Since the 1950s it is known that Ca2+-uptake can induce an almost complete loss of the mitochondrial membrane potential (ΔΨ) (1). Ca2+ and inorganic phosphate can induce large scale swelling as well (2–4). Hunter et al. called the basis of this phenomenon a “calcium-induced configurational transition from the aggregated to the orthodox state going in parallel with increased permeability” (5). Now, 36 years later, the term “permeability transition” (PT) has become established for this increase of permeability of both, the inner (IM) and the outer mitochondrial membrane (OM). There is good evidence that the increase at the IM is caused by a very large pore of unknown molecular structure dubbed permeability transition pore (PTP). The increase at the OM is additional to the permeability due to the voltage dependent anion channel (VDAC) that seems to be always present. PT was shown to be involved in different types of cell death (6) and possibly it offers various ways to pharmacologically interrupt the pathways leading to cell death (7). In spite of enormous interest in the structural basis of PT it seems far from being understood. Even some of the elements involved are still subject to speculation.
Functional Facts Known About PT
A drastically increased Ca2+-concentration definitely seems to be a key event in PT causing opening of the PTP. In non-neuronal mitochondria a Ca2+-concentration elevated beyond a threshold of about 100 μmole/mg protein causes several physiological changes attributed to PT like (i) the loss of ΔΨ (due to the loss of impermeability of the IM), (ii) mitochondrial swelling (at least in isolated mitochondria due to osmotic uptake of H2O), (iii) a reduced Ca2+-retention capacity (due to the loss of ΔΨ as a consequence of a limited substrate supply or impairments of the respiratory chain), (iv) an increased permeability for solutes smaller than about 1,500 Da including substrates of the respiratory chain, and (v) release of proapoptotic factors into the cytosol after OM rupture due to osmotic swelling (8, 9). As there are obvious effects of PT both, on OM and on IM, and as in contrast to possible membrane rupture of the OM a pore (PTP) is assumed in the IM only, we and others consider it useful to clearly separate these two terms. Nevertheless, the reader has to be aware that they may be differently used in literature. Increased permeability can cause, for example, a loss of endogenous NAD+ which as a complex I-dependent coenzyme has been used as an indicator of PT (10). It was shown in yeast mitochondria that it is no problem for NAD+ to cross the OM via VDAC (11). Thus, some of these events can clearly be attributed to the OM, some to the IM, and some to changes in both membranes. The situation in neurons may differ from that in non-neuronal cells in that a lower increase of extramitrochondrial [Ca2+] (>1 μM) may cause apoptotic cell death seemingly without causing opening of the PTP (12, 13).
There are indications that Ca2+ is acting from the inner side of the IM for inducing PT, even if it is applied from the cytosol. Thus, there is a very efficient movement of Ca2+-ions across both membranes. As it will be discussed, it is certainly no problem to cross the OM. However, electrophoretic transport of Ca2+-ions across the IM as a prerequisite of PT was explained only recently as being attributed to a voltage dependent Ca-channel and a Ca2+/H+ antiporter (14–17). These observations are of significance as it was discussed already in the 1990s that the PTP may adopt a low-conductance Ca2+-selective state under certain conditions (18, 19). It must be stressed that substances other than Ca2+, like inorganic phosphate, or fatty acids, are also able to induce PT and that there is a broad spectrum of inhibitors as well (20), for review: (9). Paticularly, the antibiotics cyclosporine A (CsA) and its nonimmunosuppressive derivative Cs9 were proven to be specific PT-inhibitors (21–24). There is agreement that CsA binds to the matrix protein cyclophiline D (Cyp-D), a peptidyl-prolyl cis-trans isomerase that is modulating the PTP at the matrix side of the IM (25).
