Endocannabinoids control hippocampal inhibitory synaptic transmission through activation of presynaptic CB1 receptors. During depolarization-induced suppression of inhibition (DSI), endocannabinoids are synthesized upon postsynaptic depolarization. The endocannabinoid 2-arachidonoylglycerol (2-AG) may mediate hippocampal DSI. Currently, the best studied pathway for biosynthesis of 2-AG involves the enzyme diacylglycerol lipase (DAGL). However, whether DAGL is necessary for hippocampal DSI is controversial and was not systematically addressed. Here, we investigate DSI at unitary connections between CB1 receptor-containing interneurons and pyramidal neurons in CA1. We found that the novel DAGL inhibitor OMDM-188, as well as the established inhibitor RHC-80267, did not affect DSI. As reported previously, effects of the DAGL inhibitor tetrahydrolipstatin depended on the application method: postsynaptic intracellular application left DSI intact, while incubation blocked DSI. We show that all DAGL inhibitors tested block slow self-inhibition in neocortical interneurons, which involves DAGL. We conclude that DAGL is not involved in DSI at unitary connections in hippocampus.
Endocannabinoids mediate both short- and long-term plasticity at many synapses in the brain. The first form of synaptic plasticity in which endocannabinoids were shown to be involved is hippocampal depolarization-induced suppression of inhibition (DSI) (Wilson and Nicoll, 2001). Hippocampal DSI occurs at a subset of inhibitory synapses onto pyramidal neurons coming from CB1 receptor (CB1R)-containing interneurons. At these synapses, a postsynaptic depolarization leads to calcium-dependent synthesis of endocannabinoids, which subsequently travel to the presynaptic inhibitory terminal. Here, endocannabinoids activate the G-protein coupled CB1R, thereby reducing presynaptic release probability for tens of seconds. The endocannabinoid 2-arachidonoyl glycerol (2-AG) (Sugiura et al., 1995) has been proposed to mediate hippocampal DSI, since inhibiting 2-AG degradation increases the duration of DSI (Kim and Alger, 2004; Makara et al., 2005; Hashimotodani et al., 2007). However, sn-1-diacylglycerol lipase (DAGL) α, the major enzyme catalyzing 2-AG formation in the adult brain (Bisogno et al., 2003) has not been found opposite symmetrical, GABAergic synapses in hippocampus (Katona et al., 2006). Furthermore, studies in which DAGLs have been targeted pharmacologically reported conflicting results. Blocking these enzymes pharmacologically suppresses hippocampal DSI in some studies (Hashimotodani et al., 2007) but not in others (Chevaleyre and Castillo, 2003; Edwards et al., 2006; Szabo et al., 2006). So far, the only study showing pharmacological evidence for DAGL involvement in hippocampal DSI was performed on cultured neurons. This raises the question whether DAGL is involved in hippocampal DSI at intact synapses. Here, by using three chemically different DAGL inhibitors, we have investigated DAGL involvement in hippocampal DSI at unitary connections between CB1R-containing interneurons and CA1 pyramidal neurons.
Materials and Methods
All experiments were approved by the Animal Ethical Committee of the VU University Amsterdam, in accordance with Dutch and European law. For DSI experiments, 14- to 21-d-old wild-type C57Bl6 mice or CB1R−/− mice (Marsicano et al., 2002) were decapitated. Transverse hippocampal slices (300 μm) were sectioned in ice-cold solution containing (mm): 126 NaCl, 3 KCl, 10 glucose, 26 NaHCO3, 1.2 NaH2PO4, 1 CaCl2, and 3 MgCl2 (carboxygenated with 5% CO2/95% O2). For slow self-inhibition (SSI) experiments, sagittal slices containing somatosensory cortex were cut from 21- to 28-d-old FVB-Tg(GadGFP) 45704Swn/J mice expressing EGFP under control of the mouse Gad1 gene promoter (The Jackson Laboratory stock no. 003718), thereby labeling a subset of somatostatin-positive interneurons in hippocampus and neocortex (Oliva et al., 2000). Slices were maintained at 33°C for ∼30 min and recorded at room temperature (20−22°C) in a similar solution, but now with 2 mm CaCl2 and 2 mm MgCl2.
Paired recordings of CB1R-containing interneurons and CA1 pyramidal neurons.
