Synthesis and Characterization of a Triphenylphosphonium-conjugated Peroxidase Mimetic

Mitochondrial production of peroxides is a critical event in both pathology and redox signaling. Consequently their selective degradation within mitochondria is of considerable interest. Here we have explored the interaction of the peroxidase mimetic ebselen with mitochondria. We were particularly interested in whether ebselen was activated by mitochondrial glutathione (GSH) and thioredoxin, in determining whether an ebselen moiety could be targeted to mitochondria by conjugating it to a lipophilic cation, and in exploring the nature of ebselen binding to mitochondrial proteins. To achieve these goals we synthesized 2-[4-(4-triphenylphosphoniobutoxy) phenyl]-1,2-benzisoselenazol)-3(2H)-one iodide (MitoPeroxidase), which contains an ebselen moiety covalently linked to a triphenylphosphonium (TPP) cation. The fixed positive charge of TPP facilitated mass spectrometric analysis, which showed that the ebselen moiety was reduced by GSH to the selenol form and that subsequent reaction with a peroxide reformed the ebselen moiety. MitoPeroxidase and ebselen were effective antioxidants that degraded phospholipid hydroperoxides, prevented lipid peroxidation, and protected mitochondria from oxidative damage. Both peroxidase mimetics required activation by mitochondrial GSH or thioredoxin to be effective antioxidants. Surprisingly, conjugation to the TPP cation led to only a slight increase in the uptake of ebselen by mitochondria due to covalent binding of the ebselen moiety to proteins. Using antiserum against the TPP moiety we visualized those proteins covalently attached to the ebselen moiety. This analysis indicated that much of the ebselen present within mitochondria is bound to protein thiols through reversible selenenylsulfide bonds. Both MitoPeroxidase and ebselen decreased apoptosis induced by oxidative stress, suggesting that they can decrease mitochondrial oxidative stress. This exploration has led to new insights into the behavior of peroxidase mimetics within mitochondria and to their use in investigating mitochondrial oxidative damage.

Production of reactive oxygen species (ROS) 1 by the mitochondrial respiratory chain contributes to a range of human diseases (1)(2)(3)(4) and is important for redox signaling (5,6). Hydrogen peroxide (H 2 O 2 ) and phospholipid hydroperoxides are particularly important in these processes (5)(6)(7)(8). Therefore, the ability to degrade mitochondrial peroxides selectively would be useful for investigating mitochondrial peroxide production in oxidative damage, apoptosis, and redox signaling and would also have therapeutic potential (9,10).
The isoselenazole derivative ebselen (2-phenyl-1,2-benzoisoselenazol-3(2H)-one) (Fig. 1A) has been used to degrade peroxides. Ebselen was first synthesized in 1924 (11), and its thiol peroxidase and antioxidant activities were first reported in 1984 (12). Since then its antioxidant properties have led to its use as a research tool to block oxidative damage in vitro and in clinical studies of stroke and inflammatory diseases (9,(12)(13)(14)(15)(16)(17)(18)(19)(20)(21). The chemistry that enables ebselen to act as a peroxidase mimetic depends on its selenium (Se) atom, as indicated by the loss of activity on replacing Se with sulfur (S) (12). The chemistry of ebselen is surprisingly complex (9,19,22), and our current understanding is outlined in Fig. 1. Ebselen (Fig. 1A) reacts readily with glutathione (GSH) to form the selenenylsulfide (B), which is reduced by GSH to give the selenol (C) and glutathione disulfide (GSSG) (19,(22)(23)(24). Ebselen (Fig. 1A) is more efficiently reduced to the selenol (C) by thioredoxin (Trx(SH) 2 ) and is also reduced directly by thioredoxin reductase (TrxR) (21). The pK a of non-aromatic organic selenols ranges from 5.2 to 7.3 (25,26), and that for benzeneselenol is ϳ4.6 (27). Therefore, although the pK a of the ebselen-derived selenol (C) is not known, it is expected to be primarily in the selenolate form in vivo. This favors nucleophilic attack on peroxides and explains why the selenol (Fig. 1C) is the most peroxide-reactive form of ebselen (21)(22)(23). Oxidation of the selenol (Fig. 1C) by peroxides is thought to regenerate ebselen via an intermediate selenenic acid (D) that undergoes rapid intramolecular condensation (10,22) thereby avoiding irreversible oxidation to a seleninic or selenonic acid. Although ebselen (Fig. 1A) itself is oxidized directly by peroxides to the selenoxide (E) (22)(23)(24), this reaction is much slower than that between selenol (C) and peroxide (24). However, the oxidation of ebselen ( Fig. 1A) to the selenoxide (E) by peroxynitrite is fast and may be of significance in vivo (20,28). Under normal conditions the selenoxide (Fig. 1E) is unlikely to persist because it is readily reduced back to ebselen (A) by GSH (24) and TrxR (29). Reac-  2 , directly by TrxR, or by GSH via a selenenylsulfide (B). The selenol (C) reacts directly with peroxides to form a selenenic acid (E), which undergoes a rapid intramolecular condensation to ebselen (A). In addition, the selenol reacts with the selenenylsulfide (B) or ebselen (A) to form a diselenide (F), which is reduced to the selenol (C) by Trx(SH) 2 or TrxR. Ebselen (A) also reacts directly with peroxides to form a selenoxide (E), which is reduced back to ebselen by GSH or TrxR. R ϭ H or lipid. tion of the selenol (Fig. 1C) with ebselen (A) or with the selenenylsulfide (B) produces the diselenide (F) (19,30). The diselenide (F) is efficiently reduced to the selenol by Trx(SH) 2 or TrxR (30) but less so by GSH. This is presumably because of the low E°Ј for the ebselen diselenide/selenol couple; the value is not known, but that of the related selenocysteine diselenide/ selenocysteine is Ϫ488 mV (26), well below the E h of GSH in cells (Ϫ230 to Ϫ260 mV (31)). In contrast, the reduction potential for Trx(SH) 2 is lower, more closely linked to that of NADP/ NADPH (E h Ϫ374 to Ϫ390 mV (31)). In summary, the thiol peroxidase activity of ebselen (Fig. 1A) is thought to be dependent on its efficient reduction to the selenol (B) by thiols and subsequent oxidation of the selenol by peroxide (B) (19,21,22,30).
However, the details of many of these reactions are uncertain due to the difficulties of analyzing the reaction intermediates. Although ebselen has been used to a limited extent in mitochondrial studies, both toxic and protective effects have been reported, and it is unclear if ebselen is activated by mitochondrial GSH and Trx (32)(33)(34). In addition, there are indications that ebselen may bind covalently to proteins in vivo, but the nature and extent of this putative binding is unclear (35,36). A peroxidase mimetic that was accumulated selectively by mitochondria would be of interest. We have developed a series of mitochondria-targeted antioxidants by covalently coupling an antioxidant to the lipophilic triphenylphosphonium (TPP) cation (37)(38)(39)(40)(41). These compounds were selectively accumulated several hundred-fold by mitochondria due to the mitochondrial membrane potential and thereby selectively blocked mitochondrial oxidative damage (40 -50). As a natural progression we sought to develop an antioxidant that was selective for mitochondrial peroxides by making a TPP derivative of ebselen. There are a number of further advantages of conjugating an ebselen moiety to a TPP cation. The fixed positive charge of the TPP cation should facilitate analysis by mass spectrometry of ebselen chemistry. Furthermore, the uncertainties about the extent and nature of ebselen binding to proteins should be resolved using a TPP-ebselen conjugate, as antiserum against the TPP moiety can be used to detect binding to proteins (51). Here we describe the synthesis of a TPP-conjugated derivative of ebselen (MitoPeroxidase) and report on an extensive series of experiments comparing it with ebselen. We have investigated the reaction of the ebselen moiety with GSH and peroxides by mass spectrometry of its binding to proteins and its uptake and antioxidant efficacy within mitochondria and cells. Together these data enable us to draw a number of conclusions about the mechanism of action of the ebselen moiety within mitochondria.

