The general anaesthetic etomidate inhibits the excitability of mouse thalamocortical relay neurons by modulating multiple modes of GABAA receptor-mediated inhibition

Modulation of thalamocortical (TC) relay neuron function has been implicated in the sedative and hypnotic effects of general anaesthetics. Inhibition of TC neurons is mediated predominantly by a combination of phasic and tonic inhibition, together with a recently described ‘spillover’ mode of inhibition, generated by the dynamic recruitment of extrasynaptic γ-aminobutyric acid (GABA)A receptors (GABAARs). Previous studies demonstrated that the intravenous anaesthetic etomidate enhances tonic and phasic inhibition in TC relay neurons, but it is not known how etomidate may influence spillover inhibition. Moreover, it is unclear how etomidate influences the excitability of TC neurons. Thus, to investigate the relative contribution of synaptic (α1β2γ2) and extrasynaptic (α4β2δ) GABAARs to the thalamic effects of etomidate, we performed whole-cell recordings from mouse TC neurons lacking synaptic (α10/0) or extrasynaptic (δ0/0) GABAARs. Etomidate (3 μm) significantly inhibited action-potential discharge in a manner that was dependent on facilitation of both synaptic and extrasynaptic GABAARs, although enhanced tonic inhibition was dominant in this respect. Additionally, phasic inhibition evoked by stimulation of the nucleus reticularis exhibited a spillover component mediated by δ-GABAARs, which was significantly prolonged in the presence of etomidate. Thus, etomidate greatly enhanced the transient suppression of TC spike trains by evoked inhibitory postsynaptic potentials. Collectively, these results suggest that the deactivation of thalamus observed during etomidate-induced anaesthesia involves potentiation of tonic and phasic inhibition, and implicate amplification of spillover inhibition as a novel mechanism to regulate the gating of sensory information through the thalamus during anaesthetic states.


Introduction
The ability of general anaesthetics to impair consciousness, and induce analgesia, amnesia and immobility, has been exploited clinically for over a century (Rudolph & Antkowiak, 2004). Yet, despite their routine clinical use, the neuroanatomical substrates of anaesthetic actions remain enigmatic, with several components of the circuitry regulating the sleep-wake cycle implicated to date (Franks, 2008). Nevertheless, electrophysiological, neuroimaging and modelling studies consistently pinpoint the thalamus as an important neural locus for anaesthetic-induced hypnosis (Angel, 1991;Fiset et al., 1999;Alkire et al., 2000;White & Alkire, 2003;Ching et al., 2010;Andrada et al., 2012). Such observations are consistent with the classical role of the thalamus in controlling transitions between conscious states (Steriade et al., 1993).
At the cellular level, c-aminobutyric acid (GABA) A receptors (GABA A Rs) represent key molecular targets for several general anaesthetics, including the intravenous agents etomidate and propofol (Hales & Lambert, 1991;Orser et al., 1994;Uchida et al., 1995). Importantly, anaesthetic-induced modulation of inhibition may be influenced by the molecular identity of GABA A Rs (Belelli et al., 1996;Bonin & Orser, 2008), which are composed of five subunits drawn from a pool of 19 isoforms (Olsen & Sieghart, 2009). Etomidate is unique in this respect, displaying selectivity for b2or b3-containing receptors vs. those incorporating the b1 subunit (Hill-Venning et al., 1997). Subunit composition additionally influences the functional properties and spatiotemporal expression of GABA A Rs. Thus, synaptic GABA A Rs, which mediate brief phasic inhibition, are composed of axbxc2 subunits, whereas receptors incorporating a d rather than a c2 subunit, are localized to extra-or peri-synaptic locations, and mediate persistent tonic inhibition (Farrant & Nusser, 2005). However, a substantial number of GABA A Rs with a typically 'synaptic' subunit composition also exist extrasynaptically, although the functional role of such receptors remains to be established (Kasugai et al., 2010). d-GABA A Rs have been proposed as important general anaesthetic targets, a suggestion supported by their abundant expression in brain regions implicated in some anaesthetic behaviours, including thalamocortical (TC) relay nuclei (Belelli et al., 2005;Jia et al., 2005). However, while etomidate enhances both phasic and tonic currents recorded from TC neurons (Belelli et al., 2005), it is unknown how these respective actions influence the excitability of relay neurons.
Recently, we demonstrated that in addition to generating tonic inhibition, d-GABA A Rs of TC neurons may be recruited via a spilloverlike mechanism to prolong phasic inhibition generated in response to presynaptic spike bursts (Herd et al., 2013). This observation may have particular relevance to anaesthetic effects in thalamus, given that nucleus reticularis thalami (nRT) neurons, which provide the major source of GABA-ergic innervation to TC neurons, exhibit prominent spike bursts during unconscious states. Whether and how anaesthetics may influence this 'burst-mediated spillover' mode is not known.
Here, we investigated the relative contribution of synaptic and extrasynaptic GABA A Rs to the inhibitory effects of etomidate on relay neuron output. Using transgenic mice in which TC relay neurons lack synaptic (a1 0/0 , Peden et al., 2008) or extrasynaptic (d 0/0 , Herd et al., 2009) GABA A Rs, we report that, although enhancement of tonic inhibition dominates the inhibitory effect of etomidate on TC neuron excitability, the synaptic a1bc2 GABA A R population also significantly influences the effects of the anaesthetic. Moreover, we demonstrate that etomidate greatly prolongs burst-mediated spillover inhibition, providing a novel mechanism whereby a putatively relevant network may be influenced by anaesthetics.

