Feed-forward inhibition shapes the spike output of cerebellar Purkinje cells
1 Wolfson Institute for Biomedical Research and Department of Physiology, University College London, Gower Street, London WC1E 6BT, UK
2 Max-Planck Institute for Neurobiology, Am Klopferspitz 18, 82152 Martinsried, Germany
Abstract
Although the cerebellum is thought to play a key role in timing of movements on the time scale of milliseconds, it is unclear how such temporal fidelity is ensured at the cellular level. We have investigated the timing of feed-forward inhibition onto interneurons and Purkinje cells activated by parallel fibre stimulation in slices of cerebellar cortex from P18–25 rats. Feed-forward inhibition was activated within 1 ms after the onset of excitation in both cell types. The rapid onset of feed-forward inhibition sharply curtailed EPSPs and increased the precision of the resulting action potentials. The time window for summation of EPSPs was reduced to 1–2 ms in the presence of feed-forward inhibition, which could inhibit the efficacy of asynchronous EPSPs for up to 30 ms. Our findings demonstrate how the inhibitory microcircuitry of the cerebellar cortex orchestrates synaptic integration and precise timing of spikes in Purkinje cells, enabling them to act as coincidence detectors of parallel fibre input.
, 百拇医药
Introduction
The cerebellum plays a central role in the timing and coordination of movements. Damage to the cerebellum can produce increased temporal variability of voluntary movement, and can also affect non-motor tasks that require precise representation of temporal information (Ivry, 1997; Spencer et al. 2003). This suggests that the cerebellar cortex must be capable of implementing timing information on the time scale of milliseconds to seconds (Ivry, 1997; Ivry & Spencer, 2004). However, it is still largely unknown how temporal information is processed within the cerebellar cortical network. An early hypothesis interpreted the long parallel fibre axons as a ‘delay line’ which could generate sequences of movements by sequentially timing the activation of Purkinje cells along a parallel fibre ‘beam’ (Braitenberg & Atwood, 1958; Meek, 1992). More recent work has implicated dynamic synchrony in the climbing fibre input to Purkinje cells as a timing system (Welsh & Llinás, 1997). An alternative suggestion is that the Golgi cell–granule cell loop may generate rhythmic synchronous activity in the cerebellar network, with information about the timing of movements encoded by the timing of action potentials in relation to the oscillation (Vos et al. 1999).
, 百拇医药
All theories of timing involving the cerebellum require that the output of the cerebellar cortex, encoded in the axons of the Purkinje cells, can be precisely timed in response to sensory stimulation. Although Purkinje cells can respond precisely to sensory stimulation acting via granule cells (Bower & Woolston, 1983), the way in which this is achieved is not clear. The classical work of Eccles and colleagues showed that activation of a beam of parallel fibres is rapidly followed by synaptic inhibition (Eccles et al. 1967). Recent anatomical and physiological work indicates that molecular layer interneurons project to Purkinje cells and interneurons that lie on the activated parallel fibre beam (‘on-beam’; Pouzat & Hestrin, 1997; Sultan & Bower, 1998) which suggests that a Purkinje cell may be both directly excited and then inhibited via molecular layer interneurons activated by the same set of active parallel fibres (feed-forward inhibition; FFI). Both the excitatory input to Purkinje cells and molecular layer interneurons from parallel fibres and the inhibitory input to Purkinje cells from interneurons are well characterized (Perkel et al. 1990; Llano et al. 1991; Barbour, 1993; Vincent & Marty, 1996; Clark & Cull-Candy, 2002; Isope & Barbour, 2002). Modelling studies, supported by dynamic clamp experiments, have also shown using asynchronous (i.e. temporally uncorrelated) activation of excitatory and inhibitory synapses that inhibition can play an important role in timing action potentials in Purkinje cells (Jaeger et al. 1997; Jaeger & Bower, 1999) and interneurons (Carter & Regehr, 2002). However, much less is known about the precise temporal interaction of correlated excitatory and inhibitory inputs in Purkinje cells and molecular layer interneurons during parallel fibre activation. Given the importance of feed-forward inhibitory pathways in regulating the timing of neuronal responses in cortical and hippocampal pyramidal cells (Buzsaki, 1984; Pouille & Scanziani, 2001), and the finding that activity in single interneurons can influence spike timing in Purkinje cells (Husser & Clark, 1997) and interneurons (Husser & Clark, 1997; Carter & Regehr, 2002), we directly investigated the precision of spiking in Purkinje cells activated by parallel fibre input and its regulation by feed-forward inhibition.
, http://www.100md.com
Methods
We recorded from visually identified neurones in 300 μm thick slices of the cerebellar vermis. Sprague-Dawley rats (18–25 days old, or 56–60 days old) were deeply anaesthetized via isoflurane inhalation and decapitated. All procedures conformed with the UK Animals (Scientific Procedures) Act 1986. Slices were cut in the coronal plane using a Leica vibrotome (Leica VT 1000S, Nussloch, Germany) to preserve the parallel fibres. Slices were continuously perfused with artificial cerebrospinal fluid (ACSF) containing (mM): 125 NaCl, 25 NaHCO3, 25 glucose, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 1 MgCl2; bubbled with 95% O2–5% CO2. Slices were viewed with infrared differential interference contrast optics on an upright microscope (Axioskop, Zeiss). All experiments were performed at 34.2 ± 0.2°C.
, http://www.100md.com
Electrophysiological recordings
Electrophysiological recordings from Purkinje cells and molecular layer interneurons were made using a Multiclamp 700A amplifier (Axon Instruments, Union City, CA, USA). For whole-cell voltage-clamp recordings we used 2–4 M pipettes and obtained series resistances of 4–12 M. The electrode solution contained (mM): 115 methanesulphonic acid, 10 Hepes, 10 Cs4BAPTA, 5 QX-314, 2 Na2ATP, 2 MgATP, 0.3 Na2GTP, 1 KCl, 0.5 CaCl2; pH 7.3 with CsOH. For whole-cell current-clamp recordings, pipettes of similar resistance were filled with solution containing (mM): 130 methanesulphonic acid, 10 Hepes, 7 KCl, 2 Na2ATP, 2 MgATP, 0.4 Na2GTP, 0.05 EGTA, biocytin (0.4%); pH 7.3 with KOH. Cell-attached patch-clamp recordings were performed with electrodes containing ACSF. Liquid junction potentials were not corrected for. Parallel fibres were stimulated with pipettes containing ACSF placed at a distance > 150 μm from the recorded cell to avoid direct stimulation of interneuron axons. In some experiments EPSPs were simulated by injection of biexponential currents (rise 0.3–1 ms, decay 3–7 ms), with the time course adjusted to generate EPSPs that matched evoked PF EPSPs recorded in the same cells in the presence of SR95531 (Sigma). Data were low-pass filtered at 4–10 kHz and sampled at 20–50 kHz using a 1321A Digidata A/D converter (Axon Instruments).
, http://www.100md.com
Analysis
Data were analysed offline using Igor Pro (Wavemetrics, Lake Oswego, OR, USA). EPSP integrals were calculated from the stimulus artifact until the voltage stably relaxed to baseline. Undershooting phases contributed negatively and thus reduced the total integral. For determining changes in EPSP integral, S.E.M. and statistical significance were calculated from the normalized absolute changes of the individual cells. Cell-attached spikes were detected using a threshold-crossing algorithm. To determine the number of extra spikes evoked by an input, poststimulus time histograms (PSTH) were computed and integrated (Fetz & Gustafsson, 1983). A linear fit to the baseline range of the integral (< 200 ms) was extrapolated over the entire duration and subtracted from the integral to yield the corrected cumulative spike probability. The number of additional evoked spikes was estimated by averaging over a 100–200 ms period following stimulation. Temporal dispersion in first spike latencies (spike jitter) was determined by fitting a latency histogram to a Gaussian of the form:
, 百拇医药
The width is the product of the standard deviation and , and was used as a measure of spike jitter.
For quantification of EPSP summation experiments we measured the residual amplitude of EPSP 1 contributing to EPSP 2 by subtracting the amplitude of EPSP 2 alone from the combined amplitude and dividing by the amplitude of the EPSP 1. Inputs were considered to be stimulated independently at an interval > 600 ms. Pooled data are expressed as means ± S.E.M. and statistical significance was determined using Student's paired t test (unless otherwise indicated).
, 百拇医药
Results
Feed-forward inhibitory synaptic currents activated by parallel fibres
To investigate FFI in the cerebellar cortex, we used a coronal slice orientation to preserve PFs, allowing stimulation of PFs at distances up to 500 μm from the recorded cell. PF stimulation evoked a sequence of inward current followed by outward current in Purkinje cells (Fig. 1A; cf. Delaney & Jahr, 2002) and molecular layer interneurons (Fig. 1B) when the cell was voltage clamped at a holding potential between the IPSC and EPSC reversal potentials. The outward current was abolished by the selective GABAA receptor antagonist SR95531 (10 μM; Hamann et al. 1988) in both cell types (Fig. 1A and B; n = 4 Purkinje cells and interneurons). The AMPA-type glutamate receptor antagonist NBQX (6 μM) abolished both the inward current and the outward current in Purkinje cells (n = 10) and interneurons (n = 8). These results confirm that the inhibitory component represents FFI, and rule out direct stimulation of interneurons. The simplest microcircuit consistent with these results is shown in Fig. 1E, illustrating the direct activation of Purkinje cells and interneurons by the PFs, as well as the indirect, disynaptic pathway for activation of FFI.
