Microgeographic Epidemiology of Malaria Parasites in an Irrigated Area of Western Kenya by Deep Amplicon Sequencing

Abstract

To improve food security, investments in irrigated agriculture are anticipated to increase throughout Africa. However, the extent to which environmental changes from water resource development will impact malaria epidemiology remains unclear. This study was designed to compare the sensitivity of molecular markers used in deep amplicon sequencing for evaluating malaria transmission intensities and to assess malaria transmission intensity at various proximities to an irrigation scheme. Compared to ama1, csp, and msp1 amplicons, cpmp required the smallest sample size to detect differences in infection complexity between transmission risk zones. Transmission intensity was highest within 5 km of the irrigation scheme by polymerase chain reaction positivity rate, infection complexity, and linkage disequilibrium. The irrigated area provided a source of parasite infections for the surrounding 2- to 10-km area. This study highlights the suitability of the cpmp amplicon as a measure for transmission intensities and the impact of irrigation on microgeographic epidemiology of malaria parasites.

Keywords: Plasmodium falciparum; amplicon sequencing; infection complexity; irrigation; microgeographic epidemiology; transmission intensity.

Figure 1.

Study site localities in relation to the Oluch irrigation scheme, Kenya. Abbreviations: Hi, high; Lo, low; Me, medium.

Figure 2.

Polymerase chain reaction (PCR) positivity rate among transmission risk zones and season, Kenya. Square indicates overall positivity rate among clusters by risk zone. Lines indicate 95% confidence intervals. Asterisks indicate statistical significance (χ ² test with Bonferroni correction, P ≤ .05). February–March occurs after the short rains (October–December), and June–July occurs after the long rains (March–May).

Figure 3.

Simulated sample sizes to assess power of hypothesis testing by molecular markers, Kenya. A, Mean unadjusted P value by simulated sample size. Dashed red line indicates α = .167. Dotted red line indicates α = .05. B, Minimum sample size for unadjusted P value to be less than α in ≥80% of simulations. C, Empirical P value compared to distribution of simulated P values at equivalent sample sizes. Diamonds indicate empirical values. Box plots indicate distribution of simulated values. Lower and upper hinges correspond to the first and third quartiles. All results are based on Wilcoxon rank-sum tests among 10 000 simulations per sample size. Sample sizes were simulated from Poisson distributions with λ = empirical average multiplicity of infection for a given molecular marker and transmission zone.

Figure 4.

Genetic indices among transmission risk zones by cpmp amplicon sequencing. A, Multiplicity of infection among risk zones. Dots indicate individual data points. Triangles indicate average values. B, Nucleotide diversity among risk zones. C, Linkage disequilibrium among risk zones. Squares indicate average values among 1000 coalescent simulations. Lines indicate 95% confidence interval from simulations. Asterisks indicate statistical significance.

Figure 5.

Allele sharing among parasite isolates by cpmp amplicon sequencing, Kenya. A, Highly related isolates by geographic locality. Dots indicate individual parasite isolates. Lines connect isolates from different clusters that share >98% of polymorphic alleles. Thin lines indicate 98%–99.9% allele sharing; thick lines indicate 100% allele sharing. B, Highly related pairs by risk zones. Bar height indicates the proportion of highly related pairs among all isolate pairs for each risk zone combination. Labels indicate the total count number of highly related pairs for each risk zone combination. For x-axis labels, letters indicate risk zone combinations (H, high; L, low; M, medium). Asterisks indicate statistical significance (χ ² test with false discovery rate correction, P ≤ .05). Four pairs of highly related isolates originated from the same cluster, and so are not visible on the map: high/high (1 pair), low/low (2 pairs), and medium/medium (1 pair). C, Histogram of highly related pairs by geographic distance between pairs. Solid green line indicates density plot. Vertical dashed line indicates mean geographic distance between isolate pairs. “All” indicates geographic distances between all isolate pairs. Geographic distances were not significantly different between allele sharing categories (unpaired t test with Bonferroni correction, P ≥ .13 for all comparisons).

Figure 6.

Migration rates among risk zones by cpmp sequencing, Kenya. Thickness of arrows is proportional to the estimated mutation-scaled migration rate (M). Diameter of circles is proportional to the estimated mutation-scaled population sizes (θ). Numbers indicate the estimated effective number of migrants per generation (Nm, calculated as θM/2) between populations. Black arrows indicate predominant direction of migration.