The new role of F₁F_o ATP synthase in mitochondria-mediated neurodegeneration and neuroprotection

Abstract

The mitochondrial F₁F_o ATP synthase is one of the most abundant proteins of the mitochondrial inner membrane, which catalyzes the final step of oxidative phosphorylation to synthesize ATP from ADP and Pi. ATP synthase uses the electrochemical gradient of protons (Δμ_H⁺) across the mitochondrial inner membrane to synthesize ATP. Under certain pathophysiological conditions, ATP synthase can run in reverse to hydrolyze ATP and build the necessary Δμ_H⁺ across the mitochondrial inner membrane. Tight coupling between these two processes, proton translocation and ATP synthesis, is achieved by the unique rotational mechanism of ATP synthase and is necessary for efficient cellular metabolism and cell survival. The uncoupling of these processes, dissipation of mitochondrial inner membrane potential, elevated levels of ROS, low matrix content of ATP in combination with other cellular malfunction trigger the opening of the mitochondrial permeability transition pore in the mitochondrial inner membrane. In this review we will discuss the new role of ATP synthase beyond oxidative phosphorylation. We will highlight its function as a unique regulator of cell life and death and as a key target in mitochondria-mediated neurodegeneration and neuroprotection.

Introduction

Mitochondria originated from endosymbiotic alphaproteobacteria more than one billion years ago (Roger et al., 2017). Motility is one of the many features that mitochondria inherited from their ancestors: mitochondria are extremely dynamic organelles; they can move long distances based on cellular energy requirements. The regulation of mitochondrial motility, as a response to different environmental and metabolic changes is crucial for maintaining energy homeostasis to support cellular function. Neurons are especially susceptible to disruptions in mitochondrial motility and quality control mechanisms. Quality control processes include removal of damaged mitochondria by mitophagy and dynamic processes of mitochondrial fusion and fission. Recent studies indicate that dysregulation of mitochondrial dynamics and mitochondrial dysfunction comprise an important underlying cause of aging, different neurological and degenerative diseases, highlighting further the central role of mitochondria in cell death. Mitochondria regulate cell death through Bax/Bak-mediated mitochondrial outer membrane permeabilization (MOMP), which can lead to caspase-dependent apoptosis or caspase-independent cell death (Chipuk et al., 2008; Finucane et al., 1999), as well as by another major mechanism termed mitochondrial permeability transition (mPT). mPT is traditionally defined as a phenomenon associated with the sudden change in the mitochondrial inner membrane permeability to different solutes with a size of <1.5 kDa (Haworth and Hunter, 1979; Hunter and Haworth, 1979a, b; Hunter et al., 1976). The mPT causes osmotic dysregulation of mitochondria leading to mitochondrial swelling, rupture of the outer mitochondrial membrane and cell death (Haworth and Hunter, 1979; Hunter and Haworth, 1979a, b; Hunter et al., 1976). These events are linked to the opening of a mitochondrial inner membrane channel, the so-called mitochondrial permeability transition pore (mPTP) (Kinnally et al., 1989; Lohret et al., 1996; Petronilli et al., 1989; Szabo and Zoratti, 1992; Zoratti and Szabo, 1995; Zorov et al., 1992). mPTP is a high conductance, voltage-gated, non-selective protein channel of controversial molecular content (Kinnally et al., 1989; Lohret et al., 1996; Petronilli et al., 1989; Szabo and Zoratti, 1992; Zoratti and Szabo, 1995; Zorov et al., 1992). A high level of matrix Ca²⁺ was shown to be one of the main factors to induce mPTP opening, along with inorganic phosphate, CypD and oxidants (Bernardi, 1992, 1996; Bernardi et al., 1992; Carraro and Bernardi, 2016; Petronilli et al., 1994; Szabo et al., 1992). mPTP opening can be inhibited by other divalent and trivalent cations, such as, Mg²⁺, Ba²⁺, Gd³⁺, as well as by H⁺ and adenine nucleotides (Bernardi, 1996; Mnatsakanyan et al., 2019; Nicolli et al., 1993; Szabo et al., 1992).

The exact structure and molecular composition of mPTP has been extensively studied and various mitochondrial proteins, the adenine nucleotide translocator (ANT) (Halestrap and Brenner, 2003), the voltage-dependent anion channel (VDAC) (Szabo et al., 1993; Szabo and Zoratti, 1993), the phosphate carrier (PiC) (Kwong et al., 2014) and the translocator protein (TSPO) (Chelli et al., 2001; Li et al., 2007; Pastorino et al., 1994) were suggested to constitute the channel of mPTP. However, further studies revealed the important role of these proteins as channel regulators rather than as a key structural element of mPTP (Baines et al., 2007; Gutierrez-Aguilar et al., 2014; Kokoszka et al., 2004; Sileikyte et al., 2014). The mitochondrial ATP synthase dimers (Carraro et al., 2018; Carraro et al., 2014; Giorgio et al., 2013; Guo et al., 2019; Urbani et al., 2019) or c-subunit ring alone (Alavian et al., 2014; Azarashvili et al., 2014; Bonora et al., 2013; Mnatsakanyan et al., 2019; Morciano et al., 2018; Neginskaya et al., 2019; Pavlov et al., 2005) were suggested and still remain strong candidates for housing the mPTP channel according to several recent studies. Here we will discuss the role of ATP synthase c-subunit leak channel (ACLC) in human health and disease, in mitochondrial homeostasis and remodeling, synaptic plasticity and metabolic efficiency of neurons, brain ischemia and excitotoxicity, neurodegenerative disease and most importantly as a therapeutic target in mitochondria-mediated neuroprotection.

The structure, oligomeric state of ATP synthase and its arrangement in the mitochondrial inner membrane

The overall structure of ATP synthase is highly conserved in all unicellular and multicellular organisms. ATP synthase is composed of two multi-subunit subcomplexes, F₁ and F_O, which drive ATP synthesis and hydrolysis in a unique and efficient manner, by using the rotation of their own subunits (Abrahams et al., 1994; Menz et al., 2001). The mitochondrial ATP synthase has by far the most complex structure and subunit composition compared with its simplest versions found in bacteria. The F_O subdomain consists of subunits a, b, c, d, F6, e, f, g, A6L, 6.8PL, DAPIT (diabetes-associated protein in insulin sensitive tissues) and OSCP (oligomycin sensitivity-conferring protein) (Kuot et al., 2019; Zhou et al., 2015). The c-subunit ring is completely embedded in the membrane and has 8 copies in mammalian mitochondria (Watt et al., 2010). The F₁ subdomain consists of subunits α, β, γ, δ, ε and the regulatory protein IF1, which are located in the mitochondrial matrix. The three copies of α and β subunits alternate with each other to form the hexameric catalytic head of F₁, which does not rotate during ATP synthase catalysis (Menz et al., 2001). Subunits γ, δ and ε form the central stalk of ATP synthase by connecting F₁ with F_O. The central stalk rotates along with the c-subunit ring during ATP synthesis or hydrolysis. F_O subunits b, d, F6 and OSCP form the peripheral stator stalk of ATP synthase by connecting the F₁ α₃β₃γ complex with the a subunit of F_O (Zhou et al., 2015). The additional function of the peripheral stator stalk is to provide the elasticity necessary for efficient coupling of F₁ and F_O motors (Murphy et al., 2019).

