Population genetic analysis of the Plasmodium falciparum 6-cys protein Pf38 in Papua New Guinea reveals domain-specific balancing selection

The Papua New Guinean provinces of Madang and East Sepik are areas of intense perennial malaria transmission that have long been the focus of malaria research and ongoing control efforts. As part of a large cross-sectional malaria survey in March 2006, venous blood samples were collected from asymptomatic human volunteers of all ages, including the three villages of Dimer, Karkum and Matukar/Bunu, within the catchment of the Mugil Health Centre, on the North Coast some 50 km from the urban population centre of Madang Town. A similar survey took place in the Wosera region of East Sepik Province in September 2005, including the villages of Gwinyingi and Nindigo. This allowed the examination of a broad range of parasites circulating in the community, unbiased by symptomatic outcomes. Genomic DNA was extracted from the whole blood samples using the 96 well QiaQuick DNA extraction kit (Qiagen). Ethical approval to conduct the study was granted by the PNG Institute of Medical Research Institutional Review Board, the Papua New Guinea Medical Research Advisory Committee and the Alfred Hospital Research and Ethics Unit.

To identify P. falciparum positive samples and concurrently estimate prevalence and multiplicity of infection (MOI), samples were screened by genotyping the highly polymorphic gene encoding merozoite antigen protein 2 (msp2), as previously described [18]. This approach utilized a nested multiplex PCR to simultaneously amplify both 3D7 and FC27 allele families from genomic DNA. Fluorescently labelled products specific to allele-family were then analysed on an ABI sequencer (Applied Biosystems). As described previously, P. falciparum positive samples found to contain a single msp2 allele were then screened further at ten microsatellite loci to confirm single infections [17].

Thirty-six single infection gDNA samples from Mugil and twenty-three from the Wosera were amplified for the Pf 38 gene, using a hemi-nested PCR technique. First round amplification was performed with oligonucleotide primers F2: 5'-ggt caa tca tta cag gaa tcg-3' and R1: 5'-cgc aaa tga aat ttc ttc tc-3' and second round with oligonucleotide primers F2: 5'-ggt caa tca tta cag gaa tcg-3' and R3: 5'-gct tgt taa ctc caa tac ttc-3'. Amplification was performed in 15 μL volumes, containing 1.5 mM MgCl₂, 200 μM each dNTP, 0.33 μM oligonucleotide primer, 5% PCR grade DMSO (Sigma-Aldrich, MO, USA) and 0.5 units hotstart Taq polymerase (5 PRIME, GmbH Germany) The cycling conditions were 94°C for 2 min, 35 cycles of 94°C for 20 sec, 54°C for 30 sec and 60°C for 1 min, followed by a final extension at 60°C for 7 min. Identical protocols for the PCR reaction mix and cycling conditions were used for both rounds of amplification. The resulting 921bp fragment was then purified and sequenced by the commercial sequencing service provided by Macrogen Inc. (Seoul, Korea) on an ABI 3730XL DNA analyser.

Raw sequence data chromatograms were edited and aligned into contigs in Sequencher version 4.0 (Gene Codes). Single nucleotide polymorphisms (SNPs) were identified by comparing each set of sequences for an individual isolate to that of the 3D7 reference genome [19] (Plasmo DB Id: PFE0395c). Individual SNPs were confirmed if they were also identified in other isolates or if detected in at least two independent PCR products from the same isolate. Determination was made of the total number of polymorphic sites (S); the numbers of synonymous (dS) and non-synonymous (dN) polymorphisms; the nucleotide diversity (P), which is the average number of nucleotide differences per site between two sequences [20]; the number of distinct haplotypes (h) and the haplotype (gene) diversity (Hd) which is analogous to heterozygosity (Hd = [n/(n-1)][(1-Σ(f _i )²)] where n is the sample size and f is the frequency of the i ^{t h} allele) [20]. The number of haplotypes (h) is strongly influenced by sample size therefore we also calculated the allelic richness (R _s ) which is normalized on the basis of the smallest sample size and based on the rarefaction method developed by Hurlbert [21] and implemented in FSTAT version 2.9.3 software [22]. The nucleotide divergence (K) was measured as the average proportion of nucleotide differences between species [20] using the P. reichenowi Pf38 orthologue (GenBank accession: EF123272).

