Synapse type-independent degradation of the endocannabinoid 2-arachidonoylglycerol after retrograde synaptic suppression

Abstract

The endocannabinoid 2-arachidonoylglycerol (2-AG) mediates retrograde synaptic suppression. Although the mechanisms of 2-AG production are well characterized, how 2-AG is degraded is less clearly understood. Here we found that expression of the 2-AG hydrolyzing enzyme monoacylglycerol lipase (MGL) was highly heterogeneous in the cerebellum, being rich within parallel fiber (PF) terminals, weak in Bergman glia (BG), and absent in other synaptic terminals. Despite this highly selective MGL expression pattern, 2-AG–mediated retrograde suppression was significantly prolonged at not only PF-Purkinje cell (PC) synapses but also climbing fiber-PC synapses in granule cell-specific MGL knockout (MGL-KO) mice whose cerebellar MGL expression was confined to the BG. Virus-mediated expression of MGL into the BG of global MGL-KO mice significantly shortened 2-AG–mediated retrograde suppression at PF-PC synapses. Furthermore, contribution of MGL to termination of 2-AG signaling depended on the distance from MGL-rich PFs to inhibitory synaptic terminals. Thus, 2-AG is degraded in a synapse-type independent manner by MGL present in PFs and the BG. The results of the present study strongly suggest that MGL regulates 2-AG signaling rather broadly within a certain range of neural tissue, although MGL expression is heterogeneous and limited to a subset of nerve terminals and astrocytes.

Endogenous cannabinoids (endocannabinoids) are lipid mediators that are released from postsynaptic neurons in activity-dependent manners (1–3). These endocannabinoids travel backward to presynaptic terminals, bind to cannabinoid CB₁ receptors, and induce transient or persistent suppression of neurotransmitter release (1, 3). Anandamide (4) and 2-arachidonoylglycerol (2-AG) (5, 6) are known as two major endocannabinoids in the CNS. Genetic deletion of the 2-AG synthesizing enzyme diacylglycerol lipase-α (DGLα) in mice results in elimination of endocannabinoid-mediated retrograde suppression of synaptic transmission in the cerebellum (7), striatum (7), and hippocampus (7, 8). Thus, 2-AG produced by DGLα is regarded as a major endocannabinoid that mediates retrograde signaling at central synapses.

The endocannabinoid 2-AG is known to be produced by strong depolarization of postsynaptic neurons and the following elevation of Ca²⁺ concentration [Ca²⁺-driven endocannabinoid release (Ca²⁺-ER)] (9–11), strong activation of postsynaptic G_q/11-coupled receptors at basal Ca²⁺ level [basal receptor-driven endocannabinoid release (basal RER)] (12), or combined Ca²⁺ elevation and G_q/11-coupled receptor activation (Ca²⁺-assisted RER) (13, 14). Previous studies have clarified detailed subcellular localizations of 2-AG–producing molecules, including group I metabotropic glutamate receptors (mGluRs) (15), G_q/11 (16), phospholipase Cβs (PLCβs) (17), and DGLα (18–22). These molecules are essentially targeted to dendritic spines on which glutamatergic excitatory synapses are formed, suggesting that 2-AG is efficiently produced by excitatory synaptic activity (23). However, CB₁ receptor expression is generally higher at inhibitory synapses than at excitatory synapses in various brain regions (18, 24). Thus, specificity and efficiency of 2-AG–mediated retrograde synaptic suppression are thought to depend not only on the expression level of CB₁ receptors at presynaptic terminals but also on the distance from the postsynaptic 2-AG production sites to the CB₁ receptors on presynaptic terminals (18–22).