Separation of Events from OM from Events from IM in Apoptosis
Mitochondrial research has gained considerable interest in the 1990s again when it was discovered that the organelles are significantly involved in programmed cell death. It is well established now that the intrinsic signalling chain of apoptosis, which includes the mitochondria, contains proteins of the Bcl 2-family linking events in the plasma membrane and cytosolic events to those in the mitochondria (26). Stimulatory members of the Bcl 2-family are causing changes of the OM enabling the release of several proapoptotic factors from the intermembrane space (8), review: (27). The nature of these alterations is still a matter of debate. There is general agreement, however, that the presence of a high phosphorylation potential is a prerequisite for apoptosis while at a low phosphorylation potential necrosis will happen (28).
Two classes of the Bcl 2-family members are distinguished, the stimulating proteins like Bax, BAD, and Bak on one side and the inhibiting proteins like Bcl 2 itself, Bcl-xL on the other. All these are thought to primarily act at the OM, promoting or preventing the release of proapoptotic factors like cytochrome c, apoptosis inducing factor (AIF), Smac/DIABLO and others, review: (27). It has been shown that two distinct mechanisms may lead to cytochrome c release: one stimulated by Ca2+ and inhibited by CsA, the other Bax dependent, Mg2+ sensitive but CsA-insensitive (29). However, it is still enigmatic how the proapoptotic factors are released across the OM; and, it is not clear, as well, how the events at the IM are modulated by the proteins of the Bcl 2-family though it was shown that they are (30, 31).
Functional Test of the PTP by Single-Channel Measurement
A useful test for an experimentally modulated PTP was discovered in three different labs independently (32–34). It is possible to record from mitoplasts, that is, from vesicles of the IM, single channel events fulfilling all criteria required of a PTP or, as it was also called, mitochondrial megachannel (MMC). These criteria are primarily Ca2+-activation, CsA-blockade, and an increased open probability during depolarization, which were demonstrated by dose–response curves (34–36). Additionally, a very large single channel conductance of >1 nS and smaller substates of the fully open state with independent kinetics must be visible. In particular, voltage dependence should differ from the bell shaped behaviour described for the VDAC (37).
It turned out that the observed single-channel recordings allowed testing of the PTP for the action of biochemicals, for example, (38). Bcl 2-family members can also affect ion channels of the IM like the PTP, Kv1.3 or the Ca2+-activated potassium channel (BK-type) that is able to modulate the PTP (31, 39–41). Also in this case, detailed knowledge about the precise mechanism is missing, particularly about transduction of the signal between OM and IM.
There were early indications that Bax and Bcl 2 showed channel characteristics in their structure and could thus by themselves be responsible for permeabilization of the OM (42–44). To our knowledge functional channels formed of Bcl 2-family proteins in functional mitochondria has never been really proven. Instead, it was found that compounds blocking the release of fluorescein from liposomes by recombinant Bax were able to directly close a mitochondrial OM channel in FL5.12 cells named mitochondrial apoptosis-inducing channel (MAC) (45). The cytochrome-c release was blocked as well. Further, the upstream cascade of apoptosis was shown to activate acid sphingomyelinase in T-lymphocytes stimulating ceramide synthesis. Ceramide can form ion channels by itself by self-assembly, a process that can be strongly influenced by Bcl 2-family proteins (46).
It was thought for a while that the VDAC would be responsible for the release because VDAC could be stimulated and inhibited by proteins of the Bcl 2-family and because anti-VDAC antibodies inhibited the Ca2+-induced PT (47–49). Additionally, it had been found that modulation of PT at the IM could be achieved by substances modulating the VDAC, as well (reviewed by ref.50), and that purified VDAC and adenine nucleotide translocator (ANT) were able to permeabilize liposomes in the presence of Cyp-D (25). This effect could be reversed by CsA. It had been also suggested that the contact sites between OM and IM could be the favoured sites of PT because it was thought to be most likely that proteins located within the contact sites would allow a coordinated permeabilization of both membranes (51). Again, VDAC in the OM and ANT in the IM were most suspicious to be involved. Hexokinase and creatin kinase at the OM were thought to be important, additionally (51–53). However, also doubts are raised that purified hexokinase, VDAC, and ANT would be responsible for the formation of CsA-inhibitable pores (9). Additionally, Bernardi's group could inhibit swelling of isolated mitochondria by Ro 68-3400, a high affinity inhibitor of PTP. In parallel, they found that Ro 68-3400 also labels one isoform of VDAC. However, the properties of the PTP in VDAC1−/− mitochondria were indistinguishable from those of wild-type mice, including inhibition by Ro 68-3400. It means that labeling of VDAC and the inhibition of the PTP are different effects. Therefore, the authors conclude that similarities in the properties of VDAC and PTP are not sufficient to conclude that VDAC is part of the PTP (9, 54, 55).
