Heteromer formation of δ2 glutamate receptors with AMPA or kainate receptors

Molecular Brain Research 110 (2003) 27–37 www.elsevier.com / locate / molbrainres

Research report

Heteromer formation of d2 glutamate receptors with AMPA or kainate receptors Kazuhisa Kohda a,b , Yoshinori Kamiya a , Shinji Matsuda a , Kunio Kato b , Hisashi Umemori c , Michisuke Yuzaki a , * a

Department of Developmental Neurobiology, St. Jude Children’ s Research Hospital, 332 North Lauderdale, Memphis, TN 38105, USA b Department of Neuropsychiatry, Faculty of Medicine, University of Tokyo, Tokyo 174 -0064, Japan c Department of Neurobiology, Washington University, St Louis, MO 63130, USA Accepted 11 October 2002

Abstract The d2 glutamate receptor (GluRd2) is predominantly expressed in the postsynaptic densities of parallel fiber-Purkinje cell synapses and plays a crucial role in cerebellar function. However, the mechanisms by which GluRd2 participates in cerebellar functions are largely unknown because GluRd2 does not bind glutamate analogs. We investigated the possibility that GluRd2 may be involved in channel formation together with other glutamate receptor families. We transiently expressed lurcher mutant AMPA receptor GluR1 Lc and kainate receptor GluR6 Lc in HEK293 cells. Cells expressing these constitutively active channels displayed a rectifying current–voltage (I–V ) relationship. However, when cells were co-transfected with GluRd2 Lc , which had the arginine residue in the channel pore region, cells displayed a linear I–V relationship, a result that indicates GluRd2 Lc formed functional heteromeric channels with GluR1 Lc or GluR6 Lc . Assembly of GluRd2 with GluR1 or GluR6 was further confirmed by co-immunoprecipitation assays in HEK293 cells. In addition, GluRd2 receptors were partially co-immunoprecipitated from cerebellar synaptosomal fractions by antibodies against GluR2 or KA2. In contrast to lurcher channels, expression of wild-type GluRd2 significantly reduced the glutamate-induced current of the wild-type GluR1 receptors without affecting channel properties, such as current kinetics, dose–response relationship, and single-channel conductance. Thus, the heteromeric channel created by the association of wild-type GluR1 and GluRd2 may not be gated by glutamate and does not participate in glutamate-induced currents. These results suggest that GluRd2 and AMPA or kainate receptors can assemble to form heteromeric receptors in vitro and could modify glutamate signaling in vivo. These findings may help explain the role of GluRd2.  2002 Elsevier Science B.V. All rights reserved. Theme: Neurotransmitters, modulators, transporters, and receptors Topic: Excitatory amino acid receptors: physiology, pharmacology and modulation Keywords: Assembly; Orphan receptor; Immunoprecipitation; Purkinje cell; GluRd2

1. Introduction In a phylogenetic comparison, the d2 glutamate receptor (GluRd2) family is positioned equidistant from three other ionotropic glutamate receptor families: the a-amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) family, the kainate family, and the N-methyl-D-aspartate (NMDA) family [2,18]. GluRd2 is predominantly expressed in the postsynaptic densities of parallel fiber-Purkinje cell *Corresponding author. Tel.: 11-901-495-2144; fax: 11-901-4952270. E-mail address: [email protected] (M. Yuzaki).

synapses [14] and plays a crucial role in cerebellar function. For example, antisense oligonucleotides specific for genes that encode GluRd2 impair long-term depression (LTD) in Purkinje cells [2,18]; LTD is a putative cellular model of cerebellar information storage [2,18]. Mice that lack the gene that encodes GluRd2 display ataxia and impaired LTD [11]. However, the mechanisms by which GluRd2 participates in cerebellar functions are largely unknown. GluRd2 does not form functional glutamategated ion channels when expressed, either alone or with other glutamate receptors (GluRs), in heterologous cells [2,18], nor does it bind glutamate analogs [18]. Recently, the ataxic lurcher mouse (Lc) was shown to

0169-328X / 02 / $ – see front matter  2002 Elsevier Science B.V. All rights reserved. PII: S0169-328X( 02 )00561-2

28

K. Kohda et al. / Molecular Brain Research 110 (2003) 27–37

result from a point mutation in the gene that encodes GluRd2 [28]; the mutant GluRd2 Lc channels are constitutively active in the absence of ligand. The mutation is located in a highly conserved motif at the end of the putative third transmembrane domain (TM3). We previously demonstrated in heterologous cells that GluRd2 Lc forms functional homomeric channels [12]. Currents through GluRd2 Lc display double rectification when a glutamine (Q) residue is present in the TM2, a feature that is similar to those of currents through wild-type AMPA or kainate receptors with Q at the Q / R site in the TM2 [4]. In addition, currents through GluRd2 Lc were reversibly blocked by 1-naphthylacetylspermine (Nasp), a polyamine analog that is an open-channel blocker of AMPA and kainate receptors [4]. Thus, we proposed that the channel properties of GluRd2 Lc are similar to those of AMPA and kainate receptors [12]. The question remains: Why isn’t homomeric wild-type GluRd2 in heterologous cells gated by glutamate analogs? One possible explanation is that GluRd2 may require an additional subunit (i.e. an accessory protein or additional subunit of the glutamate receptor family) to display channel functions. This is the case with the NMDA receptor NR3A [6,7,27] and NR3B [19]. Although NR3A alone does not bind ligand [6], NR3A / B in association with NR1 and NR2 forms ligand-gated channels in vitro [19,24] and in vivo [7]. Since GluR2 / 3, GluR6, and KA2 were not co-immunoprecipitated from cerebellar lysates by anti-GluRd2 antibody [20], the possibility that GluRd2 may assemble with AMPA or kainate receptor has not been further investigated. However, it is not uncommon that receptors and receptor-related molecules cannot be coimmunoprecipitated with their partner proteins from brain tissues even if they form functional complexes in vitro and in vivo. For example, NR1-associating protein Yotiao, NR2-associated protein NIL-16, AMPA receptor-associated Ca 21 channel subtype Stargazin, and GluRd2-associated protein delphilin were not co-immunoprecipitated from brain tissues [5,13,16,21]. In contrast to the coimmunoprecipitation assay, an immunogold electronmicroscope analysis has revealed that Glud2 is colocalized with GluR2 / 3 in Purkinje cell spines [14]. The aim of the present study was to investigate whether GluRd2 could assemble with AMPA or kainate receptors, and if so, how it affected the properties of heteromeric channels in heterologous cells. We also tested the association between GluRd2 and AMPA and kainate receptors in vivo.

