Salvatore Nesci, Fabiana Trombetti, Vittoria Ventrella, Alessandra Pagliarani*

ABSTRACT
Based on recent advances on the Ca2+-activated F1FO-ATPase features, a novel multistep mechanism involving the mitochondrial F1FO complex in the formation and opening of the still enigmatic mito- chondrial permeability transition pore (MPTP), is proposed. MPTP opening makes the inner mitochon- drial membrane (IMM) permeable to ions and solutes and, through cascade events, addresses cell fate to death. Since MPTP forms when matrix Ca2+concentration rises and ATP is hydrolyzed by the F1FO- ATPase, conformational changes, triggered by Ca2+ insertion in F1, may be transmitted to FO and locally modify the IMM curvature. These events would cause F1FO-ATPase dimer dissociation and MPTP opening. © 2018 Elsevier B.V. and Société Française de Biochimie et Biologie Moléculaire (SFBBM). All rights reserved.

Keywords:F1FO-ATPase;Calcium ion;Mitochondrial permeability transition pore Conformational mechanism

1.Introduction
Oxidative phosphorylation,which features mitochondria in eukaryotes, is based on the cooperation and interplay between multiple enzyme complexes.Briefly, these complexes are de- hydrogenases which transfer electrons according to the electro- chemical gradient from reduced respiratory substrates,namely NADH and FADH2, to the inal acceptor molecular oxygen, and, by pumping protons in the intermembrane space, generate a H+ current through the inner mitochondrial membrane (IMM). Finally, the transmembrane electrochemical gradient of H+ (ΔμH+) created by respiratory chain substrate oxidation drives ATP synthesis by the ATP synthase [1]. The formation of a large channel in the IMM, namely the so-called mitochondrial permeability transition pore (MPTP), dissipates the ΔμH+ and, differently from the accepted bases of chemiosmotic hypothesis [2], eludes ATP production and causes loss of substrates and nucleotides from the mitochondrial matrix [3e5]. MPTP opening, by dramatically changing the IMM electrophysiological features, leads to mitochondrial dysfunction. The MPTP regulation and role in different forms of cell death, including autophagy, and in various pathologies have been the subject of intense and fruitful research, sustained by the hope to exploit this mitochondrial event to ight cancer, ischemic damage and neurodegeneration [6]. On the other hand, recent studies suggest that the MPTP may also play a relevant role in mitochon- drial function, cell differentiation and development [7].

The MPTP structure has long remained a mystery, even if its identity was intensively searched for among known membrane components, above all membrane-bound proteins. At irst, the voltage- dependent anion channel (VDAC) and the peripheral benzodiaze- pine receptor on the outer mitochondrial membrane (OMM) together with the IMM adenine nucleotide translocase (ANT) seemed the most likely candidates to take part in the enigmatic mechanism of MPTP formation [8]. In this putative mechanism, ANT was thought to constitute the MPTP fulcrum since the ANT inhibitors atractyloside (ATR) and bongkrekic acid (BGK) modu- lated the MPTP. In detail, BGK inhibited the MPTP by locking ANT in the M conformation (closed MPTP), while ATR maintained it in the C conformation (open MPTP)[9]. However, the ANT channel showed a similar conductance to that of the MPTP [10]. Subsequent indings pointed out that MPTP formation involved a supra- molecular complex, namely the assembly of different proteins [11]. Accordingly, differently localized proteins, namely hexokinase bound to the cytosolic surface of OMM, creatine kinase and nucleoside diphosphate kinase in the intermembrane space, and cyclophilin D (CypD) in the matrix apparently contributed to form the MPTP. An alternative model, in which the Pi carrier by inter- acting with ANT and CypD induced MPTP opening, was depicted [12]. However, all the models proposed over 40 years of studies did not fully match the electrophysiological MPTP features [6] or were undermined by genetic deletion tests, which, one by one, excluded that any of these proteins are essential for MPTP formation [13e16].ANT, the Pi carrier and the F1FO-ATPase may mutually interact through cardiolipin which would somehow connect these proteins to form the ATP synthasome. Consistently, conformational changes triggered by Ca2+ within the ATP synthasome may perturb the interface between these structures and produce the pore [17]. The ATP synthasome dynamics is ruled by the metabolic demand and is CypD-dependent [18]. Moreover, changes in the contact sites be- tween the inner and outer mitochondrial membranes could inter- vene in MPTP opening [17]. At present, it seems likely that the MPTP may coincide with a conserved mitochondrial protein of key role in mitochondria. Recently, the F1FO-ATPase -a splendid mo- lecular machine- [19] has been proposed to form the pore structure [20,21].

