Neutralization of Plasmodium falciparum merozoites by antibodies against PfRH5

Abstract

There is intense interest in induction and characterization of strain-transcending neutralizing antibody against antigenically variable human pathogens. We have recently identified the human malaria parasite Plasmodium falciparum reticulocyte-binding protein homologue 5 (PfRH5) as a target of broadly-neutralizing antibodies, but there is little information regarding the functional mechanism(s) of antibody-mediated neutralization. Here, we report that vaccine-induced polyclonal anti-PfRH5 antibodies inhibit the tight attachment of merozoites to erythrocytes, and are capable of blocking the interaction of PfRH5 with its receptor basigin. Furthermore, by developing anti-PfRH5 monoclonal antibodies (mAbs), we provide evidence that i) the ability to block the PfRH5-basigin interaction in vitro is predictive of functional activity, but absence of blockade does not predict absence of functional activity; ii) neutralizing mAbs bind spatially-related epitopes on the folded protein, involving at least two defined regions of the PfRH5 primary sequence; iii) a brief exposure window of PfRH5 is likely to necessitate rapid binding of antibody to neutralize parasites; and iv) intact bivalent IgG contributes to but is not necessary for parasite neutralization. These data provide important insight into the mechanisms of broadly-neutralizing anti-malaria antibodies and further encourage anti-PfRH5 based malaria prevention efforts.

INTRODUCTION

Despite recent progress in malaria control, current estimates of deaths per year from Plasmodium falciparum infection remain unacceptably high (1). The prospects of artemisinin-resistant parasites and pyrethroid-resistant mosquitoes mandate on-going research into novel cost-effective interventions to control P. falciparum. The RTS,S pre-erythrocytic vaccine has shown only modest levels of efficacy in the target infant population in Phase III trials: no other subunit vaccine candidate, however, has achieved a greater level of efficacy, confirming the urgent need for the assessment of novel approaches to malaria vaccination (2).

Vaccines targeting the parasite’s asexual blood-stage have a long history of success in animal models, but have not yet achieved significant efficacy against a primary endpoint in any Phase IIa/b clinical trial (3). One of the central reasons for this is the level of polymorphism in the small group of antigens which have reached field trials. Possible strain-specific efficacy has been reported in Phase IIb clinical trials of P. falciparum apical membrane antigen 1 (PfAMA1) and merozoite surface protein 2 (PfMSP2)-based vaccines (4, 5). Strain-specific efficacy has also been apparent with PfAMA1 and PfMSP1 vaccines in non-human primate models (6, 7). We recently reported that vaccines based upon the antigen reticulocyte-binding protein homologue 5 (PfRH5) were capable of inducing antibodies that neutralized multiple parasite laboratory lines, as well as recently culture-adapted field isolates, in the widely used assay of growth inhibitory activity (GIA) (8-10). Quite unlike previous blood-stage vaccine candidate antigens, PfRH5 does not appear to be a major target of naturally-acquired immunity to P. falciparum (8, 11). Moreover, PfRH5 is highly conserved across parasite lines, with only five non-synonymous SNPs identified at frequencies >5% in at least one geographical region and among 227 sequenced field parasite isolates (9, 10).

Merozoite invasion of erythrocytes is a complex process involving a series of steps, proceeding through initial binding, reorientation, and committed attachment, followed by moving junction (MJ) motility and vacuole formation (12). Among other functions, proteins on the merozoite surface or secreted from the apical organelles mediate binding to host receptors and/or trigger subsequent steps of invasion through poorly defined signal transduction mechanisms (13-15). Such proteins are accessible to antibody which may interfere with these functions and hence inhibit invasion.

The interaction of PfRH5 with the erythrocyte surface protein basigin (BSG/CD147) is essential for merozoite invasion into erythrocytes, and blockade of this interaction by monoclonal antibodies (mAbs) to the BSG host receptor can inhibit invasion (16). Although it is tempting to speculate that the mechanism of action of vaccine-induced anti-PfRH5 antibody may be comparable, previous studies of antibodies against other blood-stage P. falciparum antigens have identified rather more complex and mixed mechanisms of action. For example, there appear to be at least three distinct mechanisms of action of antibody against PfMSP1 (17); whilst a recent study found that two anti-PfAMA1 mAbs act via blockade of the interaction of PfAMA1 with rhoptry neck protein 2 (PfRON2) (18), but additional actions of polyclonal anti-PfAMA1 antibodies are likely (19).

Although we have shown that antibodies against PfRH5 can effectively neutralize parasites, no more detailed description has yet been provided of the mechanism of action of these antibodies. Here, we characterize the effect of polyclonal antibodies against PfRH5, and report the development of a panel of mouse mAbs which we describe in detail, including epitope and kinetic analyses. Although we demonstrate that blockade of the PfRH5-BSG interaction by anti-PfRH5 mAbs in vitro consistently results in merozoite neutralization, we report further results suggesting additional actions may contribute to parasite neutralization by anti-PfRH5 antibody.

METHODS

Vaccines, animals and polyclonal antibody generation

BALB/c mice and New Zealand white rabbits were immunized intramuscularly (i.m.) with human adenovirus serotype 5 (AdHu5) and modified vaccinia virus Ankara (MVA) viral vectors at an 8 to 17 week interval, expressing full-length PfRH5 (3D7 allele); a PfRH5 fragment (amino acids 191 – 359, NSIY … IRYH); a bi-allelic PfAMA1 insert; or no malaria antigen as previously described (8). The latter control immunizations of mice used viruses expressing ovalbumin (20), whereas rabbits were immunized with viruses lacking an antigen insert (8). Sera were collected 14 days after the poxvirus booster immunization.

To raise antisera against PfRH2a/b, PfRH4 and PfRipr for use in the IFA assays, vaccines were generated as follows. AdHu5 and MVA vaccines were generated as previously described encoding a sequence shared between PfRH2a and PfRH2b and previously termed PfRH2A9 (8); or encoding the sequence amino acids 1329-1607, NHIK...NAYY, from PfRH4 from the 3D7 clone (NCBI XM_001351509) – the most C-terminal region prior to the transmembrane domain (21). BALB/c mice were immunized intradermally (i.d.) with 10¹⁰ viral particles (vp) AdHu5 diluted in endotoxin-free PBS (50μL total; 25μL per ear) on day 0 and boosted i.d. with 10⁷ plaque forming units (pfu) MVA encoding the same antigen on day 56. Serum was harvested by cardiac bleed two weeks later. For PfRipr, a recombinant ‘tet repressed’ AdHu5 vaccine was generated as previously described (8), and a recombinant chimpanzee adenovirus serotype 63 (ChAd63) vaccine was generated in the same manner except the final cloning step involved recombination of the antigen into the ChAd63 genome (22). Viruses were produced in Trex293 cells (8) and purified as previously described (23). The transgene in both viruses encoded amino acids 21-1086 of PfRipr from the 3D7 clone (DLIE...SNQN; PlasmoDB PF3D7_0323400). BALB/c mice were immunized i.m. with 10¹⁰ vp AdHu5-PfRipr on day 0 and boosted i.m. with 10¹⁰ vp ChAd63-PfRipr on day 56, according to previously established immunization regimes for use of heterologous adenoviral vaccines (24). Serum was harvested by cardiac bleed two weeks later. All animal work was approved by the University of Oxford Animal Care and Ethical Review Committee (in its review of the application for the UK Home Office Project Licence PPL 30/2414 and PPL 30/2889).

Assays of GIA and merozoite attachment

3D7 and FVO P. falciparum in vitro parasite maintenance, synchronization and assays of GIA (with the exception of those depicted in Figure 4B) were performed as previously reported (9). For the experiment depicted in Figure 4B (invasion inhibition with anti-basigin mAbs), a flow cytometry-based assay was employed, as described previously (16) with the exception that parasites were detected using 1:5,000 SYBR Green I (Invitrogen). SYBR Green I stained samples were excited with a 488 nm UV laser (20 mW) on a BD FACSCalibur flow cytometer (BD Biosciences) and detected with a 530/30 filter. BD FACS Diva software (BD Biosciences) was used to analyse 50,000 events for each sample. FSC and SSC voltages of 423 and 198, respectively, and a threshold of 2,000 on FSC were applied to gate the erythrocyte population. The data collected were further analyzed with FlowJo (Tree Star).

