Background
Plasmodium falciparum malaria has an enormous public health impact, infecting millions and killing hundreds of thousands of people each year [
1]. Drug resistance further magnifies the burden of this disease, as resistant malaria parasites have been selected by nearly every anti-malarial drug used to date. Reports of parasites with reduced susceptibility to artemisinin combination therapy (ACT) [
2,
3] underscore the importance of closely monitoring parasite drug responses and optimizing control strategies to quickly identify and prevent the spread of resistant parasites, particularly on the African continent [
4].
Malaria drug resistance monitoring involves directly measuring parasite drug responses, or indirectly measuring the prevalence of resistance-associated mutations within a parasite population.
Ex vivo drug resistance assays measure drug response in parasites taken directly from infected patients, without prior culture adaptation. These assays allow the components of combination therapies to be tested individually against parasites, and they can detect decreases in drug efficacy before resistance becomes clinically evident and widespread [
5]. Many assays have been developed to test parasite drug resistance in both laboratory and field settings [
6‐
13]. In addition, mutations in a number of parasite genetic loci have been shown to contribute to anti-malarial drug resistance, including
pfcrt and
pfmdr1, among others [
14]. Monitoring the prevalence of these mutations consistently over several years can reveal trends in allele selection within a population over time, and can extend the therapeutic life of current and future treatments [
15,
16].
The motivation for this study was to ask whether the malaria parasites circulating in Thiès, Senegal were becoming more or less resistant to anti-malarial drugs over time, and whether changes in parasite drug response could be explained by known drug resistance-associated mutations. Drug use in Senegal changed from chloroquine monotherapy to sulphadoxine, pyrimethamine and amodiaquine in 2003, and again to ACT (predominantly artesunate-amodiaquine in Thiès) in 2006 [
17]. Previous drug resistance monitoring efforts in Senegal have focused on directly testing parasite drug sensitivity [
11,
18], measuring the prevalence of resistance-associated mutations [
19,
20], or both. The aim of this study was to measure both parasite drug sensitivity and resistance mutation prevalence over time, in order to understand how parasites in Senegal may be changing in response to drug treatment.
Methods
Study population
Individuals seeking treatment for uncomplicated P. falciparum malaria at the Section de Lutte Antiparasitaire (SLAP) clinic in Thiès, Senegal, during the fall transmission seasons of 2008-2011 were tested for malaria infection by microscopy and rapid diagnostic test (RDT). Plasmodium falciparum-positive patients were eligible for screening if they met the criteria described below and the patient or legal guardian provided informed written consent or assent. Eligibility criteria were: patients older than two years, axillary temperature above 37.5°C or history of fever within the preceding 24 hours, infection with only P. falciparum, no recent anti-malarial drug use, and a haemoglobin level greater than 6 g/dL. Patients with symptoms of severe malaria were excluded and referred to the Thiès regional hospital for appropriate care. Study protocols and informed consent documents were approved by the Institutional Review Boards of the Senegal Ministry of Health IRB Committee and the Harvard School of Public Health (Senegal Protocol #16330; Harvard Protocol #P10256-127).
Among the 831 patients with uncomplicated malaria screened at the SLAP clinic between 2008 and 2011, a subset of 397 patient samples were tested for parasite drug response using the DAPI
ex vivo assay (Table
1). The subset of patients that were tested was comparable to the larger set of screened patients with respect to demographic parameters (age, gender) and clinical characteristics (temperature, haematocrit, weight) with the exception of parasitaemia.
Table 1
Clinical parameters in screened patients and the subset tested using the DAPI
ex vivo
assay
Number | 831 | 397 | - |
Gender (% male) | 66 | 64 | 0.59 |
Age (years) | 20 (15, 28) | 20 (14, 26) | 0.24 |
Weight (kg) | 55 (42, 65) | 55 (39, 65) | 0.46 |
Temperature (°C) | 38.2 (37.2, 39.7) | 38.4 (37.3, 40.0) | 0.25 |
Haematocrit (%) | 38 (32, 40) | 38 (32, 40) | 0.79 |
Parasitaemia (%) | 0.50 (0.20, 1.00) | 0.61 (0.40, 1.10) | <.0001 |
Sample collection and DAPI ex vivo testing
From each subject, 5-10 mL venous blood were collected and processed on the same day. Approximately 1 mL of blood was spotted onto Whatman FTA™ filter paper cards for subsequent DNA extraction; the remaining blood was spun at 1,500 rpm for 10 minutes, plasma and buffy coat were removed, and infected red blood cells were washed twice with unsupplemented RPMI media. Aliquots of each sample were cryopreserved in Glycerolyte 57 (Fenwal) supplemented with AB+ serum, for subsequent culture adaptation and in vitro repeat drug testing.
