*4.1. The F1FO-ATP Synthase from the Energy Production to Mitochondrial Morphology*

In the energy-transducing mtIM, the synthesis of ATP is based on the coupling of the rotary mechanisms of the two main sectors, namely the hydrophobic F<sup>O</sup> and the hydrophilic F<sup>1</sup> [131,132]. The protonmotive force (∆*p*), built by substrate oxidation in the respiratory chain, drives H<sup>+</sup> translocation through F<sup>O</sup> domain by generating torque and allowing conformational changes in F<sup>1</sup> which leads to ADP phosphorylation to yield ATP. In mitochondria, the bi-functional enzyme can also work in reverse when the ∆*p* drops. In this case, ATP hydrolysis by the F<sup>1</sup> domain fuels H<sup>+</sup> pumping by F<sup>O</sup> which reenergizes the mtIM [133]. This energy-dissipating mechanism has also been associated with a variety of pathological conditions [134]. Both in the forward and in the reverse reaction, H<sup>+</sup> translocation across the mtIM is due to the reversible protonation/deprotonation of carboxylic sites, which also converts H<sup>+</sup> flux into F<sup>O</sup> rotation. The disconnection between

these two matched activities, namely H<sup>+</sup> -translocation and catalysis, "uncouples" the F-ATP synthase and often leads to mitochondrial dysfunctions. The enzyme sensitivity to the antibiotic oligomycin, which by binding to the enzyme covers the H<sup>+</sup> binding sites and blocks the *c*-ring rotation and enzyme catalysis in either directions [135], is taken as a parameter of the efficient coupling between the two domains.

This amazing rotary enzyme complex offers a good example of how molecular structure and function are tightly linked features [130], which cannot work without each other.

**Figure 3.** Structure of the mitochondrial F1FO-ATPase in mammals. The enzyme subunits are drawn as ribbon representations obtained from modified PDB ID code: 6TT7. The differently colored letters identify the subunits, drawn in the same color as the letter.

#### *4.2. Structural Basis of the Bioenergetic Mechanism*

Δ Δ Among all F-ATPases, the mitochondrial enzyme complex has the most complex subunit composition [136]. F<sup>1</sup> protrudes in the mitochondrial matrix as an asymmetric hexagonal globular assembly of α and β subunits, arranged as (αβ)<sup>3</sup> around the γ subunit of central stalk which connects F<sup>O</sup> to F1. The hexamer, in which α and β subunits alternate, hosts three catalytic and three non-catalytic sites. The three catalytic β-sites can adopt three conformations, namely "open", "closed", and "semi-closed", defined as βE, which is empty, βTP, which contains Mg-ATP or Mg-ADP and βDP, which hosts Mg-ADP. Each of the three non-catalytic α subunits binds Mg-ATP [137] and also Mg-ADP [128]. According to the binding change mechanism, the β conformations interconvert each other as the *c*-ring rotates and transmits the rotation to central stalk [138], which when it meets the catalytic sites, changes their conformation and affinity for nucleotides. So, during a complete turn (360◦ ) each catalytic site undergoes (stepwise) all the three different conformations leading to synthesize/hydrolyze three ATP molecules. Moreover, Mg-ATP bound to the non-catalytic sites allows ADP release from the βDP site and removes the Mg-ADPdriven enzyme inhibition during multiple ATP hydrolysis turnover [139]. Interestingly, the mammalian γ subunit shows two isozymes, a "heart isoform" and a "liver isoform", which differ by a single amino acid, an additional aspartate at the C-terminus of the liver enzyme. The former is expressed in tissues with a high and variable energy demand such as heart, muscle, diaphragm, while the latter is expressed in brain, thyroid, spleen, pancreas, kidney, testis and liver, which, apart from brain, show a low ATP consumption and a steady energy demand. Even if the C-terminus Asp in the liver γ-subunit may form salt linkages and decrease the enzyme efficiency, no differences in enzyme kinetics were found, so the two γ isoforms probably only have a regulatory function [140].

Other than by the γ subunit, the functional coupling of the two main sectors F<sup>O</sup> and F<sup>1</sup> in mammalian mitochondria is ensured by the peripheral stalk and especially by OSCP [134].

