α5 Subunit-containing GABAA receptors mediate a slowly decaying inhibitory synaptic current in CA1 pyramidal neurons following Schaffer collateral activation

GABAA receptors that contain the α5 subunit (α5GABAARs) are highly expressed in the hippocampus, and have been implicated in learning and memory processes. They generate a tonic form of inhibition that regulates neuronal excitability. Recently it was shown that α5GABAARs also contribute to slow phasic inhibition of CA1 pyramidal neurons following local stimulation in the stratum lacunosum moleculare. However, it is unknown whether α5GABAARs can also be recruited indirectly by stimulation of Schaffer collaterals. Here, we studied GABAergic currents evoked by stimulation in the stratum radiatum of CA1 in the presence and absence of CNQX to block AMPA receptor-mediated excitation. We tested their sensitivity to gabazine and two drugs acting at the benzodiazepine site of α1/α2/α3 or α5GABAARs (400 nM zolpidem and 20 nM L-655,708, respectively). IPSCs evoked by stimulation in the stratum radiatum in the presence of CNQX were potentiated by zolpidem, blocked by 1 μM gabazine and were relatively insensitive to L-655,708 consistent with the lack of α5GABAARs. In contrast, IPSCs evoked by stimulation of Schaffer collaterals had a significant gabazine-insensitive component. This component was attenuated by L-655,708 and enhanced by burst stimulation. Furthermore, the L-655,708-sensitive current was absent in recordings from mice lacking α5GABAARs (gabra5−/− mice). These results show that α5GABAAR-mediated phasic inhibition is activated by the Schaffer collateral pathway and provide evidence for activity pattern-dependent participation of α5GABAARs in inhibition.


Introduction
Synaptic inhibition in the hippocampus plays a crucial role in balancing and synchronising the activity of excitatory cells.
g-aminobutyric acid (GABA) released by inhibitory interneurons activates GABA A receptors (GABA A Rs), and in most mature neurons, GABA causes a reduction of the postsynaptic cell excitability via hyperpolarising and/or shunting inhibition (for review, see Mann and Paulsen, 2007). GABA A Rs are Cl À permeable, pentameric ionic channels that are formed from the combination of distinct subunits (a1e6, b1e3, g1e3, d, 3, q, p, r1e3), and the majority of native combinations identified to date have a common 2:2:1 a/b/g stoichiometry (reviewed in Wafford, 2005).
The targeting of pyramidal cells by inhibitory interneurons follows a highly organised pattern, and the vast majority of GABAergic interneurons target either the perisomatic or specific dendritic domains of pyramidal cells (Klausberger and Somogyi, 2008). However, the role of specific GABA A R subtypes expressed in distinct CA1 pyramidal cell compartments is still poorly understood. There is some evidence of a high correlation between presynaptic interneuron type and their specific GABA A Rs subunit targets (Nusser et al., 1996;Thomson et al., 2000). For example, in the neocortex, recordings from synaptically-connected pairs between GABAergic interneurons and pyramidal cells have demonstrated that dendritic targeting inhibitory neurons preferentially activate a5GABA A Rs, whereas those targeting the soma activate a1GABA A Rs (Ali and Thomson, 2008).
By identifying the GABA A R subtypes in different inhibitory pathways, it may be possible to pharmacologically target specific GABAergic networks in the hippocampus. Recent studies have started to dissect the contribution of GABA A Rs subtypes to different behaviours. Pharmacological tools in this area include the use of benzodiazepine derivatives or genetic modifications targeted at the benzodiazepine site, which is located at the interface of g2 and a subunits (reviewed in Wafford, 2005). By altering the kinetics of single GABA A R channels, benzodiazepines enhance the effect of GABA and a behavioural readout can be obtained to interpret the function of targeted a subunits. For example, mice with a point mutation in the a1GABA A R subunit (His101Arg), which rendered it insensitive to diazepam, did not display the sedative and amnestic effects of benzodiazepines Rudolph et al., 1999). Conversely, inverse agonists acting at the benzodiazepine site (Tenen and Hirsch, 1980) inhibit the effect of GABA, and have effects opposite to those of the classical benzodiazepines. Using this approach, systemic application of the a3 inverse-agonist a3IA promoted anxiety-related behaviours in rodents (Atack et al., 2005).
