Uncoupling proteins 1 and 2 (UCP1 and UCP2) from Arabidopsis thaliana are mitochondrial transporters of aspartate, glutamate, and dicarboxylates

The Arabidopsis thaliana genome contains 58 members of the solute carrier family SLC25, also called the mitochondrial carrier family, many of which have been shown to transport specific metabolites, nucleotides, and cofactors across the mitochondrial membrane. Here, two Arabidopsis members of this family, AtUCP1 and AtUCP2, which were previously thought to be uncoupling proteins and hence named UCP1/PUMP1 and UCP2/PUMP2, respectively, are assigned with a novel function. They were expressed in bacteria, purified, and reconstituted in phospholipid vesicles. Their transport properties demonstrate that they transport amino acids (aspartate, glutamate, cysteine sulfinate, and cysteate), dicarboxylates (malate, oxaloacetate, and 2-oxoglutarate), phosphate, sulfate, and thiosulfate. Transport was saturable and inhibited by mercurials and other mitochondrial carrier inhibitors to various degrees. AtUCP1 and AtUCP2 catalyzed a fast counterexchange transport as well as a low uniport of substrates, with transport rates of AtUCP1 being much higher than those of AtUCP2 in both cases. The aspartate/glutamate heteroexchange mediated by AtUCP1 and AtUCP2 is electroneutral, in contrast to that mediated by the mammalian mitochondrial aspartate glutamate carrier. Furthermore, both carriers were found to be targeted to mitochondria. Metabolite profiling of single and double knockouts shows changes in organic acid and amino acid levels. Notably, AtUCP1 and AtUCP2 are the first reported mitochondrial carriers in Arabidopsis to transport aspartate and glutamate. It is proposed that the primary function of AtUCP1 and AtUCP2 is to catalyze an aspartateout/glutamatein exchange across the mitochondrial membrane and thereby contribute to the export of reducing equivalents from the mitochondria in photorespiration.

Mitochondrial carriers (MCs) 6 are a large family of membrane proteins that transport nucleotides, amino acids, carboxylic acids, inorganic ions, and cofactors across the mitochondrial inner membrane (1)(2)(3). Many metabolic pathways and cellular processes with complete or partial localization in the mitochondrial matrix are dependent on transport steps catalyzed by MCs (e.g. oxidative phosphorylation, metabolism of fatty acids and amino acids, gluconeogenesis, thermogenesis, mitochondrial replication, transcription, and translation) (3). The protein sequences of the MC family members have a characteristic three times tandemly repeated 100-residue domain (4), which contains two hydrophobic segments and a signature sequence motif PX(D/E)XX(K/R)X(K/R) (20 -30 residues) (D/E)GXXXX(W/Y/F)(K/R)G (PROSITE PS50920, PFAM PF00153, and IPR00193) (5). In atomic resolution 3D structures of the only MC family member determined to date (the carboxyatractyloside-inhibited ADP/ATP carrier) (6,7), the six hydrophobic segments form a bundle of transmembrane ␣-helices with a central substrate translocation pore, and the three PX(D/E)XX(K/R) motifs form a gate toward the matrix side. In most cases, the MC signature motif has been used to identify family members in genomic sequences; Homo sapiens has 53 members, Saccharomyces cerevisiae has 35, and Arabidopsis thaliana has 58. About half of these carriers have been identified and characterized in terms of substrate specificity, transport proteins, and kinetic parameters by direct transport assays (1,8,9).
In the current study, we investigated the potential transport properties of the two closest homologs of hUCP2 in Arabidopsis: AtUCP1 and AtUCP2, also known as PUMP1 and PUMP2. Previously, AtUCP1 was shown to be localized to mitochondria and display an uncoupling activity similar to that of hUCP1 (35)(36)(37). By contrast, very little is known about AtUCP2; in a proteomic study, it was detected in the Golgi (38), but in another, it was detected at the plasma membrane (39). The results presented here demonstrate that AtUCP1 and the lessstudied AtUCP2 are mitochondrially localized isoforms and have a broad substrate specificity, transporting a variety of substrates, including aspartate, glutamate, malate, oxaloacetate, and other metabolites. Characterization of metabolite profiles of T-DNA insertional knockout mutants, including a ucp1/ ucp2 double mutant, revealed clear changes in organic acid levels, some of which were exacerbated by the application of salt stress.

