|Home | About | Journals | Submit | Contact Us | Français|
The subtle mechanisms of post-traumatic epileptogenesis remain unknown, although the incidence of chronic epilepsy after penetrating cortical wounds is high. Here, we investigated whether the increased frequency of seizures occurring within 6 weeks following partial deafferentation of the suprasylvian gyrus in cats is accompanied with a change in the ratio between the number of excitatory and inhibitory neurons. Immuno-histochemical labeling of all neurons with neuronal-specific nuclear protein (NeuN) antibody, and of the GABAergic inhibitory neurons with either gamma-aminobutyric acid (GABA) or glutamic acid decarboxylase (GAD 65&67) antibodies, was performed on sections obtained from control and epileptic animals with chronically deafferented suprasylvian gyrus. Quantification of the labeled neurons was performed in control animals and at 2, 4, and 6 weeks following cortical deafferentation, in the suprasylvian and marginal gyri, both ipsi- and contra-lateral to the cortical trauma. In all epileptic animals, the neuronal loss was circumscribed to the deafferented suprasylvian gyrus. Inhibitory GABAergic neurons were particularly more sensitive to cortical deafferentation than excitatory ones, leading to a progressively increasing ratio between excitation and inhibition towards excitation, potentially explaining the increased propensity to seizures in chronic undercut cortex.
Traumatic head injuries are frequently followed by a progressive increase of neuronal excitability, finally leading to seizures and epilepsy in animal models in vitro (McKinney et al., 1997; Prince et al., 1993), in vivo (D'Ambrosio et al., 2004; Nita et al., 2006, 2007; Topolnik et al., 2003a), and in humans (Dinner, 1993; Salazar et al., 1985). Normal brain function depends on a fine balance between excitation and inhibition, which could easily be disrupted following injury. Therefore, a reduced inhibition is thought to be particularly involved in the pathophysiology of epilepsy (Bernard et al., 2000; Sloviter 1987). The reduction of inhibition could result either from a loss of inhibitory synapses (Bloom et al., 1971; Ribak et al., 1982a,b), from alterations in GABA receptors (Bianchi et al., 2002; Wallace et al., 2001), or from a decreased number of GABAergic neurons (Buckmaster et al., 1999; Dinocourt et al., 2003; Hendry et al., 1986).
Several studies reported that GABAergic neurons might be selectively vulnerable to various injuries such as hypoxia (Romijn et al., 1988; Sloper et al., 1980), epilepsy induced by convulsive agents (Obenaus et al., 1993; Ribak et al., 1982a), excessive electrical stimulation (Sloviter 1987, 1992), and neocortical isolations (Ribak et al., 1982b). On the other hand, some studies suggested that GABAergic neurons are selectively spared following some other insults (Mathern et al., 1995; Tecoma et al., 1989). Nevertheless, the fact that epilepsy may be treated using drugs that enhance GABA receptor mediated inhibition (Fueta et al., 2005; Yamauchi et al., 2006) and that seizures can be triggered with GABA receptor blockers, such as penicillin and bicuculline (Karlsson et al., 1992), suggests that altered inhibition might represent an important pathogenetic mechanism of chronic epileptogenesis.
Anatomical studies showed that the inhibitory GABA system is remarkably plastic and can be up- or down- regulated under conditions such as deafferentation or excessive stimulation (Hendry et al., 1988, 1990; Micheva et al., 1995). This indicates that there also might be temporal variations of inhibition during the development of a chronic epileptogenic esion that would give quite different results at two time points (Franck et al., 1985, 1988; Sloviter 1992; Whittington et al., 1994). Therefore, it is essential to study the ratio between excitation and inhibition at several different time delays following an injury that could promote cortical hyperexcitability and epilepsy.
