Investigation of widespread neocortical pathology associated with hippocampal sclerosis in epilepsy: A postmortem study

Authors


Address correspondence to Maria Thom, Department of Clinical and Experimental Epilepsy, Division of Neuropathology, Institute of Neurology, Queen Square, London WC1N 3BG, U.K. E-mail: m.thom@ion.ucl.ac.uk

Summary

Purpose: One possible cause for surgical failure following temporal lobectomy for the treatment of epilepsy due to classical hippocampal sclerosis (CHS) is the presence of more widespread cortical changes. Neocortical changes in CHS shown by quantitative neuroimaging studies may involve hippocampal projection pathways. Our aim was to quantitate neocortical pathology using a postmortem series of brains from patients with epilepsy and CHS.

Methods: Sections from 13 cortical regions from both left and right hemispheres, including hippocampal projection pathways, were examined from nine epilepsy patients with unilateral CHS (4), bilateral CHS (2), non-CHS (3), and non–epilepsy controls (4). Using GFAP, CD68, and NPY immunohistochemistry as markers of acquired neocortical pathology, quantitative analysis of the staining fractions in the cortex and white matter was carried out.

Key Findings: Higher staining fractions were observed for all markers in both cortex and white matter in CHS patients, which was significantly different for CD68 and NPY compared to controls (p < 0.05) but not to non-CHS epilepsy cases. There was no significant difference between staining fractions in left and right hemispheres for unilateral CHS cases. Regional analysis showed preferential gliosis and microgliosis of temporal poles, frontal poles, and orbitofrontal cortex in epilepsy cases.

Significance: This study supports acquired neocortical pathology in epilepsy patients both with and without CHS. Cortical pathology does not show lateralization to the side of CHS. Preferential involvement of the temporal and frontal poles may relate to other factors, such as cortical injury associated with seizures, rather than involvement through hippocampal pathways.

Classical hippocampal sclerosis (CHS) is associated with epilepsy, in particular mesial temporal lobe epilepsy (mTLE), and in many cases appears to be an isolated, unilateral pathology. One-third of patients continue to have seizures following temporal lobe surgery (Engel, 1996; Foldvary et al., 2000; McIntosh et al., 2001; Spencer & Huh, 2008). There are several possible explanations including incomplete resection of the epileptogenic region or extratemporal neocortical areas becoming independently epileptogenic (Lin et al., 2007; Bertram, 2009; Keller et al., 2009).

Quantitative neuroimaging studies have identified more widespread cortical changes in patients with mTLE (Cormack et al., 2005; Bonilha et al., 2007; Lin et al., 2007; McDonald et al., 2008; Riederer et al., 2008; Bernhardt et al., 2009; Keller et al., 2009; Labate et al., 2009). However, the cellular basis of extratemporal neocortical neuroimaging changes has never been investigated. Even in surgical temporal lobectomy specimens adjacent to CHS, where more direct radiologic–pathologic correlation is possible, data regarding the pathologic basis of varied neocortical neuroimaging volume and signal changes are inconclusive. Reduction of cortical neuronal number (Eriksson et al., 2009), myelin abnormalities (Meiners et al., 1999; Mitchell et al., 1999), and white matter dysplasia (Choi et al., 1999) have been proposed to relate to imaging measures. Further evaluation of the cellular basis and distribution of any extrahippocampal pathology in patients with CHS is essential to the understanding of how it might contribute to independent epileptogenic foci.

Postmortem tissues from patients with long histories of drug-resistant epilepsy with CHS as the only identified epilepsy-specific pathology provide an opportunity to evaluate extratemporal neocortical pathology. Our aim was to carry out such an investigation using quantitative immunohistochemistry.

Methods

Case selection

Postmortem cases were selected from the archives of the Division of Neuropathology at the National Hospital for Neurology and Neurosurgery, Queen Square, London. The local ethics committee of the National Hospital for Neurology and Neurosurgery and the Institute of Neurology has given approval for neuropathologic epilepsy studies, and era-appropriate consent was obtained from patients’ next of kin. All patients had histories of drug-resistant epilepsy and all but one patient (EP031) had been resident at the National Society for Epilepsy, Chalfont Centre (Bucks, United Kingdom).

Six cases (age range 31–84 years; mean age 54 years) with confirmed CHS, three cases without CHS (age range 41–48 years; mean age 45 years), and four non–epilepsy controls (age range 36–78 years; mean age 57 years) without significant neurologic disease or epilepsy were included (Table 1). The CHS was confirmed as unilateral CHS in four cases and bilateral in two using a stereologic quantitative method as described previously (Thom et al., 2005) (Fig. 1C,D, Table 2). In addition, mossy fiber sprouting was demonstrated using dynorphin immunohistochemistry for confirmation of epilepsy-associated reorganization (Thom et al., 2009b) and to exclude other causes of acquired hippocampal atrophy (Table 2, Fig. 1E,F). We excluded any case with significant second or acquired pathology, for example, contusional brain injury, cerebrovascular disease, or neurodegenerative disease. We included two cases (EP019, EP038) in which there was macroscopic and magnetic resonance imaging (MRI) impression of ipsilateral hemispheric cortical, mammillary body and thalamic atrophy, in addition to CHS (Fig. 1A,B). One patient (EP294) with postmortem proven bilateral CHS had previous unsuccessful temporal lobe surgery for right CHS (Table 2). The controls were selected from the same era so that fixation times were comparable. Mean fixation times were 6.6 years (1–14 years) for the CHS cases, 6.6 years (3–9 years) for non-CHS cases, and 10.25 years (10–11 years) for controls (Table 2).

