Antimalarial compounds with dual therapeutic and transmission-blocking activity are desired as high-value partners for combination therapies. Here, we report the identification and characterization of hexahydroquinolines (HHQs) that show low nanomolar potency against both pathogenic and transmissible intra-erythrocytic forms of the malaria parasite Plasmodium falciparum. This activity translates into potent transmission-blocking potential, as shown by in vitro male gamete formation assays and reduced oocyst infection and prevalence in Anopheles mosquitoes. In vivo studies illustrated the ability of lead HHQs to suppress Plasmodium berghei blood-stage parasite proliferation. Resistance selection studies, confirmed by CRISPR–Cas9-based gene editing, identified the digestive vacuole membrane-spanning transporter PfMDR1 (P. falciparum multidrug resistance gene-1) as a determinant of parasite resistance to HHQs. Haemoglobin and haem fractionation assays suggest a mode of action that results in reduced haemozoin levels and might involve inhibition of host haemoglobin uptake into intra-erythrocytic parasites. Furthermore, parasites resistant to HHQs displayed increased susceptibility to several first-line antimalarial drugs, including lumefantrine, confirming that HHQs have a different mode of action to other antimalarials drugs for which PfMDR1 is known to confer resistance. This work evokes therapeutic strategies that combine opposing selective pressures on this parasite transporter as an approach to countering the emergence and transmission of multidrug-resistant P. falciparum malaria.
Malaria remains a global health calamity, with an estimated 212 million cases and 429,000 deaths1 in 2015, that occurs mostly in sub-Saharan Africa and results from infection with the Apicomplexan parasite Plasmodium falciparum (Pf). Improved malaria treatment and control has reduced morbidity and mortality worldwide by 40–50% since 2000 (ref. 1), due largely to highly effective artemisinin-based combination therapies and mosquito vector control measures1,2. An important strategy to further advance the malaria elimination agenda is to include transmission-blocking agents in combination therapies3. Such agents inhibit the development of parasite gametocytes (GAMs), which represent an alternative form of intra-erythrocytic development to the pathogenic asexual blood stage (ABS) parasites and that mature over five stages4. Only stage V GAMS, which may survive days to weeks in circulation5, can continue the Pf lifecycle in Anopheles mosquitoes following ingestion of a parasitized blood meal. In the mosquito midgut, male GAMs exflagellate and produce microgametes that can fuse with female macrogametes to produce zygotes. Following subsequent development as ookinetes and oocysts, sporozoites are formed. Once inoculated in a new host, they will develop in the liver and initiate a new ABS cycle6.
Primaquine is currently the only available drug with potent activity against stage V GAMs7,8, but it can be toxic for glucose-6-phosphate dehydrogenase-deficient patients8. Hence, there is a compelling need to develop alternative transmission-blocking drugs. Recent advances in GAM-specific high-throughput screen technology9,10,11,12,13,14,15,16,17,18,19, combined with functional validation assays and drug action/resistance studies19,20,21, provide a powerful foundation to identify leads with Pf transmission-blocking activity.
In this study, we report a screen of the Novartis-GNF (The Genomics Institute of the Novartis Research Foundation) Malaria Box of ABS-active compounds22 on Pf GAM stages and describe the identification and antimalarial properties of hexahydroquinolines (HHQs), a chemical class with considerable transmission-blocking potential.
Identifying HHQs as potent gametocytocidal inhibitors
Using a luciferase (LUC)-based assay, we screened 3,825 compounds from the Novartis-GNF Malaria Box for activity against NF54Pfs16 early-stage GAMs. This screen yielded 797 compounds exhibiting ≥50% inhibition at 2.6 μM. Re-screening confirmed activity for 683 compounds, of which 176 showed ≥50% inhibition at 0.65 μM (Fig. 1a). Screening the Novartis-GNF Malaria Box against late-stage GAMs using high-content imaging identified 572 hits with ≥50% inhibition at 2 µM in at least two of the three replicates (Fig. 1a and Supplementary Fig. 1a).
Late-stage gametocytocidal hits were then prioritized for half-maximum inhibitory concentration (IC50) determination using both early- and late-stage GAM assays. Limited availability restricted the number of compounds to 149 (Supplementary Fig. 1b), yielding 71 and 60 compounds with IC50 values below 1 µM for early- and late-stage GAMs, respectively (Supplementary Table 1). Among these, 20 early-GAM and 16 late-GAM active compounds had IC50 values of ≤200 nM. In total, 43 compounds showed sub-micromolar activity against both stages (Fig. 1b). Reference drug controls showed expected values (Supplementary Table 2).
After deprioritizing chemical classes related to current antimalarial drugs or having liabilities likely to impede potential development, we identified four compounds with potent early- and late-stage gametocytocidal activity: GNF-Pf-5640, GNF-Pf-5660, GNF-Pf-5668 and GNF-Pf-5310 (Fig. 1c and Supplementary Table 3). These compounds also showed potent ABS activity (IC50 values <25 nM) against drug-sensitive 3D7 and multidrug-resistant Dd2 parasites, including a PfATP4 mutant line (Supplementary Fig. 2 and Supplementary Table 3). GNF-Pf-5640, GNF-Pf-5660 and GNF-Pf-5668 share a common scaffold belonging to the HHQ class of compounds (Fig. 1d), whereas GNF-Pf-5310 is a muscarinic-like compound (Supplementary Fig. 3). Cytotoxicity assays showed no inhibition at 40 μM with GNF-Pf-5660, while the other three compounds caused <50% inhibition at that concentration (Supplementary Fig. 2), indicating minimal toxicity against mammalian cells.
HHQs demonstrate potent in vitro transmission-blocking activity
Using an in vitro Pf dual gamete formation assay (DGFA), we investigated whether these four potent GAM-active compounds would also inhibit the development of cultured female and male gametes. The results indicated a potent inhibitory effect of the three HHQs on exflagellation when GAMs were exposed for 24 h and the compound carried over during the subsequent period of male gamete induction (‘carry-over’, Table 1). This effect persisted when compounds were washed out after 24 h, that is, before gamete induction (‘wash-out’, Table 1). HHQs were thus effective against both mature male GAMs and developing male gametes, but showed no activity on female gamete formation. The muscarinic-like compound GNF-Pf-5310 showed no activity against either male or female gametes, contrasting with our transmission-blocking control methylene blue23,24 (Table 1).