The membrane-embedded F<sup>O</sup> domain, which allows H<sup>+</sup> flow across the mtIM, hosts the *a* subunit, which channels H<sup>+</sup> , and the *c*n-ring, namely the rotor, which consists of a cylindrical palisade of n subunits. The number n varies in the range 8–15 among the species, is constant in the same species and determines the ATP bioenergetic cost, which is obtained from the ratio between the number of translocated H<sup>+</sup> and the three ATP molecules constantly produced in a complete (360◦ ) rotation of the rotor. Accordingly, the number of transported H<sup>+</sup> per complete rotation of the *c*-ring is related to the number of H<sup>+</sup> binding site(s) on each *c* subunit. In general, an increase in the *c*-ring size and consequently in the number of H<sup>+</sup> -binding *c*-subunits, implies an increase in the bioenergetic cost of ATP, since more H<sup>+</sup> should flow downhill across the mtIM to produce one ATP molecule. Mammals which show a small *c*-ring (n = 8) are highly efficient by dissipating less H<sup>+</sup> to yield one ATP, namely they obtain ATP at a low bioenergetic cost [141]. So, the number of *c* subunits, and of H<sup>+</sup> binding sites, can be taken as a rotor efficiency parameter. In the OXPHOS system the transmembrane driving force ∆*p* which links substrate oxidation to ADP phosphorylation to yield ATP, consists of an electrical ∆Ψ (membrane potential difference) and of a chemical component, ∆pH (pH gradient) between the positive and negative side of the mtIM. During evolution, species have adapted the *c*-ring size to the prevailing electrochemical parameter, ∆Ψ or ∆pH. Accordingly, when ∆Ψ prevails and the pH gradient is low, the *c*-ring is small [142]. This is the case of mammals, generally vertebrates and some invertebrates [143]. Conversely, a prevailing ∆pH is associated with the large *c*-rings (n = 11–15), typical of chloroplasts and bacteria [144,145].

The H<sup>+</sup> pathway within F<sup>O</sup> only recently has been elucidated. It is built step by step within the *a* subunit by selected amino acid side chains, which form a quite unexpected route along the mtIM allowing membrane crossing in a perpendicular way to the two aqueous half-channels. The two half-channels are discontinuous and open one in the matrix and the other in the lumen and both contain two horizontal highly conserved α-helices, named H5 and H6, which lie along the mtIM and are juxtaposed to the *c* subunits, thus providing access to the *c*-ring H<sup>+</sup> binding sites [146]. This α-helix arrangement in the mtIM features all rotary A-, F- and V-type ATPases [147]. Each *c* subunit is hairpin-shaped and spans the mtIM. These gathered hairpins form a sort of hourglass, seen laterally from the membrane side, whose concavity hosts the H<sup>+</sup> -binding sites on the outer C-terminal α-helices of *c* subunits. Therefore, the horizontal α-helix arrangement perfectly fits the rotor concavity [148].

The H<sup>+</sup> transfer across the mtIM in either direction requires subsequent protonation/deprotonation steps of H<sup>+</sup> binding sites that consist of amino acid side chains that interconvert between the protonated and deprotonated states. Interestingly, in mitochondria at the edge of the *cristae*, the pH value is higher than that in the IMS opposing the inner boundary membrane (IBM), as well as the *cristae* ∆Ψ generated by the respiratory chain [149] is higher than in the IBM [150]. The H<sup>+</sup> flow through F<sup>O</sup> for ATP synthesis occurs from the lumen or *intracrista* space of the half-channel formed by the hydrophilic cavity created by the end of H5 and H6 helices of *a* subunit, and H3 helixes of *b* and *f* subunits. The H<sup>+</sup> pathway starts from *a*His-168 and *a*His-172 on H5 helix and ends at *a*Glu-203 on H6 helix of *a* subunit, which acts as intermediate H<sup>+</sup> -donor to *c*Glu-58 of the *c*-ring [128,151]. The Glu-His interactions establish multiple H-bridges in the membrane half-channel that, by changing the carboxylic group p*K*a, allow its protonation. On the

opposite half-channel on the luminal side viewed from the *intracrista* space, the *a*Glu-145 on H5 (matrix) helix acts as "de-protonator" which restores the original p*K*<sup>a</sup> of *c*Glu-58 allowing the H<sup>+</sup> detachment. The half-channel arises where a conserved *a*Pro-153 bends the *a*H5 helix [151]. Then, the folding is enforced by DAPIT subunit and H4 helix of *a* subunit in order to make the half-channel hydrophilic core conformationally adapt to the *c*-ring shape (Figure 4).

