GABA_B autoreceptor-mediated cell type-specific reduction of inhibition in epileptic mice

Significance

Metabotropic GABA_B receptors control synaptic transmission and excitability in neuronal circuits of the brain. Although effects of these receptors are predominantly inhibitory at both cellular and network levels, application of the agonist baclofen can promote excitability and induce seizures in patients and animal models of epilepsy. Here we demonstrate that proepileptic effects of baclofen are concentration dependent and result from disinhibition. Although at high doses, baclofen reduces network excitability due to its combined pre- and postsynaptic inhibitory effects in pyramidal cells, at low doses, it leads to an enhanced presynaptic suppression of the synaptic output of a specific set of inhibitory neurons. This disinhibitory effect promotes high-frequency oscillations and the emergence of pathological discharges in the epileptic hippocampal network.

Abstract

GABA_B receptors (GABA_BRs) mediate slow inhibitory effects on neuronal excitability and synaptic transmission in the brain. However, the GABA_BR agonist baclofen can also promote excitability and seizure generation in human patients and animals models. Here we show that baclofen has concentration-dependent effects on the hippocampal network in a mouse model of mesial temporal lobe epilepsy. Application of baclofen at a high dose (10 mg/kg i.p.) reduced the power of γ oscillations and the frequency of pathological discharges in the Cornu Ammonis area 3 (CA3) area of freely moving epileptic mice. Unexpectedly, at a lower dose (1 mg/kg), baclofen markedly increased γ activity accompanied by a higher incidence of pathological discharges. Intracellular recordings from CA3 pyramidal cells in vitro further revealed that, although at a high concentration (10 µM), baclofen invariably resulted in hyperpolarization, at low concentrations (0.5 µM), the drug had divergent effects, producing depolarization and an increase in firing frequency in epileptic but not control mice. These excitatory effects were mediated by the selective muting of inhibitory cholecystokinin-positive basket cells (CCK⁺ BCs), through enhanced inhibition of GABA release via presynaptic GABA_BRs. We conclude that cell type–specific up-regulation of GABA_BR-mediated autoinhibition in CCK⁺ BCs promotes aberrant high frequency oscillations and hyperexcitability in hippocampal networks of chronic epileptic mice.

Neuronal activity in the hippocampus shows oscillations in behavior-relevant frequency ranges including γ frequencies (30–80 Hz) (1). γ activity is prominent in the aroused brain and has been implicated in higher-level brain functions, such as sensory binding, perception (2), and storage and recall of information (3, 4). At the same time, γ frequency oscillations are also prevalent in epileptic patients and are most often observed at seizure onset during in depth EEG recordings (5). The GABAergic system plays a pivotal role in the generation of γ oscillations (6–8). However, it remains to be resolved how distinct GABAergic receptor subtypes, in particular GABA_B receptors (GABA_BRs), contribute to the generation and modulation of pathological network oscillatory activity.

GABA_BRs mediate slow inhibitory effects and control synaptic transmission and the excitability of neurons in cortical networks. GABA_BRs are expressed both postsynaptically in somato-dendritic compartments and presynaptically in axon terminals, in excitatory principal cell and inhibitory interneurons (9–11). The effects of GABA_BR activation on the network are dominated by inhibition leading to an overall dampened population activity. However, if GABAergic interneurons are effected dominantly, as observed for example, during high-frequency stimulation, GABA_BR activation can produce disinhibition in principal cells (12, 13). Accordingly, the role of GABA_BRs in epilepsy and seizure generation remains ambiguous. GABA_BRs are expected to have an overall antiepileptic effect, and indeed, the receptor KO animals show an epileptic phenotype (14). However, there is also evidence that the receptor agonist baclofen can induce seizures in patients after intrathecal application (15, 16). The picture is further complicated by the fact that GABA_BR expression can be altered in both epileptic patients, e.g., in mesial temporal lobe epilepsy (mTLE) (17), and animal models (18). Thus, cell type–specific alterations in GABA_BR expression may change network excitability during the progression of mTLE.

