Phage-displayed peptides bind to the malarial protein apical membrane antigen-1 and inhibit the merozoite invasion of host erythrocytes.

Apical membrane antigen-1 (AMA1) is a transmembrane protein present on the surface of merozoites that is thought to be involved in the process of parasite invasion of host erythrocytes. Although it is the target of a natural immune response that can inhibit invasion, little is known about the molecular mechanisms by which AMA1 facilitates the invasion process. In an attempt to identify peptides that specifically interact with and block the function of AMA1, a random peptide library displayed on the surface of filamentous phage was panned on recombinant AMA1 from Plasmodium falciparum. Three peptides with affinity for AMA1 were isolated, and characterization of their fine binding specificities indicated that they bind to a similar region on the surface of AMA1. One of these peptides was found to be a potent inhibitor of the invasion of P. falciparum merozoites into human erythrocytes. We propose that this peptide blocks interaction between AMA1 and a ligand on the erythrocyte surface that is involved in a critical step in malarial invasion. The identification and characterization of these peptide inhibitors now permit an evaluation of the essential requirements that are necessary for efficient neutralization of merozoite invasion by blocking AMA1 function.

According to World Health Organization reports, malaria infects 300 -500 million people/year worldwide and causes 2-3 million deaths annually, mainly in children Ͻ5 years of age. Currently, significant efforts are directed toward the development of a vaccine based on recombinant apical membrane antigen-1 (AMA1), 1 a surface-exposed integral membrane protein that is thought to play a crucial role in invasion of eryth-rocytes by malarial parasites (1). Vaccine strategies that target molecules on the surface of the invasive merozoite such as AMA1 are a high priority in the search for an effective malaria vaccine. Our lack of understanding of the molecular mechanisms associated with the invasion process may hinder the achievement of this goal. A comprehensive ultrastructural description has emerged of a highly organized series of steps of attachment, reorientation, and junction formation leading to the complete encapsulation of the parasite within the erythrocyte (2,3). Constituents of organelles at the apical end of the merozoite have been implicated in the cascade of events leading to invasion and post-invasion events (4,5). For example, rhoptries and micronemes, flask-shaped organelles at the apical end of the merozoite, have been implicated in invasion, whereas dense granules, also part of the apical complex, appear to be involved in events immediately following invasion (6,7). Molecules that mediate the invasion process have been found to be located within apical organelles and also on the merozoite surface. Indeed, some molecules such as AMA1 may be initially localized in micronemes and later migrate to the rhoptries (reviewed in Ref. 8). They are then relocated to the merozoite surface around the time of invasion. It was the timing of this redistribution that first suggested a potential role for AMA1 in the invasion process (9 -11).
Evidence that AMA1 plays an important role in invasion comes from vaccine studies in monkey and mouse models, which showed that immunization with either purified or recombinant AMA1 could induce a protective immune response when immunized animals were challenged with the corresponding species of Plasmodium (12)(13)(14)(15). Reports that monoclonal antibodies directed against AMA1 could also inhibit merozoite invasion provided further evidence that AMA1 has a central role in the invasion process (14 -19).
Other important vaccine candidates that have been shown to induce antibodies that inhibit or block merozoite invasion in vitro include the rhoptry-associated proteins RAP1 and RAP2 (20 -22). However, targeted gene disruption studies of the RAP1 gene performed by Baldi and co-workers (23) revealed normal parasite growth and invasion of human erythrocytes in vitro, suggesting that RAP1 does not play a crucial role in merozoite invasion. It has been suggested that the inhibitory activity of anti-RAP1 antibodies is a result of steric hindrance of the invasion process rather than direct inhibition of the function of this protein. In contrast, attempts to "knockout" the Plasmodium falciparum AMA1 (PfAMA1) gene have not been successful (24), suggesting that unlike other apically located proteins, AMA1 is essential for erythrocyte invasion. * This work was supported in part by the Australian Research Council and the National Health and Medical Research Council of Australia, and by travel awards from the Australian Society of Parasitology, the Royal Society, and the Wellcome Trust (United Kingdom) (to A. D.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Although the precise steps involved in merozoite invasion are not well understood, Chitnis and Blackman (25) have put forward some suggestions for the possible roles of various merozoite surface antigens in the overall invasion process. A possible scenario is that merozoite surface protein 1 (MSP-1) mediates the initial attachment of merozoites to the surface of the erythrocyte, a process that may be mediated by relatively low affinity interactions between MSP-1 and components of the erythrocyte membranes. The role of AMA1 may be to facilitate the reorientation of the merozoite after initial attachment so that the apical complex, consisting of rhoptries and micronemes, is closely apposed to the erythrocyte surface (25). It is feasible that AMA1, which gradually redistributes from the apical organelles to the merozoite plasma membrane, might form a concentration gradient on the merozoite surface that could mediate this reorientation of the parasite. Although there is circumstantial evidence for these suppositions, a clearer understanding of the structure and function of AMA1 relies on further detailed molecular studies.
