High-Throughput Assay and Discovery of Small Molecules that Interrupt Malaria Transmission

1Malaria is a vector-borne disease caused by apicomplexan eukaryotic protozoa of the genus Plasmodium. The parasites have a complex life cycle involving vertebrates and anopheline mosquitoes. Their asexual replication and destruction of erythrocytes give rise to the symptoms of malaria, including fever and chills. In response to cues that are not well understood, a subset of asexual parasites differentiate into male and female gametocytes, a process that takes ∼8–12 days in P. falciparum. During this period, the parasites metabolize the host red cell hemoglobin, while progressing through five morphologically distinct stages that can be identified by light microscopy. Commitment to sexual development occurs well before parasites show morphological changes, and male and female gametocytes are produced at a ratio of 1 to 3–5 with females maturing slightly later. In the human body, immature gametocytes sequester in different host tissues and emerge only when fully mature. An infected individual may carry gametocytes for up to 55 days, and mature gametocytes are the only form that can survive in the mosquito midgut, mate, undergo meiosis, and give rise to the next generation of parasites to be transmitted to a new human host.

2Current first-line treatment of falciparum malaria is artemisinin combination therapies (ACTs), which do not block transmission. Follow-up treatment with 8-aminoquinolines like primaquine or tafenoquine is needed to block transmission. However, 8-aminoquinolines can be toxic to individuals with glucose-6-phosphate dehydrogenase deficiency, a genetic condition with a high prevalence in malaria-endemic regions.

3Even though assays are available to detect compounds with transmission-blocking potential, most of them are not adapted for very large chemical libraries due to multiple purification steps or lower throughput formats. In addition, some assays rely on the use of gametocyte reporters that may restrict their use to genetically modified parasites. Here we describe high-throughput assays that overcome these issues. We apply the assays to characterized and uncharacterized chemical libraries. Our analysis reveals features of chemical compounds that are likely to block malaria transmission and may serve as starting points for unique transmission-blocking drugs.

4To create a homogeneous, stage-specific gametocyte population, we optimized a previously described protocol and induced gametocytogenesis in asexual, triple synchronized P. falciparum NF54 parasites by high parasitemia and partly spent media (Figure 1A; Experimental Procedures). Microscopic staging of gametocytes collected over the 12 days of development according to description by showed purities upward of 75% per stage (Figure 1C) with a reproducible parasitemia of 1.2%–1.6% over the screening period (data not shown).

5To detect viability, we used the dye MitoTracker Red CMXRos (MTR Red), which fluoresces at ∼600 nm in parasites with intact mitochondrial membrane potential (Figure 1B). Parasites were detected using automated microscopy and showed a good correlation (R2 = 0.99) between the number of viable parasites added per well and the number of MitoTracker Red CMXRos positive objects (Figure 1E).

6To reduce the number of liquid transfer steps and make the assay more robust and less costly for use with large, unbiased libraries, we experimented with the use of saponin, an amphipathic glycoside that creates pores in red cell membrane bilayers, leading to red cell lysis. We found that treating gametocyte cultures with 0.13% saponin caused red blood cells in serum-free media to lyse, simplifying the identification of parasites with automated microscopy. Gametocytes at a parasitemia of 0.5% to 0.75% and a hematocrit of 1.25% created a monolayer on the bottom of the well. After MTR Red staining ∼1,000 objects could be counted per DMSO control well (1,536) (Figure 1E). This allowed compound exposure and imaging in the same plate without an additional transfer step. We refer to this serum-free one-step protocol as Saponin-Lysis Sexual Stage Assay (SaLSSA). We found SaLSSA to be more sensitive to few compounds like the amino alcohols (see below). Thus, an older, serum-containing assay (two-step sexual stage assay or TSSA) was used in some cases.

7The quality of the assay was found to be robust at all gametocyte stages: Z prime scores calculated with infected, DMSO-treated red blood cells and uninfected red blood cells ranged from 0.71 to 0.80 for the one-step protocols (Figure 1D). We evaluated the fluorescence intensity over time and did not observe significant differences between 30 and 360 min (data not shown).

8To further benchmark SaLSSA, we evaluated 50 compounds currently used as antimalarials or antimalarial tool compounds in dose response against individual gametocyte stages. EC50 values of different chemical classes showed distinct patterns of activity for the different gametocyte stages as summarized below (Tables 1 and S1). Most compounds yielded higher EC50 against stage V gametocytes.