EXPERIMENTAL PROCEDURES
Materials-Thioredoxin was from Escherichia coli, thioredoxin reductase was from E. coli, or rat and glutathione reductase (GR) was from bakers' yeast; all were from Sigma. Hydrogen peroxide was from BDH, and stock solutions were quantitated using ⑀ 240 ϭ 43.6 M Ϫ1 cm Ϫ1 . Tert-butylhydroperoxide (tBHP; 7.3 M stock) and soybean lipoxidase (type V, catalog number L6632) were from Sigma. Ebselen (CAS registry number (60940-34-3)) was from Calbiochem, and the selenenylsulfide of ebselen with GSH S-(2-phenylcarbamoylbenzene-selenyl)glutathione was synthesized as reported (24). (4-Iodobutyl)triphenylphosphonium iodide was synthesized as reported (51). Stock solutions of 2-100 mM ebselen or MitoPeroxidase were prepared in Me 2 SO and were stored at Ϫ20°C. MitoPeroxidase was readily soluble in Me 2 SO but was less soluble in ethanol or aqueous buffer than ebselen. Working solutions of MitoPeroxidase up to 100 M were usually prepared in aqueous buffer using Me 2 SO as a co-solvent, usually 0.2% v/v except where indicated otherwise.
Chemical Syntheses-Preparative column chromatography was performed using silica gel (Merck type 60, 200 -400 mesh, 40 -63 m). Analytical thin layer chromatography (TLC) was performed using silica gel (Merck, 60F 254). Infrared spectra were acquired using a PerkinElmer Life Sciences BX FT-IR spectrometer, and samples were prepared as KBr discs. Nuclear magnetic resonance (NMR) spectra were acquired using a Varian 300 or 500 MHz instrument. Chemical shift (␦) data are reported in units of ppm relative to an external or internal reference. Tetramethylsilane was used an as internal reference for 1 H NMR spectra acquired in CDCl 3 , and solvent peaks were used as an internal reference for 1 H NMR acquired in other solvents. Solvent peaks were used as the internal reference for 13 C NMR, and phosphoric acid (30%) was used as an external reference for 31 P NMR. Ebselen was used as an external reference for 77 Se NMR (24). Melting points were measured using a Kofler Heizbank melting point bench and are uncorrected. Scanning UV-visible absorption spectra were measured at 20°C in a quartz cuvette using a Cary 500 Scan UV-visible-Near InfraRed spectrophotometer with base-line correction. Low resolution positive electrospray ionization mass spectra were acquired using a Shimadzu LCMS-QP800X liquid chromatography mass spectrometer, and data are reported as m/z values. The overall synthesis of MitoPeroxidase is outlined in Scheme 1 and is described below.
N-(4-Hydroxyphenyl)-benzamide (1)-A solution of benzoyl chloride (12.9 g, 91.7 mmol) in dry tetrahydrofuran (THF) (250 ml) was added dropwise to a solution of 4-aminophenol (10.0 g, 91.7 mmol) and Et 3 N (9.4 g, 93 mmol) in dry THF (700 ml). The mixture was stirred for 2 days under a dry atmosphere, and the solvent was then removed in vacuo.  (2)-A solution of tert-butyldimethylsilylchloride (9.6 g, 63.9 mmol) in dry THF (60 ml) was added dropwise to a solution of 1 (13.6 g, 63.8 mmol) and imidazole (10.9 g, 63.8 mmol) in dry THF (300 ml) at 0°C under a dry atmosphere. The suspension was warmed to room temperature and stirred overnight under an argon atmosphere. The mixture was then diluted with 5% NaHCO 3 solution (300 ml) and extracted with CH 2 Cl 2 (2 ϫ 300 ml). The organic extracts were combined, dried over MgSO 4 , and filtered. Solvent was removed in vacuo to give a thick liquid, which was dissolved in boiling hexane (500 ml). The hot solution was filtered from the insoluble solid, cooled to room temperature, and then on ice to give white solids. The solids were collected by filtration, washed with cold hexane, and dried in air, giving 2 as white crystals (14. (3)-To a solution of 2 (9.20 g, 28.5 mmol) in dry THF (150 ml) at Ϫ15°C under an argon atmosphere n-butyllithium (1.6 M in hexane, 35 ml, 56.0 mmol) was slowly added via a cannula over 45 min. The resultant orange solution was stirred at Ϫ15°C for 45 min, selenium powder (2.22 g, 28.1 mmol) was then added, and the mixture was then stirred for 45 min at Ϫ15°C to give a deep red solution. The solution was cooled to Ϫ78°C, and CuBr 2 (12.6 g, 56.2 mmol) was added in 3 portions over 15 min. The suspension was stirred for 1 h at Ϫ78°C then removed from the cooling bath and stirred for 22 h. The resultant brown mixture was poured into 1% aqueous acetic acid (600 ml) and extracted with CH 2 Cl 2 (3 ϫ 500 ml). The organic fractions were combined, filtered, and dried over Na 2 SO 4 . Solvent was then removed in vacuo to give a brown greasy solid. The solid was chromatographed on silica gel prepared in CH 2 Cl 2 , and elution with CH 2 Cl 2 gave crude 3 as a pale tan solid. The solid was recrystallized from EtOH, filtered, washed with hexane, and dried in air, giving 3 as pale yellow crystals (5. 2-[4-(4-Triphenylphosphoniobutoxy)-phenyl]-1,2-benzisoselenazol)-3(2H)-one Iodide (MitoPeroxidase)-A 60% dispersion of NaH (42 mg, 1.05 mmol) in oil was washed with pentane three times, and the solid residue dried in vacuo. Dry N,N-dimethylformamide (2 ml) was then added, and the resultant suspension was stirred for 10 min at room temperature. A solution of 4 (0.255g, 0.88 mmol) in dry N,Ndimethylformamide (3 ml) was then added dropwise via cannula. The resultant solution was stirred for 90 min at room temperature, and a solution of (4-iodobutyl)triphenylphosphonium iodide (0.503 g, 0.88 mmol) in N,N-dimethylformamide (3 ml) was then added dropwise. A yellow-brown solution formed, which was left to stir at room temperature for 48 h. Water (2 ml) was then carefully added, and the solution then diluted with CH 2 Cl 2 (50 ml). The organic phase was washed with water (3 ϫ 50 ml), dried over Na 2 SO 4 , and filtered. Solvent was removed in vacuo to give a brown oily residue. The residue was triturated three times from CH 2 Cl 2 (1-2 ml) with diethyl ether (50 ml), and removal of residual solvent in vacuo gave a pale yellow solid. The solid was chromatographed on silica gel (20 g) prepared in CH 2 Cl 2 , and Octan-1-ol/PBS Partition Coefficients-Solutions of ebselen (4 mM) or MitoPeroxidase (100 M) were prepared from stocks in Me 2 SO in PBSsaturated octan-1-ol. Aliquots (100 l) of the stock solutions were added to 10 ml of octan-1-ol-saturated PBS in a glass Kimax tube (49). The Me 2 SO content was 0.04% (v/v). After shaking for 1 min the tubes were then centrifuged (1000 ϫ g for 5 min) to separate the phases, and the absorbance at 268 nm of the upper octan-1-ol phase was determined. The aqueous phase (10 ml) was transferred to a clean Kimax tube, the compounds were re-extracted into 1 ml of fresh octan-1-ol, and the absorbance at 268 nm of the organic phase was then determined. Subsequent extractions with octan-1-ol had no absorbance at 268 nm, SCHEME 1. Synthesis of MitoPeroxidase (6). Bu, butyl; tBDMSCl, tertbutyldimethylsilylchloride; OtBDMS, tert-butyldimethyl; IBTP, 4-iodobutyl)triphenylphosphonium iodide. and the partition coefficient of ebselen was the same when 100 M was used as the initial concentration. Decreasing the Me 2 SO content to 0.01% did not affect the partition coefficients. The partition coefficients, determined as the means Ϯ S.E. of 3-6 determinations, were: ebselen, 660 Ϯ 4; MitoPeroxidase, 220 Ϯ 60. The partition coefficient for TPMP was 0.35 Ϯ 0.02, as determined previously (40).