, http://www.100md.com
Purkinje cells and interneurons show sequences of inward–outward currents in response to PF stimulation. A and B, the GABAA antagonist SR95531 (10 μM) reversibly blocks the outward current in Purkinje cells and interneurons. C and D, the AMPA-type glutamate receptor antagonist NBQX (6 μM) reversibly blocks both inward and outward currents. E, schematic drawing of the microcircuit underlying feed-forward inhibition. PC, Purkinje cell; IN, interneuron.
Timing of feed-forward IPSCs
, 百拇医药
In order to precisely determine the relative timing of excitatory and inhibitory inputs, EPSCs and feed-forward IPSCs were separated by voltage clamping at their respective reversal potentials. PF stimulation strength was adjusted so that only relatively small EPSCs/IPSCs were recruited (Purkinje: EPSC –89 ± 16 pA; IPSC 356 ± 50 pA; interneuron: EPSC –150 ± 33 pA; IPSC 115 ± 21 pA), corresponding to only a few unitary synaptic inputs (Pouzat & Hestrin, 1997; Carter & Regehr, 2002; Isope & Barbour, 2002). Feed-forward IPSCs had rapid rise times (10–90% rise time in Purkinje cells: 2.2 ± 0.3 ms; interneurons: 1.0 ± 0.2 ms). We used the respective 10% rise time points to determine the EPSC–IPSC delay, which was 1.4 ± 0.2 ms (range 0.7–2.5 ms, n = 11) for Purkinje cells (Fig. 2A and B) and 1.4 ± 0.1 ms (range 1.0–1.8 ms, n = 11, P > 0.4) for interneurons (Fig. 2C and D). To test whether this delay is similar in older animals, we performed control experiments in Purkinje cells in slices from P56 to P60 rats. In this experiment the delay between EPSC and IPSC (1.4 ± 0.1 ms, n = 3) was indistinguishable from the delay in P18 to P25 Purkinje cells (P > 0.5).
, http://www.100md.com
A and C, inhibitory and excitatory synaptic currents were separated by voltage clamping the cell to the reversal potential of the IPSC (Purkinje cell: –60 ± 2 mV, interneuron: –60 ± 1 mV) or EPSC (Purkinje cell: 6 ± 2 mV, interneuron: 4 ± 1 mV). The peak of the averaged currents was measured and the 10% level was detected (red lines). B and D, currents are shown on an expanded time scale. The mean delay between the 10% rise time point of the EPSC and feed-forward IPSC was 1.4 ± 0.2 ms for Purkinje cells and 1.4 ± 0.1 ms for interneurons (red symbols); individual data points from different cells are shown as open circles, mean values as red bars. Note that the 10% level more accurately reflected the delay of the onset of currents than the peak since it minimizes the contribution of differences in EPSC and IPSC rise times.
, 百拇医药
Feed-forward inhibition shapes subthreshold synaptic potentials
The rapid onset of FFI suggests that it will significantly influence the time course of EPSPs triggered by PF activation. We investigated the effect of FFI on PF EPSP time course with current clamp recordings from Purkinje cells. We injected hyperpolarizing current to avoid spontaneous spiking, holding the cells at membrane potentials between –60 and –65 mV. PF stimulation evoked brief EPSPs (half-width: 13.1 ± 2.6 ms) that were often followed by a pronounced hyperpolarization (Figs 3A and 6A). Blocking inhibition with SR95531 more than doubled EPSP amplitude (from 1.4 ± 0.2 mV to 3.2 ± 0.6 mV in SR95531; P < 0.01, n = 26; Fig. 3B), led to a substantial increase in EPSP half-width (from 13.1 ± 2.6 ms to 43.7 ± 3.1 ms in SR95531; P < 0.01, n = 26) and increased the EPSP integral by an order of magnitude (from 15 ± 8 mV ms to 141 ± 33 mV ms; P < 0.01, n = 26). The effect of SR95531 on EPSP amplitude and time course was much larger for evoked EPSPs than for EPSPs simulated by somatic injection of currents shaped like EPSCs in the same cells (change in evoked EPSP amplitude 1.9 ± 0.7, half-width 3.4 ± 1.7; integral 6.4 ± 4.0 times greater than for injected EPSPs; n = 3). This indicates that the effect of SR on evoked PF EPSPs in these experiments is mediated primarily via block of FFI rather than via an effect on ‘tonic’ background inhibition (Husser & Clark, 1997).
, http://www.100md.com
Subthreshold voltage responses to parallel fibre stimulation were recorded in Purkinje cells hyperpolarized to –60 to –65 mV by steady current injection to prevent spontaneous spiking. A, Purkinje cells often displayed very brief EPSPs in response to PF stimulation in control conditions. In this example, blocking inhibition with SR95531 increased the amplitude (2.9–6.7 mV), halfwidth (13.2–67.7 ms) and integral (–88–235 mV ms) of the PF EPSP. B, changes in EPSP amplitude, half-width and integral following block of inhibition (all changes significant; P < 0.01). Individual cells are shown in grey, average values in black. C, interneuron EPSPs were more rapid and, when inhibition was blocked, showed smaller changes. In this cell, blocking inhibition increased EPSP amplitude from 9.2 to 11.6 mV, half-width from 13.8 to 18.5 ms and integral from 197 to 294 mV*ms. Inset shows the same traces on an expanded time scale (scale bar 2 mV, 10 ms). D, increases in EPSP amplitude, half-width and integral in all cells following block of inhibition (all changes significant; P < 0.01). E, summary of changes in EPSP properties caused by blocking inhibition in Purkinje cells and interneurons. The dotted line indicates no change; *P < 0.05; **P < 0.01, unpaired t test.
, 百拇医药
A, two independent parallel fibre inputs with strong feed-forward inhibitory components were activated separately (input 1, blue trace; input 2, red trace) or with delays of 0, 1, 2, 5, 10, 20, 30, 40, 50, 60 or 80 ms (black traces). Traces are averages of 20–30 sweeps. Stimulus artefacts and the 1 ms delay traces were omitted for visual clarity. Sweeps were truncated 5 ms after stimulus 2. The peak EPSP amplitude resulting from activation of input 2 and the linear sum of both EPSPs are indicated (red and black dashed lines, respectively). Note that the response to the second input is only enhanced by the preceding input when activated with a delay of 0–2 ms. B, blocking inhibition substantially lengthened this narrow window for EPSP summation by prolonging the time course of the EPSPs. C, EPSPs are not increased by preceding inputs unless they fall within a 2 ms window. Blocking inhibition widens that window to about 40 ms. Asterisks indicate significance (P < 0.05, n = 6).
, http://www.100md.com
Similar experiments were performed on interneurons, which showed consistent but smaller changes to those observed in Purkinje cells (Fig. 3C). EPSP amplitude (4.6 ± 0.9 mV to 6.1 ± 1 mV), half-width (10.5 ± 2.2 ms to 13 ± 2.1 ms) and integral (68 ± 22 mV ms to 120 ± 36 mV ms, Fig. 3D) all increased significantly (P < 0.01, n = 15) in the presence of SR95531. These changes were significantly smaller in interneurons than in Purkinje cells (Fig. 3E; P < 0.05).
Feed-forward inhibition regulates the level and precision of spike output
, http://www.100md.com
Because FFI dramatically shortens PF EPSPs in Purkinje cells, we expected that this would be reflected in the number and timing of action potentials (APs) evoked by PF stimulation. We tested this hypothesis using non-invasive cell-attached recordings. Because Purkinje cells generate spontaneous APs (Husser & Clark, 1997), PF stimulation led to a brief increase in spike probability on top of the baseline firing rate (Fig. 4A and B). To measure the net spike output in response to the PF input we integrated the resulting PSTH to yield the cumulative spike probability (Fetz & Gustafsson, 1983), which we then corrected for the spontaneous spike rate (see Methods; Fig. 4C). Blocking inhibition with SR95531 strongly increased the spike output as evident in raw traces, the PSTH and the corrected cumulative spike probability (Fig. 4A–C, right panels). On average, the number of additional spikes evoked in control conditions nearly tripled when inhibition was blocked (from 0.38 ± 0.09 spikes to 0.99 ± 0.18 spikes; P < 0.01, n = 21).
, http://www.100md.com
A, cell-attached recording from a Purkinje cell, showing spontaneous spikes and the short-latency spikes observed in response to parallel fibre stimulation. The number of spikes observed shortly after stimulation increased when SR was applied (right panel). B, probabilities calculated from 200 sweeps in control (left) and SR95531 (right). C, to measure the number of spikes evoked by the stimulation we calculated the cumulative sum, which was corrected for spontaneous firing rate. To estimate the additional evoked spikes, we averaged over a 100 ms window (red line) starting 100 ms after stimulation. In the cell shown parallel fibre stimulation evoked 0.9 extra spikes over baseline, increasing to 2.5 spikes when inhibition was blocked. Note the different time scale in B and C.
, http://www.100md.com
The shortening of EPSPs by FFI (Fig. 3) suggests that spikes triggered by these inputs will fall into a narrow time window defined by FFI. In cell-attached recordings from Purkinje cells, this could indeed be observed in control conditions, but not when inhibition was blocked (Fig. 5A). We fitted a Gaussian to the spike latency distribution to provide a measure for the latency of the spike response (location of peak) and for the temporal dispersion (half-width) of the spike response (Fig. 5B). In the example cell shown in Fig. 5, blocking inhibition with SR95531 only modestly increased the mean latency of the first spike from 3.6 to 4 ms, but doubled the half-width of the Gaussian fit from 0.7 to 1.4 ms. On average, latency increased from 3.4 ± 0.1 to 3.7 ± 0.2 ms (P < 0.01, n = 21) whereas spike jitter almost doubled, going from 0.96 ± 0.1 ms to 1.6 ± 0.3 ms (P < 0.01, n = 21) when inhibition was blocked.
