Effects of transmission reduction by insecticide-treated bed nets (ITNs) on parasite genetics population structure: I. The genetic diversity of Plasmodium falciparum parasites by microsatellite markers in western Kenya

This study was part of a two-phase community-based ITN trial conducted in Asembo area (Bondo District) of western Kenya, where malaria is holoendemic. The design and characteristics of the ITN trial have been detailed elsewhere [31‐33]. Briefly, biannual population censuses and annual cross-sectional surveys were conducted in 60 villages in the 200 km² trial area during the rainy season of March to May between 1996 and 2001 to determine the effects of ITNs on malaria morbidity and all cause mortality in children below five years of age. During each cross sectional survey, blood samples were collected from children and parasitological, clinical, and demographic information were documented. Entomological monitoring of Anopheles density and sporozoite infection rates were conducted regularly throughout both phases of the trial. Approximately 98% of malaria infections were due to P. falciparum in the trial area. Before the ITN trial, entomological inoculation rate (EIR) was reported at 61.3 infective bites per person per year [34] and prevalence of parasitaemia was about 70% in children aged less than five years [31]. After the introduction of ITNs, it was estimated that ITNs reduced transmission by 90% at the early stage of the two-phase trials [6]. The EIR and prevalence of parasitaemia in May 2001 were 1.3 infective bites per person per year and about 34% in children less than five years old, respectively [33]. For the study presented here, the samples collected in the 1996 survey, just prior to ITN introduction for baseline measurements, and the samples from the 2001 follow-up survey, five years post-ITN intervention were utilized. Microscopically confirmed malaria positive blood samples were randomly selected in a subset of villages from the 1996 survey and were further matched by the villages in 2001 survey. Calculation of sample size was based on a hypothesized significant difference in overall heterozygosity (H_e) between baseline and post-ITN parasite populations using a confidence level at 95% and margin of error at 5% [35]. Assuming a H_e of 0.7 observed in high transmission areas [17] for baseline and a conservative change in H_e to 0.5 observed in areas with medium transmission intensity [17] for ITN post intervention, a sample size of approximately 63 (+ 10-15%) was deemed adequate allowing for failure in laboratory testing. In total, a sampling frame of 69 samples from baseline and 74 samples from the post-ITN survey was achieved for this study. Parasite genomic DNA extraction from the stored blood samples was by the QIAamp DNA Mini Kit (Qiagen, CA, USA). Extracted DNA was stored at -20C until use. The study protocol was approved by the Ethical Review Committee of the Kenya Medical Research Institute, Nairobi, Kenya, the Institutional Review Board of Michigan State University, and the Institutional Review Board of Centers for Disease Control (CDC) Atlanta, Georgia.

Eight single copy microsatellite markers (MS) were used for genotyping as listed in Table 1. Broadly, the microsatellites included: 1) five putatively neutral MS (Poly-α, PfPK2, ADL, TAA60, and TAA 109), 2) one MS (Pfg377) linked to the protein gene exclusively expressed during maturation of gametocytes transmission stage of parasites, and 3) two MS (EBP and P195) linked to the genes of asexual stage antigens under possible natural immune selection [36]. Among these MS markers, Poly-α, PfPK2 and Pfg377 are in the coding region [37]. The selected neutral MS markers were described earlier and have been used previously to study changes in genetic structure of Plasmodium parasites [17, 27, 37]. All amplifications were carried out using single reaction PCR with thermocycling conditions described elsewhere [38]. Fluorescent-labelled primers incorporated with either HEX (green) or FAM (Blue) dyes were used and the PCR products read on ABI (Applied Biosystems 3100) capillary sequencer. GeneMapper software (ABI) was used to automate measurement of microsatellite base-pair length and quantify peak height. Allele identity per locus was obtained directly after allocation of all peaks above 200 fluorescent units. Multiple alleles were quantified based on a method described previously [17] with identification of minor alleles set at peak heights of ≥30% of the predominant allele. Where amplification failed for any of the microsatellites, data was reported as missing and not used for haplotype definition.

