Defining the molecular mechanisms of the mitochondrial permeability transition through genetic manipulation of F-ATP synthase

F-ATP synthase is a leading candidate as the mitochondrial permeability transition pore (PTP) but the mechanism(s) leading to channel formation remain undefined. Here, to shed light on the structural requirements for PTP formation, we test cells ablated for g, OSCP and b subunits, and ρ0 cells lacking subunits a and A6L. Δg cells (that also lack subunit e) do not show PTP channel opening in intact cells or patch-clamped mitoplasts unless atractylate is added. Δb and ΔOSCP cells display currents insensitive to cyclosporin A but inhibited by bongkrekate, suggesting that the adenine nucleotide translocator (ANT) can contribute to channel formation in the absence of an assembled F-ATP synthase. Mitoplasts from ρ0 mitochondria display PTP currents indistinguishable from their wild-type counterparts. In this work, we show that peripheral stalk subunits are essential to turn the F-ATP synthase into the PTP and that the ANT provides mitochondria with a distinct permeability pathway.

T he permeability transition (PT) is a Ca 2+ -dependent permeability increase of the mitochondrial inner membrane to ions and solutes with molecular mass up to about 1500 Da [1][2][3] . It is today generally accepted that the PT is due to the opening of a channel, the PT pore (PTP), as first proposed by Haworth and Hunter in 1979 1-3 . This channel, also called mitochondrial megachannel (MMC), was later identified in patch-clamp experiments in mitoplasts, which defined its maximal conductance (as high as 1.3-1.5 nS) and a number of distinctive, smaller subconductance states 4,5 . The PTP and the MMC are considered to be the same molecular entity because they respond in the same way to the same set of agonists and antagonists 6,7 . The latter include cyclosporin (Cs) A [8][9][10][11] , which desensitizes the PTP to opening after binding cyclophilin (CyP) D in the matrix 12 .
The molecular nature of the PTP is the matter of debate. The first potential candidate has been the adenine nucleotide translocator 1 (ANT), which was later shown to form Ca 2+ -activated channels with conductance of 0.3-0.6 nS that are activated by CyPD and inhibited by ADP 13,14 . The selective ANT inhibitors atractylate 15 (ATR) and bonkgrekate 16 (BKA) have opposite effects on the PT. ATR, which locks the ANT in the "c" conformation (nucleotide binding site facing the cytosol) shows a PT-stimulating effect; while BKA, which locks the protein in the "m" conformation (nucleotide binding site facing the matrix) instead shows a PT-inhibiting effect, suggesting that pore opening and closure could be related to a specific, Ca 2+ -dependent conformational change of the ANT 1 . Given that mitochondria from Ant1 −/− Ant2 −/− and Ant1 −/− Ant2 −/− Ant4 −/− mice still undergo a CsA-sensitive PT, albeit at increased matrix Ca 2+ loads; 17,18 and considering that deletion of the Ppif gene (which encodes CyPD) in the Ant1 −/− Ant2 −/− Ant4 −/− background totally prevents the PT 18 , at least another CyPD-sensitive channel must mediate PTP formation 19 . The second major candidate for PTP formation is the F-ATP synthase. This hypothesis was put forward after the demonstration (i) that CyPD binds to, and modulates the F-ATP synthase in a CsA-sensitive manner; 20 and (ii) that partially purified F-ATP synthase generates Ca 2+ -activated channels with the features expected of the corresponding PTPs in bovine 21 , human 22 , yeast 23 , and drosophila 24 mitochondria. Investigations based either on knockdown 25 or on selective ablation of individual subunits of F-ATP synthase [26][27][28] have generated conflicting results, since both persistence [26][27][28] and inhibition 25,29 of the PT have been reported to occur. In yeast, absence of the "dimerization" subunits e and g, and of the Nterminal segment of subunit b 30 , which closely interacts with subunit g 31 , dramatically decreases both size of the PTP and channel conductance of F-ATP synthase 32 . Furthermore, point mutations that do not affect either assembly of the enzyme complex or ATP synthesis did cause specific changes in the channel properties of the PTP 22,32-37 . Finally, highly purified F-ATP synthase preparations displayed the features expected of the PTP in electrophysiological experiments 38,39 . In order to address the many open questions about the role of F-ATP synthase in channel formation, we have studied the features of the PTP by in situ techniques and by patch-clamp recordings of mitoplasts deriving from HeLa cells ablated for subunit g (Δg), from HAP1 cells individually ablated for subunit b (Δb) and subunit OSCP (ΔOSCP) 28 and from ρ 0 cells derived from 143B osteosarcoma cells lacking mitochondrial (mt) DNA, and therefore devoid of subunits a and A6L 40 .
