Molecular and Functional Analysis of a Novel Neuronal Vesicular Glutamate Transporter*

,

Neurotransmission depends on the regulated exocytotic release of vesicular transmitter molecules to the synaptic cleft, where they interact with postsynaptic receptors that subsequently transduce the information. Two types of neurotransmitter transporters have been identified based on membrane localization on plasma membrane or vesicular membrane. Removal of the transmitter from the synaptic cleft results in termination of the signal, and this requires destruction of transmitter or reuptake of transmitter back to the presynaptic terminal or glial cells via a sodium-dependent uptake system on the plasma membrane (1). Packaging and storage of neurotransmitters into specialized secretory vesicles in neurons ensures their regulated release. This storage is also crucial for protecting the neurotransmmitter molecules from leakage or intraneuronal metabolism and for protecting the neuron from possible toxic effects. This process is mediated by specific transporters on the vesicular membranes. At least four different types of vesicular transporters have been functionally identified that are specific for transport of classic neurotransmitters: monoamines, acetylcholine, ␥-aminobutyric acid (GABA), and glutamate (2,3). Unlike the plasma membrane transporters, which rely on a sodium gradient across the plasma membrane, all of these vesicular transport processes depend on the proton electrochemical gradient (⌬ Hϩ ) 1 generated by a Mg 2ϩ -activated vacuolar H ϩ -ATPase (V-ATPase) on the vesicular membrane (4). When protons are pumped into the vesicular lumen, a proton gradient (⌬pH) and a membrane potential (⌬) occur across the membrane to form ⌬ Hϩ, which favors the exchange of luminal protons for cytoplasmic transmitter. The transport of monoamines and acetylcholine rely predominantly on ⌬pH (5,6), whereas the accumulation of GABA depends on both ⌬ and ⌬pH (7,8). In the case of glutamate, there is disagreement over whether the transport of glutamate is driven by ⌬ only (9,10,11) or by both ⌬ and ⌬pH components of ⌬ Hϩ (12,13).
In addition to the differences in driving force, kinetic properties, substrate specificity, and ion dependence also clearly distinguish vesicular glutamate transport from high affinity plasma membrane glutamate transport. The vesicular ATPdependent glutamate transporter is specific for glutamate, is stimulated by millimolar concentrations of chloride, and has a low affinity for uptake (K m about 1-2 mM) (12,14). These functional characteristics are in contrast to those of the plasma membrane glutamate transporter, which is sodium dependent, accepts aspartate as well as glutamate, and has a high affinity for glutamate with K m in the 3-20 M range (15,16).
Although vesicular transporters for other neurotransmitters have been intensively studied at mechanistic, biochemical, and molecular levels, molecular cloning of vesicular glutamate transporters has only occurred recently. BNPI, a brain-specific sodium-dependent phosphate cotransporter originally characterized as a plasma membrane transporter, was recently localized on vesicular membranes of small synaptic vesicles in neurons (17,18) and further functionally characterized as a vesicular glutamate transporter called VGLUT1 (18,19). Uptake of glutamate by VGLUT1 has all of the functional characteristics previously reported for vesicular glutamate transporter with ⌬ as the predominant driving force. However, only a subset of glutamatergic neurons expresses VGLUT1 (17,20). Moreover, in Caenorhabditis elegans, loss-of-function muta-tions in the VGLUT1 orthologue eat-4, lead to impairment but not to a complete loss of glutamate-mediated neurotransmission (21,22). These findings imply that an additional vesicular glutamate transporter(s) may exist. To identify new vesicular glutamate transporter isoforms, we searched the Genbank TM Expressed Sequence Tags (EST) data base using human and rat VGLUT1 sequences (23,24), and we found a mouse EST clone, which has ϳ49% homology with human and rat VGLUT1. We report here the molecular cloning and functional characterization of VGLUT2, the second identified vesicular glutamate transporter with different transport characteristics and cellular localization as compared with VGLUT1.

EXPERIMENTAL PROCEDURES
Chemicals-N,NЈ-dicyclohexylcarbodiimide (DCCD) was purchased from Fisher Scientific. All other chemicals were obtained from Sigma.
