Physiol Rev AJP: Cell Physiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Physiol. Rev. 78: 969-1054, 1998;
0031-9333/98 $15.00
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by PALACÍN, M.
Right arrow Articles by ZORZANO, A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by PALACÍN, M.
Right arrow Articles by ZORZANO, A.

PHYSIOLOGICAL REVIEWS   Vol. 78 No. 4 October 1998, pp. 969-1054
Copyright ©1998 by the American Physiological Society

Molecular Biology of Mammalian Plasma Membrane Amino Acid Transporters

MANUEL PALACÍN, RAÚL ESTÉVEZ, JOAN BERTRAN, AND ANTONIO ZORZANO

Departament de Bioquímica i Biologia Molecular, Facultat de Biologia, Universitat de Barcelona, and Departament de Criobiologia i Terapia Cellular, Institut de Recerca Oncològica, Barcelona, Spain

I. INTRODUCTION
    A. Amino Acid Transport Systems in the Plasma Membrane of Mammalian Cells
    B. Strategies Used to Identify Mammalian Amino Acid Transporters as Yet Uncloned
II. CURRENT KNOWLEDGE OF THE MOLECULAR STRUCTURE OF AMINO ACID TRANSPORT SYSTEMS
    A. Cationic Amino Acid Transporters
    B. Superfamily of Sodium- and Chloride-Dependent Neurotransmitter Transporters
    C. Superfamily of Sodium-Dependent Transporters for Anionic and Zwitterionic Amino Acids
    D. Putative Subunits of Sodium-Independent Cationic and Zwitterionic Amino Acid Transporters
III. INHERITED DISEASES OF PLASMA MEMBRANE AMINO ACID TRANSPORT
IV. PROSPECTS
REFERENCES

    ABSTRACT
Top
References

Palacín, Manuel, Raúl Estévez, Joan Bertran, and Antonio Zorzano. Molecular Biology of Mammalian Plasma Membrane Amino Acid Transporters. Physiol. Rev. 78: 969-1054, 1998. --- Molecular biology entered the field of mammalian amino acid transporters in 1990-1991 with the cloning of the first GABA and cationic amino acid transporters. Since then, cDNA have been isolated for more than 20 mammalian amino acid transporters. All of them belong to four protein families. Here we describe the tissue expression, transport characteristics, structure-function relationship, and the putative physiological roles of these transporters. Wherever possible, the ascription of these transporters to known amino acid transport systems is suggested. Significant contributions have been made to the molecular biology of amino acid transport in mammals in the last 3 years, such as the construction of knockouts for the CAT-1 cationic amino acid transporter and the EAAT2 and EAAT3 glutamate transporters, as well as a growing number of studies aimed to elucidate the structure-function relationship of the amino acid transporter. In addition, the first gene (rBAT) responsible for an inherited disease of amino acid transport (cystinuria) has been identified. Identifying the molecular structure of amino acid transport systems of high physiological relevance (e.g., system A, L, N, and x-c) and of the genes responsible for other aminoacidurias as well as revealing the key molecular mechanisms of the amino acid transporters are the main challenges of the future in this field.

    I. INTRODUCTION
Top
Next
References

Amino acid transport across the plasma membrane mediates and regulates the flow of these ionic nutrients into cells and, therefore, participates in interorgan amino acid nutrition. In addition, for specific amino acids that act as neurotransmitters, synaptic modulators, or neurotransmitter precursors, transport across the plasma membrane ensures reuptake from the synaptic cleft, maintenance of a tonic level of their extracellular concentration, and supply of precursors in the central nervous system (for review, see Refs. 93, 96, 97, 505). Transfer of amino acids across the hydrophobic domain of the plasma membrane is mediated by proteins that recognize, bind, and transport these amino acids from the extracellular medium into the cell, or vice versa. In the early 1960s, after the pioneer work of Halvor N. Christensen's group, different substrate specificity transport systems for amino acids in mammalian cells (mainly erythrocytes, hepatocytes, and fibroblasts) were identified (reviewed in Ref. 96), and general properties of mammalian amino acid transport were revealed: stereospecificity (transport is faster for L-stereoisomers) and broad substrate specificity (several amino acids share a transport system). Since these initial studies, the main functional criteria used to define amino acid transporters have been 1) the type of amino acid (acidic, zwitterionic, or basic as well as other characteristics of the side chain) the protein moves across the membrane, according to substrate specificity and kinetic properties, and 2) the thermodynamic properties of the transport (whether the transporter is equilibrative or drives the organic substrate uphill). This functional classification has been retained to date, since structural information on higher eukaryote amino acid transporters is incomplete.

Isolation of the first brain GABA transporter (184) in 1990 and the identification of the first cationic amino acid transporter in 1991 (281, 590) represent the starting points for the cloning of mammalian amino acid transporter genes. Several strategies have been used to identify mammalian amino acid transporters. During the 1980s, several groups attempted the purification of these transporters by different methods. Purification strategies have yielded few structural data, although for a couple of transporters related to neurotransmission (the GABA transporter GAT-1 and the glutamate transporter GLT-1), these data allowed microsequencing or generation of specific antibodies that have been used to isolate cDNA clones (184, 424). Alternative strategies and serendipity allowed the identification of up to 22 cDNA (including splice variants, but not species counterparts) for mammalian amino acid transporters or related proteins (see sect. II). This structural information is not complete. The genes identified seem to correspond to eight classic transport systems and their variants, whereas another eight of the major amino acid transport systems are unknown at the molecular level (see Table 1). An excellent review by Kilberg's group (342) describes the molecular biology of the amino acid transporters cloned up to 1995. 

 
View this table:
[in this window] [in a new window]
  		TABLE 1.   Best-known amino acid transport systems present in the plasma membrane of mammalian cells

The molecular identification of amino acid transporters or related proteins leads to ongoing studies on the structure-function relationship and the molecular genetics of the pathology associated with these transporters. In this review, attention is paid to the molecular biology, structure-function relationship, physiological role, and human genetics of amino acid transporters. Regulation of plasma membrane amino acid transport in mammals is beyond the scope of the present review and has been extensively reviewed (96, 278, 350).

A. Amino Acid Transport Systems in the Plasma Membrane of Mammalian Cells

Functional studies based on saturability of transport, substrate specificity, kinetic behavior, mode of energization, and mechanisms of regulation performed in perfused organs, isolated cells, and purified plasma membranes led to the identification of a mutiplicity of transport agencies in the plasma membrane of mammalian cells (for review, see Refs. 26, 94-96). The properties of some of the best-characterized amino acid transport systems are summarized in Table 1. From these studies it is evident that a particular transport system carries different amino acids and that amino acid transport systems show overlapping specificities. Different cells contain a distinct set of transport systems in their plasma membranes, as a combination of common or almost ubiquitous (e.g., systems A, ASC, L, y+ and X -AG ) and tissue-specific transport systems (e.g., systems Bo,+, Nm, and bo,+ as well as variants of common transport systems). It has been proposed that this arrangement permits both fine regulation of substrate and cell-specific amino acid flows and economy in the number of structures mediating cellular and interorgan amino acid fluxes (96, 97).

1. Common systems for zwitterionic amino acids

The zwitterionic amino acid transport systems A, ASC, and L are present in almost all cell types. Systems A and ASC mediate symport of amino acid with small side chain with sodium, and system L mediates uniport of amino acids with bulky side chains (i.e., branched and aromatic groups). System L has a high exchange property (i.e., it shows trans-stimulation, stimulation by substrates in the trans-compartment) and has been postulated to serve in many circumstances to mediate efflux from cells rather than influx into cells (96). Preferred substrates for system A are alanine, serine, and glutamine, and for system ASC alanine, serine, and cysteine (99, 101). In contrast, amino acids with a small, nonbranched side chain are poor substrates for system L, for which the analogs 2 - a m i n o e n d o b i c y c l o - [ 2, 2, 1 ] - h e p t a n e - 2 - c a r b o x y l i c   a c i d (BCH) and 3-aminoendobicyclo-[3,2,1]-octane-3-carboxylic acid (BCO) are model substrates in the absence of sodium. System L has been subdivided into subtypes L1 (substrate affinity in the micromolar range) and L2 (substrate affinity in the millimolar range), which are present at different ages in hepatocytes and hepatoma cell lines (597). Therefore, zwitterionic amino acids with bulky side chains are not cotransported with sodium in many cell types. This is different in epithelial cells, where broad specificity systems (e.g., systems Bo, Bo,+) accumulate these amino acids by coupling the electrochemical gradient of sodium. One criterion for discrimination of system A is the fact that it transports N-methylated amino acids (101); in many instances, sodium-dependent transport of alanine inhibitable by the model analog N-methylaminoisobutyric acid (MeAIB) or the sodium-dependent transport of MeAIB is used for determining system A transport activity. System A is highly pH sensitive. Intracellular histidyl residues may form part of the pH sensor of the transporter (43, 384). In contrast, system ASC is relatively pH insensitive. Additional differential characteristics of systems A and ASC are that the former is electrogenic and shows the property of trans-inhibition (i.e., inhibition by substrates in the trans-compartment), and the latter may be electroneutral and shows trans-stimulation, probably as a consequence of a high property of exchange. Therefore, whereas system A is considered a true sodium-amino acid cotransporter, system ASC may be an antiporter of amino acids necessarily associated with the movement of sodium in both directions (for review, see Ref. 185). Transporters ASCT1-2 from the superfamily of sodium- and potassium-dependent transporters of anionic and zwitterionic amino acids may represent transporter variants of system ASC (see sect. II). The molecular identity of systems A and L is unknown.

An important distinguishing feature of system A is that in many cell types its activity is highly regulated. This includes upregulation during cell-cycle progression and cell growth in many cell types, and hormonal control (insulin, glucagon, catecholamines, glucocorticoids, and growth factors and mitogens) through a wide variety of mechanisms. Thus glucagon and epidermal growth factor short-term stimulate system A in hepatocytes, whereas insulin upregulates system A in a gene transcription-dependent manner in hepatocytes and through a rapid mechanism in skeletal muscle. This is independent of protein synthesis and the electrochemical gradient of sodium, suggesting direct stimulation of preexisting system A transporters or recruitment of these to the plasma membrane. Paradoxically, situations associated with insulin deficiency or insulin resistance (e.g., diabetes, starvation, pregnancy) are associated with upregulation of system A activity in both hepatocytes and skeletal muscle. Two additional long-term maximum velocity (Vmax) regulation processes of system A have been demonstrated only in culture or incubated cells: adaptative regulation (i.e., repression and derepression of system A in the presence or absence of amino acids) and upregulation by hyperosmolarity. In both types of regulation, there is indirect evidence for the presence of system A regulatory proteins. For a detailed description of the putative mechanisms involved in the regulation of system A, the reader is directed to excellent recent reviews (185, 201, 350, 414) and to specific articles on insulin action in skeletal muscle (188-190, 561). The complexity of the regulation of system A emphasizes the importance of this transport activity in maintaining the accumulation of small side-chain zwitterionic amino acids as a source for precursor molecules, energy, and perhaps even as osmolytes. Unfortunately, the lack of structural information on the system A transporter (see sect. IB1) prevents further understanding of the molecular mechanisms underlying system A regulation.

2. Tissue-specific systems for zwitterionic amino acids

Additional sodium-dependent zwitterionic transport systems of restricted distribution have been described (Table 1). Transport of L-glutamine, L-histidine, and L-asparagine in hepatocytes has been demonstrated to occur via a sodium-dependent transport system named N (276). A system with similar properties to system N was defined in skeletal muscle and termed Nm (222). System Gly, specific for glycine and sarcosine, occurs in several cell types (134). Transporters GLYT1-2, from the superfamily of sodium- and chloride-dependent transporters of neurotransmitters, may represent variants of system Gly (see sect. II). A transport system for beta -alanine, taurine, and GABA (system BETA) has been characterized in several tissues and with differences in substrate affinity and specificity (369). Transporters GAT1-3, BGT-1, and TAUT, which belong to the superfamily of sodium- and chloride-dependent transporters of neurotransmitters, may be considered variants of system BETA (see sect. II).

Several zwitterionic transport systems seem to be specific to the apical pole of epithelial cells. In intestinal brush-border membrane vesicles, a sodium-dependent transport system (NBB, which stands for neutral brush border) serving for neutral amino acids is present. This system was renamed B for consistency with other broad-specificity systems (e.g., Bo,+, bo,+, b+; see below) (337). More recently, in bovine renal brush-border membranes, a sodium-dependent system for neutral amino acids was described that was termed Bo (329). Most probably the transport systems B (NBB) and Bo represent the same transport agency distributed in epithelial cells, as suggested after the cloning of the putative transporter for system Bo (ATBo, from the superfamily of sodium- and potassium-dependent transporters for anionic and zwitterionic amino acids; see sect. II). This broad-specificity system is thought to be responsible (together with the dipeptide and tripeptide transport systems; for review, see Refs. 122, 299) for the bulk of renal reabsorption and intestinal absorption of zwitterionic amino acids. Therefore, system Bo and the ATBo cDNA could represent the transport activity and the corresponding transcript that are altered in Hartnup disease (304), an inherited hyperaminoaciduria of neutral amino acids (see sect. III). Additional transport systems (see Table 1) seem to be specific to brush-border membranes. System IMINO, which catalyzes sodium-dependent transport of proline and N-methylated glycines and is inhibited by MeAIB, has been detected in kidney and intestine and in other epithelia (192, 376, 514). The proline transporter PROT, from the superfamily of sodium- and chloride-dependent transporters of neurotransmitters, might represent a brain-specific high-affinity [Michaelis constant (Km) in the low micromolar range] variant of system IMINO; this is not yet clear (see sect. II). The broad-specificity (cationic and zwitterionic amino acids) transport systems Bo,+ and bo,+ are described in section IA3, together with cationic amino acid transport systems.

3. Cationic amino acid transport systems

Five transport systems that mediate the uptake of cationic amino acids are known (Table 1). One corresponds to the widespread classical system y+, whereas the other four were discovered in the late 1980s and early 1990s (systems y+L, b+, bo,+, and Bo,+) and at present have been described only in specific tissues. The activity of all these systems can be distinguished by their affinity for the cationic amino acids, by their dependence on sodium, and by their capacity to share transport with zwitterionic amino acids (for a short review, see Ref. 105). System y+ catalyzes high-affinity (Km in the micromolar range) sodium-independent transport of cationic amino acids and the transport of zwitterionic amino acids with low affinity (Km in the large millimolar range; affinity increases with chain length) only in the presence of sodium; this system is electrogenic and accumulates its substrates by coupling with the cell plasma membrane potential (98, 455, 601). System y+L was discovered in human erythrocytes (127) and has also been described in placenta (135, 146). It is possible that system y+L is widely distributed among different tissues; thus it has also been detected in human fibroblasts (D. Torrents and M. Palacín, unpublished data). This system transports cationic amino acids with high affinity (Km in the micromolar range) with no need for sodium in the external medium, but it transports both small and large zwitterionic amino acids with high affinity (Km in the micromolar range) in the presence of external sodium; in the absence of sodium, transport of zwitterionic amino acids is of very low affinity. In addition to this, system y+ and y+L activities could be discriminated, at least in erythrocytes and placenta, by N-ethylmaleimide (NEM) treatment, the former being sensitive and the latter resistant. Systems y+ and y+L show very high capacity for trans-stimulation (i.e., exchange). It is thought that exchange via system y+ allows equilibration of cationic amino acids across the plasma membrane, whereas heteroexchange between cationic and zwitterionic amino acids plus sodium via system y+L catalyzes the efflux of cationic amino acids against the membrane potential, the driving force being provided by the sodium ion concentration gradient (14, 90). Systems y+ and y+L have been suggested to be candidate transport activities affected in the inherited disease lysinuric protein intolerance (LPI) (for review, see Ref. 497). This is discussed in section III. The cDNA for up to five potential y+ transporters have been isolated (CAT-1, CAT-2, the splice variant CAT-2a, CAT-3, and possibly the recently identified CAT-4), which form the cationic amino acid transporter family (see sect. II). Several groups described expression of system y+L transport activity in oocytes by the 4F2hc surface antigen, which shows homology with another protein, rBAT, also related to broad-specificity amino acid transport. Because 4F2hc and rBAT are less hydrophobic than typical transporter proteins and form disulfide-bound heterodimers with unidentified proteins, it has been suggested that 4F2hc may represent a putative subunit of system y+L transporter; this ascription is not yet clear (see sect. II).

Systems Bo,+, bo,+, and b+ were discovered in mouse blastocysts (576-578; for review, see Ref. 573). Among the systems that form the series of transport activities for cationic amino acids, the embryonic sodium-independent systems b+ (subtypes b1+ and b2+, which differ in the embryonic stage expression and sensitivity to cationic amino acid inhibition) show the narrowest specificity, serving only for cationic amino acids (576). Systems Bo,+ and bo,+ show very similar broad specificity with high affinity (Km in the micromolar range) for cationic and small and large zwitterionic amino acids. As a distinguishing feature, the former is sodium dependent and inhibitable by BCH and BCO, and the latter prefers bulky alpha ,beta -unbranched zwitterionic amino acids. More probably, both systems have a wide distribution on epithelial cells. In fact, system bo,+ (or a variant, bo,+-like) has been detected in renal epithelial cells and in Caco-2 cells (374, 557). Expression of rBAT, a protein homologous to the cell surface antigen 4F2hc, is needed for system bo,+-like transport activity; it is believed that rBAT acts as a subunit of this transporter (see sect. II). Mutations in rBAT/system bo,+-like transport activity cause cystinuria type I (for reviews, see Refs. 170, 408, 409), an inherited hyperaminoaciduria due to defective renal reabsorption and intestinal absorption of cationic amino acids and cystine (for review, see Ref. 487). The role of rBAT in cystinuria is discussed in section III.

4. Anionic amino acid transport systems

L-Glutamate and L-aspartate are accumulated in many cells (e.g., neurons and glial cells, hepatocytes, enterocytes, fibroblasts, and placental trophoblasts) by the high-affinity (Km in the micromolar range) sodium- and potassium-dependent system X -AG (165) (for review, see Ref. 185). This system shows identical affinity for the D- and L-stereoisomers of aspartate (165). Variants of this transport systems occur in neural tissues (100, 147). It is believed that the five glutamate transporters EAAT1-5 from the superfamily of sodium- and potassium-dependent transporters of anionic and zwitterionic amino acids represent variants of system X -AG (see sect. II).

Several cell types (e.g., hepatocytes, fibroblasts, and embryonic cells) transport L-glutamate (specifically anionic amino acids with 3 or more carbon atoms in the side chain) and L-cystine (as tripolar ion) via the sodium-independent antiport system x -C (28, 533). This system has an apparent Km in the 100-200 µM range, is insensitive to membrane potential, and presents trans-stimulation (for review, see Ref. 185). Bannai's group (27) proposed that this system participates in a glutamine-cystine cycle that helps cells to resist oxidative stress; glutamine, entering the cell via systems ASC and A, is converted to glutamate, which is exchanged for cystine via the oxidative stress-induced system x -C; accumulated cystine then nourishes glutathione synthesis, which protects cells against oxidative insult (27, 29, 596; for review, see Ref. 228).

B. Strategies Used to Identify Mammalian Amino Acid Transporters as Yet Uncloned

As described in section IA, the present explosion of cloned cDNA related to plasma amino acid transport in mammals is revealing an intriguing range of structural diversity within amino acid transporter proteins. This diversity is already even more pronounced than that shown by the sodium-dependent glucose transport (for review, see Ref. 608) and the facilitated glucose transporter isoforms (for review, see Refs. 381, 556). The lack of high-affinity inhibitors for mammalian amino acid carriers and their low abundance in plasma membranes complicate their structural identification and isolation. Because of this, there are many amino acid transport systems not yet identified at a molecular level (see Table 1). Among these systems, there are the highly regulated system A and the well-characterized systems L, N, and x -C. In this section, we focus on the strategies used to identify these four transporters and also speculate, in some cases, about why some strategies have failed.

1. System A

One of the goals of several laboratories has been to reconstitute and purify system A transporter. Kilberg's group (195) reported the solubilization, reconstitution, and partial purification of system A (70-fold over plasma membrane vesicles). They then used this protein fraction to immunize mice for the generation of monoclonal antibodies. Some of these antibodies specifically coprecipitate fodrin and system A transport activity. Because the protein ankyrin often binds directly to integral membrane proteins and fodrin, the authors tested whether an antibody against ankyrin could immunoprecipitate system A transport activity, and it did. McGivan's group (437) partially purified system A activity from rat liver with concanavalin A-affinity chromatography. This demonstrated that either system A or a protein bound to the carrier is a glycoprotein. McCormick and Johnstone (348) purified system A transport activity from Ehrlich ascites with a 30-fold enrichment. Three major peaks were eluted from the system A-purified Ehrlich cell preparation: high-molecular-mass aggregates, a low-molecular-mass band (~40 kDa), and the most conspicuous band of 120-130 kDa. Interestingly, polyclonal antibodies against the 120- to 130-kDa purified fraction immunoprecipitated system A transport activity. More recently, NH2-terminal sequence analysis of the 120- to 130-kDa peptide revealed a sequence similar to that of the alpha 3-subunit of the alpha 3beta 1-integrin (349). Further purification of these extracts using lectin columns resulted in the separation of most of the alpha 3beta 1-integrin from the system A activity, indicating that this integrin is not essential for amino acid transport. Moreover, the fact that transfection of alpha 3-integrin into K562 or RD cells increased system A transport activity provides evidence that this protein could modulate this transporter. In summary, from these studies of reconstitution, one may conclude that several distinct proteins contribute to the entire system A transport activity. Probably in the near future the microsequencing of some of the proteins present in these purified extracts will result in the isolation of all the components necessary for system A function.

Another strategy used to identify system A is the functional expression in Xenopus laevis oocytes. Expression of system A has been claimed (409, 546) based on the following (for review, see Ref. 277). 1) The effect of glucagon on system A in vivo was maintained after mRNA extraction and injection into the oocyte; glucagon is known to stimulate system A transport activity in liver, at least in part, through mRNA and protein synthesis-dependent mechanisms. 2) The apparent substrate affinities reported by these authors were in the same range as those described for the transport of the substrates tested via system A. 3) The cis-inhibition of the transport of L-alanine induced by MeAIB suggested that at least part of this expressed activity was due to system A. 4) The expressed transport activity, in contrast to the endogenous uptake, was inhibited by an extracellular pH of 6.5. 5) Messenger RNA from the Chinese hamster ovary (CHO) cell line alar-H3.9, which overexpresses system A activity (371, 372), resulted in higher transport rates than mRNA from the parental cell line CHO-K1. More recently, Lin et al. (313) presented evidence that mRNA of differing sizes (2.2 and 4.2 kb) from two cell lines (GF-14 cells, Ehrlich cells) increase the expression of system A transport upon injection into Xenopus oocytes. However, only the synthesis of the 2.2-kb transcript is raised by insulin, which is consistent with the idea that there are variants of system A transporter (see below). Expression cloning of this transport activity has not yet been achieved, possibly due to the high background (i.e., basal oocyte endogenous activity), and perhaps because the system A transporter would need the expression of different proteins that form part of the whole transporter or are upregulators of its activity (e.g., as shown by the adaptative or osmotic regulation of system A; see above, for review see Ref. 350). In the same line of functional expression of system A transport activity, Lin et al. (312) restored normal growth of a mutated yeast cell line incapable of growth in minimal medium with proline by transfection with a cDNA (E51) from mouse Ehrlich cells. This cDNA is 90% homologous to gamma -actin. Similarly, this cDNA increases sodium-dependent amino acid uptake when expressed in oocytes and in a mutated mammalian lymphocyte cell line (GF-17), deficient in system A transport activity. This suggests that the gamma -actin-like protein coded for by E51 cDNA may play a significant regulatory role in sodium-dependent amino acid transport. In summary, reconstitution-purification and functional expression studies suggest that a multiplicity of proteins might be involved in the functional expression and modulation of system A transport activity.

Another approach is the development of cell lines that show mutations in amino acid transport activity (reviewed in Refs. 113, 603). In this way, Englesberg's group (371) has isolated constitutive or derepressed mutants for system A activity from CHO-K1 cells by alanine-resistant selection for proline uptake (alar4 mutant) and by a stepwise selection (hydroxyurea treatment and resistance to increased alanine concentration) (alar4-H3.9 mutant). In comparison to the wild type, these mutants showed higher system A activity in isolated plasma membrane vesicles and higher mRNA-induced sodium-dependent aminoisobutyric acid uptake in Xenopus oocytes (371, 546). These mutants have increased levels of peptides banding at 62-66 kDa and 29 kDa. Sequencing the NH2 terminus of the 62- to 66-kDa peptide shows between 80 and 100% identity with the mammalian mitochondrial 60-kDa heat-shock protein (HSP60) (235). Whether these proteins are components of system A carrier is at present unknown.

Other approaches involve the chemical modification of specific residues by covalent reagents. Hayes and McGivan (202) identified a 20-kDa protein as a putative component of sodium-dependent alanine transport in liver plasma membrane vesicles. On the other hand, the presence of histidine residues critical for activity in the system A from rat liver (i.e., sensitive to diethyl pyrocarbonate inactivation) has also been demonstrated using this approach (43). Thiol reagents have been used to reveal structural differences between these carriers and between normal and transformed cells. It has been suggested that structural differences occur in system A transporters of transformed cells, based on the much greater sensitivity to NEM inactivation of liposome-reconstituted system A activity from normal hepatocytes than that from hepatoma cell lines (132). Moreover, all these studies should be critically evaluated, since these specific reagents can modify, with different affinities, a variety of different amino acid residues in proteins. In any case, no further structural information on system A has been achieved with this strategy.

2. System L

Using the strategy of functional expression in Xenopus oocytes, Oxender's group (521) reported the expression of a sodium-independent L-leucine transport system shared with dibasic amino acids, by injection of mRNA from CHO cells. This suggests that the expressed transport could correspond to that induced by the expression of rBAT (i.e., system bo,+) and not to system L. In any case, this ascription is not yet clear, since the data from Oxender's group (521) and rBAT-expressed amino acid transport activity (549) differ in the sensitivity to inhibition by L-tryptophan. More recently, Broër et al. (68) also expressed sodium-independent isoleucine transport activity from mRNA of rat brain in oocytes. The sodium-independent component of isoleucine transport was inhibited by leucine, phenylalanine, and BCH, consistent with the expression of a system L-like transporter. However, the isolated cDNA responsible for this activity was rat 4F2hc, which also expresses cationic amino acid transport in oocytes (69). As discussed in section II, several groups proposed that 4F2hc expresses a system y+L-like transport activity in oocytes. In our view, expression of system L in oocytes has not been conclusively demonstrated.

Another interesting strategy to assess the structure of this transporter has been developed by Oxender's group (136). It is based on the fact that the transport activity of system L can be derepressed by severe "starvation" for leucine or by increasing the temperature of culture in mutant cell lines with temperature-sensitive leucyl-tRNA synthetase (reviewed in Ref. 603). Oxender's group (136) transformed a temperature-sensitive leucyl-tRNA synthetase mutant CHO cell line (CHO-025C1) with human DNA from a cosmid library. Subsequent selection of transformants for inability to grow above the permissive temperature in the presence of low leucine concentration allowed the isolation of cells with higher (<2-fold) leucine uptake activity. To date, no report has described the rescue of the human DNA sequences responsible for the above-mentioned transformation.

More recently, Segel's group (606) has developed a new strategy to identify the carrier protein(s) responsible for mammalian L-system amino acid transport. In chronic lymphocytic leukemia (CLL), B lymphocytes have markedly disminished L-system transport, which is restored after treatment with 12-O-tetradecanoylphorbol 13-acetate (TPA). These authors identified six candidate L-system-related proteins in TPA-treated CLL cells using an L-system photoprobe (iodoazidophenylalanine) and ultra-high-resolution two-dimensional gel electrophoresis. Two of these six proteins were microsequenced and show sequence similarity to the mitochondrial heat-shock protein (HSP60). This report and the data published by Englesberg's group (235) implicate the family of heat shock proteins in the regulation of some transport processes. How these proteins develop their function in systems A and L transport activity is unknown.

3. System N

Kilberg's group (143), with the same approach of aggregation and differential solubility used to purify system A, achieved a 600-fold enrichment for system N amino acid transport activity in reconstituted proteoliposomes (540). They identifed a 100-kDa protein involved in system N amino acid transport activity by generating monoclonal antibodies against the purified fraction that immunoprecipitate system N transport activity (539). These tools may allow the identification of system N transporter.

Rennie's group (551) has reported the expression of rat liver glutamine transporters after injection of rat liver mRNA into Xenopus oocytes. They attributed part of the L-glutamine-induced transport activity to system N based on a characteristic feature of this transport system, that is, the toleration of lithium by sodium substitution and the inhibition by L-histidine in lithium medium. By size-fractioning the mRNA, they found three different induced transport activities: one sodium independent induced by 2.8-3.6 kb mRNA, another sodium-dependent, lithium sustitution intolerant induced by 1.9-2.8 kb mRNA, and one induced by a lighter fraction (<1.9 kb) that is sodium or lithium dependent and that could correspond to system N.

4. System x -c

Bannai's group (227), working with manually defolliculated oocytes, has reported the expression of a mouse-macrophage cystine transporter with characteristics of system x -c, as it was sodium independent and glutamate inhibitable. Fractions of mRNA of 1.5-2.9 kb are responsible for this induction. Although oocytes seem to express an endogenous system x -c (574), expression of this system correlates with injection of mRNA from x -c-rich cells (macrophages stimulated by diethylmaleate in culture), but not from x -c-poor cells (noncultured macrophages and mouse leukemia L1210 cells). In addition, cystine uptake expressed by diethylmaleate-stimulated macrophage mRNA was, in contrast to the endogenous cystine uptake, pH sensitive, highly temperature sensitive, and inhibitable by glutamate. This line of research may lead to the identification of system x -c transporter.

    II. CURRENT KNOWLEDGE OF THE MOLECULAR STRUCTURE OF AMINO ACID TRANSPORT SYSTEMS
Top
Previous
Next
References

In this section, the cDNA clones that have been related to plasma membrane amino acid transport in mammals are discussed from the structural and functional points of view. With regard to the primary structures elucidated to date, it can be summarized that two main types of membrane proteins are involved in amino acid transport: those that present multiple (i.e., 10-14) transmembrane domains and are therefore considered as putative transporters and those that do not fit this general model and are considered as activators or as components of oligomeric transporters. The first group with typical structure of transporters is arranged in families of related cDNA: 1) the family of sodium-independent cationic amino acid transporters (CAT); 2) superfamily of sodium- and chloride-dependent neurotransmitter transporters, which includes amino acid transporters; and 3) the superfamily of sodium-dependent, and in some cases potassium-dependent, anionic or zwitterionic amino acid transporters. The second group corresponds to the family of proteins rBAT and 4F2hc that induce sodium- and chloride-independent amino acid transport with broad specificity (i.e., for dibasic and zwitterionic amino acids) in X. laevis oocytes. The members of this family are less hydrophobic than typical transporters and show heteroligomeric structure.

Another protein related to amino acid transport, whose primary structure was elucidated many years ago, is the anion exchanger band 3 (290). Although its function as an anion exchanger is well accepted, it has been shown to be associated with the transport of some amino acids (e.g., glycine, taurine, and beta -alanine), especially under conditions of hyposmotic stress; in response to swelling, erythrocytes recover their initial volume by releasing organic osmolytes via a pathway with a pharmacology similar to that of band 3 (114, 203, 308). This amino acid transport has the properties of a volume-sensitive size-limited anion channel (171). Interestingly, expression of trout band 3 in oocytes resulted in anion-exchange activity but also in chloride channel activity and taurine transport (150). At present, it is not known whether band 3 is involved in the amino acid channel formation or in its regulation. The role of band 3 in amino acid transport is outside the scope of this review. This and the molecular biology of band 3 have recently been reviewed (10, 380, 574).

A. Cationic Amino Acid Transporters

Four homologous human and rodent genes defining a family of cationic amino acid transporters (CAT-1, -2, -3, and -4) have been, or are in the process of being, identified (Table 2). First, expression in oocytes revealed the ecotropic murine leukemia virus receptor (9) (now named CAT-1) as a putative cationic amino acid transporter (281, 590). Second, full-length cDNA cloned from a previously identified murine T-lymphoma cell line cDNA (Tea, for "T early activation" gene; Ref. 334) showed significant homology with CAT-1 and cationic amino acid transport expression in oocytes (106, 108, 242, 448). These cDNA, now named CAT-2 and CAT-2a, represent splice variants (the "high-affinity" or "T-cell" variant CAT-2 and the "low-affinity" or "liver" variant CAT-2a) (see Table 2) of a single gene (336). It has been suggested that the three putative proteins (CAT-1, CAT-2, and CAT-2a variants) may contain 14 transmembrane domains (TM) (9, 106) (Figs. 1 and 2, see below). Very recently, with strategies based on sequence homology with CAT-1, the mouse and rat counterparts (95% amino acid sequence identity between them) of a new brain-specific cationic amino acid transporter, CAT-3, have been isolated (219, 229). Very recently, Sebastio and co-workers (484, 485) identified a human EST sequence (Table 2) with significant homology to the 5'-end of the open reading frame of CAT-1 and CAT-2/2a cDNA (Fig. 1). Screening of the full-length human CAT-4 cDNA [we renamed HCAT-3 (485) as CAT-4 after the reported cloning of rat and mouse CAT-3 (219, 229)] has been completed, and the putative protein shows 34-38% identity with human CAT-1, -2, and -3 (G. Sebastio, personal communication). At present, there are no reported data on the transport activity associated with CAT-4 expression. Screening, performed by us (DBEST, December 1996), for additional expressed sequence tag sequences homologous to CAT transporters indicative of additional members of the family was negative.

 
View this table:
[in this window] [in a new window]
  		TABLE 2.   Human cationic amino acid transporters


View larger version (91K):
[in this window]
[in a new window]
  		FIG. 1.   Amino acid sequence comparison of human cationic amino acid transporters (CAT). HO5853 sequence corresponds to best EST sequence in database DBEST that showed homology with human CAT-1 and CAT-2, and rat CAT-3, and that served for human CAT-4 cloning (G. Sebastio; personal communication). Amino acid residues present in all CAT transporter sequences known to date are indicated by gray boxes. Straight lines over sequences indicate putative transmembrane (TM) domains (I-XIV). Between TM V and TM VI, two potential N-glycosylation sites are conserved (open boxes). Dash lines indicate gaps for sequence alignment, which corresponds to that reported by Closs (105).


View larger version (69K):
[in this window]
[in a new window]
  		FIG. 2.   Fourteen TM transmembrane topology model for murine mCAT-2 transporter. Extracellular (EL) and intracellular (IL) loops are numbered. Amino acid residues conserved in all known CAT transporters are indicated in gray circles, and residues that are specific for known CAT-2 and CAT-2a transporters are indicated in white over a black circle. At bottom, 41- to 42-amino acid segment corresponding to splice variant exon of mCAT-2 and mCAT-2a transporters are compared with homologous mCAT-1 and mCAT-3 regions (residues present in at least 3 of sequences are contained in gray boxes, and those specific for low-affinity CAT-2a isoform are in white on black boxes). In EL3 loop, amino acid sequence that binds viral envelope glycoprotein gp70, present in murine and rat CAT-1, is boxed. Three residues, a Glu in TM III and 2 glycosylated (Y) Asn residues in EL3, are marked by squares, indicating homologous residues that have been analyzed by site-directed mutagenesis studies of mCAT-1 sequence. Human CAT-1 gene structure has been published (619), and it has been reported that mouse CAT-1 gene contains 8 coding exons (419). To our knowledge, gene structure of CAT-2 has not been reported.

Figure 1 compares the amino acid sequences of the putative human CAT-1 and CAT-2a proteins, the mouse CAT-3 protein, and the putative NH2-terminal fragment of the human CAT-4 transporter. Two potential N-glycosylation sites, conserved in all known sequences, are located in the extracellular loop EL3 in the 14-TM model (Figs. 1 and 2). The known amino acid sequences of CAT-2 and CAT-2a (98% identity between the murine counterparts; see Fig. 2) are ~60% identical to that of CAT-1 (106, 448). The rat and mouse CAT-3 show 53-58% amino acid sequence identity with the isolated CAT-1, CAT-2, and CAT-2a cDNA (219). Two regions of extensive amino acid sequence identity (7E80%) of ~150 and 200 amino acid residues, comprising the first three transmembrane domains and transmembrane domains VI-X, are present in the CAT-1, -2, and -3 proteins (334; see Figs. 1 and 2). Murine CAT-2 and CAT-2a differ only in a 41- to 42-amino acid segment located in this highly conserved region (intracellular loop IL4 between TM domains VIII and IX of the 14-TM model) (Fig. 2). Because a single genomic fragment contains both exons, the isoforms result from mutually exclusive alternate splicing of the primary trancript (unpublished data from MacLeod and co-workers quoted in Ref. 336). This amino acid sequence region has a role in substrate binding, as demonstrated by the expression of CAT-2/CAT-2a chimeric transporters (backbone and the 42-amino acid domain) (108).

The CAT transporters are homologous to a family of transporters specific for amino acids, polyamines, and choline (APC family) that catalyze solute uniport, solute/cation symport, or solute/solute antiport in yeast, fungi, and eubacteria (448). Marked sequence divergence of these proteins was observed mainly in the hydrophilic NH2 terminals that precede the first transmembrane helices and in the COOH-terminal regions (448). Southern blot studies revealed that all vertebrates examined hybridize to the probes of CAT-1 and CAT-2, indicating a high conservation of these proteins among vertebrates (448). Thus the human CAT-1 protein (7, 619) and the rat CAT-1 protein (433, 610) are 86 and 95% identical, respectively, to the mouse CAT-1, and the human analogs of CAT-2 and CAT-2a proteins are ~90% identical to the murine counterparts (218; unpublished data quoted in Ref. 105).

The chromosomal location of the human CAT genes is showed in Table 2. The possible involvement of CAT genes in LPI, a human inherited hyperaminoaciduria that seems to result from an impairment of a system y+-like activity (497), is discussed in section III.

Two excellent recent reviews described functional and structural data as well as available data on the regulation of the expression of CAT-1 and CAT-2/2a transporters (105, 336).