9Endoperoxides were characterized by low nM EC50 values for stages I–IV with most failing to generate a dose-response curve for stage Vs, in agreement with standard membrane-feeding data: Although DHA, artesunate, and OZ439 have been reported to have some activity in standard membrane-feeding assays (SMFAs), none completely eliminated oocysts at 100 nM, and only OZ439 eliminated oocysts at 1 μM, a concentration well above the concentrations in blood when being used against blood-stage infections. Interestingly, these results correlate well with previous publications reporting that hemoglobin digestion ends at stages III to IV, supporting the endoperoxides’ activity against this process.

10The 4-aminoquinolines (chloroquine, piperaquine, pyronaridine, napthoquine, hydroxychloroquine, and AQ13) demonstrated low nM EC50 values for stages I and II but showed a drop-off in activity beginning in stage III and were ineffective against stage Vs. It is generally thought that 4-aminoquinolines interfere with the formation of hemozoin, resulting in the death of the parasites, and previous reports have shown that chloroquine is only active against early-stage gametocytes in line with low activity in SMFAs.

11The 8-aminoquinolines showed consistent but high (μM) EC50 values across all gametocyte stages. Primaquine, the only drug approved for blocking transmission, exhibited a 6.5 μM EC50 against stage V gametocytes in vitro. Tafenoquine, a primaquine derivative, had an EC50 of 2.4 μM against stage V gametocytes. The substantially higher EC50 than the ED50 values were expected because 8-aminoquinolines need to be metabolized for activity. The mechanism of action of primaquine as well as the identity of the active metabolites remains unknown.

12Amino alcohols are not used to prevent transmission, but some amino alcohols (lumefantrine, mefloquine, and halofantrine, but not quinidine and quinine) showed activity across all stages using SaLSSA. We note, however, that this sensitivity could be completely reversed by the addition of human serum to cultures during compound incubation using the TSSA. Mefloquine only shows SMFA activity at high concentrations (10 μM), ∼20× above the asexual growth inhibition values (50 nM). Testing some compounds in the presence of human serum would be expected to give more accuracy and eliminate false positives, but at the expense of efficiency.

13The antifolates and sulfonamides, which interfere with nucleic acid synthesis and include dihydropteroate synthase (DHPS) and dihydrofolate reductase (DHFR) inhibitors (pyrimethamine, P218, and chlorproguanil), were inactive across all gametocyte stages, except for chlorproguanil, which had μM EC50 values against all stages. Antibiotics used to treat malaria, such as doxycycline and azithromycin, which act against the apicoplast, were also inactive. As expected, atovaquone was inactive (Table 1). Evidence that this compound has some transmission-blocking activity in animals after repeated exposure may be because it inhibits ookinete formation.

14Some antimalarial compounds that are not used in humans did show activity against stage V gametocytes in our assay (Table 1). Thiostrepton is a macrocyclic thiopeptide antibiotic that inhibits prokaryotic translation and has been reported to dually target the proteasome and apicoplast. Thiostrepton had μM EC50 values at all stages. The mode of action of methylene blue remains controversial, but it may inhibit glutathione reductase or hemozoin formation. It showed low nM EC50 values for stages I to IV, with some minor loss of activity at stage V, consistent with reports that methylene blue reduced transmission by 99% in SMFAs at 38 nM. Pentamidine, which is clinically used for treatment and prophylaxis of Pneumocystis carinii pneumonia (PCP) and sleeping sickness but not malaria, inhibited gametocytes of all stages with an EC50 between 0.39 and 2.14 μM. Its mechanism of action is unknown, although it has been reported to inhibit hemozoin formation in Plasmodium by interaction with ferriprotoporphyrin IX.

15Newer classes of compounds in development, including the spiroindolones, imidazolopiperazines, imidazopyrazines, and quinoline-4-carboxamides, all have reported transmission-blocking activity, and members of these compound classes were tested using SaLSSA (Table 1). KAF246, a spiroindolone closely related to the clinical candidate KAE609 (also known as cipargamin or NITD609) that acts against the plasma membrane ATPase PfATP4, showed the expected activity (EC50 = 1 to 2 nM). GNF179, an imidazopiperazine closely related to the clinical candidate KAF156, showed the expected low nanomolar activity in the SaLSSA and complete transmission-blocking activity in SMFAs at physiologically relevant concentrations of 15 nM (Figures 2A–2C). The PI(4)K-inhibitor KDU691, which inhibits transmission at 1 μM in SMFAs, had submicromolar EC50 values across all five gametocyte stages, similar to the values seen against blood stages (∼200 nM). The Plasmodium falciparum translation elongation factor 2 (eEF2)-inhibitor, DDD107498, likewise showed potent activity, in line with reported activity in other cellular and standard membrane feeding assays.

