Mitochondrial Dysfunction Is a Primary Event in Glutamate Neurotoxicity

Abstract

Excitotoxic neuronal death, associated with neurodegenerative disorders and hypoxic insults, results from excessive exposure to excitatory neurotransmitters. Glutamate neurotoxicity is triggered primarily by massive Ca²⁺ influx arising from overstimulation of the NMDA subtype of glutamate receptors. The underlying mechanisms, however, remain elusive. We have tested the hypothesis that mitochondria are primary targets in excitotoxicity by confocal imaging of intracellular Ca²⁺([Ca²⁺]_i) and mitochondrial membrane potential (ΔΨ) on cultured rat hippocampal neurons. Sustained activation of NMDA receptors (20 min) elicits reversible elevation of [Ca²⁺]_i. Longer activation (50 min) renders elevation of [Ca²⁺]_i irreversible (Ca²⁺ overload). Susceptibility to NMDA-induced Ca²⁺ overload is increased when the 20 min stimuli are applied to neurons pretreated with electron transport chain inhibitors, thereby implicating mitochondria in [Ca²⁺]_ihomeostasis during excitotoxic challenges. Remarkably, ΔΨ exhibits prominent and persistent depolarization in response to NMDA, which closely parallels the incidence of neuronal death. Blockade of the mitochondrial permeability transition pore by cyclosporin A allows complete recovery of ΔΨ and prevents cell death. These results suggest that early mitochondrial damage plays a key role in induction of glutamate neurotoxicity.

Excitotoxicity, the process by which overactivation of excitatory neurotransmitter receptors leads to neuronal cell death, must be understood in terms of the intervening cascade of events (Choi, 1987, 1988, 1995). Compelling evidence supports a key role for massive Ca²⁺ influx through the NMDA receptor (NMDAR) channel as a trigger of glutamate neurotoxicity: (1) selective blockade of NMDARs markedly reduces the extent of excitotoxic damage (Choi et al., 1988; Michaels and Rothman, 1990); (2) neurons survive when challenged in absence of extracellular Ca²⁺ or presence of intracellular Ca²⁺([Ca²⁺]_i) chelators (Choi, 1987; Tymianski et al., 1993a,b); and (3) glutamate receptor overstimulation induces elevations of [Ca²⁺]_i that precede cell death (de Erausquin et al., 1990; Michaels and Rothman, 1990; Randall and Thayer, 1992; Tymianski et al., 1993a,b; Witt et al., 1994). Oxidative stress is thought to be a contributing factor in glutamate neurotoxicity through generation of oxygen radicals that disrupt cell functions (Dykens et al., 1987; Dawson et al., 1991; Coyle and Puttfarcken, 1993; Lafon-Cazal et al., 1993; Lipton et al., 1993). However, the link between elevation of [Ca²⁺]_i, free radicals, and neuronal death remains unclear (Choi, 1995).

Mitochondria are unique among cell organelles in their involvement in the concerted consumption of oxygen, production of oxygen radicals, and mobilization of [Ca²⁺]_i(Gunter and Pfeiffer, 1990; Thayer and Miller, 1990; Miller, 1991;Beal, 1992; Coyle and Puttfarcken, 1993; Rizzuto et al., 1993;Jouaville et al., 1995; White and Reynolds, 1995). Given that excessive Ca²⁺ accumulation in mitochondria uncouples electron transfer from ATP synthesis (Beatrice et al., 1980; Gunter and Pfeiffer, 1990; Bernardi et al., 1994), and considering that impairment of energy metabolism (Wang et al., 1994) increases generation of free radicals (Halliwell and Gutteridge, 1989; Coyle and Puttfarcken, 1993), mitochondria emerge as a plausible link between elevation of [Ca²⁺]_i and glutamate neurotoxicity. The pivotal role of mitochondria in excitotoxicity has been highlighted recently by the deterministic influence of mitochondrial function in the decision between apoptotic or necrotic cell death (Ankarcrona et al., 1995; Bonfoco et al., 1995) and by the protective action of transforming growth factors presumably conferred by enhancing mitochondrial energetics (Mattson et al., 1993; Prehn et al., 1994).

Here, we focus on the following question: is the mitochondrion the sensor that converts elevation of [Ca²⁺]_ifrom a physiological modulator into a trigger for cell death? We show that mitochondria are involved in Ca²⁺ sequestration during an excitotoxic insult (Schinder et al., 1995) and that this Ca²⁺ current induces sustained mitochondrial depolarization, which closely parallels the incidence of neuronal death. Our findings suggest that neuronal injury follows from excessive influx of Ca²⁺ via the NMDAR that, in turn, attenuates the mitochondrial membrane potential (ΔΨ) leading to the opening of the permeability transition pore (PTP), further collapsing ΔΨ. Blockade of the PTP with cyclosporin A (CsA) leads to the recovery of ΔΨ and promotes cell survival. The results outline a novel way of thinking about excitotoxic cell death, shifting the focus of attention from a global event staged by intracellular Ca²⁺ toward a more compartmentalized question centered on mitochondrial homeostasis.