In this study, we used the model of partially isolated suprasylvian gyrus (Avramescu et al., 2008; Nita et al., 2006, 2007; Topolnik et al., 2003a,b) to reveal anatomical changes that might explain the increased frequency of seizures observed in cats following cortical undercut. We hypothesized that chronic deafferentation triggers major cortical reorganization and possibly a shift in the balance of excitation-inhibition towards excitation. This would contribute to the epileptogenetic mechanisms, which might explain the high rate of epilepsy observed in patients with severe head trauma, and also the progressive nature of this process.
Some parts of the present data have been previously reported in an abstract form (Avramescu et al., 2007).
All experimental procedures were performed in accordance with the guidelines of the Canadian Council on Animal Care and of the NIH Guide for the Care and Use of Laboratory Animals, and were approved by the Committee for Animal Care of Laval University. All efforts were made to minimize the number of animals used and their suffering.
Experiments were performed on 20 adult cats of both sexes. Surgical procedures were carried out in sterile condition, following a premedication with acepromazine (0.3mg/kg i.m.), butorphanol (0.3mg/kg i.m.), atropine (0.05mg/kg i.m.), and ketamine (20mg/kg i.m.), under sodium thiopental anesthesia (30mg/kg i.v.). The level of anesthesia was continuously monitored by electroencephalography (EEG), heart rate, oxygen saturation of the arterial blood (aiming over 90%), and end-tidal CO2 (~3.5%). General surgical procedures included cephalic vein cannulation for systemic liquid delivery (lactated Ringer's solution 5–10ml/kg/h), and lidocaine (0.5%) was used for infiltration of all pressure points or incision lines. Body temperature was maintained at 37–39°C using a heating pad. All substances used in our experiments were provided by Sigma-Aldrich, Canada, unless otherwise specified.
Craniotomy was used to expose the cerebral cortex and a large undercut of the white matter below the suprasylvian gyrus (13–15mm postero-anteriorly and 3–4mm medio-laterally) was used to produce partial cortical deafferentation. A custom-designed knife was inserted in the posterior part of suprasylvian gyrus perpendicular to its surface for a depth of ~3mm, then rotated 90° and advanced rostrally along the gyrus parallel to its surface for a total distance of 13–15mm, then moved back, rotated 90° and removed from the same place where it was entered. Thus, the anterior part of the undercut cortex was relatively intact and the white matter below the posterior part of the gyrus was transected, creating conditions of partial cortical deafferentation (Fig. 1A). The skull was reconstituted using acrylic dental cement and the skin of the scalp sutured. Animals were kept under observation up to full recovery and received analgesic medication (anafen 2mg/kg s.c.) for the next 72h.
Field potential recordings in chronic conditions were obtained by means of concentric bipolar electrodes (Frederick Haer & Co., Bowdoinham, ME; Fig. 1B), under ketamine and xylazine anesthesia (10–15mg/kg and 2–3mg/kg, respectively, i.m.), muscle paralysis with gallamine triethiodide (20mg/kg, i.v.), and artificial ventilation (20–30 cycles/min) seeking an end-tidal CO2 concentration of 3.5±0.4%. The craniotomy holes exposed the cerebral cortex and allowed the insertion of recording electrodes. Body temperature was maintained at 37–39°C, and glucose (5% solution) was administered i.v. every 3–4h during experiments.
The analysis of electrophysiological data was performed using Wavemetrics's Igor Pro software. In order to evaluate the progression of paroxysmal activity in the undercut suprasylvian gyrus, we computed the Fast Fourier Transform (FFT) of field EEG on 10-s epochs. Because in both early and late stages of the undercut the changes of EEG spectra occurred mostly at 0–4Hz, we quantified the power in this slow-wave activity range by the area under the graph of the FFT during slow oscillation (SO), as well as during spike wave (SW)/poly-spike wave (PSW) periods.
