Uncoupling Protein 3 Catalyzes the Exchange of C4 Metabolites Similar to UCP2

2.1. Chemicals

Unless otherwise mentioned, all chemicals were purchased in highest purity from Sigma Aldrich GmbH, Merck Millipore, or Carl Roth GmbH. Lipids 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-dioleoyl-sn-glycero-3-phospho-ethanolamine (DOPE) and cardiolipin (CL) were from Avanti Polar Lipids Inc. (Alabaster, AL, USA). Radioactive substrates [2,5′,8-³H]-ATP, ¹⁴C-L-malic acid and ³²P-phosphate were obtained from Perkin Elmer (Waltham, MA, USA).

2.2. Homology Modelling of Mouse UCP3 and Homology Score Calculation

Homology model of mUCP3 was made using the SWISS-MODEL Workspace/ GMQE [45]. The tertiary structure of protein was computed using the amino acid sequence of mUCP3 (Uniprot #: P56501) and the crystallographic structure of bovine ADP/ATP carrier (PDB:1OKC, [46]). Sequence homology scores between mUCP3 and mUCP2 (Uniprot #: P70406) and mANT1 (Uniprot #: P48962) were calculated in 10 amino acid increments from the sequence alignments (Figure S1), using the Sim Alignment Tool (https://web.expasy.org/sim/) and coloured as depicted in Figure 1.

2.3. Production and Reconstitution of Recombinant Proteins into Liposomes

Expression, induction, inclusion body isolation and reconstitution into liposomes of mUCP3, mUCP2 and mANT1 were performed as described elsewhere [24,47,48]. In brief, the open reading frame of mUCP3, mUCP2 or ANT1 was added to the E. coli expression strain Rosetta. Inclusion bodies containing the target protein were extracted by cell disruption using a French press and resuspended into TE/G buffer (100 mM Tris at pH 7.5, 5 mM EDTA and 10% v/v glycerin) containing 2% w/v lauroylsarcosine. For protein reconstitution, 1 mg protein from inclusion bodies was solubilized in TE/G-buffer containing 2% SLS and 1 mM DTT and mixed gradually with 50 mg lipid mixture (DOPC:DOPE:CL - 45:45:10 mol%) dissolved in TE/G-buffer with the addition of 1.3% Triton X-114, 0.3% octylpolyoxyethylene, 1 mM DTT and 2 mM GTP. After overnight incubation, the mixture was concentrated using Amicon^® Ultra-15 filters (Millipore, Schwalbach, Germany), dialyzed for 2 h against 1 L of TE/G- buffer containing 1 mg/ml BSA and 1 mM DTT and then dialyzed twice overnight against 1 L of TE/G without DTT. Buffer exchange was performed by three dialyses against 1 L of assay buffer (50 mM Na₂SO₄, 10 mM MES, 10 mM Tris, 0.6 mM EGTA at pH = 7.35). Aggregated and unfolded proteins were removed by centrifugation of the dialysate at 14,000×g and application of a 0.5 g hydroxyapatite column (Bio-Rad, Munich, Germany). GTP, initially added for the protein stabilization, was diluted 7-fold, so that the amount of GTP in the proteoliposomes was negligible.

Remaining detergent was removed by incubating twice with Bio-Beads SM-2 (Bio-Rad, Germany). The protein content of proteoliposomes was determined by Micro BCA Protein Assay (Perbio Science Deutschland GmbH, Bonn, Germany). The protein purity was verified by silver staining (Figure S2). Proteoliposomes were stored at − 80 °C until used, but not longer than two months.