MATERIALS AND METHODS

Hippocampal cultures. Mixed hippocampal neuronal/glial cultures were established as described (Schinder and Montal, 1993). Briefly, hippocampi from E17–E19 rat embryos were incubated in basic saline solution (BSS) containing (in mm): 137 NaCl, 3.5 KCl, 0.4 KH₂PO₄, 0.33 Na₂HPO₄ · 7 H₂O, 5 trimethylaminoethanesulfonic acid, pH 7.4, and 10 glucose with 0.25% trypsin for 15 min at 37°C. Tissue was washed and dissociated by trituration with a Pasteur pipette. Cells were centrifuged for 5 min at 200 × g, and the pellet was resuspended in BSS. Cells were plated (10⁵ cells/cm²) and maintained as described.

Cell death assays. In vitro excitotoxicity is commonly investigated using agonist exposures that range from 5 min to 24 hr (Choi, 1987; Weiss et al., 1994). In our conditions, mild and severe types of neuronal death were selectively achieved with exposures of 20 and 50 min, respectively. Culture medium was removed, neurons were rinsed with 10 ml of BSS + 1.1 mm CaCl₂ + 10 μm glycine, and treated with 200 μm NMDA (dissolved in the same solution) for either 20 (de Erausquin et al., 1990; Michaels and Rothman, 1990; Lafon-Cazal et al., 1993; Witt et al., 1994) or 50 min (Tymianski et al., 1993a,b) at 23 ± 1°C. Treatments were terminated by addition of culture medium (Schinder and Montal, 1993) supplemented with 20 μm dizolcipine [MK-801] (Research Biochemicals, Natick, MA) to prevent activation of NMDAR attributable to residual NMDA or glutamate released from synapses (Hartley and Choi, 1989). Cultures were returned to the incubator and cell death was assessed 18–24 hr later by trypan blue staining (0.4%). When indicated, CsA (Sandoz, Basel, Switzerland) was dissolved in ethanol or dimethyl sulfoxide (DMSO; 42 mm) and added to solutions held at 37°C under vigorous stirring to a final concentration of 1.6 μm. To mimic those conditions used in imaging experiments, CsA was added 40 min before the NMDA challenge and removed after 2 hr. A minimum of 500 cells were counted blind on each culture dish. For each experiment, two identical culture dishes were treated for control and 20 and 50 min exposure to NMDA, or, in the CsA experiments, control and 20 min NMDA in absence and presence of CsA.

Calcium imaging. Neurons were Fluo-3-loaded for 75 min at a bath concentration of 5.5 μm Fluo-3 AM (Molecular Probes, Eugene, OR) dissolved in DMSO (Kao et al., 1989). Imaging was performed with an argon laser confocal system (Bio-Rad MRC600) equipped with 488 nm excitation/515 nm emission filters on a Zeiss microscope (Thornwood, NY) with a 20× water immersion objective (numerical aperture 0.4). Neurons were constantly superfused at a rate of 3 ml/min with BSS + 1.1 mm CaCl₂ + 2 mm MgCl₂ + 10 μm glycine (control solution). NMDA was applied in Mg²⁺-free solution. Images were sampled and collected at 8–15 sec intervals for 2 hr (up to 4 hr for 50 min NMDA experiments) and recorded on videotape. Slower sampling rates (1 image/2 min) and lower laser intensity (1/6 of standard power) yielded identical results, suggesting that phototoxic effects are minimal. Signals obtained during Ca²⁺ overload were not attributable to saturation of dye, as fluorescence increased further after addition of 5 μm ionomycin (Calbiochem, La Jolla, CA) in the presence of 20 mm CaCl₂, pH 8.0. Digital images were acquired either on-line or off-line with a Macintosh computer equipped with a frame grabber (Data Translation) using Image 1.47 software (W. Rasband, National Institutes of Health). A total of 20–25 neurons was studied in each experiment; somatic regions of ∼170 pixels were chosen for quantitation of fluorescence intensity. Regions without neurons or processes were used for background subtraction. For calibration experiments (Kao et al., 1989; Vergara and DiFranco, 1992), [Ca²⁺]_i was calculated as [Ca²⁺] = K_d(F − F_min)/(F_max − F), where K_d was assumed to be 400 nm, F_max was measured at the end of the experiment by treating neurons with 5 μm ionomycin,F_min was derived fromF_max, and background fluorescence was obtained after addition of 0.02% saponin.

