Mitochondria and neuronal glutamate excitotoxicity

Biochimica et Biophysica Acta 1366 (1998) 97^112

Mitochondria and neuronal glutamate excitotoxicity David G. Nicholls *, Samantha L. Budd

1

Neurosciences Institute, Department of Pharmacology and Neuroscience, University of Dundee, Dundee DD1 9SY, UK Received 5 January 1998; accepted 17 February 1998

Abstract The role of mitochondria in the control of glutamate excitotoxicity is investigated. The response of cultured cerebellar granule cells to continuous glutamate exposure is characterised by a transient elevation in cytoplasmic free calcium concentration followed by decay to a plateau as NMDA receptors partially inactivate. After a variable latent period, a secondary, irreversible increase in calcium occurs (delayed calcium deregulation, DCD) which precedes and predicts subsequent cell death. DCD is not controlled by mitochondrial ATP synthesis since it is unchanged in the presence of the ATP synthase inhibitor oligomycin in cells with active glycolysis. However, mitochondrial depolarisation (and hence inhibition of mitochondrial calcium accumulation) without parallel ATP depletion (oligomycin plus either rotenone or antimycin A) strongly protects the cells against DCD. Glutamate exposure is associated with an increase in the generation of superoxide anion by the cells, but superoxide generation in the absence of mitochondrial calcium accumulation is not neurotoxic. While it is concluded that mitochondrial calcium accumulation plays a critical role in the induction of DCD we can find no evidence for the involvement of the mitochondrial permeability transition. ß 1998 Elsevier Science B.V. All rights reserved. Keywords: Mitochondrion; Glutamate; NMDA; Excitotoxicity; Calcium; Neuron

1. Introduction Even in neurones, where the ATP requirement for ionic homeostasis is particularly high, the safety margin between cellular ATP supply capacity and demand is considerable as long as glucose and oxygen are available in excess. However even a brief interruption in ATP generation, as in transient global ischaemia, can initiate neurodegeneration, which depending on the severity of the insult can show ne* Corresponding author. Fax: +44 (1382) 667120; E-mail: [email protected] 1 Present address: CNS Research Institute, LMRC First Floor, 221 Longwood Avenue, Brigham and Women's Hospital, Boston, MA 02115, USA.

crotic or apoptotic characteristics and can occur with a delay of a few minutes to 1^2 days. Furthermore, a less severe, but chronic, restriction in neuronal ATP generation capacity may underlie a range of neurodegenerative disorders, including Alzheimer's, Huntington's and Parkinson's diseases, amyotrophic lateral sclerosis and AIDs-related dementia as well as encephalopathies associated with mitochondrial mutations (reviewed in [1]). The excitatory neurotransmitter glutamate plays a central role in neuronal cell death associated with these neurodegenerative disorders, and the interaction between the NMDA-selective glutamate receptor and the mitochondrion forms the basis of this review, which will focus on necrotic cell death in neuronal culture. The sequence of events culminating in excitotoxic

0005-2728 / 98 / $19.00 ß 1998 Elsevier Science B.V. All rights reserved. PII: S 0 0 0 5 - 2 7 2 8 ( 9 8 ) 0 0 1 2 3 - 6

BBABIO 44667 30-7-98

98

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

cell death is initiated by the entry of Ca2‡ through predominantly NMDA-selective glutamate receptors activated by elevated extracellular glutamate. Glutamate is not excitotoxic to anoxic cells as long as NMDA receptors are inactivated prior to the restoration of oxygen [2]. The two conditions of greatest risk are ¢rstly the period when circulation is restored following transient global ischaemia and before the highly active glutamate transporters can re-accumulate the amino acid into neurones and glia, and secondly in the penumbra surrounding a focal ischaemia where respiring neurones are exposed to high glutamate concentrations di¡using from the ischaemic core. This latter region is of particular signi¢cance, since it demonstrates that glutamate exposure can destroy the bioenergetic integrity of neurones continuously supplied with glucose and oxygen. It is convenient to divide the excitotoxic process into two stages: the initial exposure to glutamate; and the subsequent `latent' period, where the continued presence of glutamate is not obligatory and which culminates in the failure of cytoplasmic Ca2‡ homeostasis, termed by Tymianski `delayed Ca2‡ deregulation' [3] and subsequent cell death (Fig. 1). Our

analysis will focus on neuronal bioenergetics and will attempt to deconvolute the interactions between mitochondrial ATP synthesis, Ca2‡ accumulation and the generation of reactive oxygen species in cultured neurones exposed to excitotoxic concentrations of glutamate. 2. The initial glutamate exposure In primary neuronal culture, the excitotoxic cascade can be initiated by as little as 5 min exposure to high glutamate concentrations in the absence of Mg2‡ (to prevent voltage-dependent block of the NMDA receptor). There is overwhelming evidence that the entry of Ca2‡ into the cells predominantly through NMDA-selective glutamate receptors during this period triggers the subsequent neurodegeneration; however Ca2‡ loading via voltage-activated Ca2‡ channels during KCl-depolarisation is much less excitotoxic and the reason for this selectivity is the subject of current controversy. The possibilities are that the absolute amount of Ca2‡ entering through the NMDA receptor may greatly exceed that through voltage-activated Ca2‡ channels [4,5], or that Ca2‡ entering through the NMDA receptor is focussed onto a vulnerable excitotoxic locus within the cell [3,6]. When determined with high-a¤nity Ca2‡ indicators, such as fura-2, little di¡erence is seen in the apparent elevation in [Ca2‡ ]c evoked with glutamate or KCl [7^9]; however the use of low a¤nity indicators of free Ca2‡ capable of reporting far higher concentrations before saturating has indicated that glutamate-evoked [Ca2‡ ]c elevations may be considerably underestimated and may exceed 5 WM, considerably in excess of levels achieved with KCl-depolarisation [7,10]. 2.1. Neuronal mitochondria sequester cytoplasmic [Ca 2+]c transients

Fig. 1. Phases of acute glutamate excitotoxicity. Neurones exposed to glutamate show a transient elevation in cytoplasmic free Ca2‡ , [Ca2‡ ]c . Following this initial response, the signal decays to a plateau as NMDA receptors partially inactivate. The plateau is maintained for a variable time (`latent period') until a secondary, irreversible, increase in [Ca2‡ ]c occurs (delayed Ca2‡ deregulation).

At the termination of a transient glutamate exposure, [Ca2‡ ]c tends to return to baseline as the cation is sequestered into internal organelles or is extruded across the plasma membrane. Our early studies with isolated mitochondria from liver and brain demonstrated that the activity of the mitochondrial Ca2‡

BBABIO 44667 30-7-98

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

uniporter was highly dependent on the extramitochondrial Ca2‡ concentration, [Ca2‡ ]o [11] and exceeded the activity of the independent mitochondrial Ca2‡ e¥ux pathway (coupled to H‡ and Na‡ in liver and brain, respectively [12]) when [Ca2‡ ]o rose above the `set-point' at which uptake and e¥ux balanced [13]. Mitochondria would therefore be predicted automatically to bu¡er [Ca2‡ ]c above the set-point and to release it back into the cytoplasm when [Ca2‡ ]c recovered to below this value (reviewed in [12]). With a predicted set-point of 0.3^0.5 WM [13] even the Ca2‡ transients measured with high-a¤nity indicators during KCl stimulation rise well above these values and should therefore be in£uenced by mitochondrial Ca2‡ sequestration. In 1981, we obtained direct evidence of mitochondrial sequestration of depolarisation-evoked Ca2‡ loads by measuring mitochondrial 45 Ca2‡ pools within synaptosomes: Ca2‡ loading of these intact isolated terminals by KCl-activation of voltage-activated Ca2‡ channels resulted in the large majority of the accumulated Ca2‡ being further transported into the matrix [14,15]. In 1990, Thayer and Miller [16] showed that brief KCl-mediated depolarisation of dorsal root ganglion cells was followed by a partial recovery in [Ca2‡ ]c to a plateau which varied from 200 to 600 nM, but was absent in cells depolarised in the presence of protonophore. This plateau was interpreted to be a consequence of the slow release from the mitochondrion of Ca2‡ temporarily accumulated at the peak [Ca2‡ ]c . Addition of the protonophore during the plateau produced a large transient elevation in [Ca2‡ ]c consistent with a rapid release of the mitochondrial pool. A similar plateau was observed in these cells following trains of s 25 action potentials [17]; again the level of the [Ca2‡ ]c plateau during recovery (450^550 nM) correlated well with the set-point predicted for isolated brain mitochondria [13]. A comparable protonophore enhancement of the KCl-evoked [Ca2‡ ]c transient and abolition of the subsequent plateau has also been reported for bullfrog sympathetic neurones [18]. These results were consistent with earlier modelling with isolated mitochondria [19] and have recently been reproduced in whole-cell patch-clamped chroma¤n cells following depolarisation-evoked activation of voltage-activated Ca2‡ channels [20]. In 1990, the group of Duchen [21] also reported enhanced cytoplasmic Ca2‡ transients in dorsal root

