Effect of Single Nucleotide Polymorphisms in Cytochrome P450 Isoenzyme and N-Acetyltransferase 2 Genes on the Metabolism of Artemisinin-Based Combination Therapies in Malaria Patients from Cambodia and Tanzania

Abstract

The pharmacogenetics of antimalarial agents are poorly known, although the application of pharmacogenetics might be critical in optimizing treatment. This population pharmacokinetic-pharmacogenetic study aimed at assessing the effects of single nucleotide polymorphisms (SNPs) in cytochrome P450 isoenzyme genes (CYP, namely, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP3A4, and CYP3A5) and the N-acetyltransferase 2 gene (NAT2) on the pharmacokinetics of artemisinin-based combination therapies in 150 Tanzanian patients treated with artemether-lumefantrine, 64 Cambodian patients treated with artesunate-mefloquine, and 61 Cambodian patients treated with dihydroartemisinin-piperaquine. The frequency of SNPs varied with the enzyme and the population. Higher frequencies of mutant alleles were found in Cambodians than Tanzanians for CYP2C9*3, CYP2D6*10 (100C→T), CYP3A5*3, NAT2*6, and NAT2*7. In contrast, higher frequencies of mutant alleles were found in Tanzanians for CYP2D6*17 (1023C→T and 2850C→T), CYP3A4*1B, NAT2*5, and NAT2*14. For 8 SNPs, no significant differences in frequencies were observed. In the genetic-based population pharmacokinetic analyses, none of the SNPs improved model fit. This suggests that pharmacogenetic data need not be included in appropriate first-line treatments with the current artemisinin derivatives and quinolines for uncomplicated malaria in specific populations. However, it cannot be ruled out that our results represent isolated findings, and therefore more studies in different populations, ideally with the same artemisinin-based combination therapies, are needed to evaluate the influence of pharmacogenetic factors on the clearance of antimalarials.

INTRODUCTION

Use of artemisinin-based combination treatment (ACT) is the current World Health Organization (WHO) recommendation for uncomplicated malaria (1). While generally very effective and safe, variations in ACT efficacy are found across the areas in the world where malaria is endemic. Parasite, human, and drug factors combined contribute to such variance. Both drug efficacy and safety depend upon appropriate drug levels being achieved and maintained long enough (2–4); insufficient exposure is associated with risk of failure and emergence of resistance (5), while excessive exposure carries the risk of toxicity (1, 2, 6–8). All other factors being equal, human genetics can alter drug exposure. Understanding the role of genetic polymorphisms in genes encoding proteins and enzymes involved in drug absorption, distribution, metabolism, elimination, and action is important for better understanding the interindividual differences in drug pharmacokinetics and pharmacodynamics (9). Depending on the alleles carried by an individual, drug metabolism can be either enhanced, decreased, or abolished (10). The cytochrome P450 enzyme family (CYP genes) and the phase II N-acetyltransferase 2 enzyme (NAT2 gene) are involved in the metabolism of various antimalarial drugs (Table 1). For some antimalarial drugs (i.e., piperaquine [11], pyrimethamine [12, 13], and pyronaridine [14]), the metabolic pathways are still not well understood, while for others essentially no metabolism takes place at all (i.e., atovaquone [15] and doxycycline [16]).

Table 1.

Cytochrome P450 isoenzyme genes and N-acetyltransferase 2 genes involved in metabolism of antimalarial drugs

Enzyme(s)	Antimalarial drugs (reference[s])
CYP2A6 and CYP2B6	Artemisinins (25, 28, 62–67)
CYP2C8	Amodiaquine (25, 28, 68–70), chloroquine (25, 64, 71, 72), and dapsone (64)
CYP2C9 and NAT2	Dapsone (25, 64), sulfamethoxazole (69, 72, 73), and sulfadoxine (74)
CYP2C19	Artemether (75), dapsone (63), and proguanil (23, 25, 62, 76, 77)
CYP2D6	Chloroquine (5, 63, 71) and halofantrine (78)
CYP3A4	Artemisinins (25, 28, 62, 64, 65, 76), chloroquine (25, 63, 70, 71), dapsone (25, 63), halofantrine (25, 76, 78), lumefantrine (28, 48, 76), mefloquine (25, 28, 79–81), primaquine (25), and quinine (25, 75, 80–84)
CYP3A5	β-Arteether (25, 64), artemether (48), chloroquine (25, 70), halofantrine (25), lumefantrine (25), mefloquine (79), and quinine (80)

With varied frequencies of polymorphisms in different ethnic groups, variations in drug metabolism and hence drug efficacy and safety are expected between distinct populations (17). In this study, we analyzed and compared malaria patients who were enrolled in Cambodia and Tanzania. Artemether-lumefantrine (AL) has been the first-line treatment for uncomplicated malaria since the end of 2006 and displays good efficacy in Tanzania (18). In Cambodia, however, its efficacy since 2002 has been reported in all but one study to be considerably and consistently lower (19, 20, 21).