GNF-Pf-5660, GNF-Pf-5668 and GNF-Pf-5310 were also tested in a membrane-feeding assay using Anopheles stephensi mosquitoes. This assay more closely mimics the natural parasite lifecycle than the DGFA, because compound-exposed Pf GAMs initiate the entire mosquito vector cycle from gametes to midgut oocysts. Pre-treatment of GAMs with GNF-Pf-5660 or GNF-Pf-5668 at 3× their late-stage IC50 caused a 97% and 57% reduction, respectively, in the infection prevalence (that is, the percentage of infected mosquitoes, Fig. 2a), whereas the oocyst density per infected mosquito midgut was reduced by 95–99% (Fig. 2b). GNF-Pf-5660- and GNF-Pf-5668-exposed GAM cultures used for feeding to A. stephensi showed a 98% and 70% reduction, respectively, in their male GAM exflagellation rates (Fig. 2c), confirming earlier DGFA results. GNF-Pf-5310 was less active at similar exposure levels, causing no reduction in infection prevalence, a 73% reduction in oocyst density, and no effect on male GAM exflagellation rates (Fig. 2). As a result, GNF-Pf-5310 was deprioritized for further studies. The positive control methylene blue displayed maximal transmission-blocking activity (Fig. 2). Pre-treatment of GAMs with compounds at 1× their late-stage IC50 caused a lower to no effect on the parameters assessed (Fig. 2 and Supplementary Table 4a). The vehicle control DMSO showed no reductions compared to untreated controls.
In vivo activity of HHQs
HHQ antimalarial activity was assayed in Plasmodium berghei (Pb)-infected mice using the 4-day Peters test25. Intraperitoneal injection of 70 mg kg–1 of GNF-Pf-5660 or GNF-Pf-5668 reduced parasitaemias by 22% or 43%, respectively, compared to vehicle-treated mice (Fig. 2d). At 100 mg kg–1, these agents reduced parasitaemia by 65% and 91%, respectively (Fig. 2d and Supplementary Table 4b). To assess the in vivo transmission-blocking activity of HHQs, infected mice were administered only two 100 mg kg–1 doses of GNF-Pf-5660 or GNF-Pf-5668. Following blood meals on GNF-Pf-5660-dosed mice, we observed a 25% and 21% reduction in the number of infected mosquitoes and their oocyst burdens, respectively (Fig. 2a,b). No effect was observed in the GNF-Pf-5668 mosquito cohort, although it displayed a 31% reduction in the number of male exflagellation centres at the time of mosquito feeding (Fig. 2c and Supplementary Table 4c). Limited in vivo transmission-blocking activity, as compared with our in vitro results, was also observed upon treating mice with two doses of dihydroartemisinin (DHA). Those assays showed 65% and 75% reductions in the number of infected mosquitoes and the oocyst burden, respectively, and found no reduction in the numbers of male exflagellation centres (Fig. 2a–c). These data are consistent with this two-dosing regimen having less impact on ABS parasitaemia than the 4 days of dosing used in the Peters test (Fig. 2d).
Resistance to HHQs is mediated by mutations in PfMDR1 and results in sensitization to clinical antimalarials
To gain insight into the mode of action of HHQs, we selected for Pf ABS resistance in vitro in the Dd2-B2 parasite line (the B2 clone of the Dd2 line), an approach that has proven effective for multiple antimalarial agents26. Whole-genome sequence analysis of the resulting resistant clones identified the PfMDR1 (P. falciparum multidrug resistance gene-1) S1075I mutation in the GNF-Pf-5640-selected clone 2C4C3 and PfMDR1 Y290F in the GNF-Pf-5640-selected clone 3H9C7. These resistant clones showed a 77- and 79-fold increase in IC50 values for GNF-Pf-5640 compared to the parental Dd2-B2 line, respectively (Fig. 1e). The allele balance (AB) of the mutations in these clones was 50–59%, consistent with their acquisition in one of the two pfmdr1 copies present in Dd2-B2. The GNF-Pf-5660-selected clone 2B6H1 and the GNF-Pf-5668-selected clone 3C4D2 showed IC50 increases of 248- and 37-fold, respectively, and had acquired the PfMDR1 Y290F (AB 100%) or F1072L (AB 67%) mutations, respectively (Fig. 1e). Shifts in IC90 values were similar to IC50 increases (Supplementary Tables 5 and 6).
Whole-genome sequencing data indicated that the GNF-Pf-5668-selected 3C4D2 clone also carried four copies of pfmdr1, consistent with further gene amplification compared to the Dd2-B2 parent (which has two pfmdr1 copies in tandem). We were unable to determine whether the F1072L mutation (AB 67%) was mainly located in the additional amplicons. Susceptibility testing of resistant clones indicated that each of the three observed PfMDR1 mutations conferred cross-resistance to all HHQs (Fig. 1e).
Because HHQ resistance was the highest in lines harbouring the Y290F mutation, we used a previously derived NF54 line27 that harboured this mutation (NF54selY290F) to verify whether Y290F also conferred resistance at the GAM stage. NF54selY290F early-stage GAMs showed a 40–74-fold IC50 shift for all HHQs compared to wild-type NF54 (Supplementary Table 7). High levels of resistance were also observed at the late GAM stage, but because only very low numbers of NF54selY290F late GAMs could be obtained, accurate late-stage IC50 values could not be calculated.
Structural modelling of PfMDR1 localized amino acid residue 290 to transmembrane domain (TMD) 5 and residues 1072 and 1075 to TMD 12. All three residues resided in the predicted drug-binding pocket cavity of PfMDR128, where they might mediate contact with substrates (Fig. 3a,b). Docking studies suggested that each mutation could significantly alter transporter interactions with HHQs. Interestingly, F1072L and Y290F seemed to decrease the minimum estimated energy required for docking (Supplementary Tables 8–12 and Supplementary Fig. 4). This effect was less pronounced for S1075I, which could be due to this residue not being fully exposed to the drug-binding pocket.
