In situ respiration and bioenergetic status of mitochondria in primary cerebellar granule neuronal cultures exposed continuously to glutamate.

Mitochondria play a central role in neuronal death during pathological exposure to glutamate (excitotoxicity). To investigate the detailed bioenergetics of the in situ mitochondria, a method is described to monitor continuously the respiration of primary cerebellar granule neuron cultures while simultaneously imaging cytoplasmic Ca(2+) and mitochondrial membrane potential. Coverslip-attached cells were perfused in an imaging chamber with upstream and downstream flow-through oxygen electrodes. The bioenergetic consequences of chronic glutamate exposure were investigated, including ATP supply and demand, proton leak, and mitochondrial respiratory capacity during chronic glutamate exposure. In 25 mM K(+) medium supplemented with 10% dialyzed serum, cells utilized 54% of their respiratory capacity in the absence of receptor activation (37% for ATP generation, 12% to drive the mitochondrial proton leak, and the residual 5% was nonmitochondrial). glutamate initially increased mitochondrial respiration from 51 to 68% of capacity, followed by a slow decline. It was estimated that 85% of this increased respiration was because of increased ATP demand, whereas 15% was attributable to a transient mitochondrial proton leak. N-Methyl-D-aspartate receptor activation was only responsible for 62% of the increased respiration. When adjusted for cell death over 3 h of glutamate exposure, respiration of the viable cells remained near basal and protonophore stimulated respiration to the same extent as control cells. Pyruvate-supplemented media protected cells from glutamate excitotoxicity, although this was associated with mitochondrial dysfunction. We conclude that excitotoxicity under these conditions is not because of an ATP deficit or uncoupling. Furthermore, mitochondria maintain the same respiratory capacity as in control cells.

Although glutamate is the dominant excitatory neurotransmitter in the central nervous system, excessive glutamate exposure can cause neuronal cell death by a process known as glutamate excitotoxicity (1,2). This phenomenon amplifies brain damage following stroke, traumatic brain injury, and epilepsy where massive glutamate release from neurons in the affected region causes pathological activation of N-methyl-D-aspartate (NMDA) 1 -selective glutamate receptors and contributes to neuronal cell death in the surrounding tissue (3). In primary neuronal culture models, necrosis is not instantaneous but rather occurs after a latent period followed by an uncontrolled rise in cytoplasmic free Ca 2ϩ , ([Ca 2ϩ ] c ), known as delayed Ca 2ϩ deregulation (DCD) (4,5).
A central role for mitochondria in cerebellar granule neuron (CGN) excitotoxicity is firmly established, based on fluorescent monitoring of changes in mitochondrial membrane potential (⌬ m ) (6 -8), mitochondrial Ca 2ϩ accumulation (9, 10) and morphology (11), and on the protection afforded by mitochondrial depolarization prior to glutamate under conditions where glycolysis is adequate (12)(13)(14). However, there is very limited information on in situ mitochondrial respiration (15)(16)(17)(18), even though the classic oxygen electrode chamber has proven to be the single most important apparatus for monitoring the function of isolated mitochondria. This is primarily because of the requirement for large numbers of cells in suspension when using conventional oxygen electrodes.
The approach taken in the present study was to develop a "cell respirometer" that retains the cells on their coverslips, reduces the dead volume as far as possible, and continuously superfuses media over the cells, detecting the difference in the upstream and downstream oxygen concentrations with sensitive flow-through micro oxygen electrodes. Because the assembly is mounted on a digital imaging microscope, a representative sample of the entire cell field can be continuously imaged to monitor cell morphology and survival together with fluorescent signals from one or more indicators, such as FURA-PE3 for [Ca 2ϩ ] c and TMRM ϩ for mitochondrial and plasma membrane potentials. In experiments to monitor mitochondrial function during potentially excitotoxic glutamate exposure, it is important to distinguish between a change in population respiration because of stochastic cell death and one that reflects a uniform decline in single cell respiration. By imaging a representative field, cell viability can be continuously assessed and used to correct the response.
The present study tested hypotheses that chronic glutamate exposure increases ATP demand, that oxidative phosphorylation becomes rate-limiting, that mitochondrial uncoupling is an early event (7,19), and that mitochondrial respiratory capacity is impaired (15).

EXPERIMENTAL PROCEDURES
Reagents-TMRM was from Molecular Probes (Eugene, OR). Fura-PE3-AM was from Tef Labs (Austin, TX). 10 kDa of dialyzed fetal bovine serum and all other reagents were from Sigma. Cerebellar Granule Neuron Preparation and Indicator Loading-CGNs were prepared from 5-to 7-day-old rat Wistar pups essentially as described previously (20). 3 ϫ 10 6 cells were allowed to settle for 30 -40 min onto 22 ϫ 40-mm coverslips previously coated with 33 g/ml polyethyleneimine and then maintained at 37°C in a 95/5% air/CO 2 incubator. Cells were used after 7-10 days in culture. Prior to assembling the perfusion chamber, cells in culture media were incubated at 37°C with 3 nM TMRM ϩ plus 1 M tetraphenylboron for 2-2.5 h and with 2 M FURA-PE3-AM for 50 -80 min. Under these conditions, TMRM ϩ fluorescence was in nonquench mode (21), meaning that a decrease in either plasma membrane potential (⌬ p ) or ⌬ m is reflected in a decrease in whole-cell fluorescence. Fura-PE3-AM was used to monitor [Ca 2ϩ ] c because it was retained sufficiently by the granule neurons to provide reliable 340/380 nm signal over 2-5 h, whereas FURA2 AM was essentially lost after 3 h.
