Superoxide in Apoptosis MITOCHONDRIAL GENERATION TRIGGERED BY CYTOCHROME c LOSS*

Activation of apoptosis is associated with generation of reactive oxygen species. The present research shows that superoxide is produced by mitochondria isolated from apoptotic cells due to a switch from the normal 4-electron reduction of O 2 to a 1-electron reduction when cytochrome c is released from mitochondria. Bcl-2, a protein that protects against apoptosis and blocks cytochrome c release, prevents superoxide production when it is overexpressed. The switch in electron transfer provides a mechanism for redox signaling that is concomitant with cytochrome c -dependent activation of caspases. The block of cytochrome c release provides a mechanism for the apparent antioxidant function of Bcl-2.

Activation of apoptosis is associated with generation of reactive oxygen species. The present research shows that superoxide is produced by mitochondria isolated from apoptotic cells due to a switch from the normal 4-electron reduction of O 2 to a 1-electron reduction when cytochrome c is released from mitochondria. Bcl-2, a protein that protects against apoptosis and blocks cytochrome c release, prevents superoxide production when it is overexpressed. The switch in electron transfer provides a mechanism for redox signaling that is concomitant with cytochrome c-dependent activation of caspases. The block of cytochrome c release provides a mechanism for the apparent antioxidant function of Bcl-2.
Reactive oxygen species (ROS) 1 are important regulators of apoptosis. Pro-oxidants and redox cycling agents, such as H 2 O 2 (1), diamide (2), etoposide (3), and semiquinones (4), can induce apoptosis. Other apoptotic stimuli, such as treatment with TNF␣ (5), lipopolysaccharide (6), ceramide (7), growth factor withdrawal (8), and human immunodeficiency virus infection (9), can stimulate the ROS production. Studies with TNF␣ and ceramide have shown that these agents activated ROS generation by mitochondria (7). The recent proposed model of p53induced apoptosis also placed ROS as a central signaling event (10,11). Antioxidants and thiol reductants, such as NAC (12), overexpression of thioredoxin (13), and MnSOD (14), can block or delay apoptosis. On the other hand, ROS can provide protective mechanisms under some conditions, such as the activation of NFB by TNF-induced ROS production (15,16). Thus ROS, and the resulting cellular redox change, can be part of the signal transduction pathway during apoptosis. Yet the mechanisms of ROS generation and its relationship with the well studied caspase activation have not been resolved.
A role of ROS during apoptosis was initially proposed based upon the observation that Bcl-2, a general inhibitor of apoptosis in mammalian cells, has an apparent antioxidant function (17,18). Bcl-2 is largely localized to the mitochondrial outer membrane, and when overexpressed, it protects cells from lipid peroxidation and thiol oxidation induced by menadione and hydrogen peroxide (17,18). Bcl-2 also protects 0 cells from apoptosis under conditions that ROS generation is limited (19,20). Thus, its anti-apoptotic function is not limited to the antioxidant property. Nonetheless, this apparent antioxidant property of Bcl-2 has not been explained.
When human myeloid leukemia HL 60 and U-937 cells undergo apoptosis induced by staurosporine (21), etoposide (21), cytosine ␤-D-arabinofuranoside (22), and ionizing radiation (23), there is an early release of cyt c from mitochondria, and the released cyt c participates in the activation of caspase 3. Bcl-2 family proteins prevent cyt c from entering the cytosol either by blocking cyt c release (21,22) or by binding directly with cyt c (23). Cyt c is localized in the mitochondrial intermembrane space and is part of the mitochondrial electron transport chain. A variety of mitochondrial poisons are well known to stimulate ROS generation by a mechanism that involves inhibition of electron transfer, accumulation of reducing equivalents in the middle portion of the electron transfer chain, and direct one-electron transfer to O 2 to produce superoxide (for review, see Ref. 24). This switch appears to occur principally at the level of coenzyme Q (ubiquinone/ubiquinol) and thus could also occur as a consequence of cyt c loss.
