Expansion of a Specific Plasmodium falciparum PfMDR1 Haplotype in Southeast Asia with Increased Substrate Transport

Global efforts to eliminate malaria depend on the continued success of artemisinin-based combination therapies (ACTs) that target Plasmodium asexual blood-stage parasites. Resistance to ACTs, however, has emerged, creating the need to define the underlying mechanisms. Mutations in the P. falciparum multidrug resistance protein 1 (PfMDR1) transporter constitute an important determinant of resistance. Applying gene editing tools combined with an analysis of a public database containing thousands of parasite genomes, we show geographic selection and expansion of a pfmdr1 gene amplification encoding the N86/184F haplotype in Southeast Asia. Parasites expressing this PfMDR1 variant possess a higher transport capacity that modulates their responses to antimalarials. These data could help tailor and optimize antimalarial drug usage in different regions where malaria is endemic by taking into account the regional prevalence of pfmdr1 polymorphisms.

I n 2018, an estimated 228 million cases of malaria, predominantly in sub-Saharan Africa, resulted in 405,000 deaths worldwide, mostly in children less than five years old. Plasmodium falciparum is responsible for most of the global malaria burden. Over the past 2 decades, impressive gains have been made in the reduction of global malaria morbidity and mortality rates, which stood at an estimated one million deaths per year in the early 2000s (1). The widespread adoption of artemisinin (ART)-based combination therapies (ACTs) and increased population-level coverage of insecticide-treated bed nets to protect against infective Anopheles mosquito vectors have been important contributors in the fight against malaria (2,3). Unfortunately, the emergence and spread of ART-and ACT partner drug-resistant P. falciparum lineages in the Greater Mekong subregion of Southeast Asia (4)(5)(6)(7)(8)(9) and in South America (10,11) during the past decade pose a significant threat to malaria control and elimination. Efforts are under way to contain the spread of ART resistance into Africa and India (12)(13)(14) where it is predicted to have devastating results (15), as happened earlier with the emergence and global spread of chloroquine (CQ) resistance (8,16,17).
One important mediator involved in multidrug resistance to ACTs is the ATP-binding cassette (ABC) transporter P. falciparum multidrug resistance protein 1 (PfMDR1). PfMDR1 is localized in the membrane of the digestive vacuole (DV) of the parasite and possesses the ability to influx antimalarial drugs toward the lumen of this organelle (18)(19)(20). Transport studies using heterologous expression systems have shown that PfMDR1 variants have the ability to modulate the transport of the antimalarial drugs halofantrine and quinine (QN) (21). A surrogate assay of PfMDR1 activity using the fluorochrome Fluo-4 revealed its subcellular distribution and showed that PfMDR1mediated solute import into the parasite DV is modulated by at least three C-terminal codon variants (S1034C, N1042D, and D1246Y) (22,23). The pfmdr1 N-terminal allelic variants N86Y and Y184F have also been implicated in multidrug-resistant phenotypes. In Africa and Asia, the N86 allele is relevant for in vivo and in vitro parasite antimalarial responses against aryl-amino alcohols such as mefloquine (MFQ) (24,25) and lumefantrine (LMF) (26)(27)(28). In clinical trials of artemether-LMF or artesunate plus amodiaquine (ADQ), an increased risk of posttreatment recrudescent P. falciparum infections was associated with parasites harboring the N86 wild-type or 86Y mutant allele, respectively (29). These in vivo observations have been supported by recent allelic exchange approaches (30). Mutations in pfmdr1 have a drug-specific resistance contribution being "pro and con" depending on the antimalarial. Compared to pfmdr1 N86Y, the impact of the Y184F mutation on parasite responses to ACTs remains less studied (31). This mutation is one of the five canonical PfMDR1 mutations that was earlier considered to be associated with chloroquine resistance (32) and is now widespread in Southeast Asia and Africa (33)(34)(35)(36). The capacity of 184F to influence parasite antimalarial responses is thought to be related to alterations in protein structure through allosteric effects that reduce drug binding or alter transport kinetics (20).pfmdr1 copy number variation (CNV) is another important factor that impacts parasite susceptibility to several antimalarial drugs (18). P. falciparum parasites harboring pfmdr1 amplifications are widely distributed in South America (37) and Southeast Asia (24,25). Amplification of pfmdr1 is associated with an increased risk of treatment failure with therapies combining an aryl-amino alcohol such as MFQ or LMF with an ART derivative (25,38,39).
Using data from the updated MalariaGEN Plasmodium falciparum Community Project (40), we explored the prevalence and distribution of pfmdr1 copy numbers encoding specific haplotypes. We show that pfmdr1 gene amplification is especially common in Southeast Asia, with amplifications being selected along with the N86 and 184F alleles. Applying genome editing tools on the 86Y-and Y184-harboring Southeast Asian Dd2 strain, we demonstrate that the N86/184F haplotype in the context of amplified pfmdr1 imparts a higher transport capacity that directly affects parasite antimalarial responses.

