From ATP to PTP and back. A dual function for the mitochondrial ATP synthase

Abstract

Mitochondria play a fundamental role in heart physiology, but are also key effectors of dysfunction and death. This dual role assumes a new meaning following recent advances on the nature and regulation of the permeability transition pore, an inner membrane channel whose opening requires matrix Ca²⁺ and is modulated by many effectors including reactive oxygen species, matrix cyclophilin D, Pi and matrix pH. The recent demonstration that the F-ATP synthase can reversibly undergo a Ca²⁺-dependent transition to form a channel that mediates the permeability transition opens new perspectives to the field. These findings demand a reassessment of the modifications of F-ATP synthase that take place in the heart under pathological conditions and of their potential role in determining the transition of F-ATP synthase from and energy-conserving into an energy-dissipating device.

Introduction

Besides their central role in ATP synthesis and oxidative metabolism of nutrients, mitochondria contribute to intracellular Ca²⁺ homeostasis and generation of reactive oxygen species (ROS), and are determinant for both cell survival and cell death. Important for any cell type, these functions are crucial for cardiac myocytes. Mitochondrial processes are highly compartmentalized due to the existence of 2 limiting membranes and to the quite selective localization of proteins, nucleotides and coenzymes in the intermembrane and matrix spaces.

The outer mitochondrial membrane (OMM) is the interface between mitochondria and other cellular components and represents a physical barrier preventing the release of mitochondrial proteins involved in apoptosis, such as cytochrome c. OMM proteins are involved in apoptosis¹, intracellular signaling, tethering to the ER/SR,² autophagy^3,4 and mitophagy.⁵ OMM permeability to metabolites and small peptides (< 5kDa) is regulated by the voltage-dependent anion channel (VDAC) and other transporters including the translocator protein of 18 kDa (TSPO).⁶

The inner mitochondrial membrane (IMM), whose permeability to solutes is controlled by highly specific transporters and tightly regulated channels, is the site of coupling between substrate oxidation and ATP synthesis in the process of oxidative phosphorylation. Mitochondria operate a sequence of energy conversion processes through which the exergonic flow of electrons along the respiratory complexes supports the endergonic pumping of protons from the matrix to the intermembrane space. The resulting proton motive force (Δp) drives the rotation of the F_o sector of ATP synthase leading to the synthesis of ATP in the F₁ sector. The two components of Δp, namely Δψ_m and ΔpH, are also utilized for the uptake of ADP and Pi, respectively, and for the release of ATP into the cytosol in exchange for ADP. In addition, Δp is also utilized by mitochondria for their biogenesis and for the maintenance of ion homeostasis, which is crucial for Ca²⁺.^7,8 In particular, the mitochondrial Ca²⁺ uniporter (MCU)^9,10 catalyzes Ca²⁺ uptake down the Ca²⁺ electrochemical gradient, which is largely driven by the inside-negative Δψ_m, whereas release from the matrix to the intermembrane space, which is catalyzed by the mitochondrial Na⁺/Ca²⁺ exchanger, ultimately depends on ΔpH. Indeed, the exchanger is functionally coupled with the mitochondrial Na⁺/H⁺ exchanger, so that mitochondrial Ca²⁺ release is paralleled by H⁺ reuptake.¹¹

A rise in intramitochondrial [Ca²⁺] from a resting value of about 0.1 μM to ≤ 1 μM activates key enzymes of oxidative metabolism, such as pyruvic, isocitric, and oxoglutaric dehydrogenases.¹² Therefore, the increase in ATP demand dictated by cytosolic Ca²⁺-activated processes, especially contractile activity, is matched by increased ATP synthesis.¹³ Higher values of matrix [Ca²⁺] could reflect the ability of mitochondria to buffer undesired elevation of cytosolic [Ca²⁺], but may also promote cell injury after opening of the permeability transition pore (PTP), an IMM channel, as discussed in detail in the following Sections. More controversial is the contribution of mitochondrial Ca²⁺ homeostasis to cytosolic Ca²⁺ oscillations related to myocardial contractile activity. Based upon kinetic analysis of mitochondrial Ca²⁺ uptake and release in vitro, beat-to-beat variations of matrix [Ca²⁺] should not occur because MCU exhibits high transport capacity but low affinity, so that even for values of cytosolic [Ca²⁺] measured at peak systole (≅2 μM) mitochondrial Ca²⁺ uptake should be negligible. However, due to the close proximity between mitochondria and SR, MCU might be exposed to local domains of [Ca²⁺] higher than those occurring in the bulk cytosolic space.¹⁴ As far as Ca²⁺ release is concerned, the Na⁺/Ca²⁺ exchanger may be too slow to cope with very fast cytosolic oscillations, such as those occurring in a beating mouse heart.⁷ Nevertheless, clear evidence of beat-to-beat changes of mitochondrial [Ca²⁺] in intact cardiomyocytes has been provided by different laboratories.^15–17 This issue is still the matter of debate,¹⁸ and besides methodological questions a major point concerns the high energetic cost (or low convenience) that would be determined by beat-to-beat variations in mitochondrial [Ca²⁺]. An adequate pathway for Ca²⁺ release could be provided by transient openings of the PTP,¹⁹ which have been documented both in isolated mitochondria and in situ.^20–23 The advantage of an unselective channel of high conductance would be to provide charge compensation within the channel itself, allowing mitochondrial Ca²⁺ release at zero potential, i.e. even for small Ca²⁺ gradients.⁷ Support for this idea comes from the finding that treatment of cardiomyocytes with CsA (which desensitizes the PTP) increased the uptake of ⁴⁵ Ca²⁺ in mitochondria,²⁴ and that mitochondrial [Ca²⁺] is elevated in hearts with genetic ablation of mitochondrial cyclophilin (CyP) D, the mitochondrial target of CsA.²⁵ The reader is referred to a recent review for a general discussion of this hypothesis²⁶ while the links between PTP and cardiac disease will be discussed further below.

