Identification and Functional Characterization of a Novel Low Affinity Aromatic-preferring Amino Acid Transporter (arpAT)

We have identified in silico arpAT, a gene encoding a new member of the LSHAT family, and cloned it from kidney. Co-expression of arpAT with the heavy subunits rBAT or 4F2hc elicited a sodium-independent alanine transport activity in HeLa cells. l-Tyrosine, l-3,4-dihydroxyphenylalanine (l-DOPA), l-glutamine, l-serine, l-cystine, and l-arginine were also transported. Kinetic and cis-inhibition studies showed a Km = 1.59 ± 0.24 mm for l-alanine or IC50 in the millimolar range for most amino acids, except l-proline, glycine, anionic and d-amino acids, which were not inhibitory. l-DOPA and l-tyrosine were the most effective competitive inhibitors of l-alanine transport, with IC50 values of 272.2 ± 57.1 and 716.3 ± 112.4 μm, respectively. In the small intestine, arpAT mRNA was located at the enterocytes, in a decreasing gradient from the crypts to the tip of the villi. It was also expressed in neurons from different brain areas. Finally, we show that while the arpAT gene is conserved in rat, dog, and chicken, it has become silenced in humans and chimpanzee. Actually, it has been recently reported that it is one of the 33 recently inactivated genes in the human lineage. The evolutionary implications of the silencing process and the roles of arpAT in transport of l-DOPA in the brain and in aromatic amino acid absorption are discussed.

Heteromeric amino acid transporters (HATs) 1 are composed by disulfide-linked heavy (HSHAT) and light (LSHAT) subunits (1,2). The heavy subunit is a type II membrane glycoprotein with a large extracellular domain and is involved in trafficking of the heterodimer to the plasma membrane. The light subunit is a polytopic membrane protein that confers transport function and specificity (1,2). Two heavy subunits are known, rBAT and 4F2hc. The heterodimer b 0,ϩ AT-rBAT induces system b 0,ϩ activity (3). 4F2hc forms two different L transporters with LAT1 and LAT2, and two different y ϩ L transporters with y ϩ LAT1 and y ϩ LAT2. System asc and system x C Ϫ are composed of the heterodimers asc1-4F2hc and xCT-4F2hc, respectively (4 -9). There are two LSHAT subunits, asc2 and AGT1, which do not bind 4F2hc or rBAT (10,11). Almost all these amino acid transporters are obligatory antiporters with a 1:1 stoichiometry (12,13).
The LSHATs are associated with a wide variety of physiological processes. LAT1 expression is associated with cell proliferation and is regulated by the availability of some amino acids (14,15). Cystine transport through xCT in nonepithelial cell lines is a rate-limiting step in glutathione biosynthesis, contributing to redox control (16). Human asc1 (and probably also mouse asc2) may participate in osmotic adaptation at the distal nephron (10,17), and y ϩ LAT2 could release dibasic amino acids in non-epithelial tissues (18). Apical system b 0,ϩ mediates most of the apical (re)absorption of cystine and part of the (re)absorption of dibasic amino acids, acting in concert with the heterodimers y ϩ LAT1-4F2hc and LAT2-4F2hc (19), which mediate the net efflux of dibasic amino acids and cysteine across the basolateral membrane to the blood, respectively (20 -22). Neutral amino acid (re)absorption is performed by B 0 AT, a member of the Na ϩdependent neurotransmitter transporter family, and mutations in this transporter cause Hartnup disease (23,24). To maintain the vectorial transport of amino acids through the intestine and kidney epithelia a net basolateral efflux of amino acids is needed. This cannot be provided by LSHATs, as they are obligatory exchangers. A candidate for this efflux transporter is TAT1, which may contribute to the basolateral exit of aromatic amino acids (25). Several HATs have a role in brain physiology. In this sense, it has been proposed that the localization of xCT in meninges and some circumventricular organs is well suited for the maintenance of the redox state in the cerebrospinal fluid (26). asc1 is localized to neurons in the brain, where it can mediate the synaptic clearance of D-serine (27). Finally, LAT1 is found at the blood-brain barrier microvessels and in regions were neurogenesis occurs in mouse adult brain (28 -30), indicating a role in the transport of large essential neutral amino acids across the blood-brain barrier and in neuron proliferation.
