Golgi Cell-Mediated Activation of Postsynaptic GABAB Receptors Induces Disinhibition of the Golgi Cell-Granule Cell Synapse in Rat Cerebellum

In the cerebellar glomerulus, GABAergic synapses formed by Golgi cells regulate excitatory transmission from mossy fibers to granule cells through feed-forward and feedback mechanisms. In acute cerebellar slices, we found that stimulating Golgi cell axons with a train of 10 impulses at 100 Hz transiently inhibited both the phasic and the tonic components of inhibitory responses recorded in granule cells. This effect was blocked by the GABAB receptor blocker CGP35348, and could be mimicked by bath-application of baclofen (30 µM). This depression of IPSCs was prevented when granule cells were dialyzed with GDPβS. Furthermore, when synaptic transmission was blocked, GABAA currents induced in granule cells by localized muscimol application were inhibited by the GABAB receptor agonist baclofen. These findings indicate that postsynaptic GABAB receptors are primarily responsible for the depression of IPSCs. This inhibition of inhibitory events results in an unexpected excitatory action by Golgi cells on granule cell targets. The reduction of Golgi cell-mediated inhibition in the cerebellar glomerulus may represent a regulatory mechanism to shift the balance between excitation and inhibition in the glomerulus during cerebellar information processing.


Introduction
The cerebellum is a brain structure important for the precise execution of motor sequences. It performs critical functions required for error or novelty detection by processing differences between predictions elaborated by the cortex and incoming stimuli conveyed by the senses.
Different parts of cerebellum fulfill distinct physiological functions. The vestibulo-cerebellum, constituted by the flocculonodular lobe and adjacent vermis, regulates equilibrium and vestibulo-ocular reflexes. The spino-cerebellum, including the vermis and the intermediate part of hemispheres, is involved in movement execution including feedback adjustments. The cerebro-cerebellum, represented by the lateral part of the cerebellar hemispheres, plays an important role in preparation, initiation and timing of motor acts via the dentate nuclei.
Cerebellar networks can be subdivided into three layers: an input (granular) layer, an intermediate processing (molecular) layer and an output (Purkinje) layer connected to the deep cerebellar nuclei. The granular layer and the molecular layer form the cortical part of the cerebellum. The deep cerebellar nuclei complex, which is part of the precerebellar nuclei, represents the only output pathway of the cerebellar cortex. The granular layer is composed of three main classes of neurons: granule cells, Golgi cells, and Lugaro cells. In the vestibular cerebellum, a fourth neuron type is represented by the unipolar brush cell (UBC). The mossy fibers make excitatory glutamatergic synapses with all these cell types. The Golgi cells make inhibitory connection to granule cells and UBCs and the UBCs inhibit Golgi cells. The granule cells send excitatory inputs to the Purkinje cells and to molecular layer interneurons. In turn, the Golgi cells provide the only inhibitory input to the granular layer, generating a complex combination of feed-forward, feedback and lateral inhibition responses.
Golgi cells are GABAergic interneurons that modulate transmission through the cerebellar glomerulus, thereby regulating the input-output relationship and the gain at the synapses between mossy fibers and granule cells [1,2]. Thus, Golgi cells do not simply inhibit granule cells. For example, an in vivo investigation in the ventral paraflocculus (VPFL) of the alert squirrel monkey has shown that Golgi cells operate as statespecific temporal filters at the mossy fiber-granule cell input during a variety of vestibular and oculomotor behaviors [3]. Furthermore, a paradoxical excitatory action has been reported at the Golgi cell -granule cell synapse mediated by presynaptic metabotropic glutamate receptors [4]. A number of in vitro studies have also characterized novel properties of Golgi cell function that challenge the classical view of their roles in regulating transmission to the cerebellar cortex. Golgi cell discharge not only evokes synaptic IPSCs but also generates pronounced tonic inhibition [5,6,7]. This tonic response reflects the activation by GABA spillover of high affinity extrasynaptic receptors containing the a6 subunit, and the accumulation of ambient GABA at submicromolar concentrations in the glomerulus [8,9,10]. In addition to the GABA A receptors expressed by granule cells, the glomerulus also contains GABA B receptors localized in the somatodendritic compartment of granule cells and on the terminals of Golgi cells [11]. The postsynaptic GABA B receptors on cerebellar granule cells have been shown to mediate inhibition of a rectifier current [12]. The GABA B receptors on Golgi cell terminals, which exhibit high affinity, are tonically activated by ambient GABA [13] resulting in a decrease in release probability at the onset of Golgi cell discharge and thus in a modulation of inhibitory signaling.
In the present study, we investigated whether phasic increases in GABA release are also capable of modulating inhibitory synaptic transmission. Following the stimulation of Golgi cell axons with a brief train of high frequency pulses (100 Hz), comparable to the frequency of in vivo activity recorded in Golgi cells [14,15,16], we observed an inhibition of both phasic and tonic GABA A receptormediated responses. Interestingly, we find that this phenomenon depends primarily on the activation of postsynaptic GABA B receptors of granule cells.

