Glycine receptor clusters and synapses - Semantic Scholar

2783

Journal of Cell Science 113, 2783-2795 (2000) Printed in Great Britain © The Company of Biologists Limited 2000 JCS1263

Formation of glycine receptor clusters and their accumulation at synapses Jochen Meier1, Claire Meunier-Durmort2, Claude Forest2, Antoine Triller1 and Christian Vannier1,* 1Laboratoire de Biologie Cellulaire de la Synapse Normale et Pathologique, INSERM U497, Ecole Normale Supérieure, 46 rue d’Ulm, 75005 Paris, France 2CEREMOD, CNRS UPR 9078, 9 rue Jules Hetzel, 92190 Meudon, France

*Author for correspondence (e-mail: vannier@wotan.ens.fr)

Accepted 23 May; published on WWW 10 July 2000

SUMMARY The glycine receptor is highly enriched in microdomains of the postsynaptic neuronal surface apposed to glycinergic afferent endings. There is substantial evidence suggesting that the selective clustering of glycine receptor at these sites is mediated by the cytoplasmic protein gephyrin. To investigate the formation of postsynaptic glycine receptor domains, we have examined the surface insertion of epitope-tagged receptor α subunits in cultured spinal cord neurons after gene transfer by polyethylenimineadenofection. Expression studies were also carried out using the non-neuronal cell line COS-7. Immunofluorescence microscopy was performed using wild-type isoforms and an α mutant subunit bearing the gephyrin-binding motif of the β subunit. In COS-7 cells, transfected glycine receptor α subunits had a diffuse surface distribution. Following cotransfection with gephyrin, only the mutant subunit formed cell surface clusters. In contrast, in neurons all subunits were able to form cell

surface clusters after transfection. These clusters were not colocalized with detectable endogenous gephyrin, and the GlyR β subunit could not be detected in transfected cells. Therefore, exogenous receptors were not assembled as heteromeric complexes. A quantitative analysis demonstrated that newly synthesized glycine receptor progressively populated endogenous gephyrin clusters, since association of both proteins increased as a function of time after the onset of receptor synthesis. This phenomenon was accelerated when glycine receptor contained the gephyrin-binding domain. Together with previous results, these data support a two-step model for glycinergic synaptogenesis whereby the gephyrin-independent formation of cell surface clusters precedes the gephyrin-mediated postsynaptic accumulation of clusters.

INTRODUCTION

respectively (Essrich et al., 1998; Nusser et al., 1998). Diffuse GluR accumulates at synapses during the process of synaptogenesis (Mammen et al., 1997; Rao et al., 1998). Although the existence of diffuse GlyR was suggested (Kirsch et al., 1993; St John and Stephens, 1993), it could not be observed during development in spinal cord neurons (Béchade et al., 1996; Colin et al., 1996), though it has been described during early neocortical development (Flint et al., 1998). Postsynaptic receptor clustering was first studied for the muscular nicotinic acetylcholine receptor (nAChR), emphasizing the role of rapsyn (Mr 43000) (for a review, see Sanes, 1997). Cytoplasmic proteins interacting with postsynaptic receptors have also been identified in the CNS. Metabotropic glutamate, NMDA and AMPA receptors bind to Homer (Brakeman et al., 1997), PSD-95/SAP90 (Kornau et al., 1995) and GRIP (Dong et al., 1997), respectively. For GlyR, gephyrin serves a similar function (Kirsch et al., 1993), mediated through its binding to the M3-M4 intracellular loop of the GlyR β subunit (Kirsch et al., 1995; Meyer et al., 1995). Postsynaptic stabilization may involve a linkage of the receptor-gephyrin complex to subsynaptic microtubules (Kirsch et al., 1991, 1993). This may also hold for GABAAR

The inhibitory glycine receptor (GlyR) is a ligand-gated chloride channel abundantly expressed in spinal cord and brainstem (Aprison and Daly, 1978; Langosch et al., 1988). It is composed of two distinct transmembrane subunits, α1 (Mr 48000) and β (Mr 58000), arranged as an α3β2 pentamer, and of a peripheral membrane protein, gephyrin (Mr 93000) (Schmitt et al., 1987; Vannier and Triller, 1997 and references therein). GlyR was the first neurotransmitter receptor shown to accumulate at postsynaptic membrane areas apposed to inhibitory presynaptic endings (Triller et al., 1985, 1987; Altschuler et al., 1986; Todd et al., 1996; Colin et al., 1998). This notion was further extended to GABA (GABAAR; e.g. Baude et al., 1992; Todd et al., 1996) and glutamate (GluR; e.g. Jones and Baughman, 1991; Baude et al., 1993; Nusser et al., 1994) receptors. Neurotransmitter receptors are also found at non-synaptic plasma membrane locations, as documented for GABAAR (Nusser et al., 1998) and GluR (Baude et al., 1993; Nusser et al., 1994). In cerebellar granule cells, γ2 or δ subunitcontaining GABAAR are postsynaptic or extrasynaptic,

