[1]
135w
www.nature.com/reviews/neuro R E V I E W S of GABA transporters and postsynaptic receptors in relation to the site of transmitter release. By governing the time course and peak concentration of GABA to which receptors are exposed, these variables will not only influence occupancy, but will also dictate the subsequent changes that lead to channel opening. To account for the elementary functional properties of GABA A receptors, various kinetic schemes of microscopic gating have been proposed on the basis of steady-state and/or transient receptor activation [13][14][15][16][17][18] . It is envisaged that conformational changes in GABA receptors result in transitions between various closed, open (ion-conducting) and desensitized (relatively long-lived, agonist-bound closed) states. The time spent in each of these states is determined by the intrinsic properties of the channels and the temporal profile of GABA exposure.
[2]
178w
When recorded from compact cells, which afford voltage-clamp measurements of high temporal resolution, spontaneously occurring miniature inhibitory postsynaptic currents (mIPSCs), generated by GABA released from a single synaptic vesicle, have a rapid onset, with rise times of a few hundred microseconds 3,4,19 . This reflects the proximity of the receptors to the site of GABA release and the speed of the closedto-open transition 15,18,20,21 . If the time course of the GABA concentration transient is brief, the decay of the IPSC is dominated by the ion channel closure that follows ligand removal, a macroscopic phenomenon known as deactivation. The speed of this process is determined by the microscopic kinetics of the receptors and reflects the various transitions -most notably, entry into and exit from agonist-bound desensitized states that effectively trap GABA on the receptor before the eventual unbinding step 15,17,22,23 . The expression of receptor subtypes that incorporate different subunits (BOX 2) is proposed to contribute to the differences observed in the decay of IPSCs at different stages of development 24,25 and in different cell types [26][27][28] .
[3]
383w
The above description of phasic receptor activation addresses only the most straightforward situation, in which a single vesicle is released from an active zone and the liberated transmitter activates only those receptors that are clustered in the underlying postsynaptic density (FIG. 1a). In reality, there are further levels of complexity, which might include the activation of receptors at adjacent postsynaptic densities within the same synaptic bouton, or interactions between multiple vesicles that are released from a single synaptic specialization at a short interval (≤1 ms), from several nearby synapses or following repeated synaptic activation. If an action potential triggers the release of multiple vesicles at a single active zone (multivesicular release), the postsynaptic receptors will be exposed to a different GABA concentration transient. The time course of the synaptic GABA concentration change will be modified substantially if this vesicle release is temporally dispersed (asynchronous release). Following diffusion from its release site(s), GABA might activate adjacent (perisynaptic) receptors, receptors at other postsynaptic densities made by the same bouton, more remote extrasynaptic receptors or receptors at A defining feature of this phasic mode of receptor activation is the short duration of the GABA transient to which the postsynaptic receptors are exposed. Experiments using low-affinity competitive antagonists indicate that the synaptic GABA concentration decays with a time constant of <500 µs 5 , and related studies with agents that slow down the binding of GABA to its receptors indicate a time constant of synaptic GABA clearance of ~100 µs 6,7 . The short dwell time of GABA within the cleft can be attributed to its rapid diffusion away from the release site 5 . Efficient gating of GABA A receptor ion channels requires the receptor to be occupied by two agonist molecules 8 , and, for GABA, the binding rate is slow relative to diffusion 9 . Although the peak concentration of GABA might be higher than that required to produce maximal receptor activation at steady state, the short exposure time means that not all postsynaptic receptors will necessarily be fully occupied. Although postsynaptic receptor saturation (full occupancy) following the release of GABA from a single vesicle does occur at certain synapses, the degree of receptor occupancy often varies between synapses on different neurons, and can even vary between synapses on a single neuron 3,6,[10][11][12] .
[4]
38w
The time course of the GABA transient in the synaptic cleft might be influenced by variations in vesicle size and content, the nature of vesicle fusion, the geometry of the synaptic cleft, and the number and spatial arrangement
[5]
173w
Responses generated by ionotropic receptors result from the dissipation of transmembrane ionic gradients, which are produced by ion pumps and carriers. GABA A (γ-aminobutyric acid type A) receptors are permeable to chloride and bicarbonate anions 179,180 . The functional outcome of receptor activation depends on the transmembrane distribution of these two anions and on the membrane potential of the cell. In most mature neurons, the activity of the chloride-extruding potassium-chloride co-transporter KCC2 (REF. 181) results in a chloride equilibrium potential that is more negative than the resting membrane potential (V m ). The equilibrium potential for bicarbonate is more positive than V m , but bicarbonate is much less permeable than chloride. Therefore, GABA A receptor activation typically results in the net entry of anion, and the classically described hyperpolarizing inhibitory postsynaptic potential (IPSP). In this case, both the increase in conductance (that causes shunting of excitatory inputs) and the hyperpolarization (that sums with depolarizations) contribute to the 'inhibitory' effect of GABA, thereby reducing the probability that an action potential will be initiated.
[6]
184w
However, this model is an oversimplification. A hyperpolarizing GABA response might not be inhibitory if it triggers hyperpolarization-activated excitatory conductances to produce rebound spikes 182,183 . Moreover, the response to GABA itself can be depolarizing. This is true for most immature neurons that lack KCC2 and instead accumulate chloride by way of the sodium-and potassium-coupled cotransporter NKCC1 (REFS 181,184), and is also true for some mature neurons 185,186 . Although inhibition can still occur owing to the shunting effect of the increase in conductance 78,187 , the effect of the IPSP depends on its location and timing in relation to excitatory inputs, and the interplay between the respective conductance and voltage changes 188 . In immature neurons, the depolarization might be sufficient to trigger calcium influx, a phenomenon that is implicated in GABA-mediated modulation of neuronal proliferation, migration, growth and synapse formation [189][190][191] . It has recently been suggested that even when GABA-induced depolarization is insufficient to activate voltage-gated calcium channels, calcium entry might still occur following sustained receptor activation in response to the osmotic load that is induced by chloride entry 182 .