The animals were sacrificed 2, 4, and 6 weeks, respectively, after the initial trauma (W2 n=5, W4 n=5, W6 n=5), while control animals were sacrificed immediately following sham cortical deafferentation. Briefly, animals were deeply anesthetized using a combination of anesthetics containing ketamine (60mg/kg), xylazine (8.6mg/ml), and acepromazine (0.9mg/kg), and then perfused with cold saline (0.9%) followed by a fixative solution containing 4% paraformaldehyde, 0.05% glutaraldehyde in 0.1M phosphate-buffered saline (PBS; pH 7.4). After perfusion, brains were removed and stored in 30% sucrose in PBS for cryoprotection. Coronal brain slices were sectioned with a freezing microtome (Jung Histoslide 2000; Leica, Germany) at 80μm, through the middle of the suprasylvian gyri (7–10mm posterior to the anterior limit of the gyrus). Sections obtained from control animals, and 2W, 4W, and 6W groups were then either processed for immunohistochemistry or stained with thionine (Nissl staining) to ascertain the extension of the undercut and to visualize the general cortical cytoarchitecture.
Free-floating sections were first incubated for 30min in 1% H2O2 and 50% ethanol to inhibit endogenous peroxidase activity, than washed with PBS. Non-specific staining was minimized by incubating slices in 3% normal goat serum (NGS; Vector Laboratories, Burlingame, CA) for 1h at room temperature. Primary antibodies were diluted in 3% NGS, and the sections were incubated with primary antibodies for 48h at 4°C. We assured a better penetration of the antibody by using 0.3% Triton both during pre-incubation with the NGS and during incubation with the primary antibody. Anti-NeuN antibody was selected as a neuronal marker to reveal differences in the overall number of neurons among control and chronically deafferented cortical areas, while GABAergic neuronal populations were marked using anti-GAD 65&67 (glutamic acid decarboxylase) or anti-GABA antiserum. We employed both GAD and GABA staining because the presence of GAD and GABA in the same proportion indicates that GABA is not only accumulated, but also synthesized by the cells (Nishimura et al., 1986). After incubation with mouse anti-NeuN antibody (dilution of 1:200; Chemicon, Temecula, CA), sections were rinsed with PBS and incubated with their respective peroxidase conjugated secondary antibody (goat anti-mouse IgG; Chemicon) for 2h, at a dilution of 1:200. Finally, sections were incubated in avidin-biotin complex (ABC; Vector Laboratories) for 2h, and the reaction was revealed using 3,3′-diaminobenzidine (DAB; Sigma-Aldrich, Canada) as a chromogen in the presence of nickel ammonium sulphate and cobalt chloride (Sigma-Aldrich, Canada). Subsequently, sections marked with NeuN were double labeled with rabbit anti-GAD65&67 antiserum (dilution of 1:1500; Chemicon) or rabbit anti-GABA antiserum (dilution of 1:120; Chemicon), after re-blocking against non-specific staining and peroxidase inactivation. We used both GAD/NeuN and GABA/NeuN labeling in all animals included in the study. Following incubation with the biotinylated secondary antibody (biotinylated goat anti-rabbit IgG; dilution of 1:1500; Chemicon) for 2h, sections were incubated in ABC (Vector Laboratories) for 2h, and developed using DAB as a chromogen. Doubled-stained neurons were visualized as brown-colored cytoplasmic bodies (due to GAD65&67 or GABA reactivity) with black-blue nuclei (due to NeuN reactivity; see Fig. 3B below). In each experiment, some sections were incubated without the primary antibody to determine staining specificity (negative controls). Several sections were labeled only with anti-GAD65&67, or anti-GABA, or NeuN antibodies (see Fig. 3A below). Sections were mounted onto gelatin-coated slides, dehydrated in graded ethanol (50%, 70%, 90%, 95%, 100%), cleared in xylene and coverslipped for storage and analysis.