, 百拇医药
A, cell-attached recording (20 superimposed sweeps) from a spontaneously firing Purkinje cell showing the response to parallel fibre stimulation. The first evoked spike occurred with a short latency (3.37 ± 0.08 ms) in a narrow time window after the stimulation (left panel). Blocking inhibition widened the window in which evoked spikes occurred (right panel). B, a Gaussian was fitted to the latency histogram and the width was taken as a measure of spike jitter. Blocking inhibition widened the latency distribution from 0.7 to 1.4 ms in the example shown.
, 百拇医药
Feed-forward inhibition shortens the time window for summation of independent input pathways
In the behaving animal neurones are thought to receive a constant barrage of synaptic inputs (Destexhe et al. 2003). Since FFI curtails PF EPSPs (Fig. 3), we investigated how summation of EPSPs is affected by strong FFI. We stimulated two PF inputs with separate stimulation electrodes and tested independence of both excitatory inputs by checking for cross-facilitation (40–80 ms interval; Perkel et al. 1990). Inputs that showed such cross-facilitation were excluded. In half of the cells tested the decay of the second potential was slower when activated shortly after the first (Fig. 6A, see also Fig. 4 in Brunel et al. 2004). This could reflect partial overlap between the populations of interneurons mediating FFI activated by the two inputs, or it could be a consequence of FFI in the inhibitory network.
, 百拇医药
In the cell shown in Fig. 6, synchronous activation of both inputs led to a peak EPSP amplitude that equalled the linear sum of the individual EPSP amplitudes. When introducing a brief delay between the two stimuli, the membrane potential reached at the peak of the second EPSP was less depolarized than that reached when the second EPSP was activated on its own (red trace and dotted line). This reduction tracked the time course of the undershooting inhibitory component of the first input and lasted for about 30 ms. When FFI was blocked with SR95531, the second EPSP summated with the first, following its time course over a window of about 50 ms (Fig. 6B). On average, effective EPSP summation only took place when both inputs were activated in a narrow window of about 1–2 ms. With delays of 5 ms or longer no summation of EPSPs was observed (Fig. 6C). This was significantly different from the very effective EPSP summation when inhibition was blocked (P < 0.05, n = 6; Fig. 6C).
, http://www.100md.com
Feed-forward inhibition reduces spike generation by asynchronous inputs
We tested whether the efficacy of a PF input in triggering spike output can be reduced by FFI engaged by previously active PFs. To assess this issue, we made cell-attached recordings from Purkinje cells and used two separate stimulation electrodes to activate independent PF inputs. Input size was adjusted such that a mean maximum spike probability of 28 ± 6% (bin = 1 ms; n = 5) was achieved. After collection of cell-attached spike data the recording pipette was removed and the cell was re-patched in whole cell mode to confirm the independence of the two inputs by checking for cross-facilitation of the resulting EPSCs (40–80 ms interval; Perkel et al. 1990). Inputs that showed such cross-facilitation were excluded. The spike generation efficacy of a PF input was strongly reduced if it was preceded by an independent PF input (Fig. 7A). The suppression of the spike response to the second input lasted 50 ms, similar to the time course of the inhibitory component of the EPSP–IPSP sequence (Fig. 6). When FFI was blocked with SR95531, little or no inhibition of the second input was observed (Fig. 7B). We compared across cells by calculating the average spike probability within a 10 ms window after each stimulus. The spike responses for delays up to 30 ms were significantly smaller when inhibition was intact (P < 0.05, n = 5; Fig. 7C). These results show that FFI evoked by a PF input can penalize succeeding PF inputs in terms of their efficacy to generate Purkinje cell output.
, 百拇医药
A, the spike probabilities in response to separate stimulation of two independent parallel fibre inputs are shown (response to stimulus 1, blue bins; response to stimulus 2, red bins). Spike probabilities in response to stimulus 2 delivered at delays of 5–80 ms from stimulus 1 are also shown (black traces). Note that the spike probabilities in response to stimulus 2 are suppressed for delays up to 50 ms. Probabilities are normalized to response 2 alone (red dotted line). An exponential (grey trace) is fitted to the peak probabilities (red dots) and constrained to 100%. B, blocking inhibition leads to a faster recovery. C, mean spike probability over 10 ms after stimulus 2 is significantly smaller for delays up to 40 ms if inhibition is intact (P < 0.05, n = 5).
, 百拇医药
Discussion
We have provided the first quantitative investigation of FFI triggered by PF activation in the cerebellum. Our results demonstrate that FFI is triggered extremely rapidly by PF input, generating a very narrow time window over which PF inputs summate to generate Purkinje cell output. These findings indicate that Purkinje cells can act as coincidence detectors for PF input, and suggest that the temporal pattern of PF activation will be critical for regulating the output of the cerebellar cortex.
, http://www.100md.com
A microcircuit for precisely timed feed-forward inhibition
We have demonstrated directly that FFI is activated by PF input in just over 1 ms on average, with delays of less than 1 ms observed in some neurones. This is even faster than the delay that has been observed for FFI currents in hippocampal pyramidal neurones (1.9 ± 0.2 ms; Pouille & Scanziani, 2001). The FFI delay was indistinguishable for interneurons and Purkinje cells, indicating that it arises from a hard-wired microcircuit common to both cell types (see Fig. 1E). In order to be activated so rapidly, the elements of this microcircuit must be optimized for speed. Thus, PF EPSC kinetics in molecular layer interneurons are extremely rapid (Fig. 2; Llano & Gerschenfeld, 1993; Carter et al. 2002; Clark & Cull-Candy, 2002), and single PF inputs can reliably trigger spike outputs in interneurons (Barbour, 1993; Carter & Regehr, 2002) with submillisecond delays (Carter & Regehr, 2002; Suter & Jaeger, 2004). Interneurons have rapid membrane time constants (Husser & Clark, 1997) and brief EPSPs (this study; Carter & Regehr, 2002) which are normally not prolonged by postsynaptic NMDA receptor activation (Clark & Cull-Candy, 2002). The short interneuron axon collaterals making contacts with neighbouring Purkinje cells and interneurons (Palay & Chan-Palay, 1974) help to minimize axonal delay associated with FFI activation. Finally, the rapid kinetics of IPSCs in Purkinje cells (Vincent et al. 1992) help ensure a rapid onset for FFI. Taken together, these considerations indicate that both the anatomical and physiological properties of the FFI microcircuit are tuned for maximum speed.
, 百拇医药
Feed-forward inhibition regulates spike output of Purkinje cells
The rapid onset of FFI curtails the PF EPSP in Purkinje cells to a fraction of its original duration. We have shown directly that the shape of the resulting biphasic EPSP–IPSP waveform, originally reported by Eccles and colleagues (Eccles et al. 1967) in vivo and also observed in voltage-sensitive dye signals from the molecular layer of isolated cerebellum (Cohen & Yarom, 1998), can be explained by the kinetics and latencies of the underlying EPSC and feed-forward IPSCs. The shortening of EPSP time course by FFI has several important consequences. First, it enforces precisely timed output spikes in response to a single synchronous input. Our results show that FFI allows Purkinje cell output in response to PF stimulation to be timed to a precision of less than a millisecond, similarly to the effect of FFI in hippocampal pyramidal cells (Pouille & Scanziani, 2001). Second, it limits the time window for summation of multiple asynchronous inputs to only a few milliseconds (see also Brunel et al. 2004), indicating that the time window for synaptic integration in Purkinje cells can be much shorter than the membrane time constant in these neurones (Husser & Clark, 1997). Furthermore, inputs falling outside this time window are penalized, and have a lower-than-normal efficacy of spike generation. This provides a mechanism to further enhance the signal-to-noise of precisely timed inputs at the expense of inaccurately or asynchronously timed ones.
, 百拇医药
In contrast to pyramidal cells, cerebellar Purkinje cells show spontaneous activity, often at high rates, both in vitro (Husser & Clark, 1997) and in vivo (Armstrong & Rawson, 1979). As a consequence, both excitation and feed-forward inhibition are superimposed on a background of activity, and can act not only by initiating individual spikes, but also by accelerating or delaying spontaneous spikes. The proximity to threshold also means that even small inputs can influence spike output (Husser & Clark, 1997; Carter & Regehr, 2002; Suter & Jaeger, 2004). We have measured the effect of FFI on subsequent spike generation to be on the order of 30 ms. Therefore, even small PF inputs which do not on their own trigger spikes can influence spike output by reducing the efficacy of subsequent inputs, or by preventing spontaneous spikes. In a situation where the Purkinje cell receives hundreds or thousands of PF inputs, FFI could mediate ‘gain control’ of asynchronous ‘tonic’ excitation (Marr, 1969), whereas synchronous inputs could overcome FFI and result in fast-rising EPSPs that trigger spikes with high temporal precision due to subsequent FFI.
, http://www.100md.com
Mutual feed-forward inhibition in the inhibitory network
Molecular layer interneurons are highly interconnected (Husser & Clark, 1997; Kondo & Marty, 1998), and single spikes in presynaptic interneurons can powerfully regulate spike timing in postsynaptic interneurons (Husser & Clark, 1997; Carter & Regehr, 2002). We have demonstrated that molecular layer interneurons, in addition to generating FFI onto Purkinje cells, also receive FFI from each other. Interestingly, the level and effect of FFI in interneurons seems much weaker than that observed in Purkinje cells, measured both by the apparent underlying inhibitory conductance relative to that of excitation and by the effect of FFI on the EPSP waveform. This could either reflect differences in the relative size of unitary PF and inhibitory conductances in Purkinje cells and interneurons, or alternatively PFs could activate fewer feed-forward interneurons onto interneurons than onto Purkinje cells.
