Abstract

Electrophysiological oscillations are thought to create temporal windows of communication between brain regions. We show here that human cortical slices maintained in vitro can generate oscillations similar to those observed in vivo. We have characterized these oscillations using local field potential and whole-cell recordings obtained from neocortical slices acquired during epilepsy surgery. We confirmed that such neocortical slices maintain the necessary cellular and circuitry components, and in particular inhibitory mechanisms, to manifest oscillatory activity when exposed to glutamatergic and cholinergic agonists. The generation of oscillations was dependent on intact synaptic activity and muscarinic receptors. Such oscillations differed in electrographic and pharmacological properties from epileptiform activity. Two types of activity, theta oscillations and high gamma activity, uniquely characterized this model—activity not typically observed in animal cortical slices. We observed theta oscillations to be synchronous across cortical laminae suggesting a novel role of theta as a substrate for interlaminar communication. As well, we observed cross-frequency coupling (CFC) between theta phase and high gamma amplitude similar to that observed in vivo. The high gamma “bursts” generated by such CFC varied in their frequency content, suggesting that this variability may underlie the broadband nature of high gamma activity.

Introduction

Changes in amplitude and synchrony of electrophysiological oscillations measured from the brain accompany a myriad of cognitive, vegetative, and pathological states (Niedermeyer and Lopez da Silva 1999; Buzsaki 2006). The modulation of such oscillations with behavior suggests that they may be a marker/substrate for both local processing, and communication between spatially dispersed neuronal assemblies (Singer and Gray 1995; Fries 2005; Buzsaki 2011; Fell and Axmacher 2011; Siegel et al. 2012). Such “rhythms of the brain” (Buzsaki 2006) thus represent a potential window through which to understand information processing and neuronal communication, while also providing insights into potential pathological mechanisms (Beenhakker and Huguenard 2009). In vitro animal studies have contributed greatly to the understanding of the cellular and microcircuit properties that underlie such rhythms. In these studies, rhythms with similar frequencies to those observed in vivo have been induced through electrical stimulation (Traub et al. 1996; Whittington et al. 1997), or activation of cellular elements through application of agonists for glutamatergic and/or cholinergic receptors (Whittington et al. 1995; Buhl et al. 1998; Fisahn et al. 1998, 2002; Blatow et al. 2003; Oke et al. 2010). Such experiments have revealed that the generation of rhythmic activity likely involves an interplay between excitatory and inhibitory neurotransmission (Blatow et al. 2003; Fisahn 2005; Hasenstaub et al. 2005; Bartos et al. 2007; Szabadics et al. 2007; Yamawaki et al. 2008; van Aerde et al. 2009; Isaacson and Scanziani 2011), may require electrotonic communication (Buzsaki 2002; Hormuzdi et al. 2001; Roopun et al. 2006), and to some extent may reflect the resonant properties of individual cells (Pike et al. 2000). Currently, our best approximations of the cellular, microcircuit, and pharmacological profiles that generate rhythmic activity in the human brain come from animal studies—there are no equivalent human studies. Therefore, although very elegant work exists in human cortical tissue, a dynamic perspective, notwithstanding induced epileptiform activity (Kohling and Avoli 2006; Huberfeld et al. 2011), remains wanting.

It has been shown that cortical oscillations/activity ranging from theta to high gamma can be generated locally within the human cortex (Crone et al. 1998; Kahana et al. 2001; Raghavachari et al. 2001, 2006; Tallon-Baudry et al. 2001; Canolty et al. 2006). We therefore surmised that the requisite cortical circuitry exists within human neocortical slices to generate oscillations/activity similar to those observed physiologically, and that this activity is distinct from the pathological oscillations manifesting as epileptiform activity. We addressed this question in temporal neocortical slices of the middle temporal gyrus (MTG) that were obtained during surgery from subjects with mesial temporal lobe epilepsy (mTLE) (Valiante 2009). Our approach was based on the following considerations. First, these subjects form the majority of resective surgeries performed at our institution, affording a relatively large and stereotypical source of brain tissue. Second, the MTG is routinely removed during surgery for mTLE and thus no modification to surgical technique is required to obtain this tissue (Valiante 2009; Mansouri et al. 2012). Third, such temporal neocortical tissue has been shown to display relatively normal histological (Marco et al. 1996) and electrophysiological characteristics (Foehring et al. 1991). Fourth, the epileptogenic onset zone is within the mesial temporal structures, far removed from the temporal neocortex (Valiante 2009; Mansouri et al. 2012). Lastly, the temporal neocortex is implicated in various cognitive tasks (Penfield and Rasmussen 1950; Ojemann et al. 1989) and has been shown to be activated during performance of such tasks (Canolty et al. 2007; Edwards et al. 2010; Pei et al. 2011). Indeed, resection of the temporal neocortex ipsilateral to a sclerotic hippocampus can result in naming deficits in the dominant hemisphere, suggesting preservation of neocortical function despite its proximity to the epileptogenic zone (Jones-Gotman 1993).

We report here that human brain tissue maintained in vitro can generate seemingly physiological and pathological oscillatory activity when exposed to glutamatergic and cholinergic agonists. Our data indicate that human temporal neocortical slices maintain the local circuitry necessary for mimicking oscillations observed in vivo and display activities that have not been readily observed in in vitro animal models. We show that these slices demonstrate robust theta oscillations that are distinct from epileptiform activity, and appear to coordinate activity across deep and superficial cortical layers. These oscillations are shown to be locally associated with gamma oscillations through cross-frequency coupling (CFC). We exploit this theta-high gamma CFC to investigate the nature of broadband high gamma activity observed here, and suggest it may arise from the temporal average of variable frequency high gamma bursts.

Materials and Methods

Subjects

Written informed consent was obtained for all subjects in this study as part of a research protocol approved by the University Health Network Research Ethics Board in accordance with the Declaration of Helsinki. All subjects (Table 1) had medically refractory mTLE. Subjects had been deemed to be candidates for temporal lobectomy through a standard presurgical assessment by the multidisciplinary Epilepsy Program at the Toronto Western Hospital (Mansouri et al. 2012).

Table 1

Demographic data

Age (years) Gender MRI finding Pathology Sz duration (years) Sz onset age Medications 
25 Male Right MTS MTS 20 Carbamazepine, Topiramate 
35 Female Right MTS MTS 28 Levetiracetam, Oxcarbazepine 
32 Female Left MTS MTS 27 Divalproex, Lamotrigine, Clonazepam 
48 Female Left MTS MTS 18 30 Phenytoin, Valproate, Clonazepam 
35 Female Left MTS MTS 10 25 Levetiracetam, Lamotrigine 
34 Female Left MTS Normal hippocampus 14 20 Levetiracetam, Lamotrigine 
60 Female Right MTS MTS 20 40 Carbamazepine, Gabapentin 
25 Male Left MTS MTS 22 Clobazam, Lamotrigine 
65 Male Right MTS MTS 35 30 divalproex, Primadone, Levetiracetam 
39 Female mTL lesion Ganglioglioma 19 20 Carbamazepine 
33 Female Left MTS MTS 30 Tegretol, Divalproex 
27 Male Left MTS MTS 25 Clobazam, Lamotrigine 
52 Female Left MTS MTS 38 14 Phenytoin, Valproate 
36 Female Right MTS MTS 30 Carbamazepine, Lamotrigine 
42 Female Right MTS MTS 40 Levetiracetam, Lamotrigine 
49 Male Left MTS MTS 19 30 Lamotrigine, Clobazam 
46 Male Left MTS MTS 38 Carbamazepine, Lamotrigine, Clobazam 
21 Female Left MTS MTS 16 Levetiracetam, Carbamazepine, Clobazam 
23 Male Normal studies MTS 12 11 Lamotrigine, Topiramate, 
27 Male Left MTS MTS 20 Lamotrigine 
55 Female Left MTS MTS 30 25 Levetiracetam, Carabmazepine, Lamotrigine 
39 Male Left mts MTS 32 Lamotrigine, Topamax 
45 Female Right MTS MTS 38 Clobazam, Carbamazepine 
Age (years) Gender MRI finding Pathology Sz duration (years) Sz onset age Medications 
25 Male Right MTS MTS 20 Carbamazepine, Topiramate 
35 Female Right MTS MTS 28 Levetiracetam, Oxcarbazepine 
32 Female Left MTS MTS 27 Divalproex, Lamotrigine, Clonazepam 
48 Female Left MTS MTS 18 30 Phenytoin, Valproate, Clonazepam 
35 Female Left MTS MTS 10 25 Levetiracetam, Lamotrigine 
34 Female Left MTS Normal hippocampus 14 20 Levetiracetam, Lamotrigine 
60 Female Right MTS MTS 20 40 Carbamazepine, Gabapentin 
25 Male Left MTS MTS 22 Clobazam, Lamotrigine 
65 Male Right MTS MTS 35 30 divalproex, Primadone, Levetiracetam 
39 Female mTL lesion Ganglioglioma 19 20 Carbamazepine 
33 Female Left MTS MTS 30 Tegretol, Divalproex 
27 Male Left MTS MTS 25 Clobazam, Lamotrigine 
52 Female Left MTS MTS 38 14 Phenytoin, Valproate 
36 Female Right MTS MTS 30 Carbamazepine, Lamotrigine 
42 Female Right MTS MTS 40 Levetiracetam, Lamotrigine 
49 Male Left MTS MTS 19 30 Lamotrigine, Clobazam 
46 Male Left MTS MTS 38 Carbamazepine, Lamotrigine, Clobazam 
21 Female Left MTS MTS 16 Levetiracetam, Carbamazepine, Clobazam 
23 Male Normal studies MTS 12 11 Lamotrigine, Topiramate, 
27 Male Left MTS MTS 20 Lamotrigine 
55 Female Left MTS MTS 30 25 Levetiracetam, Carabmazepine, Lamotrigine 
39 Male Left mts MTS 32 Lamotrigine, Topamax 
45 Female Right MTS MTS 38 Clobazam, Carbamazepine 

Tissue Acquisition and Preparation

A standard anterior temporal lobectomy (Pilcher and Ojemann 1993; Valiante 2009) was performed in 23 patients (ages 25–65) under general anaesthesia using volatile anaesthetics. Subjects ingested their usual anticonvulsants with sips of water prior to surgery. Temporal lobectomy involved resection of the first 4.5 cm of temporal neocortex followed by the resection of the mesial temporal structures. Prior to resection of the temporal neocortex, 1 cm3 of brain was removed from the MTG ∼3 cm from the temporal pole using sharp dissection and local cooling with ∼4 °C TissueSol®. The tissue block was then immediately immersed in a cold (∼4 °C) precarbogenated (95% O2 and 5% CO2) dissection solution (in mM: sucrose 248, KCl 2, MgSO4 3, CaCl2 1, NaHCO3 26, NaH2PO4 1.25, D-glucose 10, and kyneurenic acid 1). The total time including transportation and slicing was kept to a maximum of 20 min (Kohling and Avoli 2006). Using a vibrotome (Leica 1200 V), 500 μm-thick brain slices were sectioned, with an orientation perpendicular to the pial surface (Kostopoulos et al. 1989), After sectioning, the slices were stabilized in standard artificial cerebrospinal fluid (ACSF) (in mM: NaCl 123, KCl 4, CaCl2 1, MgSO4 1, NaHCO3 25, NaH2PO4 1.2, and D-glucose 10), pH 7.40, at 35 °C for 30 min, with the addition of 2 mM kynurenic acid (a general antagonist for ionotropic glutamate receptors (Stone 2001)) which has been shown to decrease anoxic injury in hippocampal slices (Clark and Rothman 1987). Following this incubation, slices were rinsed twice and maintained in standard ACSF at 22–23 °C for at least 1 h before recording. Stabilization of brain slices in warmed ACSF has been shown to increase preservation of neuronal processes (Bischofberger et al. 2006) and improve intrinsic network activity in hippocampal slices of aged mice (El-Hayek et al. 2013).

Electrophysiology

Recordings were obtained in a submerged chamber in which the ACSF was perfused at a high rate (15 mL/min), over both the top and bottom surfaces of the slice (Wu et al. 2002, Wu et al. 2005). In addition, humidified gas of 95%O2–5%CO2 was allowed to pass over the perfusate to increase local oxygen tension. Previous studies (Wu et al. 2002; Wu et al. 2005; Zhang et al. 2008; Hajos and Mody 2009; Hajos et al. 2009) have shown that fast, top, and bottom perfusion of the slice is important for maintaining in vitro spontaneous population activity under submerged conditions. The recording temperature was set at 36 °C and continuously monitored for the duration of the experiments. Using glass electrodes (thin-wall borosilicate tubes with filaments, World Precision Instruments, Sarasota, FL, USA) filled with a solution containing 150 mM NaCl or the standard ACSF (tip resistance of 1–2 MΩ), local field potentials (LFPs) were recorded simultaneously in superficial (layers II/III, 200–700 μm from pial surface) and deep (layers V–VI, 2000–3500 μm from pial surface) layers (Kohling and Avoli 2006). Electrical stimulation was performed by injecting isolated current pulses of 10–150 µA generated by a Grass stimulator (S88, Grass Instruments, Quincy, MA, USA) through a stainless steel electrode (125 μm diameter), 300–500 μm away from the superficial or deep layer recording site. Cell attached and whole-cell patch recordings from putative interneurons and pyramidal neurons were performed in the superficial cortical layers. In the majority of experiments, cells for such recordings were visualized under an Olympus BX51WI upright microscope (Olympus Optical Co., NY, USA) equipped with a ×40 water immersion lens with differential interference contrast and infrared optics. For patch-clamp experiments, glass electrodes were filled with a solution containing (in mM) K-gluconate 135, NaCl 10, MgCl2 1, Na2ATP 2, GTP 0.3, HEPES 10 (pH balanced to 7.4 and tip resistance of 5–6 MΩ; as per Zhang et al. 2012). Input resistance was measured from voltage responses to hyperpolarizing current pulses (25–50 pA and 0.5–1 s). Only cells with a stable resting membrane potential <−50 mV were included in data analysis. Relative isolation of inhibitory postsynaptic currents (IPSCs) and excitatory postsynaptic currents (EPSCs) was done in V-clamp mode while holding the membrane potential at 0 and −70, respectively. The reversal potential of the synaptic activity accompanying spontaneous slow rhythms was calculated as described elsewhere (Zhang et al. 2012). Signal acquisition, storage, and part of the data analyses were done using a package from Axon Instruments/Molecular Devices Corporation (Sunnyvale, CA, USA), including amplifiers (Multiclamp 700A or B amplifiers), digitizers (Digidata 1322A), and PClamp software (version 10.2). Sample acquisition rate was 10 KHz with low-pass filtering at 5 KHz. Pharmacology and drug application: (2S,3S,4S)-3-(Carboxymethyl)-4-prop-1-en-2-ylpyrrolidine-2-carboxylic acid (kainic acid), 2-{(aminocarbonyl)oxy}-N,N,N-trimethylethanaminium chloride (carbachol) and atropine were obtained from Sigma-Aldrich and prepared in distilled water as stock solutions and directly diluted into the perfusate to the concentrations indicated in results (Buhl et al. 1998). 5,5-Diphenylimidazolidine-2,4-dione (Phenytoin) was obtained as clinically used injectable vials (Sandoz Canadian, Inc.) and directly diluted in ACSF to the desired concentration. Statistical tests (t-test, and 1-way ANOVA) were performed using Sigmaplot (version 11.1), where statistical significance was considered at P values <0.05. Results are presented as mean ± standard error of mean (SEM) throughout the text.

