Malaria Vaccine-Related Benefits of a Single Protein Comprising Plasmodium falciparum Apical Membrane Antigen 1 Domains I and II Fused to a Modified Form of the 19-Kilodalton C-Terminal Fragment of Merozoite Surface Protein 1

Abstract

We show that the smallest module of Plasmodium falciparum AMA1 (PfAMA1) that can be expressed in the yeast Pichia pastoris while retaining the capacity to induce high levels of parasite-inhibitory antibodies comprises domains I and II. Based on this, two fusion proteins, differing in the order of the modules, were developed. Each comprised one module of PfAMA1 (FVO strain, amino acids [aa] 97 to 442) (module A) and one module of PfMSP1₁₉ (Wellcome strain, aa 1526 to 1621) (module Mm) in which a cystine had been removed to improve immune responses. Both fusion proteins retained the antigenicity of each component and yielded over 30 mg/liter purified protein under fed-batch fermentation. Rabbits immunized with purified fusion proteins MmA and AMm had up to eightfold-higher immune responses to MSP1₁₉ than those of rabbits immunized with module Mm alone or Mm mixed with module A. In terms of parasite growth inhibition, fusion did not diminish the induction of inhibitory antibodies compared with immunization with module A alone or module A mixed with module Mm, and fusion outperformed antibodies induced by immunization with module M or Mm alone. When tested against parasites expressing AMA1 heterologous to the immunogen, antibodies to the fusion proteins inhibited parasite growth to a greater extent than did antibodies either to the individual antigens or to the mixture. These results suggest that compared with the individual modules delivered separately or as a mixture, fusion proteins containing these two modules offer the potential for significant vaccine-related advantages in terms of ease of production, immunogenicity, and functionality.

The annual malaria burden of 300 to 500 million clinical cases results in an estimated mortality for up to 2 million people, predominantly sub-Saharan African children under 5 years of age (52). A malaria vaccine would make a significant contribution to reducing the enormous socioeconomic burden caused by this disease. A number of vaccine approaches, targeting various stages of the complex parasite life cycle, are being investigated (21). Apical membrane antigen 1 (AMA1) and merozoite surface protein 1 (MSP1) are potential vaccine components, and a number of vaccines using elements of these molecules are currently in early clinical evaluation. Previous research has indicated that a combination of MSP1 and AMA1 has vaccine-related advantages over either antigen alone (3, 55). Both molecules are essential components of the asexual blood-stage merozoite (50, 60), the developmental stage of the parasite stage responsible for invasion of erythrocytes. They are also both present on merozoites that emerge from infected liver cells, and AMA1 has also been identified as a sporozoite protein (51). Plasmodium falciparum AMA1 (PfAMA1) is a polymorphic protein; over 10% of its amino acid residues can change without obvious effects on its function in invasion. With few exceptions, polymorphic residues are bi- or trimorphic, and all are located on the outside of the molecule, predominantly on one face (47). One strategy to tackle any potential negative effect of polymorphism in vaccine development is to combine PfAMA1 with other targets that are not, or are less, polymorphic, such as MSP1₁₉ (59). A single-protein vaccine has cost, speed, and potential functionality benefits compared with vaccines prepared from mixtures of proteins. We have therefore investigated how minimal elements of AMA1 and MSP1, each retaining the ability to induce growth-inhibitory antibodies, can be incorporated into fusion proteins that allow the development of single-protein, multitarget malaria vaccines.

Micronemes are organelles of the merozoite apical complex, a structure intimately associated with the invasion of erythrocytes. AMA1 is initially trafficked to micronemes as an 83-kDa type 1 integral membrane protein; subsequently, the N-terminal prodomain is proteolytically cleaved prior to relocalization to the merozoite outer membrane (43). Further cleavage, proximal to the transmembrane region, then releases the ectodomain from the parasite surface (26). AMA1 contains 16 conserved cysteine residues that form eight intramolecular disulfide bonds (20). The recently elucidated three-dimensional structure of AMA1 (47) confirms that after cleavage of the prodomain, the ectodomain essentially comprises three interacting domains (DI, DII, and DIII), as originally proposed based on cystine patterns (19). The immunization of rabbits and mice with PfAMA1 induces high levels of antibodies that inhibit parasite growth in vitro (1, 8, 11, 16, 30, 33). Aotus, rhesus, and squirrel monkeys have been protected by immunization with PfAMA1, Plasmodium knowlesi AMA1, and Plasmodium fragile AMA1, respectively (6, 9, 55). Humans in areas of endemicity have high circulating titers of anti-AMA1 antibodies (7, 28, 57) that may correlate with protection (49).

MSP1 is initially expressed as an ∼200-kDa molecule linked by a glycosyl phosphatidylinositol anchor to the merozoite surface membrane (reviewed in reference 22). MSP1 is proteolytically cleaved into four fragments that are assembled into a complex with other molecules (23, 25, 29) and held on the surface through the C-terminal 42-kDa fragment (MSP1₄₂). At invasion, the complex is shed from the surface by the action of a parasite protease (a process called secondary processing), except for a 19-kDa C-terminal fragment (MSP1₁₉) that remains on the merozoite surface. Some antibodies that bind to MSP1₁₉ inhibit secondary processing and erythrocyte invasion, whereas others (called blocking antibodies) facilitate invasion in the presence of inhibitory antibodies. Following elucidation of the structure of MSP1₁₉, the epitopes of several specific antibodies have been investigated using a number of methods, including site-directed mutagenesis, to change individual and multiple amino acids (10, 61). One such modified MSP1₁₉ (C12I, C28W) lacks one disulfide bond and at least one epitope for blocking antibodies and shows improved immunogenicity (17). Other structural features of this variant are similar to those of the native molecule (W. D. Morgan, unpublished results).

