Interactions of Mitochondria-targeted and Untargeted Ubiquinones with the Mitochondrial Respiratory Chain and Reactive Oxygen Species

Antioxidants, such as ubiquinones, are widely used in mitochondrial studies as both potential therapies and useful research tools. However, the effects of exogenous ubiquinones can be difficult to interpret because they can also be pro-oxidants or electron carriers that facilitate respiration. Recently we developed a mitochondria-targeted ubiquinone (MitoQ10) that accumulates within mitochondria. MitoQ10 has been used to prevent mitochondrial oxidative damage and to infer the involvement of mitochondrial reactive oxygen species in signaling pathways. However, uncertainties remain about the mitochondrial reduction of MitoQ10, its oxidation by the respiratory chain, and its pro-oxidant potential. Therefore, we compared MitoQ analogs of varying alkyl chain lengths (MitoQn, n = 3–15) with untargeted exogenous ubiquinones. We found that MitoQ10 could not restore respiration in ubiquinone-deficient mitochondria because oxidation of MitoQ analogs by complex III was minimal. Complex II and glycerol 3-phosphate dehydrogenase reduced MitoQ analogs, and the rate depended on chain length. Because of its rapid reduction and negligible oxidation, MitoQ10 is a more effective antioxidant against lipid peroxidation, peroxynitrite and superoxide. Paradoxically, exogenous ubiquinols also autoxidize to generate superoxide, but this requires their deprotonation in the aqueous phase. Consequently, in the presence of phospholipid bilayers, the rate of autoxidation is proportional to ubiquinol hydrophilicity. Superoxide production by MitoQ10 was insufficient to damage aconitase but did lead to hydrogen peroxide production and nitric oxide consumption, both of which may affect cell signaling pathways. Our results comprehensively describe the interaction of exogenous ubiquinones with mitochondria and have implications for their rational design and use as therapies and as research tools to probe mitochondrial function.

Mitochondria are the major site of reactive oxygen species generation (ROS) 1 within cells (1,2). When ROS production exceeds the capacity of detoxification and repair pathways, oxidative damage to protein, DNA, and phospholipid occurs, disrupting mitochondrial oxidative phosphorylation and leading to cell damage and death. This contributes to a number of human pathologies including Parkinson disease, Alzheimer disease, Friedreich ataxia, ischemia-reperfusion injury, diabetes, and aging (2,3). In addition to this pathological role, ROS can also act as redox signaling molecules (4,5). Hence, mitochondrial ROS production and oxidative damage are attractive targets for pharmacological intervention for both therapeutic and investigative purposes (6 -10).
Antioxidants have the potential to block oxidative damage and redox signaling, and exogenous ubiquinones have been widely used for this purpose in mitochondrial studies. These molecules are based on the predominant human form of endogenous ubiquinone, coenzyme Q 10 (CoQ 10 ; Fig. 1A), which is synthesized in the mitochondrial inner membrane and comprises a ubiquinone head group attached to a tail of 10 fivecarbon isoprenoid units (11). The ubiquinone moiety is redoxactive, accepting two electrons and two protons in its reduction to a ubiquinol, while the extremely hydrophobic tail ensures that within the cell it is almost exclusively associated with phospholipid bilayers (Fig. 1B). The redox activity of the ubiquinone moiety enables it to act as a mobile electron carrier in the mitochondrial inner membrane where it is reduced to a ubiquinol by several membrane bound dehydrogenases and oxidized back to a ubiquinone by complex III. Furthermore, the reduced ubiquinol form of CoQ 10 has an important protective function as a chain breaking antioxidant, terminating lipid peroxidation in phospholipid bilayers (12,13). Therefore, ubiquinone supplementation is an attractive therapeutic strategy in human pathology, as it could both stimulate oxidative phosphorylation by complementing any defects in respiration (14) and protect against oxidative damage (15). However, this duality complicates experimental interpretation, as any effects of CoQ 10 can result from its interaction with oxidative phosphorylation, oxidative damage, or redox signaling pathways.
While positive therapeutic effects have been observed with CoQ 10 supplementation in humans (16 -19), its oral bioavail- FIG. 1. Ubiquinones used in this study. A, structures. B, octan-1-ol/PBS partition coefficients: a, calculated using Advanced Chemistry Development (ACD) Software Solaris v4.67 as described in Smith et al. (31); b, determined experimentally in this study; c, determined experimentally in Kelso et al. (29); d, determined experimentally in Asin-Cayuela et al. (30). C, a schematic of MitoQ 10 in a phospholipid bilayer composed of C18 saturated fatty acids. The TPP cation and the alkyl chain favor the polar surface and hydrophobic core, respectively. The preferred position of the uncharged but moderately hydrophilic ubiquinone moiety is less clear. ability is poor due to its extreme hydrophobicity (Fig. 1B). Consequently, only a small fraction of orally administered CoQ 10 reaches the circulatory system, and augmentation of mitochondrial CoQ 10 content is lower still (20 -22). Therefore the beneficial effects of exogenous CoQ 10 require high doses and long term administration, and only subjects whose CoQ 10 levels have been depleted by defective synthesis, age, or disease are responsive (18, 20 -24). The negligible water solubility of CoQ 10 and its poor diffusion to mitochondria in cultured cells also hinder its usefulness as a tool to study mitochondrial oxidative damage and redox signaling in vitro. As a result there is considerable interest in developing artificial ubiquinones with better bioavailability and pharmacokinetic properties. Idebenone is one such compound and it comprises a ubiquinone head group attached to a ten carbon alkyl tail with a terminal hydroxyl (Fig. 1A) (25). Clinical trials with idebenone have shown it can ameliorate cardiomyopathy in Friedreich ataxia patients (26 -28).
Although decreasing hydrophobicity improves overall bioavailability, it would also be of benefit to target ubiquinones specifically to mitochondria, as they are the main site of ubiquinone utilization but represent only a small fraction of the cell volume. Lipophilic cations, such as methyltriphenylphosphonium (TPMP; Fig. 1A), are accumulated several hundred-fold by the large membrane potential (negative inside) generated by mitochondria during oxidative phosphorylation (6,9). We have exploited this property by covalently attaching a ubiquinone moiety to the lipophilic triphenylphosphonium (TPP) cation generating a mitochondria-targeted ubiquinone (MitoQ 10 ; Fig.  1A), which is selectively accumulated within isolated mitochondria, and within mitochondria in cells and in vivo (9, 29 -31). The interaction of amphipathic alkyltriphenylphosphonium cations with phospholipid bilayers occurs as follows: the TPP lipophilic cation is bound as a monolayer in a potential well at about the level of the phospholipid fatty acid carbonyls, while the hydrophobic alkyl chain is inserted into the hydrophobic core of the membrane (31)(32)(33)(34). This is illustrated for MitoQ 10 in Fig. 1C. Although the large ionic radius and hydrophobicity of the TPP cation allows molecules such as MitoQ 10 to permeate phospholipid bilayers readily, their steady-state concentration within the hydrophobic core of the membrane is low. Furthermore, within energized mitochondria the membrane potential causes most MitoQ 10 to be adsorbed to the matrix surface of the inner membrane. MitoQ 10 is a particularly effective antioxidant against lipid peroxidation (29,30) and has been used in a range of studies of mitochondrial dysfunction and oxidative stress where its mitochondrial localization has enabled the site of intracellular redox signaling to be inferred (35)(36)(37)(38)(39)(40)(41)(42). However, the details of the mitochondrial processes affected by MitoQ 10 remain unclear because it could act as an antioxidant (thereby blocking oxidative damage and redox signaling) or as an electron carrier in the respiratory chain (thereby stimulating oxidative phosphorylation). A further consideration is that, while ubiquinols are potent antioxidants, there are also partially reduced and/or partially protonated intermediate forms that can act as pro-oxidants through interacting with oxygen to form superoxide (O 2 . ) (13,43,44). The implications of autoxidation are unclear as stress response pathways that boost antioxidant defenses are often switched on by mild oxidative stress (a process termed hormesis). As a result autoxidation could paradoxically contribute to conditioning and protection against subsequent oxidative stress (4,45,46). In summary, ubiquinone augmentation is an attractive therapy for degenerative diseases as it has the potential to both stimulate oxidative phosphorylation and protect against lipid peroxidation. However this combination of effects makes inter-pretation difficult, limiting the design of rational therapies. Furthermore, while MitoQ 10 has been widely used to infer the existence of mitochondrial ROS-dependent signaling pathways, aspects of its mechanism of action remain uncertain. To clarify these issues, we set out to determine how MitoQ 10 and a number of related short-chain exogenous ubiquinones interact with both the mitochondrial respiratory chain and ROS. For this we used a range of MitoQ analogs that differ in the number of carbons linking the ubiquinone to the TPP moiety (MitoQ n ; n ϭ 3, 5, 10, or 15 CH 2 groups) (29,30) and compared their interactions with those of the untargeted short-chain ubiquinone analogs, CoQ 2 , decylQ, and idebenone (Fig. 1A). This work has led to a better understanding of how MitoQ and other exogenous ubiquinones interact with the respiratory chain and ROS and has considerable implications for their rational design and use as therapies and as tools to probe mitochondrial oxidative damage and redox signaling.

