|Home | About | Journals | Submit | Contact Us | Français|
A prolonged period of ischaemia followed by reperfusion irreversibly damages the heart. Such reperfusion injury (RI) involves opening of the mitochondrial permeability transition pore (MPTP) under the conditions of calcium overload and oxidative stress that accompany reperfusion. Protection from MPTP opening and hence RI can be mediated by ischaemic preconditioning (IP) where the prolonged ischaemic period is preceded by one or more brief (2–5 min) cycles of ischaemia and reperfusion. Following a brief overview of the molecular characterisation and regulation of the MPTP, the proposed mechanisms by which IP reduces pore opening are reviewed including the potential roles for reactive oxygen species (ROS), protein kinase cascades, and mitochondrial potassium channels. It is proposed that IP-mediated inhibition of MPTP opening at reperfusion does not involve direct phosphorylation of mitochondrial proteins, but rather reflects diminished oxidative stress during prolonged ischaemia and reperfusion. This causes less oxidation of critical thiol groups on the MPTP that are known to sensitise pore opening to calcium. The mechanisms by which ROS levels are decreased in the IP hearts during prolonged ischaemia and reperfusion are not known, but appear to require activation of protein kinase Cε, either by receptor-mediated events or through transient increases in ROS during the IP protocol. Other signalling pathways may show cross-talk with this primary mechanism, but we suggest that a role for mitochondrial potassium channels is unlikely. The evidence for their activity in isolated mitochondria and cardiac myocytes is reviewed and the lack of specificity of the pharmacological agents used to implicate them in IP is noted. Some K+ channel openers uncouple mitochondria and others inhibit respiratory chain complexes, and their ability to produce ROS and precondition hearts is mimicked by bona fide uncouplers and respiratory chain inhibitors. IP may also provide continuing protection during reperfusion by preventing a cascade of MPTP-induced ROS production followed by further MPTP opening. This phase of protection may involve survival kinase pathways such as Akt and glycogen synthase kinase 3 (GSK3) either increasing ROS removal or reducing mitochondrial ROS production.
The heart is one of the most energy demanding tissues in the body and is totally dependent upon oxidative phosphorylation to supply the large amount of ATP required for beat-by-beat contraction and relaxation. If the blood flow to the heart is impaired (ischaemia), as occurs when a blood clot occludes a coronary artery (coronary thrombosis) or where cardiac surgery requires the heart to be disconnected from the blood supply, the source of oxygen is removed leading to the cessation of oxidative phosphorylation. This causes tissue ATP and creatine phosphate concentrations to decrease with a concomitant rise in ADP, AMP and Pi concentrations. Although glycolysis is activated, it is unable to meet the demand of the beating heart for ATP. Consequently, the heart rapidly ceases to beat as the contractile machinery is inhibited by elevated Pi and ADP, combined with the decreasing pH that accompanies the accumulation of glycolytic lactic acid [1–3]. The heart can usually survive a short period of ischaemia and then recover upon reperfusion. Although the performance of the heart may be impaired initially (stunning), given time recovery is complete. However, if the period of ischaemia is too long, the tissue becomes irreversibly damaged. Hence, if the heart is to be salvaged, it is important to restore the blood flow as soon as possible. Yet, paradoxically, such reperfusion can exacerbate the damage occurring during the ischaemic period. This is known as reperfusion injury and is accompanied by enzyme release and morphological changes characteristic of necrosis [1–3]. The extent of damage can be visualised as an area of necrotic tissue known as the infarct whose area can be determined to provide a quantitative measure of injury. Quantification of damage may also be provided by measuring the release of intracellular proteins such as lactate dehydrogenase or troponin I . In addition to the necrotic cell death that represents the major damage to the reperfused heart there is also evidence that some myocytes around the periphery of the infarct die by apoptosis [5,6]. Understanding the causes of reperfusion injury and devising ways of preventing it is of major clinical importance in cardiac surgery and the treatment of coronary thrombosis.
There is increasing evidence that mitochondrial dysfunction plays a central role in mediating both the necrotic and apoptotic components of reperfusion injury, and that one of the most effective ways of protecting hearts from such injury, known as ischaemic preconditioning (IP), acts to attenuate this [1,3,7–10]. This review will address the causes of reperfusion injury, emphasising the role of the mitochondrial permeability transition pore (MPTP) and prevention of its opening by IP. The proposed signalling mechanisms through which IP may exert its effects will be discussed, including the proposed role of mitochondrial potassium channels which has been a major but controversial area of research (see [11–13]. However, first it is necessary to explain the phenomenon of preconditioning.
Ischaemic preconditioning (IP) involves exposing the heart to brief periods of ischaemia (typically 2–5 min) interspersed with periods of normal perfusion prior to the prolonged ischaemia initiated within an hour of the preconditioning protocol. This protocol was first shown to offer strong protection against reperfusion injury in the dog heart by Murry et al. in 1986  and has since been confirmed in all species investigated, including humans [15–17]. However, it should be noted that significant reduction in infarct size only occurs when the optimal duration for the test period of ischemia is chosen and is lost if the ischaemic period is too long. Preconditioned hearts exhibit a smaller infarct size and intracellular enzyme release (indicators of necrotic cell death) and fewer arrhythmias [14,18], whilst contractile function is preserved . Typically, 2–3 brief (2–5 min) cycles of ischaemia and reperfusion are used in IP, there being little advantage of additional cycles . If ischaemia is initiated more than 1–2 h after the preconditioning protocol, protection is lost but re-emerges again after about 24 h and lasts for up to 3 days. This is termed the second window of preconditioning [15,20–22]. A wide variety of drugs that activate different signalling pathways thought to be involved in mediating IP (see Section 6) can also induce protection and thus preconditioning is often used as a generic term to encompass any protocol applied before prolonged ischaemia that protects the heart during reperfusion. Not to be confused with such preconditioning is a more recently described phenomenon known as post-conditioning. Here hearts are given several very brief ischaemic interludes (10 s) during the early phase of reperfusion which also improves the recovery of the heart and reduces infarct size [23–25]. This protocol has considerable clinical potential since intermittent reperfusion can be induced during angioplasty prior to permanent restoration of blood flow. It is beyond the scope of this review to consider the similarities and differences between preconditioning and postconditioning in any detail. However, many of the mechanisms and signalling pathways mediating protection are thought to be the same  and where effects on mitochondria have been demonstrated in postconditioning they will be noted.
Increases in cellular [Ca2+] and reactive oxygen species (ROS), initiated in ischaemia and then amplified upon reperfusion, are thought to be the main causes of reperfusion injury. Mitochondria are involved both in the production of ROS and as targets for the damaging action of both ROS and calcium [1,3,9,26].
During ischaemia, the increased rate of glycolysis causes lactic acid to accumulate and the intracellular pH (pHi) to drop rapidly. This activates the Na+/H+ antiporter as the cell endeavours to restore pHi. However, the rapidly declining ATP concentrations cause inhibition of the Na/K ATPase and lead to a rise in intracellular [Na+]. This in turn reduces the ability of the cell to restore its pHi and increases intracellular [Ca2+] since the Na+/Ca2+ antiporter that usually pumps Ca2+ out of the cell, is inhibited or reversed. The conversion of ATP to ADP and AMP is rapid and reversible. AMP is slowly converted into adenosine and then inosine and xanthine through a purine degradation pathway. These nucleosides leak out of the cell (and may have vasodilator effects through purinergic receptors) and lead to a gradual depletion of adenine nucleotide which may contribute to the reduced cardiac performance (stunning) seen on reperfusion.
Another feature of ischaemia that is thought to contribute to the damage seen during reperfusion is the production of reactive oxygen species (ROS). There is a small increase in ROS production immediately the heart becomes ischaemic which remains stable for 20–25 min but then increases dramatically . This large increase is thought to play a major role in damaging the heart during ischaemia and sensitising it to reperfusion as described below. However the source of the ROS is unclear and might involve complex 1 and complex 3 of the respiratory chain or perhaps more likely xanthine oxidase acting on xanthine formed from the degradation of adenosine as noted above [26,28–31]. The depletion of ATP combined with elevated [Ca2+] and ROS leads to a gradual decline in cellular integrity as degradative enzyzmes are activated and ATP-dependent repair processes are unable to operate [1,3,26,32]. If the tissue remains ischaemic for only short periods and the mitochondria remain sufficiently intact to generate the ATP, tissue damage is slight and can be repaired by ATP-dependent processes upon reoxygenation. However, a critical point is reached when recovery is not possible and reperfusion actually causes further damage to the heart, causing extensive necrosis with associated enzyme release [1,3].
