Crystal Structure of the Malaria Vaccine Candidate Apical Membrane Antigen 1

See allHide authors and affiliations

Science  15 Apr 2005:
Vol. 308, Issue 5720, pp. 408-411
DOI: 10.1126/science.1107449


Apical membrane antigen 1 from Plasmodium is a leading malaria vaccine candidate. The protein is essential for host-cell invasion, but its molecular function is unknown. The crystal structure of the three domains comprising the ectoplasmic region of the antigen from P. vivax, solved at 1.8 angstrom resolution, shows that domains I and II belong to the PAN motif, which defines a superfamily of protein folds implicated in receptor binding. We also mapped the epitope of an invasion-inhibitory monoclonal antibody specific for the P. falciparum ortholog and modeled this to the structure. The location of the epitope and current knowledge on structure-function correlations for PAN domains together suggest a receptor-binding role during invasion in which domain II plays a critical part. These results are likely to aid vaccine and drug design.

Apical membrane antigen 1 (AMA1) is currently in clinical trials as a vaccine against P. falciparum, the species causing the most serious forms of malaria in humans. AMA1 is present in all Plasmodium species examined (1), and orthologs exist in other Apicomplexa, including Toxoplasma (2) and Babesia (3). Although little is known about its molecular function, genetic evidence indicates a role in maintaining parasite growth during the blood-stage cycle (4). Antibodies raised against AMA1 can inhibit erythrocyte invasion and protect against the disease in animal-model systems of malaria (5-9). Furthermore, invasion-inhibitory antibodies to AMA1 have been affinity-purified from human sera of donors from malaria-endemic regions (10). AMA1 is stored in the microneme organelles after synthesis and is translocated to the parasite surface just before or during invasion (11). It is a type I integral membrane protein with an N-terminal ectoplasmic region, a single transmembrane region, and a small C-terminal cytoplasmic domain (1). In all characterized Plasmodium AMA1 genes, 16 invariant Cys residues are encoded in the ectoplasmic region, and analysis of the disulfide-bond pattern suggests a division into three distinct domains (12, 13). Correct disulfide pairing in the recombinant protein is essential for inducing a protective immune response (6). The effectiveness of an AMA1-based vaccine therefore depends critically on maintaining authentic conformational epitopes. We report here the crystal structure of the AMA1 ectoplasmic region from P. vivax (PvAMA1) and the epitope mapping of an invasion-inhibitory monoclonal antibody specific for the P. falciparum ortholog PfAMA1. Together, these results suggest that domain II is important for the biological function of AMA1.

The crystal structure of the PvAMA1 ectoplasmic region, comprising residues Pro43 to Leu487, was determined in two crystal forms, monoclinic and orthorhombic (14), and refined at 1.8 Å and 2.0 Å, respectively (15). The structure confirms the division of this region into three structural domains, denoted I, II, and III (Fig. 1 and figs. S1 and S2). Several breaks occur in the main-chain tracing of all three domains as a result of absent or poorly defined electron density; these regions correspond to 15% of the polypeptide chain (fig. S2). Of particular note, no electron density was present for the 40-residue segment 295 to 334 in domain II, showing that this region is disordered or mobile. We refer to this region as the domain II loop.

Fig. 1.

A stereo view of the ectoplasmic region of orthorhombic PvAMA1 in ribbon representation showing domain I (green), domain II (blue), and domain III (magenta). Residues Gln294 and Phe335 indicate the limits of the disordered domain II loop. Figures were generated using MOLSCRIPT (27) and Raster3D (28).

