N-Methyl-d-aspartate Receptor-dependent Regulation of the Glutamate Transporter Excitatory Amino Acid Carrier 1*

The neuronal transporter excitatory amino acid carrier 1 (EAAC1) is enriched in perisynaptic regions, where it may regulate synaptic spillover of glutamate. In this study we examined potential interactions between EAAC1 and ionotropic glutamate receptors. N-Methyl-d-aspartate (NMDA) receptor subunits NR1, NR2A, and NR2B, but not the α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor subunit GluR2, were co-immunoprecipitated with EAAC1 from neuron-enriched hippocampal cultures. A similar interaction was observed in C6 glioma and human embryonic kidney cells after co-transfection with Myc epitope-tagged EAAC1 and NMDA receptor subunits. Co-transfection of C6 glioma with the combination of NR1 and NR2 subunits dramatically increased (∼3-fold) the amount of Myc-EAAC1 that can be labeled with a membrane-impermeable biotinylating reagent. In hippocampal cultures, brief (5 min), robust (100 μm NMDA, 10 μm glycine) activation of the NMDA receptor decreased biotinylated EAAC1 to ∼50% of control levels. This effect was inhibited by an NMDA receptor antagonist, intracellular or extracellular calcium chelators, or hypertonic sucrose. Glutamate, α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid with cyclothiazide, and thapsigargin mimicked the effects of NMDA. These studies suggest that NMDA receptors interact with EAAC1, facilitate cell surface expression of EAAC1 under basal conditions, and control internalization of EAAC1 upon activation. This NMDA receptor-dependent regulation of EAAC1 provides a novel mechanism that may shape excitatory signaling during synaptic plasticity and/or excitotoxicity.

Glutamate is an excitatory amino acid that elicits physiological and excitotoxic responses in the nervous system. Extracellular glutamate is normally maintained at low levels, allowing bursts of glutamate to generate postsynaptic responses during synaptic transmission. Tight regulation of extracellular glutamate is required to allow essential activity and prevent excitotoxicity (for reviews, see Refs. [1][2][3]. Metabotropic (mGluR) 3 and ionotropic (AMPA, kainate, and NMDA) receptors mediate the effects of glutamate. mGluRs are G-protein-coupled receptors that activate a variety of second-messenger systems (for review, see Ref. 4), whereas AMPA, kainate, and NMDA receptors are ligand-gated ion channels that also activate intracellular signaling pathways. NMDA receptors have a high affinity for glutamate, are highly permeable to calcium, and do not readily desensitize, all of which contribute to the central role of NMDA receptors in excitotoxicity (5,6).
Removal of extracellular glutamate in the forebrain is controlled by three major excitatory amino acid transporters (EAATs): EAAT1 (GLAST), EAAT2 (GLT-1), and EAAT3 (EAAC1) (for reviews, see Refs. 1 and 7). Glutamate transporters rapidly clear extracellular glutamate, thereby protecting neurons from excitotoxicity (for reviews, see Refs. [1][2][3]. Although GLAST and GLT-1 are mainly glial, EAAC1 is mostly neuronal (8). In area CA1 of the hippocampus, a neuronal transporter limits synaptic spillover of glutamate and activation of NMDA receptors (9). Because EAAC1 is the only neuronal transporter found in this location, this observation suggests that EAAC1 may control activation of some subtypes of NMDA receptors. A similar functional interaction is also observed in Xenopus oocytes, where co-expression of EAAC1 attenuates glutamate-evoked NMDA receptor currents (10). EAAC1 has also been linked to pathology in the nervous system. For example, antisense-mediated down-regulation of EAAC1 causes glutamate-induced cell death in the hippocampus (11,12). However, some studies suggest that glutamate transporters contribute to excitotoxicity through transporter reversal during insults such as ischemia (13,14). Based on these studies, EAAC1 may play important roles in NMDA receptor-mediated excitotoxicity.
Whereas NMDA receptors are localized to presynaptic, synaptic, and extrasynaptic areas (15), EAAC1 is enriched perisynaptically (16,17), placing EAAC1 in close proximity to subpopulations of NMDA receptors. NMDA receptor activation can mediate both long-term potentiation (LTP) and long-term depression (LTD) (18,19). Recent preliminary data presented in an abstract suggest that knockdown of EAAC1 may impair NMDA receptor-induced LTP (20). EAAC1 activity may, therefore, influence NMDA receptor function. Conversely, LTP increases glutamate uptake activity, which has pharmacological properties consistent with EAAC1 (21). This effect is associated with an increase in EAAC1 immunoreactivity in a subcellular fraction enriched in plasma membranes and is blocked by NMDA receptor antagonists (21). Therefore, EAAC1 and ionotropic glutamate receptors may be regulated by similar mechanisms that are normally controlled by NMDA receptor stimulation.
In this study we sought to identify associations between ionotropic glutamate receptors and EAAC1. We found evidence for physical (co-immunoprecipitable) interactions between EAAC1 and the NMDA receptor in a variety of systems. In addition, co-expression of NMDA receptors with an epitopetagged variant of EAAC1 increased the amount of biotinylated (cell surface) EAAC1 immunoreactivity. Finally, robust activation of the NMDA receptor rapidly (within 5 min) decreased the amount of biotinylated EAAC1 in hippocampal neuronenriched cultures. This effect was blocked by chelation of intracellular or extracellular calcium or by hypertonic sucrose. These studies identify novel physical and functional interactions between NMDA receptors and the neuronal transporter EAAC1.

