Artemisinin-Based Drugs Target the Plasmodium falciparum Heme Detoxification Pathway

Artemisinin (ART)-based antimalarial drugs are believed to exert lethal effects on malarial parasites by alkylating a variety of intracellular molecular targets. Recent work with live parasites has shown that one of the alkylated targets is free heme within the parasite digestive vacuole, which is liberated upon hemoglobin catabolism by the intraerythrocytic parasite, and that reduced levels of heme alkylation occur in artemisinin-resistant parasites.

ABSTRACT

Artemisinin (ART)-based antimalarial drugs are believed to exert lethal effects on malarial parasites by alkylating a variety of intracellular molecular targets. Recent work with live parasites has shown that one of the alkylated targets is free heme within the parasite digestive vacuole, which is liberated upon hemoglobin catabolism by the intraerythrocytic parasite, and that reduced levels of heme alkylation occur in artemisinin-resistant parasites. One implication of heme alkylation is that these drugs may inhibit parasite detoxification of free heme via inhibition of heme-to-hemozoin crystallization; however, previous reports that have investigated this hypothesis present conflicting data. By controlling reducing conditions and, hence, the availability of ferrous versus ferric forms of free heme, we modify a previously reported hemozoin inhibition assay to quantify the ability of ART-based drugs to target the heme detoxification pathway under reduced versus oxidizing conditions. Contrary to some previous reports, we find that artemisinins are potent inhibitors of hemozoin crystallization, with effective half-maximal concentrations approximately an order of magnitude lower than those for most quinoline-based antimalarial drugs. We also examine hemozoin and in vitro parasite growth inhibition for drug pairs found in the most commonly used ART-based combination therapies (ACTs). All ACTs examined inhibit hemozoin crystallization in an additive fashion, and all but one inhibit parasite growth in an additive fashion.

TEXT

The two parasites responsible for most human malaria infections are Plasmodium vivax and Plasmodium falciparum, the latter being responsible for the majority of the ∼405,000 deaths due to malaria in 2018 (1). Artemisinin (ART)-based combination therapies (ACTs) are the current front-line treatment for P. falciparum malaria; however, the spread of evolving ART resistance (ARTR) that compromises ACT use in some settings is a deeply worrying phenomenon. ARTR was first reported over a decade ago in Southeast Asia (2, 3) and is clinically characterized as an infection with a clearance half-time of >5 h after ACT treatment (4). This “delayed clearance phenotype” (DCP) has been associated with mutations in the pfk13 gene (1, 5) that result in amino acid substitutions in the PfK13 protein propeller domain; however, although progress has been made, a molecular mechanism for PfK13-mediated ARTR remains elusive. Recent insights include the identification of altered kinetics of intraerythrocytic parasite development for parasites harboring PfK13 mutations (6, 7) and the identification of multiple mutant PfK13 interaction partners that appear to mediate altered endocytic traffic for ARTR parasites (8), altered hemoglobin (Hb) traffic in ARTR parasites (8–10), and reduced formation of ferriprotoporphyrin IX (FPIX) heme:dihydroartemisinin (DHA) adducts within live ARTR parasites (11, 12). Since free FPIX is released upon Hb catabolism by the intraerythrocytic parasite, and since FPIX heme iron is likely necessary for activating the toxic alkylating ability of artemisinin-based drugs, these observations related to PfK13-mediated ARTR appear to be coalescing toward a unified model of reduced endocytic Hb traffic and the resultant Hb catabolism during early parasite development within the human red cell, resulting in the slower release of free FPIX and, hence, the lower activation of artemisinin-based drugs by FPIX iron at key times during intraerythrocytic parasite development.

Several studies have found that a plethora of parasite proteins can be alkylated by activated ART-based drugs (13–16); however, there is little overlap between the various proteins identified in these studies (12), suggesting that protein alkylation by ART-based drugs is stochastic (12, 14). For alkylation to occur within live parasites, the essential endoperoxide bridge of ART-based drugs (17, 18) must first be activated via reduction by reduced iron. Since the most available and accessible form of reduced iron within live malarial parasites is reduced FPIX heme released upon Hb catabolism within the digestive vacuole (DV) of the parasite (7, 11), several studies have suggested that DV FPIX is necessary for ART-based drug pharmacology (11, 15). Previous in vitro studies by Meshnick et al., Posner et al., and Robert and colleagues clearly showed that once activated by ferrous FPIX, ART-based drugs may then form covalent drug-FPIX adducts, at least under laboratory solution conditions (19–23). Recently, the suspicion that such adducts might form within live parasites has been confirmed, and reduced levels of these adducts have been quantified for live ARTR parasites (7, 11). Heme has also been found to be an alkylation target of synthetic ozonide (OZ) antimalarials (24), which rely on a similar mode of endoperoxide pharmacophore activation (25). Due to the proximity of the FPIX porphyrin ring upon FPIX iron activation, FPIX is likely to be one of the more common alkylation targets of activated ART-based drugs. Since FPIX heme released upon Hb catabolism must be detoxified by crystallization to inert nontoxic hemozoin (Hz) for the parasite to survive, and since drug-FPIX adducts are predicted to inhibit that crystallization, the reduced alkylation of FPIX for ARTR parasites (11) suggests that inhibition of FPIX-Hz chemistry may be an important facet of ART-based drug molecular pharmacology.

