Probing the Antimalarial Mechanism of Artemisinin and OZ277 (Arterolane) with Nonperoxidic Isosteres and Nitroxyl Radicals

Abstract

Peroxidic antimalarials such as the semisynthetic artemisinins are critically important in the treatment of drug-resistant malaria. Nevertheless, their peroxide bond-dependent mode of action is still not well understood. Using combination experiments with cultured Plasmodium falciparum cells, we investigated the interactions of the nitroxide radical spin trap, 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO), and four of its analogs with artemisinin and the ozonide drug development candidate OZ277. The antagonism observed for combinations of artemisinin or OZ277 with the TEMPO analogs supports the hypothesis that the formation of carbon-centered radicals is critical for the activity of these two antimalarial peroxides. The TEMPO analogs showed a trend toward greater antagonism with artemisinin than they did with OZ277, an observation that can be explained by the greater tendency of artemisinin-derived carbon-centered radicals to undergo internal self-quenching reactions, resulting in a lower proportion of radicals available for subsequent chemical reactions such as the alkylation of heme and parasite proteins. In a further mechanistic experiment, we tested both artemisinin and OZ277 in combination with their nonperoxidic analogs. The latter had no effect on the antimalarial activities of the former. These data indicate that the antimalarial properties of peroxides do not derive from reversible interactions with parasite targets.

The semisynthetic artemisinins are critically important antimalarials in the treatment of drug-resistant malaria and are recommended for use in combination with other antimalarial drugs (32) to increase efficacy and preclude or delay drug resistance. The discovery of artemisinin led to an investigation of diverse classes of synthetic peroxides as potential antimalarial agents (17, 27). One such peroxide, the ozonide OZ277 (arterolane) (31), has now entered phase III clinical trials in the form of an arterolane maleate-piperaquine phosphate combination (22). A working hypothesis (16, 19, 23) put forth to account for the antimalarial specificity (18) of natural-product and synthetic peroxides is that the pharmacophoric peroxide bond undergoes reductive activation by heme released by parasite hemoglobin digestion (9, 11). The irreversible redox reaction between antimalarial peroxides and heme produces carbon-centered radicals or carbocations that alkylate heme (5, 20, 24, 25) and proteins (2, 3, 8, 33), leading to the perturbation of lipid components of the parasite digestive vacuole (6, 13). Although artemisinin and OZ277 are nearly equipotent inhibitors of Plasmodium falciparum growth in vitro (18), their very different 50% inhibitory concentrations (IC₅₀s) (79 and 7,700 nM) (30) against one putative target enzyme, the Sarcoendoplasmic reticulum Ca²⁺-ATPase PfATP6 (8, 12), reveal that the precise mechanism of action of antimalarial peroxides is still not well understood.

In this study, we report the results of two different types of in vitro P. falciparum combination experiments with artemisinin and OZ277 designed to better understand the mechanisms of action of these two drugs. In the first set of experiments, we assessed if the nitroxide free radical 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO) and several of its analogs (21) could antagonize the action of either artemisinin or OZ277 (Fig. 1). Nitroxide free radicals such as TEMPO are very efficient carbon-centered radical spin traps (21). In the second set of experiments, we used combinations of either artemisinin or OZ277 and their nonperoxidic counterparts deoxyartemisinin (4) and carbaOZ277 (18) (Fig. 1) to assess if reversible binding interactions with parasite targets play a role in the antimalarial properties of peroxides such as artemisinin and OZ277.

FIG. 1.

Structures of artemisinin, deoxyartemisinin, OZ277, carbaOZ277, and TEMPO derivatives.

MATERIALS AND METHODS

Materials.

Deoxyartemisinin, OZ277, and carbaOZ277 were synthesized as previously described (4, 18, 31). The TEMPO analogs were purchased from Sigma-Aldrich, [8-³H]hypoxanthine was purchased from ANAWA Trading SA, and artemisinin was purchased from Saokim Pharma. The test compounds were dissolved in dimethyl sulfoxide (DMSO) at 10 mg/ml or, where required, at 50 mg/ml. The stock solutions were kept at 4°C for less than 6 months.

Parasite assays.