We chose to apply the powerful phage display technology to identify novel peptides with affinity for AMA1. Random peptide libraries displayed on phage have been used to isolate mimotopes against clinically important antibodies (26), peptides that recognize DNA sequences (27), peptides that mimic carbohydrate structures (28), and peptides that target organ-specific molecules (29,30). By panning peptide libraries on the receptors for erythropoietin and thrombopoietin, peptides have been isolated that are able to act as both agonists and antagonists (31,32). After modification by mutagenesis, these peptides were found to perform most of the functions of the native hormones such as receptor binding, dimerization, and downstream signaling leading to biological activity. These peptides exhibit high potency, in some cases as potent as the natural cytokine (32). Furthermore, analysis of peptides selected on natural ligands has provided insights into the natural binding partners of these ligands (33)(34)(35). In view of the broad success of this approach, we panned a 15-residue random peptide library expressed on gene product (GP) III of filamentous phage against the recombinant AMA1 ectodomain. One of the peptides that we have isolated specifically binds to recombinant AMA1 and recognizes the native protein in malarial parasites. Binding of this peptide to AMA1 was found to inhibit the merozoite invasion of host erythrocytes, and alanine scanning has defined a small set of amino acid side chains that are essential for AMA1 binding. These peptides represent defined reagents that will help explore the structure of AMA1 and illuminate its function within the parasite life cycle and may provide lead compounds for future therapies based on inhibition of AMA1 function.

EXPERIMENTAL PROCEDURES
Parasites-The P. falciparum cloned lines 3D7, D10, FAC-8, K1, and HB3 were continuously cultured essentially using the method of Trager and Jensen (36), except that the human serum supplement to the culture medium was substituted with 0.5% Albumax, and the gas conditions were 1% O 2 , 5% CO 2 , and 94% N 2 . Late stage parasites were purified from synchronized cultures on a Percoll cushion (37).
Phage Library Preparation-The 15-mer phage peptide library was kindly provided by George Smith (University of Missouri, Columbia, MO) (38). Phage were amplified by infecting a log-phase culture of Escherichia coli K91 and shaking overnight at 37°C in LB medium containing 25 g/ml tetracycline (39). The supernatant was twice clarified by pelleting the cells at 8000 ϫ g for 15 min, and a 20% volume of a solution of 20% polyethylene glycol 8000 and 2.5 M NaCl was added to precipitate the phage. The sample was incubated on ice at 4°C for at least 2 h before being centrifuged at 10,000 ϫ g for 50 min. The phage pellet was resuspended in 1 ml of phosphate-buffered saline (PBS; 3 mM KCl, 1.5 mM KH 2 PO 4 , 137 mM NaCl, and 8 mM Na 2 HPO 4 , pH 7.5) and stored at Ϫ20°C in 0.02% NaN 3 .
Panning the Phage Library-The panning technique adopted by Parmley and Smith (40) was modified and used to screen the phage peptide library on PfAMA1. Four rounds of panning were performed on E. coli cell-expressed and refolded AMA1 from the 3D7 strain of P. falciparum (41). The wells of a 96-well enzyme-linked immunosorbent assay (ELISA) plate (Maxisorp, Nunc International) were coated with AMA1 (1 g) in 100 l of coating buffer (0.1 M NaHCO 3 , pH 8.5), sealed, and incubated overnight at 4°C. The wells were blocked for at least 2 h at room temperature with 300 l of blocking solution (0.5% bovine serum albumin (BSA) and 0.1 M NaHCO 3 , pH 8.5). Following blocking, the wells were washed three times with PBS. Phage (ϳ10 11 particles) were added to the wells in 100 l of probing solution (0.5% BSA in PBS) and left for 2 h at room temperature with gentle agitation. After incubation, the wells were washed twice in the first round, four times in the second round, and eight times in subsequent rounds of panning with PBS-T (0.5% Tween 20 in PBS) to remove non-binding phage. Phage that bound to PfAMA1 were eluted with 100 l of elution solution (0.1 M glycine HCl, pH 2.2) for 15 min at room temperature and neutralized with 7 l of 2 M Tris. The titer of eluted phage was estimated, and an aliquot of the eluted fraction was used to infect E. coli K91 cells for amplification. The amplified phage was titered, and 10 11 particles were used in the next round of panning.
Phage Titer Determinations-Phage were subjected to serial 10-fold dilutions with 90 l of LB medium and 10 l of phage suspension in a 96-well microtiter plate (Nunc International). To each of the phage dilutions was added 90 l of log-phase E. coli K91 cells, and the mixture was incubated at room temperature for 20 min to allow the phage to infect the E. coli cells. A 50-l aliquot of each dilution was spread onto LB agar plates containing 25 g/ml tetracycline and incubated overnight at 37°C. Phage infection of bacteria confers resistance to tetracycline, and such colonies were counted and expressed as colony-forming units/ml.