P. falciparum MSP1 (PfMSP1) and, in particular, the C-terminal regions MSP1₄₂ and MSP1₁₉ (54) are leading blood-stage vaccine candidates. Immunization with full-length PfMSP1, MSP1₄₂, and MSP1₁₉ protects mice from an otherwise lethal infection of Plasmodium yoelii (24, 58). This protection is largely antibody mediated (18, 36). Aotus monkeys were protected from the virulent FVO strain of P. falciparum by immunization with MSP1₁₉ (12, 15, 34). MSP1₁₉-specific antibodies confer passive immunity in rodent models (4, 53) and inhibit merozoite invasion in vitro (2). Human populations in regions where malaria is endemic have strong anti-MSP1₁₉ antibody responses, and these antibodies are a major component of the parasite-inhibitory activity of the serum (44).

For many proteins, and in particular those in which disulfide bonds are critical to function, eukaryotic expression systems such as yeasts offer several advantages. We and others (33, 40) have successfully used the Pichia pastoris expression system to produce clinical-grade material for malaria vaccine trials. We wished to design and produce a single protein containing modules from more than one malaria vaccine target molecule by using P. pastoris. The largest heterologous protein that has been expressed successfully in P. pastoris is 170 kDa (5); this suggests that producing the protein should be feasible. With a view to ultimately developing a fusion protein effective against a broad range of parasite strains, by inclusion of protein encoded by selected AMA1 and MSP1₁₉ alleles, we report here on studies to determine the smallest region of PfAMA1 that retains the ability to induce high-titer parasite inhibitory antibodies and we demonstrate vaccine-related benefits of the fusion of this region with MSP1₁₉.

MATERIALS AND METHODS

Parasites.

P. falciparum lines 3D7, FCR3, and HB3 were maintained in vitro using standard culture techniques in an atmosphere of 5% CO₂, 5% O₂, and 90% N₂. Parasite line identity was verified by AMA1 gene sequence analysis. Apart from alterations made to remove potential N-glycosylation sites (see below), FCR3 AMA1 (GenBank accession no. M34553) differs by only one amino acid from the FVO AMA1 molecule used in these studies (GenBank accession no. AJ277646). 3D7 AMA1 (the GenBank accession no. for the 3D7 clone of NF54 is U33274) and HB3 AMA1 (GenBank accession no. U33277) differ at 29 and 22 amino acid positions, respectively, from FVO.

There are two common types of MSP1₁₉; the 3D7 sequence (GenBank accession no. CAD51981) is of the “E-TSR” type and differs at four positions from that of the Wellcome “Q-KNG” type used in the fusion protein reported in this study (GenBank accession no. X02919) (59). MSP1₁₉ sequences for FCR3 and HB3 (amplified by PCR) were of the Q-KNG type.

Selection and cloning of AMA1 and MSP1₁₉ modules and fusions for expression in P. pastoris.

In preliminary studies, a series of AMA1 regions based on the domain structure (Table 1) were generated for expression in P. pastoris. DNA was PCR amplified using an FVO AMA1 synthetic gene designed with P. pastoris codon usage and lacking N-glycosylation sites (33). PCR products of ama1 domains were cloned into the EcoRI-XbaI sites of the pPICZαA vector (Invitrogen, Groningen, The Netherlands) and used to transform Escherichia coli DH5α cells. Forward primers comprised an EcoRI restriction site sequence preceded by some overhang and followed by 15 to 20 bp of coding sequence. Reverse primers comprised an XbaI restriction site sequence preceded by some overhang, two additional cytosines to allow in-frame cloning and 15 to 20 bp complementary to the coding strand. Subsequent studies used an A module, initially comprising PfAMA1 amino acid residues 97 to 442 and later residues 106 to 442 (FVO; GenBank accession no. AJ277646), corresponding to DI and DII. The Mm module (modified MSP1₁₉) comprised MSP1₁₉ (amino acids 1526 to 1621 of Wellcome line), optimized for P. pastoris codon usage (W. D. Morgan, unpublished data) and modified to replace Ser with Ala at residue 3 (to prevent N-glycosylation), and cysteine residues 12 and 28 with Ile and Trp, respectively, to remove a disulfide bond. A near-wild-type MSP1₁₉ (module M) that retained the disulfide bond was also prepared.

TABLE 1.

Characterization of P. pastoris-expressed AMA1 proteins, antigenicity in reactivity with MAb 4G2, immunogenicity as assessed by IFA, and functional analysis by parasite growth inhibition activity of induced antibodies

Construct name	Description (domain content)a	Amino acids	Reactivity withb:		GIA activityc
			MAb 4G2	IFA
3mH	P+I+II+mH	25-442	+	+	+
4mH	P+I+II+III+mH	25-545	+	+	+
8mH	II+mH	303-442	−	+	−
9mH	II+III+mH	303-545	−	+	−
10mH	III+mH	419-545	−	+	−
14.0	I+II+III	97-545	+	+	+
15mH	I+II+mH	97-442	+	+	+

^amH, myc epitope-hexa-His tag, derived from pPicZα A vector. P, prosequence; I, domain I; II, domain II; III, domain III.

