Several lines of evidence suggest that characteristic pathologic changes are present in the entorhinal cortex (EC) in patients with hippocampal sclerosis (HS) and that this region may have importance in either the initiation of temporal lobe seizures or development of HS. Neuroimaging studies have reported volume reduction of parahippocampal gyral structures in TLE, mainly ipsilateral to the seizures (1–4). Abnormal epileptiform activity has been recorded in the EC region (5,6), which may sustain seizures. Anatomically, the EC has reciprocal connections with the hippocampus. Neurones from superficial layers (mainly layers II and III) send glutamatergic afferents, via the perforant pathway, to the dentate granule cells and CA1 neurones. Subicular and CA1 pyramidal neurones have feedback connections to the deeper layers of the EC. In pathologic studies of animal models of TLE, and after status epilepticus, vulnerability of neurones in layer III has been shown (7–10). However, relatively few neuropathologic studies of the EC have been performed in human tissues, and quantitative data are lacking. Our aim was to carry out a stereologic quantitative estimation of neuronal densities in the EC region to assess the degree of any neuronal loss in patients with HS in comparison to other causes of TLE. We aimed to compare neuronal loss in the EC with loss in CA1 and subicular regions as well as the neocortex to identify any stereotypical patterns characteristic of HS.
Summary: Purpose: Clinical, radiologic, and experimental evidence indicates that the entorhinal cortex (EC) region may be linked to the pathophysiology of hippocampal sclerosis (HS) in patients with temporal lobe epilepsy. Few neuropathologic studies of this region have been undertaken in patients with HS undergoing surgery, some suggesting preferential loss of layer III neurones.
Methods: We carried out a quantitative analysis in 26 patients with HS, nine patients with lesional temporal lobe epilepsy (LTLE), and eight postmortem controls. We measured neuronal densities in EC by using a three-dimensional cell-counting technique on NeuN immunostained and Nissl-stained sections. We also quantified the density of calretinin-positive interneurones in this region and the density of neurones in adjacent subiculum and CA1 subfields. We also assessed the patterns of gliosis in the EC in the patient groups and the presence of any neocortical neurone loss.
Results: No significant difference was found in the mean neuronal densities in the EC region between HS and LTLE groups or postmortem controls. Laminar gliosis in midcortical layers was seen in a proportion of HS cases but also in the LTLE group. No significant difference was seen in the density of calretinin interneurones and no correlation between the presence of neocortical neuronal loss and EC neuronal densities.
Conclusions: A stereotypical pattern of neuronal loss and gliosis in the EC region in patients with HS is not confirmed that distinguishes this pathologic process from that in patients with lesional TLE.
METHODS AND MATERIALS
Cases were selected from the pathology archives in the Division of Neuropathology, National Hospital for Neurology and Neurosurgery, of patients undergoing anterior temporal lobectomy and hippocampal resection for the treatment of refractory TLE. This neuropathologic study has been approved by the National Hospital for Neurology and Neurosurgery and Institute of Neurology Joint Research Ethics Committee. From the period from 1994 to 2000, 26 cases with HS were selected, representing a 14% sample of all HS cases during this period. In all cases, 3–5 cm of anterior temporal lobe was resected with separate resection of 2–3.5 cm of hippocampus as a second specimen. In all these cases, routine neuropathologic diagnosis on serial coronal sections stained with Nissl (cresyl violet/LFB), glial fibrillary acidic protein (GFAP) and NeuN-stained sections was carried out, and a diagnosis of classic HS was made with neuronal loss in CA1 and to a lesser extent in the hilus. In addition, nine patients were selected who had undergone hippocampal resection for TLE due to a “lesional” neocortical temporal lobe pathology (LTLE) (Table 1) and eight postmortem control cases from patients without an epilepsy history as comparison groups. The age range of the patients was 20–56 years (mean, 34.3 years) in the HS group, 15–63 years (mean, 32.1 years) in the LTLE group, and 34–85 years (mean, 68 years) in the nonepilepsy group. All surgical cases were selected from the file according to whether regions of the entorhinal cortex (EC), subiculum (SC), and CA1 were present in resected tissues.
