Histopathology and reorganization of chandelier cells in the human epileptic sclerotic hippocampus

Abstract

Impairment of GABA‐mediated inhibition is one of the main hypotheses invoked to explain seizure activity, both in experimental models and in human epilepsy. We have studied the distribution and the neurochemical characteristics of certain GABAergic circuits in the normal and epileptic human sclerotic hippocampal formation. We have focused our attention mainly on chandelier cells because, together with basket cells, they are considered to have powerful effects on spike generation. Chandelier cells represent a unique type of interneuron whose axon terminals (Ch‐terminals) form synapses with the axon initial segments of cortical pyramidal cells and granular cells of the dentate gyrus. Different neurochemical subpopulations of chandelier cells have been identified by immunocytochemistry, mainly in the neocortex. Markers for Ch‐terminals include the GABA transporter 1 (GAT‐1), the polysialylated form of the cell‐surface glycoprotein neural cell adhesion molecule (PSA‐NCAM) and the calcium‐binding proteins parvalbumin (PV) and calbindin D‐28k (CB). In the normal hippocampal formation, GAT‐1‐ and PV‐immunoreactive (‐ir) Ch‐terminals were identified in the granular and polymorphic layers of the dentate gyrus, in the strata pyramidale and oriens of the CA fields, and in the pyramidal layer of the subicular complex. In addition, and in contrast to the hippocampus and dentate gyrus, subsets of Ch‐terminals in the upper pyramidal layer of the normal subiculum express CB and PSA‐NCAM. The sclerotic hippocampus of epileptic patients presented an impressive morphological and neurochemical reorganization of Ch‐terminals and basket formations. This was apparent in the dentate gyrus and hippocampal formation, but not in the subiculum, which appeared to remain unaltered. Principally, numerous and more complex PV‐ and CB‐ir Ch‐terminals, as well as dense PV‐ir basket formations, appeared in some hippocampal segments, whereas in other regions there was a lack of labelled elements. These changes varied considerably not only between different patients, but also within different hippocampal fields in a given patient. In general, the changes were not correlated with the clinical characteristics or degree of histopathological alterations observed in the patients, such as granular cell dispersion, neuron loss and proliferation of mossy fibres. However, some surviving neurons in the regions adjacent to the areas of neuron loss were consistently innervated by dense basket formations and complex Ch‐terminals. These results indicate that, in the human epileptic hippocampus, GABAergic circuits are more highly modified than previously thought. When considered along with other extrahippocampal alterations, we suggest that these changes are important in the pathophysiology of temporal lobe epilepsy associated with hippocampal sclerosis.

epilepsy; interneurons; inhibition; calcium‐binding proteins; GABA transporter

AIS = axon initial segment; CB = calbindin; Ch‐terminal: chandelier cell axon terminal; GAT‐1 = GABA transporter 1; ir = immunoreactive; NeuN = neuron‐specific nuclear protein; PSA‐NCAM = polysialylated neural cell adhesion molecule; PV = parvalbumin

Introduction

Hippocampal sclerosis is typically observed in patients with temporal lobe epilepsy (Honavar and Meldrum, 1997). In general, it is characterized by gliosis and neuronal loss, most prominently in the CA1 field of the hippocampus, followed by the hilus, CA4 and CA3 fields. Neurons in the dentate granular cell layer and in the CA2 field are relatively unaffected. In addition, there is often a dispersion of the dentate granular cell layer, ectopic neurons being found in the molecular layer (Houser, 1990; Houser et al., 1992). Neuronal loss is accompanied by axonal reorganization involving both excitatory and inhibitory neurons (e.g. de Lanerolle et al., 1989; Sutula et al., 1989; Babb et al., 1991; Sloviter, 1991; Mathern et al., 1995a).

The basic mechanisms by which seizure activity is related to hippocampal damage or alterations in excitatory and inhibitory systems remain unclear (e.g. Sloviter, 1991; Mathern et al., 1995b; Bernard et al., 1998, 2000; Houser, 1999; Blümcke et al., 2002). One of the main hypotheses to explain the hyperexcitability of hippocampal principal cells and seizure activity is that GABA‐mediated inhibition is impaired. A large variety of morphological and neurochemical types of GABAergic neurons innervate different regions of principal cells (for a review see Freund and Buzsaki, 1996). Indeed, a major GABAergic input to the somata (and proximal dendrites) and to the axon initial segment (AIS) of principal cells originates in the so‐called large basket cells and chandelier cells, respectively (DeFelipe, 1999). Chandelier cells are characterized by the vertical rows of boutons at the terminal portion of their axons (Ch‐terminals), which contribute exclusively to symmetrical synaptic contacts with the AIS of pyramidal cells and dentate granular cells (reviewed in Somogyi et al., 1982; Freund and Buzsaki, 1996; DeFelipe, 1999). The AIS is a region critical for the control of cell excitability, the generation of axon potentials and the axonal output of principal cells (Stuart and Sakmann, 1994; Colbert and Johnston, 1996). Therefore, Ch‐terminals, together with axon terminals from basket cells, are considered to exert a very strong GABAergic inhibitory influence on principal cells, in contrast to interneurons that target dendrite membrane compartments (Miles et al., 1996).

In recent years it has become increasingly clear that inhibitory inputs onto the AIS of principal cells are not homogeneous across the different cortical regions, layers, neuronal populations and species. In addition to the synapses from other types of interneurons that occasionally contact the AIS of pyramidal cells (e.g. Gonchar et al., 2002), a single AIS may be innervated by one or a few chandelier cells (reviewed in DeFelipe, 1999). In addition, chandelier cells in the neocortex are chemically heterogeneous. Across species, layers and neuronal populations, they contain selected combinations of different substances, including the GABA transporter GAT‐1, the calcium‐binding proteins parvalbumin (PV) and calbindin D‐28k (CB), and the polysialylated form of the cell surface glycoprotein neural cell adhesion molecule (PSA‐NCAM). As a result, all of these markers are used to visualize Ch‐terminals (reviewed in DeFelipe, 1999; see also Arellano et al., 2002). Previous electron microscopic studies using PV immunocytochemistry reported the presence of Ch‐terminals in the human hippocampus (Seress et al., 1993; Wittner et al., 2001). However, no systematic description of the general distribution and chemical heterogeneity of Ch‐terminals in the human hippocampus has been performed. Therefore, one of our main objectives was to determine the distribution and chemical features of Ch‐terminals in the normal human hippocampus and subicular complex using immunocytochemistry for GAT‐1, PV, CB and PSA‐NCAM.