99

ganglion cells exposed to a range of metabolic inhibitors, including protonophores, cyanide and glucose removal, but noted that these e¡ects could be a consequence of impaired Ca2‡ extrusion from the cells as well as inhibited mitochondrial sequestration. In a detailed series of papers, White and Reynolds [9,22,23] have analysed the pathways of Ca2‡ removal from the cytoplasm following acute, non-toxic, glutamate exposure. Restoration of baseline [Ca2‡ ]c following as little as 15 s exposure to 3 WM glutamate was highly dependent upon both vim and extracellular Na‡ ; thus no recovery occurred during washout of glutamate by a Na‡ - free medium containing protonophore. Interestingly, when the same Ca2‡ transient was generated by the combination of high KCl and veratridine (activating voltage-activated Ca2‡ channels and allowing Na‡ entry via voltageactivated Na‡ channels), recovery of [Ca2‡ ]c to baseline could still occur. This was interpreted as indicating a fundamental di¡erence in the handling of glutamate-evoked versus KCl/veratridine-evoked Ca2‡ loads. While the reason for this di¡erence was not directly investigated, a likely explanation was considered to be a greater uptake of Ca2‡ into the cell (and hence the mitochondrion) in the former condition. 2.2. In situ mitochondrial membrane potential (vim ) during acute Ca 2+ loading of neurones The key parameter determining the energetic status of mitochondria is the proton-motive force, vp, across the inner membrane (reviewed in [24]). In the presence of physiological concentrations of Pi and Mg2‡ , isolated mitochondria from liver, heart or brain maintain a total vp of some 220 mV of which the membrane potential vim comprises 150^ 180 mV with a vpH of 30.5 to 31 pH unit contributing the remaining 30^60 mV [25]. The determination of both components of vp for in situ mitochondria is exceedingly complex; in collaboration with Hoek and Williamson [26], we quanti¢ed vp for in situ hepatocyte mitochondria: vim was determined from the Nernstian distribution between the cytoplasm and matrix of the lipophilic cation TPMP‡ , while vpH was measured by weak acid distribution; a value for vp of s 200 mV was consistent with

BBABIO 44667 30-7-98

100

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

isolated mitochondria. Subsequently, we quanti¢ed the vim component for mitochondria within isolated nerve terminals arriving at an estimate of 150 mV for glucose-maintained synaptosomes [27]. Direct determination of the Nernstian gradient between matrix and cytoplasm by confocal microscopy is extremely di¤cult due to the small size of the mitochondria and the enormous dynamic range required to quantify a 300-fold gradient. Instead, determinations at the population or single-cell level of resolution membrane-permeant cationic £uorescent indicators report their concentration within the mitochondrial matrix either by their £uorescent quenching (e.g. rhodamine-123 [28^34] or tetramethylrhodamine esters [35^37]) or by a change in their emission spectra (e.g. JC-1 [9]). An essential control, which is not always performed, is to establish that the signal is not a¡ected by a change in plasma membrane potential, for example high [K‡ ] or glutamate in the absence of external Ca2‡ , see [9,28]. With this important proviso in mind, £uorescent changes indicative of mitochondrial depolarisation have been reported following acute glutamate exposure of cultured neurones [9,28,30,31,33,34,37], although particularly with JC-1 the pattern of response can be complex, with both hyperpolarizing and depolarizing responses being detected [9]. The ATP synthase inhibitor oligomycin did not prevent the KClevoked vim decrease [28]. The in£uence of mitochondrial Ca2‡ loading on vim is complex and has been analysed in detail only for isolated mitochondria: in the presence of excess Pi a single addition of Ca2‡ causes a transient depression in both the vim and vpH components of vp, the extent of which is dependent on the Ca2‡ concentration [38^40]. When net accumulation of Ca2‡ is complete both components of vp are restored to their initial values. The transient depression in vp can be su¤cient to interrupt ATP synthesis or even to causes reversal of the ATP synthase and hydrolysis of ATP ^ consistent with early observations that Ca2‡ accumulation could take precedence over ATP synthesis [41]. The signi¢cance of phosphate is frequently overlooked: however, the anion is co-accumulated into the matrix in parallel with Ca2‡ where it forms an osmotically inactive, but rapidly dissociable, calcium phosphate complex [12]. In the absence of added Pi , and particularly following depletion of

endogenous Pi by an ADP/glucose/hexokinase trap [11] the capacity of isolated mitochondria to accumulate Ca2‡ is greatly restricted. Thus, while the same reversible drop in vp occurs during Ca2‡ accumulation, the vpH generated by protons extruded by the respiratory chain is no longer neutralised by Pi accumulation and a progressive build-up of vpH occurs. This can ultimately result in an equipartition between the two components: a vp of 220 mV being comprised of 110 mV of vim and almost 2 pH units of transmembrane gradient [42,43]. Finally, induction of the mitochondrial permeability transition (MPT, see below) will lead to a collapse in both vim and vp. It is not yet clear which of the above mechanisms underlies any decrease in vim in glutamate-exposed neurones. 2.3. Deconvolution of vim e¡ects on mitochondrial Ca 2+ accumulation and ATP synthesis In intact cells with maintained in situ metabolism, Ca2‡ homeostasis cannot be separated from cellular metabolism and bioenergetics and should be considered as part of an integrated picture (Fig. 2). Neurones are almost entirely dependent on exogenous glucose, and the consequent supply of pyruvate to the mitochondrion and its subsequent oxidation via the tricarboxylic acid cycle drives both ATP synthesis and Ca2‡ accumulation. Since the two processes compete for the proton circuit they should be considered together in an analysis of the e¡ects of cellular Ca2‡ loading. While Ca2‡ accumulation by the mitochondria may a¡ect ATP synthesis, alterations in ATP synthesis will in turn a¡ect the activity of ion pumps responsible for removing Ca2‡ from the cytoplasm. An analysis of these interactions is one of the main goals of our laboratory [31,41]. While protonophore addition has been widely employed to investigate mitochondrial Ca2‡ pools in neurones (see above), the collapse in vim will not only inhibit mitochondrial Ca2‡ accumulation and release the cation into the cytoplasm, but should also instantaneously reverse the mitochondrial ATP synthase resulting in rapid hydrolysis of cytoplasmic ATP [41]. Unless the maximal rate of glycolytic ATP synthesis comfortably exceeds the cellular ATP demand during Ca2‡ loading plus the rate of ATP hydrolysis by the uncoupled mitochondria, the cells will

BBABIO 44667 30-7-98

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

become depleted of ATP and ATP-dependent processes will be compromised. We have employed rat cerebellar granule cells as our experimental model; these cells have a high density of mitochondria both along their neurites and packed around the nucleus in the relatively thin annulus of cytoplasm [31]. We initially determined whether the granule cells showed the same response to protonophores during KCl-depolarisation as the dorsal root ganglion cells investigated by Thayer and Miller [16]. Cerebellar granule cells maintain a [Ca2‡ ]c of 6 100 nM, well below the estimated setpoint for rat brain mitochondria (0.3^0.5 WM [13]) and would thus be predicted to be largely depleted of Ca2‡ . This was con¢rmed by the failure of protonophore addition to initiate a transient release of Ca2‡ into the cytoplasm. In contrast, addition of protonophore during the plateau of [Ca2‡ ]c following KCldepolarisation resulted in a transient spike in the fura-2 signal, consistent with release from the mitochondrial compartment. When protonophore was added prior to 50 mM KCl, the size of the depolarisation-evoked [Ca2‡ ]c spike was greatly enhanced, consistent with earlier studies [18]. While this could be interpreted as a direct consequence of a failure of mitochondrial sequestration, a rapid secondary elevation in [Ca2‡ ]c led us to suspect that the protonophore was causing a collapse in ATP levels, preventing Ca2‡ extrusion from the cells. This depletion was con¢rmed by analysis of ATP/ADP ratios within the cells: within 5 min, 2 WM CCCP decreased the ratio from a control value of s 7 to 6 3 [41]. There is thus ambiguity as to whether the enhanced cytoplasmic [Ca2‡ ]c response is due a failure of mitochondrial Ca2‡ accumulation due to the collapsed vim , or a failure of Ca2‡ extrusion from the cell in response to a lowered ATP/ADP ratio. Seven-day in vitro granule cells incubated in the presence of 15 mM glucose do not have a su¤ciently active glycolytic pathway to be able to maintain a high ATP/ADP ratio and low basal cytoplasmic Ca2‡ in the presence of protonophore, when glycolysis is required to maintain the cell's normal functions in the face of the freely reversing mitochondrial ATP synthase. However, when the ATP synthase is inhibited by oligomycin, preventing both synthesis and hydrolysis of ATP by the mitochondria, glycolysis can maintain high ATP/ADP ratios and pro-