Table 1.   Clinical seizure history, epilepsy syndrome, and cause of death
GroupPM numberSexAge at death (years)Duration of epilepsyAge of onsetIPIEpilepsy syndromeSeizure type(s)
Frequency prior to death
EEG summary (most recent EEG report prior to death)Cause of death
  1. CHS, classical hippocampal sclerosis; CPS, complex partial seizures; SPS, simple partial seizures; SGTCS, secondary generalized seizures; SE, status epilepticus; IPI, initial precipitating injury for seizures; UK, unknown; HS, hippocampal sclerosis; mTLE, mesial temporal lobe epilepsy; SUDEP, sudden and unexpected death in epilepsy; IHD, ischemic heart disease; PMC, postmortem control; EP, epilepsy postmortem.

CHSEP019M7574 years1.5 yearsPostvaccination encephalitisUKSPS, CPS, SGTCS
Well controlled 50 years prior to death
Nonspecific EEG changes (24 years)Congestive cardiac failure
EP038M7562 years13 yearsFollowing trauma?Symptomatic focal epilepsy (extratemporal)CPS, SGTCS, SE
GS well controlled 4 years prior to death
Interictal epileptiform discharges, Right temporal> Left. No seizure recorded. (27 years)Pulmonary edema, esophageal stricture
EP266M3831 years7 yearsUKmTLECPS, GS
Refractory epilepsy
No reports availableSUDEP (awaiting epilepsy surgery)
EP141M8255 years26 yearsFollowing trauma?UKCPS, SGTCS
Well controlled prior to death
Right and left temporal interictal epileptiform discharges. No seizure recorded. (13 years)IHD, pancreatic cancer
EP294M5430 years24No IPImTLE? (right temporal lobectomy age 43 years)CPS, SGTCS
CPS continued year prior to death
Right temporal interictal focus/nonspecific abnormalities over right centro-temporal region. No seizure recorded. (2 years)No neuropathological cause of death (limited PM to head only)
EP286M31UKUKUKPartial epilepsy with left CHSCPS, SGTCS, SE
Frequent seizures prior to death
Left temporal intermittent interictal slow activity, bilateral frontotemporal interictal epileptiform discharges. No seizure recorded. (8 years)SUDEP
No CHSEP210M4129 years12 yearsNo IPIUKGS, myoclonic seizures
Refractory epilepsy
Right frontotemporal interictal epileptiform discharges. No seizure recorded. (5 years)SUDEP
 EP290F4847 years6 monthsNo IPIRefractory partial epilepsyPS, SGCTS
Refractory epilepsy prior to death
Right temporal interictal epileptiform discharges. No seizure recorded on EEG. (9 years)SUDEP
EP039F4648 years3 monthsNo IPIDravet’s syndromeCPS, GS, SE
Increased frequency prior to death
Interictal epileptiform discharges, multifocal, bilateral. No seizure recorded on EEG. (1 year)SE
ControlsPMC1F57   Not applicable  Pancreatitis
PMC2F78Pancreatic cancer
PMC3F58IHD
PMC4F36Cardiac arrest
Figure 1.


(A) Coronal slice in case EP038 at the level of the red nuclei with confirmed bilateral hippocampal sclerosis and evidence of atrophy of the left hemisphere (shown on left side), particularly involving the temporal lobe. (B) Coronal slice in EP019 at the level of the pulvinar with unilateral left hippocampal sclerosis, more subtle ipsilateral hemispheric atrophy, and more visible atrophy of the ipsilateral thalamus. (CF) EP141 with unilateral hippocampal sclerosis on the left side (C, E) with normal hippocampus on the right (D, F) as shown with cresyl violet (C, D) and dynorphin immunohistochemistry (E, F) confirming mossy fiber sprouting (grade 3) on the sclerotic side and a normal pattern on the right. (G) Section of left temporal lobe pole with GFAP staining showing the pattern of superficial gliosis in case EP294 and in (H) case EP038 where a more laminar pattern of gliosis was noted in the temporal pole in layers II/III. (I) GFAP staining showing white matter gliosis with autodetection of the same fields shown in (J). (K) Labeling with CD68 in the white matter and (L) NPY-immunolabeling in the cortex. Bars = CF = 1,000 microns, G = 100 microns, H = 100 microns, IL = 25 microns.

Table 2.   Neuropathologic details of postmortem hippocampal sclerosis and control cases
Case no.PatternHippocampal pathology on qualitative inspectionHippocampal neuronal densitiesa
CA1 × 104/mm3
CA4 × 104/mm3
Dynorphin staining patternb
Other significant neuropathologic findingsFixation time (year)
LeftRightLeftRight
  1. AD, Alzheimer’s disease; MFS, mossy fiber sprouting; WM, white matter; HS, hippocampal sclerosis; CHS, classical HS (neuronal loss in CA1 and CA4); EFS, end folium sclerosis (neuronal loss in CA4 only); IHD, ischemic heart disease; SUDEP, sudden and unexpected death in epilepsy; PMC, postmortem control; EP, epilepsy postmortem.

  2. aStereologic counts of hippocampal neurons to confirm neuronal loss, carried out as previously described (Thom et al., 2005) in some cases.