In this study, we used staurosporine-treated HL 60 cells to study the mechanism of ROS generation during apoptosis. Results show that the cellular redox change occurred following mitochondrial release of cytochrome c and was in parallel to the caspase activation. The release of cytochrome c was associated with inhibited mitochondrial respiration and stimulated mitochondrial superoxide production, and overexpression of Bcl-2 inhibited all of these processes.
Measurement of Intracellular and Mitochondrial Redox Change-After treatment, cells were pelleted and resuspended in 5% perchloric acid/saturated boric acid, supplemented with 5 M ␥-glutamylglutamate as internal standard. Intracellular GSH and GSSG were measured with high pressure liquid chromatography as described (25), and the amount of acid-insoluble protein was measured by the Bradford method. Cell volume and pH were measured radiochemically (26), and the redox potential was calculated using the Nernst equation with E 0 adjusted to cell pH (Ϫ59 mV/pH unit). E 0 at pH 7.0 was taken as Ϫ240 mV (27). Mitochondrial GSH and GSSG were measured similarly, except cells were first permeabilized with digitonin as described below.
Measurement of Mitochondrial Generation of ROS-After treatment with 1 M staurosporine for 4 h, cells were stained with 15 M dihydrorhodamine (DHR) for 20 min, washed with phosphate-buffered saline, and loaded onto a Becton-Dickinson FACStation. FL-1 fluorescence was recorded, and cell debris was electronically gated out based on the forward light scatter. 300 M tert-butylhydroperoxide treatment was used as a positive control. The DHR fluorescence was also recorded with confocal microscopy. Cells double-stained with 60 nM DiOC 6 and 15 M DHR were washed and resuspended in 15 l of phosphatebuffered saline and placed on a no. 1 thickness coverslip. The green and red fluorescence were simultaneously recorded with a Bio-Rad MRC 1024 laser scanning confocal microscopy system, with the excitation wavelength set at 488 nm. Images were acquired and processed with Laser-Sharp software, using identical background and gain settings. * This work was supported by National Institutes of Health Grants ES 09047 and EY 07892. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Six to eight areas were randomly picked, and representative ones are presented in Fig. 2.
Measurement of Substrate-stimulated Oxygen Consumption-After treatment with staurosporine for the indicated periods of time, 2 ϫ 10 6 cells were washed with phosphate-buffered saline and resuspended in mitochondrial respiration buffer (30). O 2 consumption was measured in a 2.0-ml chamber with a Clark-type oxygen electrode (Yellow Springs) calibrated with air and sodium dithionite (30). Cells were permeabilized with 0.01 mg/ml digitonin for 1 min, and substrate-stimulated respiration was recorded by adding 2.5 mM succinate and 0.25 mM ADP. The substrate stimulated O 2 consumption rate over the rate after digitonin permeabilization was calculated as the respiratory control ratio.
Measurement of Superoxide Generation in Isolated Mitochondria-Mitochondria were isolated from neo and bcl-2 cells with nagarsedigitonin treatment (31). Measurements were performed in a SLM/ Aminco DW-2000 spectrophotometer at 30°C. 40 M acetylated cytochrome c was included in the incubation buffer, and the absorbance change was monitored at 550 nm (32). Reaction was initiated by adding 3 mM succinate. KCN (0.5 mM) was used to inhibit normal electron flow and stimulate superoxide production (positive control), and bovine erythrocyte superoxide dismutase (SOD) (Sigma; 100 units) was added to verify that the reduction of acetylated cyt c was due to superoxide generation.
Western Blot Analysis of Mitochondrial cyt c Release-Mitochondrial and cytosolic fractions were prepared with digitonin-nagarse treatment. Each lane was loaded with 50 g of protein, and Western blot was performed as described (21), using a gift antibody from Dr. R. Jemmerson (University of Minnesota).