Here, we show that peripheral stalk subunits are essential to turn the F-ATP synthase into the PTP and that the ANT provides mitochondria with a distinct permeability pathway. Our results resolve a number of outstanding questions about PTP formation by F-ATP synthase and about the role of ANT in the occurrence of the PT, and open new perspectives in understanding this central mystery of mitochondrial biology.

Results
The permeability transition in HeLa-Δg cells. To test its role in PTP formation, we generated HeLa cells where the g subunit of F-ATP synthase had been deleted by CRISPR/Cas9 technology ( Fig. 1a; see Supplementary Fig. 1 for the structure of F-ATP synthase). Absence of subunit g also drastically lowered the level of subunit e, which was virtually undetectable (Fig. 1b), indicating that expression of these two proteins is coordinated. Other components of the lateral stalk were also affected by subunit g ablation, with decreased expression of peripheral stalk subunits b, OSCP and f, while the levels of subunit c were normal (Fig. 1b). Clear native-PAGE analysis revealed that in the absence of subunits g (and e) the complex migrated at lower molecular weight, with the appearance of a species (Fig. 1c, asterisk) which may represent a "vestigial" form of the enzyme 28 . Expression of ANT2 and ANT3, the two major isoforms of the translocator expressed in proliferating cells, was not altered (Fig. 1d). Deletion of subunit g had a strong impact on respiration, which was drastically reduced and became insensitive to oligomycin but could be stimulated by FCCP (Fig. 1e). Note that FCCP-stimulated respiration in wild-type (WT) cells was lower than basal, a toxic effect due to the combination with oligomycin, which indeed was not seen with FCCP alone (Supplementary Fig. 2a). Lower respiration of Δg cells was matched by a dramatic decrease in the expression of respiratory complexes I, III, and IV ( Supplementary Fig. 2b), which has also been observed in HAP1 cells after deletion of peripheral stalk subunits and of the c ring [26][27][28] . A possible explanation is that respiratory chain complex I and F-ATP synthase share the assembly factors TMEM70 and TMEM242, which could be part of a negative regulatory mechanism connecting the levels of complex I to those of F-ATP synthase 41 . Cell growth was also impaired in HeLa-Δg cells ( Supplementary Fig. 2c), a finding that can be explained by the severe defects of respiration and ATP synthesis. Δg mitochondria exhibited a significant reduction in the fraction undergoing swelling upon treatment with a Ca 2+ bolus (Fig. 1f), and consistently showed an increased Ca 2+ retention capacity (CRC) (Supplementary Fig. 2d).
To further explore the effect of subunit g ablation on the PTP, we studied mitochondria in living cells. HeLa cells were loaded (i) with calcein followed by Co 2+ to quench the cytosolic calcein signal and allow detection of mitochondrial calcein 42 and (ii) with the potentiometric probe TMRM to detect changes of mitochondrial membrane potential 43 (Fig. 2a). CsA-sensitive loss of mitochondrial calcein fluorescence detects the occurrence of the PT even for very short PTP open times 42 while TMRM release requires longer-lasting PTP openings associated with the release of cytochrome c 44 . We added a cell-permeant hexokinase (HK) 2 peptide that displaces HK2 from the outer mitochondrial membrane 45 and rapidly increases mitochondrial matrix Ca 2+ by direct transfer from the endoplasmic reticulum, causing PTP opening 46,47 . This is a useful tool to selectively increase Ca 2+ transfer to mitochondria without perturbing ion gradients across all membranes. Within 2 min of peptide addition, only HeLa WT cells underwent rapid loss of calcein and TMRM fluorescence, suggestive of PTP opening that was confirmed by the full protection exerted by CsA (Fig. 2b). In good agreement with resistance to PTP opening, HeLa-Δg cells maintained unaltered levels of calcein and TMRM fluorescence throughout the recording period. Importantly, (i) expression of HK2 was maintained and even somewhat increased in Δg cells (Supplementary Fig. 3a), possibly due to a more glycolytic phenotype; and (ii) the rise of mitochondrial Ca 2+ elicited by HK2 peptide was not significantly different in WT and Δg cells, indicating that resistance to opening was not due to reduced matrix Ca 2+ load ( Supplementary Fig. 3b). The features of the PTP at the single channel level were tested next. High-conductance stable currents with multiple subconductance states were induced by 0.3 mM Ca 2+ in WT HeLa mitoplasts (i.e., mitochondria devoid of the outer membrane) and these currents were fully blocked by Ba 2+ (Fig. 2c, left panel and Supplementary Fig. 3c). In striking contrast, under the same conditions, channel activity was never detected in 19 independent experiments with HeLa-Δg mitoplasts, even after increasing [Ca 2+ ] to 0.9 mM (Fig. 2c, right panel and Supplementary Fig. 3d). Channel activity in WT mitoplasts was inhibited by CsA ( Supplementary Fig. 3e) but not by BKA ( Supplementary Fig. 3f).