MATERIALS AND METHODS
Yeast Incubations-The Saccharomyces cerevisiae strains used were WT (CEN.PK2-1C MATa his3 leu2 trp1 ura3) and ⌬COQ2 (CEN.PK2-1C coq2::HIS3), kindly supplied by Prof. Catherine Clarke, UCLA (47). Coq2 codes for the enzyme para-hydroxybenzoate:hexaprenyl transferase, which catalyzes the transfer of the polyisoprenoid chain to 4-hydroxybenzoic acid in CoQ biosynthesis. ⌬COQ2 is auxotrophic for CoQ and fails to grow on non-fermentable carbon sources, such as glycerol. Yeast were cultured in 10 ml of YPG (1% (w/v) yeast extract, 2% (w/v) peptone, 3% (v/v) glycerol). The initial cell density was adjusted to A 600 ϳ0.1. To achieve a reproducible transition to a respiratory phenotype, the medium was supplemented with 0.05% (w/v) glucose. This allowed fermentative growth of ⌬COQ2 up to A 600 ϳ0.4 after which growth rapidly ceased unless ubiquinone supplementation restored oxidative phosphorylation. Yeast were incubated in the dark at 30°C with mechanical shaking at 250 rpm. Growth was monitored spectrophotometrically by measuring A 600 over 120 h. For all yeast experiments, ubiquinones and other hydrophobic compounds were added in Me 2 SO to 1% (v/v) of the total culture volume, which did not affect the growth of the WT and ⌬COQ2 strains on glucose (data not shown).
⌬COQ2 yeast cultures for mitochondrial isolation were grown aerobically at 30°C to mid-logarithmic phase (A 600 ϳ1) in lactate medium (2% (v/v) DL-lactic acid, 0.3% yeast extract, 0.2% glucose, 0.05% CaCl 2 ⅐2H 2 O, 0.05% NaCl, 0.06% MgCl 2 ⅐6H 2 O, 0.1% KH 2 PO 4 , 0.1% NH 4 Cl (all w/v) (pH 5.5, NaOH). ⌬COQ2 yeast can use D-lactate as a respiratory substrate because it donates electrons to oxidative phosphorylation at complex IV via the reduction of cytochrome c (cyt c). For WT yeast, the level of glucose was kept at 0.05% (w/v). Mitochondria were isolated according to published protocols (48,49). The protein concentration was measured by the bicinchoninic acid assay using BSA as a standard (50). Aliquots of the mitochondrial preparation were mixed with 10 mg⅐ml Ϫ1 fatty acid-free BSA as a cryoprotectant, snap-frozen on dry ice, and stored at Ϫ80°C. Upon thawing the mitochondria retained a membrane potential that was indistinguishable from that of freshly isolated yeast mitochondria as confirmed by the uncoupler-sensitive uptake of [ 3 H]TPMP (data not shown) (29).
Oxygen consumption was measured with a Clark-type oxygen electrode (Rank Brothers, Bottisham, Cambridge, UK) in a 1-ml stirred chamber at 30°C. Aliquots of frozen ⌬COQ2 yeast mitochondria were thawed rapidly, washed, and resuspended in mannitol buffer (0.6 M mannitol, 10 mM Tris maleate, 5 mM KP i , 0.5 mM EDTA (pH 6.8, KOH)) at 0.2 mg protein⅐ml Ϫ1 . The mitochondria were energized with 5 mM glycerol 3-phosphate (G3P) and uncoupled with 1 M carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP). Ubiquinones (1-20 M in Me 2 SO) were titrated in successively, followed by 1 M myxothiazol to determine non-mitochondrial oxygen consumption. For some experiments mitochondria were sonicated (3 ϫ 5 s, setting 4; Misonix XL-2020 with microtip) in an ice bath prior to the measurement of respiration and addition of ubiquinone. Uptake of MitoQ analogs by yeast mitochondria was measured using an electrode selective for TPP cations (30). For these experiments WT yeast mitochondria (0.4 mg protein⅐ml Ϫ1 ) were incubated in 2.5 ml of mannitol buffer in the presence of 2 M MitoQ analog at 30°C, and the uptake of MitoQ in response to energization with 5 mM ethanol was measured.
Mammalian Mitochondrial Preparations-Bovine heart mitochondrial membranes (BHM) were prepared from isolated bovine heart mitochondria as described previously (51,52). Rat liver mitochondria were prepared by homogenization followed by differential centrifugation as described previously (53). Rat heart mitochondria were prepared by tissue disruption using an Ultra-Turrax (5 s), followed by differential centrifugation as described previously (53). Protein concentration was determined using the biuret assay with BSA as a standard (54).
Ubiquinone Reduction and Oxidation by Respiratory Complexes-Assays were based on previously described methods for measuring respiratory complex activity (55). However, for all assays except complex III the spectrophotometric decrease in A 275 (at this wavelength ubiquinone absorbs more strongly than ubiquinol) was monitored instead. Assays were performed in KP i buffer (50 mM KP i -KOH, 100 M EDTA, 100 M diethylenetriaminepentaacetic acid (pH 7.8) unless stated otherwise) at 37°C. For complex I, the buffer was supplemented with 100 g protein⅐ml Ϫ1 BHM, 100 M NADH, and 2 mM KCN, and the reaction was started by addition of 50 M ubiquinone. For complex II, the buffer was supplemented with 100 g protein⅐ml Ϫ1 BHM, 5 mM succinate, 8 g⅐ml Ϫ1 rotenone, and 2 mM KCN, and the reaction was started by addition of 50 M ubiquinone. For glycerol-3-phosphate dehydrogenase (G3PDH), EDTA and diethylenetriaminepentaacetic acid (DTPA) were omitted as Ca 2ϩ may be required for activity (56), and the buffer was supplemented with 200 g protein⅐ml Ϫ1 BHM, 2 mM KCN, and 50 M ubiquinone. The reaction was started with the addition of 10 mM G3P. Rotenone (8 g⅐ml Ϫ1 ) and malonate (20 mM) did not affect the rate of ubiquinone reduction in the presence of G3P. For complex III, the buffer was supplemented with 50 g protein⅐ml Ϫ1 BHM, 50 M bovine cyt c, 8 g⅐ml Ϫ1 rotenone, and 2 mM KCN, plus or minus 400 nM myxothiazol. The reaction was started with the addition of 50 M ubiquinol, and cyt c reduction was measured by an increase at A 550 (55). The myxothiazol-insensitive rate was measured in parallel and subtracted as there is a significant non-enzymatic rate of cyt c reduction by ubiquinols (43).
