Multiple Forms of Endocannabinoid and Endovanilloid Signaling Regulate the Tonic Control of GABA Release

Persistent CB1 cannabinoid receptor activity limits neurotransmitter release at various synapses throughout the brain. However, it is not fully understood how constitutively active CB1 receptors, tonic endocannabinoid signaling, and its regulation by multiple serine hydrolases contribute to the synapse-specific calibration of neurotransmitter release probability. To address this question at perisomatic and dendritic GABAergic synapses in the mouse hippocampus, we used a combination of paired whole-cell patch-clamp recording, liquid chromatography/tandem mass spectrometry, stochastic optical reconstruction microscopy super-resolution imaging, and immunogold electron microscopy. Unexpectedly, application of the CB1 antagonist and inverse agonist AM251 [N-1-(2,4-dichlorophenyl)-5-(4-iodophenyl)-4-methyl-N-1-piperidinyl-1H-pyrazole-3-carboxamide], but not the neutral antagonist NESS0327 [8-chloro-1-(2,4-dichlorophenyl)-N-piperidin-1-yl-5,6-dihydro-4H-benzo[2,3]cyclohepta[2,4-b]pyrazole-3-carboxamine], significantly increased synaptic transmission between CB1-positive perisomatic interneurons and CA1 pyramidal neurons. JZL184 (4-nitrophenyl 4-[bis(1,3-benzodioxol-5-yl)(hydroxy)methyl]piperidine-1-carboxylate), a selective inhibitor of monoacylglycerol lipase (MGL), the presynaptic degrading enzyme of the endocannabinoid 2-arachidonoylglycerol (2-AG), elicited a robust increase in 2-AG levels and concomitantly decreased GABAergic transmission. In contrast, inhibition of fatty acid amide hydrolase (FAAH) by PF3845 (N-pyridin-3-yl-4-[[3-[5-(trifluoromethyl)pyridin-2-yl]oxyphenyl]methyl]piperidine-1-carboxamide) elevated endocannabinoid/endovanilloid anandamide levels but did not change GABAergic synaptic activity. However, FAAH inhibitors attenuated tonic 2-AG increase and also decreased its synaptic effects. This antagonistic interaction required the activation of the transient receptor potential vanilloid receptor TRPV1, which was concentrated on postsynaptic intracellular membrane cisternae at perisomatic GABAergic symmetrical synapses. Interestingly, neither AM251, JZL184, nor PF3845 affected CB1-positive dendritic interneuron synapses. Together, these findings are consistent with the possibility that constitutively active CB1 receptors substantially influence perisomatic GABA release probability and indicate that the synaptic effects of tonic 2-AG release are tightly controlled by presynaptic MGL activity and also by postsynaptic endovanilloid signaling and FAAH activity. SIGNIFICANCE STATEMENT Tonic cannabinoid signaling plays a critical role in the regulation of synaptic transmission. However, the mechanistic details of how persistent CB1 cannabinoid receptor activity inhibits neurotransmitter release have remained elusive. Therefore, electrophysiological recordings, lipid measurements, and super-resolution imaging were combined to elucidate those signaling molecules and mechanisms that underlie tonic cannabinoid signaling. The findings indicate that constitutive CB1 activity has pivotal function in the tonic control of hippocampal GABA release. Moreover, the endocannabinoid 2-arachidonoylglycerol (2-AG) is continuously generated postsynaptically, but its synaptic effect is regulated strictly by presynaptic monoacylglycerol lipase activity. Finally, anandamide signaling antagonizes tonic 2-AG signaling via activation of postsynaptic transient receptor potential vanilloid TRPV1 receptors. This unexpected mechanistic diversity may be necessary to fine-tune GABA release probability under various physiological and pathophysiological conditions.


Introduction
Neurons set their synaptic strength by using several signaling mechanisms that dynamically adjust parameters of synaptic transmission, including neurotransmitter release probability (Branco and Staras, 2009). Endocannabinoid signaling is one of the most widespread and efficient mechanisms to control release probability via presynaptic CB 1 cannabinoid receptor activation (Lovinger, 2008). Accordingly, CB 1 receptors are indispensable for various phasic forms of synaptic plasticity (Castillo et al., 2012) and have an essential function in the tonic control of neurotransmitter release (Losonczy et al., 2004;Hentges et al., 2005;Zhu and Lovinger, 2005;Oliet et al., 2007). However, although the molecular mechanisms accounting for phasic endocannabinoid signaling have been outlined in considerable detail previously (Ohno-Shosaku and Kano, 2014), our understanding of how tonic regulation of neurotransmitter release is organized and controlled by specific regulatory mechanisms is rather limited.
The lipid messengers anandamide and 2-arachidonoylglycerol (2-AG) are both released from postsynaptic neurons and act retrogradely on presynaptic CB 1 receptors (Ohno-Shosaku and Kano, 2014). Their synaptic levels are determined by complex enzymatic interactions with numerous synthesis and degradation routes identified in the brain . Because anandamide and 2-AG are known to participate differentially in specific behavioral processes (Long et al., 2009a), it is generally assumed that this molecular and functional complexity reflects division of labor in setting synaptic strength in the underlying brain circuits. However, it is not clear whether it is 2-AG or anandamide, which primarily contributes to the tonic control of neurotransmitter release at given circuit locations. For example, previous studies using culture preparations and pharmacological tools with limited target selectivity arrived at different conclusions concerning the identity of the endocannabinoid responsible for tonic action at hippocampal GABAergic synapses (Hashimotodani et al., 2007;Kim and Alger, 2010). Moreover, emerging in vitro and in vivo observations raised the possibility that constitutive receptor activity may also influence axonal CB 1 distribution and tonically attenuate GABA release in the forebrain (Leterrier et al., 2006;Meye et al., 2013;Thibault et al., 2013).
