Endocannabinoids Generated by Ca2+ or by Metabotropic Glutamate Receptors Appear to Arise from Different Pools of Diacylglycerol Lipase

The identity and subcellular sources of endocannabinoids (eCBs) will shape their ability to affect synaptic transmission and, ultimately, behavior. Recent discoveries support the conclusion that 2-arachidonoyl glycerol, 2-AG, is the major signaling eCB, however, some important issues remain open. 2-AG can be synthesized by a mechanism that is strictly Ca2+-dependent, and another that is initiated by G-protein coupled receptors (GPCRs) and facilitated by Ca2+. An important question is whether or not the 2-AG in these cases is synthesized by the same pool of diacylglycerol lipase alpha (DAGLα). Using whole-cell voltage-clamp techniques in CA1 pyramidal cells in acute in vitro rat hippocampal slices, we investigated two mechanistically distinct eCB-mediated responses to address this issue. We now report that pharmacological inhibitors of DGLα have quantitatively different effects on eCB-mediated responses triggered by different stimuli, suggesting that functional, and perhaps physical, distinctions among pools of DAGLα exist.


Introduction
The cannabinoid system affects behavior and regulates many synaptic functions. There are two major endogenous ligands for CB1R (the main cannabinoid receptor in the brain): the eCBs Narachidonoyl-ethanolamine (anandamide [1]) and 2-arachidonoylglycerol (2-AG) [2,3,4]. Anandamide and 2-AG have different synthetic and degradative pathways, and the eCB-dependent regulation of neuronal communication will be determined by the identity and subcellular sources of the eCB involved. Evidence is converging on the conclusion that 2-AG is the primary phasic signaling eCB at numerous synapses in the brain [5], whereas anandamide may regulate tonic eCB actions [6]. Strong support for the former inference comes from recent molecular genetic studies in which the primary synthetic enzyme for 2-AG, DAGLa, was knocked out in lines of mutant mice [7,8], causing a reduction of ,80% in basal 2-AG levels. Purely Ca 2+ -dependent eCB signaling -depolarization-induced suppression of inhibition, DSI [9,10], and excitation, DSE [11] -and eCB signaling mediated by GPCRs, including group I metabotropic glutamate receptors (mGluRs), i.e., (eCB mGluR ) [12,13] were essentially abolished by DAGLa deletion. Yet, additional issues remain unresolved. For example, it is not known if the same DAGLa source (pool) provides 2-AG for both DSI and eCB mGluR .
eCBs mediate different forms of synaptic plasticity [14], hence knowledge of the cellular source(s) of eCBs is an important issue, yet one that cannot be addressed with a global knock-out strategy. Accordingly, we have taken a pharmacological approach, using two DAGL inhibitors to determine whether the pools of Ca 2+and mGluR-dependent of 2-AG are distinguishable. If eCB responses to both stimuli were equally sensitive to the inhibitors, it would argue that the sources of 2-AG are the same, whereas marked differences in sensitivity would indicate that on a functional, and perhaps physical, level they differ. We report that the DAGL that mediates hippocampal DSI and eCB mGluR , can be functionally separated into two pools. Understanding the differences in subcellular regulation of 2-AG may lead to new modes for controlling eCB actions.

Results
While recent molecular biological evidence supports the conclusion that 2-AG is the signaling eCB, pharmacological tools can be useful in teasing apart subtle features of the DAGLa/2-AG system that are not revealed by constitutive knock-out strategies. To test the hypothesis that both DSI and eCB mGluR are mediated by the same source of 2-AG, we began by bath-applying DAGL inhibitors to voltage-clamped hippocampal CA1 cells in acute slices in which inhibitory post-synaptic currents (IPSCs) were pharmacologically isolated (see Methods). External application of the selective and potent inhibitor, OMDM-188 [15], 5 mM, or the less-selective inhibitor, tetrahydrolipstatin (THL), 10 mM, abolished DSI of evoked IPSCs (eIPSCs). As a percentage of baseline (100%) level, eIPSCs in the various conditions were: Vehicle: 60.264.0%, n = 20; OMDM-188: 95.761.5%, n = 34; THL: 92.861.4%, n = 35 (Fig. 1). We also tested two inhibitors of the 2-AG degradative enzyme, monoglyceride lipase, as these inhibitors do not affect anandamide. Both JZL 184 [16], 1 mM, and OMDM-169 [15], 2 mM, significantly prolonged t decay of DSI (cf [17]), thus providing an independent cross-check on the hypothesis that DSI is mediated by 2-AG (Fig. 2).
