Inhibition of ATP synthase reverse activity restores energy homeostasis in mitochondrial pathologies

ATP hydrolysis by CV is thought to only occur when mitochondria depolarize. However, given the differences in membrane potential between individual cristae and between cristae and inner boundary membrane within each mitochondrion (Wolf et al, 2019), the question of whether hydrolysis can occur concurrently with synthesis arises. To develop an assay for the measurement of ATP hydrolysis, we took advantage of the proton generated as part of the chemical reaction of ATP hydrolysis (see Appendix Supplementary Methods). The proton produced during this reaction can be measured as the acidification rate of the sample. If membranes are intact, as in fresh coupled mitochondria, protons will leave the mitochondria as part of the reverse activity of CV. When ATP synthase works in reverse, one proton is net transported per 2.67 molecules of ATP hydrolyzed (Hochachka & Mommsen, 1983; Robergs et al, 2004; Ferguson, 2010). To concurrently measure ATP synthesis and hydrolysis in the same sample, we used the capacity of the Seahorse (SH) XF96 Analyzer to measure oxygen consumption and pH changes (acidification rates) in different channels. ATP synthesis is determined by measuring oxygen consumption induced by saturating concentrations of adenosine diphosphate (ADP) (State 3 respiration), while ATP hydrolysis is quantified by measuring acidification via the protons released by CV when hydrolyzing ATP (Fig 1A and B). Since we are using isolated mitochondria, the changes in pH cannot be attributed to glycolysis (Divakaruni et al, 2014), but to proton release by CV hydrolysis.

We isolated intact mitochondria from mouse heart and measured oxygen consumption and proton release in fresh preparations. Measurements were carried out either in the presence of substrates and ADP (respiration linked to ATP synthesis or State 3) or in the presence of substrates without ADP (respiration linked to proton leak or State 4).

The substrates fueling respiration were pyruvate plus malate (Pyr + Mal, Fig 1B and C). The procedure starts with the addition of ADP, stimulating State 3 respiration. A control group, that remains in State 4 without the addition of ADP shows low oxygen consumption rates (OCR), confirming that the mitochondria are coupled. State 3 respiration slowly depletes ADP, leading to State 4 where OCR rates are lower. At this point, we added ATP generating the “State 4 ATP” condition. Remarkably, the addition of ATP resulted in a sharp increase in proton release that was oligomycin sensitive, confirming its origin from ATP hydrolysis and indicating that ATP hydrolysis occurs in coupled mitochondria (Figs 1B and D, and EV1A).

To verify that acidification was a result of CV reverse activity, we depolarized the same mitochondria with carbonyl cyanide p‐(tri‐fluoromethoxy)phenyl‐hydrazone (FCCP), which reverses all CV molecules, and observed an additional increase in acidification (Fig 1B and D, Also, see Appendix Supplementary Methods). Furthermore, oligomycin, a specific inhibitor of forward and reverse CV activities, decreased acidification (Fig 1B and D), which effectively demonstrated that indeed CV activity was the major determinant of proton release in FCCP‐treated mitochondria (Divakaruni et al, 2014). This acidification induced by ATP was not a result of the presence of damaged mitochondria, as the mitochondrial preparations were highly coupled, demonstrated by their high State 3/State 4 OCR ratio. Additionally, we further investigated whether an unexpected presence of broken mitochondria or extramitochondrial ATPases contributed to the acidification detected. Since exposed ATPases and mitochondria with broken membranes are sensitive to proteinase K (PK) treatment, we measured State 3 OCR and acidification rates after ATP injection in the presence of PK. Addition of PK did not change respiration nor acidification, confirming that protons were released from CV of intact mitochondria (Fig EV1B and Appendix Fig S1).

The injection of ATP increased acidification rates of mitochondria respiring under State 3 or State 4 (Figs 1D and EV1A). The observation of a similar acidification response to ATP injection under either State 3 or State 4 indicated that ATP hydrolysis can be detected in the presence of high‐ADP/ATP exchange activity, as the latter can only occur in State 3 mitochondria (State 3 mitochondria have ADP in the matrix prior to ATP injection).

Coupled mitochondria keep a PMF that is expected to prevent ATP hydrolysis by CV. However, the recent discovery that microdomains within the mitochondrial inner membrane have reduced PMF, provided a scenario for hydrolysis to still occur in coupled mitochondria. The assay described here provides a direct measurement of this phenomenon.

Based on these results, we conclude that a population of CV molecules in coupled mitochondria can hydrolyze ATP, while others can produce ATP. The rate of ATP hydrolysis in coupled mitochondria was approximately 20% of the maximal hydrolytic capacity measured in mitochondria uncoupled with FCCP. This percentage suggests that the PMF is responsible for preventing 80% of CV hydrolytic capacity in coupled mitochondria.

ATPase inhibitor (ATPIF1) binds CV and regulates ATP hydrolysis and synthesis (Nesci et al, 2015). Considering that ATPIF1 expression levels vary by tissue and environmental conditions (Campanella et al, 2008, 2009), we hypothesized that decreased ATPIF1 expression levels could play a role in the pathophysiology of mitochondrial disease models. We quantified ATPIF1 levels in skinfibroblasts derived from patients with a variety of mitochondrial diseases. Fibroblasts with mutations causing dysfunction in Complex I (CI; NDUFA6), Complex III (CIII; BCS1L and TTC19), ATP synthase (CV; TMEM and ATP6), and MICOS (MIC26) were analyzed by western blot. We observed a decrease in total ATPIF1 protein levels in all mutants tested (Fig 1E), suggesting that increased levels of ATP hydrolysis could be a shared pathogenic process among patients with mitochondrial diseases. If a reduction in ATPIF1 translates to increased ATP hydrolysis, we would expect that the maximal hydrolytic capacity per CV content would increase.

To quantify ATP hydrolytic capacity in cells, we developed an approach to measure maximal ATP hydrolysis by CV in samples that were previously frozen (Fig EV1C) named Hydrolysis in Frozen Samples or HyFS (Fernandez‐del‐Rio et al, 2023). An increase in maximal CV–ATP hydrolytic capacity was observed in CIII mutants (BCS1L and TTC19), MICOS‐deficient patients (MIC26), and a CV mutant (ATP6) (Fig 1F).

These data show that ATPIF1 downregulation together with increased ATP hydrolytic capacity is a mitochondrial phenotype shared among human fibroblasts from patients with mutations in mitochondrial proteins.

Our results in mitochondrial disease cell models highlight the need for pharmacological interventions to modulate CV hydrolytic activity. The identification of compounds selectively decreasing CV hydrolytic activity would allow us to establish whether the increase in ATP hydrolysis observed in fibroblasts from patients with mutations in mitochondrial proteins is an adaptive or maladaptive event. To find possible molecules that selectively regulate CV hydrolytic activity, we performed an in‐silico screen (Dataset EV1) of 7,441 natural or synthetic derivative compounds from the COlleCtion of Open Natural ProdUcTs database (CoconutDB) (Sorokina et al, 2021). Compounds were categorized for predicted desired properties including free radical scavenging activity (antioxidant) and creatine kinase (CK) inhibition, as well as for docking to ATP synthase. From these, 6,827 ligands were filtered out for predicted undesired activities. Since ATPIF1 binds to the F₁ head of ATP synthase, we narrowed the search by analyzing the ability of the remaining 614 ligands to directly dock with the F₁ head. Within the resulting 445 molecules that were found to have at least −8.5 Kcal/Mol affinity to the selected region, we found a group of 28 compounds belonging to the family of catechins that had a particularly high affinity to the F₁ head.

