Regulation and pharmacology of the mitochondrial permeability transition pore

Abstract

The ‘mitochondrial permeability transition', characterized by a sudden induced change of the inner mitochondrial membrane permeability for water as well as for small substances (≤1.5 kDa), has been known for three decades. Research interest in the entity responsible for this phenomenon, the ‘mitochondrial permeability transition pore’ (mPTP), has dramatically increased after demonstration that it plays a key role in the life and death decision in cells. Therefore, a better understanding of this phenomenon and its regulation by environmental stresses, kinase signalling, and pharmacological intervention is vital. The characterization of the molecular identity of the mPTP will allow identification of possible pharmacological targets and assist in drug design for its precise regulation. However, despite extensive research efforts, at this point the pore-forming core component(s) of the mPTP remain unidentified. Pivotal new genetic evidence has shown that components once believed to be core elements of the mPTP (namely mitochondrial adenine nucleotide translocator and cyclophilin D) are instead only mPTP regulators (or in the case of voltage-dependent anion channels, probably entirely dispensable). This review provides an update on the current state of knowledge regarding the regulation of the mPTP.

1. Introduction

The outer and inner mitochondrial membranes (OMM and IMM), being in many ways significantly different from each other, are each composed of phospholipid bilayers that contain numerous integral proteins. The OMM is much more permeable as a result of an abundance of the principal OMM protein, the voltage-dependent anion channel (VDAC), which forms relatively large internal channels (about 2–3 nm) and allows ions, metabolites, and certain small molecules to move between the cytoplasm and mitochondria. In contrast, the IMM is highly impermeable, and the transport of ions, molecules, and metabolites across it is supported by specialized membrane transporters and exchangers; it is also rich in cardiolipin, an unusual phospholipid which is typically found in bacterial plasma membranes. The IMM is a very good electric insulator: at a thickness of about 10 nm, this membrane can keep a potential difference of ∼180 mV, thereby creating an electric field strength of about 1.8 × 10⁵ V/cm.¹ Proton pumping driven by substrate oxidation constantly supports mitochondrial membrane potential (ΔΨ_m) and compensates for proton and other ion leakage, due to ATP synthesis and other regulated and unregulated ion transport.

The abrupt change of the IMM permeability, characterized by the sudden loss of ΔΨ_m, is known as the ‘permeability transition’ and the entity responsible for this phenomenon was named the ‘mitochondrial permeability transition pore (mPTP)’. mPTP induction causes the non-selective traffic between the mitochondrial matrix and cytosol of not only small-charged particles but also of water and substances up to 1.5 kDa in size, a phenomenon that was discovered by Haworth and Hunter.^2–4 Owing to the progress of patch-clamping studies applied to the IMM, the term ‘mitochondrial megachannel’ has also been utilized for probably the same phenomenon.^5,6

The molecular structure and identity of the mPTP is not yet known. A complicating factor in the characterization of the mPTP relates to the circumstances in which the permeability transition phenomenon was discovered and initially studied, e.g. in isolated mitochondria,^2–4 and that it was not studied in intact cells until techniques became available almost two decades later.^7,8 These two experimental systems have often failed to exhibit the expected parallelism in terms of behaviour, e.g. there are apparent differences in the role and sensitivity to Ca²⁺ as an mPTP trigger, and in the effects of mPTP inhibitors under specific conditions, suggesting potential differences in the mPTP size, its thresholds, and architecture.^9–13 It has been estimated that a single mPTP per mitochondrion can discharge the ΔΨ_m very quickly.¹⁴ Yet unresolved difficulties with isolation and identification of the mPTP may be related to the fact that (i) the assembly of the mPTP from its constituent components, and its induction, is likely a rare event and (ii) it may be assembled from components that, while relatively abundant, generate only a single macromolecular cluster that exists only transiently. There may even be several different regulatory entities assembled around the main core components, yielding somewhat different behaviours in different cells or under different conditions. This may partially explain the use of the terms ‘regulated’ and ‘unregulated’ by different authors to describe mPTP, or the in vivo vs. in vitro differences in Ca²⁺ sensitivities, demonstrating the inherent complexities in observing the obscure nature of this phenomenon.¹⁵ Furthermore, a number of recent genetic knockout studies have convincingly shown the irrelevance of certain previously ‘indisputable’ elements as belonging to the core architecture of the mPTP complex since the mPTP phenomenon persists even in their absence,^16–20 as will be discussed later in this review.

The relevance of the permeability transition to the ‘life or death’ decision of the cell^21,22 underlies the necessity to understand the detailed structure and behaviour of this phenomenon. Crompton and Costi²³ implicated, for the first time, the mPTP-induction as a critical contributing factor to ischaemia/reperfusion injury in a cell-based model. They hypothesized that reoxygenation after prolonged ischaemia could result in excessive mitochondrial damage due to the mPTP induction and that ATP depletion and oxidative stress upon reperfusion would provide additional stimuli to promote further mPTP activation and consequently exacerbate the damage. Proof that mPTP occurs after ischaemia/reperfusion was experimentally obtained shortly thereafter in an in vivo setting using isolated rat hearts.²⁴ Importantly, under normal circumstances, oxidative phosphorylation in mitochondria results in both the majority of cellular ATP production and relatively low levels of background reactive oxygen species (ROS) production; however, once the mPTP is induced, the situation is drastically reversed with the mitochondria becoming not only the producers of high levels of cytotoxic ROS but also ATP consumers due to the futile cycle in an effort to restore the Δψ_m, thereby further contributing to cellular damage. In intact cardiomyocytes we have demonstrated that in individual mitochondria the mPTP induced by an elevated level of ‘triggering ROS’ is followed by an additional ROS burst that occurs simultaneously with mitochondrial Δψ_m dissipation, a phenomenon which we named, ROS-induced ROS release (RIRR). Furthermore, RIRR can propagate and thus can be responsible for a large fraction of the oxidative damage detected in cells after reperfusion.²⁵ It has also been found that ROS rather than Ca²⁺ appears to be the more important mPTP trigger in excitable cells such as cardiomyocytes and neurons; additionally, cell survival has negatively been correlated to the fraction of cellular mitochondria that undergo mPTP induction.^9,10 Finally, it should be emphasized that ageing is also a critical factor contributing to the ability of tissues to resist damaging stress (see Juhaszova et al.²⁶ for review, and citations therein). Specifically, there is an age-related impairment of the intrinsic capacity of the heart to resist damaging stress, such as ischaemia/reperfusion injury (an effect aggravated by co-morbidities such as diabetes, but potentially attenuated by caloric restriction and physical conditioning/exercise). Indeed, cardiomyocytes from aged hearts have a markedly lower mPTP–ROS threshold and consequently a higher probability of mPTP induction when compared with those of young animals.²⁷ It also appears that the stimulus threshold for triggering endogenous protection signalling mechanisms increases with age, e.g. the ability to induce ischaemic preconditioning is significantly blunted in the old heart (an effect potentially aggravated by commonly used medications that can block certain key steps in protection signalling, including sulfonylureas, antioxidants, partial fatty acid oxidation (PFOX) inhibitors, and COX2 inhibitors, but also likely attenuated by caloric restriction and exercise) (see Juhaszova et al.²⁶ for review, and citations therein). Additionally, mutations in mitochondrial transfer RNA have been correlated with hypertension and hypercholesterolemia, major age-related cardiovascular risk factors.²⁸

