Presence of the vesicular inhibitory amino acid transporter in GABAergic and glycinergic synaptic terminal boutons

Neurons secrete a variety of signalling molecules by regulated exocytosis. In contrast with neuropeptides, whose nascent polypeptide precursors enter the secretory pathway via the endoplasmic reticulum (long before their organelles bud from the trans-Golgi), small molecule neurotransmitters are imported into synaptic vesicles by an active uptake from the cytosol. This process allows the vesicles to replenish with transmitter after exocytosis and to undergo several cycles of secretion during sustained firing of the neuron.

The uptake of small molecule neurotransmitters into synaptic vesicles is ensured by two proteins of the vesicle membrane: a V-type H⁺-ATPase (Nelson and Klionsky, 1996), which acidifies and positively charges the vesicle lumen, and a neurotransmitter transporter, which exchanges lumenal protons for cytosolic transmitters. The vesicular transporter determines, at least partly, which transmitter is loaded into the vesicle. The best characterized of these transporters is the vesicular monoamine transporter (VMAT), responsible for the vesicular storage of catecholamines, serotonin and histamine (Henry et al., 1998). Two isoforms, VMAT1 and VMAT2, encoded by distinct genes, have been identified by cDNA expression cloning (Erickson et al., 1992; Liu et al., 1992). Later on, the cloning of the Caenorhabditis elegans unc-17 gene revealed an homologous protein that was identified as the vesicular acetylcholine transporter (Alfonso et al., 1993; Varoqui et al., 1994; Varoqui and Erickson, 1996).

However, recent studies indicated that the transporters of amino acid transmitters belong to a different family of proteins. A breakthrough in the identification of the γ-aminobutyric acid (GABA) transporter was provided by insightful studies in C. elegans. The nematode propels itself by sinusoidal movements, in which GABAergic neurons ensure the cross-inhibition of opposing dorsal and ventral body muscles. McIntire et al. (1993a) ablated these neurons with a laser microbeam and observed that treated animals shrank along their body axis when prodded, because of the simultaneous contraction of dorsal and ventral muscles. Interestingly, earlier genetic studies had described a similar shrinker phenotype for five uncoordinated mutants of the nematode (Hodgkin, 1983). Further examination of these mutants showed that they were selectively impaired in GABAergic transmission (McIntire et al., 1993b). One gene, unc-47, was unambiguously assigned to a presynaptic deficit in the transmission occurring beyond the GABA synthesis step and was thus proposed to code for the vesicular GABA transporter.

Recently, the unc-47 gene has been identified by in vivo (McIntire et al., 1997) or in silico positional cloning (Sagné et al., 1997). The UNC-47 sequence revealed no homology with previously known neurotransmitter transporters. Tagging of UNC-47 with green fluorescent protein in the nematode showed a localization within GABAergic axonal varicosities (McIntire et al., 1997). Moreover, in both studies, a rodent homologue of unc-47 was cloned and shown to induce an intracellular uptake of GABA when expressed in COS-7 or PC12 cells. The pharmacological and bioenergetical profile of the uptake (McIntire et al., 1997) as well as the mRNA distribution in rat brain (Sagné et al., 1997) further suggested an identification with the vesicular GABA transporter.

Biochemical studies have shown that the mammalian vesicular transporters of GABA and glycine, another fast inhibitory transmitter, are functionally similar, since the two transmitters compete for their uptake into synaptic vesicles (Burger et al., 1991; Christensen et al., 1991; for a different view, see Kish et al., 1989). The two above-mentionned cloning studies investigated their possible identity but reached opposite conclusions in this respect. McIntire and colleagues (1997), who used a direct assay for transport, failed to observe a significant accumulation of glycine and suggested the existence of distinct transporters. On the other hand, our group showed that the unc-47-like mRNA was expressed in both GABAergic and glycinergic nuclei of the rat brain, suggesting rather the existence of a common ‘Vesicular Inhibitory Amino Acid Transporter’ (VIAAT, Sagné et al., 1997). To address this issue, we have raised a polyclonal antibody against VIAAT and studied the localization of the protein in neurons of rat spinal cord and cerebellum by immunocytochemistry at light and electron microscopy levels. Our results show a presynaptic localization of VIAAT, in GABA- and in glycine-containing terminal boutons, but also in boutons showing immunoreactivity for both amino acids, facing glycine receptor components. These data further support the view of a common transporter that may account for the corelease of GABA and glycine at some synapses, as initially suggested by anatomical studies (Triller et al., 1987; Todd et al., 1996) and recently demonstrated by an electrophysiological study in the spinal cord (Jonas et al., 1998).