2. Material and methods

2.1. Clones and transfection AMPA receptors GluR1 and GluR2 were the ‘flop’ version. Kainate receptor GluR6 encoded valine and cysteine at the RNA editing site on TM1. For NMDA receptor studies, we used NR1-4a. Site-directed muta-

genesis was carried out as previously described [12]. The expression vectors pTracer-EGFP and pCAGGS were used unless otherwise stated [12]. For coexpression studies, the pTracer-mitCFP was created by replacing the gene that encodes super green fluorescent protein (GFP) in pTracer (Invitrogen, Carlsbad, CA, USA) with a gene that encodes cyan fluorescent protein (CFP) fused with the mitochondrial localization signal, and the pTracer-nucYFP was created by replacing the same gene with one that encodes yellow fluorescent protein (YFP) fused with the nuclear localization signal (Clontech, Palo Alto, CA, USA). GluR clones were transfected into HEK293 cells (American Type Culture Collection, Rockville, MD, USA) by a calcium-phosphate precipitation method (CellPhect, Pharmacia, Piscataway, NJ, USA) or using LipofectAMINE reagents (Life Technologies, San Diego, CA, USA). DNA (1.5 mg) was used to transfect HEK293 cells that were prepared on an 18-mm plastic coverslip (Aclar, Pro Plastics, Linden, NJ, USA) for electrophysiological recordings. Cotransfection was carried out with plasmid DNAs (ratio of plasmids, 1:1) that expressed the clones of interest. When single transfection was used as a control, the total amount of DNA was kept constant by replacing the plasmid DNA encoding the clone of interest with that encoding YFP or CFP. For immunoprecipitation studies, cells were plated on a 6-cm dish and transfected with 5 mg of total plasmid DNA per dish. Two days after transfection, the cells were used for experiments.

2.2. Electrophysiological recordings of transfected HEK293 cells Transfected cells were identified by the fluorescence of EGFP, CFP, or YFP. Whole-cell patch-clamp recordings were carried out at room temperature using an Axopatch 200B (Axon Instruments, Foster City, CA, USA). Thinwall borosilicate glass pipettes (TW150F, World Precision Instruments, Sarasota, FL, USA) had resistances of 2 to 4 MV when filled with intracellular solution (150 mM CsCl, 1 mM MgCl 2 , 10 mM HEPES, and 10 mM EGTA [pH 7.3]). The series resistance was less than 10 MV and was compensated by 60 to 70%. The extracellular solution was composed of 150 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 20 mM HEPES, and 20 mM D-glucose (pH 7.3). To obtain I–V profiles, ramp stimulation of 2100 to 100 mV was applied. To apply drugs, cells were lifted to a flow pipe constructed from theta glass tubing (tip diameter, 250 mm; Sutter Instruments, Novato, CA, USA) and placed in the control-solution stream close to the interface between continuously flowing control and drug solutions. Solutions were gravity-fed into each of the two lumens of theta glass tubing at a rate of 50–70 ml / min. Solution exchange was made by rapidly moving the theta glass with a Piezo translator (LSS-3100, Burleigh Instruments, Fishers, NY, USA). In this system, the time required for solution exchange was typically 300 ms, as determined by measure-

K. Kohda et al. / Molecular Brain Research 110 (2003) 27–37

ments of open-tip junction potentials after the cell was blown off the tip of the electrode at the end of every experiment. Responses were filtered at 10 kHz with an eight-pole Bessel filter, which was digitized at 30–100 kHz. Data acquisition and analysis were done by custom software. Noise analysis was carried out as previously described [12].

2.3. Immunoprecipitation from HEK293 cells Transfected HEK293 cells were washed with PBS and suspended in 1 ml of a buffer solution that contained 150 mM NaCl, 50 mM Tris–Cl, 2 mM EDTA (pH 8.0), and a cocktail of protease and phosphatase inhibitors (PPI) that contained (final concentrations) 50 mM NaF, 10 mM pepstatin A, 0.5 mM phenylmethylsulfonyl fluoride, 10 mM leupeptin, 10 mM b-glycerophosphate (Sigma, St Louis, MO, USA), and 10 mM okadaic acid (RBI, Natick, MA, USA). The following agents (final concentrations) were added to the cell suspensions: Triton X-100 (0.5%), deoxycholate (DOC; 0.05%; Sigma) and sodium dodecylsulfate (SDS; 0.1%). Cell membranes were solubilized by rocking at 37 8C for 30 min and then at 4 8C for 30 min. The lysate was centrifuged at 100 0003g for 30 min at 2 8C. We added 4 mg of the antibody to the samples and incubated them for 90 min at 4 8C. Protein G-agarose (50 ml; Roche, Indianapolis, IN, USA) was added to the lysate and incubated for 90 min at 4 8C. Protein G-agarose beads were thoroughly rinsed four times with a buffer containing 150 mM NaCl, 50 mM Tris–Cl (pH 6.8), 5 mM EDTA, 0.5% NP40, 0.01% DOC, and PPI cocktail. Proteins were boiled for 4 min and eluted with 50 ml of 23 SDS gel-loading buffer. We confirmed that each antibody efficiently precipitated the target protein by performing immunoblot analysis on the supernatant fraction. We omitted preabsorption in the immunoprecipitation experiments, because preliminary tests indicated that this procedure did not affect the results. As a negative control, each clone of interest was transfected independently and equal volumes of each cellsuspension solution were mixed together before solubilization. These postmixture samples were subjected to the same procedures that cotransfected samples were subjected to. Equal volumes of the membrane fraction (1% of this fraction) and the pellet fraction (8% of this fraction) were examined by SDS–PAGE (7.5% acrylamide). Proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Immobilon-P, Millipore, Bedford, MA, USA), allowed to react with immunoreagents, and visualized by the chemiluminescence detection system ECL Plus (Amersham Pharmacia Biotech, Buckinghamshire, UK).

2.4. Surface biotinylation assay Cell-surface molecules of transfected HEK293 cells were biotinylated by 1 mg / ml Sulfo-NHS-biotin (Pierce

29

Chemical Co., Rockford, IL, USA) in PBS that contained 4.5 mg / ml D-glucose for 15 min at room temperature. Cells were rinsed three times with ice-cold PBS and solubilized in a buffer solution containing 0.5 M NaCl, 20 mM PBS (pH 7.5), 1 mM EDTA, 1% NP40, 0.5% DOC, 0.1% SDS, and PPI cocktail for 60 min at 4 8C. After centrifugation at 11 5003g for 20 min at 4 8C, the supernatant was subjected to immunoprecipitation by antiGluR1 antibody and immunoblotted with avidin. For recapture of biotinylated proteins, the pellet of Protein G-agarose was washed four times with the buffer containing 150 mM NaCl, 50 mM Tris–Cl (pH 6.8), 5 mM EDTA, 0.1% Triton X-100, and PPI cocktail. After the samples cooled to room temperature, 1 ml of the following buffer solution was added: 150 mM NaCl, 50 mM Tris–Cl (pH 8.0), 2 mM EDTA, 10 mM iodoacetamide (Sigma), 1% Triton X-100 and PPI cocktail, and samples were incubated at room temperature for 10 min. Recapture was done by incubating the samples in 50 ml of immobilized Streptavidin agarose (ImmunoPure, Pierce Chemical Co.) for 90 min at 4 8C. After the samples were washed four times with the buffer containing 150 mM NaCl, 50 mM Tris–Cl (pH 6.8), 5 mM EDTA, 0.1% Triton X-100, and PPI cocktail, proteins were eluted with 50 ml of 23 SDS gel-loading buffer.

