Oxygen Tolerance and Coupling of Mitochondrial Electron Transport*

Oxygen is critical to aerobic metabolism, but excessive oxygen (hyperoxia) causes cell injury and death. An oxygen-tolerant strain of HeLa cells, which proliferates even under 80% O 2 , termed “HeLa-80,” was derived from wild-type HeLa cells (“HeLa-20”) by selection for resistance to stepwise increases of oxygen partial pressure. Surprisingly, antioxidant defenses and susceptibility to oxidant-mediated killing do not differ between these two strains of HeLa cells. However, under both 20 and 80% O 2 , intracellular reactive oxygen species (ROS) pro- duction is significantly ( (cid:1) 2-fold) less in HeLa-80 cells. In both cell lines the source of ROS is evidently mitochondrial. Although HeLa-80 cells consume oxygen at the same rate as HeLa-20 cells, they consume less glucose and produce less lactic acid. Most importantly, the oxy-gen-tolerant HeLa-80 cells have significantly higher cytochrome c oxidase activity ( (cid:1) 2-fold), which may act to deplete upstream electron-rich intermediates responsi-ble for ROS generation. Indeed, preferential inhibition of cytochrome c oxidase by treatment with n -methyl protoporphyrin (which selectively diminishes synthesis of heme a in cytochrome c oxidase) enhances ROS production and abrogates the oxygen tolerance of the HeLa-80 cells. Thus, it appears that the remarkable oxygen tolerance of these cells derives from tighter coupling of the electron transport chain. the centrifuged at 14,000 (cid:2) g 20 min at 4 obtain crude mitochon- drial pellets. The total carbonyl content of mitochondrial proteins was determined by reaction with 2,4-dinitrophenylhydrazine using a slight modification of the method of Levine et al. (28). The mitochondrial pellets containing (cid:1) 300 (cid:1) g of protein were incubated with 500 (cid:1) l of 10 m M 2,4-dinitrophenylhydrazine in 2 M HCl at room temperature for 1 h with intermittent mixing. Following addition of 500 (cid:1) l of 20% (w/v) ice-cold trichloroacetic acid, the samples were centrifuged at 11,000 (cid:2) g for 3 min. The supernatant was discarded and the pellets were washed 3 times with 1 ml of ethanol:ethyl acetate (1:1) to remove unreacted 2,4-dinitrophenylhydrazine. The protein pellet was redissolved in 0.6 ml of 6 M guanidine with 20 m M K 2 HPO 4 (pH adjusted to 2.3 with 20% trichloroacetic acid) at 37 °C and any insoluble material was removed by centrifugation at 11,000 (cid:2) g for 3 min. The carbonyl content in the supernatant was measured spectrophotometrically at 375 nm. Results were calculated using (cid:2) (cid:4) 22,000 M (cid:6) 1 cm (cid:6) 1 and normalized for protein concentration. Cells (cid:1) 5 (cid:2) 6 trypsin/EDTA HeLa cells (initial 5 (cid:2) 10 per 10-cm dish) were treated with 5 (cid:1) M NMP for 9 days under both normoxia and hyperoxia. Medium was changed every 4 days. A previous study showed that the enzymatic activity of COX can be decreased by as much as 85% in cells grown in the presence of NMP (37). Twelve h after initial plating, cells were treated with 5 (cid:1) M NMP in complete medium under normal conditions (20% O 2 ) for 24 h, prior to exposure to 80% O 2 . Control cultures were carried out under 20% O 2 to detect possible drug effects on cell growth. M DCFDA in HBSS; *, p (cid:9) 0.01, n (cid:4) 6). Under 80% O 2 , HeLa-20 cells produced more ROS compared with that under 20% O 2 (#, p (cid:9) 0.05, n (cid:4) 6). B , ROS production detected by the oxidation of the fluorescent probe dihydroethidium. These results are very similar to those obtained with DCFDA ( A ) (4 h incubation with 100 (cid:1) M dihydroethidium in complete culture medium; *, p (cid:9) 0.01, n (cid:4) 4). Again, more ROS generation was detected in HeLa-20 cells under 80% O 2 (#, p (cid:9) 0.05, n (cid:4) 4). C , ROS production as assessed by the oxidation of Amplex® Red. Under hyper- oxia, ROS production increased (cid:1) 2-fold compared with normoxia in the wild-type HeLa-20 cells (#, p (cid:9) 0.05, n (cid:4) 4) and was significantly higher in HeLa-20 versus HeLa-80 cells under both 20% O 2 and 80% O 2 (3 h incubation with 50 (cid:1) M Amplex® Red in Opti-MEM medium supple- mented with 4% fetal bovine serum; *, p (cid:9) 0.05, n (cid:4) 4).

