Background
Regardless of the emergence and spread of resistance to artemisinin-based combination therapy (ACT) in Southeast Asia (SEA) [
1‐
5], ACT is still widely used as the first-line treatment for uncomplicated malaria worldwide [
5] and has maintained high efficacy in sub-Saharan Africa (SSA) with fast clearance rates [
5,
6]. SEA is historically considered to be major foci of anti-malarial drug resistance, where resistance to drugs such as chloroquine and sulfadoxine–pyrimethamine (SP) may have originated before eventually spreading to Africa [
7,
8]. There is now heightened concern that resistance to ACT in SEA might follow the same pattern in expansion globally as previously did for chloroquine and SP [
9]. With this in mind, routine monitoring of the therapeutic efficacy of ACT is critical in detecting early changes in
Plasmodium falciparum sensitivity to anti-malarial drugs, and deemed necessary for timely enactment of changes to treatment policy [
10]. Indeed, there is concerted effort to scale-up monitoring therapeutic efficacy of ACT in SEA [
1,
2,
10‐
12], SSA [
13], and the rest of the world [
4].
Artemether–lumefantrine (AL) is the most commonly used ACT for the treatment of uncomplicated
P. falciparum malaria worldwide [
14]. AL was introduced as the first-line treatment for uncomplicated malaria in Kenya in 2006 due to widespread resistance to chloroquine and SP, in 1998 and 2006 respectively [
15‐
17]. ACT has maintained adequate clinical and parasitological response (ACPR) in Kenya, with a recent study reporting more than 97% ACPR for AL and dihydroartemisinin–piperaquine (DP) in treatment of uncomplicated falciparum malaria in western Kenya [
18]. However, studies have shown AL selects for single nucleotide polymorphisms (SNPs) in the
P. falciparum chloroquine resistance transporter gene (
pfcrt) and the
P. falciparum multidrug resistance gene 1 (
pfmdr1) in recurring parasites [
19‐
24]. The genotypes associated with recurrent infections are K76 in
pfcrt, and N86, 184F and D1246 (NFD) in
pfmdr1. Reduced susceptibility to lumefantrine and mefloquine has also been linked to NFD and increase in
pfmdr1 copy numbers [
25‐
27]. Chloroquine resistance is associated with
pfcrt 76T [
28], and modulated by
pfmdr1 86Y, Y184 and 1246Y (YYY) [
29]. The
pfcrt 76 and
pfmdr1 86 alleles are the most important indicators of chloroquine susceptibility [
30]. Longitudinal studies have shown the prevalence of
pfcrt 76T and
pfmdr1 86Y reached over 90% in western and coastal regions of Kenya before the introduction of ACT, and reversed to the sensitive genotypes with the withdrawal of chloroquine and the introduction of AL [
31‐
34]. This reversal to sensitive genotype in Kenya can be attributed to the release of chloroquine drug pressure and the introduction of lumefantrine drug pressure. Recent studies have suggested changes in the prevalence of these alleles can be a sensitive indicator of selection of parasite populations by AL which can be used to signal early reduced susceptibility [
30,
35].
Kenya has a wide variation in malaria prevalence, with some regions free of malaria to those with more than 40% endemicity [
36,
37]. Most regions in western Kenya are endemic lowland with high stable transmission whereas the highlands are characterized by unstable and high transmission variability which results in epidemics during periods of suitable climatic conditions [
36]. To effectively monitor the emergence and spread of resistance to ACT, it is important not only to monitor the prevalence of these alleles (
pfcrt K76,
pfmdr1 N86, 184F and D1246), but also to monitor their origin and spread. Information on the evolutionary dynamics resulting in selection of these alleles in different parts of the country with different transmission intensities and different drug resistant profiles is important in guiding strategies to control, and prevent the emergence and spread of resistance to AL.
Microsatellites are important genetic markers used to identify regions in the genome that are under selection [
38]. Genetic hitchhiking is driven by the selection process which results in reduction of heterozygosity at both the selected locus and neutral flanking microsatellite loci [
39]. When the mutation eventually gets fixed in a population due to continuous selection, sequence diversity is reduced around the selected locus leading to selective sweeps [
38].
