Rhythmic neuronal activity in S2 somatosensory and insular cortices contribute to the initiation of absence-related spike-and-wave discharges


  • Thomas W. Zheng,

    1. Departments of Medicine, Surgery and Neurology, The Royal Melbourne Hospital, The University of Melbourne, Parkville, Victoria, Australia
    2. INSERM U666, Pathophysiology and Psychopathology of Schizophrenia, Strasbourg, France
    3. University of Strasbourg, Faculty of Medicine, Strasbourg, France
    Search for more papers by this author
  • Terence J. O’Brien,

    1. Departments of Medicine, Surgery and Neurology, The Royal Melbourne Hospital, The University of Melbourne, Parkville, Victoria, Australia
    2. INSERM U666, Pathophysiology and Psychopathology of Schizophrenia, Strasbourg, France
    3. University of Strasbourg, Faculty of Medicine, Strasbourg, France
    Search for more papers by this author
  • Margaret J. Morris,

    1. Department of Pharmacology, University of New South Wales, Kensington, New South Wales, Australia
    Search for more papers by this author
  • Christopher A. Reid,

    1. Florey Neuroscience Institutes, Melbourne Brain Centre, Melbourne, Australia
    Search for more papers by this author
  • Valentina Jovanovska,

    1. Departments of Medicine, Surgery and Neurology, The Royal Melbourne Hospital, The University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Patrick O’Brien,

    1. Departments of Medicine, Surgery and Neurology, The Royal Melbourne Hospital, The University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Leena van Raay,

    1. Departments of Medicine, Surgery and Neurology, The Royal Melbourne Hospital, The University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Arun K. Gandrathi,

    1. Departments of Medicine, Surgery and Neurology, The Royal Melbourne Hospital, The University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Didier Pinault

    1. INSERM U666, Pathophysiology and Psychopathology of Schizophrenia, Strasbourg, France
    2. University of Strasbourg, Faculty of Medicine, Strasbourg, France
    Search for more papers by this author

Address correspondence to: Terence J. O’Brien, Department of Medicine, The Royal Melbourne Hospital, University of Melbourne, Royal Parade, Parkville 3050, Vic., Australia. E-mail: obrientj@unimelb.edu.au


Purpose:  The origin of bilateral synchronous spike-and-wave discharges (SWDs) that underlie absence seizures has been widely debated. Studies in genetic rodent models suggest that SWDs originate from a restricted region in the somatosensory cortex. The properties of this initiation site remain unknown. Our goal was to characterize the interictal, preictal and ictal neuronal activity in the primary and secondary cortical regions (S1, S2) and in the adjacent insular cortex (IC) in Genetic Absence Epilepsy Rats from Strasbourg (GAERS).

Methods:  We performed electroencephalography (EEG) recordings in combination with multisite local field potential (LFP) and single cell juxtacellular recordings, and cortical electrical stimulations, in freely moving rats and those under neurolept-anesthesia.

Key Findings:  The onset of the SWDs was preceded by 5–9 Hz field potential oscillations, which were detected earlier in S2 and IC than in S1. Sustained SWDs could be triggered by a 2-s train of 7-Hz electrical stimuli at a lower current intensity in S2 than in S1. In S2 and IC, subsets of neurons displayed rhythmic firing (5–9 Hz) in between seizures. S2 and IC layers V and VI neurons fired during the same time window, whereas in S1 layer VI, neurons fired before layer V neurons. Just before the spike component of each SW complex, short-lasting high-frequency oscillations consistently occurred in IC ∼20 msec before S1.

Significance:  Our findings demonstrate that the S2/IC cortical areas are a critical component of the macro-network that is responsible for the generation of absence-related SWDs.

The neural mechanisms underlying the electrogenesis of the generalized absence seizure related spike-and-wave discharges (SWDs) have been the subject of many in vitro and in vivo studies (McCormick & Contreras, 2001; Pinault & O’Brien, 2005), giving rise to several hypotheses that are still widely debated. Two of the most common animal models used to study this are the Genetic Absence Epilepsy Rats from Strasbourg (GAERS) (Vergnes et al., 1987), and the Wistar Albino Glaxo/Rijswijk (WAG/Rij) rats (Coenen et al., 1992). These are genetic models of absence epilepsy that manifest spontaneous, generalized absence-type seizures accompanied by SWDs on electroencephalography (EEG). Recent data from these models have suggested that SWDs arise from a topographically localized “focus,” more specifically the primary somatosensory (S1) neocortex (Meeren et al., 2002; Polack et al., 2007). In addition, selective pharmacologic deactivation of the S1 perioral cortex in GAERS suppressed SWDs, confirming an involvement of this region in seizure initiation (Sitnikova & van Luijtelaar, 2004; Polack et al., 2009). However, recent data from our group have also implicated the secondary somatosensory cortex (S2) in seizure initiation in GAERS. Maximal suppression of SWDs following focal injections of the endogenous anticonvulsant neuropeptide Y occurred when injected into the S2, compared to S1 or M1 cortices (van Raay et al., 2012).

