CYP2D6 Polymorphisms and the Safety and Gametocytocidal Activity of Single-Dose Primaquine for Plasmodium falciparum

Single-dose primaquine (PQ) clears mature gametocytes and reduces the transmission of Plasmodium falciparum after artemisinin combination therapy. Genetic variation in CYP2D6, the gene producing the drug-metabolizing enzyme cytochrome P450 2D6 (CYP2D6), influences plasma concentrations of PQ and its metabolites and is associated with PQ treatment failure in Plasmodium vivax malaria.

class of 8-aminoquinolines that has been on the market for more than 70 years. In recent years, the addition of PQ to ACTs has received considerable interest because of its ability to rapidly clear mature P. falciparum gametocytes and reduce the infectious period compared to ACT alone (1)(2)(3)(4)(5)(6).
Cytochrome P450 2D6 (CYP2D6) is a human enzyme involved in the metabolization of 20 to 25% of all prescribed medicines (7)(8)(9)(10). Hundreds of different CYP2D6 alleles have been discovered, some of which influence the activity of the produced enzyme (11). Bennett and colleagues first associated genetic CYP2D6 variation with relapses of Plasmodium vivax malaria after PQ treatment (12). More recently, genetic CYP2D6 variation was found to be strongly associated with an increased risk of relapses among Indonesian patients with clinical P. vivax malaria (7). There is also evidence in mice that enzymes in the CYP2D family produce the active metabolite of PQ against Plasmodium berghei liver stages (13), but metabolic activation of PQ may not be necessary to eradicate blood stages (14).
The implications of genetic CYP2D6 variation for the use of PQ in P. falciparum infections have never been explored. One of the factors that has hindered the widespread adoption of PQ for P. falciparum transmission reduction is its safety profile, notably in individuals with genetic deficiencies in glucose-6-phosphate dehydrogenase (G6PD) production (12,(15)(16)(17). G6PD is an enzyme involved in the pentose phosphate pathway in human red blood cells (18), and G6PD deficiency (G6PDd) is associated with hemolysis following treatment with PQ. Despite safety concerns related to the hemolytic activity of PQ in individuals with G6PDd, a single low dose of PQ is considered safe in individuals with the most common African G6PDd variant (G6PDd A variant) (19)(20)(21). Since genetic variation in CYP2D6 influences the pharmacokinetics of single-low-dose PQ in humans (22), this variation may have implications for PQ efficacy or safety at doses targeting P. falciparum gametocytes.
Here, we determine the impact of genetically inferred CYP2D6 metabolizer status on the gametocytocidal and hemolytic effect of single-dose PQ in 8 clinical trials conducted across Africa.

RESULTS
CYP2D6 genotyping with the OpenArray technology used here requires high-quality DNA, ideally 50 ng/l, a condition that was not always met. CYP2D6 genotyping was thus successful in 72% (774/1,076) of all samples; success varied considerably between sample types, with good success rates for saliva samples (Ն98%) and large-volume blood samples (Ն0.5 ml blood) (success rate of Ն87%) but low success rates for different sample types (1 to 68%) ( Table 1; see also Data Set S1 in the supplemental material). As a result of differences in sample collection methods between sites, genotyping was successful for Յ58% of samples from Uganda and Balonghin, Burkina Faso, but for Ն80% of samples for other sites (Table 1 and Data Set S1). Inference of the CYP2D6 activity score (AS) from genotypes was successful in 68% (731/1,076) of samples and is presented for the different sites in Fig. 1. The CYP2D6 AS inference allowed classification of sample donors as poor metabolizer (PM) (activity score of 0), intermediate metabolizer (IM) (activity score of 0.5 to 1.0), extensive metabolizer (EM) (activity score of 1.5 to 2.0), or ultrarapid metabolizer (UM) (activity score of Ͼ2.0). For other samples, a range of ASs could be inferred that allowed classification into EM/UM classes (AS Ն 1.5; n ϭ 137) (Data Set S2). CYP2D6 PM status was inferred for a minority of individuals (2.6%; 19/731); CYP2D6 IM status was inferred for 38.2% of individuals (279/731).
A total of 544 participants from 5 studies who had gametocytes by molecular methods on the day of initiation of treatment, completed treatment, and had complete outcome measures were included in the efficacy analysis. The prevalence of CYP2D6 PM/IM status among these individuals was 31.4% (171/544) overall and ranged from 26% to 41% by study. Compared to ACT alone, PQ was effective in reducing gametocyte prevalence on day 7 or 10 in both CYP2D6 EM/UM (odds ratio [  P ϭ 0.129) in models adjusted for country, PQ dose, and baseline gametocyte density. We thus found no evidence for selection bias in our efficacy and safety outcome assessments due to variation in CYP2D6 genotyping success.