Histology

Formalin-fixed paraffin-embedded brain sections (5 µm thickness) were obtained from control postmortem samples (n = 5) with no prior history of epilepsy, and tissues resected from mTLE patients (n = 15). The sections were treated with a standard procedure (DeFelipe et al. 1993) and incubated overnight with mouse anti-Parvalbumin antibody (Sigma, mouse monoclonal, Clone PARV-19, P 3088) at 1/1000 dilution. Staining was done using a Super-Sensitive Polymer HRP kit (Biogenex, BGQD440-XAK) as instructed by the vendor. Adjacent sections were also used for hematoxylin staining via a standard protocol. Parvalbumin staining was assessed by 2 different methods. The first methods employed an “image scope pixel count.” Slides were scanned with an Aperio slide scanner at ×20, and Image Scope software for image analysis was used to outline the cortex. We ran a pixel count computing the number of immunopositive pixels over the total number of pixels—this was repeated on 5 sections from the same case and the results were expressed as the mean of immunopositive pixels/total number of pixels ± SD. This method thus reports a unit-less ratio of immunopositive pixels to total pixels. The second method employed outlining 5 randomly selected 0.5 × 0.5 mm areas. The outlined areas included mainly cerebral cortex (all layers while excluding the underlying white matter) and, within each area, the number of parvalbumin–positive (PV+) cells were counted under the microscope.

Data Analysis

All analyses, unless otherwise stated were performed with in house scripts developed in MatLab (MathWorks, Natick, MA, USA).

Power Spectrum

Data were decimated to 2.5 kHz, after first notch filtering at 60 Hz (and all harmonics <10 kHz) and low-pass filtering at 400 Hz. All filters were 10 000 order finite impulse response (FIR) filters that were run forward and in the reverse direction (filtfilt.m) and thus without phase distortions. Thirty seconds of representative data were selected for analyses. Power spectra computed using a multitaper method (pmtm.m) with 2 tapers, over 1 s intervals, obtained 30 spectra per condition. Within-slice statistics were computed between the various pharmacological manipulations and baseline utilizing the Mann–Whitney U(MWU) test at each frequency. This obtained a P value for each frequency that was then corrected for multiple comparisons utilizing false discovery rate (FDR) correction (Benjamini and Hochberg 1995), with α = 0.05. For FDR corrections, significant Pvalues fulfilled the following criterion:  

(1)
p(i)<αiNc
where p(i) are the sorted Pvalues in ascending order and c=j=1i1/j where iN, N being the number of observations, and α the chosen level of significance (Bonferroni corrected for the number of pairwise comparisons between conditions). Each slice was analyzed separately in this way. If there was any significant increase in power for either kainate, and/or kainate + carbachol conditions the slice was included in the group statistic calculations. “Group statistics” were computed similar to individual slice statistics, however prior to admitting the power spectra to statistical testing they were normalized by the baseline for that slice. The normalized spectra for each condition and for all slices were then pooled, and significance obtained as above. For both individual slice and group statistics, a frequency range was only considered significant if it spanned 5 Hz or more (Gaona et al. 2011). This condition was relaxed for frequencies below 30 Hz, as the modal frequency ranges are narrower at lower frequencies. Power spectra are displayed normalized to the baseline condition (aka “Normalized Power Spectrum”).

Interlaminar Synchronization

Electrophysiological signals were decomposed in time and frequency by convolution with a complex Morlet wavelet W (Bruns 2004):  

(2)
S(t,a)=x(τ)1aW(τta)dτwhere W(t;fB,fC)=1πfBe2iπfCtet2/fB
where a is the scale of the wavelet analysis, determining the effective frequency, fB = 5 is the frequency bandwidth or wavenumber, and fC = 1 is the center frequency. Such decomposition yields a complex time and effective frequency series S(t,a) from which phase (φ) and power can be extracted.

Phase Coherence

Phase coherence (PC), was used to investigate the synchrony between cortical layers. PC is an amplitude-free measure of the association of the phases between 2 given signals at a given frequency. The phase of a periodic signal is defined as the relative location of the signal upon the period of its underlying oscillation. It can be calculated for any frequency component (f) and at any point in time (t):  

(3)
ϕ(t,f)=tan1((x(t,f))(x(t,f)))
where ()and() denote the imaginary and real parts, respectively.

The PC at time t, a given frequency f, and a window size δ is defined as:  

(4)
PC(f,t)=1δ|τ=tδ/2t+δ/2ei[ϕ1(τ,f)ϕ2(τ,f)]|
where ϕ1(τ,f)ϕ2(τ,f) is the phase difference between the cortical layers at a given point in time–frequency space. PC measures the degree to which the phase relationships between 2 signals remain constant throughout the window δ. It varies on a scale of 0–1 where 0 indicates no coherence and 1 indicates perfect coherence (Lachaux et al. 2000).

We utilized bootstrap statistics (Davison and Hinkley 1997) to determine significance for PC analyses. Surrogates were created by randomizing the phases of the signal. This was accomplished by randomizing the phases of the fast Fourier transform of one of the signals, and performing the inverse fast Fourier transform to transform back into the time domain (Theiler et al. 1992). PC was then applied to all surrogates. One thousand surrogates were created, and the P value was then determined as the number of times the original PC was greater than a surrogate value PCsurr divided by the total number of surrogates (nsurr) +1. This was repeated for PC being less than the surrogate values for the equivalent of a 2-tailed test:  

(5)
p1=i=1nsurrI{PC>PCsurr,i}nsurr+1,p2=i=1nsurrI{PC<PCsurr,i}nsurr+1
where I{cond} is the indicator function for the condition cond, being 0 when cond, is false and 1 when cond is true. If the PC at a given frequency and time was determined significant with α = 0.05 for either p1or p2, then it was deemed to have passed the surrogate test. This test is equivalent to testing the null hypothesis H0, that the PC measured was due only to a random relationship between 2 signals statistically similar to the original signals but with random phase associations.

Significance of the PC trends was further tested when compared with PC during baseline. Thirty seconds of each treatment condition (Baseline, Kainate, and Kainate and Carbachol) was used. PC analysis was averaged across slices considered previously for inclusion in group power statistics. As with the power spectra, MWU test with FDR correction were used to compare the distribution of PC values for each frequency in all slices in comparison to baseline recordings. An α of 0.05 was used for all coherence statistics.

Wavelet Coherence

To compare intracellular and extracellular recordings, wavelet coherence (WC) was used. This method was chosen since it incorporates both phase and amplitude, and intracellular recordings are likely to contain information on all frequencies in the extracellular fields both through field effects as well as synaptic mechanisms. Low amplitude field effects may corrupt the analysis of a statistic based purely on phase. WC is defined as the wavelet cross-spectrum (Sxy) normalized by the individual spectra (Sxx and Syy) (Lachaux et al. 2000):  

(6)
WC(t,f)=|Sxy(t,f)|Sxx(t,f)Syy(t,f)
 
(7)
Sij=tδ/2t+δ/2Si(τ,f)Sj(τ,f)dτ
where the individual wavelet spectra Si have been defined previously (see above). To determine the significance of the WC results, MWU test with FDR correction were used. The WC matrices were divided into 1 s bins and averaged across those bins. The resultant distribution of WC values for each frequency was compared with baseline recordings.

Phase Statistics

To further explore the phase relationships for the PC results around the peak value of PC, we obtained phase distributions across all slices for a given treatment. After band pass filtering the signals around the frequencies which showed the largest PC values (7–10 Hz), the Hilbert transform was applied in order to calculate the instantaneous phase. The Hilbert transform takes a signal that is assumed to have only one dominant frequency and transforms it into a complex signal from which phase can be determined as in equation (1). The Hilbert transform is defined as:  

(8)
h[x(t)]=x(τ)1π(tτ)dτ

This was repeated for signals in both deep and superficial layers after which a phase difference ϕ1(τ,f)ϕ2(τ,f) was obtained. A histogram of phase differences was then built and a von Mises distribution was fit to the resultant histogram. The von Mises distribution is defined at a given phase (φ) by 2 parameters: a mean (μ) and a concentration (k):  

(9)
P(ϕ;μ,κ)=12πIo(κ)e[κcos(ϕμ)]
where Io(κ) is a Bessel function of the first kind of zeroth order. The Rayleigh test was used to determine whether the given histograms were significantly different from a uniform distribution, assuming that they originated from a unimodal von Mises distribution. All were deemed significant with P < 0.05. Resultant means of the circular distributions were then compiled into a histogram for each condition and frequency band. Maximum likelihood estimates of the von Mises distribution parameters and the Rayleigh test were performed using the circular statistics toolbox written by Philipp Berens (Berens 2009).

Cross-Frequency Coupling

CFC was estimated using the phase of low-frequency oscillations (2–30 Hz in 1 Hz steps) to index the high-frequency (60–200 Hz by 2 Hz steps) amplitude (Tort et al. 2010). The phase of the low-frequency oscillation and the amplitude of the high-frequency activity were obtained from wavelet transformation of the time series as described above. The phase interval between –π and π was then divided into 12 equal bins. Within these bins, the phases of the low-frequency oscillation were used to bin the associated high-frequency amplitudes. The modulation index was computed from this distribution as the Kullback–Leibler distance from a uniform distribution normalized by log10(12) (Tort et al. 2010). The preferred phase was obtained by fitting amplitude distribution to a von Mises distribution, and obtaining the angle at which the amplitude was maximal. For statistical analysis, 200 time-shifted surrogates were generated for each phase-amplitude frequency pair and the modulation index computed from these surrogates. P values were then computed, and FDR corrected with α = 0.05. To average across all slices, the MI was first masked by regions of significance, then normalized by the largest value for each slice, and lastly all slices were averaged together.

Spectrograms

Time series were decomposed using a complex Morlet wavelet (“cwt” function in MatLab) using fB = 5 (cmor5-1) (see above). The amplitude of this transform was obtained from its absolute value. At each frequency the mean and standard deviation of the amplitude was computed over the entire time series—usually the entire experiment. Using the mean and standard deviation so obtained, the Z-score was computed for each time point and frequency, and plotted as a function of time and frequency.

High Gamma “Burst” Analyses

Signals were filtered into 2 frequency ranges: theta (4–10 Hz) and high gamma (60–180 Hz) using a 10 000 order FIR filter, and the “filtfilt” MatLab function. The high gamma envelope (amplitude) was obtained from the absolute value of the Hilbert transform of the high gamma range filtered signal, as has been done previously (Canolty et al. 2006) when analyzing high-frequency oscillations. Local maxima were detected and constrained in minimal distance by the period of an 8-Hz theta oscillation (125 ms), the requirement of being larger than 2SD of the high gamma signal, and not <5 μV in amplitude. We will refer to these events as “high gamma bursts.” The maxima were then used to index in time, the time–frequency decomposition of high gamma obtained using an fB = 20. This wave number was used to improve the frequency resolution of the transformation. The time index was used to determine the frequency of the burst, while the amplitude of the burst was obtained from the high gamma amplitude envelope. Correlation between high gamma burst frequency, size, and incidence were performed using Spearman's rank correlation. The time index was also used to extract the phase of the theta cycle at which the burst occurred, which was obtained as the angle from the theta signal obtained above. Rayleigh's test of uniformity (Berens 2009) was used to determine if there was phase clustering.

Results

Characterization of Human Cortical Slices

Baseline Synaptic and Cell Activities

Prior to pharmacological activation, a number of baseline characterizations were performed to confirm the viability of human cortical slices. Electrical stimulation within the superficial layer generated evoked potentials of 0.53 ± 0.03 mV (n = 81) in amplitude. Paired-pulse paradigms revealed short-term plasticity within superficial cortical layers (n = 67), displaying both facilitation of 27 ± 3% (n = 47), and depression by 22 ± 4% (n = 20) (Fig. 1A). Single and paired stimuli did not elicit epileptiform discharges (Fig. 1A) in any of the slices tested, suggesting that these slices were not innately hyperexcitable as has been previously shown in drug naïve human neocortical tissues (Avoli and Williamson 1996).

Figure 1.

