Background

A long-standing perception is that upon activation of glutamate receptors followed by a robust Ca²⁺ influx, in situ mitochondria generate reactive oxygen species (ROS) [1–6]. These studies inferred that mitochondrial Ca²⁺ sequestration is a prerequisite for production of ROS: abolition of mitochondrial membrane potential (ΔΨm) by mitochondrial poisons, and thus, electrophoretic calcium uptake or direct inhibition of the uniporter with ruthenium red prevented ROS generation. Parallel to these reports, the response of isolated mitochondria to calcium loading in terms of ROS production has also been scrutinized; it was found that mitochondrial Ca²⁺ uptake led to free radical production [7–12]. On the other hand, it was shown that ROS formation depends steeply on ΔΨm [13–15], and from a thermodynamic point of view, Ca²⁺ uptake occurring at the expense of membrane potential should result in a decrease in ROS production (in the absence of respiratory chain inhibitors), as it has also been demonstrated (reviewed in [16,17]). Nevertheless, brain mitochondria also generate ROS in a ΔΨm-independent manner [18–20]. The reason behind the opposing observations that mitochondrial ROS production increases or decreases upon Ca²⁺ uptake is not entirely clear; a plausible explanation lies in the condition in which mitochondria are probed for ROS, specifically whether or not the organelles undergo permeability transition pore (PTP) formation. Among the many features accompanying mitochondrial permeability transition (for a full list see [16] and references therein) loss of glutathione, cytochrome c, substrates and pyridine nucleotides are characteristic. This leads to an increase in ROS production from the impaired mitochondria by multiple means: (a) loss of glutathione from the matrix decreases the antioxidant capacity resulting in a net ‘steady-state’ increase in the amount of ROS [21]; (b) loss of cytochrome c impairs the flow of electrons in the respiratory chain inducing over-reduction of the complexes, favouring the generation of ROS [16,17,22]; (c) reduction in the matrix concentration of electron acceptors, i.e. NAD⁺, results in ROS emission from the α-ketoglutarate dehydrogenase complex (KGDHC) [23,24].

Mitochondrial formation of ROS-the role of KGDHC

The first observation of ROS production in mitochondrial fragments was reported in 1966 by Jensen [25]. Subsequent studies by Britton Chance's group, established that mitochondria generate ROS [26,27]. The sites of ROS formation within the organelle have been extensively reviewed elsewhere [17,20,28]. Among them, complex I [29–31] and III [32–35] of the respiratory chain have attracted most attention. However, in light of recent results on the substantial contribution of matrix enzymes (especially KGDHC) on ROS generation, we believe that in addition to the respiratory chain, the components of the Krebs cycle should also be considered as a possible important source of ROS in mitochondria.

Almost all studies have used respiratory chain inhibitors as tools to maximize and to identify potential sites of ROS production in isolated mitochondria. They revealed that inhibition of complexes I and III, respectively, with specific mitochondrial toxins such as rotenone and antimycin A, results in high rates of ROS production [29,36,37]. For complex I in particular, the ‘reverse electron transport’ mode of ROS production has gained momentum throughout the past four decades [38]; reverse electron transport requires high ΔΨm and is abolished by the complex I inhibitor, rotenone [18], but the pathophysiological relevance of this mode of ROS generation is questionable. Similar approaches have been used successfully to study ROS production in in situ brain mitochondria present in isolated nerve terminals (synaptosomes) [39], but no information is yet available regarding the specific sites or mechanisms of ROS generation in the absence of respiratory chain inhibitors.

Numerous reports in isolated or in situ mitochondria support complex I being regarded as a major site of ROS production, however, a lingering assumption remains that all ROS production caused by complex I inhibitors occurs at the complex I site. There are other sources of ROS within the mitochondrial matrix that are in equilibrium with the ratio NAD(P)H/NAD(P)⁺, such as the dihydrolipoyl dehydrogenase (Dld) component of KGDHC [40]. In intact mitochondria, complex I inhibition by any means, inevitably results in over-reduction of most if not all NAD⁺-linked matrix enzymes.