Measurement of the Ubiquinone Redox State-Spectrophotometric measurements were made at 275 nm. The ubiquinone redox state was measured at 37°C in KP i buffer supplemented with 100 g protein⅐ml Ϫ1 BHM, 8 g⅐ml Ϫ1 rotenone, and 10 M ubiquinone. BHM were also supplemented with 5 M bovine cyt c, as this can be lost during membrane isolation. For succinate (5 mM) oxidation, fumarase (5 units⅐ml Ϫ1 ) was also added as fumarate absorbs at A 275 (local maximum at 240 nm). Furthermore, a background trace in the absence of ubiquinone was subtracted as fumarase does not completely abolish fumarate accumulation. For NADH oxidation, NADH absorption interfered with ubiquinone redox changes at A 275 , therefore an NADH regeneration system (5 mM lactate and 2 units⅐ml Ϫ1 lactate dehydrogenase) was used. Nucleotides were added as NAD ϩ (50 M) because unlike NADH, it rapidly reached a steady-state NADH/NAD ϩ ratio. Further additions of lactate dehydrogenase had no effect on this ratio. A background trace acquired in the absence of ubiquinone was subtracted. For HPLC measurements, the incubation was as above, but fumarase was omitted. Two min after the addition of succinate, 1 volume of dichloromethane:diethylether (1:2) was added, and the mixture was then vortexed under argon. The phases were separated by centrifugation (1 min at 1,000 ϫ g), and the upper organic phase was removed to a test tube overgassed with N 2 . The lower aqueous phase was re-extracted as before with dichloromethane:diethylether (1:2), and the second organic phase was combined with the first. This was evaporated to dryness under a flow of N 2 , then resuspended in 1 ml of 45% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid. Samples were analyzed by reverse-phase HPLC (Gilson 321 pump, 1 ml⅐min Ϫ1 ) on a C18 column (Jupiter 300 Å, Phenomenex), using a staged gradient (4.5% acetonitrile for 5 min, 4.5-54% acetonitrile over 5 min, 54 -72% acetonitrile over 20 min, 72-90% acetonitrile over 5 min) containing 0.1% (v/v) trifluoroacetic acid and detected at 220 nm using a Gilson UV/VIS 151 spectrophotometer. Peaks were assigned by comparison with ubiquinone or ubiquinol standards. Spectrophotometric measurements of ubiquinone in rat liver mitochondria were made at 275 nm. Mitochondria (200 g protein⅐ml Ϫ1 ) were incubated at 25°C in a stirred cuvette containing 250 mM sucrose, 5 mM Tris, 1 mM EGTA (pH 7.4, HCl) supplemented with 8 g⅐ml Ϫ1 rotenone and 5 M ubiquinone. 5 mM succinate, 400 nM FCCP, and 400 nM myxothiazol were added as indicated.
Ubiquinol Oxidation by Peroxynitrite-Ubiquinol oxidation by ONOO Ϫ was measured at 37°C in KP i buffer supplemented with 100 g protein⅐ml Ϫ1 BHM, 8 g⅐ml Ϫ1 rotenone, 5 M cyt c, and 10 M ubiquinone. 5 mM succinate was added to energize the BHM, and two additions of 50 M ONOO Ϫ were made. This was followed by 20 mM malonate and a further addition of 50 M ONOO Ϫ . Ubiquinone redox changes were measured at 275 nm. ONOO Ϫ has a strong absorbance maximum at 302 nm (⑀ 302 ϭ 1.67 mM Ϫ1 ⅐cm Ϫ1 (57)) leading to a transient spike in A 275 until it has decayed. The decay back to base-line absorb-ance took ϳ8 s in KP i buffer. For cis-parinaric acid (PA; Molecular Probes, Eugene, OR) oxidation, the incubation was performed using a Shimadzu RF-5301PC fluorimeter in a stirred 3-ml cuvette thermostatted at 37°C. PA was excited at 324 nm, and its fluorescence was monitored at 413 nm. The assay was in KP i buffer supplemented with 100 g protein⅐ml Ϫ1 BHM, 8 g⅐ml Ϫ1 rotenone, 5 M cyt c, and 10 M ubiquinone. BHM were supplied with 5 mM succinate, then after 1 min 2 M PA was added, followed by 20 M ONOO Ϫ and each minute thereafter. ONOO Ϫ was prepared as described previously (58).
Measurement of Reactive Oxygen Species and Autoxidation-Reduced MitoQ 10 was prepared as described previously (29) Aconitase inactivation within mitochondria was measured spectrophotometrically by a coupled enzyme assay linking isocitrate production by aconitase to NADP ϩ reduction by isocitrate dehydrogenase (⑀ 340 NADPH ϭ 6.22 mM Ϫ1 ⅐cm Ϫ1 ) (59,60). Aliquots of frozen WT (EG103 MAT␣ his3 leu2 trp1 ura3) yeast mitochondria (61) were thawed rapidly, then washed in 0.6 M mannitol buffer and resuspended in this buffer at 0.3 mg of protein/ml. The mitochondria were incubated in the presence of 5 mM G3P for 30 min in a shaking water bath at 30°C. Samples were removed every 5 min, snap-frozen on dry ice, and thawed prior to assaying. Aconitase activity was assayed in a 96-well plate with a 10-l sample added to 190 l of assay buffer (50 mM Tris-HCl, 0.6 mM MnCl 2 , 5 mM sodium citrate, 0.2 mM NADP ϩ , 0.1% v/v Triton X-100, and 0.4 unit⅐ml Ϫ1 isocitrate dehydrogenase (pH 7.4)) at 30°C. A 340 readings were carried out at 15 s intervals over 7 min in an EL x 808 Ultra Microplate Reader (Bio-Tek Instruments). Each time point was measured in quintuplicate, and plotting the natural logarithm of activity versus time linearized the time course of aconitase inactivation. The slope of the line corresponded to the pseudo-first order rate constant of aconitase inactivation. The background rate of NADPH formation was determined in the presence of fluorocitrate (100 M), a competitive inhibitor of aconitase, and was always less than 10% of the initial rate.
Efflux of H 2 O 2 from isolated rat heart mitochondria was measured using a Shimadzu RF-5301PC fluorimeter in a stirred 3-ml cuvette thermostatted at 25 or 37°C. Mitochondria (200 g protein⅐ml Ϫ1 ) were incubated in 120 mM KCl, 3 mM HEPES-KOH, 1 mM EGTA, and 0.01% (w/v) fatty acid-free BSA (pH 7.2), containing 4 units⅐ml Ϫ1 horseradish peroxidase, 50 M Amplex Red (Molecular Probes) and 100 units⅐ml Ϫ1 SOD. 5 mM succinate and 8 g⅐ml Ϫ1 rotenone were added as indicated. Amplex Red was excited at 560 nm, and its fluorescence was monitored at 590 nm.
Partition Coefficients-Octan-1-ol/PBS partition coefficients (the concentration of ubiquinone in octan-1-ol relative to the concentration in PBS) were determined as described previously (30). Membrane/PBS partition coefficients (the concentration of ubiquinone in membranes relative to the concentration in PBS) were estimated by incubating 2 ml of PBS containing 250 g protein⅐ml Ϫ1 BHM and 50 M ubiquinone for 5 min at 37°C. BHM were pelleted by centrifugation (30 min at 16,000 ϫ g), after which the supernatant was removed to a test tube and back-extracted with 1 volume of octan-1-ol. One extraction was sufficient for all ubiquinones except MitoQ 5 and MitoQ 3 , which were extracted with 2 ml of octan-1-ol two and three times, respectively. The pellet was fully aspirated to remove as much water as possible, and then all ubiquinones were extracted four times with 100 l of octan-1-ol. To estimate the partition coefficient, the inner membrane surface area per unit volume of rat heart mitochondria (61 m 2 ⅐m 3 ) was used along with values for membrane thickness (6 nm) and mitochondrial volume (0.6 l⅐mg protein Ϫ1 ) (62). This gave an estimated membrane volume of 0.22 l⅐mg protein Ϫ1 .
Thiol Oxidation-Thiol oxidation by H 2 O 2 was measured at 37°C in KP i buffer containing 10 mM glucose, 500 M GSH, 200 M NADPH, 0.4 unit⅐ml Ϫ1 glutathione reductase, and 0.02 unit⅐ml Ϫ1 glutathione peroxidase. GSH oxidation was detected by glutathione reductase-dependent NADPH oxidation. After A 340 stabilized, 0.006 unit⅐ml Ϫ1 glucose oxidase was added to generate a stable flux of H 2 O 2 . Subsequently ubiquinone or NaBH 4 -reduced ubiquinol (50 M) was added. The response was linear with H 2 O 2 over the range 1-100 M and with glucose oxidase from 0.003 to 0.03 unit⅐ml Ϫ1 . The rate of NADPH oxidation was low in the absence of glutathione peroxidase and glutathione reductase. Glutathione peroxidase activity (0.02 unit⅐ml Ϫ1 ) was in excess of glucose oxidase activity, yet kept to a minimum to allow for reaction of H 2 O 2 with MitoQ 10 and CoQ 2 .
For opening of the mitochondrial permeability transition pore (PTP) by t-butylhydroperoxide (tBHP) in isolated rat liver mitochondria, the final centrifugation during mitochondrial isolation was in 250 mM sucrose, 5 mM Tris (pH 7.4, HCl). Measurements of extra-mitochondrial Ca 2ϩ were made in a stirred 3-ml cuvette thermostatted at 25°C using a Shimadzu RF-5301PC fluorimeter. Calcium green-5N (Molecular Probes), a membrane-impermeant Ca 2ϩ probe, was excited at 506 nm, and emission was monitored at 532 nm. Fluorescence was measured in 250 mM sucrose, 5 mM Tris, 10 M EGTA (pH 7.4, HCl) supplemented with 0.5 mg protein⅐ml Ϫ1 rat liver mitochondria, 8 g⅐ml Ϫ1 rotenone, and 1 M calcium-green-5N. Mitochondria were supplied with 5 mM succinate, and after 1 min 5 M CaCl 2 Ϯ 5 M tBHP were added. 1-5 M MitoQ 10 and/or 500 nM cyclosporin A (CsA) were also added to some incubations.