At the cellular level, paired cortical and thalamic juxtacellular recordings have demonstrated that S1 layer VI corticothalamic neurons play a leading role in the generation of SWDs, triggering synchronized rhythmic activities in related thalamocortical (TC) and thalamic reticular nucleus (TRN) neurons (Pinault et al., 2001; Pinault, 2003). The study also demonstrated that S1 layer VI neurons fire before layer V neurons during the generation of absence-related SWDs, and that preictal 5–9 Hz rhythmic activity occurs first in layer VI of S1 relative to the interconnected somatosensory thalamus.

The nature of the cellular and network activity in the somatosensory cortical region that leads to the onset of seizure-related SWDs remains uncertain. This study aimed to characterize, in GAERS, the spatiotemporal dynamics of intracortical network and cellular activities in S1 cortex, and in the neighboring S2 and insular cortices (IC), before and during seizures. The findings in GAERS were compared to that of the nonepileptic control (NEC) rat strain, which was derived from the same original Wistar colony but selectively bred not to have any seizures (Vergnes et al., 1987). We revealed the presence of cellular and network activities in S2 and IC cortices that lead S1 cortical activity prior, and during, the generation of absence-related SWDs.



Experiments were conducted in inbred adult male Wistar rats (30 GAERS and 13 NEC rats). All procedures were approved by Université de Strasbourg (AL/01/23/11/07) and the University of Melbourne Animal Ethics Committee (AEC #0810999) and performed in accordance with the guidelines published by the Australian NHMRC and European Union for use of animals in research. GAERS and NECs were originally bred from the same Wistar colony in Strasbourg (Vergnes et al., 1987). All rats were progenies of the original strains and were born and housed in standard conditions. Rats were at least 13 weeks of age and weighed between 250 and 350 g at the time of the experiments. Every precaution was taken to minimize stress and the number of animals used for each series of experiments.

Multisite depth EEG recordings and electrical stimulation in freely moving rats

GAERS (n = 5) and NEC rats (n = 3) were chronically implanted with multisite stainless steel insulated electrodes (diameter 0.127 mm) using stereotactic guidance under general anesthesia (isoflurane, 5% induction, 1.5–2.5% maintenance). Each electrode bundle contained four insulated wires with tips exposed at 1.0, 3.0, and 5.0 mm from the most distal contact. Electrodes were implanted bilaterally [anteroposterior (AP) +0.2 mm, mediolateral (ML) ±3.6 mm and dorsoventral (DV) 6 mm, 20-degree angle] so that the contacts reached the S1FL, S1ULp, S2, and the IC (Fig. 1A) (Paxinos & Watson, 1998). Separate electrodes were implanted into the motor cortices (AP −2.7 mm, ML ±2.6 mm and DV 1.8 mm). Extradural electrodes were implanted into the parietal bone (AP −8 mm, ML ±4) for ground and reference points. All rats were individually housed postsurgery and given a 7-day recovery period. At least two 60-min recordings were acquired for each rat at 256 Hz (Compumedics EEG, Melbourne, Vic., Australia). Bilateral recordings were also performed in GAERS (n = 3) for high-frequency oscillations in the S1 cortex (sampling rate: 10 kHz; bandpass: 0.1–800 Hz).

Figure 1.

SWDs are first detected within the S2 somatosensory cortex. (A) Sites of electrode placements of the multisite EEG recordings acquired in freely moving rats (Paxinos & Watson, 1998). Electrode coordinates with reference to bregma—S1FL, S1ULp, S2 and IC entry point: Anteroposterior (AP), +0.2 mm, mediolateral (ML), ±3.6 mm, angle 20 degrees. Bilateral M1 electrodes: AP +1.0 mm, ML ±2.6 mm. (B) Typical EEG recording trace demonstrating the oscillatory rhythmic discharge commences unilaterally in at the one S2 electrode (arrowed) before spreading superiorly to the S1ULp and inferiorly to the IC electrodes, and then to the S1FL and then finally the M1 and the electrodes in the contralateral hemisphere. (C) Amplitude coherence analysis using bootstrap analysis method by averaging 200 randomly selected recording episodes to form a surrogate distribution, with significance level set to p = 0.05. Black arrow represents the start of generalized SWD. Fast Fourier transformation analysis (Blue) shows the dominant frequency of 5–9 Hz. An increase in amplitude coherence (Green trace and contour plot) was first observed between the S2 and S1 ULp, as well as S1 ULp and S1 FL electrodes, seconds before SWD initiation (black arrow).