Western Blotting-The harvested parasites were diluted in sample buffer (10% glycerol, 63 mM Tris, pH 6.8, 2% SDS, and 0.0025% bromphenol blue), and incubated at 100°C for 5 min. The parasite extracts were then centrifuged for 10 min at high speed to remove insoluble material. 3 g of recombinant PfAMA1 or Plasmodium chabaudi AMA1 (PcAMA1) was diluted in sample buffer and incubated at 100°C for 5 min. Parasite-derived and recombinant material was separated on SDS-polyacrylamide gels (8% acrylamide) under nonreducing conditions. Separated proteins were then transferred to a polyvinylidene difluoride transfer membrane (PVDF-Plus, Millipore Corp., Bedford, MA); and the membrane was blocked overnight in 5% Blotto (5% skim milk powder in PBS), rinsed for 5 min in PBS, and probed with phage (10 10 particles/ml) or a rabbit polyclonal antiserum to PfAMA1 (1:1000 dilution in 5% Blotto) for 1 h at room temperature with gentle agitation. The membrane was washed every 10 min for 4 h with PBS-T. Horseradish peroxidase (HRP)-conjugated anti-M13 IgG (Amersham Biosciences, Quarry Bay, Hong Kong) and HRP-conjugated anti-rabbit IgG (Amersham Biosciences Pty. Ltd.) antibodies were used as secondary antibodies, and binding was detected by chemiluminescence (Pierce).
Microtiter Plate Binding Assays-Binding assays were carried out using a process similar to that described by Harlow and Lane (42). Briefly, 96-well Maxisorp microwell plates were coated with PfAMA1, monoclonal antibody (mAb) 18/2, BSA, or PcAMA1 (all at 1 g in 100 l of 0.1 M Na 2 HCO 3 , pH 8.5, per well) overnight at 4°C. Wells were blocked for at least 2 h at room temperature with 300 l of 0.5% BSA in PBS and then washed three times with PBS. Phage (diluted in 0.5% BSA in PBS) were added to the wells and incubated for 1 h at room temperature. The wells were washed five times with PBS-T, and bound phage were detected with peroxidase-conjugated anti-M13 antibody (1:3000 dilution in PBS) using o-phenylenediamine as a color reagent. For competition experiments, 10 10 phage particles were added to the PfAMA1-coated wells (1 g/well) in the presence of increasing amounts of synthetic peptide or mAb 4G2, and the phage were detected with HRP-conjugated anti-M13 antibody. In binding assays involving the detection of synthetic peptides, 96-well plates were coated with 10 g of synthetic peptide in 100 ml of coating buffer (15 mM Na 2 CO 3 and 35 mM NaHCO 3 , pH 9.6). AMA1 (1 g in 100 l of probing solution) was added to the wells, and bound AMA1 was detected with rabbit polyclonal antiserum raised against AMA1 followed by HRP-conjugated anti-rabbit IgG antibody as described above.
PCR Amplification-The region of the phage genome encoding the displayed peptide sequence was amplified using the following primers: 5Ј-primer (GAT AAA CCG ATA CAA TTA AAG) and 3Ј-primer (CAC AGA CAA CCC TCA TAG). In a 50-l reaction volume, 2 units of Taq polymerase (Promega) was used to amplify 2 l of template phage DNA solution (released from E. coli K91 cells by boiling) using 250 nM primers, 200 M dNTPs, and 2 mM MgSO 4 . After an initial 1-min denaturation, the reaction was cycled at 95°C for 30 s, 45°C for 30 s, and 72°C for 30 s for 30 cycles. A final elongation step was carried out at 72°C for 7 min. The resultant PCR product was purified using the QIAquick 8 PCR purification kit (QIAGEN Pty. Ltd.).
DNA Sequence Analysis-DNA was sequenced by automated dye terminator cycle sequencing (SUPAMAC, Centre for Proteome Research and Gene Product Mapping, Eveleigh, New South Wales, Australia). Sequences were analyzed with DNASIS Version 2.1 computer software (Hitachi Software Engineering Co., Ltd.).
Alanine Scanning Mutagenesis of the F1 Peptide-15 derivatives of the F1 peptide (GWRLLGFGPASSFSM) in which each residue was replaced with L-alanine (alanine was replaced with L-glycine) were synthesized as cleavable pepspots. These 15 mutated peptides were solubilized in Me 2 SO followed by PBS to a final concentration of 4% Me 2 SO and analyzed for binding to PfAMA1 using the competition assay with F1 phage as described above.
