Changing Pattern of Plasmodium falciparum pfmdr1 Gene Polymorphisms in Southern Rwanda

ABSTRACT

Plasmodium falciparummultidrug resistance-1 gene (pfmdr1) polymorphisms associate with altered antimalarial susceptibility. Between 2010 and 2018/2019, we observed that the prevalence of the wild-type allele N86 and the wild-type combination NYD increased 10-fold (4% versus 40%) and more than 2-fold (18% versus 44%), respectively. Haplotypes other than NYD or NFD declined by up to >90%. Our molecular data suggest the pfmdr1 pattern shifted toward one associated with artemether-lumefantrine resistance.

INTRODUCTION

Treatment of Plasmodium falciparum malaria relies on artemisinin-based combination therapies (ACTs), comprising a fast-acting artemisinin derivative and a slowly eliminated partner drug. P. falciparum kelch-13 (pfkelch13) single-nucleotide polymorphisms (SNPs) associate with decreased artemisinin susceptibility. When conferring reduced in vitro sensitivity and delayed parasite clearance in vivo, they are termed validated mutations, common in Southeast Asia (1). Recently, these were detected in East Africa (Rwanda) and associated with delayed parasite clearance (2–4). Although ACT failure remains rare in Sub-Saharan Africa (1), the emergence of non-artemisinin partner drug resistance is feared. Susceptibility to these antimalarials, including lumefantrine (LF) and amodiaquine (AQ), is influenced by the Plasmodium falciparum multidrug resistance-1 gene (pfmdr1) SNPs N86Y, Y184F, and D1246Y (5–9). Individual allele combinations, or haplotypes (e.g., N86-Y184-D1246, NYD, wild-type haplotype), exhibit specific susceptibility phenotypes (9). Notably, pfmdr1 86Y associates with increased sensitivity to LF, mefloquine, and dihydro-artemisinin and decreased chloroquine and AQ sensitivity (5). Resistant strains spread under drug pressure but may decline without (10, 11). Rwanda has used AL as a first-line antimalarial since 2006 (12). In 2010, we reported a predominant pfmdr1 pattern (NFD) suggestive of intense AL pressure in mostly asymptomatic preschool children in Huye, Rwanda (13). Almost a decade later, we reassessed pfmdr1 alleles in symptomatic and largely adult patients in Huye and compared them to the 2010 findings.

In March–June 2018 and September–December 2019, we recruited 295 uncomplicated malaria patients at Sovu Health Centre and Kabutare District Hospital, Huye district, Rwanda. All reported fever in the preceding 48 h or were febrile (164/276; ≥37.5°C, axillary). The study was approved by the Rwanda National Ethics Committee, and participants or caregivers provided informed written consent. Patients were clinically examined, malaria was microscopically confirmed, and venous blood was collected into EDTA. Following DNA extraction (QIAamp DNA blood minikit; Qiagen, Germany), Plasmodium species was confirmed by PCR (14) in 2018 and by real-time PCR (TIB MolBiol, Germany) in 2019. Two pfmdr1 regions (codons 61 to 236 and 1023 to 1288) were PCR amplified (15), sequenced (Eurofins Genomics, Germany), and aligned to reference PF3D7_0523000 (PlasmoDB; https://plasmodb.org/plasmo/app/record/gene/PF3D7_0523000) using CodonCode Aligner 4.2.5. Plasmodium infections can contain genetically distinct parasites and have a multiplicity of infection (MOI) of >1. To include all isolates, also those with evidence of MOI of >1 (i.e., both wild-type and mutant alleles present), we grouped alleles into combinations. In a secondary analysis, we estimated haplotype frequencies using a Bayesian model (16), which integrates unknown MOIs. The same priors were used as described previously (17). SNP and allele combination prevalence as well as haplotype frequencies in 2010 and 2018/2019 were compared using Fisher’s exact test. A P value of <0.05 was considered significant. We used R 3.6.3 for statistical analyses.

By PCR, 90.3% (234/259) of the malaria patients in 2018/2019 had P. falciparum infection. Of these, 50.6% (118/233) were female, median age was 17.5 years (range, 1 to 73), and mean temperature was 37.2°C (standard deviation [SD], ±1.3°C). Good-quality sequencing reads for both pfmdr1 regions were obtained from 90.6% (212/234) of isolates. Evidence of an MOI of >1 was present in 17.9% (38/212) of samples. The observed mutation prevalence was the following: 86Y, 3.8%; 184F, 53.8%; and 1246Y, 2.4% (Table 1). As for observed allele combinations, NFD (50%) dominated over wild-type NYD (44%). Considering haplotype frequency estimates, the reverse was seen (i.e., 39% versus 56%) (Table 2). In any case, >90% of isolates showed NFD or NYD in 2018/2019. Other nonsynonymous polymorphisms were T199S (n = 4), V207I (n = 2), T222I (n = 8), and Q1198K (n = 1), but not S1034C or N1042D.

TABLE 1.