^bIFA at 1:800 dilution against P. falciparum schizont-infected erythrocytes.

^cParasite GIA at 1.5 mg ml⁻¹ IgG; +, >50% inhibition; −, <10% inhibition.

Fusion genes were generated using strategy in which two PCR products with an overlap were fused via a further PCR. For module AMm, the first PCR used a forward primer comprising an EcoRI restriction site sequence (with extra overhang), followed by 21 bp for AMA1 DI (nucleotides 289 to 309); the reverse primer comprised 18 bp complementary to the DNA sequence for the end of DII (nucleotides 1309 to 1326), followed by 21 bp complementary to the start of MSP1₁₉, using AMA1 as the template. The second PCR used a forward primer comprising the last 18 bp coding for AMA1 DII, followed by the first 21 bp coding for MSP1₁₉; the reverse primer comprised 16 bp complementary to the end of MSP1₁₉ and two cytosines for in-frame cloning with the yeast vector myc-his tag, followed by the complement of an XbaI restriction site (plus extra overhang), with the modified MSP1₁₉ gene as the template.

The fusion PCR used these products with AMA1 forward and MSP1 reverse primers. Module MmA was created in a similar fashion. After the transformation of E. coli DH5α cells, plasmids were isolated, checked for the presence of expected restriction sites, and then used to transform P. pastoris KM71H following manufacturer's protocols. A positive clone was selected by small-scale cultivation (10 ml), essentially as described before (31).

Protein production.

Fermentation used either 3- or 7-liter fermentors (Applikon, Schiedam, The Netherlands), with initial starting volumes of 1 and 2 liters, respectively. Fermentation was essentially as described in the manufacturer's manual for fermentation of P. pastoris Mut^s cells (www.invitrogen.com/content/sfs/manuals/pichiaferm_prot.pdf) at 30°C and pH 6.0, with slight modifications in the air/oxygen sparging regimen. During batch phase, air was sparged at 1 liter per minute per liter starting volume of medium; during glycerol feed, oxygen was sparged to 20% saturation, and during methanol induction, air was sparged at 0.5 liter per min per liter starting volume, using oxygen to maintain saturation at 20%. Sparging with pure oxygen prevented induction of gene expression (B. Faber, unpublished observation). The batch phase was 24 h, the fed-batch phase was 20 h, and the induction phase was 24 h. After the induction phase, medium was adjusted to pH 7.8 and cooled to 15°C. Cells were removed by centrifugation (5,000 × g, 25 min, 4°C) and filtration across a hollow fiber 0.22-μm filter on a QuixStand (GE Healthcare, Etten-Leur, The Netherlands). Subsequently, the culture supernatant was loaded (8 ml/min) on an immobilized metal affinity chromatography column charged with nickel (Fast Flow immobilized metal affinity chromatography [IMAC] Sepharose, GE Healthcare) equilibrated with 20 mM sodium phosphate, pH 7.8, 500 mM NaCl, and 5 mM imidazole. The column was washed (20 mM sodium phosphate, pH 7.8, 500 mM NaCl, 10 mM imidazole), and then bound protein was eluted (20 mM sodium phosphate, pH 7.8, 500 mM NaCl, 100 mM imidazole). Buffer replacement was achieved by three consecutive concentration-dilution steps (1:15) in phosphate-buffered saline (pH 7.4) by using Amicon concentrators (Millipore, Etten-Leur, The Netherlands).

Antigenicity and identification of products.

The antigenicity of expressed proteins was confirmed by Western blotting using rabbit serum raised against complete FVO AMA1 ectodomain (33); rat monoclonal antibody (MAb) 4G2dc1, recognizing a reduction-sensitive conserved inhibitory epitope (32); rat MAbs 58F8 and 28G2, recognizing linear epitopes at the N terminus and the C terminus, respectively, of AMA1 (43); and the anti-MSP1₁₉ mouse MAb 12.10 (38). Alkaline phosphatase-conjugated secondary antibodies (anti-rabbit, anti-rat, and anti-mouse, antibodies as appropriate) were used to detect the primary antibodies using bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium as the substrate. Antigenicity of a full-length ectodomain protein (P. falciparum 4mH) was further evaluated by enzyme-linked immunosorbent assay (ELISA) using PfAMA1-specific MAbs 4G2, 58F8, and 28G2 and human serum from Guinea-Bissau, an area where malaria is endemic (56).

ELISA and indirect IFA.

ELISA was performed in duplicate on serum samples in 96-well, flat-bottomed, microtiter plates (Greiner, Alphen a/d Rijn, The Netherlands), coated with 500 ng ml⁻¹ purified AMA1 or MSP1₁₉ wild-type (module M) antigens according to published methods (33). The sequence of reagent addition was the following: protein, primary antibody to AMA1, alkaline phosphatase conjugate, substrate. A standard curve was plotted for each plate, and the concentrations of antibodies in the samples were calculated using a four-parameter fit. An immunofluorescence assay (IFA) was performed as previously described (33) using schizont-infected erythrocytes isolated from in vitro culture (P. falciparum 3D7 or FCR3). Secondary antibodies were fluorescein isothiocyanate-conjugated goat anti-rabbit immunoglobulin G (IgG) (heavy and light chains) (Kirkegaard & Perry, Gaithersburg, MD).