Based upon a recent quantitative analysis, mitochondria do not appear to represent a significant dynamic buffer of cytosolic Ca²⁺ under physiological conditions while they might shape cellular [Ca²⁺] dynamics upon prolonged elevations of cytosolic [Ca²⁺].²⁷ A limited contribution of mitochondrial Ca²⁺ homeostasis to cardiac function appears to be supported by the lack of a specific cardiac phenotype in MCU knockout mice.²⁸ This experimental model seems also to rule out an involvement of mitochondrial Ca²⁺ overload in ischemia/reperfusion (I/R) injury.

Mitochondria are double-edged swords

Coupling between the proton gradient and the F-ATP synthase, which is essential for life, can become a life-threatening process under conditions of cell injury, especially when oxygen availability is curtailed. In the absence of oxygen electron flow stops, the mitochondrial membrane potential collapses and mitochondrial ATP synthesis can no longer occur. If cytosolic ATP is available it is taken up by mitochondria and hydrolyzed by F-ATP synthase, with regeneration of the proton gradient if the IMM permeability is low.²⁹ If the IMM permeability increases following opening of the PTP, ATP hydrolysis is maximal and ATP depletion precipitates rigor contracture. This also explains the increased rate of glycolysis, the massive accumulation of lactate and the drop in intracellular pH determined by anoxia or any defect of the respiratory chain. These concepts support the inhibition of ATP hydrolysis as an effective strategy for myocardial protection against I/R injury that is elicited by both self-defense mechanisms and pharmacological treatments.³⁰ Indeed, the rigor contracture was prevented by exposure to the F-ATP synthase inhibitor oligomycin, emphasizing the power of F-ATP synthase as an ATP consumer.^31,32

Decreased flow of electrons along the respiratory chain promotes the partial reduction of oxygen, increasing ROS production. Complete reduction of dioxygen to H₂O requires four electrons, a process occurring sequentially since oxygen accepts one electron at a time. Consequently, oxygen reduction inevitably implies the formation of partially reduced intermediates. The large majority of oxygen delivered to mitochondria is fully reduced to water at the level of complex IV. In this terminal step of the mitochondrial respiratory chain, oxygen is reduced sequentially, yet reaction intermediates remain bound to complex IV so that only the final product, i.e. water, is released. However, electrons flowing through the respiratory chain can be donated to oxygen at other sites where reduction is not complete resulting in ROS formation.³³ Very significant amounts of ROS are also produced within mitochondria at sites other than the IMM such as monoamine oxidase in the OMM and p66^Shc in the intermembrane space.³⁴ In brain mitochondria the high rate of H₂O₂ formation from the respiratory chain elicited by the complex III inhibitor antimycin A, is still much lower than that originating from activity of monoamine oxidase.^35,36

ROS generated within mitochondria may serve a function in signaling processes that are crucial for the optimal response to physiological and pathological stimuli.^37–39 Most of the effects elicited by ROS can be explained by post-translational modifications of proteins, especially at the level of Cys residues⁴⁰ that might impact also on protein (de)phosphorylation by modulating protein kinases⁴¹ and phosphatases.⁴² On the other hand, ROS can become detrimental, in a transition that has been dubbed mitohormesis.^43,44 There is little doubt that formation of ROS can determine mitochondrial dysfunction, and accelerate evolution of cell injury towards necrosis and/or apoptosis;^45,46 and that ROS formation is an essential element favoring onset of the permeability transition.⁴⁷

The mitochondrial permeability transition pore

The PTP of mammalian mitochondria is an IMM channel with a radius of about 1.4 nm⁴⁸ and a conductance ranging between 0.9 and 1.5 nS in symmetrical 150 mM KCl.^49,50 It frequently enters a half-conductance substate that may be due to its dimeric composition,^51–53 which will be further discussed below (see also a recent review by Szabò and Zoratti).⁵⁴ Occurrence of mitochondrial permeabilization and its inhibition by adenine nucleotides had been appreciated since the early 1950s,^55,56 but the potential relevance of this event to heart pathophysiology was first proposed by Haworth and Hunter,^57–60 who coined the term permeability transition and provided the unique insight that it could be due to reversible opening of a proteinaceous pore of the IMM. The nature of the PTP has remained a mystery for 60 years.⁶¹ Theories about pore formation by the IMM adenine nucleotide translocator and Pi carrier, the OMM VDAC and TSPO and by a variety of additional proteins did not stand the test of genetic inactivation of each of these putative pore component(s) and will not be covered here, but for discussion the reader is referred to a recent review where the importance of the pore as a target for cardioprotection is also addressed.⁴⁷

A turning point of research on the PTP was the discovery that the pore can be inhibited by the immunosuppressant drug CsA^62,63 through its binding to CyPD,⁶⁴ a matrix protein that like all CyPs has peptidyl-prolyl cis-trans isomerase activity that assists in protein folding.^65,66 Through the use of CsA the role of the PTP in cardiac pathophysiology could be established in cellular models,⁶⁷ in hearts subjected to ischemia-reperfusion injury⁶⁸ and in a pilot study in infarcted patients.⁶⁹ It should be stressed that CsA is not a blocker of the PTP, nor an inhibitor in the strict sense of the word. Indeed, at higher Ca²⁺ loads and/or higher concentrations of inducers the permeability transition can still occur in the presence of CsA,⁷⁰ and a more appropriate term to describe the effects of CsA on the pore is desensitization.⁷¹ This key point was firmly established by the characterization of mitochondria from CyPD-null mice, which like CsA-treated wild-type mitochondria underwent PTP opening when the matrix Ca²⁺ load was increased^72–75 and possessed a channel with electrophysiological features identical to those found in wild-type mice.⁷⁶