The total absence of Ca 2+ -induced channels was surprising, as we would have predicted the appearance of channels mediated by the ANT as observed in HAP1-Δc cells 29 . We therefore tested the effects of ATR and BKA on PT occurrence in cells and PTP opening in patched mitoplasts. As Ca 2+ is needed for channel formation by ANT 13,14 , we treated cells with the ionophore A23187. In these protocols the ionophore was preferred to the HK2 peptide because it allows to calibrate the PTP response to Images are representative of three independent blots. b Western blot analysis of isolated mitochondria from wild-type (WT) and Δg cells for the indicated F-ATP synthase subunits. Histogram refers to the quantification of protein levels relative to citrate synthase (CS) and represents mean ± SEM of 3 (for subunit a) or 4 (for all other subunits) independent blots, *p < 0.05, two-sided Student's t-test. Gray bars, WT and cyan bars, HeLa-Δg cells. c Clear native-PAGE analysis and subsequent immunoblotting against ATP synthase subunits β, c, and against SDHA of WT and Δg mitochondria in the presence of indicated amount of digitonin (g digitonin/g protein). Images are representative of two independent blots. d Western blot on isolated mitochondria for ANT2 and ANT3. Histogram refers to the quantification of protein levels relative to citrate synthase (CS) and represents mean ± SEM of three independent blots. Two-sided Student's t-test. e Mitochondrial oxygen consumption rate (OCR) was evaluated in intact cells by Seahorse XF Analyzer before and after the addition of oligomycin (O), FCCP (100 nM), rotenone (R), and antimycin A (AA). Traces are average of four independent experiments for WT (gray trace) and Δg cells (cyan trace). Basal and oligomycin-sensitive OCR is expressed as mean ± SEM of four independent experiments, *** p < 0.001 with two-sided Student's t-test. f Swelling assay in isolated mitochondria in the presence (black traces) or absence (red traces) of Ca 2+ . PTP opening was induced with 50 μM Ca 2+ , and alamethicin (ala) was added where indicated. Histograms refer to the fraction of swollen mitochondria after about 9 min of Ca 2+ addition and are mean ± SEM of six independent experiments, *** p < 0.001, two-sided Student's t-test.
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25161-x ARTICLE Ca 2+ . We determined the A23187 concentration that does not activate PTP opening per se, and then treated cells with ATR. We confirmed that calcein release is only marginal in Δg cells at 5 μM A23187, i.e., a concentration that readily activates the PTP in WT cells ( Supplementary Fig. 3g). Preincubation with ATR significantly accelerated the rate of calcein signal loss after A23187 administration both in WT and Δg cells, a process that was prevented by BKA and CsA in WT cells, and only by BKA in Δg cells (Fig. 2d). Patch-clamp measurements in isolated mitoplasts confirmed the presence of an ATR-induced channel in HeLa-Δg mitoplasts (Fig. 2e).