To test whether these point mutations in PfMDR1 were causal for resistance, we implemented a pfmdr1-specific CRISPR–Cas9 strategy27. Transfection of parental Dd2-B2 parasites yielded the PfMDR1edF1072L and PfMDR1edS1075I clones that expressed the F1072L and S1075I mutations, respectively, observed in both pfmdr1 copies. These clones showed an essentially equivalent gain of resistance to HHQs as the original drug-pressured lines, confirming the causal role of the PfMDR1 mutations in mediating resistance (Fig. 1e). Despite repeated efforts, we were unable to introduce the Y290F mutation into the parental line.
Testing KD1mdr1, a pfmdr1 knockdown line previously generated in FCB parasites, showed that a decreased pfmdr1 copy number did not affect HHQ susceptibility (Fig. 1e). Similarly, increased pfmdr1 copy number also did not noticeably affect HHQ susceptibility (Fig. 1e), as shown by the similar IC50 values of GNF-Pf-5668-3C4D2 (harbouring an estimated four copies of pfmdr1 in addition to the F1072L mutation) and PfMDR1edF1072L (with the F1072L mutation engineered into the wild-type Dd2-B2 line with two pfmdr1 copies).
The impact of the three PfMDR1 mutations on Pf ABS susceptibility to known antimalarials was determined using PfMDR1edF1072L, PfMDR1edS1075I and GNF-Pf-5660-2B6H1 (also referred to as PfMDR1selY290F) (Fig. 3c). For all three lines, no changes in susceptibility were observed for piperaquine, chloroquine (CQ), the active metabolite monodesethyl-CQ (md-CQ), md-amodiaquine (md-ADQ) or ferroquine (Fig. 3 and Supplementary Tables 13 and 14). In contrast, PfMDR1edF1072L and PfMDR1selY290F showed a higher susceptibility to both quinine (QN) (3.6- and 3.7-fold, respectively) and mefloquine (MQ) (3.6- and 2.0-fold, respectively), whereas PfMDR1edS1075I showed a 5.1-fold increased QN susceptibility but only a 1.4-fold increased susceptibility to MQ. Lumefantrine (LMF) showed a similar trend, with PfMDR1edF1072L and PfMDR1selY290F showing a higher LMF susceptibility (3.1- and 1.9-fold, respectively), whereas PfMDR1edS1075I remained unchanged. All three mutations significantly increased Pf susceptibility to DHA by ~1.6-fold. Parasite susceptibility to the experimental ACT-451840 compound, a known inhibitor of PfMDR127, was decreased by 16-, 5.5- and 45-fold for PfMDR1edF1072L, PfMDR1edS1075I and PfMDR1selY290F, respectively (Fig. 3c).
To assess the effects of pfmdr1 copy number variation in combination with the F1072L mutation, we compared the antimalarial susceptibilities of GNF-Pf-5668-3C4D2 (harbouring four pfmdr1 copies +F1072L (AB 69%)) to that of PfMDR1edF1072L (with two pfmdr1 copies +F1072L (AB 100%)). The results indicated a decreased susceptibility to QN, MQ and LMF (that is, higher IC50 values) in GNF-Pf-5668-3C4D2 compared to PfMDR1edF1072L (Fig. 3c). Importantly, these pfmdr1 copy number-induced changes in GNF-Pf-5668-3C4D2 did not fully compensate for the effects introduced by the F1072L mutation, as this line still showed a increased susceptibility to QN (1.7-fold) and MQ (1.4-fold) compared to the wild-type Dd2-B2 control (Fig. 3c). These data provide evidence that the F1072L mutation had a major effect on increasing parasite susceptibility to QN, MQ and LMF that was only partially reversed by increases in pfmdr1 copy number.
HHQs inhibit haemoglobin endocytosis
Previous studies have shown that mutant PfMDR1 isoforms can modulate the degree of Pf ABS resistance to several antimalarial drugs, including CQ and ADQ, in lines expressing mutant pfcrt 29,30,31. These 4-aminoquinolines are known to inhibit haemozoin (Hz) formation and cause a build-up in free haem32,33. In light of the finding that PfMDR1 confers Pf resistance to HHQs, we performed subcellular haem fractionation assays to assess whether these agents might also affect Hz formation in ABS parasites (Fig. 4 and Supplementary Table 15).
In parental Dd2-B2 parasites, we observed that haemoglobin (Hb) and free haem concentrations per parasite remained unchanged across all GNF-Pf-5660 exposure levels (up to 2× IC50) (Fig. 4). Hz levels, however, showed a 1.5-fold decrease when parasites were exposed to as little as 0.1× the GNF-Pf-5660 IC50 value, as compared with untreated controls. A 2.3-fold reduction in Hz levels was observed at higher exposure levels (0.25× to 2× IC50, Fig. 4). Total haem iron levels also showed similar reductions with increasing concentrations of GNF-Pf-5660 (Fig. 4). A similar profile of reduced Hz and total haem iron content per parasite was also observed in PfMDR1selY290F, albeit at >160-fold higher GNF-Pf-5660 exposure levels than in Dd2-B2, consistent with the fold increase in IC50 in the mutant line (Fig. 4).
For comparison, we also assessed the haem fractionation profiles of MQ and LMF. In the Dd2-B2 line, MQ and LMF showed little or no effect compared to untreated controls, with the exception of a 1.6-fold decrease in Hz levels in parasites exposed to 2× the MQ IC50. Intriguingly, PfMDR1selY290F showed MQ profiles similar to those observed with GNF-Pf-5660, that is, reductions in Hz and total haem iron levels with increasing MQ concentrations. In PfMDR1selY290F, LMF also caused a reduction in Hz and total haem iron amounts at 1× to 2× IC50 concentrations (Fig. 4).
