In this study, we present evidence that a glycine transporter, GLYT1, is expressed in neurons and that it is associated with glutamatergic synapses. Despite the presence of GLYT1 mRNA in both glial cells and in glutamatergic neurons, previous studies have mainly localized GLYT1 immunoreactivity to glial cells in the caudal regions of the nervous system. However, using novel sequence specific antibodies, we have identified GLYT1 not only in glia, but also in neurons. The immunostaining of neuronal elements could best be appreciated in forebrain areas such as the neocortex or the hippocampus, and it was found in fibers, terminal boutons and in some dendrites. Double labeling confocal microscopy with the glutamatergic marker vGLUT1 revealed an enrichment of GLYT1 in a subpopulation of glutamatergic terminals. Moreover, through electron microscopy, we observed an enrichment of GLYT1 in both the presynaptic and the postsynaptic aspects of putative glutamatergic terminals that established asymmetric synapses. In addition, we demonstrated that GLYT1 was physically associated with the NMDA receptor in a biochemical assay. In conclusion, the close spatial association of GLYT1 and glutamatergic synapses strongly supports a role for this protein in neurotransmission mediated by NMDA receptors in the forebrain, and perhaps in other regions of the CNS.
Beside its well-characterized role as an inhibitory neurotransmitter, glycine is a co-agonist of NMDA receptors that is necessary both for ion channel opening and probably for the internalization of the receptor from the cell surface (Kemp and Leeson, 1993; Nong et al., 2003). While it was initially believed that the concentration of glycine in the synaptic cleft would be high enough to saturate the glycine site on the NMDA receptor, recent pharmacological and electrophysiological evidence indicates that this might be not the case. Indeed, the levels of glycine in the synaptic cleft could be close to the threshold levels for activation of the receptor (Supplisson and Bergman, 1997; Berger et al., 1998; Bergeron et al., 1998; Kew et al., 1998, 2000). The concentration of glycine in the synaptic cleft is regulated by the glycine transporters GLYT1 and GLYT2 (for a review see López-Corcuera et al., 2001). GLYT2 is predominantly found in the spinal cord and the brainstem, associated with glycinergic neurotransmission (Luque et al., 1995; Zafra et al., 1995a,b; Jursky and Nelson, 1996). However, while GLYT1 is also expressed highly in glycinergic areas, GLYT1 mRNA was also found in areas devoid of inhibitory glycinergic neurotransmission where it might play a role in NMDA-mediated neurotransmission (Smith et al., 1992). In support of this hypothesis, N[3-(4′-fluorophenyl)-3-(4′-phenylphenoxy)propyl]sarcosine (NFPS), a specific inhibitor of GLYT1, potentiates the responses to NMDA both in vitro and in vivo (Chen et al., 2003; Kinney et al., 2003). Recent evidence obtained in knockout mice for glycine transporter genes confirmed the involvement of both GLYT1 and GLYT2 in glycinergic inhibitory neurotransmission. In these mutant mice, glycinergic neurotransmission was largely altered, leading to the suggestion that GLYT1 might remove glycine from the synaptic cleft, whereas GLYT2 would replenish the presynaptic pool of glycine (Gomeza et al., 2003a,b). The possible role of GLYT1 in glutamatergic neurotransmission in these GLYT1 deficient mice was not addressed since they died shortly after birth. However, heterozygous animals that expressed only 50% of GLYT1 had an enhanced hippocampal NMDA receptor function and memory retention. These animals were also protected against an amphetamine disruption of sensory gating, suggesting that drugs which inhibit GLYT1 might have both cognitive enhancing and antipsychotic effects (Tsai et al., 2004).
In order to fully clarify the role of GLYT1 in glutamatergic neurotransmission, it is necessary to determine the precise cellular and subcellular localization of this transporter. Previous histochemical studies revealed a discrepancy between the distribution of GLYT1 mRNA and protein. While GLYT1 mRNA transcripts were detected in glia and many types of neuron throughout the nervous system, the protein was only detected in glial cells (Zafra et al., 1995a,b).
Here, we present detailed immunohistochemical data showing that, in addition to the previously described expression of GLYT1 in glia, this transporter is also expressed at the pre- and postsynaptic aspects of glutamatergic synapses. This association was especially evident in the neocortex and hippocampus, areas where GLYT1 formed immunoprecipitable complexes with NMDA receptors. These results indicate a close relationship between GLYT1 and glutamatergic neurotransmission.
Material and Methods
Oligonucleotides were synthesized by Isogen (Utrecht, NL). Bovine serum albumin (BSA), ethylene glycol-bis-aminoethyl ether N,N,N′,N′-tetraacetic acid, N-2-hydroxyethylpiperazine-N′-2-ethane sulfonic acid (HEPES), ovalbumin, diaminobenzidine, phenylmethanesulfonyl fluoride (PMSF), Triton X-100, Trizma base, Freund's adjuvant and glutathione were from Sigma (St Louis, MO). Dextran-70, pGEX2, Glutathione Sepharose 4B, standard proteins for sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) (Rainbow markers), biotinylated donkey anti-rabbit immunoglobulin Gs (IgGs), streptavidin-biotinylated horseradish peroxidase and the enhanced chemioluminescent detection method (ECL) were from Amersham-Pharmacia (Amersham, UK). Isopropylthiogalactoside, EcoRI, BamHI and T4 ligase were from Roche (Germany). Lipofectamine-PLUS was from Invitrogen and Taq polymerase from Perkin Elmer Cetus. Affi-Gel 15 (N-hydroxysuccinimide esters of crosslinked agarose) and nitrocellulose sheets were from BioRad (Richmond, CA). Newborn calf serum was from Gibco (Paisley, UK). Guinea pig anti-vesicular glutamate transporters 1 and 2 (anti-vGLUT1 and anti-vGLUT2) were from Chemicon (Temecula, CA) and mouse anti-NMDA from Phamingen. Gold-coupled antibodies were from Auroprobes. Goat anti-rabbit, goat anti-guinea pig and goat anti-mouse coupled to Alexa Fluor 488 or Alexa Fluor 594 were from Molecular Probes (Eugene, OR). TAAB 812 was from TAAB laboratories (Reading, UK) and Lowicryl HM20 from Agar Scientific Ltd (Stanstead, UK). All other reagents were analytical grade.