|Case||Main neuropathologic diagnosis||Severity of hippocampal damage||Localization of lesional pathology|
|1||Cavernoma||1||Lateral temporal lobe; intracortical|
|2||Old cortical infarction||2||Temporal and frontal|
|3||Old cortical infarction||3||Parietal lobe and temporal lobe gliosis|
|4||Chronic encephalitis (Rasmussen-like end stage)||3||Temporal and parietal|
|5||Low grade glioneuronal tumor||3||Uncus|
|6||Cavernoma||1||Lateral temporal lobe|
|7||Old cortical infarction||4a||Superior temporal gyrus and central resection|
|8||Old cortical scar||2||Temporal and frontal cortex|
Sections from paraffin blocks in each case were cut at 25-μm thickness and stained with Luxol fast blue/cresyl violet, GFAP (Dako, Cambridge, U.K.; polyclonal 1:1,500) counterstained with Nissl (cresyl violet), neuronal nuclear antigen (NeuN; Chemicon, Temecula, CA, U.S.A., monoclonal 1:2,000), and calretinin (CR) (Sigma, St. Louis, MO, U.S.A., monoclonal, 1:4,000). Endogenous peroxidase activity was blocked by incubation with 10% hydrogen peroxide for 10 min. Antigen retrieval was performed by boiling sections in 0.01 M citrate buffer (pH 6.0) in a microwave oven, and the immunostaining detected by using streptavidin peroxidase (Dako) and diaminobenzadine (DAB) as the chromagen. On analysis of new sections, all three regions (EC, SC, CA1) were suitable for quantitative analysis in 18 of 26 HS cases and seven of nine non-HS cases. In a further four HS cases, only EC and CA1 regions were suitable for analysis, and SC and CA1 regions in a further four cases; in two of nine of the non-HS cases, only SC and CA1 regions were analysed.
An image-analysis system (Histometrix, Kinetic Imaging, Liverpool, U.K.) was used for the quantitative analysis linked to Zeiss Axioskop microscope with a motorized stage, using a Plan Apo oil-immersion objective lens of magnification ×63, numeric aperture 1.4. The anatomic regions EC, SC, and CA1 were first outlined at ×2.5 magnification on NeuN-stained sections. The EC was identified by its distinctive laminar architecture and prominent clusters of neurones in layer II (Figs. 1 and 2A). The full thickness of the cortex from the pial margin to the white matter was included, and the entire area of EC outlined on one section per case was analyzed. In each case, the regions of SC included in this analysis were defined as the pyramidal cell layer just medial to superficial neuronal clusters of the parasubiculum (11), with the hippocampal fissure forming the upper, and white matter, the deep boundaries (Fig. 1). The area of CA1 included the pyramidal cell layer adjacent to the lateral ventricle (with care not to include adjacent CA2 sector or SC bordering either side) (Fig. 1). A 3D cell-counting method was used on the NeuN-stained sections (optical disector method) and in pilot studies on the EC, CA1, and SC in four cases, a 10% uniform randomized unbiased field sampling gave reproducible neuronal density estimations both on recounting by a single observer (S.D.) and between two observers (S.D. and M.T.). Numeric values were expressed as neurones per mm3. In eight HS cases and postmortem controls, neuronal densities in the EC region also were estimated on Nissl-stained sections by using a similar method. Automated 2D cell counts also were carried out on NeuN sections of SC region and on CR-stained sections of both SC and EC regions, with a ×10 objective and 100% sampling of the delineated areas of SC and EC; all immunopositive cells within this region were counted and expressed as cell number per mm2.
The GFAP/Nissl sections were assessed qualitatively in the EC region for the presence and pattern of gliosis in each lamina. In addition, five HS and five LTLE cases were selected for assessment of neuronal density and size in the superficial EC layers. Cortical layers I–III were outlined, and neuronal density and mean neuronal volumes in this region were estimated by using the nucleator method with a mean disector number of 46 and a mean of 109 neurones measured per case. Neurones were identified as cells with prominent nucleoli, open chromatin, and Nissl-positive cytoplasm; small glial cells and GFAP-positive cells were excluded. In all HS cases, the severity of neuronal loss and gliosis also was assessed in the lateral temporal neocortex on Nissl-, NeuN-, and GFAP-stained sections. Statistical analysis to compare neuronal densities was carried out by using SPSS for Windows, version 10, by using the independent t test and Pearson's correlation.