Retrograde signaling mediated by 2-AG is generally determined by the balance between the production and clearance of 2-AG. About 85% of 2-AG is reported to be degraded by a serine hydrolase, monoacylglycerol lipase (MGL) (25). The amount of 2-AG in the brain is markedly increased in global MGL knockout (MGL-KO) mice (26, 27). Depolarization-induced suppression of inhibition/excitation (DSI/DSE) as a result of Ca²⁺-ER and mGluR1-mediated synaptic suppression caused by basal RER are significantly prolonged in the hippocampus and cerebellum after treatment of MGL inhibitors (28–30) or in global MGL-KO mice (31, 32). These lines of evidence indicate that MGL significantly contributes to the regulation of 2-AG signaling. Because MGL is found in the cytoplasm of presynaptic terminals (30, 33, 34), it is thought that 2-AG is degraded in a “synapse-specific” manner within the cytoplasm of nerve terminals at which 2-AG suppresses transmitter release by activating CB₁ receptors. However, because not all CB₁⁺ nerve terminals express MGL (35), it is unclear whether MGL regulates 2-AG–mediated suppression only at synaptic terminals that express MGL.

To address this issue, we used cerebellar Purkinje cells (PCs) that receive two distinct excitatory synaptic inputs, parallel fibers (PFs) and climbing fibers (CFs), and inhibitory inputs from two types of GABAergic interneurons, basket cells (BCs) and stellate cells (SCs). These four types of synapse are known to undergo 2-AG–mediated retrograde suppression following depolarization of PCs (DSE or DSI) (7, 36, 37). We found that MGL was expressed richly in PF terminals and weakly in Bergmann glia (BG), but was absent in terminals of CFs, SCs, and BCs. Despite this highly heterogeneous MGL expression pattern, DSE was significantly prolonged not only at PF-PC synapse but also at CF-PC synapse in global MGL-KO mice and in granule cell (GC)-specific MGL-KO mice. DSE at PF-PC synapses was shortened when MGL was expressed into the BG of cultured slices from global MGL-KO cerebellum. Furthermore, in global MGL-KO mice, DSI was significantly prolonged at SC-PC synapses that are surrounded by PFs and located close to BG processes, but not at BC-PC synapses that are remote from MGL-rich PFs. These results suggest that, in the cerebellum, after binding to CB₁ receptors on presynaptic terminals of PF, CF, and SC, 2-AG is degraded in a synapse-nonspecific manner mainly by MGL in PFs and the BG.

Results

MGL Is Expressed Richly in PF Terminals and Weakly in BG.

We began by examining cellular and subcellular distribution of MGL in the cerebellar cortex at postnatal day 12 (P12) and P60 by immunohistochemistry (Fig. 1). The specificity of MGL antibodies has been verified by blank labeling in global MGL-KO brains in our previous studies (22, 35). By double immunofluorescence for MGL and carbonic anhydrase 8 (Car8), a marker of PCs (38), punctate MGL labeling was prominent in the neuropil of the molecular layer (ML), but was absent in Car8-labeled PC dendrites and somata (Fig. 1A). Bright MGL puncta in the ML were often overlapped with small boutons that were immunopositive for type 1 vesicular glutamate transporter (VGluT1), a marker of PF terminals (arrowheads in Fig. 1B) (39). We also noted faint MGL signals in the neuropil, which were not overlapped with but distributed around VGluT1⁺ PF terminals. These MGL signals were found to overlap with signals of antiglutamate/aspertate transporter (GLAST), a glial glutamate transporter particularly rich in BG (Fig. 1C) (40). In marked contrast, no immunoreactivity for MGL was detected in large boutons that were immunopositive for VGluT2, a marker for CF terminals (arrows in Fig. 1B) (39). MGL immunoreactivity was also absent in GABAergic terminals of ML interneurons on the soma and dendrites of PCs, which were labeled with the antibody against vesicular inhibitory amino acid transporter (VIAAT) (Fig. 1D).

Fig. 1.

