Ramona Frida Moroni and Francesca Inverardi contributed equally to the work.
Address correspondence to Carolina Frassoni, Unit of Clinical Epileptology and Experimental Neurophysiology, Fondazione I.R.C.C.S., Istituto Neurologico “C. Besta,” via Celoria 11, 20133 Milano, Italy. E-mail: firstname.lastname@example.org
Purpose: Cortical dysplasia (CD) represents a wide range of histopathological abnormalities of the cortical mantle that are frequently associated with drug-resistant epilepsy. Recently, carmustine (1-3-bis-chloroethyl-nitrosurea [BCNU]), given to pregnant rats on embryonic day (E) 15, has been used to develop an experimental model mimicking human CD. The aim of this study was to characterize cytological and histological alterations in this model, and compare the results with those observed in human CD.
Methods: Pregnant rats were given intraperitoneal injections of BCNU on E15. Sections of cerebral cortex from adult BCNU-treated rats were cytoarchitecturally and immunohistochemically analyzed using anti-SMI311, anticalbindin (CB), and antiparvalbumin (PV) antibodies. The density of the PV-immunoreactive (PV-ir) interneurons was quantitatively assessed by means of a two-dimensional cell-counting technique, and the spatial distribution of PV-ir neurons was evaluated by using the Voronoi tessellation.
Results: The morphological features included reduced cortical size, laminar disorganization, and heterotopic clusters of neurons. We also identified large, disoriented SMI311-positive pyramidal neurons, and dysmorphic neurons intensely immunostained for neurofilaments, similar to those observed in human dysplastic cortex. An altered distribution of PV-immunoreactive cortical interneurons was also present.
Conclusions: Although some of the cytoarchitectural abnormalities found in BCNU-exposed cortex are similar to those found in other CD models, we identified new alterations that recall the neuropathological description of type IIA (Taylor's type) CD. BCNU-treated rat could therefore be a useful additional model for investigating the pathogenic mechanisms involved in this CD.
Derangements in cortical development usually lead to a spectrum of malformation phenotypes that largely depend on the timing of the defect in corticogenesis, that is, the affected developmental step (Barkovich et al., 2005). Nonetheless, cortical development is an extremely complex process, achieved through ordered temporospatial events, such as cellular proliferation, apoptosis, neuronal migration, and differentiation, which are temporally overlapping. Therefore, injuries occurring at a particular time in embryonic or postnatal life may affect more than one developmental event and thus lead to a wide range of alterations. Focal cortical dysplasia (FCD) is a subtype of cortical developmental disorders frequently associated with drug-resistant epilepsy (Tassi et al., 2002; Colombo et al., 2003; Guerrini et al., 2003). Although mainly characterized by cerebral cortex laminar disorganization, it is not a single entity as its histopathological appearance is heterogeneous. In fact, heterotopic neurons in the white matter, or giant and/or dysmorphic neurons and/or balloon cells, may be associated with abnormal cortical lamination and lead to different categories of FCD (Tassi et al., 2002; Palmini et al., 2004; Lawson et al., 2005).
A new experimental rat model of CD obtained by administering carmustine (1-3-bis-chloroethyl-nitrosurea [BCNU]) in a pregnant rat's uterus has recently been proposed to study the physiology of dysplastic cortex (Benardete & Kriegstein, 2002). BCNU, a DNA-alkylating agent that affects the proliferating pool of neuronal and glial precursors when delivered on embryonic day 15 (E15), induces histological alterations suggestive of CD such as cortex thinning and laminar disorganization, heterotopia, and abnormal/cytomegalic neurons. With field-recording techniques, under condition of partial GABA blockade, the dysplastic cortex of BCNU-treated rats showed a significant increase of spontaneous epileptiform events and excitability compared with cortex from control animals (Benardete & Kriegstein, 2002). To further clarify the cytological and histological alterations characterizing this experimental model, we immunocytochemically investigated morphological changes in the principal cells and interneurons. We concentrated our investigation on the somatosensory cortex because it is the most affected area. We found that the cortical cytoarchitectural abnormalities in BCNU-treated rats are similar to those observed in type IIA human FCD associated with intractable epilepsy (Tassi et al., 2002; Palmini et al., 2004). Preliminary results have previously been presented in abstract form (Moroni et al., 2005).
