In Vitro Sensitivity of Plasmodium falciparum from China-Myanmar Border Area to Major ACT Drugs and Polymorphisms in Potential Target Genes

Drug resistance has always been one of the most important impediments to global malaria control. Artemisinin resistance has recently been confirmed in the Greater Mekong Subregion (GMS) and efforts for surveillance and containment are intensified. To determine potential mechanisms of artemisinin resistance and monitor the emergence and spread of resistance in other regions of the GMS, we investigated the in vitro sensitivity of 51 culture-adapted parasite isolates from the China-Myanmar border area to four drugs. The 50% inhibitory concentrations (IC50s) of dihydroartemisinin, mefloquine and lumefantrine were clustered in a relatively narrow, 3- to 6-fold range, whereas the IC50 range of artesunate was 12-fold. We assessed the polymorphisms of candidate resistance genes pfcrt, pfmdr1, pfATP6, pfmdr6 and pfMT (a putative metabolite/drug transporter). The K76T mutation in pfcrt reached fixation in the study parasite population, whereas point mutations in pfmdr1 and pfATP6 had low levels of prevalence. In addition, pfmdr1 gene amplification was not detected. None of the mutations in pfmdr1 and pfATP6 was associated significantly with in vitro sensitivity to artemisinin derivatives. The ABC transporter gene pfmdr6 harbored two point mutations, two indels, and number variations in three simple repeats. Only the length variation in a microsatellite repeat appeared associated with altered sensitivity to dihydroartemisinin. The PfMT gene had two point mutations and one codon deletion; the I30N and N496– both reached high levels of prevalence. However, none of the SNPs or haplotypes in PfMT were correlated significantly with resistance to the four tested drugs. Compared with other parasite populations from the GMS, our studies revealed drastically different genotype and drug sensitivity profiles in parasites from the China-Myanmar border area, where artemisinins have been deployed extensively for over 30 years.


Introduction
The development and spread of multidrug resistant (MDR) Plasmodium falciparum has led to the adoption of artemisinin-based combination therapies (ACTs) as the first-line treatment for falciparum malaria in most malaria-endemic countries of the world [1]. Artemisinin and its derivatives are by far the most potent antimalarial drugs [2], and at present, our last line of defense against multidrug resistant parasites. Therefore, the recently confirmed emergence of artemisinin resistance in western Cambodia is a major threat for current initiatives to control and eliminate malaria [3][4][5]. Because this exact same area has been the origin of both chloroquine (CQ) and sulfadoxine-pyrimethamine resistance, both of which have subsequently spread to Africa [6,7], the consequence of a similar spread of artemisinin resistance will be catastrophic. While limited evidence suggests that artemisinin resistance has not yet spread to other areas [8], the World Health Organization (WHO) is coordinating a large-scale elimination campaign in this region aiming to contain artemisinin resistance [9,10]. Apparently, containment efforts require better resistance surveillance [11], but this is hampered due to the lack of convenient molecular surveillance tools for detecting artemisinin resistance. At this moment, the most reliable way of detecting artemisinin resistance is through rigorous clinical efficacy studies, which are expensive and time-consuming.
