for nearly all disease-causing isolates (4, 42). In mice, serum antibodies elicited by recombinant fHbp vaccines had complement-mediated bactericidal activity largely against strains with fHbp from the variant group or subfamily homologous to that of the vaccine antigen (3, 38, 50).
As noted above, mutant fHbp molecules that are engineered not to bind to human fH may be superior vaccine candidates in humans (7, 8). Previous studies of mutants with decreased fH binding focused on fHbp ID 1 (7, 49), in variant group 1. Given the high prevalence of meningococcal isolates with fHbp from variant group 2 (11, 43, 56), it is important to identify non-fH binding mutants from this variant group, which was the purpose of the present study.
MATERIALS AND METHODS
Sequence analysis.We used a combination of approaches for analysis of 11 fHbp sequences which included representatives of the six prevalent modular groups (Fig. 1). Sequences were aligned with MUSCLE version 3.7 (http://www.ebi.ac.uk/Tools/muscle/index.html) (19, 20) configured for highest accuracy. The accuracy of the alignments was confirmed by visual inspection and using the program JALVIEW (21). Alignments also were performed on the individual modular variable segments located between the blocks of invariant residues previously described (18). Networks were generated using SplitsTree, version 4.0 (22), with default parameters. Statistical tests for branch support were performed using the bootstrapping method (1,000 replicates).
Distribution of residue R41 among representative fHbp sequences. (A) Schematic of the six prevalent fHbp modular groups. Each fHbp contains five variable segments, designated A to E, which are flanked by invariant sequences. Each variable segment can be derived from one of two lineages, designated 1 (gray) or 2 (white). A segments from lineage 1 (present in modular groups I, III, and VI) contain an invariant arginine residue at position 41. (Reprinted from Microbiology [4] with permission of the publisher.) (B) Network representation of similarity of fHbp sequences from prevalent modular groups using SplitsTree 4 (29). Eleven fHbp sequences were selected for modeling to represent the three sequence variant groups (v.1, v.2, and v.3) and six prevalent modular groups (4, 42). Individual sequences are annotated with their respective ID number (see the text) followed by the modular group classification (Roman numerals). Modular groups with A segments derived from lineage 1, which contain an arginine at position 41, are shown within the dashed ovals (modular groups I, III, and VI).
Structural modeling.We used the Swiss Model server (http://swissmodel.expasy.org) (1, 13) to generate homology models of fHbp sequence variants. The fHbp models were generated using the alignment mode option and the MUSCLE alignments of each fHbp sequence variant with templates 2W80_D for the fHbp-fH complex (PDB ID 2W80/2W81) (49) and either 3KVD_D (16) or 2Y7S_A (46) depending on the percent identity for the individual proteins (see Table S1 in the supplemental material). We generated five models of each fHbp sequence and five additional models of fHbp in complexes with fH through energy minimization, docking, and molecular dynamics simulations using RosettaDock (34). The reliability of the models generated was assessed using the qualitative model energy analysis (QMEAN) scores (12). The QMEAN Z-score provides an estimate of the quality of a structural model by relating the model to a reference set of structures solved by X-ray crystallography. The score directly indicates how many standard deviations a model's QMEAN score differs from expected values for experimental structures (12). The QMEAN global score reflects the predicted model reliability ranging from 0 (low) to 1 (highest). The highest-scoring models were visually inspected using PyMOL molecular graphics software (http://www.pymol.org). Coulombic surface calculation was performed using Chimera, version 2.1 (45). Calculations of the solvent accessible surface area (ASA), buried accessible surface area, and variation in the energy of solvation (ΔGsolv and ΔΔGsolv) were made using the Protein Interfaces, Surfaces and Assemblies service at the European Bioinformatics Institute PDBePISA server (http://www.ebi.ac.uk/msd-srv/prot_int/pistart.html) (32).
