Plasmodium falciparumpopulations, transmission dynamics and infection origins across Papua New Guinea

Abstract

The successful implementation of pathogen genomic surveillance demands rapid, low-cost genotyping solutions for tracking infections. Here we demonstrate the capacity of single nucleotide polymorphism (SNP) barcodes to generate practical information for malaria surveillance and control. The study was conducted in Papua New Guinea (PNG), a country with a wide range of malaria transmission intensities. A panel of 191 candidate SNPs was selected from 5786 SNPs with minor allele frequency greater than 0.1, identified amongst 91Plasmodium falciparumgenomes from three provinces of PNG. We then genotyped 772P. falciparumisolates from a 2008 nationwide malaria indicator survey and 31 clinical infections from an outbreak of unknown origin. We assessed the performance of SNP panels with different allele frequency characteristics, and measured population diversity, structure and connectivity using both whole genome data and the SNP barcode. The full SNP barcode captured similar patterns of population structure evident with 5786 ‘whole genome’ SNPs. Geographically informative SNPs (iSNPs,FST>0.05) show increased population clustering compared to the full barcode whilst randomly selected SNPs (rSNPs) and SNPs with similar allele frequencies (FST<0.05) amongst different countries (universal, uSNPs) or local PNG populations (balanced, bSNPs) indicated little clustering. Applied to samples from all endemic areas of PNG, this barcode identified variable transmission dynamics, and eight major populations. Genetic diversity was high throughout most areas, however, in the southern region, isolates were either closely related, suggesting highly inbred or near-clonal populations; or, they shared ancestry with other parasite populations, consistent with importation. Applied to outbreak samples, only the full barcode, the iSNPs and bSNPs distinguished between locally acquired and imported infections. The full barcode contains more than 100 SNPs prevalent in other endemic regions, allowing the transfer of this tool to other settings. SNP barcodes must be validated in local settings to ensure they capture the diversity and population structure of the target population. Subsets of geographically informative SNPs will be essential for predicting geographic origins but may bias analyses of population structure and gene flow if used alone.AUTHOR SUMMARYPathogen genomic surveillance is a sensitive approach for mapping pathogen transmission dynamics to support decisions about how to prevent and control infectious diseases. High throughput genotyping tools known as single nucleotide polymorphism (SNP) barcodes are used to measure relationships between individuals and connectivity between populations, however, the barcode design may influence these results. We used whole genome sequences from the malaria parasitePlasmodium falciparumto design a barcode that captures the diversity both within and between parasite populations of Papua New Guinea (PNG), where transmission is variable amongst provinces. By investigating the performance of different panels of SNPs, we show that validation for use in the target population is crucial to correctly identifying population genetic structure. Applying the validated SNP barcode on hundreds of samples from all endemic provinces of PNG, high levels of variability in local transmission dynamics, and regions of population subdivision were observed. Some geographic areas show evidence of interrupted transmission, and the substantial genetic differentiation between the northern, eastern and island populations, presents an opportunity to design targeted, subnational control efforts. Application of the barcode to outbreak samples classified cases into imported and locally acquired infections, with substantial local transmission indicating control efforts were not sufficient to prevent the spread of infections. SNP barcodes are useful tools that can be used to supplement existing malaria surveillance tools however careful validation of their effectiveness in different settings is recommended.

Systematic bias in malaria parasite relatedness estimation

“…The error structure in sparse data HMM estimates is heteroskedastic: underestimation is most pronounced for parasite pairs with low relatedness. At the other end of the spectrum, genomic epidemiology applications typically focus on highly related parasite pairs, often using a threshold-based approach [9,14,16,48,49,50]. Since relatedness is systematically underestimated using sparse data, false positives (pairs with low levels of relatedness that appear highly related) are relatively rare.…”

“…in Figure 4C. When relatedness thresholds are sufficiently high, sparse marker data in the intermediary regime suffice; sensitive classification for low relatedness thresholds, however, requires higher resolution data [50]. We are reluctant to posit marker density thresholds necessary to sensitively identify highly-related parasite pairs in generality, because the degree of linkage structure within a set of sampled parasites depends on the distribution of shared IBD segment lengths -which, in turn, is driven by demographic processes that we neither fully understand, nor control.…”