Functional Properties and Cellular Distribution of the System A Glutamine Transporter SNAT1 Support Specialized Roles in Central Neurons*

Glutamine, the preferred precursor for neurotransmitter glutamate and GABA, is likely to be the principal substrate for the neuronal System A transporter SNAT1 in vivo. We explored the functional properties of SNAT1 (the product of the rat Slc38a1 gene) by measuring radiotracer uptake and currents associated with SNAT1 expression in Xenopus oocytes and determined the neuronal-phenotypic and cellular distribution of SNAT1 by confocal laser-scanning microscopy alongside other markers. We found that SNAT1 mediates transport of small, neutral, aliphatic amino acids including glutamine (K0.5 ≈ 0.3 mm), alanine, and the System A-specific analogue 2-(methylamino)isobutyrate. Amino acid transport is driven by the Na+ electrochemical gradient. The voltage-dependent binding of Na+ precedes that of the amino acid in a simultaneous transport mechanism. Li+ (but not H+) can substitute for Na+ but results in reduced Vmax. In the absence of amino acid, SNAT1 mediates Na+-dependent presteady-state currents (Qmax ≈ 9 nC) and a nonsaturable cation leak with selectivity Na+, Li+ » H+, K+. Simultaneous flux and current measurements indicate coupling stoichiometry of 1 Na+ per 1 amino acid. SNAT1 protein was detected in somata and proximal dendrites but not nerve terminals of glutamatergic and GABAergic neurons throughout the adult CNS. We did not detect SNAT1 expression in astrocytes but detected its expression on the luminal membranes of the ependyma. The functional properties and cellular distribution of SNAT1 support a primary role for SNAT1 in glutamine transport serving the glutamate/GABA-glutamine cycle in central neurons. Localization of SNAT1 to certain dopaminergic neurons of the substantia nigra and cholinergic motoneurons suggests that SNAT1 may play additional specialized roles, providing metabolic fuel (via α-ketoglutarate) or precursors (cysteine, glycine) for glutathione synthesis.

Glutamine uptake is crucial to neurotransmission in central excitatory and inhibitory neurons. Neuronal glutamine is derived from astrocytes and is the major precursor for neurotransmitter glutamate and ␥-aminobutyric acid (GABA) 1 (1)(2)(3)(4). Glutamine uptake into neurons is mediated largely by a System A-like transporter (5) and is a crucial step in the "glutamate-glutamine cycle" (6 -8). The neuronal enzyme phosphateactivated glutaminase (PAG) catalyzes the hydrolysis of glutamine to glutamate (9), which can also serve as a direct GABA precursor (10). Specific vesicular transporters package glutamate or GABA for synaptic release. Extrasynaptic glutamate is cleared by specific transporters (EAATs) found on astrocytes and at postsynaptic neuronal targets (11,12). The predominance of astrocytic rather than neuronal glutamate uptake systems (12) and the lack of EAAT expression on presynaptic membranes highlight a requirement for a glutamateglutamine cycle. However, GABA inactivation transporters are also present on presynaptic membranes (13). In astrocytes, the predominant fate of glutamate is conversion to glutamine by the astrocyte-specific glutamine synthetase. Completing the cycle, glutamine export from astrocytes is likely to be mediated by the System N or ASCT2 transporters (8,14,15). 15 N/ 13 C NMR studies in vivo have demonstrated that the fluxes through glutamine synthesis and glutamate/GABA-glutamine cycling are both substantial and necessary to replenish neurotransmitter pools. Furthermore, these fluxes represent the major fraction of glucose utilization and are central to the normal energetics of the brain (16,17). Glutamine may also constitute a significant fuel for many neuronal populations, with glutamate being oxidized to ␣-ketoglutarate, which enters a truncated tricarboxylic acid (TCA) cycle (18,19).
The sodium-coupled neutral amino acid transporter SNAT1, which was renamed as proposed in a recent review (20), was the first member of the System A amino acid transporter subfamily to be identified at the molecular level (21). The rat, human, and mouse cDNAs have each been isolated (21-24) and were formerly referred to as ATA1, SA2, SAT1, NAT2, or GlnT. SNAT1 mediates the Na ϩ -dependent transport of glutamine and is highly enriched in neurons (21). In the present study we explored the functional properties of SNAT1 by applying voltage-clamp and radiotracer techniques in cRNA-injected Xenopus oocytes. Using confocal laser-scanning double-immunofluorescence microscopy, we examined the cellular localization of SNAT1 and the phenotypes of the neuronal populations in which SNAT1 is expressed. We find that the functional properties and cellular distribution of SNAT1 support a key role within the glutamate/GABA-glutamine cycle, and consider additional roles for SNAT1 in amino acid metabolism in the brain.

EXPERIMENTAL PROCEDURES
Oocytes were isolated from Xenopus laevis (under 2-aminoethylbenzoate anesthesia), treated with collagenase A (Roche Diagnostics) and stored at 18°C in modified Barths' medium (25). SNAT1 (Slc38a1) cDNA was subcloned into the pTLNii vector under SP6, and SNAT1 cRNA was synthesized in vitro with the use of the Ambion mMESSAGE mMACHINE kit. SNAT2 cRNA was prepared as described (26). (The product of the Slc38a2 gene, SNAT2 was previously named SA1, ATA2, or SAT2.) Oocytes were injected with Ϸ50 ng of cRNA and incubated 2-3 days before radiotracer or voltage-clamp experiments were performed at 22 to 23°C (except as noted in Fig. 3H). Standard Na ϩ uptake medium comprised 100 mM NaCl, 2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES (pH 7.5 with Tris base). For low pH media, we lowered the HEPES concentration to 5 mM and added 5 mM MES. For Na ϩ -free or low Na ϩ medium, NaCl was replaced by equimolar choline chloride (ChoCl). For low Cl Ϫ medium, Cl Ϫ was replaced by equimolar gluconate. Solutions containing niflumic acid (where specified) also contained 0.1% dimethyl sulfoxide.
A two-microelectrode voltage clamp was used to measure currents in control oocytes and oocytes expressing SNAT1 or SNAT2. Microelectrodes (resistance 0.5-5 M⍀) were filled with 3 M KCl. Voltage-clamp experiments comprised three protocols: (i) Continuous current recordings were made at a holding potential (V h ) of Ϫ70 mV, Ϫ50 mV, or ϩ10 mV, low-pass filtered at 20 Hz, and digitized at 20 Hz. (ii) Oocytes were clamped at V h ϭ Ϫ50 mV, and step changes in membrane potential (V m ) were applied (from ϩ50 mV to Ϫ150 mV in 20-mV increments), each for a duration of 100 ms, before and after the addition of substrate. Current was low-pass filtered at 500 Hz and digitized at 5 kHz. Steady-state data were obtained by averaging the points over the final 16.7 ms at each V m step. (iii) A 1-s voltage ramp was applied from Ϫ100 mV to ϩ60 mV, preceded by a 20-ms settling phase at Ϫ100 mV. Current was filtered at 500 Hz and digitized at 5 kHz. Test solutions were washed out with substrate-free medium (100 mM ChoCl, pH 7.5) for at least 2 min. Steady-state data from protocols (i) or (ii) were fitted to a modified Hill relationship (Equation 1) for which I is the evoked current, I max the derived current maximum, S the concentration of substrate S (Na ϩ , Li ϩ or amino acid), K 0.5 S the substrate concentration at which current was half-maximal, and n H the Hill coefficient for S.