Using a chronic kainate (KA) model of mTLE, which reproduces major electrophysiological and histopathological characteristics of human mTLE (19, 20), we studied the role of GABA_BRs in altered hippocampal network activity. Our results suggest that enhanced and persistent GABA_BR activation in epileptic mice suppresses the inhibitory output from hippocampal interneurons, in particular cholecystokinin (CCK)-expressing basket cells (BCs) onto pyramidal cells (PCs). This reduction in the inhibitory output of interneurons, in turn, leads to disinhibition in hippocampal networks, enhances γ activity, and promotes the transition to pathological hyperexcitability.

Results

Dose-Dependent Effects of GABA_BR Activation on Hippocampal Network Activity in a Mouse Model of mTLE in Vivo.

We first analyzed the effects of GABA_BR activation on network activity in vivo in chronic epileptic mice. EEG recordings were carried out from the ventral Cornu Ammonis area 3 (CA3) area of the hippocampus of freely moving control and KA-injected mice during exploratory behavior (Fig. 1). In good agreement with our previous observation (20), in epileptic mice, γ activity was augmented, and sporadic pathological discharges were observed (Fig. 1 A and B). To characterize the effects of GABA_BR activation in these animals, we applied the GABA_BR agonist baclofen by i.p. injection. Because contrasting effects of baclofen have been reported in patients at different doses, we compared effects at low and high doses of the drug. Consistent with its dominant inhibitory effect, at a high dose (10 mg/kg), baclofen strongly suppressed both the network γ (from 26.2 × 10⁻⁶ ± 4.2 × 10⁻⁶ to 12.4 × 10⁻⁶ ± 2.8 × 10⁻⁶ mV²/Hz) and the occurrence of pathological discharges (from 3.0 × 10⁻¹ ± 0.6 × 10⁻¹ to 0.7 × 10⁻¹ ± 0.3 × 10⁻¹ Hz; Fig. 1 E–H) in chronic epileptic mice. In contrast, a low dose (1 mg/kg), which did not affect the network γ activity in control mice, significantly increased the γ power in epileptic mice from 29.8 × 10⁻⁶ ± 2.4 × 10⁻⁶ to 40.4 × 10⁻⁶ ± 3.8 × 10⁻⁶ mV²/Hz (P < 0.05, n = 3, Fig. 1 B–D). Moreover, this effect was accompanied by an increased frequency of pathological discharges in the epileptic animals from 2.3 × 10⁻¹ ± 0.3 × 10⁻¹ to 7.3 × 10⁻¹ ± 2.3 × 10⁻¹ Hz (P < 0.05, n = 3; Fig. 1 A and D), suggesting an increased susceptibility of the hippocampal network to epileptiform activity following activation of GABA_BRs in these mice.

Fig. 1.

Dose-dependent effects of i.p. baclofen on γ activity and interictal events in hippocampal CA3 network of epileptic mice in vivo. (A) Representative electroencephalogram from a freely moving mice before (Left) and 2 (Center) and 4 h (Right) after i.p. injection of a low dose (1 mg/kg) of baclofen. Note a transient increase in pathological activity (Center) following baclofen application. (B) The power spectra of γ activity (band-pass filtered 30–80 Hz) before (gray) and after (black) injection of baclofen. Note the markedly higher power under the influence of baclofen. (Inset) Wide-band power spectra (1–1,000 Hz); arrow indicates the γ peak. (C) Peak power of γ oscillation plotted against time. The arrow indicates the time point (0 min) of baclofen injection. (D) Summary bar charts of peak γ power (n = 3) and the frequency of interictal spike-and-wave events (n = 3). (E–H) Corresponding data obtained in epileptic mice following injection of a high dose (10 mg/kg) of i.p. baclofen. Note that γ power and the frequency of interictal spike-and-wave events decreased after application of the high dose of baclofen (n = 4 and n = 3, respectively). In D and H, asterisks indicate significant differences (P < 0.05, Student t test).

Concentration-Dependent Effects of GABA_BR Activation and Deactivation on Network γ Activity in Vitro.

To investigate cellular, synaptic, and network mechanisms of GABA_BR effects under controlled conditions, we continued our experiments in vitro using acute hippocampal transverse slice preparations. As shown before (20–22), bath application of a low concentration of KA (200 nM) reliably induced γ frequency oscillatory activity in the CA3 area, which was significantly stronger in slices from epileptic mice than in those from control animals (peak power in control: 3.1 × 10⁻⁴ ± 5.7 × 10⁻⁵ mV²/Hz, n = 13; epileptic: 9.4 × 10⁻⁴ ± 2.2 × 10⁻⁴ mV²/Hz, n = 8; P = 0.0027), whereas the frequency did not differ significantly (control: 41.4 ± 0.9 Hz, n = 13; epileptic: 36.0 ± 3.5 Hz, n = 8, P = 0.084).