DISCUSSION
In the present study, we utilized samples from clinical trials across Africa to explore the effect of genetically inferred CYP2D6 metabolizer status on PQ efficacy and safety. Compared to ACT alone, the addition of single-dose PQ resulted in a marked reduction in gametocyte carriage across populations with different CYP2D6 metabolizer statuses. Nevertheless, CYP2D6 PM/IM individuals were more likely to have persisting gametocytes until day 7 or 10 following initiation of treatment with ACT-PQ.
While the transmission-blocking effect of PQ may precede the gametocyte-clearing effect and gametocytes persisting after PQ may not result in onward transmission to mosquitoes (1,6,23), the results of the present study suggest that the efficacy of low-dose PQ may be affected by CYP2D6 metabolizer status. We previously demonstrated that PQ pharmacokinetics are influenced by genetically inferred CYP2D6 metabolizer status (22), suggesting that lower concentrations of the PQ active metabolites may occur in CYP2D6 PM/IM individuals. While CYP2D6 metabolizer status and concentrations of active PQ metabolites have direct implications for P. vivax-infected patients by affecting cure rates (12), the effect on P. falciparum-infected patients is indirect, potentially increasing the number of secondary cases arising from a PQ-treated gametocyte carrier.
We observed no effect of CYP2D6 metabolizer status on Hb concentrations after PQ treatment of G6PDd individuals. We hypothesized that G6PDd individuals with CYP2D6 PM/IM status would be relatively protected from hemolysis, but this was not observed. While we combined data from safety studies to maximize the number of observations in G6PDd individuals, it is possible that our study population size was insufficient to detect subtle effects on hemolysis. Interstudy variation may also have obscured effects of CYP2D6 status, although study site was incorporated into our multivariate regression models.
There are several limitations to this study. We worked with available samples from several clinical trials, not specifically collecting material for extensive human genotyping. The variable quality and quantity of samples affected our genotyping success rate but are unlikely to have affected the validity of our comparisons between populations with successful genotyping results. Similarly, the present study did not allow us to detect possible differences in effects between ACTs. CYP2D6 activity and PQ metabolism may be influenced differently by dihydroartemisinin-piperaquine (DP) (24) and artemether-lumefantrine (AL) (25). While we combined findings from trials with different ACTs, this is unlikely to have affected the validity of our findings, and we adjusted for study effects. Another limitation is that we inferred CYP2D6 metabolizer status from the CYP2D6 genotype. There has been a series of publications describing situations where the commercially available TaqMan assays, also used here in the OpenArray format, have not worked as expected and have been redesigned (26)(27)(28)(29). Most significantly, one assay variant detecting CYP single nucleotide polymorphisms (SNPs) (*15 allele; C_32407245_40) suffers from interference from the sequence of the pseudogene CYP2D7 to the extent that these results were not included in the analysis (27). Some additional assays have been replaced with new and improved ones during the course of this study (*7 assay [C_32388575_30 with C_32388575_A0], *8 assay [C_30634117C_20 with C_30634117C_K0], and *14 assay [C_30634117D_30 with C_30634117D_M0]) (30). In addition, a copy number variation (CNV) assay targeting intron 2 (Hs04083572_cn) may not always give the correct result due to intronic polymorphisms, and CNV assays in general work only with high sample quality (and not after product preamplification). These challenges in genetic analysis underline the complexity of the locus and the need for more sequencing of CYP2D6. Especially in African populations for which pharmacogenetic data are lacking, additional data are needed (31). Such future studies may purposefully collect select samples for human genotyping. In our studies, 0.5 to 1 ml blood collected in ethylenediaminetetraacetic acid (EDTA)-coated tubes or Oragene saliva samples resulted in high genotyping success rates (Table 1). Another option is to perform CYP2D6 phenotyping experiments, where a probe substrate to assess CYP2D6 activity is used. Although substrate specificity may complicate extrapolation of data from such assays to PQ metabolism, an unquestionable advantage of phenotyping is that it would take into consideration environmental factors influencing CYP2D6 activity. These include, but are not limited to, comorbidities, concomitant medication, and food intake (32,33).
Despite limitations, including the modest number of observations from individuals with the genetically inferred CYP2D6 PM phenotype, we present evidence that CYP2D6 PM/IM status is associated with prolonged gametocyte carriage after treatment. It is currently unclear whether this has implications for the transmission-blocking effects of PQ at the population level in malaria elimination settings. A clinically meaningful effect of genetically inferred CYP2D6 metabolizer status on PQ-induced hemolysis in G6PDd individuals is unlikely.

MATERIALS AND METHODS
Study samples. Samples from 8 published clinical trials were used for separate analyses on the impact of genetically inferred CYP2D6 metabolizer status on PQ safety and efficacy. For analyses of the impact of CYP2D6 inferred metabolizer status on PQ efficacy, we included samples from 5 PQ efficacy studies. Gametocyte detection was performed following treatment with a single dose of 0.1 to 0.75 mg/kg PQ in combination with either artemether-lumefantrine (AL) (Coartem as a standard 6-dose regimen over 3 days; Novartis Pharma, Switzerland) in Burkina Faso (3) and Uganda (2) or dihydroartemisinin-piperaquine (DP) (Eurartesim as a standard 3-day regimen; Sigma-Tau, Italy) in Mali (1), the Gambia (4), and Kenya (5). Analyses on the impact of CYP2D6 inferred metabolizer status on hemolysis were restricted to G6PD-deficient (G6PDd) individuals; we included two additional studies that specifically assessed PQ safety in G6PD-deficient individuals in Mali (20) and the Gambia (19), using 0.25 to 0.5 mg/kg PQ in combination with DP. In all studies, hemoglobin (Hb) concentrations in grams per deciliter were measured by a self-calibrating HemoCue photometer (Ängelholm, Sweden). Study details are summarized in Table 1.
Extraction of nucleic acids. An automated MagNA Pure LC 2.0 instrument (Roche, Switzerland) was used for extraction of total nucleic acid (NA) or DNA. For the samples from Uganda as well as the parasitology samples from Mali, a MagNA Pure LC high-performance total nucleic acid isolation kit was used. For samples from Burkina Faso, Kenya, and the first season of the trial in the Gambia (both full blood in EDTA and saliva samples), MagNA Pure LV DNA isolation kits were used. The saliva samples collected after the second season of the trial in the Gambia were extracted using a Maxwell 16 instrument (Promega, USA) and Maxwell 16 DNA purification kits. Concentration measurements were done using a NanoDrop device (Thermo Fisher, USA) (only DNA from full blood in EDTA) and a Qubit fluorometer (Thermo Fisher, USA) with the Qubit HS (high-sensitivity) kit, which is specific for double-stranded DNA (dsDNA).