Monopolar stimulus trains (1 ms square pulses, 2 s, 7 Hz) were delivered to each region in a randomized order (stimulator isolator A365, accupulser A360; World Precision Instruments, Sarasota, FL, U.S.A.) via the recording electrodes. The intensity of the stimuli was initially 50 μA. If the train did not trigger typical generalized SWDs of >1 s, the intensity of the stimuli was increased by increments of 25–100 μA (up to 600 μA) until a seizure was triggered. Stimulations of the same intensities were repeated three times with intervals of at least 30 s. The lowest stimulus intensity that could trigger a typical SWD in at least two of the three trials was determined. At the end of the recording, a DC current (2 mA for 5 s) was passed through the recording electrodes to induce a lesion. Rats were euthanized with pentobarbitone (Lethabarb, 200 mg/kg, i.p.) and perfused transcardially with phosphate buffered saline (PBS), followed by 4% paraformaldehyde and 0.5% potassium ferricyanide in PBS to visualize the electrode recording tip positions on cryostat cut axial sections (Adamec et al., 2001).

Electrophysiological recordings under neuroleptanalgesia

Experiments were performed in 25 GAERS and 10 NEC rats. The initial surgical procedures were made under general anesthesia (pentobarbital 40 mg/kg i.p., Sanofi, Libourne, France; and ketamine 50 mg/kg, i.m., Merial, Lyon, France). Catheterization of the penile vein and tracheotomy were performed before the rats were placed in a stereotaxic frame. Craniotomy and durotomy procedures were performed for optimal recording conditions (Pinault, 2005). At the completion of surgeries, neuroleptanalgesia was initiated and maintained by continuous intravenous injection (0.5 ml/h for a 300 g rat) of the following mixture: d-tubocurarine chloride (0.4 mg, Sigma-Aldrich, Saint-Quentin Fallavier, France), fentanyl (1 μg; Janssen, Boulogne-Billancourt, France), Haldol (100 μg; Janssen), and glucose (25 mg). Rats were artificially ventilated (SAR-830; CWE, Ardmore, PA, U.S.A.) in pressure mode (8–12 cmH2O; 60–65 breath/min). Depth of anesthesia was assessed by continuous monitoring of the EEG and heart rate; injection rate was adjusted accordingly. Body temperature was maintained at physiologic level (37.5–38.3°C). The electrophysiologic recordings commenced at least 3 h following the induction of the neuroleptanalgesia.

Glass micropipettes were filled with 1.5%N-(2-amino-ethyl) biotin amide hydrochloride (Neurobiotin; Vector Labs, Burlingame, CA, U.S.A.) dissolved in 1 m CH3COOK. Pipettes were connected to an intracellular recording amplifier (NeuroData IR-283; Cygnus Technology Inc., Delaware Water Gap, PA, U.S.A.) and were lowered with stepping microdrivers (Burleigh, Fishers, NY, U.S.A.) for monopolar recordings. Paired local field potential (LFP) recordings were also performed in the S1 (AP −0.2, ML +4.8 mm, angle 0 degree, DV 1.8–2.0 mm) and S2/IC (AP −0.2, ML +3.4 mm, angle 20 degrees, DV 4.9–5.1 mm) in seven GAERS. For single-unit juxtacellular recordings, dual pipettes (15–50 MΩ) were lowered into the S1, S2, and IC in 15 GAERS and 5 NEC rats. A ground wire was inserted into the neck muscle. Data were processed with band passes of 0.1–800 Hz for the EEG, and 0.1–6,000 Hz for cellular activity (Cyber-Amp 380; Axon Instruments, Foster City, CA, U.S.A.). A sampling rate per channel of >2.5 kHz were used for the EEG and LFP, 20 kHz for single cell recordings (Digidata 1200B; Axon Instruments). All recordings were performed simultaneously with surface EEG. At the end of the recording sessions, Neurobiotin was applied to the recording site (500–600 nA, 200 msec on/off current, 5–10 min) or individual neurons using juxtacellular technique (Pinault, 1996). In all single-cell recording experiments, only the last two recorded neurons were labeled.

Histology and region/cellular identification

At the end of the experiments, rats were euthanized with pentobarbitone overdose (>40 mg/kg, i.v.) and were transcardially perfused with PBS followed by 4% paraformaldehyde (PFA and 0.5% glutaraldehyde in PBS. Brains were postfixed overnight in the same fixative without glutaraldehyde before being processed using standard histologic techniques (Pinault, 1996). Tracer-filled regions and/or neurons were examined with a light microscope and the location of the marked region, and neurons were ascertained by consulting the rat brain atlas (Paxinos & Watson, 1998).

Sitnikova et al. (2008) described changes in Granger causality within 5 s after the last spike of SWD. Based on this study, and to avoid any overlapping between interictal and preictal periods, we analyzed only the SWDs that were at least 20 s after the last seizure, where the preictal period was characterized by low-voltage, desynchronized EEG.