To clarify the mechanisms underlying the divergent effects of baclofen application in vivo, we next evaluated the influence of GABA_BR activation on network activity at a low and high concentration. In good agreement with the in vivo results, bath application of baclofen had a concentration-dependent effect in vitro also. At a high concentration (10 µM), baclofen ab initio completely abolished γ frequency field potential oscillations recorded in slices from control mice. Unexpectedly, in slices from epileptic mice, we observed an initial increase in power of network γ activity [from 7.2 × 10⁻⁴ ± 2.7 × 10⁻⁴ (31.9 ± 3.7 Hz) to 2.3 × 10⁻³ ± 7.8 × 10⁻⁴ mV²/Hz, n = 7, P = 0.04 (32.9 ± 5.3 Hz, P = 0.87)] lasting between 5 and 10 min. Only following this transient increase did baclofen attenuate γ activity in epileptic mice (Fig. S1 D–F). At a low dose (0.5 µM), baclofen reduced the power of γ oscillations in control mice [from 4.2 × 10⁻⁴ ± 3.0⁻⁶ (32.9 ± 1.8 Hz) to 1.8 × 10⁻⁴ ± 1.2 × 10⁻⁶ mV²/Hz (30.9 ± 4.3 Hz), n = 5, P < 0.05]. In contrast it increased, albeit not significantly, network activity in slices from chronic epileptic mice at this low concentration [from 8.8 × 10⁻⁴ ± 6.4 × 10⁻⁴ (33.2 ± 1.8 Hz) to 1.3 × 10⁻³ ± 1.1 × 10⁻⁴ mV²/Hz (34.3 ± 0.1 Hz), n = 4; Fig. S1 A–C]. Furthermore, in slices obtained from epileptic animals, baclofen persistently (1 of 7 slices at 0.5 µM; Fig. S1B) or transiently (10 of 12 slices at 10 µM) converted enhanced γ oscillations to a pathological discharge pattern.

To further examine the impact of GABA_BRs, we studied effects of receptor blockade on the network oscillations by bath applying the selective antagonist (2S)-3-[[(1S)-1-(3,4-dichlorophenyl)ethyl]amino-2-hydroxypropyl](phenylmethyl)phosphinic acid CGP55845 (2 µM; Fig. S2). As expected, CGP had opposite effects to baclofen on the ongoing γ oscillations: it increased the power of γ activity in control slices (from: 2.5 × 10⁻⁴ ± 6.1 × 10⁻⁵ to 4.8 × 10⁻⁴ ± 1.1 × 10⁻⁴ mV²/Hz, n = 8, P < 0.05), indicating the presence of a tonic level of GABA_BR activation. In contrast, CGP decreased the power of oscillations in slice from epileptic animals (from 7.9 × 10⁻⁴ ± 2.8 × 10⁻⁴ to 3.0 × 10⁻⁴ ± 1.0 × 10⁻⁴ mV²/Hz, n = 9, P < 0.05). The frequency of γ oscillation was not significantly changed in either of the two groups: (41.6 ± 0.9 vs. 43.1 ± 0.9 Hz, n = 8 and 39.4 ± 1.3 vs. 40.2 ± 1.1 Hz, n = 9, P > 0.05). In contrast to baclofen, we did not observe a biphasic effect in responses to CGP application, plausibly reflecting that there is only a low level of tonic activation of presynaptic receptors under our conditions.

These results confirm in vivo observations that low levels of GABA_BR activation can enhance γ frequency oscillations and also increase the frequency of pathological discharges in the epileptic hippocampus.

Effects of Baclofen on the Firing and Membrane Properties of PCs.

