Acting locally but sensing globally: impact of GABAergic synaptic plasticity on phasic and tonic inhibition in the thalamus


Corresponding author S. G. Brickley: Biophysics Section, Imperial College London, South Kensington Campus, London SW7 2AZ, UK. Email:


We have discovered that adult thalamocortical relay neurones exhibit a sustained enhancement of synaptic inhibition triggered by transient action potential firing of a single thalamic relay neurone. The sustained activity-dependent increase in IPSC frequency (+48.3 ± 11.4%, n= 32) was blocked by chelating calcium inside an individual cell, by scavenging nitric oxide or by blocking NMDA receptor activation in the thalamus. Surprisingly, the tonic inhibition that is known to result from extrasynaptic GABAA receptor activation in these cells was unaffected by this local form of plasticity. However, tonic inhibition was increased (+131.9 ± 56.5%, n= 13) following widespread changes in GABA release across the thalamus. These data suggest that thalamocortical sleep-state oscillations requiring membrane hyperpolarization will be influenced by global sensing of GABA release acting through extrasynaptic GABAA receptors. In contrast, local changes in GABA release of the type observed following this novel form of activity-dependent plasticity will influence local integration of sensory information without changing levels of tonic inhibition.

The ability to switch a thalamocortical (TC) relay neurone into burst firing mode is a key step in generating specific sleep states (Steriade, 2005). Interactions between the various intrinsic conductances responsible for generating TC burst firing critically depend upon membrane hyperpolarization. One of the most widely accepted synaptic mechanisms for bringing about this membrane hyperpolarization involves increased potassium ion permeability due to GABAB receptor activation (Williams et al. 1995). It is now apparent that, in addition to the conventional forms of transient synaptic responses that underlie phasic conductances (IPSCs), a tonic GABAAR-mediated conductance (GGABA) is also present in some thalamic nuclei (Porcello et al. 2003; Belelli et al. 2005; Cope et al. 2005; Jia et al. 2005; Chandra et al. 2006; Bright et al. 2007). Tonic inhibition of this type refers to the continuous activation of extrasynaptic δ subunit-containing GABAA receptors (δ-GABAARs) by low concentrations of ambient GABA (Brickley et al. 1996, 2001). It is now widely accepted that δ-GABAARs are ideally suited to the generation of this form of tonic inhibition in a variety of brain regions but the physiological function of GGABA is only beginning to be appreciated (Farrant & Nusser, 2005). It has recently been demonstrated (Cope et al. 2005) that enhancing tonic inhibition due to extrasynaptic δ-GABAA receptor activation can also favour burst firing of relay neurones due to the resulting membrane hyperpolarization (Chandra et al. 2006). The importance of understanding the modulation of GGABA in the thalamus is further indicated by the fact that δ-GABAARs are emerging drug targets in the treatment of epilepsy (Maguire et al. 2005), stress (Shen et al. 2007) and sleep disorders (Winsky-Sommerer et al. 2007).

The functional significance of IPSCs in the thalamus is clear but the significance of GGABA for controlling TC relay neurone excitability is not fully appreciated and the impact of raising GABA release on GGABA has not been established in any brain region. What we do know is that GABA release clearly contributes to GGABA in both cerebellar granule neurones (Brickley et al. 1996; Wall & Usowicz, 1997; Hamann et al. 2002) and dentate granule neurones of the hippocampus (Glykys & Mody, 2007), and that GGABA can be pharmacologically enhanced by drugs that act on δ-GABAARs such as neurosteroids (Stell et al. 2003), anaesthetics (Belelli et al. 2005) and the sleep-promoting drug gaboxadol (Chandra et al. 2006). A vesicular independent form of GGABA has also been described in cerebellar granule neurones that can be enhanced by blocking GABA uptake (Wall & Usowicz, 1997; Rossi et al. 2003). In contrast, GGABA in TC relay neurones appears to be entirely vesicular dependent (Bright et al. 2007). However, it has yet to be determined whether increases in vesicular GABA release can alter GGABA in adult TC neurones although mechanisms for increasing vesicular GABA release have been reported. For example, addition of group 1 metabotropic glutamate receptor agonists can stimulate GABA release on to TC neurones (Govindaiah & Cox, 2006) and increased GABA release at thalamic synapses has also been documented following application of nitric oxide (NO) donors (Yang & Cox, 2007). These studies highlight possible candidate mechanisms for increasing GABA release in the thalamus, but none of these studies has described a physiologically relevant stimulus for triggering this process. In the current study we describe a novel activity-dependent mechanism for enhancing GABA release in the adult thalamus and quantify the impact of this plasticity on both phasic (activation of synaptic receptors; sIPSCs) and tonic inhibition (activation of extrasynaptic receptors; GGABA).