Key words: Glycine receptor, Gephyrin, Spinal cord neuron, Cell culture, Postsynaptic anchoring, Clustering, Transfection

2784 J. Meier and others (Essrich et al., 1998), perhaps through an interaction with the GABAAR-associated protein (GABA-RAP; Wang et al., 1999). It is believed that interactions between receptors and the subsynaptic cytoplasmic proteins are crucial for tethering receptors as clusters within postsynaptic scaffolds (Sanes, 1997; O’Brien et al., 1998; Kim and Huganir, 1999). Multiple interactions probably take place in these scaffolds (Kim et al., 1996): NMDA receptor is still retained postsynaptically even if the function of PSD-95 is impaired (Migaud et al., 1998). Consequently, relationships between postsynaptic localization and clustering cannot be simply determined. Gephyrin is required for the synaptic localization of GlyR; however, antisense (Kirsch et al., 1993) and gene disruption (Feng et al., 1998b) experiments did not permit discrimination between clustering and retention at synaptic loci. To investigate this issue we analyzed the fate of transfected tagged α1 and α2 subunits, and one α1 (α1-βgb) subunit bearing the minimal gephyrin-binding motif of the β subunit (Meyer et al., 1995). In COS-7 cells, coexpression studies with gephyrin showed that surface clusters formed only if the βgb motif was present. In neurons, α1 and α2 formed clusters independently of the presence of this motif and of interactions with detectable gephyrin. In the absence of gephyrin binding sites in these clusters, GlyR was able to populate postsynaptic loci. The presence of the βgb motif increased the rate of accumulation of GlyR at gephyrin-containing postsynaptic differentiations. Altogether these results indicate that cluster formation is not coupled to the gephyrin-mediated retention of receptor in synaptic loci.

MATERIALS AND METHODS Cell culture Cell lines African green monkey kidney (COS-7) cells and human embryonic kidney 293 (HEK 293) cells were plated either on glass coverslips or into 10 cm plastic dishes and grown in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal calf serum (FCS; Gibco) at 37°C and 7.5% CO2. Preparation of neuronal primary cultures of rat spinal cord Spinal cord neurons from Sprague-Dawley rats were prepared at day 14 of gestation (E14) as previously described (Béchade et al., 1996). Glass coverslips (12 mm diameter) were coated with 15 µg/ml polyDL-ornithine (Sigma) in water and then incubated with 5% heatinactivated FCS in Leibovitz medium (L15, Gibco). Routinely, neurons were plated at a density of 7.5×104 cells/cm2 in 16 mm wells. After the neurons had attached, coverslips were transferred (cell side down) to dishes containing a glial cell monolayer. They were then cultured for up to 14 days, in a 5% CO2 atmosphere at 37°C. Preparation of glial cultures of rat spinal cord Glial cell suspensions were obtained at E14. Cells were plated at a density of 4×104 cells/cm2 in 35 mm dishes (Nunc) coated with 15 µg/ml poly-DL-ornithine and grown to confluency (2-3 weeks) at 37°C and 5% CO2 in complete L15 culture medium (Henderson et al., 1995) supplemented with 10% horse serum (Gibco). Culture medium was changed after 7 and 14 days. 1 day before neuron plating, the growth medium was replaced with serum-free Neurobasal medium supplemented with B27 (Gibco) to optimize neuronal survival (Brewer et al., 1993).