Among the NAD⁺-linked dehydrogenases that generate ROS, KGDHC deserves special attention. KGDHC is a mitochondrial enzyme tightly bound to the inner mitochondrial membrane on the matrix side [41]. It (as well as other but not all dehydrogenases) binds to complex I of the mitochondrial respiratory chain [42] and may form a part of the TCA cycle enzyme supercomplex [43]. Mammalian KGDHC is composed of multiple copies of three enzymes: α-ketoglutarate dehydrogenase (E1; EC 1.2.4.2), dihydrolipoamide succinyltransferase (E2; EC 2.3.1.61), and dihydrolipoamide dehydrogenase (E3 or Dld; EC 1.8.1.4). Dld is also a part of other multienzyme complexes such as the pyruvate dehydrogenase complex (PDHC), the branched chain ketoacid dehydrogenase complex, and the glycine cleavage system [44–47]. The catalytic mechanism of the α-ketoacid dehydrogenase complex was reviewed by Bunik [40].

Isolated KGDHC [23] as well as PDHC [24] in isolated and in in situ mitochondria respectively produce superoxide and H₂O₂. Quantitatively, it seems likely that KGDHC generates the majority of ROS among dehydrogenases: under conditions of maximum respiration induced with either ADP or an uncoupler, α-ketoglutarate supports the highest rate of H₂O₂ production [24]. The Dld component of KGDHC, and to a lesser degree of PDHC, generate ROS in isolated mouse brain mitochondria [24]. The reasons behind this quantitative discrepancy among the Dld-containing dehydrogenases regarding ROS production are at present, unknown. The isolated Dld subunit is able to form H₂O₂ and superoxide radical, accompanying NADH oxidation [40,48,49]. This observation is important as to the mechanisms and sites of ROS production in mitochondria because the flavin of the Dld subunit is abundant and possesses a sufficiently negative redox potential (Em 7.4 = −283 mV) to allow superoxide formation [50,51]. Moreover, H₂O₂ production by brain mitochondria isolated from heterozygous knockout mice deficient in Dld is significantly diminished, as compared to wild-type littermates [24].

Within KGDHC, it is the flavin or the neighbouring disulfide bridge in the catalytic centre of the Dld component that could act as an electron donor for superoxide formation [52]. KGDHC is activated by low concentrations of Ca²⁺ and matrix ADP [53–56]. Considering that KGDHC-mediated ROS production requires a fully active complex with all the cofactors and substrates (except NAD⁺), the fact that the enzyme activity is stimulated by Ca²⁺ and ADP may perhaps account for previous findings that mitochondrial ROS production was increased by Ca²⁺[7–11,14] and ADP [30]. Results obtained in our laboratory [23] demonstrate that Ca²⁺ activates ROS production by isolated KGDHC both in the presence and in the absence of pyridine nucleotides. Still, the reduced Dld subunit is the most likely source of ROS under conditions of an elevated NADPH/NADP⁺ ratio in the mitochondrial matrix [23,24]. The conditions promoting KGDHC-mediated ROS production may be any that increase the intramitochondrial NADH/NAD⁺ ratio (e.g. inhibition of oxidative phosphorylation or inhibition of any segment of the mitochondrial electron transport chain). This hypothesis is favoured by our results showing that ROS production by isolated KGDHC is strongly dependent on the NADH/NAD⁺ ratio [23].

The relationship of ROS to KGDHC is extended in an ‘ouroboros’ fashion to the self-inactivation of the enzyme by ROS. We demonstrated previously, that KGDHC is sensitive to inhibition by H₂O₂[57]. That inevitably leads to a decrease in complex I function, as repeatedly demonstrated [57–61], since KGDHC which is the rate-limiting step of the TCA cycle provides NADH as a substrate for the respiratory chain complex.

It is difficult to establish the extent of contribution of KGDHC and other enzymes to overall ROS production in mitochondria, as this is prone to be condition-dependent (e.g. choice of substrate), in addition to heavily reliant on non-Krebs cycle enzyme mediated ROS formation through the respiratory chain; i.e. both complex I and KGDHC are in equilibrium with the NAD(P)H/NAD(P)⁺ ratio, and therefore interdependent on each other concerning ROS formation. Thus, in organello it might not be possible to accurately estimate the degree of contribution of each ROS-forming site, because inhibition of ROS production in the one may aggravate ROS formation in the other, and vice versa.