Tissue preparation

Acute brain slices were obtained from mature (at least 1 month postnatal) male C57Bl/6 mice, prepared in accordance with the Animals (Scientific Procedures) Act 1986. In brief, following cervical dislocation, brains were rapidly removed and placed in ice-cold slicing solution composed of (mm): NaCl 85, KCl 2.5, CaCl2 1, MgCl 4, NaH2PO4 1.25, NaHCO3 26, sucrose 75, glucose 25, pH 7.4 when bubbled with 95% O2 and 5% CO2. Coronal slices (250 μm thick) were cut (DSK Super Zero 1; Dosaka EM, Kyoto, Japan) at the level of the hippocampus and slices were incubated at 37°C for 30 min, after which the high sucrose slicing solution was replaced with normal recording solution. Following electrophysiological recording, slices were fixed in phosphate-buffered saline (PBS) containing 4% paraformaldehdye for 24 h. Slices were then washed in PBS, before being mounted in Vectashield mounting medium (Vector Laboratories, Burlingame, CA, USA). The fluorescent dye Lucifer yellow or Alexa 488 was included in the electrode solution (0.5 mg ml−1; Sigma, Poole, UK) to allow for later confocal imaging of filled neurones. A Zeiss LSM 510 confocal microscope (Carl Zeiss, Jena, Germany) was used to visualize filled cells using the 458 nm line of an argon laser. Neurones were z-sectioned into 1 μm optical slices with signal-to-noise improved by averaging at least two scans for each z-section. Three-dimensional morphological reconstruction of confocal images was performed using NIH-image.


Slices were perfused at physiological temperatures (37–38°C) with recording solution containing (mm): NaCl 125, KCl 2.5, CaCl2 2, MgCl 1, NaH2PO4 1.25, NaHCO3 26, glucose 25, pH 7.4 when bubbled with 95% O2 and 5% CO2. TC neurones in the dLGN were visually identified using a Zeiss Axioskop FS microscope (Jena, Germany). Whole-cell recordings were made under voltage-clamp and current-clamp configuration using an Axon Multiclamp 700B amplifier (Molecular Devices, Union City, CA, USA). Recording electrodes were pulled from thick-walled borosilicate glass (GC-150F-10; Harvard Apparatus, Edenbridge, UK) and had a resistance of 4–6 MΩ when filled with intracellular solution, containing (mm): KCl 140, MgCl2 5, Hepes 10, EGTA 5, phosophocreatine (disodium salt) 8, Na-ATP 3, Na-GTP 0.1; the pH was adjusted to 7.3 with KOH. In some experiments EGTA was replaced by 10 mm BAPTA in order to chelate Ca2+. During whole-cell recording, GABAA receptor-mediated conductances were pharmacologically isolated by inclusion of the ionotropic glutamate receptor blocker kynurenic acid (1 mm; Sigma) in the recording solution. For the experiments where cells were depolarized under voltage clamp, we also added 2 mm QX-314 (Tocris) to the intracellular solution to prevent the firing of action potentials (APs). In other experiments, bath application of tetrodotoxin (TTX; Sigma) was used to block APs. Additional drugs, 2-(4-carboxyphenyl)-4,4,5,5-tetramethylimidazoline-1-oxyl-3-oxide potassium (PTIO); 2R-amino-5-phosphonopentanoate (AP5), and S-nitroso-N-acetylpenicillamine (SNAP) were all purchased from Ascent Scientific (UK) and bath applied at the given concentration during whole-cell recording.