The characterization of the molecular identity of mPTP should allow identification of potential pharmacological targets and assist in rational drug design for therapeutic aims requiring precise mPTP regulation. When solved, this could provide tools for either preventing unwanted cell death when appropriate or for promoting the elimination of unwanted cells, for example in the situation when a small, damaged, but still viable region of the heart, if not eliminated, becomes a focus of abnormal automaticity leading to pathological disturbances in cardiac rhythm (including ventricular tachycardia and ventricular fibrillation). The direct link between mPTP induction and oxidative stress-related cell death in cardiomyocytes and other vital organs during ischaemia/reperfusion injury requires developing better strategies to prevent or limit the possibility of damage.

In this review, we will present what is known about certain key elements of the permeability transition phenomenon and possible targets for pharmacological intervention. We will discuss what is known about the essential (core) pore-forming elements of the mPTP and other regulatory components that control the activity of the mPTP or its sensitivity to modulating factors (such as Ca²⁺, proteins, lipids, etc.) (Figure 1). Some of the components that are known or have been suspected to be involved in the formation and/or regulation of the mPTP include adenine nucleotide translocator (ANT) (also known as ATP/ADP carrier, AAC), cyclophilin D (Cyp-D, encoded by the Ppif gene), VDAC (also known as mitochondrial porin), hexokinase, creatine kinase, mitochondrial peripheral benzodiazepine receptor (PBR; translocator protein, TSPO), Bcl-2, glycogen synthase kinase-3β (GSK-3β), and cytochrome c.^9,29–31 Since some of the tentative elements of the mPTP complex belong to the OMM, some reside in the intermembrane space and some are located in the IMM, it has been hypothesized that the mPTP complex could be organized by a structure previously known as the mitochondrial ‘contact site’.^32,33 However, this hypothesis has never been experimentally validated and remains only a speculation (comprehensively discussed by Bernardi et al.³⁴). The important role of a principal lipidic counter-partner of the mPTP, cardiolipin,³⁵ will also be discussed.

Figure 1.

Proposed mPTP complex architecture: (A) Classical view. The mPTP structure is formed by the VDAC–ANT–CyP-D complex, which is located at the ‘contact sites’. Hexokinase II (HKII), mitochondrial creatine kinase (CK), benzodiazepine receptor (PBR), and Bcl-2-family members (Bcl-2, Bcl-x_L, and Bax) are included as putative regulatory components. (B) Current view. The core elements comprising the mPTP itself (denoted ‘PTP’ for permeability transition pore) are presently unidentified, but are probably regulated by the adjacent elements as indicated. Note that VDAC, portrayed as a ‘shadow,’ is no longer seen as an essential pore component or even a regulator based on recent genetic evidence. Question mark symbols signify where important open questions remain as far as the participation as a regulator of the mPTP (see text; modified from Juhaszova et al.⁹).

2. Adenine nucleotide translocator

ANT, a relatively small protein (about 300 AA residues) and the most abundant protein of the IMM, catalyses the selective, electrogenic, and reversible exchange of ADP for ATP with a 1:1 stoichiometry. ANT transports nucleotides at their full ionization state, i.e. the antiport is accomplished by the exchange of ATP⁴⁻ for ADP³⁻.³⁶ The direction of nucleotide transport is determined by cell conditions (under normal physiological conditions ATP⁴⁻ is transported out of the mitochondria and ADP³⁻ is transported in, while the opposite is true under anoxic/ischaemic conditions). This nucleotide exchange is limited by the availability of free nucleotides that are typically at much lower concentration than their Mg²⁺-bound forms. ANT also transports other solutes such as phosphoenolpyruvate,³⁷ pyrophosphate,³⁸ and creatine phosphate,³⁹ although with very low efficiency.

Substantial evidence supports the involvement of ANT in fatty acid-induced mitochondrial uncoupling. It has been suggested that ANT can transport anionic forms of fatty acids, which is typically the rate limiting step in the mitochondrial uncoupling process.⁴⁰ Similarly, ANT may facilitate the transport of anionic forms of weak uncouplers (such as 2,4-dinitrophenol) across the IMM.⁴¹ This highlights the role of ANT (as well as other mitochondrial carriers like the glutamate/aspartate carrier or UCPs) in the regulation of ΔΨ_m and ROS production, the latter being tightly correlated with the ΔΨ_m level.⁴²

Known inhibitors of ANT are highly specific for the inhibition of the nucleotide transport. Atractyloside and its derivative carboxyatractyloside are competitive inhibitors of nucleotide transport by ANT.^43,44 The binding of these inhibitors on the cytosolic side causes the transition of ANT into its c-conformation, while binding of another competitive inhibitor, bongkrekic acid,⁴⁵ causes a transition into the m-conformation when applied from mitochondrial matrix side.

Among other ANT inhibitors, acyl-CoA, particularly long-chain derivatives such as palmytoyl- and oleyl-CoA,⁴⁶ and rhodamine 6G⁴⁷ are well known (consequently, the rhodamine-based ΔΨ_m probes should be used with proper precautions). The search for natural modulators of ANT led to discovery of the small water-soluble protein CyP-D, which has a binding site on the matrix side of ANT.⁴⁸

Bcl-2 family proteins are also known ANT modulators.⁴⁹ Viruses have evolved complex pro- and anti-apoptotic strategies (regarding the host cells they infect) for their own benefit, often acting specifically at the mitochondrial level. The interaction of viral proteins with different members of Bcl-2 family proteins as well as mPTP regulators has been demonstrated. For example, the interaction of HIV-1 viral protein R (Vpr) with ANT has been implicated in mPTP induction and cell death.⁵⁰ Some viruses encode products resembling Bcl-2 protein family domains that may operate via interaction with mPTP structures (reviewed in Galluzzi et al.⁵¹).

