A Retrospective Analysis of Hippocampal Pathology in Human Temporal Lobe Epilepsy: Evidence for Distinctive Patient Subcategories


Address correspondence and reprint requests to Dr. N.C. de Lanerolle at Department of Neurosurgery, Yale University School of Medicine, 333 Cedar Street, New Haven, CT 06520-8082, U.S.A. E-mail: nihal.Delanerolle@yale.edu


Summary:  Purpose: This study is a retrospective analysis of the pathology of the hippocampus from patients with medically intractable temporal lobe epilepsy. We attempted to relate neuronal density, immunohistochemistry, electrophysiologic data, and surgical outcome.

Methods: Immunostaining patterns for neuropeptide Y, somatostatin, substance P, and dynorphin defined the immunohistochemical characteristics of the hippocampi. Neuronal densities were determined by microscopic cell counts. Sharp electrode recordings from dentate granule cells determined measures of inhibition and excitation.

Results: Patient hippocampi without evidence of sclerosis generally resembled autopsy controls on the basis of neuronal densities of hippocampal subfields and patterns of immunostaining. The nonsclerotic hippocampi were divisible into two subgroups on the basis of neuronal density correlations between hippocampal subfields, the excitability of dentate granule cells, etiology, and surgical outcome. Hippocampi with sclerosis were divisible into those with significant neuronal loss confined to area CA1 and those with neuronal loss throughout the hippocampus and dentate gyrus. In the former, the dentate gyrus resembled in morphology the nonsclerotic hippocampi but with slightly increased excitability of the dentate granule cells. The hippocampi with more extensive neuronal loss had changes in immunostaining patterns associated with the dentate gyrus, correlated with significant hyperexcitability of dentate granule cells. The surgical outcome, with the exception of one group, was good in ∼70–90%.

Conclusions: Hippocampi from patients with intractable temporal lobe epilepsy can be assigned to several groups on the basis of pathophysiology. Different pathologies may represent differing causative mechanisms of intractable temporal lobe epilepsy and be predictive of surgical outcome.

A proper neuropathologic definition of temporal lobe epilepsy (TLE) continues to be an important endeavor in understanding the causation and maintenance of seizures in this disorder. The hippocampus, located in the mesial temporal lobe, is commonly regarded as the locus of seizure origin in TLE because depth electrode studies show seizures to originate from this area, and surgical removal of the hippocampus produces good seizure control. Beginning in the early 1800s, several investigators contributed important insights into the understanding of hippocampal pathology in TLE [reviewed in (1)].

The earliest pathological descriptions were based on macroscopic examination of autopsy brain specimens and described the hippocampus as hardened (indurated) and small (atrophied) (2). This condition came to be known as Ammon's horn sclerosis. The first microscopic examination and description of the epileptic hippocampus was published by Sommer (3). He observed that pyramidal neurons were lost in the hippocampus not diffusely but preferentially in a portion of the hippocampus that has come to be called Sommer's sector and corresponds to area CA1 (4). Sommer briefly noted that some neurons were lost in the hilus of the dentate gyrus as well.

In 1899, Bratz (5) provided a more detailed description of the TLE hippocampus. While confirming the loss of pyramidal neurons in area CA1, he observed that neurons were preserved in the subiculum, presubiculum, stratum oriens, and the granule cell layer of the dentate gyrus. He also described a smaller loss of pyramidal neurons in the hilus (“end folium”) of the dentate gyrus, and in a region corresponding to CA2 that “resists destruction the longest.” Bratz further described an abundance of blood vessels in the atrophic sectors, their patency, and a lack of any pathologic alterations in their walls [see discussion in (1)]. This latter feature has received little attention in recent studies. The term hippocampal sclerosis was therefore originally used to describe a shrunken and hardened hippocampus, which histologically displayed preferential neuronal loss with secondary astroglial proliferation (6). This pathology has been reported in 20–80% of hippocampi in autopsy series of brains from those with epilepsy [Table 2 in (6)], and in 50–70% of hippocampi surgically resected from patients with TLE [Table 10 in (6)].

Table 2.  Patient characteristics
TypeGenderAge at surgery
Age at first seizure
Seizure duration
  1. In each box for columns 2 to 4, the first number represents the mean ± standard deviation. The range is given in parenthesis. In the group comparisons, the number given is the p value calculated by the Mann–Whitney U, two-tailed test.

  2. Seizure duration, years from first intractable seizure to date of surgery; F:M, female to male ratio; NA, not applicable; NS, not significant; MTLE, mesial temporal lobe epilepsy; PTLE, paradoxical TLE: MTLE/DYN, MTLE dynorphin negative; MaTLE, mass-associated TLE.

Autopsy (n = 26)13F:13M30.5 ± 9.0 (14–50)NANA
PTLE (n = 18)10F:8M32.5 ± 9.5 (15–50)9.4 ± 6.8 (0.2–19)22.7 ± 12.8 (5.8–47)
MaTLE (n = 42)16F:26M27.8 ± 9.7 (11–53)14.7 ± 11.4 (0.1–48)13.7 ± 8.2 (0.12–33)
MTLE (n = 72)34F:38M29.9 ± 9.0 (7–51)4.4 ± 7.1 (0.3–36)25.8 ± 10.8 (2.0–48)
MTLE/DYN (n = 10)3F:7M28.1 ± 7.8 (18–47)6.5 ± 9.0 (0.1–28)21.5 ± 6.7 (10–29.3)
CA1 group (n = 9)3F:6M37.8 ± 13 (10–53)7.9 ± 8.8 (0.1–28)29.9 ± 14.3 (6–48)
MaTLE/MTLE  <0.0001<0.0001
MaTLE/MTLEDYN  0.0120.005
MaTLE/CA1  NS0.003
MTLE/CA1  0.05NS