Here we have functionally characterized arpAT, a new member of the LSHAT family. Its coexpression with both rBAT and 4F2hc elicits sodium-independent, low affinity transport activity for neutral amino acids and, with lower affinity, for dibasic amino acids. It shows preferential transport of the aromatic amino acids L-DOPA, L-tyrosine, and most likely L-tryptophan. Its expression in the brain and in the intestine suggests roles in dopamine homeostasis and aromatic amino acid absorption.
We also report that despite conservation of the arpAT gene in vertebrates, it has become silenced during the evolution of primates. A preliminary evolutionary analysis of the arpAT gene in vertebrates is presented. During completion of this article the final euchromatic sequence of the human genome was reported (31), containing a list of the 33 recently arising non-processed pseudogenes. Confirming our results, a partial arpAT sequence containing two inactivating mutations was found.

EXPERIMENTAL PROCEDURES
Identification and Cloning of Mouse arpAT-From an initial sequence comparison (TBLASTN (32)) of all the members of the LSHAT family with the mouse genome (33), we detected two novel exonic regions with significant similarity to two subregions of the b 0,ϩ AT protein. By maximizing the identity to the known members of the LSHAT family we partially predicted the coding structure of this putative novel gene, which was used to define primers for the complete identification and cloning of the corresponding cDNA (later named as arpAT) as described below.
First strand cDNA was synthesized from 3.2 g of total RNA purified from mouse kidney using the SuperScript II kit (Invitrogen). Forward primer cr12F6, 5Ј-ATGGAGCGAAGTGAGGAG-3Ј, and reverse primer, cr12R5 5Ј-CCCAGTCAATGATCTTTAGTG-3Ј, corresponding to positions 1-18 and 1452-1472 nucleotides of the predicted cDNA were synthesized and used in the first PCR round. This amplification was carried out in a PerkinElmer 9600 thermocycler, and conditions were hot start for 2 min at 94°C; 15 cycles of denaturing (94°C for 30 s), annealing (62°C for 30 s, reducing 1°C every 3 cycles), and extension (68°C for 2.2 min); 20 cycles of denaturing (94°C for 30 s), annealing (57°C for 30 s), and extension (68°C for 2.2 min); and a final extension of 7 min at 68°C. PCR-amplified DNA fragments with the expected length were subcloned into pGEM-T easy vector (Promega) and sequenced. The DNA sequences obtained were then compared with DNA and the protein sequence of the protein constructed in silico, which we later named arpAT. All amplified products were sequenced with D-Rhodamine Dye Terminator Cycle Sequencing Ready Reaction (PerkinElmer). The sequence reactions were analyzed with Abi Prism 377 DNA sequencer. The arpAT cDNA was subcloned in the mammalian expression vector pCDNA3.1 (Invitrogen) for use in functional expression experiments. The arpAT cDNA was also detected by PCR amplification in the previous conditions from total RNA purified from mouse white adipose tissue, brain, and small intestine (data not shown).
Uptake and Efflux Measurements in HeLa-transfected Cells-Transfections with mouse arpAT, human b 0,ϩ AT, human rBAT, and human 4F2hc cDNAs in HeLa cells were performed as described previously (34). For uptake measurements, cells were washed twice in 1 ml of the transport medium (137 mM N-methyl-D-gluconate, 2.8 mM CaCl 2 , 1. 5 Ci/ml)) was added and cells were incubated for 1 min (linear conditions, results not shown). After incubation, the uptake medium was removed and cells were washed three times in 1 ml of cold (4°C) transport medium. Nonspecific binding was assessed by addition and immediate removal of the uptake solution at 4°C. After washing, cells were lysed by addition of 250 l of 0.1 M NaOH, 0.1% SDS per well; 100 l was used for radioactivity counting and 20 l was used to measure the protein content (BCA Protein Assay kit (Pierce Biotechnology, Rockford, IL)). Radioactivity was measured in a ␤-scintillation counter (Beckman LS 6000TA; Beckman Instruments). For efflux measurements, HeLa cells were preloaded by incubation in transport medium containing 0.5 mM L-[ 3 H]alanine or L-[ 3 H]tyrosine or L-[ 3 H]DOPA for 5 min. The cells were then washed three times in ice-cold transport medium and incubated in efflux solution in the presence or absence of different amino acids (1 mM). Aliquots of 50 l of the efflux medium were removed at 0, 60, and 120 s and the radioactivity was measured by adding the 50 l to 3 ml of scintillation liquid. The radioactivity remaining in the cells was also measured after lysis of the cells, as above.
