Acute Glutathione Depletion Restricts Mitochondrial ATP Export in Cerebellar Granule Neurons*

Decreases in GSH pools detected during ischemia sensitize neurons to excitotoxic damage. Thermodynamic analysis predicts that partial GSH depletion will cause an oxidative shift in the thiol redox potential. To investigate the acute bioenergetic consequences, neurons were exposed to monochlorobimane (mBCl), which depletes GSH by forming a fluorescent conjugate. Neurons transfected with redox-sensitive green fluorescent protein showed a positive shift in thiol redox potential synchronous with the formation of the conjugate. Mitochondria within neurons treated with mBCl for 1 h failed to hyperpolarize upon addition of oligomycin to inhibit their ATP synthesis. A decreased ATP turnover was confirmed by monitoring neuronal oxygen consumption in parallel with mitochondrial membrane potential (Δψm) and GSH-mBCl formation. mBCl progressively decreased cell respiration, with no effect on mitochondrial proton leak or maximal respiratory capacity, suggesting adequate glycolysis and a functional electron transport chain. This approach to “state 4” could be mimicked by the adenine nucleotide translocator inhibitor bongkrekic acid, which did not further decrease respiration when administered after mBCl. The cellular ATP/ADP ratio was decreased by mBCl, and consistent with mitochondrial ATP export failure, respiration could not respond to an increased cytoplasmic ATP demand by plasma membrane Na+ cycling; instead, mitochondria depolarized. More prolonged mBCl exposure induced mitochondrial failure, with Δψm collapse followed by cytoplasmic Ca2+ deregulation. The initial bioenergetic consequence of neuronal GSH depletion in this model is thus an inhibition of ATP export, which precedes other forms of mitochondrial dysfunction.

A balance between the formation of reactive oxygen species, as normal byproducts of mitochondrial respiration (1), and the actions of antioxidants prevents oxidative stress and is crucial to neuronal survival (2). Together with the overactivation of glutamate receptors (excitotoxicity), oxidative stress is a result of the bioenergetic crisis that characterizes ischemia and plays a central role in the pathophysiology of the consequent neuronal damage (3). Neurons are particularly sensitive to oxidative damage and can be strongly sensitized to other injurious stimuli by levels of oxidative stress that are nontoxic per se (4). Similarly, Ca 2ϩ homeostasis is lost more quickly in cerebellar granule neurons (CGNs), 2 which show higher superoxide levels prior to the application of toxic concentrations of glutamate (5).
The tripeptide glutathione is a key antioxidant that maintains protein thiols in a reduced state and scavenges H 2 O 2 in a reaction catalyzed by glutathione peroxidase (6,7). In vivo, mitochondrial glutathione is partially lost during ischemia (8), and its supplementation in the form of glutathione ethyl ester can reduce the infarct size (9). Although mitochondria within cells deprived of GSH can eventually release cytochrome c and undergo opening of the permeability transition pore (PTP) (10 -12), the initial bioenergetic consequences of GSH depletion remain unclear. Thermodynamic and kinetic factors predict that the steady-state thiol redox potential will be very sensitive to changes in total glutathione pool size (7,13). This may lead to the oxidative damage of key mitochondrial proteins. Indeed, either an inhibitor of GSH synthesis or the direct removal of reduced GSH with ethacrynic acid causes the inactivation of complexes I (14) and II/III and IV (15,16) that becomes apparent several hours from the beginning of treatment. An intriguing target for thiol oxidation is represented by the adenine nucleotide translocator (ANT), which functions as a dimer and can be progressively inhibited as the intermolecular oxidation of thiol groups increases (17). This loss of ADP/ATP exchange activity precedes the formation of intramolecular cross-linking, which seems to require stronger oxidative conditions and may be related to mitochondrial PTP opening (18,19).
In this study, we rapidly depleted neuronal glutathione by conjugation with monochlorobimane (mBCl), a glutathione S-transferase (GST) substrate that is extensively used to assay glutathione pools (20,21). The intracellular thiol redox potential was detected with a redoxsensitive green fluorescent protein variant (22), whereas population cell respiratory rates were monitored in parallel with cytoplasmic free Ca 2ϩ concentration ([Ca 2ϩ ] c ) and mitochondrial membrane potential (⌬ m ) (23), allowing glutathione depletion to be correlated with a rapid inhibition of ATP turnover that became evident within the first hour of treatment. The data suggest that a restriction of mitochondrial ATP export to the cytoplasm, consistent with ANT inhibition, is the first event occurring in this model of GSH depletion.
