Neuronal and Astrocytic Monoacylglycerol Lipase Limit the Spread of Endocannabinoid Signaling in the Cerebellum123

Abstract Endocannabinoids are diffusible lipophilic molecules that may spread to neighboring synapses. Monoacylglycerol lipase (MAGL) is the principal enzyme that degrades the endocannabinoid 2-arachidonoylglycerol (2-AG). Using knock-out mice in which MAGL is deleted globally or selectively in neurons and astrocytes, we investigated the extent to which neuronal and astrocytic MAGL limit the spread of 2-AG-mediated retrograde synaptic depression in cerebellar slices. A brief tetanic stimulation of parallel fibers in the molecular layer induced synaptically evoked suppression of excitation (SSE) in Purkinje cells, and both neuronal and astrocytic MAGL contribute to the termination of this form of endocannabinoid-mediated synaptic depression. The spread of SSE among Purkinje cells occurred only after global knock-out of MAGL or pharmacological blockade of either MAGL or glutamate uptake, but no spread was detected following neuron- or astrocyte-specific deletion of MAGL. The spread of endocannabinoid signaling was also influenced by the spatial pattern of synaptic stimulation, because it did not occur at spatially dispersed parallel fiber synapses induced by stimulating the granular layer. The tetanic stimulation of parallel fibers did not induce endocannabinoid-mediated synaptic suppression in Golgi cells even after disruption of MAGL and glutamate uptake, suggesting that heightened release of 2-AG by Purkinje cells does not spread the retrograde signal to parallel fibers that innervate Golgi cells. These results suggest that both neuronal and astrocytic MAGL limit the spatial diffusion of 2-AG and confer synapse-specificity of endocannabinoid signaling.

As a diffusible lipophilic messenger, 2-AG may spread from its point of release to unstimulated synapses. In the hippocampus, DSI induced in a CA1 pyramidal neuron could spread to a non-depolarized neuron whose soma is separated by ϳ20 m (Wilson and Nicoll, 2001). In the cerebellum, DSE spreads from one Purkinje cell (PC) to a neighboring PC at room temperature (24°C), but not at physiological temperature (34°C; Kreitzer et al., 2002); mGluR-driven endocannabinoid-mediated synaptic depression does not spread to neighboring PCs or to nearby unstimulated synapses on the same PC (Brown et al., 2003;. Thus, endocannabinoid signaling is spatially restricted at most types of synapses but can spread at a short distance at other synapses. Given the pivotal role of MAGL in 2-AG metabolism, it stands to reason that MAGL limits the spread of the 2-AG signal and therefore confers cell-and synapse-specificity of endocannabinoid-mediated synaptic suppression. In recent years, there has been an increasing appreciation that glial cells are integral parts of the endocannabinoid system (Navarrete and Araque, 2008;Han et al., 2012). MAGL is not only expressed in neurons, but also in astrocytes (Tanimura et al., 2012;Viader et al., 2015). Using conditional knock-out mice in which MAGL was deleted globally and specifically in neurons and astrocytes, recent studies have shown that neuronal and astrocytic MAGL coordinately regulate brain 2-AG content and contribute to termination of DSE at the parallel fiber (PF)-PC synapses in the cerebellum and DSI in CA1 pyramidal neurons in the hippocampus (Viader et al., 2015). These results suggest that both neuronal and astrocytic MAGL contribute to the termination of endocannabinoid-mediated retrograde synaptic depression.