Real-time PCR Analysis-Total RNA from the tissues of 5 adult male CD1 mice was obtained with TRIzol reagent according to the manufacturer's instructions (Invitrogen). First strand cDNA was synthesized from 2 g of total RNA using the SuperScript II kit (Invitrogen), and subsequently diluted with nuclease-free water to 1.7 ng/l cDNA. Real-time PCR amplification mixtures (16 l) contained 10 ng of template cDNA, 8 l of 2ϫ SYBR Green I Master Mix buffer (Applied Biosystems), and 400 nM forward and reverse primers. Reactions were run on an ABI PRISM 7900 Sequence Detector (Applied Biosystems). The cycling conditions comprised a 10-min polymerase activation at 95°C and 40 cycles at 95°C for 15 s and 60°C for 1 min. Each assay included a standard curve of five serial dilution points of small intestine cDNA (ranging from 10 ng to 625 pg in duplicate), a no template control (in triplicate; one control for each oligonucleotide set), and 10 ng of each test cDNA. All PCR efficiencies were above 95%. The SDS 2.1 software was used for the quantifications. Mouse peptidylprolyl isomerase A was used to normalize expression of arpAT in different tissues. Peptidylprolyl isomerase A did not vary more than 5-fold between tissues, and arpAT C T values ranged from 23.5 in the small intestine to 32 in the kidney. The primers used for arpAT were the pair arpAT1F 5Ј-ACCTTTGCCTGTCTTCTG-TATTTG-3Ј and arpAT1R 5Ј-AACCTTGTAAGTGTGGGGGAGATTC-3Ј; and the pair arpAT2F 5Ј-GGGGCTGTGGTCCTTTGATG-3Ј and 5Ј-AGGTTCTGCTTTGGATTCTTGAG-3Ј. For peptidylprolyl isomerase A the primers were CYPF, 5Ј-CAAATGCTGGACCAAACA-CAAAC-3Ј, and CYPR, 5Ј-TGCCATCCAGCCATTCAGTC-3Ј.
Northern Blot Analysis-Total RNA (30 g) from mouse intestine was separated on 1.2% denaturing formaldehyde agarose gels, transferred to nylon membranes, and hybridized overnight with a 32 P-labeled 403-bp probe (nucleotides 1-403 of the arpAT coding region). After hybridization, the blot was washed twice at 65°C in 0.1ϫ SSC (15 mM NaCl, 1.5 mM sodium citrate) and 0.1% SDS (each for 30 min). In addition, one mouse poly(A ϩ ) membrane from Clontech (Palo Alto, CA) was also hybridized with the same probe according to the manufacturer's instructions and washed identically.
Western Blot Analysis of HeLa Cell Proteins-The Western blot analyses of HeLa cells transfected with the cDNAs of mouse arpAT, human b 0,ϩ AT, human xCT, human rBAT, and human 4F2hc cDNAs were performed as described previously (35).
In Situ Hybridization-Riboprobes corresponding to the nucleotide sequence coding for two fragments of the arpAT protein were prepared from in vitro transcription and used for in situ hybridization. In preparation for the riboprobe, 196-bp (5Ј-probe, nucleotides 69 to 264 of the arpAT coding region) and 224-bp (3Ј-probe, nucleotides 1088 to 1311 of the arpAT coding region) fragments were amplified using the following two sets of oligonucleotides: cr12F10 (5Ј-GGGGCTGATGCTGAAGAG-3Ј) and cr12R8 (5Ј-GCCCAGTTCAGCATAGCAC-3Ј) for the 5Ј-probe; and cr12F11 (5Ј-CGGCTGTGGCTTTACTGCTG-3Ј) and cr12R10 (5Ј-CGGGTGGTCAATGATGGGTG-3Ј) for the 3Ј-probe. The amplified fragments were purified, ligated to a pGEM-T Easy vector (Promega), and sequenced to determine sequence orientation. The cRNA were labeled with digoxigenin-11-UTP (Roche Molecular Biochemicals) by transcription in sense and antisense orientations using the mMessage mMachine SP6 and T7 in vitro transcription system (Ambion). The quality and the quantity of the riboprobes were evaluated using denaturing polyacrylamide gel electrophoresis and spectrophotometric reading, respectively.
