Fig 2: EPI competes with ATPIF1 on binding to CV (see also Fig EV2) A–G Binding of purified recombinant ATPIF1‐GFP to CV. Mitochondrial membranes were disrupted by freeze–thaw and exposed to ATPIF1‐GFP at different ratios of mitochondrial (mt) protein/ATPIF1 purified protein. Binding was assessed in BNGE by quantifying ATPIF1‐GFP fluorescence. (A) BNGE assessed for ATPIF1‐GFP binding (top); CV hydrolytic activity (middle) and protein loading (bottom). (B) Quantification of in gel CV–ATP hydrolytic activity shown in panel (A). On the left‐hand side are the levels of CV–bound ATPIF1 and on the right‐hand side is CV–ATP hydrolytic activity normalized to the levels of fully assembled CV (n = 3). (C, D) Representative acidification rate profile showing the effects on CV–ATP hydrolytic activity (C) and its quantification (D) (n = 4). (E–G) Binding competition assay between ATPIF1‐GFP and EPI to CV. (E) BNGE assessed for ATPIF1‐GFP binding (top); CV hydrolytic activity (middle) and protein loading (bottom). (F, G) Quantification of gels shown in panel (E). (F) Quantification of ATPIF1 bound to monomeric CV under the indicated treatments (n = 3). (G) Quantification of in gel ATP hydrolytic activity under the indicated treatments (n = 3). Note that EPI reduces ATPIF1 binding while maintaining the same inhibitory effect on ATP hydrolysis.H Plot illustrating the lack of correlation of ATPIF1 expression levels with EPI effect in inhibiting maximal CV–ATP hydrolytic activity. Each point represents a biological sample replica. Data information: For each biological replicate, technical replicates were averaged. Data represent average ± SEM. (B, D, G) Two‐way ANOVA followed by Šídák's multiple comparisons test shows statistical differences depicted by P‐value. (F) Unpaired t‐test show statistical differences depicted by P‐value. (H) Simple linear regression including R 2 and P‐value. Source data are available online for this figure.

Fig 3: In mitochondrial disease models, EPI inhibits ATP hydrolysis, increasing cellular ATP content without affecting respiration (see also Fig EV3) A, B PLA of CV and ATPIF1 produces a red fluorescence signal where anti‐ATPIF1 and anti‐ATP5A1 antibodies are localized at a distance < 10 nm of each other. Anti‐TOMM20 (green) was used to reveal mitochondrial architecture. (A) Representative high‐resolution Airyscan confocal images showing control fibroblasts treated with EPI for 24 h. Maximum intensity projection is shown. Scale bars: 20 and 5 μm in zoomed image. Note that PLA reveals the spatial distribution of ATPIF1 binding along the mitochondrial length. (B) Analysis of ATPIF1‐CV complexes occurrence quantified as PLA dots per μm3of mitochondria. Effect of EPI is shown relative to untreated (n = 2 from > 50 replicates per n).C Effect of EPI on CV ATP hydrolytic capacity in cells overexpressing either ATPIF1WT or its continuously active form, ATPIF1H49K. Hydrolytic capacity is measured by the acidification rates and is normalized to maximal oxygen consumption on succinate + rotenone (Max OCR on SR). Effect of EPI is shown as % of control cells, untreated with EPI. Cells were treated for 24 h with 50 nM EPI (n = 4). Note that in cells overexpressing ATPIF1, EPI treatment does not result in further inhibition of ATP hydrolysis.D Effect of EPI on the mitochondrial matrix ATP pool (mtATP) under conditions where mitochondrial respiration is inhibited by Antimycin A (AA). mtATP content measured as fluorescence intensity per mitochondria area and shown as % of control untreated (n = 3). Control fibroblasts were incubated for 30 min with DMSO or EPI in different concentrations of AA. Mitochondrial area was quantified using MTG staining.E mtATP content measured in human fibroblasts from control and mitochondrial disease patients, after 30 min treatments with increasing concentrations of EPI as indicated. Analysis shows the mean intensity of mtATP probe in MTG area normalized per % of correspondent untreated cells (n = 3).F–H Effects of EPI treatment on fibroblasts with a CIII assembly defect (P1:BCS1L) and fibroblast with a CV ATP synthesis defect (ATP6). (F) Total cellular ATP content shown as % of control untreated cells. Cells were treated for 24 h (n = 3). (G) Cell proliferation rates shown as % change from 0 h (untreated) to 72 h after treatment (n = 3). (H) Basal OCR vs ECAR in Control, CIII‐ and CV‐deficient patient fibroblasts treated with and without 50 nM EPI (n = 3). Data information: (A–H) In all cases, for each biological replicate, technical replicates were averaged. Data represent average ± SEM. Two‐way ANOVA followed by Šídák's multiple comparisons test shows statistical differences depicted by P‐value. (G) Unpaired t‐test versus untreated conditions shows statistical differences depicted by P‐value. Source data are available online for this figure.