Oxygen is crucial to aerobic metabolism, but excess oxygen or, more likely, reactive oxygen species (ROS) 1 generated under hyperoxic conditions, will cause cell injury and death. Damage to cells and tissues caused by hyperoxia is clinically important. For example, both lung damage and retrolental fibroplasia occur in premature infants given oxygen as therapy for pulmonary insufficiency. The risk of these complications is further amplified by the immaturity of cellular antioxidant defenses in premature infants (1). Similar pulmonary damage is also ob-served in adults who, on prolonged exposure to high partial pressures of inhaled O 2 , exhibit cough, shortness of breath, decreased vital capacity, and increased alveolar-capillary permeability (2)(3)(4).
Despite decades of work, it is still not known precisely how hyperoxia causes damage to cells and tissues. Nonetheless, it is commonly believed that free radicals play a key role in the pathophysiology of oxygen toxicity and cellular damage is probably mediated by increased production of ROS (5). This excessive production of ROS likely derives from the mitochondria that, under conditions of high oxygen, exhibit increased electron leak from the electron transport chain (5)(6)(7). We have approached the question of the nature of hyperoxic cell damage using a special line of HeLa cells selected for resistance to stepwise increases in the partial pressure of O 2 . These oxygentolerant HeLa cells are able to survive and grow under 80% O 2 , a partial pressure of oxygen under which normal HeLa cells and most other mammalian cells not only stop growing but die (8).
If enhanced ROS production is, in fact, at the root of oxygen toxicity, then there are at least two ways in which cells might avoid oxygen toxicity: hypertrophied antioxidant defenses or lesser production of ROS. However, important antioxidant defenses, which include glutathione and related enzymes (glutathione reductase, glutathione peroxidase, thioredoxin/thioredoxin reductase, and peroxiredoxins), ascorbic acid, vitamin E, catalase, superoxide dismutases (both the cytoplasmic Cu/Znsuperoxide dismutase and the mitochondrial Mn-superoxide dismutase) in the oxygen-tolerant HeLa cells are nearly identical to those in the wild-type HeLa-20 cells (8). This has two major implications. First, it indicates that the tolerance of these cells to high oxygen is probably not because of enhanced defenses against ROS. Second, many of these antioxidant enzymes are known to be induced by ROS. The fact that these enzymes display normal activities even in the tolerant HeLa cells exposed to high oxygen suggests that these cells somehow manage to avoid exaggerated ROS production even when exposed to high partial pressures of oxygen.
Even under normoxic conditions, as much as 2-4% of the reducing equivalents escape the respiratory chain, initially as superoxide (O 2 . ) and, hence, hydrogen peroxide (H 2 O 2 ) (6) (although other estimates of the extent of this electron leak are lower (9 -11)). Under hyperoxic conditions, mitochondrial ROS generation increases as a linear function of the oxygen tension (12) and this has been implicated in cell dysfunction and death associated with hyperoxia (13). Therefore, in attempting to understand the basis of oxygen tolerance in HeLa-80 cells, we focused our attention on ROS production. We find that the oxygen-tolerant HeLa cells produce less ROS under both normoxic and hyperoxic conditions. In both wild-type and oxygentolerant HeLa cells, the protonophoric uncoupler carbonyl cyanide m-chlorophenylhydrazone (CCCP) (a lipid soluble weak acid that carries protons across the mitochondrial inner membrane, thereby dissipating the proton gradient) almost completely blocks ROS production. This is probably because the consequent acceleration of electron transport leads to depletion of electron-rich intermediates (i.e. ubisemiquinone) at complexes III and I. This suggested to us that something similar might have somehow occurred in the oxygen-tolerant HeLa-80 cells. Indeed, although the activities of complexes I-III are very similar in HeLa-20 versus HeLa-80 cells, cytochrome c oxidase (COX) is ϳ2-fold higher in the oxygen-tolerant HeLa-80 cells. Furthermore, preferential blockade of this terminal complex of the mitchondrial electron transport chain, by treatment with nmethyl protoporphyrin (which selectively diminishes synthesis of heme a in cytochrome c oxidase), restores oxygen sensitivity of the HeLa-80 cells. These observations indicate that the oxygen tolerance of HeLa-80 cells arises from tighter "coupling" of the electron transport chain, thereby explaining both the decreased ROS production and increased oxygen tolerance of these unusual cells.
Conditions of Cell Culture-HeLa cells were routinely cultured in Ham's F-10 medium supplemented with 1 mM glutamine and 10% (v/v) heat-inactivated fetal bovine serum under 20% O 2 (normoxia) or 80% O 2 (hyperoxia) with 5% CO 2 at 37°C. For routine passage, cells were washed with phosphate-buffered saline and lifted with 0.05% trypsin, 0.02% EDTA in phosphate-buffered saline. Stock cultures were grown in polystyrene T-75 culture flasks in a Forma Scientific incubator under an atmosphere of 20% O 2 , 5% CO 2 . Exposure to hyperoxia was performed with cells grown in 10-cm dishes under an atmosphere of 80% O 2 , 15% N 2 and 5% CO 2 contained in a specially designed gas-tight modular incubator chamber (Billups-Rothenberg, Inc., Del Mar, CA). Because susceptibility to hyperoxia may vary with cell density (14), the initial cell number was routinely adjusted to ϳ1 ϫ 10 5 in F-10 medium. The sealed chamber was placed in a standard tissue culture incubator and the gas was replenished every 48 h. The survival and growth of cells were assessed over a period of 10 days by counting cell numbers in marked sectors of the culture dishes using light microscopy.