a5GABA A Rs are of particular interest as they are highly expressed in the adult hippocampus both at synaptic and extrasynaptic sites (Sperk et al., 1997;Sur et al., 1999), in stark contrast to low expression levels in other brain areas. Consistent with their hippocampal localisation, behavioural studies using a5 subunitspecific inverse agonists and a5 subunit knock out mice strongly implicated a5GABA A Rs in the modulation of learning and memory (Collinson et al., 2002;Atack et al., 2006;Ballard et al., 2009). Therefore, a5GABA A Rs are currently considered as relevant targets for memory blocking drugs (Martin et al., 2009) and cognitive enhancing drugs with clinical applications such as in Alzheimer's disease patients, whose a5GABA A Rs are well preserved . However, the precise mechanisms underlying the regulation of hippocampal function by a5GABA A Rs are not known.
It is well established that extrasynaptic a5GABA A Rs can mediate a large component of tonic inhibition in the hippocampus (Caraiscos et al., 2004;Scimemi et al., 2005;Glykys and Mody, 2006;Prenosil et al., 2006). In contrast, the role of a5GABA A Rs in phasic inhibition remains poorly understood. Studies comparing spontaneous and locally-evoked inhibition between mice lacking a5GABA A Rs (gabra5 À/À ) and wild type (WT) mice suggested a negligible contribution of a5GABA A Rs to phasic inhibition (Collinson et al., 2002;Glykys and Mody, 2006). Other studies have suggested that a5GABA A Rs mediate a slowly decaying component of synaptic inhibition (GABA A,slow ) (Prenosil et al., 2006;Zarnowska et al., 2009). Evoked GABA A,slow potentials have only been observed following local extracellular stimulation at or near the stratum lacunosum moleculare (SLM) of the hippocampus (Pearce, 1993;Ouardouz and Lacaille, 1997;Zarnowska et al., 2009). Thus, GABA A,slow has been proposed to modulate the activity of distal dendrites in hippocampal CA1 pyramidal neurons, and to mediate a component of synaptic inhibition activated by the direct input from the entorhinal cortex to the hippocampus at the SLM . In order to understand the underlying mechanisms of a5GABA A Rs targeting cognitive enhancing drugs, it becomes important to establish whether in addition to their SLM activation, a5GABA A Rs are recruited by CA3 output via Schaffer collateral activity. Local stimulation at the stratum radiatum (SR) in CA1 has been reported to produce fast decaying IPSCs via perisomatic targeting inhibitory cells (Ouardouz and Lacaille, 1997) mediated by a1/a2/a3GABA A Rs (Thomson et al., 2000). However, under conditions of local stimulation, excitatory synaptic transmission is usually blocked with glutamate receptor antagonists. Feed-forward inhibition requires activation of afferent fibres to interneurons which in turn release GABA onto pyramidal cells (Alger and Nicoll, 1982). The Schaffer collaterals are likely to stimulate directly or indirectly a wide variety of interneurons that would not be reached by local stimulation. In the present study, we compared locallyevoked and Schaffer collateral-stimulated inhibitory currents. To determine whether a5GAB A Rs contribute to the evoked IPSCs we used the inverse-agonist L-655,708 in rats and gabra5 À/À mice. The results show that stimulation of Schaffer collaterals can activate a slowly decaying component of GABAergic inhibition, mediated by a5 subunit-containing GABA A receptors, particularly following bursts of high-frequency stimulation of Schaffer collateral afferent input.

Methods
Animals were housed in groups with access to food and water ad libitum. The holding facilities maintained a temperature of approximately 22 C, humidity of 60e70%, and a 12-h light/dark cycle. All animal care and experimental procedures were in accordance with the UK Home Office regulations under the Animals (Scientific Procedures) Act of 1986, and the Animal Care Committee of the University of Toronto.