NMR-Reduced MitoQ 10 was prepared by reaction with NaBH 4 in methanol under argon. Excess NaBH 4 was quenched with 10% (v/v) methane sulfonic acid, and the resulting mixture was extracted with dichloromethane, washed with water, and the solvent evaporated in vacuo. The 1 H NMR spectra of the product indicated it was ϳ90% reduced. Solutions of the ubiquinol (10 mM) in D 2 O Ϯ 10 mM H 2 O 2 and/or 100 M EDTA were prepared in 5-mm NMR tubes. Oxidation of reduced MitoQ 10 was monitored using 1 H NMR of the ring methoxy peaks at 3.9 and 4 ppm for the ubiquinol and ubiquinone, respectively. This showed that 1 day exposed to air oxidized 3-6% of the ubiquinol, and this was identical in the presence or absence of H 2 O 2 .
Visualization of Respiratory Complexes-Structural files of complex II from Escherichia coli (Protein Data Bank code 1NEK) (63) and complex III from Bos taurus (Protein Data Bank code 1NTZ) (64) were visualized with PyMOL (DeLano Scientific, San Carlos, CA). Structural files for MitoQ analogs were generated with CORINA (Computer Chemistry Center, University of Erlangen-Nuremberg, Erlangen, Germany). To position MitoQ derivatives relative to the enzymes, three carbon atoms from the ring of the ubiquinone head group were pair-fitted with three corresponding atoms from the ubiquinone or inhibitor bound in their active sites.

Mitochondrial Respiration in CoQ-deficient Yeast Is Restored by Untargeted Exogenous Ubiquinones but Not by MitoQ 10 -
The first question we addressed was whether MitoQ 10 could act as an effective electron carrier in the mitochondrial respiratory chain. For this we used yeast that cannot synthesize CoQ (⌬COQ2); consequently oxidative phosphorylation is inactive, and this yeast strain cannot grow on non-fermentable media (47). This lack of growth could be rescued by supplementation with exogenous CoQ 6 (the endogenous form of CoQ in yeast) ( Fig. 2A). Although supplemented growth was slower than for the WT strain, the maximum cell density achieved was similar. Therefore exogenous ubiquinones can reach mitochondria within cells and restore oxidative phosphorylation. The effectiveness of various CoQ analogs at restoring growth exhibited a dependence on concentration and on the length of the isoprenoid tail, with CoQ 6 being the most effective (Fig. 2B). Interestingly, ubiquinone analogs with a saturated alkyl tail did not restore cell growth (Fig. 2C). MitoQ 10 at concentrations of 100 nM to 10 M also failed to stimulate cell growth. WT yeast growth was inhibited at 10 M MitoQ 10 , but as similar growth inhibition was observed with 10 M decyl-TPP, but not with 10 M TPMP, toxicity is probably due to the higher local concentration of more hydrophobic TPP cations within phospholipid bilayers (data not shown). CoQ 10 , the predominant CoQ in humans, did not support cell growth in non-fermentable media at 5 and 50 M (data not shown). However, this was caused by poor uptake of CoQ 10 probably due to its insolubility, as CoQdeficient yeast engineered to synthesize CoQ 10 grow well on non-fermentable substrates (65).
It is possible that MitoQ 10 , decylQ, and idebenone are actually effective respiratory substrates but that their mitochondrial accumulation by yeast is poor. Therefore we determined whether MitoQ 10 , decylQ, idebenone, and the CoQ analogs could stimulate uncoupled respiration by mitochondria isolated from ⌬COQ2 yeast. CoQ 2 and decylQ were the most effective at restoring respiration, followed by idebenone and CoQ 1 (Fig.  2D). Therefore the failure of decylQ and idebenone to complement the respiratory defect in intact ⌬COQ2 yeast (Fig. 2C) was a result of insufficient mitochondrial accumulation. In contrast, MitoQ 10 was still ineffective at stimulating respiration in isolated ⌬COQ2 mitochondria, even though it was rapidly accumulated by energized yeast mitochondria (data not shown). This confirms that MitoQ 10 is an ineffective electron carrier for respiration and explains its failure to restore yeast growth in non-fermentable media. Surprisingly, CoQ 6 was also ineffective at restoring respiration, despite being able to restore growth. However, this was due to its slow uptake by isolated mitochondria as respiration more than doubled (ϩ110% versus intact mitochondria with CoQ 6 ) if mitochondria were sonicated before CoQ 6 additions were made (data not shown). Sonication did not increase respiration with CoQ 0 or MitoQ 10 . Therefore, CoQ 6 is presumably too hydrophobic to diffuse rapidly through the mitochondrial outer membrane and stimulate respiration in these short term experiments. In summary we observe three classes of exogenous ubiquinone interaction: CoQ 1 , CoQ 2 , and CoQ 6 , which all migrate to mitochondria in intact yeast and restore respiration; decylQ and idebenone, which can restore respiration but do not accumulate sufficiently in mitochondria in yeast cells; and MitoQ 10 , which fails to restore respiration even after it accumulates in mitochondria. We next focussed on determining why MitoQ 10 was unable to complement the respiratory defect in the CoQ-deficient mitochondria.
MitoQ Analogs Are Not Oxidized by Complex III but Are Reduced by Complex II and Glycerol-3-Phosphate Dehydrogenase-The failure of MitoQ 10 to stimulate respiration in CoQdeficient mitochondria was general to both yeast and mammals, as it did not complement respiration in CoQ-deficient human fibroblasts, which could be rescued by decylQ. 2 Therefore MitoQ 10 is either poorly oxidized by complex III or poorly reduced by mitochondrial ubiquinone reductases (Fig. 3A). To find out which we determined whether the reduced form of MitoQ 10 and other short-chain ubiquinols were oxidized by complex III in bovine heart mitochondrial membranes (BHM). The ubiquinol form of MitoQ 10 was an ineffective substrate for complex III, while those of CoQ 2 , decylQ, and idebenone were all rapidly oxidized (Fig. 3B). We then measured how effectively the oxidized forms of MitoQ 10 CoQ 2 , decylQ, and idebenone were reduced by complex I, complex II, and G3PDH. MitoQ 10 was an ineffective substrate for complex I (Fig. 3C) but was well reduced by complex II (Fig. 3D) and G3PDH (Fig. 3E). In contrast, CoQ 2 , decylQ, and idebenone were effective substrates for all three ubiquinone reductases (Fig. 3, C-E).
We next determined how the carbon chain length between the TPP and ubiquinone moieties of MitoQ analogs affected their interaction with the respiratory chain (Fig. 3, B-E). None of the MitoQ analogs reacted effectively with complexes I and III. In contrast, MitoQ analogs did react with complex II and G3PDH in a manner that was sensitive to alkyl chain length, with the reduction rate slowing as the chain length decreased from 10 to 3 carbons. Although increasing the carbon chain length to 15 led to an apparent decrease in reduction rate relative to MitoQ 10 , this may be an artifact due to the low solubility of MitoQ 15 as reduction of MitoQ 15 at a lower concentration (10 M rather than 50 M) by complex II is comparable with MitoQ 10 (Fig. 4A).