Materials and methods
Use of a1 0/0 and d 0/0 mice as models to investigate the thalamic effects of etomidate Compensatory adaptations represent a significant caveat to the interpretation of experiments utilizing constitutive knockout mice. Indeed, biochemical and immunohistochemical evidence revealed compensatory up-regulation of a2 and/or a3 subunits in the neocortex and cerebellum of a1 0/0 mice (Sur et al., 2001;Kralic et al., 2002a,b). However, in the thalamus of a1 0/0 mice, the ablated a1 subunits are not replaced by a2 or a3 subunits (Kralic et al., 2006) and a3 expression remains unaltered in the presynaptic nRT (Kralic et al., 2006). Thus, at the functional level, the lack of any clear upregulation of the remaining palette of GABA A R subunits is reflected by a complete lack of miniature inhibitory postsynaptic currents (mIPSCs) in mature ventrobasal (VB) neurons, but unaltered mIPSC properties in the nRT (Peden et al., 2008). Moreover, extrasynaptic GABA A Rs do not appear to compensate for the lost phasic inhibition, as tonic inhibition is not significantly altered in a1 0/0 VB neurons, at least at the limited age range studied here . In the d 0/0 mouse (Mihalek et al., 1999), increased expression of c2 subunits has been reported (Tretter et al., 2001;Peng et al., 2002). Importantly, however, receptors incorporating a1, a2, a3 or a5 subunits do not appear to substitute for the lost d-containing receptors (Peng et al., 2002). In agreement, we did not observe any clear alterations in the properties of relay neuron mIPSCs that could compensate for the greatly reduced tonic inhibition in d 0/0 mice . Thus, the lack of clear adaptive changes in the properties of phasic and tonic inhibition at the thalamic level in the a1 0/0 and d 0/0 mice validates their utility as models to investigate the influence of etomidate on relay neuron excitability.
The a1 0/0 and d 0/0 mice were generated on a mixed C57BL6/ J-129SvEv (a1 0/0 ) or single C57BL6 (d 0/0 ) background at the Merck Research Laboratories at the Neuroscience Research Centre in Harlow and at the University of Pittsburgh, respectively, as described previously (Mihalek et al., 1999;Sur et al., 2001). Brain slices were prepared from the first two generations of a1 0/0 and d 0/0 and their respective wild-type (WT) mice from breeding pairs derived from the corresponding heterozygous mice bred at the University of Dundee. Measures of inhibition (phasic and tonic) and neuronal excitability [input resistance (IR) and action potential rheobase] did not differ between a1 +/+ and d +/+ (data not shown). Results obtained from the respective WT strains were thus pooled.

Electrophysiology
Electrophysiological recordings were performed from thalamic slices maintained in warmed ECS (30°C), using an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA, USA). Thalamocortical relay neurons of the VB complex were visually identified with an Olympus BX51 microscope (Olympus, Southall, UK), equipped with differential interference contrast/infrared optics and a charge-coupled device camera. Whole-cell voltage-and current-clamp recordings from VB neurons were obtained using patch pipettes prepared from thick-walled borosilicate glass (King Precision Glass, Claremont, CA, USA) using a Narishige PC-10 vertical puller, and had open tip resistances of 4-6 MΩ when filled with a solution containing [in mM): 130 K-gluconate, 2 KCl, 2 NaCl, 10 HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid), 0.2 EGTA (ethylenediaminetetraacetic acid), 2 Mg adenosine triphosphate, 0.5 Na guanosine-5 0 -triphosphate, 10 Tris-phosphocreatine, adjusted to pH 7.2-7.3 with KOH, 280-290 mOsm. Series resistance was intermittently monitored, compensated for (up to 80% for voltage-clamp, 100% for current-clamp) and recordings discarded if values changed by more than 25%, or increased beyond 25 MΩ. Voltage-clamp recordings were performed at a holding potential of À50 mV (GABA E revÀ 85 mV), and thus responses mediated by GABA A Rs were outwardly directed. Inhibitory responses were evoked in VB neurons using a bipolar tungsten electrode (World Precision Instruments, Stevenage, UK), placed within the borders of the nRT. Stimuli were delivered (20-ls duration, 0.1 Hz) using a Digitimer DS3 Isolated Stimulator (Digitimer, Welwyn Garden City, UK), with stimulus intensity adjusted to the minimum value that avoided failures, but still generated IPSC bursts of relatively consistent amplitude [i.e. for each recording, the excitatory IPSC (eIPSC) peak amplitude standard deviation (SD) was <15% of the mean peak amplitude obtained from that cell; WT = 8.5 AE 1.4%, n = 7; a1 0/0 = 10.5 AE 1.4%, n = 7; d 0/0 = 8.9 AE 0.8%]. At the stimulation intensities utilized, the number of peaks observed in response to each stimulus was relatively consistent from trial to trial (SD = 0.94 and 1.30 for WT and d 0/0 , respectively). Although we find GABA B R activation is minimal under these recording conditions/stimulation parameters (data not shown), we performed the experiments in the presence of CGP 55845 (1 lM) to block GABA B Rs. Holding currents (I hold ) were continuously monitored throughout voltage-clamp recordings to determine the effect of etomidate on tonic inhibition. Current-clamp recordings were also performed to determine the effect of burst-mediated inhibition on VB neuron spiking. Thus, evoked inhibitory postsynaptic potentials (eIPSPs) were evoked at regular intervals (0.09-0.11 Hz) during tonic trains of action potentials. Tonic spike trains were