Evidence for the Involvement of Carbon-centered Radicals in the Induction of Apoptotic Cell Death by Artemisinin Compounds*

Artemisinin and its derivatives are currently recommended as first-line antimalarials in regions where Plasmodium falciparum is resistant to traditional drugs. The cytotoxic activity of these endoperoxides toward rapidly dividing human carcinoma cells and cell lines has been reported, and it is hypothesized that activation of the endoperoxide bridge by an iron(II) species, to form C-centered radicals, is essential for cytotoxicity. The studies described here have utilized artemisinin derivatives, dihydroartemisinin, 10β-(p-bromophenoxy)dihydroartemisinin, and 10β-(p-fluorophenoxy)dihydroartemisinin, to determine the chemistry of endoperoxide bridge activation to reactive intermediates responsible for initiating cell death and to elucidate the molecular mechanism of cell death. These studies have demonstrated the selective cytotoxic activity of the endoperoxides toward leukemia cell lines (HL-60 and Jurkat) over quiescent peripheral blood mononuclear cells. Deoxy-10β-(p-fluorophenoxy)dihydroartemisinin, which lacks the endoperoxide bridge, was 50- and 130-fold less active in HL-60 and Jurkat cells, respectively, confirming the importance of this functional group for cytotoxicity. We have shown that chemical activation is responsible for cytotoxicity by using liquid chromatography-mass spectrometry analysis to monitor endoperoxide activation by measurement of a stable rearrangement product of endoperoxide-derived radicals, which was formed in sensitive HL-60 cells but not in insensitive peripheral blood mononuclear cells. In HL-60 cells the endoperoxides induce caspase-dependent apoptotic cell death characterized by concentration- and time-dependent mitochondrial membrane depolarization, activation of caspases-3 and -7, sub-G0/G1 DNA formation, and attenuation by benzyloxycarbonyl-VAD-fluoromethyl ketone, a caspase inhibitor. Overall, these results indicate that endoperoxide-induced cell death is a consequence of activation of the endoperoxide bridge to radical species, which triggers caspase-dependent apoptosis.

Artemisinin (ART, 2 1; see Fig. 1) is a sesquiterpene lactone endoperoxide found in the traditional Chinese medicinal plant Artemisia annua (1). The activity against multidrug-resistant malaria parasites and a rapid therapeutic response (2) reported for ART and semi-synthetic ART compounds combined with an absence of significant toxicity in patients (3,4) have led the World Health Organization to recommend the use of ARTbased combination therapies to all countries experiencing resistance to conventional monotherapies.
The cytotoxic properties of ART compounds in cancer cell lines were first observed against Ehrlich ascites tumor cells (5) and have subsequently been reported in many other human cell lines of the 60 cell line test panel at the NCI, National Institutes of Health (6 -12). ART and its derivatives exhibit selective cytotoxicity toward rapidly proliferating neoplastic cells, with the highest activity reported against colon and leukemia cell lines and the least against non-small cell lung cancer cells (10). The most sensitive cell lines are characterized by their rapid proliferation, often accompanied by a high intracellular iron concentration to sustain continued proliferation (13,14). It has been hypothesized that iron activation of the endoperoxide bridge is an essential event in cytotoxicity; compounds without the endoperoxide bridge do not display cytotoxicity (6,8); and the addition or chelation of iron can potentiate or inhibit cytotoxicity, respectively (15)(16)(17)(18). However, the intracellular chemistry of the endoperoxides and its link with the induction of cell death have yet to be defined.
As an antimalarial it is widely held that iron-mediated, oneelectron reduction of the endoperoxide bridge produces C-centered parasiticidal radical species and that heme is the probable source of ferrous iron (19 -21). The chemistry of endoperoxide bridge degradation has been studied extensively, using Fe(II) salts and EPR spectroscopy, to provide a framework of well defined free radical pathways ( Fig. 1) (19,20,22,23). Association of Fe(II) with oxygen 1 or 2 of the peroxide bridge results in two oxyl radicals (1a and 1d) that can rearrange to a primary or secondary C-centered radical (1b and 1e), respectively. Rearrangement of these radicals produces stable end product as follows: a ring-contracted tetrahydrofuran acetate (THF acetate, 1c) and a hydroxydeoxo isomer (1f), which have been used as surrogate markers of radical formation in vitro and in vivo (22)(23)(24)(25)(26). Although the Fe(II) chemistry has been defined, the mechanism of biological action remains unclear (27)(28)(29).
