In Vivo Selection of Plasmodium falciparum Parasites Carrying the Chloroquine-Susceptible pfcrt K76 Allele after Treatment with Artemether-Lumefantrine in Africa

Abstract

Background

Artemether-lumefantrine (AL) is a major and highly effective artemisinin-based combination therapy that is becoming increasingly important as a new first-line therapy against Plasmodium falciparum malaria. However, recrudescences occurring after AL treatment have been reported. Identification of drug-specific parasite determinants that contribute to treatment failures will provide important tools for the detection and surveillance of AL resistance.

Method

The findings from a 42-day follow-up efficacy trial in Tanzania that compared AL with sulfadoxine-pyrimethamine (SP) were analyzed to identify candidate markers for lumefantrine tolerance/resistance in the chloroquine resistance transporter gene (pfcrt) and multidrug resistance gene 1 (pfmdr1). The findings were corroborated in vitro with genetically modified isogenic P. falciparum parasite lines.

Results

Treatment with AL selected for the chloroquine-susceptible pfcrt K76 allele (P < .0001) and, to a lesser extent, the pfmdr1 N86 (P = .048) allele among recurrent infections. These genotypes were not selected during SP treatment. No pfmdr1 gene amplifications were observed. Isogenic pfcrt-modified parasite lines demonstrated a 2-fold increase in susceptibility to lumefantrine, which was directly attributable to the K76T mutation.

Conclusions

Our findings suggest that the pfcrt K76T mutation is a drug-specific contributor to enhanced P. falciparum susceptibility to lumefantrine in vivo and in vitro, and they highlight the benefit of using AL in areas affected by chloroquine-resistant P. falciparum malaria.

The development and spread of Plasmodium falciparum antimalarial drug resistance has spurred a global change in policy from the use of the former first-line antimalarials chloroquine (CQ) and sulfadoxine-pyrimethamine (SP) to the use of artemisinin-based combination therapies (ACTs). In 2006, Tanzania changed its drug policy to the use of artemether-lumefantrine (AL; Coartem [Novartis]), which is currently the most favored ACT used in Africa.

AL is highly effective on the African continent [1–3]. Nevertheless, AL has been associated with a relatively high frequency of reinfection events during the post-treatment period, when lumefantrine, the longer-lasting partner drug (with a plasma half-life of 4–5 days vs. 3–11 h for artemether), remains in the blood in subtherapeutic concentrations. During this period, selection of less-susceptible parasites (which have the ability to cause infection in the presence of plasma drug concentrations higher than those associated with susceptible parasites causing infection) may occur.

One gene that has been found to influence susceptibility to lumefantrine is the P. falciparum multidrug resistance gene 1 (pfmdr1), which encodes a transporter located on the digestive vacuole (DV) membrane of the erythrocytic stages. Allelic exchange experiments with culture-adapted lines, as well as field studies of clinical isolates, have shown that pfmdr1 single-nucleotide polymorphisms (SNPs) influence the in vitro response to various antimalarial drugs, such as the artemisinin derivatives and the arylaminoalcohols lumefantrine, mefloquine (MQ), and halofantrine (HF) [4–7]. An increased pfmdr1 copy number has also been associated with treatment failure after short 4-dose AL therapy [8], as well as after MQ or artesunate-MQ therapy [9]. So far, these associations have been observed only in Southeast Asia. However, an increased number of pfmdr1 copies has been found in a few patient samples in Gabon, subsequent to a trial with low-dose MQ treatment, thereby suggesting the potential of selection of multicopy pfmdr1 in this region [10]. Hence, the pfmdr1 copy number may also become important on the African continent after wide-scale deployment of AL.

Our previous studies in Zanzibar showed that treatment with AL selected for reinfecting parasites harboring pfmdr1 alleles encoding the N86, 184F, and D1246 residues [11, 12]. These findings are consistent with reports from other East African field sites [13, 14] and suggest the selection of lumefantrine-tolerant parasites, which may constitute a first step toward full resistance and associated clinical failure [15]. Nonetheless, it remains unclear whether this selection is the result of specific influences on the drug target or is caused by nonspecific selection of parasites that are generally more fit and able to thrive in the presence of environmental stressors. This is an important distinction, considering that mutations at codons 1034, 1042, and 1246 in pfmdr1 have been reported to incur fitness costs in vitro [16].