In patients suffering temporal lobe epilepsy, the genesis or the maintenance of epileptic activity has been related to alterations in the number and the distribution of chandelier cells in the neocortex and/or hippocampal formation (DeFelipe, 1999; Wittner et al., 2001). However, in the epileptic hippocampal formation there are no detailed studies comparing the different patterns of neuronal loss and gliosis with the possible alterations in the neurochemical profiles and the morphology and distribution of Ch‐terminals and basket formations. Thus, we also examined the morphological and chemical alterations of basket formations and Ch‐terminals in sclerotic hippocampi surgically removed from patients with temporal lobe epilepsy.

Material and methods

Human brain tissue was obtained from two sources: from autopsies (kindly supplied by Dr R. Alcaraz, Forensic Pathology Service, Basque Institute of Legal Medicine, Bilbao, Spain) and postoperative tissue (Department of Neurosurgery, Hospital de la Princesa, Madrid, Spain). Tissue obtained at autopsy (2–3 h post‐mortem) was from three normal males (aged 23, 49 and 63 years) who had died in traffic accidents. The brains were sectioned in the coronal plane and 1.5 cm thick slices were immersed in a cold solution of 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, for 24–36 h. Biopsy brain tissue was obtained from 14 patients suffering pharmacoresistant temporal lobe epilepsy, showing hippocampal sclerosis and with a well‐documented medical history (Table 1). The patient’s consent was obtained in all cases according to the Declaration of Helsinki (BMJ 1991; 302: 1194), and all protocols were approved by the institutional ethical committee (Hospital de la Princesa, Madrid, Spain). In addition, hippocampal biopsies from two epileptic patients showing no histopathological alterations were used as biopsy controls.

Surgical treatment of the epileptic patients required anterior temporal cortical resection, which included the posterior amygdala, the anterior portion of the hippocampus (1–3 cm) and the adjacent cortex. In all cases, video‐EEG recording from the scalp and bilateral foramen ovale electrodes was employed to localize the epileptic focus. Furthermore, epileptogenic regions were identified through subdural recordings with a 20‐electrode grid (lateral neocortex) and a four‐electrode strip (uncus and parahippocampal gyrus) at the time of surgery. Electrocorticographic recordings were performed at least two or three times during surgery for ∼5–10 min each. In all cases, the visual inspection of intraoperative electrocorticographic recordings revealed spiking activity predominantly localized in the mesial electrodes, but in most patients the lateral cortex also demonstrated a significant amount of activity.

Immediately after removal, biopsy samples were fixed in cold 4% paraformaldehyde for 24–36 h. Small blocks (∼15 × 10 × 10 mm) were taken from the neocortex of the medial and inferior temporal gyri (Brodmann areas 20 and 21) and throughout the rostrocaudal extent of the hippocampal formation, from both autopsy and biopsy material. Coronal sections (50 µm) were obtained from the blocks with a vibratome.

Immunocytochemistry

Sections from the hippocampal formation of autopsy and biopsy tissue were batch‐processed with standard immunocytochemical techniques (Muñoz et al., 2002) using the following antisera: rabbit‐anti‐GAT‐1 (1 : 500; Chemicon, Temecula, CA, USA), mouse‐anti‐PV (1 : 4000; Swant, Bellinzona, Switzerland), mouse‐anti‐CB (1 : 4000; Swant), mouse‐anti‐neuron‐specific nuclear protein (NeuN) (1 : 2000; Chemicon), mouse‐anti‐PSA‐NCAM (1 : 10 000, IgM; Developmental Studies Hybridoma Bank, University of Iowa, USA) or rabbit‐anti‐dynorphin A (1 : 20 000; Peninsula, San Carlos, CA, USA). The sections were then processed using the avidin–biotin method, with secondary goat‐anti rabbit (GAT‐1 or dynorphin A immunostaining) or horse anti‐mouse (PV, CB, NeuN and PSA‐NCAM immunostaining) biotinylated antibodies (1 : 200; Vector Laboratories, Burlingame, CA, USA) and the Vectastain ABC immunoperoxidase kit (Vector Laboratories) with 3,3′‐diaminobenzidine tetrahydrochloride (Sigma‐Aldrich, St Louis, MO, USA) as a chromogen. The sections were dehydrated, cleared with xylene and coverslipped. In control sections, processed either with primary antibodies that had been preadsorbed with PV or CB protein (Swant), or without the primary or secondary antibodies, no significant staining was detected. No differences were observed in the immunocytochemical staining of sections from autopsy material and sections from the two control biopsies. Adjacent sections stained with thionine or immunostained with the neuronal marker NeuN were used to reveal the borders between different areas and layers.

Histopathological analyses

To evaluate the degree of neuronal loss in the hippocampus, Nissl‐stained sections were used to estimate neuronal density in all subfields of epileptic patients and controls. Images obtained with a Hitachi video camera attached to a Leitz microscope were used for counting on a computer screen. A 100× oil‐immersion objective (final magnification 4125×) was used for the dentate gyrus, whereas a 40× objective (final magnification 1650×) was used for the remaining fields. A stage, motorized in the z‐axis, was used to determine section thickness. The number of granular cells per radial column (width 64 µm) in straight segments of the dentate gyrus was estimated by counting nucleoli in 30 µm thick optical disectors (West and Gundersen, 1990). Six optical disectors were performed per case. Adjacent microscopic fields, extending from the hilar aspect of the dentate gyrus to the most dispersed cells in the molecular layer, were studied. Dispersion of granular cells was inferred from 10 measurements of the width of the granular layer in straight segments. The proliferation of mossy fibres was estimated in dynorphin A‐immunostained sections (Houser et al., 1990) by measuring the radial extent of dynorphin staining (n = 4 per case) in the granular and molecular layers in straight segments of the dentate gyrus. Neuronal density in the hilus and in the CA4 field was estimated by counting nucleoli in nine randomly selected fields (160 × 120 µm) per case. Neuronal density in the CA3, CA2 and CA1 fields was estimated by counting nucleoli in columns (width 160 µm) including the strata oriens and pyramidale. Six fields for CA3 and CA2 and nine fields for CA1 were studied per case, along the mediolateral extent of the CA fields. Neurons with more than one nucleoli were rarely found, and in these cases only one nucleolus was counted. Statistical analysis of the differences between material from patients and controls, in terms of both estimates of neuron density and granular cell dispersion, were performed by analysis of variance (one–sided) with Dunnett post hoc comparisons. Nissl‐stained sections from two blocks of the lateral temporal neocortex per patient were analysed for study of the possible cytoarchitectonic alterations.

