Reversible glutathionylation of complex I increases mitochondrial superoxide formation.

Increased production of reactive oxygen species (ROS) by mitochondria is involved in oxidative damage to the organelle and in committing cells to apoptosis or senescence, but the mechanisms of this increase are unknown. Here we show that ROS production by mitochondrial complex I increases in response to oxidation of the mitochondrial glutathione pool. This correlates with thiols on the 51- and 75-kDa subunits of complex I forming mixed disulfides with glutathione. Glutathionylation of complex I increases superoxide production by the complex, and when the mixed disulfides are reduced, superoxide production returns to basal levels. Within intact mitochondria oxidation of the glutathione pool to glutathione disulfide also leads to glutathionylation of complex I, which correlates with increased superoxide formation. In this case, most of this superoxide is converted to hydrogen peroxide, which can then diffuse into the cytoplasm. This mechanism of reversible mitochondrial ROS production suggests how mitochondria might regulate redox signaling and shows how oxidation of the mitochondrial glutathione pool could contribute to the pathological changes that occur to mitochondria during oxidative stress.

Increased production of reactive oxygen species (ROS) by mitochondria is involved in oxidative damage to the organelle and in committing cells to apoptosis or senescence, but the mechanisms of this increase are unknown. Here we show that ROS production by mitochondrial complex I increases in response to oxidation of the mitochondrial glutathione pool. This correlates with thiols on the 51-and 75-kDa subunits of complex I forming mixed disulfides with glutathione. Glutathionylation of complex I increases superoxide production by the complex, and when the mixed disulfides are reduced, superoxide production returns to basal levels. Within intact mitochondria oxidation of the glutathione pool to glutathione disulfide also leads to glutathionylation of complex I, which correlates with increased superoxide formation. In this case, most of this superoxide is converted to hydrogen peroxide, which can then diffuse into the cytoplasm. This mechanism of reversible mitochondrial ROS production suggests how mitochondria might regulate redox signaling and shows how oxidation of the mitochondrial glutathione pool could contribute to the pathological changes that occur to mitochondria during oxidative stress.
The mitochondrial respiratory chain is the major source of superoxide for most cells (1)(2)(3). Within the mitochondrial matrix, manganese superoxide dismutase converts superoxide to hydrogen peroxide, which can diffuse from mitochondria to the cytoplasm (4,5). Increased production of these reactive oxygen species (ROS) 1 by mitochondria is a major contributor to oxidative damage in pathological situations (2). In many cases the cell responds to this oxidative damage by up-regulating cellular defenses, usually through redox-sensitive transcription factors, phosphatases, or kinases (5). However, there is now considerable evidence that, as well as being damaging agents, ROS can also act as redox signals that are involved in the decisions by mammalian cells to undergo apoptosis, proliferation, or senescence in response to cytokines, growth factors, DNA damage, hypoxia, or oxidative stress (4 -9). Although increased mitochondrial ROS production is thought to have an important role in both the pathology of oxidative damage and in cell signaling (8,10), the mechanisms by which mitochondrial ROS production increase are uncertain.
Complex I (NADH-ubiquinone oxidoreductase) is a major source of superoxide, making it a candidate for increased mitochondrial ROS production and redox signaling (1,11). This complex is the largest (ϳ980 kDa; Ն45 subunits) and least well understood component of the respiratory chain (12). Its primary function is to oxidize NADH in the mitochondrial matrix, thereby reducing ubiquinone to ubiquinol and pumping protons across the inner membrane to drive ATP synthesis (13). There are also suggestions that complex I is involved in nitric oxide physiology, induction of the permeability transition, and the regulation of apoptosis (14,15). Furthermore, the selective loss of complex I activity contributes to Parkinson's disease and Huntington's disease (16,17). One link between these processes may be complex I thiols, which alter their redox state upon oxidation of the mitochondrial glutathione pool, thereby modulating complex I function (18,19). ROS often act by altering the ratio of glutathione (GSH) to glutathione disulfide (GSSG), which can change the activity of proteins by forming mixed disulfides with critical thiols (5,9,20,21). Nitric oxide also causes reversible inhibition of complex I, probably due to the formation of S-nitrosothiols on complex I, whereas longer exposure leads to peroxynitrite formation, glutathione oxidation, and irreversible inhibition (22). These findings suggest that complex I has thiols that are affected by glutathione oxidation and nitric oxide, but their location within complex I, how they are altered by oxidative stress or redox signals, and the functional consequences of these changes are unclear.