Reperfusion is associated with a burst of reactive oxygen species (ROS) production , but here too the source of the ROS is debated. Although some may be produced by xanthine oxidase and NADPH oxidase, it is probable that most is formed by complex 1 and complex 3 of the respiratory chain [26,28–30]. When the respiratory chain is inhibited by lack of oxygen and then re-exposed to oxygen, ubiquinone can become partially reduced to ubisemiquinone. This can then react with the oxygen to produce superoxide that is reduced to hydrogen peroxide by superoxide dismutase. Hydrogen peroxide is removed by glutathione peroxidase or catalase, but if ferrous ions (or other transition metals such as copper) are present it will form the highly reactive hydroxyl radical through the Fenton reaction . Mitochondrial proteins are especially susceptible to ROS induced damage and this is reflected in the impaired respiratory chain activity of mitochondria isolated from ischaemic hearts (see [26,31]). Thus ROS have direct effects on several respiratory chain components, most especially on complex 1 but also on complex 3, and other iron sulphur proteins such as aconitase. ROS can also cause thiol oxidation and inhibition of the ATPase and adenine nucleotide translocase. In addition, ROS cause oxidation of glutathione that may then form mixed disulphides with proteins. Such protein modification is thought to have inhibitory effects on ion pumps and therefore exacerbate the effects of ATP deprivation on ionic homeostasis [33–35]. ROS also cause peroxidation of the unsaturated fatty acid components of the phospholipids, and especially cardiolipin of the inner mitochondrial membrane, and this leads to further inhibition of respiratory chain activity [36,37]. Furthermore, lipid peroxidation causes the release of reactive aldehydes such as 4-hydroxynonenal that can modify membrane proteins . Overall, it is thought that the combined effects of ROS and elevated [Ca2+] play a critical role in the transition from reversible to irreversible reperfusion injury, and that mitochondria are the major target of these agents. In particular, they lead to the opening of the mitochondrial permeability transition pore, that is now widely accepted to play a critical role in reperfusion injury [1,3,10,26,39].
Under normal physiological conditions, the mitochondrial inner membrane is impermeable to all but a few selected metabolites and ions; this is essential to maintain the membrane potential and pH gradient that together drive ATP synthesis through oxidative phosphorylation. However, under conditions of high matrix calcium, especially when this is accompanied by oxidative stress, high phosphate and low adenine nucleotide concentrations, a non-specific pore opens in the inner mitochondrial membrane known as the mitochondrial permeability transition pore (MPTP). The properties and regulation of the MPTP are reviewed extensively elsewhere [3,40,41]. Once open, this pore allows free passage of any molecule of < 1.5 kDa and so disrupts the permeability barrier of the inner membrane. This has two major consequences. First, it allows unrestricted proton movement across the inner membrane, causing oxidative phosphorylation to be uncoupled. Not only does this prevent ATP synthesis but it also enables the proton-translocating ATPase to reverse direction and so actively hydrolyse ATP rather than synthesise it. Under such conditions, intracellular ATP concentrations rapidly decline, leading to the disruption of ionic and metabolic homeostasis and activation of degradative enzymes such as phospholipases, nucleases and proteases [3,9,26]. Unless pore closure occurs, these changes will cause irreversible damage to the cell resulting in necrotic death.
It should be noted that opening of just a single pore in one mitochondrion is likely to cause its immediate depolarisation. This will then activate further pore opening in the same mitochondrion since MPTP opening of calcium loaded mitochondria is activated by depolarisation . Thus mitochondria are either fully open or closed, and it is the fully open state that leads to the second consequence of MPTP opening; the mitochondria swell extensively . This occurs as all small molecular weight solutes equilibrate across the inner membrane, leaving the high concentration of matrix proteins to exert a colloidal osmotic pressure that leads to the uptake of water and matrix swelling. Although unfolding of the cristae allows the matrix to expand without rupture of the inner membrane, the outer membrane will break and lead to the release of proteins in the intermembrane space such as cytochrome c and other factors that play a critical role in apoptotic cell death [44,45].
The molecular identity of the mitochondrial permeability transition pore remains uncertain [40,41,46], but it is generally accepted that an inner membrane component undergoes a calcium-triggered change in conformation that is facilitated by cyclophilin D (CyP-D), a peptidyl-prolyl cis-trans isomerase [46,47]. The role of CyP-D was first suggested by the discovery that cyclosporin A (CsA) acts as a potent inhibitor of pore opening . Further studies revealed that the potency of different CsA analogues to inhibit pore opening correlates with their ability to inhibit the peptidyl-prolyl cis-trans isomerase activity within the matrix [49,50] that was subsequently identified as CyP-D [51,52]. Extensive work from many laboratories confirmed the critical role of CyP-D and this was finally put beyond doubt by the demonstration that MPTP opening in liver mitochondria from CyP-D knockout mice is much less sensitive to calcium than normal mitochondria, and is no longer inhibited by CsA [53–55].
The identity of the membrane component of the MPTP is less certain. However, the most widely accepted view is that the adenine nucleotide translocase (ANT) normally fulfils this role and extensive circumstantial data supports this view (see [9,56]). Thus opening of the MPTP is inhibited by adenine nucleotides with a similar concentration dependence and specificity as they exhibit when acting as substrates for the ANT, and this inhibition is overcome by the specific inhibitor of the ANT, carboxyatractyloside (CAT) that traps the ANT in its “c” conformation. By contrast, another inhibitor of the ANT, bongkrekic acid, that causes the carrier to take up the alternative “m” conformation, inhibits pore opening . The ANT can also account for the sensitisation of the MPTP to calcium by oxidative stress and the vicinal thiol reagent phenylarsine oxide (PAO) . Thus cysteine residues 160 and 260 of rat ANT2 can be cross-linked by oxidative stress or PAO, with modification of Cys160 alone being sufficient to prevent the inhibition of MPTP opening by adenine nucleotides, so stimulating pore opening . Further evidence for an important role for the ANT is the ability of the ANT to bind to CyP-D in a CsA-sensitive manner [58,59]. In addition, when the purified ANT is reconstituted into proteoliposomes high calcium concentrations can induce the formation of non-specific channels  and this process is sensitised to [Ca2+] by the addition of purified cyclophilin . Nevertheless, despite the strong evidence in favour of the ANT being the critical membrane component of the MPTP, it is unlikely to be essential. Thus in an elegant study, that has yet to be confirmed by others, mitochondria from mouse livers in which ANT1 and ANT2 had been knocked out were found to exhibit MPTP opening that was inhibited by CsA . However, pore opening in the ANT-knockout mitochondria required much higher calcium concentrations than did wild-type mitochondria, and was no longer sensitive to adenine nucleotides confirming that the ANT is at the very least playing a critical regulatory role. One possibility is that the ANT is the normal membrane component of the MPTP but that other less abundant members of the mitochondrial carrier family can fulfil this role in the absence of the ANT . An alternative possibility, proposed by Lemasters, is that unfolded membrane proteins form the MPTP and that the ANT is the most abundant of these .
Many other proteins have been proposed to be components of the MPTP, including the peripheral benzenediazipine receptor, creatine kinase and the voltage dependent anion channel (VDAC) [3,40,41,65]. Of these perhaps the strongest candidate is VDAC since an interaction between VDAC and the ANT has been demonstrated and there is evidence that the MPTP may be associated with contact sites between the inner and outer mitochondrial membranes [3,40,65]. Furthermore, initial reports suggested that VDAC might be the locus for inhibition by ubiquinone analogues . However, the recent demonstration that liver mitochondria from mice lacking VDAC1 exhibit normal MPTP opening and inhibition by ubiquinone analogues  make an essential role for VDAC1 unlikely. See note added in proof.
We have developed a technique to demonstrate that the MPTP is kept firmly closed in the heart under normal physiological conditions, but opens upon reperfusion following a period of ischaemia. This technique involves measuring the extent of [3H]-2-deoxyglucose entrapment in mitochondria that have undergone the permeability transition . Others have used fluorescence microscopy to measure MPTP opening in isolated cardiac myocytes and have also confirmed that the pore opens under conditions of simulated ischaemia and reperfusion . It might be expected that the pore would also open after prolonged ischaemia and some have reported this to be the case [70,71]. However, these studies relied on cytochrome c release as a measure of MPTP opening, but it is known that this can occur independently of the MPTP as a result of Bax translocation to mitochondria during ischaemia . By contrast, our own direct measurements  and those of others  do not support MPTP opening in ischaemia. Since the pore has been shown to be powerfully inhibited by low pH (< 7)  and the time course of opening during reperfusion correlates with the return of pH to normal , we have suggested that it is the low pH accompanying ischaemia that prevents pore opening despite the conditions of oxidative stress, low adenine nucleotide and elevated phosphate and calcium . Indeed, as would be predicted, maintaining a low pH during the initial phase of reperfusion is known to protect hearts from reperfusion injury [76–78] and may play a role in the cardioprotective effects of inhibitors of the Na+/H+ exchanger such as amelioride and cariporide (see [3,79,80]). A similar mechanism has also been implicated recently in the protection offered by postconditioning [81,82].
It has also been proposed that the rapid energisation of mitochondria at reperfusion will lead to electrogenic calcium uptake into the mitochondria. Since the calcium has accumulated in the cytosol during ischaemia, this might be expected to induce mitochondrial calcium overload and hence MPTP opening [1,9]. However, measurements of mitochondrial [Ca2+] in isolated cardiac myocytes imply that it is the mitochondrial [Ca2+] at the end of ischaemia rather than during reperfusion that correlates better with cell injury . Another factor that will reinforce MPTP opening at reperfusion is the surge of ROS production that is known to occur [27,84] and this may well be the most important factor in determining the outcome of reperfusion . Indeed, there is increasing evidence that the extent of MPTP opening is a critical factor in the transition from reversible to irreversible reperfusion injury [1,3,9,65]. More accurately, it is the extent to which pores remain in an open state that correlates with damage, since we have demonstrated that some mitochondria in which the MPTP opens initially, subsequently undergo pore closer as reperfusion continues [75,85]. It is possible that myocytes in which the majority of mitochondria have undergone such reversible MPTP opening will have swollen sufficiently to release pro-apoptotic proteins such as cytochrome c and thus initiate apoptosis as noted in Section 4.1. These myocytes are likely to be at the periphery of the infarct where the ischaemic insult was incomplete and thus would account for the ring of apoptotic myocytes that have been observed to surround the necrotic core of the infarct [5,6]. It should be noted that mice in which CyP-D had been knocked out demonstrate normal development and cells from them respond to a range of apoptotic stimuli in the same manner as wild-type mice [54,55]. Thus MPTP opening cannot the mechanism by which apoptosis is mediated in the healthy animal, but this does not preclude apoptosis being induced in this way under pathophysiological situations such as those described above .