α

α

ΔΨ

α α

**Figure 4.** Proton translocation pathway within F<sup>O</sup> during ATP synthesis. On the top the H<sup>+</sup> entry and exit into and from the half channels, viewed from the IMS. On the left and right sides, the H<sup>+</sup> outlet and inlet half-channels, respectively, after the rotation angles above the arrows are shown. The horizontal helices H5 and H6 of *a* subunit (violet) and four *c* subunits (gray), drawn as ribbon representations, were obtained from modified PDB ID code: 6TT7. The amino acids of *a* subunit involved in H<sup>+</sup> translocation are drawn as ball and stick models.

During the H<sup>+</sup> translocation through the membrane-embedded F<sup>O</sup> domain, a putative participation of hydronium ion (H3O<sup>+</sup> ) was also suggested [152]. Accordingly, in prokaryotes and eukaryotes the bell-shaped pH profile of the F1FO-ATPase inhibition by dicyclohexyl-carbodiimide (DCCD), which covalently binds to *c*-subunit carboxylates, is not compatible with carboxylate protonation of H<sup>+</sup> binding sites, but it may be explained by H3O<sup>+</sup> coordination [143,153]. Indeed, on considering the position of molecules able to establish hydrogen bonds in the half-channels, whose core is hydrophilic, H<sup>+</sup> translocation may be coordinated by water molecules by the Grotthuss mechanism, namely by proton jumping from covalent to H-bonds and vice versa [128,154]. During ion translocation, the *c*Glu-58 adopts different conformations, namely the H<sup>+</sup> -unlocked anionic form (carboxylate) to face the aqueous environment of the half channels, and the H<sup>+</sup> -locked form of the protonated carboxylic group to face the hydrophobic environment of the mtIM. In the locally hydrated luminal half-channel, the *c*Glu-58 carboxylate is oriented in an outwardfacing (H<sup>+</sup> -unlocked) conformation before protonation and turns to an inward-facing closed (H<sup>+</sup> -locked) conformation when protonated. On the opposite half-channel, which faces the matrix, the *c*Glu-58 in the H<sup>+</sup> -locked conformation turns to the H<sup>+</sup> -unlocked conformation after deprotonation. The *c*-ring key carboxylates embedded in the mtIM

are always in the H<sup>+</sup> -locked conformation to enter the lipophilic mtIM [155] and make the *c*-ring rotate due to Brownian motion [131]. The rotation direction when viewed from F<sup>O</sup> toward F<sup>1</sup> is clockwise during ATP synthesis and counterclockwise during ATP hydrolysis [133,156]. The low pH in the luminal half-channel favors *c*Glu-58 protonation and the consequent locked conformation, pushed by ∆*p* that favors the entry to the mtIM. When an almost entire rotor rotation is completed, the *c*Glu-58 reaches the half-channel on the matrix side where the low H<sup>+</sup> concentration as well as the negative charge of the "H<sup>+</sup> -releasing site" on H5 helix of *a* subunit, favors its deprotonation associated with the H<sup>+</sup> -unlocked carboxylate form. The positive charge of a crucial arginine, *a*Arg-159, acts as electrostatic barrier between the two half-channels, prevents H<sup>+</sup> short circuiting and attracts the negatively charged *c*Glu-58 carboxylates [157]. Moreover, *a*Arg-159 helps the *c*Glu-58 de-protonation during H<sup>+</sup> translocation and prevents salt bridges between *a* and *c* subunits which would block the rotation. Interestingly, *a*Arg-159 and the H<sup>+</sup> -unlocked *c*Glu-58 are too far (about 4.5 Å) and cannot interact [144], so the two half-channels for H<sup>+</sup> entry and exit at the *a*/*c*-ring interface are spatially offset to allow the rotation direction adjust to ∆*p* [135].

During ATP hydrolysis the transition between the three β conformations in F<sup>1</sup> drives the rotations of the γ subunits and of the *c*-ring and increases the ∆*p* by pushing H<sup>+</sup> uphill [134].

The coupling of energy transduction and H<sup>+</sup> translocation is allowed by conformational adaptations of the F<sup>1</sup> and F<sup>O</sup> domains [158] that adapt to each other, helped by flexible regions in the peripheral stalk. The latter undergoes torsion during catalysis, but also the *c*-ring shows conformational fluctuations when it contacts the *a* subunit. Therefore, coordinated conformational changes within the enzyme proteins produce and synchronize the torsion.