Indirect Immunofluorescence Assay-Indirect immunofluorescence microscopy was performed essentially as described previously by Bianco et al. (43) with phage displaying the F1 or F2 peptide as the primary reagents followed by rabbit anti-M13 antibody. After incubation and washing, fluorescein isothiocyanate-labeled anti-rabbit IgG (Sigma) was used in the final detection step.
Peptide Inhibition of Merozoite Invasion of Erythrocytes-The P. falciparum cloned lines 3D7 and HB3 were grown to a parasitemia of ϳ10% late stage (schizont). Following Percoll purification (37), schizonts were mixed with uninfected erythrocytes and aliquoted into microwells containing the test/control solutions. A reference smear was examined and retained (ϳ3-4% initial parasitemia). After ϳ20 h in culture, smears were made to determine the number of invaded erythrocytes (cells containing ring stage parasites). Parasitemias were determined by counting 1000 cells from methanol-fixed, Giemsa-stained thin blood films.

Isolation of AMA1-binding Peptides from a Random Peptide
Library-To identify peptides that have affinity for AMA1, a phage library displaying random 15-residue peptides was panned against immobilized AMA1. A dramatic enrichment of phage with affinity for the antigen was observed after the third round of panning (Fig. 1A). These pools of phage showed no binding to the irrelevant proteins BSA and the ring infected erythrocyte surface antigen (RESA) (Fig. 1B), but did bind to PcAMA1, which shares 52% amino acid sequence identity with PfAMA1.
Individual phage clones were examined for their ability to bind either PfAMA1 or PcAMA1. It is clear that some clones were able to bind only to PfAMA1 ( Fig. 2A, clones 2 , 4 -6, 8, and 9), whereas other clones bound to both PfAMA1 and PcAMA1 (clones 1, 3, and 7). The binding activity was conferred by the displayed peptides because phage lacking a peptide (control (C)) and two phage clones picked at random from the unpanned peptide library (lib1 and lib2) were unable to bind to either PfAMA1 or PcAMA1 ( Fig. 2A). None of the clones examined displayed any binding to the irrelevant RESA protein.
Sequencing the DNA inserts of over 30 phage clones that bound PfAMA1 and translation of the corresponding peptide sequences allowed us to classify all of binding clones into one of three groups (Fig. 2B). The majority of binding clones consisted of the sequence GWRLLGFGPASSFSM (F1 peptide), whereas the remainder consisted of either TRLFRVPVLPSGVTS (F2 peptide) or PFARAPVEHHDVVGL (F3 peptide). Realignment of the latter two sequences revealed a common motif; RX-PVXXXXV, where represents a hydrophobic residue and X represents any residue.
Representative clones from each group were selected, and their binding properties were examined further. Phage displaying a peptide of the F1 class ( Fig. 2A, clones 2, 4 -6, 8, and 9) recognized only PfAMA1 and had no reactivity with PcAMA1. Phage expressing peptides from the F2 and F3 groups ( Fig. 2A,   clones 1, 3, and 7) bound to recombinant PfAMA1 and PcAMA1. Phage clones displaying each of the three peptides bound to PfAMA1 in a dose-dependent manner, although the F1 and F3 peptides appeared to have an ϳ10-fold higher relative affinity compared with the F2 peptide (Fig. 2C). Absolute affinities were difficult to estimate from these data because the presence of up to five copies of peptide on each phage particle may impart avidity effects that are difficult to predict. Phage containing a peptide picked at random from the unpanned library and consisting of the sequence GDVWLFKTSTSHFAR (F5 peptide) were unable to bind to PfAMA1 even at phage concentrations of 10 11 colony-forming units/ml (Fig. 2B).
Phage Displaying the F1 Peptide Recognize Native Antigen Expressed in Parasites-To examine whether the isolated peptides can recognize native as well as recombinant AMA1, we used the peptide-displaying phage as reagents in fluorescence microscopy and Western blot assays. The presence of the phage particle attached to the peptide enabled us to use an anti-phage antibody followed by a secondary antibody conjugated to fluorescein isothiocyanate or HRP to assess the binding of peptides to AMA1, as shown schematically in Fig. 3A. When phage displaying the F1 peptide were incubated on thin blood films of the P. falciparum 3D7 strain, a distinct merozoite apical fluorescence was observed in trophozoite and schizont stage parasites and was indistinguishable from that obtained with rabbit antiserum raised against PfAMA1 (data not shown) and similar to that found previously by other workers (1,44,45). These data are consistent with the reported apical location of AMA1 in mature parasites, followed by a reorganization of AMA1 to the merozoite surface. As expected, when this assay was carried out using phage displaying an irrelevant 15-residue peptide, no fluorescence was seen (data not shown).