Data analysis

Recordings were reviewed with Compumedics or Clampex. Ictal periods were classified as SWDs of duration of >0.5 s, an amplitude of more than three times baseline (>0.5 mV) and a frequency of 7–12 Hz in freely moving rats and 5–9 Hz in rats under neuroleptanalgesia. The preictal period was defined as 10 s periods before the start of an SWD. Interictal periods were defined as the time from the end of an ictal event to the beginning of the next ictal event, where the EEG shows desynchronized background (<0.2 mV). To avoid overlaps of interictal and preictal periods (Sitnikova et al., 2008), only the SWD that occurred at least 20 s after the last SWD episode was analyzed. The start of preictal seizure precursor oscillatory activity is defined as the beginning of a rising phase of an oscillatory event that is at least twice the amplitude of the desynchronized background at SWD frequency. Event detections and autocorrelograms of multi- and single-unit activity were computed using DataWave software (SCIWORKS, v5.2; DataWave Technologies, Berthoud, CO, U.S.A.). Data are presented as means ± standard error of the mean (SEM) and were evaluated for statistical significance with appropriate tests with the significance level set to 0.05.

Amplitude coherence analysis between channels was performed using EEGLAB (time frequency analysis toolbox), with the amplitude coherence mathematically defined (Delorme & Makeig, 2004). Power spectra were calculated using a sinusoidal wavelet short time Fourier transform. Statistical significance of cross-correlation and coherence analysis (significance set to p = 0.05) was assessed using bootstrap analysis method. An average of 200 randomly selected SWD episodes were used to form a surrogate distribution, from which significance levels of deviations in coherence were assessed.


The results of this study are based on surface and depth EEG recordings from freely moving GAERS (n = 12) and EEG and juxtacellular recordings in GAERS (n = 19) and NECs (n = 5) under neuroleptanalgesia. All GAERS developed spontaneous, bilaterally synchronous SWDs. Neuroleptanalgesia maintains the rats in a stress-free and pain-free state and has been shown to facilitate SWD generation without altering the electroclinical characteristics of SWD-related seizures (Pinault et al., 2001; Pinault, 2003) other than slowing the cycle frequency compared to those recorded in freely moving rats (5–9 vs. 7–11 Hz, Fig. S1).

Preictal oscillations in S2/IC cortical areas lead S1 and motor cortical regions at the onset of generalized SWDs

Multisite depth EEG recordings in freely moving GAERS revealed typical generalized SWDs at 5–9 Hz accompanied by sudden behavior arrests. In 60% of seizures (60 of 100 SWDs, 20 SWDs examined from each of five rat), 5–9 Hz medium voltage (0.2–0.5 mV) oscillations could be observed during the interictal-ictal transition from either the left, or right S2, or both. Preictal oscillations were initially seen at the S2 electrode in four of five rats and at the IC electrode in the other. Oscillations were then detected progressively, over 1–2 s, from the S1ULp, S1FL electrodes to the M1 electrodes as the discharge transformed into a typical, high-voltage (0.5–1.5 mV) generalized SWDs, which are considered to be the hallmark of the absence-like seizures in GAERS (Fig. 1B). On average, 5–9 Hz oscillations began 1.40 ± 0.08 s (maximum 3.70 s) in S2, before SWDs become generalized and detected on all electrodes. Amplitude coherence analysis was performed for each adjacent electrode contacts (Fig. 1C). Fast Fourier transformation analysis (Fig. 1C, blue trace) reveals that oscillations and SWDs occur at the frequency of 5–9 Hz. An increase in amplitude coherence (Green trace and yellow band on contour plot) was observed between S2 and S1ULp, as well as S1 ULp and S1 s before the onset of generalized SWDs (black arrow). This is followed by a dramatic increase in coherence at the start of the SWDs in all electrode pairs (Fig. 1C).

SWDs could also be observed to start abruptly and simultaneously from all recording electrodes from both hemispheres with no apparent temporal difference between the recorded regions (Fig. S2A, 40% of seizures, 40 of 100 SWDs). It is also interesting to note that oscillations could occur bilaterally in S2 without developing into generalized SWDs (Fig. S2B).

To further investigate the timing of oscillatory activity onset in S2/IC versus S1 during SWDs, paired LFP recordings were performed simultaneously in the same cortical layers of S1 and S2/IC (layers V and VI, Fig. 2D) along with the EEG over the frontoparietal cortex in neurolept-analgesia rats (Fig. 2C, n = 7). LFP recordings revealed short-lasting high-frequency oscillations (HFOs) at between 350 and 450 Hz (Fig. 2D) in S1, S2, and IC, which always appeared before each SW complex on the EEG (Fig. 2). The time delay between the HFO and the EEG spike was measured for both channels. HFOs in S2/IC occurred at each cycle of the SWDs ∼20 msec before those in S1 (Fig. 2B,C, p < 0.001, n = 140). HFOs were not specific to the experimental conditions, since they were recorded in free-moving GAERS (Fig. S3).

Figure 2.