Binding site exploration simulations were performed with and without ATPIF1 using the Protein Energy Landscape Exploration (PELE) webserver (Fig 1G). We ran this analysis using the ligands identified in the CoconutDB study and included (+)‐Epicatechin, a catechin derivative shown to protect from myocardial ischemia/reperfusion injury (Yamazaki et al, 2008, 2010) by preserving mitochondrial function (Yamazaki et al, 2014). PELE identified a low‐energy binding site for Quercetin, (‐)‐Catechin, and (+)‐Epicatechin in the vicinity of the ATPIF1 binding groove on the surface of the F₁ head. The three compounds bound with similar affinities: AutoDock Vina binding score: −5.3 for (+)‐Epicatechin, −5.7 for Quercetin and −5.5 for (‐)‐Catechin (Fig 1G, top panel).

When comparing the different epicatechin isomers, the resulting pose of (+)‐Epicatechin indicated that the B‐ring catechol hydroxyls form hydrogen bonds with the side chains of E454 and Q456, and the A‐ring resorcinol hydroxyl forms hydrogen bonds with the side chain of H451. The remaining interactions were hydrophobic in nature, with the dihydropyran heterocycle and B‐ring benzyl near L452 and P453 that form the ATPIF1 binding groove. The 3‐OH of (+)‐Epicatechin is positioned away from the surface unlike the same group in (‐)‐Epicatechin, which sterically hinders the interaction (Fig 1G, middle panel). Finally, our data indicated that ATPIF1 and (+)‐Epicatechin align to the same position and orientation in the ATPIF1 binding groove (Fig 1G, bottom panel).

Our results suggest that epicatechin could mimic ATPIF1 and decrease CV‐mediated ATP hydrolysis (Lang & Racker, 1974; Gledhill et al, 2007).

Based on the predicted high‐affinity binding, we chose to further explore the effect of EPI on CV hydrolysis. HyFS analysis in isolated heart mitochondria confirmed that EPI inhibits CV–ATP hydrolysis at a rate of 30% of the maximal ATP hydrolytic capacity. As a positive control, we used oligomycin, which inhibits ATP hydrolysis completely (100%) (Fig 1H and I). To explore EPI specificity, we tested the effect of EPI on electron transport chain (ETC) complexes. EPI did not inhibit respiration on either Complex I (CI) or Complex II (CII) substrates (Fig EV1D). This is in contrast to Quercetin, which has been reported to promote changes in oxidative respiration and redox status (de Oliveira et al, 2016). We next tested whether EPI could modulate respiration driven by CV–ATP synthetic activity (ATP‐linked respiration). EPI at 100 nM promoted a slight increase in State 3 respiration driven by Pyr + Mal without changing maximal respiration rates (MRRs) (Fig 1J). An increase in State 3 respiration that is not accompanied by an increase in MRR or proton leak (Appendix Fig S2), supports our interpretation that EPI specifically acts on ATP synthase.

To determine whether the effects of EPI on ATP hydrolysis were mediated by a direct action on the ATP synthase as the docking data supported, we employed different approaches where EPI was added either to intact mitochondria, broken mitochondria or directly to the gel loaded with mitochondria (Fig 2). First, we treated intact mitochondria with increasing concentrations of EPI. These intact mitochondria treated with EPI were then analyzed by blue native gel electrophoresis (BNGE), to determine whether EPI modulated ATP synthase assembly and in‐gel hydrolytic activity (Acin‐Perez et al, 2020a). EPI decreased in gel ATP hydrolytic activity in a dose‐dependent manner (Fig 2A and B), without altering the assembly of CV (Fig 2C). These results show that EPI inhibition of ATP hydrolysis does not involve disruption of CV assembly. In gel activity allows us to specifically determine that the ATP hydrolytic activity measured is a product of fully assembled CV.

To assess if EPI needs to access the mitochondrial matrix in order to inhibit ATP hydrolysis or whether EPI can bind directly to CV moieties in the intermembrane space, we added EPI to mitochondria permeabilized with digitonin, prior to protein separation by BNGE. EPI inhibited hydrolysis in permeabilized mitochondria with the same efficacy as observed in intact mitochondria (Fig 2D and E). Thus, we can conclude that the effect of EPI is not limited by the rate of its diffusion into the mitochondrial matrix.

To assess whether the actions of EPI on CV were caused by direct interaction of EPI with CV, we added EPI in the ATP hydrolysis reaction buffer in which the blue native gel was incubated. In this way, EPI is incubated directly with the separated CV located in the gel. We found that in‐gel treatment by EPI decreased ATP hydrolysis (Fig 2F and G). Note that higher concentrations of EPI were needed to penetrate the gel and block hydrolysis. These data suggest that EPI directly binds CV to decrease ATP hydrolytic activity, rather than by modulating an endogenous CV regulator.

Given the ability of EPI to mimic ATPIF1 function of blocking CV–ATP hydrolysis and given the predicted binding site of EPI in the ATPIF1‐binding pocket, we hypothesized that EPI could be preventing ATPIF1 from binding CV. To test this hypothesis, we measured the binding of a recombinant purified ATPIF1‐green fluorescent protein (GFP) protein to CV in the presence and absence of EPI. We permeabilized mitochondria by freeze–thaw cycles to enable the direct access of purified ATPIF1‐GFP to CV. We then incubated these permeabilized mitochondria with increasing concentrations of ATPIF1‐GFP, allowing ATPIF1‐GFP to bind CV. The free and CV‐bound ATPIF1‐GFP were then separated by BNGE. Gels were then excited at 488 nm and binding was then quantified as the levels of green fluorescence emitted by ATPIF1‐GFP from the CV band in the BNGE. Quantification shows that ATPIF1‐GFP binds CV in a dose‐dependent manner (Fig 3A, top image). To confirm that ATPIF1 binding to CV was decreasing ATP hydrolysis as expected, we subjected the same gel to an ATP hydrolytic in‐gel activity assay (Fig 3A, middle image). ATPIF1‐GFP inhibited in‐gel hydrolysis in a dose‐dependent manner, confirming the functionality of ATPIF1‐GFP in this procedure. Incubation with ATPIF1‐GFP had no effects on the levels or integrity of CV (Fig EV2A). To further validate that inhibition of hydrolysis by recombinant ATPIF1 can be achieved when CV is within mitochondrial membranes, we measured CV hydrolytic activity using HyFS in frozen and thawed mitochondria. Increasing concentrations of ATPIF1 resulted in a dose‐dependent inhibition of hydrolysis measured by HyFS as well (Fig 3C and D), without affecting ETC complex activities (Fig EV2B–D).

