The cristae modulator Optic atrophy 1 requires mitochondrial ATP synthase oligomers to safeguard mitochondrial function

1Cristae are a pleomorphic subcompartment of the inner mitochondrial membrane (IMM) essential for biological energy conversion and for regulation of mitochondrial apoptosis. Cristae shape and mitochondrial function are intimately connected: changes in cristae morphology affect stability of respiratory supercomplexes (RCS),, functional quaternary assemblies of respiratory chain complexes (RCC),–. In a reductionist view, cristae shape can be defined by the curvature of two regions: the cristae junctions (CJs), narrow tubular structures connecting cristae to the inner boundary membrane; and the cristae lumen proper. CJs limit the diffusion of cytochrome c from the cristae, and their enlargement (i.e., the transition from the negative curvature of a shape resembling a solid hyperboloid to the null curvature of a cylinder or to the positive curvature of a spheroid) triggered by proapoptotic BH3-only BCL-2 family members BID, BIM-S, or BNIP3 allows cytochrome c redistribution,–. During this process of cristae remodeling, cristae lumen width is also altered: the transition to a more positive CJ curvature results in overall cristae widening. Functionally, cristae widening destabilizes RCS and reduces mitochondrial oxidative phosphorylation efficiency. In sum, cristae shape is a key morphological parameter that influences mitochondrial apoptosis and respiration.

2A central modulator of cristae curvature is the IMM dynamin-related protein Optic atrophy 1 (OPA1): OPA1 oligomers maintain a negative CJ curvature, controlling cytochrome c redistribution and release, and stabilizing RCS to increase respiratory efficiency,,. Because of these pleiotropic effects on mitochondrial function, controlled OPA1 overexpression is beneficial against a variety of pathological conditions, ranging from ischemia-reperfusion to massive hepatocellular apoptosis to muscular atrophy, and even to deletion of genes essential for assembly of RCC. This remarkable protective effect suggests that OPA1 does not work alone in the regulation of mitochondrial structure and hence function.

3Other key players in cristae morphology include the MICOS complex, a multiprotein structure conserved from yeast to mammals and the mitochondrial F1Fo-ATP synthase, whose oligomers are retrieved on the edges of the cristae and contribute to define cristae curvature–. In mammals, OPA1 and the core MICOS component Mic60 physically interact and cooperate to stabilize the negative CJs curvature; however, CJ and cristae diameter, the key parameters defining mitochondrial apoptosis and respiration, are solely controlled by OPA1, ruling out a role for MICOS in the mitochondrial protection afforded by OPA1. Another potential candidate is the mitochondrial ATP synthase that also can physically interact with OPA1,. The ATP synthase utilizes the mitochondrial proton electrochemical gradient (ΔμH+) generated by the RCC,, to recycle ATP from ADP and Pi,. This enzyme also forms the permeability transition pore (PTP), a large conductance channel whose prolonged openings collapse ΔμH+ and cause cell death. Finally, the ATP synthase can also sustain the ΔμH+ when RCC are inhibited, by running in its reverse mode, hydrolyzing ATP to pump protons across the IMM.

4However, the interplay between ATP synthase and OPA1 in mitochondrial ultrastructure and function is unclear. Here, we provide evidence that OPA1 requires ATP synthase oligomers to protect mitochondria from respiratory chain inhibition.

5OPA1 overexpression is beneficial in vivo against primary and secondary mitochondrial dysfunction, because of its ability to blunt mitochondrial apoptosis and to promote RCS stability. However, the extent and mechanism of this mitochondrial protection are unclear. We therefore capitalized on models of Opa1 mild overexpression and conditional ablation to investigate in real time how OPA1 levels influenced the mitochondrial electrochemical gradient in response to complex III blockage by the inhibitor antimycin A (AA),.