101

longed cytoplasmic Ca2‡ homeostasis [31,41]. As in other cells [27,44], the addition of oligomycin causes a slight hyperpolarisation of the in situ mitochondria [45] consistent with a `state 3^state 4' transition [24]. Since the entire ATP generation by the cell is glycolytic in the presence of oligomycin experiments can be designed in which vim can be collapsed, for example by the further addition of a mitochondrial respiratory chain inhibitor, such as rotenone [31,41], without a priori a¡ecting cellular ATP generation (Fig. 2). If accumulation of Ca2‡ into endoplasmic reticulum is ignored, the change in free Ca2‡ concentration reported by a £uorescent probe is the net resultant of cytoplasmic Ca2‡ chelation, accumulation and release from mitochondria and uptake and e¥ux across the plasma membrane, which in neurones is due to both the plasma membrane Ca2‡ -ATPase and Na‡ /Ca2‡ exchanger [21,46,47]. In view of the evidence discussed above that mitochondria sequester much of the Ca2‡ load imposed by KCl depolarisation, it would be predicted that abolition of the mitochondrial pool would enhance the cytoplasmic Ca2‡ transient. This is not, however, what we observe: KCl-depolarisation of granule cells in the presence of rotenone plus oligomycin actually produces a smaller peak [Ca2‡ ]c elevation than in the presence of oligomycin alone [41]. Thus, mitochondrial depolarisation by protonophores enhances the KCl-evoked [Ca2‡ ]c transient, while mitochondrial depolarisation by rotenone plus oligomycin decreases the transient. The trivial explanation that rotenone might directly inhibit voltage-activated Ca2‡ channels was eliminated since when both rotenone and protonophore were present before KCl-depolarisation, the peak height was enhanced to the same extent as with FCCP alone [41]. There are two explanations for this counter-intuitive result: mitochondrial depolarisation must either reduce Ca2‡ entry via the VACCs responsible for the initial transient, or Ca2‡ extrusion from the cell must be enhanced. Application of the same protocol to granule cells exposed to glutamate showed that while oligomycin did not signi¢cantly a¡ect the peak and initial plateau following maximal stimulation of the NMDA receptor, the combination of rotenone plus oligomycin to depolarise the mitochondria prior to receptor activation decreased both the initial [Ca2‡ ]c transient

BBABIO 44667 30-7-98

102

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

and the subsequent plateau [31]. In further experiments (Budd and Nicholls, unpublished), we have con¢rmed that metabolic restriction by the complex III inhibitor antimycin A, protonophore addition and inhibition of succinate dehydrogenase by malonate each enhanced Ca2‡ deregulation following NMDA receptor activation, while the same inhibitors delayed deregulation in the presence of oligomycin. Thus, as in the case of KCl depolarisation, inhibitors which will decrease ATP levels enhance

glutamate-evoked cytoplasmic Ca2‡ transients, while removal of the mitochondrial matrix as a Ca2‡ sink without depleting ATP result in a decrease in [Ca2‡ ]c . Furthermore, the total 45 Ca2‡ accumulated during exposure to glutamate was also decreased by rotenone plus oligomycin [31]; thus mitochondrial depolarisation must restrict Ca2‡ entry into the cell and/or enhance e¥ux of the cation. The mechanism of this control is currently being investigated. One hypothesis is that mitochondria control the

BBABIO 44667 30-7-98

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

103

Fig. 2. Schematic inter-relationships between the NMDA receptor, mitochondrial Ca2‡ transport and cellular bioenergetics. Mitochondria in granule cells utilize pyruvate to generate vp, synthesize ATP and accumulate Ca2‡ . Ca2‡ extrusion pathways driven by ATP (e.g. the Ca2‡ -ATPase) maintain a low [Ca2‡ ]c . In each ¢gure, the predicted activity of pathways is represented by the line thickness. In each scheme, the initial response to NMDA receptor activation is depicted. (A) In the absence of inhibitors, ATP synthesis is predominantly via the ATP synthase. Ca2‡ entering the cell during NMDA receptor activation is largely accumulated into the mitochondria. Chronic receptor activation can lead to mitochondrial Ca2‡ overload. (B) In the presence of oligomycin, glycolysis supports the total cellular ATP demand. Both ATP synthesis and hydrolysis by the mitochondria are inhibited. The mitochondria still load with Ca2‡ and despite the presence of the inhibitor cellular ATP falls and Ca2‡ homeostasis is lost. Thus Ca2‡ -loading of the mitochondrion is excitotoxic by a process independent of the oligomycin-sensitive ATP synthase. (C) The complex I inhibitor rotenone prevents the respiratory chain from pumping protons. Glycolysis now supplies ATP both for the cell and for the ATP synthase which reverses to generate a reduced proton gradient. Even in the absence of glutamate, the cell can only maintain a low [Ca2‡ ]c for a limited period. When glutamate is added the reversal of the ATP synthase accelerates as it attempts to compensate for the inward £ux of Ca2‡ into the matrix and the resulting ATP depletion leads to an immediate loss of cytoplasmic Ca2‡ homeostasis. (D) The combination of rotenone plus oligomycin allows the proton gradient to decay preventing mitochondrial Ca2‡ loading. Cytoplasmic ATP is maintained by glycolysis; the decreased 45 Ca2‡ uptake into the cells, together with the maintained low cytoplasmic Ca2‡ , each demonstrate that Ca2‡ in£ux through the NMDA receptor is balanced by e¥ux across the plasma membrane when mitochondria can no longer sequester Ca2‡ , due either to a feed-back inhibition of the NMDA receptor or an enhanced Ca2‡ e¥ux. (E) The protonophore FCCP collapses cellular ATP by reversal of the ATP synthase.

6

net Ca2‡ £ux across the plasma membrane by rapid changes in ATP availability. ATP depletion of granule cells by simple protonophore addition enhances the KCl-evoked peak Ca2‡ elevation and prevents the maintenance of a subsequent plateau [21,41]. Glucose depletion greatly enhances the [Ca2‡ ]c transient in dorsal root ganglion cells depolarised with KCl [21]. Lowered glucose availability, both in the presence and absence of oligomycin, enhances both the peak and subsequent plateau fura-2 signal of granule cells during glutamate exposure (Budd, Ward and Nicholls, unpublished). However, since the rationale of our experiments in the presence of oligomycin has been to eliminate e¡ects of mitochondrial ATP synthesis, how could mitochondria with inhibited ATP synthase activity still in£uence cytoplasmic ATP/ADP ratios? The experimental observation is that glutamate causes an equivalent decrease in granule cell ATP/ADP ratios in the presence and absence of oligomycin [31]. One factor which is frequently overlooked is the availability of phosphate (Pi ), which is co-accumulated into the mitochondrial matrix together with Ca2‡ [12] and the rapid accumulation of the cation into the mitochondrion following NMDA receptor activation could temporarily deplete the cytoplasm of Pi , which would in turn decrease the ATP/ADP ratio which could be maintained by glycolysis. Thus it has been observed that low extracellular Pi decreases ATP levels in cultured neurones and sensitises cells to glutamate-evoked death [48].

3. Delayed Ca2+ deregulation The initial peak in [Ca2‡ ]c following glutamate addition and the subsequent recovery to a plateau largely re£ects the activation and subsequent partial desensitisation of the NMDA receptors. Once the plateau is established, the cytoplasmic free Ca2‡ of glutamate-exposed cells can remain relatively low and stable [49] until a sudden, essentially irreversible, increase occurs (Fig. 3A). Individual cells survive for varying periods before this ¢nal Ca2‡ deregulation and survival ¢ts a single exponential curve, consistent with a stochastic process [50]. Continued NMDA receptor activation is not required during this period, since transient exposure of neurones to glutamate for as short a period as 5 min can lead to subsequent cell death, although prolongation of the glutamate exposure does increase toxicity [3]. Plasma membrane integrity is initially maintained, since £uorescent probes are still retained in the cytoplasm. While the ATP/ADP ratio falls in populations of deregulating cells [31,51] it has not been possible to determine the ratio in an individual cell during the process. Thus it is not easy to determine whether there is an upstream failure of ATP synthesis or a downstream inhibition of ATP-dependent ion pumps; for example, the plasma membrane Ca2‡ -ATPase or Na‡ /K‡ -ATPase, due to proteolytic or oxidative damage. Deregulation of [Ca2‡ ]c in our preparation appears to be a result of failed e¥ux rather than enhanced in£ux, since the rise in [Ca2‡ ]c is not blocked by a

BBABIO 44667 30-7-98

104

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

Fig. 3. Delayed Ca2‡ deregulation following the addition of glutamate/glycine to rat cerebellar granule cells. (A) Addition of 100 WM glutamate plus 10 WM glycine to cells results in a stochastic Ca2‡ deregulation. (B) In the presence of antimycin A to inhibit the respiratory chain, glutamate/glycine addition leads to an immediate deregulation which can be accounted for by insu¤cient ATP synthesis (see Fig. 1C). (C) In the presence of oligomycin plus antimycin A to depolarize the mitochondria without collapsing ATP, glutamate/ glycine-evoked Ca2‡ transients are decreased and no Ca2‡ deregulation occurs within the timespan of the experiment. Each trace represents [Ca2‡ ]c in an individual cell soma determined with fura-2. For further details see [31].