, 百拇医药
What is the functional role of this FFI onto interneurons One possibility may be to ensure the brevity of FFI onto Purkinje cells by shortening the excitation of interneurons. The time course of the FFI IPSC in Purkinje cells (Fig. 2A) is comparable to that of IPSCs evoked by direct electrical stimulation (Vincent et al. 1992), suggesting that activation of interneurons contributing to the FFI is rapidly terminated. This could occur via a combination of FFI onto interneurons themselves and the prominent reset caused by the AHP following single interneuron spikes (Carter & Regehr, 2002). The fact that PF synapses onto both interneurons and Purkinje cells facilitate (Perkel et al. 1990; Clark & Cull-Candy, 2002) suggests that FFI can keep pace with PF activation even during bursts of activity (Chadderton et al. 2004).
, http://www.100md.com
Functional implications for cerebellar processing
Several lines of evidence have converged to support the idea that the cerebellum is involved in representation of precise temporal information (Ivry & Spencer, 2004). We have now identified FFI as a key cellular mechanism that can help to preserve timing of cerebellar output to within millisecond precision. These properties may help Purkinje cells act as coincidence detectors responding only to well-timed PF inputs. Furthermore, the rapid onset of FFI may have special relevance for the functional architecture of the PF network. There has been considerable controversy regarding the role of PF ‘beams’ in cerebellar processing, with little direct evidence available for beam-like activation of Purkinje cells in the intact cerebellar cortex (Cohen & Yarom, 1998; Jaeger, 2003). Our results may provide new insights into this debate. Experimental and simulation studies have suggested that there may be considerable asynchrony of propagation across multiple PFs due to variability in fibre diameter (Bernard & Axelrad, 1991; Garwicz & Andersson, 1992). Since interneurons can be activated by only one or a few inputs (Barbour, 1993; Carter & Regehr, 2002), the extremely rapid onset of FFI and its long aftereffects suggest that the early arriving PF inputs can recruit FFI that will dampen the response to later ones, making the ‘on-beam’ response of the Purkinje cells weak or undetectable (Bower, 2002). Thus, regulation of the kinetics and strength of FFI can have important consequences for the spatial spread of activity patterns across the cerebellar cortex.
, 百拇医药
Footnotes
W. Mittmann and U. Koch contributed equally to this study.
References
Armstrong DM & Rawson JA (1979). Activity patterns of cerebellar cortical neurones and climbing fibre afferents in the awake cat. J Physiol 289, 425–448.
Barbour B (1993). Synaptic currents evoked in Purkinje cells by stimulating individual granule cells. Neuron 11, 759–769.
, 百拇医药
Bernard C & Axelrad H (1991). Propagation of parallel fiber volleys in the cerebellar cortex: a computer simulation. Brain Res 565, 195–208.
Bower JM (2002). The organization of cerebellar cortical circuitry revisited: implications for function. Ann N Y Acad Sci 978, 135–155.
Bower JM & Woolston DC (1983). Congruence of spatial organization of tactile projections to granule cell and Purkinje cell layers of cerebellar hemispheres of the albino rat: vertical organization of cerebellar cortex. J Neurophysiol 49, 745–766.
, http://www.100md.com
Braitenberg V & Atwood RP (1958). Morphological observations on the cerebellar cortex. J Comp Neurol 109, 1–33.
Brunel N, Hakim V, Isope P, Nadal JP & Barbour B (2004). Optimal information storage and the distribution of synaptic weights: perceptron versus Purkinje cell. Neuron 43, 745–757.
Buzsaki G (1984). Feed-forward inhibition in the hippocampal formation. Prog Neurobiol 22, 131–153.
, 百拇医药
Carter AG & Regehr WG (2002). Quantal events shape cerebellar interneuron firing. Nat Neurosci 5, 1309–1318.
Carter AG, Vogt KE, Foster KA & Regehr WG (2002). Assessing the role of calcium-induced calcium release in short-term presynaptic plasticity at excitatory central synapses. J Neurosci 22, 21–28.
Chadderton P, Margrie TW & Hausser M (2004). Integration of quanta in cerebellar granule cells during sensory processing. Nature 428, 856–860.
, 百拇医药
Clark BA & Cull-Candy SG (2002). Activity-dependent recruitment of extrasynaptic NMDA receptor activation at an AMPA receptor-only synapse. J Neurosci 22, 4428–4436.
Cohen D & Yarom Y (1998). Patches of synchronized activity in the cerebellar cortex evoked by mossy-fiber stimulation: questioning the role of parallel fibers. Proc Natl Acad Sci U S A 95, 15032–15036.
Delaney AJ & Jahr CE (2002). Kainate receptors differentially regulate release at two parallel fiber synapses. Neuron 36, 475–482.
, 百拇医药
Destexhe A, Rudolph M & Pare D (2003). The high- conductance state of neocortical neurons in vivo. Nat Rev Neurosci 4, 739–751.
Eccles JC, Ito M & Szenthágothai J (1967). The Cerebellum as a Neuronal Machine. Springer-Verlag, Berlin.
Fetz EE & Gustafsson B (1983). Relation between shapes of post-synaptic potentials and changes in firing probability of cat motoneurones. J Physiol 341, 387–410.
, 百拇医药
Garwicz M & Andersson G (1992). Spread of synaptic activity along parallel fibres in cat cerebellar anterior lobe. Exp Brain Res 88, 615–622.
Hamann M, Desarmenien M, Desaulles E, Bader MF & Feltz P (1988). Quantitative evaluation of the properties of a pyridazinyl GABA derivative (SR 95531) as a GABAA competitive antagonist. An electrophysiological approach. Brain Res 442, 287–296.
Husser M & Clark BA (1997). Tonic synaptic inhibition modulates neuronal output pattern and spatiotemporal synaptic integration. Neuron 19, 665–678.
, 百拇医药
Isope P & Barbour B (2002). Properties of unitary granule cell–Purkinje cell synapses in adult rat cerebellar slices. J Neurosci 22, 9668–9678.
Ivry R (1997). Cerebellar timing systems. Int Rev Neurobiol 41, 555–573.
Ivry RB & Spencer RM (2004). The neural representation of time. Curr Opin Neurobiol 14, 225–232.
Jaeger D (2003). No parallel fiber volleys in the cerebellar cortex: evidence from cross-correlation analysis between Purkinje cells in a computer model and in recordings from anesthetized rats. J Comput Neurosci 14, 311–327.
, 百拇医药
Jaeger D & Bower JM (1999). Synaptic control of spiking in cerebellar Purkinje cells: dynamic current clamp based on model conductances. J Neurosci 19, 6090–6101.
Jaeger D, De Schutter E & Bower JM (1997). The role of synaptic and voltage-gated currents in the control of Purkinje cell spiking: a modeling study. J Neurosci 17, 91–106.
Kondo S & Marty A (1998). Synaptic currents at individual connections among stellate cells in rat cerebellar slices. J Physiol 509, 221–232.
, 百拇医药
Llano I & Gerschenfeld HM (1993). Inhibitory synaptic currents in stellate cells of rat cerebellar slices. J Physiol 468, 177–200.
Llano I, Marty A, Armstrong CM & Konnerth A (1991). Synaptic- and agonist-induced excitatory currents of Purkinje cells in rat cerebellar slices. J Physiol 434, 183–213.
Marr D (1969). A theory of cerebellar cortex. J Physiol 202, 437–470.
Meek J (1992). Why run parallel fibers parallel Teleostean Purkinje cells as possible coincidence detectors, in a timing device subserving spatial coding of temporal differences. Neuroscience 48, 249–283.
, 百拇医药
Palay SL & Chan-Palay V (1974). Cerebellar Cortex: Cytology and Organization. Springer-Verlag, Berlin.
Perkel DJ, Hestrin S, Sah P & Nicoll RA (1990). Excitatory synaptic currents in Purkinje cells. Proc R Soc Lond B Biol Sci 241, 116–121.
Pouille F & Scanziani M (2001). Enforcement of temporal fidelity in pyramidal cells by somatic feed-forward inhibition. Science 293, 1159–1163.
, 百拇医药 Pouzat C & Hestrin S (1997). Developmental regulation of basket/stellate cell–Purkinje cell synapses in the cerebellum. J Neurosci 17, 9104–9112.
Spencer RM, Zelaznik HN, Diedrichsen J & Ivry RB (2003). Disrupted timing of discontinuous but not continuous movements by cerebellar lesions. Science 300, 1437–1439.
Sultan F & Bower JM (1998). Quantitative Golgi study of the rat cerebellar molecular layer interneurons using principal component analysis. J Comp Neurol 393, 353–373.
, 百拇医药
Suter KJ & Jaeger D (2004). Reliable control of spike rate and spike timing by rapid input transients in cerebellar stellate cells. Neuroscience 124, 305–317.
Vincent P, Armstrong CM & Marty A (1992). Inhibitory synaptic currents in rat cerebellar Purkinje cells: modulation by postsynaptic depolarization. J Physiol 456, 453–471.
Vincent P & Marty A (1996). Fluctuations of inhibitory postsynaptic currents in Purkinje cells from rat cerebellar slices. J Physiol 494, 183–199.
Vos BP, Maex R, Volny-Luraghi A & De Schutter E (1999). Parallel fibers synchronize spontaneous activity in cerebellar Golgi cells. J Neurosci RC6, 1–5.