[7]
125w
recordings from rat cerebellar granule cells 41 . The GABA A receptor antagonists bicuculline and SR-95531 (gabazine) not only blocked spontaneously occurring IPSCs, but also decreased the 'holding' current that was required to clamp the cells at a given membrane potential. This reduction in the input conductance was associated with a reduction in current variance, consistent with a decrease in the number of open GABA A receptor channels [41][42][43][44] . Subsequent studies have indicated that GABA-mediated tonic conductances exist in granule cells of the dentate gyrus 45 , thalamocortical relay neurons of the ventral basal complex 46 , layer V pyramidal neurons in the somatosensory cortex 47 , CA1 pyramidal cells 48 and certain inhibitory interneurons in the CA1 region of the hippocampus 49 .
[8]
163w
Identifying the GABA source, or sources, is important if we are to understand how tonic receptor activation is modulated. Although certain recombinant 23,[50][51][52] and native 53 GABA A receptors have been shown to open spontaneously with low probability in the absence of agonists, most GABA A receptors require the binding of agonist molecules to promote entry into open states. Accordingly, the most parsimonious explanation for the presence of a tonic conductance is that GABA (or some other GABA A receptor agonist) must be present in the extracellular space at a sufficiently high concentration to cause persistent receptor activation. On the basis of theoretical considerations of GABA transporter stoichiometry 54 , or solute recovery during in vivo microdialysis [55][56][57][58] , estimates of the concentration of ambient GABA vary from tens of nanomolar to a few micromolar. This range probably reflects the uncertainties involved in different methods of estimation, but it also underscores the likelihood of genuine regional and temporal variations in extracellular GABA concentration.
[9]
105w
In the postnatal brain, the origin of GABA has been investigated most extensively in cerebellar granule cells. In the juvenile animal, action potential-dependent vesicular release clearly underlies the maintenance of the ambient GABA concentration that is responsible for tonic receptor activation, and one factor that contributes to this process is the presence of many GABAreleasing Golgi cell axon terminals in the cerebellar GLOMERULUS 41-44 . For mature granule cells, the situation is less clear: vesicular release does have a role 59,60 , but it has been suggested that there is also a non-vesicular source of GABA 44,61 , although this has not yet been identified.
[10]
175w
The concentration of GABA in the extracellular space reflects not only the number and 'activity' of GABA-releasing elements, but also the action of GABA transporters. These sodium and chloride symporters, which are normally responsible for removing GABA from the extracellular space, can also operate in the reverse direction, so, in certain circumstances, can themselves provide a source of GABA 54 . However, following pharmacological blockade of transport 44,45,49,61 , and in transporter-deficient mice 62 , the magnitude of the tonic current increases, which indicates that reversed transporter activity does not usually contribute to ambient GABA. nearby synapses. In this case, the GABA waveform to which any particular receptor is exposed will be determined by its location relative to the release site, the geometry and spatial arrangement of the neighbouring cellular elements, diffusional barriers and the proximity of GABA transporters in neurons and astroglia 5,29,30,31 . It is important to note that the currents that result from GABA spillover can still be considered phasic, in the sense that they are temporally related to the release event.
[11]
88w
Tonic receptor activation. Phasic activation of synaptic receptors is fundamental to information transfer in the brain. However, it is recognized that neurotransmitters that are traditionally considered to participate in rapid point-to-point communication through the activation of ionotropic receptors might also participate in slower forms of signalling 32 . At the extreme, this might include the tonic activation of receptors. Underlying this PARACRINE activity is the widespread presence of receptors in somatic, dendritic and axonal regions of neuronal membrane that are distant from sites of neurotransmitter release 33 .
[12]
84w
Tonic activation of GABA A receptors is evident in certain embryonic neurons before synapse formation has taken place [34][35][36][37] . In mature neurons that display IPSCs, the tonic activation of GABA A receptors, as opposed to the superimposition of high-frequency phasic events [38][39][40] , was first identified in voltage-clamp ALTERNATIVE SPLICING During splicing, introns are excised from RNA after transcription and the cut ends are rejoined to form a continuous message. Alternative splicing allows the production of different messages from the same DNA molecule.
[13]
39w
A signalling process that involves the secretion from a cell of molecules that act on other cells expressing appropriate receptors in the immediate neighbourhood, rather than acting on the same cell (autocrine signalling) or on remote cells (endocrine signalling).
[14]
98w
Like other members of the cysteine-loop ligand-gated ion channel family, such as nicotinic acetylcholine, glycine and 5-hydroxytryptamine type 3 (5-HT 3 ) receptors 192 , GABA A (γ-aminobutyric acid type A) receptors are pentameric assemblies of subunits that form a central ion channel. Nineteen GABA A receptor subunits (α1-6, β1-3, γ1-3, δ, ε, θ, π and ρ1-3) have been cloned from the mammalian CNS, with further variation resulting from ALTERNATIVE SPLICING (for example, for the γ2 subunit 193 ). The combinatorial co-assembly of these various subunit proteins allows a potentially enormous molecular heterogeneity of GABA A receptor subtypes.
