The Sequence, Bacterial Expression, and Functional Reconstitution of the Rat Mitochondrial Dicarboxylate Transporter Cloned via Distant Homologs in Yeast and Caenorhabditis elegans *

The dicarboxylate carrier (DIC) belongs to a family of transport proteins found in the inner mitochondrial membranes. The biochemical properties of the mammalian protein have been characterized, but the protein is not abundant. It is difficult to purify and had not been sequenced. We have used the sequence of the distantly related yeast DIC to identify a related protein encoded in the genome of Caenorhabditis elegans. Then, related murine expressed sequence tags were identified with the worm sequence, and the murine sequence was used to isolate the cDNA for the rat homolog. The sequences of the worm and rat proteins have features characteristic of the family of mitochondrial transport proteins. Both proteins were expressed in bacteria and reconstituted into phospholipid vesicles where their transport characteristics closely resembled those of whole rat mitochondria and of the rat DIC reconstituted into vesicles. As expected from the role of the DIC in gluconeogenesis and ureogenesis, its transcripts were detected in rat liver and kidney, but unexpectedly, they were also detected in rat heart and brain tissues where the protein may fulfill other roles, possibly in supplying substrates to the Krebs cycle.

The dicarboxylate carrier (DIC) 1 is a component of the inner membranes of mitochondria that plays an important role in gluconeogenesis, urea synthesis, and sulfur metabolism, particularly in liver, by transporting dicarboxylates such as malate and succinate across the inner membrane in exchange for phosphate, sulfate, sulfite, or thiosulfate (1)(2)(3)(4). The transport and biochemical characteristics of the rat protein have been studied extensively (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14), but because the protein is not abundant, neither its sequence nor the coding sequence has been established until now. To solve this problem, as described below, we took advantage of a protein encoded in Saccharomyces cerevisiae which has been shown to be a DIC (15,16). Its sequence has features characteristic of members of a family of membrane proteins that are responsible for the transport of metabolites involved in oxidative phosphorylation and in other important functions in mitochondria. The yeast genome contains 35 members of this family (15). In all species that have been investigated, the sequences of family members are made of three related domains repeated in tandem about 100 amino acid in length, each probably being folded into two antiparallel transmembrane ␣-helices linked by an extensive hydrophilic sequence. The three repeats are joined together by shorter hydrophilic sequences. The repeats in the various family members are all related to each other, and several characteristic sequence features are conserved (17)(18)(19)(20).
Comparisons of sequences of members of the family of known biochemical function in yeast with those of their mammalian counterparts show that the homologous pairs are too distant for the yeast sequences (protein or DNA) to provide a feasible basis for cloning the mammalian proteins. Therefore, the sequence of the S. cerevisiae DIC has been compared with the proteins encoded in the available genomic sequences of Caenorhabditis elegans. Currently, it encodes 32 members of the mitochondrial transporter family, and the two most closely related proteins (named K11G12.5 and B0432.4) were identical to the yeast DIC in 37% and 30% of their residues, respectively. The K11G12.5 protein has been expressed in Escherichia coli, reconstituted into phospholipid vesicles, and demonstrated to be a DIC. Then the sequence of the C. elegans DIC was compared with protein sequences encoded in mammalian expressed sequence tags. In this way, two partial murine expressed sequence tags (AA199557 and AA041737) were found to encode a fragment of a closely related protein, and synthetic oligonucleotides based on their sequences were used to amplify a DNA segment from rat liver cDNA. Using this segment, the complete rat cDNA was then obtained by PCR experiments. The encoded protein was shown to have transport specificity and other biochemical characteristics that were identical to those of the DIC isolated from rat liver mitochondria.
Synthetic Oligonucleotides-Adaptor primers AP1, AP2, and CDS were purchased from CLONTECH (Palo Alto, CA). Oligonucleotides were made with a DNA synthesizer (Applied Biosystems model 320B). Some of them had additional 5Ј-linkers and were employed as primers in PCRs. Others were used as hybridization probes for the reaction products, as described before (21). Some unique 17-base oligonucleo-tides were used as sequencing primers.
