Microsatellite genotyping and genome-wide single nucleotide polymorphism-based indices of Plasmodium falciparum diversity within clinical infections

Patients between the ages of one and 9 years presenting with uncomplicated P. falciparum malaria, as confirmed by rapid diagnostic test (Paracheck, Orchid Biomedical systems, India), were recruited from local health facilities within a 25-km radius of the town of N’Zerekore in the Republic of Guinea, between March and May 2011. Written informed consent was obtained from a parent or guardian of each child sampled, and all patients diagnosed with malaria were treated with artesunate-amodiaquine regardless of study participation. Ethical approval for the sample collection for parasite sequencing and genotypic analysis was obtained from the Comité d’Ethique National pour la Recherche en Santé, République de Guinée (National Ethics Committee for Health Research, Republic of Guinea). Following detection of P. falciparum infection, up to 5 mL of venous blood was collected in EDTA (ethylenediaminetetraacetic acid) vacutainers from consenting subjects and leukocyte depletion of each sample was performed using CF11 cellulose column filtration [21], and the erythrocytes were then frozen at −20 °C prior to shipment on dry ice to The Gambia. DNA extraction was carried out at the MRC Laboratories in The Gambia, using a QIAamp Blood Midi Kit (Qiagen) and DNA was quantified using a NanoDrop ND-1000 v3.3 spectrophotometer (NanoDrop Technologies, USA). Separate aliquots of DNA from each individual sample were then used for microsatellite genotyping and whole genome sequencing. The illumina paired-end, short-read genome sequencing of these samples has been previously described [22].

Microsatellite genotyping was performed on 95 isolates, out of the 100 for which genome sequence data were separately published [22]. Ten polymorphic microsatellite loci were genotyped using previously described heminested PCR methods [23], with dye-labelled internal primers as specified for each locus in a recent analysis of other samples from West Africa [16]. The PCR products were run electrophoretically on an ABI 3130XL Genetic analyzer alongside a GeneScan™ 500 LIZ internal size control standard. For each of the microsatellite loci for which multiple product sizes were detected within an isolate, the majority and minor alleles for each isolate were recorded [16]. The number of different genotypes detectable within an isolate was counted as the maximum number of alleles found at any individual microsatellite locus. The proportion of loci that had more than one allele detectable within each isolate was also counted as a separate measure of mixedness. For analyses of population-wide frequencies the majority allele at each microsatellite locus within each isolate was counted.

The whole genome illumina paired-end, short-read sequence data from the P. falciparum clinical isolates genotyped here are published and available at the European Nucleotide Archive as previously described [22]. Quantification of within-host parasite diversity within each individual isolate relative to the overall local population diversity was performed using the F _ws metric [19, 20]. For this process, using a set of 50,082 biallelic SNPs, allele frequencies were calculated at every SNP position for each isolate individually, with p and q representing the proportions of read counts for the minor and major alleles. All SNPs were then assigned to ten minor allele frequency (MAF) intervals representing the proportional frequency of the minor allele at each SNP across the Guinean population, with the ten equally sized intervals ranging from 0–5 % up to 45–50 %. Levels of within-host (H _w) and local parasite sub-population (H _s) heterozygosity for each SNP were calculated as H _w = 2*p _w*q _w and H _s = 2*p _s*q _s. The mean H _w and H _s of each MAF interval were then computed from the corresponding heterozygosity scores of all SNPs within that particular interval. The resulting—plot of H _w against H _s for each isolate was produced and a linear regression model was used to determine a value for the gradient H _w/H _s−, with F _ws = 1−(H _w/H _s). All F _ws analyses were performed using custom scripts in R.

F _ws indices were then derived from sampling a small number (between one and 20) of randomly selected SNPs and compared with the genome-wide indices previously determined. A Pearson’s correlation coefficient with the genome-wide SNP estimate of F _ws was determined across all isolates. For each limited SNP selection of n SNPs, 100 random samples of n SNPs were analysed and the mean Pearson’s coefficient observed for these was calculated. This analysis was repeated and comparisons performed between sets of SNPs with different ranges of overall minor allele frequencies.