Quinoline-based antimalarial drugs such as chloroquine (CQ) inhibit the formation of Hz (26, 27). This largely involves the formation of noncovalent drug-heme complexes that then inhibit the growth of Hz and presumably lead to a buildup of toxic free FPIX within the DV (28). It has been suggested that ART-based drugs might act in a similar fashion (29); however, this was later challenged (30). Many studies have continued to examine the ability of antimalarial drugs to inhibit Hz formation (30–47); however, none have quantified the potency of Hz crystallization inhibition for different ART-based drugs. In previous studies that examined ART-based drug inhibition of Hz formation, only putative noncovalent interactions were examined since inhibition was measured under oxidizing conditions that prevent endoperoxide activation.

In this study, we modify a previously reported β-hematin inhibition assay (BHIA) that quantifies drug inhibition of FPIX-Hz chemistry under close-to-physiological conditions (33) in order to quantify inhibition of Hz crystallization under reducing versus oxidizing conditions for the commonly used ART-based drugs artesunate (ATS), artemether (ATM), and dihydroartemisinin (DHA). Although organelle-specific concentrations are currently unknown, glutathione (GSH) is the principal reductant found within most eukaryotes, including Plasmodium (48–50). In the presence of GSH, Fe(III) protoporphyrin IX (PPIX) is reduced to Fe(II)PPIX, which is then is capable of activating these ART-based drugs (11, 22). Under biological conditions, once activated, these drugs rearrange to yield a C₄ radical that may form a covalent ART-FPIX adduct, during which ferrous FPIX is reoxidized back to ferric FPIX (11, 21). FPIX-heme adducts formed in this fashion could then act as Hz crystallization poisons since the bulky adduct would be unable to be incorporated within the growing Hz crystal. Here, we quantify the relative potencies of several different ART-based drugs to inhibit Hz formation in the presence versus the absence of common ACT partner drugs.

RESULTS

β-Hematin inhibition assay.

We examined several ART-based drugs (Fig. 1, top) for their ability to inhibit the formation of Hz in vitro under either oxidized (O-BHIA) (33) or reduced (R-BHIA) conditions. The reduction of Fe(III)PPIX in dimethyl sulfoxide (DMSO) or DMSO-water mixtures is nearly instantaneous (51); therefore, under the conditions of the modified R-BHIA (see Materials and Methods), most heme iron is in the Fe(II) state before the addition of the drug, whereas in the absence of the GSH reductant (O-BHIA), FPIX heme remains as Fe(III)PPIX (7, 51). By measuring the absorbance at 405 nm, corresponding to the FPIX Soret peak, we quantify the disappearance of free FPIX during the formation of Hz localized to a bicarbonate-insoluble pellet produced in the BHIA (see Materials and Methods) (33).

FIG 1.

Structures of the antimalarial drugs used in this study. These drugs can be categorized as “artemisinin derivatives,” “quinolines,” and “other.” Common artemisinin combination therapies (ACTs) include dihydroartemisinin (DHA) plus piperaquine (PPQ), artesunate (ATS) plus amodiaquine (AQ), and artemether (ATM) plus lumefantrine (LF).

In our O-BHIAs, we found no inhibition of Hz formation even at the highest ART-based drug concentrations tested (1 mM) (Fig. 2, left two panels, open symbols). In contrast, control experiments with CQ (Fig. 2, right, open symbols) showed potent inhibition, as reported previously. This inhibition is due to the ability of CQ to form noncovalent and dative complexes with FPIX under oxidized conditions (28).

FIG 2.

Representative β-hematin inhibition curves. Data ± standard errors of the means (SEM) are shown for one experiment performed in triplicate at pH 5.2 under reduced (filled circles) versus oxidized (empty circles) conditions (see the text). For artesunate (left) and artemether (center), inhibition of Hz formation is observed only under reducing conditions, whereas chloroquine (right) inhibits Hz formation under both oxidized and reduced conditions, with a slight reduction in the EC₅₀ for the latter (see the text).

In contrast, once FPIX is preincubated with GSH to form Fe(II)PPIX, clear inhibition of Hz formation is observed for ART-based drugs (Fig. 2, closed symbols), with 50% effective concentration (EC₅₀) values in the micromolar range (Table 1). We also note a significant, but much less profound, increase in CQ Hz inhibition potency under reducing conditions (Fig. 2, right, closed versus open symbols, and Table 1). By comparing Hz formation under reduced and oxidized BHIA conditions, these data further support previous conclusions that the reduction of Fe(III)PPIX is a crucial step in heme activation of ART-based drugs (11, 19–23) and support the notion that ART-based drugs are potent inhibitors of Hz formation.