Induction of GABAergic synaptic plasticity

GABAergic synaptic plasticity was induced in a number of ways during this study. Burst firing of TC relay neurones was performed in current-clamp configuration. Following standard bridge-balancing techniques, a series of hyperpolarizing current pulses of 300 ms duration were delivered to each cell. A typical switch to current-clamp is shown in Fig. 1A. The size of the hyperpolarizing current steps was incrementally increased in 10 pA steps and low threshold rebound bursts were elicited from the relay neurone at 50 pA above threshold (see Fig. 1F). Each burst contained a maximum of five APs occurring at a mean frequency of 282.2 ± 25.6 Hz (n= 9). The total number of APs delivered during burst firing protocols ranged from 5 to 64. Tonic AP firing was also elicited in current-clamp configuration using steady-state depolarizing current injection. A range of 14 to 103 APs were generated during tonic firing experiments at a mean frequency of 178.9 ± 3.0 Hz (n= 7). Plasticity was also induced by applying depolarizing voltage steps to the relay neurone under voltage clamp using ten 250 ms voltage steps to 0 mV, delivered at a frequency of 2 Hz. In some experiments, we applied below-threshold hyperpolarizing steps in current clamp and these experiments led to the ‘no firing’ data presented in Fig. 1F. Finally, bath application of the NO donor PTIO (100 μm) was also used to elicit GABAergic synaptic plasticity across the entire dLGN.

Figure 1.

Activity-dependent plasticity in dLGN relay neurones
A, continuous record illustrating the rapid switch between voltage-clamp (−60 mV) and current-clamp during which burst firing was induced in the relay neurone using hyperpolarizing current injection protocols. B, plot of spontaneous IPSC peak amplitude against time indicating the duration of the switch to current clamp and the block of all sIPSCs by SR95531 at the end of the experiment. C, superimposed average sIPSC waveforms constructed before and after the induction of GABAergic synaptic plasticity. Average sIPSC waveforms were used to calculate the 10–90% rise-times, the peak amplitude and the weighted decay constant. The inset illustrates that IPSC peak amplitude and IPSC decay do not alter following this plasticity (n= 9). D, continuous current record illustrating the block of GGABA by bath application of 100 μm SR95531. On the left of this panel are two all-point histograms (1 pA bin widths) constructed from 1 s epochs at the start and end of the illustrated current trace with superimposed Gaussian fits. The mean current level estimated from these fits were used to calculate GGABA. E, scatter plot of normalized sIPSC frequency against GGABA in relay neurones after induction of GABAergic synaptic plasticity. F, examples of burst and tonic firing patterns that were used to elicit the GABAergic synaptic plasticity. The left hand traces show a TC relay neurone's response to a subthreshold hyperpolarizing current step of 130 pA followed by a 230 pA current injection that was 50 pA above threshold for burst firing. The right hand trace shows tonic firing of another TC relay neurone in response to steady-state depolarization. G, quantification of the normalized sIPSC frequency against the number of APs used to induce this plasticity using either the burst (n= 8) or tonic (n= 6) AP firing protocol. ‘No firing’ refers to control experiments in which the TC neurone was switched to current-clamp without eliciting APs (n= 8). For comparison, the grey filled area illustrates the sIPSC frequency increase (±s.e.m.) following voltage-clamp-induced depolarization (n= 18).

Data analysis

On-line data acquisition and offline analysis were performed using the Stathclyde Electrophysiology software WinEDR/WinWCP (John Dempster, University of Strathclyde, Glasgow, UK). Current and voltage records were low-pass filtered at 10 kHz and acquired at 20 kHz using a National Instruments board (NI-DAQmx, PCI-6221). Seal resistance was calculated from the steady-state current response to hyperpolarizing steps applied in the cell-attached configuration. Recordings were only subjected to further analysis if the seal resistance was greater than 1 GΩ. In whole-cell configuration, series and input resistances were calculated using peak and steady-state current responses to a 10 mV hyperpolarizing voltage step. The membrane capacitance was determined using a weighted time constant for the decay of the capacitive current transient, calculated as the integral of the transient divided by the peak current. Spontaneous IPSCs were detected using the WinEDR software with amplitude- and kinetics-based criteria (events were accepted when they exceeded a threshold of 4–6 pA for 0.5 ms). Average IPSC waveforms were constructed using events detected over 30 s epochs (at least 50 uncontaminated IPSCs), and in the case of drug applications, these waveforms were calculated at least 60 s after the application. Weighted decay constants were calculated as the integral of the IPSC divided by the peak current. GGABA was defined as the current blocked by SR95531 at the end of each experiment. The mean current values used to obtain GGABA estimates were calculated from Gaussian fits to all-point histograms (1 pA bin-width) constructed from 1 s current regions devoid of sIPSCs. To account for cell-to-cell variability, GGABA was normalized to membrane capacitance (pS pF−1). The percentage change in GGABA following any given manipulation was based upon the mean current amplitude before and after the procedure expressed relative to the current recorded in SR95531 at the end of each experiment. All values are given as mean ±s.e.m. In all subsequent analysis, statistical tests were performed using STATISTICA 5.1 (StatSoft, Tulsa, OK, USA) and considered significant at P < 0.05. A Shapiro–Wilk test was used to determine whether measures were normally distributed and differences between groups were examined using the Student's t test. When distributions were not normal the Mann–Whitney U test or Wilcoxon matched-pair test was used. Correlations were assessed using the Spearman Rank test.


