The antimalarial drug artemisinin alkylates heme in infected mice

Abstract

Heme alkylation by the antimalarial drug artemisinin is reported in vivo, within infected mice that have been treated at pharmacologically relevant doses. Adducts resulting from the alkylation of heme by the drug were characterized in the spleen of treated mice, and their glucuroconjugated derivatives were present in the urine. Because these heme-artemisinin adducts were not observed in noninfected mice, this report confirms that the alkylating activity of this antimalarial drug is related to the presence of the parasite in infected animals. The identification of heme-artemisinin adducts in mice should be considered as the signature of the alkylation capacity of artemisinin in vivo.

Since the beginning of the last decade, many research groups have been working on the mechanism of action of antimalarial drugs and have prepared new drug candidates to fight against the comeback of this parasitic disease (1-3). Many strains of Plasmodium falciparum, responsible for severe malaria, have become resistant to classical drugs like chloroquine (4). Quinoline-based drugs (quinine and chloroquine) are able to inhibit, via a strong stacking, the aggregation of heme liberated during the digestion of hemoglobin by the parasite into the redox inactive hemozoin (5). The mechanism of action of artemisinin, a nonquinoline antimalarial drug containing a 1,2,4-trioxane motif (1, Fig. 1), has been the matter of intense debates for the last decade (6-10).

Fig. 1.

Alkylation of heme by artemisinin within P. vinckei-infected mice.

Among the different hypotheses on the mechanism of action of artemisinin, one concerns the reductive activation of its endoperoxide function by the iron(II) center of free heme, generating a C4-centered radical that alkylates the meso positions of protoporphyrin-IX (2, Fig. 1) (11-15). Many antimalarial trioxanes are also able to alkylate heme or heme models (8, 16, 17).

Is the formation of heme-artemisinin adducts a laboratory curiosity or is it biologically relevant and related to the mechanism of action of the drug in vivo? We addressed this question by treating Plasmodium vinckei-infected mice with artemisinin at pharmacological doses by i.p. or oral route. The covalent adducts heme-artemisinin 2 and 3 were identified in the spleen of all treated mice. The hydroxylated and glucuroconjugated derivatives of heme-artemisinin adducts 4 and 5 were identified in the urine of treated mice. These adducts were not detected after treatment of healthy mice by artemisinin.

Materials and Methods

Chemicals. β-Glucuronidase from bovine liver (EC 3.2.1.31, type B-1, 1.24 × 10⁶ units/g) was purchased from Sigma.

Animal Treatment. Female Swiss albino mice weighing 25-30 g were inoculated i.p. with 6 × 10⁷ erythrocytes parasitized with P. vinckei petteri strain. Parasitemia was determined by microscopic method on Giemsa thin blood smears. When it was higher than 40% (usually 3 days), animals were treated once, via i.p. or oral route, with artemisinin diluted in DMSO (100 μl) at doses of 200 or 100 mg/kg (i.p. route) or 100, 50, or 20 mg/kg (oral route).

One lot of four infected mice was treated with artemisinin at 200 mg/kg in 100 μl of DMSO by i.p. route for urine analyses. Three lots of two, one, and one mice, respectively, were treated at 100 mg/kg by i.p. route for urine and spleen analyses. Three lots (n = 2 for each lot) were treated at 100, 50, and 20 mg/kg, respectively, by oral route for urine and spleen analyses. For control experiments, infected mice received either no treatment, pure DMSO (100 μl, n = 2 by i.p. route, n = 1 by oral route), or heme-artemisinin adducts 2 + 3 (200 mg/kg in 100 μl of DMSO by i.p. route, n = 2). Healthy mice were treated with artemisinin (200 mg/kg, i.p. route, n = 2; 100 mg/kg, i.p. route, n = 1; or 100 mg/kg, oral route, n = 1), adducts 2 + 3 (200 mg/kg, i.p. route, n = 2), or pure DMSO (100 μl, n = 2 by i.p. route; n = 1 by oral route). The spleens of healthy mice without any treatment (n = 3) were also examined. Urine was collected by lot. Each spleen was analyzed individually.