Another question arising from our previous studies is the role of the chloroquine resistance transporter gene (pfcrt), which encodes a transmembrane protein that localizes to the membrane of the DV and is a key determinant of the chloroquine response of P. falciparum [17]. The pfcrt K76T mutation has been consistently associated with CQ resistance in vitro and in vivo [18 –20]. This putative transporter also appears to influence the parasite response to arylaminoalcohols. Accordingly, mutations at the pfcrt amino acid 76 position have been suggested to play a role in MQ susceptibility [19 –21]. In addition, in vitro selection of HF resistance has been associated with a novel pfcrt S163R mutation, which also has been shown to affect the in vitro response to MQ [22]. In light of these data, we hypothesized that pfcrt SNPs might also influence susceptibility to lumefantrine.

To evaluate the influence of pfmdr1 and pfcrt mutations on the mechanism of resistance to lumefantrine, we searched for selection of these mutations in samples collected during a clinical trial in Fukayosi, Tanzania, in which children with uncomplicated P. falciparum malaria were treated with AL or SP [2]. SP inhibits the synthesis of folate by targeting the dihydropteroate synthetase and dihydrofolate reductase enzymes, respectively. This treatment arm therefore enabled us to test whether any evidence of selection for pfcrt or pfmdr1 mutations in the AL arm was drug dependent and was not a result of general fitness issues associated with the parasite response to any antimalarial drug treatment. Furthermore, the availability of isogenic parasites resulting from allelic exchange experiments at pfcrt amino acid position 76 [19, 20] allowed us to evaluate the specific influence of mutant pfcrt alleles (in particular, the K76T mutation) on the response of P. falciparum to lumefantrine.

SUBJECTS, MATERIALS, AND METHODS

Subjects

The study was conducted in Fukayosi, Bagamoyo District, Tanzania, during April–July 2004, as reported elsewhere [2]. In brief, 106 children with uncomplicated P. falciparum malaria were enrolled and randomly allocated to receive either SP (Fansidar; Roche) or AL (Coartem). Both drugs were administered according to body weight. Children who were allocated to receive SP and who had body weights categorized as 5–10 kg, 11–20 kg, and 21–30 kg were given a single dose of sulfadoxine/pyrimethamine of 250 mg/12.5 mg, 500 mg/25 mg, and 750 mg/37.5 mg, respectively, whereas children who were allocated to receive AL and who had body weights categorized as 5–14 kg, 15–24 kg, and 25–34 kg were given a twice-daily dose of artemether/lumefantrine of 20 mg/120 mg, 40 mg/240 mg, and 60 mg/360 mg, respectively, for 3 days. After enrollment in the study and initiation of treatment, the children were checked routinely on days 1, 2, 3, 7, 14, 21, 28, 35, and 42 or on any day of recurrent illness occurring during the 42-day follow-up. Blood samples were collected on filter paper (3MM; Whatman) for molecular analysis. Written informed consent was obtained from the parents or guardians of the children. The study was approved by the ethics committees of the Muhimbili University College of Health and Allied Sciences (Dar Es Salaam, Tanzania) and Karolinska Institutet (Stockholm, Sweden).

Molecular analysis

DNA was extracted from blood samples on filter paper by use of the ABI Prism 6100 Nucleic Acid PrepStation (Applied Biosystems). The status of recurrent infections (recrudescence or reinfection) was available from a previous study [2] that used stepwise genotyping of the polymorphic genetic markers P. falciparum merozoite surface protein (pfmsp)–2 and –1 to classify treatment outcomes. In the present study, we analyzed the first baseline day (D₀) and the initial day of recurrent infection (R₀).