Tissue preparation
Parasagittal slices containing the hippocampus were obtained from male Sprague Dawley rats (supplied by Harlan, Bicester, UK), or from gabra5 À/À mice (Collinson et al., 2002) and wild type (WT) littermates ranging from postnatal day 14 to 28. Rodents were anaesthetized with 5% isofluorane until breathing slowed down to approximately one breath per second, and stimulation of the limb withdrawal reflex no longer elicited a response. After decapitation, the brain was quickly removed into ice-cold artificial cerebrospinal fluid (aCSF), containing (in mM): NaCl, 126; KCl, 2.5; NaHCO 3 , 26; CaCl 2 , 2; MgCl 2 , 2; NaH 2 PO 4 , 1.25; glucose, 10, saturated with 95% O 2 /5% CO 2 , with a final pH of 7.2e7.4. Slices were prepared at 350 mm thickness using a Leica VT1000S microtome. Slices containing the hippocampal formation were trimmed from other brain regions and were maintained and recorded at room temperature (22e27 C).

Electrophysiological recordings
After transferring a single slice to a submerged-style recording chamber, a monopolar stainless steel stimulation electrode (A-M Systems, Sequim, WA, USA) was placed into the SR of CA1 50e100 mm away from the stratum pyramidale (SP) for synaptic stimulation. Stimulation in the SR was carried out under two different conditions: firstly, to record GABA A,local , AMPA receptor-mediated excitation was blocked with CNQX while recording from a pyramidal cell. The stimulation electrode was placed approximately 100 mm lateral to the recorded cell to ensure stimulation of local interneurons. Secondly, to record inhibition elicited by the Schaffer collaterals (GABA A,SC ) CNQX was not included. The stimulation electrode was placed approximately 300 mm lateral to the recorded cell to reduce local stimulation of GABAergic neurons in addition to afferent stimulation.
Experiments were performed in voltage clamp mode. The intracellular solution contained (in mM): Gluconic acid 70; CsCl 10; NaCl 5; BAPTA free acid 10; Hepes 10; QX-314 10; GTP 0.3; Mg-ATP 4; pH was titrated to 7.25 AE 0.05 with CsOH. The estimated final Cs concentration for the intracellular solution was w120 mM. The final osmolarity was 280 AE 5 mOsmol l À1 . BAPTA was used to prevent Ca 2þ dependent changes while measuring synaptic activity at depolarised membrane potentials. QX-314 blocks GABA B receptor-mediated currents in addition to Na þ channels (Nathan et al., 1990). All voltage values were corrected for the liquid junction potential measured as 13 mV.
Whole-cell patch clamp recordings were obtained with 2e4 MU borosilicate pipettes from putative CA1 pyramidal cells identified by their location in the SP and by their shape.
The calculated E Cl at room temperature was À56 mV, and AMPA receptormediated currents reversed near 0 mV. For this reason GABAergic currents were recorded at 0 mV, both for local and Schaffer collateral stimulation, so as to isolate them from AMPA receptor-mediated currents in the latter case. For recordings at voltages other than À70 mV, a voltage step from À70 mV to the test potential started 5 s before synaptic stimulation. Whole-cell recordings were made using an Axon Multiclamp 700B amplifier (Molecular Devices, Union City, CA, USA). Recordings were low-pass filtered at 2 kHz and digitised at 20 kHz with a National Instruments A/D board (Austin, TX, USA) using Ginj 1.0 software (courtesy of Hugh P. C. Robinson) for acquisition from within Matlab (Mathworks Ltd, Natick, MA, USA). Postsynaptic currents were evoked using a stimulus isolator unit (ISO-flex, A.M.P.I. Jerusalem, Israel) which delivered pulses of 100 ms duration in current mode; stimulation intensities ranged between 20 mA and 70 mA and the computer-controlled stimulation interval was 60 s. Drugs were applied after a stable baseline of 6e10 min (<10% drift allowed). Series resistance was not compensated during recordings. Series resistance was measured before each stimulation with a 5 mV, 50 ms step pulse. Recordings were terminated if series resistance (16 AE 6 MU) changed by more than 20%.