Several studies have attempted to define hippocampal sclerosis further by neuron counts (7–11). The correlation of these quantitative studies with the qualitative patterns observed microscopically has been difficult, mainly because of the pronounced interindividual variability of neuron loss and the absence of an agreed-on quantitative measure of “sclerosis.” Bruton (12) identified three degrees of neuronal loss recognizable by qualitative microscopy. “Classic” Ammon's horn sclerosis is characterized by a severe to total loss of neurons in areas CA1 and the hilus (CA4), with lesser loss in the granule cell layer, and least in CA2. “Total” Ammon's horn sclerosis is characterized by a “global” destruction of hippocampal neurons (i.e., a pronounced loss of neurons in all fields of the Ammon's horn and dentate gyrus). “End folium sclerosis” involves neuron loss only in the end folium (hilus). Babb (13) pointed out the difficulty of correlating cell counts with these three types. The different quantitative criteria (percentage of cell loss) chosen to define sclerosis resulted in different classifications of the same group of patients. This was especially true for hippocampi other than those with end folium sclerosis. Kim (14), opting for a statistical approach, chose as a quantitative definition of sclerosis a neuronal density of <60% of the control group mean. This percentage represents the control mean value minus two standard deviations averaged throughout all CA fields.

More recently, immunohistochemical studies of hippocampi, removed in the surgical treatment of drug-refractory TLE, examined changes in chemically defined neuron systems within the hippocampus. These studies were reviewed recently (15). In particular, immunohistochemical studies on the localization of the peptides neuropeptide Y (NPY), somatostatin (SOM), substance P (SP), and dynorphin (DYN) in the TLE hippocampi show distinct patterns among patients (16,17). The loss of populations of peptidergic neurons in the subgranular zone of the dentate hilus and sprouting of several peptidergic axons into the dentate molecular layer (a cluster of features taken as evidence of “reorganization”) seem to be associated with hippocampi that also show sclerosis. Do such reorganizational features define the epileptogenic hippocampus better than the presence of hippocampal sclerosis, or do they together help to provide better criteria for classification of hippocampal pathology? The electrophysiologic properties of the dentate granule cells in surgically removed hippocampal specimens provide another functional index for the characterization of hippocampi in TLE patients (18). The goal of the present study was to examine hippocampi from a large population of patients with intractable TLE to determine the relations between immunohistochemical characteristics, neuronal densities, granule cell excitability, and surgical outcome. Such an analysis will facilitate the discussion of whether intractable TLE is a single entity or one with several subtypes, and help provide further insight into the pathophysiologic mechanisms underlying TLE.


Subjects and tissue collection

The human hippocampal samples analyzed in this study, other than the autopsy controls, were obtained through the Yale Epilepsy Surgery Program. The hippocampi were surgically removed from patients with medically intractable TLE. Tissue used in these studies was obtained after informed consent and with approval of the institutional Human Investigations Committee (HIC). The criteria for the selection of patients for surgery and the surgical method for en bloc resection of this tissue are described elsewhere (19,20). Hippocampi from 151 patients collected from 1990 to 2000 were used for analysis in this study. All these patients were followed up for ≥1 year after surgery to determine their postoperative seizure status. The patient hippocampi were divided into several groups. Those with atrophy and signal change seen on magnetic resonance imaging (MRI) were usually associated with the syndrome of mesial TLE (MTLE). In this group, the seizure activity was thought to originate from one hippocampus either on the basis of concordant noninvasive localization or depth electrode and/or subdural recordings. Hippocampi from this patient group were further subdivided, based on immunohistochemical criteria (described in Results), into typical MTLE (n = 72), MTLE dynorphin negative (MTLE/DYN, n = 10), and the CA1-only group, which had qualitative evidence of neuron loss in area CA1 without immunohistochemical reorganization of the dentate gyrus (n = 8). Patients with mass-associated TLE (MaTLE, n = 43; i.e., having extrahippocampal mass lesions [low-grade gliomas and cavernomas] located in the temporal lobe) had the mass lesions removed along with portions of the hippocampus to ensure lesion-free margins. Hippocampi selected for resection based on intracranial recording of ictal onset but with no visual evidence of sclerosis or a mass lesion were classified as paradoxical TLE (PTLE, n = 18) (21). Hippocampi used as normal controls were obtained from autopsy of subjects with no known history of neurologic disease (n = 26).

Neuronal counts

Neuronal counts were carried out as described in Kim et al. (10). The hippocampus was divided into areas CA1 through CA4 and the dentate granular layer (Fig. 1). Neuronal nuclei were counted in each unit area of a 200 × 400 μm2 rectangle for the CA fields, and a 50 × 100 μm2 rectangle for the granule cell layer by using an ocular grid calibrated with a stage micrometer. The neuronal numbers obtained were adjusted with Abercrombie's formula (22), and the neuronal density expressed as mean neuronal number/mm3. The numeric parameters used in applying Abercrombie's formula were section thickness of 20 μm, with 8 μm nuclear diameter for pyramidal neurons and 6 μm for granular neurons. The stereologic “disector”(23) was not used in this study because only relative estimates of neuronal density, and not estimates of total number of neurons, were obtainable because the entire hippocampus was not available for serial random sampling, as required by the disector method.