Figure 4. Anti-BSG mAbs which do not block the PfRH5-BSG interaction by AVEXIS are capable of in vitro parasite invasion inhibition.

(A) Anti-BSG mAbs were tested for their ability to inhibit the PfRH5-BSG interaction in the AVEXIS assay. mAbs MEM-M6/6 and TRA-1-85 potently blocked the PfRH5-BSG interaction. mAbs P2C2-1-D11, MEMM6/1, and 8J251 did not. Assays for mAbs MEM-M6/6, TRA-1-85, MEM-M6/1, and 8J251 were performed simultaneously in triplicate; points and error bars indicate median and range of replicate wells. The assay using P2C2-1-D11 was performed separately in singlet. (B) Anti-BSG mAbs were tested at a range of concentrations in a flow cytometric invasion inhibition assay against 3D7 clone parasites. All tested mAbs potently inhibited invasion. Points and error bars indicate median and range of three replicate wells. Data for the BSG-PfRH5 interaction blocking mAbs MEM-M6/6 and TRA-1-85 are in agreement with previously published data (16).

Free merozoite preparation was performed as described elsewhere (12, 25). For merozoite attachment assays, merozoites isolated by filter-disruption of purified E64-treated schizonts were incubated in the presence of 1μM cytochalasin D and 10mg/mL protein-G purified total rabbit IgG for 3 min on a plate shaker, before addition of fresh, uninfected erythrocytes for a further 10 min. The cells were then washed with RPMI 1640 medium, fixed with methanol, and stained with Giemsa. At least 2000 erythrocytes per experiment were counted manually to enumerate the percentage of erythrocytes with attached merozoites.

Indirect immunofluorescence assays (IFA)

For IFA in Figure 1A and Figure S2, 3D7 clone P. falciparum late-stage schizonts and free merozoites were purified by 70% v/v Percoll centrifugation, smeared onto glass slides, dried and fixed in 4% paraformaldehyde for 30 min followed by permeabilization with 0.1% Triton X100 in PBS for 10 min. When treated with E64, this was done as described previously (12, 25). After overnight blocking in 3% BSA in PBS, slides were incubated with primary antibodies at room temperature (RT) for 1 h followed by washing in PBS for 30 min and incubation with secondary Alexa Fluor (488 and 594) labelled antibodies at 1:5000 for 1 h. Nuclei were stained with DAPI (4′,6-diamidino-2-phenylindole) at 0.1 μg/mL for 1 h. Slides were finally mounted in Vectashield and viewed on a Zeiss Axioplan 2 imaging system with Plan Apochromat 100x/1.4 oil immersion objective. Images were captured using Axiovision 4.6.3 software and edited using Adobe Photoshop.

Figure 1. Localization of PfRH5 by indirect IFA.

Localization of PfRH5 was assessed by indirect IFA using anti-PfRH5 polyclonal rabbit serum (green). (A) Fixed and permeabilized schizonts with (+) or without (−) E64 treatment or free merozoites (inset), of 3D7 clone P. falciparum were co-stained with mouse mAbs (red) to mark various organelles: PfAMA1 (microneme), PfRAP1 (rhoptry body), or PfRON4 (rhoptry neck); or polyclonal mouse serum against further antigens: PfRH4 and PfRipr. Figures show the merge of the dual staining antibodies and nuclei stained with DAPI (blue), as well as the brightfield view. Scale bars = 1 μm. (B) Localization of PfRH5 and PfAMA1 was assessed by indirect IFA using antigen-specific polyclonal rabbit serum (green) on the surface of free, fixed merozoites isolated from E64-treated schizonts either permeabilized with Triton X100 (+ TX100) or non-permeabilized ( − TX100). Nuclei were stained with DAPI, and the merge of the images is shown. (C) Localization of PfRH5 and PfAMA1 using antigen-specific polyclonal rabbit serum (green) on the surface of non-fixed non-permeabilized merozoites adherent to RBC in the presence of 1μM cytochalasin D. Scale bars = 1 μm. (D) Localization of PfRH5 was assessed by indirect IFA using PfRH5-specific polyclonal rabbit serum (green) on fixed and permeabilized merozoites adherent to RBC in the presence of 1μM cytochalasin D. Co-staining with mouse antibodies (red) was used to identify PfRON4, PfRipr and PfRAP1. Nuclei were stained with DAPI, and the merge of the images together with the brightfield view is shown. RBCs, though present, are only detected as faint membrane structures due to the loss of hemoglobin during paraformaldehyde fixation and Triton X100 permeabilization.

In Figure 1A anti-PfRAP1 mouse mAb 7H8/50 was used at 1:200 (26); polyclonal rat anti-PfEBA175 region VI (PfEBA175_RVI) was obtained from MR4 and used at 1:500 (data not shown); anti-PfRON4 mouse mAb 24C6 (27) (a kind gift from Jean-Francois Dubremetz) was used at 1:1000; mouse anti-PfAMA1 (a kind gift of C. Collins, NIMR, UK) was used at 1:100 dilution; polyclonal mouse anti-PfRH4, anti-PfRH2a/b (data not shown) and anti-PfRipr were used at 1:100, 1:300 and 1:50 dilution respectively, and were obtained as described above; purified polyclonal anti-PfRH5 rabbit IgG (20mg/mL) was used at 1:8000.

In Figure S2, polyclonal anti-PfAMA1 rabbit serum (28) (a kind gift from Alan Thomas) was used at 1:2000; polyclonal anti-PfRhopH2 rabbit serum (29) was used at 1:4000; polyclonal anti-PfRON12 rabbit serum (30) was used at 1:5000; and anti-PfRH5 mAb 9AD4 was used at 10μg/mL.

For IFA of free fixed merozoites (Figure 1B), these were isolated as described for the attachment assay, washed in RPMI, and then settled onto glass slides, before fixing in 4% paraformaldehyde in PBS. 0.1% Triton X100 in PBS was used to permeabilize the parasites (if stated), before blocking the slides for 90 min at RT with 3% BSA in PBS. Primary antibodies were applied for 1 h at RT and included anti-PfRH5 or anti-PfAMA1 (bi-allelic) purified polyclonal rabbit IgG (20 mg/mL stock solution diluted 1:200 and 1:1000 respectively in blocking solution). After washing in PBS, secondary goat anti-rabbit Alexa Fluor 488 diluted 1:500 in blocking solution was applied for 1 h at RT. Slides were washed again in PBS, then mounted in Vectashield mounting medium with DAPI (Vector Labs) before viewing under a Leica DMI3000 microscope.

For IFA of free non-fixed merozoites (Figure 1C), these were isolated as described for the attachment assay, incubated in the presence of RBC and 1μM cytochalasin D for 10 min at 37°C with shaking. Cells were collected by centrifugation at 1800 x g for 5 min, and then blocked in 3% BSA in PBS for 10 min. Primary antibody diluted in blocking solution (polyclonal rabbit anti-PfRH5 at 1:2000 or polyclonal rabbit anti-PfAMA1 at 1:1000) was added for 15 min at RT before addition of Alexa Fluor 488-labelled secondary antibody (1:5000) and DAPI diluted in blocking solution (0.1 μg/mL) for 30 min at RT. Parasites were collected by centrifugation and washed in 3% BSA in PBS. Imaging of surface labelled parasites was taken immediately as for Figure 1A except no Vectashield was used. Positive (anti-PfMSP2 mAb 113.1 (31)) and negative control antibodies (rabbit polyclonal anti-Pf myosin A tail domain-interacting protein (PfMTIP) (32)) were included in every experiment (data not shown).

For IFA of fixed and permeabilized merozoites (Figure 1D), these were isolated as described for the attachment assay, incubated in the presence of RBC and 1μM cytochalasin D for 10 min at 37°C with shaking. Following attachment the cells were smeared onto glass slides, dried and fixed using 4% paraformaldehyde for 30 min followed by permeablization with 0.1% Triton X100 for 10 min. Slides were blocked overnight in 3% BSA in PBS. Incubation with primary antibodies (polyclonal rabbit anti-PfRH5 at 1:5000; anti-PfRON4 mouse mAb 24C6 at 1:1000; polyclonal mouse anti-PfRipr at 1:50; anti-PfRAP1 mouse mAb 7H8/50 at 1:800) for a duration of 1 h at RT was followed by extensive washes, incubation with AlexaFluor labelled secondary antibodies at 1:5000 and DAPI at 0.1 μg/mL. Slides were mounted in Vectashield and viewed on a Zeiss Axioplan 2 imaging system with Plan Apochromat 100x/1.4 oil immersion objective.