6In mouse adult fibroblasts (MAFs) from Opa1 transgenic (Opa1tg) mice, OPA1 levels and oligomers were 1.5-fold higher (Supplementary Fig. 1a, b) and cristae accordingly ~30% narrower (Supplementary Fig. 1c, d), as previously reported. Real-time measurements of the fluorescence of the potentiometric dye tetramethylrhodamine methylester (TMRM) indicated that Opa1tg MAFs were surprisingly protected by AA-induced mitochondrial depolarization (Fig. 1a, b). We further tested whether OPA1 was able to prevent matrix acidification in the same experimental conditions, by measuring matrix pH with mtSypHer, a ratiometric genetically encoded pH sensor targeted to the mitochondrial matrix. Matrix pH was superimposable in WT and Opa1tg MAFs, irrespective of whether the cells were cultured in media containing glucose or galactose (to force mitochondrial ATP production, Supplementary Fig. 1e). Real-time mtSypHer imaging indicated that OPA1 overexpression blunted also the matrix acidification caused by AA (Fig. 1c, d). We next turned to Opa1flx/flx MAFs where 48 h after transfection with CRE recombinase OPA1 was almost completely absent (Supplementary Fig. 2). In this same timeframe, Opa1 deletion does not modify mitochondrial DNA (mtDNA) content or translation,,. Real-time mtSypHer imaging revealed that matrix acidification induced by AA was more severe upon Opa1 deletion (Fig. 1e, f). Thus, OPA1 protects from electrochemical gradient loss upon CIII blockage.

7We wondered how a dynamin-related protein involved in cristae morphogenesis and membrane fusion could regulate mitochondrial electrochemical gradient. Because reversal of ATP synthase activity can sustain mitochondrial membrane potential upon RCC inhibition, we tested if OPA1 facilitated reversal ATP synthase activity to extrude protons from the matrix and maintain ΔμH+. The protective effect of OPA1 overexpression on AA-induced matrix acidification and depolarization was fully abolished by the ATP synthase inhibitor oligomycin (Fig. 2a–d). Oligomycin also equalized AA-induced pH changes in empty vector (EV) and CRE-transfected Opa1flx/flx cells (Fig. 2e, f). The effects of OPA1 on pH could be a consequence of decreased proton leak, or of increased proton pumping by the reversal mode of the ATP synthase. If proton leak were reduced in Opa1tg MAFs, ATP synthase inhibition would result in a higher ΔμH+ compared to WT cells. However, membrane potential and matrix pH changes were superimposable in WT and Opa1tg MAFs treated with oligomycin (Supplementary Fig. 3a-c), indicating a similar proton leak in these two cell lines. Thus, the differences in ΔμH+ upon CIII inhibition recorded in WT and Opa1tg MAFs could be due to different stimulation of the reversal ATP synthase activity.

8To directly measure if OPA1 affected ATP synthase activity in situ, we monitored ATP synthase-dependent hydrolysis of mitochondrial ATP upon complex III inhibition in real time in living cells. The FRET probe ATeam1.03 was correctly targeted to the mitochondrial matrix and revealed that basal mitochondrial ATP content was similar in WT and Opa1tg MAFs (Supplementary Fig. 3d). Following complex III inhibition, ATeam1.03 fluorescence decayed 3-fold faster in Opa1tg MAFs, suggesting faster ATP hydrolysis. These fluorescence changes were indeed due to ATP synthase activity, because they were abolished by the ATP synthase inhibitor oligomycin (Fig. 2g–i and Supplementary Movie 1). Notably, oligomycin did not affect OPA1 oligomerization in WT or Opa1tg MAFs (Supplementary Fig. 4). Thus, OPA1 overexpression stimulates reversal ATP synthase activity.

9How does OPA1 overexpression impinge on ATP synthase to protect mitochondria from AA? Because OPA1 overexpression favors mitochondrial RCS assembly by impinging on cristae shape, we verified if the same was also true for ATP synthase oligomerization. Blue Native Gel Electrophoresis (BNGE) indicated that oligomeric ATP synthase was stabilized in mitochondria from Opa1tg cells (Fig. 3a, b). When we acutely deleted Opa1 by expressing Cre recombinase in Opa1flx/flx MAFs, ATP synthase dimers and monomers were less abundant, with an increase in free F1 subunit and a reduction in ATP synthase activity and total protein levels (Fig. 3c–e). Thus, OPA1 levels correlate with ATP synthase oligomerization and stability.