Respiration Measurements Using the Cell Respirometer-The respirometer is shown schematically in Fig. 1. A closed RC-30 imaging chamber (Warner Instruments, Hamden, CT) was assembled submerged in standard buffer without serum (to avoid cell drying or bubble trapping) with a lower 22 ϫ 40-mm coverslip with attached cells, a 250-m gasket (yielding a chamber volume of 130 l), and a blank upper coverslip. The chamber was mounted on an Olympus IX 70 inverted fluorescence microscope equipped with a 40ϫ oil-immersion objective. Miniature polarographic Clark-type oxygen electrodes constructed for perfusion systems with 1/16 inch fittings (Microelectrodes, Inc., Bedford, NH) monitored upstream and eluant oxygen tension. A peristaltic pump placed distal to the downstream oxygen electrode was used to set the system flow rate, which was typically 30 -60 l/min to provide adequate oxygen depletion for detection by the downstream electrode. The standard perfusion medium used for respiration measurements consisted of (in mM) 105 NaCl, 25 KCl, 20 TES, 15 glucose, 1.3 MgCl 2 , 1.3 CaCl 2 , 1.2 Na 2 SO 4 , 0.4 KH 2 PO 4 , 0.2 NaHCO 3 , 10% dialyzed heat-inactivated fetal bovine serum, 3 nM TMRM ϩ , and 1 M tetraphenylboron, pH 7.3, at 37°C. Variations of the standard buffer are specified under "Results." The upstream electrode controlled for changes in oxygen tension in the inflow and was used to check the downstream electrode base line at the beginning and end of the perfusion, via a tubing shunt that allowed buffer flow to by-pass the chamber (Fig. 1).
A dual-pen chart recorder was wired to each OM-4 oxygen meter (which delivered the polarizing voltage to the electrodes (Microelectrodes, Inc., Bedford, NH)) to provide a continuous trace of respiration by the cells (see Fig. 2A). Electrodes were calibrated using air-saturated buffer and buffer equilibrated with 50% air, 50% nitrogen, assuming that air-saturated buffer contains 351 nmol of O/ml at 37°C.
The microscope was equipped with an Olympus Ultraview digital imaging system. TMRM ϩ was excited via a Chroma HQ548/20x filter, and dual wavelength FURA-PE3 excitation was obtained with Chroma D340/10x and D380/13x filters. The emission from both fluorophores was collected through a 73100bs dichroic and 73101 dual band emission filter.
Once assembled into the chamber, cells were equilibrated in the standard medium for 30 min. Control data were collected for 30 min, with TMRM ϩ , FURA-PE3, and transmission images taken every 10 min. The perfusing medium could be changed to allow responses to various additions to be followed over the next 1-5 h, as detailed in the figures.
The choice of tubing was important. TMRM ϩ adsorbed to Tygon tubing to such an extent that this tubing could not be used upstream of the chamber, otherwise the TMRM ϩ concentration exposed to the cells was difficult to assess and potentially variable; silastic tubing was used upstream of the chamber as it did not adsorb TMRM ϩ . However, because silastic tubing is highly permeable to oxygen, it was not appropriate to connect the chamber to the downstream electrode, otherwise oxygen in the air rapidly equilibrated with the oxygen-depleted buffer exiting the chamber. Tygon tubing, which is less permeable to oxygen, was used for this connection to minimize oxygen diffusion into the oxygen-depleted buffer, because probe binding was immaterial downstream of the chamber. Additionally, the tubing length between chamber and downstream electrode was kept as short as possible.
Most important, the oxygen electrodes exhibited significant re-sponses to temperature variations. To avoid this, a custom-made acrylic incubator was constructed to surround and maintain the entire microscope stage, objective and incubation reservoirs at 37°C, eliminating buffer cooling and temperature artifacts. Statistics-Analyses were performed using GraphPad Prism version 4.0 for Windows (GraphPad Software, San Diego, CA). Tests used were two-way ANOVA, one-way ANOVA with Tukey's post test, and unpaired Student's t test. The level of statistical significance was taken as p Ͻ 0.05.

RESULTS
Assessing Neuronal Bioenergetics Using Flow-through Oxygen Electrodes- Fig. 2 illustrates some of the considerable FIG. 1. Schematic illustrating essential features of the apparatus used to measure respiration of CGNs cultured on coverslips. 3 ϫ 10 6 CGNs isolated from the cerebella of 7-day-old rats were plated onto 22 ϫ 40-mm polyethyleneimine-coated coverslips. After 7-10 days of incubation, each coverslip with attached cells was preloaded with 2 M FURA-PE3 and 3 nM TMRM ϩ in standard perfusion medium and then mounted into the base of a submerged Warner RC30 closed-bath chamber. The chamber was assembled submerged in standard buffer using a 250-m gasket, resulting in an ϳ130-l chamber volume. The inlet and outlet ports of the chamber were connected to upstream and downstream flow-through polarographic oxygen electrodes (Microelectrodes Inc.) resting on the microscope stage. Each oxygen meter delivered a polarizing voltage to each electrode and had outputs that were connected to a dual-pen chart recorder for monitoring the signals over time. A tubing manifold was used to connect different buffers to the system. Cells were perfused with a peristaltic pump at 30 -60 l/min in the standard buffer (or supplemented according to each experiment) for a 30-min equilibration period and then a 30-min basal period to establish base-line respiration, FURA-PE3, and TMRM ϩ . During the 30-min equilibration period, buffer flow to the chamber was redirected to the shunt tubing for 3-5 min to check that both oxygen electrodes gave the same air-saturated buffer readings. The shunt was also opened at the end of each experiment to assess downstream electrode drift over time. Cells were imaged with a 40ϫ oil-immersion objective, and a single field-of-view was kept throughout each experiment. advantages of the respirometer over traditional closedchamber, stirred-suspension oxygen electrodes to monitor coverslip-attached cell respiration. First, the cells remain attached to their coverslips; second, the perfusion mode permits long term experiments and medium exchange; and third, simultaneous monitoring of cell viability and function can be performed.
