Relationship between Antimalarial Activity and Heme Alkylation for Spiro- and Dispiro-1,2,4-Trioxolane Antimalarials

Abstract

The reaction of spiro- and dispiro-1,2,4-trioxolane antimalarials with heme has been investigated to provide further insight into the mechanism of action for this important class of antimalarials. A series of trioxolanes with various antimalarial potencies was found to be unreactive in the presence of Fe(III) hemin, but all were rapidly degraded by reduced Fe(II) heme. The major reaction product from the heme-mediated degradation of biologically active trioxolanes was an alkylated heme adduct resulting from addition of a radical intermediate. Under standardized reaction conditions, a correlation (R² = 0.88) was found between the extent of heme alkylation and in vitro antimalarial activity, suggesting that heme alkylation may be related to the mechanism of action for these trioxolanes. Significantly less heme alkylation was observed for the clinically utilized artemisinin derivatives compared to the equipotent trioxolanes included in this study.

Peroxide antimalarials based on artemisinin are currently the most important treatment option for malaria, a disease which afflicts over 500 million people annually (24). The worldwide prevalence of multidrug-resistant parasites, the potent antimalarial activity of artemisinin against multidrug-resistant parasites, and the lack of clinical drug resistance observed for artemisinin derivatives have resulted in the WHO recommendation of artemisinin-based combination therapy as first-line treatment for uncomplicated Plasmodium falciparum malaria (9). Extensive utilization of artemisinin derivatives has been limited by cost, availability, and pharmacokinetic limitations; however, the development of fully synthetic peroxides may enable production of highly active antimalarials with improved biopharmaceutical properties at a lower cost (11, 27).

Despite the known potency of artemisinin and the numerous semisynthetic and fully synthetic analogues, the exact mechanism of action remains unresolved (9, 10). It is known that the peroxide bond is essential for activity, and it is generally accepted that the peroxide reacts with intraparasitic iron to form free radicals, carbocations, or other reactive species (15). The intraerythrocytic environment of the Plasmodium parasite may give rise to various sources of iron, including iron-containing proteins such as hemoglobin, free heme released following hemoglobin digestion, or free iron from the cytosolic labile iron pool (22). The mechanism(s) by which the reactive species kill the parasite may involve alkylation of specific parasite proteins, such as PfATP6 (8) or TCTP (1), or nonspecific damage to various intracellular targets, including proteins and membranes (15). Alternatively, after activation by heme iron, the resultant free radicals can react directly with the heme porphyrin, to form an adduct that may be responsible for antimalarial activity (19, 20).

Development of the fully synthetic trioxolane OZ277 (trioxolane 1) (Fig. 1) has provided a potent peroxide antimalarial with a significantly different molecular structure than artemisinin (30). The mechanism of action for trioxolane 1 is assumed to be analogous to that of the artemisinin derivatives; however, the only evidence reported for this mechanism is the generation of free radicals by inorganic iron salts (4) and antagonism of the antimalarial effect by the iron-chelating agent, desferrioxamine (29). Despite recent evidence supporting a role for PfATP6 in the mechanism of action of artemisinin (14), trioxolane 1 showed minimal inhibition of this intraparasitic target (29), indicating that other targets may be more significant for this and similar trioxolanes. It has been shown that iron-mediated reactivity is required for the antimalarial activity of trioxolanes (4); however, like the artemisinin derivatives, there was no correlation between iron-mediated reaction rates and antimalarial activity for the trioxolanes tested. For some artemisinin derivatives, a relationship has been shown between antimalarial activity and heme binding (17) or porphyrin alkylation (18) and it is thought that heme may be the most biologically relevant source of iron within the Plasmodium parasite (20).

FIG. 1.

Structures of artemisinin and trioxolane 1 (OZ277 as tosylate salt).

The aim of this study was to investigate the heme-mediated reactivity of a series of trioxolane antimalarials. Standardized reducing conditions were used to compare degradation profiles and product formation. Structurally similar trioxolanes with a wide range of antimalarial activities were investigated to determine whether the reaction with heme is related to antimalarial activity.

MATERIALS AND METHODS

Materials.