10We next verified if OPA1 could stabilize ATP synthase oligomers also during apoptotic cristae shape changes induced by the caspase-8 cleaved, active form of the proapoptotic Bcl-2 family member BID. To this end, we compared the effects on ATP synthase levels and oligomerization of WT BID, of a BID mutant (BIDKKAA) that permeabilizes the mitochondrial outer membrane (OMM) but does not remodel cristae and of an OMM-permeabilization deficient BID mutant (BIDG94E) that causes cristae remodeling but not cytochrome c release,. While total levels of ATP synthase were not affected by treatment of WT or Opa1tg mitochondria with the different BID mutants (Fig. 3f, g), BID and BIDG94E destabilized ATP synthase oligomers only in WT mitochondria. As expected, treatment with the cristae remodeling deficient BIDKKAA mutant did not affect the ATP synthase oligomerization pattern (Fig. 3f, h).

11To further corroborate these biochemical experiments, we turned to an approach of proteomic profiling of high molecular weight (HMW) complexes (complexomic) that allowed us to identify dynamic changes of MICOS during cristae remodeling. We therefore performed a similar analysis focused on ATP synthase complexes. First, we confirmed by immunoblotting and complexomic analysis of native protein complexes separated by BNGE from purified heart mitochondria that treatment with WT BID, but not BIDKKAA, caused the disassembly of OPA1-containing HMW complexes (Fig. 4a–c). We next focused our attention on the oligomeric ATP synthase assemblies. We found that in the BNGE regions corresponding to ATP synthase dimers and oligomers, the spectral counts corresponding to the F1 core subunits ATP5A (α), ATP5B (β), ATP5C1 (γ), ATP5D (δ) and ATP5O (OSCP; see details on the analyzed subunits in supplementary Data 1) were reduced in mitochondria treated with WT BID but not with BIDKKAA (Fig. 4d). The accuracy of this complexomic profiling was further supported by the analysis of the ATP5K (e) subunit essential for ATP synthase dimerization–: ATP5K was retrieved only in the BNGE region corresponding to dimers and its spectral counts dropped to ~40% of control mitochondria upon BID-treatment; the drop was caused by cristae remodeling, because the spectral counts remained ~72% of the untreated in mitochondria challenged with BIDKKAA (Fig. 4e), as further confirmed by immunoblotting (Fig. 4f). Overall, the analysis of the median of ATP synthase subunits spectral counts confirmed that ATP synthase dimers and oligomers were reduced in BID- but not in BIDKKAA-treated mitochondria (Fig. 4g). In conclusion, independent biochemical approaches indicate that ATP synthase oligomers are stabilized upon OPA1 overexpression and destabilized when Opa1 is deleted, or cristae remodeled with concomitant OPA1 HMW oligomers disassembly.

12Because OPA1 levels are proportional to ATP synthase oligomerization, we wished to understand how OPA1 sustained ATP synthase oligomerization. Levels of the ATP synthase inhibitory factor (IF1), whose oligomerization is influenced by matrix pH and that in its oligomeric form can stabilize ATP synthase dimers, were similar in WT and Opa1tg MAFs (Supplementary Fig. 5), suggesting that other mechanisms were in place.

13We first asked whether OPA1 could directly interact with ATP synthase. Chemical crosslinking using 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC) resulted in the appearance of a ~130 KDa band immunoreactive for both ATP5B (β-subunit) and OPA1 (Fig. 5a). Moreover, ATP5B and OPA1 reciprocally co-immunoprecipitated (Fig. 5b, c), confirming and extending previous results that identified several ATP synthase subunits as OPA1 interactors. To verify if OPA1 could also directly promote or stabilize the oligomeric forms of ATP synthase, we turned to an in vitro system of purified proteins and liposomes. We prepared proteoliposomes containing highly pure ATP synthase stabilized in the high affinity lipid like detergent Lauryl Maltose Neopentyl Glycol, as confirmed by EM (Fig. 5d). We also established a protocol to produce and purify a recombinant form of soluble OPA1 (lacking the import sequence and the transmembrane domain, supplementary Figure 6a) by adding a 6-His-Tag to its C-terminus. Following induction of expression in competent bacteria we purified recombinant soluble OPA1 (rOPA1) by affinity chromatography on Ni-NTA beads. Since a contaminant protein with ATPase activity bound to rOPA1, we removed it by extensive washes with ATP, before we eluted by increasing imidazole concentrations obtaining moderate yields of rOPA1 devoid of the contaminating ATPase activity (Supplementary Fig. 6b). Enzymatic activity of the dialyzed rOPA1 determined by reverse phase chromatography was in the range of other dynamin-related GTPases, confirming that rOPA1 was active and stable. We therefore incorporated rOPA1 into the proteoliposome lumen, mimicking the relative topology of OPA1 and ATP synthase in mitochondria (Supplementary Fig. 6c). When we analyzed these proteoliposomes by BNGE, we did not observe any effect of rOPA1 on ATP synthase oligomerization (Fig. 5e, f). In a further in vitro experiment, we added increasing concentrations of rOPA1 to native purified ATP synthase in a digitonin based buffer. However, also in these conditions rOPA1 failed to stimulate ATP synthase dimerization; if anything, in the presence of rOPA1, ATP synthase was mostly monomeric (Fig. 5g, h). Thus, despite its interaction with ATP synthase subunits, OPA1 does not directly stimulate or stabilize ATP synthase oligomerization.