2.5. Immunoprecipitation of cerebellar lysates Four- to 6-week-old C57 / BL6 wild-type mice were anesthetized with ether and decapitated, and their cerebella were dissected. Grid2 ho- 5 J mice (The Jackson Laboratory, Bar Harbor, ME, USA) were used as GluRd2 null mice because they do not express GluRd2 proteins (Y. Wang, S. Matsuda, Y. Kamiya, and M. Yuzaki, manuscript in preparation). We homogenized the cerebella in 0.32 M sucrose in Buffer A (10 mM Tris–Cl [pH 8.0], 1 mM EDTA, and PPI cocktail at 4 8C) using a glass–Teflon homogenizer (10 stokes at 800 rpm) and centrifuged at 9003g for 5 min. The supernatant was centrifuged again at 11 5003g for 20 min and washed once with a solution of Buffer A and PPI cocktail. The pellet was suspended in the following buffer solution: 50 mM Tris–Cl (pH 8.0), 20 mM EDTA, 1% NP40, 0.5% DOC, and 0.1% SDS with PPI cocktail and solubilized by rocking for 90 min at 4 8C. Immunoprecipitation procedures used for cerebellar lysates were the same as those described above for HEK293 lysates, except for the washing buffer, which contained 50 mM Tris–Cl (pH 8.0), 20 mM EDTA, 0.1% NP40, and 0.05% DOC with PPI cocktail.

2.6. Antibodies Polyclonal antibodies anti-GluR1, anti-GluR2, antiGluR2 / 3 and monoclonal anti-GluR2 / 4 were purchased from Pharmingen. Polyclonal anti-GluRd1 / 2 antibody was obtained from Chemicon International (Temecula, CA, USA); polyclonal anti-GluR6 / 7 from Upstate Biotechnol-

30

K. Kohda et al. / Molecular Brain Research 110 (2003) 27–37

ogy (Lake Placid, NY, USA); polyclonal anti-Erk1 / 2, from StressGen Biotechnologies (Victoria, BC, Canada); and polyclonal anti-actin, from Sigma.

3. Results

3.1. GluRd2 Lc changed channel properties of coexpressed GluR1 Lc or GluR6 Lc channels Properties of homomeric and heteromeric glutamate receptors are determined by the presence of glutamine (Q) or arginine (R) at the Q / R site in the TM2. The presence of R dominantly determines the properties of heteromeric channels. If heteromeric receptors contain a subunit that has R at the Q / R site, they show linear I–V curves, low Ca 21 permeability, and insensitivity to Nasp [4]. We used human embryonic kidney (HEK) 293 cells to test whether GluRd2, in which the Q→R mutation is introduced at the Q / R site, can form functional channels with other GluRs that have Q at the Q / R site. To directly test the ability of GluRs to form heteromeric channels without considering the ligand-binding site, we used lurcher (Lc) mutant

versions of GluRs, which are constitutively active in the absence of ligand [12]. HEK293 cells expressing the AMPA receptor GluR1 Lc that has Q at the Q / R site [GluR1 Lc (Q)] displayed strong inward and weak outward rectification; the AMPA receptor GluR2 Lc (R) showed no rectification. Like cells with wildtype receptors, cells cotransfected with GluR1 Lc (Q) and GluR2 Lc (R) showed nonrectifying I–V curves (Fig. 1A, top panels), a finding that confirmed that the R site dominantly determines the properties of lurcher mutant channels. When we coexpressed GluR1 Lc (Q) and GluRd2 Lc (R), the cells showed I–V curves with very weak rectification (Fig. 1A, middle panels). If GluR1 Lc (Q) and GluRd2 Lc (R) had not formed heteromeric channels, the summation of each I–V curve would have displayed much stronger rectification (Fig. 1A,C, red lines). We obtained similar results with the coexpression of the kainate receptor GluR6 Lc (Q) and GluRd2 Lc (R) (Fig. 1A, bottom panels). These results indicate that GluRd2 Lc (R) formed functional heteromeric channels with GluR1 Lc or GluR6 Lc receptors in vitro. I–V curves in cells expressing GluR1 Lc (Q) and GluR2 Lc (R) had essentially no rectification. However, we noticed that cells coexpressing GluR1 Lc (Q) and

Fig. 1. Cotransfection of GluRd2 Lc (R) changes the rectification of current–voltage (I–V ) curves. (A) Mean I–V curves were determined from recordings of HEK293 cells that expressed lurcher (Lc) mutant glutamate receptors (GluRs) with either glutamine (Q) or arginine (R) residues at the putative channel pore. Left panels show the mean I–V curves from cells that expressed only GluR1 Lc (Q) (n512) or only GluR6 Lc (Q) (n512). Middle panels show mean I–V curves from cells that expressed GluR2 Lc (R) (n57) or GluRd2 Lc (R) (n513). Right panels show the mean I–V curves from cells that were cotransfected with GluR1 Lc (Q) and GluR2 Lc (R) (n511), GluR1 Lc (Q) and GluRd2 Lc (R) (n516), or GluR6 Lc (Q) and GluRd2 Lc (R) (n513). Red lines indicate the I–V curves expected from the simple summation of I–V curves from each receptor. (B) The mean I–V curves from cells that were cotransfected with GluR2 Lc (Q) and GluRd2 Lc (R) (n58). Red lines indicate the I–V curves expected from the simple summation of I–V curves from each receptor. (C) Rectification indices (RI) of I–V curves from these cells. RI was calculated by dividing slope conductance ( g) at 280 mV ( g280 ) by that at 40 mV ( g40 ) in each cell. Each bar represents the mean6S.E.M. Red dotted lines indicate the expected RI of I–V curves calculated by simple summation of I–V curves from each receptor.

