Metabolic Dysregulation Induced in Plasmodium falciparum by Dihydroartemisinin and Other Front-Line Antimalarial Drugs

Abstract

Detailed information on the mode of action of antimalarial drugs can be used to improve existing drugs, identify new drug targets, and understand the basis of drug resistance. In this study we describe the use of a time-resolved, mass spectrometry (MS)–based metabolite profiling approach to map the metabolic perturbations induced by a panel of clinical antimalarial drugs and inhibitors on Plasmodium falciparum asexual blood stages. Drug-induced changes in metabolite levels in P. falciparum–infected erythrocytes were monitored over time using gas chromatography–MS and liquid chromatography–MS and changes in specific metabolic fluxes confirmed by nonstationary [¹³C]-glucose labeling. Dihydroartemisinin (DHA) was found to disrupt hemoglobin catabolism within 1 hour of exposure, resulting in a transient decrease in hemoglobin-derived peptides. Unexpectedly, it also disrupted pyrimidine biosynthesis, resulting in increased [¹³C]-glucose flux toward malate production, potentially explaining the susceptibility of P. falciparum to DHA during early blood-stage development. Unique metabolic signatures were also found for atovaquone, chloroquine, proguanil, cycloguanil and methylene blue. We also show that this approach can be used to identify the mode of action of novel antimalarials, such as the compound Torin 2, which inhibits hemoglobin catabolism.

Plasmodium, malaria, antimalarial, artemisinin, metabolomics, chloroquine, drug screening, metabolite, torin, atovaquone

Identifying novel antimalarials remains a priority for effective malaria control and reducing the burden of disease caused by Plasmodium spp. Although the number of lead antimalarials entering the development pipeline is increasing and novel classes of compounds that inhibit Plasmodium growth are becoming available [1–4], information on the mode of action of these compounds is limited in most cases. Understanding the mode of action of current and newly identified antimalarials is crucial in directing hit-to-lead optimization, informing appropriate dosing and selection of combination partners, and developing strategies to preempt the emergence of resistance.

A number of approaches have been used to identify novel antimalarial targets and resistance mechanisms, including resistance screening, whole-genome sequencing, and analysis of changes in the transcriptome and/or proteome [5–8]. Metabolomic approaches provide an alternative and potentially complementary approach for elucidating drug targets [9]. Metabolomics is well suited to the identification of drug mechanisms in microbial pathogens, because many antimicrobials, including common antimalarials, directly target metabolic enzymes [10]. In addition, metabolic fluxes are highly sensitive to changes in other processes, such as cell division, cellular stress (eg, oxidative stress), and nonenzymatic processes. Metabolomics approaches have therefore proved effective at identifying specific drug targets and the wider metabolic consequences of drug action in protozoan pathogens [11].

In Plasmodium falciparum, the causative agent of the most severe human malaria, targeted metabolite profiling has been used to investigate polyamine inhibitors, the 1-deoxy-D-xylulose 5-phosphate pathway (DOXP) inhibitor fosmidomycin, and a novel mitochondrial respiratory chain inhibitor [12–14]. In this article, we describe an untargeted, dual mass spectrometric (MS) approach to map drug-dependent perturbations in the metabolic network of the P. falciparum–infected erythrocyte. Coverage of the parasite/host cell metabolome was maximized using both gas chromatography (GC)–MS and liquid chromatography (LC)–MS. To facilitate the identification of on-target phenotypes and avoid nonspecific “death” phenotypes, dynamic time-course analyses were undertaken, and the predicted modes of action were confirmed using nonstationary [¹³C]-glucose stable isotope labeling. This approach provides crucial information on the speed of action of antimalarials and the hierarchy of metabolic dysregulation induced by compounds with pleiotropic modes of action, potentially allowing us to dissect early specific effects of inhibition as well as downstream metabolic consequences.

Before any mode of action study of novel antimalarials can be performed, it is necessary to characterize the metabolic signatures of existing antimalarials, first to confirm the validity of the approach and second to identify metabolic signatures that might aid target identification of novel compounds. Here we define the metabolic response to a range of existing antimalarials, including chloroquine, atovaquone, proguanil, cycloguanil, methylene blue, and dihydroartemisinin (DHA), as well as new classes of inhibitors with antimalarial activity (fosmidomycin, the glucose transport inhibitor 3361, and the mammalian target of rapamycin [mTOR] inhibitor, Torin 2). Treatments with each of these drugs was shown to induce distinct metabolic perturbation in the short and long terms, highlighting potentially novel modes of action and/or downstream consequences of disrupting known pathways.