14In Opa1tg mitochondria ATP synthase oligomers are more abundant and matrix pH is maintained upon complex III inhibition. We therefore turned to a genetic approach to test whether OPA1 required ATP synthase oligomers to protect from mitochondrial CIII blockage. ATP synthase subunit e (ATP5k) is required for ATP synthase dimerization–, but its ablation does not compromise monomer function,,. Efficient ATP5k silencing in WT and Opa1tg MAFs (Fig. 6a) yielded a superimposable reduction in ATP synthase dimers/total ATP synthase ratio (Fig. 6b, c). Because subunit e downregulation causes mitochondrial ultrastructural defects,,, we inspected cristae shape in ATP5k silenced mitochondria. We morphometrically evaluated three parameters: cristae width, cristae number, and the ratio between CJ and cristae number, an indicator of the presence of arched or septate cristae, i.e., cristae connecting the IMM at two or more CJ. While ATP5k downregulation did not affect cristae width (Fig. 6d, e), it reduced cristae number and it increased the CJ/cristae number ratio (Fig. 6d, f, g), indicating the formation of arched and septate cristae with less tips. In Opa1tg MAFs the formation of these arched/septate cristae was blunted (Fig. 6g). Interestingly, ATP5k silencing reduced OPA1 oligomers in WT but not in Opa1tg MAFs (Supplementary Fig. 7a, b), suggesting that the formation of arched cristae caused by ATP5k silencing in WT cells depends on disassembly of OPA1 oligomers that are more stable upon OPA1 overexpression,. Our results indicate that OPA1 can compensate for the ultrastructural defects caused by reduced ATP synthase dimerization. A diametrically different scenario emerged when we delved into the mechanism of Opa1tg cells protection from AA-induced mitochondrial depolarization. ATP5k downregulation did not reduce resting ATP content in WT or Opa1tg cells (Supplementary Fig. 8a),. Conversely, it completely abolished the effect of OPA1 on protection from AA-induced depolarization (Fig. 7a–c and Supplementary Movie 2), as well as on stimulation of ATPase reversal function, measured by following matrix ATP hydrolysis in real time (Fig. 7d, e; Supplementary Fig. 8b). Thus, OPA1 requires ATP5k and hence efficient ATP synthase dimerization to protect from mitochondrial dysfunction. If this model was correct, OPA1 should inhibit cell death induced by complex III inhibition and ATP5k should be essential for this cytoprotective effect. Indeed, WT MAFs grown in galactose-containing media rapidly died when challenged with AA, whereas Opa1tg MAFs were protected; downregulation of ATP5k equalized death levels between Opa1tg and WT cells (Fig. 7f). Our results indicate that the protection provided by OPA1 overexpression requires ATP5K and efficient ATP synthase oligomerization to sustain mitochondrial ∆pH and curtail cell death following complex III Inhibition.

15How the IMM fusion and cristae biogenesis protein OPA1 preserve mitochondrial function from a plethora of tissue-damaging insults and from respiratory chain inhibition is unclear. Here, multiple lines of evidence point to a key role for ATP synthase oligomerization status and reversal (i.e., ATP hydrolase) activity.

