Presynaptic, extrasynaptic and axonal GABA_A receptors in the CNS: where and why?

Abstract

Although GABA_A receptors are widely distributed at inhibitory synapses on dendrites and cell bodies of neurons, they also occur in other places, in particular at synapses made on axons and in extrasynaptic membranes. This review summarises some of the evidence that presynaptic receptors modulate transmission not only at primary afferents in the spinal cord, but also at a variety of sites in the brain, including hippocampal mossy fibres. These receptors modulate transmitter release via several different mechanisms. Another form of unconventional GABA_A receptor-mediated signalling is the mediation of a tonic conductance, seen in granule cells of the cerebellum and dentate gyrus and also in hippocampal interneurons. Tonic signalling appears to be mediated by extrasynaptic receptors. The adaptive significance of this form of signalling remains poorly understood.

1. Introduction

GABA_A receptors are widely distributed in the mammalian CNS, where they are conventionally thought to mediate fast synaptic inhibition. The textbook model of inhibition holds that GABA_A receptors at synapses on cell bodies or dendrites open an anion conductance, which results in both postsynaptic hyperpolarisation (mediated by Cl⁻ influx) and shunting of excitatory currents. This leads to a transient decrease in the probability of initiation of an action potential.

This model may not always be correct because GABA_A receptors can sometimes depolarise neurons, and possibly even have a net excitatory action: the reversal potential of the mixed Cl⁻/HCO₃⁻ conductance is frequently more positive than the resting membrane potential (Cherubini et al., 1991), especially in some immature neurons prior to the expression of the main Cl⁻ extrusion system, the KCC2 transporter (Rivera et al., 1999). Under some circumstances, GABA_A receptor activation can result in the opening of voltage-gated Ca²⁺ channels, and even in the initiation of action potentials.

Another important question surrounding the role of GABA_A receptors is to what extent they occur at sites other than postsynaptic elements of inhibitory synapses on cell bodies and dendrites. Other locations, such as on presynaptic boutons, extrasynaptic membranes on postsynaptic neurons or even on axons of the white matter, have been demonstrated with a variety of methods. The function and adaptive significance of these receptors remain incompletely understood. This brief review attempts to summarise some of the evidence that such receptors are abundantly expressed in some parts of the CNS, to examine what effects these receptors have on neuronal signalling, and to speculate about their possible adaptive significance.

The broader subject of presynaptic ionotropic receptor function in the CNS has been considered recently by Engelman and MacDermott (2004). The role of tonically active GABA_A receptors has been considered by Semyanov et al. (2004).

2. Presynaptic GABA_A receptors in the CNS

2.1. Primary afferent depolarisation and presynaptic inhibition

That GABA receptors can occur on presynaptic structures was suggested early by work on the pharmacological sensitivity of presynaptic inhibition in the cat spinal cord. Eccles et al. (1963) reported that picrotoxin (but not the glycine receptor antagonist strychnine) reduced both presynaptic inhibition of spinal monosynaptic reflexes and the associated phenomenon of primary afferent depolarisation (Frank and Fuortes, 1957), which is thought to reflect depolarisation of afferent axon terminals. This led to the proposal that GABA was a depolarising transmitter at axo-axonic synapses. Because picrotoxin is inactive at metabotropic GABA_B receptors, which were subsequently described, this evidence points to ionotropic (GABA_A) receptors. Further evidence linking GABA_A receptors to presynaptic inhibition came from a demonstration that iontophoresis of GABA could evoke primary afferent depolarisation, and that this phenomenon could be blocked by picrotoxin and bicuculline (Rudomin et al., 1981).

The fact that presynaptic inhibition was associated with a depolarisation of primary afferent terminals led to the hypothesis that the reversal potential for GABA, E_GABA, is relatively depolarised. A more direct demonstration of such a phenomenon came from iontophoretic applications of GABA with or without a GABA_A antagonist, close to the site of antidromic action potential initiation (Curtis and Lodge, 1982).