cocktail of channel inhibitors (Budd and Nicholls, unpublished), while the rate of Mn2‡ quenching of fura-2 £uorescence, indicative of Ca2‡ entry, does not increase as cells deregulate [52]. There may however be a contribution to deregulation from mitochondrial dumping of Ca2‡ into the cytoplasm as a result of mitochondrial depolarisation and/or the permeability transition. 3.1. The latent period prior to deregulation The processes which occur during this intermediate `latent' period culminate in cell death which may have apoptotic [53,54] necrotic [55^57] or mixed [51,58,59] characteristics depending on the severity of the initial insult. Thus, within a culture of cerebellar granule cells transiently exposed to glutamate, a subpopulation were found to depolarise their mitochondria and undergo necrosis, while the surviving cells subsequently repolarised their mitochondria and regenerated ATP, but underwent later apoptosis [51]. Apoptosis is out of the scope of this review, however, while activation of the caspase cascade is a common ¢nal pathway for a range of factors inducing granule cell apoptosis [60^62], caspase inhibitors are unable to prevent necrosis [63^66]. 3.2. Restricted bioenergetics facilitates delayed Ca 2+ deregulation NMDA receptor activation places a multiple en-

ergy demand upon a neurone. The increased cytoplasmic Ca2‡ and Na‡ concentrations will result in activation of the plasma membrane Ca2‡ -ATPase and Na‡ /K‡ -ATPase, respectively. The latter may be quantitatively more important, since even though NMDA receptors display some selectivity for Ca2‡ over Na‡ , the much higher concentration of the latter ion in the extracellular medium means that the in£ux of Na‡ may exceed that of Ca2‡ [67]. However, glutamate is still excitotoxic in Na‡ -free media [3]. In addition to these plasma membrane e¡ects, the peak values of [Ca2‡ ]c in excess of 5 WM during glutamate exposure reported in recent experiments with low a¤nity Ca2‡ indicators [7,10] would be predicted from isolated mitochondrial studies [40] to result in such a rapid uptake into the mitochondrial matrix that vp could be transiently lowered below the value required for ATP synthesis. Depletion of cytoplasmic ATP will lead to a catastrophic deregulation of [Ca2‡ ]c by inhibiting the plasma membrane Ca2‡ -ATPase and also limiting the activity of the two ATP-requiring enzymes in the glycolytic pathway, hexokinase and phosphofructokinase [68]. In our studies, cultured granule cells treated with FCCP, rotenone or antimycin A (Fig. 3B) show massive Ca2‡ deregulation upon exposure to glutamate [31]. In this condition, glycolysis is required not only to generate ATP for the activated plasma membrane ion pumps, but also to supply the reversed mitochondrial ATP synthase, which attempts to regenerate vp which is lowered by the up-

BBABIO 44667 30-7-98

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

take of Ca2‡ into the matrix. It is evident that these conditions create an instant ATP de¢cit and immediate failure of cytoplasmic Ca2‡ homeostasis. There is an extensive literature demonstrating the synergistic e¡ect of metabolic restriction upon glutamate excitotoxicity both in vivo, e.g. [69^72], and in vitro [37,59,73^75]. This synergism can also be observed in in vivo models of neurodegenerative diseases: thus the behavioural and morphological effects of Huntington's disease following injection of or 3-nitropropionate into the striatum can be diminished by NMDA receptor antagonists [76^78] as can the motor e¡ects of 1-methyl-4-phenylpyridinium (MPP‡ ) inhibition of mitochondrial complex I in animal models of Parkinson's disease [71]. 3.3. Delayed Ca 2+ deregulation is not causally dependent upon a failure of oxidative phosphorylation While Ca2‡ -deregulation of glutamate-exposed cultured cerebellar granule cells is an imperfect model for glutamate excitotoxicity, it does allow some simple hypotheses to be tested. The ¢rst is that a failure of mitochondrial ATP synthesis is necessary and su¤cient for deregulation. This might be expected in view of the synergistic e¡ects of energy restriction and NMDA receptor activation discussed above, as well as the possibility that mitochondria might be damaged by excessive Ca2‡ accumulation (discussed below). Since granule cells have su¤ciently active glycolysis to maintain Ca2‡ homeostasis when mitochondrial oxidative phosphorylation is inhibited from the outset in the presence of oligomycin [41], this hypothesis would predict that glutamate would not be excitotoxic in the presence of the inhibitor. However, no statistically signi¢cant di¡erence is observed in the time or extent of Ca2‡ deregulation in granule cells exposed to glutamate in the presence or absence of the inhibitor [31]. In other words, delayed Ca2‡ deregulation occurs in cells whose mitochondria are polarised, transporting electrons and accumulating Ca2‡ , but are not required to synthesise (or hydrolyze) ATP. A decline in ATP/ ADP ratio is seen in granule cell populations exposed to glutamate for 60 min in the presence of oligomycin [31], but it is currently unclear whether this is the cause or e¡ect of the Ca2‡ deregulation in individual

cells. A further implication of glycolysis can supply su¤cient cell's energy requirements even hanced during NMDA receptor

105

this ¢nding is that ATP to supply the when these are enactivation.

3.4. Delayed Ca 2+ deregulation is not causally dependent upon a failure of glycolysis The requirement of glycolysis for ATP at two steps (hexokinase and phosphofructokinase) implies that a cell whose ATP levels are declining, even transiently, could su¡er an irreversible collapse as glycolysis becomes limiting, exacerbating a further decrease in ATP. In 1986, we found that synaptosomes showed an initial enhancement of glycolysis, consistent with a Pasteur e¡ect, following respiratory inhibition by rotenone, but that this was followed by a progressive failure [79]. This glycolytic failure has been subsequently analysed in more detail [68] under conditions where the bioenergetic safety margin was eroded by increased energy demand (e.g. ionophore addition) and inhibited ATP generation (including rotenone addition). It was concluded that the ATP requirement for hexokinase was the limiting factor during glycolytic failure [68]. Interestingly, this raises the question as to how the neurone restarts glycolysis on recovery from severe ATP depletion. In this context, it has been demonstrated that lactate, which accumulates during anoxia, can be utilised by severely ATP-depleted cells, generating pyruvate and allowing mitochondrial oxidative phosphorylation to regenerate ATP allowing glycolysis to occur [80]. If delayed Ca2‡ deregulation occurs in a neurone at a stage where the capacity of the cell to generate ATP is insu¤cient to meet its energy demands, it might be predicted that an additional substrate supply might be able, at least temporarily, to rescue cells. Lactate and pyruvate are e¡ective neuronal substrates [81,82] and Eimerl and Schramm [83] have reported that pyruvate plus phosphate added after glutamate exposure decreases cell death. In recent experiments (Budd, Ward and Nicholls, unpublished), we have found that granule cells incubated in glucose-free media and maintained by lactate or pyruvate undergo glutamate-evoked delayed Ca2‡ deregulation after a similar delay as cells maintained by glycolysis. Thus, a time-dependent inhibition of the glycolytic pathway following glutamate exposure

BBABIO 44667 30-7-98

106

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

is not an inherent component of the delayed Ca2‡ deregulation observed in our preparation. 3.5. In situ mitochondria undergo progressive bioenergetic deterioration in glutamate-exposed cells A number of key catabolic mitochondrial enzymes are readily inhibited by oxidative damage, including pyruvate dehydrogenase [84] and aconitase [85]. The mitochondrial electron transport chain itself is a further potential locus for inhibition since complexes I, III and IV each exert some control over the rate of oxygen consumption of isolated brain mitochondria [86]. There is some disagreement as to the bioenergetic consequences of partial inhibition of these complexes: by comparing the titration of respiratory chain inhibitors with intact mitochondria and single complexes Davey and Clark [86] concluded that the individual complexes had to be inhibited by at least 60% before signi¢cant e¡ects were observed on mitochondrial respiration and ATP synthesis. However repopulation of mitochondrially depleted rho-0 cells with mitochondria from patients with either Parkinson's disease displaying 26% de¢ciency in complex I activity [87] or Alzheimer's disease displaying 52% de¢ciency in complex IV activity [88] both resulted in a slowed recovery of [Ca2‡ ]c following Ca2‡ loads. A relatively direct demonstration that in situ mitochondria undergo a progressive deterioration of bioenergetic function following glutamate exposure is the ¢nding by Atlante et al. [89] that mitochondria isolated from glutamate-exposed cerebellar granule cells display a decrease in state 3 respiration which was proportional to the time for which the cells had been exposed to glutamate. After a 5-h exposure, the subsequently isolated mitochondria displayed no signi¢cant respiratory control. Inhibited state 3 respiration implies a metabolic or respiratory chain limitation rather than a loss of respiratory control, although the locus of inhibition was not established in this study. 3.6. Delayed Ca 2+ deregulation may be triggered by mitochondrial Ca 2+ accumulation The total 45 Ca2‡ accumulated by cortical neurones [4] or cerebellar granule cells [5] in the presence of

