L-α-phosphatidylcholine (L-α-PC), extracted from dry egg yolk was purchased from Sigma-Aldrich (St. Louis, MO). This mixed lipid extract contains more than 60% (w/w) PC; the remaining 40% consists mostly of phosphatidylethanolamine (PE) and small amounts of other lipids. Cardiolipin [18:1; 1′,3′-bis [1,2-dioleoyl-snglycero-3-phospho]-glycerol (sodium salt)] was obtained from Avanti Polar Lipids (Alabaster, AL), octyl glucopyranoside (OG) was from BioShop Canada Inc. (Burlington, ON), and octyltetraoxyethylene (C₈E₄) was from Bachem (Bubendorf, Switzerland). The fluorescent probe dye, SPQ (6-methoxy-N-(3-sulfopropyl) quinolinum; 99% purity) was obtained from Biotium Inc. (Fremont, California). Carboxyatractyloside (potassium salt) was purchased from Cayman Chemical (Michigan, USA). All other chemicals were purchased from Sigma (St. Louis, MO) unless otherwise indicated.

cDNAs encoding proteins of interest were cloned into the pET26b (+) expression vector, such that recombinant versions of bovine ANT1 and human P_iT-B, UCP2 and UCP4 (UniProt accession codes: ANT1 - P02722, P_iT-B - Q00325-2, UCP2 - P55851, and UCP4 - O95847), were produced with a pelB leader sequence, followed by a His₆-tag, fused at the N-terminus. Expression of recombinant fusion proteins was achieved in E. coli BL21 (DE3) or BL21 CodonPlus (DE3)-RIPL cells using a modified autoinduction method as previously described (Hoang et al., 2013; Tabefam et al., 2024; Studier, 2005). Briefly, the bacterial cells were grown overnight in 500 mL of autoinduction culture media (1% tryptone, 0.5% yeast extract, 1 mM MgSO₄, 0.5% glycerol, 0.05% glucose, 0.2% lactose, 25 mM (NH₄)₂SO₄, 50 mM KH₂PO₄, 50 mM Na₂HPO₄) at room temperature. After 22 h, the culture was centrifuged at 4000 g for 1 h at 4 °C using a JLA 10.500 rotor (Beckman Coulter) and the cell pellets were resuspended in lysis buffer [500 mM NaCl, 5 mM MgCl₂, 20 mM Tris-HCl pH 8.0, one-quarter of a tablet of enediaminetetraacetic acid (EDTA)-free protease inhibitor (Roche Applied Science), one small scoop of DNase, and lysozyme (BioShop Canada Inc.) at a final concentration of 0.1 mg/mL]. A high-pressure cell disruptor (Constant Systems Limited, Daventry, U.K.), operating at 20 kPsi, was used to lyse the cells. The cell lysate was centrifuged at 20,000 × g for 20 min in a JA 25.5 rotor (Beckman Coulter) to remove intact cells, inclusion bodies or aggregated proteins. The cleared supernatant was then ultracentrifuged at 50,000 × g (MLA 80 rotor, Beckman Coulter) for 1 h to obtain the bacterial membranes in the pellet fraction.

Purification of proteins in their monomeric form from the bacterial membrane fraction was achieved using immobilized metal ion affinity chromatography (IMAC) via the procedure described in detail previously (Tabefam et al., 2024). Briefly, binding buffer containing 10 mM imidazole, 1% (w/v) lauryldimethylamine oxide (LDAO detergent), 500 mM NaCl, 1 mM Tris (hydroxypropyl) phosphine (THP), and 20 mM Tris-HCl, pH 8.0 was used to solubilize the bacterial membrane pellets containing the His-tagged recombinant protein and the suspension was incubated overnight at 4 °C on a rotating mixer. The solution was then incubated while mixing with 2 mL of nickel-nitrilotriacetic acid (Ni-NTA) resin (Bio-Rad, Waltham, Massachusetts) that had been equilibrated and suspended in binding buffer in a column for 1 h. The resin was allowed to settle, the flow-through was collected, and the resin was washed with 8 mL of binding buffer. Detergent exchange (from 1% LDAO to 1% OG) and further washing was accomplished by applying 5 mL of wash buffer containing 30 mM imidazole, 1% (w/v) OG, 500 mM NaCl, 1 mM THP, and 20 mM Tris-HCl, pH 8.0. Stepwise elution was achieved by applying elution buffer (1% (w/v) OG, 500 mM NaCl, 1 mM THP, and 20 mM Tris-HCl, pH 8.0) containing increasing concentrations of imidazole [100 mM (2 mL), 250 mM (2 mL), and 400 mM (6 mL, collected as six 1 mL fractions)]. Econo-Pac 10DG desalting spin columns (Bio-Rad, Hercules, California) were used to achieve removal by buffer exchange of 400 mM imidazole from the elution fractions containing the recombinant protein. For CD spectroscopic measurements the proteins were desalted into CD buffer (20 mM Tris-HCl, pH 8.0, 50 mM NaCl, 1% OG and 1% glycerol). For proton transport assays the desalting buffer was composed of 1.5X internal buffer [120 mM TEA₂SO₄, 45 mM TEATES (TEA: tetraethylammonium, TES: N- [tris(hydroxymethyl)methyl]-2-aminoethanesulfonic acid anion), 1.5 mM EDTA, pH 7.2] supplemented with 1% OG. The identity of the purified proteins was confirmed by Western blotting and MS analyses as described previously (Tabefam et al., 2024); furthermore, MS was used to confirm that the pelB leader sequence had been removed, presumably by the endogenous protein trafficking system in E. coli (Tabefam et al., 2024). Purity, and concentration of the purified proteins was assessed using semi-native sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and modified Lowry assay (Peterson, 1977), respectively. Freshly desalted proteins were used immediately for functional and structural analyses.