Because the cannabinoid tone plays a major role in interneuron-type-specific and activity-dependent adjustment of GABAergic synaptic strength, a better understanding of the underlying signaling processes is important (Losonczy et al., 2004;Ali and Todorova, 2010;. In addition, recent studies have reported selective, long-term pathological alterations in the tonic endocannabinoid control of GABA release in rodent models of major brain disorders, including autism, Huntington's disease, and seizures (Chen et al., , 2007Dvorzhak et al., 2013;. Therefore, we used paired patchclamp recordings in acute brain slices, in combination with mass spectrometry, stochastic optical reconstruction microscopy (STORM) super-resolution imaging, and immunogold electron microscopy, to identify the specific mechanisms that mediate the tonic control of GABA release in the mouse hippocampus. The findings indicate that constitutively active CB 1 receptors play a considerable role in the regulation of GABA release probability and elucidate a functional interaction between intercellular 2-AG endocannabinoid and intracellular anandamide endovanilloid signaling in setting the strength of synaptic inhibition.

Materials and Methods
Animals. All protocols related were approved by the Institutional Animal Care and Use Committee of the University of California, Irvine, and the Hungarian Committee of the Scientific Ethics of Animal Research (license number XIV-1-001/2332-4/2012), and all animal experiments were performed according to the Hungarian Act of Animal Care andExperimentation (1998, XXVIII, Section 243/1998), which are in accordance with the European Communities Council Directive of November 24, 1986 (86/609/EEC; Section 243/1998). All efforts were made to minimize pain and suffering and to reduce the number of animals used. Adult male C57BL/6 mice aged 4 -7 weeks were used in the electrophysiological, anatomical, and analytical chemistry experiments. In the electrophysiological experiments testing the role of TRPV1, adult C57BL/6 wild-type and TRPV1 Ϫ/Ϫ mice of either sex (The Jackson Laboratory) aged 4 -7 weeks were used.
Electrophysiological recordings. The procedures of electrophysiology experiments are similar to those described previously by . Briefly, coronal dorsal hippocampal slices (300 m; anteroposterior, Ϫ1.9 to Ϫ2.8 mm from bregma) were prepared from deeply anesthetized mice. After cutting, slices were incubated in sucrose-containing artificial CSF (ACSF) for 1 h (in mM: 85 NaCl, 75 sucrose, 2.5 KCl, 25 glucose, 1.25 NaH 2 PO 4 , 4 MgCl 2 , 0.5 CaCl 2 , and 24 NaHCO 3 ). All electrophysiological recordings were made at 33°C in ACSF solution (in mM: 126 NaCl, 2.5 KCl, 26 NaHCO 3 , 2 CaCl 2 , 2 MgCl 2 , 1.25 NaH 2 PO 4 , and 10 glucose). Slices were visualized with an upright microscope (Olympus BX61WI or Nikon Eclipse FN1) with infrared differential interference contrast optics. Whole-cell recordings were obtained from the interneurons with patch pipettes (3-5 M⍀) filled with internal solution containing 126 mM K-gluconate, 4 mM KCl, 10 mM HEPES, 4 mM ATP-Mg, 0.3 mM GTP-Na, 10 mM phosphocreatine, and 0.2% biocytin, pH 7.2, 290 mOsm. The interneurons were located in the stratum radiatum of the CA1 subfield. Pyramidal cells (whole-cell voltage-clamp configuration, holding potential set to Ϫ70 mV) were recorded with internal solution containing the following (in mM): 40 CsCl, 90 K-gluconate, 1.8 NaCl, 1.7 MgCl 2 , 3.5 KCl, 0.05 EGTA, 10 HEPES, 2 MgATP, 0.4 Na 2 GTP, and 10 mM phosphocreatine, pH 7.2 (290 mOsm). In specific experiments, the calcium chelator BAPTA (10 mM) was also added to the internal solution. All drugs were obtained from Cayman Chemical, Tocris Bioscience, or Sigma-Aldrich. Recordings were made using MultiClamp700B amplifiers (Molecular Devices). Signals were filtered at 3 kHz using a Bessel filter and digitized at 10 or 20 kHz with Digidata 1440A or 1550 analog-todigital interfaces (Molecular Devices). Series resistances were carefully monitored, and recordings were discarded if the series resistance changed Ͼ20% or reached 20 M⍀. The recorded traces were analyzed using the Clampfit 10 software (Molecular Devices). Action potentials in presynaptic interneurons were induced in current clamp by injecting 2 ms square pulses of 1.5-2 nA at 10 Hz frequency. Trains of 50 action potentials were performed once every minute. Values for the so-called "effective" unitary IPSC (euIPSC; amplitudes included both successful events and failures. To examine drug effects, unless stated otherwise, brain slices were treated with a drug or a mixture (JZL184, JZL184 ϩ AM251, PF3845, PF3845 ϩ JZL184, PF3845 ϩ JZL184 ϩ AM9810, or AM251 ϩ AMG9810, dissolved in the recording ACSF solution) for 40 min before paired recordings were performed (note that the lipid level measurements were also performed after 40 min treatment of slices with one of the drug conditions or 0.1% DMSO control solution; for detailed information, see below, Endocannabinoid level measurements by liquid chromatography/tandem mass spectrometry). In specific experiments, acute drug effects (AM251, NESS0327, JZL184, JZL184 ϩ NESS0327, WIN55,212-2, dissolved in the recording ACSF solution) were quantified by averaging three series of 50 trials each in the respective drug conditions. Depolarization-induced suppression of inhibition (DSI) of uIPSCs was produced by using a 0.5-s-long depolarizing pulse from Ϫ70 to 0 mV, and the IPSCs were compared between the pre-DSI period (Ϫ2 to 0 s; the latter being the start of the depolarizing pulse) and the DSI period (0 -2 s after the end of the pulse).