The mode of action of artemisinins in malaria parasites is still a debated topic and the molecular basis of reduced artemisinin susceptibility is unclear [12][13][14]. To date, a few genes have been postulated to be associated with artemisinin resistance. The P. falciparum multiple drug resistance 1 (pfmdr1) gene has received the most attention, beacuse several mutations (N86Y, Y184F, S1034C, N1042D, and D1246Y) occurring in PfMDR1 from field isolates are associated with altered sensitivity to multiple structurally unrelated antimalarials such as CQ, mefloquine (MQ), quinine (QN), halofantrine (HF), and artemisinins [15][16][17][18][19]. In addition, pfmdr1 amplification is a key determinant for both in vivo and in vitro resistance to MQ and HF [19][20][21][22][23]. Increased pfmdr1 copy number, which is more prevalent in west Cambodia, is associated with increased risk of therapy failures of artesunate (AS)-MQ, the major ACT deployed in Thailand and Cambodia [24][25][26][27][28]. The sarco/ endoplasmic reticulum calcium-dependent ATPase (SERCA) homologue PfATP6 has been considered as a specific target of artemisinins, since artemisinin derivatives inhibit this enzyme expressed in Xenopus oocytes [29] and this inhibition was abolished by the introduction of the L263E mutation in the predicted transmembrane domain 3 of PfATP6 [30]. Another mutation (S769N) has been linked to artemether resistance in P. falciparum field isolates from French Guiana [31]. However, the L263E mutation has not been found in field isolates from most malaria endemic areas, and S769N is very rare [32][33][34][35][36][37][38][39][40]. Though multiple new single nucleotide polymorphisms (SNPs) have been detected in pfatp6, their associations with artemisinin resistance have not been established. Mutations in several other genes also have been suggested to be responsible for artemisinin resistance. Mutations in the UBP1 gene encoding a deubiquitination enzyme have been identified to confer artemisinin resistance in the rodent malaria parasite P. chabaudi [41]. However, the equivalent mutations have not been found in P. falciparum field isolates from Cambodia and Thailand [42]. Recently, using a genome-wide association approach, Mu et al. detected signs of positive selection at several putative transporter genes in parasite populations including one coding the ABC transporter pfmdr6 and another gene coding the metabolite/drug transporter pfMT [43,44]. So far, none of the candidate genes have been conclusively shown to be responsible for artemisinin resistance. Therefore, further research is needed to identify the causal mutations for artemisinin resistance.
Malaria is still a serious public health problem in the Greater Mekong Subregion (GMS), which includes Cambodia, Laos, Thailand, Vietnam, Myanmar and China [45]. Within this region, malaria transmission is particularly intense along international border areas. The GMS is historically a hotspot of antimalarial drug resistance, and resistance management remains a high priority for malaria control. The China-Myanmar border area has the longest history of artemisinin monotherapy; it has been used for over three decades and in vitro studies have detected reduced susceptibility to artemisinins [46]. Therefore, close surveillance of artemisinin resistance in this area is necessary to detect and deter resistance development. Here we report an assessment of in vitro sensitivity of clinical P. falciparum isolates from the China-Myanmar border area to four antimalarial drugs, AS, dihydroartemisinin (DHA), lumefantrine (LMF) and MQ. Using a candidate gene approach, we want to determine whether polymorphisms in these genes are associated with decreased drug sensitivity.

Materials and Methods
Parasite Samples and in vitro Culture P. falciparum clinical isolates were collected in 2007-2009 from symptomatic patients presenting with uncomplicated P. falciparum infections attending a malaria clinic in Laiza township near the China-Myanmar border. Malaria infections were diagnosed by microscopic examination of Giemsa-stained thick and thin blood films. To confirm monoclonal infections, parasites were genotyped at three polymorphic genes, merozoite surface protein 1 (msp1), msp2 and glutamate-rich protein (glurp), using previously described methods [47][48][49]. A total of 21, 9, and 21 parasite isolates were adapted to continuous in vitro culture for the year 2007, 2008, and 2009, respectively [50]. The human subject protocol for this study was approved by the Institutional Review Board of Kunming Medical University. Parasite routine culture was maintained in type O + human red blood cells (RBCs) in complete medium supplemented with 6% AB human serum under an atmosphere of 90% N 2 /5% O 2 /5% CO 2 [51].