TABLE 1.

EC₅₀ values for Hz inhibition under oxidized (O-BHIA) versus reduced (R-BHIA) conditions^a

Compound	Avg EC50 of compound ± SEM
	pH 5.2		pH 5.6
	O-BHIA (μM)	R-BHIA (μM)	O-BHIA (μM)	R-BHIA (μM)
DHA	>1,000	22.5 ± 0.1	>1,000	14.1 ± 2.3
ATM	>1,000	31.6 ± 3.2	>1,000	9.6 ± 1.7
ATS	>1,000	33.5 ± 2.6	>1,000	9.2 ± 2.2
CQ	194.9 ± 5.6	66.7 ± 4.2	52.9 ± 8.0	22.2 ± 2.8
AQ	70.1 ± 1.3	19.1 ± 1.5	22.1 ± 0.9	7.3 ± 3.5
MQ	—b	—	—	24.4 ± 4.5
PPQ	15 ± 0.3	7.7 ± 0.5	—	4.0 ± 1.2
LF	—	27.2 ± 2.0	—	9.9 ± 1.5

^aDifferent pHs correspond to the pHs of the DV in CQS/ARTS (pH 5.2) versus CQR/ARTR (pH 5.6) strains of P. falciparum. All values are the averages from three separate experiments, each performed in triplicate, ± the standard errors of the means.

^b—, poor solubility at pH 5.2.

Following the validation of the O-BHIA versus the R-BHIA, and noting that (to our knowledge) all ARTR P. falciparum parasites are derived from chloroquine-resistant (CQR) P. falciparum isolates reported to have lower DV pHs (52–54), we tested whether inhibition of Hz by ATM, ATS, or DHA (Fig. 1) is affected by the small shifts in pH previously noted for CQ-sensitive (CQS) versus CQR and ARTR parasites (52–54). As summarized in Table 1, the data suggest that ATS, ATM, and DHA remain inactive at both pHs under oxidizing conditions but are slightly more potent Hz formation inhibitors at the higher pH characteristic of CQS parasites under reducing conditions. Thus, we suggest that the lower pH characteristic of the CQR parasite DV (also measured for parasites that are both CQR and ARTR [K. Iyengar and P. D. Roepe, unpublished data]) may provide an additional modest layer of protection versus ART-based drugs. Similarly, as previously reported (33), we found that under oxidizing conditions, all quinoline-based drugs tested retained activity at both pHs but were similarly slightly less potent at the lower pH characteristic of the CQR DV (53, 54). Also, where such measurements were possible, quinoline-based drugs had lower EC₅₀ values under reduced than under oxidized conditions (Table 1). While we do not yet have a simple explanation for the increased potency of quinoline drugs under reduced conditions, based on the geometry of CQ-FPIX complexes, we suggest that it is possible that the heme binding affinity for quinoline drugs is higher for Fe(II)PPIX than for Fe(III)PPIX (55, 56) (see Discussion).

ACT fixed-ratio isobologram analysis.

In the clinic, ART-based drugs are used as ACTs in combination with “partner drugs” such as amodiaquine (AQ), piperaquine (PPQ), and lumefantrine (LF), with the most common ACTs in use today being ATM-LF, DHA-PPQ, and ATS-AQ (Fig. 1). To study the effect of common ACT partner drugs on Hz inhibition by ART-based drugs, according to the Chou-Talalay 1-point titration method for assessing drug synergy (57), we performed O-BHIAs and R-BHIAs at different ART drug/partner drug ratios. These ratios were normalized to the drugs’ respective Hz inhibition EC₅₀s measured for each drug alone (meaning that “1” corresponds to 1× EC₅₀ for the drug and that “2” corresponds to 2× EC₅₀ [Table 1], etc.) Representative isobologram plots are shown in Fig. 3, and as summarized in Table 2, all combinations tested were found to be additive (fractional inhibitory concentration [FIC] index of ∼1) with respect to their ability to inhibit Hz formation (58, 59).

FIG 3.

Isobologram analysis for Hz inhibition of common ACTs. The black dashed line indicates an FIC value of 1, or absolute additivity. The effect of the common ACTs is additive with respect to Hz inhibition, suggesting that the partner drug and the ART-based drug act on the same molecular target. Red circles, DHA-PPQ; green diamonds, ATM-LF; gray triangles, ATS-AQ; blue squares, ATS-MQ.

TABLE 2.