Determination of free cysteines.

The free cysteine content of proteins was determined using a free thiol determination kit (Molecular Probes, Leiden, The Netherlands) according to the manufacturer's instructions.

Rabbit immunizations.

Rabbits were housed and immunized, and blood was sampled by Eurogentec, Seraing, Belgium, under national animal welfare regulations. Animals were prescreened for the absence of antibodies to AMA1 and MSP1₁₉, and preimmune serum samples were obtained as controls. In initial studies comparing AMA1 regions, groups of two or three animals received 100 μg purified protein intramuscularly in Freund's complete adjuvant (our standard antibody induction regimen at that time) on day 0 and further injections of 100 μg in Freund's incomplete adjuvant on days 14, 28, and 56. Two weeks later, serum was prepared for analyses.

Two immunization studies were undertaken to compare the immunogenicity of the fusion proteins with that of the individual proteins. Protein for both studies was formulated in a final 0.5-ml volume with Montanide ISA 720 (Seppic, Paris, France), according to the manufacturer's instructions, and delivered intramuscularly on days 0, 28, and 56 to groups of five rabbits (unless otherwise stated). Protein dosage was adjusted to provide approximately equimolar quantities of each immunogen based on the difference in mass between AMA1 DI and DII (40 kDa) and MSP1₁₉ (11 kDa). In study 1, rabbits were immunized with modules M (20 μg), Mm (20 μg), and A (80 μg) and a mixture of module Mm plus module A (20 and 80 μg, respectively), or AMm fusion protein (100 μg). In study 2, rabbits were immunized with module Mm (20 μg), module Mm plus module module A (20 and 80 μg, respectively), AMm fusion protein (100 μg), or MmA fusion protein (100 μg, n = 10 animals). Antisera obtained 2 weeks after the final injection were used in all assays reported here. For use in parasite inhibition assays, antibodies were purified from serum using protein A columns (Sigma, St Louis, MO) under standard conditions, dialyzed extensively against RPMI 1640, filter sterilized, and stored at 4°C. IgG concentrations were determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE).

In vitro parasite growth inhibition assay (GIA).

In preliminary studies of antibodies induced against different regions of AMA1, a radioactive incorporation assay was used to measure parasite growth (33). The analysis of antibodies in the two fusion protein studies used a parasite lactate dehydrogenase (pLDH) growth assay as previously described (46). Briefly, protein A-purified IgG, isolated from the sera of rabbits immunized with fusion and control proteins, was added to triplicate well cultures of P. falciparum schizont-infected erythrocytes at a starting parasitemia of 0.2 to 0.4%, a hematocrit of 2.0%, and a total volume of 100 μl RPMI 1640 containing 10% normal human serum and 20 μg ml⁻¹ gentamicin in 96-well, flat-bottomed plates (Greiner, Alphen a/d Rijn, The Netherlands). After 40 h, cells were resuspended and 50 μl was transferred into 200 μl ice-cold phosphate-buffered saline. The cells were then harvested by centrifugation, and the supernatant was discarded. Lysis buffer was added to the cells, and the plates were frozen prior to estimation of parasitemia by using the pLDH assay. Parasite growth inhibition, reported as a percentage, was determined by the following formula: 100 − [average (absorbance_experimental − absorbance_background)/average (absorbance_control − absorbance_background) × 100]. Control IgG was isolated from rabbits that had been immunized with adjuvant alone.

Statistical analysis.

The IgG titer and percent parasite inhibition values for the various IgG preparations were compared and evaluated using analysis of variance. For IgG titers, the results were log transformed to normalize the data and are presented as a ratio relative to an index group for each antigen; for the AMA1 titers, the index group was AMA1; and for the MSP1₁₉ titers, the index group was that for immunogen Mm. Parasite GIA data were not transformed; effects are expressed as the difference between the index (AMA1 only) group and the group under investigation.

RESULTS

Production of PfAMA1 domains.

To determine the smallest region of AMA1 that could be expressed in P. pastoris and elicit high levels of functional antibodies, we transformed P. pastoris with constructs representing all possible consecutive combinations of domains in the extracellular part of the molecule (Table 1). All constructs designed to express DI alone failed to produce protein and are not described here. Constructs were designed to express protein by using the information on the known natural processing sites wherever possible, i.e., residue 25 for the start of the prodomain and residue 97 (27) for the start of DI. Other N-terminal residues were chosen to be immediately C terminal to the final cysteine of the preceding domain; in comparison with other nearby starting positions, this choice routinely gave equivalent or better yields. The final residue for each domain was chosen to be just prior to the first cysteine of the next domain. All constructs that included DIII ended at residue 545. Most of these proteins contain a vector-encoded C-terminal myc epitope, followed by a hexa-His tag, and these were isolated by Ni-IMAC. As further confirmation of identity, reactivity with antibodies to the myc epitope was determined by Western blot analysis (data not shown) and the sequence of AMA1 elements inserted into the yeast genome was confirmed on PCR-amplified material. Repeated transformations gave the same results. All products had less than 5 mol percent free cysteine (data not shown), indicating that over 95% of cysteines were oxidized. All proteins containing DI and DII were recognized by the reduction-sensitive MAb 4G2 (Table 1). The 3mH protein was cleaved during expression, removing the prodomain and essentially leaving DI-DII as the main product (data not shown).