A wealth of evidence has since supported the hypothesis that a major (although probably not unique (Rudomin, 2000)) mechanism of presynaptic inhibition of afferent signalling in the mammalian spinal cord is via a depolarisation of terminals by GABA_A receptors. Ultrastructural studies have revealed an anatomical counterpart for this phenomenon: axo-axonic synapses are seen on primary afferents in lamina VI (see, for instance, Maxwell and Bannatyne, 1983). Surprisingly, glycine immunoreactivity is sometimes seen in the boutons presynaptic to afferent axons, albeit at a lower level than GABA immunoreactivity (Watson and Bazzaz, 2001). A role for glycine in presynaptic inhibition has not been reported. Although metabotropic GABA_B receptors play a role in presynaptic inhibition in the spinal cord (Stuart and Redman, 1992), they do not appear to contribute to primary afferent depolarisation, and will not be discussed further.

How does GABA_A receptor-mediated depolarisation of afferents affect transmitter release from primary afferents? Although there is general agreement that the probability of transmitter release is decreased (Clements et al., 1987), how this comes about is not entirely clear. Two main hypotheses have received most attention. First, depolarisation could affect transmitter release via an action on other ion channels at or very close to the release site: inactivation of Na⁺ and/or Ca²⁺ channels could reduce the amplitude of the action potential and decreased Ca²⁺ influx would be expected to reduce the probability of exocytosis of neurotransmitter. Depolarisation could, moreover, decrease the driving force for Ca²⁺ influx immediately after the action potential, further contributing to the reduction in exocytosis. The main competing hypothesis is that opening of GABA_A receptors more remote from the release sites interferes with the propagation of the action potential into the presynaptic terminals of the afferent axon by causing a decrease in membrane resistivity (Segev, 1990). Although primarily based on computer simulations, this shunting hypothesis has been supported by experimental evidence from large unmyelinated axons in crayfish (Cattaert and El Manira, 1999). Axons in the mammalian spinal cord are difficult to record from, but Verdier et al. (2003) have recently reported that GABA_A receptors can affect action potential propagation in branches of afferents in brainstem slices.

The technical difficulty of recording from small, frequently unmyelinated, terminals has limited the interpretation of computer simulations, because they are critically dependent on assumptions about the GABA_A reversal potential, synaptic conductance, resting membrane potential of afferent arborisations, and kinetics of Na⁺ channels among other parameters. Nevertheless, Graham and Redman (1994) have applied a simulation approach to argue in favour of inactivation of Na⁺ channels as the predominant mechanism of presynaptic inhibition, rather than shunting, on the basis that in order to prevent action potential conduction, an unrealistically high steady-state GABA_A conductance would be required.

An alternative approach to choose between these hypotheses is to examine where axo-axonic synapses occur in relation to boutons in synaptic contact with dendrites of target neurons, or branch points where the safety factor for action potential propagation is expected to be lower than on the shaft of the axon. Evidence for both types of spatial arrangements has been found (Walmsley et al., 1995): although axo-axonic synapses can occur in association with a branch point, a more common arrangement appears to be a ‘triad’ composed of an afferent bouton, the dendrite of a target cell and a presynaptic GABAergic bouton in contact with the afferent bouton. This provides circumstantial evidence to suggest that action potential propagation failure may not be the main mechanism of presynaptic inhibition in the mammalian spinal cord.

2.2. Presynaptic GABA_A receptors in other subcortical structures: posterior pituitary, brainstem, cerebellum, retina and hypothalamus

The mechanisms of presynaptic GABA_A receptor-mediated inhibition in the spinal cord remain incompletely resolved. However, a similar phenomenon has been identified in another part of the CNS, which provides opportunities for far more direct experimental investigation. The secretory endings of hypothalamic neurons projecting to the posterior pituitary express GABA_A receptors, and activation of these receptors decreases the release of oxytocin and arginine vasopressin and other hormones (Saridaki et al., 1989; Zhang and Jackson, 1993). Direct recordings from terminals in vitro have shown that activation of GABA_A receptors can prevent the propagation of action potentials (Zhang and Jackson, 1995). However, this does not appear to be primarily because of shunting of the depolarising wave-front, but because of inactivation of Na⁺ channels (Jackson and Zhang, 1995).