In vitro Drug Sensitivity Assay
A SYBR Green I-based fluorescence assay was used to measure parasite sensitivity to four antimalarial drugs: AS, DHA, MQ, and LMF [52]. MQ was purchased from Sigma (St Louis, MO, USA), while the other three drugs were obtained from Kunming Pharmaceutical Co. (Kunming, Yunnan, China). The stock solution of AS (0.260 mM), DHA (0.352 mM), MQ (2.56 mM), and LMF (7.56 mM) were prepared in ethanol and 2-fold dilutions were made in complete medium. Cultures were synchronized by two rounds of sorbitol treatment, and late ring or early trophozoite stage parasites were assayed for drug sensitivity in 96-well microtiter plates at 5% hematocrit and 0.3% parasitemia. For drug assays, 90 ml of parasite culture were seeded into each culture well, and 10 ml of diluted drugs in complete medium solutions were dispensed into each well to obtain a desired final concentration. The plates were incubated at 37uC in a CO 2 incubator for 72 h. The plates were then frozen at 280uC, thawed, and mixed with 100 ml of lysis buffer. The plates were incubated in the dark for about 1 h and fluorescence data were acquired using the Fluoroskan Ascent FL microplate fluorometer (Thermo Scientific, Waltham, MA). For each parasite isolate and drug concentration, the assay was performed in three biological replicates and each with three technical replicates. To reduce the variations between plates, the standard laboratory clone 3D7 was included as a reference. The in vitro drug response data were entered into the SPSS data editor, and the geometric mean of the half-maximal inhibitory concentration (IC 50 ) was calculated for all isolates using a regression-probit analysis. Standard deviation was calculated using the mean values of the three biological replicates of each parasite isolate.

Molecular Analysis of Candidate Genes
Parasite genomic DNA was isolated from cultured parasites using a proteinase K digestion and phenol/chloroform extraction procedure [53]. By using a PCR and sequencing approach, we assessed SNPs in pfcrt, pfmdr1, pfatp6, a putative ABC transporter PF13_0271 (pfmdr6), and a putative metabolite/drug transporter PF14_0260 (pfMT). Two pfcrt fragments covering codons 76 and 220 and two pfmdr1 fragments including codons 86, 184, 1034, 1042 and 1246 were amplified as described previously [54]. Two regions of pfatp6 gene (28-3065 and 3207-3952 bp) were amplified using primers and conditions from previously published work [36,55] (see Table S1). The coding sequences of pfmdr6 and pfMT genes, shown to be under strong positive selection [43], were amplified using primers shown in Table S1. To minimize errors introduced in the sequences during PCR amplification, we used Advantage HD DNA Polymerase Mix (Clontech, Mountain View, CA), which has efficient 39 R 59 exonuclease activity for high fidelity. PCR products were purified using a PCR Purification kit (Qiagen, Valencia, CA) and used for direct sequencing. Overlapping sequences were obtained by using sequencing primers. For singletons, the sequence was confirmed from two independent PCR reactions from the same DNA templates. For a highly ATrich region in the pfMT gene, the primers MT_873 and mt_1433R were used to amplify this fragment and the PCR products were cloned into the pGEM-T Easy Vectors system (Promega, Madison, WI). Plasmid DNA was extracted by Zyppy Plasmid Miniprep Kit (Zymo research, Orange, CA). For accuracy, at least two clones were sequenced for each sample. Alignment of DNA sequences were performed using the BioEdit program with the 3D7 sequence as the reference. All new sequences were deposited in GenBank under the accession numbers (JN983240-JN983290 for pfatp6, JN983291-JN983341for pfmdr6, and JN983342-JN983392 for pfMT).
Pfmdr1 copy number in field parasite isolates was determined by real-time PCR using FastStart Universal SYBR Green Master Mix on an ABI 7300 real-time PCR system (Applied Biosystems). DNA from the reference strain 3D7, which has a single copy pfmdr1 gene, was included for calibration. The single copy b-tubulin gene served as an internal control for estimating the copy number of pfmdr1. Each reaction was performed in a final reaction volume of 20 ml. Copy number of pfmdr1 gene was calculated using the 2 2DDCt method as described earlier [56]. The efficiency of each PCR (pfmdr1 and b-tubulin) was determined using a scale dilution of DNAs from the laboratory lines 3D7 (containing 1 copy of pfmdr1) and Dd2 (containing 3-4 copies of pfmdr1). Determination of the copy number was done by comparison of the ratio of pfmdr1/btubulin taking into account the efficiency of each PCR. Each sample was analyzed in triplicate. Assays were repeated if one of the following two results was obtained: C t value .35, or copy number as 1.3-1.6 or ,0.7.