Anti-PfRH5 mouse mAbs

Monoclonal antibodies against PfRH5 were raised by vaccinating BALB/c mice with the AdHu5-MVA vaccines encoding full-length PfRH5 as above, but with an 8 week prime-boost interval. Spleens were harvested three days after the MVA boost. Splenocytes from one mouse were fused with Sp2/0 myeloma cells (ECACC, HPA, UK), and plated in semi-solid methylcellulose-based medium (ClonacellHY, Stemcell Technologies, Canada) in accordance with the manufacturer’s instructions. Two successive rounds of limiting dilution cloning were performed to select hybridomas which secreted IgG capable of binding mammalian-cell expressed recombinant PfRH5 (16) by ELISA (performed as described previously for rabbit anti-PfRH5 ELISA, with the exception that an alkaline-phosphatase-conjugated goat anti-mouse IgG secondary antibody [Sigma] was used) (8).

A second set of mAbs were generated using the proprietary Direct Selection of Hybridomas (DiSH) system (Abeome Inc, Georgia USA) (33). Briefly, splenocytes were fused with Sp2ab cells, and immediately subcultured into nine 25cm² tissue culture flasks. After 7 days, the resulting B cell receptor-overexpressing cells were stained with enzymatically mono-biotinylated PfRH5 produced in HEK293E cells (16), streptavidin-phycoerythrin (PE), and goat-anti-mouse IgG-allophycocyanin (APC) (both from Invitrogen). Cells staining positively with both fluorophores were selected and deposited singly into medium in 96-well plates by a MoFlo cytometer with CyClone module (Beckman Coulter) (Supplemental Figure 1A). Following culture to expand cells and ELISA screening (as above), one ELISA-positive clone originating from each subculture flask was selected for further study.

Selected hybridomas were grown in medium with ultra-low immunoglobulin foetal calf serum (Gibco, UK), using CellLine1000 two-compartment bioreactors for larger preparations (Integra Biosciences, Switzerland). Monoclonal antibody was purified from supernatant by affinity chromatography using protein G (Pierce). IgG isotypes of the selected mAbs were defined by ELISA, with pre-adsorption of purified mAb onto Maxi-Sorp plates (Nunc, UK), followed by detection using biotinylated mouse isotype-specific antibodies (eBioscience, UK) and alkaline-phosphatase conjugated streptavidin (Sigma Aldrich, UK).

Western blotting

Proteins from 1 × 10⁶ late-stage purified schizonts or uninfected erythrocytes solubilized in Laemmli sample buffers (reducing and non-reducing) were separated by SDS-PAGE on precast NuPAGE polyacrylamide gels (Invitrogen). Proteins were transferred to Protran nitrocellulose membrane, blocked overnight in 5% skimmed milk in PBS/Tween-20 and incubated with rabbit anti-PfRH5 (full-length) polyclonal antibody at 1:10,000 or anti-PfRH5 mAb at 1:200 (QA5, 9AD4, 2AC7, RB3) for 1 h. Detection was carried out by incubation with HRP-conjugated secondary antibodies (Bio-Rad Laboratories) followed by visualization with enhanced chemiluminescence Western blotting detection reagents (GE Healthcare).

Assays of PfRH5-BSG interaction

Avidity-based Extracellular Interaction Screen (AVEXIS) was performed using plate-captured biotinylated recombinant PfRH5 and pentameric β-lactamase-conjugated BSG (BSG-S, the short isoform which predominates on erythrocytes), in accordance with a previously published method (16, 34), with the exception that, between capture of the PfRH5 bait protein on the plate and the application of the BSG prey, a sample of serum or mAb diluted in PBS was applied for 1 h, followed by 6 washes with PBS. The assay was also performed in the reverse orientation using plate-captured biotinylated BSG and pentameric β-lactamase-conjugated PfRH5; in this case, antibody was pre-incubated with the PfRH5 for 90 min at RT prior to application to the plate. Where ELISA was performed for comparison with AVEXIS, the method was as described above.

Blockade of the PfRH5-BSG interaction by anti-BSG mAbs was also tested by AVEXIS. Antibodies were obtained from the following sources: P2C2-1-D11 (Advanced Targeting Systems), 8J251 (Lifespan Biosciences), MEM-M6/1 (Abcam), MEM-M6/6 (AbD Serotec) and TRA-1-85 (R&D systems). Briefly, β-lactamase-tagged BSG prey pentamers were incubated for 1 h at RT with serial dilutions of anti-BSG mAbs prior to presentation to mono-biotinylated PfRH5 baits immobilized on a streptavidin-coated plate, and development of the assay as described previously (16).

Mapping of linear mAb epitopes

Biotinylated peptides were synthesized by Mimotopes Pty Ltd, Australia. Peptides were resuspended in DMSO at 50 mg/mL, diluted in PBS to a working concentration of 10 μg/mL, 50 μL of this solution applied to streptavidin-coated plates and ELISA performed as above, using 50 μL samples of mAb at 5 μg/mL. Initial mapping used a set of 20mer peptides based upon the 3D7 PfRH5 sequence (running from amino acid (αα) residue E26 to Q526), each offset by 8 αα from the previous peptide (i.e. overlapping by 12 αα). For each mAb, responses in excess of the mean OD₄₀₅ plus three standard deviations (across all tested peptides) were regarded as positive. Subsequent mapping of minimal linear epitopes for mAbs QA5 and 9AD4 used peptides progressively truncated from each terminus of the 20mer recognized by each mAb. Finally, ELISA was performed using a third set of peptides based upon the minimal linear epitopes but each with one αα mutated to alanine (or glycine, where the original αα was alanine).

Surface plasmon resonance (SPR) assays of mAb epitope overlap

SPR measurements were made using a Biacore T100 machine, using T100 version 2 control and evaluation software (all from GE Lifesciences, Amersham, UK). All experiments were conducted at an analysis temperature of 37°C. Running buffer comprising HBS, 3 mM EDTA, 0.05% Tween-20 (‘HBS-EP+’), was prepared and adjusted to pH 7.4.

Sandwich binding assays to assess overlap of mAb epitopes were performed using a CM5 chip and mouse antibody capture kit (both from GE Lifesciences). The assay configuration is illustrated in Figure S3A; throughout, the first (primary, or capture) mAb applied is referred to as ‘mAb A’, while the secondary mAb is described as ‘mAb B’. All injections were at 5 μL/min. Both flow cells were covalently coated with anti-mouse IgG capture reagent. First, ‘mAb A’ was injected at 40 μg/mL for 150 s. A second injection of 250 μg/mL polyclonal protein-G purified IgG from a vaccine-naïve mouse was injected across both flow cells for 120 s, to block further capture of antibody by the anti-mouse IgG capture reagent. Third, an injection of 10 μg/mL PfRH5 protein was applied to the active flow cell only for 120 s. Fourth, a repeat injection of ‘mAb A’ at 40 μg/mL over both flow cells was applied for 40 s, to confirm whether any further PfRH5-specific binding of this mAb could be detected (no binding would be expected onto monomeric protein captured by the same mAb). Finally, an injection of ‘mAb B’ at 40 μg/mL over both flow cells was applied for 60 s: binding of ‘mAb B’ to the active flow cell in excess of that to the reference cell thus represented PfRH5-specific binding, indicating that the epitopes of ‘mAb A’ and ‘mAb B’ did not overlap. To assess binding inhibition, levels of binding were compared to an estimated level of non-inhibited binding for that antibody pairing, defined as the lesser of either the maximum level of binding of that same secondary antibody to PfRH5 captured by any primary mAb, or the maximum level of binding of any secondary antibody to PfRH5 captured by the same primary mAb. Complete inhibition of binding was defined as the absence of any PfRH5-specific binding, partial inhibition as PfRH5-specific binding of <30% of the non-inhibited level, and non-inhibited binding as ≥30% of the non-inhibited level.