Imaging of ΔΨ. Neurons were incubated for 30 min in the presence of 0.1 μm tetramethylrhodamine ethylester (TMRE; Molecular Probes). TMRE was added to all solutions used during these experiments, which typically lasted 2 hr. When indicated, 1.6 μm CsA was present in all solutions throughout the experiment, with a preincubation time of 40 min before the NMDA challenge. Three to seven neurons were studied in each experiment. Imaging (Farkas et al., 1989) was as described above, except that samples were collected at 1 image/90 sec and a 514/550 nm filter and a 40× water immersion objective (numerical aperture 0.9) were used. The confocal pinhole was closed to 1/5 to 1/3, and signals were amplified 1.5× with a digital zoom. Fluorescence intensity was quantified in a perinuclear ring (∼1000 pixels) in which mitochondrial clusters are located (see Fig. 4B). Baseline fluorescence (F₀) was measured as average normalized fluorescence emitted during the initial 5 min of the experiment. Control experiments showed that selective depolarization of the plasma membrane (35 mm KCl, 0 Ca²⁺) evoked minor changes in mitochondrial fluorescence. Results indicate that decrements of mitochondrial fluorescence primarily reflect mitochondrial depolarization and not reequilibration of dye in response to NMDA-induced plasma membrane depolarization. This assertion was confirmed by the finding that 100 μm carbonyl cyanidem-chlorophenylhydrazone, uncoupler of oxidative phosphorylation, decreases TMRE fluorescence (Farkas et al., 1989) to an extent comparable to that produced by the NMDA treatments.

Fig. 4.

NMDAR overstimulation collapses ΔΨ.A, Digital images of neurons exposed to TMRE before removal of NMDA (top frame indicated by 0:00:00 time), during 20 min stimulation with 200 μm NMDA, and after removal of NMDA (wash). Collapse of ΔΨ is shown as a decrease in fluorescence intensity during exposure to NMDA. Partial repolarization occurs after the stimulus is removed. Pseudocolor scale represents arbitrary fluorescence intensity values ranging from 0 to 255. Scale bar, 20 μm. B, Time course of ΔΨ measured in arbitrary fluorescence units for neurons stimulated with NMDA for 0 (control), 0.5, 20, and 50 min and recorded during 2 hr. Each trace represents mitochondrial fluorescence from an individual neuron normalized to its initial baseline value (F₀), and each set of recordings belongs to a single representative experiment.Inset, Perinuclear ring of mitochondrial fluorescence used for quantitation. C, Mitochondrial fluorescence signals, as shown in B, are quantified as Peak Depolarization = F_min/F₀, an indicator of transient mitochondrial depolarization, and Recovery Ratio = F_end/F₀, an indicator of mitochondrial repolarization after cessation of the NMDA stimulus. Bars plotted for 0, 0.5, 20, and 50 min NMDA represent mean ± SEM for n = 28, 28, 74, and 84, respectively. For both bar charts, ANOVA analysis revealed that populations are different with p < 0.00001. **, Statistically significant differences (p < 0.0001) compared with all other treatments using the post hoc test. D, Effect of [Ca²⁺]_ext on depolarization of ΔΨ elicited by 20 min NMDA. Bar chart for Peak Depolarization and Recovery Ratio was calculated and plotted as in C. *, Statistically significant difference with p < 0.002. Inset, Comparison of the time courses of ΔΨ in the presence of [Ca²⁺]_ext = 1.1 or 5.5 mm. Traces are averages from all recordings (5.5,n = 33; 1.1,n = 74).

All experiments reported were performed at 23 ± 1°C on neurons cultured for 14–17 d.

Statistics. Pooled data are presented as mean ± SEM. For comparisons between two groups, a two-tailed Student’st test was used. When three or more groups were compared (Figs. 1

[Ca²⁺]_i (4) followed by a slow decrease to a lower level lasting several minutes (5), which gradually increases to achieve a higher plateau [F_plateau, (6)]. [Ca²⁺]_i returns to baseline levels shortly after the NMDA stimulus is removed (7). Record obtained from the neuron marked with an arrowhead. Scale bar, 40 μm. B, Same as A, but neurons were stimulated for 50 min with NMDA. A similar time course was recorded, but heightened [Ca²⁺]_i(6) failed to return to baseline levels [Ca²⁺ overload, (7)]. Calibration of [Ca²⁺]_i yielded the following values: (1) [Ca²⁺]_baseline = 29 ± 2 nm; (2) [Ca²⁺]_K⁺ = 0.91 ± 0.10 μm; (4) [Ca²⁺]_peak = 1.5 ± 0.2 μm; (5) [Ca²⁺]_valley = 0.29 ± 0.03 μm; (6) [Ca²⁺]_plateau = 2.1 ± 0.2 μm; n (number of cells) > 47. C, Extent of Ca²⁺ overload versus duration of the NMDA challenge. Ca²⁺ overload was calculated as the ratio of the fluorescence value at the end of the experiment [F_end, (7)] to the highest value achieved during the challenge [F_plateau, (6)]. Data are mean ± SEM. ANOVA analysis revealed that populations are different with p < 0.00001. *, Statistically significant difference (p < 0.005) obtained using the post hoc test compared with all other treatments; **p < 0.0001. Numbersin parentheses denote population sizes. {/ANNT;64064n;;66880n;72864n}

C, 4C), a single-factor ANOVA, followed by a post hoc Bonferroni t test with correction for multiple comparisons, was used.