Antibodies from rabbits immunized with these products reacted with the native form of AMA1 as assessed by IFA (Table 1). However, only IgG from rabbits immunized with proteins containing both DI and DII provided significant inhibition of parasite growth in in vitro cultures of P. falciparum FCR3. Neither the prosequence nor DII or DIII expressed individually or together induced appreciable levels of parasite inhibitory antibody. From these studies, we concluded that the smallest region of P. falciparum FVO AMA1 that could be expressed well in P. pastoris and that could retain the ability to induce parasite inhibitory antibody comprised DI and DII.

Production of fusion proteins of PfAMA1 DI-DII and modified MSP1₁₉.

The region of AMA1 comprising DI and DII (defined by 15mH and referred to here as module A) was fused either C terminally or N terminally to modified MSP1₁₉ (C12I, C28W) (defined herein as module Mm), with the aim of comparing this fusion with the individual components and allowing selection between AMm and MmA fusion proteins on the basis of expression level, product stability, and immunogenicity. In an original MmA construct, module A started at residue 97 in AMA1. This protein was cleaved almost quantitatively into two fragments by a yeast protease (data not shown). One cleaved product reacted with MAb 4G2, and the other reacted with MAb 12.10, suggesting that the cleavage site lay close to the fusion point of modules Mm and A. By protein sequencing, the N-terminal amino acid of the product reactive with 4G2 was shown to lie at Ser103 within the N terminus of the A module. As a result, a second MmA fusion was developed in which the A module starts at residue 106 in the AMA1 sequence; this new fusion protein was not cleaved near the point of fusion and was used for all further analyses.

One- and 2-liter fed-batch fermentations were undertaken to determine the reproducibility of protein expression and to generate MmA and AMm proteins for further analysis. After purification from the supernatant by single-pass Ni-IMAC, averages of 48 mg liter⁻¹ for AMm and 30 mg liter⁻¹ for MmA were obtained. Both fusion proteins migrate on SDS-PAGE, with a mobility corresponding to the expected mass of 50 kDa (Fig. 1A), and both proteins react with MAbs specific for both AMA1 (Fig. 1B) and MSP1₁₉ (Fig. 1C). The determination of the N-terminal amino acid sequence of AMm showed that the protein had also been cleaved proteolytically to remove the N-terminal amino acids derived from AMA1 (residues 97 to 102), leaving serine 103 at the N terminus of the AMA1 sequence. Upon reduction and denaturation, both fusion proteins showed evidence of a small amount of internal proteolytic cleavage at a single site (Fig. 1D). Size exclusion chromatography of both fusion proteins under reducing conditions indicated that less than 10% of each product was cleaved in this way (data not shown). This observation was reproducible, and the size of the cleaved products suggested that the cleavage site is the same as that reported by Kennedy et al. (30), between residues 376 and 377 within the DII loop. This was confirmed by N-terminal sequencing of the product (data not shown).

FIG. 1.

Analysis of purified AMA1-MSP-1₁₉ fusion proteins. (A) Nonreduced SDS-PAGE of fusion products, 10 μg protein per lane. Lane 1, module Mm; lane 2, module A; lane 3, module AMm; lane 4, module MmA. (B) Western blot with monoclonal antibody 4G2 specific for PfAMA1. Lane 1, module Mm; lane 2, module A; lane 3, module AMm; lane 4, module MmA. (C) Western blot with monoclonal antibody 12.10 specific for MSP-1₁₉. Lane 1, module Mm; lane 2, module A; lane 3, module AMm; lane 4, module MmA. Western blot gels were loaded with 1 μg/lane. (D) Reduced SDS-PAGE of fusion proteins, 10 μg protein per lane. Lane 1, module MmA; lane 2, module AMm.

Preliminary accelerated stability studies on the purified fusion proteins were performed at 30°C over a period of 39 days. Neither product showed a propensity for aggregation or further multimerization. The proportion of cleaved AMm protein remained unchanged throughout, while after 30 days some further cleavage of MmA protein gradually became apparent (data not shown).

Immune responses to fusion proteins.