Further experiments to assess mAb epitope overlap using a non-sandwich competition binding assay employed a Biotin CAP chip. The assay configuration is illustrated in Figure S4A. The biotin-CAP reagent supplied with the CAP chip was diluted six-fold in HBS-EP+. In line with the manufacturer’s recommendations, the HBS-EP+ running buffer was modified to reduce non-specific binding by the addition of 1 mg/mL salmon DNA (Sigma) and 2mg/mL carboxymethyl-dextran (Sigma). PfRH5 protein (enzymatically mono-biotinylated at the C-terminus) was produced by transient transfection of HEK293E cells, as previously described (16), and extensively dialyzed against PBS to remove free biotin. All injections were at 5 μL/min. Initially, 170RU of PfRH5 were captured. Secondly, either buffer or primary mAb (‘mAb A’, 20 μg/mL) were applied for 600 s, achieving saturation binding where mAbs were applied. A third and final injection of either buffer or a second mAb (‘mAb B’, also at 20 μg/mL and 5 μL/min) was then applied for 180 s. PfRH5-specific binding of ‘mAb B’ was calculated by double subtraction of the binding of antibody to a reference flow cell coated only with the biotin capture reagent and the change in reference-subtracted binding when the second injection comprised buffer. PfRH5-specific binding of ‘mAb B’ following a preceding primary ‘mAb A’ was then compared to a maximum non-inhibited level of binding (the level of binding of ‘mAb’ B when the preceding injection comprised only buffer).

Measurement of mAb-PfRH5 interaction kinetics

A Biacore T100 machine, software and HBS-EP+ buffer were used, as above. Experiments were conducted at an analysis temperature of 37°C. Single-cycle kinetic measurements of mAb-PfRH5 interactions were performed using a CM5 chip and mouse antibody capture kit (both from GE Lifesciences). Levels of antibody capture were 100-150 RU, consistent with those recommended by the manufacturer for a 150 kDa bivalent ligand and 80 kDa analyte. PfRH5 protein was produced with C-terminal rat CD4d3+4 and His₆ tags and purified as previously described (16). Five-fold increasing concentrations of PfRH5 from 0.1nM to 62.5nM were successively injected over the chip for 240 s each at a flow rate of 30 μL/min, followed by a final 1800 s dissociation phase. Double-referencing was performed by subtraction of responses on the non-antibody-coated flow cell, and responses after injection of buffer. A 1:1 binding model was fitted, with the bulk-shift correction fixed at zero, in accordance with the observed lack of step change in the sensorgrams during injection. Replicate measurements were made in separate experiments, using independently prepared PfRH5 dilution series and antibody-coated chip surfaces.

Multi-cycle kinetic measurements were used to compare the kinetics of PfRH5 interaction with Fab and intact mAb samples used in GIA. These experiments used the biotin CAP chip and HBS-EP+/dextran/DNA buffer described above. 50 RU fresh PfRH5 protein was captured prior to each application of antibody. Three to five serially diluted samples of each antibody or Fab were prepared; binding of each sample was measured during a 120 s association phase and 800 s dissociation phase. A bivalent analyte binding model was fitted globally for intact IgG samples; a 1:1 binding model was fitted globally for Fab samples. Calibration-free concentration analysis (CFCA) of the concentration of PfRH5-binding mAb and Fab in the samples used for the assays of GIA was performed as described elsewhere (9).

Production of Fab fragments

Fab fragments were prepared using immobilized papain for murine IgG2a (mAbs 9AD4 and RB3) and polyclonal rabbit IgG, and immobilized ficin for murine IgG1 (mAbs QA5, 2AC7, and isotype control IgG1 [eBioscience, clone P3.6.2.8.1]), following the manufacturer’s protocols (both from Pierce). Undigested IgG and Fc fragments were removed by binding to protein A agarose (Pierce), followed by buffer-exchange into incomplete P. falciparum culture medium (for GIA) or PBS (for SPR) using Amicon centrifugal concentrators (Millipore). Primary quantification was performed by A₂₈₀ measurement (Nanodrop), with confirmatory CFCA (see above).

RESULTS

PfRH5 is not detectable on the merozoite surface prior to contact with erythrocytes or after moving junction formation

Proteins that mediate interactions with the host erythrocyte are either constitutively located on the merozoite surface or released during invasion in an ordered schedule from apically-located organelles such as the micronemes and rhoptries. Indirect immunofluorescence assays (IFA) have previously located PfRH5 at the apical end of the merozoite, possibly within the rhoptry body (35-37). Some immuno-electron microscopy (IEM) images have also suggested that PfRH5 may be located in the rhoptry body in late schizonts and on the surface of free merozoites (36, 38). PfRH5 has also been shown to co-localize with the PfRH5-interacting protein (PfRipr) in merozoites, as well as with markers of the MJ in merozoites arrested during red blood cell (RBC) invasion (36, 38).

Here we initially assessed PfRH5 localization by IFA. We observed that polyclonal anti-PfRH5 rabbit antibodies (8) do not co-localize with conventional markers of the rhoptry bulb, rhoptry neck or micronemes (rhoptry associated protein 1 [PfRAP1], PfRON4, and PfAMA1 respectively) in permeabilized schizonts or free merozoites of 3D7 clone P. falciparum parasites (Figure 1A). Similar results were observed with an alternative microneme marker – erythrocyte binding antigen 175 kDa (PfEBA175) (data not shown). We noted minimal co-localization in schizonts with PfRH2a/b (data not shown) and also with PfRH4, described as a marker of the rhoptry tip (39), but significant co-localization with PfRipr in merozoites and late-stage schizonts, especially following treatment with the cysteine protease inhibitor E64, which prevents merozoite release by inhibiting schizont rupture (40).

Using the same polyclonal rabbit antibody, we also observed that PfRH5 does not appear to be accessible on the surface of free merozoites isolated from E64-treated schizonts in incomplete culture medium. In contrast to the result with anti-PfAMA1 polyclonal rabbit antibodies, no detectable staining was observed when anti-PfRH5 antibodies were applied to fixed free merozoites without prior permeabilization (Figure 1B). We were also unable to detect PfRH5 in preparations of unfixed and unpermeabilized merozoites isolated from E64-treated schizonts and adherent to RBCs but with invasion arrested by cytochalasin D treatment (Figure 1C). Similar experiments performed using merozoites cytochalasin-arrested during RBC invasion, followed by fixation and permeabilization, showed PfRH5 close to the PfRON4 marker of the MJ (Figure 1D), as previously noted in a similar study (36). PfRH5 also co-localized with PfRipr and, in agreement with a previous report (12), release of PfRAP1 through the MJ was observed.

Anti-PfRH5 polyclonal antibody inhibits tight attachment of merozoites to erythrocytes

Although PfRH5 was not detected on the surface of either free merozoites isolated from in vitro culture or merozoites which had already reoriented and committed to RBC attachment, it has previously been shown that inhibition of the function of PfRH-family members and related proteins can impair merozoite committed attachment and subsequent formation of the MJ. For example, PfRH4-null W2mef P. falciparum parasites form few MJs when in contact with neuraminidase-treated erythrocytes (12), and antibodies to PfRipr are also reported to inhibit attachment of merozoites to erythrocytes (38). These reports suggest that PfRH5 may be released to the merozoite surface at a stage upstream of MJ formation. In order to further clarify the stage of the invasion process inhibited by polyclonal anti-PfRH5 rabbit antibodies, we performed similar inhibition of merozoite attachment assays. These assays demonstrated that anti-PfRH5 purified IgG could markedly reduce attachment of merozoites to erythrocytes (Figure 2A).

Figure 2. Inhibition of merozoite attachment and effect of anti-PfRH5 antibodies on the PfRH5-BSG interaction measured by AVEXIS.

