γ-aminobutyric acid (GABA) is the major inhibitory neurotransmitter in the central nervous system and has a widespread distribution in the adult brain. The rapid termination of GABA transmission is achieved through high-affinity GABA transport into both GABAergic neurons and glial cells (Schousboe et al., 1983; Schousboe and Westergaard, 1995).
The cloning of GABA (Guastella et al., 1990) and norepinephrine (Pacholczyk et al., 1991) transporters allowed the subsequent isolation of related cDNAs which constitute the Na+/Cl−-dependent neurotransmitter transporter superfamily. The members of this superfamily share a similar membrane topology arranged in 12 transmembrane domains, N- and C-termini on the cytoplasmic side, and a potential glycosylation sequence between transmembrane helices III and IV (for review, see Kanner, 1994; Nelson and Lill, 1994; Uhl and Johnson, 1994). Molecular cloning studies have revealed the existence of three high-affinity subtypes of GABA transporters in the rat and human brain: GAT1, GAT2, and GAT3 (Borden et al., 1992, 1995a; Guastella et al., 1990), and one of lower affinity, BGT-1 (Yamauchi et al., 1992).
Tissue and Cell Localization
Several studies have addressed the tissue and cellular distribution of GABA transporters. The GABA transporters (GATs) GAT1 and GAT3 are expressed exclusively in the central nervous system (CNS), whereas GAT2 and BGT1 are also present in peripheral tissues, mainly liver and kidney (Clark et al., 1992; Borden et al., 1994; Guastella et al., 1990; Jursky et al., 1994; Liu et al., 1993; Nelson et al., 1990; Rasola et al., 1995; Yamauchi et al., 1992). The presence of GABA uptake systems in glial cells was first demonstrated by autoradiographic studies showing [3H]β-alanine uptake in cortical slices and synaptosome preparations (Schon and Kelly, 1975), as well as in astrocyte primary cultures (Hertz et al., 1978; Balcar et al., 1979); pharmacological studies also showed that GABA transport in cultured astrocytes from different brain regions (Hosli and Hosli, 1979) as well as to membrane vesicles derived from cortical astrocytes, is highly sensitive to β-alanine (Mabjeesh et al., 1992) and other GABA analogues (Schousboe et al., 1978).
GABA transporters do not follow a specific cell type expression pattern. Although GAT1 has long been considered a neuronal GABA transporter (Iversen and Kelly, 1975; Mabjeesh et al., 1992), recent studies have clearly shown its presence in distal astrocytic processes of all the rat brain regions examined so far (cerebral cortex: Minelli et al., 1995; retina: Johnson et al., 1996; hippocampus: Ribak et al., 1996a; cerebellum: Rattray and Priestley, 1993; Ribak et al., 1996b; thalamus: DeBiasi et al., 1998). GAT1 has been also localized to astrocytic processes in human and in monkey cerebral cortex (Conti et al., 1998); therefore, the astrocytic localization of GAT1 seems to be a consistent feature in the mammalian cortex. GAT3 is mainly expressed by glial cells in the brain and the retina, but neuronal expression has also been demonstrated, particularly in the latter (Clark et al., 1992; Durkin et al., 1995; Yang et al., 1997; Johnson et al., 1996; Ribak et al., 1996a; Durkin et al., 1995; Borden et al., 1995a). GAT2 was proposed to have a nutritional role and to perform non-neuronal functions since it was first localized to cells of the leptomeninges, choroid plexus, and ependyma in the brain (Ikegaki et al., 1994; Durkin et al., 1995), and to the pigment and ciliary body epithelia in the retina (Honda et al., 1995; Johnson et al., 1996), but its presence in neurons and glial cells has been further demonstrated (Borden et al., 1995a; Conti et al., 1999; Voutsinos et al., 1998; Obata et al., 1997; Zhao et al., 2000; Redecker, 1999). A role for GAT2 in development has also been proposed, since mRNA for this protein is more abundant in neonatal than in adult mouse brain (Liu et al., 1993). BGT1 is most probably a glial transporter, since its transcripts were observed in type 1 and type 2 astrocyte cultures, but not in neuronal cell cultures (Borden et al., 1995a); its presence in the brain has been related with osmoregulation (Borden et al., 1995b; Lopez-Corcuera et al., 1992; Rasola et al., 1995; Bitoun and Tappaz, 2000).