Unlike DSI, eCB mGluR -dependent eIPSC suppression (e.g., Fig. 3a) was highly resistant to external application of DAGL inhibitors. Responses to initial applications of DHPG were not different (p.0.05) whether the slices were treated with vehicle, or DAGL inhibitors. Even when evoked by repeated 4-min bathapplications of a high concentration of the group I mGluR agonist, DHPG., 50 mM, eCB mGluR was only slightly, though statistically significantly, reduced by OMDM-188 ( [12,13], that eCB mGluR is CB1Rdependent. Because intracellular application could conceivably be more effective on eCB mGluR [18], we tested the DAGL inhibitors by applying them intracellularly via infusion through the whole-cell pipette. We observed dose-dependent reduction in DSI with internal OMDM-188 (2-20 mM) (Figs. 4a, 4b; n = 94), and similar significant reductions caused by internal application of THL (10 mM, n = 19) (Fig. 4b). Internal DAGL inhibition did not alter eCB mGluR in the same way (Figs. 4c, 4d). We examined the effects of OMDM-188 in detail and found that, even in the same cells in which DSI was reduced to negligible levels (#5% eIPSC reduction, n = 28/35 cells; see Fig. 4d, dotted oval), eCB mGluR still suppressed eIPCSs by ,50%, i.e., OMDM-188 had almost no effect on eCB mGluR . Internal infusion of 5 mM OMDM-188 (filled triangles in Fig. 4d), reduced either eCB mGluR or DSI only slightly. Interestingly, data from the cells in which 10 or 20 mM OMDM-188 was least effective fell along a regression line around which the 5 mM data also scattered. This could mean that in these cases diffusion of 10 or 20 mM OMDM-188 out of the pipettes was incomplete, resulting in a lower-than-expected internal concentration of the drug.
It was not clear whether the passage of time alone accounted for the increased inhibitor efficacy, or whether suppression of eCB mGluR by DAGL inhibition was use-dependent, i.e., whether it was enhanced by repetitive stimulation. To distinguish the effects of longer OMDM-188 infusions from those of repeated DHPG application, in another group of cells we delayed the 1 st DHPG application until 30-40 min after break-in, i.e., it was given at the same relative time after break-in as the 2 nd DHPG application in the original group (cf, Figs. 5b, 5d). The eCB mGluR eIPSC suppression was the same in the two 1 st DHPG groups: (30-40 min post-break-in: to 50.564.3% of baseline, n = 11; 10-20 min post-break-in: to 49.562.5% of baseline, n = 16; n.s. p.0.1). A 2 nd DHPG application (i.e., 50-60 min post-break-in) given to cells receiving a delayed 1 st DHPG application, induced less eIPSC depression (2 nd DHPG: to 73.464.3% of baseline; 1 st DHPG: to 46.861.7% of baseline; n = 5, p,0.01; Figs. 5c, 5d). As a final check, we compared the magnitudes of the 2 nd DHPG responses, obtained either 30-40 or 50-60 min post-break-in, and found that they were indistinguishable (p.0.1, Figs. 5b, 5d). Hence, gradual diminution in the eCB mGluR response depended on the presence of both the DAGL inhibitor and repeated DHPG stimulation, and could not be explained simply by the duration of the inhibitor application.
The decline in eCB mGluR just described might reflect usedependent depletion of a pool of 2-AG. Two predictions would  follow from this hypothesis: 1) evoking eCB mGluR with a low concentration of DHPG should cause less of, or a slower onset of, a reduction in the DHPG effect in the presence of a DAGL inhibitor, and 2) inhibiting 2-AG synthesis by blocking another major synthetic enzyme in this pathway, phospholipase C b [19], should also give rise to a use-dependent decline in eCB mGluR .
