Crystal Structure of Plasmodium knowlesi Apical Membrane Antigen 1 and Its Complex with an Invasion-Inhibitory Monoclonal Antibody

The malaria parasite Plasmodium knowlesi, previously associated only with infection of macaques, is now known to infect humans as well and has become a significant public health problem in Southeast Asia. This species should therefore be targeted in vaccine and therapeutic strategies against human malaria. Apical Membrane Antigen 1 (AMA1), which plays a role in Plasmodium merozoite invasion of the erythrocyte, is currently being pursued in human vaccine trials against P. falciparum. Recent vaccine trials in macaques using the P. knowlesi orthologue PkAMA1 have shown that it protects against infection by this parasite species and thus should be developed for human vaccination as well. Here, we present the crystal structure of Domains 1 and 2 of the PkAMA1 ectodomain, and of its complex with the invasion-inhibitory monoclonal antibody R31C2. The Domain 2 (D2) loop, which is displaced upon binding the Rhoptry Neck Protein 2 (RON2) receptor, makes significant contacts with the antibody. R31C2 inhibits binding of the Rhoptry Neck Protein 2 (RON2) receptor by steric blocking of the hydrophobic groove and by preventing the displacement of the D2 loop which is essential for exposing the complete binding site on AMA1. R31C2 recognizes a non-polymorphic epitope and should thus be cross-strain reactive. PkAMA1 is much less polymorphic than the P. falciparum and P. vivax orthologues. Unlike these two latter species, there are no polymorphic sites close to the RON2-binding site of PkAMA1, suggesting that P. knowlesi has not developed a mechanism of immune escape from the host’s humoral response to AMA1.

Introduction steps in invasion-were observed. Thus AMA1 comes into play downstream from the initial, reversible attachment of the merozoite to the RBC, consistent with the currently accepted model where the AMA1-RON complex is a key component of the tight junction [24,[36][37][38].
We have recently tested PkAMA1 as a vaccine in monkeys [39]. These experiments showed that repeated immunization with PkAMA1 controlled parasitemia in five out of six monkeys, with the sixth monkey showing a significant delay in the onset of the parasitemia. On the basis of these data, we believe that PkAMA1 is a good vaccine candidate for P. knowlesi that could be specifically used in Southeast Asia. Here we report the crystal structure of Domains 1 and 2 of the ectoplasmic region of PkAMA1 and of its complex with the Fab fragment of R31C2 which shows that the mAb binds to the hydrophobic groove. In addition, the D2 loop (residues Pro295 to Ser332), which is displaced upon binding the RON2 receptor, remains fixed in the same conformation as the unbound PkAMA1 even though the antibody makes a significant number of direct contacts with this region of the antigen. We also examine the polymorphism of PkAMA1 in the light of these structural data.

Methods and Materials
Cloning and expression of PkAMA1 The recombinant PkAMA1 construct comprises residues Pro43 to Pro387 from the PkAMA1 sequence of the P. knowlesi H strain (Domains 1 and 2; residue numbering from the first residue of the signal sequence, GeneBank accession no. XM_002259303) and an additional 23 Cterminal residues that include the c-myc tag and a hexa-His tag. The synthetic gene was adapted to P. pastoris codon usage and four potential N-glycosylation sites were mutated as follows: Asn107Lys, Ser178Asn, Asn189Glu and Ser240Arg. (Residues Lys107 and Arg240, Asn178, and Glu189 are present in the P. chabaudi, P. falciparum and P. vivax sequences, respectively). The recombinant gene was maintained in frame with the c-myc-hexahistidine tag by the addition of two bases to the restriction site (XbaI); the sequence at the end of the gene sequence thus codes for an additional glycine [39].
The culture protocol was similar to that reported for the P. vivax AMA1 construction [40], except for the addition of AEBSF (a serine protease inhibitor, 4-(2-Aminoethyl)benzenesulfonylfluoride hydrochloride) during the induction with methanol in BMMY' at 0.1 mM. A protease inhibitor was also present (0.4 mM) after recovery of the supernatant. The protein was purified by a metallo-affinity procedure on a ProBond resin and eluted with 20 mM sodium phosphate at pH 6.0, 500 mM NaCl and 500 mM imidazole after prior washing with the same buffer containing 25 mM imidazole to remove contaminating proteins. Crystals were grown by vapour diffusion using the hanging-drop technique at 291 K. The antigen alone was crystallized in a mixture consisting of sodium citrate 0.86 M and Hepes 0.1 M pH 8.4. Crystals were flashcooled in liquid nitrogen after brief soaking in a cryoprotectant consisting of 80% of reservoir added with 20% glycerol.