Summary of FIC indices from isobologram analyses of common ACTs^a

Drug combination	Avg FIC index ± SEM at ratio of:
	3:1	1:1	1:3
DHA-PPQb	1.20 ± 0.03	1.27 ± 0.03	1.11 ± 0.24
ATM-LFb	1.05 ± 0.07	1.09 ± 0.11	1.06 ± 0.15
ATS-AQb	1.00 ± 0.12	0.98 ± 0.10	1.01 ± 0.05
ATS-MQc	0.92 ± 0.09	0.94 ± 0.11	0.93 ± 0.08

^aSee Fig. 3. An FIC index of <1 is synergistic, ∼1 is additive, and >1 is antagonistic. Values are the averages from 2 or 3 independent experiments, each performed in triplicate, ± the standard errors of the means.

^bPerformed at pH 5.2.

^cPerformed at pH 5.6 due to solubility limits.

We also examined the effects of these drug combinations versus growth inhibition for three different P. falciparum strains, HB3 (CQS and ARTS), Cam WT (CQR and ARTS), and Cam WT-C580Y (CQR and ARTR) (Table 3). Again, most of the combinations were approximately additive, with the exception that ATM-LF showed high antagonism for the CQS strain (HB3) and highly synergistic effects for the CQR/ARTS (Cam WT) and CQR/ARTR (Cam WT-C580Y) strains at a 1:1 EC₅₀ ratio. These trends in antiparasitic versus Hz inhibition activities merit additional study at additional ACT drug ratios that mimic currently unknown blood concentrations for each drug when given via the relevant ACT.

TABLE 3.

Summary of FIC indices for parasite growth inhibition at 1:1 EC₅₀ ratios^a

Drug combination	Avg FIC index ± SEM
	HB3	Cam WT	Cam WT-C580Y
DHA-PPQ	1.36 ± 0.15	0.79 ± 0.03	1.07 ± 0.01
ATM-LF	2.75 ± 0.04	0.25 ± 0.02	0.33 ± 0.02
ATS-AQ	0.98 ± 0.01	1.04 ± 0.02	0.92 ± 0.06
ATS-MQ	0.86 ± 0.02	1.41 ± 0.03	1.29 ± 0.00

^aHB3 is CQS and ARTS, Cam WT is CQR and ARTS, and Cam WT-C580Y is CQR and ARTR. An FIC index of <1 is synergistic, 1 is additive, and >1 is antagonistic. Values are the averages from three independent experiments performed in triplicate ± the standard errors of the means.

DISCUSSION

Over the past several years, the list of possible ART-based drug molecular targets has continued to grow (13–16, 19–23). Several studies have identified a long list of proteins that are alkylated by activated ART-based drugs (13–16), but there is relatively little overlap among the proteins identified in these studies, suggesting that protein alkylation is stochastic (12, 14). Other nonprotein targets for alkylation are also possible, for example, the drugs that have been shown to alkylate heme in vitro (19–23) and in vivo (11). However, even though early in vitro data show that ART-based drugs can in theory alkylate heme, the data examining whether ART-based drugs inhibit Hz formation have conflicted. By modifying the BHIA (33) to inspect reducing as well as oxidizing conditions, we have quantified relative Hz inhibition potencies for different ART-based drugs. Many studies have shown that quinoline-based antimalarial drugs inhibit Hz formation in vitro and in vivo, and some of these have quantified the potencies for different quinoline-based drugs (e.g., see references 28, 29, 31, 33, 34, 36–40, and 99), but to our knowledge, none have examined the effects of ART-based and quinoline drug combinations. This is important since leading ACTs combine ART-based and quinoline drugs (Fig. 1).

Spatial diffusion of the ART radical.

As mentioned above, there is relatively little overlap between ART-based drug protein alkylation targets identified in several previous proteomic studies (13–16). We suggest that the relative alkylation of any target is proportional to its abundance and/or proximity to the activated ART radical. Following this model, alkylation of FPIX is also expected and is indeed observed (11) since it is immediately proximal following the activation of ART-based drugs within the DV.

Using what is currently known about P. falciparum protein subcellular localization, we note that for any study reported to date, the localization of ART-based drug protein targets may span nearly the entire parasite (12). However, the most efficient alkylation must satisfy two criteria: the lifetime of the ART radical must be long enough for the radical to reach the target, and the target must be located proximal to the DV. Several common protein targets identified in multiple studies satisfy the second criterion (12). We propose that the most prominent molecular targets are those that are both in the highest abundance and closest to (or within) the DV.

