Except from the extracted mitochondrial lipids used in this work, we purchased lipids 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (PC), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (PE), bovine brain sphingomyelin (SM) and bovine heart cardiolipin (CL), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoinositol (PI), 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (PS), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(lissamine rhodamine B sulfonyl) (RhPE) and cholesterol (Chol) from Avanti Polar Lipids (Alabaster, AL, United States).

The isolation and separation of the outer (OMM) and inner mitochondrial membranes (IMMs) is based on mitochondrial rupture in the absence of detergents by swelling in the presence of inorganic phosphate, physical rupture by shear stress and the use of discontinuous sucrose gradients. First, intact mitochondria were isolated from fresh homogenized fresh pig heart (Slaughterhouse Toledo, Spain) in 0.25 M sucrose, 10 mM Tris-HCl (pH 7.4) and washed twice in the same medium before fragmentation as previously described (Smith, 1967). The washed mitochondrial pellet (20 mg protein/mL) is homogenized at 0°C–4°C in a Potter-Elvehjem homogenizer with a loose-fitting pestle in 10 mM potassium phosphate (pH 7.4) and then diluted in the same medium to a final concentration of 1 mg protein per ml and allowed to swell for 20 min. In addition, the mitochondria were broken physically by shear stress with a 1 ml syringe equipped with a 25G needle. To collect the total membrane fraction, the suspension was centrifuged for 30 min at 336,896 g (Beckmann MLA80 rotor) to remove the proteins of the mitochondrial intermembrane space and matrix. The pellet, which contains outer and inner membranes, was resuspended in up to 500 μl of 8.5% (w/v) of sucrose dissolved in 10 mM Tris-HCI (pH 7.4) and applied on a discontinuous sucrose density gradient [1.5 ml of 25%, 38%, 52%, and 61.5% of sucrose (w/v) in 10 mM Tris-HCI (pH 7.4) at a final volume of 1.5 ml each] and centrifuged for 12 min at 170,000 g (Beckmann MLA80 rotor). The outer membranes are recovered from the top sucrose layer fractions (between 8.5% and 25% w/v) and IMMs are recovered from the boundary of sucrose layer fractions 38% and 52% (w/v). The fractions containing the OMM and the IMM were eventually diluted with 10 mM Tris-HCl (pH 7.4) until fractions contain ∼15% of sucrose and centrifuged for 30 min at 336,896 g (Beckmann MLA80 rotor). The obtained pellets of the OMM and IMM were resuspended in 100 μl of 10 mM Tris-HCI (pH 7.4), flash frozen and stored until use in −80°C.

Lipid extraction of the isolated OMM and IMM membranes was performed according to Bligh and Dyer (Bligh and Dyer, 1959), where 0.8 volume of membranes at a total protein concentration of 1 mg/ml (OMM or IMM) were mixed with one volume of chloroform and two volumes of methanol [chloroform/methanol/water mixture (1:2:0.8; v/v/v)]. After thoroughly vortexing, one additional volume of chloroform is added to the suspension and again thoroughly vortexed. Next, one volume of distilled water is added [chloroform/methanol/water mixture at (2:2:1.8; v/v/v)] and after mixing the suspension incubated for 30 min at room temperature. To completely phase separate the aqueous from the organic phase, the suspension is centrifuged for 10 min at 12,000 × g (Beckmann F301.5 rotor) at 4°C and recovered the lower organic phase containing the lipids and dried under a stream of nitrogen and stored until further use.