16To determine whether the loss of activity against stage V parasites was typical or reflects the historical focus on compounds derived from quinine and artemisinin, 400 compounds from the MMV malaria box were examined. These compounds, which were all identified in asexual blood stage screens, were first examined at a single dose of 12.5 μM against each gametocyte stage with TSSA. As expected, the highest number of compounds were active against early-stage gametocytes: 216 compounds inhibited the viability of stage I gametocytes by more than 70%, 78 compounds inhibited stage III gametocytes, and 79 compounds inhibited stage V gametocytes (Table S2).

17Dose-response analysis against stage I, III, and V gametocytes for the 50 most active compounds confirmed activity <5 μM for 28 of the 50 compounds with the TSSA. EC50 values for stage V were significantly higher than for stage I gametocytes for 42 of the 50 compounds (Figure 2D). A few compounds showed EC50 values of ≤1.5 μM against stage V gametocytes in both TSSA and SaLSSA (Table S2), including MMV665941 (stage V EC50 1.04 μM, SaLSSA, Figure 2B) followed by MMV019918 (stage V EC50 1.46 μM, SaLSSA). SMFA studies with MMV665941 using GNF179 as a control at a concentration of five and ten times the EC50 calculated from the above described stage V gametocyte assay showed that mosquitoes fed on the compound-exposed gametocytes had no oocysts in their midguts, whereas the DMSO control group did (Figure 2C), most likely because the compound-treated gametocytes did not exflagellate (data not shown).

18We further investigated a subset of 18 compounds (Table 2) with reported activity against gametocytes and available luciferase-SMFA data using 1,536-well SaLSSA against gametocytes stages I, III, and V. Of this control set, 14 of the 18 were active in at least one stage in SaLSSA with an EC50 of less than 10 μM, and 13 of the 14 were active in SMFAs (Figure S3; Table S5). A possible false-negative compound was MMV665882 (Figure S3G), which showed an incomplete curve in SaLSSA but little activity in the SMFA. This compound showed some activity in viability readouts by others. The four potential false positives included MMV020492 (Figure S3A), which was previously reported to have low activity against male gametes but gave no inhibition in SMFAs and was inactive in SaLSSA. Two other false positives were MMV665827 (Figure S3C) and MMV007116 (Figure S3D), which had been shown to reversibly inhibit male gamete formation. The fourth compound, MMV666021 (Figure S3B), which had been reported active in luciferase assays with late-stage gametocytes, showed some inhibition in the single-point studies but was not reconfirmed in dose response. This compound was weakly active in luciferase-based SMFAs (complete inhibition at 10 μM, partial at 1.6). The reasons for the discrepancies for this compound are unclear, but it is possible that compound source, solubility, or parasite genetic background could play a role, especially as literature-reported blood-stage values (3D7) vary from 90 nM to 2 μM.

19To further investigate the rate at which transmission-blocking compounds would be identified in sets of compounds with known blood-stage activity (EC50s of less than 10 μM), we investigated the GNF malaria box. This set of 3,558 compounds was created after screening proliferating asexual parasites at a final compound concentration of 1.25 μM. The set was screened at a single concentration of 1.25 μM against stage V gametocytes with the 384-well SaLSSA. Of these, 145 compounds (4.07%) inhibited stage V gametocytes at greater than 72.3% (Figure S1A). Dose-response analysis showed 108 of the 145 compounds reconfirmed as having activity of less than 1 μM with 22 compounds giving EC50 values below 100 nm against stage V gametocytes. Unlike the clinical antimalarials (Table 1), most of which showed a steep drop-off in activity with mature gametocytes, these scaffolds were almost all equipotent against asexual blood stages and stage V gametocytes. Some of the most active scaffolds (Figure S1B) were carbamazide thioureas as well as naphthoquinones, a compound class known to be active against gametocytes. Several of these scaffolds were also active against a P. yoelii hepatocyte development and invasion assay that predicts causal prophylactic activity (Figure S1C; Table S3).

20In order to determine the fraction of active compounds that would be found in a larger library that was not preselected for activity against asexual parasites, we tested compounds from the diversity-oriented synthesis (DOS) library. This library was designed to populate chemical space broadly with small molecules having both skeletal and stereochemical diversity.