K. Kohda et al. / Molecular Brain Research 110 (2003) 27–37

GluRd2 Lc (R) or GluR6 Lc (Q) and GluRd2 Lc (R) showed I–V curves with very weak rectification (Fig. 1A,C). In addition, we occasionally observed strong rectification in cells expressing GluR1 Lc (Q) and GluRd2 Lc (R) (three of 17 cells) or GluR6 Lc (Q) and GluRd2 Lc (R) (three of 14 cells). In contrast, cells cotransfected with GluR1 Lc (Q) and GluR2 Lc (R) never showed rectification (0 of 12 cells), a finding that suggests that most GluR1 Lc (Q) had formed heteromers with GluR2 Lc (R). Therefore, GluRd2 Lc (R) may have had weaker affinity for GluR1 Lc (Q) than did GluR2 Lc (R). If GluR1 Lc (Q) or GluR6 Lc (Q) existed partially in a homomeric form, the currents through those homomeric receptors would have contributed to the rectification of the I–V curves. In Purkinje cells, native AMPA receptors are formed by GluR1, 2, and 3 [3,9,25] and kainate receptors are formed by GluR5, KA1 and KA2 [9]. Because the properties of channels formed by GluR1 and GluR6 have been best characterized, we used GluR1 as a representative of AMPA receptors and GluR6 for kainate receptors. Because subunits in the same subfamily (AMPA or kainate receptors) assemble each other in a similar fashion [23], we consider that results obtained with GluR1 and GluR6 can be applied to other subunits. To confirm this hypothesis, we examined if GluRd2 could make functional heteromers with GluR2. As expected, when we coexpressed GluR2 Lc (Q) and GluRd2 Lc (R), the cells showed I–V curves with very weak rectification (Fig. 1B,C).

3.2. GluRd2 Lc changed NASP sensitivity of coexpressed GluR1 Lc or GluR6 Lc channels One disadvantage of cotransfection experiments is that some cells may be transfected with only one of the vectors. To ensure that recorded cells were transfected with two vectors, we used an expression vector for GluRd2 Lc that also expressed yellow fluorescent protein with a nuclear localizing signal (nucYFP), and we used an expression vector for GluR1 Lc that expressed cyan fluorescent protein with a mitochondrial localization signal (mitCFP). Although the expression levels of the two fluorescent proteins varied among cells (Fig. 2A), most showed both signals, verifying their transfection with both vectors. By recording from cells that clearly showed both fluorescent signals, we confirmed that cells coexpressing GluR1 Lc (Q) and GluRd2 Lc (R) showed very weak rectification (Fig. 2B). To further confirm functional assembly of GluR1 Lc (Q) and GluRd2 Lc (R) in cotransfected cells, we assessed the sensitivity of currents to Nasp. As previously reported [12], treatment with 10 mM Nasp substantially reduced currents in cells that expressed only GluR1 Lc (Q); currents were reduced in a membrane potential-dependent manner (Fig. 2B, left). However, Nasp did not alter currents in cells that expressed only GluRd2 Lc (R) (Fig. 2B, center). In cells that coexpressed GluRd2 Lc (R) and GluR1 Lc (Q),

31

currents became essentially insensitive to Nasp (Fig. 2B, right). If GluR1 Lc (Q) and GluRd2 Lc (R) had not formed heteromeric channels, the current component mediated by GluR1 Lc (Q) would have been blocked by Nasp, and total currents would have been more sensitive to Nasp (Fig. 2C, red line). These results further support the view that GluRd2 Lc (R) forms functional heteromeric channels with GluR1 Lc (Q). Although GluRs and fluorescent proteins were encoded by the same expression vectors, their expression was controlled by independent promoters. Thus, it is theoretically possible that cotransfected cells that showed nonrectifying I–V curves did not express GluR1 Lc (Q); however, this possibility is unlikely because application of kainate to cells cotransfected with GluR1 Lc (Q) and GluRd2 Lc (R) induced large currents (Fig. 2D, right). Because cells expressing only GluRd2 Lc (R) did not respond to kainate, these results clearly indicate that GluR1 Lc (Q) was expressed in cells cotransfected with GluR1 Lc (Q) and GluRd2 Lc (R). Interestingly, kainate-induced currents in cells cotransfected with GluR1 Lc (Q) and GluRd2 Lc (R) displayed rectification, though the constitutive currents showed no rectification (Fig. 2D, right). Thus, GluR1 Lc (Q) existed partially in a homomeric form, but the currents through those homomeric receptors may have minimally contributed to the rectification of constitutive currents. However, because kainate-induced currents were much larger than constitutive currents, rectification may have become apparent. To confirm this hypothesis, we applied kainate to cells expressing GluR1 Lc (Q) alone (Fig. 2D, left). Kainate induced an approximately 4.4-fold increase in currents at 260 mV in these cells (Fig. 2E, left), a result consistent with the view that kainate-induced currents in cells cotransfected with GluR1 Lc (Q) and GluRd2 Lc (R) originated from homomeric GluR1 Lc (Q) channels. Because kainate-induced increase in current in cells coexpressing GluR1 Lc (Q) and GluRd2 Lc (R) is approximately 20% of that in cells expressing GluR1 Lc (Q) alone (Fig. 2E), 20% of Lc GluR1 (Q) may exist as homomeric channels in cells coexpressing GluR1 Lc (Q) and GluRd2 Lc (R).

3.3. Co-immunoprecipitation studies of GluRd2 assembly To further confirm the association of GluRd2 with GluR1 or GluR6, we cotransfected HEK293 cells with vectors encoding wild-type GluRd2 and GluR1 and immunoprecipitated cell lysates with the antibody against each protein. As expected, antibodies against GluR1 coimmunoprecipitated GluRd2, and antibodies against GluRd2 co-immunoprecipitated GluR1 (Fig. 3A). When lysates from cells that had been singly transfected with either GluRd2 or GluR1 were mixed together, co-immunoprecipitation was not observed (Fig. 3A); a finding that indicates that co-immunoprecipitation was not caused by