Observed prevalence of pfmdr1 polymorphisms and allele combinations in Huye, Rwanda, in 2010 and in 2018/2019

pfmdr1 allele or allele combinationa	2010 (13), n = 104 [% (n)]	2018 and 2019,b n = 212 [% (n)]
86Y	39.4 (41)	3.8 (8)*
184F	51.9 (54)	53.8 (114)
1246Y	12.5 (13)	2.4 (5)*
N86-Y184-D1246	18.3 (19)	44.3 (94)*
N86-184F-D1246	38.5 (40)	50.0 (106)
86Y-Y184-D1246	23.1 (24)	1.9 (4)*
86Y-184F-D1246	7.7 (8)	1.4 (3)*
86Y-Y184-1246Y	5.8 (6)	0.5 (1)*
N86-184F-1246Y	2.3 (3)	1.9 (4)*
86Y-184F-1246Y	2.3 (3)	0
N86-Y184-1246Y	1.0 (1)	0

^aMutations are presented in boldface.

^b*, significantly different from the respective proportion in 2010.

TABLE 2.

Estimated haplotype frequency in Huye, Rwanda, in 2010 and in 2018/2019^a

pfmdr1 haplotype	2010 (13) [% (95% credibility interval)]	2018 and 2019 [% (95% credibility interval)]
N86-Y184-D1246	24.0 (17.0, 32.3)	56.3 (49.6, 62.9)
N86-184F-D1246	38.0 (29.6, 47.0)	39.2 (32.8, 45.9)
86Y-Y184-D1246	24.3 (17.4, 32.3)	1.2 (0.4, 2.6)
86Y-184F-D1246	3.4 (1.2, 7.2)	0.9 (0.3, 2.2)
86Y-Y184-1246Y	4.6 (1.9, 8.6)	0.4 (0.0, 1.3)
N86-184F-1246Y	2.0 (0.5, 5.1)	1.1 (0.3, 2.4)
86Y-184F-1246Y	0.7 (0.0, 2.9)
N86-Y184-1246Y	1.5 (0.2, 4.5)	0.5 (0.1, 1.6)

^aHaplotype frequencies are estimated by a Bayesian model accounting for multiplicity of infection (17). Mutations are presented in boldface.

Compared to 2010 data from the same region, the 2018/2019 pfmdr1 allele pattern has changed: the prevalence of the 86Y mutation declined 10-fold and that of 1246Y 5-fold, whereas 184F remained basically unchanged (Table 1). Consequently, both the observed prevalence and the estimated frequency of wild-type haplotype NYD more than doubled between 2010 and 2018/2019. Allele combinations or haplotypes other than NYD or NFD present in 2010 declined by up to >90% or disappeared, resulting in reduced genetic diversity (Tables 1 and 2).

The trends in our study accord with observations across Africa, i.e., a shift toward pfmdr1 N86 and D1246, where AL is the major antimalarial (11). The N86 wild-type allele confers decreased LF susceptibility and increased AL failure (6, 8). So far, AL treatment failure is rare in Rwanda (1), possibly due to partial immunity and clinical artemisinin effectiveness. However, susceptibility to dihydro-artemisinin is linked to pfmdr1 86Y (5, 8), which almost vanished from the local parasite population. Moreover, a validated marker of artemisinin resistance, pfkelch13 R561H, occurs in 4.5% of P. falciparum isolates in the same population (2). This molecular constellation, the emergence of an artemisinin resistance allele together with >95% of pfmdr1 N86, indicates a shift toward AL-resistant genotypes in this region.

Since almost 20% of samples had evidence of an MOI of >1, we modeled haplotype frequencies, which differed from observed allele combination prevalence. This illustrates that considering one mutated allele in samples with an MOI of >1 as mutated genotypes should not be mistaken as haplotype frequency. Using MOI in analyzing temporal and/or regional allele patterns is recommended to increase comparability (16).

As limitations, we assessed pfmdr1 alleles at two time points only, in a confined region, and lack susceptibility data. Moreover, we did not type pfmdr1 copy number or the P. falciparum chloroquine resistance transporter gene, which also interfere with artemether and LF sensitivity (6, 7). We compared randomly selected, mostly asymptomatic children (13) to symptomatic, largely adult patients. Manifestation associates with pfmdr1 SNPs (18), but not to the extent observed in our study. A strength is the comparison of molecular markers in the same district almost a decade apart.

Fifteen years after the implementation of AL as a first-line antimalarial, our study suggests the pattern in pfmdr1 SNPs shifts toward AL resistance-associated genotypes in the Huye region. The recently demonstrated independent emergence of artemisinin-resistant P. falciparum strains at two sites in Rwanda underlines the importance of focal surveillance (3, 4). These developments could be the first sign of an imminent health threat to the African continent, and, in the absence of novel antimalarials, triple ACTs might be considered (19).

Data availability.

Data will be made available in the WWARN repository, as a .csv file, containing Pfmdr1 genotypes and basic patient characteristics. The doi will be available when the repository is confirmed. Code for the implementation of the haplotype frequency estimation model is available at https://github.com/welmoedvl/ARTHUR_pfmdr1_haplotypefreqest.