DNA Sequence Analysis-An automatic DNA sequenator (Applied Biosystems, Prism 377) was employed with dye terminator kits (Perkin-Elmer). Sequence errors from PCRs were avoided by sequencing at least three independent clones in both directions from each reaction. Ambiguities were resolved by sequencing more clones. Data bases were compiled and analyzed with the program Autoassembler (Perkin-Elmer).
Rat cDNA Fragments for the DIC-Touchdown PCRs (22) were conducted in a thermal cycler (Perkin-Elmer model 480). The reaction mixtures (70 l) contained adaptor-ligated double-stranded rat liver cDNA (1 ng, CLONTECH) with primers (final concentration 0.5 M). After an initial incubation at 94°C for 2 min without Taq polymerase, the PCR cycle was denaturation at 94°C (30 s) and annealing at 72°C (4 min). These steps were repeated 5 times. Then the cycle was repeated for 5 and 20 times with both annealing and extension temperatures at 70°C or 68°C, respectively. The entire procedure was repeated once again with each of the previous final PCR mixtures (1 l) as template, with two nested primers (final concentration 1 M), and 2 units of Taq polymerase. The cDNA sequence was extended to the 5Ј-end with primers AP1 and AP2 (modified by substituting the NotI restriction site by EcoRI) and to the 3Ј-end with the primer CDS. The products were identified and recovered from agarose gels as described before (21). The EcoRI and BamHI sites in the linkers were digested, and the products were cloned into the pUC19 vector.
Construction of the Rat and C. elegans DIC Expression Plasmids-Full-length cDNAs for the rat and C. elegans DIC were amplified by 60 or 30 cycles of PCR, respectively, as described before (21), except that the annealing temperature was 60°C. The templates were rat liver cDNA (CLONTECH) and for C. elegans, phage DNA (1 ng) from clone cm15f11 (23). The forward and reverse primers were nucleotides 34 -55 and 869 -894 of the rat cDNA (see Fig. 2), with 5Ј-NdeI and EcoRI restriction sites, respectively, and nucleotides 15-38 and 863-886 of clone cm15f11, with additional 5Ј-NdeI and BamHI restriction sites, respectively. The amplified products, 0.9 kb in length, were purified (with Qiaquick from Qiagen), cloned into the pMW7 expression vector, and the constructs were transformed into E. coli DH5␣ cells. Transformants selected on 2xTY plates containing ampicillin (100 g/ml) were screened by direct colony PCR and by restriction digestion of purified plasmids. The sequences of inserts were verified.
Bacterial Expression-The rat and C. elegans DICs were overexpressed at 37°C in E. coli CO214(DE3), a mutant of E. coli C41(DE3) selected for its ability to overexpress the mitochondrial phosphate carrier without toxicity, 2 as described before for improvement of host strains (24). Control cultures with the empty vector were processed in parallel. Inclusion bodies were purified according to Fiermonte et al. (25). The pellets were washed twice with buffer containing 10 mM PIPES, pH 7.0, Triton X-114 (3%, w/v), and 1 mM EDTA and then once with TE buffer. Then the proteins were solubilized in a solution of 10 mM Tris-HCl, pH 7.0, sarkosyl (1.8%, w/v), and 0.1 mM EDTA. A small insoluble residue was removed by centrifugation at 258,000 ϫ g for 1 h at 4°C.
Samples taken at various points of growth of bacterial cultures and purified inclusion bodies were examined by SDS-polyacrylamide gel electrophoresis in 17.5% gels (26). Proteins were either stained with 2 G. Fiermonte and J. E. Walker, unpublished work. Coomassie Blue dye or transferred to polyvinylidene difluoride membranes, stained with Coomassie Blue dye, and their NH 2 termini sequenced (Applied Biosystems, model 477A). The yield of purified protein was estimated by laser densitometry (LKB 2202 Ultroscan) of SDS-polyacrylamide gels stained with Coomassie Blue dye, with carbonic anhydrase as standard.
