Direct effect of ceramide on the mitochondrial electron transport chain leads to generation of reactive oxygen species. Role of mitochondrial glutathione.

Ceramide is a sphingolipid that is generated in the signaling of inflammatory cytokines such as tumor necrosis factor (TNF), which exerts many functional roles depending on the cell type where it is produced. Since TNF cytotoxicity is mediated by overproduction of reactive oxygen species from mitochondria, we have examined the role of ceramide in generation of oxidative stress in isolated rat liver mitochondria. The present studies demonstrate that addition of N-acetylsphingosine (C2-ceramide) to mitochondria led to an increase of fluorescence of dihydrorhodamine 123 or dichlorofluorescein-stained mitochondria, indicating formation of hydrogen peroxide. Such effect was significant at 0.25 μM and maximal at 1-5 μM C2, decreasing at greater concentrations. This inductive effect of ceramide was mimicked by N-hexanoylsphingosine at the same concentration range, whereas the immediate precursor of C2, C2-dihydroceramide increased hydrogen peroxide at 1-5 μM. Sphingosine generated hydrogen peroxide at concentrations ≥10 μM, whereas diacylglycerol failed to increase hydrogen peroxide. The increase in hydrogen peroxide induced by C2 was not triggered by mitochondrial permeability transition as C2 did not induce mitochondrial swelling. Blocking electron transport chain at complex I and II prevented the increase in hydrogen peroxide induced by C2; however, interruption of electron flow at complex III by antimycin A potentiated the inductive effect of C2. Depletion of matrix GSH prior to exposure to ceramide resulted in a potentiated increase (2-fold) of hydrogen peroxide generation, leading to lipid peroxidation and loss of activity of respiratory chain complex IV compared with GSH-repleted mitochondria. Mitochondria isolated from TNF-treated cells showed an increase (2-3-fold) in the amount of ceramide compared with mitochondria from untreated cells. These results suggest that mitochondria are a target of ceramide produced in the signaling of TNF whose effect on mitochondrial electron transport chain leads to overproduction of hydrogen peroxide and consequently this phenomena may account for the generation of reactive oxygen species during TNF cytotoxicity.


From the Instituto Investigaciones Biomédicas, Consejo Superior Investigaciones Científicas and the Liver Unit and Servicio de Bioquímica, Department of Medicine, Hospital Clinic i Provincial, Universidad de Barcelona, Barcelona 08036, Spain
Ceramide is a sphingolipid that is generated in the signaling of inflammatory cytokines such as tumor necrosis factor (TNF), which exerts many functional roles depending on the cell type where it is produced. Since TNF cytotoxicity is mediated by overproduction of reactive oxygen species from mitochondria, we have examined the role of ceramide in generation of oxidative stress in isolated rat liver mitochondria. The present studies demonstrate that addition of N-acetylsphingosine (C 2 -ceramide) to mitochondria led to an increase of fluorescence of dihydrorhodamine 123 or dichlorofluorescein-stained mitochondria, indicating formation of hydrogen peroxide. Such effect was significant at 0.25 M and maximal at 1-5 M C 2 , decreasing at greater concentrations. This inductive effect of ceramide was mimicked by N-hexanoylsphingosine at the same concentration range, whereas the immediate precursor of C 2 , C 2 -dihydroceramide increased hydrogen peroxide at 1-5 M. Sphingosine generated hydrogen peroxide at concentrations 10 M, whereas diacylglycerol failed to increase hydrogen peroxide. The increase in hydrogen peroxide induced by C 2 was not triggered by mitochondrial permeability transition as C 2 did not induce mitochondrial swelling. Blocking electron transport chain at complex I and II prevented the increase in hydrogen peroxide induced by C 2 ; however, interruption of electron flow at complex III by antimycin A potentiated the inductive effect of C 2 . Depletion of matrix GSH prior to exposure to ceramide resulted in a potentiated increase (2-fold) of hydrogen peroxide generation, leading to lipid peroxidation and loss of activity of respiratory chain complex IV compared with GSH-repleted mitochondria. Mitochondria isolated from TNF-treated cells showed an increase (2-3-fold) in the amount of ceramide compared with mitochondria from untreated cells. These results suggest that mitochondria are a target of ceramide produced in the signaling of TNF whose effect on mitochondrial electron transport chain leads to over-production of hydrogen peroxide and consequently this phenomena may account for the generation of reactive oxygen species during TNF cytotoxicity.
Tumor necrosis factor (TNF) 1 is a cytokine produced by a wide variety of cell types whose production is up-regulated in a number of stressful and pathological conditions (1)(2)(3). TNF exerts a pleiotropic mode of action on multiple cell functions including regulation of immune responses, host defense reactions, and gene regulation. In addition, its role as a mediator of cytotoxicity on certain susceptible transformed cell lines has been well documented (4 -7). Upon binding to its receptor subtypes, TNF evokes a complicated array of intracellular signals, including G-coupled activation of phospholipase A 2 , release of arachidonic acid, DAG production, and activation of protein kinase C, some of which may participate in the chain of reactions that result in cell killing (5)(6)(7). An overproduction of ROS has been proposed as an important mechanism to mediate the cytotoxic and gene regulating effects that TNF exerts on tumor cells (8, 9 -12).