2.The F1FO-ATPase: from an old to a new story as pore former
From its discovery around the middle of the 20th century, the F1FO-ATPase has undergone a sort of on-going evolution, stimu- lated by the increasing development of techniques and of knowl- edge, which lead to a continuous re-evaluation of the roles of this intriguing enzyme complex [19,22]. At present, we can say that new and up to now unsuspected roles for this ubiquitous enzyme are emerging in mitochondria. As widely known, in eukaryotic mito- chondria the F1FO-ATPase constitutes the amazing molecular ma- chine that exploits the electrochemical energy produced by the respiratory chain in the form of Mitchell’s proton motive force (Δp) to produce ATP via a chemo-mechanical coupling mechanism [23]. Even if ATP synthesis represents the classical enzyme task, the catalytic mechanism is long known to work also in reverse to energize the IMM by ATP hydrolysis [22,24,25]. In practice, the direction of catalysis depends on Δp, with ATP synthesis consuming Δp. and, conversely, ATP hydrolysis re-building Δp. The F1FO-ATPase structure is quite complex and can be roughly deined as an olig- omer structurally composed by a hydrophilic F1 catalytic domain and by a membrane-embedded FO domain. These two domains are joined by a central and a peripheral stalk (Fig. 1). In turn the F1 sector, which protrudes in the mitochondrial matrix, shows a a3β3Yδε subunit composition and stoichiometry [26,27]. The three a subunits alternate with three β subunits to form the F1 globular hexamer. The adenine nucleotide binding sites, namely three non- catalytic sites on a subunits and three catalytic sites on the β sub- units, open at the interfaces between the a and β subunits of this spherical complex [28,29].

The membrane-embedded FO sector is also formed by multiple proteins, namely the a subunit, the short amphipathicb subunit with the two transmembrane a-helices, e, f, g, A6L, DAPIT (diabetes-associated protein in insulin-sensitive tissue) subunits, a 6.8 kDa proteolipid and the cn-ring, in which the subunit number is species-dependent [30,31]. The Y subunit ex- tends from the center of the (aβ)3structure of F1to the FO domain where it joins the δ and the ε subunits to form the foot of central stalk [32]. The core of FO is formed by the c-ring, which is directly attached to the central stalk and constitutes the enzyme rotor, which transmits the rotational energy to F1. Laterally, the b, d, F6 and OSCP subunits form the peripheral stalk, which not only links the (aβ)3-catalytic structure to the a subunit in the FO domain, forming the integral enzyme stator, but also plays the role of resisting the torque generation of rotor [33]. Actually, the F1FO- ATPase/synthase is a rotary engine which matches rotation to catalysis. The clockwise rotation (seen from the intermembrane space) is driven by Δp which makes H+ downhill translocate across the IMM through the a subunit/c-ring complex interface. This rotation transmitted from FOto F1 produces one ATP molecule per each β subunit, namely three ATP molecules are built in a 360。 cycle. The opposite rotation, which pumps H+, in the intermem- brane space and re-constitutes Δp, is coupled to ATP hydrolysis.

Fig. 1. Subunit composition and structural arrangement of the F1FO-ATPase monomer. Protein subunits are drawn as ribbon representations (modiied PDB ID codes: 5ARA and 6B2Z). Olive, a subunits; red, β subunits; blue, Y subunit; fuchsia, δ subunit; tur- quoise, ε subunit; orange, ring of c subunits; violet, a subunit; purple, A6L subunit; gold,f subunit; green, b subunit; pink, d subunit; sky-blue, F6 subunit; grey, OSCP subunit. e and g subunit drawn in ball and stick mode, are blue and light blue, respectively. DAPITand the 6.8 kDa proteolipid, still undeined membrane subunits, are not represented nucleotide binding in the catalytic site requires the coordination of the essential cofactor Mg2+, which contributes to ATP synthesis/ hydrolysis and to the asymmetry of the three catalytic sites, which produces the differences in afinity for nucleotides [34]. Accord- ingly, each β subunit is asymmetric and during the rotation, by interacting with the Y subunit, undergoes three distinct confor- mational states βE(always empty),βDP, which contains bound MgADP and βTP which binds MgATP [35].Interestingly, in mitochondria, the F1FO-ATPases are assembled in supra-molecular dimeric complexes by the transmembrane FO domain [36] which form extensive rows [37,38] distributed along the tightly curved ridges of the IMM cristae [39]. This localization exploits the higher H+ density on the surface in the curved mem- brane regions [40] created by the respiratory complexes crowding at either side of the rows [41]. Structure, localization and function are tightly connected. Accordingly, the F1FO-ATPase structural arrangement and localization have relevant implications for the mechanism of mitochondrial energy transduction [39] and sub- stantiate the F1FO-ATPase active role in membrane bending and cristae formation[40,42],thus contributing to mitochondrial morphology.

