Expression and functional characterization of human P-gp and PfMDR1 in Xenopus oocytes

Since human transporters are relatively straightforward to express in oocytes [53], human P-gp was first expressed in Xenopus oocytes to optimize the expression conditions and to develop a transport assay. The optimized methods were then applied to PfMDR1. Direct measurements of transport via human P-gp were undertaken by microinjecting a tritiated version of vinblastine ([³H]vinblastine; a known substrate of human P-gp) into the oocyte and measuring its efflux (Fig 1A). Given that the relationship between the quantity of cRNA microinjected into the oocyte and the level of [³H]vinblastine transport measured was approximately linear between 2.5 and 10 ng (Fig 1B), all subsequent experiments used oocytes microinjected with 10 ng of human P-gp cRNA. Next, we determined that the transport of [³H]vinblastine via human P-gp was approximately linear with time for at least 2 hours (Fig 1C). Hence, all of the subsequent transport measurements were made at 1.5 hours. To confirm that the [³H]vinblastine transport detected in these assays was mediated by human P-gp (and was not due to the nonspecific leakage of [³H]vinblastine from the oocyte), compounds known to interact with the transporter (nicardipine [46,54], PSC833 [55,56], vanadate [57,58], and verapamil [46,59]) were shown to inhibit [³H]vinblastine efflux (Fig 1D and S2 Fig). Furthermore, [³H]vinblastine transport was unaffected by amantadine—a drug that human P-gp does not interact with [46–48] (Fig 1D). Finally, we showed that other known substrates of human P-gp, such as rhodamine B, quinacrine, and methylene blue, are also transported by human P-gp in our oocyte system (S1 Data). The negative controls used in these experiments were nonexpressing oocytes (ne) and those expressing an unrelated transporter, the Plasmodium falciparum nucleoside transporter 1 (PfNT1; S1 Fig) [60,61]. The low level leakage of [³H]vinblastine from the negative control oocytes is most likely due to the simple diffusion of the neutral species of the drug (Fig 1C and 1D). Together, these datasets confirmed that we had achieved the functional expression of human P-gp in Xenopus oocytes, with the optimization process yielding relatively high rates of transport for vinblastine (304 ± 10 fmol/oocyte/h) and for other known substrates of human P-gp (S1 Data), as well as a high transport signal-to-background ratio (approximately 28-fold).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Heterologous expression of functional human P-gp and PfMDR1 in Xenopus oocytes.

(a) Schematic showing the transport of a [³H]drug via human P-gp or PfMDR1 in the Xenopus oocyte system. (b) The relationship between the quantity of human P-gp cRNA injected into the oocyte and the level of human P-gp–mediated [³H]vinblastine transport measured. (c) The transport of [³H]vinblastine via human P-gp was approximately linear with time for at least 2 hours. (d) The human P-gp–mediated efflux of [³H]vinblastine from oocytes was reduced by known inhibitors of the transporter (nicardipine, PSC8333, vanadate, and verapamil) and was unaffected by amantadine (a drug that does not interact with human P-gp). (e) The field isoforms of PfMDR1 characterized in this study. Residues that differ from the wild-type amino acid sequence are shaded gray. (f) Immunofluorescence microscopy images confirmed that PfMDR1 localized to the oocyte plasma membrane. The expression of 3xHA-tagged PfMDR1^NYSND resulted in a fluorescent band external to the pigment layer, indicating that the protein was expressed at the oocyte surface. The band was not present in ne. The length of the scale bar is 50 μm. Images showing the presence of the other HA-tagged PfMDR1 isoforms at the oocyte surface are shown in S4 Fig. (g) Semiquantification of PfMDR1 protein levels in the membranes of oocytes expressing different 3xHA-PfMDR1 isoforms indicated that the 7 isoforms were expressed at similar levels. (h) The transport of [³H]vinblastine via PfMDR1. (i) The efflux of [³H]lumefantrine from oocytes expressing PfMDR1^NYSND was reduced by nicardipine, PSC8333, vanadate, and verapamil and was unaffected by amantadine. (j) The relationship between the quantity of PfMDR1^NYSND cRNA microinjected into the oocyte and the level of PfMDR1^NYSND-mediated [³H]lumefantrine transport measured. (k) The transport of [³H]lumefantrine via PfMDR1^NYSND was approximately linear with time for at least 2 hours. (l) Microinjection of additional ATP into the oocyte greatly stimulated [³H]lumefantrine transport via PfMDR1^NFCDY. The data are the mean of n = 4 to 9 independent experiments, each yielding similar results and overlaid as individual data points in panels d, g, h, and i, and the error is the SEM. Where not visible, the error bars fall within the symbols. The asterisks denote a significant difference from human P-gp (panel d), 3xHA-PfMDR1^NYSND (panel g), or PfMDR1^NYSND (panels h and i); ***P < 0.001, ns, not significant (1-way ANOVA). The data underlying this figure is supplied in S3 Data. ne, nonexpressing oocytes; PfMDR1, Plasmodium falciparum multidrug resistance protein 1; P-gp, P-glycoprotein.

https://doi.org/10.1371/journal.pbio.3001616.g001

The optimized method and assays were then applied to wild-type PfMDR1 (PfMDR1^NYSND), 5 field mutant isoforms of PfMDR1 (PfMDR1^NFSDD, PfMDR1^YYSND, PfMDR1^FYSND, PfMDR1^YFSND, and PfMDR1^NFCDY), and a catalytically inactive version of PfMDR1^NYSND (Fig 1E, S1 Table, S3 Fig, and S1 File). The latter contains mutations in the protein’s nucleotide-binding domains (NBDs) that impede ATP hydrolysis. Immunofluorescence assays confirmed the presence of hemagglutinin (HA)-tagged versions of each isoform at the oocyte plasma membrane (Fig 1F and S4 Fig). Furthermore, PfMDR1 was shown to adopt the same orientation in the oocyte plasma membrane as it does in the membrane of the parasite’s DV [36], with its amino terminus and carboxyl terminus located in the cytosol (S4 Fig). Semiquantitative western blot analyses and densitometric evaluations of total protein confirmed that each of the PfMDR1 isoforms were expressed at comparable levels in the oocyte membrane (Fig 1G, S5 Fig, and S1 Text). Hence, any differences in drug transport activity between these PfMDR1 isoforms can be attributed to differences in their transport properties (and the mutations they carry), rather than to differences in expression levels.

All 6 field isoforms of PfMDR1 transported [³H]vinblastine, albeit to varying degrees (Fig 1H). The catalytically inactive version of PfMDR1^NYSND exhibited a very low level of [³H]vinblastine transport activity, consistent with the mutations in the NBDs hampering ATP hydrolysis. The ability of PfMDR1^NYSND to transport [³H]lumefantrine was then tested in the absence or presence of human P-gp inhibitors (Fig 1I). The PfMDR1^NYSND-mediated transport of [³H]lumefantrine was inhibited by nicardipine, PSC833, vanadate, and verapamil, but was unaffected by amantadine (Fig 1I and S2 Fig). The relationship between the quantity of PfMDR1^NYSND cRNA microinjected into the oocyte and the level of [³H]lumefantrine transport achieved was very similar to that obtained for [³H]vinblastine transport via human P-gp (with 10 ng of PfMDR1^NYSND cRNA falling within the linear range) (Fig 1J) and the transport of [³H]lumefantrine via PfMDR1^NYSND was also approximately linear with time for at least 2 hours (Fig 1K). Furthermore, as was the case for human P-gp, a high rate of vinblastine transport (275 ± 3 fmol/oocyte/h) and a high transport signal-to-background ratio (approximately 29-fold) were obtained for PfMDR1^NYSND. The microinjection of ATP into the oocyte greatly stimulated [³H]lumefantrine transport via PfMDR1^NFCDY but had a milder effect on PfMDR1^NYSND (Fig 1L). This is consistent with previous reports that D1246Y (located in NBD 2 of PfMDR1^NFCDY) impedes ATP hydrolysis [39] and that PfMDR1^NFCDY has a lower basal ATPase activity than the wild-type protein [39,40].

Field isoforms of PfMDR1 differ significantly in their capacities for antimalarial drug transport

Our achievement of a robust expression system for PfMDR1 enabled direct characterizations of its function and of how this varies between different isoforms of the protein. We found that the 6 field isoforms of PfMDR1 exhibited varying capacities for the transport of a broad range of drugs and related compounds (Figs 2–4, S1 Data, and S2 Text). All of the antimalarial drugs tested—lumefantrine, mefloquine, chloroquine, quinine, quinidine, amodiaquine, piperaquine, dihydroartemisinin, methylene blue, and quinacrine—were shown to be substrates of the PfMDR1 isoforms included for study. The PfMDR1 isoforms also transported vinblastine and rhodamine B (known substrates of human P-gp), but none of the proteins transported amantadine (Fig 2 and S1 Data). The rates of antimalarial drug transport via PfMDR1^NYSND (in fmol/oocyte/h) were as follows: lumefantrine (731 ± 22), mefloquine (657 ± 23), amodiaquine (552 ± 11), chloroquine (531 ± 18), piperaquine (491 ± 17), quinidine (439 ± 11), quinine (432 ± 13), quinacrine (356 ± 11), methylene blue (333 ± 10), and dihydroartemisinin (293 ± 10). The low level of drug efflux observed from the negative control oocytes was most likely due to the simple diffusion of the neutral drug species.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Field isoforms of PfMDR1 differ significantly in their capacities for antimalarial drug transport.

