ArticlePDF Available

Inhibitory neurons in the human epileptogenic temporal neocortex - An immunocytochemical study

Authors:

Abstract

Immunocytochemical methods were used to study alterations in inhibitory neuronal circuits in human neocortex resected during surgical treatment of intractable temporal epilepsy associated or not with brain tumours. The epileptogenic cortex was characterized and divided into spiking or non-spiking zones by intraoperative electrocorticography (ECOG). The resected cortex was cut into blocks, sectioned and stained immunocytochemically for visualization of glutamic acid decarboxylase (GAD), the calcium-binding protein, parvalbumin (PV) and glial fibrillary acidic protein (GFAP). A variety of alterations in cortical neuronal circuits as revealed by immunocytochemical and histological methods were found. Similar alterations in inhibitory neuronal circuits appear to occur independently of the primary epileptogenic site and pathology associated with epilepsy, which suggests that there is possibly a common basic underlying mechanism that leads to seizure activity. These changes were apparently unrelated to ECOG findings at surgery, which bring into question the value of the use of interictal epileptic discharges recorded by ECOG to guide cortical resections. The most conspicuous and common change was the loss of chandelier cells. The finding that these cells are among the most vulnerable types of GABAergic interneurons in the epileptogenic temporal cortex indicates that they might be of great functional importance, since the axon terminals of chandelier cells are likely to exert powerful regulation of impulse generation in cortical pyramidal cells. Therefore, these cells might represent a key component in the aetiology of human epilepsy.
Brain (1996), 119, 1327-1347
Inhibitory neurons in the human epileptogenic
temporal neocortex
An immunocytochemical study
Pilar Marco,1 Rafael G. Sola,2 Paloma Pulido,2 Maria T. Alijarde,3 Alicia Sanchez,3
Santiago Ramon y Cajal4 and Javier DeFelipe1
1 Instituto Cajal, Departments of2Neurosurgery and
^Neurophysiology, Hospital de la Princesa and the
^Department of
Pathology,
Clinica Puerto de Hierro,
Madrid
Correspondence to: Dr J. DeFelipe, Instituto Cajal (CS1C),
Avenida Dr Arce, 37, Madrid 28002, Spain
Summary
Immunocytochemical methods were used to study alterations
in inhibitory neuronal circuits in human neocortex resected
during surgical treatment of intractable temporal epilepsy
associated or not with brain tumours. The epileptogenic
cortex was characterized and divided into spiking or
non-spiking zones by intraoperative electrocorticography
(ECOG).
The resected cortex was cut into blocks, sectioned
and stained immunocytochemiccdly for visualization of
glutamic acid decarboxylase (GAD), the calcium-binding
protein, parvalbumin (PV) and glial fibrillary acidic protein
(GFAP).
A variety of alterations in cortical neuronal circuits
as revealed by immunocytochemical and histological methods
were
found.
Similar alterations in inhibitory neuronal circuits
appear to occur independently of the primary epileptogenic
site and pathology associated with epilepsy, which suggests
that there is possibly a common basic underlying mechanism
that leads to seizure activity. These changes were apparently
unrelated to ECOG findings at surgery, which bring into
question the value of the use of interictal epileptic discharges
recorded by ECOG to guide cortical resections. The most
conspicuous and common change was the loss of chandelier
cells. The finding that these cells are among the most
vulnerable types of GABAergic interneurons in the
epileptogenic temporal cortex indicates that they might be
of great functional importance, since the axon terminals of
chandelier cells are likely to exert powerful regulation of
impulse generation in cortical pyramidal cells. Therefore,
these cells might represent a key component in the aetiology
of human epilepsy.
Keywords: focal epilepsy; brain tumours; glutamic acid decarboxylase; parvalbumin; gliosis
Abbreviations: ECOG = electrocorticography; GAD = glutamic acid decarboxylase; GFAP = glial fibrillary acidic protein;
GluR2/3 = glutamate receptor subunits 2 and 3; GluR5/6/7 = glutamate receptor subunits 5, 6 and 7; PV = parvalbumin;
WHO = World Health Organization
Introduction
Temporal lobe epilepsy is one of the most frequent types
of human focal epilepsy. There is a variety of epileptogenic
lesions and the most important hypotheses regarding the
basic mechanisms of epilepsy in both humans and
experimental animals are based on alterations (anatomical
and/or chemical) of glutamatergic (excitatory) and
GABAergic (inhibitory) cortical neuronal systems (Lloyd
et al., 1986; Sherwin and van Gelder, 1986; van Gelder,
1987;
Houser, 1991; Ribak, 1991; Avanzini et al, 1992;
Hamberger and van Gelder, 1993). However, it is not yet
© Oxford University Press 1996
known why cortical tissue becomes epileptogenic and why
some patients can be well controlled on antiepileptic drugs,
whereas others (the minority) are uncontrollable. In the latter
group, a relatively high percentage can be surgically treated;
the most common surgical procedure being the removal of
the anterior temporal cortex, most of the amygdala and the
anterior portion of the hippocampus (Olivier, 1992). The
most frequent pathological change found in the resected
tissue is so-called sclerosis of
the
hippocampus (e.g. Falconer,
1974;
Meldrum and Bruton, 1992; Babb and Pretorius, 1993),
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
1328
P.
Marco et al.
B
Fig. 1 (A) MRI of Patient HI7 showing a pilocytic astrocytoma localized mesially in the right temporal lobe (arrow). (B) Lateral view of
resected anterior temporal lobe from Patient HI7, which includes the tumoural mass (arrow).
but the lateral temporal cortex is usually considered to be
normal (e.g. Babb et al., 1984). However, it has been recently
shown that in lateral human epileptogenic cortex with normal
appearance in routine histopathological preparations there
may be small regions or patches of decreased immunostaining
for the calcium-binding protein PV and/or for the glutamate
receptor subunits 2 and 3 (GluR2/3) and 5, 6 and 7 (GluR5/
6/7) of the D. L-a-amino-3-hydroxy-5-methyl-4-isoxazole
propionic acid/kainate and kainate receptor subtypes
(DeFelipe et al., 1993, 1994). Furthermore, using correlative
light and electron microscopic immunocytochemical methods,
these focal decreases have been shown to correspond to a
fine disorganization in synaptic circuits which consists of an
increase and decrease of presumptive excitatory and inhibitory
synapses, respectively (DeFelipe et al., 1993; P. Marco and
J. DeFelipe, unpublished work). In addition, PV immuno-
cytochemistry labels a subpopulation of GABAergic inter-
neurons which includes chandelier and basket cells. Because
these two types of cells innervate the somata (and proximal
dendrites) and axon initial segments of pyramidal cells,
respectively, they are considered to be (in particular chandelier
cells) the most powerful inhibitory interneurons in controlling
pyramidal cell excitability (for review, see DeFelipe and
Farinas, 1992). Thus, PV immunoreactivity can be utilized
as a useful tool to study abnormal synaptic circuits in the
human epileptogenic cortex that may be particularly relevant
to epileptogenesis (DeFelipe et al., 1993). Moreover,
immunocytochemical studies of focal epilepsy using
antibodies directed against the GABA-synthesizing enzyme,
GAD in monkeys made experimentally epileptic, have shown
a preferential loss of GABAergic neurons and axon terminals
at epileptic foci (Ribak et al., 1979, 1982; Houser et al.,
1986;
Houser, 1991). It was hypothesized that this loss of
inhibition could lead to epileptic activity of cortical pyramidal
neurons (Ribak et al., 1979).