#### *4.3. The ATP Synthase Supramolecular Arrangement and the Mitochondrial Shape*

The mitochondrial F1FO-ATPase functions are ensured by the enzyme supermolecular arrangement. By self-assembling in dimers and tetramers, the F1FO-ATPase rules most bioenergetic and structural functions in eukaryotic mitochondria [159]. Accordingly, the occurrence of *sms*, which lack in bacteria and chloroplasts, in F<sup>O</sup> is linked to the *sms* involvement in supercomplex arrangement and in the mtIM ultrastructure [130]. Consistently, no F1FO-ATPase dimers were found in prokaryotes and chloroplasts, while some differences between yeast and mammalian F1FO-ATPases occur in the building of contact sites. It seems to have been ascertained that, while the basic subunits allow the overall functionality of the enzyme complex, these *sms* are required to join the monomers and to bend the membrane. The *sms* were differently named according to the taxa: A6L, *e*, *f*, *g*, and DAPIT, which corresponds to the *k* subunit in yeasts, and mammalian 6.8PL, which corresponds to the orthologue *i/j* in yeasts [128]. The membrane subunits are mainly encoded by nuclear genes, apart from *a*, A6L and *c* subunit (identified also as *6*, *8* and *9* subunit) for yeast F1FO-ATPase and only *a* and A6L in mammals which are encoded by the mitochondrial DNA. Notably, different *sms* structurally contribute to the F1FO-ATPase dimerization in yeasts and in mammals [127,160]. Three different isoforms occur in the mammalian *c* subunit.

F1FO-ATPase monomers form dimers arranged in long rows on the mtIM and MICOS (mitochondrial contact site and *cristae* organizing system) complex at *cristae* junctions cooperate to yield the *cristae* morphology in agreement with their opposing effects on membrane curvature [161–163]. Moreover, recent work shows that the F1FO-ATPase dimerization factor subunit *e* interacts with the MICOS component Mic10, which stabilizes the enzyme oligomers and possibly temporal orchestration of cristae biogenesis. Moreover, the F1FO-ATPase dimerization and the MICOS component Mic60 induce opposite membrane curvature, namely the former promotes a negative curvature and the latter a positive curvature (concave and convex as viewed from the matrix, respectively). Most likely, the interaction between MICOS and dimeric F1FO-ATPase, which implies a high mobility of

the enzyme monomers in the *crista* membrane, enables spatial and temporal coordination during *crista* biogenesis and dynamics [164].

Interestingly, the dimerization interface in different species shows different subunit composition with no apparent homology [136].

Four different types of F1FO-ATPase dimers were described in mitochondria. In mammals (e.g., *Bos taurus*, *Sus scrofa domesticus,* and *Ovis aries* heart) and yeasts (*Saccharomyces cerevisiae*) V-shaped dimers, named type I dimers, are localized on the rim of the *cristae* with an mtIM convexity of about 90◦ [137,149]. Type II, III and IV dimers, which differ in the angle between monomers, were only found in algae, protozoa and small invertebrates [148,159].

Mammalian F1FO-ATPases can also form a H-shaped tetramer, which in turn consists of two assembled type I dimers that lie antiparallel to each other. The dimers are joined by two IF1 subunits, an endogenous protein inhibitor [127], evolutionary conserved throughout eukaryotes, which only blocks ATP hydrolysis but not ATP synthesis. Most likely, IF1 normally is not essential for cell survival, but can be crucial under pathological or stress conditions [165], as suggested by its overexpression associated to the decrease in OXPHOS in some cancer types [166]. However, IF1 not only prevents ATP dissipation by ATP hydrolysis, but also contributes to mitochondrial morphology [127,167,168]. The IF1 dimer, which becomes active at acidic pHs in the matrix and stems from the dissociation of the tetramer, inhibits ATP hydrolysis when ∆*p* collapses. In the tetramer, IF1 links two F<sup>1</sup> domains of two opposite dimers; it binds to the catalytic interface between the αDP and βDP subunits in loose binding conformation. Therefore, the two monomers of the laterally opposite dimer are connected by the same IF1; these two IF1 proteins make the tetramer steady and block ATP hydrolysis by a ratchet-like action on the rotor. The F1FO-ATPase dimers are also maintained in rows by a long-range attractive force that stems from the relief of the overall elastic strain of the mtIM [169]. Other than by interactions with IF1 dimers, the mammalian tetramer is maintained by *e*-*e* and *g*-*g* interactions between two opposed and diagonally arranged monomers [170]. Moreover, the monomers are joined side-by-side in a row through DAPIT interaction with *a* subunit at the matrix space and with *g* subunit both at the matrix side and the intra*crista* space. The dimers are linked by interactions of one monomer 6.8PL with *f* and *e* subunits of the other monomer, and vice versa, at the matrix and intra*crista* space, respectively. In addition, the subunits *a*-*a* in the middle of mtIM, the *f*-*f* at the matrix and *e*-*e* at the intra*crista* space establish multiple contacts between each monomer pair [128].

