Plasmodium falciparum is evolving to escape malaria rapid diagnostic tests in Ethiopia

Abstract

In Africa, most rapid diagnostic tests (RDTs) for falciparum malaria recognize histidine-rich protein 2 antigen. Plasmodium falciparum parasites lacking histidine-rich protein 2 (pfhrp2) and 3 (pfhrp3) genes escape detection by these RDTs, but it is not known whether these deletions confer sufficient selective advantage to drive rapid population expansion. By studying blood samples from a cohort of 12,572 participants enroled in a prospective, cross-sectional survey along Ethiopia's borders with Eritrea, Sudan and South Sudan using RDTs, PCR, an ultrasensitive bead-based immunoassay for antigen detection and next-generation sequencing, we estimate that histidine-rich protein 2-based RDTs would miss 9.7% (95% confidence interval 8.5-11.1) of P. falciparum malaria cases owing to pfhrp2 deletion. We applied a molecular inversion probe-targeted deep sequencing approach to identify distinct subtelomeric deletion patterns and well-established pfhrp3 deletions and to uncover recent expansion of a singular pfhrp2 deletion in all regions sampled. We propose a model in which pfhrp3 deletions have arisen independently multiple times, followed by strong positive selection for pfhrp2 deletion owing to RDT-based test-and-treatment. Existing diagnostic strategies need to be urgently reconsidered in Ethiopia, and improved surveillance for pfhrp2 deletion is needed throughout the Horn of Africa.

Fig. 1. Distribution of P. falciparum-positive RDT results and discordant profiles suggestive of pfhrp2/3 gene deletions.

a, Aggregated results from both RDTs, CareStart Pf/Pv (HRP2/Pv-LDH) RDT and SD Bioline Malaria Ag P.f. (HRP2/Pf-LDH) RDT, displayed by region for all P. falciparum infections (n = 2,714). The ‘2 HRP2−, 1 Pf-LDH+’ discordant RDT profile indicates potential infection by pfhrp2/3-deleted P. falciparum. Triangles represent the enrolment sites, including 11 districts and the Kule refugee camp within the Itang district in Gambella. b, Percentage of study participants identified with P. falciparum infection by RDT who had the discordant RDT profile, by district.

Fig. 2. Study samples and assays performed.

Samples were randomly selected by EPHI for molecular and antigen assays and for sequencing to expedite analyses.

Fig. 3. Deletion profiling using MIP sequencing of pfhrp2 (chromosome 8), pfhrp3 (chromosome 13) and flanking regions applied to 375 field samples.

Samples are grouped by subtelomeric structural profile, with control strains denoted CT, as labelled along the right y axis. LC (low concentration) and HC (high concentration) controls include mixtures of 1% HB3, 10% DD2, 89% 3D7 strains at densities of 250 and 1,000 parasites per µl, respectively. Columns represent each MIP target segment, rows represent individual samples and the colour scale represents log₁₀(UMI) depth-of-coverage at each location. Columns are labelled by the midpoint of each probe’s target region.

Fig. 4. Comparison of MIP and WGS pfhrp2/3 deletion calls and breakpoint regions.

Among the 14 clinical samples subjected to both methods, each sample is represented by two adjacent rows representing WGS (top) and MIP (bottom) coverage results. WGS coverage is displayed as the log₁₀ median number of aligned reads per 1 kb window. MIP results are coloured by whether each probe captured its target, with intervening regions not targeted in the MIP panel uncoloured. Sample numbers (lab_ID) are provided at left. The locations of pfhrp2, pfhrp3 and flanking genes are shown in black with non-genic regions in grey. The asterisk indicates that it could not be matched with PCR and RDT data.

Fig. 5. Extended haplotype homozygosity and bifurcation diagrams.

a,b Extended haplotype homozygosity (bottom) and the bifurcation diagrams showing haplotype branching (top) centromeric to the pfhrp2 (a) and pfhrp3 (b) deletions based on MIP data. Vertical dashed lines indicate the centromeric end of deletions. No variant calls were made within the 15.5 kb region on chromosome 13 which is duplicated on chromosome 11, demarcated by the grey box (b). Mb, megabase.