It is important to note that the oxygen tolerance of HeLa-80 is a stable characteristic. Even after these cells have been continuously passed in normoxic culture for more than two months, they retain resistance to 80% O 2 . To avoid artifacts that might be introduced by the tendency of cell cultures to "drift," cultures of both cell types were replenished from liquid nitrogen stabilates every 30 days and the oxygen tolerance of each new culture was checked.
Peroxide-mediated Cytotoxicity-To determine whether oxygen tolerance might be associated with differences in oxidant sensitivity, the cytotoxic effects of both H 2 O 2 and t-butyl hydroperoxide were measured. For these experiments, wild-type HeLa-20 and oxygen-tolerant HeLa-80 cells were grown in 48-well plates (initial density ϳ2 ϫ 10 4 cells per well in 0.5 ml of complete medium) for 3 days. When cells were Ͼ90% confluent, the medium was replaced by 1.0 ml of Hank's balanced salt solution (HBSS) and the cells were exposed to H 2 O 2 or t-butyl hydroperoxide for 2 h at 37°C without CO 2 . The HBSS was then replaced with complete medium. Following a 10-h culture under normal conditions (20% O 2 , 5% CO 2 at 37°C), Alamar Blue was added and, after a 4-h incubation, fluorescence of the reduced dye was measured at excitation of 530 nm and emission of 590 nm (15-17) using a spectrofluorometric plate reader (Molecular Devices Corp., Sunnyvale, CA). Cell viability was calculated as the ratio of surviving cells to the original cell population.
Estimation of ROS Production-ROS production was assessed by the oxidation of DCFDA, dihydroethidium, and Amplex® Red to ensure that the same proportionate differences were evident regardless of the probe employed. To estimate ROS production with DCFDA, cells were plated onto 48-well plates at an initial density of 2 ϫ 10 4 cells per well and grown for 3 days. After cells were Ͼ90% confluent, they were washed 3 times with HBSS and, in some cases, treated with various inhibitors of respiration in 0.5 ml of HBSS for 5 min. Following the addition of 10 M DCFDA (final concentration), the appearance of DCF fluorescence was followed continuously using a thermostated plate reading spectrofluorometer, typically for 1 h, at excitation of 486 nm and emission of 530 nm. Cell protein in each well was measured using the bicinchoninic acid reaction (18) (Pierce) and the DCF fluorescence was corrected for variations in total protein between wells.
Generation of ROS was also evaluated under both 20 and 80% O 2 using the oxidation of dihydroethidium (19). Cells were plated and grown as for experiments with DCFDA (above). When the cells were Ͼ90% confluent, fresh medium was added containing 100 M dihydroethidium with or without 10 M of the protonophoric uncoupler, CCCP. Dihydroethidium oxidation permits estimates of ROS formation over longer periods of time in complete culture medium and the product, ethidium, is held within the cell via intercalation into nucleic acids. After incubation with dihydroethidium for 4 h, cells were washed 3 times with HBSS and ethidium fluorescence was detected at excitation of 520 nm and emission of 610 nm. Again, the relative fluorescence was corrected for variations in cell protein between individual wells. A recent report suggests that the ethidium fluorescence specifically representing superoxide is best measured with excitation at 480 nm and emission at 567 nm (20). To ensure that the results obtained at our chosen wavelengths with excitation of 520 nm and emission of 610 nm were accurate, we directly compared the two conditions. We found a linear relationship between the fluorescence readings of these two pairs of wavelengths (excitation of 480 nm and emission of 567 nm versus excitation of 520 nm and emission of 610 nm, r ϭ 0.99; n ϭ 4).
A third fluorescent dye, Amplex® Red, was also used to estimate cellular ROS generation. In the presence of horseradish peroxidase and H 2 O 2 , Amplex® Red is oxidized to the highly fluorescent product, resorufin (21). For these experiments, cells were incubated with assay medium (phenol-free Opti-MEM containing 4% fetal bovine serum) in the presence of 50 M Amplex® Red and 0.1 unit/ml horseradish peroxidase under 20 or 80% O 2 for 3 h. Fluorescence was detected at excitation at 530 nm and emission at 590 nm. A background fluorescence measured in the absence of cells was subtracted from experimental values and the relative fluorescence was corrected for measured cell protein as above.