Localization of MGL in the cerebellar cortex. (A–D) Immunofluorescence for MGL in the cerebellum of P12 mice. (A) Double immunofluorescence for MGL and a PC marker, Car8, showing prominent MGL labeling in the neuropil of the ML and its negativity in Car8-labeled PCs. GCL, granule cell layer. (B) Triple immunofluorescence for MGL, VGluT1, and VGluT2 showing strong MGL labeling in VGluT1-labeled PF terminals (arrowheads) but not in VGluT2-labeled CF terminals (arrows). (C) Double immunofluorescence for MGL and GLAST showing weak MGL labeling in GLAST-positive BG (arrowheads). (D) Triple immunofluorescence for MGL, Car8, and VIAAT showing no immunoreactivity for MGL in VIAAT-labeled interneuron terminals (arrows) located on the surface of Car8-labeled PC dendrites and somata. (E–H) Preembedding silver-enhanced immunogold for MGL in the adult (2 mo of age) cerebellum showing intracellular MGL labeling in PF terminals (E and G) and BG processes (BG, colored in purple; E), but not in CF terminals (F), interneuron terminals (In, G), or PC spines (Sp, E–G) and dendrites (Dn, G). (H) Summary bar graph demonstrates the mean number of metal particles per 1 μm² of each element of wild-type and global MGL-KO mice. The number above each column indicates the number of electron micrographs used for statistical analysis. Error bars represent SEM. [Scale bars: 10 μm (A), 5 μm (B–D); 500 nm (E–G).] *P < 0.05, **P < 0.01 (Mann–Whitney U test).

By preembedding immunogold electron microscopy in the adult cerebellum, metal particles for MGL were distributed intracellularly in PF terminals forming asymmetrical synapses with PC spines (Fig. 1E). In addition, labeling was observed in thin lamellate processes of the BG (Fig. 1E). In contrast, metal particles were rarely detected inside CF terminals (Fig. 1F), interneuron terminals (Fig. 1G), or PC spines, dendrites, and somata (Fig. 1 E–G). The density of MGL labeling was calculated as the mean number of metal particles per 1 μm² of each element in wild-type and global MGL-KO mice. Statistically significant MGL labeling was found only in PF terminals and the BG (Fig. 1H). Notably, the density in PF terminals was 2.6-times higher than that of the BG (Fig. 1H). Taken together, these immunohistochemical results demonstrate that MGL is expressed strongly in PF terminals and weakly in BG processes, but is absent in CF, SC, and BC terminals.

MGL Expressed in PF Terminals Facilitates Termination of CF-DSE.

Next, we examined how genetic deletion of MGL affected endocannabinoid-mediated retrograde synaptic suppression in the cerebellum. A recent study showed that DSE at PF-PC and CF-PC synapses were prolonged in global MGL-KO mice (32). We confirmed that both PF-DSE and CF-DSE were significantly prolonged in global MGL-KO mice, compared with their wild-type littermates (Fig. S1), with no changes in presynaptic CB₁ receptor sensitivity (Fig. S2A), depolarization-induced Ca²⁺ transients (Fig. S2B), basic properties of synaptic transmission (Table S1), and expression and localization of key molecules for retrograde 2-AG signaling (mGluR1α, PLCβ4, DGLα, CB₁) (Fig. S3).

These results on global MGL-KO mice raise a question as to why DSE is prolonged at CF-PC synapses, whereas no MGL expression is detected in their presynaptic terminals (Fig. 1). One possibility would be that MGL, which is expressed strongly in PF terminals and weakly in BG, hydrolyzes 2-AG around CF terminals and regulates the duration of DSE at CF-PC synapse. To clarify the contribution of MGL in PF terminals to the regulation of DSE, we generated mice with GC-specific deletion of MGL (GC-specific MGL-KO mice) and examined DSE at CF-PC and PF-PC synapses. To obtain a GC-specific Cre recombinase expression, we used a E3CreN line (GluN2C^+/iCre), the Cre gene of which was expressed in GCs under the control of a GluN2C (GluRɛ3) promoter (41). By intercrossing MGL-floxed (MGL^lox/lox) mice with the E3CreN line, we created GC-specific MGL-KO mice. We confirmed that MGL immunoreactivity in the ML was decreased greatly in GC-specific MGL-KO mice compared with wild-type mice (Fig. 2 A, B, D, and E). However, weak signals for MGL remained compared with global MGL-KO mice (Fig. 2 B, C, E, and F). Triple immunofluorescent labeling experiments revealed that VGluT1-labeled PF terminals (double arrowheads in Fig. 2 G–I) exhibited intense immunoreactivity for MGL in wild-type (Fig. 2G3), but not in GC-specific MGL-KO (Fig. 2H3) or global MGL-KO (Fig. 2I3) mice. In contrast, GLAST-labeled BG processes (arrowheads in Fig. 2 G–I) exhibited weak immunoreactivity for MGL in wild-type (Fig. 2G3) and GC-specific MGL-KO mice (Fig. 2H3), but not in global MGL-KO mice (Fig. 2I3). These results indicate that MGL immunoreactivity was eliminated from PF terminals but present in BG processes in GC-specific MGL-KO mice.

