Identification of a Novel Na+-independent Acidic Amino Acid Transporter with Structural Similarity to the Member of a Heterodimeric Amino Acid Transporter Family Associated with Unknown Heavy Chains*

We identified a novel Na+-independent acidic amino acid transporter designated AGT1 (aspartate/glutamatetransporter 1). AGT1 exhibits the highest sequence similarity (48% identity) to the Na+-independent small neutral amino acid transporter Asc (asc-type amino acid transporter)-2 a member of the heterodimeric amino acid transporter family presumed to be associated with unknown heavy chains (Chairoungdua, A., Kanai, Y., Matsuo, H., Inatomi, J., Kim, D. K., and Endou, H. (2001) J. Biol. Chem. 276, 49390–49399). The cysteine residue responsible for the disulfide bond formation between transporters (light chains) and heavy chain subunits of the heterodimeric amino acid transporter family is conserved for AGT1. Because AGT1 solely expressed or coexpressed with already known heavy chain 4F2hc (4F2 heavy chain) or rBAT (related tob 0,+-amino acidtransporter) did not induce functional activity, we generated fusion proteins in which AGT1 was connected with 4F2hc or rBAT. The fusion proteins were sorted to the plasma membrane and expressed the Na+-independent transport activity for acidic amino acids. Distinct from the Na+-independent cystine/glutamate transporter xCT structurally related to AGT1, AGT1 did not accept cystine, homocysteate, andl-α-aminoadipate and exhibited high affinity to aspartate as well as glutamate, suggesting that the negative charge recognition site in the side chain-binding site of AGT1 would be closer to the α-carbon binding site compared with that of xCT. The AGT1 message was predominantly expressed in kidney. In mouse kidney, AGT1 protein was present in the basolateral membrane of the proximal straight tubules and distal convoluted tubules. In the Western blot analysis, AGT1 was detected as a high molecular mass band in the nonreducing condition, whereas the band shifted to a 40-kDa band corresponding to the AGT1 monomer in the reducing condition, suggesting the association of AGT1 with other protein via a disulfide bond. The finding of AGT1 and Asc-2 has established a new subgroup of the heterodimeric amino acid transporter family whose members associate not with 4F2hc or rBAT but with other unknown heavy chains.

In the past, a large number of amino acid transport systems in mammals have been distinguished based on differences in substrate selectivity and ion dependence (1). For the last decade, molecular cloning approaches have revealed the molecular nature of amino acid transport systems (2). The amino acid transporters identified so far exhibit a variety of substrate selectivity and are composed of the members of several transporter families. Among them, the heterodimeric amino acid transporter family, a subfamily of SLC7, is unique in two aspects (3,4). First, the members of this family are linked via a disulfide bond with single membrane spanning type II membrane glycoproteins such as 4F2hc (4F2 heavy chain) and rBAT (related to b 0,ϩ -amino acid transporter) (3,4). 4F2hc is the heavy chain of the cell surface antigen 4F2 (CD98) (5,6). The 4F2 antigen is a heterodimeric protein composed of two subunits, an ϳ80-kDa glycosylated heavy chain and a ϳ40-kDa nonglycosylated light chain (5,6). The 4F2 light chain has been revealed to be an amino acid transporter. Six proteins have so far been identified to be 4F2 light chains to form transporters subserving systems L, y ϩ L, x Ϫ C , or asc (7)(8)(9)(10)(11)(12)(13)(14)(15)(16). In addition, a member of the heterodimeric amino acid transporter family has been identified that couples with the other type II membrane glycoprotein rBAT to form a system b 0,ϩ amino acid transporter (17)(18)(19). The conserved cysteine residue in the predicted extracellular loop between transmembrane domains 3 and 4 is responsible for the disulfide bond formation between transporter proteins (light chains) and the heavy chains (20). Second, the heterodimeric amino acid transporter family is distinctive for its diversity in the substrate selectivity of its members. As already mentioned, they include transporters for neutral amino acids (systems L and asc), acidic amino acids as well as cystine (system x Ϫ C ), and both neutral and basic amino acids (systems y ϩ L and b 0,ϩ ) (4).
Recently, we identified a transporter designated Asc-2 (asctype amino acid transporter 2) that exhibited relatively low but significant sequence similarity to the members of the heterodimeric amino acid transporter family (21). Asc-2, however, does not associate with 4F2hc or rBAT and is presumed to link to unknown heavy chains. Although Asc-2 itself is not sorted to the plasma membrane when expressed in Xenopus oocytes, the fusion proteins in which Asc-2 is connected with rBAT or 4F2hc appeared on the plasma membrane and exhibits the functional properties corresponding to those of the transporter subserving Na ϩ -independent small neutral amino acid transport system asc (21). In the present study, we have identified a novel transporter protein structurally related to Asc-2. The transporter is also proposed to associate with an additional protein, presumably through a conserved cysteine residue to form a functional complex. We have generated fusion proteins in which the transporter protein is connected with rBAT or 4F2hc and shown that they appear on the plasma membrane and exhibit the Na ϩ -independent transport activity with distinct selectivity for acidic amino acids.