generated by depolarizing the recorded neuron to a supra-threshold membrane potential (V m ) using depolarizing direct current (DC) injection delivered via the recording electrode. The amplitude of injected DC was carefully adjusted to maintain a tonic spike frequency of~4-8 Hz. In this paradigm, eIPSPs generated a transient cessation of action potential discharge. eIPSPs were also recorded at or near the resting potential, and in some cells a rebound low-threshold Ca 2+ potential was observed, which was commonly crowned by a high-frequency spike burst. In such recordings, the latency to rebound, relative to eIPSP onset, was determined for each event before and after etomidate application. VB neuron excitability was also examined in currentclamp experiments by monitoring action potential output in response to a family of current steps. The stimulation protocol consisted of a set of hyperpolarizing and depolarizing current pulses (À200 to +300 pA, 400-ms duration, in increments of 50 pA, every 15 s), delivered via the recording electrode before and after the bath application of etomidate. Current stimulations were delivered from a holding potential of À66 mV to isolate the tonic spike mode during depolarizing stimulations. The number of action potentials generated in response to positive current steps, and within low-threshold Ca 2+ current (I T )-mediated rebound bursts following release from hyperpolarization, were counted for each condition. Membrane IR was calculated according to Ohms's law following delivery of small (À20 pA, 300 ms) hyperpolarizing current steps from the resting membrane potential. A measured liquid junction potential of 11 mV was corrected offline. Recordings were filtered at 2 or 5 kHz for voltage-and current-clamp recordings, respectively, and acquired directly to PC using a NI-DAQmx analog-to-digital interface (National Instruments UK, Newbury, UK) for subsequent offline analysis.

Data analysis
All electrophysiological data were analysed offline using the Strathclyde Electrophysiology Software, WinEDR/WinWCP (J. Dempster, University of Strathclyde, UK) software package. For voltage-clamp experiments, eIPSCs were analysed with respect to peak amplitude, 10-90% rise times and charge transfer. The decay time course following the final peak of multi-peak eIPSCs was adequately described by fitting a double exponential function (y(t) = A 1 e (Àt/s1) + A 2 e (Àt/s2) ) using the least squares method, where A is amplitude, t is time and s is the decay time constant. Analysis of the SD of residuals and use of the F-test was used to compare goodness of fit, confirming an improved fit of the decay was achieved when employing a double-rather than a mono-exponential function. Thus, a weighted decay time constant (s w ) was also calculated according to the equation, s w = s 1 P 1 + s 2 P 2 , where s 1 and s 2 are the decay time constants of the first and second exponential functions and P 1 and P 2 are the proportions of the synaptic current decay described by each component. To determine the effect of etomidate on VB tonic inhibition, etomidate-induced outward shifts of the holding current (I hold ) were quantified by sampling sequential 51.2-ms epochs for a 2-min period before, and after application of etomidate (Belelli et al., 2005). Epochs containing evoked or spontaneous synaptic currents were excluded. The mean holding current in each condition was determined from Gaussian fits to the distribution of holding current values sampled from each epoch.
For current-clamp experiments investigating the effect of etomidate on VB neuron excitability, input-output curves were constructed by plotting the number of action potentials generated as a function of stimulus amplitude. To quantify etomidate-induced shifts in inputoutput curves, data were fit with a Boltzman sigmoidal curve according to the equation y(x) = N max /1 + e (X -X 50 )/S + N max ), where N max is the number of action potentials generated at the maximum stimulus amplitude (in this case 300 pA), X is the stimulus amplitude, X 50 is the stimulus amplitude producing half-maximal spike output (hereafter 'EA 50 ') and S is the slope factor. For experiments investigating the effect of eIPSPs on tonic spike trains, inter-spike intervals (ISIs) were measured throughout the spike train. For baseline measurements, ISIs were determined for a 5-s period preceding each eIPSP (minimum of three stimulations) to ensure spike rates recovered to stable frequencies following inhibition. The ISI during an IPSP was simply the interval between the final spike prior to delivery of an eIPSP and the first spike following recovery from inhibition. The transient suppression of tonic spike activity induced by an eIPSP was observed as an immediate and obvious increase in the ISI (see Figs 2 and 3).
All results are reported as the arithmetic mean AE standard error of the mean (SEM). Statistical significance of the mean data was assessed with a Student's t test (paired, or unpaired), one-way analysis of variance (ANOVA) or a two-way mixed-design ANOVA as appropriate, using the SPSS (SPSS Inc., Chicago, IL USA) software package. Sphericity was assessed with Mauchly's test, and degrees of freedom were adjusted using a Greenhouse-Geisser correction if sphericity was violated. Post-hoc multiple comparisons were conducted using Tukey's test. Unless explicitly stated, statistical significance was assessed using Student's t-test. Boltzmann curve fits were performed in Origin v7 (OriginLab, Northampton, MA, USA).