Urine Treatment. Mice urine was collected for 7 h after treatment and aliquoted by 50 μl of fractions. One fraction was diluted with a solution of β-glucuronidase (1,600 units) in CH₃COONa, 100 mM, pH 5.2 (50 μl), and allowed to stand at 37°C for 16 h. Two other fractions were diluted with CH₃COONa 100 mM, pH 5.2 (50 μl), and kept at either -20°C or 37°C for 16 h.

Spleen Treatment. Spleen was removed 7 ± 0.5 h after treatment. Each whole mouse spleen was crushed with sand in a mortar then extracted with glacial acetic acid (1 ml).

HPLC and Liquid Chromatography-MS (LC-MS) Analyses. HPLC analyses were performed by using a 10-μm C18 Nucleosil column (250 × 4.6 mm). Eluents were (A) methanol/water/formic acid, 60/40/1, vol/vol and (B) methanol/water/formic acid, 80/20/1, vol/vol. The elution gradient was linear from A/B = 100/0 to A/B = 0/100 in 15 min, followed by 15 min at A/B = 0/100. Flow rate was 1 ml/min, and UV-Vis detection was at 406 nm.

MS Analyses. Positive-ion electrospray mass spectra were obtained by LC-MS by using an API 365 Sciex PerkinElmer instrument. The detection was scan mode (total ionic current) with a step size of 0.15 atomic mass units (amu) and scan range of 500-1,200 amu or single ion monitoring mode, with a dwell time of 150 msec for better sensitivity. Quantification of adducts 2 + 3 was made by the following method: HPLC allowed quantification of 2 + 3 + heme-Cl (m/z = 898.5, 838.5, and 650.3, respectively), estimating that these hemin derivatives have similar extinction coefficients; the ratio 2 + 3/heme-Cl was then calculated from the relative intensities of the corresponding ionic currents, considering these two species have similar volatilities.

Hydrogen Peroxide-Mediated Chlorination of Heme. Hemin (3.5 mg, 5.4 μmol) was incubated in DMSO (0.5 ml) with nBu₄N⁺Cl^- (6 mg, 22 μmol) and H₂O₂ 100 mM (10 μl) for 30 min at room temperature. A new aliquot of H₂O₂ was then added (10 μl), and the solution was allowed to stand at room temperature for a further 30 min. For analysis, an aliquot of the reaction mixture was 10-fold diluted with glacial acetic acid.

Results and Discussion

When the parasitemia was as high as 40-75% (typically 3 days after infection), mice were treated once with artemisinin doses at 100 or 200 mg/kg by i.p. route or at 20, 50, or 100 mg/kg by oral route (solutions in DMSO, 100 μl). These doses are within the range or below the curative dose of CD₅₀ = 140 mg/kg reported for artemisinin in the Rane test (single s.c. dose on day + 3 after infection) (18). Mice with the same treatment were then housed with a standard laboratory diet and gavaged with pure water (3 × 200 μl) after 1.5, 3.5, and 5.5 h. Urines were pooled for mice that had the same treatment and analyzed by LC-MS. After 7 ± 0.5 h, mice were killed, and spleens were individually crushed, extracted with acetic acid, and analyzed by HPLC and LC-MS.

Covalent Adducts Heme-Artemisinin in the Spleen. The LC-MS analysis of the spleen extracts (without any additional treatment) of an infected mouse treated with artemisinin (100 mg/kg, i.p. route) is reported in Fig. 2. The ionic currents of m/z = 898.5 and 838.5, corresponding to heme-drug adducts 2 and 3, respectively, were detected along with unreacted heme (m/z = 616.3) (Fig. 2 b-d). Compound 2 is a mixture of four regioisomers resulting from the alkylation of the four meso positions of heme by the C4-centered radical generated by the drug activation by iron(II)-heme, as described (8, 12-14). For this reason, the ionic current of 2 appears as three partly resolved peaks having the same mass spectrum (R_t = 17.5, 18.2, and 19.3 min, m/z = 898.5, Fig. 2c). Compound 3 (four regioisomers) appeared as two chromatographic peaks (R_t = 18.4 and 18.9 min, m/z = 838.5, Fig. 2d). It is formed by intramolecular addition of the hydroxyl function at C12a onto the carbonyl at C10, followed by the loss of acetic acid and a transesterification process. A similar reaction has already been reported for heme alkylation by a C10-substituted artemisinin derivative (15). The chromatographic behavior and mass spectra (including isotopic pattern) of adducts 2 and 3 detected in the mouse spleen were consistent with those of chemically prepared adducts.§

Fig. 2.

