Shape features of epileptic spikes are a marker of epileptogenesis in mice




To identify reliable biomarkers for quantitatively assessing the development of epilepsy in brain.


In a kainate mouse model of temporal lobe epilepsy, we performed long-term video–electroencephalography (EEG) monitoring (several weeks) of freely moving animals, from kainic acid injection to chronic epileptic stage. Using signal processing techniques, we automatically detected single epileptic spikes (ESs), and we quantified the evolution of shape features during the epileptogenesis process. Using a computational model of hippocampal activity (neuronal population level), we investigated excitatory-related and inhibitory-related parameters involved in morphologic changes of ESs.

Key Findings

The frequency of ESs increases during epileptogenesis. Regarding shape features, we found that both the initial spike component and the wave component of opposite polarity of ESs gradually increase during epileptogenesis. These very specific alterations of the shape of ESs were reproduced in a computational physiologically relevant neuronal population model. Using this model, we disclosed some key parameters (related to glutamatergic and γ-aminobutyric acid [GABA]ergic synaptic transmission) that explain the shape features of simulated ESs. Of interest, the model predicted that the decrease of GABAergic inhibition is responsible for the increase of the wave component of ESs. This prediction (at first sight counterintuitive) was verified in both in vivo and in vitro experiments. Finally, from aforementioned electrophysiologic features, we devised a novel and easily computable index (wave area/spike amplitude ratio) indicative of the progression of the disease (early vs. late stage).


Results suggest that dendritic inhibition in hippocampal circuits undertake dramatic changes over the latent period. These changes are responsible for observed modifications in the shape of ESs recorded in local field potential (LFP) signals. The proposed index may constitute a biomarker of epileptogenesis.

Both diagnostic and therapeutic procedures would considerably benefit from the discovery of reliable biomarkers in epilepsy, starting from the severity of the disease up to the quantitative assessment of its development in the brain (Engel, 2011). Epileptic spikes (ESs) are transient signals (a few hundred msec) that have long been recognized as electrophysiologic markers of pathologic processes occurring in the epileptic brain (Matsumoto & Ajmone-Marsan, 1964). At chronic stage (defined by the recurrence of seizures), interictal ESs are often observed not only in human partial epilepsies but also in most experimental models of focal epilepsy (Schwartzkroin & Wheal, 1984). A number of studies conducted in experimental models have also reported the appearance of ESs during epileptogenesis (defined as the structural and functional changes leading a normal brain to produce recurring seizures), during the latent period (time from initial insult to spontaneous seizures) (Avoli et al., 2006; Staley & Dudek, 2006). Two types of ESs are usually encountered in electrophysiologic data: sustained discharges of ESs and single (or isolated) ESs. For the former type, it has been established that the frequency increases as epilepsy progressively develops (Riban et al., 2002; Heinrich et al., 2011). More recently, some shape changes were also reported (Chauvière et al., 2012). For single ESs, a recent report also demonstrated that the frequency increases (White et al., 2010). However, the evolution of their morphologic features during epileptogenesis has not yet been studied in detail.

It is commonly admitted that single ESs are polymorphic events (Alarcon et al., 1994). Nevertheless, to a large extent, these events usually present with typical shape features: a more or less sharp initial component (often referred to as the “spike”) followed by a more or less prominent slower component of opposite polarity (often referred to as the “wave”). So far, the potential value of these particular shape features as markers of underlying cell-related and network-related changes occurring in neuronal networks during epileptogenesis has never been described.

In this study, we quantified the occurrence frequency and shape features of ESs, and we analyzed the long-term changes in an experimental model of mesial temporal lobe epilepsy (MTLE) (Suzuki et al., 1995). Specifically, using signal processing techniques (automatic detection and characterization), we showed that the shape features of ESs evolve concomitantly with the progression of epilepsy. To provide a pathophysiologic interpretation for this evolution, we combined a computational modeling approach with in vivo and in vitro experiments. Using this approach, we showed that the progressive alteration of phasic γ-aminobutyric acid (GABA)ergic inhibition over the latent period (typically 2–3 weeks) leads to specific and, at first sight unexpected, changes in the morphology of ESs. Therefore, we established a link between changes observed at the level of local field potential (LFP) signals and underlying functional alterations occurring within the neuronal circuits of the hippocampus predominantly involved in seizures at chronic stage. Finally, a novel index derived from the morphologic features of ESs and predictive of the disease progression is proposed.