The cerebellar cortex forms an array of well defined neural circuits and expresses one of the highest levels of cannabinoid receptors (CB 1 Rs) in the brain (Kawamura et al., 2006). PCs, the single output neurons in the cerebellar cortex, receive excitatory inputs from PFs and climbing fibers (Palay and Chan-Palay, 1974;Konnerth et al., 1990). Multiple forms of endocannabinoid-mediated retrograde synaptic suppression can be induced in the cerebellum (Safo et al., 2006;Heifets and Castillo, 2009;Marcaggi, 2015). Among them, SSE induced by brief tetanic stimulation of PFs provides an excellent model for investigating the mechanisms of the spread of endocannabinoid signaling. Previous studies have shown that blocking calcium-dependent and independent endocannabinoid release from the recorded PC with intracellular dialysis of calcium chelator BAPTA and G-protein inhibitor GDP-␤S abolished SSE Brown et al., 2003). Because the tetanic stimulation activates PF terminals that innervate a number of PCs, these studies suggest that 2-AG signaling is highly spatially restricted and does not spread from neighboring PCs to the recorded PC. In the present study, we explored factors that may allow the spread of SSE among PCs because it would provide novel insight into the mechanisms that confer spatial restriction of endocannabinoid signaling. We examined the extent of spread of SSE following global or cell-type-specific disruption of MAGL or pharmacological blockade of glutamate transporters. We find that the 2-AG signal could spread among PCs after global disruption of MAGL or glutamate uptake, whereas neuron-and astrocyte-specific deletion of MAGL had no significant effect. Thus, hydrolysis of 2-AG by MAGL and glutamate reuptake by glutamate transporters limit the spread of endocannabinoid signaling among PCs.

Animals
All animal use was in accordance with protocols approved by our Institution's Animal Care and Use Committee. Wild-type and MAGL global and cell-type-specific knock-out mice were generated and validated based on previous studies (Viader et al., 2015). Total MAGL knockout (MAGL-TKO) and wild-type littermates were gener-ated by breeding homozygous Mgll loxP/loxP mice with Rosa26-Cre (Otto et al., 2009). Neuron-specific (MAGL-NKO) and astrocyte-specific (MAGL-AKO) were generated by crossing Mgll loxP/loxP to Eno2-Cre mice (Frugier et al., 2000) and GFAP-Cre mice (Tao et al., 2011;Sofroniew, 2012) respectively, and then backcrossing the resulting double-heterozygotes (Cre ϩ/-, Mgll ϩ/loxP ) to Mgll-loxP/loxP to produce cell-type-specific MAGL knock-out mice (Cre ϩ/-, Mgll loxP/loxP ) and wild-type littermates (Cre -/-, Mgll loxP/loxP ). In the cerebellum, GFAP is exclusively expressed in Bergmann glia, a type of astrocytes that have their cell bodies in the PC layer and processes that extend into molecular layer (Nolte et al., 2001). Using a Creinducible Rosa26-tdTomato reporter line, previous studies showed that efficient Cre-mediated recombination occurs selectively in neurons and astrocytes throughout the brain of Eno2-and GFAP-Cre mice, respectively, and that MAGL is selectively deleted in targeted cell types (Viader et al., 2015). Genotyping was performed by PCR using a DNA sample obtained from the tail or ear.

Electrophysiology
Whole-cell voltage-clamp recordings were made from PCs and Golgi cells in cerebellar slices. Cells were visualized using infrared-differential interference contrast optics (Nikon Eclipse FN1 and Olympus BX51WI) and 40ϫ water immersion lens. PF-PC EPSCs were evoked by placing a bipolar tungsten stimulation electrode (WPI) in molecular layer in most experiments but in the granular layer in experiments presented in Figure 5. PF-PC EPSCs showed graded responses and exhibited paired-pulse facilitation (Konnerth et al., 1990;Kreitzer and Regehr, 2001). GABA A receptor blocker picrotoxin (50 M) was present in the ACSF.
Golgi cells were identified and distinguished from other cell types in the granular layer based on their relative large size (8 -25 m), biexponential capacitive currents, and the presence of Na ϩ current and monosynaptic EPSCs evoked by molecular layer stimulation (Dieudonne, 1995;Bureau et al., 2000;Beierlein et al., 2007). Glass pipettes (2-3 M⍀) were filled with internal solutions containing the following (in mM): 140 K-gluconate, 5 KCl, 10 HEPES, 0.2 EGTA, 2 MgCl 2 , 4 Mg-ATP, 0.3 Na 2 GTP, and 10 Na 2phosphocreatine, pH 7.3 with KOH. In experiments required for buffering intracellular calcium and blocking mGluR signaling, the internal solution contained the following (in mM): 80 K-gluconate, 5 KCl, 10 HEPES, 20 BAPTA, 2 MgCl 2 , 4 Mg-ATP, 2 GDP-␤S (or 1 GTP-␥S), and 10 Na 2 -phosphocreatine, ϳpH 7.3 with KOH. We found that storage of stock solution of GDP-␤S or GTP-␥S even at Ϫ80°C led to reduction of their effectiveness in blocking SSE. GDP-␤S or GTP-␥S powder was weighted and freshly added into intracellular solution just before the experiments and was used within 3 h. All recordings were performed at 32 Ϯ 1°C by using an automatic temperature controller (Warner Instruments).