Data analysis
Data analysis was done using Matlab. Statistical testing was done using Matlab and SPSS software. Charge transfer was calculated by integrating the current responses from 5 to 750 ms following synaptic stimulation after leak subtraction (or 5e35 ms and 50e750 ms to separate early and late components in Fig. 1Ci). For comparison across experiments, synaptic peak current values or total charge transfer were normalised relative to the mean of baseline values obtained 5 min before drug application. For statistical comparison of drug effects, the average value of the last 4 min of recording was used. Data are presented as mean AE standard error of the mean (SEM) and are displayed in two-minute bin intervals. N values refer to the number of slices recorded. Example traces are the average of 3e5 traces (Gaussian-filtered at a corner frequency of 2 kHz). Statistical significance was assessed using Student's two-sample two-tailed t-test, one-way ANOVA or repeated measures (RM) ANOVA, with Bonferroni post-hoc corrections for multiple comparisons where appropriate. P < 0.05 was considered statistically significant.

Results
To measure synaptic GABA A R-mediated currents in CA1 pyramidal neurons in hippocampal slices, single cells were voltage clamped and brief extracellular stimuli delivered in the SR. First, to activate local GABAergic currents (referred to as GABA A,local ), AMPA receptor-mediated excitation was blocked using 10 mM CNQX.
GABA A,local reversed close to E Cl at À54 AE 3 mV (n ¼ 5; data not shown; for example traces see Fig. 1Ai) and was seen as an outward current at 0 mV. At this holding potential, IPSCs had a time constant of decay (s decay ) of 163 AE 16 ms (n ¼ 12).
Next, IPSCs were elicited by stimulating the Schaffer collaterals. For these recordings, CNQX was omitted from the extracellular solution. Biphasic responses comprising both glutamatergic and GABAergic currents were observed ( Fig. 1Ai and Aii). At À70 mV, the slowly decaying inhibitory current was fully blocked by 10 mM gabazine leaving only a fast decaying AMPA current blocked by 10 mM CNQX). The excitatory component reversed near 0 mV as expected, and the inhibitory component reversed at À45 AE 1 mV (n ¼ 5; Fig. 1Aii). The isolated GABA A R-mediated component was recorded at a holding potential of 0 mV (referred to as GABA A,SC , Fig. 1Ai). The decay time was significantly longer than that for GABA A,local (GABA A,SC , s decay ¼ 277 AE 29 ms, n ¼ 10; t-test, P < 0.001). The ratio of current observed at 500 ms over 15 ms after stimulation (I 500ms /I 15ms ) showed that GABA A,local current had decayed to approximately 5% of its peak value after 500 ms, while a substantial fraction of GABA A,SC could still be observed (GABA A,local , I 500ms /I 15ms ¼ 0.05 AE 0.01, n ¼ 11; GABA A,SC , I 500ms / I 15ms ¼ 0.32 AE 0.06, n ¼ 6; t-test, P < 0.001).
Both GABA A,local and GABA A,SC decayed significantly faster at À70 mV than at 0 mV (s decay ¼ 70 AE 5 ms, n ¼ 11, and 180 AE 10 ms, n ¼ 6, respectively, t-test for both, P < 0.01). The slower decay of GABA A,local and GABA A,SC at 0 mV is consistent with previous reports showing that GABA A R currents decay more slowly at depolarised potentials as receptors unbind agonist at a slower rate (Mellor and Randall, 1998;Burgard et al., 1999). The currents observed at À70 mV were slower than those reported in some previous studies. Two factors might have contributed to the slower kinetics observed in our experiments: the use of BAPTA and the use of Cs-gluconate. Slower IPSC decay has been observed both during recording with BAPTA  and with Cs-gluconate (Stepanyuk et al., 2002).
a5GABA A Rs are widely expressed in the SR subfield of the hippocampal CA1 area (Sperk et al., 1997;Sur et al., 1999), however their synaptic contribution has not been observed while stimulating locally in the SR. We therefore asked whether a5GABA A Rs can contribute to GABA A,SC . As a first pharmacological approach we tested the gabazine sensitivity of GABA A R currents, as it has been previously reported that a tonic inhibitory conductance mediated by a5GABA A Rs is resistant to 1 mM gabazine (Bai et al., 2001;Caraiscos et al., 2004).