Preparation of CGNs-CGNs were prepared from 7-day-old Wistar rats as described previously (24), with modifications. Cells were plated onto Lab-Tek 8-well chambered coverglasses at a density of 380,000 cells/well or onto 22 ϫ 40-mm coverslips at a density of 3 ϫ 10 6 cells/ coverslip. Coverslips had previously been coated with 33 g/ml polyethyleneimine. Cultures were maintained in minimal essential medium supplemented with 10% fetal bovine serum, 30 mM glucose, 20 mM KCl, 2 mM glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin. 24 h after plating, 10 M cytosine arabinoside was added to inhibit growth of non-neuronal cells. Cell cultures were maintained at 37°C in an incubator with a humidified atmosphere of 5% CO 2 and 95% air and used for experiments at 7-9 days in vitro.
Experimental Buffers-Standard "high KCl" buffer consisted of 100 mM NaCl, 25  Functional Confocal Microscopy-Imaging of single neurons plated onto Lab-Tek 8-well chambered coverglasses was performed in a Pascal confocal system (Carl Zeiss AG, Oberkochen, Germany) using an Axiovert 100M inverted microscope with a ϫ20 air objective and argon (488 nm) and helium/neon (543 nm) lasers. Control and experimental wells were imaged in parallel using a computer-controlled motorized stage and Physiology software, which allows parallel or sequential acquisition of time courses. For simultaneous detection of [Ca 2ϩ ] c and ⌬ m (see Fig.  7), neurons were loaded with the low affinity Ca 2ϩ indicator Fluo-4FF (0.5 M; as acetoxymethyl ester; Ca 2ϩ K d ϭ 9.7 M) and TMRM ϩ (5 nM) for 30 min at 37°C. This concentration of TMRM ϩ was insufficient for aggregation in the matrix (i.e. the experiment was performed in "nonquench mode") (25), meaning that a decrease in either the plasma membrane or mitochondrial membrane potential was reflected in a reduction in whole cell fluorescence. In experiments in which [Ca 2ϩ ] c transients following KCl-induced plasma membrane depolarization were detected, neurons were loaded with 0.5 M Fluo-5F (Ca 2ϩ K d ϭ 2.3 M). Fluo-4FF and Fluo-5F were excited at 488 nm, and emission was collected between 505 and 530 nm.
TMRM ϩ was excited at 543 nm, and the emitted fluorescence was collected between 560 and 615 nm. Under both conditions, TMRM ϩ was always present in the buffer throughout the experiment (25).
Cell Respirometer-Respiration of intact neurons in situ was measured as described previously (23). A 22 ϫ 40-mm glass coverslip with attached cells was assembled in a closed RC-30 imaging chamber (Warner Instruments, Hamden, CT) and then mounted on an Olympus IX81 inverted fluorescence microscope equipped with ϫ20 and ϫ40 objectives. A miniature polarographic Clark-type oxygen electrode for perfusion systems with 1 ⁄ 16-inch fittings (Microelectrodes Inc., Bedford, NH) was used to monitor the oxygen tension in the eluted buffer. The flow rate was typically kept at 40 -60 l/min to provide sufficient oxygen depletion for detection. In experiments in which brevetoxin was employed, the buffer was also supplemented with 1 M MK-801 to ensure that the results observed did not originate from the activation of N-methyl-D-aspartate receptors. 3 nM TMRM ϩ was present in the buffer (non-quench mode conditions). A chart recorder was wired to an OM-4 oxygen meter (Microelectrodes, Inc.) to provide a continuous trace of the oxygen depletion (and hence respiration) of the cells. The microscope was equipped with a CoolSNAP HQ CCD camera (Roper Scientific, Tucson, AZ) and MetaFluor and MetaMorph imaging software (Universal Imaging Corp., Downingtown, PA). TMRM ϩ and mBCl were excited via S555/20x and D380/13x filters, respectively (Chroma Technology Corp., Rockingham, VT). The emission from both fluorophores was collected through a 73101 dual band emission filter. roGFP2 was excited via a D480/30x filter, and the emission was collected through the 73101 filter.
Transfection of roGFP2-Cultured neurons were transfected following a published protocol (26), with modifications. On day 0, following attachment to their coverslips, CGNs were transferred to culture medium without penicillin and streptomycin and incubated with 0.8 g of roGFP2 cDNA and the transfection reagent Lipofectamine 2000. After 2 h, neurons were returned to the usual culture medium and used for experiments after day 6.