To generate figures, images were captured with a digital camera (Olympus DP50) attached to an Olympus light microscope, and Adobe Photoshop 6.0 software (Adobe Systems, San Jose, CA, USA) was used to generate figure plates. Correlation analysis of the degree of histopathological alterations (Table 2) and clinical characteristics (Table 1) with changes of inhibitory circuits was performed using the χ² test or Spearman’s test of correlation.

Results

Characteristics of the tissue and nomenclature of the fields of the hippocampal formation

In this study, we compared in detail the immunocytochemical staining for a variety of markers of GABAergic circuits in both control and epileptic temporal lobe tissue (Figs 1 and 2). The relatively short post‐mortem period of the autopsy material (2–3 h) and the rapid processing of the surgically removed tissue contributed to a high degree of anatomical preservation of the brain tissue. In the present study we followed the nomenclature of Amaral and Insausti (1990) with respect to the hippocampal formation, but considering the existence of the CA4 field according to Rosene and van Hoesen (1987). We use the term hippocampus in its general sense, i.e. including the dentate gyrus and hippocampus proper (cornu ammonis fields). The nomenclature of Braak (1980) was followed for the subicular complex.

Neuropathological assessment

Standard neuropathological assessment of the surgically removed tissue showed hippocampal sclerosis in 14 patients, as revealed in Nissl‐stained or NeuN‐immunostained sections (Fig. 1). In two patients, no histopathological alterations were found in the hippocampal formation and this material was used as control biopsy tissue. In this material, the pattern of immunostaining for all neuronal markers analysed was similar to that found in autopsy material. Unless otherwise stated, the term ‘patients’ will refer only to subjects with hippocampal sclerosis.

In the dentate gyrus, neuronal loss was significant in five of the 14 patients studied (Table 2). However, granular cell dispersion and mossy fibre proliferation were found in all patients except patient H44, who did not show granular cell dispersion (Fig. 3 and Table 2). The extent of granular cell dispersion and the proliferation of mossy fibres (in sections processed for dynorphin A) were variable between patients and within different regions in the same subject (Fig. 3; Table 2). In the CA fields, different degrees of neuronal loss and gliosis were found, mainly in the stratum pyramidale (Fig. 1B; Table 2). In the subicular complex, no apparent histological alterations were observed. Furthermore, small rounded or elongated, abnormal masses (ranging from 500 to 2500 µm in diameter) consisting mainly of glial cells were detected around blood vessels in six patients (H57, H61, H109, H123, H136 and H138). These focal glial scars appeared mainly in the white matter of the angular bundle, affecting in some cases the subiculum and entorhinal cortex. In addition, the cytoarchitecture of the lateral temporal neocortex was altered in some regions in five (H57, H75, H84, H108 and H115) of the 14 patients.

Immunocytochemistry for GABAergic circuits

Normal tissue

The general distribution of GAT‐1, PV, CB and PSA‐NCAM in the human hippocampal formation has been described in previous studies (Braak et al., 1991; Sloviter et al., 1991; Seress et al., 1993; Mikkonen et al., 1998; Mathern et al., 1999). Unless otherwise specified, in the present study we will focus mainly on axon terminals innervating the soma and the proximal dendrites (basket formations) and, in particular, on the Ch‐terminals.

GAT‐1 immunocytochemistry. Immunocytochemistry for GAT‐1 labelled numerous terminal‐like puncta in the neuropil surrounding unlabelled cell bodies. Staining of Ch‐terminals was widely distributed across the hippocampus and subicular complex, but in particular region‐ and lamina‐specific patterns. In the dentate gyrus, GAT‐1‐ir Ch‐terminals were observed in the granular layer and in the polymorphic layer. However, in the granular layer GAT‐1‐ir Ch‐terminals were difficult to visualize and were only occasionally distinguished from the punctate staining (Figs 4A and 5A). In the stratum pyramidale of the CA4, CA3 and CA1 fields, GAT‐1‐ir Ch‐terminals were clearly distinguished (Fig. 6A) but were sparse in the stratum oriens. In CA2, Ch‐terminals were only occasionally observed. In contrast, no labelled Ch‐terminals were observed in the alveus and the strata radiatum and lacunosum‐moleculare. In the subicular complex, GAT‐1‐ir Ch‐terminals were found in the pyramidal cell layer of the subiculum, presubiculum and parasubiculum. The highest density of GAT‐1‐ir Ch‐terminals was found in the clusters of the upper pyramidal layer of the subiculum, where Ch‐terminals showed intense staining (Fig. 7A, C). GAT‐1‐ir Ch‐terminals were infrequently found in the deeper portion of the pyramidal layer. In the presubiculum and parasubiculum, GAT‐1‐ir Ch‐terminals were also intensely stained.

PV immunocytochemistry. Immunocytochemistry for PV stained non‐pyramidal cell bodies and a dense plexus of axonal and dendritic processes in the principal cell layers of the different subfields of the hippocampus and subicular complex (Fig. 2A). The axonal plexus was composed of axon processes running in all directions, and axon terminals that included Ch‐terminals and basket formations. In these regions, PV‐ir Ch‐terminals showed morphological, regional and laminar distribution patterns similar to those stained for GAT‐1. However, Ch‐terminals were more difficult to visualize with PV immunocytochemistry than with GAT‐1, especially in the granular layer of the dentate gyrus (Figs 4D and 5C). PV‐ir Ch‐terminals in the CA1 field are shown in Fig. 6C, while in Fig. 7B and D Ch‐terminals in the subiculum are illustrated.

CB immunocytochemistry. Immunocytochemistry for CB stained cell bodies and dendritic and axonal processes in the hippocampus and subicular complex (Fig. 2C). The labelled cell types included, in addition to non‐pyramidal cells, dentate granular cells (Figs 4G and 5E) and a population of pyramidal neurons in the CA1 and CA2 fields. In general, no CB‐ir basket formations or Ch‐terminals were observed in the hippocampus or in the presubiculum and parasubiculum. However, in the subiculum, CB‐ir Ch‐terminals were found in the upper pyramidal layer, associated with the clusters of pyramidal cells (Fig. 7E, G). In the deeper aspect of the subicular pyramidal layer, a lower density of CB‐ir Ch‐terminals was observed. These Ch terminals were less intensely stained than those in the upper pyramidal layer.