Transient AP firing induces a sustained increase in vesicular GABA release

The functional properties of TC relay neurones in the dorsal lateral geniculate nucleus (dLGN) of adult mice (60 ± 3 postnatal days, n= 24) were studied at physiologically relevant temperatures of 37–38°C (Bright et al. 2007). In control conditions, the mean sIPSC frequency recorded from these adult TC relay neurones was 8.1 ± 1.6 Hz (n= 35) and the mean holding current blocked by 100 μm SR95531 at the end of each experiment was −67.5 ± 20.5 pA (n= 74). As shown in Fig. 1, following AP burst firing, we observed a clear increase in sIPSC frequency (+47.4 ± 7.7%, n= 8) that was maintained during the subsequent recording period (up to 20 min). On average (n= 9), there was no change in the 10–90% rise-time (356 ± 28 μ before versus 472 ± 51 μ after), peak amplitude (−18 ± 3 pA before versus−18 ± 3 pA after) or decay (3.2 ± 0.3 ms before versus 3.4 ± 0.2 ms after) of sIPSCs following this GABAergic synaptic plasticity. Surprisingly, there was little change in the holding current following this increase in release, consistent with no significant change in GGABA (+28.6 ± 29.3%). Sustained AP firing during steady-state depolarization (see Fig. 1F) also resulted in a significant increase in the sIPSC frequency (+53.2 ± 24.8%, n= 6) with no significant change in GGABA (−17.4 ± 19.0%). Using a Spearman Rank test, there was no correlation apparent between the number of APs and the magnitude of the increase in sIPSC frequency for either burst or sustained AP firing. Moreover, AP firing was not essential for this plasticity since simply depolarizing the relay neurone under voltage-clamp with ten 250 ms voltage steps to 0 mV, delivered at a frequency of 2 Hz, also induced a significant increase in sIPSC frequency (+46.9 ± 18.8%n= 18). Once again there was no significant change in GGABA (−2.6 ± 8.0%). After pooling data from all plasticity experiments (n= 32), it was clear that the 48.3 ± 11.4% increase in IPSC frequency was not associated with an increase in GGABA (−13.4 ± 11.5%).

Raising local GABA release does not increase GGABA

The vesicular dependence of GGABA in adult TC neurones (Bright et al. 2007) makes it somewhat surprising that GGABA was not altered when GABA release frequency was raised. Figure 2 illustrates a recording from a typical TC relay neurone in which both phasic and tonic inhibition were reduced by the AP blocker TTX. Given that TTX application will reduce GABA release across all synapses in the thalamus it is clear that ambient GABA concentrations can be reduced in the vicinity of the recorded relay neurone. In contrast, blocking GABA uptake can increase GGABA in the thalamus (Chandra et al. 2006) but it is still conceivable that spatially localized increases in GABA release are not sufficient to overcome this efficient uptake, explaining why the enhanced GABA release we observe following TC plasticity does not result in increased GGABA. Therefore, we examined in more detail the relationship between sIPSC frequency and GGABA to test the hypothesis that local changes in GABA release do not dictate the magnitude of GGABA. Analysing data from a large number of TC relay neurones (Fig. 2B) demonstrates that there is little correlation between the magnitude of GGABA and sIPSC frequency. Obviously, the total number of GABA release sites is not known and we are also unaware of the extrasynaptic δ-GABAAR occupancy in the slice preparation. However, in any given neurone, there is no obvious correlation between the instantaneous sIPSC frequency and the holding current immediately prior to each recorded sIPSC (Fig. 2E). This analysis would further suggest that it is not possible to increase GGABA with local changes in release. Previous pharmacological methods for elevating GABA concentrations have involved manipulating release across the entire thalamus (Govindaiah & Cox, 2006; Yang & Cox, 2007) or blocking uptake (Chandra et al. 2006). In contrast, we have examined the impact of a sustained enhancement in local GABA release, without pharmacological intervention, to show that ambient GABA concentrations are not sensitive to local changes in GABA release.

Figure 2.