In vivo experiments

Experiments were conducted in the kainate mouse model of epilepsy (Suzuki et al., 1995) in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC). Readers may refer to the Data S1 for details about in vivo experimental procedures. Twenty-four, 80 ± 5-day-old C57BL/6J male mice were used for this study (3 controls + 21 animals that received intrahippocampal injection of kainic acid). Mice were separated into the following three protocols: (1) monitoring (n = 9) of LFPs during epileptogenesis, (2) LFP recordings (n = 12) for testing the predictive value of the wave-area to spike-amplitude ratio (WA/Sa), and (3) in vivo assessment (n = 3) of computational model predictions about changes in epileptic spikes (ESs).

Monitoring of epileptogenesis

To quantify LFP signals during epileptogenesis, video–electroencephalography (EEG) monitoring was performed in six kainic acid (KA)–treated mice and three control mice at regular intervals (3, 6, 9, 12, 15, 18, 21, 24, 27, and 30 days after surgery). Recording sessions were performed between 1 p.m. and 5 p.m. in freely moving animals. The total duration of the video-EEG analyzed in this study was 270 h (KA-treated mice: 180 h; saline-treated mice: 90 h).

Prediction of the epileptogenic state using the WA/Sa ratio

To test the predictive value of the WA/Sa ratio, 12 mice provided from 6 different inbred strains (C57BL/6J, 129/SvTer, DBA/2J, BR/Orl, FVB/N, and CBA/H) were recorded one to four times between 2 and 28 days after a KA intrahippocampal injection. This test was performed with mice derived from six inbred strains, and not only with C57BL/6J mice for the following reason: working with several mice issued from one single inbred strain can be considered as working on a single individual that is replicated indefinitely. In fact, an inbred strain represents a unique individual among a whole population (species). Under no circumstances, should a large number of mice from the same inbred strain represent a population. Only a great number of inbred strains can be considered as representing a population of individuals. Therefore, in our case, testing several inbred strains allowed us to check whether the predictive value WA/Sa ratio can be generalized to the mouse species, that is, is not restricted only to the C57BL/6J strain.

Randomly chosen recordings (1-h duration) were processed (computation of the WA/Sa ratio) with the automatic method in a blinded fashion (i.e., without a priori knowledge of the epileptic status of the animal). The WA/Sa ratio was plotted a posteriori against the epileptogenic state (early or late stage). We used the presence of hippocampal paroxysmal discharges (HPDs) to differentiate the early stage (no HPDs) and the late stage (HPDs) of epileptogenesis. These HPDs are described as nonconvulsive seizures that consist of sharp waves (3–5 Hz) followed by a sustained rhythmic activity (10–14 Hz) that lasts at least 15 s (Riban et al., 2002).

In vivo assessment of computational model prediction

To assess the computational model prediction about morphologic changes of ESs, three KA-treated mice were recorded for video-EEG during 5 h between 6 and 9 weeks after surgery. Three hours after the beginning of this recording, they were injected with a solution of picrotoxin (2 mg/kg, i.p.; Sigma-Aldrich, Saint-Louis, MO, U.S.A) in 0.9% NaCl.

Epileptic spike (ES) detection

LFP signals were first visually inspected, with video control, in order to separate the exploratory behavior from resting state of the animals. An automatic detection method was designed to be sensitive and specific to the occurrence of single ESs (Fig. 1A). In brief, it is based on the detection of abrupt changes in the mean of a random quantity (CUSUM algorithm, Hinkley, 1970). These jumps correspond to transient increases of the signal energy in the 30–70 Hz frequency sub-band that characterize single ESs. It could be verified that the detector was not sensitive to sharp wave discharges and HPDs, which are characterized by a different time-frequency signature.

Figure 1.