We also measured the amount of Hb endocytosed by late rings exposed to GNF-Pf-5660, WR99210, MQ or LMF, using western blot analysis of wild-type Dd2-B2 or HHQ-resistant PfMDR1selY290F parasites (Fig. 5 and Supplementary Table 16). Results showed that the Hb levels within parasites were reduced by 32% in GNF-Pf-5660-exposed Dd2-B2, but remained unchanged in PfMDR1selY290F. Dd2-B2 parasites showed no significant changes in Hb levels when exposed to WR99210, MQ or LMF.
Here, we describe the antimalarial properties of HHQs, a class of potent gametocytocidal antimalarials with ABS activity that were identified by screening the Novartis-GNF Malaria Box against early- and late-stage Pf GAMs. Functional assays with Pf GAMs and gametes confirmed potent transmission-blocking activity. Moderate activity with our initial hits was also observed in vivo in Pb mouse models. Mode of action and resistance studies implicated a key role for the digestive vacuole (DV) transporter PfMDR1. We also provide evidence that HHQs reduce levels of Hz and total haem iron in treated parasites and inhibit the initial uptake of Hb from host red blood cells (RBCs). Cross-resistance assays also revealed that HHQ-resistant parasites became sensitized to several antimalarials in current clinical use, emphasizing the potential added value of HHQs for combination treatments.
The confirmed hit rates of 18% and 15% obtained from our compound screens of early- and late-stage GAMs, respectively (Fig. 1a), highlight the value of screening ABS-active compound libraries as a source for identifying gametocytocidal compounds. The limited overlap (20%) between early- and late-stage GAM hits (Supplementary Figure 1b) suggests that GAMs at different developmental stages express distinct sets of druggable targets, provided that technology-related differences such as incubation times, sensitivity or specific bias for selecting certain chemical classes were minimal in both screens.
Our HHQ hits GNF-Pf-5640, GNF-Pf-5660 and GNF-Pf-5668 showed potent activity against ABS parasites as well as early- and late-stage GAMs (Fig. 1c), with only minimal toxicity against mammalian cells. These compounds were inactive against cultured Pb liver-stage parasites16, possibly because Hb import and metabolism does not occur in parasitized hepatocytes.
The results provided herein also document potent in vitro HHQ activity on male GAMs and gametes, which translated into marked transmission-blocking activity in mosquitoes (Table 1 and Fig. 2). In Pb-infected mice, HHQs showed moderate efficacy: the 4-day Peters tests indicated a potent inhibition of ABS growth at 100 mg kg–1, whereas the two-dose assay showed only a limited impact on ABS parasitaemia and parasite transmission (Fig. 2). The high doses required in vivo, compared to the lower doses at which activity was observed in vitro, suggest that HHQs might be rapidly metabolized or otherwise inactivated or eliminated by the host. The ester group present in HHQs, for instance, is probably sensitive to host esterases34. These HHQs are the first hits from this class and structure–activity relationship studies and chemical optimization will be required. Enhancing HHQ transmission-blocking activity in vivo is an important goal, in light of the limited efficacy observed to date (Fig. 2). We note that earlier studies of Pb reported that a 30% reduction in oocyst numbers per infected mosquito sufficed to eliminate the parasite from a murine population after several transmission cycles35. In light of these data, HHQs constitute a promising class of compounds with transmission-blocking potential that merit pharmacokinetic and pharmacodynamic optimization.
In vitro HHQ resistance selection studies identified three mutations in PfMDR1 (Y290F, F1072L and S1075I), which were confirmed by CRISRP/Cas9 editing to be causal for ABS resistance (Fig. 1e). PfMDR1-associated resistance also extended to the GAM stage (Supplementary Table 7). We note that the Pf3k collection of over 2,500 Pf genomes collected worldwide (www.malariagen.net/apps/pf3k) contained no mutations with an AB > 0.2 at these PfMDR1 positions, suggesting that they are very rare, if not non-existent, in the field.
Structural modelling studies indicated that all three amino acids co-localize in the predicted drug-binding pocket of PfMDR1 (Fig. 3) and generally caused a closer interaction between HHQs and mutant PfMDR1 (Supplementary Tables 8–12). Further studies are required to determine the nature of the interactions between HHQs and wild-type versus mutant PfMDR1 and how this relates to HHQ potency.
PfMDR1, an ATP-binding cassette (ABC) transporter, can mediate the transport of several antimalarial drugs across the DV membrane30,36,37. In the highly acidic DV compartment, Pf proteases degrade Hb from the host RBC, providing a source of the amino acids required for protein synthesis and parasite growth38. This catabolic process triggers the formation of highly reactive haem moieties, which Plasmodium normally detoxifies through biomineralization into chemically inert Hz crystals in the DV. Several antimalarials act either by binding toxic haem species and inhibiting Hz formation (CQ or ADQ), or in the case of artemisinins by apparently being activated through interaction with haem iron39. Mutations in PfMDR1, or the chloroquine resistance transporter PfCRT that also resides on the DV membrane, are known to affect in vitro parasite susceptibility to these drugs, presumably by affecting drug transport, and are associated with an increased risk of parasite recrudescence or treatment failure in patients30,40,41,42.
Haem fractionation studies showed a different profile in parasites treated with GNF-Pf-5660 as compared to CQ32, ADQ32, MQ or LMF (Fig. 4), indicating a distinct mode of action. HHQs did not directly inhibit haem conversion into Hz, but were found to reduce the total amounts of intracellular haem iron. These data suggest that HHQs act upstream of the Hb digestion pathway by interfering with parasite-mediated uptake of host Hb, which delivers the substrate for Hz formation. Western blot data showed a reduction of total Hb levels in HHQ-treated wild-type Dd2-B2 but not HHQ-resistant PfMDR1selY290F parasites (Fig. 5), supporting an effect of HHQs on parasite-mediated Hb endocytosis.
Parasite lines resistant to CQ, pyrimethamine and PfATP4 inhibitors remained sensitive to HHQs (Supplementary Table 3), whereas HHQ-resistant lines showed an increased Pf ABS susceptibility to QN, MQ, DHA and LMF (except for PfMDR1edS1075I, which showed no difference for LMF). Susceptibility to CQ, md-CQ, md-ADQ, ferroquine and piperaquine remained unaltered in HHQ-resistant parasites (Fig. 3 and Supplementary Table 13). QN, MQ, DHA and LMF are thought to interact at some level with haem but also act outside of the DV, raising the possibility that they might engage with the Hb import and catabolism process that begins with endocytosis33,43,44,45. We speculate that the PfMDR1 mutations causing HHQ resistance might sensitize parasites to MQ and LMF by preventing their sequestration by wild-type PfMDR1 in the DV, away from their primary site of action46.