Sections were examined using a light microscope (Olympus BX61) equipped with the appropriate light sources (Olympus BX-UCB). Bright field images were acquired using a color CCD camera (Optronics, Goleta, CA), a motorized stage (MAC5000; Ludl Electronic Products, Exton, PA), and a computer system equipped with Neurolucida software (V8.0; MicroBrightField, Williston, VT) and Image-Pro Plus (V6.2; MediaCybernetics, Bethesda, MD). Slices considered for cell quantification were obtained from the middle region of the undercut, 5–10mm anterior from the entry point of the knife used for deafferentation. Architectonic boundaries of the suprasylvian gyrus were determined using Nissl-stained sections anterior and posterior to the immunostained sections. The regions of interest were outlined in the gray matter at the external surface of the gyrus, between the two perpendicular lines on the extremities of the undercut on frontal sections, in the area of maximal deafferentation (Fig. 2A, right panel). Double-labeled (NeuN/GAD, NeuN/GABA) or single-labed (NeuN) neurons were counted using Image-Pro Plus software. Staining was defined after thresholding for intensity, objects of interests were identified using predefined ranges of image analysis parameters from the Image-Pro Plus menu (such as color, diameter, and roundness); and parameters were equally applied to Z-stack images acquired at 5μm through the thickness of the slice (see Fig. 6B below). The imposed parameters (nuclear diameter 2–12μm; cellular diameter 10–30μm; cellular area>nuclear area; nuclear round-ness 1–1.2; cellular roundness 1–2) were consistent between control and chronic undercut slices. Although GABA antibodies are notorious for their difficult penetration (Broman et al., 1988), the immunostaining technique we used assured antibody penetration through the entire thickness of the slice (see Fig. 6A below). Therefore, we also used Z-stack images for quantification, to minimize the underestimation of GABAergic neurons. Equally, by quantifying both stack images and individual slices we accounted for some possible stereological errors related to neuronal changes in size or shape throughout the sections. Parameters were adjusted to obtain optimal sensitivity and specificity, by testing them on single-labeled slices. In addition to the absolute densities of neurons, we chose to illustrate neuronal densities also as percentages relative to the values obtained from control (see Fig. 6D below), in order to eliminate the possible errors in the automatic detection. All double-stained cells were considered inhibitory GABAergic neurons. We calculated the density of non-GABAergic excitatory neurons by subtracting the number of double-labeled inhibitory cells from the total number of neurons marked with NeuN.
The data were tested statistically using the Student's t-test and the analysis of variance (ANOVA) with the post hoc Tukey test adjusted for multiple comparisons. The test used is specified in the legend of each figure. Significant differences for all statistical testing were defined by p<0.05. Numerical data are represented as mean±standard deviation (SD). All statistical tests were performed using statistical analysis software (SPSS 14.0; SPSS, Inc., Chicago, IL).
All animals with chronic partially deafferented cortex showed progressive paroxysmal patterns of the cortical slow oscillation (Fig. 1B, left panels), similar to previous reports (Nita et al., 2006). In control (sham-operated) animals, EEG was characterized by cortical depression immediately after cortical undercut and was followed by slow-oscillating EEG patterns (<1Hz) in 30% of animals and paroxysmal discharges in the remaining animals, in accordance with earlier published data (Topolnik et al., 2003b). While in early stages of the undercut (2W group), the EEG field recording showed regularly alternating patterns of slow-oscillation, later stages were characterized by frequently occurring spikes and sharp transitions between the active and silent phases of the slow oscillation, giving rise to activities in the 3–4-Hz frequency range (Fig. 1C). The power spectra of the EEG signal also changed from a unimodal distribution at 2W with a peak at ~1Hz, corresponding to the slow oscillation frequency, to a multimodal distribution at 6W with components at 3–4Hz (Fig. 1C, left panel). Indeed, in the suprasylvian gyrus, the power in the 0–4-Hz range increased during SO (p<0.05, Student's t-test) at 4W (0.13±0.041mV2×Hz) and 6W (0.2±0.046mV2×Hz) compared to 2W (0.08±0.03mV2×Hz).