Preparation of Glutathione S-transferase–Glycine Transporter Fusion Proteins
The 111 bp cDNA fragment encoding the amino terminus of GLYT1b was synthesized by polymerase chain reaction using the full-length clone as template. According to available experimental evidence, the N-terminal domain of GLYT1 is intracellularly placed (Olivares et al., 1997). The specific oligonucleotide primers were CGGGATCCATGGCTGTGGCTCACGGA and GGAATTCACTCGATCTGGTTGCCCCA. Each primer consisted of 18 nucleotides of the GLYT1b sequence plus the restriction sites either for BamHI or EcoRI in their 5′ ends. The BamHI site was placed such that the inserts were in the correct reading frame after insertion in BamHI/EcoRI sites of the pGEX-2 plasmid. These restriction sites are located in the carboxyl terminus of the glutathione S-transferase (GST). After ligation of the inserts to pGEX-2, Escherichia coli XL-1 blue transformants were selected in LB (10 g/l bacto-tryptone, 5 g/l bacto-yeast extract, 10 g/l NaCl, pH 7.0)–agar plates containing 50 μg/ml of ampicillin. Individual colonies containing the pGEX-GLYT1b-NT were identified and sequenced using standard procedures. Fusion protein expression was induced by isopropylthiogalactoside (IPTG) as follows: bacteria were grown overnight at 37°C in 50 ml of LB medium containing 50 μg/ml of ampicillin. Afterwards, bacteria were diluted with 200 ml of the same growth medium and incubated at 37°C for 1 h. Induction was carried out for 5 h by adding IPTG (1 mM final concentration). The induced bacteria were collected by centrifugation resuspended in phosphate-buffered saline (PBS) medium (137 mM NaCl, 0.9 mM CaCl2, 2.68 mM KCl, 1.47 mM KH2PO4, 0.49 mM MgCl2, 7.37 mM Na2HPO4, pH 7.4) containing 10 mM EDTA, 0.2 mg/ml lysozyme and 1 mM PMSF, and lysed by sonication. The lysate was cleared by centrifugation and the supernatant was passed through a Glutathione Sepharose 4B column. The fusion protein was eluted with 10 mM glutathione in 50 mM Tris–HCl, pH 7.5. A BLAST search in non-redundant and EST sequences using as bait the N-terminal GLYT1 sequence revealed the absence of any significative homology with other proteins in sequence data banks, except for GLYT1 from different species.
Immunization was carried out in MDL New Zealand rabbits (Navarra, Spain). The aforementioned fusion protein (200 μg) was mixed with 0.8 ml of water, emulsified with 1 ml Freund's complete adjuvant and injected into two rabbits. The rabbits were reimmunized every second week as above, but using Freund's incomplete adjuvant.
Immobilization of the Fusion Protein
Fusion protein (1 mg) was dissolved in Na-HEPES buffer, pH 8.0, and coupled to Affi-Gel 15 (5 ml gel) following the manufacturer's instructions. A 10 mg aliquot of GST was immobilized (as above) on 25 ml Affi-Gel 15.
The purification of antibodies was performed as described (Zafra et al., 1995a). The anti-fusion protein antisera were absorbed with immobilized GST, concentrated on a protein A–Sepharose column and finally affinity purified on a column with immobilized fusion protein. BSA (1 mg/ml) and NaN3 (1 mg/ml) were added to stabilize the antibodies.
Preparation of SDS Extracts
SDS extracts from whole rat brains or specific regions were prepared by homogenizing and solubilizing the tissue in PBS with SDS (10 mg/ml) and 1 mM PMSF, and removing unsolubilized material by centrifugation (50 000 g, 10 min, 10°C).
Electrophoresis and Blotting
SDS–PAGE was done in the presence of 2-mercaptoethanol. The gels were run slowly (overnight) at constant current starting at 30 V. After electrophoresis, samples were transferred by electroblotting onto a nitrocellulose membrane in a semidry electroblotting system (LKB) at 1.2 mA/cm2 for 2 h. The transfer buffer consisted of 192 mM glycine and 25 mM Tris–HCl, pH 8.3. Nonspecific protein binding to the blot was blocked by the incubation of the filter with 5% non-fat milk protein in 10 mM Tris–HCl, pH 7.5, 150 mM NaCl for 4 h at 25°C. The blot was then probed with the indicated dilutions of crude antisera or purified antibodies overnight at 4°C. After washing, blots were then probed with an anti-rabbit IgG peroxidase linked, and bands were visualized with ECL and quantified by densitometry (Molecular Dynamics Image Quant v. 3.0).