In all cases, it was not possible to appreciate neuronal loss clearly, particularly laminar neuronal loss, on either Nissl or NeuN sections in the EC by qualitative analysis alone (Fig. 2A and B). This was in contrast to the more obvious neuronal loss from CA1 in HS cases and the temporal neocortical neuronal loss in a proportion of cases (Fig. 2C). After quantitative analysis, significantly lower neuronal densities in CA1 in the HS group compared with the LTLE cases were confirmed (Table 2), with mean densities of 0.85 compared with 1.81 × 104/mm3. However, no significant difference in the mean neuronal densities was measured in the EC between the groups, with slightly higher densities in the HS group of 2.6 compared with 2.4 × 104/mm3 in the LTLE group. Analysis of EC neuronal densities in eight HS cases and postmortem controls on Nissl-stained sections also showed no significant difference between these groups, with mean densities of 3.02 and 2.7 × 104/mm3, respectively (Table 2). A significant correlation in neuronal densities was measured in these eight HS cases on Nissl- and NeuN-stained sections, with mean NeuN neuronal densities in this selected group of 2.9 × 104/mm3. The analysis of neuronal densities in the superficial EC laminae also showed no significant difference between HS and LTLE groups (Table 3). Measurements of neuronal volume did, however, confirm a significant difference between the groups, with greater mean neuronal volumes in the HS group (1.26 vs. 0.68 × 103μm3) (Table 3). No difference was seen in the mean EC neuronal densities for lobectomies involving the left or right side in either HS or all TLE cases (2.4 × 104/mm3 on the right, 2.7 × 104/mm3 on the left for all cases). Similarly, no significant difference was noted in the mean neuronal densities in the SC region between the two groups (2.04 in the HS group vs. 1.88 × 104/mm3 in the LTLE group). However, a correlation was found between NeuN neuronal density measurements in the EC and CA1 in all cases (p < 0.05). A correlation also was found between neuronal density measurements made with 2D and 3D counting techniques in the SC on NeuN-stained sections (p < 0.01; Table 2).
Mean (SD); range × 104/mm3
Mean (SD) range × 104/mm3
Mean (SD) range × 104/mm3
|CA1 NeuN||0.85 (0.59)¶; 0.12–2.62||1.81 (0.82); 0.32–2.81|
|Subiculum (SC) NeuN||2.04 (0.67); 0.86–3.78||1.88 (0.79); 1.03–3.62|
|3.5 (0.78)a; 2.15–5.35||3.0 (0.43)a; 2.4 –3.67|
|Entorhinal cortex (EC) NeuN||2.6 (0.7); 2.36–4.26||2.4 (0.64); 1.28–3.18|
|Calretinin SC||0.22 (0.13)a||0.17 (0.14)a|
|Calretinin EC||0.25 (0.1)a; 0.2–0.51||0.35 (0.7)a; 0.15–0.6|
|EC Nissl||3.02 (0.91); 1.6–4.6||2.7 (0.58); 2.2–3.9|
|HS (n = 5)||LTLE (n = 5)|
|Neuronal density (EC: layers I–III): Nissl|
|Mean (SD); range × 104/mm3||3.2 (0.74); 2.3–4.7||3.8 (2.03); 1.3–6.9|
|Neuronal volume (EC: layers I–III): Nissl|
|Mean (SD); range × 103/μm3||1.26 (0.35); 0.96–1.76a||0.68 (0.23); 0.33–0.9a|
|Total EC neuronal density: NeuN|
|Mean (SD); range × 104/mm3||2.3 (0.51); 1.67–2.81||2.4 (0.81); 1.2–3.8|
Variable patterns of gliosis were seen in both HS and LTLE cases in the EC. In some cases, gliosis predominated in the superficial laminae around clusters of neurones in layer II (Fig. 3A), and in others, gliosis affected deeper cortical layers, merging with the white matter (Fig. 3B). In nine cases, including one LTLE case, a prominent laminar band of astrogliosis involved layer III/IV (Fig. 3C). The presence of this feature correlated with lower EC NeuN neuronal densities (p < 0.05), but in none of these nine cases was there evidence of obvious laminar neuronal loss or gliosis of the lateral temporal neocortex. Conversely, only three of all HS cases studied showed evidence of obvious neocortical neuronal loss (Fig. 2C), and in these cases, EC neuronal densities were not the lowest values measured (2.81–3.72/mm3 compared with a mean of 2.6/mm3 for all HS cases).