We next addressed the hypothesis that by acting as a competitor of CV, EPI could prevent ATPIF1 binding to CV using either heart mitochondria or bovine‐purified CV tetramers (Jiko et al, 2015; Urbani et al, 2019). In both preparations, we performed binding competition assays between EPI and ATPIF1‐GFP (Figs 3E–G and EV2E and F). In both purified tetramers and isolated mitochondria, EPI prevented the binding of ATPIF1‐GFP (F488) to CV (Fig 3E, top panel, Figs 3F and EV2F). While EPI prevented ATPIF1 binding to CV (Fig 3E, top panel, and Fig 3F), EPI mimicked the ability of ATPIF1 to inhibit in‐gel hydrolysis activity (Fig 3E, middle panel, and Fig 3G) without changing the levels of CV (Fig EV2E), supporting that EPI functions as an ATPIF1 competitive mimetic for CV binding.

Finally, to investigate whether the capacity of EPI to decrease CV‐mediated ATP hydrolysis is dependent on ATPIF1 expression, we studied the in‐gel CV hydrolytic activity in diverse preparations of purified bovine CV (Jiko et al, 2015; Urbani et al, 2019) that contain different levels of ATPIF1, as well as oligomeric, tetrameric and monomeric forms of CV (Fig EV2G). We found that EPI inhibited ATP hydrolysis independently of the amount of ATPIF1 (Fig EV2H–J and Appendix Fig S4). Interestingly, both EPI and oligomycin inhibited ATP hydrolysis, not only in the fully assembled CV and its oligomers, but also in the F₁ subcomplex (Fig EV2H and J and Appendix Fig S4). To further confirm that EPI inhibition of ATP hydrolysis does not require ATPIF1, we analyzed the efficacy of EPI treatment in tissues expressing different amounts of ATPIF1 using HyFS (Fig EV2K). EPI had the same efficacy in decreasing ATP hydrolysis in tissues with low‐ATPIF1 expression (Fig 3H). Consequently, and in agreement with the data in purified CV, we conclude that the inhibitory action of EPI on ATP hydrolytic activity does not require the presence of ATPIF1.

To determine whether ATPIF1 binds to CV in intact cells, we utilized the Proximity Ligation Assay (PLA) to quantify the binding of the endogenous ATPIF1 with the alpha subunit of CV F₁ head (Fig 4A). PLA detects the events of proximity between two antibodies. The two antibodies carry complementary PCR primers that allow for a PCR reaction to occur when the two antibodies bind two targets that are located within 10 nm or less of each other. The PCR reaction produces a fluorescent signal, and the localization and frequency of proximity events can be quantified. To develop the PLA assay between CV and ATPIF1, we used an anti‐ATPIF1 antibody and an anti‐ATP5A1 antibody. PLA analysis in control human fibroblasts treated with EPI for 24 h shows that EPI treatment decreases ATPIF1 binding to CV in control cells (Fig 4B) for all the concentrations tested, further supporting that in intact living cells EPI can either prevent or displace ATPIF1 binding to CV. BNGE analysis of control cells cultured for 24 h demonstrated that EPI treatment results in a decrease in the levels of ATPIF1 bound to assembled CVI, although the observed trend did not reach statistical significance (Fig EV4C and D).

To distinguish if EPI prevents ATPIF1 binding to CV or if EPI can displace the already bound ATPIF1 from CV, we ran displacement and competition assays side by side. To test for displacement, we incubated purified CV with ATPIF1‐GFP for 5 min to allow for binding to take place. EPI was then added to the mix for an additional incubation of 10 min before the sample was analyzed by BNGE, to determine the amount of ATPIF1‐GFP bound to CV (Fig EV2F). In a parallel competition assay, we incubated purified CV tetramers with ATPIF1‐GFP in the presence of EPI. We found that EPI treatment did not displace ATPIF1‐GFP that was already bound to CV. However, if EPI and ATPIF1‐GFP were incubated concurrently, ATPIF1 binding to CV was decreased. These results suggest that EPI can prevent ATPIF1 binding to CV, but cannot displace an already bound ATPIF1 from CV.

In a living cell, EPI can both prevent ATPIF1 binding and inhibit ATP hydrolysis, however the net effect on inhibition of ATP hydrolysis under conditions in which both mechanisms are in competition has yet to be explored. To address this question, we overexpressed ATPIF1 (Fig EV3A and B) in the presence or absence of EPI and measured maximal ATP hydrolytic capacity (Fig 4C). We also tested EPI competition with the most active form of ATPIF1, a constitutively active mutant of ATPIF1, H49K, whose binding is insensitive to changes in pH (Schnizer et al, 1996; Garcia‐Aguilar & Cuezva, 2018).

Both ATPIF1‐WT and ATPIF1‐H49K overexpression decreased maximal ATP hydrolytic capacity in control fibroblasts (Fig 4C). Addition of EPI to cells overexpressing either ATPIF1‐WT or ATPIF1‐H49K, did not result in any further inhibition beyond the effect induced by the overexpression of the two forms of ATPIF1. These data suggest that by competing with ATPIF1 on CV binding, EPI does not produce the same inhibition of hydrolysis that can be produced by ATPIF1. These results suggest that EPI is more likely to be efficacious in conditions where ATPIF1 levels are reduced.

Previous studies have demonstrated that ATP hydrolysis by CV is required to preserve mitochondrial membrane potential under conditions of impaired respiratory function when PMF is lost (Connolly et al, 2018). To test the effect of EPI on mitochondrial membrane potential, we adapted the previously well‐described Antimycin A (AA) model to induce CV–ATP hydrolysis. This model takes advantage of the opposite effect of oligomycin in ATP‐producing versus ATP‐hydrolyzing mitochondria. While oligomycin treatment in cells with intact oxidative phosphorylation (OXPHOS) results in mitochondrial membrane potential hyperpolarization, the addition of oligomycin to cells pretreated with AA, to block respiration, results in depolarization. This shift in oligomycin response, from hyperpolarization to depolarization termed “oligomycin null point” represents the state at which membrane potential relies on CV–ATP hydrolysis (Rego et al, 2001; Vesce et al, 2005; Nicholls et al, 2013).

To determine if under complete inhibition of respiration EPI can induce membrane potential depolarization, we first reproduced the oligomycin null‐point phenomenon in human fibroblasts using AA and oligomycin. To ensure that effective concentrations of AA and oligomycin are used, we ran a parallel analysis of OCR with each of the compounds separately. OCR analysis determined that AA at 50 nM results in complete inhibition of respiration (Fig EV3C, left panel) with a concomitant increase in extracellular acidification rate (ECAR) indicating a glycolytic shift in ATP production (Fig EV3C, right panel), while oligomycin showed a dose‐dependent effect on respiration between 50 nM and 5 μM (Fig EV3D).

To produce an oligomycin null‐point phenomenon, we treated fibroblasts with increasing concentrations of AA for 30 min in the presence or absence of oligomycin. We monitored membrane potential using the fluorescence membrane potential indicator, tetramethylrhodamine ethyl ester (TMRE), and we used Mitotracker Green (MTG) (membrane potential independent) to control for mitochondrial mass. As expected, while oligomycin treatment alone led to hyperpolarization, the combination of AA and oligomycin at 5 μM resulted in a statistically significant depolarization (Fig EV3E and F). Lower concentrations of oligomycin trended toward a dose‐dependent depolarization (Fig EV3F), although did not reach statistical significance. This result indicates that complete inhibition of CV is required to induce the oligomycin null‐point. Moreover, this assay established that treatment of these cells with 5 μM AA for 30 min is required to produce the oligomycin null‐point.