The trioxolanes were synthesized via Griesbaum coozonolysis and post-ozonolysis modifications as previously described (6, 7, 26, 30-33). Artemisinin (98%), dihydroartemisinin (97%), hemin (≥80%) and sodium hydrosulfite (sodium dithionite) (∼85%) were obtained from Sigma-Aldrich (St. Louis, MO). Artemether and sodium artesunate were obtained from TDR/WHO (Geneva, Switzerland). High-performance liquid chromatography (HPLC)-grade acetonitrile (ACN) and Milli-Q water were vacuum degassed prior to use to reduce dissolved oxygen concentrations. Hemin stock solution (0.5 mM) was freshly prepared in 0.1 M NaOH, and stock solutions of peroxide antimalarials (1.85 mM) were prepared in ethanol.

Reaction conditions.

Hemin (final concentration, 10 μM) and 50% ACN-water (final volume, 2 ml) were added to a sealed spectrophotometer cell containing excess sodium dithionite (5 mg) with careful elimination of air by a continuous argon flow. The reduced heme solution was stirred continuously in the spectrophotometer cell holder, which was kept at 20 ± 0.3°C. The reaction was initiated by rapid addition of the test peroxide (final concentration, 10 μM) while continually monitoring the absorbance of the heme Soret band (418 nm). Low reactant concentrations allowed direct analysis by visible spectrophotometry and HPLC-mass spectrometry (LCMS), while maintaining near-ideal solution kinetics. Excess dithionite was required to completely remove dissolved oxygen, thus retaining heme entirely in the reduced Fe(II) state and preventing oxidative degradation of heme (2). All reactions were conducted in triplicate.

Visible spectrophotometry.

Spectra were measured using a Cary 300 spectrophotometer (Varian, Mulgrave, Victoria, Australia) with a 10-mm special optical glass cell (Starna, Baulkham Hills, NSW, Australia). Measurements of heme absorbance were taken at 0.1-s intervals at the Soret wavelength of 418 nm, and the heme concentration was then determined from a linear calibration plot (ɛ = 1.26 × 10⁵ M⁻¹ cm⁻¹). The extent of heme loss was measured at the completion of each reaction (t = 30 s); product formation was also monitored as A₄₇₂ and normalized according to the initial heme absorbance (418 nm) for each reaction. Full spectra of reduced heme and the final product were obtained at a 1-nm resolution from 350 to 600 nm.

LCMS analysis.

Trioxolane concentrations were determined by LCMS as previously described (4), using a 50- by 2-mm, 5-μm Phenomenex Luna C₈ reversed-phase column. Test solutions of trioxolanes with either reduced Fe(II) heme, Fe(III) hemin, or sodium dithionite were monitored for 15 h at 20°C and compared to control solutions in 50% ACN-water.

RESULTS

Peroxide reactivity with heme.

Each of the trioxolane antimalarials (Fig. 1 and 2 and Table 1) underwent complete and rapid degradation by Fe(II) heme under reducing conditions with excess dithionite in 50% ACN-water. Analysis of each reaction mixture by LCMS did not detect any remaining trioxolane after 5 min (the first practical time point). The necessity for heme to be in the reduced state was confirmed by repeating the reaction for each trioxolane in the absence of the reducing agent, sodium dithionite. No trioxolane degradation was observed by LCMS after 15 h at 20°C with Fe(III) hemin alone. The peroxides were also found to be stable in ACN-water in the presence of excess sodium dithionite at 20°C over 15 h.

FIG. 2.

Structures of spiro and dispiro-1,2,4-trioxolanes. Refer to Table 1 for definitions of R, X, and R′.

TABLE 1.