Monitoring ⌬ m with Rhodamine 123-Rhodamine 123 in quench mode is the most sensitive means of detecting small changes in ⌬ m (25). Neurons were incubated in low KCl buffer and exposed to 2.2 M rhodamine 123 for 15 min at 37°C (to attain quench mode conditions). Cells were then rinsed with fresh buffer and imaged with the Olympus IX81 microscope. Rhodamine 123 was excited via a D480/20x filter, and the emitted fluorescence was collected through a 51000m filter (Chroma Technology Corp., Rockingham, VT). Given the relatively low membrane permeability, rhodamine 123 leaks out of the cells very slowly and therefore was not present in the media during the experiments.
Determination of ATP/ADP Ratios-Total cellular ATP/ADP ratios were determined using a luciferase chemiluminescence assay (Calbiochem). Cells were incubated in low KCl buffer under the relevant conditions and lysed, and extracts were assayed for ATP according to the manufacturer's instructions. Pyruvate kinase (2 units/assay) was added to the buffer supplied with the kit. The increase in chemiluminescence was recorded in a TD-20/20 luminometer (Turner BioSystems, Sunnyvale, CA). Once the ATP signal was detected, 0.5 mM phosphoenolpyruvate was added, and the further increase in chemiluminescence due to the conversion of ADP to ATP was determined.
Immunocytochemistry-Neurons plated onto Lab-Tek 4-well chambered slides were washed with 50 mM Tris and 150 mM NaCl (pH 7.5), fixed with 4% paraformaldehyde, and permeabilized in 0.1% Triton X-100. A purified mouse anti-cytochrome c monoclonal antibody (Pharmingen) that recognizes the native form of cytochrome c (1:100 dilution) was added to the coverslips (overnight, 4°C). The cells were then incubated with a fluorescein isothiocyanate-conjugated antimouse secondary antibody. Nuclei were stained with TOTO-3 before mounting the slides on ProLong Gold medium (both reagents from Molecular Probes). The samples were imaged with a Nikon PCM 2000 confocal system equipped with argon (488 nm) and helium/neon (633 nm) lasers.
Statistical Analysis-All data analyzed were collected from at least three independent experiments and are expressed as means Ϯ S.E. Data from two populations were compared with the unpaired two-tailed Student's t test. One-way analysis of variance followed by Tukey's test was used in the analysis of three or more experimental groups. p values Ͻ0.05 were considered significant.

Conjugation of Endogenous GSH with mBCl Oxidizes the Thiol Redox
Potential-mBCl becomes fluorescent when conjugated to GSH by intracellular GST. The reaction is highly selective for glutathione over other intracellular thiols (27), but is not a good indicator of subcellular distribution because GSTs are present in both the cytoplasm and the mitochondrion (28), and the probe eventually accumulates in the nucleus ( Fig. 1) (29). Another substrate for GST, ethacrynic acid, has been frequently used to deplete cells of GSH (16,30). However, ethacrynic acid also directly inhibits mitochondrial complex II (31) and, in our hands, caused a partial drop in ⌬ m when added to CGNs. An additional advantage of mBCl is that the reaction course can be followed by the development of fluorescence, as mBCl depletes GSH by forming the adduct GSH-mBCl (Fig. 1A), with the reaction usually reaching completion within 60 min.
A major advance in the ability to monitor thiol redox potentials has come as a result of the development of roGFP analogs by substituting surface-exposed residues of the chromophore with cysteines (22). Notably, an analog of roGFP2 (redox-sensitive yellow fluorescent protein) expressed in the cytoplasm of yeast cells appears to directly equilibrate with the cytoplasmic glutathione couple (32). Although the redox state of roGFPs can be monitored ratiometrically by excitation at 400/480 nm (22), the presence of the mBCl adduct interferes with the emission from 400 nm excitation, and so only 480 nm excitation was employed. When excited at 480 nm, oxidation of roGFP2 will result in a lower fluorescence emission, whereas its reduction will cause fluorescence enhancement.
The validity and sensitivity of the construct in CGNs were established by applying 500 M H 2 O 2 to elicit exogenous oxidative stress. The treatment resulted in a rapid decrease in roGFP2 emission (Fig. 1B) that was reversed by the thiol-reducing agent dithiothreitol. The endogenous oxidative stress induced by 100 M mBCl caused a slower but equally extensive oxidation of roGFP2 that was reversed by dithiothreitol, confirming the ability of GSH depletion to cause an oxidative intracellular thiol redox potential shift.
