Fine Mapping of an Epitope Recognized by an Invasion-inhibitory Monoclonal Antibody on the Malaria Vaccine Candidate Apical Membrane Antigen 1*

Antibodies that inhibit red blood cell invasion by the Plasmodium merozoite block the erythrocytic cycle responsible for clinical malaria. The invasion-inhibitory monoclonal antibody (mAb) 4G2 recognizes a conserved epitope in the ectodomain of the essential Plasmodium falciparum microneme protein and vaccine candidate, apical membrane antigen 1 (PfAMA1). Here we demonstrate that purified Fab fragments of 4G2 inhibit invasion markedly more efficiently than the intact mAb, suggesting that the invasion-inhibitory activity of this mAb is not due solely to steric effects and that the epitope lies within a functionally critical region of the molecule. We have taken advantage of a synthetic gene encoding a modified form of PfAMA1, and existing x-ray crystal structure data, to fully characterize this epitope. We first validate the gene by demonstrating that it fully complements the function of the authentic gene in P. falciparum.We then use it to identify a group of residues within the previously described domain II loop of PfAMA1 that are critical for recognition by mAb 4G2 and demonstrate that the epitope lies exclusively within this loop with no contributions from residues in other domains of the molecule. This is the first complete characterization of a conserved invasion-inhibitory epitope on PfAMA1. Our results will aid in the design of subunit vaccines designed to generate a broadly effective, focused anti-PfAMA1 protective immune response and may help elucidate the function of PfAMA1.

Antibodies that inhibit red blood cell invasion by the Plasmodium merozoite block the erythrocytic cycle responsible for clinical malaria. The invasion-inhibitory monoclonal antibody (mAb) 4G2 recognizes a conserved epitope in the ectodomain of the essential Plasmodium falciparum microneme protein and vaccine candidate, apical membrane antigen 1 (PfAMA1). Here we demonstrate that purified Fab fragments of 4G2 inhibit invasion markedly more efficiently than the intact mAb, suggesting that the invasion-inhibitory activity of this mAb is not due solely to steric effects and that the epitope lies within a functionally critical region of the molecule. We have taken advantage of a synthetic gene encoding a modified form of PfAMA1, and existing x-ray crystal structure data, to fully characterize this epitope. We first validate the gene by demonstrating that it fully complements the function of the authentic gene in P. falciparum. We then use it to identify a group of residues within the previously described domain II loop of PfAMA1 that are critical for recognition by mAb 4G2 and demonstrate that the epitope lies exclusively within this loop with no contributions from residues in other domains of the molecule. This is the first complete characterization of a conserved invasion-inhibitory epitope on PfAMA1. Our results will aid in the design of subunit vaccines designed to generate a broadly effective, focused anti-PfAMA1 protective immune response and may help elucidate the function of PfAMA1.
Over half of the world's population is exposed to malaria (1). Until recently, the disease was controlled in or eradicated from a number of areas. However, the emergence of insecticide-resistant mosquito vectors and multidrug-resistant forms of the causative agent has contributed to resurgences of the disease, which in some cases have resulted in near catastrophic epidemics. As a result malaria remains a global problem, affecting many of the poorest nations and representing a serious threat to travelers. The development and implementation of effective new drugs and a vaccine is of paramount importance.
Clinical malaria results from replication of protozoan parasites of the genus Plasmodium in circulating erythrocytes. Like most apicomplexan pathogens, the malaria merozoite invades its host cell in a multistep process initiated by reversible binding to receptors on the erythrocyte surface, followed by high affinity attachment via the apical end of the merozoite, and finally entry into a parasitophorous vacuole. Invasion is facilitated by the discharge of apical secretory organelles called micronemes and rhoptries. The type I integral membrane microneme protein apical membrane antigen 1 (AMA1) 2 is widely regarded as a leading candidate for inclusion in a malaria vaccine (2)(3)(4)(5). Identified initially in the simian malaria species Plasmodium knowlesi as a target of a monoclonal antibody (mAb) that prevented erythrocyte invasion (6), homologues of AMA1 have been identified in all species of Plasmodium (6 -13) and in all other apicomplexan genera studied (14 -17). Numerous reports have confirmed that antibodies to AMA1 interfere with host cell invasion in vitro or protect against blood stage growth in vivo (6, 18 -24). AMA1 is also expressed in sporozoites, the form of the parasite injected into the vertebrate host by the mosquito vector, and the recent demonstration that anti-AMA1 antibodies inhibit hepatocyte invasion by sporozoites suggests that an AMA1-based vaccine may be effective against both pre-erythrocytic and blood stage phases of infection (25).
