Functional analysis of monoclonal antibodies against the Plasmodium falciparum PfEMP1-VarO adhesin

89F5 Palo Alto VarO parasites were cultivated as described in O⁺ RBC in RPMI 1640 medium containing l-glutamine and 25 mM HEPES (called hereafter RPMI) and supplemented with 10 % AB⁺ human serum [13]. Rosetting parasites were enriched once a week on ice-cold Ficoll and rosetting rate was kept >90 % by panning on a specific monoclonal antibody as described [13, 30]. Procedures and reagents for monovariant culture, and rosette purification have been published in detail [31].

For inhibition of rosette formation, monovariant VarO cultures were treated with 0.3 M Alanine (in 10 mM Hepes) to eliminate all mature stages. The cultures were then adjusted to 5 % parasitemia, 2 % hematocrit in RPMI, 10 % human AB⁺ serum and 100 µL aliquots (ring stages) were incubated in 96-well plates at 37 ℃ for 24 h with serial dilutions of monoclonal IgG or control reagents (diluted in RPMI, 10 % human AB⁺ serum). After this 24-h cultivation, parasite nuclei were labelled with 10 µg mL⁻¹ Hoechst 33342 (Molecular Probes^® Thermo Fisher Scientific) for 10 min at 37 ℃ and an aliquot of the culture was examined using a fluorescence microscope. The number of iRBCs engaged in rosettes were scored by examining 200 iRBCs. The rosetting rate was calculated as the percent of mature stage iRBCs engaged in rosettes. Details are described in [31].

The rosette disruption assay has been described in detail elsewhere [32, 33]. Briefly, monovariant Palo Alto 89F5 VarO rosette preparations, adjusted to 5 % parasitemia, 5 % hematocrit in RPMI, 10 % human AB⁺ serum were incubated 30 min at 37 ℃ with serial dilutions of monoclonal IgG or control reagents (diluted in RPMI, 10 % human AB⁺ serum). At the end of incubation, mature stages engaged in rosettes or free of bound RBCs were numerated by microscopy (for details see [31, 32]). The rosetting rate was then evaluated as above.

The recombinant domains used to immunize mice and isolate mAbs are shown in Table 1, which indicates their type alongside some sequence features. Cloning of the re-codoned DBL3 and DBL5 domains (which did not elicit surface-reacting antisera) has been described previously [20]. The various DBL1 mutants have been described in [20, 28]; the Genbank accession number of the wild type protein sequence is EU908205. Recombinant proteins were stored in aliquots at −80 ℃ and thawed before use. This, however, caused internal cleavage in some proteins.

MAbs were obtained from OF1 mice immunized with bDBL1, eHead, pCIDR1, bDBL2, and eDBL4 (10 µg/dose injected subcutaneously at 3-week intervals, mixed with complete Freund’s adjuvant for the first dose and incomplete for the boosters) [33]. Mice with high specific IgG titres against the immunogen received an intraperitoneal boost immunization 4 days before being sacrificed for splenic B cell fusion, which was performed as described [34]. For bDBL1, culture supernatants were screened for VarO iRBC surface reactivity by flow cytometry (hybridomas D15-50, D15-68, E20-76). For pCIDR, eHead, bDBL2, and eDBL4, culture supernatants were first screened by ELISA on the recombinant protein and positive culture supernatants were then tested by surface-IFA/flow cytometry against VarO iRBCs. Positive hybridomas were cloned by limiting dilution and screened for ELISA and VarO iRBC surface reactivity. Hybridomas BD20E4 and BDEE10 were custom-made by RD Biotech (Besançon, France), who performed the screening by ELISA on the recombinant protein. Clones were then selected on the basis of VarO iRBC surface reactivity by flow cytometry in the authors’ laboratory. Monoclonal IgG were precipitated with 50 % ammonium sulfate from ascitic fluid, centrifuged and dialyzed against PBS. Monoclonal Ig class and sub-class were determined by ELISA using the isotrip mouse mAb isotyping kit (Roche). Monoclonal IgG were purified using the Melon™ gel monoclonal IgG purification kit (Thermo Fisher Scientific) as recommended by the manufacturers. D15-50, E20-76 and BD20E4 IgG were biotinylated using EZ-link sulfo-NHS-biotin (Thermo Fisher Scientific) and centrifuged through Zeba™ spin desalting columns (Thermo Fisher Scientific) according to the supplier’s recommendations. The concentration of the biotinylated antibodies was evaluated using the Pierce BCA protein assay kit (Thermo Fisher Scientific).