When the electron transport chain complexes are fully functional to build the Δμ_H⁺, the downhill flow of protons through F_O drives the rotation of c-subunit ring and the consequent rotation of γδε-subunits of F₁ to synthesize ATP. During the reverse reaction, ATP hydrolysis in β catalytic centers of F₁ powers the rotation of γδε-subunits and the subsequent rotation of the c-subunit ring to drive proton translocation across the membrane (Nicholls, 2013). Cellular conditions inducing the reverse catalytic reaction of ATP synthase will be discussed in later sections of this review.

Recent advances in structural biology, especially in the fields of cryo-electron microscopy and cryo-electron tomography have significantly increased our understanding of ATP synthase structure and its oligomeric state. The high-resolution structure of mitochondrial ATP synthase dimers from green alga Polytomella sp, determined by single-particle electron cryomicroscopy at 6.2 A° resolution revealed a most unexpected horizontal location of a-subunits (Allegretti et al., 2015). This structure provides unique insights into the proton translocation mechanism through F_O and how the passage of protons through the membrane drives ATP production (Allegretti et al., 2015).

Another recent structure of mitochondrial ATP synthase from porcine heart shows the high-resolution structure of ATP synthase tetramer in its inhibited state with IF1 molecules bound to it (Gu et al., 2019). Interestingly, c-subunit lumen in this structure was filled by a 40 amino acid-long alpha-helical protein instead of lipids or detergents as previously reported (Matthies et al., 2009; Meier et al., 2001; Oberfeld et al., 2006). The authors assigned 6.8PL proteolipid subunit of ATP synthase to this density although further studies are required to confirm the identity of this protein. According to this structure, the C-termini of the four F_O e subunits are bent toward the four c-subunit rings in the tetramer, which allows them to interact with the C-terminal ends of the 6.8PL helices occupying the central pore of the four c-rings (Gu et al., 2019). This interaction between subunits c, 6.8PL and e might be important for preventing dissipation of membrane potential under normal physiological conditions by keeping the ACLC closed or substantially reducing the probability of its opening.

Cryo-electron tomographic analysis of mitochondria revealed the morphology of mitochondrial inner membrane cristae and the in situ oligomeric state of ATP synthase (Davies et al., 2012; Davies et al., 2014). Recent reports have shown that ATP synthase molecules are organized in rows of dimers (Blum et al., 2019). Dimeric conformation of ATP synthase is thought to be essential for the formation of the lamellar morphology of the inner membrane cristae (Blum et al., 2019; Davies et al., 2012). ATP synthase F_O subunits e and g are proposed to be crucial for causing the curvature of cristae because a mutant strain of yeast lacking subunit e or g also lacks the normal lamellar shape of the cristae and has balloon-shaped cristae instead (Davies et al., 2012). The role of ATP synthase in shaping the inner membrane cristae and controlling mitochondrial fission and fusion processes will be discussed in later sections.

The role of ATP synthase in human health and disease

A. ATP synthase c-subunit leak channel as a main contributor to mPT

Mitochondrial F₁F_O ATP synthase was suggested to house the channel of mPTP according to several recent studies (Alavian et al., 2014; Azarashvili et al., 2014; Bonora et al., 2013; Carraro et al., 2018; Carraro et al., 2014; Giorgio et al., 2013; Guo et al., 2019; Mnatsakanyan et al., 2019; Morciano et al., 2018; Neginskaya et al., 2019; Pavlov et al., 2005; Urbani et al., 2019). Bernardi’s laboratory first described the specific interaction of ATP synthase OSCP subunit with the peptidyl-prolyl cis-trans isomerase cyclophilin D (CypD) (Giorgio et al., 2009), a well-known regulator of mPTP (Baines et al., 2005; Karch and Molkentin, 2014), which allowed them to hypothesize that ATP synthase is a key player in mPTP formation. Numerous studies performed in Bernardi’s group suggested that the channel is formed by the ATP synthase dimers and subunits e and g, required for ATP synthase dimerization, are also crucial for forming the channel (Carraro et al., 2018; Carraro et al., 2014; Giorgio et al., 2013; Guo et al., 2019; Urbani et al., 2019). The experiments with yeast mutant strains without subunits e and g displayed a remarkable resistance to channel opening (Carraro et al., 2018; Carraro et al., 2014). Additionally, channel formation was observed in lipid bilayer recordings of purified reconstituted ATP synthase tetramers and dimers but not in monomers isolated from bovine and porcine hearts, indicating that ATP synthase dimerization is required for pore formation (Giorgio et al., 2013; Urbani et al., 2019).

Our studies have recently shown that ATP synthase monomer forms a high, multi-conductance channel with the biophysical characteristics of mPTP, suggesting that the channel is not located between e and g subunits (Mnatsakanyan et al., 2019). We reported that the channel resides in the membrane embedded c-subunit ring (Alavian et al., 2014). Several other reports suggest that the c-subunit comprises the pore (Alavian et al., 2014; Azarashvili et al., 2014; Bonora et al., 2013; Mnatsakanyan et al., 2019; Morciano et al., 2018; Neginskaya et al., 2019; Pavlov et al., 2005). The c-subunit was found as the main component of the chloroform extract of mitochondrial membranes from rat liver; the channel activity of this extract closely resembled the behavior of the mPTP (Azarashvili et al., 2014; Pavlov et al., 2005). We found that the affinity purified c-subunit indeed forms a large, voltage-sensitive, multi-conductance channel in patch-clamp recordings (Alavian et al., 2014). However, the activity of the c-subunit channel was not sensitive to the well-known mPTP regulators, cyclosporine A (CsA) and Ca²⁺, since these regulators bind to ATP synthase F₁ subunits or F_O peripheral stalk subunits located in the matrix (Alavian et al., 2014). Depletion of the c-subunit attenuated Ca²⁺-induced depolarization of the inner mitochondrial membrane of primary hippocampal neurons, while its overexpression sensitized neurons to cell death. We also reported that c-subunit ring undergoes measurable conformational changes by enlarging its size upon activation of the channel (Alavian et al., 2014). Additionally, mutations that loosen the packing of the ring and expand its diameter led to significantly increased conductance compared to WT (Alavian et al., 2014). These mutations also sensitized cells to death. We also reported that CypD/Ca²⁺-mediated dissociation of ATP synthase F₁ subunits from F_O is required for unmasking of the c-subunit ring and initiation of its channel conductance (Alavian et al., 2014). Further evidence that F₁ becomes uncoupled from F_O and leads to uncoupling of ATP hydrolysis from H⁺ pumping in the presence of high Ca²⁺ was reported recently by another group (Nesci et al., 2017). Interestingly, age-dependent decrease in F₁ subunit level compared to Fo, reducing the F₁/F_O ratio, have been reported in rat brain and heart mitochondria (Guerrieri et al., 1992a; Guerrieri et al., 1992b).

However, the important role of the c-subunit ring in mPTP formation has been questioned in a recent study (He et al., 2017). The ATP synthase c-subunit knockout HAP1-A12 cells were shown to have the same sensitivity of the mPT to calcium as wild-type (WT) cells in calcium retention capacity (CRC) experiments (He et al., 2017). This allowed the authors to rule out the possibility that ATP synthase c-subunit ring could form the channel of mPT (He et al., 2017). The CRC experiment, however, may only indicate a loss of membrane potential; it is not a direct measure of a channel activity, nor does it assess mPT-induced swelling (He et al., 2017; Mansson et al., 2007; Morota et al., 2013). Patch-clamp recordings, which are a direct measure of channel activity revealed that these same c-subunit knockout mitoplasts (mitochondria lacking the outer membrane) from HAP1-A12 cells lack the high conductance CsA-sensitive mPTP channel activity recorded in WT cells (Neginskaya et al., 2019), indicating the prominent role of c-subunit ring in forming the mPTP channel.