During steady-state perfusion, the respiratory rate is the product of the flow rate multiplied by the difference between the upstream and downstream oxygen tensions. Fig. 2 illustrates basic bioenergetic information obtained from a respirometer trace ( Fig. 2A), and the associated images obtained by using the potentiometric probe TMRM ϩ and the Ca 2ϩ probe FURA-PE3 (Fig. 2B). The following respiratory rates were determined: basal (V basal ), maximal uncontrolled respiration in the presence of 2 M FCCP (V FCCP ), oligomycin (0.2 g/ml)-insensitive (V oligo ), and residual nonmitochondrial (V rot/myx ) respiration in the presence of 2 M myxothiazol plus 1 M rotenone. The concentration of FCCP was just sufficient to produce uncontrolled respiration, although this rate often was not sustained, probably because of ATP depletion.
Assuming the myxothiazol plus rotenone-insensitive rate is nonmitochondrial, mitochondria accounted for 89 Ϯ 2% ((V basal Ϫ V rot/myx )/V basal ) of basal respiration. Initial estimates (not taking account of the voltage dependence of the mitochondrial proton leak (22)) are that ATP synthesis is responsible for 69 Ϯ 4% ((V basal Ϫ V oligo )/(V basal Ϫ V rot/myx )) of basal mitochondrial respiration, whereas the remaining 31 Ϯ 4% oligomycin-insensitive rate ((V oligo Ϫ V rot/myx )/(V basal Ϫ V rot/myx )) drives H ϩ leaks across the mitochondrial inner membrane. The mitochondria exhibit a pseudo-respiratory control ratio of 6.3 Ϯ 0.9 ((V FCCP Ϫ V rot/myx )/ (V oligo Ϫ V rot/myx )) and in the basal state operate at 51 Ϯ 3% ((V basal Ϫ V rot/myx )/(V FCCP Ϫ V rot/myx )) of maximal respiration. Fig. 2C shows the effects of these additions on FURA-PE3 and TMRM ϩ fluorescence of a representative single CGN soma. Brief FCCP perfusion decreased TMRM ϩ fluorescence by ϳ50% because of mitochondrial release of TMRM ϩ and consequent re-equilibration across the plasma membrane (6), and modestly increased the FURA-PE3 340/380 nm ratio. Cytoplasmic Ca 2ϩ levels rapidly recovered during FCCP washout, whereas TMRM ϩ fluorescence took 50 min to return close to pre-FCCP values. The kinetics of TMRM ϩ recovery are primarily because of slow re-equilibration of probe across the plasma membrane, which has a low surface-to-volume ratio compared with the mitochondrial inner membrane (6,21).
Under the present conditions (elevated KCl in the presence of serum) ATP synthase inhibition by oligomycin perfusion resulted in rapid Ca 2ϩ deregulation. Thus glycolysis alone is not sufficiently active to maintain ATP levels necessary for cytoplasmic Ca 2ϩ regulation. The oligomycin-insensitive respiration was stable for only 5-10 min, after which it slowly declined (Fig. 3A, solid line). The decline is a consequence of cell death, as correcting for dead cells produced a constant state 4 rate (Fig. 3A, dashed line; see next section for an explanation of this correction). Cell viability assessed morphologically was about 62% at the time of FCCP addition. However, even after this correction respiration only reached 84% of basal, contrasting with 184% of basal when protonophore was added in the absence of oligomycin ( Fig. 2A and Fig. 4A). Because two steps in glycolysis require ATP and because ATP appears to be depleted in the presence of oligomycin, it is possible that the decreased uncontrolled respiration seen in Fig. 3A reflects compromised glycolytic pyruvate supply to the mitochondrion rather than a defect in mitochondrial substrate oxidation. To test this, the experiment was repeated in the presence of 10 mM exogenous pyruvate (Fig. 3B). Viable cell respiration following oligomycin addition was stable for a longer period and following FCCP was increased to 116% of basal. However, ATP depletion was still apparent judged by the failure to maintain cytoplasmic Ca 2ϩ homeostasis and by the slow decline in TMRM ϩ signal (Fig. 3D). Under these nonquenching conditions the TMRM ϩ signal is equally affected by changes in ⌬ p and ⌬ m values (6). Because there is no increase in state 4 respiration following oligomycin (Fig. 3B), this indicates that the decline in TMRM ϩ fluorescence largely reflects plasma membrane depolarization because of ATP deficiency rather than mitochondrial depolarization. This is confirmed by the retained respiratory control revealed by the FCCP addition, although maximal respiratory capacity is reduced. Thus oligomycin addition in the absence of exogenous pyruvate causes an ATP depletion that results in impaired glycolysis and subsequent mitochondrial respiratory inhibition because of a failure of substrate supply. This can be prevented by exogenous pyruvate even though the substrate cannot contribute to cellular ATP generation in the presence of oligomycin.

Glutamate Stimulates Granule Neuron Respiration-Switch-
ing from the standard perfusate to one supplemented with 250 M glutamate and 25 M glycine (subsequently referred to as glutamate) rapidly and robustly increased respiration by 32 Ϯ 4% ((V glutamate Ϫ V basal )/V basal ), well below the maximal respiratory capacity observed with FCCP ( Fig. 4, A versus B). Specifically, a peak glutamate response utilized 40 -50% of the reserve CGN ATP synthesis capacity. This increase was followed by a slow decline in respiration over the remaining perfusion period (Fig. 4B, solid line). Two possible explanations for the slow decline are as follows: a gradual reduction in mitochondrial activity within all CGNs or a complete loss of mitochondrial function in some neurons but not in others. Transmission, TMRM ϩ , and FURA-PE3 images containing ϳ80 -120 CGNs indicated that some cells maintained relatively normal morphology, low FURA-PE3 340/380 nm ratios, and some TMRM ϩ fluorescence, whereas other cells showed gross morphological changes, high FURA-PE3 340/380 nm ratios (followed by inevitable loss of FURA-PE3 after a further 20 -40 min), and no TMRM ϩ signals (Fig. 4, C-E). Because of these two distinct cell populations, the slow respiration decline seems better explained by the complete loss of mitochondrial function (i.e. no substrate oxidation) in the latter cells rather than a uniform reduction in mitochondrial activity in all cells. It is notable that the synchronous Ca 2ϩ deregulation of cells in the presence of oligomycin (Fig. 3) is associated with respiratory failure with no intervening uncontrolled respiration. CGNs divided into three groups as follows: those that survived 3 h of continuous glutamate (C), those that Ca 2ϩ -deregulated immediately (D), and those that recovered from initial Ca 2ϩ load imposed by glutamate but deregulated after a variable lag period (E). Twenty cells from each experiment were randomly selected for quantitative fluorescence analysis. Fractions are the number of cells in each group.