Natural selection was measured using the McDonald-Kreitman test of the neutral hypothesis [11, 23], based on a comparison of synonymous and nonsynonymous variation within and between species. Under neutrality, the ratio of replacement to synonymous fixed substitutions (differences) between species should be the same as the ratio of replacement to synonymous polymorphisms within species. Significant increases or decreases in this ratio were determined by two-tailed Fishers exact test. We also calculated the D test statistic proposed by Tajima [13], for testing the hypothesis that all mutations are selectively neutral [23]. The D test is based on the differences between the number of segregating sites (S) and the average number of nucleotide differences between sequence pairs. Positive values of Tajima's D indicate balancing selection while negative values indicate purifying or directional selection. A sliding window approach was used to highlight specific regions of Pf38 that deviate from neutral expectations. Measurement was made of the F statistic proposed by Fu and Li [14] for testing the hypothesis that all mutations are selectively neutral [23] with the P.reichenowi Pf38 sequence as an outgroup; the Hudson, Kreitman and Aguadé's [10] ratio (HKAr) was calculated as P/K and is a relative index used to compare different loci. The test is based on the Neutral Theory of Molecular Evolution [23] prediction that regions of the genome that evolve at high rates will also present high levels of polymorphism within species. All of the above statistics were determined using DnaSP version 5.0 [24]. We also used codon-based Z tests, averaging over all sequences, to calculate the probability of departure from neutrality (d_N = d_S), positive selection (d_N > d_S) or negative selection (d_N < d_S). Values of p less than 0.05 were considered significant at the 5% level. The variance of the difference was computed using the bootstrap method (1000 replicates). Analyses were conducted using the Nei-Gojobori method in MEGA4 [20, 25]. The interpopulation differentiation (F _ST ) at specific polymorphic sites was measured using GENEPOP version 4.0 ("GENEPOP on the web") [26].

As the Pf38 gene has two distinct 6-cys domains (Figure 1A) in addition to defining the above statistics for the full sequence, analysis was also performed on the two 6-cys domains separately, to investigate domain specific selective pressure (domain I: 85-466 bp, domain II: 467-892 bp).

High quality, full-length sequences were obtained for a total of 59 PNG isolates (GeneBank accession#: JF944833-JF944891). These were compared to previously published data from the Gambia (n = 88) and that available for established laboratory isolates (n = 18) [9]. The full dataset is included in the supporting material (Additional File 1). A total of 19 SNP's were identified in the region analysed (85-892 bp). Of these, 13 polymorphic sites were found in the Gambia, 10 were found in Mugil and 8 in Wosera (Figure 1). Polymorphic sites unique to specific populations were also identified, including 5 in the Gambia, 4 in Mugil and 3 in laboratory isolates. No unique sites were present in the Wosera. All of the polymorphisms identified in PNG were non-synonymous mutations, and the only synonymous mutation was found in a single isolate from the Gambia. Although there was a modest level of nucleotide diversity (∏), there were several haplotypes (h) and a high haplotype diversity (Hd) (Table 1). The majority of common polymorphisms were confined to domain I (Figure 1A) while those found in domain II were found at low frequencies being represented in only a few haplotypes (Additional File 2). Accordingly, the ∏ and Hd values for domain I are substantially higher than domain II and much higher than would have been assumed from the whole-gene calculation.

Signatures of balancing selection are strong indicators of immune selection and a number of the analyses in Table 1 address this question (HKAr, Z test, M-K, TjD and Fu & Li F). All of these tests showed the same trend, with the full gene demonstrating limited evidence of balancing selection. However, there was a strong contrast observed between domain I, with significant departure from neutrality in a number of tests, and domain II, which showed neutral values. The specific location of the selective pressure on the gene is clearly illustrated by the Tajima's D plot in Figure 1B, where positive values represent balancing selection and negative values represent purifying or directional selection. Low levels of interpopulation differentiation (F_ST) for several sites in domain I further support the activity of balancing selection on this region compared to domain II where higher levels of F_ST were observed (though small sample sizes may influence these results).

A total of 37 different Pf 38 haplotypes encoding distinct amino acid sequences were observed among the combination of the PNG sequences generated in the current study and the Gambian and laboratory isolates from previous investigations [9] (Table 2 and Additional File 3). The Gambian Pf38 sequences showed the greatest diversity with 28 of the total 37 haplotypes represented. This increased diversity was also observed when the allelic richness (R_S), which provides a normalized value on the basis of the smallest sample size (n = 23), was measured (R_S: Gambia = 18.449, Mugil = 8.604 and Wosera = 8.00). Three haplotypes were unique to laboratory isolates and not observed in the natural populations analysed. The sequences from the Wosera and Mugil contained far fewer haplotypes, with 8 and 9 respectively, which will at least in part be the result of a smaller sample size than the Gambia. Interestingly, only four haplotypes (Haps 2,3,5 and 7) were found to account for over 60% of the PNG isolates and almost half of those from the Gambia. In the sector outside these common globally distributed haplotypes there is considerable diversity, not only between Africa and PNG, but also between the two PNG locations. In Mugil, a single polymorphism seen at position 416 and found in Haps 33,35,36 and 37 is present in 30% of samples, but is not seen in either the Wosera or Gambian samples. Likewise in Wosera, a polymorphism at position 205 and found in Hap13 is seen in 20% of samples, yet not seen in the Wosera and only found in one isolate from the Gambia.