Endocannabinoid level measurements by liquid chromatography/tandem mass spectrometry. Acute hippocampal slices were prepared and treated exactly as hippocampal slices used for electrophysiological experiments. To reduce potential variability in endocannabinoid levels attributable to the different circuit architecture of dorsal and ventral hippocampal sections, five 300-m-thick acute hippocampal slices were pooled in a single measured sample. To exclude potential hemispheric differences, these slices were collected in an alternate manner from the left and right hemisphere. Thus, endocannabinoid levels were always compared between two hippocampal tissue samples derived from an individual animal, with the tissue being treated by either an enzyme inhibitor or its corresponding vehicle control. Animals were housed, and brain slices were prepared in exactly the same manner and at the same time of the day from one liquid chromatography/tandem mass spectrometry (LC/MS/MS) experiment to the next, to further reduce biological variability in endocannabinoid levels. After pharmacological treatment, the hippocampal samples were frozen on dry ice and stored at Ϫ70°C until the LC/MS/MS measurement. Sample preparation for chromatography started with tissue incubation on ice for 30 min in 1 ml of methanol containing the deuterated internal standards 2-arachidonoylglycerol-d 5 (100 ng/ml) and arachidonoyl-ethanolamide-d (1 ng/ml). After ultrasonic homogenization, samples were centrifuged on an Eppendorf miniSpin microtube centrifuge at 13,400 rpm for 15 min. The supernatant was transferred into a plastic tube and was diluted with 3 ml of ultrapure water. The samples were subjected to solid-phase extraction (SPE) sample clean-up according to the following protocol. First, the SPE cartridges (SUPELCO Discovery DSC-18 1 ml tubes, 100 mg) were conditioned with 2 ml of methanol and 2 ml of ultrapure water. Diluted samples were loaded at ϳ0.5 ml/min flow rate to apply a gentle vacuum. In the consecutive washing step, 2 ml of ultrapure water and 2 ml of methanol/water (50:50 v/v%) were percolated through the sorbent material. Elution was performed with 0.7 ml of methanol. Eluates were diluted to initial HPLC eluent composition with 10 mM ammonium formate solution.
To measure endocannabinoid levels, we used a PerkinElmer Life and Analytical Sciences HPLC Series 200 system, which was coupled to an Applied Biosystems/Sciex 4000 QTRAP triple quadrupole/linear ion trap tandem mass spectrometer operated in positive electrospray ionization mode. The electronspray ionization ion source parameters were set as follows: curtain gas, 10; ion spray voltage, 5000 V; temperature, 500°C; collisionally activated dissociation gas, medium; gas 1, 50; gas 2, 40. Chromatographic separation was achieved with a Phenomenex Kinetex C18 column (50 mm ϫ 3.00 mm, 2.6 m, 100 Å) using methanol and 10 mM ammonium formate as elution solvents at a flow rate of 500 l/min. The injection volume was 50 l. The initial eluent condition was 80% methanol/20% buffer that was increased to 85% organic phase during 3 min and then further elevated to 95% during 2 min and was kept at this condition for 2 min. Afterward, the column was equilibrated to the initial condition. Analytes were detected in multiple reaction monitoring (MRM) mode at the following ion transitions and parameter settings: ( 2), declustering potential (81 V), collision energy (41 V, 77 V), cell exit potential (6 V). The peak areas were determined with Analyst 1.4.2. software. The quantity of the analytes was calculated by comparing their peak areas with those of the deuterated internal standards, and it was normalized to the sample weight.
All endocannabinoid standards were purchased from Cayman Chemical. HPLC gradient-grade methanol was supplied by Merck. Ammonium formate was from Riedel-de Haën. Water was purified with a MilliQ Direct 8 system (Millipore).
Immunostaining for cell-type identification. After in vitro recording with pipettes containing biocytin, slices were transferred into a fixative solution containing 4% paraformaldehyde and 0.2% picric acid in 0.1 mM phosphate buffer (PB). Slices were resectioned into 40 -70 m thin sections and tested for CB 1 immunopositivity (Fukudome et al., 2004; CB 1 -GP-Af530-1; 1:5000, guinea pig; Frontiers Science). A secondary antibody conjugated to Alexa Fluor 488, raised in donkey against guinea pig (Invitrogen), were used to detect the location of the primary antibody, and streptavidin conjugated to DyLight 594 (1:500; Jackson Im-munoResearch) was used to visualize biocytin. All primary and secondary antibodies were diluted in Tris-buffered saline (TBS) containing 2% normal goat serum. The sections were then mounted in Vectashield (Vector Laboratories) and analyzed with a fluorescent microscope (Zeiss Axioscope 2 plus or Nikon Eclipse 80i) at low magnification and then with a confocal microscope (Zeiss LSM700 microscope or Nikon A1R confocal laser-scanning system built on a Ti-E inverted microscope and operated by NIS-Elements AR software) at high magnification. To evaluate interneuron morphology, Z-stacks were acquired throughout the whole axonal arborization of the recorded cells. In some cases, the biocytin-filled cells were visualized with 3,3Јdiaminobenzidine tetrahydrochloride (0.015%) using a standard ABC kit (Vector Laboratories). Identification of CB 1 -expressing interneurons was done based on the distinct axonal morphology (perisomatic interneurons: axons branching almost exclusively the pyramidale layer; dendritic interneurons: axons ramifying predominantly in the stratum radiatum and to a lesser extent in the oriens) and the immunopositivity for CB 1 (Vida et al., 1998;Cope et al., 2002;Pawelzik et al., 2002;Lee et al., , 2014Dudok et al., 2015).