Cloning, expression, and purification of recombinant fHbp.The genes encoding fHbp were subcloned into expression plasmid pET21b as described previously (38, 42). The R41S substitution was introduced as described previously (7). Other mutations were made using the QuikChange II site-directed mutagenesis kit (Agilent Technologies) or the Phusion site-directed mutagenesis kit (Finnzymes); the oligonucleotide primers used are listed in Table 1.
Oligonucleotides used for mutagenesis
Characterization of recombinant fHbp.Recombinant fHbp proteins were purified by Ni2+ affinity chromatography as previously described (3, 7), followed by ion-exchange chromatography using a HiTrap SP column (5 ml; GE Healthcare). Based on SDS-PAGE (NuPAGE; Invitrogen), the respective proteins were >90% pure. Binding of human fH and anti-fHbp monoclonal antibodies (MAbs) to recombinant fHbp was measured by enzyme-linked immunosorbent assay (ELISA) as described previously (6, 10), with the only difference being that the conformational MAbs JAR 4 and JAR 41 were incubated overnight at 4°C instead of for 2 h at room temperature. Characteristics of the six anti-fHbp MAbs used are listed in Table 2.
Properties of the murine anti-fHbp MAbs
Mouse immunogenicity.The immunogenicity of recombinant fHbp vaccines was evaluated in 6- to 8-week-old CD-1 mice (n = 10 mice per vaccine group). Three doses of fHbp vaccine, each containing 10 μg of recombinant fHbp adsorbed with 600 μg of aluminum hydroxide as the adjuvant, were administered intraperitoneally (i.p.) at 2-week intervals. A control group of five mice received the adjuvant alone. Blood samples were obtained 3 weeks after the third injection; sera were separated and stored frozen for subsequent serologic analyses. The experiments in mice were done in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (41a). The protocol was approved by the Children's Hospital & Research Center Oakland Institutional Animal Care and Use Committee.
Measurement of serum antibody responses.Serum IgG anti-fHbp Ab responses were measured by ELISA, which was performed as described previously (9) except that the recombinant fHbp ID 22 was used as the coating antigen instead of ID 1. Serum bactericidal antibody responses were measured as previously described (7, 22). In brief, the bacteria were grown for approximately 2 to 2.5 h to mid-log phase in Mueller-Hinton broth supplemented with 0.25% glucose and 0.02 mmol/liter CMP N-acetylneuraminic acid (Sigma-Aldrich). The bacteria were pelleted by centrifugation and resuspended in Dulbecco's phosphate-buffered saline containing 0.1 g/liter CaCl2 and 0.1 g/liter MgCl2 (Mediatech) (pH 7.4) with 1% (wt/vol) bovine serum albumin (Equitech-Bio). The mouse sera were heated for 30 min at 56°C to inactivate internal complement. The 40-μl bactericidal reaction mixture contained ∼300 to 400 CFU of bacteria, dilutions of the mouse sera, and 20% human serum depleted of IgG as a complement source, which was performed as described previously (7). The serum bactericidal titer was defined as the serum dilution that resulted in a 50% decrease in CFU/ml after 60 min of incubation in the reaction mixture compared with the CFU/ml in negative-control wells at time zero. Typically, the bacteria incubated with negative-control sera (from mice immunized with aluminum hydroxide alone) and complement, or with immune serum and heat-inactivated complement, showed a 150 to 200% increase in CFU/ml during the 60 min of incubation (22, 60).
One of our long-term goals is to assess the fHbp vaccine potential of fHbp against capsular group A, W-135, and X strains causing epidemics in sub-Saharan Africa (43). The majority of the W-135 isolates from this region expressed fHbp ID 22 or 23 from variant group 2. Therefore, we used a capsular group W-135 isolate, CH130W (also referred to as Ghana 7/04), as one of the test strains for the bactericidal assay. This isolate was a relatively high expresser of fHbp ID 23 (43), which differed by one amino acid from fHbp ID 22 in the vaccine; the PorA VR type was 5,2, and the multilocus sequence type was ST-11. As a second test strain we used a capsular group B isolate from California, 03S-0658 (5), which also expressed fHbp ID 23. This strain had the PorA VR type 7-2, 13-1, and the multilocus sequence type ST-1364 (ST-32 clonal complex).