Five adult rats (3 months old) (Wistar strain) were deeply anesthetized with pentobarbital (100 mg/kg) and perfused through the left ventricle–aorta with a fixative solution containing formaldehyde (4 or 0.5%) in 0.1 M sodium phosphate pH 7.4 (500 ml, room temperature), preceded by a brief flush (10–15 s) of Dextran-70 (MW 70000; 20 mg/ml) in the same buffer without fixative. The right atrium was cut open at the start of the perfusion. The perfusion liquids were delivered by a peristaltic pump at 50 ml/min, starting within 20 s after thoracotomy. The brains were postfixed (overnight) in the same fixative. The tissue was stored at 4°C in a storage solution consisting of 1 part fixative and 9 parts 0.1 M sodium phosphate until it was processed for light and electron microscopic immunocytochemistry or cryoprotected with sucrose (30%). The formaldehyde was freshly depolymerized from paraformaldehyde (PFA). Vibratome or cryostat sections (40 μm thick) were cut and stored (4°C, up to 3 weeks) in 0.1 M sodium phosphate with NaN3 (0.2–1 mg/ml). Then the sections were rinsed in 0.1 M sodium phosphate, incubated (30 min) in 1 M ethanolamine with 0.1 M sodium phosphate, washed (3 × 1 min) in buffer A (0.135 M NaCl, 0.01 M sodium phosphate), incubated in buffer B (0.3 M NaCl, 0.1 M Tris–HCl, pH 7.4) with 10% (v/v) newborn calf serum and NaN3 (1 mg/ml), and then incubated (12–48 h, 4°C at room temperature) with antibodies (5–10 μg/ml) diluted in buffer C (buffer B with 1% newborn calf serum). The sections were washed (3 × 1 min and 2 × 10–20 min) in buffer C, incubated (1 h) with biotinylated donkey anti-rabbit Ig (1:100) in buffer C, washed (3 × 1 min and 2 × 15 min) in buffer C, incubated (1 h) with streptavidin-biotinylated horseradish peroxidase complex in buffer C and washed (3 × 1 min and 2 × 15 min) in buffer C. Following this, the sections were washed (3 × 1 min) in buffer A, incubated for 6 min in 0.1 M sodium phosphate with H2O2 (0.1 mg/ml) and diaminobenzidine (0.5 mg/ml) after 6 min of preincubation in sodium phosphate/diaminobenzidine without H2O2, and finally the reaction was stopped with 0.1 M sodium phosphate (2 × 3 min). Triton X-100 (0.1%) was included only when stated. For light microscopy, sections were mounted in glycerol–gelatin. Control experiments were performed after immunosorption of the antibody for 4 h with the fusion protein immobilized on agarose (as indicated above) after the antibody had been diluted to the final working concentration in buffer C. These controls resulted in complete suppression of staining of tissue sections with the antibody used in the present study. Staining was also absent when primary antibody was omitted. Brain areas were identified referring to the atlas of Paxinos and Watson (1982). For double labeling immunohistochemistry, sections were treated as above but secondary antibodies were coupled to the fluorescent dyes Alexa Fluor 488 or Alexa Fluor 594 and the immunofluorescence was visualized in an Axioscop2 (Zeiss) coupled to a confocal Microradiance (BioRad) or to a digital camera.
Pre-embedding Electron Microscopy
Immunoperoxidase reacted sections were treated with osmium (30–45 min, 10 mg/ml in 0.1 M sodium phosphate), washed (3 × 1 min) in 0.1 M sodium phosphate, dehydrated in graded ethanol (50, 70, 80 and 96% for 1 × 5 min and 100% for 3 × 10 min) and propylene oxide (2 × 5 min), and embedded in TAAB 812. Ultrathin sections were cut at right angles to the thick (40 μm) ones in order to be able to study the parts of the tissue that had been in immediate contact with the reagents. The sections were contrasted (with 10 mg/ml uranyl acetate for 10–15 min and 3 mg/ml Pb-citrate for 1–2 min) and examined in a Jeol 1010 electron microscope. For controls, primary antibody was substituted by pre-immune IgG, or immune IgG freed of specific IgG by immunosorption with the fusion protein. Immunosorption of the antibody was performed as indicated above. These controls resulted in complete suppression of staining of tissue sections with the antibody used in the present study.
Post-embedding Electron Microscopy
Small pieces of brain tissue were cryoprotected in increasing concentrations of glycerol in PB (10, 20 and 30%; 30 min in each) and store overnight at 4°C in the 30% glycerol solution, followed by 100% glycerol. The sections were slam-frozen on a copper plate cooled in liquid nitrogen (MM80E, Leica Microsystems Inc.) to preserve ultrastructure, freeze-substituted in propanol, then embedded in resin (Lowicryl HM20) according to the manufacturer's instructions. The resin blocks were polymerized under UV light, and ultrathin sections (60 nm) were cut using an Ultracut E microtome and collected on nickel slot grids. The sections were etched briefly (3 s) in a saturated solution of NaOH in absolute ethanol, washed in distilled water and blocked in 1% fetal calf serum in PBST for 1 h. The sections were then incubated in primary antibodies in blocking solution overnight, washed in PBS (3 × 10 min), and incubated in secondary antibodies (5 or 10 nm gold-conjugated anti-rabbit IgG at 1/50 dilution and 10 nm gold-conjugated anti-mouse IgG at 1/50 dilution) in blocking solution containing 0.5% polyethylene glycol for 2 h. The sections were washed as described above, postfixed in 1% glutaraldehyde, counterstained in 2% uranyl acetate and lead citrate according to standard techniques, and examined as described above. Controls were performed as stated for pre-embedding technique. For quantitative analysis pictures of distinct and transversally cut membranes were printed at ×80 000 magnification and the distance from the reference line to the centre of gold particles was measured manually (Matsubara et al., 1996, Landsend et al., 1997). For analysis of the perpendicular distribution of particles (presynaptic versus postsynaptic) we considered a line parallel to the postsynaptic density drawn along the synaptic cleft centre and we determined manually the shorter distance from each gold particle to this line.