RESULTS AND DISCUSSION
To determine whether an oxidant signal precedes or follows mitochondrial release of cyt c, we measured the redox state of the intracellular GSH pool as a function of time following treatment of neo cells with 1 M staurosporine (Fig. 1). Under these conditions, cyt c release was detectable already at 1 h (Fig. 1B), but oxidation of cellular GSH/GSSG did not occur until after 2 h (Fig. 1A). By 6 h, the redox potential E h was oxidized by 72 mV (from Ϫ239 Ϯ 6 to Ϫ167 Ϯ 9 mV), a change similar to the 86 mV oxidation in apoptotic HT29 cells (data not shown). In cells overexpressing Bcl-2, cyt c release was blocked (Fig. 1B), and the redox change was inhibited (Fig. 1A). The measured mitochondrial GSH/GSSG ratio changed with a similar time course (Fig. 1A, inset). Thus, cyt c loss preceded the generation of an oxidant signal and as previously reported, Bcl-2 blocked cyt c release (21) and had an apparent antioxidant function (17,18), as measured by E h of the GSH/GSSG pool and the mitochondrial GSH/GSSG ratio.
Antioxidants, such as NAC, can inhibit apoptosis in some systems (12). Although the staurosporine-induced caspase activation and subsequent DNA fragmentation could be completely inhibited by overexpression of Bcl-2, they were not sensitive to NAC (Fig. 1C). Thus in staurosporine-induced apoptosis, the caspase activation seems to be redox-insensitive, and the effects of Bcl-2 cannot be attributed to its apparent antioxidant function.
Because the redox signal was not observed until after cyt c release, we then examined whether the mitochondrion is a possible source of the redox signal. For this, we used DHR, a non-fluorescent compound that accumulates in the mitochondria and can be oxidized to the fluorescent rhodamine (33). FACScan showed that DHR was oxidized in staurosporinetreated neo cells but not in cells overexpressing Bcl-2 (Fig. 2, A  and B). Confocal microscopy confirmed that this fluorescence was mainly localized to the mitochondria, together with the mt⌬-driven dye DiOC 6 (Fig. 2C). Thus, the results suggest that ROS may be generated in the mitochondria upon cyt c release and that this is prevented by Bcl-2 in association with its inhibition of cyt c release.
Babior and co-workers (30,34) found that substrate-dependent mitochondrial O 2 consumption was inhibited with loss of cyt c in a Fas-activated model of apoptosis. Measurements in mitochondria of neo cells treated with staurosporine also showed that substrate-stimulated O 2 consumption was inhibited (Fig. 3A). Thus, the loss of cyt c is associated with an interruption of normal electron flow that could divert electron transfer to the generation of superoxide.
Activation of the mitochondrial permeability transition (MPT) is a major controlling mechanism in some apoptotic systems (for review, see Ref. 35), and this could also contribute to cyt c release (data not shown) and superoxide production (36,37). To determine whether the MPT had occurred in neo cells treated with staurosporine, we used FACS analysis with rhodamine 123 or DiOC 6 . Results showed that ⌬ was largely retained under conditions where cyt c was released and the redox signal was generated (Fig. 3, B and C). At later times, the ⌬ was lost (Ref. 21 and data not shown), indicating that the redox change preceded the MPT and was not a result of MPT.
To determine whether mitochondria that have lost cyt c during apoptotic signaling have increased substrate-dependent superoxide production, we measured stimulated superoxide production in mitochondria isolated from staurosporine- Superoxide in Apoptosis 11402 treated cells. Results showed that superoxide generation was enhanced in mitochondria isolated from staurosporine-treated neo cells. By 4 h, the rate was 5.28 Ϯ 0.33 nmol/mg protein per min (Fig. 4A). This was dependent upon respiratory substrate and was inhibited by superoxide dismutase.
To determine whether superoxide production could be blocked by preventing cytochrome c release, superoxide production by mitochondria from staurosporine-treated bcl-2 cells was measured. Results showed that mitochondria from cells overexpressing Bcl-2 had no increase in superoxide production following staurosporine treatment (Fig. 4B). Addition of cyanide, a direct mitochondrial electron transport inhibitor, resulted in superoxide generation in mitochondria from Bcl-2-overexpressing cells at a rate similar to that from neo cells. Thus, the results show that the effect of Bcl-2 was not due to a direct electron-scavenging or superoxide-metabolizing activity of Bcl-2 itself but rather to the prevention of O 2 . production.