We continued our analysis by performing intracellular recordings from CA3 PCs and investigated the effects of baclofen on firing and membrane properties at high (5–10 µM) and low (0.5 µM) concentrations. We first analyzed the impact of baclofen on the discharge of PCs during KA-induced oscillations. During oscillations, baclofen (10 µM) reduced the firing frequency of PCs in slices from control mice (from 3.38 ± 0.18 to 1.18 ± 0.47 Hz between 5 and 10 min and 0.12 ± 0.07 Hz at 20 min, n = 5). In contrast, in cells from epileptic mice, baclofen transiently increased (from 7.06 ± 0.98 to 29.93 ± 8.70 Hz, between 5 and 10 min after baclofen application, n = 4) and subsequently decreased the discharge frequency (to 11.07 ± 7.92 Hz, after 20 min, n = 4; Fig. 2G). Low doses of baclofen (0.5 µM) consistently increased the firing frequency (Fig. 2A) in PCs from epileptic mice.

Fig. 2.

Baclofen produces depolarization in CA3 PCs from epileptic mice. (A and B) Bath application of a lower concentration of baclofen (0.5 µM, arrows) induced membrane potential depolarization in PCs from epileptic (n = 5) but not control (n = 4) animals. (C) Bath application of a higher concentration of baclofen (10 µM) suppressed KA-induced field oscillations (Upper), produced hyperpolarization (Lower), and decreased the firing frequency of PCs in acute slices from control mice (D). (E and F) In contrast, baclofen transiently increased the network oscillations, depolarized, and increased the firing frequency of PCs in epileptic animals. Actions potentials are truncated for clarity. (G) Mean values of firing frequency of PCs from control (n = 5) and epileptic mice (n = 4) at different time points following the application of KA and baclofen (Bacl).

We next examined the time-dependent changes in the membrane potential (MP) following bath application of baclofen at a high concentration (Fig. S3 A and B) and observed an initial transient depolarization of the MP (mean: +4.3 ± 0.85 mV; peak time: 7–10 min; n = 5) in PCs in epileptic slices. This effect was followed by a strong sustained hyperpolarization (−9.5 ± 0.9 mV, n = 5). Importantly, the depolarization of the cells showed a tight temporal correlation with the increases in the firing frequency and the power of network γ oscillations (see above). In comparison, in control slices, the initial depolarization was absent, and the amplitude of the hyperpolarization was significantly larger (−17.0 ± 2.6 mV, n = 5, P = 0.04). The hyperpolarization was associated with a reduction in the input resistance of PCs in both control and epileptic mice, consistent with the activation of an inhibitory conductance.

In contrast to the high concentration, 0.5 µM baclofen had solely a depolarizing effect on the MP in epileptic slices (+6.4 ± 2.4 mV, n = 5), albeit with slower kinetics (peak amplitude was at 15.9 ± 0.4 min, n = 5) than those induced by 10 µM baclofen (Fig. 2A). In control slices in the presence of low doses of baclofen, no depolarization was detected; instead, a small and slowly developing hyperpolarizing effect was observed (−1.0 ± 0.5 mV, peak time at 16.0 ± 1.0 min, n = 4; Fig. 2A).

To assess the possible contribution of glutamatergic transmission to the observed changes in MP, in a set of experiments, baclofen (10 µM) was applied in the presence of ionotropic glutamate receptor antagonists. Under this condition, application of baclofen also led to an initial MP depolarization (mean: +5.6 ± 0.5 mV, n = 5) and subsequent hyperpolarization (−8.0 ± 1.5 mV, n = 5) in PCs in epileptic slices. In control slices, baclofen invariably resulted in strong MP hyperpolarization (−10.8 ± 1.8 mV, n = 5, P > 0.05 in comparison with the epileptic sample).

To examine the effect of baclofen on inhibitory transmission, we next monitored spontaneous inhibitory postsynaptic currents (sIPSCs) in PCs in slices from control and epileptic mice in the presence of ionotropic glutamate receptor antagonists. Analysis of sIPSC frequency showed a decrease in the presence of a low dose of baclofen (0.5 µM) in both groups. However, this effect was significantly stronger in epileptic (from 15.5 ± 1.4 to 1.9 ± 1.5 Hz) than in control animals (from 10.9 ± 1.4 to 6.9 ± 0.9 Hz, 8 cells in each group, P < 0.05). In contrast, the mean amplitude of IPSCs was not changed significantly following application of baclofen in the two groups (63.9 ± 13.3 vs. 40.0 ± 5.4 pA and 53.5 ± 3.2 vs. 46.5 ± 4.2 pA before vs. after baclofen in epileptic and control mice, respectively, P > 0.05). These results indicate an enhanced presynaptic regulation of the inhibitory synaptic output of interneurons by GABA_BRs and point to its role underlying the changes observed in the MP and excitability of PCs.