In reconstituted systems, purified ANT formed leaky channels and pore-like structures.⁵² Single channel current measurements in these systems demonstrated Ca²⁺-dependent high conductances with multiple sublevels.³⁵ On the basis of the similarity of the ANT-derived channel activity to the tentative mitochondrial multiconductance-level channel and regulation of the mPTP phenomenon by ANT ligands,^2,53 it was originally suggested that ANT might be a core element of the mPTP complex.^48,54 Later studies, however, cast serious doubt on this conclusion because in an ANT double knockout mouse (ANT1−/− and ANT2−/−) typical Ca²⁺- and cyclosporin A (CsA)-sensitive mPTP-like behaviour of the IMM was observed, though with a higher Ca²⁺ threshold.¹⁹ However, considering the importance of mitochondrial ATP/ADP exchange in the cell, other as-yet undetermined compensatory system(s) may exist, such as other unrecognized ANT isoforms or another carrier taking charge of mitochondrial nucleotide transport. On the other hand, unless new evidence comes to light that such compensatory mechanisms exist and lead to the expression of a protein that mimics all the functions of the ANT, it should become a settled matter that the ANT is a regulatory element of the mPTP rather than (part of) its core structure. Figure 2 summarizes the current view of the mPTP regulation by ANT and CyP-D.

Figure 2.

Regulatory roles of ANT, CyP-D, and P_i. The right side of the figure indicates symbolically the threshold for mPTP induction by oxidant stress, whereas the left side indicates mPTP regulatory elements. The middle row (i.e. examined horizontally) indicates the basal state; the regulatory mechanisms shown symbolically in the upper row lead to facilitation of mPTP-induction, while those in the lower row indicate inhibition. Middle row depicts the basal state of ANT and CyP-D as they relate to the basal threshold for mPTP induction by oxidant stress. Top row reflects factors that facilitate mPTP induction: atractyloside, Ca²⁺, and indirect effects of P_i. Bottom row includes factors that are known to inhibit mPTP induction: genetic deletion of ANT (ANT is dispensable for mPTP formation per se; inhibition of CyP-D by CsA remains protective), ADP or bongkrekic acid (requirement/role of CyP-D under these conditions is unknown), CsA and genetic deletion of CyP-D in the presence of P_i (atractyloside, CsA and Ca²⁺ are no longer effective when compared with WT). Note the opposing mechanisms of P_i in mPTP-induction: (i) P_i as a direct mPTP desensitizer (bottom row) is opposed by CyP-D binding (top row), whereas (ii) P_i may also act as an indirect mPTP sensitizer (through regulation of Mg²⁺ and/or polyphosphate levels; top row; see text). Note also that Ca²⁺ is not a major factor in mPTP induction in cardiomyocytes and neurons. Ppif is the gene encoding CyP-D in mouse.

Additionally, during ischaemia and hypoxia, intracellular concentrations of ADP and P_i rise as ATP is used up. This results in a marked increase in mPTP opening probability, likely caused by the prevalence of ANT in c-conformation under these conditions. Consequently, factors restoring intracellular ATP level are considered protective, opposing mPTP induction.

3. Cyclophilin D and phosphate

Experiments employing a Cyp-D affinity column demonstrated a direct interaction between Cyp-D and ANT. Furthermore, pretreatment of the column with CsA prevented the binding of ANT to the column, thereby confirming that Cyp-D is in fact the target of CsA.⁵⁵

After the sensitivity of the mPTP to CsA was recognized,⁵⁶ there were attempts to establish a case that the mPTP is always blocked by CsA. Currently, it is believed that rather than completely blocking the mPTP, CsA desensitizes and delays its opening. Originally, a simplified scheme proposed that CyP-D, residing in the mitochondrial matrix, undergoes Ca²⁺-dependent binding to the matrix site of ANT, thereby favouring the ‘conversion’ of ANT into the mPTP.⁴⁸ The relevance of the peptidyl–prolyl cis–trans isomerase (PPIase or rotamase) activity of CyP-D (essential for CyP-D chaperone-like activity) to mPTP induction is not yet clear. One of the earlier hypotheses suggested that the mPTP is formed by aggregation of misfolded membrane proteins.¹⁵ According to this scheme, chaperone-like proteins blocked mPTP conductance by binding and refolding these proteins. CsA interaction with the hydrophobic pocket of Cyp-D resulted in inhibition of both mPTP opening and PPIase activity. It has also been proposed that CsA may compete for Cyp-D with some unknown substrate necessary for mPTP induction.⁵⁷ Once the effects of CsA and diethylpyrocarbonate on PPIase activity and mPTP-induction, respectively, were compared—although both inhibited the PPIase activity of Cyp-D—the first agent inhibited mPTP, while the second one induced it—it was noted that PPIase activity may be irrelevant to mPTP-induction.⁵⁸ In contrast, it has been reported that PPIase activity is essential for protection against apoptotic stimuli in HEK293 and rat glioma C6 cell systems overexpressing CyP-D.⁵⁹ This observation suggests that CyP-D may play a differential role in mitochondrial injury triggered by mPTP induction and by apoptotic stimuli, respectively (see below).

Similar to the situation with ANT, recent genetic studies have demonstrated that CyP-D is also not an essential component of the mPTP but rather plays a regulatory role¹⁸ (Figure 2). CyP-D deletion renders the mPTP insensitive to CsA, an observation which has confirmed that the target of this drug in modulating the mPTP is indeed CyP-D. Two other studies demonstrated that cells missing CyP-D are resistant to oxidative stress and Ca²⁺ overload,¹⁶ and that although they remain quite sensitive to apoptotic stimuli, they become resistant to necrotic death stimuli.²⁰ This lead to the overall conclusion that mPTP induction may be important for necrosis but not necessarily for apoptosis.