These observations demonstrate that the antioxidant activity of Bcl-2 is indirect and clarify an important issue concerning Bcl-2 function. However, this finding also raises a fundamental question about the redox signaling in cells. The redox state of the GSH/GSSG pool is substantially oxidized (Ϫ239 mV) compared with the NADPH pool (Ϫ394 mV) (38) that drives GSSG reduction. If the steady state E h of GSH/GSSG reflects a balance between the rate of GSH oxidation by mitochondrially derived oxidants (e.g. superoxide and H 2 O 2 ) and the rate of GSSG reduction by NADPH, then it places Bcl-2 at a pivotal point in regulation of the cellular thiol disulfide redox state by controlling the rate of mitochondrial generation of ROS (Fig. 5).
The present demonstration of a mechanism for ROS generation during apoptosis does not resolve whether ROS has a role in apoptosis signaling. Despite the fact that an oxidant signal occurs during apoptosis by diverse stimuli, so far the available pieces of evidence suggest that caspase activation is irrelevant to the redox state. In vitro data using cell-free systems indi- Superoxide in Apoptosis 11403 cated that caspase 3 activation was independent of the redox state of cyt c (39,40). Similarly, when we added 5 mM NAC to the neo cells, we obtained no inhibition of caspase activation, as measured by DNA fragmentation (Fig. 1C) and phosphatidylserine translocation (data not shown). Thus when ROS is generated concomitantly with a direct activating mechanism of caspases, it might be only a side effect that is unrelated to the key signaling events leading to the caspase-executed cell death. However, a range of other possible functions for ROS could exist beyond the activation of death proteases. Caspases are inhibited by oxidants (41,42). The superoxide generated could function as a mechanism to inactivate caspases and prevent their destruction of cells that phagocytose apoptotic bodies (41). ROS may also counteract the lethal effects of caspases in nonapoptotic cells. Caspase 3 has recently been found to function in the processing of pro-interleukin 16, a non-apoptotic function that would appear to require some low level of constitutive activation (43). If low levels of cyt c release are involved in this process, a concomitant superoxide signal may activate protective mechanisms, such as the activation of the redox-sensitive transcription factor NF-B (44,45). In TNF-␣-induced apoptosis, NF-B has been shown to control the expression of several protective genes, such as c-IAP2 and MnSOD (13,14). Despite the phosphorylation of I-B, the intracellular redox state is known to be another regulatory mechanism of NF-B (46).
Alternatively, the switch of mitochondrial electron transport from its normal pathway through cyt c, which does not result in free radical production, to one in which electrons are transferred to O 2 with production of superoxide, may provide a fail-safe mechanism that complements cyt c-dependent caspase activation for the execution of cell death. The dramatic 75-mV oxidation of the cellular GSH/GSSG pool (Fig. 1A) is a sufficient signal to result in a 500-fold change in the function of systems regulated by vicinal dithiols with E o in an appropriate range (47). Thus, global oxidation due to the electron transfer switch triggered by cyt c release could provide a general backup mechanism for a redox-activated molecular machinery of apoptosis. One example of such a mechanism could be the activation of MPT. The MPT pore has critical thiols, which upon oxidation can trigger its opening (48) and release the apoptosis-inducing factor (35,49). Although not purified yet, apoptosis-inducing factor seems also to be a cysteine protease and can induce nuclear apoptosis in cell-free systems. Oxidants, such as diamide and tert-butylhydroperoxide, can induce its release from isolated mitochondria (49). The time course of mt⌬ change (Fig. 3) indicated that MPT does not contribute significantly to the generation of superoxide in staurosporine-induced apoptosis in HL 60 cells, but rather superoxide production may subsequently activate the MPT in a final common pathway of cell death.