Mouse tissues were embedded in paraffin after whole animal fixation with paraformaldehyde, 0.1 M phosphate buffer. 5-m sections were cut in a Leica RM 2135 microtome, mounted on silanized slides (PerkinElmer), and deparafined. In situ hybridization was performed as described elsewhere (6). Slides were examined on an Olympus microscope and the images were captured in a Leica IM50 Image manager at the Serveis Científico Tècnics (University of Barcelona).
Comparative Analysis of the Evolution of arpAT Genes in Vertebrates-The alignment for the phylogenetic tree has been performed with the T-Coffee program (36).
To evaluate the patterns of evolution associated with the arpAT genes of different vertebrates we have used available tools for sequence analysis and sequence evolution. To detect arpAT orthologs, we have used the mouse arpAT protein sequence as query in BLAST comparisons against the genome of rat (assembly of June 2003), dog (assembly of June 2004), chicken (assembly of February 2004), human (build35), and chimp (assembly of November 2003). For each of the identified regions, we predicted the structure and the resulting peptide sequence using GENEWISE (37), which in addition, reports the position of all possible coding disablements found in-frame (stop codons or frameshits). These regions were considered true arpAT orthologs because: (a) flanking genes were also similar between compared genomes, and (b) the genomic regions identified appeared as first and best match when using mouse arpAT as query (tBLASTN), FIG. 1. The mouse arpAT ORF. The sequence of mouse arpAT has been deposited in GenBank TM with accession number AY795939. A, neighbor-joining tree of the mouse LSHAT family (bootstrap values are shown for each node). The tree was rooted using the YHFM transporter from Escherichia coli (NP_417829). B, the alignment of mouse arpAT and mouse b 0,ϩ AT proteins is shown. Conserved (*) and similar (:) amino acid residues are noted. The two predicted N-glycosylation sites and the conserved cysteine involved in the intermolecular disulfide bridge are shaded in gray. Straight lines span the 12 putative transmembrane domains, absolutely conserved in the LSHAT family. Also shown are 21 and 845 nucleotides of the putative 5Ј and 3Ј untranslated regions, respectively. The stop codon and the first canonical polyadenylation signal are shaded. and the comparison using tBLASTx of the predicted peptides back to the mouse genome found again the arpAT gene as the best matching region (best reciprocal match).
The levels of selective constraints associated with a certain genomic region can be inferred by the ratio of silent (synonymous, d S ) to amino acid replacement (non-synonymous, d N ) nucleotide substitutions. d N /d S ratios of non-functional regions and those of the vast majority of genes are generally different, as mutations in genes causing amino acid replacements with functional consequences are selected against (i.e. d N values are lower), in contrast to mutations occurring in dispensable regions, such as silenced genes or pseudogenes (with d N /d S close to 1). We have calculated d N /d S in different steps. We first aligned the predicted peptide sequences of all identified arpAT orthologs using CLUSTALW (38) followed by manual refinement. We next removed all regions containing gaps in at least one of the peptide sequences. Based on aligned positions we replaced each of the amino acids by the corresponding coding codon, obtaining a protein-based DNA alignment, which was then use to calculate (through maximum likelihood using codeml: unrooted tree and model 1; (39)) the specific rates of synonymous (d S ) and non-synonymous (d N ) nucleotide substitutions for each of the sequences.