Moreover, the association of two F<sup>O</sup> domains depends on *e* and *g* subunits and on a putative conserved dimerization GxxxG motif in the transmembrane α-helices of both subunits [171,172]. The substitution of a glycine residue by leucine into the *e* subunit motif led to the loss of *g* subunit, destabilized the dimer and resulted in an onion-like structure in the *cristae* [171]. The individual α-helix of *e* subunit and the H3 of *g* subunit interact with their respective GxxxG motif and joined to H2 of *b* subunit form the triple transmembrane helix bundle (TTMHB). The TTMHB tilt is driven by a U-turn structure formed by H2 and H3 helices of *b* subunit. This subunit assembly creates a reminiscent BAR-like domain that bends the mtIM [128,173] to form the apex of *cristae*, but *e* and *g* subunits also keep the monomers together. Accordingly, disulfide cross-link experiments showed the interactions between *e*-*g* or *e*-*e*/*g*-*g* increase the stability of F1FO-ATPase dimers and oligomers, respectively, in mitochondrial digitonin extracts [172].

In mammals, subunit–subunit interactions produce the F1FO-ATPase supermolecular arrangement, in which monomer pairs form dimers and dimer pairs form tetramers, which in turn form long rows of oligomers. This structural arrangement produced by subunit– subunit interactions, also forces the mtIM to be convex at the apex of the *cristae*. The enzyme subunits involved in oligomerization, though not directly involved in catalysis, contribute to the maintenance of *cristae* architecture, which is essential for an efficient respiration by acting as proton sink. A mutation in subunit *k*/DAPIT results in aberrant *cristae* formation, defective F1FO-ATPase dimerization and a mild Leigh syndrome phenotype [164]. *Cristae* remodeling and F1FO-ATPase dimerization may be also involved in cell differentiation [174].

Interestingly, in some forms of cancer, the overexpression of IF1, which leads to the F1FO-ATPase dimerization and *cristae* formation, could contribute to neoplastic degeneration and evasion of apoptosis. The absence of IF1 in Luft's disease, a rare mitochondrial disease, is associated with densely packed mitochondrial *cristae* [175]. Moreover, the acidic phospholipid CL is critical for oligomerization of F1FO-ATPase multimeric complexes either by direct interaction with the enzyme or by inducing the proper membrane curvature. CL, acting like a non-bilayer forming phospholipids, interacts with the F1FO-ATP synthase and may reduce the free energy of the extreme mtIM curvature to stabilize high-curvature folds [164,176]. From these few examples it seems clear that any change in at least one among the multiple factors which rule the mitochondrial morphology can easily lead to pathology. The F1FO-ATPase can work in monomeric and dimeric form; however, the loss of the supramolecular organization of the F1FO-ATPase causes aberrant membrane morphology results in respiratory defects.

On these bases CL defects, as in Barth Syndrome (see Section 3.1.3) can affect the supramolecular arrangement of the F1FO-ATPase, the mitochondrial shape and, as depicted in the following section, the mitochondrial permeability transition.

During aging, the typical rim shape of the *cristae* disappears and the F1FO-ATPase dimers dissociate into monomers. Consistently, the mitochondrial morphology changes, showing a vesicular mtIM, leading to mitochondrial dysfunction and cell death [177]. Even if most results come from in vitro studies on experimental models and should be deepened and confirmed in vivo, recent advances suggest that mitochondrial changes are a driving force, rather than a consequence, of the aging process and neurodegeneration [176]. The connection among the F1FO-ATPase supramolecular arrangement, the mitochondrial shape and some human diseases, as pointed out by these few examples, is also made clearer by the enzyme involvement in the mitochondrial permeability transition pore (mPTP), the topic of the following section.