Ceramide has attracted considerable attention due to its role as an intracellular effector molecule that mimics some of the biological effects exerted by inflammatory cytokines such as TNF (13)(14)(15)(16)(17)(18). In addition to its de novo biosynthesis, which is initiated by the condensation of serine and palmitoyl-CoA, ceramide can also be generated by sphingomyelin hydrolysis. Thus, enzymes that hydrolyze sphingomyelin such as sphingomyelinases stand as regulators of intracellular ceramide levels and consequently ceramide-mediated functions. These enzymes are key components of the so-called sphingomyelin pathway, an ubiquitous system that functions in transducing the signals of cytokines to the cell interior (13)(14)(15).
Sphingomyelinase is known to exist in two forms depending on their intracellular localization and pH optima (13)(14)(15)(16). A Mg 2ϩ -dependent membrane-bound with a neutral pH optima initiates signaling by generating ceramide at or near the vicinity of the plasma membrane. In addition to the membraneassociated enzyme, another cytosolic neutral SMase independent of Mg 2ϩ has been identified and partially purified, which appears to hydrolyze intracellular sphingomyelin stores to initiate signaling (19). The signal initiated by these enzymes is then transmitted further down in the signaling cascade by activation of ceramide-activated protein phosphatases and ceramide-dependent protein kinases (13)(14)(15). In addition to the neutral SMase forms, an acidic SMase form has also been identified, displaying an pH optima around 5, the bulk of which seems to be located at the lysosomes/endosomal compartment. Although it appears that acidic SMase plays a role in signaling, the molecular mechanism of its activation and recruitment during signaling is unclear. Indirect evidence have suggested that DAG generated by PC-PLC activates the acidic enzyme at or near the plasma membrane since inhibitors of PC-PLC prevent activation of the acidic SMase. This hypothesis, which implies a redistribution of the enzyme from the lysosomal compartment to or near the plasma membrane, requires further verification (20,21). Despite the existence of the neutral and acidic SMases, an alkaline form of the enzyme has been described recently, although its role in signaling remains to be defined (22).
There is compelling evidence to propose ceramide as a second messenger in the sphingomyelin pathway similar to DAG in the glycerophospholipid pathway. The role that ceramide fulfills within the cell are numerous and of varied nature (13)(14)(15)(16). It has been shown that ceramide plays a critical role in apoptosis, proliferation, cellular senescence, and gene regulation through activation of transcription factors such as NF-B (20,21). However, the possibility that ceramide may interact with mitochondria leading to production of ROS has not been documented to our knowledge, and constitutes the basis of the present report.
Mitochondria are one of the most important cellular sources of ROS due to its quantitative consumption of molecular oxygen. Since ceramide appears as an important mediator of the effects elicited by TNF and due to the participation of mitochondria in the TNF-induced ROS production (9 -12), the purpose of the present work was to analyze the effects of ceramide and other sphingolipids, on the production of hydrogen peroxide in isolated mitochondria from rat liver. Furthermore, since reduced GSH is the only defense provided to metabolize peroxides generated from the electron transport chain through GSH redox cycle (23), we determined the role of mitochondrial GSH in modulating the production of hydrogen peroxide and its consequences upon incubation of mitochondria with ceramide. Our studies demonstrate for the first time that addition of ceramide to mitochondria results in a dose-dependent increase in hydrogen peroxide, which is prevented when complex I and II of respiration are inhibited. Furthermore, mitochondria from TNF-treated hepatocytes displayed an increased level of ceramide supporting the role of ceramide as an intermediate in the TNF-induced ROS generation from mitochondria. Depletion of matrix GSH prior to exposure to ceramide results in an additional increase of hydrogen peroxide, which peroxidizes lipids from mitochondria resulting in loss of mitochondrial function. These results suggest that ceramide produced in the signaling of TNF is responsible, at least in part, for some of the TNF-induced cytotoxic effects.
Isolation and Culture of Hepatocytes-Hepatocytes were isolated as described previously, plated on rat tail collagen, and cultured in DMEM/F12 (24,25). Cell numbers were determined using a Coulter counter, model multisizer II (Coulter Electronics) and verified by hemocytometer. Cell viability was determined by trypan blue exclusion and by the measurement in the medium of glutathione S-transferase.
Fluorescence Microscopy-Cultured hepatocytes in the presence or absence of TNF or ceramide were incubated with fluorescence probes DCFDA or DHR (2 M) for 1 h, followed by washing to remove excess probes. To quantify the generation of peroxides by fluorescence microscopy, a Nikon Diaphot 300 (Tokyo, Japan) inverted microscope equipped with a CF Fluor 40ϫ objective was used. Fluorescence intensity was detected using a 3CCD camera (model DXC-930P, Sony, Tokyo, Japan) with an attached MPU-F100P intensifier (Sony). Images were recorded on a SR-SS368E (JVC, Tokyo, Japan) video cassette recorder and analyzed with a PC computer equipped with 24-bit Movie Machine II graphics. The limits of the hepatocyte monolayer in the area of observation were traced by a clear field transillumination image. Thereafter, the mean fluorescence intensity of the delimited area was measured in the corresponding fluorescence image. Five random images were analyzed for every experimental condition. Values of fluorescence are the result of subtracting background fluorescence (measured in cultures in the absence of fluorescent probes) from the values obtained in each image referred to a grayscale (0 -255).
