In vivo depth EEG recordings
Six adult rats were anaesthetized with a mixture of ketamine (1 mg kg−1) and xylazine (0.5 mg kg−1). One bipolar steel hippocampal electrode (3.5 mm posterior, 2.5 mm lateral and 3.7 mm deep with respect to bregma) and four monopolar stainless-steel cortical electrodes (screwed on the skull, above the right and left frontal and occipital cortex) were stereotaxically implanted. After 1 week recovery, animals were continuously recorded with a video-EEG system (Deltamed, Paris, France) 3 days before SE (control period) and over 1 month after SE. Animals were housed in individual cages; water and food were available ad libitum. At the end of the recording time, animals were anaesthetized as above and intracardiacally perfused with 4% paraformaldehyde, and the position of the hippocampal electrode was checked histologically. Its position within the CA1 region of the hippocampus was confirmed.
Electroencephalogram was analysed post hoc. For each rat, it was divided into 86 400 contiguous epochs, each of 30 s, for a total duration of 24 h before (control period), during the latent period, and up to 10 days after the first spontaneous seizure (chronic period). Within each epoch, sharp waves were visually detected. They were considered as epileptic sharp waves if their amplitudes were more than twice the amplitude of the biggest sharp waves recorded during the control period. Epochs containing movement artefacts (confirmed using the video that was recorded concomitantly with the EEG) were removed. Epileptic seizures were characterized by an initial increase in the sharp wave frequency, followed by a rapid low-voltage activity that evolved into sustained high-frequency oscillations and high-voltage polyspikes. The first spontaneous seizure involved the limbic structures first, since motor behaviour was either absent or occurred more than 5 s after the beginning of the electrographic seizure (not shown). Data are presented as means ± s.e.m and analysed with Student's paired t test.
In vitro electrophysiology
Hippocampal slices (380 μm thick) were prepared according to local INSERM guidelines from sham, latent and chronic animals, which were anaesthetized by i.p. injection of chloral hydrate (800 mg kg−1). Animals were perfused intracardially with cold artificial cerebrospinal fluid (ACSF) in which NaCl was substituted with an equimolar concentration of choline. Animals were then killed by decerebration, and slices were cut in modified ACSF. Slices were then transferred to a holding chamber at room temperature in normal ACSF. Artificial CSF contained (mm): NaCl, 126; KCl, 3.5; CaCl2, 2; MgCl2, 1.3; NaH2PO4, 1.2; NaHCO3, 26; and d-glucose, 10; and was continuously aerated with 95% O2 and 5% CO2 (pH 7.3). Neurons were visualized by infrared video microscopy using an upright Leica DM LFS microscope equipped with a ×40 objective (Leica, Bensheim, Germany) or a Zeiss FSII microscope equipped with a ×60 objective (Zeiss, Jena Germany). Patch pipettes were pulled from borosilicate glass tubing (2.0 mm outer diameter, 0.5 mm wall thickness) and filled with internal solutions containing (mm): Cs-gluconate, 135; MgCl2, 10; CaCl2, 0.1; EGTA, 1; Na2- adenosine triphosphate, 2; Hepes, 10; and 0.5% biocytin, pH 7.25. For whole-cell somatic and dendritic recordings, the pipette resistance was 3–8 and 6–12 MΩ, respectively. Uncompensated series resistances were 6–30 (soma) and 15–35 MΩ (dendrites). Series resistances and membrane properties were not different between sham, latent and chronic animals (not shown), as previously reported (Bernard et al. 2004). Access resistance and holding current were continuously monitored for stability; a 20% variation led to a rejection of the experiment. The recording temperature was 32–34°C. Signals were fed to an EPC9 (HEKA, Heidelberg, Germany) or Multiclamp 700A (Axon Instruments), digitized (10 kHz) with a Labmaster interface card to a personal computer and analysed with MiniAnalysis 5.1 program (Synaptosoft, Decatur, GA, USA). Spontaneous GABAA receptor-mediated currents (IPSCs) were measured at the reversal potential for glutamatergic events (+10 mV). Bicuculline, a GABAA receptor antagonist, was applied at the end of the experiments to verify that the currents were indeed GABAergic, as described (Cossart et al. 2001). Spontaneous glutamatergic currents (EPSCs) were measured at the reversal potential for GABAA receptor-mediated events (−60 mV) and were blocked by 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and d-aminophosphonovalerate (d-APV). Single events recorded (∼200 events per cell during a continuous recording session) were fully characterized: rise times (10–90%), amplitudes and decay time constants were calculated using MiniAnalysis 5.1. More than 95% of the synaptic events were fitted with a single exponential decay. In some instances, the frequency of sIPSCs was high enough for individual synaptic events to overlap most of the time. The analysis was performed on events that could be clearly isolated, which involved longer duration recordings.