METHODS

P. falciparum In Vitro Cultivation

P. falciparum 3D7 parasite line was maintained in continuous culture, according to established methods, in Roswell Park Memorial Institute 1640 medium (RPMI 1640) containing 0.5% Albumax II, supplemented with 25 mmol/L HEPES, 100 µmol/L hypoxanthine, and 10 µg/mL gentamycin with O-positive blood (kindly provided by the Australian Red Cross Blood Service) under controlled atmospheric conditions (5% carbon dioxide, 1% oxygen_, and 94% nitrogen). Cultures were synchronized with sorbitol. At the trophozoite stage, P. falciparum–infected cultures were enriched to >95% purity via magnetic enrichment using a custom magnetic enrichment apparatus ([15] and 3-dimensional printing by International Seal Company Australia [ISCA]) and cell separation (CS) columns (Miltenyi Biotech Australia). Purity and cell number were determined via Giemsa-stained smears and hemocytometer counting, and the >95% infected erythrocyte cell suspensions were allowed to recover for 0.5–1 hour in fresh medium at 37°C before experimentation.

Recovery experiments were performed after 3361 treatment of infected cultures (6 hours at 240 µmol/L). Approximately 1% parasitemia cultures were incubated with or without 3361 under standard culturing conditions. After drug treatment, both cultures were washed with RPMI 1640 and then resuspended in fresh RPMI and allowed to recover for 48 hours. Smears were then obtained, and the parasitemia for each condition was determined by cell counting. Three independent experiments were performed on separate days to determine the percentage recovered compared with untreated control cultures.

Antimalarial Time Series and Metabolite Extraction

Parallel time series were initiated with the addition of 3–50 times the median inhibitory concentration (IC₅₀) for a panel of antimalarials (and untreated for the control condition) to >95% infected erythrocyte cultures (1 × 10⁸ cells per sample at 0.4% hematocrit) in complete RPMI 1640 (with 0.5% Albumax II). The concentrations for each antimalarial were as follows: 20 nmol/L atovaquone (IC_50, 1.3 nmol/L in 3D7 [16]), 12 µmol/L proguanil (the highest concentration that does not disrupt the mitochondrial membrane potential; IC_50, 55 µmol/L in 3D7 [16]), 20 nmol/L atovaquone/12 µmol/L proguanil, 100 nmol/L chloroquine (IC_50, 9 nmol/L in 3D7 [17]), 40 nmol/L DHA (IC_50, 3.5 nmol/L in 3D7 [18]), 10 µmol/L fosmidomycin (IC_50, 1 µmol/L in 3D7 [19]), 240 µmol/L glucose transport inhibitor 3361 (equivalent to 3 × IC₅₀ in 3D7 and chosen to minimize the amount of vehicle added [20]), 40 nmol/L methylene blue (IC_50, 3.6 nmol/L in D6 [21]), 200 nmol/L cycloguanil (IC_50, 4 nmol/L in 3D7 [22]), and 40 nmol/L Torin 2 (IC_50, 2.8 nmol/L in 3D7 [8]). Cultures were returned to controlled atmospheric conditions (5% carbon dioxide, 1% oxygen, and 94% nitrogen) and incubated at 37 °C. At predetermined intervals (0.5, 1, 2, 4, and 6 hours), parasite cultures were processed for metabolite analysis [23]. (See Supplementary Materials for further details.)

For long-term untargeted analysis, tightly synchronized parasites were exposed to either methylene blue or chloroquine for 6, 12, and 24 hours. Infected erythrocytes were magnetically purified with complete RPMI 1640 containing the appropriate antimalarial, and infected erythrocytes were allowed to recover for 1 hour (in complete RPMI 1640 containing methylene blue or chloroquine). Aliquots (1 × 10⁸ cells) were metabolically quenched, and metabolite extraction was performed [23]. Smears were obtained to confirm that parasites under each drug treatment were developmentally similar to untreated control cultures, and findings were further validated by the fact that comparable amounts of infected erythrocytes were magnetically purified from treated and untreated cultures.

Nonstationary [¹³C]-U-Glucose Labeling

Purified infected erythrocytes were incubated with different antimalarials (or as untreated controls) as described above for 2 hours before stable isotope labeling. Parasite cultures were spiked with 1 mol/L [¹³C]-U-glucose to give a 1:1 final ratio of ¹²C/¹³C-U-glucose (glucose-free RPMI powder and [¹³C]-U-glucose, both from Sigma). A “spike in” approach was taken to avoid perturbation of parasite cultures during drug treatment and labeling. Parasite cultures were then returned to standard culturing conditions for 30 minutes before metabolite extraction. Because chloroquine treatment did not result in a clear metabolic phenotype throughout the initial time course, for stable isotope labeling we extended chloroquine pretreatment to 6 hours before stable isotope labeling to maximize the possibility of detecting a perturbation of metabolic turnover. The theoretical maximum of ¹³C labeling for any metabolite species was 50% (due to the 1:1 addition of [¹²C/¹³C]-glucose). Medium was collected and analyzed by GC-MS to confirm the ratio of [¹²C]-glucose to [¹³C]-glucose (¹³C/¹²C glucose) ratio. (The supplementary materials contain the noncorrected values; for final data presentation, the labeled species were corrected for complete labeling.)