Data analysis and statistics
The EPSC amplitude was normalized to the baseline. The decay time constant () of SSE was measured using a single exponential function of y ϭ y 0 ϩ k ϫ exp(Ϫx/), in which y is the magnitude of SSE, y 0 is the peak magnitude of SSE, k is the constant multiplier, and x is the time. The magnitude of SSE (%) was calculated as follows: 100 ϫ [(mean amplitude of 2 EPSCs after tetanic stimulation/ mean amplitude of 5 EPSCs before the tetanic stimulation)]. Values of two to three trials were averaged for each neuron. Data are presented as the mean Ϯ SEM. Results were analyzed with one-way ANOVA or Student's t test.
Results were considered to be significant at p Ͻ 0.05.

Both neuronal and astrocytic MAGL contribute to the termination of SSE
A brief tetanic stimulation of PFs induces transient suppression of EPSCs in PCs; this synaptically evoked SSE is mediated by synaptic activation of mGluR1 and subsequent recruitment of endocannabinoid signaling Brown et al., 2003;Attwell, 2005, 2007;Tanimura et al., 2009). We examined the effects of MAGL-TKO, -NKO, and -AKO on SSE at PF-PC synapses. A bipolar stimulation electrode was placed in the molecular layer to evoke EPSCs at 4 s intervals. SSE was induced by a brief tetanic stimulation (50 Hz, 1 s) of PFs in the molecular layer while the PC was voltageclamped at Ϫ70 mV (Tanimura et al., 2009). In wild-type slices, tetanic stimulation of the molecular layer induced SSE of PF-EPSCs (Fig. 1A,B). One-way ANOVA showed that genetic deletion of MAGL significantly prolonged the time course of SSE, as shown by increases in the decay time constant () of SSE (F (3,37) ϭ 22.46, p Ͻ 0.001). Tukey's post hoc tests indicated that the of SSE was significantly prolonged in MAGL-TKO slices compared with that of wild-type slices (p Ͻ 0.001; Fig. 1B,C). MAGL-NKO and MAGL-AKO produced SSE with less dramatic but significantly prolonged duration (NKO vs WT, p ϭ 0.007; AKO vs WT, p ϭ 0.036; Fig. 1B,C). Interestingly, in wild-type, MAGL-NKO and AKO slices, SSE peaked immediately after the tetanic stimulation; but SSE has a slow onset in MAGL-TKO slices (Fig. 1A,B), suggesting that total loss of MAGL causes delayed 2-AG mobilization. Global or cell-type-specific deletion of MAGL did not significantly alter the magnitude of SSE (F (3,37) ϭ 0.09, p ϭ 0.966; Fig. 1D). In the presence of the CB 1 R antagonist AM 251 (2 M), the tetanic stimulation induced indistin-guishable post-tetanic potentiation (PTP) of EPSCs in wild-type and MAGL-TKO slices (Student's t test: t 16 ϭ 0.38, p ϭ 0.708; Fig. 1E), confirming that the tetanic stimulation induced endocannabinoid-mediated suppression of PF-EPSCs. These results indicate that both neuronal and astrocytic MAGL made comparable contributions to the termination of SSE at PF-PC synapses.