The unexpectedly complex role of inorganic phosphate (P_i) in the modulation of the mPTP was recently highlighted. Careful observations now suggest that P_i may apparently exert dual actions on the mPTP depending on the functional state presence of CyP-D. As pointed out by Bernardi,⁶⁰ P_i can either be an important mPTP sensitizer, probably acting by decreasing matrix-free Mg^{2+ 61} and/or by the formation of polyphosphates,⁶² or play a role in desensitizing the mPTP to Ca²⁺ (in the presence of CsA, or under circumstances of CyP-D ablation). Accordingly, in this model the presence of CyP-D tends to prevent the intrinsic inhibitory action of P_i by blocking the P_i regulatory site on the mPTP (thus, CsA or genetic deletion of CyP-D act by unmasking the direct mPTP-desensitizing action of P_i; Figure 2).⁶⁰

The therapeutic application of CsA as a desensitizer of the mPTP has been hindered for many years because of a variety of factors, including a very narrow optimal concentration range²⁴ (possibly related to the P_i levels, as discussed above), the potential occurrence of cancer,⁶³ and nephrotoxicity.⁶⁴ Nevertheless, CsA was found to be effective in limiting infarct size during acute myocardial infarction,⁶⁵ and in correcting mitochondrial dysfunction in collagen VI myopathies.⁶⁶ The protective properties of CsA in different pathologies have recently been comprehensively reviewed.³⁴ Besides CsA, more specific CyP-D inhibitors have also been investigated that do not exhibit undesirable side effects (such as the inhibition of the protein phosphatase calcineurin). In particular, non-immunosuppressive CsA analogues, e.g. methylAla(3)ethylVal(4)-cyclosporine (Debio 025)⁶⁷ and (N-methyl-4-isoleucine cyclosporine) NIM811,⁶⁸ and Sanglifehrin A (a compound unrelated to CsA),⁶⁹ have shown promise as potent mPTP desensitizers with less adverse side effects in vivo when compared with CsA.

4. Voltage-dependent anion channel

Both the high permeability of the OMM and the channel properties of VDAC have been known for many decades.^70,71 Based on purified protein reconstitution experiments, the major regulator of VDAC was thought to be ΔΨ_m which when increased results in the closed state.^71,72 The observed voltage dependence of VDAC was surprising, because the OMM does not contain voltage generators. Two alternative explanations have been investigated to explain this fact: (i) it has been suggested that the close proximity of the VDAC in ‘contact sites’ to the IMM may result in sensing of the high voltage existing across the IMM and consequent VDAC closure. In isolated mitochondria the number of ‘contact sites’ depends on the mitochondrial functional state (in state 3 the number of ‘contact sites’ is higher than in state 4⁷³). Thus, depending on the metabolic state, the total VDAC conductance may vary. (ii) Although, in the reconstituted system, the voltage gating of VDAC is apparent, in living cells, other regulators of its conductance, e.g. regulatory proteins, may be involved (discussed below). It was suggested that VDAC is the first check-point responsible for control of the transport through the OMM and that its conductance is the major factor in regulation of the mitochondrial inward and outward traffic of oxidative substrates, nucleotides, amino acids (AA), small peptides, signalling molecules, etc.⁷⁴

Specific VDAC inhibitors or modulators have not been described yet. VDAC readily interacts with anionic rather than with cationic structures and in the open state VDAC preferentially conducts anions. One of the first described artificial regulators of VDAC is the Konig's synthetic polyanion.⁷⁵ In an in vitro system, this construct induced VDAC closure, resulting in impermeability of the OMM, for example, to ATP and ADP, and consequent block of mitochondrial metabolism. Another very effective group of compounds that induce VDAC closure are phosphorothioate oligonucleotides.⁷⁶ Additionally, it was suggested that protection against apoptosis elicited by the BH4 domain of Bcl-x_L may be partially mediated by VDAC closure.⁷⁷

The well-known VDAC modulator, NADH, causes a steep voltage dependence of VDAC behaviour.⁷⁸ A similar steep voltage-dependent mode of action was also induced by an unidentified ‘soluble protein modulator’ that behaves as a physiological controller of VDAC gating, and thus its involvement in the governing of metabolic flux across the OMM has been suggested.⁷⁹ It has been proposed that this modulator may be responsible for matching VDAC conductance to changes in the cell's environment. The identity of this modulator was unknown then but the molecular mass (around 90 kDa)⁸⁰ closely matches that of a cytosolic, VDAC interacting protein, the hexokinase (HK).

VDAC possesses a GSK-3β phosphorylation consensus motif at amino acids 51–55. Two protective mechanisms triggered by GSK-3β-inhibition-mediated VDAC dephosphorylation were suggested: (i) a resulting increase in HKII binding to mitochondria,⁸¹ although additional experiments found that insulin did not protect against the effects of forced HKII detachment caused by the peptide, HKII-VBD,⁸² and (ii) VDAC dephosphorylation-induced VDAC closure and consequent inhibition of the ATP influx through VDAC during ischaemia.⁸³ However, the direct relationship of VDAC phosphorylation status, and the role played by HKII (if any), to the mPTP has yet to be unambiguously demonstrated.

Although VDAC had originally been recognized as a key component of the mPTP,^29–31 recently, after demonstration that the main features of the permeability transition are still preserved in mitochondria isolated from the VDAC1–VDAC3 knock-out mouse and in VDAC 1/3−/− cells in which VDAC2 has been silenced,¹⁷ the necessity of VDAC in the mPTP has also been thoroughly disputed. Indeed, fibroblasts lacking all three VDAC isoforms exhibited similar stress-induced mPTP induction and cell death in response to the ‘pro-death’ Bcl-2 family members as did wild-type (WT) cells, suggesting not only the complete dispensability of VDAC as an element of mPTP but also its irrelevance to the Bax- and tBid-induced cell death.¹⁷ Furthermore, recent experiments in cells lacking VDAC1 and 3 (and without detectable binding between HK II and residual VDAC2) demonstrated also the dispensability of VDAC for cell death elicited by HKII detachment from mitochondria by a HKII N-terminal peptide (HKII-VBD, see below).⁸² Thus, VDAC is not an mPTP element and currently, beyond speculation there is no compelling direct evidence for its mPTP-regulatory role either (Figure 3).

Figure 3.

Regulatory roles of VDAC and HKII. Figure layout scheme is the same as in Figure 2. Middle row of the scheme depicts the basal state of VDACs (WT and genetically deleted) and HKII, and their relationship to the basal state and oxidant stress-induced mPTP. mPTP is similar in WT controls as well as in mitochondria lacking VDACs, thus VDAC is dispensable for mPTP-induction. Top row represents facilitation of the ROS-induced mPTP by HKII-VBD peptide: however, because VDAC is dispensable, it is possible that HKII dissociation from some other site may be relevant. Note that this still does not prove that the HKII dissociation from any site by HKII-VBD peptide is casually related to facilitation of mPTP-induction. Bottom row: GSK-3β inhibition results in VDAC dephosphorylation which has been linked with cell protection,^81,83 so it remains possible that VDAC phosphorylation may be involved in mPTP regulation.

