Apicomplexans such as Plasmodium are eukaryotes that have extracellular proteins that are disulfide-linked and glycosylated; therefore, we produced HAP2 proteins in insect and mammalian cells which are competent for such modifications. C-mannosylation and O-glycosylation in Plasmodium have been verified; however, while Plasmodium species have been suggested to have unusually short N-glycans, N-glycosylation has yet to be verified with any specific protein (Bushkin et al., 2010; Macedo de et al., 2010; Bandini et al., 2019; Swearingen et al., 2016; Swearingen et al., 2019). To elicit antibodies, we expressed in Drosophila S2 cells a P. berghei HAP2 D3 construct (residues 477–621, containing four putative N-glycosylation sites) containing a C-terminal His tag. After Ni-affinity chromatography, the material was nearly homogenous, as shown by SDS-PAGE and Coomassie blue staining (Figure 2A, lane 1). Shaving the N-glycans with endoglycosidase D (Endo D) followed by gel filtration decreased size and heterogeneity in SDS-PAGE and confirmed N-glycosylation of D3 (Figure 2A, lane 2).

To produce monoclonal antibodies (mAbs), mice were immunized with the glycan-shaved D3. Antibodies elicited to the shaved protein, which contains one N-glycan residue at each N-glycosylation site, would be expected to react with D3 without being influenced by its glycosylation status. Five IgG mAbs reacted with the immunogen with nanomolar EC50 values in ELISA (Figure 2B). Surface plasmon resonance (SPR) showed that Fabs of two mAbs, 2/6.14 and 2/1.12, bind with affinities of 5–10 nM (Figure 2—figure supplement 1). The antibodies also immunoprecipitated a shorter D3 (502–617) with all putative N-glycosylation sites mutated out (Figure 2C, top panel). Immunoprecipitation was comparable to the His-tag antibody, showing that the epitopes of the 5 mAbs resided within residues 502–617 and were unaffected by N-glycosylation. One antibody (1/5.13) that reacted with the C-terminal His tag and is comparable in specificity and sensitivity to commercial His tag antibodies is described in Materials and methods (Figure 2—figure supplement 2).

HAP2 D3 was also expressed and shaved with Endo D from Plasmodium species that can infect humans: P. knowlesi, P. vivax, P. malariae, P. ovale, and P. falciparum (Figure 2A). As their D3 domains share 60–70% amino acid sequence identity with P. berghei, we tested cross-reactivity with the mAbs to P. berghei D3 using immunoprecipitation (Figure 2C). We quantified immunoprecipitation (Figure 2D) and also measured cross-reactivity using ELISA (Figure 2E and Figure 2—figure supplement 3). mAbs 2/6.14 and 2/1.12 were especially efficient at recognizing D3 from multiple species (Figure 2C). mAb 2/6.14 pulled down about 70%, 30%, and 20% of P. knowlesi, P. ovale, and P. malariae D3, respectively, and mAb 2/1.12 was even more reactive, pulling down 60–100% of D3 from all species except P. falciparum D3 (Figure 2C and D). In contrast, mAb 2/1.40 showed strong reactivity to P. ovale, but not to D3 of other species, while mAbs 2/3.3 and 2/4.36 were highly specific for P. berghei (Figure 2C and D). mAbs 2/6.14 and 2/1.12 bound all species of D3 tested, as shown by ELISA EC50 values in the 1–1000 nM range (Figure 2E and Figure 2—figure supplement 3). Strong binding of P. knowlesi D3 to both 2/6.14 and 2/1.12 was observed with EC50 of <10 nM.

Two individual assays with gametes, immunofluorescent staining with mAbs and ookinete conversion, were carried out in vitro. The change of environment from the bloodstream to the mosquito midgut triggers the development of gametocytes in infected erythrocytes into mature, highly motile ‘male’ microgametes and more sessile ‘female’ macrogametes and gamete emergence from erythrocytes. These events can be mimicked in vitro by reducing the temperature or pH and adding xanthurenic acid to the medium (ookinete medium). Each male gametocyte gives rise to eight microgametes, which look like a number of waving cilia as they emerge from the infected erythrocyte; this process is thus termed ‘exflagellation’.

Reactivity of mAbs to native HAP2 in P. berghei microgametes was determined using cultured, infected erythrocytes undergoing exflagellation. After cultures were fixed, gametes adsorbed to slides were incubated with mouse D3 mAb and rabbit anti-alpha-tubulin followed by staining with fluorochrome-conjugated secondary antibodies specific for mouse and rabbit IgG and DAPI. Anti-tubulin and DAPI were used to identify microgametes by their highly elongated microtubule cytoskeletons and nuclei. All 5 D3 mAbs specifically stained P. berghei microgametes (Figure 3A).