Figure 1.

Photomicrograph of a section stained with hematoxylin & eosin of the hippocampus from a neurologically normal autopsy subject. The lines demarcate our identification of the limits of hippocampal subregions for making cell counts. The line between the CA1 region and the subiculum is positioned at the beginning of the prosubiculum where CA1 and prosubiculum overlap. This was done to avoid accidentally counting prosubicular neurons with CA1 neurons.


Immunohistochemistry was carried out on adjacent blocks of tissue. The tissue blocks, obtained at surgery, were fixed and prepared as described previously (21). In brief, tissue blocks were immersed in a picric acid and paraformaldehyde solution in phosphate buffer (PB), pH 7.4, for 1–2 h, and then in 5% acrolein in PB for 3–4 h. Fifty-micrometer sections were immunostained with the indirect peroxidase–antiperoxidase (PAP) method (24) or the avidin–biotin complex (ABC) method (25) by using antibodies to NPY, SOM, SP, and DYN (21). Tissue controls by antibody omission and antibody specificity controls by immunoadsorption with synthetic peptides were carried out, and resulted in no immunostaining. The immunohistochemical categorization of patients for the analyses reported here were done by one of us (N. de L.). A second investigator (T.E.), who was trained in the criteria used, independently evaluated the immunostained sections and assigned them to the subcategories described. The interobserver reliability score was 98%.


Electrophysiologic recordings were made from dentate granule cells in 400 μm-thick hippocampal slice preparations. Recordings were made with an Axoclamp II amplifier (Axon Instruments, Union City, Calif.) by using sharp microelectrodes. The electrodes were filled with 4 M potassium acetate and had resistances of 40–80 MΩ. Only cells with membrane potentials hyperpolarized below –55 mV, input resistances >20 MΩ, and with action potentials that overshot 0 mV were included in this study.

Orthodromic synaptic stimuli were delivered to the outer molecular layer by using a monopolar tungsten stimulating electrode to activate the incoming perforant path fibers. Synaptic intensity ranged from 0.03 to 1.5 mA, with a pulse duration of 10 μs. The minimal intensity to evoke a single action potential, doublets, and bursts (responses greater than three action potentials) was measured. In addition, the extent of postsynaptic inhibition was assessed by examining the afterpotentials after synaptic stimulation superthreshold to action potential generation when the membrane potential was held at potentials depolarized more than –70 mV with injected current.

The data were collected by using pClamp or Axodata software and analyzed off-line by using Axograph software (Axon Instruments). To quantify the changes in synaptic excitability, four parameters were examined: the ability of the cells to fire evoked epileptiform doublets or bursts, the presence or absence of evoked inhibitory potentials, the presence of polysynaptic excitatory potentials, and the presence of spontaneous activity. These physiologic data were analyzed initially without knowledge of the pathologic category or clinical history of the patient. To compare the electrophysiologic variable between patient groups, the data were quantified by using a previously described scoring system (26).

The sum of the number of spikes evoked by 1, 1.5, and 2 times the action potential threshold was used to assess the ability to fire multiple spikes (bursts). A simpler scale was used to measure the degree of inhibitory postsynaptic potential (IPSP) loss: 0, normal inhibition after synaptic stimulation; 1, some inhibitory potentials, unusually small or of an unusual morphology; and 2, no evocable inhibition at any stimulus intensity. The degree of polysynaptic activity was assessed by counting the number of presumed polysynaptic events occurring on the falling phase of 32 consecutive excitatory postsynaptic potentials (EPSPs). Finally, spontaneous activity was assessed as the frequency of spontaneous events occurring during a 10-s period and was expressed in Hertz.

Statistical analysis

In comparing the neuronal density and measures of electrophysiologic activity between patient groups, initially an analysis of variance (nonparametric Kruskal–Wallis test, two-tailed) was performed. The Mann–Whitney U, two-tailed test (27), was used to estimate paired group differences if the Kruskal–Wallis test was significant.


Immunohistochemical characteristics of patient groups

The immunohistochemical staining patterns observed in the 151 hippocampi conformed to those previously described in epileptogenic and control hippocampi (17,21). Because the patterns were fully described and illustrated in these earlier publications, the patterns for patients included in this study are described only in summary fashion.

Autopsy, PTLE, and MaTLE

The patterns of immunostaining in the hippocampi of the autopsy, PTLE, and MaTLE groups were similar to each other. NPY, SOM, and SP immunostaining was in populations of interneurons that were multipolar and bipolar in shape. These neurons were found throughout the hilus, with a particularly dense population in the subgranular polymorphic zone. The mean number of immunostained neurons in the hilus observed in a 50-μm-thick cross section of the hippocampus were as follows. For MaTLE: NPY, 89 ± 13; SOM, 71 ± 17; SP, 81 ± 16. For PTLE: NPY, 95 ± 29; SOM, 68 ± 25; SP, 79 ± 12 (Fig. 2A). Such a pattern is referred to as representing “no neuron loss.” Peptidergic neurons also were distributed throughout the fields of Ammon's horn and the subiculum, and were located particularly in the pyramidal layer (stratum pyramidale) and oriens layer (stratum oriens).

Figure 2.