The immunohistochemistry for CR in the EC and SC regions in both groups showed a staining pattern as described in normal controls. In the EC region, positive cells predominated in the superficial cortical layers, these being mainly small bipolar interneurones with radial processes, and larger immunoreactive multipolar neurones were observed in deeper cortical laminae. After quantitative evaluation, no significant difference was observed in the number of neurones in the HS and LTLE groups in EC and SC regions, although we noted a trend for fewer inhibitory interneurones in the EC in HS cases (Table 2).
In 1993, Du et al. (9) reported the finding of striking neuronal loss and gliosis in layer III and, to a lesser extent, layer II of the EC in patients with HS and a reduction of the thickness of this cell layer. This was suggested to represent a consistent neuropathologic pattern and to reflect a role of the EC in the genesis of HS. However, this study was based on just four cases, and it remains unclear how common this neuropathologic finding is and how specific this pathologic feature is for HS, as distinct from other TLE-associated pathologies. For example, a more recent study of the EC by Yilmazer-Hanke (12) suggested more variable patterns of gliosis and neuronal loss in the EC in patients with TLE and that neuronal loss of an entire lamina was relatively uncommon. In their study, qualitative evaluation suggested neuronal loss in seven of nine patients with HS and in three of five patients without. An earlier quantitative study of the parahippocampal gyral cortex in patients with HS had suggested no significant neuronal loss from this region compared with controls (13).
Our aim was to carry out a stereologic quantitative analysis of the EC region in surgical resections from patients with TLE. In the cases examined from patients with either HS or LTLE, which were selected randomly from our archive, evidence of neuronal damage or loss in the EC was not easily confirmed by qualitative examination alone, either in conventional Nissl- or NeuN-stained sections. NeuN, a pan-neuronal marker that labels all neurones, including interneurones, has been shown as useful in the assessment of cytoarchitecture as well as for determining neocortical cell loss (14,15). Our quantitative analysis failed to discriminate any differences in the degree of neuronal loss in the EC in the HS group compared with LTLE cases. As this region is rarely removed in other surgical procedures, a nonepilepsy surgical control group was not available to compare the extent of any neuronal loss that may occur in TLE. We noted in comparison with postmortem control tissues that no significant difference appeared in neuronal densities in the EC compared with HS cases. Gliosis, however, was a common finding in the EC in TLE patients, which indirectly supports damage to this region, and in patients with the most severe gliosis, lower neuronal densities in this region were noted. The patterns and distribution of gliosis in laminae of the EC in HS cases was, however, variable, and again, not distinctive from the LTLE group. This would also argue against a stereotypical or laminar-specific vulnerability of this region in HS. Variability in the gliosis of the EC region has been previously noted (12). Similar to earlier quantitative studies, we also did not observe any difference in neuronal densities in the SC in the HS compared with the LTLE control group (13,16).