Characterisation of mAb R31C2
R31C2, a monoclonal antibody raised in an A0 rat against the W1 P. knowlesi strain [6], belongs to the IgG2a isotype. R31C2 was sequenced from total RNA extracted from a hybridoma cell pellet (Fusion Antibodies Ltd, Belfast, Northern Ireland). cDNA was produced from the RNA by reverse transcription with an oligo(dT) primer. Both amplified heavy-chain variable domain (V H ) and light-chain variable domain (V L ) PCR products using variable domain primers were cloned into the Invitrogen sequencing vector pCR2.1 and transformed into TOP10 cells. Positive clones were identified by colony PCR. The sequences have been deposited in GenBank with entry number KM225619 for V L and KM225620 for V H .

Preparation of Fab R31C2 and its complex with PkAMA1
The Fab fragment of R31C2 was obtained by papain digestion and purified on a DEAE-Sephacel support in batch, then on a mono Q column to resolve different crystallizable isoforms. The complex was obtained by 1 h incubation in a 1:1 molar ratio of the recombinant protein with the Fab fragment of R31C2. A volume of 1 μl of the complex was mixed with 1 μl of buffer containing 8% PEG 6000, 40 mM Hepes pH7 and 80 mM NaCl. The drop was set up over a reservoir of 500 μl of buffer and stored at 17°C. Crystals were flash-cooled in liquid nitrogen after brief soaking in a cryoprotectant consisting of 20% PEG 6000, 50 mM Hepes pH 7, 100 mM sodium chloride and 20% glycerol. Diffraction data collection and structure determination X-ray diffraction data for PkAMA1 were collected on beamline BM14 at the ESRF (Grenoble, France), and for the PkAMA1-Fab R31C2 complex on beamline PROXIMA-1 at synchrotron SOLEIL (St Aubin, France). The diffraction images were integrated with the program XDS [41] and crystallographic calculations were carried out with programs from the CCP4 program suite [42]. The PkAMA1 structure was solved by molecular replacement with PHASER software [43] using PvAMA1 (PDB code 1W81) as template. The PvAMA1 coordinates were also used as the antigen template in the R31C2 complex. The V H coordinates from PDB entry 2GCY and the V L coordinates from 1D5I, which showed the best sequence identity with the respective domains of R31C2, were used for the R31C2 variable dimer, and the C H and C L domains from PDB entry 1IGF were used as the search model for the C dimer. Two independent protein molecules were identified in the crystallographic asymmetric unit for PkAMA1 and a single complex for PkAMA1-Fab R31C2. The structures were refined with the program BUSTER [44] and manual adjustments were made to the models with Coot [45]. The crystals parameters, data statistics and final refinement parameters are shown in Table 1 and Table 2. All structural figures were generated with PyMOL (http://www.pymol.org) [46]. The structures were deposited in the Protein Data Bank under the accession codes 4UV6 (PkAMA1) and 4UAO (PkAMA1-Fab R31C2).

Structure of PkAMA1
The free PkAMA1 protein crystallized in space group C2 with two molecules (A and B) in the asymmetric unit and the structure was refined at 2.45 Å resolution ( Table 1, Table 2). The polypeptide chain of molecule A was traced from Ser52 to Glu397 (thus including eight residues of the c-myc tag), with main-chain gaps Ala131-Asp132, Ser212-Lys215 and Gly328-Asn331 due to conformational mobility. Molecule B (Fig 1) was built from Ser52 to Asp398 with mainchain gaps Ser212-Ala217 and Gly328-Ser332. The two molecules are very similar, with an r.m. s.d. of 0.34 Å in Cα positions over the 332 common residues in the PkAMA1 sequence. The D2 loop of PkAMA1, which is displaced upon binding of the RON2 component of the RON receptor complex in the homologues TgAMA1 and PfAMA1 [27,28], could be completely traced in both molecules except for the short region Gly328-Ser332 (Fig 2A). Nonetheless, the high temperature factors of this region indicate a propensity to structural mobility.