Materials and Methods
The experiments were undertaken in accordance with the guidelines defined by the European Communities Council Directive (directive 86/609/EEC), and every effort was made to limit the number of animals used.
Twelve pregnant Sprague-Dawley rats (Charles River Italia, Calco, Italy) were intraperitoneally (i.p.) injected with 20 mg/kg of BCNU solution (5% sterile glucose in water at 4 mg/ml) on E15 (considering the day of vaginal plug as E1), at the time of peak cortical neurogenesis. On the same gestational day, control pregnant rats were injected i.p. with 5% glucose alone. The day of birth of the rat pups was designed postnatal day (P) 0.
All of the pups of five litters died soon after birth; mortality in the surviving litters ranged from 10% to 45%. The rats surviving the first 2 days of life grew to adulthood without any apparent problem except for a lower body weight in comparison with the control animals.
Perfusion, histology, and immunohistochemistry
To investigate the effects of BCNU on brain development, we used 28 rats from seven litters exposed to BCNU in utero and 10 controls (aged from P21 to P90). The BCNU-treated animals belonged to four groups of seven rats each, aged P21, P30, P40, and P >80. Two additional rats (BCNU-treated and control) were used to study the cortical fiber pathways.
The rats were anesthetized with 4% chloral hydrate 1 ml/100 g of body weight administered i.p., and then transcardially perfused with 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB), pH 7.2, by means of a peristaltic pump. Their brains were removed from the skull, immersed in 4% paraformaldehyde in PB for 1–3 days, cut into 50-μm-thick serial coronal sections by means of a Vibratome VT1000S (Leica, Heidelberg, Germany), stored in PB and processed for histological and immunocytochemical examination or embedded in paraffin, and sectioned (7 μm) on a microtome. Sections adjacent to those processed for immunocytochemistry were stained with thionin (0.1% in distilled water) for the cytoarchitectonic analyses.
For the immunohistochemical analyses, selected free-floating vibratome sections were preincubated for 30 min in 3% H2O2 in phosphate-buffered saline (PBS), pH 7.4, in order to inactivate endogenous peroxidase. After rinsing in PBS, nonspecific sites were blocked in PBS containing 10% normal goat serum (NGS) or normal horse serum (NHS) and 0.2% Triton-X100, and the sections were incubated overnight at 4°C with the following antisera diluted in 1% NGS (or NHS) in PBS: antiparvalbumin (PV) 1:10,000 (Swant, Bellinzona, Switzerland), anti-Calretinin (CR) 1:5,000 (Swant), anti-SMI311 1:1,000 (Sternberger Monoclonals Incorporated, Lutherville, MD, U.S.A.) or anticalbindin (CB) 1:5,000 (Swant). After several rinses in PBS, the sections were incubated in biotinylated goat antirabbit IgG or biotinylated horse antimouse IgG (Vector Laboratories, Burlingame, CA, U.S.A.) diluted 1:200 in 1% NGS (or NHS) in PBS. The avidin-biotin-peroxidase protocol (ABC; Vector) was followed, with 3,3′-diaminobenzidine tetrahydrochloride (DAB; Sigma, St. Louis, MO, U.S.A.) being used as chromogen. Finally, the sections were mounted, dehydrated, and coverslipped with DPX (BDH, Poole, Dorset, U.K.). In order to check antibody specificity, some sections were processed omitting the primary antibody or replacing it with NGS diluted 1:100. No specific staining was detected.
In order to study the cortical fiber pathways, paraffin sections were dewaxed in xylene, rehydrated, and immunolabeled with the polyclonal antisera antimyelin basic protein (MBP) 1:500 (DakoCytomation, Glostrup, Denmark) following the same protocol of the vibratome sections or stained with luxol fast blue (0.1% in 95% ethyl alcohol).