Heme alkylation following reaction of peroxide antimalarials with Fe(II) heme^a

Peroxideb	R, X	Loss of heme (%)	Adduct absorbancec	IC50 (nM)d
Trioxolanes
1		83	0.89	1.8
2	R = H, X = H	86	0.84	3.8
5	R = CH2NH3+e, X = H	83	0.87	1.0
6	R = X = O	81	0.88	1.7
7	R = H, X = CH2OH	83	0.78	2.6
8	R = CH2OH, X = H	83	0.89	2.8
9	R = C3H7, X = H	85	0.80	3.6
10	R = CH3, X = CH3	82	0.77	7.9
11	R = C6H4Cl, X = H	84	0.84	20
12	R = C6H5, X = H	77	0.79	6.5
13	R′ = CH3	67	0.63	10
14		61	0.57	20
15	R = C(CH3)3, X = H	34	0.25	200
16	R = C(CH3)2OH, X = H	30	0.23	210
17		17	0.03	470
18	R′ = H	16	0.13	1500
19	R = C6H5, X = C6H5	11	0.10	100
3		9	0.03	340
20	R = COOH, X = H	74	0.89	49
21	R = CH2COOH, X = H	83	0.87	110
22	R = CH2SO4Na, X = H	81	0.85	250
Other peroxides
Artesunate		44	0.39	3.2
Dihydroartemisinin		28	0.22	0.77
Artemether		24	0.18	2.5
Artemisinin		15	0.12	5.7

^aPeroxide antimalarials (10 μM) were reacted with 10 μM Fe(II) heme in 50% ACN-H₂O with excess sodium dithionite under argon at 20°C (n = 3; standard deviation, <5%).

^bRefer to Fig. 1 and 2 for peroxide structures.

^cProduct A₄₇₂ (A₄₆₇ for artemisinin derivatives) at completion of reaction.

^dIn vitro activity against chloroquine-resistant (K1) Plasmodium falciparum (4, 6, 7, 30-33) converted to nM concentrations.

^eMesylate salt.

Heme reaction profile.

The kinetic properties of the reactions were further investigated by monitoring the decrease in heme concentration over a much shorter time frame using visible spectrophotometry. The reaction of the biologically active trioxolane 2 with Fe(II) heme resulted in a rapid decrease in the absorbance of the heme Soret band (418 nm; Fig. 3). The reaction was complete within 10 s, with no further change in absorbance, and resulted in a total of 86% decrease in the initial heme concentration (Table 1). A similarly rapid reaction was observed for the biologically inactive trioxolane 3, although the extent of heme loss (9%) was much lower (Fig. 3). Confirmation of complete trioxolane degradation by LCMS indicated the presence of a parallel heme-mediated trioxolane degradation pathway for trioxolane 3 that did not involve alteration of heme itself. All trioxolanes in this study, except trioxolane 4, exhibited similar fast reaction kinetics, with reactions being essentially complete within 20 s, as indicated by a plateau in the Soret absorbance. A much slower reaction profile was observed for trioxolane 4 (Fig. 3), which took over 4 min to complete the reaction to 60% heme loss.

FIG. 3.

Kinetic profiles of the percentage of heme remaining versus time for trioxolanes 2 (solid line), 3 (dashed line), and 4 (dotted line) in 50% ACN-H₂O at 20°C in the presence of excess sodium dithionite under argon.

Product formation.

Under reducing conditions, the major heme degradation products formed by all of the trioxolanes utilized in this study exhibited an absorption peak at 472 nm (Fig. 4). The observed bathochromic shift in the spectrum is likely to result from alkylation of the porphyrin ring of heme (13, 21).

FIG. 4.

Visible spectra of unreacted Fe(II) heme (dotted line) and reaction product (solid line) from reaction of heme (10 μM) with trioxolane 12 (10 μM) in 50% ACN-water with excess dithionite under argon.

HPLC-tandem MS analysis of the reaction mixture allowed separation from sodium dithionite and revealed the oxidized Fe(III) form of the heme adduct with a λ_max of 430 nm and m/z 782.3, which underwent collision-induced fragmentation to give a product ion at m/z 616.2 (heme). This suggested the presence of an alkylated heme adduct (trioxolane 23; Fig. 5) resulting from addition of the adamantane-derived radical (mass, 167.1) known to form during Fe(II)-mediated trioxolane degradation (28). More detailed structural characterization of the reaction product was not undertaken.

FIG. 5.

Proposed reaction mechanism for alkylation of Fe(II) heme by trioxolanes. For clarity, only alkylation at the β meso position is shown for the heme adduct 23 (m/z 782.3); the product is likely to be a mixture of isomeric adducts (21).