Presynaptic localization of GAT1 by in situ hybridization and immunocytochemistry in GABAergic neurons has been shown, correlated or not with the presence of GABA and GAD67 (glutamic acid decarboxylase; Radian et al., 1990; Durkin et al., 1995; Augood et al., 1995).
In the cerebral cortex of adult rats, GAT1 exhibits the highest level of expression, followed by GAT3 and GAT2. In addition to neurons, GAT1 mRNA and GAT1 immunoreactivity are also localized to some distal astrocytic processes (Minelli et al., 1995). GAT3 localizes exclusively in astrocytic processes (Minelli et al., 1996), and GAT2 is expressed by epithelial, glial, and neuronal cells (Conti et al., 1999). GAT1 is also expressed robustly in the human and monkey cerebral cortex, where it localizes to both neurons and astrocytic processes near axon terminals forming GABAergic synapses, and scattered in the neuropil (Conti et al., 1998). Morphology and distribution of distal astrocytic processes labeled for GAT2 are similar to those labeled for GAT1 and GAT3 (Minelli et al., 1995, 1996); however, only GAT2 immunoreactivity is present in astrocytic cell bodies and their proximal processes (Conti et al., 1999). GAT2 mRNA is also expressed in vitro by O-2A/type 2 astrocytes, supporting the idea that cortical glial GABA transport is mediated also by GAT2. Although the role of BGT1 in the cortex remains to be established, its presence has been reported in type 1 astrocytic cultures derived from rat brain (Borden et al., 1995a). Rat cortical astrocytes, therefore, express GAT1, GAT2, GAT3, and BGT1. These observations raise several issues regarding the relative contribution of each of these transporters to overall GABA uptake by glial cells in the cortex, and of the functional significance of multiple GABA uptake systems. Based on the distinct distribution and degree of expression of GATs abovementioned, although glial GABA uptake in the cortex seems to be mediated largely by GAT3, GAT1, and GAT2 playing a minor role in the removal of extrasynaptic GABA, because of the differential ionic dependency, inhibitor sensitivity (Guastella et al., 1990; Borden et al., 1992; Clark et al., 1992; Keynan et al., 1992), and modulation of the three high-affinity GATs (Gomeza et al., 1991; Corey et al., 1994; Quick et al., 1997), it is reasonable to think that their relative contribution to glial GABA uptake is dynamically regulated, providing for a great adaptability in the control of extracellular GABA levels.
Astrocytic GATs neighboring GABAergic synapses are placed strategically to take up locally released GABA, thus contributing to the termination of GABA-mediated inhibitory synaptic transmission and to the modulation of the synaptic action of GABA. In contrast, the function of astrocytic GATs, such as GAT1, located extrasynaptically, may be the control of the paracrine spread of GABA to excitatory and inhibitory neighboring synapses (Isaacson et al., 1993; Thomson and Gahwiler, 1992; Rossi and Hamann, 1998). Regarding this suggestion, the label for GAT1 and GAT3 was detected in astrocytic processes enveloping several axon terminals, together with their postsynaptic dendrites in the thalamus (DeBiasi et al., 1998), but not in cerebral cortex (Minelli et al., 1995, 1996) or hippocampus (Ribak et al., 1996a), although in the cerebellum a dense glial labeling for GAT3 envelops Purkinje axon terminals (Itouji et al., 1996; Ribak et al., 1996b). A role for astrocytes in the insulation of synapses is supported by the expression of GABA transporters in thalamic and cerebellar glial processes, but not in GABAergic terminals, indicating the participation of astrocytes in the modulation of GABA transmission in these areas. Astrocytic processes labeled for GAT1 and GAT3 are also found scattered in the neuropil, distant from GABAergic terminals and occasionally neighboring axon terminals from excitatory synapses, where GABA uptake by glia could limit GABA action on distant nonsynaptic GABAA receptors (Spreafico et al., 1993; Alvarez et al., 1996). Moreover, GABA uptake by astrocytes has been proposed to regulate the action of GABA at GABAB presynaptic receptors located on excitatory terminals (Soltesz and Crunelli, 1992; Ulrich and Huguenard, 1996), aimed to inhibit synaptic transmission via G protein-mediated modulation of presynaptic Ca2+ channels (Isaacson, 1998; reviewed in Isaacson, 2000). GABA taken up by glial cells is rapidly metabolized by GABA transaminase (Iversen and Kelly, 1975) which displays high activity in astrocytes. Additionally, GABA transporters in astrocytes can also mediate GABA release (see below; Gallo et al., 1991).