For CD spectroscopic and proton transport measurements, native-like monomeric forms of purified recombinant His-tagged ANT1, P_iT, UCP2 and UCP4 were reconstituted into lipid vesicles to form proteoliposomes using a previously described detergent-mediated reconstitution procedure (Tabefam et al., 2024; Jabůrek and Garlid, 2003), with minor modifications. Liposomes with two different lipid compositions were used in this study. All liposomes contained L-α-PC from egg yolk extract (∼60% PC) (EYPC), either in the absence or presence of 18:1 cardiolipin (CL) added to a final concentration of 2.5 mol%. Briefly, a thin layer of lipid film, dissolved in methanol/chloroform (1:3) was formed in a round-bottomed flask under vacuum overnight. The lipid was then rehydrated in the desired reconstitution buffer to form multilamellar liposomes. For CD spectroscopic analysis, the reconstitution buffer contained 20 mM Tris-HCl, pH 8.0, 50 mM NaCl, 1% glycerol. For proton transport experiments, 1X internal buffer supplemented with the fluorescent SPQ probe at a final concentration of 3 mM was used at pH 7.2. The multilamellar liposomes were then solubilized in C₈E₄ to a final detergent/phospholipid ratio of 2.5 (w/w). Purified, desalted monomeric proteins were then added to the mixed lipid/detergent micelles. Unilamellar proteoliposomes were formed spontaneously following the removal of C₈E₄ detergent using SM-2 Biobeads (Bio-Rad). For proton transport assays, external SPQ probe was removed using a coarse Sephadex G25-300 (GE Healthcare) spin column. For CD spectroscopy studies, the final protein:phospholipid molar ratio was on the order of 1:1000 and in proton transport experiments this ratio was 1:10,000. In each experiment, protein-free liposomes were prepared in parallel as a control.

Semi-native SDS-PAGE was used to assess the oligomeric state of purified recombinant proteins before and after reconstitution into liposomes. In this technique, the amount of sodium dodecyl sulfate (SDS) is significantly reduced as compared to more traditional denaturing SDS-PAGE to provide the “semi-native” conditions (Voulhoux et al., 2003). Specifically, SDS is omitted from sample loading buffer, and is only included in the gel and running buffer at a concentration of 0.1% w/v (as compared to 1% w/v in fully denaturing conditions). In addition, protein samples were not heated and only incubated at room temperature for 5–10 min prior to loading on the gel. Gels were run at 110 V, stained with 0.2% (w/v) Coomassie Brilliant Blue R-250 in methanol: acetic acid: water (45:10:45 by volume) for 30–60 min and destained for 2 h.

The size and homogeneity of liposomes and proteoliposomes were determined by dynamic light scattering (DLS) using a Zetasizer Nano ZS (Malvern Instruments, Worcestershire, U.K.). The average radii of prepared proteoliposomes for the proton transport assays and CD measurements were about 80 nm. On average, the radii of the blank liposomes were ∼10% larger than the proteoliposomes. The results were the average of 3-5 measurements.

Far-UV CD measurements were performed at 25 °C, at 1 nm resolution, on an AVIV 215 spectropolarimeter (Aviv Biomedical, Lakewood, New Jersey). Quartz cells with 0.1 cm path length were used for measuring the CD spectra of proteins in OG detergent or reconstituted into liposomes. Ellipticities were converted to mean residue ellipticity, [θ]. All individual CD spectra were an average of at least two independent preparations each measured 3 times, for a total of 6 measurements. The concentrations of proteins were 8–10 µM for samples in 1% OG and ∼1 µM in proteoliposomes.

Proton efflux mediated by proteins reconstituted in lipid vesicles was measured by the anion-sensitive fluorescence quenching method, in which anions quench the fluorescent dye SPQ by a dynamic collision mechanism, as described previously (Jabůrek and Garlid, 2003; Jabůrek and Garlid, 2003; Hoang et al., 2012). The rate of proton efflux is measured indirectly through the quenching of SPQ by TES⁻. In a low pH environment, TES (pK_a ∼7.4) is fully protonated and does not affect the fluorescence of SPQ. However, when a proton is lost at high pH, the anionic TES (TES⁻) quenches SPQ fluorescence (Orosz and Garlid, 1993). These steady-state fluorescence measurements were performed in a Cary Eclipse spectrophotometer (Varian, Palo Alto, CA) with a bandwidth slit of 5 nm and a scan speed of 600 nm/min. Excitation and emission of the SPQ fluorescence signal were at 347 and 442 nm, respectively. The fluorescence scans were performed at 25 °C. Each proton transport analysis was an average of at least 5 independent measurements. For each proton transport assay, 40 µL of (proteo-) liposomes was added to 1.96 mL of external buffer [K₂SO₄ (80 mM), KTES (30 mM), EDTA (1 mM), pH 7.2]. The internal buffer of the liposomes consisted of TEA₂SO₄ (80 mM), TEATES (30 mM), and EDTA (1 mM), pH 7.2. The osmotic pressure was also kept equal across the membrane at the beginning of the measurements. Palmitic acid (as palmitate, PA) or oleic acid (as oleate, OA) were dissolved in 70% ethanol (with sonication, as necessary) and were added to a final concentration of 100 µM to activate proton efflux by the reconstituted carrier protein(s) in proteoliposomes (Hoang et al., 2012). The addition of the K⁺ ionophore valinomycin (val) facilitates the influx of K⁺, which in turn triggers fatty acid-dependent proton efflux by carrier proteins to offset the osmotic discrepancy in membrane potential. Proton efflux results in deprotonation of TES buffer (to form TES⁻), which quenches SPQ fluorescence. The rate of decrease in the SPQ fluorescence signal was correlated to the rate of fatty acid-activated proton transport by the protein under study. The latter was measured by monitoring the decrease in SPQ’s fluorescence signal intensity within the first 30s after the addition of valinomycin (Supplementary Figure S1) (Hoang et al., 2013). All proton transport data in liposomes were corrected for basal nonspecific leak by subtracting the proton efflux measured for blank (protein free) liposomes. Moreover, the proton transport data were calibrated for the SPQ fluorescence response, internal volume of proteoliposomes, and protein concentration (Jabůrek and Garlid, 2003). The concentration of protein in proteoliposomes was determined using a modified Lowry assay, as described previously (Peterson, 1977; Hoang et al., 2012). No significant proton transport/leakage was observed for the proteoliposomes in the absence of valinomycin. The influence of purine nucleotides ATP and GTP and the ANT-specific inhibitor CATR on proton transport was assessed by incubating proteoliposomes with 500 µM ATP, 500 µM GTP or 5 µM CATR for 2.5 min prior to addition of valinomycin.