LC-MS of spleen extracts of a P. vinckei-infected mouse treated with artemisinin by i.p. route (100 mg/kg). (a) UV-visible (UV-Vis.) trace (b-e) X ionic current (XIC) traces for m/z = 616.5, 898.5, 838.5, and 650.5 amu, respectively. Mass detection is in scan mode.

In several spleen extracts, compound 3 is the major product and should be considered as a strong marker of the alkylation of heme by artemisinin in vivo. Adduct 3 was detected in the spleen of all infected mice treated either by oral or i.p. routes (number of mice = six and three, respectively). It should be noted that the m/z = 838.5 ionic current was detected in the spleen of mice treated by oral route at doses as low as 50 and 20 mg/kg (Fig. 3), using single ion monitoring mode electrospray ionization (ESI)⁺-MS. As control experiments, infected mice received pure DMSO (100 μl, oral route), healthy mice received either no treatment, artemisinin (100 mg/kg i.p. or oral routes), or pure DMSO (100 μl, oral). In all control mice, compounds 2 and 3 were not detected. On the bases of HPLC and MS, the amount of heme-artemisinin adducts 2 + 3 in the spleen of infected mice treated at 100 mg/kg (i.p. route) was estimated to be 30-100 nmol, corresponding to 0.4-1.1% of the injected artemisinin dose. This value can be considered only as indicative.

Fig. 3.

LC-MS of mice spleen extracts, XIC traces for m/z = 838.5 amu. (a) Healthy mouse without treatment. (b) P. vinckei healthy mouse treated with artemisinin by oral route (100 mg/kg in 100 μl of DMSO). (c-e) Infected mice treated with artemisinin in DMSO (100 μl) by oral route at 100, 50, and 20 mg/kg, respectively (the broad signal at 16.5 min is assigned to residual ionic current of a multicharged ion centered at 838.2 that belongs to a protein). Mass detection is in single ion monitoring mode.

In addition, it should be mentioned that all of the spleen extracts contained highly variable amounts of a compound with m/z = 650.3, which corresponds to a monochlorinated heme derivative [M(heme-H+Cl)] (Fig. 2e). This product was present even in the spleen of healthy mice in the absence of any treatment. It is therefore not related to the disease or the treatment. This compound, named heme-Cl for short, is probably generated by a peroxidase-type reaction of heme not been described until now.¶

Glucuroconjugated Derivatives of Heme-Artemisinin Adducts in Urine. The urine of malaria-infected mice treated with artemisinin at 100 or 200 mg/kg (i.p. route) or at 50 mg/kg (oral route) was analyzed by LC-MS. Peaks corresponding to heme derivatives more polar than heme itself were evidenced (Fig. 4a). Compounds 4 and 5 (Fig. 1) have a blue-shifted absorption with respect to heme (λ_max = 406-408 nm, compared with 398 nm for heme), as observed for adducts 2 and 3. Their m/z values were 1,090.5 and 1,030.5 amu, corresponding to the hydroxylated and glucuroconjugated derivatives of 2 and 3, respectively. When the artemisinin dose was 100 mg/kg (i.p.), the quantity of heme-drug adducts in urine was 33-81 nmol, corresponding to 0.4-0.9% of the injected artemisinin dose (evaluated by HPLC).

Fig. 4.

LC-MS of urine of two P. vinckei-infected mice treated with artemisinin by i.p. route at 100 mg/kg. (a) Urine kept at -20°C. (b) Urine heated at 37°C for 16 h. (c) Urine heated at 37°C for 16 h in the presence of β-glucuronidase. Mass detection is in scan mode.

To confirm the identification of compounds 4 and 5, urine samples were treated with β-glucuronidase, known to cleave specifically 1β-d-glucuronides, at 37°C for 16 h in an acetic acid medium (20). Adducts 4 and 5 then disappeared by loss of a glucuronic acid unit, giving rise to less polar compounds with m/z = 914.5 and 854.5 amu (Fig. 4c), which correspond to the expected hydroxylated heme-artemisinin adducts 6 and 7 (Fig. 1).