Colombini's group summarized their view about the role of VDAC in PT by stating that: VDAC is only present in the OM, VDAC functions as a monomer, normally with or without Ca2+, it does not mediate the flux of proteins (peptides and unfolded proteins are excluded from this statement). Closure of VDAC, not opening, leads to permeabilization of OM and apoptosis (56). However, the authors leave it open whether cytochrome c passes through a gap formed under contribution of VDAC together with Bax or if alternatively a closed VDAC would interrupt the purin nucleotide-exchange which would cause cytochrome c release by unidentified pores in the OM (see Fig. 3 in ref.56). Some of these points, for example, that VDAC is located in the OM and that it works as a monomer had been questioned on the basis of single channel experiments on mitoplasts by Szabó et al. (57). Finally, results with VDAC knock out mice indicate that VDACs are dispensable for both MPT and Bcl-2 family member-driven cell death (58).
Peripheral Benzodiazepine Receptor
The peripheral benzodiazepine receptor (PBR) has to be distinguished from the benzodiazepine receptor in the central nervous system. PBR is located in the OM and had been attributed to PT. It had been shown to be associated with VDAC and ANT (59, 60). In order to prove that, PBR was overexpressed in an E. coli expression system and reincorporated into liposomes (61). The PBR inserted into liposomes was able to bind both the PBR drug ligand isoquinoline carboxamide PK 11195 and cholesterol with nanomolar affinities. However, it was shown that neither VDAC nor ANT was a prerequisit for binding of the ligands. Instead, structural characterization of the protein indicated that the protein monomer is the minimal functional unit (61). Inhibition of the PTP (=MMC) of the IM by PBR ligands was reported from patch-clamp experiments on rat heart mitoplasts (62). However, it is not clear how the signal is transferred when the OM is missing in mitoplasts. In other studies, the PTP was rather promoted by PBR ligands. These discrepancies are reviewed in more detail by (9). An easy solution seems not available, however, if VDAC is not an essential part of the PTP, a modulating effect of the PBR or an effect on the expression of VDAC does not necessarily speak for PBR as part of the PTP (63).
How Do Apoptotic Factors Leave Mitochondria? Is PT Reversible?
It has always been a riddle how the apoptotic factors could cross the OM. Because of early suggestions, the release could be either the result of opening of a proteinaceous pore (43) or the result of membrane damage (64). Several of the facts summarized in 2.1 spoke in favor of a contribution of VDAC to the permeabilization of the OM (65). However, as outlined by Bernardi et al. it was difficult to understand how, for example, cytochrome c with a mole weight of about 12,000 Da should pass the VDAC said to pass substances up to a mole weight of 1,500 Da (9).
Four ways of release of proapoptotic factors from the mitochondria have been debated: (i) the release via any kind of known channel. As VDAC is too small, it was also thought that VDAC could contribute to a pore formed next to the channel and possibly under its contribution (see “Structural elements of PT at the OM” section). (ii) The Bcl 2-proteins could form channels (see “Structural elements of PT at the OM” section). (iii) PT-induced swelling could rupture the outer membrane (not the IM) and would release the factors by this way. (iv) Finally, the heat-shock proteins of the protein import machinery are spanning both membranes and could be involved. Whether one of these pathways is identical with the so-called apoptosis-induced channel (MAC) in the OM is unknown (66). A final decision has not been made. However, the debate has consequences for another important feature of PT, namely is it reversible?