EXPERIMENTAL PROCEDURES
cDNA Cloning of AGT1-The cDNA for a mouse expressed sequence tag (GenBankTM accession number AI314100) showing nucleotide sequence similarity to rat BAT1 (17) was obtained from the Integrated and Molecular Analysis of Genomes and their Expression (IMAGE). The 1.8-kb XhoI fragment was excised from the cDNA (IMAGE cDNA clone number 1907807), labeled with [ 32 P]dCTP ( T7 Quick prime; Amersham Biosciences), and used as a probe for screening a mouse kidney cDNA library (22). The oligo(dT)-primed cDNA library was prepared from mouse kidney poly(A) ϩ RNA using the Superscript Choice System (Invitrogen) (23). The synthesized cDNA was ligated to ZipLox EcoRI arms (Invitrogen). Screening of the library and the isolation of positive plaques were performed as described elsewhere (22). The cDNAs in positive ZipLox phages were rescued into plasmid pZL1 by in vivo excision in accordance with the manufacturer's instructions (Invitrogen). The cDNA insert was subcloned into pcDNA3.1(ϩ) (Invitrogen) at a NotI site. The cDNA was sequenced in both directions by the dye terminator cycle sequencing method (PerkinElmer Life Sciences and Applied Biosystems). Transmembrane regions of proteins were predicted based on the SOSUI algorithm (24).
Construction of Fusion Proteins-Fusion proteins were constructed as described elsewhere with some modifications (21). To generate a AGT1-rBAT fusion protein, AGT1 cDNA fragment was amplified by PCR using a sense primer corresponding to nucleotides 4 -23 of AGT1 cDNA sequence extended at its 5Ј end by adding a HindIII restriction site and GCGCG (5Ј-GCGCGAAGCTTACCTATAGGCAGAAACATTC-3Ј) and a reverse primer corresponding to the end of the coding sequence extended at its 5Ј end by adding a NotI restriction site and ATAT (5Ј-ATATGCGGCCGCACTTTCTTCATGTATGTGGT-3Ј). The PCR product was digested with HindIII and NotI and ligated to the HindIII and NotI sites of a mammalian expression vector pcDNA3.1(ϩ) (Invitrogen). Mouse rBAT cDNA was amplified using a sense primer corresponding to the coding sequence starting just after the start codon (ATG) extended at its 5Ј end by adding NotI restriction site and ATAT (5Ј-ATATGCGGCCGCAGATGAGGACAAAGGCAAGAG-3Ј) and a reverse primer corresponding to nucleotides 2228 -2251 of mouse rBAT cDNA sequence (GenBank TM accession number NM009205) (25) extended at its 5Ј end by adding by XbaI restriction site and GCGCGC (5Ј-GCGCGCTCTAGAAATGCTTTAGTATTTGGCATAATC-3Ј).
The PCR product was digested with NotI and XbaI and then introduced into the vector containing AGT1 (see above) precleaved with NotI and XbaI. The amino acid sequence around the junction of the resultant fusion protein was reduced to EESAAADED. Three amino acids (AAA) were inserted between the C terminus of AGT1 and the N terminus of rBAT in which the first methionine residue was omitted.
For the AGT1-4F2hc fusion protein, mouse 4F2hc cDNA was amplified using a sense primer corresponding to the coding sequence starting just after the start codon (ATG) extended at its 5Ј end by adding a NotI restriction site and ATAT (5Ј-ATATGCGGCCGCAAGCCAGGACAC-CGAAGTGGA-3Ј) and a reverse primer corresponding to nucleotides 1820 -1838 of mouse 4F2hc cDNA sequence (GenBank TM accession number AB023408) (16) extended at its 5Ј end by adding an XbaI restriction site and GCGC (5Ј-GCGCTCTAGACATGAGGCAGGGGT-GATGTTTT-3Ј). The PCR product was digested with NotI and XbaI and introduced into the vector containing AGT1 (see above) predigested with NotI and XbaI. The amino acid sequence around the junction of the AGT1-4F2hc fusion protein was reduced to EESAAASQD. Three amino acids (AAA) were inserted between the C terminus of AGT1 and the N terminus of 4F2hc, in which the first methionine residue was omitted.
Xenopus Oocyte Expression-cRNA for AGT1 was obtained by in vitro transcription using T7 RNA polymerase for cDNA of AGT1 in pcDNA3.1(ϩ) (Invitrogen) linearized with HindIII as described elsewhere (26). cRNAs for AGT1-rBAT fusion protein and AGT1-4F2hc fusion protein were obtained by in vitro transcription using T7 RNA polymerase for cDNAs of AGT1-rBAT fusion protein and AGT1-4F2hc fusion protein in pcDNA3.1(ϩ) (Invitrogen) linearized with ApaI. The Xenopus oocyte expression studies and uptake measurements were performed as described previously (27). The uptake of 14 C-labeled amino acids were measured 2 days after injection of cRNAs. Twentyfive nanograms of cRNAs were injected to each oocyte. For coexpression of AGT1 and mouse 4F2hc or mouse rBAT, 12 ng of AGT1 cRNA and 13 ng of mouse 4F2hc cRNA (16) or mouse rBAT cRNA (25) were injected into oocytes.
Amino Acid Uptake Measurements in Xenopus Oocytes-Groups of six to nine oocytes were incubated in 500 l of standard uptake solution (100 mM NaCl, 2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, and 5 mM Tris, pH 7.4) or Na ϩ -free uptake solution in which NaCl in the standard uptake solution was replaced by choline chloride containing 0.5-3.0 Ci of the radiolabeled compounds (12). For Cl Ϫ -free uptake solution, Cl Ϫ in the standard uptake solution was replaced by gluconate anion. Preliminary experiments to determine the time course of [ 14 C]Laspartic acid (300 M) uptake into oocytes expressing the AGT1-4F2hc and AGT1-rBAT1 fusion proteins indicated that the uptake was linearly dependent on incubation time up to 30 min (data not shown). Therefore, in all of the subsequent experiments, the uptake levels were measured over 30 min, and the values were expressed as pmol/oocyte/min.