(a–i) Field isoforms of PfMDR1 transported the antimalarial drugs [³H]lumefantrine (a), [³H]mefloquine (b), [³H]chloroquine (c), [³H]quinidine (d), [³H]amodiaquine (e), [³H]piperaquine (f), [³H]dihydroartemisinin (g), methylene blue (h), and quinacrine (i). (j–l) All of the field isoforms of PfMDR1 also transported the human P-gp substrates rhodamine B (j) and [³H]vinblastine (k), but none transported [³H]amantadine (l). The transport of methylene blue, quinacrine, and rhodamine B was detected using the intrinsic fluorescence of these compounds and a fluorescence-based transport assay (see S2 Text). The data are the mean of n = 4 independent experiments (each yielding similar results and overlaid as individual data points), and the error is the SEM. The asterisks denote a significant difference from PfMDR1^NYSND; *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant (1-way ANOVA). The data underlying this figure is supplied in S3 Data. ne, nonexpressing oocytes; PfMDR1, Plasmodium falciparum multidrug resistance protein 1.

https://doi.org/10.1371/journal.pbio.3001616.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Characterization of lumefantrine, mefloquine, and chloroquine transport via PfMDR1.

(a–c) There were significant differences in the apparent kinetic parameters of lumefantrine (a), mefloquine (b), and chloroquine (c) transport between field isoforms of PfMDR1. The concentration dependence of PfMDR1-mediated drug transport was calculated by subtracting the leakage from ne from that of oocytes expressing a PfMDR1 isoform at each drug concentration. (d–f) The transport of [³H]vinblastine via PfMDR1 was inhibited by lumefantrine (d), [³H]mefloquine (e), and [³H]chloroquine (f). The data are the mean of n = 4 independent experiments (each yielding similar results), and the error is the SEM. The asterisks denote a significant difference from PfMDR1^NYSND; *P < 0.05, **P < 0.01, ***P < 0.001, ns nonsignificant (1-way ANOVA). The data underlying this figure is supplied in S3 Data. ne, nonexpressing oocytes; PfMDR1, Plasmodium falciparum multidrug resistance protein 1.

https://doi.org/10.1371/journal.pbio.3001616.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Quinine transport via PfMDR1.

(a) Field isoforms of PfMDR1 transported [³H]quinine, with PfMDR1^NFSDD exhibiting the highest level of quinine transport activity. (b) There were significant differences in the apparent kinetic parameters for quinine transport via field isoforms of PfMDR1. The concentration dependence of PfMDR1-mediated quinine transport was calculated by subtracting the leakage from ne from that of oocytes expressing a PfMDR1 isoform at each drug concentration. The data are the mean of n = 4 independent experiments (each yielding similar results and overlaid as individual data points in panel a), and the error is the SEM. The asterisks denote a significant difference from PfMDR1^NYSND; **P < 0.01 and ***P < 0.001 (1-way ANOVA). The data underlying panels a and b of this figure is supplied in S3 Data. (c) Inward-open homology model of PfMDR1 based on the C. elegans P-gp crystal structure (PDB 4F4C) showing the location of the mutations found in the 5 field PfMDR1 isoforms characterized in this study (orange). A putative binding pose of quinine in the central cavity is shown (pink). (d) A comparison of the inward-open model and outward-open models (based on the crystal structure of human P-gp; PDB 6C0V) as viewed from the side of the protein facing into the DV lumen. The N86Y mutation is part of a cluster of 3 residues forming the extracellular gate of the transporter (the participating residues are shown). These 3 residues are in close proximity in the inward-open state but move apart in the outward-open conformation. (e) Putative quinine binding site in PfMDR1^NFSDD. Amino acids interacting with quinine are indicated (apart from L71, I1071, and F1072 that are removed to clearly view the binding pose). Atoms are shaded as follows: carbon in quinine, pink; carbon in the 1042D residue, orange; nitrogen, blue; oxygen, red; hydrogen, white. ne, nonexpressing oocytes; PfMDR1, Plasmodium falciparum multidrug resistance protein 1; P-gp, P-glycoprotein.

https://doi.org/10.1371/journal.pbio.3001616.g004

In almost all cases, the introduction of mutations into PfMDR1^NYSND caused a significant reduction in drug transport activity. The most striking example of this observation was PfMDR1^NFCDY, which had the lowest level of transport activity for all of the compounds tested (Figs 2–4 and S1 Data). The rates of antimalarial drug transport via PfMDR1^NFCDY (in fmol/oocyte/h) and the corresponding fold change relative to drug transport via PfMDR1^NYSND were as follows: lumefantrine (299 ± 23, approximately 0.4-fold), chloroquine (286 ± 11, approximately 0.5-fold), piperaquine (223 ± 9, approximately 0.5-fold), quinidine (191 ± 8, approximately 0.4-fold), quinine (188 ± 10, approximately 0.4-fold), mefloquine (175 ± 7, approximately 0.3-fold), amodiaquine (155 ± 9, approximately 0.3-fold), quinacrine (121 ± 4, approximately 0.3-fold), methylene blue (108 ± 3, approximately 0.3-fold), and dihydroartemisinin (92 ± 3, approximately 0.3-fold).

No 2 drugs produced the same profile of transport activities across the 6 field isoforms of PfMDR1. There were, however, strong similarities between some of the datasets (see S3 Text for an extended analysis). For example, commonalities were evident within the following groups (1) lumefantrine and mefloquine; (2) chloroquine and piperaquine; (3) mefloquine, quinine, quinidine, and dihydroartemisinin; and (4) lumefantrine, methylene blue, and quinacrine (Figs 2 and 4, and S3 Text).

Characterization of lumefantrine, mefloquine, and chloroquine transport via PfMDR1

Having identified that the 6 field isoforms of PfMDR1 exhibit different capacities for the transport of lumefantrine, mefloquine, and chloroquine (Fig 2A–2C), we characterized the interactions of these 3 drugs with the transporter in more detail. A kinetic analysis of the PfMDR1-mediated transport of lumefantrine, mefloquine, and chloroquine revealed that the apparent Michaelis constant (app K_m) and the apparent maximal velocity (app V_max) values decreased in the order PfMDR1^NYSND ≥ PfMDR1^NFSDD > PfMDR1^YYSND > PfMDR1^NFCDY (Fig 3A–3C). Hence, PfMDR1^NFCDY has both the lowest app K_m and the lowest app V_max for all 3 antimalarials. PfMDR1^NFCDY is the only isoform included in this study that contains D1246Y. This mutation is located upstream of the Q-loop within NBD 2, which is thought to be involved in coordinating Mg²⁺ for ATP hydrolysis and/or in facilitating conformational changes of the NBDs in human P-gp [62–65]. Thus, it is possible that D1246Y disrupts ATP hydrolysis and/or the coupling of ATP hydrolysis to drug transport in PfMDR1, and that this contributes to the lower drug transport activity exhibited by PfMDR1^NFCDY.

A subset of experiments measuring the abilities of lumefantrine, mefloquine, and chloroquine to inhibit [³H]vinblastine transport via PfMDR1^NYSND and PfMDR1^NFCDY was performed to ascertain whether these drugs are potent inhibitors of the transporter. None of the antimalarials were potent inhibitors of transport via either isoform of the protein, with all 3 drugs producing IC₅₀s that are 20- to 11-fold higher than those obtained with nicardipine (Fig 3D–3F and S2 Fig). Moreover, in all cases, the antimalarials were slightly more effective in inhibiting transport via the mutant field isoform. This is not the trend that would be expected if these drugs exerted an antimalarial effect against wild-type PfMDR1 (e.g., by blocking the transport of its (currently unknown) natural substrates) that is alleviated by the introduction of mutations into the transporter.

Quinine transport via PfMDR1

Although the addition of mutations to PfMDR1^NYSND reduced drug transport activity in most cases, there were 2 exceptions to this trend. The introduction of the Y184F and N1042D mutations—yielding PfMDR1^NFSDD—increased the protein’s capacity for the transport of both quinine and its stereoisomer quinidine (Figs 2D and 4A). A kinetic analysis of quinine transport via PfMDR1 revealed several valuable insights (Fig 4B). First, the range of app K_m values obtained for quinine transport via the different PfMDR1 isoforms (5.5 ± 0.1 μM to 21.9 ± 0.9 μM) was similar to those observed for the transport of lumefantrine, mefloquine, and chloroquine (e.g., the lumefantrine app K_m values ranged from 2.9 ± 0.03 μM to 21.9 ± 0.9 μM). By contrast, the app V_max values for quinine transport (0.4 ± 0.01 to 5.0 ± 0.2 pmol/oocyte/h) were lower than those obtained for the other 3 drugs (e.g., the lumefantrine app V_max values were 3.7 ± 0.1 to 10.3 ± 0.4 pmol/oocyte/h). In general terms, these observations indicate that quinine occupies the transporter’s binding cavity for a much longer period than lumefantrine, mefloquine, or chloroquine. It is possible that in the parasite, this relatively low rate of quinine translocation impedes the transport of the natural substrates of PfMDR1.