Preparation of Mitochondria and Incubation with Ceramides-Rat liver mitochondria were isolated by differential centrifugation (26). Enrichment of mitochondria was ascertained by the specific activity of succinic dehydrogenase found in mitochondria relative to that of homogenate. Mitochondrial integrity was determined by the acceptor control ratio as oxygen consumption in states 3 and 4 of respiration using a Clark oxygen electrode with glutamate/malate or succinate as substrates for respiratory sites for complex I or II as described previously (25,27).
Mitochondria were incubated in a shaker bath a 25°C under ambient air in the presence of mediators of TNF or inhibitors of respiratory complexes for up to 1 h as detailed in the figures. When fluorescence probes were present, incubation was carried out in the dark.
Stock solutions of sphingolipids were made up in dimethyl sulfoxide and stored at Ϫ80°C under nitrogen; when added to aqueous reaction mixtures containing mitochondria, the final concentration of the carrier solvent did not exceed 0.5%. Control mitochondria contained only carrier solvent whose presence did not affect the fluorescence of DCF.
Flow Cytometry Analysis of Mitochondria-All measurements of mitochondria fluorescence and side light scatter (SSC, 90°angle) were made for at least 10,000 events/test using a FACStar flow cytometer (Becton Dickinson, San Jose, CA). Data on mitochondrial fluorescence and light scatter were obtained using a 5-watt argon ion laser tuned at 488 nm and 250 milliwatts. Fluorescence of DCF from oxidation of DCFDA was measured through a 530-nm bandpass filter placed in front of the green photomultiplier tube using a four-decade log amplifier. The mean intensity of the green fluorescence caused by mitochondria incubated with DCF in the presence or absence of ceramide was determined using the Cell Quest software program and expressed as fluorescence channels (scale from 0 to 10,000 arbitrary units). Graphics were plotted using the Cell Quest software program (Becton Dickinson).
Depletion of Mitochondrial GSH-GSH was depleted in vitro by preincubation of mitochondria with DEM (0.2-0.8 mM) or ethacrynic acid (250 M) for 10 -15 min, followed by removal of the agent by washing (2-3 times). Alternatively, mitochondria depleted of GSH from rat liver were prepared by in vivo treatment with BSO (3 mmol/kg/day, intraperitoneal) for 4 days or chronically feeding ethanol in liquid diet for 4 weeks as described (25,28). These treatments decrease mitochondrial GSH to about 50% of control (data not shown). Reduced and oxidized GSH were determined by high performance liquid chromatography as described (24,25).
Determination of Hydrogen Peroxide and Lipid Peroxidation-Hydrogen peroxide measurement was determined spectrofluorometrically using DCFDA. Mitochondria were incubated with the fluorescent probe, 2 M, in the absence or presence of ceramide or other electron transport inhibitors (see figure legends). Fluorescence was determined at 529 nm for emission and 503 nm for excitation, with slit widths of 10 and 5 nm, respectively (27,29). Fluorescence of DCF was correlated with increasing concentrations of hydrogen peroxide allowing determination of hydrogen peroxide as described (27).
Lipid peroxidation was determined by quenching of fluorescence of cis-parinaric acid as described previously (30). Mitochondria treated with C 2 were incubated with parinaric acid (5 g/ml) and fluorescence reading determined at 318 nm for excitation and 410 nm for emission, respectively.
Measurement of Mitochondrial Membrane Permeability Transition-Large amplitude swelling was measured spectrophotometrically by recording absorbance at 540 nm. An increase in mitochondrial swelling results in a decrease in optical density. Isolated rat liver mitochondria were suspended in a buffer consisting of 200 mM sucrose, 10 mM Tris-MOPS, 5 mM succinate, 1 mM potassium phosphate, 2 M rotenone, 1 g/ml oligomycin, 10 M EGTA, pH 7.4, at 25°C. Ceramides (C 2 , C 6 , or C 2 DH) were added to mitochondrial suspension at 1-5 M, and absorbance at 540 nm was determined at 25°C over time. Opening of the pore was induced by the adenine nucleotide translocator ligand atractyloside or by incubation of mitochondria with tert-butylhydroperoxide and Ca 2ϩ and prevented by preincubation with cyclosporin A (5 M) and trifluoperazine (100 M).