Absorbance and Fluorescence Measurements-For coupling peroxidase turnover to NADPH (⑀ 340 ϭ 6,200 M Ϫ1 cm Ϫ1 ) consumption by GR or TrxR, a Shimadzu UV-2501PC spectrophotometer was used with 1-ml matched quartz cuvettes. The concentration of H 2 O 2 was measured using a H 2 O 2 -selective electrode (Apollo 4000, World Precision Instruments). To quantitate the oxidation of GSH or Trx(SH) 2 by MitoPeroxidase samples were incubated in 3-ml cuvettes in an Aminco DW2000 spectrophotometer, and GR or TrxR was injected into cuvettes within the spectrophotometer through a light-tight septum in the chamber lid to enable measurement of absorbance changes without disrupting recording. Fluorescence measurements of Trx(SH) 2 were obtained using a 3-ml stirred cuvette in a Shimadzu RF5301PC fluorometer with excitation and emission slit widths of 3 nm. For scanning spectra, Excitation was 290 nm, and the Emission was scanned from 300 to 500 nm, with a fast scan rate and a sampling interval of 1 nm. To measure kinetic changes Excitation was 290 nm, and the change in intensity of Emission (345 nm) was monitored. There was no fluorescence from the buffer under these conditions.
Mitochondrial Preparations-Rat liver mitochondria were prepared by homogenization followed by differential centrifugation in 250 mM sucrose, 5 mM Tris-HCl, 1 mM EGTA, pH 7.4 (52). Bovine heart mitochondrial membranes were prepared by disruption of mitochondria in a blender followed by collection and washing by centrifugation (53). These mitochondrial membrane preparations had negligible matrix contamination (54). Before use in the MitoPeroxidase binding experiments, the mitochondrial membranes were suspended in incubation buffer (50 mM K 2 HPO 4 , 1 mM EGTA, 100 M N,N-bis (2-bis[carboxymethyl]aminoethyl)glycine, pH 8.0) and reduced with 10 mM dithiothreitol (DTT) at 37°C for 10 min, pelleted by centrifugation at 10,000 ϫ g for 10 min, washed twice in buffer without DTT, and suspended at ϳ1 mg of protein⅐ml Ϫ1 . Protein concentrations were determined by the biuret assay (55) or by the bicinchoninic acid assay (56) using bovine serum albumin as a standard.
Mitochondrial Incubations-To measure the effects of MitoPeroxidase on respiration, rat liver mitochondria (1 mg of protein⅐ml Ϫ1 ) were incubated at 25°C in a 1-ml oxygen electrode chamber (Rank Brothers, Bottisham, Cambridgeshire, UK) in 120 mM KCl, 10 mM HEPES, 1 mM EGTA, 1 mM potassium phosphate, pH 7.2, supplemented with 10 mM succinate and 4 g⅐ml Ϫ1 rotenone or 5 mM each of glutamate and malate. Coupled respiration was measured for 2 min, then phosphorylating respiration was measured for 1 min after the addition of 200 M ADP, and finally, uncoupled respiration was measured after the addition of 1 M carbonylcyanide-p-trifluoromethoxyphenylhydrazone (FCCP).
The mitochondrial membrane potential was measured by incubating mitochondria in 120 mM KCl, 10 mM HEPES, 1 mM EGTA, pH 7.2, containing 0.5 M methyltriphenylphosphonium (TPMP) supplemented with [ 3 H]TPMP (100 nCi⅐ml Ϫ1 ) at 25°C for 5 min. At the end of the incubation the mitochondria were pelleted by centrifugation, and the amounts of TPMP in the pellet and supernatant were quantitated by scintillation counting (57). The membrane potential was calculated assuming a mitochondrial volume of 0.5 l⅐mg Ϫ1 protein, and that 60% of intramitochondrial TPMP was membrane-bound (58, 59).
For some experiments the mitochondrial glutathione and nicotinamide nucleotide pools were depleted by inducing the mitochondrial permeability transition (MPT). To do this rat liver mitochondria (5 mg of protein⅐ml Ϫ1 ) were incubated in a 1-ml spectrophotometer cuvette containing 120 mM KCl, 10 mM HEPES, 1 mM EGTA, pH 7.2, supplemented with 10 mM succinate, 8 g⅐ml Ϫ1 rotenone, 300 nmol of CaCl 2 , and 5 mM potassium phosphate at 25°C for 10 min. Induction of the MPT was assessed by measuring mitochondrial swelling from the decrease in absorbance at 540 nm due to decreased light scattering. At the end of the incubation the mitochondria were pelleted by centrifugation (4000 ϫ g for 3 min) and washed once in 100 mM KCl, 10 mM Tris-HCl, pH 7.6. Control mitochondria were incubated in the absence of calcium and phosphate. This treatment decreased glutathione content by 95% as measured by the glutathione recycling assay (60), confirming loss of low molecular weight matrix components during the MPT.
Ion-selective Electrode Measurements-An electrode selective for the TPP cation moiety was constructed as described (61,62). The electrode was characterized for TPP cations by filling with 10 mM TPMP in KCl medium (120 mM KCl, 10 mM HEPES, 1 mM EGTA, pH 7.2) and soaking in this medium overnight. The output from the electrode was passed to a MacLab TM data acquisition system (62). The electrode response for MitoPeroxidase was close to that predicted by the Nernst equation for a singly charged cation, giving a slope of 62-64 mV per decade at MitoPeroxidase concentrations above ϳ4 M.
