Reticular nucleus-specific changes in α3 subunit protein at GABA synapses in genetically epilepsy-prone rats

Abstract

Differential composition of GABA_A receptor (GABA_AR) subunits underlies the variability of fast inhibitory synaptic transmission; alteration of specific GABA_AR subunits in localized brain regions may contribute to abnormal brain states such as absence epilepsy. We combined immunocytochemistry and high-resolution ImmunoGold electron microscopy to study cellular and subcellular localization of GABA_AR α1, α3, and β2/β3 subunits in ventral posterior nucleus (VP) and reticular nucleus (RTN) of control rats and WAG/Rij rats, a genetic model of absence epilepsy. In control rats, α1 subunits were prominent at inhibitory synapses in VP and much less prominent in RTN; in contrast, the α3 subunit was highly evident at inhibitory synapses in RTN. β2/β3 subunits were evenly distributed at inhibitory synapses in both VP and RTN. ImmunoGold particles representing all subunits were concentrated at postsynaptic densities with no extrasynaptic localization. Calculated mean number of particles for α1 subunit per postsynaptic density in nonepileptic VP was 6.1 ± 3.7, for α3 subunit in RTN it was 6.6 ± 3.4, and for β2/β3 subunits in VP and RTN the mean numbers were 3.7 ± 1.3 and 3.5 ± 1.2, respectively. In WAG/Rij rats, there was a specific loss of α3 subunit immunoreactivity at inhibitory synapses in RTN, without reduction in α3 subunit mRNA or significant change in immunostaining for other markers of RTN cell identity such as GABA or parvalbumin. α3 immunostaining in cortex was unchanged. Subtle, localized changes in GABA_AR expression acting at highly specific points in the interconnected thalamocortical network lie at the heart of idiopathic generalized epilepsy.

The idiopathic generalized epilepsies are characterized by abrupt losses of consciousness during which the electroencephalogram (EEG) exhibits paroxysmal, high-amplitude spike and wave complexes at ≈3Hz and lasting from a few seconds to <1 min. The loss of consciousness is referred to as an absence seizure or petit mal. Spike and wave activity reflects paroxysmal discharging of neurons in the network of reentrant thalamocortical and corticothalamic connections (1–3). Linkage studies in humans suggest involvement of genes encoding GABA_A receptor (GABA_AR) subunits of the α, β, γ, and δ families in absence epilepsy (4–7), and experimental studies in animals point to molecular genetic abnormalities in GABA_AR signaling (8, 9). The forms by which genetic anomalies in GABA_AR manifest themselves are multiple and varied. They can involve receptor synthesis and trafficking within neurons, changes in ratios of alternatively spliced mRNAs, and translocations of receptor subunits to membrane sites not normally occupied (10–14).

The GABAergic neurons of the thalamic reticular nucleus (RTN) play a key role in synchronizing activity in the thalamocortical network during states of consciousness (15, 16), and GABA_AR antagonists applied to its cells can transform 7- to 14-Hz sleep spindle oscillations generated in the RTN into paroxysmal low-frequency oscillations resembling spike and wave activity (17, 18). Mice null for the gene encoding the β3 GABA_AR subunit, one of the principal GABA_AR subunits expressed in the RTN of rodents (16, 19, 20), exhibit hypersynchronized oscillations in the thalamocortical network because of reduced intra-RTN inhibition (21, 22).

Altered GABA_AR function has been described in the cerebral cortex and thalamus of two inbred strains of rats that exhibit spontaneous spike and wave seizures and are recognized as genetic models of absence epilepsy: genetic absence epilepsy rats from Strasbourg (GAERS) and Wistar albino Glaxo/Rij (WAG/Rij) rats (23–27). In GAERS rats, GABA_AR anomalies appear to be specific to the RTN and can be identified in recordings from RTN cells as inhibitory postsynaptic currents (IPSCs) of increased amplitude and faster decay times (23), leading to unusually strong bursts of action potentials in the neurons (28). Comparable changes in GABA-mediated IPSCs, however, could not be recorded in the upper layers of the cortex or in the ventral posterior nucleus (23). In WAG/Rij rats, GABAergic inhibition is reduced in upper-layer cortical neurons (26), and GABA_AR fast IPSPs show decreased peak conductances in deeper-layer neurons (25). These results suggest that subtle changes in GABA_AR function acting locally can have global effects on the thalamocortical network and that these changes may be differentially located in the two models of absence epilepsy. We show that GABA_ARs are made up of different combinations of subunits in somatosensory thalamus and that, in WAG/Rij rats, loss of a particular subunit in the reticular nucleus alone is part of the molecular phenotype of this model of absence epilepsy.