In situ hybridization studies in the cerebellum revealed GAT1 mRNA predominantly localized to the molecular and Purkinje cell layers, whereas GAT3 is found in the deep cerebellar nuclei (Clark et al., 1992; Durkin et al., 1995; Rattray and Priestly, 1993). GAT2 mRNA, first reported exclusively in the leptomeninges, was later found in the cerebellum, predominantly in the granular layer (Voutsinos et al., 1998), in agreement with the localization of its cognate protein (Ikegaki et al., 1994). At the cellular level, GAT1 mRNA was found predominantly in cell bodies of GABAergic basket and stellate neurons; GAT1 mRNA and immunoreactivity were also detected in cell bodies and glial processes ensheathing Purkinje cells somata and dendrites, respectively, (Voutsinos et al., 1998; Morara et al., 1996), which most likely correspond to Bergmann glia (Rattray and Priestly, 1993; Ribak et al., 1996b). GAT3 mRNA is predominantly confined to glial cells (Voutsinos et al., 1998), consistent with immunocytochemical studies localizing this transporter to cerebellar glial processes (Itouji et al., 1996). The dense glial labeling for GAT3 surrounding GABAergic Purkinje axon terminals seems to compensate for the apparent lack of GABA transporters in these neurons (Ribak et al., 1996b). In the developing rat cerebellar cortex, GAT3 immunoreactivity appears initially in somata and primary processes of postatal day (P)7–21 astrocytes, and is later identified in distal processes in the adult (Yan and Ribak, 1998). GABA acting as an excitatory transmitter plays an important neurotrophic role in brain development (Cherubini et al., 1991; Staley et al., 1995); hence, the early expression of GATs in astrocytes suggests their involvement in the differentiation and maturation of developing neurons, probably through the release of GABA.
BGT1 was originally cloned from Madin-Darby canine kidney cells (Yamauchi et al., 1992), and subsequently its expression has been demonstrated in most mammalian tissues including the CNS (Rasola et al., 1995; Borden et al., 1995b, 1996). BGT1 mRNA has been detected in all mouse and human brain regions (López-Corcuera et al., 1992; Rasola et al., 1995). Since BGT1 transcripts were identified in type 1 and 2 astrocytes, but not in neurons in culture, BGT1 was proposed to be mainly glial, although its presence in glial cells from rat brain slices has not been shown (Borden et al., 1995b), and pharmacological data indicate that BGT1 makes only a minor contribution to GABA transport in these cells. The distribution of BGT1 mRNA in the brain does not correlate with GABAergic pathways, discarding a role in the termination of GABA transmission; BGT1 could, however, be involved in the removal of GABA diffused from synaptic regions. On the other hand, BGT1 might contribute to volume regulation in the CNS (Yamauchi et al., 1992; Borden et al., 1995b) since betaine, which has been assigned a role in osmoregulation, is a substrate for BGT1 (Heilig et al., 1989). A recent study on this matter (Bitoun and Tappaz, 2000) demonstrates a significant increase in mRNA levels of BGT1 in cultured cortical astrocytes exposed to hyperosmotic conditions; increased transcription of BGT1 gene is likely mediated by the osmotic responsive element (ORE) recently identified in the 5' flanking region of the BGT1 gene (Miyakawa et al., 1998).
Immunocytochemical studies in the mammalian retina using specific antibodies for the distinct GABA transporters have shown GAT1 localized mainly to neurons including amacrine, displaced amacrine, interplexiform, and ganglion cell processes throughout the inner plexiform layer (IPL), and to lower extent in Müller cells (Brecha and Weigmann, 1994; Ruiz et al., 1994; Durkin et al., 1995; Johnson et al., 1996). Immunoreactivity for GAT3 is expressed mainly in Müller cells (Brecha and Weigmann, 1994; Brecha et al., 1995; Honda et al., 1995) and amacrine cells at the inner nuclear layer (INL; Brecha et al., 1995; Johnson et al., 1996), whereas GAT2 immunostaining was found in the pigment and ciliary epithelia in the mammalian retina (Honda et al., 1995; Johnson et al., 1996).