Identification of the closest homologs of AtUCP1 and AtUCP2 in various species
The protein sequences of AtUCP1 and AtUCP2 homologs were collected, aligned, and analyzed (Fig. S1). AtUCP1 and AtUCP2 share 72% identical amino acids. Their sequences are much more similar to each other than to any other Arabidopsis protein; in Arabidopsis, the closest relative to AtUCP1 and AtUCP2 is AtDIC2, which shares 41 and 42% sequence identity with AtUCP1 and AtUCP2, respectively. In humans and S. cerevisiae, the closest homologs are hUCP2 (34), having 51 and 44% identical amino acids, and yeast Dic1p (40), exhibiting 30 and 33% sequence identity with AtUCP1 and AtUCP2, respectively. Putative orthologs with high sequence identity with AtUCP1 and AtUCP2 above 75% were found in several plant species. Moreover, from structural sequence alignments using the X-ray structure of the bovine ADP/ATP carrier (6) as a template, it can be deduced that 85 and 54% of the residues lining the surface of the substrate translocation pore are identical between AtUCP1 and AtUCP2 and between AtUCP1, AtUCP2, and hUCP2, respectively. These results suggest that AtUCP1 and AtUCP2 are isoforms, and their closest homolog with identified substrates is hUCP2.

Bacterial expression of AtUCP1 and AtUCP2
AtUCP1 and AtUCP2 were expressed in Escherichia coli BL21(DE3) strains (Fig. 1, lanes 4 and 7). They accumulated as inclusion bodies and were purified by centrifugation and washing (see "Experimental procedures"). The apparent molecular masses of purified AtUCP1 and AtUCP2 (Fig. 1, lanes 5 and 8) were ϳ31 kDa, which is in good agreement with the calculated value of 33 kDa for both AtUCP1 and AtUCP2. The identities of the recombinant proteins were confirmed by MALDI-TOF mass spectrometry, and the yield of the purified proteins was about 10 and 2 mg/liter of culture for AtUCP1 and AtUCP2, respectively. The proteins were detected neither in non-  . Samples were taken immediately before induction (lanes 1 and 2) and 5 h later (lanes 3, 4, 6, and 7). The same number of bacteria were analyzed in each sample. Lanes 5 and 8, purified AtUCP1 protein (5 g) and purified AtUCP2 (3 g) derived from bacteria shown in lanes 4 and 7, respectively.

Functional characterization of recombinant AtUCP1 and AtUCP2
Recombinant AtUCP1 and AtUCP2 were reconstituted into liposomes, and their transport activities for various radioactive substrates were tested in homo-exchange experiments (i.e. with the same external (1 mM) and internal (10 mM) substrate). In a first set of homo-exchange experiments, time-dependent uptake of several radioactive substrates (aspartate, malate, and glutamate for reconstituted AtUCP1 and AtUCP2; malonate and sulfate for AtUCP1; and 2-oxoglutarate for AtUCP2) demonstrated typical curves for carrier-mediated transport (Fig. 2,  A and B). Both AtUCP1-and AtUCP2-mediated homo-exchanges between external [ 14 C]aspartate and internal aspartate were temperature-dependent (Fig. 2, C and D), as would be expected for protein-catalyzed transport. Furthermore, no [ 14 C]aspartate/aspartate or [ 14 C]malate/malate exchange activity was detected if AtUCP1 and AtUCP2 had been boiled before incorporation into liposomes or if proteoliposomes were reconstituted with lauric acid/sarkosyl-solubilized material from bacterial cells lacking the expression vector for AtUCP1 and AtUCP2 or harvested immediately before induction of expression (data not shown). In all of these experiments, a mixture of pyridoxal-5Ј-phosphate and bathophenanthroline was used to block the AtUCP1-and AtUCP2-mediated transport reactions at various time points. In addition, AtUCP1 and AtUCP2 were found to catalyze homo-exchanges of glutamate, malonate, malate, succinate, and P i , whereas no or very low transport was observed with glutamine, arginine, phenylalanine, threonine, valine, proline, ␥-aminobutyrate, citrate, ATP, GTP, S-adenosylmethionine, or glutathione (Fig. 3).

Transport properties of AtUCP1 and AtUCP2
glutathione (Fig. 4, A and B). The activity in the presence of these substrates was approximately the same as that observed in the presence of NaCl and no substrate.
The effects of other mitochondrial carrier inhibitors on the [ 14 C]aspartate/aspartate exchange reaction catalyzed by reconstituted AtUCP1 and AtUCP2 were also examined. This transport activity was inhibited strongly by bathophenanthroline, pyridoxal-5Ј-phosphate, and tannic acid and markedly by mersalyl, HgCl 2 , and butylmalonate (Fig. 5). Phenylsuccinate and p-hydroxymercuribenzoate strongly inhibited AtUCP1 and partially inhibited AtUCP2, whereas bromcresol purple and ␣-cyano-4-hydroxycinnamate caused partial inhibition of both carriers. By contrast, carboxyatractyloside, bongkrekic acid, and N-ethylmaleimide had little or no effect on either AtUCP1 or AtUCP2 activity.