  3. bThe presence of mossy fiber sprouting using dynorphin was graded as previously described (Thom et al., 2009a,b): Grade 1 = No MFS, Grade 2 = mild, patchy MFS, Grade 3 = Marked MFS.

EO019Unilateral HSCHSNo HS0.219 × 104/mm3
0.16 × 104/mm3
Dynorphin – MFS (Grade 3)
2.33 × 104/mm3
1.36 × 104/mm3
Dynorphin – MFS (Grade 2)
Macroscopic impression of mild left hemiatrophy13
EP038Bilateral HSCHSCHS0.86 × 104/mm3
0.09 × 104/mm3
Dynorphin – MFS (Grade 2)
0.34 × 104/mm3
0.006 × 104/mm3
Dynorphin – MFS (Grade 3)
Macroscopic impression of mild left hemiatrophy 6
EP266Unilateral CHSCHSNo HS0.13 × 104/mm3
0.001 × 104/mm3
Dynorphin – MFS (Grade 3)
1.6 × 104/mm3
ND
Dynorphin – (Grade 1)
  4
EP141UnilateralCHSCHSNo HS0.47 × 104/mm3
0.33 × 104/mm3
Dynorphin – MFS (Grade 3)
1.85 × 104/mm3
0.10 × 104/mm3
Dynorphin – (Grade 1)
Focal, mild WM vascular pathology, age-related AD pathology14
EP294Bilateral CHSCHSCHS (in residual hippocampus)0.08 × 104/mm3
0.6 × 104/mm3
Dynorphin – MFS (Grade 3)
0.02 × 104/mm3
ND
Dynorphin MFS – (Grade 3)
Previous temporal lobe surgery 1
EP286Unilateral CHSCHSEFS0.104 × 104/mm3
0.3 × 104/mm3
Dynorphin – MFS (Grade 3)
1.0 × 104/mm3
0.73 × 104/mm3
Dynorphin – MFS (Grade 2)
  2
EP210No CHSNo HSNo HSDynorphin – MFS (Grade 1)Dynorphin – MFS (Grade 1)Scar in brainstem. No cortical pathology 9
EP290No CHSNo HSNo HSDynorphin – MFS (Grade 1)Dynorphin – MFS (Grade 1)Mild cerebellar atrophy and amygdala gliosis 3
EP039No CHSNo HSNo HSDynorphin – MFS (Grade 1)Dynorphin – MFS (Grade 2)Cerebellar atrophy, Vacuolar myelopathy 8
Controls PMC1-4No HSNo HSNo HS1.45 × 104/mm3 (mean)
0.58 × 104/mm3 (mean)
1.03 × 104/mm3 (mean)
0.4 × 104/mm3 (mean)
 10.2 (mean)

Block selection and staining rationale

In all cases and controls an identical set of 13 paired blocks was taken from specified cortical regions of both the left and right hemisphere from the archived formalin-fixed brain tissue (26 samples in total; Table 3). In some cases, specific areas were not available and the mean number of blocks per sampled case was 24. Regions were selected according to known hippocampal afferent or efferent projection sites (Duvernoy & Cattin, 2005) and regions shown as the most abnormal areas in quantitative neuroimaging studies (Table 3) (Cormack et al., 2005; Lin et al., 2007; Keller & Roberts, 2008; McDonald et al., 2008; Bernhardt et al., 2009).

Table 3.   Rationale for block sampling protocol
Regions sampled in current studyRationale for block sampling based on known hippocampal projection regions and MRI studies
RegionsLobeCortical regionBrodmann areaKnown hippocampal pathway projections: Direct/Indirect (Duvernoy & Cattin, 2005)Recent neuroimaging studies in TLEQuantitative neuroimaging measure
Ipsilateral changes (I)
Contralateral changes (C)
Bilateral changes (B)
Patient group in MRI studies
  1. aIn this review cortical regions were grouped into areas and the frequency of involvement in 18 published series shown as percentage.

  2. VBM, voxel-based morphometry; GMV, gray matter volume; TLE, temporal lobe epilepsy; I, MRI changes ipsilateral to TLE; C, MRI changes contralateral to TLE; B, MRI changes bilateral; TLE/HA, TLE associated with hippocampal atrophy reported on MRI; mTLE, mesial temporal lobe epilepsy; cTLE, cryptogenic temporal lobe epilepsy; mTLE/HS, mTLE with hippocampal sclerosis; rTLE, refractory TLE (no MTS).

  3. Thirteen cortical regions were selected based on known hippocampal projection regions (Duvernoy & Cattin, 2005) and on MRI studies in TLE as reviewed in (Keller & Roberts, 2008)a and more recently published VBM and other MRI quantitative studies.