32

K. Kohda et al. / Molecular Brain Research 110 (2003) 27–37

Fig. 2. Confirmation of coexpression of GluR1 Lc (Q) and GluRd2 Lc (R) and changes in sensitivity to Nasp. (A) Cells were transfected with pTracer-mitCFPGluR1 Lc (Q), which expressed GluR1 Lc (Q) and cyan fluorescent protein in the mitochondria and with pTracer-nucYFP-GluRd2 Lc (R), which expressed GluRd2 Lc (R) and yellow fluorescent protein in the nucleus. Although most cells were cotransfected, cells showing only one fluorescent signal (arrowhead) were occasionally observed. (B) Sensitivity to the AMPA receptor antagonist Nasp. The left panel shows the mean I–V curve from cells cotransfected with pTracer-mitCFP-GluR1 Lc (Q) and pTracer-nucYFP (n512). Nasp (10 mM) blocked strong rectifying currents. The second panel shows the mean I–V curve from cells cotransfected with pTracer-mitCFP and pTracer-nucYFP-GluRd2 Lc (R) (n56). Currents did not show clear rectification and were not blocked by Nasp. The right panel shows those from cells cotransfected with pTracer-mitCFP-GluR1 Lc (Q) and pTracer-nucYFP-GluRd2 Lc (R) (n513). Currents did not show strong rectification and were mostly insensitive to Nasp. (C) The mean percentage (6S.E.M.) of Nasp-induced reduction of currents at 260 mV is presented. The dotted red line indicates the percentage of reduction expected from the simple summation of each current. (D) Evidence of expression of GluR1 Lc (Q) in cells that showed a nonrectifying I–V curve. Left panel shows the representative I–V curve from cells that were cotransfected with pTracer-mitCFP-GluR1 Lc (Q) and pTracer-nucYFP-mock (n514). Right panel shows the representative I–V curve from cells that were cotransfected with pTracer-mitCFP-GluR1 Lc (Q) and pTracer-nucYFP-GluRd2 Lc (R) (n513). Blue lines indicate the I–V curves during application of 3 mM kainate (1kai). Although cells coexpressing GluR1 Lc (Q) and GluRd2 Lc (R) initially showed nonrectifying I–V curves, application of kainate induced a rectifying current. (E) The mean constitutive and kainate-induced currents (6S.E.M.) at 260 mV is presented.

the cross-reaction of antibodies with other receptors. We obtained the same results when this experiment was repeated with cells cotransfected with GluR6 and GluRd2 (Fig. 3A). Thus, in HEK293 cells, GluRd2 formed closely associated protein complexes with GluR1 or GluR6; a finding that is consistent with the results of our electrophysiological studies. After immunoprecipitation with the anti-GluR1 antibody, we noticed that some GluRd2 was recovered in the supernatant fraction (Fig. 3A). In contrast, GluR2 proteins were never observed in the supernatant fraction after

immunoprecipitation with anti-GluR1 antibody (data not shown), a finding that indicates that the GluR1–GluR2 complex was more stable than the GluR1–GluRd2 complex. These data are also consistent with the electrophysiological data indicating the occurrence of GluR1 Lc homomeric channels in cells coexpressing GluR1 Lc and GluRd2 Lc (Figs. 1A, 2D). To test whether receptors that do not normally associate with each other would form protein complexes when overexpressed, we coexpressed GluR1 and GluR6 in HEK293 cells. As reported earlier [15], GluR1 and GluR6

K. Kohda et al. / Molecular Brain Research 110 (2003) 27–37

33

Fig. 3. Co-immunoprecipitation of GluRd2 with AMPA or kainate receptors expressed in HEK293 cells. (A) Solubilized membranes from cells transfected with indicated cDNAs were immunoprecipitated (ip) with the following antibodies: anti-GluR1, anti-GluRd2, or anti-GluR6. Equal amounts of solubilized membrane (M), supernatant (S), and pellet (P) fractions were electrophoretically separated and analyzed by immunoblotting with antibodies against the indicated proteins. To ensure that co-immunoprecipitated protein complexes were formed in cotransfected cells (co), immunoprecipitation was also carried out on a mixture of cell lysates prepared from cells singly transfected with each receptor (mix). Representative results from five independent experiments are shown. (B) Partial co-immunoprecipitation of GluR1 and GluR6. Solubilized membranes from cells transfected with GluR1 and GluR6 were immunoprecipitated by anti-GluR6 or anti-GluR1 antibodies and analyzed by immunoblotting, as described above. (C) Association of GluRd2 and GluR1 on the cell surface. HEK293 cells that coexpressed GluR1 and GluRd2 were treated with the membrane-impermeable biotinylation reagent. Solubilized membranes were immunoprecipitated by anti-GluR1 antibody, recaptured (re-ip) by streptavidin-conjugated agarose beads, and analyzed by immunoblotting with antibodies against the indicated proteins. Biotinylated GluRd2 was detected in the recaptured fraction (re-P), whereas actin or Erk1 / 2 was not detected in the P fraction. Representative data from three independent experiments are shown. (D) Association of GluRd2 with GluR2 or KA2 receptors in the cerebellum. Solubilized mouse cerebellar membranes were immunoprecipitated by anti-GluR2 or anti-KA2 antibody. Equal amounts of solubilized membrane (M), supernatant (S), and pellet (P) fractions were electrophoretically separated and analyzed by immunoblotting with anti-d2 antibody. To ensure that the observed co-immunoprecipitation was specific to antibodies (1Ab), reactions were also carried out without antibodies (2Ab). Co-immunoprecipitation of GluRd2 from GluRd2 null mice by anti-GluR2 or anti-KA2 antibodies was also shown as a negative control. Representative results of five independent experiments are shown.

co-immunoprecipitated only weakly from the lysates of these cells (Fig. 3B). Thus, the association of GluRd2 with GluR1 and of GluRd2 with GluR6 was much stronger than that of GluR1 with GluR6; therefore, the co-immunoprecipitation of GluRd2 with GluR1 or GluR6 cannot be explained by the overexpression of these proteins. We noticed that the band for GluRd2 occasionally appeared slightly smaller in the co-immunoprecipitated fraction by anti-GluR1 or anti-GluR6 antibodies than in the supernatant fraction (Fig. 3A). This observation raised the possibility that co-immunoprecipitated heteromer complexes were not efficiently transported to the membrane surface and were underglycosylated. We biotinylated surface molecules on plasma membranes and confirmed that the

GluR1–GluRd2 complex was located on the cell surface (Fig. 3C). The specificity of the biotinylation of membrane proteins was confirmed by the absence of labeling of actin or Erk1 / 2, which are localized in the cytosol (Fig. 3C). Taken together, these findings indicate that in HEK293 cells, GluRd2 forms heteromers with GluR1 or GluR6, and these heteromers exist at least partly on the cell surface. It has been reported that GluR2 / 3, GluR6, and KA2 are not co-immunoprecipitated from cerebellar lysates by antiGluRd2 antibody [20]. We obtained similar results with immunoprecipitation from cerebellar lysates by antiGluRd2 antibody (data not shown). However, immunoprecipitation efficiency often depends on each antibody and its epitope. In addition, the failure to co-immuno-

34

K. Kohda et al. / Molecular Brain Research 110 (2003) 27–37

precipitate may have been caused by the fact that GluRd2 is expressed only in Purkinje cells, which represent an extremely small percentage of cells in the cerebellum. Thus, we examined whether GluRd2 could be co-immunoprecipitated by anti-GluR2 or anti-KA2 antibody. AntiGluR2 antibody partially co-immunoprecipitated GluRd2 from a solubilized membrane fraction of the cerebellum (Fig. 3D). Co-immunoprecipitation of GluR2 and GluRd2 did not occur when the anti-GluR2 antibody was omitted or when GluRd2-null cerebellum was used (Fig. 3D). Similarly, anti-KA2 antibody partially co-immunoprecipitated GluRd2 from the wild-type cerebellum, whereas they did not co-immunoprecipitate when anti-KA2 antibody was omitted or when GluRd2-null cerebellum was used (Fig. 3D). These findings indicate that at least some proportion of GluRd2 forms a close association with either AMPA or kainate receptors in vivo.

