Two human antibodies to a meningococcal serogroup B vaccine antigen enhance binding of complement Factor H by stabilizing the Factor H binding site

Microbial pathogens bind host complement regulatory proteins to evade the immune system. The bacterial pathogen Neisseria meningitidis, or meningococcus, binds several complement regulators, including human Factor H (FH). FH binding protein (FHbp) is a component of two licensed meningococcal vaccines and in mice FHbp elicits antibodies that inhibit binding of FH to FHbp, which defeat the bacterial evasion mechanism. However, humans vaccinated with FHbp develop antibodies that enhance binding of FH to the bacteria, which could limit the effectiveness of the vaccines. In the present study, we show that two vaccine-elicited antibody fragments (Fabs) isolated from different human subjects increase binding of complement FH to meningococcal FHbp by ELISA. The two Fabs have different effects on the kinetics of FH binding to immobilized FHbp as measured by surface plasmon resonance. The 1.7- and 2.0-Å resolution X-ray crystal structures of the Fabs in complexes with FHbp illustrate that the two Fabs bind to similar epitopes on the amino-terminal domain of FHbp, adjacent to the FH binding site. Superposition models of ternary complexes of each Fab with FHbp and FH show that there is likely minimal contact between the Fabs and FH. Collectively, the structures reveal that the Fabs enhance binding of FH to FHbp by altering the conformations and mobilities of two loops adjacent to the FH binding site of FHbp. In addition, the 1.5 Å-resolution structure of one of the isolated Fabs defines the structural rearrangements associated with binding to FHbp. The FH-enhancing human Fabs, which are mirrored in the human polyclonal antibody responses, have important implications for tuning the effectiveness of FHbp-based vaccines.

Introduction Like many microbial pathogens, the bacterium Neisseria meningitidis, which is also known as meningococcus, uses surface proteins to recruit soluble complement inhibitors. One of these inhibitors is Factor H (FH), which down-regulates complement activation on the bacterial surface [1,2]. FH binding to the opsonin C3b promotes decay of the alternative pathway (AP) C3 convertase, C3bBb, by competing for the Bb binding site, a process known as decay acceleration activity [3,4]. FH also serves as a cofactor for Factor I to cleave active C3b into inactive iC3b, which prevents the formation of the AP C3 convertase [5,6]. Together, these FH functions effectively block the amplification loop of the AP and limit the generation of C3b, which binds and activates the classical pathway C3 convertase, C4bC2b [7].
Meningococci produce three proteins that function as receptors for human complement FH [8][9][10]. One of these proteins, which had been independently discovered as GNA1870 and LP2086 [11,12], was found to bind FH and renamed Factor H binding protein (FHbp) [8]. The other two FH receptors are Neisserial surface protein A (NspA) and Porin B2 (PorB2) [9,10,13]. Among the three FH receptors, FHbp is generally considered to be the most important for most meningococcal strains. FHbp specifically binds human and non-human primate FH [14,15], but not rabbit or rodent FH [14,16]. FHbp was developed as a key antigen in two licensed meningococcal serogroup B vaccines [17,18].
Other investigators hypothesized that binding of a host protein decreases the immunogenicity of a microbial vaccine antigen [2]. Subsequently, we and our colleagues showed that human FH transgenic mice produced lower protective antibody responses to FHbp and NspA than did wild-type mice [16,[19][20][21]. In the case of FHbp, the higher protective antibody responses of wild-type mice were associated with the ability of the antibodies to inhibit FH binding, and the lower responses of human FH transgenic mice were associated with the antibodies enhancing FH binding [19].
In a subsequent study, we characterized ten human anti-FHbp Fab fragments from three humans vaccinated with one of the licensed vaccines, MenB-4C [22]. By flow cytometry, four of the Fabs enhanced FH binding to the bacteria by factors of three-to six-fold, five Fabs were neutral with respect to FH binding, and one had~2-fold inhibition. In the absence of vaccineelicited antibodies that inhibit FH binding, antibody enhancement of FH binding has the potential to limit vaccine effectiveness by increasing the amount of FH bound to the bacterial surface. Therefore, studies of these antibodies are important for understanding mechanisms of protection by microbial vaccine antigens that interact with host complement proteins. In the present study, we report immunological and biophysical studies of two vaccine-elicited human antibodies that enhance binding of complement FH to the meningococcal virulence factor and vaccine antigen FHbp.

