Acute effect of carbamazepine on corticothalamic 5–9-Hzand thalamocortical spindle (10–16-Hz) oscillations in therat

Thomas W. Zheng,1,2,3 Terence J. O’Brien,3 Sofya P. Kulikova,1,2 Christopher A. Reid,4 Margaret J. Morris5 andDidier Pinault1,21Neuropsychologie cognitive et physiopathologie de la schizophr�enie, INSERM U1114, Strasbourg, France2F�ed�eration de M�edecine Translationnelle de Strasbourg (FMTS), NeuroPole de Strasbourg, Facult�e de m�edecine, Universit�e deStrasbourg, INSERM U1114, 11 rue Humann, Strasbourg, 67085, France3Department of Medicine, The Royal Melbourne Hospital, The University of Melbourne, Parkville, Vic., Australia4Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, Vic., Australia5Department of Pharmacology, University of New South Wales, Kensington, NSW, Australia

Keywords: 5–9-Hz oscillations, drowsiness, EEG, intracellular, sleep spindles, somatosensory system

Abstract

A major side effect of carbamazepine (CBZ), a drug used to treat neurological and neuropsychiatric disorders, is drowsiness, astate characterized by increased slow-wave oscillations with the emergence of sleep spindles in the electroencephalogram(EEG). We conducted cortical EEG and thalamic cellular recordings in freely moving or lightly anesthetized rats to explore theimpact of CBZ within the intact corticothalamic (CT)–thalamocortical (TC) network, more specifically on CT 5–9-Hz and TC spindle(10–16-Hz) oscillations. Two to three successive 5–9-Hz waves were followed by a spindle in the cortical EEG. A single systemicinjection of CBZ (20 mg/kg) induced a significant increase in the power of EEG 5–9-Hz oscillations and spindles. Intracellularrecordings of glutamatergic TC neurons revealed 5–9-Hz depolarizing wave–hyperpolarizing wave sequences prolonged byrobust, rhythmic spindle-frequency hyperpolarizing waves. This hybrid sequence occurred during a slow hyperpolarizing trough,and was at least 10 times more frequent under the CBZ condition than under the control condition. The hyperpolarizing wavesreversed at approximately �70 mV, and became depolarizing when recorded with KCl-filled intracellular micropipettes, indicatingthat they were GABAA receptor-mediated potentials. In neurons of the GABAergic thalamic reticular nucleus, the principal sourceof TC GABAergic inputs, CBZ augmented both the number and the duration of sequences of rhythmic spindle-frequency bursts ofaction potentials. This indicates that these GABAergic neurons are responsible for the generation of at least the spindle-frequencyhyperpolarizing waves in TC neurons. In conclusion, CBZ potentiates GABAA receptor-mediated TC spindle oscillations. Further-more, we propose that CT 5–9-Hz waves can trigger TC spindles.

Introduction

Carbamazepine (CBZ) is a widely prescribed drug for the treatmentof focal epileptic seizures, as well as several other neurobiologicaldiseases, such as schizophrenia, bipolar disorders, and neuropathicpain (Post, 1982; Simhandl & Meszaros, 1992; Leucht et al., 2007).CBZ is ineffective in the treatment of typical absence epilepsy,aggravating the seizures in some patients (Genton, 2000) and in ani-mal models (Liu et al., 2006), and has even been reported to triggerthe de novo generation of epileptic discharges in humans (Monjiet al., 2004). Owing to the tricyclic structural features of CBZ, it isalso has a broad spectrum of action, resulting in significant side

effects, including drowsiness, weight gain, and seizure aggravation(Levy et al., 1985; Brady, 1989; Liu et al., 2006). Apart from itsprimary therapeutic mechanism of use-dependent Na+ channelblockage, CBZ interacts directly with several molecular targets,including GABAA receptors (Granger et al., 1995; Liu et al., 2006),adenosine A1 receptors (Biber et al., 1999), and Ca2+ currents(Walden et al., 1993). Importantly, CBZ has been shown to directlypotentiate GABAA currents in recombinant receptors expressed inhuman embryonic kidney cells (Granger et al., 1995) and Xenopusoocytes (Liu et al., 2006). Deciphering the mechanisms of CBZaction in living intact neural networks is likely to help increase ourunderstanding of its efficacy and side-effect profiles.The thalamocortical (TC) system is an appropriate model with

which to study the effects of neuroactive substances on GABAA recep-tor-mediated oscillations (von Krosigk et al., 1993; Ulrich & Hugue-nard, 1997; Crunelli et al., 2011). The TC neurons are glutamatergic

Correspondence: Didier Pinault, 2F�ed�eration de M�edecine Translationnelle de Stras-bourg (FMTS), INSERM U1114, Universit�e de Strasbourg, as above.E-mail: pinault@unistra.fr

Received 19 June 2013, revised 21 October 2013, accepted 4 November 2013

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons Ltd

European Journal of Neuroscience, pp. 1–12, 2013 doi:10.1111/ejn.12441

European Journal of Neuroscience

and are the principal cells of the dorsal thalamus. They are reciprocallyconnected with both the neocortex and the GABAergic thalamic reticu-lar nucleus (TRN), this latter structure being an important source ofGABA receptor-mediated inhibition in TC neurons (Pinault, 2004). Inthe rodent, at least two types of GABAA receptor-mediated oscillationhave been investigated: wake-related corticothalamic (CT) 5–9-Hzoscillations and sleep-related TC spindles. Sleep spindles are a uniqueEEG signature of the early stages of sleep, and start to emerge duringdrowsiness (De Gennaro & Ferrara, 2003). They may be involved inprocesses involved in memory consolidation (Tamminen et al., 2010).They are characterized by short-lasting (< 2 s) bouts of medium-volt-age (< 100 lV) 12–14-Hz oscillations in humans, 7–14-Hz oscilla-tions in cats (Contreras & Steriade, 1996), and 10–16-Hz oscillationsin rodents (Pinault et al., 2001). They generally have a singular sinu-soidal waveform that progressively increases and then graduallydecreases in amplitude. Sleep spindles are generated in TC neurons,via GABAA receptor-mediated mechanisms driven by TRN neurons(von Krosigk et al., 1993; Steriade, 1995; Ulrich & Huguenard, 1997).Wake-related CT 5–9-Hz oscillations have been recorded in

rodent sensory cortices, especially during quiet immobile wakeful-ness (Pinault et al., 2001) and sensory cortical responses (Tort &Katz, 2010). The function of this type of rhythmic activity is stillunknown. The thalamic intracellular events associated with 5–9-Hzand spindle oscillations are distinguishable (Pinault et al., 2006). InTC neurons, 5–9-Hz oscillations are characterized by a rhythmicexcitatory postsynaptic potential (EPSP) wave–inhibitory postsynap-tic potential (IPSP) wave sequence that occurs during a slow hyper-polarizing trough. The EPSP wave (summation of unitary EPSPs)can trigger a low-threshold Ca2+ potential crowned by a high-fre-quency burst of action potentials (Pinault, 2003). Oscillations of5–9 Hz usually emerge from a membrane potential that is moredepolarized, with more fast rhythmic activities than the membranepotential from which spindle oscillations emerge (Pinault et al.,2006). In TC neurons, the rhythmic 5–9-Hz and spindle IPSPs aremediated by activation of GABAA receptors.Therefore, our aim was to test the hypothesis that CBZ-induced

drowsiness is associated with CT–TC oscillations involving GABAA

receptor-mediated mechanisms. The hypothesis was tested in the ratsomatosensory CT–TC system, by the use of intracellular recordingof glutamatergic TC neurons and extracellular recordings of GAB-Aergic TRN neurons under light anesthesia. The present results indi-cate that CBZ potentiates GABAA receptor-mediated TC spindles.

Materials and methods

Animals

Experiments were conducted in inbred adult male rats (n = 39),which originate from a colony of Wistar-derived non-epileptic con-trol rats that are selectively bred in our laboratory so that they donot show absence-related spike-and-wave discharges (Vergnes et al.,1982). All procedures were approved by the Comit�e r�egionald’�ethique en mati�ere d’exp�erimentation animale (CREMEAS), theUniversit�e de Strasbourg (AL/01/23/11/07), and the University ofMelbourne Animal Ethics Committee (AEC #0810999), and wereperformed in accordance with the guidelines published by theAustralian National Health and Medical Research Council (7th Edn,2004) and the European Union (directive 2010/63/EU) for the useof animals in research. All experimental procedures also compliedwith the ARRIVE guidelines (Kilkenny et al., 2010). Rats were atleast 13 weeks of age, and weighed 250–350 g at the time of theexperiments. Rats were housed and kept under controlled

environmental conditions (12-h light/dark cycle; lights on at07:00 h), with food and water available ad libitum. Recordings wereperformed in the middle of the light phase (10:00 to 16:00 h).