The study was carried out in strict accordance with the recommendations in the guide for the care and use of laboratory animals of the Institut Pasteur and complied with the French [35] and European Union guidelines for the handling of laboratory animals [36]. Animal care and handling was approved by the Ministère de l’Agriculture et de la Pêche (Ref 107503056792, issued to OMP) and the protocols and procedures approved by the Direction Départementale des Services Vétérinaires du Préfet de Police de Paris (clearance number C75-273 issued to OMP). All animal experiments were planned and executed in order to minimize animal suffering.

ELISA plates (Nunc maxisorp) were coated with 100 μL/well of a solution of 0.2 μg mL⁻¹ recombinant protein in PBS, overnight at 4 ℃. Non-specific absorption was blocked with 200 μL blocking buffer (PBS, 5 % w/v non-fat milk) for 1 h at 37 ℃, 100 μL serial dilutions of sera (in PBS, 2.5 % non-fat milk, 0.05 % Tween-20) were incubated 1 h at 37 ℃. Plates were then washed three times with washing buffer (PBS, 0.1 % Tween-20) and incubated for 1 h at 37 ℃ with horseradish peroxidase-conjugated goat anti-mouse IgG (H + L) (Promega) diluted at 1/3000 or anti-rabbit IgG (Promega) diluted 1/3000 in washing buffer. Plates were washed three times before incubation for 5 min at RT with 100 μL TMB/H₂O₂ substrate (KPL). The enzymatic reaction was stopped by addition of 100 μL 1 M H₃PO₄. Absorbance was measured at 450–655 nm. Each serum dilution was tested in duplicate.

For competition ELISA with biotinylated IgG, plates coated with 1 μg mL⁻¹ bDBL1 were incubated with unlabelled monoclonal D15-50, D15-68, E20-76, BD20E4, BDEE10, M21-17 and M21-30 IgG at saturating concentration (50 μg mL⁻¹) for 2 h at 37 ℃. After four washes with PBS, 0.1 % Tween-20, biotinylated D15-50, E20-76 or BD20E4 IgG were added to individual wells at a concentration previously determined to generate a signal of approximately 1 OD after incubation at 4 ℃ for 20 min. After this 20-min incubation and four successive washes with washing buffer, streptavidin-conjugated horseradish peroxidase (Thermo Fisher Scientific) diluted 1/1000 was added to each well and incubated for 60 min at 37 ℃ and after four successive washes with washing buffer, the bound enzyme was revealed as above.

Immunoblots of non-reduced recombinant proteins were prepared as described [33]. Briefly, proteins were fractionated on 4–12 % SDS-PAGE gradient and transferred to a nitrocellulose membrane (Amersham). The membranes were blocked with PBS, 5 % non-fat milk for 1 h at room temperature and probed with diluted sera in PBS, 2.5 % non fat milk, 0.05 % Tween-20. After three washes with PBS, 0.1 % Tween-20, the blot was incubated with anti-mouse IgG (H + L) conjugated to alkaline phosphatase (Promega) diluted 1/10,000 in PBS, 2.5 % non fat milk, 0.05 % Tween-20 for 1 h at room temperature, washed and binding was revealed using Western Blue^® stabilized substrate (Promega) as recommended by the supplier.

Immunoblots of Palo Alto 89F5 VarO crude extracts were prepared as described [37] from rosettes selected on ice-cold Ficoll and enriched on magnetic columns (VarioMacs and CS MACS separation columns, Miltenyi Biotec) to obtain a preparation containing >95 % iRBCs at the mature stage [31]. For SDS-extracts preparations, the highly enriched parasite preparation was first resuspended in 20 volumes of Triton buffer (PBS, 1 % Triton X-100, 2 mM EGTA, complete EDTA-free protease cocktail inhibitors from Roche diagnostics) for 30 min at 4 ℃ and centrifuged at 16,100g at 4 ℃ for 30 min. The pellet was then washed twice with Triton buffer and insoluble proteins were extracted with PBS, 2 % sodium dodecyl sulfate (SDS) supplemented with protease inhibitors for 30 min at RT under a rotation movement and centrifuged at 16,100g for 30 min at 4 ℃. The supernatant was collected and diluted tenfold with PBS, 10 % Triton. The SDS-soluble protein extract was kept at −80 ℃ until use. Immunoblots prepared from reduced or non-reduced SDS-extracts were processed as above.