Preparation of Brain Slices and Solutions for Recordings
Patch-clamp recordings in acute cerebellar slices were obtained as previously described [12,17,18,19]. All experimental procedures were approved by the Ethics Committee of the University of Pavia. Briefly, 17-23 day-old Wistar rats were anesthetized with halothane inhalation (Aldrich, Milwaukee, WI), beheaded, and the cerebellar vermis was dissected out and placed into ice-cold Krebs solution. Vibratome sagittal sections from cerebellar vermis (220 mm thickness) were allowed to recover for at least 30 min at room temperature in Krebs solution continually bubbled with oxygen/carbon dioxide (95%/ 5%) before being transferred to a 1.5 ml recording chamber mounted on the stage of an upright microscope (OLYMPUS BX51WI, Japan). Slices were superfused with Krebs solution (flow rate 2 ml/min) and maintained at 30uC with a Peltierfeedback device (TC-324B, Warner Instr. Corp. Hamden, CT, USA). Krebs solution for slice cutting and recovery contained (in mM): NaCl 120, KCl 2, MgSO 4 1.2, NaHCO 3 26, KH 2 PO 4 1.2, CaCl 2 2, glucose 11, and was equilibrated with 95% O 2 and 5% CO 2 (pH 7.4). The n value referring to the number of the cells and, on average we recorded two cells per animal.

Whole-cell Patch-clamp Recordings
We recorded spontaneous inhibitory postsynaptic currents (sIPSCs), evoked inhibitory postsynaptic currents (eIPSCs) and I tonic from cerebellar granule cells using the blind whole-cell patchclamp recording technique [20]. Recordings were obtained from lobule VI of the cerebellar vermis.

Perforated Patch-clamp Recordings
We also recorded inhibitory currents from cerebellar granule cells using the perforated patch-clamp recording technique [21]. The recordings were obtained with 5-8 MV electrodes filled with a solution containing nystatin (100 mg ml 21 ) and (mM) K 2 SO 4 84, NaCl 10, glucose 30, HEPES buffer 5 (pH was adjusted to 7.2 with KOH). The series resistance averaged 23.664 MV.

Golgi Cell Axon Stimulation
Golgi cell axons were stimulated with a glass pipette through an isolation unit. The minimal stimulation intensity was determined to be , 8 V by comparing evoked IPSPs with spontaneous IPSPs. Experiments were then performed with suprathreshold strength by increasing the stimulation intensity by 20%.