[15]
313w
Of the many subunit combinations that are theoretically possible, only a few dozen have been shown to exist, reflecting the differential distribution of subunit types among brain regions and neuronal populations [194][195][196] , but also implying several basic 'rules' of assembly 106,197 . The most abundantly expressed receptor subtype in the brain is formed from α1, β2 and γ2 subunits [198][199][200] . The likely stoichiometry is two α, two β and one γ subunit 201,202 , with the subunits arranged pseudo-symmetrically around the ion channel in the sequence γ-β-α-β-α, anticlockwise when viewed from the synaptic cleft 196,203 . Other common assemblies also contain α, β and γ2 subunits (for example, α 2 β 3 γ 2 , α 3 β 3 γ 2 , α 4 β x γ 2 , α 5 β 3 γ 2 and α 6 β x γ 2 ), whereas receptors in which the γ2 subunit is replaced by γ1, γ3, or δ are less abundant. Further variability arises from the fact that individual pentamers might contain two different α or two different β subunit isoforms (reviewed in REF. 198). In some cases, the γ subunit can be replaced by an ε, δ or π subunit, and the π and θ subunits might also be capable of co-assembling with α, β and γ subunits to form receptors that contain subunits from four families [204][205][206] . Finally, although the ρ1 subunits can form homomeric receptors that share certain properties with the GABA C (GABA type C) subfamily of ionotropic GABA receptors 207,208 , there is evidence that they can also form receptors with γ2 subunits 209 or with both α1 and γ2 subunits 210 . This molecular heterogeneity has important functional consequences for GABA A receptor subtypes: subunit composition dictates not only the properties of the receptors, but also their cell surface distribution and dynamic regulation 106,198,211 .
[16]
126w
A notable example is provided by the action of the cortical and hippocampal basket cells that innervate the perisomatic regions of pyramidal cells. By phasing and synchronizing the activity of a large population of pyramidal cells, these interneurons have an essential role in generating and maintaining THETA and GAMMA FREQUENCY NETWORK OSCILLATIONS 65,68,69 . This action requires the mutual interconnection of interneurons by chemical and electrical synapses 69,70 . For GABA A receptormediated postsynaptic conductances, a rapid time course (~5 ms) is essential for synchronization at high frequencies (for example, gamma frequency 71,72 ). A role for phasic inhibition in the generation or regulation of synchronous population activity has also been shown in several other brain regions, including the thalamus 73 and olfactory bulb 74 .
[17]
52w
The exact location of GABA-releasing synapses, and the temporal relationship between their activation and that of other synaptic or voltage-gated conductances, is also important in the control of regenerative electrical activity in dendrites 75,76 . Synapse location also affects the impact of phasic GABA-mediated input on synaptic integration. For example, the selective
[18]
72w
The main feature of phasic GABA A receptor-mediated inhibition is the rapid synchronous opening of a relatively small number of channels that are clustered at the synaptic junction, whereas tonic inhibition results from random, temporally dispersed activation of receptors that are distributed (albeit in a potentially non-uniform manner) over the neuronal surface. This distinction implies a profound difference in the control of neuronal network activity by phasic and tonic forms of inhibition.
[19]
83w
Preventing overexcitation of neurons, and thereby avoiding the development of pathological states of network activity, is an essential task of GABA-releasing interneurons and GABA A receptors in the adult CNS. However, it is clear that interneurons have more complex roles than the provision of generalized inhibition, and depend crucially on synapse location and IPSC timing [63][64][65][66][67][68] . One important function of phasic inhibition, the effectiveness of which is determined by both of these variables, is the generation of rhythmic activities in neuronal networks.
[20]
477w
Axon terminals end in various configurations within the neuropil. The most common is en passant or de passage, in which axons make simple synapses as they pass dendrites or cell bodies. By contrast, some axons end in -or produce strings ofenlargements that are often packed with synaptic vesicles. These glomerular-type endings might synapse with large numbers of dendrites. In the cerebellum, each large excitatory mossy fibre terminal contacts dendrites from many granule cells and, together with inhibitory Golgi cell axon terminals, forms a glomerular structure that is wrapped with glia. evoked release from several terminals promotes GABA 'spillover', and activates both synaptic receptors and perisynaptic or extrasynaptic receptors (blue). The current record shows the larger and much slower averaged waveform of IPSCs evoked by electrical stimulation. The area of the mIPSC is superimposed for comparison. c | A low concentration of ambient GABA, which persists despite the activity of the neuronal and glial GABA transporters (GAT1 and GAT3), tonically activates high-affinity extrasynaptic receptors. The trace shows the 'noisy' tonic current that results from stochastic opening of these high-affinity GABA A receptors, with superimposed phasic currents (in this case, the synaptic events would be arising at sites not depicted in the schematic diagram). A high concentration (10 µM) of the GABA A antagonist gabazine (SR-95531) blocks the phasic IPSCs and tonic channel activity, causing a change in the 'holding' current and a reduction in current variance. The infrequent phasic events that remain in SR-95531 are glutamatergic excitatory postsynaptic currents. The shaded area beneath the current record before SR-95531 application represents the charge carried by tonically active GABA A receptors. The frequency of spontaneous IPSCs is relatively low and the tonic receptor activity generates a conductance several-fold larger than the averaged conductance that is carried by phasic IPSCs 42,61 phasic events, when viewed from the soma, becomes indistinguishable from a tonic conductance. However, there are clear differences between the two phenomena. In the case of high-frequency phasic transmission, the signals that result from discrete vesicular release events are integrated in the postsynaptic cell. In the case of tonic inhibition, 'integration' takes place in the extracellular space, where GABA is pooled to achieve an averaged ambient concentration, albeit one that can still change over time. This has implications for the way in which the two processes reflect network activity, and might also have significant energetic considerations. If the contributing IPSCs occur at different synapses that are distributed on a complex dendritic tree, the input is discrete and phasic for each dendritic location and could still participate in temporally precise local processing. Moreover, for neurons with many inputs, it might be feasible to achieve a sustained conductance through the integration of IPSCs from many sources, whereas for neurons with few inputs, such as cerebellar granule cells, this might be achieved only by sharing a restricted number of GABA-releasing elements 42 .