Determination of Ceramide-Mitochondria from TNF-treated hepatocytes were isolated by Percoll gradient as described (25). Ceramide was quantified by the diacylglycerol kinase assay as described previously (31,32). Lipids from mitochondria were extracted and dried under nitrogen and resuspended in 100 l of 150 g of cardiolipin, 280 M diethylenetriaminepentaacetic acid, 51 mM octyl-␤-glucopyranoside, 50 mM NaCl, 51 mM imidazole, 1 mM EDTA, 12.5 mM MgCl 2 , 2 mM dithiothreitol, 0.7% glycerol, 70 M ␤-mercaptoethanol, 1 mM ATP, 10 Ci of [␥-32 P]ATP, and 35 g/ml E. coli diacylglycerol kinase at pH 6.5. After 30 min at room temperature, the reaction was stopped by extraction of lipids with 1 ml of chloroform:methanol:1 N HCl (100:100:1) and 170 l of PBS. Major lipid products of the phosphorylation reaction, phosphatidic acid (from diacylglycerol) and ceramide 1-phosphate (from ceramide) were resolved by thin-layer chromatography on Silica Gel 60 plates (Whatman) using chloroform:acetone:methanol:acetic acid:water (10:4: 2:2:1, v/v) as solvent and detected by autoradiography. Incorporated 32 P was quantitated by liquid scintillation counting. The level of ceramide was calculated by comparison with a standard curve generated using known amounts of ceramide type III.
Statistical Analyses-Statistical analyses for comparison of mean values for multiple comparisons between mitochondrial preparations were made by one-way analysis of variance (ANOVA) followed by Fisher's test.

RESULTS AND DISCUSSION
Ceramide Leads to Generation of Reactive Oxygen Species in Intact Cells and Isolated Mitochondria-Ceramide has drawn attention since the description of its role as a sphingolipid second messenger whose levels are increased in cells stimulated by inflammatory cytokines such as TNF. Since one of the characteristic features of the TNF-induced cytotoxicity is mediated by overproduction of ROS, we first determined the effect of direct addition of permeable ceramide analogues, such as C 2 -ceramide, to cultured hepatocytes to monitor its effect on the fluorescence of probes that are sensitive to oxidative stress, such as DHR. Primary cultured hepatocytes were labeled with DHR, washed to remove excess fluorochrome, and analyzed for changes in fluorescence assessing ROS production. Fig. 1 shows a representative fluorescence microscopic photograph of hepatocytes labeled with DHR. Upon incubation of cells with C 2 (5 M), we observed a significant increase (2-3-fold versus control) in the fluorescence of DHR. Because peroxides are the species specifically monitored by such fluorescent probe, these results indicate that in this paradigm, there was a burst of hydrogen peroxide induced by C 2 compared with control cells in the absence of the sphingolipid. Similar results were obtained when hepatocytes were incubated with DCFDA, a non-fluorescent probe, which upon oxidation, mainly by peroxides, is converted to the highly fluorescent derivative DCF (22, 24) (data not shown). Hepatocytes remained viable under these conditions, indicating that the increase in ROS was not a consequence of cell dysfunction. The fluorescence microscopic appearance of hepatocytes incubated with C 2 were reminiscent of the effect that TNF produced in hepatocytes, 2 suggesting that ceramide reproduced in parenchymal cells the increase in ROS that TNF evokes on multiple cell types (9 -12).
Thus, data from intact cells incubated with C 2 leading to overproduction of ROS suggest but do not demonstrate that such effects were either mediated by direct action of ceramide nor were they originated from mitochondria. Involvement of substrates downstream in the signaling of ceramide could have been the effectors of the increased production of ROS. In this regard, activation of transcription factor NF-B by ceramide (20,21,33) would lead to increased expression of genes that contain B sites in their promoter/enhancer, such as nitric oxide synthetase. Rising nitric oxide levels could contribute to generation of other potent oxidants, thus potentially participating in the generation of ROS observed in cells (34).
Evidence has recently been provided that mitochondria from cells exposed to TNF are the main source of ROS generation produced by the cytokine (9 -12). Therefore, we hypothesized that ceramide may directly affect mitochondria leading to production of hydrogen peroxide. To test this hypothesis, and to demonstrate a direct effect of ceramide on mitochondria, we isolated mitochondria from rat liver and examined the effect that incubation with C 2 exerts when monitoring the generation of hydrogen peroxide by flow cytometry using DCFDA as fluorescent probe to follow its conversion to DCF (22,24). Mitochondria were incubated with succinate to drive electron flow directly at succinic dehydrogenase complex. As shown in Fig.  1C, flow cytometric profile of mitochondria labeled with DCFDA displayed a greater fluorescence intensity of DCF upon addition of C 2 (2-fold), reproducing the phenomena observed with intact cells. Similar results but of lesser magnitude were observed when NAD-linked substrates were used instead of succinate (data not shown). Oxygen consumption at states 3 and 4 of respiration (acceptor control ratio) did not differ significantly between control or ceramide-treated mitochondria, indicating that the increased fluorescence of DCF was not the result of unspecific effects due to loss of mitochondrial integrity (data not shown).