Kinetic characteristics of recombinant AtUCP1 and AtUCP2 proteins
In Fig. 6, the kinetics of 1 mM [ 14 C]aspartate (A and B) or 1 mM [ 14 C]malate (C and D) uptake into proteoliposomes catalyzed by recombinant AtUCP1 (A and C) or AtUCP2 (B and D) and measured either as uniport (with internal NaCl) or as exchange (in the presence of 10 mM substrates) are compared. The [ 14 C]aspartate/aspartate and [ 14 C]malate/malate exchanges followed first-order kinetics (rate constants 1.6 and 1.4 min Ϫ1 (AtUCP1) or 0.27 and 0.23 min Ϫ1 (AtUCP2); initial rates 14 and 11 mmol/min ϫ g of protein (AtUCP1) or 1.9 and 1.3 mmol/ min ϫ g of protein (AtUCP2), respectively), isotopic equilibrium being approached exponentially. By contrast, with internal NaCl and no substrate, very low uptake of [ 14 C]aspartate or [ 14 C]malate was observed by liposomes reconstituted either with AtUCP1 or AtUCP2, suggesting that the two proteins catalyze a minor unidirectional transport (uniport) of substrates. In addition, Fig. 6 (A-D) illustrates the time courses of several AtUCP1-mediated and AtUCP2-mediated hetero-exchanges between [ 14 C]aspartate or [ 14 C]malate and other transported substrates. The data of Fig. 6 (A and C) show that AtUCP1 transports cysteate much better than D-aspartate and dicarboxylates with the following order of efficiency: malate Ͼ oxaloacetate Ͼ malonate Ͼ succinate (these substrates better than P i ). Similarly, the data of Fig. 6 (B and D) demonstrate that cysteate is transported slightly better than D-aspartate and malate more efficiently than oxaloacetate by AtUCP2. The uniport mode of transport was further investigated by measuring the efflux of [ 14 C]aspartate or [ 14 C]malate from preloaded active proteoliposomes because it provides a more convenient assay for unidirectional transport (41). In the absence of external substrate, significant efflux of [ 14 C]aspartate (Fig. 7, A and B) or [ 14 C]malate ( Fig. 7, C and D) catalyzed by both AtUCP1 and AtUCP2 was observed. However, in the presence of external substrates, the efflux transport rates were at least one order of magnitude higher. These experiments demonstrate that AtUCP1 and AtUCP2 are capable of catalyzing both a rapid antiport of substrates and a slow uniport transport.
The kinetic constants of AtUCP1 and AtUCP2 were determined from the initial transport rates of homo-exchanges at various external labeled substrate concentrations in the presence of a constant saturating internal substrate concentration. The Michaelis constants (K m ) of the two recombinant proteins for aspartate were about 0.8 mM, and for glutamate and malate, they were between 1.9 and 2.5 mM. The maximal activities (V max ) for aspartate, glutamate, and malate varied between 24 and 33 mmol/min ϫ g of protein for AtUCP1 and 4.2 and 4.5 mmol/min ϫ g of protein for AtUCP2 (Table 1). Glutamate, malate, cysteine sulfinate, cysteate, oxaloacetate, ␣-ketoglutarate, and sulfate were competitive inhibitors of the AtUCP1and AtUCP2-mediated [ 14 C]aspartate/aspartate exchanges, as

Transport properties of AtUCP1 and AtUCP2
they increased the apparent K m without changing the V max (not shown). The inhibition constants (K i ) of these compounds are listed in Table 2.

Influence of membrane potential and pH gradient on the AtUCP1-and AtUCP2-mediated exchange reactions
Given that the mammalian aspartate glutamate carriers AGC1 and -2 have been shown to catalyze an electrophoretic exchange between aspartate Ϫ and glutamate Ϫ ϩ H ϩ (30), we investigated the influence of the membrane potential on the [ 14 C]aspartate/glutamate exchange catalyzed by recombinant AtUCP1 and AtUCP2. A K ϩ -diffusion potential was generated across the proteoliposomal membranes with valinomycin in the presence of a K ϩ gradient of 1:50 (mM/mM, in/out), corresponding to a calculated value of about 100 mV positive inside ( Table 3). The rate of the [ 14 C]aspartate out /glutamate in heteroexchange was unaffected by valinomycin in the presence of the K ϩ gradient. By contrast, the aspartate out /glutamate in exchange, mediated by recombinant AGC2 C-terminal domain (AGC2-CTD) (30), was stimulated under the same experimental conditions. These results indicate that the AtUCP1-and AtUCP2-mediated aspartate/glutamate hetero-exchange is not electrophoretic but electroneutral, suggesting that AtUCP1 and AtUCP2 transport either aspartate Ϫ for glutamate Ϫ or both together with a H ϩ . Also, the AtUCP1-and AtUCP2-mediated aspartate/aspartate, malate/malate, and malate out /aspartate in (Table 3) and malate out /glutamate in (data not shown) were unaffected by valinomycin in the presence of a K ϩ gradient of 1:50. In view of the different charges carried by the amino

Transport properties of AtUCP1 and AtUCP2
acids aspartate and glutamate and dicarboxylates at physiological pH levels, we explored whether the charge imbalance of the malate/aspartate and malate/glutamate hetero-exchanges catalyzed by AtUCP1 and AtUCP2 is compensated by proton movement. A pH difference across the liposomal membranes (basic inside the vesicles) was created by the addition of the K ϩ /H ϩ exchanger nigericin to proteoliposomes in the presence of a K ϩ gradient of 1:50 (mM/mM, in/out). Under these conditions, the uptake of [ l4 C]malate in exchange for internal aspartate or glutamate increased (Table 3), whereas the uptake of [ l4 C]malate in exchange for internal malate or 2-oxoglutarate was unaffected (data not shown). Therefore, the charge imbalance of the substrates exchanged by AtUCP1 and AtUCP2 is compensated by the movement of protons.