If the opening of the MPTP is a critical event in reperfusion injury then it would be predicted that inhibition of pore opening should protect hearts from injury. Many studies have confirmed that this is the case. Protection was first demonstrated with CsA which, when added prior to global ischaemia, was shown to improve haemodynamic function and ATP levels on reperfusion and to decrease necrotic damage as reflected in intracellular enzyme release . Subsequent studies have shown that other CsA analogues, and the alternative CyP-D inhibitor sanglifehrin A (SfA) also provide protection under these conditions [68,87,88] whilst mitochondria isolated from these hearts showed less pore opening in response to a given calcium load [88,89]. In addition, CsA and SfA can reduce the infarct size of hearts in which a coronary artery is occluded and then re-opened to mimic the treatment of a coronary thrombosis. In this model of reperfusion injury, protection is observed even when the drug is added only at reperfusion [90,91].
Less direct approaches to inhibit MPTP opening are also effective at protecting the heart from reperfusion injury. Thus free radical scavengers are well known to be protective and in the case of one such agent, the anaesthetic propofol, we have used the 2-deoxyglucose entrapment technique to confirm inhibition of the MPTP directly . In addition, we have demonstrated that propofol provides protection from reperfusion injury in a pig model of open heart surgery . Inhibition of the sodium proton exchanger with cariporide is also known to protect the heart from reperfusion injury  and here too inhibition of MPTP opening correlates with this protection . This protection probably reflects less sodium and hence calcium loading of the cardiac myocytes during ischaemia, coupled with a lower pH during the initial phase of reperfusion, both of which will reduce MPTP opening. The most powerful agent for inhibiting MPTP opening and protecting the heart is pyruvate [75,95]. The presence of 10 mM pyruvate during ischaemia and reperfusion can almost totally protect rat hearts from 30 min global ischaemia and this is accompanied by greatly reduced MPTP opening followed by total pore closure . Pyruvate has three beneficial effects that may contribute towards its inhibition of MPTP opening: it is a free radical scavenger, it maintains a lower intracellular pH during the early phase of reperfusion and it is an excellent fuel for ATP synthesis that bypasses the ATP-requiring steps of glycolysis [75,96].
In view of the critical role of the MPTP in reperfusion injury and the proven ability of inhibitors of the MPTP to protect from injury it might be predicted that ischaemic preconditioning (IP) would inhibit pore opening. Both direct and indirect techniques have been used to confirm this. Our own experiments have used the 2-deoxyglucose entrapment technique to demonstrate that less MPTP opening occurs at reperfusion in preconditioned hearts and that as reperfusion continues the majority of mitochondria that originally opened close again in the IP hearts but not in control hearts . Others have used confocal microscopy to monitor mitochondrial membrane potential in an isolated cardiac myocyte model of preconditioning and shown that IP desensitises the mitochondria to pore opening induced by oxidative stress [69,97]. Recent experiments performed with two-photon microscopy in the perfused heart have confirmed these data . Yet other studies have investigated the sensitivity of isolated mitochondria from control and IP hearts to pore opening in response to added calcium [85,99,100]. Both light scattering and mitochondrial calcium retention techniques demonstrated that when mitochondria were isolated immediately after the preconditioning stimulus, those from IP hearts were not any less sensitive to calcium-induced pore opening than those from control hearts. Indeed, if anything they were more sensitive [85,100]. When isolated at the end of ischaemia or during the first few minutes of reperfusion mitochondria were more sensitive to MPTP opening than those isolated prior to ischaemia but at these time points mitochondria from IP hearts were considerably less sensitive to calcium-activated pore opening than those from control hearts [99–101]. These data clearly show that the IP protocol alone is not having a direct effect on the MPTP but suggest that IP inhibits a process occurring during ischaemia and reperfusion that is responsible for sensitising the MPTP to calcium. This is an important fact to bear in mind when evaluating the extensive literature surrounding the mechanisms by which preconditioning exerts its protective effects as will be discussed further below (Section 6).
Temperature preconditioning, where hearts are exposed to three brief hypothermic (26 °C) episodes prior to normothermic ischaemia is also thought to protect through inhibition of the MPTP  as is post-conditioning . Thus, mitochondria isolated from such hearts during reperfusion are less sensitive to MPTP than those isolated from control hearts with temperature preconditioning being even more protective than IP .
There is a plethora of data implicating many different signalling pathways in preconditioning and the relevant role of each remains hotly debated (see [16,30,103–107]). Many of these studies have utilised pharmacological agonists and antagonists of components of the different signalling kinase cascades and their potential targets. The complex interactions that exist between different signalling pathways makes interpretation of these data complicated and this is made worse by the questionable specificity of some of the pharmacological interventions. The reader is referred elsewhere for a detailed consideration of the many potential pathways that have been proposed [30,103–107]. Here we will only provide a brief overview before focussing on how they may exert their effects on the mitochondria.
There is extensive evidence that protein kinase C (PKC) plays a central role in preconditioning since inhibition of PKC has been shown to block the protection afforded by IP and pharmacological activators of PKC are cardioprotective (see ). There remains some controversy over which of the many PKC isoforms may be involved in IP, whether they translocate to the particulate fraction and how they exert their effects . Nevertheless, there is a large body of evidence to implicate PKCε as an important player in IP . Thus hearts from PKCε knockout mice do not exhibit IP  whereas transgenic mice with cardiac-specific over-expression of PKCε or expression of an activator of PKCε are protected from reperfusion injury [109–111]. Some studies have reported PKCε translocation to the particulate fraction, including mitochondria [112–114] and have suggested a direct inhibition of the MPTP by PKCε involving phosphorylation of components of the MPTP such as the voltage dependent anion channel (VDAC) [113–115]. However, in our own studies we were unable to detect PKCε translocation to the mitochondria of IP hearts (Fig. 1A) whilst others have observed that PKC translocation to the particulate fraction following IP is transient and lost during progressive brief cycles of ischaemia and reperfusion .
Although a role for PKCε in IP seems established, the mechanism(s) by which it exerts its effects are less clear as will become apparent below (see sections 7–10). It is also uncertain how ischaemic preconditioning activates PKCε, although several pathways may be involved. Factors released during the brief ischaemic periods such as adenosine, bradykinin, noradrenaline and opioids may bind to their G protein-coupled receptors to stimulate phospholipase C, producing diacylglycerol that activates PKC. Indeed, all of these factors can pharmacologically precondition the heart . Accumulating evidence supports a role for the modest increase in ROS that occurs during IP protocol in the activation of PKC [27,117,118]. Thus ROS are known to activate PKC in the isolated heart [118–120] and IP can be prevented if free radical scavengers are present during the preconditioning phase [100,101,118,121] Oxidation of critical cysteine residues on PKC isoforms is known to cause their activation [35,122] and thus provides a mechanism by which ROS could activate PKC.
Data from several laboratories have implicated activation of cyclic GMP dependent protein kinase (PKG), perhaps by nitric oxide, in the signalling pathway for IP. Thus it has been reported that nitric oxide donors and PKG activators can induce preconditioning whilst NO scavengers and PKG inhibitors prevent preconditioning [106,123,124]. Pharmacological studies from Garlid's laboratory have led them to conclude that PGK and PKCε work in concert to induce IP through an effect on the putative mitochondrial ATP-dependent potassium channel (mitoKATP — see Section 9 below). They propose that cGMP activates PKG localized at the cytosolic surface of the mitochondrial outer membrane and that this phosphorylates some target protein which in turn can somehow activate PKCε residing in the intermembrane space of mitochondria. This PKCε then would phosphorylate the mitoKATP channel to mediate preconditioning . In a subsequent paper they proposed that activation of the mitoKATP channel increased ROS formation and that this ROS activated a second pool of PKCε that inhibited the MPTP . Our own data do not support this (see Sections 7). It should be noted that an alternative or additional mechanism has been proposed by which nitric oxide may protect the heart when present during the initial stages of reperfusion (see ). Cytochrome oxidase is competitively inhibited by nitric oxide and this would lead to an inhibition of respiration and hence a reduction in the mitochondrial membrane potential. It is suggested that one consequence of this would be less mitochondrial calcium accumulation and hence less MPTP opening. However, another consequence of a lower mitochondrial membrane potential would be impaired ATP production which would not be predicted for hearts showing improved recovery. Similar arguments have been made for the proposed depolarisation caused by mitochondrial KATP channel opening and these will be considered in more detail in Section 10.2.
Several laboratories have presented evidence to implicate activation of pro-survival kinases such as Akt (protein kinase B) in preconditioning [107,126]. It is proposed that tyrosine kinases are activated during IP through some ill-defined interaction with G-protein coupled receptors such as the adenosine and bradykinin receptors and that this causes activation of phosphatidyl inositol 3 phosphate kinase (PI-3-kinase). The resulting phosphatidylinositol 3,4-bisphosphate then activates phosphoinositide-dependent kinase 1 (PDK1) that in turn phosphorylates and activates Akt. However, it is also possible that activation of tyrosine kinases or inhibition of tyrosine phosphatases and the phosphatidyl inositol 3,4,5,-trisphosphate phosphatase PTEN (Phosphatase and TENsin homolog) may be mediated by the ROS produced during the preconditioning stimulus [107,127]. In support of this mechanism it has been demonstrated that phosphorylation of Akt is enhanced by IP, especially during reperfusion, and that pharmacological inhibitors of either Akt or PI-3-kinase prevent preconditioning [126,128–130]. It has further been proposed that following phosphorylation, activated Akt phosphorylates endothelial nitric oxide synthase to produces NO that acts via PKG as noted above [106,107].