Photomicrographs to illustrate immunohistochemical criteria adopted in categorizing patients. A: Low-magnification photomicrograph of a cross section of the hippocampus from a patient of the CA1-only group immunostained for neuropeptide Y (NPY). In this group, the dentate gyrus resembles that of mass-associated temporal lobe epilepsy (MaTLE) and paradoxical TLE (PTLE) patients in having a large population of NPY-immunoreactive neurons in the hilus (small arrows) located largely in the subgranular zone. The inner molecular layer appears as a clear band with a distinct border with the middle molecular layer, indicated by arrows with a dot on the tail. The CA1 area [as in mesial temporal lobe epilepsy (MTLE) patients] is shrunken and has a heavy loss of neurons. Nonimmunoreactive neurons are not seen, as it was not counterstained with a Nissl stain, so as to show better the small immunoreactive neurons. Fine dotted lines demarcate inner and outer borders of the granule cell layer. Thick arrows, boundaries of the fields of Ammon's horn. Calibration bar, 1 mm. B, C: A portion of the dentate granule cell layer (gc), hilus (hil), and molecular layer from an MaTLE and MTLE hippocampus, respectively, immunostained for NPY. Note the greater number of subgranular neurons in the hilus in (B) compared with the same region in (C) and the relatively pale inner molecular layer (iml) in (B) having few immunostained fibers compared with (C), where this region has a greater density of stained fibers that also extend to the outer ML. D, E: Granule cell layer, hilus, and molecular layer of an MaTLE and MTLE hippocampus, respectively, which were immunostained for dynorphin. Note the dynorphin immunostain in the IML in (E) compared with the absence of stain in (D). Bar, 100 μm.

Fibers immunoreactive for the peptides NPY, SOM, and SP were found throughout the hippocampal formation and had distinctive patterns of distribution in the dentate molecular layer (ML). In the ML, NPY and SOM immunoreactive fibers were found in low density in the inner ML (IML). The IML appears as a relatively clear zone with a distinct boundary with the middle molecular layer (best seen at low magnification) (Fig. 2A). More fibers were found in the middle and outer ML, with the greater density in the outer ML, where the fibers ran parallel to the granule cell layer. SP fibers formed dense bands on the hilar and ML borders of the granule cell layer and among the granule cell somata. SP fibers were scattered more sparsely in the middle and outer ML. DYN immunoreactivity was found in the granule cell bodies and their mossy fibers/terminals in the dentate hilus and area CA3. No DYN immunoreactivity occurred in the dentate ML [a feature we describe as dynorphin negative (DYN); Fig. 2D] or other areas of Ammon's horn. These patterns are more fully illustrated in de Lanerolle et al. (16,21).


In the MTLE hippocampi, in comparison to the MaTLE, PTLE, and autopsy controls, the most prominent feature was the loss or reduction in the population of NPY, SOM, and SP immunoreactive neurons in the subgranular polymorphic zone of the dentate hilus. Counts of stained neurons in the hilus of a 50 μm-thick cross section yielded mean values of 15 ± 14 for NPY, 14 ± 14 for SOM, and 13 ± 12 for SP. By comparison with a MaTLE, PTLE, or autopsy section, the MTLE hilus appears to have very few neurons, indeed often no neurons appeared in individual sections. This appearance, which does not even require cell counts, is described as severe cell loss. In area CA1, even when there was extensive pyramidal neuron loss, many NPY, SOM, and SP immunoreactive neurons along with some immuoreactive fibers were detected. Visual estimations suggested that the density of these neurons in CA1 was similar to that in the autopsy controls, MaTLE, and PTLE patients.

The other most prominent feature was the increase in density of NPY and SOM immunoreactive axonal processes scattered throughout the dentate ML. The general appearance is that most of these fibers extend into the IML perpendicular to the granule cell layer. Because of the abundance of stained fibers in the IML, it does not appear as a clear zone, as was the appearance in MaTLE and PTLE. In the case of SP immunoreactivity, the band of immunostained fibers along the hilar border of the granule cell layer was lost in MTLE, but the immunoreactivity at the dentate ML border extended well into and often throughout the dentate ML. This appearance of immunoreactive fibers in MTLE is described as evidence of NPY, SOM, and SP sprouting (Fig. 2C). Further, a prominent band of DYN immunoreactivity confined to the IML of the dentate was another distinctive characteristic of the MTLE hippocampi. This pattern is described as evidence of DYN sprouting (Fig. 2E).

The MTLE/DYN group of patients resembled the MTLE in all features except that they did not show DYN immunoreactivity in the dentate IML.

CA1 only group

The distinctive characteristic of the CA1 only group was the loss, often extensive, of pyramidal neurons in area CA1 along with increased gliosis, but with no reorganization of the dentate gyrus. The dentate gyrus resembled in all respects that in the MaTLE and PTLE hippocampi, as described earlier (Fig. 2A).

The immunohistochemical characteristics of the patient groups described earlier are summarized in Table 1.

Table 1.  Brief summary of immunohistochemical characteristics of patient hippocampi
  1. Interneuron loss, loss of neuropeptide Y (NPY), somatostatin (SOM), and substance P (SP) immunoreactive interneurons in the subgranular zone of the hilus; dynorphin (DYN) sprouting, DYN immunoreactivity in the inner molecular layer of the dentate; NPY/SOM/SP sprouting, immunoreactive fibers increased throughout the molecular layer of the dentate; MTLE, mesial temporal lobe epilepsy; PTLE, paradoxical TLE; MTLE/DYN, MTLE dynorphin negative; MaTLE, mass-associated TLE.