5. Hexokinases

As a key glycolytic enzyme, HK provides a functional link between mitochondria and cytosol by governing the preferential utilization of mitochondrial ATP for glucose phosphorylation. In the late 1970s and early 1980s it was demonstrated that the unknown VDAC-binding protein is indeed the HK^84,85 and since then this glycolytic enzyme has been recognized as the main VDAC modulator. The current standing regarding the role of HKII, one of the mitochondrial-bound isoforms, in mPTP regulation is summarized in Figure 3.

Mammalian cells express four isoforms of HK: types I, II, III, and IV (glucokinase),⁸⁶ all of which exhibit different affinities for ATP and glucose. Multiple isoforms often coexist in a given tissue. The N-terminal sequence of the mitochondrial HKI and II has high affinity to VDAC,⁸⁷ whereas HKIII and IV (cytosolic, preferentially expressed in non-muscle cells) lack affinity for VDAC. Mitochondria-bound isoform I is the ‘housekeeping’ enzyme expressed in all mammalian tissues. HKII, highly expressed in cardiac and skeletal muscle, appears to represent the principal regulated (e.g. insulin-responsive) isoform in many cell types. HKII expression is significantly increased in rapidly growing cancer cells where it is involved in the increased rate of glycolysis. It has been proposed that HKI is tightly bound not only to VDAC, but also specifically to VDAC located in the ‘contact sites’.⁸⁸ While HKI exists essentially in a mitochondria-bound form, HKII has been found both in mitochondria-bound as well as in soluble cytosolic form. Wilson speculated that HKII may also play an anabolic role by generating glucose-6-phosphate that is utilized in the pentose phosphate pathway to supply NADPH for lipid synthesis.⁸⁹

Glucose and ATP act both as catalytic substrates and modulators of HK-VDAC binding. Considering the ability of the mitochondrial ‘contact sites’ to modulate the amount of bound HK in response to mitochondrial metabolic state⁷³ manipulations of the metabolic state (for instance, via applying deoxyglucose) can have a dramatic effect on both the HK catalytic reaction rate and overall cell metabolism.

Three of the HK isoforms (I, II, and III) are inhibited by the product of their catalysis, glucose-6-phosphate, while type IV is insensitive to this product.⁹⁰ The reaction product of the first three isoforms is also capable of releasing HK from mitochondria. The ‘solubilizing’ activity of glucose-6-phosphate is highly variable and seems to be species dependent, which may reflect the evolution of multiple VDAC isoforms with different affinities for HK binding.⁹¹

In addition, different drugs may facilitate the release of HK from mitochondria, thereby mimicking the natural effect of glucose-6-phosphate. Antifungal drugs such as clotrimazol and bifonazol are effective hexokinase-releasing agents.⁹² Similar properties were assigned to the peptides corresponding to the N-terminal 29 AA residue domain common for both the HKI and HKII (HKII-VDB; VDAC-binding domain).^93,94 Recent findings indicate that forced release of HKII by this peptide invokes a cell death signal that engages mPTP opening.⁸² Surprisingly, this peptide was also fully effective in VDAC1/3−/− fibroblasts (where the VDAC2 isoform was also found not to be bound to HKII) proving the complete dispensability of VDACs for the induction of apoptosis in these cells (see previous section). Thus, HKII detachment from another site might be the signal to initiate apoptosis by relay of a conformational change from OMM to IMM and to components/modulators of the mPTP, independently of VDAC's presence.

6. Benzodiazepine receptor

Attempts to identify the nature of benzodiazepine binding to the OMM resulted in the discovery of a mitochondrial peripheral-type benzodiazepine receptor which is also known as the 18 kDa translocator protein (TspO).⁹⁵ TspO binds a number of ligands of varying nature,⁹⁶ but its physiological role remains obscure. Some speculative functions assigned to this receptor include, for example, steroid transport,⁹⁷ porphyrin transport,⁹⁸ and oxygen (redox) sensing.⁹⁹ TspO can form a multimeric complex with VDAC and ANT⁹⁵ in the OMM. Additionally, patch clamping studies have revealed that a Ca²⁺-dependent mitochondrial multi-conductance channel, presumably the mPTP, was equally sensitive to CsA⁶ and to the TspO ligand, Ro5-4864,¹⁰⁰ thereby demonstrating that TspO may be relevant to the mPTP entity or its regulation. Indeed, a proteinaceous supercomplex isolated from mitochondrial ‘contact sites’ demonstrated catalytic activity sensitive to this TspO ligand.³¹ In light of these findings, several tentative models of the mPTP complex have included TspO/PBR as a module attached to VDAC (Figure 1).

Among many natural TspO ligands, the putative endogenous diazepam-binding inhibitor DBI, an 11 kDa polypeptide of 86 AA, is well known.¹⁰¹ DBI could be cleaved into several biologically active fragments, including triakontatetra-neuropeptide (TTN; AA 17–50), octadeca-neuropeptide (ODN; AA 33–50), and octapeptide (OP; AA 43–50) that also act as TspO ligands.^102,103 The beneficial effect of TspO ligands in the prevention of ischaemia/reperfusion-induced damage is well established.^104,105 In a recent report, the TspO ligand, Ro5-4864, administered at the onset of reperfusion, protected against post-ischaemic contractile impairment, including reperfusion arrhythmias, probably by acting on a target which is located upstream of mPTP¹⁰⁶ thus suggesting that TspO is probably a regulator, but likely not an mPTP component.