Results

EEG Analysis of WAG/Rij Rats.

Of nine WAG/Rij rats whose brains were subjected to immunostaining two showed >10 spike and wave discharges (SWDs) per hour (10 and 21), four showed 5–9 SWDs (5, 8, 9, and 9) per hour, and three showed <1 SWD (0, 0, and 1) per hour. The incidence, “morphology,” intraspike frequency and the mean duration of SWD were in agreement with previous data obtained in rats of the same age (29).

Cellular Expression in Nonepileptic Thalamus.

Immunostaining for α1 subunits was intense throughout dorsal thalamus, including VP [supporting information (SI) Fig. 5]. Labeling was significantly less dense in RTN, with only a small number of lightly stained neurons in the dorsal part of the nucleus and few or no stained processes. These few labeled cells could be costained for parvalbumin (SI Fig. 5 A–C).

Immunostaining for α3 subunits was absent from dorsal thalamus except for intralaminar nuclei and intense in RTN, with labeled cells and processes throughout the entire nucleus (SI Fig. 5D). Labeled processes in RTN were associated with punctate labeling along the plasma membrane.

Immunostaining for β2/β3 subunits, was dense in dorsal thalamus, including VP, and less dense in RTN although greater than for α3 subunits (SI Fig. 5E).

Synaptic Localization in Nonepileptic Thalamus.

α1 subunit.

VP.

Symmetrical synapses identified at the EM level were typically formed by medium-to-large-sized presynaptic terminals (2–3 μm) containing synaptic vesicles and mitochondria and associated with a thin PSD and symmetrical pre- and postsynaptic densities. Many symmetrical synapses contained >1 ImmunoGold particle at PSDs, where the majority of particles were associated, although a few were in the synaptic cleft or at the presynaptic membrane (Fig. 1 A and B). Gold particles consistently found in consecutive thin sections were in the same subcellular compartment (Fig. 1 A and B). Large symmetrical synapses on somata or proximal dendrites had multiple PSD segments each associated with clusters of gold particles (Fig. 1 A and B). In 129 synapses, 70% of the ImmunoGold particles indicative of α1 subunit labeling were concentrated at PSDs (Fig. 1C), 20% were within the synaptic cleft, and <10% were located at presynaptic membranes. Approximately 50% of the synapses were labeled by one or more than one particle (Fig. 1D); the majority contained two to three particles per PSD, with a few PSDs exhibiting >10 particles. The mean was 6.1 ± 3.7 particles per PSD (Fig. 2F).

Fig. 1.

Synaptic localization of GABA_AR α1 subunit immunoreactivity in VP and RTN of nonepileptic Wistar rats revealed by ImmunoGold electron microscopy. (A and B) Serial electron micrographs showing ImmunoGold particles representing α1 subunits associated with PSDs of two axon terminals (t1, t2) forming symmetrical synapses on a proximal dendrite (d) of a relay cell in VP; arrows indicate PSDs associated with ImmunoGold particles. Insets in A and B are high-power images showing ImmunoGold particles at the synaptic membranes. (C) Graph showing the spatial distribution of particles representing α1 subunits along synaptic membranes. Vertical lines define the presynaptic and postsynaptic membranes; space between the lines is the synaptic cleft (≈20 nm). Zero is defined as the middle of the synaptic cleft. Negative numbers denote the presynaptic side, and positive numbers denote the postsynaptic side. The majority of particles are concentrated at PSDs. (D) Bar graph showing frequency distribution of ImmunoGold particles representing α1 subunits at single PSDs of synapses in random thin sections. (E) Electron micrograph showing few ImmunoGold particles representing α1 subunits at a symmetrical synapse (t) contacting a dendrite (d) in RTN. (F) Bar graph showing frequency distribution of α1 particles at single PSDs of RTN synapses. Approximately 90% of the synapses lack gold particles. [Scale bars: 0.5 μm (A and B); 0.25 μm (E); 0.1 μm (Insets).]