We next tested the ability of the D3 mAbs to inhibit P. berghei microgametes from fertilizing macrogametes and forming ookinetes in vitro. Mouse blood infected with P. berghei gametocytes was mixed with D3 mAbs or control mouse IgG diluted in ookinete medium. After 24 hr, ookinetes and unfertilized female gametocytes were identified by their morphology and staining with a fluorescent antibody, counted, and ookinete conversion rates were calculated. Only a single microgamete is required to fertilize a macrogamete, and each ookinete observed comprises a successful fertilization event. In control IgG, conversion rates were ~80% (Figure 3B). Strikingly, mAb 2/6.14 at 500 µg/ml completely inhibited ookinete conversion and at 250 µg/ml inhibited by 96% (p < 0.001). In contrast, none of the other mAbs inhibited by >50%, although mAb 2/1.40 showed significant inhibition at 500 µg/ml but not at 250 µg/ml (Figure 3B).

We then tested mAb 2/6.14 for its ability to block transmission in vivo in mosquitoes. Female Anopheles stephensi mosquitoes were allowed to feed through membranes on blood from P. berghei infected mice mixed with mAb 2/6.14 or control IgG. After 14 days, mosquito midguts were dissected and oocysts/midgut were counted, that is, oocyst intensity was determined. Oocyst intensity was significantly reduced by mAb 2/6.14, particularly at 250 and 100 µg/ml (Figure 3C). We also measured the prevalence of infection among the fed mosquitoes, i.e. the % of mosquitoes with at least one oocyst. Prevalence was significantly reduced by mAb 2/6.14 at both 250 and 100 µg/ml (Figure 3—figure supplement 1). Summing the results over all mosquitos in the three independent experiments showed significant inhibition of intensity at all three mAb 2/6.14 concentrations, significant inhibition of prevalence at 250 and 100 µg/ml of mAb 2/6.14, dose-dependent reduction of both measures over the three antibody concentrations, and a reduction of intensity by 85% at 250 µg/ml (Figure 3D). Overall, the results show that all five antibodies react with microgametes, that mAb 2/6.14 exhibits potent ability to block P. berghei fertilization in vitro and in vivo, and thus that purified HAP2 D3 has potential as a transmission-blocking vaccine.

To obtain structural insights relevant to rational development of a transmission-blocking vaccine, we obtained crystal structures of the P. berghei HAP2 D3 502–617 fragment with its three N-linked sites mutated out in complex with Fab fragments. Diffraction data was collected and refined to resolutions of 2.8 and 2.1 Å with Rfree of 29% and 23% for complexes with the transmission-blocking Fab 2/6.14 and cross-reactive Fab 2/1.12, respectively (Table 1). Each structure has two independent D3-Fab complexes in the asymmetric unit, giving us four examples of D3. The two Fab fragments bind to opposite faces of D3 (Figure 4A and B).

D3 contains seven β-strands labeled A to E that are arranged into two β-sheets containing the ABE and DCFG β-strands (Figure 4B and C). These β-sheets associate through hydrophobic faces to form a β-sandwich. The way in which the sequence folds into this arrangement classifies it as a fibronectin type III (Fn3) domain. The D3 β-strands connect to one another through loops at each end of the domain. The A-B, C-D, and F-G loops link adjacent β-strands within a sheet and the B-C, D-E, and E-F loops link the two sheets.

Comparisons among the four examples of D3 show markedly different conformations for the loops at the C-terminal end of D3, which abut the polypeptide segment that connects to the plasma membrane (Figure 4C). Quantitation of flexibility in these loops, A-B, C-D, and E-F, shows high root mean square deviation (RMSD) values or lack of comparison because of residues missing in density (dashes, Figure 4F). In contrast, the loops at the N-terminal end that abut D1 show very similar conformations in all independent crystallographic examples. These loops, B-C, D-E, and F-G, show low RMSD values (Figure 4F), consistent with the presence of several backbone-backbone or sidechain-backbone hydrogen bonds that stabilize each loop. Three sidechains that support the hydrogen bond networks in these loops, Asn-531 and Asn-532 in the B-C loop and Asp-598 in the F-G loop (overlined in Figure 4F), are invariant in the 6 Plasmodia species.

The two antibodies bind largely to opposite faces of D3, with 2/1.12 binding primarily to the ABE sheet and 2/6.14 binding primarily to the DCFG sheet (Fab contacts are colored in green and red, respectively in Figure 4B). The overall shapes of the contact surfaces are shown in open book views in which the Fab and D3 are rotated apart like two facing pages in an open book (Figure 4D and E). In each Fab, both the heavy and light chain variable domains contribute significantly to the contact with D3. The sidechains with major contacts ( > 10 Ų of buried solvent accessible surface area or H-bonds) are shown as large colored circles over the alignment in Figure 4F and shown in stick in Figure 4G and H.