With regard to the first criterion, to our knowledge, no measurements or estimates of any ART-based drug radical’s lifetime have yet been made (note that the radical lifetime is not equivalent to the biological half-life). To estimate the distance that the radical is able to diffuse, consideration of Einstein-Brownian motion (60) and the lifetimes of other organic small-molecule radicals is helpful. Using an abbreviated Stokes-Einstein equation (equation 1) (60), where λ is the mean displacement along one axis, D is the diffusion coefficient, and t is time, we can estimate the diffusion distance of a radical with a specific lifetime:

$$\text{λ}\approx \sqrt{2\mathit{Dt}}$$ (1)

The diffusion coefficient, which is proportional to the size of the molecule in question, is approximately 10⁻⁵ cm² s⁻¹ for small molecules with molecular masses of approximately hundreds of daltons (61). When comparing radical lifetimes, we must also consider the environment in which such measurements are made. Upon the activation of an ART-based drug, the oxygen-centered (alkoxy) radical is initially formed but rearranges to form the carbon-centered C₄ radical (17). The C₄ radical is the more reactive alkylating species (62) and is shorter-lived than the alkoxy radical. If we therefore assume that the alkoxy radical is limiting for diffusion, the lifetime of activated ART intracellularly is on the order of 10⁻⁶ s (63–66). Thus, the distance traveled by the ART radical will be 10 nm to hundreds of nanometers. This can be compared to the diffusion distance of other short-lived species, such as peroxynitrite, which has a half-life of ∼10⁻³ s and a calculated diffusion distance ranging from 0.5 to 5 μm in erythrocytes or blood plasma, respectively (67). We suggest that the molecular targets of ART-based drugs, proteins or otherwise, are likely to be found either within or close to the DV. This is consistent with the in vivo alkylation of free FPIX within the DV (11), proteins implicated in Hb trafficking to the DV (8), and proteins involved in Hb catabolism (8).

FPIX abundance and PfK13 polymorphisms.

Polymorphisms most commonly associated with ART-based drug resistance phenomena confer amino acid substitutions within the propeller domain of the PfK13 protein. Two of the most commonly observed are C580Y and R539T PfK13 (9, 68–70), each of which appears to impair Hb uptake (8). This predicts reduced alkylation of FPIX and of proteins implicated in Hb trafficking or Hb catabolism. A consequence of the latter two predictions would be reduced availability of ferrous FPIX, yielding reduced activation of ART-based drugs, which would further contribute to reduced alkylation of FPIX (11). Whether reduced alkylation of some protein targets in ARTR parasites occurs is yet to be examined, but if this occurs, it would presumably be an indirect consequence of reduced ART-based drug activation due to reduced FPIX upon altered Hb uptake, traffic, and/or catabolism at key stages of parasite intraerythrocytic development.

Via multiple studies, PfK13 has been localized throughout the parasite (8, 10, 69, 71–74), including endocytic vesicles and/or cytostomes (8, 10, 72, 73). PfK13 expression is reduced in ARTR parasites relative to ARTS parasites, ARTR parasites can be resensitized to ART by increasing PfK13 expression (73), and ARTS parasites can be desensitized to ART by PfK13 mislocalization (10). However, deletion of PfK13 is lethal to the parasite (72, 75). An interesting observation then is that PfK13 has never been identified as an ART-based drug alkylation target (13–16). Taken together, these data suggest that while certain PfK13 mutations confer some ARTR as measured by the ring stage survival assay (RSA) (5), the mechanism is indirect, with one indirect consequence being impaired Hb catabolism (8–10). A consequence would then be altered Hz inhibition in ARTR parasites, as has also been measured (7). Interestingly, the introduction of a substitution in the Toxoplasma gondii K13 protein orthologue, analogous to the C580Y PfK13 substitution, also confers reduced ART susceptibility in T. gondii (76).

Effect of the iron oxidation state on quinoline-FPIX binding.

We found that Hz inhibition EC₅₀s for quinolines were lower under reduced conditions, suggesting that Fe(II) promotes stronger quinoline drug-FPIX binding. This is reasonable based on previous studies that examined the binding of nitrogen-containing compounds to iron, including CQ binding to monomeric FPIX (56). We note that there are two binding motifs expected for quinoline-FPIX interactions: dative covalent and noncovalent (28, 55, 56, 77).

The most commonly observed quinoline-FPIX interactions are pseudo-face-centered stacking π-π interactions between the porphyrin ring and the aromatic quinoline heterocycle. Over the past decade, several X-ray crystal structures have been solved, showing features of these interactions for quinine, quinidine (78), and halofantrine (79) that are consistent with structures derived previously via nuclear magnetic resonance (NMR) studies (55, 56, 80, 81). While we have no complete explanation for how the Fe oxidation state for FPIX might influence the strength of quinoline binding, we propose a simple model based on electrostatic differences (82) between Fe(II)PPIX and Fe(III)PPIX. The quinolines tested here have an electron-withdrawing chlorine on the aromatic ring system (Fig. 1). Fe(III) withdraws electrons from the porphyrin ring more strongly than Fe(II), resulting in a more-electron-dense ring for Fe(II)PPIX, which would then interact more strongly with an electron-deficient quinoline ring (82). To our knowledge, data testing this concept for antimalarial drugs are limited. Future computational studies may prove useful, along with experiments using quinoline derivatives containing electron-withdrawing versus -donating groups.