Salts and drugs
R-(+)-etomidate (supplied by Merck, Glasgow, UK) was prepared as a concentrated aqueous stock and diluted 1000-fold in ECS to a final concentration of 3 lM. CGP 55835 (Abcam, Cambridge, UK) was prepared in dimethyl sulphoxide (DMSO) as a 1000-fold concentrated stock solution and diluted in ECS to the final desired concentration. The final DMSO concentration (0.1%) had no effect on any of the measured parameters (data not shown). Etomidate was applied via the perfusion system (2-4 mL/min) and allowed to infiltrate the slice for a minimum of 10 min. All other salts and drugs were obtained either from Sigma-Aldrich (Poole, UK), Tocris Bioscience (Bristol, UK) or VWR (Lutterworth, UK).

Etomidate prolongs eIPSCs by potentiating synaptic and extrasynaptic GABA A Rs
In VB neurons, etomidate prolongs mIPSCs and enhances tonic currents mediated by a1b2c2and by a4b2d GABA A Rs, respectively (Belelli et al., 2005). However, our recent demonstration that thalamic d-GABA A Rs may influence the kinetics of action potential-dependent phasic inhibition, particularly in response to spike bursts (Herd et al., 2013), provides an additional route whereby etomidate may influence VB inhibition. Thus, we investigated the effects of etomidate on electrically evoked IPSCs generated in response to extracellular stimulation of the nRT of WT, a1 0/0 and d 0/0 mouse brain slices.
In voltage-clamp recordings utilising K + -gluconate-based pipette solutions (GABA A R E rev~À 85 mV, V h = À50 mV), extracellular stimulation of the nRT in WT brain slices generated outwardly directed eIPSCs composed of multiple peaks, and a bi-exponential decay phase following the final peak (Table 1, Fig. 1A). The multipeak nature of the eIPSCs is consistent with the generation of I T -dependent spike bursts in presynaptic neurons. Application of etomidate (3 lM) greatly increased the eIPSC peak amplitude (Table 2, Fig. 1A and D, n = 7, P < 0.001), and s w ( Table 2, Fig. 1A and E, P < 0.001), resulting in a significant enhancement of eIPSC charge transfer (Table 2, Fig. 1A and F, P = 0.001). Consistent with our previous report (Herd et al., 2013), robust residual eIPSCs were recorded from a1 0/0 VB neurons, despite the absence of synaptic GABA A Rs (Table 1, Fig. 1B). The slow kinetics of the rising and decaying phase of a1 0/0 eIPSCs, together with their sensitivity to modulation by the d-selective compound DS2 (Wafford et al., 2009;Herd et al., 2013;Jensen et al., 2013), indicate that such residual events are mediated by recruitment of d-containing extrasynaptic (e)GABA A Rs in response to nRT spike bursts. Importantly, acute application of etomidate greatly enhanced the peak amplitude ( Table 2, Fig. 1B and D, n = 7, P < 0.001) and s w ( Table 2, Fig. 1B and E, P < 0.001) of a1 0/0 eIPSCs, leading to a significant increase in their charge transfer (Table 2, Fig. 1F, P < 0.001). However, the overall effect of etomidate on eIPSC charge transfer was reduced for a1 0/0 relative to WT neurons ( Fig. 1F, drug 9 genotype interaction, F 2,19 = 7.21, P = 0.005, mixed ANOVA), suggesting that a significant proportion of the effects of etomidate on WT eIPSCs are mediated by synaptic GABA A Rs. In support, etomidate also enhanced eIPSCs recorded from d 0/0 neurons (Fig. 1C). In agreement with previous findings (Herd et al., 2013), eIPSCs recorded from d 0/0 VB neurons exhibited faster decay kinetics, due to the abolished eGABA A R component normally engaged by neurally released GABA acting on WT neurons (Table 1). Nevertheless, despite the absence of the eGABA A R contribution to eIPSC properties, etomidate (3 lM) significantly increased eIPSC peak amplitude (Table 2, Fig. 1C and D, n = 8, P = 0.003), s w (Table 2, Fig. 1C and E, P < 0.001) and charge transfer (Table 2, Fig. 1F, P = 0.003). However, as observed for a1 0/0 recordings, deletion of the d subunit significantly reduced the overall effect of etomidate on eIPSC charge transfer relative to WT (Fig. 1F, drug 9 genotype interaction, F 2,19 = 7.21, P = 0.005, mixed ANOVA).
Etomidate prolongs phasic suppression of VB tonic firing and delays post-inhibitory rebound burst firing During conscious brain states, thalamocortical relay neurons exist in a relatively depolarized state, and respond to incoming excitatory inputs in the tonic 'relay' mode (Steriade et al., 1993). Given the shared roles of synaptic and extrasynaptic GABA A Rs in regulating the kinetics of inhibitory synaptic transmission (see above), we hypothesized that these receptor populations should combine to regulate the effect of etomidate on phasic inhibition of tonic VB spike output.