In the present study, the major aims were (i) to investigate
further possible alterations in the inhibitory neuronal circuitry
that might occur in the human neocortex, using immuno-
cytochemistry for PV and GAD on samples obtained during
surgical treatment of temporal lobe epilepsy associated or
not with brain tumours; and (ii) to compare these patterns of
immunostaining with each other, and with intraoperative
ECOG recordings (i.e. spiking or 'active' and non-spiking
or 'non-active regions'). In addition, these patterns of
immunostaining and ECOG findings were also compared
with cytoarchitectural features in thionin-stained sections
adjacent to those used for immunocytochemistry, and with
the patterns of immunostaining for GFAP. Immuno-
cytochemistry for GFAP was used for detection of reactive
astrocytes, which are a frequently detected by histo-
pathological methods in epileptogenic cortex (e.g. Meldrum
and Bruton, 1992).
Material and methods
Tissue preparation
Human cortical tissue was obtained, during surgical treatment,
from 13 patients with intractable temporal lobe epilepsy
associated (n = 4) or not associated (n = 9) with brain
tumours. Informed consent was obtained in all cases. The
line of excision was based on clinical evaluations, EEG
studies (including foramen ovale electrodes recording and
chronic monitoring from subdural grid electrodes in some
cases),
neuropsychological assessment, MRI signals, CT
scans (in some cases), intraoperative ECOG, and the
appearance of the tissue during surgery. MRI signals were
normal in the group of patients without brain tumours. The
probable primary epileptogenic site was localized (using
the electrophysiological data) in the amygdalo-hippocampal
region, where most interictal epileptiform activity and all the
ictal activity were recorded, in all cases but one (Patient
H10),
in whom it was localized in the neocortex. However,
there was a significant interictal activity arising from the
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
Neocortical inhibitory cells in epilepsy 1329
lateral temporal cortex in all cases. The four patients with
brain tumours showed abnormal MRI signals either near
the amygdalo-hippocampal region (Patients H16 and HI7)
(Fig. 1) or in the lateral neocortex and white matter (Patients
H9 and HI
1).
The clinical data are summarized in Tables 1
and 2. Surgical procedures were performed under general
anaesthetic. In the cases without brain tumours (except Patient
H10),
the resection was a tailored anterior temporal cortical
resection including most of the amygdala and 1-3 cm of the
hippocampus. In Patient H10 only a cortical resection was
performed. In the tumour cases, the surgical procedure was
directed both to mass lesion removal and to improve seizure
control by cortical resection tailored to the extent of the
epileptogenic zone determined by interictal epileptiform
abnormalities on the intraoperative ECOG.
Tissue preparation has been described in detail in DeFelipe
et al. (1993). Briefly, the neocortex was characterized and
marked as spiking or non-spiking by extracellular recordings
with grid electrodes at the time of surgery. All operative
procedures and preparation of tissue samples were recorded
with a video camera. The resected brain tissue was
immediately immersed in a cold solution of 4% para-
formaldehyde in phosphate buffer (0.1 M, pH 7.4), then
photographed and cut into small blocks about 15x8x8 mm
that included either spiking or non-spiking regions, as guided
by the video camera recording. The blocks were then
postfixed, sectioned and stained immunocytochemically for
GAD,
PV or GFAP as indicated below.
Immunocytochemistry
The blocks were transferred to a second fixative solution of
4%
paraformaldehyde in phosphate buffer in which they
remained for 24 h at 4°C. Thereafter, they were cut serially
at 100 |im on a Vibratome. Sections were processed for
immunocytochemistry for GAD, PV and GFAP. The sections
were pretreated with a solution of ethanol and hydrogen
peroxidase in phosphate buffer to remove endogenous
peroxidase activity, washed in phosphate buffer, and then
preincubated in 3% normal serum and 0.05% Triton X-100
in phosphate buffer for 3 h at room temperature. Then, the
sections were transferred to the latter solution to which either
anti-PV antibody (mouse monoclonal antibody from Swant,
Bellinzona, Switzerland) (Celio et al., 1988), diluted
1:5000,
anti-GAD antibody (rabbit polyclonal antibody, K-2 from
Chemicon, Temecula,
Calif.,
USA) (Kaufman et al., 1991),
diluted
1:1000,
or anti-GFAP antibody (rabbit polyclonal
antibody from Sigma, St Louis, Miss., USA), diluted 1:50,
were added. Sections were incubated for 24 h at 4°C. They
were rinsed in phosphate buffer and transferred to a
solution containing biotinylated horse anti-mouse or
biotinylated goat anti-rabbit immunoglobulins (Vector Labs,
Burlingame,
Calif.,
USA) for 1 h at room temperature. The
sections were washed several times in phosphate buffer and
incubated at room temperature in avidin-biotin-peroxidase
complex (Vector Labs) for 1 h. After another series of
washes, they were transferred to a solution of 0.05% 3,3'-
diaminobenzidine tetrahydrochloride and 0.01% hydrogen
peroxide in phosphate buffer for 2-3 min. The sections were
then washed and osmicated in 0.02% osmium tetroxide,
washed again and finally dehydrated, cleared with xylene
and coverslipped with DePex mounting medium (BDH,
Poole, UK). Sections adjacent to those used for immuno-
cytochemistry were stained with thionin. Some tissue samples
from medial temporal structures and lateral neocortex were
embedded in paraffin, sectioned and stained with
haematoxylin and eosin for the neuropathological evaluation
of the removed tissue. Control sections for immuno-
cytochemistry were processed as above but with primary
antiserum replaced with normal serum, or in the case of PV
immunocytochemistry with primary antiserum adsorbed with
an excess of PV (Swant). No significant staining was detected
under these control conditions.