Free-floating vibratome sections from the BCNU-treated rats and human tissue were blocked in 10% NGS and 0.2% Triton-X100 in PBS for 45 min and then incubated overnight at 4°C in SMI311 antibody diluted 1:1,000 in PBS containing 1% NGS. For double-immunostaining, some sections were incubated in a mixture of SMI311 and PV primary antibodies, respectively, diluted 1:1,000 and 1:5,000 in PBS containing 1% NGS. After several rinses in PBS, the sections were incubated in the secondary fluorescent antibody Cy2-conjugated goat antimouse IgG (1:200; Jackson Immunoresearch Laboratories, West Grove, PA, U.S.A.) or, for double-immunostaining, in a mixture of the secondary fluorescent antibodies, Cy2-conjugated goat antimouse IgG (1:200) and indocarbocyanine (Cy) 3-conjugated goat antirabbit IgG (1:600; Jackson Immunoresearch Laboratories). After further rinsing in PBS, the sections were mounted with Fluorsave (Calbiochem, San Diego, CA, U.S.A.) and examined through a confocal laser scanning microscope (Bio-Rad, Hemel Hempstead, U.K.) equipped with an argon/krypton mixed gas laser, and mounted on a light microscope (Eclipse E600, Nikon, Tokyo, Japan). The confocal images were recorded through separate channels at excitation peaks 510 nm (for Cy2) and 550 nm (for Cy3), with the double-immunostained sections being merged using Bio-Rad Lasersharp 2000 software. All the parameters were standardized to ensure the best quality and comparability (for details, see Frassoni et al., 2005).
Human surgical specimens
We also examined surgical samples of human temporal cortex from two patients operated on for drug-resistant epilepsy secondary to type IIA FCD (Palmini et al., 2004) at the “Claudio Munari” Surgery Centre for Epilepsy, Milan (patient 1: a 5-year-old girl with complex partial seizures with type IIA FCD in the right parietal cortex; patient 2: a 32-year-old man with complex partial seizures with type IIA FCD in the left parietal cortex). The surgery was performed for strictly therapeutic reasons after the patients had given their informed consent. Brain specimens were immersed in a freshly prepared fixative solution of 4% PFA for 3 h at room temperature, and for a further 24 h at 4°C. Fifty-micrometer vibratome sections were cut. For cytoarchitectonic controls, alternate sections adjacent to those processed for immunocytochemistry were stained with thionin. The immunohistochemistry and immunofluorescence assays were performed under the same conditions as those used for the animal slices (for further details see Garbelli et al., 1999).
Cortical thickness measurements and semiquantitation of PV-positive cells
Cortical thickness was determined using a 4X objective and ACT-2U software (Nikon, Tokyo, Japan) on two thionin-stained coronal sections taken at approximately the rostral and caudal levels of the somatosensory cortex of each of 25 BCNU-treated rats and 10 controls. Mean cortical thickness ± SD in the experimental and control groups was calculated and the results compared using the Student t-test, with a p value of <0.05 being considered statistically significant.
The number of cells in PV-immunoreacted sections from 12 BCNU-treated rats and seven controls was counted using a two-dimensional cell-counting technique. Two sections taken at approximately rostral and caudal levels for each animal were examined through a 4X objective, and Image Pro-Plus 5.1 software (Media Cybernetics, MD, U.S.A.) was used to capture a series of four or five contiguous images covering the entire area of the somatosensory cortex at high resolution, which were subsequently tiled together. Zoom factor 4 was applied to the captured images, and all of the strongly immunopositive neurons within this area were manually counted. The differences in cell density (cells/mm2± SD) between the two groups were assessed by means of the t-test.
Moreover, in order to assess the total neuronal density and the ratio of PV-positive cells to total neurons in BCNU-treated and control rat, cell counting was performed on paraffin sections immuoreacted with anti-PV antibody and counterstained with thionin. Eight sections taken at approximately rostral and caudal levels from one BCNU-treated and one control rat were examined through a 10X objective, and Image Pro-Plus 5.1 software was used to capture a series of five or six contiguous images covering the entire area of the somatosensory cortex at high resolution, which were subsequently tiled together. PV-immunopositive and thionin-stained cells were manually counted.