CA1 onlyNoneNoNo

Age, gender, and seizure history of patient groups

The comparison of gender, age at surgery, age at first intractable seizure, and years from first intractable seizure to date of surgery (duration of seizures) between patient groups and with autopsy controls are given in Table 2. In general, about equal numbers of male and female patients were studied, and the mean age did not differ significantly between groups. The age at first seizure was significantly shorter in the MTLE group (mean, 4.4 years) compared with the MaTLE (mean, 14.7 years) or PTLE (mean, 9.4 years) groups. Whereas the age at first seizure for the MTLE/DYN group (mean, 6.5 years) did not differ from that of the MTLE or PTLE, it was significantly shorter than that of the MaTLE group (mean, 14.7 years).

The age at first seizure in the CA1 group (7.9 years) was significantly older than that in the MTLE group (4.4 years). These data suggest that the initial pathologic event or initial precipitating injury (28) occurs early in the MTLE group and later in the CA1 group, even though both groups would be considered to show hippocampal sclerosis according to classic criteria. The seizure duration in the MTLE and CA1 group did not differ, and thus the difference in hippocampal injury cannot be attributed to this criterion. The shorter seizure duration in the MaTLE group compared with the other groups from which it significantly differs (PTLE, MTLE, MTLE/DYN, and CA1) may merely reflect the ease with which the source of pathology (mass lesion) can be detected by imaging in the former, and thus their coming to surgery earlier.

Neuronal densities associated with immunohistochemically distinguished groups

Autopsy, PTLE, and MaTLE

Just as the autopsy, PTLE, and MaTLE groups were indistinguishable immunohistochemically, their neuronal densities also were minimally different. The mean neuronal densities in different subfields of the hippocampus are given in Table 3. In comparison to the autopsy group, the PTLE and MaTLE hippocampi had only about a 20–25% reduction in neuronal densities across all hippocampal areas. The neuronal densities in the subregions of the hippocampus in the PTLE and MaTLE groups were indistinguishable from one another (Table 4).

Table 3.  Neuronal densities in hippocampal subfields
  1. Values expressed as group mean ± SD.

  2. DG, dentate granule cells; CA4 to CA1, subfields of Ammon's horn: the group types are defined in the text; SD, standard deviation from mean; MTLE, mesial temporal lobe epilepsy; PTLE, paradoxical TLE; MTLE/DYN, MTLE dynorphin negative; MaTLE, mass-associated TLE.

  3. The significant differences between these groups are shown in Table 4.

Autopsy (n = 26)313,603 ± 44,4418,880 ± 2,60916,729 ± 2,67120,497 ± 3,08914,393 ± 3,172
PTLE (n = 18)262,753 ± 44,6947,215 ± 1,84013,362 ± 4,11717,546 ± 3,98511,628 ± 2,560
MaTLE (n = 42)250,507 ± 63,8956,708 ± 2,35312,723 ± 4,04315,553 ± 4,15510,986 ± 2,948
MTLE (n = 72)138,449 ± 40,6132,723 ± 1,8025,984 ± 2,53010,409 ± 2,9013,365 ± 1,711
MTLE/DYN (n = 10)180,622 ± 47,5494,287 ± 2,3178,392 ± 3,21811,566 ± 3,3143,403 ± 800
CA1 group (n = 9)243,746 ± 53,8134,768 ± 1,4699,716 ± 2,14312,684 ± 2,3653,305 ± 315
Table 4.  Comparison of cell counts in regions of hippocampal formation
  1. Each number in this table represents the percentage of neurons in the group within parentheses, compared with the one without parentheses. All differences shown were statistically significant at p < 0.05.

  2. ND, no difference; CA1–CA4, fields of Ammon's horn; GC, dentate granule cells; MTLE, mesial temporal lobe epilepsy; PTLE, paradoxical TLE; MTLE/DYN, MTLE dynorphin negative; MaTLE, mass-associated TLE.


MTLE, MTLE/DYN, and CA1 groups

The MTLE hippocampi exhibited the greatest difference in neuronal densities in comparison with the autopsy, PTLE, and MaTLE groups. The MTLE hippocampus had neuronal densities generally <50% of those in the aforementioned three groups in all subfields. The greatest neuronal loss was in area CA1, and the least in area CA2 and the dentate granule cell layer (Table 4).

The differences in neuronal density in MTLE/DYN compared with autopsy, PTLE, and MaTLE hippocampi were similar to those of the MTLE group. The greatest decrease was in area CA1 and of the same magnitude as in MTLE (∼70% loss) compared with a 30–40% loss in other fields. However, in areas CA3, CA4, and the granule cell layer, the MTLE/DYN group retained more neurons (60–70%) than the MTLE group (40–60%).

In the CA1 group, a small decrease in neuronal density was seen in the granule cell layer (∼20%) compared with the autopsy group (Tables 3 and 4). This difference was of the same order as the difference between autopsy and PTLE or MaTLE in the same region. Thus granule cell density in the CA1 group was indistinguishable from that in PTLE or MaTLE. The neuronal density in area CA1 of the CA1 group was ∼70% less than that in PTLE or MaTLE (resembling the differences in the MTLE and MTLE/DYN and these groups), but was much smaller, ∼20–30%, in areas CA4, CA3, and CA2 (unlike MTLE but like MTLE/DYN).

The MTLE/DYN and CA1 hippocampi had similar neuronal densities in all areas of Ammon's horn. However, MTLE/DYN hippocampi had about a 25% lower neuronal density than the CA1 group in the dentate gyrus. The neuronal densities in areas CA1 were similar in the three groups (MTLE, MTLE/DYN, CA1), whereas MTLE had only ∼80% of the neurons of the CA1 group in area CA2 and 60% of the density of neurons in MTLE/DYN and the CA1 group in areas CA3 and CA4. The difference in granule cell densities between MTLE and MTLE/DYN was smaller (20% less) than between MTLE and CA1 group hippocampi (40% less) (Table 4).