Another binding mode is dative covalent coordination between a nitrogen electron lone pair and the FPIX iron (56). Two studies show higher-affinity binding of N-containing compounds to ferric, but not ferrous, iron, which is the opposite of what we observe (83, 84). However, the published computational work (84) does not examine binding under aqueous conditions, as is also the case for work by Chiavarino et al. (85). In contrast, other studies that examine the binding of NO to various enzymes, done in aqueous environments, show significantly stronger NO binding to ferrous iron, in fact, stronger by up to an impressive 6 orders of magnitude (86, 87). A summary of these data, where binding was measured for biologically relevant iron in aqueous media (86, 88, 89), is shown in Table 4, as revised from data reported previously by Wanat et al. (87). Taken together, previous data show that the observed EC₅₀ shifts for quinoline-based drugs are easily attributed to a higher affinity for Fe(II)PPIX.

TABLE 4.

Summary of previously determined binding constants for NO with reduced versus oxidized iron within heme^a

Protein	Binding constant (M−1)		Reference(s)
	Fe2+	Fe3+
Mb	1.4 × 1011	1.4 × 104	86, 88
Hb	5.3 × 1011	4.0 × 103	88, 89
Cyt c	2.9 × 105	1.6 × 104	86
Fe	1.0 × 109	1.1 × 103	86

^aMb, myoglobin; Hb, hemoglobin; Cyt c, cytochrome c; Fe, FeTPPS [iron-5,10,15,20-tetrakis(4-sulfonatophenyl)porphyrinato].

ACT partner drug effects.

Perhaps not surprisingly then, we also find that common ACT drug pairs act largely in an additive fashion for both Hz inhibition (Table 2) and parasite growth inhibition (Table 3), with the exception of ATM-LF activity versus live parasites. We note that in the R-BHIA, there is one target to act upon (FPIX), which presumably enhances competitive ART-based drug versus partner drug binding (for which we then expect additive behavior), whereas LF may have multiple targets within the cell, promoting deviation from additivity for ATM-LF versus growth inhibition. Further studies are needed to elaborate upon these conclusions.

Summary.

Overall, these data and conclusions can be incorporated into a consolidated model for the mechanism of ART-based drugs (Fig. 4). Hb from the red cell cytosol (Fig. 4, “1”) is transported to the parasite DV (“4”), in part mediated indirectly by PfK13 (green star) (8, 10, 72, 73). Hb is catabolized by plasmepsins and falcipains within the DV (90–92), which releases toxic FPIX. Initially, FPIX is in the reduced form but rapidly oxidizes to Fe(III)FPIX. DHA or other clinically useful ART-based drugs that are quickly converted to DHA (93) are activated by Fe(II)PPIX to generate the DHA C₄ radical (Fig. 4, “DHA*”) (17), which can react with a variety of targets (11–16). One such target, the highly abundant and proximal-to-DHA*-protoporphyrin ring of FPIX, is efficiently alkylated, leading to both decreased Hz formation and an increased concentration of toxic DHA-FPIX adducts that may also diffuse to the cytosol (Fig. 4, “3”), as has also been postulated for quinoline drug-FPIX adducts (94). These lead to increased oxidative stress throughout the parasite. DHA radicals also alkylate proteins within the DV, the DV membrane, or the cytosol in a stochastic fashion (13–16), with the preferred protein targets being those that are most abundant and most proximal to DV Fe(II)PPIX (12). Altered Hb catabolism and Hb uptake in ARTR parasites (7, 10) are predicted to reduce the concentration of free FPIX, thereby reducing the efficiency of DHA activation, Hz inhibition, and the abundance of DHA-FPIX and DHA-protein adducts (11, 12). Although reduced FPIX adduct abundance has indeed been reported for ARTR parasites (11), no data comparing protein adduct abundances for ARTS versus ARTR parasites are yet available.

FIG 4.

Proposed antimalarial mechanism of ART-based drugs. Hemoglobin (Hb) is catabolized by plasmepsins and falcipains within the DV (“4”) to produce short peptides and amino acids, liberating ferriprotoporphyrin IX (FPIX) in the process. Dihydroartemisinin (DHA) is activated through a Fenton reaction with reduced ferriprotoporphyrin IX [Fe(II)PPIX] in the parasite DV, after which the DHA radical reacts with multiple targets, including proximal proteins in a stochastic fashion and FPIX itself. This chemistry has at least three effects: protein adducts may then (i) impact the proteasome pathway, while DHA-FPIX adducts will both (ii) inhibit FPIX detoxification to hemozoin (Hz) and (iii) cause oxidative stress. One consequence of the first effect would be the induction of parasite autophagy (58, 95, 96), which interestingly appears to be altered in CQR and ARTR parasites (58, 96). Altered abundance, or propeller domain mutation, of the PfK13 protein (green stars) partially lowers Hb uptake (8–10) and the amount of liberated FPIX, thereby decreasing the amount of activated DHA at key stages of parasite intraerythrocytic development. Green stars, PfK13 locations (see the text); 1, red blood cell cytosol; 2, parasitophorous vacuole; 3, parasite cytosol; 4, parasite digestive vacuole; 5, apicoplast; 6, Golgi apparatus; 7, endoplasmic reticulum; 8, nucleus; 9, mitochondrion; 10 (red box), DHA-FPIX adducts; 11 (green box), DHA-protein adducts. Abbreviations: AA, amino acids; Fe(III)PPIX, oxidized ferriprotoporphyrin IX.