GC-MS and LC-MS Analysis

GC-MS analysis was performed using methods described elsewhere [23]. For LC-MS analysis, metabolite samples were separated on a SeQuant ZIC-pHILIC column (5 µmol/L; 150 × 4.6 mm²; Millipore) using a binary gradient (acetonitrile-water with 20 mmol/L ammonium carbonate) with a 1200 series high-performance liquid chromatography system (Agilent) coupled to a 6520 Q-TOF mass spectrometer (Agilent). (See Supplementary Materials for further details.)

RESULTS AND DISCUSSION

Untargeted Metabolite Profiling Using a Dual GC-MS and LC-MS Approach

The analytical metabolomics pipeline for investigating drug-induced perturbations of the P. falciparum metabolome is outlined in Supplementary Figure 1A. P. falciparum–infected cultures were magnetically purified to >95% parasitemia before being exposed to 3–50 × IC₅₀ drug concentrations. Infected erythrocytes were sampled at multiple time points to identify early drug-induced perturbations in metabolism and distinguish drug-specific effects from a general death phenotype. After rapid quenching, metabolites were extracted and analyzed by GC-MS and LC-MS, providing complementary coverage of the metabolic network of the P. falciparum–infected erythrocytes (Supplementary Figure 1B).

This approach was initially used to assess the metabolic impact of 4 clinically used antimalarials. These were atovaquone, which targets the bc₁ complex of the mitochondrial electron transport chain (mETC) [24, 25]; proguanil, which is synergistic with atovaquone but has an unknown mode of action [26, 27]; chloroquine, an inhibitor of digestive vacuole function and heme detoxification [28]; and DHA, for which a number of targets have been proposed [29, 30]. We also tested several metabolic inhibitors with known modes of action to investigate the global impact of these inhibitors on parasite metabolism. These included fosmidomycin, an inhibitor of the apicoplast isoprenoid biosynthetic pathway (the DOXP pathway) [31]; cycloguanil, an inhibitor of the fused enzyme thymidylate synthase/dihydrofolate reductase in folate biosynthesis [32]; 3361, a selective inhibitor of the P. falciparum plasma membrane hexose transporter [20]; and methylene blue, which is thought to inhibit glutathione reductase activity and heme detoxification [33]. Finally, we investigated the metabolic phenotype induced by the potent antimalarial compound Torin 2 (developed as an inhibitor of mammalian mTOR), which is active against liver, asexual, and sexual blood-stage parasites but has an unknown mode of action in P. falciparum [8, 34].

All drug treatments were performed at concentrations that corresponded to 3–50 × IC₅₀. To confirm that the parasites remained viable over the 6-hour time course, the steady-state level of a range of glycolytic intermediates was measured over this treatment period. Glycolysis is the primary source of adenosine triphosphate (ATP) generation in P. falciparum–infected erythrocytes, and the levels of glycolytic intermediates provide a sensitive measure of the permeability and overall metabolic activity of intracellular parasite stages. None of the clinically used antimalarial drugs resulted in loss of glycolytic intermediates at the concentrations and time periods used (Figure 1).

Figure 1.

Metabolic effect of antimalarials on Plasmodium falciparum–infected erythrocytes. Total metabolite pools (measured as ion counts) across a 6-hour exposure were determined relative to untreated controls. Changes in total metabolite pools are expressed as log₂ ratios of antimalarial-exposed cultures (+drug) to untreated controls (−drug). Data are presented as the means of 3–5 independent experiments. The antimalarials tested include the glucose transport inhibitor 3361 (3361), atovaquone (Atv), atovaquone +proguanil (A + P), proguanil (Pro), cycloguanil (Cyclo), dihydroartemisinin (DHA), Torin 2 (Tor2), chloroquine (CQ), methylene blue (MB), and fosmidomycin (Fos). GC (gas chromatography) and LC (liquid chromatography) indicate the data source for each metabolite row, and plus signs indicate that the metabolite assignment was confirmed with an authentic standard. Abbreviations: Arginine met, arginine metabolism; Hb Cata, hemoglobin catabolism; P, phosphate in glycolytic and PPP metabolites; peptide composition designated using IUPAC one letter amino acid notation; PPP, pentose phosphate pathway; TCA, tricarboxylic acid cycle.

