Decrease in Inhibition in Dentate GranuleCells from Patients with Medial

Temporal Lobe EpilepsyAnne Williamson, PhD, Peter R. Patrylo, PhD, and Dennis D. Spencer, MD

Alterations in synaptic inhibition are associated with epileptiform activity in several acute animal models; however, it isnot clear if there are changes in inhibition in chronically epileptic tissue. We have used intracellular recordings fromgranule cells of patients with temporal lobe epilepsy to determine whether synaptic inhibition is compromised. Twogroups of patients with medial temporal lobe epilepsy were used, those with medial temporal lobe sclerosis (MTLE), andthose with extrahippocampal masses (MaTLE) where the cell loss and synaptic reorganization that characterize MTLE arenot seen. Although the level of tonic inhibition at the somata was not significantly different in the two patient groups,there was a reduction in the conductance of polysynaptic perforant path–evoked fast and slow inhibitory postsynapticpotentials (IPSPs) (53% and 66%, respectively). We found that there was a comparable decrease in the monosynapticIPSP conductances examined in the presence of glutamatergic antagonists as that seen for the polysynaptically evokedIPSPs. These data suggest that the decrease in inhibition seen in normal artificial cerebrospinal fluid in MTLE granulecells cannot be solely explained by a decrease in excitatory input onto inhibitory interneurons and may reflect changesat the interneuron–granule cells synapse or in the number of specific inhibitory interneurons.

Williamson A, Patrylo PR, Spencer DD. Decrease in inhibition in dentate granule cells from patientswith medial temporal lobe epilepsy. Ann Neurol 1999;45:92–99

Medial temporal lobe sclerosis (MTLE) is one of themost common forms of medically intractable temporallobe epilepsy and can often be successfully treated withtemporal lobectomy. This pathology is characterized byextensive cell loss in the pyramidal cell layers of theCA1 and CA3 regions as well as in the hilar region.In addition, there is extensive synaptic reorganizationof both the granule cell axons (mossy fibers) andof several peptidergic systems.1–4 Although the g-aminobutyric acid (GABA)ergic basket cells in the den-tate appear relatively well preserved in this tissue,5

many GABAergic cells with somata in the hilus are lostin MTLE.2

Epileptic patients with extrahippocampal temporallobe masses (MaTLE; primarily tumors) do not showthis characteristic pattern of cell loss and synaptic re-organization.1,6,7 Based on these differences, and as notrue controls exist for the human material, we haveused the MaTLE group as a comparison population forthe MTLE group. This comparison allows us to ad-dress the consequences of synaptic reorganization andcell loss independent of the effects of chronic temporallobe seizures and exposure to antiepileptic drugs.

There is a controversy as to whether a decrease in

GABAergic inhibition does contribute, in part, to theepileptogenesis seen in MTLE and in animal models ofMTLE. This hypothesis stems from the observationsthat blocking inhibition can produce epileptiform ac-tivity in the hippocampus of normal control rodents.Furthermore, studies in animal models of MTLE haveshown that there is a decrease in inhibition within thehippocampus,8,9 although whether this decrease is re-gion specific is, as yet, unknown. However, in severalchronic animal models there is an increase in inhibi-tion.10–12 Finally, many of the anticonvulsants cur-rently in use enhance GABAergic inhibition.13,14

Previous slice studies on the hippocampi of MTLEpatients have shown that the dentate granule cells arehyperexcitable in response to perforant path stimula-tion.15–17 This hyperexcitability takes the form of threeor more action potentials triggered by single stimulidelivered to the outer molecular layer. Possible mech-anisms for this hyperexcitability include alterations inglutamate receptors,18,19 reorganization of excitatory fi-ber systems including mossy fiber sprouting,20–23 andalterations in inhibitory drive. However, in in vivostudies in epileptic patients, extracellular recordingshave provided data suggesting that inhibition remains

From the Department of Neurosurgery, Yale University School ofMedicine, New Haven, CT.

Received May 12, 1998, and in revised form Jul 16 and Aug 7.Accepted for publication Aug 7, 1998.

Address correspondence to Dr Williamson, Department of Neuro-surgery, Yale University School of Medicine, PO Box 208082, NewHaven, CT 06520-8082.

92 Copyright © 1999 by the American Neurological Association

intact or is even enhanced in epileptogenic regions ofthe hippocampus.24 Thus, there is conflicting data onwhether a decrease in inhibition could be involved inthe hyperexcitability seen in tissue from patients withMTLE.