Statistical Analysis
Spearman's correlation coefficients were used to investigate the degree and significance of the relation between the IC 50 values of the four drugs. It was chosen instead of Pearson's in order to provide robust correlation estimates despite potential outliers in the data. We then used t-tests to look for significant differences in mean IC 50 values between these field isolates and 3D7, and between the years. Finally, we used multiple t-tests and a generalized linear model to investigate potential, significant differences in mean IC 50 values between haplotypes and point mutations for each gene and for each drug. The Bonferroni correction was used to control for the increased probability of false positives in multiple comparisons. All statistical analyses were done using SAS version 9.2 (http://www.sas.com/) and R version 2.14.2 (http://cran.rproject.org/).

In vitro Drug Assays
To determine whether artemisinin resistance has emerged at the China-Myanmar border area, we assessed in vitro sensitivity of 51 culture-adapted P. falciparum strains originating from this region to two artemisinin derivatives AS and DHA and two aminoalcoholic drugs MQ and LMF. These parasites were collected in 2007-2009 and genotyped to confirm monoclonal infections [54]. The IC 50 values of individual parasite isolates are shown in Table S2. Parasite isolates were generally sensitive to AS with a mean IC 50 of 5.8 nM, but the range (1.4-16.5 nM) was relatively wide with an almost 12-fold difference between the most and least susceptible isolates ( Table 1). The mean IC 50 s of the samples to DHA ranged from 9.6 to 40.3 nM with a mean IC 50 of 23.0 nM. The mean IC 50 for MQ was 50.5 nM, ranging from 16.2 to 96.2 nM, with a six-fold difference in IC 50 between the most and the least susceptible isolates. The parasites had a mean IC 50 of 5.9 nM for LMF with a fairly narrow range of 3.0-11.5 nM. Compared with the mean values of the control laboratory clone 3D7, only MQ sensitivity showed a significant difference after Bonferroni correction (P,0.05). When the IC 50 data were stratified by years, slight variations were observed between the years ( Figure 1). However, none of the comparisons were statistically significant (P.0.05, t-test). Also noteworthy is the detection of several outlier isolates (greater than 2 standard deviation), which showed much higher IC 50 values to all drugs.
Correlations between the IC 50 s of the four antimalarial drugs were evaluated using the Spearman's test (Table 2). There was a highly significant correlation between responses to the two artemisinin derivatives (P,0.0001). When all samples were considered, we did not detect significant correlation between the two aminoalcohol drugs MQ and LMF (P.0.05). However, when only the samples collected in 2009 were compared, a modest correlation was found between MQ and LMF (P,0.05) (data not shown).

Polymorphisms in Candidate Target Genes
Pfmdr1. We genotyped pfmdr1SNPs by PCR and sequencing in the 51 field isolates, focusing on the known amino acid substitutions at codons 86 (N/Y), 184 (Y/F), 1034 (S/C), 1042 (N/ D), and 1246 (D/Y). Wild-type haplotype was prevalent and accounted for ,65%. Except codon 184, mutations at other codons were rare (Table 3). Specifically, mutations at codons 1034 and 1246 were not observed. The N86Y mutation, which is associated with CQ resistance and increased sensitivity to MQ and artemisinin [57], was detected only in one (2%) isolate of the samples. The N1042D mutation associated with QN resistance was also rare and found only in 4 (7.8%) isolates of the samples [18]. The C-terminal mutation haplotype 1034C/1042D/1246Y, which is prevalent in South America and enhances parasite susceptibility to MQ and artemisinin, was absent in our samples (Table 3). We did not detect pfmdr1 amplification in any samples; the 51 isolates had a mean pfmdr1 copy number value of 1.0260.14 ( Figure S1).
Pfcrt. The key determinant of CQ resistance K76T was ubiquitously present in all samples. Further, the A220S mutation almost reached fixation and was found in 50/51 parasite isolates (data not shown).
