Pregnancy and CYP3A5 Genotype Affect Day 7 Plasma Lumefantrine Concentrations

Pregnancy and pharmacogenetics variation alter drug disposition and treatment outcome. The objective of this study was to investigate the effect of pregnancy and pharmacogenetics variation on day 7 lumefantrine (LF) plasma concentration and therapeutic responses in malaria-infected women treated with artemether-lumefantrine (ALu) in Tanzania. A total of 277 (205 pregnant and 72 nonpregnant) women with uncomplicated Plasmodium falciparum malaria were enrolled. Patients were treated with ALu and followed up for 28 days. CYP3A4, CYP3A5, and ABCB1 genotyping were done. Day 7 plasma LF concentration and the polymerase chain reaction (PCR) – corrected adequate clinical and parasitological response (ACPR) at day 28 were determined. The mean day 7 plasma LF concentrations were significantly lower in pregnant women than nonpregnant women [geometric mean ratio = 1.40; 95% confidence interval (CI) of geometric mean ratio (1.119–1.1745), P < 0.003]. Pregnancy, low body weight, and CYP3A5*1/*1 genotype were significantly associated with low day 7 LF plasma concentration (P < 0.01). PCR-corrected ACPR was 93% (95% CI = 89.4–96.6) in pregnant women and 95.7% (95% CI = 90.7–100) in nonpregnant women. Patients with lower day 7 LF concentration had a high risk of treatment failure (mean 652 vs. 232 ng/ml, P < 0.001). In conclusion, pregnancy, low body weight, and CYP3A5*1 allele are significant predictors of low day 7 LF plasma exposure. In turn, lower day 7 LF concentration is associated with a higher risk of recrudescence. SIGNIFICANCE STATEMENT This study reports a number of factors contributing to the lower day 7 lumefantrine (LF) concentration in women, which includes pregnancy, body weight, and CYP3A5*1/*1 genotype. It also shows that day 7 LF concentration is a main predictor of malaria treatment. These findings highlight the need to look into artemether-LF dosage adjustment in pregnant women so as to be able to maintain adequate drug concentration, which is required to reduce treatment failure rates in pregnant women.


Introduction
Artemether-lumefantrine (ALu) has been the first-line drug for the treatment of uncomplicated malaria in Tanzania since the end of 2006. It is also the antimalarial drug of choice for the treatment of uncomplicated malaria in pregnant women in the second and third trimesters (http://www.who.int/selection_medicines/country_lists/Tanzania_STG_ 052013.pdf). Day 7 lumefantrine (LF) concentration has been shown to be a potent predictor of the treatment outcome in different study populations and correlates with the area under the concentration time curve (White et al., 2008; WorldWide Antimalarial Resistance Network (WWARN) Lumefantrine PK/PD Study Group, 2015). To be able to achieve adequate clinical and parasitological response (ACPR) during the course of ALu treatment, adequate LF plasma concentration after complete elimination of artemether and dihyroartemisinin needs to be achieved and maintained long enough. Sufficient LF plasma concentration is essential for the drug to be able to clear all residual malaria parasites and reduce the risk of the treatment failure and drug resistance among patients.
Lower LF plasma exposure has been described in young children (WorldWide Antimalarial Resistance Network (WWARN) Lumefantrine PK/PD Study Group, 2015;Kloprogge et al., 2018), pregnant women (Kloprogge et al., , 2018Mosha et al., 2014;Mutagonda et al., 2016), smokers (Tarning et al., 2009), or when ALu is taken unsupervised (WorldWide Antimalarial Resistance Network (WWARN) Lumefantrine PK/PD Study Group, 2015), without fat meal (Borrmann et al., 2010;Jain et al., 2017) or with concurrent use of efavirenz (Byakika-Kibwika et al., 2012;Huang et al., 2012;Maganda et al., 2015;Banda et al., 2018) and rifampicin (Lamorde et al., 2013;Olafuyi et al., 2017). Physiologic, immunologic, and hormonal changes during pregnancy not only increase the risk of acquiring malaria but also alter the pharmacokinetic profile of drugs used for treatment of malaria (Rogerson et al., 2007). The metabolism of drugs catalyzed by CYP3A4 and CYP3A5 is increased during pregnancy, and these enzymes metabolize both artemether and LF (Anderson, 2005;Isoherranen and Thummel, 2013). LF is extensively metabolized into desbutyl-benflumetol in the liver mainly by CYP3A4 enzymes (Staehli-Hodel et al., 2013) and eliminated through the bile via p-glycoprotein, a major biliary efflux pump encoded by ABCB1 (Oga et al., 2012;Wahajuddin et al., 2014). Due to the considerable sequence similarity and overlapping substrate specificity with CYP3A5, it is difficult to determine the independent effect of CYP3A4 variability on the safety and efficacy of CYP3Amediated drug metabolism, particularly in black population where CYP3A5 enzyme is primarily expressed (Williams et al., 2002;Klein and Zanger, 2013). CYP3A5 expression level and enzymatic activity display significant variation between populations due to genetic polymorphisms. Most whites are non-CYP3A5 expressors because of the high frequency of CYP3A5*3 allele in this population. In contrast, about 70% of the black African population are CYP3A5 expressers, as they carry one or two CYP3A5*1 alleles. CYP3A5 genetic variation may therefore contribute significantly to the variation in metabolism of CYP3A substrates in black Africans.