All data analyses were performed using SAS version 9.4 (SAS Institute, 2012). Residuals were visually assessed to confirm homogeneity of variance, and normality was tested using the Shapiro-Wilk and Kolmogorov-Smirnov tests. Proton efflux rates were statistically compared among 1:1 binary protein combinations, inhibition treatments, stoichiometric ratios, and between fatty acid activators and cardiolipin presence or absence with general linear models (e.g., ANOVA, t-tests) and general linear mixed models. General linear mixed models were used when there were random effects, which included experiment replicate (all tests), and in some cases cardiolipin presence/absence, type of fatty acid activator, and inhibition treatment. Interactions between fixed factor variables were initially included, and non-significant terms (including interactions) were removed from the models using backward selection. Significant main (fixed) effects with more than two treatments were further analyzed using post hoc Tukey’s tests to evaluate pairwise differences.

Subcloning the cDNAs encoding the four proteins of interest (UCP2, UCP4, ANT1 and P_iT) into the pET26b (+) expression vector that incorporates the pelB leader sequence followed by a six-histidine affinity tag (His₆) in-frame with the N-termini of the proteins, combined with the application of a modified autoinduction expression method, resulted in targeting of the proteins to the inner membrane of E. coli and enabled purification of the proteins using immobilized metal affinity chromatography. The pelB leader sequence is a signal peptide that typically targets secreted proteins to the periplasmic space where it is cleaved by signal peptidase (Tabefam et al., 2024). However, the hydrophobic nature of transmembrane domains of the carrier proteins used here hinders complete translocation, resulting in insertion into the membrane due to their hydrophobic interactions with the bacterial membrane (Hoang et al., 2013). The membrane fractions were then isolated using differential velocity centrifugation. Mild detergents were used to extract the proteins from the membrane and resulted in copurification of proteins with closely associated membrane lipids. This lipid shield protects proteins from potential denaturing interactions with the solubilizing detergent and maintains the proteins in a relatively folded/native state (Hoang et al., 2013; Tabefam et al., 2024). The His₆-tagged proteins were purified using nickel-containing affinity columns as monomers in the presence of OG detergent. OG was selected based on our previous study on conformational and thermal stability analysis of UCP1 in different detergents which revealed that the overall integrity of the protein was most stable when purified and stored in 1% OG micelles (Hoang et al., 2013). Purity of the proteins was confirmed by semi-native SDS-PAGE (Figure 2, lanes 1–4). The identity of each protein was confirmed by Western blotting, using monoclonal antibodies for UCP4 and P_iT and a polyclonal antibody for ANT, and MS spectrometry as described previously (Tabefam et al., 2024).

As shown in Figure 2, all proteins existed in their monomeric forms when purified from E. coli membrane extracts in the presence of mild OG detergent. As shown previously, ANT, PiT and UCP4 migrate similarly on semi-native PAGE, despite their modest molecular weight differences, which can be detected when the proteins are compared using fully-denaturing SDS-PAGE (Tabefam et al., 2024). To investigate the behaviour of each protein when reconstituted into liposome membranes, purified proteins were initially reconstituted individually in liposomes made using an egg yolk lipid extract in either the absence or presence of added cardiolipin (EYPC ±2.5% CL) for structural and proton transport functional analyses. Egg yolk extract was chosen over other lipid systems as it contains a majority of PC (∼60%) along with a mixture of other phospholipids (such as PE) that are found in the MIM in similar proportions (Horvath and Daum, 2013). The effect of added CL was examined as it is a characteristic lipid found exclusively in the MIM (Daum, 1985). The semi-native electrophoretic analysis revealed interesting findings regarding the behavior of each protein when reconstituted into liposomes. The high molecular weight bands observed in Figure 2 (lanes 5, 6, 9, and 10) show that UCP2 and UCP4 tend to self-associate into tetramers when reconstituted into liposomes regardless of the presence or absence of added CL. Conversely, reconstituted P_iT, in both the absence and presence of CL, remained mainly in its monomeric form, consistent with its state in the original OG detergent (Figure 2, lanes 8 and 12). The behavior of ANT1, on the other hand, varied depending on the lipid composition; when reconstituted in liposomes composed of EYPC in the absence of added CL, ANT1 remained predominantly in its monomeric state, with a weak indication of self-association into dimeric forms. However, when reconstituted in the presence of 2.5% CL, the protein appears to remain exclusively monomeric as the faint shadow of the dimeric form disappeared (Figure 2, compare lanes 7 and 11). It is assumed that all proteins had closely-associated native lipids when extracted and purified from E. coli membranes, which may have included an unknown amount of CL; therefore proteins reconstituted into liposomes without additional CL likely contained some residual CL, and those reconstituted in the presence of 2.5% CL may have had a final concentration of CL that was slightly higher than 2.5%.