Antibodies (IgG) reacting with the PfEMP1-VarO protein displayed on iRBC surface were monitored by indirect immunofluorescence using flow cytometry or fluorescence microscopy as described [33]. Immunofluorescence staining was analysed as described [33].

Hybridomas were generated from mice immunized with bDBL1, pCIDR, bDBL2, eDBL4 or the eHead domain (Table 1). Initial screening for surface reactivity identified mAbs E20-76, D15-50 and D15-68 from mice immunized with bDBL1. Screening of subsequent fusions was done stepwise, first by ELISA on the immunizing antigen and then by VarO iRBC surface reactivity. Overall seven anti-bDBL1, two anti-pCIDR, two anti-bDBL2 and one anti-DBL4 mAbs were studied in detail. All anti-bDBL1 mAbs and the anti-pCIDR mAbs were IgG1, while mAb G8-49 was classified as IgG2b and mAbs B12-15, B12-42 and D18-94 were IgG2a (Table 2).

MAbs M21-17 and M21-30 reacted with bDBL1 and eDBL1 by ELISA but not with the Palo Alto 89F5 VarO iRBC surface (Fig. 1a). All other mAbs studied were surface-reactive as well as reacting by ELISA on the cognate domain. Figure 1a shows profiles of surface reactivity assessed by flow cytometry and Fig. 1b the titration curves of individual mAbs by flow cytometry.

All mAbs reacted by ELISA with the cognate domain irrespective of the expression system (Fig. 2a). Immunoblotting showed that the anti-DBL1 mAbs D15-50, E20-76 and BDEE10 reacted with the cognate recombinant antigen and not with the other domains (Fig. 2b). There was very faint cross-reactivity of the anti-CIDR mAb N6-37 with eDBL1; this was not observed with mAb G8-49 (raised to eHead), which presented faint cross-reactivity to eDBL4. MAb B12-42 (raised to bDBL2) showed some faint cross-reactivity with eDBL1 while mAb D18-94 (raised to eDBL4) cross-reacted with pCIDR.

All mAbs reacted on immunoblots of non-reduced total extracts of Palo Alto 89F5 VarO iRBCs but only mAbs BDEE10, G8-49 and D18-94 reacted with the reduced extract (Fig. 3a shows the pattern of reactivity of seven mAbs with the non-reduced total extract, see also Table 2). The surface-displayed PfEMP1 antigen is a type I membrane protein that can be solubilized in 0.5 % SDS. Probing of immunoblots of SDS-extracts showed that mAbs M21-17 and M21-30, which reacted with PfEMP1-VarO on immunoblots of non-reduced total parasite extracts (Fig. 3a, lanes 1 and 2, respectively), failed to react on the SDS-solubilized, membrane-anchored PfEMP1-VarO (Fig. 3b, c, lanes 1 and 2, respectively). This suggests that these mAbs recognize an epitope that is not present on the membrane-anchored PfEMP1-VarO. All surface-reacting mAbs tested reacted with the SDS-extracted PfEMP1-VarO and this reaction was abolished upon reduction of the parasite proteins (Fig. 3b, c). Table 2 summarizes ELISA titres, parasite immunoblot reactivity and surface reactivity of the various mAbs.

Competition ELISAs using individual biotinylated anti-bDBL1 mAbs were used to classify the mAbs. Biotinylation reduced the ELISA titre of each mAb on the cognate antigen by approximately threefold as indicated by the shift of the antigen-dose dependent titration curves (see Additional file 1: Figure S1) but as its impact was similar for the three mAbs, competition ELISAs were carried out. This distributed the mAbs into three non-competing sets: (i) D15-50 and E20-76; (ii) BD20E4 and BDEE10; and (iii) M21-30 and M21-17, although each biotinylated mAb competed best with its non-biotinylated cognate mAb (see Additional file 2: Table S1). As M21-30 and M21-17 did not react with the iRBC surface, they were not studied further with regard to functionality.