[21]
214w
Several groups have investigated how tonic inhibition in cerebellar granule cells affects their excitability. If step current injections of increasing amplitude are used to evoke action potentials in granule cells, blockade of tonic inhibition with GABA A antagonists decreases the current that is required to achieve a given firing ratethe input-output relationship is shifted to the left 42,81,82 . The same reduction in firing rate is seen at all levels of excitation -equivalent to a subtractive mathematical operation (FIG. 2). Mitchell and Silver 83 used a dynamic clamp to restore tonic conductance to granule cells bathed in GABA A receptor antagonists, and their results were complementary to those seen with pharmacological blockade. However, they also showed that the effect of the tonic inhibitory conductance depended on the nature of the excitatory input. If, instead of a step excitation, random trains of synaptic conductances were used to excite the cells, shunting inhibition no longer simply shifted the input-output relationship to the right, but also decreased its slope, which corresponds to a change in gain (a divisive mathematical operation). The slope of the input-output relationship depends on the variability of the input conductance: with tonic inhibition, a higher frequency of synaptic excitatory input (with higher variance) is required to achieve a given output rate [83][84][85] .
[22]
200w
An increase in the slope of the input-output relationship on blockade of the tonic conductance was also seen when mossy fibres, which provide excitatory synaptic input to granule cells, were stimulated at high frequency 81 . Changes in tonic GABA A receptor activation, through changes in Golgi cell firing 42,59 , will therefore modify the sensitivity of the granule cell to changes in the frequency of mossy fibre input, and contribute to the sparse coding of sensory input by granule cells that is thought to be necessary for effective motor control 86,87 . Network models of the cerebellar cortex indicate that tonic inhibition of granule cells might also limit the oscillatory behaviour that is entrained by phasic feedback inhibition from Golgi cells 88 . activation of somatically terminating interneurons during feed-forward inhibition of hippocampal pyramidal cells produces a requirement for precise COINCI- DENCE DETECTION of excitatory input at the soma, whereas dendrites can integrate synaptic input over longer time periods 77 . Finally, whether the depolarizing action of GABA that persists in mature cortical pyramidal cells is inhibitory or excitatory (BOX 1) depends on the location of synapses and the timing of their activity relative to excitatory inputs 78,79 .
[23]
239w
These examples show the importance of spatially restricted IPSPs in enabling certain neuronal behaviours. In cells that receive spatially segregated plastic inhibitory input from several sources, it is also important to appreciate that this input might be subject to exquisite modulation, either through changes in the activity of the parent interneurons, or by the regulation of transmitter release from their terminals. Cortical and hippocampal interneurons are known to express, in a cell type-specific manner, receptors for various neurotransmitters and neuromodulators, including GABA, glutamate, serotonin (5-hydroxytryptamine or 5-HT), opioids, monoamines, acetylcholine and endocannabinoids, and they respond uniquely to alterations in the levels of these neuromodulators with changes in their firing frequency or GABA release 80 . By changing the activity or output reliability of specific interneuron types, this modulation allows sophisticated control that is much more refined than a simple change in the frequency, amplitude or duration of all IPSCs in the cell. For example, if cortical axo-axonic cells were selectively silenced, there would be little change in the total IPSC frequency in postsynaptic target cells, but their output would be profoundly affected. Similarly, if basket cellevoked perisomatic IPSCs were desynchronized, there would be no change in the total phasic inhibition, but rhythmic network activity would probably collapse. So, small spatially and/or temporally restricted alterations in interneuron activity that do not greatly affect total phasic inhibition might crucially alter the way in which subcortical information is conveyed to cortical networks.
[24]
118w
Compared with phasic GABA A receptor activation, we might expect tonic GABA A receptor activation to be much more limited in its capabilities and less susceptible to modulation. Tonic activation of GABA A receptors has one straightforward outcome: a persistent increase in the cell's input conductance. This affects the magnitude and duration of the voltage response to an injected current, and increases the decrement of voltage with distance. So, for a given excitatory input (excitatory postsynaptic current or EPSC), the size and duration of the excitatory postsynaptic potential (EPSP) will be reduced, and the temporal and spatial window over which signal integration can occur will be narrowed, making it less likely that an action potential will be generated.
[25]
57w
How is asynchronous tonic receptor activation distinct from a high frequency of synaptic input? If we consider the postsynaptic conductance in isolation, then at some point the integrated response to many COINCIDENCE DETECTION A situation in which two different subthreshold excitatory inputs are sufficiently closely timed that they summate to trigger the generation of an action potential.
[26]
128w
In the hippocampus, the contribution of some interneurons to the generation of network oscillations might also be affected by tonic inhibition 49 . Moreover, it has been suggested that differences in the tonic conductance of interneurons and pyramidal cells might contribute to the homeostatic regulation of phasic inhibition in pyramidal cells 49,85 . Although tonic GABA A receptor activity is present in developing pyramidal cells 37 , it is not generally seen in brain slices from adult animals unless GABA uptake or degradation are blocked [89][90][91] , or GABA receptor affinity is increased 92,93 (but see REF. 48). So, under normal conditions, pharmacological blockade of tonic inhibition selectively enhances the excitability of interneurons, leading to an increase in the frequency of IPSCs in CA1 pyramidal cells 49 .