Single epileptic spikes (ESs) during epileptogenesis. (A) Typical ES recorded in the mouse (27 days after kainic acid – KA – injection) occurring 4 s after a sharp wave discharge (dashed line). Right: an example of interictal ES recorded from hippocampus with a depth electrode (monopolar) in a patient with mesial temporal lobe epilepsy. (B) Top: evolution of ESs automatically detected in one KA-treated mouse during 1 month of monitoring. All detected events are superimposed and plotted in gray, whereas the mean signal is represented by a black solid line. Bottom, events detected in a control mouse. Recordings with no event detected are denoted by NED. (C) Occurrence rate (mean ± standard error of the mean [SEM]) of detected events in KA-treated (black dots, n = 6) and in control (white dots, n = 3) mice. (D) Total amplitude (difference between the minimum and the maximum) of ESs detected in KA-treated mice. Bottom of (C, D) shows the time where the first HPDs appear (only in KA-treated mice).

Shape features extraction

As presented in Fig. 2A, two features were calculated from in vivo ESs detected during epileptogenesis: the Spike amplitude (Sa, mV) and the Wave Area (WA, mV·msec). To avoid confusion between “amplitude” and “Area” in both abbreviations, we used a lower case “a” to denote the amplitude and an upper case “A” for the Area. The Sa is the difference between the peak of the ES and the baseline, whereas the WA is the absolute value of the sum of signal samples during the wave component. These features were calculated with an automatic processing method implemented in MATLAB for both in vivo real ESs and simulated ESs.

Figure 2.

Morphologic changes of ESs during epileptogenesis. (A) Quantified features of ES shape: Spike amplitude (Sa) defined from baseline to peak (blue) and Wave Area (WA) defined as the absolute value of the signal integral as a function of time during the wave component (green). (B) Evolution of the Sa (mean ± SEM) during epileptogenesis. (C) Evolution of the WA (mean ± SEM) during epileptogenesis. (D) Evolution of the WA/Sa ratio during epileptogenesis. (E) Left: WA/Sa ratio values (mean ± SEM) computed on 27 recordings performed in 12 mice. The WA/Sa ratio is indicative of the progression of the disease (early: yellow dots; late: red triangles). Best separation of distributions is obtained for a threshold value approximately equal to 25. Right, boxplots for both stages, middle line is the median. Nonparametric Mann-Whitney U test, **p < 0.01.

In vitro experiments

The computational model prediction about ES changes was also assessed in vitro, using field recordings on organotypic hippocampal slices superfused with bicuculline (Tocris, 0, 10, or 50 μm). Slice cultures were prepared as described previously (Gahwiler et al., 1997) and briefly in the Data S1. Experiments were conducted in accordance with Swiss law and the Ethics Committee of the Veterinary Department of the Canton of Zurich. To impose different degree of excitability of the slice, low-Mg2+–modified artificial cerebrospinal fluid (ACSF; 0.3 mm) combined with different concentrations of GABAA receptor antagonist (bicuculline: 0; 10; 50 μm) were superfused. Procedures for field recordings are described in Data S1.

Computational modeling

The conditions needed to reproduce the shape features of ESs as recorded in vivo during epileptogenesis were also studied in a biology-inspired computational model (Wendling et al., 2000), as described in the Data S1. A parameter sensitivity analysis was performed to (1) determine necessary conditions to reproduce, in the model, observed signals and (2) uncover parameters that have the strongest effect on the shape of ESs. This procedure consisted in extensive simulations of ESs when model parameters were varied. For each simulated ES, selected features (spike amplitude and wave area) were computed using the same routines as those used to quantify actual ESs. Finally, results were plotted as color-coded maps displaying shape features versus model parameters.

Modeling at the neuronal population level was chosen for two reasons. Firstly, this level allows for direct comparison of model output (summation of postsynaptic potentials generated at the level of pyramidal cells) with actual LFPs. Secondly, this level of modeling is deep enough to provide insights into the relationship between excitation-related and inhibition-related model parameters and the shape features of ESs observed in LFPs.


Assessment of epilepsy, hippocampal sclerosis, and LFP recording site

In all mice injected with KA, histologic data of brain slices revealed a diminution of the hippocampal volume, a loss of pyramidal cells (mainly in CA1 and in the hilus), and a dispersion of granule cells of the dentate gyrus (DG). As shown in Fig. S1A,B, reorganization patterns remained variable from one mouse to another. Figure S1B shows the trace of the bipolar electrode and the recording sites where single epileptic spikes (ES) are observed. In most cases, the polarity of ESs was found to be reversed on the electrode contacts located above or below the pyramidal layer of CA1, indicating that ESs are recorded from CA1 pyramidal cells. In addition, all ESs recorded from a given mouse kept the same polarity along the whole epileptogenesis process.