The comparison of our mouse VIAAT cDNA sequence (Sagné et al., 1997) with its rat homologue (McIntire et al., 1997) suggested that the mouse clone lacked the authentic ATG start codon. We thus performed reverse transcriptase-polymerase chain reaction (RT-PCR) experiments to amplify mouse cDNAs extended at the 5′ end. PCR was performed on whole brain Marathon-Ready cDNAs from male, 9-to 11-week-old BALB/c mice (Clontech), using Advantage-HF polymerase mix in the high-fidelity buffer (Clontech). Amplification with primers 5′-CGTCCCCAGACCCTTCTGTCCTTTTC, derived from the rat 5′-untranslated sequence, and 5′-CCACATCGA-AGAAGACCTGGTGC, antisense to the mouse coding sequence, yielded a 1529 bp product. This PCR product was characterized by TOPO TA cloning into pCR2.1-TOPO (Invitrogen) and by automated sequencing (ABI) of four clones derived from independent amplifications. A novel full-length mouse clone, designated pcDNA3-mVIAAT98, was constructed by sub-cloning into the pcDNA3-mVIAAT plasmid (Sagné et al., 1997) at restriction sites HindIII (upstream from the 1529 bp cDNA) and AflII (in the VIAAT coding sequence). The sequence, accessible in an update of the EMBL Data Bank entry #AJ001598, codes for a 521 amino acid protein, which differs from the rat protein by three point mutations and a divergent C terminus. The novel first ATG of the mouse sequence is in a better context for initiation of translation (Kozak, 1996) than was the previous one (now corresponding to M27). Consistently, we observed that the ‘long’ mVIAAT clone was translated in vitro at much higher levels than the ‘short’ one (data not shown), suggesting that the novel first ATG represents the actual initiation codon.

A plasmid encoding a glutathione S-transferase (GST) fusion protein containing the first 127 amino acids of the mouse VIAAT was constructed as follows. A cDNA coding for the N-terminal domain of mVIAAT was generated by PCR on a pCR2.1-TOPO clone (see above) using Taq Polymerase (Qiagen) and the primers MT20-32S (5′-GTGGATCCCCGGCCACCCTGCTCCGCAGC) and MT20-33A (5′-GGGAATTCTCAATTTGTCACGTTCCAGCCCGC). The sense primer MT20-32S contains a BamHI restriction site followed by a mutation of the first methionine to proline (both underlined) to ensure the fusion to the GST coding sequence of the pGEX4T1 vector (Pharmacia). The antisense primer MT20-33A contains a STOP codon followed by an EcoRI site for cloning (both underlined). The resulting 402-bp PCR product was purified by agarose gel electrophoresis and QIAquick gel extraction (Qiagen), and cloned into pGEX4T1 at the BamHI and EcoRI restriction sites. The resulting plasmid, pGEX4T1-VIAATNter, was verified by automated sequencing on both strands.

Expression of the GST-VIAAT fusion protein was induced by treating an Escherichia coli DH5α [pGEX4T1-VIAATNter] culture in the exponential phase of growth with 1 mM isopropyl β-D-thiogalactoside for 2 hours. The protein, which appeared as a 44 kDa band on SDS-polyacrylamide gels, was extracted and purified from the bacterial pellet by affinity chromatography on glutathione Sepharose 4B (Pharmacia) according to the manufacturer’s recommendations. Elution was performed either with glutathione, to recover the intact fusion protein, or with thrombin, to cleave the VIAAT N-terminal domain from the GST moiety bound to the column.

New Zealand white rabbits were immunized according to standard procedures. They received a first injection of 250 μg fusion protein followed by three injections of 250 μg of the N-terminal domain of VIAAT (without GST) every 2 weeks. The serum used in this study was collected from one of the rabbits at day 63.

Plasmids were purified using Qiagen anion-exchange resin. COS-7 cells were grown at 37°C under 5% CO₂ in glucose-rich Dulbecco’s Modified Eagle Medium (Gibco BRL) supplemented with 7.5% foetal bovine serum (Gibco BRL). 10⁶ cells in 100 μl of ice-cold phosphate buffered saline, pH 7.4 (PBS) were transiently transfected with 10 μg of VIAAT plasmid by electroporation with a PS10 electropulsator (Jouan). After addition of the plasmid, cells were immediately subjected to 10 square pulses (250 V, 3 ms) delivered at 1 Hz by 4 mm-spaced electrodes, diluted in 10 ml of culture medium and incubated. Transfection with a VMAT2 plasmid (Gasnier et al., 1994) was used as a control.