Characterization of
P. falciparum parasite genetic backgrounds using microsatellite loci flanking genes associated with resistance to chloroquine and SP was critical in defining the geographic origins and dissemination of chloroquine and SP resistant parasites. It has been reported that these resistant parasites originated in a few places before eventually spreading to the rest of the world [
38,
40‐
43]. More recently, characterization of microsatellite loci flanking
pfmdr1 gene were used to provide comprehensive data on the distribution of alleles in this gene and the pattern of selective sweeps in four sites in Cambodia. These sites had different levels of transmission and drug resistance profiles [
38]. The study established that
pfmdr1 184F mutant allele was under selection in this parasite population whereas copy number variation of
pfmdr1 gene occurred on multiple genetic backgrounds. Given the importance of polymorphisms in
pfmdr1 gene in response to ACT, it is warranted to investigate the selective sweep and genetic lineages of
pfmdr1 alleles in SSA. This study set out to investigate evidence of selective sweep and genetic lineages in
pfmdr1 genotypes associated with AL treatment in Kenya. Microsatellite loci flanking
pfmdr1 gene in parasite population from different regions of Kenya with different malaria transmission intensities were characterized.
Discussion
The data in this study shows differential site and region specific prevalence of SNPs associated with drug resistance in the
pfmdr1 gene. The overall prevalence of
pfmdr1 N86, 184F and D1246 were 86.4%, 47.5% and 93.9%, respectively. However, when analysed based on the region comparing western Kenya (Kisumu, Kericho and Kisii) vs. coastal Kenya (Malindi), the prevalence of
pfmdr1 N86, 184F and D1246 was 92.9% vs. 66.7%; 53.5% vs. 24.2%; 96% vs. 87.9%, respectively. The
pfmdr1 N86, 184F and D1246 genotypes are associated with AL selection [
19‐
22]. Ingasia et al. [
53] recently showed that parasites from western Kenya have high parasite genetic diversity compared to those from the coastal region of Kenya. This coincides with the reports of reduction of malaria infections and transmission in the coastal region [
54‐
57]. The high prevalence of N86, 184F and D1246 genotypes in western Kenya compared to coastal Kenya is consistent with AL selection.
Haplotype analysis have shown lumefantrine susceptibility decreases in the order of N
FD, NYD,
YY
Y and
YYD [
27], with parasites gradually acquiring tolerance, starting with N86, followed by the combination of N86 + D1246 and thereafter, the combination of N86 + 184F + D1246. This observation has been corroborated by field studies [
22,
24,
58,
59]. Similarly, in this study, N
FD haplotype was the most prevalent haplotype followed by NYD and then
YYD, depicting the role of lumefantrine drug pressure in the Kenyan parasite population. When analysed per region, western Kenya had N
FD and NYD prevalence of 51% and 39%, respectively, compared to the coastal region which was 25% and 37.5%, respectively. A previous study that analysed samples collected in 2012/13 from coastal region showed N
FD and NYD at a prevalence of 31.9% and 66%, respectively [
34]. This data suggests that there might be less AL selection pressure in parasites in coastal region of Kenya compared to western region of Kenya.
Soft sweeps are selective events in which there have been multiple origins of the beneficial alleles [
60‐
62]. Soft sweeps have variation in markers flanking selected alleles with multiple origins when mutations are high and populations are large [
47,
60‐
62]. The significantly low mean
He surrounding
pfmdr1 compared to the mean
He at the neutral loci imply that the gene has undergone selection in Kenya. The reduction of mean
He around mutant alleles compared to the respective wild type alleles is an indication of positive directional selection. Analysis of data from the four field sites indicated there was no statistical difference in mean
He between NYD and N
FD haplotypes. However, there was a statistical significant reduction of mean
He surrounding
YYD compared to NYD. When compared per study site, each site indicated unique selection pressure. In Kisumu, there was no difference in mean
He between NYD and N
FD, whereas in Kericho there was a statistical significant reduction of mean
He surrounding NYD compared to N
FD. In Kisii, the selection pressure was the opposite of what was seen in Kericho and in Malindi, reduction was only present in the mean
He surrounding
YYD compared to either NYD or N
FD.