EXPERIMENTAL PROCEDURES
Hippocampal Primary Culture-Neuronally enriched cultures were prepared from the hippocampus of 18 -19-day Sprague-Dawley rat embryos (E18-E19) as described (22). Briefly, the hippocampi from rat embryos were trypsinized (0.027%) for 20 min at 37°C in 7% CO 2 and washed with Hanks' buffered saline solution. The tissue was triturated in warm media (10% defined, heat-inactivated fetal bovine serum, 10% Ham-s F-12 medium, and 80% Dulbecco's modified Eagle's medium with penicillin/streptomycin), and cells were plated onto 60-mm poly-D-lysine-coated plates at a density of 6 ϫ 10 5 viable cells per 35-mm culture dish. Neurons were maintained at 37°C, 5% CO 2 with neurobasal media containing B27 (2%). Cultures, which include a heterogeneous mixture of pyramidal and nonpyramidal hippocampal neurons and a small percentage of astrocytes (generally Ͻ10%), were maintained for at least 17 days in vitro (DIV17) before experimentation unless otherwise indicated. In some studies cultures were maintained for shorter periods of time (8 -9 days, DIV8 -9) to model a less "mature" phenotype of synaptic development. At this earlier time, NMDA receptors are primarily composed of the NR1 and NR2B subunits with minimal levels of NR2A and minimal levels of the scaffolding protein PSD-95; this profile is similar to that observed early in development in vivo (23)(24)(25). Neuronal cultures maintained for 17 days express combinations of NMDA receptor subunits (NR1, NR2A, and NR2B) and scaffolding proteins (such as PSD-95) similar to those observed in adult animals and, therefore, may model the more mature synaptic milieu.
Cell Culture-The rat C6 glioma cell line was grown and maintained as described (26). Under normal conditions this cell line only expresses the EAAC1 subtype of glutamate transporter (27)(28)(29). We and others have used this cell line to study EAAC1 trafficking (28 -31). This system has an advantage of being relatively easy to transfect. In addition, the effects of pro-tein kinase C activation, platelet-derived growth factor, and the constitutive trafficking of EAAC1 on and off the plasma membrane display similar characteristics to those observed in neuronal cultures (32)(33)(34). Therefore, this system is analogous to differentiated NIH3T3 cells routinely used to study regulated trafficking of the GLUT4 subtype of glucose transporter (35).
Incubation in NMDA Receptor Agonists and Selective Antagonists-Agonists and antagonists were acquired from Sigma-Aldrich unless otherwise indicated. To activate NMDA receptors, NMDA (10 or 100 M) and glycine (10 M) were diluted to a 5ϫ concentration in conditioned neurobasal media with or without applicable inhibitors and then added directly to culture media for 5 min. BAPTA-AM, W-7, FK506, KN62, 7-nitroindazole, MDL28170, MG132, SB203580, PD98059, and 2-aminoethoxydiphenyl borate were dissolved in Me 2 SO, and experiments involving these inhibitors were compared with a 0.1% Me 2 SO vehicle control. Although dantrolene, nifedipine, -agatoxin IVa, -conotoxin GVIA, LY294002, and wortmannin were also dissolved in Me 2 SO, the highest concentration of Me 2 SO for these experiments was 0.02%, which did not affect EAAC1 distribution. Cultures were maintained at 37°C and 5% CO 2 during preincubation and treatment.
Transient Transfection of C6 Glioma-C6 glioma were grown to 40 -60% confluency on 35-or 60-mm plates and transfected using GenePorter (Genlantis, San Diego, CA), a lipid-based transfection kit. 35-mm plates were transfected with 3-4 g of cDNA, and 60-mm plates were transfected with 5-7 g of cDNA with a 5:1 ratio of GenePorter reagent to cDNA, diluted in Dulbecco's modified Eagle's medium as per the manufacturer's recommendations. Cultures were maintained for 18 h after transfection. All transfections included cDNAs for NR2A, NR2B (⑀2), and Myc-EAAC1. This combination of cDNAs was co-transfected with either NR1 or pRC/CMV (control vector). During and after transfection, C6 glioma cells were incubated in Dulbecco's modified Eagle's medium with 500 M ketamine, a concentration that prevents tonic toxic activation of NMDA receptors and consequent excitotoxicity (36,37). Ketamine was included in all parallel comparison groups to exclude the possibility of nonspecific effects of ketamine with targets unrelated to NMDA receptors. Experiments were standardized for expression levels of Myc-EAAC1. In 2 of 5 experiments Myc-EAAC1 immunoreactivity in the lysate fraction was barely detectable; these Western blots were not quantitated.
Transfection of HEK293 Cells-HEK293 cells were grown to 50 -70% confluency and transfected using calcium phosphate precipitation (37). cDNA combinations including Myc-EAAC1, NR1, NR2A, NR2B (⑀2), NMDA receptor subunit deletions, and/or pRC/CMV were transfected with a total of 3 g of cDNA per 35-mm plate. Experiments contained equal total amounts of cDNA. During and after transfection, HEK293 cells were incubated in minimum essential medium with 500 M ketamine. MK-801 was not used to prevent toxicity because of the slow off-rate. D-APV was not used because in the presence of serum, high concentrations do not completely block toxicity, and the experiments can become quite expensive (for discussion, see Refs. [35][36][37]. Cultures were maintained for 18 h after transfection before experimentation. Co-immunoprecipitation-Cells were washed with ice-cold phosphate-buffered saline solution and incubated in radioimmune precipitation assay buffer (150 mM NaCl, 1 mM EDTA, 100 mM Tris-Cl, 1% Triton X-100, 0.5% sodium deoxycholate) with a mixture of protease inhibitors (Calbiochem). Lysates were centrifuged at 16,600 ϫ g to remove nucleic acids and cellular debris. Supernatants from 2 or more plates were pooled and precleared with protein G beads (Invitrogen) for 2 h at 4°C. Protein lysates were separated into aliquots (150 -225 g of protein) and incubated with 3 g of one of the following antibodies unless otherwise specified: EAAC1 (Alpha Diagnostics International, San Antonio, TX), EAAC1 (affinity purified, from Dr. J. D. Rothstein), Myc (BD Biosciences), NR1 (BD Pharmingen), NMDAR2C (1 g; Invitrogen), rabbit IgG (Invitrogen), or mouse IgG (Invitrogen). After an overnight incubation at 4°C, protein G beads (25 l) were added and incubated at 4°C for 2 h. Beads were isolated by centrifugation and subsequently washed four times with radioimmune precipitation assay buffer before the addition of 2ϫ Laemmli sample buffer.