Statistical analyses.Serum antibody titers were log10 transformed prior to statistical analysis. For calculations of geometric mean antibody titers, titers below the detection limit of the assay were assigned a value half of the lowest dilution tested. Two-tailed Student's t tests were used to compare the geometric mean antibody titers between two independent groups of mice. Probability values of less than or equal to 0.05 were considered statistically significant.
RESULTS
Substitution of serine for arginine at residue 41 eliminates binding of fH in fHbp amino acid sequence variants from variant group 1 but not from variant group 2.fHbp has been classified into modular groups based on unique combinations of five variable segments (A, B, C, D, and E), each encoded by genes related to one of two sequence lineages (1 or 2 [Fig. 1A]). Arginine at residue 41 is invariant among fHbp sequence variants with A segments derived from lineage 1. Sequence variants that contain R41 include modular group I (which are expressed by the majority of strains with fHbp in variant group 1) and modular groups III and VI (all strains with fHbp in variant group 2). In contrast, arginine 41 is not present in sequence variants from modular group II or V (all strains with fHbp in variant group 3, which have A segments from lineage 2). The sequence relationships of 11 naturally occurring fHbp sequences are presented in Fig. 1B as a split network, which indicates possible recombination events (29). The sequence variants included both of the known subfamilies, the three variant groups, and the six most prevalent modular groups. The sequence variants with A segments derived from lineage 1, which contain R41, are shown within the dashed ovals (modular groups I, III, and VI). Together, sequence variants with A segments from lineage 1 accounted for >95% of fHbp sequence variants from capsular group B isolates from the United States (42, 56) and capsular group A, W-135, and X isolates from epidemics in sub-Saharan Africa.
In a previous study, replacement of arginine at position 41 by serine (R41S) eliminated fH binding to fHbp ID 1 (variant group 1) (7). To determine whether the R41S substitution decreased fH binding in other amino acid sequences from variant group 1, we prepared R41S mutants of fHbp IDs 4, 9, and 74 (Fig. 1B). In each example, the R41S mutation decreased fH binding (Fig. 2A to C). The control anti-fHbp MAb showed similar levels of respective binding for all three mutants compared with the respective wild-type fHbp variants, which indicated that similar amounts of the proteins were adsorbed to the wells of the microtiter plate (Fig. 2E). We also prepared a mutant of fHbp ID 1 (variant group 1) in which R41 was replaced by alanine (R41A) instead of serine, which also eliminated fH binding (Fig. 2D). This mutant also showed anti-fHbp MAb binding similar to that of the wild-type ID 1 protein (Fig. 2F). Thus, specific replacement of R41 by S was not required for elimination of fH binding.
Effect of replacing arginine at position 41 on binding of fH to fHbps in variant group 1 as measured by ELISA. (A to C) Binding of fH to R41S mutants of fHbp ID 4, 9, and 74, respectively. (D) Binding of fH to R41A mutant of fHbp ID 1. (E and F) Binding of control anti-fHbp MAb JAR 5 to R41S and R41A mutants, respectively. Closed circles, wild-type fHbp (WT); open circles, R41 mutants except as otherwise indicated. Mean values for optical density (OD) at 405 nm and respective ranges are plotted.
We next prepared R41S mutants of three fHbp sequence variants from variant group 2 (IDs 19, 22, and 77 [Fig. 1B]). In contrast with impaired fH binding by this substitution in fHbp sequences from variant group 1, the R41S mutation did not affect fH binding in recombinant fHbp sequences from variant group 2 (Fig. 3A to C). The control anti-fHbp MAb, JAR 31, showed similar respective levels of binding with each of the mutants compared with the respective wild-type fHbp variants (Fig. 3D).