 1FrontalPrefrontal cortex/pole10Output region (Direct)Riederer et al. (2008)
Bernhardt et al. (2009)
Lin et al. (2007)
VMB/reduced GMV (I)
Cortical thickness ↓ (B)
Cortical thickness ↓ (B)
Left mTLE (I), cTLE
TLE/HA
mTLE/HS
 2Orbitofrontal cortex11 Riederer et al. (2008)
Bernhardt et al. (2009)
Lin et al. (2007)
McDonald et al. (2008)
VMB/reduced GMV (B)
Cortical thickness ↓(B)
Cortical thickness ↓ (B)
Cortical thickness ↓ (B)
cTLE
TLE/HA
mTLE/HS
mTLE/HS
 3Anterior cingulate24Output region (Indirect)Riederer et al. (2008)
Keller & Roberts (2008)a
VMB/reduced GMV (C)
VBM/reduced GMV (C-35%, I-41%)
mTLE (left), cTLE
TLE
 4Primary motor4 Riederer et al. (2008)
Labate et al. (2009)
Lin et al. (2007)
VBM/reduced GMV (B)
VBM/reduced GMV (B)
Cortical thickness ↓ (B)
cTLE
rTLE/mTLE
mTLE/HS
 5ParietalPrimary sensory3, 1, 2 Riederer et al. (2008)
Labate et al. (2009)
Lin et al. (2007)
McDonald et al. (2008)
VBM/reduced GMV (B)
VBM/reduced GMV (B)
Cortical thinning (B)
Cortical thickness ↓ (B)
cTLE
rTLE/mTLE
mTLE/HS
mTLE/HS
 6Posterior cingulate23Output region (Indirect)Bernhardt et al. (2009)Cortical thickness ↓ (C)TLE/HA
 7Retrosplenial cortex29/30Output region (Indirect)   
 8Retrosplenial cortex26Output region (Indirect)   
 9Parietal association cortex27Input region (Indirect)   
10Parietal cortex/angular gyrus39Input region (Indirect)Bernhardt et al. (2009)
Keller & Roberts (2008)a
Cortical thickness ↓ (B)
VBM/reduction GMV (C-52.9%, I-47%)
TLE/HA
TLE
11TemporalPosterior temporal Assn. cortex/temporooccipital37Output region (Direct)Bernhardt et al. (2009)Cortical thickness ↓ (C)TLE/HA
12Temporal pole38Output region (Direct)Bernhardt et al. (2009)
Keller & Roberts (2008)a
Cortical thickness ↓ (I)
VBM/reduced GMV (C-5.8%, I-23.5%)
TLE/HA
TLE
13OccipitalVisual (calcarine) cortex17 Riederer et al. (2008)
Lin et al. (2007)
VBM/reduced GMV
Cortical thickness ↓ (B)
cTLE (C)
mTLE/HS

We selected immunomarkers to detect acquired neocortical pathology including gliosis (GFAP) and microgliosis (CD68). In addition, staining for NPY fiber networks was carried out because an increase in cortical NPY networks has been shown in epilepsy surgical tissues and experimental models (Thom et al., 2003; Kharlamov et al., 2007) in addition to the hippocampus (Mathern et al., 1995; Thom et al., 2009b) and is, therefore, potentially a more specific marker of epilepsy-induced cortical changes. We were limited to immunomarkers in which we demonstrated robust staining with long fixation times (Liu et al., 2010); this precluded the use of NeuN for estimation of cortical neuronal numbers. Furthermore, due to variability in the plane of the section in the samples it was not possible to get a reliable measure of cortical thickness to recapitulate neuroimaging measurements.

Tissue preparation

For each case, 7-μm–thick formalin-fixed paraffin-embedded sections were dewaxed, rehydrated through graded alcohols, and immersed in distilled water. Endogenous peroxidase was quenched with 3% hydrogen peroxide in water. After relevant antigen treatments, sections were stained for 1 h at room temperature with the following antibodies: polyclonal GFAP (1:1,500; Dako, Glostrup, Denmark) with proteinase K for enzyme digestion; monoclonal CD68 (1:100; Dako) with heat pretreatment; and polyclonal NPY (1:5,000; Sigma, St. Louis, MO, U.S.A.) with heat pretreatment. Labeling was detected with horseradish peroxidase kit (Dako Envision) and DAB+ for visualization. Nuclei were counterstained with light cresyl violet. Between each step, sections were washed with phosphate-buffered saline (PBS) buffer with 0.5% Tween 20. All the sections within a case for a single antibody were stained in the same run to minimize staining variability.

Quantitative analysis of stained sections

A commercial image analysis system (Histometrix; Kinetic Imaging, Liverpool, United Kingdom) with a Zeiss Axioskop microscope (Carl Zeiss, Oberkochen, Germany) was used for field fraction analysis. For analysis of the GFAP-stained sections two regions of interest (ROIs), the cortex (gyral crown) and underlying subcortical white matter, were outlined on the image analyzer program at ×2.5 magnification. The cortical ROI included all layers (pial margin to layer VI) and the full breadth of the gyral crown. We were careful not to include any cortical tissue in the underlying white matter ROI. Both ROIs were also analyzed on the CD68-stained section but only the cortical ROI in the NPY cases. The quantitative analysis (field fraction analysis) was carried out by two observers (FB, IL) with good interobserver repeatability. Field fraction analysis relies on immunostaining intensity (Eriksson et al., 2007) and measures the approximate percentage of specific immunostaining per field (Fig. 1I,J). RGB (red–green–blue) detection thresholding was set for the first field to detect the majority of either GFAP, CD68, or NPY immunostaining but minimizing nonspecific detection. Light intensity and the RGB detection thresholds were then kept constant for all subsequent fields and for all sections in a case. Within the ROI, 20% of the high power fields (using ×40 objective lens) were systematically randomly sampled. An initial pilot study showed 20% sampling of the ROI gave reproducible measurements and an average coefficient error of <0.1. The percentage area immunostained within each ROI was calculated. Data analysis was carried out using SPSS version 16.0 for Windows (SPSS, Chicago, IL, U.S.A.).