Most commonly, cell death proceeds via one of two different mechanisms, apoptosis or necrosis (30). Necrosis is defined as passive, poorly regulated cell death characterized by cellular swelling, a loss of membrane integrity, and the disruption of neighboring tissue (31,32). Apoptosis is an active, internal cell death program that can be distinguished by its characteristic morphology: the maintenance of membrane integrity, the formation of membrane blebs with decreasing cell volume, nuclear chromatin condensation, DNA degradation, and ultimately the formation of apoptotic bodies that are engulfed by phagocytosis in vivo (33). Upon the induction of apoptosis, a conserved family of cysteine aspartic acid-specific proteases, known as caspases, is activated in a cascade mechanism (34,35). The caspases comprise the following two functional groups: the initiator caspases, -2, -8, -9, and -10, activated at the onset of apoptosis; and the effector caspases, -3, -6, -7, which, when activated, target a subset of cellular proteins for proteolysis leading to cell death (36).
It has been demonstrated that endoperoxide-mediated cytotoxicity in cancer cells proceeds via the induction of caspase-depend-ent apoptosis, but the chemical basis of this induction has not been defined (13,(37)(38)(39). We have therefore investigated the cellular metabolism of the endoperoxides to elucidate the chemical species that lead to cell death and the molecular pathways initiated, using a combination of chemoanalytical and biomolecular techniques. Dihydroartemisinin (DHA, 2) and two synthetic analogues, 10␤-(pbromophenoxy)dihydroartemisinin (PBrDHA, 3) and 10␤-(p-fluorophenoxy)dihydroartemisinin (PFDHA, 4), which have increased metabolic stability at the acetal function (23), and the deoxygenated counterpart of PFDHA, deoxy-10␤-(p-fluorophenoxy)dihydroartemisinin (dPFDHA, 5), which lacks the endoperoxide bridge, were selected as investigative tools to define the mechanism of cell death induced in two leukemia cell lines, HL-60 and Jurkat, and normal freshly isolated human peripheral mononuclear cells (PBMC), which were included as control monocytic cells. HL-60 cells can terminally differentiate into monocytes (40).

EXPERIMENTAL PROCEDURES
Materials-Fetal bovine serum was from BioWhittaker Europe (Verviers, Belgium), and human AB serum was from Sera Laboratories (Milton Keynes, Buckinghamshire, UK). Lymphoprep was purchased from Nycomed (Birmingham, UK). The human cell lines HL-60 and Jurkat were obtained from the European Collection of Cell Cultures (Salisbury, Wiltshire, UK). The cytotoxicity detection kit (LDH) was purchased from Roche Applied Science. Polyacrylamide gel was purchased from National Diagnostics (Hessle, Yorkshire, UK). Caspase-3 and caspase-7 rabbit polyclonal antibodies were raised as described previously (41). Goat anti-rabbit IgG secondary antibody conjugated to horseradish peroxidase was purchased from DakoCytomation (Ely, Cambridgeshire, UK). Bradford reagent was purchased from Bio-Rad. Protease inhibitor mixture was purchased from Roche Diagnostics. All other materials and chemicals were purchased from Sigma.
Cell Culture and the Isolation of PBMC-HL-60 and Jurkat cell lines were maintained in RPMI 1640 medium supplemented with fetal bovine serum (10% v/v) and L-glutamine (1% w/v). The cells were incubated under humidified air containing 5% CO 2 at 37°C. Cell density was kept below 1 ϫ 10 6 cells/ml to ensure exponential growth and to avoid differentiation of HL-60 cells, which were only used between passages 5 and 15. Freshly drawn venous blood from healthy volunteers was collected in heparinized tubes, and PBMC were isolated as described previously (42). PBMC were then diluted to the required density in RPMI 1640 medium supplemented with AB human serum (10% v/v), penicillin/streptomycin solution (1% v/v), L-glutamine (1% w/v), and HEPES (1 M, 2.5% v/v). Cell viability was above 95% for all experiments based on trypan blue exclusion (43). Drug stock solutions were made up in Me 2 SO, and the final solvent concentration was below 0.5% v/v in each incubation.