Minor changes in inhibitory strength may not be re-solved in in vivo experiments using extracellular re-cording techniques. Therefore, we have examined syn-aptic inhibition in tissue from patients with MTLEand MaTLE by using intracellular recordings to deter-mine whether inhibition is compromised in reorga-nized human tissue and the possible sites where anyobserved decrease could occur.

Patients and MethodsPatient DataThese data are from a total of 91 granule cells from 35MTLE hippocampi and 59 granule cells from 21 MaTLEhippocampi. The MTLE patients were classified based onquantitative cell counts of the resected hippocampus showinga greater than 2 SD reduction in neuron number in the CA1and CA3 pyramidal cell regions and the dentate hilus relativeto autopsy controls. In addition, immunocytochemical stud-ies showed that there was synaptic reorganization.1,4,21 TheMaTLE patients had a wide variety of extrahippocampalmasses, which included astrocytomas (11 patients), arterio-venous malformations (3 patients), gangliogliomas (3 pa-tients), oligodendroglioma (1 patient), gliosis (1 patient),harmartoma (1 patient), and a cyst (1 patient). Quantitativecell counts show a mild (,20%) cell loss in all cell fields inthese patients in contrast to the extensive cell loss seen inspecific regions of MTLE hippocampi.3,6 In addition, immu-nocytochemical studies show that there is no evidence forsynaptic reorganization in this patient group. Patients withmasses that extended into the pes hippocampi were excludedfrom this study. All tissue was resected for clinical reasonsand these experiments were approved by the Yale UniversityHuman Investigation Committee.

Tissue PreparationThe methods for tissue preparation have been previously de-scribed.25 In brief, the hippocampi were resected en bloc anda 5-mm slab was prepared in the operating room. The tissuewas kept in ice-cold, oxygenated artificial cerebrospinal fluid(ACSF) and transported to the laboratory where 400-mmslices were prepared on a Vibratome (Ted Pella, Redding,CA). The slices were maintained at 34.0°C in an interfaceslice chamber (Fine Science Tools, Foster City, CA) andwere perfused with oxygenated ACSF containing (in mM)NaCl 124, KCl 3.5, MgSO4 2, NaH2PO4 1.2, NaHCO3 26,CaCl2 2.0, and dextrose 10, and was maintained at pH 7.4. Insome experiments, 5-aminophosphonovalerate (APV) and6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) (50 mM each)were bath applied. APV and CNQX were obtained from ei-ther Sigma (St Louis, MO) or from RBI (Natick, MA).

Intracellular recordings were performed by using micro-electrodes formed on a Brown-Flaming electrode puller (Sut-ter Instruments, Novato, CA). The microelectrodes wereusually filled with 4 M potassium acetate. In some experi-

ments, the electrodes were filled with KCl. Under these re-cording conditions, the Cl2 gradient is reversed, making itpossible to observe spontaneous inhibitory postsynaptic po-tentials (IPSPs).

Polysynaptic circuits were activated in normal ACSF, us-ing orthodromic stimuli delivered to the outer two-thirds ofthe molecular layer by using either monopolar or twisted bi-polar stimulating electrodes. We were able to evoke excita-tory postsynaptic potentials (EPSPs), which could trigger ac-tion potentials in all cells included here, indicating thatperforant path fibers contacted the cells being studied. Theelectrode was placed at a distance from the recording site toavoid direct activation of the recorded cell. We examinedfunctional inhibition by assessing whether perforant path-evoked inhibitory responses could slow the rate of action po-tential firing produced by a depolarizing current step.26 Inthese studies, the stimulus intensity used was at least twiceaction potential threshold to ensure that both feedforwardand feedback inhibition was activated.

Monosynaptic IPSPs were evoked in ACSF containingAPV and CNQX (50 mM each) by placing the stimulatingelectrode in the inner third of the molecular layer. This stim-ulation protocol allows for the direct activation of inhibitorycells and thus the examination of inhibitory events indepen-dent of excitatory input onto the interneurons. The stimulusintensity was adjusted to elicit the maximal amplitude mono-synaptic IPSP; often several electrode placements were exam-ined to ensure that the maximal amplitude was achieved.Stimulus intensities ranged from 10 to 800 mA with a pulseduration of 300 msec for both types of stimulation.

The location of the stimulating electrode may be an im-portant variable when making comparisons between mono-synaptically and polysynaptically evoked IPSPs, as several re-ports indicate that there may be dendritic retraction inMTLE hippocampi.27,28 In an attempt to control for thisvariable, we have limited our results and discussion on IPSPsprimarily to data gathered from those granule cells in whichpolysynaptic stimulation produced an IPSP.