Currents obtained over the range of temperatures 21-33°C (Fig. 3H) were fitted to an integrated Arrhenius function (Equation 2), for which E a is the Arrhenius activation energy, A the y-intercept, R the universal gas constant (1.987 cal⅐mol Ϫ1 ⅐K), T the absolute temperature, and I the current evoked by 100 mM Na ϩ or the 10 mM L-alanine-evoked current in Na ϩ .  (25)(26)(27). The amino acid-evoked current was integrated with time to obtain the amino acid-dependent charge (Q Ala or Q Gln ) and converted to a molar equivalent (assuming monovalency) using the Faraday.
Following step-changes in V m using protocol ii, we obtained presteady-state currents in oocytes expressing SNAT1. These were isolated from capacitive transients (which decayed with half times of 0.5-0.9 ms) by subtracting the currents obtained for the same oocyte in the presence of 10 mM L-alanine and subtracting the final steady-state level. The compensated currents thus obtained were integrated with time to obtain charge movement (Q) and fitted to the Boltzmann relationship (Equation 3) for which maximal charge Q max ϭ Q dep Ϫ Q hyp (where Q dep and Q hyp represent the charge at depolarizing and hyperpolarizing limits), V 0.5 is the V m at the midpoint of charge transfer, z is the apparent valence of the movable charge, and F, R, and T have their usual thermodynamic meanings.
Confocal laser-scanning double-immunofluorescence microscopy was performed as previously described (28,29). In brief, deparaffinized frontal serial sections of adult rat brain fixed in Bouin-Hollande solution were incubated overnight at room temperature with the polyclonal rabbit antibody to SNAT1 (1:400 dilution) together with one of the following antibodies as markers: polyclonal guinea pig antibody to VGLUT1 (1:800 dilution), polyclonal guinea pig antibody to VGLUT2 (1:800 dilution), monoclonal mouse antibody KLON SY 38 (2 g⅐ml Ϫ1 ; Roche Applied Science) to the synaptic vesicle marker synaptophysin, or a monoclonal mouse antibody MAB3418 (25 g⅐ml Ϫ1 ; Chemicon International) to microtubule-associated protein-2 (MAP-2), an established marker of neuronal dendrites. Antibodies to VGLUT1 and VGLUT2 were raised against the same fusion proteins as our rabbit antisera to VGLUT1 and VGLUT2 (29). Immunoreactivities were visualized with indocarbocyanine(Cy3)-conjugated species-specific secondary antibodies diluted 1:200 and applied for 45 min at 37°C, resulting in a red-orange fluorescence labeling, or with biotinylated IgG (Dianova) diluted 1:200 and applied for 45 min at 37°C, followed by incubation with Alexis 488-conjugated streptavidin (MoBiTec) for 2 h at 37°C, resulting in a green fluorescence. Sections were analyzed with the Olympus Fluoview confocal laser-scanning microscope (Olympus Optical) and documented as false color confocal images.
Somatodendritic labeling by SNAT1 antibody in all areas of the brain was blocked in the presence of 10 M GST-SNAT1 fusion protein but not by GST protein alone (data not shown). Our SNAT1 antibody was previously characterized in heterologous expression systems and cerebellar granule neuronal cultures and by Western analysis of brain homogenates and cultured cells (21). Our VGLUT1 and VGLUT2 antibodies were previously characterized by double-fluorescence deconvolution microscopy and Western blot analysis (30). Fluorescence labeling was blocked by preadsorption of the antisera with 10 M GST fusion proteins, and immunoreactive bands were absent following preadsorotion with 1 M GST fusion proteins (data not shown).

RESULTS
SNAT1 Is a Na ϩ -dependent, Neutral Amino Acid Transporter-We explored the characteristics of SNAT1 by measuring radiotracer uptake and currents associated with the expression of SNAT1 in Xenopus oocytes. We found that SNAT1-mediated amino acid transport was saturable, Na ϩ -dependent, and rheogenic. SNAT1 stimulated the uptake of 100 M [ 14 C]MeAIB, a characteristic inhibitor of System A, in Na ϩ but not in choline. The Na ϩ -dependent component of MeAIB transport mediated by SNAT1 was 44-fold that in control oocytes (Fig. 1A). A range of amino acids evoked large inward currents of up to Ϫ2,000 nA in voltage-clamped oocytes expressing SNAT1, whereas amino acid-evoked currents in control oocytes were Ͻ Ϫ3 nA (see also Ref. 31). Typical current recordings are shown at a holding potential (V h ) of Ϫ70 mV (Fig. 1B). The addition of Na ϩ and subsequently 10 mM L-alanine had negligible effects in control oocytes, but both Na ϩ alone and Na ϩ plus L-alanine evoked inward currents in oocytes expressing SNAT1. Similar currents were obtained previously for SNAT2 except that the currents evoked by Na ϩ alone were significantly smaller for SNAT2 (26).
Substrate Profile of the System A Transporters SNAT1 and SNAT2-To examine the substrate profile of SNAT1 we measured the currents evoked by a range of amino acids (each at 10 mM) at Ϫ70 mV ( Fig. 2A). Currents were normalized to those evoked by L-alanine for which the half-maximal concentration (K 0.5 Ala ) was 0.3 Ϯ 0.02 mM (see Fig. 3B). SNAT1 preferred L-alanine over the D-isomer, primarily due to reduced apparent affinity (data not shown). SNAT1 transported small, neutral aliphatic amino acids that were ranked (based on absolute currents): L-cysteine, L-alanine, L-serine, L-asparagine, L-glutamine, L-histidine, glycine, L-methionine, L-threonine, MeAIB, and the imino acid L-proline. The 24 Ϯ 5% reduction in current evoked by L-glutamine compared with L-alanine was a V max effect and not due to a reduction in apparent affinity (K 0.5 Gln ϭ 0.3 Ϯ 0.03 mM, data not shown). However, the ratio I max /K 0.5 as an index of substrate selectivity was similar for L-alanine and L-glutamine (Fig. 2B). L-Threonine, an efficient substrate for System ASC, did not rank highly among SNAT1 substrates ( Fig. 2A). SNAT1 did not exhibit significant reactivity with the charged amino acids L-arginine, L-aspartate, and L-lysine. L-Glutamate evoked very small currents but the very low affinity (K 0.5 Glu Ϸ 11 mM, data not shown) and selectivity (Fig. 2B) indicate that SNAT1 is unlikely to represent a low affinity glutamate transporter described in the literature (12). SNAT1 essentially excluded the branched-chain amino acids (isoleucine, L-leucine, and L-valine, although L-valine showed marginal reactivity), L-phenylalanine, L-tryptophan, and 2-aminobicyclo (2,2,1)heptane-carboxylate (BCH), a characterizing analogue for the System L and System B 0ϩ amino acid transporters (31)(32)(33).
