Cell death disguised: The mitochondrial permeability transition pore as the c-subunit of the F₁F_O ATP synthase

Abstract

Ion transport across the mitochondrial inner and outer membranes is central to mitochondrial function, including regulation of oxidative phosphorylation and cell death. Although essential for ATP production by mitochondria, recent findings have confirmed that the c-subunit of the ATP synthase also houses a large conductance uncoupling channel, the mitochondrial permeability transition pore (mPTP), the persistent opening of which produces osmotic dysregulation of the inner mitochondrial membrane and cell death. This review will discuss recent advances in understanding the molecular components of mPTP, its regulatory mechanisms and how these contribute directly to its physiological as well as pathological roles.

Graphical Abstract

Mitochondria at the center of cell metabolism and cell death

Mitochondria are complex organelles responsible for producing energy in the form of ATP for most eukaryotic cells. They regulate several other essential processes including calcium (Ca²⁺) homeostasis, heme and steroid biosynthesis. In addition, the mitochondrion lies at the center of the cellular response to stress and the control of cell death.

To produce energy in the form of ATP, mitochondria utilize substrates produced in the cytosol by carbohydrate, lipid and protein metabolic pathways. These products, particularly acetyl co-enzyme A, enter the tricarboxylic acid cycle. Turns of the TCA cycle synthesize NADH and FADH₂ that donate their electrons to the electron transport chain. The energy of the bonds of NADH and FADH₂ is used to pump H⁺ ions out of the matrix by the NADH dehydrogenase and other electron transport complexes, creating a proton motive force that in turn drives the F₁F_O ATP synthase [1]. Upon kinetic repositioning of the ATP synthase rotor, ATP is synthesized from ADP and P_i [2]. The machinery required for ADP/ATP exchange between the cytoplasm and the matrix including the outer membrane voltage dependent anion channel (VDAC) and the adenine nucleotide transporter (ANT) at the inner membrane are intimately linked to that of the ATP synthase [3].

Mitochondrial inner membrane leak-regulator of metabolic rate and uncoupling

There are two currents that complete the current loop of the proton pumping activity of the electron transport complexes. First, ATP is formed by hydrogen ion (H⁺) translocation through the ATP synthase in the opposite direction to that of the electron transport complexes. Second, an apparently wasteful leak in the inner mitochondrial membrane provides a pathway for uncoupling of oxidation from phosphorylation as H⁺ ions enter the matrix through channels independent of ATP production. Classically, uncoupling proteins carry out this role. Known physiological functions of uncoupling proteins are to generate heat for organisms with large surface to volume ratio, to depolarize mitochondria in order to temper oxidative damage and to regulate metabolic rate during hibernation and at other times [4–7]. In addition to uncoupling proteins, however, intrinsic uncoupling exists within other inner mitochondrial membrane channels and transporters and within the F₁F_O ATP synthase [8,9].

Mitochondrial inner membrane Ca²⁺ cycling regulates cellular Ca²⁺ dynamics: the example of neuronal short term synaptic plasticity

Mitochondrial inner membrane depolarization occurs not only through proton movement but also via the flux of other ions including Ca²⁺ across mitochondrial membranes. Ca²⁺ movement into the mitochondrial matrix is a physiological event that takes place in response to increased cytosolic Ca²⁺ levels. Ca²⁺ buffering is frequently employed by mitochondria in cells that experience rapidly changing cytosolic Ca²⁺ levels.

Mitochondria regulate cytosolic levels of Ca²⁺ and the release of Ca²⁺ and metabolites through an intricate system involving several ion channels. The discovery of the molecular structure for the Ca²⁺ uniporter ion channel (MCU) at the mitochondrial inner membrane has generated increasing interest in mechanisms of Ca²⁺ management within the cell body of many types of cells and also in the presynaptic terminals of neurons [10–13]. Additional isoforms of MCU and its helper MICU that confer tissue specificity and other behaviors have added to our understanding of the mechanisms of activity dependent energy production by mitochondria [14,15]. Mitochondrial Ca²⁺ release also appears to be highly regulated, involving both exchangers and channels, but, unlike the MCU, the molecular components of a Ca²⁺ release channel were only recently discovered and form the main focus of this review.

Ca²⁺ re-release from mitochondria determines short term synaptic plasticity in certain neuronal synapses [16]. During neuronal activity, Ca²⁺ influx across the plasma membrane occurs through glutamate receptors and voltage gated Ca²⁺ channels. After Ca²⁺ enters the cytosol, Ca²⁺ clearance is performed by the actions of Ca²⁺ ATPases at the plasma membrane and by buffering through uptake by intracellular stores including the endoplasmic reticulum (ER) and mitochondria [12,17]; these processes reset the normally low Ca²⁺ levels present in resting neurons or neuronal synapses. The Ca²⁺ that is buffered by intracellular stores is eventually re-released, providing, for example, for residual Ca²⁺ in presynaptic endings. Residual Ca²⁺ increases the amount of Ca²⁺ available for synaptic vesicle fusion, enhancing the amount of neurotransmitter released for a given stimulus [18,19].