HHQ-resistant ABS parasites showed no substantial change in their susceptibility to the haem-binding drugs CQ or md-ADQ that act primarily in the DV41. Interestingly, PfMDR1 Y290F was also recently identified in selections using ACT-451840, a piperazine-based pre-clinical candidate with low nanomolar activity against Pf ABS parasites and GAMs27. Studies with the ACT-213615 analogue earlier showed direct binding to PfMDR1, identifying it as a potential target47. The unusually high levels of resistance to both HHQs and ACT-451840 conferred by each of the three mutations (Figs 1e and 3c) suggest that both compound series probably share a similar mode of action. These data also evoke the possibility that PfMDR1 might itself be a target of HHQs, a hypothesis that can be investigated using HHQ analogues to study protein interactions with PfMDR147 in cultured parasites or heterologous expression systems48.
Stage-specific profiling showed that pfmdr1 is maximally transcribed around the time of merozoite egress (www.PlasmoDB.org), suggesting that PfMDR1 is already present in early- to mid-stage rings, well before ABS parasites have formed their mature DV. This timing, supported by western blot data49, suggests that PfMDR1 might be involved in vesicle-mediated endocytosis of host Hb from the RBC en route to the developing DV. Such a role for PfMDR1 would be consistent with our observations that HHQ inhibition reduces intra-parasitic total haem iron and Hz levels in trophozoites and reduces Hb levels in late-stage rings (Figs 4 and 5).
Further studies into the stage-dependency and subcellular site of action of HHQs and of drugs that are impacted by HHQ-selected mutations in PfMDR1, such as the first-line agent LMF, should yield interesting insights into how these agents act and how to leverage this information into identifying promising drug combinations. Ideally, such combinations would exert opposing selective pressures on mutant and wild-type PfMDR1, in an effort to stall the emergence and spread of multidrug resistance. This approach is currently being used in clinical trials where artemether + LMF is being combined with ADQ (Clinical Trial NCT02453308), based on earlier reports that LMF and ADQ exert opposing selective pressures on wild-type and mutant forms of both pfmdr1 and pfcrt 40,50. In addition, the potent activity of HHQs against both ABS parasites and GAMs highlights the promise of developing this chemical class into a partner for future antimalarial combination therapies with transmission-blocking activity.
Parasite cultures and strains
Pf ABS parasites were cultured at 3–4% haematocrit (Hct) in human O+ RBCs in RPMI-1640 media, supplemented with 25 mM HEPES, 50 µM hypoxanthine, 2 mM l-glutamine, 2 g l–1 sodium bicarbonate, 0.5% (wt/vol) AlbuMAXII (Invitrogen) and 10 µg ml–1 gentamycin, in 5% O2, 5% CO2 and 90% N2 at 37 °C. For GAM cultures, Albumax was replaced with 10% human serum. 3D7 and Dd2 parasites were originally obtained from T. Wellems (NIAID, NIH). Dd2-B2 is a genetically homogenous parasite line that was cloned from Dd2 by limiting dilution. The NITD609-resistant parasites (generated in the Dd2 background) and the NF54Pfs16 line have been reported previously23,51. The NF54selY290F and FCB pfmdr1 knockdown line (KD1mdr1) and their corresponding wild-type lines were obtained in two earlier studies27,52. A list of all parasite lines generated in this study is provided in Supplementary Table 17. Parasite lines were screened by PCR for Mycoplasma every 2–3 months.
Compound screening against gametocytes
The Novartis-GNF Malaria Box consisted of 3,825 known inhibitors of ABS Pf parasites with a median IC50 of 0.53 μM (25th percentile, 0.23 μM; 75th percentile, 1.09 μM), as determined by a SYBR Green-based 72 h growth inhibition assay53, and was obtained as 10 µl of 1 mM stock solution in 100% DMSO. These compounds were independently screened against early- and late-stage GAMs (stages I and IV, respectively), as described below in the two subsequent sections. Screening was performed with the NF54Pfs16 line that expresses a green fluorescent protein (GFP)–LUC fusion under the pfs16 promoter that is active throughout gametocytogenesis23. Earlier studies have reported the optimization of high-throughput early- and late-stage GAM screens using this line in LUC- or GFP-based assays54,55,56. For the LUC-based screens with early-stage GAMs, compounds were diluted 1.5-fold in 100% DMSO and further diluted in water to a final assay concentration of 2.6 µM in 0.4% DMSO. Hits, defined as compounds that showed ≥50% inhibition at 2.6 µM, were retested for confirmation at 2.6, 1.3 and 0.65 µM. For the GFP high-content imaging-based screens with late-stage GAMs, compounds were diluted to a final assay concentration of 2 µM in 0.4% DMSO and screened in triplicate. Due to limitations in the availability of the screening hits, not all hits could be retested for confirmation. Fresh batches of 125 hits from the late-stage screens were obtained as 2 mM 100% DMSO stock solutions from GNF-Novartis, and 24 additional hits were purchased as solids from various commercial sources (Supplementary Table 1).