2.4. Formation of Unilamellar Liposomes

DOPE, DOPC and CL lipids were mixed in chloroform in 45:45:10 mol% concentration and evaporated under nitrogen flow until a fine film is formed on the wall of the glass vial. To ensure full evaporation of chloroform, the vial was connected to a vacuum pump for 30 min. Assay buffer (50 mM Na₂SO₄, 10 mM TRIS, 10 mM MES, 0.6 mM EGTA; pH = 7.34) was added to the lipids and the solution is vortexed to dissolve the lipids. Lipids were mixed with mUCP2 -, or mUCP3 - or ANT1 - containing liposomes to reach a final lipid concentration of 4 mg/ml and a protein to lipid ratio of 4 µg/(mg of lipid). For empty liposomes, the lipid mixture was dissolved to 4 mg/ml with assay buffer. Liposomes were then incubated with 2 mM of phosphate (Pi) or malate and 10 to 50 µL ³²P-Pi or 10 µL ¹⁴C-malate as tracer. Unilamellar liposomes were formed with an extruder (Avanti Polar Lipids, Inc., Alabaster, Alabama, US) using a 400 nm polycarbonate pore filter followed by membrane filter with a 100 nm pore diameter. Unincorporated substrates were removed via size exclusion chromatography. The homogeneity of the liposomal size was continuously verified by dynamic light scattering (Zetasizer, Malvern Panalytical Ltd, UK).

2.5. Radioactive Transport Assay

Release of intraliposomal substrates was initiated by adding 2 mM of the chosen non-radioactive substrate to the extraliposomal volume. Substrate exchange was stopped at t = 10 s, 60 s and 600 s by adding the sample to 30 ml size exclusion resin. Initial radioactivity of the (proteo-) liposomes was assessed by adding the sample to the resin before adding extraliposomal substrates. For each time point, the first 20 ml from the flow-through was collected in 4 x 5 ml fractions to (i) measure the peak signal, corresponding to the remaining amount of radioactive substrate in the (proteo-)liposomes and (ii) to collect the background signal as a quality control. Peak radioactivity was usually found in fraction 2 and 3, while the background signal was obtained in fraction 1 and 4. To each fraction, 10 mL of scintillation cocktail was added, and radioactivity measured by liquid scintillation counting (Tri-Carb 2100TR, Perkin Elmer). Inhibitors were added prior to liposome formation and were present intraliposomal and in the bulk solution to account for the random orientation of UCP3 in the membrane.

2.6. Determination of Exchange Rates

Data is represented by the remaining radioactive substrate, which is the counts per minute (CPM) of the radioactive substrates over time divided by the CPM from t = 0 s. The data was then fitted to an exponential function, where y₀, a and b are the fitting parameters.

$$y\left(t\right)=y_0+a·e^{-b·t}$$

(1)

$$y_0+a=1$$

(2)

Equation (2) is required to meet the starting condition. The exchange rate k is then derived from the initial slope of the exponential function, where [X₀] = 2 mM is the starting concentration of the respective radioactive substrate:

$$k=-\left[X_0\right]\frac{\partial }{\partial t}y\left(t\right)=\left[X_0\right]·a·b$$

(3)

2.7. Site-Directed Mutagenesis

In vitro site-directed mutagenesis was carried out on expression plasmids containing the cDNA of mUCP3 as templates. The mutation was introduced with designed oligonucleotides to alter mUCP3 Arg84 (CGC) to Gln (CAG) (R84Q) using a QuikChange II site-directed mutagenesis kit (Agilent, Austria). Successful mutations of the expression plasmids were confirmed by sequencing. Mutant mUCP3 expression plasmids were transformed in the E. coli expression strain Rosetta. Expression, induction, inclusion body isolation and reconstitution into liposomes of mUCP3 mutant were performed as described above.

2.8. Docking of Phosphate and Malate to UCP3

The structure of UCP3 was prepared for docking by removal of non-standard residues, non-complexed ions, addition of hydrogen, and assignment of Gasteiger charges in UCSF Chimera [49]. The ligand structures (phosphate and malate) were retrieved from the ZINC database (https://zinc.docking.org/, accessed last on 27 February 2023). Static docking was performed using AutoDock [50,51], embedded in UCSF Chimera. The grid box was set around putative binding site. The docking was performed in 10 iterations per substrate, with one iteration picked based on binding energy and distance between arginine residues and substrate molecule. The docking image was visualized in PyMOL (PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC).

2.9. Statistical Analysis

Data are presented as the mean and standard deviation of at least 3 independent experiments.