Lactate Production-Cells cultured as above were incubated for periods of up to 24 h, and lactate accumulation in the medium was determined. These assays were carried out using lactate dehydrogenase-mediated conversion of lactate to pyruvate and hydrazine as a trap for the product (22). The change in absorbance at 340 nm was measured spectrophotometrically and results were expressed as nanomole of lactate formed per milligram of cell protein per unit time.
Glucose Consumption-The consumption of glucose by the same cultures was measured using hexokinase-mediated formation of glucose 6-phosphate followed by glucose-6-phosphate dehydrogenase-dependent reduction of NADP to NADPH. The change in absorbance at 340 nm was measured spectrophotometrically and results were expressed as nanomole of glucose consumed per milligram of cell protein per unit time (23).
Glutamine Consumption-The consumption of glutamine was measured as described by Passonneau and Lowry (24) in a two-step enzymatic reaction. The first reaction was carried out by glutaminasemediated formation of glutamate at pH 4.9. The second reaction was followed by glutamate dehydrogenase-dependent reduction of NAD ϩ to NADH. The change in absorbance at 340 nm was measured spectrophotometrically and results were expressed as nanomole of glutamine consumed per milligram of cell protein per unit time (25).
Oxygen Consumption-HeLa cell cultures in T-75 flasks (Ͼ90% confluent) were washed with phosphate-buffered saline, detached with trypsin/EDTA, and washed (1,000 ϫ g x 10 min) in complete culture medium. Oxygen consumption was measured using a Gilson Oxygraph with a Clark electrode (Yellow Springs Instruments Co., Yellow Springs, OH). Respiration rates were measured using 3-5 ϫ 10 6 cells suspended in a total volume of 3.0 ml of Ham's F-10 medium containing 10% fetal bovine serum and supplemented with 17 mM glutamate at 37°C. A starting O 2 concentration of 240 M was assumed based on O 2 solubility at sea level at 37°C (26).
Estimation of Mitochondrial Protein Carbonyls-Mitochondria were prepared by differential centrifugation using a modification of the procedure described by Fuller et al. (27). Briefly, cells grown on 10-cm dishes were washed twice with cold mitochondrial isolation buffer (250 mM sucrose, 10 mM Tris-HCl, pH 7.8, 1 mM EDTA, pH 8.0). The cells were diluted in this isolation buffer to 8 ϫ 10 6 cells/ml and disrupted by nitrogen cavitation at 4°C using an ice-cooled cell disruption bomb (Parr Instrument Co., Moline, IL). Following removal of large debris by centrifugation at 500 ϫ g for 10 min at 4°C, the supernatants were centrifuged at 14,000 ϫ g for 20 min at 4°C to obtain crude mitochondrial pellets. The total carbonyl content of mitochondrial proteins was determined by reaction with 2,4-dinitrophenylhydrazine using a slight modification of the method of Levine et al. (28). The mitochondrial pellets containing ϳ300 g of protein were incubated with 500 l of 10 mM 2,4-dinitrophenylhydrazine in 2 M HCl at room temperature for 1 h with intermittent mixing. Following addition of 500 l of 20% (w/v) ice-cold trichloroacetic acid, the samples were centrifuged at 11,000 ϫ g for 3 min. The supernatant was discarded and the pellets were washed 3 times with 1 ml of ethanol:ethyl acetate (1:1) to remove unreacted 2,4-dinitrophenylhydrazine. The protein pellet was redissolved in 0.6 ml of 6 M guanidine with 20 mM K 2 HPO 4 (pH adjusted to 2.3 with 20% trichloroacetic acid) at 37°C and any insoluble material was removed by centrifugation at 11,000 ϫ g for 3 min. The carbonyl content in the supernatant was measured spectrophotometrically at 375 nm. Results were calculated using ⑀ ϭ 22,000 M Ϫ1 cm Ϫ1 and normalized for protein concentration.