Luciferase-based early-stage gametocyte inhibition assay
LUC-based screening with stage I GAMs was conducted as previously described13. Briefly, NF54Pfs16 parasites were synchronized by two cycles of 5% sorbitol treatment, and gametocytogenesis was induced by applying multiple known stimulant stress factors (a drop in Hct, higher parasitaemia and spent media). After one replication cycle under these conditions, the resulting stage I GAMs and asexual rings were separated from residual Hz-containing and later-stage GAMs by paramagnetic purification over a MACS column (Miltenyi Biotech). Medium containing 50 mM N-acetyl-d-glucosamine (to eliminate ABS parasites) was provided and refreshed daily. On day 1 of gametocytogenesis, which corresponds to day 0 of the early-stage GAM assay, the parasitaemia was adjusted to 10% and the Hct was reduced to 0.5%. Parasites were seeded under sterile conditions to a volume of 45 μl per well in 384-well white luminescence plates (CulturPlate, PerkinElmer) using a Multidrop dispenser (Thermo Scientific). Experimental compounds were added to the plates at the desired concentrations in a volume of 5 μl with a final DMSO concentration of 0.4% per assay well. Two columns of each plate were used as in-plate controls, containing 0.4% DMSO (negative control, 0% inhibition) or 5 μM puromycin (positive control, 100% inhibition), respectively. Plates were sealed with gas-exchange membranes (Breathe-easy, Sigma) and incubated for 72 h at 37 °C in 5% O2, 5% CO2 and 90% N2. On day 4 of gametocytogenesis, 25 μl of supernatant was removed from each well and replaced with 15 μl Steadylite solution (luciferase assay kit, PerkinElmer), using a Biomek FX (Beckman Coulter) high-throughput liquid handling system. Plates were sealed and incubated at room temperature for 60 min to allow for complete cell lysis, after which the plates were read using a MicroBeta Trilux luminometer (PerkinElmer).
Image-based late-stage gametocyte inhibition assay
The effect of compounds on stage IV GAMs was determined using an image-based assay54. Briefly, at day 8 post-induction, stage IV NF54Pfs16 GAMs were purified over a MACS column and adjusted to 10% parasitaemia and 0.1% Hct. Parasites were dispensed in 45 µl volumes into 384-well imaging plates containing 5 µl of diluted compounds per well, with a final DMSO concentration of 0.4%. After 72 h incubation at 37 °C in 5% O2, 5% CO2 and 90% N2, 5 µl of 0.07 µg ml–1 MitoTracker Red CM-H2XRos (Invitrogen) in PBS was added and incubated overnight. Parasites were imaged on an Opera QEHS confocal imaging system (PerkinElmer). Image analysis using an Acapella-based script determined the number of viable mature GAMs per imaged area of each well, as described previously54. Controls included 5 µM puromycin. Experiments were performed with triplicate technical replicates on three independent occasions.
Stage-specific potency data were generated using LUC and imaging-based assays for early- and late-stage GAMs, respectively, as described above. Relevant parasite stages were exposed to serial dilutions of compounds, typically with a maximal concentration of 40 µM, in 14 to 21 serial dilution steps with three concentrations per log dose. Assays were repeated on two independent occasions. Reference compounds (puromycin, DHA, CQ, pyrimethamine and primaquine) were used as positive and negative controls and yielded IC50 values (Supplementary Table 2) that were consistent with earlier reports13,55.
Imaging-based ABS high-throughput assay
Hit compounds with confirmed gametocytocidal activity were also evaluated for ABS activity using our imaging-based high-throughput assay57. Briefly, sorbitol-synchronized Pf parasite lines were seeded at 2% parasitaemia and 0.3% Hct in 384-well clear-bottom imaging plates (CellCarrier, PerkinElmer). For this work we used 3D7 (drug-sensitive), Dd2 (resistant to CQ, MQ and pyrimethamine) and NITD609-RDd2 clone #2 (resistant to the PfATP4 inhibitor now termed KAE60951). Parasites were incubated with compounds for 72 h at 37 °C in 5% O2, 5% CO2 and 90% N2. Parasites were then permeabilized using 0.01% Triton-X100 and 0.001% saponin plus 5 mM EDTA and stained using 0.5 μg ml–1 4′,6-diamidino-2-phenylindole (DAPI). Plates were kept at room temperature in the dark for a minimum of 6 h and were then imaged on an Opera confocal microplate system (PerkinElmer) using a 20× water objective with 405 nm excitation and 420–490 nm bandpass emission filter and an exposure time of 80 ms. The experiment was performed with two independent cultures. Parasite numbers, calculated based on spot size, intensity and fluorescence over background, were determined using Acapella software57.
Cytotoxicity of the selected gametocytocidal hits was assessed against the HEK293 human embryonic kidney cell line (ATCC #CRL-1573), maintained in DMEM medium (Life Technologies) supplemented with 10% fetal calf serum (Gibco). Cells were trypsinized, seeded into sterile 384-well plates (Falcon) at 2,000 cells per well in 45 μl medium and incubated overnight at 37 °C in 5% CO2 and 95% humidity to allow adhesion. A resazurin-based metabolic assay was used to determine cell viability58. Briefly, test compounds were serially diluted and added to the plates to generate a dose-response with a maximal concentration of 40 μM, in wells containing a final DMSO concentration of 0.4%. Plates were incubated for 72 h under standard conditions, after which the medium was removed and replaced with 35 μl of serum-free DMEM containing 44 μM resazurin. The plates were then incubated for a further 4–6 h under standard conditions before being read on an EnVision plate reader (PerkinElmer) using 530 nm/595 nm fluorescence excitation/emission settings. Two biological replicates were performed. The cells were certified by ATCC as being Mycoplasma-free at source.
Dual gamete formation assay
The effects of GNF-Pf-5310, GNF-Pf-5640, GNF-Pf-5660 and GNF-Pf-5668 on the formation of male and female gametes were determined using the DGFA56. Compounds were tested across a range of 12 concentrations (0.84 nM to 20 μM) and IC50 values were calculated using nonlinear regression (Graphpad Prism 6). Briefly, stage V GAMs were exposed to compounds for 24 h, after which gamete formation was triggered in the presence (carry-over format) or absence (wash-out format, three washes) of compound. Male gamete formation was assessed by automated microscopic detection of exflagellation and female gamete formation was evaluated by live immunostaining for Pfs25 surface expression59. Data were calculated from four biological replicates, with methylene blue included as a control.
