fHbp is an important virulence factor and a key component of vaccines designed for the prevention of serogroup B N. meningitidis
infection. Furthermore, fHbp-based vaccines could provide coverage irrespective of serogroup by either combining it with other antigens or using proteins from different variant families 
. The antigen has been delivered as a recombinant protein in vaccines undergoing Phase II and III clinical trials, but can also be overexpressed in OMV vaccines by genetic modification of strains used for vaccine production 
. Here we characterised members of the three variant families by identifying amino acids that are critical for fH binding and through structural analysis, to inform future vaccine design and to understand the basis of the interaction of fHbp with fH.
The three variant family fHbps we examined all exhibited nM KD
s with fH67
, which is lower than for any human ligand of this important complement regulator. Although fHbps have been found in clinical isolates with reduced affinity for fH 
, none have displayed increased binding. Despite this, we were able to identify several single amino acid substitutions that led to a substantial increase in affinity with fH67
, suggesting that selection may not favour tighter binding, indicating that there could be circumstances when uncoupling of fHbp from fH is beneficial for the bacterium. It is possible that fHbp has other functions 
which are impaired by the presence of fH. Alternatively disengagement from fH could promote colonisation of different sites in the human host, similar to the way modification of pili facilitates disaggregation of bacteria on the surface of cells 
Characterisation of the two Glu residues in V1 fHbp that form salt bridges with fH, and their equivalent residues in V2 and V3 proteins (i.e.
) indicated that these residues make independent contributions to high affinity fH binding and that different variant family proteins engage fH in distinct ways. To identify residues that are necessary for high affinity fH interactions, we performed extensive Ala substitution mutagenesis to produce a catalogue of amino acids in each variant family that contribute to binding to fH, and could be modified in vaccine design. This adds to the three residues already described for V1 fHbp and for V2 fHbp which are required for high affinity binding, although the affinity of the modified V2 proteins for fH was not reported 
. Our findings illustrate differences in the precise mechanisms by which fH engages fHbp from different families, even though the same face of fHbp is involved. This is emphasised by the finding that Ala substitution of amino acids at the same position (i.e.
V1 and V2 Arg106
, and V3 Pro106
) have profoundly different effects, markedly reducing, not affecting, or increasing fH affinity for V1, V2 and V3 fHbps, respectively. This demonstrates that it is not possible to extrapolate data from one variant family protein to others. The dramatic increase in tightness of binding on mutating Pro to Ala at this position in V3 probably results from a kinetic effect, suggesting that in the unbound fHbp the loop containing this residue adopts a different conformation which must be refolded into the conformation seen in the complex. It may be that the Pro converts less readily to the structure required for binding than the loop bearing an Ala at this position Despite these distinctions, all amino acids from V1, V2, and V3 fHbps necessary for high affinity binding are located at the interface previously identified in the V1 fHbp:fH co-complex 
We determined the first structures of the entire V3 fHbp and the C-terminal β barrel of V2 fHbp. There is a striking conservation in the overall structure of the V1 and V3 proteins despite their relatively low level of sequence identity. Although amino acids that contribute to high affinity interactions are grouped in the same regions of these proteins, the precise interactions required to achieve the same affinity and overall interaction with fH differ. Such plasticity could permit the bacterium to alter the fH recognition site for immune evasion whilst retaining the same biological function.
The instability of V2 fHbp and its susceptibility to proteolysis are not desirable in a vaccine antigen, and might explain why it has not been included in any vaccines in clinical trials to date 
. Such instability is less likely to present an issue in the context of the protein on the exterior of bacteria where interactions with surrounding molecules are likely to stabilise the structure, rendering it competent for binding fH; however it might explain why more C-terminal residues appear to be critical for fH binding compared with fHbp from other variant families. Further work is on-going to define the basis of the instability of V2 fHbp, as there is no obvious explanation for this by molecular modelling using the V1 and V3 structures (not shown).