To compare the gabazine sensitivity of locally-evoked and Schaffer collateral-stimulated currents, IPSCs were recorded in the presence of increasing concentrations of gabazine (Fig. 1Bi). GABA A, local currents were abolished by 1 mM gabazine (initial amplitude, 863 AE 144 pA, n ¼ 11) while a significant fraction of GABA A,SC remained under these conditions (19 AE 3% of GABA A,SC of control charge transfer remaining in 1 mM gabazine; n ¼ 6, initial amplitude, 1261 AE 57 pA). This component is referred to as GABA A,SC (gz) Example traces are shown in Fig. 1Bii.
As bursting input from the Schaffer collateral could be necessary for the firing of dendritic targeting interneurons (Maccaferri and Dingledine, 2002), we next asked whether bursting activity in this pathway could enhance GABA A,SC (gz). Indeed this component became prominent with burst stimulation (3e4 stimuli at 100 Hz) as 49 AE 7% of inhibitory charge transfer remained in 1 mM gabazine (Fig. 1Ci). Remarkably, the early component (5e35 ms after stimulation) of burst responses was almost as sensitive to 1 mM gabazine as GABA A,local (fraction remaining in 1 mM gabazine: GABA A, local , 0.4 AE 1%, n ¼ 6; GABA A,SC (burst, 5e35 ms), 1 AE 5%, n ¼ 6; t-test, P ¼ 0.73; example traces in Fig. 1Cii). A significant effect of gabazine inhibition was observed for all recording conditions (GABA A,local , GABA A,SC and GABA A,SC(burst) ; RM ANOVA. Charge transfer as the between-subjects factor, and dose as within-subject factor; F 2,19 ¼ 25; P < 0.001; post-hoc comparisons showed that at 1 mM gabazine, each group was different from the others, P < 0.001).
To test whether a5GABA A Rs contribute to the inhibitory currents described above we next investigated the effect of L-655,708, which is an inverse-agonist selective for a5GABA A R subunits in rats, mice and humans at concentrations below 20 nM (Atack et al., 2006). After a stable baseline recording, 20 nM L-655,708 was added to the extracellular solution. The effects of L-655,708 on charge transfer and peak amplitude were estimated for GABA A,local , GABA A,SC and GABA A,SC (gz), with either single or burst stimulation for GABA A,SC (gz) ( Fig. 2A). In all conditions, the average charge transfer after addition of L-655,708 was significantly different from baseline, but stronger effects were observed for GABA A,SC (gz) for both single and burst stimulation (percentage of reduction: GABA A,local , 13 AE 5%, n ¼ 7; GABA A,SC , 12 AE 5%, n ¼ 9; GABA A,SC (gz) (single), 30 AE 2%, n ¼ 6; GABA A,SC (gz) (burst), 33 AE 3%,  n ¼ 7; Fig. 2B). A significant effect on peak was also observed for GABA A,SC (gz) (single) (t-test, P < 0.01), but did not reach significance for GABA A,SC (gz) (burst) (t-test, P ¼ 0.07).