High-frequency oscillations (HFOs) occur in S2/IC before S1 during seizures. (A1) Trace of filtered (300–800 Hz) local field potential (LFP) recordings of S2/IC and S1 and EEG of the related cortex. Gray band shows a 1-s recording expanded in A2. HFOs occurs earlier in S2 compared to the S1 region (−66.4 ± 2.8 vs. −40.7 ± 2.4 msec, with reference to surface EEG, n = 140, t-test, p < 0.001). (B) An expansion of the gray band from A2, showing a single spike during a typical seizure showing nonfiltered (0.1–6 kHz) and filtered (300–800 Hz) LFP from S2/IC and the S1. (C) Time between the onset of HFO and the EEG spike in S2/IC and S1 (66.02 ± 3.3 msec vs. 46.18 ± 3.4 msec, respectively, *paired t-test, p < 0.001). (D) Power spectrum computed from fast Fourier transformation analysis show that the HFOs occur at frequencies between 350 and 450 Hz. (E1) Site of paired S1 vs. S2/IC recordings were labeled on a rat brain atlas. (E2) An example of histologic staining revealing site of labeling (black labeling) and site of recording (white stars).

SWDs can be triggered by 7-Hz electrical stimulus trains in the cortex at a lower intensity in S2 than S1 in GAERS

To address the question of whether the existence of precursor neuronal activities in S2 and IC relative to S1 and motor cortex renders S2 and IC regions are more hyperexcitable, a 2-s train of 7-Hz electrical stimuli was applied to each of the cortical regions. In GAERS, SWDs could be consistently evoked by a single stimulation train delivered to all four cortical regions in all rats (n = 7). The SWDs evoked by the stimulation were identical in morphology and frequency to that of typical, spontaneously occurring SWDs (Fig. 3A,B). They were also accompanied by the same behavioral characteristics, including immobility and head-nodding, to those of spontaneous seizures. In addition, the duration of the evoked SWDs was not different from that of the spontaneous events (Fig. 3C). It is noteworthy that the seizures could only be induced when rats were in a state of quiet wakefulness accompanied by desynchronized EEG, and not when they were in active or sleep states. It is important to note that a significantly smaller current was required at the S2 region to evoke an SWD compared to the adjacent S1ULp region (Fig. 3D). There was no difference in the current needed to stimulate seizures in S2 and IC. In contrast, identical stimulations in the NEC rats (100–600 μA, n = 3) could not induce SWDs or the accompanying behavioral manifestations. Stimulations evoked 2–3 cycle (<0.5 s) medium-voltage 5–9 Hz oscillations before returning to a baseline EEG (Fig. 3E).

Figure 3.

Cortical electrical stimulation in GAERS and NEC rats. (A) A 10-s trace showing simultaneous depth EEG recording of S1Fl, S1Ulp, and IC before, during, and after electrical stimulation of S2. Typical SWD events were induced. (B) Mean frequency of spontaneous versus evoked SWDs in GAERS (7.3 ± 0.1, n = 10 vs. 7.6 ± 0.1, n = 10, respectively, unpaired t-test, p = 0.15). (C) Mean duration of spontaneous vs. evoked SWDs (mean ± SEM, respectively, 12.0 ± 1.35 s, n = 45, vs. 9.5 ± 0.79 s, n = 49, unpaired t-test, p = 0.12). (D) Mean current (μA) required to consistently evoke seizures in the S1Fl, SlUlp, S2, and IC of GAERS. A smaller current is needed in the S2 to evoke SWDs compared to the adjacent S1Ulp (146.4 ± 30.6 vs. 257.1 ± 56.4, n = 7, unpaired t-test, *p = 0.025). (E) Electrical stimulations in the NECs evoke short episodes of small to medium voltage 7-Hz oscillations but no SWD events occur (n = 3)

Preictal 5–9 Hz rhythmic activities in S2

To understand the spatiotemporal dynamics of the preictal oscillatory network activities in S1, S2, and IC, multiunit and LFP recordings were conducted in GAERS (n = 10) under neuroleptanalgesia. Using glass electrodes (diameter 5–8 μm), rhythmic field potential oscillations at SWD frequency were recorded in S2 >10 s before the SWD (Figs 4 and S4). Rhythmic activity could be observed to commence many seconds before seizure onset and could be sustained for minutes throughout interictal periods, whereas the surface cortical EEG remained at a low-voltage (<0.2 mV) desynchronized background level (Fig. 4). This is similar to the preictal oscillations observed in freely moving rats, where seizure precursor activity was first observed in the S2 before propagating to the surrounding networks.

Figure 4.

Sustained interictal multiunit activity from the S2 cortex. (A) 30-s multiunit recording trace of the S2 cortex simultaneously with the EEG of the related frontal parietal somatosensory cortex. (B1 and B2) A 6-s selection of (A) preictal and ictal periods, respectively, showing multiunit activity during interictal periods and increase in amplitude during seizures. (C1 and C2) Autocorrelograms (2-msec bin width) of interictal and ictal periods confirming the highly rhythmic nature of the multiunit discharge.