Our recent identification of the F-ATP synthase complex as the likely channel-forming component of the PTP⁵¹ was made possible by two sets of critical observations. The first was that CyPD interacts with the F-ATP synthase, as shown by blue-native gel analysis and co-immunoprecipitation.⁷⁷ Binding occurred at the lateral stalk of the complex, was favored by Pi (which promotes the permeability transition in mammalian mitochondria) and was counteracted by CsA with matching effects on the catalytic activity. Indeed, Pi-dependent CyPD binding decreased the catalytic activity of F-ATP synthase by about 30%, and the enzyme was fully reactivated by CsA-induced detachment of CyPD.⁷⁷ It is of note that the interactions with the PTP are affected by posttranslational modifications (PTM) of CyPD including phosphorylation by GSK3β,^78,79 acetylation⁸⁰ and S-nitrosylation⁸¹.

The second insight was the identification of subunit OSCP as the binding partner of CyPD,⁵¹ and the observation by Glick and coworkers that OSCP is also the binding site of the F-ATP synthase inhibitor Bz-423.⁸² Following the demonstration that Bz-423 is a PTP inducer,⁵¹ we could show that purified dimers of F-ATP synthase form channels activated by Ca²⁺, Bz-423 and oxidative stress, and inhibited by Mg²⁺/ADP in preparations from bovine hearts,⁵¹ S. cerevisiae⁵² and D. melanogaster.⁵³ Channel formation by F-ATP synthase was also observed by Alavian et al.,⁸³ and a role for the enzyme is also supported by a study where translation of the c subunit was partially inhibited by siRNAs resulting in PTP inactivation.⁸⁴

The mechanism through which the F-ATP synthase forms the PTP is the subject of intense investigation.⁴⁷ We have proposed that the channel forms from dimers of F-ATP synthase after a conformational change that would follow replacement of Mg²⁺ with Ca²⁺ at the catalytic site, which is consistent with earlier literature on the F-ATP synthase. Indeed, when Ca²⁺ replaces Mg²⁺ ATP hydrolysis is not coupled to formation of a H⁺ gradient,^85,86 suggesting that a conformational change takes place that leads to the apparent uncoupling of chemical catalysis from H⁺ transport. We suspect that the conformational change actually leads to PTP opening, and thus that the lack of measurable H⁺ translocation is due to H⁺ backflow through the channel. In mammalian mitochondria, binding of CyPD to OSCP would cause a conformational change affecting the accessibility of Me²⁺ to the catalytic binding sites, resulting in onset of the permeability transition if matrix [Ca²⁺] is sufficiently high. The conformational change would be independently favored by ROS-dependent thiol oxidation and counteracted by thiol reduction, making replacement of Mg²⁺ with Ca²⁺ possible also in mitochondria lacking CyPD. Once the conformational change has occurred, permeation would take place at the interface between dimers, consistent with the effect of genetic ablation of the e and g subunits in yeast.⁵² The F-ATP synthase would then return to its basal coupled state when the catalytic site is reoccupied by Mg²⁺, consistent with the full reversibility of the permeability transition in populations of isolated mitochondria⁸⁷ as well as in individual organelles^20,21 and in intact cells.²² If the opening is long-lasting, however, a set of events takes place (ATP depletion, loss of substrates and pyridine nucleotides, matrix swelling causing IMM remodeling and eventually OMM rupture, release of cytochrome c and other proapoptotic factors) that critically contributes to cell death.

Regulation of the PTP

Ca²⁺-dependent channel formation under condition of oxidative stress (thiol oxidation or cross-linking) and its inhibition by Mg²⁺, adenine nucleotides and acidic pH appear to be general features of F-ATP synthases, which have so far been detected in mitochondria from B. taurus, S. cerevisiae and D. melanogaster.^51–53 On the other hand, marked differences exist in channel unit conductance, sensitivity to Pi, presence of a mitochondrial CyP and inhibition by CsA. In bilayer experiments with purified dimers of the F-ATP synthase, channel conductance was 53 pS in Drosophila, 300 pS in yeast and 500 pS in mammals, suggesting structural differences in the channel-forming subunit(s) of the enzyme.^51–53 PTP opening is strongly favored by Pi in mammalian mitochondria (in spite of the Pi-dependent decrease of free matrix [Ca²⁺])⁸⁸ while it is inhibited by Pi in yeast^52,89,90 and Drosophila.^53,91 Mammalian and S. cerevisiae mitochondria possess matrix CyPs (CyPD and CPR3, respectively) yet the PTP is inhibited by CsA only in mammalian mitochondria. On the other hand, Drosophila does not possess a mitochondrial CyP, yet expression of human CyPD in Drosophila S₂R⁺ cells sensitizes the PTP to Ca²⁺ in a process that is insensitive to CsA.⁵³ These studies suggest that regulatory interactions between CyPD and the F-ATP synthase only emerged in mammals. This interaction may also contribute to explain the unique inducing effects of Pi in mammalian mitochondria, which could (in part at least) be a consequence of increased binding of CyPD to the F-ATP synthase.⁷⁷