MAGL and glutamate uptake limit the spread of SSE
In cerebellar slices, the SSE is restricted to synaptic inputs activated by the brief high-frequency stimulation and does not spread to neighboring synapses 20 m apart from the site of stimulation on the same Purkinje cell (Brown et al., 2003). Furthermore, intracellular dialysis of pharmacological agents that block endocannabinoid release from the recorded PC abolishes SSE (Brown et al., 2003;. These results suggest that SSE does not spread among neighboring PCs or among independent PFs innervating the same PC. What might be the mechanisms for synapse-specificity of SSE? One possibility is that hydrolysis of 2-AG by MAGL limits spatial diffusion of 2-AG along CB 1 R-expressing PF axonal terminals. In addition, the rapid uptake of glutamate by glutamate transporters prevents spillover of glutamate and subsequent activation mGluR1 on neighboring synapses (Brasnjo and Otis, 2001;Marcaggi and Attwell, 2005). We first explored whether global knock-out of MAGL allowed the 2-AG signal to spread among PCs during SSE. The tetanic stimulation of PFs triggers 2-AG release not only from the recorded PC but also from neighboring PCs. Non-hydrolysable analogs of GTP (GTP-␥S) and GDP (GDP-␤S) cause persistent activation and inactivation of G-proteins, respectively. Previous studies have shown that intracellular dialysis of calcium chelator BAPTA and GDP-␤S or GTP-␥S abolished SSE and mGluR agonist DHPG-induced depression of EPSCs through blockade of calcium-dependent and -independent endocannabinoid release from the recorded PC Brown et al., 2003). If SSE remains following intracellular dialysis of BAPTA and GDP-␤S or GTP-␥S into the recorded PC, then it is suggestive that 2-AG produced from neighboring PCs spreads to PF terminals that innervate the recorded PC.
Having shown that genetic deletion of MAGL globally enables the spread of the endocannabinoid signal, we next determined whether acute pharmacological blockade of MAGL produced a similar effect. Toward this end, we used a highly selective and potent MAGL inhibitor JZL184, which enhances 2-AG levels in the brain following in vivo injection . JZL184 substantially extends DSE at PF-and CF-to PC synapses (Pan et al., 2009) and SSE at PF-PC synapses (Bergerot et al., 2013) in wild-type mouse cerebellar slices. The following experiments were performed in wild-type control slices or slices treated with JZL184 (0.3 M; Pan et al., 2009). In PCs loaded with control internal solution, bath perfusion of a saturating concentration of JZL184 (0.3 M) substantially prolonged SSE (Student's t test, t 15 ϭ 7.50, p Ͻ0.001) without significantly changing the magnitude of SSE (t 15 ϭ 0.50, p ϭ 0.626; Fig. 3A,B,D). Intracellular dialysis of BAPTA (20 mM) and GDP-␤S (2 mM) blocked SSE in control slices, resulting in PTP of EPSCs (t 17 ϭ 8.74, p Ͻ 0.001; Fig. 3A,D); in contrast, intracellular BAPTA and GDP-␤S decreased the magnitude of SSE in JZL184-treated slices (t 14 ϭ 6.85, p Ͻ 0.001; Fig. 3B,D), and the remaining SSE was abolished by the CB 1 receptor antagonist AM251 (2 M; Fig. 3B), suggesting that it is mediated by endocannabinoid release from neighboring PCs because intracellular BAPTA and GDP-␤S abolished SSE in wild-type control slices. These results further confirm the spread of SSE following functional loss of MAGL. It has been shown that JZL184 does not significantly alter PF-DSE in cerebellar PCs in MAGL global knock-out mice . Consistent with this study, we found that JZL184 (0.3 M) did not significantly alter the magnitude of SSE (vehicle, 43.92 Ϯ 4.83%, n ϭ 9; JZL184, 41.42 Ϯ 7.48%, n ϭ 9, t 15 ϭ 0.29, p ϭ 0.778), nor did it alter the decay time constant of SSE (vehicle, 123.46 Ϯ 6.16 s, n ϭ 9; JZL184, 135.14 Ϯ 8.58 s, n ϭ 8, t 15 ϭ 1.12, p ϭ 0.279). These results indicate that MAGL global knock-out occludes the effects of JZL184, further confirming that JZL184 prolongs SSE in wild-type mice by inhibiting MAGL.