Oxidative Damage Assays-To measure thiobarbituric acid reactive species (TBARS), liver mitochondria (2 mg of protein⅐ml Ϫ1 ) were incubated at 37°C for 15 min in 100 mM KCl and 10 mM Tris-HCl, pH 7.6, supplemented with 10 mM succinate, 8 g/ml rotenone, and oxidative stress was induced by the addition of 50 M FeCl 2 and 500 M H 2 O 2 in the presence or absence of 1 M peroxidase mimetics or TPMP. After the incubation 0.8-ml aliquots of the mitochondrial suspension were mixed with 0.1% butylated hydroxytoluene, 200 l of 1% thiobarbituric acid, and 200 l of 35% HClO 4 , heated at 100°C for 15 min, diluted with 3 ml of water, and extracted once into 3 ml of n-butanol. TBARS were determined fluorometrically ( Excitation ϭ 515 nm, Emission ϭ 553 nm) in a Molecular Devices Spectra Max Gemini XS fluorometric plate reader, compared with a standard curve of 1,1,3,3-tetraethoxypropane, and expressed as nmol of malondialdehyde by comparison with standard solutions (40).
To measure phospholipid hydroperoxides, at the end of the incubation the mitochondrial membranes were pelleted by centrifugation and then suspended in 200 l of 100 mM KCl and 10 mM Tris-HCl, pH 7.6. Water (300 l) and methanol (250 l) were added and mixed vigorously, and then ethyl acetate (500 l) was added, and after mixing the suspension was separated into aqueous and organic layers by centrifugation (5 min at 3000 ϫ g). The upper organic layer was transferred to a glass test tube, and the aqueous layer was re-extracted with ethyl acetate (500 l). The pooled organic layers were concentrated by evaporation under a stream of nitrogen to ϳ100 l. The hydroperoxides were detected by mixing 100 l of extracted phospholipids with 900 l of FOX reagent (100 M xylenol orange, 250 M ferrous ammonium sulfate, 4 mM butylated hydroxytoluene, 25 mM H 2 SO 4 in 90% methanol). The mixture was incubated at room temperature for 30 min, and the absorbance was then read at 560 nm and compared with a H 2 O 2 standard curve (0 -100 M).
Electrophoresis and Immunoblotting-After incubations the mitochondria or the mitochondrial membranes were washed twice in incubation buffer, and the pellet was then suspended in 100 l of SDS-PAGE loading buffer. For blotting for cytochrome oxidase or manganese-superoxide dismutase, conventional loading buffer was used (50 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 0.1% bromphenol blue, 100 mM DTT). For assessing the binding of MitoPeroxidase, loading buffer without DTT and supplemented with 50 mM N-ethylmaleimide (NEM) was used. After heating at 100°C for 4 min, 20 g of protein was separated on a 12.5% SDS-polyacrylamide gel using a Bio-Rad Mini Protean system. The proteins were then transferred to 0.2-m nitrocellulose using a Bio-Rad Mini Trans Blot system and blocked with 2% fat-free milk powder in Tris-buffered saline (5 mM Tris-HCl, pH 7.4, 20 mM NaCl). Antibody binding was detected using a 1:2000 dilution of antitriphenylphosphonium rabbit antiserum (51) followed by a 1:20,000 dilution of rabbit IgG conjugated to horseradish peroxidase (Bio-Rad). Antibody binding was then visualized by enhanced chemiluminescence (ECL, Amersham Biosciences).
To investigate the intramitochondrial localization of MitoPeroxidase, liver mitochondria were incubated with MitoPeroxidase with and without 1 M FCCP for 10 min at 37°C. The mitochondria (1 mg of protein) were pelleted and resuspended in 100 l KCl of buffer containing 2 mg⅐ml Ϫ1 Thesit (0.2 mg of Thesit/mg of protein) containing 1 mM NEM and incubated on ice for 1 h. Matrix and membrane-enriched fractions were then prepared by pelleting the membranes in a Beckman Airfuge (100,000 ϫ g for 10 min) at 4°C. The matrix fraction was retained, and the membrane fraction was washed once in KCl buffer containing 1 mM NEM. Both the membrane and matrix fractions were then resolved on a 12.5% SDS-PAGE gel and immunoblotted for the presence of the TPP moiety as described before. Control immunoblots were against an anti mouse manganese-superoxide dismutase antibody (mouse monoclonal diluted 1:1000, Transduction Laboratories) showed complete separation of the mitochondrial matrix from the membranes.
Cell Culture-A rat basophilic leukemia cell line (RBL-2H3) was cultured at 37°C under humidified 95% air, 5% CO 2 in Eagle's minimal essential medium containing Earle's balanced salt solution and supplemented with 2 mM Glutamax, penicillin (100 units⅐ml Ϫ1 ), streptomycin (100 g⅐ml Ϫ1 ), and 15% heat-inactivated fetal calf serum. To test toxicity cells were grown to 90% confluence and incubated for 24 h with Eagle's minimal essential medium, 5% serum containing the compounds, and then cell death was determined by the release of lactate dehydrogenase or by propidium iodide labeling. These measurements showed that MitoPeroxidase and TPMP had negligible toxicity up to 10 M, but at 25 M and above toxicity was evident. Some ebselen toxicity was evident from 10 M and above.
Apoptosis Assay-Apoptosis was induced by incubating RBL-2H3 cells in Eagle's minimal essential medium containing 5% fetal calf serum with 100 mM 2-deoxyglucose (2DG) or 100 M H 2 O 2 . For measuring caspase-3 activity, 10 5 cells were plated overnight in 24-well tissue culture plates and then incubated with 2DG or H 2 O 2 with and without compounds. At the end of the incubation the cells were washed with PBS and lysed in 50 l of lysis buffer (50 mM HEPES, 10% sucrose, 0.1% Triton X-100, 10 mM DTT, pH 7.5) for 20 min on ice. The lysates were clarified by centrifugation (10,000 ϫ g for 10 min), and supernatant aliquots (50 l) were transferred to 96-well plates and incubated with 50 l of 50 M 7-amino-4-methylcoumarin-labeled peptide (DEVD-AMC) in 150 l of lysis buffer. Caspase activity was measured fluorometrically by the release of fluorescent aminomethylcoumarin on cleavage of the peptide in a Spectra Max Gemini XS fluorimeter plate reader ( Excitation ϭ 355 nm, Emission ϭ 460 nm).

RESULTS AND DISCUSSION
Synthesis of MitoPeroxidase-The synthesis of MitoPeroxidase is shown in Scheme 1 and the details are given under "Experimental Procedures." The basic synthetic approach involved the alkylation of a phenoxide moiety with a -haloalkyl phosphonium salt ((4-iodobutyl)triphenylphosphonium iodide) as reported previously (63). The benzamide derivative 1 was prepared from benzoyl chloride and 4-aminophenol. The direct synthesis of the isoselenazole 4 from 1 using a modification of a reported ortholithiation, selenation, and oxidative cyclization methodology (64) was not successful. Therefore, we protected the phenol group of 1 using tert-butyldimethylsilylchloride to give 2. Ortholithiation of 2 followed by selenation and oxidative cyclization gave the isoselenazole derivative 3. The silyl group was easily removed from 3 using standard tetrabutylammonium fluoride methodology to give 4. Reaction of the phenoxide of 4 with (4-iodobutyl)triphenylphosphonium iodide gave MitoPeroxidase.