To determine if, as oligomycin, EPI can induce depolarization under complete inhibition of respiration, we repeated the same experiment, but this time replaced oligomycin with EPI. Remarkably, EPI did not lead to membrane potential depolarization (Fig EV3E). Altogether, we conclude that in order to induce membrane potential depolarization under AA, a complete inhibition of ATP hydrolysis is required, and that partial inhibition by low‐oligomycin concentrations or by EPI is not sufficient to overcome the compensatory mechanisms that maintain membrane potential.

To determine the effect of inhibition of CV–ATP hydrolysis on mitochondrial matrix ATP content (mtATP), we monitored mtATP levels by using BioTracker™ (Wang et al, 2016) staining detected by fluorescence microscopy. Treating fibroblasts with AA for 30 min induced a dose‐dependent depletion of the mitochondrial matrix ATP pool (Fig 4D). Remarkably, EPI treatment restored mtATP in AA‐treated cells (Fig 4D). This result is in agreement with recent studies by Rieger et al (2021), showing that mtATP content specifically responds to changes in ATPIF1 expression.

We next aimed to determine whether EPI treatment effectively decreases ATP hydrolysis in fibroblasts from patients with mitochondrial diseases, caused by mutations in mitochondrial proteins. EPI treatment for 30 min induced a dose‐dependent increase in mtATP in each of the mutant fibroblasts tested (Fig 4E). To rule out that EPI was increasing mitochondrial respiration to restore mtATP levels, we monitored OCR before and after the addition of EPI. No differences in OCR were observed in intact cells after EPI injection (Fig EV3G). Such a rapid increase in mtATP is consistent with EPI directly inhibiting CV hydrolytic activity in isolated mitochondria and purified CV (Fig 2).

To determine whether the EPI‐mediated increase in mtATP content initiated a restoration in cellular ATP homeostasis, we measured the effects of a 24 h EPI treatment on total cellular ATP content (Figs 4F and EV3H). Remarkably, treatment with EPI resulted in a significant increase in ATP in two of the mutant fibroblasts tested, a CIII deficiency (CIII‐BCS1L) and a CV deficiency (CV‐ATP6) (Fig 4F). Other mutants (P2:BCS1L, MIC26, and TMEM) only showed a slight increase in total cellular ATP, as did the control fibroblasts (Fig EV3H). Furthermore, CIII‐deficient cells treated with EPI showed a statistically significant, mild improvement in proliferation rates after 72 h of treatment, while control and CV‐deficient cells showed a similar trend, although not statistically significant (Fig 4G).

Altogether, these results demonstrate that impaired OXPHOS and CV–ATP hydrolysis both contribute to the depletion in cellular ATP observed in mutant fibroblasts.

The above results suggest that the inhibition of hydrolysis alone is sufficient to improve ATP availability in conditions of impaired respiration. However, since the observed increase in ATP content could be the result of improved respiration, we further examined the effect of EPI on respiratory chain structure and function.

We chose to examine the effect of EPI on respiration in two mutant cells: CIII‐BCSL1 and CV‐ATP6. Both have reduced ATP content, but through different mechanisms: in BCS1L mutant, ATP depletion is caused by the combination of impaired respiration and increased ATP hydrolysis, whereas in the CV‐ATP6 mutant, reduction in ATP content is primarily caused by impaired ATP synthesis by CV. We examined basal respiration, a parameter that directly responds to increased OXPHOS. In this assay, we did not treat with oligomycin or FCCP to avoid perturbation of ATP hydrolysis.

When compared with control cells, BCS1L mutant cells showed a decrease in basal, maximal, and ATP‐linked respiration while the ATP6 mutants showed only an increase in maximal respiration (Fig 4H and Appendix Fig S5), as previously reported (Fernandez‐Vizarra et al, 2007; D'Aurelio et al, 2010). Treatment with EPI did not result in any changes to OCR. ECAR was increased in both mutants when compared to control cells, in agreement with the glycolytic shift observed in mitochondrial diseases. EPI treatment showed a tendency to decrease lactate levels only in CIII‐deficient cells, supporting a decrease in ATP demand induced by treatment with EPI (Fig EV3I). The absence of such a trend in CV‐deficient cells supports that ATP synthesis is the major contributor to decreased ATP availability induced by CV‐ATP6 mutation.

EPI did not affect the structure or content of CIII or CV in these cells as determined by BNGE analysis, further confirming that the increase in ATP content is not mediated by changes in complex assembly or content (Fig EV4A–D).

We next assessed the amount of ATPIF1 bound to ATP synthase in CIII‐BCS1L and CV‐ATP6 mutant and control fibroblasts by PLA (Fig 5A–C), which revealed that ATPIF1 binding to ATP synthase was reduced in both mutants (Fig 5B). EPI treatment further decreased ATPIF1 binding both in control and in CV‐ATP6 mutant cells, while the already low levels of ATPIF1 bound in CIII‐BCS1L cells remained unchanged (Fig 5C). Measurements of ATP hydrolysis revealed a negative correlation between ATPIF1 binding to CV and the levels of ATP hydrolysis in these cells (Fig 5D). This negative correlation between ATPIF1 binding and ATP hydrolysis was absent in cells treated with EPI, where ATP hydrolysis was markedly decreased independent of the levels of ATPIF1 binding to CV.

Similar to fibroblasts from patients with other mitochondrial diseases (Fig 1E), we found that CIII‐BCS1L fibroblasts express low levels of ATPIF1, suggesting the possibility that the lack of endogenous hydrolysis inhibition contributes to EPI efficacy in these cells. We, therefore, compared the effect of restoring ATPIF1 expression and activity with the effect of EPI treatment. We produced BCS1L cells stably overexpressing either WT ATPIF1 or the continuously active mutant of ATPIF1, H49K (Fig EV3A and B). Both ATPIF1‐WT and ATPIF1‐H49K expression decreased CV–ATP hydrolytic capacity in CIII‐BCS1L cells, mimicking the effects of EPI in these same mutants (Fig 5E). In marked contrast to EPI treatment, the inhibition of hydrolysis by ATPIF1 overexpression failed to improve either mtATP or total cellular ATP content (Fig 5F and G). These data are consistent with the hypothesis that ATPIF1 blocks the rotation of ATP synthase in both forward and reverse directions, and explains why ATPIF1, in contrast to EPI, does not increase ATP content in CIII‐BCS1L cells.

To determine the role of ATPIF1 content and activity in the efficacy of EPI, we analyzed the efficacy of EPI after restoring ATPIF1 levels by overexpressing WT ATPIF1, or ATPIF1 H49K (Figs 4E, 5D and EV3A and B) (Schnizer et al, 1996; Garcia‐Aguilar & Cuezva, 2018). Addition of EPI in combination with the overexpression of ATPIF1 or ATPIF1 H49K did not result in any additional inhibition of ATP hydrolytic capacity beyond the level achieved with either one of the constructs. Furthermore, addition of EPI in cells already overexpressing WT or H49K ATPIF1 did not improve mtATP or total cellular ATP content.