To confirm that specific disruption of parasite glycolysis does indeed result in loss of glycolytic intermediates, P. falciparum–infected erythrocytes were treated with the compound 3361 (3 × IC₅₀; 240 µmol/L), previously shown to be a specific inhibitor of the parasite, but not the erythrocyte, glucose transporters [20]. Treatment with 3361 resulted in a rapid decrease in glucose-6-phosphate (eg, 8-fold reduction within 30 minutes) and other glycolytic intermediates (Figure 1 and Supplementary Table 1). Significantly, glucose levels in the infected erythrocyte increased in the presence of 3361, consistent with continued glucose transport across the erythrocyte plasma membrane and buildup in the host cell after inhibition of glucose uptake into the parasite (Supplementary Figure 2A).

Nonstationary [¹³C]-U-glucose labeling was also performed (30 minute incubation) in the presence of drugs to determine the turnover of biosynthetic intermediates. Similar levels of [¹³C]-labeling in glycolytic intermediates were observed for all control (no drug treatment) and drug treatments, with the exception of 3361 (Supplementary Figure 2B and Supplementary Table 1). These experiments demonstrate that parasites remain viable and metabolically active during the drug treatment and that any perturbations detected are not due to cell death.

Treatment with 3361 caused disruption to several metabolic pathways beyond glycolysis (Figure 1). Disruption of tricarboxylic acid (TCA) cycle fluxes and purine was not surprising and probably reflects the loss of the energy balance in the cell and/or the loss of new carbon skeletons for biomass production. However, 3361 treatment also resulted in a decrease in the intracellular levels of several hemoglobin-derived peptides, indicating disruption to hemoglobin digestion. Inhibition of glycolysis would be expected to lead to a global decrease in ATP levels, which in turn could lead to disruption of ATP-dependent acidification of the food vacuole and/or transport of proteases to this compartment, both of which could lead to decreased proteolysis (Supplementary Figure 2C).

To confirm that 3361 treatment does eventually lead to cell death, P. falciparum–infected cultures were incubated with 3361 (240 µmol/L) for 6 hours and then resuspended in fresh medium in the absence of drug for 48 hours to assess the level of recovery. The parasitemia of 3361-treated cultures were reduced (by a mean [standard error] 23.5% [3.6%] of untreated controls), confirming that 3361 treatment, at the concentration used, leads to nonrecoverable parasites.

Metabolic Pathways Affected by Atovaquone, Proguanil, and Cycloguanil Treatment

Atovaquone is a selective inhibitor of the mETC bc₁ complex [24, 25]. The mETC is not required for oxidative phosphorylation in asexual stages but is essential for the activity of dihydroorotate dehydrogenase (DHODH), which catalyzes a key step in de novo pyrimidine biosynthesis [16]. Atovaquone is used in combination with proguanil to prevent the development of resistance, which arises rapidly if atovaquone is used alone [35]. Proguanil is a prodrug that must be metabolized by liver cytochrome p450 into the active cycloguanil [36]. However, proguanil itself exhibits strong synergistic action with atovaquone [27, 37], but is not a potent antimalarial by itself (IC₅₀, approximately 55 µmol/L) [16]. Although the mode of action of proguanil is unknown, it is proposed to inhibit a still-unidentified mitochondrial membrane protein that maintains the mitochondrial membrane potential when the mETC is blocked [16, 26, 27]. To gain further insights into the potential mode of action of proguanil, we treated infected erythrocytes with 12 µmol/L proguanil (the highest concentration at which proguanil does not disrupt the inner-mitochondrial membrane potential [27]), with or without atovaquone.

As expected, atovaquone treatment alone resulted in the rapid accumulation of pyrimidine biosynthetic intermediates upstream of DHODH (Figure 2A and Supplementary Figure 3) and a converse drop in the intracellular levels of uridine monophosphate (UMP), consistent with inhibition of the mETC bc₁ complex and concomitant loss of DHODH activity [16]. Curiously, the parasite's orotate pool is maintained during DHODH inhibition. It is unclear which mechanism contributes to this phenotype, but it may involve allosteric regulation of the enzymes immediately downstream of DHODH. Significantly, atovaquone treatment also led to a wider disruption of mitochondrial metabolism.

Figure 2.