Isolation of gephyrin and GlyR cDNAs Total RNAs (Chirgwin et al., 1979), prepared from adult rat spinal cord and from dissociated cells of embryonic spinal cord (E14), were used to synthesize cDNAs corresponding to gephyrin and GlyR α1 subunit, and to GlyR α2B subunit, respectively. Oligonucleotide primers were designed according to the known rat GlyR α1-sequence (5′ AACATCTTTGCACCCCCATAACTC 3′ and 5′ CCACCCCGTCCCAGAGCCTTCA 3′) and rat GlyR α2B-sequence (5′ TTGTCCTGGTCTTCTTTCTGGAATCA 3′ and 5′ GGATACATCTATTTCTTGTGGACATCT 3′) (Grenningloh et al., 1987; Kuhse et al., 1991) and used to generate cDNAs by RT-PCR. The α2B isoform will henceforth be referred to as α2 subunit. Gephyrin cDNA corresponding to the ORF of clone P1 (containing cassette C2 only; Prior et al., 1992) was amplified by RT-PCR using oligonucleotides 5′ TTCTCCCGGCTCCTGTCA 3′ and 5′ CATGCATCGAAACTTTCTCC 3′ and selected using cassette C2 specific oligonucleotides. All cDNAs were subcloned using standard methods and sequences were checked by DNA-sequencing using Thermosequenase (Amersham). DNA constructs The isolated cDNAs of the GlyR α1 and α2 subunits were modified by the insertion of the sequence encoding either the 10-amino-acid cmyc peptide EQKLISEEDL (Evan et al., 1985) using the oligonucleotides 5′ GCTGACGCTGCCCGA-GAG-CAA-AAGCTG-ATT-TCT-GAG-GAG-GAT-CTG-TCTGCACCCAAGCCT 3′ and 5′ GACCATG ACTCCAGG-GAG-CAA-AAG-CTG-ATT-TCTGAG-GAG-GAT-CTG-TCTGGAAAACATCCC 3′ for α1 and α2, respectively, or the 9-amino-acid peptide YPYDVPDYA from hemagglutinin (HA) of the influenza virus (Hamsikova et al., 1987) using the oligonucleotides 5′ GCTGACGCTGCCCGC-TAT-CCCTAT-GAC-GTG-CCC-GAC-TAT-GCT-TCTGCACCCAAGCCT 3′ and 5′ CAAAGACCATGACTCCAGG-TAT-CCC-TAT-GAC-GTGCCC-GAC-TAT-GCT-TCTGGAAAACATCCCTCG 3′ for α1 and α2, respectively. The epitopes were positioned by site-directed mutagenesis between the second and third amino acids, and between the sixth and seventh amino acids, of the mature protein α1 and α2, respectively (Kunkel, 1985). A chimeric α1 subunit (α1-βgb) bearing the gephyrin-binding motif of the GlyR β subunit (Meyer et al., 1995) was constructed by site-directed mutagenesis. The 18-residue gephyrin recognition sequence was inserted into the c-myc- and HAtagged α1-ORF between amino acids 363 and 364 (Grenningloh et al., 1990b) using the oligonucleotide 5′ CCAACAACAACAACACCACGA-GAT-CAA-ATG-ATT-TCA-GCA-TTG-TAG-GCA-GCTTAC-CAA-GAG-ATT-TTG-AAT-TAT-CCA-ACCCCGCTCCTGCAC 3′, by site-directed mutagenesis. The green fluorescent protein (GFP, obtained from the pEGFP-N1 plasmid; Clontech) was fused to the gephyrin C terminus with a GGS spacer sequence using standard recombination methods. cDNA of all epitope-tagged (HA and/or cmyc, respectively) forms of the α1, α1βgb and α2 subunits, including the gephyrin-GFP chimera, were subcloned into a eukaryotic expression vector derived from pEGFP-N1 (Clontech) with a cytomegalovirus (CMV) promotor. The primary structure of the various constructs was verified by DNA sequencing. Transient transfection protocols All plasmid DNAs were prepared by anion-exchange chromatography (Qiagen resin). PEI-adenofection of neurons Neurons were transfected 8 days after plating. To this end, they were transferred to 4-well plates containing 300 µl of fresh serum-free Neurobasal medium, supplemented with 0.25 mM L-glutamine (Gibco) and equilibrated at 37°C and 7.5% CO2. Transfection was performed by polyethylenimine (PEI)-mediated DNA transfer as described previously for other differentiated cells (Meunier-Durmort et al., 1997). Briefly, complexes of PEI/DNA were formed by mixing PEI (800 kDa, Fluka) and plasmid in 0.15 M NaCl at a molar charge

