Association of mutations in the Plasmodium falciparum Kelch13 gene (Pf3D7_1343700) with parasite clearance rates after artemisinin-based treatments-a WWARN individual patient data meta-analysis

Abstract

Background: Plasmodium falciparum infections with slow parasite clearance following artemisinin-based therapies are widespread in the Greater Mekong Subregion. A molecular marker of the slow clearance phenotype has been identified: single genetic changes within the propeller region of the Kelch13 protein (pfk13; Pf3D7_1343700). Global searches have identified almost 200 different non-synonymous mutant pfk13 genotypes. Most mutations occur at low prevalence and have uncertain functional significance. To characterize the impact of different pfk13 mutations on parasite clearance, we conducted an individual patient data meta-analysis of the associations between parasite clearance half-life (PC_1/2) and pfk13 genotype based on a large set of individual patient records from Asia and Africa.

Methods: A systematic literature review following the PRISMA protocol was conducted to identify studies published between 2000 and 2017 which included frequent parasite counts and pfk13 genotyping. Four databases (Ovid Medline, PubMed, Ovid Embase, and Web of Science Core Collection) were searched. Eighteen studies (15 from Asia, 2 from Africa, and one multicenter study with sites on both continents) met inclusion criteria and were shared. Associations between the log transformed PC_1/2 values and pfk13 genotype were assessed using multivariable regression models with random effects for study site.

Results: Both the pfk13 genotypes and the PC_1/2 were available from 3250 (95%) patients (n = 3012 from Asia (93%), n = 238 from Africa (7%)). Among Asian isolates, all pfk13 propeller region mutant alleles observed in five or more specific isolates were associated with a 1.5- to 2.7-fold longer geometric mean PC_1/2 compared to the PC_1/2 of wild type isolates (all p ≤ 0.002). In addition, mutant allele E252Q located in the P. falciparum region of pfk13 was associated with 1.5-fold (95%CI 1.4-1.6) longer PC_1/2. None of the isolates from four countries in Africa showed a significant difference between the PC_1/2 of parasites with or without pfk13 propeller region mutations. Previously, the association of six pfk13 propeller mutant alleles with delayed parasite clearance had been confirmed. This analysis demonstrates that 15 additional pfk13 alleles are associated strongly with the slow-clearing phenotype in Southeast Asia.

Conclusion: Pooled analysis associated 20 pfk13 propeller region mutant alleles with the slow clearance phenotype, including 15 mutations not confirmed previously.

Fig. 1

Study site locations. The numbers in the pins correspond to the study sites listed in Table 1

Fig. 2

Study profile. Definitions for specific exclusions are listed at the right of the figure detailing the number of isolates included in each analysis. Unsatisfactory fit was defined as pseudo-R² statistics < 0.8. Insufficient parasite data includes patients with too few observations to fit the model and patients with only daily counts

Fig. 3

Distribution of parasite clearance half-life by individual wild type or pfk13 mutant codons in isolates from Africa. Left panel (a) shows violin plots for wild type or mutant codons with five or more isolates. The number of individual isolates tested is at the right of each violin plot. The right panel (b) shows dot plots for mutant codons with < 5 isolates. The red line shows a half-life of 5.5 h. The median is shown as a green circle, the red bar corresponds to the interquartile range, and the curve represents kernel estimate of the density function

Fig. 4

Distribution of parasite clearance half-life by pfk13 mutant codons in isolates from Asia. Left panel (a) shows violin plots for mutant codons with five or more isolates. The number of individual isolates tested is at the right of each violin plot. The right panel (b) shows dot plots for mutant codons with < 5 isolates. The red line shows a half-life of 5.5 h. The median is shown as a green circle, the red bar corresponds to the interquartile range, and the curve represents kernel estimate of the density function. * 22 P527H and 1 P527L with PC1/2 = 5.8h; ** 6 P667T and 1 P667R without valid PC1/2 measurement, so not considered