Aconitase Assay-Cells (ϳ5 ϫ 10 6 ) were detached with trypsin/EDTA and washed with HBSS (1,000 ϫ g for 10 min). The cells were suspended in 1 ml of 50 mM Tris-HCl, pH 7.4, containing 0.6 mM MnCl 2 and disrupted by 10 1-s bursts from a microtip sonic oscillator (Fisher Scientific, Pittsburgh, PA). The lysate was transferred to a chilled 1.5-ml Eppendorf tube. MnCl 2 was included in the lysis buffer to limit the inactivation of aconitase by O 2 . during extract preparation (29). Cell extracts were clarified by centrifugation for 10 min at 2,000 ϫ g, and the supernatants were promptly assayed for total aconitase activity, following the linear absorbance change at 340 nm at 25°C in a 1.0-ml reaction mixture containing 50 mM Tris-HCl, pH 7.4, 5 mM sodium citrate, 0.6 mM MnCl 2 , 0.2 mM NADP, 2 units of isocitrate dehydrogenase, and 50 g of extract protein. One milliunit of aconitase activity was defined as the amount catalyzing the formation of 1 nmol of isocitrate per min. An initial lag in NADPH formation was seen in samples with low aconitase activity, which presumably reflects the delayed accumulation of cisaconitate (30). In these cases, the linear rates during the latter half of a 60-min assay were used for the determinations of aconitase activity. Results were corrected for measured protein concentrations. COX Assay-Lysates of fresh cells (directly counted and adjusted to 6 ϫ 10 6 cells/ml) were prepared by suspension in 0.25 M sucrose, 40 mM potassium chloride, 2 mM EGTA, 1 mg/ml bovine serum albumin, and 20 mM Tris-HCl, pH 7.2, and disruption by 10 1-s bursts from a microtip sonic oscillator (Fisher Scientific, Pittsburgh, PA). The lysate was centrifuged at 2,000 ϫ g for 10 min. The pellet was discarded and the supernatant was used for COX assay (31). Assays contained ϳ50 g of protein and were performed at 37°C in 1-ml reaction volumes. The assay involved the addition of 60 M ferrocytochrome c in an isosmotic medium (10 mM KH 2 PO 4 , pH 6.5, 1 mg/ml bovine serum albumin, 0.3 M sucrose) containing 2.5 mM n-dodecylmaltoside to permeabilize mitochondrial membranes (31). The activity was calculated from the rate of decrease in absorbance of ferrocytochrome c at 550 nm (⑀ ϭ 19.1 mM Ϫ1 cm Ϫ1 ) (32) and results were normalized for protein. In addition, because accurate estimations of COX activity could be compromised by variations in mitochondrial yield and integrity, results were also normalized using an invariant marker of mitochondrial enzyme activity, citrate synthase (see below).
Citrate Synthase Assay-This assay was performed as described by Williams et al. (33) with minor modifications. It is based on the reaction of oxaloacetate and acetyl-CoA to produce coenzyme A (CoASH), the latter being detected by dithiobis(nitrobenzoic acid), which reacts with sulfhydryls in CoASH producing free thionitrobenzoate. Eight hundred l of 50 mM potassium phosphate buffer (K 2 HPO 4 and KH 2 PO 4 , pH 7.4) was warmed to 37°C and then 10 l of crude mitochondrial extract was added along with dithiobis(nitrobenzoic acid) and acetyl-CoA in final concentrations of 100 M each. The reaction was started by addition of 100 M oxaloacetate. The change in absorbance at 405 nm was monitored for 30 min (reflecting the formation of thionitrobenzoate) and results were expressed as nanomole/min/mg of protein. The citrate synthase activity was calculated from the change in absorbance using ⑀ ϭ 13.6 mM Ϫ1 cm Ϫ1 (34).
Cell Culture in the Presence of n-Methyl Protoporphyrin (NMP)-NMP, an inhibitor of ferrochelatase, preferentially decreases the synthesis of heme a and, therefore, the activity of cytochrome c oxidase (35,36). HeLa cells (initial density ϳ5 ϫ 10 4 per 10-cm dish) were treated with 5 M NMP for 9 days under both normoxia and hyperoxia. Medium was changed every 4 days. A previous study showed that the enzymatic activity of COX can be decreased by as much as 85% in cells grown in the presence of NMP (37). Twelve h after initial plating, cells were treated with 5 M NMP in complete medium under normal conditions (20% O 2 ) for 24 h, prior to exposure to 80% O 2 . Control cultures were carried out under 20% O 2 to detect possible drug effects on cell growth.

RESULTS
HeLa-80 Cells Are Tolerant to Hyperoxia-We began by affirming the previously reported oxygen tolerance of the HeLa-80 cells (8). As shown in Fig. 1 (and previously reported  (8)), when cultured under 80% O 2 and 5% CO 2 at 37°C, wildtype HeLa-20 cells ceased growing after 5-6 days and then underwent extensive cell death. In contrast, the HeLa-80 cells grew normally throughout 10 days of exposure to 80% O 2 . As also shown in Fig. 1, both HeLa-20 and HeLa-80 cells exhibited normal replication under normoxic conditions.
HeLa-20 and HeLa-80 Cells Are Equally Susceptible to Oxidant-mediated Killing-We reaffirmed earlier reports (8,38) that there are no significant differences in antioxidant enzymes such as superoxide dismutases, catalase, reduced glutathione, glutathione reductase, and glutathione peroxidase, between these two strains of HeLa cells (results not shown). In view of earlier reports that elevated heme oxygenase-1 might protect against hyperoxia-induced lung damage (39, 40), we measured total heme oxygenase activity. Once again, no significant differences were found (data not shown). As a more direct test of the overall oxidant defenses of these two cell lines, we tested Oxygen-tolerant HeLa-80 Cells Exhibit Decreased ROS Production in Both Normoxia and Hyperoxia-The lack of significant differences in antioxidant defenses between these two cell lines directed our attention to possible changes in the production of ROS by HeLa-20 versus HeLa-80 cells. This was first estimated using the fluorescent probe DCFDA. This cell-permeable non-fluorescent probe enters cells by passive diffusion whereupon the two acetate groups are esterolytically cleaved. Within the cell, the probe can be oxidized in a reaction that requires H 2 O 2 and a peroxidase or pseudoperoxidase (such as cytochrome c or redox-active iron) yielding highly fluorescent 2Ј,7Ј-dichlorofluorescein. The results (Fig. 3A) indicate that the steady-state production of ROS by the HeLa-80 cells is substantially less than that of the HeLa-20 cells.