The conclusions derived from posterior pituitary recordings do not necessarily apply to other sites in the CNS where presynaptic GABA_A receptors occur. Turecek and Trussell (2002) recorded from the calyceal synapse in the medial nucleus of the trapezoid body in the auditory brainstem. Activation of presynaptic GABA_A receptors (or of glycine receptors that are also expressed) depolarised the presynaptic terminals and enhanced glutamate release, which the authors interpreted as being mediated by an increase in basal intracellular [Ca²⁺] and/or an increase in Ca²⁺ influx (Turecek and Trussell, 2001). There are several possible explanations for the striking contrast between the effect of presynaptic GABA_A receptors at this synapse and in the hypothalamus and spinal cord, including differences in the properties of voltage-gated Ca²⁺ channels, Ca²⁺ buffers and extrusion systems and Ca²⁺ sensors responsible for exocytosis.

Presynaptic GABA_A receptors have also been studied in inhibitory interneurons of the cerebellar cortex. Recently, Pouzat and Marty (1999) used an ingenious method to argue that, in the developing rat cerebellum, stellate and basket cell axons have a high density of GABAergic autoreceptors. They evoked an escape action current by delivering a brief depolarising step command via a somatic patch-pipette. Immediately following this current they observed a relatively slow current that had a pharmacological profile typical of GABA_A receptors. They argued that this was not an autaptic response (arising from synapses made by axon branches back onto the same cell’s dendrites) on the grounds that the kinetics of the current were very different from those of conventional IPSCs. Moreover, with a high intracellular Cl⁻ solution, the current sometimes evoked regenerative responses, consistent with a location on the axon. It is unclear whether this phenomenon arises from activation of GABA_A receptors at the presynaptic boutons, or whether GABA spills out of the synaptic cleft to reach more remote receptors, which could be diffusely distributed in the axon or even clustered at axo-axonic synapses. Moreover, what effect these receptors have on transmitter release from axonal varicosities has yet to be explored.

Presynaptic ionotropic GABA receptors have also been reported in the retina (Tachibana and Kaneko, 1987; Lukasiewicz and Werblin, 1994). A large fraction of the receptors that modulate transmission from bipolar to ganglion and amacrine cells are not strictly speaking GABA_A receptors, because they are resistant to bicuculline. Instead, they have the pharmacology of GABA_C receptors composed of ρ subunits (Qian and Ripps, 2001).

Presynaptic GABA_A receptors have also been described in the ventromedial hypothalamus (Jang et al., 2001), although relatively little is know about their effects on action potential-evoked transmitter release.

2.3. Presynaptic GABA_A receptors in hippocampal mossy fibres

In contrast to the abundant evidence for presynaptic GABA_A receptors in a variety of subcortical structures, it is far from clear to what extent they play a role in modulating transmission in the cerebral cortex. A presynaptic role in the cerebral cortex would have extensive implications for higher information processing.

Axo-axonic GABAergic synapses are widely distributed at the axon initial segments of a wide variety of neurons. This is a strategically important location because it is the most important site of initiation of action potentials, and expresses a high density of Na⁺ channels. Chandelier cells in neocortex project preferentially to this site, implying that they may have a special function in temporally precise modulation of the output of principal neurons (Somogyi et al., 1985). However, GABA_A receptors at axo-axonic synapses are arguably postsynaptic receptors that happen to occur on the axon, rather than truly presynaptic, associated with transmitter release sites, and will therefore not be considered further.

GABA_A receptors have been reported to modulate the release of neurotransmitters from synaptosomes (Fung and Fillenz, 1983; Fassio et al., 1999), and seizure-like activity in vitro can initiate ectopic axonal action potentials via GABA_A receptors in Schaffer collaterals of hippocampal CA1 pyramidal neurons (Stasheff et al., 1993). Both these phenomena are, however, open to alternative interpretations. In particular, GABA_A receptor-mediated responses can be accompanied by efflux of HCO₃⁻ and K⁺ (Voipio and Kaila, 2000). Indeed, extracellular K⁺ accumulation has been proposed to account for GABA_A receptor-dependent initiation of ectopic action potentials during epileptiform discharges (Avoli et al., 1998). Thus, there is little compelling evidence to date that presynaptic GABA_A receptors can directly affect transmitter release at cortical synapses.