Immunofluorescence in Cultured Cells
COS cells were transfected with expression vectors for GLYT1a or GLYT1b by using lipofectamine-PLUS following the manufacturer's instructions. After 48 h the cells were rinsed with PBS and fixed for 20 min with 4% paraformaldehyde in PBS. After washing in PBS, the cells were permeabilized and non-specific binding sites were blocked in PBS containing 1% BSA and 0.02% digitonin for 1 h at room temperature. Cells were incubated overnight at 4 °C with anti-GLYT1 antiserum in PBS containing 0.1% BSA and 0.02% digitonin. After washing three times in PBS, the cells were incubated with goat anti-guinea pig conjugated with Alexa 594 (1:250) for 1 h at room temperature. Cells were washed three times with PBS and mounted in Vectashield (Vector, Burlingame, CA).
Cerebral neocortex and hippocampus from 3-week-old rats were solubilized in ice-cold lysis buffer (50 mM Tris–HCl, pH 8.0, 150 mM NaCl and 2 mM EDTA, 0.1% SDS 1% NP-40 and 0.5% deoxycholate) for 30 min at 4°C. The solubilized material was centrifuged at 10 000 g for 20 min, and the supernatants were incubated with anti-GLYT1 antibodies (2 μg/ml) or with control antibodies overnight at 4°C. Then, 40 μl of a suspension of protein A cross-linked to agarose beads (Sigma) was added, and the mixture was incubated for 2 h with constant rotation at 4°C. The beads were washed twice with ice-cold lysis buffer and three times with PBS. Then, 50 μl of SDS–PAGE sample buffer was added to each sample. Bound proteins were dissociated by boiling for 5 min and resolved by SDS–PAGE on 10% gels as described above.
Characterization of a Novel Anti-GLYT1 Antibody
Despite the expression of GLYT1 mRNA by both neurons and glia in glycinergic and non-glycinergic areas (Smith et al., 1992; Borowsky et al., 1993; Zafra et al., 1995b), the antibodies currently available have only been able to detect the protein in glia (Zafra et al., 1995a). In an attempt to demonstrate that GLYT1 protein is indeed present in neurons, we have generated a novel polyclonal antiserum against a fusion protein containing the 37 N-terminal amino acids of the GLYT1b isoform. This affinity-purified antibody (antibody 224) was first tested by immunoblotting for immunoreactivity against a variety of GST fusion proteins. The serum reacted strongly with the GST-GLYT1Nt protein (Fig. 1A, lane 2), but it did not recognize GST whether alone (Fig. 1A, lane 1), fused to the C terminus of GLYT1 (Fig. 1A, lane 3) or fused to the N terminus of GLYT2 (Fig 1A, lane 4).
To test whether the antibody was also able to recognize the full-length protein, COS cells were transfected with expression vectors containing GLYT1 and cell extracts were again analyzed by immunoblotting. The antibody recognized a smear typical of a heavily glycosylated protein of between 70 and 100 kDa, similar in size and aspect to that recognized by previously described antibodies against GLYT1 (Fig. 1B). A second, 52–55 kDa, protein that corresponded in size to the partially glycosylated intracellular forms of the transporter reported previously (Olivares et al., 1995) was also recognized in COS cells overexpressing GLYT1. Moreover, a single band of 46–50 kDa was recognized in extracts from cells transfected with a construct in which the four glycosylation sites located in the large extracellular loop had been mutated, this representing the unmodified transporter peptide (Fig. 1B). None of these bands were observed in mock transfected COS cells.
Although the antigen against which the serum was generated corresponded to the N-terminus of the GLYT1b isoform, the antibody also recognized the GLYT1a isoform both on immunoblotting (Fig. 1B) and in immunofluorescence of transfected COS cells (Fig. 1C). This cross-reactivity was not surprising, since both isoforms share 27 of the 37 N-terminal amino acids in GLYT1. Nevertheless, the 224 antibody reacted more strongly with GLYT1b than GLYT1a in immunoblots (Fig. 1B, lanes 3–4). As expected, the C-terminal antibody (Ab 176) reacted identically with both isoforms which share this domain of the transporter (Fig. 1B, lanes 7–8). No cross-reactivity was observed with closely related transporters, such as GLYT2 or the GABA transporter (GAT1) transfected in COS cells (not shown). In addition, the 224 antibody recognized a single broad band in extracts from rat nervous system (Fig. 1B, lane 5), with an identical size and aspect to that recognized by the antibody 176 (Fig.1B, lane9). Indeed, as reported previously, the protein from native tissue was considerably smaller than that in transfected cells, ranging from 65 to 70 kDa. It has previously been shown that this difference in size is caused by differential glycosylation, since enzymatic deglycosylation of either the transfected or the native GLYT1 yielded the 46 kDa band (Olivares et al., 1995). Neither the 224 antibody nor 176 antibody reacted with proteins in other tissues such as the lung, liver, pancreas or kidney (not shown).