Glycine receptor clusters and synapses 2785 ratio r +/−=3 (assuming that every sixth nitrogen in PEI is positive at pH 7.3). After 10 minutes, adenovirus (replication-deficient, Ad-RSVnlsLacZ) was added to the complex, and the ternary complex (10 µl) was diluted in the medium of the wells. Typically, neurons were treated using 200 ng of DNA combined with 150 plaque-forming units (pfu) of Ad-RSV-nlsLacZ per cell. Contact was allowed for 2 hours, then the neurons were returned to the glial cell monolayer for exogene expression for times ranging from 4 to 24 hours. Transfection of cell lines For COS-7 cells, experiments were performed on subconfluent cultures (60-70% confluency) using the DEAE-Dextran method. HEK 293 cells used for immunoblotting and electrophysiological analyses were grown to 80% confluency and transfected using the calcium phosphate method (Chen and Okayama, 1987). Usually, 2 or 10 µg of plasmid DNA were added to 35 or 100 mm dishes, respectively. Transient protein expression was allowed to proceed for 24-48 hours at 37°C and 7.5% CO2. Antibodies The following antibodies were used: mouse anti-gephyrin monoclonal antibody (mAb7a) at a dilution 1/100 (Pfeiffer et al., 1984; Boehringer Mannheim); anti-c-myc antibody (clone 9E10) at a concentration of 10 µg/ml (Boehringer Mannheim). The HA epitope was detected with a rat monoclonal antibody (clone 3F10) at a concentration of 10 µg/ml (Boehringer Mannheim). MAP2 protein was detected with a 1/300 dilution of a specific rabbit polyclonal antibody (Sigma). Secondary antibodies were used at a dilution of 1/200: carboxymethylindocyanine-3 (Cy3)-conjugated affinity-purified goat anti-mouse IgG (depleted in anti-rat IgG activity), fluorescein (FITC)conjugated affinity-purified goat anti-rat IgG (depleted in anti-mouse IgG activity) and (Cy3)-conjugated affinity-purified goat anti-rabbit IgG were from Jackson Laboratories. Immunofluorescence Immunofluorescence was performed as described previously (Lévi et al., 1998). Cells were fixed in 4% (w/v) paraformaldehyde in phosphate-buffered saline (PBS) for 15 minutes, then washed with PBS and permeabilized with PBS containing 0.12% (w/v) Triton X100 and 0.12% (w/v) gelatin for 4 minutes. Cells were then incubated (1 hour, room temperature) with primary antibodies in PBS containing 0.12% (w/v) gelatin. After extensive washes with the same buffer, secondary antibodies were reacted for 45 minutes. Detection of surface antigens was performed on unfixed cells. This approach allows identification of external epitopes only (Misek et al., 1984; Pfeiffer et al., 1985). For this, cells were incubated directly with primary antibodies for 30 minutes on ice, either in PBS containing 2 mg/ml bovine serum albumin (BSA), 1 mM Ca2+ and 1 mM Mg2+ (COS-7 cells) or in air-equilibrated L15 medium supplemented with 20 mM glucose and 1 mg/ml BSA (neurons). After extensive washes on ice, cells were fixed in paraformaldehyde as above, then reacted with secondary antibodies. Immunoblotting Transfected HEK 293 cells were washed with ice-cold PBS containing 2 mM EDTA, scraped and directly solubilized in non-reducing Laemmli sample buffer. Proteins (25 µg) were separated by SDSpolyacrylamide gel electrophoresis (SDS-PAGE, 10% acrylamide), and transferred onto nitrocellulose (BA85 Schleicher and Schuell, 0.45 µm) using 20% methanol, 0.1% SDS, 20 mM Tris-base, 150 mM glycine, pH 8.3. Nitrocellulose membranes were blocked by incubating in 10% (w/v) low-fat milk powder in PBS for 1-2 hours at room temperature (RT), then incubated overnight with mAb 9E10 (1.5 µg/ml; Boehringer Mannheim) at 4°C, and with a rabbit polyclonal anti-GFP antibody (1.5 µg/ml; Clontech) in PBS containing 1 mg/ml BSA and 0.3% (w/v) Tween 20. Blots were washed with the same buffer and reacted with the appropriate horseradish peroxidase (HRP)-

conjugated goat secondary antibody. They were finally processed for enhanced chemiluminescence (ECL, Amersham). Quantitative analysis The colocalization index in transfected cells was determined by visual analysis using a standard fluorescence microscope (Leica; objective ×63). Double-labeled cells (FITC/Cy3) were digitized and superimposed using image display (Molecular Dynamics) software. For determination of colocalization, spots were counted manually for each cell. Values are expressed as means ± s.e.m. of 12 independent cells per receptor subunit and time point. A mask corresponding to gephyrin-immunoreactive spots was created. It allowed automatic discrimination between gephyrin-associated and gephyrin-negative GlyR clusters. The threshold intensity for detection and measurements was set manually at the limit in order to avoid coalescence of spots. The surface area of peripheral clusters was determined using NIH 1.52 software on the above cells. Both channels were analyzed separately. The statistical analysis of the results was carried out with StatView F.4.11 software. Confocal microscopy Confocal analysis were carried on a Leica microscope equipped with an Argon Kripton laser. Excitation of the FITC and Cy3 fluorochromes was performed at 488 and 568 nm, respectively, and the intensities were set to abolish cross-excitation. Appropriate filters allowed the detection of FITC and Cy3 fluorescence without any crosstalk. Data were acquired with an ×100 (n.a. 1.4) objective. The digitized planes were transformed using a 3×3 Gaussian filter, and superimposed using false colors.