Acquisition and Analysis
IPSCs were digitally filtered at 1.5 kHz and analyzed off-line with pCLAMP8 (Axon Instruments). Peak amplitude and decay time constants were analyzed. Series resistance was monitored by measuring passive current transients induced by 10 mV hyperpolarizing voltage steps from a holding potential of 260 mV. Cerebellar granule cells have a compact structure and behave like a single electrotonic compartment [17,18,22,23]. Accordingly, the transients were reliably fitted with a mono-exponential function yielding a membrane capacitance of C m = 4.460.3 pF, membrane resistance R m = 2.860.4 GV and series resistance R s = 13.561.3 MV (n = 29). During whole-cell recordings capacitative transients were measured to monitor potential changes in series resistance. The 23 dB cell + electrode cut-off frequency was f VC = (2pR s C m ) 21 = 3.160.2 kHz (n = 29). Accepted deviations from these parameters in current transients recorded over the time-windows used for statistical analysis were less than 10%. The arrows indicate the time of stimulation, defined as time 0 on the time scale. Experimental traces are shown without baseline adjustments: dashed lines indicate the 0 pA level. Where indicated, relative amplitude was calculated by normalizing data just before HFS application.
Data are reported as means6SEM and statistical comparisons were done using Student's paired t-test. The mean % change reported for different experimental conditions was calculated by averaging values from single cells. Traces were averaged digitally for 5 to 10 IPSCs. Averaged time courses are the means from different cells as indicated in the legends.

High Frequency Stimulation of Golgi Cell Axons Induces Transient Inhibition of Phasic and Tonic GABAergic Responses in Granule Cells
To determine whether brief discharge of Golgi cells modulates inhibitory transmission at the Golgi cell-granule cell synapse, we stimulated Golgi cell axons with an electrode positioned in the granular cell layer, close (,100 mm) to a voltage-clamped granule cell. In all experiments, inhibitory responses were pharmacologically isolated by superfusing the slice with antagonists for ionotropic (CNQX, 10 mM; D-APV, 100 mM; 7Cl-Kyn, 50 mM) and metabotropic (AIDA, 50 mM) glutamate receptors (Fig. 1). Golgi cell axons were stimulated at a basal frequency of 0.1 Hz using a suprathreshold intensity, which at holding potential of 210 mV evoked eIPSCs occurring with a delay of 2.260.16 ms, consistent with a monosynaptic response (n = 24; Fig. 2A). All evoked activity was abolished with application of 1 mM tetrodotoxin (TTX) or 10 mM bicuculline (Fig. 1C) indicating that the responses were synaptic and mediated by GABA A receptors. These results confirm the absence of slow GABA B receptor-mediated responses in phasic granule cell inhibitory currents [12,13]. Beside affecting synaptic currents, bicuculline and TTX caused an inward current shift revealing the presence of a tonic GABA A receptor-mediated leak conductance accompanied by a decrease in background noise due to reduced stochastic channel openings [9,10,24,25,26]. Local perfusion of bicuculline caused an inward current shift of 13.4 pA corresponding to a conductance change of 239.2638.6 pS (n = 8, P,0.001 paired t-test, Fig. 1B). TTX caused an inward current shift of 10.9 pA corresponding to a conductance change of 194.6632.2 pS (n = 20, P,0.00019 unpaired t-test, Fig. 1C). After TTX application, subsequent application of bicuculline (Fig. 1C) caused a non-statistically significant current shift (n = 5, Fig. 1C). Note the rare TTXinsensitive spontaneous events (Fig. 1C, inset in center panel) that were identified as inhibitory miniature synaptic currents (minis). These effects match previous reports of bicucullinesensitive tonic inhibition in granule cells, which includes a conspicuous component sustained by spontaneous Golgi cell autorhythmic activity.
After a 10 min control period we delivered a train of high frequency stimulation (HFS, Fig. 2) consisting of 10 impulses at 100 Hz elicited at 270 mV, close to the neuronal resting membrane potential, to mimic physiological conditions. Follow-ing HFS, the amplitude of eIPSCs was immediately reduced, as was outward I tonic , but these values reached their peak after 15 min ( Fig. 2A,C,D). During control conditions, the average eIPSC peak amplitude was 4164.3 pA, corresponding to a peak conductance of 732.1625.4 pS (n = 24) and I tonic was 18.763.3 pA (n = 22). The average eIPSC rise time was 1.3860.17 ms and the decay was fitted by a double exponential function with a Tau 1 of 7.161 ms and Tau 2 of 43.867 ms. Fifteen minutes after HFS, the eIPSC amplitude was 15.961.9 pA (P,0.001; Fig. 2C,D, n = 12), the peak conductance of eIPSCs was 322.1618.7 pS (P,1.8610 25 ) and outward I tonic was decreased to 6.061.2 pA (P,0.00083, n = 10; Fig. 2E,F). The kinetics of the IPSCs were not significantly changed with a rise time of 1.660.21 ms (P = 0.32) and a decay Tau 1 of 762.3 ms and a Tau 2 of 4364.5 ms. In five cells where stable recordings were maintained for over one hour, eIPSC amplitude partially recovered to 82.865.1% and the I tonic to 8063.1% of control in 4066.5 minutes. Therefore, this form of transient inhibition of GABA A -receptor mediated responses is not a form of LTD.
To exclude a possible role of Group II and Group III mGlu receptors on this form of disinhibition we repeated experiments using a cocktail of antagonists that included LY341495 (100 mM), a broad spectrum inhibitor of mGlu receptors at this concentration. After a 10 min control period we delivered a HFS (Fig. 3). Following HFS, the amplitude of eIPSCs was immediately reduced, as was the outward I tonic , but these values reached their peak after 15 min (Fig. 3A,C,D). During control conditions, the average eIPSC peak amplitude was 45.264.2 pA, and I tonic was 19.764.5 pA (n = 4). Fifteen minutes after HFS, the eIPSC amplitude was 20.163.6 pA (P,0.001; Fig. 3C,D, n = 4) and outward I tonic was decreased to 7.663.1 pA (P,0.0001, n = 4; Fig. 3E,F).The differences in the percentage decrease for both phasic and tonic current in the presence of additional LY341495 are not statistically significant in comparison with the experiments without LY341495 (P.0.3).
To confirm that the inhibition is not the result of intracellular dialysis resulting from the whole-cell recording configuration, we performed experiments using the perforated patch-clamp technique [21]. With this approach, HFS still induced a decrease in eIPSCs similar to that obtained using the conventional whole-cell recording technique, with a peak current decrease of 52.161.4%, P,0.05 and a reduction in I tonic by 29.761.1%, P,0.01 (n = 5, data not shown).