Data AnalysisThe data on membrane properties (membrane potential, in-put resistance, time constant, and action potential ampli-tude) for all the cells from a given patient were averaged toobtain representative values for each patient. The frequencyof spontaneous IPSPs recorded with KCl-filled electrodes wasassessed over a 1-minute period; events were counted if theyhad amplitudes larger than twice the baseline noise level. Theamplitudes of these events were not quantified because of thevariability associated with the degree to which the Cl2 gra-dient was shifted. IPSP conductances were determined by us-ing methods previously described.29,30

Data are presented as mean 6 SEM values. Statistical sig-nificance was determined by using a two-tailed Student’s ttest or an analysis of variance and was set at p , 0.05.

ResultsMembrane Properties of Dentate Granule Cellsfrom Human HippocampiNo differences were observed in the resting membranepotential, input resistance, time constant, and action

Williamson et al: Inhibition in Human TLE 93

potential amplitude in MTLE versus MaTLE dentategyri as shown in Table 1. Therefore, differences in syn-aptic properties between the MTLE and MaTLE tissueare not likely to be due to differences in the membraneproperties of the cells being studied.

Decreased Polysynaptic, But Not Tonic, Inhibition inthe MTLE Dentate GyrusWe examined the degree of tonic inhibition onto gran-ule cells in both MTLE and MaTLE tissue by usingKCl-filled electrodes in normal recording medium. Wefound that there was a high frequency of spontaneousactivity in all cells studied from MTLE hippocampi. In10 cells from six hippocampi, we found a postsynapticpotential frequency of 4.2 6 0.4 Hz. This was verysimilar to the level of spontaneous synaptic eventsfound in the MaTLE tissue (5.0 6 0.7 Hz; 7 cells, 4patients). These values were not significantly different.The amplitudes of these events were also not signifi-cantly different (MTLE 5 4.1 6 0.69 mV; MaTLE 54.7 6 1.2 mV). Although there was no apparent dif-ference in the amplitudes of these events, this variableis difficult to interpret as it depends on the degree towhich the Cl2 gradient is shifted by diffusion of KClinto the recorded neuron. The events recorded underthese conditions are predominantly IPSPs, because thefrequency and amplitude of these events was dramati-cally reduced in the presence of bicuculline (Fig 1).Moreover, under control recording conditions usingpotassium acetate electrodes in 12 cells from these sixMTLE hippocampi, we noted an average frequency ofspontaneous excitatory events of only 0.85 6 0.4 Hz.There was no evidence for prominent spontaneous ex-citatory activity in cells from the MaTLE hippocampi.Taken together, these data suggest that there is nogross change in tonic inhibitory synaptic input ontothese cells as assays at the level of the somata.

We also examined the ability of perforant path stim-ulation to evoke IPSPs in the two types of tissue. Weused a range of stimulus intensities so that responsesabove and below the action potential threshold couldbe examined. The response to both subthreshold andsuperthreshold stimulation at membrane potentialsboth depolarized and hyperpolarized to the IPSP rever-sal potential were examined to ensure that there was a

sufficient driving force to allow us to observe anyGABAA receptor–mediated potentials.

We were able to evoke an EPSP–IPSP sequence inonly 7.5% of the cells examined (4 of 53) from theMTLE tissue and in 42.8% of cells from the MaTLEtissue (15 of 35 cells). When stimulation intensitiesthat evoked at least one action potential were used,clear biphasic IPSPs could be evoked in only 19.1% ofthe cells from 15 MTLE hippocampi as shown in Fig-ure 2B (18 of 91 cells). We could not evoke IPSPs inthe remaining 20 hippocampi. By contrast, in granulecells from MaTLE hippocampi we were able to evokebiphasic IPSPs in 83% of the cells studied from 19hippocampi (49 of 59 cells) as shown in Figure 2A.

In those MTLE cells in which IPSPs could not beobserved, we were also unable to observe functional in-hibition (n 5 6). However, the rate of spike firing wasdecreased in all MaTLE cells (n 5 5). In addition, wewere unable to “unmask” IPSPs in MTLE granule cellswhen the late component of the EPSP was blockedby APV, suggesting that late glutamatergic excita-tory inputs do not shunt inhibition in MTLE tissue(n 5 10).