The second key observation from the kinetic analysis of quinine transport was the finding that PfMDR1^NFSDD has a higher app V_max (5.0 ± 0.2 pmol/oocyte/h versus 3.4 ± 0.3 pmol/oocyte/h) as well as a slightly higher app K_m (21.9 ± 0.9 μM versus 19.1 ± 0.2 μM) than the wild-type protein (Fig 4A and 4B). We sought a structural explanation for this unusual result by generating a homology model of PfMDR1 to determine how quinine interacts with the transporter’s binding cavity.

Two homology models of PfMDR1 have previously been generated using either the bacterial MsbA protein or mouse P-gp as the template sequences [66,67]. However, the low sequence identity between PfMDR1 and these sequences (26% to 29%) reduces the versatility of these homology models. We identified the recently elucidated structures of Caenorhabditis elegans P-gp [68] and human P-gp [69], which have slightly higher sequence identity (33%), as better templates for inward-open and outward-open homology models of PfMDR1 (S1 File), although structural predictions should still be made with caution at this level of sequence identity. The locations of the 5 field mutations are depicted in the inward-open model of PfMDR1 (Fig 4C). Consistent with previous reports, both S1034 and N1042 line the central cavity (where they could interact with substrates), Y184 faces into the membrane, and D1246 is located in NBD 2 of the protein [66,67]. However, contrary to previous models that suggested that N86 forms part of the substrate-binding cavity [66,67], our inward-open model proposes that N86 is located on the side of the protein that faces into the DV lumen, forming part of the gate and obscuring the central cavity of the protein from this compartment (Fig 4D). A comparison with the outward-open model shows that the residues in this gate (N86, F818, and D1061) move far apart to expose the substrate-binding cavity to the DV lumen (Fig 4D). Thus, it is plausible that mutation of N86 could alter these interactions and consequently stabilize the inward-open conformation.

Since the Y184F mutation in PfMDR1^NFSDD protrudes into the membrane (rather than into the binding cavity), it is unlikely to be involved in coordinating quinine. Thus, it is more likely that the N1042D mutation is the cause of the enhanced quinine transport activity exhibited by PfMDR1^NFSDD. To investigate this hypothesis, we used molecular dynamics simulations of PfMDR1 containing N1042D to examine the likely binding poses of quinine in the binding cavity. Seven simulations were performed, each with a different random starting position and orientation of quinine inside the substrate-binding cavity. In 4 of the 7 simulations, quinine quickly moved to interact with 1042D (within 20 ns; S6 Fig). This interaction with the negatively charged carboxylate oxygen of the aspartate residue occurred through the protonated nitrogen on the cyclic amine ring of quinine. Indeed, quinine adopted a very stable position in a predominantly hydrophobic pocket in which 3 of the 4 polar atoms on the molecule found hydrogen bonding partners on the protein (stabilized by the nitrogen on the cyclic amine ring, as well as the central hydroxyl group on the quinine molecule forming hydrogen bonds with 1042D, in addition to the ether oxygen of quinine forming a hydrogen bond with Y810; Fig 4E). Thus, the stability of the putative quinine binding site (as determined in our simulations) could plausibly represent an alternative binding site for the drug in PfMDR1^NFSDD that is not present in the other isoforms. This could alter the rate of quinine transport by influencing the rate of the protein conformational change or, perhaps less likely, ligand unbinding.

Correlations between rates of drug transport via PfMDR1 in Xenopus oocytes and in vitro drug resistance indices

We used the PfMDR1-mediated rates of lumefantrine, mefloquine, chloroquine, and quinine transport obtained in the Xenopus oocyte system (S1 Data) to determine relationships between the drug transport capacities of the different PfMDR1 field isoforms and the in vitro drug response of parasite strains that express these proteins. In vitro drug resistance indices were calculated using IC₅₀ values collated from the literature [16,23,24,26,70–86] (S3 Table) and then plotted against the PfMDR1-mediated transport rate for the relevant drug and parasite strain. We found strong positive correlations for lumefantrine (Pearson correlation coefficient = 0.76; Fig 5A and S3 Text) and mefloquine (Pearson correlation coefficient = 0.75; Fig 5B and S3 Text), and a strong negative correlation for chloroquine (Pearson correlation coefficient = −0.86; Fig 5C and S3 Text). No correlation was found between the rate of quinine transport via PfMDR1 and the in vitro quinine response (Fig 5D and S4 Text).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Correlations between rates of drug transport via PfMDR1 in Xenopus oocytes and in vitro parasite drug resistance indices.

The rates of PfMDR1-mediated transport of lumefantrine, mefloquine, chloroquine, and quinine (Figs 2 and 4, and S1 Data) were plotted against the in vitro resistance index for the relevant drug and parasite strain (S3 Table). Where not shown, error bars fall within the symbols. (a) A positive correlation was observed between the rate of lumefantrine transport via PfMDR1 and the in vitro lumefantrine response (Pearson correlation coefficient = 0.76, r² = 0.58, P = 0.078). (b) A positive correlation was observed between the rate of mefloquine transport via PfMDR1 and the in vitro mefloquine response (Pearson correlation coefficient = 0.75, r² = 0.56, P = 0.008). The data point for Dd2 is an outlier and its removal significantly improved the correlation (Pearson correlation coefficient = 0.97, r² = 0.94, P = 0.006). Dd2 parasites typically harbor 2 to 4 copies of PfMDR1 and have been reported to express greater levels of PfMDR1 than the other parasite strains included in the analysis [85,87]. Given that pfmdr1 amplification has been associated with mefloquine resistance [30–32,88,89], it is possible that the overexpression of PfMDR1 in Dd2 parasites imparts a higher level of resistance to this strain (see S3 Text for an extended analysis). (c) An inverse correlation was observed between the rate of chloroquine transport via PfMDR1 and the in vitro chloroquine response. The removal of the 7G8 data point (see S3 Text for an extended analysis) resulted in a stronger correlation between the rate of chloroquine transport via PfMDR1 and the parasite’s chloroquine response in vitro (Pearson correlation coefficient = −0.86, r² = 0.74, P = 0.054). (d) There was no correlation between the rate of quinine transport via PfMDR1 and the in vitro quinine response (see S4 Text for an extended analysis). The data for the in vitro resistance indices are the mean of n = 2 to 18 published studies, and the data for the PfMDR1-mediated drug transport rates are the mean of n = 4 independent experiments. The error is the SEM except for the in vitro resistance indices that were calculated from 2 studies, in which case the error is the range/2. The data underlying this figure is supplied in S3 Data. PfMDR1, Plasmodium falciparum multidrug resistance protein 1.

https://doi.org/10.1371/journal.pbio.3001616.g005

Transport of lumefantrine and mefloquine via PfCRT in Xenopus oocytes and in situ

We sought to elucidate the role of PfCRT in modulating the parasite’s response to lumefantrine and mefloquine by measuring its ability to transport these drugs in Xenopus oocytes (see S2 Table and S3 Fig for the positions of the mutations in the PfCRT isoforms studied here). Wild-type PfCRT (PfCRT^3D7) did not transport lumefantrine or mefloquine, whereas 2 chloroquine resistance-conferring isoforms of PfCRT (PfCRT^7G8 and PfCRT^Dd2) exhibited significant transport of both drugs (Fig 6A and 6D and S7 Fig). The transport of lumefantrine and mefloquine via PfCRT^7G8 and PfCRT^Dd2 was inhibited by compounds known to interact with the transporter, including verapamil [6,18], chlorpheniramine [90], and saquinavir [10,91], but was unaffected by histidine and the dipeptide LH (neither of which interact with PfCRT [10,91]) (Fig 6A and 6D, and S7 Fig). Lumefantrine and mefloquine transport via PfCRT^7G8 and PfCRT^Dd2 was saturable, with PfCRT^Dd2 having a slightly lower affinity and a higher capacity for drug transport relative to PfCRT^7G8 (Fig 6B and 6E). When compared with chloroquine transport via PfCRT^Dd2 (K_m = 232 ± 11 μM; V_max = 61 ± 6 pmol/oocyte/h [7]), the transport of lumefantrine and mefloquine is a high-affinity, low-capacity process.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Transport of lumefantrine and mefloquine via PfCRT in Xenopus oocytes and in situ.