To study PTP activity in the absence of peripheral stalk subunits b and OSCP we used HAP1 cells (a kind gift of Prof. Sir John E. Walker), as these have been thoroughly characterized 28 and thus allow a meaningful comparison to be made with the electrophysiological features of the pore, which in these cells have not been tested before. As already seen for subunits g and e, expression of subunits b and OSCP appears to be coordinated. Indeed, Δb mitochondria had a considerable decrease in the expression of subunit OSCP and vice versa ( Supplementary  Fig. 4a). The expression level of subunits e and g was also dramatically decreased in both mutants ( Supplementary Fig. 4a) and these subunits could not be detected in F-ATP synthase complexes in clear-native gels ( Supplementary Fig. 4b). We tested the occurrence of the PT in situ in cells loaded with calcein/Co 2+ and TMRM as before. To our surprise, within 5 min of the addition of We then tested the features of the PTP at the single channel level. High-conductance, Ca 2+ -activated stable currents were observed in mitoplasts from all HAP1 cell lines (Fig. 3c), although the frequency was lower in mitoplasts from the deletion mutants. No significant differences were observed in maximal conductance (G max ), mean conductance (G mean ), and net charge passing through the channel during its maximal activity (Q 4s )  ANT mediates permeability transition in HAP1 cells lacking OSCP and b subunits. Given that HAP1-ΔOSCP and HAP1-Δb cells lack lateral stalk subunits (including subunits g and e, which according to the results in HeLa-Δg cells should have caused a loss of PTP activity), the results described in the preceding paragraph were unexpected. It has been reported that HAP1-Δc cells lack a PTP but show a CsA-sensitive channel activated by Ca 2+ and inhibited by BKA, which might be mediated by ANT 29 .
In order to assess whether PTP formation in HAP1-ΔOSCP and HAP1-Δb cells could occur through ANT, we tested the effects of BKA, which does not inhibit channel activity of the purified F-ATP synthase 38 . In intact WT HAP1 cells occurrence of the PT triggered by HK2 detachment was not prevented by preincubation with BKA (Fig. 4a, left panel and Fig. 4b), which instead substantially decreased the mitochondrial calcein and fluorescence loss in HAP1-Δb and HAP1-ΔOSCP cells (Fig. 4a, central and right panel, respectively, and Fig. 4b). Consistent with these findings, in WT mitoplasts currents were insensitive to BKA (Fig. 4c, left panel and Fig. 4d) while BKA blocked the currents in Δb (Fig. 4c, middle panel and Fig. 4d) and in ΔOSCP mitoplasts (Fig. 4c, right panel and Fig. 4d). Two hundred second recordings are provided in Supplementary Fig. 5. Taken together, these data strongly suggest that both F-ATP synthase and ANT can contribute to the permeability transition in HAP1 cells.
The permeability transition pore in ρ 0 cells. ρ 0 cells lack mtDNA-encoded proteins 40 (Supplementary Fig. 6a) and therefore their ATP synthase lacks the a and A6L subunits. We analyzed calcein and TMRM fluorescence in living cells and the effects of treatment with the HK2 peptide as described above.
Although ρ 0 mitochondria have no respiratory activity and their ATP synthase does not pump protons, they maintain a membrane potential by hydrolyzing ATP and thus allowing the electrogenic exchange of extramitochondrial ATP for matrix ADP 48,49 . The HK2 peptide caused CsA-sensitive loss of mitochondrial calcein and TMRM fluorescence in both ρ + and ρ 0 cells, consistent with the onset of permeabilization (Fig. 5a, b and Supplementary Fig. 6b) and in keeping with previous results 46 .
The time-course analysis revealed that: the process of fluorescence decrease occurred rapidly, being nearly complete within about 2 min of addition of the HK2 peptide; it was very similar for ρ + and ρ 0 cells; and the effect of CsA was somewhat more complete in ρ + cells, particularly for TMRM (Fig. 5a, b). While these in situ measurements are useful to detect the occurrence of the PT in a population of mitochondria, they do not provide information on whether the absence of subunits a and A6L has affected pore conductance. We therefore tested the features of the pore at the single channel level. The addition of 0.3 mM Ca 2+ elicited high-conductance stable currents with multiple subconductance states in 11 out of 17 experiments for ρ + cells and in 10 out of 12 experiments for ρ 0 cells (Fig. 5c). In both cell types, currents were completely inhibited by Mg 2+ /ADP ( Fig. 5c and Supplementary Fig. 6c, d) and by CsA ( Supplementary Fig. 7a) as well as by Ba 2+ (Supplementary Fig. 7b). On the contrary, both currents were insensitive to BKA ( Supplementary Fig. 7c). Statistical analysis of single channel activity revealed no significant differences between ρ + and ρ 0 cells in the maximal or mean conductance, nor in the net charge passing through the channel during its maximal activity (Fig. 5d). We conclude that subunits a and A6L do not contribute to PTP formation.