High-frequency stimulation causes spillover of glutamate, which activates extrasynaptic mGluR1 to induce endocannabinoid release (Brown et al., 2003;Marcaggi and Attwell, 2005). We investigated whether the glutamate transporter inhibitor TBOA enables the spread of SSE among PCs. Experiments were performed from wildtype control slices or slices that were perfused with TBOA (100 M). In PCs loaded with control internal solution, bath perfusion of TBOA substantially prolonged SSE (t 16 ϭ 6.798, p Ͻ 0.001) without significantly changing the magnitude of SSE (t 16 ϭ 0.82, p ϭ 0.426; Fig. 3C,D). Intracellular dialysis of BAPTA and GDP-␤S blocked SSE in control slices (t 17 ϭ 8.74, p Ͻ 0.001) but did not significantly alter SSE in TBOA-treated slices (t 16 ϭ 1.76, p Ͻ 0.099; Fig. 3C,D). These results imply that 2-AG produced by neighboring PCs diffuses to PF synapses onto the recorded PC to sustain the bulk of SSE when glutamate reuptake is blocked.
Interestingly, SSE induced in wild-type slices treated with JZL184 (Fig. 3B) or in MAGL-TKO slices had a slow onset that peaked ϳ20 s after the tetanic stimulation (Fig.  1A,B). What might be the mechanisms for the slow onset of SSE following the loss of function of MAGL? Given that SSE is mediated by mGluR-driven endocannabinoid release (Brown et al., 2003), delayed activation of mGluR might account for the slow onset of SSE. A short burst of high-frequency synaptic stimulation induces mGluR1mediated slow EPSCs in PCs (Batchelor and Garthwaite, 1997;Brasnjo and Otis, 2001). We therefore compared the latency of mGluR1-mediated slow EPSCs in slices treated with vehicle or JZL184 (0.3 M). The latency (the time between the last stimulus and onset of slow EPSCs) should be increased if the activation of mGluR1 was delayed by JZL184 treatment. mGluR1-EPSCs were evoked by placing the stimulating electrode in the dendritic branches of the recorded PC at a distance of ϳ200  9 -11). B, Intracellular BAPTA and GDP-␤S or GTP-␥S attenuated but did not completely block SSE in MAGL-TKO slices (n ϭ 9 -10), suggesting that spread of 2-AG signal from neighboring PCs to the PC that was under study. C, The application of CB 1 R antagonist AM 251 (2 M) to MAGL-TKO slices blocked the remaining SSE induced in PCs loaded with BAPTA and GDP-␤S (n ϭ 9). D, E, Intracellular BAPTA and GDP-␤S abolished SSE in MAGL-NKO (D; n ϭ 8 -9) and -AKO mice (E; n ϭ 8 -10). F, Summary of changes in EPSC amplitude under experimental conditions shown in A-E (n ϭ 8 -11; ‫‪p‬ءء‬ Ͻ 0.01, ‫‪p‬ءءء‬ Ͻ 0.001). m with a fixed stimulating intensity (100 A), and AMPA receptor antagonist CNQX (20 M) and GABA A receptor blocker picrotoxin (100 M) were present in the ACSF, and the recordings were made blind to drug treatment history of the slices. We found that a brief burst of electrical stimulation (100 Hz, 5 stimuli) of the molecular layer induced slow EPSCs in PCs in both vehicle-and JZL184treated slices. Bath application of mGluR1 antagonist CPCCOEt (100 M) abolished slow EPSCs within 5 min (Fig. 4A), confirming that the slow EPSCs are mediated by mGluR1. The amplitude of the mGluR1-EPSCs varied greatly among different PCs in either vehicle-or JZL184treated slices (Fig. 4B). On average, the mean amplitude of slow EPSCs was not significantly different between vehicle-and JZL184-treated slices (t 27 ϭ 0.28, p ϭ 0.781; Fig. 4B). In order to have more precise qualification of the latency, rise time, and decay time constant of mGluR1-EPSCs, we compared these parameters in PCs in vehicleand JZL184-treated slices that exhibited mGluR1-EPSCs with peak amplitude Ն50 pA. The latency (t 15 ϭ 0.22, p ϭ 0.833), the rise time (t 15 ϭ 0.35, p ϭ 0.731) and decay time constant (t 15 ϭ 0.48, p ϭ 0.638) were not significantly different between in vehicle-and JZL-treated slices (Fig.  4C-E). These results indicate that JZL184 treatment does not cause significant delayed activation of mGluR1 in PCs. Thus, mechanisms downstream of mGluR1 activation may account for the observed delayed onset of SSE following loss of MAGL function.