While the influence of nongenetic factors, such as food intake, age, or body weight, on the pharmacokinetics of antimalarial drugs has received some attention, only more recently have the effects of polymorphisms in genes encoding enzymes responsible for antimalarial drug metabolism (such as CYP, NAT2, and UDP-glucuronosyltransferase [UGT] gene), received attention (22–28). Better knowledge of the ethnic variability of genes encoding metabolizing enzymes, and of the impact of this variability on the pharmacokinetic profiles of antimalarial drugs, could help tailor treatments for specific populations.

The objectives of the present study were thus to characterize the allelic frequencies of the main genetic variants of the CYP and NAT2 genes involved in currently used ACTs in Tanzania (artemether-lumefantrine almost exclusively) and Cambodia (artesunate-mefloquine and dihydroartemisinin-piperaquine) and to evaluate whether genetic polymorphisms influence the pharmacokinetic profile of ACTs, in addition to nongenetic factors.

MATERIALS AND METHODS

Study areas, design, and populations.

This study was part of an in vivo drug efficacy study which enrolled mostly unrelated patients with uncomplicated Plasmodium falciparum malaria who were of all ages in northern and western Cambodia (64 patients in 2007 at Phnom Dék Health Centre, Rovieng District, Preah Vihear Province, and 61 patients in 2008 at Pramoy Health Centre, Veal Veng District, Pursat Province) and in central Tanzania (149 patients in 2008 at Kibaoni Health Centre, Kilombero District, Morogoro Region). Details of the study protocol will be reported elsewhere by E. M. Staehli Hodel et al. (submitted for publication). In brief, all suspected malaria cases were screened either by a rapid diagnostic test (RDT) or microscopy. After clinical examination and informed consent, venous blood samples (anticoagulated using EDTA) were taken before and during supervised treatment (days 1, 2, and 7 in all locations, and also in Cambodia 1 h after the first dose and on day 14) with the standard national first-line treatment against uncomplicated P. falciparum malaria. In Tanzania, patients were given six doses of artemether-lumefantrine (Coartem; Novartis Pharma, Switzerland) over 3 days. In Cambodia, patients received three doses of artesunate (Arsumax; Sanofi-Aventis, France) and mefloquine (Eloquine; Medochemie Ltd., Cyprus) over 3 days in Phnom Dék and three doses of dihydroartemisinin-piperaquine (Duo-Cotecxin; Zhejiang Holley Nanhu Pharmaceutical Co., Ltd., China) over 3 days in Pramoy.

In Cambodia, more than 90% of the people are classified in the Khmer ethnic group. The ethnic minorities include Chinese, Chams (Muslim descendants of Cham refugees who fled to Cambodia after the fall of Champa), Khmer Loeu (a collective term for Mon-Khmer- or Austronesian-speaking hill tribes), and Vietnamese. In the Preah Vihear and Pursat Provinces, Khmer Loeu people form the most represented minority, and they tend to live in separate villages. The study in Tanzania was conducted at Kibaoni Health Center in Ifakara town in Kilombero District. The town lies in the flood plain of the Kilombero River Valley at approximately 36.41°E 8.8°S. The total population of the study area was approximately 90,000. The inhabitants of the Kilombero River Valley live in widely scattered households in the rice field plains. Most people are subsistence farmers of various ethnic groups, and recently highly mobile pastoralist ethnic groups, i.e., Masaai, Barbaig, and Sukuma people, have migrated to the area.

As ethnicity is a sensitive issue and was not necessary for the population pharmacokinetic-pharmacogenetic analysis, this information was not collected. In order to simplify the genetic epidemiological analyses, subsequent testing for Hardy-Weinberg equilibrium showed that the ethnic distributions among patients at both Cambodian study sites were similar.

Laboratory procedures.

Blood samples were kept on ice for no longer than 6 h after withdrawal, then aliquoted into whole blood, plasma, and pellet and immediately stored in liquid nitrogen or in a −80°C freezer. Plasma drug concentrations of 14 antimalarial drugs and their metabolites, i.e., artemether (AM), artesunate (AS), dihydroartemether (DHA), amodiaquine, N-desethyl-amodiaquine, lumefantrine (LF), desbutyl-lumefantrine (DLF), piperaquine (PPQ), pyronaridine, mefloquine (MQ), chloroquine, quinine, pyrimethamine, and sulfadoxine, were determined simultaneously by using a liquid chromatography-tandem mass spectrometry (LC-MS/MS) method that required 200 μl of plasma (29).