Metabolic signature induced by atovaquone treatment. A, Changes in the pool size of dihydroorotate after exposure to atovaquone (Atv), proguanil (Pro), atovaquone and proguanil (A + P), and chloroquine (CQ). B, Mitochondrial electron transport chain (mETC) and tricarboxylic acid (TCA) cycle of Plasmodium falciparum. Proteins associated with the mETC include glycerol-3-phosphate dehydrogenase (G-3-PDH), type II NADH (nicotinamide adenine dinucleotide, reduced) dehydrogenase (NDH), malate-quinone oxidoreductase (MQO), and dihydroorotate dehydrogenase (DHODH), and the mETC components include succinate dehydrogenase (SDH or complex II), cytochrome bc1 (complex III), cytochrome c (CytC) and cytochrome c oxidase (complex IV). Adenosine triphosphate synthase (complex V) is excluded. C, Change to the intracellular fumarate pool after drug treatment. D, Carbon 13 (¹³C) glucose labeling into citrate after drug treatment. Infected erythrocytes were treated with atovaquone alone, proguanil alone, or atovaquone and proguanil combined, for 2 hours before [¹³C]-glucose labeling (30 minutes with drug present). Chloroquine was pretreated for 6 hours before [¹³C]-glucose labeling. Top panel in D represents the total metabolite pool (combining all isotopes); bottom panel, percentage of the total citrate pool present as the M + 2 species, indicating [¹³C]-acetyl-CoA incorporation. Data are presented as mean and standard errors (SEs) from 3 biological replicates. Data in A and C are presented as log₂ ratios of drug-treated cultures (+drug) to untreated controls (−drug) and as means and SEs from 3–5 biological replicates. Statistical significance was determined using “within-model” testing and 1-way analysis of variance with Tukey post hoc testing (complete results presented in Supplementary Table 1). Abbreviations: 2-Oxo, 2-oxoglutarate; Ac-CoA, acetyl-CoA; Cit, citrate; Fum, fumarate; Iso, isocitrate; Oxo, oxaloacetic acid; Suc, succinate; Suc-CoA, succinate-Co enzyme A.

The TCA cycle is intimately linked to the mETC (Figure 2B), and atovaquone treatment led to an accumulation of the TCA cycle intermediate fumarate (Figure 2C). The accumulation of fumarate could indicate inhibition of the malate-quinone oxidoreductase complex that participates in both the TCA cycle and mETC (via blocking the bc₁ complex and inhibiting ubiquinone turnover) [38] or reflect a generalized perturbation to TCA cycle flux. To investigate TCA cycle activity, infected erythrocytes were metabolically labeled with [¹³C]-U-glucose, and labeling into the TCA cycle intermediate citrate measured in the presence or absence of atovaquone (Figure 2D). Flux through the TCA cycle is significantly reduced under atovaquone treatment as the intracellular levels of citrate and the percentage labeling derived from [¹³C]-glucose were reduced 2- and 3-fold, respectively, compared with untreated controls (Figure 2D). These findings are in good agreement with those of Ke et al [39], who recently demonstrated that aconitase activity is essential during gametocytogenesis and that atovaquone blocks [¹³C]-U-glucose incorporation into TCA cycle intermediates [39]. The secondary action of atovaquone on inhibiting TCA cycle activity could thus help prevent the development of transmission-competent gametocyte stages [40].

Despite being a prodrug and relatively nontoxic to P. falciparum asexual stages, proguanil treatment caused an accumulation of intracellular arginine levels (Figure 3A). Proguanil contains a guanidinium-like structure that resembles the side chain of arginine; raising the possibility that proguanil is a competitive inhibitor of arginase (or another arginine consuming enzyme or transporter). The parasite arginase has been proposed to be required for virulence in vivo (although not in vitro culture) [41], and inhibition of this enzyme may underlie the weak antimalarial effect of proguanil. Interestingly, there was no metabolic synergism when proguanil was added in combination with atovaquone (Supplementary Figure 3), consistent with proguanil having a primary inhibitory effect on mitochondrial membrane potential function when mETC activity is blocked by atovaquone.

Figure 3.

The metabolic phenotype of proguanil and cycloguanil exposure. A, Change to the intracellular arginine pool after proguanil (Pro) treatment. B, Disruption to thymidylate synthesis during cycloguanil exposure. DHA, dihydroartemisinin. Data are presented as log₂ ratios of drug-treated cultures (+drug) to untreated controls (−drug) and as means and standard errors from 3–5 biological replicates. Statistical significance was determined using “within-model” testing and 1-way analysis of variance with Tukey post hoc testing (complete results presented in Supplementary Table 1). Abbreviations: A + P, atovaquone and proguanil; Atv, atovaquone; CQ, chloroquine; Cyclo, cycloguanil; dTMP, deoxythymidine monophosphate; dTTP, deoxythymidine triphosphate; dUDP, deoxyuridine diphosphate; dUMP, deoxyuridine monophosphate; UDP, uridine diphosphate.