Immunohistochemical Localization of GLYT1 in the Rat Brain
General Regional Distribution
The regional distribution of GLYT1 was examined in rat brain sections with the anti-GLYT1 antibody. When the immunoperoxidase staining was visualized by low-power light microscopy (Fig. 2), in the nervous system the strongest immunostaining was found in the gray matter of the spinal cord (Fig. 2A). Intense staining was also observed in several nuclei of the pons and the medulla oblongata, such as the pontine, the spinal trigeminal, the facial, the hypoglossal, the inferior olive, the dorsal and ventral cochlear, the lateral lemniscus, the reticular, the vestibular or the cuneate nuclei, as well as in the superior olivary complex (Fig. 2B–D). In the cerebellum, dense staining was associated with the deep cerebellar nuclei, whereas in the granular and the molecular layers it was only moderate to dense (Fig. 2C,D). At the midbrain level, GLYT1 staining was intense and widespread, with the strongest labeling in the inferior and superior colliculi, in the red nucleus and in the periaqueductal gray (Fig. 2E,F). In the substantia nigra staining was moderate to dense, being most intense in the compact part (Fig. 2E).
At the level of the diencephalon (Fig. 2E–H), widespread GLYT1 immunostaining was found. In the thalamus, staining was in general moderate to intense, with the strongest labeling being detected in the principal sensory relay nuclei such as the ventroposteromedial, ventroposterolateral, and lateral geniculate and medial geniculate, whereas it was weaker in the posterior nucleus (Fig. 2E–H). Strong GLYT1 expression was also seen in some association nuclei, especially in the anterodorsal nucleus (Fig. 2H). In the hypothalamus, immunolabeling was more intense in the periventricular and medial regions than in the lateral ones, either at the preoptic, anterior, tuberal or mammilary levels (Fig. 2H). In the epithalamus, staining was more intense in the lateral than in the medial habenula (Fig. 2G). Also at the level of the diencephalon, we found strong GLYT1 expression in the subthalamic nucleus (Fig. 2G), in the pretectal area and in the nuclei of the optic tract (Fig. 2F).
In the forebrain, the septal region stained most intensely in the medial septum and in the nucleus of the diagonal band (Fig. 2I). In the hippocampus, the stratum lucidum and the hilar region of the dentate gyrus exhibited were moderately stained (Fig. 2E–G), as was the neocortex, although slightly stronger GLYT1 immunoreactivity was seen in layers 2–3 (Fig. 2E–I). In the olfactory bulb (not shown), the immunostaining was fairly dense in the periglomerular area and the external plexiform layer, while the granular layer was moderately stained.
Cellular and Subcellular Distribution of GLYT1 in the Forebrain. Light Microscopy
Whereas at low resolution the distribution of GLYT1 revealed by the 224 antibody did not appear to differ significantly from that observed with previously used antibodies, important differences were revealed at higher magnification, particularly in the forebrain regions. Thus, the staining with this antibody was analyzed in more detail in the neocortex, the hippocampus and associated regions. The labeling observed in the stratum lucidum of the hippocampus (Fig. 3A) was compatible with the presence of GLYT1 in mossy fiber terminals, since immunoreactivity was found in irregularly shaped structures, either delineating the dendrites of the principal CA3 pyramidal cells or scattered between their cell bodies (Fig. 3B). In addition to these terminals, a number of much smaller structures produced punctate staining in other areas of the hippocampus such as the stratum oriens and in the stratum radiatum of the CA3 area (Fig. 3C). Indeed, this punctate pattern also extended to these strata in CA1 (Fig. 3D). These structures were evenly distributed in the neuropil and outlining neuronal cell bodies, suggesting that they corresponded to terminal boutons.
It should be noted that under milder fixation procedures (0.5% PFA) and in the presence of Triton X-100, staining was also appreciated in fibers of the fimbria, the corpus callosum, the alveus or the lacunosum-moleculare (data not shown), and this staining was reduced, albeit not eliminated, when we used standard fixation in 4% PFA. However, the mild fixation protocol was detrimental to high resolution microscopy, and in general we used 4% PFA. Some scattered cell bodies were labeled, dispersed throughout all the hippocampus. In general, these cells had an appearance of non-pyramidal neurons (arrows in Fig. 3A). The immunostaining of the dentate gyrus was similar to that described for the stratum oriens and radiatum of the hippocampus, i.e. small punctate staining in the neuropil and delineating the neuronal cell bodies in the hilus, and in the granular and molecular layers, with a higher density of boutons in the hilar region (Fig. 3E).
In the neocortex, the distribution of GLYT1 staining was similar to that in the hippocampal formation: in all layers small boutons were labeled in the neuropil and around cell bodies (Fig. 3F–I). The highest concentration of boutons was observed in layer 1 (Fig. 3F) and the lowest in layer 6 (Fig. 3I). Pyramidal and non-pyramidal cells were contacted by immunostained boutons in all cortical layers. Some dendrites that were cut along their longitudinal and transversal axis presented a diffuse intracellular staining (Fig. 3G) and in addition, some stained astrocytes were seen, especially in layer 6 (Fig. 3I). This pattern of staining was repeated in the basal ganglia, the thalamus, and the hypothalamus: abundant stained boutons in the neuropil and delineating the cell bodies. An example, corresponding to the hypothalamus is presented in the Figure 3J. However, in the brainstem and spinal cord strong glial staining was evident in the neuropil and around cell bodies that was similar to that previously described for C-terminal antibodies (data not shown, Zafra et al., 1995a). This glial staining made it difficult to detect any punctate immunoreactivity. Also in keeping with other antibodies, we observed specific staining of amacrine cells in the retina (not shown, Zafra et al., 1995a).