Control tissue sections consisted of post-mortem human
neocortical tissue (area 38 of Brodmann) from two individuals
(Cl and C2) with no known neurological disease. These
were kindly supplied by Dr C. Bouras (Geneva, Switzerland).
The characteristics of these human samples were Cl, 70-
year-old male (2 h post-mortem delay); C2, 91-year-old male
(6 h post-mortem delay). These samples were fixed and
processed as descibed for the biopsies.
Results
Immunocytochemistry for GAD and PV in the
normal neocortex fixed by immersion: controls
The patterns of immunostaining for GAD and PV in sections
from post-mortem human neocortical tissue (control tissue
sections) which were processed identically and in parallel to
those sections obtained at biopsy from epileptic patients,
were consistently homogeneous; i.e. no patches of decreased
immunostaining were observed. Also the pattern of labelling
of neurons and axonal plexuses was similar to that previously
described by other authors in the neocortex of a variety of
species (see Discussion).
Epileptogenic neocortex cases not associated
with brain tumours
Histopathology of medial temporal structures
While cutting brain tissue into small blocks before
postfixation (see Materials and methods), it was noticed that
the cortex and white matter of the parahippocampal gyrus in
the sample from Patient H12 were indurated, whereas the
texture of the other samples was normal. Neuropathological
examination of the removed hippocampus and/or para-
hippocampal gyrus showed significant histopathological
findings in Patients H7, HI2, H25, H26 and H35 (see Table
1).
In Patients H27 and H18 no significant pathological
changes were found (although in the case of Patient H18
the hippocampus was not available for pathological assess-
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
Table
1
Summary of clinical and surgical data in cases not associated with brain tumours o
I
Patient Age
(years)/
sex
Side Age of Possible
onset aetiology
(years)
Seizure Seizure Neurological Anti-epileptic Neuropsychological Date of
frequency type examination drugs examination surgery
(lQ/hemispheric (month/year)
dominance)
Length* (cm)/ Neuropathological Surgical
weight (g) findings^ results
of removed
brain tissue
H7
HIO
HI2
HIS
H25
13/M
38/M
35/M
31/F
Right
Left 15
Right 19
Left 24
UnknownWeekly Partial
complexDeaf and dumb, CBZ
right-handed
44/M Right 28Unknown Weekly Partial Normal, PHT
complex right-handed
H26
H27
H34
H35
32/M
29/M
19/F
30/F
Right
Right
Left
Right
2
1
II
7
?Viral
encephalitis
Complicated
delivery;
multiple
febrile
seizures
Unknown
Unknown
Daily
Daily
Weekly
Daily
Partial
complex.
secondary
generalized
tonic-clonic
Partial
complex,
secondary
generalized
tonic-clonic
Partial
complex
Partial
complex
Normal,
right-handed
Normal,
right-handed
Normal,
right-handed
Normal,
right-handed
PB,
PHT
CBZ, PB
VPA
PHT, PR
CBZ, VC
<7O/left
Unknown Weekly Partial Normal, PHT, PRM 120/left
complex. right-handed
secondary
generalized
tonic-clonic
Complicated Daily Partial Normal, CBZ, VPA 120/left
delivery complex, right-handed
secondary
generalized
tonic-clonic
Complicated Daily Partial Normal, CBZ, PB 110/left
delivery complex,secondoigtit-handed
generalized
tonic-clonic
90/left
7O/left
105/left
110/left
110/left
2/1992
7/1992
9/1992
1/1993
6/1993
6/1993
6/1993
10/1994
11/1994
5.0/24.3
2.8/7.0
4.4/25.2
Neuronal loss and
gliosis in the
hippocampus and
parahippocampal
gyrus
Seizure-free
Exceptional
seizures
Seizure-freeNeuronal loss and
gliosis in the
hippocampus and
parahippocampal
gyrus
4.3/26.1
No significant Seizure-free
alterations in the
parahippocampal
gyrus (hippocampus
not analysed)
4.2/27.01iosis Seizure-free
in the
hippocampus;
calcifications
in the
parahippocampal
gyrus
5.3/-
Neuronal loss in the Seizure-free
parahippocampal
gyrus (hippocampus
not analysed)
4.2/24.0
3.3/11.4
6.7/31.8
No significant
alterations
No data available
Gliosis in the
parahippocampal
gyrus
Important
seizure
frequency
decrease
Seizure-free
Seizure-free
Patient HIO was surgically treated only with a cortical resection. Therefore, the medial temporal structures were left intact. CBZ = carbamazepine; PB
PRM = primidone; VGB = vigabatrin; VPA = valproate. 'From the temporal pole; ^in the hippocampus and parahippocampal gyrus.= phenobarbital; PHT = phenytoin;
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
Table 2 Summary of clinical and surgical data in cases associated with brain tumours
PatientAge
(years)/
sex
SideAge of
onset
(years)
Possible
aetiologySeizure
frequencySeizure
typeNeurological
examinationAnti-epileptic
drugsNeuropsychological
examination
IQ/hemispheric
dominance)
Date of surgery
(month/year)Length* (cm)/
weight (g)
of removed
brain tissue
Neuropathological
findings*
Surgical
results
H9
Hll
HI6
HI7
19/M
30/F
23/F
IS/M
Lett
Right
Right
Right
5
7
14
17.5
Tumour
Tumour
Tumour
Tumour
Daily
Daily
Daily
Monthly
Partial
complex
Partial
complex
Partial
complex
Partial
complex,
secondary
generalized
tonic-elonic
Normal,
left-handed
Normal,
right-handed
Normal,
right-handed
Normal,
right-handed
PRM
PHT
CBZ, PB,
PHT, CLB
PHT
89/right
70/left
IIO/left
125/left
6/1992
7/1992
2/1992
2/1992
5.4/18.1
5.0/20.5
4.5/32.0
5.5/22.8
Ependymoma
Anaplastic
astrocytoma
(WHO grade III)
Pilocytic
astrocytoma
(WHO grade I)
Pilocytic
astrocytoma
(WHO grade 1)
Seizure-free
Unchanged
Exceptional
seizures
Seizure-free
CBZ = carbamazepine: CLB = clobazam: PB = phenobarbital; PHT = phenytoin; PRM = primidone. *From temporal pole; *type of tumour.
I'
S-1
Co
S'
8
I
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
1332 P. Marco et al.
*;# -.*»:
••
•* ;
*<•••"%
>'.
Fig. 2 (A) Low power photomicrograph of a section through the parahippocampal gyrus stained with
haematoxylin and eosin from Patient H12 showing neuronal cell loss and gliosis (satellitosis, arrows).