PSA‐NCAM immunocytochemistry. The distribution of PSA‐NCAM was sparse in non‐pyramidal cell bodies, but a relatively high density of immunostained axonal and dendritic processes was found throughout the hippocampal formation. However, as occurred with CB immunocytochemistry, no PSA‐NCAM‐ir Ch‐terminals were found in the dentate gyrus or in the CA fields, the presubiculum or the parasubiculum. In contrast, in the subiculum, PSA‐NCAM‐ir Ch‐terminals were detected in the upper pyramidal layer associated with clusters of pyramidal cells (Fig. 7F, H).

Epileptic tissue

Dentate gyrus. In the granular layer of the dentate gyrus of some patients there were dramatic changes in the GAT‐1, PV and CB immunostaining of Ch‐terminals (Figs 4, 5 and 8; Tables 3–5). The complexity of Ch‐terminals in these patients was augmented in comparison with those in controls, and a subpopulation of Ch‐terminals expressed CB (Figs 4H and 5F). These alterations were heterogeneous between patients and no significant correlation could be found between the degree of granular cell loss, cell dispersion or the extent of mossy fibre proliferation and the changes in the labelling of Ch‐terminals in specific zones of the dentate gyrus (Tables 2–5). The most notable changes were found in sections stained for PV and CB; therefore, we will describe the results mainly in this material.

Immunocytochemistry for PV showed a general reduction in the density of immunoreactive fibres in the granular cell layer. The degree of reduction was rather heterogeneous between cases (Fig. 4E, F). In addition, each patient showed differences in the density of immunostained processes within different zones of the dentate gyrus (Fig. 8). However, in the dentate granular layer of all patients, PV‐ir Ch‐terminals were clearly distinguishable in some portions (Figs 4E, F, 5D and 8C). Two subtypes of Ch‐terminals were observed with regard to their size and the density of terminals, and will be referred to as simple and complex (Fig. 9). Simple Ch‐terminals were similar to those found in controls. They were made up of one or two rows of labelled boutons (Fig. 9A). In contrast, complex (or hypertrophic) Ch‐terminals consisted of tight cylinder‐like structures made up of multiple rows of boutons (Fig. 9B). Complex PV‐ir Ch‐terminals were found in 11 out of 14 patients in discrete zones of variable sizes throughout the extent of the granular layer (Fig. 5D; Tables 4 and 5). The relative proportions of simple and complex Ch‐terminals varied between patients and between different zones of the granular layer in the same patient. Complex Ch‐terminals were associated more frequently with the non‐migrated granular cells than with those that had dispersed into the molecular layer. GAT‐1 immunostaining also revealed the presence of complex Ch‐terminals in the granular layer of five out of 14 epileptic patients (Fig. 5B; Tables 3 and 5).

The distribution of CB was also altered in the dentate gyrus where CB immunostaining of granular cell bodies was generally reduced in epileptic patients compared with controls (Figs 4G–I and 5E, F). This reduction in CB labelling was especially evident in five patients (H61, H84, H108, H109 and H115) and affected mainly the granular cells located in the hilar aspect of the dentate gyrus (Fig. 4I). Patches of complex CB‐ir Ch‐terminals (Fig. 5F) were observed in five patients (Tables 3 and 5), four of whom showed a significant reduction in CB immunostaining of the granular cell bodies. The analysis of alternate serial sections immunostained for CB and PV revealed that the areas containing complex CB‐ir Ch‐terminals were frequently included within larger areas containing complex PV‐ir Ch‐terminals. The complex CB‐ir Ch‐terminals did not coincide with those that contained GAT‐1 (Tables 3 and 5). Complex Ch‐terminals immunostained either for GAT‐1, PV or CB were more frequently observed in patients with a significant decrease in the number of granular neurons and with less proliferation of mossy fibres, but no significant correlation was found. No simple or complex PSA‐NCAM‐ir Ch terminals were found in the granular layer of the epileptic dentate gyrus.

Finally, in the polymorphic layer of the dentate gyrus a variety of changes was found. In PV‐immunostained sections the density of labelled processes was very heterogeneous between different zones of the same patient, and between patients. For example, patient H48 showed intense PV immunostaining of Ch‐terminals and basket formations in discrete areas, whereas other zones presented a complete lack of PV‐positive fibres (Fig. 10). In some cases these terminals were immunoreactive for GAT‐1 but not for CB or PSA‐NCAM.

CA fields. To examine the possible alterations in inhibitory circuits, PV immunostaining of neuropil, basket formations and Ch‐terminals was analysed in detail (Table 4). A general reduction in the density of immunostained processes was observed in the CA fields, CA4 and CA3 being the most severely affected. It was also common that, within zones where a general decrease in PV immunostaining was observed, some PV‐ir basket formations and Ch‐terminals remained which were considerably more complex than in controls (Figs 6C, D, 10C–E and 11D, E).

A remarkable feature of the reorganization of the PV‐ir processes in all CA fields was that the alterations were heterogeneous both between patients and within the same patient (Table 4). Thus, they were apparently not associated with any particular cytoarchitectonic characteristic. However, in all the CA fields, PV‐ir dense basket formations were consistently found associated with some surviving neurons at the border of regions adjacent to the areas of neuronal loss (Fig. 11C, D, E). Within the regions of severe neuronal loss in CA1, and despite the presence of some PV‐ir cell bodies and axonal processes, PV‐ir Ch‐terminals or basket formations were rarely found.

GAT‐1 immunostaining in the CA fields revealed the presence of both simple and complex Ch‐terminals (Fig. 6A, B; Table 3). A high density of complex Ch‐terminals was revealed by GAT‐1 staining, associated with the surviving pyramidal cells of CA1, close to the subiculum. In this region, CB immunostaining labelled simple Ch‐terminals in all patients and complex Ch‐terminals in four patients (Table 3). In the CA2 field of three patients, some intensely stained complex CB‐ir Ch‐terminals were also found (Table 3). No Ch‐terminals immunoreactive for PSA‐NCAM were observed in any of the CA fields.

Subicular complex. In the subicular complex, there were no apparent differences in the density and distribution of basket formations or Ch‐terminals expressing GAT‐1, PV and CB, or in the pattern of immunostaining for PSA‐NCAM between epileptic patients and controls (not illustrated).

Discussion

In this study we investigated the distribution of Ch‐terminals labelled for GAT‐1, PV, CB and PSA‐NCAM in the normal and sclerotic human hippocampus. We show for the first time that in the normal hippocampal formation GAT‐1‐ir Ch‐terminals can be identified in the granular and polymorphic layers of the dentate gyrus, in the strata pyramidale and oriens of the CA fields, and in the pyramidal layer of the subicular complex. We have also shown that Ch‐terminals in the upper pyramidal layer of the normal subiculum also express CB and PSA‐NCAM. However, this is not the case in the CA fields and dentate gyrus.