glutamate has been considered to be a reliable predictor of the extent of subsequent cell death. However, this has recently been challenged by the report that NMDA-evoked uptake is still much more toxic than an equivalent extent of KCl-evoked 45 Ca uptake [6], suggesting that the extreme toxicity of glutamate-evoked Ca2‡ uptake may be due to the selective direction of the Ca2‡ onto an excitotoxic locus in the close vicinity of the intracellular face of the NMDA receptor [3]. However, since the mitochondrial matrix is the only compartment capable of sequestering this Ca2‡ , it is of central importance to establish the role which matrix Ca2‡ accumulation plays a role in excitotoxicity. An additional factor, which has not generally been considered, is that the initial rate of mitochondrial Ca2‡ loading, rather than the absolute amount accumulated may be critical. Thus, low a¤nity Ca2‡ probes indicate very large [Ca2‡ ]c transients in response to glutamate [7,10] which would be predicted to divert all mitochondrial proton pumping to Ca2‡ accumulation [12]. Indeed, the net accumulation of 45 Ca2‡ by granule cells is largely complete within 5 min (Budd and Nicholls, unpublished). Addition of NMDA or glutamate to granule cells in low K‡ medium in the absence of Mg2‡ (the usual experimental protocol) depolarises the plasma membrane [90]; however, glutamate addition to granule cells predepolarised in high KCl medium has been reported to show much decreased excitotoxicity [34]. Since NMDA receptors partially desensitise during prolonged depolarisation, this implies that the initial Ca2‡ entry in polarised cells may be critical in triggering excitotoxicity. It is notable that no lag can be detected between an increase in [Ca2‡ ]c and [Ca2‡ ]m following NMDA receptor activation, whereas some lag is apparent for KCl-evoked Ca2‡ loading [91]. 3.7. Glutamate is not excitotoxic to cells maintaining a high ATP with depolarised mitochondria In the presence of oligomycin, in situ mitochondria maintain a high vim and thus retain the ability to accumulate Ca2‡ . Consistent with this, 45 Ca2‡ is accumulated in the 20 min following addition of glutamate to the same extent as in the absence of the inhibitor [31]. As in the case of depolarisationevoked Ca2‡ loading discussed above, inhibition of

BBABIO 44667 30-7-98

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

the respiratory chain of oligomycin-treated cells should collapse vim without a¡ecting ATP synthesis by glycolysis. Respiratory chain inhibitors under these conditions will, however, collapse vim and abolish mitochondrial Ca2‡ accumulation [31] in addition to blocking electron transport at de¢ned sites and thus potentially a¡ecting the generation of reactive oxygen species. The total accumulation of 45 Ca2‡ within cultured cerebellar granule cells exposed to glutamate can approach 20 nmol/Wl of cell volume (equivalent to 20 mM!) before the cells die [5]. The mitochondrion is the only organelle capable of accumulating such amounts of Ca2‡ , and loss of the mitochondrial sink for Ca2‡ entering via the NMDA receptor should therefore logically result in a rapid saturation of cytoplasmic Ca2‡ chelators and a massive elevation in [Ca2‡ ]c . It was therefore remarkable that granule cells exposed to glutamate in the presence oligomycin plus rotenone [31] or antimycin A (Fig. 3C) show a relatively small peak Ca2‡ elevation, maintain a low subsequent Ca2‡ plateau and can survive for up to 5 h in a Mg2‡ -free medium in the continuous presence of 100 WM glutamate, 10 WM glycine. Any cell death which occurred in the presence of these two inhibitors was independent of the presence of glutamate. Thus, mitochondrial Ca2‡ accumulation may be an essential stage in the chain of events associated with glutamate excitotoxicity. Such a surprising result requires careful controls, the ¢rst is to investigate the possibility that rotenone or oligomycin have previously unrecognised direct inhibitory actions at the NMDA receptor. Oligomycin alone is not neuroprotective [31], while as discussed above, rotenone addition in the absence of oligomycin results in immediate glutamate-evoked Ca2‡ deregulation [31]. This would actually be predicted on bioenergetic grounds: inhibition of the respiratory chain places a severe demand upon the glycolytic capacity of the cell: in addition to the normal cellular house-keeping functions which require ATP, respiratory chain inhibition leads to a reversal of the mitochondrial ATP synthase as glycolytic ATP is utilised to regenerate vim (Fig. 2). 3.8. Elevated [Ca 2+]c is not itself neurotoxic In this analysis of the role of mitochondria in the

107

control of glutamate excitotoxicity, it is important to distinguish between a relatively stable and reversible elevation in [Ca2‡ ]c , which can be a result of enhanced in£ux and/or restricted e¥ux from the cell and the catastrophic and irreversible rise signalling the onset of Ca2‡ deregulation. Thus, Khodorov et al. [92] exposed 7^8 DIV cerebellar granule cell cultures to the combination of antimycin A and oligomycin to depolarise the in situ mitochondria by a similar mechanism to our earlier experiments [41]. These authors found that the combination of inhibitors resulted in a delayed secondary rise in [Ca2‡ ]c after glutamate exposure, in contrast to our ¢ndings that the same inhibitors decreased both peak and plateau responses to continuous glutamate [31], leading them to conclude that `mitochondria play a dominant role in the protection against neuronal Ca2‡ overload induced by excitatory amino acids' [92]. Apparently opposite ¢ndings with the same preparation require some comment: the initial recovery followed immediately by a secondary rise in [Ca2‡ ]c after glutamate found by Khodorov et al. is very similar to the time course of [Ca2‡ ]c in our cells in the presence of protonophore, and suggests an energy limitation; these authors worked at room temperature and employed 5 mM glucose (in contrast to our 37³C and 15 mM glucose). Furthermore, no controls were performed with glutamate in the presence of oligomycin alone to control for ATP depletion under these conditions where the maximal rate of glycolytic ATP synthesis and Ca2‡ extrusion will be restricted, but NMDA receptor activity will be virtually unchanged. Since di¡erent experimental or culture conditions will a¡ect the complex balance between NMDA receptor activity, Ca2‡ extrusion capacity, glycolysis and the capacity for mitochondrial ATP synthesis, it will clearly not be possible to separate e¡ects of mitochondrial ATP synthesis and Ca2‡ sequestration by using conditions which deplete ATP as well as depolarizing the mitochondria. The experiments we performed with rotenone/oligomycin [31] decreased both the mitochondrial Ca2‡ accumulation and the cytoplasmic [Ca2‡ ]c elevations in response to glutamate. While we interpreted the neuroprotection a¡orded by this combination to the abolition of the matrix accumulation, it was possible that the decreased cytoplasmic [Ca2‡ ]c could account

BBABIO 44667 30-7-98

108

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

for the neuroprotection. It is therefore of considerable signi¢cance that the increased [Ca2‡ ]c and decreased matrix Ca2‡ produced by protonophores in cultured hippocampal neurones [8], cortical neurones [93] and an immortalised hippocampal neuronal cell line [94] is not only not neurotoxic, but a¡ords protection against glutamate excitotoxicity [8,93,94]. Putting our results together with these, one can conclude that mitochondrial Ca2‡ loading is potentially excitotoxic, but that elevated cytoplasmic Ca2‡ , while controlling the extent of mitochondrial Ca2‡ loading and signalling delayed Ca2‡ deregulation, is not responsible for the intermediate events leading to a progressive deterioration in cell function. 3.9. The role of mitochondrially initiated oxidative damage Glutamate excitotoxicity requires oxygen; thus hippocampal neurones exposed to glutamate under hypoxic conditions show no more cell death than due to hypoxia alone [2]. Additionally, NMDA antagonists, which protect against excitotoxic damage to cortical neurones following chemical ischaemia, need only be present during reperfusion [95]; in other words, the return of oxidative metabolism triggers a critical period of toxic NMDA receptor activation. Reoxygenation is associated both with mitochondrial Ca2‡ accumulation and the generation of reactive oxygen species (ROS) and a central question concerns the causal relationships between these two parameters and the ultimate death of the cell. We wish to focus on just one aspect of a complex topic, namely the role of the in situ mitochondrial membrane potential in the generation of potentially toxic superoxide …Oc3 2 † radicals. The predominant source of …Oc3 2 † in most tissues is the mitochondrial respiratory chain (reviewed in [96]). Complex III, and in particular the level of reduction of ubisemiquinone at the Qp site, has generally been considered to be the main source of reactive oxygen species in mitochondria. Thus, with succinate as substrate, the generation of reactive oxygen species by isolated heart mitochondria is relatively high in state 4, but virtually abolished when ADP is added and vim decreases to state 3 levels [97,98]. However, in a recent paper, Herrero and Barja [98] failed to detect peroxide generated during succinate