16OPA1 overexpression counteracts multiple insults including ischemia, atrophy and death-receptor mediated apoptosis that converge on mitochondria,, and most remarkably it corrects electron transport chain defects in vivo. Part of its protective action can be attributed to the inhibition of the apoptotic cristae remodeling pathway, resulting in reduced cytochrome c release and apoptosis,–, but how OPA1 improves mitochondrial function when RCC are genetically impaired is less clear. Negative cristae, i.e., concave curvature at cristae junctions, correlate with increased RCS stability and increased residual RCC function. While this could explain the correction of mitochondrial function in mitochondria lacking the complex IV assembly factor Cox15, it does not explain the improvement observed in mice where the essential complex I subunit Ndufs4 is deleted. Our results indicate that cristae shaping by OPA1 fosters ATP synthase oligomers and reversal activity, providing a unifying mechanism for the ability of OPA1 to sustain mitochondrial electrochemical gradient and function when respiratory chain is inhibited.

17Dynamic complexomic analysis of mitochondria undergoing cristae remodeling and OMM permeabilization is a powerful tool to inspect HMW complexes reorganization during cristae shape changes. This method can also identify unexpected OPA1 partners in cristae morphogenesis, such as the MICOS components Mic60 and Mic19 and the SLC25A solute carriers family members that relay respiratory substrates availability to OPA1 to trigger the orthodox to condensed cristae transition. By comparative complexomics we discovered that ATP synthase oligomers are very dynamic and that are affected by cristae shape changes.

18A proteomic profiling of OPA1 interactors identified different subunits of ATP synthase, suggesting a possible mechanism for the stabilization of the latter. Indeed, OPA1 co-immunoprecipates with ATP synthase and both are retrieved in high-order crosslinked complexes, where OPA1 could directly stabilize ATP synthase. However, a reductionist approach of purified recombinant OPA1 and ATP synthase containing proteoliposomes failed to prove that OPA1 directly stimulates or stabilizes ATP synthase oligomerization. These experiments suggest that the cristae curvature, promoted by OPA1, might itself stabilize ATP synthase oligomers. This effect would complement that of ATP synthase dimers on membrane curvature, in a feed forward loop that ultimately leads to stabilization of cristae shape by its multiple molecular determinants.

19Traditionally, ATP synthase dimerization has been regarded as a cornerstone in cristae morphogenesis: rows of ATP synthase dimers at the edge of the cristae maintain the orthodox structure of the latter,. In yeast, genetic dimerization disruption results in aberrant mitochondrial ultrastructure,,. This organization has also functional consequences: the ATP synthase dimers on the cristae edge contribute to create a proton gradient surrounding the enzyme,, ultimately favoring ATP synthesis. By using apoptotic and genetic manipulations, we demonstrate that ATP synthase oligomerization is also affected by cristae curvature: ATP synthase dimers, like supercomplexes, are destabilized during cristae remodeling and when Opa1 is ablated, and stabilized by OPA1 overexpression.

20ATP synthase dimers stabilization improves mitochondrial ultrastructure,, raising the question of whether the effects of OPA1 on cristae shape are secondary to the stabilization of ATP synthase dimers. Ablation of ATP5k (subunit e), a supernumerary ATP synthase subunit required to bend the IMM and enable dimer formation, but not essential for cell growth, results in loss of cristae tips, formation of cross-sectional septa and occasional onion-like cristae,,,,,. OPA1 overexpression reverted these shape changes, indicating that OPA1 can modulate cristae shape independently of ATP synthase dimers that shape cristae edges and tips,,. Our genetic analysis places OPA1 upstream also of the ATP synthase in the pathway that controls cristae width, junctions and number. The combination of in vivo and in vitro experiments presented here suggest that ATP synthase dimerization can also be stabilized by cristae shape, possibly by favoring superassembly in regions of negative curvature generated by OPA1. Conversely, when ATP synthase dimerization was genetically hampered, OPA1 failed to preserve mitochondrial function, indicating a reciprocal functional and structural crosstalk.

21Opa1tg cells are protected from CIII inhibition in galactose-supplemented media, where the contribution of glycolysis to membrane potential maintenance is marginal,,. Mechanistically, OPA1 sustains mitochondrial function by stimulating ATPase reverse activity, which is increased in Opa1tg mitochondria: pharmacological inhibition of ATPase activity abolishes the protective function of OPA1. Similarly, genetic inhibition of ATP synthase dimerization abolishes the ATPase activity stimulation and the maintenance of mitochondrial function by OPA1. In this context, it is conceivable that the reduced matrix acidification observed in Opa1tg MAFs upon complex III inhibition prevents the binding of the ATPase inhibitor IF1 to the enzyme, which optimally occurs at pH 6.5, thereby delaying a IF1-mediated ATPase inhibition and cristae shape stabilization,.