Mutations in pfmdr1 and pfcrt were investigated to explore their association with AL tolerance. All infections observed at baseline were examined, as were all infections that recurred after AL treatment. The pfmdr1 N86Y and D1246Y SNPs, as well as the pfcrt K76T and S163R SNPs, were analyzed by polymerase chain reaction (PCR) restriction fragment–length polymorphism (RFLP) analysis, as described elsewhere [12, 23]. The pfmdr1 Y184F SNP was assayed using pyrosequencing (Pyro-Gold Reagents PSQ 96MA and PyroMark MD System; Biotage AB) with the primer 5'-CCAGTTCCTTTTTAGGTT-3'. The pfmdr1 S1034C and N1042D SNPs were examined through direct sequencing of PCR products (Macrogen). For infections recurring after SP treatment, only the pfmdr1 N86Y and pfcrtK76T mutations were assessed. The pfmdr1 gene copy number was assessed through the use of TaqMan probe quantitative PCR–based protocols [12]. All primers were synthesized by Thermo Electron. RFLP analysis products were separated by electrophoresis in 2% agarose gels with ethidium bromide and were visualized under UV transillumination (GelDoc System; Bio-Rad).

Statistical analysis

Data from patient samples were analyzed using GraphPad QuickCalcs software (GraphPad Software). Proportions were compared using Fisher’s exact test for the association of treatment type with SNPs. The complete set of baseline data (D₀ for the AL arm + D₀ for the SP arm) was used as a common comparator group for both arms of the drug trial. Statistical significance was denoted by P ≤ .05. When mixed infections were evaluated, all calculations were performed by comparing carriers of the selected allele (including pure allele carriers and mixed infections) with parasites carrying the alternative allele in pure form.

P. falciparum propagation in vitro

All parasite lines were grown at 37°C in human red blood cells at a hematocrit of 3%–4%, by use of RPMI 1640 medium supplemented with 0.5% Albumax, 0.25% sodium bicarbonate, and 0.01 mg/mL gentamicin. Cultures were maintained in an atmosphere of 90% N₂, 5% O₂, and 5% CO₂.

In vitro drug assays

Parasite susceptibilities to lumefantrine were measured in vitro using 72-h ³H-hypoxanthine assays with serial drug dilutions, as described elsewhere [24]. IC₅₀ and IC₉₀ values were determined by linear extrapolation. For statistical analyses, 2-tailed unpaired Student’s t tests, as well as nonparametric Mann-Whitney U tests, were performed.

RESULTS

In total, 38 of the 50 children who were treated with AL experienced recurrent infections during the 42-day follow-up. Of these 38 children, 37 had sufficient parasite material to be included in the analysis. Of the 56 children included in the SP treatment arm, 39 had recurrent infections [2].

In the AL arm, only 2 recrudescences were recorded, which precluded a meaningful analysis of recrudescence versus reinfection. Hence, all breakthrough infections in the AL arm were simply considered to be recurrent and were treated equally. As for the SP arm, 16 recrudescences were identified. No selection for either pfcrt or pfmdr1 variants was observed when SP recrudescences were compared with reinfections. These infections were also grouped together as recurrent infections.

The frequency of parasites carrying the pfcrt K76 allele (in pure form or mixed with 76T) increased from a baseline prevalence of 48.0% to 86.5% (P < .0001) among all recurrent infections in persons in the AL arm (table 1 and figure 1) and to 85.7% (P = .0001) among only reinfections occurring in persons in the AL arm. A statistically significant selection was also observed for pfmdr1 N86, which displayed an increase in frequency from 55.9% at baseline to 75.7% among all recurrent infections (P = .048) (table 1 and figure 1), whereas the frequency increased to 74.3% among only reinfections occurring in the AL arm (P = .071). The frequencies of the pfmdr1 184F and D1246 alleles remained unaltered (table 1 and figure 1). The SP arm revealed no changes in the frequency of pfcrt K76T or pfmdr1 N86Y in recurrent infections (table 1). The pfmdr1 1034 and 1042 SNPs, as well as the pfcrt 163 SNPs, showed no variation in either treatment.

Table 1.

Frequencies of the analyzed single-nucleotide polymorphisms (SNPs), both before the administration of the drug (D₀) and in recurrent infections (R₀).