21Two sets of compounds from the DOS compound library were screened against stage V using 1,536-well SaLSSA at 2.5 μM in duplicate (Figure S2A). The first was an “informer set,” which includes 9,886 compounds selected to represent a sampling of the structural diversity of all of the DOS scaffolds while also capturing preliminary structure-activity relationships (SARs) and stereochemical structure-activity relationships (SSARs). 25 compounds inhibited stage V gametocytes in both replicates by >30% (Table S4, hit rate 0.25%), and 17 were inconclusive (active in one of two replicates). To reconfirm and investigate the SARs as well as SSARs, 41 compounds were retested in dose-response along with 37 stereoisomers and seven analogs of select compounds. 13 of the hits and one inconclusive exhibited EC50s < 5 μM upon retest, resulting in a retest rate of 23% (or 54% for hits only).

22A second compound set included 89 compounds (representing 17 scaffolds) that had previously been shown to have activity in a blood-stage assay against P. falciparum Dd2 (EC50 < 2 μM; N.K., unpublished data). These compounds had not been further assessed for mechanism of action or additional stage-specific activity against Plasmodium prior to this study. An identical screening pipeline was used for the blood-stage active compounds; in this case, 15 compounds were identified as hits (hit rate 16.9%). These hits and two additional stereoisomers were retested at dose-response, whereupon ten of the compounds exhibited EC50s < 5 μM (retest rate 67%).

23Taken together, the 35 hits encompassed 12 different scaffolds, with five singletons and seven scaffolds with two or more representatives. Representatives from six of these scaffolds are shown in Figure S2C; one scaffold was eliminated due to a lack of SSARs and SARs in the hits. While the activity needs to be validated with resynthesized compounds, some of these do show interesting patterns of activity, including one compound with greater activity against gametocytes (BRD0608), four compounds with activity across all three parasite stages, and one compound (BRD1260) with activity against just gametocytes and the asexual blood stages (Figure S2B). Additional studies will be of interest to validate these data and investigate the mechanisms of action of these compounds.

24To further validate the screens and to identify compounds that can serve as starting points for the development of the transmission-blocking drugs, all compounds that had been screened were hierarchically clustered based on their scaffold similarity. We then identified clusters of structurally related compounds that showed enrichment in the sexual-stage active set at rates higher than expected by chance (Figure 3). For example, a cluster with the highest enrichment score (enrichment log10p = −16.11) consists of 13 compounds related to GNF-Pf-3202 and GNF-Pf-3600 (dioxonapthalen-acetamides), with 10 out of 13 compounds being active in the gametocyte assay. Another enriched cluster contains 21 compounds structurally similar to GNF-Pf-5511 and GNF-Pf-5386 (tetrahydroisoquinoline-4-carboxamide scaffolds that are related to the PfATP4 inhibitor, (+)-dihydroisoquinolones, (+)-SJ733) with seven of the 21 compounds being active (log10p = −7.01). This is not unexpected, given that other PfATP4 inhibitors are active against late-stage gametocytes and (+)-SJ733 potently blocks transmission. A final scaffold family that is highly overrepresented contains four of the five 2-furancarboxamides (GNF-Pf-1329, GNF-Pf-3542, GNF-Pf-783, GNF-Pf-1696, and GNF-Pf-2740) in the library (log10p = −7.35), all of which showed moderate activity against stage V gametocytes (0.61 to 0.484 nM) but weaker activity against asexual blood stage parasites (1.72 to > 10 μM). To our knowledge, this compound class has not previously been associated with blocking transmission. These data suggest that screens of very large libraries will likely yield starting points for the discovery of transmission-blocking drugs.

25The present data differ from those that have been reported by a number of other laboratories. Some assays have shown that compounds such as artemether and OZ439 have late-stage gametocytocidal activity of less than 1 μM. The consensus is that mature gametocytes are resistant to endoperoxides. Most previous reports combined gametocyte stages for late-stage gametocyte testing (stages III-V or IV-V), which might account for conflicting compound activity. Another consideration is that the readout of different gametocytocidal assays might vary with the mode of action of certain drugs, depending on which biological pathway the specific assay is interfering with.

26Overall, our data suggest that low-cost SaLSSA gives few, if any, false positives, compared to available SMFA data. On the other hand, the SaLSSA assay may give a few false-negatives, and will likely miss reversible inhibitors of gamete formation.