1. Tissue expression

The tissue distribution of CAT genes has been examined by Northern blot analysis (for mCAT-1; Refs. 9, 242, 281; for hCAT-4, G. Sebastio, personal communication) and by RT-PCR (specific detection of mCAT-2 and mCAT-2a; Refs. 151, 336). Tissue distribution and transcript size of these genes are indicated in Table 3. Semi-quantitative data for the murine genes were summarized by MacLeod and Kakuda (336); all tissues or cell types examined express at least one of the mCAT genes, and in some tissues, both genes (CAT-1 and CAT-2) are expressed. Liver is the only tissue that expresses only mCAT-2a and not mCAT-1 or mCAT-2. Kidney, small intestine, resident macrophages and quiescent splenocytes, and T cells only express the mCAT-1 gene but neither of the two mCAT-2 splice variants. Upon activation, these cell types express the mCAT-2 variant. The rat and mouse CAT-3 gene is expressed specifically in brain, as a transcript of ~3.4 kb (219, 229). The human CAT-4 gene is expressed mainly in pancreas, skeletal muscle, heart, and placenta, and brain, lung, liver, and kidney show a faint band in Northern analysis (Sebastio, personal communication).

 
View this table:
[in this window] [in a new window]
  		TABLE 3.   Tissue distribution and transport characteristics of expressed CAT transporters

The tissue and subcellular distribution of CAT transporter isoforms is less known than their transcript distributions. Expression of mCAT-1 protein has been studied by exploiting its function as a viral receptor (infectivity or viral glycoprotein gp70 binding assay; Refs. 108, 281, 590, 610) and by using anti-mCAT-1 antisera (106, 108, 605). Recently, approaches to the detection of mCAT-1 based on its function as a viral receptor have been validated in null knockout CAT-1 mice (419); primary embryo fibroblasts from these mice were completely resistant to ecotropic retrovirus infection (i.e., mCAT-1 is the sole receptor for ecotropic murine leukemia virus). The lack of constitutive expression of CAT-1 in human, murine, and rat liver has been demonstrated by virus infectability studies (610) and immunofluorescence studies (605). These strategies served to demonstrate the induction of surface expression of CAT-1 in the murine liver after partial hepatectomy and after insulin and dexamethasone treatment (610).

Antibodies directed against the COOH terminus of CAT-2a, which do not distinguish between CAT-2 and CAT-2a, detected a peptide band of 70-80 kDa in liver (106). Very recently, MacLeod's and Closs' groups obtained antibodies capable of distinguishing between CAT-2 and CAT-2a. This is not an easy task, since the main difference between the two mCAT-2 splice variants is an eight-amino acid segment with five substitutions (see Fig. 2), but full reports on this issue have not yet appeared (C. MacLeod and E. Closs, personal communication). Therefore, confirmation, at the protein level, of the tissue distribution and regulation of the expression of the CAT-2 isoforms is expected to be reported soon. No antibodies have been reported against the rodent CAT-3 or the human CAT-4 proteins.

The specific tissue distribution and regulation of the expression of the different CAT transporter isoforms (see sect. IIA5) suggests a high level of regulation of the expression of the corresponding gene promoters and splice variants (336). MacLeods's (151) group has addressed this question for the mCAT-2 gene. The promoter region of mCAT-2 is extremely complex, with several 5'-untranslated regions (UTR) expanded in a genomic region of 19 kb from the first 5'-coding exon. For more detailed information, the reader is directed to the review by MacLeod and Kakuda (336). MacLeod's lab (151) described five exon 1 isoforms (1a, 1b, 1b/1c, 1c, and 1d) that splice into a common sequence 16 bp 5' of the AUG start methionine codon (151). Kavanaugh et al. (267) described from a liver clone a putative complex additional 5'-UTR (named 1e) of 515 bp, with six initiation and termination codons that precede the translation start codon, which is subjected to posttranscriptional regulation. Promoter 1a (at the far end of the complex promoter region) predominates over the others in every cell type and tissue examined (336). The promoter of exon 1a is a TATA-less one with staggered initiation, GC rich, and with several SP1 and CAC boxes. Liver (where only the mCAT-2a variant is expressed) and activated macrophages (where only the mCAT-2a variant is expressed) use promoter 1a exclusively (151); this demonstrates that promoter usage does not dictate the splicing events that render the mCAT-2 and mCAT-2a splice variants (336).

Posttranscriptional regulation of CAT-1 gene expression has been reported in liver regeneration (25), a model that induces system y+/CAT-1 isoform expression (see sect. IIA5). This study shows that accumulation of CAT-1 mRNA in the regenerating rat liver is due to posttranscriptional regulation, and there is no increased transcriptional activity (run-on experiments) of the gene; this posttranscriptional regulation is sensitive to cycloheximide. In the rat, CAT-1 produces two transcripts (7.9 and 3.4 kb in length), which represent alternative polyadenylation signal usage (the long transcript uses a consensus polyadenylation sequence, whereas the shorter uses a noncanonical signal); both transcripts accumulate in the regenerating liver. The specific 3'-UTR of the long transcript contains destabilizing AU-rich sequences that are associated with a shorter half-life (90 and 250 min for the long and short trancripts, respectively); interestingly, the longer transcript accumulates to higher levels than the shorter transcript. All this suggests that protein factors control the stability of CAT-1 mRNA through the long-specific 3'-UTR and through common sequences for both transcripts. In fact, cycloheximide administered in vivo to control rats upregulates the levels of the long transcript in several tissues but unfortunately not in liver, suggesting tissue-specific regulation of the half-life of CAT-1 transcripts. Similarly, in rat hepatoma FTO2B cells, where CAT-1 expression decreases with confluency, the relative abundance of the two transcripts also varies with confluency; the shorter transcript decreases faster than the longer (610). All this suggests that the 3'-UTR sequences of CAT-1 transcripts may be involved in the regulation of polyadenylation and/or stability of the transcripts (25). Unfortunately, in these studies showing differential expression of the two CAT-1 transcripts, no attempt was made to correlate transcript levels and CAT-1 protein abundance to assess translation efficiency. As an additional regulation mechanism of CAT-1 gene expression, in analogy with CAT-2 gene, these authors quoted unpublished results that suggest multiple promoter usage for the rat CAT-1 gene, but no report or confirmation of this is yet available.

To our knowledge, information on the promoter regulation of CAT genes to explain the modulation of CAT-1 and CAT-2 expression (see sect. IIA5) has not been reported.

2. Transport properties of CAT transporters

The characteristics of the amino acid transport acitivities elicited by the murine CAT transporters mCAT-1, mCAT-2, and mCAT-2a have been studied in Xenopus oocytes and are summarized in Table 3.

1) There is sodium-independent transport of cationic amino acids (e.g., L-arginine, L-lysine, and L-ornithine) with high affinity (Km in the micromolar range) by mCAT-1 and mCAT-2 (106, 108, 242, 267, 281, 590) and with low affinity (Km in the millimolar range for L-arginine) by mCAT-2a (106, 267). It is worth mentioning that Km values for L-ornithine influx are higher than those for L-arginine and L-lysine via mCAT-2, suggesting true differences in the extracellular recognition of these substrates (105). As shown in Table 3, there are discrepancies between labs as to the Km values of cationic amino acids in mCAT-1 and mCAT-2 transporters. Transport of cationic amino acids via mCAT-1 is voltage dependent; hyperpolarization increases the Vmax and decreases the apparent Km for influx (the reverse is true for efflux) (264). Closs (105) argued that differences of oocyte membrane potential in different labs and experimental conditions could underlie the variation reported for Km values.

2) For mCAT-1, mCAT-2, and mCAT-2a (242, 267, 590), the transport expressed has been shown to be electrogenic (positive charge follows the cationic amino acid flux) and stereospecific (i.e., Km in the millimolar range for D-cationic amino acids).

3) mCAT-1, mCAT-2, and mCAT-2a present trans-stimulation of arginine uptake, with mCAT-1 being more sensitive to this phenomenon (106, 108). mCAT-2a transport activity is largely independent of trans-side substrate (Table 3) (106, 108, 267).

4) Electrophysiological studies (242, 590) showed that mCAT-1 and mCAT-2 transport the zwitterionic amino acids homoserine and cysteine only in the presence of sodium (242, 590). Expression of low-affinity histidine uptake is also elicited by mCAT-1 and mCAT-2; for mCAT-1, it has been shown to be partially dependent on the presence of cis-sodium. At low pH, when histidine is protonated, this amino acid becomes a better substrate, demonstrating selectivity of CAT transporters for dibasic amino acids. For mCAT-2a (106), transport of zwitterionic amino acids in the presence of sodium has not been observed, although in these studies transport activity was measured by radioactive amino acid uptake, a less sensitive method than electrophysiological measurements.

At present, there are no data available on the amino acid transport activity expressed by hCAT-4 (but expression of hCAT-4 in oocytes resulted in increased L-arginine uptake; Sebastio, personal communication), and there are two reports on rat and mouse CAT-3 (Table 3). Transient expression of rCAT-3 in COS-7 cells resulted in sodium- and chloride-independent transport of radiolabeled L-arginine with an apparent Km of ~100 µM, inhibitable by cationic amino acids and dependent on membrane potential, as expected for a cationic amino acid transporter (219). Expression of mCAT-3 resulted in high-affinity, sodium-independent transport of dibasic amino acid, which shows trans-stimulation (229). Therefore, CAT-3 together with CAT-1 and CAT-2 transporters are high-affinity cationic amino acid transporters in contrast to the CAT-2a isoform.

It is worth mentioning that the proposed channel mode of action described for the sodium-dependent amino acid transporters, like members of the superfamilies of neurotransmitters and of excitatory amino acid transporters (see the corresponding sections), has not been described for the CAT transporters. It is interesting that these transporters and the proteins rBAT and 4F2hc, which essentially do not mediate sodium-dependent transport, do not seem to have a channel mode of action (90, 110).

Closs et al. (108) obtained surprising data on the accumulation capacity of mCAT-1, -2 and -2a transporters in oocytes at a nonphysiological extracellular concentration of 10 mM L-arginine. Incubation of oocytes in a high L-arginine concentration (10 mM) for 6 h, assuming an oocyte space distribution of ~180 nl/oocyte (90), leads to 0.6-fold accumulation in mCAT-1-expressing oocytes, 1.4-fold in mCAT-2-expressing oocytes, and 6-fold in mCAT-2a-expressing oocytes. These differences have been interpreted as the consequence of an apparent intracellular substrate affinity of mCAT-2a smaller than that of mCAT-1 and mCAT-2 (105). In our opinion, thermodynamic gradients are unlikely to be the result of substrate affinity differences. The regular oocyte membrane potential (-30 to -50 mV) is valid for an accumulation gradient of a positive charged substrate (i.e., L-arginine) of six- to eightfold. Interestingly, accumulation of 10 mM L-arginine in mCAT-2a-expressing oocytes from a sodium-free medium tends to this value (108). Why do mCAT-1 and mCAT-2 not reach the same accumulation gradient? In the experiments by Closs et al. (108), the membrane potential was not clamped, and therefore, the impact of the high L-arginine flux (10 mM extracellular concentration) was not controlled. Additional work at different extracellular substrate concentrations and at constant membrane potentials is needed to characterize the accumulation capacity of these transporters and the transport mechanisms underlying any possible difference between them.

For the human CAT-1, -2, and -2a counterparts, as for the mouse analogs, oocyte expression showed cationic amino acid transport of high affinity that was sensitive to trans-stimulation for hCAT-1 and hCAT-2 and cationic amino acid transport of low-affinity that was only slightly sensitive to trans-stimulation for hCAT-2a (unpublished data quoted in Ref. 105).

On the basis of all these characteristics (transport properties and tissue distribution), these cDNA (CAT-1, CAT-2, CAT-2a, and CAT-3) have been attributed to system y+ and its variants (242, 281, 590). In summary, transport activity elicited by mCAT-1, mCAT-2, and rCAT-3 expression is similar but with subtle differences (105, 219): sodium-independent high-affinity transport for cationic amino acids, with a slightly higher apparent affinity for mCAT-1, which is more sensitive to trans-stimulation. No data are reported on trans-stimulation via CAT-3. The transport properties and tissue distribution of CAT-1, CAT-2, and CAT-3 are consistent with subtle variants of system y+ (reviewed in Refs. 126, 600), the common cationic amino acid transport activity of mammalian cells. Most probably, CAT-2a represents a low-affinity liver variant of y+ activity. White and Christensen (601) described a low-affinity transport of L-arginine in primary hepatocytes not subjected to trans-stimulation and concluded that the classical y+ activity was absent or altered in hepatocytes. However, Van Winkle (574) suggested that the transport activities expressed by CAT-1 and CAT-2 could fit those of systems b+. Van Winkle et al. (579) also reassessed mCAT kinetic data, suggesting the presence of both a high- and a low-affinity component for each protein (mainly for mCAT-2a) when expressed in oocytes. At present, it is not clear whether this complex kinetic behavior represents an artifact of the expression model, different conformation or oligomeric states, or interaction with endogenous proteins. A more careful characterization of these amino acid transport activities based on inhibition by amino acids and analogs is needed to clarify this issue. Similarly, cell knockout or antisense experiments, like those reported for system bo,+-like in opossum kidney (OK) cells (374), would clarify the contribution of CAT transporters to y+/b+ transport activity in cells. In this sense, uptake studies in cells derived from the null knockout CAT-1 mice (419) may help to clarify this issue.

3. Protein structure of CAT transporters

Structural information on CAT transporters is scarce. All CAT transporters identified lack a characteristic signal peptide, and therefore, the NH2 terminus is considered to be cytoplasmic (see Refs. 105 and 336 for review and Ref. 219 for rCAT-3). Most of the additional information available has been obtained from mCAT-1, and it has been extrapolated to CAT-2 and CAT-3 transporters since they show almost identical hydrophobicity profiles (105, 219). These profiles initially suggested two membrane topology models for CAT transporters: 12 TM (according to MacLeod's and Saier's groups) or 14 TM (according to Cunningham's group) (9, 106, 334, 448). The two models differ in the middle TM domains (TM domains VII and X of the 14-TM model are considered to be intracellular in the 12-TM model; see Fig. 2). The 12-TM model is supported by the fact that CAT transporters belong to the APC transporter superfamily of yeast, fungi, and eubacteria, which presumably contain 12 TM domains (448), whereas the proposed membrane topology of the first 8 TM domains of the homologous yeast and fungi permeases argue in favor of the 14-TM model (105, 508). Mutational analysis showed that the viral binding site of mCAT-1 (see Fig. 2) is located between TM V and VI, confirming the extracellular location of extracellular loop (EL) 3 in both models (8). Evidence in favor of the 14-TM model has been obtained: 1) antibodies directed against peptides of the EL3 and EL4 loops of mCAT-1 in the 14-TM model immunostained nonpermeabilized cells (605). This confirmed the extracellular location of these protein regions; the 12-TM model predicts an intracellular location for the EL4 protein region. 2) The glycosylation of CAT transporters has been demonstrated by endoglycosidase F (endo F) treatment of immunodetected CAT-1 (i.e., antiserum raised against the COOH terminus of murine CAT-1) and CAT-2/CAT-2a (i.e., antiserum raised against COOH terminus of murine CAT-2a; a region that is identical to CAT-2) from mammalian cells or expressed in oocytes. These studies showed a broad glycosylated moiety of 3-9 kDa for mCAT-1 and mCAT-2a transporters (106, 280). Mutation in mCAT-1 of the two putative N-glycosylation sites Asn-223 and Asn-229 to histidine, conserved in all CAT transporters characterized (Fig. 1), results in a protein with identical SDS-PAGE mobility to the endo F-treated wild-type mCAT-1; mutation of either Asn residue results in an intermediate mobility. The 12-TM model predicts an additional, unconserved extracellular N-glycosylation site in mCAT-1 (Asn-373, located in the intracellular IL4 of the 14-TM model). Mutation of Asn-373 to histidine does not affect glycosylation of mCAT-1. These studies (280) demonstrated that Asn-223 and Asn-229 are the glycosylated residues of mCAT-1, and therefore extracellular, like the loop EL3 in the 14-TM model (Fig. 2), and that Asn-373 (in the IL4 loop of the 14-TM model) might not be located extracellularly, which thus favors the 14-TM model.

To our knowledge, extensive studies in search of evidence for the 14-TM model, like those performed with the GABA transporter GAT1 (38), the glycine transporter GLYT1 (401), and glutamate transporters (585, 498), have not been reported.

4. Structure-function relationship

Swapping chimeras with the divergent amino acid segment of CAT transporters, mutational analysis and studies related to the interaction with the murine ecotropic leukemia virus provided the core of our knowledge of the structure-function relationship for CAT transporters.

The apparent substrate affinity, maximum transport rate, trans-stimulation, and accumulation capacity are distinctive features of mCAT-2 and mCAT-2a (see Table 2, CATS). This suggests that these differential transport capacities are determined by the variant exon coding for the 41- to 42-amino acid residue divergent segment of the two proteins (see Fig. 2). Closs et al. (108) performed elegant studies, in which chimeric transporters with the backbone of mCAT-1 were completed with the divergent domain of mCAT-2 and mCAT-2a, and the backbone of mCAT-2 was completed with the corresponding domain of mCAT-1 (see Fig. 2). The transport characteristics (apparent Km, Vmax, trans-stimulation, and accumulation of L-arginine) of these chimeras expressed in oocytes are similar to those of the divergent region. Interestingly, the recently cloned rat CAT-3 expresses high-affinity (Km ~100 µM) L-arginine uptake in COS-7 cells, and its divergent domain is more similar to that of CAT-1 and CAT-2 than to that of CAT-2a (219) (see Fig. 2). These data suggest that the divergent protein domain of CAT transporters has an impact on all transport properties, and therefore, it might have a role in substrate recognition, turnover number, and the translocation mechanism. As indicated in Fig. 2, the divergent domain corresponds to the intracellular loop IL4 (including a few amino acid residues of TM domains VIII and IX) in the 14-TM domain model of CAT transporters.

Residues involved in the reported mutational analyses of mCAT-1 are indicated in the homologous position in the mCAT-2 protein model depicted in Figure 2. N-glycosylation is not required for transport function of mCAT-1 (280); the unglycosylated mutant (double Asn to His mutation at positions 223 and 229: see Fig. 1) expresses an unaffected transport activity in oocytes. The glutamate residue at position 107 of mCAT-1 is conserved in the TM domain III of all known CAT transporter sequences (see Figs. 1 and 2) and also in the yeast transporters for arginine, histidine, and choline of the APC family (589). This residue is required for transport activity in mCAT-1 protein expressed in mink CCL64 lung fibroblasts (589). Substitution by aspartate led to a loss of transport activity; interestingly, substitution by the uncharged glutamine residue did not affect transport activity (data by Kim and Cunningham quoted in Ref. 105). All these substitutions led to mCAT-1 protein expressed in the plasma membrane of the transfected cells as demonstrated by its role as a virus receptor (infectivity and viral glycoprotein gp70 binding). All this suggests that the carbon backbone size but not the negative charge of residue glutamate 107 of mCAT-1 determines transport function for the CAT transporters.

Meruelo and co-workers (621) and Cunninngham and co-workers (8) identified by domain swapping and mutational analyses the sequence NVKYGE (amino acid residues 232-237 in mCAT-1) within EL3 (see the corresponding position of this protein segment in Fig. 2) as essential for virus envelope binding and infection; swapping the above-mentioned sequence into the human CAT-1 conferred infectivity and virus binding (8). This sequence is also present in the rat CAT-1 counterpart (610), but not in hCAT-1 or the known CAT-2 proteins (see Figs. 1 and 2). Interestingly, both mCAT-1 and rat CAT-1 serve as a receptor for the virus. Detailed description of the amino acid residues within the EL3 loop required for binding of the viral protein envelope gp70 and permissivity to infection (8, 267) has been reviewed by Closs (105). It is worth mentioning that none of the mutations examined to determine viral interaction with CAT transporters affected their transport activity. In contrast, interaction of mCAT-1 with the virus reduces its transport activity. Coexpression of mCAT-1 and glycoprotein gp70 resulted in a specific reduction of mCAT-1 glycosylation and transport activity; a decrease in transport activity also occurs with the unglycosylated double Asn to His mutant at residues 223 and 229 (280). Binding of glycoprotein gp70 to the transfected mCAT-1 results in noncompetitive inhibition of amino acid import via the murine CAT-1 with no effect on amino acid export (588). The effects of gp70 on transport kinetics led the authors to suggest that gp70 binding represents a steric hindrance that slows the rate-limiting step of the amino acid import cycle, a conformational transition of the empty transporter in which the binding site moves from the inside back to the outside of the cell, and that gp70 has no effect on the rate-limiting step of the amino acid export cycle. A similar mobile carrier hypothesis has been suggested for the two-directional operation of the system y+ (600). The above-mentioned data suggest that amino acid transport and virus receptor functions may be coupled; conformational changes of the transporter may lead to membrane fusion of virus and host cell (105). Interestingly, the transport defective mCAT-1 mutant glu107asp mediates binding of glycoprotein gp70 and virus infection (589). This suggests that transport and receptor function may be uncoupled, but this is still an open question, since conformational changes for the transport-defective mutant have not been ruled out.

5. Physiological role of CAT transporters

An intriguing question is why there is such a variety of CAT transporters in mammalian cells. Cationic amino acids are needed for protein synthesis, urea synthesis (arginine), and as precursors of bioactive molecules (arginine and ornithine are substrates for the synthesis of urea and nitric oxide as well as polyamines, respectively). Then, what does each CAT transporter isoform contribute to the supply of substrates for these purposes? The contribution of other protein structures to cationic amino acid transport, like rBAT (system bo,+-like) and 4F2hc (system y+L-like) are discussed in section IID. The nearly ubiquitous CAT-1 isoform most probably corresponds, as discussed before, to the classical system y+, a high-affinity and electrogenic cationic amino acid transport system that allows accumulation of these substrates within the cells for general metabolic purposes. Consistent with this, the null knockout CAT-1 mice are smaller (limited accretion) at birth (419). In this sense, at a physiological extracellular concentration of L-arginine (50 µM), a high expression level of mCAT-1 in oocytes allows the maintenance of an L-arginine gradient across the plasma membrane of ~19-fold (90).

The known examples of upregulation of CAT-1 transporter expression favor this general role of the classical system y+/CAT-1 isoform. The expression of this gene is enhanced in proliferating cells (e.g., T and B lymphocytes activated by concanavalin A and bacterial lipopolysaccharide, rapidly growing cells infected with Friend leukemia virus, and a variety of tumor cells of different origin) (620). In liver regeneration, CAT-1 expression (both mRNA and protein) is induced a few hours after hepatectomy (25, 610). Recently, Hatzoglou's group (25) showed that CAT-1 could be considered as a delayed early growth response gene in the regenerating liver that requires protein synthesis for its upregulation. In keeping with this, ecotropic retroviruses infect hepatocytes during fetal development and liver regeneration, but not in adult hepatocytes (198). This supports a role of CAT-1 transporter in accretion, and as discussed by Wu et al. (610), a role of this transporter in the supply of ornithine for polyamine synthesis. Interestingly, the key enzyme in polyamine synthesis, ornithine decarboxylase, peaks during the G1 phase (145), and ornithine levels rise after partial hepatectomy (149). In addition to proliferation, hormone treatment (insulin and dexamethasone) also induces mCAT-1 expression in liver (610). A recent report also links CAT-1 expression with cell proliferation. Perkins et al. (419) reported that the homozygous null knockout CAT-1 mice develop anemia and die after birth. Erythroid maturation is defective in these mice because of a specific defect in cell proliferation and/or differentiation. In addition, this suggests that CAT-1 transporter is the main contributor to the cationic amino acid supply to erythroid progenitor cells. The specific contribution of CAT-1 transporter to the intracellular accumulation of cationic amino acids in those cells that express additional CAT isoforms (e.g., in brain, heart, skeletal muscle, uterus, ovary, testis, and placenta) is difficult to assess by indirect determinations, like transcripts and protein levels (see below). Specific knockout and antisense experiments are needed to delineate the contribution of each CAT transporter to the macroscopic amino acid flux through the plasma membrane of the cells.

The low-affinity high-capacity transport properties, the accumulation capacity through the plasma membrane at high extracellular substrate concentrations, and the exclusive expression of CAT-2a isoform mRNA, but not CAT-1 (both protein and mRNA) or CAT-2 (mRNA) isoforms, in liver have been envisaged as constituting a kinetic barrier between the hepatic urea cycle and extracellular arginine (105). Furthermore, on the basis of the low intracellular concentration of arginine in liver, it is unlikely that this amino acid is released through the activity of CAT-2. All this is consistent with the lack of activity of the classical high-affinity system y+ in the hepatocyte plasma membranes, which protects extracellular L-arginine from hydrolysis by hepatic arginase (600, 601). The hepatocyte CAT-2a transporter would allow rapid accumulation of cationic amino acids only at high plasma concentrations, leaving sufficient substrates in circulation for cells expressing the high-affinity CAT isoforms (105). In keeping with this, expression of CAT-1 isoform occurs in liver when the urea-cycle enzymes are downregulated (e.g., liver regeneration, insulin treatment, low-protein diet) (610). Interestingly, stress conditions (partial hepatectomy, surgical trauma, and fasting) upregulate mCAT-2a in skeletal muscle (K. D. Finley, quoted in Ref. 336); the CAT-2/-2a isoforms are prevalent in this tissue (242). Some of these stress situations have a muscle proteolytic state in common (309, 324). The possible physiological role of CAT-2a upregulation induced by stress conditions such as fasting in skeletal muscle is far from understood. MacLeod and Kakuda (336) suggested that this is a mechanism to prevent depletion of those amino acids from the tissue with active proteolysis. In this regard, it should be mentioned that the rate of release of lysine or arginine in the rat perfused hindquarter preparation is not modified in response to 48 h of fasting (464). Brief starvation in humans has been reported to enhance the release of lysine, but not of arginine, from skeletal muscle (429). Studies should be performed to determine the role of CAT-1, CAT-2a, and CAT-4 in the metabolism of cationic amino acids in skeletal muscle.

Ashcroft and co-workers (504) studied the transport mechanisms involved in the stimulation of insulin secretion by L-arginine in mouse pancreatic beta -cells. This work suggests that L-arginine raises the intracellular concentration of Ca2+ and stimulates insulin secretion as a consequence of its electrogenic transport into this cells. The expression of mCAT-2 and mCAT-2a in beta -cells was demonstrated by RT-PCR. L-Arginine produced a dose-dependent increase in the intracellular concentration of calcium, which suggests that the low-affinity mCAT-2a is the cationic amino acid transporter responsible for the secretagogue action of this amino acid. Specific mCAT-2a knockout experiments in beta -cells or in the whole animal are needed to demonstrate the role of mCAT-2a in the insulin secretagogue action of L-arginine.

The CAT-1 mRNA is constitutively expressed in mature resting and activated T cells and splenocytes and resident macrophages (242, 336). Activation of these cells mainly induces the CAT-2 transporter isoform. SL12 thymoma cell clones, a model system of thymocyte differentiation, show developmental regulation of mCAT-2 during thymocyte maturation (602). The mCAT-2 gene is downregulated in normal and mature thymocytes until it is activated by mitogens or antigens (151, 242, 334). Peripheral blood lymphocytes and quiescent splenocytes exhibit little transport of lysine via systems y+ and y+L (64, 127). Upon activation, T cells rapidly increase system y+ transport activity and mCAT-2 mRNA levels in parallel (64). In addition, a transient increase in mCAT-1 expression has been reported (336). Human CAT-1 antisense experiments in phytohemagglutinin-induced lymphocytes reduce the induced system y+ transport activity only partially (88). Combined experiments with antisense sequences of CAT-2 and CAT-1 have not been addressed.

The release of nitric oxide is an important mediator of macrophage function (389, 536). Activation of macrophages by bacterial lipopolysaccharide (LPS) and interferon-gamma (IFN-gamma ) produces the parallel induction of the expression of CAT-2 (108) and the nitric oxide type II synthase (inducible nitric oxide synthase or NOS II), and nitric oxide production (129, 519, 612). Unpublished data from MacLeod's lab showed specific induction of mCAT-2 but not of mCAT-2a after activation of macrophages (quoted in Ref. 336). Nitric oxide synthase II uses L-arginine as a substrate (373), and its activity is dependent on extracellular L-arginine (55, 211, 230, 611). In parallel with macrophage activation, there is an increase in arginine transport rate (55, 473), due to system y+ (36). More recently, concomitant induction of NOS II and CAT-2 transcripts (>20-fold) (no CAT-1 or CAT-2a transcripts) and nitric oxide production (>30-fold) and high-affinity y+ activity (3- to 4-fold) has been observed in brain astrocytes treated with LPS/IFN-gamma (512). Interestingly, actinomycin D blocks the increase in system y+ activity and nitric oxide production due to LPS/IFN-gamma . This indicates that the contribution of CAT-1 to arginine uptake for nitric oxide production is negligible compared with that of CAT-2. Similar to macrophages and astrocytes, vascular smooth muscle cells respond to cytokine (interleukin-1beta and tumor necrosis factor-alpha ) treatment by parallel increase in mCAT-2 and NOS II mRNA levels, without any effect on mCAT-1 mRNA levels (167). In contrast, angiotensin II increases mCAT-1 mRNA levels in vascular smooth muscle cells without induction of the nitric oxide secretion pathway (323). No attempt was made in any of these studies to corroborate changes in CAT-2 and CAT-1 expression at the protein level. All this suggests that CAT-2 transporter accounts for this increased uptake. Knockout experiments are needed to elucidate the participation of CAT-2 and CAT-1 isoforms in the supply of arginine for nitric oxide synthesis by NOS II. If this hypothesis is verified, the mechanism by which arginine supplies nitric oxide synthesis by NOS II will be the induction of a specific y+ isoform (with transport characteristics very similar to the widespread y+ isoform) but through the induction of its specific gene promoter. In addition, whether differential transport properties or specific location in plasma membrane domains of CAT-1 and CAT-2 isoforms channel arginine for NOS II has not been addressed. An interesting experiment would be to induce NOS II activity in the context of knockout of CAT-2 gene and overexpression of CAT-1 gene.

B. Superfamily of Sodium- and Chloride-Dependent Neurotransmitter Transporters

The cloning of rat and human brain GABA transporters (GAT1) in 1990 (184, 391) and the human norepinephrine transporter (NET) in 1991 (405) was the starting point for the isolation of related cDNA that constitute the sodium- and chloride-dependent neurotransmitter transporter superfamily (for review, see Refs. 54, 253, 390, 565). Rat brain GABA transporter was purified and microsequenced by Kanner's group (439), and then related oligonucleotide probes were used, in a collaborative effort between Nelson's, Lester's, and Kanner's groups, to isolate rat GAT1 from a rat brain cDNA library (184). Amara's group (405) isolated the NET cDNA from the human neuroblastoma cell line SK-N-SH by expression cloning, after accumulation of the norepinephrine analog [125I]iodobenzylguanidine in COS-1 cells (405). In 1992, Burg and Handler and co-workers (614) reported the isolation of the canine betaine transporter (BGT-1) cDNA, which was obtained by expression cloning in oocytes from a size-fractionated cDNA library constructed from MDCK cells maintained in hypertonic medium; mRNA from MDCK cells maintained in hypertonic medium showed increased induction of betaine uptake in oocytes (452). Different strategies based on homology with these previous and the on-coming cDNA sequences from different cDNA sources resulted in the cloning of the rest of transporters of this superfamily (for references for the amino acid transporters, see Table 4).

 
View this table:
[in this window] [in a new window]
  		TABLE 4.   Sodium- and chloride-dependent amino acid transporters within the superfamily of neurotransmitter transporters

The members of this superfamily have a common membrane topology prognosis of 12 TM domains, and as a general characteristic that defines this superfamily, the transport activity of these carriers has been found to be sodium and chloride dependent. Substitution of sodium by lithium is not tolerated, and substitution of chloride by other anions ranked the potency order as follows: chloride >=  Br- > NO -3 > gluconate > acetate (reviewed in Refs. 44, 260, 302, 465). These transporters (15 putative cDNA have been identified without considering species counterparts) share a high level of homology (30-65%). Homologous transporters have been identified in insects, worms, and yeast (13, 318; reviewed in Ref. 390). On the basis of substrate specificity and extensive amino acid sequence identity, this superfamily has been divided into two major subfamilies (reviewed in Refs. 253, 390, 565). The first major subfamily corresponds to the sodium- and chloride-dependent neurotransmitter transporters, which comprises three minor subfamilies: 1) biogenic monoamine transporters, which include dopamine, norepinephrine, and serotonin transporters that show an amino acid sequence identity between 40 and 47% with the members of the other subfamilies (53, 213, 279, 405, 494, 567, 570; for review, see Refs. 54, 253, 390, 565) and two subfamilies for amino acid transporters; 2) GABA and taurine transporters (see Tables 4 and 5), also including creatine transporters (2 have been isolated, one of which corresponds to the previously identified choline transporter as revealed by expression studies and cellular distribution; Refs. 34, 172, 186, 346, 388, 480, 509), with an amino acid sequence identity that ranges between 49 and 69%; and 3) glycine and proline transporters that show a similar level of homology with each other and with the rest of the superfamily (amino acid sequence identity ranges from 43 to 48%). The second major subfamily comprises three "orphan transporters" with structural characteristics that differ from the previous major subfamily (large second and fourth putative extracellular domains with a canonical N-glycosylation site; Refs. 139, 317, 345, 566; for review, see Ref. 565).

 
View this table:
[in this window] [in a new window]
  		TABLE 5.   Tissue distribution and transport characteristics of expressed Na+- and Cl--dependent amino acid transporters

In the present review, only the amino acid transporters of this superfamily are considered. Excellent reviews concerning the monoamine, creatine, and orphan transporters are available (11, 22, 54, 302, 565). Because of the high level of confusion with different names for the same transporter cloned from different species (e.g., see different names for the GABA transporters in rat and mouse in Table 4), in the present review we use the nomenclature based on the rat GABA transporter cDNA (see Table 4). In this section, attention is paid to tissue distribution, transport characteristics, protein structure with relation to transport function, and the physiological role of these transporters in both neural and peripheral tissues. The amino acid transporters of this superfamily of neurotransmitter transporters do not necessarily have a role in neurotransmission, since, for example, the GABA transporter (GAT-2), the betaine/GABA transporter (BGT-1), the taurine transporter (TAUT), and the glycine transporter GLYT1 are expressed in nonneural tissues (see Table 5).

1. Tissue expression

The tissue distribution of GAT-1 was monitored by immunocytochemistry before the "cloning era" of these transporters (442). For the rest of the amino acid transporters of this superfamily, the tissue distribution and the transcript size were initially studied by Northern blot (see Table 5) and in situ hybridization analysis. All these transporters are expressed in the central nervous system (CNS), and some of them are also expressed in peripheral tissues. The GABA transporters GAT-1 and GAT-3, the glycine transporter GLYT2, and the proline transporter PROT seem to be specific to the CNS (57, 103, 155, 184, 240, 314, 315, 391, 489).

A) GABA TRANSPORTERS. Several studies address the distribution of the four GABA transporters (GAT-1, -2, and -3 and BGT-1) in the CNS, including a recent review by Borden (56). The three high-affinity (Km values from 1 to 20 µM) GABA transporters GAT-1, -2, and -3 are expressed in brain and retina (60, 65). Antibodies raised to mouse and rat GAT-1 showed a relatively even distribution of this GABA transporter in all parts of the brain, in parallel with the distribution of GABA (330, 442). This transporter is present mainly in the neuropil, but also in processes of glial cells in the cerebral cortex, cerebellum, hippocampus, and retina (364, 375, 442, 450). Similarly, both in neuronal and type 2 astrocyte cultures, most of GABA transport (~75%) shows a pharmacological profile specific for GAT-1 (e.g., inhibition by the GAT-1-selective ligand NNC-711) and in parallel GAT-1 transcript levels are abundant (61). The distribution pattern of GAT-1 makes it a good candidate for the GABA transporter that functions in the GABAergic synapses (see review in Ref. 240). Thus immunofluorescence, electron microscopy, and confocal immunolocalization of GAT-1 in rat hippocampus, retina, and cerebellum correlate with its involvement in the termination of the action of GABA by its uptake from the extracellular space into GABAergic axon terminals and astrocytes, where it could also play the additional role of regulating the extracellular concentration of GABA (216, 375, 450). In addition, there is generally good correlation between GAT-1 (mRNA and protein) and glutamic acid decarboxylase-like immunoreactivity (GAD67; i.e., the enzyme responsible for GABA synthesis) in striatum and cerebral cortex (24, 364, 525). However, the expression of GAT-1 is not restricted to neurons involved in GABAergic synapses. Thus, in cerebral cortex, the GAT-1 uptake system (mRNA and protein) is more extensive than the GABA synthesizing system (364, 525) in the cerebellar cortex, GAT-1 and GAD67 mRNA do not correlate in Purkinje cells (447), and in the rat spinal motor neurons, there is expression of GAT-1 mRNA but not GAD67 (506). All these results support the hypothesis that GAT-1 plays its presynaptic role in GABAergic synapses but, in addition, may have postsynaptic roles, and it may regulate the extracellular concentration of GABA in glial cells.

Northern blot analysis and in situ hybridization histochemistry showed a complementary distribution of GAT-3 and GAT-1 (103, 315); GAT-3 has a strong expression in spinal cord, brain stem, thalamus, and hypothalamus and is weakly expressed in cerebellum, hippocampus, cerebral cortex, and striatum, where GAT-1 transcripts are abundant. In many instances, the localization of GAT-3 (mRNA and protein) correlates well with characterized populations of GABAergic neurons and glutamic acid decarboxylase immunoreactivity, as in the medial septum-diagonal band complex, but in others, the distribution is dissimilar, such as in the cerebellum (103) and cerebral cortex (363). The distribution of GAT-3 in brain cells is controversial. Initially, GAT-3 was described predominantly as a neuron transporter (103), but later its presence in astroglial processes was demonstrated by immunocytochemistry (450), and in cerebral cortex it localized exclusively to astrocytic processes (363). Similarly, GAT-3 mRNA is present in neuronal and in type 2 astrocyte cell cultures (61). Recently, in electron microscopic immunolocalization studies in the developing rat brain (i.e., embryonic and early postnatal stages), GAT-1 and GAT-3 protein showed coordinate expression (239): GAT-1 was detected in gray matter and growing axons, and GAT-3 in radial glial cell fascicles oriented perpendicularly to the axons expressing GAT-1. Therefore, at present, it seems that GAT-3 is mainly a glial transporter.

