Neuron, Vol. 29, 717–727, March, 2001, Copyright 2001 by Cell PressRetrograde Inhibition of Presynaptic CalciumInflux by Endogenous Cannabinoidsat Excitatory Synapses onto Purkinje CellsAnatol C. Kreitzer and Wade G. Regehr*Department of NeurobiologyHarvard Medical SchoolBoston, Massachusetts 02115SummaryBrief depolarization of cerebellar Purkinje cells wasfound to inhibit parallel fiber and climbing fiber EPSCsfor tens of seconds. This depolarization-induced suppression of excitation (DSE) is accompanied by alteredpaired-pulse plasticity, suggesting a presynaptic locus. Fluorometric imaging revealed that postsynapticdepolarization also reduces presynaptic calcium influx. The inhibition of both presynaptic calcium influxand EPSCs is eliminated by buffering postsynaptic calcium with BAPTA. The cannabinoid CB1 receptor antagonist AM251 prevents DSE, and the agonist WIN55,212-2 occludes DSE. These findings suggest thatPurkinje cells release endogenous cannabinoids in response to elevated calcium, thereby inhibiting presynaptic calcium entry and suppressing transmitter release. DSE may provide a way for cells to use theirfiring rate to dynamically regulate synaptic inputs. Together with previous studies, these findings suggesta widespread role for endogenous cannabinoids inretrograde synaptic inhibition.IntroductionRetrograde signaling is used by the nervous system toconvey information about the activity of neurons backto cells that innervate them. Retrograde messengersplay an important role in synapse formation (Fitzsimondsand Poo, 1998) and in the control of synaptic strengthon rapid timescales (Alger and Pitler, 1995; Kombian etal., 1997; Zilberter et al., 1999). Short-term retrograderegulation of synapses was first described in the hippocampus and in the cerebellum, where depolarizationof either CA1 pyramidal cells or Purkinje cells inhibitspresynaptic GABA release onto these cells for tens ofseconds (Llano et al., 1991b; Pitler and Alger, 1992,1994; Vincent and Marty, 1993). Depolarization-inducedsuppression of inhibition (DSI) depends upon a rise inpostsynaptic calcium that triggers the release of a retrograde messenger that in turn acts on the presynapticnerve terminal through a G protein–dependent mechanism (Pitler and Alger, 1994). However, inhibition of excitatory inputs by the retrograde messenger that givesrise to DSI has not been described. We find a depolarization-induced suppression of excitatory inputs (DSE) incerebellar Purkinje cells, indicating that both excitatoryand inhibitory synapses can be modulated on rapidtimescales.* To whom correspondence should be addressed (e-mail: [email protected]).Studies of DSE in the cerebellum can provide insightinto several fundamental issues regarding retrogradesynaptic inhibition. A number of retrograde messengershave been identified at various synapses, including glutamate, GABA, and neuropeptides (Glitsch et al., 1996;Kombian et al., 1997; Morishita et al., 1998; Zilberteret al., 1999; Zilberter, 2000), as well as endogenouscannabinoids (R. I. Wilson and R. A. Nicoll, unpublisheddata). It is not known how widely these messengersserve to retrogradely inhibit synapses throughout thebrain. Another unresolved issue is how these retrogrademessengers inhibit release from presynaptic terminals.Several possible mechansims include branchpoint failure and incomplete action potential invasion of presynaptic terminals (Hatt and Smith, 1976; Alger et al., 1996),inhibition of presynaptic calcium channels (Anwyl, 1991),or direct effects on the release apparatus (Thompsonet al., 1993; Alger and Pitler, 1995; Chen and Regehr,1997). The excitatory synapses we study are amenableto optical recording methods, which allow us to differentiate between these possibilities.Here, we find that Purkinje cell depolarization retrogradely inhibits both parallel and climbing fiber inputs.This DSE requires elevated postsynaptic calcium andfollows a time course similar to DSI. We find that DSEis due to an inhibition of presynaptic calcium influx. DSEis prevented by blocking cannabinoid CB1 receptors,which are located in the molecular layer of the cerebellum and can inhibit both parallel fibers and climbingfibers (Mailleux and Vanderhaeghen, 1992; Matsuda etal., 1993; Levenes et al., 1998; Takahashi and Linden,2000). Thus, a transient elevation of postsynaptic calcium in Purkinje cells results in the release of endogenous cannabinoids, which inhibit afferent excitatory inputs for tens of seconds by modulating presynapticcalcium entry. Taken together with the widespread distribution of CB1 receptors in the brain and the observation that CB1 receptors are required for DSI in the hippocampus (R. I. Wilson and R. A. Nicoll, unpublished data),this suggests a general role for endogenous cannabinoids in the retrograde inhibition of both excitatory andinhibitory synapses.ResultsIn this study, we examine two excitatory glutamatergicinputs to Purkinje cells (PC) with distinctive properties:parallel fiber (PF) and climbing fiber (CF) synapses (Eccles et al., 1966a, 1966b; Palay and Chan-Palay, 1974;Ito, 1984; Konnerth et al., 1990; Perkel et al., 1990; Silveret al., 1998). PF synapses are the connections betweencerebellar granule cells and PCs. Each PC receives tensof thousands of PF synapses from cerebellar granulecells, with most granule cells making only a small number of contacts. PF synapses have a low probability ofrelease and display prominent paired-pulse facilitation.By contrast, PCs are typically innervated by a singleCF that originates in the inferior olive. Each CF makeshundreds of synaptic contacts, which have a high probability of release and exhibit paired-pulse depression.