For understanding the significance of the PT-related modifications it was of interest whether the permeability increase is irreversible. It had been observed in early experiments that PT induced swelling could be reversible (3), see also “Modulators of the PTP” section and (5). Later, Hausenloy et al. concluded from experiments on isolated perfused rat hearts and on adult rat myotubes that the PT had opened transiently to allow protection and that the opening was to a low conductance state and was induced by reactive oxygen species (18). By studying protection of the heart by pyruvate in the perfusion medium prior to ischemia and during reperfusion Kerr et al. (67) concluded that the MPTP can close again after it has opened. However, it is required that resealing follows rapidly. Thus, there is little doubt about reversibility of the IM related events of opening and closure of the PTP. Reversibility should be possible in principle but depends on the status of the modulators (Ca2+, Pi, and ATP). The advantage of such mechanism might be that at a PT-induced reduction of ΔΨ would cause an outward driving force for Ca2+ which could leave the mitochondria until ΔΨ is restored and the PTP closed again. After release of cytochrome c, however, there seems to be a point of no return, that is, that from hereon the cell is determined to die.
Structural Elements of PTP in the IM
Most discrepancies in the existing literature are concerning the structure of the PTP. It has been proposed that the “PTP-complex” is consisting of ANT, member(s) of the Bcl 2-family, and cyclophiline D and it could thus be blocked by CsA and a CsA-derivative as measured by ceased cytochrome c- or AIF-release and by bongkrekic acid (52). These three inhibitors also prevented the effects of Bax on ΔΨm reduction in isolated mitochondria. Bongkrekic acid is well known as an inhibitor of ANT (68). Fitting to these results, highly purified ANT from rat heart mitochondria was incorporated into artificial membrane vesicles where it was able to release trapped ATP, malate, or AMP after increasing the external [Ca2+] (69). > 20 μM ADP inhibited the release but N-methyl-Val-4-cyclosporin did not. So, it was concluded that ANT itself is a pore-like structure under conditions known to induce the PTP (69).
What appeared to be a dogma was seriously questioned when in mitochondria of mouse liver two isoforms of ANT were genetically inactivated. PTP activation in isolated mitochondria and the induction of cell death in hepatocytes was then analysed (70). It turned out that mitochondria lacking ANT could still undergo PT induced by various initiators of cell death and as indicated by the release of cytochrome c. Only, a higher [Ca2+] was required for activation of the PTP and the pore could no longer be regulated by ANT ligands (70). The conclusion would be that ANT may be a component of the PTP but not an essential one.
Another component of the PTP is the cyclophilin Cyp-D, a matrix-soluble protein that contains the binding site for CsA (25). Also for this protein it is controversially discussed whether its contribution to PT is essential or not. Tsujimoto et al. published that knockout mice devoid of Cyp D did not show Ca2+-activation of the CsA-sensitive PTP thus seemingly demonstrating that Cyp D is essential for the occurence of PT (71); review: (72). They found some tissue differences. Furthermore, it has been stated for heart mitochondria that cyclophilin D and the PT are required for mediating Ca2+- and oxidative damage-induced cell death, but not Bcl-2 family member-regulated death (73). However, it was reported that there was still Ca2+-stimulation of PT in liver mitochondria though two times more Ca2+ was required to open the pore, measured as the Ca2+-retention capacitiy (CRC) (74). CsA-sensitivity was missing in their experiments after Cyp-D ablation. In our experiments on liver mitoplasts of ko-mice the PTP did not become completely CsA-insensitive. The dose–response curve was shifted to higher concentrations, however, by 1.5 orders of magnitude as compared with liver mitochondria of wild-type mice and of rat (Fig. 1). The conclusion is that Cyp-D considerably sensitizes the PTP for CsA, however, at higher concentrations CsA may block the PTP even in the absence of Cyp-D.