Binding of Fabs to FHbp
We tested commercially produced lots of two human anti-FHbp Fabs and two control mousehuman chimeric Fabs for concentration-dependent binding to purified FHbp by enzymelinked immunosorbent assay (ELISA). The human Fabs were two of the four that enhanced binding of FH to FHbp by flow cytometry in our previous study [22]. The control Fabs were derived from two mouse anti-FHbp monoclonal antibodies (MAbs), JAR 5, which inhibits binding of FH to FHbp, and JAR 4, which does not [8,23]. All four Fabs exhibited sigmoidal binding to the nominal FHbp antigen, ID 1 (http://pubmlst.org/neisseria/fHbp), the sequence variant that is in the licensed MenB-4C vaccine and that was used to raise the mouse MAbs JAR 4 and JAR 5 (Fig 1A). The two human Fabs and the JAR 5 control Fab gave binding curves with midpoints at similar concentrations, whereas the JAR 4 control Fab had approximately 10-fold lower binding to FHbp.
We further validated the newly produced Fabs by testing their cross-reactivity with different FHbp sequence variants. The four Fabs bound to the three FHbp variants with the highest sequence identity to the nominal antigen, ID 1 (100% identity), ID 13 (93% identity) and ID 14 (91% identity) (S1 Fig). In addition, control Fab JAR 4 cross-reacted with FHbp ID 22 (69% identity) and human Fab 7B10 cross-reacted with ID 55 (87% identity). The binding patterns for the control Fabs are consistent with the known locations of the epitopes recognized by these Fabs [24,25], and with the modular architecture of FHbp [26].

Modulation of binding of FH to FHbp by human anti-FHbp Fabs
Next, we tested the ability of the human Fabs to modulate the binding of human complement FH to FHbp by ELISA using a fixed concentration of Fab and variable concentrations of human FH (Fig 1B). Compared to negative control reactions in the absence of Fab (filled circular symbols), the control Fab JAR 4 had no effect on the binding of human FH to FHbp. In contrast, control Fab JAR 5 (asterisks) inhibited FH binding~100-fold. The two human Fabs showed enhancement of FH binding by approximately two-fold for 1A3 (open squares) and ten-fold for 7B10 (open diamonds). In a converse experiment in which the concentration of FH was constant and that of the Fabs was varied, binding of Fab JAR 4 also had no effect and JAR 5 showed almost complete inhibition of FH binding to FHbp (Fig 1C). In this assay, the human Fabs 1A3 and 7B10 enhanced FH binding by~2.8-and 4.5-fold, respectively. Thus, each of the two human Fabs enhanced binding of FH to purified FHbp, similar to our previous measurements of FH binding to live bacteria [22].
To further investigate the kinetic effects, we measured binding of different concentrations of solution-phase FH to immobilized FHbp in the absence or presence of each of the human Fabs by surface plasmon resonance (SPR). An increase in Response Units (RU) was observed for binding of FH to FHbp in the presence of the human Fabs, which was additional evidence that these Fabs did not inhibit FH binding (S2 Fig). Human Fab 7B10 decreased both the association rate constant (k a ) and the dissociation rate constant (k d ), which resulted in a slight increase in the apparent dissociation constant (K D ) (Fig 2 and S1 Table). Human Fab 1A3 also decreased the k a , but had little effect on the k d , and resulted in a~10-fold increase in K D . Therefore, the two Fabs appeared to alter the FH binding kinetics by different mechanisms.