Biotinylation Assays-Biotinylation of cell surface proteins was performed using hippocampal cultures or C6 glioma grown on 35-or 60-mm tissue culture dishes, as previously described (33). Briefly, after rinsing cells with ice-cold phosphate-buffered saline supplemented with 0.1 mM calcium chloride and 1 mM magnesium chloride (phosphate-buffered saline (Ca 2ϩ / Mg 2ϩ )), plates were incubated with 1 mg/ml N-hydroxysulfosuccinimidobiotin (NHS-biotin, Pierce) in phosphate-buffered saline solution (Ca 2ϩ /Mg 2ϩ ) for 30 min at 4°C with slight agitation. After incubation, biotin was quenched using phosphatebuffered saline solution (Ca 2ϩ /Mg 2ϩ ) containing 100 mM glycine for 30 min at 4°C. Cells were lysed using radioimmune precipitation assay buffer containing 0.1% SDS and a mixture of protease inhibitors. Lysates were sonicated and centrifuged at 16,600 ϫ g to remove nucleic acids and debris. Half of the lysate fraction was saved for analysis; the remaining lysate was agitated overnight at 4°C with an equal volume of UltraLink immobilized monomeric avidin beads to isolate biotinylated proteins. The lysate-bead mixture was centrifuged, and the supernatant was retrieved (nonbiotinylated fraction) and saved for Western blot analysis. Beads were subsequently washed four times, and protein was eluted into SDS buffer with 2-mercaptoethanol (biotinylated fraction). The lysate, nonbiotinylated, and biotinylated fractions were all diluted in sample buffer such that the sum of the immunoreactivity in the biotinylated and the nonbiotinylated fractions should equal that observed in the lysate fraction if the yield from extraction is 100%.
Western Blot Analyses-Western blot analyses were performed as previously described (38). Proteins were separated on an 8% SDS-PAGE gel. Antibodies directed toward these targets were utilized: EAAC1 (1:250; affinity-purified, from Dr. J. D. . Primary antibody incubation was followed by mouse monoclonal or rabbit polyclonal horseradish peroxidase secondary antibody (1:2500; Cell Signaling Technology, Danvers, MA). Immunoreactivity was visualized using enhanced chemiluminescence reagent (Pierce) and exposure on x-ray film. Bands were quantitated using densitometry normalized to actin (lysate) and then expressed as a percentage of control (no treatment) for each fraction. Glutamate transporters routinely form irreversible aggregates upon solubilization that are not dissociated with standard SDS-containing sample buffer (39). Similar aggregates were observed in the present study. The amounts of monomeric and multimeric EAAC1 were quantitated and analyzed separately. Both species responded similarly, and therefore, data presented represent changes in the sum of the immunoreactivity of EAAC1 observed. Actin immunoreactivity in the biotinylated fraction was used to test for potential cell lysis. In all studies only a small percentage of actin immunoreactivity was detected in the biotinylated fraction (between 0 and 15%). In some cases data were represented as a percentage of transporter or receptor on the cell surface, calculated as (biotinylated/(biotinylated ϩ nonbiotinylated)) ϫ 100. When detecting more than one protein in a single experiment, nitrocellulose membranes were stripped (0.1 M glycine, pH 2.3 for 1 h) and then reprobed with the next antibody. Secondary antibodies were completely stripped before new antibody incubation. In instances where proteins of similar molecular weight were visualized (such as NR2A and NR2B), either primary antibodies requiring two different secondary antibodies were used, more than one gel was run with identical samples, or primary antibody was stripped off (confirmed by reincubating in secondary antibody and redeveloping).
Statistical Analyses-Multiple comparisons were completed by analysis of variance followed by t test with Bonferroni corrections. Paired t test was used to directly compare values of two groups. When there were only two groups (vehicle and treatment) data were analyzed by one sample t test. For two-group comparisons with significantly different variances, data were analyzed using Mann-Whitney non-parametric tests. Statistical significance was set at a p Ͻ 0.05, computed using InStat (GraphPad Software, San Diego, CA).

Interaction of EAAC1 and NMDA Receptor Subunits-Be-
cause EAAC1 and ionotropic glutamate receptors may overlap in distribution (15,17) and be regulated through similar mechanisms (Ref. 21; for a recent discussion, see Ref. 33), we examined potential interactions between EAAC1 and ionotropic glutamate receptors in primary hippocampal cultures. Using hippocampal culture lysates, a commercially available antibody specific to EAAC1 immunoprecipitated NMDA receptor subunits NR1, NR2A, and NR2B but not the AMPA receptor subunit GluR2 (Fig. 1A). Although a small amount of immunoreactivity of each NMDA receptor subunit was non-specifically detected with an IgG control, anti-EAAC1 always immunoprecipitated greater quantities of NMDA receptor subunit immunoreactivity. To verify the interaction between EAAC1 and NMDA receptors, immunoprecipitation was also performed using a different affinity purified EAAC1 antibody (from Dr. J. D. Rothstein). This antibody also immunoprecipitated NR1 (Fig. 1B). Because anti-EAAC1 immunoprecipitated NR1, NR2A, and NR2B, we aimed to identify the subpopulation of NMDA receptors with which EAAC1 may interact by immunoblotting for post-synaptic density 95 (PSD-95), a scaffolding protein of NMDA receptors in the postsynaptic density (40). In these experiments one of the anti-EAAC1 antibodies (from Dr. J. D. Rothstein) immunoprecipitated PSD-95 (n ϭ 2), whereas the other anti-EAAC1 antibody did not (n ϭ 4). The former anti-EAAC1 antibody also immunoprecipitated more NR1 than the latter, although both antibodies immunoprecipitated comparable amounts of EAAC1 (n ϭ 2).
To determine whether this interaction can be observed in another system, we utilized a model in which NMDA receptor subunit composition could be manipulated. C6 glioma endogenously express EAAC1 (28) but not NR1 (Fig. 1C, first lane). Small amounts of endogenous NR2A and NR2B were observed in C6 glioma (data not shown), but NR1 is also required for NMDA receptor assembly and function (for a recent discussion, see Ref. 41). Although we routinely detected small amounts of cross-reacting immunoreactivity of the approximate molecular weight of NR1 with the rabbit polyclonal anti-NR1 antibody, this was not observed with a monoclonal anti-NR1 antibody (data not shown). Taking advantage of this expression profile, C6 glioma were transfected with cDNAs for Myc-EAAC1, NR2A, NR2B, and either NR1 or pRC/CMV (control vector). Anti-Myc immunoprecipitated NR2A and NR2B only when NR1 was co-expressed, indicating that NR1 expression or NMDA receptor assembly was required for the interaction of EAAC1 with the NMDA receptor (Fig. 1C). All experiments were performed in the presence of an NMDA receptor antagonist (ketamine) to block receptor activation, suggesting that NMDA receptor activation is not required for this interaction. Although NR2A and NR2B were observed in isolates of anti-NR1 immunoprecipitation, Myc immunoreactivity (Myc-EAAC1) was barely detectable. This likely reflects the inefficiency of the NR1 antibody for immunoprecipitation (anti-NR1 immunoprecipitation depleted NR1 supernatant immunoreactivity by only 20%; data not shown). Overall, these data suggest that whereas EAAC1-NMDA receptor interactions could be detected in C6 glioma (as in hippocampal cultures), only a minority of EAAC1 and NR1 molecules can be co-immunoprecipitated.