In a subset of these recordings, eIPSPs recorded at or slightly depolarized to the resting membrane potential were of sufficient amplitude to generate a rebound low-threshold Ca 2+ potential and associated spike burst at the offset of the eIPSP (von Krosigk et al., 1993). We therefore investigated the effect of etomidate on the postinhibitory rebound properties by measuring the latency to rebound. Consistent with the prolongation of evoked phasic inhibition, etomidate significantly delayed the timing of rebound bursts in WT recordings ( Fig. 4A and C, n = 4, P = 0.009) at similar V m (pre-eIPSP V m = À68.3 AE 1.2 and À69.0 AE 1.2 mV for control and etomidate, respectively). Similarly, in d 0/0 recordings, the latency to post-inhibitory rebound was significantly increased in the presence of etomidate ( Fig. 4B and C, n = 5, P < 0.001) at equivalent V m (pre-eIPSP   , a1 0/0 (dark grey bars, n = 7) and d 0/0 (light grey bars, n = 8) thalamic slices. **P < 0.01, ***P < 0.001 vs. control, paired t-test. † P < 0.05, † † P < 0.01 vs. WT, mixed ANOVA, Tukey's post-hoc test. Etom, etomidate.
V m = À67.8 AE 0.3 and À68.3 AE 0.2 mV for control and etomidate, respectively). eIPSPs recorded from a1 0/0 slices did not elicit rebound bursts, although in all seven recordings, the duration of eIPSPs recorded from equivalent membrane potentials were prolonged (data not shown). These observations imply that both synaptic and extrasynaptic receptor modulation can influence post-inhibitory rebound timing.
Deletion of the a1 or d subunit influences the etomidateinduced enhancement of the VB tonic current While quantifying the effects of etomidate on eIPSCs, we simultaneously monitored the VB holding current in each mouse strain to further understand the contribution of a1bc2 and a4bd GABA A Rs Table 3. Comparison of resting membrane potential (V m ) and input resistance recorded from WT, a1 0/0 and d 0 to the effects of the anaesthetic on VB tonic inhibition. Bath application of 3 lM etomidate induced a clear outward shift in VB holding current (37.2 AE 5.0 pA, n = 8; Fig. 5A and G), with a time course similar to that observed for potentiation of eIPSCs (Fig. 5B, panels i and ii). In previous work, we found that deletion of a1-containing receptors did not significantly influence baseline tonic current amplitudes, consistent with the hypothesis that VB tonic inhibition is mediated predominantly by a4b2d receptors (Chandra et al., 2006;Herd et al., 2009). Thus, it was surprising that while etomidate application induced a clear outward current in a1 0/0 VB recordings (20.7 AE 2.9 pA, n = 9) concurrently with potentiation of residual eIPSCs ( Fig. 5C and D), the magnitude of the outward current shift was significantly reduced relative to WT (Fig. 5G, drug 9 genotype interaction, F 2,21 = 20.11, P < 0.001, mixed ANO-VA). In recordings from VB neurons of d 0/0 mice, bath application of 3 lM etomidate induced only a minimal outward current (5.0 AE 1.0 pA, n = 7; P < 0.001 vs. WT; P = 0.012 vs. a1 0/0 , mixed ANOVA, Tukey's post-hoc test; Fig. 5E and G), while robustly potentiating eIPSCs (Fig. 5F).
Deletion of a1 or d subunits attenuates the effect of etomidate on VB neuron excitability Given that etomidate modulation of VB tonic inhibition is significantly reduced in a1 0/0 and d 0/0 neurons, such genetic manipulations are thus predicted to influence the effect of the anaesthetic on cellular excitability. To investigate this hypothesis, we examined the effect of etomidate on VB input-output (I-O) relationships generated in response to a family of current steps (À200 to +300 pA, 400 ms duration, 50-pA increments every 15 s). Current steps were delivered while neurons were held at a membrane potential of À66 mV; at such potentials, I T is largely inactivated (Coulter et al., 1989). First, we compared I-O relationships across WT, a1 0/0 and d 0/0 under control conditions to determine if basic electrical excitability was affected by the mutations (Fig. 6). Across all three strains, in accordance with classical TC neuron physiology, hyperpolarizing currents steps of sufficient amplitude generated a sagging, hyperpolarization-activated cation current (I H )-dependent membrane potential response, followed by a rebound low-threshold Ca 2+ potential, crowned by a high-frequency spike burst (spikes per burst, WT = 2-10, n = 23; a1 0/0 = 1-9, n = 12; d 0/0 2-10, n = 22). Although the amplitude of the hyperpolarizing current step significantly influenced the mean number of spikes per rebound burst (F 3,162 = 104.80, P < 0.001, mixed ANOVA), there was no significant effect of genotype on this relationship (F 6,162 = 1.00, P = 0.43, mixed ANOVA, Fig. 6D). Supra-threshold depolarizing current steps of increasing amplitude generated tonic spikes of increasing frequency in VB neurons derived from each mouse strain (Fig. 6A-C and E). Although visual inspection of the resulting I-O curves (Fig. 6E) suggests a modest left-ward trend of the relationship towards increased excitability for d 0/0 relative to WT recordings, this did not reach statistical significance (F 5,215 = 1.46, P = 0.20, mixed ANOVA). Indeed, estimation of the current amplitude generating half maximal spike output (EA 50 ) from Boltzmann sigmoidal curve fits (see Methods) indicated a small, but non-significant effect of d subunit deletion on the I-O EA 50 (WT = 173 AE 7 pA, n = 23; a1 0/0 = 168 AE 10 pA, n = 12; d 0/0 = 158 AE 8 pA, n = 22, F 2,54 = 1.00, P = 0.37, one way ANOVA). Similarly, neither the resting membrane potential nor IR of VB neurons was significantly influenced by deletion of the d subunit, although again a trend towards an increased IR was observed relative to WT (IR, WT = 288 AE 13 MΩ, n = 38; d 0/0 = 328 AE 20 MΩ, n = 27, P = 0.1, unpaired t-test, Table 3).