3. Results

3.1. mUCP3 Exchanges Malate against Phosphate

First, we investigated whether mUCP3 reconstituted into liposomes transports radioactively labelled ¹⁴C-malate against phosphate, using mUCP2 as a positive control [33] and ANT1 as a negative control (Figure 2A). Empty liposomes, in which the exchange should be driven by diffusion over the membrane only were used as an additional negative control (Figure 2A). By comparing the efflux of ¹⁴C-malate from proteoliposomes measured over time we concluded that mUCP3 catalyses the malate/phosphate exchange similarly to mUCP2 (Figure 2B). While ANT1 catalyses the exchange of ATP against ADP (Figure S3), it did not induce any significant malate/phosphate exchange, as expected from its tight substrate specificity [52,53]. Thus, we conclude that both, mUCP3 and mUCP2, are active transporter of malate and phosphate. We calculated exchange rates of 9.57 ± 4.39 µM/s (corresponding to 12.0 ± 5.5 mmol/min/g_protein) and 14.5 ± 3.5 µM/s (18.1 ± 4.4 mmol/min/g_protein) for mUCP3 and mUCP2, while the rates for empty and ANT1-containing liposomes were 0.19 ± 0.01 µM/s and 0.23 ± 0.09 µM/s, respectively (Figure 2C and Table 1). We also tested whether UCP3-mediated phosphate/malate exchange is inhibited by GTP, GDP and the substrate analogue phenylsuccinate (PS). We measured the amount of ³²P-phosphate remaining in the liposomes after 600 s compared to the initial amount at the beginning of the experiment. We found inhibition values of 76.3 ± 18.2 % in the presence of 10 mM GTP, 47.0 ± 2.0 % for 10 mM GDP and 92.2 ± 4.6 % for 20 mM PS (Figure 2D and Figure S4).

3.2. The Homoexchange of Malate and Phosphate by mUCP3

Next, we examined the homoexchange of malate and phosphate mediated by mUCP3. We filled mUCP3 containing liposomes either with ¹⁴C-malate or ³²P-phosphate and measured their release against malate or phosphate, respectively (Inserts in Figure 3A,B). For the comparison, we measured these exchange rates for mUCP2 as well. We found that the malate exchange is higher for mUCP2 (16.6 ± 3.5 µM/s) than for mUCP3 (7.94 ± 1.88 µM/s), while the phosphate exchange is higher for mUCP3 (11.7 ± 3.2 µM/s) compared to mUCP2 (8.00 ± 1.40 µM/s) (Figure 3C and Table 1).

3.3. mUCP3 Catalyses the Heteroexchange of Phosphate against Various Substrates

In the next step, we analyzed mUCP3-mediated heteroexchange of phosphate against a selection of mitochondrial relevant substrates that were tested for UCP2. We filled UCP3-containing liposomes with ³²P-phosphate and measured the release of phosphate (Figure 4A and Figure S5). Since the proteins are randomly inserted into the liposomes (approximately 50:50), we can only globally determine the transport rate but not its directionality. We consider this by assuming that only half of the proteins are involved in the specific exchange. If the exchange of substrates is independent of the protein orientation, the exchange rates determined here would be half the value. We found that the exchange of phosphate against aspartate (19.1 ± 4.6 µM/s), sulfate (14.0 ± 4.1 µM/s) and sulfite (12.9 ± 4.5 µM/s) is higher than against malate (Figure 2). Lower transport rates were observed for oxaloacetate (5.07 ± 1.92 µM/s), thiosulfate (2.70 ± 0.36 µM/s), malonate (1.96 ± 0.72 µM/s), succinate (1.16 ± 0.17 µM/s) and citrate (1.10 ± 0.39 µM/s). In contrast, little to no transport was measured for asparagine (0.40 ± 0.04 µM/s), oxoglutarate (0.37 ± 0.04 µM/s), glutamate (0.29 ± 0.07 µM/s) and glutamine (0.27 ± 0.12 µM/s). All values are also shown in mmol/min/g protein in Table 1. Here, we decided to take twice the “apparent” malate/phosphate exchange rate of empty liposomes (Figure 2) to define non-transported substrates. We tested the same set of substrates on the mUCP2-mediated exchange with phosphate and found strong differences to mUCP3 (Figure 4B and Figure S5). Besides malate, the highest exchange rates were determined for sulfate (18.4 ± 3.7 µM/s) and malonate (16.5 ± 2.0 µM/s). Much lower exchange rates had aspartate (1.31 ± 0.19 µM/s), oxaloacetate (1.26 ± 0.18 µM/s) and citrate (1.14 ± 0.38 µM/s). Very low to no exchange was measured for thiosulfate (0.54 ± 0.36 µM/s), succinate (0.52 ± 0.19 µM/s), glutamate (0.36 ± 0.06 µM/s), oxoglutarate (0.32 ± 0.04 µM/s), asparagine (0.27 ± 0.04 µM/s), sulfite (0.24 ± 0.06 µM/s) and glutamine (0.24 ± 0.02 µM/s) (see also Table 1). While most of the substrates are not transported at all or only at a very slow rate (Figure 4C, white bars), aspartate, oxaloacetate, succinate, sulfite and thiosulfate have a significantly higher exchange rate for mUCP3 compared to mUCP2, with a ratio of exchange rate (k_UCP3/k_UCP2) of 14.6 ± 4.1, 4.0 ± 1.6, 2.23 ± 0.88, 53.8 ± 23.0 and 5.0 ± 3.4, respectively (Figure 4C, green bars). In contrast, malate and malonate are transported faster by mUCP2, with a ratio of the exchange rates (k_ucp3/k_ucp2) of 0.66 ± 0.31 and 0.12 ± 0.05, respectively (Figure 4C, orange bars). For citrate (0.96 ± 0.47) and sulfate (0.76 ± 0.27), the exchange rate is almost equal for both proteins (Figure 4C, grey bars). In the absence of any external substrate, we did not observe any significant ³²P-phosphate efflux, ruling out phosphate uniport for UCP3 and UCP2 (“None” in Figure 4 and Figure S7).