The continued interest of the potential participation of the EC in the pathophysiology of HS and TLE has largely arisen from animal studies. In several animal models of TLE, for example, after administration of excitotoxins, loss of neurones in layer III of the EC is demonstrated, with preferential survival of inhibitory interneurones (8,10,17,18). This region has been considered to play a critical role in the kindling of seizures, as well as potentially in the propagation or generalization of temporal lobe seizures (10,19). In support of this, abnormal discharges have been demonstrated in the EC in patients with TLE and in experimental systems, which may sustain or initiate seizures (5,6,20). The EC at the junction between hippocampus and neocortex acts as a conduit for incoming information and for reciprocal outgoing signals. It also contributes to local signal processing and modulation, and intercortical networks between the deep and superficial cortical layers have been shown (21). Layer II afferents project to the dentate gyrus via the perforant pathway and layer III neurones to CA1 neurones. Fibres from CA1 and SC project mainly to deeper layers of the EC, laminae V, and VI (22). Any layer-specific neuronal loss in the EC region may therefore result in alterations to these local circuits, which may potentially influence seizure spread (18). In this present study, however, we failed to demonstrate preferential loss of superficial entorhinal cortical neurones in HS versus LTLE cases.
In surgical specimens, the laminar and myeloarchitecture of the EC is easily distinguished from the neocortex. In postmortem studies, the human EC has been further divided into subregions based on subtle variations in this cytoarchitecture (23,24). In the surgical specimens, only a small region of the medial EC, adjacent to the hippocampus, is available for analysis. We cannot exclude that a variation exists in the extent of any neuronal loss in different EC subregions associated with epilepsy, as has been suggested in MRI studies (1–4). The discrepancy between the findings in our study and that of Du (9), regarding the presence of laminar specific neuronal loss, may therefore relate to different regions of EC being included in the resection specimen. These issues could be addressed in the future by postmortem studies of the entire EC region in patients with HS.
It is possible that reduction in mesial temporal tissue volume in the HS, compared with that in the LTLE group, may have artificially increased neuronal density measurements in our study using the optical dissector method. It is not possible from the surgically resected specimen to get an accurate measure of EC volume from which to calculate total neuronal number. However, we noted from our results that cases with more severe EC gliosis, an indirect measure of tissue sclerosis, had lower neuronal densities. It also has been suggested that neuronal hypertrophy occurring in TLE may bias density measurements (25). In the present study, we did confirm significantly greater neuronal size in the HS compared with LTLE cases, which may also have influenced results. Furthermore, the older age range of postmortem controls compared with surgical cases may be another possible factor influencing our findings, in relation to any age-related neuronal loss in this region.
In the analysis of CR-immunopositive neurones, we adopted 2D counting methods, because of the relatively lower numbers of these interneurones. We noted a good correlation between 3D and 2D counting methods on NeuN-stained sections, which validated our technique. The overall distribution and morphology of CR-immunopositive interneurones was as described in normal human EC (26), and although a suggestion was noted for lower numbers in HS cases compared with LTLE cases, the differences were not significant, and the ratio of CR- to NeuN-positive neurones was not significantly different between the groups (data not shown). Although evidence exists from animal epilepsy models of preferential survival of inhibitory interneurones in the EC, it is still possible that abnormal activity of the layer II projection neurones results from reduced inhibition (19). Interestingly, it has been shown in the EC of rat that a subpopulation of CR-positive interneurones is excitatory rather than inhibitory (27). Establishing whether similar interneurones are present in human EC or vulnerable in HS may important to the understanding of any alterations in local circuitry and neuronal excitability.
Similar to the study by Du et al. (9), in our patients we found no evidence to correlate the presence of gliosis or neuronal loss in the EC with the presence of neocortical neuronal loss, which typically involves cortical layers II/III (15,28,29). This suggests a different pathogenesis for loss of neurones in these two regions, which may be dependent on clinical parameters including duration of seizures, seizure type, and electrophysiologic characteristics, in addition to any differences in antiepileptic drug regimens, all of which were not explored in this present study. Future similar larger pathological studies with clinical and quantitative radiologic correlations could address these issues. In conclusion, in this neuropathological study, we demonstrated that laminar neuronal loss in the EC is not invariably seen in patients with HS, is less common than observed in animal models, and that overall neuronal densities and patterns of gliosis of this region are not different from those in patients with other causes of TLE.
Acknowledgment: The Kinetic Imaging, Image Analysis system in the department of Clinical and Experimental Epilepsy was purchased through a grant funded by the Wellcome Trust.