Fig. 1.

NMDAR overstimulation evokes persistent elevation of [Ca²⁺]_i in hippocampal neurons. A, Time course of changes in [Ca²⁺]_i of a neuron in response to K⁺-induced membrane depolarization (35 mm KCl, 30 sec, K⁺), followed by 20 min exposure to 200 μm NMDA. Values are plotted as relative pixel fluorescence intensity (F) normalized to baseline fluorescence values (F₀); upward deflection denotes an increase in [Ca²⁺]_i. Horizontal bars(gray background) indicate duration of treatment.Numbers indicate critical points in the Ca²⁺signal and code panels 1 to 7 that are pseudocolor digital images of Fluo-3 fluorescence (Kao et al., 1989).Pseudocolor scale represents arbitrary fluorescence intensity values ranging from 0 to 255. The KCl control stimulus elicits a rapid and reversible increase of [Ca²⁺]_i (2). NMDA evokes a fast elevation of

RESULTS

NMDAR overstimulation induces neuronal cell death

To assess the effects of NMDAR overstimulation on mitochondrial Ca²⁺ sequestration and cell death, experimental paradigms that elicit either moderate or severe extents of neuronal death were developed. Neurons were exposed to 200 μm NMDA for variable time intervals, and cell death was determined 18–24 hr later using the trypan blue exclusion assay. Neuronal death in hippocampal cultures increased with the duration of the NMDA challenge: 39 and 67% cell death were elicited by NMDA pulses of 20 and 50 min (Table1), respectively. This finding is consistent with the notion that cellular damage increases with the duration of Ca²⁺ influx through the NMDAR (Tymianski et al., 1993a).

Table 1.

Neuronal damage assessed by cell death, Ca²⁺overload, and mitochondrial depolarization

	20 min NMDA	50 min NMDA
Fraction of dead cells	0.39 ± 0.04 (23,601)	0.67 ± 0.06 (11,375)
Ca2+overload	0.13 ± 0.06 (107)	0.71 ± 0.06 (200)
Mitochondrial depolarization	0.37 ± 0.04 (74)	0.76 ± 0.02 (84)

The fraction of dead cells was calculated from counts of cultures stained with trypan blue 24 hr after treatments. Cell death in cultures treated with vehicle (0.07 ± 0.01; n = 18,846) was subtracted. Ca²⁺ overload was determined as in Figure1. Mitochondrial depolarization was derived as 1-Recovery Ratio (Fig.4). Values denote mean ± SEM; numbers in parentheses indicate the population of cells examined.

NMDA produces transient or sustained increases in [Ca²⁺]_i depending on pulse duration

To evaluate the time course of change in [Ca²⁺]_i evoked by NMDAR overstimulation, time-lapse fluorescent imaging was performed on hippocampal neurons loaded with the Ca²⁺ indicator Fluo-3 AM. As shown in Figure 1A, 20 min application of NMDA evokes a fast rise in [Ca²⁺]_i to a peak that decays slowly to an intermediate level before rising to a higher plateau. Calcium returns to baseline levels after withdrawal of the stimulus (recovery). However, a 50 min pulse of NMDA induces Ca²⁺ overload (Fig.1B), as indicated by the persistence of Ca²⁺ elevation. The extent of Ca²⁺ overload increases with duration of the NMDA challenge (Fig. 1C). Comparison of the incidence of cell death with that of Ca²⁺overload produced by these treatments, especially at 20 min (see Table1), reveals an apparent disparity: the incidence of cell death is significantly higher than that of Ca²⁺ overload. This indicates that recovery of [Ca²⁺]_i to baseline levels is not necessarily indicative of cell survival and suggests that additional steps intervene between elevation of [Ca²⁺]_i and cell death.

The susceptibility to NMDA-induced Ca²⁺ overload is augmented by mitochondrial function inhibitors