(A) Three independent assays of merozoite attachment to erythrocytes were conducted. Lines link each observation of the percentage of RBCs with attached merozoites (left axis) from a single assay in the presence of 10mg/mL purified IgG pooled from rabbits immunized with a non-malaria antigen (‘Control’), and in the presence of 10mg/mL purified IgG that was pooled from three representative PfRH5-immunized rabbits (‘Anti-PfRH5’). Right-hand column indicates the percentage reduction in attachment induced by anti-PfRH5 IgG (calculated from the same data; right axis). (B,C) An AVEXIS assay was performed with plate-bound PfRH5 protein and soluble pentameric basigin. Prior to application of basigin, wells were incubated with serial dilutions of antisera from (B) mice or (C) rabbits immunized with viral vector vaccines expressing full-length PfRH5, a previously described PfRH5 fragment, PfAMA1, or a non-malaria antigen. Points and error bars indicate median and range of three replicate wells. Upper horizontal dashed line indicates the mean minus three standard deviations (SD) of results in control wells with PfRH5, BSG and substrate but no antibody. Lower dashed line indicates mean plus three SD of results in control wells with PfRH5 and substrate but no BSG. (D) Each sub-panel depicts the effect of an individual mAb tested at a range of concentrations by AVEXIS (x-axis). Two sub-panels are shown for each mAb, depicting the results obtained with different orientations of the bait and prey proteins used in the assay: the left hand sub-panels show the assay when PfRH5 protein was immobilized to the plate prior to application of soluble pentameric BSG, while the right hand sub-panels show data obtained with the reverse orientation. The solid lines (both sub-panels, left-hand y-axis) show the OD₄₈₅ readouts for the AVEXIS assay following pre-incubation of the PfRH5 protein either on the plate (left-hand panels) or in solution (right-hand panels) with a dilution gradient of each mAb. To permit comparison of the effect in the AVEXIS assay with the level of binding of mAb, total IgG ELISA was also carried out using replicate wells with an identical gradient of concentration of each mAb (broken line, left hand sub-panels, right-hand y-axis). Error bars for each line indicate the mean and range of the OD readings for n = 2-3 replicate wells within an experiment.

In vitro blockade of the PfRH5-BSG interaction by polyclonal and monoclonal antibodies to PfRH5

Having shown that anti-PfRH5 antibodies achieved a reduction in attachment of merozoites to erythrocytes, we hypothesized that this activity may be related to the ability of the antibodies to block the interaction of PfRH5 with its receptor BSG (16). The interaction can be detected in vitro using recombinantly-expressed proteins in the AVEXIS (AVidity-based EXtracellular Interaction Screen) assay (34), enabling us to investigate whether anti-PfRH5 polyclonal antibody was capable of blocking this interaction. We found that polyclonal sera from mice (Figure 2B) and rabbits (Figure 2C) vaccinated with the full-length PfRH5 antigen (but not those vaccinated with a previously-used fragment of PfRH5 (8, 36), PfAMA1 nor control sera), were capable of complete blockade of this interaction at dilutions up to 1:300 for mouse and 1:900 for rabbit.

In light of these results, we continued to investigate whether antibodies blocking the PfRH5-BSG interaction were causally responsible for the neutralization of merozoites. We generated two panels of mouse mAbs, totalling eight demonstrably unique mAbs, which were capable of binding PfRH5 as assessed by ELISA (Figure 2D, broken lines). All mAbs gave strong ELISA readings, except for RB3 which may suggest a low-affinity interaction in this case or the absence of the RB3 epitope from some of the plate-bound protein. Five of the eight hybridomas were generated using a proprietary system which enables flow-cytometric sorting of single antigen-specific hybridoma cells (Figure S1A) (33). Basic characteristics of these eight mAbs are summarized in Figure S1B. We initially tested a subset of the mAbs on Western blots of 3D7 P. falciparum schizont lysate (Figure S1C-D). Polyclonal anti-PfRH5 rabbit antibodies, as well as the mouse mAbs QA5, 9AD4 and RB3 all detected the bands at 63kDa (full-length) and 45kDa (processing product), previously described by others (35-37). mAb RB3 detected additional bands at around 28 and 31kDa, although the 28kDa band appeared dominant. A band of similar size has also been previously described (37), although these data were inconsistent with the other two reports and in this published study the 45kDa product was not detected. One of the mAbs, 9AD4, which stained parasites by IFA, was also selected to further explore antigen-localization, in addition to the work previously performed with polyclonal sera. The staining pattern of mAb 9AD4 in IFA was in agreement with that of rabbit polyclonal antibody (Figure S2) – once again PfRH5 did not co-localize with polyclonal antibody markers against the rhoptry neck (PfRON12) (30), rhoptry body (high molecular weight complex rhoptry protein 2, PfRhopH2) or micronemes (PfAMA1).

Having generated and characterized the mouse anti-PfRH5 mAbs, we proceeded to test their ability to block the interaction of PfRH5 with BSG by AVEXIS assay (Figure 2D, solid lines). The assay was performed in two orientations: with plate-captured (immobilized) PfRH5 and soluble pentameric BSG, and vice versa. Pre-incubation of immobilized PfRH5 protein with mAbs QA1, QA5 and 6BF10 blocked the PfRH5-BSG interaction (irrespective of assay reagent orientation), whereas mAbs 4BA7, 8BB10, RB3, 2AC7 and 9AD4 did not inhibit the interaction in either orientation of the assay. Some 2AC7 and 9AD4 preparations did exhibit weak and inconsistent blocking activity at the highest concentrations tested (data not shown). The data shown in Figure 2D are most representative of the outcome of multiple replicate assays.

Relationship of parasite growth inhibition by mAbs to in vitro blockade of the PfRH5-BSG interaction

Having assessed activity of the anti-PfRH5 mAbs in the AVEXIS assay of PfRH5-BSG binding blockade, we next tested the ability of each of the mAbs to neutralize 3D7 clone P. falciparum parasites in a GIA assay. We found that all three mAbs which were capable of blockade of the PfRH5-BSG interaction in the AVEXIS assay (QA1, QA5 and 6BF10) were capable of in vitro parasite neutralization, although with varying levels of potency (achieving 38%, 63% and 30% GIA respectively at 500μg/mL) (Figure 3A). Three of the mAbs which had no effect in the AVEXIS assay (4BA7, 8BB10 and RB3) also neutralized parasites ineffectively – achieving no more than 20% activity in the GIA assay at 500μg/mL. Interestingly, however, two of the mAbs which did not block in the AVEXIS assay were highly potent in the assay of GIA, with EC₅₀ values of 62μg/mL (9AD4) and <15μg/mL (2AC7).

Figure 3. Effects of anti-PfRH5 mAbs in the assay of GIA.

Anti-PfRH5 mouse mAbs were tested in the in vitro assay of GIA at a range of concentrations against (A) 3D7 clone P. falciparum parasites and (B) FVO parasites. Results show the mean of two experiments with triplicate wells. Error bars indicate inter-well SEM.

Of 18 laboratory-adapted parasite lines for which the PfRH5 gene has previously been sequenced, FVO is the most divergent from the 3D7 clone upon which our vaccine antigen was based, with the two sequences differing at four amino acid positions (37). We have previously found that purified polyclonal anti-PfRH5 rabbit IgG raised by the 3D7-based antigen was effective in assays of GIA against FVO parasites (8). However, in light of the possibility that changes to crucially-positioned amino acids may affect individual antibody epitopes, we also proceeded to test the anti-PfRH5 mAbs in assays of GIA against FVO parasites. mAbs which neutralized the 3D7 clone also remained effective against FVO, and in fact tended to be more effective against the heterologous strain (Figure 3B), as we have previously seen with polyclonal anti-PfRH5 rabbit IgG (8).

In light of the above results, and given anti-BSG mAbs that block parasite invasion have also been described (16), we sought to investigate whether blockade of the PfRH5-BSG interaction was necessary when targeting the host receptor rather than the parasite adhesin. In a similar manner to 2AC7 and 9AD4, we observed that anti-BSG mAbs which did not block the interaction between the recombinant proteins in the AVEXIS assay (Figure 4A) could show potent inhibition of invasion (Figure 4B). Together these data confirmed that evidence of mAb inhibition of the PfRH5-BSG interaction by AVEXIS assay was consistently associated with in vitro GIA, but absence of blockade did not predict absence of GIA activity, suggesting the in vitro binding assay may not fully reflect the dynamics of the in vivo invasion process or that a number of mAbs may function by an alternative mechanism.