Extended Data Fig. 1. Parasite density distribution.

Pfldh quantitative PCR (qPCR) results used to assess parasite density and determine which samples were eligible for pfhrp2/3 deletion genotyping using a series of PCR assays. Pfhrp2/3 deletions were only called in 610 samples with >100 parasites/µL (solid line).

Extended Data Fig. 2. Pfhrp2 and pfhrp3 PCR results by RDT profile.

A) Concordance between RDT profile and PCR pfhrp2/3 result for 602 P. falciparum samples with >100 parasites/µL. b) Pfhrp2/3 PCR results for participants with the discordant RDT profile and sufficient DNA for molecular analysis (n = 176), by study region.

Extended Data Fig. 3. Pfhrp2/3 PCR results versus qPCR parasite density.

Comparison of pfhrp2 (top) and pfhrp3 (bottom) PCR results versus pfldh qPCR parasite densities. Note that final pfhrp2/3 PCR genotyping calls were only made in samples with qPCR parasite densities >100 parasites/µL and a positive P. falciparum beta-tubulin real-time PCR result to avoid misclassification of deletions in the setting of low DNA target concentrations.

Extended Data Fig. 4. Successful MIP deletion calls versus qPCR parasite density.

Comparison of MIP call results and qPCR parasite densities suggests a project-specific threshold for MIP calling of approximately 925 p/µL of whole blood (dashed line).

Extended Data Fig. 5. Disease severity by subtelomeric structural profile.

Smoothed distribution of disease severity for each of the broader deletion breakpoint haplotypes along chromosomes 8 (A) and 13 (B) identified by MIP genomic enrichment among MIP samples with matching clinical data (n = 338). The total number of symptoms with which participants presented was used to estimate disease severity, with all participants evaluated for: fever, headache, joint pain, feeling cold, nausea, and loss of appetite. Profile chr13-P2 was excluded due to its small sample size (n=1).

Extended Data Fig. 6. Chromosome 13 telomeric-end coverage (aligned reads/locus) plot of WGS control strains and 25 published Ethiopian genomes from 2013–2015 (MalariaGEN).

The location of pfhrp3 is indicated by vertical red lines. Inset showing pfhrp3 and 1 kilobase flanking genomic region. Large subtelomeric deletions containing pfhrp3 are apparent in the laboratory strain HB3, as well as 11 samples from Ethiopia.

Extended Data Fig. 7. Chromosome 8 telomeric-end coverage (aligned reads/locus) plot of WGS control strains and 25 published Ethiopian genomes from 2013–2015 (MalariaGEN).

The location of pfhrp2 is indicated by vertical red lines. Inset showing pfhrp2 and 1 kilobase flanking genomic region. Large subtelomeric deletions compared to the reference strain 3D7 are apparent in most samples. None of the deletions involve pfhrp2 except the DD2 strain.

Extended Data Fig. 8. Extended haplotype homozygosity centromeric to chromosome 8 subtelomeric structural profiles P1-P4 using MIP data.

The only pfhrp2-deleted profile (chr8-P4) showed sustained EHH, whereas EHH quickly broke down for the pfhrp2-intact profiles (P1-P3). A vertical dashed line on the right marks the centromeric end of the profile 4 (pfhrp2) deletion.

Extended Data Fig. 9. Extended haplotype homozygosity (bottom) and the bifurcation diagrams showing haplotype branching (top) centromeric to the pfhrp2 (A) and surrounding pfhrp3 (B) deletions based on WGS data.

Vertical dashed line indicating the centromeric end of the chromosome 8 deletion (a). Gray box demarcating the chromosome 13 deletion (b). Abbreviations: Mb, mega-base.

Extended Data Fig. 10. Extended haplotype homozygosity centromeric to chromosome 13 subtelomeric structural profiles P1, P3 and P4 using MIP data.

Chr13-P2 profile was observed in only one sample and not included in the haplotype analysis. EHH quickly brown down for all profiles (P1: pfhrp3-intact, P3-P4: pfhrp3-deleted). A vertical dashed line on the right marks the centromeric end of the P3 and P4 deletions. No variants in the duplicated segment (gray box) were used in the EHH analysis.