To test whether mitochondria lie on the main pathway to excitotoxicity, we studied the effects of specific inhibitors of mitochondrial function on cell death induced by 20 min stimulation with NMDA. Antimycin A and rotenone, specific inhibitors of complex III and complex I of the electron transport chain, were used to block electron transfer and proton extrusion mechanisms, thereby decreasing ΔΨ (see Fig. 3) and ATP synthesis. Antimycin A (10 nm) increased the susceptibility of hippocampal neurons to NMDA-induced cell death, confirming results obtained with cyanide, the specific inhibitor of complex IV (Dubinsky and Rothman, 1991). We next examined its effects on elevation of [Ca²⁺]_i in response to 20 min pulses of NMDA. Antimycin A at 1 nmabolishes the early decrease in [Ca²⁺]_i, hereafter defined as Ca²⁺ sequestration, yet recovery to baseline is achieved after the stimulus is removed (Fig.2A). At 100 nm, Ca²⁺ sequestration is abolished and, moreover, [Ca²⁺]_i remains tonically elevated. Antimycin A inhibits Ca²⁺ sequestration (•) and recovery (Δ) in a dose-dependent manner, with IC₅₀ = 470 ± 130 pm and IC₅₀ = 15.9 ± 4.5 nm, respectively (Fig. 2B). IC₅₀ values are 30-fold apart, suggesting that Ca²⁺ sequestration and recovery are distinct mechanisms, both associated with mitochondria. The primary involvement of mitochondria was confirmed using rotenone (Fig. 2C,D); pretreatment produces a similar biphasic profile, abolishing Ca²⁺ sequestration at even lower concentrations (IC₅₀ = 48 ± 19 pm) and blocking recovery at somewhat higher doses (IC₅₀ = 1.5 ± 0.7 μm). These results suggest that although mitochondrial Ca²⁺ uptake is responsible for Ca²⁺ sequestration, ATP-dependent plasma membrane transporters may account for recovery by Ca²⁺ extrusion (Thayer and Miller, 1990; Miller, 1991; White and Reynolds, 1995; Budd and Nicholls, 1996). This conjecture was addressed using oligomycin, which specifically inhibits mitochondrial ATP synthase without altering the redox-generated ΔΨ. In accordance with this prediction, neurons treated with oligomycin before stimulation fail to recover from NMDAR overactivation but show only minor changes in Ca²⁺sequestration (Fig. 2E,F). By contrast, thapsigargin (100 nm ≤ thapsigargin ≤ 500 nm), a specific blocker of endoplasmic reticulum (ER) Ca²⁺ ATPases (Thastrup et al., 1990), does not affect Ca²⁺ sequestration or recovery, indicating that ER mechanisms are not major determinants of this process.

Fig. 3.

Effects of mitochondrial function inhibitors on ΔΨ. Pictures show digital images of neurons exposed to TMRE (Farkas et al., 1989) before and after treatment with 10 μmantimycin A or oligomycin. Collapse of ΔΨ is shown as a decrease in fluorescence intensity after exposure to antimycin A. Pseudocolor scale represents arbitrary fluorescence intensity values ranging from 0 to 255. Scale bar, 20 μm. Traces (right column) indicate time course of changes in ΔΨ measured in arbitrary fluorescence units for neurons treated with antimycin A or oligomycin. Arrows indicate initiation of treatments. Each trace represents mitochondrial fluorescence from an individual neuron normalized to its initial baseline value (F₀).

Fig. 2.

Inhibitors of mitochondrial function modulate homeostasis of [Ca²⁺]_i elicited by a 20 min NMDA stimulus. A, Time course of change in [Ca²⁺]_i for neurons pretreated with antimycin A (anti) for 15 min immediately preceding stimulation with 200 μm NMDA. At 1 nm, antimycin selectively inhibits Ca²⁺ sequestration; at 100 nm, it abrogates both Ca²⁺ sequestration and recovery. Calibration of [Ca²⁺]_i on neurons pretreated with 1 μm antimycin A yielded the following values: [Ca²⁺]_baseline = 23 ± 2 nm; [Ca²⁺]_peak = 1.33 ± 0.12 μm; [Ca²⁺]_plateau = 2.7 ± 0.2 μm; n (number of cells) > 36. This indicates that lack of Ca²⁺ sequestration is not because of saturation of the dye. B, Dose–response for Ca²⁺ sequestration (•) and recovery (▵). Ca²⁺ sequestration is quantified as the area above the Ca²⁺ signal (hatched area inA, top panel) normalized to the rectangular area delimited by the baseline and the peak of the waveform during 20 min NMDA application. Recovery is defined as the fraction of cells that exhibitF_end/F_plateau ≤ 0.3. C, Neurons pretreated for 15 min with rotenone (rote). Similar to antimycin, 10 nm rotenone selectively reduces Ca²⁺ sequestration, whereas 10 μm abolishes both sequestration and recovery.D, Dose–response for Ca²⁺ sequestration (•) and recovery (▵). E, Neurons pretreated for 15 min with oligomycin (oligo). At ≥10 nm, oligomycin inhibits recovery but has a minor effect on Ca²⁺sequestration. F, Dose–response for Ca²⁺sequestration (•; IC₅₀ ≫ 10 μm) and recovery (▵; IC₅₀ = 11.1 ± 2.8 nm). Curves represent best fits to the function y = A₀ + A₁ · (IC₅₀/([drug] + IC₅₀))^h.