In accordance with the large potentiation of tonic inhibition, bath application of 3 lM etomidate hyperpolarized the resting V m , reduced membrane IR and exerted a clear reduction in the excitability of WT VB neurons (Table 4, Fig. 7A-C). Thus, etomidate significantly reduced the number of spikes per rebound burst upon release from hyperpolarizing stimulations, although the effect was particularly evident for smaller amplitude current steps ( Fig. 7A  curves was significantly increased in the presence of the anaesthetic (EA 50 : control = 170 AE 12 pA; + etomidate = 195 AE 12 pA, n = 12, P < 0.001). By contrast, although etomidate also hyperpolar-ized the resting V m and decreased the membrane IR of a1 0/0 VB neurons (Table 4), the consequent reduction of their excitability was less than for WT neurons (Fig. 7D-F). These findings are consistent with  and d 0/0 (C) VB neurons, illustrating voltage responses to a family of hyperpolarizing and depolarizing current steps (À200 to +300 pA, 50 pA increments). For clarity, the responses to only a selection of tested current steps are shown, as indicated in the stimulation protocol (bottom). (D) Graph summarizing the number of spikes occurring within rebound bursts upon offset from hyperpolarizing current steps in recordings from WT (black squares, n = 23), a1 0/0 (dark grey circles, n = 12) and d 0/0 (light grey triangles, n = 22) VB neurons. (E) Graph comparing the number of spikes generated in response to depolarizing current steps of increasing amplitude in VB neurons derived from WT, a1 0/0 and d 0/0 mice. Symbols and n numbers are as indicated for D. The input-output relationship for each strain is fitted with a Boltzmann sigmoid curve. the reduced enhancement of tonic inhibition by the anaesthetic following a1 subunit deletion. This was particularly evident upon inspection of the I-O relationship generated from depolarizing stimulations, where the rightward shift in the curve induced by etomidate (EA 50 : control = 166 AE 11 pA, + etomidate = 178 AE 13 pA, n = 11, P = 0.03) was significantly reduced relative to WT recordings ( Table 5). Deletion of the d subunit virtually abolished the effect of etomidate on VB burst and tonic firing (Fig. 7G-I), consistent with the limited etomidate-induced outward current observed in d 0/0 voltage-clamp experiments. In addition, for d 0/0 neurons, etomidate had minimal effect on the resting V m , or on the membrane input resistance (Table 4). Thus, with the exception a small effect at the smallest current step, which was significantly reduced relative to that observed in WT recordings (P = 0.019 unpaired t-test), etomidate did not significantly influence rebound intra-burst spike number in response to hyperpolarizing current steps (Fig. 7H). Similarly, for d 0/0 neurons, the I-O curve generated in response to positive current steps remained relatively unaffected by etomidate (EC 50 : control = 163 AE 14 pA, + 3 lM etomidate = 163 AE 15 pA, n = 9, P = 0.95, Fig. 7I; Table 5).

Discussion
In this study, we investigated the relative contribution of synaptic and extrasynaptic GABA A Rs to the etomidate-induced inhibition of neuronal activity in a network putatively involved in general anaesthetic mechanisms. Specifically, deletion of the d subunit reduced the effect of etomidate on VB tonic inhibition. Consequently, the ability of etomidate to inhibit burst and tonic spiking was virtually abolished in d 0/0 mice, suggesting that etomidate facilitation of tonic inhibition exerts a dominant effect on TC cell output. Unexpectedly, deletion of the a1 subunit also reduced the impact of etomidate on VB tonic inhibition and the relay neuron I-O relationship, thus implicating the a1-GABA A R population during anaesthetic actions in thalamus. Importantly, etomidate also greatly prolonged the duration of nRT burst-mediated phasic inhibition, by potentiating both synaptic and extrasynaptic GABA A Rs. Thus, further to conventional phasic and tonic inhibition, burst-mediated spill-over may represent a novel mechanism whereby anaesthetics can modulate circuit function.