A physiological role for redox changes to complex I thiols is an interesting possibility. The regulated formation of mixed disulfides between protein thiols and glutathione disulfide in response to glutathione redox changes has the potential to act as a reversible switch in much the same way as phosphorylation (9,20,21). In support of such a function for thiol changes in mitochondria, it is known that oxidation of mitochondrial glutathione and protein thiols correlate with induction of the permeability transition and with apoptosis (23). Whether similar changes contribute to the responses of mitochondria to oxidative damage and redox signaling has not been investigated. Therefore, we set out to locate redox-active thiols on complex I and to determine the physiological consequences of changes in their redox state. To do this we used a mitochondriatargeted thiol reagent (4-iodobutyl)triphenylphosphonium (IBTP) (24). This molecule is selectively accumulated into mitochondria, driven by the membrane potential, where it labels protein thiols. The formation of mixed disulfides or Snitrosothiols blocks protein thiols and prevents their reaction with IBTP, thus changes in protein thiol redox state can be determined (24). Using this approach we have located redoxsensitive thiols within complex I, shown how they alter redox state in response to glutathione oxidation and nitric oxide, and correlated these changes with increased superoxide production by the respiratory chain. These findings suggest a model for the regulation of ROS formation by mitochondria that has implications for our understanding of mitochondrial oxidative damage and redox signaling.
Preparation and Incubation of Mitochondria, Membranes, and Complex I-Rat liver mitochondria were prepared in STE (250 mM sucrose, 5 mM Tris-HCl, 1 mM EGTA, pH 7.4) (26). Rat heart mitochondria were prepared in STE by treatment of hearts with protease (type VIII, Sigma), followed by homogenization (27). Bovine heart mitochondria were prepared as described previously (28). Bovine heart mitochondrial membranes were prepared by disruption of mitochondria in a blender and collection by centrifugation (29). Complex I was prepared by solubilization of membranes with dodecyl-␤-D-maltoside (DDM, Anatrace, OH) followed by ion-exchange chromatography (30). Pooled fractions were further purified by ion-exchange separation, ammonium sulfate precipitation, and gel filtration, and the pure complex I was stored in buffer containing 0.1% DDM and 10% ethylene glycol at Ϫ80°C.
Electrophoresis and Immunoblotting-For SDS-PAGE samples were pelleted by centrifugation (mitochondria and membranes) or precipitated with acetone (complex I), and gels (usually 12.5% acrylamide) were run using a Bio-Rad Mini Protean system and transferred to nitrocellulose using a Bio-Rad Trans-Blot Semi-Dry Transfer Cell. The blot was incubated with antiserum followed by secondary antibodyhorseradish peroxidase and visualized by enhanced chemiluminescence (Amersham Biosciences).
For two-dimensional electrophoresis samples were separated on a 7-cm Immobilon Dry Strip IPG (pH range 3-10, linear or non-linear) using an Amersham Biosciences IPGphor isoelectric focusing system. The gel strips were then incorporated into a SDS-PAGE mini gel, and proteins were separated in the second dimension, transferred to nitrocellulose, and probed with antiserum.
For BN-PAGE mitochondria or membranes were pelleted and then separated on a 5-12% gradient gel in a Bio-Rad Mini Protean system (31). Proteins were transferred to PVDF using the Bio-Rad Trans-Blot Semi-Dry Transfer Cell and probed with antisera. To isolate proteins after BN-PAGE, bands were excised and proteins electroeluted (32). For fluorography of 35 S-labeled proteins, gels were soaked in fluor (Amplify, Amersham Biosciences), dried, and exposed to film at Ϫ80°C.
N-terminal Sequence Analysis-Complex I (10 g of protein) was separated by SDS-PAGE and transferred to PVDF. After staining with Coomassie Blue N-terminal sequences were obtained by Dr S. Peak-Chew by automated Edman degradation using a model 494 Procise protein sequencer (Applied Biosystems) at the Medical Research Council Laboratory of Molecular Biology, Cambridge, UK.
Assays-Enzyme assays were performed at 30°C. Complex I in bovine heart mitochondrial membranes (90 g of protein/ml) was measured as the rotenone-sensitive oxidation of NADH (⑀ 340 ϭ 6.22 mM Ϫ1 ⅐cm Ϫ1 ) in 250 mM sucrose, 1 mM EDTA, 50 mM Tris-HCl, pH 7.4, 2 mM KCN, 300 nM antimycin, 100 M NADH, and 50 M ubiquinone-1. The rotenone-sensitive rate was typically ϳ90% of the uninhibited rate. The complex II/III activity of bovine heart mitochondrial membranes (40 g of protein/ml) was measured in 40 mM potassium P i , 0.5 mM EDTA, pH 7.4, 20 mM succinate, and 2 mM KCN. After 10-min preequilibration 30 M cytochrome c was added, and its rate of reduction was measured (⑀ 550 ϭ 21 mM Ϫ1 ⅐cm Ϫ1 ). The background rate in the presence of 2 M antimycin was negligible.