Measurement of Cytotoxicity Using the 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium Bromide (MTT) Assay and LDH Release-HL-60 cells (2.5 ϫ 10 4 /well) and Jurkat cells (1 ϫ 10 5 /well) were plated, in triplicate, in flat-bottom 96-well plates and were exposed to 0.01 to 100 M of each compound for 72 h (5% CO 2 at 37°C). Following incubation, cell viability measurements using the MTT assay were carried out by the addition of 20 l of MTT solution (5 mg/ml in HBSS) to each well and incubating for 2 h at 37°C. Thereafter, 100 l of a lysing solution (20% w/v sodium dodecyl sulfate; 50% v/v N,N-dimethylformamide) was added to each well, to dissolve the formazan crystals, and incubated for an additional 4 h. The absorbances of the samples were measured at a test wavelength of 570 nm and a reference wavelength of 590 nm with a plate reader (MRX, Dynatech Laboratories) lactate dehydrogenase (LDH) release was measured using the cytotoxicity detection kit (LDH) according to the manufacturer's instructions. All results are expressed as a percentage of vehicle-only cells. The IC 50 values were calculated from individual inhibition curves plotted by Grafit software.
Synthesis of Chemical Markers of Bioactivation-The degradation of PBrDHA (3) by Fe(II) was carried out according to a novel method optimized for solvent composition and iron salt (44). Ferrous gluconate (0.611 g, 1.36 mmol, 2 eq) was added to a stirred solution of PBrDHA (0.300 g, 0.68 mmol) in N,N-dimethylformamide (5 ml) and distilled water (5 ml) under a nitrogen atmosphere at room temperature. The reaction mixture was stirred for 7 h before dilution with CH 2 Cl 2 (10 ml) followed by extraction with CH 2 Cl 2 (three times, 20 ml). The combined organic extracts were dried over anhydrous magnesium sulfate and concentrated in vacuo. Flash chromatography using 10% ethyl acetate and 90% n-hexane as the eluent afforded the THF acetate isomer (6, 0.162 g, 54%) and recovered starting material (3, 0.038 g, 13%).
Data for 6 are as follows: 1  LC-MS Analysis and Quantification of Intracellular Endoperoxide Bioactivation-Intracellular iron-related activation of the endoperoxides was monitored using LC-MS to analyze cell extracts prepared from drug-cell incubations. Either PBMC or HL-60 cells (100 ml of 1 ϫ 10 6 cells/ml) were incubated with PBrDHA (10.0 M, 24 h, 37°C) in the appropriate cell culture medium. Following incubation, the cells were extracted with CHCl 3 (four times, 60 ml). The extracts were combined and dried over anhydrous magnesium sulfate. The magnesium sulfate and cell debris were removed by filtration through a scintered glass funnel, and the solvent was removed in vacuo. The residue was dissolved in CH 3 OH (100 l) immediately before analysis by LC-MS. Aliquots were eluted from a Symmetry 5-m C 8 column (150 ϫ 3.9 mm; Waters, Milford, MA) with acetonitrile (75%) in 10 mM ammonium acetate at a flow rate of 0.9 ml/min. Eluate splitflow to the LC-MS interface was ϳ40 l/min. Positive-ion electrospray mass spectra were acquired between m/z 100 and 1050, over a scan time of 5 s, with a Quattro II mass spectrometer (Waters Corp, Manchester, UK). The source temperature was 80°C. The cone and capillary voltages were 30 V and 3.8 kV, respectively. All data were processed with Mass-Lynx 3.5. The amounts of unchanged PBrDHA and PBrDHA THF acetate extracted from incubations of PBrDHA with HL-60 cells were quantified by selected ion monitoring of m/z 458 ([M ϩ 1] ϩ ). The dwell time and cycle time were 200 and 231 ms, respectively. Calibration curves of peak area versus analyte mass (2.5-50 nmol) were generated from solutions of synthetic PBrDHA and PBrDHA THF acetate in acetonitrile, and the limit of detection was calculated to be 1.0 nmol. The efficiency of PBrDHA and PBrDHA THF acetate recovery was calculated by the immediate extraction of authentic standard compounds (1 mol) from HL-60 cells.