Studies largely carried out in the 1990s defined key mechanisms of PTP modulation, and critical residues involved in its response to pathophysiological effectors. Pore opening requires the “permissive” presence of matrix Ca²⁺ which acts through a site that can bind other Me²⁺ ions like Mg²⁺, Sr²⁺ and Mn²⁺, resulting in PTP inhibition.⁹² A second, external binding site for Me²⁺ has been defined, whose occupancy results in PTP inhibition even when Ca²⁺ is bound.⁹² Pore opening is strongly promoted by oxidation of pyridine nucleotides and of matrix dithiols at discrete sites.^93,94 These PTP-regulating dithiol-disulfide interconversions can be blocked by 1-chloro-2,4-dinitrobenzene and by monofunctional thiol reagents like N-ethylmaleimide and monobromobimane.^95,96 A second class of regulatory thiols has been identified based on the effects of the impermeant oxidant copper-o-phenanthroline, which promotes PTP opening in a process that, unlike that mediated by matrix Cys residues, cannot be inhibited by monobromobimane.^94,97 This site is located on the outer surface of the IMM, since mitoplasts displayed an identical inducing effect of copper-o-phenanthroline as intact mitochondria.⁹⁸ The permeability transition is strongly affected by matrix pH. In deenergized mitochondria the pH optimum for PTP opening is 7.4, while the open probability decreases sharply below pH 7.4 through reversible protonation of His residues that can be prevented by diethylpyrocarbonate.^99,100 The relevant His residues are not located on CyPD because PTP modulation by matrix pH is indistinguishable in mitochondria from CyPD-null and wild type animals.¹⁰¹ It should be kept in mind that in energized mitochondria an acidic pH can promote rather than inhibit PTP opening due to an increased rate of Pi uptake, an effect that could worsen PTP-dependent tissue damage in ischemic and postischemic acidosis.¹⁰² Finally, the PTP is affected by the transmembrane voltage, in the sense that depolarization increases the probability of pore opening^95,103,104 in a process where Arg residues may play a critical role.^105–107

Given that F-ATP synthase forms channels that are indistinguishable from the bona fide PTP, it may be useful to briefly present the structural features of the heart F-ATP synthase before discussing potential points of regulation, and residues where the effectors described above may exert their effects.

Structure of heart F-ATP synthase

The mitochondrial F_OF₁ ATP synthase (F-ATP synthase or complex V of the oxidative phosphorylation system) is a well conserved multi-subunit complex located in the IMM^108,109 that is responsible for the majority (over 90%) of the aerobic ATP production in the heart. This large complex is a molecular motor and normally catalyzes the synthesis of ATP from ADP and Pi using the transmembrane proton-motive force generated by respiration. As mentioned in the Introduction, however, the complex is an ancestral ATPase that can act as an ATP-consuming device when the proton gradient is dissipated.¹¹⁰

Heart mitochondria of mammals contain high amounts of F-ATP synthase, which has been estimated to be 0.38–0.45 nmol/mg protein,¹¹¹ and have been widely used as enzyme source for structural and functional studies. The complex is composed by a soluble catalytic F₁ subcomplex protruding in the mitochondrial matrix and by a membrane-embedded F_O subcomplex through which the protons flow (Figure 1, where the complex is shown in its physiological dimeric form). These subcomplexes are connected by central and peripheral stalks. The whole complex has a molecular mass that varies between 540 and 585 kDa depending on the source.¹¹² Numerous atomic structures have revealed that F₁ comprises three copies of each of the nucleotide-binding subunits α and β, which alternate around the central α-helical coiled coil of the γ subunit.^113–117 Together with the F₁ subunits δ and ε, subunit γ forms the central stalk, which is connected to the F_O c-ring structure formed by 8 copies of subunit c (Figure 1) whose central region is probably occupied by phospholipids.^118,119 The remaining F_O part consists of the subunits abdefg(A6L)F6, two subunits of which - a and A6L - are encoded by mitochondrial DNA. The remaining F_O and F₁ subunits are encoded by nuclear genes, and complex assembly is a stepwise process assisted by several factors, which are still being actively investigated.¹²⁰

Figure 1. Structure of bovine F-ATP synthase.

Left monomer, the F₁ and Fo sectors are highlighted. Right monomer, the F₁ and Fo subunits are shown. In the F₁ sector the front α and β subunits have been removed to reveal the F₁-rotor (central stalk). The F₁ α and β subunits are colored in red and yellow, respectively. The F₁-rotor γ, δ and ε subunits are colored in shades of blue, the peripheral stalk subunits b, d, F₆ and OSCP in shades of green, and the c-ring in purple. The remaining Fo subunits a, e, f, g, A6L, whose structure has not been defined yet, are located in the Fo subcomplex colored in light blue. The image (lateral view) has been built starting from the yeast dimer molecular model¹³³ (PDB id. 4b2q) and superimposing the cryo-electron microscopy map of bovine F-ATP synthase¹⁴¹ (EMD id. EMD-2091). The fit of molecular models to cryo-electron microscopy map was performed using the program ADP_EM¹⁹⁸. The molecular model for bovine F-ATP synthase was obtained by superimposing the 3D structure of the bovine F₁-c-ring complex (PDB id. 2xnd) onto each corresponding monomer of the yeast dimer. The superposition was performed using the swiss pdb viewer routine Iterative magic fit¹⁹⁹. The lateral stalk was taken from the yeast dimer (PDB id. 4b2q) which contains the bovine subunits.