Isolation of cDNA Encoding Mouse VGLUT2-The human and rat VGLUT1 sequences were used to search the GenBank TM EST data base. One query gene (GenBank TM number AI 841371) was identified from the mouse EST data base that has ϳ49% nucleotide identity with human and rat VGLUT1. This EST clone was obtained from Research Genetics (Huntsville, AL) and sequenced. The sequence was assembled and analyzed, and sequence comparisons were run using Omega software (version 1.1.3) and GenBank TM BLAST searches. The full-length cDNA insert was excised with EcoRI and NotI and subcloned into the same sites of the pSPORT1 vector (Lifetech) and further subcloned into the pIRES2-EGFP vector (CLONTECH, Palo Alto, CA) at EcoRI and BamHI sites, resulting in VGLUT2-pIRES2-EGFP (the name VGLUT2 was given after the determination of transport characteristics).
Expression of VGLUT2 in PC12 Cells-A rat pheochromocytoma cell line, PC12 cell (ATCC), which was used to characterize glutamate transport by VGLUT1 (19), was cultured in Dulbecco's modified Eagle's F-12 medium, supplemented with 20% fetal bovine serum, and 1% penicillin and streptomycin. VGLUT2-pIRES2-EGFP or pIRES2-EGFP without a cDNA insert (10 g) was transfected into cells by the lipofection method (25). Transfected cells were selected with 800 g/ml G418, and the most highly expressing cells were identified by fluorescence microscopy and selected for further analysis.
Preparation of Membrane Fractions from Transfected Cells and Glutamate Uptake-Stably transfected cells were washed twice with icecold phosphate-buffered saline and homogenized in 0.32 M sucrose and 10 mM HEPES-KOH (pH 7.4) containing protease inhibitors. The resulting homogenate was pelleted at 1000 ϫ g for 5 min to remove nuclei and at 27,000 ϫ g for 35 min to remove heavier membranes. The supernatant containing lighter membranes, including synaptic-like microvesicles, was sedimented at 210,000 ϫ g for 1 h, and the pellet was resuspended in 0.32 M sucrose, 4 mM MgSO 4 , and 10 mM HEPES-KOH, pH 7.4 (uptake buffer) at a final concentration of 10 mg of protein/ml. Plasma membranes were prepared as previously described (26) and were also resuspended in uptake buffer at 10 mg protein/ml.
To start the transport assay, 100 g of vesicular membrane or plasma membrane proteins in uptake buffer were mixed with 4 mM ATP, 4 mM KCl, and 50 M L-[ 3 H]glutamate (potassium salt, 0.4 Ci/mmol, Amersham Pharmacia Biotech) at room temperature for varying times, with other additions when indicated. Uptake was stopped by the addition of 4 ϫ 2 ml of ice-cold 0.15 M KCl and immediate filtration. Radioactivity was determined with a liquid scintillation spectrophotometer.
Northern Blot Analysis-A mouse, multiple-tissue Northern blot was purchased from Origene (Rockville, MD). Each lane was equally loaded with 2 g of mRNA from different tissues. The blot was hybridized with [ 32 P]dCTP-labeled VGLUT2 cDNA-specific probes generated by polymerase chain reaction (nucleotides 2216 -2489 from the 3Ј-untranslated region, which has low homology with VGLUT1). The filter was hybridized overnight at 42°C in 50% formamide and washed under high stringency conditions (0.1ϫ sodium chloride/sodium citrate, 0.1% SDS at 65°C) for 1 h (27).
In Situ Hybridization of Mouse Brain-In situ hybridization was performed as previously described (27), using paraffin-embedded mouse brain sections from Novagene (Madison, WI). The same polymerase chain reaction products used for Northern blots were subcloned in both orientations into the pCR II plasmid vector using the TA cloning kit (Invitrogen). cRNA probes were transcribed from these templates with T7 RNA polymerase using the MAXIscript in vitro transcription kit (Ambion, Austin, TX). Depending on the orientation of the insert, either sense (used as a negative control) or antisense (used as experimental) probes were generated. The cRNA transcripts were then labeled with biotin using the BrightStar Psoralen-Biotin labeling kit (Ambion). The biotinylated probes were hybridized to brain sections following the mRNAlocator-hyb kit protocol (Ambion). Briefly, mouse brain sections were rehydrated and predigested with 40 g/ml of proteinase K for 30 min and then hybridized with 5 g/ml of either antisense or sense cRNA probes at 55°C for 4 h. Posthybridization washes were carried out at 55°C three times for 4 min. The probes were detected using Ambion's mRNAlocator-biotin kit, and slides were photographed by standard light microscopy.