Reduction of MitoQ 10 by succinate is via complex II as it is completely inhibited by the competitive inhibitor malonate and other electron carriers, such as O 2 . , do not mediate it as reduction occurred at a similar rate under anaerobic conditions and in the presence of SOD (data not shown). Direct electron transfer from the reduced endogenous CoQ 10 pool is also unlikely to contribute to the reduction of MitoQ 10 as electron transfer between a ubiquinol and a ubiquinone occurs by sequential deprotonation/electron transfer reactions that cannot occur within the phospholipid bilayer (66). Therefore we conclude that the predominant sources of electrons for MitoQ 10 in mitochondria are the active sites of ubiquinone reductases. In summary, MitoQ analogs cannot complement defects in respiration because they are poorly oxidized by complex III. The longer chain MitoQ analogs are extensively reduced by complex II and G3PDH but not by complex I. MitoQ 10 Remains Reduced under Conditions Where CoQ 2 , DecylQ, and Idebenone Are Oxidized-In addition to transferring electrons in oxidative phosphorylation, the ubiquinol form of CoQ 10 also acts as a chain breaking antioxidant in lipid peroxidation through donation of a hydrogen atom to a carbon or oxygen-centered radical (12). Therefore if an exogenously added ubiquinone is to be an effective antioxidant, its redox state is critical. The poor reactivity of MitoQ 10 with complex III FIG. 2. Utilization of ubiquinone analogs by CoQ-deficient (⌬COQ2) yeast. Yeast growth was monitored by following A 600 over time. Cultures were prepared in YPG ϩ 0.05% glucose, and the initial cell density was adjusted to A 600 ϳ0.1. Supplementation of the glycerol medium with 0.05% glucose allows fermentative growth of ⌬COQ2 yeast to A 600 ϳ0.4 (indicated by the horizontal line in B and C). A, growth of ⌬COQ2 yeast can be restored by supplementation with CoQ 6 . a-c, growth curves for WT yeast (a), ⌬COQ2 yeast (b), and ⌬COQ2 yeast supplemented with 50 M CoQ 6 (c). Traces are means Ϯ ranges of duplicate determinations for a typical experiment, which was repeated three times. B, effect of CoQ isoprenoid chain length on growth of ⌬COQ2 yeast. A 600 was measured after 120 h, by which time maximum cell density had been reached for all growth conditions. The values are the means Ϯ ranges of two independent experiments. C, MitoQ and non-isoprenoid untargeted ubiquinones are ineffective at restoring growth in ⌬COQ2 yeast. Conditions are the same as described for B. D, MitoQ 10 cannot restore respiration in mitochondria isolated from ⌬COQ2 yeast. Mitochondria from ⌬COQ2 yeast were incubated with 5 mM G3P and 1 M FCCP. a-g, the myxothiazol-insensitive respiration rate determined in the presence of the indicated concentrations of CoQ 2 (a), decylQ (b), idebenone (c), CoQ 1 (d), MitoQ 10 (e), CoQ 6 (f), and CoQ 0 (g). Data are the means Ϯ S.D. of three independent experiments.  implies that MitoQ 10 may be persistently reduced and consequently a better antioxidant than CoQ 2 , decylQ, or idebenone. To investigate this, we measured the steady-state ubiquinone/ ubiquinol ratio for exogenous ubiquinones in the presence of BHM respiring on succinate (Fig. 4A). Under these conditions CoQ 2 remained largely oxidized (Fig. 4A, trace a), as did decylQ and idebenone (data not shown). In contrast, MitoQ 10 and MitoQ 15 (Fig. 4A, traces d and e) were rapidly reduced. MitoQ 3 and MitoQ 5 (Fig. 4A, traces b and c) were slowly reduced, consistent with their lower rate of reduction by complex II (Fig.  3D). For CoQ 2 and MitoQ 10 , the relative amounts of ubiquinol and ubiquinone were determined by HPLC (Fig. 4, B and C), and the ratios of their peak areas are shown in Fig. 4D (open bars, CoQ 2 ; filled bars, MitoQ 10 ). CoQ 2 and MitoQ 10 were both largely in the oxidized form on incubation with BHM alone, but on addition of substrate, CoQ 2 remained oxidized, while MitoQ 10 was reduced to its ubiquinol form. The complex III-inhibitor myxothiazol caused CoQ 2 to become reduced but had no effect on MitoQ 10 as it was already in the ubiquinol form. The ratio of reduced to oxidized idebenone was qualitatively similar to CoQ 2 (data not shown), but co-eluting peaks present in BHM alone prevented precise quantification at 220 nm by HPLC.
That CoQ 2 , decylQ, and idebenone were all largely oxidized on incubation with mitochondrial membranes was somewhat unexpected as endogenous CoQ 10 is ϳ75-90% reduced in isolated mitochondria during State 4, dropping to 50 -60% reduced in State 3 (67,68). To determine whether this was specific to using succinate as an electron donor, we used NADH to drive reduction through complex I. Under these conditions, MitoQ 3 and MitoQ 5 , as well as CoQ 2 , decylQ, and idebenone, remained predominantly in the oxidized form (Fig. 4E), while MitoQ 10 and MitoQ 15 were slightly reduced (Fig. 4E, traces d  and e). Ubiquinone reduction by NADH was possible as the respiratory inhibitor cyanide led to the rapid reduction of CoQ 2 , decylQ, and idebenone and to the gradual reduction of the MitoQ analogs, consistent with their slow reduction by complex I. Therefore the relatively oxidized steady state of the exogenous untargeted ubiquinones is independent of the electron donor and is determined by the relative rates of electron entry to and efflux from the ubiquinone pool.
We next investigated whether the membrane potential in intact rat liver mitochondria largely prevented the oxidation of exogenous ubiquinones by complex III. In contrast to BHM, CoQ 2 , idebenone, and MitoQ 10 were all largely reduced in mitochondria energized with succinate (Fig. 4F). When the membrane potential was collapsed with the uncoupler FCCP, CoQ 2 and idebenone rapidly became oxidized but were re-reduced on addition of myxothiazol. DecylQ behaved in a manner similar to CoQ 2 and idebenone (data not shown). Therefore the reduced state of exogenous ubiquinones in coupled mitochondria was due to the membrane potential slowing their oxidation by complex III. This behavior was in marked contrast to MitoQ 10 as its negligible oxidation by complex III meant that it remained reduced even when the mitochondria were uncoupled. Interestingly, a partial decrease in the membrane potential upon addition of ADP (State 3) did not lead to large scale oxidation of CoQ 2 (data not shown), thus ATP synthesis does not cause extensive oxidation of exogenous untargeted ubiquinones, but complete collapse of the membrane potential does.
In summary, the equilibrium redox state of an exogenous ubiquinone is determined by its relative rates of reduction and oxidation (Fig. 3A). The oxidation of MitoQ 10 by complex III is negligible, while it is rapidly reduced by complex II, hence MitoQ 10 is fully reduced under most conditions. This is not the case for CoQ 2 , decylQ, and idebenone: although they are fully reduced in coupled mitochondria, they are rapidly oxidized by complex III when the membrane potential is low. This may be critical during pathological conditions where depolarization occurs such as during ischemic injury or following induction of the PTP.
The Ineffective Oxidation of MitoQ 10 by Complex III Enhances Antioxidant Protection against Peroxynitrite-The antioxidant efficacy of MitoQ 10 is due to its conversion to a ubiquinol, as the ubiquinone is inactive (29,30). To see if the greater tendency of MitoQ 10 to remain in the reduced form enhanced its antioxidant ability, we examined its interaction with the biologically significant oxidant peroxynitrite (ONOO Ϫ ), which is produced in vivo by the reaction of O 2 . with nitric oxide (NO ⅐ ) (69). Among the oxidizing reactions of ONOO Ϫ is the one electron oxidation of ubiquinol to a ubisemiquinone radical which then dismutates (70) (Reactions 1 and 2).

REACTIONS 1 AND 2
An advantage of using ONOO Ϫ as the ubiquinol oxidant is that at pH 7 it decays within a few seconds to unreactive end products (69), leading to a pulse of ubiquinol oxidation after which regeneration of its antioxidant function by re-reduction can be assessed. CoQ 2 incubated with BHM respiring on succinate remained in its ubiquinone form (Fig. 5A, trace a), while MitoQ 10 was reduced by the respiratory chain (Fig. 5A, trace b). MitoQ 5 was also reduced but to a lesser extent due to its slower reaction with complex II (Fig. 5A, trace c). Addition of ONOO Ϫ led to a sharp upward spike in A 275 for all three ubiquinones. For CoQ 2 (Fig. 5A, trace a) (57)), which decayed away over ϳ8 s. For MitoQ 5 (Fig. 5A, trace c) there was a transient spike in A 275 due to both ONOO Ϫ itself and the formation of oxidized MitoQ 5 . After the ONOO Ϫ had decayed away, A 275 decreased as the ubiquinone was slowly reduced back to the ubiquinol. For MitoQ 10 , there was also a dramatic spike in A 275 , due to both ONOO Ϫ and ubiquinol oxidation, but in this case the ubiquinone was reduced back to the ubiquinol rapidly by complex II (Fig. 5A,  trace b). This oxidation and re-reduction of MitoQ 10 by ONOO Ϫ could be repeated several times and two such reaction cycles are shown in Fig. 5A. The re-reduction of the ubiquinone was by complex II, as malonate prevented reduction of MitoQ 5 and MitoQ 10 after addition of ONOO Ϫ (Fig. 5A). The slower reduction of MitoQ 5 by complex II enabled ONOO Ϫ decay and ubiquinone reduction to be easily distinguished in Fig. 5A as a biphasic change in A 275 after addition of ONOO Ϫ . The biphasic nature of MitoQ 10 re-reduction after ONOO Ϫ addition was not obvious so the traces from Fig. 5A were expanded to clearly show that the decay in ONOO Ϫ differs from the re-reduction of MitoQ 10 (Fig. 5B). The base lines of the traces have been aligned to emphasize the relative changes in A 275 . For CoQ 2 Ϯ malonate (traces c and d) and for MitoQ 10 ϩ malonate (trace b), the addition of ONOO Ϫ leads to an increase in A 275 that decays back to base line over ϳ8 s due to the breakdown of ONOO Ϫ . In contrast, addition of ONOO Ϫ to MitoQ 10 in the presence of uninhibited BHM (trace a) is biphasic with an initial decay in A 275 due to ONOO Ϫ that is followed by a slower decrease in A 275 due to reduction of the ubiquinone formed by ONOO Ϫ oxidation. Idebenone behaved in the same way as CoQ 2 (data not shown).