We compared the substrate profile of SNAT1 with that of SNAT2 (26), a second System A subtype of the SLC23 gene family, exhibiting apparent affinities (e.g. K 0.5 Ala ϭ 0.2 Ϯ 0.03 mM) similar to those for SNAT1. We found only subtle differences between SNAT1 and SNAT2 in terms of substrate profile ( Fig. 2A). SNAT2 displayed greater stereospecificity for the L-isomer (at least for alanine) than did SNAT1 but transported L-glutamine and MeAIB with less efficiency. The imino acid L-proline however is a better substrate for SNAT2 than for SNAT1 (see also Refs. 21,26,and 34). L-Leucine is not a substrate for SNAT1 but displayed marginal reactivity with SNAT2.
Uptake of [ 14 C]MeAIB ( Fig. 1A) was only about one-tenth of the uptake of L-[ 3 H]alanine (data not shown) and MeAIB always evoked a current much smaller than that evoked by L-alanine in oocytes expressing SNAT1 or SNAT2 ( Fig. 2A). This was primarily a result of a reduction in V max , since 10 mM MeAIB was effectively saturating despite a modest reduction in apparent affinity (K 0.5 MeAIB ϭ 1.1 Ϯ 0.4 mM, data not shown). Selectivity for MeAIB was low compared with L-alanine or L-glutamine (Fig. 2B). Our data suggest that MeAIB binds to the System A transporters with reasonable specificity (see also Ref. 26) but that architectural constraints significantly reduce the velocity at which MeAIB is transported. Whereas inhibition of radiotracer uptake by MeAIB (when used at appropriate concentrations) is a specific test for System A, our data demonstrate that MeAIB is not itself a model or paradigm substrate per se for System A since it is poorly translocated.
Saturation Kinetics and Voltage Dependence of SNAT1-mediated Amino Acid Transport-We examined the voltage dependence and concentration dependence of the amino acidevoked currents mediated by SNAT1 (Fig. 3). Amino acidevoked currents are voltage-dependent, as illustrated for L-alanine and L-glutamine (Fig. 3A). Evoked currents varied with V m in a linear fashion between Ϫ150 mV and Ϫ50 mV. Further slight reductions in current were observed at depolarized V m (Ϫ30 mV, Ϫ10 mV), but currents persisted at positive V m (up to ϩ50 mV). Whereas L-glutamine is likely to be the principal substrate for the neuronal SNAT1 in vivo, we chose to conduct our subsequent experiments with L-alanine because of the appreciably larger currents obtained with L-alanine and its relative stability in aqueous solution.

FIG. 2. Substrate profile of the System A transporters SNAT1
and SNAT2. A, currents (mean Ϯ S.E., n ϭ 3-9 oocytes) evoked by a range of amino acids (10 mM) at Ϫ70 mV were normalized to the current evoked by L-alanine for SNAT1 (black bars, Ϫ368 Ϯ 43 nA, n ϭ 9) and for SNAT2 (hatched bars, Ϫ608 Ϯ 75 nA, n ϭ 7). Except where indicated for D-alanine, all test substrates were L-isomers; BCH, 2-aminobicyclo (2,2,1)heptane-carboxylate; MeAIB, 2-(methylamino)isobutyrate. Mean currents were compared between SNAT1 and SNAT2 with Student's t test: *, p Ͻ 0.05; ***, p Ͻ 0.001. B, I max /K 0.5 (normalized to that for L-alanine) as an index of substrate selectivity. presence of Na ϩ at Ϫ70 mV (Fig. 3B). The Hill coefficient for alanine (n H Ala ) was 1 and did not differ with V m . K 0.5 Ala in Na ϩ was independent of V m in the range Ϫ150 mV to Ϫ50 mV (Fig. 3C). The increase in K 0.5 Ala beginning at Ϫ30 mV and rising to 1.3 Ϯ 0.2 mM by ϩ50 mV should be expected since 100 mM Na ϩ was no longer sufficient to activate the system maximally at these V m (see Fig. 3G). As for SNAT2 (data not shown), Li ϩ also supported alanine-evoked currents in the absence of Na ϩ but resulted in a reduction in maximal current (I max Ala ) of 45 Ϯ 4% at Ϫ70 mV (Fig. 3B). Although the K 0.5 Ala was also higher in Li ϩ than in Na ϩ at Ϫ70 mV, it again reflected the subsaturating cation concentration (see Fig. 3G), whereas the K 0.5 Ala in Li ϩ did not differ from that in Na ϩ between Ϫ90 mV and Ϫ150 mV (Fig. 3C). We conclude that alanine binding per se is independent of voltage and independent of the driving ion (Na ϩ or Li ϩ ).
The relationship of I max Ala to V m in Na ϩ medium was linear between Ϫ150 mV and Ϫ30 mV and linear again between Ϫ10 mV and ϩ50 mV (Fig. 3D). For any given V m , the alanineevoked currents were greater at higher Na ϩ concentrations and the current/voltage relationships were shifted to the right (Fig. 3E). Amino acid transport mediated by SNAT1 is therefore driven by the Na ϩ electrochemical gradient. Alanineevoked currents were saturable with Na ϩ concentration (Fig.  3F), and the Hill coefficient for Na ϩ (n H Na ) was 1, regardless of V m . SNAT1 displayed extremely high apparent affinity for Na ϩ . K 0.5 Na was exquisitely voltage-dependent, rising steeply from 2 Ϯ 0.1 mM at Ϫ150 mV to almost 30 mM at Ϫ10 mV (indeterminable at positive V m ). We conclude that Na ϩ binding is voltage-dependent.