Reaction of MitoPeroxidase with Glutathione or Dithiothreitol-To determine whether the isoselenazole moiety of ebselen and MitoPeroxidase was converted to a selenenylsulfide by reaction with GSH, we reacted 50 M GSH with 50 M ebselen or MitoPeroxidase. The spectrum changed rapidly to that of the selenenylsulfide and was stable for at least 30 min (data not shown). This demonstrates that ebselen or MitoPeroxidase react rapidly with equimolar GSH to produce the selenenylsulfide. We next determined if MitoPeroxidase was reduced to the selenol by excess GSH or DTT. When ebselen or MitoPeroxidase was treated with excess GSH or DTT the absorbance at 340 nm decreased, and there was significant absorbance around 400 nm (data not shown). These changes suggested that a selenol had been produced (22,23,30). However, the UVabsorption spectra of ebselen and its derivatives in the 340 -400-nm region overlap with the following extinction coefficients (M Ϫ1 ⅐cm Ϫ1 at 340 nm): ebselen (4,000), ebselen selenol (ϳ2,000) (21), ebselen diselenide (21,000) (30), ebselen-glutathione selenenylsulfide (2,000). Consequently it was not possible to identify the products and the extent of reaction of MitoPeroxidase with excess GSH or DTT from changes in the absorption spectra alone. Therefore, we analyzed the products of the reaction of MitoPeroxidase with thiol reductants by low resolution positive ion ESMS. This analysis was facilitated by the fixed positive charge of the TPP moiety of MitoPeroxidase.
The structures and masses of MitoPeroxidase and pertinent derivatives are shown in Fig. 2A. The mass spectrum of 27 M MitoPeroxidase showed a peak with the expected m/z of 608 for the predominant Se isotope, 80 Se (Fig. 2B). The inset to Fig. 2B shows the m/z distribution primarily due to the Se isotopes, which was similar to that calculated for MitoPeroxidase (www.  (8)). Analysis of a 1:1 mixture of MitoPeroxidase and GSH by ESMS showed a greatly diminished peak at 608 m/z and a new, albeit weak peak at 915 m/z (data not shown) that was assigned to the selenenylsulfide ( Fig.  2A, iv). The low sensitivity of the selenenylsulfide to positive ion ESMS is presumably due to ionization of the carboxylic acid groups. When MitoPeroxidase was treated with excess (50ϫ) GSH, ESMS of the mixture showed a dominant peak at 610 m/z, and the observed isotope distribution (Fig. 2C) was similar to that calculated for the selenol (Fig. 2A, ii). Similar spectra were observed when MitoPeroxidase was treated with equimolar or excess DTT (data not shown). To further confirm that the addition of excess GSH to MitoPeroxidase gave the selenol, we examined the effect of adding 1-chloro-2,4-dinitrobenzene (CDNB) to trap the selenol as a stable selenide adduct ( Fig. 2A, vi; m/z ϭ 776) (23). ESMS of this mixture showed an intense peak at 776 m/z (Fig. 2D), consistent with reduction of MitoPeroxidase to the selenol. Similar spectra were observed when MitoPeroxidase was treated with 2 eq of GSH or with DTT followed by the addition of CDNB (data not shown). MitoPeroxidase or its diselenide did not react with CDNB, and the diselenide was reduced to the selenol by excess GSH (data not shown). In addition reaction of the diselenide with one equivalent of GSH also gave selenol (data not shown), presumably via thiol-selenol exchange.
To determine whether the reaction of glutathione with Mi-toPeroxidase to form the selenol involved the oxidation of 2 mol of GSH to GSSG (Reaction 1), we incubated 60 nmol of MitoPeroxidase with excess GSH in the presence of NADPH.
Subsequent addition of GR resulted in a decrease in A 340 (Fig.  3). A further addition of GR had no effect on A 340 (Fig. 3), confirming that the change was due to the reduction of accumulated GSSG. The magnitude of the decrease in A 340 indicated that 58 Ϯ 2 nmol of GSSG formed on incubating 60 nmol of MitoPeroxidase with excess GSH, consistent with the stoichiometry of Reaction 1.
Oxidation of MitoPeroxidase Selenol by Peroxides-We next used ESMS to determine whether the selenol derived from MitoPeroxidase reacted readily with H 2 O 2 or tBHP to regenerate MitoPeroxidase. The selenol was produced by reacting 27 M MitoPeroxidase with equimolar DTT for 2 min and was stable in solution over 2-30 min, as confirmed by CDNB trapping. When the selenol was treated with 200 M H 2 O 2 , CDNB trapping followed by ESMS showed the loss of selenol over 5 min and a concomitant increase in intensity of peaks in the 608 m/z region (Fig. 4A, i-iii). The pattern in the 608 m/z region indicated that a mixture of MitoPeroxidase (608 m/z) and its diselenide (609 m/z) may have formed. This was investigated further by UV-visible spectroscopy of the CDNBfree reaction mixture. After 5 min the UV-visible spectrum largely resembled that of 13.5 M MitoPeroxidase diselenide, as confirmed by independent synthesis of the diselenide. 2 Formation of the diselenide could occur under these conditions by the selenol reacting with accumulated MitoPeroxidase, as reported for ebselen (30), although we cannot exclude the possibility of reaction of the selenol with a selenenic acid intermediate. At higher H 2 O 2 concentrations the loss of selenol was faster, and the isotope distribution in the 608 m/z region was similar to that calculated for MitoPeroxidase. Similar results were obtained when using tBHP as the oxidant, although the rate was considerably slower (data not shown). Despite extensive monitoring of these reactions, a peak indicative of a selenenic acid intermediate (626 m/z Fig.   1D) was not observed for either peroxide, presumably because of the short lifetime of this species due to its rapid intramolecular condensation.  (Fig. 4B). The addition of excess (50ϫ) GSH to the selenoxide reduced the selenoxide to the selenol, as was indicated by CDNB trapping followed by ESMS (data not shown). MitoPeroxidase was also oxidized to the selenoxide by tBHP, albeit more slowly.
In summary, these data indicate that the selenol reacts with peroxides to regenerate MitoPeroxidase. In the absence of reductants, the reaction of the selenol with relatively low concentrations of peroxide ultimately generates the diselenide. In the presence of physiological concentrations of GSH or Trx(SH) 2 the reaction of MitoPeroxidase with its selenol is unlikely to compete significantly with its reaction with thiols. Therefore MitoPeroxidase and ebselen should degrade peroxides catalytically in the presence of GSH or Trx(SH) 2 . MitoPeroxidase itself reacts with peroxides to form a selenoxide, but this reaction is much slower that its reaction with thiols and is, therefore, not likely to be significant under normal conditions in vivo. If the selenoxide is formed in vivo via another pathway it is unlikely to persist under normal conditions as it is readily reduced by GSH.