To confirm that the ubiquinol forms of idebenone and CoQ 2 could also react with ONOO Ϫ , we repeated these experiments in myxothiazol-inhibited BHM. Addition of succinate led to the complete reduction of idebenone (Fig. 5C, trace a) and MitoQ 10 (Fig. 5C, trace b). ONOO Ϫ rapidly oxidized these ubiquinols, and this was reversed by the respiratory chain within ϳ20 s in a malonate-sensitive fashion. CoQ 2 behaved similarly to idebenone (data not shown). Thus, the ubiquinol forms of all exogenous ubiquinones can be oxidized by ONOO Ϫ and then recycled by the respiratory chain.
We next tested whether MitoQ 10 would be more protective against ONOO Ϫ -induced lipid peroxidation than CoQ 2 and idebenone in uninhibited BHM. For this we used PA, a conjugated polyunsaturated fluorescent fatty acid that loses its fluorescence upon peroxidation. Sequential additions of ONOO Ϫ to BHM respiring on succinate caused step decreases in PA fluorescence (Fig. 5D, trace c), and MitoQ 10 protected against this loss (Fig. 5D, trace a). Idebenone also protected against the loss of PA fluorescence (Fig. 5D, trace b); however, the protection was significantly less than that given by MitoQ 10 . The background decay of PA was unaffected by ubiquinones, suggesting that it is not related to lipid peroxidation (Fig. 5D, traces d-f). CoQ 2 behaved like idebenone, while MitoQ 10 in the absence of succinate offered no protection (data not shown).
In summary, the reduced form of MitoQ 10 is an effective antioxidant against ONOO Ϫ , and its slow oxidation by complex III makes it a more effective antioxidant than untargeted formation and subsequent dismutation (Fig. 6A). As these reactions can affect the steady-state concentration of O 2 . , they have implications for the use of exogenous ubiquinones as therapies and for investigating ROS signaling pathways. Indeed, the potential for ROS generation, particularly from idebenone autoxidation, has been raised as a potential concern about the therapeutic use of ubiquinones (71). Therefore we have analyzed the production and consumption of O 2 . by exogenous ubiquinones.
There is the possibility that ubiquinols can consume O 2 . , probably by reaction with its protonated form (HO 2 ⅐ , pK a 4.8), in a mechanism analogous to their chain terminating reaction in lipid peroxidation.  . caused slow oxidation of reduced MitoQ 10 that could be fully blocked by SOD (Fig. 6B). Although O 2 . can react with ubiquinol in aqueous buffer, spontaneous dismutation to hydrogen peroxide (H 2 O 2 ) appears to dominate as the rate of ubiquinol oxidation was low relative to the rate of O 2 . production. Reaction 3 is also likely to occur within phospholipid bilayers and could thereby provide a mechanism for detoxifying HO 2 ⅐ that is inaccessible to SOD. This would be expected to be important in tissues such as the heart, where cristae phospholipids occupy a volume similar to the aqueous mitochondrial matrix (62). To investigate this we added ϳ100 M potassium superoxide (KO 2 ) to a 50 M concentration of the reduced form of MitoQ 10 in PBS-saturated octan-1-ol. KO 2 rapidly oxidized reduced MitoQ 10 (Fig. 6C, trace a). Oxidation of reduced MitoQ 10 was specific to O 2 . /HO 2 ⅐ as it was not oxidized by carrier or by KO 2 that was previously decomposed to H 2 O 2 (Fig. 6C,  traces b and c). Furthermore, reduced MitoQ 10 in aqueous buffer was not oxidized by KO 2 due to its rapid dismutation to H 2 O 2 (Fig. 6C, (66)). Instead, electron transfer from ubiquinol requires an initial deprotonation to a ubiquinolate anion (pK a 11.3 (66)), and this is the likely electron donor to oxygen (UQH ⅐ /UQH Ϫ , E m,7 ϭ 190 mV (66); see Reactions 4 and 5).

REACTIONS 4 AND 5
The ubisemiquinone radical formed can also react with oxygen to form O 2 . , or it can dismutate, but it is the initial reaction between the ubiquinolate and oxygen that is likely to be ratelimiting for autoxidation (Fig. 6A). To see if this model would account for ROS production by exogenous ubiquinols, we assessed the pH sensitivity of ubiquinol autoxidation by measuring the increase in ubiquinone absorption at 275 nm. There was negligible oxidation of the reduced form of MitoQ 10 in aqueous solution at pH 6.8, but autoxidation increased a little at pH 7.8 and dramatically at pH 8.3 (Fig. 7A). Consistent with Fig. 6A . generation during autoxidation of complex II-reduced Mito-Q 10 , we measured cyt c acet reduction in myxothiazol-inhibited BHM (Fig. 7B). This showed that reduction of MitoQ 10 by complex II also caused cyt c acet reduction and that this rate increased with pH from pH 6.8 to 8.3. This pH dependence was not due to changes in MitoQ 10 reduction by complex II as this rate was identical at pH 6.8 and 8.3 (data not shown). Therefore O 2 . production from exogenous ubiquinols produced chemically or by mitochondrial respiration is pH-dependent, consistent with ubiquinol deprotonation being critical for autoxidation.
As deprotonation creates two charged species, autoxidation will predominantly occur in the aqueous phase rather than within phospholipid bilayers and its rate should be inversely proportional to ubiquinol hydrophobicity. To investigate this we measured ubiquinol autoxidation in myxothiazol-inhibited BHM using a range of MitoQ and non-targeted ubiquinone analogs with a spectrum of hydrophobicities (Fig. 7C). All the exogenous ubiquinones reduced cyt c acet (Fig. 7C), and in all cases this rate of reduction was about ϳ50% inhibitable by SOD (data not shown). The rate of O 2 . production from autoxidizing untargeted ubiquinones was inversely proportional to their octan-1-ol/PBS partition coefficients (Fig. 1B). The lowest levels of O 2 . were generated by the most hydrophobic ubiquinol, decylQ, with the most water-soluble, idebenol, producing the most O 2 . and CoQ 2 being intermediate. There was a similar inverse correlation with hydrophobicity for the MitoQ analogs from MitoQ 5 to MitoQ 15 . MitoQ 3 was an exception to this trend, possibly due to its slow reduction by complex II leading to a lower ubiquinol concentration (Fig. 3C). While this inverse relationship between autoxidation and hydrophobicity held within the two groups, the MitoQ analogs were less prone to autoxidation than more hydrophobic non-targeted ubiquinones (e.g. MitoQ 10 versus idebenone). MitoQ analogs are charged cations, while untargeted ubiquinones are neutral, so partitioning into octan-1-ol may not accurately reflect binding to phospholipid bilayers. Therefore we measured the relative binding of MitoQ analogs and untargeted ubiquinones to mitochondrial membranes (Fig. 7D) and found that increased binding of MitoQ analogs to phospholipid bilayers could explain their lower rate of autoxidation when compared with equivalent untargeted ubiquinols.
In summary, all ubiquinols are autoxidized, but this requires an initial deprotonation to a ubiquinolate anion, which is unfavorable within the hydrophobic environment of a phospholipid bilayer. Therefore there are two major determinants of the degree to which ubiquinol will autoxidize: its extent of reduction and its hydrophobicity. Furthermore, the charged TPP moiety of MitoQ analogs may be a better way of lowering overall hydrophobicity and improving pharmacokinetics without enhancing the tendency to autoxidation, as it leads to adsorption onto phospholipid bilayers and insertion of the ubiquinol moiety into the hydrophobic core of the membrane.