In additional experiments, we also measured dihydroethidium and Amplex® Red oxidation. In the case of dihydroethidium oxidation, even greater differences in intracellular ROS generation were observed (Fig. 3B). To further confirm these results, we measured the H 2 O 2 -dependent oxidation of Amplex® Red in the extracellular space of these cells under both normoxic and hyperoxic conditions. Because of the extremely rapid diffusion of H 2 O 2 (41), peroxide produced intracellularly should be detected as extracellular H 2 O 2 . As shown in Fig. 3C Further Evidence of Differences in Intracellular ROS Generation in HeLa-20 and HeLa-80 Cells-If HeLa-80 cells avoid the cytotoxic and cytocidal effects of hyperoxia by virtue of diminished mitochondrial production of ROS, then this latter should also be reflected in other measures of intracellular ROS production. Two additional parameters generally support this idea. First, we measured aconitase activity under normoxic and

FIG. 2. Susceptibility of wild-type HeLa-20 and oxygen-tolerant HeLa-80 cells to oxidant-mediated cell death. HeLa-20 and
HeLa-80 cells at a density of 2 ϫ 10 4 cells per well in 48-well plates and grown 3 days at which time the cells (ϳ90% confluent) were exposed to bolus doses of H 2 O 2 (A) or t-butyl hydroperoxide (B) in HBSS at 37°C without CO 2 for 2 h. Upon return to complete culture medium, cell viability was estimated 10 h later using reduction of Alamar Blue (f ϭ HeLa-20; ‚ ϭ HeLa-80). hyperoxic conditions. Given the susceptibility of aconitase to inactivation by ROS (30), changes in the activity of this enzyme can be used as an indirect measure of intracellular ROS production. As shown in Fig. 4A, exposure of the wild-type HeLa-20 cells to hyperoxia for 24 h caused a ϳ70% decrease in total aconitase activity, whereas the aconitase activity in oxygen-tolerant HeLa-80 cells was relatively unaffected. The activity of a reference enzyme, citrate synthase, was unchanged following similar exposure of either HeLa-20 or HeLa-80 cells to hyperoxia (data not shown). We also measured mitochondrial protein carbonyl content of the cells as an indication of the extent of protein oxidation. As shown in Fig. 4B, protein carbonyls increased significantly in HeLa-20 cells after a 24-h culture in 80% O 2 . In contrast, HeLa-80 cells showed no significant change in protein carbonyls.

ROS Produced by Both HeLa-20 and HeLa-80 Cells Have a Mitochondrial
Origin-Additional experiments indicate that the majority of ROS being detected in both cell strains is almost certainly of mitochondrial origin. As shown in Table I, in both cell types, the addition of CCCP (10 M) inhibited most ROS production under both 20 and 80% O 2 , strongly supporting a mitochondrial origin of this reactive oxygen. Furthermore, a combination of two other inhibitors of mitochondrial respiration, rotenone (a complex I inhibitor) and thenoyltrifluoroacetone (a complex II inhibitor), reduced ROS production by 50 and 40% in HeLa-20 and HeLa-80 cells, respectively (Fig. 5A). Interestingly, antimycin-A, a complex III inhibitor, induced greatly increased ROS production in both cell lines (Fig. 5B) as was expected from earlier observations (42). This probably reflects the accumulation of ubisemiquinone at complex III, the likely source of most mitochondrial ROS production (12).

HeLa-80 Cells Show a More Efficient Mitochondrial Metabolism Under Both 20 and 80% O 2 -
The diminished ROS production by HeLa-80 cells could potentially reflect either an overall suppression of mitochondrial metabolism or an enhanced efficiency of mitochondrial metabolism. In the former case (and assuming an equivalent demand for glucose-derived energy in both cell types), the less efficient mitochondrial metabolism might be met by enhanced glucose consumption and production of lactic acid (reflecting increased need for glucose consumption and reduced oxidative carbohydrate metabolism by the mitochondria). As shown in Table II, it appears that both glucose consumption and lactate production are lower in the  FIG. 5. "Upstream" inhibitors of mitochondrial electron transport, but not antimycin A, decrease ROS production in both HeLa-20 and HeLa-80 cells as detected by DCFDA under normoxic culture conditions. A, ROS production was diminished in the combined presence of rotenone (complex I inhibitor) and thenoyltrifluoroacetone (TTFA) (complex II inhibitor) (*, p Ͻ 0.01 in both cases, n ϭ 5). B, ROS production was significantly increased in the presence of antimycin A (complex III inhibitor) (*, p Ͻ 0.05, n ϭ 4).