Because C6 glioma are a central nervous system-derived cell line that expresses low levels of NR2A and NR2B, we tested the NMDA receptor subunit requirements for these interactions in a system lacking endogenous NMDA receptors, HEK293 cells Anti-EAAC1 consistently immunoprecipitated more NR1, NR2A, and NR2B than IgG control. Anti-EAAC1 and IgG immunoprecipitations exhibited identical trace immunoreactivity for GluR2. Based on calculations of 100% efficiency of this immunoprecipitation and adjusting for amount of protein loaded, we found that 0.93% Ϯ 0.01% (n ϭ 4) of total NR1 co-immunoprecipitates with EAAC1. B, immunoprecipitation in hippocampal cultures was performed with two EAAC1 antibodies, EAAC11-A and another affinity-purified antibody (Dr. J. D. Rothstein; indicated as EAAC1-B). Both EAAC1 antibodies immunoprecipitated more NR1 than control IgG (n ϭ 2). EAAC1-B immunoprecipitated PSD-95 (n ϭ 2), but EAAC11-A did not (n ϭ 4). Both EAAC1 antibodies immunoprecipitated comparable amounts of EAAC1. The EAAC1 in the lysate could not be visualized in the representative Western blot, since overexposure of immunoprecipitated lanes was avoided. C, C6 glioma were co-transfected with Myc-EAAC1, NR2A, NR2B, and either NR1 or pRC/CMV (vector control). Cells were lysed 18 h after transfection, and co-immunoprecipitations were performed with anti-Myc, IgG (control), or anti-NR1 antibodies. Anti-Myc immunoprecipitated NR1, NR2A, and NR2B only when NR1 cDNA was included in the transfection mixture. Anti-NR1 immunoprecipitated NR2A and NR2B, but only trace amounts of Myc-EAAC1, similar to that of control IgG. Presented blots are from a single experiment and are representative of three independent experiments (n ϭ 3). (42). We transfected HEK293 cells with Myc-EAAC1, NR2A, and NR2B with either NR1 or pRC/CMV. In this system, anti-Myc antibodies were unable to immunoprecipitate significant levels of any NMDA receptor subunit nor were anti-NR1 antibodies able to immunoprecipitate EAAC1 (data not shown). However, an NR2 antibody (directed toward the conserved N terminus of NR2) immunoprecipitated Myc-EAAC1 in the presence but not absence of NR1 ( Fig. 2A).
To isolate the NMDA receptor subunit requirements, HEK293 cells were transfected with NR1 and Myc-EAAC1 with NR2A, NR2B, or pRC/CMV (excluding the other NR2 subunit or both NR2 subunits). The anti-NR2 antibody immunoprecipitated Myc-EAAC1 in the absence of either NR2A or NR2B but not when both NR2 subunits were omitted (Fig. 2B). These data suggest that this interaction was independent of NR2 subtype. To test the NR1 domain required for this interaction, we transfected HEK293 cells with biologically relevant alternatively spliced forms of NR1. The above described transfections included NR1a, the NR1 splice-variant containing the C1 and C2 cassettes of the C terminus. Cells were transfected with Myc-EAAC1, NR2A, NR2B, and NR1e, the shortest splice variant form of NR1. Anti-NR2 immunoprecipitated Myc-EAAC1 in transfections containing NR1e (Fig. 2C). Furthermore, transfection of NR1 truncated at amino acid 838, which deleted a major portion of the remainder of the C terminus (beyond that of NR1e), in combination with NR2A, NR2B, and Myc-EAAC1 did not alter this immunoprecipitation. These data suggest that the intracellular C termini of NR1 are not responsible for the interaction with EAAC1, but the interaction required both NR1 and NR2 subunits.
Based on the ability of one anti-EAAC1 antibody to coimmunoprecipitate PSD-95 from cultures, we assessed if cotransfection of PSD-95 would alter NMDA receptor-EAAC1 interactions in HEK293 cells. Although difficult to quantitate due to the variability of transfection efficiency, co-transfection of PSD-95 modestly increased the ability of anti-NR2 to immunoprecipitate EAAC1 (Fig. 2D). These data suggest that, whereas PSD-95 is not necessary for NMDA receptor-EAAC1 interactions, it may facilitate or stabilize this interaction.
EAAC1 cell surface expression in the presence and absence of NMDA receptor expression. C6 glioma have been used to model cell surface trafficking of EAAC1 as observed in neurons (see "Experimental Procedures"), whereas HEK293 cells do not exhibit similar trafficking mechanisms. 4 Therefore, we transfected C6 glioma cells with Myc-EAAC1, NR2A, and NR2B with NR1 or with pRC/CMV; cell surface proteins were batchextracted after biotinylation with a membrane impermeant biotinylating reagent. In these studies, NR1 (but essentially no actin) immunoreactivity was observed in the biotinylated fraction, showing that NMDA receptors are appropriately targeted to the plasma membrane (Fig. 3A).