Fig. 2.

PF-specific deletion of MGL in GC-specific MGL-KO mice. (A–F) Immunofluorescence for MGL in parasagittal cerebellar sections in wild-type (A and D; WT), GC-specific MGL-KO (B and E; GC-KO), and global MGL-KO (C and F; KO) mice at P20–P25. D–F are enlarged images in the cerebellar cortex. PCL, Purkinje cell layer; GCL, granule cell layer. (G–I) Triple immunofluorescence for MGL (red), GLAST (green), and VGluT1 (blue) in the ML of WT, GC-specific MGL-KO, and global MGL-KO mice. Single and double arrowheads depict VGluT1-labeled PF terminals and GLAST-labeled BG processes, respectively. [Scale bars: 1 mm (A–C), 20 μm (D–F), 2 μm (G–I).]

We examined DSE at the CF-PC synapse following depolarization of PCs (5-s duration, −70 mV to 0 mV). As shown in Fig. 3A, DSE at the CF-PC synapse was significantly prolonged in GC-specific MGL-KO mice compared with wild-type mice. Interestingly, the DSE prolongation in GC-specific MGL-KO mice was less prominent compared with global MGL-KO mice (Fig. 3A). Although the peak magnitudes of DSE measured at 5–10 s after depolarization were similar among the three mouse strains, the DSE magnitude of GC-specific MGL-KO mice at 40–50 s after depolarization was significantly larger than those of wild-type mice, but significantly smaller than global MGL-KO mice (Fig. 3B). Consequently, the duration of DSE measured as the time to reach 50% recovery from the peak DSE was in the order of global MGL-KO mice, GC-specific MGL-KO mice, and wild-type mice (Fig. 3C).

Fig. 3.

DSE is prolonged at CF-PC and PF-PC synapses in GC-specific and global MGL-KO mice. (A and D) Sample traces and averaged time courses of CF-EPSCs (A) and PF-EPSCs (D) from PCs of wild-type (open circles, P25–P30), GC-specific MGL-KO (GC-KO) (gray circles, P25–P30), and global MGL-KO (KO) (closed circles, P25–P30) mice before and after depolarization. Traces obtained before (1, 4, 7, black), 5–10 s after (2, 5, 8, blue), and 40–45 s after (3, 6, 9, red) depolarization are superimposed. (B and E) Summary bar graph showing the magnitude of CF-DSE (B) and PF-DSE (E) at the time points 5–10 s (Peak) and 45–50 s after (50 s) depolarization. (C and F) Duration of DSE (half-width) measured as the time to reach 50% recovery from the maximum magnitude of DSE. In B, C, E, and F, the number of cells for each experiment is indicated in parenthesis. Calibration bars: 2 nA and 5 ms for (CF-DSE: WT and GC-KO), 1 nA and 2 ms for (CF-DSE: KO), 50 pA and 5 ms for (PF-DSE: WT and GC-KO), and 0.1 nA and 5 ms for (PF-DSE: KO). Data are presented as means ± SEM. *P < 0.05, **P < 0.01 (one-way ANOVA with Fisher least-significant difference multiple-comparison post hoc test).