Proguanil is also metabolized by cytochrome p450 into cycloguanil, a potent inhibitor of the dual enzyme dihydrofolate reductase/thymidylate synthase, which mediates the conversion of deoxyuridine monophosphate (dUMP) into deoxythymidine monophosphate (dTMP) [32]. Cycloguanil treatment causes a rapid accumulation of dUMP and a progressive decrease in the downstream intermediates, dTMP and deoxythymidine triphosphate (dTTP) (Figure 3B). This phenotype is consistent with the selective inhibition of thymidylate synthase by cycloguanil (but not proguanil), preventing the synthesis of dTMP and the coincident buildup of the enzyme's substrate.

Pleiotropic Effect of DHA on Parasite Metabolism

The artemisinin class of antimalarials has been intensively studied because of their importance to global malarial control. Much research has focused on determining the mechanism of activation, mode of action, and mechanism of resistance to this class of drugs (reviewed in [29]). It has been established that artemisinin or DHA (the active form of the clinically used artemisinins) requires reductive cleavage of the characteristic endoperoxide bridge for antimalarial activity (as reviewed in [30]). The free radicals formed via the opening of the bridge are then thought to readily react with a variety of protein targets. The majority of artemisinin is activated within the digestive vacuole of the parasite, primarily by heme and reduced iron (Fe²⁺) [42, 43], and the activated artemisinin likely acts on targets proximal to the sources of activation. Given the proposed nonspecific targets of activated artemisinin and several reports of alternative modes of action [18, 44, 45], we sought to investigate what metabolic phenotype is elicited after DHA treatment.

Hemoglobin catabolism is rapidly perturbed after DHA treatment, as indicated by a decrease in the abundance of hemoglobin-derived peptides within 1 hour of exposure (Figure 4). However, the temporal profile of peptide depletion induced by DHA treatment was clearly distinct from the loss of energy metabolic phenotype induced by 3361 (Figure 1). These findings are consistent with DHA uptake and activation within the digestive vacuole, leading to alkylation and inactivation of digestive vacuole proteins and resulting in inhibition of hemoglobin catabolism.

Figure 4.

Perturbations to hemoglobin-derived peptides after dihydroartemisinin (DHA) and Torin 2 (Tors) treatment. The disruption to hemoglobin catabolism is represented by hemoglobin-derived peptides prolyl-glutamate (PE), prolyl-aspartate (PD), prolyl-glutamyl-glutamate (PEE) and aspartyl-leucyl-histidine (DLH). Peptide pool sizes are presented as the log₂ ratios of drug-treated cultures (+drug) to untreated controls (−drug) across 0.5–6 hours of exposure, means and standard errors (3–5 biological replicates). One-way analysis of variance with Tukey post hoc testing was performed to determine the significance of differences between treated time courses; complete results are presented in Supplementary Table 1. Abbreviations: A + P, atovaquone and proguanil; Atv, atovaquone; Fos, fosmidomycin; CQ, chloroquine; Cyclo, cycloguanil; MB, methylene blue; Pro, proguanil.

DHA treatment also leads to disruption of pyrimidine biosynthesis (Figure 5) but with a different metabolic signature from that induced by atovaquone (which results in a buildup of carbamoyl aspartate; Supplementary Figure 3). Thus, in addition to its effect on digestive vacuole function, DHA may also directly or indirectly inhibit enzyme(s) in pyrimidine biosynthesis.

Figure 5.

The perturbation of pyrimidine biosynthesis during dihydroartemisinin (DHA) exposure. Metabolite pool size changes are presented as the log₂ ratios of drug-treated (+drug) to untreated controls (−drug) across 0.5–6 hours of exposure and as means and standard errors (3–5 biological replicates). Two-tailed Student t tests were used to determine the significance of differences between DHA- and chloroquine (CQ)–treated time courses. Abbreviation: UMP, uridine monophosphate.

The [¹³C]-glucose labeling experiments confirmed that DHA had no effect on labeling of glycolytic intermediates in P. falciparum–infected erythrocytes (Supplementary Figure 2B and Supplementary Table 1). However, the labeling of UMP and uridine triphosphate (UTP) from ribose incorporation (isotopomer corresponding to monoisotopic mass plus five atomic mass units; M + 5) was reduced (Figure 6). In contrast, labeling into aspartate and malate was increased in DHA-treated parasites, consistent with inhibition of pyrimidine biosynthesis and redirection of carbon into malate production under DHA treatment.

Figure 6.

Dihydroartemisinin (DHA)–induced changes to metabolic turnover in Plasmodium falciparum–infected erythrocytes. [¹³C]-glucose labeling into the pyrimidine biosynthetic pathway in the absence of drug or the presence of atovaquone (Atv), the glucose transport inhibitor 3361 (3361), or DHA. For each metabolite presented, the top panel represents means (and standard errors [SEs]) for the total pool as ion counts; bottom panel, percentage labeling for each metabolite, as mean and SE. Red bar graphs represent M + 3 isotopomer species (monoisotopic mass plus three atomic mass units) for each metabolite (derived from triose phosphate); blue bar graphs, M + 5 species for UMP and UTP (derived from the ribose group). Abbreviations: Oxo, oxaloacetic acid; PEP, phosphoenolpyruvate; UMP, uridine monophosphate; UTP, uridine triphosphate.