Late-stage GAMs (12 days post GAM induction) were treated for 3 days with GNF-Pf-5660, GNF-Pf-5668 or GNF-Pf-5310 at 1× or 3× their late-stage GAM IC50 values (with daily medium changes including compound refreshment) and then maintained without compound for another 4 days. Methylene blue was used as a positive control at a concentration of 3× the GAM IC50. GAMs treated with 0.25% DMSO were used as the negative control23. Stage V GAM cultures at 50% Hct and 0.01–0.16% parasitaemia were exposed to A. stephensi mosquitoes (4–6 days post-emergence) for 30 min using an artificial membrane feeding apparatus. For each compound and controls, two containers holding 25 to 35 parasite-exposed mosquitoes each were maintained at 26 °C, 78% relative humidity with a 12 h day/night cycle. Seven or eight days post feeding (for experiments 1 and 2, respectively), the midguts from blood-fed female mosquitoes (n = 17–21 per container) were dissected and stained with 0.2% mercurochrome to count oocyst numbers. Infection prevalence was defined as the percentage of mosquitoes harbouring one or more midgut oocysts, and infection intensity was defined as the mean number of oocysts per mosquito. Midguts of mosquitoes fed with DMSO vehicle-treated GAMs showed an average of 19 oocysts. All samples were blinded for both dissection and oocyst counting.
In vivo potency of HHQs on ABS parasites
For our 4-day Peters test, four female CD-1 mice (Taconic; 5 weeks old, weighing 22–25 g) per group were infected intraperitoneally with 1 × 106 P. berghei ANKA MRA868 ABS parasites. Treatments with 70 or 100 mg kg–1 of GNF-Pf-5660 or GNF-Pf-5668, 30 mg kg–1 DHA or solvent-only (20% DMSO in water) were administered intraperitoneally 2 h, 24 h, 48 h and 72 h after infection. Blood-stage parasitaemias were assessed at 96 h post infection by microscopy. Plasma compound concentrations were not measured. All experiments involving mice were reviewed and approved by the Columbia University Institutional Animal Care and Use Committee. Sample size was determined in accordance with published literature60,61 and to minimize the number of animals under experiment. All research animals were randomly allocated into treated and control groups and groups were blinded for collection and analysis of data. This research is compliant with federal regulations detailed in the Animal Welfare Act (AWA) (7 U.S.C. 2131, seq.) and the Health Research Extension Act of 1985 (Public Law 99-158).
In vivo transmission-blocking effect of HHQs
To assess the in vivo transmission-blocking effect of HHQs, four female CD-1 mice were treated with two doses of 100 mg kg–1 GNF-Pf-5660 or GNF-Pf-5668, 30 mg kg–1 DHA, or solvent only. Doses were spaced 12 h apart, starting when blood-stage parasitaemia after infection with 5 × 106 P. berghei ANKA MRA868 parasites reached 1%, to measure drug impact only on GAMs. We note that in P. berghei GAMs develop and become transmissible within a 24 h period. Twelve hours after the last dose, both blood-stage parasitaemia and the level of inhibition of male GAM exflagellation were assessed. For the exflagellation test, a tail blood sample was collected in a heparinized tube, mixed with ookinete medium (RMPI medium, 25 mM HEPES, 50 mg l–1 hypoxanthine, 2 g l–1 sodium bicarbonate and 0.5% Albumax) containing 100 μM xanthurenic acid and evaluated by microscopy after a 10 min incubation at 20 °C. Shortly afterwards, 25–30 A. stephensi mosquitoes were fed on individual mice that were anaesthetized by intraperitoneal injection of 200 μl of a mixture of ketamine (17.5 mg ml–1) and xylazine (2.5 mg ml–1). Unfed mosquitoes were removed, and engorged mosquitoes were fed daily with a 10% sugar solution and maintained at 20 °C at 80% humidity for 10 days. Midguts of 10–20 mosquitoes fed per mouse were then dissected and stained with 0.5% mercurochrome for oocyst counting by microscopy. Data were analysed in a similar manner as for the membrane-feeding assay (see section 'Membrane-feeding assay').
Selections for HHQ resistance
To select for Pf resistance to GNF-Pf5640, GNF-Pf-5660 and GNF-Pf-5668, triplicate flasks of 109 Dd2-B2 parasites were exposed to compounds at 5× their ABS IC50 values. Briefly, medium containing the compound of interest was refreshed daily during the first 7 days, while cultures were carefully monitored to avoid overgrowth and ensure that sensitive parasites were effectively cleared. RBCs (0.5% Hct) were added on day 7 and the total culture volume was reduced by 25% each subsequent week until a 5 ml volume was obtained. Fresh RBCs (1% Hct) were added on a weekly basis from day 14 onwards, always maintaining a total Hct level of 3–4%. Once cultures appeared negative, the medium containing the compound of interest was refreshed every other day until at least 1% parasitaemia was reached, after which medium was replenished daily. Resistant clones were obtained by limiting dilution62. DNA was extracted with the DNAeasy Blood & Tissue kit (Qiagen) and sequenced using paired-end reads on a Illumina HiSeq 2500 platform27. Whole-genome sequence analysis was performed at UC San Diego and Columbia University, using previously described customized pipelines and the GATK toolkit26,63,64.
pfmdr1 point mutations were introduced into Dd2-B2 parasites using a two-plasmid CRISPR–Cas9 targeted gene-editing strategy27. The Cas9 plasmid encoded the Cas9 nuclease, a single guide RNA (5′-GAAAATTAATGTCCTTAAA for F1072L and S1075I and either 5′-ATTTTAGTTTCTTATGCATT or 5′-TTACATATAGGTTTAATAAA for Y290F) that complexes the Cas9 nuclease with the target sequence, and a human dhfr cassette that mediates resistance to the WR99210 selection agent27. The donor plasmid contained the homologous pfmdr1 sequence (nucleotides 1871–3421 of the Pf3D7_0523000 coding sequence for the mutations at amino acid position 1072 and 1075, and nucleotides 383–1034 for the mutation at amino acid position 290) harbouring the mutation of interest, as well as eight synonymous mutations to prevent the Cas9 nuclease from cleaving the edited gene sequence, and the blasticidin S deaminase gene cassette that mediates resistance to blasticidin. Ring-stage cultures were transfected with the Cas9 and donor plasmids as described in ref. 27. At 24 h after electroporation, blasticidin (2 μg ml–1) and WR99210 (2.5 nM) selection was initiated. The medium containing these drugs was refreshed daily for 12 days. From day 13 onwards, cultures were maintained under blasticidin pressure until parasites were microscopically observed. pfmdr1-modified clones were obtained by limiting dilution, and gene editing was verified by PCR and Sanger sequencing. Transfections were performed in duplicate on one to three occasions until editing was achieved.