(B) Higher magnification of A, showing a neuron surrounded by glia (arrow). Scale bar = 50 urn for
A, 23 urn for B.
ment).
No data were available for Patient H34. These
changes consisted mainly of neuronal cell loss and/or gliosis
(Fig. 2A and B) in the hippocampus and/or parahippocampal
gyrus.
In Patient H25, the pathology found was neuronal cell
loss and gliosis in the hippocampus and calcifications in the
parahippocampal gyrus (Fig. 3). The amygdala were only
available for pathological examination in Patients H7 and
H25 and, in both cases, neuronal cell loss and gliosis
were found. ,
Cytoarchitecture and patterns of immunostaining
for
PV,
GAD and GFAP in the lateral neocortex
Tissue samples used for immunocytochemistry were from
anterior portions of the temporal lobe (areas 38, 20 and
21 of Brodmann). In thionin-stained sections from the 36
cortical blocks from the nine patients examined (Table 3),
25 blocks presented a normal cytoarchitecture, eight displayed
gliosis and three showed small regions with cell loss (focal
cell loss) (Fig. 4). Within the regions showing cell loss, many
of the remaining neurons were darkly stained and shrunken
(Fig. 4D) compared with adjacent normal cortex (Fig. 4C)
and some of them showed pericellular incrustations (Fig. 5).
Sections from 32 blocks were processed for GFAP
immunocytochemistry. Numerous GFAP-positive cells were
found in sections from only seven blocks in at least one
cortical layer as compared with the other blocks which
showed few GFAP-positive cells in the grey matter. Therefore,
we distinguished two patterns of GFAP immunostaining in
the grey matter: pattern 0 (normal, see Discussion) when few
or no GFAP-positive cells were present, and pattern + +
(abnormal) when numerous cells were stained. It was notable
that only a few blocks showed gliosis in Nissl stained sections
and also only a few blocks showed pattern ++ of GFAP,
and that different cortical layers were affected (layer VI and
layers I—III, respectively) {see Table 3).
We distinguished four main patterns of PV-immunostaining
(patterns A, B, C and D) (Fig. 6), as have previously been
described in human epileptogenic neocortex (DeFelipe
et al., 1993). Pattern A was characterized mainly by the
labelling of numerous nonspiny interneurons (including
chandelier cells and basket cells) and processes and by the
labelling of a dense immunoreactive band in the middle
layers (Fig. 6A). This band is made up of immunoreactive
cell somata, processes and puncta [originating, in part, from
the axon terminals of nonspiny interneurons and, in part,
from terminals of thalamocortical neurons {see DeFelipe
et al., 1993; del Rio and DeFelipe, 1994)]. The three
other main patterns (Fig. 6B-D) of immunostaining showed
decreased immunoreactivity to a variable extent and were
considered abnormal. In the present study, the analysis of a
larger number of cortical blocks from more patients has
allowed us to make a more detailed definition compared with
our previous studies of these altered patterns, particularly
pattern C. Patterns B (Fig. 6B) and D (Fig. 6D) were
characterized by a reduction in immunoreactivity (in intensity
of staining and number of immunoreactive elements) only in
the layers above the superficial half of layer III or in all
layers (from layer I to layer VI), respectively. These two
patterns and, in particular, pattern B, usually affected
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
Neocortical inhibitory cells in epilepsy 1333
\
Bf
c *
Fig. 3 (A) Low power photomicrograph of a section through the parahippocampal gyrus stained with
haematoxylin and eosin from Patient H25, showing calcifications (arrows) in the deeper layers of the
parahippocampal gyrus. (B and C) Higher magnification of
A.
These calcium deposits (arrows) display
a concentric rings structure. Scale bar = 236 |im for A, 10 um for B and C.
relatively large segments of cortex (often >1000 (im wide).
Pattern C was the most outstanding because this pattern of
decreased immunoreactivity appeared in rather small regions
(often 200-1000 um wide), while the surrounding cortex
showed a normal pattern of immunostaining (Fig. 6C). The
decreased immunoreactivity consisted of a virtual absence of
immunoreactive neurons (whereas in patterns B and D there
are relatively numerous immunoreactive neurons, although
lightly stained) and a reduction in immunoreactive processes,
affecting mainly the chandelier cell axon terminals.
Furthermore, the decreased immunostaining can affect a
single or several of the superficial, middle or deep cortical
layers. (In the present study we referred to these patches of
decreased immunostaining as pattern C with a subscript
indicating the layers showing that decrease.) Finally, block
la from Patient H18 showed a pattern of immunostaining
that, at first glance, was indistinguishable from pattern A,
but a closer inspection showed a virtual lack of chandelier
cell axon terminals in layers II and III. This kind of pattern
was commonly found in cases associated with brain tumours
(see below). We referred to this pattern as A (-Ch) with a
subscript indicating the layers with such reduction. In general,
in patterns B and D all types of PV-positive cells appear to
be equally affected, whereas in pattern C the consistent lack
of chandelier cell terminals is notable while the number of
terminal-like puncta may vary from a moderate or severe
reduction to normal. Therefore, chandelier cells are among
the most vulnerable. As shown in Table 3, of the 36 blocks
processed for PV immunocytochemistry, 29 showed a single
pattern (27 pattern A; two pattern B), whereas the other
seven blocks displayed a mixture of patterns [five patterns
A and C; one pattern A(-Ch) and C; one pattern B and D].
Furthermore, pattern A is the predominant pattern in the
blocks showing patterns A and C. Thus, the normal pattern
of PV immunostaining is the most frequently found.