Fig. 2.

Synaptic localization of α3 subunits in RTN of nonepileptic Wistar rats. (A and B) Serial electron micrographs showing ImmunoGold particles representing α3 subunits localized at PSDs of a symmetrical synapse formed by a terminal (t) on a dendrite (d). Insets in A and B are higher-power images. (C) Electron micrograph showing ImmunoGold particles representing α3 subunits associated with multiple PSD segments of a symmetrical synapse (t). Inset in C is higher-magnification view. (D) Graph showing distribution of α3 ImmunoGold particles in relation to synaptic membranes. The majority of ImmunoGold particles are located at the PSD. (E) Bar graph showing frequency distribution of ImmunoGold particles at PSDs in random thin sections. [Scale bars: 0.5 μm (A–C); 0.1 μm (Insets).] (F) Bar graph showing mean number of ImmunoGold particles representing α1 (black bar) or β2/β3 subunits (gray bar) per PSD profile of symmetrical synapses in VP. (G) Bar graph shows ImmunoGold particles representing α3 (dark gray bar) or β2/β3 subunits (light gray bar) per PSD profile of symmetrical synapses in RTN.

RTN.

Large numbers of synapses in RTN lacked α1 ImmunoGold particles (Fig. 1 E and F). Of 38 synapses ≈90% had no associated particles. ≈10% had a small number (Fig. 1 E and F) often not repeated in serial sections.

α3 subunit.

VP.

A few particles were found near synapses or in cytoplasm of dendrites, but labeling was insignificant compared with immunocytochemical controls (data not shown).

RTN.

ImmunoGold particles were associated with synaptic membranes of many symmetrical synapses (Fig. 2 A and B). In large synapses with several PSD segments (Fig. 2C), particles were present at each segment. Approximately 55% of particles were at or within 30 nm of the edge of the PSD (Fig. 2D). Twenty percent of particles were associated with presynaptic membranes (Fig. 2D). Of 110 symmetrical synapses, >70% were labeled by one or more particle per PSD. Most synapses showed three to five particles, and some showed >10 particles per PSD (Fig. 2E). The mean number was 6.6 ± 3.4 particles per PSD (Fig. 2G).

β2/β3 subunits.

VP.

ImmunoGold particles were found at many symmetrical synapses (SI Fig. 6 A and B). More than 60% were localized at PSDs, 10–15% were in the synaptic cleft, and 10% were associated with presynaptic membranes (SI Fig. 6D). Of 43 symmetrical synapses, >50% were labeled; the mean number of particles per PSD was 3.7 ± 1.3 (Fig. 2F).

RTN.

Of 51 symmetrical synapses, >50% were associated with one or more particle (SI Fig. 6I); the number was 3.5 ± 1.2 per PSD (Fig. 2G). Approximately 70% of particles were concentrated at PSDs and ≈30% were distributed in the synaptic cleft or at presynaptic membranes (SI Fig. 6 E–H and J).

Selective Loss of α3 Subunit Protein but Not mRNA in RTN of WAG/Rij Rats.

Epileptic WAG/Rij rats (n = 9) showed a dramatic loss of immunostaining for α3 subunits in RTN (Fig. 3 A and B). The loss of α3 subunit immunoreactivity did not result from cell loss, because immunostaining for GABA and parvalbumin showed no significant difference between animal groups (Fig. 4 E and F).

Fig. 3.

Alteration of immunostaining for GABA_AR α3 subunits in the thalamus of WAG/Rij rats. (A and B) Comparison of immunoperoxidase staining forα3 subunit in the thalamus of a nonepileptic Wistar rat (A) and a WAG/Rij rat (B). Note selective loss of α3 subunit immunostaining in RTN of the WAG/Rij rat. [Scale bars: 0.5 mm (A and B). (C) Electron micrograph taken from a thin section of a control Wistar rat RTN showing α3 subunit ImmunoGold particles (arrow) specifically associated with the PSD of a symmetrical synapse formed by an axon terminal (t1) on a dendrite (d). (Inset) High-magnification image of the ImmunoGold particles concentrated at the PSD. (Scale bar: 0.2 μm (C); 0.1 μm (Inset). (D) Electron micrograph from the RTN of a WAG/Rij rat showing absence of α3 subunit ImmunoGold particles at a symmetrical synapse. (Scale bar: 0.2 μm.)