Because most of the original MmA protein was cleaved into two separate polypeptides, a redesigned product was required. In the interim, to evaluate whether a fusion protein approach would compromise the quality of antibody being induced, a first immunogenicity study was initiated. Groups of five rabbits were immunized with equimolar amounts of modules AMm, Mm, M, and A and a mix of module Mm plus module A in Montanide ISA 720. Sera collected from rabbits 2 weeks after the final immunization were assessed by ELISA against module M and against P. falciparum 11.0, comprising prosequence-DI-DII-DIII of AMA1 FVO, both derived from P. pastoris (Table 2). Overall, levels of anti-AMA1 IgG were not significantly influenced by mixing or fusing the components; in contrast, levels of IgG induced against MSP1₁₉ were. In this first study, mixing modules A and Mm resulted in a decreased IgG response to M (0.36-fold, P = 0.021). Conversely, IgG responses to module M were boosted (1.6-fold, P = 0.26) when module Mm was fused to module A (Table 2). As predicted from studies of a rodent model (17), module Mm alone induced significantly higher antibody MSP1₁₉ levels than did module M alone. IgG titers to module M were significantly higher (4.4-fold, P < 0.002) in animals immunized with AMm fusion compared to those immunized with a mixture of modules A and Mm. To assess the functionality of the antibodies, parasite growth inhibition assays were performed at four concentrations of IgG (Fig. 2A) with parasites of the FCR3 strain (the strain expressing AMA1 most homologous to the PfAMA1 immunogen) as well as with HB3 and 3D7, strains heterologous for AMA1. This gave evidence that the fusion of AMm did not compromise the induction of parasite inhibitory antibody compared to that induced by a mixture of modules A and Mm and the fusion had improved responses compared to those of module Mm alone. It also demonstrated that inhibitory responses to module Mm were, in general, no worse than those to module M, providing further evidence that the modification had not caused the loss of critical epitopes.

TABLE 2.

IgG titers to AMA1 and MSP-1₁₉ for study 1

Immunogen	Value for AMA1a			Value for MSP-119
	Geometric mean titer (95% CI)	Fold increase (95% CI)	P valueb	Geometric mean titer (95% CI)	Fold increase (95% CI)	P valuec
A	1,092 (332-3,592)	1		0 (0-0)	0.00 (0.00-0.00)	<0.001
A + Mm	1,247 (352-4,421)	1.14 (0.70-1.86)	0.57	92 (11-773)	0.36 (0.15-0.84)	0.021
AMm	1,059 (265-4,227)	0.97 (0.60-1.58)	0.90	408 (112-1,492)	1.61 (0.69-3.75)	0.26
Mm	0 (0-0)	0.00 (0.00-0.00)	<0.001	254 (80-804)	1
M				99 (17-570)	0.39 (0.16-0.96)	0.042

^aCI, confidence interval.

^bP value for comparison with A.

^cP value for comparison with Mm.

FIG. 2.

Functional analysis of immune responses; in vitro parasite growth inhibition. Rabbit IgG raised against the indicated antigens was incorporated into parasite cultures; growth inhibition was determined relative to cultures containing IgG from rabbits receiving adjuvant alone. (A) Study 1. (B) Study 2. Open circles, module A; closed circles, module A and Mm mixed; open squares, module AMm; closed squares, module MmA; closed triangles, module Mm; open triangles, MSP1₁₉ wild type.

When module MmA that had been modified to remove the cleavage site at the start of the AMA1 sequence became available, study 2 was initiated; groups of five rabbits (except where noted) were immunized (using the same regimen as for study 1) with equimolar amounts of modules MmA (n = 10 animals), AMm, Mm, and A and a mix of module Mm plus module A. Sera collected from rabbits 2 weeks after the final immunization were assessed by ELISA (Table 3). As for study 1, IgG responses to MSP1₁₉ were boosted when it was fused to AMA1, while no significant effect on AMA1 responses could be detected. In contrast to study 1, mixing modules A and Mm did not decrease responses to MSP1₁₉. Compared to IgG responses induced by immunization with module Mm alone, immune responses to module M were 4-fold higher (P = 0.032) and 4.6-fold higher (P = 0.009) in animals immunized with the AMm and MmA fusion proteins, respectively. Immunofluorescence analysis using mature schizont-infected erythrocytes showed the expected pattern of staining with the different antibodies (Fig. 3). Rabbits immunized with module Mm (alone, in fusion, or in a mix) showed a uniform staining of the merozoite circumference, while antibody from rabbits immunized with module A alone showed the typical apical localization expected for AMA1. When both antigens were present in the immunogen, the MSP1₁₉ character of the staining tended to prevail, although strong apical staining was often also evident. Parasite inhibition by the antibodies was assessed for three parasite strains (Fig. 2B). Antibodies against modules AMm and MmA functioned comparably against FCR3 and HB3, with a tendency for module AMm to perform better against 3D7 at low IgG concentrations. As for study 1, there was a tendency for the fusions to perform as well as or better than the mixture and to outperform the individual antigens. Notably, anti-Mm GIA responses were proportionately greater against strains expressing an AMA1 heterologous to the immunogen, suggesting potential added benefits of immunization with both module A and module Mm in improving responses to a broad range of parasite strains.

TABLE 3.

IgG titers to AMA1 and MSP-1₁₉ for study 2

Immunogen	Value for AMA1a			Value for MSP-119
	Geometric mean titer (95% CI)	Fold increase (95% CI)	P valueb	Geometric mean titer (95% CI)	Fold increase (95% CI)	P valuec
A	256 (33-2021)	1		0 (0-0)	0.00 (0.00-0.00)	<0.001
A + Mm	180 (50-651)	0.70 (0.38-1.30)	0.25	37 (8-168)	2.03 (0.66-6.30)	0.21
AMm	245 (76-795)	0.96 (0.51-1.78)	0.88	146 (46-467)	8.11 (2.61-25.15)	<0.001
MmA	282 (128-624)	1.10 (0.64-1.90)	0.72	170 (74-391)	9.43 (3.47-25.58)	<.001
Mm	0 (0-0)	0.00 (0.00-0.00)	<0.001	18 (0-2511)	1

^aCI, confidence interval.

