**2. Results**

#### *2.1. ANT1-Mediated Substrate Transport*

To evaluate whether recombinant murine ANT1 was correctly refolded in proteoliposomes, we performed ADP/ATP transport measurements using proteoliposomes initially filled with radioactively labeled 3H-ATP [26]. After adding ADP to the bulk solution, we measured the release of 3H-ATP with time (Supplementary Figure S2). The determined ADP/ATP exchange rate depends on ANT1 content (kANT = 5.53 ± 0.74 mmol/min/g) and corresponds well to the reported results for ANT reconstituted into liposomes (Supplementary Table S1) [26–29].

#### *2.2. Basal Proton Leak*

We further investigated the controversially discussed ANT1 involvement in the basal leak [20,21]. For this, we formed planar bilayer lipid membranes from proteoliposomes reconstituted with recombinant ANT1 [26]. Figure 1 demonstrates that the total specific conductances, G m and G0, of bilayer membranes made from DOPC, DOPE, and cardiolipin were similar in the presence and absence of ANT1 (G m =10.0 ± 2.0 nS/cm<sup>2</sup> and G0 =8.4

± 2.3 nS/cm2) if no purine nucleotides (PN) were added. The addition of ATP and ADP on both sides of the membrane led to a substantial G m increase, which was directly proportional to the applied membrane potentials, reaching G m = 39.5 ± 4.9 nS/cm<sup>2</sup> at 190 mV (Figure 1, Supplementary Figure S3 and Supplementary Table S2. This increase vanished after adding the specific inhibitor of ADP/ATP transport—CATR (Figure 1) and can be explained by the electrogenic shift due to ATP/ADP exchange by ANT1. This experiment showed that ANT1 has no measurable impact on the proton leak without FAs.

**Figure 1. ANT1 does not contribute to the basal proton leak.** Total membrane conductance (G m) was measured at different membrane potentials (ΔΦ) and membrane compositions (s. legend). Planar bilayer membranes were made of 45:45:10 mol % PC:PE:CL. Lipid concentration was 1.5 mg/(mL of buffer solution). Protein concentration measured by BCA assay was 4 μg/(mg of lipid). Buffer contained 50 mM Na2SO4, 10 mM Tris, 10 mM MES and 0.6 mM EGTA at pH = 7.34 and T = 306 K. ADP, ATP and CATR were added at concentrations 2 mM, 2 mM and 100 μM. Lines represent the least square regression fit of an exponential function to the data. Data are the mean ± SD of at least three independent experiments.

#### *2.3. ANT1-Mediated Proton Transport in the Presence of FA*

The addition of polyunsaturated arachidonic acid (AA) to the membrane in the absence of ANT1 led to a potential-dependent increase in G m. It confirms FA's importance as weak uncouplers, especially at high potentials (G m was one order of magnitude higher at 190 mV) relevant for mitochondrial membranes [30]. The reconstitution of ANT1 in the membrane increased G m in the presence of AA 4-fold (G m ANT, AA/G m AA) (Figure 2a). At 190 mV G m ANT, AA, G m AA and G0 were equal to 1750 ± 220 nS/cm2, 440 ± 135 nS/cm<sup>2</sup> and 20.4 ± 3.4 nS/cm2, respectively (Supplementary Figure S4 and Supplementary Table S2). Notably, ANT1-mediated G m depended on the structure of FAs. It increased with the elongation of FA chain length in order palmitic (PA, 16:0) → arachidic (ArA, 20:0) acid and was the highest by unsaturated AA (20:4) (Figure 2b).

**Figure 2. Fatty acids are required to activate ANT1-mediated proton transport.** (**a**) Total membrane conductance (G m) of lipid bilayers in the presence of AA (gray squares), ANT1 (red triangles), ANT1 and AA (dark red diamonds) and in the absence of AA and ANT1 (white circles) at different membrane potentials (ΔΦ m). Lines represent the least square regression fit of an exponential function to the data. (**b**) Dependence of total membrane conductance (G m) on fatty acid chain length and unsaturation in the presence (red) and absence (gray) of ANT1. PA, SA, ArA, LA, and AA indicate palmitic, stearic, arachidic, linoleic, and arachidonic acids. In all measurements, planar bilayer membranes were made of 45:45:10 mol % PC:PE:CL reconstituted with 15 mol % FA, except indicated otherwise. Lipid concentration was 1.5 mg/(mL of buffer solution). Protein concentration measured by BCA assay was 4 μg/(mg of lipid). The buffer solution contained 50 mM Na2SO4, 10 mM Tris, 10 mM MES and 0.6 mM EGTA at pH = 7.34 and T = 306 K. Data are the mean ± SD of at least three independent experiments.

#### *2.4. Proton Turnover Number of ANT1*

To determine the H+ turnover number of ANT, we recorded current-voltage characteristics in the presence and absence of a transmembrane pH gradient (Figure 3, insert) [31].