Among the high-affinity GABA transporters, GAT-2 is the only one expressed in peripheral tissues in addition to brain and retina (60, 223, 315). Initial Northern blot analysis revealed that GAT-2 mRNA levels in brain are developmentally regulated, being more abundant in the brains of newborn mice than in the adult or fetal brain (315). Two studies compared the distibution of GAT-2 with the other two high-affinity GABA transporters in retina and brain (216, 223). Immunoreactivity of GAT-2 was faint throughout the brain but was concentrated in the arachnoid and ependymal cells, a completely different distribution from that of GAT-1 and GAT-3 proteins (see above). Similarly, in the retina, GAT-2 protein localizes to the retinal pigment epithelium layer, nerve fiber layer, and cilliary body epithelium, whereas GAT-1 and GAT-3 localize to amacrine neurons and Muller glial cells, respectively. In rat brain-derived cultures containing O-2A progenitor cells and type 2 astrocytes, GAT-2 mRNA is the second most abundant after GAT-1 transcripts; in contrast, GAT-2 transcripts are not detected in neuronal cultures or type 1 astrocyte cultures (61). These results suggest that GAT-2 may be related to nonneuronal function in brain and retina.

Northern blot analysis showed an even distribution of BGT-1 transcripts in mouse (~5 kb in length) and human brain (>= 3 kb and a less conspicuous band of 4.1 kb) (59, 322, 446). BGT-1 is most probably a glial transporter; its transcripts were observed in type 1 and type 2 astrocyte cultures, but not in neuronal cell cultures (61). The presence of BGT-1 mRNA in human and mouse brain (59, 322, 446) suggests that, at least in these species (BGT-1 was not shown in canine brain; Ref. 614), this transporter could participate in brain osmoregulation (59) (see sect. IIB6). Betaine is present in brain at low concentrations, but levels increased after salt loading (207). Handler's group (537) showed the presence of different 5'-UTR in canine BGT-1 transcripts due to splice variants and the use of three tissue-specific promoters.

The GABA transporters GAT-2 and BGT-1 are expressed in peripheral tissues (59, 60, 315, 322, 446, 614): mouse and rat GAT-2 are expressed in liver and kidney and human and mouse BGT-1 are expressed in kidney medulla and liver. In addition, one report also showed expression of human BGT-1 in placenta, heart, and skeletal muscle (446). In situ hybridization and RT-PCR of microdissected nephron segments revealed that BGT-1 is predominantly expressed in the medullary thick ascending limbs of Henle's loop and the inner medullary collecting ducts (365). In the polarized epithelial cell model MDCK, Handler's group (613) showed that BGT-1 localizes to the basolateral membrane, consistent with its role in protecting cells in the renal medulla from hypertonicity (385, 387). Caplan and co-workers (5, 423) addressed the polarized expression of the four GABA transporters (GAT-1, -2, and -3 and BGT-1) expressed in MDCK and in freshly isolated hippocampal neurons: 1) GAT-1 localized exclusively to the axons of cultured neurons and to the apical pole of transfected MDCK cells, consistent with its axonal localization in vivo (364, 375, 442, 450). 2) GAT-3 was expressed in the apical membrane of transfected MDCK cells and in both axons and somatodendritic membranes of hippocampal neuron cultures. 3) In contrast, expressed GAT-2 and BGT-1 localized to the basolateral membranes of transfected MDCK cells; expression of BGT-1 in hippocampal neurons in culture localized to somatodendritic membranes. These studies are subsidiary to those of Simons and co-workers (130, 131, 221) showing a polarized correlation between axons and the epithelial apical pole, and somatodendritic membranes and epithelial basolateral membranes. Following this line, Caplan and co-workers (5) suggested that GAT-2 may have a basolateral location in kidney and liver and a dendritic location in neurons, but to our knowledge, the subcellular distribution of GAT-2 in epithelia and brain has not been determined. Very recently, Pietrini, Caplan, and co-workers (417), working with MDCK cells BGT-1, rat GAT-1 and human nerve growth factor receptor chimeras, suggested the presence of basolateral sorting information in the cytosolic COOH-terminal domain of MDCK cells BGT-1; a short segment within this domain (residues 565-572), rich in basic residues well conserved in human BGT-1 but not in rat GAT-1, contains information necessary for exit from the endoplasmic reticulum and for the basolateral localization of MDCK cells BGT-1 in these cells.

B) TAURINE, GLYCINE, AND PROLINE TRANSPORTERS. To our knowledge, the tissue distribution of the taurine transporter TAUT has been studied only by Northern blot analysis or RT-PCR. The TAUT transcripts are widespread, and its abundance varies between different studies and species. In general, however, it has been shown to be present in kidney cortex and medulla, small intestinal mucosa, brain, lung, retina, liver, skeletal muscle, heart, placenta, spleen, and pancreas (232, 316, 445, 503, 562). Reverse transcriptase-polymerase chain reaction also showed TAUT mRNA in human ovary, colon, and thyroid (232).

The tissue distribution of the glycine transporters, as revealed by Northern blot analysis, indicates that GLYT1 is expressed in the CNS and peripheral tissues, whereas GLYT2 is specific to the CNS (see Table 5). A recent review by Zafra et al. (622) summarizes the cellular localization studies of glycine transporters at the protein and mRNA levels in brain (2, 62, 183, 237, 314, 328, 502, 623, 625). Both GLYT1 and GLYT2 are mainly expressed in caudal areas, and in addition, GLYT1 has a moderate expression in forebrain areas. The distribution of GLYT2 is consistent with the distribution of the inhibitory glycine receptor (immunocytochemistry and strychnine binding studies) and with neurons with high glycine content (17, 237, 623): high expression in the dorsal and ventral horn of the spinal cord, auditory system, and nuclei of the cranial nerves, and low expression in cerebral hemispheres. The distribution of GLYT1 in caudal areas is more widespread than that of the glycine receptors detected by strychnine binding (623). This indicates first, association of both transporters with the inhibitory glycinergic neurotransmission and a role in the termination of the glycinergic action, and second, additional roles for GLYT1 in nonglycinergic regions. Electron microscopic immunolocalization showed a glial (perikarya and processes) location for GLYT1 and an axonal (presynaptic terminals) location for GLYT2 (623). The location of GLYT1 in neurons is at present controversial: some authors did not observe neuronal expression of GLYT1 mRNA (2, 183), whereas others (62, 502, 625) reported in situ hybridization signal in neurons of the spinal cord, brain stem, and cerebellum and also in forebrain regions (cortex, hippocampus, thalamus, hypothalamus, and olfactory bulb). Immunocytochemistry studies failed to detect the neuronal form of the protein clearly (except in the retina; GLYT1 protein was localized in the amacrine neurons, Ref. 623), but they confirmed the glial expression of GLYT1 (238, 623). This controversy might be because of the use of probes with different isoform specificity and antibodies that do not react with the neuronal form of the protein (622). The specific mRNA distribution of GLYT1-1a and 1b/1c (GLYT1-1b probes used in this study do not distinguish between 1b and 1c variants) isoforms has been studied by Borowsky et al. (62): GLYT1-1a is found only in gray matter, whereas GLYT1-1b/1c is found exclusively in fiber tracts. Furthermore, GLYT1-1b/1c is found in all white matter, whereas GLYT1-1a distribution parallels the distribution of mRNA for the strychnine-sensitive and strychnine-insensitive glycine receptor in olfactory bulb, hippocampus, cerebellum, and spinal cord. For the presence of GLYT1 in areas without glycine receptor expression, two explanations have been offered (622): 1) Smith et al. (502) suggested a role of GLYT1 in modulating glutamatergic transmission through the activity of some N-methyl-D-aspartate (NMDA) receptors (see sect. IIB6), and 2) association of GLYT1 with non-strychnine-sensitive glycine receptors; in this sense, the mRNA of the beta -subunit of the glycine receptor and GLYT1 show similar distribution (62, 625).

GLYT1 is expressed in peripheral tissues (see Table 5). The GLYT1-1a variant, but not 1b/1c variants, is expressed in rat peripheral tissues: liver, spleen > lung > stomach, uterus > pancreas, kidney (i.e., relative abundance of transcripts) (62). Expression of GLYT1-1a in lung, spleen, liver, and thymus occurs in macrophages (62).

The mRNA coding for the brain-specific high-affinity L-proline transporter PROT was shown to be expressed in subpopulations of putative glutamatergic neurons in the olfactory bulb, cerebral cortex, and hippocampus (155). Western blot studies revealed the presence of PROT in enriched synaptic plasma membrane preparations and its absence from postsynaptic membranes (489, 580). This suggests a presynaptic regulatory role of L-proline or the PROT transporter in specific excitatory pathways in the CNS.

2. Transport characteristics

A) GENERAL CHARACTERISTICS. The characteristics of the transport activity associated with the expression of the amino acid transporters of this superfamily are summarized in Table 5. Sodium and chloride dependence has been reported for all these transporters (see Table 5).

B) GABA TRANSPORTERS. Of the eight amino acid transporters in this superfamily, only four transporters, GAT-1, -2, and -3 and BGT-1, induce uptake of GABA when expressed in cultured cells or in oocytes; GAT-1 to -3 are high-affinity transporters (Km values from 1 to 20 µM), and BGT-1 is a low-affinity GABA transporter (when expressed in oocytes it showed an apparent Km of <100 µM for GABA, even lower than that for betaine) (see Table 5). Pharmacological studies helped to distinguish the transport activity of the four GABA transporters (see Table 5). GAT-1 shows the highest sensitivity to cis-3-aminocyclohexanecarboxylic acid (ACHC), 2,4-diaminobutyric acid (L-DABA), nipecotic acid (NIP) (also OH-NIP), and guvacine, and it is not inhibited by beta -alanine (60, 103, 117, 273, 315). These are the pharmacological characteristics of the neuronal GABA transport (see references in Refs. 103, 184). In addition, biochemical evidence demonstrated that GAT-1 corresponds to the neuronal high-affinity subtype GABAA transporter, sensitive to ACHC (for review, see Ref. 260). Lipophilic derivatives of piperidencarboxylic acid [tiagabine, SKF-89976A, CI-966, and NNC-711; the latter with an inhibition constant (Ki) of 6 nM] are highly selective for GAT-1 transport activity (58, 102, 103). Interestingly, these inhibitors have anticonvulsant properties (103, 524).

The pharmacologies of GAT-2 and GAT-3 are similar to each other; GAT-3 shows higher sensitivity to NIP (see Table 5) and to a new triarylnipecotic acid derivative, 4(S) (128). The most characteristic BGT-1 inhibitor is betaine (see Table 5). Kilimann and co-workers (187) listed the pharmacological characteristics of the GABA transporters isolated from different species and expressed in different systems and pointed out the dissimilarities in the data reported (e.g., BGT-1 inhibition by L-DABA and betaine, TAUT inhibition by GABA, and interaction of beta -alanine with GAT-3; see Table 5). One of the most relevant dissimilarities is the role of beta -alanine in GAT-3 transport activity: an apparent Km of ~100 µM for mouse GAT-4 (i.e., GAT-3) (315), and no transport of beta -alanine up to 500 µM via rat GAT-B (i.e., GAT-3), but an apparent Ki of ~7 µM for beta -alanine inhibiting GABA transport (103). More recent studies (102) showed consistent interaction of GAT-3 permanently expressed in LLC-PK1 cells with beta -alanine (i.e., similar apparent Km and Ki values of ~30 µM). The reasons for these discrepancies are unknown. Interaction of beta -alanine with GAT-3 is consistent with its presence in glial cells, but not with its expression in neurons (61, 103, 216, 363, 450).

In summary, the pattern of expression in brain and the pharmacological characteristics of the four GABA transporters strongly indicate that GAT-1 corresponds to the most typical neuronal presynaptic GABA transporter, whereas GAT-2 and GAT-3 characteristics fit the glial transporter activity. In addition, the neuronal expression of GAT-3 and its brain distribution implies a complementary role of GAT-3 and GAT-1 in GABAergic synapses. The beta -alanine sensitivity of neuronal GAT-3, a glial characteristic (103), might be explained by this complementary expression of GAT-1 and GAT-3, and the higher expression of GAT-1 in neurons, as described in neuronal cell cultures (61).

C) SS-AMINO ACID TRANSPORTERS. Four transporters of this superfamily mediate beta -amino acid transport (GAT-2, GAT-3, and TAUT) or are inhibited by beta -alanine (BGT-1) (see Table 5). Murine, canine, rat, and human TAUT are highly homologous (~90% identity), have a wide transcript distribution, including kidney and small intestine, and express a very similar transport activity for the sulfur-containing beta -amino acids taurine and beta -alanine in oocytes (232, 316, 368, 445, 503, 562). Apart from the difference in the length between the mouse (590 amino acid residues; Ref. 316) and the human, rat, and canine proteins (620, 621, or 655 amino acid residues, respectively; Refs. 232, 368, 445, 562), because of different COOH termini, all these transporters could be considered as the counterparts of the same gene for these species. Expression of these transporters in oocytes results in sodium- and chloride-dependent high-affinity transport of taurine (Km, 5-12 µM) and beta -alanine (Km, ~50 µM); hypotaurine inhibits transport with a Ki that is probably in the micromolar range, whereas GABA, L-alanine, and MeAIB have no effect at 100 µM (see Table 5). Thus it seems that the taurine/beta -alanine transporter is a beta -amino acid-specific carrier. A specific system for beta -amino acids with high affinity for taurine (Km, 10-14 µM) and similar transport characteristics has been described in luminal plasma membranes from small intestine and renal proximal straight tubules; furthermore, in renal luminal membranes, other transport systems with lower affinity (Km values in micromolar and millimolar ranges) for taurine have been reported (231, 370, 496). A similar transport activity has been reconstituted from brush-border membranes of human placenta (444). Handler and co-workers (562) proposed that the cloned taurine carrier may correspond to the basolateral transporter described in MDCK cells, since the levels of the carrier mRNA increase after incubation in hypertonic medium. However, this issue is not clear, since the Km (~50 µM) reported for the basolateral carrier in MDCK cells is higher than that reported for the cloned carrier and for the apical membranes of MDCK cells (~10 µM) (563). Both GAT-2 and GAT-3 encode for high-affinity GABA transporters (see above). These two transporters also carry beta -alanine with high affinity (Km values are ~30 and ~100 µM, respectively) and taurine with a lower affinity (Km values are ~500 and ~1,500 µM, respectively) when expressed in oocytes. Homology between these two transporters and the taurine transporter is very high (i.e., amino acid sequence identity ranges from 62 to 66%; amino acid residues present in beta -amino acid interacting transporters: GAT-2, GAT-3, TAUT, and BGT-1 are idicated in Fig. 3). As for the taurine transporter, GAT-2 and GAT-3 transport activity is inhibited by micromolar concentrations of hypotaurine (562). Therefore, the substrate specificity of GAT-2 and GAT-3 transporters covers GABA and the beta -amino acids. As indicated above, GAT-2 and GAT-3 represent tissue/cell-specific isoforms with very similar transport activity: GAT-3 is neural tissue specific, whereas GAT-2 is highly expressed in kidney and liver, and in lower amounts in adult brain.


View larger version (146K):
[in this window]
[in a new window]
 

View larger version (95K):
[in this window]
[in a new window]
  		FIG. 3.   Amino acid sequence comparisons of sodium- and chloride-dependent amino acid transporters. Sequences multialigned correspond to human transporters but for rat GAT-2 and GLYT2. Horizontal lines over sequences indicate roughly position of theoretical (i.e., hydrophobicity plots) TM domains I-XII. Potential N-glycosylation sites are indicated by open boxes between TM domains III and IV. Gray boxes indicate amino acid residues present in all mammalian amino acid transporters (negative or positive charged residues are considered together) of this superfamily, in 4 GABA transporters (GAT-1, GAT-2, GAT-3, and BGT-1), in beta -alanine-interacting transporters (GAT-2, GAT-3, BGT-1, and TAUT), or in GABA, betaine, and taurine transporter subfamily. Finally specific amino acid residues for glycine and proline subfamily of transporters are shown in white inside black boxes. For all amino acid sequence identity shown here, all mammalian cDNA clones indicated in Table 4 are considered. Dash lines indicate gaps for sequence multialignment obtained with Clustal Multiple Sequence Alignment from Baylor College of Medicine.

BGT-1, with high homology (amino acid sequence identity ranges from ~60 to 70%) to the beta -amino acid transporters (i.e., TAUT, GAT-2, and GAT-3), encodes for a sodium- and chloride-dependent betaine (fully methylated glycine) and GABA transporter (see sect. IIB2B). Interestingly, beta -alanine interacts weakly with BGT-1 (i.e., 2 mM beta -alanine inhibits ~50% of transport). BGT-1 cDNA from mouse, dog, and human most probably represents species counterparts of the same transporter because of their homology (~90% amino acid sequence identity) and similar transport characteristics (see Table 5), in addition to their reported different tissue distribution in brain and peripheral tissues (see above). The murine and human BGT-1 are expressed in kidney, liver, and brain, but the canine transporter is mainly expressed in kidney medulla but not in liver or brain (59, 322, 446, 614). Most probably, the canine BGT-1 transporter represents the major hypertonicity-modulated transporter of the nonperturbing osmolyte betaine that operates in kidney medulla, normally the only hypertonic tissue in mammals (292). As discussed by Yamauchi et al. (614), luminal membranes from small intestine and renal proximal tubules transport betaine in a sodium-dependent manner, which is shared with high-affinity transport of proline. Interestingly, canine BGT-1 transporter expressed in oocytes has a low affinity for proline (Km in the millimolar range) (614).

In summary, transporters of this superfamily can be grouped according to substrate specificity. From strict GABA transporters to strict beta -amino acids (i.e., beta -alanine) transporters, the proteins could be ranked as follows (see Table 5): 1) GAT-1, high affinity and GABA specific; no interaction with beta -alanine; neural specific; 2) BGT-1, inhibition by beta -alanine at millimolar concentration; 3) GAT-2 and the neural-specific GAT-3, substrate affinity in the micromolar range for GABA and beta -alanine; and 4) TAUT transporter, substrate affinity in the micromolar range for beta -alanine and taurine and weak, if any, interaction with GABA. The concept of classical system BETA, defined in Ehrlich ascites cells with a low affinity for beta -alanine (Km in millimolar range) (96, 495), might correspond to the variety of different transporter isoforms described here and to carrier isoforms not yet identified. For example, in mammalian kidney, a complex interaction between reabsorption systems for beta -amino acids and GABA has been shown, including shared transport systems for amino acids and GABA as well as more GABA-specific carriers (496). TAUT, BGT-1, GAT-2, and GAT-3 transporters, which show overlapping specificities for beta -amino acids and GABA, are expressed in kidney. The challenge is now to relate the function of these transporters with the physiological fluxes of these amino acids in vivo, through the regional and cellular localization of these transporters and the "knockout" of the respective genes.

D) GLYCINE TRANSPORTERS. Two glycine transporter genes, GLYT1 and GLYT2, belong to this superfamily (see Table 5). GLYT1 presents three variants (1a, 1b, 1c) that are transcribed from a single gene (2, 62, 282). The three protein variants differ in their NH2-terminal sequences. Two promoters are responsible for 1a and 1b variants, and 1c isoform is an alternative splicing variant of 1b transcript with a 54-amino acid-long exon toward the NH2 terminus (2) (see Fig. 4B). The three variants show no differences in their transport characteristics; in fact, truncated proteins, constructed by elimination of the differential amino acid residues, retain the transport characteristics of the intact proteins, and only their cellular processing is affected (282). GLYT2 (~50% amino acid sequence identity with GLYT1 variants) most probably corresponds to the 100-kDa reconstituted and purified glycine transporter from pig brain (314, 320). GLYT2 is a larger protein than GLYT1 variants, due to a longer NH2 terminus (see Fig. 3). GLYT1 and GLYT2 transport activities can be distinguished by the higher sensitivity of GLYT1 variants to sarcosine (N-methylglycine) (see Table 5). Specific high-affinity (apparent Km values from 20 to 100 µM) transport systems for glycine have been identified in synaptosomes and glial cells (for review, see Ref. 622). The uptake process is electrogenic (inward positive charge flux) with a stoichiometry of 1 glycine, 2 sodium, and 1 chloride (16, 624); these coupling coefficients permit a glycine gradient through the plasma membrane that is strong enough to maintain an extracellular glycine concentration of 0.2 µM (23). The characteristics (substrate affinity, sodium and chloride dependence, and pharmacology) of the glycine transport via GLYT1 and GLYT2 in oocytes and mammalian cells are consistent with the neuronal and glial transport activity (622). Because of their brain distribution, GLYT2 most probably represents the neuronal glycine transporter, and GLYT1 the glial transporter, but the presence of GLYT1 transcripts in neurons (see sect. VB1) suggests its contribution to neuronal glycine uptake.


View larger version (61K):
[in this window]
[in a new window]
 

View larger version (70K):
[in this window]
[in a new window]
  		FIG. 4.   A. new 12-TM domain membrane topology model for rat GAT-1 transporter. This model has been recently proposed by Kanners and co-workers (38), and it is in full agreement with that proposed by Aragón, Giménez, and co-workers (401) for GLYT1 transporter (see Fig. 4B). Notice that hydrophobic domain I corresponds to a "reentrant" membrane loop. Hydrophobic external (EL) and internal (IL) loops are numbered. A 44-amino acid hydrophobic stretch, from Leu-118 to Trp-161, is proposed to contain TM domain III, IL2 loop, and new TM domain III*; sequence within TM domain III* is only tentative. Amino acid residues conserved in all mammalian amino acid transporters of this superfamily are indicated in gray circles, whereas only 6 specific amino acid residues for GABA transporters (GAT-1, -2, -3, and BGT-1) are indicated in white over a black circle. Amino acid residues within squares are those subjected to site-mutagenesis studies (see text for details). Amino acid sequences indicated by arrows connected with a line in EL4 and EL5 are corresponding ones cross-mutated between mouse GABA transporters and proposed to be involved in substrate specificity and affinity in Reference 542 (see text for details). Y, three putative N-glycosylation sites in EL2. Exon-intron boundaries of mouse GAT-1 transporter gene (318; corrected in Ref. 390) within putative rat protein sequence are indicated, and exons are numbered. Exon 1 is an untranslated sequence (318). In general, position of exon-intron boundaries is conserved between transporters of this superfamily (390) (see Fig. 4B for murine GLYT1 transporter gene).

B. new 12 TM domain membrane topology for mouse GLYT1a transporter. This model has been proposed by Aragón, Giménez, and co-workers (401) for rat GLYT1 transporter and is in full agreement with model proposed by Kanner and co-workers (38) for rat GAT-1. See Fig. 4A for description of model. Amino acid residues conserved in all mammalian amino acid transporters of this superfamily are indicated by gray circles, whereas those specific for glycine transporters are indicated in white over a black circle. N-glycosylation sites (Y) used for rat GLYT1 are indicated by a square (see text for details). Murine exon-intron boundaries within putative protein sequence are indicated, and exons are numbered. Inset: alignment of NH2 terminals of 3 mouse GLYT1 variants (1a, 1b, and 1c). At present, full-length murine exon 1c is unknown (2), but human one has been reported (282).

The transport characteristics of GLYT1 (sodium-dependent glycine transport inhibited by sarcosine, but not by MeAIB or beta -alanine) resemble that of system Gly (274, 572). Notwithstanding, the expressed transport activity of GLYT1 presents a much higher affinity for glycine than the classical transport system (62, 183, 282, 314, 319, 502) and is mainly found in neural tissues with very low levels of mRNA in liver or kidney (2, 62, 183, 282, 319, 502). Thus the encoded protein might correspond to a transporter isoform of the widespread system Gly. GLYT1 gene expression (1a variant; see sect. IIB1) in kidney, although low, might be responsible for the high-affinity (Km in the micromolar range) sodium-dependent glycine uptake component in luminal membranes from proximal straight tubules (reviewed in Ref. 483). This can now be tested by immunolocalization of the GLYT1 in kidney. To this end, it is important to assess whether GLYT1 in kidney is expressed in the epithelial cells or in macrophages, as has been reported for lung, spleen, and liver (62).

E) PROLINE TRANSPORTER. The proline transporters (PROT) isolated from a rat and human brain libraries show ~45% identity to GLYT transporters at the amino acid sequence level (155, 489). The protein is encoded by a 4-kb mRNA that is present in the excitatory pathways of rat brain. This pattern of expression suggests that the encoded protein does not represent an ubiquitous transporter that might have a general metabolic role but rather supports a specific role for L-proline or this transporter in excitatory amino acid neurotransmission (see sect. IIB6). The transporter expresses sodium- and chloride-dependent uptake of L-proline with very high affinity (Km is ~10 µM) in HeLa cells (155, 489). The expressed sodium-dependent uptake of proline is sensitive to sarcosine, norleucine, phenylalanine, histidine, and cysteine, with Ki ranging from 30 to 90 µM. Therefore, in common with GLYT transporters, PROT transporters tolerate N-methyl derivatives, since both show high-affinity interaction with sarcosine (N-methylglycine). This high-affinity L-proline transport is unique to nervous tissue; in contrast, a lower affinity, sodium-dependent L-proline transport has also been described in neural tissues and renal, intestinal, and choroid plexus brush-border membrane vesicles through various systems (e.g., IMINO, Bo,A) (reviewed in Refs. 483, 513, 505).

3. Mechanisms of transport and uncoupled ion fluxes

The transport characteristics of the first member of this superfamily of transporters to be discovered, GAT-1, have been studied more extensively than the other members (see Table 5). As discussed above, GAT-1 corresponds to the neuronal high-affinity ACHC-sensitive GABA transporter present in synaptosomes and synaptic plasma membranes (260, 273). The transport of GABA by GAT-1 in different expression systems, including permanent expression in mammalian cells, showed the following characteristics (for references, see Table 4): 1) a single high-affinity (Km of ~4 µM) component of transport, in full agreement with values obtained in synaptic plasma membranes (250) and reconstituted systems (439); 2) absolute requirement of sodium, but not of chloride; capacitive charge movement and transport-associated current studies indicate that chloride facilitates the binding of sodium and that this limits the overall transport rate at saturating GABA concentrations (339, 325); and 3) exchange mechanism of transport (i.e., efflux of GABA is trans-stimulated by external GABA). This represents a part of the translocation cycle of the transporter (257, 393). Stable expression in mammalian cells revealed unique properties for GAT-3 (102), a Km values for chloride of 78 mM, which is severalfold higher than that for GAT-1, and the Km for GABA depends on the chloride concentration (i.e., a decrease in chloride concentration results in an increase in the apparent Km for GABA).

Stoichiometry of the coupled ions and GABA via cloned GAT-1 was first deduced from the corresponding Hill equations. GAT-1 permanently expressed in mouse LtK- cells (273) shows a sodium concentration dependence with sigmoidal behavior, whereas the dependence of GABA and chloride concentrations was hyperbolic. This allowed a proposed stoichiometry of cotransport of 1 GABA, > 1 (probably 2) sodium, 1 chloride, in accordance with data obtained previously with synaptic plasma membrane vesicles and the purified transporter (257, 272, 440). Thus direct measurements of 22Na+, 36Cl-, and [3H]GABA fluxes using proteoliposomes into which a partial purified preparation of the neuronal high-affinity ACHC-sensitive GABA transporter (most probably GAT-1; see above) was reconstituted yielded the following stoichiometry (272): 1 GABA >=  2 sodium, 1 chloride. This stoichiometry suggests that GABA transport through GAT-1 is electrogenic (i.e., external GABA induces positive inward current). Expression of rat GAT-1 in oocytes resulted in electrogenic uptake of GABA, with an apparent affinity constant of ~5 µM for GABA and Hill coefficients of 0.7 for chloride and 1.7 for sodium (265). Concentration dependence studies of capacitive charge movements support the interaction of two sodium with GAT-1 (339). Surprisingly, correlation of induced currents and radiolabeled GABA uptake gave a ratio of one positive charge flux per molecule of GABA transported; this indicates a stoichiometry of 1 GABA, 2 sodium, and 1 chloride (265, 341). Similarly, kinetics of GABA uptake induced by rat GAT-1 transfected in HeLa cells gave Hill coefficients compatible with this stoichiometry (451). This fits with the proposed stoichiometry for the rat GAT-3, canine BGT-1, and human dopamine transporter DAT (507, 102), but not with other transporters of this superfamily; a stoichiometry of 1 serotonin, 1 sodium, and 1 chloride for the rat serotonin transporter SERT (180) and 1 norepinephrine, 1 sodium, and 1 chloride for the human transporter NET (161, 181). Therefore, as discussed below for the sodium- and potassium-dependent transporters of glutamate and zwitterionic amino acids, the stoichiometry of the coupled ions does not appear to be conserved between the members of the superfamily of sodium- and chloride-dependent neurotransmitter transporters.

The transport stoichiometry of the neurotransmitters and co-ions just described leads to electrogenic transport. However, the conducting properties of these transporters go well beyond this (301, 507). The expression of these transporters displays several conducting states that resemble single-channel openings and that occur both in the presence of substrate (substrate-gated channel activity) and in its absence (leakage) (301, 507). Uncoupling between transport flux and substrate-evoked current in steady state has been observed for rat SERT, human DAT, human NET, and rat GAT-1 (160, 161, 340, 451). Several recent reviews address this issue (301, 507, 609). This is similar to the uncoupled chloride channel mode of action of several transporters of the superfamily of sodium- and chloride-dependent glutamate and zwitterionic amino acid transporters (see sect. IIC), but the charge-carrying ions responsible for these uncoupled currents have not been identified (507). Comparisons of the transport flux and the currents associated with the expression of the biogenic amine transporters (SERT, NET, and DAT) and the GAT-1 GABA transporter revealed ratios greater than the proposed stoichiometries (discrepancies ranged from a factor of 3 to >100), but the apparent Hill coefficients for sodium and chloride for the activation of the substrate-induced currents is consistent with the uptake studies and with the proposed stoichiometry (for review, see Refs. 301, 507). Thus, for rat SERT expressed in oocytes, a ratio of charge flux to serotonin molecule of 5-12 positive charge flux was found, whereas the proposed transport stoichiometry (1 serotonin, 1 sodium, 1 chloride) predicted a ratio of 1 positive charge flux/serotonin molecule (340). Electrophysiological studies with rat GAT-1 permanently transfected in HEK 293 cells showed that the stoichiometry between GABA and co-ions flux was not fixed (e.g., outward and inward currents require different ions on each side of the membrane) (81). DeFelice and co-workers (451) reported that the sodium-dependent GABA-induced currents due to GAT-1 expression in HeLa cells do not show saturation of the current-voltage curve, as fixed stoichiometry would predict, but it shows saturation of the current-external GABA concentration curve at a fixed voltage; these data suggest that saturable binding of external GABA gates a channel through the GAT-1 transporter or an associated channel. Mager et al. (341) estimated that GAT-1 is expressed in oocytes at a plasma membrane density of 104 transporters/µm2, and with a transport turnover rate of ~10 s-1 (at 300 µM GABA, 96 mM NaCl, -80 mV and 22°C). To explain the maximal currents due to the expression of GAT-1 in HeLa cells on the basis of the 1 GABA, 2 sodium, 1 chloride-coupled model, either the density of transporters at the cell surface or the turnover rate is >10-fold higher than the values estimated in oocytes (451). In summary, all this suggests substrate-activated conduction, which is not thermodynamically coupled to transport. It is at present unknown, as for the members of the superfamily of sodium- and potassium-dependent glutamate and zwitterionic amino acids (see sect. IIC), whether the substrate-gated channel occurs through the transporter itself or via interaction with an extrinsic channel protein. Fluctuation analysis indicates that the substrate-gated channel current events (-0.2 to -0.5 pA at -80 to -100 mV), associated with the expression of human NET and rat serotonin transporter (SERT) in HEK 293 cells and oocytes, respectively, are brief (~1 ms) and have an extremely low open probability (10-3 to 10-6); interestingly, the probability of opening increases with substrate concentration (160, 507). This is interpreted, at least for several members of the superfamily of neurotransmitter transporters (SERT, NET, DAT, and GAT-1), as indicating that these transporters show brief channel openings whose probabilities depend on the concentration of the substrate and co-ions (e.g., 1 opening/350-700 transport cycles; Ref. 301). The channel openings would account for the macroscopic current that exceeds the charge flux expected from transport flux and substrate/co-ion stoichiometry (301, 507).

Transporters of this superfamily produce cation-permeable channel activity in the absence of substrate. This channel mode of action is due to the expression of the corresponding transporter, since it is absent in the nontransfected cells or in uninjected oocytes and shows the same pharmacological sensitivity as the transporters (507). Constitutive transporter leak currents (i.e., flux of driving ions down their electrochemical gradient in the absence of substrate) have been detected for the expressed human NET, rat SERT, human DAT, and rat GAT-1 (81, 82, 161, 340, 520). Similarly, other sodium-dependent cotransporters show leak currents, like the glutamate transporters EAAT1 and EAAT3 (see sect. IIC), and the sodium/glucose cotransporter SGLT1 (reviewed in Ref. 609). Interestingly, for many of these transporters (rat GAT-1, rat SERT, and human DAT), the ion selectivity of the leak conductance is similar but not identical to the cotransporter activity (alkali metal cations like sodium, potassium, or lithium produce the leak conductance, whereas potassium and lithium do not couple the transport of the corresponding substrates; reviewed in Ref. 507). This raises the question as to whether the leak conductance and the translocation of the co-ions use the same permeation pathway. In any case, the leak pathway depends on the presence of substrate: application of substrate inhibits this conductance associated with the expressed rat GAT-1, rat SERT, and human DAT (82, 340, 507). Patch-clamp studies by Cammack and Schwartz (82) in HEK 293 cells, which permanently express rat GAT-1, suggest that there are two distinct subpopulations of GAT-1 transporters: 1) the vast majority of the transporters act as normal c-transporters that yield only small currents, and 2) a small "channel-like" population produces leak conductance in bursts (~1 pA at -50 mV) and is sensitive to the GABA GAT-1-specific inhibitor SKF-89976A. For these authors, the large discrepancy between the number of channels and transporters indicates that they function independently. The question is then, What mechanisms determine whether a protein functions as a channel or as a transporter (82)? Posttranslational modifications, the state of aggregation or interation with other proteins, like cytoskeletal elements, or even silent channels, might be the molecular basis for the channel behavior. Purification and reconstitution of these transporters are clearly needed to demonstrate that they could work intrinsically as channels.

4. Protein structure

Figure 3 shows the alignment of the human amino acid transporters of this superamily, with the exception of rat GAT-2 and rat GLYT2. They show high homology (at least 40%) in their primary structures, with ~150 well-conserved amino acid residues. As general features (see references in Table 4), all these transporters lack a signal peptide, a good prognosis for 12 TM domains, with the NH2 and COOH termini located intracellularly, where most putative phosphorylation sites are located, and one to four putative N-glycosylation sites between the theoretical TM domains III and IV. The regions of highest homology are TM domains I, II, and IV-VIII; in contrast, the lowest level of homology occurs in the NH2 and COOH termini (see Fig. 3 and Ref. 12 for review). Regions of extensive homology may be related to general functions for these transporters. In general, there is good conservation of proline, glycine, tyrosine, and tryptophan residues in TM domains (see Fig. 3). Thus every theoretical TM domain has at least one conserved glycine residue; proline residues are conserved in TM domains I, II, V-VIII, XI, and XII; and tyrosine residues are conserved in TM domains I-III, V, and X. Transmembrane domains I and IV present one conserved positive amino acid residue. The structure-function relationship of some of these conserved structural features has been addressed experimentally (see sect. IIB5).

These transporters are N-glycoproteins. The GABA transporter GAT-1 was reconstituted and purified to homogeneity from rat brain (441, 439) and later cloned (184). The transporter was resolved in SDS-PAGE as a polypeptide band of 80 kDa that also dimerizes to an apparent molecular mass of 160 kDa, as revealed by cross-reactivity with polyclonal antibodies (439). The deglycosylated GABA transporter has an electrophoretic mobility compatible with a molecular mass of ~67 kDa, deduced from the cloned GAT-1 transporter (184, 259). Other studies have also demonstrated that GAT-1 is N-glycosylated (37, 273, 410, 439). Similarly, N-glycosylation has been demonstrated for DAT, SERT, NET, and GLYT1 transporters (70, 307, 357, 358, 400, 468, 415, 547637).

Purification to apparent homogeneity of a pig brain glycine transporter showed it to be a glycoprotein (it binds to lectins) with the appearance of a broad band with an average size of ~100 kDa in SDS-PAGE (320). Physicochemical characterization of the native glycine transporter indicated a monomeric structure of this size (321). Treatment with peptide N-glycosidase F (PNGase F), but not with endoglycosidase F or O-glycanase, produced a dramatic electrophoretic mobility change, which shows that ~30% of the transporter mass corresponds to the carbohydrate moiety; neuraminidase produces a slight reduction of its apparent mass (395). The size of the deglycosylated and glycosylated glycine transporter purified from pig brain suggests that it most probably corresponds to the GLYT2 transporter (314, 395). Interaction with succinylated wheat germ agglutinin lectin, but not with concanavalin A-lectin, and the previous glycosidase treatments indicate a tri- to tetra-antenary complex structure with terminal sialyc acid residues for the carbohydrate moiety of the purified GLYT2 transporter (395). With a similar approach, the carbohydrate moiety of the human DAT transporter was shown to lack high-mannose residues (306).