^bP value for comparison with A.

^cP value for comparison with Mm.

FIG. 3.

IFA with anti-AMA1 and anti-MSP-1₁₉ rabbit sera. Methanol-fixed, thin blood films of schizont-infected P. falciparum erythrocytes were reacted with rabbit sera (1:800 dilution) as indicated and were stained with fluorescein isothiocyanate-conjugated secondary antibodies.

DISCUSSION

A number of single and combination target vaccines are in development for P. falciparum malaria (13, 37, 39, 45). Simultaneous immunization with a number of antigens in combination vaccines has several potential advantages: the induction of responses to different stages of parasite development, thereby offering protection throughout the infection cycle; the simultaneous attack on multiple parasite targets with the possibility of additive or synergistic effects; the better response to one target than to another in some individuals; and the presentation of a range of variants for any given target antigen with potential benefits in breadth of immunity. Drawbacks of such approaches include the potential need to develop and test each component individually and the complexity of any ultimate combination. With a view to targeting the asexual malaria parasite that is responsible for the clinical manifestations of the disease, we have initially focused on the AMA1 and MSP1 molecules as two merozoite surface proteins for which there is evidence from animal models of inhibitory antibodies that can be induced by vaccination. Our aim is to simplify and substantially reduce vaccine development costs by the production of a fusion protein combining the smallest elements from each molecule capable of inducing a strong inhibitory antibody response. In the present study, this has been restricted to the fusion of two elements. However, given the allelic diversity of AMA1 and the potential for combination with other vaccine candidates in addition to MSP1, the use of a fusion protein strategy may have considerable benefits. Indeed, in preliminary work, we have already extended these findings to more complex fusions, indicating that the strategy offers a realistic way forward.

From data presented here, we conclude that the smallest region of P. falciparum FVO AMA1 that can be expressed in P. pastoris and that is capable of inducing high levels of parasite inhibitory antibody is defined by DI-DII. Using a prokaryotic expression system and refolding of the expressed protein, Lalitha et al. reached similar conclusions (35). Others have reported that DIII is also a target for parasite inhibitory antibodies by analyzing antibodies against a peptide loop (41) and antibodies directed to sites of mutation within DIII (42). We have been unable to obtain significant levels of inhibitory antibody to DIII when DIII is used either on its own or in combination with DII; this result is in agreement with the findings of Pan et al. (45), although intriguingly, in their study it was reported that inhibitory responses to DIII could be induced when it was fused with MSP1₁₉.

We and others have evaluated sera collected from areas where P. falciparum malaria was endemic and shown that most exposed individuals do not respond strongly to DIII (7, 49) and such antibodies appear to be only weakly inhibitory to the parasite (42). It is possible that the recombinant DIII used for these analyses was not folded correctly into the native conformation. Most AMA1 epitopes recognized by antibodies in human sera are discontinuous (56, 57), and the formation of correct disulfide bonds in the protein is required for the induction of significant levels of inhibitory antibody following immunization (1). The solution structure of E. coli-expressed and refolded DII and DIII has been determined by nuclear magnetic resonance methods (14, 42). Both domains are substantially more disordered than the corresponding regions in the structure of the entire ectodomain established by X-ray crystallography of P. pastoris-expressed P. vivax AMA1 (47). Thus, DIII may only fold correctly in the context of DI and DII. However, we suspect that in the context of the AMA1 ectodomain, DIII is poorly immunogenic or presents only a limited target for inhibitory antibodies because we have been unable to show significant parasite inhibitory activity in IgG raised against the entire ectodomain following the depletion of antibodies reactive with DI and DII (unpublished results).

Even without optimization of fermentation and purification, the AMm and MmA fusion proteins were produced in good yield and with a main band purity of over 95% as judged by SDS-PAGE. The main product for each fusion reacted with monoclonal antibodies to both AMA1 and MSP1₁₉ that recognize discontinuous epitopes dependent on disulfide bond formation and correct protein folding (32, 61). This suggests that both modules retain appropriate conformation in the fusion proteins. As assessed by size and reactivity with these two antibodies, the other minor polypeptides in the products were predominantly multimers of the fusion proteins (Fig. 1B and C).

Because of potential benefits for vaccine development in the production and use of a fusion protein over the A or Mm modules used individually or as a mixture, we assessed whether the immunogenicity of either component, or its ability to induce functional parasite inhibitory antibodies, was compromised by fusion. Antibodies were raised in rabbits in two separate studies; these studies used the same immunization regimen, differing only in the incorporation of the MmA fusion protein in the second study. Antibodies induced in both studies bound as expected to native parasite molecules in the micronemes, and on the merozoite surface, showed appropriate specificity by ELISA. Overall, the relative ranking of responses by ELISA and GIA, and the relative susceptibilities of the three strains tested were comparable between the two studies.