The chloroquine-sensitive NF54 isolate and the chloroquine- and pyrimethamine-resistant K1 isolate of P. falciparum, provided by F. Hoffmann-La Roche Ltd. (Basel, Switzerland), were used for the in vitro assays. Parasite cultivation in RPMI 1640 medium (10.44 g/liter) supplemented with HEPES (5.94 g/liter), Albumax II (5 g/liter), hypoxanthine (50 mg/liter), sodium bicarbonate (2.1 g/liter), and neomycin (100 mg/liter) was performed according to a method described previously by Trager and Jensen (29), in an atmosphere of 93% N₂, 4% CO₂, and 3% O₂ at 37°C. P. falciparum growth was assessed by measuring the incorporation of the nucleic acid precursor [³H]hypoxanthine (7). In detail, compounds were dissolved in DMSO (10 mg/ml) and diluted in hypoxanthine-free culture medium. Infected erythrocytes (100 μl per well with 2.5% hematocrit and 0.3% parasitemia) were added to each drug titrated in 100-μl duplicates over a 64-fold range. After 48 h of incubation, 0.5 μCi of [³H]hypoxanthine in 50 μl of medium was added, and plates were incubated for an additional 24 h. Parasites were harvested onto glass-fiber filters, and radioactivity was counted by using a Beta-Plate liquid scintillation counter (Wallac, Zurich, Switzerland). The results were recorded as counts per minute per well at each drug concentration and expressed as a percentage of the untreated controls. Fifty-percent inhibitory concentrations (IC₅₀) were estimated by linear interpolation (15). The experimental DMSO concentrations had no inhibitory effect on parasite growth.

Combination experiments.

The fixed-ratio method (26) uses serial dilutions (1:2) of fixed ratios of two compounds. Initially, the IC₅₀s of the test compounds alone were determined. Subsequently, compound solutions were diluted with hypoxanthine-free culture medium to initial concentrations of at least 1.5 times the predetermined IC₅₀s. The solutions were combined in IC₅₀ ratios of 1:1, 1:3, and 3:1. Single and combination compound solutions were then introduced into 96-well plates to give duplicative rows. The IC₅₀s of the compounds in combination were expressed as fractions of the IC₅₀s of the compounds alone and were recorded as fractional inhibitory concentrations (FICs) for compounds A and B, respectively. These data were expressed numerically as the sum FICs (ΣFICs) of FIC_A and FIC_B. ΣFIC values indicate the nature of the interactions as follows: ΣFIC of <0.8 is synergism, ΣFIC of 0.8 to 1.4 is additive, and ΣFIC of >1.4 is antagonism. Cutoff values of 0.8 for synergism and 1.4 for antagonism were determined by mixing the same compound at various ratios and accounting for experimental variation. Statistical significance was determined by using the Student t test. Compound interactions were also assessed by using a checkerboard combination assay (10) in which one compound was kept at a constant subinhibitory concentration and the other compound was serially diluted (1:2). To assess dose-response relationships, the subinhibitory concentrations were increased in four steps from 0 to approximately 0.5 times the respective IC₅₀s.

RESULTS

Table 1 shows the IC₅₀s of all tested compounds. Against P. falciparum strain NF54, the IC₅₀s of artemisinin (2.7 ± 1.2 ng/ml) and OZ277 (0.54 ± 0.29 ng/ml) were approximately 10,000-fold lower than the IC₅₀s of their respective nonperoxidic analogs, deoxyartemisinin (27,000 ± 3,200 ng/ml) and carbaOZ277 (5,200 ± 380 ng/ml). The IC₅₀s of these four compounds against the K1 strain of P. falciparum varied by no more than 2.3-fold from those against the NF54 strain. The IC₅₀s for TEMPO and the functionalized TEMPO derivatives against the NF54 strain of P. falciparum ranged from 10,000 to 35,000 ng/ml; the most potent of these was 4-amino-TEMPO (Table 1). The K1 strain of P. falciparum was similarly affected by all TEMPO derivatives.

TABLE 1.

IC₅₀s against the NF54 and K1 strains of P. falciparum^a

Compound	Mean IC50 (ng/ml) ± SD
	NF54	K1
Artemisinin	2.7 ± 1.2	1.6 ± 0.84
Deoxyartemisinin	27,000 ± 3,200	19,000 ± 2,500
OZ277	0.54 ± 0.29	0.57 ± 0.35
CarbaOZ277	5,200 ± 380	2,200 ± 190
TEMPO	25,000 ± 8,900	16,000 ± 4,000
4-Oxo-TEMPO	28,000 ± 5100	12,000 ± 5,100
4-Hydroxy-TEMPO	34,000 ± 2,800	36,000 ± 4,300
4-Carboxy-TEMPO	33,000 ± 4,400	34,000 ± 5,800
4-Amino-TEMPO	10,000 ± 2,900	8,200 ± 770

^aValues are the means and standard deviations from ≥5 experiments.

The ΣFIC values (NF54 strain) for artemisinin and OZ277 in combination with the TEMPO analogs are shown in Table 2. For artemisinin and OZ277, an antagonistic tendency was evident for 4/5 and 2/5 of the TEMPO analogs, respectively. However, antagonism with artemisinin was statistically significant only for 4-oxo-TEMPO and 4-amino-TEMPO, and antagonism with OZ277 was statistically significant only for 4-amino-TEMPO. ΣFIC data (not shown) obtained for the K1 strain were very similar to the NF54 data shown in Table 2.