As indicated before, the cloning of the first two members of this superfamily, GAT-1 and NET, suggested a common 12 TM domain model that was extended to the other members of the family (see Fig. 3). This model initially received some experimental confirmation. Site-directed mutagenesis showed that the glycine transporter GLYT1 is heavily glycosylated at four Asn residues within the loop EL2 (see Figs. 3 and 4B), demonstrating the extracellular location of this hydrophilic loop (400, 401). N-glycosylation within this loop has also been demonstrated for SERT (547), NET (70, 358), and GAT-1 (38) transporters. Immunofluorescence and electron microscopic studies in permeabilized and nonpermeabilized cells expressing GLYT1 and NET (623, 16, 70), and N-glycosylation scanning mutagenesis for GAT-1 and GLYT1 (38, 401) confirmed the intracellular location of the NH2 and COOH terminals of these transporters. Similar evidence obtained from experiments using antibodies located the hydrophilic loop of human NET connecting TM domains VIII and IX intracellullarly (357), and those connecting TM domains III and IV and VII and VIII (loops EL2 and EL4 in Fig. 4, A and B) extracellularly (70). N-glycosylation scanning mutagenesis unambiguously showed the extracellular location of loops EL3 and EL6 (see Fig. 4A) of GAT-1 (38) and of loop EL3 of rat GLYT1 (401). N-glycosylation of loops EL4 and EL5 (see Fig. 4B) has been shown for rat GLYT1, but with an impaired transport function (401). For rat GLYT1, all this has been confirmed by in vitro glycosylation reporter fusion; TM domains VI, VIII, and X act as stop transfer signals for the preceeding hydrophobic TM domains V, VII, and IX, which act as membrane anchor signals (401). The extracellular location of the loop EL6 of rat GLYT1 is supported by similar experiments showing that TM domain XII acts as a stop transfer signal when placed after TM domain XI and the loop EL6 (401). In summary, the theoretical model of 12 TM domains between the extracellular loop EL2 and the COOH terminus is supported by strong experimental evidence.

Very recently, two back-to-back reports examined in depth the membrane topology of GAT-1 (38) and GLYT1 (401) and by different strategies reached the same conclusion: the highly conserved theoretical TM domain I is probably immersed in the plasma membrane, but does not span it completely. In addition, a new TM domain III* is proposed. This new 12 TM domain model is depicted in Figure 4, A and B. Bennet and Kanner (38) show that the previous intracellular loop connecting TM domains II and III (i.e., new loop EL1, see Fig. 4A) can be glycosylated in vivo, as well as loop EL2, which contains the naturally occurring N-glycosylation sites, whereas loop IL2 is not glycosylated; therefore, two TM domains should be placed in between the TM domains III and III* shown in Figure 4A. There is a hydrophobic segment of 44 amino acid residues within these transporters that could accommodate both TM domains (see Fig. 4A). In addition, studies with permeant and impermeant methanesulfonate reagents indicate that the cysteine residue 74 (see Fig. 4A) should be located intracellularly, as proposed by the new model, and not extracellularly, as proposed by the previous theoretical model; consistent with this, the hydrophilic loop containing cysteine-74 (i.e., the previous EL1) is not glycosylated (38). Aragón and Giménez and co-workers (401) also used N-glycosylation scanning mutagenesis to show that the loop connecting the TM domains II and III can be glycosylated in vivo, and must thus be extracellular. These results suggest the model proposed in Figure 4B for rat GLYT1, where the TM domain III* should contain a stop transfer signal. Unfortunately, this has not been demonstrated in studies of fusion reporter glycosylation (401). This suggests that the proposed TM domains III and III* constitute a single "reentrant loop" as proposed for the first highly conserved hydrophobic domain. In this sense, it is not easy to conceive two TM domains within the hydrophobic segment of 44 amino acid residues as proposed by Bennet and Kanner (38) (see Fig. 4A). Indeed, no clear support for the intracellular location of the hydrophilic loop IL2 has been offered; in fact, the construct of glycosylation reporter fusion of GLYT1 spanning the NH2 terminus to the TM domain III is glycosylated in vitro (401).

At present, the only evidence for the membrane interaction of the reentrant hydrophobic loop I (see Fig. 4A) is that alkaline stripping does not release from microsomes a GLYT1 construct having only the COOH terminus, the reentrant domain I, and a small part of the new IL1 loop (401). Reentrant hydrophobic loops (or "pore loops") have been described for receptors and channels (39, 215, 362) and are involved in ion permeation of voltage- or ligand-dependent ion channels (for review, see Ref. 333). This might help us to understand the channel mode of action described for several transporters within this superfamily. In addition, several amino acid residues that are critical for the transport function of the transporters of this superfamily have been described within this hydrophobic domain (see sect. IIB5). It is therefore important to clarify, by different strategies, the membrane topology of these transporters within their NH2-terminal third.

5. Structure-function relationship

Site-directed mutagenesis with rat GAT-1 and GLYT1 transporters by Kanner's, and Aragón's and Giménez's groups (16, 37, 50, 271, 288, 331, 410), a cross-mutation study between the external loops of mouse GAT-1, -2, and -3 and BGT-1 by Nelson and co-workers (542), examination of the role of the carbohydrate moiety for the glycine transporters by Aragon and co-workers (395, 400), and a substrate-protected proteolytic cleavage study of rat GAT-1 by Mabjeesh and Kanner (332) are the bases for our present knowledge of the structure-function relationship of the sodium- and chloride-dependent amino acid transporters. These data are discussed here in the light of the new membrane topology model described for GAT-1 and GLYT1 transporters (see sect. IIA). These studies show that the NH2-terminal third of these proteins may be involved in common transport functions, whereas the external hydrophilic loops affect substrate discrimination.

In a series of studies, Kanner's group (37, 256, 331) identified domains of rat GAT-1 that are not required for its transport function. Papain- or pronase-treated purified rat GABA transporter, which has probably lost the NH2 and COOH termini, nevertheless retains all the transport characteristics of the intact GAT-1 in a reconstituted system (259, 331). In a parallel study, deletion mutants of rat GAT-1, where most of the hydrophilic NH2 and COOH terminals have been eliminated, also retain all the transport characteristics of the intact transporter when expressed in HeLa cells (37). These results demonstrate that neither terminus of the GABA transporter is needed for its transport function. These protein domains might be relevant for other functions, like regulation of the transport function, or interaction with cystoskeleton, as described for other transporters (e.g., erythrocyte anion transporter, sodium/proton exchanger; Refs. 290, 472). These results cannot be generalized to other transporters of this superfamily. In contrast to GAT-1, the COOH terminus of rat GLYT1 is necessary for the correct trafficking of the protein to the plasma membrane (399). Deletion of the first 30 amino acid residues of GLYT1a (i.e., NH2 terminus) does not alter the transport of glycine by the transporter when it is expressed in COS cells. Deletion of the last 34 amino acid residues (i.e., COOH terminus; see Fig. 4B) does not impair the transport function of GLYT1a, but longer deletions within the COOH terminus result in a progressive decrease of its transport expression. Immunofluorescence and transport reconstitution studies showed that this impaired transport expression was due to both a defect in trafficking to the plasma membrane and a loss of the intrinsic transport activity of these mutants (399).

Two studies analyzed the role of the hydrophilic loops in the function of the GABA transporters (542, 256). Kanner et al. (256) examined the effect of deletions of randomly picked nonconserved segments, or even deletion of single residues, within the EL4 and IL5 loops of rat GAT-1 (single residues deleted are indicated by a square in Fig. 4A). All deletions, of 1-22 residues, produce a loss of transport activity. This could not be explained by a low level of synthesis (i.e., normal [35S]methionine labeling) or by a missorting to the plasma membrane (i.e., there is no rescue of transport activity when the transporter is solubilized and reconstituted from HeLa cells expressing the mutants); only the deletion mutant of valine-348 shows a missorting defect, which is partially responsible for the loss of its transport activity. Interestingly, conserved and nonconserved (mutations to glycine) substitutions of these nonconserved residues within the superfamily of transporters in rat GAT-1 retain significant transport activity when expressed in HeLa cells (256). These results suggest that a minimal length of the hydrophilic loops EL4 and IL5 is important for the transporter functional structure (256). In agreement with this, insertion of an Asn residue between valine-348 and threonine-349 ensures significant transport activity (256).

As discussed above (transport characteristics section), GAT-1 transports GABA, whereas GAT-2 and GAT-3 and BGT-1 transport beta -alanine in addition to GABA; all four GABA transporters show different apparent affinity for GABA (see Table 5), and there are specific amino acid residues for the beta -alanine transporters (i.e., GAT-2, GAT-3, BGT-1, and TAUT) between TM domains VII and VIII, IX and X, and XI and XII (see Fig. 3). Nelson and co-workers (542) examined the role of the external loops EL3 to EL6 (see Fig. 4A) in substrate selectivity by cross-mutation between the four GABA transporters (GAT-1 to -3 and BGT-1) (i.e., swapping amino acid residues between these transporters) and expressing the mutants in oocytes (542). The most relevant results in this study are as follows.

1) Expression of a mutant produced by swapping amino acid residues of the EL4 loop of GAT-3 (also very similar to GAT-2 sequence) into GAT-1 sequence resulted in an apparent affinity change for GABA that mimics GAT-3 transport (Km values were ~9, ~1, and ~2 µM for mouse GAT-1, GAT-3, and the cross-mutant GAT-1/GAT-3-EL4 loop/GAT-1, respectively).

2) The cross-mutant GAT-1/BGT-1-EL6/GAT-1 has increased affinity for GABA (Km ~35 µM) that mimics BGT-1 transport (Km <100 µM).

3) The cross-mutant GAT-1/GAT-2-EL5/GAT-1 acquired similar beta -alanine sensitivity to that of GAT-2, which is not present in GAT-1. Cross-mutations within the other loops do not confer beta -alanine sensitivity, and swapping amino acid residues of the EL5 loop of GAT-1 into the GAT-2 sequence reduced beta -alanine sensitivity.

4) Cross-mutations within the EL3 loop produce no significant changes.

Therefore, this elegant study strongly suggests that the three external loops EL4-6 (see Fig. 4A) might form a pocket on the transporters into which the substrates bind (542). Interestingly, a similar conclusion was reached in studies with chimeric biogenic monoamine transporters (71, 169). More recent studies (72) also showed that the protein regions spanning TM domains V-VII and I-III of NTE and DAT are involved in the selective inhibition by antidepressants and psychomotor stimulants.

Four studies by Kanner's group look for critical amino acid residues for the transport activity of rat GAT-1, within charged residues (410) or conserved tryptophan or tyrosine residues located in the theoretical TM domains (50, 288), or conserved negatively charged residues adjacent to putative TM domains (271) (see Fig. 4A). Interestingly, the three studies identified amino acid residues that may be critical for the transport function within the first NH2-terminal part of the protein (see Fig. 4A), the membrane topology of which has recently been revised (see sect. IIB4). Of the five charged residues within theoretical TM domains, only Arg-69, located in the highly conserved reentrant loop (38, 401), is critical for the radiolabeled transport of GABA; the arginine residue itself is critical and not merely the associated positive charge (410). The loss of transport function when the Arg-69 mutants are expressed in HeLa cells is due to intrinsic transport defects and not to protein synthesis or missorting effects (410). The role of Arg-69 is unknown, but its positive charge and the fact that it is conserved in all the transporters of this superfamily prompted the authors to suggest that it might be involved in the binding of chloride (410).

The lack of negatively charged residues conserved in the theoretical TM domains of rat GAT-1 that would be critical for the transport function, and therefore candidates for the binding of sodium (410) fostered the study of the role of tryptophan residues that might interact with their P-electrons with this co-ion within the putative TM domains (288). Of the 10 conserved tryptophan residues within putative TM domains, only Trp-68, Trp-222 and Trp-230 resulted in impaired transport function when mutated to serine or leucine residues; Trp-230 mutants showed missorting, whereas mutations in the other two tryptophan residues had intrinsic transport activity defects (288). Interestingly, Trp-68, contiguous to the critical residue Arg-69, is conserved in all the transporters of this superfamily, and the GAT-1 transport activity tolerates substitutions of Trp-68 to other aromatic residues. This suggested that these two contiguous residues might interact with the transport co-ions sodium and chloride (288). Moreover, capacitive charge movement studies in oocytes suggest that the Trp68Leu mutant "locks" sodium onto the transporter by preventing (or slowing) the intracellular release of the substrate (339). As in all types of site-directed mutagenesis study, it is also possible that both residues influence the transport function by affecting the transporter structure.

The study of negatively charged residues adjacent to TM domains identified Glu-101, within the newly proposed loop EL1 (see Fig. 4A), as being critical for transport function (271). Replacement of Glu-101 by Asp reduced the transport activity of GAT-1 by 99%. Substitutions of Glu-101 do not show the sodium transient currents, which are indicative of conformational changes during the transport cycle. This has led to the proposal that this residue is critical for the transport cycle-associated conformational changes of GAT-1 transporter (271).

Several lines of evidence highlight the relevance of the reentrant hydrophobic loop I and the adjacent protein regions for the transporters of this superfamily (i.e., it might be close to the permeation pore):

1) This domain shows high conservation within the transporters of this superfamily and contains the critical Trp-68 and Arg-69 residues.

2) Deletion mutations of the NH2 terminus of rat GAT-1 and GLYT1 affecting amino acid residues adjacent to this domain abolish transport function (16, 37).

3) The cysteine residue 74, located in the adjacent IL1 loop (see Fig. 4A), is partially responsible for the inactivation of rat GAT-1 by (2-aminoethyl)methanethiosulfonate (38).

4) N-glycosylation scanning mutagenesis within the extracellular loop EL1 (see Fig. 4A) resulted in almost complete loss of transport function.

5) The glutamate residue 101, conserved in all the transporters of this superfamily and located in this EL1 loop, has been found to be essential for the transport function of GAT-1 (271).

6) Moreover, it is noticeable that the biogenic amine transporters have a specific aspartate residue within this domain (Asp-79 in the human DAT sequence; the amino acid transporters of this superfamily have a conserved glycine residue in this position; Gly-63 in rat GAT-1; see Fig. 4A), which is critical for the binding of the amine group by human DAT (286).

In section IIB4, the presence of reentrant hydrophobic domains in channels was discussed. It is thus important to determine whether mutations within these domains (e.g., Arg-69, Trp-68) affect the channel mode of action of these transporters. In other words, do Arg-69, Trp-68, or adjacent domain (e.g., Glu-101) mutants show the characteristic channel mode of activity of rat GAT-1? If, finally, transporters of this family have intrinsic channel-like activity, strategies developed for the study of the conducting pathways of channels will be useful. As discussed by Lester et al. (301), substitution mutants of critical residues in or near the permeation pathway should affect single-channel conductance, open-channel blockade, or ion selectivity. Only upon purification and reconstitution will it be possible to demonstrate the intrinsic channel mode of action of these transporters. Until then, the residues that are critical for transport and channel activities could be interpreted as being involved in the substrate-transporter conformational changes needed to activate an extrinsic channel activity. The study of mutants able to dissociate between transport and channel functions may also reveal the mechanism of the channel mode of action of these transporters. Lester et al. (301) proposed a variation of the "alternating access model" to explain the putative intrinsic channel activity of these transporters: the transporter is alternatively gated at the intracellular and the extracellular face during a transport cycle, but with low probability, whereas during its channel activity events both gates remain open, allowing a continuous aqueous phase through the transporter. To explain why the ion selectivity does not always coincide between the transport and the channel activities (either for the present transporters, see above and Ref. 507, or for the chloride conductance of the sodium, potassium-glutamate transporters; see sect. IIC) the substrates and co-ions themselves might become part of the ion-selective pathway, as proposed for the glutamate transporters (582).

The critical residues Trp-222 and Trp-230 are located within TM domain IV (see Fig. 4A). Tryptophan-222 is conserved in the amino acid transporters, but not among the monoamine transporters of this family (see Fig. 3 and Ref. 253). This fosters the hypothesis that Trp-222 might be involved in the binding of the amino group of these substrates (288). Further support for this role of Trp-222 has been obtained by substitution mutants of Trp-222 that do not bind radiolabeled tiagabine, a NIP derivative inhibitor of GAT-1 that is not transported (253). As discussed above, the biogenic amine transporters have a specific aspartate residue within the reentrant hydrophobic loop that seems to mediate the binding of the amine substrates (286). Of the 12 conserved tyrosine residues predicted to be located in the membrane according to the theoretical topological model (see Fig. 3), Tyr-140 (located in TM domain III or in the intracellular loop IL2, depending on the topology model; see Figs. 3 and 4A) is the only one that does not tolerate, in rat GAT-1, replacement by either phenylalanine or tryptophan (50). A detailed study on substitution mutants of residue Tyr-140 of rat GAT-1 expressed in HeLa cells or oocytes demonstrated that this residue is a specific determinant on GABA binding (i.e., sodium and chloride binding are unimpaired) (50). Interestingly, this tyrosine residue is conserved throughout this superfamily of transporters (including the transporters for biogenic amines). This allowed the authors (50) to suggest that Tyr-140 may be involved in the binding of the amino group, the moiety which is common to all the substrates of this superfamily of transporters. Until the membrane topology of the transporters of this superfamily is resolved (see sect. IIB4), it is impossible to know which other group might be located close to Tyr-140 (50), but the results discussed here (50, 286, 288) suggest that this residue and Trp-222 (rat GAT-1 sequence) or Asp-79 (human DAT sequence) are determinants of the binding of the amino group of GABA or biogenic amines, respectively. This suggests proximity for these residues within the three-dimensional structure of these transporters. In any case, all these results confirm the relevance of the NH2-terminal third of these transporters for their transport function.

The N-glycosylation of the transporters of this superfamily, with the possible exception of the "orphan transporters," occurs within the external loop EL2 (see Figs. 3 and 4, A and B). Aragón and co-workers (395, 400) analyzed the role of the carbohydrate component of the pig brain purified glycine transporter and the expressed rat GLYT1 glycine transporter. Treatment of the purified and reconstituted pig brain glycine transporter (GLYT2) with PNGase F results in the loss of ~30% of its apparent mass and a substantial decrease in its transport activity (395). These results do not reveal whether the carbohydrate moiety is important for substrate binding or for stabilizing the active conformation of the transporter. This study cannot be extended to other transporters of this family. Indeed, N-glycosidase F treatment of the purified and reconstituted GLYT1 glycine transporter expressed in COS cells results in the loss of its carbohydrate moiety without loss of its transport function; N-glycosylation does not appear to be essential for the transport activity of GLYT1 (400). On the other hand, the carbohydrate moiety of GLYT1 is indeed necessary for its proper trafficking to the plasma membrane. Progressive disruption of the four naturally occurring N-glycosylation sites of GLYT1 in loop EL2 (see Fig. 4B) results in a progressive decrease in the transport activity when the mutants are expressed in COS cells; surface biotinylation and immunofluorescence analysis demonstrated missorting to the plasma membrane of the unglycosylated GLYT1 mutants (400). This is consistent with the behavior of nonglycosylated NET mutants, which upon expression in COS cells had reduced protein stability and surface trafficking, which could explain the loss of transport expression. Interestingly, the residual transport activity conserves substrate and antagonist recognition (358), and loop EL2 of SERT (and probably NET) transporter does not appear to participate in substrate or antidepressant binding. Swapping half of this loop of NET into SERT transporter does not alter substrate or drug affinity but slows the transport rate; this suggests that loop EL2 may be involved in the translocation mechanism (511).

The last item to be discussed on the structure-function relationship of these transporters is the evidence in favor of transport cycle-associated conformational changes. An elegant study by Mabjeesh and Kanner (332) provides experimental evidence of this for rat GAT-1 transporter. The GABA transporter from brain membranes or purified and reconstituted into liposomes is cleaved by proteases (pronase and trypsin), and its transport activity is abolished. The presence of GABA, together with the transport co-ions (sodium and chloride) on the same side of the membrane, almost entirely blocked the action of the proteases, both in the proteolysis of the GAT-1 transporter and in its transport function (332). The concentration of GABA necessary for the half-maximal protection, the partial protection by the GAT-1-specific inhibitor ACHC, and the lack of protection by beta -alanine (noninteracting with GAT-1 transporter) (see Table 5) in addition suggest that the GAT-1 transporter undergoes conformational changes after binding of its substrate and co-ions. Mabjeesh and Kanner (332) propose that the formation of the translocation complex (GABA plus co-ions) induces conformational changes that render the transporter resistant to digestion by these proteases. As for other substrate-protection strategies, it is not possible to distinguish between protection due to steric hindrance by the substrate or conformational changes hiding protein domains. This discussion is relevant, since Lester and co-workers (520) recently developed a new model of ion-coupled cotransport for GAT-1, SERT, and the sodium-glucose cotransporter (SGLT1) based on substrate-substrate interactions that does not require the global conformational changes characteristic of the paradigmatic alternating access model. In any case, the study of the capacitive properties of the GAT-1 transporter also provides some clues to the conformational changes associated with the binding of sodium and GABA (see Ref. 301 for review). The slow component of capacitive currents is associated with sodium binding, and it is compatible both with sodium ions entering the membrane dielectric field and with reorientation of dipoles after sodium binding (i.e., conformational changes) (301). The fast capacitive currents due to stable expression of GAT-1 in HEK 293 cells show an increased current noise in the membrane patch after addition of GABA that increases with frequency (82), suggesting that GABA binding facilitates sodium binding by eliminating barriers to the entry of ions from solution (301). Similarly, jumps in the sodium concentration produce charge movements that reflect the same population of charges that move during voltage-jump relaxations (339). All this suggests that the binding of GABA and sodium induces conformational changes in the transporter (301).

6. Physiological role

A) TRANSPORTERS FOR NEUROTRANSMITTER AMINO ACIDS. Sodium-dependent transporters located in the presynaptic and glial membranes remove transmitter molecules (e.g., GABA, glycine, biogenic amines, and glutamate) from the synaptic cleft and thus terminate transmission. This has been directly demonstrated for the dopamine transporter in the DAT knockout mouse; in the homozygotic mice, dopamine persists ~100 times longer in the extracellular space (168). gamma -Aminobutyric acid is the major CNS inhibitory neurotransmitter, and glycine has this major role in spinal cord and brain stem. As discussed above, three high-affinity GABA transporters are expressed in the CNS, GAT-1, -2, and -3. Biochemical, pharmacological, and immunolocalization studies demonstrated that GAT-1 corresponds to the neuronal subtype GABAA transporter (see Ref. 260 for review). Therefore, GAT-1 should be resposible for the presynaptic removal of GABA for terminating transmission in the GABAergic synapses. The blockade of GABAergic transmission precipitates epileptic seizures (for review, see Refs. 56, 157). Interestingly, lipophilic derivatives of piperidencarboxylic acid, which have anticonvulsant properties (524), are highly specific inhibitors of GAT-1 transporter (58, 102, 103); most probably, GAT-1 transporter is the site of action of the anticonvulsant drug tiabagine (56). The GAT-1 uptake system is more extensive than the GABA synthesizing system, and it is also expressed in glial cells of the cerebral and cerebellar cortex and spinal cord (364, 375, 447, 506). Then, GAT-1 should play additional roles to its presynaptic function. In this sense, it might contribute to the regulation of the cerebrospinal concentration of GABA. To our knowledge, antisense or knockout experiments on GAT-1 to demonstrate this have not been reported.

The channel mode of action of GAT-1 suggests a role for this transporter in intracellular signaling. Sonders and Amara (507) have recently reviewed the potential physiological role of the GABA transporter-associated electric activity; at present, evidence in favor of this role is less documented than that for the excitatory amino acid transporters (see sect. IIC). Application of GABA to glial cells elicited depolarization in a sodium-dependent manner, which is therefore more attributable to transporters than to receptors (291). Depolarization caused by the transport of GABA can trigger intracellular signals. In isolated skate retinal horizontal cells, GABA-evoked transport currents (as determined by the pharmacology and ion dependence) depolarize cells and open voltage-sensitive Ca2+ channels (200).

The pharmacological characteristics of the GAT-2 and GAT-3 transporter fit the GABA glial transporter activity (see above), but GAT-3 is present in glial and neuronal cells (61, 103, 450). Studies in the developing brain (239) showed a coordinate expression of GAT-1 and GAT-3 that suggested a role for the latter in the termination of GABAergic synapses. In contrast, GABA uptake inhibitors with anticonvulsant properties do not inhibit uptake via GAT-3 (102). Nelson and co-workers (240) noticed that the general pattern of GAT-3 immunoreactivity is similar to that of GLYT1, which suggested a possible role of GAT-3 in neurons with GABA termini known to be apposed to glycine receptors. As discussed above, the localization of GAT-2 in brain, retina, and peripheral tissues points to a nonneuronal function for this transporter.

There is codistribution in brain of the neuronal-specific glycine transporter GLYT2, the inhibitory glycine receptor (immunocytochemistry and strychnine binding), and neurons with high glycine content (17, 237, 623). In contrast, codistribution of GLYT1 and inhibitory glycinergic neurotransmission is not complete (for review, see Ref. 622). First, this suggests participation of both glycine transporters in the termination of glycinergic synapsis. Second, GLYT1 might have additional roles in the CNS. The presence of GLYT1 (mRNA and protein) in neurons is controversial, but its presence in areas without glycine receptor expression suggests additional roles (622). Based on the mRNA distributions, Smith et al. (502) postulated a role for GLYT1 in the modulation of the glutamatergic transmission through NMDA receptors. This receptor requires two coagonists, glutamate and glycine, to activate its channel (234, 287). Administration of glycine or related agonits potentiates NMDA receptor function in several models in vivo (reviewed in Ref. 622). The concentration of glycine needed to restore normal NMDA electric responses is controversial, and this is an important issue in assessing the role of the glycine transport in the modulation of glutamatergic transmission; 0.1 µM glycine is enough to restore normal NMDA responses, but it is estimated that glycine concentrations in the cerebrospinal fluid are >10 µM (622). Two findings leave room for a role of glycine transport in NMDA receptor function. First, the affinity of the NMDA receptor for glycine varies in different neurons (432), depends on the type of NR2 subunits within the receptor (584), and increases with extracellular calcium (182). Second, the concentrative glycine transport present in brain synaptic plasma membrane and glial cells, with a cotransport stoichiometry of 1 glycine, 2 sodium, and 1 chloride (16, 624) (to our knowledge the stoichiometry for the expressed GLYT1 transporter has not been reported) is able to maintain a concentration of 0.2 µM in the synaptic cleft (23). The finding of brain regions with high NMDA receptor expression without glycinergic terminals also points to a role of the glycine transporter in the regulation of glutamatergic synapsis. The GLYT1 transporter might regulate glycine concentration in the synaptic cleft, not only by the uphill uptake of glycine, but also by the efflux of glycine via the transporter working in the opposite direction (622). This is similar to the nonvesicular release of GABA and glutamate (23). Thus, with synaptic depolarization, the intracellular concentration of sodium increases to levels at which the operation of the glycine transporter is reversed (15). Additional studies are needed to demonstrate the role of the GLYT1 transporter in changing the glycine concentration in the synaptic cleft and thus modulating the NMDA glutamate receptor.

Transporters for two other amino acids, taurine and proline, have been identified in brain. The physiological role of taurine and beta -alanine, two substrates of the TAUT transporter, in the CNS remains obscure. As reviewed by Nelson and co-workers (316) and Weinshank and co-workers (503), taurine is one of the most abundant amino acids in brain, with a function that is best understood as an osmoregulator. Because taurine is degraded slowly, uptake is a relevant way to regulate its extracellular concentration; taurine can be released from neurons and glial cells in response to changes in cell volume. Taurine reaches millimolar concentration in excitable tissues that generate oxidants, a decrease in taurine uptake has been implicated in retinitis pigmentosa, and depletion of taurine results in retina degeneration; this supports a role of taurine in neuronal survival (316, 368, 503). Although most animals can synthesize taurine, this is not sufficient and the supply relies on dietary sources (503). This role for taurine transport is consistent with the presence of TAUT transporter in the blood retinal barrier (retinal pigment epithelium) (368) and in the placenta (445). A more precise cellular localization in the brain of the high-affinity TAUT transporter (see sect. IIB1) is needed to understand its role in the CNS.

The cloning of PROT transporter, a sodium- and chloride-dependent high-affinity L-proline transporter expressed in CNS, suggested a role for L-proline in neurotransmission. Circumstantial evidence implicates L-proline as a putative synaptic regulatory molecule (see Refs. 155 and 489 for review).

1) As for neurotransmitters, high-affinity sodium-dependent uptake of L-proline has been described in rat brain synaptosomes and slices, and L-proline is released from brain slices and synaptosomes by potassium-induced depolarization. In general, the pharmacology (potency of inhibitors) of the L-proline transport of the human PROT expressed in HeLa cells is consistent with the uptake of L-proline in brain slices.

2) L-Proline and its high-affinity synaptosomal transport show heterogeneous regional distribution in the CNS.

3) There is a synaptosomal L-proline biosynthetic pathway from ornithine.

4) L-Proline produces complex electrophysiological actions when iontophoresed onto neurons.

5) Intracerebral injections of L-proline are neurotoxic and disrupt memory processes. Interestingly, the neurotoxic effects of L-proline are blocked by antagonists of glutamate receptors.

Transcripts of PROT are concentrated in subpopulations of putative glutamatergic neurons, and the protein is detected in synaptosomes (155, 489), and a specific postsynaptic role of L-proline at glutamate receptors has been suggested (for references, see Ref. 155). In this scenario, several roles have been proposed for the uptake of L-proline via the neuronal PROT transporter (489): 1) limitation, or modulation, of the extracellular concentration of L-proline to prevent inadequate activation of, or modulate, inhibitory synapses (glutamate and glycine receptor); and 2) a nutritional role for feeding the neuronal tricarboxylic acid cycle with intermediates or for presynaptic accumulation of L-proline as a precursor of L-glutamate. Recently, a novel nonopioid action of enkephalins has been described (156): enkephalins, which are not PROT transporter substrates, competitively inhibit this transporter. Subcellular localization, design of specific inhibitors (156, 489), and knockout strategies for the PROT transporter should help to elucidate the role of L-proline and its transporter in CNS.

B) OSMOLYTE TRANSPORTERS. Both BGT-1 and TAUT are established osmolyte transporters. Hypertonicity produces loss of cell water; cells shrink within seconds to equalize intra- and extracellular osmolarity. An immediate consequence, within minutes, is the elevation of the intracellular concentration of electrolytes (e.g., K+), which perturbs the function of macromolecules (for review, see Refs. 193, 292, 293). Cells of all phyla have a fundamental response, the accumulation of nonperturbing small organic osmolytes (e.g., betaine, taurine, myo-inositol, and sorbitol) (618). This response occurs in mammalian cells within hours or days. Accumulation of compatible osmolytes and keeping intracellular electrolyte levels isotonic is critical for adaptation to hypertonic stress (for references, see Ref. 292). The kidney medulla is the only mammalian tissue that is normally hypertonic, due to the concentration mechanisms of urine; depending on the hydration status of the animal, osmolarity varies, and in humans, it easily surpasses 1,000 mosM (193, 292). As part of the renal mechanisms for water conservation, cells in the medula accumulate betaine, myo-inositol, taurine, and sorbitol; induction of transport for the first three osmolytes and synthesis of sorbitol are the mechanisms that produce osmolyte accumulation (386, 387).

In a series of elegant studies, Handler's and Burg's groups demonstrated the role and the mechanisms of regulation of BGT-1 transport activity in the adaptation of renal cells to osmolarity changes (reviewed in Refs. 194, 292, 293).

1) Hypertonic media induce betaine accumulation in the canine renal MDCK cells (385).

2) The MDCK cells rely on transport as a source of betaine; the transport activity of BGT-1 is the rate-limiting step in the accumulation of betaine (387, 614).

3) Addition of betaine at physiological concentration (100 µM) restores the MDCK cell growth and survival challenged by hypertonicity (563).

4) In MDCK cells, hypertonicity induces the parallel and progressive turn-on of BGT-1 transcription, which raises BGT-1 transcripts and transport activity (~6-fold increase); hypotonicity reverses this process (387, 564, 614).

5) The BGT-1 transporter seems to play the same role in renal medulla: renal preparations show sodium- and chloride-dependent betaine uptake (292), the increase in medular tonicity due to dehydration results in an increase in BGT-1 transcript levels (quoted in Ref. 292), and intraperitoneal administration of NaCl in rats rapidly increases BGT-1 transcript abundance in renal medulla, mainly in the thick ascending limbs of Henle's loop (365). In these whole animal models, the expected increase in transcriptional activity of BGT-1 gene has not been reported.

After cloning the complete BGT-1 gene (537), Handler and co-workers (538) identified the first-known tonicity-responsive sequence element (TonE; TGCAAAAGTCCAG) 50-62 bp upstream of the first exon of the BGT-1 gene. Electrophoretic mobility-shift assays with TonE-containing DNA and nuclear extracts from MDCK cells revealed the presence of a specific binding protein (TonEBP) (538), whose binding activity is induced by hypertonicity with a similar time course to the transcription of the BGT-1 gene (292). To our knowledge, the cloning of TonEBP has not been reported. Hypertonicity also results in an increase in transcript abundance of TAUT and SMIT (sodium/inositol) transporters (615, 562) and aldose reductase for the synthesis of sorbitol (499). A tonicity-responsive sequence element has recently been identified in the mouse aldose reductase promoter, and it is similar to the TonE in the BGT-1 promoter (123). The cell tonicity sensor and its signal transduction pathway (e.g., involvement of mitogen-activated protein kinases) to TonE-dependent transcription in mammalian cells is not yet clear (73, 294). Hormones, cumulative substrate uptake, and oxidative stress also produce the swelling/shrinking reaction in a variety of mammalian cells; in addition to renal medulla, osmolytes have been identified in astrocytes, hepatocytes, lens epithelia, and macrophages (for a short review, see Ref. 594). In mouse macrophages (RAW 264.7 cells), as in MDCK cells, BGT-1 transcript abundance and betaine uptake is strongly dependent on extracellular osmolarity (593). In contrast, in H4IIE rat hepatoma cells and primary hepatocytes, taurine is the most prominent osmolyte, and its TAUT transporter is strongly regulated by tonicity (594). To our knowledge, whether the L-proline (PROT) and glycine (GLYT1) transporters, homologous to BGT-1 and TAUT transporters, are regulated by tonicity has not been reported.

C. Superfamily of Sodium-Dependent Transporters for Anionic and Zwitterionic Amino Acids

This family of transporters comprises five anionic amino acid transporters (EAAT1 to EAAT5; for excitatory amino acid transporter isoforms 1 to 5) and three zwitterionic amino acid transporters (ASCT1, ASCT2, and ATBo; for system ASC transporter isoforms 1 and 2, and for amino acid transporter for system Bo) that are sodium dependent (18, 144, 245, 269, 424, 488, 518, 568). The human homologs for all these transporters appear to have been cloned (18-20, 144, 249, 268, 343, 488, 490, 491, 568). Indeed, most probably ASCT2 is the murine counterpart of human ATBo (see below). The cDNA accession numbers, chromosome location, and protein length for the human transporters are shown in Table 6. The simultaneous cloning from different labs resulted in a confusing array of clone names (see Table 6). For clarity, in this review, the nomenclature used by Amara's group (11) has been adopted for the glutamate transporters (i.e., EAAT). On the basis of sequence homology, two subfamilies could be interpreted: 1) the human EAAT isoforms (the anionic transporters) show 36-65% amino acid sequence identity between them, and 2) the human zwitterionic amino acid transporters ASCT1 and ATBo show 57% between them. Amino acid sequence identity between human ATBo and mouse ASCT2 is very high (~80%). Amino acid sequence identity between members of the two subfamilies ranges between 39 and 44%. Homologous prokaryotic proteins (26-32% identity) involved in the transport of glutamate and other dicarboxylates (140, 233, 558, 559, 586) are evolutively related to this amino acid transporter superfamily (205, 247, 252).

 
View this table:
[in this window] [in a new window]
  		TABLE 6.   Sodium-dependent transporters for anionic and zwitterionic amino acids

Three independent labs cloned the first three members of this superfamily almost simultaneously in 1992, using different approaches. Stoffel and co-workers (518), during the isolation of a galactosyltransferase from rat brain, purified a 66-kDa hydrophobic glycoprotein which upon protein microsequencing and oligonucleotide probe cloning resulted in the isolation of GLAST cDNA (EAAT1 in Table 6); the putative protein showed significant homology with prokaryotic glutamate and dicarboxylate transporters, and its expression in oocytes resulted in anionic amino acid transport activity. Kanner and co-workers (120) purified to apparent homogeneity a rat glial 70- to 80-kDa glycoprotein that upon reconstitution showed L-glutamate transport activity; they raised antibodies against the protein (121) and used them to isolate GLT-1 cDNA (EAAT2 in Table 6). Transfection of EAAT2 in HeLa cells resulted in a sodium-dependent glutamate uptake that was thereafter reconstituted in liposomes (424). Finally, Hediger's lab (245) isolated EAAC1 cDNA (EAAT3 in Table 6) from rabbit small intestine by expression cloning in oocytes, after a sodium-dependent L-glutamate uptake expression signal. The other transporters of this superfamily were cloned based on homology with these seminal sequences. The transport properties of EAAT1, -2, and -3 did not fulfill all the pharmacologically distinguishable glutamate uptake activities in the cerebellum and retina (119, 152, 454). This was pursued by Amara's group to isolate EAAT4 cDNA from human cerebellum by using degenerate oligonucleotides from two conserved regions of EAAT1-3 (i.e., one between TM domains III and IV, the other in the hydrophobic long stretch toward the COOH terminus of the proteins; see Fig. 5) as the starting point of a RT-PCR-based cloning strategy. Expression in oocytes confirmed the sodium-dependent L-glutamate transport activity of EAAT4 (144). More recently, the same group isolated EAAT5 from human retina (18), based on a previous glutamate transporter cDNA isolated from salamander retina by homology strategies (not reported). Independently, two groups cloned human ASCT1, the first zwitterionic amino acid transporter of the family (also named SATT; see Table 6). Amara and co-workers (20), using a degenerated oligonucleotide from a conserved region of the long hydrophobic stretch of EAAT1-3 (see Fig. 5), isolated ASCT1 from motor cortex, and Fremeau et al. (155) isolated SATT from hippocampus taking advantage of a human expressed sequence tag (EST) (1) that showed slight homology with an Escherichia coli glutamate/aspartate transporter (586). Both proteins upon expression in HeLa cells or oocytes exhibit sodium-dependent zwitterionic amino acid transport activity with characteristics of system ASC (20, 488). After the corresponding corrections, sequences ASCT1 (accession no. L14595) and SATT (accession no. L19444) are identical. Serendipity again added to our structural knowledge of amino acid transport. While screening a 3T3-L1 adipocyte cDNA library for clones encoding protein tyrosine phosphatase HA2, Liao and Lane (311) isolated a cDNA (AAAT in Table 6) that showed significant homology with the previous members of this family of amino acid transporters. Independently, Kanai's group (144), after a strategy based on RT-PCR amplification using degenerate oligonucleotide from similar conserved regions to those of Amara's group (20), isolated ASCT2 cDNA from mouse testis (568). Expression in oocytes and transfection in 3T3-L1 preadipocytes of ASCT2 and AAAT resulted mainly in sodium-dependent zwitterionic amino acid uptake, also with characteristics of system ASC (311, 568). Finally, a low-stringency screening of a human placental choriocarcinoma cell cDNA library with a human ASCT1 probe (20) allowed Ganapathy and co-workers (269) to isolate ATBo cDNA. Expression in HeLa cells and oocytes showed a sodium-dependent uptake of zwitterionic amino acids with characteristics of system Bo (269). This group screened a mouse kidney cDNA library using rat ATBo cDNA as a probe, and all the positive clones obtained turned out to be ASCT2 (V. Ganapathy, personal communication). This, and the homology and the similarities of the amino acid transport associated with the expression of ASCT2 and ATBo (see below) strongly suggest that both cDNA are species counterparts of the same gene. A search in dbEST performed by the authors of the present review (December 1996) found no other EST clones indicative of new members of this superfamily of transporters in mammals.