oxidation by non-synaptic rat brain mitochondria, but did detect the reactive oxygen species with substrates utilizing complex 1. No decrease in peroxide production was found following a state 4^state 3 transition; thus it was proposed that complex I could be a source of reactive oxygen species in state 3. Continued production of …Oc3 2 † in state 3 by isolated brain mitochondria is also indicated by the study of Dykens [99], where Ca2‡ loading of malate/glutamate supplemented mitochondria in the presence of ADP increased production of highly reactive hydroxyl radicals. Con£icting consequences of in situ mitochondrial depolarisation have been reported on the generation of reactive oxygen species in cultured neurones: the production of hydrogen peroxide detected by dichloro£uorescin is inhibited by protonophores [100], in accordance with results with isolated mitochondria [97,101], although the pH sensitivity of the signal from the oxidised product, dichloro£uorescein, creates complications during the large cytoplasmic acidi¢cation produced both by protonophores [102] and by NMDA receptor activation [102,103]. The non-£uorescent membrane permeant hydroethidine (HEt) is oxidised within neurones to the £uorescent ethidium cation by …Oc3 2 † and has been widely used to detect the formation of the radical in a large number of neuronal and non-neuronal studies [104^108]. Similar studies have been performed with dihydrorhodamine-123 [109], which is oxidised to cationic rhodamine-123 predominantly by peroxynitrite [110]. Since both cationic products are membrane permeant, they will accumulate in the matrix of polarised mitochondria, even if they were the product of non-mitochondrial oxidation; thus they do not automatically signal the site of free radical production in imaging studies [111]. In addition, generated ethidium undergoes £uorescent quenching when accumulated within the mitochondria matrix and conversely shows a £uorescent dequenching upon mitochondrial depolarisation [45]. This can be di¤cult to separate from the inherent radical-induced oxidation, particularly in experiments involving glutamate- or protonophore-mediated mitochondrial depolarisation. Thus, a number of studies have reported that protonophore treatment of neurones to depolarise in situ mitochondria causes an enhanced generation of ROS detected by dihydrorhodamine-

BBABIO 44667 30-7-98

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

123 [109] or hydroethidine [104,108], in contrast to results with isolated mitochondria. We have re-examined the behaviour of the latter probe in cultured cerebellar granule cells and found that the £uorescent yield of the ethidium generated by hydroethidine oxidation is enhanced by mitochondrial depolarisation, consistent with £uorescent dequenching as the mitochondria release ethidium and compounded by further £uorescent enhancement as the released ethidium binds to non-mitochondrial nucleic acids [45]. This vim sensitivity could be minimised by the use of very low concentrations of HEt, such that the ethidium remains bound within the mitochondria and is not released on depolarisation, or by extraction and quanti¢cation of the generated ethidium at the end of the experiment. With these precautions, we con¢rm reports that NMDA receptor activation enhances the generation of …Oc3 2 † by granule cells, but ¢nd no enhancement upon protonophore addition, in agreement with the above studies using dichloro£uorescin but in contradiction to those using dihydrorhodamine-123 or HEt. It may, therefore, also be necessary to re-examine the sequence of events discussed in the context on apoptotic signalling, where a decrease in vim is proposed to initiate an increase in the generation of reactive oxygen species, since this has been, in part, based upon experiments using HEt [112]. 3.10. The role of the mitochondrial permeability transition While a consensus appears to be emerging that the MPT, which can be readily observed with isolated mitochondria, occurs in intact cells as an obligatory component of apoptotic and necrotic neuronal cell death, we would wish to counsel caution. While there is evidence in support of an activation of the MPT in myocardial reperfusion injury [113^115] and during oxidative [116] and anoxic [117] stress of hepatocytes, a role in glutamate-evoked neuronal excitotoxicity is currently more speculative. Isolated brain mitochondria incubated in the presence of adenine nucleotides and Pi are able to accumulate large amounts of Ca2‡ and to maintain a stable set-point [13]. Cyclosporin A is an e¡ective inhibitor of the MPT in isolated liver mitochondria exposed to Ca2‡ and Pi (reviewed in [118]) but has

109

been reported to be much less e¡ective against brain mitochondria [119]. The inhibitor has been employed in an attempt to identify a role for the MPT in intact neurones, but while some measure of protection against glutamate is generally observed [9,36,37] but see [30], it is likely that it is the ability of the agent to inhibit calcineurin, rather than the MPT, which underlies most of the observed e¡ects. Thus, while Ankarcrona et al. [120] found that cyclosporin A protected granule cells against both early necrosis and delayed apoptosis induced by glutamate, a similar protection was a¡orded by the more selective calcineurin inhibitor FK-506, which does not interact with the mitochondrial permeability transition pore. Calcineurin inhibition by cyclosporin A has multiple e¡ects, including an increase in spontaneous action potential ¢ring [121] and an inhibition of NMDA receptors [122], while chronic cyclosporin A induces neuronal apoptosis in cortical cultures [123,124]. In our experiments, cyclosporin A, the more selective methylvaline-4-cyclosporin [125] and bongkrekic acid [126] each only a¡orded a slight delay before the onset of delayed Ca2‡ deregulation (Castilho, Budd and Nicholls, unpublished), although a more e¡ective protection has been reported for hippocampal neurones [34]. Even if the MPT is an essential stage in both apoptotic and necrotic cell death, there are many questions which remain to be answered. Since increasing excitotoxic stress causes a shift from apoptosis to necrosis in a single cell preparation [51], this would suggest that the proportion of mitochondria within a neurone undergoing the transition might de¢ne the fate of the cell: a small proportion swelling and releasing the proposed pro-apoptotic factors cytochrome c [127,128] and apoptosis-inducing factor [112] might initiate apoptosis with the residual mitochondria maintaining su¤cient ATP levels for apoptosis to occur, while a more powerful insult would disrupt the majority of mitochondria and lead to death by ATP depletion and subsequent necrosis. However, delayed Ca2‡ deregulation still occurs in cells where mitochondria are prevented from synthesizing ATP by oligomycin [31]. Secondly, with isolated mitochondria the MPT occurs within seconds of Ca2‡ overload; in contrast delayed Ca2‡ deregulation can occur hours after a transient glutamate exposure, when the mitochondria would have un-

BBABIO 44667 30-7-98

110

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

loaded their Ca2‡ . Finally, the MPT as normally investigated in isolated mitochondria is inhibited by Mg2‡ , adenine nucleotides and low pH and is di¤cult to see with NAD-linked substrates [129]. The conditions within the cytoplasm of a glutamate exposed neurone would on each of these counts be predicted to be resistant to the MPT and it is currently unclear what factor would induce the transition in situ. 4. Conclusions The mechanism of glutamate-induced neuronal necrosis is far from understood. The mitochondrion occupies centre stage in three roles: as the prime generator (and possible dissipator) of cellular ATP; as the main sink for the Ca2‡ accumulated via the NMDA receptor and as the major source of potentially excitotoxic reactive oxygen species. These three roles are all interconnected, and we may have made a contribution to unravelling the interactions by highlighting the role of mitochondrial Ca2‡ accumulation. Acknowledgements SLB was supported by a Medical Research Council studentship. The work was supported by a grant from the Wellcome Trust. References [1] M.F. Beal, N. Howell, I. Bodis-Wollner, Mitochondria and Free Radicals in Neurodegenerative Disease, Wiley-Liss, New York, 1997. [2] J.M. Dubinsky, B.S. Kristal, M. Elizondo-Fournier, J. Neurosci. 15 (1995) 7071^7078. [3] M. Tymianski, M.P. Charlton, P.L. Carlen, C.H. Tator, J. Neurosci. 13 (1993) 2085^2104. [4] D.M. Hartley, M.C. Kurth, L. Bjerkness, J.H. Weiss, D.W. Choi, J. Neurosci. 13 (1993) 1993^2000. [5] S. Eimerl, M. Schramm, J. Neurochem. 62 (1994) 1223^ 1226. [6] R. Sattler, M.P. Charlton, M. Hafner, M. Tymianski, Soc. Neurosci. Abstr. 23 (1997) 894.4. [7] K. Hyrc, S.D. Handran, S.M. Rothman, M.P. Goldberg, J. Neurosci. 17 (1997) 6669^6677.