RESULTS
Temporal distribution of PfMDR1 haplotypes and copy number variants in Southeast Asia. We determined the worldwide frequencies of pfmdr1 N86Y and Y184F mutations and pfmdr1 copy number using a collection of 7,113 parasite genomes from the MalariaGEN P. falciparum Community Project version 6 (https://www.malariagen .net/resource/26) (40). Genomic data were available from 73 countries where malaria is endemic with information on pfmdr1 N86Y and Y184F mutations, as well as pfmdr1 copy number, in a subset of 5,003 samples. From that subset, 650 had pfmdr1 amplifications. Of these, 53 were excluded from further frequency analysis due to the presence of multiple alleles at codon positions 86 and/or 184, indicating the presence of mixed infections (see Table S1 in the supplemental material).pfmdr1 allele frequency data were extracted from the MalariaGEN Community Project during the period from 2002 to 2015, revealing that the pfmdr1 N86Y mutation frequency has remained relatively constant in Southeast Asia, with the N86 allele nearly fixed at a frequency of 96 to 98% (N86 allele frequency, 98% between 2002 and 2007, 98% between 2008 and 2010, 99% between 2011 and 2013, and 96% between 2014 and 2015). In Africa, the N86 allele frequency has increased, whereas 86Y has decreased over the same time period (see Fig. S1 in the supplemental material). In contrast to the fixation of the N86 allele in Southeast Asia, the frequency of the 184F variant has increased within the same time frame (21% between 2002  The MalariaGEN Community Project data also showed that pfmdr1 amplifications are the most prevalent in Southeast Asia, with only one African isolate reported in 2009 from Ghana harboring two pfmdr1 copies encoding a N86/Y184 haplotype (Table S1). In Southeast Asia, infections with P. falciparum isolates harboring pfmdr1 amplifications were most prevalent in Thailand (52%), followed by Cambodia, Myanmar, and Vietnam (15%, 14%, and 6%, respectively) (Fig. 1A). In Cambodia, the frequency of isolates with pfmdr1 amplifications decreased progressively from 48% ( Table S1).
Together, these results show that pfmdr1 amplification is especially common in Southeast Asia, with amplifications encoding the N86 and 184F residues being selected.
Amplified pfmdr1-edited parasites at codons 86 and 184. We developed a targeted gene editing approach to study the impact of haplotype-specific increases in pfmdr1 copy number on P. falciparum. Parasite lines derived from Southeast Asian P. falciparum isolates from the 1980s, like Dd2, characteristically contain the 86Y allele. To investigate the advantage of the predominant genotypes observed in Southeast Asia described above (pfmdr1 amplifications containing haplotypes N86/184F or N86/Y184), we performed gene editing in all pfmdr1 gene copies present in Dd2 parasites. Our Dd2 line represents a MFQ-resistant clone carrying four pfmdr1 copies isolated from the W2mef line, originally derived from the Southeast Asian CQ-resistant parasite strain W2 (harboring a single copy of pfmdr1) upon extended culture with MFQ for 96 weeks (43). Dd2 and W2 parasites express the PfMDR1 86Y/Y184 haplotype and the CQ resistanceconferring CVIET haplotype of the P. falciparum chloroquine resistance transporter (PfCRT) ( Table 1) (44,45).
Using the previously described zinc-finger nuclease (ZFN) strategy to engineer pfmdr1 mutations at codons 86 and 184, we successfully generated the Dd2 edited lines, NF Dd2 and NY Dd2 , each maintaining four copies of pfmdr1 encoding the N86/184F and N86/Y184 haplotypes, respectively (Table 1 and Fig. 2A and B). These edited lines represent the two major geographic variants of pfmdr1 present in Southeast Asia and include gene amplifications. The presence of pure N86/184F and N86/Y184 haplotypes in NF Dd2 and NY Dd2 lines, respectively, was confirmed by Sanger sequencing (Fig. 2C). Quantitative PCR analysis demonstrated that the edited NF Dd2 and NY Dd2 lines maintained four copies of pfmdr1 as present in the unedited Dd2 parental strain (Fig. 2D). Attempts to engineer parasites expressing the 86Y/184F haplotype failed. We verified that this haplotype combination was not found among the 597 parasite genomes harboring amplified copies of pfmdr1 in the MalariaGEN Community Project data set  Table S1 in the supplemental material) and stratified into three intervals of time (2002 to 2007, 2008 to 2010, and 2011 to 2013). No samples were available for the years 2014 and 2015. The N86/184F haplotype frequency is shown with 95% confidence intervals in parentheses below each pie chart. Empty pie charts represent time periods in countries where no data were available.
a Plasmids were previously constructed (30). CNV, copy number variation. (Table S1). These findings suggest that the presence of the 86Y/184F haplotype in amplified pfmdr1 copies might render the parasites nonviable or might confer a survival disadvantage to the parasite population. Fluo-4 transport into the parasite DV is altered by amplified PfMDR1 haplotypes. We evaluated the influence of pfmdr1 polymorphisms at codons 86 and 184 on PfMDR1 functional activity by measuring the accumulation of Fluo-4 in the DVs (22) of the gene-edited parasite lines (NF Dd2 and NY Dd2 ) compared to the Dd2 parental control strain.