The K m and V max values of amino acid substrates were determined using an Eadie-Hofstee plot based on the amino acid uptakes mediated by the AGT1-4F2hc and AGT1-rBAT fusion proteins measured at 1, 3, 10, 30, 100, and 300 M. The amino acid uptakes mediated by AGT1-4F2hc fusion protein or AGT1-rBAT fusion protein were calculated as differences between the means of the uptakes by the oocytes injected with cRNAs for the fusion proteins and those of the control oocytes injected with water. For the uptake measurements in the present study, six to nine oocytes were used for each data point. Each data point in the figures represents the mean Ϯ S.E. of uptake (n ϭ 6 -9). To confirm the reproducibility of the results, three separate experiments using different batches of oocytes and in vitro transcribed cRNA were performed for each measurement. The results from the representative experiments are shown in the figures.
Functional Expression in COS-7 Cells-cDNAs for AGT1 were subcloned in the mammalian expression vector pcDNA3.1(ϩ) (Invitrogen) as described above. cDNAs for mouse 4F2hc and mouse rBAT were also subcloned into the pcDNA3.1(ϩ) at EcoRI site and at EcoRI and ApaI sites, respectively. For functional expression, 1 g of these plasmids were solely transfected or cotransfected to COS-7 cells, which were grown in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, by using LipofectAMINE TM 2000 reagent (Invitrogen) as described elsewhere (28). Amino acid uptake measurements were performed at 48 h after transfection as described elsewhere (17).
Anti-peptide Antibody-An oligopeptide (CIPDVSDDHIHEES) corresponding to amino acid residues 465-478 of AGT1 was synthesized. The N-terminal cysteine residue was used for conjugation with keyhole limpet hemocyanin. An anti-peptide antibody was produced as described elsewhere (29).
Immunofluorescence and Confocal Laser Microscopy-The immunofluorescence detection of AGT1 and the epitope of 4F2hc in AGT1-4F2hc fusion protein expressed in Xenopus oocytes was performed as described previously (21) with minor modifications. Briefly, 2 days after injection of cRNAs, Xenopus oocytes were fixed with 4% paraformaldehyde in phosphate buffer at 4°C overnight. The samples were dehydrated in graded alcohol series, embedded in paraffin, and cut into 3-m thick sections that were then deparaffinized and equilibrated with 0.05 M Tris-buffered saline containing 0.1% Tween 20. The sections were then treated with 5% goat serum as a blocking agent for 45 min at room temperature and were then washed and incubated overnight with anti-AGT1 antiserum (1:250) or affinity-purified anti-4F2hc antibody (21) (1:100) at 4°C. Thereafter, they were treated with Alexa Fluor TM 488conjugated goat anti-rabbit IgG (Molecular Probes, Inc.; diluted 1:200) for 2 h at room temperature. The sections were then washed three times with 0.05 M Tris-buffered saline containing 0.1% Tween 20 and mounted with fluorescent mounting medium (DAKO, Carpinteria, CA).
The images were acquired using an Olympus Fluoview (FV 500) laser scanning confocal microscope (Olympus Optical, Tokyo, Japan). An Argon laser beam was used for excitation at 488 nm for Alexa Fluor TM 488 visualization. Emission from Alexa Fluor TM 488 was detected via BA505IF filter (30). For absorption experiments, the sections were treated with the primary antibodies in the presence of antigen peptides (200 g/ml) (21).
Immunohistochemistry-Three-micrometer paraffin sections of mouse kidney were processed for light microscopic immunohistochemical analysis as described previously (17). For immunostaining, the sections were incubated with anti-AGT1 antiserum (1:1,000) overnight at 4°C. Thereafter, they were treated with Envision (ϩ) rabbit peroxidase (DAKO) for 30 min. To detect immunoreactivity, the sections were treated with diaminobenzidine (0.8 mM) (17). For absorption experiments, the tissue sections were treated with the primary antibodies in the presence of antigen peptides (50 g/ml) (21). The sections were counterstained with hematoxylin.
Western Blot Analysis-The protein samples from mouse kidney were prepared as described elsewhere (21), with minor modifications. Briefly, the mouse kidney was homogenized in 9 volumes of 50 mM Tris-HCl (pH 7.5), 25 mM KCl, 1 mM MgCl 2 , 1 mM phenylmethylsulfonyl fluoride, and 0.25 M sucrose, with 15 strokes of a Dounce homogenizer. The homogenate was centrifuged for 10 min at 8,000 ϫ g, and the supernatant was centrifuged further for 1 h at 100,000 ϫ g. After centrifugation the membrane pellet was resuspended in 0.25 M sucrose, 100 mM KCl, 5 mM MgCl 2 , and 50 mM Tris (pH 7.4). The protein samples were heated at 100°C for 5 min in the sample buffer either in the presence or absence of 5% 2-mercaptoethanol and subjected to SDSpolyacrylamide gel electrophoresis. The separated proteins were transferred electrophoretically to a Hybond-P polyvinylidene difluoride transfer membrane (Amersham Biosciences). The membrane was treated with nonfat dried milk and diluted anti-AGT1 antiserum (1: 10,000) and then with horseradish peroxidase-conjugated anti-rabbit IgG as a secondary antibody (Jackson ImmunoResearch Laboratories, Inc.). The signals were detected with an ECL plus system (Amersham Biosciences). To verify the specificity of immunoreactions by absorption experiments, the membranes were treated with primary antibodies in the presence of antigen peptides (50 g/ml) (17).