Neuron718Figure 1. Postsynaptic Depolarization Inhibits Excitatory PurkinjeCell Afferents(A) Stimulus protocol with the holding potential of the postsynapticcell (hp; upper) and the stimulation timing (stim; below). Parallel fiber(B) and climbing fiber (C) EPSC amplitudes are plotted over timefor control responses with no preceding prepulse to 0mV (opencircles) and test responses following Purkinje cell depolarization(closed circles). Average parallel fiber (B) and climbing fiber (C)EPSCs are shown at the right. Stimulus artifacts are blanked forclarity. Parallel fiber and climbing fiber responses are from tworepresentative experiments. The duration of the depolarization to0mV was 50 ms for parallel fiber experiments and 1 s for climbingfiber responses. The test stimulus followed the depolarization by t 5 s.Effects of Postsynaptic Depolarizationon Excitatory Synaptic TransmissionThe effects of brief postsynaptic depolarizations on excitatory Purkinje cell afferents were studied in transverserat cerebellar slices. PF EPSCs were evoked with anextracellular electrode placed in the molecular layer.After obtaining a stable synaptic response, we assessedthe influence of stepping the voltage of the postsynapticPC from 60mV to 0mV (Figure 1A). This greatly decreased the PF EPSC amplitude, as shown for a teststimulus that followed the postsynaptic depolarizationby a t of 5 s (Figure 1B, lower right). This EPSC inhibitionrecovered within 90 s and was repeatedly and reliablyelicited over the duration of an experiment (Figure 1B,left) where EPSCs following depolarization (closed circles) and control EPSCs (open circles) are both plotted.Depolarization of the PC also reliably inhibited CF synapses (Figure 1C). CF synapses appeared to be somewhat less sensitive to postsynaptic depolarization, and,therefore, the duration of the depolarization prior to atest CF EPSC was 1 s, compared to 50 ms for PF stimulation. We refer to this depolarization-induced suppression of excitation as DSE.We next examined the time course of DSE by systematically varying t following a postsynaptic depolarizingprepulse (Figures 2Aa and 2Ba). DSE is small at earlytimes ( t 1 s) and then approaches a maximum at5–10 s and decays with a t1/2 of ⵑ15–20 s. The timecourse of DSE is similar at both the PF and CF synapses,suggesting that a common mechanism may underlieboth phenomena.To test whether DSE is expressed as a presynapticor postsynaptic change, we assessed the effects ofpostsynaptic depolarization on paired-pulse plasticity.Most synapses, including both PF and CF synapses,display prominent short-term synaptic plasticity, whichcan provide insight into the probability of release. PFand CF synapses behave very differently in response topairs of stimuli. The ratio of the amplitudes of the EPSCevoked by the second and first stimuli (A2/A1) is typicallyabout 160% for PF synapses, and they are said to facilitate, which is consistent with their low initial probabilityof neurotransmitter release. CF synapses have a highinitial probability of release and display paired-pulsedepression, with A2/A1 of about 40%. At both of thesesynapses, short-term plasticity is thought to be presynaptic in origin (Atluri and Regehr, 1996; Dittman andRegehr, 1998), and a decrease in the initial probabilityof release increases the ratio A2/A1. Therefore, if theinhibition we observe following depolarization reflectsa presynaptic change in the probability of release, A2/A1 should increase following depolarization. Most postsynaptic mechanisms are not consistent with suchchanges in short-term plasticity.We found that postsynaptic depolarization affectsshort-term plasticity at both PF and CF synapses (Figures 2Ab and 2Bb). In the experiment shown, the pairedpulse ratio A2/A1 at the PF synapse increased from150% in control conditions to 260% at 5 s after postsynaptic depolarization. At the CF synapse, the A2/A1 ratioincreased from 30% to 80% at t 5 s. Moreover, thetime course of these changes in A2/A1 at both the PFand CF synapses matches the time course of DSE (Figures 2Aa and 2Ba). These increases in A2/A1 suggestthat