Part of the difficulty of studying cortical neurons is that their dendritic and axonal arborisations frequently overlap spatially. Thus, in contrast to spinal afferent axons or neurosecretory axons, for instance, it is difficult to be certain that stimuli delivered in vivo or in vitro are restricted either to dendrites or to cell bodies. Among several exceptions to this rule are granule cells of the dentate gyrus, which project their axons, mossy fibres, to stratum lucidum of the hippocampus proper, but whose dendrites are confined to the molecular layer of the dentate gyrus. We have recently obtained evidence for presynaptic or axonal GABA_A receptors in these axons (Ruiz et al., 2003). Antibodies against the α₂ subunit of GABA_A receptors labelled mossy fibre boutons as well as extrajunctional membranes. To confirm that such receptors can modulate the excitability of mossy fibres, we delivered extracellular stimuli to stratum lucidum while recording from granule cells in voltage clamp mode (Fig. 1). The GABA_A receptor agonist muscimol reduced the success rate for evoking an antidromic action potential when a threshold-straddling stimulus was used. The opposite effect was obtained when the GABA_A receptor antagonist gabazine (SR95531) was applied, implying that the receptors are tonically active. The effects of manipulating GABA_A receptors did not depend on changes in the ionic environment, because experimentally increasing the intracellular [Cl⁻] in the recorded cell converted the effect of muscimol from decreasing axon excitability to increasing it. Stimuli delivered through a second electrode, designed to activate GABAergic axons in the vicinity, mimicked the effect of exogenous muscimol, implying that GABA_A activation potentially has a physiological role in modulating transmission from the dentate gyrus to the hippocampus proper.

Fig. 1.

Evidence for GABA_A receptors modulating mossy fibre excitability. Top: granule cells were recorded in cell-attached mode in acute rodent hippocampal slices, and a threshold-straddling stimulus was applied via an electrode positioned in stratum lucidum. This resulted in intermittent antidromic action potentials, recorded as action currents (AC). Local pressure application of the GABA_A agonist muscimol close to the stimulation site decreased the axon excitability (summary plot showing mean AC success rate and SEM for four cells). Bottom: following ‘break-in’ with a pipette containing a high [Cl⁻], designed to depolarise the GABA_A reversal potential, the same application of muscimol caused an increase in axon excitability (30 min were allowed to elapse following break-in to allow Cl⁻ to equilibrate). The individual sweeps on the right were obtained from one example cell, showing intermittent failures to evoke an AC. Reproduced from Ruiz et al. (2003).

Importantly, we obtained no evidence for action potential conduction failure because muscimol was active when pressure-applied locally to the site of stimulation, but bath perfusion of muscimol was ineffective with supramaximal stimulation. This agrees with evidence that action potential propagation in the axons of cortical principal cells in the hippocampus has a high safety factor (Cox et al., 2000; Koester and Sakmann, 2000; Emptage et al., 2001). Thus, if presynaptic or axonal GABA_A receptors affect orthodromic transmission, they are unlikely to do so by preventing invasion of distal boutons, although this does not rule out an effect on invasion of filopodial side-branches of mossy fibres that synapse with interneurons (Acsady et al., 1998).

2.4. Actions of GABA_A receptors on mossy fibre Ca²⁺ signalling

A technical limitation of studying mossy fibre transmission is that the connectivity between individual granule cells and postsynaptic CA3 pyramidal neurons is extremely low. Because manipulating GABA_A receptors may have effects both on the recruitment of axons with extracellular stimuli, and on transmitter release, this makes it difficult to determine directly whether the receptors depress or enhance transmitter release. Instead, we imaged fast action potential-induced Ca²⁺ influx at axonal varicosities of mossy fibres while voltage clamping the granule cell, and examined the effect of activating axonal GABA_A receptors with muscimol or blocking tonically active receptors with gabazine (Ruiz et al., 2003). GABA_A receptor activation both increased the resting Ca²⁺-dependent fluorescence and decreased the incremental fluorescence upon action potential propagation. When the incremental fluorescence was corrected for the increased baseline occupancy of the indicator, the results still pointed to a net decrease in action potential-evoked Ca²⁺ influx. Surprisingly, blocking tonically active GABA_A receptors also decreased the action potential-induced Ca²⁺-dependent fluorescence transient, although without an effect on the baseline fluorescence. This paradoxical result (decrease in action potential-induced Ca²⁺ influx with either an increase or a decrease in GABA_A receptor activation) can be explained by postulating that there is a biphasic relationship between Ca²⁺ influx at an axonal varicosity and its membrane potential.