While anaesthetics probably exert their behavioural effects by altering neuronal activity across multiple brain regions (Franks, 2008), human neuro-imaging studies have identified a consistent thalamic deactivation during anaesthetic-induced sedation and/or loss of consciousness (Fiset et al., 1999;Alkire et al., 2000;White & Alkire, 2003;Hofbauer et al., 2004;Gili et al., 2013). Such studies are consistent with the canonical view of the thalamus as a sensory 'gate' that governs the transmission of multi-modal information to the cortex (Steriade, 2003). In vivo, anaesthetics may modulate thalamic activity directly by influencing the intrinsic excitability of resident thalamic neurons or indirectly by altering the activity of extrinsic excitatory inputs (e.g. descending cortico-thalamic, or ascending reticular activating system inputs). Indeed, single-unit recordings from cats showed a reduction of the spontaneous tonic firing rate of TC neurons by etomidate, alongside similar decreases in the neocortex and reticular formation (Andrada et al., 2012). Computer simulations utilizing thalamocortical network models have also provided important insights, revealing that elevated inhibition of cortical and thalamic neurons imposes a slowed, hyper-synchronous rhythmicity upon the thalamocortical loop that is reminiscent of electroencephalography dynamics during anaesthetized states (Talavera et al., 2009;Ching et al., 2010;Purdon et al., 2013). Collectively, such studies favour a co-ordinated action of anaesthetics across key interconnected networks, rather than a specific localized effect. Nevertheless, our data support the hypothesis that etomidate mediates at least a component of its actions by a direct reduction of TC neuron output, resulting from enhanced synaptic and extrasynaptic GABA A R function. In agreement, propofol and the volatile agent isoflurane also inhibit relay neuron tonic spiking in a manner that is at least partially GABA A R-dependent (Ying & Goldstein, 2005a,b;Jia et al., 2008).
That some general anaesthetics exert their spectrum of behavioural effects via a selective interaction with the GABA A R is well established (Hales & Lambert, 1991;Franks & Lieb, 1994). In support, mutant mice harbouring point-mutated GABA A Rs engineered to be etomidate-insensitive (b2N265S; Belelli et al., 1997;Reynolds et al., 2003) or propofoland etomidate-insensitive (b3N265M; Pistis et al., 1999;Jurd et al., 2003) display dramatically altered behavioural responses to these anaesthetics, revealing that specific anaesthetic behaviours can be assigned to individual GABA A R populations. Specifically, the sedative and immobilizing effects of both etomidate and propofol are mediated by b2and b3-containing receptors, respectively, whereas both receptor populations contribute to the hypnotic component (Jurd et al., 2003;Reynolds et al., 2003). However, these studies did not address which b2or b3expressing neuronal populations may be involved in the respective behaviours. Thus, it is noteworthy that the effects of etomidate on relay cell phasic and tonic currents are greatly reduced in b2N265S slices (Belelli et al., 2005). In light of the present evidence for a reduced etomidate-induced inhibition of relay neuron spiking in a1 0/0 and d 0/0 mice, such studies implicate synaptic (a1b2c2) and extrasynaptic (a4b2d) GABA A Rs as important molecular targets for the thalamic actions of this anaesthetic.
The effect of etomidate on VB tonic inhibition was virtually abolished in d 0/0 mice, leading to a reduced inhibitory effect on burst and tonic spiking. Such observations are consistent with the tonic conductance generating an inhibitory charge which greatly outweighs that produced by brief, spontaneous phasic inhibition. In agreement, computational studies utilizing a thalamocortical network model suggest that enhanced tonic inhibition is critical for the etomidate-induced depression of thalamic firing (Talavera et al., 2009). Thus, the significant effect of the a1 subunit deletion (albeit lesser than for the d 0/0 ) was unexpected. One potential interpretation Table 4. Comparison of the effect of etomidate on the resting membrane potential (V m ) and input resistance of WT, a1 0/0 and d 0/0 VB neurons WT (n = 19) a1 0/0 (n = 12) d 0/0 (n = 11) suggests that a subpopulation of d-GABA A Rs may contain the a1 subunit instead of a4. Indeed, a1 can form a functional partnership with the d subunit in some hippocampal interneurons (Glykys et al., 2007). However, while a1 subunit deletion abolishes spontaneous phasic inhibition in VB (Peden et al., 2008), we did not detect a significant reduction of baseline tonic inhibition, or the current induced by the d-selective agonist THIP (4,5,6,7-tetrahydroisoxazolo [5,4-c]pyridin-3-ol), which does not discriminate between a1bd and  a4bd receptors (B. Ebert, personal communication; Herd et al., 2009), suggesting that a1bd receptors are not abundantly expressed in VB neurons. Alternatively, the loss of on-going spontaneous phasic inhibition, which is robustly prolonged by etomidate in WT neurons, may partially account for the reduced effect of the anaesthetic on spike discharge for a1 0/0 relay neurons. Additionally, a substantial pool of a1bc2 receptors may exist extrasynaptically (Kasugai et al., 2010). While this receptor population does not appear to contribute to baseline tonic inhibition, etomidate, which increases GABA A R channel open probability in response to low GABA concentrations (Yang & Uchida, 1996), may enhance the activation of extrasynaptic a1b2c2 receptors by ambient GABA. Finally, we cannot exclude that direct activation of a1b2c2 receptors by etomidate (Hill-Venning et al., 1997) may influence VB neuron excitability. Using paired nRT-VB recordings, we recently reported that peri/ extrasynaptic a4bd receptors are recruited during nRT burst firing, leading to a substantial prolongation of burst-mediated phasic inhibition (Herd et al., 2013). In support of this, despite the absence of synaptic GABA A Rs, we observed a residual eIPSC in a1 0/0 VB neurons, which was greatly potentiated by the d-selective positive allosteric modulator DS2 (Herd et al., 2013). Here, we observed a similar potentiation of the residual eIPSC upon application of etomidate to a1 0/0 slices. Consequently, the remarkable ability of the residual eIPSPs to suppress tonic spiking of a1 0/0 neurons was greatly prolonged by etomidate. In a4 0/0 and d 0/0 mice, eIPSCs are shorter due to the abolished spill-over component, and are insensitive to DS2, consistent with the loss of a4bd receptors (Herd et al., 2013). In contrast to DS2, in d 0/0 recordings, etomidate significantly prolonged eIPSCs and therefore the eIPSP-induced suppression of tonic spiking. Curiously, for the latter paradigm, the duration of spike suppression induced by the eIPSP in the presence of etomidate was not significantly different from WT in either a1 0/0 or d 0/0 neurons (Fig. 3D), despite a clear effect of both gene deletions under control conditions (Fig. 2D). In current-clamp recordings the duration of IPSPs is determined partially by the kinetic properties of GABA A Rs, but also by the ensuing activation of voltage-dependent conductances (e.g. I H and I T ), which may act to truncate the IPSP time course (Thomson et al., 2000). Thus, we speculate that the duration of the IPSP may be maximized in the presence of the anaesthetic and that this 'ceiling effect' may be achieved via the potentiation of either synaptic or extrasynaptic GABA A Rs in the absence of the complementary receptor population. Nevertheless, the results suggest that etomidate facilitates burst-mediated inhibition via a combined action at synaptic and peri/extrasynaptic GABA A Rs, leading to an amplified inhibitory effect on VB tonic firing.