We have discussed all the recent concepts and controversies regarding ACLC, as well as proposed new models describing the physiological gating of ACLC and its pathological openings in our recent publication (Mnatsakanyan and Jonas, 2020) and, partially, in the later sections of this review.

B. Role of ATP synthase in mitochondrial homeostasis and cristae remodeling

More than a century ago it was described that organelles later understood to be mitochondria undergo shape changes rapidly and in response to signals (Chan, 2012). Mitochondrial structural alterations occur during cell stress and death but are also relevant to normal metabolism, cell growth, differentiation, aging and senescence. Mitochondrial quality control is maintained by the mitophagy of defective mitochondria and by the dynamic processes of mitochondrial fusion (the joining of two or more mitochondria) and fission (division), identifying these processes as important targets in neuroprotection.

Mitochondrial fission was found to be carried out by a family of proteins first identified for their role in endocytosis, the dynamin super-family (De Camilli et al., 1995). This family is comprised proteins that hydrolyze GTP to fuse or separate membranes. Drp1 was first described in yeast (DNM1, encoding the protein Dnm1p) (Gammie et al., 1995) and subsequently found to regulate mitochondrial shape and distribution (Bleazard et al., 1999; Otsuga et al., 1998) (Baker et al., 2019; Chan, 2006, 2012; Flippo and Strack, 2017; Smirnova et al., 2001). In neurons, Drp1 regulates Ca²⁺ dependent shape change and movement arrest necessary for synaptic signaling (Hollenbeck, 2005; Rintoul et al., 2003) and for cell death. It is required for normal brain and synapse development and for normal synaptic growth and plasticity (Ishihara et al., 2009). (Berman et al., 2009a; Guo et al., 2005; Li et al., 2008; Verstreken et al., 2005).

Mitochondrial fusion is brought about by several proteins, but key are the mitofusins (1 and 2) and optic atrophy protein 1 (OPA1) (Giacomello et al., 2020). Mitofusin 1 (Mfn1) interacts with OPA1 to bring about fusion on the outer mitochondrial membrane. Mfn2 also tethers mitochondria to other organelles (de Brito and Scorrano, 2008); mutations in Mfn cause Charcot-Marie-Tooth disease, a sensory motor neuropathy (Giacomello et al., 2020). OPA1 loss or mutation leads to genetic diseases including optic nerve atrophy (Giacomello et al., 2020). Upregulation of OPA1 in mice protects against muscular atrophy, ischemic damage in the heart and brain and hepatocyte death, since OPA-1 prevents the release of pro-apoptotic signals (Civiletto et al., 2015; Varanita et al., 2015).

Morphological changes in isolated mitochondria occur during metabolic changes in in vitro experiments. Healthy mitochondria switch between an “orthodox” configuration, in the absence of substrate or ADP, to a “condensed” configuration when performing electron transport and producing ATP (Hackenbrock, 1966). In the orthodox configuration, the matrix is electron light and the cristae appear small and disconnected from each other or the inner boundary membrane. The inner boundary membrane also appears tightly juxtaposed to the outer membrane (Fig. 1). In the condensed configuration, the matrix takes up less of the space of the outer limiting membrane even though the overall volume of the mitochondria is not changed. The matrix becomes exceedingly dense while the cristae spaces enlarge and appear interconnected and connected to an enlarged space between the outer membrane and the inner boundary membrane (Fig. 1). Often one or more enlarged cristae spaces appear within the center of the mitochondria. These changes occur within seconds of the supply of substrate and ADP to mitochondria (Hackenbrock, 1966) and are truly a remarkable example of extreme metabolic plasticity linked to structural changes of mitochondria (Fig. 1).

Figure 1.

The interplay between metabolic plasticity and structural reorganization of mitochondria. Mitochondria switch between an A. orthodox configuration when they have no substrate or ADP, to a B. condensed folded arrangement of cristae while they perform electron transport and produce ATP. These changes occur within seconds of the supply of substrate and ADP to mitochondria (Hackenbrock, 1966). The orthodox configuration of cristae is characterized by the electron light matrix; the cristae appear small and disconnected from each other, and from the inner boundary membrane. In the condensed configuration, the matrix becomes exceedingly dense while the cristae spaces enlarge and appear interconnected and connected to an enlarged space between the outer membrane and the inner boundary membrane. Cristae, C, limiting membrane, LM; inner membrane, IM; outer compartment OC, intracristal space, IS. x 50,000. (Reproduced from Hackenbrock, 1966, by permission of Journal of Cell Biology).

The observed rapid structural changes linked to ATP production raise questions about the underlying molecular mechanisms. Cristae remodeling requires forces driven by the dynamin-like proteins and adapter proteins including OPA-1 and ATP synthase. OPA-1 is a master regulator of the mitochondrial contact site and cristae organizing system (MICOS), in that it participates in opening of the cristae junction (Glytsou et al.,2016) (Fig. 2). MICOS is composed of MIC10, MIC12, MIC19 MIC25, MIC26, MIC27 (also known as MOMA1 and APOOL) and MIC60 (also known as mitofilin); these are subdivided into two independent protein assemblies, MIC27–MIC10–MIC12 and MIC60–MIC19, linked via MIC19 (Friedman et al., 2015). MIC60 is an essential component of the complex; its depletion leads to loss of normal cristae morphology and disconnection of the cristae from the inner boundary membrane. MIC60 KO cells also have abnormalities in oxidative phosphorylation and mitochondrial RNA translation (John et al., 2005; Rabl et al., 2009; Zerbes et al., 2012), implicating MIC60 in organization of electron transport and ATP production. Overexpression of OPA-1 tightens the cristae junction and removal of OPA-1 dissolves it. The inner membrane-anchored form of OPA1 (L-OPA1, long-OPA1) can be cleaved by proteases, generating transmembrane-free, short form of OPA1 (S-OPA1) (Ishihara et al., 2006). Many enzymes participate in L-OPA-1 cleavage, including OMA-1, which may create an imbalance between long and short forms favoring mitochondrial fission and causing mitochondrial and cellular injury (Baker et al., 2014). The S-OPA1 was thought to be incompetent in mitochondrial fusion and energetic maintenance for a long time, however, a recent study suggested that S-OPA1 maintains energetic activity (Anand et al., 2014; Lee et al., 2017). Cells lacking OPA1 not only show highly fragmented mitochondria, but also widened and dysfunctional mitochondrial cristae, and an increased susceptibility to cell death (Frezza et al., 2006; Olichon et al., 2003). Mild overexpression of OPA1 enhances mitochondrial supercomplex formation and respiration efficiency by increasing cristae tightness (Cogliati et al., 2013). In contrast, OPA1-KO cells show no or very few cristae and are unable to support the maturation of the ATP synthase: a significant amount of F₁ can be detected without assembly into the mature ATP synthase complex (Lee et al., 2017).

Figure 2.