Northern Blot Analysis-RNA was prepared from the tissues of 4 -5week-old Jcl:ICR male mice and placenta of mice with late pregnancy by the guanidinium isothiocyanate method using cesium trifluoroacetic acid (Amersham Biosciences) in accordance with the manufacturer's instructions. Poly(A) ϩ RNA (3 g/lane) selected by oligo(dT) cellulose chromatography (Amersham Biosciences) was separated on a 1% agarose gel in the presence of 2.2 M formaldehyde and was blotted onto a nitrocellulose filter (Schleicher & Schuell) as described elsewhere (31). The polymerase chain reaction-amplified fragment of AGT1 cDNA, corresponding to 43-1836 base pairs, was labeled with 32 P using the T7 QuickPrime kit (Amersham Biosciences). Hybridization was performed for 20 h at 42°C in 50% formamide. The final stringent wash of the filter was performed in 0.1ϫ SSC, 0.1% SDS at 65°C three times for 20 min.

Structural
Features of AGT1-A mouse cDNA clone with a 2,141-bp insert was isolated from a mouse kidney cDNA library. It contained an open reading frame from nucleotides 59 to 1,495 encoding a putative 478-amino acid protein, designated as AGT1 (aspartate/glutamate transporter 1). The start of the coding sequence was defined by the first ATG and the surrounding sequences (CTCTCAATGG) corresponding to the Kozak consensus translation initiation sequence (32). The cDNA includes the poly(A) tail (16 As), which starts 23 nucleotide downstream from a typical polyadenylation signal AATAAA at the nucleotide 2,103. The amino acid sequence of AGT1 was identical to that of BCO14684, which was in a GenBank TM data base but not functionally characterized. The AGT1 amino acid sequence exhibited remarkable sequence identity to that of mouse system asc transporter Asc-2 (48% identity), which is presumed to be associated with unknown heavy chains (21). AGT1 also exhibits sequence identity to rat system L transporters, LAT1 (35% identity) (7) and LAT2 (37%) (12), the mouse system asc transporter, Asc-1 (37%) (16), the y ϩ L transporters, rat y ϩ LAT1 (37%) (33) and human KIAA0245/y ϩ LAT2 (36%) (14,34), the mouse system x Ϫ C transporter, xCT (37%) (15), and the rat system b 0,ϩ transporter, BAT1 (36%) (17), all of which are associated with either 4F2hc or rBAT (Fig. 1). AGT1 also exhibited significant sequence identity to the system y ϩ transporters, CAT1-4 (30%) from mice and humans (35) and to the amino acid permeases from bacteria and yeast (e.g. 30% identity to Saccharomyces cerevisiae methionine permease MUP1 (36)).
As shown in Fig. 1, 12 transmembrane regions were predicted on the AGT1 amino acid sequence. There is a conserved cysteine residue (AGT1 amino acid residue 129) in the putative extracellular loop between predicted transmembrane domains 3 and 4, through which LAT1, LAT2, Asc-1, y ϩ LAT1, y ϩ LAT2, xCT, and BAT1/b 0,ϩ AT are proposed to link to 4F2hc or rBAT via a disulfide bond (20). Protein kinase C-dependent phosphorylation sites and a tyrosine phosphorylation site are predicted in the putative intracellular domains. A cAMP-dependent phosphorylation site is predicted in the putative intracellular loops that is conserved between AGT1 and Asc-2 (see legend for Fig. 1).
Functional Expression of AGT1-The expression of AGT1 did not induce functional activity in Xenopus oocytes (Fig. 2a) or COS-7 cells (Fig. 2b). Therefore, AGT1 was coexpressed with 4F2hc or rBAT, because AGT1 exhibited structural similarity to the members of the heterodimeric amino acid transporter family. The coexpression of AGT1 and 4F2hc, however, did not induce amino acid transport activity in Xenopus oocytes (Fig.  2a). This result was confirmed in COS-7 cells in which the coexpression of AGT1 with 4F2hc or rBAT did not induce amino acid transport activity (Fig. 2b). Then, following the functional characterization of Asc-2 (21), we generated fusion proteins in which the C terminus of AGT1 was connected with the N terminus of 4F2hc or rBAT. When expressed in Xenopus oocytes, AGT1-4F2hc and AGT1-rBAT fusion proteins exhibited [ 14 C]L-aspartate uptake (Fig. 2a).
To examine whether the fusion proteins were expressed in the oocyte plasma membrane, we performed confocal immunofluorescence microscopic analysis using specific antibodies raised against C-terminal parts of AGT1 and 4F2hc. As shown in Fig. 3 (a and b), these antibodies did not exhibit specific staining in the control oocytes injected with water instead of cRNAs. When AGT1 was solely expressed, AGT1 protein did not appear on the plasma membrane (Fig. 3d). In contrast, when the AGT1-4F2hc fusion protein was expressed in Xenopus oocytes, both anti-4F2hc antibody and anti-AGT1 antibody recognized the immunoreactivity on the plasma membrane (Fig. 3, e and f), indicating that the AGT1-4F2hc fusion protein is expressed in the plasma membrane. In the absorption experiments in which the tissue sections were treated with the primary antibodies in the presence of their antigen peptides, the immunostainings were not detected, confirming the specificity of the immunoreactions (data not shown).
