Pf7: an open dataset of Plasmodium falciparum genome variation in 20,000 worldwide samples

Abstract

We describe the MalariaGEN Pf7 data resource, the seventh release of Plasmodium falciparum genome variation data from the MalariaGEN network. It comprises over 20,000 samples from 82 partner studies in 33 countries, including several malaria endemic regions that were previously underrepresented. For the first time we include dried blood spot samples that were sequenced after selective whole genome amplification, necessitating new methods to genotype copy number variations. We identify a large number of newly emerging crt mutations in parts of Southeast Asia, and show examples of heterogeneities in patterns of drug resistance within Africa and within the Indian subcontinent. We describe the profile of variations in the C-terminal of the csp gene and relate this to the sequence used in the RTS,S and R21 malaria vaccines. Pf7 provides high-quality data on genotype calls for 6 million SNPs and short indels, analysis of large deletions that cause failure of rapid diagnostic tests, and systematic characterisation of six major drug resistance loci, all of which can be freely downloaded from the MalariaGEN website.

Keywords: data resource; genomic epidemiology; genomics; malaria; plasmodium falciparum.

Figure 1.. Geographic distribution of sampling locations and population structure.

( A) Map shows the centres of the 97 first-level administrative divisions from where samples were collected. Points are coloured according to the major sub-population to which the location is assigned ( Table 1). ( B) First two components of a genome-wide principal coordinate analysis. The first axis (PC1, 17.6% of variance explained) captures the separation of African and South American from Asian and Oceanian samples. The second axis (PC2, 2.4% of variance explained) captures finer levels of population structure particularly in the eastern SE Asia population. Each point represents a QC pass sample and the colour legend is the same as in ( A). ( C) First and fifth (0.7% of variance explained) components of a genome-wide principal coordinate analysis. Here there is an approximate mapping between the principal components and the geographic location (latitude and longitude).

Figure 2.. Heterogenity of chloroquine resistance in west Africa.

Inferred resistance levels to chloroquine between 2013 and 2016 in different administrative divisions within West Africa. We only include locations for which we have at least 25 samples with an unambiguous inferred chloroquine resistance phenotype. Note the very different chloroquine resistance profiles in nearby locations, e.g. Volta, Ghana vs Atlantique, Benin.

Figure 3.. Newly emerging Dd2 background mutations in crt.

(a) Top panel - frequency of different haplotypes with a genetic background identical to the lab strain Dd2. Dd2 is derived from an isolate taken from a patient in Indochina in 1980. Middle panel - breakdown of samples by major sub-population for each haplotype. Lower panel - amino acid mutations in the haplotypes (with respect to 3D7 reference). Mutations found in the Dd2 haplotype are shown in grey, all other mutations are shown in black. (b) Bar plots showing changing frequency of newly emerging Dd2 background crt haplotypes in different locations in the eastern SE region. Newly emerging Dd2 background haplotypes are defined as all haplotypes that have all mutations seen in Dd2 plus additional mutations.

Figure 4.. Analysis of c-terminal of csp.

(a) Upper panel - frequency of different haplotypes of c-terminal of csp. Haplotypes found in some lab strains are named and highlighted in red. Haplotypes are ordered as per lower panel. Lower panel - global mean distance (number of amino acid differences) to all other haplotypes. (b) Histograms of number of amino acid differences between samples in each major sub-population and the 3D7 haplotype (upper plot) and Dd2 haplotype (lower plot).

Figure 5.. HRP deletion breakpoints.

We see five different breakpoints resulting in the deletion of hrp2. Four of these are within exon 2 of the gene whereas the fifth is found between hrp2 and the pseudogene PF3D7_0831750. For all five events we see evidence of telomeric healing from reads that contain part Pf3D7_08_v3 sequence and part telomeric repeat sequence (GGGTTCA/GGGTTTA). We see 16 different breakpoints resulting in the deletion of hrp3. For fourteen of these we see evidence of telomeric healing. Note that many of these events result in the deletion of other genes in addition to hrp3. For twenty samples from Cambodia and a single sample from Vietnam we see evidence of a recombination with chromosome 5 which results in a hybrid chromosome comprising mostly chromosome 13 sequence but a small inverted section of an internal portion of chromosome 5 containing the gene mdr1. We also see evidence of a recombination with chromosome 11 which results in a hybrid chromosome comprising mostly chromosome 13 sequence but also a section of the 3’ end of chromosome 11. This is the most common deletion type, being seen in 151 samples from 14 different countries. Because the recombination occurs between highly similar sequences of a set of three orthologous ribosomal RNA genes found on both chromosomes, it is not possible to identify the exact breakpoints.