PfATP6. We obtained complete sequences of the PfATP6 coding region in 51 samples and compared them with that from 3D7. Mutations were rare and sporadic with a total of five identified (Table 3). Three mutations have been described previously (I89T, A438D and N465S), and two were novel mutations (I226V and E710K). Except for the I89T and A438D mutations, which occurred in 5 (9.8%) and 3 (5.9%) isolates of the samples, respectively, the remaining mutations were only detected once in the samples. The L263E and S769N mutations, which have been proposed to confer artemisinin resistance [30,31], were not detected in our samples. A total of seven haplotypes were detected; 43 (84.3%) were wild-type, whereas six and two samples contained single and two mutations, respectively.  Pfmdr6. Analysis of the coding region of the pfmdr6 gene from 51 parasite samples revealed mutations at residues 175 and 353 ( Table 3). The Y175S mutation was found in 9 (17.7%) isolates, whereas L353W mutation was only found in one sample. In addition, this gene harbored two deletions. The tri-peptide SIN corresponding to positions 735-737 in 3D7 was deleted in four isolates, while the N residue at residue 823 was deleted in 45 (88.2%) isolates. Other polymorphisms all occurred in repeat sequences. For the microsatellite sequence corresponding to positions 103-110 in 3D7 (referred to as repeat R1), the number of the amino acid N ranged from 6 (in 3D7) to 12, with 8 as the predominant one occurring in 39 (76.5%) isolates. 3D7 had two NI repeats at positions 267-270 (referred to as repeat R2). Whereas 50 (98.0%) isolates had two NI repeats, only one field isolate had three NI repeats. The third repeat sequence, referred to as repeat R3, contains NIN repeats at positions corresponding to 717-734 in 3D7. Whereas 46 (90.2%) isolates had 6 NIN repeats, 4 (7.8%) and 1 (2.0%) isolates had 5 and 4 NIN repeats, respectively. These polymorphisms gave rise to a total of 13 haplotypes (Table 3), but none of the field isolates had the same haplotype as 3D7. The most prevalent haplotype, which differs from 3D7 by 8 N residues in microsatellite repeats R1 and a codon 819 deletion, occurred in 25 (49.0%) parasite isolates.
PfMT. Sequencing of the pfMT gene detected a total of two amino acid substitutions at positions 30 and 286. The I30N mutation was found in all but one parasite isolate, whereas the S286C mutation was rare and found only in one isolate. In addition, the codon for amino acid N496 was deleted in the majority (94.1%) of the isolates. A total of four amino acid haplotypes were recognized with one (N30/286S/496-) accounting for 94.1% parasite isolates (Table 3). In comparison, all other haplotypes only occurred once in the samples.

Correlation between Polymorphisms and in vitro Drug Sensitivities
Since the pfcrt K76T mutation was fixed and the A220S mutation almost reached fixation (50/51), and no pfmdr1amplification was detected in the studied parasite populations, associations could not be evaluated between these markers and drug responses. Thus, we analyzed the potential associations between point mutations in pfmdr1, pfATP6, pfmdr6, and pfMT genes and drug responses. None of the mutations in the pfmdr1 and pfATP6 genes had a significant association with altered responses to AS and DHA. Yet, the N1042D mutation in pfmdr1was found to be associated with increased susceptibility to MQ ( Figure 2, Table  S3). SNPs in pfmdr6 and pfMT genes that caused amino acid substitutions were rare. None of them was associated with significantly altered sensitivity to the four drugs tested (P.0.05). Neither did the deletion mutations (S7352/I7362/N737-and N823-in pfmdr6, as well as N496-in pfMT) affect the sensitivity to the four drugs. There are three simple amino acid repeats in pfmdr6. Whereas no significant differences in drug sensitivity were detected between the parasites carrying different numbers of repeats in repeats R2 and R3, parasites carrying 9 repeats in repeat R1 were associated with significantly increased resistance to DHA (P,0.05) ( Figure 2, Table S3). The difference between parasites carrying 8 and 9 R1 repeats remained significant after Bonferroni correction (Figure 2). In addition, 7 repeats in R1 were also significantly correlated with reduced sensitivity to LMF (P,0.05) (Table S3).