Because of polymorphic expression and wide interethnic variations in variant allele frequency distributions, CYP3A5 genotype may be the most important contributor to interindividual and interethnic differences in CYP3A-dependent drug disposition (Mukonzo et al., 2010). Indeed, previous studies in different populations reported significant influence of CYP3A5 genotype for variability in metabolism of quinine (Mukonzo et al., 2010), a known CYP3A4 probe drug, and 4b-hydroxycholesterol (Diczfalusy et al., 2008;Gebeyehu et al., 2011), an endogenous marker for CYP3A4 activity.
Recently, few studies highlighted relevance of pharmacogenetics variations in determining patient variability in plasma LF exposure Mutagonda et al., 2017;Vos et al., 2017). In a small sample size study, it was reported that pregnant women who were carriers of functional CYP3A5 and CYP3A4*1B allele had lower day 7 LF concentrations and recrudescence, respectively, compared with the carriers of defective alleles (Mutagonda et al., 2017). However, a study in nonpregnant population did not yield significant associations between LF pharmacokinetic parameters with CYP3A and ABCB1 single-nucleotide polymorphisms (Staehli-Hodel et al., 2013).
Increased CYP3A enzyme activity during pregnancy due to induction (Jeong, 2010) may alter LF disposition in a significant manner. Previous studies reported the relevance of genetic variation on CYP3A enzyme induction (Habtewold et al., 2013;Ngaimisi et al., 2014). This raises the hypothesis that genetic variations in drug-metabolizing enzymes and transporters might significantly contribute to variability in day 7 LF concentrations and treatment outcome in pregnant and nonpregnant women. Therefore, the aim of this study was to examine the effects of pregnancy and role of genetic variations in CYP3A4, CYP3A5, and ABCB1 on day 7 LF concentration and malaria therapeutic response.

Materials and Methods
Study Design, Population, and Procedures. This was an observational prospective cohort study conducted at Kisarawe, Kibiti, Mkuranga, and Rufiji districts located in the Coast Region in Eastern Tanzania. The reported prevalence of malaria in pregnancy during the study period was 8.1% and Plasmodium falciparum was the predominant species (Mutagonda et al., 2016). The study was carried out at Kisarawe, Mkuranga, and Utete district hospitals and at Mohoro and Kibiti health centers from May 2014 to August 2017. This study was approved by the Muhimbili University of Health and Allied Sciences (MUHAS) and National Institute for Medical Research ethical committees. All women consented to the study before enrollment. Women's identification number was used throughout the data collection and analysis period so as to ensure confidentiality. All patients' information was filled in the confidential case report forms.
Pregnant and nonpregnant women diagnosed with uncomplicated P. falciparum malaria infection detected by microscopy were recruited and enrolled from the Reproductive and Child Health clinic and outpatient department. Inclusion criteria were women aged $18 years (married women ,18 years were also included), resident of study areas, and with hemoglobin level of $8 g/dl. Exclusion criteria were pregnant women in the first trimester, allergic to ALu, unable to take oral medication, or vomited the medication within 1 hour of taking the dose, and reported intake of any antimalarial drug within the past 28 days. Also, patients with a history of renal, liver, or heart problems, or with severe malaria were excluded.