Sollott et al.  have suggested that all these kinases may converge to phosphorylate and inhibit glycogen synthase kinase 3β (GSK3β) that they report translocates to the mitochondria. Inhibition of GSK3β by Akt-induced phosphorylation would then mediate protection. Although it is known that GSK3β is pro-apoptotic and that its phosphorylation inhibits its activity and stimulates cell survival , the mechanism of any mitochondrial effects remain to be elucidated. However, it may be significant that GSK3β has been reported to phosphorylate VDAC, and that this phosphorylation is associated with a sensitisation to cytotoxic drugs that stimulate pore opening . VDAC was widely regarded as an integral component of the MPTP, but recent data showing that mitochondria from the VDAC1 knockout mouse have a normal MPTP cast doubt on this . See note added in proof.
Yet another protein kinase, AMP-activated protein kinase (AMPK), is activated by a large variety of cellular stresses that deplete ATP and increase AMP, including glucose deprivation, hypoxia, ischaemia, oxidative stress and hyperosmotic stress . It could be argued that the many pharmacological agents known to induce preconditioning, from KATP channel openers, to uncouplers and respiratory chain inhibitors (see Section 10), might all represent insults to the heart that increase AMP and so activate AMPK as the focus of a common signalling mechanism to protect the heart . Indeed, transgenic mice expressing a kinase dead form of AMPK in the heart have been shown to exhibit greater myocardial necrosis and apoptosis after ischaemia/reperfusion  and cannot be preconditioned . Furthermore, Nishino et al.  have shown that PKC inhibitors abolished AMPK activation by ischaemic episodes in the myocardium, suggesting that AMPK might be activated by IP in a PKC-dependent manner. However, our own data fail to show any abrogation of the IP-induced protection by the inhibitor of AMPK, compound C, under conditions where the inhibitor of PKC, chelerythrine, did overcome the protection .
Taken together, all these data suggest that several signalling pathways may interact or act in parallel to induce IP, but the ultimate target of their action remains unclear. Since IP involves inhibition of the MPTP the signalling pathway must ultimately inhibit MPTP opening, but this could be achieved either by direct phosphorylation of a component of the MPTP or indirectly by influencing factors that enhance pore opening such as by reducing oxidative stress or calcium overload.
As noted above, there are several reports that protein kinases may migrate to the mitochondria in response to IP suggesting that phosphorylation of a component of the MPTP may be responsible for its inhibition. Baines et al. have reported that the MPTP of isolated heart mitochondria was inhibited following incubation for 15 min with purified PKCε plus phorbol ester . However, such experiments are hard to interpret since incubating isolated mitochondria in vitro can change the sensitivity of MPTP opening to calcium, especially if ATP is added since this is an inhibitor of MPTP opening in its own right [3,40]. Similar criticisms could be levelled against experiments in which incubation of isolated rat liver mitochondria with a cGMP analogue in the presence of a cytosolic extract and ATP caused inhibition of the calcium-induced mitochondrial permeability transition , although here reversal of the effect by the PKG inhibitor KT5823 makes the data more convincing. However, no data are available to show changes in phosphorylation of any mitochondrial protein following IP that might account for inhibition of the MPTP. Furthermore, when we and others measured MPTP opening in mitochondria isolated immediately after the preconditioning stimulus we observed no reduction in sensitivity to calcium as might have been predicted if a component of the MPTP had been phosphorylated [85,100]. Indeed, it is unclear how activated cytosolic protein kinases could cross the outer membrane to phosphorylate and inhibit components of the MPTP, let alone cross the inner membrane into the matrix. Although there are reports to suggest that this occurs [97,113,114], our own attempts to demonstrate the translocation of PKCε or GSK3β to mitochondria have consistently failed as illustrated in Fig. 1. A key aspect of our experiments was the use of Percoll gradient centrifugation to remove plasma membrane contamination which we have shown is significant in less pure mitochondrial fractions. This was demonstrated by the presence of monocarboxylate transporter 1 (MCT1) in the crude but not the purified mitochondria.
Our inability to detect protein kinase translocation into the mitochondrial matrix is entirely consistent with extensive work from this laboratory more than 20 years ago in which high specific activity 32P labelling was used to investigate the phosphorylation of mitochondrial matrix proteins. Such studies demonstrated the presence of two proteins within the mitochondrial matrix that could be phosphorylated in Percoll-purified mitochondria from both heart and liver. These were pyruvate dehydrogenase (PDH) and branched chain 2-oxoacid dehydrogenase (BCDH) . This was true whether the mitochondria were from control or glucagon-treated (raised cyclic AMP) rats , yet mitochondria from glucagon-treated rats were found to exhibit enhanced calcium retention characteristic of reduced sensitivity of the MPTP to calcium-induced opening [140,141].
Although these data do not support a role for protein kinases within the mitochondrial matrix in the inhibition of the MPTP following preconditioning, they do not exclude the possibility of regulation by phosphorylation of proteins exposed to cytosolic kinases, such as those in the outer mitochondrial membrane or bound to it using scaffolding proteins [142,143]. Indeed, work from this laboratory many years ago demonstrated that when mitochondria were isolated from hepatocytes incubated with high specific activity 32P, or isolated mitochondria were incubated with γ-[32P]-ATP, additional phosphorylated proteins were observed. Two of these proteins, both 30–35 kDa, demonstrated an increase in phosphorylation with glucagon treatment [139,144]. However, a weakness of the 32P labelling technique is that it will not detected proteins whose phosphorylation turnover is small and this may account for the larger number of phosphorylated proteins within mitochondria detected using a phosphoprotein-specific dye . Nevertheless, a lack of turnover might suggest that such a phosphorylation does not play an important role in signalling.
There have been an increasing number or reports in which proteomic and immunodetection methods have been used to detect a variety of other kinases associated with the mitochondria [142,146]. However, the extent to which these represent real mitochondrial kinases rather than non-specific mitochondrial contamination of the mitochondrial fraction is hard to assess. We have recently used a proteomics approach to identify any mitochondrial proteins whose phosphorylation might be altered by preconditioning. Mitochondria were isolated from control and IP hearts immediately after the preconditioning stimulus as well as the end of ischaemia and after three min reperfusion and phosphor-proteins detected using 2-D gel electrophoresis and staining with ProQ Diamond phospho-protein fluorescent dye. Although a significant number of phospho-proteins were detected in addition to the predominant spots representing PDH and BCDH, no consistent changes were detected in any of these in response to IP (unpublished data of S.J. Clarke, I. Khaliulin, Kate Heesom and A.P. Halestrap). Typical data are shown in Fig. 2. Overall, our data do not support a role for mitochondrial protein phosphorylation in mediating the inhibition of MPTP opening by IP. Rather, our data suggests that the inhibition of the MPTP by preconditioning may be secondary to either diminished ROS production or calcium overload as described below.
As noted above, opening of the MPTP is triggered by calcium overload especially when this is accompanied by oxidative stress and these two parameters appear critical determinants of the recovery of isolated heart cells from simulated ischaemia. Indeed, the extent of recovery has been reported to correlate inversely with mitochondrial matrix calcium concentrations at the end of ischaemia [83,147] and to be triggered by the rise in ROS during reperfusion . It is well established that preconditioning reduces ROS production both at the end of ischaemia and during reperfusion [27,84,148,149] and decreases mitochondrial calcium overload [150–153]. Thus it is quite possible to account for the IP-mediated inhibition of MPTP opening at reperfusion merely through these indirect effects without having to invoke phosphorylation of any mitochondrial protein.
Our own data are entirely consistent with this mechanism. Opening of the MPTP in mitochondria isolated immediately following the preconditioning stimulus showed the same sensitivity to calcium as in control mitochondria [85,100]. Mitochondria isolated at the end of ischaemia, or during reperfusion, were more sensitive to calcium-induced MPTP opening than those isolated prior to ischaemia. However, at these times the mitochondria from preconditioned hearts were less sensitive to MPTP opening than were mitochondria isolated from control hearts at the same time [99–101]. This difference in sensitivity to calcium correlated with the exposure of the mitochondria to oxidative stress as reflected in the carbonylation of mitochondrial proteins [100,101], a surrogate marker for oxidative stress . These data confirm that the oxidative stress mitochondria experience at the end of ischaemia and during reperfusion is significantly attenuated in preconditioned hearts. As such they provide strong evidence in favour of IP exerting its primary effect on the reduction of ROS levels during ischaemia and reperfusion with the secondary consequence that MPTP opening is diminished. If this is the case, the key signalling pathways in preconditioning will be those that mediate the decrease in ROS levels at the end of ischaemia and during reperfusion. This could be the result of decreased ROS production or of improved ROS removal, but the ability of preconditioning to protect against oxidative stress mediated by exogenous hydrogen peroxide [155,156] argues in favour of the latter. However, an additional effect of IP to reduce ROS production cannot be ruled out.