	Frequency, (pure + mix)/all [95% CI]a
		R0
SNP	D0	AL	SP
pfcrt K76	(26 + 23)/102 = 0.480 [0.386–0.576]	(22 + 10)/37 = 0.865 [0.716–0.946]	(12 + 7)/34 = 0.559 [0.394–0.711]
pfmdr1 N86	(27 + 30)/102 = 0.559 [0.462–0.651]	(21 + 7)/37 = 0.757 [0.597–0.868]	(8 + 6)/33 = 0.424 [0.272–0.592]
pfmdr1 184F	(0 + 25)/100 = 0.250 [0.175–0.344]	(3 + 10)/34 = 0.382 [0.239–0.550]	NA
pfmdr1 D1246	(50 + 24)/98 = 0.755 [0.661–0.830]	(21 + 9)/35 = 0.857 [0.702–0.942]	NA

NOTE. The D₀ data for both treatment arms were pooled at baseline, because no significant differences were found between the 2 groups for any of the analyzed SNPs. AL, artemether-lumefantrine; CI, confidence interval; NA, not analyzed; SP, sulfadoxine-pyrimethamine.

^a“All” denotes the total no. of successful analyses. The 95% CI was determined using the modified Wald method (GraphPad Software; GraphPad).

Figure 1.

Set of comparisons (conducted using Fisher’s exact test [see Subjects, Materials, and Methods]) between infections at baseline (D₀) and all recurrent infections (R₀) concerning the chloroquine resistance transporter gene (pfcrt) K76 allele and the multidrug resistance gene 1 (pfmdr1) N86, 184F, and D1246 alleles. The pfcrt S163 residue, as well as the pfmdr1 S1034 and N1042 residues, had a prevalence of 100% in this setting, as did single-copy pfmdr1. AL, artemether-lumefantrine; SP, sulfadoxine-pyrimethamine.

Our sequencing procedures also revealed a novel nonsynonymous pfmdr1 W1031R SNP (TGG→CGG) located in a predicted transmembrane domain [25]. The new SNP, which was observed in 2 separate PCR amplification and sequencing studies, was present in 1 infection at baseline but was not detected among the recurrent infections in the group given AL. Finally, an increased number of pfmdr1 gene copies was not detected in any of the 89 infections noted at baseline or in the 29 recurrent infections noted in the AL treatment arm that were successfully analyzed in this Tanzanian cohort.

To directly test the contribution of mutant pfcrt to susceptibility to lumefantrine under controlled in vitro conditions, we assayed P. falciparum lines that had been genetically engineered to express distinct pfcrt alleles, rather than the endogenous pfcrt allele in isogenic background. One set was produced in the CQ-susceptible GC03 line—a progeny of the HB3 × Dd2 genetic cross [26]. This set enabled us to compare clones expressing the wild-type CQ-susceptible allele (C2^GC03), the CQ-resistant Dd2 allele typical of Asian and African CQ-resistant strains (C4^Dd2), and the CQ-resistant 7G8 allele observed in South America, the Oceanic region, and India (C6^7G8) [19]. The Dd2 and 7G8 alleles contain 8- and 5-point mutations (M74I/N75E/K76T/A220S/Q271E/N326S/I356T/R371I and C72S/K76T/A220S/N326D/I356L), respectively, compared with the wild-type allele. To test the observed clinical association between the K76T mutation and susceptibility to lumefantrine, we also assayed clones generated in the Dd2 and 7G8 lines, which expressed either the parental allele (C-1^Dd2 and C-1^7G8) or an allele in which the mutant 76T codon had been replaced with the wild-type K76 codon (known as the T76K-1^Dd2 and T76K-1^7G8 “back mutant” clones, respectively). Recombinant clones were assayed in vitro on 5–8 separate occasions, and parental controls (GC03, Dd2, and 7G8) were assayed on 7–12 separate occasions.

These in vitro assays revealed that CQ-resistant lines with mutant pfcrt alleles carrying the K76T mutation were uniformly more susceptible to lumefantrine (table 2). C4^Dd2 and C6^7G8, expressing the 2 most prevalent alleles of mutant pfcrt worldwide, both showed a ~35% decrease in lumefantrine IC₅₀ values, compared with C2^GC03 (P < .05 and P < .01, respectively, by the Mann-Whitney U test). The results achieved using the back-mutant clones demonstrated that the K76T mutation itself is a large mediator of this increased susceptibility. Both the T76K-1^Dd2 and T76K-1^7G8 clones were 2-fold less susceptible to lumefantrine than were their recombinant counterparts (for both, P < .01, by the Mann-Whitney U test). In the parental lines, we also observed substantially lower lumefantrine IC₅₀ values in Dd2 and 7G8, compared with GC03.