The reaction of the artemisinin derivatives with reduced heme also resulted in an absorption peak at 467 nm, in agreement with previous reports under similar conditions (17). Robert et al. (21) have previously shown that heme alkylation occurs by addition of the artemisinin-derived primary carbon-centered radical to any one of the four meso positions on the porphyrin ring. A similar mechanism is expected for addition of the trioxolane-derived secondary carbon-centered radical from these trioxolanes.

Relationship between heme reactivity and biological activity.

A number of trioxolanes with known in vitro antimalarial activity against Plasmodium falciparum (4, 6, 7, 30-33) were tested for their ability to alkylate Fe(II) heme. Greater than 80% loss of heme was observed for the highly biologically active trioxolanes 1, 2, and 5 to 10 (Table 1). The kinetic profiles and heme reactivity rates induced by each these trioxolanes were similar to those seen with trioxolane 2 (Fig. 3). The other potent biologically active trioxolane, 12, exhibited a marginally slower reaction rate; however, this still resulted in a total of 77% loss of heme within 20 s.

The less biologically active trioxolanes, 3 and 15 to 19, also reacted completely within 5 to 20 s, but this resulted in less than 40% loss of heme (Table 1), indicating that heme alkylation is not the preferred reaction pathway for these less active trioxolanes.

A strong relationship can be seen between the in vitro antimalarial activity and the observed extent of heme alkylation by the trioxolanes used in this study (Fig. 6). The most potent antimalarial trioxolanes appeared to efficiently alkylate heme, while the trioxolanes capable of alkylating only small amounts of heme exhibited poor biological activity in vitro. A correlation (R² = 0.86; Fig. 6A) was obtained between the antimalarial activity (log₁₀ 50% inhibitory concentration [IC₅₀]; nM) and the observed loss of heme for the neutral and basic trioxolanes in this study. A similar correlation was also obtained if the extent of heme alkylation was measured by relative product A₄₇₂ at the completion of the reaction (R² = 0.88; Fig. 6B). Attempts to relate the rate of heme alkylation to antimalarial activity were precluded by the inability to accurately determine reaction rates, especially when there was minimal heme alkylation for the poorly active compounds.

FIG. 6.

Relationship between heme alkylation and in vitro antimalarial activity against P. falciparum (4, 6, 7, 30-33) for neutral and basic trioxolanes (filled circles), acidic and anionic trioxolanes (open circles), and artemisinin derivatives (open triangles). (A) Relative loss of heme. (B) Heme adduct A₄₇₂ (or A₄₆₇ for the artemisinin derivatives). The solid lines represent the regression lines for the neutral and basic trioxolanes.

Interestingly, the negatively charged trioxolanes 20 to 22 with low antimalarial activity appeared to alkylate heme to an extent comparable to that of the highly active trioxolanes. It is likely that biological interactions unrelated to heme reactivity may limit the antimalarial activity of these trioxolanes, and this may be linked to the physicochemical properties of the acidic and anionic functional groups in trioxolanes 20 to 22.

Reactions of Fe(II) heme with artemisinin and the clinically utilized derivatives DHA, artemether, and sodium artesunate indicated that the artemisinin derivatives mediated much less heme alkylation than the equipotent trioxolanes under the conditions used in this study (Table 1). The reactions were also very rapid, and complete degradation of each artemisinin derivative occurred within 5 s. The stability of each artemisinin derivative was confirmed in 50% ACN-H₂O controls with excess sodium dithionite and also in Fe(III) hemin solution (neutralized with HCl) for 15 h at 20°C.