Immunostaining for STORM imaging. Immersion-fixed 300-m-thick acute brain slices containing biocytin-filled cells were developed with Alexa Fluor 488 -streptavidin, covered in Vectashield, and imaged on a Zeiss LSM780 confocal microscope using a Plan Apochromat 20ϫ objective (0.8 numerical aperture), and the cell types were evaluated based on axonal morphology. After imaging, the slices containing interneurons selected for STORM imaging were washed and embedded in 2% agarose in distilled water, and 20-m-thick coronal sections were cut with a Leica VT-1000S Vibratome in PB and transferred into 24-well tissue culture plates. Immunostaining was performed in a free-floating manner as described previously (Dudok et al., 2015). Briefly, sections were washed in TBS, blocked with 1% HSA (albumin from human serum; Sigma), and incubated overnight with an affinity-purified primary antibody (guinea pig anti-CB 1 , 1:1000; Fukudome et al., 2004). Sections were then washed in TBS and incubated with a secondary antibody (2 g/ml) for 4 h, which was prepared for STORM imaging by using an unlabeled affinitypurified anti-guinea pig IgG (Jackson ImmunoResearch), and was tagged with Alexa Fluor 647 as reporter and Cy3 as activator. After washing in TBS and PB, sections were mounted and dried on coverslips. Finally, the samples were covered with freshly prepared imaging medium containing 5% glucose, 0.1 M mercaptoethylamine, 1 mg/ml glucose oxidase, and catalase (2.5 l/ml aqueous solution from Sigma; ϳ1500 U/ml final concentration) in Dulbecco's PBS (Sigma). Coverslips were sealed with nail polish and transferred onto the microscope after 10 min incubation in imaging medium. STORM imaging was performed for up to 3 h after covering the specimens.
Combined confocal/STORM image acquisition and processing. Correlated STORM and confocal imaging, image processing, and analysis were performed as described previously (Dudok et al., 2015). Images were taken using a Nikon Ti-E inverted microscope equipped with CFI Apo TIRF 100ϫ objective (1.49 numerical aperture), Nikon N-STORM system, a Nikon C2 confocal scan head, and an Andor iXon Ultra 897 EMCCD. First, the field of view containing biocytin-labeled axon terminals was selected, and a confocal stack was captured. Next, the astigmatic 3D STORM image of the anti-CB 1 immunostaining was acquired, using 1000 cycles of one activator and three reporter frames, 30 ms each. The low-power activation and the similar imaging depth (ϳ5 m) in tissue for all boutons enabled quantitative evaluation of the number of detected localization points (LPs), as demonstrated previously (Dudok et al., 2015).
The corresponding deconvolved confocal images and STORM molecule lists were overlaid using custom-written macros in NIH ImageJ. The outline of the analyzed bouton was determined using the confocal image of biocytin, and LPs from the STORM data belonging to the same bouton were identified. To quantify the internalization of CB 1 receptors, the distance (d) of each CB 1 LP was first measured from the 3D center of mass of all LPs within the given axon terminal. The radius of the terminal was calculated as the radius of a circle with the same area as the 2D area measured on the corresponding confocal image of the biocytincontaining axon terminal. The ratio of internalized receptors was estimated as the ratio of predicted plasma membrane-bound receptors over total receptors in untreated animals based on previous immunogold electron microscopy data (88%; Dudok et al., 2015). Subsequently, we determined the 88% percentile of the d/r values of all individual LPs from boutons of untreated cells and used this as a threshold to identify internalized LPs in boutons from control and treated samples. Visualization of STORM data was achieved by using the N-STORM module in NIS-Elements AR 4.2 software for 2D images and by using the Visual Molecular Dynamics 1.9.2 software for 3D renderings. The axon terminal surface was approximated by fitting a 3D convex hull on the LPs. LPs were visualized as atoms, and convex hull edges were visualized as bonds.