Table 2.

Lumefantrine susceptibility levels in chloroquine resistance transporter gene (pfcrt)–modified and control lines.

Line	LMF IC50, nmol/La	Recombinant	Parental line	PfCRT haplotype									CQb		ARTM IC50,b nmol/L	MQ IC50,b nmol/L
				72	74	75	76	220	271	326	356	371	IC50, nmol/L	Phenotype
GC03	85.5 ± 7.2	No	…	C	M	N	K	A	Q	N	I	R	29.2 ± 2.0	Susc.	39.4 ± 6.9	52.6 ± 5.4
Dd2	55.2 ± 9.6	No	…	C	I	E	T	S	E	S	T	I	195.1 ± 25.3	Res.	46.3 ± 7.6	46.8 ± 2.0
7G8	27.0 ± 4.1	No	…	S	M	N	T	S	Q	D	L	R	140.4 ± 15.0	Res.	21.6 ± 6.3	9.6 ± 1.3
C2GC03	74.0 ± 7.7	Yes	GC03	C	M	N	K	A	Q	N	I	R	22.9 ± 2.8	Susc.	39.1 ± 7.0	54.7 ± 1.5
C4Dd2	48.7 ± 4.3c	Yes	GC03	C	I	E	T	S	E	S	T	I	147.4 ± 13.3d	Res.	12.6 ± 1.8d	33.1 ± 3.1d
C67G8	46.3 ± 4.6d	Yes	GC03	S	M	N	T	S	Q	D	L	R	126.9 ± 17.2d	Res.	17.2 ± 3.0d	27.3 ± 3.4d
T76K-1Dd2	73.2 ± 10.5d	Yes	Dd2	C	I	E	K	S	E	S	T	I	22.1 ± 2.1d	Susc.	40.4 ± 11.7	27.2 ± 8.0c
C-1Dd2	33.7 ± 5.0	Yes	Dd2	C	I	E	T	S	E	S	T	I	160.7 ± 28.9	Res.	32.9 ± 6.6	36.9 ± 4.2
T76K-17G8	68.3 ± 11.7d	Yes	7G8	C	I	E	K	S	Q	D	L	R	23.9 ± 4.6e	Susc.	13.8 ± 3.2	3.6 ± 0.7
C-17G8	33.6 ± 5.5	Yes	7G8	S	M	N	T	S	Q	D	L	R	108.8 ± 15.9	Res.	7.4 ± 2.1	3.1 ± 0.9

NOTE. ARTM, artemisinin; CQ, chloroquine; LMF, lumefantrine; MQ, mefloquine; Res., resistant; Susc., susceptible.

^aLMF IC₅₀ values are expressed as the mean value (±SE) for pfcrt–modified and control lines. Data were derived from 5–12 independent assays performed in duplicate. (Results from Mann-Whitney U tests compared mutants and recombinant controls).

^bCQ, ARTM, and MQ IC₅₀ values (from at least 3 independent assays performed in duplicate) are expressed as the mean value (±SE) and are reproduced from [19] for the top 6 lines and from [20] for the bottom 4 lines.

^cP < .05.

^dP < .01.

^eP < .001.

DISCUSSION

Coartem, a fixed formulation of artemether and lumefantrine, is currently the most successful and widely implemented ACT and recently has been adopted as first- or second-line chemotherapy for the treatment of uncomplicated malaria by a large number of national malaria control programs in sub-Saharan Africa [27]. This combination, referred to as “AL” in the present study, has consistently demonstrated high efficacy. Nevertheless, the high levels of parasite transmission in most parts of sub-Saharan Africa dictate that reinfections will often occur after cessation of AL treatment [1–3]. This creates a risk for selection of reinfecting lumefantrine-tolerant/-resistant parasites after treatment, when the slowly eliminated lumefantrine remains at subtherapeutic concentrations. Previous data from some of the investigators involved in the present study [11]—data subsequently confirmed by others [13, 14]—have shown a consistent pattern of detectable increases in the frequency of drug resistance–associated pfmdr1 haplotypes among recurrent infections after AL treatment. Because these recurrent parasites represent reinfections in all but a few cases, the data have been interpreted as a lumefantrine-driven selection for variant pfmdr1 alleles.