Since the ITN intervention is one of long-term goals for malaria control programmes, the sampling strategy in the current study aimed at testing changes in genetic diversity of P. falciparum after an extended use of ITN while minimizing any spatial effects on the population. Potential biases were envisioned in the data analyses as allele frequencies might change over time in finite populations [39]. However, the prediction in this study was that the ITN-mediated transmission reduction would precipitate changes in the parasite population from panmictic (higher parasite diversity, expected for high transmission intensity) to a more clonal structure (lower parasite diversity, expected for low transmission intensity). Changes in genetic diversity of the parasites between the baseline and post-ITNs were, therefore, assessed by quantifying and comparing multiple infections, allele frequencies, H_e, LD, and F_ST.

Initial microsatellite data checking and data conversion was done using Excel Microsatellite Tool Kit, an add-in programme used to format raw microsatellite data in Microsoft Excel for consequent use in different genetics softwares [40]. Analysis of multiple infections was based on both predominant and minor alleles. Outcome measures for this analysis were the proportion of infections with more than one allele, the mean allele count for individual microsatellite loci, the overall proportion of infections with at least two alleles and the overall mean of the highest number of allele count detected by any of the microsatellites. The difference in proportions of infections with at least two alleles and the difference in the mean allele count between the baseline and post-intervention parasite populations were assessed using Pearson's chi-square and Wilcoxon tests respectively. Conversely, only the predominant allele defined by the highest peak in each electropherogram per locus was used to analyze the allele frequency and allele richness, H_e, LD, and F_ST. There are possible biases from using predominant allele techniques to determine allele frequency in multiple infections [41, 42]. However, the method applied here was shown to be appropriate in previous studies using some of the microsatellite targets selected in this study [17, 37, 43, 44].

The infinite allele model which is more appropriate for analysis of the complex patterns observed in microsatellite loci in P. falciparum was used for genetic analysis [45]. Allele number, frequency and richness per locus in each parasite population were obtained by FSTAT2 [46]. The measure for allele number, frequency and richness was to illustrate the composition and distribution of alleles in the population. Unbiased H_e at each locus was calculated as H_e = [n/(n-1)][1-Σⁿ_{i = 1}p ² _i ] where n is the number of isolates sampled and p _i is the frequency of the i th allele while sampling variance for H_e (V_s(He)) was calculated as V_s(He) = 2(n-1)/n³[2(n-2)][Σpi³-(Σpi²)²] [17, 22]. The difference in single locus heterozygosities between the two parasite populations was tested with standard error (SE) of the sampling variance by the method described earlier [22] using the z absolute values to obtain the p-values.

The prediction of the ITN-mediated decrease in transmission intensity and consequent reduction in genetic diversity was a possible increase in LD in the post-ITN parasite population. LD measures the degree of association between or among gene loci under the null hypothesis of no association [47]. As such,

individual pair-wise LDs for baseline and post-ITN parasite populations were first obtained respectively by the Fisher's exact test adapted for haploid data using ARLEQUINS 3.11 programme [48]. Multilocus LDs were further assessed from the overall index of association (I ^S _A ) using LIAN programme [49] for both baseline and post-ITNs respectively. The multilocus LD test measures non-random association among all loci. The test compares the variance of differences at LD with the variance expected in LD derived from 10,000 simulated data sets, H₀: V_D = V_e. A significant LD was when the observed variance (V_D) was greater than expected (V_e). Index of association was expressed as I ^S _A = (V_D/V_e-1)(r-1) where r is the number of loci tested. A 95% confidence limit was determined by Monte Carlo simulation.

In order to further assess genetic diversity before and after ITNs intervention, genetic differentiation was tested using the F_ST estimator [50] implemented by FSTAT programme [46]. F_ST is a comparison of the sum of genetic variability within and between populations based on allele frequency differences in populations. Interpretation of the F_ST values at each locus was based on three categories defined earlier as no differentiation (0), low genetic differentiation (0 > 0.05), moderate differentiation (0.05-0.15) and great differentiation (0.15-0.25) [51].

When multiple tests were conducted, significant levels of p-value for the comparisons were adjusted using Bonferroni's correction [52].

