A decrease in the number of GABAergic somata is associated with the preferential loss of GABAergic terminals at epileptic foci.

Previous studies have indicated that a loss of GABAergic terminals occurs at epileptic foci. The present study was undertaken to investigate if this loss is associated with a loss of GABAergic neuronal somata. Seven juvenile monkeys (M. mulatta) received alumina gel injections to the pre-central gyrus of the left cerebral hemisphere to produce epileptic foci. Four of these monkeys were chosen for further quantitative study. One was sacrificed prior to seizure onset ('pre-seizure'), one had seizures for 3 days ('acute'), and two had a seizure record of one month ('chronic'). Sections of tissue from the epileptic cortex and from the contralateral, non-epileptic cortex were processed for glutamate decarboxylase (GAD) immunocytochemistry at the light microscopic level. Quantitative analysis revealed that a loss of GAD-positive neuronal somata ranging from 24 to 52% occurred at epileptic foci for all monkeys. This decrease was significant (P less than 0.01) for the two chronic monkeys. There was also a slight decrease in GAD-positive neurons 1 cm distal to the focus ('parafocus') in the chronic monkeys, but not in the acute or pre-seizure animals. In addition, small GAD-positive somata (50-150 micron2) were more severely decreased in number at epileptic foci than larger ones (200-250 micron2). As an experimental control, an additional monkey was given a surgical lesion in area 4 of one cerebral hemisphere. It did not display seizure activity prior to sacrifice and did not show a loss of GAD-positive neurons proximal to the control lesions. The results of this study indicate that a loss of GABAergic neuronal somata is associated with a loss of GABAergic terminals at epileptic foci, and that this loss may be more specific for the small GABAergic neurons.


INTRODUCTION
Numerous studies from our laboratory have indicated that a loss of GABAerg~c terminals occurs at epileptic loci 20-23 This loss was shown to be preferentlal for GABAerglc, symmetric synapses with quantitative electron microscopic methods21. The results of biochemical studies support these findings in that indices for other neurotransmitters were not reduced as severely as those for GABA 1,2,29. In fact, these data obtained from experimental animal models of focal epilepsy have been predictwe for the pathophyslology of human epilepsy because biochemical studies of human epileptic foci also indicate a GABA deficit 14 Therefore, a loss of GABAerglc terminals is a hallmark of focal epilepsy.
The present study was undertaken to mvesngate if th~s loss of GABAerglc terminals was associated with a loss of GABAerglc somata. Since the terminals of both cortical basket and chandeher cells are dramatically reduced by over 80% at the epileptic focus, it was predicted that this loss was due to a neuronal loss and not simply a pruning back of the axonal plexus of these GABAergic neuronal types20, 21. In addition, the results of an analysis of Nlssl-stamed preparations of alumina gel epileptic foci have demonstrated a small neuronal loss 6 Therefore, it would be important to know if GABAerglc somata were decreased m number at epllepnc foci To explore this point, we utthzed an anti-glutamate decarboxylase (GAD) serum that detects somal GAD m many types of neurons without the use of colchtclne 15. This antmerum would facilitate a better comparison of epileptic and non-epileptic nssue because other antlsera that require colchlcme provide non-umform laminar staining of GABAerglc neuronal somata that is depend-Correspondence C E Rlbak, Department of Anatomy, University of California at Irvlne, Irvme, CA 92717, U S A ent on the site of colchicme rejection H. In addition, the damage caused by the injection may cause nonspecific damage to the cortex that could obscure the actual data on the number of GABAergic somata. The results of this study are consistent with the notion that a loss of GABAergic somata is associated with the preferential loss of GABAergic terminals at epileptic foci.