In these same studies, co-expression of NR1 with Myc-EAAC1 increased biotinylated Myc-EAAC1 levels nearly 3-fold compared with pRC/CMV transfections (Fig. 3, A  and B; p ϭ 0.015 by the Mann Whitney test, n ϭ 3). The amount of Myc-EAAC1 at the cell surface was 9 Ϯ 3% in control transfections and 24 Ϯ 4% in NR1-containing transfections (Fig. 3A). This effect was not accompanied by a change in the total amount of Myc-EAAC1 in cell lysates nor by an increase in the amount of biotinylated actin. Although no significant change in nonbiotinylated Myc-EAAC1 immunoreactivity was observed under these conditions (Fig. 3B), this likely reflects the large amount of the transporter found in the nonbiotinylated fraction. In the absence of NMDA receptors (NR1), a 3-fold increase in biotinylated immunoreactivity would correspond to only a 16% loss in the nonbiotinylated transporter, which would be difficult to detect. Similar increases in biotinylated EAAC1 were observed with co-transfection of NR1e (Fig.  3A), suggesting that all biologically relevant NR1 splice variants can increase in EAAC1 cell surface expression. In these experiments cells were maintained in ketamine (500 M) to prevent NMDA receptordependent excitotoxicity (36,37) from the beginning of transfection until the cells were incubated with biotin. Under these conditions, we observed no evidence of cell death (floating cell debris), suggesting that this concentration of ketamine was sufficient to prevent excitotoxicity and that ketamine was not toxic under these conditions. Furthermore, because both control-transfected (pRC/CMV) and NR1-transfected cells were incubated with ketamine, these studies strongly suggest that the increase in biotinylated Myc-EAAC1 results from the expression of the NR1 subunit and not nonspecific effects of ketamine. Together these studies provide strong evidence that the expression of NR1 is sufficient to dramatically increase the fraction of EAAC1 found at the cell surface in a manner that is independent of NMDA receptor activation.
if such changes in protein expression in cultured hippocampal neurons were associated with changes in EAAC1 cell surface expression. Total EAAC1 immunoreactivity (lysate) did not change between DIV8 -9 and DIV17 (Fig. 3, C and D), consistent with previous reports of EAAC1 expression levels over the course of development in vitro (44). However, the amount of biotinylated EAAC1 dramatically increased with ϳ25% of total EAAC1 immunoreactivity localized to the cell surface at DIV8 -9 and ϳ80% on the cell surface at DIV17 (Fig. 3, C and  E). Similarly, NR1 immunoreactivity in the biotinylated fraction increased from ϳ55% (at DIV8 -9) to ϳ90% (at DIV17) of the total NR1 immunoreactivity. This was accompanied by a modest increase in total NR1 expression (Fig. 3, C and D). There were no changes in total actin levels or the amounts of biotin-ylated actin, indicating that these changes in biotinylated transporter/ receptor are not related to increased cell lysis. These data suggest that the increased cell surface expression of EAAC1 at DIV17 is associated with an increase of NR1 protein and cell surface expression, an effect that correlates with synaptic development.
NMDA Receptor Activation Decreases the Amount of Biotinylated EAAC1-One NMDA receptor-dependent model of LTP induction increases plasma membrane levels of EAAC1 (21), suggesting that EAAC1 levels are subject to NMDA receptor activation-dependent alterations in synaptic strength. To test this possibility, we examined EAAC1 cell surface expression in a complementary paradigm of NMDA receptor activation-dependent synaptic restructuring in DIV17 hippocampal cultures. A brief (5 min) bath application of NMDA (100 M) and glycine (10 M) has been used to "chemically" induce LTD (45). This treatment in hippocampal cultures significantly decreased (ϳ25%) the amount of biotinylated GluR2 (as observed previously) (45) and the amount of biotinylated NR1 (Fig. 4,  A and B). In these same experiments, this brief treatment with NMDA and glycine dramatically decreased (ϳ50%) the amount of biotinylated EAAC1 (Fig. 4, A and  B). To confirm the linearity of the signal observed in the biotinylated fraction as measured by film exposure to chemiluminescence, the amount of EAAC1 immunoreactivity after treatment with NMDA was compared with that observed when half of control biotinylated material was loaded on the gel. In these analyses the amount of EAAC1 immunoreactivity was comparable in these two samples ( Fig. 4A for an example), showing that NMDA treatment decreased biotinylated EAAC1 by ϳ50% of control within 5 min. NMDA did not alter total (lysate) receptor or transporter immunoreactivity. Furthermore, the effects of NMDA on GluR2, NR1, and EAAC1 were associated with consistent increases in the amounts of nonbiotinylated protein (Fig. 4A). However, under basal conditions only a small amount of nonbiotinylated protein was noted, making it difficult to represent increases in nonbiotinylated protein as a percentage (e.g. see NR1 immunoreactivity under control conditions in Fig. 4A). Still, these increases in nonbiotinylated EAAC1, GluR2, and  4). B, quantitation of changes in GluR2, NR1, and EAAC1 immunoreactivity in lysate, biotinylated, and nonbiotinylated fractions with NMDA treatment. All conditions were standardized to lysate actin and expressed as a percent of vehicle control (ϮS.E.). No changes were noted in lysate fractions. Differences between biotinylated fractions for all comparisons were statistically significant (p ϭ 0.0002 by analysis of variance; n ϭ 4). The decreases in GluR2 and NR1 immunoreactivity in the biotinylated fraction were similar (*, p Ͻ 0.05, by t test with Bonferroni correction). EAAC1 immunoreactivity was decreased in the biotinylated fraction (**, p ϭ 0.0009 by t test with Bonferroni correction), significantly more than GluR2 (p ϭ 0.032, by uncorrected t test) (n ϭ 4). C, representative immunoblots of GLAST and GLT-1 immunoreactivity in response to NMDA treatment, as compared with EAAC1. Neither GLAST nor GLT-1 immunoreactivity was altered by NMDA treatment in any fraction (n ϭ 4). D, top, representative immunoblot of EAAC1 immunoreactivity in DIV8 -9 cultures with NMDA treatment (100 M, 5 min). EAAC1 immunoreactivity in the biotinylated fraction did not change with NMDA treatment (n ϭ 4). Bottom, representative immunoblot showing NR1 immunoreactivity with anti-EAAC1 (EAAC11-A) co-immunoprecipitation (IP). The amount of NR1 immunoreactivity in the immunoprecipitates was not altered by 5 min 100 M NMDA (n ϭ 3).