A sequencing strategy described in detail by Hodel et al. (30) was adopted to assess the following SNPs: CYP2A6*2, CYP2B6*5, CYP2B6*6 (only 516G→T, also called CYP2B6*9), CYP2C8*3 (only 416G→A), CYP2C9*3, CYP2C9*5, CYP2C19*3, CYP2D6*4, CYP2D6*10 (100C→T and 4180G→C), CYP2D6*17 (1023C→T and 2859C→T), CYP3A4*1B, CYP3A5*3, NAT2*5, NAT2*6, NAT2*7, and NAT2*14.

Sample size.

Since prevalence rates of the different polymorphisms varied from 0 to 100%, there was no a priori sample size that could be estimated. The sample size of 150 patients represents one of the largest sample sizes so far reported for population pharmacokinetic and pharmacogenetic studies of antimalarial agents.

Data analysis.

Sequences were analyzed using the Prism AutoAssembler version 1.4.0 (Applied Biosystems) for assembly. The genotype of each patient was then assessed visually. Hardy-Weinberg equilibrium was tested using the χ² Hardy-Weinberg equilibrium test calculator for biallelic markers of the Online Encyclopedia for Genetic Epidemiology studies (http://www.oege.org; last accessed 12 January 2012). Differences of allele frequencies between populations were tested using two-by-two tables and Fisher's exact test. A Bonferroni correction for multiple comparisons was performed for both tests, and a P value of <0.003 was considered significant. The fixation index (F_ST) was calculated using Arlequin version 3.1 (31) in order to measure population differentiation based on the investigated SNPs.

Population pharmacokinetic modeling.

The population pharmacokinetic-pharmacogenetic analysis was performed using NONMEM version 6 (32). This method uses mixed (fixed and random) effects regression to estimate population means and variances of the pharmacokinetic parameters and to identify factors that influence them. The structural and statistical models for AM, LF, MQ, and PPQ were adopted from our models to be described elsewhere, as elaborated for the same study population and elsewhere by Staehli Hodel et al. (submitted). In brief, the population pharmacokinetic model for LF, AM, their metabolites DLF and DHA, and MQ were described using a one-compartment model with first-order absorption, whereas a two-compartment model was used for PPQ. A fixed allometric power function of body weight was used on clearance (CL) and the volume of distribution of the central compartment (V_C) to capture difference in kinetics according to age. Except for body weight, which influenced CL and V_C, only CYP-inhibiting medications taken concomitantly showed a significant effect on CL. The average estimates from the final model, along with interindividual variability (Staehli Hodel et al., submitted) are summarized in Table 2. Owing to the limited number of samples for AS and its metabolite, DHA, no pharmacokinetic analysis could be performed for this drug and its metabolite.

Table 2.

Parameters estimated for the base models describing the population pharmacokinetics of artemether, lumefantrine, mefloquine, and piperaquine and estimates from the bootstrap evaluation in 200 replicates