Baseline characterization of the human MTG brain slice preparation. (A) Schematic of the typical recording arrangement with electrodes in the superficial (layers II/III) and deep (V/VI) layers. The stimulating electrode (Stim) was positioned in the superficial layer to generate single or paired evoked potentials. The red arrows represent measures used to quantitate responses to paired-pulse stimuli. (B) Whole-cell recordings from putative pyramidal cells and interneurons, displayed typical spiking (regular spiking vs. fast spiking) and spike features for the respective cell types (see Table 2). (C) Voltage clamp recordings revealed both evoked IPSCs (left, red) and EPSCs (right, black) at their respective holding potentials (Vm). (D) Parvalbumin stained mesial temporal lobe from a control and subject with mTLE.

Figure 1.

Baseline characterization of the human MTG brain slice preparation. (A) Schematic of the typical recording arrangement with electrodes in the superficial (layers II/III) and deep (V/VI) layers. The stimulating electrode (Stim) was positioned in the superficial layer to generate single or paired evoked potentials. The red arrows represent measures used to quantitate responses to paired-pulse stimuli. (B) Whole-cell recordings from putative pyramidal cells and interneurons, displayed typical spiking (regular spiking vs. fast spiking) and spike features for the respective cell types (see Table 2). (C) Voltage clamp recordings revealed both evoked IPSCs (left, red) and EPSCs (right, black) at their respective holding potentials (Vm). (D) Parvalbumin stained mesial temporal lobe from a control and subject with mTLE.

The viability of our tissue was further characterized using patch-clamp recordings of superficial layer neurons. Based on electrophysiological characterizations, these recordings revealed the presence of 2 functionally different neuronal types; putative pyramidal cells (n = 49) and putative interneurons (n = 13) (Fig. 1B). The electrophysiological characteristics of these distinct cell types (Table 2) were similar to those described previously for human cortical cells (Avoli and Olivier 1989; McCormick 1989; Foehring et al. 1991; Menendez de la Prida et al. 2002; Molnar et al. 2008). Regular spiking cells were considered pyramidal in nature, displayed a wider action potential, and smaller spike after hyperpolarization (Fig. 1B) than putative interneurons. In current clamp mode we did not observe bursting type responses as a result of extracellular stimulation. To isolate EPSCs and IPSCs, cells were held in voltage clamp mode at −70 and 0 mV, respectively. Stimulation in superficial layers revealed the presence of both EPSCs (1070 ± 170 pA, range = 130–3100 pA, n = 22) and IPSCs (760 ± 90 pA, range = 190–1400 pA, n = 18) (Fig. 1C) as has been previously observed in human cortex (Avoli and Olivier 1989; McCormick 1989; Avoli et al. 1991; Menendez de la Prida et al. 2002; Szabadics et al. 2006).

Table 2

Biophysical properties of putative pyramidal cells and interneurons

Parameter Putative pyramidal mean ± SEM (nPutative interneuron mean ± SD (n
RMP (mV) −69 ± 1 (48) −66 ± 1 (12) 
Input resistance (MΩ) 75.9 ± 4 (49) 100 ± 13 (12)* 
AP amplitude (mV) 73.8 ± 1 (49) 93.6 ± 14 (12)* 
AP half-width (ms) 1.15 ± 0.4 (49) 0.56 ± 0.2 (13)* 
AP decay (mV/ms) 68 ± 3 (48) 144 ± 8 (13)* 
f/I curve slope (spikes/pA) 0.05 ± 0.004 (49) 0.131 ± 0.02 (13)* 
Bursting cells 24/49 4/13 
Spike frequency adaptation 37/49 8/13 
Adaptation ratio 0.63 ± 0.03 (37) 0.57 ± 0.04 (7) 
Tonic firing 12/49 5/13 
Parameter Putative pyramidal mean ± SEM (nPutative interneuron mean ± SD (n
RMP (mV) −69 ± 1 (48) −66 ± 1 (12) 
Input resistance (MΩ) 75.9 ± 4 (49) 100 ± 13 (12)* 
AP amplitude (mV) 73.8 ± 1 (49) 93.6 ± 14 (12)* 
AP half-width (ms) 1.15 ± 0.4 (49) 0.56 ± 0.2 (13)* 
AP decay (mV/ms) 68 ± 3 (48) 144 ± 8 (13)* 
f/I curve slope (spikes/pA) 0.05 ± 0.004 (49) 0.131 ± 0.02 (13)* 
Bursting cells 24/49 4/13 
Spike frequency adaptation 37/49 8/13 
Adaptation ratio 0.63 ± 0.03 (37) 0.57 ± 0.04 (7) 
Tonic firing 12/49 5/13 

AP is action potential. The f/I curve is the firing frequency as a function of injected current. Spike frequency adaptation indicates the number of cells that showed apparent slowing during a 500-ms depolarizing constant current injection of 50 pA steps (e.g., Fig. 1B where both cells show this). This is quantified by the adaptation ratio of the first interspike interval over the last interspike interval elicited by the constant current pulse. Tonic firing means continuous firing throughout the current pulse.

Parvalbumin Immunoreactivity is Not Decreased in MTG Slices

Inhibitory neurotransmission is an important contributor to cortical oscillations (Whittington et al. 2000; Isaacson and Scanziani 2011), a major component of which are PV+ interneurons (Ascoli et al. 2008). Relative preservation of PV+ interneurons, as well as histological normalcy of analogous tissue from patients with epilepsy, has been previously reported (Babb et al. 1984; Marco et al. 1996). To determine if this holds true in our experiments, we quantified the relative amount of PV immunoreactivity in a subset of MTG tissues from which electrophysiological recordings were obtained (n = 15 patients) and compared these with corresponding tissues of 5 postmortem controls with no prior history of seizures. No significant difference in the ratio of PV positive pixels/total pixels was found between the 2 groups of MTG tissue (mTLE = 0.23 ± 0.01, control = 0.23 ± 0.02, P = 0.9) (Fig. 1D). Furthermore, by the second method of quantification, the number of PV+ cells per 0.5 × 0.5 mm region was not significantly different in MTG of mTLE patients when compared with controls (mTLE = 55 ± 4 PV+ cells, control = 54 ± 5 PV+ cells, P = 0.97). These observations are consistent with the paucity of histological changes previously observed in cortex from epilepsy patients (Babb et al. 1984; DeFelipe et al. 1993).

Sharp Wave Activity

As previously described in slices of human temporal neocortex (Kohling et al. 1998, 1999), we observed spontaneous sharp waves in almost 50% of the slices (n = 55) with a mean frequency of 0.9 ± 0.02 Hz (range from 0.27 to 1.6 Hz, n = 27 slices) (Fig. 2A). Such activity could be recorded in one or the other layer in all slices, and simultaneously from both superficial and deep layers in 9 slices. The temporal relation of these regional activities appeared to be variable from slice to slice. In 6 slices, the spontaneous slow wave activity of the deep layer preceded corresponding events in the superficial layer, as the former led in time by 37 ± 5 ms (calculated from 44 spontaneous events; e.g., Fig. 2A inset); whereas in another 3 slices, the deep layer activity lagged in time by 44 ± 19 ms (measured from 16 events). There was no statistical difference between the delays for slow waves originating in either deep or superficial layers (t-test, P = 0.42). Furthermore, there was no statistical difference in the amplitudes of the evoked field potentials in tissues with or without spontaneous slow wave activity (0.39 ± 0.4 mV, n = 14 vs. 0.48 ± 0.25 mV, n = 15 t-test; P = 0.43) and both synaptic facilitation (20.7 ± 6%, n = 4) and depression (21.7 ± 7%, n = 3) could be seen in these neocortical slices.

Figure 2.

Characterization of spontaneous slow waves. (A) Simultaneous superficial and deep layer extracellular recordings demonstrating spontaneous slow waves in the standard recording solution prior to the addition of kainate and carbachol. Right, fast sweep of one spontaneous slow wave demonstrating the time delay between layers. (B) Representative trace of spontaneous slow waves during simultaneous intracellular (IC) voltage clamp of a pyramidal cell and extracellular (EC) recording (n = 6) to demonstrate the reversal potential of the intracellular currents correlated with the sharp wave activity. By holding at various potentials, the reversal potential of the spontaneous outward currents was estimated from the potential at which the fitted (solid line) intersected with zero. In this cell this was estimated to be −65 mV.

Figure 2.

Characterization of spontaneous slow waves. (A) Simultaneous superficial and deep layer extracellular recordings demonstrating spontaneous slow waves in the standard recording solution prior to the addition of kainate and carbachol. Right, fast sweep of one spontaneous slow wave demonstrating the time delay between layers. (B) Representative trace of spontaneous slow waves during simultaneous intracellular (IC) voltage clamp of a pyramidal cell and extracellular (EC) recording (n = 6) to demonstrate the reversal potential of the intracellular currents correlated with the sharp wave activity. By holding at various potentials, the reversal potential of the spontaneous outward currents was estimated from the potential at which the fitted (solid line) intersected with zero. In this cell this was estimated to be −65 mV.

Simultaneous extracellular and whole-cell recordings were used to explore the synaptic activity underlying such activity. As shown in Figure 2B, putative pyramidal neurons exhibited evident synaptic currents in temporal association with these sharp waves. The synaptic currents were largely inward at negative potentials (near −70 mV) and outward synaptic currents became dominant at holding potentials more positive than −50 mV (Fig. 2B). The calculated reversal potentials of these synaptic currents were −62 ± 2 mV, suggestive of strong GABAergic mechanisms underlying the sharp wave events. Indeed, such sharp waves were shown to be abolished by the application of bicuculline and picrotoxin in human temporal lobe slices (Kohling et al. 1998), suggesting preservation of inhibitory mechanisms in human temporal neocortex from patients with epilepsy (Foehring et al. 1991; Menendez de la Prida et al. 2002).

Both the presence of functional excitatory and inhibitory cells, and the ability to generate spontaneous activity associated with inhibitory and excitatory synaptic currents suggested to us that the human neocortical slices used in our experiments may possess the cellular and microcircuit constituents necessary for the generation of “physiological” oscillations in vitro.

Oscillations Induced by Kainate and Carbachol

Basic Features of Induced Activity

Various frequencies of oscillatory activity have been induced in rodent neocortex by pharmacological activation with kainate, carbachol, or a combination of these agonists (Lukatch and MacIver 1997; Buhl et al. 1998; Roopun et al. 2006; Yamawaki et al. 2008; van Aerde et al. 2009; Oke et al. 2010). To establish a stable protocol that could reliably induce activity akin to what has been observed in rodent cortex, we screened a large number of slices (n = 448, from 23 patients, Table 1) by using various concentrations of agents either alone or in combination. The protocol we ultimately settled on was bath application of 50 nM kainate for 10–15 min, followed by the addition of 50 μM of carbachol. This protocol induced theta oscillations and gamma activity (observed as increases in spectral power) in 74% (41 of 55) of slices tested and ictal events in a subset (31%, 17 of 55) (Table 3).

Table 3

Activation patterns induced by kainate and kainate plus carbachol in human MTG cortical slices

Total slices (n = 55) Kainate (nKainate and carbachol (n
Activated 33 41 
Peak PS 10 33 
Delta (0.5–4 Hz) 
Theta (4–10 Hz) 18 
Alpha (10–15 Hz) 
Beta (15–30 Hz) 19 
Activated with later ictus 17 
Total slices (n = 55) Kainate (nKainate and carbachol (n
Activated 33 41 
Peak PS 10 33 
Delta (0.5–4 Hz) 
Theta (4–10 Hz) 18 
Alpha (10–15 Hz) 
Beta (15–30 Hz) 19 
Activated with later ictus 17 

The oscillations observed within these tissues did not persist for long periods of time, appearing to emerge from little or no obvious antecedent activity, and then dissipated in an equally abrupt fashion (Fig. 3). They appeared gradually over several minutes after the addition of carbachol to kainate (4 ± 1 min, range 1.5–16.6 min, n = 16), and persisted, usually intermittently, for an average of 2.7 ± 0.5 min (n = 25). The same slice at times could generate both oscillatory activity and ictal events (for a comparison of these different types of activities, please see Figs 3 and 4). Similar oscillations could be induced in the same slices by a second application of kainate and carbachol (n = 5 slices), again with frequencies ranging from theta to high gamma. Such oscillations were usually observed to occur concurrently within the superficial and deep cortical layers, and although they could wax and wane at different times, there appeared to be a temporal relation between superficial and deep layers as measured by wavelet PC (Figs 3C and 9). Activity was particularly conspicuous within the theta (4–8 Hz), gamma (30–60 Hz), and high gamma frequency ranges (60–200 Hz). Such activity, unlike ictal activity (see Fig. 4), did not change its frequency characteristics either at onset or offset (Fig. 3B). Although in some slices there was some fluctuation in frequency within the theta range, such oscillations for the most part remained confined to the “modal” frequency range in which they started (Fig. 3). Quite conspicuously, theta oscillations appeared to have a rather symmetrical rise and fall to them, and associated high-frequency “burst” of activity could be observed. This visually apparent CFC between theta and gamma activity (Figs 3C), is not unlike that which has been described in the intact human brain (Canolty et al. 2006) (see below). The different waveforms observed between superficial and deep layers, as well as the time lag between them (Fig. 3D) suggests that the activity was not volume conducted.

Figure 3.

Evanescent oscillatory activity in human cortical slices induced by kainate and carbachol. (A) Extracellular traces, high pass filtered at 1 Hz, recorded from superficial (black) and deep (green) layer of a cortical slice. The voltage scale is the same for all the raw data traces. Corresponding spectrograms of varying temporal scales during application of kainate and kainate + carbachol are shown below the traces. The spectrograms are computed from the raw trace, not the high pass filtered data. The first color bar ZS on the right corresponds to the various spectrograms. The vertical red line demarcates the transition between kainate and the addition of carbachol. Horizontal red bars under each spectrogram represent the time range expanded immediately below. (B) Expanded traces and corresponding spectrograms. (C) Wavelet phase coherence with its own color bar shown (ZPC) to the right of it. (D) Further expanded traces and corresponding spectrograms. Top, the superficial and deep layer traces are superimposed and the area under each curve colorized to show the temporal relation between the activities in the 2 layers. Although the peak of the activity in the superficial layer leads the deep layer, buildup of activity in the deep layer precedes the peak in the superficial layer.