The current study focused on assessing whether binary and ternary combinations of these proteins, in various stoichiometric ratios, would associate and form hetero-oligomers upon reconstitution in lipid vesicles. As illustrated in Figure 3 (lanes 1–6), regardless of lipid composition, all 1:1 stoichiometric binary combinations of carriers that were tested (UCP2:UCP4, UCP2:ANT1, UCP2:P_iT, UCP4:ANT1, UCP4:P_iT, and ANT1:P_iT) formed tetramers. Interestingly, binary combinations including ANT1 and P_iT, which form mainly monomers when reconstituted on their own (Figure 2, lanes 7-8 and 11-12, respectively), whether reconstituted with each other or with UCP2 or UCP4, also led to the formation of tetramers (Figure 3, lanes 2–6). That ANT1 and P_iT do not form homo-tetramers, and that tetramers are only observed when these proteins are co-reconstituted in binary combinations with other carrier proteins, indicates that the tetramers that are formed from binary combinations include both proteins (i.e., they are heterotetramers). Reconstitution of three proteins together in 1:1:1 stoichiometric combinations of proteins (UCP2:ANT1:P_iT and UCP4:ANT1:P_iT) also resulted in the formation of tetramers (Figure 3, lanes 7 and 8).

Circular dichroism (CD) spectroscopy is a sensitive technique for studying protein conformations and monitoring structural changes in proteins. In this study, CD was utilized to characterize and compare the conformations of UCP4, ANT1 and P_iT proteins in OG and liposomes of different lipid compositions (EYPC ±2.5% CL) (Figure 4). CD spectra of the proteins in the far UV region (below 250 nm), dominated by the electronic transitions of the amide (peptide) bond chromophore, give information about protein secondary structures. Irrespective of the environment (detergent/lipid membrane), the CD spectra of all proteins exhibited distinctive spectral features of α-helical backbone structures [negative maxima at 208 nm and 222 nm, and a positive maximum around 193 nm (not shown), corresponding to π → π* (193 and 208 nm) and n → π* (222 nm) transitions of the peptide bond] (Greenfield, 2006). It is worth mentioning that light scattering and flattening effects (Hoang et al., 2012), that limit the CD studies of membrane proteins, did not distort the spectra in the wavelength range shown in Figure 4.

UCP4 tetramers in EYPC liposomes (Figure 4a, inset) displayed a notable reduction in the far-UV CD signal intensity compared to the monomeric protein in OG detergent, indicating a reduced degree of helicity (Figure 4a). However, inclusion of 2.5% CL in the lipid membrane led to an enhancement in helicity and in the tetramers comparable to the helicity observed for the monomers in OG, albeit with differences in the CD spectra shape (Figure 4a). These results suggest a significant conformational change of the protein from OG (monomers) to EYPC vesicles (more densely-packed monomeric units in tetramers) and subsequently from EYPC to EYPC +2.5% CL vesicles (more relaxed or less densely packed monomeric units in tetramers), as evidenced by a notable decrease in negative ellipticity and a reversal of the relative intensity of minima (θ₂₀₈/θ₂₂₂) from 1.18 (in OG) to 0.98 (in EYPC +2.5% CL liposomes) and 0.6 (in EYPC liposomes). These findings underline the importance of the lipid environment in the folding and structural stabilization of UCP4.

The differences between tetrameric UCP4 in liposomes and its monomeric form in OG are further evident from the shape of the CD spectra of the two preparations. In particular, when reconstituted in EYPC liposomes, the CD spectrum of UCP4 tetramers displayed a distinct shoulder-like π → π* parallel transition band around 208 nm, replacing the very intense minimum observed for monomers in OG. Additionally, a more intense n → π* transition band emerged at 222 nm (Figure 4a). This shift is reflected in the θ₂₀₈/θ₂₂₂ ellipticity ratio, which transitioned from 1.18 for the UCP4 monomer in OG to 0.60 upon reconstitution in EYPC liposomes (Figure 4a). A ellipticity ratio below 1.0 is consistent with previous findings regarding the packing density in helical bundle motifs and self-associated oligomers of human UCPs (Hoang et al., 2015). The CD spectra for UCP4 tetramers in this study align well with spectra reported in previous studies (Hoang et al., 2015).

Compared with UCP4, the transition of the θ₂₀₈/θ₂₂₂ ellipticity ratio for P_iT from 1.27 in the original state (OG) to 0.87 after reconstitution in EYPC vesicles, and further decrease to 0.77 in the presence of 2.5% CL, together with apparent changes in the shape of the CD spectra, indicate a conformational shift of the monomeric form when reconstituted into the lipid environment (Figure 4b). These alterations suggest a potential change in the compactness of the helices within each monomeric unit in the lipid environment.

For ANT1, the θ₂₀₈/θ₂₂₂ ellipticity ratio changed from 1.11 in OG detergent to 0.81 when reconstituted in EYPC vesicles, and 1.07 in the EYPC +2.5% CL lipid environment. These observations correlate with the semi-native gel results (Figure 4c, inset), which indicate that reconstitution in EYPC vesicles led to some degree of oligomerization (evidenced by a faint dimeric band). However, ANT1 remained monomeric when reconstituted in liposomes that included 2.5% CL in the lipid environment similar to what was observed in OG, but with a reduced helical intensity (Figure 4c). The reduced helical intensity of the reconstituted protein is indicative of denser packing of helices in monomeric units.