The Dirichlet tesselation (also called Voronoi tessellation) was used to evaluate the spatial relationship between the PV-ir neurons in the BCNU-treated rats and controls (Duyckaerts & Godefroy, 2000; Eriksson et al., 2003) and to obtain the confidence interval of the mean neuronal density. The distribution of PV-ir neurons was characterized on the previously acquired digital images of the anterior somatosensory cortex of three controls and three BCNU-treated rats. Separate charts of PV-ir neurons were obtained from rectangular frames with a fixed horizontal dimension of 1,000 μm and a vertical dimension adjusted according to cortex thickness (between layer II and white matter). The map of the distribution of the PV and CR-ir neurons was built by measuring the X and Y coordinates for each neuron using Neuron_Morpho plugin for ImageJ (freely available at http://www.maths.soton.ac.uk/staff/D'Alessandro/morpho). Voron software (which can be downloaded from the Web site http://mendel.neuromorpho.chups.jussieu.fr/~charles/) was used to perform the Voronoi tessellation of the maps, which consisted of drawing a polygon around each point (PV-ir neuron) delineating “the territory of the map that is closer to the point than to any other point of the map” (Duyckaerts et al., 1994); the sides of the polygons were medians between the point and its neighbors. The polygons located at the medial and lateral periphery of each selected area were removed in order to decrease the border effects (Duyckaerts & Godefroy, 2000), whereas upper and lower polygons were included because of the “natural” boundaries given by the layer II and white matter. Three frames covering most of the anteroposterior extent of the somatosensory areas were considered for each animal and calculated the coefficient of variation (CV) of each, that is, the ratio between the standard deviation and mean value of the polygon areas. The CV provides an indication of the spatial distribution of neurons: CVs of 0.33–0.64 are associated with a random distribution, those of less than 0.33 with a regular distribution, and those of more than 0.64 with a clustered distribution. Repeated-measure analysis of variance (ANOVA) was performed to evaluate the effect of groups and frames, using a 5% level of significance.
Macroscopic and microscopic features of BCNU-treated brains
On gross examination, the brain of all the adult animals exposed to BCNU in utero appeared microcephalic, with clear hypoplasia of the telencephalon leaving the collicular plate entirely exposed. In comparison with the controls (Figs 1A and 1B), the thionin-stained coronal sections of the treated rats showed a reduced brain size, partial agenesis of the corpus callosum, and ventriculomegaly that was particularly marked caudally (Figs 1D and 1E). The neocortex of the affected brains was clearly thinned (Figs 1C and 1F) as confirmed by cortical thickness measurements. Mean cortical thickness in the control group was 2.00 ± 0.12 mm anteriorly and 1.66 ± 0.18 mm posteriorly, while comparable regions of the BCNU-treated rats measured on average 1.37 ± 0.15 mm rostrally and 1.26 ± 0.16 mm caudally. The between-group differences in both measurements were significant (p < 0.0001).
The CD induced by BCNU spread throughout the rostrocaudal extent of the neocortex, but varied in severity. In all cases, alterations were more frequently observed in the somatosensory cortex, the area specifically examined in the present study. The most prominent histopathological feature was the disorganization of cortical layering, which was observed in almost all the animals: in the milder forms (39% of the rats), layers V and VI were recognizable whereas the supragranular layers appeared discontinuous, with groups of neurons separated by areas of low cell density (Fig. 2A); in more severe forms (61% of the rats), layering was hardly recognizable, being disorganized by the presence of large nodular clusters of neurons in the superficial (Fig. 2B) or deep cortex (Fig. 2C). In most of the animals, hippocampal hypoplasia was observed (Fig. 1E) with a disruption of CA1 or CA2 areas in about 45% of the animals owing to the presence of nodules of heterotopic neurons (data not shown). No detectable alterations in the thalamus were found.