In view of the evidence that ceramide acts as a messenger in transmitting the signaling of TNF (13)(14)(15)(16)(17)(18) and on the direct effect of ceramide on mitochondria leading to ROS production, we sought to determine if treatment of hepatocytes with TNF increases the level of ceramide in mitochondria. We first verified that treatment of cultured hepatocytes with TNF resulted in generation of ROS. As seen in Fig. 2, the generation of hydrogen peroxide determined by fluorescence of DCF in hepatocytes labeled with DCFDA increased upon treatment with TNF. Subsequently, mitochondria from these cells were isolated and the level of ceramide determined by the diacylgycerol kinase assay. The mitochondrial fraction was enriched in succinic dehydrogenase (3-4-fold) and de-enriched in lactic dehydrogensase relative to intact cells. Compared with mitochondria isolated from control cells, the mitochondrial fraction from cells treated with TNF revealed a significant increase (2-3fold) in the amount of ceramide (Fig. 2B). Therefore, these data correlate the increase in ceramide in mitochondria of cells treated with TNF with its ability to overproduce ROS.
Our findings demonstrating direct effect of ceramide in mitochondria have extended previous related observations that mitochondria isolated from septic rats generated ROS to a greater extent than mitochondria from control rats mainly from FAD-linked substrates (35). However, these studies did not examined the role of TNF or identify its mediators as causal effectors for the increased hydroxyl radical generation from septic mitochondria. Taken together these data demonstrate that inflammatory cytokines such as TNF leads to increased generation of ceramide associated with mitochondria which by interacting with mitochondria may account for by the increased generation of ROS in intact cells. To our knowledge, the direct effect of ceramide on mitochondria has not been previously described, and therefore we sought to further characterize such phenomena in terms of specificity and mechanism(s).
Structural Specificity of Sphingolipids in Generating Hydro-gen Peroxide from Mitochondria-Since C 2 is a permeable analogue of natural ceramides, we tested if other analogs of C 2 including N-hexanoylsphingosine (C 6 -ceramide) also led to generation of ROS. As shown in Fig. 3, the magnitude of generation of ROS by C 6 was similar to that observed by C 2 . Similar results were also obtained when N-palmitoylsphingosine was used as the effector lipid (data not shown). These results indicate that N-fatty acyl-sphingosine derivatives mediate the increased generation of hydrogen peroxide from mitochondria, regardless of the length of the alkyl moiety. One of the features of sphingolipids such as ceramide is the presence of a trans double bond in atom 4 of sphingosine. Ceramide is formed from dihydroceramide by the introduction of the trans-4,5-double bond (reviewed in Ref. 15). Therefore, we tested if the immediate precursor of ceramide, dihydroceramide, which lacks the trans double bond, mimic the effect of C 2 regarding its ability to generate hydrogen peroxide in mitochondria. As shown in Fig.  3, the dose-dependent effect of dihydroceramide was shifted to the right compared with C 2 ; the increased production of hydrogen peroxide by C 2 -dihydroceramide at 1 M was of similar potency to that of C 2 . However, at lower concentrations (0.25-0.5 M) compared with the effect elicited by C 2 -ceramide, dihydroceramide did not result in generation of hydrogen peroxide. The ability of dihydroceramide to mimic the effect induced by ceramide in the generation of hydrogen peroxide is an intrigu- ing finding. Although in most cases dihydroceramide is considered as an inactive derivative often used as a negative control, there are instances where dihydroceramide evokes cellular responses similar to those exerted by permeable analogues. Thus, studies examining the effect of ceramides on the apoptotic response in P388 cells showed that dihydroceramide induced apoptosis compared with untreated cells, but of less magnitude than that induced by C 2 , suggesting that dihydroceramide is not an inert molecule (36). Other examples illustrating the ability of dihydroceramide to mimic the cellular response elicited by permeable ceramides exist. For instance, hepatic cytochrome P450 2C11 has been shown to be down-regulated by both C 2 and C 2 -dihydroceramide with similar potency (37). Although these studies showed that incubation of hepatocytes with C 2 -dihydroceramide increased the endogenous level of ceramide suggesting that the effect of the former was not direct, this possibility seems unlikely since in this paradigm C 2 -dihydroceramide failed to affect the expression of ␣ 1 -acid glycoprotein, an acute phase protein, whose mRNA levels are up-regulated by C 2 (37). Furthermore, the presence and configuration (cis versus trans) of the 4,5-double bond of sphingolipids does not appear critical for stimulation of cell proliferation in Swiss 3T3 cells (38,39). Thus, these data suggest that the ability of dihydroceramide to reproduce some of the effects elicited by cytokines and permeable ceramides may be dependent on the cellular response studied and the type of cells used (36 -38).
The effect of C 2 -ceramide was dose-dependent, displaying a bifunctional effect starting at 0.25 M and reaching the maximum at 0.5-5 M (Fig. 3B). The mechanism underlying this behavior is unclear, although similar effects have also been seen in other cell types. Hence, studies describing the regulation of fMLP-induced superoxide anion generated by neutrophils found that C 2 at concentrations below 1 M potentiated the generation of this reactive species, whereas at concentrations greater than 1 M inhibited its production (40). In addition, the ability of ceramide to activate and phosphorylate protein kinase C , an atypical protein kinase C isoform, has been shown to be bifunctional (33).
Sphingolipids contain sphingosine as the sugar backbone to which a fatty acid is linked through an amide bond at carbon 2.