(a, d) The transport of [³H]lumefantrine (a) and [³H]mefloquine (d) via PfCRT^Dd2 and PfCRT^7G8 was reduced by verapamil (VP; 100 μM). The asterisks denote a significant difference from the ne (gray asterisks), the PfCRT^7G8 control (orange asterisks), or the PfCRT^Dd2 control (red asterisks). (b, e) The transport of lumefantrine (b) and mefloquine (e) via PfCRT^Dd2 and PfCRT^7G8 was saturable. The asterisks denote a significant difference from PfCRT^Dd2 (red asterisks). (c, f) The effects of unlabeled lumefantrine (c) and unlabeled mefloquine (f) on [³H]VDPVNF transport via PfCRT^3D7, PfCRT^Dd2, and PfCRT^7G8. The asterisks denote a significant difference from PfCRT^3D7 (blue asterisks). (g) Lumefantrine and mefloquine (2.5 μM) increased the rate of DV alkalinization in the chloroquine-resistant C6^7G8 and C4^Dd2 parasite lines but not in the chloroquine-sensitive C2^GC03 line. Unless labeled ns, P < 0.05 relative to the absence of a test solute. (h) The increase in the rate of DV alkalinization caused by lumefantrine and mefloquine was inhibited by VP (50 μM). The asterisks denote a significant difference from the relevant C2^GC03 (blue asterisks), C6^7G8 (orange asterisks), or C4^Dd2 (red asterisks) treatments in the absence of verapamil. (i, j) Lumefantrine (i) and mefloquine (j) increased the rate of DV alkalinization in a concentration-dependent manner. The data are the mean of n = 4 independent experiments, each yielding similar results and overlaid as individual data points in panels a, d, g, and h, and the error is the SEM. Where not visible, the error bars fall within the symbols. **P < 0.01, ***P < 0.001, ns, not significant (1-way ANOVA). The data underlying this figure is supplied in S3 Data. ne, nonexpressing oocytes; PfCRT, Plasmodium falciparum chloroquine resistance transporter.

https://doi.org/10.1371/journal.pbio.3001616.g006

Lumefantrine and mefloquine were shown to inhibit the transport of a natural substrate of PfCRT (the hemoglobin-derived hexapeptide VDPVNF [10]) via all 3 isoforms (Fig 6C and 6F). However, all of the resulting IC₅₀s were above 45 μM, which is substantially higher than the IC₅₀s previously obtained for the quinine dimer “Q₂C” (IC₅₀s = 0.01 to 0.6 μM)—a drug that has been established as exerting an antiplasmodial effect via its potent inhibition of PfCRT’s natural function [10,92]. The lumefantrine and mefloquine IC₅₀s are also significantly higher than those observed for 2 other inhibitors of PfCRT—verapamil (IC₅₀s = 6.8 to 17.5 μM) and chlorpheniramine (IC₅₀s = 5.0 to 9.8 μM)—but are very similar to the IC₅₀s obtained with chloroquine (e.g., 110 ± 5.9 μM against PfCRT^3D7 and 53 ± 2.8 μM against PfCRT^Dd2) [10]. It seems unlikely, therefore, that lumefantrine and mefloquine possess significant anti-PfCRT activity.

We then tested whether PfCRT^7G8 and PfCRT^Dd2 transport lumefantrine and mefloquine within their native environment in the DV membrane of live parasites. The transport of lumefantrine and mefloquine via PfCRT was measured in situ using the “H⁺-efflux assay” with a set of P. falciparum transfectants—the C2^GC03, C6^7G8, and C4^Dd2 lines—that are isogenic except for the pfcrt allele. These lines express either PfCRT^3D7 (C2^GC03), PfCRT^7G8 (C6^7G8), or PfCRT^Dd2 (C4^Dd2) [17]. Lumefantrine and mefloquine enter the DV via simple diffusion of the neutral species and also via PfMDR1 (Figs 2 and 3), where they become protonated in the acidic lumen. The H⁺-efflux assay uses a fluorescent pH-sensitive probe to provide an indirect method of detecting the efflux of protonated drugs from the DV via PfCRT, which manifests as an increase in the rate of DV alkalinization. The positive controls were chloroquine and saquinavir (a peptide mimic that is a substrate of all 3 PfCRT isoforms [10]). Consistent with previous reports [9,10,20–22], chloroquine increased the rate of DV alkalinization in the C6^7G8 and C4^Dd2 lines and was without effect in the C2^GC03 line, whereas saquinavir increased the rate of DV alkalinization in all 3 parasite lines (Fig 6G). We found that lumefantrine and mefloquine induced a H⁺ leak in the C6^7G8 and C4^Dd2 lines, but not in the C2^GC03 line. These observations are consistent with both drugs being substrates of PfCRT^7G8 and PfCRT^Dd2, but not PfCRT^3D7 (Fig 6G–6J). Furthermore, verapamil inhibited the lumefantrine- and mefloquine-induced H⁺ leaks (Fig 6H), providing further evidence that the drug-induced efflux of H⁺ from the DV was mediated by PfCRT^7G8 and PfCRT^Dd2. Both drugs increased the rate of DV alkalinization in a concentration-dependent manner, with PfCRT^Dd2 again exhibiting a higher capacity for lumefantrine and mefloquine transport than PfCRT^7G8 (Fig 6I and 6J). Together, these findings support the datasets obtained with the Xenopus oocyte system by providing an in situ demonstration that lumefantrine and mefloquine accumulate within the DV and that chloroquine resistance-conferring isoforms of PfCRT transport both of these drugs back out into the parasite’s cytosol.

If lumefantrine and mefloquine exert their primary antimalarial activities on targets outside of the DV, the PfCRT-mediated efflux of these drugs from the DV of the C6^7G8 and C4^Dd2 lines should heighten parasite sensitivity to lumefantrine and mefloquine, and verapamil should reduce this effect. We tested this hypothesis by measuring the susceptibility of the isogenic parasite lines to lumefantrine, mefloquine, and chloroquine in the presence or absence of verapamil. The resulting IC₅₀s confirmed that the C6^7G8 and C4^Dd2 lines are more susceptible to lumefantrine and mefloquine than the C2^GC03 parasites (Table 1), a finding that is consistent with previous reports [2,17]. Furthermore, verapamil decreased the sensitivity of the C6^7G8 and C4^Dd2 lines to lumefantrine and mefloquine, while simultaneously increasing their sensitivity to chloroquine. Together, our in situ datasets show that chloroquine resistance-conferring isoforms of PfCRT can mediate the transport of lumefantrine and mefloquine from the DV into the parasite cytosol and that this “gain of transport function” underlies the heightened sensitivity of chloroquine-resistant parasites to lumefantrine and mefloquine.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. In vitro antiplasmodial activities of chloroquine, lumefantrine, and mefloquine against P. falciparum pfcrt transfectant lines.

https://doi.org/10.1371/journal.pbio.3001616.t001

Xenopus laevis frogs

Ethical approval of the work performed with female X. laevis frogs was obtained from the Australian National University Animal Experimentation Ethics Committee (Animal Ethics Protocol Numbers A2013/13 and A2019/26) in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. The frogs were purchased from Nasco USA (catalog no. LM00535M) and were housed in the Xenopus Frog Facility (at the Australian National University Research School of Biology) in compliance with the relevant institutional and Australian Government regulations. X. laevis frogs are fully aquatic animals held in large tanks (8 to 20 frogs per tank). The water is percolated and filtered in a closed system, and the water temperature is maintained at 19 to 20°C. The pH, hardness, ammonia, and other water parameters were checked daily and kept at levels reported to be necessary for the production of healthy oocytes [103,104].

Preparation of coding sequences and cRNA synthesis

The coding sequence of human P-gp was purchased from GenScript (catalog no. OHu23307; NM_000927; New Jersey USA) and inserted into the pGEM-He-Juel oocyte expression vector [105]. The PfMDR1^NYSND coding sequence was codon-harmonized to facilitate the correct folding, and thus the functional expression, of the transporter in Xenopus oocytes. This codon-harmonized sequence was custom synthesized and then inserted into the pGEM-He-Juel oocyte expression vector by GenScript. The remaining isoforms of PfMDR1, as well as the version of PfMDR1^NYSND carrying a double carboxyl-terminal HA tag (PfMDR1^NYSND-2xHA), were generated from the PfMDR1^NYSND template using site-directed mutagenesis [7] and the primers listed in S4 Table. The resulting PfMDR1 coding sequences were verified by sequencing (undertaken by the ACRF Biomolecular Resource Facility, Australian National University). Amino-terminally HA-tagged versions of these PfMDR1 variants, as well as the version of PfMDR1^NYSND that carries a triple HA tag at the carboxyl terminus (PfMDR1^NYSND-3xHA), were generated by GenScript.

The oocyte expression vectors containing PfNT1, PfCRT^3D7, PfCRT^Dd2, or PfCRT^7G8 were made previously [6,7,106]. The PfCRT coding sequences were codon-harmonized and encode a version of PfCRT that is free of endosomal–lysosomal trafficking motifs, thereby resulting in the expression of the transporter at the plasma membrane of Xenopus oocytes [6,7].

The plasmids containing human P-gp, PfMDR1, PfCRT, or PfNT1 coding sequences were linearized with SalI (Thermo Fisher Scientific, Massachusetts, USA; catalog no. ER0642), and 5′-capped cRNA was synthesized using the mMESSAGE mMACHINE T7 transcription kit (Ambion, Texas, USA; catalog no. AM1344). The in vitro transcription reactions were conducted as per the manufacturer’s instructions with the following minor modification for the synthesis of human P-gp and PfMDR1 cRNA: in a 40 μL reaction, 500 ng of linearized template plasmid was used, and 30 nmol GTP was added halfway into the 2-hour incubation. The resulting cRNA samples were purified with the MEGAclear kit (Ambion, catalog no. AM1908) and adjusted to the desired concentration using RNase-free elution buffer. The quality of the cRNA was assessed via agarose gel electrophoresis.