We tested this interpretation by deliberately manipulating the granule cell voltage, while measuring the Ca²⁺-dependent fluorescence at an axonal varicosity in the proximal part of the mossy fibre. We argued that the somatic membrane potential would propagate electrotonically into the proximal axon, so that experimental de- or hyper-polarisation would be sensed at least qualitatively in the imaged varicosities. Depolarising the granule cell led to an increase in baseline fluorescence and a decrease in action potential-induced fluorescence increment, qualitatively identical to the effect of muscimol. Hyperpolarising the granule cell decreased the incremental fluorescence with no effect on baseline Ca²⁺, similar to the effect of gabazine. These results thus support the hypothesis that the Ca²⁺ influx into an axonal varicosity is related to its membrane potential in a biphasic manner, and that either activating or blocking GABA_A receptors can push the varicosity away from a state that corresponds to a near-maximum Ca²⁺ influx.

Thus, we conclude that GABA_A receptors depolarise mossy fibres, which is consistent with evidence that the GABA_A reversal potential in granule cells is consistently positive to their generally very negative resting membrane potential (Misgeld et al., 1986). This gives rise to another puzzle: if GABA_A receptors normally depolarise mossy fibres, why do they also elevate their threshold for antidromic action potential initiation? A possible resolution of this paradox is that opening GABA_A receptors also has a shunting effect on the tissue surrounding the stimulating electrode, an effect which outweighs the effect of depolarisation on axonal excitability when the intracellular [Cl⁻] is low. This shunting effect is presumably insignificant when a much larger depolarisation is obtained by activating GABA receptors in the presence of a high intracellular [Cl⁻].

A limitation of the imaging study is that the granule cell voltage and its intracellular ionic composition were experimentally perturbed, and that only proximal varicosities were studied. Nevertheless, the biphasic relationship between Ca²⁺ influx and axonal voltage is consistent with evidence that gradual depolarisation of mossy fibres with increasing extracellular [K⁺], or by activating kainate receptors, first enhances orthodromic transmission to CA3 pyramidal neurons and then depresses it (Schmitz et al., 2001).

What are the possible mechanisms of the changes in basal and action potential-dependent intracellular Ca²⁺ upon changes in membrane potential? Depolarisation may lead to a decrease in Ca²⁺ influx because of inactivation of Na⁺ and/or Ca²⁺ channels, similar to what has been postulated to occur in the spinal cord and posterior pituitary. However, the accompanying increase in baseline fluorescence suggests that depolarisation may also lead to the accumulation of intracellular Ca²⁺, possibly because of activation of low-threshold Ca²⁺ channels, or because of impairment of Ca²⁺ extrusion mechanisms. Which, if any, of these mechanisms explains the change in baseline fluorescence remains to be established.

What about the decrease in action potential-evoked Ca²⁺ influx with hyperpolarisation? The explanation for this is not obvious, but a possibility is that hyperpolarisation leads to deinactivation of K⁺ channels, leading to a narrowing of the action potential waveform.

Many questions remain to be addressed regarding the consequences of GABA_A receptor activation for Ca²⁺ kinetics in mossy fibre boutons, and downstream events such as release of transmitter onto target neurons. It will be important to see whether the novel phenomena uncovered at mossy fibres also apply to other sites in the CNS where presynaptic GABA_A receptors occur. Whether mossy fibres are unique among cortical axons with respect to presynaptic actions of GABA cannot be answered at present.