The relative neuronal densities in each of the hippocampal subfields, discussed earlier, for the several patient subgroups are summarized in Table 5.

Table 5.  Summary of the relative neuronal densities in each of the hippocampal subfields across subject groups
  1. MTLE, mesial temporal lobe epilepsy; PTLE, paradoxical TLE; MTLE/DYN, MTLE dynorphin negative; MaTLE, mass-associated TLE.

Granule cells: Autopsy > PTLE/MaTLE/CA1 > MTLE/DYN > MTLE
Area CA4: Autopsy > PTLE/MaTLE > CA1 > MTLE/DYN > MTLE
Area CA3: Autopsy > PTLE/MaTLE > CA1 > MTLE/DYN > MTLE
Area CA2: Autopsy > PTLE/MaTLE > CA1/MTLE/DYN/MTLE
Area CA1: Autopsy > PTLE/MaTLE > CA1/MTLE/DYN/MTLE

Electrophysiologic characteristics associated with immunohistochemically distinguished groups

The four electrophysiologic measures of excitability for each patient group are given in Table 6. The IPSP score is a rough measure of the strength of inhibition, whereas the other three characteristics are indicators of excitability. These data showed that the MaTLE and PTLE hippocampi, although having intact inhibition and no signs of excessive excitability such as bursting activity, were nevertheless not identical (Table 6). The PTLE group showed greater excitability, based on EPSP score and spontaneous activity, than that seen in MaTLE. However, compared with the MaTLE and PTLE groups, the MTLE group was much more excitable, having a significant loss in inhibition (IPSP score) and strong indicators of excitability including bursting activity (Table 6). Compared with the MaTLE and PTLE groups, the MTLE/DYN group (like MTLE) had reduced inhibition (IPSP score) and signs of increased excitability (EPSP score, bursting). However, unlike MTLE, the MTLE/DYN did not have spontaneous activity. This was supported by a comparison of the MTLE and MTLE/DYN groups, which appeared similar on many parameters, but the MTLE group showed a tendency (p = 0.08) toward a higher frequency of spontaneous activity. The CA1 group, like the MaTLE and PTLE groups, showed that inhibition was not statistically (at p ≤ 0.05 level) different. However, there appeared to be a tendency to a reduction in IPSP scores (p = 0.08 and 0.09, respectively; Table 6). The CA1 group also had more excitation (burst, EPSP, and spontaneous activity) than the MaTLE group but was only slightly more excitable than the PTLE group (burst ratio).

Table 6.  Electrophysiologic characteristics of patient groups
  1. Values expressed as mean ± SD. The p values only for comparisons showing significant differences are indicated. In parentheses are the p values for group comparisons that are not statistically significant but may indicate a trend toward being so.

  2. IPSP, inhibitory postsynaptic potential; EPSP, excitatory postsynaptic potential; MTLE, mesial temporal lobe epilepsy; PTLE, paradoxical TLE; MTLE/DYN, MTLE dynorphin negative; MaTLE, mass-associated TLE.

MaTLE (n = 17)0.41 ± 0.63.06 ± 1.743.89 ± 630.47 ± 0.38
PTLE (n = 9)0.41 ± 0.43.12 ± 0.965.0 ± 2.410.91 ± 0.57
MTLE (n = 41)1.47 ± 0.654.9 ± 3.2416.73 ± 13.41.82 ± 1.05
MTLE/DYN (n = 8)1.58 ± 0.414.9 ± 2.169.02 ± 3.91.08 ± 0.80
CA1 group (n = 7)0.85 ± 0.574.36 ± 1.77.24 ± 3.81.21 ± 0.42
MaTLE/CA1 group(0.075)0.020.0150.003
PTLE/CA1 group(0.09)(0.056)(0.08)
MTLE/CA1 group0.0270.048
MTLE.DYN/CA1 group0.024

Dual pathology

The issue of what constitutes dual pathology is complicated because no good agreement exists among pathologists as to what pathologic entities must exist with hippocampal sclerosis to be classified as dual pathology. We consider here those patients who have a temporal lobe mass lesion along with hippocampal sclerosis as showing dual pathology. The criterion for hippocampal sclerosis we have chosen is a neuronal density of <60% of controls averaged across all areas of the hippocampus (29). By these strict criteria, only three MaTLE hippocampi of our patient group would have dual pathology. These three also had <60% neurons compared with controls in CA1. Two other hippocampi had averaged over all areas values of 63 and 61% but 49 and 47%, respectively, in area CA1. When considered separately, these five cases show several features in common. In all, the mass lesion was very close to the hippocampus. In three, the tumor had infiltrated the body of the hippocampus. In these three cases, the hippocampus had immunohistochemical reorganization, as described earlier. They also had a 70–80% neuron loss. In another patient, the tumor was in the amygdala and head of the hippocampus. This case showed no immunohistochemical reorganization, with only a 40% neuron loss. In one other patient, the mass lesion was in the amygdala. Immunohistochemistry in this case showed hilar interneuron loss but no sprouting, with ∼45% cell loss.