In the present study, we report significant selection of the pfcrt K76 allele after AL treatment, which, to our knowledge, represents the first report of a significant in vivo association of this gene with an ACT containing an arylaminoalcohol (a group of drugs that includes lumefantrine, MQ, and HF). This selection was not observed in our preceding studies, probably because of the rarity of the K76 allele in Zanzibar [1, 11, 12], as opposed to the present study site on mainland Tanzania, where ~50% of the analyzed infections harbored K76 allele carriers. As discussed elsewhere [12], the high efficacy of AL precludes the collection of statistically meaningful numbers of recrudescent parasites for analysis. Hence, the conclusions of the present study are mainly considered to be applicable to parasites causing reinfection (i.e., formally “lumefantrine-tolerant” parasites, as opposed to “lumefantrine-resistant” parasites [15]).

The selection for pfcrt K76 in the recurrent infections observed in the present study is consistent with earlier in vitro data supporting a role for this gene in P. falciparum susceptibility to arylaminoalcohols. After allele exchange at the pfcrt amino acid 76 position, isogenic clones have been shown to acquire different susceptibilities to MQ and HF. In the first of these experiments, exchanging the K76 for 76T (in the CQ-susceptible GC03 background) increased parasite in vitro susceptibility to MQ by 2- to 3-fold [19]. In a set of reverse experiments, the replacement of 76T with K76 decreased the susceptibility of the parasites to MQ, albeit to a lesser extent (1.5- to 2-fold) [20] (table 2). In another study, P. falciparum lines selected in vitro for HF resistance were shown to have acquired pfcrt mutations that included S163R, which was also associated with MQ resistance [22]. Finally, in vitro development of resistance to CQ in the Sudanese 106/1 clone (which already harbored 7 of the 8 Dd2 pfcrt mutations, lacking only K76T) resulted in novel pfcrt SNPs at position 76 (K76I/N) [17, 21]. These novel K76I/N SNPs were shown to significantly increase susceptibility to MQ and HF [21].

Our present research extends these observations by documenting a 2-fold increase in susceptibility to lumefantrine in parasites expressing the K76T mutation (compare C-1^Dd2 with T76K-1^Dd2 and compare C-1^7G8 with T76K-1^7G8). Interestingly, 7G8 was 2-fold more susceptible than Dd2, a finding that, on the basis of our data, cannot be attributed to differences between these pfcrt alleles. These lines also differ in their pfmdr1 point mutations (with Dd2 having the 86Y/Y184/S1034/N1042/D1246 haplotype and with 7G8 having the N86/184F/1034C/1042D/1246Y haplotype) as well as their copy number (3–4 copies in our Dd2 line, compared with 1 copy in 7G8). Both types of genetic changes in pfmdr1 are known to affect susceptibility to lumefantrine. Other, as yet undefined mediators may also differ between 7G8 and Dd2. Overall, our data clearly define the pfcrt K76T mutation as a major determinant that contributes to decreased susceptibility to lumefantrine on distinct genetic backgrounds.

Our results also show that the 2 globally most prevalent mutant pfcrt alleles, Dd2 and 7G8, both increase susceptibility to lumefantrine (compare C4^Dd2 and C6^7G8 with the isogenic CQ-susceptible C2^GC03 line). In this regard, lumefantrine is an ideal ACT partner drug in areas of high resistance to CQ. Interestingly, mutant pfcrt alleles have also previously been found to significantly increase susceptibility to artemisinin [19]. This finding indicates an important dual benefit of using AL in areas with a high prevalence of mutant pfcrt conferring resistance to CQ.

Mutant PfCRT is thought to efflux compounds out of the DV into the cytoplasm. For CQ, which interferes with the detoxification of heme moieties inside the DV, resistance is thought to be achieved by mutant PfCRT-mediated export of drug away from its hematin target [25]. CQ resistance has been shown to critically depend on the pfcrt K76T mutation [20]. The observation that parasites carrying the pfcrt K76T mutation are more susceptible to lumefantrine and MQ suggests that those compounds might also be transported by mutant PfCRT out of the DV and, consequently, might exert their action in the parasite cytoplasm.