Although the 2/1.12 Fab primarily engages the ABE β-sheet, it extends over the edge of this sheet to also contact the D strand, where it overlaps with the 2/6.14 epitope (Figure 4B and F). Fab 2/1.12 has the most contact with four residues in β-strands A, B, and E of D3 (Figure 4G). The sidechains of Thr-508 from D3 β-strand A and His-528 from strand B form hydrogen bonds with the sidechain of Tyr-32 from light chain complementarity-determining region 1 (L-CDR1). In adjacent interactions, Tyr-577 in β-strand E forms a sidechain hydrogen bond to the backbone of Gly-91 in L-CDR3 while the Tyr-577 backbone hydrogen bonds to the sidechain of Asn-104 in H-CDR3. Another close contact and backbone-backbone hydrogen bond between D3 E β-strand residue Ala-575 and H-CDR3 residue Tyr-102 further strengthens interaction (Figure 4G). All four of these major epitope residues are invariant among the species examined here (Figure 4F), correlating with excellent cross-reactivity of mAb 2/1.12.

Although mAb 2/6.14 contacts residues in each β-strand of the D3 DCFG β-sheet, the major contacts are formed by Arg-565 in strand D and six residues in the F-G loop (Figure 4F and H). The F-G loop forms a large bulge on the D3 surface that fits into a cavity in the Fab centered between the three H chain CDR loops (Figure 4E). The Arg-565 sidechain forms bidentate hydrogen bonds to the backbone of Tyr-91 in L-CDR3 and a cation-pi interaction with Tyr-32 in L-CDR1. All three Fab H chain CDR loops surround the D3 F-G loop with each CDR forming at least one of a total of four mainchain-mainchain hydrogen bonds. A network of two sidechain-sidechain and two sidechain-backbone hydrogen bonds forms around FG loop residue Glu-603 and H-CDR2 residue Asn-56. Hydrogen bonds between Lys-597 and Ser-57 of H-CDR2 and salt bridge between Lys-602 and Asp-31 of H-CDR1 further extend the interaction with the F-G loop (Figure 4H).

D3 contains six cysteine residues, all of which are conserved among Plasmodium species; however, only four of these form disulfide bonds (Figures 4F and 5A). Cys-513 in the A-B loop forms a short-range disulfide to Cys-523 in the B strand that likely stabilizes the residues between Cys-513 and the beginning of the B strand, which are highly polar and mostly disordered. Cys-546 in the C strand and Cys-592 in the F strand form a long-range disulfide bond to bridge these two adjacent β-strands in the middle of the DCFG β-sheet. Two D3 cysteines are free. Free cysteines buried in hydrophobic regions are occasionally present in extracellular proteins. Cys-542 is well buried in the hydrophobic core. Cys-604 is exposed in D3-Fab complexes but in the full ectodomain may be buried by residues at the interface between D1 and D3, including residues N-terminal to D3 residue 502, which were not present in the crystallization construct (Figure 5A and B).

HAP2 D3 in P. berghei is only 9% and 18% identical in sequence to C. reinhardtii and A. thaliana HAP2 D3, respectively; however, their three-dimensional structures are highly conserved (Figure 5B). Their seven β-strands superimpose with very low RMSD (Figure 5C). The long-range disulfide bond that connects the C and F strands is also conserved (Figure 5C). Both C. reinhardtii and P. berghei but not A. thaliana have long insertions in their A-B loops and a disulfide bond that may help to stabilize these long loops, which orient quite differently (Figure 5B,C).

There is little variation in the orientation and length of the BC, DE and FG loops at the N-terminal end of D3. These loops are stabilized by an Asn in the BC loop, an Asp in the FG loop, and the hydrogen bonds they make. Asn-531 in P. berghei (asterisked in Figure 5C), forms hydrogen bonds to both the BC and DE loops near the D1 junction (Figure 5D and E). Invariance in the other two species, despite the evolutionary distance between the unicellular algae, the plant, and protozoan parasite, is of particular interest because Asn-531 is in an N-glycosylation NX(S/T) sequon in P. berghei but its equivalents in the other species are not (Figure 5C). The two other species are known to be N-glycosylated but Plasmodium species are not. In D3 constructs mutated to remove N-linked sites, the N was left unchanged because it was much more conserved than the (S/T) position in other Plasmodium species. The B, D, and E strands that connect in these loops all contribute residues important in the mAb 2/1.12 epitope (Figure 5E).