[27]
55w
The question of whether axonal or presynaptic GABA A receptors are tonically activated has received less attention, so the possible functional consequences of such activation are poorly understood. Although axonal receptors are unlikely to influence integration at the somato-dendritic level, there is evidence that they might modulate action potential conduction and transmitter release 33 .
[28]
172w
Although variations in the subcellular locations of receptor subtypes undoubtedly contribute to their selective participation in tonic and phasic forms of activity, this distinction alone is not sufficient to account for their differential activation. Given the variations in and atextrasynaptic and perisynaptic locations in hippocampal dentate gyrus granule cells 104 . In the cerebellum, gold particles labelling the δ subunit were typically hundreds of nanometres from the edge of the nearest postsynaptic density, whereas in the hippocampus they were concentrated in a perisynaptic position, just outside the postsynaptic density (within 30 nm). The δ subunit forms receptors specifically with the α6 and β2/3 subunits (α 6 β 2/3 δ and α 1 α 6 β 2/3 δ) in cerebellar granule cells and with the α4 and βx subunits (α 4 β x δ) in several areas of the forebrain, including the thalamus, neostriatum and dentate gyrus 105 . For each of these receptor subtypes, the lack of a γ subunit is probably responsible for their failure to be incorporated at the synapse.
[29]
122w
The postsynaptic density of GABA-releasing synapses contains many proteins that have been postulated to have roles in the targeting and stabilization of GABA A receptors [106][107][108] . The precise molecular architecture remains to be established, but it is clear that the γ2 subunit has a central role in the clustering of synaptic GABA A receptors. Deletion of the γ2 gene in mice leads to a profound reduction in the clustering of both GABA A receptors and the GABA A receptor-associated protein gephyrin, which is paralleled by a reduction in mIPSC frequency 109 . This requirement for the γ2 subunit exists not only in embryonic neurons that are undergoing synapse formation, but also in more mature neurons with existing synaptic contacts 110,111 .
[30]
236w
δ subunit-containing receptors seem to be purely extrasynaptic, but other subtypes might also be present predominantly, if not exclusively, outside synapses. In hippocampal pyramidal cells, the α5 subunit (which probably forms α 5 β 3 γ 2 receptors) shows diffuse surface labelling at the light microscopic level without detectable synaptic clustering, as judged by the lack of co-localization with gephyrin 103,112 (FIG. 4). In this case, the presence of the α5 subunit seems to override the ability of the γ2 subunit to promote synaptic localization. Notably, in hippocampal slices from mice lacking the α5 subunit, the amplitudes of action potentialdependent or evoked IPSCs in CA1 pyramidal cells were reduced compared with those in wild-type mice 113 . This difference was not seen for mIPSCs in cultured CA1 neurons 90 . Although the reduction in evoked IPSC amplitude could reflect compensatory changes in either the probability of GABA release or the number of release sites without any change in quantal size, it is also consistent with the idea that phasic activation of α5-containing receptors might require synchronous multivesicular release and spillover of GABA onto receptors located beyond the synaptic cleft, as described for extrasynaptic or perisynaptic α6 and α4 subunitcontaining receptors in granule cells of the cerebellum 114 and dentate gyrus 104 . This could be one of the mechanisms that contribute to the generation of slow dendritic IPSCs that are seen in CA1 neurons 115 .
[31]
159w
Overall, these findings indicate that receptors containing a γ2 subunit in association with α1, α2 or α3 subunits (α 1 β 2/3 γ 2 , α 2 β 2/3 γ 2 and α 3 β 2/3 γ 2 ) are the predominant receptor subtypes that mediate phasic synaptic inhibition. Receptors that contain α4, α5 or following exposure to agonist, the rate and extent of desensitization of the current in the continued presence of the agonist and the deactivation of the current following agonist removal. These measurements reflect various microscopic parameters attributed to the receptors, which include the agonist binding and unbinding rates, and the rate constants of the transitions to and from open and desensitized states. Other measurable features that depend on the interplay of these various gating processes include the mean open and closed times and the mean burst durations of the channels, as well as the probability of channels being open when the receptors are fully occupied.
[32]
234w
A key property of any ligand-gated ion channel is its sensitivity to endogenous agonist -that is, how much ligand is required to produce a given response. This reflects both the affinity of the receptor for its ligand (the equilibrium constant for the binding step) and the efficacy of the ligand (how effectively it promotes ion channel gating) 118 . For recombinant receptors that contain α, β and γ subunits, sensitivity to GABA is most strongly affected by the type of α subunit that is present, with α3 subunits conferring the highest and α6 subunits the lowest EC 50 values [119][120][121][122][123] . Across studies, the absolute EC 50 values for specific subunit combinations -and the relative differences between the various combinationsare variable, but, in studies in which α subunits have been compared, the rank order was shown to be α6<α1<α2<α4<α5<<α3 (REF. 121). Replacing the γ2 subunit in α 4 β 3 γ 2 assemblies with a δ subunit decreases the EC 50 for GABA 124 , but makes no difference to α 1 β 3 γ 2 and α 6 β 3 γ 2 assemblies 122 . Overall, α 6 β 3 δ or α 4 β 3 δ combinations have the lowest EC 50 s for GABA (~0.3-0.7 µM), whereas for α 1 β 3 γ 2 or α 2 β 3 γ 2 subtypes they are an order of magnitude higher (~6-14 µM).