In experiments thus far, we used 50 mM DHPG, which is a high concentration. To test the prediction that, in the presence of a DAGL inhibitor, weaker stimulation of mGluRs would induce less decline in eCB mGluR , we used 10 mM DHPG, with 20 mM OMDM-188 in the internal solution. In this case, we did not observe significant reduction in eCB mGluR , even with four applications of DHPG given to the same cell (1 st DHPG: to 69.968.3% of baseline; 2 nd DHPG: to 72.967.5% of baseline; 3 rd DHPG: to 70.169.9% of baseline; 4 th DHPG: to 74.669.2% of baseline; n = 5, p.0.1, data not shown). Hence, the declines in eCB mGluR seen with the higher DHPG concentration were not caused simply by repetitive activation of mGluRs.
If DSI and eCB mGluR arise from the same source of 2-AG, then strong activation of one mechanism could alter the response of the other. Because strong stimulation of mGluRs with bath-applied DHPG has persistent effects on eIPSCs and DSI [14,20] that could confound interpretation, we tested this prediction by determining if repetitive elicitation of DSI would affect eCB mGluR . Pipettes contained normal intracellular solution. We used two 4-min applications of 50 mM DHPG separated by a 4-min period during which DSI was elicited with 1-s depolarizing steps given at 12-s intervals. The 12-s interval is too short to permit full recovery from each DSI episode, with the result that eIPSCs are continually suppressed by the DSI mechanism for the period of stimulation (cf, [20]). As we have reported, an even longer period of repetitive elicitation does not persistently diminish DSI [20], nevertheless, it was possible that repetitive DSI stimulation could reduce the magnitude of the subsequent eCB mGluR interval if a common 2-AG pool was being tapped. Nevertheless, we found no evidence that this occurred. The eIPSCs were suppressed to 49.865.8% of baseline by 50 mM DHPG before the repetitive DSI stimulation and to 48.464.8% of baseline afterwards; n = 6, p.0.1; Figs. 6c, 6d). We also tested the possibility that repetitive DSI stimulation might somehow alter eCB mGluR if DSI expression was first blocked by OMDM-188. Again, after a 4-min period of repetitive DSI, 50 mM DHPG induced an eIPSC suppression to 48.665.2% of baseline, not significantly different from the 1 st DHPG responses obtained either 10-20 or 30-40 min post break-in, p.0.5 (data not shown).

Discussion
During a previous investigation [20], we noticed differences in the efficacy of THL on DSI or eCB mGluR , however, in view of the non-specific effects of THL, no definite conclusions could be drawn. Moreover, it was unclear if both DSI and eCB mGluR were mediated by the same eCB. In showing that DAGLa, and by implication 2-AG, are involved in both processes, the recent studies on DAGLa 2/2 mice [7,8] prompted an examination of whether or not the same sources of 2-AG mediate eCB responses evoked by different stimuli. The hypothesis that a unitary pool of DAGLa supplies 2-AG for DSI and eCB mGluR predicts they should be similarly affected by pharmacological inhibitors of DAGLa. Using different inhibitors and modes of drug application, we observed marked quantitative distinctions between the responses produced by DSI and eCB mGluR . In particular: 1) eCB mGluR is much less sensitive to block by DAGL-inhibitors than is DSI, and 2) repetitive activation of the eCB mGluR system enhanced the effect of DAGL inhibition, whereas such usedependence was not a feature of the block of DSI.
The differences in sensitivity to the DAGL inhibitors were obvious even when both responses were recorded in the same cell, ruling out systematic differences between experiments. Differences in inhibitor-enzyme interactions are also ruled out, as DAGLa mediates both responses. A reasonable interpretation is that different pools of DAGLa provide 2-AG in the two cases. The hypothetical pools would not simply represent differences in spatial localization along the pyramidal cell: with external application the inhibitors have equal access to the surface of the cells, but had only slight effects on eCB mGluR , despite abolishing DSI. A plausible explanation for the differing efficacy of internal and external application on eCB mGluR is that the DAGLa responsible for eCB mGluR is much less accessible to externally applied inhibitor.