Development of an AMA1-based vaccine will be aided by a detailed understanding of the function of the protein and the mechanism by which protective antibodies interfere with it. An essential role for AMA1 was first indicated by unsuccessful attempts to disrupt its gene in both P. falciparum (26) and the related apicomplexan Toxoplasma gondii (14). An elegant electron microscopic study in P. knowlesi using the mAb referred to above indicated that AMA1 is required for the reorientation step that follows the initial interaction between merozoite and host cell (27). Using a conditional expression system in T. gondii, Mital et al. (28) confirmed that AMA1 plays a role in the later stages of host cell invasion and obtained evidence that it may mediate regulation of rhoptry discharge. Determination of the x-ray crystal structure of AMA1 of P. vivax (PvAMA1) (29) and P. falciparum (PfAMA1) (30) confirmed previous indications (31) that the AMA1 ectodomain comprises three disulfide-constrained domains and showed that two of these (domains I and II) form PAN modules, often implicated in protein-protein and protein-carbohydrate interactions. Consistent with this, two reports have suggested an erythrocyte-binding role for AMA1 (32,33). More recent studies in T. gondii and P. falciparum have demonstrated that AMA1 functions as a complex with at least two distinct proteins (RON2 and RON4), which initially reside in the rhoptry neck, and that this complex contributes to formation of the moving junction that translocates over the invading parasite surface as it moves into the nascent parasitophorous vacuole (34,35). Despite all the above data, the molecular details of AMA1 function remain obscure. The biosynthesis and trafficking of PfAMA1 has been characterized in some detail. Maximally expressed during the latter stages of schizont maturation, PfAMA1 is targeted to the micronemes of developing merozoites (36,37) where it is subject to a proteolytic processing event in which a short N-terminal "prosequence" is removed, converting the 83-kDa precursor protein (PfAMA1 83 ) to a 66-kDa product called PfAMA1 66 (38,39). Upon schizont rupture PfAMA1 66 (but not its PfAMA1 83 precursor) is released from the micronemes to distribute across the surface of the free merozoite (38,39). At or around the point of invasion the molecule is then quantitatively shed by a membrane-bound subtilisin-like "sheddase" called PfSUB2, which translocates across the parasite surface, possibly in association with the moving junction (40,41). Shedding is mediated by a single cleavage just 29 residues distal to the transmembrane domain, resulting in release of essentially the entire AMA1 ectodomain in the form of a 48-kDa polypeptide called PfAMA1 48 ; an additional low level processing event within this molecule generates two disulfide-linked fragments, the larger of which is called PfAMA1 44 (41). Following invasion only the juxtamembrane extracellular PfAMA1 "stub," still membranebound via its cognate transmembrane domain and cytoplasmic tail, can be detected in the young ring stage parasite (42). The role of AMA1 shedding is unknown. However, some antibodies that prevent invasion also interfere with shedding, raising the possibility that shedding is essential for invasion. Its inhibition may represent one mechanism by which anti-AMA1 antibodies exert invasion-inhibitory activity (43,44).
PfAMA1 exhibits significant polymorphism, and polymorphic residues have been identified at 52 different positions within the ectodomain (45). In most cases these are dimorphic, being represented by one or the other of two alternative residues, but more extensive diversity is evident at a number of positions. It has been suggested that the polymorphisms are a result of selective pressure exerted by host immune responses. A number of observations support this, including the fact that all the polymorphic residues are surface-exposed and that invasion-inhibitory anti-PfAMA1 antibody responses are often strain-specific. Remarkably, the distribution of the polymor-phic sites is highly biased to one side of the PfAMA1 ectodomain (45), but the significance of this is unknown.
The availability of PfAMA1-specific mAbs that inhibit invasion provides opportunities to identify functionally important regions of the molecule. A recent study using phage display of PfAMA1 fragments showed that recognition of PfAMA1 by a strain-specific, invasion inhibitory mAb called 1F9 required the presence of a Glu residue at position 197, the most polymorphic position in PfAMA1 (46). This site is flanked by a number of other polymorphic sites. In our earlier preliminary work we used expression of a range of PfAMA1 mutants to show that recognition by a different invasion-inhibitory mAb called 4G2 involved interactions with an exposed flexible loop of domain II, composed of residues that are conserved in all P. falciparum isolates. Here we extend this work. We show that purified Fab fragments of 4G2 potently inhibit invasion, suggesting that the epitope lies within a functionally important region of the ectodomain. Next, we demonstrate by genetic complementation that a wholly synthetic gene designed for heterologous expression of PfAMA1 is fully functional in P. falciparum. Finally, we make use of this gene along with all the available high resolution structural data and a greatly expanded library of mutants to map the location of the 4G2 epitope with high precision. We show conclusively that binding of mAb 4G2 involves exclusively interactions with side chains of residues within the domain II loop, with no contribution by any residues from domain I or domain III.

EXPERIMENTAL PROCEDURES
Antisera, Monoclonal Antibodies, and Fab Fragments-Mouse antisera were raised against the mature ectodomain of PfAMA1 (Ile 97 -Lys 546 ) expressed in Pichia pastoris from a synthetic gene (FVO ama-1syn) designed to replace the AϩT-rich codon usage of the P. falciparum FVO strain (23,47). Mice were immunized with the purified recombinant protein (called rPfAMA1) either in its native state or following reduction and alkylation as described previously (29), generating antisera N5 and R5, respectively. Antiserum N5 was then adsorbed with reduced and alkylated immunogen to remove antibodies reactive with this form of the protein. The product was an antiserum called N5Ads that recognizes only native, disulfide-dependent conformations of PfAMA1 and shows no cross-reactivity with the denatured form.