There was no significant difference in the reversalpotentials between IPSPs in the two different types oftissue (fast IPSP, 263.6 6 3.3 and 271.9 6 1.6; slowIPSP, 294.9 6 1.2 and 292.8 6 0.1, for the MTLEand MaTLE tissue, respectively). However, when wecompared the conductances of both the fast and theslow polysynaptically evoked IPSPs in those cells inwhich an IPSP could be evoked, we found that therewas a significant difference in the conductance of thefast and slow IPSPs between the MaTLE and MTLEtissue as shown in Table 2. This difference was greaterfor the slow than the fast IPSP. For these studies, thestimulus intensity that produced the maximal ampli-tude fast IPSP was used. Therefore, the strength of theinhibitory tone appeared to be compromised even inthose cells where a biphasic IPSP could be evoked.

Monosynaptic IPSPs Also Exhibit a ReducedConductance in MTLE DentatePolysynaptic activity from newly formed recurrent fi-bers could also shunt any inhibitory input. In addition,disconnection of inhibitory interneurons from their ex-

Table 1. Membrane Properties of MTLE and MaTLE Granule Cells

MembranePotential (mV)

Input Resistance(MV)

Time Constant(msec)

Action PotentialAmplitude (mV)

MaTLE (n 5 21) 268.7 6 1.3 41.6 6 2.9 7.9 6 0.5 97.4 6 1.7MTLE (n 5 35) 269.2 6 1.2 39.1 6 2.6 8.4 6 0.72 89.9 6 1.8

There were no significant differences in the cellular membrane properties in these two patient populations. Data are mean 6 SEM values; n 5number of patients.

MTLE 5 medial temporal lobe sclerosis; MaTLE 5 medial temporal lobe epilepsy with extrahippocampal masses.

94 Annals of Neurology Vol 45 No 1 January 1999

citatory inputs could also cause the observed decreasein polysynaptic IPSP conductance. To test these hy-potheses, we recorded monosynaptic IPSPs in granulecells from both MTLE and MaTLE tissue in ACSFcontaining the glutamate receptor antagonists APV andCNQX. Examining monosynaptically evoked IPSPs al-lowed us to study the interneuron–granule cell synapse

in isolation and, therefore, to test the hypothesis thatthe decrease in polysynaptic IPSP conductance is pri-marily due to an alteration at the inhibitory interneu-ron–granule cell synapse.

We were able to evoke monosynaptic IPSPs in all ofthe cells studied from both types of tissue, includingtissue in which we were unable to evoke polysynaptic

Fig 2. Decrease in evoked inhibitory postsyn-aptic potentials (IPSPs) in medial temporallobe sclerosis (MTLE) dentate granule cells.Examples of perforant path–evoked responsesfrom MaTLE (A) and MTLE (B) dentategranule cells. Even at depolarized membranepotentials, only a very small IPSP could beevoked in the MTLE cell, but that a bipha-sic IPSP could be generated in the cell froman MaTLE hippocampus. In most cells fromMTLE hippocampi, no IPSP could beobserved.

Fig 1. Spontaneous inhibitory postsynaptic potentials (IPSPs) in medial temporal lobe sclerosis tissue. This figure shows examples ofspontaneous IPSPs recorded by using a KCl-filled microelectrode. These events were reversibly blocked after the application of bicu-culline (middle). The cell was held at 275 mV. A similar frequency of spontaneous depolarizing events was seen in granule cellsfrom hippocampi of epileptic patients with extrahippocampal temporal lobe masses.

Williamson et al: Inhibition in Human TLE 95

IPSPs in control ACSF. As mentioned in Patients andMethods, the stimulus intensity and stimulating elec-trode placement were adjusted to produce a maximalmonosynaptic IPSP. It was noteworthy that the fastIPSP could be evoked reliably in all cells studied fromboth groups at low stimulus intensities whereas theslow IPSP could only be seen at higher stimulus inten-sities or at depolarized membrane potentials. There wasno difference in the apparent reversal potential of thefast and slow monosynaptic IPSPs compared withpolysynaptically evoked responses. However, as withthe polysynaptic IPSPs, we noted a significant decreasein the IPSP conductances in the MTLE tissue relativeto MaTLE cells (Fig 3; see Table 2), suggesting thatthe decrease in inhibition is at least in part due to achange at the level of the interneuron–granule cell syn-apse. Across the population studied, we noted no sig-nificant differences in the rise or fall times of the fastcomponent of the monosynaptic IPSP between the twogroups, although there was a trend toward longer falltimes for the IPSPs studied from MaTLE tissue. Ex-amples of these responses are shown in Figure 3.