Depletion of GSH Affects Mitochondrial ATP Synthesis-Because the active synthesis of ATP by the mitochondrial ATP synthase slightly decreases ⌬ m by utilizing the proton current, it follows that addition of the ATP synthase inhibitor oligomycin will acutely hyperpolarize the mitochondria. Conversely, mitochondria that have inhibited respiratory chains or that have leaky inner membranes can reverse the ATP synthase to maintain a suboptimal ⌬ m by utilizing glycolytic ATP. In this case, oligomycin will depolarize the inner membrane. This "oligomycin null point" assay (33,34) can therefore be used to investigate the bioenergetic status of the mitochondrial population within a single cell body. Fig. 2A (panels i-iii) shows responses of representative neurons whose ⌬ m hyperpolarized, was unchanged, or depolarized in response to the inhibitor, respectively. Because rhodamine 123 is in "quench mode," hyperpolarization will cause further sequestration and quenching of the dye inside the matrix and decrease the whole cell fluorescence. The final addition of carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) to totally release the dye into the cytoplasm confirms that the experiment is performed under quench mode conditions.
As expected, oligomycin decreased the fluorescence in a majority of control neurons (53%) ( Fig. 2A, panel i). By contrast, the large majority of mBCl-treated cells (83%) did not undergo a detectable change in fluorescence following oligomycin addition, whereas 15% of the cells showed mitochondrial depolarization upon addition of oligomycin, indicating that, in some mitochondria, the membrane potential was sustained by the ATP synthase running in reverse. The unchanged rhodamine 123 fluorescence observed in most mBCltreated cells upon oligomycin addition suggests that the mitochondria are close to respiratory state 4 (35), i.e. with little ATP turnover. This could be due either to a decreased cellular ATP demand or to an inhibition of the generation or export of ATP from the mitochondrion to the cytoplasm. The former would result in an increased total cellular ATP/ADP ratio, whereas the latter would decrease the ratio. A 75-min exposure to mBCl caused a significant decrease in the ATP/ADP ratio compared with control cells (2.76 Ϯ 0.29 versus 4.83 Ϯ 0.36) (Fig. 2B). This decrease was similar to that induced by oligomycin (2.54 Ϯ 0.18). ATP synthase reversal induced by the inhibition of complex I with rotenone and by the protonophore FCCP caused a greater reduction in the ATP/ADP ratio (to 1.9 Ϯ 0.16 and 1.2 Ϯ 0.23, respectively). The decreased ratio is thus consistent with inhibited generation or export of ATP from the matrix.
Mitochondria in CGNs Exposed to mBCl Approach State 4-To confirm and quantify the above observations, the O 2 consumption of intact neurons attached to their coverslips was determined (Fig. 3) with a cell respirometer assembly as described previously (23). Cells were either perfused with standard high KCl buffer ϩ 0.1% Me 2 SO (control) until mitochondrial inhibitors were added or switched to buffer additionally containing 100 M mBCl after basal respiration was assessed. Rates of respiration were calculated from the perfusion rate times the difference between the downstream oxygen tension during chamber perfusion and that recorded when the chamber was bypassed with a shunt.
The basal respiration of control neurons was stable until oligomycincontaining medium was perfused. The inhibitor decreased respiration to 44 Ϯ 6% of the basal level, the residual value reflecting the inherent proton leak across the inner mitochondrial membrane (Fig. 3A). Thus, almost half of the basal respiration was used to compensate for the endogenous mitochondrial proton leak, whereas the remainder reflected basal ATP turnover. The effect of 100 M mBCl was to initiate within 15 min a slowing of the basal respiration that progressed as GSH was conjugated to mBCl over a 60-min period (Fig. 3B). At that point, ATP turnover in the cells (calculated as the oligomycin-sensitive component of respiration) had decreased to Ͻ50% of that in control cells. Notably, mBCl caused no significant change in state 4 respiration (following oligomycin addition); thus, there was no uncoupling or permeability transition induction following 60 min of mBCl exposure, and the decrease in total respiration could be ascribed to an inhibition of mitochondrial ATP synthesis.
The respirometer also allows maximal respiratory capacity to be quantified. Following the addition of the protonophore FCCP, mBClexposed neurons increased their respiration to the same extent as control cells (203 Ϯ 42% over the state 4 rate in control cells versus 180 Ϯ 30% in mBCl-treated cells) (Fig. 3D), demonstrating that this acute GSH depletion did not detectably restrict substrate delivery or electron transfer activity. This contrasts with long-term partial glutathione depletion that manifests as an inhibition of complex I activity in PC12 cells (36).
The respirometer was mounted on an epifluorescence microscope, which allows imaging of multiple fluorescence indicators at a single cell level (see "Materials and Methods"). Thus, when 100 M mBCl was added, we could confirm that the fluorescence of the intracellular adduct GSH-mBCl reached a plateau within 1 h (see Fig. 1A). Also the fluorescence from cells equilibrated with 3 nM TMRM ϩ indicated that there was no detectable depolarization of either plasma or mitochondrial membrane potentials during this period (Fig. 4).