Effect of replacing arginine at residue 41 on binding of fH to fHbps in variant group 2 as measured by ELISA. (A to C) Binding of fH to R41S mutants of fHbp ID 19, 22, and 77, respectively (all in variant group 2). (D) Binding of control anti-fHbp MAb JAR 31. Closed triangles, wild-type fHbp (WT); open triangles, R41S mutant. Mean values for optical density (OD) at 405 nm and respective ranges are plotted.
In order to identify mutations in fHbp in variant group 2 that decreased fH binding, we generated structural models of 11 representative fHbp sequence variants (Fig. 1B), both alone and in complexes with human fH (see Materials and Methods). Each model was visually inspected, and the quality of each one was estimated using QMEAN global and Z-scores. The models received QMEAN global scores of 0.73 or higher (range is from 0 to 1, with 1 being the highest quality), which indicated that the models were of good quality. As a benchmark, the corresponding score for the starting model based on the crystal structure was 0.52. Additional data on the models are presented in Table S1 in the supplemental material.
Figure 4A depicts a model showing the region of fHbp ID 1 (variant group 1) that interacts with fH based on the crystal structure of the complex (49). Thus, the region is inferred to be exposed on the surface of the bacteria. The amino acid residues that are identical across the 11 different sequences are shown in gray. The divergent amino acid residues are shown in green and are located in a central portion of the surface of the fHbp molecule that binds fH and also in three patches located peripherally. The inferred charge distribution on the surface of fHbp ID 1 where the interaction with fH takes place (Fig. 4B) differed significantly from that of fHbp ID 22, which is representative of variant group 2 (Fig. 4C). For example, the middle section of the surface of fHbp ID 1 shows a positively charged region (depicted in blue in Fig. 4B), which was mostly negatively charged in fHbp ID 22 (depicted in red in Fig. 4C). Thus, the models predicted considerable variability in electrostatic potential of the fH-binding surface across different fHbp variant groups and suggested that the residues participating in interactions with fH likely varied between fHbp sequences from different variant groups. These differences may explain in part why amino acid substitutions at R41 (shown by arrows in Fig. 4B and C) eliminated fH binding in fHbp sequences from variant group 1 but not from variant group 2.
Structural variability in the fH-binding surface of fHbp models from different sequence variants. (A) fH-binding surface of fHbp ID 1 (variant group 1). Amino acid positions shown in gray are conserved among the 11 fHbp sequence variants shown in Fig. 1B. Amino acid positions in green depict residues that differ among the sequences. (B and C) Electrostatic potential of the fH-binding surfaces of structural models for fHbp sequence variants ID 1 (variant group 1 [B]) and ID 22 (variant group 2 [C]). Blue, positively charged residue; red, negatively charged residue. The arrow indicates the location of amino acid R41, which was critical for fHbp-fH interaction in variant group 1 fHbp but not in variant group 2 (Fig. 2 and 3).
Prediction of residues contributing to fH binding in different fHbp variants.We used the structural models to calculate the per-residue empirical solvation energy (ΔG; Fig. 5A) and buried surface area (Fig. 5B) of fHbp alone versus fHbp in the complex with fH (see Materials and Methods). Solvation energy and buried surface area values for each residue were calculated, and their average values for all of the models were plotted with the standard errors of the means. We used these data to identify residues that potentially affected the fHbp binding surface directly or indirectly, and we generated a list of all residues with a buried surface area value over 5% of their total solvent accessible area and all residues with a nonzero ΔG. We visually inspected each model for the location of each residue and identified 63 residues that were predicted either to be in direct contact with fH or to affect the binding surface through indirect effects (see Table S2 in the supplemental material).
Per-residue changes in solvation energy and buried surface area upon binding fH. (A) Mean values of the solvation energy effect (ΔG) calculated per residue. The effect on solvation energy when comparing free versus bound surfaces is expected to be negative when the residue strongly contributes to the complex formation. (B) Mean values for the buried surface area (BSA) calculated per residue. Values represent how much of the solvent-accessible surface of a given residue changes between free and fH-bound fHbp models. Higher buried surface area values suggest a role in the interaction with fH. For both sets of calculations, the mean value and the standard error for the 11 models is shown.