Incubation of mitochondria with sphingosine or its precursor sphinganine at 1 M, concentration at which C 2 elicited a maximal increase in hydrogen peroxide, did not result in production of hydrogen peroxide (Figs. 3A and Fig. 4). Only sphingosine at Ն10 M induced a significant increase in hydrogen peroxide. Addition of other sphingolipids, such as sphingomyelin did not increase fluorescence of DCF (Fig. 3A). Incubation of mitochondria with the enzyme responsible for sphingomyelin hydrolysis, SMase that leads to generation of ceramide in cells did not exert any effect in DCF-labeled mitochondria (data not shown). These results suggest that ceramide is not locally produced within mitochondria by action of SMases acting on the sphingomyelin, implicating that even the small fraction of sphingomyelin of the mitochondrial membrane is not accessible to hydrolysis by SMase or that the ceramide that would have been generated in situ had not built up to exert any significant effect on mitochondria.
Although ceramide is one of the lipid mediators that reproduce many of the effects exerted by cytokines such as IL-1 or TNF (13)(14)(15)(16), other lipid molecules such as DAG and arachidonic acid also arise within cells in response to these cytokines. These lipid signals, DAG or arachidonic acid, accumulate in cells in response to cytokines by the action of PC-PLC and phospholipase A 2 , respectively. Incubation of mitochondria with short chain diacylglycerol, 1,2-diacylglycerol, over the same range of concentrations tested for C 2 , did not result in significant generation of hydrogen peroxide (Fig. 4). Arachidonic acid at concentration up to 10 M failed to result in any significant change on the generation of hydrogen peroxide (Table I). Nevertheless, this fatty acid at concentrations exceeding 20 M decreased state 3 while increasing state 4 respiration (data not shown). Similarly to these results with isolated mitochondria, incubation of hepatocytes with either DAG or arachidonic acid did not rise fluorescence of DCF or DHR (data not shown).
The lack of effect of DAG in comparison with ceramide is of interest and adds as another example illustrating the divergent functional behavior of these mediators. In this regard, protein kinase C isoenzyme, which is insensitive to phorbol ester or DAG, becomes activated by ceramide (14,33). Although these differential functions described here for DAG and ceramide might have been predicted based on different structures between glycerolipids (DAG) and sphingolipids (ceramide), there are examples of enzymes that recognize either lipid as substrates. Sphingomyelin synthase, a mammalian enzyme responsible for sphingomyelin synthesis, transfers the phosphocholine to ceramide, generating sphingomyelin from phosphatdidylcholine. The lack of effect of TNF itself or its lipid mediators in generating ROS (Table I) strengthens the hypothesis, supported by our findings, that ceramide is an important link between TNF binding to its receptor at the plasma membrane and the distally evoked generation of ROS from mitochondria (11).

Potentiated Effect of Ceramide and Antimycin A in the Generation of Hydrogen Peroxide and Effect of Blocking Electron
Flow at Complex I and II of Respiration-The generation of hydrogen peroxide from mitochondria arises from superoxide anion upon its dismutation catalyzed by Mn-superoxide dismutase. The production of superoxide anion originates from the ubiquinone, Q cycle, of complex III where one electron from ubisemiquinone is transferred directly to molecular oxygen. Reduction of Q to ubiquinol occurs at NADH dehydrogenase and succinate dehydrogenase complexes. The transfer of one electron from ubiquinol to the cytochrome bc 1 complex catalyzed by the Rieske iron-sulfur center generates ubisemiquinone. Ubisemiquinone by transferring a second electron to cytochrome b 566 is oxidized to Q. Cytochrome b 566 can transfer this electron to cytochrome b 562 reducing ubisemiquinone to ubiquinol. When this process is blocked by AA, the electron is passed directly to molecular oxygen, which is expected to result in generation of superoxide anion. Therefore, we determined the effect of interruption of electron flow from cytochrome b 562 to ubisemiquinone by AA on the magnitude of increase of hydrogen peroxide produced by C 2 . As previously shown, AA led to an increased generation of hydrogen peroxide (2-3-fold) determined as fluorescence of DCF of similar magnitude to that of C 2 (Fig. 5) (27). Interestingly, the addition of C 2 to AAsupplemented mitochondria resulted in an additive production of hydrogen peroxide compared with either of these separately.
To further support the view that inhibition of electron flow at the ubiquinone pool of complex III is the major site of hydrogen peroxide generation by C 2 , we inhibited electron flow at complexes I and II with known blockers of electron transfer at these sites, i.e. rotenone and TTFA, respectively, separately or in combination. When mitochondria energized with succinate were incubated with rotenone and TTFA, the fluorescence of DCF did not increase, indicating lack of production of hydrogen peroxide. However, inhibition of electron flow at these complexes significantly prevented the increase in hydrogen peroxide resulting from incubation of mitochondria with C 2 . Similarly, blocking electron flow at complexes I and II did also partially prevent the increase of hydrogen peroxide resulting from C 2 plus AA, compared with the combined presence of the two. This phenomena has been described also in the generation of ROS of rat liver mitochondria when incubated with AA as well as for splenic T lymphocytes committed to programmed cell death, where generation of ROS was diminished by inhibiting mitochondrial electron transport with rotenone, highlighting the role of the Q cycle as the electron source for ROS (27,41). The additive effect of C 2 and AA and the similar effect of complexes I and II blockers in preventing the effect of C 2 in inducing hydrogen peroxide suggest that C 2 favors the electron transfer to molecular oxygen at or near same center where AA acts in the Q cycle of complex III. Further evidence in favor of this site as the main generator of DCF increase came when mitochondria were incubated with myxothiazol. This compound inhibits oxidation of ubiquinol to ubisemiquinone by the Rieske iron-sulfur center of cytochrome bc 1 complex and is expected to block superoxide anion formation. Accordingly, myxothiazol prevented the increase in DCF caused by C 2 -ceramide (Fig. 5), indicating that the Q cycle of complex III is a significant source of ROS produced by ceramide.