For detection of hemagglutinin (HA) epitope-tagged proteins in Western blot and immunofluorescence (IFA), the HAspecific mAb 3F10 (Roche Applied Science) was used. The mAb 1E1 (48), which recognizes an epitope in the C-terminal domain of merozoite surface protein-1 (MSP1 19 ), was conjugated to Alexa Fluor 594 (Molecular Probes, Eugene, OR) as described previously (40). The mAb 4G2 (23, 49) was affinitypurified from hybridoma culture supernatants using Protein G-Sepharose (GE Healthcare). For production of 4G2 Fab fragments, purified mAb at 12.3 mg ml Ϫ1 in 0.1 M NAOH/citrate solution, pH 3.5, was digested with pepsin (1 mg ml Ϫ1 ) at 37°C for 2 h. The reaction was terminated by the addition of an equal volume of 1 M Tris-HCl, pH 8.2. The resulting F(ab) 2 was purified by chromatography on a Superdex 200 HR gel filtration column equilibrated in 20 mM Tris-HCl, 150 mM NaCl, pH 8.2.
The recovered F(ab) 2 was reduced and alkylated by addition of 10 mM L-cysteine and incubation under nitrogen for 2 h at 37°C, followed by the addition of iodoacetamide to 130 mM and incubation in the dark at room temperature for a further 2 h. Fab fragments were finally purified from this mixture by a second round of gel filtration as described above, then concentrated by ultrafiltration and stored at Ϫ70°C.
Construct Synthesis-A construct based on pSecTag2A (Invitrogen) for transient, cell surface expression of a region of ama-1syn, encoding the entire PfAMA1 except for its native secretory signal sequence and pro-sequence (called sgP-fAMA1 66 ) in COS-7 cells, has been described previously (29). Mutations were introduced into regions encoding domains I and II (DI-II) of the PfAMA1 ectodomain by QuikChange sitedirected mutagenesis (Stratagene). The resulting mutants were expressed in COS-7 cells and analyzed by Western blotting (29).
For expression of ama-1syn in P. falciparum, a single HA epitope tag (YPYDVPDYA in single letter amino acid code) (50) was first introduced into the juxtamembrane stub region (42) of the PfAMA1 ectodomain. To do this, sequence encoding 535IPEHKPTYD543 was modified by a two-step site-directed mutagenesis approach using QuikChange (Stratagene). Forward primers used for this were 2ndHalfHAfor (5Ј-GAG CAT AAG CCT gac Tat GcT AAC ATG AAG-3Ј) plus its reverse complementary primer (called 2ndHalfHArev), followed by FullHAfor (5Ј-GAA ATC GCT GAT taT CCA tAc gAT gtTC CTg acT AtG cTA ACA GTA AG-3Ј) plus its reverse complementary primer (FullHArev). Mutations required to introduce the tag are indicated in lowercase. For episomal expression of the resulting tagged gene (called ama-1synHA) in P. falciparum, the unique internal XhoI site was removed by site-directed mutagenesis leaving the encoded amino acid sequence unaltered, using forward primer XhoIrepfor (5Ј-CCA AGA AAG AAC CTg GAa AAC GCA AAG TTC GGT CTG TGG-3Ј) and its corresponding reverse complementary primer XhoIreprev (the original position of the XhoI site is indicated in bold, and the mutations introduced to remove this are shown in lowercase). The entire coding sequence was then amplified by PCR using primers XhoIfor (5Ј-ACC GAG ctc gag TTC ATG AGG AAG-3Ј) and XhoIrev (5Ј-GCC AGT ctc gag TTA GTA GTA AGG C-3Ј), and the product was cloned into pHAM-ACPGFP (unpublished; a kind gift of J. Healer, Walter and Eliza Hall Institute of Medical Research, Melbourne, Victoria 3050, Australia) pre-digested with XhoI to excise the acpgfp gene. The resulting plasmid (called pHAM-sgPfa1/HA) contains an expression cassette comprising the ama-1synHA gene between 1.4 kb of pfama1 5Ј-flanking sequence containing the pfama1 promoter, and the 3Ј untranslated region of the P. berghei dihydrofolate reductase-thymidylate synthase gene, present to ensure correct transcription termination and polyadenylation of the gene.
Plasmid pHH1-5ЈsgPfa1/HA was designed to disrupt the endogenous pfama1 gene by single crossover homologous recombination into the locus, simultaneously inserting the ama-1synHA expression cassette into the chromosome. A 900-bp region from the 5Ј-end of pfama1 (starting 60 bp downstream of the translational start ATG site) was amplified from P. falciparum T9/96 genomic DNA using primers PfaI-5Јfor (5Ј-CAT A TA TGA aga tct TTG GAA GAG GAC AG-3Ј) and PfaI-5Јrev (5Ј-ctc gag GTA TAT CTT CTC AAT TTC CAT CGA CCC-3Ј) and cloned via BglII and XhoI sites (indicated in lowercase) into pre-digested plasmid pHH1 (51) producing plasmid pHH1-5ЈPfaI. The reverse primer was designed to incorporate a stop codon (indicated in bold) at the 3Ј-end of this gene-targeting fragment. The ama-1synHA expression cassette was excised from pHAM-sgPfa1/HA by digestion with HindIII and NotI, blunted with T4-DNA polymerase, and cloned downstream of the 900-bp target sequence into XhoI-digested, blunted pHH1-5ЈPfaI producing construct pHH1-5ЈsgPfa1/HA.