Mapping of inhibitory mAb epitopes

To further explore the mechanism of action of anti-PfRH5 antibodies, and to shed light upon the regions of PfRH5 required for BSG binding, we attempted to map the binding sites of all eight mAbs using a panel of overlapping 20mer linear peptides. We successfully mapped the epitopes bound by four of the mAbs. Each of these mAbs bound to a distinct linear epitope (Figure 5A) – either a single peptide, or an overlapping pair of peptides. There was no detectable binding of mAbs QA1, 2AC7, 6BF10 or 8BB10 to any peptide (data not shown). RB3 bound two overlapping peptides: E50 (EKDDIKNGKDIKKEIDNDKE) and K58 (KDIKKEIDNDKENIKTNNAK). QA5 bound the peptide Y194 (YHKSSTYGKCIAVDAFIKKI). 4BA7 bound two overlapping peptides: Y242 (YDINNKNDDSYRYDISEEID) and D250 (DSYRYDISEEIDDKSEETDD). 9AD4 bound the peptide Y346 (YNNNFCNTNGIRYHYDEYIH). Of the four peptide-binding mAbs, two had been shown to neutralize parasites in the GIA assay: QA5 and 9AD4. Interestingly, the non-neutralizing mAb 4BA7 bound a peptide in between the linear epitopes of the two neutralizing mAbs, whilst RB3 bound a peptide in the N-terminal region (suggesting the 28 kDa band observed in the Western blot data was probably the N-terminal processing product).

Figure 5. Mapping of linear and minimal epitopes for mAb binding.

(A) Assessment by ELISA of mAb binding to overlapping PfRH5-derived 20mer peptides. Binding to native PfRH5 protein is also shown as a control. Peptides are indicated in this graph by the single-letter code and numerical position of their first αα, e.g. E26 indicates the peptide ENAIKKTKNQENQLTLLPIK, beginning with glutamic acid 26 (the first amino acid residue after PfRH5’s 25 amino acid residue signal peptide). * indicates significant binding (OD₄₀₅ values greater than 3 SD above the mean for an individual mAb). (B) Binding of mAb QA5 to peptides progressively truncated from the N-terminus (white) and the C-terminus (red) of its 20mer-epitope (beginning with Y194: YHKSSTYGKCIAVDAFIKKI). Results are the mean of two replicate assays. Truncation of the N-terminus beyond Y200 abrogated binding; truncation of the C-terminus beyond I213 progressively reduced binding, which was completely abrogated by truncations beyond K211. Binding to a 20mer with the FVO parasite strain sequence (YHKYSTYGKYIAVDAFIKKI; green) was similar to that to the 3D7-clone 20mer Y194 (blue). Other controls shown in blue include a non-recognized 20mer peptide (beginning with H178) and the 9AD4 20mer binding peptide (beginning Y346). (C) Binding of 9AD4 to progressively truncated peptides derived from its 20mer epitope beginning with Y346 (YNNNFCNTNGIRYHYDEYIH), as for (B). (D) Binding of QA5 to peptides with each of the internal amino acids in the sequence YGKCIAVDAFIKK progressively mutated to alanine (or glycine in the case of the two alanines in the native sequence). Results are the mean of two replicate assays. Results obtained with these ‘alanine walk’ peptides are shown in grey; results with peptides from the N- and C-terminal truncation set are shown in white and red respectively; results obtained with the originally identified 20mer peptide are shown in blue. The lettering "YGK(C/Y)IAVDAFIKKI" indicates the inferred linear epitope for QA5, with sizes of lettering proportional to the reduction in binding resulting from mutation of each amino acid and (C/Y) indicating the polymorphism between 3D7 (red) and FVO (green) parasites. (E) Binding of 9AD4 to the series of ‘alanine walk’ peptides within the inferred minimal epitope. Colours as in (D); one of the original 20mer peptides (beginning NGIRY...; bar second from left; blue) was used to infer the role of the initial threonine in the epitope.

We continued to define the minimal linear epitopes bound by the two inhibitory mAbs using series of truncated peptides (Figure 5B,C). Interestingly, each of these minimal epitopes contained one of the four amino acids which differ between the 3D7 and FVO parasite sequences, but the antibodies remained capable of binding to equivalent peptides representing the FVO allele (green bars, Figure 5B,C). In order to evaluate which amino acids within these minimal peptides were particularly important for antibody binding, we assessed the ability of the mAbs to bind a further series of peptides in which each residue was sequentially mutated to alanine (‘alanine walking’; Figure 5D,E), thus defining the minimal epitopes as shown.

The remaining three growth-inhibitory mAbs (QA1, 2AC7 and 6BF10) and the non-growth inhibitory mAb 8BB10 did not bind to any of the tested peptides, suggesting that they recognize conformation-sensitive or discontinuous epitopes. This was consistent with the Western blot data where 2AC7 failed to bind following SDS-PAGE (unlike the other three mAbs tested by Western, which bind linear peptide epitopes). To localize the binding sites of these mAbs relative to the mapped QA5, 4BA7, and 9AD4 epitopes, and to assess the relationship of the QA5 and 9AD4 epitopes relative to each other on the folded antigen, we performed two SPR assays to assess the ability of pairs of mAbs to bind simultaneously to PfRH5 protein.

Initially, a sandwich-configuration assay was used to assess overlap of epitopes bound by the seven mAbs 2AC7, 4BA7, 6BF10, 8BB10, 9AD4, QA1 and QA5 (Figure 6A and Figure S3). For certain mAb pairs, either complete inhibition of binding of the second mAb or substantially reduced binding was evident. These data were used to construct a network of pairs of mAbs which inhibit each other’s binding (Figure 6B). For each of the seven mAbs, at least partial inhibition of binding by at least one other mAb was evident. The non-parasite-neutralizing mAb 4BA7 appeared most loosely linked to this network, with its binding not completely inhibited by any other antibody. Each of the five parasite-neutralizing mAbs (2AC7, 6BF10, 9AD4, QA1 and QA5) was capable of complete inhibition of the binding of at least one of the others from this group. We next used a second SPR assay, similar to a competition ELISA, to confirm the relationship of the epitopes of five mAbs (2AC7, 4BA7, 9AD4, QA1 and QA5) (Figure S4A-G). The resulting network of pairs of mAbs with competitive binding (Figure S4C) closely resembled that obtained with the sandwich-configuration assay, although some minor differences were observed.

Figure 6. Analysis of anti-PfRH5 mAb epitope overlap by sandwich SPR binding assay.

A sandwich SPR-based assay was used to assess overlap of epitopes bound by mAbs 2AC7, 4BA7, 6BF10, 8BB10, 9AD4, QA1 and QA5. (A) Levels of binding (response units) of each mAb when injected as a secondary mAb (‘mAb B’) over PfRH5 captured onto chip by binding to a primary mAb (‘mAb A’). See Figure S3A for illustration of assay configuration, and Figure S3B-H for the data from which these binding levels were calculated. To assess binding competition, levels of binding were compared both to the maximum level of binding of that specific ‘mAb B’ to PfRH5 captured by any ‘mAb A’, and to the maximum level of binding of any ‘mAb B’ to PfRH5 captured by the same ‘mAb A’. Complete competition (red), partial competition (orange) and non-competitive binding (green) were defined as in Methods. (B) Schematic model of the arrangement of epitopes on the PfRH5 surface, with summary of inhibitory binding interactions: solid lines indicate complete competition; dashed lines indicate partial competition; absence of a line indicates non-competitive binding. The blue-outlined area indicates the putative BSG-binding region, as defined by ability of these mAbs to block the PfRH5-BSG interaction in AVEXIS (Figure 2D); the red-outlined area indicates the region susceptible to neutralizing antibodies (Figure 3).

Overall, these results demonstrate that mAbs recognizing at least two distinct regions of the PfRH5 primary sequence are capable of parasite neutralization, but suggest that in the folded PfRH5 protein, the epitopes of parasite-neutralizing mAbs lie together in a region which overlaps with the region involved in the interaction of PfRH5 with basigin.

Kinetic characterization of inhibitory mAbs

Although the fine specificities of the neutralizing mAbs QA1, QA5, 2AC7 and 9AD4 are distinguishable, the competition binding experiment suggested that their epitopes are spatially related. We therefore hypothesized that a factor other than epitope specificity may explain their differing potency in the assay of GIA. Antigen-antibody binding kinetics have previously been shown to be important determinants of antibody potency in neutralization of other pathogens (41). We therefore proceeded to examine the kinetics of the interactions between PfRH5 protein and selected mAbs by SPR (Figure 7A-E).