Because respiratory capacity is not rate-limiting, the approach to state 4 following mBCl addition could be due to a decreased capacity for mitochondrial ATP synthesis (ATP synthase inhibition) or to inhibited export to the cytoplasm (ANT inhibition), whereas a decreased cellular ATP demand (e.g. inhibited ion pump activity) would be inconsistent with the decreased ATP/ADP ratio observed (Fig. 2B). Because of reports showing ANT sensitivity to thiol oxidation (18,19,37), a partial inhibition of this protein appears a plausible mechanism for the observed effects. In support of this view, incubation with the cell-permeant ANT inhibitor bongkrekic acid (10 M) induced a decline in . Mitochondria in CGNs exposed to mBCl approach state 4. A, representative respirometer trace for control CGNs. 3 ϫ 10 6 cells were equilibrated in standard high KCl buffer for 30 min. Where indicated by the arrows, buffers containing 2 g/ml oligomycin (O) to quantify ATP turnover and proton leak or 500 nM FCCP (F) to determine maximal respiratory capacity were perfused in sequence. B, parallel experiment with 100 M mBCl added where indicated. C, parallel experiment with 10 M bongkrekic acid (Bk) added where indicated. D, quantitation of oligomycin-induced respiratory inhibition and FCCP-induced respiratory stimulation relative to the rate in the presence of oligomycin (State 4) for control and mBCl-and bongkrekic acid-exposed cells. Also included are means of experiments in which mBCl (60 min before oligomycin addition) and bongkrekic acid (30 min before oligomycin addition) were both present. Data are from six independent experiments per treatment. Oligomycin-insensitive respiration was not significantly changed by mBCl or bongkrekic acid alone or in combination. *, p Ͻ 0.05 versus other groups. basal respiration; but, as with 60 min mBCl exposure, the approach to state 4 was incomplete, and an additional effect of oligomycin was apparent (Fig. 3C). If bongkrekic acid and mBCl were affecting cell respiration via alternative mechanisms, their combined effects should be additive and produce a greater decrease in respiration. To investigate this, we perfused cells with mBCl for 30 min, switched to buffer containing mBCl ϩ bongkrekic acid for an additional 30 min, and then assessed the oligomycin-insensitive respiration. Interestingly, oxygen consumption was reduced to a similar extent by bongkrekic acid alone or in combination with mBCl. (Respiration was 55.4 Ϯ 7.9% above the oligomycin rate in cells exposed to bongkrekic acid versus 53.1 Ϯ 13.8% in cells treated with both mBCl and bongkrekic acid.) Bioenergetic Failure of mBCl-treated Neurons following Increased Cytoplasmic ATP Demand-The plasma membrane Na ϩ /K ϩ -ATPase is potentially the major ATP-utilizing reaction in the neuron (38). However, its activity is usually restrained by the limited Na ϩ cycling due to the low basal Na ϩ conductance of the plasma membrane. To activate Na ϩ cycling and to increase the cytoplasmic ATP demand, the voltageindependent Na ϩ channel activator brevetoxin was used. Because of its ability to promote extensive release of vesicular glutamate and thus activate N-methyl-D-aspartate receptors (39), brevetoxin was perfused in the presence of 1 M MK-801. In control neurons, brevetoxin robustly enhanced respiration, utilizing ϳ50% of the "reserve capacity" of the mitochondria and increasing respiration by 32 Ϯ 5% above basal levels (Fig. 4A). After washout of the neurotoxin, basal respiration was restored, and the neurons responded to FCCP, confirming that the mitochondria possessed sufficient ATP-generating capacity to meet this enhanced demand. In contrast, when brevetoxin was added to neurons that had been incubated with mBCl for 60 min, little respiratory enhancement was seen (11 Ϯ 2% above that immediately prior to toxin addition) (Fig. 4B). The loss of TMRM ϩ signal immediately upon addition of the toxin indicates a loss of plasma or mitochondrial membrane potential, and the subsequent minimal response to FCCP, together with examination of cell morphology (data not shown), indicates extensive cell damage consistent with a cellular ATP deficit.
