Highly multiplexed molecular inversion probe panel in Plasmodium falciparum targeting common SNPs approximates whole-genome sequencing assessments for selection and relatedness

Abstract

Introduction: The use of next-generation sequencing technologies (NGS) to study parasite populations and their response and evolution to interventions is important to support malaria control and elimination efforts. While whole-genome sequencing (WGS) is optimal in terms of assessing the entire genome, it is costly for numerous samples. Targeted approaches selectively enriching for the sequence of interest are more affordable and have higher throughput but sometimes lack adequate information content for key analyses.

Methods: We have developed a highly multiplexed molecular inversion probe (MIP) panel (IBC2FULL) targeting 4,264 single-nucleotide polymorphisms (SNPs) with ≥5% minor allele frequency (MAF) in Sub-Saharan African regions from publicly available Plasmodium falciparum WGS (n = 3,693). We optimized the panel alone and in combination with antimalarial drug resistance MIPs in laboratory P. falciparum strains at different parasitemias and validated it by sequencing field isolates from the Democratic Republic of Congo, Ethiopia, Ghana, Mali, Rwanda, Tanzania, and Uganda and evaluating the population structure, identity-by-descent (IBD), signals of selection, and complexity of infection (COI).

Results: The new panel IBC2FULL consisted of 2,128 MIPs (containing 4,264 common SNPs) spaced by 5.1-18.4 kb across the entire genome. While these microhaplotypes were developed based on variations from Sub-Saharan African WGS data, 59.3% (2,529) of SNPs were also common in Southeast Asia. The MIPs were balanced to produce more a uniform and higher depth of coverage at low parasitemia (100 parasites/μL) along with MIPs targeting antimalarial drug resistance genes. Comparing targeted regions extracted from public WGS, we observed that IBC2FULL provided a higher resolution of the local population structure in Sub-Saharan Africa than current PCR-based targeted sequencing panels. For sequencing field samples (n = 140), IBC2FULL approximated WGS measures of relatedness, population structure, and COI. Interestingly, genome-wide analysis of extended haplotype homozygosity detected the same major peaks of selection as WGS. We also chose a subset of 305 high-performing MIPs to create a core panel (IBC2CORE) that produced high-quality data for basic population genomic analysis and accurate estimation of COI.

Discussion: IBC2FULL and IBC2CORE panels have been designed to provide an improved platform for malaria genomic epidemiology and biology that can approximate WGS for many applications and is deployable for malaria molecular surveillance in resource-limited settings.

Keywords: Plasmodium falciparum; genomic epidemiology; malaria; molecular inversion probe; molecular surveillance; targeted sequencing.

FIGURE 1

IBC2FULL panel design. (A) Summary of the IBC2FULL panel development pipeline. (B) Genome-wide distribution of microhaplotype loci targeted by the IBC2FULL panel. Each blue band represents a microhaplotype captured by IBC2FULL. LD, linkage disequilibrium; MAF, minor allele frequency; VCF, variant call format; Pf6, MalariaGEN’s Pf6 release (MalariaGEN et al., 2021).

FIGURE 2

IBC2FULL panel optimization. (A) Distribution of microhaplotype heterozygosity by probe. Heterozygosity was calculated from the Pf6 dataset in African field samples. (B) Sequencing depth variation by parasitemia represented by log₁₀ transformation of the unique molecular index (UMI) count across samples. Each dot indicates the UMI count of one probe. (C) UMI depth distribution of the unfiltered panel containing 2,490 probes after one and two NextSeq runs, and one run of the filtered and rebalanced panel with 2,128 probes remaining. Second NextSeq was performed to verify if the low read depth of some probes is due to their intrinsic performance rather than insufficient sequencing. After this additional sequencing, 362 probes still showed low read depth and were removed from the final panel. UMI depth normalized per million (UPM). (D) IBC2FULL and main antimalarial drug resistance genes showing similar sequencing depth (log10 transformation of UMI) after spiking with drug resistance MIPs. The test was conducted in lab controls at 1,000 parasites/μL.

FIGURE 3

Local population structure in Sub-Saharan Africa based on SNPs extracted from the Pf6 dataset. First three dimensions were displayed for (A) IBC2FULL; (B) whole-genome sequencing (WGS) after selecting SNPs with ≥1% minor allele frequency; (C) IBC2CORE; (D) MAD⁴HatTeR; (E) SpotMalaria panels (n = 3,693). The first three dimensions (Dim 1, 2, and 3) of the PCA with the proportions of variance explained are shown. Each dot represents a sample colored by region defined in Sub-Saharan Africa.

FIGURE 4

Measures of relatedness and signatures of positive selection across the genome in recent field isolates from Tanzania. (A) Correlation of pairwise identity-by-descent (IBD) sharing between IBC2FULL and whole-genome sequencing (WGS) (n = 52). (B) Direct comparison of IBC2FULL and WGS pairwise IBD. (C) Histogram showing sample pair distribution by IBD range for IBC2FULL versus WGS. (D) Concordance between IBD-based sample clustering by IBC2FULL and WGS data. Four clusters were found in total by each of the tools at IBD ≥0.5. Each dot represents a sample assigned to a cluster name on the x-axis, and the vertical lines link samples that belong to the same cluster. (E) Manhattan plot showing p-values (formula is shown on the y-axis where Φ represents the Gaussian distribution function) of the integrated haplotype homozygosity score (iHS) across the 14 chromosomes of Plasmodium falciparum for IBC2FULL versus WGS (n = 52). Major signals detected by both tools were similar and are displayed, including ama1, trap, PF3D7_0711500 (putative regulator of chromosome condensation), PF3D7_1035100 (unknown function), and PF3D7_1475900 (KELT protein). Each dot represents an SNP.

FIGURE 5

Complexity of infection and population structure in Africa based on molecular inversion probe (MIP) sequencing. (A) Comparison of the complexity of infection (COI) between IBC2FULL, IBC2CORE, and WGS. Both x- and y-axes show COI values but for different panels. Dots represent field isolates from Tanzania (n = 52), and the dot size is proportional to the number of samples with the same COI. (B) Prevalence of polygenomic infections (n = 140) in African countries studied using IBC2CORE. (C and D) Population structure in Africa based on MIP sequencing using recent field samples (n = 140) for IBC2FULL and IBC2CORE, respectively. First three dimensions are shown. Each dot represents a sample colored by country.