Figure 7. Kinetics of mAb – RH5 interactions.

(A-D) Binding data from single cycle kinetic measurements of the interactions between mAbs 2AC7, 9AD4, QA1 and QA5 are depicted respectively. Curves represent captured mAb-specific binding of PfRH5 in solution (after subtraction of the PfRH5 binding to the reference cell and refractive index changes due to dummy injections of buffer). In each case, replicate results are shown, obtained on two different days with independently captured mAb chip surfaces (and independently prepared PfRH5 dilution series). Solid lines indicate observed binding; dashed lines indicate model-fitted binding. The maximal level of achievable binding (R_max) varied between replicates; these differences in R_max were proportionate to differences in the quantity of captured mAb, which varied between 130RU and 230RU for different mAbs and on different days. (E) Results of single cycle kinetic measurement of interactions of mAbs with PfRH5. To aid interpretation, equilibrium dissociation constants, K_D, are shown both in molar units, and expressed as the μg/mL concentration of mAb which would be expected to achieve 50% saturation of antigen at equilibrium (assuming monovalent binding; the molar K_D multiplied by 1.5×10⁸). To aid interpretation of the dissociation rate constant, k_d, the interaction half-life in minutes is also shown (−ln[0.5]/[60k_d]). Standard errors (SE) of model-fitted kinetic parameters and chi-squared measures of model goodness-of-fit (<10% of R_max in all cases) are also presented.

The equilibrium dissociation constant, K_D, for an antibody-antigen interaction reflects the concentration of antibody required to bind 50% of available antigen at equilibrium. All four mAbs had K_D values in the range 0.3 to 1.4 nM, such that even the lowest concentrations present during the GIA assay (15μg/mL = 100nM) would be approximately 50-fold in excess of those required to achieve 50% binding at equilibrium. The concentrations of antibody present in the GIA assay would thus be more than sufficient to bind most of the antigen if there was sufficient time available to reach equilibrium. However, the period of extracellular exposure of a merozoite prior to erythrocyte invasion is likely to be brief (approximately 1 minute) (42), and hence it may be more relevant to consider the time taken to reach equilibrium rather than percentage antigen saturation at equilibrium.

Interestingly, association rate constants, k_a, of the most potent antibodies (2AC7 and 9AD4) were relatively high, and substantially higher than the value of 1 × 10⁵ M⁻¹s⁻¹ used to illustrate a previous discussion of the kinetics of merozoite neutralization by antibody (43). The measured k_a values of the two less potent parasite neutralizing mAbs studied (QA1 and QA5) were respectively 2-fold and 4-fold lower than those of 2AC7 and 9AD4. In contrast, the dissociation rate constants for all four mAbs were sufficiently slow to result in interaction half-lives of 10 minutes or more, well in excess of the likely period of extracellular exposure of a viable merozoite.

Overall these data suggest that the on-rate of IgG binding, as well as the epitope specificity, may be associated with merozoite neutralization potency.

Anti-PfRH5 Fabs are capable of merozoite neutralization

The experiments above focussed upon the interactions of PfRH5 with antibody and BSG without addressing the potential contribution to parasite neutralization of the bivalent nature of the IgG molecule and its Fc portion. We therefore aimed to compare the effectiveness of anti-PfRH5 monovalent Fabs and intact IgG in the assay of GIA. Previous studies of Fabs derived from polyclonal and monoclonal antibodies against PfAMA1 have suggested that these neutralize parasites at least as effectively as intact IgG (19, 44). In contrast, here we observed that anti-PfRH5 Fabs consistently had higher GIA EC₅₀ values than intact IgG (approximately 10-fold) when the Fabs were derived from the QA5, 2AC7 and 9AD4 mAbs (Figure 8A-C), but this was not apparent for polyclonal IgG (Figure 8D). At higher concentrations, Fabs proved just as effective as intact IgG for both polyclonal rabbit anti-PfRH5 IgG and the mAbs.

Figure 8. In vitro GIA with anti-PfRH5 Fabs as compared to GIA with intact IgG.

GIA was assessed against 3D7 clone P. falciparum parasites with (A) mAb 2AC7, (B) mAb 9AD4 and (C) mAb QA5, and (D) purified polyclonal rabbit anti-PfRH5 IgG. In each case, GIA with intact IgG is plotted as well as GIA with the respective Fabs. GIA using control purified rabbit IgG and control Fab samples is also shown. A control IgG1 mouse mAb was used for 2AC7 and QA5, whilst RB3 (an IgG2a mAb) was used for 9AD4. The RB3 Fab showed negative GIA within the range of concentration (up to 250μg/mL) in which intact RB3 mAb is GIA-negative (Figure 3). Results show the mean of two assays, each with three replicate wells; error bars plot inter-well SEM. Concentrations are shown as ‘IgG equivalent’. For Fabs, this is the weight/volume concentration of IgG which would have the same molar concentration of antigen-binding sites (i.e. 1.5x the w/v Fab concentration).

Coomassie-stained gels of the Fab preparation were used to demonstrate purity (data not shown). To exclude the possibility that the digestion process had reduced the PfRH5 binding activity of the Fabs, we measured the concentration and affinity of ‘bio-active’ QA5 and 9AD4 Fabs (i.e. the concentration of Fabs capable of binding PfRH5) and the on-rate of the interaction using SPR multi-cycle kinetic and calibration free concentration analyses (9) techniques. The measured values were within 50% of those obtained with comparator samples of intact IgG (data not shown), a difference insufficient to explain the 10-fold discrepancy in the GIA assay mAb results. Neutralization was thus still possible with the Fabs but (especially for the mAbs) tended to be less effective than with intact IgG, suggesting a moderate but non-essential contribution of the Fc to invasion inhibition.

DISCUSSION

PfRH5 is now the leading candidate antigen for inclusion in a next-generation anti-merozoite blood-stage malaria vaccine based on potent cross-strain anti-PfRH5 activity in assays of GIA and the knowledge that its interaction with BSG is essential and universally required for erythrocyte invasion (8, 16). Importantly, anti-PfRH5 antibody-mediated GIA has subsequently been shown to be strongly associated with in vivo protection (Douglas et al., manuscript in preparation). Although we acknowledge the likelihood that other effector mechanisms could contribute to protection induced by other blood-stage antigens, it appears unlikely that PfRH5 – a relatively low abundance protein with brief exposure on the merozoite surface – is a good target for merozoite opsonizing antibody. We have therefore continued, with the work reported here, to focus upon the GIA assay, in the hope that a better understanding of this process, by which anti-PfRH5 antibodies neutralize RBC invasion by P. falciparum parasites, could assist in the design of improved strategies to inhibit erythrocyte invasion by the parasite.

Previous IFA data for PfRH5 have shown a speckled pattern in schizont stages and an apical localization within merozoites (35-37), whilst IEM data have located the antigen to the rhoptries (35, 36, 38). Our IFA data using both polyclonal anti-PfRH5 rabbit IgG as well as the 9AD4 mAb are in agreement with these studies, and also confirm previous reports that PfRH5 does not co-localize with markers of the micronemes, PfRON markers of the rhoptry neck or with the PfRhopH2 and PfRAP1 markers of the rhoptry bulb (35-37). Limited co-localization of PfRH5 with PfRH2a/b has been reported, but supporting IEM data were not conclusive (36). Here we also observed limited co-localization of PfRH5 with other PfRH-family members, PfRH4 or PfRH2a/b, in late-stage schizonts. Although not assessed here, it also seems unlikely from the staining patterns observed that PfRH5 would localize to other secretory organelles such as the “thread-like” mononemes (45) or exonemes that are discharged during egress (46). In contrast, PfRH5 was confirmed to co-localize with PfRipr, especially in very late-stage schizonts following E64 treatment, and this was maintained in merozoites arrested during RBC invasion.