Although the K 0.5 for Li ϩ was even more steeply dependent upon V m than was K 0.5 Na (Fig. 3G), the K 0.5 for Li ϩ approached that for Na ϩ at hyperpolarized V m and was not significantly different at Ϫ150 mV. We also noted that, at any given V m , I max Ala was always lower in Li ϩ medium than in Na ϩ (Fig. 3D). (I max Ala did not saturate at hyperpolarized potentials in Na ϩ but did in Li ϩ , even at potentials at which Li ϩ is in excess (Ϫ150 mV to Ϫ110 mV). We conclude that the V max of the transporter is significantly limited when Li ϩ is the driving ion compared with Na ϩ . SNAT1 is pH-sensitive, with amino acid transport being greatly inhibited at low pH (21, 22,34). We found that alanine evoked smaller currents at pH 5.5 than at pH 7.5 and no currents in the absence of Na ϩ at either pH (data not shown) indicating that H ϩ cannot activate amino acid transport in SNAT1. Consistent with this observation, SNAT1-mediated glutamine transport, although pH-sensitive, was independent of the H ϩ electrochemical gradient, as demonstrated by the application of a proton ionophore in transfected CV1 fibroblasts (21).
SNAT1 Transports Na ϩ and Amino Acid by a Simultaneous, Ordered Mechanism-The current evoked by 10 mM alanine in oocytes expressing SNAT1 (at Ϫ70 mV) was markedly temperature-dependent, ranging from Ϫ312 nA at Ϸ 21°C to Ϫ1387 nA at Ϸ 33°C. Arrhenius transformation of the temperature dependence data (Fig. 3H) yielded an activation energy (E a ) of 21.8 Ϯ 2.0 kcal⅐mol Ϫ1 , slightly higher than estimates of System A activation energy (15-16 kcal⅐mol Ϫ1 ) derived from other cell preparations (35,36), although it is not known which SNAT Ala . E-G, cation dependence of the currents evoked by L-alanine presented at a subsaturating concentration. E, I/V relationships for 0.2 mM L-alanine at 2, 10, and 100 mM Na ϩ . F and G, Na ϩ and Li ϩ saturation kinetics were determined from the currents evoked by 0.2 mM L-alanine at 1, 2, 5, 10, 20, 35, 50, 75, and 100 mM Na ϩ or Li ϩ . At V m ϭ Ϫ70 mV (F), the derived kinetic parameters for Na ϩ -activated currents were K 0.5 Na ϭ 9.5 Ϯ 1.3 mM, I max Na ϭ Ϫ98 Ϯ 4 nA, and n H Na ϭ 0.9 Ϯ 0.1 (r 2 ϭ 1.0); whereas for Li ϩ , K 0.5 Li ϭ 18.7 Ϯ 2.8 mM, I max Li ϭ Ϫ69 Ϯ 4 nA, and n H Li ϭ 0.9 Ϯ 0.1 (r 2 ϭ 1.0). The value of n H for either cation did not vary with V m (data not shown). G, voltage dependence of the half-maximal cation concentrations. H, temperature dependence of the currents associated with the Na ϩ /L-alanine cotransport and Na ϩ leak activities mediated by SNAT1 at Ϫ70 mV. Cotransport (Na ϩ /L-Ala, filled circles) is given as the current evoked by 10 mM L-alanine in 100 mM NaCl (pH 7.5). Na ϩ leak (inverted gray triangles) is given as the current induced by switching from Na ϩ -free (100 mM ChoCl) medium to 100 mM Na ϩ (pH 7.5) in the absence of amino acid. Data were fitted to the Arrhenius equation (Equation 2) with activation energy (E a ) of 21.8 Ϯ 2.0 kcal⅐mol Ϫ1 for Na ϩ /L-alanine cotransport (r 2 ϭ 0.97, ln A ϭ 43 Ϯ 4) and 7.8 Ϯ 0.7 kcal⅐mol Ϫ1 for the Na ϩ leak (r 2 ϭ 0.97, ln A ϭ 18 Ϯ 1).
isoforms were represented in those preparations. The high E a of Ϸ 22 kcal⅐mol Ϫ1 for SNAT1 is consistent with a "carrier" model in which a series of ligand-induced conformational changes execute Na ϩ /alanine cotransport (Fig. 9).
The mechanism of cotransport and the order of substrate binding was determined by analyzing the saturation kinetics for both Na ϩ and alanine as a function of the cosubstrate concentrations (Fig. 4). Our data indicated that both substrates are transported simultaneously, since the half-maximal concentrations (K 0.5 ) for both substrates (Na ϩ and alanine) rose substantially at low cosubstrate concentrations and were minimal when cosubstrate concentration was saturating (Fig. 4, A  and B). In contrast, for a consecutive carrier model, the K 0.5 is expected to rise as cosubstrate concentration is increased, since both substrates are effectively competing for the available carriers (37). I max for Na ϩ (I max Na ) was exquisitely dependent upon the cosubstrate (alanine) concentration whereas I max for alanine (I max Ala ) did not substantially differ with Na ϩ concentration (Fig. 4, C and D), i.e. the presence of saturating alanine was always sufficient to drive the transporter at maximal velocity regardless of Na ϩ concentration, whereas I max Na was always limited by the available alanine. Considering a carrier model in dynamic equilibrium (Fig. 9), these data demonstrate that alanine binds after the Na ϩ binding event.
SNAT1 Presteady-state Currents Are Observed in the Absence of Amino Acid-After step changes in V m were applied in the absence of amino acid, we observed presteady-state currents in oocytes expressing SNAT1, but not in control oocytes. The compensated presteady-state currents (Fig. 5, inset) were integrated with time, and the relationship of charge (Q) to V m could be described by a single Boltzmann function with apparent valence (z) of Ϫ0.7 and maximal charge (Q max ) Ϸ 9 nC (Fig. 5). We did not observe presteady-state currents associated with SNAT1 in the absence of Na ϩ (data not shown). Our observation of Na ϩ -dependent presteady-state currents in the absence of amino acid is consistent with our conclusions that voltagedependent Na ϩ binding precedes alanine binding in SNAT1. SNAT1 presteady-state currents decayed with half-time () of 4 -5 ms (data not shown); was distributed over V m with maximum of 4.6 Ϯ 0.1 ms at Ϫ26 Ϯ 8 mV.
SNAT1 Mediates a Discrete Cation Leak Pathway-SNAT1 exhibited an uncoupled cation transport pathway (leak) in the absence of amino acid substrate. Its presence was first apparent from the large inward currents (I ⌬Na ) observed in oocytes, expressing SNAT1 upon switching from Na ϩ -free to 100 mM NaCl in the absence of amino acid (see Fig. 1B). At Ϫ50 mV, I ⌬Na in oocytes expressing SNAT1 was 22-fold that observed for control oocytes whereas I ⌬Na was only doubled for SNAT2 (Fig.  6A). We found that I ⌬Na was concentration-dependent but did not saturate at physiological Na ϩ concentration (Fig. 6B). To determine the ionic species underlying the leak current in SNAT1, we applied a voltage-ramp protocol to measure reversal potentials (V r ) under varying ionic conditions (Fig. 6C). At 100 mM Na ϩ , we observed inward currents at polarized potentials and outward currents at positive potentials, with V r of Ϫ25 mV. Reducing the Na ϩ concentration reduced the magnitude of the inward current and shifted V r to more polarized potentials. The Nernstian shift in V r of ϩ55.4 Ϯ 4.0 mV per 10-fold increase in Na ϩ concentration indicated that the leak current was a Na ϩ conductance. Currents in Li ϩ were identical to those we we obtained for Na ϩ (Fig. 6C). Increasing K ϩ concentration from 2 mM to 10 mM (by equimolar exchange with choline in the presence of 100 mM Na ϩ ) or reducing pH from 7.5 to 5.5 (in the absence of Na ϩ ) had no significant effect upon V r (data not shown), indicating that neither K ϩ nor H ϩ are permeant.