Structural bases for effects of Fab binding on FH-FHbp interaction
To determine the structural bases for the complex effects of the two human Fabs on FH binding to FHbp, we crystallized and determined the structures of the two human Fabs in complexes with FHbp by X-ray crystallography. We refined the structures of the Fab 1A3 and 7B10 complexes with FHbp at high-resolution limits of 1.7 and 2.0 Å, respectively; the data collection and refinement statistics are shown in Table 1. The overall structures, the complementarity determining region (CDR) loops, and representative electron density maps of the two Fab complexes are presented in Fig 3. Despite the differences between the two Fabs in their effects on FH binding and kinetics, and their difference in cross-reactivity with one of the FHbp variants tested, they bound to similar epitopes on the amino-terminal domain of FHbp. Although the epitopes bound by the two Fabs were similar, the buried surface areas between the two Fabs and FHbp were different, 916 Å 2 for the 1A3 complex and 817 Å 2 for the 7B10 complex. Moreover, there were differences in the polar and charged intermolecular interactions between the two paratopes and their respective epitopes (S2 Table). The different interactions corresponded with differences in the amino acid sequences of the CDRs of the heavy and light chains of the two Fabs (S3 Fig). Most notably, there were two salt bridges in the interface of the Fab 7B10-FHbp complex, while there were none in that of the 1A3-FHbp complex. A superposition of the Fab molecules from the two complexes identifies the largest structural differences in CDR-H3 and -L1 (S4A Fig). CDR-H3 has backbone shifts of as much as 3 Å between the two complexes. CDR-L1 has a disordered segment in the Fab 7B10 complex that is ordered in the 1A3 complex; there are also backbone differences of up to 5 Å. To determine whether there were conformational changes associated with Fab binding to FHbp, we crystallized Fab 7B10 alone, and solved its structure to 1.5-Å resolution ( Table 1). An electron density map in the region of CDR-H3 is shown in S5 Fig. A superposition of Fab 7B10 alone with that in its complex with FHbp indicates that the largest rearrangements are in CDR-L1 and -L2 (S4B Fig). In the structure of the FHbp complex, CDR-L1 has five disordered residues, which are ordered in the Fab alone. CDR-L2 shows a backbone shift of up to 3 Å between the two structures.
We also examined the effects of the human Fabs on the conformation and flexibility of loop regions of FHbp using structural superpositions and visualization of crystallographic thermal (B-) factors. Notably, the Fabs affected the conformations and mobilities of two prominent surface-exposed loops compared with the structure of FHbp alone (Fig 4A-4C). Each of the human Fabs displaces these two loops of FHbp away from the Fab and decreases the B-factors of backbone atoms in these loops, which is seen as a decrease in the radius of the tube representations (Fig 4B and 4C). Interestingly, binding of FH to FHbp also affects the conformations and mobilities of these two loops even though FH binds to an adjacent surface of FHbp compared with the human Fabs (Fig 4D).
Finally, to evaluate the structural bases for the effects of the Fabs on FH binding to FHbp, we constructed a superposition model using the two Fab-FHbp complexes and that of a fragment of human FH in a complex with FHbp [27] (Fig 5). These analyses suggest that the human Fabs are likely to be in direct contact with FH, because the closest polar atoms between Fab and FH in the superposition models are 3.7 and 2.4 Å, for 1A3 and 7B10, respectively. Although these are near and within H-bonding distance, the indirect effects of the human Fabs on the conformation of the FH-binding site might be more important for FH enhancement, which we discuss below.