HeLa-80 cells under both normoxia and hyperoxia. However, earlier reports suggest that glutamine may be the major energy source for cultured HeLa cells (25). Indeed, we find that under identical conditions, glutamine utilization is significantly higher in HeLa-80 cells under both normoxia and hyperoxia (Table II).
COX Activity Is Significantly Higher in HeLa-80 Cells-The above results suggested tighter coupling of mitochondrial metabolism in the HeLa-80 cells. In view of the marked reduction in ROS generation caused by CCCP (Table I), we hypothesized that accelerated movement of electrons down the electron transport chain would, by depleting the electron-rich intermediates such as ubisemiquinone, decrease incidental one-electron reduction of oxygen. Indeed, as shown in Fig. 6, it appears that the activity of the terminal complex, COX, is substantially higher in HeLa-80 versus HeLa-20 cells. In contrast, activities of complexes I-III showed no significant differences (results not shown).
Preferential Inhibition of COX Activity Abrogates the Oxygen Tolerance of HeLa-80 Cells-If HeLa-80 cells are protected against the cytostatic and cytotoxic effects of hyperoxia by increased COX activity, then suppression of COX should restore susceptibility to hyperoxia. Heme deficiency induced by NMP (an inhibitor of ferrochelatase) selectively decreases the activity of COX by preferentially inhibiting heme a synthesis (35). Following 1 week of culture in the presence of 5 M NMP, COX activity was significantly decreased in both cell lines (Fig.  7A). In contrast, only slight decreases were found in catalase and cytochrome c reductase (which contain hemes b and c, respectively) (data not shown). Importantly, NMP-treated HeLa-80 cells lost resistance to hyperoxia, stopped growing, and progressively died after 6 days exposure to 80% O 2 (Fig.  7B). Furthermore, NMP-treated HeLa-80 cells produced more ROS than did control HeLa-80 cells under 80% O 2 as assessed by DCF (46.53 Ϯ 16.07 fluorescence units/min/mg of protein in control HeLa-80 cells and 90.21 Ϯ 24.52 fluorescence units/ min/mg of protein in HeLa-80 cells treated with 5 M NMP for 6 days; p Ͻ 0.05, n ϭ 4).
HeLa-80 Cells Show a Highly Significant Increase in "Respiratory Potential"-Although steady-state oxygen consumption was similar between these two cell lines, the rate of oxygen consumption was disproportionately increased in HeLa-80 cells exposed to CCCP compared with that observed in HeLa-20 cells (Fig. 8). Given that the activities of complexes I-III were not significantly different in HeLa-20 versus HeLa-80 cells, whereas COX activity was elevated ϳ2-fold in the latter, these results lend direct support to recent suggestions that COX may actually be rate-limiting in electron transport and respiration (43)(44)(45). DISCUSSION High partial pressures of oxygen can cause respiratory, cardiovascular, nervous, and gastrointestinal toxicities (2, 46 -49). The clinical importance of this phenomenon did not come to prominence until the 1950s when an epidemic of retrolental fibroplasia was finally ascribed to the therapeutic use of high O 2 tensions for neonatal pulmonary insufficiency. Despite extensive research, the mechanisms involved in oxygen toxicity remain incompletely understood. The present investigations were conducted in the hope that an improved understanding of how cells manage to be tolerant of hyperoxia might yield clues as to the basic mechanisms involved in hyperoxic cell damage and death.
Hyperoxic damage is almost certainly because of enhanced ROS formation (5). The excessive production of ROS, in turn, likely derives from the mitochondria (5,6,50). We are aware of earlier reports that the endoplasmic reticulum can also be the source for exaggerated ROS production (51,52). However, CCCP, not known to affect ROS production by endoplasmic reticulum, blocked ϳ80% of the ROS generation by both HeLa-20 and HeLa-80 cells under both normoxic and hyperoxic conditions (Table I) with the combination of thenoyltrifluoroacetone and rotenone being somewhat less effective. This strongly supports a mitochondrial origin of the ROS detected in the present investigations.