RESULTS Construction and expression in non-neuronal cells of tagged GlyR subunits cDNAs encoding the full-length subunits of GlyR, namely α1 and α2, and of gephyrin, were obtained as described in Materials and Methods. They were modified by adding the coding sequence of either HA or c-myc epitope-tags and Aequorea victoria GFP, respectively (Fig. 1A). A chimeric α1 subunit (α1-βgb) was also created by inserting the 18-aminoacid sequence from the β subunit (gb) responsible for gephyrin binding in the cytoplasmic loop (Meyer et al., 1995) (Fig. 1A). All modified cDNAs were subcloned into an expression vector allowing a constitutive expression driven by the CMV promotor. The sizes of the corresponding proteins were checked by transfection of HEK 293 cells followed by SDS-PAGE and immunoblotting performed on whole cell lysates (Fig. 1B). As expected, each of the various α constructs gave rise to a single species of Mr 48,000-49,000, detected after reaction with the anti-c-myc antibody. The gephyrin-GFP fusion protein, reacted with a GFP-specific antibody, appeared as a homogenous species of Mr 105,000. From the known topology of the GlyR subunits (Betz, 1991), it was anticipated that positioning epitope tags at or close to the N terminus of the polypeptide would lead to exclusive detection at the cell surface in non-permeabilized, intact living cells. This approach (see Materials and Methods) proved successful with both non-neuronal cells and neurons. The behavior of the tagged proteins was analyzed in transfected COS-7 cells using immunofluorescence microscopy. Results obtained with c-myc-tagged subunits are reported in Fig. 2. They were identical to those obtained with HA-tagged subunits

2786 J. Meier and others

Fig. 1. Structure and expression of constructs used in this study. (A) Diagram of the structures of the tagged α1 and α2 subunits and of gephyrin. In α1, α2 and α1-βgb sequences, the black box is either the c-myc or the HA sequence inserted downstream of the indicated N-terminal amino acid residues, and the hatched boxes represent the four transmembrane domains (M1-M4). In the α1-βgb molecule, the black hatched box between M3 and M4 is the gephyrin-binding site from the GlyR β subunit inserted after residue 363. The gephyrin-GFP chimera contains GFP (black bar) fused to the tripeptide spacer GGS, added after residue 736 of gephyrin. (B) Immunoblotting analysis of transfected HEK 293 cells. Cells transfected with gephyrin-GFP (lane 1) or myc-tagged subunits α1 (lane 2), α2 (lane 3) or α1-βgb (lane 4), were lysed 24 hours after transfection. Following SDS-PAGE and electroblotting (see Materials and Methods), transferred proteins were probed with anti-GFP-antibody and anti-c-myc-antibody and revealed using ECL.

(not shown). 24 hours after transfection, the α1, α2 and α1-βgb subunits were expressed at high levels, as judged from the bright perinuclear staining seen in permeabilized cells reacted with the anti-c-myc antibody (Fig. 2A2-C2). This pattern was distinct from that obtained following exposure of living cells to the anti-c-myc antibody at 0°C (Fig. 2A1-C1). In this case, the diffuse labeling of the plasma membrane demonstrated both proper transport of the modified subunits to the cell surface and their acquisition of the expected topology. In agreement with this, patch-clamp analysis in the whole-cell configuration (Hamill et al., 1981) on HEK 293 cells expressing tagged α1 or α1-βgb, revealed that glycine (40 µM, bath application) elicited chloride currents characterized (not shown) by parameters similar to those already reported for the transfected subunit (Bormann et al., 1987; Sontheimer et al., 1989; Grenningloh et al., 1990a). These results indicated that subunit activity was not grossly altered by the tag sequence. A feature of gephyrin, when transfected in HEK 293 cells, was the formation of cytoplasmic aggregates, probably reflecting concentration-dependent homophilic interactions, as observed by others (Kirsch et al., 1995). These aggregates were also observed in transfected COS-7 cells, and do not seem to perturb surface insertion of either α1 or α2 upon cotransfection