A comprehensive study on UCP2, which included CD analysis, was recently published by our group, and suggested that the increased ellipticity observed for tetrameric UCP2 compared to UCP2 in OG was likely due to a higher degree of helicity after reconstitution in liposomes (Ardalan et al., 2021a; Ardalan et al., 2021b). In summary, based on our experimental conditions the differences between the CD spectra of UCP4, P_iT and ANT1 show that these proteins acquire different molecular forms and different degrees of helical packing after reconstitution in vesicles of different lipid compositions. For example, in EYPC liposomes, the lower θ₂₀₈/θ₂₂₂ ratio of UCP4 compared to ANT1 and P_iT (0.60 vs. 0.81 and 0.87) is consistent with the semi-native SDS-PAGE data suggesting that UCP4 forms tetramers when reconstituted into liposomes, whereas ANT1 and P_iT remain mainly monomeric.

To study the potential formation and function of heteromeric carrier protein complexes, the rate of proton efflux was compared for individually reconstituted proteins and combinations of the carrier proteins that were reconstituted at different stoichiometric ratios. Proton efflux was measured using a fluorescence quenching based assay, as described previously (Hoang et al., 2012). The proton efflux rate (PER) is expressed as µmol proton efflux/min/mg protein (for instance, PER = 16.3 ± 0.3) (Supplementary Figure S1) for proteins reconstituted in liposomes.

In several experiments, proton efflux was quantified and compared between experimental systems with and without added cardiolipin [EYPC ±2.5% CL]. Data were also collected and compared in the absence or presence of two different free fatty acids (FAs), palmitic acid (as sodium palmitate, PA) or oleic acid (as sodium oleate, OA). PER for all proteins reconstituted into liposomes in the absence of added FAs was comparable to blank liposomes (data not shown). The inhibitory effects of purine nucleotides (ATP and GTP) and the ANT c-state-specific substrate inhibitor, CATR, on selected proteins and/or selected combinations of proteins were also examined.

As has been shown previously for UCP2 (Brand et al., 2005; Lee et al., 2015), ANT1, P_iT-B and UCP4 formed α-helical conformations in OG detergent and retained their overall helical conformations when reconstituted into lipid environments (Figures 2, 4). UCP2 and UCP4 spontaneously self-associated to form tetramers in lipid membranes (Figures 2, 9a), as was observed previously (Ardalan et al., 2021a; Tabefam et al., 2024). These homotetramers were functional in transporting protons across the lipid bilayers - a general characteristic of all UCPs (Brand et al., 2005; Hoang et al., 2012; Orosz and Garlid, 1993; Kreiter et al., 2021). For both UCPs, proton transport activity was higher in the presence of added CL (Figures 5a,b). CD analysis indicated that the inclusion of CL in the membrane may lead to a more relaxed helical conformation of the monomeric subunits of UCP4 (and UCP2) within the homotetramers (Figure 4a). This suggests that conformationally flexible UCP tetramers could facilitate proton translocation (Figures 4a, 5). On the other hand, neither P_iT nor ANT1 (both of which have comparable three-dimensional structures to UCPs) formed tetramers when reconstituted into lipid membranes (Figures 2, 9a). This finding is consistent with previous studies (Krämer, 1996; Ardalan et al., 2021a; Tabefam et al., 2024). P_iT monomers exhibited a basal level of proton transport, which increased significantly in the presence of CL. Unlike UCP4, this increase in the proton transport rate, in the presence of CL, could not be attributed to a more relaxed helical structural arrangement in a tetramer. A minor conformational reorganization likely plays a key role in increasing the proton transport activity of dominantly monomeric P_iT (Figures 2, 4b, 5). Reconstituted ANT1 in EYPC lipid membranes also exhibited proton transport activity both as a monomer (with CL) and as a mixture of monomers and dimers (without added CL) (Figure 2). The presence of CL in the membrane could modify the conformation of ANT1 monomers (CD spectra in Figure 4c), thereby enhancing proton transport function. This observation is consistent with previous studies suggesting that CL acts as a “grease” at the dynamic interface of monomeric units (Kreiter et al., 2021; Beck et al., 2007). However, one of these earlier reports indicated that CL stabilizes interactions of ANT with itself (self-association) and possibly other partners (Claypool, 2009), whereas the other proposes that CL could allow for the close packing of ANT monomers by preventing interactions with other proteins (Ruprecht et al., 2014). A more general conclusion has been suggested in a recent review suggesting that CL is required to accommodate the close association of proteins (including self-association) in the highly crowded and protein-rich environment of the MIM (Schlame, 2021).

For all 4 MC proteins tested, proton transport occurred only in the presence of long-chain free FAs (Figures 5a–d). Four mechanistic models have been suggested to explain the protonophoric activity of the monomeric form of these carriers in the presence of FAs: (i) the FA cycling model (Engstová et al., 2001; Kreiter et al., 2021; Pohl et al., 2025); (ii) the buffering/cofactor model (Winkler and Klingenberg, 1994); (iii) the long-chain FA shuttling model (Fedorenko et al., 2012); and (iv) the induced transition fit (ITF) model (Bertholet and Kirichok, 2017). In the FA cycling model, protons are transported from the IMS into the matrix by a quick diffusion and flip-flop of protonated FA across the MIM. Meanwhile, the passage of the anionic form of the FA (FA⁻) back to the IMS is supported by MCs, such as ANT, P_iT and UCPs (Engstová et al., 2001; Kreiter et al., 2021; Beck et al., 2007; Pohl et al., 2025; Garlid et al., 1996; Garlid et al., 1998; Berardi and Chou, 2014; Kreiter et al., 2023). The buffering/cofactor model, originally proposed for UCP1 (Winkler and Klingenberg, 1994), suggests that the deprotonated FA (FA⁻) is bound to the translocon-like channel of the carrier, facilitating proton translocation via their carboxylate group to buffering amino acids of the carrier through the center of the protein. The shuttling model proposes that hydrophobic tails of long-chain FAs are lodged within the carrier, establishing hydrophobic interactions with the protein, and the carboxylate head group then moves through the protein’s cavity, shuttling back and forth between the IMS and matrix. In this model, the head group of the FA becomes protonated in the IMS and deprotonated in the matrix while the FA hydrophobic tail remains bound to the protein (Fedorenko et al., 2012; Bertholet and Kirichok, 2017). A modified version of the shuttling model is called the ITF model (proposed for UCPs) in which FAs induce alternating conformational change in the protein. In this model, as the polar carboxylate head group of a FA moves across the MIM, the UCPs can alternate between the two conformational states in which the protein cavity is either open to the IMS side (c-state) or the matrix side (m-state) of the membrane (Winkler and Klingenberg, 1994; Bertholet and Kirichok, 2017). This so-called alternating access mechanism has been suggested to be a common mechanism of transport in response to substrate binding for all MC protein family members (Ruprecht and Kunji, 2020).