Results

Qualitative inspection confirmed that the cytoarchitecture from each sampled region on cresyl violet stain was compatible with the respective Brodmann area. There was no specific pathology noted in any of the sections: no cortical dysplasia and no hemosiderin deposition. Qualitative inspection of GFAP staining in epilepsy cases showed Chaslin’s gliosis of variably intensity in all cases, with radial fibers extending into layers I and II and more variable pancortical gliosis (Fig. 1G). In two cases (EP019 and EP038), patchy laminar gliosis in layer II–III was noted in the temporal pole on the left (Fig. 1H), as well as in orbitofrontal and frontal pole regions on the left in EP038. In patient EP294 with a previous right temporal lobectomy, only patchy Chaslin’s and cortical gliosis were noted in the blocks sampled, with no evidence of postoperative scarring. In most regions in epilepsy cases, gliosis appeared more severe in the white matter than in cortex. The qualitative impression was also of more CD68-positive microglial cells in white matter than in cortex (Fig. 1K). Only occasional microglial clusters were present. NPY showed labeling of numerous fibers and processes primarily in the cortex in epilepsy cases (Fig. 1L). NPY staining in the white matter was restricted to isolated single cells. For this reason field fraction analysis was carried out only on the cortical ROI for NPY.

Quantitative analysis of CHS and non-CHS cases

Quantitative analysis showed higher mean field fraction measures for each marker (averaged over all 26 regions from left and right hemispheres) in CHS cases compared to non–epilepsy controls in both cortex and white matter (Table 4); this was significant for NPY in the cortex and CD68 in the white matter (at p < 0.05). The non-CHS epilepsy cases were not significantly different from controls or CHS cases for any of the markers, although the mean NPY staining fraction in this group appeared the highest. In epilepsy cases the white matter showed significantly more GFAP reactivity compared to the cortex (p < 0.05) and similarly for CD68 (p < 0.01), confirming the qualitative impression.

Table 4.   Mean field fraction staining in all cortical regions and white matter regions from both hemispheres for individual cases in the three groups: classical hippocampal sclerosis, no hippocampal sclerosis, and controls
MarkerGFAPCD68NPY
RegionCortexWhite matterCortexWhite matterCortex
  1. All values are shown as percentage of staining (field fraction).

  2. PMC, postmortem control; EP, epilepsy postmortem.

  3. aResults significantly different from controls (p < 0.05).

Classical hippocampal sclerosis
 EP 0192.863.890.070.290.61
 EP0385.8812.620.290.461.22
 EP2665.4323.070.190.440.54
 EP14120.930.900.080.210.77
 EP29426.8437.361.092.481.31
 EP28616.9627.900.480.922.03
 Average13.1522.620.370.80a1.08a
Epilepsy cases
 No hippocampal sclerosis
  EP2102.395.570.040.10.47
  EP0393.9110.020.220.290.65
  EP2907.7118.010.510.533.34
  Average4.6711.20.250.31.48
Control cases
 PMC14.7312.370.270.350.54
 PMC28.8111.730.150.200.66
 PMC34.7813.280.0720.060.52
 PMC40.721.240.110.120.52
 Average4.769.660.150.180.56

Lateralization of neocortical pathology

We compared GFAP, CD68, and NPY staining fractions from the cerebral cortical regions ipsilateral to unilateral CHS cases to the corresponding paired contralateral regions. Overall no significant or consistent patterns emerged to indicate greater labeling on the side of the sclerosis (Table 5). In patients with impression of left hemiatrophy on MRI (EP019 and EP038), high field fractions for GFAP were present in the left frontal pole, temporal pole, and orbitofrontal regions, but there were no significant overall left–right hemispheric differences. Similarly, no significant interhemispheric differences were noted in non-CHS cases or non–epilepsy controls (Table 5).

Table 5.   Interhemispheric differences. Comparison of field fraction immunostaining between paired hemispheric blocks on left and right sides in unilateral classical hippocampal sclerosis (CHS), epilepsy patients without CHS, and controls
Ratio left:right hemisphereUnilateral left CHS casesNo CHS epilepsy casesControls
EP019EP266EP141EP286EP210EP290EP039PMC1PMC2PMC2PMC4
  1. The data are shown as mean values for each hemisphere expressed as ratio between left:right. There was no significant difference in the left:right hemisphere ratios in each group.

  2. PMC, postmortem control; EP, epilepsy postmortem.

GFAP cortex1.51.10.851.030.711.181.071.040.911.630.85
GFAP white matter1.90.871.11.191.220.750.861.121.380.980.87
CD68 cortex1.20.421.00.361.00.931.01.351.01.111.35
CD68 white matter0.960.761.361.231.00.931.01.210.971.01.07
NPY cortex0.81.161.050.391.00.550.851.821.10.680.9
Significance between hemispheres (Wilcoxon signed-rank test)p = 0.263p = 0.08p = 0.296

Regional distribution of cortical pathology

Over the 13 cerebral cortical regions, field fractions of GFAP, CD68 in both the cortex and white matter, and NPY in the cortex showed significant differences in values in CHS cases compared to non–epilepsy controls (p < 0.01) (Table 6), often with two- to threefold higher values for GFAP and CD68 in epilepsy cases. NPY was higher in all regions in CHS, except the retrosplenial cortex. The cortical regions consistently showing the highest labeling values with all markers in CHS were the temporal pole, prefrontal cortex (frontal pole), and orbitofrontal and calcarine cortex (Table 6).