Slices were processed for the detection of biocytin-filled neurons. They were fixed overnight at 4°C in a solution containing 4% paraformaldehyde in 0.12m phosphate buffer (PB, pH 7.4). After fixation, slices were rinsed in PB, cryoprotected in sucrose and quickly frozen on dry ice. To neutralize endogenous peroxidase, slices were pretreated for 30 min in 1% H2O2. After several rinses in saline potassium phosphate buffer (0.01 m KPBS, pH 7.4), slices were incubated for 24 h at room temperature, in 1:100 avidin–biotin peroxidase complex (Vector Laboratories, Inc., Burlingame, CA, USA) diluted in KPBS containing 0.3% Triton X-100. After 30 min rinses in KPBS, slices were processed with 0.04% 3,3′- diaminobenzidine-HCl (DAB; Sigma, St Louis, MO, USA) and 0.006% H2O2 diluted in KPBS. Based on their characteristic morphological features, neurons were morphologically identified post hoc as CA1 pyramidal cells. All dendritic recordings were performed farther than 300 μm from the soma (range, 300–450 μm). The recording sites in the dendrites appeared as holes (Fig. 3A) or dents, and were measured post hoc at a distance of 320 ± 28 (sham, n= 8), 342 ± 47 (latent, n= 6) and 335 ± 72 μm (chronic, n= 8) from the soma, i.e. close to the stratum radiatum–lacunosum moleculare border. CA1 pyramidal cells did not display axonal sprouting during the latent period (not shown), as reported in the kainic acid model (Smith & Dudek, 2001). Sprouting was present in epileptic animals, as reported previously (Esclapez et al. 1999; Smith & Dudek, 2001).
Figure 3. Persistent increase of the excitatory drive during the late latent period and gradual erosion of the GABAergic drive in CA1 pyramidal cell dendrites A, photomicrograph of a biocytin-filled pyramidal cell during the late latent period. The recording site (hole) is shown in the inset. Abbreviations: O, stratum oriens; P, stratum pyramidale; R, stratum radiatum; LM, stratum lacunosum moleculare; and M, molecular layer of the dentate gyrus. Scale bars represent 50 μm in main panel and 5 μm in the inset. B, traces are recordings of spontaneous EPSCs in dendrites in sham, late latent (cell shown in A) and epileptic animals. Note the increase in frequency during the late latent period, an increase that persists during the chronic period, as shown in the histogram on the right (*P < 0.002, **P < 0.01 versus sham). The graphs are cumulative probability plots of amplitude, 10–90% rise time, decay time constant and charge transfer of EPSCs in all recorded cells. Note the permanent increase in EPSC charge transfer during the late latent period as a result of the increase in amplitude, whilst rise times and decay time constants remain unaltered. Normalized average EPSCs are displayed in the inset. C, the traces at the top are recordings of spontaneous IPSCs in dendrites in sham, late latent (cell shown in A) and epileptic animals. Note the permanent decrease in frequency during the late latent period, as shown in the histogram on the right (***P < 0.05 versus sham, Student's unpaired t test). The graphs are cumulative probability plots of amplitude, 10–90% rise time, decay time constant and charge transfer of IPSCs in all recorded cells. There was no modification in the distribution of amplitudes and rise times, whereas decay time constants, hence IPSC charge transfers, were transiently increased during the late latent period. Normalized average IPSCs are displayed in the inset.
Download figure to PowerPoint
Computational modelling of field potential activity recorded in vivo during control, latent and chronic periods
Sufficient conditions to reproduce signals recorded during control, latent and chronic periods were studied in a neuronal population model described in detail previously (Wendling et al. 2002). Modelling at the neuronal population level was chosen for two reasons. First, this macroscopic level allows for direct comparison of model output (average postsynaptic activity generated by the subset of pyramidal cells) with field activity recorded in vivo and arising from large assemblies of cells. Second, this level of modelling is deep enough to provide insights into the relationship between excitation/inhibition-related model parameters and extracellular field activity recorded in vivo during control, latent and chronic periods.