View larger version (113K):
[in this window]
[in a new window]
  		FIG. 5.   Amino acid sequence comparison of excitatory amino acid transporter (EAAT) and related zwitterionic amino acid transporters. Alignment of amino acid sequence of 7 human transporters of this superfamily corresponds to that published by Arriza et al. (18) for EAAT1-5 isoforms. Amino acid residues present in at least 6 of listed sequences are indicated by gray boxes. Amino acid residues that are specific to glutamate transporters or zwitterionic amino acid transporters of this superfamily are in white on black boxes. Because of weak homology present in NH2 and COOH termini of these transporters, conserved amino acid residues in those domains are not indicated. Straight lines over sequences indicate putative TM domains (I-VI). Dashed line over sequence delimits highly conserved, long hydrophobic stretch of amino acid residues with a controversial topology. Between TM III and TM IV, 2 or 3 potential N-glycosylation sites are conserved (open boxes). Dashes indicate gaps for sequence alignment.

Since the identification of the first members of this transporter superfamily, several excellent reviews have appeared related to these glutamate transporters (205, 243, 247, 252, 253). The study of the transporters of this superfamily has revealed intringuing concepts, as follows.

1) As a general rule, all these transporters show potassium dependence in addition to sodium dependence when expressed in different cell system, with the possible exception of the zwitterionic amino acid transporters ASCT1 (627), ASCT2 (568), and ATBo (269) (see Table 7). The characteristics of their expressed transport activity suggest that EAAT1-5 isoforms are neural and nonneural variants of the anionic amino acid transport system X -AG, and that ASCT1 and ASCT2 might be variants of system ASC transport activity, or that ATBo and ASCT2 correspond to the epithelial system system Bo.

 
View this table:
[in this window] [in a new window]
  		TABLE 7.   Tissue distribution and transport characteristics of expressed EAAT and related zwitterionic amino acid transporters

2) Several labs provide evidence that some of these transporters (EAAT1-5, ASCT-1, and perhaps ATBo) have a substrate (amino acid and sodium)-gated chloride channel mode of action in addition to their amino acid transport mode of action (18, 144, 270, 582, 583, 627; reviewed in Ref. 507).

3) The topology of all these transporters in the plasma membrane is difficult to imagine on the basis of hydrophobicity algorithms (6-10 TM domains have been suggested, including beta -sheets TM domains; Refs. 245, 424, 518), and it is still not evident after experimental work, from three different labs, attempting to elucidate this issue for EAAT1, EAAT2, and a prokaryotic glutamate transporter (140, 498, 585).

4) Specific antisense experiments, both in vivo and in cell culture, represent the first experimental evidence of the physiological role of EAAT isoforms in brain (458). More recently, Stoffel and co-workers (416) obtained the null knockout EAAT3 mice, and Tanaka and co-workers (544) obtained the null knockout EAAT2 mice.

These aspects as well as structure-function relationship studies on these transporters are discussed in the following sections.

1. Tissue expression

The tissue distribution of the transporters of this superfamily was studied initially by Northern analysis and in situ hybridization. Table 7 shows the mRNA tissue distribution and the mRNA size for these human transporters, except for the mouse ASCT2. Rat EAAT1-2 and rabbit EAAT3 show a similar tissue distribution and transcript size (245, 424, 518); for the rabbit EAAT3, an additional transcript band of 2.5 kb has been reported (245). For the human ASCT1, Shafqat et al. (488) reported two hybridization bands of ~2 kb (2.8 and 2.2 kb) instead of a single band of ~2.5 kb (20).

The human EAAT glutamate transporters have specific tissue distribution. The EAAT2 isoform is specific to the CNS, EAAT5 is mainly, if not solely, expressed in the retina, and EAAT4 is mainly expressed in cerebellum, whereas the other isoforms are also expressed in peripheral tissues (Table 7). Rat EAAT1 is also brain specific (see below). Tissue and subcellular distribution of EAAT1-3 transporters in the CNS has been studied at the protein level (85, 121, 298, 305, 461), largely confirming previous in situ hybridization analyses (205, 245). These studies showed that rat EAAT1 and EAAT2 are present in astroglial cells, whereas EAAT3 is present in neurons. Glial EAAT2 is distributed throughout the brain and spinal cord (most abundant in Bergmann glia in the cerebellar molecular layer); neuronal EAAT3 is also generalized, but it is more prominent in hippocampus than cortex and striatum and less abundant in cerebellum; and glial EAAT1 is most prominent in cerebellum (cerebellum >> hippocampus > cortex > striatum) (85, 298, 461, 581). Ultrastructural analysis revealed that glial EAAT1 is localized to processes that most probably envelop glutamatergic synapses, suggesting a role in synaptic plasticity (461). In contrast, glial EAAT2 localized to cell bodies and processes (461). Neural EAAT3 localized to axons, dendrites, and presynaptic terminals in the deep cerebellar nuclei, but in the cerebellar granule cells it is not expressed presynaptically (461). Astrocyte membranes facing nerve terminals, axons, and spines are higher in EAAT1-2 glutamate transporters than those facing capillaries, pia, or white matter (85). Interestingly, EAAT3 protein is present in, but not restricted to, glutamatergic neurons (e.g., some but not all cortical pyramidal neurons), and it is enriched postsynaptically within GABAergic Purkinje cells in cerebellum (461). This was confirmed by in situ hybridization studies (205, 581), where an additional localization of EAAT3 to cholinergic alpha -motor neurons of the spinal cord was shown. These results indicate that neuronal EAAT3 is not the presynaptic glutamate transporter for all glutamatergic synapses and that it is also present in nonglutamatergic neurons. Additional glutamate transporters may be present presynaptically in many glutamatergic neurons. In this sense, the EAAT4 isoform may be a good candidate for those neurons in cerebellum. The presence of neuronal EAAT3 in GABAergic synapsis suggests that it could transport glutamate intracellularly as a precursor for GABA synthesis (461). The general role of EAAT1-3 isoforms in brain in the maintenance of low extracellular glutamate concentration has been addressed by partial and specific knockout of these isoforms (458) and the complete knockout of EAAT3 (416) and EAAT2 (544) (see sect. IIC8). To our knowledge, human EAAT4 and EAAT5 distribution has been studied only by Northern analysis, showing an almost specific cerebellar and retinal localization, respectively (144). It is worth mentioning that the human EAAT5 has a putative synaptic localization: the COOH terminus of EAAT5 contains a sequence motif (Glu-Ser or Thr-X-Val-COOH) found in synaptic membrane proteins and that interacts in the yeast two-hybrid assay to PDZ (a modular protein-binding motif) domains of the postsynaptic density-95 protein (18).

The peripheral distribution of EAAT1 (brain > heart > skeletal muscle > placenta and lung in human tissues, Ref. 19; specifically expressed in brain for the rat, Ref. 518), EAAT3 (small intestine > kidney > brain > cerebellum, lung, placenta, heart; similar for human and rat tissues, Refs. 19, 245), EAAT4 (cerebellum > > placenta, for human tissues, Ref. 144) and EAAT5 (weak signals in liver and of different transcript size in skeletal muscle and heart from human tissues, Ref. 18) isoforms has been, to our knowledge, only addressed by Northern analysis. It is noticeable that rat EAAT1 seems to be specific to brain (518), whereas peripheral human tissues express this isoform (19). Kanai and Hediger (245) performed in situ hybridization studies with EAAT3 in the small intestine and showed that the transporter is expressed in epithelial cells. This, together with its expression in kidney, suggests a participation of this glutamate transporter in the reabsorption of anionic amino acids (see sect. IIC8).

The tissue distribution of the zwitterionic amino acid transporters ASCT1, ASCT2, and ATBo has been examined by Northern blot analysis (20, 269, 488, 568). Human ASCT1 transcripts (~5, ~4, and ~2.4 kb; Ref. 20; and ~4.8, ~3.5, ~2.8, and 2.2 kb; Ref. 488) were detected in skeletal muscle, pancreas > brain > placenta > heart, lung, kidney, liver. The mouse ASCT2 transcript (Table 7) shows a complementary tissue distribution: lung, large intestine, kidney > skeletal muscle, testis and white adipose tissue (568). Expression of the two suspected transporter variants of system ASC (ASCT1-2) in liver and placenta is very low. If this low level of expression also occurs at the protein level, to our knowledge not yet studied, different ASCT variants or transporters not related to this superfamily may be responsible for the high system ASC transport activity present in these tissues (488) (see sect. IIC8). The human ATBo transcript is present in epithelial tissues (placenta, lung > kidney, pancreas), very scarce in skeletal muscle, and absent, according to Northern analysis, in heart, brain, and liver. It has been proposed that ATBo corresponds to the apical (i.e., brush-border membrane) system Bo (Ref. 269; see below) and is therefore expected to be expressed in small intestine, where this transport activity has been shown (329, 337, 515). Indeed, rabbit ATBo cDNA has been isolated from jejunum (270). It is worth mentioning that the tissue distribution of ATBo is similar to that of ASCT2, again favoring the hypothesis that both cDNA are species counterpart of the same gene. Immunolocalization studies to demonstrate the presence of ATBo in epithelial brush-border membranes are not yet reported.

2. Transport properties

The transport properties of the glutamate (EAAT1-5) and zwitterionic amino acid transporters (ASCT1-2 and ATBo) have been studied upon expression in oocytes or transfection into mammalian cells followed by both radioisotopic and electric measurements (see Table 7). Reconstitution studies from the expressed protein have been reported only for the glutamate transporter EAAT2 (424, 426, 630). All these transporters share several transport properties (see Table 7). 1) The glutamate EAAT transporters are sodium cotransporters and potassium countertransporters. Potassium dependence (countertransport) has been demonstrated for all of these transporters, except EAAT4 and EAAT5, for which it has not been reported (18, 144). 2) The zwitterionic amino acid transporters (ASCT1, ASCT2, and ATBo) are probably electroneutral sodium-dependent amino acid exchangers that do not interact with potassium (162, 627), but to our knowledge, the putative exchange mechanism of transport for ASCT2 and a role of potassium for ATBo and ASCT2 has not been properly tested. 3) In addition to their transport mode of action, for EAAT1-5 and ASCT1 cumulative evidence has been obtained that they also have a chloride channel mode of action (144, 582, 583, 627). Very recent data also suggest that this channel activity might be associated with ATBo (270). To our knowledge, this has not been reported or tested for ASCT2, and the only available study shows no electric activity associated with this transporter (568) (see below).

Substrate specificity has been described for the EAAT isoforms and the zwitterionic amino acid transporters of this superfamily (see Table 7). In constrast to these zwitterionic amino acid transporters, EAAT transporters express high-affinity sodium- and potassium-dependent transport of L-glutamate (micromolar range). In contrast, the interaction of these transporters with zwitterionic amino acids (i.e., either as a substrate or as a cis-inhibitor) is of low affinity (millimolar range). All this information is consistent with a large body of accumulated knowledge on the native glutamate transport properties in cells, membrane preparations, synaptosomes, and purified transporters of neural origin (for review, see Refs. 251, 394). Interestingly, these EAAT isoforms do not interact with neurotransmitters that are substrates of the superfamily of sodium- and chloride-dependent transporters of neurotransmitters (e.g., GABA, dopamine, norepinephrine, serotonin) (reviewed in Ref. 252). The five glutamate transporter isoforms are stereospecific for glutamate but do not discriminate the enantiomers of aspartate, a characteristic of sodium-dependent glutamate transport activities in many tissues (185, 262, 483, 505, 513); only for human EAAT5 is the apparent Km for L-aspartate approximately fivefold lower than that for D-aspartate (18). The five glutamate transporters form a continuum of transport activities with a different tissue-specific distribution, and they are difficult to distinguish with the usual functional criteria, other than by their pharmacology (see below). Their activity corresponds to system X -AG, which represents various transport agencies that carry anionic amino acids, with the described characteristics, into different cells (41). Moreover, as expected for system X -AG (41), threo-3-hydroxy-DL-aspartate (THA) is a competitive substrate of EAAT1-3 (19, 518). When the five human EAAT isoforms are expressed in oocytes, the highest affinity is shown by EAAT4 [apparent Km of 2-3 µM for L-glutamate, L- and D-asparate, and the analog L-trans-pyrrolidine-2,4-dicarboxylic acid (PDC)]; the other isoforms showed apparent Km values for glutamate in the range between 10 and 64 µM (18, 19). These Km values (and those shown in Table 7) obtained in oocytes should be interpreted with caution since they depend on the expression system used: human EAAT2 has an apparent Km for L-glutamate of 18 µM in oocytes (19), whereas for rat EAAT2 (i.e., GLT1) expressed in HeLa cells, the Km is 10 µM, and 2 µM when it is reconstituted in proteolisosomes (424). This latter value fits the affinity of L-glutamate when assayed in rat brain membrane preparations or in rat brain purified and reconstituted glutamate transporter (120, 255, 263, 425). Endogenous L-glutamate diluting the radiolabeled tracer, membrane lipid composition, posttranslational modifications, or other factors may affect the apparent affinity of these transporters in different expression systems (19, 424). Interestingly, the apparent Km values determined for rat EAAT1 in oocytes when measured by radiolabeled glutamate (~80 µM) are higher than those based on measurement of the induced glutamate-dependent current (11 µM) (289), favoring the impression that endogenous substrate diluting the radiolabeled substrate increases the apparent Km values in those studies. It is worth mentioning that for human EAAT3, dependence of the apparent Km on membrane potential has been described (249).

Pharmacological studies helped to distinguish between the transport activity of the five isoforms (see Table 7). Sequence comparisons strongly indicate that human EAAT1-3 corresponds to the rat GLAST1 (96% amino acid sequence identity; Ref. 518), rat GLT1 (95% amino acid sequence identity to the corrected sequence; Refs. 253 and 424), and rabbit EAAC1 (92% amino acid sequence identity; Refs. 245, 249). Because of this, in the present review, they are considered to be counterparts of the glutamate transporter isoforms in different species. In contrast, pharmacological similarities between species are less clear. Interaction of different inhibitors with the transport activity of the human EAAT isoforms is shown in Table 7. Human EAAT1 and EAAT3 show a similar pharmacology, which is different from that of human EAAT2 (19), EAAT4 (144), and EAAT5 (18). Human EAAT2 is the isoform sensitive to dihydrokainic acid (DHK) and kainic acid (KA), insensitive to L-alpha -aminoadipate (Lalpha AA), and relatively insensitive to L-serine-O-sulfate (SOS); human EAAT1 and EAAT3 are highly sensitive to SOS and insensitive (i.e., Ki >3 mM) to DHK, KA, and Lalpha AA; and human EAAT4 is the isoform that is sensitive to Lalpha AA and insensitive to DHK and KA (Ki >3 mM). In contrast to the other EAAT isoforms, for which THA and PDC are competitive substrates, for human EAAT5 these analogs act as blockers (Ki ~1-6 µM) and not as substrates (18). In constrast to the human EAAT transporters, rat GLT-1 (rat EAAT2 counterpart) and rabbit EAAC1 (rabbit EAAT3 counterpart) are very sensitive to Lalpha AA (Ki in the micromolar range) (245, 424). These results suggest pharmacological differences between the human and other EAAT isoforms. Brain regional differences in pharmacology have suggested at least four glutamate transporters subtypes (454); this is consistent with the four mammalian brain EAAT transporter isoforms isolated.

The zwitterionic amino acid transporters of this superfamily share as general transport properties high affinity for zwitterionic amino acids and low affinity for L-glutamate and sodium dependence, whereas the potassium dependence is controversial (20, 269, 488, 541, 627). A weak potassium dependence was described initially for ASCT1 (alanine-induced current, Ref. 20) and ASCT2 (alanine uptake, Ref. 568) upon expression in oocytes. To our knowledge, potassium dependence for ATBo has not been examined. Later on, Zerangue and Kavanaugh (627) stated the contrary, i.e., that alanine transport (both influx and efflux) was independent on the external potassium concentration in oocytes expressing ASCT1 under voltage clamp. One explanation of these dissimilar results, obtained with electric or radiolabeled L-alanine uptake studies, within the same group could be that alanine transport via ASCT1 is electroneutral, but sodium and the substrate evoked a chloride channel activity associated with ASCT1, which is thermodynamically uncoupled from the amino acid transport (627). It has been suggested that the sodium dependence of ASCT1 is not the result of sodium cotransport, since transport of L-alanine via ASCT1 is electroneutral and potassium independent, and the ligand-evoked chloride current shows a Hill coefficient of 1 for external sodium (627). In contrast, the electroneutral uptake of radiolabeled L-alanine via ASCT1 involves the electroneutral exchange of extracellular and intracellular amino acid substrates (627); influx and efflux of L-alanine gave a 1:1 flux stoichiometry with the substrate specificity of ASCT1. Interestingly, external sodium is needed for this exchange. Very recently, Ganapathy and Leibach (162) reported exchanger transport activity for ATBo, but a full paper is still lacking on this issue. In addition, a substrate-induced current that reverses at -30 to -40 mV has been associated with rabbit ATBo (270). These results offer a new view of the transport mechanisms of these zwitterionic amino acid transporters: 1) electroneutral exchange of zwitterionic amino acid which is dependent on extracellular sodium; 2) ligand-evoked chloride currents, which are thermodynamically uncoupled from the amino acid transport, as for the EAAT transporters of this superfamily; and 3) lack of interaction with potassium. Interestingly, these zwitterionic amino acid transporters do not conserve the glutamate residue identified as crucial for potassium countertransport of the EAAT transporters of this superfamily (see sect. IIC7). Additional studies are needed to clarify these transport mechanisms.

1) Does ASCT1 operate as an obligatory exchanger for a wide range of amino acid and sodium gradients through the plasma membrane? Unfortunately, in the Zerangue and Kavanaugh's study of the amino acid exchanger activity of ASCT1 (627), influx and efflux rates were only compared at a fixed concentration of external L-alanine and sodium; it is therefore possible that this exchange reflects ASCT1 transport near equilibrium and not a real obligatory exchanger mechanism. A similar transport mechanism, but concentrative, was described for system ASC in fibroblasts (76).

2) What is the role of sodium in the ASCT1 exchanger mode of action proposed by Zerangue and Kavanaugh (627)? Is the sodium ion really translocated through the membrane, or is its action like that of an allosteric modulator? Uptake measurements of 22Na are needed to clarify this issue. Then, if sodium is translocated through ASCT1, what mechanism explains the lack of concentrative transport activity of ASCT1 located along the large sodium electrochemical gradient of the plasma membrane?

3) Is the exchange mechanism of transport of ASCT1 and ATBo extensive to ASCT2? It has been reported that mouse ASCT2 shows no electric activity evoked by amino acid substrates and sodium even in the presence of chloride (568). This suggests an electroneutral amino acid exchange mechanism like that of ASCT1, and a not very conspicuous ligand-evoked channel mode of action, but to our knowledge, amino acid exchange via ASCT2 has not been examined. On the other hand, the lack of electric activity of ASCT2 (627) might be a consequence of charge flux compensation between the transport mode of action of the carrier and a yet unknown chloride channel associated activity? If so, what is the electric activity of ASCT2 in a chloride-free system (i.e., extracellular chloride = 0 and chloride-depleted oocytes)?

4) Finally, is the ligand-evoked chloride channel activity associated with the glutamate transporters of this superfamily and to ASCT1 extensive to ASCT2 and ATBo? The ligand-evoked current associated with rabbit ATBo reverses at -30 to -40 mV (270). Is this a reflection of an associated ligand-evoked channel activity? If so, its ion selectivity might be different from the channel activity associated with the other members of this superfamily that tends to reverse current at the equilibrium potential of chloride in oocytes (i.e., -23 mV).

Transporters ASCT1 and ASCT2 upon expression in HeLa cells or oocytes have transport properties similar to system ASC (185) (see Table 7), i.e., high-affinity (micromolar range) and sodium-dependent transport for neutral amino acids (including small ones like alanine, serine, and cysteine) and interaction with anionic amino acids at pH <7.4 (i.e., the protonated amino acid species is probably the substrate), and insensitive to inhibition by dibasic amino acid and the analog MeAIB (20, 488, 541, 568). This suggested that these two carriers represent isoform variants of system ASC. Indeed, system ASC showed significant variability in substrate specificity; e.g., threonine is a better substrate than cysteine in rat liver, but the converse is true in the hepatoma cell line HTC (165, 196, 275, 569). Northern analysis showed a very low expression of ASCT1 and no expression of ASCT2 in liver, a tissue with high system ASC activity. This suggests that neither ASCT1 nor ASCT2 represents the liver system ASC transporter. Work by Kilberg's group aiming to identify a liver homolog of ASCT1-2 failed to isolate a new transporter cDNA (M. S. Kilberg, personal communication). System ASC is concentrative, voltage dependent, and electroneutral (185). It is necessary to clarify whether ASCT1 and ASCT2 mediate a concentrative transport (see above) of their substrates before ascribing ASCT1 and ASCT2 transporters to variants of system ASC.

The transport characteristics of ATBo upon expression in HeLa cells and oocytes, as well as the epithelial distribution of ATBo transcripts (269), suggested that this transporter corresponds to the epithelial system Bo, the major apical sodium-dependent transport for zwitterionic amino acids with broad specificity (excluding methylated amino acids and dibasic and anionic amino acids) (329, 337, 515). The new view that ATBo and ASCT2 might be species counterparts of the same transporter with an electroneutral amino acid exchange mechanism of transport (162, 270; V. Ganapathy, personnal communication) compromises the initial expected role of ATBo in the active epithelial uptake of neutral amino acids.

3. Stoichiometry

The stoichiometry of glutamate or zwitterionic amino acids and the corresponding coupled ions, as well as the participation of pH-changing ions in the transport cycle, varies among the members of the present superfamily. In general, for many cell types, glutamate uptake is electrogenic and driven by cotransport of sodium and countertransport of potassium with a first-order dependence on external L-glutamate and internal potassium and a sigmoidal dependence on external sodium, which suggests a stoichiometry of 1 glutamate:3 sodium:1 potassium (30, 263). In addition, movement of pH-changing ions (cotransport of H+ or countertransport of OH-) occurs during transport (63, 141, 392). In this instance, the stoichiometry of 1 glutamate:2 sodium:1 potassium:1 OH- would still be electrogenic. As discussed by Kanner (252), a stoichiometry of 1 glutamate:>= 2 sodium is favored by direct experimental evidence obtained by kinetic and thermodynamic methods (141, 510).

The stoichiometry of the cotransported ligands of EAAT transporters has been examined for rat EAAT1 (289) and human and rabbit EAAT3 (246, 628) expressed in oocytes, and the models proposed are different for each. Zerangue and Kavanaugh (628) offered evidence that the neuronal human EAAT3 cotransports one glutamate, one hydrogen, and one sodium and countertransports one potassium, with a glutamate-to-charge ratio of 1:2 (2 positive charges accompany the movement of glutamate and in the same direction), and with this flux-coupling model, a transmembrane gradient of glutamate through the transporter of 106 is predicted. This is based on the following.

1) Because of the uncoupled chloride channel activity associated with the expression of human EAAT3 in oocytes (582), this study was conducted in a chloride-free medium and chloride-depleted oocytes, or with oocytes clamped at chloride equilibrium potential (-25 mV) to avoid the contribution of this channel activity to the fluxes coupled to the transport of glutamate.

2) Flux of the pH-changing ion is thermodynamically coupled to transport, and it is not the consequence of permeation through the uncoupled chloride channel activity of the transporter (137, 144, 295, 582), because in voltage-clamp conditions uptake of glutamate resulted in intracellular acidification of the oocyte even with an inwardly directed electrochemical gradient for OH-.

3) Most likely, hydrogen is cotransported with glutamate as a carboxylate ion pair, instead of the countertransport of OH-, because transport of L-cysteine, a structurally related substrate that is transported as a neutral zwitterion (626), resulted in a clearly lower intracellular acidification than transport of an equivalent amount of glutamate. Interestingly, this is at odds with evidence that countertransport of anions (such as OH- or HCO-3) is responsible for the pH-changing activity of glutamate transport in salamander retinal glia cells (63).

4) Countertransport of potassium is coupled with glutamate transport because, in chloride-free medium and chloride-depleted oocytes expressing human EAAT3, superfusion of glutamate and sodium produced an inward current, and superfusion with potassium produced an outward current; both currents are inhibited by 5 mM KA, a competitive antagonist of EAAT isoforms (see sect. IIC2). Changes in the reversal potential caused by altering glutamate and potassium membrane gradients are consistent with the countertransport of potassium and glutamate.

5) By applying the changes in the glutamate transport reversal potential due to varying glutamate and ion gradients (i.e., for Na+, H+, or K+) to a zero-flux equation relating the membrane potential to the transmembrane ion gradients, the coupling coefficient for each substrate was similar to 1 glutamate:1 hydrogen:3 sodium:1 potassium.

6) Consistent with the previous data, the Hill coefficient for external sodium concentration at a fixed external glutamate concentration (10 µM) was between two and three, whereas this was not different from one when the external concentrations of glutamate, hydrogen, and potassium were varied.

7) This stoichiometry is consistent with a measured inward positive charge transfer of two during uptake of tracer L-glutamate (100 µM) under voltage clamp.

8) Finally, uptake of tritiated 1 µM L-glutamate in oocytes expressing human EAAT3 reaches a steady state of <10 nM external glutamate concentration, which with an estimated intracellular concentration of anionic aminio acids in oocytes of 12 µM (626) represents a transmembrane concentration gradient >106; this value is compatible with cotransport of three sodium per glutamate, but not with two sodium per glutamate.

Surprisingly, Zerangue and Kavanaugh's data (628) on human EAAT3 are different from those previously obtained in rabbit EAAT3-expressing oocytes by Hediger and co-workers (246). Initial studies aiming to elucidate the stoichiometry of rabbit EAAT3 revealed a Hill coefficient for external sodium that is dependent on the external concentration of glutamate, suggesting that the affinity for the sodium sites depends on the glutamate concentration; interestingly, the Hill coefficient was 2 at 200 µM glutamate and >2 at 10 µM glutamate (246). This covers the glutamate concentration range studied by Zerangue and Kavanaugh (628). Parallel measurements of 22Na and [14C]glutamate flux, intracellular pH changes, and substrate-evoked currents gave a stoichiometry of cotransport of one glutamate for two sodium ions, countertransport of one potassium ion, and either cotransport of one hydrogen or countertransport of one hydroxide (246). Unfortunately, these studies were conducted without clamping the oocyte membrane potential. In any case, it will be interesting to determine the accumulation capacity of both EAAT transporters at different glutamate concentrations (this has not been reported for the rabbit counterpart), as well as 22Na uptake measurements and determination of ion coupling coefficients at high glutamate concentration (e.g., 200 µM and up) for the human counterpart.

Consistent with Kavanaugh's stoichiometry for human EAAT3, the glial rat EAAT1 expressed in oocytes showed a Hill coefficient for the kinetics of external sodium and glutamate of approximately three and approximately one, respectively (289). In contrast to human EAAT3, the current associated with glutamate transport via rat EAAT1 was not changed by reducing external pH from 7.4 to 6.0, suggesting that glutamate is transported as an anion (289). These results allowed Stoffel's group (289) to propose a stoichiometry for rat EAAT1 of 1 glutamate, 3 sodium/1 potassium. This is at odds with the increased glutamate transport current obtained by decreasing the external pH in glial cells and in oocytes expressing the neuronal human EAAT3 (48, 628). In contrast, it is in agreement with data in fibroblasts and salamander glial cells (165, 482). This result might reflect the participation of different pH-changing ions for glutamate transporters of glial origin. In fact, studies of electrogenic uptake of glutamate into glial cells gave Hill coefficients between 2 and 3 (31, 482). In this sense, it will be very interesting to know if the stoichiometry and the pH-changing ions for the glial EAAT2 fit the data reported by Attwell and co-workers (63) in salamander retinal glia cells. To our knowledge, these studies have not been reported either for EAAT2, EAAT4, or the retinal EAAT5.

Initial data on the stoichiometry and ion dependence of the zwitterionic amino acid transport via ASCT1 and ASCT2 indicate variability of ion coupling among these transporters. Recent strong data supporting the idea that human ASCT1 is a sodium-dependent zwitterionic amino acid transporter exchanger (with a 1:1 stoichiometry of exchange) might help to clarify this issue (see above; Ref. 627). Because of the lack of precise information on the potassium dependence of ASCT2 and the mechanism of transport of ASCT2 and ATBo, the suggested substrate stoichiometry (568) is at present speculative.

4. Mechanism of transport

The mechanism of substrate translocation has been studied in membrane preparations of neural origin by Kanner's group (for review, see Ref. 252), for the expressed rabbit and human EAAT3, human EAAT2, and human ASCT1 (246, 249, 583, 627). Studies with synaptic plasma membrane vesicles from rat brain demonstrated that influx or efflux of glutamate is coupled to efflux or influx of potassium, respectively (261); the same situation has been demonstrated with salamander retinal glia cells (23, 529). In addition, in the absence of potassium, the glutamate transporter catalyzes exchange of its anionic amino acid substrates (255). This suggested that the glutamate translocation step is distinct from that of potassium. The glutamate transport cycle occurs in two parts: 1) translocation of sodium and glutamate, and 2) reorientation of the binding sites upon binding and translocation of potassium (252). When a comparison was made of the ion dependence of net flux with that of exchange, the binding order of the substrates during influx was shown as follows (255, 425): 1) an ordered binding of the coupled two or three sodium ions before the binding of one glutamate ion to the extracellular face, 2) translocation of the complex, 3) release of glutamate and the sodium ions (it is not clear whether the release or binding at cytosolic face is also ordered or random), 4) binding of potassium on the inside, and 5) translocation and release of potassium to start a new cycle. If necessary, the glutamate transport-coupled pH-changing ions could be translocated concomitantly with either glutamate (i.e., hydrogen) or potassium (i.e., hydroxide) steps (252). Additional evidence for the translocation of sodium and potassium ions in different steps comes from the recent demonstration that rat EAAT2 Glu404Asp mutant is able to mediate exchange of D-aspartate and sodium but not countertransport of potassium (266). The kinetics of human EAAT2 have been examined upon expression in oocytes by analysis of nonlinear capacitance (i.e., pre-steady-state or transient currents) (583), after the pioneer work of Wright and co-workers (411, 412), who aimed to analyze the conformational changes and kinetic properties of the sodium/glucose cotransporter. Human EAAT2 transient currents are compatible with the free carrier being in voltage-dependent equilibrium with the sodium-transporter complex and to a charge movement due to the binding of sodium to a site of the carrier within the membrane electric field (583). Again, EAAT isoforms show different mechanisms of transport, since these currents are not detectable with rabbit or human EAAT3 expressed in oocytes (see below; Refs. 246, 249). Analysis of the charge movement estimated a glutamate turnover of 4-27 s-1 over the voltage range 0 to -140 mV (583). Interestingly, these values are similar to those estimated with reconstituted glutamate transporters from rat brain, which were the basis for the cloning of rat EAAT2 (120). In contrast, these values are smaller than those obtained with salamander retinal glia cells (>103 s-1) (482). Again, this most probably reflects that these cells express EAAT transporters other than EAAT2 (e.g., the EAAT5 isoform).

Hediger and co-workers (246, 249) examined the mechanism of transport for the rabbit and human neuronal EAAT3 glutamate transporters. Expression in oocytes of these two EAAT3 counterparts by two independent labs demonstrated that the reverse mode of action (glutamate efflux; potassium influx) is a true reversal of the overall forward reaction (246, 628); thus both reactions have the same stoichiometry, basic mechanism of transport, and rate-limiting step (246), in agreement with the model proposed by Kanner (252). Pre-steady-state currents, substrate-specific current-voltage relationship, and sodium Hill equation-extracellular glutamate concentration relationship studies with human and rabbit EAAT3 expressed in oocytes (246, 249) allowed Hediger and co-workers to propose that 1) at low extracellular sodium concentration (less than Km) binding of sodium to the extracellular face of the transporter becomes rate limiting. 2) At high extracellular sodium concentration (more than Km), the rate-limiting translocation step depends on the extracellular L-glutamate concentration. At low concentration (less than or equal to mean affinity constant; the apparent L-glutamate affinity depends on the membrane voltage, Refs. 249, 246), L-glutamate binding to the extracellular face becomes rate limiting, whereas at high concentration of extracellular glutamate (more than mean affinity constant), the charge translocation step becomes rate limiting (i.e., translocation of the fully loaded carrier, the carrier plus L-glutamate plus 2 sodium ions). These results are consistent with previous studies on the high-affinity L-glutamate transport in renal brush-border membrane vesicles (208). In addition, cooperativity of sodium and glutamate binding for rabbit and human EAAT3 strongly suggests an ordered mechanism of binding: first one sodium binds with low affinity, second glutamate binds, and third the second sodium binds with high affinity. The transport cycle then progresses as in the model of Kanner (252). The authors hypothesize that the sodium leak (i.e., transport of sodium in the absence of glutamate) (see Fig. 6) through EAAT3 is a consequence of the translocation of the carrier with two sodium ions bound (246). All this suggests that the mechanism of the glutamate transport cycle is different for the neuronal and renal EAAT3 transporter (246, 249) and the glial EAAT2 transporter (252). In one aspect, the model proposed by Hediger's group for EAAT3 is in conflict with the suggestion that this carrier transports glutamate as a carboxylate ion pair together with one hydrogen ion, which acts as the pH-changing ion (628). Human and rabbit EAAT3 lack significant current relaxation in response to voltage jumps (246, 249). This suggests that the empty EAAT carrier is electroneutral (i.e., any conformational change of the carrier during the transport process does not involve movement of charge residues within the membrane electric field) and that the translocation complex (i.e., carrier plus glutamate plus 2 Na+) has a positive charge and the relocation complex (i.e., carrier plus K+ plus OH-) is electroneutral. It will be interesting to know the kinetic behavior of EAAT3 using cysteine as a substrate, the strategy used by Zerangue and Kavanaugh (628) to identify hydrogen as the pH-changing ion in EAAT3, and the behavior of EAAT3 (e.g., pre-steady-state currents, potassium-specific current-voltage relationship, and potassium Hill equation) in the reverse mode of action (i.e., absence of glutamate and presence of potassium in the external medium).


View larger version (70K):
[in this window]
[in a new window]
 

View larger version (74K):
[in this window]
[in a new window]
  		FIG. 6.   A. 10 TM transmembrane topology model for rat EAAT2 transporter. EL and IL loops are numbered, according to model proposed by Slotboom et al. (498). Amino acid residues conserved in all known transporters of this superfamily are indicated in gray circles, and residues that are specific for EAAT transporters are indicated in white over a black circle. Nine residues (in IL1, IL2, TM V, TM VI, IL3, TM VII, and TM X) and two glycosylated (Y) Asn residues in EL2 are marked by a square, indicating homologous residues that, after analysis by site-directed mutagenesis studies, have been shown to be critical for transport activity (see text for details). Shaded area delimits protein segment involved in kainate binding for human EAAT2 (571). To our knowledge, gene structure of rat EAAT2 has not been reported. Coding region of human EAAT2 gene is composed of 10 exons (360; not yet accessible in data bases). Human EAAT1 gene is composed of 10 exons (accession numbers in NCBI: Z31713, Z31703-Z31710; W. Stoffel, M. Dueker, R. Mueller, and K. Hofmann).

B. 6 TM alpha -helix plus 4 TM beta -sheet transmembrane topology model for human ASCT1 transporter. Amino acid sequence presented corresponds to that of human ASCT1. EL and IL loops are numbered according to model proposed by Wahle and Stoffel (585) for rat EAAT1. Amino acid residues conserved in all known transporters of this superfamily are indicated in gray circles, and residues that are specific for zwitterionic transporters are indicated in white over a black circle. Y, glycosylated residues. Shaded area delimits protein segment homologous to that involved in kainate binding for human EAAT2 (571). Exon boundaries within putative protein sequence are indicated, and exons are numbered (214).

5. Ligand-gated chloride channel activity associated with the transporters

As stated in the introduction to this section, for the human EAAT1-5 isoforms and the human zwitterionic amino acid transporter ASCT1, evidence for a mode of action as a sodium/amino acid-gated chloride channel has been offered by Amara's and Kavanaugh's groups (18, 144, 582, 583, 627; for review, see Ref. 507). In a recent review, Wright et al. (609) argued that this, in addition to a mode of action as water channel of the sodium-glucose SGLT1 cotransporter and water cotransport of SGLT1 and GAT1, and sodium leak by glutamate transporters and SGLT1, GAT1, and NET1, is indicative of the multifunctional (i.e., cotransporters, uniporters, channels, and water transporters) behavior of these cotransporters.

Let us examine the evidence for the sodium/amino acid-gated chloride channel activity of human glutamate EAAT1-5, rat EAAT2, and human ASCT1 transporters (18, 144, 266, 582, 627).