The structure and arrangement of the membrane subunits has not been defined yet except for subunit a, which associates with the c-ring peripherally, possibly forming an angle of approximately 70° relative to the c-ring helices, as recently resolved in the green alga Polytomella.¹²¹ Subunit a provides two half transmembrane water channels for H⁺ to access the conserved carboxylic residues present in each c subunit. Indeed, proton translocation into the mitochondrial matrix through these (half) channels is concurrent with the clockwise rotation of the F_O c-ring (when viewed from the matrix side), which drives γ subunit rotation within the α₃β₃ F₁ subcomplex at about 100 revolutions per second. Rotation of subunit γ takes each of the 3 catalytic sites through at least 3 major functional states denoted as βE, βDP and βTP, thereby supporting the synthesis of 3 ATP molecules for each 360° rotation.^109,122,123 When the enzyme works in the direction of ATP hydrolysis, the counterclockwise rotation of the γ subunit and of the c-ring cause proton traslocation into the intermembrane space. This activity is selectively inhibited by reversible binding of the endogenous inhibitor factor 1 (IF₁) to F₁ with a 1:1 stoichiometry (Figure 2), resulting in full inhibition.^124,125 Accordingly, the ability of IF₁ to block the ATP-driven rotation of human F₁ has been recently visualized¹²⁶. The peripheral stalk, comprising the F_O subunits oligomycin sensitivity conferring protein (OSCP), b, d and F6, is important to prevent the co-rotation of α and β subunits with γ and therefore the loss of enzyme efficiency.¹²⁷ As already mentioned, this subcomplex is also the target of CyPD, which binds to the OSCP subunit (Figure 2), possibly sharing a common binding site with the F-ATP synthase inhibitor Bz-423.^128,129 CyPD binding to OSCP is favored by Pi and counteracted by CsA;⁷⁷ CyPD sensitizes the PTP to Ca²⁺, promoting the transition of the F-ATP synthase to a channel that may form at the interface between monomers^51–53,61 (Figure 2). Interestingly, OSCP has also been recently recognized as the binding site of 17β-estradiol, whose binding promotes an intrinsically uncoupled state of ATP synthase,¹³⁰ as well as of regulatory proteins, such as Sirtuin3,¹³¹ highlighting its ability to sense metabolic signals.

Figure 2. Interactors of F-ATP synthase and PTP formation.

Reversible binding of CyPD and IF₁ to F-ATP synthase is shown, together with the putative region of channel formation (broken arrow). Factors promoting binding/release of interactors are also indicated.

Electron cryomicroscopy of mitochondria from different sources¹³² clearly revealed that F-ATP synthase is organized in dimers, which are essential to maintain a high local curvature of the IMM¹³³ and normal cristae morphology,¹³⁴ and form long rows of oligomers at the cristae edges of mitochondria. An additional proposed role of F-ATP synthase dimers/oligomers is higher catalytic efficiency.^133,135 Biochemical and genetic studies provided evidence that, for dimer formation, preferential interactions occur between F_O subunits a,¹³⁶ b,¹³⁷ e,^138,139 and g,¹⁴⁰ forcing the peripheral stalks (which are turned away from each other) and the two F₁ heads at fixed angles of > 70°.^133,141 In mammals the dimers have been reported to be further stabilized by interaction with IF₁ favoring ATP synthesis in cells overexpressing IF₁,^142,143 but this proposal has been criticized based on biochemical studies in isolated ox heart membranes,¹⁴⁴ as well as on the lack of effect on dimer stability of IF₁ silencing in human osteosarcoma cells.¹⁴⁵ Moreover, dimers seem to be stabilized by binding of the matrix metalloprotein Factor B (or subunit s of F-ATP synthase),¹⁴⁶ which gets in contact with the e and g subunits as well as with the ADP/ATP carrier. The existence of a second interface (the oligomerization interface) through e/e and g/g interactions originally proposed in yeast is still debated because the distance between dimers appears variable in electron cryotomography of mitochondria from different sources, which would make direct protein contacts difficult.¹³² An interesting aspect of the dimeric/oligomeric structure is that the stabilizing contribution of the different subunits seems to be additive. Indeed, mutants lacking one or more of the above mentioned subunits, such as ρ⁰ cells (which lack mitochondrial DNA¹⁴⁷ and therefore do not synthesize subunits a and A6L)¹³⁶ and human cells totally or partially depleted of e and g subunits,¹⁴⁸ still form lower amounts of F-ATP synthase dimers and oligomers as detected by native gel electrophoresis.¹³⁶

Catalysis requires the binding of the nucleotide as a complex with Mg²⁺,¹⁴⁹ which plays a key role in shaping the high affinity catalytic sites located in the β subunits, where Mg²⁺ is hexa-coordinated by βThr₁₆₃ (bovine numbering), by the oxygen atoms βO₂ and γO₂ of ATP, and by 3 ordered water molecules (hydrogen-bonded to βArg₁₈₉, βGlu₁₉₂ and βAsp₂₅₆)¹¹⁷ (Figure 3). Mg²⁺ can be replaced by other divalent cations, including Ca²⁺, and the ionic radius is the chief determinant of the ability to support catalysis.¹⁵⁰ However, Ca²⁺ ions support γ subunit rotation and ATP hydrolysis, but not H⁺ translocation and ATP synthesis, as documented both in bacteria⁸⁵ and in mammals,⁸⁶ strongly suggesting that the catalytic site has a different conformation state when it is occupied by Ca²⁺.

Figure 3. Map of Mg²⁺/Ca²⁺ binding sites and Ca²⁺-dependent interactions in bovine F-ATP synthase.