For immunofluorescence analysis, the transfected cells were recovered on the following day by mild trypsinisation, plated on sterile, uncoated 12 mm glass coverslips and cultured. 2 days after transfection, cells were fixed for 10 minutes with 4% paraformaldehyde (PFA), washed and permeabilized for 1 hour with 0.2% Triton X-100 PBS containing 100 μM CaCl₂, 100 μM MgCl₂ and 5% donkey serum, before being incubated for 1 hour with the polyclonal anti-VIAAT serum (1:5,000), rinsed, incubated for 1 hour with a carboxy-metylindocyanin-3 (CY3)-coupled donkey anti-rabbit IgG (1:600, Jackson ImmunoResearch Laboratories) and rinsed again. Coverslips were mounted in Fluoromount-G (Southern Biotechnology Associates) and examined with a standard epifluorescence microscope (Leica DMRD).

COS-7 cells were scraped 3 days after transfection and homogenized with a glass-Teflon device in ice-cold PBS containing 5 μg/ml each of leupeptin, pepstatin and aprotinin, 1 mM EGTA and 1 mM of extemporaneously added PMSF. After centrifugation for 5 minutes at 100 g to remove cellular debris, total membranes were collected by centrifugation for 1 hour at 200,000 g. Pellets were resuspended in the same ice-cold buffer. Samples were frozen in liquid nitrogen and stored frozen at −80°C. Neuronal and non-neuronal tissues were dissected from a 2-month-old female Sprague-Dawley rat, minced with a scalpel and homogenized as described above in PBS containing protease inhibitors. COS-7 cell membranes (1-10 μg of protein per lane) or rat tissue homogenates (10 μg of protein) were analyzed by SDS-PAGE on a 10% acrylamide gel and electrotransferred onto a nitrocellulose membrane according to standard procedures. The VIAAT immunoreactivity on the membrane was detected using the anti-VIAAT serum (1:5,000) and an anti-rabbit IgG goat antibody conjugated to horseradish peroxidase (1:300,000; Sigma) as primary and secondary antibodies, respectively. The peroxidase activity was detected on Kodak X-OMAT AR films using SuperSignal ULTRA chemiluminescent substrate (Pierce).

For comparison, plasmids were also transcribed and translated in vitro in the presence of [³⁵S]methionine (1000 Ci/mmol, Amersham), using TNT-coupled reticulocyte lysate systems (Promega) and canine pancreatic microsomal membranes (Promega) according the manufacturer’s recommendations. The translation products were analyzed by SDS-PAGE and exposure to X-OMAT AR films for autoradiography.

The polyclonal anti-VIAAT antibody (1:200 and 1:500, for detection in cell culture and tissue sections, respectively) was combined to several monoclonal antibodies, directed against synaptophysin (1:20 and 1:200, Boehringer Mannheim), GAD65 (1:200; Boehringer Mannheim) and gephyrin (mAb7a, 1:200, Boehringer Mannheim). The mAb4a, recognizing the α- and β-subunits of the glycine receptor (GlyRα?β)? was used for detection in cell cultures only (1:100, gift from Dr H. Betz, Frankfurt). For this antibody, an additional fixation in a mixture of methanol:acetic acid (95:5) for 10 minutes at −20°C was necessary. Primary antibodies were detected with a combination of CY3-coupled goat anti-mouse IgG (1:200; Jackson ImmunoResearch Laboratories) and fluorescein isothiocyanate (FITC)-coupled goat anti-rabbit IgG (1:200; Jackson ImmunoResearch Laboratories). All monoclonal antibodies used in this study have been extensively characterized by other groups: anti-synaptophysin (Wiedenmann and Franke, 1985), anti-GAD65 (Gottlieb et al., 1986), mAb7a and mAb4a (Pfeiffer et al., 1984). In our controls, no staining was observed when the primary antibodies were omitted.