22Our work unravels a bioenergetic mechanism accounting for OPA1 protection against mitochondrial failure and suggests that its overexpression might be beneficial to sustain bioenergetics in mitochondriopathies–, germline defects and neurodegeneration, whereas OPA1 might be a target to correct oxidative phosphorylation changes in cancer–.

23WT, Opa1flx/flx, Opa1tg MAFs were generated in the Scorrano lab and grown in DMEM supplemented with 10% FBS. Unless otherwise stated, glucose in the medium was substituted with 0.9 mg/ml galactose to force ATP production by the respiratory chain.

24Transient transfection with scramble or ATP5k (F1Fo-ATP synthase sub. e; NM_007507.2) shRNA encoding SureSilencing® plasmids (Qiagen, KM31364H, plasmid #3) were performed using Transfectin (Biorad). Following overnight transduction, the rate of GFP expression was typically around 60–70%, as determined by flow cytometry. Co-transfections of SureSilencing® shRNA plasmids or pcDNA3.1 vectors harboring WT or mutant tBID cDNAs were performed at a 3:1 ratio to empty pIRES2-eGFP plasmid (Clontech) or vectors encoding for pSypHer-dMito or wt (AT1.03) or inactive (AT1.03R122K/R126K) ATeam. Acute Opa1 ablation in Opa1flx/flx MAFs was performed by co-transfections with the Cre-recombinase under the control of a PGK promoter (pPGK-Puro, Addgene) and subjected to analysis after 48 h incubations.

25For live imaging, cells (5 × 104) seeded onto 24-mm round glass coverslips and incubated in Ca2+/Mg2+ supplemented HEPES buffer (HBSS, Invitrogen) were transferred onto the stage of an Olympus IX81 inverted microscope (Melville, NY) equipped with a CellR imaging system and a beam-splitter optical device (Multispec Microimager; Optical Insights). Images were acquired using a 40×, 1.4 NA objective (Olympus) and the CellR software. Analysis of fluorescence was performed following background subtraction over mitochondrial regions of interests (ROIs), using the multi-measure plug-in of Image J (NIH). Representative still frames are pseudocolor coded.

26For real time imaging of mitochondrial ΔpH, cells (5 × 104) seeded onto 24-mm round glass coverslips were transfected with pSypHer-dMito and analyzed after 24 or 48 h after transfection to express Cre recombinase. Ratiometric images of the 535-emission fluorescence were acquired every 10 s by alternate excitation of cells at 430 and 500 nm for 100 ms. Mean fluorescence ratios of selected ROIs matching mitochondria were measured and expressed as mtSypHer (430/500 nm) ratios.

27Mitochondrial ATP content was determined by FRET image analysis of cells transfected with pcDNA-ATeam1.03. Sequential images of the 525 and 475 nm fluorescence emission after alternate excitation at 435 nm for 100 ms were acquired every 30 s.

28For TMRM fluorescence analysis, sequential images were acquired every 30 s.

29Electron microscopy (EM) imaging of cells was performed as described on a Tecnai G2 (FEI) transmission electron microscope operating at 100 kV. Images were captured using a Veleta (Olympus Imaging System) digital camera (pixel size 13 × 13 µm; pixel size at a 46,000X magnification with screen magnification of 3 × 0.1 × 0.1 nm). For morphometric analysis of mitochondrial cristae, performed in a blind fashion on at least five mitochondria/cell from six randomly selected cells (n = 3 independent experiments), maximal cristae width was measured using the ImageJ Multimeasure plug-in; the number of horizontal cristae and cristae junctions were quantified manually.