Immunogold staining for electron microscopy. Two wild-type and two TRPV1 Ϫ/Ϫ mice (5 weeks old) were anesthetized deeply with ketaminexylazine (25 mg/ml ketamine, 5 mg/ml xylazine, and 0.1% w/w pipolphen in H 2 O; 1 ml/100 g, i.p.). Animals were perfused transcardially with 0.9% saline for 3 min, followed by 100 ml of 4% paraformaldehyde fixative in 0.1 M PB, pH 7.4, for 20 min. After perfusion, the brains were postfixed for 2 h and washed in PB. Hippocampal blocks were sliced into 50-m-thick coronal sections with a Leica VTS-1000 vibratome. After slicing and extensive washing in 0.1 M PB, the sections were incubated in 10% sucrose for 15 min and 30% sucrose overnight, followed by freezethawing over liquid nitrogen four times. Subsequently, all washing steps and dilutions of the antibodies were performed in 0.05 M TBS, pH 7.4. After extensive washing in TBS, the sections were blocked in 1% human serum albumin for 45 min and then incubated in goat anti-TRPV1 (1: 1000; sc-12498; Santa Cruz Biotechnology) for a minimum of 48 h at 4°C. After the primary antibody incubation, the sections were washed extensively in TBS before incubation in 0.8 nm gold-conjugated donkey antigoat secondary antibody (1:50; Aurion) overnight at 4°C. Then the sections were silver intensified using the R-GENT SE-LM kit according to the protocol (Aurion). After development, the sections were treated with 1% osmium tetroxide in PB for 10 min on ice in the dark and dehydrated in an ascending series of ethanol and acetonitrile, before being embedded in Durcupan (ACM; Fluka). During dehydration, sections were treated with 1% uranyl acetate in 70% ethanol for 15 min on ice in the dark. Areas of interest from the CA1 stratum pyramidale or stratum radiatum were reembedded and resectioned for additional electron microscopic analysis. Ultrathin (60 nm) sections were collected on Formvar-coated single-slot grids and stained with lead citrate. Electron micrographs were taken at 40,000ϫ magnifications with a Hitachi 7100 electron microscope. Finally, the images were collected and analyzed by an experimenter blind to the animal genotype.
To establish the precise subcellular distribution of TRPV1 at inhibitory symmetrical synapses in the hippocampal CA1 stratum pyramidale and stratum radiatum, high-resolution quantitative evaluation was performed in a population of Ͼ250 immunogold-labeled synapses in each animal. Superficial ultrathin sections were collected at a similar tissue depth (ϳ5 m) to ensure equal probability of antibody penetration. Images were taken randomly of clearly identifiable symmetrical synapses projecting onto the soma or onto the distal thin dendrites of CA1 pyramidal cells. If a gold particle was present in the image, its location was determined within one of six compartments: (1) presynaptic (within the synaptic bouton); (2) postsynaptic (directly adjacent to the bouton on the postsynaptic cell soma); (3) somatic cytoplasmic (within the postsynaptic soma but not adjacent to the synaptic bouton); (4) mitochondrial (within mitochondria); (5) nuclear (within the cell nucleus); or (6) unknown. The 2 tests for these subcellular compartments revealed similar numbers of gold particles in the two animals within each genotype in all six compartments. Therefore, data from the two animals belonging to the same genotype were pooled and compared with data obtained from the other genotype. The 2 tests showed significantly more gold particles in the postsynaptic compartment of wild-type mice compared with TRPV1 knock-out animals in the case of perisomatic synapses, whereas dendritic symmetrical synapses did not contain specific TRPV1 immunolabeling. Gold particle numbers also did not differ between genotypes in the categories of presynaptic, somatic cytoplasmic, mitochondrial, and unknown. More gold particles were found over the nuclei of TRPV1 knock-out mice than wild-type mice, which likely reflects the fact that the antibody can bind to off-targets in a nonspecific manner when its real target protein is absent in the tissue.
To assess the subsynaptic location of these gold particles, we measured their distance from the postsynaptic membrane and allocated them to corresponding 100 nm "bins" to assess their proximity to the synapse. Distance measurements were performed using NIH ImageJ Software.
Statistics. To analyze the data obtained from electrophysiological recordings, paired or unpaired (as appropriate) two-tailed Student's t tests were used. In cases in which the data did not show normal distribution based on the Shapiro-Wilk test or D'Agostino-Pearson test, then the Wilcoxon's signed-rank or Mann-Whitney U tests for paired and unpaired data, respectively, were used. Pearson's 2 tests were performed for the validation of immunogold staining, and Kolmogorov-Smirnov test was applied for the subcellular distribution analysis of TRPV1positive immunogold particles. ANOVAs were followed by Tukey-Kramer or Dunnet's test for mean comparisons. Data are presented as mean Ϯ SEM. A p value Ͻ0.05 was considered as significant. Statistical analyses were performed using Origin Pro 8 and STATISTICA analysis software.

Tonic cannabinoid control on GABA release at perisomatic synapses in the mouse hippocampus
Paired whole-cell patch-clamp recordings were performed in acute slice preparations from identified presynaptic CB 1 -positive interneurons and from postsynaptic pyramidal neurons located in the CA1 region of the mouse hippocampus (n ϭ 164 pairs). The presynaptic interneurons were selected based on the position of their cell bodies in the stratum radiatum, their large soma size and multipolar shape, and their characteristic regular action potential firing pattern showing prominent accommodation in response to suprathreshold depolarizing current pulses (Fig. 1A). Importantly, each presynaptic interneuron was filled with biocytin during the electrophysiological recording and then visualized and identified after the recording (Fig.  1A). Post hoc immunostaining confirmed the presence of CB 1 receptors on the axon terminals of all recorded interneurons used in this study (Fig. 1A).
CB 1 -positive interneurons with the typical regular spiking pattern represent a morphologically heterogeneous population of several distinct GABAergic cell types in the hippocampus, which can be best characterized on the basis of the layer-specific distribution of their axonal arbors (Cope et al., 2002;Pawelzik et al., 2002;Klausberger and Somogyi, 2008;Nissen et al., 2010;Szabó et al., 2014b;Dudok et al., 2015;Lenkey et al., 2015). Recently, we reported that phasic and tonic forms of synaptic endocannabinoid signaling are markedly different at the output synapses of perisomatically versus dendritically targeting interneurons in the CA1 . Therefore, we characterized each interneuron according to the distribution of its axons, and only those were included in this study that could be unequivocally classified either as a perisomatic or a dendritic cell having an axon arbor predominantly distributed in either the stratum pyramidale or stratum radiatum, respectively.