Figure 3.

Evanescent oscillatory activity in human cortical slices induced by kainate and carbachol. (A) Extracellular traces, high pass filtered at 1 Hz, recorded from superficial (black) and deep (green) layer of a cortical slice. The voltage scale is the same for all the raw data traces. Corresponding spectrograms of varying temporal scales during application of kainate and kainate + carbachol are shown below the traces. The spectrograms are computed from the raw trace, not the high pass filtered data. The first color bar ZS on the right corresponds to the various spectrograms. The vertical red line demarcates the transition between kainate and the addition of carbachol. Horizontal red bars under each spectrogram represent the time range expanded immediately below. (B) Expanded traces and corresponding spectrograms. (C) Wavelet phase coherence with its own color bar shown (ZPC) to the right of it. (D) Further expanded traces and corresponding spectrograms. Top, the superficial and deep layer traces are superimposed and the area under each curve colorized to show the temporal relation between the activities in the 2 layers. Although the peak of the activity in the superficial layer leads the deep layer, buildup of activity in the deep layer precedes the peak in the superficial layer.

Figure 4.

Ictal events display unstable frequency characteristics over time. (A, B) Spectrograms of two types of ictal events observed in human cortical tissue exposed to kainate and carbachol. The event in (A) is recorded from the superficial cortical layer, with no accompanying activity in the deep cortical layer.

Figure 4.

Ictal events display unstable frequency characteristics over time. (A, B) Spectrograms of two types of ictal events observed in human cortical tissue exposed to kainate and carbachol. The event in (A) is recorded from the superficial cortical layer, with no accompanying activity in the deep cortical layer.

From a synaptic perspective, the induction of oscillations with this protocol was associated with significant decreases in evoked synaptic field potentials (0.48 ± 0.06 mV in baseline vs. 0.27 ± 0.05 mV, (n = 31, P < 0.001, 1-way ANOVA) and no evoked epileptiform discharges. The amplitude of the evoked synaptic field potential recovered (0.43 ± 0.05 vs. 0.41 ± 0.1 mV) after 3.5 ± 0.7 min of kainate and carbachol wash out.

Given the above observations, increases in spectral power was used as a measure of induced activity (Fig. 5A). Only superficial cortical layers displayed significant power increases with kainate application. This occurred over a wide range of frequencies (4–400 Hz) without any notable peaks in the power spectrum. We refer to this type of power increase as “broadband” activity. The addition of carbachol-induced further broadband increases in power (4–300 Hz) above that of kainate, with the appearance of peaks (“narrowband” increases) in the low-frequency range, particularly within the theta frequency range (Fig. 5A). This was observed in both layers. In fact, on a slice-by-slice basis, spectral peaks were observed most commonly in the theta and beta frequency range (37 of 55 slices) (Table 3). These peaks were taken to represent oscillatory activity, as can be evidenced in the raw data (Fig. 3). The beta peak may however be an harmonic of the theta activity, and thus potentially artifactual. Although both layers displayed power increases (Fig. 5), the power increase for theta frequency range was almost 10-fold higher for superficial cortical layers. Interestingly, carbachol was incapable of inducing power increases in the absence of kainate (n = 7).

Figure 5.

Kainate and carbachol induce broadband and narrowband increases in spectral power in human MTG slices in vitro. (A) Group averaged normalized spectra; the bars above the plots indicate significant power increases during application of kainate (blue), and kainate plus carbachol (green) from the baseline condition. The red bar below indicates the frequency range over which application of kainate + carbachol induced power increases greater than the kainate condition. Insets show non-normalized power spectra for the 3 different conditions: Baseline, Kainate, Kainate + carbachol.

Figure 5.

Kainate and carbachol induce broadband and narrowband increases in spectral power in human MTG slices in vitro. (A) Group averaged normalized spectra; the bars above the plots indicate significant power increases during application of kainate (blue), and kainate plus carbachol (green) from the baseline condition. The red bar below indicates the frequency range over which application of kainate + carbachol induced power increases greater than the kainate condition. Insets show non-normalized power spectra for the 3 different conditions: Baseline, Kainate, Kainate + carbachol.

Induced Oscillations Differ from Ictal Events

In addition to the induced activity described above, application of kainic acid and carbachol was found to provoke delayed ictal events in 31% of slices in our experiments (Figs 4 and 6Ai). These ictal events only occurred after a delay of 9.5 ± 7.1 min (n = 11) after the addition of both kainate and carbachol. Ictal events differed from induced oscillations on several counts. First, the amplitude of induced oscillations were smaller than ictal events; 51 ± 33 μV (range: 13–135 μV, n = 13) in superficial layers and 26 ± 18 μV (range: 9–73 μV, n = 14) in deep layers for oscillations, while the ictal events had amplitudes of 226 ± 48 μV (range: 132–277 μV, n = 7) in superficial layers and 246 ± 49 μV (range 113–311 μV, n = 5) in deep layers, (1-way ANOVA P < 0.001 for both layers). Furthermore, the induced oscillations were more robust in the superficial layer relative to the deep layers, whereas ictal events were equally expressed in both layers (Fig. 6Aii). Second, the induced oscillations persisted in the presence of phenytoin, albeit reduced in amplitude, whereas phenytoin abolished ictal events in all slices examined (n = 8, Fig. 6C,D). Lastly, the induced oscillations were associated with frequent EPSCs and/or rhythmic IPSCs in putative pyramidal cells, whereas the ictus correlated with intense discharges in pyramidal cells (n = 4, Fig. 6B). The latter is in keeping with previous studies in human and murine cortical slices (Gutnick et al. 1982; Dudek and Spitz 1997; McCormick and Contreras 2001; Beenhakker and Huguenard 2009; Lasztoczi et al. 2009; Huberfeld et al. 2011).

Figure 6.

Electrophysiology of ictal events induced by kainate plus carbachol, and distinguishing features from other activated states. (Ai) Top panel shows a representative trace of 2 ictal events induced by kainate plus carbachol in the human MTG. In the middle and bottom, faster sweeps of the electrographic activities observed during 3 distinguishable stages of the ictal events; preictal (blue marker), ictal (red marker), and late-ictal (green marker). (Aii) Bar graphs showing the difference in amplitudes between layers of the activities recorded and classified as ictal events and rhythms. (B) Representative trace showing the intense firing pattern of a putative pyramidal cell during the ictus, and their apparent correlation with the ongoing field potential. (C) Representative trace of an experiment demonstrating the effect of phenytoin on the kainate + carbachol induced ictus. Note that in the presence of phenytoin, physiological oscillations appear under kainate + carbachol, while the ictal events do not recur. (D) Power spectra and representative traces of slices pretreated with phenytoin under kainate + carbachol. The bars above the plots indicate significant power increases during application of kainate (blue), and kainate + carbachol (green) from the baseline condition. The red bar below indicates the frequency range over which application of kainate + carbachol during preapplication of phenytoin-induced power increases above the baseline condition.

Figure 6.

Electrophysiology of ictal events induced by kainate plus carbachol, and distinguishing features from other activated states. (Ai) Top panel shows a representative trace of 2 ictal events induced by kainate plus carbachol in the human MTG. In the middle and bottom, faster sweeps of the electrographic activities observed during 3 distinguishable stages of the ictal events; preictal (blue marker), ictal (red marker), and late-ictal (green marker). (Aii) Bar graphs showing the difference in amplitudes between layers of the activities recorded and classified as ictal events and rhythms. (B) Representative trace showing the intense firing pattern of a putative pyramidal cell during the ictus, and their apparent correlation with the ongoing field potential. (C) Representative trace of an experiment demonstrating the effect of phenytoin on the kainate + carbachol induced ictus. Note that in the presence of phenytoin, physiological oscillations appear under kainate + carbachol, while the ictal events do not recur. (D) Power spectra and representative traces of slices pretreated with phenytoin under kainate + carbachol. The bars above the plots indicate significant power increases during application of kainate (blue), and kainate + carbachol (green) from the baseline condition. The red bar below indicates the frequency range over which application of kainate + carbachol during preapplication of phenytoin-induced power increases above the baseline condition.

Of potentially greater specificity than the differences mentioned above, ictal events did not display stable frequency characteristics either at their onsets or offsets, or during the event. Shown in Figure 4 are examples of 2 types of events that we classified as being ictal in nature. These patterns somewhat resemble the predominant types of seizures observed in intracranial recordings from subjects with temporal lobe epilepsy (Velasco et al. 2000). The first pattern was represented by intermittent “preictal bursts,” followed by a ramp in frequency from delta up into the high-beta range (20–30 Hz), then a slowing over the ictus, followed by a postictal state of intermittent disorganized bursts (Fig. 4A). This pattern in some ways resembles the low-voltage fast onset type seizure (Velasco et al. 2000). The second observed pattern was charaterized by a slow increase in frequency from the delta into the theta frequency range (Fig. 4B), which appeared oscillatory at onset resembling the hypersynchronous-type seizure (Velasco et al. 2000). The morphology of such “pathological oscillations” when viewed on an expanded time scale (Fig. 4B, lower panel) was quite distinct from that observed with physiological oscillations (Fig. 3). High-frequency activity can be seen early on within the oscillatory cycle, creating an asymmetric appearance to the pathological activity (Fig. 4B). A detailed analysis of these seizures types in underway and will be reported separately.

Induced Oscillations Depend on Muscarinic Receptors and Synaptic Activity and Correlate with Rhythmic IPSCs in Pyramidal Cells

Carbachol is a mixed agonist of both muscarinic and nicotinic receptors. To determine which subtype of receptor is necessary for the induced oscillations described above, atropine a selective blocker of muscarinic receptors was preapplied to slices that had displayed previous activation with kainate and carbachol. Atropine (10 μM) completely abolished carbachol activation (n = 4, Fig. 7A). The blockade was not specific to a specific frequency band: instead, all spectral power increases above baseline were abolished. Conversely, application of nicotine (1, 5, 10, and 100 μM) in the presence of kainate was not sufficient to induce rhythms (n = 6, data not shown).

Figure 7.

Synaptic transmission and cholinergic activation are essential for rhythm induction in human MTG. (A) Effects of 10 μM atropine on the rhythms induced by kainate and carbachol. (Ai) Grouped normalized spectral analysis (n = 4) showing significant decrease in the activity induced by kainate and carbachol after the application of the muscarinic receptor antagonist atropine. The only significant power increases were observed during application of kainate and carbachol (green line). (Aii) Representative extracellular recordings under various conditions. (Bi) Grouped normalized spectral analysis (n = 5) showing the effect of synaptic transmission blockade on the activation pattern induced by kainate. Such blockade induced by low calcium (0.5 mM) and high magnesium (3 mM), significantly reduced the broadband power changes induced by kainate (blue bar). (Bii) Representative traces under varying experimental conditions.

Figure 7.

Synaptic transmission and cholinergic activation are essential for rhythm induction in human MTG. (A) Effects of 10 μM atropine on the rhythms induced by kainate and carbachol. (Ai) Grouped normalized spectral analysis (n = 4) showing significant decrease in the activity induced by kainate and carbachol after the application of the muscarinic receptor antagonist atropine. The only significant power increases were observed during application of kainate and carbachol (green line). (Aii) Representative extracellular recordings under various conditions. (Bi) Grouped normalized spectral analysis (n = 5) showing the effect of synaptic transmission blockade on the activation pattern induced by kainate. Such blockade induced by low calcium (0.5 mM) and high magnesium (3 mM), significantly reduced the broadband power changes induced by kainate (blue bar). (Bii) Representative traces under varying experimental conditions.

The above-described increase in broadband power by kainate, which was further augmented by carbachol, has been suggested to arise from asynchronous spiking activity (Ray et al. 2008; Miller 2010). If such power changes were solely a result of increased spiking (Ray et al. 2008) we would expect to see power changes in the absence of synaptic transmission. We therefore attempted to attenuate synaptic transmission by perfusing slices using low Ca2+/high Mg2+ perfusate (Jefferys and Haas 1982; Yaari et al. 1983; Aram et al. 1991). Under these conditions broad band activity was abolished (n = 5) (Fig. 7B), suggesting that the power changes observed in human tissue requires intact synaptic transmission, and may be an emergent property of intact neocortical microcircuitry (Miller 2010).

Somatic voltage clamp recordings from pyramidal cells from held at −60mV layers during physiological oscillations demonstrated increases in spontaneous EPSCs (4.3 ± 5 Hz in baseline vs. 11.1 ± 9 Hz in kainate and carbachol, 1-way ANOVA P = 0.009, n = 14) (Fig. 8A). The increased EPSC frequency, occurred particularly during delta and theta activities. Cross-correlation analysis of the extracellular and intracellular recordings during these rhythms at −70 mV, showed entrainment with the extracellularly measured field potentials (n = 9). Similarly, under voltage clamp conditions holding at 0 mV, far from the reversal potential of Cl and at the reversal potential for EPSCs, spontaneous rhythmic IPSCs were observed (Fig. 8B), which displayed a strong positive correlation to the ongoing field activity. The average frequency of this correlation was quantified using WC. This demonstrated that the maximal coherence between intracellular IPSCs and the ongoing field potential lay within the theta frequency range (n = 4 slices) (Fig. 8C). As mentioned above the beta “oscillations” may be an harmonic of the theta activity, and thus coherence was seen at this frequency as well.

Figure 8.

Oscillations in the human MTG are associated with increased rhythmic synaptic activity: (A) Voltage clamp recording at holding potential of −60 mV during various stages of activation. Significant increase in the frequency of EPSCs occurred only during the application of kainate + carbachol (n = 14 cells). (B) Representative intracellular voltage clamp recordings (IC) at 0 mV membrane potential, and extracellular (EC) traces displaying rhythmic activity in the theta frequency range during kainate + carbachol application. Cross-correlation (right panel) analysis of a 2-s sweep between IC and EC recordings. (C) Wavelet coherence analysis of IPSCs and EC signal from (B) demonstrating significant effects during kainate + carbachol (red line) when compared with baseline. This was observed in 2 other cells.