Molecular
Cloning of Mouse VGLUT2-We performed a search of the EST data base with the nucleotide sequence of human and rat VGLUT1 (BLASTx program). One mouse EST clone from amygdala exhibited significant similarity (ϳ49%) with human and rat VGLUT1. By sequencing the clone from Research Genetics, we obtained a 2528 bp cDNA with an open reading frame of 1746 bp with 638 bp of 5Ј-untranslated sequence and 134 bp of 3Ј-untranslated sequence, which we named VGLUT2 (GenBank TM accession number AF324864). The predicted protein has ϳ75-80% amino acid identity with human and rat VGLUT1 ( Fig. 1 and data not shown). Particularly, VGLUT2 showed 97% amino acid identity to a recently cloned human differential sodium-dependent phosphate cotransporter, called DNPI (28). Hydropathy analysis suggests the presence of 8 -10 membrane-spanning domains.
Functional Characteristics of VGLUT2-We first determined the transport activity of different membranes from VGLUT2 or vector-only transfected cells. In the presence of 4 mM ATP, chloride and MgSO 4 , conditions that optimize glutamate accumulation by native synaptic vesicles (12,29), vesicle membranes from VGLUT2-IRES2-EGFP-transfected cells exhibited 5-to 6-fold increased uptake. Plasma membrane glutamate uptake from VGLUT2-IRES2-EGFP-transfected cells did not show any differences from that of IRES2-EGFP-transfected cells ( Fig. 2A). The semiquantitative estimates of initial velocities of uptake were plotted in Fig. 2B. Kinetic experiments indicated that VGLUT2-mediated transport is dose-dependent and saturable (Fig. 2B). In one experiment depicted in Fig. 2B, the maximum velocity (V max ) was 1405 pmol/min-mg protein, and the apparent affinity constant (K m ) for glutamate was 1.2 mM. In three separate batches of membranes, the V max was 1219 Ϯ 109 pmol/min-mg protein, and the K m for glutamate was 1.1 Ϯ 0.2 mM. The K m is similar to that of glutamate transport by native synaptic vesicles as well as VGLUT1 (1ϳ2 mM), whereas it is in contrast to that of plasma glutamate transporters, which exhibit a K m for glutamate between 3 and 20 M (15, 16).
In addition to kinetic properties, vesicular glutamate transport has a number of distinct characteristics from plasma membrane glutamate transport. Accumulation of glutamate into vesicular membranes depends on the vacuolar Mg 2ϩ -ATPase, which utilizes ATP, whereas plasma membrane glutamate transport depends on energy provided by the sodium gradient. VGLUT2-mediated glutamate transport is ATP-dependent, as substitution of ATP with ADP or AMP abolished glutamate transport (Fig. 2C). To specify the type of ATPase involved in glutamate uptake by VGLUT2, we tested inhibitors for different types of ATPases in transport assays. N-ethylmaleimide (NEM), DCCD, and bafalomycin A1, inhibitors for vacuolar Mg 2ϩ -ATPase dramatically inhibited glutamate uptake. The plasma membrane and mitochondrial ATPase inhibitors, vanadate, oligomycin B, and ouabain, had no effect on glutamate uptake by VGLUT2 (Fig. 2C), which is in agreement with previous work (10,30).
Another feature of the vesicular glutamate transporter that segregates it from other neurotransmitter transporters is the marked stimulation seen in the presence of low, physiologically relevant concentrations of chloride (12,13,29,30). At low concentrations (1-5 mM), chloride stimulates transport, whereas higher concentrations (above 10 mM) inhibit transport. Fig. 2D shows that low concentrations of chloride (1-4 mM) stimulated glutamate uptake by VGLUT2, whereas concentrations higher than 10 mM attenuated the stimulatory effect, which is consistent with previous findings.
Substrate specificity was also used to distinguish vesicular and plasma membrane glutamate transporters. Vesicular transporters recognize only glutamate, whereas plasma membrane transporters recognize both glutamate and aspartate. Uptake of [ 3 H]glutamate by VGLUT2 was partially inhibited by D-glutamate, but not by a number of other compounds including aspartate, glycine, GABA, glutamine, and phosphate (Fig. 2E), which is consistent with known characteristics of glutamate transport by vesicular membranes. Evan blue, a potent inhibitor of vesicular glutamate transport, significantly blocked the uptake mediated by VGLUT2. In addition, 1 M DIDS, an anion channel blocker, significantly inhibited VGLUT2-mediated glutamate uptake (Fig. 2E). Uptake by VGLUT2 is also pH-dependent, with the highest glutamate uptake occurring between pH 7.0 and 7.5 (Fig. 2F).