As a control experiment, urine was heated under the same conditions but without β-glucuronidase (urine/CH₃COONa 100 mM, pH 5.2, 1/1, vol/vol, 37°C, 16 h). Under these conditions, adduct 4 was quantitatively transformed to 8 (m/z = 1,048.5), corresponding to the hydrolysis of the lactone function and loss of an acetic acid molecule (Fig. 1). In addition, traces of adduct 9, corresponding to the loss of a glucuronic acid moiety from 8 (m/z = 872.5), were also detected (Fig. 4b). However, in the presence of β-glucuronidase, this pathway 4 → 8 → 9 was always a minor route, as attested by trace amounts of adduct 9 (m/z = 872.5, not shown in Fig. 4c).

The small quantities of hydroxylated and glucuroconjugated heme-artemisinin adducts obtained in these animal studies did not allow NMR identification of the exact hydroxylation sites on these adducts. A heme-artemisinin adduct with glucuronidation at a heme carboxyl group should have m/z = 1,074.5, not 1,090.7. The m/z value = 1,090.7 is consistent only with hydroxylation + glucuronidation. Such reaction cannot be excluded on a propionic -CH₂- of heme. However, despite the fact that the urine of infected mice contained a large amount of free heme (m/z = 616.3, Fig. 4), heme derivatives having undergone glucuronidation at the carboxyl function (m/z = 792.5), hydroxylation (m/z = 632.3), or hydroxylation + glucuronidation on a propionic -CH₂- (m/z = 808.5) were never detected (the degradation of heme in animals is mediated by heme oxygenase via an oxidation at a meso position; see ref. 21). On the contrary, the artemisinin skeleton is known to undergo hydroxylation and glucuronidation in healthy rat and human (22). The glucuronidation of heme-artemisinin adducts on the artemisinin-derived moiety is therefore the more likely pathway.

It was previously reported (22) that artemisinin derivatives injected into rats or humans undergo phase I hydroxylations at several positions by P450 enzymes, followed by phase II glucuroconjugation. Liver was the main organ of their metabolism. However, these studies were performed on healthy animals or patients, without the possibility of detecting products derived from the reactivity of artemisinin in malaria-infected erythrocytes.

Healthy mice were treated with artemisinin or chemically prepared adducts 2 + 3§ (200 mg/kg in 100 μl of DMSO, i.p.), and malaria-infected mice were treated with adducts 2 + 3 (200 mg/kg in 100 μl of DMSO) or with pure DMSO by i.p. route. In none of these control mice did the spleen and urine contain detectable amounts of adducts 2 and 3 or 4-9, respectively. This might be because of the poor absorption of adducts 2 and 3 when administered by i.p. route.

In the present work, the in vivo formation of heme-artemisinin adducts is strictly related to both the disease and the artemisinin treatment at curative doses. These adducts were not observed in healthy animals treated with artemisinin. Furthermore, nonmetabolized adducts 2 and 3 were detected in the spleen, the main organ involved in the elimination of damaged red blood cells.

Besides spleen and urine, several other organ extracts were analyzed by HPLC and/or LC-MS. No heme-artemisinin adducts could be detected in liver, gall bladder, duodenum, and kidneys of malaria mice treated with artemisinin by i.p. route (100 mg/kg). However, the presence of glucuroconjugated derivatives in the urine clearly indicates that these compounds have gone through liver and kidneys. But the metabolism of heme-artemisinin adducts formed in infected erythrocytes is a dynamic phenomenon and, among the analyzed organs, only the spleen retains adducts 7 h after treatment.

Conclusion

Because the alkylation of heme by artemisinin occurred only in infected mice treated at pharmacologically relevant doses, these heme-artemisinin adducts should be considered the signature of the strong alkylating capacity of this antimalarial drug. We have to keep in mind that such alkylating property of artemisinin is triggered by an iron(II) center; consequently, other parasitic iron-containing enzymes or proteins also have to be considered as potential activators of trioxane-based antimalarial drugs.

The results reported in this manuscript are the validation in animals of the concept of alkylating trioxanes used in the design of new antimalarial agents, including the dual molecules named trioxaquines (8, 23).