We then examined DSE at PF-PC synapse following depolarization of PCs (3-s duration, −70 mV to 0 mV) in the three strains of mice. As shown in Fig. 3D, DSE in GC-specific MGL-KO mice was significantly prolonged compared with wild-type mice, but recovered significantly faster compared with global MGL-KO mice (Fig. 3D). The magnitude and duration of DSE in GC-specific MGL-KO mice were intermediate between those of wild-type and global MGL-KO mice (Fig. 3 E and F). These results indicate that MGL in PF terminals facilitates termination of 2-AG signaling not only “homosynaptically” at PFs but also “heterosynaptically” at CFs. The difference in DSE duration and magnitude between GC-specific and global MGL-KO mice strongly suggest that MGL present in cerebellar tissues other than PF terminals significantly contributes to the termination of 2-AG signaling. Because MGL is present in the BG in GC-specific MGL-KO mice (Fig. 2H), glial MGL is considered to be important for termination of 2-AG signaling.

MGL Within the BG Facilitates Termination of 2-AG–Mediated Retrograde Signaling.

To directly examine whether MGL in the BG can influence the termination of 2-AG signaling in the cerebellum, we developed a reduced cerebellar preparation in which MGL was expressed richly in the BG but absent in PFs. We made organotypic cerebellar slice cultures from global MGL-KO mice at P9 or P10, and expressed the MGL gene into the BG by using a lentivirus vector (Fig. S4). We confirmed that immunoreactivity of MGL was overlapped with that of GLAST (Fig. S4A) but not with that of VGluT1 (Fig. S4B), indicating that MGL was richly expressed in the BG. We then made whole-cell recordings from PCs, stimulated PFs in the ML, and recorded PF-excitatory postsynaptic currents (EPSCs). Slice cultures from global MGL-KO mice with overexpression of GFP alone into the BG were used as control. We checked that CB₁ receptor sensitivity was not altered at PF-PC synapses in slice cultures with MGL expression in the BG compared with control slices (Fig. S4C). We then elicited DSE at the PF-PC synapse by applying 10 depolarization pulses (100-ms duration, −70 mV to 0 mV, repeated at 1 Hz) to PCs. We found that DSE recovered significantly faster in slice cultures with MGL expression in BG than in control cultures (Fig. 4A). Magnitude of DSE at the peak and 130 s after depolarization as well as the duration of DSE were significantly smaller in slice cultures with MGL expression than in MGL-KO cultures (Fig. 4 B and C). These results suggest that MGL in the BG can influence the duration of DSE.

Fig. 4.

MGL expressed in BG shortens PF-DSE. (A) Sample traces and averaged time courses of PF-EPSCs from PCs in cerebellar slice cultures with overexpression of GFP alone (KO, open circles) or MGL plus GFP (KO + MGL, closed circles) into BG from global MGL-KO mice before and after applying 10 depolarizing pulses to PCs (100-ms duration, from −70 mV to 0 mV, repeated at 1 Hz). Traces obtained before (1, 4, black), 5–10 s after (2, 5, blue), and 125–130 s after (3, 6, red) depolarization are superimposed. (B and C) Summary bar graph showing the magnitude (B) and duration (C) of PF-DSE illustrated similarly to Fig. 3. Calibration bars: 0.1 nA and 5 ms. Data are presented as means ± SEM. **P < 0.01, ***P < 0.001 (Mann–Whitney U test).

MGL Facilitates Termination of DSI at SC-PC Synapses.

The results so far indicate that MGL in PF terminals and the BG hydrolyzes 2-AG in a synapse nonspecific manner and facilitates termination of 2-AG–mediated retrograde signaling at both PF-PC and CF-PC synapses. MGL immunoreactivity is strong in the ML and very weak in the PC layer (PCL) and GC layer in the cerebellum (Fig. 1 A2 and D2). Because CF terminals are surrounded by PF terminals and enwrapped by BG processes, 2-AG around CF terminals is thought to be degraded by MGL in PF terminals and the BG. Therefore, spatial disposition of MGL seems to be important for its regulation of 2-AG–mediated retrograde signaling. GABAergic synapses from BCs and SCs to PCs undergo 2-AG–mediated transient suppression following depolarization of PCs, a phenomenon known as DSI (42–44). However, GABAergic nerve terminals in the ML and PCL, which represent SC and BC terminals, respectively, do not express MGL (Fig. 1 D, G, and H). Although SCs form GABAergic synapses on PC dendrites that are embedded within a cloud of PF terminals, BCs innervate PC somata that are distant from PF terminals. Therefore, DSI at SC synapses is thought to be influenced more strongly by MGL than DSI at BC synapses. We thus compared DSI at BC and SC synapses between wild-type and global MGL-KO mice.