We also identified dramatic morphological and neurochemical reorganizations of Ch‐terminals and basket formations in the sclerotic hippocampus of epileptic patients. These changes varied considerably across different hippocampal fields in a given patient, and between different patients. The changes were not correlated with the clinical characteristics or the degree of histopathological changes in these patients, such as granular cell dispersion, neuron loss and proliferation of mossy fibres. However, certain surviving neurons adjacent to areas of neuron loss were consistently innervated by dense basket formations and complex Ch‐terminals. In the subicular complex, no apparent alterations were found in epileptic patients with regard to the cytoarchitecture or the distribution of GAT‐1‐, PV‐, CB‐ and PSA‐NCAM‐immunostained elements. These results emphasize that, in the epileptic human hippocampus, GABAergic circuits are more modified than previously thought.

Chandelier terminals in the normal hippocampal formation

The presence of Ch‐terminals in the dentate gyrus and the CA fields in humans has been reported previously (Seress et al., 1993; Wittner et al., 2001). The results presented here confirm and extend these studies, indicating that Ch‐terminals show both region‐ and laminar‐specific variation in their distribution and neurochemical characteristics. These differences suggest that variations exist in the relative weights of inhibitory synaptic inputs to the AIS of the principal cells in the various hippocampal fields from the population of chandelier cells. The possible functional implications of these differences should be evaluated in further structure–function studies.

In the dentate gyrus and CA fields of both primate and non‐primate mammalian species, Ch‐terminals are CB‐negative and PV‐positive (Ribak et al., 1990; Seress et al., 1991, 1993; Freund and Buzsaki, 1996; present results). Nevertheless, at the electron microscope level CB‐ir terminals have recently been reported to make symmetrical synapses with the AIS of pyramidal cells in CA1 of the normal human hippocampus, suggesting the presence of a subpopulation of CB‐ir chandelier cells (Wittner et al., 2002). Here we could only detect CB‐ir Ch‐terminals in the subicular complex. Our failure to identify a subpopulation of CB‐ir Ch‐terminals in the CA1 field could be simply explained by the fact that they were so faintly labelled that they passed unnoticed in our preparations. However, another possibility is that the CB‐ir axon terminals that form synapses with the AIS originate from interneurons other than chandelier cells (e.g. Gonchar et al., 2002). Further studies using a larger number of controls should be performed to reach a more definitive conclusion about the distribution of CB‐ir Ch‐terminals in the hippocampal formation.

In the subicular complex, we have shown that the distribution of Ch‐terminals is different from that in the hippocampus. PV and GAT‐1 expressing Ch‐terminals were distributed throughout the subicular complex, and were denser and more intensely immunostained in the upper aspect of the pyramidal layer, associated with clusters of pyramidal cells. These clusters of pyramidal cells were associated with Ch‐terminals that stain intensely for CB and PSA‐NCAM. The similar density and distribution patterns of Ch‐terminals immunostained for these four neuronal markers suggest the presence of a distinct population of chandelier cells that express these four substances. These results indicate that the inhibitory input by chandelier cells to the AIS of pyramidal cells in these clusters might exhibit particular physiological properties (Arellano et al., 2002).

Chandelier cell axon terminals and basket formations in temporal lobe epilepsy

In the present work, we found a morphological and chemical reorganization of the Ch‐terminals that contact the AIS of some of the surviving principal cells in discrete regions of the dentate granular layer and the CA fields of certain epileptic patients (see below). It is not known whether some of these Ch‐terminals are newly formed or whether they are generated by the sprouting of previously existing Ch‐terminals which increase in complexity and relocate around surviving principal cells. We have previously suggested that the presence of PSA‐NCAM in Ch‐terminals of the entorhinal cortex and the temporal neocortex might be related to the structural remodelling of these Ch‐terminals (Arellano et al., 2002). However, it is unlikely that PSA‐NCAM plays a role in the structural remodelling of Ch‐terminals associated with dentate granular cells and CA pyramidal cells, since these terminals do not express PSA‐NCAM in either the normal or the epileptic hippocampus. Nevertheless, we cannot discard the possibility that the structural remodelling of these Ch‐terminals takes place at early stages of hippocampal sclerosis (or of the epileptic activity), and that PSA‐NCAM might be transiently expressed by these Ch‐terminals during a limited time window.

Dentate gyrus

The granular cells represent the major source of hyperexcitability in the hippocampus during seizures since they exhibit reduced inhibition and enhanced excitation (Ben‐Ari, 1985; Sutula et al., 1986, 1989; Sloviter, 1989, 1991; Heinemann et al., 1992; Mody et al., 1992; Isokawa et al., 1993). Nevertheless, an increase rather than a decrease in inhibition was reported in the dentate gyrus of some epileptic patients (Isokawa‐Akesson et al., 1989; Prince and Jacobs, 1998). In addition, alterations in the number and subunit composition of GABA_A receptor subtypes (Loup et al., 2000) and a reduction in the expression of GABA_BR1a‐b receptors (Muñoz et al., 2002) have been reported in dentate granular cells of epileptic patients. In the present study, we observed a general reduction of inhibitory terminals in the granular layer of epileptic patients. This was concomitant with a clear increase in the complexity of Ch‐terminals in discrete zones of the dentate gyrus. These results further support the existence of profound alterations of inhibitory circuits that could differentially affect the inhibitory control of dentate granular cells in epilepsy.

It has been proposed that the focal loss of PV‐ir chandelier cells (among other interneurons) in the temporal neocortex of patients suffering temporal lobe epilepsy may be related to the genesis or the maintenance of seizure activity (DeFelipe, 1999). Similarly, in the granular layer of the dentate gyrus of epileptic patients, a decrease in the number and intensity of staining of PV‐ir fibres, which presumably represent inhibitory pericellular fibres, has been reported, suggesting that basket and chandelier cells may have been lost (Sloviter et al., 1991). Wittner and colleagues reported that the number of PV‐ir cells is strongly reduced in the hilus and CA4 region of epileptic patients (Wittner et al., 2001). However, these authors found that the synaptic coverage of the granular cell AISs is increased more than threefold, even in regions of the dentate gyrus devoid of PV‐ir elements, suggesting that a loss of PV expression rather than a reduction of Ch‐terminals occurs in these patients. Nevertheless, it should be noted that changes in the distribution and complexity of PV‐ir Ch‐terminals and the degree of neuronal loss and gliosis varied considerably between patients and between different zones of the dentate gyrus within a given patient.