Isolation and microinjection of Xenopus oocytes

Adult female frogs were anaesthetized in a solution of ethyl 3-aminobenzoate methanesulfonate salt (Sigma-Aldrich, Missouri, USA; catalog no. A5040) and 1 mM NaHCO₃ (Sigma-Aldrich, catalog no. S5761) [8]. Sections of the ovary were removed from the frog using surgical grade forceps and scissors. The ovary sections were cut into pieces containing approximately 20 oocytes, transferred to a 100 mL Erlenmeyer flask, and then gently rinsed 5 times with “calcium-free oocyte ringer” buffer (OR2⁻; 82.5 mM NaCl, 2.5 mM KCl, 1 mM MgCl₂, 1 mM Na₂HPO₄, 5 mM HEPES; pH 7.8). The collagenous membrane encasing the oocytes was degraded by incubating the ovary sections in 25 mL of OR2⁻ buffer supplemented with 0.5 M Na₂HPO₄ (Sigma-Aldrich, catalog no. 342483), bovine serum albumin (250 mg/mL; Sigma-Aldrich, catalog no. A7030), and collagenase B (Roche, Basel, Switzerland; catalog no. 11088831001) for 12 to 13 hours at 16°C on an orbital shaker. The speed of the orbital shaker was adjusted according to the volume of ovary sections in the flask, such that the ovary sections were circulating gently around the flask (shaker speeds that are too low will result in incomplete defolliculation, whereas speeds that are too high cause damage to the plasma membrane of the oocytes). The amount of collagenase B added to the solution was dependent on the amount of collagen surrounding the ovary sections, which was assessed by visual inspection under a dissecting microscope. The collagenase-treated oocytes were washed 10 times with OR2⁻ buffer, 5 times with “calcium-containing oocyte ringer” (OR2⁺) buffer (OR2⁻ buffer supplemented with 1 mM CaCl₂ and 50 μg/mL penicillin and streptomycin), and then once with L-15 (Leibovitz) media with L-glutamine (used at 0.5×; Sigma-Aldrich, catalog no. L4386) supplemented with 10 mM HEPES and 50 μg/mL penicillin and streptomycin. Stage V and VI oocytes were microinjected with cRNA encoding a PfMDR1 isoform (10 ng per oocyte), a PfCRT isoform (20 ng per oocyte), human P-gp (10 ng per oocyte, these oocytes served as a positive control), or PfNT1 (20 ng per oocyte, these oocytes served as a negative control). In a subset of experiments that were performed to optimize the amount of cRNA required for human P-gp or PfMDR1 expression, oocytes were microinjected with cRNA amounts ranging from 2.5 to 50 ng. The oocytes were stored at 16 to 18°C in L-15 media supplemented with 10 mM HEPES and 50 μg/mL penicillin and streptomycin.

Radiolabeled and unlabeled drugs used in the Xenopus oocyte transport assays

The radiolabeled drugs were purchased from either Pharmaron (Beijing, China; [³H]chloroquine, 27 Ci/mmol; [³H]dihydroartemisinin, 14 Ci/mmol), American Radiolabeled Chemicals (Missouri, USA; [³H]lumefantrine, 10 Ci/mmol; [³H]mefloquine, 20 Ci/mmol; [³H]amodiaquine, 15 Ci/mmol; [³H]piperaquine, 15 Ci/mmol; [³H]quinine, 20 Ci/mmol; [³H]quinidine, 20 Ci/mmol; [³H]vinblastine sulfate, 20 Ci/mmol), Moravek (California, USA; [³H]amantadine, 137 mCi/mmol), Cambridge Research Biochemicals (Billingham, UK; [³H]VDPVNF, 20 Ci/mmol), or PerkinElmer (Massachusetts, USA; [³H]hypoxanthine, 13.3 Ci/mmol). The structures and protonation states of these drugs are shown in S8 Fig and S2 Data, respectively. The unlabeled compounds were purchased from either Sigma-Aldrich (lumefantrine, mefloquine, chloroquine, quinine, quinacrine, rhodamine B, methylene blue, amantadine, nicardipine, vanadate, verapamil, PSC833, chlorpheniramine, saquinavir, histidine, and iron chelator IV) or GenScript (the peptides LH and VDPVNF). The structures of a subset of these compounds are shown in S8 Fig.

Measurements of drug transport via human P-gp and PfMDR1 in Xenopus oocytes

The human P-gp and PfMDR1 transport assays were performed 1 day post-cRNA injection and, unless specified otherwise, were conducted over 1.5 hours at 27.5°C in ND96 buffer (96 mM NaCl, 2 mM KCl, 1 mM MgCl₂, 1.8 mM CaCl₂, 10 mM MES, and 10 mM Tris-base) that had been adjusted to pH 5.5 to mimic the pH of the DV lumen [107,108]. Since the ATPase domains of ABC transporters (such as PfMDR1 and human P-gp) are located in the cell’s cytosol [36], both PfMDR1 and human P-gp are predicted to orientate in the oocyte plasma membrane with their ATPase domains in the oocyte cytosol. An established efflux assay [10] was therefore used to microinject the radiolabeled drug under study into the oocyte and measure its efflux via human P-gp or PfMDR1. In this assay, the transport of the substrate via human P-gp from the oocyte cytosol into the extracellular solution is analogous to the P-gp–mediated transport of a substrate from the cytosol of a human cell into the extracellular space. Likewise, for PfMDR1 the direction of transport is analogous to the PfMDR1-mediated transport of a substrate from the parasite’s cytosol into the DV (S9 Fig). In all cases, 4 independent experiments were performed (on different days and using oocytes from different frogs) and within each experiment measurements were made from 10 oocytes per treatment.

The oocytes were microinjected with 35 nL of ND96 buffer (pH 8.0) supplemented with the [³H]drug under study. The estimated intracellular concentrations (calculated on the basis that the aqueous volume of stage V andVI oocytes is 400 nL [109]) were as follows: [³H]vinblastine sulfate, 4.4 μM; [³H]lumefantrine, 2.2 μM; [³H]chloroquine, 5.9 μM; [³H]quinine, 4.4 μM; [³H] mefloquine, 4.4 μM; [³H]amodiaquine, 3.1 μM; [³H]piperaquine, 3.1 μM; [³H]dihydroartemisinin, 3.1 μM; [³H]quinidine, 4.4 μM; [³H]amantadine, 30 μM. In the experiments measuring [³H]dihydroartemisinin transport, 1 mM iron chelator IV was also microinjected into the oocyte to bind any free iron present in the oocyte (free iron could activate dihydroartemisinin, which, in turn, could damage the oocytes and thus interfere with measurements of transport; S10 Fig). Ten oocytes that had resealed at the injection site were transferred to a 5 mL polystyrene round-bottom tube (Corning, New York, USA; catalog no. 352008) and washed once with 3.5 mL of ND96 buffer (pH 5.5) before being transferred to separate wells of a white 96-well plate (PerkinElmer, catalog no. 6005299). This treatment determined the amount of radioactivity within the oocytes at the beginning of the assay (T = 0). Another 10 resealed oocytes were transferred to a 5-mL tube, washed once with 3.5 mL of ND96 buffer (pH 5.5), and suspended in 100 μL of ND96 buffer (pH 5.5) to commence the assay. The incubation was terminated by removing the reaction buffer with a pipette and washing the oocytes twice with 3.5 mL of ice-cold ND96 buffer (pH 5.5). Each oocyte was transferred to a separate well of a white 96-well plate, lysed by an overnight incubation at room temperature with 20 μL of 10% (w/v) SDS, and then combined with 150 μL of MicroScint-40 microscintillant (PerkinElmer, catalog no. 6013641). The plate was covered with a TopSeal-A (PerkinElmer, catalog no. 6050185), and the radioactivity within each well was measured with a MicroBeta² microplate liquid scintillation analyzer (PerkinElmer, catalog no. 2450–0010). Drug transport was calculated by subtracting the amount of radioactivity (measured in counts per min) present within the oocytes at the end of the incubation from that measured in the oocytes sampled immediately before the commencement of the assay.