Fig. 4.

Immunostaining in cortex and mRNA levels in thalamus. (A and B) Immunofluorescent staining for α3 subunits in cerebral cortex of a control Wistar rat (A) and a WAG/Rij rat (B). The immunostaining pattern is similar in both animals; immunoreactivity for the α3 subunit is concentrated in dendritic shafts of deep layer pyramidal cells. Inset in B is an electron micrograph of a dendritic shaft (d) in the cortex of a WAG/Rij rat labeled with α3 subunit immunoperoxidase reaction product and contacted by a symmetrical synapse (t). [Scale bar: 100 μm (B); 0.5 μm (Inset).] (C and D) Film autoradiograms from frontal sections through a control Wistar rat thalamus (C) and a WAG/Rij rat thalamus (D) showing hybridization of ³³P-labeled RNA probes complementary to GABA_AR α3 subunit mRNA. Expression is relatively strong in RTN of both animals. (Scale bar: 1 mm.) (E and F) Immunoperoxidase staining for parvalbumin in the thalamus of a control Wistar rat (E) and a WAG/Rij rat (F) showing the identical pattern of strong immunoreactivity in the RTN. (Scale bar: 0.5 mm.)

Loss of α3 subunit immunostaining at the cellular level was also reflected at the synaptic level. In contrast to the abundant synaptic labeling for the α3 subunit in RTN of nonepileptic Wistar rats (Figs. 2 and 3C), neither immunoperoxidase reaction product nor a significant number of ImmunoGold particles were found at symmetrical synapses in the RTN of WAG/Rij rats (Fig. 3D and SI Fig. 7A). By contrast, no significant difference in α3 subunit immunostaining was found in the cerebral cortex of nonepileptic Wistar and WAG/Rij rats, both showing strong immunostaining for α3 subunit in apical dendrites of pyramidal neurons (Fig. 4 A and B). For β2/β3 subunits at symmetrical synapses in RTN or cortex, the mean number of particles showed no significant difference in Wistar and WAG/Rij rats (P ≤ 0.70). Immunostaining for α1 subunits showed an inconsistent reduction in the thalamus of WAG/Rij rats (data not shown).

Sections from three WAG/Rij rats showing >9 or <0.75 SWDs per hour and two Wistar rats when processed simultaneously for in situ hybridization histochemistry displayed α3 subunit mRNA expression at approximately equal levels in RTN and cortex of WAG/Rij and control Wistar rats (Fig. 4 C and D and SI Fig. 7B). Densitometry analysis of autoradiograms showed no significant differences in α3 mRNA expression in RTN (P ≤ 0.74) or cortex (P ≤ 0.14) in WAG/Rij and Wistar rats.

Discussion

We combined immunocytochemistry and in situ hybridization histochemistry with quantitative immunoelectron microscopy to compare cellular and synaptic localization of three major GABA_AR subunits in VP and RTN of WAG/Rij rats and control rats. The synaptic distribution of the α1, α3, and β2/β3 subunits reflected cellular expression patterns of the subunits in VP and RTN, with α3 being the major subunit expressed at inhibitory synapses in RTN. At the synaptic level, α1, α3, and β2/β3 subunits were exclusively localized at symmetrical synapses in VP and RTN and were concentrated at the PSD with little extrasynaptic localization. Immunolabeling for α3 subunits was selectively lost at both cellular and synaptic levels in RTN of WAG/Rij rats but with no alteration in α3 subunit mRNA and no sign of cell loss or change in other markers of GABAergic cell identity in RTN and with no comparable loss of α3 immunoreactivity in the cerebral cortex.

Composition of GABA_A Receptors and Functional Implications.