Cellular recordings were performed from 2–6 neurons simultaneously using borosilicate glass pipettes with resistance of 3–5 MΩ, containing the following (in mm): 70 K gluconate, 70 KCl, 4 Mg-ATP, 4 phosphocreatine, 0.4 GTP, 0.5 EGTA and 10 HEPES (pH 7.3, KOH). In synaptically connected neurons, suprathreshold stimulation of the presynaptic interneuron evoked a GABAAR mediated IPSP/IPSC in the postsynaptic CA1 pyramidal neuron. Because of the high intracellular chloride concentration, IPSCs were observed as inward currents, while IPSPs were depolarizing. Presynaptic cells were stimulated with a pair of two suprathreshold current pulses (at ∼10 Hz). Trains were delivered with intervals of 5 s. Recording and stimulus delivery was performed using standard electrophysiological equipment.
SSI recordings in neocortical interneurons.
For experiments on SSI, EGFP expressing interneurons in somatosensory cortex were selected using an UV light source and fluorescence imaging. Extracellular and intracellular solutions were similar as for paired recordings, but to exclude contributions from synaptic conductances DNQX (10 μm), GABAzine (100 μm), and dl-AP5 (100 μm) were added to the extracellular solution.
Some recorded interneurons were filled with biocytin (0.5%), and slices were fixed and resliced after recording. Sections were stained using the antibodies goat-anti-streptavidin Alexa Fluor 488 conjugate (1:200) and the primary antibody rabbit-anti- CB1R (diluted 1:1000 in TBS-TX) (Tsou et al., 1998). In a second step, sections were incubated with the secondary antibody Alexa Fluor 594 goat anti-rabbit IgG (1:200). Colocalization between biocytin and CB1R was determined using a Leica confocal laser-scanning microscope (Leica), the Huygens system (Scientific Volume Imaging), and the Amira isosurface module (Konrad Zuse Center for Information Technology, Berlin, Germany). Care was taken to prevent false positives, incomplete signal separation and crosstalk of dyes. Neuron reconstructions were performed under an up-right microscope (Leica) using a 40× objective with Neurolucida software (Neurolucida, MicroBrightField).
Of tetrahydrolipstatin (THL), RHC-80267 (Sigma-Aldrich) and OMDM-188 (Ortar et al., 2008) appropriate stock solutions were prepared in DMSO (Final concentration <0.2%). Slice-incubation was done in the slice chamber for at least 30 min before recordings started, with the same compound concentration present extracellularly. For intracellular application, intracellular equilibration was allowed for at least 20 min after whole-cell break-in (see Figs. 3, 4). Vehicle controls with intracellular or extracellular DMSO were interleaved with normal control experiments. Because we observed no vehicle effects, all control experiments were grouped.
All analysis was done using custom written procedures in Igor Pro (Wavemetrics). Statistical analysis was done either using two-sided Student's t test (for single comparisons; α = 0.05), or ANOVA with post hoc Dunnett's test (for multiple comparisons; α = 0.05).
Simultaneous whole-cell recordings were made from CA1 pyramidal neurons and interneurons located in stratum radiatum, close to the border with stratum pyramidale (Fig. 1). Cell bodies of selected interneurons had a multipolar appearance and their firing pattern upon current injection showed considerable spike-frequency adaptation (Fig. 1C). To confirm the presence of CB1Rs in the studied interneurons, we used biocytin staining and immunostaining for the CB1R in a subset of interneurons (Fig. 1A,B). All interneurons showed substantial colocalization of biocytin and CB1R immunoreactivity in axons (n = 5/5) (Fig. 1B).
Inhibitory connections were activated by evoking two action potentials (APs) at ∼10 Hz in the interneuron, while recording IPSCs in the CA1 pyramidal neuron (voltage-clamped at −70 mV) (Fig. 1D). In cell pairs showing an inhibitory synaptic connection we tested for the presence of DSI. Two APs were evoked in the interneuron once every 5 s (0.2 Hz). After recording 50 control responses, the subsequent 50 responses were each preceded by a depolarizing step (1 s to 0 mV) to the postsynaptic pyramidal cell (depolarization period) (Fig. 2A). For IPSCs evoked after such a depolarizing step, the IPSC amplitude was reduced to 36.0 ± 2.6% of control (Fig. 2A–C) (p = 0.0001; n = 7), meaning that there was 64.0 ± 2.6% DSI. Amplitude reduction was accompanied by a strong increase in synaptic failure rate (Fig. 2D) (control 30.0 ± 4.6%; depolarization 49.4 ± 5.1%; p = 0.024). After the depolarization period, both IPSC amplitude and failure rate partially recovered, respectively to 76.3 ± 3.1% and 37.1 ± 5.0% (Fig. 2A–D).