Figure 8.

Oscillations in the human MTG are associated with increased rhythmic synaptic activity: (A) Voltage clamp recording at holding potential of −60 mV during various stages of activation. Significant increase in the frequency of EPSCs occurred only during the application of kainate + carbachol (n = 14 cells). (B) Representative intracellular voltage clamp recordings (IC) at 0 mV membrane potential, and extracellular (EC) traces displaying rhythmic activity in the theta frequency range during kainate + carbachol application. Cross-correlation (right panel) analysis of a 2-s sweep between IC and EC recordings. (C) Wavelet coherence analysis of IPSCs and EC signal from (B) demonstrating significant effects during kainate + carbachol (red line) when compared with baseline. This was observed in 2 other cells.

Interlaminar Communication During Activation

Our superficial and deep layer recordings roughly correspond to supragranular, and infragranular compartments respectively. How activity is integrated between these layers is a critical question that underlies the computational properties of brain microcircuits. We thus asked if the activity we observe in activated slices was occurring synchronously throughout all layers, or if it remained segregated. To address this question, we employed PC (Lachaux et al. 2000) to examine the temporal dependence of the activity occurring in the superficial and deep layers. Significant synchrony between the superficial and deep layers was observed during kainate and carbachol application within the theta frequency range (Fig. 3). By collapsing the measurement over time, we obtained a single metric of PC as a function of frequency for each slice. Group level analyses revealed no significant PC during kainate application alone, but in combination with carbachol significant synchrony was observed between 5 and 12 Hz, with a peak at 8.5 Hz (Fig. 9A). The nonsignificant peaks observed in the baseline condition (as well as at very low frequencies) likely arise from the sharp waves that were present during baseline recordings, yet abated with activation. To determine if the observed PC represents propagated activity, we examined the phase lag between the 2 layers. A non-zero phase lag would suggest that the activity is likely propagated from one layer to another (Bao and Wu 2003). All slices displayed nonuniform phase distributions (Rayleigh test) around the center frequency of 8.5 Hz (actual frequency interval analyzed was 7–10 Hz) suggesting that within each slice there was a preferred phase lag between layers. The average phase lag across slices (Fig. 9B) was significantly different from zero only during the kainate plus carbachol condition (mean = 34°, 95% confidence = 2–82°), with the deep layer leading on average. Thus, not unlike what has been observed in murine neocortex, human neocortex theta activity although rhythmically organized does not occur with zero phase lag (Bao and Wu 2003). It is as well important to note that gamma activity was not phase coherent between layers suggesting independent gamma generation in the superficial and deep cortical layers.

Figure 9.

Phase coherence between superficial and deep layers is maximal at theta frequency. (A) Averaged PC both across time and across brain slices (n = 25). The red colored bar above the plot indicate statistical significance with respect to baseline conditions, using nonparametric statistical testing. (B) Rose plots for peak PC frequencies (7–10 Hz). Distributions were generated for each trial and condition. Mean phase angle are shown as red arrows. Degrees are shown in black, histogram counts are shown in gray. At theta frequencies (7–10 Hz) on average, the deeper layer leads.

Figure 9.

Phase coherence between superficial and deep layers is maximal at theta frequency. (A) Averaged PC both across time and across brain slices (n = 25). The red colored bar above the plot indicate statistical significance with respect to baseline conditions, using nonparametric statistical testing. (B) Rose plots for peak PC frequencies (7–10 Hz). Distributions were generated for each trial and condition. Mean phase angle are shown as red arrows. Degrees are shown in black, histogram counts are shown in gray. At theta frequencies (7–10 Hz) on average, the deeper layer leads.

Modulation of High Gamma Activity

The establishment of an oscillatory hierarchy (Lakatos et al. 2005) by which one oscillation modulates another, is a mechanism that may link cellular process across different time scales and across space (Canolty and Knight 2010). Although many such coupling regimes could potentially exist (Canolty and Knight 2010), the modulation of high gamma amplitude by theta phase has been described in the human brain (Canolty et al. 2006). Such CFC between theta and gamma oscillations is suggested to represent a distinct neural code (Lisman and Jensen 2013). Theta-gamma CFC was observable in the raw data traces (Fig. 3), and was present in approximately half (13 of 25) of the slices that displayed physiological oscillations (Fig. 10 and see Supplementary Fig. 2) with in both superficial and deep cortical layers.

Figure 10.

High gamma bursts are of variable frequency and may account for the “broadband” power increases seen during kainate and carbachol activation. (A) Time–frequency analysis of high gamma activity. Note the peaks in power are variable in frequency. To quantitate this, peaks in high gamma were detected from the magnitude (orange trace) of the Hilbert transformed band pass filtered signal (60–180 Hz) (light blue trace). Only those peaks ≥ 2 SD were included (red dots) as “bursts.” Superimposed is the theta oscillation (green trace). Using the identified maxima the frequency at which the power was maximal (from the trace above) was extracted for all bursts. (Bi) Histogram of the incidence of the various high gamma bursts. A strong negative correlation was found between the frequency of the burst and number of occurences. (Bii) The amplitude of each high gamma burst as a function of its frequency. No correlation was found between these 2 features. (Biii) Phase at which the bursts occurred was not uniformly distributed and was similar to that obtained from the MI calculation. (Ci) Shown plotted is the MI from the same slices as used in part A. The black contour line represents the region of significance. (Cii) The phase at which CFC was maximal. The mean phase was computed from the circular mean of all points within the significant region obtained from the MI. (Ciii) The power spectrum from the same slice.

Figure 10.

High gamma bursts are of variable frequency and may account for the “broadband” power increases seen during kainate and carbachol activation. (A) Time–frequency analysis of high gamma activity. Note the peaks in power are variable in frequency. To quantitate this, peaks in high gamma were detected from the magnitude (orange trace) of the Hilbert transformed band pass filtered signal (60–180 Hz) (light blue trace). Only those peaks ≥ 2 SD were included (red dots) as “bursts.” Superimposed is the theta oscillation (green trace). Using the identified maxima the frequency at which the power was maximal (from the trace above) was extracted for all bursts. (Bi) Histogram of the incidence of the various high gamma bursts. A strong negative correlation was found between the frequency of the burst and number of occurences. (Bii) The amplitude of each high gamma burst as a function of its frequency. No correlation was found between these 2 features. (Biii) Phase at which the bursts occurred was not uniformly distributed and was similar to that obtained from the MI calculation. (Ci) Shown plotted is the MI from the same slices as used in part A. The black contour line represents the region of significance. (Cii) The phase at which CFC was maximal. The mean phase was computed from the circular mean of all points within the significant region obtained from the MI. (Ciii) The power spectrum from the same slice.

There are currently 2 competing hypothesis for the generation of broadband gamma (see (Crone et al. 2011) for a detailed discussion of this issue). One hypothesis suggests that high gamma is an “oscillation,” and due to the spatial (and potentially temporal) averaging of a number of differently tuned (narrow band) local high gamma generators, power spectra appear to be broadband. The other hypothesis suggests that high gamma cannot be narrow band, and arises from asynchronous neuronal activity—be it spiking or postsynaptic potentials. The CFC we observed between theta and high gamma provided a convenient temporal signature to address the nature of high gamma activity. As high gamma activity appears in bursts, we hypothesized that one potential source of broadband high gamma may arise from the variability in the frequency of such bursts. By convolving the time series with a wavelet of more cycles (see Materials and Methods section), we improved the frequency estimation of the decomposition. By this method, we observed that each high gamma burst had a distinct peak frequency (Fig. 10A). The observed frequency variability in high gamma bursts might generate broadband power increases (Figs 5 and 10Ciii) by either slower frequency bursts being more frequent, of larger amplitude, or both. To examine these possibilities we plotted the number of high gamma bursts, or their amplitude as a function of their frequency. We observed that slower high gamma bursts were more frequent (Fig. 10Bi) but not larger in amplitude (Fig. 10Bii). To confirm the physiological nature of the detected high gamma bursts, we determined the average theta phase at which they occured. It can be seen that this phase corresponded identically to the phase at which CFC (Fig. 10Ci) was maximal between theta phase and high gamma amplitude (Fig. 10Cii). Performing this analysis across all slices displaying CFC, we observed a consistent pattern where slower gamma bursts were more frequent, but not larger in amplitude (Fig. 11). These findings support the possibility that broadband high gamma may arise from the temporal average of narrowband high gamma bursts of ever fluctuating frequency. What such high-frequency events represent is however speculative, although it is suggested that they represent the transient emergence and dissolution of local cell assemblies (Buzsaki 2011; Lisman and Jensen 2013).

Figure 11.

Slow high gamma bursts are more common than fast high gamma bursts, and carry the same amount of power. Group analyses of high gamma bursts from the superficial layers of all slices displaying significant theta to high gamma CFC. A total of 3066 bursts were identified across 13 slices. (A) The number of bursts as a function of frequency. The nadir at 120 Hz is due to notch filtering. There was a strong inverse relation between the incidence and frequency of burst events (R = −0.8, P = 3 × 10−8). (B) Amplitude of the burst events as a function of frequency. A weak yet significant inverse relationship was found (R = −0.1, P = 0).

Figure 11.

Slow high gamma bursts are more common than fast high gamma bursts, and carry the same amount of power. Group analyses of high gamma bursts from the superficial layers of all slices displaying significant theta to high gamma CFC. A total of 3066 bursts were identified across 13 slices. (A) The number of bursts as a function of frequency. The nadir at 120 Hz is due to notch filtering. There was a strong inverse relation between the incidence and frequency of burst events (R = −0.8, P = 3 × 10−8). (B) Amplitude of the burst events as a function of frequency. A weak yet significant inverse relationship was found (R = −0.1, P = 0).

Discussion

We report in this study that brain slices from resected human temporal neocortex can generate population rhythms following exposure to kainate and carbachol. We thus demonstrate for the first time that small volumes of human neocortex, with the appropriate pharmacological activation, have the necessary cellular constituents and microcircuitry to generate seemingly physiological spectral changes as observed in the intact human brain.

Induced Rhythms are Not Epileptic in Nature

In our experiments, temporal neocortical slices were prepared from surgically resected tissue of individuals with epilepsy. One major concern is whether or to what degree the theta/gamma rhythms we observed represent the activity of isolated epileptogenic circuitry. While we cannot rule out this possibility, several lines of evidence suggest that these rhythms are not ictal in nature. First, the ictal events displayed clearly defined preictal, ictal, and postictal stages observed as unstable frequency profiles. Second, ictal activity was larger in amplitude than nonpathological oscillatory activity. In this context, there was no difference between the magnitude of the activity in the deep and superficial layers during ictal events, whereas larger amplitude activity predominated in superficial layers during oscillations. Third, ictal activity could be blocked by phenytoin, whereas oscillatory activity, although reduced in magnitude, was not abolished. Fourth, it has been shown that seizure-like activity can be elicited with high concentrations of kainate in nonepileptogenic murine cortex (Fisahn et al. 2004), suggesting that oscillations exist in a delicate balance with ictal activity (Fisahn 2005). We thus used the lowest possible concentration of agonist to minimize the likelihood of ictal events. Lastly, physiological oscillations measured from scalp EEG or intracranial EEG (iEEG or electrocorticography) have been segregated into distinct frequency bands, beginning chronologically with the 9–14 Hz alpha oscillation (Berger 1929). Many more have been described since (Niedermeyer and Lopez da Silva 1999), each with its own distinct frequency range, and a seemingly distinct relationship to other named, “modal,” frequency ranges (Draguhn et al. 1998; Roopun et al. 2008). It has been suggested that these modal frequency ranges (i.e., delta, theta, alpha, etc.) may be related to one another by a factor equal to the golden mean (Roopun et al. 2008). This interesting hypothesis has been suggested to prevent cross-talk between different types of oscillations, since their excitatory phases will rarely meet (Pletzer et al. 2010). This reduces the potential for synchronization between frequencies, thus segregating modal frequencies into potential communication “channels.” From this, one might suggest that a physiological oscillation would be unlikely to change its center frequency in a continuous fashion. This would be different for epileptiform activity, during which shifts in frequency content might increase the probability of cross-talk between variably tuned microcircuits. Although we have seen such continuous frequency changes at low frequencies during epileptiform activity, such “glissandi” have been described within the gamma frequency range in epileptogenic cortex (Cunningham et al. 2012). Thus, from a physiological perspective, it would seem advantageous that a dedicated communication channel maintain its frequency content. In those situations where seemingly physiological rhythms have been shown to change their frequency content, it occurs as a stepwise change, and not as a continuous function of frequency (Roopun et al. 2006; Kramer et al. 2008). We therefore suggest that an additional criterion for distinguishing a physiological from a pathological oscillation is the temporal evolution of its frequency profile. As we have observed seemingly physiological oscillations emerge and disappear, as they do in vivo, with a center frequency at any moment in time within the same modal range, whereas pathological oscillation have temporal frequency profiles that are unstable and wax and wane across modal frequencies in a continuous fashion.

From a clinical perspective, the clinical–pathological entity of mTLE is well defined, and can be diagnosed with noninvasive tests. Thus, although we did not perform iEEG in the majority of patients who participated in this study, there is a large body of evidence, both electrophysiological and clinical, demonstrating that the epileptogenic zone is within the mesial temporal lobe structures and not in the neocortex. The strongest evidence is derived from iEEG, that clearly distinguishing mTLE from nTLE with seizure onsets arising from either the hippocampus or parahippocampal gyrus in the former, and from the temporal neoxortex in the latter (Wieser et al. 1993) (see Supplementary Fig. 1 for an example of this). As we selected an homogenous group of patients with mTLE, we are confident that the neocortical tissue that we examined here was not within the epileptogenic zone.