[33]
198w
For receptors that become occupied by ligand and subsequently open, the conductance of their channels will influence the size of the response. The single-channel conductance of GABA A receptors depends on their subunit composition. Compared with changes in EC 50 , the changes in conductance are more modest, and are essentially restricted to those that are produced by switching from dimeric αβ to ternary αβγ or αβδ assemblies. When recorded at room temperature in outside-out patches, αβ receptors (for example, α 1 β 1 or α 1 β 3 ) have a single-channel conductance of ~15 pS, whereas those incorporating a γ2 or a δ subunit (for example, α 1 β 2 γ 2S or α 1 β 3 δ) have a conductance of ~25-28 pS 4,120,[125][126][127] . Changing the type of α or β subunit in the assembly has little or no effect on the conductance. Although recordings from immature neurons indicate the existence of native dimeric αβ receptors 4 , most receptors in mature neurons probably contain either a γ or a δ subunit. So, we would expect that, in most mature neurons, GABA A receptors of roughly similar channel conductance underlie tonic and phasic inhibition.
[34]
88w
The magnitude of the response is not determined solely by channel conductance: the time that the channels spend in the open state is equally important, and the kinetic properties that govern this phenomenon are also influenced by subunit composition. In the case of αβγ GABA exposure that such receptors are likely to encounter, we might expect them to be endowed with distinct biophysical properties, particularly those associated with processes of binding (how the agonist interacts with the receptor) and gating (how the channel opens and closes in response).
[35]
121w
Measurable macroscopic parameters include the concentration of ligand that gives the half-maximal response (EC 50 ), the rate of activation of the current are wholly consistent with their involvement in phasic and tonic signalling. However, this distinction does not preclude the possibility that under conditions of elevated extracellular GABA, or drug-induced increases in receptor affinity, other extrasynaptic and/or synaptic receptors might contribute to the generation of a tonic conductance. Notably, studies that have addressed the biophysical properties of recombinant receptors commonly investigate the effects of high concentrations of GABA that are relevant specifically to synaptic transmission. The use of lower GABA concentrations should provide a clearer view of the behaviour of receptor subtypes that are activated by ambient GABA 17,144,145 .
[36]
153w
The pattern of phasic inhibition that a neuron receives is obviously determined by the number, variety and activity of presynaptic GABA-releasing neurons, but whether tonic inhibition is similarly determined by neuronal activity is less clear. An ability to modulate the tonic conductance would seem to be essential if this form of inhibition is to reflect dynamic network activity, as opposed to simply providing a constant brake on excitability. Synaptic transmission relies on the interplay of many tightly regulated processes that together determine the timing, magnitude and kinetics of postsynaptic responses, and each of these processes might be subject to modulation. Tonic inhibition shows far fewer degrees of freedom, yet, owing in part to the low receptor occupancy, the dynamic range of modulation is potentially much larger. In theory, both phasic and tonic inhibition could be modulated by changes in GABA release or uptake and/or by changes in the number and properties of receptors.
[37]
284w
If the GABA that is released from synaptic vesicles contributes in any way to tonic inhibition, changes in presynaptic activity or release would be expected to modify the magnitude of the tonic conductance. This has been shown to be the case in the hippocampus, where stimulation of interneuron firing by the glutamate receptor agonist kainate increases the GABA A receptor-mediated tonic conductance in both pyramidal cells 146 and interneurons 147 . Furthermore, in cerebellar granule cells, facilitation of GABA release by acetylcholine causes a calcium-dependent, action potential-independent increase in the tonic conductance 61 . As the cerebellum receives cholinergic innervation, this latter mechanism could provide a physiologically relevant modulation of granule cell excitability. Blockade of action potential firing with TETRODOTOXIN has also been shown to reduce the tonic conductance in cultured neurons from the hippocampus 145 and cerebellum 148 . GABA transporters have well-documented effects on phasic inhibition, and they also have an important and dynamic influence on ambient GABA 149 . The extracellular GABA concentration at which they are at equilibrium, and consequently the magnitude of GABA flux, will vary depending on the membrane potential and αβδ receptors, significant differences have been documented for both microscopic channel behaviours and macroscopic kinetic parameters. In α 1 β 3 γ 2 receptors, replacing the γ2 with a δ subunit results in a roughly 5-fold reduction in both the mean open time and the duration of bursts of channel openings 120 . Single-channel data for the prevalent α 4 β x δ combination 128 indicate that these receptors probably behave similarly, which is consistent with the idea that GABA has high affinity but low efficacy (GABA is a partial agonist) at δ subunitcontaining receptors 124,129,130 .
[38]
185w
The activation, deactivation and desensitization of recombinant receptors are also greatly affected by their subunit composition. For α x β x γ 2 receptors, direct comparison of receptors with different α subunits reveals up to 4-fold differences in the activation rate of currents evoked by rapid application of high concentrations of GABA, with rise times in the order α2<α1<α3 (REFS 21,131,132). The presence of a δ or γ2 subunit also influences activation, with rise times in the order α 1 β 3 γ 2L «α 1 β 3 δ≈α 1 β 3 (REF. 17). Insertion of a γ2 subunit (S but not L splice variant) into αβ receptors increases deactivation speed ~ 2-fold 133,134 , as does replacing the γ2 with a δ subunit 17 . Moreover, for both αβγ and αβδ assemblies, the rate of deactivation depends on the type of α subunit present; for example, α1-containing α x β 1 γ 2 receptors deactivate ~5-fold faster than those containing the α2 subunit 21 , and α1-containing α x β 3 δ receptors deactivate ~4-fold faster than those containing the α6 subunit 135 .