Drug and pharmacokinetics parametera	Population mean				Bootstrap evaluation
	Estimate	SEb	IIVc	SEd	Mean of 200	SE	95% CI
Artemether
CL (liters/h/kg)	24.7 × BW0.75	10%	44%	17%	24.4 × BW0.75	10%	19–29
θINH	−0.3	19%			−0.27	60%	0.5–1.1
VC (liters/kg)	129	20%			133	26%	88–232
VM (liters)	(Fixed to VC)
ka (h−1)	0.27	11%	119%	11%	0.27	14%	0.21–0.35
k23 (h−1)	5.86	21%	68%	9%	5.83	25%	3.6–9.7
CLmet (liters/h)	419	30%			440	42%	213–927
σC (μmol/liter)	74%	9%d			73%	11%d	59–86%
σM (μmol/liter)	119%	11%d			116%	9%d	88–149%
Lumefantrine
CL (liters/h/kg)	0.84 × BWθBWCL	28%	38%	14%	0.87 × BWθBWCL	24%	0.51–1.37
θBWCL	0.52	19%			0.51	14%	0.36–0.65
VC (liters/kg)	59.9 × BWθBCVC	28%	33%	11%	59.5 × BWθBCVC	24%	35.1–91.9
θBCVC	0.35	28%			0.34	19%	0.1–0.45
VM (liters)	(Fixed to VC)
ka (h−1)	0.54	31%			0.48	43%	0.11–0.88
F0 (mg)	2.53	14%	103%	14%	2.45	15%	1.58–3.28
k23 (h−1)	3.7 × 10−4	12%	38%	15%	3.7 × 10−4	9%	(30–44) × 10−4
CLmet (liters/h)	4.8	10%			4.6	13%	3.4–5.7
σC (μmol/liter)	72%	9%d			6%	3%	55–77%
σM (μmol/liter)	0.013	4%c			0.013	45%	0.010–0.016
Mefloquine
CL (liters/h/kg)	0.10 × BW0.75	5%	12%	88%	0.10 × BW0.75	5%	(0.09–0.11)
VC (liters/kg)	8.93 × BW	6%	19%	96%	9.01 × BW	6%	(8.04–10.20)
ka (h−1)	0.15	14%			0.15	14%	0.12–0.19
F0 (mg)	33.1	56%	175%	48%	31.0	43%	11.8–48.1
σC (μmol/liter)	43%	6%d			43%	6%d	0.14–0.22
Piperaquine
CL (liters/h/kg)	4.50 × BW0.75	13%	45%	61%	4.26 × BW0.75	22%	3.24–5.76
VC (liters/kg)	346 × BW	12% × BW	65%	48%	347 × BW	13%	260–432
Q (liters/h)	122	13%			126	13%	86–158
VP (liters)	18,600	22%	50%	77%	20,053	37%	8,778–28,422
ka (h−1)	0.93	28%			1.00	34%	0.35–1.52
F0 (mg)	123	18%			125	18%	75–171
σC (μmol/liter)	41%	10%d			41%	6%d	0.14–0.21%

^aCL, clearance; BW, body weight; θ_INH, inhibitor effect on CL, determined as (1 − [θ_INH × INH]); V_C, central volume of distribution; Q, intercompartment clearance; V_M, volume of distribution of the metabolite; V_P, peripheral volume of distribution; k_a, first-order absorption rate constant; F₀, residual amount from the previous treatment; k₂₃, metabolism rate constant; CL_met, metabolite clearance; σ_C, exponential residual error for the central compartment; σ_M, exponential residual error for the metabolite compartment.

^bStandard error of the estimate θ_i defined as (SE estimate)/estimate × 100%.

^cIIV, interindividual variability.

^dStandard error of the coefficient of variation defined as √[(SE estimate)/estimate] × 100%.

Genetic polymorphisms in the enzymes responsible for the metabolism of the respective antimalarial agent (Table 3) were included in the covariate analysis if their allele frequency was not equal to 0% or 100% among the study populations (see Tables 4 and 5, below). For PPQ, no data on the metabolic pathways were available, and thus all available SNPs were tested, i.e., CYP2B6*5, CYP2B6*6, CYP2C9*3, CYP2C19*3, CYP2D6*10 (100C→T and 4180G→C), CYP2D6*17 (2859C→T), CYP3A4*1B, CYP3A5*3, NAT2*5, NAT2*6, and NAT2*7. Patients lacking pharmacogenetic information for a specific SNP were excluded from the analysis of the respective SNP.

Table 3.

SNP tested in the population pharmacokinetic model

Drug	Allele
Artemether	CYP2B6*5, CYP2B6*6, CYP2C19*3, CYP3A4*1B, CYP3A5*3
Lumefantrine	CYP3A4*1B, CYP3A5*3
Mefloquine	CYP3A4*1B, CYP3A5*3
Piperaquinea	CYP2C9*3, CYP2C19*3, CYP3A4*1B, CYP3A5*3, CYP2D6*10 (100C→T and 4180G→C), CYP2D6*17 (2850C→T), NAT2*5, NAT2*6, NAT2*7

^aFor piperaquine, no data on the metabolic pathways were available, and thus all available SNP were tested if their allele frequency was not equal to 0% or 100% among the study population.

Table 4.