Table 6.   Fraction of staining in 13 sampled cortical regions for all classical hippocampal sclerosis (CHS) epilepsy cases (both left and right hemispheres) compared to non–epilepsy controls
Cortical region (Brodmann area)GFAP in cortex
Mean epilepsy (SD)
Mean control (SD)
GFAP in white matter
Mean epilepsy (SD)
Mean control (SD)
CD68 in cortex
Mean epilepsy (SD)
Mean control (SD)
CD68 in white matter
Mean epilepsy (SD)
Mean control (SD)
NPY in cortex
Mean epilepsy (SD)
Mean control (SD)
  1. The figures shown are the immunostaining fraction (percentages) and the significance between CHS cases and controls is shown (Wilcoxon signed-rank test). The highest two values for each marker in CHS cases are underlined illustrating that the temporal pole more often showed higher labeling.

Prefrontal cortex/pole (area 10)15.1 (0.60)
 5.0 (0.77)
20.0 (11.33)
10.13 (6.94)
0.53 (0.6)
0.21 (0.18)
1.09 (1.16)
0.25 (0.17)
0.96 (0.85)
0.43 (0.28)
Orbitofrontal cortex (area 11)13.6 (0.16)
 4.0% (0.84)
24.05 (15.04)
9.93 (6.9)
0.39 (0.46)
0.11 (0.1)
1.16 (1.04)
0.26 (0.19)
1.26 (0.68)
0.96 (0.08)
Anterior cingulate (area 24)14.2 (0.33)
 7.3 (4.08)
20.85 (12.29)
9.35 (5.59)
0.36 (0.39)
0.17 (0.16)
0.87 (0.78)
0.27 (0.16)
2.3 (3.54)
0.8 (0.3)
Primary sensory (area 1, 2, 3)12.8% (0.43)
 4.9% (0.75)
18.38 (12.7)
6.11 (3.47)
0.30 (0.27)
0.18 (0.12)
0.91 (0.87)
0.18 (0.13)
0.95 (0.55)
0.43 (0.15)
Posterior cingulate (area 24)12.0 (0.86)
 4.5 (0.15)
23.07 (13.34)
7.91 (4.84)
0.40 (0.48)
0.16 (0.09)
0.85 (0.79)
0.19 (0.06)
1.01 (0.55)
0.57 (0.28)
Primary motor (area 4) 9.4 (0.17)
  4.8 (0.007)
21.31 (12.52)
9.22 (6.64)
0.20 (0.19)
0.12 (0.1)
0.43 (0.21)
0.17 (0.16)
0.96 (0.6)
0.52 (0.37)
Retrosplenial cortex (area 29)13.3 (0.75)
 3.8 (0.91)
20.5 (14.47)
5.66 (3.25)
0.44 (0.51)
0.19 (0.14)
0.79 (0.87)
0.19 (01.5)
0.77 (0.36)
0.57 (0.37)
Retrosplenial (area 26)10.3 (1.17)
 6.7 (0.98)
26.86 (15.58)
16.34 (11.44)
0.34 (0.51)
0.13 (0.09)
0.83 (1.0)
0.12 (0.14)
0.53 (0.16)
0.62 (0.42)
Parietal association cortex (area 7)14.3 (0.63)
  4.65 (0.67)
21.04 (14.13)
7.68 (5.1)
0.28 (0.3)
0.10 (0.08)
0.74 (0.68)
0.0013 (0.09)
1.26 (0.95)
0.54 (0.31)
Temporal association cortex (area 37)12.0 (1.56)
  4.18 (0.32)
22.54 (18.8)
8.81 (5.02)
0.27 (0.31)
0.11 (0.1)
0.65 (0.8)
0.21 (0.21)
0.76 (0.56)
0.54 (0.25)
Parietal cortex/angular gyrus (area 39)12.25 (2.75)
 4.6 (0.09)
21.93 (12.68)
7.28 (5.34)
0.31 (0.27)
0.11 (0.09)
0.63 (0.91)
0.14 (0.11)
1.07 (0.89)
0.45 (0.34)
Visual/calcarine cortex (area 17)12.1 (3.26)
  3.3 (0.007)
31.03 (10.7)
15.87 (10.6)
0.41 (0.47)
0.20 (0.15)
0.89 (0.93)
0.23 (0.14)
0.46 (0.37)
0.28 (0.11)
Temporal pole (area 38)19.0 (3.5)
 2.8 (0.54)
30.22 (7.99)
12.53 (5.35)
0.68 (0.63)
0.00 (0)
1.55 (1.61)
0.05 (0.11)
1.64 (0.17)
0.41 (0.39)
Significance between cases and controlsp = 0.001p = 0.001p = 0.002p = 0.001p = 0.002

We also compared relative regional distribution of immunolabeling in the 13 cerebral cortical regions by ranking each region from 13 (highest field fraction value) to 1 (lowest field fraction value) within individual cases. This ranking was then compared within and between groups. In epilepsy cases (both CHS and non-CHS epilepsy cases), GFAP in the cortex was more often greatest in the temporal pole and orbitofrontal cortex and in the white matter in the retrosplenial area (area 29) compared to controls (Fig. 2). For CD68, the temporal and frontal pole cortex showed greater labeling in epilepsy cases, whereas distribution in the white matter was similar to controls, except for the temporal pole in CHS cases (Fig. 3). The distribution pattern of labeling for NPY was similar in CHS and non-CHS epilepsy cases. In comparison to controls, accentuated NPY was noted in the frontal pole and primary sensory cortex in epilepsy cases, but with a notable diminution of NPY in the retrosplenial cortex (areas 26, 29) (Fig. 4).