Preictal rhythmic firing is predominant in deep layers of S2/IC

To identify the cellular correlates of the preictal rhythmic multiunit events described above, juxtacellular recordings of single neurons were performed within the S1, S2, and IC (178 cells from 16 GAERS; 78 cells from 5 NEC rats). Based on independent observations by two investigators blinded to the location of the cell, followed by confirmation with autocorrelation analysis, two distinct types of preictal cellular firing patterns were observed (Fig. 5): (1) Rhythmic cells that predominantly fired at the SWD frequency of 5–9 Hz (21% [37 of 178], Fig. 5A1–3,C1). This firing pattern was sustained and persisted for seconds to minutes before, during and often after the occurrence of an SWD on the surface EEG (Fig. 5C1). A small number of these rhythmic cells (9 of 178) fired at slightly higher frequency (8–15) Hz during interictal periods and slowed to 5–9 Hz up to 2 s prior to the onset of SWDs. (2) Non-rhythmic cells that fired irregularly during interictal periods before the abrupt switching to 5–9 Hz rhythmic firing with the commencement of the SWDs on the EEG (79% [141 of 178], Fig. 5B1–3,C2). Despite the distinct preictal firing patterns, during SWDs all recorded neurons fired in a highly synchronous rhythmic manner, phase-locked with every SW complex (Fig. 5A,B). In addition, there is a trend for the S2 neurons to firing rhythmically before those in the S1, although this did not reach statistical significance (Fig. 5D, p = 0.0894). The stereotaxic locations of all recorded cells were plotted on the rat brain atlas (Fig. 5E). Neurons that fired rhythmically during preictal and interictal periods (black) were predominantly clustered in a restricted region centered in S2, but extends to include the dorsal IC and ventral S1Ulp regions. Rhythmic cells were found in both layers V and VI, and their principal morphologic features were not apparently different to those of the nonrhythmic cells (Fig. 5F1–2).

Figure 5.

Identification of rhythmic and nonrhythmic cell types in GAERS under neurolept-anesthesia. Juxtacellular recording were performed in the GAERS somatosensory cortex simultaneously with surface EEG during preictal and ictal periods. (A) Inherently rhythmic cells (21%)—autocorrelogram reveals the 5–9 Hz rhythmicity before and during seizures. (B) Nonrhythmic cells (79%)—autocorrelogram reveals no rhythmicity before seizures and 5–9 Hz ictal rhythmicity. (C1 and C2) Average instantaneous frequency of two successive APs from each of the two cell types 30 s before and after seizure onset (1-s bin width), time zero denotes start of an SWD event. (D) S2 neurons appeared to fire rhythmically before those in the S1 during preictal periods (median ± SD, S2 vs. S1, −0.52 ± 1.3 s, n = 46 vs. −0.32 ± 1.2, n = 38, Mann-Whitney U test, p = 0.0894). (E) Stereotaxic location of all recorded neurons with reference to the Paxino and Watson Rat Brain Atlas (Paxinos & Watson, 1998). Neurons that show interictal rhythmicity are plotted in black. (F) Microphotograph of a rhythmic (F1) and nonrhythmic (F2) cortical layer V neuron.

Of interest, a similar proportion of rhythmic cells that were found in GAERS were also identified in NEC rats, with 13 (17%) of 78 cells exhibiting preictal rhythmic firing.

GAERS layer V and VI neurons in S2 show greater burst firing than those of NEC rats

The interictal firing characteristics of the juxtacellularly recorded layer V and VI neurons were quantitatively compared between GAERS and NEC rats for S1, S2, and IC (Table S1). There was no difference in the neuronal firing rate between GAERS and NEC rats in all cortical regions examined. Neurons fire as either a single action potential (AP) or AP bursts that are defined as two or more APs occurring no more than 6 msec apart from its adjacent APs. In both GAERS and NEC rats, AP bursts consisted of up to 14 APs per burst. No difference was found in the number of APs per burst between GAERS and NECs (median ± standard deviation [SD], GAERS vs. NEC, 2.0 ± 1.03, n = 58 cells from 15 rats vs. 2.03 ± 1.2, n = 23 cells from five rats, p = 0.92). However, neurons within S2 of the GAERS showed more burst firing compared to those in the same region of NEC rats (median, 3.4% vs. 2.4%, p = 0.03). The amount of burst firing of neurons in the adjacent S1 or IC was not different between GAERS and NEC rats.

Layer VI neurons in the somatosensory cortex play a leading role during seizures

To investigate the cellular correlates of the oscillations that occurred in S2 before the neighboring cortices during preictal periods, single-cell juxtacellular recordings of the cortical layers V and VI were performed in S1, S2, and IC (n = 108 cells from 15 GAERS). No significant difference was found between the onsets of rhythmic firing between neurons recorded in S2 versus those in IC. During SWDs, all recorded cells fired rhythmically and synchronously with each surface EEG spike. S2 neurons appeared to fire before S1 and IC neurons throughout the seizures (Figs. 6 and S2, −39.5 ± 2.5 msec, vs. S1, −22.2 ± 2.6 msec and IC, 29.8 ± 3.8 msec compared to the EEG spike). When comparing time latencies between neurons of layers V and VI for both S1 and S2 (Table S2), layer VI cells were found to fire significantly ahead of layer V cells in S1 during SWDs (layer VI, −31.3 ± 3.7 msec, vs. layer V, −11.1 ± 5.12 msec, p < 0.01). However, there was no significant time lag between layers V and VI in S2 (layer V, −40.1 ± 2.5 msec vs. layer VI, −37.6 ± 4.6 msec, p > 0.05). No significant difference was found between timing of the firing of layer VI cells of S1 and S2 during SWDs (Table S2). There was no difference in the firing rate or percentage burst firing between the regions or layers.