For differentially stimulating axons of BCs and SCs, we placed stimulation pipettes near the PCL and in the outer half of the ML, respectively (Fig. S5A). Inhibitory postsynaptic currents (IPSCs) evoked by stimulation near the PCL were thought to arise mostly from BC axons forming synapses on the PC soma, whereas those evoked by stimulation at the outer ML are considered to originate mainly from SC axons contacting PC dendrites (Fig. S5A). To estimate the extent of overlap between the activated axons by stimuli at the two sites, we used homo- and heterosynaptic paired-pulse protocols. When extracellular Ca²⁺ concentration was elevated, paired stimuli delivered to each site individually (homosynaptic paired stimulation) at 30-ms intervals resulted in clear depression of the second IPSCs (Fig. S5B). In contrast, when stimulation at one site was preceded by stimulation at the other site at 30-ms intervals (heterosynaptic paired stimulation), no significant depression was observed (Fig. S5B). These results indicate that stimuli at the ML and near the PCL activated nonoverlapping population of inhibitory axons.

Because synaptic currents arising from synapses distant from the somatic recording site undergo stronger distortion and have longer rise times than those arising from the soma (45, 46), we checked the 10–90% rise times of IPSCs to estimate the sites of GABAergic synapses along the somatodendritic domain of PCs. We found that about 90% of the IPSCs by stimulation near the PCL had rise times shorter than 0.8 ms, whereas about 90% of IPSCs by stimulation at the outer ML had rise times longer than 0.6 ms (Fig. S5C). To exclude the possible overlap between the two populations, we determined IPSCs with rise times shorter than 0.6 ms as arising from putative BC axons and those with rise times longer than 0.8 ms as being evoked by stimulating putative SC axons.

We compared DSI by applying five depolarizing pulses (100-ms duration, from −70 mV to 0 mV, repeated at 1 Hz) to PCs between wild-type and global MGL-KO mice. For IPSCs arising from putative SC axons (i.e., rise time > 0.8 ms), DSI was significantly prolonged in global MGL-KO mice compared with wild-type mice (Fig. 5A). Although the peak magnitudes of DSI measured at 5–10 s after depolarization were similar between the two mouse strains, the DSI magnitude of global MGL-KO mice at 40–50 s after depolarization was significantly larger than wild-type mice (Fig. 5B). The duration of DSI measured as the time to reach 50% recovery from the peak DSI was significantly longer in global MGL KO mice than in wild-type mice (Fig. 5C). In marked contrast, for IPSCs arising from putative BC axons (i.e., rise time < 0.6 ms), DSI was not prolonged in global MGL-KO mice (Fig. 5D). Although the peak magnitude of DSI was greater in global MGL-KO mice, the magnitude at 40–50 s after depolarization and the duration of DSI were not different between wild-type and global MGL-KO mice (Fig. 5 E and F).

Fig. 5.

DSI is prolonged at SC-PC but not BC-PC synapses in global MGL-KO mice. (A and D) Sample traces and averaged time courses of IPSCs at putative SC-PC (A) and BC-PC (D) synapses in wild-type (WT) (open circles, P9–P12) and global MGL-KO (KO) (closed circles, P9–P12) mice before and after five depolarization pulses (100-ms duration, from −70 mV to 0 mV, repeated at 1 Hz). Traces obtained before (1, 4, black), 5–10 s after (2, 5, blue), and 45–50 s after (3, 6, red) depolarization are superimposed. (B, C, E, and F) Summary bar graph showing the magnitude (B and E) and duration (C and F) of DSI illustrated similarly to Fig. 3. Calibration bars: 0.1 nA and 5 ms for A and 0.5 nA and 5 ms for D. Data are presented as means ± SEM. *P < 0.05 (Mann–Whitney U test); n.s., not significant.