The mechanism by which long-chain FAs facilitate UCP2- and UCP4-mediated proton transport can be qualitatively explained by a recently proposed biphasic two-state model for proton transport by tetrameric UCPs (Ardalan et al., 2021b). In this model, a tetramer is composed of two stable dimeric units, with the monomers within each dimeric unit being in a similar conformational state that is opposite to that of both monomers in the second dimeric unit. That is, the model proposes that the monomeric subunits are functionally and conformationally correlated within a dimer. In the context of the biphasic two-state model, the dependence of UCP2 and UCP4 proton transport activity on long-chain FAs can be generally explained by a combination of translocon-like channel and ITF mechanisms. Accordingly, the monomeric subunits of UCP tetramers act as a H⁺ uniporter and the long hydrocarbon chain becomes anchored to the hydrophobic region of each monomer (four FAs are required per tetramer) by hydrophobic interactions, acting as a cofactor, and the carboxylate head group can move back and forth across the protein as it changes from the c-to m-state and vice versa. This oscillating movement of the FA within the protein opening allows protons to be transported by the exchange between UCP-bound FAs and the buffering amino acids of the protein in their path to the matrix side via a translocon-like mechanism, similar to the proton-buffering model described previously (Krauss et al., 2005). It should be also considered that only one of the dimeric units of the tetrameric protein can transport protons at a time, in accordance with the proton gradient across the membrane. In this model the monomeric subunits mutually impact one another and induce interconnected conformational changes. One dimeric unit starts in the c-state and changes to the m-state as the protonated head groups of FAs move from the IMS toward the matrix. Concurrently, the second dimeric unit starts in the m-state and changes to the c-state upon the movement of FA head groups toward the IMS. The stronger hydrophobic interactions provided by the longer, unsaturated hydrocarbon chain of OA enable it to remain bound and participate in subsequent transport cycles (Ardalan et al., 2021b).

The low level of long-chain FA-induced H⁺ translocation mediated by P_iT and ANT1 might be better explained by the FA cycling model (Engstová et al., 2001; Kreiter et al., 2021; Pohl et al., 2025; Kreiter et al., 2023). This model assumes that the negatively charged FA anion (FA⁻) is attracted to the positively charged surface of the protein and translocates along the lipid-protein interface. In this process, the FA carboxylic group remains in close contact with the positively charged amino acids on the protein surface. Longer and more unsaturated FAs diffuse faster through the lipid bilayer (Kreiter et al., 2021; Pohl et al., 2025; Kreiter et al., 2023).

The inhibitory effects of ATP and GTP on proton transport of UCP2, UCP4, ANT1 and P_iT carriers differed. It is possible that the differences can be attributed to both the different modes of proton transport and the distinct substrates that these proteins interact with. In ANT1, the nucleotide substrate binding site has a high specificity for adenine nucleotides (ATP in this study) and its c-state inhibitor, CATR, whereas the binding sites of UCP2 and UCP4 accept both ATP and GTP, with a higher affinity for GTP and no affinity for CATR. Conversely, proton transport by P_iT is not affected by either ATP or GTP (Figures 5, 8; Supplementary Tables S1, S2, S6).

The inhibitory effects of ATP and CATR on ANT1 can be explained by previous molecular dynamics simulation studies, which describe a positive electrostatic potential patch at the protein-lipid interface (Ardalan et al., 2021b). When ATP and/or CATR (both of which carry a charge of −4) bind to the positively charged substrate binding site of ANT1, the resulting local conformational change reduces this electrostatic potential. Consequently, the attraction of anionic FAs (as proton transport activators) to the ANT1 surface is reduced, leading to reduced (or inhibited) proton transport. However, when GTP binds to ANT1, it adopts a different orientation that has less impact on ANT1’s electrostatic potential (Kreiter et al., 2021). The surface electric charge of ANT1 is directly correlated with the inhibition potency of purine nucleotides and is supported by observed competition between purine nucleotides and FAs for the binding site in ANT1 (Kreiter et al., 2021; Kreiter et al., 2023).