Transport Properties-When expressed in Xenopus oocytes, AGT1-4F2hc and AGT1-rBAT fusion proteins mediated the Na ϩ -and Cl Ϫ -independent transport (Fig. 4). The uptake of [ 14 C]L-aspartate mediated by the fusion proteins was saturable and followed Michaelis-Menten kinetics (Fig. 5). The K m values for L-aspartate were calculated to be 25.5 Ϯ 5.9 M in AGT1-4F2hc fusion protein and 20.1 Ϯ 6.1 M in AGT1-rBAT fusion protein.
Substrate Selectivity-The substrate selectivity of the AGT1-4F2hc fusion protein and the AGT1-rBAT fusion protein was investigated by inhibition experiments in which the uptake of 20 M [ 14 C]L-aspartate was measured in the presence of 2 mM of nonlabeled amino acids. As shown in Fig. 6, the [ 14 C]L-aspartate uptake by the oocytes expressing AGT1-4F2hc fusion protein or AGT1-rBAT fusion protein was markedly inhibited by acidic amino acids such as L-aspartate and L-glutamate. L-Cysteine exhibited weaker but significant inhibitory effect on the [ 14 C]L-aspartate uptake. Other L-␣-amino acids including neutral amino acids and basic amino acids ␤-alanine and ␥-aminobutyric acid did not inhibit [ 14 C]L-aspartate uptake mediated by AGT1-4F2hc fusion protein (Fig. 6). D-Amino acids such as D-aspartate, D-glutamate, D-asparagine, D-glutamine, D-alanine, D-serine, and D-valine did not exhibit significant inhibitory effects on the [ 14 C]L-aspartate uptake. The [ 14 C]L-aspartate uptake was not affected by the system L-specific inhibitor 2-aminobicyclo-(2,2,1)-heptane-2-carboxylic acid or the system A-specific inhibitor ␣-(aminomethyl)isobutyric acid (Fig. 6).
Consistent with the results from the inhibition experiments, high uptake levels of 14 C-labeled L-aspartate and L-glutamate were observed for the AGT1-4F2hc fusion protein (Fig. 7). The uptake of neutral amino acids including L-cysteine and L-cystine and basic amino acids was at the trace level. [ 14 C]D-Aspartate and [ 14 C]D-glutamate were not transported by the AGT1-4F2hc fusion protein, suggesting that AGT1 is stereoselective for aspartate and glutamate (Fig. 7). The AGT1-rBAT fusion protein exhibited an amino acid uptake profile basically identical to that of AGT1-4F2hc fusion protein (data not shown). The K m and V max values of [ 14 C]L-glutamate uptake for AGT1-4F2hc fusion protein were 21.8 Ϯ 6.5 M and 0.63 Ϯ 0.10 (normalized to that of [ 14 C]L-aspartate determined on the same batch of oocytes).
The effects of acidic amino acid analogues on AGT1-mediated transport were also investigated by inhibition experiments in which the uptake of 20 M [ 14 C]L-aspartate was measured in the presence of 2 mM nonlabeled acidic amino acid analogues. As shown in Fig. 8, the [ 14 C]L-aspartate uptake by the oocytes expressing the AGT1-4F2hc fusion protein was markedly inhibited by threo-␤-hydroxyaspartate, L-serine-O-sulfate, L-cysteine sulfinate, and L-cysteate as well as L-aspartate and Lglutamate, whereas it was not inhibited by L-␣-aminoadipate, L-homocysteate, L-trans-pyrrolidine-2,4-dicarboxylate, and dihydrokainate.
Tissue Distribution of Expression-The expression of AGT1 mRNA was analyzed by Northern blotting of poly(A) ϩ RNAs from various mouse tissues. A strong 2.2-kb band was detected only in the kidney (Fig. 9).
Protein Characterization under Nonreducing and Reducing Conditions-Western blot analyses were performed on the membrane fraction prepared from mouse kidney in the presence or the absence of 2-mercaptoethanol (Fig. 10). The antibody raised against AGT1 recognized the band of 250 kDa in the absence of 2-mercaptoethanol (nonreducing condition). In the presence of 2-mercaptoethanol (reducing condition), the 250-kDa band detected in the nonreducing condition disappeared, and a 40-kDa band was detected, consistent with the predicted molecular mass of AGT1 monomer (51 kDa). The shown aligned with those of system asc transporter Asc-2 (mouse) (21), system L transporter LAT1 (rat) (7), system y ϩ L transporter y ϩ LAT1 (rat) (33), system x Ϫ C transporter xCT (mouse) (15), and system b 0,ϩ transporter BAT1 (rat) (17). Identical residues in at least two sequences are shaded. Predicted transmembrane regions of AGT1, numbered 1-12, are shown by bold lines above the sequences. The conserved cysteine residue (AGT1 amino acid residue 129) in the predicted extracellular loop, through which LAT1, y ϩ LAT1, xCT, and BAT1 are proposed to link to 4F2hc or rBAT is indicated by an asterisk. Protein kinase C-dependent phosphorylation sites are predicted on the AGT1 sequence at residues 5, 36, 282, 312, and 419, among which the ones at 5 and 312 are predicted intracellularly (labeled with ϩ). A potential cAMP-dependent phosphorylation site is located at residue 237, which is predicted to be intracellular (labeled with #). A tyrosine phosphorylation site is predicted at residue 17 (labeled with &). Although a potential N-glycosylation site is located at residue 259, it is predicted to be in the membrane-spanning region. The residue numbers indicated above the aligned sequences are in reference to those in the amino acid sequence of AGT1. bands disappeared in the presence of antigen peptides in the absorption experiment, confirming the specificity of immunoreactions (data not shown).