In sections stained immunocytochemically for GAD, two
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
Table 3 Summary of immunocytochemical results and cytoarchitecture (in adjacent serial sections unless otherwise specified) in non-spiking and spiking cortical
regions
from patients without brain tumours
Patient
H7
HIO
HI2
HIS
H25
H26
H27
H34
H35
Sample location within
the
neocortex
Block
2:
area
38
Block
7:
area
21
Block
2:
area
20
Block
4a:
area
21
Block
1:
area
38
Block
2:
area
20
Block
3b:
area
21
Block
4:
area
38
Block
5:
area
38
Block
la:
area
38
Block
2a:
area
38
Block
4:
area
21
Block
5:
area
21
Block
7:
area
21
Block
9:
area
21
Block
10:
area
21
Block
6:
area
21
Block
7:
area
20
Block
8:
area
20
Block
9:
area
21
Block
10:
area
20
Block
1:
area
38
Block
2:
area
21
Block
7:
area
20
Block
8:
area
20
Block
II:
area
38
Block
12:
area
38
Block
2:
area
21
Block
6:
area
21
Block
7:
area
21
Block
11:
area
38
Block
1:
area
38
Block
4:
area
38
Block
5:
area
21
Block
1:
area
21
Block
17:
area
20
ECOG
Non-spiking
Spiking
Non-spiking
Spiking
Non-spiking
Non-spiking
Non-spiking
Non-spiking
Non-spiking
Non-spiking
Spiking
Spiking
Spiking
Non-spiking
Spiking
Non-spiking
Spiking
Spiking
Non-spiking
Spiking
Non-spiking
Non-spiking
Spiking
Non-spiking
Spiking
Non-spiking
Non-spiking
Spiking
Spiking
Spiking
Non-spiking
Non-spiking
Non-spiking
Non-spiking
Non-spiking
Spiking
Pattern staining
GAD*
_
a
a
-
a
at,ci_ins
-
a
a
at,c,_,v§
a
a
"t,c,,i_iv§
a
a
-
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
Sample location within
the
neocortex
Block
2:
area
38
Block
7:
area
21
Block
2:
area
20
Block
4a:
area
21
Block
1:
area
38
Block
2:
area
20
Block
3b:
area
21
Block
4:
area
38
Block
5:
area
38
Block
la:
area
38
Block
2a:
area
38
Block
4:
area
21
Block
5:
area
21
Block
7:
area
21
Block
9:
area
21
Block
10:
area
21
Block
6:
area
21
Block
7:
area
20
Block
8:
area
20
Block
9:
area
21
Block
10:
area
20
Block
1:
area
38
Block
2:
area
21
Block
7:
area
20
Block
8:
area
20
Block
11:
area
38
Block
12:
area
38
Block
2:
area
21
Block
6:
area
21
Block
7:
area
21
Block
11:
area
38
Block
1:
area
38
Block
4:
area
38
Block
5:
area
21
Block
1:
area
21
Block
17:
area
20
Pattern staining
PV*
A
A
Bt,
B
A
At,
A
A
A
A(-
A
A
At,
A
A
At,
A
A
A
A
A
A
A
A
A
A
B
A
A
A
A
A
A
A
At,
At,
D
c,_,,,s
CH),]_|,it,
C,_iv
C|,i_vS
Ci_in
Cum*
c,,,«
Cytoarchitecture^
Normal
Normal
Normal
Normal
Normal
Normalt.
gliosis
(ns)
Normal
Gliosis,_vi
Normal
^
Normal
(ns)
Normal
Normal
Normalt,
focal cell
loss,,,_\
Normal
Normal
Normal
(ns)
Normal
Normalt,
gijOsiS|_V]
Normal
Normal
Normalt,
gliosisVi
Normal
Normal1'.
gliosisVi
Normal
Normal
Normalt,
g|josisVi
Normal
Normalt,
gliosisv,
Normal
Normal
Normalt,
gliosisv_vi
Normalt,
foca|
cen
loss,,.,,,
(ns)
Normal
Normal
Normalt,
foca|
ce]|
lossn_n
Normal
(ns)
GFAP
staining*
_
-
0
-
0
0
-
0
0
0
0
0
/S0
0
0
0
0
01". + + |
0
0
0t. ++1-1M
0
++1-111
0
0
0
+ +
1-1II
()t, ++,
+
+1
+
+UV1
0
0
0
0
,S
0
i
Subscripts indicate cortical layers.
*See text
for
explanation
of
patterns
of
staining;
tpredominant;
^when
the
patterns
of
staining
for
GAD
and PV and
abnormal cytoarchitecture
were
coincident;
"'Nissl staining
in
adjacent
serial sections
to
those
immunocytochemically stained
for
GAD
and
PV, except
when indicated
by (ns),
that means that
the
sections were
non-
adjacent.
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
Neocortical inhibitory cells in epilepsy 1335
m
H
III
°<
IV^
V
n
•• - :--::x\±:a
VI
Fig. 4 Photomicrographs of a 100 (im thick vibratome section through area 38 stained with thionin
from Patient H34 (block 1), showing a small region with severe neuronal cell loss: (A and B) low (A)
and higher magnification (B) photomicrographs illustrating a region with focal neuronal cell loss
(asterisks) in layer III; (C and D) photomicrographs illustrating normal looking neurons (C) from a
normal region adjacent to the area of neuronal cell loss (D) which shows fewer and abnormal looking
neurons (arrows) compared with C. Scale bar = 509 \xm for A, 280 \im for B, 50 (im for C and D.
main patterns of immunostaining (a and c) were distinguished.
Pattern a (Fig. 7) was characterized by the presence of
a relatively large number of immunostained cell somata
(particularly in layers II—IV), and numerous terminal-like
puncta throughout all layers. The unstained somata of
pyramidal cells were frequently outlined by numerous
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
1336
P.
Marco et al.
Fig. 5 High power photomicrographs of neurons within the region showing neuronal cell loss illustrated
in Fig. 4D. The neurons are darkly stained and shrunken and show pericellular incrustations (arrows).
Note the staining of proximal dendritic processes (open arrows) in B. Scale bar = 10 u.m for A and B.
immunoreactive puncta (Fig. 7C and D). This pattern was
similar to that previously described in the normal primate
neocortex (e.g. Houser et al., 1984) and was considered to
be normal. Pattern c was considered to be abnormal {see
Discussion) and was characterized by the decrease in both
immunoreactive puncta and, particularly, of cell somata
(Fig. 8). As occurred with pattern C of the PV immuno-
staining, the decrease in GAD immunostaining was found to
affect a single layer or several of the superficial, middle or
deep cortical layers and appeared in small regions of the
same size as pattern C of the PV immunostaining. Therefore,
we used the same terminology indicating the layers showing
that decrease by subscripts. It was notable that only few
patterns c were found (Table 3): only five out of 32 blocks
processed for GAD immunocytochemistry showed pattern c.
Correlation between patterns of immunostaining
for PV and GAD
As shown in Table 3, in all cases where the PV staining
pattern was normal (pattern A) in a given section, a normal
pattern for GAD immunocytochemistry (pattern a) was found
in the adjacent section, and that when pattern C of PV
immunoreactivity was present in a section, this decrease was
mirrored by coincident decrease of GAD immunoreactivity
(pattern c) in the adjacent section. Therefore, there was a
complete coincidence of patterns A and a, and C and c
for PV and GAD immunoreactivities in the same regions.
However, this correlation did not occur with patterns B and
D of PV immunostaining, since in the corresponding adjacent
sections that were stained for GAD the pattern of
immunostaining was normal (pattern a).