Ubiquinol  shown to block a number of redox signaling and apoptotic pathways, but the mechanism remains unclear (35-37, 39 -42). BHM alone did not produce H 2 O 2 when inhibited by myxothiazol (Fig. 7E, trace a) nor did BHM supplemented with CoQ 2 and succinate (Fig. 7E, trace b). However, inhibition of ubiquinol oxidation with myxothiazol (Fig. 7E, trace c) or cyanide (data not shown) led to a build up of ubiquinol and consequent H 2 O 2 production. MitoQ 10 also produced H 2 O 2 but did so in the absence of myxothiazol as it was already fully reduced (data not shown). Therefore autoxidation of ubiquinols does produce O 2 . that will dismutate to H 2 O 2 , and this could be important in redox signaling. NO ⅐ is a signaling molecule produced by NO ⅐ synthases, and their activity may be associated with mitochondria and is important for mitochondrial biogenesis (75)(76)(77). NO ⅐ interacts with mitochondria by reversibly inhibiting complex IV (78), as well as affecting the function of proteins via S-nitrosylation of cysteine residues (79). Therefore we next investigated the interaction of exogenous ubiquinols with NO ⅐ , which may react directly with ubiquinols to produce NO Ϫ or with ubiquinolgenerated O 2 . to form ONOO Ϫ (80). For this the steady-state NO ⅐ concentration produced by the NO ⅐ donor, DETA-NONOate, in the presence of BHM was measured using an NO ⅐ electrode (Fig. 7F). DETA-NONOate gave a steady-state NO ⅐ concentration of ϳ400 nM after 8 -10 min that persisted for a further 10 -20 min (Fig. 7F, trace a). BHM supplemented with CoQ 2 consumed NO ⅐ at an increased rate upon addition of succinate, and addition of myxothiazol led to the complete depletion of NO ⅐ that could be partially reversed by SOD (Fig.  7F, trace b). MitoQ 10 also consumed NO ⅐ to a similar extent, but as it is present in the reduced form, it consumed NO ⅐ rapidly even in the absence of myxothiazol (data not shown  (81) and be detected using Amplex Red and horseradish peroxidase (Fig. 8A). A significant rate of H 2 O 2 efflux was observed in the presence of rotenone and a 1 M concentration of either MitoQ 3 , MitoQ 5 , or MitoQ 10 (Fig. 8A, traces b-d). As expected, the presence of exogenous SOD had no effect on H 2 O 2 efflux from MitoQ 3 (data not shown). In BHM exogenous ubiquinols inhibit the horseradish peroxidase detec- Statistical significance was calculated relative to G3P alone using a Student's one-tailed t test. *, p Ͻ 0.05; ***, p Ͻ 0.001. tion system (data not shown), consequently the accumulation of ubiquinol may cause the apparent decrease in H 2 O 2 efflux with time. Therefore, the initial rate of H 2 O 2 efflux is the best indication of autoxidation, and this is significantly lower for the more hydrophobic MitoQ analogs. To gauge the relative magnitude of MitoQ 3 autoxidation, we compared it with other known mechanisms for generating H 2 O 2 efflux at 37°C (Fig.  8B). This showed that H 2 O 2 generation by 1 M MitoQ 3 was equivalent to high micromolar concentrations of the redox cycler Paraquat and greater than the proton motive force-dependent and rotenone-sensitive efflux of H 2 O 2 from succinate-energized mitochondria (82) (Fig. 8B).
To determine whether this level of intramitochondrial O 2 .
production was damaging, we incubated intact yeast mitochondria with MitoQ analogs and studied their effects on aconitase activity, the iron-sulfur center of which is particularly sensitive to O 2 . (60). The pseudo-first order rate constant for aconitase inactivation within mitochondria was taken as an indication of the steady-state matrix O 2 . concentration (Fig. 8C) Exogenous Ubiquinones Do Not React with Peroxides-MitoQ 10 blocks H 2 O 2 -induced apoptosis and cell death, but the details of how this is achieved remain unclear (29,41,42). H 2 O 2 can cause oxidative damage through Fenton chemistry with Fe 2ϩ or Cu ϩ as well as being a diffusible signaling molecule that reacts with protein thiols (4,5). As direct reaction of H 2 O 2 or alkyl peroxides with ubiquinones and ubiquinols has not been reported under physiological conditions, it was important to clarify whether they react with MitoQ 10 and other exogenous ubiquinones. A lack of direct reaction of the ubiquinol form of MitoQ 10 with H 2 O 2 was confirmed by 1 H NMR measurement of the oxidation of 10 mM reduced MitoQ 10 in acidified D 2 O. Even incubation with 10 mM H 2 O 2 under air over several days did not result in H 2 O 2 -sensitive oxidation of reduced MitoQ 10 (data not shown).
As many of the signaling effects of H 2 O 2 within mitochondria occur through its interactions with thiols, we next determined whether MitoQ 10 could affect the oxidation of reactive biological thiols by H 2 O 2 . To do this we incubated the ubiquinol form of MitoQ 10 in the presence of reduced GSH and glutathione peroxidase and exposed it to a flux of H 2 O 2 generated by glucose oxidase. The H 2 O 2 from glucose oxidase oxidized GSH to GSSG catalyzed by glutathione peroxidase, and GSSG was detected through NADPH consumption by glutathione reductase (Fig. 9A). Neither the reduced nor oxidized forms of MitoQ 10 ( Fig. 9A and data not shown) or CoQ 2 (data not shown) affected GSH oxidation (Fig. 9A) confirming that exogenous ubiquinols neither directly scavenge H 2 O 2 nor interfere with the reaction of H 2 O 2 with biological thiols. We next investigated the induction of the PTP by Ca 2ϩ and tBHP as this involves oxidation of critical protein thiols (83). PTP opening in Ca 2ϩ -loaded rat liver mitochondria was induced by 5 M tBHP and led to the release of accumulated Ca 2ϩ (Fig. 9B, trace b). This process could be blocked by CsA (Fig. 9B, trace c). When the experiment was repeated in the presence of either 1 M (Fig. 9C) or 5 M MitoQ 10 (data not shown), PTP opening was still triggered by 5 M tBHP (Fig. 9C, trace b) and blocked with CsA (Fig. 9C, trace c). The extent of Ca 2ϩ uptake was decreased compared to the system without MitoQ 10 , but the kinetics of PTP induction were similar. Thus MitoQ 10 cannot scavenge alkyl peroxides and prevent PTP opening caused by thiol oxidation.
In summary, MitoQ 10 does not react directly with peroxides or affect the interaction of peroxides with biological thiols. Therefore the potent blocking of exogenous peroxide-dependent reactions by MitoQ 10 (29,41,42) occurs downstream of the peroxide itself. As MitoQ 10 blocks lipid peroxidation (29,30), it is likely that the effects of MitoQ 10 on exogenous peroxides are due to chain termination of lipid peroxidation, which is probably induced by Fenton chemistry of H 2 O 2 in combination with ferrous iron. DISCUSSION This study has clarified the redox, antioxidant, and prooxidant properties of a series of mitochondria-targeted and untargeted exogenous ubiquinones. While CoQ 1 , CoQ 2 , decylQ, CoQ 6 , and idebenone could all restore respiration in mitochondria lacking CoQ, MitoQ 10 could not. The ability of the CoQ analogs to restore respiration was consistent with their fast reduction by ubiquinone reductases and their rapid oxidation by complex III. In contrast, none of the MitoQ analogs could act as electron carriers in respiration because they were not oxidized by complex III. While all the MitoQ analogs rapidly migrate through the mitochondrial inner membrane (30), the hydrophobic core is still a significant activation energy barrier for their transport, albeit a far lower one than that for equivalent hydrophilic cations (32,33). Furthermore, although MitoQ 10 has a similar octan-1-ol/PBS partition coefficient to idebenone, MitoQ analogs are amphipathic and consequently not evenly distributed within phospholipid bilayers (Fig. 1C) (32,33). Thus for MitoQ analogs the steady-state concentration of the ubiquinone moiety at a particular distance from the membrane surface will be related to the length of the alkyl chain. This is in marked contrast to the untargeted ubiquinones, which will be freely soluble within the hydrophobic core. This affinity of MitoQ analogs for the surface of the phospholipid bilayer is simply demonstrated by its limited solubility in cyclohexane, an organic solvent that mimics the membrane core: MitoQ 10 formed a separate, orange-red, oily phase, with only some slight discoloration of the cyclohexane. In contrast idebenone was freely soluble up to at least 4 mM in cyclohexane, as were MitoQ 10 and idebenone in octan-1-ol, a more polar solvent that mimics the membrane surface. Therefore, the significantly decreased reactivity of all MitoQ analogs with complex III may be a consequence of the low concentration of their ubiquinone moieties in the active sites of complex III (Fig. 10A) (64,84). These are near the center of the phospholipid bilayer, and for MitoQ analogs to dock into them, the TPP cation must be located in the hydrophobic core (Fig. 10A). In addition, enzyme shape and dimerization may further increase the effective distance between these binding sites and the surface of the phospholipid bilayer (64,84). Therefore the steady-state concentration of MitoQ analogs within the membrane core is decreased relative to other ubiquinones, and this is likely to explain their low reactivity with complex III.