To assay for superoxide (33) complex I was suspended in 50 mM potassium P i , pH 7.5, 1 mM EGTA, 100 M N,N-bis (2-bis[carboxymethyl]aminoethyl) glycine (DTPA) supplemented with 50 M acetylated cytochrome c (Sigma). NADH (200 M) was added, and superoxide formation was measured as the reduction of acetylated cytochrome c (⑀ 550 ϭ 21 mM Ϫ1 ⅐cm Ϫ1 ) and corrected using the rate in the presence of Cu,Zn-SOD (200 -600 units/ml). Superoxide production by bovine heart mitochondrial membranes was measured in the presence of 2 mM KCN and 100 nM myxothiazol. This concentration of myxothiazol was just sufficient to completely inhibit complex III without affecting complex I.
To measure H 2 O 2 efflux (34) rat heart mitochondria (1 mg of protein) were suspended in KCl buffer supplemented with 100 M DTPA, 100 M homovanillic acid, and 18 units of horseradish peroxidase at 30°C. Fluorescence was measured continuously ( excit ϭ 312 nm; emis ϭ 420 nm), and mitochondrial H 2 O 2 production was calibrated with H 2 O 2 (⑀ 240 ϭ 43.6 M Ϫ1 ⅐cm Ϫ1 ) and corrected for the background caused by respiratory substrates in the absence of mitochondria.
The uptake of IBTP by mitochondria was measured using an ionselective electrode (35,36). Oxygen consumption was measured using a Clark-type oxygen electrode (Rank Brothers, Bottisham, Cambridge), calibrated assuming 222 nmol of O 2 /ml at 30°C (37). Glutathione, glutathione disulfide, and protein-glutathione mixed disulfides were assayed as described (38). The thiol content of complex I was quantitated using a kit (Molecular Probes, catalog number T-6060). The FMN content of complex I was determined fluorometrically (39) and was 1.09 Ϯ 0.06 nmol of FMN/mg of protein: assuming a M r of 980,000, this gives 1.07 Ϯ 0.06 mol of FMN/mol of complex I. Protein concentration was determined by the bicinchoninic acid assay. Statistical significance was calculated using Student's t test for paired or unpaired data as appropriate.

Complex I Contains Redox-active Thiols-
The lipophilic cation IBTP is taken up selectively into the mitochondrial matrix where it binds to reactive protein thiols to form stable thioether conjugates (24). Following incubation of rat heart mitochondria with IBTP, complex I was isolated by blue native gel electrophoresis (BN-PAGE), and IBTP labeling was demonstrated by immunoblotting with antiserum against butyltriphenylphosphonium (␣IBTP) (Fig. 1A). Blocking the uptake of IBTP into mitochondria by dissipating the membrane potential with the uncoupler carbonyl cyanide-p-trifluoromethoxyphenylhydrazone prevented IBTP labeling (Fig. 1A). Bovine heart mitochondrial membranes also showed IBTP labeling of complex I thiols (Fig. 1B). There was a dramatic loss of IBTP labeling of complex I within intact rat liver mitochondria on treatment with tBHP ( Fig. 1C), or on exposure to the nitric oxide donor SNAP (Fig. 1D). Because tBHP is a glutathione peroxidase substrate, it primarily acts though oxidation of glutathione to glutathione disulfide; however, direct interactions of tBHP with thiols cannot be excluded. Hence complex I contains reactive thiols that become altered in response to oxidative and nitrosative stress.
Identification of Polypeptides in Isolated Complex I That Contain Redox-active Thiols-To determine which of the 45 or more polypeptides of complex I contain redox-active thiols, we incubated isolated complex I with IBTP, resolved the subunits by SDS-PAGE, and characterized the IBTP-labeled polypeptides by immunoblotting with ␣IBTP ( Fig. 2A). IBTP labeled the 75-and 51-kDa polypeptides ( Fig. 2A). There was also some labeling of a 23-kDa polypeptide ( Fig. 2A) that cross-reacted with antiserum against the TYKY subunit (data not shown). A time course showed that the 75-and 51-kDa polypeptides were labeled rapidly by IBTP, whereas the TYKY subunit reacted more slowly (Fig. 2B). A two-dimensional immunoblot of IBTPlabeled complex I gave three IBTP-labeled spots at the expected pI and molecular mass values (75 kDa ϭ 5.4, 51 kDa ϭ 7.1, and TYKY ϭ 5.2) which also cross-reacted with cognate antisera (Fig. 2C). IBTP labeled only 3 of the 45 or more complex I polypeptides, and N-terminal sequence analysis confirmed that these were the 75-kDa, 51-kDa, and TYKY subunits of complex I (Fig. 2D).