Primary cultures of spinal cord neurons from Sprague-Dawley embryos at day 14 of gestation were performed as previously described (Béchade et al., 1996). Prior to cell plating, coverslips and Petri dishes were coated with 15 μg/ml poly(D-L)ornithine (Sigma) and culture supports were incubated with medium containing 5% of inactivated fetal calf serum (Gibco). Cells were plated in the serum-free supplement B27-Neurobasal medium combination (Gibco) at a density of 4×10⁵ cells/ml on sterilized glass coverslips in 4-well plates (Nunc). Cultures were kept for 30 days at 37°C in 7.5% CO_2, the medium being changed once a week. After this period, cells were fixed with 4% PFA in PBS for 15 minutes, washed in PBS, immersed for 20 minutes in 50 mM ammonium chloride in PBS and permeabilized with PBS containing 0.1% Triton X-100 and 0.1% gelatin (PBSTg) for 5 minutes before being incubated with a combination of primary antibodies for 1 hour. Following rinses, cells were simultaneously incubated with the two secondary antibodies for 45 minutes, rinsed again and the coverslips mounted with Vectashield (Vector Laboratories), prior to examination with a standard epifluorescence microscope (Leica DMRD).

Three adult female Sprague Dawley rats (Janvier) were deeply anesthetized with pentobarbital (60 mg/kg body mass, i.p.) and intracardially perfused with 4% PFA in PBS. The cervical spinal cord segment and cerebellum were removed, postfixed for 12-15 hours in 4% PFA and cut into 40 μm-thick sections in a coronal plane. Sections were immersed for 20 minutes in 50 mM ammonium chloride in PBS, rinsed and incubated overnight at 4°C either with the anti-VIAAT antibody alone or combined with one of the monoclonal antibodies described above, in PBSTg. After rinses, sections were incubated with the secondary antibodies for 2 hours, rinsed again and mounted with Vectashield before being examined with a standard epifluorescence microscope.

For quantification of the number of terminals showing colocalization of two markers, 30 neurons from a single experiment in cell culture and 30 motoneurons from spinal cord sections from three different animals and experiments were randomly selected. Successive FITC- and CY3-fluorescent images were digitized using a standard epifluorescence microscope (Leica DMRD) and a chilled CCD camera (Hamamatsu C5985). The quantification of the number of single- or double-labelled boutons around the soma and initial portion of dendrites was achieved using ImageSpace software (Molecular Dynamics). Mean values ± s.e.m. were calculated using StatView F.4.11 software (Abacus Concepts, Inc.).

Four adult Sprague Dawley rats (Janvier) were deeply anesthetized with pentobarbital (60 mg/kg body mass, i.p.) and intracardially perfused with 4% PFA-0.5% glutaraldehyde. The cervical spinal cord segment and cerebellum were removed, postfixed for 12-15 hours in 4% PFA and then cut into 100-μm-thick sections in a coronal plane.

Sections were immersed for 20 minutes in 50 mM ammonium chloride in PBS, rinsed and incubated overnight at 4°C with the anti-VIAAT antibody (1:2,000) in PBS containing 0.1% gelatin (PBSg). Detection of the antibody was carried by the avidin-biotin complex method (Kit Elite Vectastain, Vector Laboratories). Briefly, sections were successively incubated with a goat anti-rabbit biotinylated antibody (1:200) in PBSg for 4 hours at room temperature, followed by an avidin-peroxidase complex (1:200) for 3 hours and, after a preincubation in 0.6% diaminobenzidine (DAB) in Tris-buffered saline (0.06 M, pH 7.4), sections were finally incubated with DAB and hydrogen peroxide in the same buffer (Sigma Fast DAB, Sigma). They were then postfixed for 1 hour at 4°C with 2% osmium tetraoxide in PBS, dehydrated through a graded series of ethanol and finally flat-embedded in epoxy resin (Durcupan, Fluka). Ultrathin sections (pale yellow) were collected on 400-mesh Formvar-coated nickel grids, counterstained for 10 minutes with 2% uranyl acetate and 0.2% Reynold’s lead citrate and examined in a Jeol JEM-100CXII electron microscope.

Sections were immersed for 20 minutes in 50 mM ammonium chloride in PBS, rinsed and incubated overnight at 4°C with the anti-VIAAT antibody (1:2,000) in PBSg. After rinses, section were incubated with an anti-rabbit IgG coupled to 1.4 nm gold particles for 6 hours (1:100, Nanoprobe). Gold particles were stabilized for 5 minutes in 1% glutaraldehyde in PBS and intensified for 7 minutes at 20°C with an HQ Silver Kit (Nanoprobe). After a gold-toning step (Liposits et al., 1982), sections were incubated overnight at 4°C with the anti-gephyrin antibody (1:100) in PBSg. Detection of the latter was carried by the avidin-biotin complex method as described above, but using a goat anti-mouse biotinylated antibody (1:200; Vector Laboratories). Osmification, dehydration and flat-embedding in Durcupan of sections were performed as described above.