Accumulation of Fluo-4 in the DVs of NF Dd2 , NY Dd2 , and Dd2 parasites was observed by fluorescence microscopy and measured by flow cytometry (Fig. 3). Fluo-4 accumulation in the DV of the parasite was detectable after 120 min of incubation, prior to which time the probe was distributed throughout the infected red blood cell (RBC) and parasite cytosol (Fig. S2). ZFNs induce a double-strand break (orange thunderbolt) 42 bp downstream of the pfmdr1 start codon, which is repaired by homologous recombination with the 2.4-kb pfmdr1 donor sequence present on the plasmid. The pfmdr1 donor sequence from pmdr1 NY and pmdr1 NF plasmids contains the N86/Y184 or N86/184F genotypes, respectively, and carries three synonymous mutations (nucleotides colored in green) at the ZFN binding site to prevent ZFNs from cleaving the plasmids or the pfmdr1-edited sequences. Two of the three synonymous mutations create a SmlI restriction site and abolish a BstEII restriction site present in the nonedited (genomic) template. (B) PCR-RFLP genotyping confirming the presence of pfmdr1-edited lines NY Dd2 and NF Dd2 . The silent mutations introduced into the donor sequence created a SmlI restriction site, which was used to screen PCR products (yielding a 1.3-kb band with P3 and P4 primers, which formed a doublet at 667/669 bp in edited parasites upon SmlI digestion). BstEII cuts the P3 and P4 PCR product in nonedited parasites and was used as a control; the observed 1.3-kb band was not cut, indicating the presence of correctly edited parasites. (C) Sanger sequencing chromatograms of DNA samples from the NY Dd2 and NF Dd2 edited lines at the pfmdr1 locus, confirming correct editing at codons 86 and 184. (D) Plot showing means plus standard errors of the means (SEM) (error bars) from three independent assays of pfmdr1 copy number, analyzed through quantitative PCR using ␤-tubulin as a control gene and the 3D7 strain as a calibrator with a single pfmdr1 copy.
To determine whether there were PfMDR1 haplotype-specific differences in the accumulation of Fluo-4, we measured the Fluo-4 intensity of the pfmdr1-edited lines and the W2 comparator strain after normalizing to the fluorescence intensity of Dd2 parasites after 240 min. The results showed that the N86 allele has an important role in the modulation of the PfMDR1 transporter since N86 edited parasites accumulated more Fluo-4 compared with the Dd2 parental strain containing the 86Y allele (P ϭ 0.03). Furthermore, edited NF Dd2 parasites containing the N86/184F haplotype accumulated Values below or greater than zero indicate a decrease or increase, respectively, in the percentage of accumulated fluorescence compared with Dd2 parasites. Mann-Whitney U tests were used to assess statistically significant differences between the NF Dd2 , NY Dd2 , Dd2, and W2 parasites in panels B and C (*, P Ͻ 0.05). (D) Normalized Fluo-4 intensities (as a percentage) in the absence (untreated control) and presence of elacridar (ELC, 0.1 M) and verapamil (VP, 0.8 M). Results were normalized to untreated controls (indicated in the graph as the baseline [y ϭ 0]). Values below or greater than zero indicate a decrease or increase, respectively, in the percentage of accumulated fluorescence compared with untreated controls. The means plus SEM of four independent measurements are shown. Mann-Whitney U tests were used to assess statistically significant differences between untreated and treated parasites (*, P Ͻ 0.05). more Fluo-4 in their DVs than edited NY Dd2 parasites (haplotype N86/Y184) (P ϭ 0.03). W2 parasites, which harbor a single genomic copy of the pfmdr1 gene with the 86Y/Y184 haplotype, accumulated less Fluo-4 (P ϭ 0.03) relative to the parental Dd2 control strain that expresses four copies with the same 86Y/Y184 haplotype (Fig. 3C). The influence of pfmdr1 copy number on the import of Fluo-4 into the parasite DV observed herein confirms previous reports where a parasite strain harboring only one pfmdr1 copy accumulated less Fluo-4 compared with parasites expressing two copies (31).
To determine whether the changes in Fluo-4 accumulation were related to the PfMDR1 haplotype, we used elacridar (ELC) as a tool, since it is a well-established inhibitor of P-glycoprotein type ABC transporters (46). We measured the fluorescence intensity of Fluo-4 in NF Dd2 , NY Dd2 , or Dd2 parasites in the presence of ELC. Data were normalized to the respective untreated controls (NF Dd2 , NY Dd2 , or Dd2; indicated in Fig. 3D as the baseline [y ϭ 0]). We observed a significantly decreased Fluo-4 accumulation in ELC-treated parasites compared with untreated controls (P ϭ 0.03). The addition of verapamil (VP), a calcium ion channel blocker that reverses parasite resistance to CQ, led to a significant reduction of Fluo-4 accumulation only in the edited NF Dd2 line compared to untreated NF Dd2 (P ϭ 0.03), suggesting a possible interplay between the PfMDR1 and PfCRT transporters in 184F-expressing parasites (Fig. 3D). The addition of VP produced a trend toward more Fluo-4 accumulation in NY Dd2 and Y184-harboring Dd2 parasites compared with NF Dd2 . Nevertheless, despite having a different trend, the quite large standard error of the mean (SEM) made these results inconclusive.
Amplified pfmdr1 with the N86 allele decreases parasite susceptibility to antimalarials. We determined the impact of the N86 allele in the genomic context of multicopy pfmdr1 on parasite susceptibility to antimalarial drugs. Replacement of the 86Y variant present in Dd2 with N86 in each of the four pfmdr1 copies (edited parasite line NY Dd2 ) induced a significant increase in the half-maximal in vitro growth inhibitory concentration (IC 50 ) values for MFQ (2.5-fold, P ϭ 0.02) and LMF (2.4-fold, P ϭ 0.02), as well as for dihydroartemisinin (DHA) (2.3-fold, P ϭ 0.05) ( Fig. 4 and Table 2). A reduction in the IC 50 values for MFQ (6.1-fold, P ϭ 0.05) in W2 (a single pfmdr1 copy strain with the 86Y/Y184 haplotype) compared with Dd2 parasites was observed. These data show a clear role of pfmdr1 gene amplifications, particularly those encoding the N86 allelic variant, in the multidrug resistance phenotype.