3.4. Effect of coexpression of GluRd2 and GluR1 on glutamate-induced currents Because wild-type GluRd2 (GluRd2 wt ) does not undergo RNA editing [12], it has Q at the Q / R site. Thus, unlike GluRd2 Lc (R), GluRd2 wt does not modify the channel properties (I–V curve or sensitivity to Nasp) of the heteromeric receptor. To investigate the function of GluRd2 wt in heteromeric receptors, we coexpressed GluR1 wt and GluRd2 wt (Q) in HEK293 cells and characterized the glutamate-induced currents. We found that glutamate-induced currents in cells that coexpressed GluR1 wt and GluRd2 wt were significantly smaller than those in cells that expressed GluR1 wt alone (Fig. 4A,B). Coexpression of GluR1 wt and the NMDA receptor NR1-4a, which is exported to the cell surface but does not associate with GluR1 [22], did not reduce the glutamate-induced response. In addition, coexpression of GluRd2 wt and GluR1 wt did not affect the expression of GluR1 wt on the cell surface (Fig. 4C). These results indicate that the reduction of glutamate-induced currents was caused by the formation of heteromeric channels that consisted of GluRd2 wt and GluR1 wt . The next question was whether GluR1 wt –GluRd2 wt heteromers are functional. Heteromeric receptors may have less affinity for glutamate or may have reduced singlechannel conductance and, as a result, they may show reduced macroscopic glutamate-induced currents. The glutamate dose–response curve generated in cells that coexpressed GluR1 wt and GluRd2 wt was similar to that generated in cells that expressed GluR1 wt alone (Fig. 4D). In addition, there was no difference in desensitization kinetics among these cells (Fig. 4A, scaled trace). Moreover, there was no significant difference in mean single-channel conductance among these cells (Fig. 4E). These results suggest that glutamate-induced currents observed in cells that coexpressed GluR1 wt and GluRd2 wt originated from GluR1 wt homomers, which were also present in these cells.

Thus, the heteromeric channel created by the association of GluR1 wt and GluRd2 wt may not be gated by glutamate and does not appear to participate in glutamate-induced currents. This view is also consistent with the reduction of kainate-induced currents in cells coexpressing GluR1 Lc (Q) and GluRd2 wt (R) (Fig. 2D,E).

4. Discussion

4.1. Assembly of GluRd2 with AMPA or kainate receptors in heterologous cells By using channels with the lurcher mutation, we demonstrated that GluRd2 formed heteromeric channels with the AMPA receptor GluR1 or the kainate receptor GluR6 in heterologous cells. The finding was also compatible with our co-immunoprecipitation assays. Thus, channel properties of GluRd2 are not only similar to those of AMPA and kainate receptors [12], but these receptors can form functional heteromeric channels with each other. In contrast to lurcher channels, we demonstrated that heteromers formed by wild-type GluR1 and wild-type GluRd2 were not functional: cells coexpressing GluR1 and GluRd2 had significantly smaller glutamate-induced currents than cells expressing GluR1 alone. It has been suggested that AMPA receptors have four ligand-binding sites and that the progressive occupancy of these sites leads to greater single-channel conductance [26]. Thus, heteromeric receptors containing GluRd2, which do not bind glutamate, may have a low mean channel conductance. Thus, the finding that there was no significant difference in mean single-channel conductance between cells coexpressing GluR1 and GluRd2 and cells expressing GluR1 alone suggests that heteromeric receptors probably do not regulate channel openings or do so only rarely. It was previously reported that glutamate- or kainateinduced currents among HEK293 cells that express GluR2 or GluR5 alone are comparable to those induced in cells that coexpress GluRd2 [18]. These experiments were probably designed to detect the enhancement of agonistinduced currents by coexpression of GluRd2. Because GluR2 and GluR5 receptors, especially those containing R at the Q / R site, give rise to very small agonist-induced currents, it would have been very difficult to detect the reduction of currents by coexpression of GluRd2. Furthermore, because of the nature of cotransfection experiments, some cells may have been transfected with only one vector. If the currents induced in homomeric cells were included, that data would also have made it difficult to detect the effect of coexpression of GluRd2. Thus, the discrepancy between our data and this previous report [18] may have been caused by the difference in experimental designs. GluRd2 is unique in that it can form heteromers with both AMPA and kainate receptors; it is reported that

K. Kohda et al. / Molecular Brain Research 110 (2003) 27–37

35

Fig. 4. Coexpression of GluRd2 wt reduced glutamate-induced currents without changing channel properties of GluR1 wt . (A) Glutamate (3 mM)-induced currents in cells that expressed GluR1 wt alone (left) or GluR1 wt and GluRd2 wt (right). Cotransfected cells were whole-cell voltage clamped at 260 mV. Note that cotransfection essentially did not alter desensitization kinetics. (B) Summary of the amplitude of glutamate-induced current in cells that expressed GluR1 wt alone (1mock), GluR1 and GluRd2, or GluR1 and NR1-4. Each bar represents the mean6S.E.M. (n59). **Statistically significant differences (P,0.01 as determined by the Mann–Whitney U-test). (C) Surface expression of GluR1 in cells expressing GluR1 alone (mock) or GluR1 and GluRd2 receptors. Cells cotransfected with GluR1 and GluRd2 were treated with the membrane-impermeable biotinylation reagent; solubilized membranes were immunoprecipitated by anti-GluR1 and analyzed by immunoblotting with avidin (left) or anti-GluR1 (right). Representative results of three independent experiments are shown. (D) Dose–response relationships to glutamate. To evaluate the macroscopic affinities of heteromeric receptors, we applied various doses of glutamate to cells expressing GluR1 alone (closed circles) or GluR1 and GluRd2 (open circles). For each cell, the data were normalized to the current evoked by a saturating dose. Points indicate the mean6S.E.M. of the amplitude values of the steady-state current from four to six cells; the curves indicate the best fit to the data, according to the logistical function 1 /(11(EC 50 / [agonist] n H , in which EC 50 is the concentration of glutamate that causes a 50% maximal response, and n H is the Hill coefficient. (E) Noise analysis of glutamate-induced currents in cells that coexpress GluR1 and GluRd2. Representative steady-state glutamate-induced currents in cells that express GluR1 alone (upper trace) or GluR1 and GluRd2 receptors (lower trace) are shown. Glutamate (10 mM) and cyclothiazide (CTZ; 50 mM) were applied to minimize the desensitization of current and to ensure the low open-channel probability. Variance-to-mean current plots are shown with regression lines that intersect the origin. The mean channel conductance (g ) was calculated using the slope of the regression line.