The dried extracted lipid extracts were resuspended in chloroform and the lipid concentration was estimated according to Fiske and Subbarow (King, 1932). 10 μl of lipid extract was pipetted in triplicate into glass tubes and dried at 200°C–215°C. Then the samples were digested for 25 min at 200°C–215°C in the presence of 225 μl of 9N H₂SO₄. During all incubation steps, except for the first drying one, the open glass tubes were sealed with a glass marble to prevent evaporation. After cooling to RT, the samples were supplemented with 75 μl of commercial H₂O₂ and incubated at 200–215°C for additional 30 min to remove any present coloration of the samples. For the colorimetric assay, the colourless samples were cooled down to RT and first supplemented with 1.95 ml of milliQ water and 250 μl of 2.5% (w/v) of ammonium molybdate (VI) tetrahydrate. After thoroughly vortexing, the samples were supplemented sequentially with 250 μl of 10% (w/v) of ascorbic acid and thoroughly vortexed. The samples were then incubated for 7 min at 100°C and after cooling, 200 μl of each sample was transferred to a 96-well plate to read the absorption at 750 nm with a GEMINI XS fluorescence plate reader (Spectramax). To obtain a lipid concentration, the measured absorption of the samples was then referred to the absorption of standard curve ranging from 0 to 0.228 μmoles of KH₂PO₄.

To compare the qualitative profile of inner and outer mitochondrial membranes, a single mono-dimensional thin layer chromatography (TLC) plate was prepared using a combination of solvents able of showing as many of the membrane lipid classes as possible. 10 μg was taken from the OMM and IMM extract and spotted onto a high performance-TLC silica gel plate (HP−TLC; Merck). The plate was first run with chloroform/methanol/acetic acid/water (50:37.5:3.5:2; v/v/v/v) up to the middle of the plates to resolve the phospholipid classes. Then, to concentrate the neutral lipids into a single sharp band, after drying the plate, diethyl ether was run until the solvent front was passed by 1 cm to finally run hexane/ether (80:20; v/v) up to the top of the plates to resolve the neutral lipids present in the sample. Lipid standards have been employed simultaneously to identify each spot of the OMM and IMM extracts. After running the TLC, lipid spots were assessed by charring-densitometry. The plates were dried and dipped into an aqueous solution of 3% (w/v) cupric acetate and 8% (v/v) phosphoric acid. Then, the plates were allowed to dry for 10 min at room temperature and heated for 10–15 min at 180°C to carbonize the lipids (Entezami et al., 1987). The semi-quantification of lipid stains was performed with the ImageJ software package (NIH) (Schneider et al., 2012) and the percentage of each lipid spot was calculated relative to the sum of all spots. Note that the percentage of cholesterol was estimated using the same densitometry analysis. As the total intensity of each spot is proportional to amount of material in each lipid spot, and assuming a similar density for all lipid species, the percentages given in our estimates correspond to % w.

The equilibrium surface tension of aqueous solutions of OMM and IMM lipid extracts at different phosphorous concentrations was measured at 22°C using a paper Wilhelmy plate. The standard deviation was ±0.2 mN/m.

Compression isotherms were obtained using a computer controlled Langmuir balance (611 model from NIMA, UK). The surface pressure was measured using a paper Wilhelmy plate. In order to prevent for surface perturbations induced by the air convection streams and to avoid undesirable dust contamination, the whole set-up is covered by a transparent Plexiglas case. Before each measurement, the subphase surface is cleaned by sweeping and suction. Small aliquots (10 μl typically) of the chloroform lipid solutions were spread on the aqueous surface with a Hamilton syringe. All experiments were carried out at 22°C. Compression isotherms were recorded upon compression of a diluted monolayer at a constant compression rate of 10 cm²/min. The surface pressure is plotted against the mean molecular area, as obtained by dividing the total surface area by the numerical amount of phosphorous spread on the surface.

The OMM and the IMM lipid extracts were fluorescently labelled with Rhodamine-PE (RhPE; 0.1% mol) and dried with a speedvac concentrator at room temperature. The film was resuspended in phosphorous buffered saline (PBS) to lipid final concentration of 1 mg/ml. The lipid suspension was sonicated with a tip sonicator (Vibra-Cell 75115 Ultrasonic Liquid Processor) for 30 min in 5 min on-off cycles (40% power) to form unilamellar lipsomes. Supported lipid bilayers (SLBs) were prepared in custom-made observation chambers (200 μl) on a thoroughly cleaned coverslip (Menzel, 60 mm × 60 mm). Prior to liposome spreading, coverslips were incubated with 100 μl of 10% (w/v) methoxysilane (3-[methoxy (polyethyleneoxy)9-12]-propyltrimethoxysilane, Geles) for 30 min and at room temperature to improve the wettability of the glass surface (Halliwell and Cass, 2001). After removing the excess silane, the reaction chamber was carefully washed with PBS and liposomes (60 μM final concentration) were incubated for 30 min in PBS supplemented with CaCl₂ (20 mM final concentration). Finally, SLBs were washed with PBS to remove CaCl₂ and non-spread lipid material. For microscopy, the observation chamber was finally filled with 200 μl of PBS.