Anesthesia and surgery

Adult male rats (n = 8) were chronically implanted with extraduralEEG electrodes (stainless steel screws; bilateral electrode placements– frontal parietal bone, 0.5 mm anterior to the coronal suture;ground and reference – parietal bone, 2 mm anterior to the lambdoidsuture) under general anesthesia (isoflurane, 5% induction, 1.5–2.5%maintenance). Rats were given a 7-day recovery period.The acute experiments (cortical EEG, juxtacellular and intracellu-

lar recordings) were conducted under light anesthesia (neurolepticanalgesia; 31 rats). Details of all procedures are available in previ-ous reports (Pinault, 2003; Kulikova et al., 2012). The light anesthe-sia was induced by continuous intravenous injection of thefollowing mixture (quantity given per hour for a 300-g rat): fentanyl(2 lg) and glucose (25 mg). Muscle rigidity and tremors wereblocked with intravenous administration of D-tubocurarine chloride(0.4 mg/h). The rats were artificially ventilated in the pressure mode(8–12 cm H2O; 60 beats/min) with an O2-enriched gas mixture(50% air/50% O2). The rat’s rectal temperature was maintained at itsphysiological level (37–38.3 °C) with a thermoregulated blanket.The cortical EEG and the heart rate were also continuously moni-tored to maintain a steady depth of light anesthesia, either by givinga bolus or adjusting the injection rate of the sedating mixture. Thedepth of anesthesia was ascertained by the occurrence of slowwaves in the EEG. Recording sessions started 3 h after state induc-tion. Local anesthetic (lidocaine, 2%) was infiltrated into all surgicalwounds every 2 h.

Drugs

Carbamazepine (Sigma-Aldrich) was dissolved in vehicle containing10% dimethylsulfoxide, 40% propylene glycol, and 50% saline.CBZ was injected intraperitoneally for EEG recordings in freelymoving rats (n = 8), and subcutaneously for rats under neurolepticanalgesia (n = 31). The administered doses (10–20 mg/kg) werethose that aggravate absence seizures (Liu et al., 2006).

Electrophysiology

Freely moving condition

For four rats, recordings were made with a Compumedics EEGacquisition system (Compumedics, Melbourne, Australia) with asampling rate of 256 Hz, and for another four rats, recordings weremade with ultralow-noise amplifiers (AI 402, 950; Axon Instru-ments) with a sampling rate of 10 kHz (bandpass: 0.1–800 Hz).EEG recordings were performed in a well-lit, quiet room with ratsremaining in their home cages. The recording session started after15 min of habituation to the experimental conditions; this wasfollowed by a 30-min baseline period and a 90-min post-injectionperiod. The behavior was rated every minute as immobile, active(exploration, whisking, eating, chewing, motion, grooming, etc.),sleepy, eyes closed or open, etc.

Light-anesthesia condition

Glass micropipettes were filled with a solution containing 1.5%N-(2-amino ethyl)-biotin amide hydrochloride (Neurobiotin; Vector

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

2 T. W. Zheng et al.

Laboratories, Burlingame, CA, USA) dissolved in 0.5–1 M

CH3COOK. The micropipette (25–70 MO) was stereotaxically (Pax-inos & Watson, 1998) lowered with a stepping micro-driver into themedial part of the ventral posterior thalamic nucleus and into therelated TRN (intermediate part) to record, extracellularly or intracel-lularly, a single TC or TRN neuron along with the surface EEG ofthe related somatosensory cortex. Additional recordings were per-formed in three rats with intracellular glass micropipettes filled withKCl (3 M). The EEG and cellular signals were digitized at a sam-pling rate of 20 kHz, with bandpasses of 0.1–800 Hz and 0.1–6 kHz, respectively. During the intracellular recording session, acurrent pulse in the range of �0.2 to �0.5 nA was applied every2 s to keep the Wheatstone bridge well balanced.

Histology

At the end of the recording session, the neurons were individuallylabeled by use of the juxtacellular (Pinault, 1996) or intracellular tra-cer micro-iontophoresis technique for standard histological identifi-cation and localization. After a survival period of 15–30 min, the ratwas killed with an intravenous overdose of pentobarbital, followedby transcardial perfusion with a fixative containing 4% paraformal-dehyde and 0.25% glutaraldehyde in 10 mM phosphate-bufferedsaline. Brains were kept in 4% paraformaldehyde overnight. Serialslices of 100 lm of the required region were cut with a vibratome,and processed with standard histological techniques for retrieval ofthe tracer-stained neurons or regions (Pinault, 1996).

Data analysis

The tracer-filled neurons were examined with a light microscope.Their location was ascertained by reference to the stereotaxic atlas(Paxinos & Watson, 1998). Electrophysiological recordings wereanalyzed with Axon software (CLAMPEX, v7; Axon Instruments), andDataWave software (SCIWORKS, v8; DataWave Technologies,Berthoud, CO, USA).The average membrane potential of the recorded neurons was

estimated from at least 30 values collected during periods free ofslow oscillations (< 17 Hz) at rest (no holding current). Fast Fouriertransform analysis (hamming windowing) of EEG and intracellularsignals was based on epochs of ≥ 500 ms, with a resolution of0.6–2.4 Hz. The total power of a given frequency band was com-puted as the sum of all of the corresponding fast Fourier transforma-tion values.The thalamic high-frequency bursts of action potentials recorded

in the intact brain were identified according to standard criteria(Domich et al., 1986; Pinault et al., 2006). In short, a typical burstcontains at least two action potentials with an interval of < 3 ms.The action potential discharge in the 5–9-Hz bursts shows a lessmarked acceleration–deceleration pattern than that in spindle bursts(Pinault et al., 2006).Autocorrelation histograms of EEG and intracellular oscillations

and inter-action potential interval histograms of TRN neuron firingwere computed over periods of time of > 10 s.Burst-triggered averages (BTAs) of the cortical EEG recording

and of the corresponding fast Fourier transformation were computedover a period of time of � 250 ms. The first action potential of a5–9-Hz or spindle TRN extracellular burst was the triggering event.As it is known that the rhythmic 5–9-Hz and spindle firing of TRNneurons is more reliable than that of TC neurons (Pinault et al.,2006), the BTAs were computed from the recordings of TRN neu-rons along with the related cortical EEG. Two BTAs were com-

puted, one with a 5–9-Hz burst, and the other with a spindle burst(Fig. 6A). The corresponding extracted EEG segments were sub-jected to fast Fourier transform analysis.Pearson’s correlation coefficient (implemented with Bonferroni

probability) was used to provide an index of the power coherence of5–9-Hz and spindle (10–16-Hz) oscillations between the membranepotential of thalamic neurons and the EEG recording of the relatedfrontoparietal cortex. We used the intracellular membrane potentialoscillations because they are composed of intrinsic and postsynapticpotentials, which contribute more than action potentials to extracel-lular local field potential oscillations (Buzs�aki, 2004). The linearcorrelation was estimated from paired values of the total power from500-ms recording epochs that included a hybrid 5–9-Hz spindle pat-tern in the intracellularly recorded TC neurons. From every hybridpattern, two epochs were extracted for analysis: the first epoch tar-geted the first component that included at least two 5–9-Hz waves,and the second epoch targeted the spindle component.For each set of data, normality was checked with the D’Agostino

and Pearsons omnibus normality test (a = 0.05; GRAPHPAD PRISM,V5.04). For group analysis of non-parametric data, the Kruskal–Wallis test was used followed by Dunn’s post hoc test. t-Tests wereused for parametric data, for comparison between groups. Data(mean � standard error of the mean) were considered to be signifi-cant at P < 0.05.