To determine whether certain haplotypes of pfmdr1, pfATP6, pfmdr6, and pfMT genes are associated with altered drug responses, we compared the drug responses of the mutant haplotypes with those of the wild-type isolates. Comparison of pfmdr1 haplotypes with $3 (5.9%) prevalence showed that parasites carrying the pfmdr1 1042D mutation showed increased sensitivity to MQ, whereas responses to other three drugs were not significantly different between the pfmdr1 haplotypes ( Figure 3). All mutant haplotypes of pfATP6 were below 4% in prevalence, and only 2 (3.9%) parasite isolates with the 89T/438D double mutations had significantly higher IC 50 values to AS (data not shown). For the pfmdr6 gene, we found a significant association for the haplotype carrying 9 N residues in microsatellite repeat R1 (9Y2L6+-) with significantly higher DHA IC 50 values when compared to both 3D7 and the haplotype 8Y2L6+-. These findings remained significant after Bonferroni correction ( Figure 3). Furthermore, haplotype 8Y2L6+-showed significantly lower IC 50 s to LMF when compared to the wild-type 3D7. Interestingly, all of the haplotypes showed increased sensitivity to LMF when compared to 3D7. In comparison, none of the haplotypes in pfMT were associated with altered drug sensitivity (data not shown).

Discussion
The GMS in Southeast Asia has been a breeding ground of antimalarial drug resistance, where MDR parasites have emerged. Parasites in this region appear to have the ARMD (accelerated resistance to multiple drugs) genetic background, favoring development of resistance to new antimalarial drugs [58,59]. Further reflecting such a tendency, malaria parasites from this region have recently been found to exhibit reduced susceptibility or low-grade resistance to artemisinin derivatives [3,4]. This urgent situation calls for closer surveillance, containment and heightened research on resistance mechanisms. In some areas of the GMS such as the China-Myanmar border area, artemisinins have been used for over three decades, mostly as monotherapies prior to 2005. Earlier in vitro assays have already detected a trend of declining sensitivity to AS [46]. Surveillance at the Thai-Myanmar border also detected reduced sensitivity to artemisinins, MQ, and QN [60][61][62]. Therefore, understanding whether artemisinin-resistant parasites have spread to neighboring regions or emerged elsewhere in the GMS is essential for coordinating containment efforts. Here, we evaluated this situation at the China-Myanmar border area and assessed the in vitro sensitivity of recently collected parasite isolates to artemisinins and two amino alcoholic drugs. Our data showed that although these parasites were generally sensitive to AS, there were eight parasite isolates showing mean IC 50 values above 8 nM. In addition, in vitro sensitivity to DHA also had a more than 4-fold variation. Spearman's correlation test detected significant correlations in sensitivity between the two artemisinin derivatives, suggesting that these drugs have shared cellular targets. Despite the fact that no careful clinical studies have been conducted in this region to rule out the lack of clinical resistance to artemisinins, our data suggest that reduced susceptibility to the artemisinins might have already emerged in some isolates (outliers) in the study parasite population. Therefore, close monitoring of resistance development at such sentinel sites is necessary. We have also included two aminoalcoholic drugs in the in vitro assays, since responses to MQ and LMF are often correlated with those to artemisinins [18,61]. At the China-Myanmar border area, MQ has never been deployed, but in vitro assays suggested the existence of reduced sensitivity to MQ with 37 parasite isolates showing higher mean IC 50 values than 3D7. Another perplexing finding is that the parasite isolates tested were highly sensitive to LMF, which is in sharp contrast to the parasites from Thailand that displayed more than 20-fold higher IC 50 s [63], although LMF has not been used in Thailand. The decreased sensitivity to LMF in Thailand could be related to the mass deployment of MQ in treating falciparum malaria, since sensitivities to MQ and LMF are correlated. Similarly, our samples collected in 2009 also showed significant correlation in IC 50 values between these two drugs. However, none of these two aminoalcoholic drugs showed positive correlation in IC 50 with the two artemisinin drugs. These are different from the in vitro drug studies carried out in other areas of the GMS [61,[64][65][66]. The reason for such discrepancies was not clear, but may be related to the divergent antimalarial drug policies in these areas.