To corroborate our findings on the effects of L-655,708 on GABA A R-mediated currents, and to test the specificity of L-655,708 for a5GABA A Rs, we next used gabra5 À/À mice and WT littermate control mice (Collinson et al., 2002). In this set of experiments, GABA A,SC currents from WT mice showed a significant reduction from baseline following application of 20 nM L-655,708 (GABA A,SC , 33 AE 4% reduction, n ¼ 5; GABA A,SC (gz), 36 AE 4% reduction, n ¼ 7; t-test, P < 0.001 for both), whereas L-655,708 did not significantly alter the charge transfer for GABA A,SC and GABA A,SC (gz) currents in gabra5 À/À slices (Fig. 3A, B; GABA A,SC , n ¼ 7; t-test, P ¼ 0.11; GABA A, We additionally tested whether L-655,708 could be used as a selective inhibitor of a5GABA A Rs at concentrations higher than 20 nM during our recording conditions. We found that L-655,708 at 1 mM concentration, and even at concentrations as low as 50 nM, inhibited the GABA A R current in WT mice to a similar extent to that seen in gabra5 À/À mice (no significant difference WT versus gabra5 À/À , n ¼ 7 and n ¼ 5, respectively: 50 nM, P ¼ 0.33, and 1 mM, P ¼ 0.16; Fig. 3C), suggesting that L-655,708 is not selective for a5GABA A Rs at concentrations above 20 nM, and thus limiting the selective inhibition of the a5GABA A R-mediated current to no more than w30%. Further analysis to compare GABA A,SC currents recorded from WT and gabra5 À/À mice revealed that the fraction of GABA A,SC charge transfer that persisted after the application of 1 mM gabazine was larger in WT mice than in gabra5 À/À mice (34.4 AE 4%, n ¼ 11 versus 20.6 AE 2%, n ¼ 10, respectively, t-test, P < 0.01; Fig. 4Ai, Aii, Bi). Also, the peak current was significantly different between gabra5 À/À and WT mice (t-test, P < 0.05; Fig. 4Ai, Aii, Bii).
Finally, to test whether the GABA A R currents described above are mediated in part by a1/a2/a3GABA A Rs receptors we tested the effects of 400 nM zolpidem (Fig. 5). As expected, zolpidem increased the charge transfer of GABA A,local currents (48 AE 6%, n ¼ 7). Also, both GABA A,SC and GABA A,SC (gz) charge transfer was enhanced (GABA A,SC by 33.3 AE 4%, n ¼ 5, and GABA A,SC (gz) by 16 AE 5%, n ¼ 5; Fig. 5A, B). Zolpidem had a significantly different effect on currents under all recording conditions (GABA A,local , GABA A,SC and GABA A,SC (gz); one-way ANOVA, F 2,32 ¼ 21.6, P < 0.001; post-hoc comparisons with significant differences as shown in Fig. 5Bi). A slight increase in peak amplitude from baseline values was observed in all three recording conditions with no significant difference among them (Fig. 5Bii). The large increase in charge transfer and small change in peak amplitude in response to zolpidem have been previously observed for miniature postsynaptic potentials (Perrais and Ropert, 1999). In total, our results suggest that a5GABA A Rs contribute significantly to GABA A,SC .

Discussion
Previous work investigating the subunit composition of synaptic GABA A receptors has measured spontaneous or stimulus-evoked GABAergic inhibition in the hippocampus while blocking glutamatergic excitation. In the present study, we describe the distinct pharmacology of locally-evoked versus Schaffer collateral-stimulated GABA A R currents. GABA A,local currents were eliminated by 1 mM gabazine and were markedly enhanced by 400 nM zolpidem, whereas GABA A,SC currents were less sensitive to 1 mM gabazine Fig. 3. L-655,708 (20 nM) is specific for the GABA A receptor a5 subunit. (A) Superimposed example traces of GABA A,SC (i) and GABA A,SC (gz) (ii) before and after application of 20 nM L-655,708 in wild type and gabra5 À/À mice. (B) Normalised values for GABA A R charge transfer measured during baseline and following application of L-655,708 at time 0 for GABA A,SC (i) and GABA A,SC (gz) (ii). (C) Concentration-response plot of GABA A,SC for 20, 50 and 1000 nM L-655,708 in wild type and gabra5 À/À mice. ***P < 0.001. Fig. 4. GABA A receptor-mediated currents in gabra5 À/À mice and WT mice are differentially sensitive to 1 mM gabazine. (A) Superimposed example traces of GABA A,SC currents before (grey) and after (black) application of 1 mM gabazine. Recordings from WT (i), and gabra5 À/À mice (ii). (B) Normalised values for charge transfer (i) and peak current (ii) measured during baseline and following application of 1 mM gabazine at time 0. *P < 0.05; **P < 0.01. and zolpidem. The gabazine-insensitive component of GABA A,SC showed the greatest reduction by L-655,708 and was relatively insensitive to zolpidem, suggesting a significant proportion of a5GABA A R. Furthermore, the gabazine-insensitive current was markedly enhanced following burst stimulation of Schaffer collaterals. Finally, we confirmed the specificity of L-655,708 on GABA A,SC and the gabazine-insensitive currents by recording from WT and gabra5 À/À mice.