Sample Size. The calculation of sample size for this study used the classic statistical tool recommended by the Technical Expert Group on Malaria Chemotherapy to calculate the sample size (https://apps.who.int/iris/bitstream/ handle/10665/44048/9789241597531_eng.pdf;jsessioni). The sample size was determined based on an expected proportion of treatment failures in both groups (pregnant and nonpregnant women), 95% confidence level, and 5% precision. The estimated population proportion of clinical failures in pregnant women was 18% (Mosha et al., 2014) at confidence level of 95% and precision of 5%. Therefore, a minimum sample size of 196 pregnant women would be needed for this study. For nonpregnant women, P was 5% (Mosha et al., 2014) at confidence level of 95% and precision of 5%, thus requiring 73 nonpregnant women as a minimum sample size for this study. Based on World Health Organization recommendations, in order for the efficacy study to be representative, a minimum sample of 50 patients is required, regardless of the rates of failure (https://apps. who.int/iris/bitstream/handle/10665/44048/9789241597531_eng.pdf;jsessioni). In this study, we recruited 205 pregnant and 72 nonpregnant women, hence sufficient to describe the efficacy of LF in the population.
Treatment, Clinical, and Laboratory Procedures, and Patient Follow-Up. All women received six doses of four tablets of ALu (Coartem; Novartis Pharma AG, Basel, Switzerland) (20 mg artemether and 120 mg LF) over the course of 3 days at 0, 8, 24, 36, 48, and 60 hours. For each patient, general, physical, and clinical examinations, including axillary temperature measurements and evaluation of malaria-related symptoms, were performed at enrollment and on the follow-up visits on days 2, 7, 14, 21, and 28. Full medical history, including current illnesses and medications used, was recorded. Adherence to ALu intake was assessed using a self-administered questioner and by pill counting on day 2 visit. Pregnant women gestational age was determined from the estimated first day of the last normal menstrual period and compared with clinical examination of a fundal height.
For pharmacogenetics analysis, 1 ml whole blood was taken into an EDTAcontaining vacutainer tube and stored at 280°C at MUHAS laboratory. Also, approximately 50 ml blood was collected at enrollment (day 0) and on follow-up days (day 7, 14, 21, and 28) on a filter paper (Whatman grade 3) for screening and genotyping of malaria parasite using polymerase chain reaction (PCR). These filter papers were air dried and stored separately in a plastic bag and kept frozen at 280°C. For pharmacokinetics, 3 ml venous blood was collected in heparinized tubes at the enrollment day and on day 7 (corresponding to 168 hours) following initiation of ALu treatment to be able to determine plasma LF concentrations. The whole blood was centrifuged at 2000g for 5 minutes. Plasma obtained after centrifugation was stored in cryo tubes at 280°C until analysis.
CYP3A4, CYP3A5, and ABCB1 Genotyping. In the pharmacogenetics analysis, genomic DNA was isolated in whole-blood samples using QIAamp DNA Midi Kit (Qiagen GmbH, Hilden, Germany), according to the manufacturer's instructions. Genotyping for the common functional variant alleles for ABCB1 c.3435C.T and ABCB1 c.4036A.G (rs3842), CYP3A4*1B, CYP3A5*3, CYP3A5*6, and CYP3A5*7, which have been reported to be relevant for LF disposition (German and Aweeka, 2008;Djimdé and Lefèvre, 2009), was carried out at the Division of Clinical Pharmacology, Department of Laboratory Medicine, Karolinska Institutet, Sweden, as previously described (Mutagonda et al., 2017). The Quant Studio 12 K Flex Real-Time PCR system (Life Technologies Holding, Singapore) was used for genotyping.
Parasite Screening and Genotyping. Dried blood spots on filter papers were punched, and three punched-out circles 3 mm in diameter were used for DNA extraction using QIAamp DNA blood micro kit (Qiagen GmbH) following the manufacturer's recommendations. Screening of Plasmodium parasites was performed using a species-specific PCR targeting the ssRNA gene, as previously described (Shokoples et al., 2009;Mutagonda et al., 2017). P. falciparum genotyping was performed by nested PCR using fluorescent primers in the nested allelic specific reaction, followed by capillary electrophoresis in a DNA sequencer for fragment sizing, as previously described (Liljander et al., 2009;Mutagonda et al., 2017).