Ca²⁺ -sensitive ligand gated mitochondrial channels, which are widely conserved and found in species from invertebrates to mammals, open in response to elevated Ca²⁺ within the mitochondrial matrix. In the squid presynaptic terminal, opening of a Ca²⁺ -activated mitochondrial channel is correlated with enhanced neurotransmitter release [20]. Electrophysiological recordings [21] demonstrate that within the resting presynaptic terminal, the conductance of mitochondrial membranes is low [20]. In contrast, during high frequency electrical stimulation of the presynaptic nerve, a large increase in mitochondrial membrane ion channel activity takes place [20]. The delay in onset of the mitochondrial activity and the persistence of the mitochondrial activity after stimulation are in keeping with the role of a channel and/or exchanger in re-releasing Ca²⁺ from mitochondria for short term plasticity [20,22,23]. Furthermore, mitochondrial activity and short term increases in post stimulation synaptic transmitter release are both abrogated by applying the uncoupler FCCP (carbonyl cyanide p-trifluoromethoxyphenylhydrazone), which depolarizes mitochondria, preventing Ca²⁺ handling [20].

Bcl-2 family proteins: regulators of mitochondrial outer membrane permeability and cell death

In contrast to the normal physiological role for Ca²⁺ release channels and exchangers in mitochondrial membranes, one of the main regulators of pathological mitochondrial permeability or leakiness are the proteins of the Bcl-2 family. Programmed cell death (apoptosis) in vertebrate cells may be initiated by signaling at the plasma membrane or by intracellular pathways that lead to changes in mitochondria [24]. The final common pathway for programmed cell death in many systems is mitochondrial outer membrane permeabilization (MOMP) [25–28]. Pro-apoptotic Bcl-2 family members such as Bax regulate MOMP by inducing the formation of large outer membrane pores comprised of activated oligomerized proteins, aided by other pro-apoptotic moieties [25,27,29]. In their canonical role, the anti-apoptotic Bcl-2 family proteins such as Bcl-x_L protect cells against MOMP by interacting with, and inhibiting the pore forming properties of, the pro-apoptotic family members [28,30].

MOMP leads to the release of several inter-membrane space proteins such as cytochrome c [31,32]. The resultant decrease in cytochrome c levels compromises the ability of mitochondria to maintain the mitochondrial inner membrane potential and to produce ATP [33]. In addition, cytochrome c released into the cytoplasm activates downstream cytosolic enzyme pathways including effector caspases that execute cell death [34].

The mitochondrial permeability transition and pathophysiology

In some cases, an increase in mitochondrial outer membrane permeability may also be triggered by an acute inner membrane depolarization [35], particularly after cytosolic and mitochondrial Ca²⁺ overload. Although Ca²⁺ uptake and re-release from mitochondria is a normal physiological event in cells, accumulation of Ca²⁺ in the matrix can have detrimental effects, including diminution in energy production by the ATP synthase [36]. Ca²⁺ overload can produce an uncoupling process described historically as a rapid increase in permeability of the mitochondrial inner membrane to solutes and the halting of ATP production [37–39]. This phenomenon is termed permeability transition (PT).

PT can be reversible or irreversible [35,37,40–45]. If not reversed, PT leads to an even more extreme form of catastrophic PT associated with structural breakdown of the mitochondrial matrix accompanied by outer mitochondrial membrane rupture and cell death [35]. The interaction of this kind of mitochondrial cell death with apoptotic death produced by MOMP has been debated. Although the two types of cell death seem to be overlapping, it is safe to say that pathological PT is associated with necrotic cell death such as is found in ischemia or injury whereas MOMP occurring in the presence of sufficient amounts of ATP may have a more important role in developmental and genetically predetermined death [46,47]. Inter-membrane space pro-apoptotic factors such as cytochrome c and Smac/DIABLO are released during both forms of cell death. In MOMP, outer membrane permeabilization alone leads to release of these factors, whereas in prolonged PT, rupture of the outer membrane after inner membrane swelling releases pro-apoptotic factors into the cytosol [35].

Physiological functions of the permeability transition pore

PT has been extensively studied for its role in ischemic injury in brain, heart and other organs as well as in neurodegenerative conditions [48]. In the heart, data suggest that opening of the mPTP during early reperfusion after ischemia is a harmful event that precipitates further damage to the myocardium [49]. However additional data also suggest that transient mPTP opening during preconditioning can be protective, thus serving a physiological role even during injury [43].

Further evidence of physiological opening of the mPTP has emerged over the last few years [40–45]. Flickering of the mPTP was described in the late 1990s [50,51], and an association between transient mPTP opening and “superoxide flashes” has been described in striated muscle mitochondria [44]. It has been hypothesized that transient opening of the mPTP releases mitochondrial matrix Ca²⁺ to maintain mitochondrial homeostasis [52] although this function of the mPTP has recently be questioned [53], and as mentioned above, mitochondrial matrix Ca²⁺ re-release regulates synaptic transmission [20]. Finally, as discussed in more detail below, we have observed physiological, long-term opening of the mPTP in the early embryonic heart [54].

Regulation of the permeability transition pore

The mPTP is induced by elevated mitochondrial matrix Ca²⁺, ROS, inorganic phosphate, and intracellular acidification [55,56]. In contrast, it is inhibited by ATP/ADP and Mg²⁺ [40,57]. Recent reports have also confirmed increased activity of PT by polyphosphates, chains of 10s to 100s of repeating phosphates linked by ATP-like high energy bonds [58–61]. The actions of Ca²⁺ may also require polyhydroxybutyrate (PHB), which enters mitochondria and enhances the ability of Ca²⁺ to induce PT [62].

Many signaling cascades have been proposed to cause or protect against ischemia reperfusion injury by targeting the mPTP. The RISK (Reperfusion Injury Salvage Kinase) pathway, involving PI3K, Akt, and Erk 1/2; SAFE (Survivor Activating Factor Enhancement) pathway, involving TNFα and STAT-3; and PKA, PKC, and PKG may all converge on inactivation/phosphorylation of GSK3β, which is known to regulate the mPTP [63–65]. However, the importance of these signaling cascades, in comparison to simple oxidative stress, has been questioned [64,66].