Left monomer, the subunits involved in Ca²⁺-dependent interactions are highlighted, i.e. subunits α and β, which interact with the matrix protein S100A1¹⁵⁴ (not shown in the picture), and the Fo region containing subunit e, which may interact with a hypothetical tropomyosin-like protein localized in the intermembrane space¹⁵⁵. Ca²⁺-regulatory sites located in the c-ring are also shown^156,157. Right monomer, the residues (T163, R189, E192, D256) of β subunit interacting with the catalytic metal ions are mapped onto the 3D structure of the bovine F₁-c-ring complex.

F-ATP synthase and PTP formation in the heart

The mechanism of PTP formation from F-ATP synthase remains unsolved and is the matter of considerable debate⁴⁷. It is clear, however, that all PTP effectors will eventually have to be assigned to specific sites of the F-ATP synthase. What follows is therefore a discussion of potential targets of well-characterized pore effectors in heart F-ATP synthase, and of PTM that may affect or cause the transition to the PTP.

The PT has been detected in ρ⁰ cells lacking subunit a and A6L,¹⁵¹ while it was markedly inhibited in yeast mutants lacking subunits e and/or g,⁵² suggesting a different contribution of the F_O subunits in forming the PTP. Based on the observation that PTP has been conserved from yeast to Drosophila and mammals,¹⁵² it seems reasonable to suggest that the conserved subunits b, e and g of the Fo subcomplex, which mediate the dimer/oligomer formation, may also favor PTP formation in the presence of Ca²⁺ ions and thiol oxidants, possibly also through other subunits.

In the heart Ca²⁺ ions are also regulatory elements of F-ATP synthase activity, which must match ATP utilization for muscle contraction, whose rate and force can vary in vivo up to 5–10 fold.¹²⁴ Indeed, it is now well established that F-ATP synthase does not respond simply to the bioenergetic parameters, i.e. to the levels of ATP, ADP, Pi and proton motive force, and reversible and rapid stimulation of F-ATP synthase by physiological intramitochondrial Ca²⁺ levels has been clearly demonstrated in isolated pig heart mitochondria¹⁴ and cultured rat cardiomyocytes.¹⁵³ These changes were parallel to stimulation of the Ca²⁺-sensitive dehydrogenases of the Krebs cycle and rapid enough (100 ms) to potentially support steep changes in myocardial workload.¹⁴ Intriguingly, the ability to up-regulate the F-ATP synthase is lost in hypertension and hyperthyroidism,¹²⁴ although its relationship to the genesis of these disease states is not clear. Although a direct allosteric mechanisms mediated by Ca²⁺ ions was ruled out already in the 1990s,¹²⁴ how Ca²⁺ modulates the F-ATP synthase activity has not been established yet.

In cardiomyocytes a Ca²⁺-dependent interaction of F₁ with the S100A1 protein, which is expressed predominantly in cardiac muscle, has been reported to lead to increased ATP production.¹⁵⁴ The F_O subunit e has been hypothesized to contain a putative Ca²⁺-dependent activating region exposed at the cytosolic site of F_O sector, potentially ensuring the rapid decoding of increases of cytosolic [Ca²⁺].¹⁵⁵ Indeed, residues 34–56 of e subunit are highly homologous to the Ca²⁺-dependent tropomyosin-binding region for troponin T, and their interaction with a specific antibody mediates an increase of F-ATP synthase activity. Due to exposure of the Ca²⁺-binding site to the mitochondrial intermembrane space, subunit e may be also involved in the Ca²⁺-dependent inhibition of PTP opening⁹². Subunit e is important for F-ATP synthase dimer formation,^138,139 suggesting that Ca²⁺ inhibition could target the dimer structure and function (Figure 3).

Another conserved Ca²⁺ binding site has been described in the N-terminus of F_O subunit c,^156,157 which could also represent a candidate for PTP inhibition by Ca²⁺ in the intermembrane space. Ca²⁺ binding to subunit c purified from neuronal plasma membrane and reconstituted in artificial membranes was able to block monovalent cation currents¹⁵⁷, and Ca²⁺ binding to subunit c inhibited H⁺ translocation in bacteria and chloroplasts,¹⁵⁸ suggesting that Ca²⁺ could alter the c ring conformation. However, while the Ca²⁺-binding capacity of subunit c from bacteria and chloroplasts is well established, that of the mammalian subunit c is debated.¹⁵⁸ These data emphasize the complexity of the Ca²⁺-dependent modulation of F-ATP synthase (Figure 3), and the need for further studies to clarify how different activating signals may be integrated in the normal heart and in the transition of F-ATP synthase to PTP formation.

We have already mentioned that when the proton motive force declines F-ATP synthase reverses to a proton-pumping ATPase, and that this event plays a major role in I/R injury of the heart. It is worth mentioning that this condition may also apply to mitochondrial diseases. ^120,159 Interestingly, we observed that the sense of rotation of F-ATP synthase strongly affects the threshold of Ca²⁺ required for PTP opening in intact mitochondria.^51,53 Because continuous ATP hydrolysis markedly reduces the Ca²⁺ sensitivity of PTP, the inverse rotation of F-ATP synthase during ischemia might contribute to maintain a closed pore. A tentative explanation is that rotation during ATP hydrolysis might destabilize the dimers by pulling the stators apart and vice versa, as proposed Buzhynskyy et al., who demonstrated the existence of two classes of dimers characterized by different stalk-to-stalk distance in native IMM.¹⁶⁰