Reconstitution into Liposomes-Sarkosyl-soluble proteins were diluted 10-fold with a buffer containing 2 mM PIPES, pH 7.0, 0.6% (w/v) Triton X-114, and 0.2 mM EDTA and reconstituted into liposomes in the presence of appropriate substrates, as described before (25), except that the reconstitution mixture contained cardiolipin (1 mg/ml). The extent of incorporation of the protein into liposomes was estimated as follows. The proteoliposomes were passed through a Sephadex G-75 column, centrifuged at 300,000 ϫ g for 30 min, precipitated with cold acetone, and delipidated with organic solvents (27). The protein was dissolved in 0.1% SDS, and after SDS-polyacrylamide gel electrophoresis, the amount of the DIC was measured by laser densitometry of stained samples, as described above.
Activity Assays-External substrates were removed from proteoliposomes on Sephadex G-75 (28). Transport at 25°C was started by adding [ 14 C]malate or 33 P to the eluted proteoliposomes and, unless otherwise indicated, terminated after 20 s (C. elegans DIC) or 40 s (rat DIC) by the addition of either 30 mM butylmalonate or a mixture containing 30 mM pyridoxal 5Ј-phosphate and 10 mM bathophenanthroline, (this is the "inhibitor-stop" method (28)). In controls, the inhibitor was added with the labeled substrate. Extraliposomal labeled substrate was removed from quenched samples on Dowex AG1-X8, and eluted radioactivity was measured (28). Transport activities were calculated from the experimental values minus the controls. Various other transport activities were also assayed by the inhibitor-stop method.
Northern and Southern Blotting-Total RNA was extracted from rat liver, kidney, heart, and brain according to Chirgwin et al. (29). The hybridization experiments were performed as described before (30), except that the probe was nucleotides 34 -894 of the rat cDNA sequence (see Fig. 2). Samples were autoradiographed at Ϫ70°C for 15 days with an intensifying screen. Hybridization signals were normalized with an 850-base pair probe encoding part of rat glyceraldehyde-3-phosphate dehydrogenase (31) employed under the same conditions as described above.
Samples (10 g) of rat liver genomic DNA (32) were digested for 6 h at 37°C with one of the restriction enzymes NcoI, EcoRI, or BamHI. DNA fragments were separated on 0.8% agarose gels, transferred to a Hybond-N membrane (33), and hybridized with a probe from the cDNA for the rat DIC (nucleotides 91-536; see Fig. 2) using the same procedures described for Northern blotting.
Homology Searches-Homologs of the yeast DIC in C. elegans were identified by screening the nematode (the Sanger Center, Hinxton, U. K.) and the National Center for Biotechnology Information data bases with the sequence of the S. cerevisiae protein (15) using the BLASTP program. The EMBL nonredundant expressed sequence tag data base was probed with a protein sequence encoded in the nematode clone K11G12.5 using the TBLASTN program. The same program was used to search all available sequence data bases for murine fragments coding for protein sequences related to the putative rat DIC protein sequence. Amino acid sequences were aligned with ClustalW (version 1.7).

RESULTS AND DISCUSSION
Homologs of the S. cerevisiae DIC-The nematode gene product K11G12.5 is identical to the yeast DIC sequence (15,16) and to the bovine oxoglutarate-malate carrier proteins (34) in 37 and 35% of its residues, respectively. However, the bovine oxoglutarate-malate carrier protein is 64% identical to the C. elegans protein B0432.4, which is therefore more likely to be the worm oxoglutarate-malate carrier protein than K11G12.5. Two overlapping cDNA clones (AA199557 and AA041737) from Mus musculus contain 622 nucleotides of unique sequence, and the encoded protein sequence is 47% identical to part of the C. elegans K11G12.5 protein.