Immunolocalization of AGT1 in the Mouse Kidney-Immunohistochemical analysis on mouse kidney revealed the strong immunoreactivity for AGT1 in the proximal tubules in the outer medulla and the distal tubules in the cortex (Fig. 11a). AGT1 immunoreactivity appeared to be localized on the basolateral membrane of the proximal tubule S3 segments (Fig.  11c) and the distal convoluted tubules (Fig. 11d). In the absorption experiments in which the tissue sections were treated with the primary antibodies in the presence of antigen peptides, the immunostaining was not detected, confirming the specificity of the immunoreactions (Fig. 11b). DISCUSSION In the present study, we identified a novel transporter AGT1 that exhibits structural similarity to Asc-2, a member of the heterodimeric amino acid transporter family presumed to be associated with unknown heavy chains. By generating fusion proteins with 4F2hc or rBAT, we were able to express AGT1 on the Xenopus oocyte plasma membrane. AGT1 transported acidic amino acids at high affinity in a Na ϩ -independent manner.
In the family of heterodimeric amino acid transporters, the transporter proteins are linked via a disulfide bond to the single membrane spanning heavy chains 4F2hc or rBAT (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19). Among the members of this family, a cysteine residue is conserved in the extracellular loop between predicted transmembrane domains 3 and 4 (Fig. 1). Through this cysteine residue, the transporter proteins of this family are proposed to form a disulfide bond with heavy chains (20). The cysteine residue is also conserved in AGT1 as well as Asc-2 (Fig. 1). It is therefore predicted that AGT1 is linked to the other protein(s) by forming a disulfide bond through the conserved cysteine residue.
For the members of the heterodimeric amino acid transporter family, the association with the heavy chain type II membrane glycoproteins is required for the light chain trans-porter proteins to be sorted to the plasma membrane (8,18,37). In the previous investigation, we showed that the fusion proteins in which Asc-2 was connected with 4F2hc or rBAT were sorted to the plasma membrane and exhibited their functions, although Asc-2 was not functional when solely expressed or coexpressed with 4F2hc or rBAT (21). This indicates that 4F2hc and rBAT are capable of supporting the membrane sorting of the light chain subunit transporter proteins when their fusion proteins are constructed even though not between the right partners. AGT1 is also proposed to require additional associating proteins similar to 4F2hc or rBAT but not 4F2hc and rBAT themselves, because AGT1 was not functional when solely expressed or coexpressed with 4F2hc or rBAT (Fig. 2) and also because AGT1 is not colocalized with 4F2hc or rBAT in kidney in vivo; AGT1 is expressed in the basolateral membrane of proximal straight tubules and distal convoluted tubules (Fig.  11), whereas rBAT is present in the apical membrane of proximal tubules, and 4F2hc is most densely expressed in the basolateral membrane of proximal convoluted tubules (17,19,31,38). We thus generated fusion proteins in which the C terminus of AGT1 is connected with the N terminus of 4F2hc or rBAT to examine the functional properties of AGT1. The fusion proteins in fact appeared on the plasma membrane and exhibited the transport activity when expressed in Xenopus oocytes (Figs. 2 and 3, e and f).
The fusion proteins of heavy chain and light chain subunits of heterodimeric amino acid transporters were first generated by Pfeiffer et al. (19) for the characterization of system b 0,ϩ transporter b 0,ϩ AT/BAT1 in Xenopus oocytes. They showed that the fusion protein in which the C terminus of the light chain b 0,ϩ AT/BAT1 was connected with the N terminus of its associating heavy chain rBAT was functional and exhibited the identical properties to those obtained by the coexpression of b 0,ϩ AT/BAT1 and rBAT (19). In addition, the mutant fusion protein whose light chain portion was mutated was not functional, confirming that the detected transport activity was due to the fusion protein itself and not to the associated oocyte endogenous light chain (19). For the characterization of Asc-2, we generated fusion proteins in which the C terminus of Asc-2 is connected with the N terminus of 4F2hc or rBAT (21). We showed that two Asc-2 fusion proteins connected with rBAT or 4F2hc exhibited basically identical properties in their ion dependence, affinity, and substrate selectivity, suggesting that the rBAT or 4F2hc portion of the fusion proteins does not affect the transport properties of light chain transporter subunits (21). Consistent with this, LAT1 fusion proteins connected with 4F2hc or rBAT exhibited substrate selectivity and affinity basically identical to those obtained by the coexpression of LAT1 and its partner heavy chain 4F2hc (21). In the present study, the fusion proteins AGT1-4F2hc and AGT1-rBAT were also shown to exhibit basically identical transport properties. Therefore, it is suggested that to generate fusion proteins with 4F2hc or rBAT is a useful strategy to examine the functional properties at least for the heterodimeric amino acid transporters and their related proteins.
The AGT1-4F2hc and AGT1-rBAT fusion proteins exhibited Na ϩ and Cl Ϫ -independent transport of acidic amino acids (Fig.  4). Furthermore, the AGT1-4F2hc fusion protein showed high affinity to L-aspartate and L-glutamate. For mammalian acidic amino acid transport systems, four transport systems have been characterized so far: Na ϩ -dependent X Ϫ A,G and X Ϫ A and Na ϩ -independent x Ϫ G , and x Ϫ C (1). X Ϫ A,G transports both glutamate and aspartate, whereas X Ϫ A largely excludes glutamate FIG. 3. Detection of the AGT1 and AGT1-4F2hc fusion protein expressed in Xenopus oocytes. Confocal immunofluorescence microscopic analyses were performed on the Xenopus oocytes injected with water as a control (a and b), AGT1 cRNA (c and d), or the cRNA for AGT1-4F2hc fusion protein (e and f) using anti-4F2hc antibody (a, c, and e) or anti-AGT1 antibody (b, d, and f). When the fusion protein was expressed in Xenopus oocytes, it was detected in the oocyte plasma membrane with anti-4F2hc antibody (arrows in e) and with anti-AGT1 antibody (arrows in f). When AGT1 was solely expressed, it was not detected in the oocyte plasma membrane (arrowheads in d).