Components of the Electrochemical Proton Gradient Involved in VGLUT2-mediated Glutamate Transport-To gain information about what component(s) of ⌬ Hϩ is involved in VGLUT2
transport, we determined the effect of ionophores on VGLUT2mediated glutamate transport. The doses of ionophores were selected by testing the minimally effective doses in preliminary studies (data not shown). As shown in Fig. 3, carbonyl cyanide p-(trifluoromethoxy)phenylhydrazone (FCCP), a protonophore acting as mobile carriers of H ϩ -ions and dissipating both the ⌬ and ⌬pH components of ⌬ Hϩ , abolished glutamate uptake by VGLUT2. The K ϩ ionophore valinomycin, which reduces only ⌬ without changing ⌬pH in the presence of 4 mM KCl, deceased glutamate uptake by VGLUT2. Nigericin is an electroneutral ionophore that exchanges K ϩ for H ϩ , thus eliminating ⌬pH whereas allowing ⌬ to increase when the H ϩ -ATPase is working. Surprisingly, nigericin inhibited 40% of VGLUT2mediated glutamate uptake in the presence of 4 mM KCl and subsaturating concentrations of glutamate. This observation is distinct from that of VGLUT1 in that nigericin increased VGLUT1-mediated glutamate uptake under the same conditions (19). But this is in agreement with previous results on synaptic vesicles (13,31). Furthermore, the addition of both nigericin and valinomycin together essentially abolished glutamate uptake by VGLUT2.
Tissue and Cellular Distribution of VGLUT2 mRNA-Mouse VGLUT2 expression was examined by probing poly(A) ϩ RNA from heart, brain, spleen, lung, liver, skeleton muscle, kidney, and testis. Northern blot analysis using mouse VGLUT2-specific probes detected a single transcript of ϳ3.8 kbp, with a strong hybridization signal observed only in brain (Fig. 4). In situ hybridization was then carried out to determine the regional and cellular expression of VGLUT2 mRNA in the mouse brain (Fig. 5). VGLUT2 is highly expressed in neuron-rich regions of the cerebral cortex (Fig. 5A), the hippocampus (Fig.  5B), and the hypothalamus (not shown). It is notable that VGLUT2 is highly expressed in neurons of the thalamus (Fig.  5, C and F), whereas VGLUT1 showed no or very low expression in the thalamus (20,28). Very little if any specific signal was observed with sense probe (Fig. 5, G and H). Although the signal was present throughout layers II-VI of the cortex, VGLUT2 is more abundant in layers III to V, where a distinct signal is observed in pyramidal neurons (Fig. 5D). The hippocampus is very densely labeled compared with other brain regions. At higher magnification, VGLUT2 can be seen to be concentrated in pyramidal neurons of the hippocampus (Fig.  5E). Moreover, VGLUT2 mRNA is mainly localized to the cell bodies (Fig. 5, D, E, and F). DISCUSSION Glutamate is the major excitatory neurotransmitter in the mammalian central nervous system. It has been shown that glutamate can be accumulated into synaptic vesicles to concentrations of at least 60 mM through an ATP-dependent mechanism (32,33). The vesicular transport system, in contrast to the sodium-dependent plasma membrane transport system, is specific for glutamate, has a low affinity for glutamate, and is stimulated by physiologically relevant concentrations of chloride (12,13,29,30). These unique properties of the vesicular glutamate transporter are conserved in such diverse vertebrates as fish, reptiles, amphibians, and mammals (34). Consistent with these observations, glutamate was transported by VGLUT2 in the absence of sodium. Furthermore, the transport required ATP and was inhibited by specific inhibitors of the V-ATPase. Glutamate transported by VGLUT2 is saturable with a K m of 1.1 mM, which is consistent with characteristics of the vesicular glutamate transporter. In addition, VGLUT2 is specific for glutamate and does not transport aspartate or other amino acids.