Indeed, the enzyme complex assembly locally pro- duces an extreme membrane curvature in either concave (negative curvature) where the membrane invaginates or convex (positive curvature) at the edge of the cristae, as seen from the matrix [40].The F1FO-ATPase energy-transduction mechanism of bio- energetics [24] and its modeling ability on mitochondria [29,36] turns into an energy-dissipating machinery when the mitochon- drial Ca2+concentration abruptly increases under pathological conditions [43]. In this case the F1FO-ATPase activated by Ca2+ instead of Mg2+would form a channel which matches the conductance properties of the MPTP [44,45]. The pore opening leads to transient IMM depolarization and allows the diffusion of solutes, water enters and ATP is hydrolyzed by the F1FO-ATPase [4]. Disruption of mitochondrial homeostasis induces swelling and bursting of IMM, an event which has been linked to pathways leading to cell death [46,47]. The MPTP opening has been reported to be affected by a large variety of effectors and conditions, span- ning from ion concentrations to physical changes.Accordingly, Ca2+, Mg2+, adenine nucleotides, Pi, H+ and membrane potential have been claimed as F1FO-ATPase modulators and MPTP inducers/ inhibitors [48]. On considering that the F1FO-ATPase is not only ruled by its own substrates/products, but also responds to post- translational modiications (PTMs) on different subunits [49], PTMs affecting the enzyme function or its super-complex organi- zation may greatly impact MPTP modulation [48]. Furthermore, cyclosporin A (CsA) blocks the MPTP binding to CyPD, a protein which modulates the MPTP without being an essential component of its structure [50,51].Additionally, the putative MPTP-CypD interaction involves OSCP subunit of F1 sector [52], which con- nects the catalytic (aβ)3 spherical complex to the peripheral stalk. However,since MPTP opening was Medial extrusion recently reported to remain CsA-sensitive also in the absence of OSCP [53] or pH-dependent by protonation of the unique histidine in OSCP subunit [54],the whole mechanism need to be clariied.On balance, the experimental evidence accumulated up to now points out that the MPTP-forming properties in eukaryotes [20,21,55] are apparently linked to the dual F1FO-ATPase function. Even if some points remain controversial and require to be eluci- dated, the enzyme of life that synthesizes ATP, activated by Mg2+, apparently turns into the enzyme of death when hydrolyzes ATP in the presence of Ca2+.

3.The hypotheses on the F1FO-ATPase involvement in the MPTP
From the beginning of the story, namely since cell death was associated with an abruptly increased mitochondrial permeability, it seemed quite obvious that the pore formation should involve mitochondrial membrane components whose conformational changes under certain conditions eventually made the inner membrane itself permeable to solutes. The most likely candidates were membrane-bound proteins, since the pore was always supposed to be proteinaceous [6,48]. However, the MPTP constitution was never elucidated and it was also proposed to include a non- proteinaceous ion-conducting module [56].Assumed that the F1FO-ATPase contributes to MPTP formation, as recent advances strongly hint, up to now two main mechanisms have been hypothesized: the channel forms within the c-ring of the FO sector [20,57] or, alternatively, at the interface between the contact region of the dimer [21].The “c-ring hypothesis” is sustained by intriguing hints on the channel conductance,obtained on mammalian pure c subunits reconstituted in lipid bilayers capable of generating spontaneous electrical oscillations activated by cGMP and inhibited by Ca2+ [58]. However, the current generated Lonidamine nmr through the c subunit pore was found to be cation-selective Antipseudomonal antibiotics [59], while the MPTP is known to be induced by Ca2+ and non-selective for solutes with a molecular weight of up to 1500 Da [60]. Otherwise, the dephosphorylation of a peptide related to c subunits promoted by Ca2+ and prevented by CsA emerged as MPTP inducer [61]. The formation of a voltage- sensitive channel in reconstituted c subunits or puriied F1FO- ATPase in liposomes was hypothesized. The pore would be formed by CyPD and Ca2+-dependent c-ring expansion and F1 detachment [57]. This multi-conductance channel lacked cation selectivity, but was resistant to CsA, insensitive to Ca2+ and inhibited by the β subunit of F1FO-ATPase [57]. Experiments of overexpression or depletion of endogenous c subunits by speciic siRNA down- regulation and consequent MPTP inactivation strengthened the “c- ring channel” hypothesis [20].