Our results for UCP2 and UCP4 are consistent with previous studies showing that purine nucleotides bind to the proteins in their c-state, near the center of the central cavity, occluding the UCP translocation pathway (Ardalan et al., 2021b; Bertholet and Kirichok, 2017). This binding likely obstructs (or reduces the rate of) the movement of FA headgroups between the IMS and the matrix, thereby preventing the protein’s conformational change or locking it in an intermediate state (Ardalan et al., 2021b; Jones et al., 2023). Moreover, inhibition by ATP and GTP could also be the consequence of the neutralization of the positive charges of arginine triplet residues (e.g., R88, R185, and R279 in UCP2) (Bertholet and Kirichok, 2017) through interactions with the negatively charged phosphate groups. These arginine triplet residues are suggested to be the substrate contact points in MCs and have been suggested to attract FAs toward the protein cavity to initiate proton transport (Ardalan et al., 2021b). Additionally, previous reports have suggested that ATP and long-chain FA anions might compete to bind near or within the proton translocation pathway of UCP1 (Fedorenko et al., 2012; Bertholet and Kirichok, 2017). It was also proposed that long-chain FAs could potentially counteract the inhibitory effects of ATP on UCP1 (Fedorenko et al., 2012). Given the substantial structural differences between FA anions and ATP, their binding sites on UCP1 are unlikely to be identical. However, these sites may partially overlap or be located in close proximity, leading to electrostatic repulsion and competition between the two negatively charged species (Bertholet and Kirichok, 2017). Alternatively, binding either of ATP or FA may discourage the binding of the other through negative allosteric effects. Our findings support this hypothesis for UCP2, as ATP was less effective as an inhibitor of proton transport in the presence of OA than in the presence of PA. This was also the case for the inhibition of UCP2 by GTP. However, in the case of UCP4, ATP and GTP were more effective at inhibiting proton transport in the presence of the longer chain OA, which could be the result of stronger interaction of ATP and GTP with UCP4.

Only bands corresponding to tetrameric complexes were visible on semi-native gels when UCP4, ANT1 and P_iT were reconstituted into lipid membranes together in different stoichiometric ratios, indicating that no proteins remained monomeric. Since ANT1 and P_iT are not capable of forming homotetramers, all combinations of these MC proteins must have resulted in the formation of heterotetramers (Figures 3, 9b). The current study reveals, for the first time, the possibility of the cooperative interactions and joint functional roles of these MC proteins in mitochondrial proton transport regulation. In our experimental system, the basal proton transport rates for UCP2 and UCP4 homotetramers were determined to be 3.5 ± 0.1 and 2.6 ± 0.06, respectively (Figures 5a, b; Supplementary Table S1). However, when combined in a 1:1 stoichiometric ratio to form heterotetramers the resulting proton transport rate (3.4 ± 0.3) was close to the rate of UCP2 on its own, suggesting that their combined function does not result in an additive effect. This indicates that, although UCP2 and UCP4 may cooperate when combined, their proton transport activity is not significantly enhanced or diminished compared to UCP2 alone (Figures 6, 9b; Supplementary Table S3). The similar inhibition by ATP (52.9%) (Figure 6; Supplementary Table S3) for the 1:1 (UCP2:UCP4) combination, which closely aligns with the inhibition observed for individual UCP2 (55.1%) (Figure 5a; Supplementary Table S1), further supports the idea that the proton transport activity of the heterotetramers remains comparable to that of the individual UCP2.

The significantly enhanced proton transport rate for the 1:1 stoichiometric combination of ANT1 with UCP2 (13.90 ± 0.70) (Figures 6, 9b; Supplementary Table S3), compared to individual ANT1 (0.50 ± 0.30) and UCP2 (3.50 ± 0.10) proteins (Figures 5, 9a; Supplementary Table S1), is suggestive of a strong cooperativity between the proteins in the heterotetramers. This cooperativity could possibly be explained by ANT’s role in mitochondrial energy exchange complementing UCP2’s transport capabilities.

For the UCP2:P_iT combination (1:1 stoichiometry), the transport rate (7.60 ± 0.40) (Figures 6, 9b; Supplementary Table S3) was significantly higher (approximately two times higher) than UCP2 on its own (3.5 ± 0.1), and much higher than P_iT on its own (0.46 ± 0.30) (Figures 5, 9a; Supplementary Table S1). That ATP did not significantly inhibit the UCP2:P_iT combination (30.3% inhibition) compared to other combinations, may indicate a lower affinity for ATP of this hetero-oligomeric complex (Figure 6; Supplementary Table S3).

Proton transport by the combination of UCP4 and ANT1 (1:1) (7.80 ± 0.40) (Figures 6, 9b; Supplementary Table S3) was significantly (three times) higher than UCP4 (2.57 ± 0.06) on its own and much higher than ANT1 alone (0.50 ± 0.30) (Figures 5, 9a; Supplementary Table S1), suggesting positive cooperativity between these 2 MCs. The significant inhibition by ATP of proton transport by the 1:1 combination of UCP4 and ANT1 (75.6%) (Figure 6; Supplementary Table S3) indicates that the UCP4-ANT1 heterotetramer may be regulated similarly to UCP2-ANT1 heterotetramers (79.1% inhibition by ATP) (Figure 6; Supplementary Table S3). That heterotetramers including UCPs and ANT1 respond similarly to ATP supports the idea that ANT1 cooperatively enhances the proton transport activity of UCPs, and that ATP modulates this cooperative interaction effectively. However, the proton transport rate of heterotetramers formed by the 1:1 combination of UCP4 and P_iT (2.10 ± 0.10) (Figures 6, 9b; Supplementary Table S3) was not significantly enhanced compared to UCP4 alone, showing that the positive cooperativity observed for other combinations is not a universal effect of any combination of MCs, but is specific to certain combinations (UCP2-ANT1 and UCP4-ANT1 in this instance).

Finally, the significantly higher proton transport rate by ANT1 combined with P_iT (6.0 ± 0.4), illustrates the possibility of strong cooperative interaction among MCs (Figures 6, 9b; Supplementary Table S3) other than UCPs. The significant inhibition by ATP (55.8%) reflects the cooperative yet possibly distinct regulatory mechanism of the ANT1- P_iT pair (Figure 6; Supplementary Table S3). This observation suggests that while neither ANT1 nor P_iT are effective proton transporters on their own, their combination can significantly enhance their proton transport activity and that this function can be modulated by ATP.