Results

CBZ potentiates cortical EEG 5–9-Hz and 10–16-Hz (spindle)oscillations

Under the drug-free condition, the behavior of the rats fluctuatedfrom active and quiet immobile wakefulness to drowsiness with aninclination to sleep. Rats were left undisturbed as much as possible(at times, rats were gently handled to untwist recording cables)throughout the recording sessions. During quiet immobile wakeful-ness, the frontoparietal sensorimotor cortex spontaneously showeddesynchronized low-voltage (< 0.1-mV) fast (17–160-Hz) EEGoscillations, which were at times (0.5–4/min) interrupted by theoccurrence of short-lasting (< 3-s) bouts of medium-voltage (< 0.2-mV) slower waves, the most prominent being 5–9-Hz oscillations(Pinault et al., 2001). The rats entered the first stages of sleep(curled-up posture) 20–40 min after the beginning of the recordingsessions. The sleep-related cortical EEG was characterized by ongo-ing slow-frequency (< 16-Hz) oscillations, including a large amountof delta (1–4-Hz) waves (Fig. 1). The CBZ-treated rats were behav-iorally much less reactive to gentle handling than the control rats.Indeed, under the CBZ condition, the rats often fell asleep < 2 minafter handling, whereas the control rats could stay awake and activefor > 2 min (Fig. S1).5–9 Hz and spindle oscillations were more prominent during

CBZ-induced drowsiness than during natural drowsiness (Fig. 1A,B2, and B3). There was a significant overall difference in the 5–9-Hzand 10–16-Hz frequency bands after CBZ (20 mg/kg) injection(Kruskal–Wallis test, n = 4; Fig. 1B2, P = 0.018; Fig. 1B3,P = 0.013). Dunn’s post hoc test showed significant increases in bothfrequencies following the administration of 20 mg/kg CBZ (n = 4;Fig. 1B2, P = 0.03; Fig. 1B3, P = 0.007). A significant effect wasmeasured with 20 mg/kg CBZ. On the basis of the internal frequencyand waveform, 5–9-Hz oscillations could be distinguished from spin-dles (Fig. 2A, C, and D) [see also Pinault et al. (2006)]. Followingthe administration of CBZ, spindles occurred either alone or after afew 5–9-Hz waves or a K complex (Fig. 2B), this latter event

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

CBZ potentiates thalamocortical spindles 3

representing an isolated cortical down-state (Cash et al., 2009). Curi-ously, we never observed a 5–9-Hz oscillation at the end of a typicalspindle, i.e. a frank deceleration of the spindle (10–16-Hz) oscilla-tions. During quiet immobile wakefulness, in contrast to the CBZcondition, following the administration of vehicle spindle oscillationswere not visible in EEG recordings even though a certain amount of10–16-Hz oscillations could be measured with mathematical tools(not shown).

CBZ increases GABAA receptor-mediated potentials in TCneurons

To determine whether the CBZ-induced 5–9-Hz and spindle oscilla-tions involved a GABAA receptor-mediated component, weperformed intracellular recordings of TC neurons under neurolepticanalgesia, a condition that allows recording of spontaneously occur-ring EEG 5–9-Hz and spindle oscillations (Pinault et al., 2006;Zheng et al., 2012). It is worth noting that this light anesthesia itselfmaintains the animals in a relatively stable state that is electrophysi-ologically equivalent to quiet immobile wakefulness (Pinault et al.,2001). After CBZ administration, the rats’ state consistently evolvedtowards a sleepy state accompanied by more low-frequency

A

B1 B2 B3

Fig. 1. CBZ increases 5–9-Hz and spindle oscillations in the frontoparietal cortex. (A) Top: 20-min records (15-s bouts extended just below) of the surfaceEEG under control (left records: drowsiness; +60–80 min after vehicle injection) and CBZ (right records: drowsiness; +60–80 min after 20 mg/kg injection)conditions (from the same rat, 3-day interval). Below: time–frequency spectra from 30-s epochs (fast Fourier transformation, resolution of 2.4 Hz, hammingwindowing) of the surface cortical EEG recording under control and CBZ conditions. The white dotted lines indicate the putative limits between 5–9-Hz andspindle (10–16-Hz) oscillations, although they share a common frequency band. (B1–B3): Spectral power of the 1–4-Hz (B1), 5–9-Hz (B2) and 10–16-Hz (B3)oscillations under three experimental conditions (veh, vehicle; CBZ10, CBZ 10 mg/kg; and CBZ20, CBZ 20 mg/kg). Dunn’s post hoc test shows significantincreases in both frequencies following administration of 20 mg/kg CBZ (n = 4; B2, *P = 0.03; B3, *P = 0.007).

A

C

B

D

Fig. 2. Characteristics of 5–9-Hz, spindle and ‘hybrid’ EEG oscillations.(A) Typical traces of cortical 5–9-Hz and spindle oscillations recorded infreely behaving rats. (B) A spindle (sp) can occur after either a few 5–9-Hzwaves (hybrid) or after a K complex (k). (C and D) Autocorrelation histo-grams computed from three successive bouts of 5–9-Hz and spindle oscilla-tions, respectively.

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

4 T. W. Zheng et al.

(< 20-Hz) waves, including spindle oscillations (Fig. S2), whichwere similar to those recorded in freely behaving rats (Fig. S3).Intracellular recordings were made from 10 TC neurons under the

control condition and from eight TC neurons 20–60 min after a sin-gle CBZ injection. Under the control condition, TC neurons had anestimated average resting membrane potential of �56 � 5 mV, anaverage input resistance of 26 � 9 MΩ, and action potential ampli-tude above 50 mV. After a single injection of CBZ, the membranepotential was more hyperpolarized [�61 � 5 mV; Student’s non-directional t-test (equal variances), P = 0.007], revealing subthresh-old oscillations, including rhythmic 5–9-Hz and spindle waves(Fig. 3), the latter waves being more prominent in power under theCBZ condition (Fig. 3A2, B2, and C). Cortical EEG and thalamiccellular 5–9-Hz oscillations were spatiotemporally more coherent inpower than spindles (Fig. S4). Indeed, a thalamic cellular spindlewas not systematically time-locked with a cortical EEG spindle,although spindles were increased in both the cortex and the thala-mus under the CBZ condition relative to the control condition. Allrecorded neurons showed a putative low-threshold Ca2+ potentialtopped by a high-frequency burst of action potentials (Fig. 3D).

Following the administration of CBZ, the membrane potential ofTC neurons often showed 5–9-Hz oscillations that were suddenlyprolonged by pronounced spindle oscillations, forming a 5–9-Hz–spindle hybrid sequence. This was characterized by two to fourcycles of a rhythmic 5–9-Hz depolarizing wave–hyperpolarizingwave sequence extended by a typical series of 3–10 rhythmic(10–16-Hz) short-lasting (66.5 � 1.1-ms, n = 35, from five rats)hyperpolarizations (Fig. 3A2 and B2). Both 5–9-Hz and spindleoscillations always occurred during a slow hyperpolarizing trough(Fig. 3B1 and B2). The first spindle IPSP abruptly appeared imme-diately after either a 5–9-Hz EPSP wave–IPSP wave sequence or a5–9-Hz EPSP wave (Fig. 3B2). On the other hand, intracellularspindles were never followed by 5–9-Hz oscillations. It is importantto emphasize that, like the oscillations observed in the frontoparietalEEG, intracellular TC spindles were extremely rare in control rats(one or two cases during a recording session of several tens of min-utes in two of 10 TC neurons). Such a rare case is shown inFig. 3A2. In contrast, under the CBZ condition, intracellular spin-dles frequently occurred (up to 12/min), and lasted for no more than1.5 s (average of 0.82 � 0.05 s, n = 40, in five of eight recorded

A1 B1

B2A2

C D E F

Fig. 3. Carbamazepine (CBZ) increases intracellular voltage spindle oscillations in TC neurons. (A and B) Simultaneous intracellular TC recordings and front-oparietal cortical (FPcx) EEG obtained under control (A1 and A2) and CBZ (10 mg/kg, subcutaneous; B1 and B2) conditions. Sections highlighted in gray inA1 and B1 are expanded in A2 and B2, respectively. In A2 and B2, stars indicate depolarizing waves, vertical arrows hyperpolarizing waves, and curved arrowsaction potentials (truncated). The 5–9-Hz depolarizing wave–hyperpolarizing wave sequences are indicated by a horizontal curved bar. The inset in A2 is anautocorrelation histogram of the 1-s firing period that just precedes the subthreshold spindle hyperpolarizing waves (indicated by the arrows). The photomicro-graph in B2 is from a 100-lm-thick section of the recorded neuron, revealing part of its soma and dendrites. (C) Spectral analysis (average of eight 2.5-sepochs) of fast Fourier transform data with a resolution of 0.76 Hz. The histogram shows the average (eight values from four rats) total power of 10–16-Hzoscillations (t-test, *P < 0.001). (D) A typical low-threshold Ca2+ potential topped by a high-frequency burst of action potentials triggered by a depolarizingpulse from a hyperpolarizing holding current (�79 mV). (E and F) Superimposition of three successive rhythmic depolarizations, which occur during 5–9-Hzoscillations (E) and during spindle oscillations (F) in the TC neuron. In E, one of the three 5–9-Hz depolarizations triggers a putative low-threshold Ca2+ spiketopped by a high-frequency burst of APs, whereas in F only a single action potential was evoked during one of the spindle events.