Flow Cytometric Analysis of Sub-G 0 /G 1 Population-Propidium iodide (PI) was used to quantify cellular DNA content to measure the formation of a sub-G 0 /G 1 population of HL-60 cells. Following 24 h of exposure to each compound, 5 ϫ 10 5 treated cells were washed twice in HBSS, fixed in 1 ml of icecold 70% ethanol, and frozen at Ϫ20°C. After 2 h, the 70% ethanol was removed, and the cell pellet was resuspended 1 ml of PI staining solution (phosphate-buffered saline containing 40 g/ml PI, 0.1 mg/ml RNase, and 3.8 mM sodium citrate) and  Flow Cytometric Analysis of Mitochondrial Depolarization-Mitochondrial membrane potential (MMP) was measured using tetramethylrhodamine ethyl ester (TMRE) to quantify HL-60 cells with a high MMP. Drug-treated and untreated cells (5 ϫ 10 5 ) were washed in HBSS, and the resultant cell pellet was resuspended in 500 l of TMRE solution (50 nM in HBSS) and incubated for 30 min at 37°C. A minimum of 5 ϫ 10 3 cells were analyzed by flow cytometry (Coulter Epics, XL Software), and TMRE fluorescence was measured in fluorescence channel FL-2.
Western Blot Analysis of Caspase-3 and -7 Processing-Cells were harvested at the indicated times, washed once in cold phosphate-buffered saline, and stored at Ϫ20°C prior to use. Cell lysates were prepared by resuspending frozen cell pellets in cold phosphatebuffered saline containing protease inhibitor mixture. Cell suspensions were sonicated briefly before the protein concentration was measured by the Bradford assay (45). Lysates, with equal amounts of protein (20 g per lane), were mixed with SDS-PAGE loading buffer and denatured at 95°C for 3 min prior to being resolved on 14% SDS-PAGE. Proteins were transferred to nitrocellulose membrane for Western blotting analysis as described previously (41).
Electron Microscopic Examination of Cell Morphology-Electron microscopy was used to examine ultrastructural changes in HL-60 cells following incubation with DHA (10 M) for 24 h at 37°C. Cells were prepared and examined as described previously (46).
Statistical Analysis-Values are expressed as a mean Ϯ S.D. Data were analyzed for non-normality using a Shapiro-Wilk test. Student's t test was used when normality was indicated. A Mann-Whitney U test was used for nonparametric data. All calculations were performed using Arcus Quickstat statistical software; results were considered to be significant when twosided p values were less than 0.05.

RESULTS
The Endoperoxides Exhibit Significant Cytotoxicity in Human Cell Lines, Which Is Dependent Upon the Presence of the   MARCH 30, 2007 • VOLUME 282 • NUMBER 13

JOURNAL OF BIOLOGICAL CHEMISTRY 9375
Endoperoxide Bridge-The cytotoxic activities of DHA, PBrDHA, and PFDHA in HL-60 cells, Jurkat cells, and PBMC were assessed by examining their effects upon cellular dehydrogenase activity using the MTT assay, and also by measuring LDH leakage from the cytosol of damaged cells into the supernatant in HL-60 cells ( Table 1). The novel endoperoxides PBrDHA and PFDHA displayed high levels of cytotoxicity as follows: IC 50 values Յ0.5 M toward both HL-60 and Jurkat cell lines. They were more cytotoxic than their parent compound DHA, the pharmacologically active metabolite of first-generation ether and ester derivatives of ART (47). In these tests the novel endoperoxides were not as cytotoxic as doxorubicin, a chemotherapeutic agent that has been included as a positive control for cytotoxic activity toward leukemia cells (48,49). Conversely, quiescent PBMC remained resistant to the endoperoxides. No toxicity was seen at concentrations up to 250 M, and PBMC were at least 500fold less susceptible to the endoperoxides, thus demonstrating the selective cytotoxic properties of endoperoxide compounds against rapidly dividing cell lines. An IC 50 value for doxorubicin in PBMC could not be determined because of large inter-individual variations in IC 50 values, possibly arising from polymorphic expression of the efflux pump p-glycoprotein in PBMC (50), of which doxorubicin is known to be a substrate (51). The deoxygenated counterpart of PFDHA, dPFDHA, which has an ether linkage in place of the endoperoxide bridge, displayed low levels of cytotoxicity; IC 50 values were 50-and 130-fold higher in HL-60 and Jurkat cells, respectively, than for the parent compound, thus confirming that high levels of cytotoxic activity are dependent upon the presence of the endoperoxide bridge. Both measures of cytotoxicity, cellular dehydrogenase activity and LDH leakage, produced comparable IC 50 values for all compounds in HL-60 cells. In sensitive HL-60 cells, PBrDHA and PFDHA exerted a time-dependent cytotoxicity, which was first apparent after 16 h but increased markedly by 24 h (Table 2), and these cells were used in subsequent investigations.