As noted above, the strength of synaptic inhibition arising from nRT neurons interacts with intrinsic voltage-dependent conductances to control the timing of post-inhibitory rebound bursts, and thus the period and synchrony of thalamus-dependent oscillations, which prevail in the unconscious brain (Huguenard & McCormick, 2007). In addition to an enhanced tonic inhibition, which induces a membrane hyperpolarization that is likely to favour burst firing (Cope et al., 2005), etomidate also delayed the timing of post-inhibitory rebound burst firing by prolonging the IPSP. Given the importance of burst firing during thalamocortical rhythmogenesis, the etomidate-induced increase of IPSP duration may therefore contribute to the slowed oscillatory activity of the anaesthetized brain.
Our in vitro data, together with a computational study (Talavera et al., 2009), suggest that the etomidate-induced inhibition of TC output is dependent on synaptic and extrasynaptic GABA A Rs. Thus, it is surprising that deletion of a1, a4 or d subunits exerts little or no influence on the sedative or hypnotic actions of etomidate (Mihalek et al., 1999;Kralic et al., 2003;Iyer et al., 2013). However, the interpretation of behavioural data obtained from mice harbouring global GABA A R subunit deletions is complicated by the non-uniform up-regulation of alternative GABA A R isoforms following subunit loss (Sur et al., 2001;Kralic et al., 2002a,b;Peng et al., 2002). This caveat is exemplified by studies investigating the subunit-dependence of benzodiazepine-induced behaviours; mice expressing a benzodiazepine-insensitive a1 subunit (H101R) are resistant to the sedative effects of diazepam (Rudolph et al., 1999;McKernan et al., 2000), yet a1 0/0 mice exhibit enhanced diazepaminduced sedation (Kralic et al., 2002b). Thus, caution is warranted when comparing results from a slice preparation demonstrating minimal functional GABA A R compensation to behavioural studies underpinned by complex interactions between multiple brain networks subject to varying degrees of compensation. We note that an as yet unidentified neuronal adaptation (e.g. as described by Brickley et al., 2001) is likely to compensate for the loss of tonic or phasic inhibition in relay neurons, as demonstrated by the lack of any significant alteration in membrane IR or the I-O relationship in the mutant mice. Crucially, however, such adaptations do not appear to include upregulation of the remaining palette of GABA A R subunits (Kralic et al., 2006;Peden et al., 2008;Herd et al., 2009).
In summary, we have shown that etomidate reduces the electrical excitability of thalamocortical relay neurons via a specific interaction with synaptic and extrasynaptic GABA A Rs. Given the established role of the thalamus in arbitrating conscious state transitions, our results support the putative role of these distinct thalamic GABA A R populations in general anaesthetic actions. Finally, we suggest that burst-mediated spill-over, a kinetically intermediate form of inhibition that bridges the kinetic extremes of fast phasic and slow tonic inhibition, represents a novel means by which GABA-modulatory anaesthetics may influence the relay of sensory information from thalamus to cortex. current; IPSP, inhibitory postsynaptic potential; IR, membrane input resistance; ISI, inter-spike interval; I T , current mediated by low threshold T-type Ca 2+ channels; mIPSC, miniature IPSC; nRT, nucleus reticularis thalami; TC, thalamocortical; VB, ventrobasal complex; V m , membrane potential; WT, wildtype; a1 0/0 , a1 subunit knock-out mouse; a1-GABA A R, a1 subunitcontaining GABA A receptor; b3N265M, asparagine to methionine mutation at position 265 of the GABA A receptor b3 subunit; b2N265S, asparagine to serine mutation at position 265 of the GABA A receptor b2 subunit; d 0/0 , d subunit knock-out mouse; d-GABA A R, d subunitcontaining GABA A receptor; s w , weighted decay time constant.