TABLE 2.

In vitro interactions of artemisinin and OZ277 plus TEMPO drug combinations against the NF54 strain of P. falciparum^a

Drug combination	Mean ΣFIC50 ± SD at a ratio ofb:			Mean ΣFIC50 ± SD	Interaction
	1:3	1:1	3:1
Artemisinin + TEMPO	1.3 ± 0.6	2.1 ± 1.3	2.2 ± 1.2	1.9 ± 0.4	Antagonism
Artemisinin + 4-oxo-TEMPO	2.3 ± 1.2c	2.6 ± 1.0c	2.0 ± 0.6c	2.3 ± 0.2c	Antagonism
Artemisinin + 4-hydroxy-TEMPO	1.8 ± 0.6	4.0 ± 2.2d	2.5 ± 0.6	2.8 ± 0.9	Antagonism
Artemisinin + 4-carboxy-TEMPO	1.4 ± 0.1	1.4 ± 0.1	1.2 ± 0.1	1.3 ± 0.1	Addition
Artemisinin + 4-amino-TEMPO	1.3 ± 0.1	1.9 ± 0.2c	2.8 ± 0.4c	2.0 ± 0.6	Antagonism
OZ277 + TEMPO	1.8 ± 0.2	1.5 ± 0.1	1.2 ± 0.1	1.5 ± 0.2	Antagonism
OZ277 + 4-oxo-TEMPO	1.4 ± 0.1	1.3 ± 0.1	1.1 ± 0.1	1.3 ± 0.1	Addition
OZ277 + 4-hydroxy-TEMPO	1.5 ± 0.1	1.3 ± 0.1	1.2 ± 0.1	1.3 ± 0.1	Addition
OZ277 + 4-carboxy-TEMPO	1.3 ± 0.1	1.3 ± 0.1	1.2 ± 0.0	1.3 ± 0.1	Addition
OZ277 + 4-amino-TEMPO	2.3 ± 0.3c	2.4 ± 0.2c	1.7 ± 0.0c	2.1 ± 0.3e	Antagonism

^aValues are the means and standard deviations from ≥3 experiments.

^bThe 1:1, 1:3, and 3:1 ratios refer to fixed-concentration ratios of compound A to compound B.

^cSignificant (P < 0.05) based on a cutoff value for antagonism of 1.4.

^dEven with six independent repetitions of the experiment, we obtained only two values; in four of the experiments, the 1:1 combination was not active enough, precluding the calculation of an IC₅₀.

^eP = 0.052 based on a cutoff value for antagonism of 1.4.

Representative combinations were also tested with the checkerboard assay, in which the performance of an active compound is evaluated in the presence of an inactive compound. The concentrations used were 0, 750, 1,500, and 3,000 ng/ml for 4-amino-TEMPO and 0, 5,000, 10,000, and 20,000 ng/ml for 4-hydroxy-TEMPO. The IC₅₀ of artemisinin increased with increasing concentrations of the TEMPO analogs. This is in accordance with the results from the fixed-ratio assay, where we also observed antagonism between artemisinin and the TEMPO analogs. However, contrary to the results from the fixed-ratio assay, in which we observed antagonism between OZ277 and 4-amino-TEMPO, we observed no antagonistic effect of 4-amino-TEMPO in the checkerboard assay (Fig. 2). Data for the K1 strain (not shown) paralleled those for the NF54 strain.

FIG. 2.

In vitro P. falciparum IC₅₀s of artemisinin and OZ277 in the presence of increasing 4-amino-TEMPO concentrations representative of the interactions of the peroxides and TEMPO analogs tested. IC₅₀s of artemisinin are the means from 4 independent experiments, and those of OZ277 are the means from 2 independent experiments.

The ΣFIC values (NF54 strain) from combination experiments with either artemisinin or OZ277 and its nonperoxidic analogs are depicted in Table 3. Neither artemisinin nor OZ277 was antagonized by either nonperoxidic analog; all four combinations were additive. Measured ΣFIC values for the K1 strain (data not shown) differed from those of the NF54 strain by less than 1.4-fold. We also tested the same combinations with the checkerboard method, and in accordance with the ΣFIC data shown in Table 3, we observed no increase in the IC₅₀s of artemisinin and OZ277 (data not shown) with increasing concentrations of their nonperoxidic analogs.

TABLE 3.