Mutations in variant group 2 sequences that impair fH binding.Because of limited resources, it was not possible to investigate experimentally the effect of mutations on all 63 residues. We therefore selected 12 different positions to mutate in recombinant variant 2 proteins (Table 3). Residues Q38, Q126, D201, R80, T220, and H222 were selected because of their predicted ability to form hydrogen bonds with residues in fH, while D211 and E218 were selected because of their predicted ability to form salt bridges (Table 3). Since the main goal was to select positions that affected fH binding without affecting critical epitopes for eliciting bactericidal antibodies, we did not necessarily select the most exposed residues on the surface. Residues predicted to be less exposed—E202, A235, G236, and E248—belong to this category. We made a total of 11 mutants in fHbp ID 22 (variant group 2 [Table 3]), which included one double mutant, T220A/H222A.
Summary of fHbp ID 22 mutantsa
Using purified fHbp adsorbed to the ELISA plate, five of the purified fHbp ID 22 mutant proteins showed no decrease in fH binding (Q38A and A235G), or a slight increase (Q126A, D201A, and E202A), compared to that of the wild-type fHbp ID 22 control (data not shown). The remaining six mutants, R80A, D211A, E218A, E248A, T220A/H222A, and G236I, showed impaired fH binding (Fig. 6).
Binding of human fH to mutants of fHbp ID 22 (variant group 2) by ELISA. (A) R80A and D211A mutants; (B) E218A and E248A mutants; (C) T220A/H222A and G236I mutants. WT, wild-type fHbp ID 22. All of the mutants bound equally well with representative control anti-fHbp MAbs such as JAR 31 (Fig. 7), which indicated that similar amounts of the WT and mutant proteins were adsorbed to the wells of the microtiter plate. OD, optical density.
Preservation of the epitopes among fHbp mutants with impaired fH binding.To determine whether the fHbp ID 22 mutants with decreased fH binding retained epitopes important in eliciting bactericidal antibodies, we tested each of them for binding with a panel of six anti-fHbp MAbs. All of the MAbs except JAR 64 had human complement-mediated activity when tested in combination with other anti-fHbp MAbs that individually were not bactericidal (6, 10, 59, 60). The MAbs also recognized different epitopes (Table 2) (6, 10). By ELISA, the six fHbp ID 22 mutants (R80A, D211A, E218A, E248A, T220A/H222A, and G236I) with decreased fH binding (Table 3) showed similar levels of respective binding with five of the six anti-fHbp MAbs tested (Fig. 7B to F). For the sixth MAb (JAR 4), binding was eliminated by the R80A or G236I substitutions and decreased by the E248A substitution (Fig. 7A).
Binding of anti-fHbp MAbs to fHbp ID 22 mutants with impaired fH binding activity. Murine monoclonal antibody reactivity was measured at different concentrations (x axis). (A to C) JAR 4, JAR 41, and JAR 64, respectively. JAR 4 and 41 bind to overlapping conformational epitopes located in the N-terminal domain of fHbp. JAR 64 binds an unknown epitope believed to be located in the N-terminal domain of fHbp, since it inhibits binding of JAR 41 fHbp (unpublished data). (D to F) JAR 31, 33, and 35, respectively. The MAbs recognize different epitopes in the C-terminal portion of fHbp (Table 2). OD, optical density.
Immunogenicity of fHbp mutants in mice.We immunized wild-type CD-1 mice with each of the six recombinant fHbp ID 22 mutants with decreased fH binding. In wild-type mice, whose serum fH is not bound by fHbp, the mutant fHbp vaccines were expected to elicit antibody responses similar to that of the control wild-type fHbp vaccine unless the mutation resulted in structural changes in the protein and/or loss of conformational epitopes.