It has been demonstrated that PK11195, an isoquinoline carboximide, and Bz-423 (1,4-benzodiazepine), both high affinity ligands of the PBR, can bind in micromolar concentrations to the oligomycin sensitivity conferring protein (OSCP) of the mitochondrial F1F0-ATPase (75, 76). The results suggest that the F1F0-ATPase is inhibited by these substances and some of the results on PBR mentioned in “Structural elements of PT at the OM” section may gain a new significance, for example, the results about inhibition of MMC by PBR ligands (62).
Modulators of the PTP
Inorganic phoshate has been known as a PTP inducer since long (3, 4, 21). It appears, that inorganic phosphate can induce two kinds of swelling processes, one which can be reversed by the addition of uncoupling agents, respiratory chain inhibitors, or ADP, and another which can be reversed only by external ATP. We saw activation of the PTP by Pi in our patch-clamp experiments as well (77), see their Fig. 6]. In our hands, PTP activation by Pi can counteract CsA-inhibition and inhibition by the sythetical dopamine agonist ropinirole. Phosphate activation of the MPTP opening could involve the phosphate carrier (PiC) that was proposed as the pore's cyclophilin-D binding component (78) or could even be converted into a channel by itself (79). However, the specificity of Pi over arsenate and vanadate in PTP desensitization represents an argument against the suggestion that the PiC might be involved in Pi effects on the PTP or in its composition (80). Others have reported that following CyP-D ablation Pi inhibits PTP opening directly while CsA inhibits PTP opening only when Pi is present (80, 81). However, Varanyuwatana and Halestrap demonstrate that Pi activates PTP opening under all energized and de-energized conditions tested while CsA inhibits pore opening whether or not Pi is present. Using siRNA in HeLa cells they could reduce PiC expression by 65–80% but this inhibited neither mitochondrial Ca2+-accumulation nor PTP opening (78). The latter but not Fig. 1 in ref.80 is in agreement with our own patch-clamp results from liver mitoplasts of Ppif-/- mice where we found PTP opening and CsA block without any Pi added (Fig. 1).
A recent Cell paper reported that p53 would induce PTP-dependent necrosis by translocation of the inducible p53 to the matrix where it would bind to CypD and open the PTP (82). However, only four months later another paper collects good arguments speaking against this hypothesis (83). Particularly, the independence of Ca2+, as regulator of the PTP, from p53 that was shown by Vaseva et al. (82) is speaking together with further arguments against this mechanism.
Is a Protein the Basis of PTP?
The question has been raised whether the PTP is performed by protein pores, at all. Large-conductance voltage-dependent ion channels have also been isolated by means of a chloroform extract of isolated, dehydrated rat liver mitochondria (84). They consisted of poly-3-hydroxybutyrate and calcium polyphosphate. The polyphosphate consists of hundreds of phosphates linked by ATP-like high-energy bonds (85). Several aspects of their behaviour resemble behaviour also exhibited by the PTP.
Pathogenic Consequences of PT Opening
Several reports have made it very likely that PT is deleterious for cells in diseases and that in many diseases cells die by mechanisms involving defects of mitochondria. This was found particularly in neurodegenerative diseases. There is increasing effort to cure or at least delay the progress of neurodegenerative diseases by inhibiting PT. However, also in this field attempts and ideas had to be rejected because basic hypotheses were experimentally falsified.
The antibiotic minocycline (MC) is an example for the difficulties to interpret the results of pharmacological treatment of the PTP correctly. There were seemingly clear indications that MC could inhibit the PTP and it therefore could be an ideal tool to suppress motor neuron decline in amyotrophic lateral sclerosis (86, 87). However, later it turned out that the matter was complicated by multiple actions of MC partly leading to detrimental effects so that even a major clinical trial had to be cancelled (87, 88). In rat liver mitochondria MC turned out to induce swelling by removing Mg2+ from the matrix thereby opening K+- and Cl−-pathways which can be blocked by N,N′-dicyclohexylcarbodiimide and tributyltin. This swelling occured in a KCl medium only, however, not in a sucrose medium. It was accompanied by mitochondrial depolarization, a reduced calcium retention capacity, and a decrease of state 3 and uncoupled respiration (88).