The morphology and localization of the cortical heterotopic clusters were accurately revealed by means of immunohistochemistry using an antibody against CB, a calcium-binding protein whose immunolabeling clearly distinguishes the superficial layers from layers V and VI (Sanchez et al., 1992). In the BCNU-treated rats, this pattern was often disrupted and CB-immunoreactive (ir) neurons were observed in differently shaped abnormal clusters that suggested the cells were unable to migrate to their proper position (Figs. 2 D–F). Groups of CB-ir neurons in the dorsal cortex were organized in radial columns associated with large vessels (Figs. 2D and F) or in nodules localized in the lowest part of the grey matter (Figs. 2E and 3F); in the lateral cortex they were frequently localized in the superficial layers (Fig. 3B). Other animals showed larger volumes of heterotopic neuronal clusters expanding through different layers of the cortex (Fig. 2F).
Additional information concerning cortical impairment was provided by the staining obtained using antibodies against PV, a calcium-binding protein marker of a subpopulation of inhibitory neurons and SMI311, which recognizes nonphosphorylated neurofilaments and allows the identification of pyramidal neurons. The heterotopic clusters were formed by both pyramidal and nonpyramidal neurons with no identifiable laminar pattern or consistent orientation (Figs. 3A–H). In addition to CB-ir neurons (Figs. 3B and 3F), the clusters enclosed PV-ir cells (Figs. 3C and 3G), and some SMI311-ir neurons with randomly oriented apical dendrites that were prevalently located along the borders of the clusters (Figs. 3D and 3H).
The organization of cortical fibers of BCNU-treated rats was clearly disarranged as shown in Fig. 4. Tangential bundles of fibers were present in layer I of treated animals but not in control brains (Figs. 4A and 4D); the main fibers, usually running radially through middle layers of the normal cortex (Fig. 4B), were warped and frequently crossed by tangential ones in BCNU-treated rats (Fig. 4E); in the deep layers, the heterotopic clusters appeared almost devoid of fibers (Fig. 4F).
Another striking feature of the brain of the BCNU-treated rats was the presence of ectopic SMI311-ir neurons in cortical layer I/II. These were abnormally oriented pyramidal cells whose apical dendrites were turned toward the deep layers or extended tangentially (Figs. 5A and 5B). Double-labeling experiments showed that despite their abnormal location, these pyramidal neurons were surrounded by PV-ir puncta (Fig. 5B). Similarly, ectopic pyramidal neurons were also observed in human type IIA CD (Figs. 5D and 5E). Moreover, rare dysmorphic neurons with atypical dendritic processes and rich in neurofilaments were detected in the BCNU-treated rats as well as in the specimens with type IIA CD (compare Figs. 5C and 5F).
Inhibitory neurons: altered spatial distribution of PV-positive neurons
Cell bodies and neuropil were intensely immunopositive for PV in both the BCNU-treated and control rats, although some BCNU-treated rats had areas with little or no immunolabeling (Fig. 6). Nevertheless cell counting analysis did not reveal any significant differences in density of PV-ir neurons between the control (125 ± 26 cell/mm2) and treated rats (133 ± 43 cell/mm2). This datum was confirmed calculating also the confidence interval of the mean density of PV-ir neurons by means of Voronoi tessellation (Table 2).
Table 2. Density of PV-positive neurons: comparison between two methods of cell count
No. of rat
Densitya (2D) cells/mm2
Mean density (Voronoi)b cells/mm2
Confidence intervalb cells/mm2
No. of rat
Densitya (2D) cells/mm2
Mean density (Voronoi)b cells/mm2
Confidence intervalb cells/mm2
acalculated with the standard 2D method.
bcalculated by means of Voronoi.
This result was supported by the measure of the ratio of PV-positive cells to total thionin-stained cortical neurons, which are comparable in the control (0,073) and BCNU (0,084) rats. Moreover, even if the cortex was significantly thinner in this CD model, the neuronal density of the dysplastic cortex was similar to that of the control (BCNU: 1,157 cell/mm2; control: 1,192 cell/mm2).
