Using customized ZFNs we disrupted the pfmrp1 gene in P. falciparum Dd2 and 3D7 parasites, by single crossover integration of plasmid into the gene locus (Supplementary figure 1). These two strains originate from Southeast Asia and Africa respectively, are part of genetically distinct sub-populations of P. falciparum, and harbor very distinct antimalarial drug responses (Figure 1)30–33. ZFNs were engineered to bind neighboring sites on opposite strands of pfmrp1, producing a double-stranded break 1208 bp upstream of the stop codon. Our homology-driven template consisted of a 2.5 kb pfmrp1 fragment engineered with three non-synonymous mutations at the ZFN binding site to prevent ZFNs from cleaving the plasmid, and promoted site-specific plasmid integration, thereby disrupting pfmrp1.
Red blood cells (RBCs) infected with ring-stage P. falciparum 3D7 or Dd2 parasites were electroporated with the pfmrp1 ZFN disruption plasmid (Supplementary figure 1). This plasmid expresses the blasticidin S deaminase selection marker conferring resistance to blasticidin. Parasite growth was observable by microscopy 18-22 days post drug selection. pfmrp1 disruption was confirmed by PCR, which identified the full-length gene only in non-mutated WT strains (Supplementary figure 1). PCRs spanning both ends of integration were positive for knockout (Δpfmrp1) strains, denoted 3D7Δpfmrp1 and Dd2Δpfmrp1, which were then cloned by limiting dilution (Supplementary figure 1). RNA sequencing analysis (Supplementary figure 2 and data set 1) showed that pfmrp1 was expressed at a lower level in the Dd2Δpfmrp1 line, confirming that the disruption influenced the expression of the full transcript.
pfmrp1 disruption has no impact on quinoline and dihydroartemisinin susceptibility
In vitro drug susceptibility assays were performed on WT and Δpfmrp1 lines. Tightly synchronized ring-stage parasites were incubated for 72 hours with antimalarials, using a range of 10 compound concentrations serially diluted by 2-fold. Parasitemias were quantiଁed using ଂow cytometry. IC50 values were determined for CQ, mefloquine (MQ), lumefantrine (LMF), dihydroartemisinin (DHA), PYR, TMP, WR, MTX and aminopterin (AMT). Of note, there were no alterations in IC50 values for the commonly used quinoline antimalarials CQ, MQ and LMF, nor to DHA for either 3D7Δpfmrp1 or Dd2Δpfmrp1 strains when compared to their parental strains (Figure 1A). IC50 values were in the range of previous reports. The main differences observed in CQ susceptibility between the parental Dd2 and 3D7 strains are explained by known polymorphisms in the P. falciparum chloroquine resistance transporter (PfCRT) and PfMDR114,34−36.
pfmrp1 disruption leads to resistance to folate analogs
3D7Δpfmrp1 and Dd2Δpfmrp1 were strikingly less sensitive to the antifolates MTX and AMT, compared to their respective parental strains. The IC50 values for the 3D7Δpfmrp1 and Dd2Δpfmrp1 lines were increased by 15 and 40-fold when assayed against MTX (P<0.0001) and 6 and 25-fold against AMT (P<0.0001), respectively, compared to their WT parental controls (Figures 1A and 1D). However, WT and Δpfmrp1 lines displayed comparable responses to the antifolates PYR, TMP and WR. Strain-specific differences in susceptibility to PYR, TMP and WR were observed as expected due to genotypic differences in the P. falciparum dihydrofolate reductase (pfdhfr) gene37,38. MTX and AMT are folate analogs (Figure 1B), while PYR, TMP and WR are structurally distinct to folate, which might account for the differential pattern of susceptibility changes to these antifolates. These results suggest that loss of pfmrp1 leads to parasite resistance to folate analogs. This would be incompatible with an exporter function, as the disruption should lead to increased drug concentration in the parasite cytoplasm leading to heightened drug sensitivity in the Δpfmrp1 lines. These findings suggest PfMRP1 as an importer present in the P. falciparum plasma membrane.