Redox-active Complex I Thiols in Mitochondrial Membranes and Intact Mitochondria-We next investigated whether the IBTP-labeled polypeptides in isolated complex I were also the ones that reacted with IBTP in bovine heart mitochondrial membranes. Membranes were incubated with IBTP, and proteins were separated by two-dimensional electrophoresis and probed with ␣IBTP ( Fig. 2E) or with antisera against the 75and 51-kDa polypeptides (data not shown). The 75-and 51-kDa subunits were labeled by IBTP (Fig. 2E). It was not possible to determine whether the TYKY subunit was IBTP-labeled due to comigration with other proteins (data not shown). We then incubated intact liver or heart mitochondria with IBTP and isolated complex I by BN-PAGE followed by electroelution. The individual complex I polypeptides were resolved by SDS-PAGE and probed with ␣IBTP, or with antisera against the 75-kDa, 51-kDa, or TYKY polypeptides (␣ComI) (Fig. 2F). This showed that IBTP labeled the 75-and 51-kDa subunits in mitochondria but not the TYKY subunit (Fig. 2F).
We conclude that thiols on the 75-and 51-kDa subunits of complex I are labeled by IBTP in the isolated complex, bovine heart mitochondrial membranes, and intact mitochondria. These thiols are blocked in intact mitochondria upon oxidation of the glutathione pool or upon exposure to nitric oxide, suggesting that IBTP binding indicates the redox state of critical complex I thiols. The TYKY subunit in isolated complex I reacts with IBTP after prolonged incubation but not within intact mitochondria. This subunit contains 8 Cys residues that are all components of iron-sulfur centers. Hence this is probably due to loss of an iron-sulfur cluster that is unrelated to thiol redox changes in intact mitochondria and so was not investigated further.
Response of Redox-active Complex I Thiols to Oxidative and Nitrosative Stress-To characterize how the redox-active thiols on the 75-and 51-kDa subunits of complex I can be modified, we investigated the effects of thiol reagents, oxidative stress, or nitrosative stress on the IBTP labeling of isolated complex I (Fig. 3). The thiol reductant DTT increased IBTP labeling of the 75-and 51-kDa subunits, presumably by reducing thiols oxidized during preparation, but did not cause IBTP labeling of other subunits ( Fig. 3A and data not shown). The thiol-alkylating reagent NEM, the thiol-specific oxidant diamide, glutathione disulfide (GSSG), and the S-nitrosothiol SNAP (Fig. 3A) all blocked the IBTP labeling of the 75-and 51-kDa subunits. In contrast, biologically effective concentrations of hydrogen peroxide, superoxide, or peroxynitrite did not (Fig. 3B). Incubation with NADPH (500 M), NADH (500 M), rotenone (20 g/ml), ubiquinone 1 (10 M), or a free calcium concentration of ϳ17 M did not affect IBTP binding to complex I (data not shown). We conclude that the redox-active thiols on the 75-and 51-kDa subunits of complex I react rapidly with disulfides such as GSSG and with S-nitrosothiols, but not with superoxide or hydrogen peroxide.
Few Reactive Thiols on Other Complex I Polypeptides-To determine whether the 51-and 75-kDa subunit thiols were the only available thiols on complex I, or were a particularly reactive subset, we used the thiol reagent maleimide-biotin to label all exposed thiols (Fig. 3C). The 51-and 75-kDa subunits were the most heavily labeled polypeptides (Fig. 3C). In addition, one band above and another just below the 51-kDa subunit were labeled (Fig. 3C). The identity of these bands was not pursued further, but we note that the 49-kDa and the ND5 polypeptides run in this region and that the 51-kDa subunit runs as a doublet under some conditions.
Our complex I preparations contained 19.9 Ϯ 4.1 thiols/flavin mononucleotide (FMN), and this increased to 23.6 Ϯ 5.7 thiols/ FMN upon incubation under standard conditions for 20 min (mean Ϯ range of two preparations: there is one FMN per complex I). It is difficult to relate this precisely to the number of thiols exposed in native complex I due to the loss of thiols by oxidation on storage, and the gain of thiols on denaturation of iron-sulfur centers. Even so, it is helpful to compare these data with the predicted number of thiols. There are 116 Cys residues in the 45 known subunits of complex I and correcting for putative iron-sulfur centers leaves 85 or 86 Cys. The mature 75-and 51-kDa subunits probably contain 6 and 8 of these, respectively, potentially accounting for 14 of the 20 Ϯ 4 accessible thiols on complex I. Therefore most of the accessible thiols on complex I are on the 75-and 51-kDa subunits, and these are the only ones sufficiently reactive to be labeled by IBTP.