postsynaptic depolarization decreases the probability of release from both climbing fibers and parallelfibers. Summary data is shown in Figure 3.The Role of Postsynaptic Calcium in DSEBecause postsynaptic depolarization results in an apparent suppression of presynaptic inputs, we next determined if a rise in postsynaptic calcium is required forDSE. The inclusion of the calcium chelator BAPTA (40mM) in the postsynaptic recording pipette completelyblocked the suppression of EPSCs at both the PF andCF synapses (Figure 3). At t 5 s, when DSE is maximalin control conditions, there is no sign of either EPSCdepression or a change in paired-pulse plasticity withBAPTA in the postsynaptic recording pipette (Figures
Retrograde Inhibition by Endogenous Cannabinoids719Figure 2. Presynaptic Short-Term Plasticity Is Altered Following Postsynaptic DepolarizationParallel fiber (Aa) and climbing fiber (Ba) responses to stimuli lacking a postsynaptic prepulse and in response to test stimuli followingdepolarization to 0mV, during systematic variation of t. EPSCs are shown above, and the time course of depression of the test EPSC isbelow. Short-term plasticity of the parallel fiber (Ab) and climbing fiber (Bb) EPSCs evoked by pairs of stimuli are shown. Traces are normalizedto the first EPSC of the control stimulus to aid in comparison. The amplitude of the paired-pulse ratio (A2/A1) as a function of time after thepostsynaptic depolarization is plotted below. Data in (Aa) and (Ab) are from a single representative parallel fiber experiment. Data in (Ba) and(Bb) are from a single representative climbing fiber experiment.3Aa and 3Ba). At PF and CF synapses, postsynapticBAPTA completely abolished EPSC inhibition followingdepolarization at each value of t tested (Figures 3Aband 3Bb). The changes in paired-pulse plasticity weobserved at the PF and CF synapses during inhibition incontrol conditions were also eliminated by postsynapticBAPTA at all time points (Figures 3Ac and 3Bc). Thesedata suggest that elevations in postsynaptic calciumare required for DSE at both PF and CF synapses.DSE at High TemperatureWe next examined the magnitude and time course ofDSE at 34 C, which is much closer to physiological temperatures (Figure 4). At PF synapses, both the onset anddecay of DSE is faster (Figure 4A). At t 1 s, a largeinhibition is already present, which is maximal at 3 sand completely decayed by 20–30 s. The magnitude ofDSE is slightly larger in the parallel fibers at 34 C. In theclimbing fiber, a similar increase in the speed of onsetand decay of inhibition was observed, while the magni-tude of DSE remained similar between 24 C and 34 C(Figure 4B). The prominence of DSE at 34 C suggeststhat this phenomenon is important at physiological temperatures.The Mechanism of Presynaptic EPSC SuppressionChanges in paired-pulse plasticity suggest that Purkinjecell depolarization produces a presynaptic suppressionof transmitter release. This could arise from an inhibitionof presynaptic calcium channels, branchpoint failure,or presynaptic inhibition downstream of calcium influx(Hatt and Smith, 1976; Anwyl, 1991; Thompson et al.,1993; Alger et al., 1996; Chen and Regehr, 1997). Wetherefore monitored presynaptic calcium influx to distinguish between these possibilities and to further characterize the mechanism responsible for DSE.Optical measurements of presynaptic calcium levelswere made at the climbing fiber. Each PC is innervatedby a single CF whose axonal arbor forms a dense clusterof synapses that are contained in a single plane and are
Neuron720Figure 3. Elevation of Postsynaptic Calcium Is Required for DSEParallel fiber (Aa) and climbing fiber (Ba) EPSCs evoked by pairs of stimuli. Responses to stimuli without a preceding prepulse are overlayedon responses to test stimuli, both in control conditions and in the presence of postsynaptic BAPTA (40 mM). Traces are single trials fromrepresentative experiments. Summary of the time course of EPSC inhibition following postsynaptic depolarization in control conditions (closedcircles) and in the presence of postsynaptic BAPTA (open circles) for paralle