RESULTS
In Silico Cloning of arpAT-Our goal was to identify new members of the LSHAT family. BLAST searches with the known human and mouse LSHATs protein sequences identified a region spanning 10.2 Mb on mouse chromosome 12 (12-A1.3 band, region 8.674000 -8.688000 of the June 2003 mouse genome assembly), which contained a putative new open reading frame fulfilling the criteria for a bona fide LSHAT: 1) the ORF exhibited high similarity to mammalian LSHATs along the whole sequence, without significant gaps (e.g. 43% identity, 64% similarity with mouse b 0,ϩ AT; and see Fig. 1A, where a phylogenetic tree of the mouse LSHAT family is shown); 2) its predicted topology revealed 12 transmembrane domains with intracellular N and C termini like the recently proven topology for the LSHAT xCT (40); 3) the cysteine involved in the conserved disulfide bond in the putative second extracellular loop (41) is also present. Fig. 1B shows the alignment of mouse b 0,ϩ AT and the new LSHAT member, which we named arpAT because of its functional properties (see below). The predicted open reading frame codes for a 53.1-kDa protein of 488 amino acids with two potential N-glycosylation sites. One lies in the putative 7th transmembrane domain and it is therefore unlikely to be used. The second site, located in the 3rd extracellular loop, is also present in human LAT1 where it was shown not to be used in vivo (42). A Kozak consensus sequence (5Ј-g-aaATGg-3Ј) and a putative canonical polyadenylation signal (AATAA) 837 bp downstream of the stop codon were also present (Fig. 1B). Two RIKEN cDNA clones map onto the mouse arpAT region (see Fig. A in the Supplemental Materials).
Comparative Analysis of the Evolution of arpAT in Vertebrates-We identified the arpAT orthologs in rat, dog, and chicken with complete ORFs (see Fig. B in the Supplemental Materials). The proteins showed 92 (rat), 80 (dog), and 59% (chicken) identity with the mouse arpAT. We also found the human and chimpanzee arpAT orthologous regions (chr2: 20507813-20517977 and chr12:21334250 -21344417, respectively; see Fig. A in the Supplementary Materials). Their predicted ORFs were 75 and 76% identical to mouse for human and chimpanzee, respectively, and revealed 10 frame-disrupting insertions/deletions and 4 in-frame stop codons, which should prevent the expression of functional arpAT protein.
Moreover, an Alu sequence element disrupted exon 1. Nearly all of these coding disablements (including the Alu insertion) share the same position in both organisms, indicating that silencing of the arpAT gene took place in a common ancestor during primate evolution. We did not detect close paralogs of arpAT making it unlikely that its function in primates was rescued by newly arisen gene copies. By the completion of this manuscript, the final euchromatic sequence of the human genome was published (31). Confirming our results, two inactivating mutations were found in a partial arpAT human sequence, but their positions were not reported (31). Strikingly, from a total of 20,000 -25,000 human protein genes, only 33 recently arising non-processed pseudogenes were identified. arpAT was the only transporter among them (31).
We assessed the levels of associated selective constraint, reflected by the d N /d S ratios, of the human and chimpanzee arpAT regions. As most functional genes in mammals (33,43), there was an excess of synonymous versus non-synonymous nucleotide substitutions in the mouse (d N /d S ϭ 0.3), rat (d N /d S ϭ 0.37), dog (d N /d S ϭ 0.12), and chicken (d N /d S ϭ 0.05) arpAT gene regions. In contrast, no selective constraint was found associated with the arpAT regions of human or chimpanzee (d N /d S ϭ 0.8 and d N /d S ϭ 0.85, respectively), indicating that they are neutrally evolving pseudogenes.
The rate of neutral substitutions (d S /Mya) of arpAT in the primate lineage is significantly lower than that in the other mammals: d S /Mya ϭ 0.013 in human and chimpanzee, versus d S /Mya ϭ ϳ0.022 in rodents and dog (taking 90 Mya as an approximation for the divergence time of these three lineages (44)). This disagrees with the general trend, in which pseudogenes normally present rates of substitutions that are similar or higher than functional genes (43,45).