NR1 were observed in every experiment, supporting internalization of the transporter and receptors rather than just a loss of biotinylated protein (due to protease activity or experimental error). In these same studies, there was no change in the amount of biotinylated actin. Together these studies provide strong evidence that NMDA and glycine cause a rapid and robust redistribution of EAAC1 from the plasma membrane to an intracellular compartment. This effect was selective for the EAAC1 subtype of transporter, as NMDA treatment did not alter levels of immunoreactivity in the biotinylated fractions of other glutamate transporters expressed in hippocampal cultures (GLT-1 and GLAST) (Fig. 4C) (1, 7).  ; n ϭ 4). B, representative immunoblot and summary of EAAC1 immunoreactivity after NMDA treatment with and without 1 mM D-APV. D-APV prevented NMDA-induced EAAC1 redistribution (*, p ϭ 0.0007 by paired t test; n ϭ 5). C, quantitation of EAAC1 immunoreactivity after treatment with 100 M glutamate in the absence or presence of D-APV. Glutamate significantly decreased biotinylated EAAC1 (*, p ϭ 0.014 by one sample t test), and the effect of glutamate was blocked by D-APV (p ϭ 0.0151 by paired t test; n ϭ 4). D, quantitation of EAAC1 immunoreactivity in the presence of other glutamate receptor agonists. 100 M AMPA did not significantly alter EAAC1 immunoreactivity in the biotinylated fraction (p ϭ 0.47 by one sample t test; n ϭ 4); however, concomitant treatment of AMPA and cyclothiazide (CTZ; 50 M) significantly reduced EAAC1 in the biotinylated fraction (*, p ϭ 0.0009 by one sample t test; n ϭ 5). Dihydroxyphenylglycine (DHPG) (250 M) had no effect on EAAC1 immunoreactivity in the biotinylated fraction (p ϭ 0.64 by one sample t test; n ϭ 4).
We then examined whether these effects were observed in other models beyond hippocampal cultures at DIV17. NMDA and glycine treatment did not induce changes in EAAC1 cell surface expression in hippocampal cultures at DIV8 -9 (Fig. 4D,  upper panel) or in transfected C6 glioma (data not shown, n ϭ 3), suggesting that the potentially more mature synaptic milieu observed in hippocampal cultures at DIV17 is required for the NMDA-induced effects.
Because the decrease in biotinylated EAAC1 was greater than that observed for NR1, this treatment might alter the amount of NR1 that can be co-immunoprecipitated with EAAC1. Hippocampal neurons at DIV17 were treated with NMDA as described above, and the amount of NR1 co-immunoprecipitated with EAAC1 was compared with basal co-immunoprecipitation. This treatment had no effect on the amount of NR1 immunoprecipitated (Fig. 4D, lower panel).
We then defined the pharmacological requirements for redistribution of EAAC1 by NMDA receptor stimulation. Lower concentrations of NMDA (10 M) did not reduce EAAC1 immunoreactivity in the biotinylated fraction (Fig. 5A). The competitive NMDA receptor antagonist D-APV prevented the NMDA-induced EAAC1 redistribution, indicating that the effects of NMDA depended upon NMDA receptor activation (Fig. 5B). To determine whether glutamate had a similar effect on EAAC1, hippocampal cultures were treated with glutamate (100 M, 5 min) and glycine (10 M) in the absence and presence of D-APV (Fig. 5C). Glutamate also caused a decrease in biotinylated EAAC1 similar to that noted with NMDA. This effect was significantly attenuated by D-APV, and there was no significant difference in the amount of biotinylated EAAC1 between glutamate/glycine/D-APV and control (p ϭ 0.44 by one sample t test). Together, these studies strongly suggest that NMDA receptor activation resulted in a rapid redistribution of EAAC1. To determine whether activation of other glutamate receptors can cause a redistribution of EAAC1, hippocampal cultures were treated with AMPA to specifically activate AMPA receptors or dihydroxyphenylglycine, an agonist of type I mGluRs. Neither AMPA nor dihydroxyphenylglycine significantly altered cell surface immunoreactivity of EAAC1, but the combination of AMPA and cyclothiazide, which prevents AMPA receptor desensitization and potentiates AMPA-mediated intracellular calcium rises (46 -48), induced a rapid redistribution of EAAC1 (Fig. 5D).
Because NMDA or AMPA receptor activation (in the presence of cyclothiazide) increases intracellular calcium levels, the effects of calcium chelators BAPTA-AM or EGTA on NMDA/ glycine-induced redistribution of EAAC1 were examined to determine whether intracellular or extracellular calcium, respectively, was required for this effect. BAPTA-AM and EGTA separately or together prevented the NMDA-induced redistribution of EAAC1 (Fig. 6, A, B, C), indicating that increases in both intracellular and extracellular calcium are required for the effects of NMDA. In cortical cultures, depletion of intracellular calcium stores results in a loss of EAAC1 from the cell surface (49). Thapsigargin, which depletes intracellular calcium stores but also causes an immediate rise in cytosolic calcium through inhibition of the endoplasmic reticulum Ca 2ϩ -ATPase, resulted in EAAC1 redistribution (Fig.   6D), similar to the effect noted by others in cortical cultures (49). This effect was completely blocked by BAPTA-AM, indicating that in hippocampal cultures thapsigargin-induced EAAC1 redistribution required an increase in intracellular calcium.
Because calcium is required for NMDA-induced redistribution of EAAC1, the effects of other inhibitors of specific calcium sources (i.e. voltage-dependent calcium channels, canonical transient receptor potential channels, inositol trisphosphate receptors, etc.) were assessed. None of these inhibitors altered the NMDA-induced redistribution of EAAC1 (Table 1). Furthermore, a mixture of all of these inhibitors had no effect on the NMDA-induced changes in biotinylated EAAC1 ( Table 1), suggesting that the calcium increases directly mediated by the NMDA receptor are sufficient to induce EAAC1 redistribution.
Because internalization of EAAC1, GluR2, and NMDA receptors occurred concurrently, we hypothesized that the signaling pathways previously implicated in GluR2 internalization, NMDA receptor internalization, or LTD may be involved in NMDA receptor-dependent redistribution of EAAC1. Although NR1/NR2B NMDA receptors may selectively induce LTD (19, 50 -52), the NR1/NR2B receptor subtype selective antagonist Ro25-6981 did not inhibit the NMDA-dependent redistribution of EAAC1 ( Table 2), suggesting that activation of NR2A-containing receptors is sufficient to induce redistribution of EAAC1. Hyperosmotic sucrose (0.45 M) blocks redistribution of AMPA and NMDA receptors and has been used to implicate clathrin-mediated endocytosis in redistribution of these receptors (53)(54)(55). Hyperosmotic sucrose completely blocked NMDA-induced redistribution of EAAC1 (p ϭ 0.0003 by paired t test, n ϭ 5), suggesting that this effect is clathrin-dependent. Inhibitors of calcineurin and p38 MAPK, which inhibit AMPA receptor endocytosis and LTD (56,57), did not affect the NMDA/glycine-induced redistribution of EAAC1. Although the proteasome inhibitor MG132 significantly attenuated the NMDA-induced redistribution of EAAC1 (p ϭ 0.0125, by paired t test, n ϭ 3), this effect was small, and NMDA in the presence of this inhibitor still had a significant effect on biotinylated EAAC1 levels (p ϭ 0.049, by one sample t test, n ϭ 3). This suggests that proteasomal mechanisms may play a modest or secondary role in EAAC1 redistribution.