To better understand the P. berghei HAP2 ectodomain and its binding to 2/6.14 Fab, we obtained negative stain EM images and subjected ~2000 particles to iterative alignment and classification. The P. berghei HAP2 ectodomain is rod-like with three or four oval densities arranged linearly along the rod (Figure 4I panel 1 and Figure 4—figure supplement 1). The 2/6.14 Fab - ectodomain complex is L-shaped (Figure 4I panel 2). Similar results were obtained for complexes with ectodomains with residues 43–617 and 61–611 (Figure 4—figure supplement 1). The two globules with stronger density correspond to the Fab, which has two domains per globule. The globules with weaker density correspond to the HAP2 ectodomain. Fab 2/6.14 binds essentially perpendicularly to one end of the HAP2 rod, in agreement with the crystal structure showing that the Fab binds to the side of D3 (Figure 4A). The crystal structure of the 2/6.14 Fab-D3 complex cross-correlated well with the entire ectodomain Fab complex (Figure 4I panel 3), and the ribbon cartoon in the same orientation had its N-terminus facing D1 as expected (Figure 4I panel 4).

Because of their differences in blocking transmission, we wondered whether mAbs differed in reactivity with the HAP2 ectodomain. Yields of monomeric ectodomain were lower than for D3 as described in Methods; nonetheless, gel filtration yielded a sharp peak (Figure 6A). The peak primarily contained a band at 75 kDa, in agreement with the expected size of the ectodomain; furthermore, its C-terminus was present as shown by detection of the C-terminal His-tag by western blotting (Figure 6B). Treatment with Endo D reduced the size of the ectodomain to ~71 kDa (Figure 6B).

Presence of D3 epitopes in the ectodomain was tested by immunoprecipitation. Immunoprecipitation of the ectodomain, and D3 as a control, was detected using western blotting of the His tag (Figure 6C). mAb 2/6.14 was fully reactive with the ectodomain as shown by immunoprecipitation comparable to that with the His-tag antibody. In contrast and surprisingly, the other four mAbs were only partially reactive with the ectodomain, although they reacted with D3 comparably to mAb 2/6.14.

These differences were followed up by further comparisons that focused on mAbs 2/6.14 and 2/1.12. Calculations of affinity and dissociation constants assume fully active material; if only a fraction is active then the apparent affinity is lower. In SPR, only the concentration of the analyte enters into affinity calculations. Therefore, we compared affinities measured for the two Fabs using them either as analyte or immobilized on the chip (Figure 6E and Figure 2—figure supplement 1). For Fab 2/1.12, the ratio of K_D values for D3 (analyte/immobile) was 5.2/5.9 = 0.88 ± 0.09 (Figure 6E). In contrast, these values for the ectodomain were 106/11.7 = 9.1 ± 6.2. Thus, the ectodomain appears to be only 0.88/9.1 = 9% ± 6% active in binding to Fab 2/1.12. Comparable K_D value ratios for Fab 2/6.14 were 6.0/10.2 = 0.59 ± 0.23 for D3 and 22.9/50.5 = 0.45 ± 0.04 for the ectodomain. Thus, the ectodomain appears to be 0.59/0.45 = 130% ± 52% active in binding mAb 2/6.14, within error of the expected value of 100%, while Fab 2/1.12 binding to the ectodomain was far below the expected value.

We also assessed epitope content of the ectodomain by mixing it with equimolar concentrations of Fab and separating the complexes from components in gel filtration. Gel filtration of the mixtures was compared to that of the same concentrations of unmixed components (Figure 6D). With Fab 2/6.14, much of the ectodomain was shifted to higher molecular weight and the Fab fragment was substantially depleted. With Fab 2/1.12 in contrast, only a small portion of the ectodomain was shifted to a higher molecular weight shoulder and there was little or no depletion of the Fab.

We further measured affinity with bio-layer interferometry (BLI) and complex formation with an independent preparation of the ectodomain. mAb IgG were immobilized on anti-mouse Fc capture sensors. HAP2 D3 bound well to both mAbs 2/6.14 and 2/1.12. In contrast, HAP2 ectodomain bound to mAb 2/6.14 but not to mAb 2/1.12 (Figure 6—figure supplement 1). In conclusion, mAb 2/6.14 binds well to both D3 and the ectodomain, whereas mAb 2/1.12 and three other antibodies bind well to the D3 immunogen but bind poorly to or do not recognize the ectodomain.