A proper c-ring conformation is required for MPTP opening [62] and the highly conserved c subunit Gly zipper domain apparently plays a key role in the c-ring as- sembly linked to MPTP sensitivity [57,62]. Recent insights are provided by the Ca2+-induced de novo water-permeable MPTP complex, which is inhibited by CsA, and apparently made up by c subunits associated with polyphosphate (polyPi)and poly- hydroxybutyrate (PHB) [63]. The interaction of the c subunits with polyPi and PHB, by generating a charged polymer, provides an environment compatible with the hydrophobicity of these pro- teolipids. Indeed, the c-ring ion conduction gaps were offset by the electrophysiological properties of the non-proteinaceous Ca2+-se- lective polyPi/PHB channel [56]. Otherwise, as far as we are aware, being the structural arrangement of this c-subunit assembly still unknown, the c-ring cannot form a water-illed channel as sug- gested by the atomistic simulations of two c-rings of different lumenwidth, a conditionwhich would be consistent with the MPTP conductance levels [64]. The c-ring lumen contains lipids in bac- teria [65], while in mitochondria it would be occluded by lipid molecules and forms a non-conducting channel [64]. Moreover, even in a potentially conducting state (i.e. hydrated state of the c- ring interior), molecular dynamics simulations demonstrated that the biophysical properties of such channel were not consistent with the high ionic conductance attributed to the MPTP [64].

However, atomic simulations on the c-ring do not take the polyPi/PHB model into account. Accordingly, due to the PHB amphipathic properties, lipids may localize in the c-ring hydrophobic core and even allow ion flux [63]. Even if the structural bases of the “c-ring channel” opening [57] necessarily require CyPD binding and inhibition by CsA only after Ca2+ addition, CyPD-null mice showed electro- physiological MPTP features indistinguishable from those of wild- type individuals in presence of high Ca2+ loads [66,67]. Intrigu- ingly, aging is known to be associated with an increase in mito- chondrial dysfunctions and MPTP formation [68,69]. Consistently, the decrease in F1 content with respect to FO found in aging heart mitochondria [70], shoulders the hypothesis that FO components, above all c-subunits, are directly involved in MPTP formation.The possibility that the channel may be opened by the displacement of the two main F1FO-ATPase domains has also been considered. However, strip down F1from FO sector generally occurs under drastic conditions, for instance in the presence of high urea concentrations, and leads to irreversible F1FO-ATPase denaturation, whereas the MPTP reversibly shifts between open/closed states. Moreover, the displacement of the two F1FO-ATPase sectors is not a likely mechanism to create a channel within FO, because the Y, δ and ε subunits of F1 cannot return to their native position within the hydrophobic sector. Additionally, it seems dificult to think that free β subunits can inhibit the MPTP formation because the cata- lytic subunits are linked into (aβ)3 globular hexamer resistant to denaturation, yet the hexamer does not interact with the embedded membrane sector in wild-type F1FO-ATPase [27]. Finally, human cells in which the c subunit genes are disrupted preserve the typical MPTP properties [71]. These vestigial F1FO-ATPases in cells unable to synthesize the c subunits are also structurally devoid of a and A6L subunits and cannot translocate H+, even if the CsA- sensitive MPTP formation is maintained [71]. On these bases it seems reasonable to conclude that in the absence of c subunits the MPTP could be formed and be sensitive to CypD, but it remains unclear if its conductance properties occur through an unregulated MPTP pathway or not [72].