The observation that KGDHC generates and is also self-inactivated by ROS, is of paramount importance in neuronal pathology. A compelling body of evidence indicates that mitochondria are the major source of ROS in several neurodegenerative conditions [37,62]. Also, KGDHC activity is severely reduced in a variety of neurodegenerative diseases associated with impaired mitochondrial functions, specifically, Alzheimer's disease [63–67], Parkinson's disease [68–71], progressive supranuclear palsy [72,73] and Wernicke–Korsakoff syndrome [74]. It is not known if the physical association of KGDHC with complex I (see above) plays a role in the dual deficiency of these protein complexes in Parkinson's disease. It appears that neuronal pathology is preferentially associated with KGDHC deficiency: in an animal model of diminished KGDHC activity caused by thiamine deprivation in the diet, neurons are dying, while endothelial cells, astrocytes and microglia are not affected. In fact, KGDHC activity is increased in these non-neuronal cell types [63], which might indicate that KGDHC deficiency has an etiologic role in the manifestation of some neurodegenerative diseases [75,76]. It must be emphasized that this multienzyme is the rate-limiting step of the Krebs cycle, and if altered that would inpact on the overall energy production in the affected tissue. Moreover, in vivo studies suggested that reduced activity of KGDHC predisposes to damage by toxins, such as 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) or malonate, reducing the capacity of neurons to respond to stress [77,78]. In addition, it was shown recently that reduction in the E2 subunit of KGDHC is associated with diminished growth of cells and impaired antioxidant defence systems, without a reduction in the overall activity of the complex [79]. This finding should come at no surprise: several enzymes of the TCA cycle (and at least one glycolytic enzyme [80]) have roles beyond those of just being cycle participants for the provision of reducing equivalents: aconitase, isocitrate dehydrogenase and kgd2p (a subunit of KGDHC in yeast equivalent to E2 in mammals), have two or more different functions, in addition to having supporting functions for oxidative defences [79], involving the thioredoxin system [40]. Aconitase acts also as an iron-responsive element binding protein, isocitrate dehydrogenase is an RNA-binding protein, while kgd2p is a mitochondrial DNA binding protein [81–84].

Mitochondria from different brain regions contain different amounts of KGDHC [85,86], which may account for regional vulnerability. For instance, the cholinergic neurons of the nucleus basalis of Meynert have high levels of KGDHC, and these neurons are particularly vulnerable in Alzheimer disease [64].

Nevertheless, the relationship between KGDHC activity and mitochondrial damage per se is much less clear. One can speculate that KGDHC-mediated oxidative stress predisposes the cell to succumb to concomitant adverse conditions; in addition, a diminished KGDHC activity will lead to insufficient provision of reducing equivalents, lowering the energetic capacity of the mitochondria of the affected cell. However, studies with the KGDHC inhibitor KMV (alpha-keto-beta-methyl-n-valeric acid) suggest that inhibition of the enzyme might contribute to cell death by induction of permeability transition [87].

Permeability transition pore in situ

Permeability transition pore is considered to be a channel with a large conductance provided by proteins residing in both the inner and outer mitochondrial membrane, that is activated by mitochondrial Ca²⁺ overloading and other factors including oxidative stress [88,89]. In neurons the presence of PTP in situ has not gained wide acceptance among investigators and results published in the literature support views of both its presence and absence in several in vitro models of neurodegeneration [90–98]. One of the possible reasons for this discrepancy is that sensitivity to cyclosporin A is considered pathognomonic for mitochondrial PTP (see also [90]). Cyclosporin A is a potent inhibitor of PTP in isolated liver mitochondria [99] that has been demonstrated to be effective also in situ in this and other organs [100–103]. The sensitivity of isolated brain mitochondria to cyclosporin A depends highly on the conditions: in the absence of adenine nucleotides and magnesium, cyclosporin A mitigates Ca²⁺-induced mitochondrial pore formation [104,105] however, in the presence of 3 mm ATP plus 1 mm free Mg²⁺, cyclosporin A is only marginally effective, provided that mitochondria are challenged by boluses of CaCl₂[104]. In the case that Ca²⁺ loading occurs slowly, cyclosporin A delays onset of PTP in brain mitochondria extensively, even in the presence of adenine nucleotides and magnesium [106]. The caveat here is that despite the decreased ATP levels to less than the millimolar range during ischemic deenergizing, ADP levels approximate 400 µm[107], and the K_i for inhibition of the PTP by ADP is in the low micromolar range [108]. Moreover, in situ neuronal mitochondria are exposed to bolus-like additions of Ca²⁺[109] during intense glutamate receptor stimulation for the duration of seizure activity or reversal of glutamate transporters throughout ischemia [110]. Ca²⁺ cycling across the mitochondrial inner membrane ensues subsequently [111]. On the other hand, intense stimulation of N-methyl-d-aspartate (NMDA) receptors on cultured cerebellar granule and hippocampal neurons causes major ultrastructural alterations of mitochondria, implying the activation of some form of PTP [112,113]. Mitochondrial alterations suggestive of pore opening is also demonstrated in vivo, during the postischemic period in the gerbil brain [114]. Yet, to identify these in situ mitochondrial alterations as the PTP on the basis of the functional/morphological/pharmacological criteria applied for isolated mitochondria is rather hasty.