The transport of methylene blue, quinacrine, and rhodamine B via PfMDR1 was measured using the intrinsic fluorescence of these compounds and a fluorescence-based transport assay. The oocytes were microinjected with 35 nL of ND96 buffer (pH 8.0) supplemented with the fluorescent compound under study. The estimated intracellular concentration (calculated on the basis that the aqueous volume of stage V and VI oocytes is 400 nL [109]) for quinacrine, rhodamine B, or methylene blue was 4 μM. Of the oocytes that had resealed at the injection site, 10 were transferred to a 5-mL polystyrene round bottom tube and washed once with 3.5 mL of ND96 buffer (pH 5.5) before being transferred to separate wells of a clear 96-well plate. This treatment determined the amount of fluorescence within the oocytes at the beginning of the assay (T = 0). Another 10 resealed oocytes were transferred to a 5-mL tube, washed once with 3.5 mL of ND96 buffer (pH 5.5), and suspended in 100 μL of ND96 buffer (pH 5.5) to commence the assay. The incubation was terminated by removing the reaction buffer with a pipette and washing the oocytes twice with 3.5 mL of ice-cold ND96 buffer (pH 5.5). Each oocyte was transferred to a separate well of a clear 96-well plate. In addition, the autofluorescence of oocytes that had not been microinjected with a drug was measured by transferring 10 oocytes to a 5-mL tube and washing them 3 times with 3.5 mL of ND96 buffer (pH 5.5) before transferring them to separate wells of a clear 96-well plate. The oocytes were lysed by an overnight incubation at room temperature with 20 μL of 10% (w/v) SDS. The fluorescence intensity in each well was measured using an Infinite M1000 PRO plate reader (Tecan, Mannedorf, Switzerland; catalog no. 30064852) at the following excitation and emission wavelengths: 440 nm and 510 nm for quinacrine; 546 nm and 568 nm for rhodamine B; 668 nm and 682 nm for methylene blue. Drug transport was calculated by first subtracting the autofluorescence measured from all treatments, and then subtracting the fluorescence present within the oocytes at the end of the incubation from that measured in the oocytes sampled immediately before the commencement of the assay.

A subset of assays measured the ability of unlabeled compounds to cis-inhibit [³H]drug transport via human P-gp and PfMDR1. In these experiments, the oocytes were microinjected with 35 nL of the relevant [³H]drug solution supplemented with the unlabeled test compound at a concentration that would achieve the desired intracellular concentration (0.1 to 30 μM). The test solutes included known inhibitors of human P-gp (nicardipine [46,54], PSC833 [55,56], vanadate [57,58], and verapamil [46,59]), and a compound that does not interact with human P-gp (the antiviral drug amantadine [46–48]). The abilities of unlabeled lumefantrine, mefloquine, and chloroquine (intracellular concentrations of 1 to 200 μM) to cis-inhibit [³H]vinblastine transport via PfMDR1 were also assessed. The IC₅₀ value of each inhibitor was determined in GraphPad Prism Version 8 by a least squares fit of the equation y = y_min + [(y_max − y_min)/(1 + ([test compound]/IC₅₀)^c] to the data, where y is P-gp–or PfMDR1-mediated drug transport, y_min and y_max are the minimum and maximum values of y, and c is a fitted constant.

The ATP dependence of lumefantrine transport via PfMDR1 was determined by microinjecting the oocytes with [³H]lumefantrine suspended in 35 nL of ND96 buffer (pH 8.0) containing different concentrations of ATP. Unfertilized stage V and VI oocytes typically contain approximately 2.3 mM ATP [110,111] and microinjection of the different ATP solutions is calculated to have raised the intracellular concentration by 0 to 2 mM.

Analyses of the kinetics of drug transport via PfMDR1 were undertaken by microinjecting the oocytes with ND96 buffer (pH 8.0) supplemented with the [³H]drug under study and different concentrations of the unlabeled drug (to achieve intracellular concentrations between 0 and 100 μM). The kinetic parameters for drug transport via different isoforms of PfMDR1 were determined in GraphPad Prism Version 8 by a least squares fit of the Michaelis–Menten equation to the data.

The experiments measuring drug transport via human P-gp or PfMDR1 included ne and those expressing PfNT1 [60,61] (an unrelated transporter) as negative controls. The latter control verified that the integrity of the oocyte plasma membrane had not been compromised (i.e., that the membrane was not leaky) from the combined effects of the heterologous expression of a foreign protein and the second injection event. To confirm that PfNT1 was expressed at the oocyte surface, the uptake of [³H]hypoxanthine (a known substrate of PfNT1) was measured in ne and those expressing PfNT1 on day 1 post-cRNA injection. Briefly, 10 oocytes were washed twice with 3.5 mL of ND96 buffer (pH 6.0), and the residual solution was removed by pipette. The assay commenced with the addition of 100 μL ND96 buffer (pH 6.0) containing 0.38 μM [³H]hypoxanthine and the incubation period was 30 minutes. The assay was terminated by removing the reaction buffer and washing the oocytes twice with 3.5 mL of ice-cold ND96 buffer (pH 6.0). Each oocyte was transferred to a well of a white 96-well plate and the samples were processed, and the radioactivity measured, as described above.

Measurements of drug transport via PfCRT in Xenopus oocytes

Unless specified otherwise, the PfCRT transport assays were performed 2 to 3 days post-cRNA injection and were conducted over 1.5 hours at 27.5°C. The orientation of PfCRT in the parasite’s DV membrane is such that its amino terminus and carboxyl terminus extend into the cytosol, and hence, the orientation of the protein is predicted to be the same in the oocyte plasma membrane. Hence, the transport of a substrate via PfCRT from the acidic extracellular solution into the oocyte cytosol [6] is analogous to the PfCRT-mediated efflux of a substrate out of the parasite’s acidic DV and into the cytosol [6] (S9 Fig). This is the direction in which mefloquine transport was measured in the PfCRT oocyte assays. Briefly, 10 oocytes were washed twice with 3.5 mL of ND96 buffer (pH 5.0), and the residual solution was removed by pipette. The assay commenced with the addition of 100 μL ND96 buffer (pH 5.0) containing 0.125 μM [³H]mefloquine and 0.5 μM unlabeled mefloquine and, where specified, 100 μM of either a known inhibitor of the transporter (verapamil [6,18], chlorpheniramine [90], or saquinavir [10,91]) or a compound that does not interact with PfCRT (histidine or the dipeptide LH [9,10]). Four independent experiments were performed (on different days and using oocytes from different frogs), and within each experiment measurements were made from 10 oocytes per treatment.

Analyses of the kinetics of mefloquine transport via PfCRT were undertaken by measuring the uptake of [³H]mefloquine (0.125 μM) in the presence of 0 to 50 μM unlabeled mefloquine. In all cases, the incubation was terminated by removing the reaction buffer and washing the oocytes twice with 3.5 mL of ice-cold ND96 buffer (pH 5.0). Each oocyte was transferred to a well of a white 96-well plate and the samples were processed, and the radioactivity measured, as described above. The component of transport attributable to PfCRT (i.e., PfCRT-mediated transport) was calculated by subtracting the level of [³H]mefloquine accumulation detected in the negative control oocytes (ne) from that measured in oocytes expressing PfCRT^3D7, PfCRT^7G8, or PfCRT^Dd2. The kinetic parameters for mefloquine transport via PfCRT were determined in GraphPad Prism Version 8 by a least squares fit of the Michaelis–Menten equation to the data.

We have recently identified the native substrates of PfCRT as being host-derived peptides of 4 to 11 amino acid residues in length [10]. Moreover, we found that in addition to being H⁺-dependent, peptide transport via PfCRT requires a second cosubstrate that remains to be identified, but which is naturally present in the Xenopus oocyte. The PfCRT-mediated transport of the hemoglobin peptide VDPVNF was therefore quantified by microinjecting [³H]VDPVNF into the oocytes and measuring its efflux from the oocyte [10]. Moreover, after preliminary experiments revealed that [³H]lumefantrine is likewise reliant on the presence of the second cosubstrate for transport via PfCRT ([³H]lumefantrine uptake from the extracellular solution could not be detected in PfCRT-expressing oocytes), the transport of this drug was also measured using the efflux assay. Lumefantrine transport was measured by microinjecting oocytes with 25 nL of ND96 buffer (pH 5.5) supplemented with [³H]lumefantrine and unlabeled lumefantrine (to achieve estimated intracellular concentrations of 1.6 μM and 1 μM, respectively). The measurements of [³H]VDPVNF and [³H]lumefantrine transport via PfCRT were conducted as described above for the PfMDR1 efflux assays. Four independent experiments were performed (on different days and using oocytes from different frogs), and within each experiment measurements were made from 10 oocytes per treatment.

Analyses of the kinetics of lumefantrine transport via PfCRT were undertaken by microinjecting the oocytes with [³H]lumefantrine supplemented with different concentrations of unlabeled lumefantrine (to achieve estimated intracellular concentrations between 0 and 50 μM). The kinetic parameters for lumefantrine transport via PfCRT were determined in GraphPad Prism Version 8 by a least squares fit of the Michaelis–Menten equation to the data.

The ability of unlabeled verapamil, chlorpheniramine, saquinavir, histidine, or LH to cis-inhibit [³H]lumefantrine transport was assessed by supplementing the solution containing [³H]lumefantrine and unlabeled lumefantrine with the unlabeled test compound (to achieve an intracellular concentration of 100 μM).

The ability of unlabeled mefloquine or lumefantrine to cis-inhibit [³H]VDPVNF transport was assessed by microinjecting oocytes with 25 nL of ND96 buffer (pH 5.5) supplemented with [³H]VDPVNF and unlabeled VDPVNF (to achieve estimated intracellular concentrations of 1.4 μM and 200 μM, respectively) as well as the unlabeled drug at a concentration that would achieve the desired intracellular concentration (10 to 250 μM). The IC₅₀ values were determined in GraphPad Prism Version 8 by a least squares fit of the equation y = y_min + [(y_max − y_min)/(1 + ([test compound]/IC₅₀)^c] to the data, where y is PfCRT-mediated VDPVNF transport, y_min and y_max are the minimum and maximum values of y, and c is a fitted constant.