To further elucidate the mechanism by which ATPIF1 overexpression prevents the effects of EPI, we examined if EPI could compete against ATPIF1 for binding to CV, PLA analysis showed increased ATPIF1 binding to CV in ATPIF1 WT and H49K overexpressing cells, validating that the constructs encoding for ATPIF1 WT and H49K produced proteins capable of incorporating into CV (Fig EV3B). EPI was not able to decrease the levels of ATPIF1 binding to CV in CIII‐BCS1L cells overexpressing ATPIF1‐WT or ATPIF1‐H49K (Fig EV4E and F). These data are in agreement with EPI effectively inhibiting CV hydrolytic activity in conditions where CV binding sites to ATPIF1 are available.

To determine whether inhibiting CV‐mediated ATP hydrolysis can prevent tissue dysfunction in vivo, we chose Duchenne Muscular Dystrophy (DMD) as a disease model, in which mitochondrial depolarization has been reported in association with ATP depletion (Bulfield et al, 1984; Sicinski et al, 1989; McGreevy et al, 2015). DMD is caused by a mutation in the cytoskeletal protein dystrophin. The mdx genetic mouse model of dystrophin deficiency has been used extensively to study the pathophysiology of DMD (McGreevy et al, 2015; Yucel et al, 2018; Hammers et al, 2020). Loss of dystrophin destabilizes the myofiber membrane to cause pathological calcium overload in the cytosol during physical activity (Bonilla et al, 1988; Mareedu et al, 2021). Cytosolic calcium overload leads to an excessive uptake of calcium by the mitochondria resulting in mitochondrial depolarization and permeability transition (Moore et al, 2020; Budzinska et al, 2021; Dubinin et al, 2021). While ATP depletion has been documented in mdx mice and in DMD patients (Kelly‐Worden & Thomas, 2014; Budzinska et al, 2021), the contribution of ATP hydrolysis by CV has not been tested.

To study the contribution of CV‐mediated ATP hydrolysis to dystrophic pathology, we used D2.mdx mice, harboring the mdx mutation in a DBA/2J genetic background (D2.mdx). At 12 weeks of age, D2.mdx mice were subjected to exercise‐induced gastrocnemius eccentric injury by stimulating muscle contraction using external electrodes (Blaauw et al, 2010; Khairallah et al, 2012; Call & Lowe, 2016; Hammers et al, 2020). We analyzed the extent of CV–ATP hydrolysis in gastrocnemius from WT and D2.mdx mice at rest or after exercise‐induced injury. In each mouse, the right leg was subjected to exercise‐induced injury and the left leg was kept at rest as a control. Gastrocnemius muscle was collected 1 or 24 h post‐injury (pi). We found that the gastrocnemius from mdx mice had decreased CV (ATP5A1) and CII (SDHA) content (Figs 6A and EV5A). Exercise‐induced injury further decreased CV (ATP5A1) and CII (SDHA) content 24 h after injury in mdx mice, while no decrease was observed in WT mice (Fig 6A). These decreases in CV and CII occurred in the absence of changes in mitochondrial content (Appendix Fig S6). HyFS analyses of gastrocnemius homogenates from mdx mice revealed an increase in CV–ATP hydrolytic capacity at both 1 and 24 h postinjury (Figs 6B and EV5B, respectively). In contrast, WT mice did not show any change in ATP hydrolytic capacity following the injury. Moreover, gastrocnemius of mdx mice revealed that exercise‐induced injury results in cleavage of the mitochondrial inner membrane protein OPA1 (Figs 6C and EV5C), as well as cytochrome c release (Figs 6D and EV5D). Together, these findings demonstrated that exercise‐induced injury in mdx mice resulted in mitochondrial changes characteristic of calcium overload.

The observation of a robust increase in CV–ATP hydrolysis in injured mdx mice revealed that these mice are a suitable disease model to test the impact of inhibiting CV–ATP hydrolysis in vivo on pathology. EPI administered twice a day for 2 weeks decreased susceptibility to injury in mdx mice in a dose‐dependent manner, when compared with vehicle treated mice (Fig 7A and B). Remarkably, EPI treatment decreased ATP hydrolysis in the injured muscle from mdx mice, an effect that was accompanied by a trend towards increased CV content (Figs 7C and EV6A). Analysis of muscle function demonstrated that the efficacy of EPI treatment improving the post‐injury muscle force was strongly correlated with ATP hydrolysis inhibition (Fig 7D). Additionally, EPI treatment blocked OPA1 cleavage (Fig 7E) and cytochrome c release (Figs 7F and EV6B). A dynamic process of plasma membrane damage and repair occurs when mdx mice develop exercise induced‐muscle injury, during which the material from the interstitial fluid can enter the cell. The extent and duration of plasma membrane barrier breach are tested using Evans Blue Dye (EBD), by measuring how much dye can enter the cell through regions of broken plasma membrane (Hamer et al, 2002). Twenty‐four hours prior to injury, mdx mice were injected with EBD and the number of Evans Blue positive cells was quantified post‐injury. EPI treatment decreased the number of Evans Blue positive fibers, indicating that EPI protected the integrity of gastrocnemius muscle fibers (Fig 7G). To further validate that the mechanism of improvement was through inhibition of hydrolysis, we studied human myotubes that model DMD in vitro. Our data showed that EPI was able to inhibit hydrolysis and preserved membrane integrity in DMD mutant cells (Fig EV6C–E), supporting that the beneficial effect of EPI can be cell autonomous. Principal component analysis (PCA) of different variables measuring overall muscle status, including muscle function, bioenergetics, OPA1 cleavage, cytochrome c release, and cell membrane integrity, showed that EPI triggered the improvement in muscular function in mdx mice, and restored parameters of muscle function in WT‐injured mice, that clustered with the control (non‐injured) groups (Fig 7H). These data support the hypothesis that ATP hydrolysis by CV contributes to muscle damage and accelerates mitochondrial apoptosis, via OPA1 cleavage.

D2.mdx, DBA2/J, C57BL/10ScSn.Dmd<mdx>/J (mdx) and C57BL/10ScSn/J mice were purchased from Jackson laboratories. For the EPI treatment study, mice were treated orally twice daily, 12 h apart, from doses ranging from 0.5 to 15 mg/kg/day. After 13 days of treatment, mice underwent muscle function testing and susceptibility to eccentric injury was determined, as described below. The following day, 1 h after the final dose, the mice were euthanized, and tissues were collected. All animal protocols were approved by the Animal Care and Use Committee at the University of Maryland School of Medicine. Samples were not blinded.