The increased ROS generation during exposure to hyperoxia can potentially overwhelm antioxidant defense mechanisms, eventuating in cell death (53). Furthermore, a recent report indicates that ROS generation is a requisite precedent to hyperoxia-induced cell death, occurring before the activation of the apoptotic pathway (54). With these assumptions in mind, there are at least two potential strategies that might permit  cells to be tolerant of hyperoxia, either improved defenses against ROS generated in hyperoxic conditions or metabolic changes that might suppress ROS generation in the first place. Of these two possibilities, preemptive suppression of ROS production would seem the most effective and economical solution. Indeed, in the oxygen-tolerant HeLa-80 cells under study, there appear to be no significant differences in antioxidant enzymes or compounds compared with wild-type HeLa-20 cells (8). An even more direct test of cellular resistance to oxidant stress, challenge with either H 2 O 2 or t-butyl hydroperoxide, similarly revealed no significant differences between HeLa-20 and HeLa-80 cells. These results are also in agreement with an earlier report that the wild-type and oxygen-tolerant HeLa cells were equally susceptible to the lethal effects of ␥-radiation (8,38).
However, the possibility remains that the tolerance of HeLa-80 cells to hyperoxia might be related to a relative resistance to apoptosis. We therefore examined cell death in response to two different agonists of apoptosis: serum withdrawal and etoposide exposure. Apoptotic cell death detected by both Hoechst staining (Hoechst 33258, Sigma) and Annexin V-FITC staining (BD Pharmingen) was similar in both cases (at 48 h, ϳ50% apoptosis under both serum starvation and exposure to 50 M etoposide in both cell lines). These results indicate that the oxygen tolerance of the HeLa-80 cells is probably not because of a differential susceptibility to apoptotic death.
These observations suggested that the oxygen tolerance of HeLa-80 cells might have arisen from selection for cells in which ROS production itself was suppressed under hyperoxic conditions. Assuming that hyperoxic cell damage does arise from exaggerated mitochondrial ROS production, the site of electron leak from the electron transport chain becomes an important question. Although this is still an area of some controversy, most experts agree that ubiquinone (coenzyme Q10) is a likely candidate. Ubiquinones can undergo both oneand two-electron reduction, with the one-electron reduction intermediate, ubisemiquinone, having the potential for oneelectron reduction of O 2 to O 2 . . Although ubiquinone is present at complexes I and III, it appears that the majority of electron leak may occur through the intermediacy of ubisemiquinone formed at complex III (12,42,55). In both wild-type and oxygen-tolerant HeLa cells, the results with inhibitors of electron transport are in accord with this possibility. Thus, whereas complex I and II inhibitors decreased ROS production in both cell lines, a complex III inhibitor, antimycin A, caused substantial increases in ROS generation. The latter is probably because of an accumulation of ubisemiquinone leading to associated increases in O 2 . production (42). The importance of this source of spontaneous O 2 . generation has been further supported by observations in the small annelid, Caenorhabditis elegans. In this ϳ900 cell worm, a mutant (clk) has been found that is associated with increased longevity and impaired synthesis of coenzyme Q 10 . Replenishment of normal coenzyme Q 10 levels leads to decreased longevity, probably associated with increased leak of electrons from the electron transport chain (56,57  ever, that this is by no means a complete metabolic inventory. Cultured cells not only use large amounts of glucose and glutamine but may also use other substrates, some of which may be as yet uncharacterized (58).
The observation that the protonophoric uncoupler CCCP would greatly decrease ROS production in both cell types suggested that anything that would accelerate the movement of electrons down the electron transport chain would, by depleting the electron-rich intermediates such as ubisemiquinone, decrease the incidental one-electron reduction of O 2 . In support of this, we find that although the activities of complexes I-III do not differ significantly, the activity of the terminal complex, COX, is substantially higher in HeLa-80 versus HeLa-20 cells. Enhanced activity of this complex might well cause the aforementioned depletion of electron-rich intermediates. The importance of this hypertrophied COX activity was further emphasized by the observation that selective inhibition of COX activity by NMP resulted in increased ROS production in HeLa-80 cells under hyperoxia and HeLa-80 cells treated in this way lose their resistance to hyperoxia. Although it was earlier thought that COX activity was in excess of other components of the electron transport chain, more recent reports indicate that COX is, in fact, the rate-limiting step for mitochondrial respiration (44,45). In support of this, we found that the rate of oxygen consumption was greatly accelerated in HeLa-80 cells exposed to CCCP compared with the increase observed in HeLa-20 cells.
Overall, our results support the general idea that hyperoxic cell damage derives from enhanced leak of reactive oxygen from electron-rich intermediates of the mitochondrial electron transport chain. The oxygen-tolerant HeLa-80 cells were found to have significantly less leak and this was associated with enhanced tolerance to hyperoxia. The suppression of hyperoxiainduced ROS production probably stems from increased activity of the terminal electron transport complex, COX, and selective inhibition of this complex causes enhanced ROS generation and abrogates the resistance of HeLa-80 cells to the cytostatic and cytocidal effects of hyperoxia.
We conclude that the oxygen tolerance of HeLa-80 cells derives from tighter coupling of electron transport because of higher COX activity, which depletes electron-rich intermediates within the chain, thereby diminishing the leak of ROS. These observations suggest that agents that lessen the leak of electrons from the mitochondrial electron transport chain may hold promise for the prevention of hyperoxic damage to lung and other tissues.