While it might be tempting to conclude that Hz inhibition, or alkylation of any specific protein, is the most pharmacologically relevant target of any ART-based drug, several questions are important. One question is which specific adduct(s) exerts cytostatic (parasite growth-inhibitory) versus paracytocidal (parasite kill) effects (33, 95, 96). We propose that the Hz inhibition pathway is an additional, crucial target of ART-based drugs. We suggest that quantifying differences in the abundances of various DHA adducts and their biological effects, for ARTS versus ARTR parasites, will ultimately reveal their relative importance. Regardless, considering the unique nature of Hz detoxification along with the unique very high potency of ART-based drugs versus malarial parasites compared to other cells, we suggest that poisoning Hz crystallization via the formation of DHA-FPIX adducts is a particularly important target of ART-based drugs. This conclusion is further supported by the reduced abundance of DHA-FPIX adducts for ARTR versus ARTS parasites (11).

Finally, we note that similar to ART-based drugs, ozonides (OZs) containing trioxalane groups are also cleaved by Fe(II) and behave similarly to ART-based drugs (25). Some OZs have indeed recently been shown to alkylate heme within live P. falciparum parasites (24). We therefore propose that OZ compounds will have similar redox-dependent effects versus Hz inhibition.

MATERIALS AND METHODS

All chemicals and solvents were of reagent grade or better, purchased from Sigma-Aldrich (St. Louis, MO) or Fisher Scientific (Waltham, MA), and used without further purification unless otherwise noted. Sterile black flat-bottom tissue culture or nonsterile clear polystyrene 96-well plates and 12-channel basins were obtained from Fisher Scientific. SYBR green I nucleic acid stain was purchased from Invitrogen (Carlsbad, CA). Hemin, GSH, phosphatidylcholine (PC), and sodium propionate were purchased from Sigma-Aldrich. DHA was purchased from Tokyo Chemical Industry Co. (Tokyo, Japan). CQ and AQ were purchased from Sigma-Aldrich. ATS and ATM were purchased from Cayman Chemical Co. (Ann Arbor, MI). LF was purchased from Selleck Chemicals (Houston, TX). Mefloquine (MQ) was purchased from F. Hoffmann-La Roche Ltd. (Basel, Switzerland). Red blood cells and matched human serum for parasite culture were purchased from Valley Biomedical (Winchester, VA). P. falciparum clone HB3 (Honduras) was obtained from the Malaria Research and Reference Reagent Resource Center (Manassas, VA). Cam WT and Cam WT-C580Y strains of P. falciparum were a kind gift of David Fidock (Department of Microbiology and Immunology, Columbia University, New York, NY).

P. falciparum cultures.

P. falciparum strains HB3, Cam WT, and Cam WT-C580Y were maintained essentially as described previously, with minor modifications (7, 97, 98). In brief, cultures were maintained under an atmosphere of 5% CO₂, 5% O₂, and 90% N₂ at 2% hematocrit in complete RPMI 1640 medium supplemented with 25 mM HEPES (pH 7.4), 24 mM NaHCO₃, 11 mM glucose, 0.75 mM hypoxanthine, 20 μg/liter gentamicin, and either 10% type O⁺ human serum (HB3) or 0.5% Albumax II (Cam WT/Cam WT-C580Y). Parasitemia was monitored by Giemsa staining and adjusted every 48 h by the addition of fresh erythrocytes. Before being added to cultures, erythrocytes were washed with incomplete medium (RPMI 1640, 24 mM NaHCO₃, 11 mM glucose, and 0.75 mM hypoxanthine [pH 7.4]).

Oxidized β-hematin inhibition assay.

An oxidized β-hematin inhibition assay (O-BHIA) was performed essentially as described previously (33). Ten microliters of a hemin chloride solution (2 mM) in 0.1 M NaOH was transferred to each well of a 96-well plate, followed by 180 μl of propionate buffer (1.0 M; either pH 5.2 or pH 5.6) and 10 μl of a sonicated phosphatidylcholine solution (10 mg/ml), which serves as a physiological catalyst for Hz formation at close-to-physiological temperature and pH (see references 33 and 35). The drug was serially diluted across the wells, and appropriate controls (e.g., no lipid catalyst and no drug) were included in each plate. Plates were sealed and incubated at 37°C with gentle shaking for 16 h. Hz formation was terminated by the addition of 50 μl of an SDS solution (2.5% [wt/vol] in 0.1 M bicarbonate [pH 9.1]). The contents were gently mixed and then centrifuged (3,700 rpm for 5 min) to allow undissolved Hz crystals to settle. Fifty-microliter aliquots of the supernatant were transferred to a separate plate preloaded with 200 μl/well of an SDS solution (2.5% [wt/vol] in 0.1 M bicarbonate [pH 9.1]). The absorbance of uncrystallized heme at 405 nm was quantified using a Tecan Infinite F200 fluorescence microplate reader, and conversion to Hz was quantified as described previously (33).