Plasma membrane ATP-dependent potassium channels (KATP) are strongly expressed in the heart and have long been associated with cardioprotection [12,157]. It was originally proposed that opening of the plasma membrane KATP channel might hyperpolarize the cell leading to a shorter action potential duration (APD) and hence less calcium entry during ischaemia which would protect the hearts from calcium overload. In support of this, mouse hearts whose KATP channel Kir 6.2 had been knocked out exhibited greater calcium overload and ischaemia/reperfusion injury than control hearts . A role for the sarcolemmal KATP channels in the protective mechanism exerted by preconditioning was first suggested by the observation that KATP channel blockers such as glibencamide prevent preconditioning (see [12,157]). Further support came from the demonstration that hearts from the Kir 6.2 knockout mice were insensitive to either ischaemic or pharmacological preconditioning [158–160]. However, despite this convincing evidence in favour of a role for the sarcolemmal KATP channel in preconditioning, other data argue against such a role. In particular conditions were found in which cardioprotection could be observed in the absence of APD shortening whilst a range of KATP channel openers were reported to show a poor correlation between their effect on APD and their protective effects (see [11,12,161–163]). These data led to a change of emphasis, away from the sarlcolemmal KATP channel towards a similar KATP channel proposed to reside in the mitochondrial inner membrane.
The presence of an electrogenic mechanism for K+ entry into mitochondria (K+ channel) together with a K+/H+ antiporter to pump K+ out again has been known for more than 30 years, and their relative activities are thought to play a key role in the regulation of matrix volume (see [43,164]). What is less clear is the identity of these channels and transporters since none have been purified sufficiently to allow sequencing. Nor has genomic analysis revealed any plasma membrane K+ channel isoform or spliced variant with a mitochondrial targeting sequence . Rather the presence and properties of the K+ channels and transporters have been surmised from their functional characterisation (see [165,166]).
In isolated mitochondria, potassium ion movements can be measured directly using isotopes or K+-sensitive fluorescent dyes, or indirectly by determining matrix volume changes (see [167–171]). For mitoplasts patch clamping has been employed [172–174] whilst studies have been performed on proteoliposomes containing reconstituted inner mitochondrial membrane proteins using electrophysiological measurement of single channel current or fluorescent techniques to measure net K+ transport [168,175–179]. A major concern with the patch clamping and reconstitution techniques is the possibility that any channels detected in the mitochondrial membranes may actually represent a small number of K+ channels present in contaminating sarcolemmal membranes. It is known that most preparations of isolated mitochondria are extensively contaminated with plasma membranes unless further purified by density gradient centrifugation, and even then it is difficult to achieve complete removal of contaminating membranes [180,181]. However, in the majority of published work no data were presented to discount such contamination. Thus it is important to confirm that the activity of any channels identified by such techniques can be demonstrated in mitochondria, and this can be achieved by measurements of mitochondrial matrix volume.
It is widely agreed that the matrix volume is regulated by the relative activity of the K+ channels mediating membrane potential driven potassium entry into mitochondria, and the K+/H+ antiporter catalysing proton-linked K+ extrusion activity [43,164,165,182]. Thus activation or inhibition of any K+ channel should lead to an increase or decrease in matrix volume respectively. This can be measured either directly, using an isotopic technique based around the permeability of 3H2O into the matrix and the exclusion of [14C]-sucrose, or by light scattering (see [165,182]). The latter provides a very sensitive real-time measurement of matrix volume since as the mitochondria swell, their refractive index decreases and they scatter less light. A wavelength of 520 nm is frequently employed since this represents the isosbestic wavelength at which changes in the redox state of the mitochondrial cytochromes cause no change in absorbance . However, mitochondria can exhibit light-scattering changes that are independent of changes in matrix volume, reflecting rather a change in morphology or shape. It has been known for many years that such morphological changes, often referred to in terms of orthodox and condensed conformations of mitochondria, can be induced by ligands of the adenine nucleotide translocase (ANT) that switch the carrier from the “c” to the “m” conformation [44,49,170,183,184]. Thus externally added ATP, ADP or bongkrekic acid induce the “m” conformation of the ANT and cause a contraction of the mitochondria detected as an increase in light scattering whilst carboxyatractyloside, that induces the “c” conformation, exerts the opposite effect. Yet these light scattering responses all occur without changes in matrix volume measured isotopically [44,49,170,171].
Garlid and Paucek dismiss these data, stating that “this claim cannot be supported by any theoretical argument”, but the basis on which they make this assertion is unclear. First, the claim is backed up by experimental data from two different laboratories [44,49,170,171]. Second, the curvature of the mitochondria relative to the wavelength of light determines its refractive index and hence light scattering properties . Thus a shape change without a change in volume (e.g. from a sphere to a cylinder) will affect the light scattering signal. Since the ANT represents some 25% or more of the inner membrane protein in heart mitochondria , it is not surprising that when it switches its conformation it can cause a change in mitochondrial morphology and hence light scattering. Third, Garlid and Paucek argue that changes in light scattering induced by adenine nucleotides must reflect mitochondrial volume changes because they did not observe them in potassium free medium. They go on to say “If Das et al. had carried out these simple control experiments they would have avoided spurious claims about conformational changes”. However, we  and Brustovetsky et al.  have reported other controls that confirm that the adenine nucleotide-induced light scattering changes do not reflect changes in matrix volume. Thus additions of very low concentrations of valinomycin (0.4 to 1 pmol per mg mitochondrial protein) induced changes in matrix volume of 10–25% that were readily detectable using radioisotopes, yet the magnitude of the light scattering changes were similar to those induced by ADP or ATP for which no detectable changes in matrix volume were detected (see Fig. 3 and [170,171]). We have also performed experiments in potassium free media as suggested by Garlid and Paucek, which we report in Fig. 3. These data clearly show that such adenine nucleotide induced changes in light scattering do occur in potassium free medium and similar results were obtained by Brustovetsky et al. . We suggest that experiments directed towards investigating the activity of K+ channels in intact mitochondria that rely entirely upon light scattering changes for measurement of matrix volume should be treated with caution. Parallel isotopic measurements of matrix volume provide an essential confirmation that such light scattering changes do reflect changes in volume rather than morphology.
Some 20 years ago we showed that adenine nucleotides could inhibit electrogenic mitochondrial K+ entry into mitochondria and that this was probably mediated by their binding to the ANT [49,186,187], in agreement with earlier proposals . More recent data obtained with brain mitochondria have confirmed this proposal . However, the existence of a mitochondrial KATP channel similar to that found in the plasma membrane was first formally proposed as a result of data obtained using patch clamping of giant fused mitoplasts from liver mitochondria  and light scattering experiments in isolated mitochondria . Subsequently this channel has been studied extensively in several other laboratories (see [165,166]). Experiments using reconstituted proteoliposomes reported the characterisation of these channels with respect to the specificity for nucleotide inhibition and the activity of a range of pharmacological channel openers and blockers. In many cases the properties determined are similar to those of the plasma membrane KATP channel lending support to the possibility that it is actually these channels present in contaminating plasma membranes that are being characterised, rather than true mitochondrial channels (see Section 8.2). However, there are a number of differences in the reported properties of mitochondrial and sarcolemmal KATP channels that argue against this. For example, mitochondrial KATP channels have been reported to require the presence of Mg2+ for ATP to inhibit them, which is not true of the plasma membrane KATP channel. In addition, ADP acts like ATP to inhibit the mitochondrial channel but opens the sarcolemmal channel, whilst GTP has been reported to open the mitochondrial KATP channel but not the plasma membrane one [176,177]. Furthermore, their measured conductances were reported to be distinct  and different dose response curves for a range of pharmacological openers and blockers have been described [165,168,190]. This is especially true of the KATP channel opener diazoxide that has been reported to show a sensitivity towards the mitochondrial KATP channel at least three orders of magnitude greater than for the sarcolemmal KATP channel whilst 5-hydroxydecanoate (5HD) blocks the mitochondrial channel whilst having little effect on the sarcolemmal one [167,168]. However, there are other data claiming that diazoxide will open and 5HD block plasma membrane KATP channels at the concentrations frequently used to modulate mitochondrial KATP channels when investigating their possible role in cardioprotection [191–194].
Although no molecular identity has been ascribed to the mitochondrial KATP channel at the genomic level, similarities with the components of the plasma membrane KATP channels have been proposed. These contain one of two channel components, Kir6.1 (48 kDa) or Kir6.2 (44 kDa) and the sulphohylurea receptors, SUR1 (177 kDa) or SUR 2 (174 kDa) . Two studies reported that antibodies against Kir6.1 and Kir6.2 detected proteins of the right size (about 45 kDa) in heart mitochondria, but no rigorous steps were taken to remove sarcolemmal contamination [166,195]. In heart mitochondria SUR2 antibodies detected a protein of about 25 kDa in one study  but no protein in the other study , whilst a protein of about 28 kDa was labelled by [125I]-glibencamide . Overall, the available data provide no strong evidence for the presence of a KATP channel in mitochondria that is closely related to the well-characterised plasma membrane KATP channel. Garlid's laboratory found proteins of 55 kDa and 63 kDa in a partially purified preparation from brain mitochondrial inner membranes that showed KATP channel activity when reconstituted into proteoliposomes. The 63-kDa protein showed labelling by a fluorescent glibencamide derivative  leading them to surmise that the 55-kDa protein represented the channel forming component, but no sequence data were presented to identify either component. Furthermore, because only Coomassie blue staining of proteins was used, it was not established that there were not other minor components present in the partially purified preparation used and the high activity of channel proteins would only require a small contamination to produce significant channel activity. Indeed, it should be noted that mitochondrial preparations used in this study were unpurified on density gradient and therefore unavoidably would have contained significant contamination by fragments of plasma membrane and thus plasmalemmal KATP channels. This might explain why Garlid and co-workers found 6 to 7 times larger amounts of KATP channel in their preparations of isolated brain mitochondria than reported in liver or heart mitochondria.