Further evidence that the F1 peptide binds specifically to AMA1 was obtained by Western analysis of parasite extracts using F1 peptide-displaying phage as the primary reagents. Two polypeptides of ϳ80 and 60 kDa were strongly recognized by the F1 phage probe (Fig. 3B, right panel). Polypeptides of identical sizes were recognized by rabbit antiserum raised against recombinant AMA1 (Fig. 3B, left panel). The molecular masses of the two polypeptides are in agreement with the previously reported masses of the full-length AMA1 gene product (80 -83 kDa) and the 62-63-kDa processed form (1,44,46). F1 phage and rabbit anti-AMA1 antiserum also recognized a polypeptide of ϳ40 kDa from parasite material, which is likely to result from a secondary processing event within the parasite (47). Both rabbit anti-AMA1 antiserum and F1 phage recognized the E. coli cell-expressed AMA1 ectodomain (Fig. 3B). Consistent with previous ELISA experiments, F1 phage did not bind to PcAMA1, although anti-PfAMA1 polyclonal antiserum did recognize this ortholog (Fig. 3B). F1 phage recognized AMA1 from all parasite strains examined except HB3 (Fig. 3B,  right panel). In contrast, rabbit antiserum raised against PfAMA1 bound to AMA1 from all strains examined, including HB3 (Fig. 3B, left panel). No binding of F1 phage to AMA1 was observed when Western blotting was carried out under reducing conditions (data not shown), indicating that the F1 peptide binds to a conformation-dependent epitope on the intramolecular disulfide bonds in AMA1. Examination of the deduced amino acid sequence of AMA1 from all strains revealed only seven positions that had a residue unique to HB3 (Fig. 3C). Specificity controls showed that phage expressing an irrelevant peptide (F5) did not bind to recombinant AMA1 or AMA1 from parasites as determined by Western analysis (data not shown).
Synthetic Peptides Bind to AMA1 in Close Proximity to Each Other-The displayed peptides are fused to the N terminus of GPIII (38,48,49), and it is conceivable that GPIII could influence the AMA1-binding characteristics of the phage-displayed peptide. To address this question, the F1, F2, and F3 peptides were synthesized, and their respective AMA1-binding qualities were assessed. When the F1 peptide was immobilized on wells of a microtiter plate, it was able to capture AMA1 from solution as determined by secondary capture of anti-AMA1 antibody (Fig. 4A). By contrast, wells coated with a peptide consisting of a scrambled F1 sequence (F1(s); AMSPWFRSLGFGSLG) did not capture AMA1 (Fig. 4A). The F2 and F3 peptides were also able to capture PfAMA1 in this assay (data not shown). This demonstrates that sufficient information for binding AMA1 is contained within the peptide sequences identified by panning and that the phage framework plays a negligible role in the binding affinity.
To explore further the ability of the three peptides to bind AMA1, a competition assay using F1 phage as the capture moiety was performed. As expected, the F1 peptide in solution was able to inhibit the binding of F1 peptide-displaying phage to AMA1 almost completely, with an IC 50 of 100 nM (Fig. 4B). The importance of the linear sequence of the F1 peptide in conferring AMA1 binding was evidenced by the inability of the scrambled peptide to inhibit binding (Fig. 4B). Importantly, synthetic peptides corresponding to the F2 and F3 AMA1- FIG. 2. A, fine specificity of individual phage after panning on AMA1. 16 phage clones were isolated, and each clone was examined for binding to PfAMA1 (black bars), PcAMA1 (gray bars), or RESA (white bars). The binding specificity of helper phage lacking a displayed foreign peptide (control (C)) as well as two clones picked at random from the unpanned library (lib1 and lib2) was also examined. B, deduced amino acid sequences of phage peptides that bind AMA1. Shading indicates amino acid residues common between the F2 and F3 peptides. C, binding of a representative clone from each sequence to PfAMA1 with increasing phage concentration. The binding of phage expressing an irrelevant peptide (F5) to PfAMA1 was also examined.

FIG. 3. The phage-displayed F1 peptide binds to native AMA1.
A, shown is a schematic of the assay used to detect the binding of phage displaying the F1 peptide to AMA1 expressed in parasites cultured in vitro. The secondary or tertiary antibody (Ab) was conjugated with either fluorescein isothiocyanate (FITC) or HRP for immunofluorescence and Western blot experiments. B, phage displaying the F1 peptide were incubated with nylon filters on which were immobilized mature parasite extracts after SDS-PAGE. Five parasite strains were used (HB3, K1, FAC-8, D10, and 3D7), and recombinant PfAMA1 (Pf) and PcAMA1 (Pc) were also included. Bound phage were detected using HRP-conjugated anti-phage antibodies (right panel). Similar blots were probed with antibodies to PfAMA1, and binding was detected using HRP-conjugated anti-rabbit secondary antibody (left panel). C, shown is a schematic of the locations of polymorphisms in the PfAMA1 ectodomain that are unique to HB3. binding sequences were able to inhibit the interaction between F1 phage and AMA1, albeit with a lower apparent affinity (IC 50 ϭ 100 and 10 M, respectively) (Fig. 4B). Thus, although the three peptides have very different sequences and there is no obvious homology between the F1 peptide and the other two, they appear to be able to bind to a similar region on the AMA1 surface. Clearly, the footprints of the three peptides, although not identical, do overlap sufficiently to allow cross-competition.