To further understand the function of AMA1, we mapped the epitope recognized by the invasion-inhibitory monoclonal antibody 4G2 (16), specific for PfAMA1, and related this to a model of PfAMA1 constructed by homology from the PvAMA1 structure. Different domain combinations of PfAMA1 (I-II, II-III, and I-II-III; and I, II, and III separately) expressed in Pichia pastoris (9, 15) were used to test 4G2 binding. Only the I-II and I-II-III combinations were reactive under nonreducing conditions and none were reactive under reducing conditions, which shows that 4G2 recognizes a conformational epitope requiring both domain I and domain II. We further mapped the 4G2 epitope using a series of mutations (mainly replacements by Ala) introduced into domains I and II of PfAMA1 expressed on the surface of COS-7 cells (15). Correctly folded mutants, as detected by reactivity with polyclonal antibodies specific for conformational epitopes, were then tested for recognition by 4G2. All but one of the mutations that abrogated PfAMA1 recognition by 4G2 occur in the region between residues 348 and 389 (equivalent to residues 293 to 334 in PvAMA1). This contains the domain II loop, which is disordered or mobile. The 4G2 epitope includes residues from the base of the loop at both the N- and C-terminal ends (PfAMA1 residues Asp348, Lys351, Gln352, Glu354, Gln355, His356, Phe385, Asp388, and Arg389) but not from the central region (fig. S2). Of note, all residues except Asp348 fall within the disordered region and, accordingly, were not modeled. Lys280 in domain I (Lys225 in PvAMA1) was the only 4G2-sensitive residue detected outside the domain II loop. This residue, which is invariant in Plasmodium species, is buried, and its amino group forms hydrogen bonds to the main-chain atoms in domains I and II and an invariant Asn on helix α7 in domain II (Asn338 in PfAMA1, Asn283 in PvAMA1). Replacement of Lys280 in PfAMA1 by Ala is thus likely to induce structural changes at the domain I-domain II interface. This would explain our observation that both domains I and II are necessary to maintain the conformational 4G2 epitope, because the base of the domain II loop interfaces with domain I. In addition, the Lys280Ala mutant did not show full recognition by conformation-sensitive polyclonal antibodies, indicating some effect on global folding. We conclude that the invasion-inhibitory 4G2 epitope is confined essentially to the base of the domain II loop.

Comparison of the PvAMA1 structure with other known three-dimensional structures using a distance matrix alignment procedure (DALI) (17) reveals that domains I and II are structurally similar to each other and belong to the PAN module superfamily (Fig. 2 and figs. S3 and S4). No previously determined polypeptide fold, however, was identified for domain III. PAN domains are found in proteins with diverse adhesion functions, binding to protein or carbohydrate receptors (18). They generally show low sequence identity within the superfamily and have been recognized largely by the cystine pattern and the observed or predicted secondary structure. The closest correspondence of domains I and II of PvAMA1 to known structures bearing PAN domains was found with leech antiplatelet protein (LAPP) (19). The PvAMA1 domains are compared with LAPP in Fig. 2, showing preservation of the canonical PAN secondary structure elements and the connectivity between them. The PAN disulfide-bridge pattern is less well preserved in AMA1; only Cys162 to Cys192 in domain I and Cys282 to Cys354 in domain II (fig. S4), equivalent to the disulfide bridge between β4 and the helix in LAPP, conform to the PAN motif (18). The structural similarity between AMA1 domains I and II suggests that they may originate from an ancient gene duplication. The structure of domain III from PfAMA1, expressed alone, has been analyzed in solution by nuclear magnetic resonance (NMR) (13). A significant fraction of the NMR structure is disordered, which can be explained by the absence of domains I and II that interact with these regions of domain III.

Fig. 2.