IPSCs recorded in CA1 pyramidal neurons evoked by local stimulation at the SR were likely generated by perisomatic targeting interneurons, including basket cells (Ouardouz and Lacaille, 1997).
This GABA A,local current had a s decay of 163 ms at 0 mV. In experiments designed to study Schaffer collateral-stimulated currents, the IPSCs decayed with a strikingly slow time constant (277 ms). GABA A,local and GABA A,SC also showed significantly different timeto-peak, and 10e90% rise time. Previous studies have already shown evidence for such fast and slow components of synaptic inhibition (Pearce, 1993;Banks et al., 1998). Whereas there was no significant difference in latency between GABA A,local and GABA A,SC , suggesting that stimulation at the SR could also trigger direct activation of interneurons, GABA A,SC (gz) also showed a longer latency, suggesting that under these conditions, a delay in activation of presynaptic interneurons contributed to the longer time-to-peak.
Several possible explanations could account for the late peak and surprisingly slow decay of GABA A,SC compared to GABA A,local . Firstly, using fast perfusion on excised patches from cells expressing recombinant receptors, it has been shown that subunit composition can affect decay time (Tia et al., 1996;Burgard et al., 1999). Secondly, the time course of GABA concentration at the release sites would affect the response kinetics, for example the rise time responses for extrasynaptic receptors could be slowed down as has been suggested for glutamatergic synapses (Scimemi et al., 2004). Thirdly, long exposure to GABA, either in the synaptic cleft or in extrasynaptic space, could produce the reactivation of synaptic receptors as previously shown by modifying GABA uptake kinetics (Roepstorff and Lambert, 1994). Recent studies using somatic recordings in anatomically-identified connected cell pairs suggest that the time course of inhibitory responses can indeed be determined both by subunit composition and by distinct transmitter release transients (Szabadics et al., 2007;Ali and Thomson, 2008), but we cannot exclude the possibility that dendritic filtering contributes to the slow kinetics as recorded at the soma. A further possibility with potential physiological significance is that the slow rise and decay kinetics we observed for GABA A,SC is due to the pattern of activation of GABAergic interneurons. For example, latefiring dendritic targeting interneurons have been shown to generate slow GABAergic events in CA1 pyramidal neurons (Maccaferri and Dingledine, 2002). Furthermore, bursts of GABA release produced by several action potentials or slow asynchronous release (Hefft and Jonas, 2005) could contribute to slow kinetics. GABA A,SC and GABA A,local showed different sensitivities to gabazine. We consistently found that approximately 20% of GABA A, SC charge transfer evoked by single stimulation was insensitive to 1 mM gabazine, and this fraction increased to w50% when using burst stimulation. We interpreted this finding as an indication of a component mediated by a5GABA A Rs as it has been previously shown that the tonic current mediated by these is not sensitive to 1 mM gabazine (Bai et al., 2001;Caraiscos et al., 2004). We obtained further evidence that a5GABA A Rs are activated by Schaffer collateral stimulation by studying the effects of L-655,708 on the isolated GABA A R components described above. GABA A,SC (gz) produced either by single or burst stimulation was reduced by 30% and there was no significant difference in the amount of reduction using either type of stimulation (P ¼ 0.27) as would be expected if in both cases the synaptically evoked current is mediated by a similar proportion of a5GABA A Rs. L-655,708 also produced a significant but smaller effect on GABA A,SC and GABA A,local . In previous studies, spontaneous GABA A,local currents were found not to contain an a5GABA A R-mediated component (Caraiscos et al., 2004;Glykys and Mody, 2006;Zarnowska et al., 2009), however, it is likely that the sorting of populations of spontaneous IPSCs by their decay time for analysis could limit the detection of a slow component in GABA A,local currents. Furthermore, it is likely that the extracellular stimulation used to produce GABA A,local under our experimental conditions would recruit interneurons other than perisomatic targeting interneurons. It is also possible that gabazine, as a competitive antagonist, could be displaced from GABA A receptors by released GABA under our recording conditions for GABA A,SC (gz). Although we do not have evidence to discard this possibility, the effects of 20 nM L-655,708 on GABA A,SC in both rat and mouse recordings without the use of gabazine strongly support a specific a5GABA A R-mediated component.