In contrast to mammalian retinae, Müller cells in salamander and in most of the ectotherms do not take up GABA (for review see Marc, 1992). In the tiger salamander retina, GAT1 immunoreactivity was found in bipolar, amacrine, and interplexiform cells as well as in the ganglion cell layer. No detectable staining was found in horizontal cells or in structures resembling Müller cells. GAT3 antibodies labeled fewer cells and cell types than GAT1 antibodies, localized to amacrine cells and cells in the ganglion cell layer, but not horizontal cells, bipolar cells, or Müller cell like-structures (Yang et al., 1997). Retinal horizontal cells of ectotherms are GABAergic (for review see Marc, 1992; Wu, 1992) and bear electrogenic GABA transporters (Malchow and Ripps, 1990; Cammack and Schwartz, 1993; Takahashi et al., 1995). Since GAT1 and GAT3 antisera did not detectably label horizontal cells, the presence of an unidentified GABA transporter in these species is suggested. Similar results were obtained in the salmon retina, where GAT1 immunoreactivity was present in amacrine cells and the IPL, but not in Müller cells; as opposed to the salamander retina, however, bipolar cells were not labeled in this species (Ekström and Anzelius, 1998).
Not all nonmammalian Müller cells lack the ability to take up GABA. In contrast with the abovementioned species, bullfrog Müller cells strongly express GAT1 and GAT2, but not GAT3. Somata, major processes, endfeet, and branchlets of most Müller cells expressed GAT1, whereas a moderate labeling for GAT2 was observed in main trunks and endfeet of 80–90% of these cells (Zhao et al., 2000).
Available evidence suggests that GABAergic transmission in the retina could be modulated through the uptake of GABA, not only by retinal neuronal elements but also by Müller cells in both mammalians and nonmammalians. Moreover, GATs in Müller cells could play a protective role from excessive GABA inhibition during physiological and/or pathological events.
Also regarding the visual system, the mRNAs encoding for GAT1, GAT2, and GAT3 are expressed in the optic nerve of both neonatal and adult rats (Howd et al., 1997). Since optic nerves contain mainly axons and glia, GAT1 and GAT3 mRNAs identified are likely of glial origin. It is possible, however, that axons may contain GAT1 mRNA (Gioio et al., 1994), which is predominantly expressed by neurons, while the expression of GAT2 mRNA probably corresponds to fragments of pia-arachnoid present in the tissue.
Stoichiometry and Channel Properties
GABA transporters belong to the family of Na+- and Cl−-coupled neurotransmitter transporters (for reviews, see Nelson and Lill, 1994; Uhl and Johnson, 1994). Studies on GABA transport in diverse expression systems revealed GAT1 to be strictly Na+-dependent but only partially Cl−-dependent (Mager et al., 1993; Lu et al., 1995). GAT1 expressed in mouse LtK− cells shows a calculated stoichiometry of two Na+, one Cl−, and one GABA molecule per cycle (Keynan et al., 1992), in agreement with previous data obtained in synaptic plasma membrane vesicles and for the purified transporter (Kanner, 1983; Keynan and Kanner, 1988; Radian and Kanner, 1983), as well as in electrophysiological studies (Kavanaugh et al., 1992; Mager et al., 1993, 1996; Risso et al., 1996). The cotransport of GABA with three Na+ and one or two Cl− by BGT1 has been demonstrated recently (Matskevitch et al., 1999). Importantly, the predicted amount of charge crossing the membrane during the transport cycle, based on stoichiometric ion fluxes, is smaller than the charge movement experimentally determined; thus, GABA transporters such as those for glutamate (Glu), exhibit ligand-gated ion channel properties (Mager et al., 1996; Cammack and Schwartz, 1994, 1996; reviewed in Sonders and Amara, 1996). GABA transporters also exhibit substrate-independent leak currents carried by Li+ and K+ (Mager et al., 1993; Cammack and Schwartz, 1996), blocked by the substrate and by transport inhibitors (reviewed in Sonders and Amara, 1996). These characteristics suggest that GABA transporters, in addition to a role in the termination of synaptic GABA transmission, can also influence neuronal and glial excitability. On this line, a raise in intracellular calcium due to the opening of L-type calcium channels, and a subsequent calcium-induced calcium release from intracellular stores induced by GAT-mediated depolarization, has been documented (Haugh-Scheidt et al., 1995).