Western blot analyses

Semiquantitative measurements of the level of HA-tagged PfMDR1 protein in the oocyte membrane 1 day post-cRNA injection were performed using an established method for obtaining preparations of oocyte membrane proteins [7]. The pellet containing the membrane protein fraction was solubilized in a solution comprising 10 mM dithiothreitol, 1.5% (v/v) SDS, 10% (v/v) β-mercaptoethanol, 9 mM Tris-HCl (pH 7.6), 4.5 mM NaCl, 0.45% (v/v) Triton X-100, and 27.5% (v/v) NuPage LDS sample buffer (Life Technologies, California, USA; catalog no. NP0007). The proteins present in the oocyte membrane preparations were separated in a NuPage 3 to 8% Tris-Acetate polyacrylamide gel (Thermo Fisher Scientific, catalog no. EA03752) using Tris-Acetate SDS running buffer, and then transferred to a 0.45 μM nitrocellulose blotting membrane (Bio-Rad, California, USA; catalog no. 1620115). The membranes were probed with mouse anti-HA antibody (1:5,000; Sigma-Aldrich, catalog no. H9658) followed by horseradish peroxidase-conjugated horse anti-mouse antibody (1:3,000; Cell Signaling Technologies, Massachusetts, USA; catalog. no 7076S). The band for each PfMDR1 isoform was detected with SuperSignal West Pico Chemiluminescent reagent (Thermo Fisher Scientific, catalog no. 34578), quantified using the ImageJ software [112], and expressed as a percentage of the band intensity measured for PfMDR1^NYSND. Total protein staining was used to evaluate sample loading and efficiency of transfer [7,113]. Briefly, the nitrocellulose membranes were stained with Ponceau S for 5 minutes, then de-stained using ultrapure water. The stained protein bands were visualized using the Bio-Rad ChemiDoc MP Imaging System, and densitometric analyses were performed in Image Lab version 6.0.1 software (Bio-Rad) to quantify the summed intensity of all of the protein bands between 70 and 180 kDa (the region in which PfMDR1 is detected) within each lane. At least 4 independent experiments were performed (using oocytes from different frogs), and within each experiment, measurements were averaged from 2 separate replicates.

Immunofluorescence assays

An immunofluorescence assay was used to localize HA-tagged PfMDR1 in oocytes 1 to 2 days post-cRNA injection using a method adapted from Richards and colleagues [9]. Unless otherwise specified, all incubation and wash steps were conducted at room temperature with gentle shaking or rotation. All incubations were performed using a volume of 500 μL, and all wash steps were performed using a volume of 1 mL. Six oocytes of each oocyte type were fixed with 4% (v/v) paraformaldehyde in PBS for 30 minutes. Oocytes were then washed 3 times for 10 minutes each in PBS and then permeabilized by incubation in 100% methanol at −20°C for 20 minutes without shaking. Another three 10-minute washes in PBS were performed before the oocytes were incubated for 2 hours in a blocking solution (4% (w/v) BSA, 2% (v/v) normal goat serum, and 0.1% (v/v) Triton X-100 in PBS). The solution was replaced with a second blocking solution (4% (w/v) BSA and 2% (v/v) normal goat serum in PBS) and incubated overnight at 4°C. The following day, the oocytes were incubated for an additional 4 hours at room temperature in the same blocking solution before incubation in a solution containing mouse anti-HA antibody (1:100; in a solution of PBS containing 1.5% (w/v) BSA and 0.01% (v/v) Triton X-100; Sigma-Aldrich, catalog no. H9568) for 4 hours at room temperature and then overnight at 4°C. The following day, the oocytes were washed 3 times for 10 minutes each in PBS supplemented with 1.5% (w/v) BSA. All of the remaining steps were performed in the dark at room temperature. The oocytes were incubated for 4 hours in a solution containing the Alexa Fluor 488 donkey anti-mouse antibody (1:500; in a solution of PBS containing 4% (w/v) BSA and 2% (v/v) normal goat serum; Molecular Probes, Oregon, USA; catalog no. A21202). The oocytes were then washed 3 times for 10 minutes each in PBS, fixed with 3.7% (v/v) paraformaldehyde in PBS for 30 minutes, and washed twice for 15 minutes each in PBS. The oocytes were dehydrated by a series of 15-minute incubations in solutions of increasing ethanol content: 30% (v/v) ethanol in PBS, 50% (v/v) ethanol in PBS, 70% (v/v) ethanol in ultrapure water, 90% (v/v) ethanol in ultrapure water, and 100% ethanol (3 incubations in 100% ethanol were performed). The dehydrated oocytes were embedded in acrylic resin using the Technovit 7100 plastic embedding system (Kulzer, Indiana, USA; catalog no. 64709003). Slices of 2 μm were obtained using a microtome with a glass knife, and the slices were dried on microscope slides. A drop of ProLong Gold Antifade Mountant (Life Technologies, catalog no. P36934) was placed on a #1.5 coverslip, which was then placed over the oocyte slices on each slide and sealed using clear nail polish. Images were captured using a Leica Sp5 inverted confocal laser microscope (Leica Microsystems, Wetzlar, Germany) using a 63× objective. Excitation was achieved with a 488-nm argon laser, and the emissions were captured using a 500- to 550-nm filter. Images were acquired using the Leica Application Suite Advanced Fluorescence Version 3.6 software (Leica Microsystems). At least 2 independent experiments were performed (on oocytes from different frogs) for each oocyte type, within which slices were examined from at least 5 oocytes.

Orientation of PfMDR1 in the plasma membrane of Xenopus oocytes

To determine the orientation of PfMDR1 at the surface of Xenopus oocytes, an immunofluorescence assay was developed and performed on oocytes 1 day post-cRNA injection. This method used live oocytes that were not fixed or permeabilized, thus preventing the penetration of antibody into the oocyte cytosol. Hence, under these conditions, the antibody can only bind to cognate extracellular epitopes. All incubation steps were performed at 16 to 18°C in a volume of 500 μL with gentle shaking. The experiment included the following oocyte types: ne (the negative control), oocytes expressing HA_EL2-AmDAT (the positive control), and oocytes expressing either 3xHA-PfMDR1^NYSND or PfMDR1^NYSND-3xHA. Six oocytes from each type were incubated for 6 hours in a blocking solution (3% (w/v) BSA and 1% (v/v) normal goat serum in OR2⁺ buffer). This was followed by a 2-hour incubation with the mouse anti-HA antibody (1:250 in a solution of 1.5% (w/v) BSA in OR2⁺ buffer; Sigma-Aldrich catalog no. H9568). The samples were washed 3 times for 10 minutes each with OR2⁺ buffer and then incubated for 2 hours with the Alexa Fluor 488 donkey anti-mouse antibody (1:500 in a solution of 1.5% (w/v) BSA in OR2⁺ buffer; Molecular Probes, catalog no. A21202). The oocytes were then rinsed 3 times (10 minutes each) with OR2⁺ buffer. Imaging was performed using a Leica M205 FA fluorescence stereomicroscope and Leica Application Suite Version 4.12 software (Leica Microsystems). Excitation was achieved using a 488 nm laser and the emissions were captured using a 500- to 550-nm filter. At least 2 independent experiments were performed (using oocytes from different frogs) for each oocyte type.

Culture of parasitized erythrocytes

The use of human blood for this work was approved by the ANU Human Research Ethics Committee (Human Ethics Approval Number 2017/351). The asexual intraerythrocytic stages of 3 P. falciparum pfcrt transfectant lines and 3 P. falciparum strains were studied. The 3 pfcrt transfectant lines were the chloroquine-sensitive line C2^GC03 (a recombinant control that retains the wild-type pfcrt allele, PfCRT^3D7) and the chloroquine-resistant lines C6^7G8 and C4^Dd2 (in which the wild-type pfcrt allele of the “GC03” line has been replaced with the mutant allele from the 7G8 or Dd2 strain, respectively [17]). A PCR error in the generation of the C6^7G8 line introduced an additional mutation (I351M) [22], which is not found in the sequence of PfCRT^7G8. The strains included 3D7 and HB3, which are chloroquine-sensitive parasites that carry PfCRT^3D7, but express either PfMDR1^NYSND or PfMDR1^NFSDD, respectively. The chloroquine-resistant strain was Dd2, which expresses PfCRT^Dd2 and PfMDR1^FYSND and/or PfMDR1^YYSND. The parasite cultures were maintained at 4% hematocrit and synchronized with 5% (w/v) sorbitol [114]. The pfcrt transfectant lines were maintained in the presence of the selection agents blasticidin (5 μM; Sigma-Aldrich, catalog no. 30891) and WR99210 (5 nM; MedChemExpress, catalog no. HY-116387). These selection agents were not present during the experiments.