GluR1 and GluR6 do not coassemble each other in heterologous cells [23]. It is still possible that very few GluR1–GluR6 heteromers may exist but they were not detected in this previous study, because the heteromers may not be effectively gated by agonists specific to AMPA or kainate receptors. However, GluR1 and GluR6 did not co-immunoprecipitate comparably under the same conditions in vitro (Fig. 3B). Furthermore, unlike GluRd2, NR1-4 did not reduce glutamate-induced currents when

coexpressed with the GluR1 receptor (Fig. 4B). Thus, we propose that the observed GluRd2–GluR coassembly is specific and cannot be explained by the effect of overexpression of receptor proteins in heterologous cells.

4.2. Association of GluRd2 with AMPA or kainate receptors in vivo It has been reported that GluR2 / 3, GluR6, and KA2 are

36

K. Kohda et al. / Molecular Brain Research 110 (2003) 27–37

not co-immunoprecipitated from cerebellar lysates by antiGluRd2 antibody [20]. However, we found that GluRd2 was partially co-immunoprecipitated by antibodies against GluR2 or KA2; this apparent discrepancy may have been caused by the fact that GluRd2 is expressed only in Purkinje cells, which represent an extremely small percentage of cells in the cerebellum. In addition, the association of proteins in vivo is often not detected by the immunoprecipitation assay [5,13,16]. This insensitivity is attributed to the solubilization conditions and the relatively low levels of expression of the target proteins. An immunogold electronmicroscope analysis has also revealed that Glud2 is colocalized with GluR2 / 3 in Purkinje cell spines [14]; therefore, we believe that at least a portion of GluRd2 is closely associated with AMPA or kainate receptors in vivo. The question of how GluRd2 functions by interacting with AMPA or kainate receptors in vivo remains unanswered. It is intriguing to point out the similarity between GluRd2 and the NMDA receptor NR3A / B. Like GluRd2, NR3A alone does not bind ligand [6], but when assembled with NR1 and NR2 subunits either in vitro [6,19,27] or in vivo [7], NR3A / B formed channels that has reduced glutamate-induced currents. Thus, on the basis of the results of our in vitro studies, we postulate that GluRd2 coassembles with AMPA or kainate receptors in vivo (Fig. 5B) and reduces the glutamate-induced currents of heteromeric channels. Although only a part of the GluRd2 is associated with AMPA or kainate receptors in vivo, the formation of the receptor complexes may be regulated by unknown factors. Neuronal activity causes rapid and dynamic changes in the

subunit composition of AMPA receptors in neurons [17]. Since LTD is caused by reduced synaptic glutamate responses in Purkinje cells [10], it is possible that formation of GluRd2–GluR heteromers contributes to the induction of LTD. If this hypothesis is true, glutaminergic synaptic currents in GluRd2 2 / 2 Purkinje cells would be greater than those in wild-type Purkinje cells. However, because synaptogenesis between granule cells and Purkinje cells is impaired in GluRd2 2 / 2 mice, this hypothesis cannot be tested directly. Alternatively, the close association between GluRd2 and AMPA or kainate receptors in vivo may be caused by indirect mechanisms, such as binding to the same anchoring protein or localization on the same membrane raft. For example, AMPA receptors [1] and GluRd2 [8] are anchored to the actin cytoskeleton in the dendritic spine. However, although NR1 is also anchored to the actin cytoskeleton [1], NR1 does not co-immunoprecipitate with GluR1. Even if GluRd2 does not form heteromeric channels in vivo, we believe that the proximity of these receptors to GluR1 has functional significance. For example, GluRd2 may regulate the surface expression or endocytosis of GluRs, as does the g-1 subunit of Ca 21 channels in stargazer mutant mice [5]. Interestingly, the interaction between g-1 and GluRs was also not detected by co-immunoprecipitation assay, probably because the interaction was transient [5]. The present study demonstrated that GluRd2 assembles with GluR1 or GluR6 in vitro and forms close associations with these receptors in vivo. As no functions have been assigned to GluRd2, these findings go a long way towards understanding the role of this receptor.

Fig. 5. Models of the association of GluRd2 with GluR in HEK293 cells. (A) GluRd2 Lc and GluRLc form functional heteromeric receptors. Because lurcher (Lc) mutant receptors are constitutively active in the absence of ligand, the heteromeric receptor is functional, and if it contains a subunit that has arginine (R) at the Q / R site, it changes channel properties, such as the I–V curve and sensitivity to Nasp. (B) GluRd2 wt and GluRwt form non-functional heteromeric receptors. When wild-type receptors were used, coexpression of GluRd2 and GluRs reduced glutamate-induced currents, because heteromers cannot be gated by glutamate.

K. Kohda et al. / Molecular Brain Research 110 (2003) 27–37

Acknowledgements We thank J. Boulter for cDNAs that encoded GluRd2, GluR2, or NR1-4; P.H. Seeburg, for cDNA that encoded GluR1; R. Dingledine, for cDNA that encoded GluR6. We also thank M. Ishiwata, Y. Takeyama, A. Hoshino and A. Saneyoshi for technical assistance. This work was supported by the Japan Society for the Promotion of Science (Y.K. and S.M.), the NIH grant NS36925, the Cancer Center Support Grant CA21765, and the American Lebanese Syrian Associated Charities (M.Y.).