By ELISA, the anti-fHbp IgG titers elicited by each of the mutant vaccines were not significantly different from the control wild-type vaccine when measured against the ID 22 wild-type protein (Fig. 8A). To assess functional activity of the serum antibodies, we measured human complement-dependent killing of the bacteria (Fig. 8B and C). In pairwise comparison, two of the mutant fHbp vaccines, R80A and G236I, elicited significantly lower serum bactericidal antibody titers than the control wild-type fHbp vaccine against both test strains. The respective reciprocal bactericidal geometric mean titers (GMTs) for the W-135 strain were 46 and 45, versus 221 (P < 0.05); the respective reciprocal GMTs for the group B strain were 66 and 73, versus 548 (P < 0.01). A third mutant vaccine, E248A, elicited a reciprocal bactericidal GMT of 88 against the group B strain (P = 0.055 versus wild-type vaccine) and a reciprocal bactericidal GMT of 69 against the W-135 strain (P = 0.16 versus wild-type vaccine). For the remaining three mutant vaccines, there were no significant differences in pairwise comparisons of the respective reciprocal bactericidal GMTs with the wild-type fHbp ID 22 vaccine against either of the test strains (P > 0.10).
Serum antibody responses of mice immunized with recombinant fHbp ID 22 mutant vaccines with decreased fH-binding activity. (A) Titers of IgG, measured by ELISA, to wild-type fHbp ID 22. There were no significant differences between the GMTs elicited by the mutant fHbp vaccines compared to that of wild-type ID 22 vaccine. (B) Serum bactericidal antibody responses to a capsular group W-135 strain with fHbp ID 23. (C) Serum bactericidal antibody responses to a capsular group B strain with fHbp ID 23. Bactericidal titers were defined as the serum dilution resulting in a 50% decrease in CFU/ml after 1 h of incubation with human complement compared to CFU/ml at time zero in negative controls. Each symbol represents the titer of an individual mouse, and the horizontal lines indicate the reciprocal GMTs. For pairwise comparisons of a particular group with the wild-type fHbp ID 22 vaccine, *, P < 0.05; **, P < 0.01; and #, P = 0.055. For all other pairwise comparisons between mutant and wild-type vaccines, P > 0.10. Control mice were immunized with aluminum hydroxide alone (all titers <1:10; not shown).
DISCUSSION
Microbial proteins such as fHbp that bind host complement inhibitors are important virulence factors (48, 58) and can be effective vaccine candidates (39). A potential drawback of such vaccines, however, is decreased immunogenicity from binding of the host protein to the vaccine and covering important epitopes or interfering with antigen processing. Our previous studies with human fH transgenic mice indicated that recombinant fHbp that bound human fH (7) elicited impaired serum bactericidal antibody responses compared to the respective responses of immunized wild-type mice whose mouse fH did not bind to the vaccine antigen. Further, the higher the concentration of human fH in the mouse sera, the lower the serum bactericidal responses to the recombinant fHbp vaccine that bound human fH (7). While the biological basis for the impaired fHbp immunogenicity in the presence of human fH is not known, the data provided direct evidence that meningococcal fHbp vaccines that bind human fH may not be optimal antigens.
Based on a crystal structure of fHbp in complex with an fH fragment, Schneider et al. identified two glutamate (E) residues (residues 218 and 239, based on the numbering of the mature fHbp ID 1 sequence beginning with the lipidated cysteine residue), which, when both were replaced by alanine, eliminated human fH binding (49). However, in subsequent studies, this double mutant E218A/E239A fHbp vaccine elicited much lower serum bactericidal antibody responses in wild-type mice than the control fHbp vaccine containing both of the glutamate residues (7, 9). Thus, in a mouse model in which mouse fH did not bind to the mutant or control fHbp vaccines, the two amino acid substitutions adversely affected epitopes that were needed for eliciting bactericidal antibodies and impaired fHbp immunogenicity.