[39]
191w
The entry of GABA A receptors into desensitized states is thought to be important for shaping the time course of IPSCs 5,15,17,22,23,136 . Desensitization also affects the ability of postsynaptic receptors to respond to repetitive highfrequency activation 137 (but see REFS 138,139), and is of obvious importance with regard to the effect of a persistent low concentration of GABA. Ambient GABA can promote entry of receptors into partially bound, slowly desensitizing states 15,140,141 , which potentially limits the magnitude of any tonic conductance, and also reduces the availability of synaptic receptors 140 . Consistent with the interrelation of deactivation and desensitization, the addition of a γ2 subunit to αβ receptors slows macroscopic desensitization 133,134 . Likewise, αβδ receptors desensitize more slowly and less extensively than αβγ receptors 17,137,142 .Again, for both αβγ and αβδ assemblies, the rate and extent of desensitization is influenced by the type of α subunit: receptors with an α 1 βγ subunit composition desensitize more rapidly than those containing an α5 (REF. 90) or α6subunit 143 , whereas the opposite effect is observed for substitution of the α1 with an α6 subunit in αβδ receptors 135 .
[40]
84w
In summary, data from recombinant receptors show that all macroscopic and microscopic properties of GABA A receptors depend strongly on their subunit composition. The most important differences between receptors that mediate phasic inhibition and those that have been implicated in tonic inhibition are their affinities for GABA and the speed and extent of their desensitization. These different biophysical features, together with their differential cell surface distributions, TETRODOTOXIN A potent marine neurotoxin that blocks voltage-gated sodium channels. Tetrodotoxin was originally isolated from the tetraodon pufferfish.
[41]
226w
As might be expected, differences in subunit composition between synaptic and extra-or perisynaptic receptors are reflected in the differential modulation of phasic and tonic inhibition by benzodiazepine site ligands. In dentate gyrus granule cells, zolpidem prolongs the decay of IPSCs, but has no effect on the tonic conductance 45 , which is thought to be mediated by benzodiazepineinsensitive α 4 βδ receptors. A similar differential effect is seen with diazepam in cerebellar granule cells 81 , in which the tonic conductance is mediated by α 6 βδ receptors. In cultured hippocampal neurons, IPSCs are prolonged by both midazolam 48 (Versed; Roche) and zolpidem 90 , whereas the tonic conductance that is seen in the presence of the GABA transaminase inhibitor vigabatrin (Sabril; Sanofi-Aventis) is enhanced only by midazolam 48,90,168 , consistent with a tonic activation of α 5 β 3 γ 2 receptors. By contrast, the tonic conductance in CA1 interneurons is enhanced by low concentrations of zolpidem 49 -a result that is inconsistent with the activation of α5 subunit-containing receptors in these cells. It should be noted that enhanced tonic conductance in the presence of a positive allosteric modulator is not as easy to interpret as a lack of effect, because any druginduced increase in receptor affinity might recruit receptor populations (including synaptic receptors) that are not ordinarily activated by the low ambient GABA concentration.
[42]
251w
For the experimental investigation of tonic and phasic inhibition, several competitive and non-competitive GABA A receptor antagonists have also been useful, not because of differences in their affinity for the receptors that underlie the tonic and phasic currents, but because their function depends on the affinity of the receptors for GABA and on the conditions of receptor activation. All GABA A receptor-mediated conductances are blocked by high concentrations of bicuculline, picrotoxin or SR-95531. However, a sub-micromolar concentration of SR-95531 selectively blocks phasic currents 48,49,91,168 , consistent with the underlying receptors having a lower affinity for GABA than those that mediate the tonic current 91 .A differential block of phasic and tonic currents has also been observed with the open-channel blocker penicillin 168 , and this has been proposed to reflect the low occupancy of the receptors that mediate the tonic conductance. Few GABA A receptor antagonists show clear subunit selectivity, but the diuretic furosemide (Lasix; Aventis) has ~100-fold selectivity for α6 over α1 subunit-containing receptors 123,169 , and has been used to determine the role of synaptic and extrasynaptic α6-containing receptors in cerebellar granule cells 81,170,171 . The only agonist that shows a clearly different profile of action at receptors of synaptic or extrasynaptic subtype is 4,5,6,7-tetrahydroisothiazolo-[5,4-c]pyridin-3-ol (THIP (Gaboxadol; Lundbeck/Merck)). This compound is a partial agonist at α 4 β 3 γ 2 receptors but behaves as a full or 'super' agonist at α 4 β 3 δ receptors, producing a maximum response greater than that produced by GABA 124 .
[43]
131w
Tonic inhibition also seems to be highly sensitive to modulation by various clinically relevant agents, including endogenous neuroactive steroids (FIG. 5), intravenous and inhalation anaesthetics, certain NOOTROPIC agents and the transmembrane gradients for the transported substrates (GABA, sodium and chloride). This also means that their ability to function as potential sources of GABA will be of greatest significance under pathological conditions or during exposure to drugs that increase the intracellular GABA concentration [149][150][151][152] . It is also recognized that transporters can undergo rapid redistribution between surface and intracellular compartments, and that their function can be altered by phosphorylation 153,154 or intermolecular interactions 155 . So, even in the face of unchanging GABA release, it is possible that ambient GABA concentration, and thereby tonic inhibition, could be modulated by changes in uptake.