Multiple independent lineages of
pfmdr1 allele have been previously described for parasites in Ghana [
52] and Cambodia [
38]. Similarly, the current study demonstrated the presence of independent genetic lineages for all the
pfmdr1 alleles. Interesting however, for the study that described parasites in Ghana [
52], the authors observed an increase in linkage disequilibrium among loci around
YFD haplotype, which suggested one major and a few minor lineages of this haplotype. Only one sample with the
YFD genotype was observed in current study.
F
st statistics analyses among the linked loci showed geographic distance between the field sites, and appear to play a role in selection. This was evident when western Kenyan parasites were compared to coastal Kenyan parasites, which are geographically separated by more than 800 km. The significant difference in F
st between samples from the two separate locales agree with previous studies which have showed greater genetic distance between physically isolated populations [
56]. There was however evidence of marginal genetic sharing among these populations that could be partly due to dispersal of parasites across these regions [
63]. Parasites from Kericho and Malindi were exception because they were identical despite of the distance between the two sites. Loci − 9.3 and 9.1 were comparable across all populations depicting minimal involvement in selection.
Drug pressure has been implicated as a key driver of selection [
64]. The samples clustered into seven haplotypes of the
pfmdr1 N86Y, Y184F and D1246Y showed F
st values greater than zero suggesting increasing divergence among most haplotypes. Chloroquine use before the year 2000 was shown to be the greatest force behind selection in these loci. Since chloroquine withdrawal more than 20 years ago, there has been return of wild-type at
pfmdr1 86 and 1246, but emerging
pfmdr1 184F [
31], which is associated with lumefantrine selection [
65]. These findings show divergence which appear to suggest different lumefantrine pressure in the different field sites, or presence of other factors that influence selection differently.
Since switching of the first-line, anti-malarial against uncomplicated malaria in Kenya from chloroquine to SP, and then to AL in 2006 [
15,
17], studies have shown trends of recovery of chloroquine sensitive parasites [
31,
33,
34]. These trends have been shown in other African countries as well [
35,
66‐
72]. However, the rates with which the changes occur are different from one region to another, or one country to another. This is the first study which directly compares the prevalence of
pfmdr1 alleles and genetic lineages in samples from the western Kenya to those in the coastal Kenya. Notably, the populations are structured, with those from coastal region showing significant variation in loci surrounding the allele under selection compared to those from the western Kenya loci. Study by O’Meara et al. underscored declining malaria incidence in this region [
73] which is attributed to intensified intervention [
16]. On the contrary, there are reports of sustained malaria transmission in western Kenya [
16,
74] despite similar country-wide transitions of interventions including in drug treatment policy [
75,
76]. Findings in this study which show significant variations between these populations provide evidence for differential selection pressure between the different malaria transmission regions of Kenya, especially the western region of Kenya compared to the coastal region. Indeed, a recent study showed western Kenya parasites have high genetic diversity compared to those in coastal Kenya [
53]. The difference in selection pressure can be attributed to disease prevalence, genetic diversity of the parasite population, anti-malarial drug usage and cultural behaviour of the different patient populations alongside environmental factors that modulate vector density [
77]. Adherence to anti-malarial drug treatment is a challenge as evident by a recent case report of attenuated-responsiveness to AL treatment in western Kenya [
78].
Authors’ contributions
EK, BA and HA designed the study protocol, PM and DJ performed laboratory assays, LI, LC, BO, BN, JC did data analysis and interpretation. PM, DJ and EK drafted the manuscript. PM, DJ, BA, HA, VS, JN and EK reviewed, re-wrote the final manuscript draft. All authors read and approved the final manuscript.