These maneuvers were performed in the presence of 100 M niflumic acid in order to block currents mediated by endogenous Cl Ϫ channels (38,39). In preliminary investigations with SNAT1 in the absence of niflumic acid, varying the Na ϩ concentration resulted in non-Nernstian shifts in V r of about only ϩ20 mV/decade, as observed by Chaudhry et al. (22). This result was due to activation of the endogenous Cl Ϫ channel (blocked by niflumic acid) by an undetermined mechanism, as previously observed in oocytes expressing any one of several transporters or channels (40). The narrow selectivity of the cation current mediated by SNAT1 distinguishes it from the nonselective endogenous cation current (for which K ϩ was permeant) observed in the latter study. Whereas cotransport was very temperature-dependent, the SNAT1 Na ϩ leak displayed only modest temperature dependence: the low E a of 7.8 Ϯ 0.7 kcal⅐mol Ϫ1 (Fig. 3H) is consistent with channel activity. That the SNAT1 Na ϩ leak differs so much from cotransport in terms of saturability and temperature dependence suggests that the Na ϩ leak is mediated by a pathway that (i) possesses properties of a channel and (ii) is physically discrete from the cotransport pathway.
Na ϩ /Amino Acid Cotransport Is Coupled 1:1-Net charge uptake (the integral of the alanine-evoked current over 5 min) was directly proportional to 500 M L-[ 3 H]alanine uptake (in 100 mM Na ϩ ) at either Ϫ70 mV or ϩ10 mV (Fig. 7), validating our use of the amino acid-evoked current as a direct linear index of amino acid transport mediated by SNAT1, as was the case for SNAT2 (26). From the molar ratio of net charge uptake to [ 3 H]amino acid accumulation, we derived a Na ϩ /amino acid coupling coefficient (n) of 1.4 Ϯ 0.2 at Ϫ70 mV (Fig. 7) and concluded that coupling is 1 Na ϩ per 1 amino acid in tandem with uncoupled leaks. Hill coefficients (n H ) for the driving ion (Na ϩ or Li ϩ ) were always 1, whereas n H somewhat in excess of 1 might be expected in the case of 2 Na ϩ per 1 amino acid coupling stoichiometry, depending on the degree of cooperativity between cation-binding sites. That the coupling coefficient (n) significantly exceeded 1 was not due to an amino acidstimulated increase in endogenous Cl Ϫ channel activity, since this was blocked in these experiments by 100 M niflumic acid. Similar values were obtained in the absence of niflumic acid, and the current/voltage relationship for L-alanine did not appreciably differ in the presence or absence of niflumic acid (data not shown). The value of n also did not differ: (i) over a range of alanine concentrations (200 M to 1 mM), (ii) at either 5 or 100 mM Na ϩ , or (iii) when L-[ 3 H]glutamine was used in place of alanine (data not shown).
We considered the possibility that n Ͼ 1 reflected thermodynamic coupling of 2 Na ϩ per 1 amino acid. Increasing the amino acid concentration will increase the rate k 23 describing the step 2 3 3 in our model (Fig. 9). If the Na ϩ leak were accounted for by Na ϩ uniport activity (a direct pathway between states 2 and 5), the addition of amino acid might then block the uncoupled Na ϩ leak by driving the carrier in favor of Na ϩ /amino acid cotransport, such that n derived from this experimental procedure would underestimate the real coupling stoichiometry by the extent to which the Na ϩ uniport was inhibited by the addition of amino acid. By this reasoning, a coupling stoichiometry of 2 Na ϩ per 1 glucose was inferred for SGLT1 from n ϭ 1.6 at Ϫ70 mV (27). In the case of SGLT1 however n varied with voltage, since the current/voltage relationship of the phlorizininhibitable Na ϩ uniport differed from that of the sugar-evoked current (27). In contrast, we found that the SNAT1 Na ϩ /amino acid coupling coefficient was no different (n ϭ 1.4 Ϯ 0.3) at ϩ10 mV (Fig. 7). The Na ϩ leak is reduced at ϩ10 mV significantly more than is the amino acid-evoked current (see Figs. 3 and 4). Therefore, determination of n appears to be unaffected by the Na ϩ leak in SNAT1 and the channel-like properties of the SNAT1 Na ϩ leak distinguish it from the uniport activity observed in other Na ϩ -dependent transporters such as SGLT1. We conclude that Na ϩ and amino acid are thermodynamically coupled 1:1 and that a Na ϩ leak activity, with channel-like properties, proceeds independently of cotransport. The activation of an additional, uncoupled leak in the presence of amino acid (accounting for n Ͼ 1) warrants further investigation. Cellular Distribution of SNAT1 in the Central Nervous System-We compared SNAT1 immunoreactivity patterns with those of several cellular markers in order to determine the cellular distribution of SNAT1 and the neurotransmitter phenotypes of neurons in which SNAT1 is expressed in adult rat brain. Confocal laser-scanning double-immunofluorescence microscopy localized SNAT1 protein to many presumed glutamatergic and GABAergic neurons of the adult rat brain (Fig. 8). We did not find SNAT1 in astrocytes. SNAT1 staining was absent from those cells that were labeled with glial fibrillary acidic protein (GFAP) in the hippocampal formation or thalamus (Fig. 8, A and B) and SNAT1 also did not colocalize with GFAP in cerebellar granule cell cultures (21). However SNAT1 expression in brain was not confined to neurons, since SNAT1 was also abundantly expressed in the ependymal cells lining the ventricles (Fig. 8, A and C). We observed polarized expression of SNAT1 in ependymal cells, with SNAT1 being localized to the luminal plasma membrane (Fig. 8C). SNAT1 (with K 0.5 for glutamine of 0.3 mM) is well suited to a role in the ependymal uptake of glutamine from ventricular CSF, the glutamine concentration of which is 0.4 -0.5 mM (41)(42)(43). SNAT1 acting in concert with SNAT3 (SN1), which appears to be present on the luminal and abluminal membranes of these cells (14), may account for glutamine transport between CSF and brain compartments.