Neuronal GLYT1 Localizes to Glutamatergic but not GABAergic Terminals in the Forebrain
Because the staining pattern described above was in certain areas reminiscent of the distribution of glutamatergic terminals, we compared the labeling of anti-GLYT1 antibodies with that of anti-vGLUT1 or anti-vGLUT2 antibodies. These two proteins define two subgroups of glutamatergic terminals and, together, label most of the glutamatergic terminals in the brain. It has been reported that vGLUT1 terminals are generally more abundant in the forebrain and vGLUT2 terminals more so in the caudal brain (FremEau et al., 2001; Herzog et al., 2001; Kaneko et al., 2002). When visualized in the same sections, an overlap was observed in the labeling for GLYT1 and vGLUT1 in forebrain regions such as the hippocampus (Fig. 4A–C) and amygdala (Fig. 4D–F). This co-localization could be appreciated, for example, in large mossy fiber terminals in the stratum lucidum of the hippocampus. Nevertheless, while GLYT1 frequently formed clusters that coincided with the vGLUT1 presynaptic marker, a number of smaller puncta were labeled for GLYT1 alone. On occasions, these small puncta were localized on the surface of transversally cut dendrites (arrowheads in Fig. 4A–C), compatible with them being postsynaptic clusters of GLYT1 (see below). Other puncta were intracellular and might correspond to vesicular GLYT1 being transported or recycled.
In the neocortex, co-localization of both proteins was less frequent than in the hippocampus or the amygdala, yet it was detected as white puncta in triple labeling experiments [some marked with arrowheads in layer 1 of the neocortex (Fig. 4G–J) or in layer 3 (Fig. 4K–N)]. There were a number of structures that were recognized by the anti-GLYT1 but not by the anti-vGLUT1 antibody. However, most of these structures were not synaptic boutons since they were not labeled with anti-synaptophysin (blue channel in Fig. 4G–N). Consequently they appeared as green structures in the merged images obtained in these triple labeling experiments. Of these, the larger puncta might correspond to axons cut transversally or glial cell processes (small arrows in Fig. 4K–N). While others were intracellular, a number of small puncta were also dispersed through the neuropil. Interestingly, it was unusual to find synaptic boutons containing GLYT1 and synaptophysin but not vGLUT1 in the neocortex. Notably, synaptophysin-containing puncta outlining pyramidal cells (large arrows in Fig. 4K–N), which probably corresponded to GABAergic terminals (they lacked vGLUT1), were not stained for GLYT1. The inverse was not true, and in the neocortex there were many vGLUT1 labeled terminals with low levels of GLYT1, indicating that GLYT1 is associated only with a subpopulation of glutamatergic terminals.
Consistent with a preferential association of GLYT1 with terminals containing vGLUT1, GLYT1 only occasionally co-localized with vGLUT2 in the forebrain. For instance, terminals containing vGLUT2 (Fig. 4O–Q) in the amygdala were devoid of GLYT1 immunoreactivity. These observations again indicate that GLYT1 is not ubiquitously expressed in glutamatergic terminals but rather that it is associated to specific subpopulations of excitatory terminals. Nevertheless, in caudal areas of the nervous system, like the spinal cord, vGLUT1-containing terminals were less strongly labeled by the GLYT1 antibody, whereas vGLUT2 terminals were abundantly immunolabeled (data not shown).
To investigate whether GLYT1 was also present in inhibitory terminals we compared labeling with the 224 antibody with that of an antibody raised against the vesicular transporter of inhibitory amino acids (VIAAT). In accordance with the aforementioned triple labeling experiments, GLYT1 only occasionally co-localized with VIAAT. Indeed, it was rare to find structures labeled by both antibodies in the neocortex (Fig. 4R–T). Similarly, GLYT1 did not appear to co-localize with the vesicular acetylcholine transporter (vAChT), a marker of cholinergic terminals (data not shown).
To investigate whether antibody 224 was able to recognize the glial GLYT1 in the neocortex, we performed co-localization experiments with the glial marker GFAP. As was the case with C-terminal antibodies, GLYT1 was detected on cell bodies and processes of GFAP immunoreactive cells, where the transporter formed large clusters (Fig. 5A–F). Clusters of GLYT1 were also appreciated on astrocytic profiles around blood vessels (Fig. 5G–I).
Ultrastructural Localization of GLYT1
To study the ultrastructural localization of GLYT1, we performed pre-embedding immunoperoxidase and post-embedding electron microscopy of hippocampal and cortical tissue. Pre-embedding studies revealed the presence of clumps of the peroxidase end-product precipitated on the cytosolic side of plasma membranes and on internal cellular structures, an observation consistent with the intracellular localization of the N-terminal epitope (Olivares et al., 1997). In the stratum lucidum of the hippocampus, both unmyelinated axons and terminals contained GLYT1. The labeled axons were observed within fasciculi of tightly packed fibers that probably correspond to the fasciculated grouping of the mossy fibers (arrowheads in Fig. 6A). Nevertheless, the strongest labeling in the stratum lucidum was found within the complex and multilobulated presynaptic expansions of the mossy fibers that were filled with synaptic vesicles and made synaptic contacts with the complex spines of the CA3 pyramidal cells — the thorny excrescences — (Fig. 6B). On occasions, peroxidase staining was observed within filopodial expansions of the mossy fiber terminals (not shown). In the CA1 region we also detected labeling within terminals making asymmetric synapses with simple dendritic spines, in both the stratum oriens (not shown) and radiatum (Fig. 6C). Similarly, in the neocortex we found labeling of terminals that established asymmetric synapses (Fig. 6D). GLYT1 labeling was also associated with dendrites and spines in both the hippocampus (not shown) and the neocortex (Fig. 6D,E), although the staining was in general weaker than in the terminals. Moreover, some occasional staining of glial profiles was observed in the hippocampus and neocortex (not shown; Zafra et al., 1995a). This contrasted with the strong labeling of glial profiles found in caudal regions such as the spinal cord (Fig. 6F). Nevertheless, neuronal elements such as dendrites (Fig. 6F) and terminals (not shown) were also labeled in the spinal cord. Staining was absent from control material (see Material and Methods; data not shown).