Cellular metabolic pathways regulate the PT. First, electron transport chain activity regulates the mPTP; an increase in mitochondrial energization (membrane potential) inhibits the PT, while de-energization/depolarization (a fall in membrane potential) enhances it [37,38,64]. Components of the glycolytic pathway may also control opening of the mPTP, as the binding of hexokinase II to the OMM in cardiac myocytes inhibits MOMP and PT [67]. Furthermore, a complex of ANT, VDAC, hexokinase, and mitochondrial creatine kinase (mtCK) regulates PT [68,69]. Combined, these data suggest that mitochondrial function and metabolic channeling of substrates and products (high energy phosphate bonds in the form of creatine kinase) into and out of mitochondria play a large role in the regulation of the mPTP. Finally, the major pharmacologic agents used to manipulate the mPTP target these metabolic pathways. For example, inhibitors of ANT can either attenuate (bongkrekic acid) or enhance (atractyloside) mPTP opening [38,70,71]. In addition, the pharmacological agent most efficient in inhibiting PT is cyclosporine A (CsA), an immunosuppressant drug which binds to cyclophilin D (CypD) and inhibits the channel activity associated with PT. CyPD binds to ANT, F₁F_O ATP synthase, and the phosphate carrier (discussed below) further suggesting a role for metabolic pathways in the regulation of the PTP.

The quest for the PT pore

Electrophysiologic properties of the mPTP

Thus, PT is an important event that performs both physiologic and pathophysiologic functions. Most current studies make the assumption that PT begins as the opening of a Ca²⁺ sensitive ion channel in the inner mitochondrial membrane similar to the ion channel activity initiated by mitochondrial Ca²⁺ influx occurring during physiological mitochondrial Ca²⁺ cycling. Such a Ca²⁺ release channel is heavily regulated; therefore it is assumed that only after prolonged opening does pathological PT (with MOMP) occur [72]. The conversion of a physiological Ca²⁺ extrusion mechanism into a pathological channel opening is perhaps correlated with energy failure as a result of arrest of ATP synthesizing activity and slowing of energy dependent Ca²⁺ extrusion mechanisms. Despite these hypothetical models, the factors that regulate the transition from physiological to pathophysiological events are not completely understood. Therefore, identification of the molecular structure of the pore will provide the target for regulatory activities, allowing for a greater understanding of PT modulation during health and disease.

Description of the biophysical properties of the pore that opens in the inner membrane during PT (the mPTP) provided the earliest indication that PT was initiated by the opening of an ion channel. The first patch clamp recordings of mitochondrial inner membrane were published in 1987. This early report highlighted a ~100 pS channel recorded by patch-clamping giant mouse liver mitochondria produced by cuprizone application [73]. In the late 1980s, a putative mPTP was recorded by patch-clamping mitochondrial inner membrane or mitoplast preparations [74]. The activity occurred at positive potentials of the patch pipette and was found either in whole organelle mode or in single channel recordings in the organelle-attached configuration. The activity was slightly anion over cation selective with multiple sub-conductance states ranging from 30 pS to a peak single channel conductance of 1.3 nS. Lower conductances were attributed to substates of the larger channel openings because of long periods lacking activity followed by periods of multi-conductance behavior [74]. Conductances of 550 pS were frequently observed at positive potentials. Gating was less common at negative potentials but this observation was consistent with the presence of prolonged openings and fewer sub-conductance steps at negative patch potentials contrasted with increased flickering at positive potentials. The authors concluded that conductance levels were not sharply defined, consistent with the existence of many varied conductance levels of the channel.

Also in 1989, Kinnally et al. recorded a similar mitochondrial multiconductance channel (MMC) in mouse liver mitoplasts [75]. This channel changed over time, with low activity at the onset of the recording followed by progressively higher activity at later times during the recording. The channels were sometimes open more frequently at negative potentials but at times rectification occurred in the opposite direction (more frequently open at positive potentials). Channel activity displayed multiple conductances ranging from 10–1000 pS and was weakly cation-selective. These early studies began to establish expected criteria for activity of mPTP.

Shortly after the first recordings of the putative mPTP were performed, similar inner membrane activity was found to be inhibited by CsA. In patch clamp experiments performed in liver mitochondria, channel activity was rapidly inhibited by submicromolar concentrations of CsA in a manner consistent with the expression of the binding site on the matrix side of the inner membrane. Ca²⁺ -activated large conductance channel activity up to 1.3 nS was inhibitable, but a 107 pS inner membrane conductance similar to the first recorded inner mitochondrial membrane channel was also observed in the recordings. This smaller conductance was resistant to CsA, suggesting that this activity might be due to a separate ion channel [70]. The large conductance channel was sensitive to Mg²⁺, Mn²⁺, Ba²⁺ and Sr²⁺ in that order, which inhibited the activity in a competitive manner with Ca²⁺, the main activator of the channel [55].

Characterization of a molecular complex regulating the pore

The recent identification of the molecular structure matching the biophysical properties of mPTP was aided by seemingly unrelated sets of findings. One was that Bcl-xL enhances metabolic efficiency (decreases uncoupling) by binding to the β-subunit of the ATP synthase. The second finding was that CypD, which had been known for many years to regulate PT, binds to the stator arm of ATP synthase, specifically on the OSCP subunit. The third finding suggested that closure of the mPTP is related to the level of CypD activity in a developmentally regulated manner as activity falls at the onset of respiration in mammalian heart. The final project found that ATP synthase assembles into a very large complex with other proteins that may regulate the mPTP. These findings will now be discussed in greater detail.