The kinetics and pharmacological properties of GABA_A receptors depend on subunit composition (30–34) and where subunits are assembled on plasma membranes of neurons. In the CNS, a functional GABA_AR requires a combination of two α, two β and one γ subunit, but the γ subunit can be replaced by δ, ε, or π (35). Inclusion of specific subunits can affect not only channel kinetics but also localization of GABA_ARs. GABA_AR, in the combination α6β2/3δ are located on nonsynaptic plasma membranes, have high affinity for GABA, do not desensitize in the presence of GABA, and contribute mainly to tonic inhibition of neurons. α1β2/3γ2 receptors have low affinity for GABA, show more pronounced desensitization to GABA, and are concentrated at synaptic junctions where they mediate strong, phasic inhibition, because a high concentration of GABA is present only briefly (36–38). The subunits examined in the present study were all concentrated at the core of the PSDs of inhibitory synapses.

α1, β2, and γ2 are the principal GABA_AR subunits expressed in relay nuclei of the dorsal thalamus of rodents (39) and primates (40), suggesting that most GABA_AR in those nuclei are made up of a combination of these three subunits. Significant differences appear in the intralaminar nuclei (16, 40), but their functional significance has not been explored. In the RTN of rodents, α3 is the principal subunit expressed, with lower levels of expression of β1, β2/β3, and γ2 subunits (19, 39). In monkeys and humans, the γ2 subunit is the principal subunit expressed, with lower levels of expression of α3 and β1 subunits (40, 41). The reversed levels of α3 and γ2 subunit expression in rodents and primates is of interest, given the apparent relationship of the former to the two rat models of absence epilepsy and the report of mutations in the latter in cases of human idiopathic generalized epilepsy (7).

α1 subunit-containing GABA_AR display fast decay and large current amplitudes; IPSP decay kinetics are much faster and evoked synaptic currents larger in VP relay cells than in RTN cells (42, 43). α3 subunits coexpressed with β2 and γ2 in transfected HEK293 cells form GABA_AR with slower activation, deactivation, and desensitization (34, 42, 44). Because GABA_AR at inhibitory synapses between neurons in the RTN of rodents are dominated by α3 subunits, these synapses should display similar kinetics. In mice null for the β3 subunit, spindle oscillations become converted into hypersynchronized 3-Hz oscillations in acute thalamic slices akin to those found during spike and wave discharges during an absence seizure (21). In GAERS rats, IPSCs in RTN neurons display faster kinetics than in controls (23), and this is also associated with paroxysmal spike and wave discharges. It will be interesting to repeat our investigation on the RTN of GAERS rats.

Role of GABA_AR a3 Subunits in Absence Epilepsy.

Inhibition between RTN cells affects the strength of inhibition from RTN to VP cells so that burst firing generated by deinactivating T-channels in hyperpolarized relay cells can be tightly controlled to prevent the hypersynchrony of thalamic and cortical cells observed during SWDs (45–47). When intra-RTN inhibition is blocked, normal SWDs are transformed to hypersynchronous epileptiform responses. The exact site of origin of absence seizures is still debated (48): SWDs in genetic rodent models of absence epilepsy appearing in the somatosensory cortex are propagated to other cortical areas and into the corticothalamocortical network (48, 49), where abnormal thalamic synchronization of burst firing is crucial for sustaining a seizure (50, 51).

All WAG/Rij rats have SWD by 6 months of age, and all WAG/Rij rats examined in this study had displayed seizures before the experiments commenced. Although three of the nine exhibited fewer than one SWD per hour, all displayed the marked reduction in α3 receptor polypeptide in the RTN. Therefore, lack of α3 receptor is probably not induced by absence seizures but represents part of the epileptic phenotype and predates onset of seizures. Alterations in GABA_AR kinetics in the RTN of GAERS rats also antedate onset of seizures (23).

There was no significant alteration of α3 subunit mRNA in RTN of WAG/Rij rats. This strongly indicates that the transcriptional machinery for producing α3 subunit protein is still functional in RTN of epileptic WAG/Rij rats, but that translational mechanisms may be impaired. Apart from interference with α3 subunit protein synthesis, defective synaptic targeting of the subunit, perhaps by affecting synaptic clustering proteins such as gephyrin, as demonstrated in α3-null mice (52), could prevent α3 polypeptide from reaching synapses in RTN. Although we have little or no information about specific mechanisms responsible for the highly selective reduction in α3 subunit polypeptide in the RTN of WAG/Rij rats, it seems clear from the absence of demonstrable alterations in GABA_AR in the dorsal thalamus and cerebral cortex that the electrophysiological hallmarks of generalized epilepsy exhibited by this model do not stem from general global malfunction of GABA_A receptors.