The increase in failure rate during the depolarization period is in support of a presynaptic locus of suppression of inhibition. Analysis of the coefficient of variation (CV) of the IPSC amplitudes (Faber and Korn, 1991) showed that the ratio between CV2 during DSI versus CV2 under control was proportional to the relative change in synaptic amplitude (Fig. 2E) (averaged CV−2DSI/CV−2control: 0.38 ± 0.03; amplitude ratio: 0.36 ± 0.03), confirming that the modulation is presynaptic in origin. In CB1R knock-out (CB1R−/−) mice, we did not observe a significant reduction in IPSC amplitude during the depolarization period, and failure rate was unaffected (supplemental Fig. S1, available at www.jneurosci.org as supplemental material). Together, these findings suggest that DSI at these unitary synapses occurs through a decreased presynaptic release probability by CB1R activation.
Activity-dependent endocannabinoid synthesis in several brain regions depends on the DAGL enzymatic activity (Melis et al., 2004; Safo and Regehr, 2005; Szabo et al., 2006). To test for the involvement of DAGLs in hippocampal DSI, we used several DAGL inhibitors. Slices were incubated with the novel potent DAGL inhibitor OMDM-188 (2 μm; Ortar et al., 2008) for at least 30 min before commencing paired recordings in the presence of OMDM-188. Under these conditions, significant DSI was still observed (amount of DSI: 46.2 ± 5.3%; n = 6; p = 0.016) (Fig. 3A), and was accompanied by an increase in failure rate (control: 51.7 ± 1.6%; depolarization: 67.7 ± 2.2%; p = 0.03) (Fig. 3A). The amount of DSI did not differ significantly from control conditions (p > 0.05) (Fig. 3F). Similarly, slices incubated and recorded with the established DAGL inhibitor RHC-80267 (50–70 μm) also showed significant DSI (87.5 ± 2.4%; n = 5; p = 0.009) (Fig. 3B), and an increase in failure rate (control: 54.0 ± 4.7%; depolarization: 92.0 ± 0.6%; p = 0.03) (Fig. 3B).
Recently, several studies have reported opposing effects of extracellular application of the DAGL inhibitor THL on hippocampal DSI (Szabo et al., 2006; Edwards et al., 2008; Hashimotodani et al., 2008). It has been shown that intracellular application of low concentrations of THL is very effective at inhibiting DAGL activity in striatum (500 nm) (Melis et al., 2004) and cerebellum (2 μm) (Safo and Regehr, 2005). Therefore, we added THL (5 μm) to the intracellular solution of the postsynaptic pyramidal neuron, without applying THL extracellularly. To allow intracellular diffusion of THL the depolarization protocol was started at least 20 min after establishing the whole-cell configuration at the pyramidal neuron. Intracellularly applied THL did not affect DSI (70.9 ± 1.6%; n = 15; p = 0.02) (Fig. 3C) (compared to control DSI: p > 0.05) (Fig. 3F). Accordingly, failure rate significantly increased upon depolarization (control: 24.9 ± 1.6%; depolarization: 66.8 ± 2.0%; p < 0.001) (Fig. 3C). Even with a 1 h waiting period for intracellular THL diffusion, DSI was clearly observed (Fig. 3E), suggesting that the ineffectiveness of intracellular THL was not due to incomplete intracellular diffusion. Surprisingly, when slices were incubated extracellularly with higher concentrations of THL (10 μm), DSI was abolished (17.2 ± 2.1%; n = 6; p = 0.10) (Fig. 3D), and failure rate was not affected by depolarization (control: 51.8 ± 3.8%; depolarization: 51.1 ± 3.4%; p = 0.86) (Fig. 3D). To assess whether extracellularly applied THL affected CB1R signaling downstream of the CB1 receptor, we tested the effect of THL on the suppression of IPSCs by the CB1R agonist WIN55,212-2 (2 μm). IPSC suppression by WIN55,212-2 was unaffected in slices incubated in THL, showing that THL incubation did not interfere with CB1R signaling down stream of the receptor (supplemental Fig. S2, available at www.jneurosci.org as supplemental material).