Collectively, the sensitivity of glutamate-induced neuronal damage to cyclosporin A as diagnostic for PTP occurrence is unreliable. This ambiguity is also nurtured by the complex pharmacology of cyclosporin A and its affinity to non-PTP targets [90,115] that could be involved in the manifestation of neuronal injury [116], in addition to the fact that PTP may not have a causal role in excitotoxic cell death. It is to be noted that the magnitude of the literature involving cyclosporin A unrelated to mitochondria is 12 times larger than that implicating PTP! The nonimmunosuppressant analogue, N-methyl-valine-4-cyclosporin also gave contrasting results, conferring neuronal protection against excitotoxicity in some studies [92,117,118], but not in others [94].

What could be important though, is the role of the in situ mitochondrial pore formation in dictating the type of death that the ill-fated neuron will follow. A most simplistic view is that this pore will promote apoptosis due to release of cytochrome c followed by activation of caspases [119,120], provided that pertaining conditions divert the type of cell death from the necrotic to the apoptotic pathway [121,122]. The role of mitochondria in apoptosis and necrosis has been extensively reviewed elsewhere [121,123–131]. Recently however, a blow was delivered to the conception that PTP contributes to apoptotic cell death by three almost simultaneous and independent reports using cyclophilin D knockout mice [132–134]. Cyclophilin D is a component of the PTP complex [135,136] and it is the target for cyclosporin A. As expected, mitochondria isolated from the cyclophilin D knockout mice were much less susceptible to various PTP-inducing regimes, that are otherwise sensitive to cyclosporin A treatment (see also [137]). Unexpectedly though, tissues obtained from mutant mice were not more resistant to several apoptotic stimuli than those from their wild-type littermates; however, the resistance of the mutant mice to treatments known to result in necrotic cell death was much higher than in control mice.

Mitochondrial Ca²⁺-flux pathways and relation to signal transduction

In general, the contribution of mitochondria to intracellular Ca²⁺ homeostasis is ascribed to uptake and release through the uniporter, the mitochondrial Na⁺/Ca²⁺ exchanger, the PTP (both high- and low-conductance mode) and other less well characterized pathways, such as the ‘Na⁺-independent pathway for Ca²⁺ efflux’ and a H⁺/Ca²⁺ antiporter [89,138]. With the exception of the high-conductance mode of PTP and the uniporter, none of these molecular complexities have been described to be modulated by any signal transduction mediators. High-conductance PTP is known to be affected by matrix Ca²⁺ and ROS [89]. Also the uniporter is supposed to be activated only if extramitochondrial Ca²⁺ levels exceed a certain threshold concentration, termed the ‘set-point’[139]; however, this has been challenged recently, showing that in situ mitochondria accumulate Ca²⁺ well below the set-point, in permeabilized rat adrenal glomerulosal cells [140]. Nonetheless, despite that mitochondria are increasingly viewed as active mediators of [Ca²⁺]_c regulation, the pathways that these organelles use to achieve this task are rather passive.

To this repertoire of Ca²⁺ influx and efflux mechanisms across the mitochondrial membranes, a novel Ca²⁺-efflux-only machinery has been recently added: a channel located in the inner membrane activated by diacylglycerols (DAGs) [141]. This is either a single channel with numerous substates (mean conductance ≈ 200 pS), or multiple channels with unequal conductance. DAGs cause a biphasic form of Ca²⁺ efflux in Ca²⁺-loaded mitochondria: the first wave of efflux is attributed to the activation of the DAG-sensitive nonselective cationic channels; the second wave is due to opening of the PTP. It is not yet known how activation of the former leads to induction of the latter. One is tempted to hypothesize that the initial Ca²⁺ efflux through DAG-sensitive channels causes intense Ca²⁺ cycling due to reuptake by the uniporter, leading to PTP. However, cyclosporin A fails to defend against the secondary Ca²⁺ efflux in liver mitochondria in the presence of DAGs, in which the immunosuppressant otherwise confers significant protection against PTP induction.