Parasites were drug tested using the previously described DAPI
ex vivo assay [
11]. Briefly, 180 μL of supplemented RPMI media containing parasitized erythrocytes at 2% haematocrit were distributed into 96-well plates preloaded with 20 μL serial dilutions of amodiaquine (USP 1031004), artemisinin (Sigma A5430), chloroquine (Sigma C6628) and mefloquine (Sigma M2319). Drug concentrations ranged from less than 1 nM to greater than 1 μM and each plate included 6-8 negative control wells with media only. Plates contained two wells of each drug concentration, were prepared in a single batch, and were frozen prior to use. When possible, samples with parasitaemia greater than 1% were diluted into leukocyte-free donor O
+ erythrocytes to a final plating parasitaemia of 0.4-1%.
Parasites were cultured 48-72 hours at 37°C under standard gas conditions (1% O
2, 5% CO
2, 94% N
2) before addition of 4′,6-diamidino-2-phenylindole (DAPI) solution, as described previously [
8,
11]. Data were collected by measuring relative fluorescence units (RFUs) using a Fluoroskan plate reader (Thermo Scientific; ex 358 nm, em 461 nm). 3D7 parasites were tested on each batch of drug plates to control for batch variation, and there were no consistent trends toward increased resistance among parasites tested later each season, suggesting minimal batch degradation.
DNA was extracted from 4-5 circular 6 mm punches of blood preserved on Whatman FTA™ filter paper cards using either a QIAmp DNA Blood Mini Kit (Qiagen) or a Maxwell DNA IQ Casework Sample Kit (Promega). Parasite genomic DNA was quantified by quantitative Real-Time PCR (qPCR) [
21], and clonality of infection, defined as monoclonal or polyclonal, was assessed using the 24-SNP molecular barcode [
21].
High resolution melt (HRM) technology was used to genotype a set of single nucleotide polymorphisms (SNPs) associated with reduced drug sensitivity [
19]. Mutations were detected based on changes in DNA sequence; in the text, mutations are referred to by the corresponding amino acid changes. Briefly, 0.01 ng of parasite template, as quantified by qPCR [
21], was used for each 5 μL reaction, which included 2.5× LightScanner Master Mix with LCGreen Plus dsDNA dye (BioFire Diagnostics, Inc.). HRM analysis and genotype determination was performed on a LightScanner-384 (BioFire Diagnostics, Inc.). The HRM method can determine genotypes from as little as 10 pg of parasite DNA and can detect mutant alleles present at less than 1% [
19].
Culture adaptation and in vitro drug testing
To assess whether
ex vivo drug responses were reproducible
in vitro, 16 parasite isolates derived from monoclonal infections collected in 2009 were culture adapted and re-tested
in vitro. Culturing was conducted under standard conditions [
22] with gentle shaking at 55 rpm. Parasites were
in vitro drug tested against a panel of known anti-malarials using a standard hypoxanthine incorporation assay [
7], or a SYBR Green I-based drug assay [
23] with modifications for 384-well format.
Calculation of IC50 values and data exclusion
Fluorescence data from drug assays were analysed using GraphPad Prism (San Diego, CA) through a four-parameter, log-logistic nonlinear regression of fluorescence intensity versus log10-transformed drug concentrations. To include control wells with no drug in the analysis, 1 nM was added to each concentration value. Dose-response curves were visually inspected for fit of the sigmoidal dose-response model. Among 397 patient samples tested using the DAPI ex vivo assay, 25 samples were considered assay failures due to no parasite growth or assay contamination and were excluded from further analysis. An additional two patient samples with a plating parasitaemia below 0.1% and 30 samples with a plating parasitaemia above 1.5% were excluded, because there was no clear association between plating parasitaemia and fluorescence intensity in the no-drug wells, perhaps due to insufficient growth or saturation. This left 340 patient samples from which parasite response to at least one anti-malarial drug was determined.