The spread of endocannabinoid signaling did not occur at spatially dispersed PF synapses
The axons of granule cells ascend to the molecular layer and bifurcate once to form parallel fibers (Palay and Chan-Palay, 1974). In the above experiments, PF-EPSCs were evoked by stimulating the PFs in the molecular layer, where PFs form dense and adjacent synapses onto PCs Attwell, 2005, 2007; Fig. 5A, stimulation a). Results from Figures 2 and 3 indicate that disruption of  Figure  2A for the purpose of comparison. B, Bath application of MAGL inhibitor JZL184 (0.3 M) substantially extended SSE (n ϭ 7-9; ‫‪p‬ءءء‬ Ͻ 0.001), and intracellular dialysis of BAPTA and GDP-␤S attenuated SSE (n ϭ 7-8; ‫‪p‬ءءء‬ Ͻ 0.001), and the remaining SSE was abolished by the CB 1 receptor antagonist AM251 (2 M; n ϭ 8 -9, ‫‪p‬ءءء‬ Ͻ 0.001). C, Bath application of glutamate transporter inhibitor TBOA (100 M) extended SSE (n ϭ 8 -9; ‫‪p‬ءءء‬ Ͻ 0.001), and intracellular dialysis of BAPTA and GDP-␤S did not significantly alter SSE (n ϭ 8 -9; ‫‪p‬ءءء‬ Ͻ 0.001). D, Summary of changes in EPSC amplitude under experimental conditions shown in A-C (n ϭ 7-9; ‫‪p‬ءءء‬ Ͻ 0.001, n.s. p Ͼ 0.05).
MAGL or glutamate reuptake enables endocannabinoid signaling to spread among dense and adjacent PF-PC synapses. PF-PC EPSCs can also be evoked by placing the stimulating electrode in the granular layer, which activates granule cell axons that ascend to different levels in the molecular layer and form spatially dispersed PF synapses onto PCs Attwell, 2005, 2007;Pachoud et al., 2014;Fig. 5A, stimulation b). It has been shown that tetanic stimulation in the granular layer does not recruit endocannabinoid signaling unless glutamate uptake is blocked by the glutamate transporter inhibitor TBOA Attwell, 2005, 2007). We next determined whether endocannabinoid signaling could spread among spatially dispersed PF axonal terminals that innervate the recorded PCs.
Consistent with previous studies Attwell, 2005, 2007), we found that tetanic stimulation (50 Hz, 1 s) in the granular layer induced PTP of EPSCs in wild-type slices. In the presence of glutamate transporter inhibitor TBOA (100 M), the same tetanic stimulation caused transient synaptic suppression (F (3,33) ϭ 27.63, p Ͻ0.001; TBOA vs control: p Ͻ 0.001; Fig. 5B,C). This synaptic suppression was abolished by the CB 1 R antagonist AM251 (2 M), resulting in PTP of EPSCs (p Ͻ0.001; Fig.  5B,C). Can the endocannabinoid signal spread among spatially dispersed synapses onto PCs? To test this possibility, we blocked endocannabinoid release from the recorded PC with intracellular dialysis of BAPTA (20 mM) and GDP-␤S (2 mM). We found that BAPTA and GDP-␤S abolished the transient post-tetanic synaptic suppression induced in the presence of TBOA (p Ͻ0.001; Fig. 5B,C). These results suggest that endocannabinoid signaling does not spread between spatially dispersed PF synapses even when glutamate uptake is blocked by TBOA.
We next determined whether disruption of MAGL enabled endocannabinoid-mediated synaptic suppression at spatially dispersed synapses onto PCs. In wild-type slices that were perfused with MAGL inhibitor JZL184 (0.3 M), tetanic stimulation (50 Hz, 1 s) of the granular layer induced PTP of EPSCs in PCs (F (2,24) ϭ 0.41, p ϭ 0.667; Fig. 5D). The same stimulation did not induce synaptic suppression in PF-PC synapses in MAGL-TKO slices (p Ͼ 0.05; Fig. 5D). Thus, pharmacological blockade and global knock-out of MAGL did not cause endocannabinoid-mediated synaptic suppression at spatially dispersed PF-PC synapses.