The use of transgenic mice to study human pathogens has been an important advance in infectious diseases research and prevention. For instance, introducing single amino acid changes into murine molecules 
or transgenes encoding complete cellular receptors or nutritional sources 
have allowed the study of human-specific pathogens in rodents. However care must be taken when modifying regulatory factors that govern the activity of complex pathways such as the complement system. We attempted to make minimal changes to mfH within the region that mediates high affinity interactions with fHbp, which would allow binding to the antigen without compromising the important regulatory functions of the molecule. Initial efforts to achieve this by introducing multiple amino acid changes in mfH proved unsuccessful, most likely due to the orientation of CCP 6 with 7 in mfH which would sterically inhibit interactions with fHbp. Therefore, we used a chimeric fH which was humanised through substitution of the CCPs involved in interactions with fHbp together with hfH8
in case it induced unforeseen structural changes in fH7
. This model provides a physiological assay to evaluate non-functional fHbps, rather than simply introducing an intact human transgene, and was employed to examine the immunogenicity of functional and non-functional fHbps.
Overall there were no substantial differences in the immune responses in transgenic and wild-type mice vaccinated with the same protein; both generated similar levels of IgG and SBA responses against the antigen and relevant strain. Previous work suggests that the immunogenicity of fHbpDM
is impaired compared with V1 fHbp 
. However we found that the structure of this protein is unchanged except for the loss of the side chains of Glu283/304
, and that it retained its immunogenicity in both transgenic and non-transgenic mice. We also examined the immunogenicity of fHbpI311A
which we predict reduces the affinity due to the loss of interactions with the bulkier Ile side chain in the Ala mutant. Previous work indicated that V1 fHbpR106S
exhibits a degree of enhanced immunogenicity compared with wild-type fHbp in mice possessing extra copies of hfH as well as endogenous mfH 
. SBA activity was increased by only a single dilution in mice immunised with the non-functional fHbp compared with the wild-type protein, and the effect was only seen in mice with hfH levels above a certain threshold. However we were unable to replicate this finding either with the corresponding protein, V1 fHbpR106A
, or with two other non-functional fHbps, fHbpDM
, and did not observe a relationship between fH levels and SBA titres in individual mice (Fig. S4
). This is unlikely to result from the hydroxyl side chain in Ser in fHbpR106S
compared with fHbpR106A
(used here). Potential explanations for these discrepancies in immunogenicity include differences in antigen and adjuvant preparation, immunisation schedules, and the age of mice and their genetic background (C57/Bl6 here vs
. BALB/c). Furthermore the effects on immunogenicity of the presence of both murine and human fH in a single animal, or an antigen binding hfH (which might not function efficiently in a heterologous environment) are not known.
Any rodent model of immunogenicity has inherent limitations. For instance, both we and others 
immunised mice with 20 µg of fHbp on each occasion. This is relatively a much higher dose than given to infants in current formulations (50 µg) 
, so the proportion of antigen bound by fH might be significantly lower in rodent than in humans. Additionally the route of immunisation (intraperitoneal in rodent models, subcutaneous in clinical trials) will affect delivery to and the site of immune induction, while results from inbred rodent lines will not be directly applicable to human populations. Despite these reservations, ours and other's findings demonstrate that a series of non-functional, structurally defined fHbps elicit at least equivalent responses to V1 fHbp, and provides proof in principle that these antigens merit evaluation in clinical trials which would provide the only definitive evidence of whether they offer advantages as a vaccine compared with wild-type proteins in terms of safety and immunogenicity.
The efficacy of vaccine antigens can be substantially enhanced by structure based studies to generate non-toxic derivatives of bacterial molecules or antigens with increased efficacy 
. Here we show that even though V1, 2 and V3 fHbps exhibit remarkably conserved atomic structures, differences in key amino acids necessary for interactions with fH are only revealed by functional studies. Our findings both provide a catalogue of proteins that could be included in the rational development of the next generation of vaccines containing non-functional fHbps, and could be informative about the basis of the diversity in fHbp sequences seen among clinical isolates, and the genetic susceptibility of individuals to meningococcal disease