Correlation between patterns of immunostaining
for GAD and PV with the pattern of
immunostaining for GFAP and cytoarchitecture
Comparison of sections immunocytochemically stained for
PV or for GAD with sections that had been stained with
thionin or immunocytochemically for GFAP, showed that a
given pattern of immunostaining for PV or GAD was
apparently not associated with any particular cytoarchitectural
characteristic nor with a high or small number of GFAP-
positive cells, except the correlation between patterns C and
c and focal cell loss (Table 3). Examination of adjacent serial
sections that had been stained with thionin, revealed that
the regions showing these abnormal patterns were always
associated with cortical tissue showing focal cell loss, with
surrounding tissue being of normal appearance. Recently,
this correlation with patches of decreased immunostaining
has been demonstrated more directly after examination of
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
Neocortical inhibitory cells in epilepsy 1337
**.
J
r .
VI
Bn
Fig. 6 Photomicrographs showing the main patterns of PV immunostaining observed in the human epileptogenic cortex. Pattern A
immunostaining is considered to be normal, whereas patterns B-D, showing decreased immunostaining are considered to be abnormal.
Note, in C decreased immunoreactive elements in a small region in layer MIA (asterisk). Scale bar = 200 |im for A-D.
toluidine blue stained semi-thin (2 |im thick) plastic
sections obtained after resectioning of plastic embedded
sections (P. Marco and J. DeFelipe, unpublished observa-
tions).
In thionin-stained sections, this focal cell loss was
seen as a moderate or severe decrease in neurons (Fig. 4),
and the decrease affected (at least in the severe cases) all
kinds of neurons (i.e. pyramidal cells and nonpyramidal
cells).
Often the regions with focal cell loss were so small
that they were very difficult to identify in thionin-stained
sections. However, once a region with a pattern c of GAD
or pattern C of PV immunostaining was identified (which is
a change that is easily detected), then it was readily identified
as a decrease in cells in the adjacent thionin-stained section
(Fig. 4).
Correlation of intraoperative ECOG with
cytoarchitecture and patterns of immunostaining
for GAD, PV and GFAP
As shown in Table 3, both non-spiking and spiking regions
(as determined by intraoperative ECOG) displayed either
normal or abnormal cytoarchitectural characteristics or
patterns of immunostaining for
PV,
GAD or
GFAP.
Therefore,
there was a lack of correlation between electrocortico-
graphical and anatomical findings.
Epileptogenic neocortex cases associated with
brain tumours
Histopathology
In these patients (Patients H9, HI 1, HI6 and HI7), MRI
signals revealed the existence of intracerebral masses, which
after surgical removal and neuropathological examination,
were confirmed to be tumours (Fig. 1). Tumours were
classified according to Russell and Rubinstein (1989) and
the World Health Organization (WHO) (Kleihues et al.,
1993).
In Patients H16 and HI7, the tumours were located
near the amygdalo-hippocampal region, whereas in Patients
H9 and HI
1
the tumours affected the lateral neocortex and
white matter. The tumours were identified (Table 2) as
pilocytic astrocytomas (WHO grade I) (Patients H16 and
HI7),
ependymoma (Patient H9) and anaplastic astrocytoma
(WHO grade III) (Patient Hll). The long duration of the
symptoms (23 years) in the latter case could be explained
by a malignant transformation of
a
previous low grade tumour
into a higher grade (e.g. see Wolf et al., 1993).
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
133H
P.
Memo et al.
I
II
1IIA
1MB
IV
<
VI-
* p
Fig. 7 (i \l) iiiiiiiunostaining thnniyh l>i\er>. I-\ 1 nt aica 21 tioin Patient HIS ihlm.k " i. show my
pattern ii i normal i ot Ci\I) iiiiniunovtainiiiL' i \i
I.OVK
pouei phoioniitroyraph ^ho^Mns; numerous
immunovtained cell soniatj throuiihout all Livers (Bi Higher niayiiitication photomitrogr.iph through
la\eis III-\-I\'. illustiatini; the \aiiL't\ ol moqihologKal
INIX'S
ot Ci \[)-pnMti\e neuron*, i.inousi iC
and I)i Photomicrograph through la\er I11B. showing numerous imiiiunoreacti\e puiKta Note the
unstained somata ot p\ramidal cells ipi outlined h\ immunoieacti\e puiicta Scale hai = 25s urn lor A.
KHI um toi B '2 urn lor I' and I)
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
Neocortical inhibitory cells in epilepsy 1339
I
B
t
$
*
f
Fig. 8 (A) GAD immunostaining through layers IIIA-V of area 21 from Patient HI8 (block 5),
showing pattern c (abnormal) of GAD immunostaining. Arrow indicates a focal decrease of
immunoreactivity. (B and C) Higher magnification photomicrographs of the centre (B) and periphery
(C) of the region with decreased immunostaining. Note the decrease of both immunoreactive puncta
and, particularly, of cell somata, in the central part. Scale bar = 311 urn for A, 54 jim for B, 50 |im for
C.
Cytoarchitecture and patterns of immunostaining
for PV, GAD and GFAP in the lateral neocortex
Tissue samples used for immunocytochemistry were non-
tumoural samples (confirmed by neuropathological assess-
ment) from anterior portions of the temporal lobe (areas 38
and 21 of Brodmann) which were located at a distance from
the tumour. In thionin-stained sections from the nine cortical
blocks from the four patients examined (Table 4) it was
found that four blocks displayed a normal cytoarchitecture
(Fig. 9A), two showed gliosis and three showed a small
region with focal cell loss. In sections processed for GFAP
immunocytochemistry from the seven blocks examined,
numerous GFAP-positive cells and processes were found in
all but two blocks.