To confirm that the observed resistance phenotype was mediated by PfMRP1, the MRP chemical inhibitor MK571 was used to block PfMRP1 transport in the Dd2 strain39–41. A small but significant increase (1.3-fold, P<0.001) in IC50 was observed for MTX only in MK571-treated Dd2 WT parasites (24.7 ± 1.4 nM) compared to untreated controls (18.8 ± 0.8 nM), whereas no difference was observed for the MK571-treated Dd2Δpfmrp1 line, supporting a PfMRP1-mediated effect (Figure 1C). The fold-increase in MTX IC50 values by chemical inhibition was considerably lower compared to the increase observed with the Dd2Δpfmrp1 line (Figures 1A and 1D), which might be explained by incomplete inhibition of PfMRP1-mediated transport by MK571.
We next examined the effect of inhibiting PfCRT and PfMDR1 in Dd2Δpfmrp1 parasites with verapamil and elacridar, respectively, to understand the possible crosstalk between these antimalarial drug resistance mediators and PfMRP1. Verapamil-mediated inhibition led to a small (1.2-fold, P<0.01) but significant decrease in MTX susceptibility in Dd2Δpfmrp1 parasites (896 ± 42.4 nM compared to 767 ± 42.4 nM), whereas no effect was observed in Dd2 WT (Supplementary Figure 3). These results suggest that PfCRT might interplay with PfMRP1 to augment MTX resistance (Supplementary Figure 3). This increased resistance could be due to trapping of residual MTX in the digestive vacuole, away from its cytosolic site of action, with MTX being unable to be effluxed by PfCRT in the presence of verapamil that is a known blocker of PfCRT-mediated drug transport42. No effect was observed with elacridar-mediated inhibition (Supplementary Figure 3).
pfmrp1 -disrupted parasites accumulate less fluorescein methotrexate
To understand how resistance to folate analogs occurs in the Δpfmrp1 parasites, we measured the time-dependent accumulation of fluorescein MTX (F-MTX), a fluorescent derivative of MTX, using flow cytometry. Remarkably, the Dd2Δpfmrp1 line showed a significant lack of F-MTX accumulation over time, with an average of 3.3 times less F-MTX fluorescence intensity after 3 hours of incubation compared to Dd2 WT parasites (P<0.01) (Figure 2A). The observed F-MTX phenotype seems specific to PfMRP1-mediated transport (Figure 2B). Fluo-4 is a calcium probe that enters the parasite through passive diffusion, accumulates in the parasite digestive vacuole, and can be modulated by PfMDR1-mediated transport43,44. Repeating the assay with Fluo-4 showed no differences in accumulation between Dd2 WT and Δpfmrp1 parasites (Figure 2B). These results provide evidence that PfMRP1 can import F-MTX, an effect that is much reduced in Dd2Δpfmrp1, thus explaining the resistance phenotype. The residual accumulation observed in Dd2Δpfmrp1 might occur through a distinct transport pathway, passive diffusion, or both.
To confirm the location of F-MTX inside the infected RBC, parasites were incubated with F-MTX and visualized using confocal microscopy. After a 3-hour incubation period, F-MTX accumulated in the parasite cytosol, co-localizing with MitoTracker fluorescence that labels mitochondria, but not with the DAPI nuclear stain (Figure 2C). Confocal microscopy confirmed the same accumulation pattern observed in the flow cytometry assay, with Dd2 WT exhibiting significantly more F-MTX accumulation than Dd2Δpfmrp1 (Figure 2C). The fluorescence ratio of F-MTX/MitoTracker demonstrated significantly higher (P<0.0001) F-MTX accumulation in Dd2 WT parasites (Figure 2D).
However, a drug susceptibility assay showed that F-MTX does not retain the antimalarial activity of MTX, most likely due its bulkier structure preventing binding to PfDHFR, the target of MTX (Supplementary Figure 4). Since transport proteins are less specific in their substrate binding properties compared to the interaction of a substrate with the less accessible active site of an enzyme, F-MTX might be transported into the parasite cytoplasm via PfMRP1 but unable to bind PfDHFR and impair parasite growth.
Overall, both the flow cytometry and confocal microscopy data on the transport of F-MTX into infected RBCs point to a putative import function of PfMRP1.