Data Analysis. x 2 test was used to compare the observed and expected genotype frequencies, according to the Hardy-Weinberg equilibrium. Patients with LF concentrations . LLOQ before treatment were excluded from this analysis, and the predose concentration was set to 0 if , LLOQ. LF plasma concentration data were log 10 transformed, and data were expressed as mean with 95% CI. Independent t tests were done to compare log day 7 LF between the pregnant and nonpregnant women. Univariate linear regression analyses were used to identify the individual effect of independent covariates on log-transformed day 7 LF concentration, followed by a stepwise multivariable regression analysis. The final model consisted of variables with P values ,0.05. Comparison of mean log day 7 plasma LF concentration based on the number of functional CYP3A5*1 variant alleles in all study participants and stratified by pregnancy status was analyzed using ANOVA.
Factors influencing malaria treatment outcome were evaluated using Cox regression analysis. Comparisons of treatment outcomes both PCR uncorrected and corrected ACPR and reinfection were performed using two analysis methods,

Results
Baseline Characteristics of Study Participants. A total of 277 women was enrolled in the study, including 205 pregnant and 72 nonpregnant women. Majority of pregnant women (85%) were in the second trimester, with the remainder being in the third trimester. Baseline data showed significant difference between the two groups. The median age, Hb levels, and baseline parasitemia were higher among nonpregnant compared with pregnant  women. Baseline characteristics of study participants are shown in Table 1. Genotype distribution for CYP3A4, CYP3A5, and ABCB1 was similar among pregnant and nonpregnant women, as presented in Table 2. There were no significant differences in the distribution of genotype or allele frequencies between the two groups. There was no significant deviation between the observed and expected genotype frequencies from Hardy-Weinberg equilibrium. Fig. 2. Comparison of mean day 7 LF plasma concentration based on the number of functional CYP3A5*1 alleles in all study participants and stratified by pregnancy status. Zero 5 homozygous for *3, *6, or *7; one 5 heterozygous for *3, *6, or *7; two 5 homozygous CYP3A5*1/*1 genotype. Vertical bars denote 95% CIs of the mean.
Carriers of both homozygous and heterozygous functional ABCB1 c.3435C.T alleles had lower LF concentration compared with the homozygous mutant carriers, but the results were not statistically significant (P . 0.05). Comparison of mean log day 7 plasma LF concentration based on the number of functional CYP3A5*1 variant alleles in all study participants and stratified by pregnancy status is presented in Fig. 2 using ANOVA. In a multivariable regression analysis and using backward stepwise elimination, only pregnancy, body weight, and CYP3A5 genotype were retained in the model as a significant predictor of day 7 LF plasma concentration (Table 3).
Malaria Treatment Outcome. Malaria treatment outcomes were defined following the World Health Organization protocol (https:// apps.who.int/iris/bitstream/handle/10665/44048/9789241597531_ eng.pdf;jsessioni) as ACPR, new infection, and treatment failure, designated as early treatment failure, late clinical failure, or late parasitological failure. Analyses of treatment response were performed using both intention-to-treat (ITT) population that included all enrolled patients and per protocol population, including all patients who were part of the ITT and did not deviate from the protocol for other reasons than failure.
Factors Associated with Treatment Failure. Day 7 LF concentration was significantly lower among women with therapeutic failure than those with ACPR.  )] ng/ml for those with ACPR and treatment failure, respectively (P , 0.0001). Comparison of mean log day 7 plasma concentration between patients with PCR-corrected ACPR (success) versus treatment failure stratified by pregnancy status is presented in Fig. 3.
In univariable analysis (Table 5), there was a significant association between risk of recurrent infection and day 7 LF concentration (P , 0.001), whereby, using the previous cutoff point of 600 ng/dl (Mutagonda et al., 2016), the risk of recurrence was 5.3 (95% CI) (1.483-19.066) times higher among patients with concentration ,600 ng/dl (P 5 0.010). The risk was also higher in pregnant women (1.6 times than nonpregnant), primigravid (1.5 times than multigravida), and higher among carriers of functional CYP3A4*1B (3.2 times) and ABCB1 c.3435 (1.1 times) alleles than mutants, but the association was not statistically significant (P . 0.05). Multivariable analysis could not be conducted due to lack of association of most of the study variable.