DHA may directly inhibit an enzyme or enzymes within the early steps of pyrimidine biosynthesis, such as aspartate carbamoyltransferase (which mediates the synthesis of carbamoyl-aspartate) or carbamoyl-phosphate synthase, which would inhibit carbamoyl-aspartate synthesis and lead to disruption of downstream pyrimidine biosynthetic intermediates. Alternatively, DHA may nonspecifically alkylate mitochondrial proteins, consistent with effects on both the digestive vacuole and mitochondrion [46].

The pleiotropic actions of DHA may contribute to the potency of the endoperoxide artemisinins and explain the differential susceptibility of parasite stages within the intraerythrocytic cycle to artemisinin [43]. In particular, artemisinin sensitivity increases during the cycle, commensurate with increasing hemoglobin catabolism. The exception is very early ring-stage parasites (0–4 hours after invasion), which are hypersensitive to artemisinins yet have negligible hemoglobin digestion [43]. Our finding that DHA may also perturb pyrimidine biosynthesis, which is required throughout the intraerythrocytic cycle [16], may explain the antiparasitic activity of DHA during the early stages of parasite development. The pleiotropic effect of DHA on parasite metabolism is also consistent with a mechanism of resistance involving a generalized stress response and an altered ubiquitin-proteasome system [47, 48].

Differential Disruption of Hemoglobin Catabolism After Long-Term Chloroquine and Methylene Blue Exposure

Chloroquine, methylene blue and fosmidomycin did not induce detectable metabolic signatures during initial metabolite profiling. Fomidomycin has a well-described target in isoprenoid biosynthesis, and although intermediates within this pathway are detectable via MS [14], they were below the limit of detection in this study. It is likely that a targeted LC-MS approach is necessary for a detailed analysis of secondary pathways with low-abundance intermediates (such as isoprenoid biosynthesis). We therefore undertook an expanded untargeted analysis of parasites that had been treated with chloroquine and methylene blue for 6–24 hours at 5–10× IC₅₀ to determine when (if any) metabolic disruption first occurs. Untargeted LC-MS analysis resulted in the detection of >1200 mass (m/z) features, and pairwise comparisons with untreated controls were performed to determine which features were significantly different (adjusted P < .01; Benjamini-Hochberg corrected) [49].

Methylene blue treatment resulted in no significant changes after a 6-hour exposure (consistent with Figure 1), but the number of perturbed metabolites increased significantly after 12 and 24 hours of exposure (Supplementary Figure 4). The first significant metabolic changes after methylene blue treatment occur to TCA cycle intermediates (including the related amino acids glutamine and aspartate), oxidized glutathione, and the hemoglobin-peptide, valine-asparate-aspartate (VDD) (after 12 hours of treatment; Supplementary Figure 4). After 24 hours of treatment, TCA cycle–related metabolites remain perturbed, but additional hemoglobin-derived peptides and lipid biosynthetic precursors are also disrupted (Figure 7A and 7B). In contrast, chloroquine elicited a subtle metabolic response (Figure 7C and 7D), resulting in elevated dipeptides and tripeptides after 12 and 24 hours of exposure (Supplementary Figure 4). The peptides altered during chloroquine exposure are also elevated in chloroquine-resistant parasite lines [50].

Figure 7.

Long-term methylene blue and chloroquine exposure. Parallel Plasmodium falciparum–infected cultures enriched at trophozoite stage after drug treatment (24 hours) and measured with liquid chromatography–mass spectrometry. A, Changes to the metabolome of P. falciparum after methylene blue (MB) exposure (40 nmol/L; 24 hours). Untargeted analysis was performed and each mass/charge (m/z) feature detected (>1200) is presented as the relative ion intensity of drug-treated to untreated controls. Unchanged features are plotted along the diagonal, with features increased/decreased under drug treatment plotted above/below the diagonal, respectively. Each data point represents a single m/z feature, with the size of the point corresponding to the fold change between the 2 conditions. Color intensity corresponds to the reproducibility across biological replicates [49]. B, Statistically significant changes in identified metabolites after 24 hours of methylene blue exposure (P < .01; Benjamini–Hochberg corrected). C, Changes to the metabolome of P. falciparum–infected erythrocytes after 100 nmol/L chloroquine (CQ) treatment (24 hours). Data are presented as described for A. D, Metabolites differed significantly between chloroquine-treated infected erythrocytes and untreated controls (P < .01; Benjamini–Hochberg corrected). For significantly different metabolites detected with both chloroquine and methylene blue treatments, authentic standards for 1,3 (2,3)-diphosphoglycerate, glutamine, aspartate, 2-oxoglutarate, and oxidized glutathione were performed to confirm the putative metabolite assignments. Abbreviation: GSSG, oxidized glutathione; peptides are denoted using the IUPAC one letter notation for amino acid sequences.