In line with previous observations (Celio, 1990), the somatosensory cortex of the control rats contained PV-ir neurons in all cortical layers except layer I, although they were mainly located in layers II–IV. Conversely, the distribution of PV-ir neurons in the BCNU-treated rats seemed to be disrupted, as confirmed by means of Voronoi tessellation (see an example in Fig. 7). Density was irregular in the control rats, being less in the lower part of layer VI (large polygons) and more in the other layers (small polygons), particularly layer IV, whereas the BCNU-treated rats showed a more homogeneous distribution (Fig. 7). In order to determine the type of neuronal distribution, we calculated the CV for each area (see Materials and Methods for details). All of the animals (controls and treated rats) showed a clustered distribution, with average CVs of >0.64 (above the upper limit for random distribution) (Table 3). However, the BCNU-treated rats had lower average CVs (0.66 ± 0.05 vs. 0.95 ± 0.11; ANOVA: F = 28.689, p = 0.002), thus indicating that the distribution of PV-ir cells tended to be random.
Table 3. Voronoi-derived coefficients of variation for the PV-positive cell population in the somatosensory cortex of control and BCNU-treated rats
No. of rat
CV ± SD
No. of rat
CV ± SD
PV, parvalbumin; CV, coefficient of variation; SD, standard deviation.
1.04 ± 0.10
0.70 ± 0.05
0.98 ± 0.20
0.58 ± 0.02
0.82 ± 0.07
0.66 ± 0.04
0.70 ± 0.01
0.65 ± 0.03
0.95 ± 0.11
0.66 ± 0.05
Conversely, the distribution of CR-ir neurons did not differ between control and BCNU-treated rats (Fig. 8). The examination of spatial distribution by means of Voronoi analysis validated our qualitative observation: no difference was revealed between the CV of CR-positive cells in BCNU (CV = 0.64) and control rats (CV = 0.61).
We here described the neuropathological aspects of the somatosensory cortex of BCNU-treated rats. Beside the main morphological features including reduced cortical size, laminar disorganization, and heterotopic clusters of neurons previously described by Benardete and Kriegstein (2002), we identified some novel cytoarchitectural abnormalities such as a disarrangement in the organization of cortical fibers, the improper position of some pyramidal cells and the presence of dysmorphic neurons closely resembling those observed in human type IIA CD (Taylor's type CD without balloon cells). Finally, our analysis of the spatial distribution of PV-ir cells revealed an altered location of the inhibitory interneurons in BCNU-exposed cortex.
BCNU is commonly used as a chemotherapeutic drug because its alkylating properties contribute to kill proliferative cells (Carter et al., 1972; Walker et al., 1978; Drabløs et al., 2004; Ewend et al., 2005). Its high degree of lipophilicity and central nervous system penetration make its in utero administration efficacious in affecting the proliferating pool of neuronal and glial precursors in E15 rats (Benardete & Kriegstein, 2002). Like other teratogenic methods, BCNU exposure leads to many morphologic features similar to human CD. The most widely used agents to obtain CD animal models are methylazoxymethanol acetate (MAM) and ionizing radiation (RAD) which, when given to rats on E15, induce microcephaly, cortical thickness thinning, and layering disorganization (Roper, 1998; Colacitti et al., 1999). We found that the same alterations were also the most evident in rats treated with BCNU. Microcephaly and cortical thinning (the average thickness of the somatosensory cortex in the BCNU-treated rats was about 30% less than in controls) are probably due to cell loss, which may cause a reduction in the number of fibers and consequently reduce the thickness of the white matter and corpus callosum. Layering disorganization represents the postinjury structural adaptation of neurons surviving the original insult, and is prevalently due to the formation of heterotopic clusters of neurons in the deep (VI) or superficial layers (II and III). We have also provided evidence for alterations of the cortical fiber pathways in BCNU-treated rats: we observed abnormal tangential fibers located in layer I and fibers skirting heterotopias. Similar alterations were previously reported in MAM rats (Chevassus-au-Louis et al., 1999) and in some cases of human FCD (Janota et al., 1992; Hannan et al., 1999). Although the disarrangement of axonal pathways secondary to CD is possible, its pathogenic implications remain unknown.