Discussion
In the present study, investigation on the effect of pregnancy, pharmacogenetic variations, and other clinical and sociodemographic factors on day 7 plasma LF concentration and malaria treatment outcome was conducted. The main findings include the following: 1) significantly lower day 7 LF plasma concentration in pregnant compared with nonpregnant women, 2) significant association of CYP3A5*1/*1 genotype and low body weight with low day 7 LF plasma concentration, and 3) low day 7 LF plasma concentration as a significant predictor of treatment failure. To date, there are limited studies investigating the effect of CYP3A and ABCB1 genotype on LF concentration and malaria treatment outcome Vos et al., 2017), and only one small sample size study reported effect of genotype on LF plasma concentration in pregnant women (Mutagonda et al., 2017). Therefore, the present study described the effect of pregnancy and CYP3A genotype on day 7 LF plasma concentration and malaria treatment outcome. To the best of our knowledge, this is the first study to investigate the effect of pregnancy and genotyping on day 7 LF concentration and malaria treatment outcome among pregnant and nonpregnant women living in the same area.
In this study, ALu was effective in both pregnant and nonpregnant populations with the cure rate (PCR-corrected) of .90% (lower in pregnant women than nonpregnant women). The most important determinant of malaria treatment outcome in both pregnant and nonpregnant women was day 7 LF plasma concentration, indicating that the concentration of ,600 ng/ml is associated with treatment failure. Day 7 LF concentration was lower by 27.1% in pregnant women compared with nonpregnant ones. This is similar to what has been reported previously from Africa (Tanzania, Uganda, and Rwanda) and Thailand (McGready et al., 2006;Tarning et al., 2009Tarning et al., , 2013Kloprogge et al., 2013;Mosha et al., 2014;Lohy Das et al., 2018). Lower drug concentrations in pregnant women contribute to lower cure rates and selection of parasite resistance. To achieve similar LF exposure to nonpregnant women, dosage adjustment has been suggested (Kloprogge et al., 2018). Pregnancy-related hormones such as estrogen and cortisol induce CYP3A activity, which can result in subtherapeutic plasma concentrations of drugs (Jeong, 2010). One of the determinants of day 7 LF concentration in this study was CYP3A5 genotype in gene dosedependent manner (Fig. 2). In a univariate analysis, CYP3A4*1B genotype was significantly associated with higher LF plasma concentration, but not in the multivariable analysis. The variant allele frequency of CYP3A4*1B is more than 70% in the study population, which is unlikely to account for the day 7 LF concentration observed between different CYP3A4*1B genotypes. CYP3A5 may contribute to the complexity of these unexplained differences, as almost all CYP3A4 substrates, with a few exceptions, are also metabolized by CYP3A5 (Desta and Flockhart, 2017). In this study, carriers of homozygous wildtype CYP3A5*1 allele had 15.5% times higher day 7 LF concentration compared with the homozygous wild type, and the difference was significant. Further analysis showed that pregnant women with at least one mutant CYP3A5*1 allele had a higher day 7 LF concentration compared with the carriers of homozygous wild-type alleles (Fig. 2). In HIV patients' cohort study, it was observed that patients receiving Efavirenz-based antiretoviral therapy (ART) (CYP3A inducer) had lower LF concentration compared with those receiving nevirapine-based ART and ART naive control group . Previous studies showed absence of CYP3A5*1 influence on LF concentration in nonpregnant population Vos et al., 2017), which could not be ascertained in this study due to a small sample size of nonpregnant women.
Weight of patients is among the important factors in the determination of the doses of ALu to be prescribed. For instance, patients with $35 kg Treatment of Malaria in Pregnancy body weight are given four tablets of ALu (each containing 20 mg artemether and 120 mg LF) at predefined interval (http://www.who.int/ selection_medicines/country_lists/Tanzania_STG_052013.pdf). In this study, it was observed that day 7 LF concentration was increased by 1.8% with increase per kg body weight. The effect of patients' weight on day 7 LF plasma concentration and subsequent malaria treatment outcome was reported in a systematic review and meta-analysis using individual patient data showing a substantially higher risk of recrudescent malaria and lower LF concentrations at day 7 with decreasing weight-for-age z-scores (WorldWide Antimalarial Resistance Network (WWARN) Lumefantrine PK/PD Study Group, 2015).