3.4. Arginine 84 (R84) of UCP3 Plays a Crucial Role in mUCP3-Mediated Substrate Exchange

Finally, we performed static docking experiments to identify most of the existing phosphate binding sites by using the homology model structure of mUCP3 (Figure 1). Since the aspartate/phosphate exchange had highest exchange rate, we also tested the docking of aspartate to mUCP3. We found a high probability for phosphate to bind to R84 and R184 (Figure 5A) and for aspartate to R84 and R278 (Figure 5B). Since R84 appeared in both docking results, we produced the mUCP3R84Q mutant and examined the effect of the mutation on the aspartate/phosphate heteroexchange (Figure 5C). Indeed, the exchange rate was almost 40-fold lower for the R84Q mutant (0.50 ± 0.08 µM/s, Figure 5D) and comparable with non-transported substrates (Figure 4A, Table 1). Thus, R84 of mUCP3 has a significant impact in the UCP3-mediated heteroexchange of aspartate against phosphate.

4. Discussion

Our present results show that mouse UCP3 facilitates membrane transport of negatively charged metabolites which otherwise have a rather low intrinsic membrane permeability [54]. The exchange rate decreases in the order: aspartate > sulfate ≈ sulfite ≈ phosphate ≈ malate >> oxaloacetate >> thiosulfate ≈ malonate ≈ succinate ≈ citrate. These results are similar to those obtained for recombinant human UCP3 in parallel with our study [55]. As mUCP2 and mUCP3 are highly homologous at the substrate binding site (Figure 1), we expected similarities in their substrate specificity. However, there are clear differences between the two proteins: mUCP3 showed its highest exchange rates for aspartate, sulfate and sulfite, whereas mUCP2 transported sulfate and malonate at the highest rate. Because the tissue expression of mUCP2 and mUCP3 is different [28], we suggest that the difference in substrate affinity may be explained by differences in the main tissue metabolism. For example, the export of aspartate catalyzed by UCP3 from mitochondria may play a role in attenuating the development of hypertrophy in cardiomyocytes under stress conditions when cellular metabolism shifts toward glycolysis [56]. By transporting sulfate, UCP3 could contribute to cardioprotection [57]. The highly membrane permeable signaling molecule H₂S [58] is degraded to the membrane impermeable sulfite in the mitochondrial matrix, and sulfite is in turn degraded to sulfate by sulfite oxidase, which is in the intermembrane space. As UCP3 also exhibits a high transport rate for sulfate, it may help to reduce ROS damage in the heart and skeletal muscle by exporting sulfate from the mitochondria, encouraging further sulfite/sulfate exchange. Human UCP5 and UCP6, which are less homologous to UCP3 than UCP2, have previously been proposed to contribute to ROS regulation by exporting H₂S degradation products, sulfite and thiosulfate, from mitochondria in exchange for sulfate [57]. Thus, it remains an open question whether UCP3 transports sulfoxides. Interestingly, hUCP5 and hUCP6 are also proposed to transport dicarboxylates and aspartate, although to a lesser extent. In addition, the insect UCP4 from Drosophila melanogaster was reported to catalyze the homoexchange of aspartate and, with much lower activity, the uniport transport of aspartate [38]. However, uniport transport was not observed for mammalian UCP2, UCP5 and UCP6 [33,59]. Thus, UCP3-mediated uniport transport of aspartate is not supported by the much closer mammalian homologs.