These observations raise the possibility that the other treatments were not effective in inhibiting DAGLs due to suboptimal experimental conditions such as lack of cell penetration. To exclude this possibility, we used the above procedures and inhibitors to block DAGLs in the well-established phenomenon of SSI of neocortical interneurons (Bacci et al., 2004). SSI is induced by repetitive AP firing of somatostatin-positive low-threshold spiking (LTS) interneurons in neocortex (Fig. 4). It is triggered by endocannabinoid-mediated activation of postsynaptic CB1Rs, which in turn activate a prolonged K+ conductance. Recently, it was shown that DAGL inhibition effectively abolishes SSI (Marinelli et al., 2008). We made whole-cell recordings of interneurons in mice expressing EGFP in somatostatin-positive interneurons (Oliva et al., 2000) (Fig. 4A). Upon hyperpolarizing current steps these neurons showed a clear rebound depolarization that regularly induced a rebound AP (Fig. 4A), characteristic of LTS interneurons. After recording resting membrane potential for several minutes, LTS interneurons were stimulated with 10 trains of 60 APs at 50 Hz each (Marinelli et al., 2008). This stimulation induced a clear, long-lasting drop in membrane potential (SSI; ΔVm: −4.30 ± 0.96 mV; n = 12; p = 0.001) (Fig. 4B). SSI was absent in CB1R−/− mice, showing that it is mediated by CB1Rs (supplemental Fig. S3, available at www.jneurosci.org as supplemental material). All pharmacological interventions reducing DAGL-activity effectively blocked SSI (Fig. 4,C–F). When slices were incubated and recorded in OMDM-188 (2 μm), SSI was absent (ΔVm: −0.79 ± 0.76 mV; n = 8; p = 0.33) (Fig. 4C). Incubation with RHC-80267 (70 μm) also effectively blocked SSI (ΔVm: 0.63 ± 1.34 mV; n = 5; p = 0.66) (Fig. 4D). Finally, intracellular application of THL (5 μm) also effectively blocked SSI (ΔVm: −0.35 ± 0.98 mV; n = 6; p = 0.74) (Fig. 4E).
Therefore, we conclude that the lack of effect of DAGL inhibitors on DSI at the interneuron to CA1 pyramidal neuron unitary connections that we studied was not due to suboptimal experimental conditions. Taken together, our data suggest that DSI at these unitary connections in hippocampus occurs without the involvement of DAGL activity.
Biosynthetic pathways controlling endocannabinoid formation in hippocampal DSI have remained unclear. We systematically addressed involvement of DAGLs in DSI at identified unitary synaptic connections. We show that several known DAGL inhibitors do not affect DSI at these hippocampal synapses, while they abolish SSI in neocortical interneurons (Fig. 4), a process known to require DAGL activity (Marinelli et al., 2008). Our findings strongly suggest that DAGL is not involved in hippocampal DSI.
Recently, studies have reported opposing effects of the DAGL inhibitor THL on DSI (Szabo et al., 2006; Edwards et al., 2008; Hashimotodani et al., 2008). Intracellular application of THL effectively inhibits DAGL activity in VTA, cerebellum, and neocortex (Melis et al., 2004; Safo and Regehr, 2005; Marinelli et al., 2008). We found that intracellular postsynaptic application of THL did not affect DSI. This lack of effect is unlikely to be due to experimental conditions, because intracellular THL application successfully blocked SSI, confirming a previous report by Marinelli et al. (2008). Therefore, these results confirm the findings with other DAGL inhibitors, and rule out involvement of DAGL in hippocampal DSI. This raises the question by what mechanism extracellular THL application blocks hippocampal DSI (Edwards et al., 2008; Hashimotodani et al., 2008). Hashimotodani et al. (2008) and Szabo et al. (2006) ruled out indirect effects of THL on either CB1R signaling or calcium influx necessary for DSI induction. Likewise, we show that slice incubation with THL does not affect CB1R signaling (supplemental Fig. S2, available at www.jneurosci.org as supplemental material). At present, the mechanism by which extracellular THL blocks hippocampal DSI remains elusive. Given the absence of effects of OMDM-188, RHC-80267, and intracellular THL on DSI, we argue that extracellular THL does not block DSI through inhibition of DAGL.