Glutathione Disulfide Forms Mixed Disulfides with Complex I-To see whether GSSG forms mixed disulfides with  (Fig. 3D). Two complex I polypeptides of ϳ75 and ϳ50 kDa were labeled by [ 35 S]GSSG, and this was reversed by the thiol reductant DTT, indicating a disulfide linkage (Fig. 3D). Incubation of bovine heart mitochondrial membranes with [ 35 S]GSSG followed by BN-PAGE and fluorography showed that complex I in mitochondrial membranes also formed mixed disulfides (Fig. 3E). We conclude that, upon exposure to GSSG, the IBTP-reactive thiols on the 75-and 51-kDa subunits of complex I readily form glutathione-protein mixed disulfides.
Relative Reactivity of Complex I Thiols with GSSG and SNAP-To determine the relative reactivity of the 51-and 75-kDa thiols, we incubated complex I with various concentrations of GSSG or SNAP and determined how this affected IBTP binding (Fig. 3, F and G). Both GSSG and SNAP were more effective at preventing IBTP binding to the 51-kDa subunit than to the 75-kDa subunit, suggesting that the 51-kDa subunit thiols are slightly more reactive with GSSG and S-nitrosothiols.

Redox Modification of Complex I Thiols Affects
Complex I Activity-To investigate how thiol redox changes affected complex I function, we measured the time course of inhibition of complex I in bovine heart mitochondrial membranes by GSSG or SNAP (Fig. 4A). This was compared with the combined activity of complexes II and III to determine the relative susceptibilities of complex I and other respiratory chain complexes to inhibition. Both GSSG and SNAP inhibited complex I to a greater extent than complexes II-III (Fig. 4A). The inhibition was time-dependent with 30-to 60-min incubation leading to substantial inactivation of complex I. To determine whether this inhibition was due to the formation of mixed disulfides on complex I, we exposed mitochondrial membranes to GSSG or SNAP for various times then treated them with the thiol reagents GSH, DTT, or thioredoxin to break the disulfide links (Fig. 4B). GSSG did prevent IBTP labeling under these conditions (Fig. 4C), consistent with mixed disulfide formation. Under these conditions quantitation was problematic, because GSSG caused a variable decrease in the amount of complex I resolved by BN-PAGE, presumably due to aggregation. Even so, subsequent incubation of these GSSG-treated membranes with DTT removed the glutathione disulfide and restored the was incubated with IBTP as in A before separation by two-dimensional gel electrophoresis (IEF, pH 3-10). Identical blots were probed with ␣IBTP or a mixture of antisera against the 75-kDa, 51-kDa, and TYKY complex I polypeptides (␣ComI). D, N-terminal sequence analysis of IBTP-labeled complex I polypeptides. Complex I was incubated with IBTP as in A, and identical 5-g protein samples were separated on a 10% SDS-PAGE and transferred to PVDF. One blot was probed with ␣IBTP, whereas the other was stained with Coomassie Blue, and the indicated N-terminal amino acid sequences were obtained. E, identification of IBTP-labeled complex I polypeptides in mitochondrial membranes. Bovine heart mitochondrial membranes (1 mg of protein/ml) were incubated with 100 M IBTP for 20 min. Proteins were then separated by two-dimensional gel electrophoresis (IEF, non-linear pH 3-10) and probed with ␣IBTP. Antiserum against the 75-or 51-kDa polypeptides confirmed the identity of the indicated IBTP-labeled spots. F, determination of IBTP-labeled complex I subunits in intact mitochondria. Rat liver or heart mitochondria were incubated with 10 mM succinate and 50 M IBTP. After 10 min complex I was isolated by BN-PAGE, and polypeptides were isolated by electroelution. Identical samples were then separated by SDS-PAGE and probed with ␣IBTP or with antisera against the 75-kDa, 51-kDa, or TYKY complex I subunits (␣ComI). The data for the 51-kDa and TYKY subunits are from liver mitochondria, but these mitochondria gave several bands at and below 75-kDa, which cross-reacted with antiserum against the 75-kDa subunit and ␣IBTP, presumably due to proteolytic degradation (data not shown). Therefore, data for the 75-kDa subunit were obtained using heart mitochondria in the presence of protease inhibitors. reactivity of complex I thiols with IBTP, even after incubation with GSSG for 1 h (Fig. 4C). Although this, and related treatments with thioredoxin or GSH, removed the glutathione mixed disulfides from complex I, they did not reverse the inhibition caused by incubation with GSSG for 30 -60 min (Fig.  4B). Therefore, the complex I inhibition that follows long term incubation with GSSG is not simply due to the presence of protein-glutathione mixed disulfides, but may instead be a consequence of their long term maintenance. Possible mechanisms for this gradual inactivation include reaction of complex I-glutathione mixed disulfides with adjacent protein thiols to form an internal disulfide; formation of non-reversible glutathione mixed disulfides on a small number of critical complex I thiols; degradation of iron-sulfur centers; or alterations to the quaternary structure or conformation of the complex. We conclude that upon incubation with GSSG the redox-active thiols on the 75-and 51-kDa subunits of complex I form mixed disulfides with glutathione, and this correlates with a gradual decrease in the activity of complex I over time that is not reversed on breaking the disulfides.