In the case of QN, the replacement of 86Y to N86 in multicopy pfmdr1 resulted in a decrease of IC 50 values in the NF Dd2 and NY Dd2 lines (1.6-fold, P ϭ 0.02) relative to the parental Dd2 strain.
The N86/184F PfMDR1 haplotype decreases parasite susceptibility to LMF. Analysis of the MalariaGEN P. falciparum Community Project data predicted a prevalence of the pfmdr1 184F allele in Africa of 58.6% (184F, 1,438/2,454 genomes sampled between 2002 and 2015) and 36% (184F, 858/2,403 between 2002 and 2015) in Southeast Asia, regardless of the number of pfmdr1 copies (Table S1 and Fig. S1). The frequency of the 184F allele increased worldwide over the years analyzed (Table S1 and Fig. S1). Here, in the context of amplified pfmdr1 gene copies, we observed up to 1.7-fold (P ϭ 0.002) decreased susceptibility to LMF in edited parasites expressing the N86/184F haplotype compared with isogenic parasites expressing the N86/Y184 haplotype ( Fig. 4 and Table 2). This suggests a specific effect of the 184F allele in conferring resistance to LMF in parasites that harbor amplified copies of pfmdr1.
PfMDR1-mediated resistance phenotypes are modulated by VP and ELC. We investigated the potential interplay between PfMDR1 and PfCRT transporters in modulating parasite susceptibility to antimalarials using the PfCRT and P-type ATPase transporter inhibitors VP and ELC, respectively. All the parasite strains used in this study express the CVIET variant at amino acid positions 72 to 76 of PfCRT (Table 1). In the presence of VP, the multicopy pfmdr1 Dd2 strain and the edited parasites lines NF Dd2 and NY Dd2 exhibited increased susceptibility to the 4-aminoquinolines CQ and md-ADQ compared with the respective untreated parasites, with 5-fold and 2-to 3-fold changes, respectively. Increases in susceptibility were also obtained with QN and MFQ in the presence of VP compared with untreated parasites, with N86 producing higher fold changes than 86Y (QN, NF Dd2 ϭ 2.8-fold, NY Dd2 ϭ 3.1-fold, and Dd2 ϭ 2.2-fold; MFQ, NF Dd2 ϭ 4.1-fold, NY Dd2 ϭ 5.0-fold, and Dd2 ϭ 2.9-fold) ( Table 2). The differences in fold changes observed in the presence of VP can be explained by the differential basal levels of susceptibility to QN and MFQ in the absence of VP in NF Dd2 , NY Dd2 , and Dd2 parasites (Fig. 4). For LMF, VP sensitized NF Dd2 and NY Dd2 parasites up to twofold with no effect observed on unedited Dd2 parasites compared with untreated parasites ( In the presence of the P-type ATPase transporter inhibitor ELC, the parasite lines NF Dd2 and NY Dd2 exhibited 2.8-fold and 2.5-fold increased susceptibility to MFQ, respectively, compared with the corresponding untreated parasites. It is interesting to observe that ELC treatment reverted the MFQ IC 50 values of NF Dd2 and NY Dd2 parasites to the same level as that of Dd2 in the absence of ELC, validating the impact of the N86 pfmdr1 variant. ELC treatment did not significantly impact the susceptibility of Dd2, NY Dd2 , or NF Dd2 parasites to any of the other aryl-amino alcohols tested (i.e., QN and LMF) or the 4-aminoquinolines (CQ and md-ADQ).

DISCUSSION
The propensity of P. falciparum to develop antimalarial drug resistance underscores the importance of characterizing its genetic basis. The amplification of pfmdr1 was the first candidate mechanism implicated in CQ resistance (CQR) (47). Although we now know that PfCRT is the primary determinant of CQR (48), pfmdr1 can modulate the degree of resistance (49,50). Several other studies have also shown that increased pfmdr1 copy number decreases the parasite's susceptibility to other antimalarials such as MFQ, LMF, and ART (25,38,39). The worldwide spread of the pfmdr1 N-terminal 86Y and 184F mutations have also been linked with differential responses to multiple quinoline-based antimalarials and ART, both in vitro and in vivo (33,36,(51)(52)(53)(54). The conjugation of these two types of polymorphisms (gene copy number and specific sequence haplotypes) could modulate the P. falciparum response to these antimalarials.