Subcellular localization of AtUCP1 and AtUCP2 proteins in transiently transformed Nicotiana benthamiana leaf cells
C-terminal fusion proteins of AtUCP1 and AtUCP2 with the green fluorescent protein (GFP) under the control of an Arabidopsis ubiquitin-10 promoter were transiently expressed in N. benthamiana to investigate the subcellular localization via confocal laser-scanning microscopy. Simultaneously, N. benthamiana was co-infiltrated with the mitochondrion-located Arabidopsis isovaleryl-CoA-dehydrogenase tagged with a Cterminal eqFP611 (IVD-eqFP611) under the control of the cauliflower mosaic virus 35S promoter. Two days after infiltration, protoplasts were isolated from leaf tissue and directly used for microscopy.
The C-terminal fusion proteins AtUCP1-GFP (Fig. 8A) and AtUCP2-GFP ( Fig. 8B) (shown in green) clearly overlap with almost all mitochondrial IVD-eqFP611 fluorescent signals (shown in red) in all observed protoplasts, indicating the mitochondrial localization of both proteins. 48 h after infiltration, the mitochondrial marker was generally higher-expressed than the GFP fusion proteins and was also found in the cytosol with more prominent fluorescent signals detected in mitochondria.

Isolation, generation, and metabolic characterization of AtUCP1-2 knockout mutants
After biochemically characterizing the properties of recombinant AtUCP1 and AtUCP2 proteins, we turned our attention to evaluating their physiological role in Arabidopsis. For this purpose, we acquired the individual T-DNA insertion mutants and crossed them to obtain the ucp1/ucp2 double mutant (Fig.  S2). The ucp1 line used here was extensively characterized by Sweetlove et al. (37), including functional complementation by the UCP1 genomic sequence. This mutant harbors a T-DNA insertion in the first intron and shows low residual UPC protein amounts in the mitochondria (5% of the wildtype line (37)). Congruent with this previous work, we also detected residual expression of the UCP1 gene, which was much lower than that of the wildtype (Fig. S3). The expression of the UCP2 gene was

Transport properties of AtUCP1 and AtUCP2
virtually absent in the newly isolated ucp2 mutant (Fig. S3). Having these genotypes in hand, we next assessed their metabolic phenotypes via GC-MS-based metabolic profiling both in plants grown on normal MS agar and in plants exposed to salt stress. The effects of salt stress were investigated because UCPs have been previously proposed to contribute to the abiotic stress response (36). The clearest metabolic phenotype was that observed in the organic acids; however, changes in phenylalanine, serine, the branched chain amino acids, ornithine, myoinositol, putrescine, and AMP were apparent in one or more of the genotypes (Fig. 9 and Tables S1-S3). These metabolite profiles are thus consistent with the transport assay data suggesting that AtUCP1 and AtUCP2 are important in organic and amino acid metabolism in plants. The metabolic phenotypes of the ucp1/ucp2 double mutant tend to be similar to those of ucp1, suggesting a predominant role of UCP1 in Arabidopsis, at least in the leaf tissue assayed in the work reported here. Regarding the observed changes, interestingly, for the levels of some of the metabolites, such as malate and fumarate, the imposition of salt stress exacerbated intergenotypic differences. For others, such

Table 1 Kinetic constants of recombinant AtUCP1 and AtUCP2
The values were calculated from linear regression of double reciprocal plots of the initial rates of the indicated homo-exchanges versus the external substrate concentration. The exchanges were started by adding appropriate concentrations of labeled substrate to proteoliposomes preloaded internally with the same substrate (10 mM). The reaction time was 7 and 20 s for AtUCP1 and AtUCP2, respectively. The values are means Ϯ S.E. of at least three independent experiments carried out in duplicate.

Transport properties of AtUCP1 and AtUCP2
as citrate, these differences were ameliorated (Fig. 9). The complexity of these results suggests that further research is warranted into the precise physiological role(s) of these proteins under both optimal and suboptimal conditions and in different plant tissues and developmental stages.