Discussion
Whether the PTP originates from a Ca 2+ -dependent conformation of F-ATP synthase is the matter of debate. Evidence in favor is based on reconstitution experiments from mitochondria of various origins 21-24 , on knockdown of subunit c 25,29 , on generation of point mutations that affect specific channel properties 22,[32][33][34][35][36][37] , and on reconstitution of channel activity from highly purified and fully functional F-ATP synthase from bovine and porcine hearts 38,39 . Evidence against is provided by persistence of a CsA-sensitive PT after genetic ablation of subunit c and of constituents of the F-ATP synthase peripheral stalk [26][27][28] . The findings of the present study provide a solution to this apparent discrepancy and shed new light on the molecular bases of the PT and on the mechanisms of PTP formation.
Our results indicate that in wild-type HeLa, HAP1, and 143B osteosarcoma ρ + cells, both the Ca 2+ -induced PT in situ and the high-conductance channel recorded by patch-clamp in mitoplasts is inhibited by CsA and unaffected by BKA, the selective inhibitor of the ANT. Given that the channel formed by purified F-ATP synthase is insensitive to BKA 38 , we conclude that the PT is mediated by opening of the F-ATP synthase channel (F-PTP). Since both the F-PTP and the ANT channel (A-PTP) are inhibited by CsA, their relative contribution can be inferred from the effects of BKA, which is selective for the A-PTP. Opening of both channels in wild-type cells cannot be excluded, but given the lack of inhibition by BKA we must conclude that in HeLa, HAP1, and ρ + cells the F-PTP predominates. From these experiments it is clear that cell-specific differences exist, and that the basis for these will need to be addressed in the future, including the extent to which the use of HK2 peptide compares to ionophores in inducing mitochondrial Ca 2+ (over)loading. A second important point is that HeLa-Δg cells, which lack key peripheral stalk subunits and therefore do not have a fully assembled F-ATP synthase, do not undergo a PT nor form highconductance channels after treatment with Ca 2+ , and therefore lack the F-PTP. Yet, addition of ATR elicits currents indicating that HeLa-Δg cells have a latent, inducible A-PTP. The presence of the latter could provide an explanation to the eventual Ca 2+ release seen in Δg mitochondria, although the limited respiratory capacity inevitably curbs the ability to accumulate Ca 2+ . The A-PTP was also detected in HAP1-Δb and HAP1-ΔOSCP cells, where prompt BKA-sensitive permeabilization and channel opening followed the addition of HK2 peptide in situ and of Ca 2+ at the patch-clamp. Thus, while in wild-type HAP1 cells the F-PTP predominates (as indicated by lack of inhibition by BKA), in HAP1-Δb and HAP1-ΔOSCP cells (and at variance from HeLa-Δg cells) activation of the A-PTP does not require ATR. Similar findings were obtained in HAP1-Δc cells, which also displayed currents sensitive to BKA that indicate the presence of the A-PTP 29 . It should be noted that in HeLa-Δg and in HAP1-Δb and HAP1-ΔOSCP cells the A-PTP became insensitive to CsA, a puzzling finding that will be further discussed below.
We also studied ρ 0 cells lacking mtDNA 40 . This is a very interesting model because the F-ATP synthase lacks subunits A6L and a, but has an intact peripheral stalk and undergoes full assembly with generation of dimers and oligomers 26,50 , although these may have a lower stability toward detergent extraction 50 . Mitochondria in these cells undergo a process of CsA-sensitive permeabilization consistent with PTP opening 26,46 and have CsAsensitive channels of identical conductance to those of ρ + cells, as shown here. Thus, F-PTP formation does not require a catalytically active F-ATP synthase, nor H + transport through the c ring, which cannot occur in the absence of subunit a 51 . It is intriguing that during apoptosis induction ρ 0 cells undergo mitochondrial permeabilization in a fashion that is indistinguishable from that of control ρ + cells 52 , which is entirely consistent with preservation of the PT. It will be interesting to test whether PTP-defective cells have an altered response to cell death inducers, although decreased ATP synthesis combined with respiratory inhibition make for a very challenging task.