[44]
218w
Many processes are known to modulate GABA A receptor number and function and these are likely to be relevant to both phasic and tonic inhibition. For example, the intracellular loops of β and γ subunits contain sites for phosphorylation by various protein kinases 156 , and the intracellular loop of the γ2 subunit is a substrate for PALMITOYLATION by the thioacyltransferase GODZ 157 . These reversible, post-translational modifications have been shown to affect both the properties 158,159 and subcellular location 160,161 of the receptors. GABA A receptors can cycle rapidly between surface and intracellular domains, and probably move laterally within the membrane. In addition, through interaction with various cytosolic proteins, they can cluster at synaptic and non-synaptic sites 106,156,162 . This dynamic behaviour not only allows rapid changes in receptor number, but might also induce changes in receptor properties. For example, clustering of recombinant GABA A receptors by the receptor-associated protein GABARAP changes both their kinetic behaviour and single-channel conductance 163,164 , and in cultured hippocampal neurons, cytoskeleton disruption reduces receptor clustering and alters the behaviour of both synaptic 165 and extrasynaptic 145 receptors. Finally, various pathological conditions (for example, epilepsy), hormonal fluctuations and chronic ethanol withdrawal have been shown to differentially affect the expression of subunits that have been implicated in tonic and phasic inhibition 166,167 .
[45]
144w
Just as the biophysical properties of GABA A receptors are determined by their subunit composition, so are their pharmacological properties. The most frequently cited example is the role of the α subunits in defining their affinity for benzodiazepines (widely used GABA A receptor ALLOSTERIC modulators). Receptors formed from α1, α2, α3 or α5 subunits, together with two β and one γ2 subunit, have a high affinity for diazepam, a classic benzodiazepine agonist. Receptors containing the α1 subunit also have a high affinity for the imidazopyridine zolpidem (Ambien; Sanofi-Synthelabo). Changing the α subunit to α4 or α6 eliminates diazepam and zolpidem sensitivity, as does substitution of the γ2 subunit with a δ, ε or π subunit 105 . Subunit positioning within the pentamer is also crucial; for example, benzodiazepine sensitivity is affected by the type of α subunit that neighbours the γ2 subunit 123 .
[46]
76w
The covalent attachment of a palmitate (16-carbon, saturated fatty acid) to a cysteine residue through a thioester bond. ALLOSTERIC A term originally used to describe enzymes that have two or more receptor sites, one of which (the active site) binds the principal substrate, whereas the other(s) bind(s) effector molecules that can influence the enzyme's biological activity. More generally, it is used to describe the indirect coupling of distinct sites within a protein, mediated by conformational changes.
[47]
10w
Refers to agents that enhance memory or other cognitive functions.
[48]
446w
increases the tonic conductance in granule cells of the dentate gyrus and cerebellum without modifying phasic currents 89 . Similarly, the high sensitivity of α 4 βδ receptors to ethanol 174,175 is mirrored by the selective augmentation of tonic inhibition in granule cells of the dentate and alcohol. In line with the selective enhancement of the GABA responsiveness of δ subunit-containing receptors by neurosteroids 124,129,172,173 , a low concentration of 3α,21-dihydroxy-5α-pregnan-20-one (allotetrahydrodeoxycorticosterone or THDOC) significantly de novo in neurons and glia, or generated by the metabolism of circulating precursors that originate in peripheral steroidogenic organs 212 . One pathway for glial neurosteroidogenesis is shown: cholesterol is taken into mitochondria by the action of the steroidogenic acute regulatory protein (StAR) and the peripheral benzodiazepine receptor (PBAR). Pregnenolone is formed by the action of the enzyme cytochome P450 side chain cleavage (P450scc). In this example, subsequent metabolism in the smooth endoplasmic reticulum through progesterone and 5α-dihydroprogesterone (5α-DHP) leads to the formation of 3α,5αtetrahydroprogesterone (3α,5α-THP; allopregnanolone). Secretion from the glial cell is indicated by a bold arrow. b | The steroid 3α,21-dihydroxy-5α-pregnan-20-one (allotetrahydrodeoxycorticosterone, THDOC) is secreted from the adrenal gland and is also formed in the brain from its peripherally secreted precursor, deoxycorticosterone. THDOC differentially modulates α 1 β 3 γ 2L and α 1 β 3 δ receptors when they are expressed in human embryonic kidney 293T cells. Currents evoked from α 1 β 3 γ 2L receptors are minimally affected by THDOC (1 µM), but those from α 1 β 3 δ receptors are markedly enhanced. c | Low concentrations of THDOC selectively enhance the magnitude of the tonic current that is mediated by extrasynaptic α 6 βδ receptors in cerebellar granule cells (blue receptors in a), with little effect on synaptic responses (yellow receptors in a). Current values were averaged over 10-ms epochs at 100-ms intervals. Horizontal bars indicate the application of THDOC (grey) and the GABA A (γ-aminobutyric acid type A) antagonist SR-95531 (gabazine) (black). The dashed line is the mean current observed after complete block of GABA A receptors, which was used to calculate the magnitude of the tonic GABA A receptor-mediated conductance. This conductance (top panel) is increased in the presence of both 10 and 100 nM THDOC, whereas spontaneous inhibitory postsynaptic potentials (IPSCs) (lower traces) are minimally affected. The histogram shows the effects of THDOC on the tonic conductance (green) and average charge transfer through phasic IPSCs (blue) expressed as a percentage of control values in the absence of THDOC (dashed line). Error bars denote standard error mean, and asterisks denote statistical significance. nA, nanoAmp; pA, picoAmp. Panel b modified, with permission, from REF. www.nature.com/reviews/neuro R E V I E W S