Structure of the PkAMA1-Fab R31C2 complex
The PkAMA1-Fab R31C2 complex, which crystallized in the space group C2 with one molecule of complex in the asymmetric unit, was refined at 3.1 Å resolution ( Table 1, Table 2). The main chain of PkAMA1 was traced from Met51 to Phe386, with gaps Glu301-Met302 and Phe330-Asn331 which occur in the D2 loop ( Fig 2B). The heavy and light chains of the Fab fragment were both traced in their entirety to the C-terminal cysteine residues that form the interchain disulfide bridge. (Fig 3) The Fab fragment binds to the hydrophobic groove of PkAMA1 and makes significant contact with the D2 loop. A total of 2465 Å 2 of solvent-accessible surface is buried at the antibody-antigen interface (1180 Å 2 from the antibody; 1285 Å 2 from the antigen). The heavy-chain variable domain (V H ) contributes 838 Å 2 to the buried surface while the light-chain variable domain (V L ) contributes 342 Å 2 . All three Complementarity-Determining Regions (CDR) of both V H and V L make direct contacts with PkAMA1. The shape complementarity factor (shape correlation statistic, Sc) is 0.74, which is significantly above values typical of antibody-antigen complexes [47]. The more significant participation of V H is also reflected in the number and nature of the contacts across the interface; for V H there are 116 interatomic distances less than 3.8 Å, of which 22 are polar (including three salt bridges) while for V L there are only 25 contacts, of which two are polar (Table 3, Table 4, Fig 4). The D2 loop is contacted by CDR-H1, CDR-H2 and CDR-H3 (first, second and third CDR of V H , respectively), with 46 interatomic contacts < 3.8 Å.
Conformational changes that occur in PkAMA1 upon binding R31C2 are restricted mainly to flexible loops that come into contact with the antibody. The largest changes are induced in the PkAMA1 loop Phe169-Asn178, where differences exceed 7 Å at its extremity. A single turn of α-helix is formed from residues Ala172 to Gln175, placing Glu173 and Asp174 in position to form a salt bridge with R31C2 heavy chain residues Arg-H95 and Arg-H56, respectively. This region of PkAMA1 makes the most extensive interactions with R31C2, with residues from CDR-H1, CDR-H2, CDR-H3 and CDR-L1 participating in direct contacts with the antigen. The loop 128-136 becomes more ordered in comparison to the unbound PkAMA1 structure with differences of up to 6 Å between the unbound and bound states. Although the D2 loop makes extensive interactions with R31C2, the regions in direct contact show only small differences. The largest of these differences occur close to the N-and C-termini of the loop, which

Consequences of somatic mutations in R31C2 on PkAMA1 binding
The V H and V L sequences were compared with germline sequences using the International Immunogenetics Information server [48] (Fig 5). The V L domain derives from the Vκ germline gene IGKV14S1 (98.9% identity) and the J gene segment IGKJ5. Five somatic mutations in the amino acid sequence are present in the regions encoded by the Vκ and Jκ gene segments but none of these make direct contacts with PkAMA1. The V H domain derives from the germline gene IGHV5S36 (92.7% identity), the IGHD1-6 D minigene and the IGHJ2 J gene segment. Eighteen somatic amino acid mutations are present, five of which contact PkAMA1: Tyr50Thr, Gly53Ser, Gly55Ser and Ser56Arg in CDR-H2 and Asp100bGlu in CDR-H3. The higher selection pressure on V H probably reflects the significantly more important role of this domain in binding to PkAMA1. The somatic mutation Tyr50Thr leads to a hydrogen bond between Oγ and the glutamate group of Glu173, which would not be sterically possible for the germline residue tyrosine. The mutation Gly55Ser does not lead to additional polar interactions but probably improves the complementarity at the antibody-antigen interface. The mutation Ser56Arg appears to have the most significant impact in the affinity maturation of R31C2 as this antibody residue forms a salt bridge to Glu81 and Asp174 of PkAMA1. CDR-H3 residue Arg95, which forms a salt bridge to Glu173, probably results from nucleotide deletions and additions occurring at the junction between V H and D (used in the third reading frame). PkAMA1 is shown in surface representation with the epitope residues on Domain 1 in green and the epitope residues on the D2 loop in red. In addition, residues of both PkAMA1 and R31C2 that are involved in polar interactions are drawn in stick representation with hydrogen bonds indicated by dotted lines. The stick models of the PkAMA1 residues and their labels are green for Domain 1 and red for the D2 loop. Stick models for R31C2 are blue for V H residues and yellow for V L residues with residue labels in black. (The complete list of PkAMA1 and R31C2 residues in contact is given in Table 3, Table 4). doi:10.1371/journal.pone.0123567.g004