with gephyrin-GFP (Fig. 2D,E). The distribution pattern of gephyrin in cells coexpressing the chimeric α1-βgb subunit was modified (Fig. 2F,G). In these experiments gephyrin still formed numerous intracellular aggregates of heterogeneous size, and part of the gephyrin pool (and only under these conditions, compare with Fig. 2D,E) was found under the plasma membrane. There, gephyrin formed clusters and was colocalized with α1-βgb inserted in the plasma membrane (93±0.55% of peripheral gephyrin aggregates were colocalized with cell surface GlyR clusters). The colocalization of surface clusters of GlyR α1 chimera with clusters of gephyrin is illustrated in Fig. 2F3, arrowheads. This result indicated that the gephyrin-binding domain of the β subunit was active when inserted in a heterologous polypeptide and could mediate the interaction of the two transfected proteins. This result also meant that relocation of gephyrin beneath the membrane resulted from an interaction with clustered α1-βgb subunit. Moreover, the colocalization of gephyrin and α1-βgb was not restricted to the plasma membrane since gephyrin interaction also occurred with the intracellular chimera (Fig. 2G). Altogether, these experiments indicated that: (1) the GlyR α subunits alone are not able to form clusters, (2) gephyrin alone is not able to translocate beneath the plasma membrane, (3) binding of gephyrin to a transmembrane subunit can relocate gephyrin and, finally, (4) gephyrin promotes the formation of GlyR α subunit clusters.

Clusters of exogenous subunits over the somatodendritic compartment In order to obtain a short-term expression of exogenous GlyR components in differentiated neurons, primary cultures of spinal cord cells were transfected using a ternary complex of polyethylenimine/DNA/adenovirus, as described in Materials and Methods. This method, in which LipofectAMINE can also be used in place of polyethylenimine, was initially utilized for transient expression in established cell lines (Meunier-Durmort et al., 1996, 1997). It was preferred in the present study because the low toxicity of the transfection medium ensured better neuron survival (for cells taken 7-15 days after plating) than protocols involving other polycation-mediated condensations. This was assessed by estimating morphological alterations and fragmentation of nuclei (unpublished observations). The maximum efficiency of gene transfer (up to 5% of neuronal cells were transfected) was achieved at a charge ratio of 2.5-3 with PEI 800 kDa (see Materials and Methods), whereas no transfer could be obtained with PEI 25 kDa (data not shown). It was important to determine whether plasmid DNA transferred into spinal neurons using this protocol allowed a proper expression and localization of encoded GlyR subunits. As determined by immunocytochemistry of the HA-tag on non-permeabilized living cells, all the HA-α1, HA-α2 and HA-α1-βgb subunits could be detected at the cell surface. Tags could be detected as early as 4 hours following transfection. The exclusive labeling of the exogenous subunits present at the cell surface with mAb3F10 (anti-HA) was confirmed by the membrane impermeability toward antibodies, as ascertained

Glycine receptor clusters and synapses 2787

Fig. 2. Surface expression of myc-tagged GlyR subunits in COS-7 cells. (A-C) Cells were transfected with α1, α2 or α1-βgb. After a 24 hour expression period the c-myc epitope was revealed on intact living cells (no permeabilization; A1-C1) or on fixed and permeabilized cells (A2-C2). A1,2, α1; B1,2, α2; C1,2, α1-βgb. In the absence of permeabilization, myc-tagged proteins are diffusely distributed at the cell surface (arrowheads). After fixation and permeabilization tags are also detected intracellularly (arrows). (D-G) Coexpression of gephyrin-GFP chimera and myc-tagged subunits. Localization of proteins was analyzed 24 hours after transfection by myc-immunostaining and GFP-fluorescence on living non-permeabilized (D1-F3), and on fixed and permeabilized (G1-3) cells. D1-3, gephyrin and α1; E1-3, gephyrin and α2; F1-G3, gephyrin and α1-βgb. The c-myc epitope was visualized using mAb 9E10 and a Cy3-conjugated secondary antibody (red, arrowheads; D1-G1), and gephyrin by GFP fluorescence (green, arrows; D2-G2). Higher magnifications of superimposed images are shown in D3-G3. Note the presence of clusters of both gephyrin and GlyR α1-βgb subunit at the plasma membrane (F3, arrowheads) and in intracellular compartments (G3, arrows) in coexpression experiments. Bars, 10 µm (5.4 µm in D3-G3).

2788 J. Meier and others

Fig. 3. Somatodendritic surface expression of tagged GlyR subunits in cultured rat spinal neurons. Neurons were transfected with α1, α2 or α1-βgb, 8 days after plating. After an 8 hour expression period, cell surface HA-immunofluorescence (green) was visualized on living nonpermeabilized cells. After permeabilization, somatodendritic compartments were identified using anti-MAP2 antibody (red). Note the discontinuous labeling at the cell surface for all constructs. (A) HA-GlyR α1; (B) HA-GlyR α2; (C) HA-GlyR α1-βgb. Lower panels: higher magnifications of the area delineated in the upper panels. Bars, 10 µm (upper panels) and 2.7 µm (lower panels).