Culture and Transfection of P. falciparum and COS-7 Cells-Asexual blood stages of P. falciparum clone 3D7 were cultured and synchronized as previously described (52,53). Ring stage parasites at 1-5% parasitemia (assessed by microscopic examination of Giemsa-stained thin blood films) were transfected by electroporation with 70 -100 g of plasmid DNA as previously described (40,54). After initial selection for ϳ4 weeks in 10 nM of the antifolate WR99210 (Jacobus Pharmaceuticals, Buena, NJ), parasites were subjected to repeated cycles of selection pressure followed by removal of the drug for 3 weeks. Clonal populations were eventually obtained by limiting dilution. Synchronous parasite preparations were saponin-treated and washed in phosphate-buffered saline for analysis by Western blot as described previously. For analysis of proteins released into culture supernatants, purified schizonts were cultured for 8 h in the presence of fresh human erythrocytes to allow reinvasion and shedding of invasion proteins. Parasites and uninfected erythrocytes were removed by centrifugation and the culture supernatant concentrated 50-fold using a 5000 molecular mass cut-off Vivaspin column (Vivascience) before analysis by Western blot. COS-7 cells adapted for growth under serum-free conditions were cultured in 6-well plates and transfected with constructs for expression of unmodified and mutant forms of sgPfAMA1 66 as previously described (29) using 3 g of plasmid DNA. Cells were harvested 48 h post-transfection for analysis by Western blot or IFA as described previously (29,40).

SDS-PAGE and Western
Blotting-Parasite and COS-7 cell lysates were separated by SDS-PAGE under non-reducing or reducing conditions prior to transfer to Hybond-C extra nitrocellulose membrane (Amersham Biosciences), and membranes were probed with mAbs or polyclonal antibodies as previously described (40).
Indirect Immunofluorescence Analysis-Thin films of P. falciparum cultures were fixed in acetone as described previously (40). Surface labeling of COS-7 cells was carried out using unfixed preparations as previously described (55). Slides were mounted in Citifluor (Citifluor Ltd., Canterbury, UK). Images were captured using AxioVision 3.1 software on an Axioplan 2 Imaging system (Zeiss) using a Plan-Apochromat 100ϫ/1.4 oil immersion objective and annotated using Adobe PhotoShop.
Invasion-inhibition assays-Purified mAb 4G2 and Fab fragments thereof were dialyzed against phosphate-buffered saline before use in invasion-inhibition assays. Cultures containing highly synchronous mature schizonts (2% hematocrit and 10% parasitemia) were cultured for 4 h in the presence of varying concentrations of purified mAb 4G2 or Fab fragment to allow schizont rupture and merozoite invasion. Thin blood films were Giemsa-stained, and the number of rings in 10,000 erythrocytes was counted for each condition. Invasion inhibition was determined as a percentage of the mean parasitemia observed in the absence of added antibody.
Modeling of the PfAMA1 Ectodomain-Coordinates for PfAMA1 domains I and II (PDB ID code 1Z40) and the PvAMA1 ectodomain (PDB ID code 1W81 and 1W8K) were aligned using LSQKAB (56). The PDB files were combined, and PvAMA1 to PfAMA1 mutations were inserted in sterically favorable conformations using O (57). The resulting structure extended from residues 96 to 533 and was complete apart from the domain III loop (residues 458 -469), which was inserted in a random orientation. The final model was energy-minimized using CNS (58). The figures were created using PyMOL. 3

Fab Fragments of mAb 4G2 Are Potent Inhibitors of Erythro-
cyte Invasion-Highly purified Fab fragments of mAb 4G2 (Fig.  1A) were tested for invasion inhibitory activity at a range of concentrations, in parallel with intact mAb. The Fab was consistently a significantly more potent inhibitor of invasion than the intact mAb (Fig. 1B). 4G2 Fab at a concentration of 0.5 mg/ml resulted in virtually 100% invasion inhibition, whereas at five times the concentration (2.5 mg/ml) the intact antibody only resulted in 70% inhibition. These results demonstrate that the mechanism by which mAb 4G2 interferes with invasion requires neither cross-linking of exposed PfAMA1 molecules nor agglutination of free merozoites (a phenomenon that has been observed in the case of some invasion-inhibitory antimerozoite antibodies (59)), both of which can only be mediated by divalent intact antibody. Given the small mass of Fab relative to that of intact IgG, these results also suggest that mAb 4G2 interferes with invasion not simply as a result of steric effects but because antibody binding interferes directly with the function of a critical domain of PfAMA1.