Figure 6.

(A) S2 leads S1 neuronal firing during seizures. Juxtacellular recording of single cells from layer V of S2 and S1 during seizures show that S2 cells consistently fire before the S1 cells, both of which precedes the downward spike of the SW complex on the EEG. (B) Dot plot showing the temporal relationship between cells in layers V and VI of the S1 and S2 cortices with reference to the ictal EEG spike (Time zero) during seizures. S1 layer V cells show delayed firing compared to other regions (*Kruskall-Wallis test, p < 0.0001).


The main observations of the present study are the following: (1) In GAERS, SWDs arose from brief (<2 s) 5–9 Hz oscillations, which were first observed in S2 and IC (they were also recorded in NEC rats but did not trigger SWDs); (2) self-sustained SWDs could be triggered following a short-lasting train of 7-Hz electrical stimuli in GAERS, but not in NEC rats, during periods of quiet wakefulness, with a lower stimulus intensity in S2/IC than in S1; (3) a population of neurons focused around S2/IC, extending into the adjacent S1 region, showed sustained rhythmic firing during interictal periods and physiologic 5–9 Hz oscillations; (4) interictally, neurons in the deep layers of S2 showed more burst firing in GAERS than in NEC rats; (5) during SWDs, HFO occurred in S2/IC before S1; and (6) during SWDs, S1 and S2 layer VI and S2 layer V cells fired synchronously, followed by S1 layer V cells. Together, these results demonstrate the presence of precursor cellular and network rhythmic activities in S2/IC cortices before the S1 in the lead up to SWD occurrence, and indicate that the S2/IC regions form part of the epileptogenic circuit from which seizures are initiated.

Precursor oscillations in S2/IC prior to the onset of seizures in GAERS

By linearly comparing the neural dynamics of four cortical regions (M1, S1, S2, and IC), we found that precursor oscillatory network and cellular rhythmic activities occurs first in S2 and IC relative to S1 during preictal periods. Previous studies showed that SWDs in GAERS arise from oscillations of 5–9 Hz (Pinault et al., 2001; Polack et al., 2007), but the involvement of the S2 and IC regions had not been investigated. This rhythm may be comparable to the increase in theta rhythm (4.5–8 Hz) during the 1 s epoch before SWD onset in WAG/Rij rats (Sitnikova & van Luijtelaar, 2009). Similar brief oscillations can also be found in NEC rats, suggesting that although they are required they are not in themselves sufficient to generate SWDs. In an epileptic animal, they represent a necessary element of the seizure trigger without which seizures cannot occur, but it is clear that other factors are also required for SWD initiation in GAERS. This is consistent with the finding that short bursts of 7-Hz electrical stimulation were able to trigger self-sustained SWDs in GAERS but not in NEC rats. Such factors may include molecular and cellular changes, such as increased low-threshold calcium channel expression and function (Powell et al., 2009), which make the thalamocortical circuitry hyper-resonant to the precursor oscillations, thus generating pathologic SWDs.

A subpopulation of preictally rhythmic neurons found in layers IV to VI of S1ULp, S2, and granular IC is the likely cellular basis of the preictal rhythmic oscillations. These cells sustain the 5–9 Hz rhythm between SWDs before the adjacent cells are recruited to firing at the same frequency during SWDs. The degree of rhythmicity of the neurons could depend on the strength of neuronal synchronization within the thalamocortical circuit. SWDs occur when the nonrhythmically firing cells are recruited to fire in synchrony with the rhythmic cells at 5–9 Hz, likely via direct layer VI corticocortical connections (Zhang & Deschenes, 1998).

In agreement with previous studies that identified rhythmic electrical events in nonepileptic rats (Buzsaki et al., 1990; Pinault et al., 2001), the preictally rhythmically firing cells are present in similar proportions in GAERS and NEC rats. However, we also recorded increased burst firing in S2/IC cortex in GAERS, which may increase the likelihood of seizure occurrence. Several possible mechanisms exist for this apparent cortical hyperexcitability. These include the gain-of-function mutation in the low-threshold calcium channel, Cav3.2 (Powell et al., 2009); the increase in expression of the transmembrane 2-amino-3-(3-hydroxy-5-methyl-isoxazol-4-yl) propanoic acid (AMPA) receptor regulatory proteins (TARPs) stargazin (Powell et al., 2008; Kennard et al., 2011); an increase in Nav1.1 and Nav1.6 during epileptogenesis (Klein et al., 2004), as well a decrease in Ih (Schridde et al., 2006). An N-methyl-d-aspartate (NMDA) receptor hyperfunction reported in both GAERS and WAG/Rij compared to the control rats (Pumain et al., 1992; D’Antuono et al., 2006) within the S2/IC region could also increase the likelihood of switching cortical activity from nonrhythmic state to the more proepileptogenic, rhythmic state.