1) The initial observation was that the cotransporter-coupled currents were larger than expected for the cotransporter density and turnover number, i.e., the current exceeds the charge movement due to the amino acid transport. In chloride-depleted conditions or at the chloride equilibrium potential (-23 mV in oocytes), the current due to the transport mode of action could be estimated and would represent, at a membrane potential of -60 to -80 mV, only 5% of the total current elicited by glutamate and sodium in oocytes expressing EAAT4 (144) and varies from 50 to 73% in oocytes expressing human EAAT1-3 (582). In chloride-depleted conditions, glutamate did not induce measurable current in oocytes expressing human EAAT5 (18). Thus it appears that currents elicited by glutamate and sodium in EAAT5 are due primarily to the chloride conductance. For human ASCT1, with an electroneutral mode of transport, ~100% of the ligand-evoked current is because of chloride conductance (627).

2) In steady-state flux condition measurements, the reversal potential (Erev) evoked by the amino acid and sodium due to the expression of the transporters fits with the proportion of chloride currents and the electrogenicity due to the amino acid transport activity (144, 582, 627, 18) (see Table 7): Erev values for EAAT4, EAAT5, and ASCT1 are very close to the chloride equilibrium potential in oocytes, whereas for EAAT1-3, Erev is within a positive range of membrane potential. The lowest chloride conductance contribution is found with EAAT2, for which current reverses only at +60 mV.

3) The chloride current is thermodynamically uncoupled from the amino acid flux via these transporters (18, 144, 582, 627). Thus, for all these transporters, radiolabeled amino acid flux is independent of the chloride current direction and the magnitude of the anion flux when substituting chloride, and for EAAT4 and ASCT1, it is largely or totally independent of the membrane potential.

4) The chloride conductance is evoked by sodium and the amino acid substrate (18, 144, 582, 627), and with identical apparent affinities as in the transport mode of action of the corresponding transporters (144, 582, 627); kinetic analysis of the amino acid flux via EAAT5 has not been reported.

5) For all these transporters, the sodium/amino acid-evoked chloride conductance exhibits the same chaotropic selectivity sequence: SCN- > NO -3 > I- > Cl- (144, 582, 627); for human EAAT5, only the partial sequence NO -3 > Cl- has been reported (18).

What is the nature of this sodium/amino acid-evoked chloride current? All the data could be interpreted as an intrinsic chloride channel activity of these transporters gated by sodium and the amino acid or, alternatively, as the induction, via direct or indirect interaction, of a silent chloride channel of the oocyte. The ion selectivity of the chloride conductance associated with these transporters is different from that of the endogenous calcium-dependent chloride channel activity of the oocytes, and it is not inhibited by typical oocyte chloride channel blockers (144, 582). These data indicate that the chloride channel conductance associated with these transporters is not due to activation of known endogenous chloride channels of the oocytes. Interestingly, for EAAT4, the Vmax of transport for radiolabeled L-glutamate is higher (<2-fold) than that for L-aspartate, whereas the maximum current of the associated chloride channel for L-aspartate is higher (3-fold) than that for L-glutamate (144). This has been interpreted as showing that gating of the channel occurs after binding of the sodium ions and the amino acid, and the different size of L-glutamate and L-aspartate allows a larger chloride flux in the latter case (144, 582).

In our view, however, there is still room for activation of an endogenous silent chloride channel. It is conceivable that an intermediate complex of the amino acid transport cycle activates an endogenous silent chloride channel. In this hypothesis, the identical ion selectivity of this chloride channel activity for many of the transporters of this family would reflect the activity of a single type of channel, and the different magnitude of the chloride channel activity would be due to the activation of a differential number of chloride channels by these transporters. In this instance, the differential chloride current due to L-aspartate and L-glutamate in EAAT4-expressing oocytes might indicate that activation of the chloride channel occurs through interaction with a particular intermediate complex of the transport cycle that would become more represented in the steady-state obtained with L-aspartate than the one obtained with L-glutamate. This transport complex might be the translocating complex (carrier plus sodium/amino acid). Notice that the associated chloride channel activity of the potassium independent amino acid exchange via ASCT1 (627) and via the Glu404Asp EAAT2 mutant (266) points to this intermediate. As discussed above for the neurotransmitter transporter superfamily, only a small fraction of the expressed GAT-1 and NET cotransporters behaves as ion channels in the presence of the ligands (GABA or norepinephrine and Na+) (82, 160). This could be interpreted as an extremely low open probability of the channel (reviewed in Ref. 609) or as a low fraction of the transporters interacting with the endogenous silent channels. In our view, reconstitution of the sodium/amino acid-gated chloride channel activity in proteoliposomes containing expressed and purified EAAT or ASCT1 transporters may be the final demonstration of the intrinsic channel activity of these transporters. More simply, it will be interesting to know the relationship between amino acid transport rate and the chloride conductance at different levels of transporter expression.

For two members of the present superfamily, ASCT2 and ATBo, no chloride channel activity has been described, but to our knowledge, this has not been properly tested (269, 568). Very recently, the amino acid-evoked current associated with ATBo has been reported to reverse at -30 to -40 mV (270). At present, there is no clear explanation for this Erev value in terms of an associated chloride conductance. Data from two different labs (Hediger's and Stoffel's groups) are in apparent contradiction with the chloride channel activity associated with EAAT1 and EAAT3. L-Glutamate-induced current due to rat EAAT1 expression in oocytes does not reverse up to +80 mV (289), and that due to rabbit and human EAAT3 approaches asymptotically zero at +50 mV (246, 249). Is this a consequence of a differential behavior of different species counterparts (Erev for human EAAT1 is +9 mV, see Table 7; Ref. 582), or does it reflect differential experimental protocols? In the latter sense, it is worth mentioning that Hediger's group (246) did not show currents at depolarization potentials over +50 mV, and Amara and Kavanaugh and co-workers (582) measured an Erev of +38 mV for human EAAT3.

Although the mechanism of the uncoupled chloride conductance during the transport cycle remains unknown, there is compelling evidence for a chloride conductance in parallel with sodium/glutamate cotransport in photoreceptors and bipolar cells, where the ligands increase the rate of opening of the chloride channel (137, 138, 175, 176, 295, 471, 531). The retinal EAAT5 transporter, with a putative synaptic localization (see sect. IIC1), may be responsible for this chloride conductance. Pre- and postsynaptic glutamate-gated chloride conductances may have physiological roles in vertebrate retina. 1) The light response mediated by cones in depolarizing bipolar cells in the perch retina is due to the closing of a postsynaptic chloride conductance that has properties of the glutamate transporter (i.e., similar pharmacology and ionic dependence) (175). 2) Presynaptically, the activation of a chloride conductance concomitant with glutamate transport would provide a potential mechanism to offset the depolarizing action of transmitter reuptake and reduce cell excitability. Thus, in salamander cone photoreceptors, a glutamate-evoked chloride conductance, with properties similar to the glutamate transporters, responds as an inhibitory signal (hyperpolarization) to the release of glutamate from the same cell (421). The physiological relevance of the thermodynamically uncoupled chloride conductance of the glutamate transporters in several cell types (retinal and pituitary cells) has recently been reviewed by Sonders and Amara (507).

6. Protein structure

The main common structural features (Figs. 5 and 6) among the mammalian members of this family are as follows: 1) the absence of a cleavable signal sequence, suggesting a cytosolic localization of the NH2 terminus; 2) the absence of an SOB motif, identified in a variety of sodium/solute cotransporters; 3) the presence of the sequence motif AA(I,V,L)FIAQ, probably located in a membrane-spanning domain, which is conserved throughout the evolutionary diversity of glutamate transporters from prokaryotes to mammals, and also in the zwitterionic amino acid transporters of this family; 4) a higher level of conservation in the COOH-terminal half of the proteins which exceeds the level of conservation in the NH2-terminal half by a factor of at least three; 5) the presence of six highly conserved putative membrane-spanning domains in the NH2-terminal half of the proteins; 6) the presence of two cannonical sites for N-linked glycosylation on a presumably extracellular hydrophilic loop EL2 between TM domains III and IV; and 7) a similar appearance of EAAT1-3 glutamate transporters (there is no available data for EAAT4 or the zwiterionic transporters) as broad electrophoretic bands of similar size (65-75 kDa) due to variable glycosylation (121, 298, 305, 461, 481).

Ever since the initial description of the first three members of this superfamily (EAAT1-3), the topology of these transporters in the plasma membrane has been controversial. Based on topology algorithms, Stoffel's lab (214, 518) for EAAT1 and ASCT1, Kanner's group (424) for EAAT2, and Hediger's lab (245) for EAAT3 agree on the presence of six classical alpha -helix TM domains in the first NH2-terminal part of these proteins (see Fig. 5). Controversy appears in the COOH-terminal part, and this is an important issue since homology in this part of these proteins is very high (see Fig. 5) and several amino acid residues critical for transport activity have been described within this region (see sect. IIC7). There is a long hydrophobic stretch of amino acid residues with no clear tendency to show alpha -helix structures toward this end in any of the members of this superfamily (dashed line in Fig. 5). Kanner's group (424) suggested the presence of two additional classical TM domains within this protein region for EAAT2, whereas Kanai and Hediger (245) included four additional classical TM domains. In contrast, Stoffel and co-workers (214, 518) suggested six classical TM domains and four hydrophobic beta -sheets crossing the plasma membrane. In all cases, these models positioned the COOH terminus inside the cell. Very recently, two studies offered experimental evidence on the topology of EAAT1 (585) and the glutamate transporter GltT from Bacillus stearothermophilus, a prokaryote-related member of this superfamily (498). After these studies, the controversy remains. Stoffel and co-workers (214, 518) applied "reporter glycosylation scanning" (i.e., chimeras containing EAAT1 domains and an N-glycosylated reporter peptide), expressing chimeras in oocytes, to support a model for EAAT1 (GLAST-1) with 10 TM domains (see Fig. 6B). This model has NH2 and COOH termini intracellular, six NH2-terminal hydrophobic TM alpha -helices, and four COOH-terminal short hybdrophobic domains spanning the membrane bilayer as beta -sheets. The six NH2-terminal hydrophobic TM alpha -helices correspond to those suggested for all these transporters. Site-directed antibodies used in immunofluorescence studies with permeabilized cells confirmed the intracellular location of the NH2 terminus of EAAT1 (585) and the COOH terminus of rat EAAT1 and EAAT2 (298), suggesting an even number of TM domains. Slotboom et al. (498) used alkaline phosphatase (PhoA) gene fusion technique (i.e., scanning chimeras containing GltT domains and alkaline phosphatase) to study the controversial COOH-terminal part of the prokaryotic GltT transporter. Extrapolation of their results to the eukaryotic members of the superfamily is warranted by the fact that all these transporters showed a very similar hydropathy profile in the COOH-terminal half of the protein (i.e., fragment comprised between amino acid residues 400 and 550 of the multialignment of EEAT1-4 isoforms and GltT from E. coli, Bacillus subtillis, and B. stearothermophilus) (498). The GltT topology model proposed the presence of four additional TM alpha -helices in the COOH-terminal half of the protein (see Fig. 6A) (498).

Both strategies have been used to study the membrane topology of several proteins and are considered a good technical standard. In our view, however, both studies (498, 585) lack clarity, contain inconsistent data, and in addition used an objective experimental strategy to favor previous subjective topology models. Konings' group (498) based their model on the expression of the PhoA activity toward the periplasmic space if the particular chimera positioned PhoA extracellularly. The expression of low PhoA activity is interpreted as an intracellular location of the Phoa domain in the chimeras. In all cases, they demonstrate that low PhoA is not caused by a low expression level of the particular chimera, but they do not attempt to demonstrate that these chimeras are expressed in the plasma membrane and not as inclusion bodies. In addition, the model proposed is very rigid and needs three very small loops (EL4, IL4, and EL5) which is difficult to apply to mammalian members of this superfamily because of the presence of charged residues at the extremes of TM domains VII, IX, and X (see Fig. 6A). Stoffel's model is based on very clear data for the first NH2-terminal part of the protein (TM domains I-VI); all the chimeras constructed showed glycosylation of the reporter protein domains (they used an endogenous N-glycosylated domain of EAAT1 that corresponds to a large portion of the EL2 loop) when connected to loops EL1, EL2, and EL3. Conversely, the reporter protein domain is not glycosylated when located in loops IL1, IL2, and IL3 (see Fig. 6, A and B). This part of the model is confirmed by the following evidence: 1) the NH2 terminus is intracellular since EAAT1-specific antibody immunofluorescence signal is only obtained with permeabilized cells (298, 585). 2) For EAAT1, Stoffel's group showed by peptide sequencing, endoglycosidase F treatment, and site-directed mutagenesis that Asn-206 and Asn-216 residues are the only ones in the whole protein sequence that are N-glycosylated (116, 481); these residues are located in the extracellular loop EL2. 3) The Ser-113 residue of the glutamate transporter EAAT2 is phosphorylated in vivo by protein kinase C (83); this agrees with the intracellular location of the IL1 loop (see Fig. 6A).

In contrast, Stoffel's model of the topology for the COOH-terminal part of EAAT1 (585) is based on data that appear to be inconsistent. The "reporter glycosylation scanning" data obtained with chimeras constructed with residues located in the proposed intracellular loops IL3 and IL4 and the COOH-terminal domain are clearly consistent with the model, but those with residues located in the proposed extracellular loops EL4 and EL5 are controversial. 1) The latter chimeras produce only ~50% of the protein with the reporter domain glycosylated. In addition, the fusion protein of the reporter glycosylation domain at a residue located in the proposed EL4 loop is not glycosylated at all in the reporter domain when expressed in oocytes. To explain these results, the authors need to invoke reorientation or steric hindrance for the translocation of the COOH-terminal reporter domain to the lumen of the endoplasmic reticulum because of the moderate size and hydrophobicity of TM domains VII and IX acting as anchoring sequences (see Fig. 6B). 2) A new N-glycosylation site produced by site-directed mutagenesis in the center of the proposed extracellular loop EL5 is not glycosylated when expressed either in oocytes or in a translation system in vitro (585). 3) Three GltT-PhoA fusion proteins in amino acid residues within GltT protein regions that are homologous to the extracellular loops EL4 and EL5, proposed by Stoffel's group, gave rise to a very low periplasmic PhoA activity (498).

Finally, it is interesting to notice that the two groups gave differing interpretations to results that are consistent with each other. For instance, the higly conserved motif AA(I,V,L)FIAQ is placed in Stoffel's beta -sheet TM domain IX and in Konings' alpha -helix TM domain VII. This is based on 1) a nonglycosylated reporter domain and a low periplasmic PhoA activity when the reporter domain is fused to EAAT1 Glu-406 residue (498) or to its homologous residue in GltT (585), 2) a low periplasmic PhoA activity when the fusion involves the GltT residue corresponding to EAAT1 Ile-413, and 3) a partial glycosylated reporter domain and a high periplasmic PhoA activity when the reporter domain is fused to EAAT1 Gln-425 residue (498) or to the GltT residues that are homologous to the EAAT1 421 and 426 residues (585). This is used by Stoffel's group to propose a beta -sheet (residues 407-416 of EAAT1) and by Konings' group to propose an alpha -helix (corresponding to residues 410-427 in the EAAT1 sequence) spanning the plasma membrane, respectively. Stoffel's group argues that in their studies, most probably, there is no room for an alpha -helix between the EAAT1 residues 407 and 425. Konings' group argues that detailed studies with the lactose permease LacY and the melibiose carrier MelB from E. coli have demonstrated that the NH2-terminal half of an outgoing TM helix is sufficient to export the PhoA domain fused to a membrane protein, whereas the NH2-terminal half of an ingoing TM helix is sufficient to prevent the export of the PhoA moiety to the periplasm (77, 428).

It is patently clear that the topology of these transporters stands in need of further research. The two models are quite different and could be tested with alternative strategies. Studies with limited proteolysis and peptide-directed specific antibodies could be informative. Notice that the exposed loops in the COOH-terminal half of these transporters in Stoffel's model are very conspicuous, whereas in Konings' model they are very limited (see Fig. 6, A and B). Alternatively, vectorial labeling of cystine residues reintroduced in the borders of the proposed TM domains of a cystineless transporter may also help. The establisment of the membrane topology of the COOH-terminal half of these transporters is an important issue because of the high level of homology in this region for all the members of this superfamily, and because several studies have shown that residues within this region are critical for substrate binding or translocation (see sect. IID). Finally, to date, beta -sheet TM domains have been proposed for several eukaryotic and prokaryotic membrane proteins like the acetylcholine receptor, the VDAC ion channel, and the lac permease (6, 52, 438), but they have only been demonstrated by X-ray analysis of the bacterial porins (118).

7. Structure-function relationship

Our knowledge of the structure-function relationship is based on studies with glutamate transporters of this superfamily, using chimeric proteins (between the human homologs of EAAT1 and EAAT2; Ref. 571), site-directed mutagenesis (for rat EAAT1 and EAAT2 transporters; Refs. 83, 115, 116, 426, 630), on the conformational changes associated with the transport step (for rat EAAT2; Refs. 179, 266, 583), or on the homomultimerization of these transporters (for rat EAAT1-3; Ref. 199). Part of these studies has been recently reviewed (254).

In an elegant study, Kavanaugh, Amara, and co-workers (571) prepared a human EAAT1-2-1 chimera, in which 76 amino acid residues of EAAT2, comprising most of the highly conserved long hydrophobic stretch (see Figs. 5 and 6A) were exchanged within the EAAT1 sequence. This EAAT2 protein segment, in which only 18 amino acid residues are different in the two isoforms, corresponds to part of the IL3 loop, TM domain VII, EL4 loop, and most of TM domain VIII in the 10 alpha -helix TM domain model (498) (see Fig. 6A). This segment in the EAAT1-2-1 chimera, when expressed in oocytes, confers sensitivity to inhibition by the nontransported competitive analog KA to both glutamate transport (Ki in the micromolar range, characteristic of EAAT2 isoform) and to the uncoupled glutamate-independent sodium leak current, characteristic of EAAT1 isoform. Kinetic analyses are compatible with inhibition of both processes by binding of KA to a single site (571). Interestingly, other transport characteristics of EAAT1 isoform are unchanged in the EAAT1-2-1 chimera, like the apparent affinity for the substrate analog SOS (see Table 7) and the Erev of the glutamate- and sodium-induced current [Erev = ~10 mV for EAAT1-2-1 (461); compare with the Erev for EAAT1 and EAAT2 in Table 7]. This suggests that the kinetic parameters for substrate translocation and the uncoupled chloride channel activity are determined by the EAAT1-derived sequences.

Most of the amino acid residues critical for the transport function of EAAT1 and EAAT2 transporters revealed by site-directed mutagenesis are within or near this highly conserved COOH-terminal part of glutamate transporters, which confers sensitivity to KA. Conradt and Stoffel (115) analyzed the effect of substitution of three positively charged residues (Arg-122, Arg-280, and Arg-479) and one polar residue (Tyr-405) in rat EAAT1 transporter, which are conserved in the glutamate transporters and substituted by apolar residues in the zwitterionic transporters (see Fig. 5). Mutations Arg122Ile and Arg280Val (and both together) reduce the apparent affinity for L-aspartate without affecting the kinetic parameters for L-glutamate transport, and mutants Tyr405Phe and Arg479Thr, within the highly conserved COOH-terminal part of these transporters, completely abolished the intrinsic EAAT1 transport activity (see below) (115). Kanner and co-workers (426) analyzed the role in transport of five negatively charged residues of rat EAAT2 located in hydrophobic surroundings and highly conserved within the glutamate transporter family (Asp-398, Glu404, Glu461, Asp462, and Asp-470). Only three of these residues (Asp-398, Glu-404, and Asp-470; indicated in Fig. 6A) are critical for intrinsic transport activity, which could not be explained by protein expression level or defects in trafficking to the plasma membrane. Interestingly, defective transport cannot be attributed to the mere requirement of a negative charge at this residues (i.e., transport is also affected by substitution of the corresponding charged residue, either Glu or Asp) (426).

The rat EAAT2 Glu404Asp (this residue is located in the TM domain VII of the 10 alpha -helix TM domain model; see Fig. 6A) mutant has been revealed as a powerful tool to address structure-function studies. This mutant conserves most of D/L-aspartate transport (~80%), but only a small part of L-glutamate transport (<20%) (426). The defective Glu404Asp L-glutamate transport is not because of defective binding (i.e., high-affinity L-glutamate inhibition of D/L-asparatate transport is conserved). This allows the authors to propose that the Glu-404 (conserved in all the glutamate transporter isoforms but absent from the zwitterionic transporters of this transporter family; see Fig. 5) determines the amino acid substrate permeation pathway of the glutamate transporters. The Glu-404 residue in EAAT2 together with residues Arg-122 and Arg-280 within the NH2-terminal part of EAAT1 (see above) are those already identified, which are involved in substrate specificity discrimination (115, 426). Very recent data obtained from collaboration between Kanner's and Kavanaugh's groups (266) showed that Glu-404 residue also influences the potassium transport coupling (either binding or translocation), and the rat mutant Glu404Asp EAAT2 catalyzes obligatory exchange of coupled amino acid substrate and sodium through the plasma membrane. 1) The sodium/D-aspartate transport via Glu404Asp EAAT2 is electroneutral in oocytes, 2) external potassium does not reverse transport through Glu404Asp in oocytes, and 3) in the liposome, reconstituted mutant influx and efflux of radiolabeled D-aspartate are dependent on trans-sodium/amino acid substrate but not on trans-potassium. In contrast, wild-type EAAT2 catalyzes trans-potassium-dependent influx and efflux of amino acid substrate in the presence of sodium. This is a consequence of the countertransport of potassium in the transport mechanism of these glutamate transporters (255, 529; see sect. IIC4). Because the Glu404Asp mutant is locked in an exchange mode of transport, either potassium binding or permeation, or sodium binding is affected (i.e., a significant increase in sodium binding will displace potassium binding and force the transporter toward amino acid/sodium exchange) (266). The former possibility seems to be true because apparent sodium affinity is unchanged in the Glu404Asp mutant. From all this, it is not surprising that the sodium-dependent transient currents produced by voltage jumps in human and rat EAAT2 expressing oocytes (266, 583) are hardly affected by the mutant (266). These transient currents are thought to be a reflection of either sodium binding or a subsequent conformational change of the transporter. In agreement with the EAAT1-2-1 chimera studies discussed above (571), the Glu404Asp mutant, located within the KA-binding/sensitive determining domain, does not affect the uncoupled amino acid substrate/sodium-induced chloride channel activity of the transporter (266), suggesting that this protein region does not influence this channel activity. It is remarkable that Glu-404 residue is in between two other conserved residues in all glutamate transporters of the family, comprising the sequence Tyr-Glu-Ala (see Fig. 5). Interestingly, the Glu404Asp homologous mutation in human EAAT3 also abolishes potassium-dependent efflux (266). In contrast, the zwitterionic amino acid transporters of this family have the conserved sequence Phe-Gln-Cys (see Fig. 5). Interestingly, mutation of this conserved Tyr residue to Phe (as in the zwitterionic transporters of this family) in rat EAAT1 (Tyr405Phe) abolished the intrinsic glutamate transport activity (115), and in rat EAAT2 (Tyr403Phe) abolished interaction with potassium, and resulted in an increased sodium affinity (629a). Very recently, Zerangue and Kavanaugh (627) offered evidence that ASCT1 transporter has an electroneutral exchange mode of transport for the amino acid substrate and sodium through the membrane; this mechanism of transport has also been suggested for ATBo (162). It is therefore tempting to speculate that Glu-404 within these residues (located in the VII TM domain in the 10 alpha -helix topology model, see Fig. 6A) confers coupled cotransport of sodium and countertransport of potassium, whereas its lack determines an exchange mode of transport coupled with sodium. Unfortunately, for human ASCT2 transporter, which also contains the conserved sequence Phe-Gln-Cys, the mechanism of transport and potassium dependence has not been addressed in depth. In summary, the hydrophobic, topologically controversial, and highly conserved domain located toward the COOH terminus of the glutamate transporters is involved in kainate binding and amino acid and ion (potassium coupling) permeation pathways.

Several residues within the NH2-terminal part of these transporters have been shown to be involved in their transport activity or expression (83, 116, 630), in addition to the above-mentioned EAAT1 Arg-122 and Arg-280 residues (115). Stoffel and co-workers (116) demonstrated that the deglycosylated rat EAAT1 (N-glycosylation occurs in 2 canonical sites within the loop EL2; see Fig. 6A) is fully active, and none of its kinetic parameters is affected. Kanner and co-workers (630) examined the effect of substitution of the only two positively charged residues (Lys-298 and His-326 in EAAT2; see Figs. 5 and 6A) conserved in all members of this transporter family and located within putative alpha -helix TM domains (TM domains V and VI in the 10 alpha -helix TM domain topology model; see Fig. 6A). Replacement of these residues by small hydrophilic or positively charged amino acids produces in Lys-298 mutants a partial plasma membrane targeting defect and partial intrinsic transport defect of EAAT2; His-326 mutants have an almost complete impairment of their intrinsic transport activity without a trafficking defect toward the plasma membrane (630). Zhang et al. (630) suggested two possible roles for the conserved His-326 residue. In analogy with structure-function studies of the proton-coupled lactose permease of E. coli, His-326 could either form ion pairs with negatively charged residues within TM domains that stabilize the transporter (133, 283, 466) or participate in the mechanism of hydrogen transport (241, 406). The same group (426) examined the first possibility by constructing double mutants with three conserved negatively charged residues that are critical for the transport activity of EAAT2 (Asp-398, Glu-404, Asp-470; see above). None of the double mutants (i.e., His326Asn with Asp398Asn, Glu404Asn, or Asp470Asn) regained activity, and therefore, there is no evidence for these ionic pairs within EAAT2 transporter.

A very interesting line of research is the stimulation of EAAT2 by protein kinase C. In loop IL1 of EAAT1-4 glutamate transporters isoforms, there is a protein kinase C canonical site (see Fig. 6A). Giménez and co-workers (84) showed that phorbol esters increased Vmax of sodium-glutamate cotransport in cultured glial cells. Later, these authors in collaboration with Kanner's group (83) demonstrated the following: 1) protein kinase C phosphorylates, in serine residues, pig brain purified glutamate transporter; 2) phorbol esters increase in parallel glutamate transport activity (2-fold) and phosphorylation of immunoprecipitated EAAT2 in C6 glial cells; and 3) rat EAAT2 transfected in HeLa cells is stimulated by phorbol esters, and mutation of Ser-113 to Asn abolished this stimulation without affecting transport activity expression. This is the first direct demonstration of regulation of a neurotransmitter or amino acid transporter by phosphorylation. The nature of the upstream event that stimulates glutamate transport via EAAT2 through protein kinase C is at present unknown. The authors hypothesize that elevation of the extracellular glutamate concentration would stimulate NMDA receptors in astrocytic processes, resulting in activation of the phosphatidylinositol cycle and protein kinase C activation, and therefore in a more efficient clearance of the extracellular glutamate. To our knowledge, neither this hypothesis nor the mechanism of EAAT2 stimulation has been addressed experimentally.

Conformational changes of EAAT2 have been revealed through its transport cycle (179, 266, 583). Sodium-dependent transient currents of the expressed EAAT2 transporter suggested conformational changes associated with binding of sodium (266, 583). More directly, limited proteolysis studies demonstrated conformational changes of purified EAAT2 associated with glutamate and sodium, or potassium binding; these studies suggest that EAAT2 transporter has at least two conformation states and that the transition between them is associated with the transport step (179).

Finally, EAAT1-3 glutamate transporters form homomultimers (dimers and trimers), as revealed by chemical cross-linking in intact brain membranes and solubilized transporters, or after reconstitution in liposomes (199). The original EAAT2 purification studies by Kanner and co-workers (120, 121) revealed that the monomeric 73-kDa band of the transporter correlated with glutamate transport activity. In addition, it is interesting that the fully deglycosylated EAAT1, obtained either by deletion of the two glycosylation sites or after endoglycosidase F treatment, does not homodimerize in electrophoretic gels, and it is fully active (116). These data suggest that dimerization of glutamate transporters does not affect their transport activity. In contrast, radiation inactivation studies suggest that the minimal funtional unit corresponds to an oligomer of the rat EAAT2 transporter (199). It is therefore clear that further research is needed on this issue.

8. Physiological role of the glutamate superfamily transporters

It is believed that the transporters in this superfamily have a role both in the termination of transcription in the synapsis and also in the supply of nutrients to brain and peripheral tissues (205, 244, 247, 252, 269, 568, 627). The overall process of synaptic transmission, except for acetylcholine, is terminated by high-affinity sodium-dependent transport of neurotransmitters (e.g., GABA, L-glutamate, glycine, dopamine, serotonin, and norepinephrine; see reviews in Refs. 205, 252, 253, 394). The concentration of L-glutamate, the predominant excitatory neurotransmitter of the mammalian central nervous system, is typically four orders of magnitude higher in the nerve terminals than in the cleft (estimations of 10 mM in neurons, low millimolar range in glial cells and submicromolar to low micromolar range in the glial extracellular fluid; Refs. 40, 74, 262, 458); therefore, energy input is required. The Na+-K+-ATPase generates an inwardly directed electrochemichal sodium gradient that drives uphill the the sodium- and potassium-coupled glutamate transport in neurons and glial cells (210, 262, 394). This role of glial and neuronal glutamate transport, in maintaining a low extracellular neurotransmitter concentration (<1 µM), has been postulated to be critical to protect against exocitotoxic cell damage (63). Glutamate transport blockers, which are nonselective for isoforms and the different transport activities detected in brain (454), raise extracellular glutamate, alter postsynaptic potentials, and result in neurotoxicity both in vitro (33, 226, 359, 453, 460, 470) and in vivo (326, 402, 403). This effect is blocked by non-NMDA glutamate receptor antagonists (459, 460) and is most probably because of the excessive calcium influx through NMDA receptor channels (for review, see Refs. 205, 252, 394).

Knockout studies of EAAT glutamate transporter isoforms (416, 458, 544) revealed that the glial transporters (EAAT2 and EAAT1) rather than the neuronal transporter EAAT3 control the extracellular glutamate levels in brain. Very recently, studies on the null knockout mice for EAAT2 and EAAT3 have been published (416, 544). To our knowledge, the null knockout EAAT1 mouse has not been reported. Tanaka et al. (544) studied the knockout of the widely distributed astrocytic glutamate transporter EAAT2 (also named GLT-1). These mice show lethal spontaneous epileptic seizures, selective neuronal degeneration in hippocampus, and increased susceptibility to acute cortical injury. In these mice, the estimated peak concentration and time course of free glutamate in the synaptic cleft is elevated. This indicates that glial EAAT2 is an important determinant of the clearance of free glutamate from the synaptic cleft. Thus, in the absence of EAAT2 transport activity, glutamate levels rise enough to cause epilepsy and cell death. In contrast to this, null knockout EAAT3 mice, obtained by Stoffel and co-workers (416) show, in addition to the renal phenotype (see below), a nonconspicuous brain phenotype, only characterized by reduced locomotor activity. Rothstein and Kuncl (458) addressed the contribution of the three EAAT1-3 isoforms described in rat to the maintenance of global extracellular glutamate concentrations in the cerebrospinal fluid, as well as the histological and behavioral consequences of their specific partial knockouts (in vitro and in vivo intraventricular phosphorothioate antisense administration). At present, the cerebellar EAAT4 isoform, described in humans (63) and suspected to maintain extracellular glutamate concentrations below excitotoxic levels, has not been isolated from rat tissues. The partial knockout of EAAT1-3 isoforms (458) showed that both glial transporters (EAAT1-2) contribute largely to the maintenance of the tonic cerebrospinal glutamate concentration and that the impact of the EAAT2 isoform was more conspicuous. In contrast, the contribution of the neuronal EAAT3 is negligible. In parallel, the EAAT isoform-specific partial knockout showed that glial glutamate transport sites (EAAT1-2) are more conspicuous than the EAAT3 transport sites (binding of radiolabeled D-aspartate inhibitable by DL-threo-beta -hydroxyaspartate to membranes) in the two structures studied, striatum and hippocampus. This is in agreement with greater expression of these transporters in comparison with the neuronal isoform. The EAAT2 isoform is present in astrocytes throughout the brain and spinal cord (121, 305, 461). In comparison with EAAT isoforms 1-3, partial knockout of EAAT2 resulted in the largest decrease in the glutamate transporter sites in striatum and hippocampus (458), purification through functional reconstitution from rat brain resulted in the identification of EAAT2 (120, 121, 424), and immunoprecipitation studies suggest that EAAT2 isoform is the most prevalent glutamate transporter in brain (199). The brain phenotype of knockout EAAT2 mice and the very low residual glutamate transport activity in cortical crude synaptosomes from these mice (544) have confirmed this suggestion. In agreement with this, in the sporadic form of amyotrophic lateral sclerosis (ALS), there is a specific marked reduction (up to 95%) of the expression of EAAT2 in the motor cortex and the spinal cord (463). In parallel, there is also a marked decrease in the Vmax of high-affinity glutamate uptake in synaptosomes from those brain structures and an increased cerebrospinal fluid concentration of L-glutamate and L-aspartate in ALS patients (462) (see sect. III).

The above-mentioned knockout studies showed that the glial transporters (EAAT1-2) are the more conspicuous transporters in brain, and they have a crucial role in the maintenance of the tonic extracellular glutamate concentration. Thus the tonic increase in extracellular glutamate because of EAAT1-2-specific partial knockouts explains the progressive paralysis and neurodegeneration in these rats (458). As discussed by Rothstein and Kuncl (458), glial cells have a considerably lower estimated intracellular glutamate concentration (in the micromolar range) than neurons, which suggests that EAAT1-2 transporters operate far from equilibrium, most probably due to the rapid metabolization to glutamine by glutamine synthetase, which is absent in neurons. Therefore, in addition to its larger expression, the operation of EAAT1-2 far from equilibrium may explain why the phenotype obtained after total or partial knockout of EAAT1-2 is clearer than that given by knockout of the EAAT3 isoform (416, 458, 544). It is worth mentioning that the proposed lack of role for the EAAT3 transporter in the tonic extracellular glutamate levels (458) does not imply that this transporter has no role in excitotoxicity. Reversal of glutamate transport has been proposed as a mechanism of excitotoxicity under conditions of energy failure, as in cerebral ischemia (hypoxia, stroke; Refs. 23, 246, 394). The nonlimiting transport flux via EAAT3 running in reverse could produce a significant local increase in the extracellular concentration of glutamate (i.e., >350 µM as demonstrated in salamander retinal glia cells and EAAT3 expressed in oocytes; Refs. 63, 246).

Null knockout EAAT2 mice (544) confirmed the hypothesis (394) that astroglial glutamate transporters contribute to the reuptake/termination of the glutamate synaptic transmission. Thus total loss of EAAT2 transport activity (i.e., as in the homozygous null knockout mice), but not its partial loss (i.e., as in the heterozygous null knockout mice or in chronic antisense administration to rat brain), produces epilepsy (458, 544) and increases the time course of free glutamate in the synaptic cleft (544). On the other hand, Amara and Kavanaugh and co-workers (583) and Kanner and co-workers (120) estimated a transport cycling time of 70-700 ms for human EAAT2 expressed in oocytes and rat purified EAAT2. This is significantly slower than the 1- to 2-ms time constant of glutamate decay estimated in hippocampal synapses (104, 112). This suggests that glutamate diffusion and "fast" binding to EAAT2 transporters (583) dominates the synaptic concentration decay kinetics.

What is the role of the neuronal EAAT3 glutamate transporter in the termination of synapsis? In contrast to the glial glutamate transporters, the partial knockout of EAAT3 protein produced no changes in extracellular glutamate and only mild neurotoxicity and motor phenotype, but consistent epileptic seizures (458). It is believed, although this is not completely clear (628), that the neuronal glutamate carrier EAAT3 operates at or near equilibrium, and its expression is confined to pre- and postsynaptic elements (461). It is somehow expected that the partial reduction of a plasma membrane transport activity, which is working at equilibrium (i.e., flux through this transporter does not limit the overall metabolic handling of the neuronal glutamate) and confined to the synapsis, has little or no impact in the global extracellular glutamate concentration, as the partial knockout studies showed (458). The epileptic phenotype of the partial knockout of EAAT3 suggests that a moderate rise in the intrasynaptic glutamate concentration, without global concentration changes, may cause persistent depolarization or alteration of the presynaptic transmitter release (458). In addition to glutamatergic neurons, EAAT3 has also been located in inhibitory GABAergic neurons, and because glutamate is a precursor for GABA synthesis, transport via EAAT3 could have a role in GABA neurotransmission (205, 461). Superstimulation of excitatory glutamatergic neurons and blockade of inhibitory GABAergic neurons are known to produce epilepsy (157). Unfortunately, the null knockout EAAT3 mouse model only reproduces the locomotor, but not the epileptic, phenotype (416) of the antisense-depleted EAAT3 rat model (458). It seems that overexpression of the glial glutamate transporters does not occur in the knockout EAAT3 mice (517). These apparently contradictory results raise doubts as to the contribution of the EAAT3 transporter to the termination of glutamate synapsis.

Expression of the glutamate transporter EAAT3 in GABAergic neurons (205, 461), its strong transcript expression in the small intestine, and at a lower levels in kidney, liver, and heart (245), suggest a metabolic role for this transporter. In epithelial cells, system X -AG has been described mainly in the apical plasma membrane (483, 513), and therefore, it is believed that EAAT3 mediates net absorption of glutamate and aspartate in kidney and intestine (205). This role is demonstrated by the dicarboxylic aminoaciduria developed by null knockout EAAT3 mice (416). In addition, this suggests that mutations in EAAT3 may cause dicarboxylic aminoaciduria, an inherited disease due to defective glutamate transport in kidney and intestine (see sect. III). Recent results showed that system X -AG transport activity is increased (Vmax effect) by hypertonic stress in the bovine renal cell line NBL-1 (148). Concomitantly, the EAAT3 transcript levels increase, suggesting that this glutamate isoform is responsible for system X -AG in these cells, and indicating a direct effect of hypertonic stress in the expression of this transporter isoform (148). Whether hypertonic stress increases EAAT3 gene transcription and/or mRNA stability in these cells has not been reported. This regulation of EAAT3 might be due to a role of this glutamate transporter in glutamine metabolism and pH regulation in renal cells.