Functional Complementation of the pfama1 Gene by a Synthetic Recodonised Gene-In light of the above results, we set out to accurately map the mAb 4G2 binding site within the PfAMA1 ectodomain. The nucleotide composition of the P. falciparum genome is extremely AϩT-rich, and early attempts to express the authentic pfama1 gene in eukaryotic systems resulted in truncated products due to premature transcription termination and/or inefficient translation. To overcome this problem, we previously designed a synthetic gene (called FVO ama-1syn), replacing the AϩT-rich codon usage of the FVO pfama1 gene with one optimized for expression in the methylotrophic yeast P. pastoris (23). The FVO PfAMA1 sequence contains six potential N-glycosylation sites (Asn-Xaa-Ser/Thr). These sites are not modified by N-glycosylation in the parasite (39), so to avoid inappropriate glycosylation of the gene product in P. pastoris or other heterologous expression systems the coding sequence of the synthetic gene was altered to remove all six sites. Before using this gene here for detailed analysis of the 4G2 epitope we decided to ensure that these modifications did not interfere with the function of the gene by investigating its capacity to complement the function of the authentic parasite gene.
To do this, the FVO ama-1syn sequence was first modified by replacing sequence encoding 9 amino acids (Ile 535 -Asp 543 ) in the stub region of the ectodomain with that encoding an HA epitope tag. The modified gene was called ama-1synHA. Preliminary experiments performed with plasmid construct pHAM-sgPfa1/HA, designed to express ama-1synHA from an episomal location, showed expression and correct trafficking of the gene product (data not shown). In the light of this, construct pHH1-5ЈsgPfa1/HA was produced, designed to disrupt the chromosomal pfama1 gene by single crossover homologous recombination into the coding sequence, simultaneously inserting the entire gene expression cassette for expression of  ama-1synHA under the control of the pfama1 promoter (Fig. 2,  A and B). Successful integration was predicted to result in the endogenous gene-coding sequence being interrupted about 900 bp downstream of the start ATG by a stop codon, followed immediately by the entire integrated plasmid and finally the displaced endogenous pfama1 sequence, lacking both a promoter sequence and a translational start site. Following transfection of the construct and drug cycling to select for integration events, the selected population (called 3D7-sgPfA1/HA) was cloned. Southern blot analysis of four clones showed that in all cases the plasmid had integrated as expected into the 5Ј region of the pfama1 coding sequence (Fig. 2C).
PfAMA1 is maximally expressed toward the end of the fourth nuclear division in schizogony and is targeted to the micronemes of developing merozoites (36 -38). To determine whether the ama-1synHA gene product was correctly trafficked, mature segmented schizonts of transgenic clone 3D7-sgPfAMA1/HA.F5 were analyzed by IFA using the anti-HA mAb 3F10. Fig. 3 shows that the ama-1synHA gene product was readily detectable at the apical prominence of intracellular and free merozoites of the transgenic clone. These data indicated that the synthetic gene product is correctly and specifically sorted to micronemes.
PfAMA1 is trafficked to micronemes in the form of an 83-kDa precursor protein called PfAMA1 83 . This is then N-terminally truncated to produce a 66-kDa form, PfAMA1 66 , prior to distribution onto the merozoite surface (38,39,41). Western blot analysis of 3D7-sgPfA1/HA.F5 schizonts with mAb 3F10 or polyclonal anti-PfAMA1 antibodies (Fig. 4A) showed the presence of both forms of the protein, indicating correct processing of the ama-1synHA gene product. These polypeptides migrated slightly more rapidly on reducing and non-reducing SDS-PAGE than the wild-type products (Fig.  4A, middle and right-hand panels), probably due to the above-mentioned sequence differences between the authentic 3D7 pfama1 gene and the ama-1synHA gene. Importantly, this characteristic enabled us to observe that the 3D7-sgPfAMA1/HA.F5 clone exhibited no detectable expression of the wild-type PfAMA1 (note the absence of a doublet in the F5 lanes of Fig. 4A, right-hand panel). This is consistent with the Southern blot data described above, confirming that integration of the ama-1synHA expression construct had resulted in complete loss of expression of the endogenous pfama1 gene.
During erythrocyte invasion PfAMA1 66 is shed from the merozoite surface in the form of a 48-kDa polypeptide (PfAMA1 48 ) comprising the bulk of the ectodomain, as well as a minor 44 species (PfAMA1 44 ) arising from an internal "nick" within PfAMA1 48 (41,42). To determine whether the ama-1synHA gene product was correctly shed, culture supernatants collected from clone 3D7-sgPfA1/HA.F5 were analyzed by Western blot alongside similar samples from parental 3D7 parasites. Fig. 4C shows that the expected PfAMA1 48/44 doublet was evident in both cases. Again it was notable that the 3D7-sgPfA1/HA.F5-derived polypeptides migrated slightly differently from the wild-type 3D7 proteins. These results confirm that the ectodomain of the synthetic gene product is correctly shed from the parasite surface.