Cristae architecture and interplay between OPA1, MICOS-complex and the F₁F_O ATP synthase. Schematic representation of OPA1, MICOS subunits and ATP synthase in the mitochondrial inner membrane: Mic10 interacts with ATP synthase and its e-subunit, while Mic27 was shown to be an important bridge mediating the interaction between Mic10/ATP synthase and the remaining MICOS complex, Mic60-Mic19-Mic26-Mic12. The cristae shape is formed by OPA-1 and MICOS interactions at the cristae junction and ATP synthase structural organization at the cristae tip, so the complexes work together, at opposite ends of the cristae to form its normal architecture. MICOS and the ATP synthase together with cardiolipin (CL, red circles) and other membrane lipids support the normal formation and plasticity of the cristae structure. CL interacts with MICOS, OPA-1, ATP synthase and with the other members of the electron transport chain. The 3D model of mitochondria from Microsoft Paint 3D was used.

ATP synthase is crucial for cristae modeling (Paumard et al., 2002a; Paumard et al., 2002b; Quintana-Cabrera et al., 2018). Dimerization of the synthase and the further organization of dimers into rows forms the shape and optimal function of the cristae tips (Davies et al., 2012; Davies et al., 2011). Loss of ATP synthase dimers through depletion of membrane embedded F_O subunits e and g, that are implicated in assisting in oligomerization, causes a flattening of the cristae tips opposite to the cristae junction with the inner boundary membrane, ballooning the cristae (Davies et al., 2012). OPA1 enhances the stability of ATP synthase oligomeric structures when the mitochondria are under stress (Quintana-Cabrera et al., 2018). Studies show that ATP synthase interacts directly with MICOS component MIC10 (Rampelt et al., 2017), implicating either a different location for at least part of the ATP synthase or high mobility of ATP synthase monomers in the cristae membrane. Chemical crosslink of Mic10 with ATP synthase subunit e and Mic10 to Mic27 in yeast mitochondria has been observed (Eydt et al., 2017) . Therefore, Mic27 was shown to be an important bridge mediating the interaction between Mic10/ATP synthase and the remaining MICOS complex, Mic60-Mic19-Mic26-Mic12 (Eydt et al., 2017) (Fig. 2).

MIC60/MIC19 form a different subcomplex from Mic27/Mic10/Mic12. This latter complex is selectively dependent on the presence of mtDNA, suggesting a critical role of respiratory complexes as a whole, but not any specific respiratory complex, in the assembly and/or stabilization of this MICOS subcomplex (Friedman et al., 2015). In yeast, Fcj1, the homologue of MIC60, assists in decreasing higher order oligomers of F₁F_O ATP synthase, allowing the formation of the appropriate dimerization, but subunits e and g of the ATP synthase are required for normal cristae tip formation (Rabl et al., 2009). Therefore, the cristae shape is likely formed by OPA-1 and MICOS interactions at the cristae junction and ATP synthase structural organization at the cristae tip or edge, so the complexes work together but at opposite ends of the cristae to form its normal architecture (Rabl et al., 2009). The data suggest that MICOS and the ATP synthase together with cardiolipin and other membrane lipids support the normal formation and plasticity of the cristae structure (Friedman et al., 2015; Wollweber et al., 2017) (Fig. 2).

Cardiolipin (CL) is a major lipid membrane component of the acute angle at the cristae junctions and at the ATP synthase edge. CL is necessary for interactions between inner and outer membranes. The mitochondrial fusion and fission, as well as protein transport are all dependent on CL. CL interacts with MICOS, OPA-1, ATP synthase and with the other members of the electron transport chain. CL is highly acidic. It has a glycerol head group that is esterified to two phosphatidylglyceride backbone fragments, combined with four acyl chains. This gives cardiolipin its dimeric shape, resulting in a conical structure that creates the acute bends (Paradies et al., 2014). CL is almost exclusively associated with membranes designed to produce ATP through electron transport activities (mitochondrial and bacterial membranes); it affects the efficiency and stability of respiratory enzyme supercomplexes (Paradies et al., 2014). Most of the major respiratory enzymes, metabolite carriers and ion transporters of the inner membrane are dependent on CL. CL also contributes to activity of the metabolite transporter proteins by providing a binding site to clamp the transporters in position. CL is critical for oligomerization of ATP synthase multimeric complexes either by direct interaction with ATP synthase or by producing the proper membrane curvature. CL and ATP synthase interaction may reduce the free energy of the extreme membrane curvature to stabilize high-curvature folds (Acehan et al., 2011). These properties affect cristae morphology and increase energy use and production efficiency. Cardiolipin-bound cytochrome c (cyt c) acts to perform H₂O₂-dependent CL peroxidation, essential for the release of cyt c during apoptosis. Alterations in CL structure or amount are associated with mitochondrial dysfunction in hypo and hyperthyroid states (Paradies and Ruggiero, 1990; Paradies et al., 1991), heart ischemia–reperfusion (Paradies et al., 1999), nonalcoholic fatty liver disease (Petrosillo et al., 2007), diabetes (Han et al., 2005), Barth syndrome (Schlame and Ren, 2006) and aging (Paradies et al., 2011; Semba et al., 2019).

CL likely assists in the formation of supercomplexes of the respiratory chain enzymes (Acin-Perez et al., 2008; Wittig et al., 2006). Evidence suggests that CL improves electron transfer efficiency by tightening the interaction of the supramolecular assemblies, including improving the efficiency of oxygen consumption and metabolite flux (Acin-Perez et al., 2008; Genova and Lenaz, 2013, 2014), enhancing efficiency of electron/proton exchange by as much as 35% (Claypool, 2009). These assemblies also reduce ROS production and increase the capacity for ATP production. CL has been suggested to use the phosphate head groups as a proton trap, restricting movement of pumped protons, the buffering of which may provide a structural basis for mitochondrial membrane potential (ΔΨ) by increasing the voltage component of the mitochondrial membrane potential. This also may provide an avenue for supply of protons to the ATP synthase (Haines and Dencher, 2002). CL could locally contribute to driving directional ATP export and ADP import through the ADP/ATP carrier (Klingenberg, 2009).

The assembly and disassembly of the complexes is dynamic and subject to change depending on physiological and pathological conditions (Beutner et al., 2014; Pfeiffer et al., 2003; Zhang et al., 2002). Therefore fusion, fission and structural changes to the cristae are all highly dependent on CL.

C. Role of mPTP in cristae remodeling during apoptosis

Understanding cristae remodeling during apoptosis has helped to gain an understanding of the normal remodeling that accompanies physiological alterations in metabolism. Mitochondrially mediated apoptosis is a well-orchestrated event beginning with translocation of tBID’s (truncated BID, BH-3 only Bcl-2 family member) to the mitochondrial outer membrane. tBID’s activate Bax or Bak oligomerization to release cyt c from mitochondria through a Bax/Bak dependent channel. Cyt c activates the apoptosome including caspase 9 and APAF-1, which then activate downstream caspases initiating the breakdown of cellular and nuclear components (Adams and Cory, 2007; Hardwick and Youle, 2009). The release of almost 100% of mitochondrial cyt c occurs during these events, raising the question of how the intracristal cyt c gets out without obvious matrix swelling, since only about 16% of cyt c is available for release from the intermembrane space (Scorrano et al., 2002). Later it was shown that a reorganization of the cristae occurs to release the majority of cyt c (Giacomello et al., 2020; Scorrano et al., 2002). The original studies showed that tBID-induced apoptotic signaling induced partial mitochondrial membrane depolarization and release of calcein (a matrix-entrapped fluorescent dye). Both processes were inhibited by CsA but did not cause appreciable swelling, indicating reversible activation of the mPTP during apoptotic cyt c release. CsA in the presence of tBID still resulted in 17% cyt c release, suggesting that CsA could not inhibit the release of cyt c from the intermembrane space but could block its release from the intracristal compartments. Therefore, the major portion of cyt c is contained in the intracristal space and this cyt c is not available for release unless mPTP opens. After loosening its anchor from the intracristal location, another action is required to release cyt c from the outer membrane and this proved to be Bax/Bak oligomerization since the CsA sensitive membrane depolarization and calcein release still occurred in Bax/Bak KO cells but cyt c release was abrogated (Scorrano et al., 2002). These studies confirmed that the release of cyt c during apoptosis is made up of two events: A cyt c mobilization event dependent on tBID and mPTP and a separate releasing event dependent on tBID activation of Bax/Bak in the outer membrane, since Bax/Bak deficient mitochondria were able to redistribute, but not release, cyt c.