Discussion
Meningococcal FHbp is part of two licensed vaccines for prevention of disease caused by serogroup B strains [17,18], and these vaccines also protect against disease caused by several other serogroups of meningococci [28,29] [42,43]. However, neither of these human Fabs inhibited nor enhanced FH binding. A crystal structure of a third human Fab, 4B3, in a complex with FHbp confirmed that the human antibody repertoire does include bactericidal antibodies that inhibit FH binding [44]. This Fab was derived from one of three MAbs that inhibited FH binding out of 110 human anti-FHbp antibodies tested. Thus, in conjunction with our earlier study [22], it appears that the vast majority of human antibodies to FHbp do not inhibit binding of FH, but those that do can elicit bactericidal activity individually [44].
In our previous study, four of ten human anti-FHbp Fabs enhanced FH binding to the bacterial surface by flow cytometry [22]. In the present study, we investigated two of these Fabs more extensively, and we confirmed their ability to enhance FH binding to purified FHbp in two ELISA formats. In all of these experiments, Fab 7B10 gave higher enhancement of FH binding (5-to 10-fold) than did Fab 1A3 (2-to 3-fold). Although we expected that the enhancement of FH binding would result from a higher affinity of FH for the Fab-FHbp complex compared with that of FH for FHbp alone, we observed slightly lower affinity of FH for the complex by SPR. Both of the human Fabs decreased the on-rate (k a ) for FH binding and Fab 7B10 also decreased the off-rate (k d ). Although we did not measure affinity under equilibrium conditions, it is thought that kinetic SPR experiments yield reliable measurements of affinity [45], particularly at lower coupling densities. One possible explanation for these differences is that the ELISA and flow cytometry experiments measured binding under equilibrium conditions, whereas the SPR experiments measured binding kinetics. Thus, the ELISA and flow cytometry assays might be less sensitive to differences in on-rates but more sensitive to the off-rates, and also because these experiments involve multiple washing steps to remove non-specifically bound antibodies.
Prior to determining the crystal structures of the human Fab complexes, our hypothesis was that the enhancing Fabs stabilized the FH-FHbp complex by direct interactions between Fab and FH. Based on superposition models with the structure of FHbp in a complex with an FH fragment [27], the structures of the Fab-FHbp complexes showed that the Fabs are in proximity to FH and likely are in direct contact. However, judging from the limited contact surface area, the direct interactions are not likely to explain the enhanced FH binding. Rather, it appears that the Fabs lead to structural rearrangements in two loops of FHbp that comprise part of the FH binding site. In addition, based on the crystallographic B-factors, the Fabs also constrain the mobilities of these two loops. The restricted conformational changes potentially explain the lower on-and off-rates for FH binding measured by SPR. More definitive mechanistic details on the enhancement of FH binding could be obtained through structural analyses of ternary complexes of the Fabs with FHbp and FH.
We did not anticipate that the two human Fabs, which differ in many of their properties, would bind to similar epitopes on FHbp. The two Fabs are derived from some of the same germ line immunoglobulin (Ig) genes including heavy and light chain V genes, but not D or J genes (S1 Table in ref. [22]). It is clear that the two Fabs are not clonal based on the fact that they were isolated from different human subjects and they differ in their germline Ig gene usage [22]. In addition, the Fabs contain between two and ten sequence differences in each of the six CDRs (S3 Fig). From an analysis of the polar contacts between Fabs and FHbp using PISA [46], 6 of 10 H-bond interactions are the same in both complexes (S2 Table). Further, five of the six FHbp residues mediating these polar contacts are conserved among the FHbp phylogenic variant groups, which suggests that the Fabs would bind to most or all FHbp variants. Interestingly, the two salt bridges present in the Fab 7B10 complex but not in the Fab 1A3 complex appear to be compensated for by several additional H-bonds in the latter complex, conferring similar affinities of the two Fabs for FHbp (~1 nM) [22].
Although our study is based on only two FH enhancing Fabs, it raises the question of how many different FH enhancing epitopes exist on the FHbp vaccine antigen. One mouse MAb, MAb 502, is known to enhance binding of FH to FHbp, and the epitope that it recognizes appears to be distinct [47,48] from that bound by the human Fabs described here. Structures of complexes of additional FH enhancing human Fabs would elucidate whether there are any other FH enhancing epitopes. If there were only one or several FH enhancing epitopes, our structural data raise the possibility that the FH enhancing epitopes could be eliminated by mutation of one or two of the FHbp residues that mediate conserved interactions with both Fabs. Collectively, the effects of the two anti-FHbp human Fabs on FH binding kinetics and high-resolution structural data provide new information on an important subset of the human antibody repertoire to FHbp vaccine antigens that might provide insight on strategies to redirect the antibody repertoire away from FH enhancing epitopes. Although we do not yet know the clinical significance of the FH enhancing human Fabs, further studies are needed to determine the net contributions of antibodies that enhance or inhibit binding of FH to meningococcal FHbp.

Production of human and chimeric anti-FHbp Fabs
The human Fabs used in this study have GenBank accession numbers of KP770116, KP770115, KP770118 and KP770117 for the heavy and light chains of 1A3 and 7B10, respectively and were characterized previously [22]. In previous studies, we produced human Fabs in E. coli [22,42]. For the present studies that required larger quantities of the Fabs, we had two of the human Fabs expressed in mammalian cells (TunaCHO; LakePharma, Inc.). As controls, we used two chimeric Fabs that comprised mouse variable regions (V H and V L ) derived from mouse MAbs JAR 4 and JAR 5 [23] and human constant regions (C H 1 and κ L ), which enabled detection of the human and control Fabs with the same reagents. The Fabs were purified by anti-C H 1 chromatography and tested for purity by capillary electrophoresis under denaturing conditions, and their concentrations were determined by measuring the UV absorbance at 280 nm and using the molar extinction coefficients calculated from their sequences (https://web. expasy.org/protparam/) [49].