SNAT1 Is Localized to the Cell Bodies and Proximal Dendrites of Glutamatergic and GABAergic
Neurons-We observed moderate levels of SNAT1 immunoreactivity in hippocampus, enriched in the CA3 layer and dentate gyrus granule cells (Fig.  8, A and B). Higher power magnification of the CA3 region revealed SNAT1 immunoreactivity in pyramidal neurons known to be glutamatergic (Fig. 8B). SNAT1 was detected previously in developing primary cultures of hippocampal pyramidal neurons (44). Higher power magnification of the molecular layer of the hippocampus revealed SNAT1 immunoreactivity in GABAergic interneurons (data not shown).
For the most part, we found that neuronal expression of SNAT1 was confined to cell bodies and proximal dendrites. SNAT1 was not detectable at the axon nerve terminal in paraffin-embedded sections. This was readily apparent for glutamatergic pyramidal neurons of the CA3 region of the hippocampus (Fig. 8D) and in the cerebral cortex (Fig. 8E). There, dendrites of SNAT1-positive neurons were identified with the dendritic marker MAP2, which also weakly stains cell bodies. We observed in the pallidum abundant SNAT1 immunoreactivity (Fig. 8F) that was confined to the cell bodies and proximal dendrites of the large cells, which are known to be GABAergic (45). SNAT1 immunoreactivity did not colocalize with the synaptic vesicle markers VGLUT1 (Fig. 8, G and L), VGLUT2 (Fig. 8, H, J, and K), or synaptophysin (Fig. 8I) indicating that little or no SNAT1 is present in nerve endings in vivo under these conditions, despite contrasting findings in vitro. For example, preparations of rat brain synaptosomes, which are considered to be pinched-off nerve endings, contained significant amounts of SNAT1 protein according to Western blot analysis (21), and SNAT1 was also detected at the tips of growing axons and at sites of synaptic contacts in cultured hippocampal nerve preparations (44).
In many neurons SNAT1 immunoreactivity appeared punctate (Fig. 8, E, F, I, J, L, and N), indicating the presence of large intracellular pools of SNAT1. In the cerebral cortex (Fig. 8E) and pallidum (Fig. 8, F and I), fine puncta filled the cell body and extended to the proximal dendrites. This may explain the abundance of SNAT1 protein in microsomal fractions from rat brain preparations (21). In other neurons, such as those of the trigeminal mesencephalic nucleus, SNAT1 immunoreactivity clearly delineated the plasma membrane (Fig. 8H). However, discrete areas of SNAT1 immunoreactivity were apparent within the cell bodies of these primary sensory neurons, possibly representing sites of synthesis or regulation such as the trans-Golgi network or endosomal recycling centers. These mesencephalic neurons, known to be the primary afferents supplying the masticatory muscles, receive abundant input from VGLUT2-expressing neurons and may be part of a very active excitatory VGLUT2-encoded pathway. Our data indicated that the cellular expression of SNAT1 within the cell body differs among various neuronal populations and suggest that trafficking of the SNAT1 protein between intracellular pools and the cell surface may be regulated.
SNAT1 protein was abundantly expressed in other presumed glutamatergic neurons, such as those dominating the pontine nuclei and gigantocellular reticular nucleus (Fig. 8, G and O). These neurons contain VGLUT1 mRNA (46,47) and in part account for the presence of mossy fiber terminal-like VGLUT1 immunoreactivity in the granule layer of the cerebellum (48) and for VGLUT1-encoded innervation of motoneurons in the spinal cord (Fig. 8L) (49). SNAT1 was abundantly expressed in the red nucleus (Fig. 8M), an excitatory projection to the spinal cord, and in the thalamus (Fig. 8A), where both glutamatergic and GABAergic cell bodies are numerous. SNAT1-immunoreactive cell bodies were observed in the reticular thalamus where GABAergic neurons predominate (Fig. 8K). We observed moderate VGLUT2 innervation of this region, possibly on the dendritic portions of these SNAT1-positive neurons. In cerebellum, SNAT1 protein was abundant in GABAergic Purkinje cell bodies (Fig. 8J) and SNAT1 mRNA was detectable in the Purkinje cell layer (21, 22). We detected SNAT1 immunoreactivity in interneurons, probably GABAergic, of the molecular layer of the cerebellum (Fig. 8J). SNAT1 immunoreactivity did not colocalize with the VGLUT2 immunoreactivity observed in a subset of mossy fiber terminals in the granule layer and the climbing fiber terminals that line the dendrites of the Purkinje neurons. The glutamatergic granular cell layer exhibited lower levels of SNAT1 immunoreactivity than did the GABAergic cell populations in the cerebellum. Whereas SNAT1 mRNA was detected in the granular layer of the cerebellum (21, 22), SNAT2 mRNA was the dominant isoform expressed in cerebellar granule interneurons (22,26,50). Whereas SNAT1 expression in the granule neurons in vivo appeared low, cultured cerebellar granule cells (day 20) expressed high levels of SNAT1 protein (21).

SNAT1 Is Expressed in Subsets of Cholinergic and Dopaminergic Neurons-SNAT1
was not restricted to glutamatergic and GABAergic neurons of the central nervous system but was also found in subsets of cholinergic and dopaminergic neurons. We found abundant expression in the cholinergic motoneurons of the spinal cord (Fig. 8L), a major target for VGLUT2-expressing glutamatergic neurons (29). SNAT2 mRNA is also abundantly expressed in these cells (26). Here, VGLUT1-positive synaptic terminals abut dendritic fields and cell bodies that are strongly SNAT1-positive (Fig. 8L). The high levels of SNAT1 expression in motoneurons may account for the very high glutamate content of motoneuron cell bodies and nerve terminals (51)(52)(53). Double immunostaining with the vesicular acetylcholine transporter VAChT (which stains cholinergic cell bodies and nerve terminals) revealed that SNAT1 was only present at low levels in most other cholinergic cells in the brain, including those of the striatum, globus pallidus, and basal forebrain systems (data not shown).
Catecholaminergic neurons of the A1 cell group of the brainstem (Fig. 8O) and the olfactory bulb (data not shown) did not express SNAT1. However, we identified expression of SNAT1 in subsets of dopaminergic cells of the substantia nigra, identified by tyrosine hydroxylase immunoreactivity (Fig. 8, M and  N). No appreciable SNAT1 expression was detected in the dopaminergic neurons of the ventral tegmentum or medial portion of the substantia nigra pars compacta. SNAT1 was evident, in some cases at very high levels, in dopaminergic neurons in the ventral and lateral portions of the pars compacta and the pars reticulata (Fig. 8N). The expression of SNAT1 in ventral midbrain dopaminergic neurons is consistent with the high levels of glutamate and PAG activity detected in the substantia nigra pars compacta (54,55) and raises the possibility that these neurons synthesize and secrete neurotransmitter glutamate. Cultured dopamine midbrain neurons exhibit a mixed dopaminergic-glutamatergic phenotype (56,57), and single dopaminergic nigrostriatal neurons may form two chemically distinct synaptic types (58). Notably, SNAT1 and tyrosine hydroxylase immunoreactivity in the substantia nigra pars compacta colocalized in those lateral and ventral neurons that are most susceptible in Parkinson's disease (59). The neurotoxic effects of glutamate (during impaired neuronal energy metabolism) are known to play a role in the etiology of neurodegeneration in Parkinson's disease (60), as well as in amyotrophic lateral sclerosis (ALS) (61,62).