Welsh JP & Llinás R (1997). Some organizing principles for the control of movement based on olivocerebellar physiology. Prog Brain Res 114, 449–461., 百拇医药(Wolfgang Mittmann, Ursula)
2 Max-Planck Institute for Neurobiology, Am Klopferspitz 18, 82152 Martinsried, Germany
Abstract
Although the cerebellum is thought to play a key role in timing of movements on the time scale of milliseconds, it is unclear how such temporal fidelity is ensured at the cellular level. We have investigated the timing of feed-forward inhibition onto interneurons and Purkinje cells activated by parallel fibre stimulation in slices of cerebellar cortex from P18–25 rats. Feed-forward inhibition was activated within 1 ms after the onset of excitation in both cell types. The rapid onset of feed-forward inhibition sharply curtailed EPSPs and increased the precision of the resulting action potentials. The time window for summation of EPSPs was reduced to 1–2 ms in the presence of feed-forward inhibition, which could inhibit the efficacy of asynchronous EPSPs for up to 30 ms. Our findings demonstrate how the inhibitory microcircuitry of the cerebellar cortex orchestrates synaptic integration and precise timing of spikes in Purkinje cells, enabling them to act as coincidence detectors of parallel fibre input.
, 百拇医药
Introduction
The cerebellum plays a central role in the timing and coordination of movements. Damage to the cerebellum can produce increased temporal variability of voluntary movement, and can also affect non-motor tasks that require precise representation of temporal information (Ivry, 1997; Spencer et al. 2003). This suggests that the cerebellar cortex must be capable of implementing timing information on the time scale of milliseconds to seconds (Ivry, 1997; Ivry & Spencer, 2004). However, it is still largely unknown how temporal information is processed within the cerebellar cortical network. An early hypothesis interpreted the long parallel fibre axons as a ‘delay line’ which could generate sequences of movements by sequentially timing the activation of Purkinje cells along a parallel fibre ‘beam’ (Braitenberg & Atwood, 1958; Meek, 1992). More recent work has implicated dynamic synchrony in the climbing fibre input to Purkinje cells as a timing system (Welsh & Llinás, 1997). An alternative suggestion is that the Golgi cell–granule cell loop may generate rhythmic synchronous activity in the cerebellar network, with information about the timing of movements encoded by the timing of action potentials in relation to the oscillation (Vos et al. 1999).
, 百拇医药
All theories of timing involving the cerebellum require that the output of the cerebellar cortex, encoded in the axons of the Purkinje cells, can be precisely timed in response to sensory stimulation. Although Purkinje cells can respond precisely to sensory stimulation acting via granule cells (Bower & Woolston, 1983), the way in which this is achieved is not clear. The classical work of Eccles and colleagues showed that activation of a beam of parallel fibres is rapidly followed by synaptic inhibition (Eccles et al. 1967). Recent anatomical and physiological work indicates that molecular layer interneurons project to Purkinje cells and interneurons that lie on the activated parallel fibre beam (‘on-beam’; Pouzat & Hestrin, 1997; Sultan & Bower, 1998) which suggests that a Purkinje cell may be both directly excited and then inhibited via molecular layer interneurons activated by the same set of active parallel fibres (feed-forward inhibition; FFI). Both the excitatory input to Purkinje cells and molecular layer interneurons from parallel fibres and the inhibitory input to Purkinje cells from interneurons are well characterized (Perkel et al. 1990; Llano et al. 1991; Barbour, 1993; Vincent & Marty, 1996; Clark & Cull-Candy, 2002; Isope & Barbour, 2002). Modelling studies, supported by dynamic clamp experiments, have also shown using asynchronous (i.e. temporally uncorrelated) activation of excitatory and inhibitory synapses that inhibition can play an important role in timing action potentials in Purkinje cells (Jaeger et al. 1997; Jaeger & Bower, 1999) and interneurons (Carter & Regehr, 2002). However, much less is known about the precise temporal interaction of correlated excitatory and inhibitory inputs in Purkinje cells and molecular layer interneurons during parallel fibre activation. Given the importance of feed-forward inhibitory pathways in regulating the timing of neuronal responses in cortical and hippocampal pyramidal cells (Buzsaki, 1984; Pouille & Scanziani, 2001), and the finding that activity in single interneurons can influence spike timing in Purkinje cells (Husser & Clark, 1997) and interneurons (Husser & Clark, 1997; Carter & Regehr, 2002), we directly investigated the precision of spiking in Purkinje cells activated by parallel fibre input and its regulation by feed-forward inhibition.
, http://www.100md.com
Methods
We recorded from visually identified neurones in 300 μm thick slices of the cerebellar vermis. Sprague-Dawley rats (18–25 days old, or 56–60 days old) were deeply anaesthetized via isoflurane inhalation and decapitated. All procedures conformed with the UK Animals (Scientific Procedures) Act 1986. Slices were cut in the coronal plane using a Leica vibrotome (Leica VT 1000S, Nussloch, Germany) to preserve the parallel fibres. Slices were continuously perfused with artificial cerebrospinal fluid (ACSF) containing (mM): 125 NaCl, 25 NaHCO3, 25 glucose, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 1 MgCl2; bubbled with 95% O2–5% CO2. Slices were viewed with infrared differential interference contrast optics on an upright microscope (Axioskop, Zeiss). All experiments were performed at 34.2 ± 0.2°C.
, http://www.100md.com
Electrophysiological recordings
Electrophysiological recordings from Purkinje cells and molecular layer interneurons were made using a Multiclamp 700A amplifier (Axon Instruments, Union City, CA, USA). For whole-cell voltage-clamp recordings we used 2–4 M pipettes and obtained series resistances of 4–12 M. The electrode solution contained (mM): 115 methanesulphonic acid, 10 Hepes, 10 Cs4BAPTA, 5 QX-314, 2 Na2ATP, 2 MgATP, 0.3 Na2GTP, 1 KCl, 0.5 CaCl2; pH 7.3 with CsOH. For whole-cell current-clamp recordings, pipettes of similar resistance were filled with solution containing (mM): 130 methanesulphonic acid, 10 Hepes, 7 KCl, 2 Na2ATP, 2 MgATP, 0.4 Na2GTP, 0.05 EGTA, biocytin (0.4%); pH 7.3 with KOH. Cell-attached patch-clamp recordings were performed with electrodes containing ACSF. Liquid junction potentials were not corrected for. Parallel fibres were stimulated with pipettes containing ACSF placed at a distance > 150 μm from the recorded cell to avoid direct stimulation of interneuron axons. In some experiments EPSPs were simulated by injection of biexponential currents (rise 0.3–1 ms, decay 3–7 ms), with the time course adjusted to generate EPSPs that matched evoked PF EPSPs recorded in the same cells in the presence of SR95531 (Sigma). Data were low-pass filtered at 4–10 kHz and sampled at 20–50 kHz using a 1321A Digidata A/D converter (Axon Instruments).
, http://www.100md.com
Analysis
Data were analysed offline using Igor Pro (Wavemetrics, Lake Oswego, OR, USA). EPSP integrals were calculated from the stimulus artifact until the voltage stably relaxed to baseline. Undershooting phases contributed negatively and thus reduced the total integral. For determining changes in EPSP integral, S.E.M. and statistical significance were calculated from the normalized absolute changes of the individual cells. Cell-attached spikes were detected using a threshold-crossing algorithm. To determine the number of extra spikes evoked by an input, poststimulus time histograms (PSTH) were computed and integrated (Fetz & Gustafsson, 1983). A linear fit to the baseline range of the integral (< 200 ms) was extrapolated over the entire duration and subtracted from the integral to yield the corrected cumulative spike probability. The number of additional evoked spikes was estimated by averaging over a 100–200 ms period following stimulation. Temporal dispersion in first spike latencies (spike jitter) was determined by fitting a latency histogram to a Gaussian of the form:
, 百拇医药
The width is the product of the standard deviation and , and was used as a measure of spike jitter.
For quantification of EPSP summation experiments we measured the residual amplitude of EPSP 1 contributing to EPSP 2 by subtracting the amplitude of EPSP 2 alone from the combined amplitude and dividing by the amplitude of the EPSP 1. Inputs were considered to be stimulated independently at an interval > 600 ms. Pooled data are expressed as means ± S.E.M. and statistical significance was determined using Student's paired t test (unless otherwise indicated).
, 百拇医药
Results
Feed-forward inhibitory synaptic currents activated by parallel fibres
To investigate FFI in the cerebellar cortex, we used a coronal slice orientation to preserve PFs, allowing stimulation of PFs at distances up to 500 μm from the recorded cell. PF stimulation evoked a sequence of inward current followed by outward current in Purkinje cells (Fig. 1A; cf. Delaney & Jahr, 2002) and molecular layer interneurons (Fig. 1B) when the cell was voltage clamped at a holding potential between the IPSC and EPSC reversal potentials. The outward current was abolished by the selective GABAA receptor antagonist SR95531 (10 μM; Hamann et al. 1988) in both cell types (Fig. 1A and B; n = 4 Purkinje cells and interneurons). The AMPA-type glutamate receptor antagonist NBQX (6 μM) abolished both the inward current and the outward current in Purkinje cells (n = 10) and interneurons (n = 8). These results confirm that the inhibitory component represents FFI, and rule out direct stimulation of interneurons. The simplest microcircuit consistent with these results is shown in Fig. 1E, illustrating the direct activation of Purkinje cells and interneurons by the PFs, as well as the indirect, disynaptic pathway for activation of FFI.