Bcl-x_L regulates metabolic efficiency by binding to the β-subunit of the ATP synthase

Inefficiency of metabolism is correlated with cell death under conditions of neurodegeneration or acute cellular injury such as occurs during PT [76–78]. In contrast, a highly efficient state of metabolism requires maximally decreased uncoupling of the inner membrane, and it has been found recently that Bcl-2 family proteins regulate this efficiency by binding directly to the ATP synthase [79–81]. It was known previously that Bcl-2 family proteins regulate mitochondrial outer membrane permeability to produce or inhibit synaptic degeneration and cell death [30,82–84]. Therefore, it seemed possible that Bcl-2 family proteins could form part of a large protein complex that regulates mPTP and cell death.

During the initiation of cell death an important function of Bcl-x_L, in addition to protection from MOMP, is to increase the release of ATP through enhanced VDAC opening. This decreases the probability of MOMP in cancer cell lines by providing extra ATP to overcome cell death stimuli [33,85]. In the neuronal synapse, injection of either Bcl-x_L or ATP enhances synaptic transmitter release [86], suggesting that Bcl-x_L increases ATP levels in the synapse [87]. These findings raise the possibility that Bcl-x_L might regulate not only the release but also the production of ATP.

In support of a role for Bcl-x_L in the manufacture of ATP, hippocampal neurons overexpressing Bcl-x_L show a large increase in cytoplasmic ATP levels. Surprisingly, this increase in ATP accompanies a decrease in neuronal oxygen uptake and aerobic glycolysis, consistent with the notion that Bcl-x_L overexpression increases mitochondrial bioenergetic efficiency [80,81]. Bcl-x_L depletion reverses these effects on metabolism, decreasing ATP production and increasing oxygen uptake by resting cells [80,88].

Direct interaction of Bcl-x_L with the β-subunit of the ATP synthase maximizes the efficiency of ATP production [80,81] by decreasing a leak conductance in the mitochondrial inner membrane. This was first measured by patch clamp recordings of submitochondrial vesicles (SMVs) enriched in ATP synthase [80].

Closure of the leak within the inner membrane in the presence of Bcl-x_L aids actively firing neurons to increase neurotransmitter release [88,89], consistent with a correlation between the increase in metabolic efficiency and the long-term higher efficacy of synaptic transmission found in Bcl-x_L expressing neurons. In contrast, opening of the Bcl-x_L-regulated leak decreases metabolic efficiency and predisposes neurons to death, providing yet another clue that the Bcl-x_L-regulated inner membrane leak could be mPTP. Furthermore, neurons lacking Bcl-x_L display a fluctuating mitochondrial inner membrane potential and a marked depolarization in the presence of the ATP synthase inhibitor oligomycin [81]. These data suggest that Bcl-x_L may regulate inner membrane coupling and cell death via direct effects on F₁F_O ATP synthase.

Cyclophilin D binds to ATP synthase and regulates permeability transition

Another piece of the puzzle that helped determine the molecular components of mPTP was the discovery of the interaction between CypD and ATP synthase [90]. It has long been known that CypD, a chaperone protein and peptidyl-prolyl cis-trans isomerase that resides in the mitochondrial matrix, regulates the mPTP by enhancing its sensitivity to Ca²⁺. In keeping with this, once it was found that CsA inhibits PT, CsA rapidly became one of the major tools to study the mPTP. Later, it was found that CsA attenuates mPTP by inhibiting CypD [91–93].

In 2005, experiments using CypD null mice demonstrated that CypD was not itself the pore of the mPTP but that it played an important regulatory role, particularly in the modulation of the mPTP by Ca²⁺ [94–97]. Four groups showed that deletion of CypD decreased sensitivity to ischemia-reperfusion injury in the heart and brain [94–96] and the Molkentin group suggested that mPTP’s physiologic function is to maintain “homeostatic mitochondrial Ca²⁺ levels to match metabolism with alterations in myocardial workload” [52].

CypD expression varies widely among cell types. CypD is more highly expressed in aged hearts, and these changes in expression may regulate its association with a complex of proteins that increases mPTP opening during reperfusion [63]. Furthermore, CypD activity appears to be regulated by cell signaling and metabolic pathways [64,98–101] and by developmental cues in differentiating myocytes (see below).

Although the exact mechanism of CypD’s action in the mitochondrial matrix has not been clarified, recent data support the idea that this requires its binding to F₁F_O ATP synthase. First, the propensity toward PT is regulated by ATP hydrolysis and synthesis in as much as this regulates the membrane potential. Therefore, PT requires twice the Ca²⁺ load in mitochondria that are in the process of hydrolyzing ATP (making a membrane potential) versus synthesizing ATP (dissipating the membrane potential) [38,39,56]. Second, CypD binds to proteins on the stator of ATP synthase (OSCP, b and d-subunits, [71]) as well as to the F₁F_O ATP synthase binding partners ANT [102,103] and the phosphate carrier (PiC, [52,104,105]). Recent work on the binding of CypD to ATP synthase indicates that this occurs exclusively with OSCP, and depletion of OSCP using siRNA also depletes CypD [56] suggesting that the association regulates the expression of each protein. Interestingly, CsA decreases the propensity toward PT to the same degree (half) that decreases in OSCP expression increase the propensity toward PT (double), suggesting an inhibitory effect of OSCP on PT [56]. Third, purified ATP synthase dimers produce current consistent with mPTP in artificial lipid membranes [56]. The single channel activity has a maximal conductance of 1–1.3 nS with subconductance states. The current is not stimulated by Ca²⁺ or inhibited by CsA presumably because CypD is not present in the purified dimer preparation although the enzymatic activity of the F₁F_O ATP synthase is preserved. The ATPase dimer current is also not sensitive to bongkrekic acid or atractyloside, agents that primarily affect ANT activity, making it unlikely that ANT forms part of the pore of the channel [56]. These data emphasize the idea that the regulation of the mPTP may occur via the interaction of CypD and other molecules with F₁F_O ATP synthase and its binding partners.