SW/PSW seizures continuously developing from anesthesia-induced slow-oscillation were described in 30% experimental cats (Steriade et al., 1995), as a result of the facilitating effect of ketamine and xylazine anesthesia (Boucetta et al., 2008; Steriade et al., 1998). However, in the present experiments, all anesthetized animals with chronic partial cortical deafferentation developed generalized seizures consisting in SW complexes at ~3Hz or SW and PSW complexes at lower frequency (1.5–2.5Hz) in the first weeks after deafferentation (2W–4W), intermingled in late stages with fast-runs of ~10–20Hz (Fig. 1B, right panels), similar to the waveforms seen in humans with severe epileptic syndromes (Niedermeyer 1969). The power in the 0–4-Hz range increased (p<0.05, Student's t-test) in the suprasylvian gyrus during these SW/PSW epochs from 0.15±0.086mV2×Hz at 2W, to 0.3±0.06mV2×Hz at 4W, and 0.35±0.077mV2×Hz at 6W. The fast-run activity (10–20Hz) occurred sporadically in three of five cats in the 6W group, and so we could not consistently quantify it or compare it with all the other groups. However, we interpreted the occurrence of fast runs as a sign of gravity of paroxysmal activity since it was present only in the very late stages of the undercut.
The increase in the power of SO from the early (2W) to the late stages of cortical undercut (4W and 6W) and the more severe patterns of seizures observed in chronic undercut at 6W are similar to previous reports (Nita et al., 2006) and suggest that epileptogenesis in the suprasylvian gyrus following trauma is a continuously evolving process and that in chronic stages the animals are more prone to develop severe seizures.
Normal cortex has a specific cytoarchitecture, being horizontally organized into six laminae (Baillarger, 1840) and vertically into groups of synaptically linked cells, called neocortical minicolumns, that represent the basic processing units of the mature neocortex, and which are further grouped together by short-range horizontal connections into cortical columns (Mountcastle, 1957, 1997).
Following chronic deafferentation of the suprasylvian gyrus by completely transecting the fibers in the white matter (Fig. 2A, right panel), we observed a gradual change in the normal distribution of neurons on columns and layers, particularly evident at 4W and 6W after the undercut was performed (Fig. 2B). Nissl staining is suggestive for a dis-organized cortical cytoarchitecture, lacking the expected arrangement of neurons on layers and columns typically seen in the normal hexalaminar neocortex (Fig. 2B,C), but more specific lamina stains are needed to further study subtle laminar disturbances in the chronically deafferented cortex.
Severe neuronal loss was observed in the deep layers of the suprasylvian gyrus (Fig. 2C), and the gray matter appeared to be shrunk (Fig. 2A,B,D). The few remaining neurons observed in the depth of the undercut cortex were abnormally oriented and showed cytological dystrophic properties, with small atypical and pycnotic nuclei (Fig. 2C).
The thickness of the cortical gray matter diminished progressively in the undercut suprasylvian gyrus (Fig. 2B,D) from 1.87±0.17mm in control animals to 1.72±0.16mm at 2 weeks, reaching a statistical significant difference (p<0.01, Student's t-test) at 4 weeks (1.5±0.21mm) and 6 weeks (1.45±0.18mm; Fig. 2D). The linear progression of this alteration pleads for a biopathological process and correlates with the increased propensity to seizures reported after chronic cortical deafferentation.
The Nissl-stained sections were suggestive for an important reduction in neuronal number in the undercut cortex (Fig. 2B,C). Therefore, we specifically assessed the neuronal loss and the depth distribution of excitatory and inhibitory neurons following injury using immunohistochemical staining. All neuronal nuclei were labeled with the anti-NeuN (neuronal-specific nuclear protein) antibody, while the inhibitory GABAergic neurons were labeled either with anti-GABA (gamma-aminobutyric acid) or anti-GAD 65&67 (glutamic acid decarboxylase) antibodies. The GAD, GABA, and NeuN immunolabeling in control and undercut cortex showed a dramatically reduced number of neurons in the undercut suprasylvian gyrus compared to control animals (Fig. 3A).
Double immunohistochemical labeling (both GAD/NeuN and GABA/NeuN; Fig. 3B) was used to build up depth profiles for the number of excitatory (labeled only with NeuN) and inhibitory GABAergic neurons (double-labeled). The neuronal density for each 100μm of depth in a particular example of double-labeled neurons with GAD/NeuN (Fig. 4A) and GABA/NeuN (Fig. 4B) shows a progressive neuronal loss following deafferentation.