For pfmdr1 and its encoded ABC-transporter Pgh-1, transport studies have shown that the pfmdr1 copy number is associated with Fluo-4 accumulation inside the DV [28]. Moreover, it has been shown that the pfmdr1 gene expressed in Saccharomyces cerevisiae functionally complements an exporter in the plasma membrane [29], with the nucleotide-binding domains on the cytosolic side [30]. In the DV membrane, Pgh-1 is situated with the nucleotide-binding domains in the cytoplasm [31, 32], indicating a transport direction from the cytoplasm into the DV. Thus, Pgh-1 may function as a drug importer. For drugs such as CQ that have their target inside the DV, an increased number of pfmdr1 copies might therefore be expected to result in increased drug susceptibility. In line with this hypothesis, selection of high-level CQ resistance has been observed to result in pfmdr1 deamplification [33]. For drugs that have their target outside the DV, increased numbers of pfmdr1 copies might therefore result in decreased susceptibility. The observed association between an increased pfmdr1 copy number and treatment failures after AL [8], MQ, or artesunate-MQ therapy [9] is therefore another indication that the target of lumefantrine and MQ might lie outside the DV.

The CQ susceptibility markers pfcrt K76 and pfmdr1 N86 [18] and increases in pfmdr1 copy number have all been associated with a decreased susceptibility to both lumefantrine and MQ [7–9, 11, 19–21, 34, 35]. Thus, substantial evidence points to cross-resistance between the 2 arylaminoalcohols. In Southeast Asia, decreased efficacy of artesunate-MQ treatment has been reported [36], presumably because of MQ resistance caused by an increased pfmdr1 copy number [37]. Although an increased pfmdr1 copy number seems to be rare in Africa, it has previously been detected in Gabon in a subset of patients treated with a low dose of MQ. This suggests a risk for selection and spread of multicopy pfmdr1 in Africa after the ongoing wide-scale deployment of AL [10]. This finding, combined with our observation of selection of pfcrt K76 and pfmdr1 N86 alleles, suggests that AL tolerance or resistance could emerge in Africa. This scenario would have major public health implications for AL as a malaria treatment, in terms of its ability to effectively and sustainably reduce disease-associated morbidity and mortality.

In the present study, we compared selection of pfmdr1 and pfcrt mutations that occurred in association with AL treatment with selection that occurred in association with treatment with SP, a drug for which resistance is established to be unrelated to genetic alterations in these genes. Because no selection was observed for the analyzed SNPs in the SP arm, the present study supports the argument that pfmdr1 N86 and, now, pfcrt K76 are specifically associated with a decreased susceptibility to lumefantrine, allowing a more successful reinfection of the human host in the period corresponding to slowly decreasing lumefantrine concentrations [11, 15].

The lack of pfcrt K76 selection in the SP treatment arm is in apparent contrast to the documented replacement of the 76T allele by K76 in regions where SP has been introduced to replace CQ, as has been illustrated by reports from Malawi [38, 39]. In this regard, we speculate that harboring pfcrt K76 presents a relatively small fitness advantage, compared with harboring pfcrt 76T. In the absence of CQ pressure over several years, the progressive replacement becomes detectable in the overall population, ultimately overtaking the parasite population harboring the K76T mutation [38–41].

In conclusion, we identified the pfcrt K76 allele as a further potential marker of decreased susceptibility to lumefantrine in vivo, in addition to the established pfmdr1 SNPs [10–12]. Importantly, pfcrt encoding the K76T mutation, which confers CQ resistance in vitro and is a highly sensitive marker of CQ treatment failure in vivo, also increases in vitro susceptibility to artemisinin [19]. These findings highlight a substantial benefit of using AL in areas with a high prevalence of CQ-resistant P. falciparum malaria. Defining the interrelationships between drug-determinant polymorphisms and changes in susceptibility to the various antimalarials will be important in ongoing efforts to best manage the implementation of different ACTs and reduce the dissemination and influence of drug-resistant malaria.