To further investigate the ability of mBCl-treated neurons to handle higher cellular ATP demand caused by increased Na ϩ cycling, the neurons were exposed to the excitotoxin kainate (Fig. 5), which induces massive Na ϩ entry into the cell through non-inactivating ␣-amino-3hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (34). As in the above experiments, cells were loaded with 3 nM TMRM ϩ to monitor ⌬ m in parallel with cell respiration. In control neurons, 50 M kainate induced a substantial increase in cell respiration (ϩ76 Ϯ 7.4%) that returned to basal values once the excitotoxin was washed out (Fig.  5A). Partly because of plasma membrane depolarization, TMRM ϩ fluorescence decreased significantly upon perfusion with kainate, but recovered almost completely after the washout. In contrast to control cells, neurons perfused with 100 M mBCl for 60 min increased respiration only slightly when exposed to kainate (ϩ30.1 Ϯ 10.5%; p Ͻ 0.05 versus kainate effect in control cells) (Fig. 5B). Moreover, mitochondria in mBCl-treated cells profoundly depolarized in the presence of kainate, failed to repolarize after the washout, and showed virtually no response to subsequent FCCP addition. Thus, as with brevetoxin (Fig. 4B), the experiments with kainate suggest that mitochondria within mBCltreated cells cannot increase ATP production sufficiently to cope with the increased cytoplasmic demand of Na ϩ cycling at the plasma membrane, leading to membrane depolarization and cell death.
Mitochondrial Ca 2ϩ Uptake Is Preserved in mBCl-treated Neurons-When plasma membrane Ca 2ϩ channels are activated by plasma membrane depolarization in elevated KCl, Ca 2ϩ entry elevates [Ca 2ϩ ] c sufficiently for the in situ mitochondria to accumulate the cation (40). This accumulation can be detected following the addition of protonophore as a transient "spike" in [Ca 2ϩ ] c as Ca 2ϩ is released into the cytoplasm and pumped out across the plasma membrane. Mitochondrial Ca 2ϩ accumulation is not dependent upon ATP and thus should be unimpaired if the only mBCl-induced modification to the mitochondrion is an inhibited ATP export. Fig. 6 shows that mitochondria accumulated Ca 2ϩ under these conditions and released it into the cytoplasm when the collapse of ⌬ m was induced by 250 nM FCCP. A similar pattern was observed after treatments with mBCl for 75 min, indicating that these 40  neurons were still able to accumulate Ca 2ϩ within the mitochondrial matrix. FCCP induced a 1.76 Ϯ 0.29-fold increase in Fluo-5F fluorescence in control cells compared with a 1.65 Ϯ 0.25-fold increase in mBCl-treated cultures. Ca 2ϩ accumulation is largely dependent upon the parallel uptake of phosphate (41). Although the mitochondrial phosphate transporter is sensitive to thiol-oxidizing agents, the more subtle effects elicited here by GSH depletion did not appear to affect the transporter sufficiently to influence Ca 2ϩ transport. However, in mBCltreated cells, the [Ca 2ϩ ] c plateau that followed the application of KCl was slightly elevated. This might reflect a reduced extrusion capacity of the plasma membrane Ca 2ϩ -ATPase due either to direct oxidative damage or to the fact that less ATP is available in mBCl-treated neurons.
When 250 nM FCCP was administered in the absence of 50 mM KCl, neither control nor mBCl-treated cells showed [Ca 2ϩ ] c changes (data not shown), consistent with a lack of mitochondrial Ca 2ϩ accumulation in polarized cells (40).
MBCl Treatment Leads to ⌬ m Collapse, Cytochrome c Release, and Cell Death Only at Later Stages-Previous studies conducted on cell lines have shown that the depletion of GSH may lead ultimately to the release of cytochrome c from mitochondria (10). The occurrence of this event within 1 h of exposure to mBCl would result in partial mitochondrial depolarization limited by reversal of the ATP synthase (33) and a loss of respiratory capacity, neither of which was observed. However, the majority of neurons exposed chronically to mBCl for 2.5-3 h ultimately lost their TMRM ϩ signal (Fig. 7). Notably, in 58 Ϯ 10% of the neurons, the collapse of potential substantially preceded an uncontrolled rise in [Ca 2ϩ ] c (Fig. 7B), indicating that mitochondrial depolarization could be causative in the subsequent Ca 2ϩ deregulation. This is in contrast to the events of glutamate excitotoxicity (25), where mitochondrial depolarization and Ca 2ϩ deregulation are synchronous.
The relatively sudden and dramatic mitochondrial depolarization seen after ϳ3 h could be the result of cytochrome c loss from the mitochondrion. This was confirmed by immunocytochemical localization with an antibody for native cytochrome c. CGNs either untreated or exposed to mBCl for 1 h displayed a punctate pattern of cytochrome c signal consistent with mitochondrial localization. By contrast, the staining for cytochrome c was almost entirely lost in neurons exposed to mBCl for 3.5 h. The apoptosis inducer staurosporine (1 M, 5 h) was used as a positive control to assess cytochrome c release (Fig. 7C).