The suggestion of different pools of DAGLa is in good agreement with previous observations. For example, the DAGLa involved in eCB mGluR is found in dendrites apposed to glutamate releasing nerve terminals [21,22,23]. In contrast, DAGLa has not been reported near perisomatic GABAergic synapses like those we have studied [21,22]. Since eCBs can spread longitudinally along cell structures for only ,#10 mm [24], 2-AG produced by DAGLa near excitatory synapses, which are located on CA1 pyramidal cell dendrites .50 mm from the somata, is most unlikely to account for DSI. While the failure to have detected DAGLa in pyramidal cell somata may reflect technical limitations in available morphological tools, it does highlight the possibility that different parts of a cell employ different pools of DAGLa for generating eCBs. We also note, however, that while this seems to be a parsimonious proposal, the lack of identification of the DAGLa responsible for DSI means that other possibilities are not ruled out. For example, our data would be compatible with differences in DAG (rather than DAGLa) pools. DAG can be produced by several mechanisms besides PLC [5]. We confirm the 2-AG produced by mGluRs is dependent on PLC b , but knocking out [19], or inhibiting [20] PLC b does not affect hippocampal DSI. Since DSI is dependent on DAGLa, and therefore probably on 2-AG, it could be mediated by a source of DAG that is distinct from that underlying eCB mGluR . Interestingly, use-dependence of eCB mGluR reduction was seen when PLC, rather than DAGL, was inhibited, supporting the concept that the DAGLa-PLC pathway is upstream of the depletable source of 2-AG for eCB mGluR .
Attempts to use DAGLa inhibitors to probe the eCB system have produced controversial results [18,20,25,26]. We had observed that external, though not internal, THL application affected DSI [26], and Min et al. [25] arrived at conclusions that are diametrically opposed to our present observations. Probably the use of higher concentrations of the inhibitors, longer application times, and repetitive activation of the 2-AG-dependent responses are the primary explanations for the reported variability. In particular, difficulties in bath-or internally-applying these lipophilic agents to cells in brain slices could account for the requirement for higher concentrations, especially with intracellular techniques, because restricted efflux from whole-pipettes can result from partial electrode occlusion or adherence of the drugs to pipette glass. Nevertheless, the possibility of non-specific effects must also be kept in mind.
The use-dependence of the reduction in eCB mGluR is puzzling, and while direct evidence is not available, one speculative scenario is intriguing: decoupling of 2-AG synthesis and release could partly account for the data. For example, if the pool of DAGLadependent 2-AG that is present in unstimulated cells [4,7,8] could be mobilized by mGluR activation, its release would not be closely tied to DAGLa stimulation. Release from such a pool could persist after DAGLa was inhibited, decreasing in a use-dependent way as the preformed 2-AG pool diminished. If such decoupling were greater for eCB mGluR than DSI, it could also account for their different sensitivities to DAGL inhibitors. Finally, 2-AG release from an existing pool could explain why knocking out PLC b abolishes eCB mGluR [19], while a potent PLC inhibitor was ineffective [20,26] (unless eCB mGluR is repetitively elicited, Fig. 6). Thus, our present data is consistent with, and extends, previous proposals [26,27,28].
Factors that could be responsible for maintaining the hypothetically distinct DAGLa (or DAG) pools are unknown. Physical separation of protein components of the biosynthetic cascades, perhaps sequestration on lipid rafts [29] or other factors related to the heterogeneity of the lipid bilayer could be involved [28]. In this context, it may also be worth noting that DAGLa 2/2 mice also showed a reduction of ,50% in the levels of anandamide [7,8]. Therefore, although the greater reduction in 2-AG supports the concept that 2-AG is the eCB that mediates many phasic CB1Rdependent responses, the possibility cannot be eliminated that anandamide might play a contributory role in some cases.
The model of 2-AG release from a pre-existing pool differs from the conventional ''on-demand'' model of eCB production in which synthesis and release are necessarily tightly coupled to each other. The hypothesis may account in part for some of the multiple mechanisms of eCB responses [20,26], and may help reconcile the negative pharmacological reports with DAGL inhibitors, e.g. [25], with data from the DAGL 2/2 studies [7,8]. The hypothesis would also be consistent with evidence that 2-AG release is a regulated step, perhaps involving the eCB transporter [30,31]. Testing this idea and investigating the subcellular distribution and regulation of DAGL in general will be important future tasks.