GABA_BR-Mediated Autoinhibition of the Inhibitory Output of CCK⁺ Interneurons onto PC Is Augmented in Epileptic Mice.

To further examine whether the observed changes in network activity are the result of altered properties of pre- or postsynaptic GABA_BRs in epileptic mice, we analyzed evoked monosynaptic inhibitory responses in PCs. Monosynaptic IPSCs of perisomatic origin were elicited by electrical stimulation in stratum pyramidale close to the recorded PCs in the presence of ionotropic glutamate receptor antagonists (Fig. S4). Using a paired-pulse stimulation protocol, we found depression of the second IPSCs in cells from both control and epileptic mice. However, the paired-pulse ratio of IPSC amplitudes was significantly smaller in epileptic than in control PCs (0.61 ± 0.02% in epileptic vs. 0.71 ± 0.08% in control, P < 0.05; Fig. S4). As the enhanced paired-pulse inhibition of the monosynaptic IPSCs is attributed to the enhanced presynaptic GABA_BR activity, we next studied the effects of a GABA_BR antagonist CGP55845 in the two groups (Fig. S5). CGP significantly increased the paired-pulse ratio in epileptic (from 0.67 ± 0.03% to 0.95 ± 0.05%, P = 0.01) but not in control mice (from 0.77 ± 0.04% to 0.88 ± 0.07%, P = 0.07), suggesting a stronger presynaptic inhibition mediated by GABA_B autoreceptors in epileptic mice.

Perisomatic inhibition is provided by two neurochemically distinct groups of interneurons that express either CCK or parvalbumin (PV) with distinct functional properties. To evaluate a possible cell type–specific GABA_BR-mediated modulation of perisomatic inhibition, we isolated IPSCs produced by CCK⁺ and PV⁺ BC axon terminals on the basis of their sensitivity to presynaptic modulation. We first checked the sensitivity of the monosynaptic IPSCs to agonists of presynaptic cannabinoid receptors [(R)-(+)-[2,3-Dihydro-5-methyl-3-(4–morpholinylmethyl)pyrrolo[1,2,3-de]-1,4-benzoxazin–6-yl]-1-naphthalenylmethanone mesylate WIN55.212-2] to identify whether they were mediated by CCK⁺ axons. The drug was applied briefly (3 min), and changes in the postsynaptic responses were monitored. Baclofen was applied after full recovery from the initial drug effect (≥10 min after the agonist washout; Fig. 3). We found that a low concentration of baclofen consistently suppressed WIN-sensitive IPSCs in both groups; however, the effect of baclofen was significantly stronger in epileptic than in control mice (IPSC amplitude reduced to 24.8 ± 5.8% vs. 53.8 ± 7.1% of control, P = 0.01, n = 5; Fig. 3 A–C). Indeed, the slopes of the regression lines fitted to the plots of prebaclofen vs. postbaclofen IPSC amplitudes showed a significant difference between the two groups (Fig. 3C). The suppression of inhibitory output was stronger in the presence of 10 µM baclofen (6.0 ± 5.0% vs. 36.1 ± 9.8% of control, n = 3, P = 0.03). To confirm the identity of WIN-insensitive IPSCs, we applied the M2 receptor agonist, arecaidine but-2-ynyl ester tosylate (ABET), and found that these IPSCs were reduced by this drug (five of five tested), consistent with their origin from PV-positive axons (23). In contrast to WIN-sensitive IPSCs, the suppression of WIN-insensitive and ABET-sensitive IPSCs was not different in the two groups in the presence of a low (87.7 ± 7.5% and 86.1 ± 3.2% of control; Fig. 3C) and high concentration of baclofen (50.3 ± 6.9% vs. 60.8 ± 6.7% of control, P > 0.05). These data provide direct evidence that an increased presynaptic suppression of inhibition produced by GABA_BR activation in epileptic mice is interneuron type specific. Our data further suggest that this disinhibitory effect is due to an increased sensitivity of cannabinoid (CB1) receptor–expressing CCK⁺ BC terminals to baclofen.