[8] J.M. Dubinsky, S.M. Rothman, J. Neurosci. 11 (1991) 2545^ 2551. [9] R.J. White, I.J. Reynolds, J. Neurosci. 16 (1996) 5688^5697. [10] S. Rajdev, I.J. Reynolds, Neurosci. Lett. 162 (1993) 149^152. [11] F. Zoccarato, D.G. Nicholls, Eur. J. Biochem. 127 (1982) 333^338. î kerman, Biochim. Biophys. Acta [12] D.G. Nicholls, K.E.O. A 683 (1982) 57^88. [13] D.G. Nicholls, I.D. Scott, Biochem. J. 186 (1980) 833^839. î kerman, D.G. Nicholls, Biochem. J. 192 [14] I.D. Scott, K.E.O. A (1980) 873^880. î kerman, D.G. Nicholls, Biochim. Biophys. Acta [15] K.E.O. A 645 (1981) 41^48. [16] S.A. Thayer, R.J. Miller, J. Physiol. (Lond.) 425 (1990) 85^ 115. [17] J.L. Werth, S.A. Thayer, J. Neurosci. 14 (1994) 346^356. [18] D.D. Friel, R.W. Tsien, J. Neurosci. 14 (1994) 4007^4024. î kerman, D.G. Nicholls, Rev. Physiol. Biochem. [19] K.E.O. A Pharmacol. 95 (1983) 149^201. [20] J. Herrington, Y.B. Park, D.F. Babcock, B. Hille, Neuron 16 (1996) 219^228. [21] M.R. Duchen, M. Valdeolmillos, S.C. O'Neill, D.A. Eisner, J. Physiol. (Lond.) 424 (1990) 411^426. [22] R.J. White, I.J. Reynolds, J. Neurosci. 15 (1995) 1318^1328. [23] R.J. White, I.J. Reynolds, J. Physiol. (Lond.) 498 (1997) 31^ 47. [24] D.G. Nicholls, S.J. Ferguson, Bioenergetics, Vol. 2, Academic Press, London, 1992. [25] D.G. Nicholls, Eur. J. Biochem. 50 (1974) 305^315. [26] J.B. Hoek, D.G. Nicholls, J.R. Williamson, J. Biol. Chem. 255 (1980) 1458^1464. [27] I.D. Scott, D.G. Nicholls, Biochem. J. 186 (1980) 21^33. [28] M.R. Duchen, Biochem. J. 283 (1992) 41^50. [29] V.P. Bindokas, R.J. Miller, J. Neurosci. 15 (1995) 6999^ 7011. [30] N.K. Isaev, D.B. Zorov, E.V. Stelmashook, R.E. Uzbekov, M.B. Kozhemyakin, I.V. Victorov, FEBS Lett. 392 (1996) 143^147. [31] S.L. Budd, D.G. Nicholls, J. Neurochem. 67 (1996) 2282^ 2291. [32] J.H. Prehn, V.P. Bindokas, J. Jordan, M.F. Galindo, G.D. Ghadge, R.P. Roos, L.H. Boise, C.B. Thompson, S. Krajewski, J.C. Reed, R.J. Miller, Mol. Pharmacol. 49 (1996) 319^ 328. [33] B.I. Khodorov, V. Pinelis, O. Vergun, T. Storozhevykh, N. Vinskaya, FEBS Lett. 397 (1996) 230^234. [34] J. Keelan, O. Vergun, L. Patterson, M.R. Duchen, Soc. Neurosci. Abstr. 23 (1997) 895.1. [35] L.M. Loew, W. Carrington, R.A. Tuft, F.S. Fay, Proc. Natl. Acad. Sci. U.S.A. 91 (1994) 12579^12583. [36] A.L. Nieminen, T.G. Petrie, J.J. Lemasters, W.R. Selman, Neuroscience 75 (1996) 993^997. [37] A.F. Schinder, E.C. Olson, N.C. Spitzer, M. Montal, J. Neurosci. 16 (1996) 6125^6133. î kerman, Biochim. Biophys. Acta 502 (1978) 359^ [38] K.E.O. A 366.

BBABIO 44667 30-7-98

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112 [39] G.M. Heaton, D.G. Nicholls, Biochem. J. 156 (1976) 635^ 646. [40] D.G. Nicholls, Biochem. J. 170 (1978) 511^522. [41] C.S. Rossi, A.L. Lehninger, J. Biol. Chem. 239 (1964) 3971^ 3980. [42] D.G. Nicholls, Biochem. J. 176 (1978) 463^474. [43] D.G. Nicholls, M. Crompton, FEBS Lett. 111 (1980) 261^ 268. [44] M.R. Duchen, T.J. Biscoe, J. Physiol. (Lond.) 450 (1992) 33^ 61. [45] S.L. Budd, R.F. Castilho, D.G. Nicholls, FEBS Lett. 415 (1997) 21^24. [46] C.D. Benham, M.L. Evans, C.J. McBain, J. Physiol. (Lond.) 455 (1992) 567^583. [47] J.L. Werth, Y.M. Usachev, S.A. Thayer, J. Neurosci. 16 (1996) 1008^1015. [48] M. Glinn, B. Ni, R.P. Irwin, S.W. Kelley, S.-Z. Lin, S.M. Paul, Soc. Neurosci. Abstr. 23 (1997) 895.11. [49] J.M. Dubinsky, J. Neurosci. 13 (1993) 623^631. [50] J.M. Dubinsky, B.S. Kristal, M. Elizondo-Fournier, Neuropharmacology 34 (1995) 701^711. [51] M. Ankarcrona, J.M. Dypbukt, E. Bonfoco, B. Zhivotovsky, S. Orrenius, S.A. Lipton, P. Nicotera, Neuron 15 (1995) 961^973. [52] B.I. Khodorov, D.A. Fayuk, S.G. Koshelev, O.V. Vergun, V.G. Pinelis, N.P. Vinskaya, T.P. Storozhevykh, E.N. Arsenyeva, L.G. Khaspekov, A.P. Lyzhin, N. Isaev, I.V. Victorov, J.M. Dubinsky, Int. J. Neurosci. 88 (1996) 215^241. [53] T. Nitatori, N. Sato, S. Waguri, Y. Karasawa, H. Araki, K. Shibanai, E. Kominami, Y. Uchiyama, J. Neurosci. 15 (1995) 1001^1011. [54] M. Ankarcrona, B. Zhivotovsky, T. Holmstro«m, A. Diana, J.E. Eriksson, S. Orrenius, P. Nicotera, Neuroreport 7 (1996) 2659^2664. [55] F. Dessi, C. Charriaut-Marlangue, M. Khrestchatisky, Y. Ben-Ari, J. Neurochem. 60 (1993) 1953^1955. [56] J. Ikeda, S. Terakawa, S. Murota, I. Morita, K. Hirakawa, J. Neurosci. Res. 43 (1996) 613^622. [57] B.J. Gwag, J.Y. Koh, J.A. Demaro, H.S. Ying, M. Jacquin, D.W. Choi, Neuroscience 77 (1997) 393^401. [58] E. Bonfoco, D. Krainc, M. Ankarcrona, P. Nicotera, S.A. Lipton, Proc. Natl. Acad. Sci. U.S.A. 92 (1995) 7162^7166. [59] Z. Pang, J.W. Geddes, J. Neurosci. 17 (1997) 3064^3073. [60] B.H. Ni, X. Wu, Y.S. Du, Y. Su, E. Hamilton-Byrd, P.K. Rockey, P. Rosteck Jr., G.G. Poirier, S.M. Paul, J. Neurosci. 17 (1997) 1561^1569. [61] R.P. Xiao, H.H. Valdivia, K. Bogdanov, C. Valdivia, E.G. Lakatta, H.P. Cheng, J. Physiol. (Lond.) 500 (1997) 343^ 354. [62] J.B. Schulz, M. Weller, T. Klockgether, J. Neurosci. 16 (1996) 4696^4706. [63] H. Hara, R.M. Friedlander, V. Gagliardini, C. Ayata, K. Fink, Z.H. Huang, M. Shimizu-Sasamata, J.Y. Yuan, M.A. Moskowitz, Proc. Natl. Acad. Sci. U.S.A. 94 (1997) 2007^2012.