Inhibition of Phasic and Tonic GABA A Responses in Granule Cells is Mediated by GABA B Receptors
Inhibitory neurotransmission has been reported to undergo negative modulation through GABA B receptors in various brain areas [27]. When slices were pre-incubated for 30 minutes and perfused with the GABA B receptor antagonist CGP35348 (50 mM), HFS (10 pulses at 100 Hz) no longer led to a significant change in eIPSCs and I tonic with whole cell recording (Fig. 4A). In the presence of CGP35348, eIPSCs and I tonic were no longer reduced (from 32.  n = 7; data not shown), indicating that during low frequency stimulation ambient GABA does not affect the modulation of phasic and tonic responses mediated by postsynaptic GABA B receptors in granule cells [28].

Pharmacological Activation of GABA B Receptors Mimics the Effect of HFS
We next examined whether we could reproduce the GABA B receptor-dependent effects induced by HFS with pharmacological activation of GABA B receptors. Application of the GABA B receptor agonist baclofen (30 mM) for 30 minutes significantly decreased the amplitude of eIPSCs evoked at 0.1 Hz (from 28.363.8 to 11.661.7 pA, P,0.001, n = 8, Fig. 5A-D) and the I tonic (from 20.268.2 to 11.466.8 pA, P,0.01, n = 8, Fig. 5E,F). Preincubation followed by bath-application of CGP35348, a selective GABA B -receptor antagonist, prevented the reduction by baclofen of eIPSC amplitude (from 38.366.6 to 39.262.7 pA, P.0.6, n = 5) and for I tonic (from 25.562.9 to 21.364.4 pA, P.0.1, n = 5, Fig. 5B,D,F).