In some epileptic patients, CB expression is decreased or absent in granular cells and other principal neurons (Maglóczky et al., 1997, 2000; present results). However, while the density of CB‐ir interneurons does not change in the hilus of epileptic patients with hippocampal sclerosis, their distribution, dendritic morphology and connections are altered (Sloviter, 1989; Sloviter et al., 1991; Blümcke et al., 1999; Maglóczky et al., 2000). In these studies, CB immunostaining of interneurons was reported to be either normal or increased, suggesting an increase in CB synthesis (Sloviter et al., 1991; Maglóczky et al., 1997). The present study shows that, whereas the normal human dentate gyrus lacks CB‐ir Ch‐terminals, in certain epileptic patients a high density of complex CB‐ir Ch‐terminals is seen within discrete zones of the dentate granular layer. Whether previously existing CB‐negative Ch‐terminals start to express CB in epilepsy or whether CB‐positive neurons, normally innervating dendritic profiles, reorganize and contact the AISs of granular cells is not known.

The role of CB in neurons is not well understood (Baimbridge, 1992). Many reports consider that it imparts resistance to excessive calcium‐dependent neuronal damage (e.g. Miller and Bainbridge, 1983; Sloviter, 1989; Lledó et al., 1992; Maglóczky et al., 1997). In contrast, other studies suggest that the absence of CB may alleviate calcium‐dependent neuronal damage, and it has therefore been suggested to be neuroprotective (Klapstein et al., 1998; Nägerl et al., 2000). The specific functional role that CB might have in Ch‐terminals contacting granular cells in the epileptic dentate gyrus is at the moment unknown.

CA fields and the subicular complex

As occurs in the dentate gyrus, Ch‐terminals and basket formations showed variable morphological reorganization throughout the CA fields of some epileptic patients. A remarkable feature of this reorganization of PV‐ir processes in all CA fields was that dense basket formations were consistently found associated with some surviving neurons at the borders adjacent to regions of neuron loss. Since axonal sprouting is a common feature of epileptogenic tissue (e.g. de Lanerolle et al., 1989; Sutula et al., 1989; Isokawa et al., 1993; Represa et al., 1994; Schwartzkroin, 1994), one mechanism that might explain the increase in PV‐ir perisomatic terminals could be the sprouting of basket cell axons, or of axons from other interneurons targeting surviving neurons. Alternatively, previously existing PV‐ir axon terminals that lose their targets due to cell death may relocate, provoking these dense accumulations of terminals. In addition, we have seen that, whereas the normal human CA fields lack CB‐ir Ch‐terminals, in certain epileptic patients complex CB‐ir Ch‐terminals are present in discrete zones of CA2 and within zones containing surviving neurons in the CA1 field facing the subiculum.

GABA is the inhibitory neurotransmitter used by the majority of interneurons and thus the changes observed could represent compensatory plastic mechanisms to enhance the inhibition of some pyramidal cells. However, a recent study by Cohen and colleagues raised the intriguing possibility that changes in GABAergic circuits may have excitatory effects in the sclerotic hippocampus of epileptic patients and, thus, increase the hyperexcitability of certain neurons innervated by GABA terminals (Cohen et al., 2002). Their results showed that in the damaged subiculum there is a subpopulation of pyramidal cells in which GABAergic signalling is depolarizing instead of hyperpolarizing. These cells discharge interictal‐like bursts and presumably act as pacemaker cells in generating interictal synchrony (Cohen et al., 2002). Thus, it is possible that in the parahippocampal gyrus abnormal circuits may amplify the relatively few excitatory outputs from the sclerotic hippocampus. In our patients, we did not observe morphological alterations in subicular GABAergic circuits. However, the damaged subicular region reported by Cohen and colleagues may correspond to what we have considered the most distal region of CA1, adjacent to the areas of cell loss, which showed some surviving neurons innervated by dense basket formations and, in some cases, by complex Ch‐terminals.

The degree of cell loss and the alterations in GABAergic circuits were remarkably variable between patients and between different fields of the hippocampal formation within a given patient. Indeed, we did not find a correlation between these abnormalities and any clinical parameter analysed. In a large series of epileptic patients, Mathern and colleagues also found a great deal of heterogeneity in the damage in the hippocampus, but the degree of neuronal loss was independently correlated with factors such as the initial precipitating event and seizure duration (Mathern et al., 2002). Thus, a larger series of patients should be examined to reach more meaningful conclusions regarding the possible relationships between patient history, neuropathology and changes in GABAergic circuits.

In conclusion, alterations in GABAergic circuits that arise in epileptic tissue maybe involved in the generation of epileptiform activity, albeit through different mechanisms. These alterations may include both aberrant GABAergic hyperinnervation in some neurons, and the apparent loss of GABAergic perisomatic innervation in others. Nevertheless, it should be emphasized that neuronal loss and synaptic reorganization are per se not necessarily epileptogenic, and it thus seems clear that particular circuits must be altered to induce epilepsy (DeFelipe, 1999). Therefore, further electrophysiological studies (such as those of Cohen and colleagues (Cohen et al., 2002), in conjunction with correlative microanatomical and neurochemical characterization of hippocampal preparations, will be necessary to define the significance of specific changes in GABAergic circuits in the various fields of the human epileptic hippocampal formation.

Acknowledgements

We thank A. Ortiz for technical help. This work was supported by grants awarded to J.DeF. from the Spanish Ministry of Science and Technology (DGCYT PM99‐0105) and from the Comunidad de Madrid (08.5/0036/2000 and 08.5/0027/2001).

Fig. 1 (A, B) Low‐power photomicrographs from Nissl‐stained sections showing the different fields of the hippocampal formation and subicular region in (A) a normal subject and (B) in the epileptic patient H138. Arrows in A point to the neuronal clusters in the superficial pyramidal layer of the subiculum. Note in B the neuronal loss in the CA4, CA3 and CA1 fields. (C, D) Photomicrographs of dynorphin‐immunostained sections from a (C) normal and (D) epileptic (H138) hippocampal formation, showing mossy fibre reorganization in the epileptic hippocampal gyrus (for higher magnification see Fig. 3H, I). Scale bar: 1200 µm in A, 1000 µm in B, 600 µm in C and D. DG = dentate gyrus; Sub = subiculum.