We also calculated an averaged neuronal density from five different animals for each 100μm from the cortical surface, on a total area of 0.6mm2 (Fig. 5). The number of non-GABAergic, presumably excitatory neurons, decreased progressively after the initial trauma, reaching a first statistical significant value (p<0.05, Student's t-test) at 2 weeks at ~500μm (corresponding to cortical layers II–III; Fig. 5A). The decreased density of excitatory neurons became more important at 4 and 6 weeks for all deep layers (1600–2000μm, cortical layers V–VI), but also for some of the more superficial layers (Fig. 5A). The number of inhibitory GABAergic neurons was reduced significantly (p<0.05, Student's t-test) as early as 2 weeks after deafferentation for the deep layers (especially V–VI) and also for superficial layers II and III. The diminished density of inhibitory neurons was even more constant at 4 and 6 weeks for both the deep and the superficial layers (Fig. 5B). It is important to note that, although the number of inhibitory neurons globally decreased, the excitatory neurons from the deep cortical layers were also affected, and the hexalaminar cortical structure was disrupted.
In order to quantify the degree of neuronal loss after cortical trauma, we measured the densities of neurons in the undercut suprasylvian gyrus, in the contralateral suprasylvian gyrus, as well as in the ipsi- and contra-lateral marginal gyrus in sham-operated animals and after 2, 4, and 6 weeks following injury. For easier comparison, in addition to the absolute number of neurons per area of 0.07mm2 (Fig. 6C), we also present the relative neuronal densities (Fig. 6D). The total number of neurons stained with NeuN decreased from the control value (95.2±9.5) to 2 weeks (90.1±9.41) and became statistically significant (p<0.05, ANOVA with post hoc Tukey test) at 4 weeks (82.71±7.64) and 6 weeks (78.86±6.18) in the undercut gyrus, while neuronal densities in the undamaged gyri were similar to control within the time frame of our experiments (Fig. 6C). The inhibitory GABAergic neurons were much more affected by the traumatic injury, decreasing significantly (p<0.01, ANOVA with post hoc Tukey test) as early as 2 weeks (10.33±1.1 for GAD+ and 11.68±1.2 for GABA+ neurons) following undercut compared to control (12.4±2.15 for GAD+ and 14.1±1.85 for GABA+ neurons; Fig. 6C). The number of GABAergic neurons decreased even more (p<0.001, ANOVA with post hoc Tukey test) at 4 weeks (8.49±1.2 for GAD+ and 9.59±1.5 for GABA+ neurons) and 6 weeks (7.7±1.35 for GAD+ and 9.1±1.83 for GABA+ neurons) after the initial injury, whereas again they remained unchanged in the other gyri (Fig. 6C).
When analyzing globally the number of excitatory and inhibitory neurons only in the deafferented suprasylvian gyrus, we noticed once again that the density of excitatory neurons was significantly decreased (p<0.05, ANOVA with post hoc Tukey test) at 4 weeks (89.17±9.97%) and 6 weeks (85.06±7.1%) following deafferentation (Fig. 6D). The inhibitory GABAergic neurons were far more affected by the penetrating cortical injury, declining significantly (p<0.01, ANOVA with post hoc Tukey test) after only 2 weeks (82.9±8.52) and even more dramatically (p<0.001, ANOVA with post hoc Tukey test) after 4 weeks (68.02±10.55) and 6 weeks (64.54±12.98; Fig. 6D).