G Protein-dependent Mechanism of GABA A Receptor Modulation in Granule Cells
CGP35348 is an antagonist with preferential selectivity for postsynaptic GABA B receptors [29,30,31]. We therefore performed experiments to confirm that the GABA B receptors responsible for the inhibition of GABA A receptor-mediated responses are localized postsynaptically. GABA B receptor function was blocked specifically in the postsynaptic granule cell by disrupting G-protein-dependent signaling with GDPbS (1 mM), a non-hydrolyzable analog of GDP [32], introduced into cells by passive diffusion from the intracellular patch pipette. Because of the relatively fast dialysis of granule cells with patch pipettes [12,33] it was possible to assess the effect of GDPbS in the first minutes after establishing the whole cell configuration. Furthermore, when a HFS was applied after 10 minutes of dialysis, the inhibition of eIPSCs was no longer observed. The I tonic rapidly decreased in the first minute after establishing the whole cell configuration (from 22.964.8 to 7.962.6 pA, P,0.01, n = 5): therefore it was not possible to evaluate the effect of HFS (Fig. 6A). The tonic current decrease during intracellular dialysis by GDPbS before HFS is likely to reflect the blockade of G protein-sensitive ion channels that contribute to the generation of the resting potential, as described previously [34]. Introduction of GDPbS into granule cells also reduced the effects of GABA B receptor activation in response to baclofen application (from 43.867.2 to 36.863.7 pA, P.0.2, n = 4, Fig. 6E,F). Therefore, the transduction mechanism linking GABA B receptor activation to GABA A phasic current involves G-protein dependent activation in the postsynaptic membrane.

Activation of Postsynaptic GABA B Receptors is Sufficient to Inhibit GABA A Receptor-mediated Responses in Granule Cells
To verify that selective activation of postsynaptic GABA B receptors is sufficient to inhibit granule cell eIPSCs, we blocked synaptic transmission by superfusing slices with TTX (1 mM) and activated postsynaptic GABA A receptors on a granule cells by puff application of muscimol (pipette concentration: 100 mM). Again, application of baclofen (30 mM) for 5 min significantly reduced the amplitude of the chloride currents induced in granule cells (from 30.566.3 to 15.164.7 pA, P,0.008, n = 4, Fig. 7). This effect recovered after washout of baclofen (89.267.9%, n = 4).

Discussion
Our main finding is that high frequency activation of Golgi cells leads to transient inhibition of phasic and tonic GABA A receptor currents recorded in granule cells. This effect is mediated by GABA B receptors, as it is blocked by the selective antagonist CGP35348, and is mimicked by bath-application of baclofen. CGP35348 had no effect on its own, indicating that ambient GABA concentration in cerebellar slices is insufficient to contribute to the depression of inhibition. Our data indicate that the inhibition of GABA A responses is mediated primarily by postsynaptic GABA B receptors, because 1) inactivation of Gproteins in the postsynaptic granule cell significantly reduced the effect, and 2) GABA A receptor-mediated responses selectively induced in granule cells were depressed following the activation of GABA B receptors.