Using data from the MalariaGEN P. falciparum Community Project, we determined the worldwide prevalence and temporal changes of pfmdr1 copy number with specific pfmdr1 N-terminal mutations. Furthermore, we measured the impact of these polymorphisms on solute kinetics of this transporter. We note that the MalariaGEN data from a given period (2002 to 2015 in our case) relates to samples whose locations were dependent on which partner studies were operative at the time (40). Temporal trends in aggregated data should therefore be interpreted with caution. From 73 countries where malaria is endemic with information on pfmdr1 amplifications and mutations at amino acid positions 86 and 184, between 2002 and 2015, pfmdr1 amplification was found almost exclusively in Southeast Asia, where we also observed a near fixation of the N86 allele ( Fig. 1; see also Table S1 and Fig. S1 in the supplemental material). Considering the previous strong association of pfmdr1 amplifications and the N86 allele with treatment failures that included MFQ or LMF (24,25,39), the frequencies observed could be explained by the extensive use of artesunate plus MFQ and artemether plus LMF in this region during that time period. Thailand, for example, is the country with  values (nanomolar) were derived from three to five independent drug assays. Mann-Whitney U tests were used to assess statistically significant differences between the untreated control parasites (Dd2, NY Dd2 , and NF Dd2 ) and VP-or ELC-treated parasites. NS, not significant.
the highest frequency of pfmdr1 amplifications, possibly reflecting the extensive use of artesunate-MFQ as a first-line antimalarial treatment. Its failure during the following years led to changes in treatment policy, with the adoption of dihydroartemisinin (DHA)-PPQ as first-line therapy in 2008 (1,2). This combination was associated with a rapid decline in pfmdr1 copy number (55,56), which could have resulted from PPQmediated selection against multicopy pfmdr1 and/or selective pressure against multicopy pfmdr1 parasites that show some loss of fitness and can therefore be outcompeted by single-copy pfmdr1 parasites (2,57,58). In Cambodia, the loss of DHA-PPQ efficacy led to the resumption of artesunate-MFQ use beginning in 2011, and this combination retains good efficacy with pfmdr1 copy number generally remaining at one (59). In contrast, the MalariaGEN data illustrate pfmdr1 amplifications as a rare event in Africa. The substantial use of CQ in the past and the current use of the ACT combination artesunate-ADQ possibly contribute to this fact by selecting for deamplifications (60,61). Furthermore, although artemether-LMF has been used for many years in Africa (62), the lack of MFQ usage could partially explain the lack of pfmdr1 amplifications. Nevertheless, it is conceivable that pfmdr1 amplifications may increase in prevalence in Africa due to the widespread use of artemether-LMF (63,64). The presence of the 86Y allele in Africa could also contribute to the low prevalence of pfmdr1 amplifications as this allele is present at a very low frequency worldwide when combined with pfmdr1 amplifications (6/597 genomes). Another factor could be related to the known fitness cost associated with pfmdr1 amplifications, which manifests as decreased parasite survival in the absence of drug pressure (58,65). This feature is of particular relevance in high transmission areas such as in sub-Saharan Africa. An interesting observation is that within the MalariaGEN data, pfmdr1 amplifications with the haplotype 86Y/184F were not present and despite several attempts to engineer this specific parasite line, we were unsuccessful, possibly related to a genetic background that is detrimental for survival.
Over time in Southeast Asia, we observed a clear selection of the 184F allele in combination with pfmdr1 amplifications containing the N86 allele. Cambodia was the country with the highest observed selection (25% [3/12] (Fig. 1). It is important to note that at the same time that this takeover of the 184F occurs, the prevalence of pfmdr1 duplications was actually shrinking due to the switch in first-line treatment to DHA-PPQ. To better understand this selection of the N86/184F haplotype in the context of amplified pfmdr1, we generated two edited lines representing the two major geographic variants from Southeast Asia and studied the impact of the specific haplotypes on antimalarial responses. The impact of these polymorphisms on PfMDR1 functional transport was also measured through accumulation of Fluo-4 in the DV of the parasite (Fig. 3). This approach has been previously used for pfmdr1 variants and revealed an important impact of the C-terminal amino acid 1042 on Fluo-4 transport (22,23). In these studies, although they used strains encompassing different haplotypes in regard to the pfmdr1 N-terminal mutations, their direct impact was not evident. In our Fluo-4 accumulation study with the Dd2 strain and isogenic lines differing only in their pfmdr1 N-terminal residues at positions 86 and 184, we observed that parasites expressing the mutations 86Y and 184F were capable of transporting Fluo-4 into the DV (Fig. 3A). The N86 allele by itself impacted Fluo-4 transport capacity, with significant differences obtained between the parental Dd2 strain and NY Dd2 (Fig. 3C). Transport studies in Xenopus oocytes have also provided evidence of the impact of polymorphisms at position 86 on the ability of PfMDR1 to transport different drugs. A single amino acid alteration from N86 to 86Y in PfMDR1 resulted in loss of the ability to transport QN and CQ (21). Consistent with this observation, the presence of the N86 allele induced a significant impact on antimalarial responses (Fig. 4). We were able to confirm the involvement of residue 86, specifically in the context of multicopy pfmdr1, in modulating parasite susceptibility to multiple antimalarials. Gene-edited parasites with amplified pfmdr1 harboring the N86 allele displayed decreased susceptibility to MFQ, LMF, and DHA as well as increased susceptibility to QN, compared to the unedited Dd2 parental line possessing the 86Y allele (Fig. 4). These findings are consistent with our previous allelic exchange work that used ZFN-edited parasites expressing the 86Y variant in the context of single-copy pfmdr1 (30). This observation explains the selection of the N86 allele after treatment with artemether-LMF (26)(27)(28) and corroborates with N86 being a risk factor for recrudescence following treatment with artemether-LMF compared with P. falciparum infections containing 86Y (29).