This study was designed to provide insights for developing a vaccine to block or reduce malaria transmission. This objective was addressed first by generating monoclonal antibodies against the D3 fragment of Plasmodium berghei HAP2. CB6F1 mice were immunized with the glycan-shaved D3 protein. All in vitro characterization of binding properties was carried out after a detailed antibodies screening as described in Methods. We next tested the ability of the D3 mAbs to inhibit P. berghei microgametes from fertilizing macrogametes and forming ookinetes in vitro. The results showed that all five antibodies reacted with microgametes, and that mAb 2/6.14 was outstanding for its ability to block P. berghei fertilization in vitro. Therefore, we decided to test mAb 2/6.14 for its ability to block transmission in vivo in mosquitoes. For in vivo characterization of the ability of mAb 2/6.14 to reduce malaria transmission, female Anopheles stephensi mosquitoes were randomized to group feed through membranes on blood from P. berghei infected mice mixed with mAb 2/6.14 or control IgG. After 14 days, the number of oocyst of infected mosquitoes was counted to evaluate the ability of mAb 2/6.14 to inhibit transmission.

Mouse immunization was conducted in accordance with and was approved by Boston Children’s Hospital Institutional Animal Care and Use Committee (IACUC) under protocol #14-06-2731. Animals were cared in compliance with the U.S. Department of Agriculture (USDA) Animal Welfare Act (AWA) and the Public Health Service (PHS) Policy on Humane Care and Use of Laboratory Animals.

Immunization and Plasmodium berghei infection procedures were performed in accordance with the UK Animals (Scientific Procedures) Act (PPL 70/8788) and were AWERB approved. The Office of Laboratory Animal Welfare Assurance covers all Public Health Service supported activities involving live vertebrates in the United States (no. A5634-01).

P. berghei HAP2 amino acid sequence numbering is from RefSeq accession XP_022713330.1. D3 constructs utilized residues 477–621, 497–617, or 502–617 with a C-terminal His tag; only the 502–617 construct contained N516T, S533N and N539Q mutations to abolish N-glycosylation. HAP2 D3 construct sequence ranges and accessions in other species were P. falciparum, 479–626, XP_001347424.1; P. knowlesi, 492–607, XP_002258781.1; P. vivax, 492–608, SGX77070.1; P. malariae, 498–613, SCN12177.1; and P. ovale, 493–608, SBS88209.1. These D3 constructs were expressed either transiently in Expi293F cells using pD2529-CAG vector (Atum, Newark, CA), or in the case of the P. berghei HAP2 D3 477–621 construct and ectodomain 61–611 and 43–617 constructs, and the P. falciparum D3 479–626 construct in Drosophila Schneider S2 cells, were expressed exactly as described previously (Feng et al., 2018). All constructs were codon-optimized for mammalian cells. UniProt accession Q4YCF6.1, which was used for P. berghei HAP2 ectodomain construct codon optimization, was retrospectively discovered to be deleted for S206, which locates to the βd-strand of D2.2 in the alignment to Chlamydomonas reinhardti HAP2, at the opposite end of the ectodomain from D3. All other eight HAP2 accessions recovered from the nonredundant protein database with NCBI BLAST in 2020, including at least seven distinct P. berghei strains including ANKA, are identical to one another and differ from the Q4YCF6.1 sequence only at residue S206 and the signal sequence. This error has been reported to help@uniprot.org. Our recent experience is that, among hundreds of sequences of human and mouse extracellular proteins of similar length, ~ 10% of UniProt but not RefSeq accessions have similar errors.

For purification, HAP2 fragments in 1 L of culture supernatant were adjusted to 1 mM NiCl₂ in D3 buffer (20 mM Tris, pH 8 and 300 mM NaCl) or ectodomain buffer (20 mM Tris, pH 8.5, 500 mM NaCl) and applied to 10 ml Ni-NTA-Agarose (Qiagen) columns. After washing with 15 mM imidazole in D3 or ectodomain buffer, the protein was eluted with 300 mM imidazole in the same buffers. Fragments were then subjected to gel filtration chromatography using Superdex 75 (GE life Sciences) in D3 or ectodomain buffer, concentrated, and frozen at –80 °C. Purification of P. berghei D3477–621 construct expressed stably in S2 cells was similar, except after the Ni-NTA step it was concentrated (1 mg/ml in 0.2 ml) and shaved with endoglycosidase D (10 µl, 100 unit) (New England BioLabs) for 16 hrs at 4 °C prior to gel filtration. Final yield was 22 mg/L culture supernatant. Yields for transiently expressed D3 constructs were ~5–8 mg/L culture supernatant. Ectodomain yields were ~0.2–0.3 mg/L culture supernatant.