DiscussionThe primary findings of this study are that tonic inhi-bition appears to be preserved in tissue from MTLEpatients, but that there is a decreased conductance ofboth fast and slow IPSPs evoked either polysynapticallyor monosynaptically relative to MaTLE tissue. Thesefindings are in line with prior studies on tissue fromMTLE patients where it was reported that inhibitionappears to be relatively intact, but that there maybe subtle differences in the efficiency of the inhibitorysystem.15,31

It is noteworthy that small changes in the strength ofinhibition may critically affect the overall output of asystem. The study by Chagnac-Amitai and Connors32

indicates that modest compromises in the strength ofGABAA-mediated inhibition can dramatically alter theability of neocortical circuits to initiate and spread ep-ileptiform activity, due presumably to the presence ofextensive excitatory recurrent connections. Althoughthis type of recurrent circuitry is not found in the den-tate in control tissue, it is a well-documented feature ofthe dentate gyrus in MTLE tissue.20,21 Thus, small

Fig 3. Monosynaptically evoked inhibition is reduced in medial temporal lobe sclerosis (MTLE) granule cells. This figure shows ex-amples of monosynaptically inhibitory postsynaptic potentials (IPSPs) evoked in the presence of 5-aminophosphonovalerate (APV) and6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) (50 mM each). These data are plotted on the right. There was no difference in thereversal potentials of either the fast or the slow component of the responses; however, the conductance of the fast IPSP was signifi-cantly smaller in the MTLE tissue relative to that from hippocampi of epileptic patients with extrahippocampal temporal lobemasses (MaTLE). The stimulus intensity was set to produce the largest amplitude response without evoking an antidromic spike.

Table 2. Polysynaptic and Monosynaptic IPSP Conductances Are Reduced in MTLE Dentate Granule Cells

MTLE MaTLE p

G fast polysynaptic IPSP (nS) 17.29 6 1.3 37.4 6 4.1 0.03G slow polysynaptic IPSP (nS) 4.73 6 0.4 14.3 6 1.9 0.008G fast monosynaptic IPSP (nS) 15.1 6 2.7 38.9 6 12.1 0.01G slow monosynaptic IPSP (nS) 2.8 6 2.0 6.1 6 0.5 0.03

The conductances for both the fast and slow IPSPs were significantly smaller in granule cells from MTLE hippocampi compared with thosefrom MaTLE tissue for both polysynaptic and monosynaptic IPSPs. The magnitude of the difference is similar to that seen for both types ofIPSPs. The polysynaptic data are from 8 MTLE cells and 12 MaTLE cells; the monosynaptic data are from 5 MTLE cells and 10 MaTLE cells.

IPSP 5 inhibitory postsynaptic potential; MTLE 5 medial temporal lobe sclerosis; MaTLE 5 medial temporal lobe epilepsy with hippocampalmasses; G 5 conductance; nS 5 nanosiemens.

96 Annals of Neurology Vol 45 No 1 January 1999

changes in the strength of inhibition in the epileptichuman dentate may allow for the initiation and spreadof seizure activity through a reorganized synaptic cir-cuit. Although we cannot precisely determine themechanism that produces this decrease in inhibitorystrength, several possibilities are described below.

The first point to consider is that we observed a sim-ilar decrease in the conductance of polysynaptically andmonosynaptically evoked IPSPs. Therefore, it is un-likely that the compromise in the conductance ofpolysynaptically evoked IPSPs observed in MTLEgranule cells can be solely explained by either (1) al-terations in excitatory input, specifically enhancedN-methyl-D-aspartate receptor–mediated events15,16,33

and/or newly formed recurrent excitatory inputs mask-ing inhibitory responses, or (2) glutamatergic deaffer-entation of interneurons.9,34 However, these mecha-nisms may, in part, play a role in the increasedincidence of observing monosynaptically versus poly-synaptically evoked IPSPs. We cannot rule out the pos-sibility that the IPSP is in part shunted by some otherconductance that was not blocked by the addition ofAPV and CNQX. This seems unlikely, however, asmany of the neuromodulators present in the hip-pocampus do not appear to be released at the low stim-ulus frequencies used here (0.2–0.1 Hz).

One potential hypothesis to explain the parallel re-duction of polysynaptically and monosynapticallyevoked IPSPS that we observed in MTLE tissue is thatthere is a loss of specific populations of interneurons.Although the density of glutamate decarboxylase–immunoreactive neurons at the granule cell-hilar bor-der is not significantly different in MaTLE and MTLEtissue,5 up to 90% of the hilar cells can be lost inMTLE hippocampi. Many of these cells colocalizeGABA and peptides and may, therefore, subserve aninhibitory function.35–37 The loss of a specific popula-tion of hilar interneurons, many of which receive bothgranule cell and perforant path input,35 could explainour data, especially if the vulnerable class of interneu-ron has a restricted axonal input onto the apical den-drites of granule cells. In this scenario, one would pre-dict that the level of spontaneous IPSPs recorded at thegranule cell soma would not be significantly differentbetween the MaTLE and MTLE groups as the basketcells that provide strong input to the granule cell layerappear intact in both groups.3,5 However, the responseto evoked stimulation would be reduced in the MTLEtissue, because there would be fewer cells activated by agiven input. Support for this hypothesis comes from arecent study in the CA1 region of both epileptickainate-treated and pilocarpine-treated rats.8,38 In thisstudy, as in ours, the frequency of spontaneous inhib-itory events recorded at the soma was comparable be-tween the control and epileptic tissue, yet there was a