Functional Expression of arpAT-A cDNA identical to the ORF constructed in silico (see Fig. 1) was obtained by reverse transcriptase-PCR from mouse kidney. The ORF was then subcloned in an expression vector for functional analysis in mammalian cells. The transfection of arpAT alone or in combination with either 4F2hc or rBAT in HeLa cells resulted in the induction of L-alanine transport (see also that binds 4F2hc), human b 0,ϩ AT (which binds only rBAT (3)), and mouse arpAT (Fig. 2). rBAT alone and rBAT co-transfected with xCT were barely detected as a 86-kDa band in both reducing and non-reducing conditions. This band corresponds to the immature, core-glycosylated form of rBAT ( cg rBAT), as shown previously (35,46). When rBAT was co-transfected with b 0,ϩ AT, the mature Endo-H resistant form of rBAT appeared ( m rBAT, 90 kDa), and rBAT expression increased (35,46). The same occurred when rBAT was co-transfected with arpAT. In addition, only when rBAT was transfected with b 0,ϩ AT or ar-pAT were two bands of 125 and 240 kDa detected under nonreducing conditions, which represent the heterodimer and probably a dimer of the heterodimer with the light chain (3,35,46).
Characterization of the arpAT-induced Amino Acid Transport Activity-We examined the amino acid transport activity in HeLa cells co-transfected with rBAT and arpAT. The only difference from the transport elicited by 4F2hc and arpAT is the magnitude of the transport activity (see above and data not shown). arpAT induced transport of many amino acids, including small and large neutral amino acids and dibasic amino acids (Fig. 3). Because of the high rate of L-DOPA and Ltyrosine uptake in HeLa cells, arpAT-mediated uptake was only detected at higher concentrations of these amino acids (Fig. 3). L-Alanine transport was pH-independent (5.5-8 pH range) and sodium did not modify the uptake of different arpAT substrates (data not shown). We then measured the inhibition of 50 M L-alanine transport by a 100-fold excess concentration of a wide range of amino acids, amino acid analogues, polyamines, and dopamine (Fig. 4). Transport of L-alanine was not inhibited by anionic amino acids, D-amino acids, proline, glycine, polyamines, or MeAIB, and was slightly affected by dopamine. The other amino acids and the analogues BCH and AIB partially inhibited L-alanine transport. The most potent inhibitors were the aromatic amino acids (L-DOPA Ͼ L-Tyr Ͼ L-Trp), and L-Phe and L-OHTrp showed behavior similar to L-Ala and L-Leu (60 to 80% inhibition). In general, ␤-branched amino acids were less inhibitory than their unbranched counterparts (compare L-Ala and L-Ser with L-Thr; and L-Leu with L-Val and L-Ile).
Consistent with the relatively low inhibition by a 100-fold excess substrate, kinetic analysis of the L-alanine induced uptake revealed a single low affinity component with a K m of 1.59 Ϯ 0.24 mM and a V max of 15.4 Ϯ 0.84 nmol/mg of protein/ min (Fig. 5A). The high endogenous transport rates for L-DOPA, L-Tyr, and L-Trp hampered kinetic analysis of the ar-pAT-mediated transport (data not shown). Therefore, we measured the inhibition of the induced L-alanine uptake (50 M) by L-Tyr, L-DOPA, and L-Ser (Fig. 5B). For competitive inhibitors, IC 50 can be used as a good estimate of the K I provided that it is assayed at substrate concentrations well below the K m (47). The IC 50 were 272.2 Ϯ 57.1 M, 716.3 Ϯ 112.4 M, and 2.21 Ϯ 0.19 mM, for L-DOPA, L-Tyr, and L-Ser, respectively. An estimate of the K I for each inhibitory amino acid was calculated from the data in Fig. 4: thus, the K I would be ϳ0.8 mM for L-Trp, ϳ1.6 mM for L-Leu, ϳ3.7 mM for L-Val, ϳ8.5 mM for L-Thr, and ϳ7.5 mM for L-Arg. In summary, arpAT is a low affinity amino acid transporter for neutral and dibasic amino acids, with a strong preference for aromatic amino acids, especially L-DOPA.
In agreement with the activities of other LSHATs (2), arpAT showed trans-stimulation (Fig. 6). The efflux of L-[H 3 ]alanine in HeLa cells expressing arpAT and rBAT was dependent on the presence of a substrate in the medium (e.g. L-alanine or L-tyrosine), but it was not trans-stimulated by amino acids that are not substrates (e.g. L-proline). The efflux in trans 0 conditions (no amino acid substrates in the medium) was similar in HeLa cells expressing arpAT-rBAT and in non-transfected HeLa cells (Fig. 6), suggesting that arpAT is a tightly coupled exchanger. We also observed transstimulation of L-[H 3 ]tyrosine and L-[H 3 ]DOPA efflux in the presence of L-alanine, but not in the absence of amino acids or with L-proline, in HeLa cells expressing arpAT-rBAT (data not shown).