2.5. Adaptive significance of presynaptic GABA_A receptors

What is the adaptive significance of presynaptic GABA_A receptor-mediated effects on transmitter release? This question has been addressed extensively in the spinal cord, where evidence now exists for different segmental and descending systems inhibiting interneurons that mediate primary afferent depolarisation (and hence are implicated in presynaptic inhibition, Rudomin and Schmidt, 1999). Not only can presynaptic inhibition be modulated, but there is extensive heterogeneity in the degree to which different target afferent axons are subject to this form of gating. Briefly, it is most marked for large diameter afferents, but anatomically closely related axons can be differentially affected, and possibly also distinct collaterals of individual afferents. Presynaptic inhibition has also been shown to be modulated in man (Iles, 1996). In contrast to the wealth of data on presynaptic GABA_A receptor function in the spinal cord, much less is known of the role of the phenomenon elsewhere. It could represent a form of gating or of lateral inhibition. A feedback function from amacrine to bipolar cells has been suggested in the retina (Tachibana and Kaneko, 1987). As for mossy fibres, we showed that stimuli designed to release GABA from neighbouring structures could modulate the excitability of a mossy fibre without directly activating it, implying that GABA spillover occurs (Ruiz et al., 2003). However, considerable circumstantial evidence exists that mossy fibres themselves can release GABA (Gutierrez, 2000; Walker et al., 2001; Bergersen et al., 2003). Whether GABA released from a single mossy fibre can act on autoreceptors, as suggested to occur in cerebellar basket and stellate cells (Pouzat and Marty, 1999), remains to be determined.

3. Extrasynaptic GABA_A receptors

3.1. Tonic GABAergic signalling in cerebellar granule cells

Another unconventional site for GABA_A receptors is in extrajunctional membranes of the somatodendritic compartments of neurons. Amino acid receptors for either glutamate or GABA are generally highly concentrated at synapses, as expected from the view that such receptors are designed to signal with high temporal and spatial precision. Where such receptors occur outside synapses, conventional explanations are that they act as a reserve pool (for instance for the postsynaptic expression of long-term potentiation), that they are en route to clustering at nascent synapses, or simply that there is some inefficiency in the mechanisms that underlie their synaptic clustering.

A striking observation that, instead, argues that extrajunctional receptors may actually have a separate function, is the finding that cerebellar granule cells express a form of tonic GABA_A receptor-mediated signalling that appears to be mediated by extrasynaptic receptors. This accounts for a large tonic conductance, which increases in magnitude during development, possibly partly reflecting an increase in the degree to which GABA is trapped within glomeruli (Brickley et al., 1996; Wall and Usowicz, 1997). Thus, the tonic GABA_A receptor-mediated signalling does not appear to be a developmental artefact. Evidence that it originates from activation of extrajunctional receptors comes from the fact that α₆ and δ subunits, which contribute to non-desensitising high affinity receptors (Saxena and Macdonald, 1996), occur in the extrasynaptic membrane (Nusser et al., 1998).

Among other puzzling aspects of tonic GABA_A receptor-mediated signalling in this structure is that much of the GABA appears to originate from a non-vesicular source, and is under cholinergic control (Rossi et al., 2003). The latter observation hints at an adaptive significance for the tonic GABA_A receptor-mediated conductance. Although it has been argued to reduce the transmission of information from mossy fibres to parallel fibres (De Schutter, 2002; Hamann et al., 2002), maintaining an ion gradient for a constant GABA_A receptor-mediated current appears to be a metabolically expensive way of achieving this. However, because GABA_A receptors are also targets of endogenous neuromodulators such as neurosteroids (Stell et al., 2003) and nitric oxide (Wall, 2003), this may provide a versatile way of modulating the computations performed by granule cells (Semyanov et al., 2004).

3.2. Tonic signalling in the dentate gyrus and hippocampus

Tonic GABA_A receptor activation has since been demonstrated in dentate granule cells too (Overstreet and Westbrook, 2001; Nusser and Mody, 2002; Stell and Mody, 2002). The pharmacological profile of GABA_A receptors in dentate granule cells has not been characterised to the same extent as that in cerebellar granule cells, but some indirect evidence suggests that δ subunits are present (Nusser and Mody, 2002; Stell and Mody, 2002), although α₆ subunits do not occur in this part of the brain. Because δ and γ subunits do not appear to co-assemble, and because γ subunits play an important role in synaptic targeting (Essrich et al., 1998), these findings suggest that tonic signalling in dentate granule cells is also mediated by extrajunctional δ-containing receptors. This hypothesis remains to be tested by high-resolution immunogold methods. As mentioned above, tonically active receptors are also present in hippocampal mossy fibres, although they are modulated by benzodiazepines, which distinguishes them from receptors in the somatodendritic compartment (Ruiz et al., 2003).