Fig. 3.

Cell type–specific reduction of GABA_BR-mediated inhibition in epileptic mice. (A) Bath application of a low dose of baclofen (0.5 µM) decreased the amplitude of WIN55,212-2–sensitive, monosynaptic IPSCs evoked by minimal stimulation. This effect was significantly stronger in PCs from epileptic than control mice. (B) Time course of the peak amplitude of WIN-sensitive IPSCs in a CA3 pyramidal cell (horizontal bar, 1 µM WIN) during bath application of the agonist baclofen (0.5 µm Bacl) and the antagonist CGP (2 µM CGP). The cell´s access resistance (R_a) was continuously monitored during the experiment (Lower). (C) Linear regression lines fitted to the plots of prebaclofen vs. postbaclofen IPSC amplitudes from epileptic and control mice. The slopes of the regression lines were markedly different for WIN-sensitive IPSCs (WIN-sensitive, 0.19 ± 0.05 vs. 0.45 ± 0.03, in epileptic and control mice, respectively, n = 5) but not for WIN-insensitive IPSCs (WIN-insensitive, 0.86 ± 0.03 vs. 0.84 ± 0.02, n = 7). (D) Summary bar chart of the effect of WIN on the amplitude of WIN-sensitive (27.6 ± 2.5% of control, n = 6) and WIN-insensitive (91.9 ± 3.2% of control, n = 5) IPSCs in epileptic animals.

The observed increase in presynaptic GABA_BR efficacy could be mediated by changes in the number of GABA_BRs or by changes in their functional coupling to effectors (24–27). To clarify this issue, we performed double immunofluorescent staining for the GABA_B1 subunit and CCK. Integrated fluorescence intensity ratio analysis of colocalized GABA_B1 and CCK immunoreactivities (Fig. S6 A and B, arrows and arrowheads) revealed that the GABA_B receptor subunit, associated with CCK⁺ cellular compartments, is up-regulated in the ipsilateral (i.e., the KA-injected side) CA3 region compared with the contralateral side (Fig. S6 A–D; GABA_B1/CCK, contra: 0.83 ± 0.02; ipsi: 1.24 ± 0.06; P < 0.001). These results indicate an increase in the immunoreactivity of GABA_B1Rs on CCK⁺ cellular compartments in the CA3 region and suggest that an up-regulation of GABA_BRs on CCK⁺ BCs underlies the stronger disinhibition during baclofen application in the epileptic hippocampus.

Modeling Results.

We examine whether effects similar to those obtained in the experiments can be observed in standard models of γ oscillations (SI Materials and Methods). Depending on the balance of drive between PCs (E-cells) and fast-spiking, PV⁺ BCs (I-cells) and the strength of the E-to-I synapses, pyramidal-interneuronal network γ (PING) or interneuonal network γ (ING) rhythms can arise (28). In the PING rhythm, the removal of a tonic inhibition to the E-cells with a reversal potential of −75 mV, modeling the GABA_AR-mediated inhibition originating from the activity of CCK⁺ BCs, leads to a stronger γ rhythm (Fig. 4 A and D): the percentage of E-cells participating in each γ cycle increases (from 34.5% to 46%), and the population γ frequency rises (from 36.5 to 42.0 Hz). Addition of a tonic inhibition with reversal potential of −90 mV to the E-cells, modeling the activation of postsynaptic GABA_BR-mediated inhibition (Fig. 4B), results in a marked reduction in γ power: the participation rate of the E-cells and the population γ frequency drop significantly (from 46% to 29.5% and from 42.0 to 27 Hz, respectively). Thus, the model reproduces the modulation of γ power by pre- and postsynaptic GABA_BR-mediated effects, but unlike the experimental results, changes in the γ power (a change in the fraction of E-cells participating in each population spike volley) are accompanied by substantial changes in the population frequency.

Fig. 4.