Discussion
A recent report has shown that the MC family member hUCP2, which was thought to have a UCP1-like uncoupling activity (42,43), transports aspartate, 4-carbon dicarboxylates, phosphate, and sulfate (34). The percentages of identical amino acids between hUCP2 and AtUCP1 (51%) and AtUCP2 (44%) suggest that these proteins are highly related to one another. However, it is not possible to make reliable assumptions about the substrate specificity or about the transport modes on the basis of the amino acid similarity, given that even close MC homologs, such as isoforms 1 and 2 of the human ornithine carrier having 87% identical sequences, exhibit considerable differences in substrate specificity and transport kinetics (44,45). Therefore, we decided to investigate the transport properties of AtUCP1 and AtUCP2 by recombinant expression, purification, and reconstitution into liposomes (the EPRA method (9)).
The results presented in this study demonstrate that AtUCP1 and AtUCP2 both transport aspartate, glutamate, cysteine sulfinate, cysteate, malonate, malate, oxaloacetate, and 2-oxoglutarate and, to a lesser extent, D-aspartate, cysteine, oxalate, succinate, P i , sulfate, and thiosulfate (Fig. 4). In addition, AtUCP2 also transports fumarate, glutarate, and nitrate to some extent (Fig. 4B). The substrate specificities of AtUCP1 and AtUCP2 are, therefore, (i) similar as expected in light of their high sequence identity (72%) and (ii) broader than those of previously characterized mitochondrial carriers (9), given that their substrates overlap those of hUCP2 and those of the aspar- Table 3 Influence

of membrane potential and pH gradient on the activity of reconstituted AtUCP1 and AtUCP2
The exchanges were started by adding 0.8 mM [ 14 C]aspartate or 0.8 mM [ 14 C]malate to AtUCP1-and AtUCP2-reconstituted proteoliposomes. For the measurements of the aspartate/glutamate carrier activity, 50 M [ 14 C]aspartate was added to proteoliposomes reconstituted with AGC2-CTD. K in ϩ was in included as KCl in the reconstitution mixture, whereas K out ϩ was added as KCl together with the labeled substrate. Valinomycin or nigericin was added in 10 l ethanol/ml of proteoliposomes, whereas the control samples contained the solvent alone. 20 mM or 2 mM PIPES (pH 7.0) was present inside and outside the proteoliposomes in the experiments with valinomycin or nigericin, respectively. The exchange reactions were stopped after 7, 20, and 30 s for AtUCP1, AtUCP2, and AGC2-CTD, respectively. The values are means Ϯ S.E. of four independent experiments carried out in duplicate.

Transport properties of AtUCP1 and AtUCP2
tate/glutamate and dicarboxylate carriers (24,30,40,46,47). AtUCP1 and AtUCP2 share a number of similar transport properties; for example, both proteins catalyze a highly efficient counterexchange of substrates; do not transport mono-and tricarboxylates, nucleotides, and other amino acids; respond similarly to the inhibitors tested; and have similar transport affinities (K i ) for aspartate, glutamate, and malate. However, they greatly differ for their specific activities (V max ), AtUCP1 being much more active than AtUCP2, although both activities are similar or higher than those exhibited by most mitochondrial carriers characterized to date (1,9). Furthermore, some substrates (e.g. D-aspartate and 2-oxoglutarate) are transported at higher rates by AtUCP2 than AtUCP1 compared with the respective [ 14 C]aspartate/aspartate exchanges.
The results of recombinant protein studies are largely consistent with the in vivo evaluation of the function of the proteins that was possible via the isolation and crossing of the respective knockout mutants. Thus, the metabolic phenotype of the mutants was characterized by changes in organic acid and amino acid levels, which are probably due to altered exchanges of these metabolites between the mitochondria and cytosol. However, the differences in metabolite content of the knockouts, which were dependent on salt stress, are difficult to dis-entangle, and this will probably require considerable further research effort.
Two additional remarks regarding the transport properties of AtUCP1 and AtUCP2 should be made. First, the close biochemical similarities between AtUCP1 and AtUCP2 are understandable, given the commonality of their gene structures; both genes (At3g54110 and At5g58970) share an identical exon/intron structure. We therefore assume that they derive from a common molecular ancestor, accounting for their similarities in biochemical properties. After gene duplication, independent evolution took place allowing the development of individual properties, such as the different specific activity and slightly different substrate preference. Second, our transport measurements, in agreement with the previous data on AtUCP3-6 (24) and hUCP2 (34), are in contrast with the idea that AtUCP1-6 as well as the human UCP1-6 are all "uncoupling proteins" transporting protons and dissipating the proton motive force generated by the respiratory chain. In particular, our findings show that AtUCP1 and AtUCP2 greatly differ from AtUCP4 -6 (previously demonstrated to be dicarboxylate carriers) and suggest that they also differ from AtUCP3, which displays only 35 and 37% identical amino acids with AtUCP1 and AtUCP2, respectively. Several protein sequences available in databases
Previous analyses of AtUCP1 knockout mutants revealed impaired photorespiration due to a dramatic decrease in mitochondrial glycine oxidation rate, which led to the suggestion of a physiological role of AtUCP1 in uncoupling the mitochon-drial membrane potential for fine-tuning the cell redox state concomitant with photorespiration (37). The substrate specificity and high transport activity of AtUCP1 revealed in this study shed new light on its role in photorespiration. We suggest that AtUCP1 is involved in the glycolate pathway by playing a role in the transfer of reducing equivalents across the mitochondrial inner membrane (Fig. 10). In the glycolate pathway, 2-phosphoglycolate formed from O 2 usage by Rubisco in chloroplasts needs to be transformed in a series of reactions localized in peroxisomes and mitochondria in order to return to the Calvin-Benson cycle as 3-phosphoglycerate. The mitochondrial reaction catalyzed by glycine decarboxylase (GDC), which is the most abundant enzyme in plant mitochondria, reduces NAD ϩ to NADH, and peroxisomal hydroxypyruvate reductase oxidizes NADH to NAD ϩ . This implies that the reducing equivalents of NADH are transported from the mitochondria to the peroxisomes as malate (Fig. 10). In this respect, AtUCP1 (and AtUCP2) would mainly transport aspartate and glutamate