Disruption of MAGL and glutamate reuptake did not enable 2-AG signal to spread from PCs to Golgi cells
PFs of granule cells form excitatory synapses onto PCs and Golgi cells (GCs). However, DSE and tetanic stimulation-induced depression could not be induced at PF-GC synapses even though the CB 1 R agonist WIN52122-2 induced robust depression of EPSCs. The failure to induce endocannabinoid-mediated synaptic depression could be explained by the inability of the GCs to release 2-AG (Beierlein et al., 2007). Because 2-AG could spread among PF-PC synapses following disruption of MAGL and glutamate reuptake and PFs innervate both PCs and GCs (Beierlein et al., 2007; Fig. 6A), we explored whether 2-AG released from PCs could spread to PF-GC synapses. To this possibility, we examined whether posttetanic synaptic suppression could be induced in GCs following disruption of MAGL or glutamate reuptake. Tetanic stimulation of PFs in the molecular layer (50 Hz, 1 s) induced small PTP of EPSCs in the GCs in wild-type control slices, and treating the slices with the MAGL inhibitor JZL184 (0.3 M) or glutamate transporter blocker TBOA (100 M) did not significantly alter the PTP (F (3,32) ϭ 0.64, p ϭ 0.594; Fig. 6B). Similar PTP of EPSCs was induced in the GCs in MAGL TKO slices (Fig. 6B). These results suggest that the 2-AG signal generated from PCs during the tetanic stimulation of PFs does not spread from PF-PC synapses to PF-GC synapses even after gross disruption of MAGL or glutamate reuptake.

Discussion
Here we have shown that both neuronal and astrocytic MAGL contribute to the termination of SSE at PF-PC synapses and limit the spread of endocannabinoid signaling in the cerebellum. We found that global knock-out of MAGL substantially prolonged SSE, whereas neuron-and astrocyte-specific knock-out of MAGL produced a less dramatic but significant prolongation of SSE. In addition, our results suggest that the spread of endocannabinoid signaling occurred only after global knock-out of MAGL or pharmacological blockade of either MAGL or glutamate uptake, whereas no spread was detected following neuron-or astrocyte-specific deletion of MAGL.
Pharmacological inhibition and global knock-out of MAGL substantially prolongs DSE at PF-PC synapses in cerebellar slices (Pan et al., 2009;Tanimura et al., 2012;Viader et al., 2015), whereas neuron-and astrocyte-specific deletion of MAGL produces modest but significant prolongation of DSE (Viader et al., 2015). Further, viral transduction of Bergmann glia with MAGL in a MAGL global knock-out background significantly shortens DSE (Tanimura et al., 2012). We extended these studies in an important way by examining the effects of global and cell-type-specific knock-out of MAGL on SSE. A brief tetanic stimulation of PF terminals in the molecular layer activates group I mGluRs to induce the release of 2-AG, which produces SSE in PCs (Maejima Figure 5. There was no detectable spread of endocannabinoid signaling induced in spatially dispersed PF-PC synapses. A, Stimulation electrode "a" was placed in molecular layer to activate dense and spatially adjacent PF-PC synapses, and this protocol has been used to induce SSE in Figures 1-3. Stimulation electrode "b" was placed in granular layer to activate spatially dispersed PF-PC synapses, and this protocol was used in Figure 5. B, A brief tetanic stimulation (50 Hz, 1 s) in granular layer induced PTP of EPSCs in PCs. Bath perfusion of glutamate transporter TBOA (100 M) enabled endocannabinoid-mediated synaptic suppression, which was blocked by bath perfusion of CB 1 R antagonist AM 251 (2 M) or intracellular dialysis of BAPTA (20 mM) and GDP-␤S (2 mM; n ϭ 7-10). The latter result suggests that there was no detectable spread of 2-AG signal induced in spatially dispersed PF-PC synapses even after blockade of glutamate reuptake. C, Summary of peak amplitude of EPSCs under experimental conditions shown in (B; ‫‪p‬ءءء‬ Ͻ 0.001). D, The tetanic stimulation induced PTP of EPSCs in wild-type slices treated with JZL184 (0.3 M) and in MAGL-TKO slices (n ϭ 7-10; p Ͼ 0.05). Brown et al., 2003;Marcaggi and Attwell, 2005;Tanimura et al., 2009;. We found that neuron-and astrocyte-specific knock-out produced similar, modest extension of SSE at PF-PC synapses, whereas global MAGL knock-out caused much greater extension of SSE. Thus, global and cell-type-specific knock-out of MAGL exert similar effects on DSE and SSE at PF-PC synapses (Viader et al., 2015). Immunohistochemical studies have shown that MAGL is highly enriched in the PF terminals in the molecular layer but is also expressed in Bergmann glia (Tanimura et al., 2012) and the astrocytes in the cerebellum (Nolte et al., 2001), which may explain why neuron-and astrocyte-specific knockout of MAGL extends SSE at PF-PC synapses.