Preparation of slices
All experimental protocols were reviewed and approved by the University of Maryland School of Medicine IACUC (IACUC approval #0609001), and all animal handling was conducted in accordance with national and international guidelines. The number of animals used was minimized, and all necessary precautions were taken to mitigate pain or suffering.
Hippocampal slices were obtained from 4-to 6-week-old male Sprague-Dawley rats. After rats were sedated with isoflurane and decapitated, the hippocampi were removed and 400-mm-thick slices were cut on a Vibratome (model VT1200s, Leica Microsystems, Inc., Bannockburn, IL) in an ice-cold extracellular recording solution. Slices were stored in a holding chamber on filter paper at the interface of this solution and a moist, oxygenated atmosphere at room temperature for $1h before transfer to the recording chamber (RC-27L, Warner Instruments, CT) and warmed to 30-31uC. The extracellular solution contained (mM): 120 NaCl, 3 KCl, 2.5 CaCl 2 , 2 MgSO 4 , 1 NaH 2 PO 4 , 25 NaHCO 3 , and 20 glucose, and was bubbled with 95%O 2 , 5%CO 2 (pH 7.4).

Electrophysiology
Whole-cell voltage-clamp recordings of CA1 pyramidal cells were made using the blind patch method. Pipettes were pulled from thin walled glass capillaries (1.5 O.D., World Precision Instruments, Sarasota, FL). Electrode resistances in the bath were 3-6 MV with internal solution containing (mM): 90 CsCH 3 SO 4 , 1 MgCl 2 , 50 CsCl, 2 MgATP, 0.2 Cs 4 -BAPTA, 10 HEPES, 0.3 Tris GTP and 5 QX314. If the series resistances, which was checked by -2 mV voltage steps throughout experiments, changed .20%, the data were discarded. The holding potential was -70 mV in all experiments. Monosynaptic eIPSCs were elicited by 100-ms-long extracellular stimuli delivered at 0.25 Hz with concentric bipolar stimulating electrodes placed in s. radiatum. NBQX (10 mM) and D-AP5 (20 mM) were present in all experiments to block glutamatergic EPSCs. Slices were pretreated in the holding chamber with the irreversible P/Q-type voltage-gated Ca 2+channel toxin, v-agatoxin GVIA (agatoxin, 300 nM) for $1 h to reduce the contribution of eCB-insensitive eIPSCs in all experiments [32]. Data were collected with an Axopatch 1C amplifier (Molecular Devices, Sunnyvale, CA), filtered at 1 kHz and digitized at 5 kHz using a Digidata 1200 (Molecular Devices) and Clampex 8 software (Molecular Devices).
The bath solution was oxygenated with 95% O 2 /5% CO 2 gas, and perfused continuously through the recording chamber at ,1 ml/min. For external applications, slices were preincubated for .40 min with OMDM-188, THL, OMDM-169, DMSO, or ethanol, and the drug was also present in the bath solution throughout the recording. The final concentration of the solvent, DMSO or ethanol, was 0.05% (v/v) or less for both OMDM-188 and THL. JZL184 was obtained from Cayman Chemical, OMDM-188 and OMDM-169 was synthesized by Giorgio Ortar and Enrico Morera, and all other chemicals were purchased from Sigma (St. Louis, MO).

Data analysis
To measure DSI, we evoked IPSCs at 4-s intervals and depolarized the postsynaptic cell to 0 mV for 1-5 s at 90-s intervals. The magnitude of DSI was calculated as follows: DSI (%) = 1006 [1 -(mean of 4 IPSCs after depolarization/mean of 5 IPSCs before depolarization)]. Values of 2 -3 DSI trials were averaged for a given condition. The decay time constant of DSI (t decay ) was determined by fitting the data with a single-exponential decay function in SigmaPlot 10.0. Two-tailed paired t-tests were used whenever appropriate; otherwise unpaired t-tests were used for single comparisons. Statistical tests among groups were done with oneway ANOVA. For comparison of results from repeated DHPG applications, we used one-way repeated ANOVA. The significance level for all tests was p,0.05 (*). Group means 6 SEMs are shown for display purposes. For comparison of cumulative distributions, we used the Kolmogorov-Smirnov (K-S) test, available at http://www. physics.csbsju.edu/stats/KS-test.n.plot_form.html.