The cyt c releasing event involves a major reorganization of the cristae, as noted in electron micrographs (Mannella, 2006). Especially important is that the connection to the intermembrane space, the crista junction, widens during apoptosis to about 5x its normal width, indicating that this is the size most likely needed for cyt c release (Mannella, 2006).

Studies of the ER-associated BH3 only protein BIK (BCL-2 interacting killer) indicated that BIK, similarly to tBID, activates cristae remodeling. These rapid changes in cristae morphology are dependent on BIK’s BH3 domain, Ca²⁺ uptake by mitochondria, the mPTP, and functional Drp1 (Germain et al., 2005). The order of events is that calcium enters the mitochondria from the ER through the MCU, it activates the mPTP and this induces cristae membrane remodeling at the same time that Ca²⁺ localizes Drp1 to the outer membrane and activates calcineurin to activate Drp1 GTPase activity (Germain et al., 2005). The mechanism of how calcium-dependent Drp1 activity causes cristae membrane remodeling is not understood but may involve Drp1 interactions with inner membrane proteins at the adjacent cristae junction.

During cell death Drp1 dependent remodeling requires high calcium. In contrast, the cristae remodeling that occurs during normal ATP production does not require calcium but only substrate and ADP (Hackenbrock, 1966). New studies using super resolution imaging of fluorescently labeled cristae indicate that rapid physiological changes in cristae morphology occur almost constantly even in resting cells (Kondadi et al., 2020). Therefore, rapid changes in structure brought on by normal variation in daily physiological metabolic events are the norm. The key feature observed in the early studies was cristae enlargement, albeit in the presence of abnormally high ADP/ATP ratio, probably causing a very rapid shift to ADP dependent oxidative phosphorylation (state III) (Hackenbrock, 1966). By analogy with apoptosis and cyt c release, ATP and other metabolite release through the cristae junction would be the most likely reason for the drastic remodeling of the cristae and matrix compartment during normal state III respiration. The large change in volume of the intracristal space occurring during these events is likely to be brought on by osmotic changes, particularly K⁺ influx into the crista compartment (Kaasik et al., 2007). The ions (and water) may derive from two sources: The first could be an enlarged connection formed at the crista junction with the inner boundary membrane. The second is uptake of K⁺ from the matrix, since the matrix rapidly transforms into an extremely electron dense compartment and shares the same volume, fixed by the outer membrane, with the enlarged cristae during this time. MICOS and OPA-1 must participate in the rapid changes in fusion with the inner boundary membrane but the mechanisms of K⁺ flux from the matrix at this time are completely unknown. These could involve the mPTP operating in its physiological mode, particularly as it was observed that the tBID induced, CsA sensitive changes that cause cyt c release appear not to require changes in total mitochondrial volume (Scorrano et al., 2002).

ATP synthase as a key target in mitochondria-mediated neurodegeneration: Reduced F₁/F_O ratio in neurodegenerative disease

Recent studies indicate that mitochondrial changes are a driving force, rather than a consequence, of the aging process and neurodegeneration. Abnormal accumulation of amyloid-β protein (Aβ) was found not only in the mitochondria of transgenic mouse models of AD that overexpress Aβ (Dragicevic et al., 2010; Manczak et al., 2006) but also was reported to interact directly with ATP synthase subunits and cause detrimental changes to its structure and function (Beck et al., 2016). Alzheimer’s Disease (AD) is one of the most commonly occurring neurodegenerative diseases. AD pathology causes synaptic dysfunction followed by the death of brain cells, leading to significant tissue shrinkage affecting memory, attention, speech and behavior (Cadonic et al., 2015; Selkoe, 2011). High levels of reactive oxygen species produced in mitochondria cause oxidative damage and redox imbalance, impairing mitochondrial function (Beal, 2005; Trushina and McMurray, 2007). Such changes were observed in human pathological samples of early AD (Cenini et al., 2019; Du et al., 2012; Du and Yan, 2010a, b; Nunomura et al., 2001). Interestingly, the α subunit of the mitochondrial ATP-synthase was reported as the most common lipoxidized protein in the entorhinal cortex of all AD cases at stages I/II of the disease (Terni et al., 2010). Additionally, it was reported that ATP synthase α subunit accumulates in the cytosol either alone or associated with tau aggregates of neurofibrillary tangles in degenerating neurons of AD (Sergeant et al., 2003). In the brains of AD individuals and in an AD mouse model, a decrease in ATP synthase OSCP subunit level was observed (Beck et al., 2016) and OSCP was shown to be a binding site for the aggregated protein Aβ (Beck et al., 2016). Decreases in OSCP levels were pronounced especially in neuronal mitochondria (Beck et al., 2016). Furthermore, increased F₁ dissociation from F_O and significantly high free F₁ levels were reported in neuronal mitochondria in the AD mouse (Beck et al., 2016). The loss of OSCP and reduced F₁ levels led to reduced ATP production, elevated oxidative stress and activated mPT and these abnormalities were alleviated in the mouse by OSCP overexpression (Beck et al., 2016).

Similarly, we have now found that loss of the ATP synthase F₁ β subunit and reduction in F₁/F_O ratio occurs in a DJ-1 deficient, Parkinson’s Disease (PD) mouse model (Chen et al., 2019). DJ-1 is a molecular chaperone and peptidase C56 family protein with known and uncharacterized cellular functions (Bonifati et al., 2003). Familial PD protein DJ-1 mutations are linked to early onset PD. DJ-1 mutant animals show increased sensitivity to neuronal toxins, and in different species DJ-1 is required for normal life span, motor function, and neuronal resistance to oxidative damage (Biosa et al., 2017; Flao et al., 2010; Malgieri and Eliezer, 2008). DJ-1 translocates to mitochondria from the cytosol in response to mitochondrial stress (Canet-Aviles et al., 2004; Junn et al., 2009; Weinert et al., 2019) and defects in DJ-1 alter mitochondrial morphology and function (Larsen et al., 2011). We have found that DJ-1 binds directly to the F₁F_O ATP synthase β subunit and this interaction decreases mitochondrial uncoupling and enhances ATP production efficiency (Chen et al., 2019). In contrast, DJ-1 deficient mitochondria demonstrate a leaky mitochondrial inner membrane and the leak is produced by the ATP synthase (Chen et al., 2019; Hao et al., 2010). The abnormalities of the ATP synthase in DJ-1^−/− animals are associated with reduced enzymatic function, reduced ATP levels and impaired neurite extension in isolated dopaminergic neurons, suggesting that normal ATP synthase function is critical for dopaminergic neuron growth (Chen et al., 2019). Levels of the ATP synthase β subunit are low in aged DJ-1^−/− mouse brain and in patient-derived cell lines even though levels of the c-subunit are normal, suggesting that mutant DJ-1-induced degenerative disease is associated with a reduction in F₁/F_O ratio with resultant ACLC formation.