Maximal force production of the plantar flexor muscle group was measured in vivo with a 305C muscle lever system (Aurora Scientific Inc., Aurora, Canada), as previously described (Khairallah et al, 2012; Call & Lowe, 2016). Animals were anesthetized via inhalation (~2% isoflurane, SomnoSuite, Kent Scientific), placed on a thermostatically controlled table with anesthesia maintained via nose‐cone (~2% isoflurane), the knee fixed with a pin pressed against the tibial head, and the foot firmly fixed to a footplate on the motor shaft. Contractions were elicited by percutaneous electrical stimulation of the tibial nerve (0.2 ms pulse, 500 ms train duration) at increasing frequencies. Following assessment of isometric torque, susceptibility to injury was assayed with 25 eccentric contractions as previously described (Khairallah et al, 2012; Call & Lowe, 2016) at maximal isometric torque (150 mS duration, 0.2‐mS pulse train at 150 Hz). Eccentric contractions were achieved by translating the footplate 38° backward at a velocity of 800°/S after the first 100 mS of the isometric contraction. The decrease in the peak isometric force before the eccentric phase was taken as an indication of muscle damage.

Evans Blue Dye (EBD) detection was used as previously described as an in vivo marker of muscular fiber damage (Hamer et al, 2002). Briefly, mice were injected intravenously with 1% EBD (Sigma, St Louis, MO, USA) (w/v) in phosphate‐buffered saline (PBS, pH 7.5), 24 h before they were euthanized. Tissues were harvested, fixed, and imaged as described (Hamer et al, 2002).

Mouse gastrocnemius muscles were homogenized in 0.75 ml MAS (70 mM Sucrose, 220 mM Mannitol, 5 mM KH₂PO₄, 5 mM MgCl₂, 1 mM EGTA, 2 mM HEPES; pH 7.2) using a glass–glass Dounce homogenizer. Thirty strokes with the tight pestle were applied per sample. The homogenate was centrifuged at 1,200 × g to pellet non‐broken material and the supernatant was stored at −80°C until further use.

Primary fibroblasts derived from patients were acquired from Telethon Network Genetic Biobank (NDUFA6^{A178P/c.420+784C>T}, P1:BCS1L^R73C/F368I, P2:BCS1L^R183C/R184C, P1:TTC19^L219X, TMEM70^Q197X), Coriell Institute for Medical Research (ATP6^L156R), and Control and MIC26^I117T (Beninca et al, 2021). Cells were maintained in EMEM (Eagle's Minimum Essential Medium; ATCC #30‐2003) and supplemented with 15% FBS and antibiotic–antimycotic (ThermoFisher, #15240062) in humidified atmosphere at 37°C and 5% CO₂. During the treatment with the different compounds, media was changed to phenol red‐free DMEM (Dulbecco's Modified Eagle Medium; ThermoFisher #A1443001) containing 4.5 g/l glucose or galactose supplemented with 2 mM glutamine, 1 mM sodium pyruvate, 10% FBS, and antibiotic–antimycotic. Compounds were prepared in DMSO and added directly to the growth media at the specified concentration. Lentiviral infection of pLentiATPIF1‐WT and pLentiATPIF1‐H49K (produced by Welgen, Inc) was performed using 8 μg/ml polybrene in EMEM for 30 h and positive selection performed by 1 μg/ml puromycin.

Immortalized myoblasts derived from a healthy control and a DMD patient (DMD KM1328, mutation in exon 52 of DMD gene) were grown in Skeletal Muscle Growth Medium (Promocell) containing 1% penicillin–streptomycin at 37°C with 5% CO₂ (Mamchaoui et al, 2011). Upon reaching 100% confluency, the medium was replaced with Skeletal Muscle Differentiation Medium (Promocell) containing 1% penicillin–streptomycin.

To measure cellular proliferation, fibroblasts were plated at 1,000 cells/well in duplicated 96‐well imaging plates 98 h prior to the assay. Twenty‐four hours after plating, one plate was used for imaging of Hoechst staining (1 μg/ml) for cell counting. The second plate was treated with 50 nM (+)‐Epicatechin or DMSO and kept for 72 h until cell counting. Imaging was performed in triplicates with a Perkin Elmer Operetta system (10× objective). Analysis was performed with Perkin Elmer Harmony (Harmony 4.1) software and % change from plate 1 to plate 2 was calculated for the respective cells and treatments.

The osmotic shock base solution was prepared using 5 mM HEPES, 5 mM KCl, 1 mM MgCl₂, 5 mM NaCl, 1.2 mM CaCl₂, and 1 mM glucose in DIW. Sucrose was added to the base solutions to reach 45 milliosmole (mosmol). The actual osmolarity was determined using a VAPRO vapor pressure osmometer (Wescor Inc.) Myotubes in 24‐well plates were treated for 1 and 24 h at days 4–5 of differentiation. On day 5 of differentiation, the cells were subjected to 20 min of osmotic shock at 37°C using 45 mosmol solution. The supernatant was collected and centrifuged to separate cell debris. Adherent cells were trypsinized and pelleted before lysis in 50 μl of DIW and 3 freeze–thaw cycles. The Creatine Kinase Assay (Sekisui Diagnostics) was used to measure creatine kinase (CK) levels in both the supernatant and lysate fractions. In a 96‐well plate, 8 μl of lysate or 12 μl of supernatant with 140 μl of reagent was loaded per well in triplicate. The U/l of CK was calculated as follows: (mOD/min) (total volume in ml) (dilution factor)/(6.22 M⁻¹ cm⁻¹) (light path in cm) (sample volume in ml). The percent CK release was calculated as follows: CK_{extracellular}/(CK_{extracellular} + CK_{intracellular}) × 100.

Mice were anesthetized with isoflurane followed by a cervical dislocation, the heart was immediately removed and placed in ice‐cold relaxation buffer (5 mM sodium pyrophosphate, 100 mM KCl, 5 mM EGTA, and 5 mM HEPES; pH 7.4). The heart was squeezed with tweezers to remove blood, minced with scissors, and then placed in a glass–glass Dounce homogenizer with 3 ml of HES homogenization buffer (250 mM sucrose, 5 mM HEPES, and 1 mM EDTA; pH to 7.2, adjusted with KOH). The heart tissue was homogenized with 10 strokes with the loose pestle and 15 strokes with the tight pestle. The homogenized tissue was placed in a prechilled 15 ml conical tube and centrifuged at 900 × g (4°C) for 10 min. The supernatant was removed, placed in a new tube, and centrifuged again at 900 × g for 10 min. The supernatant was then transferred to 2 ml microcentrifuge tubes and centrifuged at 10,000 × g (4°C) for 10 min. The mitochondrial pellets were resuspended in ice‐cold HES buffer and mitochondrial protein was measured with a BCA assay (Pierce). The concentrated mitochondrial pellet was stored on ice.