CB6F1 mice (Charles River, Wilmington, MA) were immunized intraperitoneally with 20 μg of N-glycan shaved P. berghei HAP2 D3 (residues 477–621) in PBS and complete Freund’s adjuvant (Sigma) and 3 weeks later with the same material in incomplete Freund’s adjuvant (Sigma). Mice were boosted both intravenously and intraperitoneally 2 weeks later with 20 μg of the same antigen in PBS. Three days later, splenocytes were fused with the murine myeloma P3 × 63Ag8.653 (CRL 1580, ATCC, Rockville, MD) as described (Springer, 1980). Hybridomas supernatants were screened by ELISA in microtiter wells coated with the immunogen. Hybridomas that produced IgG mAbs were subcloned. Five produced antibodies specific for P. berghei HAP2 D3 and are characterized in Results. mAb 1/5.13 was found to be specific to the His-tag. It reacts with proteins with C-terminal His tags, but not with N-terminal His tags. It is fully active in ELISA, western blot and immunoprecipitation and has sensitivity comparable to the THE His Tag Antibody (GenScript) in ELISA and western blot (Figure 2—figure supplement 2).

For antibody purification, hybridoma cells were adapted to and expanded in serum free medium containing 1:2 vol/vol Cell MAB medium (Life Technologies) and HyClone CDM4MAB medium (GE Life Sciences). IgG was purified using protein G affinity chromatography (Invitrogen).

Heavy and light chain V region cDNA sequences of mAbs 2/6.14 and 2/1.12 were determined by Syd Labs (Natick, MA). The mAbs each have γ1 heavy and κ light chains and unique V regions as shown by BLAST searches. A somatic mutation in the 2/6.14 VL domain that introduced an N-glycosylation site into the framework region was reversed with an N74S mutation (mature protein numbering) in recombinantly expressed Fab and IgG. Fab or intact H chains used g-Blocks encoding the murine κ chain secretion signal peptide, VH and γ1 CH1 domain, with or without hinge, CH2 and CH3 domains, followed by a Gly-Ser linker and 6xHis tag, that were assembled in EcoRV-linearized pVRC8400 vector (Barouch et al., 2005) using NEBuilder HiFi DNA reagents and protocol (New England BioLabs). κ light chains used g-Blocks encoding the murine κ chain secretion signal peptide, VL domain, and CL domain that were similarly assembled with SapI-linearized pD2529-CAG vector (Atum, Newark, CA).

Fabs and IgGs were expressed in Expi293F cells co-transfected with H and L chain constructs in 2:1 ratios. Fab fragments were purified from culture supernatant by Ni-NTA affinity purification followed by Superdex 200 gel filtration chromatography. IgG was purified by protein G affinity chromatography. 2/6.14 IgG purified from Expi293F and hybridoma cell supernatants bound to P. berghei HAP2 D3 in ELISA with comparable EC50 values.

96-well ELISA plates (Costar) were coated overnight at 4 °C with 50 µl of purified, His-tagged HAP2 D3 at 5 µg/ml in 50 mM sodium carbonate buffer, pH 9.5 and blocked with 3% BSA for 90 min at 37 °C. 50 µl of diluted hybridoma supernatants or purified mAbs in triplicate were incubated for 2 hrs at 37 °C. After three washes, 50 µl of 1:10,000 diluted HRP-conjugated goat-anti-mouse IgG (H + L) (Abcam) was added. After 1 hr at room temperature and four washes, peroxidase substrate (Life Technologies) was added and after 10 min plates were read at 405 nM on an Emax plate reader (Molecular Devices). As a control, hybridoma medium or mAb dilution buffer (1% BSA in PBS) was substituted in the antibody binding step. For measuring binding of His-tagged D3 to immobilized mAbs, ELISA plates were coated with 5 µg/ml mAb, and HRP-anti-His (Penta-His Ab, Qiagen) was used in the detection step. Titration curve fitting and EC50 measurements used the sigmoidal, 4 parameter logistic equation in GraphPad Prism 7.

Purified HAP D3 (residues 502–617 with N516T, S533N and N539Q mutations) or ectodomain (residues 43–617) fragments or Fabs 2/1.12 or 2/6.14 were either used as analytes or amine immobilized on a CM5 chip in a Biacore 3000 (GE Healthcare) according to the manufacturer’s instructions. For immobilization, protein was diluted to 5 μg/ml in 0.15 M NaCl, 20 mM Hepes pH 7.4, and injected at 10 μL/min. The surface was regenerated with a 10- to 60 s pulse of 18 mM HCl at the end of each cycle to restore resonance units to baseline. Kinetics and affinity analysis were performed with SPR evaluation software version 4.0.1 (GE Healthcare). A 1:1 Langmuir binding model, with or without a conformational change model, was applied for experimental data fitting, and kinetic parameters were fit globally to sensorgrams at different analyte concentrations.

