For a brief window of time in the asexual lifecycle of Plasmodium falciparum, the parasite is exposed to the host immune response as it seeks to engage with and invade an erythrocyte. Intervention at this point, targeting the merozoite to interrupt the mechanism of invasion, is a strongly pursued strategy of malaria vaccine development. Many different parasite proteins have been implicated in the invasion cycle of merozoite attachment, re-orientation and entry [1]. The surface coat of the merozoite is largely comprised of glycophosphatylinositol (GPI) anchored membrane proteins and their associated partners [2]. Nine of these GPI-anchored merozoite surface antigens (MSAs) have been described so far, in raft like membranes [3], and all are potentially erythrocyte ligands. The GPI-anchored proteins MSP-1 and MSP-2 have long been recognized as potential vaccine targets, and they have already been developed and trialled as asexual blood-stage vaccine candidates [4]. Three more GPI-anchored MSAs have more recently come to attention as potential vaccine candidates [5]. These proteins are members of a six-cysteine (Cys₆) family, whose 11 known members are expressed at various stages of the parasite lifecycle, mostly in the sexual stages [6]. The three asexual stage Cys₆ proteins, Pf12, Pf38 and Pf41 have been localized to the merozoite surface, with Pf38 distributing more specifically to the apical end than the other two. All three are well-recognized by human immune sera [3], and thus may be involved in the host immune response. However, antibody recognition alone is not substantial proof that an antigen is a sufficiently important target of the protective response to be worthy of prioritization as a vaccine candidate.

Evidence of selective pressure on a particular antigen in a natural parasite population is a good indicator of its importance as a target of human immunity. Using a population genetic approach, this selective pressure can be identified through a signature of balancing selection, detecting the maintenance of low to medium allele frequencies within a parasite population that result in an advantage for immune evasion [7, 8]. Recently, a prospective study was undertaken in a Gambian parasite population to identify balancing selection among 26 blood-stage surface-exposed proteins [9], including the Cys₆ proteins Pf12, Pf38 and Pf92. An initial screen of the antigens by Hudson-Krietman-Aguade (HKA) [10] and McDonald-Krietman (MK) [11] analysis, that calculate divergence from the sequence of Plasmodium reichenowi [12], led to the identification of six antigens potentially under selection. Further population-based analysis, using Tajima's D (TjD) test [13] and Fu and Li's indices [14], confirmed four of these (msp 3/6, sera 5, Pf 38 and msp 7) to have signatures of balancing selection [9].

On the basis of these findings, further exploration of the population genetics of Pf 38 in two regions of intense perennial malaria transmission in Papua New Guinea was undertaken. The Mugil area in Madang Province, on the north coast of PNG, had been the site of a previous study in 2004, looking at the risk of infection and disease in PNG children attending the local elementary schools [15]. Likewise, the Wosera region of East Sepik Province had been the subject of extensive cross-sectional surveys in 2001-2003 [16]. This previous work provided us with a detailed background of transmission and infection dynamics in these two catchments. In more recent all-age cross-sectional community-based surveys in Mugil and Wosera, infection prevalence (~40%) was found to be similar to the published literature [17]. Genotyping of ten microsatellite loci from these samples uncovered high levels of genetic diversity (allelic richness and heterozygosity) and a lack of linkage disequilibrium [17]. The catchments also had strong local parasite population structure, indicating limited gene flow. The genetic structure of these two populations suggested that this sample set would serve as an excellent indicator of the diversity and selection of Pf 38 in a malaria endemic area.

The GPI-anchored P. falciparum merozoite membrane protein Pf38 is a potentially attractive candidate for vaccine development studies. It is expressed in both sexual and asexual stages and its isolation from raft-like membranes and localization to the merozoite apical complex in the erythrocytic cycle suggests a role as a ligand in erythrocyte invasion [3]. The protein is strongly targeted by the antibodies in the serum of immune PNG adults [3] and 20-mer peptides based on Pf 38 sequence have been shown to inhibit in vitro invasion [5]. In Gambian isolates, Pf 38 was one of only 4 out of 26 genes encoding surface-exposed proteins that carried a signature of balancing selection, though this was a relatively moderate effect compared to better known vaccine candidates, such as ama1 [9]. Study of the population genetics of Pf38 from a different geographical location was needed to confirm its susceptibility to immune selection and to explore the degree of diversity and geographical variation within the gene, in order to further consider of its potential as a malaria vaccine candidate.