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

CBZ potentiates thalamocortical spindles 5

TC neurons). In contrast to spindle waves, a 5–9-Hz depolarizingwave could contribute to the generation of a putative low-thresholdCa2+ potential crowned by a high-frequency burst of action poten-tials (Fig. 3E and F).In an attempt to determine whether the hyperpolarizing waves of

the spindle oscillations were GABAA receptor-mediated IPSPs, weapplied a sustained hyperpolarizing current of increasing intensity tomove the membrane potential from rest (with a holding current of0 nA) to more hyperpolarized levels (Fig. 4). In all TC neurons thatshowed spindle oscillations (five of five rats) under the CBZ condi-tion, the polarity of the 5–9-Hz and spindle hyperpolarizationsreversed at a membrane potential of approximately �70 mV (rangebetween �67 and �73 mV; Fig. 4B1 and B2). Moreover, when themembrane potential was set below �80 mV, all of the components(EPSPs and IPSPs) of the hybrid 5–9-Hz–spindle sequence weredepolarizing (Fig. 4C2). The slow hyperpolarizing trough was, ingreat part, reversed. These observations are consistent with previous

data (Pinault et al., 1998), suggesting the reversal potential of chlo-ride ions. Indeed, in three TC neurons that were recorded for at least10 min with an intracellular KCl-filled micropipette, putative rhyth-mic 5–9-Hz and spindle IPSPs became depolarizing and triggeredhigh-frequency bursts of action potentials (Fig. S5). These observa-tions substantiate the hypothesis of the involvement of activation ofGABAA receptors in the generation of 5–9-Hz and spindle hyperpo-larizations, which are probably IPSPs.

CBZ increases spindle burst firing in TRN neurons

The TRN is the principal source of GABA receptor-mediated inhibi-tion of the TC neurons (Bal & McCormick, 1993; Pinault, 2004;Astori et al., 2011). If the TC rhythmic hyperpolarizing waves atthe spindle frequency were indeed GABAA receptor-mediated IPSPs,we would record rhythmic burst firing in TRN neurons at the samefrequency (10–16 Hz). We performed, under light anesthesia,

A1 A2

B1 B2

C1 C2

Fig. 4. Reversal of intracellular spindle hyperpolarizing waves in TC neurons under the CBZ condition. All voltage recordings are from the same TC neuronalong with the frontoparietal cortical (FPcx) EEG. (A1 and A2) In the absence of holding current (0 nA), the spindle hyperpolarizing waves are obvious at theresting membrane potential (�59 mV). (B1 and B2) At a membrane potential in the range of �65 to �72 mV (holding hyperpolarizing current of �0.8 nA),the hyperpolarizing waves are not identifiable. The curved arrow indicates a putative low-threshold Ca2+ potential topped by a high-frequency burst of actionpotentials. (C1 and C2) At a more hyperpolarized level (lower than �80 mV with a holding current of �4.6 nA), the hyperpolarizing waves are fully reversed(the last six depolarizing events in C2). In C2, the first two depolarizing events presumably include increased EPSPs plus reversed IPSPs (two cycles of 5–9-Hzoscillations extended by a series of six reversed spindle IPSPs). B2 and C2 also include autocorrelation histograms of an intracellular spindle oscillation occur-ring at �59 mV (rhythmic hyperpolarizing waves) and an intracellular spindle oscillation occurring at �88 mV (rhythmic depolarizing waves), respectively.

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

6 T. W. Zheng et al.

extracellular single-unit recordings of GABAergic TRN neuronss(five units from five rats) along with the related frontoparietal corti-cal EEG. The state of the EEG was more synchronized under theCBZ condition than under the control condition (Fig. 5A1 and A2).In all five recorded TRN neurons, inter-action potential interval his-tograms did not show any significant difference in the principalintrinsic firing features recorded between the vehicle and CBZ con-ditions (Fig. 5B), with averaged instantaneous frequencies of270 � 12 Hz (maximum, 646 Hz; 316 action potentials) and251 � 12 Hz (maximum, 667 Hz; 366 action potentials), respec-tively. The related TRN spindle bursts are, in great part, under-pinned by low-threshold T-type Ca2+ channels of the Cav3.3 type(Astori et al., 2011). Under the CBZ condition, a rhythmic spindle-frequency burst firing pattern was significantly more frequent andlonger in duration (Fig. 5A1 and A2, vertical arrows; Fig. 5C andD; Fig. 5E, spindle frequency, Wilcoxon matched-paired signedrank test, P = 0.03; Fig. 5F, spindle duration, Mann–Whitney test,U = 1056, P < 0.0001). It was also striking that the robust spindleburst firing often occurred immediately after two or three cycles of5–9-Hz high-frequency bursts of action potentials (Fig. 5A1 and

A2; vertical arrows for spindle bursts and asterisks for 5–9-Hzbursts). This hybrid 5–9-Hz–spindle sequence was accompanied bya larger amount of 5–9-Hz than of spindle waves recorded in therelated cortex (Fig. 6), supporting the hypothesis that the cortexplays a key role in the generation of CT 5–9-Hz oscillations (Pina-ult, 2003).

Discussion

The present study, carried out in vivo in the rat TC system, demon-strates for the first time that drowsiness induced by a single injectionof CBZ is principally associated with an increase in sleep spindlepower, number, and duration. Indeed, in somatosensory TC neurons,spindles constituted a manifestation of GABAA receptor-mediatedrhythmic IPSPs, which were triggered by high-frequency bursts ofaction potentials in TRN neurons. In TC neurons, spindle oscilla-tions often occurred after two or three CT 5–9-Hz EPSP–IPSPwaves. In TRN neurons, spindle bursts often occurred after a few(two or three) 5–9-Hz bursts. This is the first time that this hybrid5–9-Hz–spindle pattern has been reported. These findings supports

A1 A2

B

CD

E F

Fig. 5. CBZ increases spindle burst firing in TRN neurons. (A1 and A2) Simultaneous cortical EEG recording from the frontoparietal cortex (FPcx) and extra-cellullar recording of a TRN neuron under the control (vehicle, A1) condition and approximately 40 min after CBZ (A2) injection. Stars indicate typical 5–9-Hz burst firing, and arrows indicate typical spindle burst (10–16-Hz) firing. Below (A2) is shown a typical hybrid 5–9-Hz–spindle sequence (high-pass filtercut-off at 200 Hz) and expanded 5–9-Hz (left) and spindle (right) high-frequency bursts of action potentials. (B) Inter-action potential interval histograms undervehicle and CBZ conditions (resolution, 1 ms; 10-s epoch during synchronized EEG; CBZ, 366 action potentials; vehicle, 316 action potentials). (C and D)Autocorrelation histograms of typical 5–9-Hz burst and spindle-burst firing, respectively. (E and F) Both the frequency (E; P = 0.03, n = 5) and the duration(F; CBZ, 87 spindles from five rats; vehicle, 54 spindles from five rats) of spindle firing are significantly increased after systemic injection of CBZ (10 mg/kg,subcutaneous). *P < 0.05 (see text for detail).

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

CBZ potentiates thalamocortical spindles 7

the hypothesis that CBZ potentiates GABAA currents (Liu et al.,2006). Furthermore, we propose that the cortex can play a leadingrole in the triggering of TC spindles.