MATERIALS AND METHODS
Eight juvenile monkeys (Macaca mulatta), weighmg 2.5-4 kg, were used for this study. Each monkey displayed a normal scalp electroencephalogram (EEG) prior to surgery. Under general anesthesia a left frontal craniotomy was performed. The precentral gyrus (Brodmann's area 4) was identified and location of the hand/face area was confirmed by cortical stimulation. Seven monkeys received alumina gel applications using the Ward modification of the Kopeloff technique 13. No anticonvulsants were administered postoperatively. The eighth monkey was given a surgical control lesion which consisted of a subpial resection of approximately 5 x 5 mm down to the white matter (an area equivalent to that of a mature granuloma observed in this model). Serial scalp EEGs were subsequently performed on all monkeys on a weekly basis. Chronic seizure momtoring was similar to that previously reported 1. Of the 7 experimental monkeys, 4 were used for extensive quantitative analysis. The other 3 did not display a high enough quality of lmmunocytochemical staining for a meaningful interpretation of quantitative data. However, they were used for the qualitatwe description of the GAD-positwe neuronal somata.
Of the 4 monkeys used for the quantitative analysis, one ('pre-seizure') was sacrificed prior to the development of clinical seizures (RB-161). An electrocorticogram (ECoG) demonstrated sharp waves adjacent to the developing granuloma. One animal ('acute') was sacrificed 6 weeks following the injection of alumina gel (RB-160). This animal had 4 days of clinically evident seizure activity with a seizure frequency of 2-3 per day. Another animal ('chromc') was sacrificed two months following the alumina injection (RB-163). This ammal had 3 weeks of clinically evident seizure activity with a frequency of approximately one seizure every other day The fourth 79 monkey ('chronic') was sacrificed 2.5 months following the injection of alumina (RB-165). This animal had seizures for 6 weeks, demonstrating a seizure frequency of approximately one seizure per day. Each of the epileptic animals demonstrated an ECoG conslstent with spike and wave discharges in the area immediately adjacent to the developing granuloma. The control monkey was sacrificed 2.5 months following the surgical control lesion (RB-166). This ammal demonstrated no clinical seizure activity and the ECoG was normal.
Following the ECoG, each ammal was deeply anesthetized with sodium phenobarbital and perfused transcardially with 0.2% paraformaldehyde m balanced phosphate buffer (pH 7.3). A low paraformaldehyde concentration was used in the fixative so that tissue from the same monkey could be analyzed for receptor binding. This primary fixative is not optimal for lmmunocytochemistry and may partially explain the variability in the number of ~mmunoreactive somata between monkeys (Table I). The skull and dura were rapidly removed after the perfusion The tissue block from the left precentral gyrus was immediately removed and included the area adjacent to the alumina injection ('proximal' tissue) and tissue 1-2 cm superior to the injection site ('distal' tissue). A block from the right precentral gyrus was cut for control tissue from the homotopic cortical area because the resuits of Harris and Lockard 7 show that the contralateral homotoplc cortex in monkeys with alumina gelreduced seizures does not show independent seizure actwity. Both blocks were immediately placed m 10% formalin for further fixation and placed in a cryoprotectant solution of 30% sucrose for 24 h prior to sectioning. Sections 40 am thick were cut on a freezing microtome and placed m 0.1 M phosphate-buffered sahne for light microscopic immunocytochemlstry.
Briefly, free-floating tissue sections were processed for the immunocytochemlcal locahzat~on of GAD using the GAD antibody characterized by Oertel et al.15. Sections from the experimental and contralateral hemispheres were incubated simultaneously. A modification of Oertel's protocol15 was used, employing the avidin-biotin-horseradish peroxldase complex 12 (Vector Labs ) in a double-bridged techtuque 16. Following immunocytochemical processing, the sections were mounted on subbed glass slides,

Numbers of GAD-postttve neuronal somata m 4 regtons of cortex for 5 monkeys
In each section of tissue processed for GAD lmmunocytochemlstry (3-11 sections/animal), the number of GAD-posmve somata was counted m an area of tissue covered by 3 adjacent, 240/~m-w~de radial traverses throughout the entire depth of the cortex For the first 4 (experimental) monkeys, counts were made at the focus, parafocus, and two contralateral sites In the fifth (control) monkey, counts were made proximal and distal to the control lesion, and at two comparable contralateral s~tes  One section from each hemisphere of every monkey was processed as described above except that normal sheep serum was used instead of anti-GAD serum. Such sections displayed no staining above a diffuse background level when examined with a light microscope.
Sections from the experimental and contralateral hemispheres were examined under the light microscope with a x40 objective, at both the 'proximal' and 'distal' edges of the experimental tissue, and at corresponding sites in the contralateral tissue. A grid reticule (240/~m wide) was used to count the number of GAD-positive neuronal somata. Three adjacent traverses were made through the entire thickness of the cortex for each of the 4 sites from 3-11 sections per monkey. Cells which contained a dense, brownish reaction product within the penkaryal cytoplasm were considered to be GAD-positive. The average numbers of GAD-positive somata per 1.0 mm 2 were compared between each of the 4 tissue sites examreed. An analysis of varmnce and the Newman-Keuls multiple comparison procedure were used to determine if the numbers of somata at each site were slgmficantly different for each monkey Camera lucida drawings of about 100 GAD-positive cell bo&es per site were made at a magnification of × 700 from a plane of focus that demonstrated the greatest extent of the soma. These data were entered mto an Apple II Plus computer by tracing them onto a Houston Instrument HI-PAD dlgmzmg tablet A Bioquant I1 program for computerized morphometry was used to calculate somal areas of GAD-positwe cells