Confocal images were acquired using a Nikon Ti-E inverted microscope equipped with a Nikon C2 scanning confocal module, a Nikon Plan Apo 100X NA 1.45 oil immersion objective and filter cubes. Images were captured with Nikon NIS-Elements software package.

Standard Fluorescence recovery after photobleaching (FRAP) measurements on SLBs were performed with a Nikon Ti-E inverted microscope as described for CLSM. Random zones on SLBs were photobleached and the recovery of fluorescence intensity at the bleached zone was monitored and corrected for image alignment and photobleaching with ImageJ software package. The fluorescence recovery curves were fitted to (Yguerabide et al., 1982): f ( t ) =   f ( 0 ) + f ( ∞ ) · ( t / t 1 / 2 ) 1 + ( t / t 1 / 2 ) (1) where f ( t ) is the time normalized fluorescence intensity F ( t ) / F o , F o is the fluorescence intensity just before photobleaching, f ( 0 ) is the normalized fluorescence intensity just after photobleaching, f ( ∞ ) is the normalized maximum fluorescence intensity and t 1 / 2 is the half-time of the fluorescence recovery. As t 1 / 2 depends on the percent bleach of the sample, the diffusion coefficient, D, was then calculated applying the correction factor β as follows (Axelrod et al., 1976). D = w 2 4 · t 1 / 2 β (2) where β accounts for both the beam shape and the bleaching extent (Yguerabide et al., 1982) and w is the bleached radius.

Giant vesicles were prepared using the standard electroformation protocol (Mathivet et al., 1996). The fabrication chamber was composed of two 1-mm spaced conductor indium tin oxide (ITO)-coated slides (7.5 cm² × 2.5 cm²; 15–25 Ω/sq surface resistivity; Sigma). Briefly, giant unilamellar vesicles (GUVs) were prepared by transferring a volume of 15–20 μl of purified IMM or OMM in chloroform (1 mM phosphorous) onto each ITO slide. Samples are dried at room temperature and rehydrated in sucrose solution (300 mM) and the electrodes were connected to an AC power supply (10 Hz, 1.1 V; Agilent) for 2 h at room temperature and 50°C for IMM and OMM lipid extracts, respectively. For fluorescence microscopy, OMM and IMM lipids were supplemented with 0.2% mol RhPE.

To measure the bending modulus with the micromanipulation device, a micrometer-sized GUV (20 μm typically) is pulled out by a cylindrical micropipette in the aspiration mode (Longo and Ly, 2007). In a suction experiment, the pressure difference Δ p between the vesicle interior and the pipette determines the membrane tension, σ , as σ = Δ p R p 2 ( 1 − R p R V ) ; where R p is the micropipette radius and R V the vesicle radius. For a vesicle under suction at a given Δ p , a protrusion length, L p , is aspirated inside the pipette and the relative excess area α = A − A 0 A 0 = Δ A A 0 is given by α = Δ L p 2 R p [ ( R p R V ) 2 − ( R p R V ) 3 ] ; where Δ L p is the variation of the protrusion length from an increase in suction pressure compared to the protrusion length measured at an initial low tension state, σ 0 . For low tensions, the entropic regime dominates (Evans and Rawicz, 1990) and the bending modulus is obtained from the linear fitting of the Canham-Helfrich equation which connects the relative excess area α and the membrane tension σ through (Evans and Rawicz, 1990): ln ( σ σ 0 ) ≈ 8 π κ α k B T (3) where κ is the bending modulus of the vesicle and k B T is the thermal energy.