Consequently, respiration was corrected for viable cell number with each image taken (Fig. 4B, dashed line). In the corrected trace, viable CGNs maintain respiration at or slightly above basal for the entire 3-h glutamate exposure. Because experimental variation in viability was large, it is uncertain if the apparent rise in viable cell respiration during the final 60 min of glutamate reflects a true increase in viable cell metabolism above basal. Additionally, the maximal respiratory capacity of the viable cells with FCCP was within the range of that determined for cells not exposed to glutamate (Fig. 4A). Thus surviving cells do not undergo a loss of respiratory capacity, and contradictory reports in the literature (15) are most probably because of cell death.
Viability was significantly lower after 3 h of continuous glutamate exposure compared with 4 h of control perfusions (Table I). After correcting for viability, each replicate showed a reserve ATP synthesis capacity after 3 h of continuous glutamate, although this was variable (100 Ϯ 60%) and was not significantly different from that determined in control perfusions after 1 (84 Ϯ 10%; Fig 4A) or 4 h (76 Ϯ 15%; Table I).
The Initial Increased Respiration in the Presence of Glutamate Is a Combination of Increased ATP Turnover and Mitochondrial Proton Leak-To determine whether glutamatestimulated respiration was because of increased ATP demand or increased H ϩ leak, 10-min glutamate-stimulated CGNs were treated for 15 min with 0.2 g/ml oligomycin plus glutamate (Fig. 5). Brief oligomycin perfusion was used to minimize subsequent cell death (see Fig. 3). The oligomycin-insensitive respiration was significantly greater in CGNs acutely treated with glutamate compared with time-matched controls (50% of basal for glutamate treated versus 40% of basal for controls). Nonmitochondrial respiration was not significantly different between these experiments, indicating that the increased oligomycininsensitive rate was because of a greater mitochondrial H ϩ leak or utilization of the proton current for some process other than ATP synthesis.
In one experiment (not shown), CGNs were treated with glutamate for 10 min as in Fig. 5B and then perfused for 15 min with oligomycin in the absence of glutamate. The oligomycininsensitive rate (43%, with 11% nonmitochondrial) was close to that determined for control perfusions ( Fig. 5A; range 36 -42%), suggesting that the increased H ϩ leak may rapidly reverse as glutamate is removed.
To determine whether the increased H ϩ leak persists during longer glutamate exposure, CGNs were perfused for 140 min with glutamate followed by 15 min with oligomycin plus glutamate (Fig. 5C). After correction for nonviable cells, oligomycin-insensitive respiration returned to control levels and was significantly lower than after 25 min of total glutamate exposure (Fig. 5, C versus B). However, variation after additionally correcting for nonmitochondrial respiration meant that the H ϩ leak after 155 min of glutamate exposure was not significantly different from either the control or 25-min glutamate leak rates (one-way ANOVA, control, 33.7 Ϯ 1.8%; 25-min glutamate exposure, 44.5 Ϯ 0.6%; 155-min glutamate exposure, 35.7 Ϯ 3.7%).
The NMDA Receptor Is the Primary Contributor to Glutamate-stimulated Respiration-Although glutamate activates multiple receptors, it is the NMDA receptor that is responsible for excitotoxicity under the present conditions (23). To assess the NMDA receptor contribution to the metabolic demand imposed by glutamate, cells were sequentially exposed to glutamate and NMDA (Fig. 6A) or NMDA prior to glutamate (Fig.  6B). The NMDA and glutamate responses were independent of the order of addition (Fig. 6, A versus B). Respiration stimulated by NMDA was 62% of that stimulated by glutamate. Consistent with this finding, glutamate-stimulated respiration was reduced by 67% in the presence of the selective NMDA receptor antagonist MK801 (Fig. 6B). 5 M MK801 completely blocked the NMDA response, confirming that this concentration was saturating under the conditions used (Fig. 6A). The responses elicited via the NMDA and non-NMDA components were virtually additive compared with the total glutamate response (Fig. 6B).
Pyruvate Increases Initial Maximal Respiration and Protects against Glutamate Excitotoxicity-Supplementing the standard medium with 10 mM pyruvate did not affect basal respiration ( Fig. 7B versus Fig. 4B) but significantly increased the maximal FCCP rate from 84 Ϯ 9% (n ϭ 6; Fig. 4A) to 127 Ϯ 16% (n ϭ 3; Fig. 7A) above basal (student's t test, p ϭ 0.02), indicating that substrate supply to mitochondria via glycolysis is limiting in the fully uncontrolled state, as in the absence of mitochondrial ATP synthesis (Fig. 3).