The requirement of low Cl Ϫ concentrations is a significant property of vesicular glutamate transport. The presence of chloride at low concentrations (1-5 mM) is essential for the uptake of glutamate, with substantially lower transport activity observed at higher and lower levels (9,13,29). This biphasic effect of chloride and the DIDS-sensitive chloride dependence of glutamate uptake, have suggested that the vesicular glutamate transporter possesses an anion binding site, distinct from the substrate binding site, which regulates transporter activity (35). Chloride showed a similar effect on glutamate uptake by VGLUT2. We also observed a 70% inhibition in VGLUT2mediated glutamate uptake by 1 M DIDS, which is consistent with previous findings. Taken together, these functional characteristics suggest that VGLUT2 functions as a vesicular glutamate transporter.
Based on the high homology (97% at the amino acid level) between mouse VGLUT2 and human DNPI, we surmised that DNPI is the human homologue of mouse VGLUT2. By using Xenopus oocytes expressing hDNPI, Aihara et al. (28) observed an ϳ75% increase in sodium-dependent phosphate uptake. However, in our substrate specificity experiments, high concentrations of phosphate did not inhibit glutamate uptake by VGLUT2, suggesting that phosphate was not transported by the same mechanism as glutamate. This is a similar situation to VGLUT1, where weak sodium-dependent phosphate uptake was observed in oocyte plasma membranes expressing VGLUT1 (23,24), and relatively strong ATP-dependent glutamate uptake was observed in vesicular membranes expressing VGLUT1 (18,19). The relationship between glutamate and phosphate transport by VGLUT1 and VGLUT2 remains unclear. As has been proposed for VGLUT1 (19), VGLUT2 may be a bifunctional transporter, which functions as a phosphate transporter at the plasma membrane and a glutamate transporter in synaptic vesicles.
Whereas it is known that vesicular glutamate transport is driven by an electrochemical proton gradient generated by V-ATPase, the precise manner in which the glutamate transporter and V-ATPase operate is currently under debate. To assess the driving forces of glutamate transport mediated by VGLUT2, we tested the effect of different ionophores on VGLUT2-mediated glutamate uptake. In the presence of 4 mM KCl, the K ϩ ionophore valinomycin, which reduces only ⌬ without changing ⌬pH, deceased glutamate uptake by VGLUT2 by 50%. Nigericin, an electroneutral ionophore that exchanges K ϩ for H ϩ , eliminates ⌬pH whereas allowing ⌬ to increase in the presence of 4 mM KCl and subsaturating concentrations of glutamate, inhibited glutamate uptake by VGLUT2 by 40%. This result is consistent with previous findings that both the ⌬ and ⌬pH components are driving forces for the transport of glutamate into synaptic vesicles (13). However, this functional characteristic is distinct from VGLUT1, because ⌬ was suggested as the primary driving force for VGLUT1 (18,19). Thus, one possible explanation of the previous controversial results observed in synaptic vesicles could be the differing functions of two closely related isoforms. Further elucidation of molecular differences between these two isoforms may answer this question.
Another interesting finding is that the two vesicular glutamate transporters have different expression patterns in brain. VGLUT1 is only expressed in a subset of glutamatergic neurons in the cerebral cortex, hippocampus, and cerebellum (20), whereas VGLUT2 is expressed in cerebral cortex, hippocampus, and the thalamus (Fig. 5). Thalamic nuclei use glutamate as their neurotransmitter but they lack VGLUT1 mRNA and protein (17,20). Thus, VGLUT2 appears to be an important glutamate transporter in thalamic neuronal vesicles. In addition, DNPI, the human homologue of VGLUT2, was highly expressed in the fetal brain when VGLUT1 was not expressed (28), suggesting that VGLUT2 rather than VGLUT1 might be important in the initial steps of neuronal differentiation.
In summary, we have cloned and functionally characterized a novel vesicular glutamate transporter expressed in the mouse brain. The VGLUT2 cDNA encodes a protein of 582 amino acids with 75% identity with human VGLUT1, the first identified vesicular glutamate transporter. All of the major functional characteristics of VGLUT2, such as ATP dependence, chloride stimulation, substrate specificity, substrate affinity, and mode of energization, are consistent with that of a glutamate transporter previously characterized from synaptic, vesicular membranes. Thus, VGLUT2 functions as a vesicular glutamate transporter. Identification of two isoforms with different functional characteristics provides molecular information for structure-function studies of vesicular glutamate transporters. In addition, the presence of VGLUT2 in brain regions lacking VGLUT1 suggests that the two isoforms together might account for glutamate transport into synaptic vesicles in glutamatergic neurons.