Glutathionylation of Complex I Thiols Increases Superoxide Production-We next determined whether superoxide production by complex I was affected by the formation of glutathioneprotein mixed disulfides. Superoxide was measured as the superoxide dismutase (SOD)-sensitive rate of reduction of acetylated cytochrome c (Fig. 5A). Preincubating isolated complex I with GSSG to form mixed disulfides doubled the rate of SOD-sensitive superoxide production to 195.3 Ϯ 19.1% of controls (n ϭ 4, mean Ϯ S.E., p Ͻ 0.05). These measurements were extended to superoxide formation by complex I in bovine heart mitochondrial membranes, where the complex III inhibitor myxothiazol was used to maintain reduction of the Q pool, eliminate changes in superoxide production by complex III, and functionally isolate complex I (Fig. 5B). Preincubation with GSSG for 5 min increased SOD-sensitive superoxide formation to 134 Ϯ 2.5% of controls (n ϭ 3, mean Ϯ S.D., p Ͻ 0.001). This increase in superoxide production was reversed when the mixed disulfides on GSSG-treated bovine heart mitochondrial membranes were removed by incubation with thiol reductants (Fig. 5C). This increase in superoxide production occurred rapidly upon formation of mixed disulfides on complex I and is therefore distinct from the complex I inhibition upon long term incubation with GSSG (Fig. 4A), which was not reversed by breaking mixed disulfides (Fig. 4B). We conclude that the formation of mixed disulfides between complex I thiols and glutathione correlates with increased superoxide production and that this can be reversed by breaking the mixed disulfide.

Oxidation of Mitochondrial Glutathione Correlates with Increased Hydrogen Peroxide Production by Complex I in Intact
Mitochondria-It is important to determine whether the increased superoxide production upon formation of mixed disulfides with complex I occurs within mitochondria under physiological conditions. Direct measurement of superoxide within mitochondria is not possible; instead, superoxide is converted to hydrogen peroxide (H 2 O 2 ) by SOD in the matrix, and the subsequent efflux of H 2 O 2 from mitochondria is measured (34). Heart mitochondria were incubated with the complex I-linked substrates malate/pyruvate to generate NADH in the matrix and initiate complex I respiration. The rate of H 2 O 2 efflux from control mitochondria was indistinguishable from background ( Fig. 5D, inset, traces a and b), in agreement with others (40). We then treated mitochondria with the glutathione peroxidase substrate tBHP, which mimics the glutathione oxidation that occurs during oxidative stress in vivo. Treatment with tBHP increased H 2 O 2 production by mitochondria respiring on complex I-linked substrates (Fig. 5D, inset, trace c). Calibration and correction for background showed that tBHP led to mitochondrial H 2 O 2 production of 0.62 Ϯ 0.17 nmol of H 2 O 2 /min/mg of mitochondrial protein (n ϭ 3, ϮS.E., Fig. 5D). Exposure to the thiol-specific oxidant diamide also increased mitochondrial H 2 O 2 production (data not shown). Control experiments showed that neither trace amounts of tBHP carrying over into the assay nor inhibition of H 2 O 2 detoxification by glutathione peroxidase upon glutathione oxidation could account for this increase in H 2 O 2 production (Fig. 5D, see legend). Therefore, oxidation of the mitochondrial glutathione pool dramatically increases H 2 O 2 efflux from mitochondria respiring on complex I-linked substrates.