Fortunately, a pessimistic view is not the end of the story about the PTP. Progress might be possible following four different strategies or a combination of them. These strategies include (i) studying function of the PTP in more detail, (ii) discovering more details about structure, (iii) introduction of genetic modification or ablation of proteins suspected to be involved, and finally, (iv) studying mitochondrial diseases and pharmacological approaches to cure them. A functional property that will be characterized quite possibly in near future by single channel measurement is the pretended low-conductance Ca2+-selective state of the PTP which will have implications for mitochondria as part of the Ca2+-regulating system of the cell. A functional feature of PT that needs further explanation is the role of the space between the mitochondrial cristae for release of, for example, cytochrome c and the way it is remodeled (89). During fusion/fission of mitochondria and cristae remodeling profound alterations of the IM and the contained proteins will take place which may even affect the outer membrane (89); for review: (90). Considerable details of the protein machinery involved, have been reviewed (91). The dynamin-related protein optic atrophy 1 (OPA1), associated with the IM, can control cristae remodeling even independently from fusion, thereby inhibiting cytochrome c release and protecting from apoptosis (92). The authors conclude that OPA1 oligomers participate in the formation of the cristae junctions and would thus have implications for the mitochondrial structure. To our knowledge, a relation to PT has not been proved but seems likely. Many of the earlier discussed and then rejected hypotheses described here could be ruled out by genetic ablation of the very protein. Unfortunately, some of the proteins exist in several isoforms which have to be studied seperately or all together if not being lethal. Nevertheless, we owe enormous progress to this method and as soon as the gene(s) for the PTP are identified point mutations should help to characterize it in detail.
A wealth of literature exists on ischemia-reperfusion with the consequence that it is now much better understood. Limited space does not allow going into detail here. Also, in neuroprotection progress was made, demonstrating that the neuroprotectants melatonin in a stroke model, and pramipexol and ropinirole as dopamine agonists, well introduced in Parkinson treatment, are inhibitors of the PTP (36, 38, 77). Patients treated with the dopamine agonists may gain several months when comparing the decline of their motor scores with the decline of L-dopa treated patients. Together with CsA and Cs9 which are best known to really block the PTP, these substances seem to be really important as starting points for screening related substances for suitable new drugs and as possible tools for understanding the basic mechanism of PT (93).
When Bernardi et al. wrote their review in 2006 they found close to 2,000 publications about the PTP in Medline. Now, 2012, we found 3,434 with the steepest increase in number in the 1990s. These numbers point to the great attention this pore has attracted in the last decades. Thus, it is really surprising that no conclusive model of the pore is existing. However, it is promising that only recently the nature of two mitochondrial channels, the Ca2+-uniport and the ATP-sensitive K-channel was discovered after a long-lasting search. To date it looks as if ANT, Cyp D, VDAC, PBR, and Hexokinase are not essential parts of the PTP, though they can modulate PT/PTP. Even the question whether Pi or p53 are stimulating the PTP, there were contradicting answers published. Maybe looking at the F1F0-ATPase in more detail could be a promising approach. It seems high time to solve the problem of the unknown structure of the PTP and research is on promising ways. Knowing its molecular structure would be a basis for screening available drugs and for an easier development of new pharmaceuticals particularly in the fields of cancer and neurodegeneration.
The authors are indebted to Frank Gellerich for reading the manuscript, C. Höhne, K. Kaiser, and J. Witzke for technical assistance and to J. Molkentin, University of Cincinnati, for the Ppif-/--mice. Financial support by Deutsches Zentrum für Neurodegenerative Erkrankungen (DZNE) is greatfully acknowledged.