Intracortical electrical stimuli trigger typical SWDs in GAERS

Electrical stimuli in S1, S2, or IC could reliably trigger typical SWDs in GAERS. Our results are in agreement with a recent study in the WAG/Rij model, where SWD-like afterdischarges could be elicited with cortical stimulations, more readily in the somatosensory cortex than the motor cortex (Luttjohann et al., 2011). The level of the current required for the stimulations to trigger seizures was significantly lower in S2 than in S1, suggesting that the S2/IC stimulation sites are more epileptogenic than the S1 site. The elements that could be activated are primarily the axon branches and axon initial segments surrounding the electrode tip (Nowak & Bullier, 1998). However, the same stimulation may activate orthodromically efferent axons from the S2 to distant postsynaptic intracortical regions. Based on this experiment alone, the involvement of a distant afferent cortical or subcortical structure or a region whose axons of its neurons pass through S2 or IC for SWD initiation cannot be ruled out.

Both S1 and S2 are reciprocally connected and receive topographically organized inputs from the ventral posteromedial thalamic nucleus (VPM) and ventral posterolateral thalamus (VPL) (Liao & Yen, 2008). Cells within layer VI of S1 and S2 are known to project to the VPM and the TRN (Allen et al., 1991; Bokor et al., 2008) and are thought to initiate SWDs at least in the primary somatosensory system (Pinault, 2003). The IC receives visceral, gustatory, and somatosensory inputs and is reciprocally connected with dorsal thalamus (Allen et al., 1991). Although the IC, S1, and S2 are anatomically interconnected (Saper, 1982), their anatomofunctional properties and roles in SWD generations remain unclear. The intracortical spread and synchronization of the 5–9 Hz rhythmic events from S2/IC may not simply follow the rule of synaptic circuits within the intracortical circuitry, as rhythmic activity may be sustained within the S2/IC for minutes without spreading. The initiation of SWDs may involve a corticothalamic-induced resonant phenomenon in the thalamus (Pinault, 2003) subsequent to S2/IC rhythmic events. Such corticothalamic resonance would then recruit S1 layer V neurons, which lag behind by >20 msec during SWDs. Therefore, a mass synchronization processes involving both intracortical spread and thalamic feedback must be present to enable the transformation of the physiologic somatosensory rhythm into pathologic SWDs.

HFO occurs earlier in S2/IC than S1 during SWDs

We recorded rhythmic, short-lasting HFOs in deep cortical layers of S1, S2, and IC prior to the each spike during a seizure, with those in the IC and S2 consistently leading those in S1 by >20 msec. This time lag is remarkably long, since the paired recorded sites were only 1–3 mm apart, suggesting a synchronization mechanism other than direct intracortical recruitment. HFOs in S1 could be a result of direct volume-conducted fast activity from S2/IC or recruited by the thalamus as part of secondary synchronization mechanism. Further experiments are required to determine the cellular and network nature of these absence-related HFOs and whether they are either intrinsically generated in the cortex, or the consequence of massive hypersynchronized synaptic inputs originating in the cortex or in subcortical structures. HFOs may arise from tight electrical coupling between neuronal populations via gap junctions to trigger synchronized network activity (Draguhn et al., 1998; Traub et al., 1999) and may be analogous to somatosensory fast oscillations (Barth, 2003) or hippocampal ripples (Buzsaki et al., 1992). We also reported HFOs in association with the SWDs in freely moving, nonanesthetized GAERS, which had the same basic characteristics of those recorded in animals under neuroleptic anesthesia. To our knowledge this is the first report of HFOs occurring with absence seizures, an observation that warrants further in depth investigation.

A model for SWD initiation

On the basis of current and previous findings, we propose a model for the sequence by which physiologic somatosensory oscillations develop into absence-like SWDs: (1) S2 and IC cortical areas form part of a critical circuit where precursor rhythm cells initiate and sustain the physiologic 5–9 Hz preictal oscillations, (2) excitatory propagation spreads from the S2/IC to the interconnected S1, motor, and frontal cortical regions, possibly via a caudorostral excitatory pathway (Fujita et al., 2010), which could also involved thalamic nuclei (Gorji et al., 2011), (3) the anatomic and functional properties of S1 and S2 layer VI cells predict that they will trigger synchronous barrages of excitatory postsynaptic potentials (EPSPs) in the related thalamic relay and caudal part of the reticular nucleus (Aker et al., 2006), thereby leading to hypersynchronized resonant neuronal firing in an epileptic thalamocortical network.


None of the authors has any conflict of interest or financial interests to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.