decrease in the frequency of spontaneous IPSPs mea-sured in the dendrites of the epileptic tissue.8

An alternate hypothesis is that inhibition is affectedat the level of the inhibitory interneuron–granule cellsynapse. An alteration at this level could involve severalspecific sites, including (1) alterations in membraneproperties of a subset of interneurons38,39; (2) changesin the GABA receptor itself 40–42; (3) modulation ofthe GABA receptor by N-methyl-D-aspartate recep-tor–mediated Ca21 entry43–45; (4) modulation ofthe GABA receptor by neuromodulators includingzinc11,46,47; (5) reduction in the amount of releasableGABA48,49; and (6) presynaptic modulation of GABArelease.50,51 Although we cannot directly address thelikelihood of these different mechanisms, we have pre-viously shown that there is a decrease in GABA uptakein the dentate gyrus of MTLE patients.25,52 This alter-ation would result in an enhanced level of GABA inthe extracellular space in MTLE tissue after periods ofsynaptic activity. One possible consequence of this de-crease in GABA uptake could be enhanced bindingonto GABAB receptors located on interneuronal axons,which would reduce any additional evoked GABA re-lease.50,51 Data from Otis and Mody53 suggest thatGABAB agonists only affect the size of evoked IPSCsbut do not affect the conductance of spontaneousevents. Thus, this mechanism could produce the reduc-tion of evoked IPSPs we observed in cells from MTLEhippocampi in this study without a concomitantchange in the frequency or amplitude of spontaneousIPSPs between MaTLE and MTLE tissue.

In conclusion, although we cannot rule out the pos-sibility that alterations in the excitatory drive onto dif-ferent populations of interneurons compromise thestrength of inhibition in MTLE granule cells, our dataindicate that this cannot be the sole mechanism in-volved. Our data suggest that the reduction in inhibi-tion in MTLE hippocampi may, in part, occur at thelevel of the interneuron–granule cell synapse.

This study was supported by NIH grants NS06208 and NS30012to Dr Williamson.

We thank the patients, without whose consent these experimentscould not have been performed.

References1. de Lanerolle NC, Brines ML, Williamson A, et al. Neurotrans-

mitters and their receptors in human temporal lobe epilepsy. In:Ribak Gall CE, Mody I, eds. The dentate gyrus and its role inseizures. The Netherlands: Elsevier, 1992:235–250

2. de Lanerolle NC, Kim JH, Robbins RJ, Spencer DD. Hip-pocampal interneuron loss and plasticity in human temporallobe epilepsy. Brain Res 1989;495:387–395

3. Mathern G, Babb T, Leite J, et al. The pathogenic and pro-gressive features of chronic human hippocampal epilepsy. Epi-lepsy Res 1996;26:151–161

4. Mathern GW, Babb TL, Pretorius JK, Leite JP. Reactive syn-

Williamson et al: Inhibition in Human TLE 97

aptogenesis and neuronl densities for neuropeptide Y, soma-tostatin and glutamate decarboxylase immunoreactivity in theepileptogenic human fascia dentata. J Neurosci 1995;15:3990–4004

5. Babb TL, Pretorius JK, Kupfer WR, Crandall PH. Glutamatedecarboxylase-immunoreactive neurons are preserved in humanepileptic hippocampus. J Neurosci 1989;9:2562–2574

6. Kim JH, Guimaraes PO, Shen M-Y, et al. Hippocampal neu-ronal density in temporal lobe epilepsy with and without glio-mas. Acta Neuropathol 1990;80:41–45

7. de Lanerolle NC, Spencer DD. Neurotransmitter markers inhuman seizure foci. In: Fisher S, Coyle JT, eds. Neurotransmit-ters and epilepsy. New York: Wiley-Liss, 1991:201–217

8. Bernard C, Esclapez M, Agid F, et al. Selective loss of GABAer-gic inhibition in the apical dendrites of CA1 pyramidal cells intemporal lobe epilepsy. Soc Neurosci Abstr 1997;23:2157(Abstract)