To assess the capacity of the mAbs to interfere with VarO rosetting, two functional assays were carried out with monovariant Palo Alto 89F5 VarO cultures, i.e., cultures in which >90 % of the mature stages formed rosettes that reacted with any of the anti-VarO mAbs. Data are summarized in Table 2.

Inhibition of rosette formation was assessed by cultivating synchronized ring stages for 24 h in the presence of a mAb and numerating the rosettes formed in presence of the mAbs at the end of this time period when the parasites reached the trophozoite-early schizont stage. Only mAbs reacting with DBL1 were able to efficiently prevent the formation of rosettes (Fig. 4a). D15-50, E20-76 and BD20E4 had similar rosette inhibition profiles and prevented 95–100 % rosette formation at a concentration of 1 μg mL⁻¹. This high level of inhibition was observed with 10 μg mL⁻¹ polyclonal rabbit IgG raised to bDBL1. D15-68 less efficiently prevented formation of VarO rosettes with low inhibition rates at 1 μg mL⁻¹ and the least potent anti-DBL1 mAb was BDEE10. MAbs against the downstream domains failed to inhibit formation of rosettes at a concentration of 10 μg mL⁻¹. Data are summarized in Table 2.

The capacity to disrupt VarO rosettes, i.e., break the interactions between the iRBC and the uninfected RBC, was evaluated by incubating monovariant cultures with the mAbs and numerating the iRBCs remaining engaged in rosettes after this incubation. This showed that the three mAbs that had similar inhibition profiles differed in their capacity to disrupt rosettes (Fig. 4b). BD20E4 was the most potent disrupter, with 50 % disruption at 0.7 μg mL⁻¹ and 100 % disruption at 10 μg mL⁻¹. MAb D15-50 was 100 % disruptive at 100 μg mL⁻¹ concentration and only 60 % disruptive at 10 μg mL⁻¹. MAb E20-76, on the other hand, was only 60 % disruptive at the highest concentration tested (100 μg mL⁻¹). Surprisingly, BDEE10 displayed a much higher disruption capacity than D15-68. The other mAbs had no or quite marginal rosette-disrupting capacity (Fig. 4b; Table 2).

To investigate epitope specificity of anti-DBL1 mAbs, reactivity of four mAbs (D15-50, E20-76, BD20E4, BDEE10) with a panel of recombinant eDBL1 proteins was analysed (Table 3). Previous work has shown that the eDBL1-s construct lacking the two C-terminal cysteine residues had essentially the same immunoreactivity with polyclonal sera raised to the longer bDBL1 or eDBL1 constructs [33]. This C-terminal deletion had limited impact on RBC binding [20, 28]. It however had a significant impact on binding of BD20E4. BD20E4 titres and reactivity were substantially lower on all constructs lacking 16 residues at the C-terminus and consequently the Cys19–Cys20 disulfide bond (Fig. 5b). Such a reduced reactivity with shorter constructs was not observed with the other mAbs (Fig. 5a, c, d; Table 3).

Reactivity with a set of recombinant proteins carrying specific mutations introduced into the eDBL1-s parent sequence was then analysed. Some of these mutations reduced heparin binding [28], while others abolished RBC binding, localizing the heparin-binding site and the RBC-binding site to opposite sides of the DBL1 domain [28]. Importantly, the various mutations have no major effect on conformation, as indicated by unaltered CD spectra [28, 20]. Although none of the mutations abolished reactivity of the set of mAbs tested, some did reduce binding as shown by reduced ELISA values and titres. The impact of specific mutations on mAb reactivity showed a clear partitioning (Fig. 5). D15-50 and E20-76 had a reduced binding to Mut2 (panels a and c, respectively), whereas binding of BD20E4 and BDEE10 to Mut4 was reduced (panels c and d, respectively). Cleavage of Mut10 by Factor Xa after residue R69 greatly reduced binding of D15-50 and E20-76 (panels e and g, respectively) but had no impact on binding of BD20E4 and BDEE10 (panels f and h, respectively). These data indicate that the mAbs bind to two adjacent DBL1 surface areas involved in RBC binding, since both the mutations harbored by Mut2 and Mut4 as well as Factor Xa cleavage after residue R64 were shown to abolish RBC binding [20]. There was no significant impact of the mutations of the heparin-binding site on mAb binding (Table 3).