Cloning and Sequencing of the cDNA for the Rat DIC-Four nested oligonucleotides (1F, 2F, 1R, 2R) based on the sequence of the AA199557 and AA041737 cDNA sequences from M. musculus were used to amplify a segment from a rat liver cDNA (see PCR1 in Figs. 1 and 2). The product gave a single band of about 350 base pairs. Its nucleotide sequence is 94% identical to the M. musculus partial sequence. The rat cDNA sequence was completed by experiments PCR2 and PCR3 (see Fig. 1).
The final rat cDNA sequence of 1946 nucleotides (see Fig. 2) encodes a protein with a molecular mass of 31,455. The assignment of the proposed start codon is supported by the closely related murine sequence where the corresponding codon is preceded by an in-frame stop codon. The fragment of murine  4), with the coding sequence of the C. elegans DIC (lanes 2 and 5), and with the coding sequence of the rat DIC carrier (lanes 3 and 6). Samples were taken immediately before induction (lanes 1-3) and 5 h later (lanes 4 -6). The same number of bacteria was analyzed in each sample. Lane 7, purified C. elegans DIC (8 g); lane 8, purified recombinant rat DIC (6.5 g) obtained from the bacteria analyzed in lanes 5 and 6, respectively. protein sequence (132 amino acids) is 96% identical to the corresponding part of the rat sequence. The 5Ј-and 3Ј-untranslated regions are 33 and 1055 base pairs, respectively, and the sequence is terminated by a run of A residues separated by a 13-nucleotide sequence from the preceding AATAAA sequence, a typical signal for polyadenylation of mRNA (32).
Recent additions to sequence data bases have allowed us to assemble the entire coding sequence of the murine protein. In the alignment of the four homologous protein sequences (Fig.  3), the putative rat DIC is 96, 59, and 37% identical to the murine, C. elegans, and S. cerevisiae proteins, respectively.
Bacterial Expression of the Putative DIC Proteins from C. elegans and Rat-The overexpression of the C. elegans K11G12.5 protein and its rat homolog in E. coli C0214(DE3) (see Fig. 4, lanes 5 and 6) yielded about 60 and 45 mg of purified protein per liter of culture, respectively. Both purified proteins gave single bands on SDS-polyacrylamide gels (Fig. 4, lanes 7  and 8) with apparent molecular masses of 32 and 29.5 kDa, respectively (the corresponding calculated molecular masses including the initiator methionine are 32.23 and 31.55). Neither protein was detected in bacteria harvested immediately before induction of expression (lanes 1-3) nor in cells harvested after induction but lacking the coding sequence in the expression vector (lane 4). The determined NH 2 -terminal sequences (AEDKTKRLGR and AEARTSRWYF, respectively) of residues 1-10 of the purified 32-and 29.5-Da proteins are identical to those predicted for residues 2-11 of the C. elegans K11G12.5 protein and its rat homolog, respectively.
Transport Properties of the C. elegans K11G12.5 Protein and Its Rat Homolog-The proportion of rat and C. elegans proteins incorporated into liposomes in various experiments was usually in the range of 11-18% of the protein used in the experiment, and they were incorporated to approximately the same extent. Both proteins catalyzed a [ 14 C]malate/phosphate exchange that was inhibited by butylmalonate (see Table I), as does the DIC in rat mitochondria (1)(2)(3)(4)(5)(6)(7)(8). No such activity was found by reconstitution of sarkosyl-solubilized material from bacterial cells either lacking the expression vector, or har-vested immediately before induction of overexpression. Neither reconstituted protein catalyzed the homoexchange of pyruvate, citrate, oxoglutarate, glutamate, aspartate, glutamine, carnitine, ornithine, ADP or ATP, all being substrates for various mitochondrial transporters.