Our results also bear on the question of how the F-PTP originates from F-ATP synthase. Previous experiments highlighted the importance of an intact peripheral stalk, which is essential for enzyme dimerization 21,38 . In turn, this requirement could explain why we could not detect channel activity in monomers 21,38 and why HeLa-Δg cells lack PTP formation in spite of their normal levels of subunit c. This is in apparent contrast with recent findings reporting channel formation from monomers 39 and leading to the conclusion that the F-PTP forms from the c ring, in keeping with earlier suggestions 22,25 . At variance from the classical PTP, however, channel opening was observed in the absence of added Ca 2+ and was inhibited by oligomycin 39 , which does not block the PTP in native mitochondria 53 . As discussed in more detail elsewhere 19 , it is possible that removal of lateral stalk subunits (including e and g) by dodecylmaltoside 39 may have generated an F-PTP that is no longer regulated through subunit e, which directly contacts the lipids within the c ring 54,55 . Thus, both an intact peripheral stalk and the c ring appear to be required for F-PTP formation in situ. The Ca 2+ -induced conformational change would originate at the enzyme crown region 33 , and would be transmitted via OSCP and the peripheral stalk to the c ring through the "wedge" or "bundle" region formed by the tight association between the N-termini of e, g, and b subunits [54][55][56] . This hypothesis, first suggested by Gerle in the "death finger" model for PTP formation and recently revisited 57 , is supported by a deletion study in yeast 32 and by recent structural data on the entire enzyme complex 54 .
The sensitivity of both F-PTP and A-PTP to CsA suggests that the PT-promoting regulatory protein CyPD interacts with both F-ATP synthase 20,21 and ANT 14,58 . What remains puzzling is why the channel observed in HAP1-Δb and HAP1-ΔOSCP cells, which is sensitive to BKA and thus mediated by ANT, becomes insensitive to CsA as does the PT induced by ATR in HeLa-Δg cells. The best characterized interaction of CyPD with mitochondrial proteins is with subunit OSCP of F-ATP synthase 21 , which has been confirmed in several laboratories [59][60][61][62] . Since both HAP1-Δb and HAP1-ΔOSCP cells lack OSCP, this finding suggests that CyPD binding occurs at OSCP and that the presence of F-ATP synthase is required for the ANT to assume the A-PTP conformation, possibly through a physical interaction with F-ATP synthase in the "ATP synthasome", which may also include the Pi carrier [63][64][65] . Experiments are underway to address this hypothesis. Irrespective of the detailed mechanism through which the F-PTP and A-PTP may communicate, however, the existence of two PTPs provides a convincing explanation for the persistence of Ca 2+ -dependent permeabilization in the absence of an assembled F-ATP synthase [26][27][28] . We are confident that having solved this apparent discrepancy will further boost research on mitochondrial permeability pathways and on their role in physiology and pathology. Generation of HeLa-Δg cells. The CRISPR/Cas9 system was used to create HeLa cell lines lacking the expression of ATP5MG gene encoding ATP synthase subunit g. A pair of guide RNAs targeting exon 1 and exon 2 (see Supplementary Table 1) were subcloned into the BbsI site of px330 plasmid (Addgene). HeLa cells were grown in DMEM (Gibco 11965) supplemented with 10% FBS, 100 mg/L uridine, non-essential amino acids (Gibco), and vitamins (Gibco) in a humidified atmosphere of 5% CO 2 /95% air at 37°C to 70% confluency in 6-well plates. Cells were then transfected with 6 µl Lipofectamine 2000 with 7 µg px330 gRNA1, 7 µg px330 gRNA2, and 7 µg pAAV Syn-GFP (Addgene). The next day, transfected cells were subjected to FAC sorting based on GFP fluorescence and single cells were placed in individual wells of a 96-well plate. The single colonies were subsequently expanded and the loss of subunit g expression was confirmed by Western blot. For cell growth analysis, 10 × 10 3 WT or HeLa-Δg cells were seeded into a 6-well plate and counted after 48, 72, and 96 h.

Methods
Mitochondrial isolation. Cells grown to 90% confluence were washed twice with cold phosphate-buffered saline (Sigma-Aldrich), detached using a scraper, and centrifuged for 5 min at 600 × g. The resulting pellet was resuspended in 2 ml of 250 mM sucrose, 10 mM Tris-HCl, and 100 µM EGTA (pH 7.4). Then, cells were homogenized using a Teflon Potter and the homogenate was centrifuged at 600 × g for 5 min. The resulting supernatant was centrifuged at 7000 × g for 10 min at 4°C and the pellet containing intact mitochondria was resuspended in 50 µl of the above medium and quantified with the BCA method.