PT activity regulates cardiac development

A third line of evidence that helped unravel the identity of the mPTP was a series of studies of mitochondrial function during cardiac development. Until recently, the general consensus was that opening of the mPTP was a devastating event that triggers cell death. However, as discussed above, over the last two decades, data have emerged suggesting that transient opening of the mPTP could serve a physiologic purpose.

In the heart, physiological variations in mPTP activity play a critical role in cardiac myocyte differentiation and cardiac development [106]. Interestingly, the mPTP is open in myocytes in the early embryonic mouse heart, and this opening is not associated with any form of cell death. However, by the mid-embryonic stage, the mPTP is closed [106]. This closure coincides with activation of complex I of the electron transport chain, assembly of electron transport chain supercomplexes called respirasomes, and activation of oxidative phosphorylation [54]. These changes cause a fall in mitochondrial-derived ROS that signals the myocyte to undergo further differentiation [106]. Furthermore, pharmacologically inhibiting or genetically deleting mPTP enhances myocyte differentiation, while opening mPTP inhibits differentiation [106]. These findings have been confirmed in cardiac stem cells [107,108], and various reports have stressed the importance of the mPTP during cardiac development and myocyte differentiation [109,110].

Regulatory molecules do not form the pore of mPTP

The F₁F_O ATP synthase interacts with a large number of proteins many of which have been candidates for mPTP. ANT was an early candidate to form the mPTP since atractyloside and bongkrekic acid, which inhibit ANT, affect the mPTP [38] and ANT was found to interact with CypD [92]. VDAC was also an early candidate to form the mPTP due to its high conductance and its association with ANT in immunoprecipitation experiments [102]. In addition, it was shown that a complex of ANT, VDAC, hexokinase, and mitochondrial creatine kinase (mtCK) could form high conductance pores when reconstituted into membranes [68,69]. Finally, the PiC is a more recent candidate to form the mPTP [105].

However, genetic deletion of ANT1 and 2 and of the PiC demonstrated that these proteins were not essential to mPTP formation, although these studies still supported their regulatory roles [104,111,112]. Furthermore, deletion of VDAC did not affect pore formation [113]. Additional data suggest that the conformation of ANT may be important for regulation of the mPTP [114]. Atractyloside induces mPTP opening and is known to stabilize the “c” conformation of ANT, such that the adenine nucleotide transport site faces the cytoplasmic, or intermembrane space, side of the IMM [114]. In contrast, bongkrekic acid prevents mPTP opening and stabilizes ANT in its “m,” or matrix facing conformation [114]. However, as both ATR and BKA inhibit ANT, it is unlikely that specific effects on ADP/ATP translocation regulate the mPTP and a more likely scenario is that the conformation of ANT itself can regulate the mPTP. This is supported by a report that Ca²⁺ stabilization of ANT’s “c” conformation is related to its proline isomerization and CypD is a peptidyl-prolyl, cis-trans isomerase [115].

It remains unclear how these candidate molecules regulate the mPTP, but evidence suggests that they form large macromolecular structures with F₁F_O ATP synthase in the inner mitochondrial membrane. ANT and PiC can form a complex with F₁F_O ATP synthase called the synthasome [3]. In addition, the large complex of ANT, VDAC, hexokinase and mtCK is likely also involved in the regulation of ATP synthesis [68,69]. Therefore, each of these molecules may regulate the structure and activity of F₁F_O ATP synthase, and, in doing so, modulate the opening of the mPTP.

The mPTP, a molecular definition

The c-subunit of F₁F_O ATP synthase comprises the PT pore

These various reports all agree that F₁F_O ATP synthase is a major factor in the formation of the mPTP, and recent evidence suggests that the F_O or membrane portion of F₁F_O ATP synthase in fact forms the pore [48,116–120]. Mammalian F₁F_O ATP synthase is a ~600 kDa complex of 15 subunits. The membrane portion, or F_O, contains a ring of 8 very hydrophobic c-subunits and subunits a, b, e, f, g, and A6L. A stalk composed of the δ, ε, and γ subunits connects the c-subunit ring to the catalytic F₁ component made of a hexamer of alternating α and β subunits, where ATP synthesis and hydrolysis occur. Finally, a stator containing the b, d, F6, and OSCP subunits connects the lateral portion of F_O the top of the F₁. Movement of protons between the c-subunit and the a-subunit causes rotation of the c-subunit ring, the energy of which is transferred to F₁ to synthesize ATP [121–125].