Bio-layer interferometry (Wallner et al., 2013) experiments were performed on a ForteBio Octet RED384 instrument using anti-mouse Fc capture (AMC, 18–5090) sensors. The reaction was measured on a 384-well plate (working volume of 25 µL) in 20 mM Tris-HCl, pH7.5, 150 mM NaCl, 0.01%BSA, 0.02% Tween 20 (Assay buffer). Biosensors were hydrated in assay buffer for 10 min before starting the measurements. Each biosensor was sequentially moved through five wells with different components: (1) Assay buffer for 1 min in baseline equilibration step; (2) 5 µg/ml (33.3 nM) 2/6.14 or 2/1.12 IgG for 3 min for immobilization of the antibodies onto the biosensor; (3) Assay buffer for 3 min for another baseline equilibration; (4) indicated concentrations of PbHAP2 ectodomain (aa:43–617) or PbD3(aa: 477–621) for 10 min for the association phase measurement; and (5) Assay buffer for 10 min for the dissociation phase measurement. Each biosensor has a corresponding assay buffer reference sensor that went through the same five steps. Kinetics and affinity analysis were performed with Octet RED384 Data Analysis 11.0. A 1:1 Langmuir binding model was applied for experimental data fitting, and kinetic parameters were fit globally to different analyte concentrations for each IgG and HAP2 combination, with kon and koff as shared fitting parameters and maximum response (R_max) as individual fitting parameter.

Culture supernatants or purified HAP2 D3 and ectodomain fragments diluted in TBS (25 mM Tris, pH 8, 300 mM NaCl) containing 0.5% BSA were incubated with mAbs or control non-binding IgG overnight at 4 °C. Immunocomplexes were pulled down with protein G beads by incubation at 4 °C for 2 hr with rotation. Beads were washed 3 times with 1 ml TBS and bound proteins eluted in 1 x Laemmli sample buffer containing 5% β-mercaptoethanol and subjected to reducing SDS-polyacrylamide gel electrophoresis. Blots were probed with polyclonal rabbit anti-His (0.4 µg/ml, Cell Signaling), followed by incubation with HRP-conjugated goat-anti-rabbit (GE Healthcare) and chemiluminescence imaging using LAS-4000 system (Fuji Film). Quantitation of protein bands used ImageJ software.

For immunofluorescent staining of microgametes, tail blood (10 μl) from a mouse with high gametocytaemia was mixed with 10 μl of ookinete medium (RPMI1640 containing 25 mM HEPES, 20% FCS, 100 μM xanthurenic acid pH 7.4) to stimulate emergence of male gametes (exflagellation), which was monitored microscopically as described (Sebastian et al., 2012). Exflagellating gametocytes were fixed in solution for 15 min by diluting formalin to 4% paraformaldehyde, then added to poly-L-lysine-coated slides and incubated at 4 °C overnight. For staining, mAbs or control mouse IgG UPC10 (M7769, Sigma Aldrich) (2 µg in 4 µl of 3% BSA/PBS) were added to washed slides; rabbit anti-alpha tubulin (Abcam AB18251) was at 1:1000 dilution. Following washes, Alexa Fluor-488 anti-mouse IgG and Alexa Fluor-594 anti-rabbit IgG secondary antibodies (Molecular Probes) were added at 25 µg/ml before mounting in VectaShield with DAPI (Vector Laboratories). Fluorescence images were obtained using an epifluorescence 100 x objective on a Nikon Eclipse Ti microscope. Image handling used Adobe Photoshop CC.

In vitro ookinete conversion was assayed as described (Angrisano et al., 2017; Blagborough et al., 2013). Briefly, infected mouse blood containing P. berghei (strain ANKA 2.34) female and male gametocytes (20 µl) was mixed with ookinete medium (100 µl) containing HAP2 D3 antibodies or control IgG (500 and 250 µg/ml final concentration) and incubated at 19 °C. After 24 hr, cultures were incubated with Cy3-conjugated Pbs28 mAb 13.1, which stains both ookinetes and unfertilized macrogametes (Winger et al., 1988), for 20 min on ice. Larger, elongated, banana-shaped ookinetes were distinguished from smaller, round macrogametes and counted by fluorescence microscopy. Conversion rates were calculated as % ookinetes /(macrogametes + ookinetes). Inhibition of ookinete conversion was expressed as the percentage reduction in ookinete conversion with respect to the negative control IgG at the same concentration.

In vivo transmission-blocking activity of HAP2 antibodies was assayed using standard membrane feeding assay (SMFA) as described (Blagborough et al., 2013). Briefly, heparinized P. berghei infected blood containing gametocytes was mixed with HAP2 antibodies or negative control IgG. Female Anopheles stephensi (SDA 500 strain) were starved for 24 hr and then fed on the mixtures using membrane feeders ( > 50 mosquitoes per each blood-antibody feed). Twenty-four hr later, unfed mosquitoes were removed. Mosquitoes were maintained on 8% (w/v) fructose, 0.05% (w/v) p-aminobenzoic acid at 19–22°C and 50–80% relative humidity. Day 14 post-feeding, mosquito midguts were dissected, and oocyst numbers per midgut in each mosquito was determined by phase contrast microscopy. Reductions in oocyst intensity (number of oocysts/midgut) and prevalence (number of infected over total mosquitoes fed) in the presence of an HAP2 antibody were calculated with respect to the negative control IgG present at the same concentration in the feeds.