CB antibody staining provided additional information concerning cortical impairments. We found that the neuronal clusters in the deep layers of BCNU-treated cortices consist of CB-positive neurons and show immunolabeling typical of the superficial layers in normal rats (Celio, 1990; Sanchez et al., 1992). It could be argued that these cells are unable to migrate to their proper position, probably because they do not find a permissive environment for correct migration. It is plausible to suppose that migration patterns are affected in BCNU-treated rats as it was demonstrated in the MAM model in which CD is the consequence of a teratogen exposure (Battaglia et al., 2003).
Unlike cortical heterotopias in the deep layers, which are prevalently found in the dorsal cortex, superficial aggregates are observed in the lateral somatosensory cortex. They are separated by areas of low cell density that may be due to the selective columnar destruction of neural precursors induced by BCNU. In fact, it has been suggested that cortical neuroblast proliferation is not synchronous, but occurs through “proliferative units” (Rakic, 1988; Ferrer et al., 1993). Only some units may be damaged by BCNU, thus leading to the intermingling of cell-poor columns and areas with a normal cell composition. However, further studies are necessary in order to establish which cells are ablated by BCNU and how the pattern of migration is affected.
In addition to altered cortical layering, other typical features of BCNU-treated rats are as follows: (1) the improper position of some SMI311-positive pyramidal neurons in layer I/II which, despite their abnormal position, seemed to be surrounded by inhibitory PV-immunopositive terminals; and (2) dysmorphic neurons that are intensely immunostained for neurofilaments. Dysmorphic neurons have been observed only in few other cases of CD animal models (see Table 1), such as RAD and TSC2+/− rats (Mizuguchi et al., 2000; Marín-Padilla et al., 2003), making BCNU-treated rats a good model to study specific aspects of this FCD. As shown in Fig. 5, these cell abnormalities recall the neuropathological description of type IIA CD, characterized by the presence of dysmorphic and ectopic neurons and differing from type IIB CD by the absence of balloon cells (Tassi et al., 2002; Palmini et al., 2004). With the exception of TSC2+/−, balloon cells, a typical feature of type IIB CD and tubero sclerosis, have not been described in CD models, including BCNU-treated rats. The nature of balloon cells still remains enigmatic: they express glial and neuronal markers as well as proteins usually seen in neural stem or precursor cells, suggesting an immature differentiated state. In fact, type IIB CD is considered a malformation secondary to defects in cell proliferation (Barkovich et al., 2005), a mechanism presumably not involved in most of the injury-based models; this could explain the absence of balloon cells in BCNU and other CD animal models.
It can be argued that type II dysplasias are mainly focal in adult epileptic patients while the neuropathological alterations were diffused throughout the cortex of BCNU-treated rats. However, it is worth noting that the extent of these dysplasias are also extremely variable in human pathology, ranging from very focal lesions with a few and scattered dysplastic cells to the multilobar dysplastic lesions recently described in pediatric patients (Hildebrandt et al., 2005; Cepeda et al., 2006). It must also be remembered that cortical neurogenesis is completed in a few days in rat, but takes weeks in humans. The lesions caused by BCNU may therefore influence a wide range of developmental processes and lead to diffuse CD, whereas human focal CD may be caused by injuries affecting only a restricted number of processes.
It has been suggested that an imbalance between the excitatory (glutamatergic) and inhibitory (GABAergic) activities of cortical structures may play a role as epileptogenic mechanism and neuropathological findings suggest that impaired GABAergic mechanisms may play a pivotal role in focal epilepsies (Engel, 1996; Marco et al., 1996; Spreafico et al., 1998; Garbelli et al., 1999; Spreafico et al., 2000; Crino et al., 2001). GABAergic interneurons express calcium-binding proteins CR and PV. In particular those expressing PV are critically important in controlling pyramidal cell excitability (DeFelipe, 1999), and their number or distribution has been found altered in various types of human and experimental CD. The examination of spatial distribution by means of Voronoi analysis showed no differences in CR-immunoreactivity between control and BCNU-treated rats while it revealed altered spatial distribution of PV-ir neurons in BCNU-treated rats.