If the “c-ring hypothesis”, even if intriguing, show some in- consistencies, other recent indings corroborate the involvement of the dimeric form of the F1FO-ATPase. Accordingly, the MPTP for- mation was observed after reconstitution into lipid bilayers of gel- puriied F1FO-ATPase dimers associated with the detection of an indistinguishable channel current ascribable the MPTP electro- physiological equivalent mitochondrial mega-channel [21]. Cross- linking experiments from Bernardi’s group indicate that the OSCP subunit of peripheral stalk interacts with CyPD [52] and benzodi- azepine 423 (Bz-423), a MPTP inducer which overlaps the CyPD binding site [21]. Bz-423 inhibits the F1FO-ATPase activity similarly to the Pi-dependent CyPD which, by binding to the OSCP subunit, decreases the Mg2+-dependent ATP hydrolysis in the absence of CsA [52]. Thus, CyPD modulates MPTP opening and CyPD binding to OSCP may propagate the conformational changes of the catalytic sites through the stator to enzyme membrane portions [73]. The high matrix Ca2+ concentration features MPTP activation. Therefore the“F1FO-ATPase peripheral stalk/dimer hypothesis” is fully consistent with the occupation of the catalytic site by Ca2+in replacement of natural cofactor Mg2+[48,73]. In the catalytic site the βThr163 of the P-loop is directly linked to Mg2+, while the βArg189, βGlu192, βAsp256 residues are coordinated with Mg2+ by three water molecules respectively [74](Fig. 2b). The metal binding pocket can be occupied by other divalent cations such as Ca2+ [75]. Interestingly, experiments on prokaryotes showed that the single mutation of βThr159Ser at the catalytic site equivalent to βThr163Ser in eukaryotes is the only aminoacid substitution which allows the normal F1FO-ATPase function if sustained by MgATP, but not by CaATP[76].A recent paper [77] investigated the effects of the βThr163Ser mutation in human F1FO-ATPase. In comparison with wild type mitochondria, ATP hydrolysis driven by Mg2+ was stim- ulated, while the Ca2+-dependent F1FO-ATPase activity was nearly

Fig. 2. Detailed representation of the F1FO-ATPase subunits involved in catalysis. a) The OSCP subunit (grey) and the β subunit (red) with the “long connecting loop” and the “crown region” highlighted in olive and blue, respectively. The CaATP substrate in the catalytic binding site is depicted as ball and stick model (modiied PDB ID code: 5ARA). The cofactors Mg2+ (yellow) (b) and Ca2+ (c) (turquoise) are depicted as inserted spheres in the βTP site (modiied PDB ID code:2JDI). ATP molecule and the side chains of Thr163, Glu192, Asp256, Arg189 are drawn as ball and stick models. The Ca2+ insertion is represented assumed that the cation can form the same bonds as Mg2+. The igure illustrates the different steric hindrance of Ca2+ and Mg2+ in the catalytic site completely inhibited. Moreover, the mutation in β subunit appar- ently decreased the MPTP sensitivity to Ca2+, since higher Ca2+ levels were required to induce MPTP opening [77]. The βThr163 is probably the only aminoacid which directly binds to the metal cation in the catalytic site. Consistently, while in presence of Mg2+ the mutation in the catalytic subunit apparently favors ADP release during hydrolysis, the larger Ca2+ would cause a spatial rear- rangement that stiffens the F1 sector and limits OSCP motility. Molecular dynamics simulations suggest that the mechanical en- ergy of Ca2+ bound to β-subunits sites may be transmitted through a long connecting loop to the “crown region” of the OSCP subunit (Fig. 2a). According to this mechanism, the motion starting from the catalytic sites would be transferred through the lateralstalk to the membrane subunits where the MPTP opens [77]. Looking at the residues putatively involved in this mechanism, some concern might arise about the Ca2+ speciic conformational changes trans- mission from the catalytic binding sites to OSCP, since only the N- terminus of the three a subunits interacts with OSCP at the N-ter- minal a-helical domain.