Allele frequencies in cytochrome P450 isoenzyme genes in Cambodia and Tanzania^a

Gene	SNP	Cambodia				Tanzania				PF
		n	% Allele	χ2	Pχ	n	% Allele	χ2	Pχ
CYP2A6*2	T/T	123				148
	T/A	0	0.0			0	0.0
	A/A	0				0
CYP2B6*5	C/C	116				145
	C/T	2	3.3	66.52	<0.003	1	1.7	93.91	<0.003	0.108
	T/T	3				2
CYP2B6*6	G/G	38				56
	G/T	40	38	0.93	0.335	47	34	2.31	0.129	0.056
	T/T	16				18
CYP2C8*3	G/G	75				69
	G/A	0	0.0			0	0.0
	A/A	0				0
CYP2C9*3	A/A	109				131
	A/C	10	6.6	13.38	<0.003	0	0.0			<0.003
	C/C	3				0
CYP2C9*5	C/C	122				129
	C/G	0	0.0			2	0.8	0.01	0.920	0.268
	G/G	0				0
CYP2C19*3	G/G	117				139
	G/A	7	2.8	0.10	0.752	2	0.7	0.01	0.920	0.049
	A/A	0				0
CYP2D6*4	G/G	74				126
	G/A	0	0.0			5	4.1	37.07	<0.003	0.007
	A/A	0				3
CYP2D6*10 (100C→T)	C/C	19				119
	C/T	33	61	2.30	0.129	16	6.6	0.31	0.578	<0.003
	T/T	38				1
CYP2D6*10 (4180G→C)	G/G	17				13
	G/C	36	71	9.78	<0.003	49	74	1.79	0.181	0.065
	C/C	69				80
CYP2D6*17 (1023C→T)	C/C	66				74
	C/T	0	0.0			37	20	0.02	0.888	<0.003
	T/T	0				5
CYP2D6*17 (2850C→T)	C/C	46				9
	C/T	4	7.7	10.92	<0.003	19	61	1.10	0.294	<0.003
	T/T	2				19
CYP3A4*1B	A/A	87				8
	A/G	7	3.7	0.14	0.708	55	74	0.39	0.532	<0.003
	G/G	0				71
CYP3A5*3	A/A	16				98
	A/G	56	65	0.02	0.888	39	18	1.39	0.238	<0.003
	G/G	52				7

^an indicates the number of patients, % SNP is the mutated allele frequency, χ² is the result from the Hardy-Weinberg equilibrium test, and P_χ is the resulting one-tailed P value. P_F is the P value based on Fisher's exact test for differences between populations.

Table 5.

Allele frequencies in N-acetyltransferase 2 genes in Cambodia and Tanzania^a

Gene	Allele	Cambodia				Tanzania				PF
		n	% SNP	χ2	Pχ	n	% SNP	χ2	Pχ
NAT2*5	T/T	106				58
	T/C	16	6.6	0.60	0.439	66	36.4	0	1.000	<0.003
	C/C	0				19
NAT2*6	G/G	44				84
	G/A	59	40	0.01	0.920	49	27	3.51	0.061	<0.003
	A/A	19				15
NAT2*7	G/G	78				141
	G/A	38	21	0.24	0.624	7	2.4	0.09	0.764	<0.003
	A/A	6				0
NAT2*14	G/G	115				107
	G/A	6	2.5	0.08	0.777	27	15	7.42	0.006	<0.003
	A/A	0				7

^an indicates the number of patients, % SNP is the mutated allele frequency, χ² is the result from the Hardy-Weinberg equilibrium test, and P_χ is the resulting one-tailed P value. P_F is the P value from Fisher's exact test for differences between populations.

In these analyses, each genotype was entered solo into the model. Several models were tested and compared to the richest possible model, which assigned a separate fixed effect to each genotypic group, i.e., homozygous for the reference allele (Hom-REF) or heterozygous (Het-VAR) or homozygous (Hom-VAR) for the mutated allele, as follows: CL = θI₁ + θI₂ + θI₃, where θ is the population pharmacokinetic estimate of CL and I_i is an indicator variable that takes the value of 1 if an individual carries the ith genotypic score (i.e., I₁, Hom-REF; I₂, Het-VAR; I₃, Hom-VAR) and 0 otherwise. The impacts of genotypes were further explored to distinguish the differences between the genotypic groups by using two reduced models, in which the same genotyping group was assigned to Hom-VAR and Het-VAR or Hom-REF and Het-VAR.

The models were fitted using the first-order conditional method with interaction. Goodness-of-fit statistics were used to compare models. For goodness-of-fit statistics, NONMEM uses the objective function, which is approximately equal to −2 logs of the maximum likelihood. A ΔOFV (−2 log likelihood; approximate χ² distribution) of 3.84 points for 1 additional parameter and 5.99 for 2 additional parameters, respectively, was used for determining statistical significance (P < 0.05) of the differences between two models. Model comparison and evaluations were also assessed by goodness-of-fit plots and visual predictive checks (VPC). The figures were generated with GraphPad Prism (version 4.03).

Ethical approval.

All the applied protocols were approved by the ethics committees of the two cantons of Basel (Ethikkommission beider Basel) and the responsible local authorities (Medical Research Coordination Committee of the National Institute for Medical Research in Tanzania and the National Ethics Committee for Health Research in Cambodia). Blood samples were obtained after written informed consent in the local language (Khmer or Swahili) from the participants or their responsible guardians.