In the presence of pyruvate, the initial respiratory increase to glutamate was not significantly different compared with experiments without pyruvate (n ϭ 4; 44 Ϯ 8% versus 32 Ϯ 4%, with versus without pyruvate; p ϭ 0.22, Student's t test). Although cell  Fig. 4, 7, and 8. Data are mean Ϯ S.E. of four experiments, except ϩFCCP (mean Ϯ S.D., n ϭ 2). Since there was a significant treatment effect, 1-way ANOVA with Tukey's post-hoc test was used to determine which means differed significantly at a given time. Means sharing common superscripts are not different. viability was significantly improved by pyruvate (Table I), viable cell respiration declined steadily with time after glutamate and responded weakly to FCCP (Fig. 7B), in contrast to glutamate controls that had robust FCCP responses. Although the initial respiratory response to glutamate was not enhanced, cells in 10 mM pyruvate-supplemented buffer exhibited greater initial [Ca 2ϩ ] c responses to glutamate, with mean peak FURA-PE3 340/380 nm increasing by 0.64 Ϯ 0.03 (Fig. 7, C-E; n ϭ 80 cells). Similarly, TMRM ϩ dropped to a greater extent after glutamate (71 Ϯ 1%; n ϭ 80 cells) than when using the standard buffer without pyruvate, indicating a greater plasma and/or mitochondrial depolarization.
Whereas pyruvate relieves substrate limitation, and hence increases uncontrolled respiration in the presence of oligomycin (Fig. 3B) or in the basal state (cf. Fig. 7A and Fig. 4A), the decrease in uncontrolled respiration following 3 h of glutamate exposure in the presence of pyruvate (Fig. 7B) is an apparent contradiction. More cells survive glutamate exposure in the presence of pyruvate; however, sustained [Ca 2ϩ ] c levels are higher (cf. Fig. 7C and Fig. 4C), suggesting that the respiratory capacity of viable cells may be compromised by Ca 2ϩ .
Submaximal FCCP Facilitates Cell Death during Initial Glutamate Exposure-In the presence of 10% serum, 1 M FCCP is largely bound to albumin and is not sufficient to fully relieve CGN respiratory control (data not shown). Perfusion of cells with 1 M FCCP starting 15 min prior to glutamate exposure increased respiration 41% above basal (Fig 8B). This was associated with a significant 22 Ϯ 2% TMRM ϩ reduction but no significant FURA-PE3 340/380 change (basal, 1.59 Ϯ 0.02, n ϭ 40; 1 M FCCP, 1.62 Ϯ 0.02, n ϭ 40 cells; Student's t test). Subsequent glutamate perfusion in the continued presence of 1 M FCCP resulted in a significantly lower respiratory increase (8%, n ϭ 2) compared with experiments in the absence of protonophore (32 Ϯ 4%; n ϭ 4). Additionally, in the presence of FCCP, the average peak FURA-PE3 340/380 increase following glutamate was significantly greater than without FCCP (0.50 Ϯ 0.05, n ϭ 40 cells versus 0.32 Ϯ 0.02, n ϭ 80 cells; p Ͻ 0.0001 by Student's t test). Viable cell respiration was sustained at or very close to the initial 1 M FCCP-stimulated rate throughout the experiment. During the 1st h of glutamate/ FCCP exposure, the presence of 1 M FCCP was the only condition that resulted in a significantly lower viability when compared with controls (Table I).
Excess FCCP Leads to Rapid, Uniform Changes in Cell Morphology following Glutamate Exposure without Complete Loss of Respiration-Pretreating cells with 5 M FCCP increased respiration to the same extent as 2 M FCCP, but with much slower kinetics (Fig. 9A). Whole-cell TMRM ϩ fluorescence was reduced by 90% (Fig. 9B). Subsequent glutamate addition resulted in immediate loss of Ca 2ϩ homeostasis in all cells examined and a rapid decline in respiration (Fig. 9, A and B). After 90 min of exposure, ϳ98% of cells showed changes in morphology ( Fig. 9, C and D). Using the current criteria for dead cells (i.e. loss of TMRM ϩ and Ca 2ϩ homeostasis and changes in morphology), 98% were considered dead. At this time, respiration was 19% of peak rate, indicating that at least 19% of cells considered dead were still consuming oxygen. Although this does not change the overall conclusions from the preceding experiments, it does limit the precision with which it is possible to use these criteria to correct for cell death.
Viable Cell FURA-PE3 Ratios Show an Inverse Correlation with Respiratory Capacity-Because the cells are perfused under conditions of partial plasma membrane depolarization (25 mM K ϩ ), [Ca 2ϩ ] c is elevated because of Ca 2ϩ channel activation (24). In the absence of NMDA receptor activation (control or glutamate plus MK801) [Ca 2ϩ ] c decreased significantly during the course of the experiment (Fig. 10). At the end of the experiment, [Ca 2ϩ ] c in glutamate-exposed cells was not significantly higher than before glutamate addition, although with glutamate plus pyruvate [Ca 2ϩ ] c was significantly enhanced. There was a significant (p ϭ 0.02, r 2 ϭ 0.314) negative correlation between the FCCP respiratory response and the average viable cell FURA-PE3 340/380 nm ratio at the end of each experiment, indicating that chronically elevated [Ca 2ϩ ] c may lead to loss of cell substrate oxidation capacity (Fig. 11). This is consistent with the reduced respiratory capacity observed after synchronous Ca 2ϩ deregulation with oligomycin (Fig. 3).
After 4 h of control perfusions, viable cell TMRM ϩ was not significantly different from the basal 30 -60-min period (Fig.   10B). Perfusions with glutamate and glutamate plus pyruvate resulted in significantly lower viable cell TMRM ϩ compared with the basal 30 -60-min period prior to glutamate addition, consistent with NMDA receptor-mediated plasma membrane depolarization, while 0.5 M MK801 attenuated this effect. DISCUSSION Bioenergetic parameters measure the extents of displacement from equilibrium (e.g. Gibbs free energy, protonmotive force, membrane potential, redox state, and ATP/ADP ratios), the rates of energy flux (e.g. metabolic rates, ATP turnover, proton flux, and respiration), or combinations of these parameters (e.g. proton conductance that is determined as proton leak per mV of proton electrochemical gradient). Although there is considerable literature detailing the rates of glucose or pyruvate utilization and lactate or CO 2 production in CGN cultures (reviewed in Refs. [25][26][27], technical problems have limited the investigation of the primary mitochondrial flux parameter, namely the rate of electron transport or proton cycling as monitored by respiration. The classic oxygen electrode chamber has proven to be the single most important apparatus for monitoring the function of isolated mitochondria, because careful experimental design allows determination of ATP turnover, proton leaks, maximal electron transport activity, and other extensive parameters (for review see Ref. 28). However, a conventional cylindrical chamber is entirely unsuitable for monitoring the respiration of a neuronal monolayer attached to a substrate. This study shows that the respiration of a neuronal monolayer can reliably be measured over extended periods using flow-through polarographic oxygen electrodes.  (17,31) reported value of 3.9 -4 nmol of O/min ϫ 10 6 CGNs in 25 mM K ϩ . However, with the exception of the present study, it must be remembered that all cited investigations required scraping or trypsinization of the cells to remove them from their substrates, almost certainly resulting in some cell damage. Significantly, we are aware of no studies that report the extent of protonophore-relieved respiratory control seen in such preparations.