7. Mitochondrial creatine kinase

Mitochondrial creatine kinase (mtCK), located in the mitochondrial intermembrane space, catalyses reversible transfer of the phosphoryl group from phospho-creatine to ADP. mtCK is encoded by two different genes that are expressed in a tissue-specific manner: the sarcomeric isoform is expressed in heart and skeletal muscles, whereas the ubiquitous isoform is expressed in smooth muscle, the brain, kidney, and other tissues; however, no expression was detected for liver and lung.¹⁰⁷

Isolated mtCK exhibits an association/dissociation equilibrium of preferentially octameric and dimeric structures.¹⁰⁸ mtCK has been studied mostly in in vitro systems that typically rely on the biophysical examination of purified mtCK and its interaction with model membranes. In these experiments the cube-like octameric mtCK formed close contacts between the two membranes.¹⁰⁹ Indeed, it has been demonstrated that octameric mtCK was enriched in isolated ‘contact sites’.¹⁰⁸ mtCK forms high-molecular-weight complexes also containing VDAC and ANT, and may provide the structural basis for functional interaction between the components of the two membranes. Because of these factors, it has been proposed that mtCK may be involved in mPTP regulation.²⁹

This enzyme is very sensitive due to the low tolerance of its sulfhydryl groups to oxidative stress.¹¹⁰ This sensitivity is thought to be the mechanism by which ROS induces a shift of the equilibrium between the two mtCK oligomer structures towards the dimeric form.¹¹¹ Correspondingly, in cells containing mtCK, it has been postulated that the mPTP threshold diminished by oxygen radicals is possibly caused at least in part by the vulnerability of mtCK to oxidant stress. Thus, under these circumstances, a shift towards the octameric structure of mtCK could result in an acquisition of a high tolerance to mPTP induction, whereas a shift to the opposite mode (towards the mtCK dimeric structure) may favour mPTP-induction.¹¹² Mitochondria isolated from transgenic mice expressing ubiquitous mtCK in liver (which expresses no CK in WT) were not different vs. WT controls in their ability to undergo (Ca²⁺+atractylate)-induced mPTP; however, creatine or cyclocreatine, stabilizing mtCK in an ANT-bound octameric form, inhibited mPTP induction only in transgenic mice.¹¹² Thus, the mtCK could also potentially be involved in the mPTP regulation.

8. Essential lipids

Boundary lipids are also known to play an important role in mPTP regulation. The essential role of cardiolipin, a hallmark phospholipid of the IMM and mitochondrial ‘contact sites’, has long been recognized. Cardiolipin is tightly bound to ANT and is required for the ATP/ADP exchange activity of ANT reconstituted into bilayer membranes.¹¹³ Ca²⁺ exhibits a very high affinity for cardiolipin, and displaces this lipid from ANT upon binding, thereby promoting its conversion into a ‘pore-forming’ m-conformation. The relevance of Ca²⁺ to mPTP induction in situ seems to depend on cell type: although strict Ca²⁺ dependence of the mPTP is characteristic of isolated mitochondria and probably many types of non-excitable cells, it is essentially absent in cardiomyocytes and neurons (see below).

It has been shown that the proapoptotic factor, tBid, which contains a BH3 domain, induces cardiolipin-dependent release of cyctochrome c after translocation to mitochondria.¹¹⁴ tBid-induced changes in cardiolipin within the mitochondrial membranes result in modulation of the ANT activity.¹¹⁵ The data also suggest that inhibitors of peroxidase activity could be protective against cell death stimuli via the conservation of intact mitochondrial cardiolipin,¹¹⁶ and the control of ROS levels.

9. Other regulatory mechanisms

9.1. Calcium ions

Based on experiments with isolated mitochondria, Ca²⁺ was thought to be the most important factor in mPTP induction. Recently though, experiments in intact excitable cells such as cardiomyocytes and neurons demonstrated that the mPTP is largely insensitive to increased cytosolic Ca²⁺ (to levels exceeding an order of magnitude above basal), but becomes apparently very Ca²⁺ sensitive when these same mitochondria are separated from the normal cytoplasm. Thus, Ca²⁺ has probably been inappropriately ‘blamed’ for mPTP induction and subsequent cell death in cells subjected to injury-producing stresses such as hypoxia/reoxygenation in cardiomyocytes or exitotoxic stress in neurons^9,25 and discussed in online Supplementary material.¹⁰ Rather, oxidant stress mechanisms appear to be the dominant factors responsible for mPTP induction due to stress injury in excitable cells. While Ca²⁺ might be of little relevance as far as mPTP induction in excitable cells, the role of Ca²⁺ in the mPTP induction remains important in a variety of intact non-excitable cells. Indeed, recently, it has been demonstrated that the antibiotic minocycline protected against mPTP induction and mitochondrial injury during rat liver transplantation, which in turn preserved graft function by inhibition of mitochondrial Ca²⁺ uptake.¹¹⁷ Interestingly, the difference in Ca²⁺ sensitivity of the mPTP between these two categories of cells is abolished by gentle cell permeabilization procedures¹⁰—all cell types demonstrated Ca²⁺ sensitivity of the mPTP after permeabilization. The Ca²⁺ dependence of the mPTP after cell permeabilization or isolation of mitochondria may be explained by the existence of a yet-to-be defined soluble factor(s) involved in regulating Ca²⁺ sensitivity which is lost during cell permeabilization/mitochondrial isolation.^9,10

9.2. Mitochondrial phosphate carrier

Recently, a new model for the mPTP has been suggested in which the mitochondrial phosphate carrier (PiC, which would provide the pore-forming component instead of the ANT), facilitated by CyP-D and possibly the interaction with the ANT and it ligands, initiates the Ca²⁺-induced conformational change that triggers the mPTP. This conclusion was based, in part, on findings that PiC binds Cyp-D in a CsA-sensitive manner,¹¹⁸ and was supported by the observation that PiC silencing dramatically delays cytochrome c mobilization and apoptosis.¹¹⁹ Although it is tempting to make the association, the complex effects of P_i on mPTP induction (see above) do not necessarily implicate the PiC as a constituent of the mPTP (as discussed by Basso et al.⁶⁰). This issue of the PiC remains a subject of active debate and deserves further attention and research.

9.3. Acidosis

In vitro experiments with isolated mitochondria demonstrated for the first time that a decrease in pH delays mPTP opening.² The protective effect of acidosis, which seems to act directly or in proximity to the mPTP rather than through upstream (e.g. kinase) signalling, was later confirmed in vivo in intact cells⁹ and in situ in hearts exposed to ischaemia/reoxygenation applied at reperfusion.¹²⁰ These data suggest that careful, catheter-based approaches directing controlled acidosis to the infarct zone at the point of reperfusion could have therapeutic promise as a postconditioning strategy in the treatment of patients with acute myocardial infarction.