The substrate specificity of the C. elegans and rat recombinant proteins was examined in greater detail by measuring the uptake of [ 14 C]malate and 33 P into proteoliposomes that had been preloaded with a variety of substrates (Table I). In the absence of internal substrate, uptake of labeled malate or phosphate was not observed. The highest activities were detected in the presence of internal L-malate, phosphate, and malonate. To a somewhat lower extent, succinate, sulfate, and thiosulfate were exchanged for both external malate and phosphate. Virtually no exchange was observed with internal oxoglutarate, fumarate, glutamate, aspartate, citrate, ATP, pyruvate, carnitine, or ornithine. The residual activity in the presence of these substrates was approximately the same as the activity observed in the presence of sodium chloride. Therefore, proteoliposomes reconstituted with the C. elegans and rat recombinant proteins import external malate and phosphate only in exchange for certain dicarboxylates, inorganic sulfur-containing anions, and phosphate, similar to the rat DIC in liver mitochondria (1-8).
The [ 14 C]malate/phosphate exchange reaction of both proteins was inhibited by the substrate analogs butylmalonate,  C. elegans K11G12.5 protein or its rat homolog Proteoliposomes were preloaded with 20 mM phosphate, and transport was started by adding 0.5 mM [ 14 C]malate. The incubation time was 20 s and 40 s for the reconstituted C. elegans and rat proteins, respectively. Thiol reagents, pyridoxal 5Ј-phosphate, carboxyatractyloside, and ␣-cyanocinnamate were added 2 min before the labeled substrate; the other inhibitors and external substrates were added together with [ 14 C]malate. All inhibitors and substrates were used at a concentration of 10 mM except for organic mercurials, carboxyatractyloside and ␣-cyanocinnamate (0.1 mM) and N-ethylmaleimide (0.1 or 1 mM) benzylmalonate, and phenylsuccinate (Table II). Butylmalonate was more effective than phenylsuccinate, as with the DIC in mitochondria (3,6), whereas the reverse is true for the oxoglutarate-malate carrier protein (35). The activities of both recombinant proteins were also inhibited by organic mercurials, pyridoxal 5Ј-phosphate and bathophenanthroline, all of them inhibitors of the rat DIC, both in mitochondria (3,5,8,36) and after purification and reconstitution into liposomes (9,10,13,14). The yeast DIC also has similar properties (15,16,37). Inhibitors of other characterized mitochondrial carriers, such as carboxyatractyloside, phthalonate, and ␣-cyanocinnamate, had little or no effect on the activities of the C. elegans and the rat reconstituted proteins. In contrast to the recombinant C. elegans protein, the rat protein is more sensitive to organic mercurials and is also inhibited by 1 mM N-ethylmaleimide to some extent. The rat DIC, both in mitochondria and after purification, and the yeast DIC, are insensitive to N-ethylmaleimide (2-5, 8 -10, 14 -16, 37), but this point requires further clarification. There are at least two possible explanations for this partial inhibition of the rat recombinant protein by N-ethylmaleimide. First, the expressed rat protein might differ from the native DIC by the absence of disulfide bridges (although there is no evidence for their presence in the native rat protein). Second, both rat liver mitochondria and the DIC purified from mitochondria might contain more than one isoform of DIC, one being relatively insensitive to N-ethylmaleimide.
The ability of nonradioactive potential substrates to inhibit the reconstituted [ 14 C]malate/phosphate exchange of both the C. elegans and rat recombinant proteins was also examined. The exchange was prevented by external addition of any one of the known substrates of the DIC (Table I), and it was unaffected by substrates of other mitochondrial carriers, such as fumarate, oxoglutarate, ADP, or citrate.