In the basal state, in situ CGN mitochondria operate at about 50% of maximal respiration ( Fig. 2A). Oligomycin reduced respiration by 61%, suggesting that ATP turnover accounts for this percentage of basal cell respiration. However, this is an underestimate because oligomycin hyperpolarizes mitochondria, enhancing the activity of the highly potential dependent constitutive inner membrane proton leak (22). Rolfe et al. (32) showed that oligomycin reduced hepatocyte respiration by ϳ64%; however, when corrected for membrane potential changes in H ϩ leak, ATP turnover was calculated to account for 69% of basal cell respiration. If CGN H ϩ leak is similarly dependent on ⌬ m , then the reported oligomycin-insensitive respiration is about 20% higher than it would be in the absence of inhibitor (Fig. 12). By using this correction, a better estimate is that 68% of CGN respiration supports ATP turnover, 21% supports H ϩ leak, and the remaining 11% (in the presence of rotenone plus myxothiazol) supports nonmitochondrial reactions. Pyruvate supplementation did not significantly affect basal respiration, indicating that in this state glycolysis provides adequate substrate to the mitochondrion. However, during uncontrolled respiration glycolytic substrate supply to the mitochondria limits respiratory chain activity because pyruvate significantly enhanced the FCCP effect. Moreover, under the present incubation conditions, which prevent induction of apoptosis, these CGN preparations had insufficient glycolytic ac-tivity to maintain prolonged basal function in the presence of oligomycin (Fig. 3). This is in contrast to previous work with CGNs maintained in low K ϩ media (12,33). Most important, pyruvate decreases cell death (Fig. 3) in the presence of oligomycin even though it cannot contribute to ATP production, suggesting that mitochondrial polarization is protective, perhaps by maintaining NADPH and glutathione reduction.
Acute Effects of Glutamate-A key finding from this study is that under the present conditions glutamate exposure acutely stimulates CGN respiration submaximally. From first principles it would be predicted that NMDA receptor activation would lead to an increase in CGN mitochondrial respiration on three counts as follows: utilization of the proton gradient to drive Ca 2ϩ uptake into the matrix, increased plasma membrane Ca 2ϩ -ATPase activity to expel Ca 2ϩ entering the cell, and increased Na ϩ /K ϩ -ATPase activity in response to increased Na ϩ entry. In fact, a 32% increase in basal respiration was observed immediately following glutamate infusion (Fig. 4B). This increase utilized about 40 -50% of the reserve ATP-generating capacity of CGN mitochondria. Pyruvate supplementation did not further increase the response to glutamate (Fig. 7B). Thus, glycolysis was capable of supplying necessary substrate to sustain the higher respiratory chain activity, and oxidative phosphorylation was capable of supplying sufficient ATP to meet the energy demand imposed by glutamate.
Atlante et al. (15) exposed CGNs to glutamate/glycine in Mg 2ϩ -free low K ϩ medium for 30 min, washed, and suspended the neurons by scraping and determined respiration in the presence of 10 mM glucose. They reported a 20% decrease in respiration relative to control cells; however, because cell viability was not monitored, it is unclear whether this represented a change in respiration per viable cell. Surprisingly, they observed a total inhibition of respiration by ouabain both in Point b is the estimated respiration supporting the proton leak in the basal state, assuming that the oligomycin rate exceeds that in the basal state by 20% because of the increased ⌬ m (32). Mitochondrial membrane potential with oligomycin was assumed to be similar to that determined for mitochondria within synaptosomes (43). A 20-mV potential drop was assumed sufficient for maximal respiration (43). Because substrate oxidation rate is a linear function of membrane potential (22,44), potentials associated with points d-f were set according to this slope (i). The nonlinear control proton leak curve (ii) was drawn based on points a-c, whereas the curve after glutamate treatment (iii) was drawn based on points a and d, assuming that respiration driving the leak is similar to the control curve at potentials giving maximal respiration.
control and glutamate-exposed neurons, suggesting that their assay was insufficiently sensitive to detect the respiration under conditions of low ATP demand. In their study, classic respiratory control was performed on homogenates from parallel cell preparations, and a 50% decrease in state 3 respiration was observed with succinate as substrate with little change in state 4 rate (15). This was incorrectly interpreted as indicating rapid "uncoupling" of oxidative phosphorylation, i.e. an increased H ϩ leak, which would have manifested as an increased state 4 respiration. The decreased state 3 rate could reflect either an inhibited respiratory chain or a decreased yield of mitochondria because of cell death.
In a related study, Almeida et al. (34) exposed cortical neurons cultured in 78-cm 2 dishes to glutamate for 15 min, restored the culture medium, and monitored respiration after 24 h by trypsinizing the cells, washing, storing on ice, and assaying in a conventional Clark-type oxygen electrode chamber. They reported a modest decrease in respiration, but the ability of the preparation to respond to succinate indicated extensive damage resulting from the manipulations.