Effects of mitochondrial function inhibitors on ΔΨ

To examine the effects of mitochondrial function inhibitors on ΔΨ, the fluorescent potentiometric dye TMRE, known to partition specifically into polarized mitochondria, was used for imaging (Farkas et al., 1989). Neurons treated with antimycin A (Fig. 3,top panel) display a prominent decrease in fluorescence, reflecting mitochondrial depolarization (Farkas et al., 1989; Loew et al., 1994). As shown in the trace, a transient increase in relative fluorescence precedes the decay (see also Figs.4B, 5B); this may arise from unquenching of the dye before its loss from mitochondria (Budd and Nicholls, 1996). The extent of depolarization (D) was quantified as the ratio of the mitochondrial fluorescence 20 min after antimycin A application to fluorescence before the treatment (D = 0.632 ± 0.043; n = 23 neurons). Oligomycin does not alter ΔΨ (D = 1.008 ± 0.028; n = 19 neurons; bottom panel), however, subsequent application of antimycin A does elicit depolarization of ΔΨ. These results strongly suggest that differences between the effects of blockers of the electron transport chain (antimycin A and rotenone) and the ATP synthase (oligomycin) on Ca²⁺ sequestration (Fig. 2) are attributable to their distinct actions on ΔΨ.

Fig. 5.

Blockade of the PTP enhances recovery of the NMDA-induced collapse of ΔΨ and decreases neuronal death.A, Digital images of TMRE fluorescence of neurons exposed for 20 min to 200 μm NMDA in the presence of 1.6 μm CsA. Numbers indicate critical time points in the ΔΨ signal shown in B and code panels1 to 6 that are pseudocolored images of TMRE fluorescence. Neurons show a marked depolarization of ΔΨ evoked by the NMDA pulse; repolarization occurs after removal of the stimulus. Scale bar, 20 μm. All other conditions are as in Figure 4.B, Left panel, Time course of ΔΨ from neurons shown in A, measured in arbitrary fluorescence units. Each trace corresponds to one of the neurons in the digital images. Right panel, Comparison of the time courses of ΔΨ in response to 20 min NMDA in the presence or absence of CsA. Traces are averages from all recordings (+CsA:n = 26; −CsA: n = 74). C, Recovery of ΔΨ (Recovery Ratio) was calculated for recordings obtained in the presence of CsA (n = 26 neurons) and compared with those presented in Figure 4C for 20 min NMDA. Peak depolarization is not affected by CsA (Peak Depolarization = 0.351 ± 0.048; p = 0.1). Cell death induced by NMDA was simultaneously assessed in the presence (n = 7126 neurons) and absence (n = 7985 neurons) of CsA for six independent experiments. Fraction of dead cells in cultures treated with vehicle was subtracted (0.054 ± 0.003; n = 6547 neurons). Bars represent mean ± SEM. *, Statistically significant difference between treatments (p < 0.002); **p < 0.0001.

NMDA collapses ΔΨ

The link between NMDAR overstimulation, Ca²⁺sequestration, and impairment of mitochondrial function was substantiated by measurements of ΔΨ during the excitotoxic challenge (Fig. 4). Neurons challenged for 20 min with NMDA display a prominent depolarization (Fig. 4A) that exhibits a time scale coincident with the development of Ca²⁺ overload (Fig. 1A). Only partial recovery is observed, suggesting an irreversible process in contrast to the reversibility of [Ca²⁺]_i signals (Fig.1A). Increasingly longer NMDA stimuli induce more conspicuous depolarizations of ΔΨ, with decreasing extents of recovery (Fig. 4B,C). A 30 sec stimulus yields a shallow and reversible depolarization of ΔΨ (Fig. 4C). However, a prominent depolarization is induced by both 20 and 50 min NMDA stimuli, establishing a linkage between NMDAR overactivation and mitochondrial impairment. Moreover, recovery of ΔΨ is only partial for 20 min stimuli, and negligible for 50 min stimuli, suggesting that permanent mitochondrial damage occurred. Heightened extracellular [Ca²⁺] ([Ca²⁺]_ext) sharply decreases the extent of recovery induced by 20 min NMDA (Fig.4D), strengthening the connection between NMDA-induced Ca²⁺ influx and collapse of ΔΨ.