While the mode of action of artemisinins is not completely understood, several cellular targets have been proposed. Among them, PfATP6 has been suggested as a prime candidate target of artemisinins [29,30], and studies with field parasite isolates from French Guiana seemed to support this suggestion [31]. However, since these reports, results from subsequent studies appear to contradict the hypothesis of PfATP6 as a specific target of artemisinins. First, artemisinin treatment of P. falciparum did not alter the morphology of endoplasmic reticulum, where PfATP6 is supposed to be localized [67]. Second, purified recombinant PfATP6 did not seem to bind artemisinins, nor was it inhibited by artemisinins [68,69]. Third, allelic exchange experiments aiming to determine whether the PfATP6 L283E mutation is linked to a drastic reduction of artemisinin sensitivity only provided tangential evidence supporting this suggestion [70]. In addition, our recent allelic exchange experiment did not detect differences in sensitivity to artemisinin derivatives between parasites carrying either S769 or 769N [71]. Further, an African parasite isolate carrying the S769N mutation did not show increased resistance to artemisinins [34]. In the Thai-Cambodian border region, where resistance to artemisinins has recently emerged, these two mutations have not been detected either [42]. We and others further confirmed the lack of these mutations in parasites from the China-Myanmar border area, where artemisinins have been deployed extensively for over 30 years [38]. In fact, most parasites from our study carried the wild-type PfATP6 gene, and no significant association could be established between the rare mutations in pfATP6 and artemisinin resistance. Collectively, these data potentially suggest that PfATP6 might not play a major and direct role in artemisinin resistance.
The malaria parasite encodes many transporters and some of them such as pfcrt, pfmdr1 and pfmrp have been strongly connected with antimalarial drug resistance [72][73][74]. Parasite resistance to multiple drugs has been associated with SNPs and increased copy number of the ABC transporter family member pfmdr1. Pfmdr1 amplification is correlated with resistance to MQ, QN, HF, and artemisinins [20,75]. Laboratory experiments confirmed that an increase of pfmdr1 copy number can be induced by stepwise MQ selection [76], whereas knockdown of pfmdr1 expression results in increased susceptibility to MQ, QN, and artemisinin [77]. Moreover, recent laboratory selection studies strongly implicated pfmdr1 gene amplification as a major determinant of induced artemisinin resistance [78,79]. Pfmdr1 also harbors many point mutations: the 86Y mutation which modifies resistance to CQ is associated with increased sensitivity to MQ and artemisinin [16], whereas mutations at the C-terminal end, which are prevalent in South America, contribute to QN resistance but enhance MQ and artemisinin sensitivity [18]. Consistently, administration of artemether-LMF has selected for increased frequency of the N86 allele [32,80]. In Thailand and Cambodia, pfmdr1 gene amplification has become increasingly prevalent in field parasite populations and was responsible for resistance to MQ and declining efficacy of ACTs [19,22,23,26,28]. In this case, pfmdr1 amplification has been strongly tied to the extensive deployment of MQ in this region. At the China-Myanmar border, however, pfmdr1 gene amplification has not been observed, consistent with the fact that MQ has never been used in this region. Overall, the frequencies of pfmdr1 N86Y and C-terminal mutations which confer increased sensitivity to  50 values are shown as mean + standard deviation. * indicates significant difference (P,0.05) in sensitivity between the haplotypes after Bonferroni correction. # indicates that this pfmdr6 haplotype was significantly more sensitive to LMF than 3D7. doi:10.1371/journal.pone.0030927.g003 artemisinins were very low in the GMS [61,64]. Our results from the China-Myanmar border further verified these findings [81]. Yet, despite the implied significance of pfmdr1 in artemisinin resistance, in our study, pfmdr1 was not associated with the reduced artemisinin susceptibility recently unveiled in western Cambodia [42]. The uncommon pfmdr1 polymorphisms from these parasites offer opportunities for investigations of mechanisms other than pfmdr1.