FIG. 4. Ion dependence of the transport mediated by AGT1-4F2hc fusion protein.
The [ 14 C]L-aspartate uptake (20 M) mediated by the AGT1-4F2hc fusion protein measured in the standard uptake solution (Control) was compared with that measured in the Na ϩ -free uptake solution (Na ϩ -free) and that measured in the Cl Ϫ -free uptake solution (Cl Ϫ -free).

FIG. 5. Concentration dependence of [ 14 C]L-aspartate uptake.
The [ 14 C]L-aspartate uptake by the oocytes expressing the AGT1-4F2hc fusion protein (a) or the AGT1-rBAT fusion protein (b) was measured at 1, 3, 10, 30, 100, and 300 M of L-aspartate in a Na ϩ -free uptake solution and plotted against the L-aspartate concentration. The L-aspartate uptake was saturable and fit to the Michaelis-Menten curve. The insets show Eadie-Hofstee plots of L-aspartate uptake that were used to determine the kinetic parameters. The K m and V max values were 25.5 Ϯ 5.9 M and 0.67 Ϯ 0.013 pmol/min/oocyte for the AGT1-4F2hc fusion protein; 20.1 Ϯ 6.1 M and 0.55 Ϯ 0.052 pmol/min/oocyte for the AGT1-rBAT fusion protein.
and longer analogues. x Ϫ G transports glutamate and its analogues, largely excluding aspartate and short analogues. x Ϫ C is similar to x Ϫ G except that it transports cystine as well as glutamate (1). The functional properties of AGT1 is, however, not able to be assigned to any of these classically characterized amino acid transport systems. The reason for this is probably in its restricted localization in the basolateral membrane of renal tubules (Figs. 9 and 11). In addition, Na ϩ -dependent transport systems for acidic amino acids were present not only in the apical membrane but also in the basolateral membrane of the tubular epithelial cells (39), which may mask the contribution of Na ϩ -independent transport system in the basolateral membrane. The transporters subserving systems X Ϫ A,G and x Ϫ C have been identified so far by molecular cloning approaches (15,40). Five isoforms of Na ϩ -dependent high affinity glutamate transporters, which belong to SLC1 family, and Na ϩ -independent cystine/glutamate transporter xCT, which belongs to SLC7 family, have been identified for systems X Ϫ A,G and x Ϫ C , respectively (15,40,41). xCT is a heterodimeric amino acid transporter that is associated with 4F2hc (15). xCT accepts L-glutamate, L-homocysteate, and L-cystine. L-Aspartate is not transported at high rate by xCT (15). It is proposed that the substrate-binding site of xCT possesses a negative charge recognition site in the side chain-binding site so that xCT recognizes amino acids as anions. Thus, the length of the side chain of substrate amino FIG. 6. Inhibition of the [ 14 C]L-aspartate uptake by various amino acids. The [ 14 C]L-aspartate uptake (20 M) mediated by AGT1-4F2hc fusion protein (a) or AGT1-rBAT fusion protein (b) was measured in the presence of 2 mM nonradiolabeled indicated amino acids. The uptake was measured in the Na ϩ -free uptake solution, and the values are expressed as percentages of the control L-aspartate uptake in the absence of inhibitors ((Ϫ)). The L-aspartate uptake was inhibited by L-aspartate, L-glutamate, and L-cysteine. The asterisks indicate statistical significance. **, p Ͻ 0.01, Student's unpaired t test.
acids, namely the distance between the ␣-carbon and the negative charge on the side chain, would be an important determinant for the amino acids to be accepted by the substrate binding site of xCT (4). It is therefore understandable that glutamate and homocysteate are well accepted by xCT, whereas aspartate, which has a shorter side chain, is not. The distance between two carboxyl groups of L-cystine is probably well suited to meet the requirement, so that cystine is proposed to be recognized as an anionic amino acid to be accepted by the binding site (4).
Although AGT1 is structurally related to xCT, AGT1 exhibits a remarkable difference in the selectivity for acidic amino acid substrates. In contrast to xCT, AGT1 well accepts acidic amino acids with shorter side chains such as L-aspartate, threo-␤-hydroxyaspartate, L-serine-O-sulfate, L-cysteine sulfinate, Lcysteate, and L-glutamate (Figs. 7 and 8). AGT1 does not accept acidic amino acids with longer side chains such as L-homocysteate and L-␣-aminoadipate (Fig. 8). L-Cystine is not transported by AGT1. Again, the length of the side chain of acidic amino acids is an important determinant for substrates of AGT1. It is predicted that the negative charge recognition site in the side FIG. 7. The uptake of radiolabeled amino acids mediated by AGT1-4F2hc fusion protein. The uptake rates of 20 M radiolabeled amino acids mediated by the AGT1-4F2hc fusion protein were measured in Na ϩ -free uptake solution. The high uptake rates were observed for L-aspartate and L-glutamate. Cyst, Lcystine; MeAIB, ␣-(aminomethyl)isobutyric acid; AIB, ␣-aminoisobutyric acid; BCH, 2-aminobicyclo-(2,2,1)-heptane-2carboxylic acid.