30For SDS-PAGE experiments, proteins (30–40 µg) were separated on Any-KD (BioRad) polyacrilamide precasted gels, transferred onto PVDF membranes (BioRad) and probed using the indicated antibodies and isotype matched HRP-conjugated secondary antibodies. The following primary antibodies were employed at 1:1000 dilution: OPA1 (BD, #612607), ATP5A (ab14748), ATP5B (ab14705), IF1 (ab110277), SDHA (ab14715), ATP5 subunit e (ab54879) from Abcam; ACTIN (Chemicon) was used at a 1:30000 dilution (MAB1501 Millipore). Densitometry was performed using ImageJ gel measure tool and analyzing the optical density of selected ROIs containing ATP synthase dimers, monomers and F1. Uncropped scans of relevant blots are included in Supplementary Figure 9.

31Mitochondria were extracted from cells grown in 500 cm2 dishes. After isolation, mitochondrial protein concentration was determined by Bradford assay (BioRad) and 0.5 mg/ml protein were incubated in Experimental Buffer (EB: 150 mM KCl, 10 mM Tris MOPS, 10 µM EGTA-Tris, 1 mM KHPO4, 5 mM glutamate, 2.5 mM malate) and further incubated where indicated with 10 pmol/mg cBID for 30 min at room temperature.

32Isolated mitochondria were resuspended (0.5 mg/ml) in NativePAGE Sample buffer (Invitrogen) containing 1.1% (w/V) digitonin (Sigma) and protease-inhibitor cocktail (Sigma). After 5 min on ice, the lysate was spun at 20,000xg for 30 min at 4 °C. G250 (5%, 1 µl/100 µg protein, Invitrogen) was added to the supernatant 30–40 µg of protein were loaded onto a 3–12% native precast gel (Invitrogen).

33Where indicated, mitochondrial extracts were crosslinked in the presence of the zero-length crosslinker EDC. Proteins (50 µg) were incubated for 30 min at 37 °C in PBS supplemented with 10 mM EDC. The reaction was quenched by adding 15 mM dithiothreitol (DTT) and proteins were separated by SDS-PAGE after 15 min.

34Isolated mitochondria were lysed in 150 mM NaCl, 25 mM Tris-Cl pH 7.4, 1 mM EDTA, 5% glycerol, 0.1% Triton X-100 in the presence of Protease Inhibitory Cocktail (PIC) (Sigma). Lysates (250 μg) were precleared on 20 μl of Protein-A agarose beads (Roche) for 30 min at 4 °C and subsequently immunoprecipitated with protein-A agarose beads coupled with the indicated antibodies in lysis buffer overnight at 4 °C. The immunoprecipitated material was separated by SDS-PAGE.

35Liposomes were prepared from purified soybean asolectin (L-α-phosphatidylcholine, Sigma). Lipids were dissolved in chloroform (5 mg/ml) until a homogeneous mixture was obtained; the solvent was then evaporated on a nitrogen stream to yield a thin lipids layer on a glass tube bottom. The lipid film was thoroughly dried by placing the tube on a vacuum pump overnight to remove residual chloroform. To obtain large multilamellar vesicles (LMV) the lipid film was hydrated with 1 mL buffered solution (150 mM KCl, 10 mM Hepes, pH 7.4), containing where indicated 20 µg of purified recombinant OPA1 (rOPA1), and gently agitated at room temperature. When rOPA1 was added, liposomes were then centrifuged at 30,000×g for 5 min and resuspended in fresh buffer devoid of rOPA1. LMV were then downsized to liposomes (large unilamellar vesicles, LUV) by extrusion through a polycarbonate filter with a pore size of 100 nm (Avanti Polar Lipids).

36Intact mammalian F1FO-ATP synthase was purified from beef heart mitochondria as described and inserted into freshly made liposomes by direct incubation of the protein (20 µg) with the liposomes solution for 30 min at 4 °C. To remove non-inserted F1FO-ATP synthase complexes proteoliposomes were pelleted by centrifugation at 30,000×g for 5 min and resuspended in buffer.

37Twenty-five µl of containing freshly prepared proteoliposomes suspension were placed on a 400-mesh holey film grid, stained with 1% UO2(CH3COO)2 and observed with a Tecnai G2 (FEI) transmission electron microscope operating at 100 kV. Images were captured using a Veleta (Olympus Imaging System) digital camera.