EEG features associated with CBZ-induced drowsiness

Consistently, increased spindle power in the cortical EEG was themost important electrophysiological feature recorded here in freelymoving rats treated with a relatively high dose of CBZ (20 mg/kg),which induced drowsiness. At a lower dose (10 mg/kg), the amountof spindles remained in the control range. It is worth rememberingthat the spindle power computed with the fast Fourier transformdoes not differentiate between baseline EEG sigma (10–16-Hz)oscillations and transient spindle oscillations (De Gennaro & Ferr-ara, 2003). The increase in EEG spindle power was confirmed bythe cellular correlates recorded in TC (spindle IPSPs) and TRN(spindle bursts) neurons after a single CBZ injection (see below).Our results are consistent with a recent human study showing thatCBZ enhanced frontal lobe spindle activity and slow oscillations(Ayoub et al., 2013).The power of EEG 5–9-Hz oscillations was also significantly

increased after a single CBZ injection in freely moving (CBZ, 20 mg/kg) and in lightly anesthetized (CBZ, 10 or 20 mg/kg) rats. Theincrease in 5–9-Hz waves was not obvious in cellular recordings,

suggesting that the spectral analysis included a frequency band com-mon to both 5–9-Hz and spindle oscillations. Oscillations of 5–9 Hzare recorded in the rodent frontoparietal cortex during quiet immobilewakefulness, but not during active wakefulness (present study) [seealso Pinault et al. (2001)]. In genetic models of absence epilepsy,5–9-Hz oscillations suddenly become hypersynchronized, giving riseto generalized epileptic discharges in CT–TC systems (Pinault et al.,2001; Polack et al., 2007; Zheng et al., 2012). The physiologicalelements that underlie these 5–9-Hz oscillations remain to be deter-mined. They belong to a puzzling continuum of theta-frequency,alpha-frequency, mu-frequency and sigma-frequency oscillations (Ti-ihonen et al., 1989; Pfurtscheller et al., 1997; M€olle et al., 2002).Their function is still under debate and investigation (Tort & Katz,2010). Their frequency band is lower but overlaps with that of spin-dles, reaching a maximum frequency of 12 Hz (Pinault et al., 2001).There is also an overlap with lower-frequency delta oscillations, andboth can coexist. As 5–9-Hz oscillations are recorded in the somato-sensory system, it is tempting to suggest that they could be the equiva-lent of either the alpha rhythm (8–12 Hz) recorded in the visual CT–TC systems, or the theta rhythm, which is known to increase duringthe first stages of sleep (Silber et al., 2007). Further investigation isrequired to clarify this intriguing point.Another remarkable EEG feature consistently recorded under the

control condition (freely moving or light anesthesia) but more

A

B1 B2

C1 C2

Fig. 6. Relationship of the TRN hybrid 5–9-Hz–spindle sequence with the simultaneous EEG from the frontoparietal cortex (FPcx) under the CBZ condition.A typical example of a TRN hybrid sequence with the related cortical activity filtered at 2–20 Hz is shown in A. The two vertical arrows indicate a 5–9-Hzand spindle burst used to compute the BTA of the related EEG (B1 and B2) and corresponding fast Fourier transformation (C1 and C2). Averaging is computedfrom 10 hybrid sequences, like the one shown in A. The inserted histogram in C2 shows the total power of 5–9-Hz (black) and spindle (10–16-Hz, white) oscil-lations computed from 22 TRN hybrid 5–9-Hz–spindle sequences from four TRN cells (four rats). The histogram includes the values of the spectra in C1 andC2. Note that the power of 5–9-Hz oscillations is much higher than that of spindle oscillations during both the 5–9-Hz and the spindle TRN bursts in the hybridpattern. t-test: ns, not significant.

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

8 T. W. Zheng et al.

frequently after CBZ administration was 5–9-Hz waves followed bya robust spindle (Fig. 7). Intriguingly, the order of this stereotypedhybrid pattern seems to be fixed. Indeed, in EEG and intracellularrecordings, spindles were never followed by 5–9-Hz oscillations.Furthermore, both the autocorrelogram of the cortical EEG recordingand the intracellular recording of thalamic neurons revealed thatspindle oscillations were more stereotyped than 5–9-Hz oscillations.As 5–9-Hz oscillations are generated in the cortex (Pinault, 2003;Pinault et al., 2006) and spindles in the thalamus (von Krosigket al., 1993), we propose that CT 5–9-Hz oscillations can triggerTC spindles, sometimes under the physiological condition, and morefrequently after the administration of CBZ. This proposal is inagreement with previous in vitro and in vivo studies (Bal et al.,2000; Blumenfeld & McCormick, 2000; Bonjean et al., 2011). Inother words, CT 5–9-Hz oscillations would make the TC systemresonate at the spindle frequency. The electrophysiological hybrid5–9-Hz–spindle sequence might be a signature of CBZ effects, as itwas at least 10 times more frequent during the CBZ condition thanduring the control condition. Moreover, what is intriguing is that the5–9-Hz rhythm, which is more related to wakefulness than to sleep(Pinault et al., 2001), can immediately be prolonged by sleepspindles, suggesting that a certain amount of residual wake-related

5–9-Hz oscillations persist during CBZ-induced drowsiness. Thus,the question of whether this singular hybrid pattern has a physiolog-ical or pathophysiological function remains open.

The thalamic cellular correlates of 5–9-Hz and spindleoscillations

The present cellular data are in agreement with those from previousstudies (Pinault, 2003; Pinault et al., 2006). Here, it is further shownthat, through unknown prethalamic and/or intrathalamic mechanisms,CBZ significantly hyperpolarizes the membrane potential of thalamicneurons and increases spindle oscillations. We predict that CBZlikewise hyperpolarizes other types of cortical and subcortical neu-rons. CT neurons of layer VI, which are massive in number (Jones,1985), exert a strong facilitatory effect on thalamic neurons (Yuanet al., 1986). Therefore, it is reasonable to propose that the hyperpo-larizations recorded in thalamic neurons under the CBZ conditionmight be the result of disfacilitation of CT inputs. The CT feedbackon TC and TRN neurons is thought to play a central role in atten-tional processes (Sillito et al., 1994; Pinault, 2004). Concerning thehyperpolarizing trough that is associated with the rhythmic events,which include GABAA receptor-mediated IPSPs, the present results(partial reversal at approximately �70 mV) suggest the involvementof a cumulative effect of the repetitive activation of GABAA recep-tors. However, we cannot exclude the possibility of a contributionof the cation currents IK-leak and/or Ih to the slow hyperpolarizingtrough that is associated with rhythmic 5–9-Hz and spindle activityin TC neurons (Meuth et al., 2006). Further investigation is requiredto better understand the relative contributions of synaptically medi-ated (GABAA) and intrinsically mediated (IK-leak and Ih) currentsduring 5–9-Hz and spindle oscillations.The thalamic intracellular rhythmic events associated with 5–9-Hz

and spindle oscillations are distinguishable (Pinault et al., 2006). InTC neurons, the rhythmic 5–9-Hz and spindle IPSPs are mediatedby activation of GABAA receptors. The present study demonstratesthat the GABAA receptor-mediated rhythmic 5–9-Hz and spindleIPSPs are caused by rhythmic burst discharges in related GABAer-gic TRN neurons. The 5–9-Hz oscillations recorded in TC neuronsare very probably initiated by the glutamatergic layer VI CT neurons(Pinault, 2003), whereas the electrogenesis of spindles principallyinvolves the functional reciprocal interaction in the glutamatergicTC–GABAergic TRN circuit (von Krosigk et al., 1993; Steriade,1995). Importantly, both the initiation and the termination of sleepspindles are under the control of CT activities (Bonjean et al.,2011).Figure 7 is a sketch presenting the likely scenario of the anatomo-

functional events underlying the hybrid 5–9-Hz–spindle pattern pres-ent in the surface cortical EEG during drowsiness or the first stagesof non-rapid eye movement sleep. On the basis of the present find-ings, we conclude that CBZ significantly and steadily hyperpolarizesboth TC and TRN neurons. The glutamatergic layer VI neurons,which innervate TRN and TC neurons, sometimes rhythmically dis-charge at 5–9 Hz, imposing at that frequency a series of high-fre-quency bursts of action potentials in TRN neurons, which have ahigher propensity to burst than TC neurons, because of their reliableelectro-responsive properties (Llin�as, 1988; Huguenard & Prince,1992). The number of excitatory CT neurons is large; they areapproximately 10-fold more abundant than TC neurons (Jones,1985). In response to the CT discharges, TC neurons generate sub-threshold EPSP waves, because they become more hyperpolarizedduring the first stages of sleep (natural or CBZ-induced). However,such a TC EPSP wave can trigger a low-threshold Ca2+ potential

Fig. 7. Prediction that CT 5–9-Hz oscillations can trigger TC spindles. Thissketch predicts cortical and thalamic events that can be recorded in rats withCBZ-induced drowsiness. On the left, the three principal elements that makeup the somatosensory TC system are shown. Note that the CT neuron oflayer VI innervates both the GABAergic TRN neuron and the glutamatergicTC neuron. The surface cortical EEG recording shows a series of 5–9-Hzwaves followed by a typical spindle (sinusoidal waves at 10–16 Hz). Thissequence was recorded in a drowsy rat (freely moving condition). The TRNneuron shows a series of four 5–9-Hz high-frequency (300–650-Hz) bursts ofaction potentials (n = 3–15). The membrane potential of the TC neuronshows a few waves at 5–9 Hz followed by a spindle oscillation during asteady envelope-like hyperpolarization. A depolarizing 5–9-Hz wave can trig-ger a low-threshold Ca2+ potential crowned by a high-frequency burst ofaction potentials (left curved arrow). The following spindle oscillations aresubthreshold. Such a 5–9-Hz–spindle sequence can be terminated by arebound firing (right curved arrow), which usually appears at the terminationof the steady envelope-like hyperpolarization. The TRN and TC neuronswere recorded under light anesthesia. On the basis of the current literature,the origin of 5–9-Hz oscillations is led by CT neurons of layer VI (greenbar), whereas the generation of spindles involves reciprocal TC–TRN interac-tions (red and purple bars).