To finally confirm full functionality of the ama-1synHA gene, three of the transgenic clones (clones 3D7-sgPfA1/ HA.F5, -D8, and -C9) were selected for detailed analysis of their growth rates. Parasites were removed from drug pressure 5 days prior to commencing the assay. Cultures containing synchronous ring stage parasites were then adjusted to a parasitemia of 0.5% and maintained in the absence of drug for 6 days. Parasitemias were measured daily by microscopic examination of Giemsa-stained thin blood films, and every 2 days the cultures were diluted 1:5 into fresh medium containing fresh red cells. No significant differences in replication rates were observed between the parental 3D7 line and the transgenic clones (not shown). Collectively, these results show that the ama-1syn/HA synthetic gene product fully complements the function of the authentic pfama1 gene. A, plasmid pHH1-5ЈsgPfa1/HA was constructed by ligating the synthetic gene expression cassette comprising ama-1synthHA (black) flanked by the pfama1 5Ј flanking sequence containing the promoter (cross hatch) and the P. berghei dihydrofolate reductase-thymidylate synthase 3Ј-untranslated region (light gray dots), downstream of the targeting sequence, 0.9 kb of the authentic pfama1 coding sequence (white). The selectable marker cassette (hDHFR) is indicated (dark gray). Sizes are not to scale. B, the predicted structure of the modified locus following plasmid integration and the position of the BglII and AflIII restriction sites used to map this locus are indicated. Note that this single cross-over strategy reconstitutes a promoterless pfama1 gene, lacking the 5Ј-start site, downstream of the site of integration. C, Southern blot analysis of genomic DNA from parental 3D7 parasites and transgenic clones. DNA was restricted with BglII and AflIII and probed with a 0.5-kb fragment from within the 0.9-kb pfama1 target sequence. The resulting hybridization pattern confirms integration of pHH1-5ЈsgPfa1/HA into the pfama1 gene locus as predicted in all four clones analyzed.

The mAb 4G2 Epitope Localizes Entirely to the Domain II
Loop-We have previously demonstrated (29) by recombinant expression of truncated forms of the PfAMA1 ectodomain that mAb 4G2 recognizes an epitope within a structure requiring the combination of domains I and II (DI-II). In that preliminary work, predicted solvent accessibility of individual residues within DI-II was calculated using computer algorithms, and residues predicted to be surface-exposed were substituted to determine their contribution to the epitope. Residues identified as components of the epitope were mapped onto the modeled PfAMA1 structure and found to cluster in a region referred to as the domain II loop.
In light of the above genetic data validating the use of FVO ama-1syn for further functional and structural studies, we set out to use this gene to extend these epitope-mapping studies, taking advantage of the high resolution x-ray crystal structural data now available for PvAMA1 and PfAMA1. The footprint described by a Fab binding to its antigen is ϳ700 Å 2 (60), so a series of gene mutants were produced in which codons for all surface-exposed residues within ϳ25 Å of those previously identified as being part of the 4G2 epitope were modified by site-directed mutagenesis. Mutations took the form of point substitutions, substitutions of groups of 2-4 consecutive residues, or short deletions. In total, over 80 different residues in DI-II were modified. Substitutions were usually with Ala, but existing Ala residues were substituted with a residue known to form the equivalent position in AMA1 from a different species of Plasmodium. Each mutant was then examined for expression, correct folding, and reactivity with 4G2. To do this, individual mutants were expressed transiently at the surface of COS-7 cells, then cell extracts were analyzed by Western blot, probing with mAb 4G2 as well as two polyclonal sera: these were antiserum N5Ads, which recognizes only conformational epitopes in the PfAMA1 ectodomain and acted as an indicator of the degree to which each mutant adopted a correctly folded conformation, and antiserum R5, which recognizes non-conformational epitopes in the PfAMA1 ectodomain and which was used to provide an indication of the expression level of each mutant irrespective of its conformation, as well as to normalize loadings on Western blots (Fig. 5B). Where substitution or deletion of a contiguous group of residues resulted in an effect of 4G2 binding, these residues were then individually examined by point substitution of each. In most cases (see below) this resulted in the identification of one or more critical individual residues required for 4G2 binding. Fig. 6B shows a selection of results from the mutant screen (see supplemental Table S1 for a complete description of all the mutants). Most mutants (examples shown in Fig. 6B are 227DND/AAA, 269KRN/AAA, 305KNL/AAA, and 359DYE/ AAA) displayed levels of 4G2 and N5Ads reactivity equivalent  to the wild-type protein. We conclude that the substituted residues do not form an essential part of the 4G2 epitope, nor do they play a crucial structural role within the molecule.