38Purified ATP synthase was supplemented with Coomassie Blue G-250 (Serva) and applied to 1D 3–12% polyacrylamide gradient BNE (Invitrogen). After electrophoresis, proteins were eluted from BNE gels overnight as follows. Bands corresponding to monomers of F-ATP synthase were excised and diluted with 25 mM tricine, 7.5 mM Bis-Tris, and 1% (w/V) n-heptyl β-D-thioglucopyranoside, pH 7.0 supplemented with 8 mM ATP-Tris, and 10 mM MgSO4. Samples were incubated at 4 °C, centrifuged at 20,000 × g for 20 min at 4 °C, and supernatants were inserted into liposomes as described above.

39Liposomes were then solubilized with 4% (wt/vol) freshly prepared digitonin, supplemented with Coomassie Blue G-250 (Serva), and loaded in 2D-BNE followed by silver staining. Gels were fixed overnight with formaldehyde, rinsed with ethanol, and pretreated with a solution of 0.8 mM Na2S2O3; gels were then stained with 11.2 mM AgNO3 for 20 min, and then with 0.6 M Na2CO3 for the time required for the bands to be revealed.

40OPA1 mouse transcript variant 2 (NM_133752) was amplified from position 502 to the stop codon to produce recombinant OPA1 (rOPA1) lacking the N-terminal transmembrane domain. The PCR product was cloned into pET21 + (Novagen) which adds a His-tag to the C-terminus of the encoded protein and expressed in E. coli. Protein production was induced with 0.5 μM IPTG for 20 h at 18 °C. E. coli were collected by centrifugation and the pellet was resuspended in lysis buffer (40 mM Hepes/KOH, 500 mM NaCl, 10% glycerol, 5 mM MgCl2, 5 mM β-mercaptoethanol, 0.5% Triton-X-100, 2% Tween-20, 1 mM PMSF, 20 mM Imidazole, Roche protease-inhibitor cocktail, pH 8.0). Cells were lysed by sonication and cell debris was removed by centrifugation at 14,000xg for 45 min, 4 °C. rOPA1 was purified by Ni-NTA batch chromatography. Prior to elution, beads were washed with 1 mM ATP and 10 mM MgCl2 in lysis buffer for 30 min at room temperature followed by an intermediate washing step in wash buffer (40 mM Hepes/KOH, 300 mM NaCl, 0.05% Triton-X-100, 5 mM MgCl2, 5 mM β-mercaptoethanol, 10% glycerol, 20 mM Imidazole, pH 8.0). rOPA1 was eluted with increasing concentrations of imidazole in elution buffer (40 mM Hepes/KOH, 0.05% Triton-X-100, 5 mM β-mercaptoethanol, 10% glycerol, pH 7.4). Imidazole was removed by dialysis and the protein was concentrated in storage buffer (40 mM Hepes/KOH, 0.05% Triton-X-100, 0.3 mM TCEP, 10% glycerol) and stored at -80 °C until use.

41ATPase in-gel activity was measured directly in the gel by incubating it for 2 h in a solution containing 35 mM Tris, 270 mM glycine, 14 mM MgSO4, 0.2% Pb(NO3)2, and 8 mM ATP, pH 7.8,.

42Mass spectrometry analysis of mitochondrial complexes from mouse CD1 hearts was performed as indicated. False discovery rate (FDR) of identification was controlled as described by the algorithm Dxtractor. Median values of all identified ATP synthase subunits and representative color-contour plots of spectral counts of OPA1 peptides, core F1 ATP synthase subunits and FO complex subunit e required for dimerization were considered for analysis.

43For cell death analysis, 3.5 × 103 cells/cm2 of the indicated genotype were co-transfected with SureSilencing® shRNA and pIRES2-eGFP plasmids (3:1 ratio). After 48 h, the medium was replaced with galactose-supplemented DMEM and after 24 h cells were treated with 5 µM antimycin A for 6 h. Cell death was assessed by flow cytometry detection (FACSCalibur) of double Annexin-V-APC/PI positive events from the transfected GFP+ cell population.

44Results are expressed as the mean ± SEM values of the indicated number (n) of independent experiments. Individual data points are overlaid on the bar graphs. Statistical significance was determined by Student's t test or ANOVA between the indicated samples. P values are indicated in the legends and P < 0.05 was considered significant.

45The data that support the findings of this study are available from the corresponding author upon reasonable request. The uncropped blots can be found in Supplementary Figure 9.