Description of GABAergtc neurons m normal and epileptic monkey motor cortex
GAD-positive neurons were found in all layers of non-epileptic monkey motor cortex. They were &stributed rather homogeneously throughout the layers (Fig. 1). However, layer I and the deeper half of layer VI appeared to have slightly fewer GABAerglc neurons than the other layers.
Neurons which were stained with the GAD antiserum contained a dark brown reaction product in their perlkaryal cytoplasm. Many of the somata of these GAD-positive neurons appeared to have eccentrically-placed nuclei. The nuclei and most dendritic processes were unstained. Since the reaction product &d not extend into the dendrites and axons, neurons could not be classified on the basis of their dendrmc and axonal morphology. The shapes of the neuronal somata, however, indicated that the GAD-posmve cells were non-pyramidal. Most somata were round, oval, teardrop-shaped, or fuslform, while some were irregular in shape (Figs. 1 and 3A). A few neurons had trmngular somata, though not of a size to suggest that they were pyramidal cells. With the exception of layer I, there was no difference m shape dlstnbunon throughout the layers of cortex. Layer I neurons all had round or slightly oval somata.
The sizes of GAD-posmve somata also varied. Almost all of the GAD-positive neurons in layer I had very small somal areas. In contrast, somata m layers II-VI &splayed a wide variety of sizes Thus, small, GAD-posmve neurons were found in all cortical layers but the average somal area tended to increase from the superficml to deep laminae, such that increasing numbers of larger neurons were found m the deeper layers Large somata were only occasionally observed In more superficml layers GAD-posmve neurons located at sites both proximal and &stal to epdeptic loci were slmdar in size and shape to those m normal monkey motor cortex ( Fig  2) In addmon, they were observed throughout all of the cortical layers, although m reduced numbers as compared to the contralateral cortex No pamcular layers seemed more affected than the others Another observanon for this focal region was the dramanc loss of GAD-posmve puncta (cf Fig 3A, B) as reported m an earlier study:2

Quantitative differences between eptleptw and normal cortex
The number of GAD-positive neuronal somata at sites proximal to the alumina gel focus was less than that found at distal or contralateral sties in the brains of epileptic monkeys (Table I). For example, monkey RB-163, which had seizures for 3 weeks, showed 70 GAD-positive cells/mm 2 at the focus, 97 cells/mm 2 at the distal lpsllateral site, and 108 and 109 cells/mm 2 at two corresponding sites m the contralateral cortex Monkey RB-165, which had a 6-week history of setzures, displayed a similar dlsmbutlon o! cells Since the average numbers of GAD-positive somata m the two examined regions of the contralateral cortex of each monkey were very similar (Table  I), they were averaged together for each of the 4 epileptic monkeys The average number of cells/mm2 at sites proximal and distal to each epileptic focus was expressed as a percentage of the contralateral cortex for the 4 monkeys used in the quantltatwe analysis. This mampulatlon of the data allowed for a direct comparison between monkeys because the raw data &splayed a wide variation due to the fixation protocol (see Methods). The data (Fig. 4) indicated that the loss of GAD-positive cells at the site proximal to the focus ranged from 35 to 52% for the two chromc epdeptlc monkeys. In contrast, monkey RB-160, which had only a 4-day record of severe seizures, showed a 24% loss of GAD-posttive somata at the epileptic focus The monkey (RB-161) that had been injected with alumina gel but had not displayed seizure acttvity prior to sacrifice showed a similar loss of GAD-posttive somata (28%) proximal to the focus. The distal ipsdateral site (parafocus) in these two latter animals did not display a loss of GAD-positive somata. Instead, the number of cells at the parafocus was slightly more than that of the contralateral, control cortex. In contrast, the two monkeys with chronic seizure records (RB-163 and RB-165) displayed a loss of GAD-positive neurons at the parafocus.
Statistical analyses were made to determine ff the decrease in the number of GAD-posttive somata proximal to epileptic foci was stattstically signtficant. A two-factor analysis of variance showed that, for the 4 alumina gel-injected monkeys as a group, the numbers of GAD-posmve somata at the focus, parafocus and two control sites were significantly different (P < 0.01). The Newman-Keuls multiple compartson procedure did not replicate this finding To determine if the numbers of cells at each site were mgnificantly different for each mdwldual monkey, a one-factor analysis of variance and the Newman-Keuls procedure were used. Both tests showed that