Enyzme linked immunosorbent assay (ELISA)
Purified, recombinant FHbp (2 μg/ml) was immobilized in the wells of an ELISA microtiter plate (Immulon-2B; ThermoFisher) by incubation at 4˚C overnight. The five different FHbp variants used in this study were purified in our laboratory by Ni 2+ -affinity and ion-exchange chromatography as described previously [50]. Non-specific binding was blocked with 5% nonfat dry milk in phosphate-buffered saline (PBS; Roche) containing 0.1% (v/v) Tween-20 (Sigma) for 1 h at room temperature. After washing the plate, serial dilutions of the Fabs or purified human FH were added to the wells and the plate was incubated for 1 h at room temperature. For enhancement experiments, the Fab or FH concentration was fixed at 5 μg/ml (Fig 1B or 1C, respectively). After washing, the bound Fabs were detected with goat antihuman IgG (Fab-specific) conjugated with alkaline phosphatase (AP) (1:10,000; Sigma). Alternatively, bound FH was detected with sheep anti-human FH followed by donkey anti-sheep IgG conjugated with AP (1 h each at room temperature). The colorimetric assay was developed by adding p-nitrophenyl phosphate (1 mg/ml; Sigma) in NaCO 2 , 1 mM MgCl 2 , pH 9.8. The ELISAs were performed in three to four independent assays, and the means of triplicate measurements and 2 SE were calculated in Prism 8.3 (GraphPad).

Surface Plasmon Resonance (SPR)
We performed SPR experiments using FHbp ID 1 (200 to 400 RU) covalently coupled to CM5 biosensor chips and a Biacore X100 Plus instrument (GE Life Sciences). We conducted direct binding experiments using multiple-cycle kinetics and a concentration series of each Fab or purified human FH (Complement Technology, Inc.) from 1 to 100 nM. For experiments to detect enhancement of FH binding by human anti-FHbp Fabs, we injected 10 μM Fab for 120 s to achieve near saturation, then injected a concentration series of 1 to 100 nM of FH. The regeneration solution was 10 mM glycine-HCl, pH 1.7, which was injected for 30 s between protein solutions. The data were analyzed using Biacore Evaluation software with a 1:1 binding model. The data from two to three independent experiments, each performed in duplicate, are reported as means and 2 SE. Probability-(p-) values were obtained from two tailed t-tests, which were calculated in Prism 8.3 (GraphPad).

Crystallization, data collection and structure refinement
We formed equimolar complexes of FHbp with the human Fab 1A3, and separately with 7B10, and purified stable complexes by size-exclusion chromatography (HiLoad Superdex 16/60; GE Life Sciences) in 10 mM Tris-HCl, 25 mM NaCl, pH 7.0. We concentrated the Fab-FHbp complexes to 6 to 8 mg/ml by ultrafiltration (Corning Spin-X UF-6, 30K MWCO). We crystallized the complexes by sparse-matrix crystallization screening (Crystal Screen, PEG Ion and Index; Hampton Research Corp.) in sitting-drop crystallization plates (Art Robbins 96-well or MRC 48-well). We optimized the crystallization conditions in hanging drops in 24-well VDX plates. Crystals of at least 0.02 mm in the smallest dimension were used to collect high-resolution X-ray diffraction data on beamline 8.3.1 at the Advanced Light Source, Lawrence Berkeley National Laboratory. Useful diffraction limits for the 1A3 and 7B10 Fab complexes were 1.7 and 2.0 Å, respectively and for the 7B10 Fab alone was 1.5 Å. Although we were able to obtain small crystals of Fab 1A3 alone, these were not of sufficient quality to obtain useful diffraction data. We solved the three structures by molecular replacement using FHbp ID 1 (3KVD; [51]) and/or homology models based on the sequences of the human Fabs built with Swiss-Model (https://swissmodel.expasy.org) [52]. We solved, refined and validated the structures using the Phenix suite [53] and used Coot [54] for manual model building. Ramachandran and other geometrical outliers were inspected in the electron density maps and were found to be supported by the X-ray data. Structural analyses and generation of Figures were generated with PyMol (The PyMOL Molecular Graphics System, Version 2.4.0, Schrödinger, LLC.). The Protein Data Bank deposition identification numbers, and the data collection and refinement statistics are given in Table 1.