27In the majority of cases, our data showed that stage V gametocytes had a lower susceptibility to compounds than stage I gametocytes, suggesting decreased metabolic activity during their maturation in preparation for subsequent development in the mosquito midgut. The acquired standard membrane-feeding data, which still represent the gold standard for transmission-blocking activity, suggest that compounds that inhibit stage V gametocytes can block transmission as well. On the other hand, the SMFAs may also find compounds that our cellular assay would miss, including compounds that have a contraceptive effect. Given that fertilization occurs in the mosquito midgut over minutes while late-stage gametocytes can persist in the human body for days, stage V gametocytes are arguably more attractive therapeutic targets.

28The data from this study indicate which proteins and pathways might be targeted by transmission-blocking drugs (Figure 4). Compounds that interfere with hemoglobin digestion would be poor candidates for transmission-blocking drugs, although they could yield a reduction in gametocyte numbers as early-stage gametocytes might be killed. In addition, the process of DNA replication should probably not be targeted, nor apicoplast function (azithromycin and doxycycline). Targets for compounds that act against mature gametocytes include proteins that play a role in protein translation (the targets of puromycin, DDD107498, cycloheximide, thiostreptin) and processing, including protein secretion (GNF179), as well as protein degradation (e.g., epoxomycin; ). Interestingly, functional genomic studies had previously shown that gametocytes acquire and store RNA transcripts that rapidly convert to proteins during gamete formation, creating a particular vulnerability. Targets involved in maintaining ion homeostasis, such as PfATP4 as well as lipid kinases (e.g., PI4K), are transmission-blocking targets as well as asexual stage targets.

29It should be noted that all these targets are also essential for asexual parasites. Targeting exclusively gametocytes could be achieved through inhibiting translational repression, autophagy, sperm function, as well as meiosis. It is expected that compounds inhibiting these processes would be found at lower rates in large libraries, emphasizing the need for ultra-high-throughput screens. One example is BRD0608, whose EC50 against asexual blood stages was 15× higher than against stage V gametocytes and whose selectivity for sexual stages might be improved through medicinal chemistry.

30The advantage of compounds like BRD0608 is a reduced potential for emergence of drug resistance. There are billions of asexually replicating parasites in an infected human, each of which has the capacity to develop a drug resistance mutation and pass it on to their progeny. This has been a major reason why malaria control is so difficult.

31An open ethical question is whether drugs, which do not relieve malaria symptoms but benefit the community as a whole, should be licensed. Vaccines may be given that provide little benefit to an individual—for example, the rubella vaccine in boys, mainly recommended to cohort-protect pregnant women in order to prevent congenital rubella syndrome of newborns. Vaccines are also not without risk, and one could argue that the benefit to humanity that would be achieved with malaria eradication would outweigh the risk.

32Asexual P. falciparum parasites (NF54) were grown at 5% hematocrit in O+ human erythrocytes in serum-containing complete media (RPMI 1640, gentamicin 0.05 mg/ml, hypoxanthine 0.014 mg/ml, HEPES 38.4 mM, sodium bicarbonate 0.2% [w/v], D-glucose 0.2% [w/v], sodium hydroxide 3.4 mM, 4.3% [w/v] heat-inactivated human serum [O+] and 0.2% [w/v] AlbuMAX II) at 37°C under low-oxygen conditions (3% O2, 5% CO2, and 92% N2) and a parasitemia between 0.5% and 3%. Ring-stage parasites were triple synchronized (d-8, d-6, d-4) with 5% (w/v) D-Sorbitol, and cultures were expanded from T25 to T225 culture flasks. The hematocrit was adjusted to 5% until day −4. On day −2, only 50% fresh media was substituted during a high parasitemia of 7%–10%. Media was exchanged daily from day −1 onward. For stages I–IV, magnetically activated cell sorting (MACS) was performed on day 0 and cultures were sorbitol synchronized on day 1. All gametocytes were treated with 50 mM NAG on days 0–9. See Supplemental Experimental Procedures for more details.