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

CBZ potentiates thalamocortical spindles 9

(McCormick & Pape, 1990). In turn, the TRN 5–9-Hz burstsdirectly trigger a GABAA receptor-mediated IPSP in TC neuronsimmediately after the subthreshold EPSP wave. Intriguingly, the fewTC 5–9-Hz EPSP wave–IPSP wave sequences are suddenly pro-longed by a spindle, which corresponds to a resonant train of rhyth-mic 10–16-Hz bursts in TRN neurons and to a subsequent typicalseries of rhythmic 10–16-Hz IPSPs in related TC neurons. This TCspindle usually terminates at the end of the steady hyperpolarization,and occasionally shows rebound firing, very likely under the depo-larizing influences of a hyperpolarization-activated cation current (Ih)and of CT inputs (Bal & McCormick, 1996; Budde et al., 1997; Bon-jean et al., 2011). The hybrid 5–9-Hz–spindle pattern is, to allappearances, a hallmark of drowsiness or of the first stages of non-rapid eye movement sleep. It appears as an index of functional inter-actions between the cortex and thalamus. Moreover, it supports thehypothesis of an active contribution of the cortex to the initiation ofsleep spindles (Bonjean et al., 2011). Whether the hybrid pattern isan epiphenomenon or is implicated in a given function or dysfunctionremains an open question.As no intermediate event was apparent in between the 5–9-Hz

oscillations and the 10–16-Hz spindle oscillations, one may wonderwhat is (are) the factor(s) responsible for the sudden switch fromthe former to the latter oscillation type. First, one important factor isthe state of the system, which becomes increasingly hyperpolarized(at least in TC and TRN neurons) during 5–9-Hz oscillations(Fig. 3). Such a hyperpolarized state is a prerequisite for triggeringthe spindle pacemaker properties of TRN neurons, which synaptical-ly and mutually work with the related TC neurons (von Krosigket al., 1993). This likely scenario is supported by previous studiesshowing that cortical electrical stimulation generates a long-lastinghyperpolarization in TRN neurons, immediately followed by a spin-dle burst sequence (Contreras et al., 1996), and that activation ofthe CT pathway triggers intra-TRN inhibition (Zhang & Jones,2004). Therefore, it is tempting to put forward the concept that, afterCT 5–9-Hz oscillations, a spindle is a post-inhibitory all-or-noneresonant rebound activity in TRN (series of spindle bursts) and TC(subsequent series of spindle GABAA IPSPs) neurons. The presentfindings highlight the persuasive notion that the CT–TC system ischaracterized by plastic mechanisms during the wake–sleep cycle(Steriade, 2006; Timofeev, 2011), in which the TRN, with its pow-erful intrinsic properties, plays a pivotal role, at least in the genera-tion of sleep spindles (Steriade et al., 1987).In agreement with a previous study (Pinault et al., 2006), the

present investigation shows that, in rat TC neurons, the series ofrhythmic spindle-frequency IPSPs does not trigger rebound burst fir-ing, as would have been expected from previous in vivo and in vitrostudies (Deschenes et al., 1984; von Krosigk et al., 1993). This maybe the reason why TC spindles never triggered 5–9-Hz waves in thesomatosensory CT–TC system. The lack of spindle-frequency burstfiring in the recorded TC neurons is discussed in a previous report(Pinault et al., 2006).The cellular and ionic mechanisms underlying sleep spindles have

been well described (Beenhakker & Huguenard, 2009). Whether theCT–TC systems are the primary sites of action of CBZ, and, simi-larly, whether GABAergic transmission is the primary target ofCBZ, are questions that deserve further investigation. Although theCBZ-binding site on GABAA receptors is not known, it has beendemonstrated to be a positive allosteric modulator on GABAA

receptor subtypes that are naturally expressed in the thalamus (Gran-ger et al., 1995; Liu et al., 2006; Zheng et al., 2009). We shouldnot exclude the possibility that CBZ may have other moleculartargets. Spindle bursting in TRN neurons involves an interplay

between Cav3.3-type Ca2+ channels and Ca2+-dependent small-conductance type 2 (SK2) K+ channels (Astori et al., 2011). Asenhanced SK2 channel activity sustains sleep spindles (Wimmeret al., 2012), CBZ might directly or indirectly (e.g. via GABAreceptors) potentiate the activity of SK2 channels.

Functional implications

The present study demonstrates that CBZ is a sleep spindle-promot-ing agent, and this property is probably responsible for provokingpersistent somnolence and attention deficit, some of the drug’s majorside effects (Levy et al., 1985). These findings raise a number ofquestions requiring further investigation. Are the behavioral andneurophysiological effects of chronic CBZ treatment similar to thoseinduced by a single CBZ injection? Do the neurocognitive sideeffects of CBZ (Dias et al., 2012) result from sleep disorders, ormore specifically from an increase in sleep spindles? Could theCBZ-induced spindle increase disrupt the beneficial effects of sleep,e.g. by disturbing synaptic homeostasis (Tononi & Cirelli, 2012)? Inpatients with schizophrenia, sleep spindles are reduced (Ferrarelliet al., 2007, 2010). Therefore, another question arising from thecurrent study is whether CBZ-induced renormalization of theamount of spindles could be associated with beneficial cognitiveeffects in patients with schizophrenia.Despite the effective use of CBZ in the treatment of focal sei-

zures, the mechanisms underlying significant side effects of thedrug, such as drowsiness and severe alteration of sleep architecture(Levy et al., 1985; Yang et al., 1989), remain unclear. Pharmaco-therapies that induce drowsiness or sedation, whether as the pri-mary therapeutic mechanism of action or as an adverse effect,commonly involve GABAA receptor-mediated neurotransmission.This diverse class of drugs includes benzodiazepines, barbiturates,ethanol, and several anesthetics that have been used either therapeu-tically or as experimental tools to study sleep physiology. GABAA

agonists such as muscimol are known to induce sleep that is associ-ated with an increase in slow waves (0.5–4 Hz) and the generationof spindles (11–16 Hz) in rats (Lancel et al., 1996). Treatments forinsomnia have also been dominated by positive modulators ofGABAA receptors, such as barbiturates and benzodiazepines, whichact at different binding sites on GABAA receptors to increase spin-dle-frequency activity (Winsky-Sommerer, 2009). In short, thedirect and/or indirect mechanisms by which CBZ modulatesGABA-related oscillations in CT–TC systems remain to be furtherinvestigated.

Conclusion

The present study shows that a single systemic administration ofCBZ induces a dysfunctional state in CT–TC systems, characterizedby an excessive increase in GABAA receptor-mediated TC sleepspindles. They often appeared after a few CT 5–9-Hz waves, form-ing a unique hybrid CT 5–9-Hz–TC spindle pattern. We proposethat, at least in drowsy rodents, the rhythmic CT 5–9-Hz activitycan make the TC system resonate at the spindle frequency.

Supporting Information

Additional supporting information can be found in the onlineversion of this article:Fig. S1. Electrophysiological indication of CBZ-induced sedation.Fig. S2. Under the light anesthesia condition, CBZ increases the amountof cortical low-frequency (< 20-Hz) oscillations.

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

10 T. W. Zheng et al.

Fig. S3. Cortical 5–9-Hz and spindle (10–16-Hz) oscillations recordedunder light anesthesia are similar to those recorded in freely behavingrats.Fig. S4. Oscillations at 5–9 Hz are more coherent than spindle oscilla-tions in power between the membrane potential of TC neurons and theEEG of the related frontoparietal cortex (FPcx).Fig. S5. Thalamocortical hybrid 5–9-Hz spindle pattern recorded intra-cellularly with a KCl (3 M)-filled micropipette.

Conflict of interest

All of the authors report no biomedical financial interests or poten-tial conflicts of interest.

Author contributions

T. W. Zheng, S. P. Kulikova, and D. Pinault: designed the study, performedexperiments and data analysis, and wrote the manuscript. T. J. O’Brien, C.A. Reid, and M. J. Morris: designed the study and wrote the manuscript.