Discussion
R31C2 was the first invasion-inhibitory anti-AMA1 mAb to be characterized, providing the first demonstration that AMA1 plays a crucial role in the infection of RBCs by Plasmodium merozoites and highlighting the potential of this parasite protein as a malaria vaccine candidate [6]. P. knowlesi merozoites are able to attach to the RBC surface in the presence of R31C2 but no penetration of the host cell is observed, implying that AMA1 comes into play after initial contact by the parasite [35]. The accumulated results of many studies with Plasmodium and other Apicomplexa parasites have since confirmed this view. AMA1 has been shown to bind to RON2, a component of the parasite RON protein complex that is transferred to the RBC membrane during invasion [25,27,28]. This interaction is thought to occur at the tight junction formed between the invading merozoite and the RBC. The RON2-binding site on AMA1 comprises the hydrophobic groove and an adjacent region that becomes exposed after displacement of the D2 loop by the receptor [27,28]. In the crystal structure of the PkAMA1-Fab R31C2 complex, the antibody binds not only to the hydrophobic groove but also to the D2 loop (Fig 6A). The D2 loop makes significant contacts with R31C2 (52 out of 142 interatomic contacts <3.8 Å, including nine hydrogen bonds); it nonetheless preserves the same conformation as found in the unbound PkAMA1. The RON2-binding site is thus inaccessible to the receptor in the presence of R31C2 because interactions with the mAb keep the D2 loop firmly in place as well as blocking access to a large fraction of the hydrophobic groove (Fig 6B). Structural analyses of two other complexes of AMA1 with invasion-inhibitory anti-PfAMA1 mAbs, 1F9 [21] and the IgNAR 14I [22], have shown that these also bind to the hydrophobic groove but their epitopes are displaced towards one end of the groove and do not include the D2 loop. By contrast, epitope mapping of the anti-PfAMA1 mAb 4G2, another extensively studied invasion-inhibitory antibody, shows that this antibody does not bind to the hydrophobic groove but rather to the base of the D2 loop [17,49]. The mechanism of inhibition in this case is thus most likely indirect by preventing displacement of the D2 loop to fully expose the RON2-binding site. By targeting both the hydrophobic groove and the D2 loop, R31C2 is thus a very effective inhibitor of RBC invasion.
The D2 loop is functionally important as it must be displaced to fully expose the binding site for the RON receptor. Nonetheless, this region is more variable among the different Plasmodium homologues than is the sequence variability over the complete ectodomain [50]. Interestingly, the D2 loop adopts a different conformation in PkAMA1 to that which has been observed in PfAMA1 (Fig 7). This could be due to sequence differences between the two homologues or could arise from inherent flexibility in this region of AMA1.
Polymorphism in AMA1 has been studied most extensively in P. falciparum [51,52]. In PfAMA1, polymorphism has most likely occurred in response to the host's immune defenses since non-synonymous differences in the nucleotide sequences predominate over synonymous differences. PvAMA1 is also highly polymorphic [50,[53][54][55][56]. We have studied polymorphism in PkAMA1 to understand its effects on protein structure and function, and on the immune response of the host [57]. A total of 21 polymorphic residues were detected in the 52 isolates examined: four in Domain 1, seven in Domain 2 and five in Domain 3. The pattern of polymorphism thus differs from that found in PfAMA1 and PvAMA1, since 14 of the polymorphic sites in PkAMA1 align with sites that are non-polymorphic in the two other homologues. No polymorphisms were located around the periphery of the RON2-binding site of PkAMA1 is shown in surface representation with Domain 1 residues lining the hydrophobic groove that are invariant or well conserved across species [18] shown in cyan and the D2 loop shown in red. The CDR residues of R31C2 are shown in ribbon representation; those of V H are blue and those of V L are yellow. (B) Structure of PfAMA1 complexed with the PfRON2 peptide (PDB entry 3ZWZ) [28]. PfAMA1 is shown in surface representation and in the same orientation as PkAMA1 in (A). Species-conserved residues of Domain 1 that line the hydrophobic groove are shown in cyan. The displaced D2 loop is not visible in this structure, probably due to high mobility. Binding of the PfRON2 peptide, shown here in orange as ribbon representation, requires displacement of the D2 loop to expose the complete receptor-bing site. doi:10.1371/journal.pone.0123567.g006 Structure of PkAMA1 with an Inhibitory Antibody PkAMA1, in contrast to PfAMA1 and PvAMA1 where variability in these regions has been suggested to play an important role in immune escape from the host response [18]. Moreover, none of the residues comprising the epitope of R31C2 are polymorphic and the mAb should thus be largely cross-strain reactive. Of note, in the high frequency polymorphism Arg296Ser, the arginine residue forms a salt bridge to Glu288; these two residues correspond to Lys351 and Glu343 in PfAMA1, which do not form a salt bridge [18]. It is unclear if this PkAMA1 polymorphism influences the D2 loop conformation. Nonetheless, the absence of polymorphism close to the receptor-binding site of PkAMA1 suggests that, unlike PfAMA1 and PvAMA1, a single variant sequence may be sufficient for an effective vaccine formulation against P. knowlesi malaria.