The extent of multiple infections was assessed based on the proportion of multiple alleles and mean allele counts (Table 2). Overall, more than 90% of samples from both baseline and post-ITN surveys had at least two or more alleles detected by any of the microsatellites targets. There was no significant difference in the overall proportions of multiple alleles between the baseline and post-ITN parasite populations. However, analysis of individual loci revealed that the Pfg377 marker had a significantly lower proportion of multiple alleles in the baseline (30%) compared to post-ITN parasite population (53%). All other loci showed similar proportions of multiple alleles in the two parasite populations.

Results from the mean allele counts showed that PfPK2 had significantly higher allele counts in the baseline compared to that in the post-ITN survey. In contrast, Pfg377 had a relatively higher mean allele count (1.62 ± 0.09) in the post-ITN survey than in the baseline parasite population (1.36 ± 0.07) although it was not a statistically significant difference (p = 0.0076) after Bonferroni correction. There was no significant difference in the mean allele counts for all other microsatellite markers between the two parasite populations. Overall, mean allele count was similar in the baseline parasite population (3.4 ± 0.15) compared to the post-ITN parasite population (3.1 ± 0.12) (Table 2).

The allele size, number, frequency and richness for each MS shown in Figure 1 and Table 3 illustrated the composition and distribution of alleles. There were no differences between the allele number and richness as sample numbers used in the baseline and post-ITN surveys did not differ significantly for each locus. The allele number per locus ranged from a minimum of 4 (Pfg377) to maximum of 18 (Poly-α) in the baseline parasite population and from 3 (Pfg377) to 17 (Poly-α) in the post-ITN parasite population. This suggests marked variation between loci in the baseline and post-ITN surveys and was further tested by H_e. Overall, H_e was high and similar in the baseline (0.75 ± 0.072) and post-ITN parasite populations (0.79 ± 0.038). However, H_e of individual loci showed extensive range in gene diversity (Table 3). Poly-α locus had the highest level of H_e while Pfg377 locus showed the lowest in both parasite populations. Notably, Pfg377 locus had significantly lower H_e in the baseline parasite population compared to the H_e in the post-ITN survey (p = 0.044).

Pair-wise LD for each individual MS locus for the baseline and post-ITN parasite populations was assessed. Of the 28 possible pair-wise tests LD was highly significant in a total of 14 pairs in the baseline parasite population while only 6 pairs were significant in the post-ITN population (p = 0.0018) as shown in Table 4. Specifically, in the baseline survey MS P195 locus (Chr9) and TAA60 (Chr9) had strong LD with loci located on different chromosomes including Poly-α (Chr4), Pfg377 (Chr12), PfPK2 (Chr12), EBP (Chr13) and ADL (Chr13). This association was however, reduced in the post-ITN population where 6 pairs showed significant LD. Only LD between P195 and ADL, and between TAA60 and EBP loci were maintained while the remaining four loci pairs were new associations in the post-ITN survey.

The multilocus analysis showed significant LD in the baseline population with a variance in difference (V_D) of 1.21 and variance expected (V_e) of 1.10 (p = 0.01) compared to non-significant LD in the post-ITN population where V_D and V_e were 1.183 and 1.180 respectively (p = 0.57) as shown in Table 5. Consequently the index of association (I ^S _A ) was significant in the baseline survey (0.016) compared to the post-ITN survey (0.0003). The multilocus LD results coupled with the significant pair-wise LD observed in individual microsatellites suggest that the existing non-random association between the MS loci in the baseline was broken in the post-ITN parasite population.

Results of the overall and locus specific genetic differentiation between the baseline and post-ITN surveys are shown in Table 6. Overall F_ST was low at 0.027. In the single locus F_ST, both Pfg377 and TAA60 markers showed moderate genetic differentiation (F_ST = 0.117 and 0.137) respectively while all other remaining six loci had little differentiation. The results suggest that a moderately significant genetic variability at Pfg377 and TAA60 arose from differences between the two parasite populations while the majority of genetic variability remains within each population. Conversely, low genetic differentiation in the other six markers suggested much of the genetic variability resulted from within each parasite population.