Reduced β-hematin inhibition assay.

A reduced β-hematin inhibition assay (R-BHIA) was performed essentially as described above, with the following modifications. Ten microliters of hemin chloride (2 mM) in 80% DMSO was added to each well of a 96-well plate. Thirteen microliters of a glutathione solution (1.5 mM in 1× phosphate-buffered saline [PBS]) was added to each well such that FPIX heme and GSH were at molar equivalence. The content of the plate was mixed, and the plate was sealed and incubated at 37°C with gentle shaking for 2 h to reduce hemin to the ferrous form. Drug serial dilutions were prepared in either water (CQ, AQ, and PPQ) or DMSO (DHA, ATS, ATM, MQ, and LF) at a 100× concentration, and 2 μl of the drug solution was then added to each well in triplicate. The contents of the plate were mixed, and the plate was sealed and then incubated at 37°C with gentle shaking for 1 h to allow activation of the drug. One hundred sixty-five microliters of propionate buffer (1.0 M; either pH 5.2 or 5.6) and 10 μl of a sonicated phosphatidylcholine solution (10 mg/ml) were then added to each well, the contents were mixed, and the plate was sealed once again and incubated at 37°C with gentle shaking for 16 h. Hz formation was terminated by the addition of 50 μl of an SDS solution (7.5% [wt/vol] in 0.1 M bicarbonate [pH 9.1]). The contents were gently mixed and then centrifuged (3,700 rpm for 5 min) to allow undissolved Hz to settle. Fifty-microliter aliquots of the supernatant were transferred to a separate plate preloaded with 200 μl/well of an SDS solution (2.5% [wt/vol] in 0.1 M bicarbonate [pH 9.1]), and the absorbance was recorded as described above. Hz formation was then quantified as described previously (33).

R-BHIA fixed-ratio isobologram analysis.

To measure potentially synergistic effects between ACT drug pairs, an abbreviated isobologram analysis was performed utilizing the R-BHIA format and the Chou-Talalay method for drug combinatorial analysis (57). Drug solutions were prepared at 5 different ratios, 1:0, 3:1, 1:1, 1:3, and 0:1, scaled to the drugs’ Hz inhibition EC₅₀s [that is, “3” is 3 × (Hz inhibition EC₅₀ of the given drug alone)], and the combinations were analyzed using a fixed-ratio isobologram analysis. Fractional inhibitory concentration (FIC) values were calculated using equations 2 to 4:

$${\text{FIC}}_{\text{A}}=\frac{{\text{EC}}_{\text{50\hspace{0.17em}}}\text{of drug A in combination}}{{\text{EC}}_{\text{50\hspace{0.17em}}}\text{of drug A alone}}$$ (2)

$${\text{FIC}}_{\text{B}}=\frac{{\text{EC}}_{\text{50\hspace{0.17em}}}\text{of drug B in combination}}{{\text{EC}}_{\text{50\hspace{0.17em}}}\text{of drug B alone}}$$ (3)

$${\text{FIC}}_{\text{index}}={\text{FIC}}_{\text{A}}{\text{+\hspace{0.17em}FIC}}_{\text{B}}$$ (4)

Growth inhibition assay.

Drug stock solutions were prepared for each drug using complete cell culture medium supplemented with human serum (for HB3 parasite strains) or Albumax II (for Cam WT and Cam WT-C580Y strains). These drug solutions were prepared such that the final concentration approximately equaled the single-drug 50% inhibitory concentration (IC₅₀) following 5 to 6 2-fold dilutions.

The growth inhibition experiment was performed essentially as previously described (98). In short, 100 μl of the drug solution, prepared as described above, and 100 μl of an asynchronous parasite culture (final 1% parasitemia and 2% hematocrit) were added to a 96-well clear-bottom black plate. The plates were transferred to an airtight sterile chamber, gassed with a 5% CO₂–5% O₂–90% N₂ gas blend, and incubated at 37°C for 48 h. SYBR green I dye (50 μl; 10× in complete medium from a DMSO stock) was added, and plates were incubated for 1 additional hour at 37°C. Fluorescence was measured at 538 nm (485-nm excitation) using a Spectra GeminiEM plate reader (Molecular Devices, Sunnyvale, CA) fitted with a 530-nm long-pass filter. Background controls included fluorescence from uninfected red blood cells (98).