In sections immunocytochemically stained for GAD and
PV, the same patterns of immunostaining as in cases not
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
I
Table 4 Summary of immunocytochemical results and cytoarchitecture (in adjacent serial sections) in non-spiking and spiking cortical
regions
from patients with
brain tumours
Patient Sample location ECOG
within
the
neocortexPattern staining GAD* Sample location within Pattern staining PV* Cytoarchitecture11 GFAP staining*
the neocortex
H9
HII
HI6
HI7
Block
3:
area
38
Block
2:
area
38
Block
3:
area
38
Block
la:
area
38
Block
3:
area
38
Block
4a:
area
38
Block
1:
area
21
Block
2:
area
21
Block
7:
area
21
Spiking
Non-spiking
Non-spiking
Non-spiking
Non-spiking
Non-spiking
Non-spiking
Non-spiking
Non-spikingat,c,_,v§
Block
3:
area
38
Block
2:
area
38
Block
3:
area
38
Block
la:
area
38
Block
3:
area
38
Block
4a:
area
38
Block
1:
area
21
Block
2:
area
21
Block
7:
area
21
A(-Ch)M_vl
A(-Ch)n.vl
A(-Ch)1Mvt
C,,_vs\D
A(-Ch)IWM
A
Af,
Cm_vS
A^B
A
AT,
B,
C,_IV§
Normal
GliosiS]_vi
GliosiS|_vi
.
local cell
lossn_iv8
Normal
Normal
Normalf,
focal cell
lossm_vs
Normal
Normal
Normal^, focal cell
l0SS|,IV§
-
+ +I-VI
+ +I-VI
0
+ +1I-VI
+ + I-VI
_
+ +I-V1
0
For abbreviations
and
symbols,
see
Table
3.
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
Neocortical inhibitory cells in epilepsy 1341
••:••*'
it
i
Fig. 9 Parvalbumin immunostaining through area 38 of the human epileptogenic neocortex in cases associated with brain tumours. (A
and B) Photomicrographs of paired surfaces of adjacent sections, one stained with thionin (A) and the other for PV (B), from block la of
Patient HI6. (C) Photomicrograph of a section stained for PV from block 2 of Patient HI
1.
Asterisks indicate the dense immunoreactive
band extending from the lower half of layer III to layer IV typical of the normal temporal cortex. Arrows indicate immunoreactive
neurons in layers II and V, which are also shown in Fig. 10 at a higher magnification. Scale bar = 300 urn for A, B and C.
associated with tumours were found. It was notable that very
few blocks showed a normal pattern of immunostaining for
PV (in only two out of nine) and that pattern A (-Ch) was
commonly found (Figs 9B and C and 10) (Table 4).
Correlation between patterns of immunostaining
for PV and GAD
As occurred in the cases not associated with tumours, there
was a complete coincidence of patterns A and a. and C and
c for PV and GAD immunoreactivities. respectively in the
same regions (Table 4). For comparison between the other
abnormal patterns of immunostaining for PV and GAD,
unfortunately sections from only three blocks showing
abnormal patterns of PV immunostaining (two showing a
pattern A (-Ch) and the other a pattern B) were available
for GAD immunocytochemistry. In these sections, the pattern
of immunostaining for GAD was normal (pattern a), as in
the cases not associated with tumours.
Correlation between patterns of immunostaining
for GAD and PV with the pattern of
immunostaining for GFAP and cytoarchitecture
Comparison of sections immunocytochemically stained for
PV or for GAD with sections that had been stained with
thionin or immunocytochemically for GFAP showed a lack
of correlation between normal and altered patterns, except
the correlation between patterns C and c and focal cell loss
(Table 4). Again, these results were the same as in the cases
not associated with tumours. The only difference found
between the two groups of patients was that in all cortical
blocks from patients with brain tumours at least one of
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
1342
P.
Marco et al.
Fig. 10 Higher magnification photomicrographs of Figs 9B (A and C) and 9C (B and D), showing
patterns A(-Ch)IMn and A(-Ch)M_VI of PV immunostaining. respectively (see text for further details).
Open arrows in A and B indicate the same immunoreactive neurons in layer I! as in 9B and C. N'ote
the lack of immunoreactive chandelier cell terminals (short, vertically oriented rows of puncta). Open
arrows in C and D indicate the same immunoreactive neurons in layer V as in Fig. 9B and C. Note that
in C there is a the large number of immunoreactive chandelier cell terminals (arrows), whereas in the
same layer in D there is a dramatic decrease of these terminals. Scale bar = 100 u.m for A and B, 204
)im for C and D.
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
Neocortical inhibitory cells in epilepsy 1343
the patterns of immunostaining or the cytoarchitecture was
abnormal.
In summary, the abnormalities found in the epileptogenic
neocortex of patients with brain tumours were the same as
in the cases without brain tumours, but in the former group
of patients the alterations were more commonly found than
in the other group.
Correlation of ECOG with cytoarchitecture and
patterns of immunostaining for GAD, PV and
GFAP
All cortical regions examined except one displayed a non-
spiking (normal) activity as determined by intraoperative
ECOG. However, all blocks showed an abnormal
cytoarchitecture or altered patterns of immunostaining.
Therefore, the lack of correlation between abnormal circuitry
and abnormal ECOG in these cases was even more clear
than in the non-tumoural cases.
Discussion
In the present study we have made three main observations.
First, in human neocortex resected during surgical treatment
of intractable temporal lobe epilepsy associated or not with
brain tumours, multiple small regions with abnormal patterns
of immunostaining for PV and GAD were commonly found.
Secondly, there is a variety of changes in inhibitory circuits
and these changes may, in certain regions, affect the whole
population of inhibitory neurons, whereas in other regions
they may affect a subpopulation of
these
neurons (in particular
chandelier cells) more selectively. Thirdly, there were no
consistent changes in any of the parameters examined that
could be correlated with the normal or abnormal ECOG
characteristics of the samples at surgery. These abnormal
patterns of immunostaining were unlikely to be due to uneven
fixation in different portions of the tissue block and/or
differences in the penetration of immunocytochemical
reagents, first because control sections from post-mortem
human neocortical tissue show no such patches of decreased
immunostaining, and secondly because previous correlative
light and electron microscopic studies of regions showing
focal decreases in PV-immunoreactivity have shown that, in
these regions, there is a disorganization in synaptic circuitry
{see below).
Immunocytochemical and cytoarchitectural
changes what is normal and abnormal?
The main limitation in these kinds of studies is, of course,
that we do not have data about patterns of immunostaining
for GAD and PV from biopsy samples of the strictly normal
human neocortex. It is generally thought that cortical regions
that had been resected in order to gain access to tumours or
other abnormalities located deep in the brain are good
controls. However, possible damage to subcortical projection
pathways to the neocortex, as well as other alterations (e.g.