The six mice injected with KA developed epilepsy, since they presented recurrent hippocampal paroxysmal discharges (HPDs), typical of this model (Riban et al., 2002). We also recorded from one to three convulsive seizures in four of them. Conversely the three control subjects (which received no KA) remained free of HPDs and convulsive seizures.

Occurrence frequency of ESs during epileptogenesis

An automatic spike detection method was applied to LFP signals recorded from the right hippocampus of nine mice, over a 30-days period after KA or NaCl (control) intrahippocampal injection. Altogether, 4,664 transient events were detected. Ninety-seven percent of these events (n = 4507) occurred in KA-treated mice (n = 6) and corresponded to ESs to a large extent. Typical detected events are shown in Fig. 1B. It is noteworthy that these sporadic events (1) strongly resemble interictal ESs recorded from hippocampus in human mesial temporal lobe epilepsy and (2) morphologically differ from hippocampal paroxysmal discharges HPDs. The remaining 3% events (n = 157) were detected in control animals (n = 3). As illustrated in Fig. 1B, ESs detected in KA-treated animals were characterized by a stereotyped shape (sharp spike followed by a slow wave of opposite polarity). The occurrence frequency of these ESs was found to increase considerably during epileptogenesis (Fig. 1C), as recently reported in another experimental model (White et al., 2010). In addition, and strikingly, the shape of these ESs was found to progressively change during epileptogenesis with a noticeable increase of amplitude (Fig. 1D).

Shape changes of ESs during epileptogenesis

Distinct cellular processes are likely to underlie the spike (AMPAergic: α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) and the wave (GABAergic: γ-aminobutyric acid) components of ESs (de Curtis & Avanzini, 2001; Demont-Guignard et al., 2009, 2012). Starting from this consideration, we used separate features, namely the spike amplitude (Sa) and the wave area (WA) (Fig. 2A), to quantify shape changes of ESs during epileptogenesis. Results revealed a continuous and almost linear (Sa: r2 = 0.99, WA: r2 = 0.98) increase of both the spike amplitude and the wave area (Fig. 2B,C). As expected, these two tendencies were found to be highly correlated (r2 = 0.99). Of interest, the WA increase (578%) was found to be much higher than the Sa increase (274%). This observation led us to take account of the evolution of the WA/Sa ratio. Moreover, this ratio has the advantage to be independent from the absolute amplitude values of each component. As expected, this ratio was found to increase nonlinearly with epileptogenesis (Fig. 2D). More importantly, a blind test was performed consisting of measuring WA/Sa ratio values in randomly selected recordings of KA-treated animals (n = 12, distinct strains, WA/Sa values averaged over 1 h). As illustrated in Fig. 2E, results showed that statistically significant higher WA/Sa ratio values corresponded to animals in which the epileptogenic process was more advanced (early vs. late stage, right plot). Precisely, we could determine that ESs with a WA higher (respectively lower) than 25 times the Sa are characteristic of the late (respectively early) stage.

To interpret these changes in terms of underlying mechanisms, a macroscopic but still physiologically relevant computational model was used to mimic CA1 hippocampal ESs recorded in vivo.

Model predictions about observed changes in the shape of ESs

As shown in Fig. 3A, and in accordance with previous reports (Zetterberg et al., 1978; Wendling et al., 2000), the neuronal population model could simulate transient signals, the time course of which closely resembles that of actual ESs. This model gives access to noneasily measurable variables like the firing rate of both pyramidal cells (P) and interneurons (I) that can be plotted simultaneously with the simulated LFP (approximated as the summation of average postsynaptic potentials (PSPs) at the level of pyramidal cells) (Fig. 3B). We performed a parameter sensitivity analysis to assess the relationship between the shape features of simulated ESs and the four main parameters of the model, namely CA3-P (excitatory drive from CA3), P-P (collateral excitation among pyramidal cells), P-I (excitatory drive of pyramidal cells on interneurons), and I-P (feedback inhibitory drive of interneurons on pyramidal cells). Results are displayed as color-coded maps in Fig. 3C. Regarding the amplitude of simulated ESs, the analysis revealed that parameter CA3-P plays the key role, whereas the three other parameters have almost no effect. Regarding the wave area (WA), results showed that both parameters CA3-P and I-P have a major influence. To analyze the specific contribution of these two parameters, we superimposed the WAs as quantified on simulated ESs and on actual ESs (Fig. 3D). Of interest, this comparison demonstrated that two conditions are necessary in the model to obtain the best fit between simulated and experimental data: increase of parameter CA3-P and decrease of parameter I-P. In other words, the observed augmentation of the WA can be accurately reproduced if, and only if, the inhibitory drive of interneurons on pyramidal cells is diminished. As discussed later, this finding was at first sight counterintuitive, as the wave component is presumably mainly GABAergic. This result is illustrated in Fig. 3E, where parameters CA3-P and I-P were progressively modified, leading to noticeable changes in the shape of simulated ESs, as observed in vivo. It is worth noting that a second wave component (same polarity as the spike) also appears in the model for increased excitability. Finally, we found that the diminution of parameter I-P has also a major effect on the time to ictal-like activity in the model, as the unstable behavior is more rapidly reached for reduced inhibitory feedback (Fig. 3F).