Other interactions between the (aβ)3 spherical domain and the peripheral stalk are established, con- necting the a subunit N-terminal region to b and F6 subunits respectively and the a subunit C-terminus to d subunit [29,78]. However, He et al., [53] recently provided experimental evidence that peripheral stalk subunits (i.e. OSCP and b subunit) are not involved in the MPTP formation. Consistently, the F1 modiication induced by Ca2+ cannot be transmitted to the membrane subunits when the stator is defective. In cells with F1FO-ATPase devoid of OSCP or if the peripheralstalk lacks the transmembrane a-helices of b subunit, mitochondria retained a reduction of MPTP- dependent swelling rate responsive to CsA inhibition, thus sug- gesting that the binding site for CyPD is not provided by OSCP and in this vestigial F1FO-ATPase the channel size is affected [53]. Nevertheless, the inhibition of the MPTP opening by acidic pHs is sensitive to the protonation of the unique histidine in OSCP subunit (OSPCHis112 in humans) [54], thus leading Antoniel and colleagues suspect that more than one subunit of the peripheral stalk can transmit the full-conductance signal for MPTP opening. In addition, the overexpression of e subunit, which is known to promote the F1FO-ATPase dimerization [79,80], limits the MPTP induction by Ca2+. Briefly, even if there is still much work to be done to clarify the MPTP opening mechanism and some still unexplained contra- dictions emerge from literature data, experimental evidence gathered up to now indicates that MPTP opens when the F1FO- ATPase dimers dissociate and the c-ring maintains an adequate conformation [62].

4.A new conformational transmission model for MPTP opening
On considering the hypotheses so far proposed, some questions are still open. The MPTP, the F1FO-ATPase and Ca2+ depict an enigmatic triangle [81], in which Ca2+ apparently plays the leading role. From the available literature data and of some recent indings in our lab, we become increasingly convinced that Ca2+ by inter- acting with the F1FO-ATPase triggers subsequent conformational events which ultimately lead to form a pore in the inner mito- chondrial membrane. In our opinion, there are many clues that lead to build an intriguing and quite realistic model. Accordingly, when the mitochondrial Ca2+concentration increases, it replaces the natural cofactor Mg2+ in the catalytic site of the F1FO-ATPase [77]. As recently pointed out [82], the catalytic mechanism of ATP hy- drolysis and H+ translocation by the Ca2+ and Mg2+-dependent F1FO complexes are apparently similar, in contrast with previous reports [83e85]. Interestingly, small molecules or cofactors (e.g. nitrite or NAD+) were found to act differently on the enzyme when it is activated by Ca2+ or by Mg2+, namely they inhibit the Ca2+- dependent F1FO-ATPase without affecting the Mg2+-dependent F1FO-ATPase [86,87]. Therefore, the catalysis modulation may represent a molecular mechanism which is somehow involved in MPTP regulation. As far as we are aware, the Ca2+-activated F1FO- ATPase cannot synthesize ATP, but it is capable of ATP hydrolysis, probably by adapting the catalytic mechanism, which is compatible with the greater steric hindrance of Ca2+ with respect to Mg2+ when inserted in the β subunits [82]. So, in the presence of the larger Ca2+ radius,the coordination geometry of the cofactor- binding site of the enzyme would change from six-fold octahe- dral up to allow eight ligands, resulting into a less rigid geometry with irregular distances and angles [88]. The Mg2+ and Ca2+ stimulated enzyme activities, thus suggesting that the two cations similarly interact with the protein and the nucleotides [82,88]. Nevertheless, Ca2+ could bind to the same aminoacid residues as Mg2+ with even greater afinity (Fig. 2band c). Moreover, ATP hy- drolysis, which causes the mitochondrial ATP pool depletion associated with the MPTP [3], drives the torsional mechanism of the central stalk in the Ca2+-activated F1FO-ATPase [89].

This tor- sion is coupled to H+ pumping [82] through the IMM, even if H+ translocation is unable to re-energize the IMM [48,82]. As a result, when the F1FO-complex works in the reverse mode (ATPase) driven by Ca2+ there is ATP dissipation without membrane polarization. So, it seems reasonable to think that the Ca2+-activated F1FO- ATPase can start a multistep process resulting into MPTP formation and opening [90]. According to this model, the spatial rearrange- ment within F1would arise from Ca2+ binding to the catalytic sites, and, in the form of conformational change signal, would be trans- mitted to reach the hydrophobic FO sector, where it would promote the dissociation of F1FO-ATPase dimers into monomers and deter- mine the loss of the local curvature of cristae, thus making the MPTP open (Fig. 3). These events, namely the F1FO-ATPase activa- tion by Ca2+ and the deformation of cristae, are both associated with mitochondrial dysfunction and cell death due to MPTP opening. The F1FO-ATPase dimerization in mammalian mitochon- dria could arise from the a and e subunits, while in yeast mito- chondria also i/j and k subunits would participate in maintaining the dimer joined. These supernumerary membrane subunits in yeasts could be functional orthologs of the 6.8 proteolipid and DAPIT subunits in mammalian mitochondrial F1FO-ATPase, respectively. Interestingly, two adjacent a subunits in mammals form a “dimerization motif” in which the a subunit of each monomer has a strand planar structure that connects the two monomers. The cristae are bent by e, g, and the N-terminal portion of b subunits forming an unusual transmembrane domain [36] which, when the F1FO-ATPase dimerizes, induces the positive cur- vature of the membrane which produces the morphology of the cristae.