Folate analogs compete for PfMRP1 transport
To further explore the propensity of PfMRP1 to transport antimalarial drugs, the F-MTX probe was used as a proxy of transport capacity. We performed flow cytometry competition assays and measured the fluorescence of F-MTX in infected RBCs, after 3 hours incubation periods with varying concentrations of different compounds and a fixed concentration of F-MTX. These assays tested CQ, MQ, DHA, and amodiaquine (AQ), as well as the antifolates MTX, AMT, PYR, TMP and MK571. Competition assays showed that increases in the MTX, AMT or MK571 concentration caused concentration-dependent decreases in F-MTX accumulation in Dd2 WT parasites but not in Δpfmrp1 parasites (Figure 3). These results demonstrate that F-MTX competes for the same transport pathway as MTX, AMT and MK571. This pathway is mediated by PfMRP1, as evidenced by the lack of competition in the Dd2Δpfmrp1 line (Figure 3). These assays further suggest that residual F-MTX in the Dd2Δpfmrp1 strain is probably the result of another transport mechanism or passive transport. The reduction in accumulation was more pronounced for MTX and AMT compared to MK571. This finding is likely related to MK571 acting as a weak chemical inhibitor of PfMRP1, as supported by the IC50 results in the presence of MK571 (Figure 1C), compared to PfMRP1 role in direct transport of MTX and AMT.
Folate impacts parasite growth
Parasite growth was monitored to understand the overall impact of Δpfmrp1. Regular levels of folate in human serum are around 6-20 µg/L and folate deficiency in RBCs is considered below 151 µg/L45. Standard RPMI medium used for in vitro culture contains 1 mg/L folate, which is 10 times higher than the limit of folate deficiency in RBCs and around 50 times higher than in human serum. Therefore, we modified an assay to limit exogenous folate in order to reduce parasite folate pools and evaluate parasite reliance on external folate sources, and hence PfMRP1 dependency38. Additionally, since Δpfmrp1 impacts folate-related transport, growth was evaluated in culture medium lacking an exogenous source of folate. Parasites were cultured for 14 days in medium depleted of folate, after which parasites were tightly synchronized at 0.05% parasitemia and growth was measured in regular medium, medium depleted of folate or medium containing 100 µg/L folate. Both 3D7Δpfmrp1 and Dd2Δpfmrp1 parasites were able to achieve high parasitemias like WT control parasites under regular culture conditions, although Dd2Δpfmrp1 displayed slightly less growth at high parasitemias (>5%) (Figure 4A). Without exogenous folate supplementation the Dd2 strain displayed similar growth in the first cycle but slower growth rate in the second cycle, with around 2% parasitemia compared to 6-8% of regular culture medium. Dd2 parasites were nonetheless able to reach parasitemias of around 8% one cycle later (Figure 4A). In contrast, 3D7 parasites displayed similar growth rates in medium with or without exogenous folate (Figure 4B). There was no difference in growth between WT and Δpfmrp1 strains in medium without exogenous folate (Figure 4A and 4B). To confirm that this was not an artifact due to parasites still having access to folate pools, cultures were diluted multiple times using RBCs from the same blood donor bag and growth was monitored in medium lacking folate46. Results showed no difference between WT and Δpfmrp1 growth for both strains in continuous culture in medium depleted of folate (Figure 4C and 4D). Overall, no differences in growth between WT and Δpfmrp1 were observable for both 3D7 and Dd2 strains.
The assay with medium containing 100 µg/L folate had results comparable to medium without folate. Addition of a source of exogenous folate partly recovered Dd2 growth compared to medium without folate, with parasites able to grow to 4% on the second cycle compared to 2% in medium without folate (Figure 4A). As with medium lacking folate, there was no impact on the 3D7 growth rate (Figure 4B). These results support the impact of exogenous folate on the growth rate of Dd2, which we observed to be considerably faster than 3D7 under our in vitro conditions. Accordingly, the Dd2 growth rate without folate supplementation resembled the 3D7 growth rate (Figure 4A and 4B). Although, PfMRP1 might impact parasite growth as observed for Dd2Δpfmrp1 cultured in standard medium, likely this impact is only observable under very rich in vitro growth conditions or as a result of fitness cost at high parasitemia as previously reported20. Furthermore, the effect of folate on the overall Dd2 growth is likely independent of PfMRP1 and mediated by alternative pathways.