In vitro interactions of either artemisinin or OZ277 and the nonperoxidic analogs deoxyartemisinin and carbaOZ277 against the NF54 strain of P. falciparum^a

Drug combination	Mean ΣFIC50 ± SD at a ratio ofb:			Mean ΣFIC50 ± SD	Interaction
	1:3	1:1	3:1
Artemisinin + deoxyartemisinin	1.2 ± 0.0	1.1 ± 0.1	1.0 ± 0.1	1.1 ± 0.1	Addition
OZ277 + carbaOZ277	1.4 ± 0.1	1.3 ± 0.1	1.2 ± 0.1	1.3 ± 0.1	Addition
Artemisinin + carbaOZ277	1.1 ± 0.1	1.1 ± 0.0	1.1 ± 0.1	1.1 ± 0.0	Addition
OZ277 + deoxyartemisinin	1.2 ± 0.1	1.2 ± 0.1	1.1 ± 0.1	1.2 ± 0.1	Addition

^aValues are the means and standard deviations from ≥3 experiments.

^bThe 1:1, 1:3, and 3:1 ratios refer to fixed-concentration ratios of compound A to compound B.

DISCUSSION

The antagonism observed with combinations of either artemisinin or OZ277 with the TEMPO analogs supports the hypothesis that the formation of carbon-centered radicals is critical for the activities of these two antimalarial peroxides. At first glance, it may seem surprising that the TEMPO analogs showed a tendency toward greater antagonism with artemisinin than they did with OZ277. However, based on subtle differences in the reaction pathways of the carbon-centered radicals derived from artemisinin and OZ277, this outcome is not unexpected. First, both primary and secondary carbon-centered radicals formed from artemisinin self-quench (16, 23) in intramolecular reactions to a greater extent than did the secondary carbon-centered radical formed from OZ277 and its parent ozonide, OZ03 (28). Second, for data obtained under identical experimental conditions (5), significantly less heme alkylation was observed for artemisinin and its semisynthetic derivatives dihydroartemisinin, artemether, and sodium artesunate than for a series of equipotent ozonides, including OZ03 and OZ277.

In Fig. 3, the very different reaction outcomes between iron(II) and the semisynthetic artemisinin, 10-deoxoartemisinin (14), and OZ03 (28), the core substructure of OZ277, are illustrated. The reactions were conducted under identical conditions in two different laboratories. In contrast to the efficient reaction of OZ03 with 4-oxo-TEMPO in the presence of ferrous acetate to form the radical-trapped aminoxy adduct in a yield of 56% accompanied by a 34% combined yield of the self-quench reaction products (28), the same reaction with 10-deoxoartemisinin formed the radical-trapped aminoxy adduct in only 3% yield accompanied by an 85% combined yield of the self-quench reaction products (14). This low reaction efficiency of 10-deoxoartemisinin with 4-oxo-TEMPO derives from the greater tendency of artemisinin-derived carbon-centered radicals to self-quench, resulting in a lower proportion of radicals available for subsequent chemical reactions. This explains why radical traps such as TEMPO and its analogs have a greater antagonistic effect on artemisinin than on OZ277.

FIG. 3.

Reaction of 10-deoxoartemisinin and OZ03 with 1.5 equivalents of ferrous acetate in the presence of a 2-fold molar excess of 4-oxo-TEMPO.

Since artemisinin appears to react less efficiently with heme than OZ277, why do these two drugs have such similar potencies against P. falciparum (18)? To answer this question, one might consider that the physicochemical properties of the OZ277-heme and artemisinin-heme covalent adducts (Fig. 4) are quite different in that the OZ277-heme adduct (5) has an additional negative charge due to a carboxylate functional group, whereas the artemisinin-heme covalent adduct (24) retains the ionization characteristics of heme. The more lipophilic artemisinin-heme covalent adduct may mediate the antiplasmodial effects of artemisinin more efficiently than the more polar OZ277-heme covalent adduct mediates those of OZ277. For example, the more lipophilic artemisinin-heme covalent adduct would be expected to have a stronger interaction with relatively hydrophobic membrane lipids (6, 13) in the food vacuole.

FIG. 4.

Structures of artemisinin-heme and OZ277-heme covalent adducts.

The combination experiments with either artemisinin or OZ277 along with the nonperoxidic analogs deoxyartemisinin and carbaOZ277 revealed that the latter had no effect on the antimalarial activities of the former. This outcome is precisely what one would expect if the antimalarial properties of peroxides are derived solely from irreversible bond-forming interactions with parasite targets. If, on the other hand, this irreversible chemistry occurred only after a reversible interaction with one or more parasite targets, one would have expected to observe substantial antagonism by the nonperoxidic analogs, reminiscent of substrate protection against enzyme inactivation by suicide enzyme inhibitors (1). From these data, we can conclude that the antimalarial properties of peroxides such as artemisinin and OZ277 do not derive from reversible interactions with parasite targets. Furthermore, previous work (18) demonstrating the peroxide bond-dependent activity of artemisinin and OZ277 against hemoglobin-digesting protozoa (such as plasmodia) is consistent with the idea that the irreversible interactions of peroxides with parasite targets are iron dependent.