Blockade of the mitochondrial PTP with cyclosporin A promotes recovery of the NMDA-induced collapse of ΔΨ and prevents neuronal death

Depolarization of mitochondrial inner membrane, together with an increase in matrix [Ca²⁺] may induce opening of the PTP, a mitochondrial channel that allows ionic diffusion and thereby collapses ΔΨ (Petronilli et al., 1993). The immunosuppressant drug CsA is a potent blocker of the PTP (Bernardi et al., 1994; Nicolli et al., 1996) and presumably may prevent the NMDA-induced collapse of ΔΨ. Figure 5 shows that in the presence of 1.6 μm CsA, depolarization of ΔΨ in response to 20 min NMDA is only transient. The rate of decay of ΔΨ appears to be slower in the presence of CsA than in its absence (Fig. 5B,right panel), suggesting that opening of the PTP may accelerate mitochondrial depolarization. Remarkably, the recovery is practically complete (compare with Fig. 4B,C), suggesting that persistent opening of the PTP may be the cause for tonic mitochondrial depolarization. To examine the involvement of the PTP in excitotoxic cell death, the survival of neurons challenged in the presence of CsA or its absence was determined. CsA protects neurons from NMDA-induced death by 64.8 ± 6.3% (Fig. 5C). These findings implicate the PTP in the persistent mitochondrial depolarization in response to NMDAR overstimulation (Fig. 4) and establish a connection between mitochondrial dysfunction and excitotoxic neuronal death.

DISCUSSION

The results presented here show that (1) overstimulation of NMDARs deregulates homeostasis of [Ca²⁺]_i and induces neuronal death depending on pulse duration; (2) control of [Ca²⁺]_i during an excitotoxic treatment relies primarily on mitochondrial function; (3) NMDA-induced Ca²⁺ influx is sufficient to induce early and persistent mitochondrial depolarization; and (4) blockade of the mitochondrial PTP promotes recovery of ΔΨ and decreases the incidence of cell death. The complex interplay between elevation of [Ca²⁺]_i and mitochondrial function provides a basis to formulate a model of excitotoxic neuronal death. In essence, this model proposes that cell survival relies on a delicate balance between mitochondrial function and [Ca²⁺]_ihomeostasis. Massive Ca²⁺ influx induces an imbalance in mitochondrial homeostasis, leading to mitochondrial dysfunction, which in turn triggers neuronal cell death (Fig. 6). We turn now to discuss the experimental foundation of the model.

Fig. 6.

Schematic representation of potential pathways by which mitochondrial dysfunction could act as an effector of excitotoxic neuronal death. NMDAR overstimulation induces excessive Ca²⁺ influx and abnormal elevations of [Ca²⁺]_i. Mitochondrial Ca²⁺uptake, driven by ΔΨ, attenuates ΔΨ. This, in turn, causes a decrease in ATP synthesis and the opening of the PTP, which collapses ΔΨ. Mitochondrial dysfunction elicits a further reduction in intracellular ATP pools, increases free radical generation, and most likely activates other processes that ultimately contribute to neuronal cell death.

Time-lapse video imaging of ΔΨ shows that depolarization is a consequence of NMDAR overstimulation (Fig. 4); whereas mitochondrial depolarization evoked by 20 min NMDA does not revert to baseline levels (Fig. 4B), elevation of [Ca²⁺]_i does (Fig. 1, Table 1). These results have three major implications: (1) sustained mitochondrial depolarization is not attributable to tonically elevated [Ca²⁺]_i; (2) mitochondrial homeostasis is not linked to [Ca²⁺]_i after removal of the excitotoxic stimulus; and (3) persistent mitochondrial depolarization is an expression of mitochondrial dysfunction.

Findings on cell death, [Ca²⁺]_i homeostasis, and ΔΨ are summarized in Table 1. Results from 20 min NMDA treatments reveal an apparent mismatch between the extent of Ca²⁺ overload and the percentage of dead neurons. This suggests that although existing ATP pools may be sufficient to drive Ca²⁺ extrusion systems and recover [Ca²⁺]_i homeostasis (Fig.1A,C), early neuronal damage occurs as revealed by the collapse of ΔΨ (Fig. 4) and the induction of cell death. In contrast, the extent of Ca²⁺ overload observed for 50 min treatments (Fig. 1B) parallels the incidence of neuronal death, suggesting exhausted ATP pools. Ca²⁺overload therefore may be secondary to mitochondrial dysfunction, and not a causal event for cell death (Fig. 6). This conjecture is supported by the results presented in Figure 2, in which inhibition of mitochondrial function leads to Ca²⁺ overload; for instance, [Ca²⁺]_i responses to 50 min NMDA are mimicked by 20 min NMDA in neurons pretreated with oligomycin. Moreover, in agreement with previous findings (Tymianski et al., 1993a), Ca²⁺ overload occurs much earlier than loss of plasma membrane integrity, as indicated by the ability of neurons to retain the Ca²⁺ indicator Fluo-3 in the cytoplasm for more than 4 hr. Thus, Ca²⁺ overload may be an indicator of cell death, but recovery is not an indicator of cell survival. This notion resolves discrepancies in correlations of [Ca²⁺]_i postinsult and neuronal death (de Erausquin et al., 1990; Michaels and Rothman, 1990; Randall and Thayer, 1992; Witt et al., 1994).