Tissue Localization of arpAT mRNA-Attempts to obtain antibodies against arpAT did not succeed, and therefore we analyzed arpAT mRNA expression. Northern blot in high stringency conditions revealed a 3.4-kb band only in the small intestine and, after a much longer exposure, in the brain (data not shown). Quantitative real-time PCR showed strong arpAT expression in the small intestine. The other tissues expressed very low amounts of arpAT (from 0.2% in kidney to 6.6% in testis, compared with small intestine) (Fig. 7). By in situ hybridization, no specific signal could be detected in the liver, heart, or kidney (data not shown). In contrast, the arpAT mRNA was found both in the small intestine (jejunum) and brain (Fig. 8, A and B). In the jejunum, the arpAT mRNA was detected in the enterocytes in a decreasing gradient along the crypt to villus axis. The signal was concentrated in the basal cytoplasm (as shown for other mRNAs, such as villin, lactase, and brush-border myosin I (48, 49)) and in some crypt cells. In the brain, arpAT mRNA was found in neurons of pyriform cortex and amygdala, hippocampus (especially in the CA1 and For each experimental sample, the amount of the targets (arpAT) and endogenous reference (peptidylprolyl isomerase A) was determined from the corresponding standard curves. Then, each individual target amount was divided by its own endogenous reference amount to obtain a normalized target value. The highest arpAT normalized value (small intestine) was set to 100, and the amounts of the other tissues were corrected accordingly. The results are the mean Ϯ S.E. of five points in duplicate, each one corresponding to the tissue extracted from a different individual. The amounts are statistically different from the no cDNA negative controls, which were set to zero (as no amplification signal was detected at the last PCR cycle) (one sample t test, p Ͻ 0.001; and for the kidney, p Ͻ 0.05). For arpAT determination two different oligonucleotide pairs were used, which gave similar results. Only the results for the first set of oligonucleotides (arpAT1F and arpAT1R, see "Experimental Procedures") are shown.
CA3 regions), and in several hypothalamic (periventricular, supraoptic, and arquate) and thalamic (paraventricular and anterodorsal) nuclei. The choroid plexus was also distinctly labeled. In contrast, no signal was detected in the striatum, globus pallidus, and subfornical organ. DISCUSSION A similarity search through vertebrate genomes identified the arpAT gene, a new member of the LSHAT family. arpAT is a broad specificity amino acid exchanger with preference for aromatic amino acids. Its mRNA is expressed mainly in the small intestine and in some brain areas. In HeLa cells arpAT forms a heterodimer with rBAT and probably also with 4F2hc. In addition, we report that the arpAT gene has become silenced in human and chimpanzee genomes, and present an initial evolutionary analysis of the arpAT gene in vertebrates.
The absence of other novel LSHAT genes in the vertebrate genomes indicates that arpAT completes the LSHAT family with no less than 10 different members. With the exception of asc2 and AGT1, the LSHATs bind either rBAT or 4F2hc. It is likely that arpAT binds also rBAT or 4F2hc in vivo: (i) arpAT is as distant from asc2 and AGT1 as the other 7 LSHATs; and (ii) it covalently binds rBAT and most likely 4F2hc in HeLa cells (see Fig. 2). In epithelial cells the only known apical heterodimer is rBAT-b 0,ϩ AT (3,50); the 4F2hc-LSHAT heterodimers are basolateral (50). In vivo, the LSHATs seem to heterodimerize only with one of the heavy subunits, although they can be less selective when overexpressed in heterologous systems (51), which may apply to arpAT in HeLa cells. arpAT mRNA localization in the small intestine and brain is compatible with both rBAT and 4F2hc binding, as both heavy chains have been localized to these tissues (26,50,(52)(53)(54). Another possibility is that arpAT may heterodimerize with both rBAT and 4F2hc in vivo, depending on its localization.