Surgical outcome

The surgical outcomes in the patient groups described in this study are summarized in Table 7. The classification of outcome proposed by Engel (30) was adopted. Class I of the classification includes patients that were essentially seizure free. The highest percentage of patients with a class I outcome was in the MTLE group (84.5%). The MaTLE, MTLE/DYN, and CA1-only group had similar percentages of patients with a class I outcome (∼75%) that were a little lower than those for the MTLE group. By comparison with all other groups, the PTLE group had a markedly lower percentage of patients with a class I outcome (44%). Class II of the classification is sometimes described as a marked reduction in seizures (>90%). The percentage of patients with class I and II outcomes taken together is >90% for all patient groups other than the PTLE group (67%). The PTLE group also had the highest percentage of patients with class III and IV outcomes (∼30%). In those patients with poorer outcomes (classes III and IV), no consistent differences were identifiable in any of the parameters examined in this study.

Table 7.  Surgical outcome
Class IClass IIClass IIIClass IV
  1. Values expressed as percentage of total for the patient group. The surgical outcome is presented in terms of the four classes proposed by Engle (30). Class 1, no seizures or only auras or rare owing to withdrawal of medication; class II, moderate decrease, >90%; class III, some decrease; class IV, no worthwhile change in seizures after surgery, <75% reduction.

  2. MTLE, mesial temporal lobe epilepsy; PTLE, paradoxical TLE; MTLE/DYN, MTLE dynorphin negative; MaTLE, mass-associated TLE.

CA1 only77.822.200


Are there distinctive categories of TLE patients?

The analysis presented here demonstrates that, on the basis of conventional histopathologic criteria, surgically removed hippocampi can be divided into two groups: (a) those with no hippocampal sclerosis compared with autopsy controls (MaTLE, PTLE), and (b) those with varying degrees of hippocampal sclerosis (MTLE, MTLE/DYN, CA1 group).

The patients with no hippocampal sclerosis also had no immunohistochemical evidence of reorganization of the dentate gyrus. The MaTLE and PTLE groups differed from the autopsy group in having a small (<25%) loss of neurons throughout the hippocampus. In addition to the absence of immunohistochemical changes, the MaTLE and PTLE groups resembled each other in their hippocampal volumes (31). However, the etiology of the disease in the two groups differed (mass lesion in MaTLE; unknown etiology in PTLE). Further, a correlation analysis of neuron densities in the different hippocampal subfields revealed that whereas there were significant correlations between neuron densities in the dentate gyrus and all areas of Ammon's horn in the MaTLE group, the PTLE group did not show such correlations (16). On electrophysiologic measures, the two groups, although resembling each other in having no loss of inhibition of granule cells and no burst activity, differ in that the PTLE show some signs of increased excitability: increased spontaneous activity and EPSP scores in dentate granule cells [(31) and this article]. On the measures of spontaneous activity and EPSP scores, the PTLE group lies between the MaTLE and MTLE groups. The surgical outcome in the PTLE group is considerably poorer than that in all other groups.

The MaTLE patients are recognizable, presurgically, by the detection of a mass lesion with MRI. The absence of such lesions in PTLE patients, their relatively normal hippocampal volume (31), and absence of a clear etiology such as febrile seizures differentiate this group from the MTLE patients before surgery. More regional neocortical ictal onset is the most common alternative focus under these circumstances and requires an intracranial electrode study for clarification.

Recently, Zaveri et al. (32) showed that power spectral density (PSD) analyses of background EEG, from intracranial monitoring, can be used to distinguish PTLE from MTLE patients. The PTLE patients have greater total signal power as well as greater power in the delta and theta frequency bands. Further, it was observed (Zaveri, unpublished data) that the signal power in the theta band has a significant positive correlation with dentate granule cell polysynaptic EPSP activity (from slice electrophysiology) in the MTLE but not PTLE patients, and a significant negative correlation between signal power in the alpha band with polysynaptic EPSP in the PTLE but not MTLE group. Such data suggest that the distinctive anatomic substrates (i.e., differences in neuronal density and reorganization) of the hippocampi may be related to the characteristic electrical activity patterns in these patient groups. PSD analyses have not been carried out on MaTLE patients, as their clinical evaluation does not frequently warrant intracranial electrode placement.

These observations profile the PTLE patients as a group distinct from the MaTLE patients. More important, the PTLE patients also are different from the MTLE group. We do not think the PTLE hippocampi are an intermediate form between MaTLE and MTLE, as the differences from either are substantial.

The patient hippocampi with hippocampal sclerosis show, immunohistochemically, varying degrees of reorganization in the dentate gyrus or none at all. The hippocampi showing evidence of immunohistochemical reorganization (MTLE, MTLE/DYN) have defined patterns of neuronal density associated with the immunohistochemical patterns. The MTLE group shows the most extensive neuronal loss and reorganization. The MTLE/DYN group is similar to MTLE patients except that they show less neuronal loss in the dentate, CA4, and CA3 areas. They show all the anatomic features of MTLE except evidence of DYN staining in the IML. Electrophysiologic data from dentate granule cells also confirms the close similarity of the two groups. The surgical outcome is good in both groups but a little better for the MTLE than for the MTLE/DYN group. Although we had previously argued (17), on the basis of a smaller sample of patients for whom electrophysiologic data were available, that the MTLE/DYN group was a distinctive group, we now conclude that the MTLE/DYN group may represent an intermediate stage in the development of the MTLE type.