, http://www.100md.com
Purkinje cells and interneurons show sequences of inward–outward currents in response to PF stimulation. A and B, the GABAA antagonist SR95531 (10 μM) reversibly blocks the outward current in Purkinje cells and interneurons. C and D, the AMPA-type glutamate receptor antagonist NBQX (6 μM) reversibly blocks both inward and outward currents. E, schematic drawing of the microcircuit underlying feed-forward inhibition. PC, Purkinje cell; IN, interneuron.
Timing of feed-forward IPSCs
, 百拇医药
In order to precisely determine the relative timing of excitatory and inhibitory inputs, EPSCs and feed-forward IPSCs were separated by voltage clamping at their respective reversal potentials. PF stimulation strength was adjusted so that only relatively small EPSCs/IPSCs were recruited (Purkinje: EPSC –89 ± 16 pA; IPSC 356 ± 50 pA; interneuron: EPSC –150 ± 33 pA; IPSC 115 ± 21 pA), corresponding to only a few unitary synaptic inputs (Pouzat & Hestrin, 1997; Carter & Regehr, 2002; Isope & Barbour, 2002). Feed-forward IPSCs had rapid rise times (10–90% rise time in Purkinje cells: 2.2 ± 0.3 ms; interneurons: 1.0 ± 0.2 ms). We used the respective 10% rise time points to determine the EPSC–IPSC delay, which was 1.4 ± 0.2 ms (range 0.7–2.5 ms, n = 11) for Purkinje cells (Fig. 2A and B) and 1.4 ± 0.1 ms (range 1.0–1.8 ms, n = 11, P > 0.4) for interneurons (Fig. 2C and D). To test whether this delay is similar in older animals, we performed control experiments in Purkinje cells in slices from P56 to P60 rats. In this experiment the delay between EPSC and IPSC (1.4 ± 0.1 ms, n = 3) was indistinguishable from the delay in P18 to P25 Purkinje cells (P > 0.5).
, http://www.100md.com
A and C, inhibitory and excitatory synaptic currents were separated by voltage clamping the cell to the reversal potential of the IPSC (Purkinje cell: –60 ± 2 mV, interneuron: –60 ± 1 mV) or EPSC (Purkinje cell: 6 ± 2 mV, interneuron: 4 ± 1 mV). The peak of the averaged currents was measured and the 10% level was detected (red lines). B and D, currents are shown on an expanded time scale. The mean delay between the 10% rise time point of the EPSC and feed-forward IPSC was 1.4 ± 0.2 ms for Purkinje cells and 1.4 ± 0.1 ms for interneurons (red symbols); individual data points from different cells are shown as open circles, mean values as red bars. Note that the 10% level more accurately reflected the delay of the onset of currents than the peak since it minimizes the contribution of differences in EPSC and IPSC rise times.
, 百拇医药
Feed-forward inhibition shapes subthreshold synaptic potentials
The rapid onset of FFI suggests that it will significantly influence the time course of EPSPs triggered by PF activation. We investigated the effect of FFI on PF EPSP time course with current clamp recordings from Purkinje cells. We injected hyperpolarizing current to avoid spontaneous spiking, holding the cells at membrane potentials between –60 and –65 mV. PF stimulation evoked brief EPSPs (half-width: 13.1 ± 2.6 ms) that were often followed by a pronounced hyperpolarization (Figs 3A and 6A). Blocking inhibition with SR95531 more than doubled EPSP amplitude (from 1.4 ± 0.2 mV to 3.2 ± 0.6 mV in SR95531; P < 0.01, n = 26; Fig. 3B), led to a substantial increase in EPSP half-width (from 13.1 ± 2.6 ms to 43.7 ± 3.1 ms in SR95531; P < 0.01, n = 26) and increased the EPSP integral by an order of magnitude (from 15 ± 8 mV ms to 141 ± 33 mV ms; P < 0.01, n = 26). The effect of SR95531 on EPSP amplitude and time course was much larger for evoked EPSPs than for EPSPs simulated by somatic injection of currents shaped like EPSCs in the same cells (change in evoked EPSP amplitude 1.9 ± 0.7, half-width 3.4 ± 1.7; integral 6.4 ± 4.0 times greater than for injected EPSPs; n = 3). This indicates that the effect of SR on evoked PF EPSPs in these experiments is mediated primarily via block of FFI rather than via an effect on ‘tonic’ background inhibition (Husser & Clark, 1997).
, http://www.100md.com
Subthreshold voltage responses to parallel fibre stimulation were recorded in Purkinje cells hyperpolarized to –60 to –65 mV by steady current injection to prevent spontaneous spiking. A, Purkinje cells often displayed very brief EPSPs in response to PF stimulation in control conditions. In this example, blocking inhibition with SR95531 increased the amplitude (2.9–6.7 mV), halfwidth (13.2–67.7 ms) and integral (–88–235 mV ms) of the PF EPSP. B, changes in EPSP amplitude, half-width and integral following block of inhibition (all changes significant; P < 0.01). Individual cells are shown in grey, average values in black. C, interneuron EPSPs were more rapid and, when inhibition was blocked, showed smaller changes. In this cell, blocking inhibition increased EPSP amplitude from 9.2 to 11.6 mV, half-width from 13.8 to 18.5 ms and integral from 197 to 294 mV*ms. Inset shows the same traces on an expanded time scale (scale bar 2 mV, 10 ms). D, increases in EPSP amplitude, half-width and integral in all cells following block of inhibition (all changes significant; P < 0.01). E, summary of changes in EPSP properties caused by blocking inhibition in Purkinje cells and interneurons. The dotted line indicates no change; *P < 0.05; **P < 0.01, unpaired t test.
, 百拇医药
A, two independent parallel fibre inputs with strong feed-forward inhibitory components were activated separately (input 1, blue trace; input 2, red trace) or with delays of 0, 1, 2, 5, 10, 20, 30, 40, 50, 60 or 80 ms (black traces). Traces are averages of 20–30 sweeps. Stimulus artefacts and the 1 ms delay traces were omitted for visual clarity. Sweeps were truncated 5 ms after stimulus 2. The peak EPSP amplitude resulting from activation of input 2 and the linear sum of both EPSPs are indicated (red and black dashed lines, respectively). Note that the response to the second input is only enhanced by the preceding input when activated with a delay of 0–2 ms. B, blocking inhibition substantially lengthened this narrow window for EPSP summation by prolonging the time course of the EPSPs. C, EPSPs are not increased by preceding inputs unless they fall within a 2 ms window. Blocking inhibition widens that window to about 40 ms. Asterisks indicate significance (P < 0.05, n = 6).
, http://www.100md.com
Similar experiments were performed on interneurons, which showed consistent but smaller changes to those observed in Purkinje cells (Fig. 3C). EPSP amplitude (4.6 ± 0.9 mV to 6.1 ± 1 mV), half-width (10.5 ± 2.2 ms to 13 ± 2.1 ms) and integral (68 ± 22 mV ms to 120 ± 36 mV ms, Fig. 3D) all increased significantly (P < 0.01, n = 15) in the presence of SR95531. These changes were significantly smaller in interneurons than in Purkinje cells (Fig. 3E; P < 0.05).
Feed-forward inhibition regulates the level and precision of spike output
, http://www.100md.com
Because FFI dramatically shortens PF EPSPs in Purkinje cells, we expected that this would be reflected in the number and timing of action potentials (APs) evoked by PF stimulation. We tested this hypothesis using non-invasive cell-attached recordings. Because Purkinje cells generate spontaneous APs (Husser & Clark, 1997), PF stimulation led to a brief increase in spike probability on top of the baseline firing rate (Fig. 4A and B). To measure the net spike output in response to the PF input we integrated the resulting PSTH to yield the cumulative spike probability (Fetz & Gustafsson, 1983), which we then corrected for the spontaneous spike rate (see Methods; Fig. 4C). Blocking inhibition with SR95531 strongly increased the spike output as evident in raw traces, the PSTH and the corrected cumulative spike probability (Fig. 4A–C, right panels). On average, the number of additional spikes evoked in control conditions nearly tripled when inhibition was blocked (from 0.38 ± 0.09 spikes to 0.99 ± 0.18 spikes; P < 0.01, n = 21).
, http://www.100md.com
A, cell-attached recording from a Purkinje cell, showing spontaneous spikes and the short-latency spikes observed in response to parallel fibre stimulation. The number of spikes observed shortly after stimulation increased when SR was applied (right panel). B, probabilities calculated from 200 sweeps in control (left) and SR95531 (right). C, to measure the number of spikes evoked by the stimulation we calculated the cumulative sum, which was corrected for spontaneous firing rate. To estimate the additional evoked spikes, we averaged over a 100 ms window (red line) starting 100 ms after stimulation. In the cell shown parallel fibre stimulation evoked 0.9 extra spikes over baseline, increasing to 2.5 spikes when inhibition was blocked. Note the different time scale in B and C.
, http://www.100md.com
The shortening of EPSPs by FFI (Fig. 3) suggests that spikes triggered by these inputs will fall into a narrow time window defined by FFI. In cell-attached recordings from Purkinje cells, this could indeed be observed in control conditions, but not when inhibition was blocked (Fig. 5A). We fitted a Gaussian to the spike latency distribution to provide a measure for the latency of the spike response (location of peak) and for the temporal dispersion (half-width) of the spike response (Fig. 5B). In the example cell shown in Fig. 5, blocking inhibition with SR95531 only modestly increased the mean latency of the first spike from 3.6 to 4 ms, but doubled the half-width of the Gaussian fit from 0.7 to 1.4 ms. On average, latency increased from 3.4 ± 0.1 to 3.7 ± 0.2 ms (P < 0.01, n = 21) whereas spike jitter almost doubled, going from 0.96 ± 0.1 ms to 1.6 ± 0.3 ms (P < 0.01, n = 21) when inhibition was blocked.