LF is a substrate of P-glycoprotein, and the effect of different transporters on LF concentration has been reported. The role of ABCB1c.3435 (Multi-Drug Resistance -MDR1) in HIV patients receiving Nevirapine-based ART and that of ABCC2 ( MDR2) in children under ,5 years has been described Vos et al., 2017). It has been reported that ABCB1 is highly expressed during pregnancy as a defense mechanism to protect the fetus from toxicity starting from the first trimester to term (Sun et al., 2006). In the current study, ABCB1 c.3435 did not show the significant association with day 7 LF concentrations, but lower LF concentration was observed among patients with the functional allele compared with homozygous mutants. The effect of ABCB1c.3435 would have been described well if the proportion of patients with homozygous mutation would have been higher. Because levels of protein expression decrease with advancing gestational age, it is also likely that inclusion of pregnant women in the first trimester would have given better association of the effect of ABCB1c.3435 on day 7 LF plasma concentrations (Mathias et al., 2005;Sun et al., 2006). However, for the purpose of this study, we could only recruit pregnant women in the second and third trimesters because ALu was contraindicated to be used during the first trimester of pregnancy at the time of study. There is still limited information on the effect of ABC transporter proteins on plasma LF concentration and malaria treatment outcome. Therefore, we recommend further studies to determine the association between LF concentration and malaria treatment outcome with other ABC drug transporters such as ABCB11 (bile salt export pump), ABCG2 (breast cancer resistance protein), and ABCC (multidrug resistance protein).
The efficacy of ALu in pregnant compared with nonpregnant women showed the ACPR of greater than 90%, indicating adequate therapeutic efficacy. The efficacy of ALu for treatment of uncomplicated malaria was not significantly different between pregnant and nonpregnant women. In this study, 86 pregnant women had day 7 LF concentration below 600 ng/ml, but only 14 (16.3%) had treatment failure, whereas in nonpregnant women 20 women had concentration below 600 ng/ml, but only three (15%) had treatment failure. Despite the significant differences in day 7 LF concentration between the two study populations, the majority of women with lower day 7 LF concentrations in both groups were still cured. Higher cure rates in malaria-endemic areas have been associated with the immunity that acts in synergy with antimalarial chemotherapy (White, 1997;Rogerson et al., 2010). Also, more than half of enrolled pregnant women were multigravida, which can also explain a nonsignificant difference in treatment outcome between pregnant and nonpregnant women, as shown in Table 4. However, the mean day 7 LF concentration between those who were cured and those with treatment failure was significant, as observed in Fig. 3. Similar

1422
findings were also reported in a meta-analysis examining the pharmacokinetic and pharmacodynamics of LF, in which the main determinant of treatment failure was lower LF exposure (Kloprogge et al., 2018). There was no early treatment failure or late clinical failure throughout the follow-up period of 28 days, an indication that ALu is still effective for treatment of uncomplicated malaria in pregnant women. There was no significant effect of the study variables such as CYP3A enzymes, ABCB1 transporters, obstetrics, and other patients' demographic characteristics on malaria treatment outcome.
The limitation for this study was a relatively small sample size of nonpregnant women compared with pregnant women. Due to availability of predetermined schedules and sensitization of pregnant women to attend antenatal clinics, it was much easier to recruit and retain them for the entire period of follow-up after treatment. In contrast, nonpregnant women with uncomplicated malaria are not admitted, and they only attend at the outpatient department when feeling unwell, so it was difficult to recruit and retain them for the entire period of follow-up after treatment. Despite differences in sample size in the two groups of women, the findings of this study show the importance of genetic variation in the CYP3A5 locus for LF pharmacokinetics and treatment outcome in pregnant women and nonpregnant women.
In conclusion, this study assessed the effect of CYP3A enzymes and ABCB1 transporter on day 7 ALu concentrations and treatment outcome in pregnant and nonpregnant women. Lower day 7 LF concentration was observed in pregnant women than nonpregnant. The determinants of day 7 LF concentration were pregnancy status, the number of CYP3A5*1, and weight of patients. Slightly lower therapeutic efficacy of ALu was observed in pregnant women compared with nonpregnant women.