Figure 2.


A graph of the regional variation in GFAP staining, in the 13 sampled regions, in the cortex and white matter. In individual cases regions were ranked from 13 (highest GFAP labeling) to 1 and then overall ranks shown within each pathology group. The CHS cases are shown with highest ranks on the left and lowest on the right. Brodmann area shown in parentheses. CHS, classical hippocampal sclerosis.

Figure 3.


A graph of the regional variation in CD68 staining, in the 13 sampled regions, in the cortex and white matter. In individual cases regions were ranked from 13 (highest CD68 labeling) to 1 and then overall ranks shown within each pathology group. The CHS cases are shown with highest rank on the left. Brodmann area shown in parentheses. CHS, classical hippocampal sclerosis.

Figure 4.


A graph of the regional variation in NPY staining in the 13 sampled cortical regions. In individual cases regions were ranked from 13 (highest NPY labeling) to 1 and then overall ranks shown within each pathology group. The CHS cases are shown with highest rank on the left. Brodmann area shown in parentheses. CHS, classical hippocampal sclerosis.

Clinical pathology correlations

Because of the small study group it is not possible to draw firm conclusions regarding clinicopathologic correlations. All the patients had secondary generalized seizures. There was no correlation between duration of epilepsy and the mean field fraction values with GFAP, CD68, and NPY. There were no differences in pathology measures in patients who continued to have refractory seizures and those whose seizures had come under control in the years prior to death.

Discussion

There is accumulating quantitative neuroimaging data to support the notion that widespread neocortical abnormalities accompany TLE and that it is not a disease limited to the hippocampus. Explanations include reciprocal atrophy via hippocampal–cortical networks, an independent effect of seizures, or that these abnormalities predate seizures. Neuropathologic studies of surgical tissue in TLE have long recognized that sclerosis (neuronal loss and gliosis) is not restricted to the hippocampus but can involve the amygdala, parahippocampal gyrus, and temporal lobe (Meyer et al., 1954; Cavanagh & Meyer, 1956; Bruton, 1988). The distribution of extratemporal neocortical pathology remains relatively unexplored. An earlier postmortem study in patients with TLE and HS noted patchy cortical neuronal loss and gliosis involving frontal and occipital lobes in 22% of cases but without further specification (Margerison & Corsellis, 1966). We studied postmortem material from patients with refractory epilepsy with CHS as the main pathology and confirmed neocortical pathology, as determined with GFAP, CD68, and NPY markers, with a distribution favoring the prefrontal cortex (pole), orbitofrontal cortex, and temporal pole.

MRI voxel-based morphometry (VBM) has demonstrated that 26 brain regions show gray matter volume reduction in TLE as recently reviewed (Keller & Roberts, 2008). Commonly involved extratemporal regions include parietal, insular, cingulate, orbitofrontal, and dorsal frontal cortex (Keller & Roberts, 2008; McDonald et al., 2008; Riederer et al., 2008; Keller et al., 2009; Labate et al., 2009). Reduction in white matter has also been reported, mainly in the ipsilateral temporal lobe, but also involving frontal, parietal, cingulate, and occipital cortex (Keller & Roberts, 2008; Yasuda et al., 2010). Variability in regional involvement between studies may reflect differences in patient cohorts or methodologies (Table 3). MRI methods measuring cortical thickness have also shown a reduction in several regions including bilateral frontal poles, and orbitofrontal, lateral temporal, and occipital lobes (Lin et al., 2007) (Table 3). In addition, progressive thinning has been shown in longitudinal studies of adult TLE patients involving ipsilateral temporopolar, frontal, central, and contralateral orbitofrontal and insular cortex (Bernhardt et al., 2009). It has been proposed that MRI neocortical abnormalities primarily involve hippocampal projection pathways, with several studies favoring this reflecting limbic–neocortical network changes (Lin et al., 2007; McDonald et al., 2008; Riederer et al., 2008).

Our cortical sampling protocol aimed to include regions highlighted by these MRI studies as well as known hippocampal projection regions (Table 3). We included hippocampal output regions to the cingulate and retrosplenial cortex (indirect pathway), inferior temporal association cortex, temporal pole, and prefrontal cortex (direct pathway), as well as hippocampal input regions from the parietal cortex. We demonstrated greater gliosis and microgliosis in temporal pole and prefrontal output regions in addition to a white matter gliosis. We noted a relative reduction of NPY in the retrosplenial cortex compared to controls. Therefore, pathology measures in hippocampal–neocortical projection (input or output) regions are not all equally altered. In addition, we noted significant changes in orbitofrontal cortex, which is not a known hippocampal projection site (Duvernoy & Cattin, 2005). Furthermore, a similar distribution of cortical changes was noted in epilepsy patients, both with and without CHS, despite less overall staining in non-CHS cases. This latter finding is consistent with observations in MRI studies that have included TLE patients without CHS (Riederer et al., 2008).

Concerning laterality of damage, greater ipsilateral cortical thinning has been shown in some (McDonald et al., 2008) but not all MRI studies in HS (Lin et al., 2007). In our postmortem study, although two patients macroscopically showed mild hemispheric atrophy and microscopic asymmetric temporal lobe gliosis, overall no significant interhemispheric differences were identified. However, given that 11 of our 13 selected regions were extratemporal, this finding may reflect sampling. MRI studies have demonstrated greater volume changes in the ipsilateral temporal lobe (including the temporal pole) in contrast to extratemporal structures, which more often show bilateral changes (Keller & Roberts, 2008). Furthermore, in our study, all unilateral CHS cases were left-sided. It has been shown that patients with left HS have more extensive and bilateral damage with VBM, compared with right HS (Bonilha et al., 2007; Riederer et al., 2008). This could explain why we failed to detect any left–right cortical asymmetry in unilateral CHS cases.