The physiological role of ASCT1-2 and ATBo zwitterionic transporters is at present unclear. It is necessary to clarify whether mouse ASCT2 corresponds to human ATBo, or whether they code for different transport activities. In addition, a more precise description of the mechanism of transport for these transporters is needed. If, finally, ASCT transporters mediate concentrative uptake of their substrate coupled with the transmembrane gradient of sodium and amino acids, ASCT1-2 might be assigned as variants of the almost ubiquitious ASC system. It will be also necessary to explain the molecular basis of the hepatic ASC system, which as discussed above does not appear to be represented by either one of these ASCT isoforms. Studies with anti-ASCT1-2 antibodies and knockout experiments will be needed to estimate the role of these transporters in the macroscopic flux of amino acids in cells expressing them.

The ATBo transporter (269) might correspond to system Bo, the most conspicuous sodium-dependent uptake system for zwitterionic amino acids. This apical transport system is thought to play a major role in reabsorption of zwitterionic amino acids in kidney and small intestine (see Refs. 483, 513). Elucidation of the transport mechanism and demonstration of the apical localization of ATBo in epithelial cells may reaffirm the assignation of ATBo as system Bo transporter. Finally, demonstration of the responsibility of ATBo in Hartnup disease, an inherited neutral hyperaminoaciduria (see sect. III), may confirm that ATBo plays a role in amino acid nutrition and renal reabsorption and system Bo activity. In contrast to this view, the recent description of an amino acid exchange mechanism of transport for ATBo (162) questions the participation of this transporter in the active renal and intestinal absorption of neutral amino acids and its role in Hartnup disease.

D. Putative Subunits of Sodium-Independent Cationic and Zwitterionic Amino Acid Transporters

The last protein family related to plasma membrane amino acid transport in mammals is composed by the protein rBAT and the heavy chain of the cell surface antigen 4F2 (4F2hc) (see Table 8). Amino acid transport expression in Xenopus oocytes was used independently in three labs to clone cDNA of a putative transporter from rabbit, rat, and human kidney; homology between these proteins is very high (~85% identity) (45, 46, 110, 297, 549, 598). A partial rBAT cDNA sequence from OK cells has also been reported (374). The three labs gave different names to these cDNA: NBAT (Udenfriend and Tate's group), D2 (Hediger's group), and rBAT (ourselves). For clarity, the name rBAT will be used for all these cDNA and proteins in this review. The cDNA of human 4F2hc was cloned using a monoclonal antibody designed against a cell surface antigen from lymphoblastoid cells (327, 435, 553), and its mouse counterpart was identified by homology (413). The biological role of this antigen was unknown. The deduced rBAT protein amino acid sequences have ~30% identity (~50% similarity) with the heavy chain of the cell surface antigens 4F2 (4F2hc) (69, 413, 435, 553). Figure 7 shows the sequence homology between the human rBAT and 4F2hc proteins, and in Figure 8, the amino acid residues conserved in all rBAT and 4F2hc proteins known are indicated. Consistent with rBAT and 4F2hc being members of the same family, within the open reading frame of human rBAT and 4F2hc, introns 1 and 2 have identical locations, intron 3 in 4F2hc corresponds to intron 4 in rBAT, and intron 8 in 4F2hc to intron 9 in rBAT (174, 434, 431) (see Fig. 8). Given the homology between rBAT and 4F2hc, cRNA from 4F2hc was tested in oocytes for expression of amino acid transport activity. Expression of 4F2hc resulted in an amino acid transport activity (system y+L- like) different from that elicited by rBAT (system bo,+-like) (42, 599) (see Table 9). Interestingly, expression cloning in oocytes after a zwitterionic amino acid transport signal (68) resulted in the isolation of rat 4F2hc (named lLAT in this study for linked to L amino acid transport) (69). It is worth mentioning that rat and mouse 4F2hc proteins are very similar (91% amino acid sequence identity), whereas the human protein is only 76% identical to the rat and mouse proteins (69). More recently, using an antibody that induces apoptosis in hematopoietic progenitor cells and homotypic aggregation of lymphoid progenitor cells as a screening tool in transiently transfected COS-1 cells, the mouse 4F2hc was cloned again (592). As discussed in section IID5, 4F2hc might have multiple functions.

 
View this table:
[in this window] [in a new window]
  		TABLE 8.   Putative subunits of sodium-independent cationic and zwitterionic amino acid transporters


View larger version (94K):
[in this window]
[in a new window]
  		FIG. 7.   Amino acid sequence comparison of human rBAT and 4F2hc proteins. Thick horizontal line over sequences indicates hydrophobic segment that corresponds to first putative TM domain, whereas thin horizontal lines indicate amphipathic TM domains II-IV proposed by Tate and co-workers (378). Solid frame box in gray indicates residues that resemble catalytic site of homologous glycosidases. Here, arrows indicate position of proposed catalytic residue (aspartate or glutamate) of these glucosidases; this residue is substituted by arginine in human 4F2hc and by asparagine in rabbit rBAT (42). In gray boxes are indicated amino acid residues present in human rBAT and 4F2hc proteins. Four dash frame boxes indicate segments of 12-17 amino acid residues with high homology between rBAT and 4F2hc proteins. Solid frame boxes indicate potential N-glycosylation sites, 6 for rBAT and 2 for 4F2hc. Dash indicates gaps for sequence multialignment obtained with all known rBAT and 4F2hc sequences with Clustal Sequence Alignment from Baylor College of Medicine.


View larger version (75K):
[in this window]
[in a new window]
  		FIG. 8.   Cystinuria-specific mutants in 4 TM domain topology model of human rBAT protein. This model has been proposed by Tate and co-workers (378) (see text for details). Amino acid residues conserved in all rBAT proteins (including OK cell rBAT protein fragment corresponding to human residues 373-593) are indicated in gray circles and those present also in 4F2hc proteins in white letters on a black circle. Notice that cysteine residue 114, after first TM domain, is conserved in all these proteins. Twenty-one cystinuria-specific mutations are indicated by arrows. Number shows amino acid residue involved. For 15 missense mutations, substitution amino acid is indicated. Most of these mutations occur in conserved amino acid residues, with exception of R181Q, T652R, and F648S. ns, fs, and sm denote nonsense, frame shift, and splice mutations, respectively. In addition, four large deletion mutations have been described with the following approximate boundaries: 114-1306, 198-1575, 430-765, and a mutation affecting exons V to X; position 1 corresponds to 5'-position of first ATG codon. All mutations referred to here have been reviewed in References 170 and 408. Four of these mutations (E268K, T341A, M467T, and M467K; amino acid residues indicated by an square) have been analyzed in oocytes and show defective transport expression (78, 91, 366). Six potential N-glycosylation sites (Y) are indicated in first and second putative extracellular loops. Drawing of extracellular and intracellular loops and NH2 and COOH terminals do not indicate any type of structure. Exon-intron boundaries of human rBAT are shown as reported in References 434 and 431. Boundaries conserved in open reading frame of human 4F2hc (174) are indicated by an asterisk.

 
View this table:
[in this window] [in a new window]
  		TABLE 9.   Tissue distribution and transport characteristics of expressed rBAT and 4F2hc proteins

The relevance of rBAT in the reabsorption of cystine and dibasic amino acids in kidney and intestine has been demonstrated by the involvement of the rBAT gene (named SLC3A1 in Gene Data Bank) in cystinuria (for recent reviews, see Refs. 170, 408, 467). This is an inherited aminoaciduria due to defective renal and intestinal reabsorption of cystine and dibasic amino acids; the poor solubility of cystine causes the formation of renal cystine calculi (351, 487). Surprisingly, rBAT and also 4F2hc are not very hydrophobic, and they seem to be unable to provide an aqueous translocation pathway in the plasma membrane. This prompted the hypothesis that rBAT and 4F2hc are subunits or modulators of the corresponding amino acid transporters. In this sense, it has been suggested or demonstrated that there is an association of rBAT and 4F2hc, respectively, with a corresponding light subunit of ~40 kDa. Here attention is focused on the hypothesis that both rBAT and 4F2hc are subunits of the actual amino acid transporters corresponding to system bo,+-like and y+L-like. Structural and functional evidence in favor of this is discussed. The role of the rBAT gene in cystinuria is described in section III.

1. Tissue expression

The rBAT mRNA is expressed in the kidney and the mucosa of the small intestine (45-47, 297, 598, 617). Consistent with this, hybrid depletion with rBAT antisense oligonucleotides blocks expression of system bo,+-like by renal and intestinal poly(A)+ RNA in oocytes (45, 338, 598). Northern blot analysis of human, rat, and rabbit renal and intestinal RNA revealed two rBAT transcripts: ~2.3 kb (which corresponds to the above-mentioned cDNA) and ~4 kb. A cDNA corresponding to the long rBAT transcript was identified by expression cloning in oocytes and represents an alternative polyadenylation of the same gene (344). In situ hybridization and immunolocalization studies have demonstrated that rBAT localizes to the microvilli of the small intestinal mucosa and the epithelial cells of the proximal straight tubules of the nephron (159, 248, 422). Interestingly, the expression of rBAT is developmentally regulated in rat kidney; rBAT transcripts appear after birth, and the onset of the protein expression coincides with postnatal nephron maturation (159). Clear rBAT transcripts are also visible in human pancreas; the significance of rBAT expression in pancreas is unknown (45). In addition to kidney and intestine, brain tissues show a transcript of ~5 kb that hybridizes with rBAT cDNA probes (45, 46, 617). This long transcript is almost ubiquitous, but with a substantially lower abundance in tissues other than brain (45). The RNA protection assay studies and Western blot analysis with some but not all anti-rBAT peptide antibodies suggested that this long transcript corresponds to the expression of a gene that is homologous to rBAT (422, 617). Moreover, rBAT immunoreactivity in hypothalamus is intracellular, and it is not located in the plasma membrane as in kidney and intestine (212). One antibody directed against a peptide of the rBAT sequence labeled intracellular structures of magnocellular neurons of the supraoptic and paraventricular nuclei.

In contrast to rBAT, 4F2hc mRNA is almost ubiquitous in mouse tissues, with a higher expression level in testis, lung, kidney, brain, and spleen and without a clear pattern of developmental regulation (413). Studies previous even to the cloning of 4F2hc showed that this protein is induced after activation of human and mouse lymphocytes (reviewed in Ref. 413; see sect. IID5). In fibroblasts (NIH 3T3 and BALB/c 3T3 cells), 4F2hc expression is induced during cell activation and maintained high throughout the cell cycle in exponentially growing cells (413). This suggests that 4F2hc plays a role in proliferating and quiescent cells. The amino acid transport activities associated with 4F2hc, as described here, may be relevant for both situations.

2. Transport properties

The characteristics of the amino acid transport activity associated with rBAT and 4F2hc expression have been studied mainly in oocytes. rBAT induces, through the oocyte plasma membrane, transport of cystine (up to >100-fold over background) and dibasic and zwitterionic amino acids (up to 50-fold over background). This is a high-affinity transport with Km values in the micromolar range for amino acids such as L-cystine, L-arginine, L-lysine, L-ornithine, L-leucine, and L-histidine. Kinetic and cross-inhibition studies offered convincing evidence that rBAT induces a single amino acid transport system in Xenopus oocytes, at least in sodium-free medium (see below) (46), which is not present in stage VI oocytes (42, 46, 354). This transport activity is sodium independent, and it is very similar to the amino acid transport system bo,+ defined by Van Winkle et al. (577) in mouse blastocysts, as a sodium-independent high-affinity system for dibasic and zwitterionic amino acids. In contrast to the transport system associated with rBAT, the blastocyst bo,+ system does not transport L-cystine (L. J. Van Winkle, personal communication). For this reason, we named our human and rabbit cDNA clones rBAT, as an acronym for "related to bo,+ amino acid transporter"). Further characterization of the rBAT/system bo,+-like transport activity showed that it was independent of external potassium and chloride (75), changes in the external pH (Palacín, unpublished data), and internal ATP (110). Cystine and dibasic and zwitterionic amino acid transport with the characteristics of system bo,+-like have been described in renal and intestinal plasma membrane preparations (see sect. IID5).

Ahmed et al. (4) and Taylor and co-workers (420) propose that the expression of rBAT induces several amino acid transport systems: 1) a NEM-resistant sodium-independent transport for cationic and zwitterionic amino acids, equivalent to system bo,+-like (R. Estévez and M. Palacín, unpublished data), and in brush-border membrane preparations of chicken jejunum (560); 2) sodium-independent transport activities (perhaps two), which are sensitive to NEM treatment, and with overlapping specificties for cationic and zwitterionic amino acids; and 3) a sodium-dependent transport for L-histidine, which has a pH dependence compatible with the transport of this substrate in the nonprotonated form. Then, either as a consequence of the overexpression of rBAT in oocytes or reflecting a true mechanism of activation, rBAT induces several amino acid transport activities in the oocyte (see sect. IID4). The fact that partial knockout of rBAT in OK cells results in a specific, partial decrease in the apical system bo,+-like activity (374), and the finding that mutations in the rBAT gene cause cystinuria (for review, see Refs. 170, 408, 467) demonstrate the role of rBAT in the high-affinity reabsorption system of cystine (system bo,+-like) (see sect. IID5). In this context, the physiological relevance of the other amino acid transport activities induced in rBAT-expressing oocytes remains to be demonstrated.

In contrast to rBAT, the cRNA of human 4F2hc induces an amino acid transport activity (e.g., up to 10-fold over background for radiolabeled L-arginine), which is sodium independent with high affinity (micromolar range) for L-dibasic amino acids, but with high affinity for L-zwitterionic amino acids only in the presence of sodium; in the absence of sodium, the affinity for L-zwitterionic amino acids is dramatically reduced (46, 599). This transport activity, which does not transport L-cystine, is very similar to the system y+L, initially described in human erythrocytes by Devés et al. (127), and later described in brush-border membrane vesicles from human placenta (135); a recent review (126) describes the transport characteristics of system y+L. In the same line, Ganapathy and co-workers (146) have shown that poly(A)+ RNA from a human choriocarcinoma cell line expresses y+L transport activity in oocytes that is hybrid depleted by 4F2hc antisense oligonucleotides. Very recently, similar data have been obtained with rat lung poly(A)+ RNA (142). In contrast to this, Broër et al. (69) showed a clear induction by rat 4F2hc in oocytes of a high-affinity uptake for cationic and zwitterionic amino acids, both in the absence of sodium. The induced transport activity has combined characteristics of system bo,+-like and L-like, but not of system y+L-like (see sect. IID4).

The fact that mutations in the rBAT gene cause cystinuria (see sect. III), a defect in the renal and intestinal reabsorption of cystine and dibasic amino acids, raised an important question: how does a sodium-, potassium-, proton-, and ATP-independent transporter such as the bo,+-like system associated with rBAT participate in an active process, like the reabsorption of cystine and dibasic amino acids? The answer came from the study of the electrical activity of system bo,+-like. Busch et al. (75) studied this activity of the system bo,+-like expressed by rBAT in oocytes; as expected, in oocytes expressing rBAT, but not in control oocytes injected with water, the presence of L-arginine in the medium produces an inward positive current, most probably due to the positive charge of arginine at neutral pH. Surprisingly, exposure of rBAT-expressing oocytes to L-leucine produced an outward positive current through the plasma membrane of the oocyte. The participation of ions (e.g., Na+, K+, Cl-) in these currents was ruled out. These results prompted the hypothesis that the bo,+-like/rBAT transporter exchanges amino acids through the plasma membrane; the outward positive current produced by zwitterionic amino acids (e.g., L-leucine) would be due to the concomitant exit of dibasic amino acids from the oocyte. This was demonstrated in several laboratories by testing the dependence of amino acid efflux from oocytes expressing rBAT on the external amino acids; the efflux of L-[3H]arginine and L-[3H]leucine is totally dependent of the presence of amino acids in the medium (3, 90, 110). In fact, Coady et al. (110) have isolated a renal rabbit rBAT cDNA by expression of the electric activity of system bo,+-like/rBAT in oocytes. Additional data confirmed that system bo,+-like is an obligatory exchanger, which acts as a tertiary active transporter (90).

1) Only the system bo,+-like substrates elicited efflux via system rBAT/bo,+-like in oocytes.

2) The exchange mechanism is able to accumulate amino acid substrates in oocytes expressing rBAT; ~50-fold intracellular accumulation of 50 µM extracellular radiolabeled L-arginine, L-leucine, or L-cystine. This level of accumulation is not due to metabolism of the radiolabeled substrate in the oocytes, and it is significantly higher than that obtained in noninjected oocytes, or in oocytes expressing the cationic amino acid transporter CAT1 (system y+ activity, which shows trans-stimulation but is not an exchanger; see sect. IIA2).

3) The active transport due to rBAT/system bo,+-like expression in oocytes has a limit of accumulation, which coincides with the amount of intracellular free amino acid substrates in the oocyte (these cells contain a very high intracellular concentration of amino acids that has been estimated to be ~2,500 µM zwitterionic amino acids and 750 µM dibasic amino acids; Ref. 550).

4) As a consequence of this, prolonged incubations of the rBAT-expressing oocytes in the presence of an rBAT/system bo,+-like substrate results in the complete exchange of this amino acid within the oocyte. In this situation (homogeneous exchange), and in voltage-clamp conditions, when homoexchange is forced (e.g., L-arginine influx and efflux or L-leucine influx and efflux), the electric activity of rBAT/system bo,+-like disappears. In contrast, when homogeneous heteroexchange is forced (e.g., L-arginine influx and L-leucine efflux, or vice versa), the current elicited by the substrates in oocytes expressing rBAT is maximal, and with a direction corresponding to the exchange. At a defined membrane potential (e.g., -50 mV), the exchange L-arginine (influx)/L-leucine (efflux) is favored versus the reverse exchange. This demonstrated that the exchange of substrates via rBAT/system bo,+-like is the only electric activity of this transporter, in agreement with previous data obtained with the cut-open oocyte model (110). In addition, both influx and efflux require substrates on both sides; no electric activity is evoked by substrates in rBAT-expressing cut-open oocytes, which is entirely due to heteroexchange of substrates, if no substrates are present on the trans-side (109, 110).

5) In conditions of homogeneous exchange, either homo- or heteroexchange, radiolabeled substrate efflux equals influx. This together with estimations of Hill coefficients of approximately one both in radiolabeled and electric amino acid transport measurements (45, 47, 75, 90) indicates a stoichiometry of exchange of one amino acid (influx)/one amino acid (efflux).

The tightness of the obligatory exchange mechanism of rBAT/system bo,+-like is at present unknown. The fact that cationic amino acid-evoked currents occur in rBAT-expressing cut-open oocytes only when zwitterionic amino acids are present on the trans-side favors a very tight coupling of exchange (109). With the assumption of a theoretical absolute requirement of the transporter to be occupied by a substrate at either side (intra- or extracellularly) to translocate (this is thermodynamically impossible) the amino acid, accumulation curves via rBAT/system bo,+-like have been modeled (90). If the transport model assumes that the velocity constants of translocation of the empty (no substrate bound at either side) transporter is ~30-fold lower than for the amino acid-transporter complex, the experimental accumulation curves may still be reproduced by the model (J. L. Gelpí and M. Palacín, unpublished data); the sensitivity of the transport studies of radiolabeled substrates in oocytes expressing rBAT precludes a more precise determination. The mechanism of exchange of rBAT/system bo,+-like (sequential or concerted substrate binding at both sides) is at present unknown. Interestingly, when the analog aminoisobutyric acid is used as a substrate, the amino acid exchange via rBAT/system bo,+-like shows a variable stoichiometry of exchange: the aminoisobutyric acid-induced currents (i.e., efflux of the positively charged cationic amino acid substrates) is higher than the concomitant aminoisobutyric acid radiolabeled transport flux (109). This suggests that aminoisobutyric acid "locks" the transporter in a conformation that enables free translocation of the transporter in a fraction of transport cycles. Accumulation studies with aminoisobutyric acid (predicted to be lower than with physiological substrates) have not been reported.

The exchange mechanism of rBAT/system bo,+-like is not an oocyte artifact. It has also been shown to occur in renal cells that naturally express rBAT; the gene is expressed, and the system bo,+-like is present in the "renal proximal tubular" OK cell line (374). 1) In the apical pole, the transport of cystine and most of the L-arginine transport (~80%) has the substrate specificity of system bo,+-like. 2) The substrate efflux via system bo,+-like shows complete dependence on external amino acid substrates. 3) This transport activity is due to the expression of the rBAT gene; stable transfection with antisense rBAT sequences results in the partial and specific loss of system bo,+-like activity. This demonstrated that the exchange mechanism for rBAT/system bo,+-like also occurs in epithelial renal cells. This is in full agreement with heteroexchange of cystine and dibasic amino acids observed in renal brush-border membrane vesicles (352). The sodium-dependent L-histidine transport induced by rBAT in oocytes (4) is currently being studied in OK cells to assess the physiological relevance of this induction.

Interestingly, the y+L-like transport activity induced by 4F2hc in oocytes also behaves as an exchanger (90). 1) Efflux of radiolabeled L-arginine via 4F2hc/system y+L-like requires the extracellular presence of its substrates (e.g., cationic amino acids in sodium-free medium or zwitterionic amino acids in sodium-containing medium). 2) As a consequence of this exchange mechanism, oocytes expressing 4F2hc are able to accumulate system y+L-like substrates at higher levels than noninjected oocytes or CAT1/system y+-injected oocytes. 3) The exchange of amino acids via 4F2hc/system y+L-like is asymmetric. Efflux of radiolabeled L-leucine is not observed in oocytes expressing 4F2hc even in the presence of extracellular substrates; this is interpreted as showing that the interaction of zwitterionic amino acids with 4F2hc/system y+L-like at the low intracellular sodium concentration is very weak and therefore not visible in radiolabeled uptake studies. This suggests that exchange via 4F2hc/system y+L-like favors the efflux of cationic amino acids and sodium-dependent influx of zwitterionic amino acids. The erythrocyte/placental system y+L shows marked trans-stimulation, compatible with a ratio of velocity constants for the translocation of the occupied and empty carrier of at least 25 (14, 127, 135). It is therefore also possible that system y+L may indeed be acting as an exchanger. In this sense, as discussed previously for the modeling of the accumulation of substrates in oocytes via rBAT/system bo,+-like, the level of trans-stimulation of system y+L would allow transient accumulation of substrates similar to those observed in 4F2hc-expressing oocytes (90). It would be interesting to discern whether the asymmetric exchange (efflux of cationic amino acids/influx of zwitterionic amino acids plus Na+) observed via system y+L-like activity in 4F2hc-injected oocytes also happens in the erythrocyte/placental y+L system. What is the nature of the sodium dependence of L-leucine efflux from plasma membrane preparations in conditions in which other transport activities are blocked (i.e., NEM-treated vesicles to inhibit system y+ and in the presence of intravesicular BCH, a system L inhibitor)?

3. Protein structure

rBAT and 4F2hc proteins have no membrane leader sequence, similar hydrophobicity plots (reviewed in Ref. 407), and four regions (12-17 amino acid residues long) highly conserved (67-80% identity) (see Fig. 7). Both proteins also have a domain with significant homology with a protein family of prokaryotes and insect alpha -amylases and alpha -glucosidases (42, 46, 598). Interestingly, the catalytic site of these glucosidases is not totally conserved in rabbit rBAT or human 4F2hc; this is consistent with the fact that expression of rBAT in oocytes does not show alpha -amylase or maltase activity (598). In contrast to the well-known membrane multispanning structure of membrane transporters of substrates of polar nature (607), rBAT and 4F2 (4F2hc) are less hydrophobic and contain, depending on prognosis based on hydrophobicity algorithms, a single TM domain (i.e., a type II membrane glycoprotein; Refs. 46, 435, 553, 598) or four TM domains, with intracellular NH2 and COOH termini; Ref. 549); the more NH2-terminal hypothetical TM domain is the only one showing a clear prognosis as a TM domain (see Fig. 7). Surprisingly, these structures induce amino acid transport activity via system bo,+-like and y+L-like in Xenopus oocytes, respectively, and the involvement of rBAT in cystinuria demonstrates a role for rBAT in renal and intestinal reabsorption of amino acids. The apparent inability of these proteins to provide an aqueous translocation pathway in the plasma membrane, due to their low hydrophobicity, prompted the hypothesis that they may be modulators of transporters with a heteromeric structure (42, 46, 598).

Biochemical and immunochemical studies have demonstrated that rBAT and 4F2hc are integral membrane N-glycoproteins. The experimental evidence for rBAT can be summarized as follows: 1) in vitro translation. Addition of microsomes to the reticulocyte translation system increases (<20 kDa) the molecular mass of the protein product synthesized from rBAT cRNA (344, 598). 2) rBAT is expressed in oocytes. The protein product (~90 kDa) from rBAT cRNA in oocytes, shown by metabolic labeling with [35S]methionine, is an integral N-glycoprotein. Thus the product is not solubilized from oocyte membranes by sodium carbonate treatment. The treatment of the oocytes with tunicamycin reduces the size of the protein to ~72 kDa, compatible with the mass of the deduced protein from the cDNA (Mr <79 × 103) (45). 3) Studies with the native protein have been done. Western blot analysis using specific anti-rBAT antibodies revealed a protein band of 90-95 kDa in membrane preparations from kidney and mucosa from the small intestine (159, 379). The size of this band is reduced to ~72 kDa after endoglycosidase F treatment of renal brush-border membranes (J. Chillarón and M. Palacín, unpublished data). Because of the lack of leader peptide and because the N-glycosylation sites are toward the COOH terminus from the location of the most evident putative transmembrane domain (see Fig. 7), it was proposed that rBAT and 4F2hc were type II membrane glycoproteins (i.e., cytosolic NH2 terminus and extracellular COOH terminus) (46, 435, 553, 598). In contrast, Tate and co-workers (378) have proposed that rBAT crosses the plasma membrane at least four times, with the first transmembrane domain already mentioned and three additional amphipathic transmembrane domains (Fig. 8). This is based on studies of limited proteolysis and peptide-specific antibody detection of permeabilized cells expressing the rBAT protein (378). These highly interesting results on the rBAT protein await confirmation with different approaches; no similar studies have been conducted with 4F2hc. In any case, it seems that one or four transmembrane domains are not enough to conform a polar pore for the movement of amino acids through the plasma membrane.

rBAT and 4F2hc might be components of heteromeric amino acid transporters (42, 46, 599): rBAT and 4F2hc may be "activators" of silent bo,+-like and y+-like transporters of the oocyte, respectively. A possible mechanism for this activation could be the constitution of holotransporters with subunits present in the Xenopus oocytes. This hypothetical mechanism would be similar to the activation of the oocyte alpha -catalytic subunits of the Na+-K+-ATPase by the expression of foreign beta -subunits of the Na+-K+-ATPase (166); a similar mechanism has been described for multimeric channels (32, 206, 469). In this sense it is very interesting that the cell surface antigen 4F2 is a heterodimer (~125 kDa) composed of a heavy chain of 85 kDa (4F2hc, i.e., the homologous protein to rBAT) and a light chain of 40 kDa linked by disulfide bridges (204, 209). Unfortunately, this light subunit evidenced by 125I labeling and immunoprecipitation has not been microsequenced or cloned. In a similar way, Wang and Tate (591) have reported the presence of these complexes in brush-border preparations from kidney and intestine. In our hands, renal rBAT is immunodetected in Western blot studies in nonreducing conditions as complexes of ~240 and ~125 kDa; in two-dimensional gels (first in nonreducing conditions, then in reducing conditions), the 240-kDa and the 125-kDa bands contribute to the ~90 kDa seen in reducing conditions (408). Interestingly, in membranes obtained in the presence of NEM from oocytes expressing rBAT, complexes similar in size to those observed in kidney have been reported (591). All this suggests that similarly to 4F2 antigen, rBAT forms a heterodimeric structure (125 kDa) of a "heavy chain" (~90 kDa) linked by disulfide bridges to a putative "light chain" of 40-50 kDa.

4. Structure-function relationship

The studies on structure-function relationship on rBAT/system bo,+-like and 4F2hc/system y+L-like are scarce, and limited to, the defect in four cystinuria-specific rBAT mutants (78, 91, 366), the effect of a COOH-terminal deletion of rBAT (367), and indirect evidence that supports the hypothesis that the functional units of the systems bo,+-like and y+L-like are heterodimeric structures of rBAT and 4F2hc with their corresponding putative "light subunits" (see below; see Refs. 408, 409, 548).

Four human cystinuria-specific rBAT missense mutations have been tested for amino acid transport expression in oocytes (Met467Thr and Met467Lys, Refs. 78 and 91; E268K and T341A, Ref. 366) (see location of these amino acid residues in Fig. 8). All four show partially defective amino acid transport when expressed in oocytes. Expression in oocytes of Met467Thr, the most frequent cystinuria type I mutant known worldwide (it represents 26% of the type I cystinuria chromosomes explained; see sect. III), and the mutant Met467Lys results in a reduced Vmax of the induced system bo,+-like, without substantial effect on the apparent Km; this is not due to defects in the synthesis or degradation of the transporter (78, 91). A deeper study on the transport defect associated with Met467Thr and Met467Lys mutants revealed a plasma membrane trafficking defect. These mutants express only an endoglycosidase H-sensitive protein band in the oocytes, and the protein reaches the oocyte plasma membrane slowly and inefficiently, as revealed by surface biotinylation studies (91). Long oocyte expression periods (>3 days after injection) and injection of oversaturating amounts of mutant rBAT cRNA result in total (Met467Thr) or partial (<= 20% activity of the wild type for Met467Lys) recovery of the induced amino acid transport (91); it is interpreted that these conditions overcome the protein quality control machinery of the oocyte. Interestingly, when the amino acid transport activity induced by Met467Thr mutant is recovered, the amount of Met467Thr on the oocyte surface is only <10% of the corresponding wild-type protein; this suggests that an oocyte "factor" limits the expression of system bo,+-like activity when oversaturating amounts of rBAT cRNA are expressed (91).

In an interesting study, Miyamoto et al. (367) showed that a COOH-terminus deletion (Delta 511-685) on human rBAT, which eliminates the fourth putative TM domain as well as the fourth segment of high homology between rBAT and 4F2hc (see Figs. 7 and 8), induces in oocytes a decreased amino acid transport activity (radiolabeled amino acid transport studies) that resembles that of 4F2hc/system y+L-like (i.e., sodium-independent transport of dibasic amino acids and sodium-dependent transport of zwitterionic amino acids); expression of longer deletions in the COOH terminus of rBAT renders no transport function in oocytes. This suggests that rBAT and 4F2hc are modulators or subunits of the complete transporters, in which the substrate specificity of systems bo,+-like and y+L-like resides. In addition, this study suggests that the COOH terminus is relevant for the interaction with the putative transporter or subunit. There is one concern in this interpretation of these results. Delta 511-685 rBAT expresses substantial substrate-evoked currents (15-20 nA by 50 µM L-arginine or L-leucine at -50 mV) in the oocytes (367). In contrast, substrate-evoked currents by 4F2hc in oocytes are very small (<= 1 nA by 50 µM L-arginine or L-leucine at -50 mV) (90). In agreement with this, placental y+L activity is largely insensitive to membrane potential (135). This is interpreted as the reflection of the cotransport of sodium with zwitterionic amino acids in exchange with cationic amino acids via y+L, resulting in no electric activity (90). At present, there is no explanation as to why Delta 511-685 rBAT induces in oocytes an amino acid transport activity that is identical to 4F2hc/system y+L-like when transport is measured with radiolabeled substrates but differs in its electrical activity.

Additional evidence also points to rBAT and 4F2hc as modulators or subunits of the amino acid transporter. Transient expression of rBAT in COS cells resulted in the production of a glycosylated rBAT form that either does not reach the plasma membrane (our experiments; Ref. 408) or does so (Tate's group experiments; Ref. 378) but, in both cases, with no amino acid transport expression. Interestingly, in nonreducing conditions, the renal and intestinal characteristic ~125 kDa rBAT complex is not present; it might be that the putative "light subunit" of rBAT is not expressed in COS cells, precluding transport expression (408). Recently, we obtained additional functional evidence for the need of the putative light subunit in the 4F2hc-induced expression of system y+L-like (142): 1) there is dissociation between oocyte surface 4F2hc protein and induced amino acid transport activity (saturation of induced amino acid transport occurs at very low amounts of injected cRNA, 0.01-0.1 ng 4F2hc cRNA/oocyte); expression of larger amounts of cRNA results in more 4F2hc on the surface without increment in the induced uptake. 2) In addition, there is coexpression of system y+L-like activity upon injection of saturating doses of 4F2hc plus rat lung mRNA or plus rat lung size-fractionated mRNA; 4F2hc is necessary for this coexpression since 4F2hc antisense oligonucleotides specifically hybrid-deplete the coexpression of system y+L-like activity (Estévez and Palacín, unpublished data).

In summary, the studies with the Met467Thr rBAT mutant, the COOH-terminal deletion of rBAT and the coexpression of 4F2hc and rat-lung mRNA strongly suggest that oocyte light subunits together with expressed rBAT or 4F2hc are responsible for the expression of systems bo,+-like and y+L-like, respectively. This, together with the heterodimeic structure demonstrated for 4F2hc or indirectly evidenced for rBAT, suggests that the functional unit of these transporters is a heterodimer (rBAT or 4F2hc plus the corresponding putative light subunit). As already mentioned, two labs recently showed induction of additional amino acid transport activities with rBAT and 4F2hc. Ahmed et al. (4) showed sodium-dependent histidine uptake induced by rBAT in oocytes, in addition to the induction of system bo,+-like activity. Bröer et al. (69) showed induction by rat 4F2hc in oocytes of a type of system L-like transport activity (sodium-independent transport for cationic and some zwitterionic amino acids). This is in contrast to others who showed induction of system y+L-like activity (sodium-independent transport for cationic and sodium-dependent transport for some zwitterionic amino acids) by both human and rat 4F2hc (46, 142, 146, 599). The physiological relevance of the induction of sodium-dependent uptake by rBAT and a system L-like is at present unknown. Knockout studies similar to that reported for rBAT in OK cells (374) are necessary to assess the physiological relevance of these amino acid inductions. Nevertheless, this opens the possibility that several putative light subunits in combination with rBAT and 4F2hc, or unidentified homologous proteins, constitute different amino acid transport systems. If the hypothesis of the heterodimeric holotransporters for rBAT and 4F2 is valid, the amino acid transport systems bo,+-like and y+L-like will be the first examples of heteromeric transporters for organic substrates in mammals. Knowledge of the structure-function relationship of rBAT and 4F2hc urgently needs the isolation and cloning of the light subunit of 4F2 and the putative light subunit of rBAT. Purification of the ~125-kDa rBAT complex by classical biochemical ways and coexpression cloning strategies are currently in progress in several labs in an attempt to identify these subunits.

5. Physiological role of rBAT and 4F2hc

Our knowledge of the physiological role of rBAT is clearly greater than that of 4F2hc (see Ref. 408 for a recent review). The involvement of rBAT in classic cystinuria demonstrates the role of rBAT in the renal and intestinal reabsorption of cystine and dibasic amino acids. Because of the cellular localization of the rBAT protein and its mechanism of exchange (tertiary) active transport (see above), we proposed a model for the physiological role of transporter bo,+-like in the renal reabsorption of cystine and dibasic amino acids (90). In this model, the function of the transporter is directed toward apical reabsorption of cystine and dibasic amino acids, dissipating the intracellular gradient of zwitterionic amino acids. The negative membrane potential and the intracellular reduction of cystine to cysteine should favor this direction of the exchange. The zwitterionic amino acids released to the tubular lumen should be reabsorbed via active transporters (e.g., the sodium-dependent system neutral brush border) located in the apical plasma membrane of tubular epithelial cells. Is this model valid? The fact that mutations in the rBAT gene cause cystinuria, aminoaciduria of cystine and dibasic amino acids, but not of zwitterionic amino acids, clearly favors this model.

In addition to the cystine and dibasic amino acid reabsorption defect of classic cystinuria, the present knowledge of cystine reabsorption in kidney and intestine is very confused (for review, see Refs. 487 and 496). Work with brush-border membrane preparations from rat kidneys showed that L-cystine reabsorption is less sodium dependent than that of other zwitterionic amino acids (153, 352, 353, 449). Because of the weak sodium dependence of L-cystine reabsorption, cystine and dibasic amino acids may be accumulated across the apical membrane of kidney epithelial cells partly because of the intracellular reduction of cystine to cysteine and the negative membrane potential, respectively; basolateral transport systems would mediate the efflux of these amino acids (496). Segal and co-workers (352, 486) have provided evidence that renal brush-border membrane vesicles show two cystine transport systems: one with high-affinity (Km in the micromolar range), shared with dibasic amino acids that shows heteroexchange diffusion, and the other with low affinity and unshared. In addition, several authors have found inhibition by zwitterionic amino acids of cystine uptake, measured at low concentration (micromolar range) in renal brush-border preparations or perfused tubules, suggesting that the high-affinity system is also shared by zwitterionic amino acids (153, 158, 449, 479). In contrast to renal preparations, cystine transport in brush border from mucosa of the small intestine shows a single kinetic transport system of high affinity, shared with dibasic amino acids (404). Therefore, this high-affinity system, present in kidney and intestine, may be the system that is defective in cystinuria (111, 555). Microperfusion studies showed that this cystine high-affinity transport system is present in the proximal straight tubule (S3 segment), whereas the low-affinity system is present in the proximal convoluted tubule (S1-S2 segments) (479). Recently, Riahi-Esfahani et al. (449) reported that luminal membrane vesicles from the pars recta ("outer medulla") of rabbit kidney show a conspicuous component of cystine transport of high affinity (Km values of ~30 µM); interestingly, cystine transport in the pars recta is less sodium dependent and more sensitive to inhibition by micromolar concentrations of zwitterionic amino acids than in the pars convoluta (i.e., in apical membranes isolated from the "outer cortex").