arpAT mRNA is restricted to the small intestine and the brain. Real-time PCR analysis of brain cDNA most likely shows the combination of a low and confined expression of the transcript to specific areas, as seen by in situ hybridization. The arpAT amounts measured with real-time PCR in other tissues may reflect illegitimate transcription (55). arpAT transport specificity and tissue distribution gives few clues about its physiology. In the brain, sodium-independent low and high affinity components of L-Tyr, L-DOPA, and L-Trp transport have been identified (56,57). The low affinity component is consistent with arpAT expression. arpAT may participate in the homeostasis of dopamine, catecolamines, and 5-hydroxytryptamine. In addition, L-DOPA may act as a neurotransmitter in the rat, in a dopamine-independent manner (58). In this scenario, arpAT could be a L-DOPA transporter related to the neurotransmitter function of this amino acid in the rodent brain.
Most of the transporters involved in nutrient absorption increase their expression from the crypt to the villi (50,59,60). In contrast, arpAT and the aspartate/glutamate transporter EAAC1 (which mediates glutamate absorption), show an opposite expression gradient (61). Therefore, arpAT may participate in the absorption of aromatic amino acids at the lower part of the villi, probably by exchange with other neutral amino acids. Depending on its localization, arpAT could complement the function of the basolateral aromatic amino acid transporter TAT1 or the apical neutral amino acid transporter B 0 AT. ar-pAT may also modulate the autocrine/paracrine non-neuronal dopaminergic system (which regulates intestinal sodium absorption (62)), providing L-DOPA for dopamine production in the jejunum (63). The expression of arpAT in the small intestine and brain might link arpAT carrier activity to a putative amino acid sensor system that gets information of the levels of some essential amino acids (L-Phe ϩ L-Tyr, and L-Trp) in these two organs (64). The relatively low affinity of arpAT would ensure a linear response on a wide range of concentrations. Whatever its physiological role, arpAT silencing in primates suggests its functional rescue by a different protein or that FIG. 8. In situ hybridization of arpAT mRNA in adult mouse tissues. Serial paraffin-embedded sections of mouse small intestine (jejunum) (A), and brain (B) were incubated with antisense or sense probes (see "Experimental Procedures"). Results are representative of two independent experiments, and were performed in parallel with a different set of probes from the 5Ј and 3Ј end of the cDNA, with similar results. Only the images obtained with the 3Ј probes are shown. In A, arpAT-specific detection (arrows) is restricted to the lower part of the villi (1) and some cells of the crypts (2). In B, arpAT-specific detection (arrows) is found in cortex (1) and amygdala (2)  some biological or environmental changes rendered its function non-essential. These changes may be associated with different feeding behaviors of primates compared with other vertebrates.
The evolution of the arpAT gene shows an atypical pattern. It is conserved in rodents, dog, and chicken but the accumulation of lethal mutations prevents its functional expression in the primate lineage. arpAT belongs to the group of recently arising non-processed pseudogenes (dead genes) reported in the last refinement of the human genome (31). The group of dead genes (where arpAT is the only transporter) accounts only for 33 of 20,000 -25,000 of the human protein genes, a number that highlights its importance for lineage-specific evolution (see below) (31). The levels of selective constraint for each of the arpAT genes analyzed agree with the integrity of their genes: rodent, dog, and chicken arpAT suffer purifying selection, whereas human and chimpanzee arpAT evolve neutrally, like most of the mammalian pseudogenes (43). A surprising observation is the excess of coding disablements identified in the primate arpAT genes (most of them are frameshifts) compared with the relatively low mutational activity that these genes underwent from the rodent-primate split (d S values are approximately half of those in the other mammals). In addition to the possibility that variable insertion-deletion and substitution rates have affected these regions in primates in various periods of their evolution, the silencing of this gene may have undergone positive selection instead of simple neutral evolution because of the loss of selective advantage. Loss of function of arpAT likely reflects recent changes in the brain and intestinal metabolism of arpAT substrates in primates. Its carrier activity characterized here will help to elucidate the physiological role of arpAT in mice, which will be relevant to understand its possible role in the lineage-specific evolution of primates.