Age-dependent structural and functional alterations of F₁F_O ATP synthase in rat brain and heart mitochondria have been reported [126]. The Fo portion is found to be present at 3 months in the membranes of heart mitochondria in the absence of an equivalent complement of F₁. During the period from 3 to 12 months, the level of F₁ increases with an increase in ATPase activity and this accompanies a decrease in proton leak consequent to binding of F₁ to F_O. Decreases of F₁ content with respect to that observed for F_O are detected for aging heart in animals from 12 to 24 months suggesting the presence of lone F_O leak channels in the membranes of these aging mitochondria [126].

To determine if the membrane portion of the F₁F_O ATP synthase comprises a cell death regulatory moiety that could produce PT under cell stress, the Pinton group performed different measures of cell death and PT after c-subunit expression levels were genetically manipulated either by depletion of all three isoforms of the c-subunit or by overexpression of a tagged version of the protein in HeLa cells [116]. Given that proliferating cell lines rely on glycolysis for ATP production in normal glucose-containing medium, depleting the c-subunit did not alter ATP levels in the cells. Depletion of the c-subunit, however, prevented both CsA-sensitive PT measured by the calcein-cobalt quench technique and by mitochondrial morphological analysis. Cell death brought on by H₂O₂ was also attenuated by c-subunit depletion as was cell death in neurons brought on by glutamate-induced excitotoxicity.

Although the above study indicated that the F₁F_O ATP synthase c-subunit was an important component of the mPTP [116], this work did not directly determine what portion of ATP synthase could form the pore of the mPTP. Subsequent to that publication, work from the Bernardi group suggested that the mPTP could only form from dimers of ATP synthase [56], demonstrating mPTP-like channel activity from purified dimers, and not monomers, of ATP synthase. However, the regulation of such a purified mPTP by traditional mechanisms has not been conclusively demonstrated, and the actual pore of the mPTP in this model remained to be determined. Furthermore, it was subsequently demonstrated by our group that purified monomers and c-subunits (discussed below) can form the pore of mPTP [117]. Finally, the ability of ATP synthase inhibitory factor 1 (IF1) to promote dimer formation yet prevent PT argues against the idea that ATP synthase dimers are necessary and sufficient for mPTP formation [127,128].

Membranous pores are usually formed by integral membrane proteins, and it is interesting that molecules that regulate PT are either not membrane proteins, or, if they are integral membrane proteins (see above), they have been shown not to form the pore. For example, CypD and Bcl-x_L regulate ATP synthesis by interacting with the stator and β subunit of ATPase, respectively, and these proteins are not embedded in the inner mitochondrial membrane. However, ATP synthase contains integral membrane proteins in its F_O segment. 9 polypeptides form the F_O and the stator, but only three, a, b and c, are required for proton translocation and are evolutionarily highly conserved, like PT. In addition, ρ0 cells that lack mitochondrial DNA do not contain an a-subunit but do undergo PT. Attention focused on the c-subunit, and not the b-subunit, for a number of reasons. First, the structure of purified bacterial c-subunit rings suggests that the center of the mammalian c-subunit ring could form an ion conducting channel that would allow for uncoupling if the stalk partially or completely dissociated from it [129]. Furthermore, the c-subunit had been shown previously to express ion channel activity [130]. As a matter of fact, the mammalian c-subunit undergoes conformational changes from an α-helix to a β-sheet when in contact with water, encompassing spaces that form the walls of ion channels [131]. These ion channels have a diameter of 2.3 nm which allows molecules up to 1.5 kDa to pass, similar to PT.

Homologues of ATP synthase c-subunit are also present in the V₀ subcomplex of various vacuolar H⁺-ATPases, which are located both on the plasma membrane and on intracellular membranes [132,133]. It has been shown that transmembrane domain 1 of V₀V₁-ATPase c-subunits faces the center of the V₀ rotor structure and lines a water-accessible pore structure [134,135] V₀ c-subunits have been shown to be involved in formation of mega-channels in gap junctions between cells [136]. Therefore pore-forming ability is an important feature of all homologous c-subunits which share similar amino acid sequence [118].

The c-subunit of ATP synthase creates the high conductance mPTP pore

More recent experiments have directly tested the hypothesis that the main membrane embedded portion of mammalian F₁F_O ATP synthase, i.e. the c-subunit ring, forms the pore of the mPTP [118,137]. Indeed, electrophysiologic recordings of the purified mitochondrial c-subunit yielded a multi-conductance, voltage dependent channel with prominent subconductance states [117]. Patches contained a ~100-pS conductance, which appeared to be a subconductance state of the multiconductance activity rather than a separate conductance. Peak single channel conductances of ~1.5–2 nS were similar to activity described previously for the mitochondrial multiple conductance channel (MCC) [75]. Also consistent with MCC, channel activity often but not always demonstrated negative rectification. At very positive patch pipette potentials of over 100 mV, single channel conductances of ~1.5 nS and ~2 nS were also consistently observed. There was a higher likelihood of observing single channel events or gating at positive potentials, most likely because of the negative rectification, in keeping with similar activity observed previously for the mPTP [74].

Regarding the voltage dependence of the c-subunit recordings, it should be pointed out that voltage dependence is an inherent property of the channel [138] and is not dependent on the mitochondrial inner membrane potential or on the solutions used to record the currents. It can be measured by varying the voltage across the membrane (the command or holding voltage). Early published current-voltage (I–V) curves of mPTP [74,75] reveal that the channel shows both non-rectifying conductances and a rectification in the I–V that can be either in the positive or negative voltage range. These data are consistent with I–V plots of recent recordings of the purified c-subunit [117] (Fig. 1).

Fig. 1.