The existence of only heterotetrameric complexes of different MCs is suggestive of a tendency for heterotypic protein-protein interactions and regulatory mechanisms in mitochondrial proton transport.

A simple mechanism of proton transport by monomers within heterotetrameric complexes can also be explained by the two-state model (Ardalan et al., 2021b). The organization of monomeric subunits within heterotetramers varies according to the stoichiometric ratio of reconstituted proteins. Assuming the likelihood that homo- and heterotypic interactions are similar, when a 1:1 ratio of 2 MCs is used, the formation of two different homodimeric units is highly possible. In a heterotetramer with a 3:1 or 1:3 ratio of two different MC proteins, the resulting complexes are likely to contain one homodimer and one heterodimer. Furthermore, the proton transport activity is significantly different for individual proteins and protein combinations. Tetramers formed from 3:1 (and 1:3) combinations may consist of a mixture of homo- and heterodimers.

Analysis of 3:1 and 1:3 stoichiometric combinations of MCs is performed on the UCP4:ANT1 system. The 1:1 combination of the lesser-studied UCP4 can be compared to the 1:1 combination of UCP2 and ANT1 (Figures 5, 8). This equimolar UCP4-ANT1 tetramer most likely comprises a combination of a homodimer of UCP4 and a homodimer of ANT1. When reconstituted in a 3:1 ratio (UCP4:ANT1), the most likely (or predicted to be most abundant) structure of the heterotetramers is a UCP4 homodimer combined with a UCP4-ANT1 heterodimer. Conversely, in a 1:3 ratio (UCP4: ANT1), the most probable heterotetramer would contain a homodimer of ANT and a UCP4-ANT1 heterodimer (Figure 10). According to the biphasic transport model, proton transport occurs through an alternating mechanism involving dimeric units. During one phase, the UCP4 homodimers are in the c-state and accept protons while the ANT1 homodimers are in the m-state. In the next phase, the UCP4 homodimers switch to the m-state and release protons to the matrix as the ANT1 homodimers switch to the c-state. Although there is a possibility that the dimeric ANT1 units can also cause FA⁻ cycling, it can be assumed that most of the proton transport occurs through the UCP4 homodimers, with the ANT1 dimers enhancing this function (Figure 10). The importance of ANT in the 1:1 combination was further verified using CATR, an inhibitor of ANT’s c-state (Pebay-Peyroula et al., 2003). CATR tightly blocks ANT1 from switching between conformations, thus preventing alternating access; when part of a heterotetramer with UCP, preventing conformational switching of ANT1 between the m- and c-state prevents the entire tetrameric complex from engaging in alternating access. Consequently, the transport rate significantly decreases, approaching the activity level of individual UCP4 function (Figures 8–10; Supplementary Tables S7, S8). We also studied the inhibitory effect of CATR on co-reconstituted UCP2 and ANT1 (1:1 ratio). Comparable to the UCP4-ANT1 pair, blocking ANT1 in the UCP2:ANT1 heterotetramer with CATR prevents alternating access of the entire tetramer, leading to a drastic decrease in proton transport.

In summary, we propose that the proton transport mechanism by heterotetrameric complexes of MCs is influenced by the stoichiometric ratio of the proteins involved. Different combinations, particularly the 1:1 and 3:1 (and 1:3) ratios of UCP4 and ANT1, result in specific structural arrangements of homodimers and heterodimers within the tetramers, leading to distinct functional properties compared to individual proteins. For example, in the 1:1 (UCP4:ANT1) combination the higher proton transport rates can be explained by the cooperative role of ANT1 dimers in enhancing the transport functionality of UCP4 dimers. The critical role of ANT1 in this process is underscored by the significant decrease in transport rate when conformational switching by ANT is inhibited in UCP4:ANT1 and UCP2:ANT1 (1:1) combinations, implying the complex interplay between these MC proteins in facilitating efficient proton transport.

In this study, we explored the functional dynamics of a group of MC proteins involved in regulation of the proton motive force and ATP synthesis in mitochondria (two UCPs, ANT1, and P_iT) in an effort to elucidate their individual and combined roles in proton transport across the MIM. Our findings provide compelling evidence that these proteins not only function independently but can also form heterotetrameric complexes that exhibit distinct and enhanced proton transport activities as compared to individual proteins.

We also demonstrated that UCP2 and UCP4 form functional tetramers that facilitate proton transport, with their activity modulated by CL and long-chain FAs. Concomitantly, ANT1 and P_iT-B, while not forming tetramers on their own, showed a similar dependence on interactions with long-chain FAs for proton transport, supporting the idea that these interactions are crucial for their function. Our experiments with the 1:1 combination of UCP2-ANT1 and UCP4-ANT1 revealed a cooperative effect that led to increased proton transport rates compared to the individual proteins. Such cooperative behavior may be a general feature of UCP interactions with ANT1. Functional synergy appears to be fine-regulated by inhibitors ATP, GTP, and CATR, suggesting a broader relevance to mitochondrial bioenergetics. ANT1 could therefore have a critical role in regulating the proton transport function of UCPs in the heterotetrameric complexes. Generally, the importance of stoichiometry of proteins in determining the functional behaviour of heterotetramers is emphasized.

Finally, it is important to recognize that our biophysical experiments may provide valuable insights but do not fully capture the complexities of in vivo environments. The physiological relevance and behavior of these molecular interactions in living systems remain to be clarified. Validation of these findings in vivo will help to confirm the physiological relevance of our observations and will also reveal differences in behavior and function under more complex cellular conditions. In the scope of mitochondrial molecular physiology, we hope that the emphasis of this study on intricate functional relationships between these MC proteins that affect efficiency of mitochondrial proton transport can open new avenues for understanding mitochondrial bioenergetics, and the potential for targeting MC protein interactions in therapeutic strategies.