GABA B Receptor-mediated Long-term Modulation of Inhibitory Responses
Although synaptic plasticity was originally characterized at excitatory synapses, it has long been known that GABAergic synapses can also undergo activity-dependent modification [35,36]. Subsequent investigations identified LTP and LTD at inhibitory synapses throughout the brain, in some cases involving a presynaptic, in others, a postsynaptic mechanism of action [27]. In most of these studies it was shown that activation of pre or postsynaptic NMDA receptors initiates a complex signaling cascade culminating in the modulation of GABA A receptor function. However, at inhibitory synapses in visual cortex [37], auditory brainstem [38,39], and hippocampus [40], GABA B receptors were identified as the triggers of plasticity. In general, GABA B receptors are expressed both at presynaptic terminals as well as in the somatodendritic compartment of neurons, where they regulate myriad functions through numerous G-protein coupled signal transduction pathways [41]. Interestingly, in a previous study we observed a further example of modulation by postsynaptic GABA B receptors of cerebellar properties in which GABA B receptor action reduced a constitutive inwardly rectifying conductance in cerebellar granule cells [12].
In the cerebellar glomerulus, GABA B receptors are expressed in both presynaptic Golgi cells [11] as well as in postsynaptic granule cells [42,43]. We have previously shown that presynaptic GABA B receptors on Golgi cell terminals are tonically activated by ambient GABA resulting in a negative modulation of neurotransmitter release [13]. In contrast, in the present study we found no evidence for tonic activation of the postsynaptic GABA B receptors in granule cells. At present, the only evidence we have with regard to the transduction pathway is that the first step requires activation of a G-protein (Fig. 5). However, in a previous paper we showed that the modulation by GABA B receptors of inwardly rectifying potassium channels depends on protein phosphatase 2A [12]. Furthermore, several studies in other brain regions have shown that phosphatases [44] and kinases [45,46] regulate the function of postsynaptic GABA A receptors.
A GABAergic Paradox at the Golgi-cell-granule Cell Synapse: from Inhibition to Excitation Some unusual features are emerging from in vivo studies on cerebellar circuits during sensory information processing. Highfrequency ''bursty'' firing, a common feature in granule cells [47] leads to a decrease in Golgi cell firing through activation of mGlu2 receptors [4]. This contradicts the classical view that granule cells serve only to excite Golgi cells. The activation of metabotropic glutamatergic and metabotropic GABA B receptors with high frequency discharge thus leads to paradoxical effects at the granule cell to Golgi cell synapse [4] and at the Golgi cell to-granule cell synapse (our data), indicating more complex interactions in the cerebellar glomerulus than previously assumed.
Golgi cells reside in the granular layer and are spontaneously active thus providing the sole source of inhibition for granule cells. At low frequency background firing rates (,4 Hz), Golgi cells function like typical inhibitory interneurons to induce both phasic and tonic outward currents, whereas during high frequency firing, as occurs during sensory stimulation, spill-over in the cerebellar glomerulus and activation of GABA B receptors can inhibit phasic and tonic responses mediated by GABA A receptors. This GABA B receptor-mediated inhibition of GABAergic transmission that we have characterized is likely to be of physiological significance in regulating cerebellar circuit function. We observed the effect with a train of 10 impulses at 100 Hz, which is in keeping with in vivo responses measured in Golgi cells [3,16]. Synaptic inhibition at the mossy fiber input stage to the cerebellum is generated by short high frequency Golgi cell bursts [14,15,16], which typically consist of fast and precise responses that then slowly decrease over about 100 ms [48,49].
The switch from an inhibitory to an excitatory action occurs when a Golgi cell discharges with a high frequency burst. With low frequency discharge, a Golgi cells inhibits granule cells through both phasic and tonic inhibition leading to strong lateral inhibition. In contrast, high frequency Golgi cell discharge has two distinct effects: 1) the system acts as a high-pass filter, allowing granule cell input to be processed by Purkinje cells. This effect is described in [50], figure 5B and C. High-frequency bursts transmitted through the granular layer are not reduced by inhibition. Thus, Golgi cell activity does not prevail over mossy fiber excitation during bursting discharge. This effect is in line with in vivo observations of protracted granule cell firing in response to sustained sensory stimuli [51,52,53]. These findings indicate that Golgi cells act as an activity-dependent gate, which permits cerebellar input to Purkinje cells to bypass Golgi cell inhibition when input is strong, and 2) the system acts as spatial filter. Spatial organization can play a critical role for distributed signal processing in neuronal networks. Lateral inhibition exerted by Golgi cells at low frequency results in a patchy configuration of granule cell activity [54] with limited dispersion of activity because of inhibition through the Golgi cell loop. Consequently, granular layer responses are spatially organized so that inhibition is strongest in areas of the granular layer surrounding (or adiacent to) granule cells activated by mossy fiber stimulation. With high frequency stimulation, Golgi cell inhibition as well as lateral inhibition is decreased, suggesting a change in the spatial properties of incoming information.
Previous studies on GABA B receptor function in the cerebellar glomerulus have shown that their activation enhances excitatory transmission from mossy fibers to granule cells by inhibiting an inward rectifier in granule cells [12] and by decreasing release probability at GABAergic Golgi cell terminals [13]. Our present finding of GABA B receptor-mediated disinhibition constitutes a third mechanism to promote glomerular excitatory transmission in the cerebellum.