Figure 3.

Computational model predictions about ES shape changes. (A) Schematic diagram of the neuronal population model (CA1 subfield). Two neuronal subpopulations are represented in the model: pyramidal cells (P) and inhibitory interneurons (I). Synaptic transmission include three excitatory (AMPAergic) connections: CA3-P (input from CA3), P-P (collateral excitation), P-I (excitatory input to interneurons) and one inhibitory (GABAergic) connection: I-P (inhibitory input to pyramidal cells). Right: top trace is example of ES-like event simulated by the model. Bottom trace is real ES recorded from CA1 in a KA-treated mouse. (B) Top: time course of the average excitatory PSP and inhibitory PSP during the simulation of an ES (background activity was removed). Bottom: average firing rates of P and I subpopulations. (C) Effects of the four synaptic gain parameters on shape features. These four-dimensional maps present the value of features Sa (left) and WA (right) computed on simulated events. Small maps reveal the influence of both I-P (vertical) and CA3-P (horizontal) parameters. Stepping from one map to another reveals the influence of the P-I (vertical) and P-P (horizontal) parameters. (D) Comparison between the WA evolution in the computational model and in the in vivo model. Colored lines show the WA evolution in the model for an increase of parameter CA3-P (excitatory drive, 100–400%) at different decreasing levels of parameter I-P (inhibitory, 40–100%). Previously shown WA values (black dots) computed on in vivo data are superimposed on simulated data (colored lines). The increase of CA3-P parameter is tuned to reproduce the increase of the Sa observed in vivo. (E) Effect of I-P (green arrow) diminution on the shape of simulated ES. Top: model configuration; bottom: corresponding simulated signals. (1) I-P = 100%. (2) I-P = 75%. (3) I-P = 50% (%-ages given with respect to initial values used in configuration 1). The consequence is a retroactive increase of the excitatory drive onto interneurons (+, ++). (F) Effect of parameter I-P on the model ability to generate ictal-like activity obtained by progressively increasing the level of the excitatory input. The time to reach the unstable behavior (ictal-like activity) is reduced for reduced I-P value.

Experimental assessment of the computational model prediction

In order to verify the model prediction regarding the influence of parameter I-P on the WA, specific experiments were conducted. First, in KA-treated animals (showing frequent sporadic ESs), the intraperitoneal injection of the GABAA receptor antagonist picrotoxin induced no consistent changes of the spike amplitude. In contrast, this protocol induced a reproducible and highly significant increase of the wave area (Fig. 4A), as observed in the computational model. To complement these results and to get closer to the computational model conditions, in vitro data were recorded from organotypic rat hippocampal slices under low-magnesium modified ACSF containing either 10 μm or 50 μm bicuculline (GABAA receptor antagonist). Results showed that this protocol could induce spontaneous ESs in the CA1 subfield (LFP, stratum radiatum) in a dose-dependent manner (Fig. 4B). With higher bicuculline concentration, the wave area was twice as high compared to the lower concentration, whereas the spike amplitude was not significantly modified. ESs were not observed when control ACSF (n = 11) was used.

Figure 4.