A related explanation may account for the increasing reduction of MitoQ analogs by complex II and G3PDH on increasing the length of the carbon chain connecting the TPP and ubiquinone moieties (Fig. 3, D and E). The ubiquinone reduction site of complex II is close to the membrane surface, and it is possible to dock the ubiquinone moiety of MitoQ 10 into the active site without the TPP cation dipping into the membrane (Fig. 10B). In contrast, the alkyl linkers for MitoQ 3 and MitoQ 5 appear to be too short to span the distance from the active site to the membrane surface (Fig. 10, C and D). This offers a plausible explanation for the greater reactivity of MitoQ 10 with complex II compared with MitoQ 3 and MitoQ 5 , although with complex II it is not possible to exclude a contribution from hydrophobicity or steric hindrance. If this explanation of the observed reactivities is correct, it may lead to the development of alkyl TPP derivatives as useful probes of membrane proteins. For example, MitoQ 10 was poorly reactive with complex I but did react well with G3PDH, suggesting that the ubiquinone binding site of G3PDH may be closer to the membrane surface, while that of complex I is nearer the hydrophobic core of the membrane.
The effects of MitoQ analogs are not due to their ability to complement respiration. In contrast, untargeted ubiquinones should be able to supplement respiratory defects in vivo provided they can be delivered to mitochondria within cells. In our experiments only the ubiquinones with isoprenoid tails, CoQ 6 , CoQ 2 , and CoQ 1 , could restore non-fermentative growth in CoQ-deficient yeast, while the exogenous ubiquinones with simple saturated carbon chains were ineffective. As this difference is not due to poor electron transfer in the respiratory chain, there may be lower mitochondrial accumulation of the saturated relative to the isoprenoid exogenous ubiquinones. This is not due to differences in passive diffusion as CoQ 6 is several orders of magnitude more hydrophobic than decylQ, which in turn has an octan-1-ol/PBS partition coefficient 1,000fold higher than CoQ 1 (Fig. 1B) (31). Supporting this, the diffusion of CoQ 6 through the mitochondrial outer membrane was far slower than for decylQ or idebenone (Fig. 2D). As uptake does not appear to be passive, we speculate that in yeast there is a process for trafficking exogenous ubiquinones from the plasma membrane to mitochondria and that it is selective for ubiquinones with isoprenoid tails. Such a process is suggested by some earlier work on yeast where accumulation of exogenous CoQ 6 in mitochondria was deficient in some strains (85). It is unclear whether a similar pathway would exist in mammalian cells; however, in a clinical case where CoQ 10 supplementation ameliorated the symptoms of CoQ deficiency, subsequent switching to idebenone caused clinical and metabolic worsening, which disappeared on return to CoQ 10 supplementation (86). As the final steps of CoQ synthesis are in mitochondria, it is not clear why a putative CoQ uptake system should exist. However, it is possible there is a pathway that allows the export mitochondrial CoQ to the plasma membrane, and in the presence of exogenous CoQ it can operate in reverse to insert CoQ into mitochondria. This may explain why dietary supplementation with CoQ 10 leads to only small increases in mitochondrial CoQ 10 concentration except when there is an underlying CoQ 10 deficiency (20 -24, 87). Our results suggest that short chain ubiquinone analogs containing isoprenoid side chains may accumulate more effectively within mitochondria than saturated exogenous ubiquinones and are worth exploring as potential therapies for human diseases with CoQ deficiencies.
The autoxidation of ubiquinols is well established, but its significance for exogenous short-chain ubiquinones is uncertain. Here we confirm that autoxidation of exogenous ubiquinones occurs by enzymatic reduction to the ubiquinol followed by deprotonation to the ubiquinolate anion, which reacts with oxygen to produce O 2 . . Critically, ubiquinol autoxidation will not occur when the ubiquinol head group is within the phospholipid bilayer and is therefore proportional to ubiquinol hydrophilicity. However, while it was possible to detect autoxidation of ubiquinols in BHM and isolated mitochondria, the autoxidation of MitoQ 10 or idebenone did not lead to a significant increase in O 2 . -dependent damage to matrix aconitase.
This lack of autoxidation-dependent damage is consistent with clinical trials using idebenone that have shown it is well tolerated for at least 2 years (28, 88 -90), even though it has a high tendency to autoxidize in vitro (71). While the focus of studies on ROS is often their toxicity, it is possible that autoxidation of exogenous ubiquinones generates subtoxic levels of ROS. This may even be beneficial as ROS can increase the expression of antioxidant pathways via hormesis (4,91). The complex nature of ROS interactions is illustrated by the paradoxical fact that 20 M H 2 O 2 , autoxidation-resistant ascorbate derivatives, and MitoQ 10 all extend the lifespan of cultured cells and decrease telomere shortening (39,91). Therefore, it remains unclear whether exogenous ubiquinone autoxidation is detrimental in vivo as it is associated with a large increase in membrane associated antioxidant capacity, against both lipid peroxidation and O 2 . , and could lead to the induction of antioxidant enzymes via redox-sensitive transcription factors (92)(93)(94). Whether a large increase in protection against lipid peroxidation is on balance beneficial when coupled to a small increase in O 2 . is beyond the scope of this work and would require long term animal studies looking at pathological markers. Even so, it is apparent that increasing hydrophobicity limits the "bad" aqueous autoxidation of ubiquinols without diminishing its "good" organic phase antioxidant functions (Fig. 6). This could provide a teleological explanation for the extreme hydrophobicity of endogenous CoQ 10 (Fig. 1B). MitoQ 10 has been shown to block a number of redox signaling pathways as well as cell death induced by exogenous H 2 O 2 (29,41,42); however, the details of the processes affected are uncertain. MitoQ 10 does not react significantly with H 2 O 2 or other alkyl peroxides ( Fig. 9) but is very effective against lipid peroxidation (29,30). This suggests that the effects of MitoQ 10 against H 2 O 2 are primarily due to blocking the consequences of OH ⅐ presumably formed from iron-catalyzed Fenton chemistry. The presence OH ⅐ within mitochondria results in lipid peroxidation, which has a number of deleterious effects on mitochondrial function and leads to breakdown products, such as the reactive aldehyde 4-hydroxy-2-nonenal, which may have a role in cell damage and signaling (95). This work shows that while MitoQ 10 could interact with ROS in several ways in vivo, the most likely way is that it blocks the effects of exogenous H 2 O 2 by acting downstream of H 2 O 2 to inhibit peroxidative chain reactions.
Here we have investigated how a series of exogenous ubiquinone analogs interact with oxidative phosphorylation and with ROS metabolism. We found that while the type of reactions that MitoQ and untargeted ubiquinone analogs underwent with ROS were similar, ubiquinone hydrophobicity dramatically affected the extent of these reactions. This work clearly shows that the antioxidant reactions of exogenous ubiquinones will predominantly occur within phospholipid bilayers, while the pro-oxidant reactions require an aqueous environment. Therefore, the relative rates of these reactions can be finetuned by hydrophobicity, allowing a rational approach to the design of therapeutic ubiquinone-based antioxidants (e.g. MitoQ 10 ) or mitochondria-targeted redox cyclers (e.g. MitoQ 3 ). Furthermore, there was a sharp distinction between how MitoQ analogs and other short-chain ubiquinones interacted with the mitochondrial respiratory chain. This occurred because the TPP cation decreased the oxidation of MitoQ analogs by complex III. Thus MitoQ analogs function only as antioxidants, while other short-chain ubiquinones can also be oxidative phosphorylation substrates. When used in combination they may provide a way for separating cause and effect in mitochondrial DNA diseases, Parkinson disease, and aging itself where both a decline in oxidative phosphorylation and an increase in ROS may overlap to cause pathology.