The CA1 group hippocampi show clear evidence of classic Ammon's horn sclerosis. However, they do not show immunohistochemical evidence of dentate gyrus reorganization. On the basis of neuronal densities, the group closest to the CA1 group is the MTLE/DYN group. The former differs from the latter only in the greater density of dentate granule cells and the preservation of the populations of peptidergic neurons (SOM, NPY, and SP) in the hilar subgranular region of the former. On the basis of just these two features, the CA1 group exactly resembles the PTLE and MaTLE groups, with the densities of surviving granule cells being in the same range as those in the latter two groups. The CA1 group, in comparison to the MaTLE and PTLE groups, shows no statistically significant loss of inhibition of dentate granule cells. However, the CA1 group has other measures of excitability (burst ratios, EPSP scores, and spontaneous activity scores) that are greater than in MaTLE but not PTLE. In contrast, the MTLE group in comparison with the CA1 group shows significant loss of inhibition of dentate granule cells. The MTLE group also has a higher EPSP score than the CA1 group. The age at onset of the first intractable seizure is significantly older in the CA1 group than in MTLE patients. Although it may be inferred from the later onset of seizures in the CA1 group that the lack of dentate injury may be because of shorter seizure history, this is not so. Both the CA1 group and MTLE had similar seizure duration in our study. Thus whereas the MTLE/DYN group closely resembles the MTLE group, the CA1 group does not. Sagar and Oxbury (9) also described a group of TLE patients similar to our CA1 group. They described a group of patients showing “non-specific hippocampal sclerosis”[i.e., those with neuron loss only in the H1 field (CA1) and no dentate granule cell loss]. In their group, the age at the first convulsion was longer (older than 4 years) than in hippocampi with classic Ammon's horn sclerosis (younger than 4 years).

What is the epileptogenic mechanism in these groups?

We and others (21,33) argued previously that in the MTLE-type patients, the loss of hilar peptidergic neurons is an important event in the pathophysiology of epileptogenesis in this group. Perhaps related to this is the observed correlation only between dentate granule cell densities and hilar neuron densities but not with other CA fields in MTLE, suggesting a possible causal relation between them. The clear relation between the loss of these hilar interneurons and concomitant loss of inhibition (increased IPSP scores) and increased burst activity in MTLE compared with all groups in which these neurons are not lost, suggest their importance in normally controlling hippocampal hyperexcitability in this group. Their loss in very young (younger than 5 years) TLE patients in the absence of any prominent sprouting is additional evidence of their potential importance (15). The precise circuitry and mechanisms by which these interneurons contribute to the normal inhibition of granule cells in the human hippocampus remains unexplained. Recent evidence implicates NPY as being an inhibitory transmitter on dentate granule cells in the human hippocampus (34). However, the analyses in the present article point to a probable significant contribution of changes in area CA1 as well as to granule cell excitability. Increased bursting activity in dentate granule cells (a clear indicator of hyperexcitability) in the CA1 group (in the absence of changes in dentate IPSP scores) compared with MaTLE and PTLE (all three groups that show little or no loss of hilar interneurons) suggests a potential contribution from this region to hippocampal hyperexcitability. Such a contribution may override the normal inhibition provided by hilar interneurons, as could be inferred from the hyperexcitability in the CA1 group. The mechanisms by which area CA1, a region with extensive loss of principal neurons (in the MTLE and CA1 groups) and the presence of a substantial population of presumptive interneurons, influences hippocampal hyperexcitability remains a puzzle. A clue to its excitatory potential may be related to the significant gliosis of this area. Recent studies imply that astroglia in sclerotic hippocampi may have neuronal-like excitable properties along with altered glutamate transporter function and be contributors to hippocampal hyperexcitability (35). A fuller understanding of the role of astroglia in epileptogenic hippocampi awaits study.

It also is possible that some extrahippocampal brain region(s) may be responsible for the epileptogenicity and pathology in the CA1 group. The location of such a region, if present, is unknown. One subcortical region, the intralaminar nucleus reuniens of the thalamus, is a candidate region worthy of further examination. In rat and cat, this nucleus is shown to have a direct projection to the hippocampus, where projecting axons terminate specifically in the stratum moleculare of the CA1 region and layers I and III of the entorhinal cortex. However, no projections extend to the dentate gyrus (36,37). It has been shown that this and other nonspecific regions of the thalamus, when stimulated, resulted in widespread changes in cortical EEG (presumably via other reuniens projections to the cortex).

The location of the epileptogenic focus in the PTLE group remains unresolved. The regional neuronal densities and the correlation patterns of neuronal densities between hippocampal subregions in PTLE resemble most closely, although they are not identical to, those in the autopsy control group. If we assume that the hippocampi in the autopsy group are not spontaneously excitable, the source of the spontaneous activity and polysynaptic EPSPs must be explained. It is possible that hitherto unidentified changes in the microcircuitry of these hippocampi, responsible for this activity, may be present. Conversely, the observed excitation in the hippocampus may be driven by an extrahippocampal input to the hippocampus. The origin(s) of such inputs remains speculative at best.

The epileptogenic focus in MaTLE patients seems to be the region of tissue where the mass lesion is located. Removal of such lesions in extratemporal neocortical locations appears to produce good seizure control (38). This also may be the case in the MaTLE group in which the mass lesion is located in the temporal lobe outside the hippocampus (i.e., in the entorhinal cortex and parahippocampal gyrus) and is removed along with adjacent regions of hippocampus for seizure control. The minimal hippocampal pathology observed in the MaTLE may be a result of excitation passing into the hippocampus from the lesion. The epileptogenic mechanism in this group may best be elucidated by analysis of mass lesions and the neuropil surrounding them (39).

Acknowledgment: We thank Ms. Ilona Kovacs for her excellent assistance with all histologic studies on which this article is based. This work was supported by NIH grants NS39092 and by the Lee Foundation.