9. Bekenstein JW, Lothman EW. Dormancy of inhibitory inter-neurons in a model of temporal lobe epilepsy. Science 1993;359:97–100

10. Otis TS, De KY, Mody I. Lasting potentiation of inhibition isassociated with an increased number of gamma-aminobutyricacid type A receptors activated during miniature inhibitorypostsynaptic currents. Proc Natl Acad Sci USA 1994;91:7698–7702

11. Gibbs JW 3rd, Shumate MD, Coulter DA. Differentialepilepsy-associated alterations in postsynaptic GABA(A) recep-tor function in dentate granule and CA1 neurons. J Neuro-physiol 1997;77:1924–1938

12. Buckmaster PS, Dudek FE. Neuron loss, granule cell axon re-organization, and functional changes in the dentate gyrus ofepileptic kainate-treated rats. J Comp Neurol 1997;385:385–404

13. Meldrum BS. Update on the mechanism of action of antiepi-leptic drugs. Epilepsy Res Suppl 1996;11:67–77

14. Macdonald RL, Kelly KM, Antiepileptic drug mechanisms ofaction. Epilepsia 1995;36(Suppl 2):S2–S12

15. Isokawa M, Levesque M. Increased NMDA responses and den-dritic degeneration in human epileptic hippocampal neurons inslices. Neurosci Lett 1991;132:212–216

16. Masukawa LM, Higashima M, Hart GJ, O’Connor MJ.NMDA receptor activation during epileptiform responses in thedentate gyrus of epileptic patients. Brain Res 1991;562:176–180

17. Williamson A. Electrophysiology of epileptic human neocorticaland hippocampal neurons maintained in vitro. Clin Neurosci1994;2:47–52

18. Mody I, Reynolds JN, Salter MW, et al. Kindling-induced ep-ilepsy alters calcium currents in granule cells of rat hippocampalslices. Brain Res 1990;531:88–94

19. Kohr G, Lambert CE, Mody I. Calbindin-D28K (CaBP) levelsand calcium currents in acutely dissociated epileptic neurons.Exp Brain Res 1991;85:543–551

20. Sutula T, Cascino G, Cavazos J, et al. Mossy fiber synapticreorganization in the epileptic human temporal lobe. Ann Neu-rol 1989;26:321–330

21. Houser CR, Miyashiro JE, Swartz BE, et al. Altered patterns ofdynorphin immunoreactivity suggest mossy fiber reorganizationin human hippocampal epilepsy. J Neurosci 1990;10:267–282

22. Wuarin J-P, Dudek FE. Electrographic seizures and new recur-rent excitatory circuits in the dentate gyrus of hippocampalslices from kainate-treated epileptic rats. J Neurosci 1996;16:4438–4448

23. Patrylo PR, Dudek FE. Physiological unmasking of new gluta-matergic pathways in the dentate gyrus of hippocampal slicesfrom kainate-induced epileptic rats. J Neurophysiol 1998;79:418–429

24. Isokawa AM, Wilson CL, Babb TL. Inhibition in synchro-nously firing human hippocampal neurons. Epilepsy Res 1989;3:236–247

25. Williamson A, Telfeian AE, Spencer DD. Prolonged GABA re-sponses in dentate granule cells in slices isolated from patientswith temporal lobe sclerosis. J Neurophysiol 1995;74:378–387

26. McCormick DA. GABA as an inhibitory neurotransmitter inhuman cerebral cortex. J Neurophysiol 1989;62:1018–1027

27. Isokawa M. Preservation of dendrites with the presence of re-organized mossy fiber collaterals in hippocampal dentate gran-ule cells in patients with temporal lobe epilepsy. Brain Res1997;744:339–343

28. von Campe G, Spencer D, de Lanerolle N. Morphology ofdentate granule cells in the human epileptogenic hippocampus.Hippocampus 1997;7:472–488

29. McCarren M, Alger BE. Use-dependent depression of IPSPs inrat hippocampal pyramidal cells in vitro. J Neurophysiol 1985;53:557–571

30. Williams S, Vachon P, Lacaille JC. Monosynaptic GABA-mediated inhibitory postsynaptic potentials in CA1 pyramidalcells of hyperexcitable hippocampal slices from kainic acid-treated rats. Neuroscience 1993;52:541–554

31. Isokawa M. Decrement of GABAA receptor-mediated inhibi-tory postsynaptic currents in dentate granule cells in epileptichippocampus. J Neurophysiol 1996;75:1901–1908

32. Chagnac AY, Connors BW. Horizontal spread of synchronizedactivity in neocortex and its control by GABA-mediated inhi-bition. J Neurophysiol 1989;61:747–758