A ten-locus microsatellite analysis was used to generate genotype profiles for each of the 95 Guinean P. falciparum clinical isolates (Additional file 1). A mean of 2.8 different genotypes was detected per isolate. Multiple genotypes were detected in all except one of the isolates, with 45 isolates (47.4 %) having two genotypes, 28 (29.4 %) having three, ten (10.5 %) having four, and 11 isolates (11.5 %) having five or more genotypes (Fig. 1a). Within each isolate, the numbers of loci that had multiple alleles detected had a wide range, with a mean of 4.4 out of the ten loci (Fig. 1b). The majority allele at each locus within each of the isolates was counted to assess overall population frequencies, which were similar to those seen in a sample of isolates taken from this area in the previous year [16] (Additional file 2).

Genome-wide SNP data for each of the isolates were used to generate within-host F _ws fixation indices, for comparison to the microsatellite profiling of the same isolates. Using the illumina short-read sequence reads to estimate relative frequencies of SNP alleles within each isolate, the within-isolate diversity (H _w) was derived, and the gradient of H _w/H _s (where H _s is the diversity in the entire local population sample) was calculated to derive the F _ws index for each isolate (with a range from zero indicating maximal possible diversity, to 1.0 indicating no observed diversity within an isolate). Over all 95 clinical isolates, the mean genome-wide F _ws index was 0.79, with a range of 0.16 to 1.00 (Additional file 1). Of these, 50 isolates (52.6 %) had an F _ws index approaching 1 (values of >0.95) indicating that they each contain a single predominant haploid genome sequence, with any additional genotypes being rare or absent, or closely related to the predominant genotype. Conversely, 45 of the clinical isolates (47.4 %) had an F _ws index of <0.95 and were thus clearly genotypically mixed infections. No significant correlation was seen between average genome sequence read mapping depth and F _ws index (Pearson’s r = 0.02), indicating that there was no bias in estimating within-host diversity due to sequence coverage.

Microsatellite typing allows an assessment of the proportion of loci that have multiple alleles within an infection, as well as a minimum estimate of the number of different parasite genotypes (the highest number of alleles detected at any locus). As expected, there was a highly significant negative correlation between the F _ws fixation index derived from the SNP analysis and the number of microsatellite loci with multiple alleles detected within each infection (Pearson’s r = −0.88, P < 0.001) (Fig. 2). There was also a highly significant negative correlation between the isolate F _ws index and the number of different genotypes within each infection detected by microsatellite typing (Pearson’s r = −0.80, P < 0.001). However, many infections that appeared to be unmixed on the basis of the genome sequence data (having an F _ws index of >0.95) were shown to have multiple genotypes as assessed by PCR-based microsatellite genotyping, with some of these infections having multiple alleles at several of the microsatellite loci (Fig. 2; Additional file 1).

Calculation of F _ws indices as above was performed with genome-wide SNP data for each infection, although such data will be rarely available for new clinical samples. In order to investigate the potential of using small numbers of SNPs, data from up to 20 randomly chosen SNPs were used to estimate F _ws indices, and correlations with values derived from genome-wide SNP data were determined (Fig. 3). By selecting from all SNPs, regardless of allele frequency, correlations with the genome-wide estimate were modest (for 20 SNPs Pearson’s r < 0.71). However, by selecting SNPs with minor allele frequencies greater than a specified threshold (either <5, 10, 20, 30, or 40 %), much higher correlations were seen. Correlation for each selection began to plateau above ~ten SNPs (Fig. 3). For example, randomly choosing ten or more SNPs with minor allele frequencies of greater than 10 % gave correlations of r > 0.9 in comparisons with the genome-wide F _ws indices.