Mechanisms of SNAT1-mediated Amino Acid Transport-
Coupling of the amino acid flux through SNAT1 to the Na ϩ electrochemical gradient ensures unidirectional inward transport of amino acid at resting potential, and no reversal was observed at the V m expected during the action potential (see Fig. 3A). The SNAT1 transport cycle does not require any other ionic species; however, Li ϩ can substitute for Na ϩ in driving transport and (contrary to common thinking) classic System A is tolerant of a Li ϩ for Na ϩ substitution (63). However, we propose that Li ϩ tolerance is not in itself a useful criterion with which to classify amino acid transporters, since our data for SNAT1 illustrated that: (i) Li ϩ tolerance is incomplete and (ii) the apparent effectiveness of Li ϩ as a replacement for Na ϩ depends both on membrane potential and on the parameters (e.g. K 0.5 , I max ,or absolute current) that are used in comparisons (Fig. 3). Our data are consistent with a carrier model describing SNAT1-mediated amino acid transport as a series of temperature-dependent, ligand-induced conformational changes (Fig. 9). Na ϩ binding is voltage-dependent and precedes binding of the amino acid and its simultaneous translocation. SNAT1 exhibited presteady-state currents, which can be used to estimate transporter density (64). Using the relationship Q max ϭ N T ⅐z⅐e (for which e is the elemental charge), we estimate the number of functional transporters (N T ) in the plasma membrane at Ϸ 10 11 per oocyte. In analogy to other Na ϩ -or H ϩ -dependent transporters (65-68), we attribute presteady-state currents for SNAT1 to two steps within the transporter cycle: (i) translocation of the empty, charged carrier and (ii) binding/dissociation of Na ϩ within the membrane electric field. Presteady-state currents decayed with half times () of 4-5 ms, whereas the turnover rate for L-alanine transport was up to 200 s Ϫ1 at Ϫ150 mV. On the basis of our model (Fig.  9), we therefore anticipate that carrier reorientation (step 6 3 1) and/or Na ϩ binding (step 1 3 2) may be rate-determining at saturating voltage. However, another step in the transport cycle must be rate-determining at depolarized V m .
Heterologous expression of SNAT1 in oocytes ( Fig. 9) was also associated with stimulation of an endogenous chloride channel activity (that was blocked by niflumic acid) and a cation conductance selective for Na ϩ and Li ϩ . The exclusion of protons from both the leak and cotransport pathways in the System A transporters distinguishes this branch of the SLC38 gene family from the System N subtypes, which mediate the exchange of H ϩ for Na ϩ plus glutamine (7, 69 -71). Our conclusion that Na ϩ /amino acid coupling stoichiometry is 1:1 relies upon our finding that the Na ϩ leak is discrete from the Na ϩ flux through the cotransport pathway. Evidence to this effect includes the observation of channel-like properties of the Na ϩ leak (i.e. nonsaturability and low temperature dependence) that distinguish it from Na ϩ uniport. Further supporting the view that the Na ϩ leak is discrete from the cotransport pathway is our observation of a substantial Na ϩ leak in an SNAT1 mutant (H391A) in which amino acid transport is severely impaired; of particular note, addition of amino acid did not inhibit the Na ϩ leak in H391A-SNAT1 even though K 0.5 Ala was unaffected by the mutation. 2 Our model and the Na ϩ /amino acid coupling stoichiometry of SNAT1 could be further resolved pharmacologically if inhibitors were available to block either the leak or the cotransport pathways. For example, our model would predict that the Na ϩ leak will escape inhibition by a high affinity, non-transported amino acid analogue (antagonist) that can inhibit Na ϩ /amino acid cotransport. Meanwhile, the differences in substrate selectivity between SNAT1 and SNAT2 may permit the development of specific blockers to distinguish between the two isoforms, and specific blockers could be advantageous in investigating glutamate-GABA/glutamine cycling with NMR or immunocytochemical approaches in vivo.
We have concluded that the cation leak observed in oocytes expressing SNAT1 is a function of the SNAT1 protein ( Fig. 9) since the cation selectivity (Na ϩ , Li ϩ Ͼ Ͼ H ϩ , K ϩ ) of the leak: (i) exactly matches the ion dependence of amino acid cotransport and (ii) differs from the predominant endogenous cation channel, which is also permeant to K ϩ (40). To explore the possibility that the cation leak was an artifact of overexpression in the oocyte, we monitored the time course and cRNA-concentration dependence of the Na ϩ leak and cotransport currents, since differences may be expected if the cation leak were as a result of saturation of some component or subunit endogenous to the oocyte. The Na ϩ leak and cotransport currents each began to rise significantly at 12-24 h, peaked at 48 h, and remained unchanged at 60 h (data not shown). Also, expression of both components showed comparable dependence on the amount of cRNA injected (5, 10, 20, or 50 ng) (data not shown). The parallel expression of the Na ϩ leak and amino acid cotransport FIG. 9. Transport activities associated with overexpression of SNAT1 in Xenopus oocytes. Our data are consistent with an ordered, simultaneous transport model for SNAT1 in which Na ϩ /amino acid cotransport is coupled 1:1 and described by a series of temperature-dependent, ligand-induced conformational changes. Our data revealed that Na ϩ binding is voltage-dependent (an ion well effect) and precedes the binding of the amino acid. The order of dissociation at the internal face is assumed. In this six-state model, the rate k xy describes step x 3 y, and at least the rates k 61 , k 12 , k 21 , and k 16 should include voltage-dependent terms. We also observed significant uncoupled ion fluxes including a Na ϩ leak (g Na ) that displayed properties of a channel and appears to be discrete from the Na ϩ flux through the cotransport pathway. SNAT1 expression was also associated with stimulation of endogenous Cl Ϫ channel activity (g Cl ) by an unknown mechanism (? dotted arrow) that may involve Ca 2ϩ activation. components suggests that the cation conductance associated with SNAT1 expression in oocytes is in fact a property of SNAT1 and not an artifact of the oocyte expression system. However, until the cation leak can be demonstrated in neurons, its physiological significance will remain unknown. Nevertheless, we speculate that the presence of SNAT1 at the neuronal plasma membrane and resultant Na ϩ fluxes (whether via the leak or cotransport pathways) will place a significant energy burden on the cell and may increase the vulnerability of neurons that are metabolically compromised. This highlights the importance of the possible regulation of protein synthesis and trafficking between intracellular compartments and the plasma membrane, as illustrated for the peripheral System A isoform SNAT2 (72). Little is known about the regulation of SNAT1 in brain; however System A activity in cultured cerebellar granule cells (5) did not display the responsiveness to insulin observed for SNAT2 elsewhere (73,74). The abundance of SNAT1 within intracellular compartments of some cells (e.g. cerebral cortex and pallidum) but at the plasma membrane of others (e.g. mesencephelon and ependyma) may reflect protein synthesis de novo, plasma membrane recruitment, or recycling of SNAT1 according to the demand for glutamine, as for SNAT2 (72). Long term sensitization of glutamatergic neurotransmission in Aplysia is accompanied by increased neuronal glutamine uptake (75), suggesting that glutamatergic activity may play a role in regulation of glutamine uptake. The enrichment of SNAT1 in the cell body and proximal dendrites as opposed to the axon terminal could also be expected to contribute to the acute regulation of electrical activity at the soma and neuronal excitability as a result of the depolarizing action of the SNAT1mediated Na ϩ fluxes.