Reduction of MitoPeroxidase by Thioredoxin-A major pathway for the reduction of ebselen to the selenol in vivo is through the thioredoxin system (21,30). To determine whether Trx(SH) 2 reduced MitoPeroxidase to the selenol we monitored the quenching of the fluorescence of tryptophan residues that occurs upon oxidation of Trx(SH) 2 to Trx(SS) (21,65). Trx(SH) 2 gave a distinctive fluorescence emission spectrum between 300 and 400 nm (Fig. 5A, trace a) that was largely abolished over 5 min by equimolar MitoPeroxidase (Fig. 5A, traces b-d). The oxidation of Trx(SH) 2 was essentially complete as shown by the minimal decrease in fluorescence on adding another 5 M Mito-Peroxidase (data not shown). The subsequent addition of DTT effectively restored the spectrum to that of Trx(SH) 2 (Fig. 5A, trace e), confirming that the fluorescence changes were due to the oxidation of Trx(SH) 2 to Trx(SS) by MitoPeroxidase. Ebselen is known to oxidize Trx(SH) 2 rapidly with a rate constant of ϳ2 ϫ 10 7 M Ϫ1 s Ϫ1 (21). To compare the rates of reaction of ebselen and MitoPeroxidase with Trx(SH) 2 , we measured the kinetics of the oxidation under pseudo-first order conditions by employing an excess of the selenium compound (50 M) over Trx(SH) 2 (2 M); these measurements indicated that MitoPeroxidase reacted with Trx at about 50% of the rate of ebselen (data not shown), suggesting that the reduction of MitoPeroxidase by Trx(SH) 2 is also rapid, with a rate constant in the range of 10 6 -10 7 M Ϫ1 s Ϫ1 .
We next determined if the stoichiometry of the reaction of MitoPeroxidase with Trx(SH) 2 was equimolar (Reaction 2).
MitoPeroxidase ϩ Trx(SH) 2 3 Selenol ϩ Trx(SS) REACTION 2 When excess Trx(SH) 2 was reacted with 30 nmol of Mi-toPeroxidase in the presence of NADPH, subsequent addition of TrxR resulted in a decrease in A 340 (Fig. 5B). The addition of a second aliquot of TrxR had no effect on A 340 , confirming that the observed change was due to the reduction of accumulated Trx(SS). The magnitude of decrease in A 340 indicated that 29.8 Ϯ 0.2 nmol of NADPH had been oxidized on reduction of 30 nmol of MitoPeroxidase. Therefore, the stoichiometry of the reaction of MitoPeroxidase with Trx(SH) 2 is equimolar.
To determine whether MitoPeroxidase was reduced directly by TrxR, we incubated 20 M MitoPeroxidase with NADPH and TrxR and at various times quenched the reaction with 6 M guanidinium HCl containing 1 mM 5,5Ј-dithio-bis-2-nitrobenzoic acid to enable the production of the selenol to be quantitated spectrophotometrically (21) (Fig. 5C). This showed that TrxR in the presence of NADPH reduced ebselen but not Mito-Peroxidase (21). This may be due to steric hindrance by the TPP-alkoxy substituent preventing access of MitoPeroxidase to the active site of TrxR.
MitoPeroxidase Catalyzed Reduction of Peroxides by Glutathione and Thioredoxin-To determine whether MitoPeroxidase catalyzed the reduction of peroxides, we first assessed the ability of the isoselenazole form of MitoPeroxidase to degrade H 2 O 2 (Fig. 6A) To see if MitoPeroxidase catalyzed the degradation of H 2 O 2 by GSH we measured the effect of oxidation of GSH to GSSG by coupling it to NADPH consumption by GR (Fig. 6B). This showed that MitoPeroxidase catalyzed the oxidation of GSH by H 2 O 2 , albeit more slowly than ebselen. To determine whether MitoPeroxidase catalyzed the oxidation of Trx(SH) 2 by H 2 O 2 , we coupled the oxidation of Trx(SH) 2 to NADPH consumption by TrxR (Fig. 6C). This showed that both MitoPeroxidase and ebselen catalyzed the degradation of H 2 O 2 by Trx(SH) 2 .
To confirm that MitoPeroxidase catalyzed the oxidation of Trx(SH) 2 by H 2 O 2 , we examined the effect of MitoPeroxidase on the rate of the reaction directly. The oxidation of Trx(SH) 2 was followed by monitoring the loss of its tryptophan fluorescence (Fig. 6D). In the absence of MitoPeroxidase or peroxide there was a relatively slow aerial oxidation of 5 M Trx(SH) 2 (Fig. 6D,  trace a). The addition of 50 M H 2 O 2 increased the rate of oxidation of Trx(SH) 2 a little over aerial oxidation (Fig. 6D,  trace b). The addition of 1 M MitoPeroxidase to 5 M Trx(SH) 2 led to the oxidation of about 20% of the Trx(SH) 2 , and the fluorescence then remained stable (Fig. 6D, trace c). Subsequent addition of H 2 O 2 led to further oxidation of Trx(SH) 2 (Fig. 6D, trace c). The extent of Trx(SH) 2 oxidation stimulated by MitoPeroxidase was far greater than the amount of MitoPeroxidase added. Therefore MitoPeroxidase catalyzes the degradation of H 2 O 2 by Trx(SH) 2 .
Effects of MitoPeroxidase and Ebselen on Isolated Mitochondria-MitoPeroxidase or ebselen above 10 M decreased the membrane potential and increased the coupled respiration rates of isolated mitochondria (data not shown). This is consistent with a nonspecific increase of the proton permeability of the inner membrane. At higher concentrations (50 M) both MitoPeroxidase and ebselen significantly inhibited respiration of both isolated mitochondria and of mitochondrial membranes (data not shown). Because 50 M TPMP did not affect respira- tion, this inhibition is due to the isoselenazole moiety of Mi-toPeroxidase and ebselen. The sites of inhibition were not pursued further in this work, but we note that ebselen is known to inhibit enzymes by reacting with critical cysteine residues (9,10,34). Therefore, both ebselen and MitoPeroxidase have dual effects on mitochondrial function, with concentrations above 10 M increasing the nonspecific proton permeability of the mitochondrial inner membrane, whereas higher concentrations inhibit respiration. Consequently, in subsequent experiments MitoPeroxidase or ebselen concentrations up to 10 M could be used with minimal disruption to mitochondrial function.
Binding of MitoPeroxidase to Isolated Mitochondria-To determine whether the TPP cation of MitoPeroxidase led to its selective accumulation by energized mitochondria, we measured the MitoPeroxidase concentration continuously using a TPP-selective electrode (49,61). The additions of MitoPeroxidase to the incubation chamber led to a clear electrode response enabling the MitoPeroxidase concentration to be measured in real time (Fig. 7A). The subsequent addition of de-energized mitochondria dramatically decreased the MitoPeroxidase concentration, apparently leaving negligible free MitoPeroxidase in solution. When de-energized mitochondria were added to the incubation chamber first, subsequent additions of Mito Peroxidase up to 5 M were not detected by the electrode (data not shown), confirming that the free concentration of MitoPeroxidase in the presence of mitochondria was negligible. The lack of free MitoPeroxidase made it impossible to assess whether there was uptake of MitoPeroxidase by energized mitochondria; consequently, subsequent generation of a membrane potential by succinate or abolition of the potential by FCCP had no effect on the electrode response (Fig. 7A).