To estimate a protein to lipid ratio, we measured the number of fluorescently-labeled ANT per liposome using fluorescence correlation spectroscopy (FCS) [32] (s. Methods and Supplementary Figure S5). By comparing the number of the fluorescent particles in proteoliposomes before (NANT, none = 1.60 ± 0.01) and after (NANT, SDS = 13.83 ± 0.04) the addition of 2 % (*v*/*v*) SDS, and assuming one ANT protein per detergent micelle after micellization, we calculated 8.67 ± 0.74 ANT molecules per liposome. The protein to lipid ratio estimated according to Equation (4) was 1:12,000.

From the potential shift and proton/lipid ratio, we then estimated that ANT has a turnover rate of 14.6 ± 2.5 H+/s (Figure 3), being similar to those of uncoupling proteins (Supplementary Table S3) [6,7,9,33–35].

#### *2.5. Inhibition of ANT1-Mediated Proton Transport*

Specific inhibitors of nucleotide transport lock ANT either in its cytosolic-opened c-side (CATR) or in its matrix-opened m-side (BA) [15] and inhibit both ADP/ATP exchange and FA-mediated proton leak. The comparison of CATR and BA effect on G m (Figure 4a) showed that inhibition by CATR was more effective than by BA. It is displayed by the EC50 values of 18.9 ± 1.8 μM for CATR and 32.3 ± 11.4 μM for BA, respectively (Figure 4b and Supplementary Table S4). Maximum inhibition values (Imax = 64.2 ± 2.8% and Imax = 44.3 ± 5.7% in the presence of CATR or BA) indicate that ANT conformation in the bilayer is approximately 60% in the c-state and 40% in the m-state in our system (Figure 4c and Supplementary Table S4).

**Figure 3. The proton turnover number of ANT is similar to UCPs.** Representative current-voltage recordings of lipid bilayer membranes reconstituted with ANT1 in the presence (squares) and absence (circles) of ΔpH = 1.0 across the membrane. Lines represent a linear fit to the data. Planar bilayer membranes were made of 45:45:10 mol % PC:PE:CL reconstituted with 15 mol % AA. Buffer contained 50 mM Na2SO4, 10 mM Tris, 10 mM MES and 0.6 mM EGTA at pH = 7.34 and T = 306 K. Lipid concentration was 1.5 mg/(mL of buffer solution). Protein concentration measured by BCA assay was 4 μg/(mg of lipid). Insert: Experimental setup of the measurements to establish a transmembrane pH gradient.

To investigate the interdependence of proton transport, activated by FA, and ADP/ATP transport, we measured G m of ANT1-containing lipid bilayers reconstituted with AA in the presence and absence of purine nucleotides (PN). Adenine nucleotides inhibited H+ transport much more effectively than guanosine nucleotides (Figure 5a). The EC50 values correlated well with the known narrow substrate specificity of ANT (Figure 5b and Supplementary Table S4) [36]. ATP and ADP fully inhibited G m, whereas all other PN decreased G m by 50% (Figure 5c and Supplementary Table S4).

**Figure 5. FA activated proton leak is preferably inhibited by ADP and ATP and maintains the substrate specificity of ANT.** (**a**) Membrane conductance of lipid bilayers reconstituted with AA and ANT1 in the presence of different purine nucleotides. Lines are a least square regression fit of a sigmoidal function to the data. (**b**) EC50 and (**c**) maximum inhibition values as fit function parameters in (**a**). Values for GMP were dropped due to the low effect. For experimental conditions, see Figure 4. Data are the mean ± SD of at least three independent experiments.

#### *2.6. Analysis of the ANT*´*s Surface Electrostatic Potential using Molecular Dynamic Simulations*

The high similarity of the ANT activation pattern to those of UCP1, UCP2 and UCP3 [6,7,35] leads to the hypothesis that the H+ transport can be explained by the FA cycling mechanism. We analyzed the ANT's surface electrostatic potential in the DOPC bilayers to test whether a possible FA translocation pathway may be localized at the lipidprotein interface. The calculation revealed a large positively charged patch (Figure 6a) that might facilitate FA anion's sliding alongside the protein. The ATP binding significantly decreased the positive electrostatic potential (Figure 6b) due to its strong screening at the bottom of the cavity [37]. The existence of such a patch would explain the inhibition of H+ transport by ATP observed in electrophysiological experiments. GTP has less effect on the positive electrostatic potential (Figure 6c) because of its different orientation in the cavity (Figure 6d) and its "weaker interaction with the hydrophobic pocket that binds the adenine moiety" [36].

**Figure 6.** The purine nucleotides ATP and GTP differently modulate the electrostatic surface potential of ANT upon binding. (**<sup>a</sup>**–**<sup>c</sup>**) Electrostatic potential (Δ of ANT1 in the absence (**a**) and presence of bound ATP (**b**) and GTP (**c**) calculated by molecular dynamics simulations. The isosurface of the potential of 0.9 V is shown with the wireframe. (**d**) Different average binding location of ATP (blue wireframe) and GTP (red wireframe) in ANT1.