In view of the problems associated with the reconstitution and patch-clamping studies discussed above, and the absence of a molecular identity for the mitochondrial KATP channel, the best evidence for the activity of mitochondrial KATP channels should come from the measurement of changes in matrix volume of isolated mitochondria in response to channel openers and blockers. In this context, Garlid's laboratory has performed extensive experiments using light scattering to detect matrix volume changes in response to adenine nucleotides and a variety of KATP channel openers and blockers. However, as noted above (Section 8.2), we believe that without direct measurements of matrix volume these data must be treated with caution. Thus, although data from both this laboratory and that of Brustovetsky confirmed effects of adenine nucleotides on light scattering that are consistent with the presence of a KATP channel, neither laboratory was able to detect volume changes isotopically despite demonstrating with valinomycin that the technique was quite sensitive enough to do so should they occur [170,171]. Furthermore, neither we nor Brustovetsky et al. were able to detect any significant changes in light scattering with a variety of KATP channel openers (diazoxide, nicorandil, cromakalim, pinacidil, minoxidil) or blockers (5HD and glibencamide) [170,171]. Our own data have led us to conclude that if mitochondrial KATP channels exist, there is no strong evidence for them being active in isolated mitochondria. However, this does not preclude such channels being active within the intact cell, since we might be losing a factor during mitochondrial isolation that is required for their activity. We will consider if there is any evidence for this in the context of studies investigating the role of mitochondrial KATP channels in preconditioning (Section 9.1).
Our own laboratory was the first to suggest that calcium might regulate potassium entry into mitochondria and hence matrix volume (see [182,198]). We demonstrated that K+ entry into liver mitochondria was stimulated by hormones that activated gluconeogenesis, such as glucagon, adrenaline and vasopressin, and that this led to an increase in matrix volume that was important for the stimulation of mitochondrial respiration and ATP production [182,198]. The mechanism was shown to involve a rise in matrix calcium that inhibits pyrophosphatase leading to a rise in matrix pyroposphate (PPi). Indeed, elevating matrix [PPi] independently of Ca2+, for example by provision of butyrate, also caused modest mitochondrial swelling . Since PPi was known to bind to the ANT [199–201] and adenine nucleotide depletion had been reported to increase mitochondrial K+ uptake [202,203], we proposed that the ANT could act as a potassium channel when ATP was displaced by PPi . In support of this we confirmed earlier data  that the permeability of the inner membrane to K+ was increased in the presence of atractyloside which inhibits the ANT by trapping it in the “c” conformation . This mechanism has recently received support from studies on brain mitochondria where it was shown that carboxyatractyloside could enhance potassium loss from de-energised mitochondria .
Our own data were unable to demonstrate calcium mediated increase in either matrix [PPi] or volume in isolated rat heart mitochondria [180,204]. However, data from O'Rourke's laboratory  have suggested the presence of a calcium-activated potassium channel (KCa) in these mitochondria. These workers employed patch-clamped mitoplasts to characterise mitochondrial KCa channels that, like the sarcolemmal KCa channels were blocked with charybdotoxin. Furthermore, antibodies against sarcolemmal KCa channels detected immunoreactive proteins in the mitoplasts. One problem with these data is that, as noted above, it is well established that isolated mitochondria are contaminated with plasma membrane and thus it is possible that the immunological and patch clamping data were actually detecting the sarcolemmal KCa.
There are numerous studies reporting protection of hearts from ischaemia reperfusion injury by drugs that are claimed to be specific openers of the mitochondrial putative mitochondrial KATP channel, most notably diazoxide, and the prevention of this effect by supposed specific blockers of the channel, in particular 5HD (see [11,12,161–163]). However, when haemodynamic performance of the heart is used to assess recovery after ischaemia/reperfusion, the effect of 5HD to reverse the protective effects of diazoxide is not universally observed [205,206]. O'Rourke et al. have suggested a role for the putative mitochondrial KCa channels in preconditioning, in addition to mitochondrial KATP channels. This proposal was based on the observation that the KCa agonist NS-1619 decreased the infarct size upon reperfusion after ischaemia, and this was blocked by paxilline, a KCa blocker [13,174]. However, as noted above (Section 8.3), a major problem associated with such studies is the assumption that the pharmacological agents used are only exerting their effects on the mitochondrial KATP or KCa channels. In the absence of independent verification of this, the conclusions drawn must be treated with caution. This is especially so in the light of increasing evidence that these agents have other non-specific effects on mitochondria including inhibition of the respiratory chain and uncoupling as is discussed below.
Diazoxide has been shown by many laboratories including our own to inhibit succinate dehydrogenase at concentrations usually used (50 μM) to open the mitochondrial KATP channel [169,206–209]. Because inhibiting succinate dehydrogenase will also block the citric acid cycle in an intact cell, this may well perturb cellular energy metabolism leading to mild depolarisation of the mitochondria and flavoprotein oxidation. It has also been reported that both diazoxide and pinacidil can uncouple mitochondria although the concentrations (> 50 μM) required to produce a significant depolarisation are at the high end of the spectrum used in in vivo studies [210,211]. Interestingly these effects appear to require an interaction with the ANT  and modelling studies based around the published structure of the ANT have shown potential drug binding sites on this membrane protein .
The specificity of mitochondrial KATP blockers is also doubtful, especially for the most commonly used 5 HD which is a substituted fatty acid and has the potential to be metabolised. Indeed, our own work from and that of others has confirmed that 5 HD can be activated to its CoA derivative and then further metabolised through the β-oxidation pathway [206,209,213]. As such it can act as a poor respiratory substrate. In addition, the slow kinetics of 3,5-dihydroxydecanoyl-CoA oxidation by l-3-hydroxyacyl-CoA dehydrogenase, the penultimate step of the pathway, can produce a bottleneck that can inhibit normal fatty acid oxidation [206,214]. Interestingly, the other commonly used mitochondrial KATP channel blocker, glibencamide, also has the potential to decrease fatty acid oxidation by inhibition of palmitoyl carnitine transferase [215–217].
Overall, the specificity of the mitochondrial KATP channel openers and blockers is poor and the ability of several of the KATP channel openers used to act as uncouplers or inhibitors of respiration is especially worrying. This is because it is well established that bona fide respiratory chain inhibitors [31,218–221], succinate dehydrogenase inhibitors  and uncouplers [223,224] are all cardio-protective when applied before or during ischaemia even though they have no effects on mitochondrial KATP channels. Thus, in order to establish a role for mitochondrial potassium channels in ischaemic preconditioning more direct evidence for their opening during IP must be provided.
Two approaches can be taken to investigate whether mitochondrial K+ channel opening is occurring during preconditioning. The first is the determination of matrix volume in situ and the second is measurement of other effects that an increase in electrogenic potassium flux into the matrix might have on mitochondrial function within the cell.
Although measurement of mitochondrial volume in situ in the perfused heart is not possible we have determined matrix volume in mitochondria isolated very rapidly from the perfused heart . Our data showed a significant increase in matrix volume induced by both IP (35%) and by 50 μM diazoxide (50%) which is consistent with the opening of a mitochondrial KATP channel. However, 100 μM 5HD, the putative mitochondrial KATP channel blocker, also induced an increase in matrix volume (50%) and failed to prevent the increase induced by IP. Indeed, we found no correlation between matrix volume (pre-ischaemic, end ischaemic or during reperfusion) and the haemodynamic recovery of the heart . Sollott et al.  have used a sophisticated Fourier analysis of transmitted optics linescan imaging with a 633-nm laser in isolated cardiac myocytes to estimate changes in mitochondrial diameter from which they calculate changes in matrix volume. They report a 2.5–4% increase in matrix volume with 30 μM diazoxide and 50 μM pinacidil consistent with opening of mitochondrial KATP channels. However, their approach makes the assumption that changes in diameter necessarily mean a change in matrix volume, but as noted above (Section 8.2), this is not necessarily the case. Furthermore, Sollott et al. report that a wide variety of other agents known to protect hearts from reperfusion injury through quite distinct mechanisms, including cyclosporin A, the sarcolemmal sodium proton antiport inhibitor Hoe694 and ligands of G protein–coupled receptors such as bradykinin and Tyr-D-Ala-Gly-Phe-D-Leu also cause similar increases in matrix volume . They propose that such an increase in matrix volume is critical for providing a memory that allows protection even after removal of the primary stimulus. Their explanation as to how this is achieved is that swelling increases the threshold at which ROS induce the permeability transition. However, why this should be the case is unclear since measurements on isolated mitochondria have demonstrated that MPTP opening is enhanced at higher matrix volume . Overall, the published data on measurements of mitochondrial volume in situ provide no strong evidence either for or against the opening of mitochondrial K+ channels.