The P. berghei D3 (502–617) construct with N516T, S533N and N539Q mutations to abolish N-glycosylation was used for crystallization. D3 and Fab were mixed in 1:1.3 molar ratios and complexes were isolated by gel filtration. Complexes were crystallized at 20 °C by hanging-drop vapor diffusion with equal volumes of complex and well solution. The Fab 2/1.12 complex (4.5 mg/ml) was crystallized with 0.2 M ammonium sulfate, 25% PEG 3350, 0.1 M Bis-Tris pH5.5. The 2/6.14 Fab complex (7.5 mg/ml) was crystallized with 0.3 M proline, 22% PEG 1500, 0.1 M HEPES pH7.5; crystals were dehydrated by soaking in solutions that had the starting concentrations of components in the protein and reservoir solutions while raising the concentration of PEG 1500% to 31% in 3% steps. Crystals of Fab 2/1.12-D3 and Fab 2/6.14-D3 were cryo-protected with reservoir solution containing 15% glycerol or 15% ethylene glycerol, respectively, in 2 steps of 5% and 10% increase and plunged in liquid nitrogen. Data were collected at 100 K on GM/CA beamline 23-IDB at the Advanced Photon Source (Argonne National Laboratory) and processed with XDS (Kabsch, 2001). Structures were refined with PHENIX, built with Coot (Emsley and Cowtan, 2004) and validated with MolProbity (Davis et al., 2007). Figures were made with PyMol.

The 2/1.12 Fab-D3 complex structure was solved by molecular replacement in the Phenix suite (Adams et al., 2010) using a Fab search model (PDB ID 2A6D). The 2/6.14 Fab-D3 complex structure was solved by molecular replacement using D3 from the 2/1.12 Fab complex, the H chain of PDB ID 1IGC, and the L chain of PDB ID 5AZ2. Each crystal structure has two complexes in the asymmetric unit.

During model building and refinement of the 2/6.14 Fab complex, the Fab constant domains of one complex had good density, but the variable domains and D3 had broad but continuous density that was difficult to trace. In contrast, all domains of the other complex were easily traced. Furthermore, refinement remained stuck. Alternative space groups including those with lower symmetry or use of twin rules provided no improvement. We then realized that in the troublesome D3-Fab complex in the asymmetric unit, two alternative conformations were present for D3, VH, and VL, whereas CH1 and CL had a single conformation. The transition between dual and single conformations occurred at the elbows between VH and CH1 and between VL and CL; that is the two conformations differed in elbow angle. In further refinement, the data cutoff was changed from 2.4 to 2.8 Å, and we largely treated each of the dual conformations of D3, VH1, and VL as rigid bodies based on their structure in the single conformation of the other D3-Fab complex.

The alternate conformations allow the VL domain in chain B and the D3 domain in chain C to pack against alternate conformation symmetry mates in different complexes along adjacent edges of the unit cell. Symmetry mates with the B conformation severely clash, while those with the distinct conformations A and B have good lattice contacts. In contrast, symmetry mates with conformation A are too far apart to provide stabilizing lattice contacts. Thus, clashes in one conformation together with a lack of stabilizing lattice contacts in the other conformation may have driven the formation of a crystal lattice with dual conformations of VL, VH, and D3 in one of two Fab-D3 complexes in the asymmetric unit.

Each P. berghei HAP2 ectodomain construct (residues 43–617 or 61–611), with or without Fab at a molar ratio of 1:1.3, was subjected to Superdex 200 gel filtration in 20 mM Tris-HCl, pH8, 500 mM NaCl. Peak fractions were adsorbed to glow-discharged carbon-coated copper grids, washed with deionized water, and stained with freshly prepared 0.75% uranyl formate. Images were acquired with an FEI Tecnai-T12 transmission electron microscope at 120 kV and a nominal magnification of 52,000×. About two thousand particles were picked interactively and subjected to 2D alignment, classification and averaging using RELION-2.1 (Zivanov et al., 2018). Selected EM class averages were masked and cross-correlated using EMAN2 (Ludtke et al., 1999) with 2D projections generated at 2° intervals from the 2/6.14 Fab-D3 complex crystal structure filtered to 25 Å.

Expi293F cells and Drosophila Schneider S2 cells were used to express proteins. Expi293F cells were purchased from Thermo Fisher Scientific. The S2 cells were obtained from ExpreS2ion Biotechnologies. All cells tested negative for mycoplasma.