Activation of protein kinase C with phorbol esters increases cell surface expression of EAAC1 in C6 glioma and in primary cortical cultures, an effect blocked by protein kinase C antagonists (32,34). In these studies inhibition of protein kinase C on its own has no effect on biotinylated EAAC1. Because NMDA receptor activation can increase protein kinase C activity (58), primary hippocampal cultures were treated with the protein kinase C inhibitor BisII. In the present study, treatment of hippocampal neurons with BisII reduced biotinylated EAAC1 in both the presence and absence of NMDA (Table 2) and had no effect on total EAAC1 expression (data not shown). The effects of NMDA combined with BisII were not additive (Table 2), and a 5-min treatment with BisII mimicked the effects of NMDA (p ϭ 0.013 by one sample t test; % of control biotin immunoreactivity ϭ 49 Ϯ 10; n ϭ 4). These data suggest that protein kinase C may be constitutively active in supporting cell surface expression of EAAC1 in hippocampal cultures. Surprisingly, BisII-induced EAAC1 redistribution was blocked by BAPTA-AM pretreatment (p ϭ 0.022, by paired t test; % of control biotin immunoreactivity ϭ 99 Ϯ 18; n ϭ 4), suggesting that the effects of BisII also involve intracellular calcium. No other inhibitor of signaling molecules tested had effects on biotinylated EAAC1 or on the NMDA-induced changes in biotinylated EAAC1 (Table 2).

DISCUSSION
In the present study we identified novel physical and functional interactions between EAAC1 and NMDA receptor subunits NR1, NR2A, and NR2B in primary hippocampal cultures and C6 glioma. The biochemical association required expression of both NR1 and NR2 subunits and coincided with increased EAAC1 cell surface expression. Because interactions between EAAC1 and GluR2 were not detected, the association appears selective for NMDA receptor subunits. Independently, NMDA receptor activation induced a dramatic redistribution of EAAC1 using a paradigm identical to that previously used to chemically induce LTD (45). NMDA treatment also caused internalization of GluR2 and NR1 to a lesser extent than EAAC1, an effect mimicked by glutamate, AMPA with cyclothiazide, thapsigargin, and protein kinase C inhibition. Pharmacological manipulation with a variety of agonists and antagonists identified a requirement for increased cytosolic calcium and implicated clathrin-mediated endocytosis in this NMDA receptor-mediated redistribution of EAAC1.
Our data provide evidence that EAAC1 interacts with NMDA receptors containing NR1, NR2A, and/or NR2B in primary hippocampal cultures and two model cell systems. A proteomic anal-ysis of NMDA receptor binding partners did not previously identify EAAC1 (59), perhaps reflecting difficulties in co-immunoprecipitating EAAC1 with NMDA receptor antibodies. EAAC1 may also be found in a complex with PSD-95, since one of the anti-EAAC1 antibodies immunoprecipitated PSD-95. However, our data in HEK293 cells and in C6 glioma indicate that interaction of EAAC1 and NMDA receptors does not require PSD-95, and thus, the presence of EAAC1 in a PSD-95-containing complex may be coincidental to associations of EAAC1 and NMDA receptors. Therefore, EAAC1-PSD-95 interactions do not necessarily imply EAAC1 association with a particular subpopulation of NMDA receptor complexes. Furthermore, the interaction of EAAC1 and NMDA receptors could not be narrowed to the intracellular C terminus of NR1 or a particular NR2 subunit even though NR2 subunits have relatively low levels of homology in the C terminus (41). Although we cannot rule out more than one domain of interaction, one explanation for these findings could be that interaction between EAAC1 and NMDA receptors occur with N-terminal or transmembrane domains. This interaction also required the full complement of subunits that compose a functional NMDA receptor (both NR1 and NR2 subunits), which may reflect a stoichiometric requirement for EAAC1-NMDA receptor complexes. Although evidence for this interaction was obtained in three different types of cells (C6 glioma, HEK293 cells, and neurons), we cannot rule out the possibility that the EAAC1-NMDA receptor interaction is indirect. Future studies will be required to map the domains mediating the interaction and to determine whether the interaction is direct using purified proteins.

TABLE 1 Effects of inhibitors of specific calcium sources on NMDA receptor-mediated changes in EAAC1 cell surface expression
The percent of control in the biotinylated fraction is indicated. VDCC, voltage-dependent calcium channels; TRPC, canonical transient receptor potential. n represents the number of independent analysis.  The association of EAAC1 and NMDA receptors, although representing a small amount of the total transporter and receptor populations, correlated with a large change in EAAC1 distribution. Because NR1 expression increased the levels of biotinylated EAAC1, NR1 may inhibit the internalization or facilitate the trafficking of EAAC1 to the cell surface. Pilot experiments using reversible biotinylation assays in C6 glioma found no change in the internalization rates of Myc-EAAC1 with or without NR1 expression (data not shown, n ϭ 2). However, small differences in internalization rates over short duration experiments would be difficult to detect but could still explain large differences in surface expression over the transfection time period (18 h). At this time we cannot precisely determine the mechanism by which NR1 expression increases EAAC1 cell surface expression.