Detection of GABA- and glycine-like immunoreactivities was performed on ultrathin sections that had been reacted with the anti-VIAAT antibody as described above. Adjacent ultrathin sections were etched for 7 minutes in 1% orthoperiodic acid and incubated with either an anti-GABA (1:500, Immunotech; Séguéla et al., 1984) or an anti-glycine antibody (1:400, gift from Dr Ottersen, Oslo; Kolston et al., 1992) in Tris-buffered saline (5 mM Tris-HCl, pH 7.4) containing 9% sodium chloride, 2% bovine serum albumine, 5% sodium azide and 10% Triton X-100. Primary antibodies were detected with a goat anti-rabbit IgG coupled to 15-nm colloidal gold particles (1:50, BioCell). As for the pre-embedding method, sections were counterstained and examined in a Jeol JEM-100CXII electron microscope. The preadsorption with glutaraldehyde-coupled GABA or glycine (50 μM) of the anti-GABA or anti-glycine antibody led to the abolishment of the respective immunoreactivity. Moreover, the two antibodies have been shown not to cross-react in these experimental conditions (Colin et al., 1998).

VIAAT possesses a large, hydrophilic domain before the first transmembrane segment. This N-terminal domain (amino acids 2-127, mouse sequence) was expressed and purified as a GST-fusion protein and, after removal of the GST moiety, used as an antigen to raise polyclonal antibodies in rabbits. It might be noted that a single point mutation exists between mouse and rat in this domain. The serum was first characterized by immunofluorescence analysis of COS-7 cells transiently transfected with an mVIAAT- (pcDNA3-mVIAAT98) or bVMAT2-expression plasmid (Gasnier et al., 1994). As shown in Fig. 1B, fluorescent cells were detected in the VIAAT-transfected culture, but not in the VMAT2-transfected one (Fig. 1C). The VIAAT-immunoreactive signal appeared as punctate or reticulate material scattered in the cytoplasm (Fig. 1D-F), indicating that VIAAT was localized in an intracellular compartment. Positive results were also obtained with serum from another rabbit or transfection with the ‘short’ VIAAT clone (data not shown).

The polyclonal antibody was further characterized by western blot analysis. A strong immunoreactive band was detected in membranes from mVIAAT-but not from bVMAT2-transfected cells (Fig. 2A, left). This band had an apparent molecular mass of 57 kDa, which is consistent with values calculated from the sequence (56,814 Da) or obtained from an in vitro translation product (56 kDa; Fig. 2A, right). It might be noted that the molecular mass of the in vitro translation product was independent of the presence of microsomes, showing that VIAAT is not N-glycosylated and thus confirming a feature of the current topological model (Sagné et al., 1997; McIntire et al., 1997). Two fainter 49 and 50 kDa bands were also detected in the VIAAT-transfected cells, which might correspond to degradation products. The analysis of rat tissue homogenates revealed a strong 58 kDa immunoreactive band and a faint 57 kDa one in all central nervous system (CNS) regions except the olfactory bulb, where the 57 kDa band predominated (Fig. 2B). Similar results were obtained in mouse (data not shown). No labelling was detected in peripheral tissues with the anti-VIAAT serum (Fig. 2B), nor in brain with the pre-immune serum (data not shown). We thus concluded that the anti-VIAAT serum selectively recognizes the rodent VIAAT.

In order to study the neuronal distribution of VIAAT, we first focused on the rat spinal cord, where well-characterized inhibitory synapses are found. The distribution of VIAAT immunoreactivity was compared with those of synaptophysin, a marker of synaptic vesicles (Wiedenmann and Franke, 1985) and GAD65, a marker of GABAergic terminals (Erlander and Tobin, 1991). We also used two postsynaptic markers, the α- and β-subunits of the glycine receptor (GlyRα?β) and its anchoring protein, gephyrin (Vannier and Triller, 1997). In spinal cord sections only gephyrin was used because fixation procedures were ineffective for double anti-VIAAT and anti-GlyRα?β immunodetection.