The pfmdr1 184F mutation adds to the role of polymorphisms at position 86 in modulating parasite drug resistance and transport. We found that edited parasites harboring the N86/184F haplotype accumulated more Fluo-4 inside the DV compared to edited parasites with the N86/Y184 haplotype, accompanied by faster kinetics of PfMDR1-mediated transport activity (Fig. 3B). A previous in silico PfMDR1 model proposed that the 184F allele alters the parasite's response by an allosteric effect on transport kinetics, independent from drug-binding capacity (20). Our gene-edited parasites expressing amplified pfmdr1 with the N86/184F haplotype displayed a mild but significant decreased susceptibility to LMF (1.7-fold) compared to those harboring the PfMDR1 N86/Y184 haplotype or to the 86Y/Y184-bearing parental Dd2 parasites (Fig. 4). The mild but significant fold difference herein observed was not detected in our previous allelic exchange work for the same haplotype, possibly due to the single pfmdr1 copy present in the NF10 and KC5 strains precluding sufficient expression to display this phenotype (30). In vivo selection of the 184F allele has been reported, typically along with other pfmdr1 alleles, after artemether-LMF treatment. Most notably, the N86/184F haplotype was selected upon the regular six-dose therapy with artemether-LMF (26,27). In their analysis of two clinical trials after artemether-LMF treatment, Malmberg et al. demonstrated that recrudescent infections harboring the PfMDR1 N86/184F haplotype, in the context of a single pfmdr1 copy, were able to endure higher blood LMF concentrations (28). Together with our results that examine the role of copy number, these in vivo data suggest that although the pfmdr1 184F allele might not be the primary determinant for parasite resistance, it could provide a complementary genetic background that facilitates the acquisition of multidrugresistant phenotypes, possibly leading to the specific haplotype selection observed in Southeast Asia (Fig. 1).
The interplay between PfMDR1 and PfCRT transporters in modulating parasite susceptibility to antimalarials was also evident in this work (Table 2). Heightened susceptibilities of NF Dd2 , NY Dd2 , and Dd2 P. falciparum lines to MFQ, md-ADQ, CQ, and QN were affected by VP as previously observed in other genetic backgrounds (48,66). The role of PfMDR1 in the parasite's response to QN remains unclear, but mutant PfCRT is known to contribute to QN resistance (67). Our results with VP also suggest a role for PfCRT. We note that VP increased susceptibility to LMF only in the edited parasites and not in the unedited Dd2 control. Prior data showed selection of the pfcrt CQ-sensitive wild-type K76 allele occurs upon artemether-LMF treatment, as confirmed in vitro using allelic exchange (68). In the case of LMF, VP might have a greater effect on PfMDR1 in the context of N86, since we observed that VP decreased Fluo-4 accumulation in the parasite's DV in N86 parasites (Fig. 3D). This observation would explain the increased parasite susceptibility to LMF upon VP exposure.
In the presence of ELC, decreased MFQ IC 50 values were observed in NF Dd2 and NY Dd2 edited parasites, reinforcing the hypothesis that not only are pfmdr1 amplifications associated with MFQ resistance (24,25,39) but also that the pfmdr1 N86 allele is a key mediator of MFQ susceptibility. These findings are in line with the MalariaGEN data showing that almost all parasites that contain pfmdr1 amplifications are in the context of the pfmdr1 N86 allele.
In conclusion, our in vitro parasite drug susceptibility data using pfmdr1 gene-edited parasites reinforce the role of both pfmdr1 CNV and haplotype variation in modulating parasite responses to multiple first-line antimalarial drugs. Our investigation highlights the steady geographic expansion of a parasite population harboring a pfmdr1 gene amplification with the N86/184F haplotype in Southeast Asia, which we show has a proven higher DV transport efficacy. These data could help tailor and optimize antimalarial drug usage in different areas where malaria is endemic by taking into account the regional prevalence of pfmdr1 polymorphisms.

MATERIALS AND METHODS
Geographical distribution of P. falciparum isolates carrying pfmdr1 copy number amplifications and the mutations N86Y and Y184F. Genome analysis used data generated by the Wellcome Sanger Institute as part of the MalariaGEN Plasmodium falciparum Community Project version 6 (https:// www.malariagen.net/resource/26) (40). Sample metadata were obtained from https://www.malariagen .net/resource/26 and were combined with the genotype data for pfmdr1 polymorphisms at codons 86 and 184. pfmdr1 copy number was determined using a combination of a coverage-based approach and a method based on position and orientation of reads near discovered duplication breakpoints as described elsewhere (41). The map presented in Fig. 1 was created using mapchart.net, geo-referencing the pfmdr1 gene copy number prevalence in the region.
Parasite culture. The Dd2 strain (MRA-156, MR4-Malaria Resources) was maintained at ϳ4% hematocrit with human red blood cells in RPMI 1640 medium supplemented with 2 mM l-glutamine, 200 M hypoxanthine, 0.25 g/ml gentamicin, 25 mM HEPES, 0.2% NaHCO3, and 0.25% Albumax II (Invitrogen; ThermoFisher Scientific). Red blood cells were isolated either from whole blood or from buffy coat samples provided by healthy donors of all blood types. Parasite cultures were maintained at 37°C under a humidified controlled atmosphere of 5% O2/5% CO2/90% N2. Parasite growth was monitored by inspecting Giemsa-stained blood smears. Parasite synchronization was performed with 5% sorbitol for 15 min at 37°C. To obtain highly synchronous cultures, sorbitol was added to the culture to eliminate trophozoites. This was repeated 20 h and 44 h after the initial treatment to obtain 4-h postinvasion ring stages that then mature into trophozoites, which were used to measure Fluo-4 fluorescence.