Transport properties of AtUCP1 and AtUCP2
as components of the mitochondrial malate/aspartate shuttle (MAS), which has been suggested to exist in plants in connection to photorespiration (48,49). More precisely, we propose that the primary function of AtUCP1 (and AtUCP2) is to catalyze an exchange of aspartate out for glutamate in across the inner mitochondrial membrane, thus contributing to the export of reducing equivalents of NADH from mitochondria in conjunction with the other enzymes of MAS. In mammals, MAS transfers the reducing equivalents of NADH from the cytosol to the mitochondria (i.e. in the opposite direction of that occurring during photorespiration) because the aspartate glutamate carriers (AGC1 and -2) catalyze an electrophoretic exchange of aspartate Ϫ for glutamate Ϫ ϩ H ϩ , and hence exit of aspartate and entry of glutamate are greatly favored in active mitochondria with a positive membrane potential outside, making the aspartate/glutamate exchange and the entire MAS unidirectional. By striking contrast, the aspartate/glutamate exchanges mediated by AtUCP1 and AtUCP2 are electroneutral (Table 3) and therefore independent from the proton motive force existing across the mitochondrial membrane. The hypothesis that AtUCP1 and AtUCP2 are involved in the glycolate pathway by catalyzing an aspartate out /glutamate in exchange and thereby contributing to the export of reducing equivalents from the mitochondrial matrix is supported by the following considerations: (i) aspartate and glutamate, to the best of our knowledge, are only transported by AtUCP1 (and AtUCP2) in Arabidopsis mitochondria; (ii) these metabolites are present in the cytosol at very high concentrations (about 20 mM) (50), which are much higher than those of 2-oxoglutarate, malate, and oxaloacetate; (iii) both mitochondrial glutamate-oxaloacetate transaminase and malate dehydrogenase are involved in the regeneration of NAD ϩ by GDC in mitochondria (51), and both mitochondrial and peroxisomal malate dehydrogenases are required for optimal photorespiration rates (52,53); and (iv) as shown by the BAR Arabidopsis eFP Browser 2.0 (http:/bar.utoronto.ca), 7 AtUCP1 and AtUCP2 are expressed in many plant tissues, being more highly expressed in photosynthetic ones (Figs. S4 and S5). Interestingly, the expression of AtUCP1 is co-regulated with enzymes of the citric acid cycle, such as aconitase, isocitrate dehydrogenase, ␣-ketoglutarate dehydrogenase, and succinyl-CoA ligase, as well as with the peroxisomal transporter for NAD ϩ (21,22) (Fig. S6).
It is noteworthy that AtUCP1 (and AtUCP2) may also catalyze (i) the exchange between malate in and oxoglutarate out (i.e. the other mitochondrial membrane reaction of MAS) (Fig. 10) and (ii) an oxaloacetate out /malate in exchange, which per se would result in export of the reducing equivalents of NADH from the mitochondria. However, these exchanges are also catalyzed by other Arabidopsis MCs, such as DTC, DIC1, DIC2, and DIC3 (23,24), and the affinities of AtUCP1 and AtUCP2 for aspartate are much higher than those for the other substrates.
An additional hypothetical function of AtUCP1 and AtUCP2 in the photorespiratory glycolate pathway concerns the transfer of nitrogen equivalents across the mitochondrial membrane. During oxidative glycine decarboxylation by GDC in mitochon-dria, ammonia is released and reassimilated by the plastidial glutamine synthetase/glutamine oxoglutarate aminotransferase (GS/GOGAT) reaction. How ammonia released by GDC in mitochondria is shuttled to GS/GOGAT is still unknown, but it has been suggested that shuttling of amino acids across the mitochondrial membrane might be involved in this process (54). One possible route that would involve AtUCP1 and/or AtUCP2 would be incorporation of ammonia into 2-oxoglutarate by mitochondrial glutamate dehydrogenase, yielding glutamate, which is exported to the cytoplasm in counterexchange with external 2-oxoglutarate, thereby providing a new acceptor molecule for the glutamate dehydrogenase reaction. This hypothesis awaits experimental testing in future work.
Due to their broad substrate specificities, AtUCP1 (and AtUCP2) may be multifunctional and play further important physiological roles, depending on the metabolic conditions of the organ/tissue and the light/dark phase. For example, they might be involved in sulfur metabolism by exchanging cysteine sulfinate, cysteate, and cysteine with sulfate. Furthermore, in the dark, AtUCP1 (and AtUCP2) may catalyze the transport of glutamate into the mitochondria and the exit of aspartate to the cytosol, providing (together with the other enzymes of the malate aspartate shuttle) reducing equivalents in the form of NADH ϩ H ϩ to the mitochondrial respiratory chain.