Lack of Involvement of Mitochondrial Permeability Transition in the Burst of Hydrogen Peroxide by Ceramide-Mitochondrial membrane permeability is a phenomena that has been studied for decades in isolated mitochondria. It has been proposed that such process is a critical mechanism involved in cell damage. The mitochondrial membrane permeability is characterized by a sudden increase in the permeability of the inner mitochondrial membrane to small solutes. The permeability transition occurs through the opening of a transmembrane pore in the inner mitochondrial membrane. The opening of the pore is facilitated by loading mitochondria with calcium, pH, oxidation of thiols, and by activation of the adenine nucleotide translocator. This process collapses ion gradients across the inner mitochondrial membrane, leading to mitochondrial depolarization, loss of oxidative phosphorylation, and generation of ROS (41,42). Therefore, in further defining the mechanism(s) leading to overproduction of ROS by ceramide, we determined if the opening of the pore responsible for the mitochondrial permeability transition contributes to the generation of ROS. Compared with positive inducers of permeability transition such as atractyloside (Fig. 6) or tert-butylhydroperoxide (data not shown), which induce a fall in absorbance at 540 nm indicating opening of the pore, mitochondria incubated with C 2 (1-5 M) did not reveal any significant change in optical density at 540 nm, which was maintained over time at levels similar to control mitochondria (Fig. 6). The opening of the pore induced by atractyloside was prevented by cyclosporin A, as seen by the maintenance of the optical density at 540 nm (data not shown); however, this inhibitor did not affect the absorbance recording at 540 nm of mitochondria in the presence of C 2 , indicating lack of swelling of mitochondria incubated with C 2 (Fig. 6).
The time pattern of DCF fluorescence of mitochondria in the presence of C 2 did not parallel that of A 540 nm , since the fluorescence of DCF increased over time despite lack of opening of the pore, suggesting that the former was not caused by engagement of the latter. Beyond 30 min of incubation, there was a fall in absorbance in the presence of C 2 indicating swelling of mitochondria; however, the onset in the opening of the pore was preceded by the increase in DCF fluorescence.
These results suggest that the generation of hydrogen peroxide induced by C 2 is not the consequence of increased membrane permeability leading to the mitochondrial swelling and generation of ROS; in fact, our findings indicate that the overproduction of ROS induced by C 2 would result in activation of the pore as indicated by the time relationship of these two mutually regulated processes. The opposite has also been noticed, since it has been shown in lymphocytes committed to cell death that engagement of permeability transition leads to ROS overproduction (41). The control of the opening of the pore by ROS imply the existence of critical sulfhydryls that are subject to redox regulation. This constitute the basis for the opening of the pore in the presence of strong prooxidants such as tertbutylhydroperoxide. Accordingly, in view of the reciprocal regulation of the permeability transition and ROS, our results suggest the possibility that upon generation of ROS induced by ceramide engagement of the permeability transition would entail as an amplification wave-like mechanism contributing to the overproduction of ROS generated by ceramide in response to inflammatory cytokines.
Mitochondrial GSH Depletion Results in Loss of Mitochondrial Function by Oxidative Stress Induced by Ceramide-Hydrogen peroxide generated within the electron transport chain can undergo two possible fates: conversion to hydroxyl radical with the participation of transition metals in the Haber-Weiss reaction or reduction to water by the catalysis of GSH peroxidases with the required participation of reduced GSH as cofactor. Since mitochondrial GSH is the only defense to metabolize peroxides, depletion of GSH prior to exposure of these mitochondria to C 2 would be expected to result in a potentiating increase in hydrogen peroxide. Thus, we have determined the magnitude of hydrogen peroxide production by C 2 and the degree of lipid peroxidation as consequence of the oxidative stress induced by C 2 in mitochondria depleted of GSH. We have used several maneuvers to deplete GSH in mitochondria: by in vitro incubation of mitochondria with ethacrynic acid, mitochondrial GSH is depleted to about 50% of control levels (Fig.  7A). The same degree of depletion was achieved when DEM was used (data not shown). In addition to these in vitro maneuvers, depletion was achieved by in vivo administration of BSO, a selective inhibitor of the ␥-glutamylcysteine synthetase, which leads to a cellular depletion of GSH including mitochondria (43) or after ethanol feeding to rats, which results in a selective depletion of mitochondrial GSH as consequence of impaired transport of GSH from cytosol into mitochondria (28). GSH-depleted mitochondria incubated with C 2 revealed greater production of hydrogen peroxide (75-80%) compared with mitochondria with repleted levels of GSH. This magnitude in the generation of hydrogen peroxide was comparable to that observed by combination of C 2 and AA in GSH-repleted mitochondria (Fig. 7B).