γ frequency oscillations in PING and ING models of the epileptic network of 160 excitatory cells (E-cells) and 40 inhibitory cells (I-cells). (A) Conductance of tonic inhibition with reversal potential −75 mV, modeling the modulated output of CCK⁺ BCs by presynaptic GABA_BRs, as a function of time. (B) Conductance of tonic inhibition with a reversal potential of −90 mV, simulating the effect of postsynaptic GABA_BR, as a function of time. (C and D) PING: Mean membrane potential of all E-cells (smoothed), as a function of time (C) and spike rastergram (D). Red dots indicate spiking of E-cells, and blue dots indicate spiking of I-cells. (E and F) Similar to C and D but with modifications in parameter values turning PING into ING.

If the rhythm is ING, the discrepancy between the experiments and the simulations is eliminated (Fig. 4 E and F). Removal of the tonic inhibition, modeling the activity of CCK⁺ BCs, still raises the population frequency, but only very slightly (from 33 to 34.5 Hz). At the same time, the percentage of E-cells participating in each population cycle rises greatly (from 27.5% to 49.5%), reflecting the increased discharge frequency of the individual cells. Addition of the tonic inhibition, modeling the activation of postsynaptic GABA_BRs, reduces the population frequency slightly (to 32 Hz) while reducing the E-cell participation rate dramatically (to 5%). The mean membrane potential changes in both models is qualitatively similar to that observed in the experiments (Fig. 4 C and E; Fig. S3A).

Discussion

Our results show that the GABA_BR agonist baclofen can augment hippocampal γ oscillations and promote pathological discharge in the CA3 network of chronic epileptic mice in a dose-dependent manner. These changes are correlated with a reduced inhibition of CA3 PCs and an enhanced presynaptic GABA_BR activation at the output of CCK⁺ BCs at low concentrations of baclofen. We propose that the proepileptic effect of low doses of baclofen is related to cell type–specific altered expression and enhanced activation of presynaptic GABA_BRs on CCK⁺ BCs, leading to a decreased GABA release and consequent disinhibition and hyperexcitability of PCs.

Although GABA_BRs are mediators of slow inhibitory effects of GABA in the brain, a series of experimental and clinical observations have shown that application of the canonical agonist baclofen can induce seizures in animal models, as well as patients (15, 16, 29, 30). To address a possible mechanism, we performed in vivo experiments in a mouse model of TLE and found that baclofen acts in a dose-dependent manner: at a high dose (10 mg/kg), baclofen reduces network excitability and oscillatory activity, consistent with its inhibitory actions. At a low dose (1 mg/kg), in contrast, it increases the power of oscillations in the γ range and the frequency of pathological population discharges. These dose-dependent effects in the in vivo recordings are consistent with the concentration-dependent action of baclofen on the excitability of CA3 PCs seen in our in vitro experiments. The hyperpolarization and reduced excitability in PCs induced by baclofen at a high concentration can be explained by the activation of postsynaptic GABA_BRs in principal cells. In contrast, the depolarization, the increase in the power of γ oscillations, and the increased discharge found at a low and in the early phase of a high concentration of baclofen applications plausibly reflects an indirect action, whereby presynaptic GABA_BRs on axon terminals of interneurons reduce the release of GABA and result in a disinhibition of PCs. The disinhibitory effect is likely to involve extrasynaptic GABA_AR-mediated tonic inhibition, known to be enhanced in principal cells in the epileptic hippocampus (31).

Pre- and postsynaptic GABA_BRs differ in their molecular composition and pharmacological properties (32–34). Receptors containing the 1a splice variant are preferentially localized presynaptically and show high sensitivity to baclofen, whereas those incorporating the GABA_B1b subunit are directed to postsynaptic, somato-dendritic compartments and show low sensitivity (34, 35). Thus, the disinhibitory effect of baclofen observed in our experiments suggests an increased function of the GABA_B1a subunit containing receptors in GABAergic terminals. Higher presynaptic autoinhibition of GABAergic transmission has been observed in another animal model in the entorhinal cortex of kindled rats (18). In contrast, however, evidence for a down-regulation of presynaptic GABA_B autoreceptors was found in the CA1 and dentate gyrus (DG) in kindled animals (36, 37), suggesting that these changes might be region and cell type specific.