The glutamate-induced initial increase in respiration can be analyzed both to determine the proportion of the response because of NMDA-and non-NMDA-mediated effects and also to distinguish between increased ATP demand and increased utilization of the H ϩ current via H ϩ leaks (classic uncoupling) or to drive matrix Ca 2ϩ accumulation. Comparison of the effects of glutamate and the selective agonist NMDA (Fig. 6) shows that 62% of the increase in respiration is because of NMDA receptor activation. Clearly ion influx through this receptor stimulates ATP turnover but not to such a degree that demand exceeds supply. Thus, the protective effect of MK801 (Table I; Fig. 10) is likely a consequence of limiting one or more Ca 2ϩ -dependent processes rather than reducing ATP demand. The non-NMDAdriven respiration is additive to the NMDA effect and could be because of the metabolic effects of the metabotropic glutamate receptor, AMPA/kainate receptor activation, or glutamate uptake into neurons or contaminating astrocytes causing an increased ATP demand by the Na ϩ /K ϩ -ATPase. However, no respiratory inhibition was observed on addition of the glutamate transport inhibitor DL-threo-␤-benzyloxyaspartic acid (100 M, data not shown).
The analysis to distinguish ATP turnover from uncoupled respiration was performed. The second analysis was performed by acute addition of oligomycin to inhibit ATP synthase (Fig. 5). Again, it should be noted that under these conditions these CGN preparations do not possess sufficient glycolytic activity to maintain Ca 2ϩ homeostasis in the absence of mitochondrial ATP synthesis, and rapid cell death ensues. The slight 10% increase in oligomycin-insensitive respiration observed 25 min following glutamate addition (Fig. 5) does not persist, confirming the absence of a significant population of rapidly respiring "uncoupled" mitochondria in the glutamate-exposed cells. The transient increase could (from first principles) be due to the utilization of the H ϩ current to drive matrix Ca 2ϩ accumulation during the initial period of glutamate exposure. However, in this case a similar increase would be predicted following oligomycin alone, which causes cytoplasmic Ca 2ϩ to rise to a similar extent as that in glutamate-treated cells (Fig. 5).
The CGN preparations used in Fig. 4B and Fig. 5, B and C, responded similarly to glutamate, with a peak respiration occurring 6 min after glutamate and reaching the same extent (30 Ϯ 5% in Fig. 5B, 39 Ϯ 7% in Fig. 5C, and 32 Ϯ 4% in Fig.  4B). Oligomycin blocked oxidative phosphorylation after a short delay (Figs. 3 and 5). During this time, glutamate-stimulated respiration declined to 19 -21% above basal or about 0.18 nmol of O/min ϫ 10 6 viable cells (Fig. 5C and Fig. 4B), suggesting that 50% of the glutamate-induced increase is attributable to a mitochondrial proton leak. However, because oligomycin hyperpolarizes mitochondria and glutamate slightly depolarizes mitochondria (6), it is necessary to correct the leak for changes in ⌬ m . A more realistic estimate is that H ϩ leak accounts for about 15% of the increase with glutamate, whereas the remaining 85% would be attributable to ATP turnover (Fig. 12). Because this uncoupling rapidly reverses upon glutamate washout, it does not seem to reflect a small minority of cells entering a fully uncoupled state but rather may be a consequence of activating a regulated H ϩ conductance (e.g. an uncoupling protein).
Chronic Glutamate Exposure-An important finding of this study is that mitochondria in surviving CGNs do not show diminished respiratory capacity after 3 h of continuous glutamate exposure. Previous attempts to monitor the effects of glutamate exposure on neuronal respiration (e.g. Ref. 35) have not allowed for stochastic cell death, and so it has been unclear whether reported chronic decreases in respiration were because of cell death or a uniform respiratory inhibition. If the assumption is made that Ca 2ϩ deregulated cells having altered morphology and no TMRM ϩ fluorescence do not respire, then the 45% remaining viable cells after 3 h glutamate show no decrease in respiration, or even a statistically insignificant increase (Fig. 4B). These cells retain protonophore-releasable respiratory control. Moreover, viable cell respiration with 1 M FCCP is maintained at the pre-glutamate level throughout 3 h glutamate exposure (Fig. 8B), again suggesting no loss of mitochondrial function.
The analysis of cell respiration makes the assumption that Ca 2ϩ -deregulated, morphologically altered cells lacking TMRM ϩ fluorescence do not consume oxygen. This was tested by using oligomycin to cause a synchronous Ca 2ϩ deregulation in the whole-cell population that was monitored in the respirometer (Fig. 3). No prominent increase in oligomycin-insensitive respiration was seen despite an initial FURA 340/380 nm ratio increase from about 1.4 to 3. Furthermore, extended incubation with oligomycin resulted in further respiratory inhibition, reduction in TMRM ϩ signal, and altered cell morphology. From this perspective, the assumption that glutamate-treated cells that undergo these changes do not respire seems valid. However, when 98% of cells were considered dead (i.e. nonrespiring) following treatment with excess FCCP plus glutamate, respiration declined by 81% rather than the expected 98% (Fig. 9). This indicates that (a) some cells considered "dead" do consume oxygen, and (b) the viable cell respiration traces overestimate the true rate at which these cells consume oxygen. Although this uncertainty does not affect our conclusions that cells exposed to glutamate retain normal mitochondrial function until cell death, it is clear that more experiments are required to determine the variable(s) that more accurately differentiate an oxygen-consuming cell from one that is truly dead.
Because rat brain mitochondria possess a high Ca 2ϩ uniport activity sufficient to utilize the full capacity of the respiratory chain when extramitochondrial [Ca 2ϩ ] is elevated (36), the lack of respiratory increase associated with the increased [Ca 2ϩ ] c following oligomycin was initially surprising (Fig. 3), suggesting that elevated [Ca 2ϩ ] c was not sufficient to activate the uniporter significantly, that the uniporter was inhibited, or that mitochondrial respiration was severely limited because of lack of substrate. Glycolytic failure did occur with oligomycin because uncontrolled respiration with pyruvate supplementation was much greater than with glucose alone (Fig.  3). Despite this, the failure to observe a prominent increase in respiration after oligomycin-induced Ca 2ϩ deregulation can-not be attributed to substrate supply problems because bypassing glycolysis with exogenous pyruvate did not stimulate state 4 respiration.