RESULTS

The SNPs of interest for the CYP and NAT2 genes were determined by sequencing DNAs from 125 Cambodian and 150 Tanzanian malaria patients. The allele frequencies for each population are shown in Tables 4 and 5 (sequence data were not always available for all samples). Most allele frequencies were found to be in Hardy-Weinberg equilibrium, except for the Tanzanian study population for CYP2B6*5 and CYP2D6*4 and in the Cambodian study population for CYP2B6*5, CYP2C9*3, CYP2D6*10 (4180G→C), and CYP2D6*17 (2850C→T).

The frequency of mutations varied with the enzyme and the population (Tables 4 and 5). Higher frequencies of mutant alleles were found in Cambodians than in Tanzanians for CYP2C9*3 (6.6% versus 0%), CYP2D6*10 (100C→T; 60.6% versus 6.6%), CYP3A5*3 (64.5% versus 18.4%), NAT2*6 (39.8% versus 26.7%), and NAT2*7 (20.5% versus 2.4%). In contrast, higher frequencies of mutant alleles were found in Tanzanians for CYP2D6*17 (1023C→T; 20.3% versus 0%), CYP2D6*17 (2850C→T; 60.6% versus 7.7%), CYP3A4*1B (73.5% versus 3.7%), NAT2*5 (36.4% versus 6.6%), and NAT2*14 (14.5% versus 2.5%). Significant differences in frequencies were found in 10 SNPs, while no significant differences were observed in 8 SNPs (Tables 4 and 5). The F_ST value was 0.16 at the 5% significance level. Many studies are in agreement that F_ST values in world samples of human populations are between 0.10 and 0.15 (33). The somewhat higher figure in the present study indicates that the two populations are distinct.

The population pharmacokinetic-pharmacogenetic analysis was performed on the data collected from 143 and 135 of the total of 150 patients recruited in Tanzania for LF and AM, respectively. In Cambodia, 63 out of 64 were included in the analysis of MQ and 60 of 61 were included in the analysis of PPQ. Reasons for exclusion were low hemoglobin levels, inability to swallow the drug, impossibility to take a blood sample, and withdrawal of consent (details are provided in another report by Staehli Hodel et al. (submitted). The pharmacogenetic profile of the patients included in the analysis is summarized in Table 6.

Table 6.

Allele frequencies in patients included in the population pharmacokinetic model

Gene	Frequency of allele in patients who received indicated drug(s) (n)a
	AM/DHAb (135)	LF/DLFb (143)	AS/DHAb and MQ (63)	DHA (56)	PPQ (60)
CYP2A6*2	0 (134)	0 (142)	0 (59)	0 (55)	0 (59)
CYP2B6*5	1.9 (134)	1.8 (142)	5.2 (58)	1.8 (56)	1.7 (60)
CYP2B6*6	34 (110)	34 (116)	42 (43)	33 (44)	33 (46)
CYP2C8*3	0 (62)	0 (66)	0 (36)	0 (33)	0 (37)
CYP2C9*3	0 (118)	0 (125)	8.5 (59)	4.5 (56)	4.2 (60)
CYP2C9*5	0.9 (118)	0.8 (125)	0 (59)	0 (56)	0 (60)
CYP2C19*3	0.8 (128)	0.7 (136)	3.3 (60)	2.7 (56)	2.5 (60)
CYP2D6*4	4.1 (122)	4.3 (128)	0 (50)	0 (20)	0 (22)
CYP2D6*10 (100C→T)	6.4 (125)	6.9 (130)	65 (52)	53 (32)	51 (35)
CYP2D6*10 (4180G→C)	75 (128)	75 (136)	72 (58)	69 (56)	70 (60)
CYP2D6*17 (1023C→T)	22 (105)	21 (112)	0 (50)	0 (13)	0 (14)
CYP2D6*17 (2859C→T)	65 (36)	62 (38)	11 (19)	7.1 (28)	6. 5 (31)
CYP3A4*1B	73 (121)	73 (129)	4.7 (53)	1.3 (39)	2.4 (41)
CYP3A5*3	19 (130)	19 (138)	70 (59)	59 (56)	61 (60)
NAT2*5	36 (129)	37 (137)	5.9 (59)	8.0 (56)	7.5 (60)
NAT2*6	27 (134)	27 (142)	35 (58)	46 (56)	44 (60)
NAT2*7	2.2 (134)	2.5 (142)	23 (59)	19 (55)	20 (59)
NAT2*14	14 (127)	14 (135)	5.1 (59)	0 (56)	0 (60)

^aFrequencies are percentages of patients in whom the SNP was found, from among the total patient population that received the drug(s). Values in parentheses are numbers of patients for whom SNP data were available for each group (compared to the total n, show at the top of each column).