Action potentials in presynaptic, perisomatically projecting, CB 1 -positive interneurons evoked uIPSCs in postsynaptic CA1 pyramidal neurons ( Fig. 1 B, E). In agreement with previous findings in rats , the proportion of presynaptic action potentials that resulted in successful postsynaptic events ("successes") and the amplitudes of euIPSCs, including both successes and failures , were highly variable between individual connections, ranging from 17 to 91% and from 7 to 120.7 pA, respectively ( Fig. 1C, D, F,G). Nevertheless, independent analysis of two randomly selected datasets obtained from perisomatic connections in our two different laboratories (Irvine and Budapest) revealed consistently similar mean values for both successes and euIPSC amplitudes under control conditions. Neither the proportion of successful postsynaptic events (  Fig. 1E-G) also augmented both the success rate (predrug control, 57 Ϯ 11%; AM251, 69 Ϯ 8%; Wilcoxon's signedrank test, p Ͻ 0.05) and the euIPSC amplitude (predrug control, 49.9 Ϯ 15 pA; AM251, 74.7 Ϯ 22 pA; Wilcoxon's signed-rank test, p Ͻ 0.05), confirming that persistent CB 1 receptor activity regulates release probability and synaptic strength at perisomatic GABAergic connections in the mouse hippocampus.
In summary, these findings demonstrate that the neutral antagonist NESS0327 at a dose of 0.5 M effectively inhibits phasic CB 1 activation evoked by exogenous and endogenous cannabinoids, but it does not affect tonic cannabinoid signaling. Thus, the tonic blockade of neurotransmitter release probability is likely to be caused primarily by the presence of constitutively active CB 1 receptors on perisomatic GABAergic axon terminals.

Tonically active monoacylglycerol lipase controls 2-AGmediated endocannabinoid signaling at perisomatic GABAergic synapses
Endocannabinoid levels and the activity of their metabolizing enzymes dynamically change according to specific physiological states, e.g., strictly follow diurnal phases (Vaughn et al., 2010;Liedhegner et al., 2014;Hanlon et al., 2015). Moreover, impaired inactivation of anandamide or 2-AG attributable to reduced fatty acid amide hydrolase (FAAH) or monoacylglycerol lipase (MGL) expression, respectively, leads to an elevated endocannabinoid tone, which is often associated with pathophysiological mechanisms (Sumislawski et al., 2011;Dincheva et al., 2015). Therefore, in the next set of experiments, we aimed to determine whether anandamide, 2-AG, or the two endocannabinoids together may also be involved in tonic endocannabinoid signaling at perisomatic GABAergic synapses by using pharmacological inhibitors of endocannabinoid-degrading enzymes. As above, we used the technically challenging and labor-intensive paired whole-cell patch-clamp recording approach, because interference with any aspects of endocannabinoid signaling using pharmacological tools may trigger complex perturbations at the network level. In addition, post hoc anatomical identification of every interneuron was performed to ensure that our measurements were focused on the specific GABAergic connections of interest, because CB 1 receptors are only present on a subset of perisomatic GABAergic axon terminals (Katona et al., 1999). Therefore, the paired recording technique has distinct advantages compared with alternative approaches, such as the bulk electrical stimulation of mixed populations of fibers or the measurements of compound spontaneous inhibitory postsynaptic events (which are derived from more than one interneuron type). Electrophysiological measurements were complemented by LC/MS/MS to directly monitor changes in endocannabinoid levels in parallel with alterations in the magnitude of the synaptic endocannabinoid tone in similarly prepared and maintained control or drugtreated acute hippocampal slice preparations.
MGL is responsible for the inactivation of ϳ85% of the total 2-AG content in the brain (Dinh et al., 2002;Blankman et al., 2007). Moreover, recent studies demonstrated that this serine hydrolase plays a regulatory role in endocannabinoid-mediated short-term DSI (Straiker and Mackie, 2009;Pan et al., 2011). Therefore, we first tested the effects of JZL184 (4-nitrophenyl 4-[bis(1,3-benzodioxol-5-yl)(hydroxy)methyl]piperidine-1-carboxylate), an irreversible inhibitor of MGL (Long et al., 2009b). Although JZL184 has a 300-fold selectivity for MGL over FAAH (Long et al., 2009b), the primary degrading enzyme of anandamide, we took precautions to reduce the potential off-target effects of this compound on anandamide levels. Besides using a low dose (100 nM), 2-AG and anandamide levels were also measured by LC/MS/MS. The data showed that a 40 min pretreatment of acute hippocampal slices with JZL184 evoked a robust 105% increase in hippocampal 2-AG levels compared with vehicle-treated slices ( Fig. 3A; paired t test, p Ͻ 0.001; n ϭ 9 animals in each group), whereas anandamide levels did not change significantly ( Fig. 3A; paired t test, p Ͼ 0.05; n ϭ 9 animals in each group).