DISCUSSION

All trioxolanes used in this study reacted rapidly with heme, the most abundant source of iron in the malaria parasite (22). Reactivity with heme is thought to be an important process in the antimalarial mechanism of action for peroxide antimalarials, due to the formation of potentially toxic carbon-centered radicals. Peroxide cleavage and radical formation from these trioxolanes have been shown previously with inorganic Fe(II) iron salts (4, 28), and the present study indicates that the peroxide bond also reacts with the Fe(II) iron center of the biologically relevant heme molecule. The reactions with heme appeared much less sensitive to trioxolane structural variation than was seen with Fe(II) salts (e.g., FeSO₄), and occurred much faster than the inorganic reactions, taking seconds rather than hours to reach completion. This increased reaction rate may be attributable to the lower reduction potential of heme (E₀ ≈ −0.2 V) (23) relative to the inorganic hexa-aqua iron complex (E₀ = +0.77 V) (16); the presence of excess dithionite is also likely to contribute to the increase in reaction rate due to its coordinating and reducing properties (5). The increased reactivity was most noticeable for trioxolanes 15 and 16, which showed no measurable degradation with FeSO₄ (4) but were immediately and completely degraded by Fe(II) heme. Trioxolane 4 was also completely degraded by heme, although at a much slower rate, even though degradation of this peroxide by inorganic iron was prevented by steric shielding attributed to the two proximal adamantane substituents. This steric hindrance around the peroxide bond may explain the slower reaction with heme compared to the other more sterically accessible peroxides used in this study (Fig. 3). This slow reactivity with heme may also be the basis for the lack of antimalarial activity of trioxolane 4.

The extent of heme alkylation was highly correlated to the in vitro antimalarial activity of the trioxolanes used in this study, except for those containing acidic or anionic functional groups. The more potent trioxolane antimalarials (IC₅₀ < 20 nM) all resulted in significant heme alkylation (>60%). Furthermore, trioxolanes that alkylated heme less efficiently (<40%) displayed poor antimalarial activity (IC₅₀ > 100 nM). The correlation of in vitro antimalarial activity with loss of heme and the closely related correlation with heme adduct formation suggest that heme alkylation may be related to the mechanism of action for these trioxolane antimalarials.

The alkylated heme adduct produced from this reaction appears to result from substitution of a trioxolane-derived carbon-centered secondary radical onto the porphyrin ring. Formation of this trioxolane-derived radical has been previously demonstrated by radical-trapping experiments with inorganic iron (FeSO₄ and FeBr₂) (4, 28), and it is expected that the same radical forms after Fe(II) heme-mediated cleavage of the peroxide bond. This carbon-centered radical forms in close proximity to the porphyrin macrocycle, and substitution is likely to occur on the porphyrin ring, giving a mixture of isomeric heme adducts analogous to the artemisinin-derived adducts previously reported (21).

Exact quantitation of the alkylated heme product(s) was not possible in this system due to the absence of a pure standard for calibration. Kannan et al. (12) have reported that heme alkylation by artemisinin only marginally decreased the molar extinction coefficient of heme (27% decrease when the solvent was methanol), and it is expected that alkylation by trioxolanes would have a similar effect on absorptivity. The visible absorbance of the alkylated heme product (472 nm) from each active trioxolane in this study was equivalent to 79% (± 5%) of the decrease in heme absorbance (418 nm), suggesting highly efficient formation of the alkylated adduct.

The alkylated heme adduct may be an important intermediate responsible for the biological activity of peroxide antimalarials. Heme adducts may inhibit hemozoin formation, either by inhibition of crystal growth (25) or by inhibition of enzymes such as Plasmodium falciparum histidine-rich protein II (13). Additionally, heme adducts may cause membrane damage or competitively inhibit other heme-interacting proteins within the parasite (3), or trioxolanes may directly alkylate intraparasitic hemoproteins (20).

An alternative hypothesis is that efficient heme alkylation by active trioxolanes may be a marker of the potential for these trioxolanes to undergo the required sequence of iron-mediated peroxide cleavage, free radical formation, and alkylation of a biological target in close proximity to the generated radical. Conversely, the high yield of heme adduct resulting from active trioxolanes may suggest that the generated free radical reacts immediately with the porphyrin ring.

The heme-mediated reactivity of all peroxides in this study suggests that the low heme alkylation observed for the poorly active trioxolanes resulted from a reduced tendency of the carbon-centered radical intermediates to react directly with the porphyrin ring. The primary carbon-centered radical formed from trioxolane 17 is expected to have lower radical stability, and it is possible that the radicals derived from trioxolanes 3 and 17 (4) may escape the heme due to greater flexibility compared to the adamantane-derived radicals. The structure and orientation of the intermediate complex formed between the trioxolane and heme are likely to promote heme alkylation if the carbon-centered radical is produced in close proximity to the porphyrin ring, allowing immediate alkylation of the heme. The bulky 8′-substituents on trioxolanes 15 and 16 may inhibit the required close interaction with heme and may also decrease the preference for Fe(II) attack at the peroxide oxygen adjacent to the cyclohexane ring leading to greater formation of the cyclohexane-derived primary radical associated with lower antimalarial activity (4). It appears that trioxolanes containing a spiroadamantane moiety, but without additional steric bulk around the peroxide bond, are generally efficient heme-alkylating compounds; these attributes have previously been linked to the biological activity of these trioxolanes (6).