Reverse operation of GATs can result in the nonvesicular release of GABA by breakdown of the sodium/chloride/GABA gradient or by cell depolarization (Attwell et al., 1993). Such nonvesicular release has been identified in both neurons (Schwartz, 1987; Yang et al., 1999) and glia (Gallo et al., 1991). GABA is released from cultured striatal neurons by high [K+], veratridine, and Glu receptor agonists in the absence of calcium, or in the presence of tetanus toxin (Pin and Bockaert, 1989). In cultured hippocampal neurons, large GABAA receptor-mediated responses have been observed as a result of carrier-mediated GABA release from neuronal and/or glial neighboring cells, induced by brief alteration of Na+ or K+ electrochemical gradient (Gaspary et al., 1998). Since carrier-mediated release of GABA does not rely on ATP but on ion gradients, it may offer advantages during periods of high energy utilization, such as burst firing or seizures.
Regulation of GABA Transporters
The existence of several types of GABA transporters opens the possibility of distinct regulatory systems according to the phenotype and/or the anatomical localization of the cells expressing these transporters. Short-term regulation of neurotransmitter transporters involving phosphorylation/dephosphorylation has been studied. The inhibition of high-affinity GABA uptake in glial cells by phorbol ester activation of PKC (protein kinase C), due to a decrease in affinity, has been reported (Gomeza et al., 1991). Supported by the presence of multiple consensus sites for PKC phosphorylation on GABA transporters, these data suggested the regulation of GABA uptake by direct transporter phosphorylation, although removal of the PKC consensus sites from GAT1 failed to eliminate PKC-induced inhibition in oocytes (Corey et al., 1994). More recent studies on this matter suggest that GABA transporter function could be regulated by components of the vesicle docking and fusion machinery. In support of this idea, injection of antisense oligonucleotides directed to synaptophysin or syntaxin into oocytes expressing total rat brain mRNA, as well as inactivation of these proteins by botulinum toxins (BTXs), eliminates the regulation of GAT1 by PKC (Quick et al., 1997). Later studies have shown PKC to regulate the interaction between GAT1 and syntaxin 1A in neurons endogenously expressing all three proteins, provided that Munc18, a substrate for PKC phosphorylation, is also present (Beckman et al., 1998). Although modulation of GABA transport by PMA (phorbol 12-myristate 13-acetate) and BTX was not observed in cultured astrocytes, the presence of SNARE (soluble NSF receptors) complex proteins in glial cells including syntaxin, synaptobrevin, and SNAP-23 has recently been reported (Araque et al., 2000; Hepp et al., 1999; Madison et al., 1999). The signal which triggers PKC-mediated regulation of GAT1 is still unknown; however, specific agonists of G-protein-coupled receptors for serotonin, acetylcholine, and Glu downregulate GAT1 function in neurons (Beckman et al., 1999). Such functional inhibition has been ascribed to the redistribution of the transporter from the plasmamembrane to intracellular locations; moreover, BTX prevents the receptor-mediated inhibition, suggesting the involvement of syntaxin 1A. Extracellular GABA also regulates GAT1 by increasing transporter expression in the plasmamembrane of hippocampal neurons. Hippocampal astrocyte cultures exposed to GABA also show a marked increase in GABA uptake, although the mechanism underlying this effect has not been studied (Bernstein and Quick, 1999).