Comparison in stereo of domain I and domain II of PvAMA1 with LAPP (PDB entry 1i8n). Components characterizing the PAN motif are illustrated as follows: the helix in brown, the 5-stranded β sheet in blue, and the double β strand from the joining segments in yellow. The three domains are viewed from a common orientation: (A) domain I, (B) domain II, and (C) LAPP. Strands β4-β10-β7-β8-β6 of domain I and β11-β18-β15-β16-β14 of domain II correspond to the central 5-stranded β1-β5-β3-β4-β2 sheet of the PAN motif in LAPP, whereas α5 and α7 correspond to the PAN helix in the two respective domains. The inner and outer crossing segments, occurring between β4 and β6, and β8 and β10, respectively, in domain I, and between β11 and β14, and β16 and β18, respectively, in domain II, also conform to the PAN module organization. PvAMA1 domain I residues 92 to 238 superimpose on LAPP residues 46 to 122, with a root mean square (RMS) difference of 2.3 Å between 70 structurally equivalent Cα atoms (DALI z score 4.0). PvAMA1 domain II residues 249 to 376 superimpose on LAPP residues 47 to 22, with a RMS difference of 2.6 Å between the 70 structurally equivalent Cα atoms (z score 5.1). PvAMA1 domain I residues 93 to 243 optimally position onto PvAMA1 domain II residues 249 to 381, with a RMS difference of 2.7 Å between 83 structurally equivalent Cα atoms (z score 6.0).

The structural similarity of a protein with unknown function to one with well characterized properties can sometimes provide useful insights into its biological role (20). The splice variant of hepatocyte growth factor NK1 is the only protein of known structure with a PAN domain that has been studied in association with its receptor (21). NK1, comprising an N-terminal PAN domain and a C-terminal Kringle domain, binds heparin. The crystal structure of NK1 in complex with heparin shows that the ligand binds largely to the PAN domain. The heparin-binding site on the NK1 PAN domain is formed by β2, the helix, and the loop region connecting the helix to β3. Arg and Lys residues are involved in charged interactions with sulfate groups of heparin, and their importance has been confirmed by binding studies with mutants.

To determine whether AMA1 could accommodate ligands or receptors in a similar way, we compared the heparin-binding site on the PAN domain of NK1 with the equivalent surfaces on domains I and II. Docking of the NK1-heparin complex onto PvAMA1 domain I showed that such an interaction is sterically hindered by helix α3 and the loop that precedes β6 and thus would not form an NK1-like binding site. For domain II, however, heparin could be docked without engendering steric hindrance with the antigen (Fig. 3). This putative receptor-binding site on domain II is formed by strand β14 (equivalent to β2 in NK1) and helix α7 (the PAN domain helix). The loop connecting α7 and β15 (equivalent to the helix and β3 in NK1), by contrast, is not positioned for direct contact with the ligand as in NK1. Interestingly, however, this latter region includes the disordered domain II loop, and we do not exclude the possibility that the mobile 40-residue peptide segment closes upon the putative receptor once the latter is in place (Fig. 3). This conjecture is compatible with our observation that the epitope recognized by the invasion-inhibitory antibody 4G2 maps to the base of this loop. Although we are aware that the distribution of charged residues at this postulated binding site on AMA1 does not favor fixation of negatively charged ligands such as heparin, we speculate that other glycans or proteins might be candidates for interaction with domain II.

Fig. 3.

A stereo view of domain II of AMA1 (blue), with the N domain of NK1 (red) complexed to heparin (yellow with sulfur atoms in green) superimposed. The regions of AMA1 domain II that, by analogy with NK1, may potentially form a ligand-binding site are helix α7, strand β14, and the domain II (DII) loop. The disordered domain II loop (not modeled) could conceivably close upon this site once the putative ligand is in place.

Other micronemal proteins from Apicomplexa include PAN or Apple (a subfamily of PAN) motifs in their structural organization (22, 23). Micronemal proteins play important roles in host-cell invasion (23), and several lines of evidence imply a similar function for AMA1 (24-26). The presence of PAN motifs in AMA1 and other micronemal proteins may indeed suggest a common phylogenetic origin. Although these diverse observations are consistent with receptor binding by AMA1, direct confirmation of such a role and the nature of the putative ligand remain elusive. The presence of PAN domains in PvAMA1 and our mapping of a protecting epitope to the structure nonetheless provide support for this role by highlighting a functionally important region of the molecule. These results establish a conceptual basis for future experiments to probe the function of AMA1; moreover, they should contribute to further development of vaccine and drug design strategies against malaria.

Supporting Online Material

Materials and Methods

Figs. S1 to S4

Table S1


Movie S1

References and Notes

View Abstract

Navigate This Article