We first investigated a model of primary culture of spinal cord neurons currently used in our laboratory for studying synaptic mechanisms. By immunocytochemistry, VIAAT was shown to form bright and numerous puncta on somata and dendrites of virtually all neurons (Fig. 3A1-D1). VIAAT immunoreactivity colocalized with synaptophysin-immunoreactive profiles (Fig. 3A1-3), indicating that VIAAT is exclusively present in nerve terminals; 35.0±1.5% of the synaptophysin-positive boutons were however negative for VIAAT (Table 1). These are probably excitatory boutons (see Discussion). In sections labelled with VIAAT and GAD65, up to 85.2±1.3% of VIAAT-positive endings colocalized with GAD65 (Fig. 3B1-3). Simultaneous detection of VIAAT and gephyrin or GlyRα/β indicated that VIAAT-immunoreactive boutons were not colocalized with but apposed to gephyrin- (Fig. 3C1-3) or GlyRα/β-immunoreactive clusters (Fig. 3D1-3).

We then examined sections from the adult rat spinal cord, where VIAAT immunoreactivity similarly showed a punctate, bright labelling around cell bodies and fibers (Fig. 4A1-C1) in all layers of the grey matter and in fibers within the white matter. Labelling was restricted to nerve terminals, as shown by its exclusive colocalization with synaptophysin immunoreactivity (Fig. 4A3). There were 35.5±1.5% (Table 1) of the synaptophysin-positive terminals that were devoid of VIAAT. Particular attention was paid to the innervation of motoneurons, since both GABA and glycine exert a control upon them. We observed that most of the VIAAT-positive boutons were apposed to gephyrin-positive clusters on dendrites and soma (Fig. 4C3), whereas only a subset of them was colocalized with GAD65 immunoreactivity (Fig. 4B3). The occurrence of the latter subset (32.9±1.7%, Table 1) was in good agreement with previous quantitative data on inhibitory synapses apposed to motoneuron soma (Örnung et al., 1996).

Since purely GABAergic synapses are rare in the ventral horn and cannot be unambiguously recognized as such, we took advantage of the well-known circuitry of the cerebellum to look for the expression of VIAAT in electrophysiolocally characterized GABAergic synapses (Puia et al., 1994; Kaneda et al., 1995). VIAAT labelling was found in the three layers of the cerebellar cortex (Fig. 4D) and labelled the typically GABAergic axon terminal boutons such as those of the basket cells, which contact the Purkinje cell body and form the ‘pinceau’ around its initial axonal segment (Fig. 4E), as well as the Golgi cell terminal boutons within glomerular structures of the granular layer (Fig. 4F). In summary, light microscopy data indicate that VIAAT immunoreactivity is detected in GABAergic terminals as well as in terminals facing glycine receptor components, the latter corresponding to either glycinergic or mixed GABA- and-glycinergic terminals (see Discussion).

Ultrastructural immunoreactivity of VIAAT was first studied in the ventral horn of the adult rat spinal cord by means of a peroxidase reaction product in a pre-embedding procedure. Immunoreactivity was found within axonal synaptic boutons containing pleomorphic vesicles and establishing synaptic symmetrical contacts (Fig. 5A), but never in the afferent axon (Fig. 5B) nor in boutons containing round vesicles and showing asymmetrical synaptic contacts. Detection of VIAAT with intensified gold particles showed that the particles were preferentially located in regions enriched with vesicles (Fig. 5C1,2). Double immunolabelling of this material with the mAb7a showed gephyrin immunoreactivity in the inner face of the postsynaptic dendritic membrane in regions facing VIAAT-containing boutons, while regions of the same dendrite facing VIAAT-negative boutons were devoid of gephyrin immunoreactivity (Fig. 5C1). Hence, the ultrastructural distribution of VIAAT confirms a role of VIAAT at inhibitory terminals.

Light microscopy results suggested that VIAAT was closely related to GABAergic and glycinergic markers. We performed immunodetection of GABA and glycine, revealed with 15 nm gold particles, on serial ultrathin sections from a block previously reacted for VIAAT (detected by peroxidase product). Detection of glycine showed a high density of gold particles in boutons labelled for VIAAT (Fig. 6A1). The same boutons also appeared to contain GABA, as seen in adjacent sections reacted with anti-GABA serum (Fig. 6A2). However, in many cases boutons displayed glycine immunoreactivity only (Fig. 6B,C). An additional, striking finding was that some glycine-containing boutons were devoid of VIAAT although VIAAT-labelled terminals containing glycine were detected in the same section (Fig. 6D,E).

The subcellular localization of VIAAT was also investigated in GABAergic synapses of the cerebellar cortex, that can be easily recognized on morphological basis. As in light microscopy immunocytochemistry, VIAAT labelled the basket cell terminals (Fig. 7A,B) and the Golgi cell axonal boutons (Fig. 7C,D). Taken together, the electron microscopy results demonstrate that VIAAT is present in the CNS at the presynaptic level either in boutons containing only glycine or GABA, or in mixed GABA- and glycine-containing terminals.