111

[64] S.A. Loddick, A. MacKenzie, N.J. Rothwell, Neuroreport 7 (1996) 1465^1468. [65] F.J. Gottron, H.S. Ying, D.W. Choi, Mol. Cell. Neurosci. 9 (1997) 159^169. [66] R.C. Armstrong, T.J. Aja, K.D. Hoang, S. Gaur, X. Bai, E.S. Alnemri, G. Litwack, D.S. Karanewsky, L.C. Fritz, K.J. Tomaselli, J. Neurosci. 17 (1997) 553^562. [67] L. Kiedrowski, G. Brooker, E. Costa, J.T. Wroblewski, Neuron 12 (1994) 295^300. [68] M. Erecinska, D. Nelson, J. Deas, I.A. Silver, Brain Res. 726 (1996) 153^159. [69] J.G. Greene, J.T. Greenamyre, Prog. Neurobiol. 48 (1996) 613^621. [70] R.C. Henneberry, in: M.F. Beal, N. Howell, I. Bodis-Wollner (Eds.), Mitochondria and Free Radicals in Neurodegenerative Disease, Wiley-Liss, New York, 1997, pp. 111^143. [71] F. Blandini, R.H.P. Porter, J.T. Greenamyre, Mol. Neurobiol. 12 (1996) 73^94. [72] E. Brouillet, P. Hantraye, R.J. Ferrante, R. Dolan, A. Leroy Willig, N.W. Kowall, M.F. Beal, Proc. Natl. Acad. Sci. U.S.A. 92 (1995) 7105^7109. [73] A. Novelli, J.A. Reilly, P.G. Lysko, R.C. Henneberry, Brain Res. 451 (1988) 205^212. [74] A.M. Marini, J.P. Schwartz, I.J. Kopin, J. Neurosci. 9 (1989) 3665^3672. [75] P.G. Lysko, J.A. Cox, M.A. Vigano, R.C. Henneberry, Brain Res. 499 (1989) 258^266. [76] J.G. Greene, J.T. Greenamyre, J. Neurochem. 64 (1995) 2332^2338. [77] R. Henshaw, B.G. Jenkins, J.B. Schulz, R.J. Ferrante, N.W. Kowall, B.R. Rosen, M.F. Beal, Brain Res. 647 (1994) 161^ 166. [78] M.F. Beal, E. Brouillet, B. Jenkins, R. Henshaw, B. Rosen, B.T. Hyman, J. Neurochem. 61 (1993) 1147^1150. [79] R.A. Kauppinen, D.G. Nicholls, J. Neurochem. 47 (1986) 1864^1869. [80] A. Schurr, R.S. Payne, J.J. Miller, B.M. Rigor, Brain Res. 744 (1997) 105^111. [81] M. Villalba, A. Mart|¨nez-Serrano, P. Go¨mez-Puertas, P. Blanco, C. Bo«rner, A. Villa, M. Casado, C. Gime¨nez, R. Pereira, E. Bogo¨nez, T. Pozzan, J. Satrustegui, J. Biol. Chem. 269 (1994) 2468^2476. [82] C. Vicario, C. Arizmendi, G. Malloch, J.B. Clark, J.M. Medina, J. Neurochem. 57 (1991) 1700^1707. [83] S. Eimerl, M. Schramm, J. Neurochem. 65 (1995) 739^743. [84] T. Tabatabaie, J.D. Potts, R.A. Floyd, Arch. Biochem. Biophys. 336 (1996) 290^296. [85] M. Patel, B.J. Day, J.D. Crapo, I. Fridovich, J.O. McNamara, Neuron 16 (1996) 345^355. [86] G.P. Davey, J.B. Clark, J. Neurochem. 66 (1996) 1617^1624. [87] J.P. Sheehan, R.H. Swerdlow, W.D. Parker, S.W. Miller, R.E. Davis, J.B. Tuttle, J. Neurochem. 68 (1997) 1221^1233. [88] J.P. Sheehan, R.H. Swerdlow, S.W. Miller, R.E. Davis, J.K. Parks, W.D. Parker, J.B. Tuttle, J. Neurosci. 17 (1997) 4612^4622.

BBABIO 44667 30-7-98

112

D.G. Nicholls, S.L. Budd / Biochimica et Biophysica Acta 1366 (1998) 97^112

[89] A. Atlante, S. Gagliardi, G.M. Minervini, E. Marra, S. Passarella, P. Calissano, Neuroreport 7 (1996) 2519^2523. [90] M.J. Courtney, J.J. Lambert, D.G. Nicholls, J. Neurosci. 10 (1990) 3873^3879. [91] T.-I. Peng, J.T. Greenamyre, Soc. Neurosci. Abstr. 23 (1997) 685.19. [92] B.I. Khodorov, V.G. Pinelis, T. Storozhevykh, O.V. Vergun, N.P. Vinskaya, FEBS Lett. 393 (1996) 135^138. [93] A.K. Stout, H.M. Raphael, J.M. Scanlon, I.J. Reynolds, Soc. Neurosci. Abstr. 23 (1997) 650.6. [94] S. Tan, Y. Sagara, D. Schubert, Soc. Neurosci. Abstr. 23 (1997) 542.1. [95] J.J. Vornov, J. Neurochem. 65 (1995) 1681^1691. [96] V.P. Skulachev, Q. Rev. Biophys. 29 (1996) 169^202. [97] A. Boveris, N. Oshino, B. Chance, Biochem. J. 128 (1972) 617^630. [98] A. Herrero, G. Barja, J. Bioenerg. Biomembr. 29 (1997) 241^249. [99] J.A. Dykens, J. Neurochem. 63 (1994) 584^591. [100] I.J. Reynolds, T.G. Hastings, J. Neurosci. 15 (1995) 3318^ 3327. [101] M. Cino, R.F. Del Maestro, Arch. Biochem. Biophys. 269 (1989) 623^638. [102] G.J. Wang, R.D. Randall, S.A. Thayer, J. Neurophysiol. 72 (1994) 2563^2569. [103] Z. Hartley, J.M. Dubinsky, J. Neurosci. 13 (1993) 4690^ 4699. [104] J.M. Scanlon, E. Aizenman, I.J. Reynolds, Eur. J. Pharmacol. 326 (1997) 67^74. [105] N. Zamzami, P. Marchetti, M. Castedo, D. Decaudin, A. Macho, T. Hirsch, S.A. Susin, P.X. Petit, B. Mignotte, G. Kroemer, J. Exp. Med. 182 (1995) 367^377. [106] M. Castedo, A. Macho, N. Zamzami, T. Hirsch, P. Marchetti, J. Uriel, G. Kroemer, Eur. J. Immunol. 25 (1995) 3277^3284. [107] A. Macho, M. Castedo, P. Marchetti, J.J. Aguilar, D. Decaudin, N. Zamzami, P.M. Girard, J. Uriel, G. Kroemer, Blood 86 (1995) 2481^2487. [108] V.P. Bindokas, J. Jordan, C.C. Lee, R.J. Miller, J. Neurosci. 16 (1996) 1324^1336. [109] L.L. Dugan, S.L. Sensi, L.M.T. Canzoniero, S.D. Handran, S.M. Rothman, T.S. Lin, M.P. Goldberg, D.W. Choi, J. Neurosci. 15 (1995) 6377^6388.

[110] N.W. Kooy, J.A. Royall, H. Ischiropoulos, J.S. Beckman, Free Radic. Biol. Med. 16 (1994) 149^156. [111] L.M. Henderson, J.B. Chappell, Eur. J. Biochem. 217 (1993) 973^980. [112] P.X. Petit, S.A. Susin, N. Zamzami, B. Mignotte, G. Kroemer, FEBS Lett. 396 (1996) 7^13. [113] R.A. Altschuld, C.M. Hohl, L.C. Castillo, A.A. Garleb, R.C. Starling, G.P. Brierley, Am. J. Physiol. 262 (1992) H1699^H1704. [114] E.J. Gri¤ths, A.P. Halestrap, J. Mol. Cell Cardiol. 25 (1993) 1461^1469. [115] E.J. Gri¤ths, A.P. Halestrap, Biochem. J. 307 (1995) 93^ 98. [116] G.E. Kass, M.J. Juedes, S. Orrenius, Biochem. Pharmacol. 44 (1992) 1995^2003. [117] J.G. Pastorino, J.W. Snyder, J.B. Hoek, J.L. Farber, Am. J. Physiol. 268 (1995) C676^85. [118] P. Bernardi, V. Petronilli, J. Bioenerg. Biomembr. 28 (1996) 131^138. [119] B.S. Kristal, J.M. Dubinsky, J. Neurochem. 69 (1997) 524^ 538. [120] M. Ankarcrona, J.M. Dypbukt, S. Orrenius, P. Nicotera, FEBS Lett. 394 (1996) 321^324. [121] R.G. Victor, G.D. Thomas, E. Marban, B. O'Rourke, Proc. Natl. Acad. Sci. U.S.A. 92 (1995) 6269^6273. [122] Y.F. Lu, K. Tomizawa, A. Moriwaki, Y. Hayashi, M. Tokuda, T. Itano, O. Hatase, H. Matsui, Brain Res. 729 (1996) 142^146. [123] J.W. McDonald, M.I. Behrens, C. Chung, T. Bhattacharyya, D.W. Choi, Brain Res. 759 (1997) 228^232. [124] J.W. McDonald, M.P. Goldberg, B.J. Gwag, S.I. Chi, D.W. Choi, Ann. Neurol. 40 (1996) 750^758. [125] A. Nicolli, E. Basso, V. Petronilli, R.M. Wenger, P. Bernardi, J. Biol. Chem. 271 (1996) 2185^2192. [126] N. Brustovetsky, M. Klingenberg, Biochemistry 35 (1996) 8483^8488. [127] J. Yang, X.S. Liu, K. Bhalla, C.N. Kim, A.M. Ibrado, J.Y. Cai, T.I. Peng, D.P. Jones, X.D. Wang, Science 275 (1997) 1129^1132. [128] R.M. Kluck, E. Bossy-Wetzel, D.R. Green, D.D. Newmeyer, Science 275 (1997) 1132^1136. [129] P. Bernardi, K.M. Broekemeier, D.R. Pfei¡er, J. Bioenerg. Biomembr. 26 (1994) 509^518.

BBABIO 44667 30-7-98