9.4. Protein kinases and phosphatases

Ischemic pre- and postconditioning activate endogenous signalling mechanisms resulting in the most potent forms of protection capable of reducing cell death following prolonged periods of ischaemia (e.g. as a consequence of arterial occlusion). Activation of these endogenous mechanisms can be triggered by brief episodes of transient ischaemia and reperfusion preceding a prolonged ischaemic insult (i.e. preconditioning; originally demonstrated by Murry et al.¹²¹) or repetitive ischaemia applied during early reperfusion (i.e. postconditioning (see Zhao et al.,¹²² for comprehensive reviews, and also Skyschally et al.¹²³ and Yellon and Downey¹²⁵). Similar levels of protection could be achieved by pharmacological activation of variety of surface receptors and their downstream signalling pathways including the mitochondrial ATP-dependent K⁺ channel (mitoKATP) (see below) and kinases such as PKA, PKB/Akt, PKC, PKG, and glycogen synthase kinase-3β (GSK-3β) (reviewed in Juhaszova et al.^26,124 and Yellon and Downey¹²⁵). GSK-3β was found to serve as a critical convergence point of these signalling pathways, and in turn was responsible for conveying the signals downstream to mPTP components/modulators;¹⁰ reviewed in Juhaszova et al.¹²⁴ The role of GSK-3β in cell protection was found to be mediated by the mPTP.^10,126 Inhibition of GSK-3β by phosphorylation on Ser⁹ resulted in an increased mPTP–ROS threshold and consequent cell protection. Therefore, the application of reversible GSK-3 inhibitors has the potential to be a useful cell protection strategy (Figure 4).

Figure 4.

Proposed model of mPTP modulation by GSK-3β. The phosphorylation state of the mitochondrial-(ANT)-associated pool of GSK-3β contributes to the balance of Bcl-2 family protein effects, the result of which determines the resistance of the mPTP to oxidant stress. Basal state (top): the local GSK-3β pool is active, binds ANT in a complex with phosphorylated VDAC, CyP-D, and possibly other mPTP regulatory elements. ‘Anti-apoptotic’ BH4- and the ‘de-repressor’ BH3-only domain Bcl-2 protein family members are held in mutual check. Protection state (bottom): induced by phosphorylation/inactivation of this mitochondrial GSK-3β subdomain pool resulting in a shift in the balance between Bcl-2 family members toward unmasking the activity of ‘anti-apoptotic’ Bcl-2 family members (and possibly a shift in the interaction between the mPTP and other of its regulators) and consequently in protection of the mPTP against oxidant stress injury (modified from Juhaszova et al.⁹).

An in silico search for GSK-3β phosphorylation motifs within proteins involved in protection yielded several potential candidates, including Bcl-2, Bis, and protein phosphatase 2A.^9,10 Experimental data demonstrated that GSK-3β also phosphorylates VDAC, which leads to a decrease in HKII binding to mitochondria, and inhibition of the ATP influx into mitochondria^81,83 (Figures 3 and 4).

The putative mitoKATP has also been proposed to play a role as a mediator or a possible end-effector of protection signalling (we think the latter role is unlikely). Ischaemic pre/postconditioning and pharmacological activation of the mitoKATP has been shown to induce a protective state sensitive to inhibition by mitoKATP blockers such as 5-hydroxydecanoate (5-HD) and glybenclamide.^127–129 In isolated cardiomyocytes mitoKATP activation-induced mitochondrial K⁺ influx has been associated with regulatory mitochondrial swelling (seen also in isolated mitochondrial suspension¹³⁰), an increase in oxygen consumption, and the production of a redox signal; furthermore, these actions of diazoxide were abolished not only by 5-HD but also by the ROS scavenger N-acetylcysteine, implicating redox signal in certain types of conditioning processes.^10,131 Subsequent redox activation of PKC in mitochondria has been demonstrated.^10,132 This PKC signal, together with the mitochondrial volume activated respiration and resulting redox signals, has been implicated in maintaining the memory of the conditioning process in a positive feedback loop by sustaining the activation state of the mitoKATP.¹⁰ Preconditioning-induced translocation of PKCϵ into the mitochondria (and/or activation of resident PKCϵ) of cardiomyocytes resulted in cell protection.^10,133,134 The opposing roles of protein kinase C (PKC) isoforms, specifically PKCϵ and PKCδ, in cell protection are relatively well defined for heart tissue.¹³³ PKCϵ has been demonstrated to physically associate with the mPTP modulators ANT and VDAC (the latter's status as an mPTP modulator being questionable), and to phosphorylate VDAC in heart mitochondria.¹³⁵

The general role of kinases in mPTP protection has even been questioned recently.¹³⁶ Based on the analysis of the phosphoprotein levels in the mitochondria isolated from preconditioned hearts, it was concluded that protein kinase-mediated changes in mitochondrial phosphoprotein levels may not be required for the preconditioning-induced mPTP inhibition, but rather that preconditioning-induced reduction in oxidative stress is responsible for mPTP protection. It has been noted, however, that experimental conditions strongly influence the detection of protein phosphorylation status, so real changes may be artifactually lost.⁸³ Therefore, in light of the extensive existing evidence to the contrary, it would seem that kinases should continue to be recognized as important regulatory elements in cell protection (until and unless considerable new evidence to the contrary could be obtained).

9.5. Bcl-2 family proteins

Early on, in vitro studies of the interaction between purified Bcl-2 or BAX, and ANT in artificial lipid bilayers suggested that Bcl-2 and Bax may form with ANT channels with different behaviours. In contrast to Bax, Bcl-2 does not form active channels when incorporated into membranes with ANT. Bcl-2 also blocks the ability of BAX to form channels in reconstitution experiments with ANT. These data suggested that Bcl-2 family proteins may play a regulatory role in ‘mPTP induction’.⁴⁹ Indeed, it has been demonstrated that a membrane permeable peptide carrying the BH4 domain of Bcl-x_L (TAT–BH4 fusion peptide) increases the threshold for mPTP induction in cardiomyocytes⁹ and reduced infarct size in the heart after ischaemia/reperfusion¹³⁷ while HA14-1, a small non-peptidic inhibitor of Bcl-2, abolished the protection afforded by TAT–BH4 and sensitized the mPTP to opening.^9,138 In addition, the BH3-domain peptide of Bad, which has a high affinity for Bcl-2, completely abolished protective effects of TAT-BH4, Li⁺, insulin, and an inhibitor of the Na⁺/H⁺ exchange without significantly affecting the basal mPTP–ROS threshold.⁹ These data suggest that specific functional groups of the Bcl-2 family proteins, the ‘de-repressor’ BH3-domain-only proteins (e.g. Bad, Bnip3, Puma, etc.) together with the so-called ‘anti-apoptotic’ BH4-domain carrying proteins (e.g. Bcl-2, Bcl-x_L, etc.), acting downstream of GSK-3β, appear to be able to mediate (or at least regulate) cell-protective or protection-inhibiting functions via interaction directly with the mPTP (and/or its other regulatory elements). Specifically, in the basal state, the so-called ‘anti-apoptotic’ members of the Bcl-2 family of proteins would be held in check in a functional sense by ‘de-repressors’. According to this scheme, GSK-3β in the active (basal) state may participate by modulating the interaction between regulatory components of the mPTP, i.e. ANT and CyP-D, as well as the functional interaction and binding of various Bcl-2 family proteins. These interactions could be important in maintaining a balance between opposing influences of these members of the Bcl-2 protein family, thus producing the basal state of resistance to mPTP induction. The protection state would be associated with the phosphorylation and therefore inactivation of GSK-3β, which would lead to changes in interaction between mPTP components/regulatory elements, inducing a shift in the balance within the Bcl-2 family of proteins in favour of ‘anti-apoptotic’ proteins, resulting in an increase in the mPTP–ROS threshold (see Figure 4).