Formation of Complex I Protein-Glutathione Mixed Disulfides Correlates with Increased Mitochondrial Hydrogen Peroxide Production-Finally, we determined whether the increased H 2 O 2 production by intact mitochondria correlated with the formation of glutathione-protein mixed disulfides on complex I. Treatment with tBHP oxidized mitochondrial GSH to GSSG and led to the formation of mitochondrial protein mixed disulfides (Fig. 6A). Because of the slow uptake of GSSG by mitochondria (41)(42)(43), it was difficult to measure complex I-glutathione mixed disulfides directly. Instead, IBTP was added to mitochondria and then complex I was isolated by BN-PAGE. Treatment with tBHP decreased IBTP labeling of complex I thiols (Fig. 6B). This loss of labeling was not due to disruption of IBTP uptake (Fig. 6B, see legend). GSSG only forms mixed disulfides with the IBTP-reactive complex I subunits and in doing so blocks IBTP binding (Figs. 2 and 3). Consequently, the prevention of IBTP binding to complex I suggests that glutathione-complex I mixed disulfides form within tBHP-treated mitochondria. The rate of oxygen consumption by coupled mi-tochondria respiring on malate/pyruvate was 13.1 Ϯ 0.3 nmol of O 2 /min/mg of protein and was not affected significantly by tBHP (13.5 Ϯ 0.4 nmol of O 2 /min/mg of protein; n ϭ 2; Ϯ range). Consequently, the rate of H 2 O 2 production shown in Fig. 5D accounts for less than 2% of electron flow through the respiratory chain. We conclude that protein mixed disulfides form on complex I in intact mitochondria following oxidation of the glutathione pool and that this correlated with increased superoxide formation, which in turn leads to the efflux of H 2 O 2 from mitochondria. DISCUSSION We have found redox-active thiols on the 75-and 51-kDa subunits of complex I that react with GSSG to form glutathione-protein mixed disulfides. These mixed disulfides were reversed by thioredoxin or glutathione, suggesting that the reversible glutathionylation of complex I may occur in vivo. This production by mitochondria exposed to tBHP. Rat heart mitochondria (1 mg of protein) were incubated with 100 M DTPA for 5 min, then with 500 M tBHP for a further 5 min, before being isolated and resuspended, and H 2 O 2 production was measured upon addition of malate/pyruvate (substrates). D is a typical rate of H 2 O 2 production following incubation with tBHP, corrected for background. The number is the corrected rate in nanomoles of H 2 O 2 /min/mg of protein. The inset shows the background without mitochondria (a), the negligible production of H 2 O 2 by control mitochondria (b), and the increase in tBHP-treated mitochondria (c). AFU, arbitrary fluorescence units. In control experiments mitochondria were incubated with tBHP, re-isolated, and then pelleted, and the effects of the supernatant on the H 2 O 2 assay were shown to be negligible; therefore, the increase in H 2 O 2 production is not due to tBHP carryover into the assay. The increase in H 2 O 2 production was not due to oxidation of mitochondrial glutathione inactivating mitochondrial glutathione peroxidase, because inhibition of this enzyme with mercaptosuccinate did not increase H 2 O 2 efflux (57).
is illustrated in Fig. 7 by a model that shows that the duration and extent of complex I glutathionylation is regulated by thioredoxin or glutaredoxin and by the redox state of the mitochondrial glutathione pool. Glutathionylation of complex I correlated with a rapid increase of mitochondrial superoxide formation, which was reversed upon removal of the glutathione. Within mitochondria, superoxide is converted to H 2 O 2 , which can diffuse from the mitochondria through the lipid bilayer into the cytosol. Consequently, mitochondrial ROS production may be reversibly modulated in response to the redox state of the mitochondrial glutathione pool. This mechanism could contribute to the response of mitochondria to thiol redox changes and oxidative damage.
The 75-and 51-kDa subunits of complex I are on the large arm of complex I that protrudes into the matrix (13). The 51-kDa subunit contains the NADH binding site and the FMN prosthetic group, which is the entry site for electrons from NADH to complex I, whereas the 75-kDa subunit is close to the 51-kDa subunit (13). The location of the NADH and FMN binding sites within the sequence is currently uncertain, but the formation of mixed disulfides by thiols close to FMN or bound NADH might be expected to affect complex I activity and superoxide formation. The reversibility of the increased superoxide production by complex I suggests that this may be a regulated process, but it could also be a consequence of disruption to complex I. The site of superoxide production by complex I is disputed, with iron-sulfur centers (44), FMN (45), or the Q binding site (46) all candidates. Therefore, finding out how mixed disulfide formation increases superoxide production requires better understanding of the structure of complex I and of the pathway of electron movement through the complex.