Sequence analysis
BLAST and reciprocal BLAST were used to search for homologs of AtUCP1 and AtUCP2 (encoded by At3g54110 and At5g58970, respectively) in e!Ensamble and UniProt. Sequences were aligned with ClustalW.

Bacterial expression and purification of AtUCP1 and AtUCP2
PCR using complementary sequence-based primers was used to amplify the coding sequences of AtUCP1 from A. thaliana leaf cDNA and AtUCP2 from a custom-made synthetic gene (Invitrogen) with codons optimized for E. coli. The forward and reverse oligonucleotide primers contained the restriction sites NdeI and HindIII (AtUCP1) or XhoI and EcoRI (AtUCP2). The amplified gene fragments were cloned into pMW7 (AtUCP1) and pRUN (AtUCP2) vectors and transformed into E. coli TG1 cells (Invitrogen). Transformants were selected on LB (10 g/liter tryptone, 5 g/liter yeast extract, 5 g/liter NaCl, pH 7.4) plates containing 100 g/ml ampicillin. All constructs were verified by DNA sequencing.
AtUCP1 and AtUCP2 were overexpressed as inclusion bodies in the cytosol of E. coli BL21(DE3) (AtUCP1) and BL21 CodonPlus(DE3)-RIL (AtUCP2) as described previously (55). Control cultures with the empty vector were processed in parallel. Inclusion bodies were purified on a sucrose density gradient (56) and washed at 4°C, first with TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 7.0); once with a buffer containing 3% Triton X-114 (w/v), 1 mM EDTA, 10 mM PIPES (pH 7.0), and 10 mM Na 2 SO 4 ; and finally three times with TE buffer (57). The inclusion body proteins were solubilized in 2% lauric acid, 10 mM PIPES (pH 7.0), and 3% Triton X-114 (AtUCP1) or 1.6% sarkosyl (w/v), 10 mM PIPES, pH 7.0, and 0.6% Triton X-114 7 Please note that the JBC is not responsible for the long-term archiving and maintenance of this site or any other third party hosted site.

Reconstitution of AtUCP1 and AtUCP2 into liposomes and transport measurements
The solubilized recombinant proteins were reconstituted into liposomes by cyclic removal of the detergent with a hydrophobic column of Amberlite beads (Bio-Rad), as described previously (41). The reconstitution mixture contained solubilized proteins (about 6 g), 1% Triton X-114, 1.4% egg yolk phospholipids as sonicated liposomes, 10 mM substrate, 20 mM PIPES (pH 7.0), 1 mg of cardiolipin, and water to a final volume of 700 l. These components were mixed thoroughly, and the mixture was recycled 13 times through a Bio-Beads SM-2 column preequilibrated with a buffer containing 10 mM PIPES (pH 7.0) and 50 mM NaCl and the substrate at the same concentration used in the starting mixture.
External substrate was removed from proteoliposomes on a Sephadex G-75 column pre-equilibrated with 10 mM PIPES and 50 mM NaCl, pH 7.0. Transport at 25°C was initiated by adding the indicated radioactive substrates (American Radiolabeled Chemicals Inc. or PerkinElmer Life Sciences) to substrateloaded (exchange) or empty (uniport) proteoliposomes. Transport was terminated by adding 20 mM pyridoxal 5Ј-phosphate and 20 mM bathophenanthroline, which in combination inhibit the activity of several MCs completely and rapidly (58 -60). In controls, the inhibitors were added at the beginning together with the radioactive substrate according to the "inhibitor-stop" method (41). Finally, the external substrate was removed, and the radioactivity in the proteoliposomes was measured. The experimental values were corrected by subtracting control values. The initial transport rates were calculated from the radioactivity incorporated into proteoliposomes in the initial linear range of substrate transport. The kinetic constants K m , V max , and K i were determined from Lineweaver-Burk and Dixon plots. For efflux measurements, proteoliposomes containing 5 mM internal aspartate or malate were loaded with 5 M [ 14 C] aspartate and [ 14 C]malate, respectively, by carrier-mediated exchange equilibrium (61,62). The external radioactivity was removed by passing the proteoliposomes through Sephadex G-75 columns pre-equilibrated with 50 mM NaCl. Efflux was started by adding unlabeled external substrate or buffer alone and terminated by adding the inhibitors indicated above.