To determine whether the reactivity of the isoselenazole moiety contributed to the binding of MitoPeroxidase to mitochondria, we reacted the selenol form of MitoPeroxidase with iodoacetamide. This carboxyamidomethylated the selenol form of MitoPeroxidase (Reaction 3; R ϭ PhO(CH 2 ) 4 ϩ PPh 3 ), as evident from matrix-assisted laser desorption ionization time-offlight mass spectrometry, which showed a peak consistent with that calculated for the selenide (m/z ϭ 667.19). REACTION 3 The concentration of carboxyamidomethylated (CAM)-Mito-Peroxidase was monitored in real time using a TPP-selective electrode. CAM-MitoPeroxidase also bound to de-energized mitochondria but to a far lower extent than MitoPeroxidase (Fig.  7B). On subsequent energization of the mitochondria with succinate there was further uptake that was reversed by dissipating the membrane potential with the uncoupler FCCP (Fig.  7B). This uptake of CAM-MitoPeroxidase by de-energized and energized mitochondria was similar to that of other TPP derivatives (49) but was significantly different from that of MitoPeroxidase itself. The partition coefficients of MitoPeroxidase and CAM-MitoPeroxidase were similar (data not shown), and it was expected that the hydrophobicity of MitoPeroxidase would result in some binding to deenergized mitochondria. The octan-1-ol/PBS partition coefficient of MitoPeroxidase (220) is significantly less than that of another TPP derivative, MitoQ 10 (2760 (49)), which shows significant membrane-potential-dependent uptake by energized mitochondria and release on uncoupling (40,49). Together these findings suggested that the reactivity of the isoselenazole ring of MitoPeroxidase leads to covalent binding to de-energized mitochondria.
Covalent Binding of MitoPeroxidase to Mitochondrial Thiol Proteins-The isoselenazole moiety may lead to MitoPeroxidase binding to mitochondria through reaction with mitochondrial thiols, as ebselen is known to form selenenylsulfide bonds with protein thiols, glutathione, and cysteine (9,35,36,66,67). To determine whether MitoPeroxidase formed selenenylsulfide bonds with mitochondrial protein thiols, we incubated MitoPeroxidase with bovine heart mitochondrial membranes. These membranes are a useful model for studying the interactions of compounds with exposed mitochondrial protein thiols (54). Af- ter incubation with MitoPeroxidase the membrane proteins were separated by non-reducing SDS-PAGE to prevent selenenylsulfide bond reduction, and the presence of MitoPeroxidase was determined using antiserum against TPP. This procedure detected extensive binding of MitoPeroxidase to membrane proteins (Fig.8A). Pretreatment of the membranes with the thiol alkylating reagent NEM prevented MitoPeroxidase binding, and the thiol reductant DTT reversed MitoPeroxidase binding by reducing the selenenylsulfide bond (Fig. 8A). When intact mitochondria were incubated with MitoPeroxidase, mitochondrial proteins were also extensively labeled by MitoPeroxidase, and this was also reversed by DTT and prevented by NEM (Fig.  8B). Similar labeling of protein thiols by MitoPeroxidase was found after incubation of MitoPeroxidase with RBL-2H3 cells (data not shown). The extent of labeling was independent of the membrane potential, as it was unaffected by preincubating with FCCP to abolish the membrane potential (Fig. 8B).
TPP cations easily permeate the mitochondrial inner membrane in the presence or absence of a membrane potential (68).
To determine the intramitochondrial location of the proteinbound MitoPeroxidase, we incubated mitochondria with Mito-Peroxidase and then isolated membrane-and matrix-enriched fractions. This showed binding of MitoPeroxidase to both matrix and membrane proteins (Fig. 8C). Therefore, Mito-Peroxidase is taken up inside mitochondria where it binds extensively to thiol proteins and to low molecular weight thiols such as GSH through selenenylsulfide bonds. This extensive binding in conjunction with adsorption to membrane surfaces due to hydrophobicity is the likely cause of the complete sequestration of MitoPeroxidase by de-energized mitochondria seen in Fig. 7A.
Antioxidant Effects of MitoPeroxidase and Ebselen-Mito-Peroxidase and ebselen are rapidly reduced to their active selenol forms by GSH or Trx(SH) 2 , and the selenol form degrades peroxides. To determine whether MitoPeroxidase and ebselen could degrade mitochondrial phospholipid hydroperoxides, mitochondrial membranes were incubated with lipoxidase to form phospholipid hydroperoxides (9) (Fig. 9A). MitoPeroxidase or ebselen, in conjunction with DTT to reduce the isoselenazole to a selenol, degraded these phospholipid hydroperoxides (Fig. 9A). TPMP was ineffective, even in the presence of DTT (Fig. 9A). Therefore, MitoPeroxidase and ebselen efficiently degrade phospholipid hydroperoxides, thereby protecting mitochondrial membranes from lipid peroxidation.
We next determined whether MitoPeroxidase and ebselen could degrade phospholipid hydroperoxides in intact mitochondria and thereby protect against lipid peroxidation. Incubation of mitochondria with ferrous iron and H 2 O 2 induced extensive lipid peroxidation, as indicated by the accumulation of TBARS (Fig. 9B). MitoPeroxidase prevented this TBARS accumulation (Fig. 9B), as did ebselen (Fig. 9B), whereas the lipophilic cation TPMP did not (data not shown). To test whether the membrane potential-dependent accumulation of MitoPeroxidase by mitochondria contributed to its antioxidant efficacy, we compared its antioxidant capacity in the presence and absence of a membrane potential (the respiratory inhibitor antimycin was used to abolish the membrane potential as the uncoupler FCCP prevents lipid peroxidation). In the absence of a membrane potential, MitoPeroxidase was significantly less effective at preventing lipid peroxidation (Fig. 9B). In contrast, the antioxidant efficacy of ebselen was unaffected by the presence or absence of a membrane potential. Therefore, MitoPeroxidase is an effective antioxidant within intact mitochondria, and the membrane potential-dependent accumulation of MitoPeroxidase increases its antioxidant efficacy.
To determine whether the antioxidant efficacy of MitoPer- oxidase and ebselen in intact mitochondria (Fig. 9B) depended on their reduction to the selenol by intramitochondrial GSH and Trx(SH) 2 , we preincubated mitochondria with calcium and phosphate to induce the MPT, which removes low molecular solutes such as NADPH and GSH from the mitochondrial matrix (60), thereby inactivating the intramitochondrial glutathione and thioredoxin reduction pathways. When MPT-treated mitochondria were exposed to oxidative stress they were no longer protected against lipid peroxidation by MitoPeroxidase or by ebselen, whereas control mitochondria were (Fig. 9C). These findings confirm that the protection conferred by both MitoPeroxidase and ebselen to mitochondria against peroxidemediated lipid peroxidation was dependent on intramitochondrial thiol reduction systems.
We then determined whether MitoPeroxidase and ebselen protected mitochondrial function from the consequences of oxidative damage. We used the membrane potential as a measure of mitochondrial function because it depends on the activity of the respiratory chain and on the intactness of the mitochondrial inner membrane. Lipid peroxidation decreased the mitochondrial membrane potential (Fig. 9D), whereas MitoPeroxidase protected against this loss of function, presumably by limiting oxidative damage (Fig. 9D). Ebselen up to 1 M also protected against the loss of membrane potential; however, ebselen concentrations of 5 M and above were not protective, in marked contrast to MitoPeroxidase.
To summarize, MitoPeroxidase protects mitochondria against lipid peroxidation and the damaging effects of peroxide-mediated oxidative stress on mitochondrial function. In both cases the protection required active intramitochondrial thiol reducing systems, indicating the protective effects of Mi-toPeroxidase and ebselen were due to reduction of phospholipid hydroperoxides. MitoPeroxidase was effective at protecting against loss of mitochondrial function during oxidative stress, and this protection was enhanced slightly by the membrane potential.