Mitochondrial membranes were extracted from pig heart mitochondria by density gradient centrifugation after mechanical breakdown (Smith, 1967). Then, the lipids were extracted from the isolated OMM and IMM through liquid-liquid separation (Bligh and Dyer, 1959). To characterize the extracted lipid mixtures and to check for their lipid compositions we performed standard mono-dimensional thin layer chromatography (TLC), run with different solvents to specifically separate the polar from the non-polar lipids as described in experimental procedures. To identify the individual lipid bands we have simultaneously spotted a series of lipid standards representing the expected lipids present in OMM and IMM (Horvath and Daum, 2013). We can clearly distinguish a difference in lipid pattern for the OMM and IMM (Figure 1A). As expected, the OMM lipid extract is mainly composed of phosphatidylcholine (PC), phosphatidylethanolamine (PE), and sphingomyelin (SM), whereas the IMM contains cardiolipin (CL). Interestingly, bacterial cell membranes and the inner mitochondrial membrane have a relatively high proportion of CL—between 10% and 20% of the total lipid content of these membranes is made up of CL (Horvath and Daum, 2013), a lipid that is practically absent from other biological membranes (van Meer et al., 2008). However, no PE was detected in the IMM fraction. As PE and CL are known to run similarly in more polar solvents, a multiple-step TLC was run with three different solvents (see methods) and the PE and CL spots were clearly resolved in the IMM fraction (Figure 1B). The semi-quantification of lipid stains from three independent experiments indicated an enrichment of cholesterol, CL, PE and PC in the IMM fraction (10%, 21%, 34%, and 27%, respectively). Additionally, minor components were also detected: lysophosphatidylcholine (LPC), sphingomyelin (SM), phosphatidylserine (PS), and phosphatidylinositol (PI) (0.7%, 1.6%, 2.4%, 2.2% of the total phospholipid content, respectively). As for OMM (Figure 1C), the multiple-step TLC indicated an enrichment of PC, PE, SM, and cholesterol (35%, 41%, 6.3%, and 16.7%, respectively). The total lipid composition of OMM was completed with CL (1%). Although a complete analysis of fatty-acyl chain profile was not performed, the absence (1%) and the presence (up to 21%) of CL in the OMM and IMM fractions, respectively, were taken as an indication of a clean separation (See Table 1 for a comparative lipid composition between IMM and OMM).

The Gibbs adsorption isotherms at 20°C of OMM and IMM lipid extracts at the air-water interface were determined by measuring the surface tension of aqueous solutions with different lipid concentrations (Figure 2). At low concentration and for both OMM and IMM lipid extracts, the surface tension monotonically decreased as the bulk concentration increased, corresponding to the monomeric adsorption of lipids to the interface. At higher concentrations, the surface tension reached a plateau value, which was similar for both lipid extracts ( γ min ≈ 46   m N / m ) . However, the minimal surface tension was reached progressively for IMM whereas the OMM lipid extract behaved as an insoluble surfactant, i.e., once the minimal value of surface tension was reached the surface concentration was no longer modified. The bulk concentration leading to this transition provides the critical micelle concentration or CMC, where the bulk surfactants aggregate into micelles. The CMC was found to be 23 nM for OMM and 50 nM for the aqueous IMM lipid solution (Figure 2). Assuming that 1) the activity coefficient of both OMM and IMM extracts are close to unity for the studied concentration range and 2) the lipid mixture behaves as a single and non-ionic surfactant; the variation of the surface tension, γ , with the bulk lipid concentration could be fitted to the Gibbs-Langmuir equation of state: γ = γ 0 − R T Γ ∞ ln ( 1 + K c ) (4) where γ 0 is the surface tension of the bare air/water interface, Γ ∞ is the maximum surface concentration, K is the adsorption equilibrium constant, c is the lipid concentration, R is the gas constant and T is the absolute temperature. From the fitting curves, we obtained the adsorption equilibrium constant K , K O M M = 0.42   ±   0.3   n M − 1 and K I M M = 0.15   ± 0.04   n M − 1 . Also, the maximum surface concentration was obtained from the fit, with Γ ∞ ≈ ( 3.0 ±   0.7 ) × 10 − 6 and ( 2.4   ±   0.3 ) × 10 − 6   m o l / m 2 for OMM and IMM lipids, respectively. The mean molecular area can be calculated through A = 1 Γ ∞ N A , with A ≈ 55 and 69 Å 2 for OMM and IMM lipids, respectively. As the mean molecular area is expected to be inversely related to the average lipid packing within the monolayer, higher values of A are indicative for more expanded membranes, so that OMM lipids exhibit stronger intermolecular interactions than IMM lipids.