Comparison of somal areas for GAD-posttive neurons m 5 monkeys
For the first 4 (experimental) ammals, the range and mean somal area for GAD-posaive neurons at the focus, parafocus and two contralateral s~tes are gwen In the fifth (control) monkey, these measurements were taken proximal and distal to the control lesion, and at two comparable contralateral s~tes   (Table II) As an experimental control, a fifth monkey was given a surgical lesion in area 4 of one cerebral hemisphere. This animal did not display any seizure activity prior to sacrifice. It also did not show extensive glial scarring at the lesion site, as did the monkeys mjected with alumina gel. The numbers of GAD-posltwe neurons at the 4 sites examined in the control monkey were very similar (Table I) and indicated that GAD-positive neurons were not decreased m number at the site of the surgical lesion. Areal &strlbutlon was approximately the same in each of the 4 sites examined for this control monkey (Table II).

DISCUSSION
The results of this study are consistent with many prewous studies that have described the GABAergic neurons and terminals In the cerebral cortex These results also confirm the previously reported decrease of GABAerglc terminals at chronic epileptic foci Most importantly, the present study demonstrates a loss of GABAergic neuronal somata at epileptic foci. These findings are discussed in the following sections

GABAergic neurons in the cerebral cortex
The analysis of the shapes, sizes and distribution of GABAergic neurons in the cerebral cortex as demonstrated with immunocytochemical methods has been the subject of numerous papers In the past few years. We first demonstrated GABAergic axon terminals and a heterogeneous population of GABAerglc somata in the rat visual cortex 19 and determined on the basis of Golgi-electron m~croscoplc data that these neurons were asplnous and sparsely-spmous stellate cells 17. Our subsequent analysis in the monkey sensorimotor cortex 22 showed a similar finding. However, colchicme injections were not used in that analysis to reveal the GABAergic neuronal somata. The distribution of axon terminals and the fact that they formed symmetric synapses provided adequate assurance that these neurons were non-pyramidal cells 22. The location of many terminals adjacent to pyramidal cell bodies indicated that basket cells were a major GABAergic cell type that could provide a powerful inhibition of the projection neurons of the cerebral cortex.
Other studies have confirmed these initial findings for the rat parietal cortex 3, cat visual cortex 4,1027, and monkey visuals and sensorlmotor cortex 9-11. In addition, they have shown that chandelier cells are GABAergic cortical neurons 4A7, 25,26 and that at least 3 peptldes (somatostatm, cholecystokmln and neuropeptide Y) are co-localized to GABA neuronal somata which resemble other cortical types besides basket and chandeher cells 10,27 The most thorough quantitative analysis previously made of the sizes of GABAerglc neuronal somata in monkey motor cortex was by Houser et a1. 11 The mean areas and ranges for GABAerglc cells in our study are remarkably simdar to the ones pubhshed in that study This similarity ~s significant because different antibo&es were utdlzed for these two immunocytochemlcal studies. Houser et al Xl utdlzed an anti-GAD serum z4 produced in rabbit against purified GAD from mouse synaptosomal preparations, whereas our study used an antiserum raised against GAD in sheep 15 One further difference between these two studies was the fact that the present study did not require colchlcme rejections to produce somal staining This &fference may represent an advantage because colch~cme may cause alterauons to the internal structure of neurons (see ref 5 for a review)