The link between antimalarial activity and heme alkylation was evident for neutral and basic trioxolanes in this series; however, the acidic and anionic trioxolanes did not follow the same trend. The efficient heme alkylation seen for trioxolanes 20 to 22 does not correlate with their poor in vitro antimalarial activity, suggesting that other processes are also important for the biological activity of these antimalarials. Trioxolanes 20 to 22 are predominantly negatively charged at physiological pH, which may inhibit uptake across the erythrocyte (4), parasite, and vacuolar membranes. Furthermore, accumulation of trioxolane 20 or 21 is unlikely to occur within the heme-rich food vacuole, which at pH 5.2 (34) affords partial protonation of these weak acids, promoting diffusion toward the cytosol, where the ionized form predominates due to the higher pH. It is also possible that the charge on these less-active trioxolanes inhibits interaction with heme associated with intraparasitic membranes or proteins, which may be more mechanistically relevant than the free heme used in this study.

The relatively low level of heme alkylation observed for the highly active artemisinin derivatives in this study appears to differentiate these two pharmacophorically similar but structurally distinct groups of peroxide antimalarials (11). While extensive heme alkylation may be a determinant of trioxolane activity, it is possible that artemisinin activity is dependent on interactions other than, or in addition to, heme alkylation, such as interactions with specific parasite proteins: e.g., the SERCA orthologue PfATP6, which shows 100-fold-greater inhibition by artemisinin than by trioxolane 1 (29).

It is noted that Robert et al. (21) have reported a much higher yield of heme alkylation by artemisinin (85%) using very different reaction conditions compared to the present study. However, the conditions used in Robert's study (aimed at maximizing adduct formation) utilized 3,000-fold-higher reactant concentrations, a much lower excess of a different reducing agent (glutathione), and dimethyl sulfoxide as the solvent. We have shown previously that reaction conditions significantly affect the rate and product distribution for iron-mediated artemisinin degradation (5). It is likely that the conditions employed for the present study may offer more efficient competing quenching mechanisms for the heme-generated radical, allowing greater differentiation between test peroxides with respect to the extent of heme alkylation. This partially aqueous system with low reactant concentrations allowed for direct analysis of the reaction kinetics, and the inclusion of excess sodium dithionite was essential to ensure complete heme reduction to allow a controlled comparison for each peroxide tested.

The conditions used in this study reflect a highly simplified system compared to the reaction with biological heme within the parasite. It is expected that a majority of the intraparasitic heme exists in the Fe(III) state, which would be unreactive with the peroxides, as demonstrated in this study. However, a small percentage of the high intraparasitic heme concentration (400 mM) (22) is likely to exist in the Fe(II) state due to reduction by biological reducing agents such as glutathione or complexation with proteins such as hemoglobin; this may be sufficient to allow for rapid heme alkylation by nanomolar concentrations of trioxolanes.

Conclusions.

The reaction of antimalarial trioxolanes with reduced heme induced complete and rapid peroxide degradation and resulted in conversion of heme to an alkylated heme adduct. The heme adduct appears to result from addition of the trioxolane-derived secondary carbon-centered radical formed following iron-mediated reduction of the peroxide bond. The most potent trioxolane antimalarials were capable of highly efficient heme alkylation, and a correlation was observed between the extent of heme alkylation and in vitro antimalarial activity. This link indicates that heme alkylation may be an important process involved in the mechanism of action of these peroxide antimalarials, and inefficient heme alkylation may help to explain the poor antimalarial activity of a number of fully synthetic peroxides. Interestingly, the highly active artemisinin derivatives mediated a relatively low level of heme alkylation in this study, suggesting that the mechanism of artemisinin action may be less dependent on heme alkylation and may involve other parasite interactions specific to the more topologically complex sesquiterpene structure of artemisinin.