A molecular marker of artemisinin-resistant Plasmodium falciparum malaria

Abstract

Plasmodium falciparum resistance to artemisinin derivatives in southeast Asia threatens malaria control and elimination activities worldwide. To monitor the spread of artemisinin resistance, a molecular marker is urgently needed. Here, using whole-genome sequencing of an artemisinin-resistant parasite line from Africa and clinical parasite isolates from Cambodia, we associate mutations in the PF3D7_1343700 kelch propeller domain ('K13-propeller') with artemisinin resistance in vitro and in vivo. Mutant K13-propeller alleles cluster in Cambodian provinces where resistance is prevalent, and the increasing frequency of a dominant mutant K13-propeller allele correlates with the recent spread of resistance in western Cambodia. Strong correlations between the presence of a mutant allele, in vitro parasite survival rates and in vivo parasite clearance rates indicate that K13-propeller mutations are important determinants of artemisinin resistance. K13-propeller polymorphism constitutes a useful molecular marker for large-scale surveillance efforts to contain artemisinin resistance in the Greater Mekong Subregion and prevent its global spread.

Conflict of interest statement

The authors declare no competing financial interests. Readers are welcome to comment on the online version of the paper.

Figures

Extended Data Figure 1. SNP-calling algorithm and sequence and coverage of SNPs

a, SNP-calling algorithm of the whole-genome sequence comparison of F32-ART5 and F32-TEM. b, Sequence and coverage of SNPs in seven candidate genes differing in F32-TEM and F32 ART5.

Extended Data Figure 2. Geographic distribution of K13-propeller alleles in Cambodia in 2011–2012

Pie charts show K13-propeller allele frequencies among 300 parasite isolates in ten Cambodian provinces. Pie sizes are proportional to the number of isolates and the different alleles are colour-coded as indicated. The frequencies (95% confidence interval) of mutant K13-propeller alleles are: Pailin (95%, 88–99, n = 84), Battambang (93%, 87–99, n = 71), Pursat (89%, 67–99, n = 19), Kampot (83%, 52–98, n = 12), Kampong Som (71%, 29–96, n = 7), Oddar Meanchey (76%, 58–89, n = 33), Preah Vihear (16%, 3–40, n = 19), Kratie (71%, 44–90, n = 17), Mondulkiri (67%, 9–99, n = 3) and Ratanakiri (6%, 1–19, n = 35).

Extended Data Figure 3. Correlation between the frequency of wild-type K13-propeller alleles and the prevalence of day 3 positivity after ACT treatment in eight Cambodian provinces

The frequency of day 3 positivity is plotted against the frequency of wild-type K13-propeller alleles. Data are derived from patients treated with an ACT for P. falciparum malaria in 2010–2012 in eight Cambodian provinces (Extended Data Figure 2): Pailin (n = 86, 2011 WHO therapeutic efficacy study, artesunate-mefloquine); Pursat (n = 32, 2012 WHO therapeutic efficacy study, dihydroartemisinin-piperaquine); Oddar Meanchey (n = 32, 2010 NAMRU-2 therapeutic efficacy study, artesunate-mefloquine); Kampong Som/Speu (n = 7, 2012 WHO therapeutic efficacy study, dihydroartemisinin-piperaquine); Battambang (n = 18, 2012 WHO therapeutic efficacy study, dihydroartemisinin-piperaquine); Kratie (n = 15, 2011 WHO therapeutic efficacy study, dihydroartemisinin-piperaquine); Preah Vihear (n =19, 2011 WHO therapeutic efficacy study, dihydroartemisinin-piperaquine); Ratanakiri (n = 32, 2010 WHO therapeutic efficacy study, dihydroartemisinin-piperaquine). Spearman’s coefficient of rank correlation (8 sites): r = −0.99, 95% confidence interval −0.99 to −0.96, P < 0.0001.

Extended Data Figure 4. Schematic representation of homology between P. falciparumK13 and human KEAP1 proteins and structural 3D model of the K13-propeller domain