The small changes in peptide abundance during chloroquine treatment are in contrast to the rapid and substantial (4–8-fold) reduction in hemoglobin-derived peptides during DHA and 3361 exposure (Figure 4). Moreover, methylene blue exhibits a different effect on hemoglobin peptide abundance (in the peptides themselves and the direction of change). The data suggest that chloroquine, methylene blue, DHA, and 3361 inhibit hemoglobin digestion in distinct ways. This could reflect a difference in how each antimalarial disrupts parasite metabolism, potentially inhibiting heme detoxification and hemoglobin proteases directly or through indirect mechanisms, such as by altering the homeostasis of the digestive vacuole or inhibiting transporters on the digestive vacuolar membrane.

Exposure to methylene blue (unlike chloroquine) also resulted in significant changes to TCA cycle intermediates, lipid biosynthetic precursors, and oxidized glutathione. The elevation of oxidized glutathione is consistent with glutathione reductase being a target of methylene blue [33], although we recognize that glutathione oxidation may occur during sample preparation and analysis. Nevertheless, perturbation of glutathione metabolism indicates induction in oxidative stress, and it is conceivable that the remaining metabolic perturbations are secondary to inhibition of pathways particularly sensitive to such stress.

Investigating the Mode of Action of the Novel Antimalarial Torin 2

We wanted to investigate whether metabolite profiling could be used to understand the mode of action of new antimalarial compounds with unknown modes of action. The mTOR inhibitor, Torin 2, has recently been shown to kill liver-stage, asexual, and sexual intraerythrocytic-stage parasites [8, 34]. The low IC₅₀ for Torin 2 (1–3 nmol/L) was surprising given that there is no evidence that Plasmodium parasites have an mTOR signaling pathway [8, 34]. Strikingly, MS profiling of Torin 2–treated parasites revealed a rapid and selective reduction in hemoglobin-derived peptides (Figure 1). Decreases in multiple dipeptides and tripeptides were evident within 1 hour of treatment (Figure 4).

Torin 2 has been elsewhere shown to interact with a number of Plasmodium proteins, including phosphoribosyl pyrophosphatase synthetase (PF3D7_1325100), aspartate carbamoyltransferase (PF3D7_1344800), and a putative nutrient transporter (PF3D7_0914700) [8]. PF3D7_0914700 is a polytopic membrane protein that is expressed in the asexual stage and has a dual localization to the digestive vacuole and parasite plasma membrane (unpublished work). Although the substrate for this putative transporter is unknown, our findings raise the possibility that PF3D7_0914700 contributes to hemoglobin catabolism (either directly by exporting hemoblogin-derived peptides or indirectly by maintaining the homeostasis of the digestive vacuole).

No changes were observed in the intracellular levels of intermediates in the pentose phosphate and pyrimidine biosynthetic pathways, which contain the other 2 Torin 2–binding proteins [8]. However, we cannot discount the possibility that Torin 2 inhibits these pathways at higher concentrations or that inhibition occurs more slowly. Moreover, this finding does not contradict reports that Torin 2 affects protein export in the liver stage and may indicate a common target [34].

CONCLUSION

Although considerable progress has been made in combating the global impact of malaria, there remains a pressing need to develop new antimalarials in order to avoid overreliance on the limited number of existing front-line drugs and counter the threat of emerging drug resistance. With the success of large-scale phenotypic screens for new antimalarials [4], new methods are needed to broadly identify the mode of action of inhibitor compounds and prioritize those that are most suited for further optimization. We show that the metabolomics pipeline used here is suitable for investigating the modes of action of antimalarials with pleiotropic activities that are refractory to in vitro resistance screens but highly desirable for clinical development. Moreover, this approach can be combined with genomics and proteomics approaches to validate putative drug target identification and accelerate hit prioritization.

Notes

Acknowledgments. We thank Dr Jose Garcia-Bustos (Monash University) for stimulating discussions and feedback.

Financial support. This work was supported by the Australian National Health and Medical Research Council (project grant APP64095), R. D. Wright Biomedical Fellowship 1062504 to S. A. R., C. J. Martin training fellowship 628899 to D. J. C., and a NHMRC Principal Research Fellowship to M. J. M.).

Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.