Purification of mammalian F‐type ATP synthase was conducted as previously described (Urbani et al, 2019) with slight modifications. Fresh B. taurus hearts were obtained immediately after slaughter by an authorized slaughterhouse and fat and connective tissues were carefully removed allowing the preparation of 1,000 g of minced meat. Each 500 g was suspended in 3,250 ml of 23 mM sodium phosphate buffer (pH 7.4 at 0°C) and homogenized for 5 min at 13,000 rpm in a homogenizer (Nihon Seiki), followed by centrifugation for 20 min at 1,000 × g in a refrigerated centrifuge (Kubota Model 9810; RS‐6600 rotor). The precipitate was suspended in 3,375 ml of 22.2 mM sodium phosphate buffer (pH 7.4) and rehomogenized, followed by centrifugation for 20 min at  1,000 × g. Supernatants were then combined and centrifuged for a further 30 min at 8,000 × g with a refrigerated centrifuge (Beckman Model Avanti HP‐30I) using a JLA‐10.500 rotor. The precipitate was then suspended in 50 mM Tris–HCl buffer (pH 8.0) and pelleted for 30 min at 22,000 × g with an ultracentrifuge (Beckman Model‐7) using a 45 Ti rotor. The pellet was suspended in 50 mM Tris–HCl buffer (pH 8.0) containing 660 mM sucrose to a final protein concentration of ~23 mg/mL. The suspension was kept in a 40 mM HEPES buffer (pH 7.8) containing 2 mM MgCl₂, 0.1 mM EDTA, and 0.1 mM DTT and solubilized on ice via slow addition of deoxycholate and decyl‐maltoside to final concentrations of 0.7% (wt/vol) and 0.4% (wt/vol), respectively. The suspension was then centrifuged at 176,000 × g   for 50 min and the supernatant was applied to a sucrose step gradient (40 mM HEPES pH 7.8, 0.1 mM EDTA, 0.1 mM DTT, 0.2% wt/vol decyl‐maltoside and 2.0 M, 1.1 M, 1.0 M, or 0.9 M sucrose) and centrifuged at 176,000 × g for 15.5 h. Fractions exhibiting ATP hydrolysis activity were loaded onto a Poros‐20HQ ion‐exchange column. The detergent was exchanged to LMNG using a double gradient from 0.2 to 0% decyl‐maltoside and 0–0.05% LMNG for 80 min at 1 ml/min. Complexes were eluted by a linear concentration gradient of 0–240 mM KCl in 40 mM HEPES pH 7.8, 150 mM sucrose, 2 mM MgCl₂, 0.1 mM EDTA, 0.1 mM DTT and 0.005% (wt/vol) LMNG. For preparation of monomer and tetramer fractions, ion‐exchange chromatography was replaced by an additional overnight sucrose density gradient ultracentrifugation step (176,000 × g for 15.5 h) conducted using either 0.05% LMNG (monomer) or 0.05% GDN (tetramer) as gradient detergents.

Primary fibroblasts were plated at 6,000 cells/well 48 h prior to the assay. Twenty‐four hours after plating, standard maintenance media was replaced with high glucose (4.5 g/l) DMEM supplemented with 2 mM glutamine, 1 mM sodium pyruvate, 10% FBS, and antibiotic–antimycotic, and cells were treated with 50 nM (+)‐Epicatechin. On the day of the assay, the medium was replaced with assay medium composed of DMEM (Sigma #D5030) with 5 mM HEPES, pH 7.4 supplemented with 2 mM glutamine, and 1 mM sodium pyruvate. OCR and ECAR were measured in a SH XF96 analyzer under basal conditions as well as after injection of 2 μM oligomycin, two sequential additions of 1.5 μM FCCP, followed by 1 μM rotenone with 2 μM Antimycin A (AA). Respiratory parameters were calculated according to standard protocols (Divakaruni et al, 2014), and all rates were corrected for nonmitochondrial respiration/background signal by subtracting the OCR insensitive to rotenone plus AA.

ATP hydrolysis capacity or State 4 acidification rates were measured using Agilent Seahorse XF96 as described (Fernandez‐del‐Rio et al, 2023). Briefly, plates were loaded with 0.75–1.5 μg of mouse heart mitochondria, 20 μg of mouse gastrocnemius homogenate, or 25 μg of cell lysate diluted in MAS (70 mM Sucrose, 220 mM Mannitol, 5 mM KH₂PO₄, 5 mM MgCl₂, 1 mM EGTA, 2 mM HEPES; pH 7.2). Cell lysates were prepared by subjecting the samples to 4 cycles of free‐thaw (liquid nitrogen‐37°C water bath) before measuring protein concentration. Initial respiration of the samples was sustained by the addition of 5 mM succinate + 2 μM rotenone in the MAS after centrifugation. Injections were performed as indicated in the figure panels at the following final concentration in the well: Antimycin A (AA) (2 μM), oligomycin (5 μM), FCCP (1 μM), ATP (20 mM). To assess maximal ATP concentration, ATP was injected consecutively.

Quantification of the ratio of hydrolysis/synthesis in heart mitochondria fueled with either pyruvate plus malate or succinate plus rotenone (Succ+Rot) was determined by calculating the ratio between the ECAR rate on mitochondria started in state 4 after ATP injection (hydrolysis) and the basal OCR rate of mitochondria started in State 3 (synthesis) before any port injection.

To determine mitochondrial mass, muscle homogenate (8 μg) in 20 μl of MAS (70 mM Sucrose, 220 mM Mannitol, 5 mM KH₂PO₄, 5 mM MgCl₂, 1 mM EGTA, 2 mM HEPES; pH 7.2) was placed in a clear‐bottom 96‐well microplate. Then, 130 μl of a 1:2,000 dilution of Mitotracker Deep Red FM (MTDR, ThermoFisher), a mitochondrial selective dye insensitive to weak or mild changes in membrane potential, was added and incubated for 10 min at 37°C. Plates were centrifuged at 2,000 × g for 5 min at 4°C (no brake), and supernatant was carefully removed. Finally, 100 μl of PBS was added per well and MTDR fluorescence measured (λ_excitation = 625 nm; λ_emission = 670 nm). Mitochondrial content was calculated as MTDR signal (minus blank) per microgram of protein as previously described (Acin‐Perez et al, 2020a).

Cells were fixed with 4% paraformaldehyde in phosphate‐buffered saline (PBS) for 20 min at RT, permeabilized (0.1% Triton X‐100, 0.05% sodium deoxycholate in PBS), blocked, and stained with primary and secondary antibodies in blocking solution (5% donkey serum). Titration of primary antibodies dilution (1:5,000) was made for visualization of spaced PLA dots. Saturation of antibodies was avoided. Coverslips were mounted with Mowiol. For PLA, cells were plated on coverslips and then fixed, permeabilized, and blocked as described above. After that, we followed the manufacturer's protocol using the Duolink In Situ Red Starter Kit Mouse/Rabbit (Sigma Aldrich). After the last wash step, coverslips were incubated overnight with described primary antibodies (see below) for immunofluorescence. Image acquisition was performed using Zeiss LSM880 Confocal system with Airyscan equipped with a Zeiss Pla‐Apochromat 40×/1.2 and 63×/1.4 N.A. objective. 3D image stacks were acquired at optimal Z‐distance and reconstructed using Zen black software. Quantification of PLA dots and mitochondria volume was performed in aleatorily acquired images using Aivia v.10 software. Pixel classification was used to segment mitochondria and PLA dots. Only PLA dots localizing in mitochondria were counted.

Cells were washed twice with PBS and lysed using 0.5% Triton‐X100 in PBS. After 3 min centrifugation at 11,000 × g, 5 μl of total lysate was used in triplicates for assessment of total ATP content using the ATP Determination Kit (#A22066 Thermo Fisher) and following the manufacturer's instructions. Luminescence was monitored at ∼560 nm using a Tecan Spark luminometer.