^bParent drug/metabolite.

For LF, inclusion of genotype CYP3A4*1B did not significantly improve the fit, suggesting a very small change in CL of 13% (95% confidence interval [CI₉₅], −83% to 110%) in Hom-VAR compared to Het-VAR and Hom-REF carriers (ΔOFV = −2.2; P = 0.13). The inclusion of genotype CYP3A5*3 did not improve the model fit either (ΔOFV = −1.3; P = 0.25), with an estimated 13% decrease (CI₉₅, −48% to 230%) in CL in Hom-VAR compared to Hom-REF individuals. For AM, no statistical improvement of the model was observed after inclusion of CYP2B6*5 and CYP2B6*6 variants (ΔOFV < −2.9; P = 0.09), yielding a modest 20% increase (CI₉₅, −17% to 70%) in CL Hom-VAR compared to the other genotypic groups. No effect of CYP3A4*1B nor of CYP2C19*3 was observed either (ΔOFV = 0). Neither CYP3A5*3 nor CYP3A4*1B influenced MQ significantly (ΔOFV < −2; P = 0.16), with less than a 10% change (CI₉₅, −15% to 35%) in the CL estimated between Het-VAR and Hom-REF genotypic groups. Finally, among all tested covariates of PPQ CL, a small 15% decrease (CI₉₅, −68% to 39%) in elimination was estimated in the individuals carrying the CYP2D6*17 (2859C→T) allele, which did not reach statistical significance (ΔOFV = −0.8). No effects of other genetic variants were found to significantly affect PPQ elimination (ΔOFV < 0.6; P > 0.4).

DISCUSSION

This study investigated the effects of single nucleotide polymorphisms in CYP and NAT2 genes on the population pharmacokinetic profiles of ACTs. For this purpose, alleles of CYP and NAT2 genes involved in drug metabolism of antimalarials (CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP3A4, CYP3A5, and NAT2) were assessed in malaria patients from Cambodia and Tanzania. We found alleles that showed similar frequencies in Cambodians and Tanzanians and others that varied greatly between the two populations. The largest differences were found for CYP2D6, CYP3A4, CYP3A5, and NAT2. These results are in agreement with previous reports that allele frequencies of CYP2D6 and NAT2 generally vary considerably between continents or even countries (34, 35). However, there is little information available on allele frequencies of CYP and NAT2 gene polymorphisms specifically for Tanzanian or Cambodian populations. Sistonen et al. investigated CYP2D6 haplotype frequencies in single populations, and in Cambodia they found no polymorphisms in CYP2D6*4 or CYP2D6*17 genes but a 54.5% mutation rate in CYP2D6*10 (34), which is comparable with our data. A study on CYP2D6 genotypes in Tanzania also reported a low allele frequency for CYP2D6*4 (1%) and an intermediate frequency for CYP2D6*17 (17%) (36), while for the latter we found a much higher frequency (60.6%). As for CYP2C19*3, the observed low frequency confirmed previous findings from a study for Tanzania, where no CYP2C19*3 alleles were found (37). The allele frequencies we have reported for CYP2B6*6 (516G→T), CYP3A4*1B, and CYP3A5*3 are very similar to those previously described for three Tanzanian populations: one study from Zanzibar reported allele frequencies of 32.0% for CYP2B6*6 (516G→T), 69.2% for CYP3A4*1B, and 15.8% for CYP3A5*3, and two studies from mainland Tanzania reported allele frequencies of 73.2% for CYP3A4*1B and 19.0% for CYP3A5*3 (38–40).

In general, the allele frequencies found in our study show high concordance with data from the National Center for Biotechnology Information (NCBI) SNP home page (note: data for Asians and Africans are available only for CYP2A6*2, CYP2B6*5, CYP2B6*6 [516G→T], CYP2C8*3, CYP2C9*3, CYP2C19*3, CYP3A4*1B, and CYP3A5*3) (41). Furthermore, the F_ST value of 0.16 at the 5% significance level implies that the degree of differentiation between the two populations is moderate. The fact that the majority of samples were found to be in Hardy-Weinberg equilibrium indicates that sampling was unbiased. The few cases where population data were found not to reflect Hardy-Weinberg proportions were most likely due to the rather small sample size (and thus might represent chance findings) or due to the Wahlund effect, i.e., the reduction in heterozygosity because of highly diverging subpopulations.