These mass spectrometry and electrophysiology results indicated that 2-AG is mobilized constitutively in acute hippocampal slice preparations. To investigate more directly this possibility, we measured 2-AG levels in the absence of any pharmacological perturbation. Surprisingly, acute hippocampal slices produced 2-AG even under basal conditions, because comparison of two sets of samples kept in ACSF revealed 41% higher 2-AG levels in slices snap-frozen at a later time point (35 min; Fig. 3E). MGL can partially keep the tonic elevation of 2-AG levels under control, because the same measurement in the presence of JZL184 (100 nM) resulted in a much larger increase in 2-AG levels (89%; Fig.  3F ). Next, we aimed to test directly the effect of the tonic release of 2-AG on GABAergic transmission. After establishing stable paired recording between a CB 1 -positive perisomatic interneuron and a CA1 pyramidal neuron, the baseline levels of successes and euIPSCs were measured, and then the effect of acute administration of JZL184 (100 nM) was followed through 20 min. Notably, blockade of MGL significantly reduced the strength of perisomatic GABAergic connections within 12 min of JZL184 application ( Fig. 3F-H ). The maximal decrease in the proportion of successful postsynaptic events was down to 23 Ϯ 6% of the baseline levels (Fig. 3G), whereas the amplitude of euIPSCs was dropped to 29 Ϯ 8% of the baseline values (Fig. 3H ) at 16 min after drug application (one-way ANOVA with repeated measures; n ϭ 4 pairs, p Ͻ 0.001 for both parameters).
Previous studies provided direct evidence that tonic endocannabinoid signaling is a homosynaptic phenomenon (Hentges et al., 2005;Zhu and Lovinger, 2005;Roberto et al., 2010). In these studies, the increased GABAergic synaptic transmission observed in the presence of 10 M AM251 could be blocked by calcium chelation in the postsynaptic neuron. Therefore, we also recorded GABAergic events by using an intracellular pipette solution containing 10 mM BAPTA. Remarkably, the presence of BAPTA in the postsynaptic neuron vetoed the acute effect of MGL blockade (Fig. 3G,H ), demonstrating that the postsynaptic neuron is tonically releasing 2-AG, which then affects GABA release probability in a homosynaptic manner. Specifically, both the success rate (95 Ϯ 1% of baseline values; n ϭ 4 pairs) and the euIPSC amplitude (86 Ϯ 1% of baseline values; n ϭ 4 pairs) remained unchanged (one-way ANOVA with repeated measures; n ϭ 4 pairs, p Ͼ 0.05 for both parameters), when the MGL inhibitor was applied in the presence of BAPTA in the pipette that was used to record the uIPSCs from the postsynaptic cell. Finally, to fully exclude any potential confounding effect of inverse agonism by AM251, the neutral antagonist NESS0327 was used to prove unequivocally that tonic 2-AG activation of CB 1 receptors was necessary to diminish GABAergic synaptic transmission (Fig. 3G,H ). Indeed, when NESS0327 (0.5 M) was coapplied with JZL184 (100 nM), neither the proportion of successful postsynaptic events (control, 92 Ϯ 5% of baseline value; JZL184 ϩ NESS0327, 94 Ϯ 8%; n ϭ 4 pairs) nor the amplitude of euIPSCs (control, 82.6 Ϯ 9.2 pA of baseline value; JZL184 ϩ NESS0327, 84.1 Ϯ 15.2 pA; n ϭ 4 pairs) were changed (oneway ANOVA with repeated-measures; p Ͼ 0.05 for both parameters).
Collectively, these findings demonstrate that 2-AG is generated constitutively in acute hippocampal slice preparations and that tissue levels of 2-AG are under the tonic control of MGL activity. As a result, persistent MGL activity plays a critical role in setting the strength of GABAergic synaptic transmission at the CB 1 -expressing GABAergic perisomatic inputs to pyramidal cells.
These data together show that anandamide is also constitutively generated in hippocampal slice preparations and its tissue level is tonically controlled by FAAH activity. However, in contrast to the effects of the elevated 2-AG levels, highly increased anandamide levels have no direct effect on the strength of neurotransmission at perisomatic GABAergic synapses.

Neither MGL nor FAAH regulates GABA release at dendritic inhibitory synapses
Tonic endocannabinoid signaling is not present at the efferent synapses of dendritically projecting CA1 interneurons, despite the fact that the axon terminals of these interneurons do possess functional CB 1 receptors . This was especially surprising because diacylglycerol lipase-␣ (DGL-␣), the synthesizing enzyme of 2-AG (Bisogno et al., 2003), is concentrated mainly in the dendritic spine heads of pyramidal neurons Yoshida et al., 2006). The finding that MGL tonically controls the synaptic effects of 2-AG at perisomatic GABAergic synapses raised the possibility that elevated MGL activity in dendritically projecting interneuron boutons could mask the presence of a weak tonic 2-AG effect at dendritic GABAergic synapses. To test this possibility, paired recordings were performed from presynaptic dendritically projecting CB 1 -positive interneurons (Fig. 5A) and postsynaptic CA1 pyramidal neurons in acute hippocampal slices.
These observations indicate the absence of constitutive CB 1 activity, tonic endocannabinoid signaling, and the lack of functionally detectable tonic MGL and FAAH activities at dendritic GABAergic synapses in the mouse hippocampus. The interneuron-type specificity of tonic 2-AG signaling also indicates that the robust JZL184 effect on perisomatic GABAergic synaptic transmission was not the consequence of unidentified nonspecific network mechanisms because of a potential elevation of 2-AG throughout the hippocampal tissue.