Drug curves that did not exhibit the standard sigmoidal dose-response shape were classified as either fitting an exponential or linear decay model, and had their IC50 values estimated through these alternative models, or were excluded. When IC50 values from technical replicates could not be estimated due to a single outlier point, this point was excluded.
Data and statistical analysis
Dynamic range of the DAPI
ex vivo assay was assessed by calculating the signal-to-noise ratio (SNR) and Z’-factor of each assay. SNR was measured by dividing fluorescence signal (RFUs) from no-drug wells by fluorescence signal from maximum drug wells. The median signal-to-noise ratio (SNR) among all assays was 3:1 (Interquartile Range = 2:1, 5:1). Z’-factor was calculated using the following equation: Z’ = 1- [(3 standard deviations of positive controls + 3 standard deviations of negative controls)/absolute difference between negative and positive controls] [
24]. The median Z’-factor among all assays was 0.61; Z’-factors greater than 0 are considered acceptable, and Z’-factors greater than 0.5 are considered excellent [
24]. Reliability of the DAPI
ex vivo assay was measured by evaluating agreement between technical replicates in the untransformed scale using the intraclass correlation coefficient (ICC) for agreement. Only sigmoidal curves were analysed, to avoid biases due to lack of fitness of different IC
50 curve-fitting models. For all drugs, mean differences in IC
50 values between replicates were approximately zero, as expected. Except for a few outliers (fewer than five points for each drug), differences between replicates were small compared with the IC
50 range of each drug.
Statistical analyses were performed in GraphPad Prism (v5.0d, San Diego, CA) and R-2.11.1. IC50 values measured ex vivo were compared to in vitro IC50 values from culture-adapted parasites by calculating the intraclass correlation coefficient (ICC) for consistency (R package irr), and by linear regression. To monitor population drug sensitivity, IC50 variations over time were measured through linear regression with log10-transformed IC50 values. Primary analysis focused on non-linear trends (using indicator variables for years) but results were confirmed by assessing linear trends. Multiple regression models were used to measure whether IC50 values changed significantly over time after accounting for the effect of potential confounders (clonality, haematocrit, parasitaemia, age, and temperature). Since parasites with reduced drug sensitivity may first arise in subpopulations that exhibit larger IC50 values, the 90th percentile among all IC50 values is reported for each year. Changes in the prevalence of drug resistance markers over time were measured by Fisher-Hamilton exact test, and 95% confidence intervals for marker prevalence are based on the logit (R package binom). Finally, associations between IC50 values and the occurrence of drug resistance-associated mutations in pfcrt and pfmdr1 were assessed through the Wilcoxon rank-sum test.
Discussion
Ex vivo assays are an important tool for malaria drug resistance monitoring in direct patient samples. These assays complement in vivo studies by allowing researchers to test parasite responses to different drugs individually and in the absence of patient factors that might introduce noise or confound results. Importantly, ex vivo monitoring of malaria parasite drug responses can provide an early warning of decreased parasite sensitivity before parasites become highly resistant and cause infected patients to fail drug treatment.
The DAPI
ex vivo assay performed well, with excellent agreement between technical replicates, good dynamic range, and very good correlation between drug responses measured
ex vivo with those measured
in vitro. Furthermore, the IC
50 values observed in Thiès, Senegal between 2008 and 2011 were comparable to other
ex vivo studies of
P. falciparum drug response [
10,
18,
29]. Because parasite drug responses were measured over a four-year time span, trends in drug response in this population over time could also be assessed. The trends observed in parasite responses to amodiaquine and artemisinin suggest that malaria parasites in Thiès are becoming more tolerant to these compounds. Data from future years of
ex vivo monitoring will be critical in determining whether these trends continue.