GABA uptake by cerebellar glia is stimulated in vitro by adrenaline (Hansson and Ronnback, 1991) or by GABA-CIP (GABA-carrier inducing protein), a protein released by granule cells (Nissen et al., 1992), and inhibitory control of glial GABA uptake by serotonin has been demonstrated in ependymocytes of the subcomissural organ in vivo (SCO; Didier-Bazes et al., 1989, 1992). A direct serotonergic control of glial GABA uptake has been further demonstrated in vitro, since serotonin stimulated the activity and mRNA expression of GABA transporters in cerebellar astrocyte cultures. Serotonin regulation of glial GABA transport might be involved in some pharmacological effects of serotonergic drugs, i.e., antidepressants, known to increase extracellular serotonin levels (Voutsinos et al., 1998).
Glial GABA Transporters in Disease
A decrease in GABAergic neurotransmission has been implicated in the pathophysiology of several CNS disorders, particularly in epilepsy. Consequently, research in this area has focused on the development of pharmacological agents capable of increasing GABAergic function. Intracerebroventricular application of the glia-selective GABA transport inhibitors 4,5,6,7-tetrahydroisoxazole [4,5-c]pyridin-3-ol (THPO), and 5,6,7,8-tetrahydro-4H-isoxazolo(4,5-c)azepin-3-ol (THAO), has been shown to protect against seizures induced by drugs known to impair GABAergic neurotransmission (Krogsgaard-Larsen et al., 1987); in contrast, the inhibition of neuronal GABA uptake by L-DABA (L-2,4-diamino butyric acid) results in proconvulsant behavior (Gonsalves et al., 1989). These data have led to the proposal that selective block of glial GABA transport elevates GABA concentration in nerve terminals, whereas the selective blockage of the neuronal transporter depletes the releasable neurotransmitter pool (Wood et al., 1980; Schousboe et al., 1983). However, the lipophilic derivatives of piperidencarboxylic acid (tiagabine, SKF-89976A, CI-966, and NNC-711), highly selective for GAT1, have been reported to exhibit anticonvulsant properties (Borden et al., 1994; Clark et al., 1992; Suzdak et al., 1992). It is evident that no simple correlation exists between the pharmacological characteristics of GABA transport mediated by the cloned carriers and that of neuronal and glial GABA uptake (Schousboe and Westergaard, 1993; Borden, 1996). Although the participation of neuronal and glial GABA transporters in seizure activity has not been fully elucidated, a role for GABA has been proposed, and impaired GABA release due to a decrease in GATs has been observed (During et al., 1995). However, increases in extracellular GABA concentration during kindling (Ueda and Tsuru, 1995) and human seizures (During et al., 1995) could relate to nonvesicular GABA release due to the upregulation of GABA transporters (Hirao et al., 1998). Increases in the expression of GAT1 and GAT3 following mechanical and chemical lesions in rat cerebral cortex and FeCl3 treatment in amygdala have been shown to propagate to the contralateral side, suggesting the existence of transcellular signaling mechanisms regulating GATs expression in the cerebral cortex, probably as a protective mechanism (Yan and Ribak, 1999; Ueda and Willmore, 2000a). On the other hand, the upregulation of GATs expression in epilepsy may result in lowered GABAergic transmission and therefore contribute to the generation of seizures. Actually, one of the effects of the anitiepileptic drug, valproate, is the downregulation of neuronal and glial GAT1 and GAT3 protein expression, which results in an increased extracelluar GABA concentration (Ueda and Willmore, 2000b).
In addition to epilepsy, existing evidence supports a role for GABA transporters in several clinically related processes. Demonstration of the inhibitory action of some anesthetics on GABA uptake into striatal synaptosomes, suggests the involvement of this mechanism in the effect shown by these agents (Mantz et al., 1995). Also, postmortem analysis of schizophrenic brains has revealed a decreased number of GABA uptake sites in subcortical regions, which reveals GABAergic mechanisms are abnormal in schizophrenia (Simpson et al., 1992). Further studies are required to elucidate the participation of glial GABA uptake in these clinical conditions.
Recent work demonstrated the modulation of seizure development in a kindling model of epilepsy by transplantation of immortalized mouse neurons and glial cells genetically engineered to produce GABA by driving GAD (glutamic acid decarboxylase) expression (Thomson et al., 2000). Although the mechanism of GABA release was not investigated, it is interesting to speculate that GABA transporters could be involved.
Although recent information supports a role for glial GATs in the modulation of inhibitory neurotransmission in the CNS and the retina, further research is needed in order to establish their precise function.