Folic acid impacts the antifolate response
To understand the impact of folate on the antifolate resistance phenotype observed in Δpfmrp1 parasites, we performed 72-hour drug susceptibility assays under various folate conditions. IC50 values were measured for PYR, TMP, WR, AMT, MTX and the two non-antifolate based compounds ferroquine and DMS265, as controls. Assays were performed in the presence of diminished folate conditions, namely no exogenous folate, and a folate supplementation of 3 µg/L or 100 µg/L folic acid, the last being the folic acid concentration previously tested and that partly recovers Dd2 growth rate. Similar to growth assays (Figure 4) there was no impact on the overall growth of these strains in the 72-hour period of the assay. Results showed a slight but significant increase in susceptibility of Dd2Δpfmrp1 relative to WT parasites when assayed against PYR or TMP, relative to WT parasites, under conditions of no folate or 3 µg/L of folate, respectively, and 100 µg/L of folate (Figure 5). These results suggest that an inefficiency of Dd2Δpfmrp1 to import folate could render the drugs slightly more effective under limiting exogenous folate conditions, possibly due to lower intracellular dihydrofolate levels, the natural substrate of PfDHFR. Other folate conditions for PYR, TMP, and WR did not show significant differences but presented the same trend. An inefficiency to import folate could further explain the phenotype of the decreased growth of Dd2Δpfmrp1 under regular conditions (Figure 4)
The resistance phenotype of 3D7Δpfmrp1 and Dd2Δpfmrp1 parasites for MTX and AMT was maintained under all folate conditions (Figure 5). Surprisingly, the susceptibility to AMT was drastically reduced under lower folate concentrations, an effect that was even more pronounced for the 3 µg/L folic acid supplementation compared with no exogenous folate (Figure 5). A similar effect was observed for MTX, although with a less pronounced difference and evident only in the Dd2Δpfmrp1 and parental Dd2 lines (Figure 5). This phenotype might be multifactorial, as it seems to be independent of PfMRP1 for AMT, as the effect is observable also in WT parasites. A slight increase in IC50 was also observed for Dd2 WT at 3 µg/L folic acid for MTX (2-fold, P<0.01), additionally pointing to a PfMRP1-independent effect. These data, with Dd2 drug response being more affected by folate concentration agree with parasite growth assays, in which Dd2 was generally more impacted by folate concentrations.
Both MTX and AMT are expected to be competitors of folate for the PfDHFR binding site, and in this case a lower folate concentration was expected to render parasites more susceptible to drug action. However, the opposite was observed where less exogenous folate led to lowered susceptibility to the antifolate analogs MTX and AMT. This result suggests that this phenotype was likely independent of PfMRP1 and was instead due to another mechanism. Likewise, a similar phenotype was observed for antifolates that are structurally different from folate (PYR, TMP, WR) and even unrelated compounds that are not expected to act on the folate pathway (DSM265 and ferroquine) (Figure 5). Overall, this phenotype was only observed in the Dd2 strain, except for AMT where this was also observed in the 3D7 strain.
pfmrp1 is phylogenetically unrelated to other eukaryotic ABC transporters
ABC transporters are an ancient family present in a wide variety of species ranging from prokaryotes to eukaryotes. The results suggesting that PfMRP1 can function as an influx ABC transporter prompted us to investigate evolutionary relationships among PfMRP1 and other proteins from the same family and explore potential links between sequence similarity and function. We started by using the PfMRP1 amino acid sequence to BLASTp query the landmark database that includes a taxonomically diverse and non-redundant set of protein sequences. This resulted in 116 protein hits from 11 different species (Caenorhabditis elegans, Arabidopsis thaliana, Dictyostelium discoideum, Glycine max, Saccharomyces cerevisiae, Drosophila melanogaster, Danio rerio, Homo sapiens, Schizosaccharomyces pombe, Mus musculus and Leishmania donovani) (Supplementary dataset 2). Except for PfMRP2 that shared moderate sequence identity with PfMRP1 (49% identity and 64% query coverage), all hits from other organisms showed low identity (<25%). Phylogenetic analysis using these sequences placed PfMRP1 and PfMRP2 in an isolated branch in the phylogeny (Figure 6). These sequences were inferred to share a node, deep in the phylogeny, only with an ABC transporter from Leishmania donovani (XP_003863220.1) but with a low bootstrap support value. In contrast, several other ABC transporters from different species were found to be highly related to each other branching within well supported monophyletic clades. This finding suggests that although PfMRP1 harbors transmembrane helices typically found in the ABC transporters that function as exporters, it has a distinct protein sequence and origin. This finding is compatible with the hypothesis that PfMRP1 might have an atypical transport function.