Due to the electron dense nature of the postsynaptic densities (PSDs), pre-embedding studies could not clarify to what extent GLYT1 localized to PSDs. Thus, to improve the resolution and to quantify GLYT1, post-embedding immunogold staining on ultrathin sections of hippocampal tissue was performed. Consistent with the pre-embedding experiments, gold particles were notably enriched at asymmetric synapses, identified by their dense postsynaptic density (Fig. 6G). Indeed, particles were observed both at the pre- (Fig. 6H) and postsynaptic side of these synapses (Fig. 6I), and in the dentate gyrus, the density of grains was higher in the postsynaptic element than in the presynaptic one. A quantitative analysis of the distribution of gold particles along the perpendicular axis of the synapses in the dentate gyrus revealed that 61% of the gold particles (187 out of 305, n = 75 synapses) were localized in the postsynaptic element whereas only 39% were presynaptic (Fig. 7A). Additionally, some labeling was localized close to the synapse and intracellular structures in the presynaptic element were frequently labeled (Fig. 6H). It remains to be determined whether these intracellular structures are synaptic vesicles or recycling endosomes, as shown for other transporters (Geerlings et al., 2001; Deken et al., 2003). Double labeling of ultrathin sections for GLYT1 and the NMDA receptor revealed the co-existence of these proteins in PSDs within the dentate gyrus (Fig. 6J).
The presynaptic/postsynaptic ratio of GLYT1 in the dentate gyrus contrasted with that found in other regions analyzed for comparative and control purposes. Thus, in the cerebellum, where GLYT1 was abundantly expressed in terminals of parallel fibers, GLYT1 was enriched in the presynaptic element (Fig. 6K). A quantitative analysis of the gold particles along the perpendicular axis of the parallel fiber–Purkinje cell synapses indicated that of the 327 particles counted in these synapses (n = 62), 226 (69%) were presynaptic (Fig. 7B). These particles were found within a stretch of ∼70 nm, both in the presynaptic membrane and the adjacent cytoplasm. These experiments indicated that the high concentration of GLYT1 present in the PSDs of hippocampal synapses cannot be attributed to non-specific binding of the antibody, a common phenomenon due to the high protein concentration at the PSDs.
GLYT1 Interacts with the NMDA Receptor
In view of the close association of GLYT1 with glutamatergic synapses, we adopted a biochemical approach to determine whether GLYT1 was physically associated with the NMDA receptor. Native rat cortical and hippocampal tissue was obtained and solubilized in a buffer containing a mixture of ionic and non-ionic detergents (0.1% SDS, 1% NP-40 and 0.5% deoxycholate). The material resistant to solubilization was removed by centrifugation and the solubilized proteins were immunoprecipitated with the anti-GLYT1 antibody or with preimmune serum. Immunoblot analysis of the precipitated material revealed that anti-GLYT1 antibody but not the preimmune serum, precipitated the NR2A subunit of the NMDA receptor (Fig. 8). However, antibodies against the closely related transporter GLYT2, which is expressed exclusively in glycinergic neurons, were unable to immunoprecipitate NR2A from solubilized proteins obtained from the brainstem, which were processed in parallel to the forebrain tissue. Together, these data indicate that GLYT1 is associated with the NMDA receptor in the forebrain of the rat.
Through the generation of a novel polyclonal antibody against the amino terminal region of the GLYT1b isoform, we have highlighted the close association between the glycine transporter GLYT1 and glutamatergic synapses. Both conventional and confocal microscopy, together with ultrastructural pre- and post-embedding electron microscopy, provided anatomical evidence in support of this association. The specificity of the antibody was similar to that of antibodies raised against the carboxyl terminal sequences of GLYT1, at least when evaluated by immunoblotting or immunocytochemistry of transfected COS cells. However, in tissue sections, this new antibody revealed the presence of GLYT1 in locations where it has not been previously detected. Specifically, the protein was now found in a large number of neuronal elements as well as in glia.
The staining of neuronal profiles was best appreciated in forebrain regions, since in the brainstem and spinal cord, the intense glial staining masked the detection of neuronal elements. Double labeling experiments with the glutamatergic marker vGLUT1 and the morphological analysis of cortical and hippocampal tissue by EM revealed that many of the labeled structures correspond to synaptic boutons that probably use glutamate as a neurotransmitter. However, the association of GLYT1 with glutamatergic markers was not ubiquitous and showed regional specificity. Thus, in the forebrain there was a preferential association with vGLUT1, while it was only sparsely associated with vGLUT2 in this tissue. However, in caudal regions there was an enrichment of GLYT1 in vGLUT2-labeled terminals. In the telencephalon, we did not find evidence for the co-localization of GLYT1 with inhibitory terminals, identified with antibodies against VIAAT, even in terminals outlining cell bodies in the neocortex. These findings are compatible with a recent immunohistochemical study that demonstrated the presence vGLUT1 in terminals outlining pyramidal and non-pyramidal neurons in the neocortex and in the neuropilar region (Alonso-Nanclares et al., 2004).