The question however arises, why in some slices did we observe epileptiform activity? Among a large number of possibilities, the generation of physiological oscillations likely represents but one state along a continuum of states, where excess network drive can result in epileptiform activity (Fisahn 2005; Beenhakker and Huguenard 2009). Other possibilities include: 1) Acute anticonvulsant withdrawal, since clinically this can result in seizures (Delanty et al. 1998) and it has been shown that anticonvulsant levels are negligible in these slices (Kohling et al. 1998); 2) loss of specific subpopulations of interneurons (DeFelipe 1999) and; 3) changes in protein expression including specific excitatory receptor types (Gonzalez-Albo, Elston et al. 2001; Gonzalez-Albo, Gomez-Utrero et al. 2001). However, the potential histological, and immunohistochemical changes mentioned above are likely regional (DeFelipe et al. 1993) and may underlie the variability in observing physiological as opposed to pathological rhythms in our slices. If this is the case, the manipulations described here could be exploited to study factors that predispose to human ictogenesis, since regional epigenetic, metabolic, and protein expression profiles could be correlated to electrophysiological markers distinguishing physiological from pathological oscillations.

Theta Oscillations

Although generally thought to be a hippocampal rhythm, human intracranial recordings have revealed the clear presence of neocortical theta oscillations (Kahana et al. 2001; Raghavachari et al. 2001, 2006; Canolty et al. 2006). Our data are consistent with these in vivo human recordings, and unequivocally confirm the suspicion that local theta generators exist in human neocortex (Kahana et al. 2001; Raghavachari et al. 2001). In addition, the highest density of muscarinic receptors in the human brain are in the superficial cortical layers (Zilles 1991), and thus, muscarinic effects (Krnjevic et al. 1971) would be expected to be more profound in the superficial layers. Our data are in keeping with this view, demonstrating stronger theta rhythms in the superficial layers relative to the deep layers following carbachol application (Fig. 5).

Although theta oscillations can be generated in murine cortical slices, they typically require some form of disinhibition (Silva et al. 1991; Lukatch and MacIver 1997; Castro-Alamancos and Rigas 2002; Bao and Wu 2003). In our preparation, pharmacological disinhibition was not required to generate theta rhythms. Furthermore, the theta and beta oscillations induced in rat prefrontal cortex (PFC) were not stable at temperatures above 27 °C (van Aerde et al. 2009; Oke et al. 2010), whereas our recordings were performed at physiological temperatures. Thus, although the appearance of the observed theta oscillations was comparable to rodents, given the pharmacological differences and different sensitivity to temperature, it is possible that the mechanisms generating these oscillations are different in rodents and humans. Nonetheless, similar to in vitro neocortical theta activity in rodents (Lukatch and MacIver 1997), human neocortical theta activity (as well as all other frequency bands) was abolished by atropine, suggesting that muscarinic cholinergic receptor activation is required for inducing theta oscillations. Carbachol alone did not induce significant power changes, unlike kainate, which induced a broadband increase in power, without discernible peaks. Thus, unlike rat PFC, in human cortex, carbachol alone is not sufficient for the generation of low-frequency oscillatory activity. Given the uncertainty of the precise generators of theta, particularly in the hippocampus where this rhythm has been extensively studied (Buzsaki 2002), it would be speculative to propose a precise role that muscarinic activation has in generating theta activity in human neocortical slices. However, a number of putative mechanisms for theta generation in human cortex could be: 1) increased activity in pyramidal cells which naturally resonate at theta (Pike et al. 2000); 2) activation of multipolar bursting interneurons that burst at theta frequencies (Blatow et al. 2003), and 3) network dynamics with the appropriate synaptic delays to create synchronous population activity (Buzsaki 2002). We observed greater power increases during carbachol activation in superficial layers than in deeper layers, suggesting that the differential effect of carbachol may relate in part to the preponderance of theta generating interneurons within superficial layers (Blatow et al. 2003). This would be consistent with the correlation and coherence we observe between theta oscillations measured extracellularly, and IPSCs. However, whether such IPSCs are generated de novo from activity of interneurons, or require pyramidal cell drive, is unknown (Wang 2010), as carbachol can both activate multipolar bursting cells and pyramidal cells which intrinsically resonate at theta.

Does our data suggest a specific mechanism in the generation of theta in the human cortex? That carbachol and kainate were required to generate theta oscillations suggests that both pyramidal cell activation and interneuronal activation may be necessary for the generation of theta activity in human cortex. Although as discussed above, cholinergic activation of multipolar bursting cells (Blatow et al. 2003) by cholinergic agonists results in bursting activity within the theta range in these cells, kainate may somewhat selectively activate interneurons (Fisahn et al. 2004). Thus, it is possible that theta generation in the human cortex requires phasic interneuronal activity. Consistent with this, we observed PC between IPSCs recorded in pyramidal cells and ongoing LFP theta oscillations. However, such experiments correlate interneuronal activity to theta oscillations, but do not provide a causal link between the two. From a microcircuit perspective, we find that the deep layer on average leads the superficial layer. Oscillations of similar frequencies have been shown to originate in deep cortical layers (Silva et al. 1991; Castro-Alamancos and Rigas 2002). Consistent with what we observe here, stimulus-induced 10 Hz oscillations in disinhibited cortical slices occurred first in deep cortical layers, followed by a large sink in the superficial layers (Castro-Alamancos and Rigas 2002). Thus, although we observed larger theta oscillations in the superficial layers, it is possible that they are driven by deep layer activity. Layer V pyramidal cells appear to be possible candidates for driving superficial cortical layers, as they have recently been shown to generate up states in the cortex (Beltramo et al. 2013). Their apical dendrites, the recipients of both excitatory and inhibitory superficial layer inputs have been shown to be richly populated with Ih channels (Berger et al. 2001), possibly endowing them with the ability to oscillate at theta frequencies (Kamondi et al. 1998). Layer V pyramidal cells thus represent one potential cellular substrate for coordinating, and possibly generating theta oscillations within the human cortex (Silva et al. 1991).

High Gamma Activity

An important finding of the present study is the generation of broadband high gamma activity in isolated human cortical slices. High gamma activity measured from human neocortex using intracranial electrodes is a reliable marker of regions of eloquence (Crone et al. 1998) and its modulation accompanies a number of cognitive and motor activities, and sleep (for review, see Crone et al. 2011). Interestingly, in the murine and human cerebellum, activation of nicotinic receptors induces high gamma activity that in murine cerebellum was independent of GABAA-mediated inhibition, and required intact gap junctional communication (Middleton et al. 2008). As the gamma activity we observed from temporal neocortex was blocked by atropine, it would appear that different mechanisms may operate to generate the gamma rhythms in human temporal neocortex relative to those engaged in human cerebellar slices. To address the issue of increased synaptic activity as the putative source of broadband high gamma activity, we used the approach of blocking synaptic transmission by perfusing slices with a low-Ca2+ and high-Mg2+ ACSF and then examined the effects of applied kainate. We found that when evoked synaptic transmission was blocked, kainate did not induce broadband power increases.

There is currently no clear mechanistic understanding of how high gamma activity is generated in the human cortex (Crone et al. 2011). We have shown here that the phase of theta oscillations modulated high gamma amplitude, resulting in bursts of high gamma activity. These bursts displayed an instability in their frequency content, with slower frequency bursts being more common than faster events. We suggest this as one potential explanation for the broadband power changes we observe here, and which have been consistently shown from human intracranial recordings (Crone et al. 1998; Miller 2010). What these gamma bursts represent at the network level remains unclear. Such fast activity has been suggested to represent synchronous spiking (for review see Jefferys et al. 2012), although modeling studies suggest that given the appropriate kinetics, such high-frequency activity can be generated by postsynaptic potentials (Brunel and Wang 2003). Regardless of exactly how they arise, they likely represent transient synchronous activity in a small population of neurons (Jefferys et al. 2012), and thus the high gamma we observe here may in fact represent, as previously discussed, the superposition in either time or space of variously tuned local cell assemblies (Crone et al. 2011). However, the broadband high gamma observed during the application of kainate was abolished when synaptic transmission was abolished. Thus, it is possible that the high gamma bursts we observe are altogether different entities superimposed on broadband high gamma arising from asynchronous spiking activity and postsynaptic potentials (Miller 2010). An alternate explanation would be that upon application of carbachol, the emergence of theta oscillations heralds an organization of the disordered local spiking activity, with the high gamma bursts thus representing the temporal ordering of cellular activity.

Segregation of Activities by Cortical Layer

Superficial and deep layers not only differ histologically but also in their input–output relationships (Douglas and Martin 2004). Superficial layers are the only source of cortico-cortical projections, while deep layers project to subcortical structures (Douglas and Martin 2004). Given these anatomical specializations, it is expected that their electrophysiological properties would be distinct and could even represent separate computational units. Two observations from this study suggest this. First, we observed that cholinergic activation of the superficial layer generated greater power increases within the low-frequency range (200-fold at 8 Hz) than within deep layers at (20-fold increase at 8 Hz). The net effect of muscarinic cholinergic activation in neurons is a slow depolarization, accompanied by spiking activity (Krnjevic et al. 1971; McCormick and Prince 1985) and pyramidal cells are predisposed to resonating at theta frequencies (Pike et al. 2000). Thus, cholinergic activation in addition to kainate may recruit pyramidal cells that resonate at theta frequencies, generating the carbachol-induced theta activity. Since the highest density of muscarinic receptors in the human brain are in the superficial cortical layers (Zilles 1991), muscarinic effects (Krnjevic et al. 1971) would be expected to be more profound in the superficial layers. Second, although phase synchrony between layers was observed within the theta frequency range, this did not extend to the gamma frequency range.

Integration Across Layers

Although we have discussed the relative uniqueness of the distinct layers, and their apparent independence, what mechanism might exist to coordinate activity across cortical laminae? We have shown that the superficial and deep layers are most strongly phase synchronous within the theta frequency range. We as well observed a significant time lag between cortical layers, suggesting that the activity may be either propagated from one layer to another, or may be an emergent property of 2 coupled oscillators—the superficial and deep layers (Bao and Wu 2003). We observed a phase lag between layers of 30°, which corresponds to a 10-ms time lag at a frequency of 8 Hz. Similar interlaminar delays have been observed in human cortex in vivo (Csercsa et al. 2010), and in cat auditory cortex (Atencio and Schreiner 2010).

Conclusions

Through the activation of glutamatergic and cholinergic receptors, we have established that human cortical slices can generate rhythms of similar character to those observed in vivo. Some of the spectral changes, in particular theta and high gamma activity, while of significant importance in the human brain, are not readily reproduced in animal models. This preparation as well preserved the CFC between theta and high gamma that is observed in vivo (Canolty et al. 2006). Furthermore, our findings suggest a new role for theta oscillations as an integrative rhythm between cortical laminae in human cortex, and have provided some mechanistic insights into the generation of broadband high gamma activity. However, the precise relationship of the activity we have observed here in such abnormal tissue to those observed in normal cortex will require future in vivo and in vitro investigations.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org.

Funding

This research was supported by the Canadian Institutes of Health Research-MOP 119603.

Notes

Conflict of Interest: None declared.

References

Aram
JA
Michelson
HB
Wong
RK
Synchronized GABAergic IPSPs recorded in the neocortex after blockade of synaptic transmission mediated by excitatory amino acids
J Neurophysiol
 , 
1991
, vol. 
65
 (pg. 
1034
-
1041
)
Ascoli
GA
Alonso-Nanclares
L
Anderson
SA
Barrionuevo
G
Benavides-Piccione
R
Burkhalter
A
Buzsaki
G
Cauli
B
Defelipe
J
Fairen
A
, et al.  . 
Petilla terminology: nomenclature of features of GABAergic interneurons of the cerebral cortex
Nat Rev Neurosci
 , 
2008
, vol. 
9
 (pg. 
557
-
568
)
Atencio
CA
Schreiner
CE
Columnar connectivity and laminar processing in cat primary auditory cortex
PLoS One
 , 
2010
, vol. 
5
 pg. 
e9521
 