A recent study showed that hippocampal DSI during cholinergic receptor activation requires signaling by nitric oxide (NO) (Makara et al., 2007). This opens the possibility that under these conditions endocannabinoid synthesis takes place outside of the postsynaptic neuron, perhaps in the presynaptic terminal, and that NO is the messenger emanating from the postsynaptic neuron. If the downstream target of NO would be sensitive to THL, this might explain why extracellular application of THL blocks DSI, while postsynaptic application is ineffective. However, in the same study from Makara et al. (2007) it was shown that NO is not involved in DSI in the absence cholinergic receptor activation. In agreement with this, we find that DSI at unitary connections, in the absence of cholinergic receptor activation, is not affected by incubation with the NO-synthase inhibitor Nω-Nitro-l-arginine methyl ester (supplemental Fig. S4, available at www.jneurosci.org as supplemental material). Therefore, NO involvement cannot explain the discrepancy between the effects of extracellular and postsynaptic THL application in our experiments.
It has been suggested that DSI might involve a pre-formed 2-AG pool (Edwards et al., 2006), and that DAGLs are involved in filling this pool, not in the actual DSI process. However, when blocking DAGLs, we did not observe a reduction in the amount of DSI during 50 subsequent depolarizations, although this would be expected when depleting a pre-formed pool of 2-AG. Therefore, if DSI involves a pre-formed pool of 2-AG, this pool must be quite substantial and stable.
Recent studies suggested that 2-AG, and not anandamide, is the endocannabinoid mediating hippocampal DSI. This conclusion was based on the ability of inhibitors of monoacylglycerol lipase, the enzyme that breaks down 2-AG, to prolong hippocampal DSI (Makara et al., 2005; Hashimotodani et al., 2007). Furthermore, inhibition of fatty-acid amide hydrolase, the enzyme responsible for anandamide breakdown, does not affect the time course of hippocampal DSI (Kim and Alger, 2004; Makara et al., 2005; Hashimotodani et al., 2007). However, while studies of DAGL distribution in hippocampus have shown that sn-1-DAGL-α is perfectly positioned for production of 2-AG as retrograde messenger at excitatory synapses (Katona et al., 2006; Yoshida et al., 2006), sn-1-DAGL-α was not found opposite symmetrical, GABAergic synapses (Katona et al., 2006). Together with our results, this suggests that sn-1-DAGL-α is not involved in DSI at inhibitory synapses in hippocampus. This may suggest that 2-AG is produced by different mechanisms at these synapses, involving another, hitherto unknown, synthesis pathway, such as the lysophosphatidic acid-dependent route (Nakane et al., 2002). Alternatively, it may suggest that 2-AG is not the endocannabinoid mediating DSI at unitary inhibitory synaptic connections in hippocampus.
The Department of Integrative Neurophysiology is financially supported by NeuroBsik (www.neurobsik.nl), NWO and the VU University board. We thank B. Lutz for providing us with CB1R−/− mice, K. Mackie for providing us with the CB1R antibody, N. Lozovaya and T. Nevian for helpful discussions, and J. Timmerman and T. Lodder for excellent technical support. R.M., A.B.B., N.B., and H.D.M. designed the experiments. R.M., G.T.-S., T.S.H., and J.C.L. performed experiments; C.B.C. performed immunohistochemistry and reconstructions; T.B. and V.D.M. synthesized OMDM-188; R.M. analyzed the data; R.M. and H.D.M. wrote the manuscript; all authors commented on the manuscript.
- Correspondence should be addressed to either of the following: Huibert D. Mansvelder, Department of Integrative Neurophysiology, Center for Neurogenomics and Cognitive Research, Neuroscience Campus Amsterdam, VU University, De Boelelaan 1085, Room C-440, 1081 HV Amsterdam, The Netherlands, ; or Nail Burnashev, Institut de Neurobiologie de la Méditerranée, Parc Scientifique de Luminy, 163 Roude de Luminy, BP13, 13273 Marseille, France,