Overall, our postmortem findings could support the notion that neocortical pathology in refractory epilepsy occurs independently of hippocampal pathology and known projection pathways. Other possible explanations for this regional pattern of bilateral acquired cortical damage should be considered. All the patients in our series had frequent generalized convulsions and were, therefore, at risk for cumulative head injury. Noncontusional traumatic brain injury could be one possible mechanism, consistent with the preferential gliosis of the gyral crowns of the orbitofrontal cortex in addition to frontal and temporal poles. The orbitofrontal cortex is not a hippocampal projection site. It is, however, a common site for chronic trauma-related atrophy as observed in traumatic encephalopathy following repetitive head injuries incurred in professional sportsmen (Corsellis et al., 1973; McKee et al., 2009), although this is more typically accompanied by neurofibrillary tangle pathology.

The markers we chose to measure cortical pathology—GFAP, CD68, and NPY— were selected as these are reliable indicators for acquired cortical changes in epilepsy surgical temporal lobectomy specimens (Thom et al., 2003; Love et al., 2008). It was not possible to obtain accurate measures of cortical thickness from the tissue sections in the present study as this is dependent on the plane of section, which was highly variable and not always perpendicular to the cortical surface. Because of the known variability of immunohistochemistry staining for NeuN in autopsy tissues (Gill et al., 2005), particularly with long fixations (Liu et al., 2010), we also did not carry out measures of neuronal number. One possibility for the differences observed between MRI and our pathology studies is that we are measuring different features, for example, the extent of gliosis in tissue sections may not parallel tissue volume loss by MRI. In addition, histologic studies are limited to regional tissue sampling, whereas in MRI the whole brain is considered. Conversely, it is recognized that VBM may not detect subtle changes in localized anatomic regions (Keller & Roberts, 2008). In surgical tissues where more exact coregistration between imaging and tissue sections is possible (Eriksson et al., 2005) there is still uncertainty regarding the pathologic basis for MRI volume or signal changes (Meiners et al., 1999; Mitchell et al., 1999; Coste et al., 2002; Eriksson et al., 2007). Furthermore, in a surgical series with histologically confirmed temporal lobe sclerosis in addition to hippocampal sclerosis (Thom et al., 2009a), VBM failed to detect abnormalities in individual patients (Eriksson et al., 2009).

It is possible that the presence of more widespread neocortical pathology is the basis for failure to control seizures following surgical treatment for CHS (Bertram, 2009). Widespread atrophy of both gray and white matter has been recently associated with poorer outcomes following surgery for HS (Yasuda et al., 2010). We did not identify any specific extrahippocampal lesional pathologies, as occult dysplasias, but activation of normal resident cortical and white matter cell types. Inflammatory cellular responses, including upregulation of CD68-positive microglia number, in response to seizures, may enhance neuronal excitability (Vezzani et al., 2008a,b). In addition, the frequent observation of astrocytosis in association with epilepsy is no longer regarded as a mere marker of neuronal damage but as likely contributing to an increased propensity for seizure mediation (Eid et al., 2008; Sofroniew & Vinters, 2010). Sprouting of NPY-positive fibers from inhibitory interneurons is a frequent observation in hippocampal sclerosis (de Lanerolle et al., 2003) and may represent a cellular mechanism to counteract seizures. An excess of NPY fibers was confirmed in almost all cortical regions in epilepsy patients, with the exception of the retrosplenial cortex.

There are limitations with postmortem tissue studies. This includes the effects of fixation on tissue shrinkage and intensity of immunostaining. We have shown that GFAP and NPY maintain robust staining even with long fixation times, but the CD68 marker may show some reduction of staining (Liu et al., 2010). We attempted to overcome variability between cases by comparing left and right hemispheres and ranking the distribution of staining within a case in addition to comparison of actual field fraction values. This PM series is not as homogenous a patient group as surgical CHS series in that all did not have a clinical diagnosis of mTLE and there were not adequate MRI data in all patients for quantitative analysis, or detailed electrophysiologic data. We selected cases for which CHS was the main pathology, but we cannot exclude the possibility that confounding pathologies, including age-related vascular changes, contribute to cortical and white matter gliosis in older patients. The majority of the patients were residents at an epilepsy center dedicated to patients with more difficult to manage epilepsy and, as such, are a selected group and not representative of all people with epilepsy. Our study group is also small compared to imaging study samples because of the time-consuming nature of the histologic, quantitative methods.

Despite these limitations this initial study highlights the value of postmortem material in the assessment of the distribution and the nature of extrahippocampal pathology in TLE in order to further understand its pathogenesis, relationship to hippocampal pathology, and potential contribution to the continuation of seizures following surgery.

Acknowledgments

The work was supported by the MRC (Grant G79059). We would like to acknowledge the help of Professor Matthias Koepp at the National Society of Epilepsy. This work was undertaken at UCLH/UCL who received a proportion of funding from the Department of Health’s NIHR Biomedical Research Centres funding scheme.

Disclosure

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

Ancillary