In other studies, recombinant or native outer membrane vesicle (NOMV) vaccines with mutant fHbp in which arginine at position 41 was replaced by serine to impair fH binding elicited slightly lower serum bactericidal titers in wild-type mice than the respective control vaccines (7, 8). However, in human fH transgenic mice, the mutant vaccines elicited up to 19-fold-higher serum bactericidal antibody responses than the respective fHbp vaccines that bound human fH. Thus, the slight loss of immunogenicity from the amino acid substitution, which was evident in wild-type mice, was far outweighed by the substantial enhanced bactericidal activity elicited by the mutant vaccines in the presence of human fH. Collectively, the data indicated that human fH decreased immunogenicity of fHbp molecules that bound human fH and that fHbp mutants that did not bind human fH but which retained critical epitopes for eliciting bactericidal antibodies were promising vaccine candidates for humans.
The non-fH-binding mutant vaccines described above were made in fHbp ID 1, which is in variant group 1. In the present study, we prepared R41S mutants in three additional fHbp sequence variants from variant group 1 and an R41A mutant in fHbp ID 1. In all four mutants there was impaired fH binding. Thus, arginine at position 41 was a critical residue for fH binding not only in fHbp ID 1 but also in three other fHbp sequence variants tested from the same variant group. Furthermore, substitutions at this position with either serine or alanine eliminated fH binding. The mutant variant 1 proteins tested included ID 4, 9, and 74, which are prevalent among capsular group A, W-135, and X strains from Africa. Thus, R41S mutants of these sequences potentially could be employed for an fHbp vaccine that targeted meningococcal epidemics in Africa caused by strains with fHbp in variant group 1. The remaining gap in coverage for a vaccine for Africa is capsular group W-135 strains with ID 22 or 23 in variant group 2, which accounted for about two-thirds of W-135 isolates from Africa (42). In other geographical regions, fHbp from variant group 2 was present in 20% to >70% of invasive meningococcal isolates (25, 42, 43, 56). Since there is little cross-protective anti-fHbp bactericidal antibody response to variant groups 1 and 2 (3, 21, 38), we sought to identify mutants with impaired fH binding activity in fHbp from variant group 2.
Arginine at position 41 is conserved in all known fHbp sequence variants from variant group 2 (53). However, we were surprised that none of the R41S substitutions in variant group 2 proteins decreased fH binding. These data, and the predicted variability in the distribution of electrostatic charges between the fH-binding surfaces of variant group 1 or 2 proteins (Fig. 4B and C), suggested that fH interactions were fundamentally different between the different fHbp variant groups. In previous studies, binding affinities of fH to fHbp also were reported to be higher for sequence variants in variant groups 2 and 3 than in variant group 1 (17, 50), which also suggested that the fH-binding modes may be different between fHbp variants.
Using a structural homology modeling approach, we identified 63 residues that potentially played a role in the interaction of fHbp with human fH. We made 11 mutants in fHbp ID 22 from variant group 2 (Table 3). We chose fHbp ID 22 because, as noted above, this sequence variant was prevalent among epidemic capsular group W-135 isolates from sub-Saharan Africa, and the choice reflected our long-term goal to investigate the fHbp vaccine potential against strains causing meningococcal disease in that region (2, 43).
Six of the fHbp ID 22 mutants with substitutions at positions 80, 211, 218, 220/222, 236, and 248 had decreased fH binding. The fH-binding surface of fHbp is very large and involves residues from both the N- and C-terminal structural domains of the protein (see also Table S2 in the supplemental material for a list of the residues putatively involved in the interaction). The extensive binding surface helps explain why residues of disparate number (in the primary sequence) in both the N-terminal and C-terminal domains can influence fH binding.