Mitochondria are major buffering compartments for elevations of [Ca²⁺]_i in response to brief exposures to glutamate in cortical neurons (White and Reynolds, 1995). Inhibition of mitochondrial function by picomolar concentrations of antimycin A or rotenone abolishes Ca²⁺ sequestration in response to prolonged NMDA treatments (Fig. 2), indicating that mitochondria act as a primary buffer of [Ca²⁺]_i also under excessive insults. A unique function of mitochondria (as a main source of ATP) is highlighted by the finding that oligomycin or higher concentrations of antimycin A or rotenone eliminate the recovery phase, presumably by compromising ATP-dependent Ca²⁺ extrusion systems (Fig. 6). Thus, mitochondria, as intracellular Ca²⁺buffers, set a point of vulnerability whereby if the boundary between mitochondrial function and dysfunction is crossed, cell death ensues. As discussed below, this crossroad is probably determined by the PTP (Fig. 5).

Our population study suggests a correlation between mitochondrial dysfunction and cell death, but does not establish a causal link (Table1). If mitochondrial impairment were a primary event in the induction of excitotoxic cell death, then neuronal death might be prevented by treatments that protect mitochondrial homeostasis. As shown in Figure5, the NMDA-induced collapse of ΔΨ recovers completely in the presence of CsA. Accordingly, neuronal death is reduced by 65% under these conditions compared with neurons challenged in the absence of CsA (Fig. 5C). NMDARs are not a target of CsA, because voltage-clamp recordings show that currents elicited by NMDA remain unchanged in the presence of CsA (data not shown). The fact that complete recovery of ΔΨ is accompanied by a substantial, yet partial, reduction of cell death might indicate that pathways that are parallel and/or downstream to mitochondria are also activated by NMDAR overstimulation (Fig. 6). Immunosuppressant drugs, such as CsA, display numerous intracellular binding sites (Kunz and Hall, 1993) and have been shown to reduce excitotoxic neuronal death in cortical neurons, presumably by inhibition of a pathway that activates nitric oxide synthase (NOS) (Dawson et al., 1993). However, because CsA is a potent blocker of the PTP (Petronilli et al., 1993; Bernardi et al., 1994;Nicolli et al., 1996) and because our results show a direct effect of CsA on mitochondrial ΔΨ (Fig. 5A,B), it appears unlikely that preservation of mitochondrial function and promotion of cell survival are attributable to inhibition of NOS, yet this cannot be ruled out. Our findings suggest that persistent mitochondrial depolarization results from opening of the PTP [presumably as a consequence of combined mitochondrial depolarization and elevation of intramitochondrial Ca²⁺ (Bernardi et al., 1994)] and that maintenance of mitochondrial homeostasis can protect hippocampal neurons from excitotoxic cell death (Fig. 6).

In view of these findings, glutamate excitotoxicity can be visualized as a cascade of events that starts with overstimulation of NMDARs, which elicit massive Ca²⁺ influx and abnormally high [Ca²⁺]_i (Fig. 6). Ca²⁺ is sequestered into the mitochondrial matrix, driven by the proton electrochemical gradient generated by the electron transport chain, depolarizing ΔΨ (Akerman, 1978; Gunter and Pfeiffer, 1990; Loew et al., 1994). This reduction in the electrochemical gradient decreases ATP synthesis at a time of great demand by the plasma membrane Ca²⁺ pump and, indirectly, by the Na⁺/Ca²⁺ exchanger (White and Reynolds, 1995). Concurrent buildup of intramitochondrial [Ca²⁺] and attenuation of ΔΨ induce opening of the mitochondrial inner membrane PTP, thereby allowing ionic diffusion and collapsing ΔΨ (Beatrice et al., 1980; Gunter and Pfeiffer, 1990; Bernardi et al., 1994; Hoek et al., 1995). The drop of cellular ATP levels, in concert with the generation of free O radicals consequent to high ATP consumption (Halliwell and Gutteridge, 1989; Coyle and Puttfarcken, 1993; Reynolds and Hastings, 1995), is a primary cause of cell death (Siesjö, 1992). The boundary between cell survival and cell death relies on a subtle balance of mitochondrial function/dysfunction, centered on ΔΨ. This model predicts that neuronal lifespan may be prolonged by specific inhibition of mitochondrial Ca²⁺uptake and that susceptibility to excitotoxic death may be influenced by the energetic state of neurons before and during the insult.

Our findings establish a direct coupling between excitotoxicity and shifts of redox potentials and membrane potentials of mitochondria inside living neurons. This evidence strengthens the involvement of mitochondria in the neurotoxicity associated with mitochondrial complex I inhibitor 1-methyl-4-phenylpyridinium (Saporito et al., 1992), amyotrophic lateral sclerosis (Brown, 1995; Wong et al., 1995), apoptosis (Newmeyer et al., 1994; Vayssière et al., 1994; Zamzami et al., 1995), and age-related disorders (Luft, 1994; Brouillet et al., 1995). These findings may provide useful guidelines for the development of the next generation of drugs targeted against neurodegeneration and stroke.