Acknowledgements

We thank Anne-Sophie Bouillot (INSERM Apprenticeship, ProfessionalLicence, University Joseph Fourier, Grenoble, France) for her excellent tech-nical assistance. T. W. Zheng is the recipient of an Eiffel Doctorate Fellow-ship of Excellence. S. P. Kulikova is a fellow of the Trinational Neurex JointMaster in Neuroscience (Basel-Strasbourg-Freiburg) and of INSERM. Thiswork was supported by grants from the French Institute of Health and Medi-cal Research (INSERM), the Universit�e de Strasbourg, FMTS, Strasbourgand Neurex to D. Pinault. It was also supported by a Project Grant from theNH&MRC (Australia) (#568729) to T. J. O’Brien and M. J. Morris.

Abbreviations

BTA, burst-triggered average; CBZ, carbamazepine; CT, corticothalamic;EEG, electroencephalogram; EPSP, excitatory postsynaptic potential; IPSP,inhibitory postsynaptic potential; SK2, Ca2+-dependent small-conductancetype 2; TC, thalamocortical; TRN, thalamic reticular nucleus.

References

Astori, S., Wimmer, R.D., Prosser, H.M., Corti, C., Corsi, M., Liaudet, N.,Volterra, A., Franken, P., Adelman, J.P. & L€uthi, A. (2011) The Ca(V)3.3calcium channel is the major sleep spindle pacemaker in thalamus. Proc.Natl. Acad. Sci. USA, 108, 13823–13828.

Ayoub, A., Aumann, D., H€orschelmann, A., Kouchekmanesch, A., Paul, P.,Born, J. & Marshall, L. (2013) Differential effects on fast and slow spindleactivity, and the sleep slow oscillation in humans with carbamazepine andflunarizine to antagonize voltage-dependent Na(+) and Ca(2+) channelactivity. Sleep, 36, 905–911.

Bal, T. & McCormick, D.A. (1993) Mechanisms of oscillatory activity inguinea-pig nucleus reticularis thalami in vitro: a mammalian pacemaker.J. Physiol., 468, 669–691.

Bal, T. & McCormick, D.A. (1996) What stops synchronized thalamocorticaloscillations? Neuron, 17, 297–308.

Bal, T., Debay, D. & Destexhe, A. (2000) Cortical feedback controls the fre-quency and synchrony of oscillations in the visual thalamus. J. Neurosci.,20, 7478–7488.

Beenhakker, M.P. & Huguenard, J.R. (2009) Neurons that fire together alsoconspire together: is normal sleep circuitry hijacked to generate epilepsy?Neuron, 62, 612–632.

Biber, K., Fiebich, B.L., Gebicke-H€arter, P. & van Calker, D. (1999) Carba-mazepine-induced upregulation of adenosine A1-receptors in astrocytecultures affects coupling to the phosphoinositol signaling pathway.Neuropsychopharmacol., 20, 271–278.

Blumenfeld, H. & McCormick, D.A. (2000) Corticothalamic inputs controlthe pattern of activity generated in thalamocortical networks. J. Neurosci.,20, 5153–5162.

Bonjean, M., Baker, T., Lemieux, M., Timofeev, I., Sejnowski, T. & Bazhe-nov, M. (2011) Corticothalamic feedback controls sleep spindle durationin vivo. J. Neurosci., 31, 9124–9134.

Brady, K.T. (1989) Weight gain associated with psychotropic drugs. South.Med. J., 82, 611–617.

Budde, T., Biella, G., Munsch, T. & Pape, H.C. (1997) Lack of regulationby intracellular Ca2+ of the hyperpolarization-activated cation current in ratthalamic neurones. J. Physiol., 503(Pt 1), 79–85.

Buzs�aki, G. (2004) Large-scale recording of neuronal ensembles. Nat. Neuro-sci., 7, 446–451.

Cash, S.S., Halgren, E., Dehghani, N., Rossetti, A.O., Thesen, T., Wang, C.,Devinsky, O., Kuzniecky, R., Doyle, W., Madsen, J.R., Bromfield, E.,Eross, L., Hal�asz, P., Karmos, G., Csercsa, R., Wittner, L. & Ulbert, I.(2009) The human K-complex represents an isolated cortical down-state.Science, 324, 1084–1087.

Contreras, D. & Steriade, M. (1996) Spindle oscillation in cats: the role ofcorticothalamic feedback in a thalamically generated rhythm. J. Physiol.,490(Pt 1), 159–179.

Contreras, D., Timofeev, I. & Steriade, M. (1996) Mechanisms of long-last-ing hyperpolarizations underlying slow sleep oscillations in cat corticotha-lamic networks. J. Physiol., 494(Pt 1), 251–264.

Crunelli, V., Cope, D.W. & Terry, J.R. (2011) Transition to absence seizuresand the role of GABA(A) receptors. Epilepsy Res., 97, 283–289.

De Gennaro, L. & Ferrara, M. (2003) Sleep spindles: an overview. SleepMed. Rev., 7, 423–440.

Deschenes, M., Paradis, M., Roy, J.P. & Steriade, M. (1984) Electrophysiol-ogy of neurons of lateral thalamic nuclei in cat: resting properties andburst discharges. J. Neurophysiol., 51, 1196–1219.

Dias, V.V., Balanz�a;-Martinez, V., Soeiro-de-Souza, M.G., Moreno, R.A.,Figueira, M.L., Machado-Vieira, R. & Vieta, E. (2012) Pharmacologicalapproaches in bipolar disorders and the impact on cognition: a criticaloverview. Acta Psychiat. Scand., 126, 315–331.

Domich, L., Oakson, G. & Steriade, M. (1986) Thalamic burst patterns inthe naturally sleeping cat: a comparison between cortically projecting andreticularis neurones. J. Physiol., 379, 429–449.

Ferrarelli, F., Huber, R., Peterson, M.J., Massimini, M., Murphy, M.,Riedner, B.A., Watson, A., Bria, P. & Tononi, G. (2007) Reducedsleep spindle activity in schizophrenia patients. Am. J. Psychiat., 164,483–492.

Ferrarelli, F., Peterson, M.J., Sarasso, S., Riedner, B.A., Murphy, M.J.,Benca, R.M., Bria, P., Kalin, N.H. & Tononi, G. (2010) Thalamic dys-function in schizophrenia suggested by whole-night deficits in slow andfast spindles. Am. J. Psychiat., 167, 1339–1348.

Genton, P. (2000) When antiepileptic drugs aggravate epilepsy. BrainDev.-Jpn., 22, 75–80.

Granger, P., Biton, B., Faure, C., Vige, X., Depoortere, H., Graham, D., Langer,S.Z., Scatton, B. & Avenet, P. (1995) Modulation of the gamma-aminobutyr-ic acid type A receptor by the antiepileptic drugs carbamazepine and phenyt-oin. Mol. Pharmacol., 47, 1189–1196.

Huguenard, J.R. & Prince, D.A. (1992) A novel T-type current underlies pro-longed Ca(2+)-dependent burst firing in GABAergic neurons of ratthalamic reticular nucleus. J. Neurosci., 12, 3804–3817.

Jones, E.G. (1985) The thalamus. Plenum Press, New York.Kilkenny, C., Browne, W., Cuthill, I.C., Emerson, M. & Altman, D.G.

(2010) Animal research: reporting in vivo experiments: the ARRIVEguidelines. Brit. J. Pharmacol., 160, 1577–1579.

von Krosigk, M., Bal, T. & McCormick, D.A. (1993) Cellular mechanismsof a synchronized oscillation in the thalamus. Science, 261, 361–364.

Kulikova, S.P., Tolmacheva, E.A., Anderson, P., Gaudias, J., Adams, B.E.,Zheng, T. & Pinault, D. (2012) Opposite effects of ketamine and deepbrain stimulation on rat thalamocortical information processing. Eur.J. Neurosci., 36, 3407–3419.

Lancel, M., Cr€onlein, T.A. & Faulhaber, J. (1996) Role of GABAA receptorsin sleep regulation. Differential effects of muscimol and midazolam onsleep in rats. Neuropsychopharmacol., 15, 63–74.

Leucht, S., Kissling, W., McGrath, J. & White, P. (2007) Carbamazepine forschizophrenia. Cochrane Db. Syst. Rev., CD001258.

Levy, A., Chong, S.K. & Price, J.F. (1985) Carbamazepine-induced drowsi-ness. Lancet, 2, 221–222.