The THF Acetate Isomer Can Be Used as a Cellular Biomarker of Endoperoxide Bridge Activation-The cellular bioactivation of the endoperoxide bridge was investigated using PBrDHA in susceptible HL-60 cells and resistant PBMC. A time course of endoperoxide bridge activation, via C-centered radical formation, in HL-60 cells was followed by LC-MS analysis of chloroform extracts containing recovered endoperoxide-derived compounds. A time-dependent loss of parent compound was demonstrated over 24 h accompanied by the formation of the THF acetate isomer (6, Fig. 2), indicative of endoperoxide bridge activation via the primary carbon center radical (Fig. 1). The relevant mass chromatographic peaks were not evident in the extracts of cells incubated in medium alone (not shown), demonstrating that these measurements result from a specific PBrDHA-derived compound. Quantification of the parent compound and the THF acetate isomer, achieved using synthetic standards, revealed that following a 16-h incubation, 37 Ϯ 8% of the parent compound remained, dropping to 19 Ϯ 5% after 24 h, and 7 Ϯ 2% of the PBrDHA had undergone activation through the primary C-centered radical to form the isomer, allowing for a calculated 19% recovery efficiency. The same analysis was carried out following a 24-h incubation of resistant PBMC with PBrDHA in comparison with sensitive HL-60 cells (Fig. 3). In these studies only trace amounts of THF acetate isomer were detected, which were below the limit of quantification, although the unchanged parent compound, equaling 64 Ϯ 7% of incubated material, was recovered.
The Endoperoxides Induce Apoptosis in HL-60 Cells-The ability of the endoperoxides to induce apoptosis was assessed by flow cytometric measurement of mitochondrial membrane depolarization, Western blot analysis of caspase-3 and -7 activation, and DNA degradation, which are events associated with the apoptotic pathway. TMRE was used to label cells with high MMP (52), and any reduction in fluorescence was attributed to mitochondrial membrane depolarization (Fig. 4A). DHA, PBrDHA, and PFDHA (Fig. 4, B and C) induced time-and concentrationdependent depolarization. The endoperoxides initiated similar extents of depolarization, with significant levels reached at 0.5 M, increasing until the maximum effect was reached at 1.0 M after 24 h, when ϳ80% of cells were depolarized. PBMC did not undergo mitochondrial membrane depolarization following treatment with PFDHA (0.1-100 M, 24 h; data not shown). Flow cytometric analysis of cellular DNA content with PI staining (53) was used to measure the formation of a sub-G 0 /G 1 population of apoptotic cells with degraded DNA following endoperoxide treatment (Fig. 5A). DHA, PBrDHA, and PFDHA (Fig. 5, B and C) all induced similar degrees of time-and concentration-dependent formation of a sub-G 0 /G 1 population, which became significant from 16 h, reaching a maximum of ϳ60% of cells in the sub-G 0 /G 1 phase (1 M, 24 h). The deoxygenated counterpart of PFDHA, dPFDHA, did not induce mitochondrial membrane depolarization (Fig. 4C) or DNA fragmentation (Fig. 5C) in HL-60 cells, thus indicating that the peroxide bridge is essential for the induction of apoptotic cell death. The activation of caspases-3 and -7 in endoperoxide-treated HL-60 cells was examined by Western blotting techniques that visualized the processing of the inactive caspase proform to the catalytically active smaller units. In untreated HL-60 cells, caspase-3 was primarily present as the 32-kDa proform (Fig. 6,  A and B, lane 1). Following exposure to the endoperoxides, a concentration-dependent processing of caspase-3 was evident by the loss of the proform