We remained unable to detect PfRH5 on the merozoite surface in any assay, with staining only successful following permeabilization. The results of our merozoite attachment assay nevertheless do suggest an effect of anti-PfRH5 antibody upon the reorientation of merozoites and/or subsequent MJ formation. Indeed, tight coordination of parasite invasin and adhesin release appears to lead to the presence of these proteins on the surface of the merozoite "just in time" prior to or at the time of invasion (14, 39). Our results therefore suggest that PfRH5 has a subcellular location distinct from well-characterized markers, but that the antigen co-localizes with PfRipr towards the MJ during RBC invasion. The majority of PfRH5 is unlikely to be released to the merozoite surface by exposure to the low K⁺ extracellular environment which has previously been shown to result in externalization of the micronemal protein PfEBA175 (15). PfRH5 may therefore only be transiently accessible to antibody after the merozoite contacts the RBC. If this is the case, the time window in which anti-PfRH5 IgG must act to block invasion may be shorter than 30 seconds; indeed, if the antibody is acting between RBC contact and reorientation or MJ formation, this period may be as short as 11 seconds (42). Such a brief period of antigen exposure will place substantial kinetic constraints upon the binding of antibody (43), making it all the more surprising that anti-PfRH5 antibodies appear to be so effective.

We subsequently hypothesized that the invasion inhibitory effects of anti-PfRH5 IgG may be related to the ability of the antibodies to block the interaction of PfRH5 with its receptor BSG. Polyclonal sera against the full-length PfRH5 antigen blocked this interaction in the AVEXIS assay in accordance with its ability to show functional GIA. To address this question further, we raised eight new mouse mAbs against the PfRH5 antigen, and also used a panel of existing anti-BSG mAbs. The novel anti-PfRH5 mAbs included antibodies with GIA EC₅₀ values among the lowest of any anti-merozoite antibody described to date, and considerably lower than those of intensively-studied anti-PfAMA1 mAbs 4G2 and 1F9 (19, 47). Similar to polyclonal sera, those mAbs which showed effective GIA against vaccine-homologous parasites were also capable of neutralizing a vaccine-heterologous parasite.

Consistent with our hypothesis that antibodies capable of blocking the PfRH5-BSG interaction in vitro could neutralize parasites, all anti-PfRH5 and anti-BSG mAbs which blocked the PfRH5-BSG interaction in the AVEXIS assay also had some neutralizing effect against parasites. Three anti-PfRH5 mAbs which had no effect on the interaction also had no neutralizing activity, including mAb RB3 whose binding epitope localized to the N-terminus of PfRH5, which is known to be cleaved leaving the 45kDa product. However, although demonstration of blockade in the AVEXIS assay consistently predicted functional GIA, absence of PfRH5-BSG AVEXIS blockade did not predict absence of GIA: the two anti-PfRH5 mAbs which showed the most potent GIA (9AD4 and 2AC7) did not block effectively in AVEXIS, and three anti-BSG mAbs exhibited GIA in the absence of blocking in AVEXIS.

We have not yet attempted to identify the mechanism of invasion inhibition by mAbs which do not affect the PfRH5-BSG interaction in AVEXIS. It is possible that the in vitro AVEXIS assay does not reflect the full dynamics of the in vivo PfRH5-BSG interaction in some manner, such that a blocking effect of these mAbs does occur in vivo despite being undetectable by AVEXIS. The extremely high parasite-neutralizing potency of 2AC7 and 9AD4 is however intriguing and strongly suggests that these two mAbs are blocking a crucial part of the invasion process. Such antibodies may instead be blocking the interaction of PfRH5 with another binding partner (with PfRipr being a prime candidate) or possibly the processing of the PfRH5 protein. Alternatively, this potency may arise not from a specific block of the interaction of PfRH5 with a partner, but instead simply from steric and cross-linking effects of the binding of antibodies to critical components of the invasion machinery – ‘spanners in the works’ – a result supported to some extent by the loss of potency of 2AC7 and 9AD4 Fabs relative to intact IgG. It is also possible that the mechanisms of action of anti-PfRH5 and anti-BSG antibodies could be quite different; mAbs against BSG may, for instance, act to block invasion by sequestering or reducing mobility of BSG in the erythrocyte membrane. In vitro assays of the ability of vaccine-induced polyclonal sera to block the interaction of parasite adhesins and their receptors have previously been proposed to be useful readouts of functional antibody induction, and hence potential surrogates of in vivo protection (48). The results presented here suggest that at least for PfRH5, given the possibility of multiple mechanisms of antibody action, it is probably most relevant to study the ability of vaccine-induced antibodies to achieve GIA – a functional readout more closely related to the in vivo situation; although, for other parasite species, notably P. vivax (for which in vitro neutralization assays remain challenging) receptor-ligand blockade assays will continue to have an important role.

We also identified minimal linear epitopes recognized by the growth inhibitory mAbs QA5 and 9AD4. Interestingly each of these epitopes contained sequence coded by one of the few known SNPs within PfRH5, suggesting these sequences may be under a degree of immune pressure, although each mAb successfully recognized peptides containing the heterologous epitope. This epitope mapping suggests two regions of the PfRH5 primary structure which are likely to be surface-exposed, and, at least in the case of the QA5 epitope, likely to be closely related to the BSG binding site. The identification of the non-neutralizing 4BA7 mAb binding site between these two neutralizing sites, combined with competition binding results demonstrating overlap between the QA5 and 9AD4 epitopes, suggest that the molecule may fold in a manner which brings the QA5 and 9AD4 epitopes together, with a loop of intervening sequence which is not critical to BSG binding nor a target of neutralizing antibody.

Our observations of the mAb kinetic data also raise the possibility that binding on-rate may be an important parameter in determining neutralization potency. This seems highly plausible, because parasite neutralization is constrained to occur within a brief window of PfRH5 antigen exposure. It has been suggested in studies of humans vaccinated with tetanus toxoid that maximal on-rates can be reached by vaccination, and that this parameter is not further enhanced by booster vaccinations (49). This toxoid study did not take account of the potential effect of adjuvants upon affinity. It therefore remains unclear whether antibody on-rate is a parameter which the vaccinologist can attempt to influence through vaccine design, such as adjuvant selection. This may well be a question worthy of further investigation. However, while mAb on-rates are plausibly related to their neutralization potency, it is arguably less likely that off-rates are relevant. Modern antibody-inducing vaccine platforms are capable of inducing antigen-specific antibody concentrations well in excess of 15μg/mL (50, 51). It seems unlikely that an anti-merozoite vaccine will achieve protection without effective serum IgG concentrations above this level. In the relevant in vivo situation, therefore, antibody concentrations are well in excess of those required to bind the majority of antigen at equilibrium, and the time taken for a merozoite to complete invasion is likely to be far shorter than the dissociation half-life of an antigen-antibody complex which has formed (42). The rate at which merozoite invasion is neutralized is therefore likely determined by the antibody on-rate and concentration, rather than off-rate (43). Although an association between antibody off-rate and risk of disease was recently reported in an immuno-epidemiological study of natural immunity to malaria, such evidence is far from establishing a causal link (52).

The experiments conducted here with anti-PfRH5 Fabs suggested a reduction in the potency of parasite neutralization by Fabs as compared to intact mAb IgG. This contrasts with the effect of digestion of anti-PfAMA1 mAbs upon their neutralizing potency (19, 44), and is suggestive that there is some additional neutralizing action of intact bi-valent anti-PfRH5 IgG beyond simple occlusion of the BSG-binding site on PfRH5. It is possible that mAb-induced impedance of the movement of PfRH5 molecules relative to each other, and relative to other proteins, may be enhanced by cross-linking of PfRH5 molecules and the presence of the bulkier intact IgG molecule. Despite this potency reduction, it was clear that monovalent binding of anti-PfRH5 Fabs from the mAbs can still achieve high levels of GIA and the effect was minor for polyclonal IgG – an observation consistent with the hypothesis that blockade of the interaction of PfRH5 with its binding partners is a major contributor to parasite neutralization.

Overall, our results shed light upon the mechanism of action of anti-PfRH5 antibody. At the level of the whole parasite, it appears that the brevity of the PfRH5 exposure window is likely to necessitate rapid binding of antibody to neutralize parasites. At the level of the PfRH5 molecule, we have demonstrated that antibodies that block the PfRH5-BSG interaction show functional GIA, however, other mechanism(s) may also operate; whilst neutralizing mAbs bind spatially-related epitopes involving at least two defined regions of the PfRH5 primary sequence. These findings offer insights into which approaches may assist design of improved PfRH5 immunogens. The mAbs we have developed and characterized will be valuable tools in future studies of PfRH5 structure, function, and the wider mechanisms of erythrocyte invasion by the P. falciparum merozoite.