Indeed,F1FO-ATPase monomers are per se suficient to produce curvature in lipid bilayers (a 43。inclination is imposed by the membrane stator domain) and the detachment of dimers pre- vents the formation of the edge of the cristae, resulting in a zig-zag topology of the membrane [91]. Conversely, the positioning of di- mers along rows in a ridge is a self-association of side-by-side union of multiple dimeric F1FO-ATPase super-complexes which does not require additional protein-protein interactions [40,92].
The stiffness of the peripheral stalk holds the(aβ)3 globular hexamer in a stationary position with respect to the IMM, even if the stator has the adequate flexibility to allow the rotary catalytic cycle [93]. The dimer interface is formed at the basis of the pe- ripheralstalk. Thus, any unusual flexibility induced in this structure can compromise the stability of the dimeric F1FO-ATPase supra- molecular assembly. However, MPTP persists in the absence of the peripheral stalk subunit [53]. Interestingly, vestigial F1FO- ATPase complexes, which lack the c-ring, the a and the A6L subunit or either the b subunit or the OSCP subunit, retain the MPTP fea- tures. All these vestigial enzymes were found abundantly associ- ated with two forms of the intrinsic inhibitor protein (IF1) [53,71]: in particular a mature inactive IF1 form (IF1-M1 but not the IF1-M2 isoform) and the import precursor IF1, IF1-P (amature IF1 form with import sequence). The same proteins were also associated with the monomeric F1FO-ATPase derived from p0 cells [94]. It is still un- known whether IF1-M1 and IF1-P differ in the entry pathway to mitochondria or in ATP hydrolysis inhibition, but probably the speciic association of IF1-P with the F1FO-ATPase could prevent dimerization [94]. Moreover, the active IF1 form (an antiparallel a- helical coiled coil dimeric structure) in the matrix can associate with ive F1 subunits in different conformations [95].

Fig. 3. Putative involvement of the Ca2+-activated F1FO-ATPase complex (es) in MPTP formation. A) The dimeric form of the Mg2+-activated F1FO-ATPase super-complex is associated with a highly convex membrane curvature which protrudes into the matrix; B) the dissociation of the F1FO-ATPase dimers, produced by the mechanical signal transduction from the Ca2+-activated F1 sector to the FO sector (as detailed in the text), reduces the membrane curvature at the apex of the cristae. By this mechanism, the channel forms between two adjacent monomers. (Modiied PDB ID codes: 5ARA and 6B2Z) is not essential for dimer formation [96,97], even if it can promote the F1FO-ATPase dimeric structure super-complexes [98]. The dimeric IF1 role is the ATPase inhibition by binding with a ratchet- like mechanism to the a/βDP site [99] proximal to the peripheral stalk [29]. However, some clues suggest that IF1 may also play a role in MPTP opening [100]. The active form of IF1 occurs at pH values below 6.5 [101], when the mitochondrial ATP hydrolysis to re- energize the IMM causes a disastrous ATP drop and cellular acidi- ication. Interestingly, when pH lowers to 6.5, the protonation of histidine residue(s) near the region of dimer combination shifts the IF1 equilibrium from tetramers (inactive IF1 form) to dimers (active IF1 form) so as to interact with F1 sector and inhibit the F-ATPase activity [101]. The reversible histidine(s) protonation on the mito- chondrial matrix side [102] as well as the OSPCHis112 protonation [54] are known to inhibit the MPTP. As a matter of fact, when the matrix pH decreases below 7.0, the MPTP is generally in the closed conformation, even if MPT was reported to be induced at low pHs in brain mitochondria [103]. Dithyl pyrocarbonate (DPC) allows MPTP opening at pH 6.5 and maintains the Ca2+-dependent channel sensitivity to CsA. Accordingly, DPC reacts with histidyl residues and prevents their reversible protonation [102].