Fig 4: Screening of EPI effect and inhibition of ATP hydrolysis in mitochondrial disease models (linked to main Fig 4) A Representative SDS‐PAGE blots for ATPIF1, HA, SDHB, ATP5A, and Vinculin showing the overexpression of ATPIF1‐WT and H49K in Control (Ctrl) and CIII‐deficient cells (P1:BCS1L).B Representative confocal images showing mitochondrial localization of overexpressed ATPIF1‐WT and H49K in Ctrl and CIII‐deficient fibroblasts, labeled with anti‐HA (green) antibody and DAPI (nuclei – blue). Maximum intensity projection is shown. Scale bars: 20 μm.C Basal respiration (left) and acidification rates (right) in control fibroblasts before and after the direct injection of the indicated concentrations of Antimycin A (AA) (n = 4).D Basal respiration in control fibroblasts before and after the direct injection of the indicated concentrations of oligomycin.E, F Membrane potential measured using TMRE (average intensity) in Mitotracker Green (mitochondria area) normalized per % of untreated control: (E) control fibroblasts incubated for 30 min with AA plus DMSO, 100 nM EPI or 1 μM Oligo (n = 3); and (F) control fibroblasts incubated for 30 min with AA plus increased concentrations of Oligomycin (Oligo) (n = 3).G Basal respiration (OCR) in control fibroblasts before and after the direct injection of the indicated concentrations of EPI (n = 4).H Total ATP content measured in fibroblasts from patients with different mutations in mitochondrial proteins by luciferase assay and normalized by % of luminescence from control cells (n = 3).I Extracellular lactate levels measured as % of the untreated control from Control (Ctrl), CIII-(P1:BCS1L) and CV‐deficient (ATP6) fibroblasts treated with and without 50 nM EPI (n = 3). Control cells grown in galactose media (GAL) were included as a positive control for low levels of lactate. Data information: In all cases, data represent average ± SEM. Two‐way ANOVA followed by Šídák's multiple comparisons test shows statistical differences depicted by P‐value.

Fig 5: EPI replaces ATPIF1 to block ATP hydrolysis in mitochondrial disease models (linked to main Fig 5) A Representative BNGE blots (top) and quantification (bottom) of CIII2, CIII2 + CIV showing the distribution of CIII complexes and supercomplexes in control and CIII‐deficient cells treated with and without 50 nM EPI for 24 h. Samples were immunoblotted with UQCRC2 antibody. SDHB was used as a loading control.B Representative BNGE blots (top) and quantification (bottom) of CV monomer (CVm) supercomplex in control and CV‐deficient cells treated with and without 50 nM EPI for 24 h. Samples were immunoblotted with ATP5A antibody. SDHB was used as a loading control.C, D (C) Representative BNGE blots and quantification (D) of ATPIF1 relative intensity normalized per amount of CV and represented as % of control untreated cells in fibroblasts (n = 3). SDHB was used as a loading control.E Chart shows PLA dots/μm3 of mitochondria in fibroblasts normalized as % of Ctrl values (n = 2).F Representative confocal images showing fibroblasts treated with EPI for 24 h, labeled with anti‐ATPIF1 and anti‐ATP5A1 (PLA in red) and anti‐TOMM20 (green) antibodies. Maximum intensity projection is shown. Scale bars: 20 and 5 μm. Data information: In all cases, data represent average ± SEM. Two‐way ANOVA followed by Šídák's multiple comparisons test shows statistical differences depicted by P‐value.