It is well established that heart mitochondria of both slow- and fast-beating species possess sufficiently high levels of IF₁ relative to the F-ATP synthase to completely inhibit ATPase activity with an effect equivalent to the action of oligomycin.¹²⁴ Consistently, both in vitro and in vivo experimental models demonstrated that during ischemia the ATPase activity is (at least partially) inhibited by the reversible binding of IF₁, thus mitigating ATP depletion. Indeed, in ischemic hearts of dog,¹⁶¹ goat¹⁶² and rabbit¹⁶³ the loss of ATPase activity correlates with an increased IF₁ content. Down-regulation of F-ATP synthase activity is also observed in anoxic cardiomyocytes, although the IF₁ contents were not measured.¹⁵³ In HeLa and C2C12 cells transient IF₁ overexpression preserves ATP and protects from cell death induced by oxygen and glucose deprivation, a finding that would enlist IF₁ in prosurvival proteins.¹⁴³ Nevertheless, the response to cell injury seems to depend on the cell type, as IF₁ content varies between species, tissues, and even between cell types within a given tissue, such as in central nervous system where neurons contain much more IF₁ than astrocytes.¹²⁵ However, it is not yet clear how IF₁ expression is modulated. Two modulators seem to be implicated, i.e. hypoxia inducible factor 1α and the immediate early response gene X-1, the latter targeting IF₁ for degradation.^164,165

IF₁-F₁ interactions are governed by the factors promoting IF₁ release and rebinding, i.e. membrane potential and MgATP. The existence of a dynamic equilibrium, in which binding and release occur simultaneously involving the entire population of F-ATP synthase, has been demonstrated in energized ox heart mitochondria.¹²⁴ In addition, IF₁ binding is strongly favored by low pH when the inhibitory dimeric IF₁ is formed¹⁶⁶ so that in ischemia, when pH and membrane potential decrease, a new dynamic equilibrium is reached. In principle, the binding of IF₁ might also contribute to decrease the open probability of PTP during ischemia (Figure 2). Indeed, the crystal structure of the IF₁-F₁ complex from ox heart has shown that the initial low-affinity binding of IF₁ to the F₁ surface is followed by the final inhibitory binding in response to ATP hydrolysis, so that two Mg²⁺-ADP molecules are contained in the F₁ catalytic sites of β_DP and β_TP.¹⁶⁷ Considering that Mg²⁺-ADP is a PTP inhibitor, it is conceivable that the final inhibited IF₁-F₁ complex may be unable to properly bind Ca²⁺, which is necessary for the switch of F-ATP synthase to the PTP. Intriguingly, the His reagent diethylpyrocarbonate, which restores the ability of mammalian mitochondria to undergo PTP opening at low matrix pH,^100,168 also prevents IF₁ binding at acidic pH.¹⁶⁹ On the other hand, the inhibited IF₁-F₁ complex does not contain Pi,¹⁶⁷ which is a PTP inhibitor at low concentrations, suggesting that IF₁ binding might mask an inhibitory site favoring, rather than preventing, PTP opening.

Although direct modulation of PTP by IF₁ remains an intriguing possibility that should be further explored, IF₁ could also affect PTP formation as a consequence of its role in ischemic preconditioning. It is well established that massive PTP opening - which occurs during reperfusion when membrane potential is restored and Ca²⁺ overload occurs - is prevented by short ischemic periods before long-lasting ischemia and reperfusion, a phenomenon called ischemic preconditioning. During ischemic,^162,170 as well pharmacological preconditioning,^171,172 IF₁ seems to mediate a peculiar, long-lasting ATPase inhibition in dog, goat and rat hearts (until 30–120 min of reperfusion depending on the species or the experimental model), although contradictory results have been published, which could be explained by the different analytic methods used¹⁷³. As preconditioning results in sparing of ATP, lowering of membrane potential and Ca²⁺ accumulation, it may appear that IF₁ favors conditions that inhibit PTP formation during reperfusion. Moreover, based on the observation that the lack of IF₁ in Luft’s disease is associated with uncoupled mitochondria,¹⁷⁴ long-lasting IF₁ binding during ischemia might maintain tight coupling supporting oxidative phosphorylation during reperfusion. Clarification of the molecular events involving IF₁ in preconditioning hold great promise for understanding preconditioning and for devising pharmacological strategies for cardioprotection.

Post-translational modifications of F-ATP synthase and PTP formation

Until recently, there were few reports regarding the PTMs of F-ATP synthase, which have been associated with specific biological function/processes in a very limited number of cases. Considerable information has been obtained about oxidative modifications of F-ATP synthase. Experiments with isolated mitochondria from bovine heart demonstrated that F-ATP synthase contains a set of thiols located in F_O (Figure 4), whose environment senses membrane energization and whose oxidation results in complete and reversible mitochondrial uncoupling.^175,176 Because uncoupling was not inhibited by oligomycin, the Authors proposed that the permeability pathway is located on the intermembrane side of the oligomycin binding site.¹⁷⁵ This position would suggest involvement of the unique Cys residue of subunit c, which is located near the Glu₅₈ residue essential for proton translocation.¹¹⁸ It is tempting to speculate that oxidation of this Cys residue may favor PTP formation in another part of F_O. This does not exclude the involvement of species-specific Cys residues located in other subunits, such as subunit b in bovine and human heart (Figure 4), whose modification in the bovine enzyme also affects catalytic function.¹⁷⁷ In animal F-ATP synthase Factor B contains two vicinal Cys residues, whose oxidation leads to mitochondrial uncoupling.¹⁷⁸ However, Factor B is not present in yeast and Drosophila melanogaster, where the PTP is also modulated by SH reagents,^52,53,91 thus suggesting that Factor B is probably not involved in PTP formation. We showed that isolated F₁ from bovine heart F-ATP synthase is selectively inactivated by hydrogen peroxide through redox-active iron-protein adducts probably generating highly reactive oxygen species in proximity of the catalytic nucleotide binding sites.¹⁷⁹ Due to their conservation it is tempting to hypothesize that also these sites may be candidates for ROS-mediated PTP modulation.