The role of DAG-sensitive mitochondrial channels in physiological [Ca²⁺]_c regulation can easily be envisaged: upon phosphatidylinositol (4,5) bisphosphate (PIP₂) hydrolysis, inositol-1,4,5-triphosphate (IP₃) diffuses in the cytosol to activate IP₃ receptors on the endoplasmic reticulum releasing Ca²⁺ to the cytoplasm, followed by triggering of Ca²⁺ influx from the extracellular space [142]. The role of mitochondria in shaping Ca²⁺ transients during such events is recognized in limiting Ca²⁺ diffusion, and secondarily relieving Ca²⁺-mediated negative feedback on the Ca²⁺ flux pathways themselves [143]. However, the other obligatory metabolite of PIP₂ catabolism − DAG − may regulate the role of mitochondria in shaping those [Ca²⁺]_c transients: mitochondrial DAG-sensitive channels would re-release sequestered matrix Ca²⁺ only in the vicinity where DAGs are formed most likely in microdomains, since this second messenger is extremely lipophilic and does not diffuse into the aqueous cytosol.

Mitochondrial permeabilization and the delayed calcium deregulation

The association of ROS to a possible PTP induction prior to neuronal cell death has received much attention in relation to the delayed, irreversible rise in [Ca²⁺]_c following a prolonged glutamate stimulus, coined by Nicholls' group as ‘delayed calcium deregulation, DCD’[144] that commits a neuron to die [145–148]. DCD was originally described by Manev and colleagues [149], further characterized by the groups of Thayer [150] and Tymianski [146]. However, credit should also be given to an earlier work by Connor and colleagues, showing that a short exposure (1–3 s) of CA1 hippocampal neurons to NMDA causes an abrupt elevation in [Ca²⁺]_c that returns to baseline; a subsequent exposure to NMDA of the same duration a few minutes later leads to an irreversible and sustained increase in intracellular [Ca²⁺]_c in apical dendrites [151]. DCD is invariably demonstrated in every neuronal cell type studied, i.e. spinal [146], hippocampal [150], cerebellar granule [152], striatal [117] and cortical neurons [93,153]. The phenomenon is not observed if high extracellular K⁺ is alternatively employed to elevate [Ca²⁺]_c; this led to the proposal of a ‘source specificity’ of Ca²⁺-induced neurotoxicity [146]. However, this was subsequently challenged by studies demonstrating that activation of NMDA receptors produces much larger Ca²⁺ entry than activation of voltage-dependent Ca²⁺ channels by high extracellular K⁺[154].

This secondary [Ca²⁺]_c rise is not inhibitable by postglutamate addition of antagonists of NMDA or non-NMDA receptors [94,145,149,150], nor by blocking voltage-dependent Ca²⁺ or Na⁺ channels [145,149,150,155]. Results supporting views that DCD is comprised of an active Ca²⁺ influx pathway [93,146,149,150,155–159] as well as those indicating a failure in Ca²⁺ efflux mechanisms [160–162], are available in the literature. It is anticipated that these seemingly opposing observations represent two-facets of the same problem: even in the earliest report on DCD by Manev and colleagues [149] it was shown that during the postglutamate period neurons still accumulate ⁴⁵Ca²⁺ within 30 s exposure to the isotope, without any statistically significant difference seen in the presence or absence of N-methyl-d-aspartate receptors/non-N-methyl-d-aspartate receptors/voltage dependent Ca²⁺ channels (NMDAR/non-NMDAR/VDCC blockers). That attests to the presence of a discrete pathway for Ca²⁺ influx. Yet, it was recently demonstrated that in an almost identical paradigm of excitotoxicity, the plasmalemmal Na⁺/Ca²⁺ exchanger (in particular the NCX3 isoform) is cleaved by calpain, severing the high capacity Ca²⁺ efflux pathway in neurons [161]. Provided that the Ca²⁺ influx pathway is most likely a channel, it must saturate [163] imposing a continuous load of calcium to the neuron. The turning point upon which the cell looses the ability to buffer the incoming calcium resulting in an abrupt, sustained and irreversible increase in [Ca²⁺]_c, probably coincides with the cleavage of the exchanger (but see [164]). Therefore, inhibition of the, as yet unidentified, Ca²⁺ influx pathway or prevention of NCX proteolysis should thwart DCD. The question arises: what is the nature of the Ca²⁺ influx pathway?