It has previously been shown that disruption of the cystines in the small C-terminal region of MSP1 tested here facilitates processing for antigen presentation and increases the immunogenicity of the protein (17, 40). Data from study 1 confirm this; antibody responses were significantly higher following immunization with module Mm than with module M, and by GIA we show that rabbit antibodies to module Mm recognize functionally important epitopes. In this study, immunization with a mixture of modules A and Mm appeared to induce antigenic competition that depressed responses to module M compared to immunization with module Mm alone. This effect was not seen in study 2, a study that had quantitatively lower antibody responses than study 1, suggesting that antigenic competition to MSP1₁₉ may occur only when stronger antibody responses are induced. In both studies, in contrast to products containing AMA1, responses to modules M and Mm alone show little strain-dependent variation in GIA.

In both studies, immunization with products containing module A resulted in GIA responses where FCR3 > HB3 > 3D7. This equates with the fact that HB3 and 3D7 differ from the FCR3-type A immunogen by 14 and 21 amino acid differences, respectively. Thus, as others have reported (30), rabbits develop a significant response to polymorphic epitopes of AMA1, a tendency that may be diminished in primate immune responses (E. J. Remarque, unpublished observation; Carole Long, personal communication).

There are no established in vitro correlates for protection in malaria, and we have only a limited understanding of the immune mechanisms that may mediate parasite growth reduction following vaccination in humans. Our study has therefore been designed as a prelude to human studies to determine whether fusion was likely to cause any overall harm to the nature and quality of the immune response to the two component molecules. We can conclude that at the very least, fusion of the modules had no deleterious effect on antibody induction by AMA1 or MSP1. Indeed, fusion resulted in significantly enhanced induction of antibody to MSP1₁₉, as has previously been shown in the fusion of AMA1 DIII with MSP1₁₉ (45). Although these increased levels of antibody did not significantly enhance performance in the one-cycle growth GIA protocol when equal amounts of IgG isolated from serum were compared, they may have positive effects on mechanisms operating in vivo. There was a general trend of enhanced antibody levels and GIA activity to AMA1 following fusion; importantly, in none of the assays was there any indication that the fusion reduced capacity to induce GIA activity. Thus, the fusion is performing at least as well as the mixture, offering the practical advantage that the development of a fusion vaccine is in principle simpler and more cost effective than the development of individual components that are subsequently mixed. The fusion protein has other potential advantages: under circumstances where the contribution of module A to 3growth-inhibitory activity becomes less due to AMA1 polymorphism, the relative contribution of the Mm module increases, suggesting that the combination offers better strain coverage than vaccination with either module alone.

In conclusion, the two fusion proteins offer vaccine-related benefits over either module alone or mixed, prompting assessment of the feasibility to produce clinical grade material. Although a proportion of both fusion proteins (less than 10%) was cleaved during expression by a yeast protease at a single site, probably within a DII loop, no significant further deterioration of the products occurred during storage at 30°C for over a month. After this time, further cleavage of module MmA, probably at the same site, gradually became evident; this may have been due to the copurification of small amounts of the yeast protease, a problem that should be overcome by improved postexpression purification protocols. Thus, results presented here from simple, nonoptimized laboratory scale fermentation and purification for both MmA and AMm fusion proteins suggest that adaptation to GMP level production should not be problematic, and feasibility studies to this end are currently under way. Given the constraints in capacity and costs associated with clinical studies, the selection of one of the two fusions for advancement to phase I studies will be required. Given equal immunogenicity of the two fusion proteins produced under near-GMP conditions, the choice for production and clinical evaluation will be based on relative ease of production and stability. With a view to clinical studies, we have developed fusion proteins devoid of the myc/hexa-His tag; at small-scale fermentation levels, these proteins express at least as well as the products described here (data not shown). From previous studies comparing AMA1 molecules with and without His tags, we do not expect the presence or absence of a His tag to affect the immunogenicity profile of these fusions.

One further observation deserves further comment. We and others (37) have first-generation vaccines in current early stage clinical trials comprising almost the entire AMA1 ectodomain produced in P. pastoris. As we report here, during expression in P. pastoris, an original MmA fusion in which module A started at residue 97 was cleaved at residue 103. Expression of the 3mH construct also resulted in a product in which the prodomain had been cleaved, suggestive of the same processing event. In contrast, protein 4mH (full ectodomain) was not cleaved. The crystal structure of AMA1 shows that that the N-terminal region associates with DIII (47), raising the possibility that this interaction in some way stabilizes the region or sterically blocks its availability to the yeast protease. This stability may explain why the first-generation ectodomain products are not cleaved at this site.

We believe that the fusion protein approach has much to commend it for malaria vaccine development, and it should be possible to express products larger than these simple fusion proteins. This may be relevant because, for example, MSP1₁₉ of P. falciparum exists in essentially two forms (59), and it is probable that AMA1 polymorphism is largely maintained by immune selection (48). Recently we showed that coimmunization with a small number of recombinant proteins, containing the most frequently occurring amino acids at each polymorphic site in DI and DII of AMA1, can result in improved functional responses to a broad range of parasite lines (E. J. Remarque et al., unpublished data). We have already, albeit at somewhat lower expression levels, obtained products in which two A modules have been fused to two Mm modules, with the idea that the specificity of the immune response can be broadened by the inclusion of two forms of both AMA1 and MSP1₁₉. This approach offers the prospects of fusion proteins with an even greater opportunity to overcome the natural variation inherent in parasite populations.