Figure 4. Map of Cys residues in bovine heart F-ATP synthase.

Position of Cys residues on the specified subunits (red dots) is mapped onto the 3D structure of the bovine F₁-c-ring complex and of the bovine lateral stalk.

Information on oxidative modifications of F-ATP synthase subunits has widely increased with the development of mass spectrometry-based methods, which demonstrated that F-ATP synthase is susceptible to oxidative/nitrative stress associated with heart failure,¹⁸⁰ CNS disorders,^181,182 caloric restriction¹⁸³ and aging.^184,185 Specifically, the α subunit is S-nitrosylated in response to GSNO treatment of a mouse heart membrane fraction.¹⁸⁶ Moreover, in the same subunit Tyr nitration was increased following I/R of mouse heart¹⁸⁷ (Figure 5), suggesting that α subunit is actively involved in various oxidative modifications. Consistently, the formation of an inter-subunit disulfide bridge between αC251 and γC78 has been observed in canine dyssynchronous heart failure, which reverted following cardiac resynchronization therapy,¹⁸⁸ and disulfide formation was proposed to be involved in PTP activation.⁹⁵ However, these residues are quite distant in the assembled complex (Figure 4) suggesting that the disulfide may only form in a misfolded/aggregated enzyme. Interestingly, in mitochondria from aging cardiomyocytes, which are more prone to undergo the PT during reperfusion in spite of an impaired ability to accumulate Ca²⁺ ions during ischemia and reperfusion, quantitative proteomics revealed increased Cys oxidation at several subunits, including OSCP and d, which might potentially affect CyPD binding to the peripheral stalk and/or PTP formation.¹⁸⁹ Moreover, a pharmaco-proteomic approach demonstrated that myocardial infarction in dogs caused several nitric oxide-related chemical modifications of d subunit, i.e., nitration of Y114 and hydroxylation of K71 (numbering according to PDB id 4b2q), followed by its myristoylation.¹⁹⁰ These PTMs were abolished by the cardioprotective drug valsartan, which also induced phosphorylation of S29 (numbering according to PDB id 4b2q) of the same subunit (Figure 5). Intriguingly, δPKC, which plays key roles in I/R, is a modulator of F-ATP synthase in cardiomyocites through direct binding to subunit d, which requires cardiolipin¹⁹¹. Subunit d also contains three Trp residues, which have been classified as “hot spots” for oxidation in human heart being found in normoxic conditions¹⁹² (Figure 5). These data suggest subunit d as actively involved in various oxidative modifications, whose implication in PTP formation remains to be elucidated.

Figure 5. Posttranslational modifications of F-ATP synthase in the normal and failing heart.

Left monomer, the residues involved in PTMs in normal heart are shown, i.e. complete trimethylation of cK43¹¹⁸; oxidation of W12, W53 of subunit d¹⁹²; phosphorylation of αS184, αS419,¹⁹⁴ γY52¹⁹⁵ and γS121,¹⁹⁶ dS30.¹⁹³ Right monomer, residues involved in PTMs associated with preconditioning or heart failure are shown, i.e. phosphorylation of βS56, βT57, βT212, βS213¹⁹⁷; nitration of d Y114 (numbering according to PDB id 4b2q) and hydroxylation of dK71¹⁹⁰. Subunit α is also highlighted, being the target of tyrosine nitration¹⁸⁷.

The most extensively characterized PTM concern Lys43 of the c subunit, which is completely trimethylated in all vertebrates and in the majority of invertebrates^118,119 (Figure 5). These residues are located in the loops exposed to the matrix and possibly provide sites for the specific binding of cardiolipin. In the normal heart several phosphorylated residues have also been reported, and these are mainly located in the F₁ sector¹⁹³ where they may be functionally relevant. The modified residues include αS184, αS419,¹⁹⁴ γY52¹⁹⁵ and γS121¹⁹⁶ (Figure 5). The β subunit was found to be modified by preconditioning of rabbit myocytes with adenosine, and contained at least 5 phosphorylated residues, 3 of which (βS56, βT57, and βT212 or βS213) are accessible and therefore potentially regulated¹⁹⁷ (Figure 5).

Conclusions

Identification of the F-ATP synthase as the most likely channel-forming element of the PTP^51–53,83 has been a turning point for our understanding of mitochondrial pathophysiology. Conservation of the channel-forming ability across species points to a conserved function, and species-specific differences in conductance and regulation suggest an evolutionary role that still needs to be addressed.^51–53 We are aware that a coherent picture explaining the mechanisms through which the energy-conserving enzyme turns into an energy-dissipating device is not possible at present, but we hope that this review will help address the many open mechanistic issues and inspire the molecular analysis of appropriate F-ATP synthase mutants.

Non-standard abbreviations and acronyms

CsA

cyclosporin A

CyP

cyclophilin

IF₁

inhibitor factor 1

IMM

inner mitochondrial membrane

I/R

ischemia/reperfusion

MCU

mitochondrial Ca²⁺ uniporter

OMM

outer mitochondrial membrane

OSCP

oligomycin sensitivity conferring protein

PTM

posttranslational modifications

PTP

permeability transition pore

ROS

reactive oxygen species

TSPO

translocator protein of 18 kDa (TSPO)

VDAC

voltage-dependent anion channel