A limitation of our SNP analysis was the small number of genotyping results for some of the CYP2D6 loci, which may be ascribed to the PCR products being too short or the PCR efficiency being too low. Short PCR products tend to fail when sequencing due to stuttering at the beginning of sequencing. The PCR primers were initially designed for use in a DNA microarray, and length of the PCR product was not considered. Due to budgetary constraints, a redesign was not feasible. It is also possible that PCR efficiency was reduced by mismatches incorporated into the primer in order to avoid amplification of the pseudogene CYP2D7. For future studies, primers should be redesigned, especially for CYP2D6.

To investigate the effects of SNPs in CYP and NAT2 genes on the CL of ACTs, and thus to address the main question of this study, the individual pharmacogenetic profiles of the patients acquired by sequencing were incorporated into population pharmacokinetic models for AM, LF, MQ, and PPQ (Staehli Hodel et al., submitted).

For AM, the effects of CYP2B6, CYP2C9, CYP2C19, and CYP3A4/5 genotypes were assessed. Previous studies of AM conversion to DHA showed that the turnover was highest for human recombinant CYP2B6 (42). In the present study, CYP2B6*6 polymorphism was not associated with AM elimination. While intestinal CYP3A4 might play a role in the presystemic metabolism of AM (43, 44), interaction studies indicated that liver CYP3A4 is not important in the in vivo metabolism of AM (28). This is in line with our finding that there was no difference in CL according to genetic variations of CYP3A4. In addition, it has been shown that CYP3A4 genotyping only poorly reflects the variability of CYP3A4 activity (45). Likewise, no significant difference in the pharmacokinetic profile of AM was seen in carriers of CYP2C19*3 compared to carriers of the respective reference allele. In our study, about one-third of patients were carriers of at least one CYP3A5*3 allele, which is the most frequent and functionally important SNP for the CYP3A5 gene (46). The mutation confers low CYP3A5 protein expression as a result of improper mRNA splicing and reduced translation of a functional protein (46, 47). However, including the pharmacogenetic profile of CYP3A5 did not significantly improve the fit.

LF, the long-acting component of artemether-lumefantrine, is metabolized by CYP3A4 (48). In our findings, no influence of CYP3A4*1B was detected. Inconclusive data on alterations of the metabolism in CYP3A4*1B carriers are reported in the literature. Some investigators have suggested enhanced drug elimination associated with variations in this gene (49, 50), whereas others have reported decreased metabolism of drugs in carriers of CYP3A4*1B among cancer patients (51, 52). It is uncertain whether interindividual differences in drug metabolism can be attributed to the allelic variant CYP3A4*1B (47, 49, 53–58); it has also been suggested that a linkage disequilibrium between CYP3A4*1B and CYP3A5*1A (and thus increased CYP3A5 expression) could be the actual cause of the altered metabolism (47, 58, 59). No effect of CYP3A5*3 could be detected in our study.

For MQ and PPQ, none of the SNPs tested seemed to affect their kinetics. Only one individual carrying the Hom-VAR CYP2D6*17 showed a modest decrease in PPQ elimination. This is in line with previous study results, indicating diminished CYP2D6 activity (60, 61) in carriers of CYP2D6*17 alleles. The PPQ elimination mechanism is still unknown, and the influence of this SNP needs to be confirmed in a larger population.

To the best of our knowledge, this is the first study to assess the impact of pharmacogenetic profiles on the clearance of several antimalarial drugs currently in use in Southeast Asia and Africa. Pharmacogenetics could not explain differences in treatment outcomes with AL between Tanzanians and Cambodians (18, 19). We cannot exclude the possibility that other genetic factors not tested in this candidate approach might contribute to treatment outcome. Thus, our results suggest that pharmacogenetic data need not be included in determining appropriate first-line treatments for uncomplicated malaria in specific populations. Possible reasons for our results might be that (i) the pharmacogenetic profiles were not different enough for the panel tested, or (ii) the roles of tested CYP genes in metabolism of drugs used in our study populations are not large enough to show effects in vivo. Furthermore, DHA, the primary metabolite of AM, is more potent than the parent drug, and the very high intrapatient variability might have limited the possibility to detect genetic association with drug disposition. The limitation of this study was the low allelic frequency of poor metabolizers, which prevented us from detecting a true difference of 30% in drug elimination with most alleles. Further studies might be more relevant if blood is taken from patients with treatment failure and to compare their pharmacogenetic profiles with those of patients with an adequate clinical and parasitological response (ACPR).