Non-selective cationic channel(s) and the DCD

As mentioned above, inhibition of NMDAR/non-NMDAR/voltage-dependent Ca²⁺ or Na⁺ channels after the initial Ca²⁺ and Na²⁺ influx through the glutamate receptors, failed to prevent DCD. Yet, DCD demands the existence of a discrete pathway as it precedes, and eventually leads to, plasma membrane leakiness and cell death [145,146,148]. The notion that DCD is not attributed to the ‘traditionally’ recognized Ca²⁺ channels, such as glutamate receptor-operated or voltage-gated Ca²⁺ channels has been proposed previously [157,158]. Along this line, it was shown that a secondary activation of a nonselective cation conductance, termed postexposure current (I_pe), is induced subsequent to excitotoxic application of NMDA to hippocampal neurons that probably contributes to the delayed Ca²⁺ rise [156].

Relevant to the inability of the glutamate receptor blockers to prevent DCD, antiexcitotoxic therapy utilizing these compounds failed to produce a better outcome in clinical trials concerning stroke treatment [165–167]. To address this setback, Aarts and colleagues [159] examined the possibility that an overlooked neurotoxic process was occurring in a well-established in vitro model of excitotoxicity, by subjecting cultured neurons to oxygen–glucose deprivation. This treatment results in neuronal demise through NMDAR activation [168,169]. It was found that a member of the melastatin branch of the transient receptor potential channel (TRP) family, TRPM7 [170], mediates a lethal cation current loading the neurons with Ca²⁺ and Na⁺. This nonselective current was activated by ROS and reactive nitrogen species (RNS), and its abolition permitted the survival of neurons previously destined to die from prolonged anoxia, regardless of the presence or absence of NMDAR blockers.

In a subsequent study, we explored the hypothesis that a TRP channel contributes to the manifestation of DCD [93]. A pharmacological approach was used, applying 2-aminoethoxydiphenyl borate (2-APB) or La³⁺ to cultured cortical neurons challenged by prolonged glutamatergic stimulation. We observed that 2-APB and La³⁺ diminished the delayed Ca²⁺ rise with a 50% inhibitory concentration of 62 ± 9 µm and 7.2 ± 3 µm, respectively. Both substances are known to inhibit TRP channels in addition to acting on many other targets; 2-APB blocks store-operated Ca²⁺ (SOC) channels [171], the IP₃ receptor [172], the sarco-endoplasmic reticulum Ca²⁺ ATPase (SERCA) pump [173], voltage-dependent K⁺ channels [174], gap junctions [175] and the cyclosporin A-insensitive PTP [104], while La³⁺ blocks SOC [176] and voltage-dependent Ca²⁺ channels [177]. Almost all non-TRP targets are irrelevant or have been previously excluded concerning the origin of DCD, except for the cyclosporin A-insensitive PTP that is abolished by 2-APB in isolated brain mitochondria [104]. However, in our hands, bongkrekic acid ameliorated the cyclosporin A-insensitive PTP but not the DCD [93,104]. From this study we concluded that a TRP channel could be responsible for the Ca²⁺ influx part of DCD. In general, the two inhibitors that we used do not distinguish among individual members of the TRP family, but for reasons explained below, it is tempting to speculate that it is the TRPM7. Unfortunately, we could not achieve silencing of TRPM7 expression in our cultures with short interfering RNA (siRNA); primary neurons are notoriously vulnerable to transfection techniques, as opposed to the ease and the high efficiency of the procedure in cell lines. Hopefully, the development of novel approaches such as the conjugation of siRNA to penetratins [178,179] will assist transfection protocols and allow research on primary neuronal cultures to benefit from the tremendous potential of siRNA.

The connection of TRPM7 to DCD may lie in the observation that this channel is activated by ROS and RNS [159]. For a long time, ROS were considered to be responsible for DCD [180]; however, in a recent study it was deduced that the increased ROS production is a consequence, rather than a cause of DCD [181]. In the latter study the authors also demonstrated that the increase in superoxide radical formation is predominantly associated with extramitochondrial phospholipase A(2) (PLA₂) activation, and it does not emanate from mitochondria. That may be in contrast with previous reports claiming that ROS are the inducers of DCD. However over the years concerns have arisen as for the reliability of ROS-detecting dyes, given that some are affected by confounding parameters such as mitochondrial membrane potential (see discussion in [181]). The development of new dyes described recently will no doubt contribute to the clarification of these matters [182].