The association of GLYT1 with glutamatergic neurotransmission was further supported by ultrastructural data, indicating that GLYT1 concentrates not only in the presynaptic aspect of asymmetric glutamatergic synapses, but also in postsynaptic densities, where it co-localized with NMDA receptors. The tight association of GLYT1 with NMDA receptors was reinforced by the immunoprecipitation of GLYT1-NMDA receptor detergent insoluble complexes with anti-GLYT1 antibodies. It remains to be determined whether this association is mediated by a direct interaction between the transporter and the receptor or involves other interposed proteins. Interestingly, both the NR2A subunit and the C-terminal sequence of GLYT1 contain a PDZ-interacting motif that could tether GLYT1 to the postsynaptic density, where PDZ proteins form a complex scaffold that links NMDA receptors to regulatory proteins.
In addition, and as shown in our previous studies (Zafra et al., 1995a), GLYT1 was also detected in glia, particularly in the caudal CNS, as well as in amacrine cells of the retina. The high concentration of GLYT1 in glial processes in glycinergic areas suggests that the glial transporter is responsible for the removal of glycine from inhibitory glycinergic synapses. Indeed, recent experiments in GLYT1 deficient mice revealed a large increase of glycinergic inhibition in hypoglossal motoneurons that contributed to respiratory failure and the premature death of pups (Gomeza et al., 2003a). Since motoneurons are abundantly decorated with GLYT1 immunoreactive glia, our data are consistent with glial GLYT1 being essential for the regulation of glycinergic inhibition of motoneurons.
To date, it has not been possible to visualize the distribution of GLYT1 in neurons with the antibodies previously available (raised against the C-terminus), even using any of the different protocols applied here. This may be due to the inaccessibility of the neuronal C-terminal epitopes, a problem that does not seem to exist for N-terminal epitopes. Nevertheless, the accessibility of this epitope seems to be crucial to observe immunoreactivity in axonal fibers, since this was only optimally detected under mild fixation protocols.
The presence of a neuronal form of GLYT1 has been suggested on the basis of pharmacological experiments in synaptosomes derived from the cortex, cerebellum and spinal cord (Herdon et al., 2001; Raiteri et al., 2001). Our observations provide direct anatomical evidence for the existence of this neuronal form of the transporter and substantiate previous observations of a tight association between GLYT1 and glutamatergic neurotransmission at the mRNA level (Smith et al., 1992; Borowsky et al., 1993; Zafra et al., 1995b). Indeed, mRNA for GLYT1 was observed in many neuronal types throughout the brain, including, among others, the granular and pyramidal neurons of the hippocampus and pyramidal neurons in the neocortex that are probably the source of the glutamatergic terminals containing GLYT1 detected in the present study.
The localization of GLYT1 in excitatory synapses that probably use glutamate as a neurotransmitter is of particular interest given the controversial role of glycine in NMDA receptor function. The low micromolar levels of glycine measured in the extracellular and cerebrospinal fluids initially led to the conclusion that they would saturate the glycine sites on NMDA receptors (Westergren et al., 1994). However, glycine transporters might reduce the glycine concentration in the local environment of the NMDA receptor to below 1 μM (Attwell, 1993; Supplisson and Bergman 1997; Berger et al., 1998; Bergeron et al., 1998; Roux and Supplisson, 2000; Billups and Attwell, 2003). Moreover, NMDA populations with low affinity for glycine have been described (Kew et al., 1998). Indeed, recent pharmacological and electrophysiological evidence support a regulatory role for glycine transporters on NMDA receptor. Thus, the potent and selective inhibition of GLYT1 with NFPS has implicated this transporter in long-term potentiation in the dentate gyrus induced by the stimulation of the perforant path (Kinney et al., 2003). Furthermore, the administration of NFPS in vivo produced physiological responses consistent with those expected after potentiation of NMDA receptor function (Kinney et al., 2003). Moreover, experiments performed in mutant mice that expressed only 50% of GLYT1 (heterozygous Glyt1+/−) revealed an enhanced hippocampal NMDA receptor function. In the water maze, these mice exhibited better spatial retention. These animals were also protected against an amphetamine disruption of sensory gating, suggesting that drugs which inhibit GLYT1 might have both cognitive enhancing and antipsychotic effects (Tsai et al., 2004).
Finally, a novel role for glycine in regulating NMDA receptor internalization has been recently suggested. Apparently this novel function requires higher concentrations of glycine than those necessary to operate the ion channel (Nong et al., 2003). Thus, the transporter might regulate the turnover of the NMDA receptor at the cell surface, rather than the channel function.
In summary, the anatomical and biochemical data reported in this study reveal the presence of GLYT1 in subpopulations of glutamatergic synapses of the telencephalon, in addition to previously reported glial transporter. These data explain the partially contradictory distribution of mRNA and protein reported previously, and they support an active role of GLYT1 in regulating the glycine concentrations in the local environment of the NMDA receptor.
We thank the expert technical help of E. Núñez, and Prof. Andrew Todd for comments on the manuscript. Anti-VIAAT antibodies were kindly provided by Dr B. Gasnier. This work was supported by the Spanish Dirección General de Investigación Científica y Técnica (BMC2002-03502), Fondo de Investigaciones Sanitarias, Comunidad Autónoma de Madrid and an institutional grant from the Fundación Ramón Areces.