Drug susceptibility profiling of HHQ-resistant lines in ABS parasites
To assess antimalarial susceptibilities, we exposed ring-stage cultures (at an initial parasitaemia of 0.2% and 1% Hct) for 72 h to a range of ten concentrations (plated in duplicate), in a 96-well format. Dd2-B2 parasites were included as a parental control. Parasite survival was assessed by flow cytometry on an Accuri C6 (BD Biosciences) using SYBR Green and MitoTracker Deep Red FM (Life Technologies) as nuclear and vital dyes65. Assays were performed on three to ten separate occasions (Supplementary Table 5). Growth inhibition data were plotted using nonlinear regression to calculate IC50 values (using Graphpad Prism 6). IC50 values were compared between the selected or edited lines and their parental controls (Dd2-B2, FCB or NF54Pfs16) using an unequal variance t-test, also referred to as Welch’s t-test, to correct for differences in the number of repeats per group. A Bonferroni–Dunn correction for multiple comparisons was performed when applicable.
Drug susceptibility of HHQ-resistant lines at GAM stages
The NF54selY290F line was tested together with a wild-type NF54 parental control for GAM susceptibility to the three HHQs using a previously described parasite lactate dehydrogenase-based assay66.
The Pf 3D7 PfMDR1 amino acid sequence (accession no. XP_001351787) was used to develop a structural model. Membrane protein topology was predicted using TOPCONS67. Residues 51–350 and 781–1090, based on the topology prediction, were used for homology detection and structure prediction with the HHpred tool68. The best ranked hits (Prob = 100) were considered and the ABC transporter structure of Thermotoga maritima (PDB 4Q4H) was used to model PfMDR1, using MODELLER because of its apo state69. PfMDR1 TMDs were refined using WinCoot70 and assembled by superimposition on the 4Q4H template using the visualization software Yasara71. Docking simulations of compounds GNF-Pf-5640, GNF-Pf-5660 and GNF-Pf-5668 were performed using DockingServer72, with residues 290, 1072 and 1075 seeded in a 15 Å box at default conditions.
Haem fractionation assay
Haem fractionation analysis was performed as described in ref. 32. Briefly, early ring-stage sorbitol-synchronized Dd2-B2 and PfMDR1selY290F parasites were exposed to GNF-Pf-5660, MQ or LMF at concentrations corresponding to 0.1, 0.25, 0.5, 1 and 2× the ABS IC50 (ref. 32). After a 32 h incubation, cells were saponin-lysed and trophozoites were subjected to a series of lysis, solubilization and differential centrifugation steps to obtain fractions that contain Hb, free haem or Hz. The UV–visible spectra of haem as Fe(iii)haem-pyridine were analysed in each of these fractions and compared to a haem standard curve for accurate quantification. These quantities were subsequently normalized using the number of analysed cells, as determined by flow cytometry of an aliquot of exposed cells. Two-tailed t-tests were used to assess significance32.
Sorbitol-synchronized parasites were exposed to 5× the IC50 of GNF-Pf-5660, WR99210, MQ or LMF for 5 h, starting at the late-ring stage, that is, before parasites entered the mid-stage trophozoites that have maximal rates of Hb degradation and Hz formation. RBCs were then saponin-lysed and parasites washed five times in 1× PBS. The parasite pellet was solubilized in SDS–PAGE sample buffer and sonicated for ten cycles of 10 s. Samples were then boiled for 10 min at 90 °C and 3.5 × 106 parasites were loaded on a 4–20% Criterion TGX precast midi protein gel (Bio-Rad). After transferring the resolved proteins to a nitrocellulose membrane, membranes were blocked for 1.5 h in TBS-Tween 0.1% (TBS-T) + 5% blocking-grade blocker (Bio-Rad). Membranes were cut in two at the 37 kDa marker position, and the lower and upper parts were incubated overnight at 4 °C in TBS-T + 1% blocker and a 1:5,000 dilution of rabbit anti-human Hb antibody (Sigma #H4890, lot 124M4788V) or a 1:2,000 dilution of β-actin monoclonal antibody (Fisher #AM4302), respectively. After washing in TBS-T, membranes were incubated in TBS-T +1% blocker and a 1:10,000 dilution of goat anti-rabbit IgG conjugated to horseradish peroxidase (GE Healthcare #RPN4301) to detect Hb, or a 1:10,000 dilution of sheep anti-mouse Ig coupled to horseradish peroxidase (GE Healthcare #NXA931) to detect β-actin. Proteins bands were detected using the Pierce ECL western blotting substrate (ThermoScientific), blue basic autoradiography film (Genemate) and a Nikon X-OMAT 2000. Hb band intensities were calculated in ImageJ, normalized for loading with the β-actin signal, and compared to untreated controls. Experiments were repeated on three to four independent occasions.
All data generated or analysed during this study are included in this published article and its Supplementary Information. Requests for materials should be addressed to the corresponding author.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors thank T.T. Diagana (Novartis Institute for Tropical Diseases, Singapore) for provision of the compounds, the Red Cross (Australia and the USA) for the provision of human blood for cell cultures, and G. Stevenson for assistance with the triaging of compounds following screening. The authors acknowledge the Bill and Melinda Gates Foundation (grant OPP1040399 to D.A.F. and V.M.A. and grant OPP1054480 to E.A.W. and D.A.F.), the National Institutes of Health (grant R01 AI103058 to E.A.W. and D.A.F., grant R01 AI50234 to D.A.F, and R01 AI110329 to T.J.E.), the Australian Research Council (LP120200557 to V.M.A.) and the Medicines for Malaria Venture for their continued support. P.E.F. and M.I.V. are supported by the Northern Portugal Regional Operational Programme (NORTE 2020), under the Portugal 2020 Partnership Agreement, through the European Regional Development Fund (FEDER).