33. Mody I, Stanton PK, Heinemann U. Activation of N-methyl-D-aspartate receptors parallels changes in cellular and synapticproperties of dentate gyrus granule cells after kindling. J Neu-rophysiol 1988;59:1033–1054

34. Sloviter RS. Permanently altered hippocampal structure, excit-ability, and inhibition after experimental status epilepticus inthe rat: the “dormant basket cell” hypothesis and its possiblerelevance to temporal lobe epilepsy. Hippocampus 1991;1:41–66

35. Freund T, Buzsaki G. Interneurons of the hippocampus. Hip-pocampus 1996;6:347–470

36. Esclapez M, Houser CR. Somatostatin neurons are a subpopu-lation of GABA neurons in the rat dentate gyrus: evidence fromcolocalization of pre-prosomatostatin and glutamate decarbox-ylase messenger RNAs. Neuroscience 1995;64:339–355

37. Houser CR, Esclapez M. Vulnerability and plasticity of theGABA system in the pilocarpine model of spontaneous recur-rent seizures. Epilepsy Res 1996;26:207–218

38. Esclapez M, Hirsch JC, Khazipov R, et al. Operative GABAer-gic inhibition in hippocampal CA1 pyramidal neurons in ex-perimental epilepsy. Proc Natl Acad Sci USA 1997;94:12151–12156

39. Rempe DA, Bertram EH, Williamson JM, Lothman EW. In-terneurons in area CA1 stratum radiatum and stratum oriensremain functionally connected to excitatory synaptic input inchronically epileptic animals. J Neurophysiol 1997;78:1504–1515

40. Rice A, Rafiq A, Shapiro SM, et al. Long-lasting reduction ofinhibitory function and gamma-aminobutyric acid type A re-ceptor subunit mRNA expression in a model of temporal lobeepilepsy. Proc Natl Acad Sci USA 1996;93:9665–9669

41. Kapur J, Coulter DA. Experimental status epilepticus altersg-aminobutyric acid type A receptor function in CA1 pyrami-dal neurons. Ann Neurol 1995;38:893–900

42. Gibbs JW 3rd, Shumate MD, Coulter DA. Differentialepilepsy-associated alterations in postsynaptic GABA(A) recep-tor function in dentate granule and CA1 neurons. J Neuro-physiol 1997;77:1924–1938

43. Chen QX, Wong RK. Suppression of GABAA receptor re-

98 Annals of Neurology Vol 45 No 1 January 1999

sponses by NMDA application in hippocampal neuronesacutely isolated from the adult guinea-pig. J Physiol (Lond)1995;482:353–362

44. Stelzer A, Slater NT, Ten BG. Activation of NMDA receptorsblocks GABAergic inhibition in an in vitro model of epilepsy.Nature 1987;326:698–701

45. Isokawa M, Levesque M, Fried I, Engel JJ. Glutamate currentsin morphologically identified human dentate granule cells intemporal lobe epilepsy. J Neurophysiol 1997;77:3355–3369

46. Xie X, Smart TG. Properties of GABA-mediated synaptic po-tentials induced by zinc in adult rat hippocampal pyramidalneurones. J Physiol (Lond) 1993;460:503–523

47. Buhl E, Otis T, Mody I. Zinc-induced collapse of augmentedinhibition by GABA in a temporal lobe epilepsy model. Science1996;271:369–373

48. Petroff OA, Rothman DL, Behar KL, Mattson RH. Low brainGABA level is associated with poor seizure control. Ann Neurol1996;40:908–911

49. Lothman EW, Bennett JP, Perlin JB. Alterations in neurotrans-mitter amino acids in hippocampal kindled seizures. EpilepsyRes 1987;1:313–320

50. Mott DD, Xie C-W, Wilson WA, et al. GABAB autoreceptorsmediate activity-dependent disinhibition and enhance signaltransmission in the dentate gyrus. J Neurophysiol 1993;69:674–691

51. Davies CH, Davies SN, Collingridge GL. Paired pulse depres-sion of monosynaptic GABA-mediated inhibitory postsynapticpotentials in rat hippocampus. J Physiol (Lond) 1990;424:513–531

52. During MJ, Ryder KM, Spencer DD. Hippocampal GABAtransporter function in temporal-lobe epilepsy. Nature 1995;376:174–177 (Comment)

53. Otis TS, Mody I. Differential activation of GABAA andGABAB receptors by spontaneously released transmitter. J Neu-rophysiol 1992;67:227–234

Williamson et al: Inhibition in Human TLE 99