Materials and Methods

Control Rats.

Wistar rats (weight 200–350 g; Charles River Laboratories, Wilmington, MA) were used. All experiments were approved by the Institutional Animal Care and Use Committee. Animals were deeply anesthetized with sodium pentobarbital and perfused through the heart with normal saline, followed by 4% paraformaldehyde in 0.1 M phosphate buffer (for light microscopy) or by 4% paraformaldehyde and 0.1–0.5% glutaraldehyde in 0.1 M phosphate buffer (for electron microscopy). Brains were cut either on a sliding microtome at 15–20 μm (for light microscopy) or on a vibratome at 400–500 μm (for electron microscopy).

WAG/Rij Rats.

Three-month-old male WAG/Rij rats, born and raised in the Biological Psychology Department of Radboud University, Nijmegen, served as subjects. All procedures were approved by the Ethical Committee on Animal Experimentation of Radboud University. Rats were maintained on 12-h light/dark cycle (white light off at 7:00 a.m.) with free access to water and food. Before surgery, rats were housed five to six animals per cage. A tripolar electrode set (Plastics One, Roanoke, VA) was implanted under isoflurane anesthesia. One electrode was implanted into the right frontal cortex (AP +2.0, L +2.0), a second in the parietal/occipital region (AP −6.0, L +4.0) (coordinates with skull surface flat and bregma 0–0), and a third, placed over the cerebellum, served as ground. The length of the recovery period was 7–10 days.

EEG recording and analysis.

The rats were allowed to adapt to the recording cage and EEG leads for at least 12 h before recording started. EEG recordings were done in the dark (between 9:00 a.m. and 1:00 p.m.) in free moving animals. Signals were amplified, those <1 Hz or >100 Hz were filtered out, and then digitized (sample rate 200 Hz), monitored, and stored on optical disks for off-line analysis with WINDAQ data acquisition software. The number and duration of spike-wave discharges were identified with a custom software routine, and the identified periods were visually classified according to established criteria (53). Nine of the rats were used for immunocytochemistry.

Perfusion.

After recording, rats were deeply anesthetized with barbiturate and perfused with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.3) for light microscopy or with 4% paraformaldehyde plus 0.2% glutaraldehyde in 0.1 M phosphate buffer (pH 7.3) for electron microscopy and sectioned as above.

Immunocytochemistry.

Sections from brains of Wistar and WAG/Rij rats were stained simultaneously. Sections were incubated in preimmune serum to block nonspecific staining and were then placed in the following primary antibodies for 48 h at 4°C: anti-GABA_AR α1or α3 subunit (rabbit polyclonal antibodies, diluted 1:200 to 1:500; Alomone Labs, Jerusalem, Israel), anti-GABA_AR β2/β3 subunits (a monoclonal antibody that recognizes both β2 and β3 subunit epitopes, diluted 1:400, a gift from Dr. Angel L. de Blas, University of Connecticut, Storrs, CT). Sections were washed and incubated in biotinylated goat anti-rabbit or goat anti-mouse secondary antibodies (1:150; Vector Laboratories, Burlingame, CA), and immunoreactivity was visualized by the diaminobenzidine (DAB) reaction. For double immunofluorescent labeling, sections were incubated in a mixture containing anti-α1 subunit and anti-parvalbumin antibodies (a mouse monoclonal antibody, diluted 1:1,000; Sigma). Sections were reincubated in a solution containing fluorescein- (for α1 subunit) and Cy5- (for parvalbumin) conjugated secondary antibodies (1:200; Molecular Probes, Eugene, OR) and examined by using dual channels of a Fluoview confocal microscope (Olympus, Melville, NY). Control experiments consisted of omitting the primary or secondary antibodies and resulted in only background labeling.

Preembedding Immunoelectron Microscopy.

Some sections immunostained for GABA_AR subunits by the DAB reaction were examined under light microscope, and specific thalamic regions (VP or RTN) and the somatosensory cortex were dissected and thin sectioned and examined in the electron microscope (54).