A previous immunohistochemical study in KA-injected mice revealed that GABA_BRs initially undergo a rapid down-regulation in both principal cells and interneurons in the hippocampus (11). Subsequently, in the chronic phase, the receptors show a recovery and up-regulation in dentate granule cells and a subset of interneurons, but not in PCs of the CA areas (11). The lowered dendritic GABA_BR expression in PCs therefore could contribute to the dissociated pre- and postsynaptic effects in our experiments. The functional relevance of the changes in interneurons, and the precise cellular and subcellular expression pattern in diverse types of interneurons, however, remained unknown. The present in vitro results indicate an interneuron type–specific up-regulation of the receptors in GABAergic axon terminals. The fact that the disinhibitory effect of baclofen in the epileptic mice was occluded by CB1, but not by muscarinic type 2 receptor agonists, in these experiments points to the involvement of CCK⁺ interneurons but excludes PV⁺ fast-spiking cells (38, 39). Indeed, this conclusion was further corroborated by our immunocytochemical data demonstrating that GABA_B1Rs in CCK-immunopositive neurites are significant increased in the CA3 region. Thus, our results indicate that activation of presynaptic GABA_BRs by low concentrations of baclofen mutes the output of CCK⁺ BCs in the epileptic brain; however, we cannot fully exclude the involvement of other, dendritic inhibitory CCK-positive and -negative interneurons. The disinhibition caused by the functional change at the inhibitory output of CCK⁺ BCs observed here might be further exaggerated by a reduction in the number of these interneurons (40).

CCK- and PV-expressing interneurons differ with respect to their activity pattern and involvement in oscillations. PV⁺ BCs are thought to be the main pacemaker for γ oscillations (8, 21, 41, 42), whereas CCK⁺ interneurons provide out-of-phase inhibition to the principal cell population and are assumed to interfere with γ synchrony (43). Therefore, up-regulation of GABA_B autoreceptors in CCK⁺ interneurons and the consequently reduced inhibitory output is expected to produce not only disinhibition in PCs but also enhance γ oscillations. Indeed, the network simulations confirm that concentration-dependent dissociated activation of postsynaptic GABA_BRs in principal cells and presynaptic receptors in CCK⁺ interneuron axons that provide tonic inhibition can lead to enhanced γ oscillations. This effect is independent of whether the network activity is dominated by PING interactions or ING interactions (44).

Enhanced γ activity was previously observed in the KA-injected mouse model of TLE and was found to be dependent on altered intrinsic properties and the discharge pattern of dendritic inhibitory oriens lacunosum molecular interneurons (20, 45, 46). Our present data confirm and extend those findings, demonstrating that additional alterations in the GABAergic system affecting GABA_BR in CCK⁺ interneurons can promote γ activity and the emergence of pathological discharge in the epileptic hippocampal network. Enhanced γ frequency oscillations in epileptic patients are often observed at seizure onset in depth EEG recordings (5), suggesting that the high-frequency oscillations promote hypersynchronization and transition to pathological activity. During oscillations, due to the heightened synchronized activity of interneurons, GABA levels increase in the extracellular space and lead to activation of extrasynaptic GABA_BRs (10, 47, 48). Due to their higher affinity (34), first presynaptic receptors will be activated, leading to disinhibition in the epileptic network, which in a positive feedback manner will further enhance oscillations. In contrast, after prolonged activity, when GABA levels reach higher tissue levels, activation of postsynaptic GABA_BRs in principal cells dampens the oscillations and could contribute to the cessation of seizure activity. Thus, an altered distribution of the molecularly and pharmacologically distinct pre- and postsynaptic GABA_BR subtypes underlies the dual action of the receptors in the epileptic brain. Furthermore, our results explain the dose-dependent pro- and antiepileptic effects of the agonist baclofen in patients and may help to find new therapeutic agents for the treatment of this neurological disease.

Materials and Methods

Experiments were performed on adult C57/Bl6 mice. All animal procedures were approved by the Regional Berlin Animal Ethics Committee.

In Vivo Experiments.

In vivo KA injection and local field recording, as well as in vitro slice preparation and recordings, were similar to those described in Dugladze et al. (20). This information can be also found in SI Materials and Methods.

Intracellular Recordings.

Membrane potential changes and action potential firing properties of cells were measured by using intracellular recordings with sharp microelectrodes (49). Detailed information about the measurements of the synaptic properties using the whole-cell recordings can be found in SI Materials and Methods.

Immunofluorescence.

For analysis of mouse brain sections, confocal laser scanning microscopy was applied. Detailed information about the fluorescence measurements can be found in SI Materials and Methods.