References [1] D.W. Allison, V.I. Gelfand, I. Spector, A.M. Craig, Role of actin in anchoring postsynaptic receptors in cultured hippocampal neurons: differential attachment of NMDA versus AMPA receptors, J. Neurosci. 18 (1998) 2423–2436. [2] K. Araki, H. Meguro, E. Kushiya, C. Takayama, Y. Inoue, M. Mishina, Selective expression of the glutamate receptor channel delta 2 subunit in cerebellar Purkinje cells, Biochem. Biophys. Res. Commun. 197 (1993) 1267–1276. [3] A. Baude, E. Molnar, D. Latawiec, R.A. McIlhinney, P. Somogyi, Synaptic and nonsynaptic localization of the GluR1 subunit of the AMPA-type excitatory amino acid receptor in the rat cerebellum, J. Neurosci. 14 (1994) 2830–2843. [4] N. Burnashev, Calcium permeability of glutamate-gated channels in the central nervous system, Curr. Opin. Neurobiol. 6 (1996) 311– 317. [5] L. Chen, D.M. Chetkovich, R.S. Petralia, N.T. Sweeney, Y. Kawasaki, R.J. Wenthold, D.S. Bredt, R.A. Nicoll, Stargazin regulates synaptic targeting of AMPA receptors by two distinct mechanisms, Nature 408 (2000) 936–943. [6] A.M. Ciabarra, J.M. Sullivan, L.G. Gahn, G. Pecht, S. Heinemann, K.A. Sevarino, Cloning and characterization of chi-1: a developmentally regulated member of a novel class of the ionotropic glutamate receptor family, J. Neurosci. 15 (1995) 6498–6508. [7] S. Das, Y.F. Sasaki, T. Rothe, L.S. Premkumar, M. Takasu, J.E. Crandall, P. Dikkes, D.A. Conner, P.V. Rayudu, W. Cheung, H.S. Chen, S.A. Lipton, N. Nakanishi, Increased NMDA current and spine density in mice lacking the NMDA receptor subunit NR3A, Nature 393 (1998) 377–381. [8] H. Hirai, Clustering of delta glutamate receptors is regulated by the actin cytoskeleton in the dendritic spines of cultured rat Purkinje cells, Eur. J. Neurosci. 12 (2000) 563–570. [9] M. Hollmann, S. Heinemann, Cloned glutamate receptors, Annu. Rev. Neurosci. 17 (1994) 31–108. [10] M. Ito, Long-term depression, Annu. Rev. Neurosci. 12 (1989) 85–102. [11] N. Kashiwabuchi, K. Ikeda, K. Araki, T. Hirano, K. Shibuki, C. Takayama, Y. Inoue, T. Kutsuwada, T. Yagi, Y. Kang et al., Impairment of motor coordination, Purkinje cell synapse formation, and cerebellar long-term depression in GluR delta 2 mutant mice, Cell 81 (1995) 245–252. [12] K. Kohda, Y. Wang, M. Yuzaki, Mutation of a glutamate receptor motif reveals its role in gating and delta2 receptor channel properties, Nat. Neurosci. 3 (2000) 315–322. [13] C. Kurschner, M. Yuzaki, Neuronal interleukin-16 (NIL-16): a dual function PDZ domain protein, J. Neurosci. 19 (1999) 7770–7780.

37

[14] A.S. Landsend, M. Amiry-Moghaddam, A. Matsubara, L. Bergersen, S. Usami, R.J. Wenthold, O.P. Ottersen, Differential localization of delta glutamate receptors in the rat cerebellum: coexpression with AMPA receptors in parallel fiber-spine synapses and absence from climbing fiber-spine synapses, J. Neurosci. 17 (1997) 834–842. [15] W.D. Leuschner, W. Hoch, Subtype-specific assembly of alphaamino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor subunits is mediated by their n-terminal domains, J. Biol. Chem. 274 (1999) 16907–16916. [16] J.W. Lin, M. Wyszynski, R. Madhavan, R. Sealock, J.U. Kim, M. Sheng, Yotiao, a novel protein of neuromuscular junction and brain that interacts with specific splice variants of NMDA receptor subunit NR1, J. Neurosci. 18 (1998) 2017–2027. [17] S.Q. Liu, S.G. Cull-Candy, Synaptic activity at calcium-permeable AMPA receptors induces a switch in receptor subtype, Nature 405 (2000) 454–458. [18] H. Lomeli, R. Sprengel, D.J. Laurie, G. Kohr, A. Herb, P.H. Seeburg, W. Wisden, The rat delta-1 and delta-2 subunits extend the excitatory amino acid receptor family, FEBS Lett. 315 (1993) 318–322. [19] K. Matsuda, Y. Kamiya, S. Matsuda, M. Yuzaki, Cloning and characterization of a novel NMDA receptor subunit NR3B: a dominant subunit that reduces calcium permeability, Brain Res. Mol. Brain Res. 100 (2002) 43–52. [20] E. Mayat, R.S. Petralia, Y.X. Wang, R.J. Wenthold, Immunoprecipitation, immunoblotting, and immunocytochemistry studies suggest that glutamate receptor delta subunits form novel postsynaptic receptor complexes, J. Neurosci. 15 (1995) 2533–2546. [21] Y. Miyagi, T. Yamashita, M. Fukaya, T. Sonoda, T. Okuno, K. Yamada, M. Watanabe, Y. Nagashima, I. Aoki, K. Okuda, M. Mishina, S. Kawamoto, Delphilin: a novel PDZ and formin homology domain-containing protein that synaptically colocalizes and interacts with glutamate receptor delta 2 subunit, J. Neurosci. 22 (2002) 803–814. [22] S. Okabe, A. Miwa, H. Okado, Alternative splicing of the Cterminal domain regulates cell surface expression of the NMDA receptor NR1 subunit, J. Neurosci. 19 (1999) 7781–7792. [23] K.M. Partin, D.K. Patneau, C.A. Winters, M.L. Mayer, A. Buonanno, Selective modulation of desensitization at AMPA versus kainate receptors by cyclothiazide and concanavalin A, Neuron 11 (1993) 1069–1082. [24] I. Perez-Otano, C.T. Schulteis, A. Contractor, S.A. Lipton, J.S. Trimmer, N.J. Sucher, S.F. Heinemann, Assembly with the NR1 subunit is required for surface expression of NR3A-containing NMDA receptors, J. Neurosci. 21 (2001) 1228–1237. [25] R.S. Petralia, R.J. Wenthold, Light and electron immunocytochemical localization of AMPA-selective glutamate receptors in the rat brain, J. Comp. Neurol. 318 (1992) 329–354. [26] C. Rosenmund, Y. Stern-Bach, C.F. Stevens, The tetrameric structure of a glutamate receptor channel, Science 280 (1998) 1596– 1599. [27] N.J. Sucher, S. Akbarian, C.L. Chi, C.L. Leclerc, M. Awobuluyi, D.L. Deitcher, M.K. Wu, J.P. Yuan, E.G. Jones, S.A. Lipton, Developmental and regional expression pattern of a novel NMDA receptor-like subunit (NMDAR-L) in the rodent brain, J. Neurosci. 15 (1995) 6509–6520. [28] J. Zuo, P.L. De Jager, K.A. Takahashi, W. Jiang, D.J. Linden, N. Heintz, Neurodegeneration in Lurcher mice caused by mutation in delta2 glutamate receptor gene, Nature 388 (1997) 769–773.