Critical Binding Residues Revealed by Alanine Scanning-In an effort to identify amino acids within the F1 peptide that are critical for binding to AMA1, we performed an alanine scan of the F1 sequence. The extent of AMA1 binding by peptides with each residue in turn replaced with alanine indicates that residues 5-9 are important for binding. When any of these residues (LGFGP) were replaced with alanine, the binding of the resulting peptide to AMA1 was dramatically reduced, as assessed by the inability of these peptides to inhibit authentic F1 peptide-displaying phage from binding to immobilized AMA1 (Fig. 5). In contrast, substitution of residues N-or C-terminal to this central motif had no effect on the ability of the phage-displayed F1 peptide to bind to AMA1. To confirm that the C-terminal five residues (SSFSM) are not required for AMA1 binding, the binding to AMA1 and the invasion inhibitory activity of a truncated F1 peptide lacking the last five residues were determined. As predicted, the binding of this 10-residue peptide to AMA1 was virtually indistinguishable from that of the full-length F1 peptide (data not shown).
Synthetic Peptides Can Inhibit P. falciparum Merozoite Invasion-It has been reported that AMA1-reactive mAb 4G2 is an efficient inhibitor of P. falciparum merozoite invasion of host erythrocytes (18) and Plasmodium reichenowi (45). To determine whether this inhibitory monoclonal antibody and the F1 peptide bind to a similar region on the AMA1 surface, phage displaying the F1 peptide were incubated with immobilized recombinant AMA1 in the presence of increasing concentrations of mAb 4G2. In this assay, mAb 4G2 was able to inhibit the binding of F1 phage to AMA1 in a dose-dependent manner, and the extent of this inhibition was similar to that produced when soluble F1 peptide was included in the assay (Fig. 6A). To exclude the possibility that mAb 4G2 inhibits binding of the F1 peptide by steric hindrance, we examined the ability of soluble F1 peptide to block the binding of mAb 4G2 to AMA1. Fig. 6B demonstrates that, of the peptides tested, only the F1 peptide was able to inhibit mAb 4G2 binding to any extent. It is therefore likely that both mAb 4G2 and the F1 peptide bind to a similar (if not identical) site on AMA1.
Because these results raised the possibility that the peptides may block the merozoite invasion of host erythrocytes, we assessed the invasion efficiencies of P. falciparum parasites cultured in the presence of the corresponding synthetic peptides. 25 g/ml F1 peptide resulted in ϳ50% inhibition of invasion, whereas 50 g/ml F1 peptide showed close to 90% inhibition (Fig. 7A). By contrast, the F2 and F3 peptides were much less effective inhibitors of invasion, requiring 10-fold higher concentrations to produce an effect (Fig. 7A), with the F2 peptide being a more efficient inhibitor of invasion than the F3 peptide. These experiments were performed on a number of occasions with slight variations in parasitemias and hematocrits; however, the dose-dependent trend was always consistent, with the F1 peptide being more active at lower concentrations than the F2 or F3 peptide. The synthetic peptide corresponding to the scrambled sequence of the F1 peptide (F1(s)) showed little inhibitory activity even at a concentration FIG. 4. The synthetic F1 peptide binds to AMA1. A, synthetic peptides consisting of the F1 sequence and the scrambled F1 sequence (F1(s)) were immobilized on wells of a microtiter plate and incubated with PfAMA1. Binding of AMA1 to the immobilized peptides was detected with rabbit anti-PfAMA1 antibody, followed by incubation with HRP-conjugated anti-rabbit IgG in an ELISA format. Data are the means Ϯ individual values of duplicate measurements. B, competition phage ELISA was carried out to determine the ability of various synthetic peptides to compete with phage displaying the F1 peptide for binding to PfAMA1. Phage binding was detected by HRP-conjugated anti-phage antibodies.