The physical state of a monolayer can be also studied by monitoring its surface pressure upon compression at a particular temperature. Compression approaches the molecules to each other promoting transitions from different phases, which are characterized by particular orientation and packing of molecules on the interface. The experimental pressure-area (π-A) compression isotherms of OMM and IMM lipid extracts were obtained in a Langmuir trough at T = 25°C (Figure 3). Remarkably, IMM lipids exhibited larger areas per molecule than OMM lipids, i.e., OMM behaved as a more packed monolayer compatible with the presence of cholesterol in the lipid composition (Figure 1). At lower compression, the surface pressure remained close to 0 mN/m below ≈ 90 and ≈ 140   Å 2 for OMM and IMM, respectively. At further compression, the lateral pressure of IMM monolayers increased with a monotonic expanded-like behaviour without phase transition plateaux. Then, IMM reached the collapse regime at a mean molecular area A ≈ 80   Å 2 , characterized by a constant pressure π c ≈ 40   m N / m . The compression isotherm obtained for IMM is similar to those measured for POPC (see Figure 3) or E. coli lipids (Lopez-Montero et al., 2008), (Lopez-Montero et al., 2010). Unlike IMM or POPC, the expanded regime of OMM monolayers transitioned to a condensed-like phase at an area close to 65   Å 2 and at surface pressures typical of the bilayer packing state (Marsh, 1996). The π-A monolayer profile at the condensed-like phase was characterized by a higher slope and entered the collapse regime at a mean molecular area of ≈ 55   Å 2 . The monolayer collapse pressure of OMM was higher than IMM and achieved at π c ≈ 43   m N / m .

A useful parameter for lipid monolayer characterization is the compression modulus, C − 1 , which takes higher values for more condensed monolayers. The equilibrium compressibility modulus (Figure 3, inset), as defined by the change in monolayer pressure caused by an infinitesimal change in the molecular area, can be calculated from the numerical derivative of the experimental π-A isotherms (see Methods). Similar to POPC and E. coli monolayers, the compressibility modulus of IMM lipids displays a maximum ( C − 1 ≈ 80   m N / m ) at the bilayer lipid packing state (30–35 mN/m) and drops to zero when reaching the collapse pressure. For OMM lipids, the monolayer was highly compressible ( C − 1 ≈ 50   m N / m ) in the more expanded-like regime ( π < 10   m N / m ) but the compressibility modulus increased to a maximum value ( C − 1 ≈ 150   m N / m ) just before reaching the pseudoplateau ( π ≈ 35   m N / m ). This value is compatible with compacted monolayers (Wydro, 2012). Again, the maximal rigidity appeared at surface pressures corresponding to the biologically relevant surface packing (Marsh, 1996) and further compression reduced C − 1 down to smaller values compatible with the collapsed arrangement.