α-synuclein is the most abundant protein found in Lewy bodies, the pathological hallmark of Parkinson’s disease. Under physiological conditions monomeric α-synuclein interacts with ATP synthase α subunit and enhances ATP synthase activity and mitochondrial function (Ludtmann et al., 2016). However, α-synuclein aggregation generates β sheet-rich oligomers that localize to the mitochondria in close proximity to several mitochondrial proteins including complex I and ATP synthase. These oligomers interact with the ATP synthase β subunit, induce its selective oxidation and mitochondrial lipid peroxidation, and increase the probability of mPTP opening, triggering mitochondrial swelling and cell death (Ludtmann et al., 2018). Therefore, transition of α-synuclein from its monomeric to oligomeric state switches its role from physiological to pathological leading to mitochondrial dysfunction in Parkinson’s disease through a direct interaction of pathological α-synuclein oligomers with the ATP synthase. This highlights the role of ATP synthase in PD pathology.

ATP synthase as a key target in mitochondria-mediated neuroprotection

A. ATP synthase leak channel as a regulator of synaptic plasticity and metabolic efficiency in neurons

We have reported that overexpression of the anti-apoptotic protein Bcl-xL in neurons markedly increases ATP production, while lowering oxygen consumption (Alavian et al., 2011). This contradiction was resolved when we found that Bcl-xL is bound directly to the ATP synthase β subunit (Alavian et al., 2011; Chen et al., 2011), where it increases ATP synthase activity rate and reduces inner mitochondrial membrane leak, enhancing the coupling and efficiency of oxidative phosphorylation. Bcl-xL inhibitors ABT-737 (Alavian et al., 2011) and WEHI-539, induce ATP synthase c-subunit related inner membrane leak formation by inhibiting the binding of Bcl-xL to the ATP synthase β subunit. Bcl-xL depletion causes energy loss and instability of the mitochondrial membrane potential (Chen et al., 2011). Bcl-xL KO causes embryonic lethality just before E12.5 (Motoyama et al., 1995), but conditional KO mice lacking Bcl-xL in neuronal progenitors survive, albeit with a highly malformed cortex and severe loss of neurite growth and synaptogenesis (Chen et al., 2011).

By increasing the metabolic efficiency of neurons Bcl-xL contributes to enhancing synaptic growth and plasticity through regulating mitochondrial biogenesis. Bcl-xL increases the numbers of presynaptic vesicle clusters (Li et al., 2008), synapses and mitochondria. Bcl-xL coordinates with the mitochondrial fission protein, Drp-1 and the biogenesis factor PGC-1 alpha to achieve mitochondrial positioning, followed by mitochondrial fission to supply the enlarged synapses with newly minted mitochondria (Berman et al., 2009a, b). Bcl-xL and Drp-1 work together directly on synaptic vesicles to bring about activity dependent vesicle pool refilling within the first few minutes after high frequency action potential firing (Li et al., 2013). Despite its important anti-apoptotic functions, Bcl-xL can be cleaved to ΔN-Bcl-xL, an N-terminal, pro-apoptotic product during ischemia or excitotoxicity. The role of Bcl-xL in modulating the ACLC activity during excitotoxic conditions will be discussed in the next section.

B. Modulation of ATP synthase leak channel activity in ischemia-triggered excitotoxic conditions

Oxygen and nutrient deprivation during acute stroke causes energy crisis in neurons and cell death by necrotic mechanisms (Moskowitz et al., 2010). Depolarization of neuronal plasma membranes and failure of active transporters result in accumulation of high glutamate levels in synaptic terminals (Wang et al., 2003), creating glutamate excitotoxicity, an important cause of cell death during ischemia. Subsequent transport of calcium into the intracellular compartment and the mitochondria, results in opening of the mPTP, mitochondrial swelling, rupture of the mitochondrial outer membrane and oxidant stress (Arundine and Tymianski, 2004; Nicholls, 2004, 2008). The opening of mPTP is less dramatic and possibly reversible in the area of brain tissue that has partial blood supply, the penumbra (Astrup et al., 1981; Moskowitz et al., 2010), while it can be acute in other cases. It is during this period in the peri-infarct territory that programmed cell death mechanisms are activated. These mechanisms follow a path determined by genetically encoded molecular players carrying out predetermined roles through protein/protein and protein/membrane interactions (Adams and Cory, 2007; Broughton et al., 2009; Park and Jonas, 2017).

The Bcl-2 family proteins are critical in regulating neuronal death in the at-risk penumbra. Pro-death Bcl-2 family members, Bax and Bak, produce outer mitochondrial membrane permeabilization, leading to cyt c release and activation of downstream signaling pathways, including activation of enzymes known as cell-destructive caspases (Adams and Cory, 2007; Hardwick et al., 2012; Hardwick and Youle, 2009). The death-inducing functions of the pro-apoptotic versions of Bcl-2 family proteins are opposed by those of the anti-death proteins such as Bcl-2, Bcl-xL, Mcl-1 and others. Small molecule inhibitors of the Bcl-2 proteins such as ABT-737 and others were developed to bind into a groove in the BH1-3 domains and prevent the interaction of the anti-apoptotic proteins with the pro-apoptotic proteins, releasing Bax and Bak to permeabilize mitochondrial membranes. These powerful and selective reagents are useful for their chemotherapeutic activity (Chipuk et al., 2008; Oltersdorf et al., 2005)

The BH4-domain-containing anti-apoptotic proteins such as Bcl-xL carry out important survival functions in the brain, but during ischemia or excitotoxicity, Bcl-xL is cleaved to ΔN-Bcl-xL, an N-terminal, cleavage product (Fujita et al., 1998; Miyawaki et al., 2008; Ofengeim et al., 2012) and it activates large conductance mitochondrial channel activity (Hickman et al., 2008; Jonas et al., 2004; Ofengeim et al., 2012) similar to that of oligomerized Bax (Jonas et al., 2005a). Blocking formation of ΔN-Bcl-xL or the mitochondrial channel activity of ΔN-Bcl-xL with the small molecule inhibitor ABT-737 rescues neurons from ischemic death (Hickman et al., 2008; Ofengeim et al., 2012; Park and Jonas, 2017; Park et al., 2017). ABT-737 also blocks synaptic depression produced by hypoxia or by injection of recombinant ΔN-Bcl-xL protein into the presynaptic terminal (Hickman et al., 2008; Jonas et al., 2004; Jonas et al., 2005b). A mouse expressing a cleavage resistant form of Bcl-xL is protected from ischemic brain damage (Ofengeim et al., 2012).

In addition to its known outer mitochondrial membrane functions (to release cyt c and activate apoptosis), (Ofengeim et al., 2012) ΔN-Bcl-xL, similarly to full-length Bcl-xL, functions at the mitochondrial inner membrane where it exacerbates the effects of glutamate excitotoxicity on mitochondrial depolarization, cyt c release, caspase activation, loss of ATP production and death (Park et al., 2017). These effects are prevented by low dose ABT-737, which preferentially binds to ΔN-Bcl-xL overfull length Bcl-xL, the latter of which requires higher concentrations to cause mitochondrial dysfunction in neurons (Park et al., 2017). ΔN-Bcl-xL depolarizes the mitochondrial inner membrane most likely by opening the ACLC. It is localized to the mitochondrial inner membrane by immunocytochemistry and by membrane fractionation studies. In neurons challenged with excitotoxicity, ΔN-Bcl-xL overexpression causes cell death prevented by the relative depletion of ATP synthase c-subunit with shRNA (Park et al., 2017). Neuronal dendritic and axonal processes are highly sensitive to metabolic perturbations. Depletion of c-subunit prevents ΔN-Bcl-xL-induced neurite dysfunction and cell death by modulating ACLC activity.