Tonic inhibition is not modulated by local changes in GABA release
A, a continuous current record illustrating the reduction in sIPSC frequency and block of GGABA that occurs following application of 500 nm TTX. After electrophysiological recording, filled cells were z-sectioned into 1 μm optical slices on a confocal microscope. The subsequent reconstruction highlights the somatic, dendritic and axonal compartments of this typical relay neurone. On average (n= 6), a mean capacitance of 90.3 ± 19.6 pF was predicted following reconstruction (assuming a specific membrane capacitance of 0.01 pF μm−2) compared to 76.5 ± 3.9 pF measured during voltage-clamp experiments. Measured and predicted values were not significantly different. Therefore, in all subsequent analysis we have relied upon the measured cell capacitance to normalize conductance changes to cell size. B, scatter plot demonstrating little correlation (R2= 0.2) between sIPSC frequency and GGABA across all relay neurones (n= 100). Values have been normalized to cell size using measured capacitance values. C, a continuous current record obtained at a command voltage of −60 mV. Dashed line indicates the zero current level. D, individual sIPSCs demonstrating the measurement of holding current and baseline variance immediately prior to each sIPSC. E, scatter plots illustrating no relationship between sIPSC instantaneous frequency and GGABA (quantified both as the baseline variance and holding current) in an individual neurone. A similar lack of correlation has been observed for 20 relay neurones in which a sufficiently large number of sIPSCs (at least 1000) had been obtained.

GABAergic synaptic plasticity requires calcium influx, NO release and NMDA receptor activation

Although relatively rare compared to glutamatergic plasticity, sustained enhancement of GABAergic transmission has now been reported in a number of mammalian brain areas. As shown in Fig. 3A, the GABAergic synaptic plasticity we describe on to adult glutamatergic relay neurones is blocked by bath application of the NO scavenger PTIO (20 μm; n= 7). Chelation of calcium inside the cell (10 mm BAPTA; n= 11) and bath application of the NMDA receptor antagonist AP5 (50 μm; n= 10) also both block the enhancement of GABA release. In the case of NMDA receptor blockade it was possible to reverse the blocking action of AP5 on wash-out (n= 8). Therefore, the pharmacology of this plasticity most resembles the activity-dependent plasticity recently described in dopaminergic neurones of the juvenile rat brain ventral tegmental area (VTA) where calcium influx into glutamatergic neurones initiates increased GABA release from local interneurones following NMDA receptor-dependent production of NO (Nugent et al. 2007). However, the long-term plasticity in the VTA requires high-frequency stimulation of large numbers of afferent inputs. In contrast, the robust GABAergic synaptic plasticity we describe in the thalamus can be induced following AP firing in a single relay neurone.

Figure 3.

A global increase in GABA release does result in increased tonic inhibition
A, continuous recording from a TC relay neurone in voltage- and current-clamp (grey shading) illustrating how GABAergic synaptic plasticity is blocked by addition of the NO scavenger PTIO (20 μm) in the external solution. B, bar chart of the normalized sIPSC frequency calculated following the plasticity protocols. This plasticity is clearly blocked in the presence of external PTIO as, on average, no change in sIPSC frequency was observed. Recordings from relay neurones with 10 mm BAPTA in the internal solution also blocked this plasticity. The NMDA receptor antagonist d-AP5 (50 μm) was also shown to block this plasticity but the ability to induce GABAergic synaptic plasticity was restored following AP5 wash-out. C, continuous current record from a relay neurone illustrating that the NO donor SNAP (100 μm) induced a clear increase in both the sIPSC frequency and the holding current. On the left of this plot are all-point histograms constructed from the holding current recorded in control conditions, in the presence of the NO donor SNAP, and with the GABAA receptor antagonist SR95531 present. The data were fitted with Gaussian functions (smooth curves) to calculate the mean current level in order to estimate changes in GGABA. D, comparison of changes in sIPSC frequency and GGABA following GABAergic synaptic plasticity and SNAP application. Note the clear increase in both GGABA and IPSC frequency using the NO donor SNAP and the similar increase in sIPSC frequency observed following this activity-dependent plasticity with no significant change in GGABA.