Cloning and transient expression of GFP fusion constructs
For subcellular localization of AtUCP1 and AtUCP2, the AtUCP1-GFP and the AtUCP2-GFP fusion constructs were prepared. The AtUCP1 coding sequence was amplified via Phusion High-Fidelity DNA Polymerase (New England Biolabs) using primers BH254 and BH255 (Table S4) and cloned with the Gibson Assembly Cloning Kit (New England Biolabs) into the expression vector pTKan (63) using the restriction sites ApaI and SacII. The GFP coding sequence for C-terminal GFP fusion was inserted via SacII and SpeI into the pTKan vector. The final vector contains AtUCP1 with a C-terminal GFP (AtUCP1-GFP) under the control of an optimized Arabidopsis ubiquitin-10 promoter (64) and the terminator of the nos gene from Agrobacterium tumefaciens. The AtUCP2 coding sequence was amplified via Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientific) using primers UCP2_BPF and UCP2_BPR-s (Table S4) and cloned into pDONR207 vector by a BP recombination reaction with BP Clonase II enzyme mix (Invitrogen). The resulting entry clone was used for an LR recombination reaction by LR Clonase II enzyme mix (Invitrogen) with pK7FWG2 destination vector (65) to construct expression vector for the expression of AtUCP2-GFP under the control of the 35S promoter.

Isolation, generation, and molecular characterization of single-and double-knockout mutants of ucp1 and ucp2
T-DNA insertion lines for AtUCP1 (SAIL_536G01, referred to as ucp1 (37)) and AtUCP2 (SALK_080188, referred to as ucp2) were obtained from the ABRC. To identify homozygous T-DNA insertion lines, genomic DNA was extracted and genotyped using gene-specific primer pairs (DG8/DG9 for ucp1 and DG6/DG5 for ucp2) and a primer pair for T-DNA/gene junction (DG9/SAIL-LBa for ucp1 and DG6/SALK-LBa1 for ucp2) ( Table S4). The position of the T-DNA was checked by sequencing. Homozygous ucp1 and ucp2 were further propagated. To generate double mutants (referred to as dKO), homozygous ucp1 and ucp2 were crossed. Heterozygous plants were selected by PCR in the T1 generation. After self crossing, dKO were selected by PCR and further propagated.
Total RNA was extracted from wildtype, mutant, and transgenic Arabidopsis plants using the guanidinium thiocyanatephenol-chloroform method (70) and subjected after DNase treatment (RQ1 RNase-Free DNase, Promega) to cDNA synthesis (Superscript II Rnase H-reverse transcriptase, Invitrogen). Gene expression of AtUCP1 and AtUCP2 were analyzed using gene-specific primer pairs. The following primer sets

Transport properties of AtUCP1 and AtUCP2
were used: DG23/DG24 for AtUCP1 and DG25/DG26 for AtUCP2 (Table S4). As a control for cDNA quality and quantity, a cDNA fragment of an actin gene (ACT7, At5g09810) was amplified using ML167 and ML168. PCR conditions were as follows: 94°C for 2 min, followed by cycles of 94°C for 30 s, 58°C for 45 s, 72°C for 60 s, and a final extension for 2 min. Products were visualized on an ethidium bromide-stained 1% agarose gel.

Metabolite profiling
To obtain a broad overview of the major pathways of central metabolism, an established GC-MS-based metabolite profiling method was used to quantify the relative metabolite levels in the Arabidopsis rosette (ϳ50 mg fresh weight). The extraction, derivatization, standard addition, and sample injection were performed exactly as described previously (71). This analysis allowed the determination of 46 different compounds, representing the main classes of primary metabolites (i.e. amino acids, organic acids, and sugars).

Other methods
Proteins were analyzed by SDS-PAGE and stained with Coomassie Blue dye. The identity of the bacterially expressed, purified AtUCP1 and AtUCP2 was assessed by MALDI-TOF mass spectrometry of trypsin digests of the corresponding band excised from Coomassie-stained gels (25,72). The amount of purified AtUCP1 and AtUCP2 proteins was estimated by laser densitometry of stained samples using carbonic anhydrase as a protein standard (73). The amount of protein incorporated into liposomes was measured as described (74) and was about 25% of protein added to the reconstitution mixture. K ϩ -diffusion potentials were generated by adding valinomycin (1.5 g/mg phospholipid) to proteoliposomes in the presence of K ϩ gradients. For the formation of an artificial ⌬pH (acidic outside), nigericin (50 ng/mg phospholipid) was added to proteoliposomes in the presence of an inwardly directed K ϩ gradient.