To better illustrate the preferential loss of these inhibitory GABAergic neurons, we calculated the ratio between excitatory and inhibitory cells. Although the data obtained in stack images are an excellent source of globally estimating the relative loss of neurons in the undercut cortex compared to the intact one, they may however underestimate the absolute number of cells due to the well-known poor penetration of the GABA antibodies throughout the thickness of the sections. Therefore, we decided to compute the ratio between excitatory and inhibitory neurons using the manually quantified data. The ratio was 4.25±0.63 in control, reflecting a density of about 20% for GABAergic neurons, as previously described (Gabbott et al., 1986). The preferential loss of these inhibitory neurons was reflected by a progressively increasing ratio between the excitatory and the inhibitory neurons from control to 2 weeks (p<0.05, ANOVA with post hoc Tukey test), and then to 4 and 6 weeks (p<0.01, ANOVA with post hoc Tukey test; Fig. 6E).
In this study, we illustrate the progressive anatomical changes that might explain the increased frequency of seizures observed after penetrating cortical wounds. We used the undercut cortex as a paradigm for post-traumatic seizures and we found an important disorganization of the cortical layers following injury and a progressive loss of neurons (NeuN-positive cells) in the deafferented suprasylvian gyrus. The inhibitory GABAergic neurons (GABA/GAD 65&67-positive) were particularly more sensitive to cortical trauma than the excitatory ones, leading to a relative raise in the number of excitatory neurons that might explain the increased propensity to seizures following penetrating cortical injury.
A higher frequency of seizures following head trauma has been described both in humans (Dinner 1993; Marcikic et al., 1998; Salazar et al., 1985) and in animal models (Nita et al., 2006; Prince et al., 1993; Topolnik et al., 2003a,b). The mechanisms responsible for this outcome include changes in intrinsic properties of pyramidal neurons (Esplin et al., 1994; Topolnik et al., 2003b), enhanced excitatory synaptic conductances without altered inhibition (Bush et al., 1999; Houweling et al., 2005), but also disorganization of normal cerebral cytoarchitecture either as microgyria or cortical sclerosis (Jacobs et al., 1999a,b; Marin-Padilla, 1999; Marin-Padilla et al., 2002; Swartz et al., 2006), or neuronal death (Lowenstein et al., 1992; Marin-Padilla, 1999; Marin-Padilla etal., 2002; Swartz et al., 2006).
A prominent characteristic observed in many types of epilepsy is the massive and widespread neuronal degeneration occurring in different brain areas (Ebert et al., 2002; Garrido Sanabria et al., 2006), produced either as a consequence of the acute excitotoxic damage (Gorter et al., 2003), or following recurrent chronic seizures, both in humans and animals (Cendes 2005). However, our data demonstrate cortical atrophy and substantial neuronal loss mainly due to the massive disappearance of GAD 65&67 and GABA-positive neurons in the suprasylvian gyrus. There were no statistical differences in cortical neuronal densities between control and 2W-4W-6W animals in the other analyzed gyri. Although we never directly tested the hypothesis that selective loss of GABAergic neurons contributes to epileptogenesis, the finding that neuronal loss in cats with chronically deafferented suprasylvian gyrus is strictly limited to the undercut suggests that the cause of neuronal loss was the deafferentation and not the subsequent seizures present in these animals (Nita et al., 2007). Examining whether such changes occur prior to the onset of the seizures would have been important information in order to establish the cause of neuronal loss. Nevertheless, some of the animals displayed electographic seizures even in acute conditions after cortical undercut, similar to what has been previously described (Topolnik et al., 2003b). This finding made impossible any quantification after the deafferentation but before the occurrence of seizures.
Previous data show that following brain trauma, surviving neurons develop remarkable axonal sprouting (McKinney et al., 1997; Salin et al., 1995), which despite the decrease in total neuronal density (present study), triggers increased probability of connection among remaining neurons (Avramescu et al., 2008; Jin et al., 2006), increasing the propensity to seizures (Nita et al., 2006).
Our results are further supported by studies done both in animals and in humans, in which the deafferentation was generated in a non-traumatic manner, by sensorial deprivation, such as limb amputation, ischemic nerve injury or visual deprivation (Chen et al., 1998, 2002; Hendry et al., 1986; Ziemann et al., 1998). It was shown that deafferentation of the somatosensory (Welker et al., 1989) or visual cortex (Hendry et al., 1986) in animals led to a reduction in the number of neurons containing GABA or its synthesizing enzyme, GAD. Similarly, studies using transcranial magnetic stimulation in humans clearly demonstrate a reduction of GABAergic cortical inhibition and an enhancement of excitation following amputation of the upper limb (Schwenkreis et al., 2000) and the lower limb (Chen et al., 1998), and after nerve ischemia (Ziemann et al., 1998).