FIG . 5. Alanine scan of the F1 sequence. 15 peptides (pep1-pep15) corresponding to the F1 sequence but with the systematic replacement of each residue with alanine were synthesized. The residue replaced in each peptide is shown above the peptide on the histogram. Peptides were incubated with F1 phage and added to wells of a microtiter dish with immobilized AMA1. Binding of F1 phage was detected as described in the Fig. 4 legend. The effect of each mutated peptide was compared with that of the control parental F1 peptide, which abolished binding of phage, and the F1(s) peptide, which had no effect on phage binding. If either AMA1 (no AMA-1) or F1 peptide-displaying phage (no phage) was omitted from the assay, there was no detectable binding. of 500 g/ml. In addition, two irrelevant synthetic 15-mer peptides (P1, CFDYAPYVSAVDDIC; and P2, GWLSPSWFEP-GLASM) were found to have little effect on merozoite invasion at similar concentrations (Fig. 7A). A peptide corresponding to a scrambled version of the F2 sequence (F2(s), VDAPHVF-GVPHRLEA) also showed little inhibitory activity at 500 g/ml (data not shown). Significantly, the parasites that were able to invade despite the presence of inhibitory peptide appeared to develop normally and progressed from ring trophozoite through schizogony normally. Further evidence for the specificity of the mechanism of inhibition was obtained by noting that when the F1, F2, and F3 peptides were added to parasite cultures immediately after invasion had occurred, no observable effects on parasite development were seen (data not shown), ruling out a general toxic effect of the peptide on the parasitized erythrocytes.
The observation that the F1 peptide did not bind to AMA1 from parasites of the HB3 strain suggested that it would not inhibit the invasion of erythrocytes by HB3 parasites. As predicted, the F1 peptide was unable to block the invasion of HB3 parasites, but did reduce the invasion of 3D7 parasites grown under the same conditions (Fig. 7B). This result strongly sup-ports the proposition that it is the binding of the F1 peptide to AMA1 that is the critical event mediating the inhibition of merozoite invasion of host erythrocytes.

DISCUSSION
In this study, we have identified a set of peptides (from a random peptide library) that bind to AMA1. Low micromolar concentrations of one of these peptides effectively blocked the merozoite invasion of human erythrocytes in vitro. The three peptides with affinity for P. falciparum AMA1 were obtained by panning a phage-displayed library containing hundreds of millions of peptides on bacterially expressed, refolded AMA1, followed by deconvolution of the final round pool. Interestingly, although they were panned on PfAMA1, two of these peptides (F2 and F3) were also able to recognize recombinant AMA1 from the rodent malaria P. chabaudi, thus distinguishing them from the F1 peptide, which recognized only PfAMA1. These peptides did not recognize a variety of other proteins (Figs. 1B and 3C). The punctate pattern obtained with the F1 peptide was indistinguishable from that obtained with anti-AMA1 antiserum when used in fluorescence microscopic analysis to probe the location of AMA1 in schizont-infected erythrocytes. Moreover, the peptides were specifically reactive with AMA1 in Western blots of asexual parasite culture material subjected to electrophoresis under nonreducing conditions. We did not ob- falciparum parasites in vitro, and invasion was assessed by counting newly formed ring stage parasites. Of the peptides tested, F1 was the most efficient at inhibiting invasion. At higher concentrations of peptide, F2 was also able to inhibit invasion; however, the other peptides tested were not found to have any inhibitory activity. All peptide concentrations are given in micrograms/ml. B, the F1 peptide inhibited the 3D7 strain of P. falciparum, but not the HB3 strain. The F1(s) peptide showed no inhibitory effect on either strain of parasites. Data are the means Ϯ individual values of duplicate measurements. serve binding of F1 peptide-displaying phage to AMA1 upon Western blotting carried out under reducing conditions, indicating that the F1 peptide recognizes a binding site that is dependent on the intramolecular disulfide bonds in AMA1. Thus, although the F1 peptide was isolated by panning on recombinant protein, it is capable of recognizing authentic, parasite-derived AMA1.
The observation that the F2 and F3 peptides bound to PcAMA1 (DS strain) and that the F1 peptide bound only to PfAMA1 suggests that the F1 peptide makes different molecular contacts with AMA1 or binds to a different location on AMA1 than the F2 or F3 peptide. There is 52% amino acid identity between AMA1 from these two species, and the 16 cysteine residues present are absolutely conserved in both polypeptides. It therefore seems reasonable to postulate that these molecules share a similar folded structure and that the F2 and F3 peptides bind to a common feature in PfAMA1 and PcAMA1. It may therefore be expected that panning on AMA1 from one source will identify not only peptides that are specific for AMA1 from that species (e.g. F1), but also peptides that react more broadly across AMA1 molecules from different species (e.g. F2 and F3). When all three soluble synthetic peptides were examined for their ability to inhibit F1 phage from binding to AMA1, the F1 peptide was the most potent inhibitor; however, the F2 and F3 peptides were both able to inhibit the binding of phage displaying the F1 peptide. Taken together, these data suggest that despite the sequence diversity, all three peptides bind in close proximity on the AMA1 polypeptide, possibly making overlapping (but not identical) molecular contacts with the surface of the protein.