a, Schematic representation of the predicted PF3D7_1343700 protein and homology to human KEAP1. Similar to KEAP1, PF3D7_1343700 contains a BTB/POZ domain and a C-terminal 6-blade propeller, which assembles kelch motifs consisting of four anti-parallel beta sheets. b, Structural 3D model of the K13-propeller domain showing the six kelch blades numbered 1 to 6 from N to C terminus and colour-coded as in Supplementary Fig. 1. The level of amino-acid identity between the K13-propeller and kelch domains of proteins with solved 3D structures, including human KEAP1,, enabled us to model the 3D structure of the K13-propeller and to map the mutations selected under ART pressure (Extended data Table 5). The accuracy of the K13-propeller 3D model was confirmed by Modeller-specific model/fold criteria of reliability (see Methods). We predict that the K13-propeller folds into a 6-bladed β-propeller structure closed by the interaction between a C-terminal beta-sheet and the N-terminal blade,. The first domain has three β-sheets, the fourth one being contributed by an extra C-terminal β-sheet called β’1 in Supplementary Fig. 1. The human KEAP1 kelch propeller scaffold is destabilized by a variety of mutations affecting intra- or inter-blade interactions in human lung cancer and hypertension. The positions of the various mutations are indicated by a sphere, colour-coded as in Figs 2–4. The M476 residue mutated in F32-ART5 is indicated in dark grey. Like the mutations observed in human KEAP1,, many K13-propeller mutations are predicted to alter the structure of the propeller or modify surface charges, and as a consequence alter the biological function of the protein. Importantly, the two major mutations C580Y (red) and R539T (blue) observed in Cambodia are both non-conservative and located in organized secondary structures: a β-sheet of blade 4 where it is predicted to alter the integrity of this scaffold and at the surface of blade 3, respectively. The kelch propeller domain of KEAP1 is involved in protein–protein interactions like most kelch containing modules. KEAP1 is a negative regulator of the inducible Nrf2-dependent cytoprotective response, sequestering Nrf2 in the cytoplasm under steady state. Upon oxidative stress, the Nrf2/KEAP1 complex is disrupted, and Nrf2 translocates to the nucleus, where it induces transcription of cytoprotective ARE-dependent genes,. We speculate that similar functions may be performed by PF3D7_1343700 in P. falciparum, such that mutations of the K13-propeller impair its interactions with an unknown protein partner, resulting in a deregulated anti-oxidant/cytoprotective response. The P. falciparum anti-oxidant response is maximal during the late trophozoite stage, when haemoglobin digestion and metabolism are highest. Its regulation is still poorly understood and no Nrf2 orthologue could be identified in the Plasmodium genome.

Figure 1. Temporal acquisition of mutations in F32-ART5

F32-Tanzania parasites exposed to increasing artemisinin concentrations for 120 consecutive cycles were analysed by whole-genome sequencing at five time-points (red arrows). Loci mutated after a given number of drug-pressure cycles are shown (red boxes). The earliest time-points where three mutations were detected by PCR (black arrows) are indicated by † for PF3D7_1343700, * for PF3D7_0213400 and ‡ for PF3D7_1115700. Orange and green circles indicate RSA0–3 h survival rates for F32-ART5 and F32-TEM parasites, respectively (mean of 3 experiments each).

Figure 2. Survival rates of Cambodian parasite isolates in the RSA0–3 h, stratified by K13-propeller allele

Genotypes were obtained by mining whole-genome sequence data (n = 21) or sequencing PCR products (n = 28). Mutant parasites have significantly higher RSA0–3 h survival rates than wild-type parasites: wild type (n = 17, median 0.16%, IQR 0.09–0.24, range 0.04–0.51); C580Y (n = 26, median 14.1%, IQR 11.3–19.6, range 3.8–27.3, P < 10−6 for wild type versus C580Y, Mann–Whitney U test); R539T (n = 5, median 24.2%, IQR 12.6–29.5, range 5.8–31.3, P < 10−3 for wild type versus R539T); Y493H (51.4%); and I543T (58.0%). The RSA0–3 h survival rate (0.04%) of control 3D7 parasites is indicated by an asterisk.

Figure 3. Frequency of K13-propeller alleles in 886 parasite isolates in six Cambodian provinces in 2001–2012

Genotypes were obtained by sequencing PCR products from archived blood samples. All mutant alleles carry a single non-synonymous SNP (colour-coded, same colour codes as in Fig. 2 for wild type, C580Y, R539T, Y493H and I543T). Significant reductions (Fisher’s exact test) in wild-type allele frequencies were observed in Pailin, Battambang, Pursat and Kratie over time (see Methods).

Figure 4. Parasite clearance half-lives

a, Correlation of parasite clearance half-lives and K13-propeller alleles for parasite isolates in Pursat and Ratanakiri in 2009–2010. Wild-type parasites have shorter half-lives (median 3.30 h, IQR 2.59–3.95, n = 72) than C580Y (7.19 h, 6.47–8.31, n = 51, P < 10−6, Mann– Whitney U test), R539T (6.64 h, 6.00–6.72, n = 6, P < 10−6) or Y493H (6.28 h, 5.37–7.14, n = 21, P < 10−6) parasites. The half-life of C580Y parasites is significantly longer than that of Y493H parasites (P = 0.007). b, Correlation of parasite clearance half-lives, KH subpopulations and K13-propeller alleles for the same 150 parasite isolates. Half-lives are shown for Pursat (squares) and Ratanakiri (triangles) parasites, stratified by KH group and K13-propeller allele (colour-coded as in a). Median half-lives stratified by K13-propeller allele are KH1: wild type (2.88) and Y493H (6.77); KH2: C580Y (7.13) and Y493H (4.71); KH3: wild type (3.65), C580Y (8.73) and R539T (6.65); KH4: Y493H (6.37); and KHA: wild type (4.01), C580Y (7.09), Y493H (6.18) and R539T (5.73).