Therefore, at acidic pHs IF1is in the active form and the MPTP is closed. Conversely, at pH values > 7.0, the IF1 dimers aggregate into a tetramer which occludes the inhibitory portion of F1 sector, the F1FO-ATPase can hydrolyze ATP [104] and the MPTP is also closed. At alkaline pHs, to avoid detachment from the F1FO-ATPase under non-inhibitory conditions, the C-terminus of IF1 is kept anchored to FO [105,106], while the N-terminal region of IF1 is cross-linked to the a subunit [107]. In the presence of an incomplete structure of the peripheral stalk, the Ca2+-induced conformational change within F1 could be transmitted to the membrane subunits by the inactive form of IF1 [108]. Accordingly, IF1 by anchoring to the a subunit [107] and to a still unidentiied membrane receptor of known molecular mass (approximately Mr 5400e6400 Da) [105] corresponding to one of the membrane subunits (A6L, e, f, g, DAPIT subunits and 6.8 kDa proteolipid) [109], could create a low density bridge-like structure between the two monomers [37]. This connecting structure maybe responsible for the conformational transmission and drive the conformational change from the Ca2+-bound catalytic subunits to the membrane subunits [108]. This transmission would draw up the dimer stalks and cause the reduction or even the loss of the membrane convexity between the two matched FO sectors [110]. Briefly, if the protonated state of IF1inhibits the MPTP and F-ATPase activity,the deprotonated state of IF1 at alkaline pHs acts as transmitter and forms a connecting arm between F1and FO dimers [108]. Moreover, the MPTP inhibition at low osmotic strength is linked to changes in the IMM curvature [111].

Consistently, the pore formation between the two F1FO-ATPase monomers undoes the supra-molecular dimeric structure [62]. According to this model, the torque generation driven by ATP hydrolysis pushes the rotors of two adjacent F1FO-complexes towards opposite directions, allowing the respective monomer axis to contribute to reduce the distance between the two F1 sectors from 15 to 10 nm [112]. The consequent curvature inversion of the cristae, associated with the spatial re- arrangement of the membrane subunits responsible for the “bridge shape domain”, results in MPTP opening [113] (Fig. 4). Noteworthy, the rotation of the Ca2+-activated F1FO-ATPase [89], driven by ATP hydrolysis, can only occur in the presence of a correct structure of the c-ring or even in the absence of the c-ring. Accordingly, a proper c-ring or vestigial F-ATPases unable to build the c-ring preserve the MPTP properties [62,71]. Once transmitted through conformational changes, such rotation may be indirectly responsible for the stalk-to-stalk distance modiication involved in MPTP opening. The supra-molecular arrangement of the F1FO- ATPase complexes is consistent with the ultrastructure of the cristae and with the ideal mitochondrial bioenergetics. When the F1FO-ATPase dimeric and oligomeric forms were destabilized by mutated e subunits, the F1FO-ATPase activity was maintained.

Fig. 4. FO dimerization and IMM curvature change as related events. The membrane dimeric domain is formed by the membrane-intrinsic a-helices of b subunit, the a subunit, the A6L subunit, the f subunit, the e subunit and the g subunit of each monomer. According to the model, the reduction of the concave IMM curvature dissociates the membrane- embedded FO dimer and creates a pore at the monomer-monomer interface. Color subunits are the same as in Fig. 1. (Modiied PDB ID code: 6B2Z). The left side column illus- trates the membrane bending angle corresponding to the central portion of the igure (FO-couple), viewed from the intracristae space if unexpectedly accompanied by IMM Δ4 reduction [114]. On the other hand the e subunit overexpression, which supports the dimeric structure, was found to limit MPTP formation and prevent IMM depolarization [62].To sum up, the experimental evidence accumulated up to now strongly suggests that MPTP formation and opening result from a multi-step process in which conformational changes play a key role. The data accumulated up to now are well compatible with a conformational mechanism triggered by the Ca2+-dependent F1FO ATP hydrolysis, which implies a spatial rearrangement within F1. According to the depicted model, this conformational change would constitute a structural signal that, once transduced to the membrane subunits, promotes the monomerization ofF1FO-ATPase super-complexes and the curvature inversion at the apex of the cristae. So, the features of the Ca2+-activated F1FO-ATPase are fully consistent with the Δ4 loss and ATP hydrolysis, mitochondrial events known to be associated with the MPTP [3,90].