C. ATP synthase leak channel as a target in ischemia/reperfusion injury

One of the leading causes of death in the United States is ischemia/reperfusion (I/R) resulting from stroke or myocardial infarction (Benjamin et al., 2018). During I/R injury tissue damage and cell death occur first in response to the initial ischemic insult and then later by the subsequent reperfusion. Severe reduction in ATP levels and intracellular pH during prolonged ischemia are associated with hypoxia-induced inhibition of oxidative phosphorylation and lactate accumulation due to enhanced glycolysis. The ATP- dependent ion transport mechanisms (Na⁺/K⁺ ATPase etc.) become dysfunctional under these conditions, causing neurons to depolarize and release excessive amount of glutamate into the synaptic cleft, initiating excitotoxic injury (Chomova and Zitnanova, 2016; Kalogeris et al., 2012; Mehta et al., 2007; Nakka et al., 2008). The glutamate overflow activates post-synaptic NMDA and AMPA receptors leading to intracellular and mitochondrial calcium overload and mitochondrial swelling. The conditions after reperfusion, including elevated levels of reactive oxygen species and phosphate and increased intracellular pH orchestrate the opening of the mPTP, which is a critical factor for inducing both apoptotic and necrotic cell death during I/R injury (Halestrap et al., 2004; Kalogeris et al., 2012) (Fig. 3). Increasing evidence suggests that mPTP is a promising target for neuro and cardioprotection during I/R injury.

Figure 3.

ATP depletion, elevated pH, phosphate, and ROS levels after ischemia/reperfusion injury trigger the opening of the mPTP and induce cell death. The opening of ACLC occurs upon non-reversible dissociation of F₁ subcomplex from F_O (Alavian et al., 2014), due to the binding of Ca²⁺ to ATP synthase β subunits, CypD to OSCP and upon removal of lipid or other plugs from the c-subunit lumen. The ischemic preconditioning, stimulated by the inhibitors of ATP synthase and of other mitochondrial electron transport chain complexes prevent the opening of ACLC and thus increase the brain’s tolerance against ischemic cell death, highlighting the role of mPTP as a promising target for neuro and cardioprotection during I/R injury. ATP synthase subunits are drawn as ribbon representations (modified PDB ID code: 6J5I) (Gu et al., 2019).

Ischemic preconditioning is found to be a promising therapy for increasing the brain’s tolerance against ischemic cell death. A short, non-damaging ischemic episode, preceding sustained ischemia, stimulated by different therapeutic conditions or pharmacological agents proves useful in different animal models and in in vitro cell cultures (Kitagawa et al., 1991; Murry et al., 1986; Schurr et al., 1986). However, the underlying molecular mechanisms mediating ischemic preconditioning protection are not fully understood.

One of the protective mechanisms was reported to be related to ATP synthase inhibition during preconditioning (Vuorinen et al., 1995). A brief global ischemic episode used for preconditioning of the rat heart myocardium caused ATP synthase inhibition that persisted throughout the initial reperfusion period, resulting in a better final recovery of the preconditioned hearts (Vuorinen et al., 1995). Under normoxic conditions mitochondrial ATP synthase drives ATP synthesis. However, as the H⁺ gradient drops under hypoxic conditions the ATP synthase begins to run in reverse, contributing to mitochondrial membrane potential generation by ATP hydrolysis; this process depletes cellular ATP content and activates mPTP opening (Galkin et al., 2009). The intrinsic inhibitor protein IF1 binds to ATP synthase at low pH (under ischemic conditions) and prevents the reverse reaction of ATP hydrolysis and the ATP wastage, thus IF1 is reported to be protective against ischemic injury (Campanella et al., 2008). Meanwhile, increased pH level (under reperfusion conditions) impairs IF1 interaction with ATP synthase allowing ATP hydrolysis to occur freely (Cabezon et al., 2000a; Cabezon et al., 2000b; Campanella et al., 2008). Interestingly, lack of IF1 induces a non-thyroidal hypermetabolic syndrome in humans known as Luft’s disease (DiMauro et al., 1976; Yamada and Huzel, 1992). It is a rare mitochondrial disease, characterized by altered mitochondrial morphology (abnormally large mitochondria with packed cristae) and uncoupled mitochondrial respiration with high basal ATP hydrolysis activity. Furthermore, the rate of energy-dependent calcium uptake by isolated mitochondria is normal, but the amount and retention of accumulated calcium is much reduced suggesting that mPTP opening participates in the disease phenotype. These findings highlight the key role of ATP synthase and IF1 in mitochondrial physiology and in cell survival during ischemia.

Inhibitors of ATP synthase and of other mitochondrial electron transport chain complexes have been reported to cause preconditioning-like effects (Arbelaez-Quintero and Palacios, 2017; Della-Morte et al., 2009; Vuorinen et al., 1995) (Fig. 3). Oligomycin, which inhibits ATP synthase through its Fo domain, preserves mitochondrial ATP content and protects against or postpones cell death during ischemia. Although a 5-minute preischemic infusion of oligomycin caused initial decline in the cellular energy state due to inhibition of ATP synthesis, it significantly reduced further decreases in ATP levels during the subsequent ischemic period (Vuorinen et al., 1995). Consequently, preservation of mitochondrial ATP allows neurons to survive the long-lasting hypoxia and energy deprivation that occur during an ischemic episode. Although mPTP channel activity was not investigated in this study, we suggest that the preconditioning period of ischemia, as well as oligomycin-induced ATP synthase inhibition can lead to the closure of the mPTP channel to increase the bioenergetic efficiency of neurons (Fig. 3). This is in agreement with our recent observation of oligomycin inhibition/closure of the ACLC activity in in vitro patch-clamp recordings of porcine ATP synthase reconstituted liposomes (Mnatsakanyan et al., 2019).

In recent studies, mild hypoxic conditions in general were reported to protect cells and animal models from defects in mitochondrial electron transport chain (Ferrari et al., 2017; Jain et al., 2019; Jain et al., 2016). In these studies, oxygen deprivation was shown to activate the hypoxic response, an evolutionarily conserved adaptive mechanism allowing mammals to survive under limited oxygen conditions by switching their metabolism from oxidative phosphorylation to glycolysis (Jain et al., 2016). Neurons prefer to use mitochondrial oxidative phosphorylation over glycolysis to meet the high energy needs of the brain. Due to such reliance on oxidative phosphorylation mitochondria are critical to neuronal activities and mitochondrial dysfunction becomes a main cause of neurodegeneration. A genome-wide Cas9-mediated screening of ~18,000 genes for identification of protective factors during mitochondrial disease conditions revealed the importance of the Von Hippel-Lindau (VHL) factor, a key regulator of the hypoxic response pathway, as the most effective genetic suppressor. Activation of the hypoxia response by either genetic or pharmacological interventions protects cultured cells or zebrafish models of mitochondrial dysfunction (Jain et al., 2016). Additionally, chronic hypoxia (11% O₂) improves the disease phenotype in the Ndufs4 KO mouse model of the pediatric mitochondrial disease Leigh syndrome, the most common mitochondrial disease in children (Jain et al., 2016). The Ndufs4 gene encodes subunit 4 of complex I, the deficiency of which is the most frequent biochemical cause of Leigh syndrome. Chronic hypoxia is also reported not only to prevent but to reverse brain lesions in mice with advanced neuropathology in a mouse model of Leigh syndrome (Ferrari et al., 2017; Jain et al., 2019).