9.6. Ubiquinone 0

Fontaine et al.¹³⁹ demonstrated the existence of a quinone-binding site on the mPTP and initially proposed that mitochondrial complex I, which has two quinone-binding sites, might be a structural element of the mPTP. They also demonstrated that ubiquinone 0 is a potent inhibitor of the mPTP, although the quinone-binding site(s) involved in mPTP protection remains a subject of debate. However, the potential application of ubiquinone 0 as mPTP inhibitor in an in vivo setting is limited due to its unspecific toxic effects on mitochondrial respiration.¹⁴⁰

10. Concluding remarks: mPTP with no defined core?

Despite extensive research efforts over several decades, at this point the essential core component(s) of the mPTP responsible for IMM permeabilization remain unidentified. It is likely that none of the elements discussed above form the principal structure of the mPTP; they most likely act only as modulators/regulators. In view of the new genetic data, the models of the physical entity of the mPTP complex have been forced to undergo extensive revision and correction. Therefore, based on careful re-evaluation of available data, it is possible to conclude that components once recognized as the core structural (pore-forming) elements of the mPTP (namely, mitochondrial ANT, cyclophilin D and TspO) are instead only mPTP modulators/regulators (or, in the case of VDAC, probably entirely dispensable). Additional important protein regulators include the Bcl-2 family and certain kinase signalling cascades, including those converging and acting via a mitochondrial-associated pool of GSK-3β. Recently, a novel preliminary model for the mPTP has been hypothesized in which the PiC would provide the pore-forming component that deserves further consideration. Finally, important small molecule/ion regulation is obtained from species requiring CyP-D (such as CsA, Ca²⁺ and P_i), those requiring the ANT (e.g. ADP, BKA, and atractyloside), or from those acting at sites that remain to be defined (e.g. accompanying acidosis).

Currently, the minimum number of components involved in the mPTP phenomenon is not known. Under specific circumstances it is possible that different mPTPs, formed by different clusters of proteins, may exist. It is not known whether just one type of mPTP can be formed in a particular cell type, or whether diverse triggers result in the same or various mPTPs, or if recruitment of different mPTP elements result in different behaviours, e.g. resulting in different thresholds to Ca²⁺, ROS, and exhibiting sensitivities to CsA, etc. This phenomenon may even involve different states with very subtle differences in the overall behaviour. Additionally, since a single mPTP can likely discharge the ΔΨ_m, it is possible if not likely that the assembly of the mPTP, from its constituent components and its induction, is a rare event, and that it could be assembled from components, which while relatively abundant, generate only a single macromolecular cluster and that exists only transiently. Multiple regulatory entities assembled around the main core components resulting in somewhat different behaviours in different cells may exist. This may partially explain the appearance of the terms ‘regulated’ and ‘unregulated’ to describe mPTP, or the in vivo vs. in vitro differences in Ca²⁺ sensitivities, and thereby the inherent complexities in observing the obscure nature of this phenomenon.

What is clear, however, is that the mPTP phenomenon plays a key role in the ‘life or death’ decision of the cell. In addition to the well-established role of the mPTP induction in ischaemia/reperfusion injury, recently the mPTP has also been shown to play an important role in pathologies such as Ulrich congenital muscular dystrophy, Bethlem myopathy,⁶⁶ etc. (reviewed in Rasola and Bernardi¹⁴¹) and is likely to be discovered as playing fundamental roles in many other diseases, in the future. Finally, as we have discussed, the ageing process impairs the intrinsic capacity of the heart (and probably other vital organs)—and specifically that of the mPTP—to resist damaging stress such as excess oxidants and ischaemia/reperfusion injury; these effects are likely aggravated by comorbidities such as diabetes, but may be ameliorated somewhat by caloric restriction and physical conditioning/exercise. There are also considerable data that the ability to induce ischaemic preconditioning in aged animals and humans is impaired, and in particular, that the stimulus threshold for triggering protection increases with age.²⁷ Thus, there needs to be an appreciation that clinical studies on elderly subjects have been hampered by the added complications arising due to the fact that cardioprotection signalling is impaired in the old heart and that the design of future clinical trials testing these questions must take this into consideration (specifically, that it may take a stronger stimulus to activate protection in aging individuals compared to young subjects). These problems with endogenous protection signalling are also potentially aggravated by commonly used medications including sulfonylureas, antioxidants, PFOX inhibitors, and COX2 inhibitors (but also potentially improved by caloric restriction and physical conditioning/exercise), so it would be prudent for future studies to critically examine drugs taken by the elderly for potential interference with endogenous protection signalling (see Juhaszova et al.²⁶ for review, and citation therein). Postconditioning (both by means of the balloon catheter as well as by pharmacology) to reduce the size of myocardial infarction appears to be promising in small cohorts of carefully chosen patients,^65,142 but it would be important to examine if efficacy could be demonstrated in the elderly. Some of the age-related impairments in protection signalling could remain unresponsive to lifestyle/diet measures. Pharmacological approaches aimed at downstream targets (such as GSK-3 and the mPTP), to bypass potentially defective upstream signalling components, might provide a rational strategy to further restore protection. Therefore, a better understanding of the mPTP and its modulation will allow identification of possible pharmacological targets and assist in drug design for its precise control.

Conflict of interest: none declared.

Funding

This research was entirely supported by the Intramural Research Program of the National Institutes of Health, National Institute on Aging.