Mitoplast preparation and patch clamp. Isolated mitochondria were diluted (1:100) in a solution of 30 mM Tris-HCl, pH 7.4 and let to swell at ice-cold temperature for 10 min to obtain mitoplasts (i.e., mitochondria without the outer membrane). The suspension was then inserted in the patch-clamp chamber and washed with the recording medium. Mitoplasts were well distinguishable from debris, being characterized by a typical cap region (formed by remnants of the outer membrane); mitoplasts suitable for patch clamping, with a diameter of 2-5 µm, were visually selected. Patch-clamp recordings were performed using borosilicate pipettes (5 MΩ) in a solution of 150 mM KCl, 10 mM HEPES, 0.3 mM CaCl 2 (pH 7.4) both in the pipette and in the bath. Giga-ohm seals were established by gentle suction of the membrane section opposite to the cap; the mitoplast membrane, corresponding to mitochondrial inner membrane was maintained intact, leading to a mito-attached configuration. Data were sampled at 10 kHz and filtered at 500 Hz. Single channel currents were monitored at constant holding potential (V h ) of +20 mV. Data were acquired at 10 kHz using a L/M EPC-7 amplifier (List-Medical, Darmstadt, Germany), digitized and stored with a Digidata 1322 A and PClamp8.0 acquisition software (all from Molecular Devices). Inducers and inhibitors were added in the bath during the experiment. When indicated, atractyloside (ATR) was presented in the pipette solution and in the bath solution.
Current analysis. Data were analyzed using Clampfit software (Molecular Devices) and MATLAB 2007b (MathWorks). Maximal conductance (G max ) was calculated for every experiment as the maximal transition in channel conductance between two stable states (transition duration <10 ms) detected with a multi-Gaussian fitting of the current amplitude histogram. Mean conductance (G mean ) was calculated, after offset correction, as the average of the mean conductance measured during channel activity in 30 s before administration of the blocker for each experiment. The Q 4s parameter, representing the net charge passing through the fully open channel in a time interval of 4 s, was calculated for each experiment as the integral over 4 s of the current signal at the maximal activity. Statistical comparison of data was assessed with the two-sided Student's t-test.
Live cell imaging. For epifluorescence microscopy, cells were seeded onto 24 mm diameter round glass coverslips and grown for 1-2 days in the proper culture medium described above. Cells were incubated in DMEM without phenol red (Gibco) plus 0.8 μM cyclosporin H (CsH, Adipogen) to inhibit P-glycoprotein. Mitochondrial membrane potential was monitored with 20 nM tetramethylrhodamine methyl ester (TMRM, Invitrogen) in combination with 0.5 µM calcein-AM (Invitrogen) and 8 mM CoCl 2 to detect PTP openings as described 42 .
To test the effect of atractyloside (ATR), HeLa cells were incubated in HBSS (H1387 SIGMA) supplemented with CsH for 30 min with 2 mM CoCl 2 and for another 10 min with 0.5 μM calcein-AM. When indicated, cells were incubated since the beginning with 50 μM ATR alone or in combination with 2 μM BKA. After calcein-AM loading, cells were washed with PBS and incubated with HBSS devoid of CoCl 2 . After 1 min of calcein-AM fluorescence recording, the Ca 2+ ionophore A23187 was added as indicated in figure legends. Recordings were performed with a DMI6000B inverted microscope (Leica, HCX Plan Apo 40x oil objective, NA 1.25), while keeping cells in the incubation solution. TMRM was excited using an EL6000 lamp (Leica) combined with a 540-580 nm bandpass optical filter and a 595 nm dichroic mirror to reflect the light beam. Emission light passed through the 595 nm dichroic mirror and a 607-683 nm bandpass optical filter. Calcein was excited using the aforementioned lamp combined with a 460-500 nm bandpass optical filter and a 505 nm dichroic mirror. Emission light passed through the 505 nm dichroic mirror and a 512-542 nm bandpass optical filter. Emissions were collected by a DMC4500 CCD camera (Leica). Fluorescence emission was sampled every 30 s using LAS AF software (Leica). After background subtraction, images were analyzed with ImageJ, calculating the fluorescence emissions generated by exciting cells at 480 and 560 nm, respectively, in specific regions of interest (ROIs) comprising the entire mitochondrial network. For GCAMP6f Ca 2+ measurements, cells were transfected with a cDNA encoding mitochondrial and nuclear GCAMP6f 66 . To perform Ca 2+ measurements, medium was replaced with DMEM without phenol red supplemented with 0.8 μM CsH (Adipogen) and with 1 mM CaCl 2 . Fluorescence was recorded with an inverted microscope (Zeiss Axiovert 100, Fluar 40x oil objective, NA 1.30) in the 500-530 nm range (by a bandpass filter, Chroma Technologies). Probes were sequentially excited at 475 and 410 nm, respectively, for 180 and 300 ms, every 5 s. Excitation light produced by a monochromator (polychrome V; TILL Photonics) was filtered with a 505 nm DRLP filter (Chroma Technologies). After background subtraction, images were analyzed with ImageJ, calculating the ratio (R) between emissions generated by exciting cells at 475 and 410 nm, respectively, in specific ROIs comprising the entire mitochondrial network. Standard error of the mean for the signal is denoted by the dashed traces above and below the solid lines.