and appearance of processed caspase-3, primarily to its fully processed catalytically active p17 large subunit (Fig. 6A, lanes, 3, 4, 6, 7, 9, and 10). Similarly, the endoperoxides induced a time-dependent processing of caspase-3 to its fully processed p17 large subunit (Fig. 6B, lanes  4, 5, 8, 9, 12, and 13). In contrast, the inactive dPFDHA failed to induce processing of caspase-3 (Fig. 6, A, lanes 11-13, and B,  lanes 14 -17). Caspase-7 was also primarily present as the 35-kDa proform in untreated cells (Fig. 6, C and D, lane 1). PBrDHA, DHA, and PFDHA all caused a concentration-dependent processing of caspase-7 to its catalytically active p19 large subunit (Fig. 6C, lanes 3, 4, 6, 7, 9, and 10), whereas no processing of caspase-7 was observed with the inactive dPFDHA (Fig. 6C, lanes 11-13). Similarly, the active endoper- oxides, but not the inactive dPFDHA, also caused a time-dependent processing of caspase-7 to its active p19 large subunit (Fig. 6D).
Endoperoxides Induce Caspase-dependent Apoptosis in HL-60 Cells-A caspase inhibitor, Z-VAD-fmk, was used to examine the dependence of apoptosis upon activation of the caspase cascade in endoperoxide-treated HL-60 cells. In control HL-60 cells, the cytoplasm and nucleus appeared normal with only traces of heterochromatin lining the nuclear envelope (Fig. 7a). Following exposure to DHA or PFDHA, a wide range of morphologies was present, with few cells with normal morphology remaining (Fig. 7, b and c, and data not shown). Many cells contained ultrastructural changes characteristic of apoptosis, including pyknotic nuclear fragments or sharply defined dense crescents of condensed chromatin that abutted against the nuclear envelope. Some cells had a condensed cytoplasm, with vacuolation of the endoplasmic reticulum, but others were already undergoing secondary necrosis. These ultrastructural changes were largely prevented by Z-VAD-fmk, when most cells appeared normal but some showed a moderate condensation of their heterochromatin, which either abutted against the nuclear envelope or as clumps that extended throughout the nucleus (Fig. 7d). In cells exposed to DHA and PFDHA, in addi-tion to the characteristic ultrastructural changes of apoptotic cells, a small number of cells were observed that showed abnormalities not normally associated with apoptosis, and these were not influenced by Z-VAD-fmk (data not shown).
To further confirm that Z-VADfmk was inhibiting apoptotic changes, we examined its effects on the processing of the major effector caspase-3 induced by PFDHA and DHA. PFDHA and DHA, as well as etoposide used as a positive control, induced the processing of caspase-3 to its p17 catalytically active large subunit (Fig. 8, lanes 2, 4, and 6). Z-VAD-fmk inhibited the PFDHAand DHA-induced processing of caspase-3, resulting in the presence of more caspase-3 zymogen as well as a small amount of the p20 form of caspase-3. The latter most probably represents Z-VAD-fmk covalently bound to the initial p20 form of caspase-3 cleaved at Asp-175 between the large and small subunits and is catalytically inactive as evidenced by an inhibition of cleavage to the p19 and p17 forms. Furthermore, cells that had been preincubated with Z-VAD-fmk (1 h, 100 M) still underwent mitochondrial membrane depolarization following 24 h of treatment with DHA and PFDHA (1 M) (Fig. 9A). Conversely, the presence of Z-VADfmk (1 h, 100 M) inhibited DHA-and PFDHA-induced (1 M, 24 h) DNA fragmentation as assessed by PI staining of the sub-G 0 /G 1 population (Fig. 9B). These results indicate that mitochondrial membrane depolarization is not dependent upon caspase activity but that DNA degradation is.