Reactivity of the mAbs was further tested on a second panel of mutants derived from the eDBL1-s wild type sequence reported previously [20]. This included in particular specific mutations of the computationally derived ABO blood group binding site, introduced as single mutations (K95A) or double mutations (F145A K216A). None of these had any significant impact on the binding of the four mAbs (Additional file 3: Table S2, Figure S2).

Fine binding specificity differences could be observed within each mAb pair. The reduced binding of BD20E4 with the shorter constructs compared to the longer ones (eDBL1, Mut11, eHead) was not observed for BDEE10 and the impact of some mutations on binding differed between these mAbs (Fig. 5; Table 3). A trend for distinct mutant-specific reactivity was also observed for D15-50 and E2076. Binding of E20-76 to factor Xa-cleaved Mut10 was reduced compared to D15-50 (Fig. 5g, e respectively). Moreover, compared to D15-50, E20-76 displayed reduced binding to Mut11, Mut1, Mut2, Mut4, and Mut5, but enhanced binding to Mut8 (summarized in Table 3). In summary, binding profiles were specific for each of the two pairs identified by competition ELISA, and within each pair, specific for each mAb.

The structures of the DBL1-VarO domain (a DBL1α1.6 type) and the VarO Head protein (double domain DBL1α–CIDRγ) have been determined by X-ray crystallography (PDB entries 2xu0 and 2yk0 respectively) [20, 28]. The DBL1-VarO single domain, which was cleaved after residue R69 during expression and purification of the recombinant protein, did not bind to RBC [28]. The DBL1-VarO domain of the Head construct, by contrast, was produced intact and efficiently bound RBC [20]. Comparison of these two structures shows that the R69 cleavage leads to a significant conformational change in the polypeptide segment comprising the C-terminal region of NTS, the N-terminal region of subdomain 1 and the hinge connecting these two regions (in total, the segment between residues T51 and K95, inclusive) (Fig. 6a, b). Although this region could be completely traced in the electron density of the intact DBL1-VarO domain (PDB entry 2yk0), significant parts of this region were disordered in the cleaved protein and could not be built (PDB entry 2xu0). As this region is adjacent to the blood group binding sites on DBL1-VarO predicted by docking calculations, abrogation of RBC binding in the cleaved domain is most likely due to the conformational change (Fig. 6). Since the inhibitory and rosette-disrupting mAb D15-50 binds to the intact Head construct but poorly to the cleaved DBL1-VarO protein (See Table 3, Mut10fXa), this antibody is affected by the structural changes arising from the R69 cleavage. In addition, point mutations affecting the binding of D15-50 are displayed by Mut2 (K95A, K166A and K179A), which lie to one side of the predicted blood group A and blood group B binding sites, adjacent to the region that undergoes the conformational change induced by the R69 cleavage (Table 3; Fig. 6c, d). By contrast, BD20E4, a very effective rosette-disrupting mAb, binds less efficiently to Mut4, whose mutations (K226A, K227A and K230A) lie on the opposite side of the ABO blood group antigen binding site (Table 3; Fig. 6c, d). This is compatible with the observation that binding by BD20E4 is unaffected by the R69 cleavage and does not compete with the binding of D15-50. Although both Mut2 and Mut4 also carry the mutation K87A, the mutant Mut-K87, with only the K87A amino acid change, is fully recognized by these two mAbs (Table 3) and thus does not contribute to the epitope of either mAb. Of note, neither Mut2 nor Mut4 affected the binding of heparin, whose binding site on DBL1-VarO is predicted to be on the opposite side of the domain [28]. The ELISA results and structural epitope mapping thus present a coherent picture; the two potent rosette-disrupting mAbs, D15-50 and BD20E4, recognize two distinct epitopes adjacent to, but on opposite sides of, the predicted ABO blood group antigens-binding site.