Another approach used to investigate the activity of mitochondrial K+ channels in cardiac myocytes is based on the reasoning that if a K+ channel is opened the entry of K+ would cause depolarisation of the mitochondrial membrane potential. Marban and colleagues reported that with diazoxide this can be detected indirectly in rabbit cardiac myocytes by an oxidation of mitochondrial flavoproteins that is reversed by 5HD . Similar effects have been reported in guinea pig myocytes using NS1619 to open the putative mitochondrial KCa channel with reversal by paxilline . However, other workers have failed to reproduce the effects of diazoxide and 5HD in guinea pig and rat cardiac myocytes leading to the suggestion that the effects observed may reflect substrate deprivation [209,228]. In addition, there are theoretical reasons to question whether opening of the mitochondrial KATP channel would cause a significant depolarisation. Indeed, Garlid has argued convincingly that the fluxes of K+ predicted to occur if the channel opens would allow minimal perturbation of the membrane potential even when giving a significant increase in matrix volume [169,229]. In fact, such an increase in matrix volume might even increase membrane potential as a result of stimulating the respiratory chain itself . Direct measurements of membrane potential with the fluorescent dyes TMRE or JC-1 failed to detect either an increase or decrease [228,230]. There are, however, other mechanisms by which changes in mitochondrial membrane potential and thus flavoprotein oxidation could be induced by putative mitochondrial K+ channel openers and blockers that reflect other non-specific sites of their action. For example, in substrate deprived myocytes where diazoxide has been shown to cause flavoprotein oxidation, this could be caused by the drug inhibiting succinate dehydrogenase and thus the citric acid cycle whilst the ability of 5 HD to overcome this effect might be the result of the drug acting as a respiratory substrate as noted above (Section 9.1 and ).
For mitochondrial K+ channels to play a role in preconditioning there would have to be a mechanism by which their opening might be linked to improved recovery. Several possible mechanisms have been proposed.
It is known that an increase in matrix volume can stimulate respiration and oxidative phosphorylation [182,198] and so might improve ATP production during reperfusion leading to enhanced haemodynamic recovery [161,231]. Although we were able to measure an increase in mitochondrial volume and rates of ADP-stimulated respiration in mitochondria isolated from hearts immediately following ischaemic preconditioning or diazoxide treatment, the effects were largely lost during reperfusion when they are predicted to be most important . Furthermore, 5HD-treatment was shown to exhibit a similar increase in matrix volume but was not cardioprotective . Garlid and colleagues have proposed an alternative, but not unrelated mechanism in which it is the maintenance of the intermembrane space and the close association between the outer an inner membrane that is the critical factor resulting from KATP channel opening . They argue that during ischaemia the loss of membrane potential would decrease electrogenic K+ entry into mitochondria and that this would cause contraction of the matrix. It is proposed that that this would disrupt the interaction between VDAC in the outer membrane, octomeric creatine kinase (CrK) in the intermembrane space and the ANT in the inner membrane leading to faster permeation of ATP through VDAC into the intermembrane space. Further translocation of this ATP into the matrix through the ANT would lead to its hydrolysis, that would be detrimental to the heart. It is suggested that opening of the mitochondrial KATP channel would prevent this matrix condensation and so decrease ATP breakdown during ischaemia leading to less damage. There is some evidence that ATP decline during ischaemia is slowed by ischaemic preconditioning or diazoxide treatment, although the mechanism originally proposed was through inhibition of the F1Fo ATPase by enhanced binding of the ATPase inhibitor protein [232,233]. Furthermore, other data report the opposite effect of preconditioning, with the decline in ATP and the development of contracture being faster than in control hearts [232–235]. Garlid and colleagues further propose that during reperfusion, keeping the mitochondrial KATP channel open maintains the VDAC, CrK, ANT complex which is vital to export ATP rapidly from the mitochondria to the cytosol where it is used to drive contraction. Although this may be true, our own data showed that the slight increases in matrix volume of diazoxide-treated and ischaemic preconditioned hearts compared to control hearts was not significant at reperfusion . The more probable determinant of the efficiency of myocardial ATP production during reperfusion is the extent to which the mitochondria are damaged.
It has been proposed that opening of K+ channels would depolarise the mitochondria sufficiently during reperfusion to reduce ROS production and calcium accumulation and hence prevent MPTP opening [12,151,152,162,163,227,236]. However, as noted above, the amount of depolarisation predicted from K+ channel opening is unlikely to cause much if any depolarisation [169,229] and this is confirmed by direct measurements of mitochondrial membrane potential that failed to detect any depolarisation [228,230]. Furthermore, should significant depolarisation occur at reperfusion, ATP synthesis would be compromised leading to poorer rather than better recovery of the heart. Yet NMR spectroscopy measurements have shown that the bioenergetic state of the preconditioned heart improves during reperfusion, consistent with the better haemodynamic function [232,237,238]. Nevertheless, it is not possible to rule out the possibility that opening of mitoKATP channels might cause a minor depolarisation during reperfusion, or slightly reduce the repolarisation, sufficient to reduce ROS formation without having a significant effect on ATP production.
Although significant uncoupling during the reperfusion phase seems unlikely to be protective, it is known that adding low doses of uncoupler prior to ischaemia can precondition hearts [223,224]. Indeed, any interference with oxidative phosphorylation during the preischaemic phase seems able to exert a similar protective effect whether it is brought about by a brief ischaemic episode as in IP, by addition of a respiratory chain inhibitor [218–221] or a succinate dehydrogenase inhibitor . Since many of the putative mitochondrial K+ channel openers have also been shown to have direct uncoupling effects [209–211,239] or to inhibit components of the respiratory chain including succinate dehydrogenase [169,206–209,240], this provides a common mode of action for these agents. How this might be translated into a protective effect at reperfusion is not clear, although a signalling pathway involving AMP-activated protein kinase provides one possibility as discussed above (Section 6.4). Another potential mechanism would be through increased levels of ROS production causing PKC activation and respiratory chain inhibitors such as antimycin are known to increase ROS production which is critical in mediating its protective effects [218,241]. It has also been reported that low doses of uncoupler can increase ROS production in isolated myocytes , although how this might occur is unclear since in isolated mitochondria it is well established that even very modest uncoupling greatly decreases ROS production . In the case of antimycin, PKCε translocation has been demonstrated and its protective effects shown to be abolished in PKCε knockout mice supporting a role for ROS and PKCε activation in its protective mechanism .
Some groups have shown that diazoxide and other putative mitochondrial potassium channel openers cause an increase in ROS production by isolated heart mitochondria, cardiac myocytes and perfused hearts and that this can be blocked by 5HD [240,243–248], but this has not been reproduced in other studies [249,250] or in our own unpublished experiments. Since the protective effects of diazoxide and nicorandil can be overcome by free radical scavengers such as N-(2-mercaptopropionyl) glycine [12,100,240,245] it has been concluded by many workers that opening of the KATP channel causes ROS production and that this signals preconditioning through the pathways discussed above. Attractive though this hypothesis may be, it is difficult to formulate a mechanism by which this might occur. As noted above, even if there were modest uncoupling, this would be predicted to decrease ROS production rather than increase it. Another possible mechanism would be via an increase in matrix volume causing a stimulation of electron flow from complex 1 into complex 3 [182,251]. Indeed, there are reports that low doses of valinomycin can increase ROS formation in isolated mitochondria  and myocytes [11,163,247], whilst pretreatment of the perfused rabbit heart with valinomycin substantially decreased the infarct size on reperfusion . However, in our own unpublished experiments (H. Buckman and A.P. Halestrap, unpublished data) we have used Amplex Red to measured ROS production by isolated heart mitochondria incubated with a variety of respiratory substrates in State 4 and were unable to detect an increase in ROS production with any mitochondrial potassium channel opener tested, or when matrix volume was increased by decreasing the osmolarity of the incubation medium. Indeed, in the latter case we actually detected a decrease ROS production. This is consistent with the major locus of ROS production being at a highly reduced site on complex 1 that is oxidised by an increase in matrix volume as electron flow out of complex 1 and into complex 3 is stimulated [182,251].
Another consequence of opening a mitochondrial K+ channel would be an increase in the mitochondrial pH gradient coincident with a decrease in membrane potential . Whereas the latter would be predicted to decrease ROS formation as outlined above, there is evidence to suggest than an increase pH gradient or matrix pH can increase ROS formation from complex 1 . However, in view of the ability of many of the mitochondrial KATP channel openers to inhibit components of the respiratory chain [169,206–209,240], its is perhaps more likely that their ability to increase ROS is through a direct interaction with a redox centre in one of the respiratory complexes.
Recently, it has been proposed that connexion 43 (Cx43) may play a critical role in linking preconditioning to the mitochondria [255–258]. Cx43 is the major protein of the gap junction with hexameric assemblies forming connexons on adjacent cardiomyocytes. Its traditional role is to provide a route for intercellular communication through the propagation of action potential, signalling molecules and metabolites [258–260]. The permeability of the connexons can be regulated through many mechanisms including changes in [Ca2+], pH and phosphorylation by several protein kinases [258,259], including PKCε and PKG, both of which have been implicated in IP as noted above (see Section 6). Cx43 is normally partially phosphorylated with low conductance, but progressive dephosphorylation occurs during ischemia causing increased conductance. It has been proposed that this is important in propagating injury from one cell to another, although hearts from Cx43-deficient mice showed no difference in their sensitivity to reperfusion injury . Studies in both pig and rat hearts demonstrated that the dephosphorylation of Cx43 and electrical uncoupling during ischaemia were prevented by IP in a PKC-dependent manner and that IP increased the co-localization of several protein kinases with Cx43 during ischemia [262,263]. Furthermore, hearts from Cx43-deficient mice showed no protection from reperfusion injury by either IP  or diazoxide  and in the latter case the ability of diazoxide to increase ROS was also abolished.