We also examined whether pharmacological blockade of MGL in wild-type mice had the same effects on the two types of IPSCs. Incubation of slices with the MGL inhibitor N-arachidonoyl maleimide and 4-nitrophenyl 4-(dibenzo[d][1,3]dioxol-5-yl(hydroxy)methyl)piperidine-1-carboxylate (JZL 184) (47) (100 nM) caused a significant prolongation of DSI at putative SC-PC synapses (Fig. S6 A–C) but had no significant effect on DSI at putative BC-PC synapses (Fig. S6 D–F). Finally, we examined whether other endocannabinoid degradation enzymes, α/β-hydrolase domain 6 (ABHD6) and fatty acid amide hydrolase, contributed to the time course of DSI at putative BC-PC synapses. Incubation of slices from global MGL-KO mice with the ABHD6 inhibitor N-methyl-N-[[3-(4-pyridinyl)phenyl]methyl]-4′-(aminocarbonyl)[1,1′-biphenyl]-4-yl ester, carbamic acid (WWL70) (10 μM) and the fatty acid amide hydrolase inhibitor 3′-(aminocarbonyl)[1,1′-biphenyl]-3-yl)-cyclohexylcarbamate (URB597) (1 μM) had no effect on DSI at putative BC-PC synapses (Fig. S7). Together, these results indicate that MGL is important for the termination of DSI at SC-PC synapses, whereas none of the three endocannabinoid degradation enzymes contribute significantly to shaping DSI at BC-PC synapses.

Discussion

We examined detailed localization of MGL by immunohistochemistry and found that MGL was expressed abundantly in PF terminals and weakly in the BG, whereas MGL expression was very low or absent in presynaptic terminals of CF and inhibitory terminals of BCs and SCs. Despite this highly heterogeneous MGL expression pattern, DSE was significantly prolonged not only at the PF-PC synapse but also at the CF-PC synapse in global MGL-KO mice. Furthermore, DSE was prolonged both at PF-PC and CF-PC synapses in GC-specific MGL-KO mice in which cerebellar MGL expression was confined to the BG. This result indicates that MGL in PFs regulates 2-AG signaling not only homosynaptically at PF-PC synapses but also heterosynaptically at CF-PC synapses. Importantly, prolongation of DSE in GC-specific MGL-KO mice was less prominent than in global MGL KO mice, suggesting that MGL in BG also contributes to termination of 2-AG signaling. This notion was supported by the observation that DSE at PF-PC synapses was shortened when MGL was expressed into the BG of cultured slices from global MGL-KO cerebellum. We also found that DSI was prolonged at SC-PC synapses but not at BC-PC synapses in global MGL-KO mice. SC terminals are surrounded by PFs and BG processes in the ML, whereas BC terminals are distant from these structures. Thus 2-AG, which is released from PCs and causes retrograde suppression of PF-, CF-, and SC-PC synapses, is degraded in a synapse-type independent manner by MGL present in the cytoplasm of PF terminals and the BG.

The 2-AG that acts as a synaptic retrograde messenger is produced exclusively by DGLα, (7, 8) which is essentially targeted to dendritic spines (18–22). In PCs, DGLα is expressed densely at the base of the spine neck and sparsely on the somatodendritic membrane, but is excluded from the main body of the spine neck and head (21). On the other hand, CB₁ is highly enriched at perisynaptic region of PFs within 500 nm from the edge of synaptic junction (24), and preferentially accumulated on the synaptic side facing dendritic spines (21). CB₁ is also rich on GABAergic terminals that form synaptic contacts on dendritic shafts and somata (24), with much lower levels of DGLα expression compared with the spine neck and head (21). These molecular arrangements suggest that 2-AG travels a certain distance from its main production site (the base of the spine neck) to its target (CB₁ receptor) on presynaptic terminals of PFs, CFs, or GABAergic axons. Because the ML is packed with PF terminals and BG processes, 2-AG released from PCs may be efficiently degraded by MGL in these two structures irrespective of its target. The 2-AG around presynaptic terminals of CFs and GABAergic axons may be degraded similarly to that around PF terminals. Hence, there is no apparent synapse specificity in terms of 2-AG degradation in the ML. It is not clear how extracellular 2-AG is eventually degraded by MGL located in the cytoplasm. One possibility would be that 2-AG is incorporated into membrane, caught by MGL attached to the inner leaflet of the membrane, and then hydrolyzed into arachidonic acid and glycerol.