Increasing global GABA release does increase GGABA

Pharmacological methods for enhancing vesicular GABA release have previously been reported in the thalamus. For example, addition of group 1 metabatropic glutamate receptor agonists stimulates GABA release on to relay neurones (Govindaiah & Cox, 2006) as does application of nitric oxide (NO) donors (Yang & Cox, 2007). Neither of these studies examined the consequence of increased GABA release on GGABA. However, Fig. 3C clearly illustrates that when GABA release was enhanced on to adult dLGN relay neurones with the NO donor SNAP (100 μm), we observed a clear increase in the holding current that was blocked by application of the GABAA receptor antagonist SR95531. On average, application of 100 μm SNAP resulted in a 54.4 ± 24.0% increase in the sIPSC frequency and a significant 131.9 ± 56.5% increase in the magnitude of GGABA (n= 13).


This is the first study to demonstrate that enhancing GABA release in the thalamus can lead to an increase in tonic inhibition. More importantly, the results presented here show that a localized increase in GABA release does not alter GGABA and tonic inhibition is only likely to be increased when GABA release is enhanced in a global manner across the thalamus. This is also the first study to demonstrate the presence of an NMDA- and NO-dependent form of plasticity that is characterized by a robust and sustained increase in the frequency of GABA release on to adult TC relay neurones. This GABAergic synaptic plasticity requires calcium influx into a single relay neurone following AP firing or depolarization and is therefore quite different to the drug-induced methods of enhanced GABA release that have previously been reported in the thalamus. However, it is currently not clear how a sustained enhancement of GABA release is produced following AP firing or TC relay neurone depolarization. TC relay neurone axon collaterals are reported to give rise to large synaptic profiles which make contact with local interneurones in the cat dLGN (Bickford et al. 2008) and it has been suggested that this putative excitatory input generates feed-back inhibition on to TC relay neurones (Cox et al. 2003). There are currently no data that directly demonstrate the functional presence of this synaptic connection in the mouse dLGN but, nonetheless, these collaterals could theoretically be responsible for glutamate release on to local interneurones leading to the NMDA receptor activation and NO production that is necessary for this form of GABAergic synaptic plasticity. Gap junctions between thalamic relay neurones have also been reported in the cat dLGN (Hughes et al. 2002) and it is conceivable that this relatively sparse coupling could lead to excitation of adjacent neurones. It is also possible that the action of a retrograde messenger released from TC relay neurones is triggering an NMDA receptor- and NO-dependent mechanism that leads to a sustained increase in transmitter release from GABAergic terminals arising from either local dLGN interneurones or from the strong GABAergic reticular input into the thalamus. Obviously, future studies are required to distinguish between these and other possibilities (Parri et al. 2001) but the current study has concentrated on the impact of this activity-dependent increase in GABA release for both phasic and tonic inhibition within the thalamus.

The level of GABA release clearly contributes to GGABA in a number of brain regions (Farrant & Nusser, 2005) and GGABA can be pharmacologically enhanced by drugs that modulate δ-GABAARs directly (Stell et al. 2003; Belelli et al. 2005; Chandra et al. 2006) or by blocking GABA uptake (Wall & Usowicz, 1997; Rossi et al. 2003; Chandra et al. 2006). However, few studies have described activity-dependent mechanisms for enhancing GGABA in the brain and none have studied the impact of raising GABA release in a physiologically relevant manner on tonic inhibition. Surprisingly, the data presented in this study suggest that extrasynaptic δ-GABAARs in the thalamus are insensitive to local changes in release and instead act as global sensors of GABA release originating from more widespread synapses. Therefore, when an NO donor is used to increase GABA release more globally throughout the thalamus a large increase in GGABA is apparent. It remains to be established whether this insensitivity to local changes in GABA release reflects efficient GABA uptake or indicates that the high-affinity δ-GABAARs are saturated by ambient GABA concentrations in the thalamus. However, it appears that the potential for GGABA to hyperpolarize the membrane will only be realized once levels of GABA release increase in a concerted manner across the thalamus. This could be particularly important when considering the function of GGABA in switching relay neurones into burst firing mode during non-REM sleep (Cope et al. 2005; Chandra et al. 2006). Our data would suggest that those oscillations associated with membrane hyperpolarization would be selectively supported by this specialized sensor of global GABA release in the thalamus. In contrast, GGABA will not respond to the transient local changes in GABA release that occur during the integration of afferent sensory information (Le Masson et al. 2002).



We would like to thank Professors Nicholas Franks and William Wisden for useful comments on this manuscript. Morphological reconstructions were performed by Doran Amos. This work was supported by the Wellcome Trust.

Author's present address

D. P. Bright: Department of Neuroscience, Physiology and Pharmacology, University College London, Gower Street, London WC1E 6BT, UK.