GABAergic neurons are known to express several calcium-binding proteins (e.g., parvalbumin, calbindin, and calretinin) that can potentially be neuroprotective against the accumulation of intracellular calcium (Bouilleret et al., 2000) such as following extreme glutamate receptor activation during seizures. Consequentially, we may presume that GABAergic neurons would be less vulnerable than other phenotypes. Our data do not support this assumption; on the contrary we show a preferential loss of GABAergic neurons. Previous studies suggest that GABAergic neurons are much more dependent upon aerobic metabolism than other types of cortical neurons and they may be affected more severely by any insult to the brain (Ribak et al., 1982a; Sloper et al., 1980). Similarly, GABAergic neurons in retinal cultures were shown to be highly vulnerable to kainate neurotoxicity, probably due to the presence of AMPA/kainate receptors with high calcium permeability (Iino et al., 1990; Weiss et al., 1990). These are possibly the main two factors responsible for preferential loss of GABAergic cells in the undercut model.
We found a preferential loss of GABAergic neurons, similar to what has been described both in human (de Lanerolle et al., 1989; Robbins et al., 1991) and animal temporal lobe epilepsy (Sloviter 1987), but also a progressive disruption of the cortical hexalaminar structure analogous to morphological studies done in epileptic patients resistant to antiepileptic drugs showing a complete disorganization of the cortical lamination accompanied by abnormalities in the morphology and distribution of inhibitory neurons (Spreafico et al., 1998a,b). Recent data in humans show an important reduction of neocortical thickness and complexity in patients with mesial temporal lobe epilepsy (Lin et al., 2007). Moreover, Marín-Padilla et al. (2002) showed that in children with brain trauma, the neurons directly damaged by the initial insult or incapable of reestablishing functional connections will die, while neurons capable of reestablishing post-injury functional connections will survive and will start a slow process of reorganization resulting in progressive cortical dysplasia. They also illustrated that, in post-traumatic epileptic children (as in shaken infant syndrome), the amount of residual gray matter displays areas of complete destruction alternating with areas of dysplastic tissue without lamination, similar to what we describe in the chronically deafferented suprasylvian gyrus. Although it is unlikely that the exact same mechanisms apply in our experimental conditions, this line of evidence suggests that major brain trauma is capable of triggering important structural changes that will ultimately affect the functioning of the entire neuronal network. In the light of these data, the increased resistance to antiepileptic therapy of the patients with posttraumatic epilepsy may rely on an abnormal laminar cortical structure and neuronal death. Given that most of the new antiepileptic drugs exert their anticonvulsant action through enhancement of inhibitory-mediated neurotransmission (White, 1999), they possibly fail to control some posttraumatic epilepsy because of the low number of inhibitory neurons in the damaged cortex.
To sum up, we report that penetrating brain injuries initiate a morphological cortical disruption (loss of neurons, laminar disorganization) capable of shifting the balance between excitation and inhibition towards excitation in the deafferented cortex, contributing probably to the increased propensity to seizure of chronically injured brains. Nevertheless, additional experiments are required to determine whether this disproportionate loss of GABAergic neurons when compared to the loss of pyramidal neurons is the actual cause of the development of recurrent seizure activity in the model.
This research was supported by grants (MOP-67175 and MOP-93611) from the Canadian Institutes of Health Research, Natural Science and Engineering Research Council of Canada (grant 298475), and National Institute of Neurological Disorders and Stroke (1R01NS060870-01). I.T. is a scholar of Fonds de la Recherche en Santé du Québec, and S.A. is a Savoy Foundation fellow.
No conflicting financial interests exist.