33Gametocyte stages I–V were diluted to 0.50% gametocytemia and 1.25% hematocrit into complete media for the two step protocol (TSSA) or 0.5%–0.75% gametocytemia and 1.25% hematocrit into serum-free SALSSA screening media (RPMI 1640, gentamicin 0.05 mg/ml, hypoxanthine 0.014 mg/ml, HEPES 38.4 mM, sodium bicarbonate 0.2% [w/v], D-glucose 0.2% [w/v], sodium hydroxide 3.4mM and 0.4% [w/v] AlbuMAX II). Cultures were dispensed (40 μl versus 10 μl) into 384 or 1,536-well plates containing 50 nl or 2.5 nl of compound (final concentration of 1.25 to 12.5 μM) using a MultiFlo dispenser. Plates were incubated at 37°C for 72 hr under low-oxygen conditions. For SaLSSA 3 μl (1,536 well) or 10 μl (384 well) of 2.5 μM MitoTracker Red CMXRos and 0.13% saponin solution (w/v) in screening media was added to each well, and plates were incubated for 60–120 min at 37°C. For 384-well TSSA, 5 μl MitoTracker Red CMXRos (5 μM) in screening media was added to each well. After 20 min at 37°C, 5 μl was transferred from the assay plate to a new 384-well imaging plate, that already contained 40 μl MitoTracker Red CMXRos (500nm) in serum-free screening media. For both TSSA and SaLSSA, plates were imaged after 30 min incubation.

34Imaging of 384- or 1,536-well plates was performed using a high content imaging system (Operetta, Perkin Elmer) and Harmony software for image analysis. Viability indices were calculated by dividing the particle count of each compound-treated well by the average particle count of the DMSO wells per plate and range from 0 (active compound) to >1 (inactive). Z values were calculated using DMSO-treated gametocytes as positive and uninfected red blood cells as negative wells.

35P. falciparum NF54 parasites were grown at 0.5% parasitemia and 5% hematocrit and continuously cultured with daily media changes until they reached stage V. After incubation for 24 hr with compound in DMSO, the SMFA was performed. Briefly, 4- to 6-day-old female A. stephensi STE 2 mosquitoes were fed with the treated gametocytes for 15 min using a membrane feeding apparatus. Midguts were dissected after 8 days, and the number of oocysts counted.

36P. falciparum NF54-L1 (hsp70-luc reporter) stage V gametocytes were pre-incubated for 24 hr with compound in six dilutions in duplicate. DMSO was used as negative and DHA as positive control. The compound was washed out, and the gametocytes were fed to Anopheles stephensi mosquitoes. At day 8 post-infection, luminescence signals were determined for 24 individual mosquitoes per cage. EC50s were determined by applying a four parameter logistic regression model. The SMFAs, which were organized by a consortium of laboratories including this one, have been deposited at ChEMBL-NTD (https://www.ebi.ac.uk/chemblntd) and were performed by TropIQ in Nijmegen, The Netherlands.

37Liver stage assays were performed as previously described. Briefly, 103 P. berghei luciferase-expressing sporozoites (New York University Insectary) were used to infect HepG2-A16-CD81EGFP cells (pretreated with compound at 1.25 μM) in a 1,536-well plate. After incubation for 48 hr, 2 μl BrightGlo (Promega) was added, and the EEF growth was quantified by bioluminescence on an Envision Multilabel Reader (PerkinElmer).

3813,844 tested compounds were clustered using the Scaffold Tree algorithm. Each scaffold node was then assigned an enrichment score reflecting the degree of overrepresentation of active compounds (stage V gametocyte inhibitors). We calculated the accumulative hypergeometric p value as probability of observing at least as many hits as we observed within each scaffold. The tree was then pruned, so that only scaffolds with p values < 0.001 were retained. The final resultant tree in Figure 3 was rendered with Cytoscape (version 3.2.0). To focus on those nodes where the scaffold of a node could reasonably resemble the full structures of all associated compound members, the average Tanimoto similarity score between each scaffold node and its associated compounds were calculated based on ChemAxon topological fingerprints (ChemAxon, Kft.), and those tree nodes and leaves with at least 0.85 Tanimoto scores and with at least three hits are highlighted in colors in Figure 3.

39D.M.P. and M.W. designed experiments, performed screens and dose response assays, analyzed data, and wrote the manuscript; A.Y.D. prepared parasite materials; S.M. performed P. berghei liver stage assay; F.L., K.P., and A.L. performed standard membrane feeding assays; S.L.O. designed experiments; E.L.F. analyzed data; O.T. and Y.Z. performed compound clustering; E.C. performed data analysis; N.K. performed Dd2 blood stage assays; C.A.S. assisted with experimental design analyzed data; S.L.S. and D.L. performed data analysis; and E.A.W. designed experiments, analyzed data, and wrote manuscript. All authors edited the manuscript and contributed to writing.