Another compound, which acts as an electron transport chain inhibitor with beneficial effects in I/R injury and in Alzheimer’s disease is metformin, a common drug used to treat type 2 diabetes (Arbelaez-Quintero and Palacios, 2017; Yin et al., 2011). Metformin is a mild inhibitor of complex I. It inhibits respiration leading to activation of AMP-activated protein kinase (AMPK) and resulting in beneficial metabolic effects (Adak et al., 2018). Furthermore, metformin and other biguanides may inhibit ATP synthase activity (Bridges et al., 2014). However, their direct interaction sites with ATP synthase and the inhibitory mechanisms are currently unknown.

One of the most promising pharmacological agents used in preconditioning is resveratrol, a natural polyphenol and antibiotic found in grapes, berries and their products, such as wines (Della-Morte et al., 2009; Khoury et al., 2016; Khoury et al., 2019; Raval et al., 2006; Salehi et al., 2018). Resveratrol was shown to have neuroprotective effects in cerebral ischemia and other models of neurological disease, such as Parkinson’s, Alzheimer’s and Huntington’s (Kursvietiene et al., 2016; Salehi et al., 2018). The neuroprotective effect of resveratrol was first reported to be through the activation of Sirtuin 1 (SIRT1), which is an NAD-dependent deacetylase that removes acetyl groups from various proteins contributing to cellular regulation (Ghosh, 2008; Howitz et al., 2003; Lagouge et al., 2006). SIRT1 activates the PGC1-alpha and ERR-alpha transcription factors, which are important metabolic regulators playing a central role in transcriptional control of mitochondrial biogenesis and respiratory function (Lagouge et al., 2006; Scarpulla, 2011). SIRT1 also inhibits NF-κB-regulated inflammation related gene expression and thus extends lifespan and delays aging in mice (Lawrence, 2009). Other studies suggest that the effect of resveratrol on SIRT1 is indirect in that it may be due to the inhibition of mitochondrial respiration, which increases the AMP/ATP ratio and thus activates AMPK, the upstream activator of SIRT1 (Kulkarni and Canto, 2015).

Resveratrol preconditioning was reported to significantly reduce uncoupling protein 2 (UCP2) levels in rat brain by 35% and increase the ATP synthesis ratio per oxygen molecule (ADP/O ratio) in hippocampal mitochondria suggesting that resveratrol preconditioning increases overall metabolic and bioenergetic efficiency of mitochondria (Della-Morte et al., 2009).

In further studies, a long window of cerebral ischemic tolerance that lasts for two-weeks, mediated by resveratrol preconditioning, was reported in mice (Khoury et al., 2019). RNA-seq experiments of the preconditioned mouse cortex revealed downregulation of more than 100 genes involved in transcription, synaptic signaling, and neurotransmission. Meanwhile, the overexpression of genes involved in pyruvate uptake, TCA cycle, and oxidative phosphorylation was observed, along with increased ATP levels and mitochondrial abundance. However, both basal and maximal OCR were significantly reduced, suggesting reduced inner membrane leak and more efficient respiration (Khoury et al., 2019). This could be explained partially by the overexpression of the anti-apoptotic protein Bcl-2 and decreased expression of the pro-apoptotic proteins Bax, caspase-3 and cleaved caspase-3 observed during resveratrol preconditioning, overall resulting in the attenuation of ischemia-induced cell death (Liu et al., 2016). We have discussed the role of Bcl-xL in closing the ACLC and increasing the bioenergetic efficiency of neurons in the previous sections. However, other regulators of the ACLC in resveratrol preconditioning are yet to be investigated.

Resveratrol and other related polyphenols, quercetin and piceatannol bind directly to mitochondrial ATP synthase (Gledhill et al., 2007). Resveratrol interacts with specific residues of α, β and γ subunits by binding to the hydrophobic pocket between the C-terminal tip of the γ subunit and the β_TP subunit (Gledhill et al., 2007) (Fig. 4). Resveratrol binding prevents both ATP synthesis and hydrolysis activities by blocking the rotation of γ subunit in mammalian ATP synthase (Gledhill et al., 2007). Investigation of resveratrol effects on regulation of ACLC activity will help us understand if neuroprotective effects of resveratrol can also be explained by its direct interaction with the mitochondrial ATP synthase.

Figure 4.

The structure of bovine heart mitochondrial F₁-ATPase in 2.3-Å resolution showing the site of binding of resveratrol. PDB ID: 2JIZ (Gledhill et al., 2007). The α, β, γ subunits are blue, green, yellow respectively, and resveratrol is red. A. Side view of F₁-ATPase with resveratrol bound in a pocket between β_TP and γ subunits. For clarity, the other two α and β subunits were removed. B. Top view of F₁-ATPase showing the binding site of resveratrol between the β_TP and γ subunits. C. Chemical structure of resveratrol.

In contrast to its neuroprotective effects, resveratrol displays anti-cancer properties by inducing cell growth arrest or triggering apoptotic cell death signals in tumors. In contrast to neurons, cancer cells have different energetic needs; they exhibit an altered metabolism by switching from oxidative phosphorylation to aerobic glycolysis, known as the Warburg effect. Resveratrol exposure reverses Warburg metabolic phenotype in colon cancer cells, by causing a metabolic shift from glycolysis to oxidative phosphorylation; resveratrol slows down cell proliferation without inducing massive cell death in this study (Saunier et al., 2017). Taken together these studies suggest that resveratrol-mediated mitochondrial changes, its neuroprotective effects in ischemia and neurodegeneration and apoptotic effects in cancer cells depend on the metabolic phenotype of the cell type and more specifically on its bioenergetic profile.

Concluding remarks

F₁F_o ATP synthase has always been considered as a crucial component of the mitochondrial electron transport chain since it synthesizes ATP during oxidative phosphorylation to meet cellular energy needs. The importance of F₁F_o ATP synthase in cell physiology was further highlighted by the recent findings that it houses the cell death channel of mitochondrial permeability transition. In this review we have discussed the new role of ATP synthase beyond oxidative phosphorylation. We have described the physiological function of ACLC in enhancing the metabolic efficiency and synaptic plasticity of neurons, as well as its function in mitochondrial remodeling and shaping cristae morphology. We have highlighted the function of ATP synthase as a unique regulator of cell life and death in ischemia-reperfusion injury and in neurodegenerative disease, such as Alzheimer’s and Parkinson’s. We suggested that the ATP depletion, elevated pH, phosphate and ROS levels after ischemia/reperfusion injury trigger the opening of ACLC upon non-reversible dissociation of F₁ subcomplex from F_O, and upon removal of lipid or other plugs from the c-subunit lumen (Fig. 3). We described that the ischemic preconditioning, stimulated by the inhibitors of ATP synthase and of other mitochondrial electron transport chain complexes prevent the opening of ACLC and thus increase the brain’s tolerance against ischemic cell death, highlighting the role of ATP synthase as a promising target for neuro and cardioprotection during I/R injury.

Future structural and functional studies of ACLC will make it an irreplaceable target for the treatment of degenerative and metabolic diseases.