The ready formation of glutathione mixed disulfides on complex I has pathological significance. Complex I was markedly more susceptible to glutathionylation and irreversible inacti-vation than other respiratory chain complexes. This may explain why complex I is selectively damaged in Parkinson's disease or Huntington's disease (16,17,47) and indicates why mitochondrial oxidative stress and glutathione depletion are common features of these diseases (48,49). Redox active complex I thiols also reacted with S-nitroso compounds, probably by transnitrosation to form S-nitrosothiols. This also inhibited complex I activity and supports a pathological role for the interaction of NO with complex I (14) or possibly a physiological role for the putative mitochondrial isoform of nitric oxide synthase (14,50). Although the etiology of neurodegenerative diseases associated with complex I defects is uncertain, we suggest that mechanisms that lead to the persistent oxidation of mitochondrial glutathione should be explored and that therapies to maintain this glutathione pool may be effective.
The increase in mitochondrial ROS production in response to glutathione oxidation may be how oxidatively damaged or defective mitochondria are recognized and degraded by the cell. Damage to a mitochondrion oxidizes its glutathione pool, and, because the redox state of the glutathione pool of each mitochondrion is distinct (42), this would distinguish the damaged mitochondrion from healthy organelles. Within the damaged mitochondrion, increased ROS production from complex I would maintain glutathione oxidation and culminate in destruction of the mitochondrion through the permeability transition (51) and subsequent autophagy by lysosomes (52) or contribute to mitochondrial damage and cell death (53).
Hydrogen peroxide may be a redox signal from the mitochondrial matrix to signaling pathways in the cytosol. A number of signaling pathways involving protein kinase C, NF-B, platelet-derived growth factor, c-Jun N-terminal kinase, Ras, and p53 are activated by H 2 O 2 (4, 54), and increased H 2 O 2 production upon oxidation of mitochondrial glutathione could be a physiological mechanism for activating them. Oxidation of mi-FIG. 6. Glutathione oxidation and formation of protein mixed disulfides correlates with increased H 2 O 2 production by isolated mitochondria. A, increase in GSSG and protein mixed disulfide formation after tBHP treatment. Rat heart mitochondria (1 mg of protein/ml) were incubated with 500 M tBHP as in Fig. 5D and then assayed for GSH, GSSG, and protein mixed disulfides (Pr-SSG). Data are means Ϯ S.D. of six incubations (*, p Ͻ 0.001). B, loss of IBTP binding to complex I during glutathione oxidation. Rat heart mitochondria were incubated as above with 500 M tBHP, and then IBTP (50 M) was added and 10 min later complex I was separated by BN-PAGE and probed with ␣IBTP. The effect of tBHP on the uptake of IBTP by mitochondria was minimal with uptake by control mitochondria of 6.8 Ϯ 0.7 nmol of IBTP/mg of protein and uptake of 5.9 Ϯ 0.1 nmol of IBTP/mg of protein by tBHP-treated mitochondria.

FIG. 7.
Model for interaction of glutathione with complex I to increase mitochondrial superoxide production. This scheme shows electron transfer through complex I from NADH via FMN and iron-sulfur centers (diamonds) to ubiquinone (Q) in the mitochondrial inner membrane. Oxidative stress shifts the equilibrium of the mitochondrial glutathione pool from being largely glutathione (GSH) to increase the amount of glutathione disulfide (GSSG). GSSG can form mixed disulfides (GSSϪ) with complex I thiols (HSϪ) leading to increased superoxide (O 2 . ) formation. This increase in superoxide formation is reversed on breaking down the mixed disulfide by glutaredoxin (Grx) or thioredoxin (Trx). The superoxide is converted to H 2 O 2 by the action of manganese superoxide dismutase (MnSOD). Hydrogen peroxide can diffuse from mitochondria into the cytoplasm to act as a potential redox signal.
tochondrial glutathione can be initiated by oxidative damage or by conventional signals: for example, the inflammatory cytokine tumor necrosis factor ␣ oxidizes mitochondrial glutathione and increases mitochondrial ROS production, but how it does this is obscure (55,56). Such a pathway has the potential to act as a self-amplifying feed-forward loop, by H 2 O 2 efflux from one mitochondrion inducing glutathione oxidation and consequent H 2 O 2 generation in neighboring mitochondria. Although much remains to be explored, it is tempting to speculate that this mechanism might enable the internal redox state of mitochondria to be communicated to redox-sensitive signaling pathways in the cytoplasm and thus modulate apoptosis, cell senescence, or proliferation.
In conclusion, we have shown that reversible glutathionylation of complex I occurs within mitochondria following oxidation of the glutathione pool. This elevates superoxide production, which in turn increases the amount of H 2 O 2 that is released from mitochondria to the cytoplasm. These findings suggest a mechanism by which mitochondria may be able to reversibly signal their glutathione redox state to the rest of the cell.