These data imply a link between Cx43 and preconditioning and thus that Cx43 can in some way influence the opening of the MPTP, perhaps by a signalling mechanism involving ROS. Recent work has suggested that this may be through a translocation of Cx43 to the mitochondria following preconditioning [255–257]. A small fraction of Cx43, primarily in its phosphorylated form, was detected in purified mitochondria from rat, mouse, pig and human hearts and this was increased following preconditioning. In these studies, considerable efforts were taken to show that this Cx43 was not just the result of sarcolemmal contamination of the mitochondrial fraction and further evidence for this was provided by immunofluorescence confocal microscopy and immuno-gold electron microscopy. Furthermore, sub-fractionation of mitochondrial compartments led the authors to conclude that Cx43 is translocated to the inner mitochondrial membrane, although some Cx43 may remain associated with the outer membrane .
Although we have been able to detect Cx43 in purified mitochondria, and in preliminary experiments show that this was increased following preconditioning, we have expressed reservations about the likelihood of Cx43 translocation to the inner mitochondrial membrane . Furthermore, following more extensive experiments in which we determined the extent of sarcolemmal contamination of mitochondria by measuring the presence of MCT1, we were unable to confirm any consistent increase in the Cx43 associating with the mitochondrial fraction following IP (Fig. 4). Thus the relevance of Cx43 translocation to the mitochondria remains uncertain, although the loss of preconditioning in Cx43-deficient hearts from mice argues strongly for some pool of Cx43 playing a critical role [247,261]. Since myocytes from the Cx43-deficient mice did not show any diazoxide-induced ROS formation , this is likely to involve a ROS-mediated signalling pathway, although the mechanism linking Cx43 to ROS production is unclear. From their data, Garcia-Dorado and colleagues propose that recruitment of Cx43 to the mitochondria in IP might be required for opening of the mitochondrial KATP channel leading to ROS production that triggers preconditioning [255,257]. However, when they prevented this recruitment with geldanamycin, a blocker of heat shock protein 90-dependent protein translocation across the outer mitochondrial membrane, the cardioprotective effects of IP were not lost . Furthermore the cardioprotective effects of both diazoxide and isoprenaline were not accompanied by translocation of Cx43 to the mitochondria . Thus the balance of evidence suggests that Cx43 translocation to mitochondria is not essential for preconditioning .
Hausenloy and colleagues have reported that when the MPTP inhibitors CsA or SfA are present during IP or preconditioning induced by uncoupler and diazoxide, cardioprotection is abolished . They conclude that transient opening of the MPTP may play a critical role in triggering preconditioning and that this might be caused by the well-documented rise in ROS that occurs during preconditioning (see Section 6.1). However, we have challenged these conclusions both on theoretical and experimental grounds . First, our own data using the mitochondrial 2-deoxyglucose entrapment technique failed to detect any increase in MPTP opening immediately following the preconditioning stimulus [68,75]. Second, although the authors suggest that transient MPTP opening during the preconditioning phase prevents mitochondrial calcium overload and that this mediates protection, it is the calcium overload at the end of ischaemia and during reperfusion that mediates the MPTP opening responsible for reperfusion injury. Third, the authors argue that uncoupler also preconditions hearts by opening the MPTP, but uncouplers will depolarize mitochondria independently of MPTP opening. Thus any protection mediated by uncoupler that involves a decrease in mitochondrial calcium loading should be insensitive to both SfA and CsA, but this is not what the authors observed . Hence it seems probable that CsA and SfA are overcoming IP independently of any effects on MPTP opening. This might involve roles for cyclophilins in the preconditioning signalling pathway similar to those responsible for the immunosuppressive actions of these drugs .
The ability of preconditioning (and post-conditioning) to inhibit the opening of the MPTP during reperfusion and so provide cardioprotection seems to be well established. What is less clear is the mechanism(s) by which this is achieved. We end this review by presenting our own views on how preconditioning is most likely to be mediated. These are summarised schematically in Fig. 5.
The observation that mitochondria isolated immediately after the preconditioning protocol show no protection from MPTP opening [85,100] argues against a signalling pathway that causes phosphorylation of some regulatory component of the MPTP as an early event. Nor have we been able to observe any consistent migration of protein kinases to the mitochondria or changes in phosphorylation of a mitochondrial protein (see Sections 6 and 7.1). By contrast, mitochondria isolated at the end of ischaemia or during reperfusion do show a decreased sensitivity to calcium-mediated MPTP opening [99–101]. Here too we were unable to detect consistent migration of protein kinases to the mitochondria or changes in phosphorylation of a mitochondrial protein, arguing against this protection being mediated by phosphorylation (see Section 7.1). However, the changes in sensitivity of the MPTP do correlate with the extent of protein carbonylation, a surrogate indicator of the oxidative stress the mitochondria have experienced [100,101]. Since it is well established that oxidative stress sensitises the MPTP to calcium, this would seem to provide an adequate explanation as to how preconditioning inhibits MPTP opening at reperfusion (see Section 7.2). Furthermore, this is entirely consistent with the decrease in ROS production late in ischaemia and during reperfusion that has been observed directly in preconditioned hearts [27,265].
If this explanation is correct then the focus of research should shift towards investigating the mechanisms underlying the decreased levels of ROS seen in preconditioned hearts at the end of ischaemia and during reperfusion. This will involve consideration of both the sources of ROS, which may include mitochondria, NAD(P)H oxidase or xanthine oxidase, and the pathways that remove them including superoxide dismutase, glutathione peroxidase and catalase [26,28]. The many signalling pathways identified as potential mediators of preconditioning, including ROS, PKCε and PKG (see Section 6) may converge on one or more of these steps to decrease ROS levels and hence reduce activation of the MPTP by thiol oxidation. The pharmacological agents used to implicate these pathways may interact specifically with their proposed target, but they may also have less specific effects such as on mitochondrial respiratory chain components (see Section 9.1) to produce a modest increase in ROS levels that stimulate signalling pathways rather than cause damage. Indeed, it does seem that any stress to the myocyte's energy status, be it through respiratory chain inhibition, uncoupling or ischaemia, causes preconditioning. This may also implicate the energy sensing protein kinase, AMP-activated protein kinase (AMPK) [133,266], in the signalling pathway of preconditioning, and there are data supporting such a role for AMPK in both the heart [101,135] and the liver .
In addition to the protection seen early in reperfusion that could be explained through the mechanisms described above, there is good evidence for ongoing protection against MPTP opening being important as reperfusion progresses. Thus inhibition of MPTP opening with CsA or SfA provides effective cardioprotection, as measured by decreased infarct size, even if the drug is not present at the start of reperfusion, provided it is added within the first 15 min [90,91]. Furthermore, there is a considerable body of evidence to suggest that activation of survival kinase pathways during reperfusion is critical for protection during this phase of the insult, and that these may converge on Akt and GSK3 to mediate their effects [97,107,126]. It is possible to bring these observations together by invoking the established phenomenon of MPTP-induced ROS production [268,269]. Once MPTP opening has occurred during the initial phase of reperfusion, the resulting ROS production would lead to a progressive increase in pore opening in other mitochondria and thus an escalating number of necrotic cells as reflected in the ongoing LDH release and increased infarct size. CsA or SfA will prevent this ongoing opening, but so too might the survival kinase cascades by reducing the ROS production or increasing their removal as described below.
It is well established that MPTP opening causes mitochondrial swelling and rupture of the outer mitochondrial membrane . This leads to release of pro-apoptotic factors including cytochrome c, that activates caspase 9 and hence caspase 3 . In addition, it has been shown that the pro-apoptotic factor Bax can translocate to the mitochondria during ischaemia  and this, in conjunction with cleaved Bid (tBid), might cause the cytochrome c release observed during prolonged ischaemia  despite there being no evidence of MPTP opening [68,98]. The resulting loss of cytochrome c will slow electron transfer out of complex 3 and thus potentially cause increased ROS production in either complex 3 or in complex 1. An additional mechanism that may lead to increased ROS production during ischaemia and reperfusion is through caspase 3 mediated cleavage of the p75 component of complex 1 [270,271]. Protection from the rise in ROS that accompanies such cytochrome c release could be mediated by survival kinase cascades in two ways. They might stimulate ROS removal as described above for the ischaemic phase or they could reduce the Bax-induced cytochrome c release. Indeed, it is well established that survival kinases can block apoptosis by inhibiting cytochrome c release . This is brought about by Akt-mediated phosphorylation of the pro-apoptotic Bcl-2 family member Bad [273,274] and, via GSK3 phosphorylation and stabilisation of the anti-apoptotic Bcl-2 family member Mcl-1 .
Since submission of this article it has been reported that the properties of the mitochondrial permeability transition pore in mitochondria devoid of all VDAC isoforms are the same as in mitochondria from wild-type mitochondria (C.P. Baines, R.A. Kaiser, T. Sheiko, W.J. Craigen, J.D. Molkentin, Voltage-dependent anion channels are dispensable for mitochondrial-dependent cell death, Nat Cell Biol. 9 (2007) 550–555). This confirms that VDAC is not an essential component of the MPTP.
Work from the authors' laboratory has been funded by a Programme Grant from The British Heart Foundation. The authors wish to thank members of the Bristol Heart Institute, and especially Dr. Elinor Griffiths and Professor Saadeh Suleiman, for their continuing encouragement and support, and to Dr. Kate Heesom of The Proteomics Facility for performing the 2D gel electrophoresis.