We have recently extended this work by showing that tonically active GABA_A receptors also occur in interneurons in the CA1 region of the hippocampus, both in stratum oriens and in stratum radiatum (Semyanov et al., 2003). Strikingly, tonic signalling was not observed in pyramidal neurons unless the extracellular GABA concentration was elevated experimentally. The receptors mediating this signal show some pharmacological peculiarities that are difficult to explain on the basis of their subunit composition. Thus, they are relatively resistant to gabazine, although sensitive to picrotoxin and the benzodiazepine agonist zolpidem (Fig. 2). This is a similar profile to one reported in cultured hippocampal neurons, which have been suggested to be extrasynaptic (Yeung et al., 2003), although a definitive demonstration with anatomical methods will require the subunit composition to be established. At present, therefore, it is not possible to exclude the hypothesis that synaptic receptors mediate both tonic signalling and inhibitory postsynaptic currents (‘phasic’ signalling).

Fig. 2.

Pharmacological separation of tonic and phasic GABA_A receptor-mediated signalling in hippocampal neurons. (a) An interneuron (IN, top) and a pyramidal cell (PC) were recorded in the CA1 region of acute guinea pig brain slices with a high [Cl⁻] pipette solution, in the presence of glutamate receptor antagonists (holding voltage −70 mV). The holding current shows intermittent inward transients, reflecting spontaneous inhibitory postsynaptic currents. Picrotoxin perfusion to block GABA_A receptors abolished the synaptic currents, but also produced an outward shift in holding current in the interneuron but not in the pyramidal cell, reflecting the removal of a tonic GABA_A conductance. (b) A low concentration of the GABA_A receptor antagonist gabazine (SR95531, SR) abolished spontaneous synaptic currents, but had no effect on the tonic current. Tonic and phasic GABA_A receptor-mediated currents thus have distinct pharmacological profiles. Reproduced from Semyanov et al. (2003).

The adaptive significance of tonic activation of GABA_A receptors in dentate granule cells is difficult to understand (a similar situation to that originally encountered when studying cerebellar granule cells): the tonic activation of GABA_A receptors results in constant offset current, which entails a significant metabolic cost to maintain the Cl⁻ gradient. One can speculate that this current provides a means of setting the overall excitability of the system, for instance by altering the membrane expression of the appropriate GABA_A receptors. However, such a form of receptor plasticity has yet to be demonstrated. In the CA1 region of the hippocampus, in contrast, the distinct behaviour of interneurons and principal cells suggests a possible role of the tonic GABAergic current in regulating the overall network excitability. Elevation of extracellular GABA (either by perfusion of GABA or by blockade of GABA uptake) caused an increase in the tonic current in interneurons and a decrease in their excitability (Semyanov et al., 2003). Conversely, blockade of the tonic current increased their excitability and, although this remains to be demonstrated, we predict that a decrease in extracellular GABA should do the same. Thus, bi-directional changes in interneuron excitability by tonic GABA_A receptor activation may result in a homeostatic regulation of the overall level of GABA released into the extracellular space.

4. Conclusions

There is now overwhelming evidence that GABA_A receptors are not confined to the postsynaptic elements of axo-dendritic or axo-somatic synapses. This review has concentrated on presynaptic receptors in the spinal cord, pituitary and hippocampus, and on extrasynaptic receptors that are thought to mediated tonic signalling in granule cells and hippocampal interneurons. Among other examples that have not been considered here are extrasynaptic GABA_A receptors occurring in axons far away from GABAergic synapses. Such receptors have been demonstrated in white matter of young rats (Sakatani et al., 1991) and even in peripheral nerves (Brown and Marsh, 1978). Although some evidence exists that they can modulate action potential conduction, and that they play a role in pathological situations, there is almost no evidence on whether they are activated by GABA released from oligodendrocytes, Schwann cells, migrating astroglia or from neighbouring grey matter.

Although some roles for extrasynaptic GABA_A receptors are starting to emerge, progress is hampered by the poor and incompletely understood selectivity of available ligands that can either activate or inhibit these receptors. One intriguing finding is that heterologously expressed α₄β₂δ GABA receptors are exquisitely sensitive to ethanol (at socially relevant concentrations) (Sundstrom-Poromaa et al., 2002). These receptors may occur extrasynaptically (Nusser et al., 1998), and are candidates for mediating a tonic conductance, implying that the psychoactive effects of ethanol may be related to this unconventional form of signalling.