Heterogeneous expression of MGL is also found in other brain regions. In the dentate gyrus, MGL is expressed in astrocytes and in some GABAergic inhibitory terminals of both CB₁ receptor-positive and -negative interneurons (35). In marked contrast, MGL is absent in excitatory mossy-cell terminals, although DGLα is abundantly expressed in the spine neck of dentate granule cell (DGC) (35). Each DGC spine was found to be innervated by a single mossy-cell terminal and also contacted nonsynaptically by other mossy-cell terminals (35). This molecular and morphological configuration suggests that 2-AG produced at DGC spine neck is readily accessible to neighboring mossy cell-DGC synapses and to inhibitory synapses nearby. It is interesting to test whether DSE is prolonged in MGL-KO mice at mossy cell-DGC synapses where MGL expression is lacking.

MGL is considered to be important for determining the extent of 2-AG diffusion in brain tissues. In the cerebellum, PC depolarization suppresses the firing rates of neighboring interneurons and reduces inhibitory inputs to PCs (48). These effects are now considered to be mediated by 2-AG released by PC depolarization (49). Thus, 2-AG–mediated retrograde signaling is not strictly synapse-specific but is essentially diffusible and can affect neighboring neurons and synapses within a certain distance from the site of 2-AG production. Spatial arrangement of DGLα, CB₁ receptor and MGL is unique to each brain area and, therefore, the magnitude and the extent of diffusion of 2-AG signaling are different. The results of the present study suggest that MGL regulates 2-AG signaling rather broadly within a certain range of neural tissue, although MGL expression is heterogeneous and limited to a subset of nerve terminals and astrocytes.

Materials and Methods

Animals.

All experiments were performed according to the guidelines for the care and use of laboratory animals of the University of Tokyo, Hokkaido University, and Niigata University. Global MGL-KO mice were generated as described previously (35). GC-specific MGL-KO mice were obtained by crossing MGL^lox/lox with an E3CreN line (GluN2C^+/iCre) whose Cre gene was expressed in GCs under the control of a GluN2C (GluRɛ3) promoter (41).

Immunohistochemistry.

Under deep pentobarbital anesthesia (100 mg/kg of body weight), mice at P12, P20–25, and P60 were fixed by transcardial perfusion with 4% (wt/vol) paraformaldehyde/0.1M phosphate buffer (PB, pH 7.4) for immunofluorescence microscopy and with 4% (vol/vol) paraformaldehyde/0.1% glutaraldehyde/0.1 M PB for immunoelectron microscopy. Details of the procedures for immunofluorescence and immunoelectron microscopic analyses are described in SI Materials and Methods.

Electrophysiology.

For preparing acute cerebellar slices, mice at P8–P30 were decapitated under CO₂ anesthesia, and brains were rapidly removed and placed in chilled external solution (0–4 °C). Parasagittal cerebellar slices (250-μm thick) were prepared by a vibratome (Leica) as described (12, 14). For preparing organotypic cultures, cerebellar slices (250-μm thick) were taken from the vermis of MGL-KO mice at P9 or P10, and cultured on a membrane filter. Whole-cell recordings were made from visually identified PCs in acute or cultured cerebellar slices at 32 °C using an upright microscope (Olympus). Compositions of external and pipette solutions, protocols of stimulation and recording, the design of viral vectors and the details of their transfection to cultured cerebellar slices are described in SI Materials and Methods.