Cardiolipins

The total mitochondrial mass was determined using 10-N-Nonyl acridine orange (NAO, Invitrogen, Waltham, MA, USA), a dye that binds to cardiolipin present specifically on the mitochondrial membrane [63 (link)]. Briefly, wt and cdh1Δ cells were grown to mid-log phase in YPGal medium and incubated in culture medium containing 10 μM NAO for 30 min. Fluorescence intensity measured using the BD Accuri C6 flow cytometer. Data were analysed with FlowJo v10 software version.

SLC25A20 structural models have been inserted in a bilayer phospholipid membrane mimicking the inner mitochondrial membrane (IMM), using the web server CHARMM-GUI (http://www.charmm-gui.org, accessed on 1 December 2022) [62 (link)]. The membrane composition was based on the IMM model published by the CHARMM-GUI team available at the CHARMM-GUI Archive (https://charmm-gui.org/?doc=archive&lib=biomembrane, accessed on 1 December 2022) [63 (link)]. Different concentrations and lipid tail composition are used to better represent the inner and outer IMM leaflets. In this model membrane, phosphatidylcholine is the most represented phospholipid species, followed by phosphatidylethanolamine and cardiolipin, the latter being more abundant in the inner layer. Water molecules, from the TIP3P model, were added on both sides of the membrane, forming two layers each 22.5 Å thick. The total system charge was neutralized by adding NaCl ions, reaching a physiological concentration of 0.15 M. The CHARMM36m force field [64 (link)] and the AMBER22 package [65 ] were used to perform the MD simulations of the assembled systems (~100,000 atoms, see Supplementary Table S6 for representative compositions), following the CHARMM-GUI protocol. First, an energy minimization procedure, involving 2500 steps of steepest descent and 2500 steps of conjugate gradient, was performed. Positional restraints were applied on the protein residues (10 kcal mol⁻¹ Å⁻²) and on the membrane (2.5 kcal mol⁻¹ Å⁻²). The resulting minimized systems have then been simulated using the canonical NVT ensemble, reaching a final temperature of 310.15 K. Thereafter, isothermal-isobaric NPT ensemble simulations have been performed to equilibrate the pressure to 1 bar. During the thermalization and equilibration phases, the positional restraints have been gradually reduced. The equilibrated system, without restraints, was properly simulated using the NPT ensemble for a total of 1 μs. Each system was simulated in replica. The Langevin thermostat has been used for both NVT and NPT ensembles, while the Monte Carlo barostat with a semiisotropic pressure scaling has been used for pressure control [66 (link)]. All of the MD simulations have been performed in a periodic boundary system. For the long range non-bonded interactions, the Particle Mesh Ewald method [67 (link)] and a 12 Å cut-off, with a force switching region at 10 Å, were used. The MD simulations were performed with a time step of 2 fs, apart from the apoSLC25A20 simulation for which the hydrogen mass repartitioning (HMR) method [68 (link)] was used with a 4 fs time step.