The model design is based on experimental data describing the neuronal organization and connectivity of the CA1 region. It includes recurrent excitatory connections from pyramidal cells to pyramidal cells (Thomson & Radpour, 1991; Whittington et al. 1997). Pyramidal cells receive two types of GABA receptor-mediated currents: slow dendritic and faster perisomatic inhibitory postsynaptic currents (IPSCs). As previously proposed (Miles et al. 1996; White et al. 2000; Banks & Pearce, 2000; Banks et al. 2002), two separate classes of interneurons (possibly basket cells and dendrite-projecting interneurons, respectively, called, for simplicity, GABAA,fast interneurons and GABAA,slow interneurons) give rise to these two types of IPSCs. Both classes of interneurons interact, in that GABAA,slow cells inhibit not only pyramidal cells but also GABAA,fast interneurons (Banks et al. 2000).
The model was designed to represent this functional organization of interacting subsets of principal cells and interneurons. It consists in three subsets of neurons, namely the main cells (i.e. pyramidal cells), the slow dendritic-projecting inhibitory interneurons (GABAA,slow receptors) and the fast somatic-projecting inhibitory interneurons (GABAA,fast receptors). Interneurons receive an excitatory input (AMPA and NMDA receptor mediated) from pyramidal cells. The influence from neighbouring or more distant populations is represented by an excitatory input n(t) (modelled by a positive mean Gaussian white noise) that globally describes the average density of afferent action potentials.
In each subset, a linear transfer function is used to transform the average presynaptic density of afferent action potentials (the input) into an average postsynaptic membrane potential (the output). This transfer function models synaptic transmission in a simplified way but still takes into account both the amplitude and kinetics of synaptic responses. Its impulse response
determines the excitatory (E), dendritic inhibitory (DI) and somatic inhibitory (SI) average postsynaptic membrane potential, respectively. Lumped parameters EPSP, IPSPD and IPSPS define the amplitude of the average postsynaptic membrane potential. Although the exact relationship is not yet established, it can reasonably be assumed that these parameters aggregate both the rate and amplitude of spontaneous postsynaptic potentials (sPSP). Lumped parameters τE, τD and τS are time constant parameters that account for both the average decay time of sPSP and average distributed delays in the dendritic tree. In each subset, in turn, a static non-linear function (asymmetric sigmoid curve , where 2e0 is the maximum firing rate, ν0 is the post-synaptic potential corresponding to a firing rate of e0, and r is the steepness of the sigmoid) is used to model threshold and saturation effects in the relationship between the average postsynaptic potential of a given subset and the average pulse density of potentials fired by the neurons. Interactions between main cells and local neurons are summarized in the model by connectivity constants which account for the average number of synaptic contacts.
Finally, in order to model the electric extracellular field potential, we used the classical approximation based on the equivalent current dipole theory. As described by Lopes da Silva (2002), we assumed that the neuronal events that cause the generation of electric fields in the neuronal mass consist of ionic currents, the origin of which is mainly postsynaptic. Indeed, at the level of the single cell, synaptic activation of a neuron causes changes in membrane conductance that lead to the generation of primary currents through the membrane and cause a variation of the postsynaptic membrane potential. In both excitatory and inhibitory postsynaptic cases, extracellular currents (volume currents flowing in the surrounding medium) orientated in the same direction are generated. At the level of neuronal assemblies, these extracellular currents can only be measured by macroelectrodes at a distance from the sources if they sum together. This is typically the case for pyramidal cells, which are organized both in space and in time (parallel orientation and quasi-synchronous activation). These considerations allow us to approximate the electrical contribution of the whole neuronal population by an equivalent current dipole whose time-varying moment depends on the time variations of the average postsynaptic potential at the subset of pyramidal cells. Consequently, the temporal dynamics of the extracellular field potential activity seen by the electrode are represented, in the model output, by the summation of average excitatory and inhibitory postsynaptic potentials given, respectively, by the functions hE, hDI and hSI.
A parameter sensitivity study was conducted using an exhaustive procedure aimed at quantifying the frequency of interictal-like spikes detected in simulated signals as a function of the aforementioned model parameters. The amplitudes of glutamatergic average postsynaptic potentials (parameter EPSP) and GABAergic average postsynaptic potentials (parameters IPSPD and IPSPS) were varied step by step. At each step, a 15 s duration signal was simulated, in which the frequency of epileptic spikes was determined using an automatic detection method. The procedure was reiterated for increased average IPSP time constants (parameters τD and τS). All other parameters were kept constant. Finally, results were represented as colour-coded ‘activity maps’ to obtain a complete view of the model behaviour with respect to parameter variations. These maps were used to determine necessary conditions to reproduce, in the model, signals observed in control (normal background activity), latent and chronic periods (background activity mixed with interictal-like spikes). In addition, this procedure was carried out in a blind manner, i.e. without knowledge of the evolution of excitation and inhibition determined experimentally.