Postembedding ImmunoGold Electron Microscopy.

Vibratome sections were examined under a dissecting microscope; the VP and RTN of the thalamus and the somatosensory cortex were identified and cut out in small pieces (1 × 1 × 0.5 mm). The blocks were processed by cryofixation in a KF80 unit (Reichert, Wien, Austria) and cryoembeded in a cryoembedding unit (Leica, Wetzler, Germany) at 0°C (55). Ultrathin serial sections were cut at 70 nm on an ultramicrotome and collected on Formvar-coated single-slot nickel grids. Thin sections containing RTN or VP were incubated at room temperature in the following primary antibodies as described above: anti-α1subunit (1:100); anti-α3 subunit (1:100 to 1:200); anti-β2/β3 subunit (1:200). After washing, sections were incubated in secondary antibodies conjugated to 10 nm or 15 nm gold particles (Biocell, Cardiff, U.K.) for 1 h at room temperature. Some thin sections from VP or RTN were processed for double immunolabeling for α1 and β2/β3 subunits by using secondary antibodies conjugated to 10- or 20-nm gold particles. Grids were lightly stained with lead and uranyl acetate and examined in a CM120 electron microscope (Philips, Amsterdam, The Netherlands). Controls were carried out on the thin sections by omitting primary or secondary antibodies, and only background labeling was detected in EM. Electron micrographs were captured by a 2,000 × 2,000 resolution CCD camera (Gatan, Pleasanton, CA). Digitized images were processed by using DigitalMicrograph software (Gatan) and composed in Adobe Photoshop CS (Adobe Systems, Mountain View, CA).

EM Data Analysis.

In VP and RTN and for each receptor subunit, at least 30 typical symmetrical synapses were identified. Only gold particles localized within 20 nm of pre- or postsynaptic membranes were included in the quantitative analysis (56). Criteria for inclusion as labeled synapses were: (i) clear pre- and postsynaptic membranes; (ii) a clear PSD; (iii) at least one gold particle localized within 20 nm of the edge of the PSD; (iv) the labeling detectable in at least two serial sections. Two sets of quantitative data were generated: (i) Number of ImmunoGold particles per PSD: Randomly selected areas of VP or RTN from Wistar and WAG/Rij rats were examined, all morphologically identified synapses showing clear synaptic contacts (57) were recorded, all ImmunoGold particles associated with PSD profiles of the synapses were counted, and mean number of particles per PSD was calculated by dividing total gold particles by total number of PSDs and (ii). Spatial distribution of ImmunoGold particles at synapses: The distribution of gold particles across the pre- and postsynaptic membranes was quantified. At each labeled synapse, the distance between each gold particle and the middle of the synaptic cleft was measured by using Scion Image (Scion, Frederick, MD). Zero point was defined as the middle of the synaptic cleft, positive values denote postsynaptic to the middle of the cleft; negative values denote presynaptic to the middle of the cleft. Percentages of ImmunoGold particles at each 10-nm bin crossing the pre- and postsynaptic membranes were calculated and plotted (55). For quantification, the number of particles at single synapses was expressed as mean ± SD. To compare differences between Wistar and WAG/Rij rats, for each subunit, at least 20 synapses from each region were randomly chosen for statistical analysis by Student's t test. A P value ≤0.01 was considered significant.

In Situ Hybridization Histochemistry.

Sections from control Wistar and epileptic WAG/Rij rats were processed simultaneously for in situ hybridization histochemistry with GABA_AR α3 subunit specific [³³P]UTP labeled cRNA probes (40, 58, 59). Film autoradiograms of hybridized sections were digitized and quantified by taking optical density readings over defined areas by using a microcomputer imaging system (MCID/M6; Imaging Research, St. Catharines, Ontario, Canada). Optical density readings were converted to levels of radioactivity by reference to radioactive standards exposed on the same sheet of film (56). Student's t test was used to compare differences in density of labeling between Wistar and WAG/Rij rats. A P value ≤0.01 was considered significant.

Abbreviations

EEG

electroencephalogram

GABA_AR

GABA_A receptor

PSD

postsynaptic density

RTN

reticular nucleus

SWD

spike-wave discharge

VP

ventral posterior nucleus.