by the lack of reactivity of the anti-MAP2 antibody upon coincubation with mAb3F10 (not shown). The immunoreactivity of these subunits was not randomly localized at the cell surface. First, as illustrated in Fig. 3, the anti-HA antibody stained only the dendritic and somatic membranes and not the axons, since HA- and MAP2-immunoreactivities were found in the same processes (Fig. 3), and HA labeling was absent from compartments devoid of MAP2 (not shown). In the illustrated experiment, the HA-tag was labelled prior to permeabilization on unfixed living cells (see Materials and Methods) whereas MAP2 was labelled following permeabilization. Immunoreactivity of all three α subunit forms was restricted to the MAP2 compartment over a 24-hour period after the onset of expression (not shown). This is consistent with the known distribution of endogenous GlyR in the somatodendritic compartment (Béchade et al., 1996). Second, once detectable, the immunoreactivity of cell surface GlyR subunits appeared as spots of variable size scattered over the somatodendritic domain. The formation of these spots could result from the oligomerization of exogenous α subunits with endogenous β subunit, thus allowing tagged heteromeric GlyR to interact with endogenous gephyrin. To test the possibility of the presence of a gephyrin-binding site, we first compared the localization of exogenous α1 or α1-βgb in neurons cotransfected with GFP-tagged gephyrin. If oligomerization with endogenous β subunit occurred, both exogenous α1 and α1-βgb should be colocalized with gephyrin-GFP. The gephyrin-binding domain is a potent signal since it can relocate NMDAR to intracellular gephyrin-rich loci when incorporated in the C terminus of the NR1 subunit (Kins et al., 1999). Others used similar redistribution experiments in non-neuronal cell systems (Kirsch et al., 1995) to detect interactions with gephyrin. Therefore, a lack of colocalization of α1 subunit to gephyrin intracellular blobs would indicate that GlyR has not incorporated the β subunit. As shown in Fig. 4A,B, we found that myc-tagged α1 subunit and gephyrin-GFP exhibit very low colocalization at the cell surface and intracellularly. This indicates that almost all exogenous α1 subunits synthesized in neurons transfected 7 days after plating did not oligomerize with the β subunit. If any oligomerization were to occur with an endogenous GlyR β subunit, then the stoichiometry of the gephyrin-binding site to α1 subunit would be far below the 1:1

ratio present in the α1-βgb chimera. In accordance with this, when the βgb sequence was present, the α1-βgb and gephyrin strongly colocalized at both intracellular compartments (Fig. 4C) and cell surface (Fig. 4D). Using western blotting analysis we also ascertained that transient expression of either α1 or α2 in neurons did not induce β subunit synthesis. We used the 4a antibody, which recognizes GlyR α and β subunits, and allows their simultaneous detection in purified spinal cord heterooligomeric receptor (Pfeiffer et al., 1984). In cells transfected at 2 and 7 days after plating (Fig. 4E), the expression of exogenous α1 or α2 subunits was not accompanied by the appearance of the β subunit (Fig. 4E, lanes 1,2 and 5,6). As expected, β subunit was not synthesized in cells expressing GFP alone or in non-transfected cells (Fig. 4E, lanes 3,4 and 7,8), in agreement with published data (Hoch et al., 1989) on similar cell cultures. Our results on the expression of tagged receptors demonstrate that: (1) cell surface exogenous GlyR α subunits are localized over the somato-dendritic compartment, (2) GlyR α1 subunits form clusters that do not contain gephyrin-binding sites and (3) GlyR α subunits are not oligomerized with an endogenous GlyR β subunit. Relationship of exogenously expressed GlyR α subunits to postsynaptic loci The clusters of exogenous GlyR α subunits at the cell periphery are reminiscent of those previously reported for endogenous receptor, indicating that exogenous subunits can clusterize in the plasma membrane of transfected neurons. The postsynaptic localization of clusters of endogenous GlyR in cultured neurons is also well established (Béchade et al., 1996; Nicola et al., 1997). We therefore further characterized the distribution pattern of the exogenous subunits and determined whether the new α1, α2 and α1-βgb clusters could be associated with a known postsynaptic marker. Double immunostaining experiments were performed in order to compare the respective distributions of HA-tagged subunits and of gephyrin (Fig. 5A-C). For all three subunits, surface clusters of exogenous origin, specifically labeled with mAb3F10, were colocalized with a large fraction of the pool of endogenous gephyrin clusters visualized with mAb7a (Fig. 5A-C). However, the extent of association of gephyrin clusters with exogenous GlyR, as determined after 24-hours