In vivo and in vitro validation of the model prediction about the inhibitory mechanisms underlying the WA increase. (A) In vivo results. Changes of shape features in ESs after picrotoxin (PTX) systemic injection in KA-treated mice (n = 4). Mean ± SEM of spike amplitude (Sa, Left) and Wave Area (WA, Right) values normalized to Sa or WA values before PTX injection (100%). Paired Student t-test, *p < 0.05. (B) In vitro results. Changes of shape features observed in rat organotypic hippocampal slices for two concentrations of bicuculline (10 and 50 μm). Right: bar plots show mean ± SEM of the Sa and the WA for low (n = 4 slices) and high (n = 3 slices) bicuculline concentration. Student t-test, *p < 0.05.


Epileptic spikes have long been associated with epileptogenic processes. Over the last decades, many aspects of the quantification of these transient events have been dealt with including automatic detection (Gotman, 1999), counting (Spencer et al., 2008), shape characterization (Wadman et al., 1983), and source localization (Michel et al., 2004). Most of these studies were performed in the context of “mature” epileptogenic networks, that is, in patients or animals with chronic epilepsy. In the present report, the context is different. The objective was to quantify the evolution of ESs during epileptogenesis, as these events occur at a very early stage after the initial insult and before the appearance of recurrent seizures.

So far, two studies have reported a detailed analysis of ESs over prolonged periods after brain injury in rodent models. In the first one (White et al., 2010), in rat, the temporal features of spikes (occurrence frequency and pattern, i.e., isolated vs. clusters) were hypothesized as a potential predictive biomarker for the development of chronic epilepsy after KA-induced status epilepticus (systemic injection). Our study corroborates these results, as we also observed a significant increase of the frequency of ESs during epileptogenesis. However, it goes beyond, since we also found that the shape features of ESs (1) convey highly relevant information about underlying neuronal alterations and (2) can be used to derive an index indicative of the progress of the disease after injury. In the second publication (Chauvière et al., 2012), sustained discharges of spikes were shown to contain two types (type 1 and type 2) of events that differ depending on the presence (type 1) or absence (type 2) of a wave following the spike component. Authors showed that the number of type 1 spikes progressively decreases while the number of type 2 spikes increases until the occurence of the first spontaneous seizure. At first sight, our results are not in agreement with these findings, since we observed a gradual increase of the wave component with time. However, it is noteworthy that mechanisms underlying hippocampal paroxysmal discharges and single ESs are likely to be dissimilar, since both events correspond to distinct electrophysiologic patterns (Riban et al., 2002; Maroso et al., 2011).

Conversely to (Chauvière et al., 2012), in our experimental protocol, animals were not monitored continuously all along the epileptogenesis process, as we did not consider as crucial the fact to accurately identify the occurrence time of the very first seizure. Actually, epileptogenesis is described as an evolutive process that leads the brain to generate recurrent seizures. However, this process cannot be limited strictly to the period that precedes the first seizure (Pitkänen, 2010). Indeed, in some cases, this period can be very short (about 3–6 days postinjection), and many structural/functional alterations related to epileptogenesis are likely to occur after this time (Heinrich et al., 2011). To us, there is no more reason to believe that the first seizure is the marker of a new step in the epileptogenesis process than to think that it is the first detectable paroxysmal event in a more continuous process. Our findings on the evolution of the WA/Sa ratio suggest that, in this in vivo model, the epileptogenesis process does not stop after the first seizure occurrence (2 weeks after injection, on average), but continues as the disease becomes more severe (at least until 4 weeks).

Single ESs have a stereotyped shape characterized by an initial spike component followed by a wave component of opposite polarity. Our results showed that these events appear at an early stage of epileptogenesis and that the two aforementioned components gradually change over days and weeks that follow the initial lesion induced by the local microinjection of KA. From these electrophysiologic markers, we devised a novel index indicative of the progression of the pathology. This index is the Wave Area/Spike amplitude ratio (WA/Sa). Indeed, neither the Sa nor the WA can be taken as a standalone marker. A normalization is needed to account for recording conditions (like the electrode position, size, and impedance). For instance, in the same animal, the electrode impedance can vary over time (increase due to gliosis). Any impedance change would directly affect the ES amplitude and consequently both the Sa and the WA. The use of the ratio (WA/Sa) instead of the absolute quantities (either the Sa or the WA) has the enormous advantage of making the WA/Sa index independent from biophysical factors that can considerably affect the LFP amplitude. It allows for comparison of ESs (1) in the same animal at different stages and (2) across different animals. We finally tested this index using a double-blind procedure that showed that this index is highly correlated with the epileptogenesis progression. This predictive value of the WA/Sa index may be of broad interest, as results were obtained over different inbred strains.