P. falciparum transfections. The previously designed plasmids pmdr1 NF and pmdr1 NY (30) were used to transfect Dd2 parasites. Briefly, we introduced the desired mutations into the endogenous pfmdr1 locus of Dd2 parasites using ZFN-mediated genome editing. These customized ZFNs bind on opposite strands of pfmdr1, producing a double-stranded break 42 bp downstream of the start codon. DNA repair proceeds via homologous recombination, as P. falciparum lacks the nonhomologous end-joining pathway (69). Our homology-driven repair template consisted of a 2.4-kb pfmdr1 fragment that encompassed codons 86 and 184 (30).
Ring-stage cultures at 5% parasitemia were electroporated with ϳ50 g of plasmid DNA diluted in Cytomix (120 mM KCl, 0.15 mM CaCl 2 , 2 mM EGTA, 5 mM MgCl 2 , 10 mM K 2 HPO 4 /KH 2 PO 4 , 25 mM HEPES [pH 7.6]). The plasmids express the human dihydrofolate reductase (hDHFR) selectable marker, which confers resistance to the antifolate drug WR99210. To allow transient expression of the ZFNs, WR99210 selection was applied to transfected cultures beginning 24 h postelectroporation and maintained for 6 days to select for edited parasites. Parasites were visible microscopically 6 weeks postelectroporation and screened for editing events via blood PCR using the primers p3 and p4. Positive bulk cultures were cloned by limiting dilution in 96-well plates. After 15 days, cloning plates were screened and parasitepositive wells were screened for ZFN-mediated editing by blood PCR amplification using a supreme NZYTaq II 2ϫ Green Master Mix (NZYTech, Lisbon, Portugal) and the p3 and p4 primers (see Table S2 in the supplemental material). The donor sequence includes silent mutations that create a SmlI restriction site, which was used to screen the PCR products by restriction fragment length polymorphism (RFLP) analysis (a 1.3-kb band obtained with primers P3 plus P4 [ Table S2] that forms a doublet at 667/669 bp with SmlI in edited parasites). The BstEII restriction site is present only in the nonedited genomic pfmdr1 sequence and was therefore used as a control in the PCR-RFLP assay (Fig. 2B). Editing events at codons 86 and 184 were confirmed by Sanger sequencing using the p5 primer (Table S2). Successfully edited parasite lines (NF Dd2 and NY Dd2 ) were expanded and cryopreserved.
In this work, we have compared the edited NF Dd2 and NY Dd2 lines to an unedited Dd2 parental control harboring the 86Y/Y184 haplotype. An edited Dd2 control was not performed in this study, since the plasmids used herein to edit Dd2 parasites were the same as those used previously by our group (30).
In that study, we demonstrated that gene-edited control parasites carrying only silent ZFN binding site mutations did not alter the parasite's response to antimalarial drugs compared with unedited parental lines.
Copy number analysis of pfmdr1. Genomic DNA from parental control Dd2 and pfmdr1-edited parasites (NY Dd2 and NF Dd2 ) were extracted using the NZY Blood gDNA isolation kit (NZYTech, Lisbon, Portugal) with RNase treatment. Copy numbers of the pfmdr1 gene were determined using a NZYSpeedy Green Master Mix (2ϫ) (NZYTech, Lisbon, Portugal). The amplification reactions were carried out in 20-l volumes in a 96-well plate using 10 ng of genomic DNA, 400 nM each forward and reverse primer, and the master mix. The P. falciparum ␤-tubulin gene was used as the housekeeping control gene. The pfmdr1 and pf␤-tubulin forward and reverse primers (Table S2) were previously designed (70). For each run, 3D7 was used as a single-copy-number control for pfmdr1. Thermal cycling was performed at 98°C for 3 min, followed by 44 cycles with 1 cycle consisting of 98°C for 15 s, 59°C for 30 s, and 72°C for 25 s using the Bio-Rad CFX96 real-time system C1000 thermal cycler. Results were analyzed by the 2 Ϫ ΔΔCt (threshold cycle) method of relative quantification. The ΔΔCt calculation was used as follows ΔΔCt ϭ (Ct of pfmdr1 -Ct of Pf␤-tubulin) -(Ct of pfmdr1 -Ct of Pf␤-tubulin)y, where is sample and y is P. falciparum 3D7 strain. The average gene copy number was calculated from three biological replicates, each with technical replicates run in duplicate.

Live-cell imaging of Fluo-4 accumulation in the DV.
Fluorescence microscopy was used to evaluate the accumulation site of Fluo-4 AM (Invitrogen; ThermoFisher Scientific) in different parasite strains in trophozoite stages. Synchronized trophozoite cultures, at ϳ2% parasitemia and 2% hematocrit, were washed with RPMI 1640 medium and loaded with 5 M Fluo-4 diluted in RPMI 1640 medium. Parasites were incubated for 240 min at 37°C in a humidified controlled atmosphere of 5% O2/5% CO2/90% N2. After 240 min, the parasites were washed twice with 1ϫ phosphate-buffered saline (PBS) and transferred onto a slide. Fluorescence microscopy was performed using an Olympus BX61 microscope equipped with a visible light laser, and the images were recorded with a digital camera (DP70). The parasites incubated with Fluo-4 were excited at 488 nm with the emission in the green channel (505-nm filter), and an exposure time of 457 ms was applied to obtain the images. Single images were obtained using a 100ϫ objective lens. Regions of interest within the infected RBC, including the parasite cytosol and the parasite DV, were recorded with Cell^P software (Electro optics, UK) and the image overlays were obtained using Image J software version 1.52a.