The trends observed in the prevalence of resistance-associated mutations in
pfcrt and
pfmdr1 suggest that anti-malarial drug use is selecting for resistance-associated alleles within this population. In contrast to other studies [
11,
20,
25], resistance-associated mutations within
pfcrt remained prevalent within this population, and even appeared to increase in prevalence between 2010 and 2011. This suggests either that compensatory mutations have restored the fitness of resistant parasites, and/or that anti-malarial drug use is maintaining these mutations within the population. The continuous distribution of chloroquine IC
50 values observed in 2008 and 2011 further suggests that additional mutations affecting parasite drug response exist in this population. The finding that parasites with mutations in
pfcrt have higher amodiaquine IC
50 values is in agreement with previous studies of laboratory parasite lines [
26], and malaria-infected patients [
28]. Additionally, amodiaquine has been administered to malaria-infected patients in Senegal since 2003 [
17]. Because increased amodiaquine IC
50 values were associated with the typed
pfcrt mutations, it seems possible that use of amodiaquine is preserving these mutations within the population.
The observed trends in resistance mutations within
pfmdr1 suggest that artemisinin compounds are selecting for a combination of wild-type and mutant alleles within this gene. The N86 and 184F alleles have been previously associated with
in vivo selection by ACT [
30,
31], and two recent studies of the prevalence of drug resistance markers in Dakar, Senegal also found a high prevalence of the Y184F mutation [
32,
33]. Furthermore, Y184F has been found to be under selection among parasite populations in Cambodia [
34], where artemisinin resistance, defined as delayed parasite clearance, has been described. While the artemisinin resistance phenotype of delayed
in vivo parasite clearance does not appear to correlate well with
ex vivo or standard
in vitro assays [
2], artemisinin resistance might occur through different mechanisms in Africa as compared to southeast Asia. Furthermore, as parasites become increasingly artemisinin resistant
in vivo, they may become amenable to monitoring with
ex vivo assays such as this one. The disappearance of the N86Y and N1042D mutations, coupled with the rapid rise of the Y184F mutation, suggest that selective pressure is acting on
pfmdr1, eliminating some mutations while driving others to high prevalence within this population. Because artemisinin response was associated with all three of the typed
pfmdr1 mutations, it appears that the artemisinin derivatives used in ACT might be the selective force driving the Y184F mutation to high prevalence, while simultaneously selecting for the wild-type alleles at positions 86 and 1042.
These findings are consistent with the hypothesis that amodiaquine use in Thiès, Senegal has selected for chloroquine resistance-associated mutations within
pfcrt, while artemisinin compounds have selected for a particular combination of wild-type and mutant alleles within
pfmdr1. In both cases, alleles that make parasites better able to withstand drug pressure are likely selected. Other African countries that have used artesunate-amodiaquine have also documented sustained high prevalence of chloroquine resistance-associated mutations within
pfcrt[
35,
36]. Conversely, countries deploying ACT that does not include amodiaquine have seen a return to chloroquine-sensitivity after chloroquine was removed from the treatment arsenal [
25,
37], presumably due to the fitness costs of resistance-associated mutations. Other African countries have also documented recent increases in the prevalence of the
pfmdr1 N86 and 184F alleles [
38,
39], though this study marks the highest recorded prevalence to date of the Y184F mutation on the African continent.
Acknowledgements
We gratefully acknowledge members of the sample collection team at the SLAP clinic in Thiès, including Younous Diedhiou, Lamine Ndiaye, Amadou Mactar Mbaye, Ngayo Sy and Moussa Dieng Sarr. We also thank Meaghan Galligan, Justin Becker, Kate Fernandez, and Vishal Patel for technical assistance and thoughtful discussions. This work was supported by the National Institutes of Health [R01 AI075080-01A1 and D43 TW001503], and by the ExxonMobil Foundation.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
DVT, PDS, ON, SM, SKV, DFW and DN designed the experiments. DVT, BD, RFD, PDS, AKL, MN, AKB, YDN, EJH and SKV carried out the experiments and collected data. DVT, RFD, CV, PDS and SKV analysed the data. DVT, CV, SKV, DFW and DN wrote the manuscript. All authors read and approved the final manuscript.