Several mutations, 110WTE/AAA, 115AKYD/EAAA, 233NYK/AAA, and the point mutations 280K/A and 348D/A, consistently reduced recognition by 4G2. However all of these substitutions also reduced reactivity with N5Ads ( Fig. 6B and Table S1), indicating some loss of native structure and suggesting that the residues involved have a role in maintaining the overall architecture of the protein. In the case of the triplet mutations, it is possible that only one residue in each triplet may be responsible for this loss, but this possibility was not further explored. The side chain of the domain I residue Lys 280 is buried in the crystal structure and hydrogen bonds via its amino group with the main chain atoms of Asn 338 in domain II (29). Both positions are highly conserved in AMA1 across Plasmodium species, suggesting that these residues play an important structural role at the domain I-domain II interface. The side chain of Asp 348 is also buried, and hydrogen bonds with the side-chain atoms of two serine residues, one in domain I (Ser 125 ) the other in the domain II loop (Ser 392 ). Asp 348 , which itself lies in the domain II loop, may therefore also have an important stabilizing role, perhaps anchoring the loop to the surface of domain I.
The triplet mutation 354EQH/AAA within the domain II loop resulted in complete loss of 4G2 recognition, without any reduction in expression levels or any discernible effect on global fold (Fig. 6B). Paradoxically, however, individual substitution of these three residues to Ala had no effect on 4G2 recognition in each case (not shown). Much of the domain II loop (40 residues) displays no electron density in the PvAMA1 x-ray structure but is almost complete in the crystal structure of PfAMA1, which  lacks only the segment Gly 383 -Lys 386 . Residues Glu 354 -His 356 form a surface-exposed hydrophilic segment, and our results suggest that, although single alanine substitutions can be tolerated, the triple alanine substitution induces a structural change, perhaps because of the increased hydrophobic character of the 354EQH/AAA mutation. We produced an additional point mutation, 356H/P, into this segment designed to impose a conformational constraint on the main-chain structure. This mutant exhibited a complete absence of 4G2 recognition (Fig.  6B). To further explore this issue, deletion mutations were designed to remove significant sections of the domain II loop. While deletions Tyr 353 -Ser 377 , Tyr 353 -Lys 368 , Leu 357 -Lys 368 , and Glu 361 -Lys 368 all ablated 4G2 binding without compromising the global fold of PfAMA1, mutagenesis of these residues to alanine (either individually or in groups; see supplemental Table S1) led to no discernible loss of mAb recognition. On the other hand, deletion of a large central section within the domain II loop (deletion Asp 373 -Ser 377 ) had no effect on 4G2 recognition. These results support the notion that the overall conformation of the domain II loop is critical for 4G2 binding.
Only five additional substitutions were identified (351K/A, 352Q/A, 385F/A, 388D/A, and 389R/A; Fig. 6B) that specifically ablated recognition of PfAMA1 by mAb 4G2 without affecting the overall architecture of the protein. The residues involved (Lys 351 , Gln 352 , Phe 385 , Asp 388 , and Arg 389 ) can therefore be considered to be essential components of the 4G2 epitope. All these residues cluster around the base of the domain II loop in the PfAMA1 ectodomain (Fig. 7). Despite substitution of all the surrounding surface residues (Fig. 7), including spatially adjacent residues derived from domain I (Arg 128 , Glu 133 , Asp 134 , Glu 136 , Ser 146 , Lys 148 , Tyr 262 , Asn 264 , and Lys 265 ), no additional residues were identified as contributing to 4G2 binding. Our findings conclusively demonstrate that the 4G2 epitope is formed by residues lying within the domain II loop, with no direct contributions from any residues within domain I. Moreover, critical residues are confined to the base of the domain II loop at both its N-and C-terminal regions, because deletion of a large central segment (residues Asp 373 -Ser 377 ) does not lead to loss of 4G2 binding. . B, domain coloring is as in A except that residues analyzed by site-directed mutagenesis are red (required for 4G2 binding: Lys 351 , Gln 352 , Phe 385 , Asp 388 , and Arg 389 are visible) or lilac (no effect on 4G2 binding). The only residue so far implicated to be in direct contact with mAb 1F9 is the polymorphic site 197, which is shown in cyan. The epitopes recognized by mAbs 4G2 and 1F9 are seen to be on opposite sides of the molecule. The figures were created using PyMOL.

DISCUSSION
In this study we have examined in detail the invasion-inhibitory capacity of mAb 4G2 and the structural requirements for its recognition of PfAMA1. Purified Fab fragments of mAb 4G2 were considerably more potent in invasion-inhibition assays, relative to that of the intact antibody. Similar results were reported by Dutta et al. who examined the invasion-inhibitory activity of impure preparations of 4G2 Fab (43). This result is intriguingly reminiscent of similar findings reported in an early study of invasion-inhibition by a mAb specific for AMA1 of the simian malaria parasite P. knowlesi; in that work, the authors speculated that the enhanced potency of the Fab fragments might be a result of charge effects, removal of the Fc perhaps reducing electrostatic repulsion between the antibody and the negatively charged merozoite surface (61,62). It has been suggested that some invasion-inhibitory antibodies exert their activity by cross-linking AMA1 at the apical tip of the parasite, interfering with its redistribution across the parasite surface and possibly as a result blocking the secretion of other essential invasion-related proteins from the micronemes and/or rhoptries (44). Purified Fab fragments are unable to mediate cross-linking, so our data prove that this is not the primary mechanism of inhibition by mAb 4G2, although it cannot be ruled out as a possible mechanism for other invasion-inhibitory antibodies. Given the relatively small mass (ϳ50 kDa) of Fab, our findings suggest that mAb 4G2 interferes with the function of a critical region of the molecule. This region may be involved in identifying and capturing some host erythrocyte receptor molecule or may interact with a parasite-derived partner protein such as RON2 or RON4 (34).