FIG. 8. The effects of acidic amino acid analogues on AGT1mediated transport. The [ 14 C]L-aspartate uptake (20 M) mediated by AGT1-4F2hc fusion protein was measured in the presence of 2 mM nonradiolabeled indicated acidic amino acid analogues. The uptake was measured in the Na ϩ -free uptake solution, and the values are expressed as percentages of the control L-aspartate uptake in the absence of inhibitors ((Ϫ)). The L-aspartate uptake was markedly inhibited by threo-␤-hydroxyaspartate (THA), L-serine-O-sulfate (SOS), L-cysteine sulfinate, and L-cysteate but was not inhibited by L-␣-aminoadipate, L-homocysteate, L-trans-pyrrolidine-2,4-dicarboxylate (PDC), and dihydrokainate (DHK). The asterisks indicate statistical significance. **, p Ͻ 0.01, Student's unpaired t test. chain-binding site of AGT1 is closer to the ␣-carbon binding site than that of xCT. It is interesting that Asc-2 structurally related to AGT1 also accepts amino acids with short side chains, although Asc-2 prefers neutral amino acids (21). It is proposed that, although the substrate-binding site of AGT1 possesses similar spatial configuration to that of Asc-2, AGT1 has acquired the additional mechanisms for negative charge recognition in the course of evolution.
Na ϩ -dependent high affinity glutamate transporters transport not only L-glutamate and L-aspartate but also D-aspartate (40), whereas AGT1 is quite stereoselective for both aspartate and glutamate (Figs. 6 and 7). Although five isoforms of Na ϩdependent high affinity glutamate transporters exhibit remarkable differences in the inhibitor selectivity, L-trans-pyrrolidine-2,4-dicarboxylate is accepted by all of the isoforms and in general is regarded as a selective inhibitor for Na ϩ -dependent high affinity glutamate transporters (40,41). Dihydrokainate and L-␣-aminoadipate are selective for the isoforms GLT-1/EAAT2 and EAAT4, respectively (40,(42)(43)(44). AGT1mediated L-aspartate uptake is not affected by these inhibitors for Na ϩ -dependent high affinity glutamate transporters. AGT1 is thus supposed to possess quite distinct mechanisms of substrate recognition compared with Na ϩ -dependent high affinity glutamate transporters.
AGT1 is only expressed in kidney. In the immunohistochemical analysis, AGT1 immunoreactivity was detected in the basolateral membrane of the proximal straight tubules, particularly S3 segments in the outer medulla, and of the distal tubules in the cortex (Fig. 11, a, c, and d). In the S2 and S3 segments of the proximal tubules, Na ϩ -dependent high affinity glutamate transporter EAAC1 is present in the apical membrane (45). EAAC1 plays a critical role in the reabsorption of acidic amino acids from the luminal fluid because EAAC1 knockout mice exhibit acidic amino aciduria (46). Although the level of expression is less than that in the proximal tubules, EAAC1 is also present in the apical membrane of the distal convoluted tubules in agreement with previous physiological studies showing significant glutamate reabsorption distal to the proximal tubules, including the distal convoluted tubules (45,47,48). Although the functional role of AGT1 is not clear at this moment, considering the distribution of AGT1 along nephron segments apparently corresponding to that of EAAC1, it is speculated that AGT1 might function as an exit path for the acidic amino acids at the basolateral membrane of tubular epithelial cells in the reabsorption of acidic amino acids from the luminal fluid. It is also possible that AGT1 might contribute to provide tubular epithelial cells with metabolically important acidic amino acids from basolateral side, although Na ϩ -dependent glutamate transport systems with higher concentrating capability have been reported to be present in the basolateral membrane (39).
In the Western blot that we performed on mouse kidney, a high molecular mass band detected in the nonreducing condition shifted to the lower molecular mass band, which seems to correspond to the AGT1 monomer in the reducing condition (Fig. 10). This observation is interesting because AGT1 is proposed to be linked to the other protein by a disulfide bond through the conserved cysteine residue, although it is still unclear at this stage whether the high molecular mass band in the nonreducing condition is because AGT1 forms a heteromeric complex or because AGT1 oligomerizes with other cysteine residues.
In summary, we identified and characterized a novel amino acid transporter AGT1 with structural similarity to the members of heterodimeric amino acid transporter family particularly Asc-2 that is proposed to be associated with unknown heavy chains (21). AGT1 exhibits distinct Na ϩ -independent transport activity with substrate selectivity for acidic amino acids. Similar to Asc-2, AGT1 appears to be associated with unknown protein(s) other than 4F2hc or rBAT. The finding of AGT1 has established a subgroup of the heterodimeric amino acid transporter family, which includes transporters such as Asc-2 and AGT1 associated not with 4F2hc or rBAT but with other unidentified proteins. FIG. 11. Localization of AGT1 protein in mouse kidney. Shown are the results from the immunohistochemical analysis of mouse kidney sections showing the localization of AGT1 protein. a, at low magnification view, AGT1 immunoreactivity was located in the proximal tubules in the outer medulla (OM) and the distal tubules in the cortex (C). b, in the absorption test in which tissue sections were treated with the primary antibodies in the presence of antigen peptides, the immunostaining detected in a disappeared, indicating the specificity of the immunoreactions. c and d, high magnification view of the tubules in the outer medulla (c) and cortex (d). AGT1 immunoreactivity was detected on the basolateral membrane of the proximal straight tubules (particularly the S3 segment) in the outer medulla (arrows in c) and of the distal convoluted tubules in the cortex (arrowheads in d). Scale bars, 300 m for a and b and 30 m for c and d.