The dependence of the [ 14 C]malate/malate and 33 P/malate exchange rates on substrate concentration was investigated by changing the concentration of externally added [ 14 C]malate or 33 P at a constant internal concentration of 10 mM malate (see Fig. 5). In liposomes reconstituted with either the C. elegans or the rat protein (Fig. 5, A and B) linear functions were obtained which intersected the ordinate close to a common point. The different slopes indicate that the half-saturation constant (K m ) for phosphate is about three times higher than for malate (in liposomes reconstituted with the C. elegans protein, the mean values from four experiments are 0.54 Ϯ 0.10 and 1.52 Ϯ 0.36 mM for malate and phosphate, respectively; for the rat protein the corresponding values are 0.78 Ϯ 0.14 and 1.77 Ϯ 0.27). Thus, both the C. elegans and rat recombinant proteins are also similar in these respects to the rat liver DIC (K m values for malate and phosphate are 0.26 mM and 1.5 mM, respectively, in  The RNA samples were from liver (L), heart (H), and brain (B). About 20 g of total RNA was loaded in each slot. After 15 days of autoradiography at Ϫ70°C, the blot in A was stripped by boiling in 0.1% (w/v) SDS for 3 min and rehybridized with the glyceraldehyde 3-phosphate dehydrogenase probe. On the left side of panel A are the positions of 28 S and 18 S ribosomal RNAs (4,800 and 1,900 nucleotides in length, respectively). mitochondria (2) and 0.49 mM and 1.41 mM, respectively, after purification (11)). The K m values of the recombinant yeast DIC for malate and phosphate are 0.59 and 1.65 mM, respectively (15). In four experiments, the V max value, corrected for small differences in efficiency of reconstitution, was 149.3 Ϯ 38.6 mmol/min/g of C. elegans protein and 44.7 Ϯ 12.7 mmol/min/g of rat protein. These values are virtually the same for both the malate/malate and phosphate/malate exchanges and are independent of the type of substrate, as observed before (2,11,15). Furthermore, the V max values of the recombinant rat and C. elegans proteins are higher than those determined previously for the purified rat liver and recombinant yeast DIC (11,15). This difference may arise because the efficiency of reconstitution was not quantified in the earlier studies.
Repetitive Sequences within the DIC and Hydrophobicity Analysis-The sequences of the DICs consist of three homologous tandem repeats that are distantly related in sequence to repeats in other mitochondrial carrier proteins. Each repetitive element appears to be made of two antiparallel transmembrane hydrophobic ␣-helices linked by an extensive polar region. Therefore, they have the tripartite structure that is characteristic of other mitochondrial transporters (17)(18)(19)(20)38). Each of the three domains contains the sequence motif characteristic of the mitochondrial carriers, namely P-h-D/E-X-h-K/R-X-R/K-(20 -30 residues)-D/E-G-(4 residues)-a-K/R-G (where h and a represent hydrophobic and aromatic amino acids, respectively, and X is unknown ).
Number of Genes for the DIC in Rat-Some mitochondrial carriers are unique proteins encoded by single genes, whereas others have closely related isoforms either encoded by multiple genes (39 -48) or generated by alternative splicing from a single gene (21). Because only single hybridizing bands were detected by Southern hybridization (see Fig. 6), it seems likely that the rat DIC has a single gene, although this experiment would not detect isoforms arising by alternative splicing or isoforms with distantly related coding sequences.
Expression of the DIC in Rat-Among the tissues that were examined, the rat DIC was expressed most strongly in liver. Somewhat lower levels were detected in brain, and much less was found in heart (see Fig. 7). It is notable that the DIC transcript, estimated to be about 2.1 kb, is larger than the cDNA sequence that has been determined, indicating that the transcript may have an extensive 5Ј-noncoding region. High levels of mRNA for the DIC in liver and kidney (not shown) are probably related to gluconeogenesis and ureogenesis, which occur mainly in these tissues. At present, there is no explanation for the relatively high levels of the DIC mRNA in brain and for its presence in heart where gluconeogenesis and ureogenesis do not operate. So the DIC may have other functions in these tissues, possibly exclusively catalyzing the supply of substrates to the Krebs cycle. It remains to be demonstrated that the DIC mRNA is translated in heart and brain.