Phage displaying the F1 peptide proved to be robust reagents in both fluorescence microscopy and blot assays, giving patterns comparable to those observed using serum from a rabbit immunized with purified AMA1 (Fig. 3). This suggests that phage-displayed peptide libraries may be a source of affinity reagents that can be assessed rapidly without the need for animal immunization. The surface features of the binding site on AMA that makes contact with the F1 peptide appear to be present on AMA1 molecules from most strains tested in this study, but interestingly, are absent in AMA1 from HB3 parasites. Analysis of the sequences of AMA1 from the different strains used in this assay revealed that there are only seven positions that are unique to HB3. These polymorphisms are clustered at the N and C termini of AMA1, and four of the seven polymorphisms result in changes of charge, suggesting that the residues at these positions could have a large influence on the binding energy of the peptide. Mutational analysis could be used to define the relative contributions of each of these seven residues to the binding site of the F1 peptide. If it is assumed that more than one of these residues is involved in forming the F1 peptide-binding site, then the distribution of these residues along the AMA1 sequence implies that the binding site is formed by regions of AMA1 that are distant in the primary sequence, but brought into close proximity in the folded structure. This is consistent with the observation that F1 peptidedisplaying phage were unable to bind AMA1 that had been treated with a reducing agent prior to SDS-PAGE and Western blotting.
It was not possible to identify a motif responsible for AMA1 binding by comparison with other peptide sequences, as F1 was the only peptide isolated that bound solely to PfAMA1. To address the possibility of a subdomain or motif contained within the F1 peptide, alanine scanning of the whole peptide sequence was performed. The small size of the F1 peptide makes it particularly amenable for assessing how specific mutations affect AMA1 binding. Alanine replacement at each po-sition in the central LGFGP sequence reduced AMA1 binding compared with the wild-type F1 sequence. The mutants that had the greatest effect on binding activity were G6A and F7A, indicating that these residues may be critical for binding. In the absence of structural information, it is difficult to conclude whether these residues contact AMA1 directly or are important in maintaining the peptide in the correctly folded state. It is likely, however, that the phenylalanine at position 7 in the F1 peptide binds to a hydrophobic pocket on AMA1.
It is interesting to note that the central residues FGP in the F1 peptide are also present in a peptide described by Wrighton et al. (31) that is able to interact with the erythropoietin receptor. Structural studies on this peptide, which acts as a dimer to stimulate erythropoiesis, revealed that the GP dipeptide forms a ␤-turn on the peptide backbone. It was noticed that the residues in this ␤-turn (the GP dipeptide and the adjacent leucine) made several hydrogen bond contacts with the receptor and are important for the overall binding activity. Although the F1 peptide and the peptides described by Wrighton et al. (31) are clearly different, it might be predicted that the GP dipeptide in F1 induces a turn that is important for binding to AMA1 and ultimately in inhibiting merozoite invasion.
The observation that the F1 peptide did not bind to AMA1 from HB3 parasites and was incapable of inhibiting the invasion of erythrocytes by HB3 merozoites, together with the identification of residues on the F1 peptide that are important for AMA1 binding, provides the basis of examining the structure and function of AMA1 in molecular detail. Besides mutating the amino acid residues that are unique to AMA1 from HB3, it is also possible to create libraries of F1 peptides with mutations flanking the conserved LGFGP region to improve the affinity of interaction with AMA1. Thus, libraries of different peptide sequences can, for example, be panned to isolate peptides with higher affinity binding to AMA1 from 3D7 as well as peptides that bind to AMA1 from HB3. Sequence information from these peptides coupled with an investigation of whether these peptides inhibit merozoite invasion will enable a delineation of the features necessary for inhibition of invasion due to inactivation of AMA1 and possible rational design of a non-peptide inhibitor of the invasion process.
Examination of the primary sequence of the F2 and F3 peptides revealed a potential common motif. The core of this motif consists of an arginine followed by a small hydrophobic residue (either alanine or valine) and then proline and valine. There is also a valine at a similar position in both peptides several residues C-terminal to this cluster. Furthermore, the two positions immediately preceding the arginine are hydrophobic in both the F2 and F3 peptides. This RXPVXXXXV motif is predicted to be important for the binding of these peptides to AMA1 and may explain why both the F2 and F3 peptides are able to cross-react with AMA1 from different species. It is tempting to postulate that this F2/F3 peptide-binding site common to AMA1 from different parasite species is located close to the site on PfAMA1 occupied by the F1 peptide. This is evidenced by the ability of F2 and F3 to prevent F1 from reacting with PfAMA1.
Although the peptides isolated in this study are unlikely to be therapeutic agents in themselves, they do provide a set of tools with which to probe the structure and function of AMA1. Identification of important functional regions of AMA1 will enhance the possibility of developing "second generation" vaccines based on domains or subdomains of AMA1 rather than on the highly disulfide-bonded ectodomain. Furthermore, the interactions of the chemical groups on these peptides may provide a starting point for the screening of non-peptide drugs by, for example, the method recently described by Qureshi et al.