Ca 2+ retention capacity (CRC) and mitochondrial swelling. The CRC was evaluated with Calcium Green-5N (Molecular Probes) using a Fluoroskan Ascent FL (Thermo Electron) plate reader. Isolated mitochondria were resuspended to a final concentration of 0.4 mg/ml in 130 mM KCl, 10 mM MOPS-Tris, 10 μM EGTA-Tris, pH 7.4, 5 mM glutamate, 2.5 mM malate, 1 mM Pi, and 0.5 μM Calcium Green-5N and then subjected to a train of 2.5 µM Ca 2+ pulses. Swelling of isolated mitochondria was evaluated by measuring the absorbance at 540 nm using an Infinite M200Pro (Tecan) plate reader. Briefly, 80 μg of mitochondria were resuspended in 130 mM KCl, 10 mM MOPS-Tris, 10 μM EGTA-Tris, pH 7.4 supplemented with 5 mM glutamate, 2.5 mM malate, and 1 mM Pi. PTP opening was triggered by the addition of 50 μM Ca 2+ . At the end of the experiment, 10 μM alamethicin was added to measure maximal mitochondrial swelling. The fraction of swollen mitochondria was calculated as described 53,67 . Statistical comparison of data was assessed with the two-sided Student's t-test.
Oxygen consumption rate. Oxygen consumption rate in adherent cells was measured with an XF24 Extracellular Flux Analyzer (Seahorse Bioscience). Briefly, HeLa cells were seeded in XF24 microplates at 3 × 10 4 cells/well for WT and at 5 × 10 4 cells/well for Δg cells in 200 μl supplemented DMEM and grown at 37°C in a 5% CO 2 humidified incubator for 24 h. Before starting the assay, the growth medium was replaced with Seahorse medium (DMEM-Sigma D5030) supplemented with 143 mM NaCl, 25 mM glucose, 10 mM sodium pyruvate, 2 mM glutamine, and 15 mg/l phenol red. Cells were incubated at 37°C for 30 min to allow temperature and pH equilibration. After an oxygen consumption rate (OCR) baseline measurement, 1 μg/ml oligomycin, 100 nM FCCP, 1 μM rotenone, and 1 μM antimycin were sequentially added to each well. OCR values were normalized for the protein content and rotenone-and antimycin-insensitive respiration was subtracted. Statistical comparison of data was assessed with the two-sided Student's t-test.
Clear native-PAGE. Clear native-PAGE was performed according to a published protocol 68 . Briefly, isolated mitochondria were resuspended in 50 mM NaCl, 50 mM imidazole/HCl, 2 mM aminocaproic acid, 1 mM EDTA, pH 7.0 at a final concentration of 10 μg/μl, supplemented with the indicated amount of digitonin and subjected to an ultraspin at 100,000g for 25 min at 4°C. The resulting supernatant was collected and supplemented with 5% glycerol and 0.001% Ponceau S solution. Samples were loaded onto a Native-PAGE 3-12% gel and run in the presence of 50 mM Tricine, 7.5 mM imidazole, pH 7.0 cathode, supplemented with 0.05% deoxycholic acid sodium salt (DOC) and 0.01% n-Dodecyl β-D-maltoside (DDM). Samples were then transferred to a PVDF membrane and subjected to western blot analysis for subunit β, c, g, e, and SDHA as indicated in the figure legends.