, 百拇医药
A, cell-attached recording (20 superimposed sweeps) from a spontaneously firing Purkinje cell showing the response to parallel fibre stimulation. The first evoked spike occurred with a short latency (3.37 ± 0.08 ms) in a narrow time window after the stimulation (left panel). Blocking inhibition widened the window in which evoked spikes occurred (right panel). B, a Gaussian was fitted to the latency histogram and the width was taken as a measure of spike jitter. Blocking inhibition widened the latency distribution from 0.7 to 1.4 ms in the example shown.
, 百拇医药
Feed-forward inhibition shortens the time window for summation of independent input pathways
In the behaving animal neurones are thought to receive a constant barrage of synaptic inputs (Destexhe et al. 2003). Since FFI curtails PF EPSPs (Fig. 3), we investigated how summation of EPSPs is affected by strong FFI. We stimulated two PF inputs with separate stimulation electrodes and tested independence of both excitatory inputs by checking for cross-facilitation (40–80 ms interval; Perkel et al. 1990). Inputs that showed such cross-facilitation were excluded. In half of the cells tested the decay of the second potential was slower when activated shortly after the first (Fig. 6A, see also Fig. 4 in Brunel et al. 2004). This could reflect partial overlap between the populations of interneurons mediating FFI activated by the two inputs, or it could be a consequence of FFI in the inhibitory network.
, 百拇医药
In the cell shown in Fig. 6, synchronous activation of both inputs led to a peak EPSP amplitude that equalled the linear sum of the individual EPSP amplitudes. When introducing a brief delay between the two stimuli, the membrane potential reached at the peak of the second EPSP was less depolarized than that reached when the second EPSP was activated on its own (red trace and dotted line). This reduction tracked the time course of the undershooting inhibitory component of the first input and lasted for about 30 ms. When FFI was blocked with SR95531, the second EPSP summated with the first, following its time course over a window of about 50 ms (Fig. 6B). On average, effective EPSP summation only took place when both inputs were activated in a narrow window of about 1–2 ms. With delays of 5 ms or longer no summation of EPSPs was observed (Fig. 6C). This was significantly different from the very effective EPSP summation when inhibition was blocked (P < 0.05, n = 6; Fig. 6C).
, http://www.100md.com
Feed-forward inhibition reduces spike generation by asynchronous inputs
We tested whether the efficacy of a PF input in triggering spike output can be reduced by FFI engaged by previously active PFs. To assess this issue, we made cell-attached recordings from Purkinje cells and used two separate stimulation electrodes to activate independent PF inputs. Input size was adjusted such that a mean maximum spike probability of 28 ± 6% (bin = 1 ms; n = 5) was achieved. After collection of cell-attached spike data the recording pipette was removed and the cell was re-patched in whole cell mode to confirm the independence of the two inputs by checking for cross-facilitation of the resulting EPSCs (40–80 ms interval; Perkel et al. 1990). Inputs that showed such cross-facilitation were excluded. The spike generation efficacy of a PF input was strongly reduced if it was preceded by an independent PF input (Fig. 7A). The suppression of the spike response to the second input lasted 50 ms, similar to the time course of the inhibitory component of the EPSP–IPSP sequence (Fig. 6). When FFI was blocked with SR95531, little or no inhibition of the second input was observed (Fig. 7B). We compared across cells by calculating the average spike probability within a 10 ms window after each stimulus. The spike responses for delays up to 30 ms were significantly smaller when inhibition was intact (P < 0.05, n = 5; Fig. 7C). These results show that FFI evoked by a PF input can penalize succeeding PF inputs in terms of their efficacy to generate Purkinje cell output.
, 百拇医药
A, the spike probabilities in response to separate stimulation of two independent parallel fibre inputs are shown (response to stimulus 1, blue bins; response to stimulus 2, red bins). Spike probabilities in response to stimulus 2 delivered at delays of 5–80 ms from stimulus 1 are also shown (black traces). Note that the spike probabilities in response to stimulus 2 are suppressed for delays up to 50 ms. Probabilities are normalized to response 2 alone (red dotted line). An exponential (grey trace) is fitted to the peak probabilities (red dots) and constrained to 100%. B, blocking inhibition leads to a faster recovery. C, mean spike probability over 10 ms after stimulus 2 is significantly smaller for delays up to 40 ms if inhibition is intact (P < 0.05, n = 5).
, 百拇医药
Discussion
We have provided the first quantitative investigation of FFI triggered by PF activation in the cerebellum. Our results demonstrate that FFI is triggered extremely rapidly by PF input, generating a very narrow time window over which PF inputs summate to generate Purkinje cell output. These findings indicate that Purkinje cells can act as coincidence detectors for PF input, and suggest that the temporal pattern of PF activation will be critical for regulating the output of the cerebellar cortex.
, http://www.100md.com
A microcircuit for precisely timed feed-forward inhibition
We have demonstrated directly that FFI is activated by PF input in just over 1 ms on average, with delays of less than 1 ms observed in some neurones. This is even faster than the delay that has been observed for FFI currents in hippocampal pyramidal neurones (1.9 ± 0.2 ms; Pouille & Scanziani, 2001). The FFI delay was indistinguishable for interneurons and Purkinje cells, indicating that it arises from a hard-wired microcircuit common to both cell types (see Fig. 1E). In order to be activated so rapidly, the elements of this microcircuit must be optimized for speed. Thus, PF EPSC kinetics in molecular layer interneurons are extremely rapid (Fig. 2; Llano & Gerschenfeld, 1993; Carter et al. 2002; Clark & Cull-Candy, 2002), and single PF inputs can reliably trigger spike outputs in interneurons (Barbour, 1993; Carter & Regehr, 2002) with submillisecond delays (Carter & Regehr, 2002; Suter & Jaeger, 2004). Interneurons have rapid membrane time constants (Husser & Clark, 1997) and brief EPSPs (this study; Carter & Regehr, 2002) which are normally not prolonged by postsynaptic NMDA receptor activation (Clark & Cull-Candy, 2002). The short interneuron axon collaterals making contacts with neighbouring Purkinje cells and interneurons (Palay & Chan-Palay, 1974) help to minimize axonal delay associated with FFI activation. Finally, the rapid kinetics of IPSCs in Purkinje cells (Vincent et al. 1992) help ensure a rapid onset for FFI. Taken together, these considerations indicate that both the anatomical and physiological properties of the FFI microcircuit are tuned for maximum speed.
, 百拇医药
Feed-forward inhibition regulates spike output of Purkinje cells
The rapid onset of FFI curtails the PF EPSP in Purkinje cells to a fraction of its original duration. We have shown directly that the shape of the resulting biphasic EPSP–IPSP waveform, originally reported by Eccles and colleagues (Eccles et al. 1967) in vivo and also observed in voltage-sensitive dye signals from the molecular layer of isolated cerebellum (Cohen & Yarom, 1998), can be explained by the kinetics and latencies of the underlying EPSC and feed-forward IPSCs. The shortening of EPSP time course by FFI has several important consequences. First, it enforces precisely timed output spikes in response to a single synchronous input. Our results show that FFI allows Purkinje cell output in response to PF stimulation to be timed to a precision of less than a millisecond, similarly to the effect of FFI in hippocampal pyramidal cells (Pouille & Scanziani, 2001). Second, it limits the time window for summation of multiple asynchronous inputs to only a few milliseconds (see also Brunel et al. 2004), indicating that the time window for synaptic integration in Purkinje cells can be much shorter than the membrane time constant in these neurones (Husser & Clark, 1997). Furthermore, inputs falling outside this time window are penalized, and have a lower-than-normal efficacy of spike generation. This provides a mechanism to further enhance the signal-to-noise of precisely timed inputs at the expense of inaccurately or asynchronously timed ones.
, 百拇医药
In contrast to pyramidal cells, cerebellar Purkinje cells show spontaneous activity, often at high rates, both in vitro (Husser & Clark, 1997) and in vivo (Armstrong & Rawson, 1979). As a consequence, both excitation and feed-forward inhibition are superimposed on a background of activity, and can act not only by initiating individual spikes, but also by accelerating or delaying spontaneous spikes. The proximity to threshold also means that even small inputs can influence spike output (Husser & Clark, 1997; Carter & Regehr, 2002; Suter & Jaeger, 2004). We have measured the effect of FFI on subsequent spike generation to be on the order of 30 ms. Therefore, even small PF inputs which do not on their own trigger spikes can influence spike output by reducing the efficacy of subsequent inputs, or by preventing spontaneous spikes. In a situation where the Purkinje cell receives hundreds or thousands of PF inputs, FFI could mediate ‘gain control’ of asynchronous ‘tonic’ excitation (Marr, 1969), whereas synchronous inputs could overcome FFI and result in fast-rising EPSPs that trigger spikes with high temporal precision due to subsequent FFI.
, http://www.100md.com
Mutual feed-forward inhibition in the inhibitory network
Molecular layer interneurons are highly interconnected (Husser & Clark, 1997; Kondo & Marty, 1998), and single spikes in presynaptic interneurons can powerfully regulate spike timing in postsynaptic interneurons (Husser & Clark, 1997; Carter & Regehr, 2002). We have demonstrated that molecular layer interneurons, in addition to generating FFI onto Purkinje cells, also receive FFI from each other. Interestingly, the level and effect of FFI in interneurons seems much weaker than that observed in Purkinje cells, measured both by the apparent underlying inhibitory conductance relative to that of excitation and by the effect of FFI on the EPSP waveform. This could either reflect differences in the relative size of unitary PF and inhibitory conductances in Purkinje cells and interneurons, or alternatively PFs could activate fewer feed-forward interneurons onto interneurons than onto Purkinje cells.