Using a combined computational/experimental approach, we have also provided a pathophysiological interpretation for this biomarker. Indeed, results revealed the WA/Sa ratio increases as a function of time, since the WA increases more rapidly than the Sa. We first investigated the conditions for which we could reproduce these ES shape changes in a macroscopic neurophysiology-inspired computational model of neuronal population. Although minimalist (two subpopulations of neurons interacting via synaptic transmission), this model gave insights into key parameters. In particular, the excitatory drive on pyramidal cells explained the Sa increase and partly the WA increase. However, to achieve a closer matching of the simulated WA increase with real data, results also showed that the inhibitory drive (phasic GABAergic dendritic inhibition) on pyramidal cells should be reduced. Indeed, in the model, this drop of inhibition leads to increased excitation of pyramidal cells and therefore to increased excitatory drive on interneurons. The net effect is an augmentation of the gain in the negative feedback loop leading to a more pronounced wave following the spike component in the simulated ESs.

As this prediction was not intuitive (the ES wave is often attributed to a GABAergic effect), we carried out both in vivo and in vitro experiments to specifically test the impact, on the shape features of ESs, of modified GABAergic inhibition onto pyramidal cells. Both experimental approaches confirmed that the reduction of GABAA-receptor–mediated inhibition leads to a significant increase of the ES wave component, as predicted by the model. Consequently, our results strongly suggest that the ES wave component does indeed have a GABAergic component (conversely to a previous report, Newberry & Nicoll, 1984 in which this hypothesis was rejected under the argument that bicuculline did abolish the wave), but they also provide an explanation to an apparently paradoxical situation where this wave component increases when GABAA antagonists are being used.

According to these results, epileptogenesis is a pathologic process in which one of the main factor is the progressive erosion of the inhibitory drive onto pyramidal that is gradually altered over a week time scale. This finding corroborates a number of studies showing a decrease of GABAergic inhibition, mainly in focal temporal epilepsy (Ben-Ari & Dudek, 2010). In particular, two main processes could account for the reduction of inhibitory drive. First is the progressive loss of interneurons in epileptic tissue (Sloviter, 1987). Indeed, some dendritic-projecting GABAergic interneurons have been shown to degenerate, which results in a permanent reduction of inhibitory drive to principal neurons (Cossart et al., 2001). Second is the shift of GABAA reversal potential, because of the drop of potassium chloride cotransporter 2 (KCC2) transporter expression at the membrane, leading to depolarizing GABA in human tissue (Cohen et al., 2002; Huberfeld et al., 2007) and during epileptogenesis in animal models of temporal lobe epilepsy (Staley, 2008).

Although several preclinical studies have shown effects of novel treatments (Pitkänen & Lukasiuk, 2011), it remains unclear whether one can prevent epileptogenesis after brain trauma, regardless of the strategy being used to “interfere” with pathologic processes which gradually take place in the underlying neuronal systems. In this study, we propose an index related to the shape of ESs, which provides information about the progress of the epileptogenic process. An evident advantage of this biomarker is that ESs are easily and routinely recorded in clinics (intracranial EEG and, to some extent, scalp EEG data). In this respect, ESs differs from pathologic high frequency oscillations (pHFOs) (Jefferys et al., 2012), which could also constitute a biomarker but which detectability decrease with electrode size (Demont-Guignard et al., 2012).

Finally, as a perspective to this work, we propose to combine the WA/Sa index with other modalities like in vivo structural imaging (Nehlig, 2011). Indeed, this combination could be efficiently used to quantify the impact of “preventive” strategies aimed at slowing down or even stopping the epileptogenic process, ultimately.


This work was supported by “Region Bretagne” (CREATE 2009, “EPIGONE” project). Gabriel Dieuset holds a fellowship grant from UCB Pharma S.A. (France). We thank Prof. Urs Gerber (team leader, Brain Research of Zurich) for his invitation to Dr P Benquet to perform patch clamp experiments in his lab and L. Spassova and D. Göckeritz-Dujmovic for organotypic hippocampal slice cultures preparation.


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


  • Image of creator

    Clément Huneau PhD, postdoctoral researcher in computational and experimental neuroscience.