Endocannabinoids are thought to be produced ondemand in activated neurons (Piomelli, 2003). Interestingly, global knock-out of MAGL and the MAGL inhibitor JZL184 delayed the onset of SSE Bergerot et al., 2013) but not DSE (Pan et al., 2009;Viader et al., 2015). SSE is triggered by activation of extrasynaptic mGluR1 and subsequent recruitment of endocannabinoid signaling (Maejima et al., , 2005Brown et al., 2003;Marcaggi and Attwell, 2005;Tanimura et al., 2009;. We found that the latency, rise time, and decay time of mGluR1-mediated slow EPSCs were not significantly different between slices treated with vehicle or MAGL inhibitor JZL184. These results argue against the notion that mGluR1 activation is delayed following loss of function of MAGL. Previous studies have shown that the onset of DSE or DSI was not significantly altered by JZL184 or global knock-out of MAGL (Pan et al., 2009;Tanimura et al., 2012;Viader et al., 2015). Thus, the loss of the function of MAGL does not significantly delay the Ca 2ϩ -induced release of 2-AG, nor does it delay the CB 1 receptor response to 2-AG. One possibility is that loss of MAGL function causes delayed 2-AG release in response to mGluR activation, whereas the onset of Ca 2ϩ -induced release of 2-AG is not altered.
A comprehensive profile of brain serine hydrolases revealed that about 85% of total 2-AG in mouse brain is hydrolyzed by MAGL and the remaining 15% is hydrolyzed by other serine hydrolases including serine hydrolase alpha-beta-hydrolase domain 6 and 12 (ABHD6 and ABHD12; Blankman et al., 2007). Previous studies have shown that selective ABHD6 inhibitor WWL123 and WWL70 had no significant effect on DSE and DSI in cerebellar Purkinje cells in wild-type and MAGL global knock-out mice Tanimura et al., 2012). These results suggest that ABHD6 does not significantly contribute to the termination of endocannabinoid-mediated retrograde synaptic depression. The development of a selective ABHD12 inhibitor in the future would enable the examination of the role of ABHD12 in regulating endocannabinoid signaling.
As a diffusible retrograde messenger, endocannabinoids may spread to unstimulated synapses to cause heterosynaptic depression. The spread appears to be varied among different cell types. DSI could spread among hippocampal CA1 pyramidal neurons that are in close proximity (Wilson and Nicoll, 2001). In the cerebellum, DSE does not spread to neighboring PCs at physiological temperature, and the effect of DSI could "spread" to neighboring PCs through inhibition of interneuron firing rather than physical diffusion of 2-AG (Vincent and Marty, 1993;Kreitzer et al., 2002). mGluR stimulation-induced endocannabinoid release could spread to neighboring PCs, resulting in depression of IPSCs (Galante and Diana, 2004). Intracellular dialysis of GDP-␤S or GTP-␥S blocked mGluR agonist DHPG-induced endocannabinoid-mediated retrograde suppression at CF-PC synapses . Of particular relevance to the present study is the finding that SSE does not spread to other PFs activated by a separate stimulating electrode, and blocking endocannabinoid release in the recorded PC with BAPTA and GDP-␤S abolished SSE (Brown et al., 2003), suggesting that endocannabinoids released from neigh-