The current voltage relationship (I–V) of proteoliposome recordings of the purified c-subunit mimics that of mitoplast recordings of mPTP. A, B, C. Graphs reproduced from Kinnally et al., 1989 (with kind permission of Springer Science+Business Media) showing examples of the different I–V relationships for the mPTP found in different mitoplast (isolated inner mitochondrial membrane) recordings. D, E. Graphs reproduced from Alavian et al., 2014 showing examples of the different I–V relationships found in different proteoliposome recordings of the purified c-subunit. Shown in E are a low and high conductance mode of the channel.

F₁ regulates biophysical characteristics of the purified c-subunit

Purified c-subunit protein reconstituted into liposomes clearly lacks extrinsic regulatory moieties that are important for mPTP opening. The effects of Ca²⁺, a critical regulatory molecule, demonstrate this property. Ca²⁺ binding sites have not been detected in mitochondrial Fo, perhaps because mammalian c-subunits lack the formyl Met at the N-terminus to which Ca²⁺ binds to E. coli or chloroplast F₁F_O ATP synthase c-subunits [139]. In contrast, Ca²⁺ can bind to specific, low affinity and moderate capacity sites on the β-subunit of mitochondrial F₁ ATP synthase [140]. Other sites in other ATP synthase-interacting molecules may also be important for regulation by Ca²⁺ and other agents [68,69]. Therefore, although the new models of the pore must account for all inducers and inhibitors, these molecules may not interact directly with the c-subunit pore itself, but may instead bind to sites in the F₁ or to other molecules such as ANT that undergo structural re-arrangements to open and close the pore. Another example of this type of regulation is CypD.

In order to determine the location of the regulators, mitochondrial recordings were carried out using purified mitochondrial and F₁F_O ATP synthase preparations. In these studies, the absence of an effect of a modulator was taken as an indication that the ligand or binding site for that modulator had been removed by the purification process. These studies demonstrated that the c-subunit leak channel is regulated by the overlying F₁ and peripheral regulatory proteins [117]. In mitochondria or inner membrane preparations lacking the outer membrane such as submitochondrial vesicles (SMVs)[3], Ca²⁺ activates the c-subunit leak channel while CsA and ATP/ADP inhibit it, suggesting that the Ca²⁺ and CsA sensitive sites are present in these preparations. In contrast, removal of the F₁ and other peripheral membrane proteins by urea treatment of the inner membrane or removal of CypD by purification of ATP synthase monomers abrogates regulation of the c-subunit channel by CsA and Ca²⁺ and greatly diminishes sensitivity to ATP/ADP. These studies suggest that the CypD/Ca²⁺ binding site is contained within or associated with the F₁ portion of the ATP synthase and that a second, low-affinity ATP binding site exists in the F_O. These results are consistent with reports identifying the binding site of CypD and benzodiazepine 423, an ATP synthase-inhibitory and mPTP-sensitizing agent, on OSCP [141,142], and suggest that the assembly of F₁F_O ATP synthase into monomers, dimers, and higher order oligomers can regulate the formation of the mPTP.

Channel activity of the purified c-subunit is inhibited by the purified β-subunit of F₁, suggesting a structural rearrangement whereby the stalk and F₁ of the ATP synthase inhibits opening of the c-subunit channel, aided by ADP/ATP/Bcl-x_L binding to the β-subunit and opposed by CypD/Ca²⁺ interaction with OSCP (Fig. 2). In order to demonstrate this further, mitochondria were treated with Ca²⁺ following which the F₁F_O ATP synthase was immunoprecipitated using an antibody directed at the F₁. These studies determined that Ca²⁺ exposure destabilizes the connection between the stalk and the c-subunit, disrupting protein/protein interaction between the c-subunit and F₁ (Fig. 3). The model suggests that the channel of the mPTP forms within the c-subunit ring itself upon reversible CypD and Ca²⁺-dependent movement of the stalk away from the c-subunit [117].

Fig. 2.

Model of physiological mPTP activity: β subunit, which binds Bcl-x_L and ATP/ADP, blocks c-subunit conductance. CsA, inhibitor of CypD, blocks c-subunit conductance. Both Bcl-x_L and CsA bind to F₁ components during pore closure. The pore diameter also decreases during pore closure. The channel can switch reversibly between these open and closed states.

Fig. 3.

Model of pathophysiological mPTP activity: Ca²⁺ induces separation of F₁ from F_O during mPTP opening. The pore diameter also expands during Ca²⁺ -induced increases in mPTP conductance. The components of F₁F_O that separate from the c-subunit during this process are not yet completely known; the illustration here is schematic.

It should be emphasized that the loss of protein/protein interaction between F₁ and F_O does not require very drastic conditions; just 60 μM Ca²⁺ in the bath is enough to initiate what may indeed be reversible PT [117]. This concentration is well within the range of physiological Ca²⁺ concentrations found within the mitochondrial matrix [143–145] or concentrations measured adjacent to the mitochondria in Ca²⁺ microdomains at the plasma membrane or ER membrane [146,147]. That the loss of protein/protein interaction between F₁ and F_O is likely to be reversible has been shown upon chelation of Ca²⁺ in mitoplasts [68], intact mitochondria [148], intact neurons [20] as well as in reconstituted dimers of F-ATP synthase [149], suggesting that the F₁ and the c-subunit can recombine to close the mPTP, reforming intact F₁F_O ATP synthase and re-initiating enzymatic function [150]. However, under certain conditions, this separation may become irreversible, forming pathophysiological PT (with MOMP).