Protection against Apoptosis Induced by Mitochondrial Oxidative Stress-To see if ebselen and MitoPeroxidase could prevent mitochondrial oxidative stress within cells, we investigated apoptotic cell death in RBL-2H3 cells (13,33,69). Apoptosis in these cells occurs in response to increased mitochondrial oxidative stress induced by 2DG or by exogenous H 2 O 2 and is prevented by overexpressing the mitochondrial isoform of phospholipid glutathione peroxidase (7,8,70). Consequently, if ebselen or MitoPeroxidase can degrade peroxides within mitochondria in cells, then they should prevent apoptosis. Incubation of RBL-2H3 cells with 2DG induced apoptosis, as indicated by elevated caspase-3 activity after 3 h (Fig.  10A). MitoPeroxidase and ebselen blocked this caspase-3 activation, whereas TPMP was ineffective. The addition of H 2 O 2 also led to the induction of apoptosis that was decreased by MitoPeroxidase and ebselen but not by TPMP (Fig. 10B). For both H 2 O 2 and 2DG the protective effect of MitoPeroxidase against apoptosis was greater than that of ebselen. MitoPeroxidase and ebselen may prevent apoptosis by degrading peroxides or by blocking the apoptosis pathway elsewhere. To see if this is the case, we induced apoptosis by staurosporine, a protein kinase C inhibitor that activates the mitochondrial apoptotic pathway without causing oxidative damage. Staurosporine treatment led to extensive caspase-3 activation that was unaffected by MitoPeroxidase, ebselen, or TPMP (Fig.  10C). Therefore, the anti-apoptotic effects of MitoPeroxidase and ebselen are specific to their peroxidase activity.
Conclusions-We have investigated the interaction of the peroxidase mimetic ebselen with mitochondria and compared it with MitoPeroxidase, an ebselen derivative conjugated to the lipophilic TPP cation. The TPP cation of MitoPeroxidase facilitated the investigation of its reactions with peroxides and GSH by ESMS. From this we showed that the ebselen moiety is readily reduced to the related selenol by GSH. The selenol was oxidized to the isoselenazole moiety by H 2 O 2 and by the alkyl peroxide tBHP. Direct oxidation of the isoselenazole moiety to the selenoxide by H 2 O 2 also occurred, and this was reduced back to the ebselen moiety by GSH. MitoPeroxidase was an effective glutathione and thioredoxin peroxidase mimetic, with the selenol acting as the active peroxide reductant. Ebselen and MitoPeroxidase were effective antioxidants within mitochondria, being activated by the intramitochondrial glutathione and thioredoxin systems. The active selenol form then prevents phospholipid peroxidation by degrading phospholipid hydroperoxides. Within cells, both ebselen and MitoPeroxidase protected against apoptosis induced by increased mitochondrial oxidative stress. This indicates that both MitoPeroxidase and ebselen can be used to prevent peroxide damage to mitochondria within cells. MitoPeroxidase was at best only slightly more effective than ebselen in preventing oxidative damage to mitochondrial function and in preventing apoptosis. This lack of a dramatic increase in efficacy caused by conjugation to TPP was in contrast to that caused on conjugating other antioxi- In some cases the membranes were preincubated with 1 mM NEM for 10 min on ice before incubation with MitoPeroxidase. B, MitoPeroxidase binding to intact mitochondria. Liver mitochondria (1 mg of protein⅐ml Ϫ1 ) were incubated in 120 mM KCl, 10 mM HEPES, 1 mM EGTA, pH 7.2, supplemented with 10 mM succinate and 8 g⅐ml Ϫ1 rotenone with 5 M MitoPeroxidase for 10 min at 37°C with and without 1 M FCCP to abolish the membrane potential. For some incubations the mitochondria had been preincubated with 1 mM NEM for 10 min before the addition of MitoPeroxidase. C, submitochondrial localization of MitoPeroxidase binding. Mitochondria were incubated as described in B, and at the end of the incubation the mitochondria (1 mg of protein) were pelleted and separated into matrix and membrane-enriched fractions, and ϳ 20 g of protein was probed for MitoPeroxidase binding. dants such as ubiquinone or vitamin E to TPP (40,41,44,50). This was because the isoselenazole moiety of MitoPeroxidase decreased its uptake into mitochondria by its ability to form selenenylsulfide bonds with mitochondrial and cytosolic thiols. Consequently, most MitoPeroxidase is conjugated to thiols prewere preincubated in 100 mM KCl, 10 mM Tris-HCl, pH 7.6, supplemented with 10 mM succinate, 8 g⅐ml Ϫ1 rotenone with 10 mM succinate, and 8 g⅐ml Ϫ1 rotenone in the presence and absence of MitoPeroxidase or ebselen. Ferrous chloride (50 M) and H 2 O 2 (500 M) were then added, and after a 15-min incubation at 37°C TBARS formation was quantitated. Control incubations were carried out in the absence of ferrous chloride/H 2 O 2 . Data are the means Ϯ S.D. or range of 2-3 independent experiments. C, effect of MPT induction on antioxidant efficacy. Mitochondria that had undergone the MPT were compared with those that had been mock-incubated and re-isolated. Mitochondria were then exposed to oxidative stress and assessed as in B above. Data are the means Ϯ S.E. of three independent experiments. D, prevention of loss of membrane potential. Mitochondria were incubated as described for B and isolated by centrifugation, and their membrane potential was measured from the distribution of [ 3 H]TPMP while respiring on succinate. Control measurements were done on mitochondria that had (ox stress) or had not (no ox stress), been exposed to ferrous chloride/H 2 O 2 . Data are the means Ϯ S.D. of three separate mitochondrial preparations. *, p Ͻ 0.05; **, p Ͻ 0.01 compared with oxidative stress with no addition by a 2-tailed paired Student's t test. venting its accumulation into mitochondria. The bound and free MitoPeroxidase is likely to equilibrate through the reversible formation and reduction of the selenenylsulfide bonds, and this may increase the concentration of MitoPeroxidase within mitochondria and, thus, explain its greater efficacy over ebselen in mitochondria and cells. Despite the less than expected increase in efficacy on conjugation to TPP, MitoPeroxidase permeates the mitochondrial inner membrane and was taken up by mitochondria where it was activated by the intramitochondrial thiol reducing systems, thereby protecting mitochondria from oxidative damage.
The binding of MitoPeroxidase to mitochondrial thiol proteins through a selenenylsulfide bond under normal intracellular conditions enabled those proteins to be visualized on immunoblots using antiserum against the TPP moiety. This procedure may enable MitoPeroxidase to be used to probe reactive mitochondrial thiol proteins, especially under non-oxidatively stressed intracellular conditions, and may be of use in identifying reactive thiols involved in the response of mitochondria to ROS (51,54).
To summarize, by conjugation of the isoselenazole moiety to the TPP cation we have demonstrated how ebselen reacts with GSH and peroxides by mass spectrometry, shown that ebselen is activated within mitochondria by the GSH and thioredoxin systems and then protects mitochondria from oxidative damage, and demonstrated that the isoselenazole moiety within cells binds extensively to thiol proteins. We have also shown that ebselen and MitoPeroxidase may be useful tools for investigating mitochondrial oxidative damage within cells. These findings have implications for our understanding of the effect of ebselen and related molecules on mitochondrial oxidative stress.