Glycine receptor clusters and synapses 2789

Fig. 4. Absence of detectable β subunit in exogenously expressed GlyR in neurons after transfection. (A-D) Localization of myc-tagged GlyR α subunits and GFP-tagged gephyrin in cultured rat spinal neurons. Experiments were carried out on neurons as in Fig. 3. The tagged GlyR α subunits (red) and the exogenous gephyrin (green) are revealed as in Fig. 2. Cotransfection of GlyR α1 subunit and gephyrin (A,B) or GlyR α1-βgb and gephyrin (C,D), followed by immunostaining with (A,C) or without (B,D) permeabilization, respectively. Note that colocalization (yellow) occurs intracellularly (crossed arrows) or at the cell surface (open arrowheads) only if the βgb sequence is present. Open arrowheads and arrows indicate surface expressed tagged GlyR α1-βgb and tagged GlyR α1, colocalized or not with gephyrin-GFP, respectively; crossed and tailed arrows, intracellular tagged GlyR α1-βgb and tagged GlyR α1, colocalized or not with gephyrin-GFP, respectively; arrows, intracellular gephyrin-GFP. Bar, 5 µm. (E) Analysis of lysates of neurons transfected 2 or 7 days after plating. Cells transfected with myctagged α1 (lanes 1,5), α2 (lanes 2,6) subunits or GFP (lanes 3,7), and non- transfected cells (lanes 4,8) were homogenized 24 hours after transfection and a post-nuclear fraction was prepared. 10 µg protein fractions were used for SDS-PAGE and electroblotting. Transferred material was probed with the monoclonal anti-α/β antibody (mAb4a) and revealed using ECL. The arrowhead indicates the position expected for the β subunit (Mr 58,000).

post-transfection, depended on the transfected subunits (α1, 72%±4.51; α2, 54%±4.1; α1-βgb, 86%±0.86 (means ± s.e.m.); see Fig. 8). As shown in Fig. 5, tagged-GlyR α1 partially associates with endogenous gephyrin, whereas it was rarely associated with intracellular aggregates of exogenous gephyrin-GFP (see Fig. 4). Various explanations may account for this apparent discrepancy. First, the GFP-tagged gephyrin used in this experiment (cassette 2 only) could be different from the endogenous gephyrin recruiting GlyR at synapses (Fig. 5). Second, this difference could result from routing. Tagged-GlyR α1 would be delivered to the plasma membrane prior to association with pre-existing subsynaptic clusters of endogenous gephyrin. In cotransfected cells, it would escape trapping by exogenous gephyrin aggregates because of its low affinity for gephyrin, and therefore reach the cell surface. As a result, only associations between tagged-GlyR α1 and endogenous gephyrin would be detected. These transfection and colocalization experiments indicate that: (1) tagged GlyR α subunits are able to form clusters associated or not with postsynaptic loci, (2) in neurons only, clusters can form independently of interaction with gephyrin,

and (3) the level of association with postsynaptic loci depends on the transfected subunit. Tagged GlyR α subunit clusters can be independent of gephyrin in transfected neurons Clusters formed by the three tagged α subunits were classified into two groups, according to whether HA-epitope and endogenous gephyrin immunoreactivities were colocalized. As illustrated by the immunostaining of the two antigens on dendrites after 8 hours of expression (Fig. 6), the tagged GlyR clusters, associated or not with gephyrin, displayed size heterogeneity over the entire surface of the soma and dendrites. Quantitative analysis revealed that gephyrin-associated clusters of α1 and α2 subunits were significantly larger than gephyrinnegative ones, independent of the expression time (4-24 hours) following transfection (Fig. 7). The size distribution of α1 clusters 8 hours post-transfection is shown in Fig. 7A1,2. The colocalization with gephyrin aggregates was associated with an increase in the mean surface area, and the fraction of clusters smaller than 0.1 µm2 was 75% in the gephyrin-negative population, as compared to 38% in the positive one. Size differences were also observed for α1 and α2 at various time-

2790 J. Meier and others

Fig. 5. Relationship of tagged GlyR subunits with synaptic components in transfected neurons. Experiments were carried out as in Figs 3 and 4 on neurons transfected 8 days after plating with HA-tagged α1, α2 or α1-βgb. (A-C) HA-tagged GlyR subunits and gephyrin. HAimmunoreactivity (IR) (A1-C1), 24 hours after transfection, visualized on intact living cells prior to fixation (A1, α1; B1, α2; C1, α1-βgb). Gephyrin was immunostained following fixation and permeabilization (A2-C2). Higher magnifications (A3-C3) of superimposed images corresponding to outlined regions. Arrows, HA-associated IR; crossed arrows, gephyrin IR; arrowheads, examples of colocalization. Bar, 10 µm (5.8 µm in A3-C3).

points after transfection, as compared to α1-βgb (Fig. 7B). Independent of the expression time (4, 8 or 24 hours), α1 and α2 clusters associated with gephyrin were significantly (P