, 百拇医药
What is the functional role of this FFI onto interneurons One possibility may be to ensure the brevity of FFI onto Purkinje cells by shortening the excitation of interneurons. The time course of the FFI IPSC in Purkinje cells (Fig. 2A) is comparable to that of IPSCs evoked by direct electrical stimulation (Vincent et al. 1992), suggesting that activation of interneurons contributing to the FFI is rapidly terminated. This could occur via a combination of FFI onto interneurons themselves and the prominent reset caused by the AHP following single interneuron spikes (Carter & Regehr, 2002). The fact that PF synapses onto both interneurons and Purkinje cells facilitate (Perkel et al. 1990; Clark & Cull-Candy, 2002) suggests that FFI can keep pace with PF activation even during bursts of activity (Chadderton et al. 2004).
, http://www.100md.com
Functional implications for cerebellar processing
Several lines of evidence have converged to support the idea that the cerebellum is involved in representation of precise temporal information (Ivry & Spencer, 2004). We have now identified FFI as a key cellular mechanism that can help to preserve timing of cerebellar output to within millisecond precision. These properties may help Purkinje cells act as coincidence detectors responding only to well-timed PF inputs. Furthermore, the rapid onset of FFI may have special relevance for the functional architecture of the PF network. There has been considerable controversy regarding the role of PF ‘beams’ in cerebellar processing, with little direct evidence available for beam-like activation of Purkinje cells in the intact cerebellar cortex (Cohen & Yarom, 1998; Jaeger, 2003). Our results may provide new insights into this debate. Experimental and simulation studies have suggested that there may be considerable asynchrony of propagation across multiple PFs due to variability in fibre diameter (Bernard & Axelrad, 1991; Garwicz & Andersson, 1992). Since interneurons can be activated by only one or a few inputs (Barbour, 1993; Carter & Regehr, 2002), the extremely rapid onset of FFI and its long aftereffects suggest that the early arriving PF inputs can recruit FFI that will dampen the response to later ones, making the ‘on-beam’ response of the Purkinje cells weak or undetectable (Bower, 2002). Thus, regulation of the kinetics and strength of FFI can have important consequences for the spatial spread of activity patterns across the cerebellar cortex.
, 百拇医药
Footnotes
W. Mittmann and U. Koch contributed equally to this study.
References
Armstrong DM & Rawson JA (1979). Activity patterns of cerebellar cortical neurones and climbing fibre afferents in the awake cat. J Physiol 289, 425–448.
Barbour B (1993). Synaptic currents evoked in Purkinje cells by stimulating individual granule cells. Neuron 11, 759–769.
, 百拇医药
Bernard C & Axelrad H (1991). Propagation of parallel fiber volleys in the cerebellar cortex: a computer simulation. Brain Res 565, 195–208.
Bower JM (2002). The organization of cerebellar cortical circuitry revisited: implications for function. Ann N Y Acad Sci 978, 135–155.
Bower JM & Woolston DC (1983). Congruence of spatial organization of tactile projections to granule cell and Purkinje cell layers of cerebellar hemispheres of the albino rat: vertical organization of cerebellar cortex. J Neurophysiol 49, 745–766.
, http://www.100md.com
Braitenberg V & Atwood RP (1958). Morphological observations on the cerebellar cortex. J Comp Neurol 109, 1–33.
Brunel N, Hakim V, Isope P, Nadal JP & Barbour B (2004). Optimal information storage and the distribution of synaptic weights: perceptron versus Purkinje cell. Neuron 43, 745–757.
Buzsaki G (1984). Feed-forward inhibition in the hippocampal formation. Prog Neurobiol 22, 131–153.
, 百拇医药
Carter AG & Regehr WG (2002). Quantal events shape cerebellar interneuron firing. Nat Neurosci 5, 1309–1318.
Carter AG, Vogt KE, Foster KA & Regehr WG (2002). Assessing the role of calcium-induced calcium release in short-term presynaptic plasticity at excitatory central synapses. J Neurosci 22, 21–28.
Chadderton P, Margrie TW & Hausser M (2004). Integration of quanta in cerebellar granule cells during sensory processing. Nature 428, 856–860.
, 百拇医药
Clark BA & Cull-Candy SG (2002). Activity-dependent recruitment of extrasynaptic NMDA receptor activation at an AMPA receptor-only synapse. J Neurosci 22, 4428–4436.
Cohen D & Yarom Y (1998). Patches of synchronized activity in the cerebellar cortex evoked by mossy-fiber stimulation: questioning the role of parallel fibers. Proc Natl Acad Sci U S A 95, 15032–15036.
Delaney AJ & Jahr CE (2002). Kainate receptors differentially regulate release at two parallel fiber synapses. Neuron 36, 475–482.
, 百拇医药
Destexhe A, Rudolph M & Pare D (2003). The high- conductance state of neocortical neurons in vivo. Nat Rev Neurosci 4, 739–751.
Eccles JC, Ito M & Szenthágothai J (1967). The Cerebellum as a Neuronal Machine. Springer-Verlag, Berlin.
Fetz EE & Gustafsson B (1983). Relation between shapes of post-synaptic potentials and changes in firing probability of cat motoneurones. J Physiol 341, 387–410.
, 百拇医药
Garwicz M & Andersson G (1992). Spread of synaptic activity along parallel fibres in cat cerebellar anterior lobe. Exp Brain Res 88, 615–622.
Hamann M, Desarmenien M, Desaulles E, Bader MF & Feltz P (1988). Quantitative evaluation of the properties of a pyridazinyl GABA derivative (SR 95531) as a GABAA competitive antagonist. An electrophysiological approach. Brain Res 442, 287–296.
Husser M & Clark BA (1997). Tonic synaptic inhibition modulates neuronal output pattern and spatiotemporal synaptic integration. Neuron 19, 665–678.
, 百拇医药
Isope P & Barbour B (2002). Properties of unitary granule cell–Purkinje cell synapses in adult rat cerebellar slices. J Neurosci 22, 9668–9678.
Ivry R (1997). Cerebellar timing systems. Int Rev Neurobiol 41, 555–573.
Ivry RB & Spencer RM (2004). The neural representation of time. Curr Opin Neurobiol 14, 225–232.
Jaeger D (2003). No parallel fiber volleys in the cerebellar cortex: evidence from cross-correlation analysis between Purkinje cells in a computer model and in recordings from anesthetized rats. J Comput Neurosci 14, 311–327.
, 百拇医药
Jaeger D & Bower JM (1999). Synaptic control of spiking in cerebellar Purkinje cells: dynamic current clamp based on model conductances. J Neurosci 19, 6090–6101.
Jaeger D, De Schutter E & Bower JM (1997). The role of synaptic and voltage-gated currents in the control of Purkinje cell spiking: a modeling study. J Neurosci 17, 91–106.
Kondo S & Marty A (1998). Synaptic currents at individual connections among stellate cells in rat cerebellar slices. J Physiol 509, 221–232.
, 百拇医药
Llano I & Gerschenfeld HM (1993). Inhibitory synaptic currents in stellate cells of rat cerebellar slices. J Physiol 468, 177–200.
Llano I, Marty A, Armstrong CM & Konnerth A (1991). Synaptic- and agonist-induced excitatory currents of Purkinje cells in rat cerebellar slices. J Physiol 434, 183–213.
Marr D (1969). A theory of cerebellar cortex. J Physiol 202, 437–470.
Meek J (1992). Why run parallel fibers parallel Teleostean Purkinje cells as possible coincidence detectors, in a timing device subserving spatial coding of temporal differences. Neuroscience 48, 249–283.
, 百拇医药
Palay SL & Chan-Palay V (1974). Cerebellar Cortex: Cytology and Organization. Springer-Verlag, Berlin.
Perkel DJ, Hestrin S, Sah P & Nicoll RA (1990). Excitatory synaptic currents in Purkinje cells. Proc R Soc Lond B Biol Sci 241, 116–121.
Pouille F & Scanziani M (2001). Enforcement of temporal fidelity in pyramidal cells by somatic feed-forward inhibition. Science 293, 1159–1163.
, 百拇医药 Pouzat C & Hestrin S (1997). Developmental regulation of basket/stellate cell–Purkinje cell synapses in the cerebellum. J Neurosci 17, 9104–9112.
Spencer RM, Zelaznik HN, Diedrichsen J & Ivry RB (2003). Disrupted timing of discontinuous but not continuous movements by cerebellar lesions. Science 300, 1437–1439.
Sultan F & Bower JM (1998). Quantitative Golgi study of the rat cerebellar molecular layer interneurons using principal component analysis. J Comp Neurol 393, 353–373.
, 百拇医药
Suter KJ & Jaeger D (2004). Reliable control of spike rate and spike timing by rapid input transients in cerebellar stellate cells. Neuroscience 124, 305–317.
Vincent P, Armstrong CM & Marty A (1992). Inhibitory synaptic currents in rat cerebellar Purkinje cells: modulation by postsynaptic depolarization. J Physiol 456, 453–471.
Vincent P & Marty A (1996). Fluctuations of inhibitory postsynaptic currents in Purkinje cells from rat cerebellar slices. J Physiol 494, 183–199.
Vos BP, Maex R, Volny-Luraghi A & De Schutter E (1999). Parallel fibers synchronize spontaneous activity in cerebellar Golgi cells. J Neurosci RC6, 1–5.
Welsh JP & Llinás R (1997). Some organizing principles for the control of movement based on olivocerebellar physiology. Prog Brain Res 114, 449–461., 百拇医药(Wolfgang Mittmann, Ursula)