The resting [Ca 2ϩ ] c of granule neurons in 25 mM KCl is about 100 nM (37), well below the threshold of about 500 nM for initiating mitochondrial Ca 2ϩ accumulation (36). Even with 70 mM KCl to fully activate L-type Ca 2ϩ channels, steady-state [Ca 2ϩ ] c only reaches about 200 nM (37). Therefore, the basal rate of Ca 2ϩ entry in 25 mM KCl is quite low relative to influx necessary to achieve mitochondrial loading. In Fig. 3, C and D, it is seen that oligomycin caused FURA 340/380 nm to plateau at about 3.0; this is more readily seen in Fig. 3D with pyruvate present. The FURA plateau at 3 likely represents the critical concentration of about 500 nM at which the mitochondria began accumulating Ca 2ϩ at a slow rate that matched the continued rate of Ca 2ϩ entry across the plasma membrane (36). Subsequent FCCP addition further increased this ratio to Ͼ4 in Fig.  3D. Because FCCP depolarizes mitochondria, thus inhibiting Ca 2ϩ accumulation, the further FURA increase is consistent with the plateau at 3 representing the point at which mitochondria began accumulating Ca 2ϩ . The plateau was not sustained in the cell depicted in Fig. 3C, suggesting failure of mitochondrial Ca 2ϩ accumulation. This is consistent with the rapid loss of TMRM ϩ in this cell. Careful inspection of the respiration traces in Fig. 3, A and B, provides an indication as to the amount of oxygen utilized for Ca 2ϩ uptake. Oligomycin addition initially reduced respiration to 33 and 34% of basal rate, respectively. This was followed by a very small increase (about 4% in both cases) to a final constant state 4 rate. In both cases the increase amounted to 0.04 nmol of O/min ϫ 10 6 cells. We suggest that this small increase represents the utilization of the protonmotive force to drive Ca 2ϩ accumulation under these conditions. This is an exceedingly small rate of Ca 2ϩ uptake, consistent with the idea that Ca 2ϩ entry through L-type Ca 2ϩ channels is relatively low under the current conditions. Thus, [Ca 2ϩ ] c did not increase to such an extent to prominently activate the Ca 2ϩ uniporter.
Taken together, the data indicate that glutamate-induced DCD is associated with a loss of respiration rather than pronounced uncoupling. Does Ca 2ϩ deregulation precede or follow loss of respiration? This remains an unresolved issue because these alternatives cannot adequately be resolved in these experiments, particularly given the synchronous changes in TMRM ϩ and FURA-PE3 signals as cells begin to lose Ca 2ϩ homeostasis. Although there is no indication of a slow decline in mitochondrial function of viable cells after glutamate, it may be that mitochondrial respiration rapidly shuts down, resulting in Ca 2ϩ deregulation. This possibility cannot be distinguished from the alternative that Ca 2ϩ deregulation results in loss of respiration.
Chronic Glutamate Exposure with Pyruvate-Pyruvate supplementation significantly attenuates glutamate excitotoxicity (Table I). This protection cannot be attributed to enhanced mitochondrial substrate supply because glutamate did not maximally stimulate respiration and pyruvate did not further enhance glutamate-stimulated rate, in contrast to its effects on uncontrolled respiration. Surprisingly, cells that remained viable after 3 h of continuous glutamate exposure exhibited a pronounced reduction in maximal respiratory capacity (Fig.  7B), indicating an impairment downstream of pyruvate supply to the mitochondria. Four observations indicate that elevated cytoplasmic Ca 2ϩ levels may be involved. First, there was a significant negative correlation between [Ca 2ϩ ] c and mitochondrial responsiveness to FCCP (Fig. 11). Second, Ca 2ϩ transients immediately following glutamate addition were significantly greater with pyruvate present (Fig. 7). Third, cytoplasmic Ca 2ϩ levels after 3 h of glutamate exposure were significantly higher than before glutamate addition (Fig. 10). Fourth, oligomycin-induced Ca 2ϩ deregulation resulted in impaired uncontrolled respiration (cf. Fig. 3B and Fig. 7A). Clearly, pyruvate protection from glutamate excitotoxicity cannot be ascribed to improved mitochondrial function, as indicated by respiratory capacity. This is not surprising given that chronic glutamate exposure imposed very little energy demand above the basal state (Fig. 4). Reduction in the reserve ATPgenerating capacity is therefore not an event associated with DCD. Previous studies (18, 38 -42) have shown that pyruvatesupplemented media protect neurons from glutamate-or NMDA-mediated excitotoxicity. However, it is unclear how pyruvate paradoxically increases Ca 2ϩ influx and protects against excitotoxicity.
Conclusions-The oxygen electrode has proven over the years to be the most versatile apparatus for investigating the function of isolated mitochondria and cells that exist in suspension. The lack of a similar device for cultured neurons and other cells attached to a substrate has restricted investigations of in situ mitochondrial function in such preparations. By combining continuous flow oxygen electrodes with a digital fluorescence imager, it is possible to obtain quantitative information on mitochondrial and cellular energy supply and demand while simultaneously viewing a representative field to monitor cell viability together with parameters such as membrane potentials and [Ca 2ϩ ] c . The hypotheses that glutamate stimulates CGN respiration maximally, that the increase is entirely the result of increased ATP demand, and that excitotoxicity is associated with a loss of reserve ATP-generating capacity are not valid for the present experimental conditions, although the data support the hypothesis that the NMDA receptor is a significant contributor to the overall energy demand imposed by glutamate. It has not escaped our attention that studies to investigate neuronal bioenergetic changes associated with apoptosis and other stresses are equally feasible with this technique.