A polyclonal antibody against the N-terminal domain of the rodent VIAAT was raised; immunocytochemical and immunoblotting analysis of transfected COS-7 cells indicated that the antibody selectively recognized this protein. The distribution of VIAAT immunoreactivity in COS-7 cells indicated that the transporter was localized in an intracellular compartment, as previously suggested from GABA uptake experiments (Sagné et al., 1997). In the rat CNS, the antibody mainly labelled a 58 kDa band, in agreement with the expected molecular mass of VIAAT. The minor detected band of 57 kDa might correspond to an alternatively processed form of VIAAT, particularly abundant in the olfactory bulb, for a reason to be investigated. Since this study addressed the question of the role of VIAAT in relation to glycine, it is important to notice that the 58 kDa band predominated equally in mainly glycinergic (spinal cord, brainstem) and GABAergic regions of the CNS (cerebral cortex, cerebellum).

In the present study, the localization of VIAAT was investigated in situ by immunocytochemistry. At the light microscope level, VIAAT immunoreactivity appeared to be colocalized with that of synaptophysin, a synaptic vesicle marker, as previously observed on cultured hippocampal neurons (McIntire et al., 1997). However, 35% of the synaptophysin-immunoreactive boutons were devoid of VIAAT (Table 1). These most probably represent excitatory boutons, since their occurrence is in the same range as that obtained from direct or indirect quantification of glutamate-positive boutons in cell cultures (O’Brien et al., 1997) and spinal cord sections (Örnung et al., 1996, 1998). The presynaptic localization of VIAAT was also supported by the fact that it appeared slightly shifted in relation to the postsynaptic markers GlyRα?β or its associated protein gephyrin. At the ultrastructural level VIAAT was exclusively present in presynaptic boutons (but not in the afferent axon), at synapses that exhibited classical inhibitory features such as pleomorphic vesicles and symmetrical postsynaptic densities, further demonstrating the presynaptic localization of VIAAT. The vesicular localization of VIAAT was initially deduced from pharmacological and bioenergetical characteristics of the GABA uptake activity (McIntire et al., 1997; Sagné et al., 1997), and from the distribution of its nematode orthologue in unc-104 mutants that lack a synaptic vesicle-specific kinesin and in which the UNC-47 protein does not reach synaptic varicosities but accumulates in cell bodies (Hall and Hedgecock, 1991; McIntire et al., 1997). The present data, including the fact that VIAAT immunoreactivity was preferentially distributed in regions enriched in synaptic vesicles within the bouton, provide strong and direct support to a vesicular localization of VIAAT.

Inhibitory input on spinal cord motoneurons is mainly purely glycinergic or mixed GABA- and-glycinergic, and seldom GABAergic (Örnung et al., 1996). The co-existence of GABA and glycine and of their postsynaptic ionotropic receptors at mixed spinal synapses is supported by a large body of morphological evidence at both light microscope (Triller et al., 1987; Mitchell et al., 1993; Bohlhalter et al., 1994) and electron microscope levels (Triller et al., 1987; Taal and Holstege, 1994; Örnung et al., 1994; Todd et al., 1996), as well as by a recent electrophysiological study (Jonas et al., 1998). Moreover, it has been shown that gephyrin is detected at postsynaptic densities facing glycine (47.6%) or glycine- and GABA-containing boutons (51.2%) of the rat spinal ventral horn (Colin et al., 1998). Thus, the VIAAT immunoreactivity apposed to gephyrin-positive clusters observed in the present study in the spinal ventral horn corresponded to both pure glycinergic and mixed GABA- and-glycine boutons, but only the post-embedding detection of the amino acids could demonstrate the phenotype of the presynaptic terminal. The latter method confirmed our light microscope observations, as VIAAT immunoreactivity could be detected in boutons containing both GABA and glycine in the spinal cord but, most important, VIAAT-labelled boutons containing glycine alone were also frequently detected. In addition, a few glycine-containing boutons devoid of VIAAT immunoreactivity could be found, this finding raising the question of a distinct vesicular inhibitory amino acid transporter in some boutons.

Finally, axon terminals from the basket cells, known to be GABAergic and to contact the axon hillock and initial segment of the Purkinje cell, were found to be immunoreactive for VIAAT in the cerebellum. In conclusion, our study shows that VIAAT is expressed at virtually all inhibitory amino acid-containing terminals independently of their phenotype (GABAergic, glycinergic or mixed), in agreement with our previous in situ hybridization data (Sagné et al., 1997).