Liu, L., Zheng, T., Morris, M.J., Wallengren, C., Clarke, A.L., Reid, C.A.,Petrou, S. & O’Brien, T.J. (2006) The mechanism of carbamazepine aggra-vation of absence seizures. J. Pharmacol. Exp. Ther., 319, 790–798.

Llin�as, R.R. (1988) The intrinsic electrophysiological properties of mamma-lian neurons: insights into central nervous system function. Science, 242,1654–1664.

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

CBZ potentiates thalamocortical spindles 11

McCormick, D.A. & Pape, H.C. (1990) Properties of a hyperpolarization-activated cation current and its role in rhythmic oscillation in thalamicrelay neurones. J. Physiol., 431, 291–318.

Meuth, S.G., Kanyshkova, T., Meuth, P., Landgraf, P., Munsch, T., Ludwig,A., Hofmann, F., Pape, H.C. & Budde, T. (2006) Membrane resting poten-tial of thalamocortical relay neurons is shaped by the interaction amongTASK3 and HCN2 channels. J. Neurophysiol., 96, 1517–1529.

M€olle, M., Marshall, L., Gais, S. & Born, J. (2002) Grouping of spindleactivity during slow oscillations in human non-rapid eye movement sleep.J. Neurosci., 22, 10941–10947.

Monji, A., Maekawa, T., Yanagimoto, K., Yoshida, I. & Hashioka, S. (2004)Carbamazepine may trigger new-onset epileptic seizures in an individualwith autism spectrum disorders: a case report. Eur. Psychiat., 19, 322–323.

Paxinos, G. & Watson, C. (1998) The Rat Brain in Stereotaxic Coordinates,4th edn. Academic Press, San Diego.

Pfurtscheller, G., Stanc�ak, A. Jr & Edlinger, G. (1997) On the existence ofdifferent types of central beta rhythms below 30 Hz. Electroen. Clin.Neuro., 102, 316–325.

Pinault, D. (1996) A novel single-cell staining procedure performed in vivounder electrophysiological control: morpho-functional features of juxtacel-lularly labeled thalamic cells and other central neurons with biocytin orNeurobiotin. J. Neurosci. Meth., 65, 113–136.

Pinault, D. (2003) Cellular interactions in the rat somatosensory thalamocorti-cal system during normal and epileptic 5–9 Hz oscillations. J. Physiol.,552, 881–905.

Pinault, D. (2004) The thalamic reticular nucleus: structure, function andconcept. Brain Res. Brain Res. Rev., 46, 1–31.

Pinault, D., Leresche, N., Charpier, S., Deniau, J.M., Marescaux, C., Verg-nes, M. & Crunelli, V. (1998) Intracellular recordings in thalamic neuronesduring spontaneous spike and wave discharges in rats with absenceepilepsy. J. Physiol., 509, 449–456.

Pinault, D., Vergnes, M. & Marescaux, C. (2001) Medium-voltage 5–9-Hzoscillations give rise to spike-and-wave discharges in a genetic model ofabsence epilepsy: in vivo dual extracellular recording of thalamic relay andreticular neurons. Neuroscience, 105, 181–201.

Pinault, D., Sl�ezia, A. & Acs�ady, L. (2006) Corticothalamic 5–9 Hz oscilla-tions are more pro-epileptogenic than sleep spindles in rats. J. Physiol.,574, 209–227.

Polack, P.O., Guillemain, I., Hu, E., Deransart, C., Depaulis, A. & Charpier,S. (2007) Deep layer somatosensory cortical neurons initiate spike-and-wave discharges in a genetic model of absence seizures. J. Neurosci., 27,6590–6599.

Post, R.M. (1982) Use of the anticonvulsant carbamazepine in primary andsecondary affective illness: clinical and theoretical implications. Psychol.Med., 12, 701–704.

Silber, M.H., Ancoli-Israel, S., Bonnet, M.H., Chokroverty, S., Grigg-Dam-berger, M.M., Hirshkowitz, M., Kapen, S., Keenan, S.A., Kryger, M.H.,Penzel, T., Pressman, M.R. & Iber, C. (2007) The visual scoring of sleepin adults. J. Clin. Sleep Med., 3, 121–131.

Sillito, A.M., Jones, H.E., Gerstein, G.L. & West, D.C. (1994) Feature-linkedsynchronization of thalamic relay cell firing induced by feedback from thevisual cortex. Nature, 369, 479–482.

Simhandl, C. & Meszaros, K. (1992) The use of carbamazepine in the treat-ment of schizophrenic and schizoaffective psychoses: a review. J. Psychia-tr. Neurosci., 17, 1–14.

Steriade, M. (1995) Thalamic origin of sleep spindles: Morison and Bassett(1945). J. Neurophysiol., 73, 921–922.

Steriade, M. (2006) Grouping of brain rhythms in corticothalamic systems.Neuroscience, 137, 1087–1106.

Steriade, M., Domich, L., Oakson, G. & Deschenes, M. (1987) The deaffe-rented reticular thalamic nucleus generates spindle rhythmicity. J. Neuro-physiol., 57, 260–273.

Tamminen, J., Payne, J.D., Stickgold, R., Wamsley, E.J. & Gaskell, M.G.(2010) Sleep spindle activity is associated with the integration of newmemories and existing knowledge. J. Neurosci., 30, 14356–14360.

Tiihonen, J., Kajola, M. & Hari, R. (1989) Magnetic mu rhythm in man.Neuroscience, 32, 793–800.

Timofeev, I. (2011) Neuronal plasticity and thalamocortical sleep and wakingoscillations. Prog. Brain Res., 193, 121–144.

Tononi, G. & Cirelli, C. (2012) Time to be SHY? Some comments on sleepand synaptic homeostasis. Neural Plast., 2012, 415250.

Tort, A.B., Fontanini, A., Kramer, M.A., Jones-Lush, L.M., Kopell, N.J. &Katz, D.B. (2010) Cortical networks produce three distinct 7–12 Hzrhythms during single sensory responses in the awake rat. J. Neurosci., 30,4315–4324.

Ulrich, D. & Huguenard, J.R. (1997) GABA(A)-receptor-mediated reboundburst firing and burst shunting in thalamus. J. Neurophysiol., 78, 1748–1751.

Vergnes, M., Marescaux, C., Micheletti, G., Reis, J., Depaulis, A., Rumbach,L. & Warter, J.M. (1982) Spontaneous paroxysmal electroclinical patternsin rat: a model of generalized non-convulsive epilepsy. Neurosci. Lett., 33,97–101.

Walden, J., Grunze, H., Mayer, A., D€using, R., Schirrmacher, K., Liu, Z. &Bingmann, D. (1993) Calcium-antagonistic effects of carbamazepine inepilepsies and affective psychoses. Neuropsychobiology, 27, 171–175.

Wimmer, R.D., Astori, S., Bond, C.T., Rov�o, Z., Chatton, J.Y., Adelman,J.P., Franken, P. & L€uthi, A. (2012) Sustaining sleep spindles throughenhanced SK2-channel activity consolidates sleep and elevates arousalthreshold. J. Neurosci., 32, 13917–13928.

Winsky-Sommerer, R. (2009) Role of GABAA receptors in the physiologyand pharmacology of sleep. Eur. J. Neurosci., 29, 1779–1794.

Yang, J.D., Elphick, M., Sharpley, A.L. & Cowen, P.J. (1989) Effects of car-bamazepine on sleep in healthy volunteers. Biol. Psychiat., 26, 324–328.

Yuan, B., Morrow, T.J. & Casey, K.L. (1986) Corticofugal influences of S1cortex on ventrobasal thalamic neurons in the awake rat. J. Neurosci., 6,3611–3617.

Zhang, L. & Jones, E.G. (2004) Corticothalamic inhibition in the thalamicreticular nucleus. J. Neurophysiol., 91, 759–766.

Zheng, T., Clarke, A.L., Morris, M.J., Reid, C.A., Petrou, S. & O’Brien, T.J.(2009) Oxcarbazepine, not its active metabolite, potentiates GABAA acti-vation and aggravates absence seizures. Epilepsia, 50, 83–87.

Zheng, T.W., O’Brien, T.J., Morris, M.J., Reid, C.A., Jovanovska, V., O’Brien,P., van Raay, L., Gandrathi, A.K. & Pinault, D. (2012) Rhythmic neuro-nal activity in S2 somatosensory and insular cortices contribute to theinitiation of absence-related spike-and-wave discharges. Epilepsia, 53,1948–1958.

© 2013 Federation of European Neuroscience Societies and John Wiley & Sons LtdEuropean Journal of Neuroscience, 1–12

12 T. W. Zheng et al.