Using an allelic replacement approach we next showed that the ama-1synth gene is correctly trafficked, properly proteolytically processed, and fully functional in P. falciparum. This result proves that the known primary sequence differences between the predicted ama-1synth gene product and authentic PfAMA1 do not substantially impact on its function. In addition, the result is entirely consistent with our previous data showing that parasite-derived PfAMA1 is not modified by glycosylation (39). The ama-1synth gene lacks potential N-glycosylation sites, so we can now unambiguously infer that N-glycosylation is not a requirement for AMA1 function. This is an important conclusion given the widespread role of glycosyl moieties in pathogen-host surface protein interactions.
Encouraged by our functional validation of the ama1-synth gene, we went on to use it to explore in detail the spatial distribution of surface residues required for 4G2 binding. Our results indicate that the 4G2 epitope is primarily formed by a stretch of residues that lie within the domain II loop of PfAMA1. We cannot rule out the possibility that additional residues may contribute weakly to interactions with mAb 4G2, the substitution of which may have no detectable effect in Western blot. However, our results show that no residues in domain I are directly involved in 4G2 recognition. The PvAMA1 and PfAMA1 crystal structures show that there are extensive stabilizing interactions between domain I and domain II, and correct folding of both domains is likely mutually dependent upon the presence of the other. Consistent with this, we have previously found that both domains are required to produce a minimal structure supporting 4G2 recognition (29). The intimacy of the interactions is further supported by our observation here that a mutation resulting in loss of stabilization of the domain II loop on the surface of DI (348D/A) abrogates 4G2 recognition.
The domain II loop lies against the non-polymorphic face of the molecule (45). Although two residues of this loop (Lys 357 and Phe 367 ) form part of a hydrophobic trough in domain I that is surrounded by polymorphic residues, the epitope recognized by 4G2 lies some distance from these polymorphic residues. That polymorphic residues in PfAMA1 are the targets of protective immune responses is supported by both in vitro and field studies (63,64). For this reason, it has previously been predicted that epitopes recognized by invasion-inhibitory Abs are most likely to reside within regions of high sequence polymorphism, as has been demonstrated for mAb 1F9 (46). In support of this, the epitope recognized by mAb 1F9 lies on the opposite face of the molecule to that recognized by 4G2 (Fig. 7). Here we have demonstrated that at least one invasion-inhibitory epitope of PfAMA1 lies within a highly conserved region of the molecule, providing encouragement for the possibility of producing subunit PfAMA1-based vaccines, which have the capacity to induce strain-transcending protective antibody responses. A recent investigation using phage display of short peptides to study a protective mAb called 45B1 that recognizes the Plasmodium yoelii AMA1 (but does not recognize PfAMA1) similarly mapped its epitope to the domain II loop (65), further supporting a critical function for this region of the molecule. It will be important in future work to establish whether there is any clustering of epitopes recognized by other PfAMA1-reactive invasion-inhibitory mAbs. Our extensive library of PfAMA1 mutants will prove invaluable for this.
The precise role played by AMA1 in invasion is unknown. Attempts to explore its erythrocyte-binding capabilities have provided contradictory results, with the identification of either DI-II of P. yoelii AMA1 (32) or domain III of PfAMA1 (33) being critical for this function. This raises the possibility that either these molecules function differently in different species of Plasmodium or, more likely, that these results are at best incomplete. In binding assays using human erythrocytes, the shed forms of PfAMA1 obtained from P. falciparum in vitro cultures failed to exhibit any binding capacity (39). Furthermore, extensive attempts by us to demonstrate erythrocyte binding using PfAMA1 expressed on COS-7 cells have been unsuccessful (data not shown), despite the fact that, as described in the current study, the FVO ama-1syn gene is fully functional in P. falciparum. Despite the questions raised by these studies, there is strong evidence suggesting that AMA1 is involved in host cell specificity (27,34) in parasite reorientation and in tight junction formation. One possible mechanism by which mAb 4G2 interferes with PfAMA1 function is in preventing interactions with partner proteins such as RON2 and RON4, necessary for recruitment into the moving junction at invasion. The finding that the invasion-inhibitory mAbs 4G2 and 1F9 bind opposite faces of PfAMA1 raises the possibility that the AMA1 ectodomain accommodates two separate functional domains. These may support binding to two distinct parasite molecules, or PfAMA1 may form a bridge binding to a parasite protein via one face and an erythrocyte receptor (or a distinct parasite partner) via the other.
In summary, this study is the first complete characterization of a conserved PfAMA1 epitope recognized by an invasioninhibitory mAb. Our work provides an important step in identifying functional moieties within this essential molecule and may lead to the development of low molecular weight subunit polypeptide or peptide components of a vaccine able to induce an efficiently targeted protective immune response to this highly evolved parasite and other apicomplexan pathogens.