Two cell line authentication methods were used to verify the identities of the 6 parasite cultures. The first evaluated the chloroquine resistance phenotype of each culture by performing proliferation assays, thereby yielding chloroquine IC₅₀s for each parasite line or strain. The results were verified by comparing them to previously published data [9,10,20,21] (see the P. falciparum proliferation assays section for details). Mycoplasma detection assays were also performed periodically on each of the 6 parasite cultures via PCR with a primer mix that amplifies ribosomal DNA from various mycoplasma strains [115]. No mycoplasma infections were detected in any of the parasite cultures throughout the course of this study.

P. falciparum H⁺-efflux assays

Parasite cultures were enriched for trophozoite-stage parasites by magnetic separation and were cultured with uninfected erythrocytes loaded with the membrane-impermeant pH-sensitive fluorescent indicator fluorescein-dextran (10,000 MW; Life Technologies, catalog no. D1821). This dye is taken up into the parasite’s DV as they grow inside the erythrocyte. Following a complete asexual cycle (approximately 48 hours) mature trophozoite-stage parasites were isolated from erythrocytes using 0.05% (w/v) saponin and suspended in a saline solution (125 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 20 mM glucose, and 25 mM HEPES; pH 7.1) at a density of 1 × 10⁷ cells/mL. The fluorometry experiments were performed as outlined previously [21]. Briefly, the pH of the DV was monitored at 37°C using a Cary Eclipse fluorescence spectrophotometer (Agilent Technologies, California, USA) (excitation wavelengths 490 nm and 450 nm; emission wavelength 520 nm). The drugs were added to the parasite suspension (2.5 μM of either chloroquine, saquinavir, mefloquine, or lumefantrine), and the solution was incubated for 4 minutes (chloroquine, saquinavir, and mefloquine) or 8 minutes (lumefantrine) to allow the drugs to accumulate to a high concentration within the DV. This was followed by the addition of the V-type H⁺-ATPase inhibitor concanamycin A (100 nM; Sigma-Aldrich, catalog no. 27689). Half-times for the rate of DV alkalinization were determined in GraphPad Prism Version 8 by a least squares fit of the equation F = F₀ + F_max/[1 + (t/t_1/2)^c] where F is the fluorescence ratio, F₀ is the initial fluorescence ratio (averaged over 20 seconds immediately prior to opening the chamber of the fluorometer and adding concanamycin A), t is time, t_1/2 is the half-time for DV alkalinization, F_max is the maximal change in fluorescence ratio, and c is a fitted constant [20]. The rate of PfCRT-mediated DV alkalinization was calculated by subtracting the rate of DV alkalinization of the solvent control from that of each treatment within each parasite line. In all cases, 4 independent experiments were performed on different days.

P. falciparum proliferation assays

The parasite proliferation assays were performed in clear 96-well plates using a fluorescent DNA-intercalating dye as described previously [116,117]. Briefly, synchronous ring-stage parasite cultures (approximately 1% hematocrit and 1% parasitemia) were incubated with increasing concentrations of quinine or chloroquine for 72 hours at 37°C and under reduced O₂ conditions. The assay was terminated by freezing and thawing the samples, after which 100 μL from each well was transferred to another 96-well plate. This was followed by the addition of 100 μL of SYBR Safe DNA Gel Stain (0.2 μL/mL; Molecular Probes, catalog no. S33102) in a lysis buffer (20 mM Tris-HCl, 5 mM EDTA, 0.008% (w/v) saponin, and 0.08% (v/v) Triton X-100; pH 7.5) to each well. Fluorescence was measured using a Tecan Infinite M1000 PRO microplate reader (excitation wavelength 490 nm; emission wavelength 520 nm). For each 96-well plate, the fluorescence values for the wells containing the highest concentration of the compound were averaged, and this value was then subtracted from the fluorescence values obtained for each well. The level of parasite proliferation in the presence of each compound concentration was expressed as a percentage of the proliferation measured in the absence of the compound. The IC₅₀s were determined in GraphPad Prism Version 8 by a least squares fit of the equation y = a/[1 + ([compound]/IC₅₀)^c] to the data, where y is the percent parasite proliferation, a is the maximum change in the percent parasite proliferation, and c is a fitted constant. In all cases, 4 independent experiments were performed (on different days), and within each experiment, measurements were averaged from 3 replicates.

Construction of the MDR1 alignment and generation of the PfMDR1 homology model

Proteins belonging to the MDR1 family were retrieved from the NCBI database using a Basic Local Alignment Search Tool (BLAST) search with PfMDR1 (PF3D7_0523000) as the query sequence. A representative selection of 61 of these proteins, including those for which structures have been determined, were aligned using the ClustalW program [118], and the resulting alignment was manually edited in MacVector Version 12.7.5. Generating an alignment that includes a large number of related proteins is likely to enhance confidence in the alignment of PfMDR1 to a template structure compared with using a simple pairwise alignment. Using this large multiple sequence alignment, we identified 2 proteins related to PfMDR1 with known structures and significant sequence identity: C. elegans P-gp in the inward-open state [68] (PDB accession code 4F4C, 32.6% sequence identity) and human P-gp in the outward-open state [69] (PDB accession code 6C0V, 33.2% sequence identity). Using the automodel protocol of MODELLER Version 9.17 [119], 100 models of PfMDR1 were built from each template, and the best model was chosen according to the lowest discrete optimized protein energy (DOPE) [120] value calculated by MODELLER Version 9.17. Molecular figures were made using the Visual Molecular Dynamics (VMD) modeling program [121].

Molecular dynamics simulations of quinine in the binding pocket of the inward open PfMDR1 homology model were conducted using the NAMD Version 2.13 simulation code [122]. First, CGENFF force field [123] parameters for quinine were generated using the automated ParamChem CGENFF tool [124,125]. Seven independent simulations were conducted, with random starting positions and orientations of quinine within the large aqueous cavity of PfMDR1. Within the structure of PfMDR1, atoms further than 17 Å from the center of the binding pocket (defined by the midpoint of the backbones of L327 and F1072 that are on opposing sides of the cavity) were harmonically restrained while atoms within this range were not restrained. Protein restraints were decreased from 1 kcal/mol/Ų to 0.001 kcal/mol/Ų over 8 steps (values were 1, 0.5, 0.2, 0.1, 0.05, 0.02, 0.01, and 0.001) with the system energy minimized for 100 steps before 100 ps of dynamics at each step. Once this equilibration phase was complete, simulations were continued for 30 ns with the final harmonic restraint on the protein maintained. The central cavity of PfMDR1 was solvated with TIP3P water molecules that were restrained to stay within a sphere (radius of 20 Å) at the center of the binding pocket, using a spherical harmonic boundary potential applied to the water molecule’s oxygen atoms with a force constant of 20 kcal/mol/Ų. The CHARMM36 force field with CMAP correction [126] was used for the protein, and the simulations were performed at a temperature of 300 K using a Langevin thermostat. All bonds to hydrogen were kept fixed, allowing the use of 2 fs timesteps.

Correlations between in vitro drug resistance indices and the rate of drug transport via PfMDR1

IC₅₀s for lumefantrine, mefloquine, chloroquine, and quinine were collated from the relevant literature (see S3 Table for references). The inclusion criteria that were applied to previously published studies were: (1) that an IC₅₀ for one or more of the parasite strains of interest (i.e., HB3, Dd2, K1, GB4, or 7G8) be reported, and (2) that the IC₅₀ for the 3D7 control strain be performed pairwise. For lumefantrine, mefloquine, and quinine, 4 to 10 studies fitted these criteria and were thus incorporated into the analyses. Several of the studies that measured chloroquine IC₅₀s did not include pairwise measurements with the 3D7; instead, pairwise measurements with either HB3 or GC03 parasites were made (both of these strains express PfCRT^3D7 and PfMDR1^NFSDD, and their chloroquine IC₅₀s are typically very similar to 3D7 in pairwise studies). Since these studies measured chloroquine IC₅₀s for infrequently investigated parasite strains such as GB4, it was important to include them in our analyses. Therefore, the inclusion criteria for the chloroquine studies were altered to include those that had measured chloroquine IC₅₀s in either HB3 or GC03 parasites as the control strains, regardless of whether the 3D7 strain was included. The in vitro resistance indices for each study were calculated by dividing the IC₅₀ of each strain by the IC₅₀ of the control strain (3D7 for lumefantrine, mefloquine, and quinine; HB3 or GC03 for chloroquine). The indices across multiple studies were then averaged and the standard error of the mean calculated (or the range/2 for resistance indices that were calculated from only 2 studies). For each drug, the resistance indices were plotted against the experimentally determined transport activities of the PfMDR1 isoforms in Xenopus oocytes (S1 Data).

Quantification and statistical analyses

The statistical tests, data collection, and analyses were performed using ImageJ Version 1.8.0, GraphPad Prism Version 8, Leica Application Suite Advanced Fluorescence software Version 3.6, Leica Application Suite software Version 4.12, Image Lab version 6.0.1 software (Bio-Rad), MODELLER Version 9.17, VMD, NAMD Version 2.13, and MarvinSketch software Version 18.10. Statistical comparisons were made using 1-way ANOVAs in conjunction with Tukey multiple comparisons test. All errors cited in the text and shown in the figures represent the SEM or the range/2. Where not shown, error bars fall within the symbols. Significance was defined as P < 0.05. The key statistical details for each dataset are included in the figure legends.