Additional regulation of the mPTP may be due to the association of other molecules with F₁F_O ATP synthase. As discussed above, F₁F_O ATP synthase can complex with ANT and PiC. In the study of F₁F_O ATP synthase dimers, bongkrekic acid, which inhibits ANT, fails to attenuate the mPTP channel activity [56]; this is recapitulated in studies of SMVs [80]. Therefore, the regulation of the mPTP by ANT and PiC may occur through their association with the peripheral membrane components of F₁F_O ATP synthase. Furthermore, the association of F₁F_O ATP synthase and ANT with PiC, mtCK, VDAC and hexokinase may explain why mPTP regulation can occur via these molecules. Therefore, the layers of regulation that control formation of the c-subunit ring/mPTP are very complex and may involve many known and undiscovered proteins that associate directly or indirectly with F₁F_O ATP synthase, as suggested in previous reports [3,151,152].

Regulators of the mPTP may also work directly on the F₁F_O ATP synthase itself. F₁ has binding sites that accommodate the effects of Ca²⁺, Mg²⁺, adenine nucleotides and P_i; and through CypD (un)binding those of H⁺, CsA and possibly of oxidants [153]. Therefore, in summary, the new model of mPTP describes either direct or indirect interaction with all known inducers, inhibitors and modulators of pore function.

Structural location of the pore within the c-subunit ring

The exact location of the pore within the c-subunit is becoming increasingly understood. Although it has been proposed that the pore of the ion channel sits between the two lateral stalks of a dimer of F₁F_O ATP synthases and not within the c-subunit ring [56], there is currently no electrophysiological evidence for the formation of such a channel, and regulation of the mPTP by components of F₁ [56,117,154] argue against this. If the pore were located between dimeric membrane-associated subunits, then F₁ would need to be positioned over the dimeric link, outside of the F_O, directed to the side of the complex, in order to regulate channel activity. This seems unlikely to happen, and, thus, a model of the pore forming in the region between monomers of F₁F_O ATP synthase requires further hypothesis testing.

Rather, it is likely that the leak is located either within the central portion of the c-subunit ring, between the individual c-subunit monomers, or between the c-subunit and the other F_O subunits, although the latter is less likely given the presence of PT in ρ0 cells that lack both mitochondrial DNA and the a-subunit [116]. In two separate experiments, it has been demonstrated that the c-subunit ring expands when it conducts ions, making it likely that the pore is formed by the c-subunit ring. The first experiment used fluorescent tetracysteine display with the placement of cysteine pairs on all c-subunit monomers within the ring. These studies showed that Ca²⁺ influx into cells causes a decrease in fluorescence consistent with expansion of the diameter of the c-subunit ring, while CsA increases fluorescence consistent with a decrease in ring diameter [117]. Mutagenesis to increase the diameter of the c-subunit ring also demonstrated that ring expansion is a means to increase conductance. Mutations targeted to four highly conserved glycines within the first (N terminus) alpha-helical region of the c-subunit [155] decrease the tight packing of the ring and increase average single-channel conductance compared with WT c-subunit rings [117].

These findings support the hypothesis that the c-subunit is necessary and sufficient to produce the pore of mPTP. When viewed from the inter-membrane space, the denuded c-subunit oligomer appears as a ring with a central pore-like structure that is normally obscured by the F₁ stalk components gamma, delta and epsilon, suggesting that the pore may form within the center of the ring given the proper hydrophilic conformation [129]. Although it has been suggested that phosphlipids occupy the central cavity of the c-subunit ring in F₁F_O ATP synthases from different species [156–158] other evidence provides for formation of a proteolipid or proteophospholipid channel structure within the central lipid region [58,62,118,131,159]. Data suggest a working model whereby the c-subunit pore forms within the proteolipid milieu upon activation of mPTP (for example by elevated matrix Ca²⁺) whereupon the ring expands and F₁ shifts; the pore is closed by a decrease in diameter of the ring and inactivated by binding of the F₁ components to the ring (Figs. 2, 3). The details of these changes and their regulation remain a work in progress.

Conclusion

For many years investigators have sought the identity of the molecular structure underlying acute alterations in mitochondrial morphology and inner membrane conductance known collectively as mitochondrial PT. Early evidence asserted that PT was caused by opening of an inner membrane ion channel. More recent data have supported this idea and shown that the c-subunit of the F₁F_O ATP synthase forms a channel with similar biophysical characteristics to mPTP but lacking regulation by Ca²⁺ or CsA and with reduced sensitivity to adenine nucleotides. Depletion of c-subunit isoforms in cells blocks CsA-dependent PT and subsequent cell death. The traditional regulation of mPTP by Ca²⁺, Mg²⁺, Pi, adenine nucleotides, CsA, CypD and recently Bcl-x_L has now been assigned to sites on the F₁ including the stator complex and the enzymatic portion of ATP synthase. Inhibitors and activators may also work through peripheral regulatory moieties such as ANT, PiC and VDAC that exist in a large complex of proteins with the F₁F_O ATP synthase. Lipids and polyphosphates also may play an important role in pore gating or formation.

In this model, activators of the mPTP open the pore in a gating mechanism in which F₁ moves away from the mouth of the c-subunit ring while the ring expands (Fig. 3). This process is reversible, perhaps due to binding of F₁ components like the β-subunit to the ring or by the re-association of the entire F₁ onto the ring (Fig. 2). Although an amazing amount of information has come to light recently regarding the molecular structure and regulation of mPTP, there is still much to do to understand the details. In addition, the role of mPTP during development of oxidative phosphorylation, in aging and in supercomplex formation comprise rapidly changing fields.