Mitochondrial ATP content was assessed with the BioTracker™ ATP‐Red dye, a live cell red fluorescent imaging probe for ATP. The probe specifically reports ATP content in the mitochondrial matrix of living cells. The probe without ATP forms a closed ring structure that is not fluorescent. In the presence of the negatively charged ATP, the covalent bonds between boron and ribose in the probe are broken and the ring opens, causing the probe to be fluorescent.

Cells were incubated for 1 h with 200 nM MTG and 1 μg/ml Hoechst, washed twice with PBS, and incubated for 15 min with 5 μM BioTracker™ ATP‐red dye (Millipore) in medium at 37°C (5% CO₂).

Before imaging, the cells were washed twice with medium and fresh medium was added. Imaging was performed in triplicates and in two focus planes with a Perkin Elmer Operetta system (20× objective) at 37°C and 5% CO₂. Excitation and emission filters used for the combination of dyes: Hoechst (ex. 360–400, em. 410–480), MTG (ex. 460–490, em. 500–550) and mtATP Red (ex. 560–580, em. 590–640). Analysis was performed with Perkin Elmer Harmony (Harmony 4.1) software by measuring the average ATP red fluorescence intensity inside a region of interest (Filimonov et al, 2014) generated by the MTG area.

Cells were incubated for 1 h with 200 nM MTG, 200 nM TMRE (Biotium) and 1 μg/ml Hoeschst, washed twice with PBS, and incubated with 200 nM in medium at 37°C (5% CO₂). Imaging was performed in triplicates and in two focus planes with a PerkinElmer Operetta CLS high‐content system (20× objective) at 37°C and 5% CO₂. Analysis was performed with Perkin Elmer Harmony (Harmony 4.1) software by measuring mean TMRE fluorescence intensity inside an ROI generated by MTG area.

1 × 10⁶ cells were plated and treated for 24 h in the described media supplemented with 10% dialyzed FBS. Cells were fixed and media was frozen at −80°C until further processing. Media was thawed on ice and diluted 1:70 in PBS, 5 μl was used in triplicates for assessment of lactate using the Lactate‐Glo Assay (#J5021 Promega) following manufacturer's instructions. Luminescence was recorded using a Tecan Spark luminometer.

Mouse monoclonal anti‐ATP5A (1:1,000; #439800; Thermo Fisher Scientific), Mouse monoclonal anti‐SDHA (1:1,000, #459200, Thermo Fisher Scientific), Mouse monoclonal anti‐SDHB (1:1,000, #ab14714, Abcam), Rabbit polyclonal anti‐UQCRC2 (1:1,000, #14742‐1‐AP; ProteinTech), Rabbit polyclonal anti‐ATPIF1 (1:1,000, #13268S; Cell Signaling Technology), Rabbit monoclonal anti‐TOMM20 Alexa Fluor 488 (1:1,000; #ab205486; Abcam), Mouse monoclonal anti‐Vinculin (1:1,000, #V9131, Sigma Aldrich), Mouse monoclonal anti‐OPA1 (1:1,000, #612606, BD Transduction Laboratories), and Mouse monoclonal anti‐cytochrome c (1:5,000, #ab110325, Abcam).

Heart permeabilized mitochondria (20 μg) were run on a 3–12% native precast gel (Invitrogen). CV in gel activity was performed as previously described (Acin‐Perez et al, 2020b). CV in gel Activity was stopped in 50% methanol. Finally, gels were stained with Coomassie to correct for loading. Imaging was performed at different stages: after 3 h and O/N In Gel Activity, after fixing with 50% methanol and after Coomassie staining.

Recombinant ATPIF1 was added to bovine mitochondrial CV tetramer or isolated mouse heart mitochondria at a protein/protein ratio 1/5, 1/25, or 1/50 in MAS. Binding assay was performed at RT for 10 min. Samples were then prepared to be loaded in Native gels. Ten micrograms of CV tetramer or isolated mitochondria were used and recombinant ATPIF1 at the indicated ratio was added.

Cytochrome c (Cyt c) release was measured in mitochondrial supernatants of gastrocnemius homogenates. Fifty micrograms of gastrocnemius homogenate were centrifuged at 10,000 × g for 10 min. Supernatant was collected and cytochrome c levels were measured by western blot and ELISA (R&D Systems) using 10 and 3 μg of supernatant, respectively.

Ligands with molecular weight between 80 and 800 kDa were filtered from CoconutDB database (https://coconut.naturalproducts.net/) (Sorokina et al, 2021) using mongoDB. To reinforce the potential mitochondrial activity, PASS analysis (Filimonov et al, 2014) was used for prediction of the three‐biological activities: creatine kinase inhibitor, antioxidant, and free radical scavenger. In‐situ ligand screening was carried out using Autodock Vina 4.2 (Trott & Olson, 2010) against the bovine ATP synthase (PDB: 6ZQN) as a receptor. PYMOL structure viewer (https://pymol.org/2/) was used for visualization and selection of coordinates. The selected region contains the F1 head of the ATP synthase, more specifically with the Alpha and Beta subunits around the central stalk, where ATPIF1 is possibly interacting with the complex (Bason et al, 2011).

All simulations for the interaction of (+)‐ and (−)‐Epicatechin, quercetin and (−)‐catechin with F1‐ATP synthase with ATPIF1 were started from the following x‐ray structure Protein Data Bank ID: 1ohh (Cabezón et al, 2003). Structures of the epicatechins, quercetin and (−)‐catechin were retrieved from Pubchem (CID: 182232: (+)‐Epicatechin; 72276: (−)‐Epicatechin; 5280343: quercetin; 73160: (−)‐catechin). Structures were prepared using YASARA (Land & Humble, 2018), adding hydrogen atoms, checking the protonation states of side chains, and optimizing the hydrogen‐bond network. If necessary, loops were closed using YASARA. Simulations were carried out with ATPIF1 present and absent. PELE (Protein Energy Landscape Exploration) webserver was used to identify potential epicatechin binding sites on F1‐ATP synthase. PELE is a Monte Carlo (MC)‐based technique that consists of three main steps, i.e., ligand and protein perturbation; side chain sampling; and minimization (Borrelli et al, 2005, 2010). Resulting poses were analyzed for ligand affinities and energies using AutoDock Vina (Trott & Olson, 2010). Visualizations were analyzed using UCSF Chimera (Pettersen et al, 2004).

Principal component analysis was carried out in R studio (R version 4.1.1) using Singular value decomposition to determine the principal components and the eigenvectors for the analyzed variables.

Statistical analysis was performed with GraphPad Prism® 9.01 using one‐way, two‐way analysis of variance (ANOVA) or pairwise comparison as indicated in the figure legends. Corrections for multiple comparisons were made by Šídák's or Dunnett's multiple comparisons when appropriate. Differences were considered statistically different at P < 0.05. Statistical significance is denoted by P‐value in each figure. Individual points in a graph denote individual cell preparations, mouse samples, or biological replicates.