In light of the recent observations though, one could argue that TRPM7 is not the Ca²⁺ influx pathway of DCD, as the increase in superoxide radical appears after the secondary [Ca²⁺]_c rise. However, the exact species activating TRPM7 is not known, and the extent of ROS production necessary to activate the channel maybe less than the detection level of the probes used. In addition, ROS/RNS could be just one of the many activators of the channel [183], while others that might play a significant role could be also mobilized upon prolonged glutamate exposure. We have found that by elevating intracellular [Mg²⁺]_i DCD is abolished in cultured cortical neurons [93], and it is known that TRPM7 receives strong negative feedback by intracellular Mg²⁺[170]. In addition, TRPM7 currents induced by oxygen–glucose deprivation promote further ROS production [159], and this could partially explain the results of Vesce and colleagues, detecting an increase in superoxide formation after the delayed secondary [Ca²⁺]_c rise [181]. In our opinion, TRPM7 is one of the best possible candidates for the Ca²⁺ influx part of DCD; other good candidates are TRPM2 (see below) and the calcium-permeable acid-sensing ion channel [184] (not reviewed here).

Nonselective cationic channels and the ’Ca²⁺ paradox’

In spite of the widely accepted role of [Ca²⁺]_c deregulation in the manifestation of neurodegeneration, exactly how Ca²⁺ ions mediate neural cell death is less clear [185]. One of the most important unresolved issues is the mechanism by which [Ca²⁺]_c increases to excessively high levels in neurons following periods of intense neuronal activation. Reaching further from the possibility of the involvement of TRP channels in the delayed calcium deregulation, these proteins could participate in an additional overlooked pathway of Ca²⁺ influx that may pertain during ischemia/reperfusion or other type of pathology. Large [Ca²⁺]_c increases are known to be triggered by reintroduction of ‘normal’ Ca²⁺ concentrations to the extracellular milieu after the tissue has experienced a [Ca²⁺]_e-free challenge, or at least a severe reduction in extracellular calcium concentration, termed ‘Ca²⁺ paradox’. The free extracellular calcium concentration falls dramatically in several brain disease states: (a) during or after ischemia (0.1–0.28 mm[186–189]); (b) traumatic brain injury (0.1 mm[190]); (c) severe hypoglycemia (0.12 mm[191]); and (d) spreading depression (0.06–0.08 mm[192]). Reduction of extracellular Ca²⁺ is mostly due to robust influx of the cation to the intracellular milieu, although the appearance of lactate in the interstitium during ischemia, with the ability to chelate divalent ions significantly, also plays a role [193,194].

Mitochondrial permeabilization and a possible link to TRP channel activation

Among the known activators of some members of the TRP family, NAD⁺ and its catabolite ADP-ribose (ADPR) were described to activate TRPM2 [244–247], in addition to the fact that the channel is stimulated by ROS/RNS [236,238,246]. Furthermore, it was demonstrated that the major source of free ADPR mediating the activation of TRPM2 in cultured cells were the mitochondria [248]. One could link these observations to the fact that opening of the PTP causes the release of mitochondrial NAD⁺ followed by its hydrolysis by an extramitochondrial NAD⁺ glycohydrolase to ADPR [103,249]. It is tempting to speculate that this ADPR in conjunction with ROS produced upon loss of mitochondrial integrity, activates the nonselective TRPM2 allowing a large Ca²⁺ and Na⁺ load to enter the cytosol. Since both high [Ca²⁺]_c and ROS promote mitochondrial pore formation, it seems that the order of appearance of a pore or TRPM2 activation is trivial; what is probably more important is that activation of the one can lead to activation of the other, completing a vicious cycle. Intriguingly, silencing the expression of TRPM7 with siRNA, led to an accompanying decrease in TRPM2 expression. This suggests that the two transcripts might be coordinately regulated, raising the possibility that a fraction of the oxygen–glucose deprivation-induced current recorded earlier [159] is mediated by TRPM2 or TRPM7 heteromultimers, a structural arrangement commonly occurring among TRP channels [250,251]. Further implications of TRP channels in relation to the overall metabolic state of the cell in hypoxia have been reviewed elsewhere [252].