Quantification of Fluo-4 accumulation by flow cytometry. Flow cytometry was used to quantify Fluo-4 accumulation in NF Dd2 , NY Dd2 , Dd2, and W2 parasites. Parasite labeling and evaluation of Fluo-4 accumulation were performed as described previously (22) with minor modifications. Briefly, synchronized trophozoites were incubated with 5 M Fluo-4 AM (Invitrogen; ThermoFisher Scientific) for 240 min at 37°C in an airtight environment flushed with 5% O2/5% CO2/90% N2. The level of fluorescence in infected RBCs was recorded by flow cytometry every 15 min until 120 min, and every 30 min thereafter up to a total incubation period of 240 min (see Fig. S2 in the supplemental material). This assay was followed by live-cell imaging at all time points described above to visualize when the Fluo-4 probe accumulated in the parasite DV (Fig. S2). Fluo-4 accumulation in the DV, presented in Fig. 3, was confirmed by fluorescence microscopy before flow cytometry. Thirty minutes before the incubation ended, parasites were loaded with 1.6 M Mitotracker Deep Red FM (Invitrogen; ThermoFisher Scientific). For flow cytometry, labeled parasites were excited at 488 nm and 633 nm to detect Fluo-4 and Mitotracker Deep Red FM, respectively. Forward (FSC) and side scatterplots (SSC) were used to define the RBC populations followed by gating double-positive populations for Fluo-4 and Mitotracker Deep Red FM. Approximately 200,000 events were read per experimental condition, and the mean Fluo-4 intensity was calculated for each parasite strain. Due to intrinsic variability observed in the different replicate assays, the fluorescence intensity of NF Dd2 , NY Dd2 , and W2 was normalized to the Dd2 parental control (i.e., Dd2 corresponding to 100%; indicated in the graph as the baseline [y ϭ 0]). Values below zero indicate a decrease in the percentage of accumulated fluorescence compared with Dd2 parasites, whereas values above zero indicate an increase.
Fluo-4 accumulation in NF Dd2 , NY Dd2 , and Dd2 parasites was also measured in the presence or absence of ELC, an inhibitor of P-type ATPase transporters, including PfMDR1 (and others, e.g., ABCI3), or VP, a calcium channel blocker that functions as a CQ resistance reversal agent. Briefly, synchronized trophozoites were washed with RPMI 1640 medium and preincubated for 10 min with ELC (0.1 M) or VP (0.8 M). Sample preparation and flow cytometry were performed as described above. The fluorescence intensities of NF Dd2 , NY Dd2 , and Dd2 in the presence of VP or ELC were normalized to the respective untreated controls (indicated in the graph as the baseline [y ϭ 0]). Values below zero indicate a decrease in the percentage of accumulated fluorescence compared with untreated controls, while values above zero indicate an increase.
In vitro antimalarial drug assays. Drug susceptibility assays using CQ, DHA, MFQ, QN, LMF, md-ADQ, PPQ, and PND in the absence or presence of ELC (0.1 M) or VP (0.8 M) (48) were run for the edited parasite lines NF Dd2 and NY Dd2 , the parental control strain Dd2, and the W2 strain. These assays were performed using a published flow cytometry-based method (30). Briefly, synchronized ring-stage parasites at 0.2% starting parasitemia and 1% hematocrit were incubated in the presence of different concentrations of drug (across a dilution range of twofold; 12-point dilution series). After 72 h of incubation at 37°C, samples were stained with 1.6 M Mitotracker Deep Red FM (Invitrogen; Thermo-Fisher Scientific) and 2ϫ SYBR green (for DNA staining) (Invitrogen; ThermoFisher Scientific) in 1ϫ PBS for 30 min and analyzed by flow cytometry to determine the parasite growth rates. Approximately 100,000 events were captured per well. In vitro IC 50 values were calculated using nonlinear regression analysis performed with GraphPad Prism 6 software.
Statistical analysis. Nonparametric, two-tailed Mann-Whitney U tests were used to assess IC 50 antimalarial responses in the presence or absence of VP or ELC and Fluo-4 accumulation differences between the NF Dd2 and NY Dd2 , Dd2, and W2 parasite strains (normal distribution not assumed; performed with GraphPad Prism Software).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
This study used data from the MalariaGEN Plasmodium falciparum Community Project. Genome sequencing was done by the Wellcome Trust Sanger Institute (WTSI), and sample collections were coordinated by the MalariaGEN Resource Centre. We thank the staff of the WTSI Sample Logistics, Sequencing, and Informatics facilities for their contributions and Roberto Amato for help with data acquisition and for critically reading the manuscript. We also thank António Marques from Immuno-hemotherapy services at Hospital de Braga and Hugo Sousa and Luisa Santos from Immunohemotherapy services, Portuguese Oncology Institute, Porto (IPO-Porto) for providing red blood cells from healthy volunteers for in vitro P. falciparum cultures. The P. falciparum strains MRA-102 (3D7) and MRA-150 (Dd2) were kindly provided by the Malaria Research and Reference Reagent Resource Center (MR4) (https://www .beiresources.org/MR4Home.aspx).