These data provide a mechanism by which NMDA receptors could modulate their own activation through EAAC1-mediated glutamate clearance (10). NMDA receptor expression could promote surface expression of EAAC1, thereby attenuating NMDA receptor currents. Although other factors such as growth factor production and synaptic scaffolding could alter EAAC1 cell surface expression during neuronal development, our data in C6 glioma support NR1 expression and/or NMDA receptor assembly as a major regulator. In other studies NMDA receptors co-localize with the glycine transporter GLYT1 (60), which also may regulate NMDA receptor activity (61,62). Although EAAC1 is localized perisynaptically, GLYT1 localizes to glutamatergic synapses and is regulated by PSD-95 (63,64). A splice variant of GLYT1 also directly interacts with ␥-aminobutyric acid (GABA) C receptors, an interaction that may alter GABA C receptor responses to glycine (65). Taken together, these data suggest that neurotransmitter transporters may associate with neurotransmitter receptors and, thus, modulate receptor activation.
Brief NMDA receptor stimulation resulted in redistribution of EAAC1, suggesting that high concentrations of endogenously released glutamate (such as during excitotoxicity) may change the localization of EAAC1. Under these conditions, NMDA receptor activation would be largely responsible for EAAC1 redistribution. However, AMPA receptor activation, which our data suggest can internalize EAAC1 in the absence of desensitization, may also provide small contributions.
NMDA receptor activation resulted in a 50% loss of cell surface EAAC1 in 5 min, a time matching the half-life of EAAC1 at the plasma membrane (33). Using reversible biotinylation (33), we tried to determine whether NMDA receptor activation accelerates internalization of EAAC1. After biotinylation, NMDA did not cause a redistribution of the transporter (data not shown, n ϭ 3). Although not tested, it is possible that biotinylation prevents receptor activation. Therefore, these experiments were not informative. However, as with AMPA and NMDA receptor endocytosis, EAAC1 internalization is blocked by hypertonic sucrose, an inhibitor of clathrin-mediated endocytosis (53)(54)(55). These data suggest that endocytosis is required for NMDA receptor-mediated EAAC1 redistribution.
Several pharmacological inhibitors were utilized to assess the means by which NMDA induced EAAC1 endocytosis. NR1/ NR2B subtype receptor antagonists can prevent NMDA receptor-mediated LTD of AMPA and NMDA receptors (19, 50 -52) but did not alter NMDA receptor-mediated EAAC1 redistribution. Different forms of LTD are mediated by calcineurin (56,57), p38 MAPK (66,67), and group I mGluRs (mGluR1 and mGluR5) (66, 68 -72). Pharmacological manipulation of these pathways did not alter the distribution of EAAC1. Proteasomemediated pathways in conjunction with PSD-95 may be involved in AMPA receptor endocytosis (73)(74)(75)(76). Proteasome inhibition, however, only slightly attenuated EAAC1 redistribution. Nevertheless, the potential association of EAAC1 with PSD-95 may facilitate EAAC1 endocytosis in response to NMDA receptor activation (73,77,78). Alternatively, blocking proteasomal degradation may impair upstream mechanisms, thereby maintaining EAAC1 on the cell surface. Therefore, although there are some similarities between the internalization of EAAC1 and that of AMPA and NMDA receptors, they arise from independent mechanisms.
Neither the interaction of EAAC1 with NMDA receptors nor the activity-dependent redistribution clearly required a subpopulation of NMDA receptors. Although the interaction of EAAC1 with the NMDA receptor involved NR1, NR2A, and NR2B subunits, NR1/NR2B receptor inhibition did not prevent the internalization of EAAC1. Furthermore in DIV8 -9 hippocampal cultures, which express few NR2A subunits (24,25), NMDA did not induce internalization of EAAC1. However, the internalization of EAAC1 also occurred after large rises in cytosolic calcium caused by other sources, diminishing the likelihood for a specific requirement of NMDA receptor subtypes. Thus, at present there appears to be no NMDA receptor subtype selectivity in the processes defined here.
Recent studies demonstrate a redistribution of EAAC1 and an increase in EAAC1 activity after LTP induction (21), and preliminary data from a recent abstract suggest that knockdown of EAAC1 impairs synaptic plasticity (20). These studies suggest that EAAC1 may participate in synaptic strengthening. Our findings suggest that during synaptic development, increased NR1 expression coinciding with increased EAAC1 cell surface expression may be required for facilitation of LTP. Furthermore, NMDA receptor activity-dependent EAAC1 redistribution may decrease EAAC1 activity, thereby increasing the likelihood of synaptic depression. EAAC1 and NMDA receptors may work in concert to modulate synaptic function.
Studies also support a role for EAAC1 in excitotoxicity. EAAC1 overexpression decreases NMDA receptor currents (10), and a loss of EAAC1 results in glutamate-induced toxicity in the hippocampus (11). Mice genetically deleted of EAAC1 exhibit increased oxidative stress and age-dependent neurodegeneration, consistent with increased sensitivity to excitotoxic insults over the course of synaptic maturity (79). This protective effect of EAAC1 may involve neuronal glutathione synthesis through cysteine uptake. Therefore, NMDA receptordependent effects on EAAC1 may be essential for limiting toxic cellular mechanisms.
Alternatively, NMDA receptor-mediated EAAC1 internalization may regulate increases in extracellular glutamate that occur during acute insults to the nervous system. This increase in extracellular glutamate enhances NMDA and AMPA recep-tor responses (80) and results in excitotoxicity (81). Several studies suggest that reversed operation of glutamate transporters that can occur as a result of a collapse in the electrochemical gradients contributes to this increase in extracellular glutamate (for examples, see Refs. 13 and 14). The source of extracellular glutamate is likely neuronal since astrocytes contain glutamine synthetase, which converts glutamate to glutamine. Mice deleted of EAAC1 display a slower rise in extracellular glutamate after an ischemic insult, supporting a role for EAAC1 in the rise of extracellular glutamate under these conditions (82). Therefore, NMDA-induced redistribution of EAAC1 might serve to limit the ability of these transporters to release glutamate under conditions of robust NMDA receptor activation. Further studies are required to explore these possibilities.
NMDA receptors and EAAC1 are important in both maintaining synaptic strength, and their aberrant regulation results in excitotoxicity. NMDA receptor activation can increase synaptic strength, expand dendritic spines, and increase gene transcription. However, NMDA receptor over-activation is responsible for decreased synaptic strength, cytoskeletal reorganization, and inhibition of regulatory cellular processes (5,6). The novel physical and functional interaction between NMDA receptors and EAAC1 reported in the present study may provide a mechanism to limit NMDA receptor activation, may contribute to synaptic plasticity, or may serve to limit the ability transporters to operate in the reverse direction under conditions of robust receptor activation.