To further characterize the dynamic properties of both lipid extracts we investigated the effect of the lipid composition on the lateral mobility of a fluorescently labelled lipid (RhPE) embedded in lipid membranes built as supported bilayers (SLBs). This configuration allows for a quantitative measurement of the diffusion coefficients of fluorescent probes through Fluorescence Recovery After Photobleaching (FRAP). A zone in the lipid bilayers was selectively photobleached by a high laser power scanning. The recovery of fluorescence intensity at the bleached zone occurs through Brownian diffusion of surrounding unbleached probes (Figure 4A). The measurement of a single characteristic diffusion time enables to obtain the diffusion coefficient of the probe in the embedding bilayer (see Methods) (Axelrod et al., 1976; Yguerabide et al., 1982). Figure 4A shows typical FRAP curves for RhPE embedded in SLBs made of POPC, OMM, and IMM lipid extracts. The labelled phospholipid RhPE followed unrestricted Brownian motion with diffusion coefficients D P O P C = ( 1.8 ±   0.2 )   μ m 2 / s , D O M M = ( 0.8   ±   0.2 )   μ m 2 / s and D I M M = ( 1.0 ±   0.2 )   μ m 2 / s when embedded in POPC, OMM and IMM lipid bilayers, respectively (Figure 4B). The diffusion coefficient measured for RhPE in mitochondrial lipid bilayers was in agreement with previous studies using FRAP (Merzlyakov et al., 2006) or FCS experiments (Bockmann et al., 2003). However, the immobile fraction was higher (up to 30%) for lipid extracts than for POPC (<10%) The presence of lipid heterogeneities and defects in SLBs made of mitochondrial lipid extracts might explain the high immobile fraction.

Using a classic electroformation protocol (Mathivet et al., 1996) we succeeded in preparing GUVs from mitochondrial lipid extracts (Figure 5). Standard settings (0.5 V/mm and 10 Hz) were used for GUV electroformation. However, the electroformation chamber and the sucrose solution was kept at 50°C during GUV formation for the OMM lipid extract. The high experimental temperature is compatible with an OMM lipid composition enriched with cholesterol and/or high-melting temperature phospholipids, such as SM (see Figure 1C; Table 1). Fluid lipid membranes are required for GUV fabrication to favour lipid mixing during electroformation. We named the vesicles IMM-GUVs and OMM-GUVs for GUVs made of inner and outer mitochondrial lipids, respectively. Both OMM- and IMM-GUVs were almost spherical with a variable size ranging from 5 to 60 μ m . Unlike IMM-GUVs (2%), 90% of OMM-GUVs exhibited micron-scale liquid immiscibility (Veatch and Keller, 2003) as visualized by the selective partition of RhPE into liquid disordered (l _d ) phases. In contrast, coexisting liquid ordered (l _o ) phases excluded the fluorescent probe and were imaged as dark membrane regions (Figure 5, inset). To further characterize IMM- and OMM-GUVs, the size distribution of a population of giant vesicles was determined. For that, several z-stacks of the different observation fields of sedimented vesicles were taken. For each vesicle, the true diameter is assumed when the size of the vesicle reached its maximal size under confocal observation. The size distribution of GUVs was represented in histograms obtained from a population of tens of vesicles showing a large population of IMM-GUVs and OMM-GUVs with a typical size above 10   μ m . The histograms displayed a similar size distribution obtained for simple GUV model made of POPC (Figure 5).

GUVs are membrane models that allow their mechanical characterization under external forces. The mechanical parameters define the compliance of the membrane to be shaped when exposed to different modes of deformation. In particular, the bending modulus provides the basic energy scale for bending deformations. The bending modulus of OMM- and IMM-GUVs was measured by micropipette aspiration (Evans and Rawicz, 1990) (see Methods and Figure 6A) from the fitting of Eq. 3 to the experimental data (Figure 6B).

A first set of experiments provided the bending modulus of IMM lipids (Figure 6C), with κ I M M = 5.6 ±   2.0   k B T , (N = 14). This value agrees with previously reported measurements for cardiolipin-containing GUVs made of E. coli lipid extracts and measured with flickering spectroscopy (Almendro-Vedia et al., 2017). For OMM-GUVs, the absolute values of bending moduli for the l _d and l _o phases cannot be obtained from phase-separated vesicles (Tian et al., 2007). For simplicity, we performed the experiments on non phase-separated OMM-GUVs, obtaining a higher bending modulus of κ O M M = 7.8 ±   2.9   k B T , (N = 13). This value, although underestimated, is in agreement with previously reported measurements for cholesterol- and SM-containing fluid membranes (Gracia et al., 2010). For comparison, POPC GUVs were characterized by a bending modulus of κ P O P C = 10 ± 3   k B T in agreement with (Niggemann et al., 1995).