As might be predicted from this delayed cell damage, extensive cell loss was seen in cultures exposed to mBCl for 24 h (Fig. 8). Notably, cell death was prevented if the cell-permeant glutathione ethyl ester was present together with mBCl during and after mBCl treatment, suggesting that the former can effectively replenish intracellular GSH pools.

DISCUSSION
These experiments allowed us to detect and quantify the earliest bioenergetic consequences of acute neuronal glutathione depletion by combining several novel techniques such as the cell respirometer (Figs. [3][4][5]; membrane potential determinations, including the oligomycin null point test for mitochondrial ATP synthesis (Fig. 2); and fluorescent monitoring of the time course of GSH depletion and verification of oxidative shifts in thiol redox potential (Fig. 1). The initial events of mBCl-mediated glutathione depletion appeared to be focused at the mitochondrion, even though the GSH-mBCl conjugate concentrated in the nucleus (Fig. 1). In culture, removal of GSH by conjugation to a variety of GST substrates affects the mitochondrial pool first (30). Although the reasons for this differential depletion are unknown, one explanation might be that the matrix is slower to replenish its GSH pools, as has been reported for COS cells exposed to the GST substrate 4-hydroxynonenal (28). Mitochondria do not synthesize GSH, but replenish their pool from the cytoplasm (42). The two pools are interconnected; but, in the brain of adult rats, the GSH content of mitochondria is ϳ4-fold lower than that detected in the total tissue (8). Even greater differences have been observed between mitochondrial and cytoplasmic fractions in primary cultures (30,43). Mitochondrial GSH pools are preferentially reduced when GST substrates are used, as seen with ethacrynic acid in CGNs (43) and cortical astrocytes (30) and with 4-hydroxynonenal in COS cells (28). Mitochondrial levels of GSH are reduced following focal ischemia (8); and supplementation with glutathione ethyl ester, which increases GSH levels in mitochondria, reduces the infarct zone caused by a similar treatment (9). Kainate substantially enhanced cell respiration and caused a drop in ⌬ m that was reversed upon washout. B, after assessing basal respiration, CGNs were perfused with 100 M mBCl for 60 min before kainate was applied to the bath. Kainate increased respiration only slightly, but caused an extensive loss of TMRM ϩ fluorescence that was not reversed upon washout (note the lack of response to FCCP). The phenomenon is consistent with a collapse of both plasma membrane and mitochondrial membrane potentials. The enzyme that maintains glutathione reduction (glutathione reductase) operates away from thermodynamic equilibrium because redox potentials of the matrix and cytoplasmic NADP ϩ /NADPH pools at approximately Ϫ350 mV (44) are much lower than the glutathione redox potential in the cell at about Ϫ240 mV (7,45,46). This disequilibrium implies that the glutathione pool rapidly cycles in the cell and that glutathione reductase exerts considerable control over the thiol redox potential. There is evidence that glutathione reductase is limited by the low GSSG concentration in the cell; thus, oxidation of the GSH pool by t-butyl hydroperoxide leads to a rapid and selective oxidation of NADPH both in cultured cells (47) and in vivo (48). Although it is intuitive that reductions in the GSH pool will exacerbate the rate limitation due to glutathione reductase, the distinctive thermodynamics of the glutathione couple, where two GSH molecules condense upon oxidation to form GSSG, mean that the midpoint potential becomes progressively more oxidized as the pool decreases. To maintain the same thiol redox potential when the pool size is halved would require a 4-fold decrease in GSSG concentration (35), which is clearly at odds with the kinetic constraints discussed above.
Although mBCl has been extensively used to assay glutathione in cells (29,49,50), the simultaneous monitoring of the bioenergetic consequences during the resulting glutathione depletion is novel. A major advantage over "blind" techniques such as incubation with ethacrynic acid or L-buthionine (SR)-sulfoximine (51) is that the extent and time course of depletion can be monitored continuously by the formation of the fluorescent adduct. In addition, the fluorescent signal from transfected roGFP2 (22), a member of a class of thiol-modified fluorescent proteins that appear to equilibrate directly with the cytoplasmic thiol redox potential imposed by the glutathione couple (32), allows the resulting oxidation to be monitored. The oxidative shift (Fig. 1) confirms the predictions of the thermodynamic analysis discussed above.