Fig. 2 Low‐power photomicrographs from sections showing (A, B) the patterns of PV and (C, D) CB immunostaining in the hippocampal formation and subicular region of (A, C) a normal subject and (B, D) epileptic humans (H57 and H136 respectively). Note the general reduction of PV and CB immunostaining in both the dentate gyrus and the CA fields of the epileptic patients. See captions of Figs 4–11 for details. Scale bar: 1000 µm in A and B, 600 µm in C and D.

Fig. 3 Photomicrographs from (A–C) Nissl‐stained, (D–F) NeuN‐immunostained and (G–I) dynorphin‐immunostained sections though the dentate gyrus of (A, D, G) a normal subject and (B, C, E, F, H, I) two epileptic patients. Note the dispersion of (A–F) granular neurons and (G–I) the proliferation of dynorphin‐immunostained mossy fibres in the molecular layer in epileptic patients (B, E, H correspond to patient H115 and C, F and I to patient H94). Scale bar: 72 µm. gran = granular layer; mol = molecular layer; pol = polymorphic layer.

Fig. 4 Photomicrographs from (A–C) GAT‐1‐, (D–F) PV‐ and (G–I) CB‐immunostained sections though the dentate gyrus of (A, D, G) a normal subject and (B, C, E, F, H, I) two epileptic patients, showing the reorganization of the inhibitory elements in epilepsy. In B, C, E and F note the decrease in the density of neuropil GAT‐1 and PV immunostaining in the granular layer of epileptic patients and the appearance of intensely immunostained Ch‐terminals. In H and I note the differential alteration of CB immunostaining in the granular layer of the two epileptic patients. In H, note that the cell bodies of granular cells lack CB immunostaining, whereas in I CB‐immunostaining is retained mostly in the most‐migrated granular cells. (B, E, H correspond to patient H115 and C, F and I to patient H94). Scale bar: 66 µm. gran = granular layer; mol = molecular layer; pol = polymorphic layer.

Fig. 5 Photomicrographs of the granular layer of the dentate gyrus of (A, C, E) a control subject and (B, D, F) epileptic patient H115 immunostained for (A, B) GAT‐1, (C, D) PV and (E, F) CB, showing the reorganization of inhibitory circuits in epilepsy. Note the reduction of CB immunostaining in the granular cell somata and the presence of complex and dense Ch‐terminals intensely immunostained for the three markers (arrows). Scale bar: 13 µm.

Fig. 6 Photomicrographs of (A, B) GAT‐1‐ and (C, D) PV‐immunostained sections showing distribution of Ch‐terminals (arrows) in the (A, C) normal and (B, D) epileptic CA1 field (B corresponds to patient H123 and D to patient H75). Note that the density of labelled boutons that make up Ch‐terminals is increased, in the epileptic tissue (complex Ch‐terminals) compared with the control (simple Ch‐terminals). Scale bar: 70 µm in A and B, 10 µm in C and D.

Fig. 7 Photomicrographs from sections immunostained for (A, C) GAT‐1, (B, D) PV, (E,G) CB and (F, H) PSA‐NCAM, showing the distribution of Ch terminals in the subiculum of normal humans. C, D, G and H are higher magnifications of the boxed areas in A, B, E and F. Ch‐terminals (arrows) are more abundant and intensely stained with the four markers in the superficial than in the deep aspect of the pyramidal layer of the subiculum. Scale bar: 155 µm in A, B, E and F, 40 µm in C, D, G and H.

Fig. 8 (A, B) Photomicrographs of adjacent (A) Nissl‐stained and (B) PV‐immunostained sections of the dentate gyrus of epileptic patient H48, showing the heterogeneity of the alterations of inhibitory circuits along the extent of the dentate granular cells. (C) Higher magnification of the boxed area in B. Note that the presence of complex Ch terminals, intensely immunostained for PV, is not continuous along the granular layer; there are microzones (asterisk) that represent a reduction in PV immunostaining and in which no, or very few, PV‐ir Ch‐terminals are labelled. This reduction in immunostaining is not associated with a reduction in granular cells. Scale bar: 150 µm in A and B, 30 µm in C. gran = granular layer; mol = molecular layer; pol = polymorphic layer.

Fig. 9 Photomicrographs taken from sections though the granular layer of the dentate gyrus of patient H94 showing examples of (A) simple and (B) complex PV‐immunostained Ch‐terminals. Note that density of labelled boutons is increased in the complex Ch‐terminal compared with the simple Ch‐terminal. Scale bar: 5 µm.

Fig. 10 Photomicrographs of two adjacent sections of the hippocampus of the H48 epileptic patient (A) stained for Nissl or (B–E) immunostained for PV, showing the heterogeneity of the alterations of inhibitory circuits in the hilar/CA4 region. The boxed area in B is enlarged in C. Note that, within the same hippocampal field, the surviving neurons of adjacent zones are differentially innervated by PV‐ir axons. Arrows in C and D point to neurons surrounded by dense PV‐ir perisomatic basket formations and Ch‐terminals at the AIS. Outline arrows in C and E point to Ch‐terminals presumably contacting the AIS of neurons which lack PV‐ir basket formations. Scale bar: 485 µm in A and B, 75 µm in C, 30 µm in D and E.

Fig. 11 Photomicrographs from two adjacent sections from patient H48 immunostained for (A, C) NeuN and (B, D, E) PV. The boxed areas in A, B and D are shown at a higher magnification in C, D and E respectively. Note (A, C) the neuronal loss in the CA1 region and (D, E) the presence of abnormally dense basket formations, intensely immunostained for PV around some surviving neurons at the border between the CA1 and the subiculum. Scale bar: 585 µm in A and B, 150 µm in C and D, 15 µm in D.