DISCUSSION
ART (1) and its derivatives are an important class of antimalarial agents that are active against resistant strains of P. falciparum. These drugs are widely used, with over 100 million courses administered annually (54), despite the fact that embryotoxicity in rats and rabbits (55,56) and in vivo and in vitro neurotoxicity (57, 58) have been reported. Artemisinins can display cytotoxic activity in actively proliferating mammalian cells (5)(6)(7)(8)(9)(10)(11)(12), and it is suggested that toxicity is related to high intracellular iron concentrations (15,16,18). It is therefore essential that the chemical and molecular mechanisms of endoperoxide cytotoxicity are defined to assess the safe and effective use of this class of drugs in the established area of malaria and their potential use in the treatment of cancer. It was the aim of this study to identify the chemostructural and metabolic basis of selective endoperoxide cytotoxicity and the molecular mechanism of cell death in order to characterize the types of cells that are susceptible to the endoperoxides.
We have investigated DHA (2) because it is used clinically as an antimalarial and is the pharmacologically active metabolite of some first-generation endoperoxides (47). PBrDHA (3) and PFDHA (4) are synthetic derivatives with increased biological stability over the parent compound, DHA, because of incorporation of the metabolically stable phenoxy linkage (25). The present studies (Table 1) demonstrated that addition of the phenoxy group increased cytotoxic activity in HL-60 and Jurkat cells over the parent compound, perhaps due in part to the increased chemical stability of the phenoxy-C bond, which could minimize degradation before target biomolecules are reached. Importantly, we discovered that the endoperoxides are significantly more cytotoxic toward rapidly dividing HL-60 and Jurkat cells compared with quiescent PBMC, with IC 50 values at least 500-fold higher in PBMC (Table 1). This selectivity can be rationalized by the iron-activation hypothesis, as it is known that neoplastic cells are highly dependent upon iron to sustain their characteristic high levels of proliferation (14,59). It is interesting to note that susceptible HL-60 and Jurkat cells express high levels of transferrin receptors, whereas non-sensitive cell lines such as lymphocytes do not (60 -63). Transferrin receptors are membrane-bound non-heme iron-binding glycoproteins that control the major uptake of cellular iron and are overexpressed in proliferating cells (64).
Although reduction of MTT in mammalian cells is usually attributable to the activity of mitochondrial and cytoplasmic dehydrogenases (65), superoxide is known to reduce tetrazolium dyes in cells (66). Doxorubicin and ART compounds (13,48) are reported to induce production of reactive oxygen species in cells, but the extreme cytotoxicity of doxorubicin (Table   1) implies that any effect on tetrazolium reduction is insignificant. Moreover, measurement of LDH leakage, a superoxideindependent marker of cell death, has produced IC 50 results comparable with the MTT assay ( Table 1).
The importance of peroxide bridge activation for cytotoxic activity was demonstrated using an analogue of PFDHA, dPFDHA (5), in which the endoperoxide bridge has been replaced by an ether linkage. Without the endoperoxide bridge cytotoxicity levels were greatly diminished, by 50and 130-fold, in HL-60 and Jurkat cells, respectively. This hypothesis was further examined by probing the chemistry of peroxide bridge activation, via C-centered radical formation, in both sensitive HL-60 cells and resistant PBMC. PBrDHA was used in these studies as the presence of bromine provides a characteristic isotope pattern that facilitates LC-MS detection. In HL-60 cells LC-MS analysis revealed time-dependent loss of the parent compound accompanied by detection of the THF acetate isomer (6), which is regarded as a marker of endoperoxide activation via rearrangement of the primary C-centered radical (1b) (Fig. 2). By using synthetic standards, we quantified that ϳ7 Ϯ 2% of the PBrDHA had been activated via this radical pathway. Although these levels are low, this could be a result of intracellular reactions between the radical species and biomolecular targets that compete with the intracellular rearrangement, or it may be indicative of competing chemical pathways following peroxide bridge activation. The occurrence of such reactions with cellular targets is unconfirmed, but chemical studies have shown that ART endoperoxides can react with proteins, via thiol and amino groups (67), and can form adducts with cysteine and glutathione biomimetically (68 -71). Analysis of PBMC following 24 h of incubation with PBrDHA did not yield any quantifiable trace of the THF acetate isomer, but it did reveal that large amounts of the parent compound, 64 Ϯ 7% of the starting material, were recoverable from the cells (Fig. 3). This demonstrates that time-dependent bioactivation of the endoperoxide bridge occurs in susceptible HL-60 cells but not in resistant PBMC. By using these techniques we have provided the first chemical evidence that bioactiva-  tion via C-centered radicals may underlie the differential cytotoxicity between sensitive and resistant cells.