A Specialized Role for SNAT1 in the Glutamate/GABA-Glutamine Cycle-Whereas low-level expression of SNAT1 was detected in several tissues (21, 22, 73), SNAT1 was highly enriched in specific neuronal populations within the CNS, but SNAT1 was not present in astrocytes. We expect L-glutamine to be the principal substrate for the neuronal SNAT1 in vivo since its interstitial concentration in the brain, estimated at 0.2-1.0 mM (7,43,77), significantly exceeds that of any other amino acid. SNAT1 efficiently transports L-glutamine, and its K 0.5 Gln of Ϸ 0.3 mM is well suited to the role of glutamine uptake from the neuronal environment. These considerations, and the localization of SNAT1 almost exclusively to neuronal populations of known glutamatergic or GABAergic phenotypes, support a specialized role for SNAT1 in the glutamate/GABA-glutamine cycle. Although we observed lower levels of SNAT1 in glutamatergic granule cells of the cerebellum, these cells may upregulate SNAT1 when cultured in vitro under depolarized conditions (21) and the often punctate appearance of SNAT1 immunoreactivity in neurons in vivo implicates a role for regulated transporter recruitment to the plasma membrane. The highest levels of SNAT1 immunoreactivity were observed in the cell bodies and proximal dendrites of neurons, whereas axon terminals in adult rat appeared to lack SNAT1. Therefore the punctate SNAT1 immunoreactivity observed in most perikarya is probably not destined for nerve terminals, and the soma and proximal dendrites may represent the primary site of SNAT1-mediated glutamine uptake in these neurons. It is possible that the epitope recognized by our antibody against SNAT1 is masked at the nerve terminal in the adult brain in vivo. However, the polyclonal SNAT1 antibody was raised against a fusion protein containing 63 amino acids of the SNAT1 sequence. SNAT1 immunoreactivity in hippocampal neuronal projections in vitro (44) may be a property unique to primary culture of differentiating neurons. The abundance of SNAT1 in microsomal fractions from rat brain synaptosomes (21) is consistent with the SNAT1 immunoreactivity pattern in situ observed in this study since synaptosomal preparations include cleft and postsynaptic components (including somatodendritic synapses) in addition to presynaptic terminals (78). We do not consider expression at the nerve terminal prerequisite to a specialized role for SNAT1 in replenishing the glutamate and GABA neurotransmitter pools. That SNAT1 expressed at the cell body and proximal dendrites could fulfill this role is supported by the observations that: (i) cell bodies in many putative glutamatergic neurons possess high activity of the phosphate-activated glutaminase required for glutamate synthesis (79), (ii) the liver-type glutaminase is also localized to neuronal nuclei (80), and (iii) glutamate or GABA concentrations are uniformly high throughout the cytoplasm of glutamatergic or GABAergic neurons (54,81) suggesting that rapid diffusion occurs. A role in the glutamate/GABA-glutamine cycle would make a SNAT1 a potential target for modulating presynaptic glutamatergic/GABAergic function.
Alanine, although present in brain interstitial fluid at only one fifth the concentration of glutamine (43,77), is also a very efficient substrate of SNAT1. Modest uptake of alanine into glutamatergic neurons may also contribute to maintaining neurotransmitter glutamate pools, especially during the recovery from ischemia/hypoxia, when alanine concentration rises and glutamate concentration falls (82), since transamination of ␣-ketoglutarate plus alanine can replenish the glutamate pool released during potassium-induced depolarization (i.e. loss of neurotransmitter glutamate) in cerebellar granule cells (83).
Additional Roles for Neuronal SNAT1-The recent localization of liver-type glutaminase in neuronal nuclei (80) prompts us to consider additional neuronal glutamine requirements with discrete control pathways, and alternative roles for neuronal SNAT1. The expression of SNAT1 in ventral midbrain neurons considered to be dopaminergic may well be explained by the existence of a mixed glutamatergic-dopaminergic phenotype in those cells, as supported by studies in vitro (56 -58). A mixed phenotype has also been considered for cholinergic motoneurons (52). Alternatively, the expression of SNAT1 in these neuronal populations may point to additional roles for SNAT1-mediated uptake, not only of glutamine but also of other amino acids efficiently transported by SNAT1. For example, cysteine and glycine, together with glutamate, are required for neuronal synthesis of glutathione, a major antioxidant responsible for the detoxification of reactive oxygen species readily generated within the brain (84,85). Locally high concentrations of cysteine and glycine may result from cleavage of astrocyte-derived dipeptides (e.g. CysGly) by the ectopeptidase aminopeptidase N, which, like SNAT1, is present on the neuronal cell body (86). SNAT1 could therefore supply neurons with the components of glutathione synthesis. In fact glutamine, through this pathway, may play an important role in regulation of oxidative metabolism in brain (87). Notably, the ventral midbrain neurons that express SNAT1 are known to be particularly susceptible to oxidative stress resulting from autoxidation of dopamine. Finally, SNAT1 may also serve neurons with an elevated demand for glutamine as a metabolic fuel, possibly via ␣-ketoglutarate, which may then enter the tricarboxylic acid cycle (18,19). It is of note that there is also significant expression of SNAT1 in the heart (22,23,73), where glutamine metabolism appears to be crucial to postischemic recovery (76).
The functional characterization of SNAT1 in terms of substrate profile, ion dependence, thermodynamics, and kinetics provides the foundation for structure-function analysis of SNAT1 and the development of specific inhibitors, and a knowledge of the properties and cellular distribution of SNAT1 in the