drug therapy) which may affect the neocortex, should not be
disregarded. Furthermore, tumours may lead to changes that
eventually cause epilepsy after variable delays (Ketz, 1974;
Spencer et al., 1984; Morris and Estes, 1993). Another more
common source of neurochemical and anatomical data on
non-pathological human neocortex comes from autopsy
samples. It is true that a relatively long post-mortem delay
may severely affect levels of amino acid such as GABA
(e.g. Lloyd et al., 1986), but does not apparently affect
immunostaining for certain antigens (e.g. PV
immunostaining; see Bliimcke et al., 1990; Ferrer et al.,
1991;
Hof et al., 1991). Furthermore, there are numerous
studies on the cerebral cortex in a variety of other primates
and non-primates mammals, using the same or different
antibodies as in the present study (e.g. Celio, 1986; Ribak,
1978;
Houser et al., 1984; DeFelipe et al., 1989; Esclapez
et al., 1994). In our control material (autopsy samples
processed as the biopsies) and other studies on normal
neocortex, there is no indication of patches of decreased
immunostaining for GAD and PV as found in the epileptic
neocortex. Thus, patterns A and a of immunostaining for PV
and GAD are considered to be normal because these patterns
are virtually identical to the patterns of immunostaining
found in the normal neocortex of
a
variety of species, whereas
the patterns showing decreased of immunostaining (pattern
c of GAD immunostaining and patterns B, C and D of PV
immunostaining) are considered to be abnormal (DeFelipe
et al., 1993; see below). In the case of immunocytochemistry
for GFAP, the staining of grey matter fibrous astrocytes
(which are immunoreactive for GFAP) varies widely among
individuals (e.g. Hansen et al., 1987), but this number is
rather small in the normal brain of relatively young individuals
(Meldrum and Bruton, 1992) as for the cases in the present
study. Therefore, we considered the appearance of many
GFAP-positive cells (pattern ++) to be an abnormal pattern.
Finally, since mild neuronal cell loss or gliosis is difficult to
evaluate due to the great variation in the numbers of cortical
glial cells and neurons between individuals (e.g. Henderson
et al., 1980), the cytoarchitecture in thionin-stained sections
was considered abnormal only when clear neuronal cell loss
and/or gliosis was observed.
Changes in the cytoarchitecture and patterns of
immunostaining for GAD, PV and GFAP in
cases not associated with brain tumours
Relatively few cortical blocks showed abnormal patterns
of immunostaining for GAD, PV and GFAP or abnormal
cytoarchitecture (16%, 25%, 22% and 31%, respectively).
However, if we consider the blocks in which at least one of
the patterns of immunostaining or the cytoarchitecture was
abnormal, then the majority (56%) of the blocks presented
some kind of alteration. Furthermore, we must keep in mind
by guest on February 20, 2013http://brain.oxfordjournals.org/Downloaded from
1344
P.
Marco et al.
that only a relatively small extent of cortex was analysed for
each staining (on average four 100 \im thick sections per
block were used for each immunocytochemical or thionin
staining) and, therefore, there may be small regions with
altered patterns (particularly patterns C and c) that might be
not included in the analysis.
In conclusion, multiple, small, altered cortical regions
seem to be a common characteristic of the epileptogenic
neocortex in patients without brain tumours. Comparison of
adjacent sections stained for PV and GAD, showed that there
was a complete coincidence of normal patterns A and a, and
altered patterns C and c for PV and GAD immunoreactivities,
respectively, in the same regions. However, this correlation
did not occur apparently in regions with patterns B and D
of PV immunostaining, since a normal pattern of immuno-
staining for GAD was found in these regions.
Changes in the cytoarchitecture and patterns of
immunostaining for GAD, PV and GFAP in
cases associated with brain tumours
In the neocortex of epileptic patients with brain tumours, we
found the same changes as in the cases not associated
with brain tumours, i.e. regions showing altered patterns of
immunostaining for GAD, PV and GFAP, focal cell loss and
gliosis (Table 4). There was also a correlation between normal
and abnormal patterns of immunostaining for GAD and PV
(patterns A and a and patterns C and c, respectively).
However, abnormal patterns of immunostaining were more
frequently found in the cases associated with tumours than
without tumours.
Since within each group of patients the 'primary' pathology
associated with epilepsy varied (e.g. different types of
tumours, neuronal cell loss or not in the hippocampus, etc.)
and similar alterations in inhibitory neuronal circuits appear
to occur in all cases, this suggests that there is possibly a
common basic underlying mechanism that leads to seizure
activity. In both tumoural and non-tumoural cases, among
the inhibitory interneurons that are lost, chandelier cells
appear to be one of the most affected type of cell and
therefore they might play a crucial role in the aetiology of
epilepsy (see below). Finally, it is possible that the degree
of some of the pathological changes found in the hippocampus
(or adjacent areas) might be related to the duration of
symptoms (Mathern et al., 1995), but in the neocortex this
was apparently unrelated.
Possible significance of the
immunocytochemical changes
GAB A is the neurotransmitter used by the majority of smooth
nonpyramidal neurons (short-axon cells or interneurons).
It is well-established that different types of GABAergic
interneurons innervate both pyramidal cells and nonpyramidal
cells (Houser et al., 1984). Among the interneurons that
innervate pyramidal cells are chandelier cells and basket
cells,
which are immunoreactive for PV and are thought to
play an important role in the control of pyramidal cell
excitability (DeFelipe etal., 1989, 1993; Hendry etai, 1989;
Bliimcke et al., 1990; Lewis and Lund 1990; Ferrer et al.,
1991;
Hendrickson et al., 1991; Hof et al., 1991; Williams
et al., 1992; del Rio and DeFelipe, 1994). Co-localization
immunocytochemical studies have shown that virtually all
PV-positive cells are also immunoreactive for GABA and
that a large proportion of GABA cells also display
immunoreactivity for PV (for review, see DeFelipe, 1993).
Thus a decrease in GABA neurons in a given cortical region
could lead to a general reduction of inhibition in all kinds
of neurons (pyramidal and nonpyramidal) located in that
region, whereas a decrease in PV immunoreactive neurons
could lead to a more selective although important reduction
in the inhibitory synaptic control on pyramidal cells.
It has been shown that expression in cortical neurons of
a variety of transmitters or transmitter-related substances
(including GAD and PV) can be altered in an activity-
dependent manner (Hendry and Jones, 1986, 1988; Bliimcke
et al., 1994; for a review see Jones, 1993). Thus, it is possible
that the decrease in GAD and PV immunoreactivities in the
epileptogenic neocortex could be due to activity-dependent
fluctuations in the immunocytochemical staining, and that
these changes are reversible and dynamic. In a previous
study of similar material (DeFelipe et al., 1993), cortical
regions showing decreased immunostaining for PV were
examined (in particular cortical regions with pattern C) at
the electron microscope level. In these regions cell death
(apoptotic bodies) and an apparent decrease in axon terminals
forming symmetric synapses were found. In the present study,
regions showing patterns C and c of immunostaining for PV
and GAD, respectively, presented a clear cell loss and this
loss affected nonpyramidal cells but also pyramidal cells in
a variable degree. Immunocytochemical studies with
an