The clear link between polymorphisms of transporter genes and antimalarial drug resistance has prompted us to evaluate other genes encoding putative transporters. We have chosen pfmdr6, another ABC transporter family member, and pfMT, a putative metabolite/drug transporter, since these two genes have been shown to be under recent positive selection from a genome-wide association study [43]. Both genes were found to harbor small numbers of nonsynonymous SNPs, whereas pfmdr6 also have polymorphisms in three simple repeats. Compared with 3D7, the Y175S mutation and the 823 codon deletion had reached 17.7 and 88.2% of prevalence, respectively. However, none of the point mutations were associated with altered drug sensitivity. Yet, the microsatellite repeat R1, which has seven different alleles (ranging from 6 to 12 repeats), appeared to affect the in vitro sensitivity to DHA. Specifically, 9 Ns in R1 was associated with reduced sensitivity to DHA. This is reminiscent of the effect of the microsatellite repeats in the pfNHE1 gene on in vitro sensitivity to QN [50,82]. Of the 13 haplotypes detected within this gene, 8 were rare haplotypes, each being represented by one or two isolates. Again, the haplotype with 9 Ns in R1 was associated with significantly higher IC 50 value to DHA than others. Compared to 3D7, two mutations in pfMT (I30N and N496-) were predominant in the samples analyzed, which together gave rise to a predominant haplotype. Correlation studies did not detect any association of the mutations or haplotypes with significant altered drug sensitivity. However, we would like to exercise caution in this finding since previous in vitro studies of parasite resistance to AS have shown poor correlations with clinical in vivo studies. Furthermore, significant associations between these rare haplotypes and drug responses were at most based on a small number of parasite isolates carrying such mutations. Therefore, such an association may not hold if more samples carrying these mutations were analyzed.
Despite extensive efforts aimed at elucidating the mechanism of artemisinin resistance in P. falciparum using the candidate gene approach, most of the mutations encountered in candidate genes did not account for reduced sensitivity to artemisinins in parasites from clinical studies in western Cambodia [42]. Our in vitro studies using parasites from the China-Myanmar border area seem to convey a similar conclusion. However, analysis of parasites originating from the Thai-Myanmar border region implied that pfmdr1 copy number and a new point mutation (F1226Y) are significantly associated with in vitro response to artemisinins [65]. This discrepancy may reflect the intrinsic difference between in vitro assay and in vivo clinical studies. In addition, it may also be due to divergent genetic backgrounds of the parasite populations, since the significance of pfmdr1 point mutations for CQ and QN resistance appears to depend on the genetic background of the parasites [17,18]. Meanwhile, another very relevant finding of this study is the existence of dramatic differences in drug responses and genotypes between our studied parasite population and other populations in the GMS. As malaria control is being intensified, geographical separation of parasite populations will occur. Differences in regional drug policy will inevitably exert distinctive selection pressures on parasites. Therefore, investigations into drug resistance mechanisms using different populations may help define the bona fide genetic determinants of artemisinin resistance. Furthermore, since artemisinin resistance appears to be a multigenic trait, mutations in multiple genes may have cumulative effects on parasite's response to artemisinins. Consequently, genetic analysis at multiple loci and association studies with combined haplotypes may generate meaningful insights into the mechanisms of artemisinin resistance [65]. Genome-wide approaches such as SNP array and deep sequencing technologies may be crucial in identifying the molecular basis of the artemisinin resistance. It is also imperative that close resistance surveillance be continued at sentinel sites where artemisinins have been deployed extensively in the past so that the emergence and spread of artemisinin resistance is carefully monitored. Figure S1 Value of the pfmdr1 copy number in 51 P. falciparum isolates. Each symbol indicates the pfmdr1 copy number in one isolate. (TIF)

Author Contributions
Conceived and designed the experiments: LC ZY. Performed the experiments: HM DP JL ZZ ZW. Analyzed the data: HM DP QF HW LC ZY. Contributed reagents/materials/analysis tools: LW RZ. Wrote the paper: DP LC ZY.