GABAergtc termmals at eptlepttc ]ocl
The &stribution of GAD-posmve puncta m the normal cortex and the cortex contralateral to alumina gel apphcatlons was similar to that previously described 3, 9-11A8,192225. 27 These puncta, which have been shown to be axon terminals m electron microscopic preparations, are concentrated along the somata, proximal dendrites and axon initial segments of pyramidal neurons. In addition, they are found adjacent to non-pyramidal somata and scattered m the neuropll The &stnbutlon and number of GAD-posztwe puncta at the epileptic foo are dramatically changed. Although the present quantitative study of somata did not include an analysis of GAD-positwe puncta, the high magmficatlon light mlcrograph ( Fig  3B) of an epdeptlc focus indicates a reduction of these puncta These findings are consistent with our previous quantitative analysis 22

GA BA ergzc neuronal somata at epileptic foct
The results of this study show that monkeys with alumina gel rejections had reductions in the number of GAD-positive neuronal somata at epileptic focl. Only two monkeys demonstrated statistically slgmficant decreases and they were the chronic ones that had seizures for a period of several weeks or more. These findings are consistent with the previously described neuronal loss at alumina gel epileptic focl 6. Alumina gel must play some role in damaging GABAergic neurons because the surgical control did not display any loss of GAD-positive cells. Sloper et al. 25 have shown that terminals with features similar to GABAergic terminals are sensitive to lschemia, and it is possible that alumina gel affects the vascular supply of the cortex to cause an lschemic result that is selective for certain GABAergic neurons 21. The more active GABAergic neurons would be most susceptible to ischemia because they would have a higher demand for oxygen and nutrients. Since terminals that form synapses with pyramidal somata and axon initial segments have more mitochondrla per terminal than other terminals m the cortex 20,21, it is likely that basket and chandelier cells are selectively affected if we assume that the number of mitochondria in a terminal is related to its activity. Nevertheless, since the magnitude of the decrease in GABAergic neurons is 35-50% in the chronic monkeys, we can speculate that a critical number of functional GABAergic neurons (70-75% of total) is required for the prevention of seizures in motor cortex.
The data on the changes in the number of GABAergic neurons at the parafocus are more difficult to interpret. Only the chronic monkeys displayed a decrease in the number of GABAergic somata at this site. This decrease was greater in the monkey with a 6-week record 'of daily seizures (RB-165) than in the animal that had seizures every other day for 3 weeks (RB-163). These data suggest that GABAergic neuronal loss may spread from the focus as the length and severity of seizures increases. Such a notion would be consistent with previous data in human clinical studies which show that seizures beget more seizures7, 30. This finding indicates the importance of prescribing effective anti-epileptic drugs for patients following their initial seizures. Surgical removal of the focus may provide an excellent treatment for focal epilepsy that is not controlled with drugs because the surgery itself does not cause a seizure focus 7. In our study, the size of the control lesion that produced no reduction of GABAergic somata was as large as the alumina gel granuloma that caused a significant decrease of GABAergic neurons. Thus, a gross loss of whole brain is not epileptogenic but a preferential loss of GABAergic neurons in cerebral cortex due to ischemia or some other insult may be the basis for epileptic activity in focal epilepsy.
The analysis of the sizes of GABAergic neurons at the focus and the contralateral cortex which lacks epileptic activity indicates that the small cells are more severely lost at epileptic loci than the larger GABAergic neurons. These data do not imply that large GABAergic cells are not reduced m number. All sizes of GABAergic neurons are decreased at the epileptic loci. However, the data indicate that a greater proportion of smaller cells is lost. Since chandelier cells are small and basket cells are large, these data may indicate that both of these GABAergic cell types are decreased at epileptic loci. Such a prediction was made based on the analysis of the axon terminal loss at epileptic loci in this same model20, 21. Therefore, the data from the present study do not provide an indication of the GABAergic cell types that remain at the epileptic loci. An analysis with a combined Golgi and immunocytochemlcal method 4,26 may help to determine such reformation.
In conclusion, the results of this study add further data to demonstrate a GABAergic deficit at sites of focal epilepsy in the cerebral cortex. Also, they demonstrate that a loss of GABAergic somata is associated with the preferential loss of GABAergic terminals at epileptic loci.