We corroborated the specificity of 20 nM L-655,708 on GABA A,SC and GABA A,SC (gz) currents using gabra5 À/À mice and WT littermate controls. There was a one third reduction in the GABA A,SC and GABA A,SC (gz) currents from initial control values following application of L-655,708 in WT mice, and no significant reduction in gabra5 À/À mice. In vitro analysis of L-655,708 activity on recombinant human GABA A Rs showed a maximum inhibition of 20% of the a5GABA A R current (Atack et al., 2006). Consistent with this we found in both rat and mouse recordings that 20 nM L-655,708 inhibited approximately 30% of GABA A,SC (gz).
As a final test to probe the composition of GABA A Rs mediating the early and late components of inhibition, we used 400 nM and peak current (ii) measured during baseline and following application of zolpidem at time 0. Significant difference of charge transfers but not peak currents between baseline and the last 4 min of recording for the three recording conditions. *P < 0.05; ***P < 0.001. zolpidem which is a highly potent and selective benzodiazepine, with a5GABA A R sparing properties (a5GABA A R, Ki > 15 mM, reviewed in Sieghart, 1995). GABA A,local was markedly enhanced after application of zolpidem with a three fold increase in charge transfer compared to GABA A,SC (gz). In comparison, GABA A,SC showed only a two fold increase compared to GABA A,SC (gz). We interpret this result to suggest that local stimulation in SR near SP in the absence of fast synaptic excitation mostly activates perisomatic targeting interneurons, which selectively activate a1, a2 and a3GABA A Rs (Thomson et al., 2000), and that activation of Schaffer collaterals not only stimulates perisomatic targeting cells but also dendritic targeting interneurons, some of which specifically target a5GABA A Rs. Therefore, the combined results using 400 nM zolpidem and 20 nM L-655,708 suggest that GABA A,SC contains a significant population of a5GABA A Rs. Thus, the present observations highlight the strong activation of a5GABA A Rs following Schaffer collateral stimulation in the hippocampus and the pharmacological analysis of GABA A,SC (gz) suggests a particularly robust activation during bursting activity.
Our results are consistent with previous suggestions that slow inhibition is mediated by a specific subpopulation of interneurons and molecularly distinct receptors (Banks et al., 1998;Pearce, 1993;Zarnowska et al., 2009). The widespread localisation of a5GABA A Rs at dendritic sites (Sperk et al., 1997) suggests that they are involved in gating dendritic excitability, for example they could be involved in regulating the generation of dendritic spikes such as those observed in vivo during sharp waves (Kamondi et al., 1998). Furthermore, the time course similarity of GABA A,SC and GABA A, SC (gz) to the decay kinetics of NMDA receptors (Vicini et al., 1998) makes them well suited for inhibition of the induction of long-term potentiation by providing a shunting inhibitory effect on the NMDAR current (Staley and Mody, 1992).
The higher acquisition rate in associative memory tasks observed in mice after systemic application of the inverse-agonist L,655,708 or in gabra5 À/À mice could be the expression of plasticity produced by enhanced dendritic excitation. In order to understand the cellular and network mechanisms that lead to enhanced learning after reduction of a5GABA A Rs function it will be necessary to study the requirement for activation/silencing of inputs targeting a5GABA A Rs on hippocampal pyramidal neurons by afferent inputs to the hippocampus and their timing in relation to the timing of somatic and dendritic spikes. Here we provide evidence of activity pattern-dependent feed-forward activation of a5GABA A Rs in CA1. Testing the role of a5GABA A Rs in controlling the generation of dendritic spikes or in the fine tuning of long-term potentiation awaits the broader availability of more potent and selective drugs acting on a5GABA A Rs, such as RO4938581 (Ballard et al., 2009) or the specific activation of a5GABA A R-targeting interneurons in vivo.