To evaluate preservation of important epitopes in the six ID 22 mutants that had impaired fH binding, we measured binding with a panel of six anti-fHbp MAbs and immunogenicity in wild-type mice. Five of the MAbs showed similar levels of binding to each of the six ID 22 mutants compared to the respective binding to the wild-type ID 22 protein (Fig. 7). The exception was JAR 4, which did not bind with two of the mutants (R80A and G236I) and showed partial loss of binding with a third mutant (E248A). The fHbp molecule consists of two domains of anti-parallel β-strands connected by a five amino acid linker (15, 37, 49). JAR 4 is known to recognize a discontinuous conformational epitope located on the N-terminal domain, which may explain why JAR 4 binding was disrupted by the R80A mutation. It is not clear, however, why the G236I and E248A substitutions in the C-terminal domain affected the JAR 4 epitope.
In wild-type mice, the three mutant fHbp ID 22 vaccines that had decreased JAR 4 reactivity elicited lower serum bactericidal responses than the control wild-type ID 22 vaccine. Thus, complete or partial loss of JAR 4 binding was predictive of decreased vaccine immunogenicity. In contrast, the three fHbp ID 22 mutant vaccines that retained JAR 4 reactivity elicited serum bactericidal titers against both the capsular group W-135 and capsular group B test strains similar to that of the respective wild-type fHbp ID 22 vaccine.
Both of the test strains used in the present study expressed fHbp ID 23, which is in variant group 2. The ID 23 amino acid sequence differs by one amino acid from fHbp ID 22 in the control wild-type vaccine. The two sequences also are closely related to other fHbp sequence variants in variant group 2, such as ID 19 or ID 77 (Fig. 1B), which were prevalent among capsular group B strains causing disease in the United States or Europe (11, 41). Collectively, the data suggest that the three mutant ID 22 vaccines with decreased fH binding can elicit protective serum antibodies against both capsular group W-135 and B strains.
Note that we did not expect to observe superior immunogenicity of the mutant vaccines in wild-type mice, since serum mouse fH does not bind to either the wild-type or mutant fHbp antigens. Testing immunogenicity in wild-type mice is an important screening step for determining whether the epitopes needed for eliciting bactericidal antibodies were preserved in the mutant vaccines. However, because of the relatively small sample sizes in the present study (10 mice per vaccine group), and large variability in antibody responses among individual mice, there was only limited statistical power to detect small losses in immunogenicity that might have resulted from the amino acid substitutions. As a next step, therefore, we intend to retest immunogenicity of the most promising fH-binding mutants in larger numbers of wild-type mice. We also will test immunogenicity in human fH transgenic mice, in which we expect to see increased immunogenicity of the mutants compared with the control fHbp vaccine that binds human fH (7, 8).
In summary, in the present study, we found that amino acid substitutions for arginine at position 41 of fHbp eliminated fH binding in multiple sequence variants from variant group 1 but not in variant group 2 (7). Collectively, the results suggested that antigenic variability among divergent fHbp sequences might be balanced by coevolution of residues affecting fH binding. We also identified three new mutants of fHbp in variant group 2 that eliminated fH binding while preserving epitopes important for eliciting bactericidal antibody. Potentially one or more of these mutants will prove useful for development of the next generation of fHbp-based vaccines with improved immunogenicity and broader vaccine coverage in humans.
ACKNOWLEDGMENTS
This work was supported, in part, by Public Health Service grants R01 AI 46464 (to D.M.G.), R01 AI 82263 (to D.M.G.), and R01 AI 70955 (to P.T.B.) from the National Institute of Allergy and Infectious Diseases, NIH. The work at Children's Hospital Oakland Research Institute was performed in a facility funded by Research Facilities Improvement Program grant number C06 RR 16226 from the National Center for Research Resources, NIH.
David Vu, Serena Giuntini, Andrew Fergus, and Emily Braga provided expert technical assistance.
FOOTNOTES
- Received 31 January 2012.
- Returned for modification 21 March 2012.
- Accepted 10 May 2012.
- Accepted manuscript posted online 21 May 2012.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/IAI.00103-12.
- Copyright © 2012, American Society for Microbiology. All Rights Reserved.