The presence of VIAAT in pure glycinergic terminals suggests that it is able to translocate glycine in addition to GABA. Consistently, VIAAT-induced GABA uptake activity appeared to be sensitive to glycine. However, no accumulation of glycine could be detected in an in vitro assay of VIAAT-containing vesicles (McIntire et al., 1997), although a small but significant uptake of glycine was observed in an indirect assay (Sagné et al., 1997). This failure may originate from the sixfold higher K_M of glycine as compared with GABA, and thus the corresponding lower signal-to-noise ratio of glycine uptake. The question of whether GABA and glycine are simultaneously released at mixed synapses was elegantly addressed by a recent electrophysiological study of interneuron-motoneuron synapse (Jonas et al., 1998). Unitary postsynaptic responses were shown to contain two components, distinguished by their kinetic and pharmacological properties: a glycine receptor-mediated, fast-decaying current and a GABA_A receptor-mediated, slow-decaying one. Further on, the two components were also observed at the quanta level, unambiguously demonstrating that GABA and glycine are released from, and thus loaded into, a single synaptic vesicle.

In the cerebellum, the terminal axonal boutons of Golgi cells, which contact dendrites of granule cells in glomerular structures, were found to be labelled for VIAAT. Since Golgi boutons are known to contain both GABA and glycine (Ottersen et al., 1988), the presence of VIAAT suggests that the two neurotransmitters may be co-released at this synapse. As glycinergic inhibitory postsynaptic currents have never been recorded in the postsynaptic granular cells (Kaneda et al., 1995), glycine may interact with a receptor other than its own ionotropic receptor. The N-methyl-D-aspartate (NMDA) receptor, present on granule cell dendrites, is a good candidate (Johnson and Ascher, 1987). This would suggest that the same terminal could be inhibitory on the Golgi-to-granule cell synapse by means of GABAergic transmission and simultaneously potentiate the excitatory mossy fiber-to-granule cell synapse, the latter action being mediated by glycine via a spillover mechanism (Barbour and Hausser, 1997). This hypothesis is substantiated by the fact that membrane glycine transporters, which were shown to regulate the saturation of the glycine-binding site of the NMDA receptor in the hypoglossal nuclei (Berger et al., 1999), are abundant in the glomeruli (Zafra et al., 1995). Hence, in the cerebellum, mixed GABA- and-glycine terminals might have a different physiological significance from that in the spinal cord.

The existence of a common transporter implies that the loading of GABA and/or glycine into the vesicle is controlled by their cytosolic concentrations. This regulation could be achieved by the GABA biosynthetic enzyme (GAD) and the plasma membrane transporters for GABA (GAT1; Guastella et al., 1990) or glycine (GLYT2; Liu et al., 1993). A similar dependence of the nature of the vesicle content on the transmitter biosynthetic and reuptake systems exists for monoamines. Several mechanisms may ensure the preferential vesicular loading of GABA at GABAergic terminals. First, the activity of GAT1 will maintain a high GABA-to-glycine ratio in the cytosol. Second, the lower K_M of GABA and, possibly, the attachment of GAD to synaptic vesicles (Reetz et al., 1991), may favor the interaction of VIAAT with GABA over glycine. At glycinergic terminals, the activity of GLYT2 builds up a high cytosolic concentration of glycine and, depending on whether GAD or GAT1 are co-expressed or not, VIAAT-containing vesicles are expected to load, and release, a mixed or ‘pure’ glycine content. Several mechanisms may regulate the amount of GABA and glycine in the nerve terminal and thus control the intravesicular content of these amino acids, and therefore the kinetics of the inhibitory postsynaptic current following the opening of a given synaptic vesicle.

The authors are indebted to P. Krafft, M. Boström and C. Sagné for their contribution to RT-PCR, pGEX4T1-VIAAT-Nter construction and sequencing, respectively, and to H. Betz and O.P. Ottersen for kindly providing mAb4a and anti-glycine serum. They are grateful to S. Dieudonné for critical reading of the manuscript. This work was supported by the CNRS (ERS 575, to J.P.H.), the European Community (ERB 4061 PL 970277, to J.P.H. and BMH4CT 972374, to A.T.) and the French Spinal Cord Research Institute (IRME, to A.D.).

After submission of the manuscript, a study reaching similar conclusions was published (Chaudhry et al., 1998, J. Neurosci. 18, 9733-9750).