Avoli
M
Hwa
GG
Kostopoulos
G
Olivier
A
Villemure
JG
Electrophysiological analysis of human neocortex in vitro: experimental techniques and methodological approaches
Can J Neurol Sci
 , 
1991
, vol. 
18
 (pg. 
636
-
639
)
Avoli
M
Olivier
A
Electrophysiological properties and synaptic responses in the deep layers of the human epileptogenic neocortex in vitro
J Neurophysiol
 , 
1989
, vol. 
61
 (pg. 
589
-
606
)
Avoli
M
Williamson
A
Functional and pharmacological properties of human neocortical neurons maintained in vitro
Prog Neurobiol
 , 
1996
, vol. 
48
 (pg. 
519
-
554
)
Babb
TL
Brown
WJ
Pretorius
J
Davenport
C
Lieb
JP
Crandall
PH
Temporal lobe volumetric cell densities in temporal lobe epilepsy
Epilepsia
 , 
1984
, vol. 
25
 (pg. 
729
-
740
)
Bao
W
Wu
JY
Propagating wave and irregular dynamics: spatiotemporal patterns of cholinergic theta oscillations in neocortex in vitro
J Neurophysiol
 , 
2003
, vol. 
90
 (pg. 
333
-
341
)
Bartos
M
Vida
I
Jonas
P
Synaptic mechanisms of synchronized gamma oscillations in inhibitory interneuron networks
Nat Rev Neurosci
 , 
2007
, vol. 
8
 (pg. 
45
-
56
)
Beenhakker
MP
Huguenard
JR
Neurons that fire together also conspire together: is normal sleep circuitry hijacked to generate epilepsy?
Neuron
 , 
2009
, vol. 
62
 (pg. 
612
-
632
)
Beltramo
R
D'Urso
G
Dal Maschio
M
Farisello
P
Bovetti
S
Clovis
Y
Lassi
G
Tucci
V
De Pietri Tonelli
D
Fellin
T
Layer-specific excitatory circuits differentially control recurrent network dynamics in the neocortex
Nat Neurosci
 , 
2013
, vol. 
16
 (pg. 
227
-
234
)
Benjamini
Y
Hochberg
Y
Controlling the false discovery rate: a practical and powerful approach to multiple testing
J R Stat Soc B
 , 
1995
, vol. 
57
 (pg. 
289
-
300
)
Berens
P
CircStat: a MATLAB toolbox for circular statistics
J Stat Software
 , 
2009
, vol. 
31
 (pg. 
1
-
21
)
Berger
H
Uber das Elektrenkephlogram des Menshen
Arch Psychiatr Nervenkr
 , 
1929
, vol. 
87
 (pg. 
527
-
570
)
Berger
T
Larkum
ME
Luscher
HR
High I(h) channel density in the distal apical dendrite of layer V pyramidal cells increases bidirectional attenuation of EPSPs
J Neurophysiol
 , 
2001
, vol. 
85
 (pg. 
855
-
868
)
Bischofberger
J
Engel
D
Li
L
Geiger
JR
Jonas
P
Patch-clamp recording from mossy fiber terminals in hippocampal slices
Nat Protoc
 , 
2006
, vol. 
1
 (pg. 
2075
-
2081
)
Blatow
M
Rozov
A
Katona
I
Hormuzdi
SG
Meyer
AH
Whittington
MA
Caputi
A
Monyer
H
A novel network of multipolar bursting interneurons generates theta frequency oscillations in neocortex
Neuron
 , 
2003
, vol. 
38
 (pg. 
805
-
817
)
Brunel
N
Wang
XJ
What determines the frequency of fast network oscillations with irregular neural discharges? I. Synaptic dynamics and excitation-inhibition balance
J Neurophysiol
 , 
2003
, vol. 
90
 (pg. 
415
-
430
)
Bruns
A
Fourier-, Hilbert- and wavelet-based signal analysis: are they really different approaches?
J Neurosci Methods
 , 
2004
, vol. 
137
 (pg. 
321
-
332
)
Buhl
EH
Tamas
G
Fisahn
A
Cholinergic activation and tonic excitation induce persistent gamma oscillations in mouse somatosensory cortex in vitro
J Physiol
 , 
1998
, vol. 
513
 
Pt 1
(pg. 
117
-
126
)
Buzsaki
G
Neural syntax: cell assemblies, synapsembles, and readers
Neuron
 , 
2011
, vol. 
68
 (pg. 
362
-
385
)
Buzsaki
G
Rhythms of the brain
 , 
2006
New York
Oxford University Press
Buzsaki
G
Theta oscillations in the hippocampus
Neuron
 , 
2002
, vol. 
33
 (pg. 
325
-
340
)
Canolty
RT
Edwards
E
Dalal
SS
Soltani
M
Nagarajan
SS
Kirsch
HE
Berger
MS
Barbaro
NM
Knight
RT
High gamma power is phase-locked to theta oscillations in human neocortex
Science
 , 
2006
, vol. 
313
 (pg. 
1626
-
1628
New York, NY
Canolty
RT
Knight
RT
The functional role of cross-frequency coupling
Trends Cogn Sci
 , 
2010
, vol. 
14
 (pg. 
506
-
515
)
Canolty
RT
Soltani
M
Dalal
SS
Edwards
E
Dronkers
NF
Nagarajan
SS
Kirsch
HE
Barbaro
NM
Knight
RT
Spatiotemporal dynamics of word processing in the human brain
Front Neurosci
 , 
2007
, vol. 
1
 (pg. 
185
-
196
)
Castro-Alamancos
MA
Rigas
P
Synchronized oscillations caused by disinhibition in rodent neocortex are generated by recurrent synaptic activity mediated by AMPA receptors
J Physiol
 , 
2002
, vol. 
542
 (pg. 
567
-
581
)
Clark
GD
Rothman
SM
Blockade of excitatory amino acid receptors protects anoxic hippocampal slices
Neuroscience
 , 
1987
, vol. 
21
 (pg. 
665
-
671
)
Crone
NE
Korzeniewska
A
Franaszczuk
PJ
Cortical gamma responses: searching high and low
Int J Psychophysiol
 , 
2011
, vol. 
79
 (pg. 
9
-
15
)
Crone
NE
Miglioretti
DL
Gordon
B
Lesser
RP
Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis. II. Event-related synchronization in the gamma band
Brain
 , 
1998
, vol. 
121
 
Pt 12
(pg. 
2301
-
2315
)
Csercsa
R
Dombovari
B
Fabo
D
Wittner
L
Eross
L
Entz
L
Solyom
A
Rasonyi
G
Szucs
A
Kelemen
A
, et al.  . 
Laminar analysis of slow wave activity in humans
Brain
 , 
2010
, vol. 
133
 (pg. 
2814
-
2829
)
Cunningham
MO
Roopun
A
Schofield
IS
Whittaker
RG
Duncan
R
Russell
A
Jenkins
A
Nicholson
C
Whittington
MA
Traub
RD
Glissandi: transient fast electrocorticographic oscillations of steadily increasing frequency, explained by temporally increasing gap junction conductance
Epilepsia
 , 
2012
, vol. 
53
 (pg. 
1205
-
1214
)
Davison
A
Hinkley
D
Bootstrap methods and their application
 , 
1997
New York
Cambridge University Press
(pg. 
161
-
174
)
DeFelipe
J
Chandelier cells and epilepsy
Brain
 , 
1999
, vol. 
122
 
Pt 10
(pg. 
1807
-
1822
)
DeFelipe
J
Garcia Sola
R
Marco
P
del Rio
MR
Pulido
P
Ramon y Cajal
S
Selective changes in the microorganization of the human epileptogenic neocortex revealed by parvalbumin immunoreactivity
Cereb Cortex
 , 
1993
, vol. 
3
 (pg. 
39
-
48
)
Delanty
N
Vaughan
CJ
French
JA
Medical causes of seizures
Lancet
 , 
1998
, vol. 
352
 (pg. 
383
-
390
)
Douglas
RJ
Martin
KA
Neuronal circuits of the neocortex
Annu Rev Neurosci
 , 
2004
, vol. 
27
 (pg. 
419
-
451
)
Draguhn
A
Traub
RD
Schmitz
D
Jefferys
JG
Electrical coupling underlies high-frequency oscillations in the hippocampus in vitro
Nature
 , 
1998
, vol. 
394
 (pg. 
189
-
192
)
Dudek
FE
Spitz
M
Hypothetical mechanisms for the cellular and neurophysiologic basis of secondary epileptogenesis: proposed role of synaptic reorganization
J Clin Neurophysiol
 , 
1997
, vol. 
14
 (pg. 
90
-
101
)
Edwards
E
Nagarajan
SS
Dalal
SS
Canolty
RT
Kirsch
HE
Barbaro
NM
Knight
RT
Spatiotemporal imaging of cortical activation during verb generation and picture naming
NeuroImage
 , 
2010
, vol. 
50
 (pg. 
291
-
301
)
El-Hayek
YH
Wu
C
Ye
H
Wang
J
Carlen
PL
Zhang
L
Hippocampal excitability is increased in aged mice
Exp Neurol
 , 
2013
 
pii: S0014-4886(13)00095-2. doi: 10.1016/j.expneurol.2013.03.012.
Fell
J
Axmacher
N
The role of phase synchronization in memory processes
Nat Rev Neurosci
 , 
2011
, vol. 
12
 (pg. 
105
-
118
)
Fisahn
A
Kainate receptors and rhythmic activity in neuronal networks: hippocampal gamma oscillations as a tool
J Physiol
 , 
2005
, vol. 
562
 (pg. 
65
-
72
)
Fisahn
A
Contractor
A
Traub
RD
Buhl
EH
Heinemann
SF
McBain
CJ
Distinct roles for the kainate receptor subunits GluR5 and GluR6 in kainate-induced hippocampal gamma oscillations
J Neurosci
 , 
2004
, vol. 
24
 (pg. 
9658
-
9668
)
Fisahn
A
Pike
FG
Buhl
EH
Paulsen
O
Cholinergic induction of network oscillations at 40 Hz in the hippocampus in vitro
Nature
 , 
1998
, vol. 
394
 (pg. 
186
-
189
)
Fisahn
A
Yamada
M
Duttaroy
A
Gan
JW
Deng
CX
McBain
CJ
Wess
J
Muscarinic induction of hippocampal gamma oscillations requires coupling of the M1 receptor to two mixed cation currents
Neuron
 , 
2002
, vol. 
33
 (pg. 
615
-
624
)
Foehring
RC
Lorenzon
NM
Herron
P
Wilson
CJ
Correlation of physiologically and morphologically identified neuronal types in human association cortex in vitro
J Neurophysiol
 , 
1991
, vol. 
66
 (pg. 
1825
-
1837
)
Fries
P
A mechanism for cognitive dynamics: neuronal communication through neuronal coherence
Trends Cogn Sci
 , 
2005
, vol. 
9
 (pg. 
474
-
480
)
Gaona
CM
Sharma
M
Freudenburg
ZV
Breshears
JD
Bundy
DT
Roland
J
Barbour
DL
Schalk
G
Leuthardt
EC
Nonuniform high-gamma (60–500 Hz) power changes dissociate cognitive task and anatomy in human cortex
J Neurosci
 , 
2011
, vol. 
31
 (pg. 
2091
-
2100
)
Gonzalez-Albo
MC
Elston
GN
DeFelipe
J
The human temporal cortex: characterization of neurons expressing nitric oxide synthase, neuropeptides and calcium-binding proteins, and their glutamate receptor subunit profiles
Cereb Cortex
 , 
2001
, vol. 
11
 (pg. 
1170
-
1181
)
Gonzalez-Albo
MC
Gomez-Utrero
E
Sanchez
A
Sola
RG
DeFelipe
J
Changes in the colocalization of glutamate ionotropic receptor subunits in the human epileptic temporal lobe cortex
Exp Brain Res
 , 
2001
, vol. 
138
 (pg. 
398
-
402
)
Gutnick
MJ
Connors
BW
Prince
DA
Mechanisms of neocortical epileptogenesis in vitro
J Neurophysiol
 , 
1982
, vol. 
48
 (pg. 
1321
-
1335
)
Hajos
N
Ellender
TJ
Zemankovics
R
Mann
EO
Exley
R
Cragg
SJ
Freund
TF
Paulsen
O
Maintaining network activity in submerged hippocampal slices: importance of oxygen supply
Eur J Neurosci
 , 
2009
, vol. 
29
 (pg. 
319
-
327
)
Hajos
N
Mody
I
Establishing a physiological environment for visualized in vitro brain slice recordings by increasing oxygen supply and modifying aCSF content
J Neurosci Methods
 , 
2009
, vol. 
183
 (pg. 
107
-
113
)
Hasenstaub
A
Shu
Y
Haider
B
Kraushaar
U
Duque
A
McCormick
DA
Inhibitory postsynaptic potentials carry synchronized frequency information in active cortical networks
Neuron
 , 
2005
, vol. 
47
 (pg. 
423
-
435
)
Hormuzdi
SG
Pais
I
LeBeau
FE
Towers
SK
Rozov
A
Buhl
EH
Whittington
MA
Monyer
H
Impaired electrical signaling disrupts gamma frequency oscillations in connexin 36-deficient mice
Neuron
 , 
2001
, vol. 
31
 (pg. 
487
-
495
)
Huberfeld
G
Menendez de la Prida
L
Pallud
J
Cohen
I
Le Van Quyen
M
Adam
C
Clemenceau
S
Baulac
M
Miles
R
Glutamatergic pre-ictal discharges emerge at the transition to seizure in human epilepsy
Nat Neurosci
 , 
2011
, vol. 
14
 (pg. 
627
-
634
)
Isaacson
JS
Scanziani
M
How inhibition shapes cortical activity
Neuron
 , 
2011
, vol. 
72
 (pg. 
231
-
243
)
Jefferys
JG
Haas
HL
Synchronized bursting of CA1 hippocampal pyramidal cells in the absence of synaptic transmission
Nature
 , 
1982
, vol. 
300
 (pg. 
448
-
450
)
Jefferys
JG
Menendez de la Prida
L
Wendling
F
Bragin
A
Avoli
M
Timofeev
I
Lopes da Silva
FH
Mechanisms of physiological and epileptic HFO generation
Prog Neurobiol
 , 
2012
, vol. 
98
 (pg. 
250
-
264
)
Jones-Gotman
M
Engel
J
Neuropsychological testing for localizing and lateralizing the epileptogenic region
Surgical treatment of the epilepsies
 , 
1993
New York
Raven Press
(pg. 
203
-
211
)
Kahana
MJ
Seelig
D
Madsen
JR
Theta returns
Curr Opin Neurobiol
 , 
2001
, vol. 
11
 (pg. 
739
-
744
)
Kamondi
A
Acsady
L
Wang
XJ
Buzsaki
G
Theta oscillations in somata and dendrites of hippocampal pyramidal cells in vivo: activity-dependent phase-precession of action potentials
Hippocampus
 , 
1998
, vol. 
8
 (pg. 
244
-
261
)
Kohling
R
Avoli
M
Methodological approaches to exploring epileptic disorders in the human brain in vitro
J Neurosci Methods
 , 
2006
, vol. 
155
 (pg. 
1
-
19
)
Kohling
R
Lucke
A
Straub
H
Speckmann
EJ
Tuxhorn
I
Wolf
P
Pannek
H
Oppel
F
Spontaneous sharp waves in human neocortical slices excised from epileptic patients
Brain
 , 
1998
, vol. 
121
 