The attenuation by an iron chelator of the induction of apoptosis in HeLa cells exposed to DHA suggests that intracellular free iron plays a major role in the activation of ART endoperoxides (13). As discussed, it is widely accepted that it is liberated heme that is responsible for the activation of ART during its antimalarial action in red blood cells, and there is evidence to suggest heme-containing proteins, such as hemoglobin, have the same action (29). Both HL-60 cells and PBMC express the myeloperoxidase (72, 73), a heme-containing protein, but it is unlikely to contribute to the activation of the endoperoxides because myeloperoxidase catalyzes two-electron reductions (74), whereas peroxide activation occurs via one-electron reduction (29).
It was found that the endoperoxides induce concentrationand time-dependent apoptosis via mitochondrial membrane depolarization, caspase activation, and DNA degradation in sensitive HL-60 cells, even at concentrations up to 100 times greater than the IC 50 values (Figs. 4 -6, respectively). Furthermore, studies with dPFDHA revealed that the peroxide bridge is essential for the induction of apoptosis and implied that the low levels of dPFDHA cytotoxic activity (Table 1) proceed via the induction of necrotic cell death. The induction of apoptosis, a specific, controlled form of cell death, by the endoperoxides suggests that the cytotoxic mechanism of action induced by endoperoxide activation proceeds via a specific subcellular target.
A broad spectrum caspase inhibitor, Z-VAD-fmk (75), was used to probe the dependence of apoptosis upon caspase activity and to further probe the apoptotic mechanism. Apoptosis can be initiated by various stimulants and toxicants that results in the activation of different biochemical pathways, namely the intrinsic or chemical stressmediated pathway and the extrinsic or receptor-linked pathway, with caspase-9 or caspase-8, as the initiator caspase, respectively (76), but the morphology of cell death remains the same irrespective of the biochemical pathway initiated. In these studies Z-VAD-fmk did not inhibit mitochondrial membrane depolarization but did inhibit DNA degradation (Fig. 9), morphological changes (Fig. 7), and caspase-3 activation (Fig. 8); this difference is important and can be used to elucidate the specific apoptotic pathway triggered by the endoperoxides by considering the order of apoptotic events. The lack of inhibition of mitochondrial membrane depolarization indicates that caspase-9 and not caspase-8 is responsible for the initiation of the downstream caspases-3 and -7.

Radical Involvement in Endoperoxide-induced Cell Death
Overall, we have demonstrated that the endoperoxides induce apoptosis via the chemical stress-mediated pathway, suggesting that mitochondrial perturbation is an apical event in cytotoxicity, and that the execution of apoptotic cell death is dependent upon caspase activity. The importance of mitochondria to the cytotoxicity of endoperoxides in other cell types was the subject of a recent study, which reported the effects of ART on the mitochondria in a yeast model. ART disrupted the normal functions of the mitochondria through its ability to depolarize the membrane potential (77). The authors proposed that the electron transport chain plays a role in ART bioactivation and that the radicals formed act locally to damage the mitochondria. The exact cytotoxic trigger of the endoperoxides remains unknown although it has been reported that endoperoxide-derived radical species are able to form adducts with proteins (67) and can damage DNA (78). In addition, ART endoperoxides have been shown to trigger the production of reactive oxygen species (13), which have also been implicated in the cytotoxic mechanism of action.
In summary, by using novel ART derivatives as chemical probes, we have shown that the differential cytotoxicity of these endoperoxides between resting and dividing cells is a result of selective activation of the endoperoxide bridge via C-centered radical species in susceptible cells. We have demonstrated that selective targeting of metabolically active cells results in mitochondrial membrane depolarization, the induction of apoptosis via the chemical stress pathway, and the activation of caspase-3 and -7 culminating in the formation of hypodiploid DNA. Collectively, this provides a chemical and molecular explanation of endoperoxide-induced cell death (Fig. 10), which can be used to identify cell types that might be susceptible to these drugs and aids assessments of the potential hazards associated with clinical use of the endoperoxides.