More recently, it was demonstrated that cystine is transported through the apical pole of the "renal proximal tubular" cell line OK via system bo,+-like (i.e., sodium-independent, high-affinity transport system, shared with dibasic and zwitterionic amino acids), and it is due to the expression of rBAT; expression of antisense rBAT sequences specifically reduces this amino acid transport activity in OK cells (374). System bo,+-like activity has also been described in brush-border membrane vesicles from chicken jejunum (560) and in Caco-2 cells (557). The specific expression of rBAT in the microvilli of the S3 segment of the nephron and the cystinuria-specific mutations found in the rBAT gene allows us to propose that system bo,+-like (associated with rBAT) participates in the renal and intestinal reabsorption of cystine and dibasic amino acids of high affinity, most probably with a tertiary active transport mechanism (see above). Because of the location of rBAT in the S3 segment of the nephron, where only a part of the cystine reabsorption occurs (496), system bo,+-like could be envisaged as a low-capacity high-affinity system of physiological relevance as revealed by its alteration in cystinuria. In conclusion, most probably rBAT/system bo,+-like corresponds to the high-affinity reabsorption system of cystine described in renal ("pars recta") and intestinal preparations (see above). In contrast, the proteins responsible for the high-capacity low-affinity reabsorption of cystine in the proximal convoluted tubule are unknown.

We are far from establishing the physiological role of 4F2hc. It is even possible to imagine a multifunctional role for this protein. Before the linkage between 4F2hc and amino acid transport, it was implicated in calcium movement through the plasma membrane: 1) an anti-4F2hc monoclonal antibody (44D7) inhibited sodium/calcium exchanger activity in cardiac and skeletal muscle sarcolemmal vesicles (for review, see Ref. 303). 2) An anti-4F2hc antibody on parathyroid cells produces an increase in cytosolic free calcium concentration at low extracellular calcium levels (427). Recently, 4F2hc has also been implicated in cell fusion (396) and regulation of cell survival/death control (592). Fusion regulatory protein (FRP-1) regulates virus-mediated cell fusion and fusion of monocytes. Purification and partial sequencing of human FRP-1 revealed a strong homology of the NH2 terminus with human 4F2hc (it corresponds to the cluster of differentiation CD98) (11 of 15 amino acid residues are identical); both proteins show cross-reactivity with different antibodies, and the expression of both proteins is induced by concanavalin A or interleukin-2 treatment (396). To us, it seems that FRP-1 and 4F2hc are highly homologous, although not identical. To our knowledge, more extended sequences of FRP-1 have not been reported. In addition, treatment of monocytes with anti-4F2hc antibodies resulted in cell fusion and formation of multinucleated giant cells of Cd+U2ME-7, a CD4+U97 cell line transfected with HIVgp160 gene, whereas other anti-4F2hc antibodies suppress these induced fusion events (396, 397). Similarly, anti-FRP-1/4F2hc antibodies suppress human parainfluenza virus type 2-induced cell fusion (398). In a search for cell surface markers expressed on hematopoietic stem cells, Palacios and co-workers (592) found that Joro 177 monoclonal antibody stained these cells. A cDNA library search with this antibody resulted in the cloning of mouse 4F2hc. Interestingly, this antibody stimulates tyrosine phosphorylation of an unidentified 125-kDa protein, induces homotypic aggregation of progenitor lymphoid cells, inhibits cell survival/growth of hematopoietic cells, induces apoptosis, and prevents the generation of lymphoid, myeloid, and erythroid lineage cells. This study suggests that 4F2hc might act as a membrane receptor involved in the control of cell survival/death of hematopoietic cells (592).

The above-mentioned roles of 4F2hc in cell fusion and aggregation might involve integrin function. Very recently, 4F2hc has been implicated in the regulation of integrin function (146): 1) expression of 4F2hc (i.e., CD98) complements dominant suppression due to the overexpression of an integrin beta 1-cytoplasmic domain, 2) 4F2hc coimmunopreipitates with active beta 1-integrins, and 3) antibody-mediated cross-linking of 4F2hc stimulated beta 1-integrin-dependent cell adhesion. In this sense, anti-beta 2- and anti-beta 1-integrin antibodies blocked anti-FRP/4F2hc antibody-induced cell aggregation and antibody-induced polykaryocyte formation, respectively (530). In any case, this issue is not yet clear because other proteins also associate with 4F2hc. Thus FRP-1/4F2hc and cytoskeletal proteins (e.g., actomyosin, vimentin, and heat shock cognate protein 70) are coimmunoprecipitated by anti-FRP-1/4F2hc antibodies (522), and anti-FRP-1/4F2hc antibodies change the immunofluorescence pattern of these cytoskeletal proteins (522). It is therefore difficult at present to ascertain whether anti-FRP-1/4F2hc antibody-mediated cell fusion events are due to direct or indirect effects via changes in the cell surface distribution or conformation of other proteins. The role of the interaction of 4F2hc (or FRP-1) with other proteins (e.g., cystoskeletal proteins) in any of the putative functions of 4F2hc is also unknown.

As mentioned before, several labs have observed induced amino acid transport activity in oocytes injected with 4F2hc. One important question to resolve is the amino acid transport activity associated physiologically with 4F2hc in the cells that express 4F2hc naturally. In this sense, 4F2hc was originally described as a marker for tumor cells and activated lymphocytes (204). Stimulated lymphocytes (e.g., by concanavalin A, interleukin-2, or phytohemagglutinin) have a larger increment (~60-fold) of 4F2hc in the plasma membrane (for review, see Ref. 303); in some instances, this is due to increased transcript stability (554). Boyd and co-workers (88) addressed this question (88): 1) phytohemagglutinin induces in lymphocytes cationic amino acid transport with system y+ characteristics; and 2) transfection of antisense oligonucleotide sequences of human 4F2hc and CAT-1 (system y+; see sect. IIA), singly or in combination, inhibits the phytohemagglutinin-induced system y+ activity in human peripheral blood mononuclear cells. These results suggest a shared responsibility of 4F2hc and CAT-1 in system y+ activity. Unfortunately, all our attempts to coexpress system y+ transport activity by coinjecting human 4F2hc and mouse CAT-1 or CAT2-a failed (142). Therefore, at present, we do not known which specific cationic transport activity is physiologically related to 4F2hc [i.e., systems similar to y+L (42, 146, 599), L (69), or y+ (88)]. In any case, the obligatory exchange of amino acids via system y+L-like, associated with 4F2hc expression in oocytes, which has been discussed in the previous sections, might have important physiological consequences. It has been reported that efflux across the basolateral membrane is the rate-limiting step in the intestinal absorption of dibasic amino acids (86, 383). Furthermore, leucine at low micromolar concentration increases (6- to 10-fold) the transepithelial flux of lysine (86, 87). Countertransport between lysine (outward) and leucine (inward) or allosterism was considered to be responsible for this process. System y+L can sustain lysine-leucine exchange with an apparent Km for leucine of ~10 µM in the presence of sodium (127). If such a system is found in the basolateral membranes of intestinal or renal epithelial cells, the hypothesis that system y+L can affect countertransport will be supported (14). The surface antigen 4F2hc has a basolateral localization in renal epithelial cells from the proximal tubule (436). System y+L-like, associated with 4F2hc expression, could be responsible for the active release of dibasic amino acids through the basolateral membrane of epithelial cells. The fact that the direction of exchange that is favored is L-arginine (outward) with low micromolar concentration of leucine (inward) in the presence of sodium strongly supports this hypothesis (90). Further research is needed to elucidate the mechanism (e.g., a weak interaction of zwitterionic amino acids from inside due to the low intracellular concentration of sodium) responsible for this asymmetric exchange.

    III. INHERITED DISEASES OF PLASMA MEMBRANE AMINO ACID TRANSPORT
Top
Previous
Next
References

This section deals with the inherited pathology due to defective amino acid transport in the plasma membrane of human cells. Table 10 summarizes the characteristics of the defective amino acid transport systems and the candidate genes for eight (including subtypes) of these diseases. They are all aminoacidurias and therefore affect the tubular reabsorption of specific amino acids. For background information (clinical, genetic, biochemistry, and physiology) about these diseases, see Reference 476 and OMIM (On-line Mendelian Inheritance in Men; http://www3.ncbi.nlm.nih.gov/omim/). Only one human amino acid transporter gene, rBAT (also named SLC3A1, for solute carrier family 3, member 1; OMIM no. 104614), has been shown to be responsible for one of these inherited diseases, cystinuria type I (see below). Very recently (49, 296, 595), cystinuria type III (perhaps also type II) and LPI have been linked to chromosomes 19q13.1 and 14q, respectively. The tissue distribution and the transport characteristics associated with the expression of the sodium- and potassium-dependent zwitterionic amino acid transporter ATBo and the sodium- and potassium-dependent anionic amino acid transporter EAAT3 (see sect. II) make them good candidates for Hartnup disorder and dicarboxylic aminoaciduria, respectively (see Table 10). In addition, the dicarboxylic aminoaciduria developed by the null knockout EAAT3 mice (416) reinforces the putative role of EAAT3 in this inherited disease. For the rest of aminoacidurias due to defective renal reabsorption (i.e., isolated cystinuria, hyperdibasic aminoaciduria 1, isolated lyinuria, and iminoglycinuria), neither obvious candidate genes nor chromosomal location is known.

 
View this table:
[in this window] [in a new window]
  		TABLE 10.   Inherited diseases of plasma membrane amino acid transport

In addition to the above-mentioned inherited diseases, it is worth mentioning the putative responsibility of the EAAT2 glutamate transporter in the sporadic form of ALS. Amyotrophic lateral sclerosis is a chronic degenerative neurologic disorder characterized by the death of motor neurons in the cerebral cortex and spinal cord. About 90% of ALS is sporadic, and only 10% is familial; mutations in the superoxide dismutase-1 gene have been found in 15-20% of all familial cases (see entry no. 105400 in OMIM). Although the etiopathology of sporadic ALS is not known, it is hypothesized that glutamate excitotoxicity participates in the selective motor neuron degeneration of the disease (457). Amyotrophic lateral sclerosis is characterized by increased cerebrospinal fluid concentration of L-glutamate and L-aspartate and a marked and specific decrease (~70%) in the Vmax of high-affinity glutamate uptake in synaptosomes from motor cortex and spinal cord (462). In this sense, decreased D-aspartate binding sites have been reported in the spinal cord of ALS specimens (492), suggesting a decreased number of glutamate transporters. Rothstein et al. (463) showed that this defect is specific to the glial EAAT2 transporter, the expression of which is reduced up to 95% in the motor cortex and spinal cord of postmortem samples from ALS patients. Despite the large loss of EAAT2 protein in those brain structures, the transcript levels of EAAT2 in motor cortex are not altered in ALS (66). The first mutational analysis study showed no mutations in the EAAT2 mRNA sequence of ALS patients (347). In contrast, Rothstein found aberrant EAAT2 RNA, including exon-skipping and intron-retension species, in 65% of sporadic ALS specimens (J. D. Rothstein, personal communication). An intron-retention species has a dominant negative effect on the stability of wild-type EAAT2 protein when expressed in COS cells. At present, the origin of these aberrant RNA species is unknown, and more likely, they are not genomic but rather due to aberrant RNA processing. Further research is needed to understand the mechanism underlying the aberrant EAAT2 RNA processing in the brain of ALS patients.

1. Cystinuria

Classic cystinuria (OMIM no. 220100) is an inherited hyperaminoaciduria of cystine and dibasic amino acids (for review, see Refs. 351, 487), discovered by Wollaston (604), and described as one of the first four "inborn errors of metabolism" by Garrod (163). Cystinuria is an autosomal recessive disease with an overall estimated prevalence of 1 in 7,000 neonates; prevalence estimations range between 1 in 2,500 neonates in Israeli Jews of Libyan origin and 1 in 100,000 in Sweden (487). In our opinion, these numbers are overestimations in screening programs, because of the nonsilent hyperaminoaciduria phenotype of cystinuria types II and III (see below). Because of the poor solubility of cystine, it precipitates to form kidney calculi that produce obstruction, infection, and ultimately renal insufficiency. Three types of classic cystinuria have been described (456): type I heterozygotes present normal aminoaciduria, whereas types II and III present high and moderate hyperaminoaciduria of cystine, lysine, and, to a lesser extent, arginine and ornithine. As a consequence of the intestinal amino acid transport defect, type I and II homozygotes do not show increase in the plasma levels of cystine after an oral administration of the amino acid. In contrast, type III homozygotes show a nearly normal increase in the plasmatic levels of cystine after the oral dose. This suggests that the amino acid transport system affected in cystinuria type III is not expressed or is not very conspicuous in the intestine. Others (197, 377) divide cystinuria into two types: type I, or true recessive, and type II, or incomplete recessive (this includes the types II and III of Rosenberg; Ref. 456).

Dent and Rose (124) postulated that cystinuria may result from the defective function of a common uptake system for cystine and dibasic amino acids. Milne et al. (361) demonstrated a reduced intestinal absorption of dibasic amino acids in patients with cystinuria. Finally, transport studies in vitro demonstrated a defective accumulation of cystine and dibasic amino acids in biopsies of patients with cystinuria (111, 555). Interestingly, patients with cystinuria show no malabsorption of arginine when given in a peptide form; this suggested normal apical absorption of peptides in cystinuria and positioned the disease-associated transport defect at the apical membrane of the intestinal epithelium (21). As discussed in section IID5, there is an apical high-affinity amino acid transport system for cystine and dibasic amino acids that also shows interaction (cis-inhibition and heteroexchange) with zwitterionic amino acids, in the brush-border membranes of the epithelial cells of the proximal straight tubules of the nephron and of the small intestine. It is believed that this is defective in cystinuria (for review, see Refs. 408, 487). As described in section II, the transport characteristics of rBAT/bo,+-like system and its tissue and subcellular distribution suggested the participation of rBAT in a high-affinity reabsorption system of cystine and dibasic amino acids in kidney and intestine and postulated rBAT as a good candidate gene for cystinuria. Mutational analysis of the rBAT gene of patients with cystinuria initially revealed six missense cystinuria-specific mutations; for one of these mutations (Met467Thr; see Fig. 8), defective amino acid transport activity was shown (see sect. II); this demonstrated that mutations in rBAT cause cystinuria and that rBAT/system bo,+-like participates in the renal reabsorption of cystine and dibasic amino acids (78). Genetic analysis demonstrated linkage of cystinuria with chromosome 2p microsatellite markers (430), which colocalize with the rBAT gene locus in 2p16.3 (79). Further mutational analysis by several groups (for review, see Refs. 170, 408, 409) of the rBAT gene in Italian, Spanish, Middle Eastern, Eastern European, Canadian, Japanese, and United States populations revealed a growing number of cystinuria-specific mutations in the rBAT gene (25 mutations have been described, including missense, nonsense, splice-junction, deletions, and insertions; see Fig. 8). Four cystinuria-specific rBAT mutants have been shown to express defective amino acid transport activity (see sect. IID4 and Fig. 8; Refs. 78, 91, 366). Mutations Met467Thr and R270X [stop codon at arginine residue 270; this eliminates two-thirds of the protein toward the COOH terminus; Miyamoto et al. (367) reported that deletions affecting the COOH terminus result in defective rBAT-expressed amino acid transport activity] represent approximately one-half of the cystinuric chromosomes where mutations have been detected. These mutations have been found in homozygosis in several patients and in compound heterozygotes with other mutations.

Clinical and physiological evidence suggested heterogeneity in cystinuria (see above). 1) The oral cystine test may be indicative that in type III cystinuria the intestinal defect is not very conspicuous. 2) Most of renal reabsorption of cystine occurs in segments S1-S2 of the nephron (i.e., in a tubular region other than that in which rBAT is expressed). Thus other cystine reabsorption system(s) not present (or not very conspicuous) in the small intestine may also be cystinuria genes (see sect. IID5). Mutational analysis suggested that only patients with type I cystinuria carried mutations in the rBAT gene (164, 217). Genetic linkage analysis with markers of the genomic region of rBAT in chromosome 2 and intragenic markers of rBAT have demonstrated genetic heterogeneity for cystinuria (80). The rBAT gene is linked to type I cystinuria, but not to type III (OMIM no. 600918). A wide search through the genome, carried out independently by two groups, localized type III (and perhaps also type II) cystinuria gene in patients from Italy and Israeli Jews with a Lybian origin to 19q13.1 (49, 595). We are currently analyzing this locus for the identification of a new cystinuria gene. In these studies, the phenotype classification of cystinuria (types I, II, and III) was based on the urine excretion values of cystine and dibasic amino acids in the obligate heterozygotes.

Cystinuria type I, the most frequent worldwide (i.e., >60% of the cases), is due to mutations in the rBAT gene. As discussed in section IID, this gene codes for a protein that most probably participates as a subunit of a heterodimeric bo,+-like transporter. This activity is responsible for the high-affinity cystine and dibasic reabsorption in the S3 segment of the nephron and in the small intestine with a tertiary active transport mechanism coupled with the exchange of neutral amino acids. Interestingly, in the Italian and Spanish patients with cystinuria type I subjected to study, we have identified mutants only in ~50% of the cases. In the future, this figure may increase through the analysis of deletions in the rBAT gene (we are currently studying 3 putative new deletions), but it is possible that mutations in the rBAT gene will not explain all cystinuria type I chromosomes. In this sense, the putative light subunit of rBAT could also be envisaged as a type I cystinuria gene. From the cystinuria loading test (see above), we can speculate that the type III cystinuria gene would have a low expression in the small intestine. The transport system responsible for the high-capacity low-affinity reabsorption of cystine in segments S1-S2 of the nephron may be defective in cystinuria type III. In contrast, there is no obvious candidate gene for type II cystinuria. This is a rare (<= 5% of the cystinuria cases) type of the disease, and its ascription to the 19q13.1 locus is at present not definitive, since this linkage, although significant, is based on a small number of cases (49) (see Table 10).

Whether mutations in rBAT (chromosome 2p16.3) and in the new type III cystinuria locus (chromosome 19q13.1) lead to a full-blown type I/type III cystinuria phenotype is still an open question. Initial mutational analysis suggested this genotypic/phenotypic interaction (164, 217). Linkage analysis with both cystinuria loci is currently in progress. Preliminary data suggest that cases of type III heterozygotes within the lower range of cystine and dibasic hyperexcretion values of these carriers (173) may be due to mutations in the rBAT gene.

Finally, Brodhel et al. (67) reported isolated cystinuria (OMIM no. 238200) in two siblings of unrelated parents (see Table 10), in which urinary hyperexcretion of amino acids was restricted to cystine. This suggested that a cystine renal transporter not shared with dibasic amino acids was defective in these patients (for review, see Ref. 487). Biochemical evidence for this transporter has not been obtained in renal or intestinal transport studies (496). It is therefore possible that a rare allele either of the rBAT gene or of the cystinuria gene in 19q13.1 may be responsible for this phenotype (i.e., a mutant affecting cystinuria transport but not dibasic amino acid transport). To our knowledge, linkage and/or mutational analysis of the rBAT gene and the 19q13.1 cystinuria locus in isolated cystinuria has not been reported.

2. Other dibasic aminoacidurias

There are four diseases in which a cationic amino acid transport defect is suspected (for a review, see Ref. 497): 1) cystinuria (see above); 2) LPI, hyperdibasic aminoaciduria type 2, or familial protein intolerance (OMIM no. 222700); 3) hyperdibasic aminoaciduria type 1; and 4) isolated lysinuria. Lysinuric protein intolerance is an autosomal recessive trait. Almost one-half of the known LPI patients (~100) are from Finland, where the prevalence of the disease is 1 in ~60,000. In contrast, hyperdibasic aminoaciduria type 1 (autosomal dominant trait) and isolated lysinuria have been described only in one French Canadian pedigree and in a Japanese patient, respectively. Lysinuric protein intolerance was first described by Perheentupa and Visakorpi (418). In addition to hyperdibasic aminoaciduria, the clinical symptoms of LPI are failure to thrive, protein aversion, short stature, hepatomegaly, osteoporosis, hyperammonemia, common interstitial lung disease, and renal damage, and occasionally moderate mental retardation. It is believed that the disease is caused by a defective dibasic amino acid transport that is expressed at the basolateral membrane of the renal and intestinal epithelia, and in nonepithelial cell types (e.g., culture fibroblasts, hepatocytes) (for review, see Ref. 497). An oral loading administered to LPI patients with the dipeptide lysyl-glycine increased plasma glycine concentrations properly, but plasma lysine remained almost unchanged; this indicated unaffected apical peptide absorption and cellular hydrolysis and suggested the basolateral location of the defective cationic amino acid transport (443). In agreement with this, transport studies with jejunal LPI biopsy samples showed that the transport defect is situated at the basolateral plasma membrane (125).

In an interesting study, Scriver, Simell, and co-workers (501) reproduced in LPI fibroblast cell lines the cationic amino acid transport defect; LPI fibroblasts showed a reduced trans-stimulated efflux of cationic amino acids. This defect showed gene-dosage effect (homozygotes more affected than heterozygotes). It is believed that the defective cationic amino acid transport activity corresponds to system y+, but unfortunately, this has not been carefully characterized. Cationic amino acids are transported through the plasma membrane of human fibroblasts via systems y+ and y+L (see sect. I) (Torrents and Palacín, unpublished data).

Very recently, Simell, Aula, and co-workers (296) reported a locus on chromosome 14 for LPI in Finnish patients; linkage disequilibrium in markers within this locus suggests that LPI in these patients is due to one historical mutation. The hyperdibasic aminoaciduria characteristic of LPI has fostered studies on the involvement of the known cationic amino acid transporter in this disease. Unfortunately, none of the known proteins involved in cationic amino acid transport seems to be responsible for LPI. Indeed, human CAT-1 (chromosome 13q12-14), CAT-2 (chromosome 8p21.3-22), and CAT-4 (chromosome 22q11.2) (see Table 1) have been excluded from linkage to the LPI phenotype in Finnish patients (296). Similarly, mutational and linkage analysis excluded human CAT-1, CAT-2, and CAT-4 as LPI genes among Italian or Japanese LPI patients (128, 224, 485). The recently described rat CAT-3 (219), for which no human counterpart has been cloned (see Table 1), is expressed exclusively in brain and therefore does not represent a candidate gene for LPI. Two other proteins are known to be associated with cationic amino acid transport: rBAT and 4F2hc (see sect. IID). The rBAT gene is expressed in kidney and intestine, and it is associated with the cystinuria phenotype (see above). In contrast, the putative role of 4F2hc in the renal reabsorption and intestinal absorption (basolateral efflux of cationic amino acids by exchange with zwitterionic amino acids plus sodium; see sect. IID5) makes it a good candidate for LPI. In addition, 4F2hc is expressed in fibroblasts (Torrents and Palacín, unpublished data), where the LPI transport defect has been substantiated (501). 4F2hc does not seem to be directly involved in LPI, since, as indicated above, LPI gene localizes to chromosome 14q (296), and the human 4F2hc gene localizes to chromosome 11q12-13 (174). However, a role of 4F2hc in LPI cannot be ruled out. As mentioned in section IID, there is evidence that the functional unit of 4F2hc/system y+L-like transporter is composed by 4F2hc (heavy chain) plus an unidentified light subunit (142). This putative light subunit might be envisaged as an LPI gene. The identification of the LPI gene in the 14q locus (already restricted to 100 kb) and/or the cloning of the putative light subunit of 4F2 surface antigen may clarify this issue in the near future.

3. Hartnup disorder

This disorder (OMIM no. 234500) was first described by Baron et al. (35). It is transmitted as an autosomal recessive trait, and it is characterized by a pellagra-like light-sensitive rash (niacin deficiency), cerebellar ataxia, emotional instability, and aminoaciduria. This is a characteristic aminoaciduria that involves the zwitterionic amino acids (with the exception of cysteine/cystine, glycine, methionine, and the imino acid proline) and that occurs at a frequency of 1 in ~40,000 in urine amino acid screens (for review, see Ref. 304). Most of the hyperexcretors never display the niacin deficiency symptoms, and therefore, the Hartnup disorder is usually benign (477, 300). Urinary hyperexcretion occurs with normal amino acid plasma levels. Some patients have elevated fecal amino acid levels and secondary metabolites of the excess of tryptophan in the urine, plasma, and feces as well as reduced transient plasma levels increase after an oral load of zwitterionic amino acids. Then, the disorder appears to involve a renal reabsorption defect, and in some patients intestinal malabsorption of zwitterionic amino acids (for review, see Ref. 304). Studies with brush border of intestinal mucosa biopsies, but not with leukocytes or fibroblasts, from Hartnup patients showed defective zwitterionic amino acid transport (178, 493, 532, 545).

Scriver and co-workers (474, 477) have proposed that Hartnup disorder is an amino acid transport single gene defect affecting kidney and intestine, with a variant form affecting only the kidney; in contrast, the pathological state associated with the disease (niacin deficiency) seems to be multigenic. These authors suggest that other genes that control plasma amino acid homeostasis may influence the occurrence of clinical abnormalities with the Hartnup biochemical defect. The disease symptoms occur with low aggregate plasma amino acid levels and nutritional stress (malnutrition, diarrhea).

Very recently, Dove and co-workers (528) developed a mouse model for Hartnup disease (hyperphenylalaninemia 2; hph2) by N-ethyl-N-nitrosourea mutagenesis and screening for delayed plasma clearance of an injected load of phenylalanine. The hph2 is a recessive mutation that causes a deficient amino acid transport that is similar but not identical to Hartnup disease. Like Hartnup patients, the hph2 homozygotes show 1) specific urinary hyperexcretion of many of the zwitterionic amino acids, while plasma concentrations of these amino acids are normal; 2) a partial deficiency in the sodium-dependent uptake of glutamine in brush-border membrane vesicles; and 3) a niacin-reversible syndrome influenced by diet and genetic background. In contrast to Hartnup patients, hph2 homozygotic mice show urine hyperexcretion of arginine, a mild urine hyperexcretion of tryptophan and valine, and significant urine hyperexcretion of methionine.

Dove and co-workers (527) mapped hph2 to a region of mouse chromosome 7 synthenic with human chromosome 11q13 (see Table 10). Interestingly, 4F2hc, the putative subunit of the amino acid transport system y+L-like, also maps to this locus (see Table 1, rBAT). This amino acid transporter-related protein has been suggested as a candidate gene for the Hartnup disorder (304). In our opinion, the amino acid transport associated with 4F2hc expression in oocytes (systems y+L-like or L-like, depending on the authors; see sect. IID5), and the almost ubiquitous tissue distribution and the renal basolateral localization of 4F2hc (see sect. IID) weakens the candidature of this gene for this disorder. In any case, because of the hitherto unclear physiological role of 4F2hc in amino acid transport (see sect. II) and the dissimilar aminoaciduria phenotypes (i.e., urine excretion values of arginine, tryptophan, valine, and methionine; see above) of the hph2 mice with Hartnup patients, it will be very informative to answer the following questions: 1) What renal amino acid transport activity is defective in the the hph2 mice? 2) Does the hph2 locus contain the mouse 4F2hc gene? If the answers to these questions reinforce the candidature of 4F2hc for the hph2 phenotype, it should be assessed directly.

The transport characteristics (sodium-dependent zwitterionic amino acid transport) and the epithelial distribution of ATBo (see sect. IIC) fit those expected for the transporter responsible for the Hartnup disorder (Table 10). In contrast to the human synthenic locus of the hph2 mouse mutation (chromosome 11q13), the ATBo gene localizes to 19q13.3 (see Table 6). Therefore, ATBo does not seem to hold the hph2 mutation. In our opinion, because of the nonidentical aminoaciduria phenotype of Hartnup disorder and hph2 mutation, there is still room for a role of the ATBo gene in the Hartnup disorder. First, direct evidence for the apical localization of the ATBo transporter in renal and intestinal epithelia and for an active transport mechanism for the amino acid transport activity associated with ATBo should be offered. Then, direct genetic analysis of the ATBo gene (mutational and/or linkage studies) in Hartnup's aminoaciduria families should be addressed.

4. Iminoglycinuria

Familial iminoglycinuria (OMIM no. 242600) is a benign inherited defect of membrane transport (for review, see Ref. 89). It involves a glycine, L-proline, and hydroxy-L-proline transporter in the renal tubule and, in some cases, in the epithelial intestine. There are no reports of the prevalence of this disease, but it seems more frequent in Ashkenazim (see OMIM). As for other systems of renal reabsorption of amino acids, the reabsorption of these amino acids matures during the first months of life. The persistence of iminoglycinuria beyond 6 mo is considered abnormal. In addition to familial iminoglycinuria, this urinary hyperexcretion phenotype also occurs in familial hyperprolinemia and hyperhydroxyprolinemia, and in the generalized disturbance of membrane transport of the Fancony syndrome. In contrast to these, urine hyperexcretion of glycine, L-proline, and hydroxy-L-proline in familial iminoglycinuria is specific to these amino acids and occurs with normal levels of these amino acids in plasma. For glycine and these imino acids, the endogenous renal clearance rates are high, and the net reabsorption decreased in familial iminoglycinuria probands (reviewed in Ref. 89).

The iminoglycinuria phenotype is autosomal recessive, but in some pedigrees, there is an incomplete recessive phenotype; of 16 familial iminoglycinuria pedigrees reviewed by Chesney (89), in 9 pedigrees the obligate heterozygotes show hyperglycinuria without prolinuria. In addition, Greene et al. (177) reported a family in which the father and two sons had hyperglycinuria. The renal tubular titration curve for proline reabsorption in one of the sons was compatible with a mutation affecting the affinity of the proline transporter. This "Km" variant has been designated iminoglycinuria type II (OMIM no. 138500). It is believed that all these variants are allelic: the same renal phenotype is observed in probands inheriting two recessive mutant alleles, two hyperglycinuric alleles, or two different alleles (475).

There is evidence of two sodium-dependent proline transport systems in the brush border of human renal cortex (154): a high-affinity system shared with glycine and a low-affinity system not shared with glycine. Studies in rat, dog, and rabbit kidneys (reviewed in Ref. 89) revealed two sodium-dependent transport systems for imino acids and glycine, one of high affinity and specific for these substrates, and the other with low affinity and with broad specificity with other zwitterionic amino acids. For glycine, two apical sodium-dependent transport systems have been described: a high-affinity low-capacity system located in the proximal straight tubules and a low-affinity high-capacity system in the proximal convoluted tubule. Notice the similarity with the proposed renal reabsorption systems for cystine (see sect. IID5). It is hypothesized that the defective transport system is low-capacity high-affinity for glycine and the two imino acids in the proximal straight tubule, but there is no direct proof of this (89). Ontogeny in humans also gives clues to the amino acid transport systems serving the renal reabsorption of these amino acids (reviewed in Ref. 89): 1) maturation of the renal reabsorption of glycine and proline occurs at different times after birth. 2) In contrast to controls, iminoglycinuria homozygotes have an almost complete absence of tubular reabsorption for proline and glycine; with maturation of the tubular function, reabsorption of proline and glycine appear independently. 3) In rats, the postnatal prolinuria is associated with low activity of a high-affinity sodium-dependent nephron transport system. This and additional evidence suggest that ontogeny is associated with deficient activity of high-affinity systems for imino acids and glycine that does not include the system controlled by the familial iminoglycinuria gene (89).

Unfortunately, there is no genetic information of a chromosome locus for the familial iminoglycinuria phenotype. The transport characteristics of the expected amino acid transport system defective in this phenotype fit that of the IMINO and Proline transport systems (see sect. I). Three cDNA and their splice variants, which belong to the superfamily of sodium- and chloride-dependent neurotransmitter transporters, GLYT1, GLYT2, and PROT (see sect. II), transport glycine and/or proline with characteristics of these systems (see Table 5). At present, it seems that GLYT2 and PROT are specific to the CNS, and only a peripheral tissue distribution has been demonstrated for GLYT1 (see Table 5). The splice variant 1a of GLYT1 is expressed in kidney and other peripheral tissues; in lung, spleen, and liver, there is evidence that GLYT1-1a is expressed in macrophages and not in the parenchymal cells (62). To our knowledge, there are no data on the subcellular distribution of GLYT1-1a in kidney. If this transporter is expressed in the apical pole of the tubular epithelium, it would be a good candidate for the familial iminoglycinuria phenotype. It is worth mentioning that L-proline uptake in renal brush border is chloride dependent in addition to sodium dependent (478). The human GLYT1 gene localizes to chromosome 1p31.3-p32 (see Table 4). Linkage studies of the familial iminoglycinuria phenotype would be the first step to contrasting this hypothesis.

5. Dicarboxylic aminoaciduria

Teijema et al. (552) reported the first case of dicarboxylic aminoaciduria (OMIM no. 222730) in a female child, most probably due to an anionic amino acid transport defect in kidney and intestine. To our knowledge, only two other cases have been reported (355, 526). Swarna et al. (526) found one of these by screening for amino acid disorders in 500 mentally retarded children in India. Melancon et al. (355) detected in a neonatal screening program a boy with massive glutamic and aspartic aminoaciduria. The boy was apparently healthy at the age of 3 years. Amino acid clearance studies revealed the presence of renal wastage of dicarboxylic amino acids. Intestinal transport and in vitro oxidation of dicarboxylic amino acids were found to be intact. The same group later reported (356) reduced uptake velocities of glutamate and aspartate in dicarboxylic aminoaciduria fibroblasts.

The neuronal and peripheral high-affinity glutamate transporter EAAT3 (see Tables 6 and 7) is an obvious candidate for the transporter defective in dicarboxylic aminoaciduria (see sect. IIC). This transporter is highly expressed in kidney and in epithelial small intestinal cells (19, 245). Indeed, EAAT3 is the only known anionic transporter in kidney and intestine (see Table 7). Finally, Stoffel and co-workers (416, 517) reported that null knockout EAAT3 mice develop dicarboxylic aminoaciduria. This clearly substantiates the role of EAAT3 transporter in the renal reabsorption of anionic amino acids and in addition suggests EAAT3 gene (chromosome 9p24; see Table 6) as the immediate candidate for dicarboxylic aminoaciduria (Table 10). At this stage, because of the low number of disease cases described, the obvious next step is to search for mutations of the EAAT3 gene in patients with dicarboxylic aminoaciduria.

    IV. PROSPECTS
Top
Previous
Next
References

We are halfway toward the identification of the genes coding for the transporters that mediate the amino acid flux across the plasma membrane of mammalian cells. Relevant amino acid transport systems, like systems A, L, N, and x -c, have not been identified at the molecular level. We are just beginning to understand the molecular bases of the human inherited diseases of amino acid transport: mutations in the rBAT gene cause cystinuria type I. On the other hand, the first knockouts for amino acid transporters have been produced, by homologous recombination for the cationic amino acid transporter CAT-1 and for the glutamate transporters EAAT2 and EAAT3, and by antisense technology for the glutamate transporters EAAT1, EAAT2, and EAAT3 in brain and for the rBAT/system bo,+-like in epithelial renal cells. We can envisage a final goal in this line of research: the ascription of every amino acid transporter and the cognate transport system to the macroscopic fluxes of amino acids across the plasma membrane of mammalian cells.

The amino acid transporters cloned can be grouped in four protein families, and for many amino acid transport systems, several transporter isoforms have been identified. This has revealed a high complexity in mammalian amino acid transport. A relevant question, then, is what are the key structural elements that explain amino acid transport mechanisms at the molecular level? After the cloning of the first mammalian amino acid transporters, a growing number of studies based on site-directed mutagenesis and chimera constructions are being reported. These studies, although valuable, show a weakness, the lack of knowledge of the three-dimensional structure of these transporters sitting in the plasma membrane. There is no doubt that an enormous challenge in this line of research is the resolution of the amino acid transporter structures at the Ao scale, as for aquaporin-1 (587).

Finally, two amino acid transport systems, bo,+-like and probably y+L-like, could be a heterodimeric structure composed of rBAT or 4F2hc, respectively, plus the corresponding as yet unidentified subunit. If this hypothesis is proven, these transporters will be the first known transporters for organic solutes with a heteroligomeric structure.

In the present decade, molecular biology has reached mammalian amino acid transport; now we are on the way to explaining interorgan amino acid flux at the molecular level.

    ACKNOWLEDGEMENTS

  We thank Drs. Carol MacLeod, Baruch Kanner, Enerst Wright, Pertti Aula, Cecilio Giménez, Rosa Devés, Marçal Pastor-Anglada, Eduardo Soriano, Josep L. Gelpí, Virginia Nunes, Beatriz López-Corcuera, and David Torrents for helpful discussion for the writing of this manuscript. Our most grateful thanks to Dr. Gianfranco Sebastio for access to nonpublished information on the putative human CAT-4 transporter, to Carles Pucharcós for help in consulting data bases via internet, and to Cecilio Giménez for access, before publication, to the manuscript for Reference 622. We also thank all our collaborators from Spain, Italy, Switzerland, and Germany of the group of study of cystinuria and Robin Rycroft for editorial corrections.

    FOOTNOTES

   The work from our lab was supported in part by Direccion General de la Investigación Científica y Técnica, Spain, Grants PB90/0435, PB93/0738, and PM96/0060 and by Direcció General de Recerca Grant GR94-1040 from Catalonia.

  

    REFERENCES
Top
Previous

1.   ADAMS, M. D., M. DUBNICK, A. R. KERLAVAGE, R. MORENO, J. M. KELLEY, T. R. UTTERBACK, J. W. NAGLE, C. FIELDS, AND J. C. VENTER. Sequence identification of 2,375 human brain genes Nature 355: 632-634, 1992.

2.   ADAMS, R. H., K. SATO, S. SHIMADA, M. TOHYAMA, A. W. PÜSCHEL, AND H. BETZ. Gene structure and glial expression of the glycine transporter GlyT1 in embryonic and adult rodents. J. Neurosci. 15: 2524-2532, 1995[Abstract].

3.   AHMED, A., G. J. PETER, P. M. TAYLOR, A. A. HARPER, AND M. J. RENNIE. Sodium-independent currents of opposite polarity evoked by neutral and cationic amino acids in neutral and basic amino acid transporter cRNA-injected oocytes. J. Biol. Chem. 270: 8482-8486, 1995[Abstract/Free Full Text].

4.   AHMED, A., P. C. YAO, A. M. BRANT, G. J. PETER, AND A. A. HARPER. Electrogenic L-histidine transport in neutral and basic amino acid transporter (NBAT)-expressing Xenopus laevis oocytes. Evidence for two functionally distinct transport mechanisms induced by NBAT expression. J. Biol. Chem. 272: 125-130, 1997[Abstract/Free Full Text].

5.   AHN, J., O. MUNDIGL, T. R. MUTH, G. RUDNICK, AND M. J