To evaluate the consequences of increased level of hydrogen peroxide under these circumstances, we examined the fluorescence of cis-parinaric-labeled mitochondria to determine the degree of lipid peroxidation (30). In the GSH-repleted mitochondria, C 2 did not significantly lead to increased loss of cis-parinaric compared with control mitochondria, indicating lack of lipid peroxidation. However, the degree of lipid peroxidation was increased 2-fold upon addition of C 2 to mitochondria that have been depleted of GSH. In these conditions, GSH depletion prior to exposure to C 2 led to a significant loss of complex IV activity determined as cytochrome c oxidase as parameter subject to inactivation by ROS (44). Similar consequences in terms of generation of hydrogen peroxide, lipid peroxidation, and loss of complex IV activity by C 2 were observed when mitochondria from BSO-or ethanol-treated rats were used (data not shown). The equivalent results observed between the in vitro or in vivo-induced depletion of GSH discard the possibility that the outcome obtained from ethacrynic acid-treated mitochondria were caused by unspecific effects of the toxicant.
Recent studies by Goosens et al. (11) provided indirect evidence that mitochondrial GSH was critical in scavenging the ROS generated by TNF in murine fibrosarcoma cell line L929, based on differential effects of DEM versus BSO in accelerating cytotoxicity. However, these studies did not report the level of mitochondrial GSH in L929 cells after these maneuvers. Our findings have demonstrated the critical importance of mitochondrial GSH in scavenging the ROS produced in the organelle as consequence of interference of electron transport by C 2 at complex III. Similar conclusion were obtained when oxidative stress in isolated rat liver mitochondria was induced by blocking electron flow at complex III of respiration (27). Recently, a critical role of mitochondria has been deciphered in splenic lymphocytes committed to programmed cell death induced by a variety of stimulus, including ceramide, where a loss of mitochondrial transmembrane potential and generation of ROS constitute an important feature of early apoptosis (41), although these studies did not address the effect of ceramide on isolated mitochondria. In light of our findings, it could be speculated that depletion of GSH in mitochondria prior to exposure to ceramide could accelerate or increase the degree of apoptosis. The fact that ceramide interacts with components of the complex III of the electron transport chain favoring the production of ROS highlights the pivotal role of GSH as the primary line of defense of mitochondria due to virtual lack of catalase activity. Thus, mitochondrial GSH status will be a critical modulator of mitochondrial function and cell viability and, hence, in diseases and/or tissue injury mediated by oxidative stress in mitochondria, mitochondrial GSH depletion will accentuate the adverse effects of ROS (10,24,25,27,43,45,46). Since mitochondrial GSH arises by the existence of an ATP-dependent carrier, which translocates cytosol GSH into the matrix, it would be critical to characterize its nature and properties at a molecular level (47).
In summary, we have determined the capability of ceramide to result in increased generation of hydrogen peroxide in iso- FIG. 7. Mitochondrial matrix depletion of GSH and consequences of ceramide on generation of hydrogen peroxide and peroxidation. A, in vitro depletion of GSH was accomplished by incubation of mitochondria with ethacrynic acid (100 M) for 15 min followed by washing of mitochondria. GSH level was determined by high performance liquid chromatography as described under "Materials and Methods." B, GSH repleted or depleted mitochondria (1 mg/mg) were then incubated with DCFDA to monitor fluorescence of DCF upon addition of AA or C 2 -ceramide (5 and 1 M, respectively) or its combination for 60 min. Hydrogen peroxide was determined from DCF fluorescence as described under "Materials and Methods." C, parallel incubations of GSH repleted or depleted mitochondria were labeled with cis-parinaric acid according to the protocol described under "Materials and Methods" to determine degree of lipid peroxidation and expressed as percentage loss of initial fluorescence intensity. D, cytochrome oxidase activity of complex IV of respiration was determined spectrophotometrically by oxidation of added ferrocytochrome c in the presence of C 2 -ceramide. Results are mean Ϯ S.D. of four mitochondrial preparations. *, p Ͻ 0.005 versus control GSH-repleted mitochondria; **, p Ͻ 0.05 versus corresponding condition from GSH-repleted mitochondria. lated mitochondria. Our results clearly demonstrate that ceramide exerts a direct effect in mitochondria describing a new functional role of sphingolipids as inducers of oxidative stress. Since ceramide is an intermediate generated intracellularly upon stimulation of cells with inflammatory cytokines, our findings demonstrate a role of ceramide in mediating the cytotoxicity of TNF by increased generation of ROS. Furthermore, the present studies identify mitochondria as a primary target of ceramide leading to generation of ROS by interacting with complex III of electron transport chain. GSH in mitochondria being the only defense to cope with deleterious effects of ROS produced within mitochondria stands as a critical preventive factor whose depletion or limitation may be of significance in amplifying the cytotoxic effect of TNF.