A substantial number (n = 19) of SNPs were found amongst the isolates. The majority of these were confined to domain I and those appearing in domain II were either singletons or found at extremely low frequencies within the parasite populations. In all three populations, the overall nucleotide diversity (∏) was low, yet the haplotype diversity (Hd) was high. This has been previously observed in analysis of other merozoite surface antigen genes, such as eba175, and suggests that recombination generates a range of haplotypes even where functional constraints exist [17]. When considering domain I alone, nucleotide diversity is much higher than domain II and substantially higher than the figure for the full gene. This might suggest some functional limitation of the amount of polymorphism that can be maintained in domain II.

Signatures of balancing selection, demonstrating the maintenance of putative antigenic diversity within populations, have proven a good indication of the importance of antigens as targets of the immune response [7, 8]. It is important to consider a range of statistical tests of selection, such as those performed here, as no single test gives definitive proof, and indeed correlation of some of the tests can be quite poor in certain circumstances [9, 12]. The various methods utilized here consistently pointed towards a modest degree of balancing selection for the Pf38 gene in the PNG population when the full sequences were analysed, as had been seen previously in the Gambian isolates [9]. Of particular interest was the very considerable increase in evidence for balancing selection in domain I, compared to the full gene and domain II. The sliding window analysis of Tajima's D, with domain I demonstrating strong balancing selection and domain II showing neutral values, combined with the SNP map and observation that the majority of SNPs in domain I were common while those in domain II were rare, suggests that domain I may be the dominant target of natural host immunity.

The greater diversity seen in the Gambian isolates is confirmed by the presence of many more unique haplotypes than that observed in the PNG populations. Of the 34 haplotypes seen in field isolates, 28 are represented in the Gambian dataset. Wosera and Mugil display 8 and 9 haplotypes respectively. Of importance, in the context of the possibility of creating a vaccine with broad coverage of the worldwide parasite population, is that the majority of isolates in each natural population harboured one of just four dominant haplotypes. Aside from these common haplotypes, there is quite considerable variation, with several unique haplotypes in each population.

Taking this information as a whole, one could speculate that the antigen has a functionally restricted domain II shielded from the immune response by an exposed and highly polymorphic domain I. The hypothesis of functional conservation of domain II is given support by work from Garcia et al [5], who performed a fine mapping study of the binding of Pf38. They created a set of 20-residue-long peptides covering the complete gene sequence and performed a series of binding experiments, and erythrocyte invasion inhibition assays. Two peptides (aa141-160 and aa201-220) showed high activity binding and binding inhibition. Both of these were located inside domain II, suggesting that this region has a role in protein-protein interactions. None of the SNPs identified in the present study co-localized with these peptides.

This study of polymorphism, diversity and selection of Pf38 in two P. falciparum populations in PNG, confirmed the findings of Tetteh et al. [9] in the Gambia that Pf38 is a modestly polymorphic antigen gene that shows some evidence of balancing selection. However, the analysis of population data in a domain specific fashion here has painted a considerably different picture of the gene than that assumed from the full-length analysis. Domain I is highly polymorphic and shows evidence of significant balancing selection, suggesting that the gene is a target of natural immunity. Domain II is relatively conserved with no evidence of immune selection, supporting the hypothesis that it may be the functionally-conserved site of protein-protein interaction in the invasion process. The findings here have implications for future population genetic studies on vaccine candidates. They demonstrate that direct whole-gene comparisons may give misleading rankings of vaccine potential. One must instead consider the structural and functional context as a framework for analysis.