Although bioenergetic limitations were apparent within 15 min of mBCl addition (Fig. 3), neither induction of the permeability transition nor inhibition of complex I, which occurs following a prolonged inhibition of GSH synthesis (14), could be detected during the 1-h depletion time course. In fact, the classical criterion of bioenergetically intact mitochondria (a maintained membrane potential, low inner membrane proton conductance, high respiratory capacity, and hence a high "uncoupler respiratory control") remained unaffected until much later.  was also present in the appropriate culture medium. 24 h later, cells were exposed to 1 M propidium iodide, and the visual fields that had been selected the day before were imaged a second time.
Instead, the initial event appeared to be a progressive decrease in mitochondrial ATP turnover. The candidate targets for the decreased ATP export from the mitochondrion include the ATP synthase and ANT. Any inhibition was incomplete at 60 min because oligomycin still inhibited respiration (Fig.  3). However, it is well established that ANT is vulnerable to thiol oxidation, and ATP/ADP translocation is progressively inhibited by reaction of specific cysteines with thiol reagents (17,37,52). This inhibition is distinct from the change in ANT conformation that opens a channel conductance to form the mitochondrial PTP (18,53). PTP activation clearly did not occur during the initial 60 min of GSH depletion because there was no detectable loss of respiratory control (Fig. 3). We found no evidence for an early low conductance pore (54,55) in our experiments because this event would result in an increase in oligomycin-insensitive respiration. However, prolonged exposure to mBCl eventually triggered mitochondrial depolarization and cytochrome c release, events consistent with PTP opening, although the collapse of ⌬ m was not prevented by 10 M bongkrekic acid (data not shown) added as a cell-permeant PTP inhibitor (56). It is possible that the oxidation of mitochondrial proteins at that stage is so advanced to cause misfolding and thus render the PTP insensitive to classical inhibitors (57). PTP opening can also occur without ANT participation (58), although it is not known whether oxidative stress can trigger a similar mechanism.
Levels of oxidative stress that are nontoxic per se strongly sensitize primary cultures to other injurious stimuli: in cortical neurons subjected to sublethal mechanical stretch (4) and in astrocytes depleted of GSH, the respective addition of N-methyl-D-aspartate and a peroxynitrite generator produces far more cell death than each of the stimuli alone (30). This synergistic effect may be particularly relevant to ischemia, where glutamate excitotoxicity develops in parallel with the loss of mitochondrial GSH (8). We have shown here that Na ϩ influx via voltagedependent channels or kainate receptors imposes a higher ATP demand that mitochondria within cells acutely depleted of GSH are unable to meet. A similar scenario is likely to occur during ischemia, when Na ϩ and Ca 2ϩ flood neurons via activated glutamate receptors. Indeed, restrictions in mitochondrial ATP generation similar to those induced by mBCl sensitize neurons to excitotoxic damage (59).
At later stages (2.5-3 h), a collapse of ⌬ m preceding Ca 2ϩ deregulation was observed. This suggests that cell death triggered by mBClinduced GSH depletion originates from mitochondrial dysfunction. It is notable that the temporal relationship between mitochondrial depolarization and cytoplasmic Ca 2ϩ deregulation (Fig. 7) differs from that observed in the delayed Ca 2ϩ deregulation induced by chronic N-methyl-D-aspartate receptor activation in the same cells, where it is not possible to distinguish the precedence between mitochondrial depolarization and loss of Ca 2ϩ homeostasis (5,25). In this study, mitochondrial depolarization clearly preceded delayed Ca 2ϩ deregulation; and because mitochondria released cytochrome c at about this time (Fig. 7), this would be consistent with a redox regulation of mitochondrial cytochrome c release similar to that reported in cultured sympathetic neurons following nerve growth factor withdrawal (60). These data are particularly interesting if we consider that cytoplasmic roGFP was quickly oxidized during mBCl treatment (Fig. 1). Thus, the oxidation of extramitochondrial proteins is also likely to occur in this time frame. The slightly higher [Ca 2ϩ ] c plateau observed in 50 mM KCl may indeed suggest a partial inactivation of plasma membrane Ca 2ϩ extrusion pumps. However, any potential impairment of cytoplasmic proteins did not trigger delayed Ca 2ϩ deregulation until the collapse of ⌬ m was observed. Future experiments on sister cultures expressing either cyto-plasmic or mitochondrially targeted roGFP would help to shed light on the dynamics of thiol oxidation at a subcellular level.
In conclusion, our work provides the first analysis of mitochondrial function in intact neurons acutely depleted of GSH. We have shown that the restriction of ATP turnover is the first sign of mitochondrial dysfunction, preceding depolarization and cytochrome c release. In this experimental model, GSH depletion induces an early energy shortage that may transform the use of ATP by rapidly working plasma membrane ion pumps into a toxic challenge.