Although in the majority of human FCD cases a decrease of PV-positive interneurons in the region of dysplasia has been reported (Garbelli et al., 1999; Spreafico et al., 2000; Thom et al., 2003), only few quantitative analyses have been performed in human dysplastic tissue (Garbelli et al., 2006; Zamecnik et al., 2006). Changes in the density of PV-interneurons have been also reported in some genetic (i.e., flathead, tish rats) or injury-based (i.e., irradiation, freeze) CD animal models (Rosen et al., 1998; Roper et al., 1999; Sarkisian et al., 2001; Trotter et al., 2006), but data are contradictory or lacking in other cases (see Table 1). Although we did not find any significant differences in the total number of PV-ir neurons between BCNU-treated and control rats, the former frequently showed patches of decreased PV-immunostaining associated with an altered spatial distribution of PV-ir neurons. We evaluated PV-ir cell distribution by means of Voronoi tessellation, which has recently been used to estimate nerve cell distribution in the cerebral cortex (Eriksson et al., 2003; Carretta et al., 2004). The obtained CVs gave some indications concerning the type of distribution in terms of cell aggregation (Duyckaerts et al., 1994; Duyckaerts & Godefroy, 2000). The control rats showed a high average CV, thus indicating the clustered distribution of PV-ir cells. In fact, in line with previously published data (Celio, 1990; Sànchez et al., 1992), PV-ir neurons were prevalently located in layers II–IV, although present in all cortical layers except layer I. On the contrary, the average CV in the BCNU-treated rats (0.66) was only slightly above the upper limit for random distribution (0.64), thus indicating that they tend to be randomly distributed. This result demonstrates that the dysplastic cortex is associated with a loss of the normal PV-ir cell grouping architecture. Interestingly, abnormally distributed GABAergic interneurons, including clusters separated by gaps containing few or no GABAergic cells, have been recently described in dysplastic cortex from patients with FCD type II (DeFelipe et al., 1993; Ferrer et al., 1994; Alonso-Nanclares et al., 2005; Calcagnotto et al., 2005). These anatomical data suggest that not only a loss of interneurons but also an altered arrangement of GABAergic neurons (uneven distribution) might generate an altered circuitry, resulting in inhibitory dysfunctions. Thus, the altered distribution of PV-ir cells that we found in BCNU-treated rats may be one of the causes of the hyperexcitability demonstrated by Benardete and Kriegstein (2002).
In spite of what occurs in humans, BCNU-treated rats, as well as most of the CD animal models, do not exhibit spontaneous recurrent seizures. However, it has been previously reported that CD lowers the threshold to seizures triggered by hyperthermia or kainic acid (Germano et al., 1996; Germano & Sperber, 1998). Further studies are needed to characterize the clinical state of BCNU-treated rats more precisely, as well as their propensity to develop spontaneous behavioural seizures.
In conclusion, our results show that many of the cytoarchitectural abnormalities found in the cerebral cortex of BCNU-treated rats (alteration of the cortical layering, presence of ectopic and dysmorphic neurons) are similar to those found in human dysplastic tissue and, in particular, in type IIA CD. However, it should be emphasized that a model of CD is “a model” and in no cases an animal model can completely replicate the complex features of human pathology; therefore, caution is required to transfer experimental data to human clinical and pathological disorders. In fact, some differences between BCNU model and human type IIA CD can be pointed out, that is, the presence of nodular formations and the presence of a diffuse (not focal) dysplasia. In this regard, it should be underlined that “complex” malformations can be observed, although rarely, in patients presenting nodular formations in association with type IIA CD and, as previously mentioned, diffuse dysplasia has been recently described in pediatric patients. Therefore, despite some limitations, the development of new models might provide additional understandings on specific aspects of the etiopathogenesis of CD and may also provide good tools to assess the causal relationship between CD and epileptogenesis.
The authors wish to thank Dr. Rita Garbelli for providing the human tissue photomicrographs and for her comments on the manuscript. This work was supported by grants of the Italian Ministry of Health to C.F.
Conflict of interest: The authors have no personal, commercial, academic, or financial conflicts of interest with anybody. 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.