Table 1

Summary of clinical and surgical data

Patient	Age (years), sex, side	Age of onset, duration (years)	Possible precipitating event	Status epilepticus (age, duration)	Seizure type	Seizure frequency	Engel scale for surgical outcome (time after surgery in months)
H44	33, M, L	2, 31	Head trauma, 2 years	No	PC, gen	1–3/week	I (80)
H48	41, M, L	18, 23	None	No	PC, gen	1/week	I (79)
H57	27, M, R	13, 14	Meningitis, 6 months	12 years, 40 min	PC, gen	2–3/week	I (74)
H61	17, F, R	7, 10	Febrile seizures, 6 and 9 months	No	PC	2/week	I (59)
H75	37, M, L	13, 24	Febrile seizures, 5 months	5 months, 4 h	PS, PC	1–3/week	II (57)
H84	31, M, R	2, 29	Febrile seizures, 10 months	10 months, 5 h	PC, gen	3–4/week	I (25)
H94	27, M, L	20, 7	Birth problems, head trauma 3 weeks before seizures	No	PC, gen	3–5/week	I (43)
H104	32, M, L	12, 20	Febrile seizures, 10 months	No	PC	1–3/week	I (15)
H108	50, M, L	15, 35	None	No	PC, gen	3–4/week	III (18)
H109	22, F, R	4, 18	Birth problems, possible encephalitis	No	PC, gen	0–3/week	I (40)
H115	40, F, L	1.8, 38	None	No	PC, gen	2–5/week	III (39)
H123	24, F, L	7, 17	Birth problems, febrile seizures, 3 years	No	PC, gen	20–25/week	I (31)
H136	20, F, R	0.7, 19	Febrile seizures, 8 months	16 years, 2.5 h	PC, gen	0.5/week	I (28)
H138	41, F, L	1.4, 40	Febrile seizures, 17 months	17 months, 6 h	PC	4–6/week	II (24)

M = male; F = female; R = right; L = left; PC = partial complex; PC, gen = partial complex, sometimes secondarily generalized. Engel scale (Engel, 1987) = class I, seizure‐free; class II, rare seizures; class III, worthwhile improvement.

Table 2

Quantitative analysis of hippocampal alterations

Patient	Thickness (µm)		Neuronal density (%)
	Granular layer	Aberrant mossy fibres	Granular layer	Hilus	CA4	CA3	CA2	CA1
H44	117	113	32***	17***	33***	50***	96	56***
H48	174***	173	55*	64	69	79	79	39***
H57	260***	306	79	13***	17***	35***	55***	35***
H61	282***	310	59	114	118	99	103	37***
H75	351***	388	99	36***	38***	46***	48***	31***
H84	335***	198	28***	10***	14***	NA	14***	20***
H94	299***	271	90	37***	24***	73	101	28***
H104	318***	268	67	29***	32***	NA	55***	33***
H108	337***	198	48***	13***	14***	27***	41***	33***
H109	213***	268	87	94	82	105	97	46***
H115	278***	198	45***	22***	27***	43***	66***	23***
H123	222***	225	99	42**	80	61	79	37***
H136	316***	268	87	20***	54*	45***	45***	31***
H138	233***	225	81	47*	27***	40***	72*	38***

The table shows the thickness of the granule cell layer, the extent of aberrant mossy fibres and percentages of neuronal density in epileptic patients compared with controls. Thickness of control granular layer was 89 µm. Statistically significant differences compared with control values: *P < 0.05; **P < 0.01; ***P < 0.005. NA = no data available.

Table 3

Distribution and relative abundance of GAT‐1 and CB immunostained Ch‐terminals in the dentate gyrus and CA fields in epileptic patients

Patient	Granular layer		Hilus and CA4		CA3		CA2		CA1
	GAT‐1	CB	GAT‐1	CB	GAT‐1	CB	GAT‐1	CB	GAT‐1	CB
H44	NA	0	NA	0	NA	0	NA	0	NA	+C
H48	+C	0	+	0	+	0	++	0	+++C	+C
H57	+	0	0	0	0	0	+C	0	+++C	++
H61	0	0	0	0	0	NA	0	NA	+	NA
H75	0	0	0	0	0	0	+	0	+	++
H84	+C	+C	0	0	NA	NA	0	+	+++C	+C
H94	+C	0	+C	0	+C	+	++C	+C	++C	+
H104	++	0	+C	0	NA	NA	+C	+C	++C	++C
H108	++C	+C	+C	0	+C	0	++C	0	++C	+
H109	0	+C	+	0	++C	0	+	0	++C	+
H115	+++C	++C	0	+C	NA	NA	+C	++C	+	+
H123	+	+C	0	0	0	0	+	0	++C	+
H136	0	0	0	0	0	0	+C	0	++C	+
H138	0	0	0	0	0	0	+C	0	++C	+

Density of Ch‐terminals: 0 = no staining, +++ = high density of Ch‐terminals. Hilus and CA4 were considered together. C = presence of complex Ch‐terminals; NA = no data available.

Table 4

Relative abundance of PV immunostaining in epileptic patients

Patient

Dentate gyrus

Hilus and CA4

CA3

CA2

B

Ch

N

B

Ch

N

B

Ch

N

B

Ch

N

H44

+

+

+

NA

NA

NA

NA

NA

NA

NA

NA

NA

H48

++

++C

+

++

++C

+

+

+C

‐

++

+++C

++

H57

+

+C

++

0

0

+

0

0

0

0

+C

0

H61

++

+

+

0

+C

+

0

0

+

NA

NA

NA

H75

++

+C

++

+

+C

+

+

+C

+

+

+C

+

H84

++

++C

++

+

0

0

+

+C

+

++

++C

+++

H94

+

+C

+

+

+C

+

+

+C

+

+

+C

+

H104

+

+C

+

+

+C

+

NA

NA

NA

++

++C

+

H108

+

+C

+

0

0

0

0

0

0

0

+

0

H109

++

++C

++

+

+C

+

++

0

++

++

+

++

H115

+

++C

+

0

0

+

NA

NA

NA

+

++C

+

H123

+

+C

+

0

+C

+

0

0

0

+

+

+

H136

++

+

+

+

0

+

+

0

+

+

0

+

H138

+

+C

+

0

0

+

0

+C

0

+

++C

+

The table shows basket formations (B), Ch‐terminals (Ch) and other axonal and dendritic processes in the neuropil (N) in the dentate gyrus, hilus, CA4, CA3 and CA2. Density of immunostained elements: 0 = no staining, +++ = high density of labelled elements. C = presence of complex Ch‐terminals; NA = no data available.

Table 5

Distribution of complex Ch‐terminals in the granular layer of the dentate gyrus of epileptic patients, as revealed with different markers

Patient	GAT‐1	PV	CB
H44	NA	0	0
H48	Focal	Focal	0
H57	0	Focal	0
H61	0	0	0
H75	0	Extensive	0
H84	Focal	Extensive	Focal
H94	Focal	Extensive	0
H104	0	Focal	0
H108	Focal	Focal	Focal
H109	0	Focal	Focal
H115	Focal	Extensive	Extensive
H123	0	Focal	Focal
H136	0	0	0
H138	0	Focal	0

NA = no data available; 0 = no labelling; Focal = complex Ch‐terminals were distributed focally within an extent <50% of the total extent of the dentate gyrus; Extensive = distributed along >50% of the length of the dentate gyrus.