Pt 6
(pg. 
1073
-
1087
)
Kohling
R
Qu
M
Zilles
K
Speckmann
EJ
Current-source-density profiles associated with sharp waves in human epileptic neocortical tissue
Neuroscience
 , 
1999
, vol. 
94
 (pg. 
1039
-
1050
)
Kostopoulos
G
Drapeau
C
Avoli
M
Olivier
A
Villemeure
JG
Endogenous adenosine can reduce epileptiform activity in the human epileptogenic cortex maintained in vitro
Neurosci Lett
 , 
1989
, vol. 
106
 (pg. 
119
-
124
)
Kramer
MA
Roopun
AK
Carracedo
LM
Traub
RD
Whittington
MA
Kopell
NJ
Rhythm generation through period concatenation in rat somatosensory cortex
PLoS Comput Biol
 , 
2008
, vol. 
4
 pg. 
e1000169
 
Krnjevic
K
Pumain
R
Renaud
L
The mechanism of excitation by acetylcholine in the cerebral cortex
J Physiol
 , 
1971
, vol. 
215
 (pg. 
247
-
268
)
Lachaux
J
Rodriguez
E
Le Van Quyen
M
Lutz
A
Martinerie
J
Varela
F
Studying single-trials of phase synchronous activity in the brain
Int J Bifurcation Chaos Appl Sci Eng
 , 
2000
, vol. 
10
 (pg. 
2429
-
2440
)
Lakatos
P
Shah
AS
Knuth
KH
Ulbert
I
Karmos
G
Schroeder
CE
An oscillatory hierarchy controlling neuronal excitability and stimulus processing in the auditory cortex
J Neurophysiol
 , 
2005
, vol. 
94
 (pg. 
1904
-
1911
)
Lasztoczi
B
Nyitrai
G
Heja
L
Kardos
J
Synchronization of GABAergic inputs to CA3 pyramidal cells precedes seizure-like event onset in juvenile rat hippocampal slices
J neurophysiol
 , 
2009
, vol. 
102
 (pg. 
2538
-
2553
)
Lisman
JE
Jensen
O
The theta-gamma neural code
Neuron
 , 
2013
, vol. 
77
 (pg. 
1002
-
1016
)
Lukatch
HS
MacIver
MB
Physiology, pharmacology, and topography of cholinergic neocortical oscillations in vitro
J Neurophysiol
 , 
1997
, vol. 
77
 (pg. 
2427
-
2445
)
Mansouri
A
Fallah
A
Valiante
TA
Determining surgical candidacy in temporal lobe epilepsy
Epilepsy Res Treat
 , 
2012
 
2012
Marco
P
Sola
RG
Pulido
P
Alijarde
MT
Sanchez
A
Ramon y Cajal
S
DeFelipe
J
Inhibitory neurons in the human epileptogenic temporal neocortex. An immunocytochemical study
Brain
 , 
1996
, vol. 
119
 
Pt 4
(pg. 
1327
-
1347
)
McCormick
DA
GABA as an inhibitory neurotransmitter in human cerebral cortex
J Neurophysiol
 , 
1989
, vol. 
62
 (pg. 
1018
-
1027
)
McCormick
DA
Contreras
D
On the cellular and network bases of epileptic seizures
Annu Rev Physiol
 , 
2001
, vol. 
63
 (pg. 
815
-
846
)
McCormick
DA
Prince
DA
Two types of muscarinic response to acetylcholine in mammalian cortical neurons
Proc Natl Acad Sci U S A
 , 
1985
, vol. 
82
 (pg. 
6344
-
6348
)
Menendez de la Prida
L
Benavides-Piccione
R
Sola
R
Pozo
MA
Electrophysiological properties of interneurons from intraoperative spiking areas of epileptic human temporal neocortex
Neuroreport
 , 
2002
, vol. 
13
 (pg. 
1421
-
1425
)
Middleton
SJ
Racca
C
Cunningham
MO
Traub
RD
Monyer
H
Knopfel
T
Schofield
IS
Jenkins
A
Whittington
MA
High-frequency network oscillations in cerebellar cortex
Neuron
 , 
2008
, vol. 
58
 (pg. 
763
-
774
)
Miller
KJ
Broadband spectral change: evidence for a macroscale correlate of population firing rate?
J Neurosci
 , 
2010
, vol. 
30
 (pg. 
6477
-
6479
)
Molnar
G
Olah
S
Komlosi
G
Fule
M
Szabadics
J
Varga
C
Barzo
P
Tamas
G
Complex events initiated by individual spikes in the human cerebral cortex
PLoS Biol
 , 
2008
, vol. 
6
 pg. 
e222
 
Niedermeyer
E
Lopez da Silva
F
Electroencephalography
 , 
1999
Baltimore
Williams & Wilkins
Ojemann
G
Ojemann
J
Lettich
E
Berger
M
Cortical language localization in left, dominant hemisphere. An electrical stimulation mapping investigation in 117 patients
J Neurosurg
 , 
1989
, vol. 
71
 (pg. 
316
-
326
)
Oke
OO
Magony
A
Anver
H
Ward
PD
Jiruska
P
Jefferys
JG
Vreugdenhil
M
High-frequency gamma oscillations coexist with low-frequency gamma oscillations in the rat visual cortex in vitro
Eur J Neurosci
 , 
2010
, vol. 
31
 (pg. 
1435
-
1445
)
Pei
X
Leuthardt
EC
Gaona
CM
Brunner
P
Wolpaw
JR
Schalk
G
Spatiotemporal dynamics of electrocorticographic high gamma activity during overt and covert word repetition
NeuroImage
 , 
2011
, vol. 
54
 (pg. 
2960
-
2972
)
Penfield
W
Rasmussen
T
The cerebral cortex of man
 , 
1950
New York
The Macmillan Company
Pike
FG
Goddard
RS
Suckling
JM
Ganter
P
Kasthuri
N
Paulsen
O
Distinct frequency preferences of different types of rat hippocampal neurones in response to oscillatory input currents
J Physiol
 , 
2000
, vol. 
529
 
Pt 1
(pg. 
205
-
213
)
Pilcher
WH
Ojemann
G
Apuzzo
MLJ
Presurgical evaluation and epilepsy surgery
Brain surgery: complication avoidance and management
 , 
1993
New York
Churchill Livingstone
(pg. 
1525
-
1555
)
Pletzer
B
Kerschbaum
H
Klimesch
W
When frequencies never synchronize: the golden mean and the resting EEG
Brain Res
 , 
2010
, vol. 
1335
 (pg. 
91
-
102
)
Raghavachari
S
Kahana
MJ
Rizzuto
DS
Caplan
JB
Kirschen
MP
Bourgeois
B
Madsen
JR
Lisman
JE
Gating of human theta oscillations by a working memory task
J Neurosci
 , 
2001
, vol. 
21
 (pg. 
3175
-
3183
)
Raghavachari
S
Lisman
JE
Tully
M
Madsen
JR
Bromfield
EB
Kahana
MJ
Theta oscillations in human cortex during a working-memory task: evidence for local generators
J Neurophysiol
 , 
2006
, vol. 
95
 (pg. 
1630
-
1638
)
Ray
S
Crone
NE
Niebur
E
Franaszczuk
PJ
Hsiao
SS
Neural correlates of high-gamma oscillations (60–200 Hz) in macaque local field potentials and their potential implications in electrocorticography
J Neurosci
 , 
2008
, vol. 
28
 (pg. 
11526
-
11536
)
Roopun
AK
Kramer
MA
Carracedo
LM
Kaiser
M
Davies
CH
Traub
RD
Kopell
NJ
Whittington
MA
Temporal interactions between cortical rhythms
Front Neurosci
 , 
2008
, vol. 
2
 (pg. 
145
-
154
)
Roopun
AK
Middleton
SJ
Cunningham
MO
LeBeau
FE
Bibbig
A
Whittington
MA
Traub
RD
A beta2-frequency (20–30 Hz) oscillation in nonsynaptic networks of somatosensory cortex
Proc Natl Acad Sci U S A
 , 
2006
, vol. 
103
 (pg. 
15646
-
15650
)
Siegel
M
Donner
TH
Engel
AK
Spectral fingerprints of large-scale neuronal interactions
Nat Rev Neurosci
 , 
2012
, vol. 
13
 (pg. 
121
-
134
)
Silva
LR
Amitai
Y
Connors
BW
Intrinsic oscillations of neocortex generated by layer 5 pyramidal neurons
Science
 , 
1991
, vol. 
251
 (pg. 
432
-
435
New York, NY
Singer
W
Gray
CM
Visual feature integration and the temporal correlation hypothesis
Annu Rev Neurosci
 , 
1995
, vol. 
18
 (pg. 
555
-
586
)
Stone
TW
Kynurenines in the CNS: from endogenous obscurity to therapeutic importance
Prog Neurobiol
 , 
2001
, vol. 
64
 (pg. 
185
-
218
)
Szabadics
J
Tamas
G
Soltesz
I
Different transmitter transients underlie presynaptic cell type specificity of GABAA,slow and GABAA,fast
Proc Natl Acad Sci USA
 , 
2007
, vol. 
104
 (pg. 
14831
-
14836
)
Szabadics
J
Varga
C
Molnar
G
Olah
S
Barzo
P
Tamas
G
Excitatory effect of GABAergic axo-axonic cells in cortical microcircuits
Science
 , 
2006
, vol. 
311
 (pg. 
233
-
235
New York, NY
Tallon-Baudry
C
Bertrand
O
Fischer
C
Oscillatory synchrony between human extrastriate areas during visual short-term memory maintenance
J Neurosci
 , 
2001
, vol. 
21
 pg. 
RC177
 
Theiler
J
Eubank
S
Longtin
A
Galdrikian
B
Doyne Farmer
J
Testing for nonlinearity in time series: the method of surrogate data
Nonlinear Phenomena
 , 
1992
, vol. 
58
 (pg. 
77
-
94
)
Tort
AB
Komorowski
R
Eichenbaum
H
Kopell
N
Measuring phase-amplitude coupling between neuronal oscillations of different frequencies
J Neurophysiol
 , 
2010
, vol. 
104
 (pg. 
1195
-
1210
)
Traub
RD
Whittington
MA
Colling
SB
Buzsaki
G
Jefferys
JG
Analysis of gamma rhythms in the rat hippocampus in vitro and in vivo
J Physiol
 , 
1996
, vol. 
493
 
Pt 2
(pg. 
471
-
484
)
Valiante
TA
Lozano
AM
Gildenberg
PL
Selective Amygdalohippocampectomy
Stereotactic and Functional Neurosurgery
 , 
2009
Springer DE
van Aerde
KI
Mann
EO
Canto
CB
Heistek
TS
Linkenkaer-Hansen
K
Mulder
AB
van der Roest
M
Paulsen
O
Brussaard
AB
Mansvelder
HD
Flexible spike timing of layer 5 neurons during dynamic beta oscillation shifts in rat prefrontal cortex
J Physiol
 , 
2009
, vol. 
587
 (pg. 
5177
-
5196
)
Velasco
AL
Wilson
CL
Babb
TL
Engel
J
Jr
Functional and anatomic correlates of two frequently observed temporal lobe seizure-onset patterns
Neural Plast
 , 
2000
, vol. 
7
 (pg. 
49
-
63
)
Wang
XJ
Neurophysiological and computational principles of cortical rhythms in cognition
Physiol Rev
 , 
2010
, vol. 
90
 (pg. 
1195
-
1268
)
Whittington
MA
Stanford
IM
Colling
SB
Jefferys
JG
Traub
RD
Spatiotemporal patterns of gamma frequency oscillations tetanically induced in the rat hippocampal slice
J Physiol
 , 
1997
, vol. 
502
 
Pt 3
(pg. 
591
-
607
)
Whittington
MA
Traub
RD
Jefferys
JG
Synchronized oscillations in interneuron networks driven by metabotropic glutamate receptor activation
Nature
 , 
1995
, vol. 
373
 (pg. 
612
-
615
)
Whittington
MA
Traub
RD
Kopell
N
Ermentrout
B
Buhl
EH
Inhibition-based rhythms: experimental and mathematical observations on network dynamics
Int J Psychophysiol
 , 
2000
, vol. 
38
 (pg. 
315
-
336
)
Wieser
HG
Engel
JJ
Williamson
PD
Babb
TL
Gloor
P
Engel
JJ
Surgically remediable temporal lobe syndromes
Surgical treatment of the epilepsies
 , 
1993
New York
Raven Press
Wu
C
Luk
WP
Gillis
J
Skinner
F
Zhang
L
Size does matter: generation of intrinsic network rhythms in thick mouse hippocampal slices
J Neurophysiol
 , 
2005
, vol. 
93
 (pg. 
2302
-
2317
)
Wu
C
Shen
H
Luk
WP
Zhang
L
A fundamental oscillatory state of isolated rodent hippocampus
J Physiol
 , 
2002
, vol. 
540
 (pg. 
509
-
527
)
Yaari
Y
Konnerth
A
Heinemann
U
Spontaneous epileptiform activity of CA1 hippocampal neurons in low extracellular calcium solutions
Exp Brain Res
 , 
1983
, vol. 
51
 (pg. 
153
-
156
)
Yamawaki
N
Stanford
IM
Hall
SD
Woodhall
GL
Pharmacologically induced and stimulus evoked rhythmic neuronal oscillatory activity in the primary motor cortex in vitro
Neuroscience
 , 
2008
, vol. 
151
 (pg. 
386
-
395
)
Zhang
L
He
J
Jugloff
DG
Eubanks
JH
The MeCP2-null mouse hippocampus displays altered basal inhibitory rhythms and is prone to hyperexcitability
Hippocampus
 , 
2008
, vol. 
18
 (pg. 
294
-
309
)
Zhang
ZJ
Koifman
J
Shin
SD
Ye
H
Zhang
L
Valiante
TA
Carlen
PL
Transition to seizure: ictal discharge is preceded by exhausted presynaptic GABA in the hippocampal CA3 region
J Neurosci
 , 
2012
, vol. 
32
 (pg. 
2499
-
2512
)
Zilles
K
Mendelsohn
FA
Paxinos
G
Codistribution of receptors in the human cerebral cortex
Receptors in the Human Nervous System
 , 
1991
Orlando, FL
Academic Press

Author notes

C.M.F. and R.J.McG. contributed equally to this work.