Both mUCP3 and mUCP2 catalyze the homoexchange of phosphate and malate, but with different kinetics. While the homoexchange of mUCP3 is faster for phosphate (compared to mUCP2), the homoexchange of malate is faster for mUCP2 (compared to mUCP3). Substrate homoexchange seems strange at first, but several mitochondrial carriers have been shown to mediate homoexchange, such as the ADP/ATP carrier (ADP and ATP transport in both directions) and the phosphate carrier. Both transport modes are correlated with the metabolic needs of the mitochondria and whether the ATP synthase produces ATP from ADP and phosphate or breaks up ATP into ADP and Pi. The exchange rates determined in this study are of the same order of magnitude for the different mitochondrial carriers and their respective substrates (Table S1). The Pi/Pi homoexchange rate (in mmol/min/g protein) is slightly higher for the phosphate carrier [60] than for UCP3 (PiC: 70 –90; UCP3: 11.7 ± 3.2). In contrast, the malate/malate homoexchange mediated by the citrate carrier (CiC) is lower [61] compared to UCP3 (CiC: 1.98 ± 0.16; UCP3: 7.94 ± 1.88).

The calculated transport rates of mUCP2 in this study are comparable to those reported for other uncoupling proteins [33,38,39]. We generally found a similar substrate specificity for mUCP2, except for citrate, which was reported to be not transported by hUCP2 [33]. The transport rates of malate, malonate, and sulfate where 3- to 6-fold higher in our study, which could be due to the higher concentration of substrates used (2 mM compared to 1 mM [33]. In contrast, the transport rates of aspartate and oxaloacetate are lower [33]. However, it should be emphasized that exchange rates in different studies can be better compared by the determination of the apparent binding constant K_m and the maximum exchange rate V_max according to the Michaelis-Menten relation. With the apparent K_m of 1 mM for phosphate [24] and 2.4 ± 0.1 mM for malate [38], the transport rates determined in this paper are expected to be higher than those determined by Vozza et al. Another possible explanation for some of the differences in the results could be that Vozza et al. investigated human UCP2 by measuring ³²Pi influx, whereas in the present study mouse UCP2 was investigated by measuring of ³²Pi efflux.

Meanwhile, other orthologous uncoupling proteins have been shown to be substrate carriers. Plant UCP1 and UCP2 were functionally characterised as amino acid/dicarboxylate transporters [39], while UCP4A from Drosophila melanogaster catalyses the unidirectional transport of aspartate [38]. UCP4 from Caenorhabditis elegans was reported to transport succinate, which controls complex II-mediated oxidative phosphorylation [62]. In contrast, several mitochondrial substrate transporters from the SLC25 group such as ATP/ADP carrier (ANT), phosphate carrier (PiC), dicarboxylate carrier (DIC), oxoglutarate carrier (OGC) have been shown to have a protonophoric activity, as a part of a dual transport function.

The importance of the central arginine ring for the binding of the ATP, ADP and fatty acid anion has already been demonstrated for ANT, UCP1 and UCP3 [24,41,44,63,64]. We have hypothesised that the substrate binds in the same region of mUCP3. Indeed, the mutation of R84 resulted in impaired aspartate/phosphate exchange, demonstrating its critical role in substrate transport, and suggesting a mechanism similar to that of ANT1 [44]. The latter involves at least one of the arginines (R79) in the central binding site in the substrate and FA anion transport pathways.

Our results shed light on the possible dual transport function that UCP3 may play in the mitochondrial metabolism of the tissues in which it is expressed. They also highlight the importance of further investigation of the possible substrate transport function of members of the uncoupling protein family.