Increased endocannabinoid tone affects synaptic functions without changing the number of presynaptic CB 1 receptors
The above results were somewhat unexpected considering the general view that anandamide may be better suited to serve as a tonic endocannabinoid signal (Alger, 2012; Ohno-Shosaku and Kano, 2014; Piomelli, 2014). This notion is based in part on the slower time course of anandamide mobilization and the fact that anandamide displays partial agonism toward CB 1 , in contrast to 2-AG, which is a full agonist (Stella et al., 1997;Hillard, 2000). In accordance with this view, chronic elevations of 2-AG levels by long-term pharmacological or genetic blockade of MGL activity causes robust behavioral tolerance as a result of the massive downregulation of CB 1 receptors, whereas chronically increased anandamide levels attributable to the inhibition of FAAH do not evoke similar adaptive responses (Schlosburg et al., 2010). These findings obtained at the systems level also suggest that any experimental perturbation that increases endocannabinoid levels may also indirectly affect CB 1 receptor availability. Thus, any measured changes in synaptic strength could simply be the consequence of a change in receptor numbers.
We tested the possibility of alterations in receptor numbers after the increase in endocannabinoid tone using STORM super-resolution imaging (Rust et al., 2006;Dani et al., 2010;Dudok et al., 2015) at CB 1 receptor-expressing perisomatic synapses onto CA1 pyramidal cells, i.e., at the synapse type that exhibits tonic endocannabinoid-mediated inhibition of GABA release. CB 1 -positive perisomatic interneurons were filled with biocytin during patch-clamp recordings in either vehicle-treated or enzyme (MGL or FAAH) inhibitor-treated acute hippocampal slices. The CB 1 content of the biocytinlabeled axon terminals of interneurons was analyzed at the nanoscale level to distinguish surface and internalized receptors (Fig. 6A). Because the vast majority of CB 1 receptors are found in the plasma membrane (Nyíri et al., 2005), fitting a convex hull onto the outermost LPs representing CB 1 receptors could reliably identify the receptor population on the axon terminal surface (Dudok et al., 2015). Notably, an analysis of the nanoscale distribution of CB 1 receptors in 303 in- Figure 5. Blockade of endocannabinoid degradation does not affect synaptic functions at GABAergic synapses formed by CB 1 -expressing dendritically projecting interneurons. A, Camera lucida reconstruction illustrates a CB 1 -positive dendritically projecting GABAergic interneuron in the CA1 region of the mouse hippocampus. The dendrites (red) and most of the axons (blue) of this multipolar interneuron arborized in stratum radiatum (Rad.), but a few additional axon branches were also observed in the stratum pyramidale (Pyr.) and stratum oriens (Ori.). Scale bar, 50 m. The example voltage traces in response to depolarizing and hyperpolarizing current steps (Ϫ100 and 150 pA, respectively, from Ϫ60 mV) show a characteristic regular spiking pattern with prominent spike frequency adaptation, an electrophysiological fingerprint of CB 1 -positive interneurons. Calibration: 30 mV, 0.3 s. Confocal images demonstrate that the biocytin-filled boutons (labeled by arrows) of this interneuron are indeed CB 1 positive. Scale bar, 10 m. B, Representative traces demonstrate that a low dose of the synthetic CB 1 agonist WIN55,212-2 (20 nM) significantly reduced synaptic strength at both dendritic and perisomatic GABAergic connections. C, D, Application of WIN55,212-2 reduced the success rates and euIPSC amplitudes more strongly at perisomatic connections. E-J, Application of neither 1 M AM251 nor 10 M AM251 unmasked persistent CB 1 activity on the axon terminals of CB 1 -expressing dendritic interneurons. K, Despite the presence of functional CB 1 receptors on the axon terminals, GABAergic synaptic transmission originating from dendritic interneurons is not affected by inhibiting endocannabinoid hydrolysis by either the MGL inhibitor JZL184 (100 nM; 40 min) or the FAAH inhibitor PF3845 (1 M; 40 min). L, M, Neither the proportion of successful postsynaptic events nor the amplitude of euIPSCs were altered by JZL184 or PF3845 pretreatment, indicating that the absence of dendritic tonic inhibition is not attributable to high MGL or FAAH activity at dendritic connections. The numbers of individual experiments are shown in parentheses. Error bars represent SEM. *p Ͻ 0.05, **p Ͻ 0.01; ns, not significant. dividual axon terminals from 13 identified perisomatically projecting interneurons revealed no significant changes in the total receptor number (Fig. 6B) or in the ratio of internalized/ surface membrane CB 1 receptors (Fig. 6C) after JZL184 (100 nM) or PF3845 (1 M) application for similar amounts of time as in the electrophysiological experiments described above (40 min; Tukey-Kramer test, p Ͼ 0.05 in all four measurements, n ϭ 7 vehicle-treated cells compared to 3-3 JZL184-or PF3845-treated cells). These observations demonstrate that the pretreatment protocol used in the present study does not cause a change in CB 1 receptor numbers on perisomatic GABAergic axon terminals, despite the strong increase in endocannabinoid levels attributable to the blockade of endocannabinoid degradation.

Tonic anandamide signaling interacts with tonic 2-AG signaling
The biological rationale of why the brain needs two endocannabinoid messengers is not fully understood (Di Marzo and De Petrocellis, 2012). A potential clue may be that the remarkable biochemical diversity of endocannabinoid synthesis routes (Blankman and Cravatt, 2013) provides an opportunity for specific signaling mechanisms to selectively recruit endocannabinoid synthesis under certain physiological or pathophysiological conditions. This also raises the possibility that coincident physiological signals may sometimes trigger the mobilization of both endocannabinoids at the same time, which then interact with each other. Therefore, in the next set of experiments, we asked the question whether tonic 2-AG signaling at perisomatic GABAergic synapses could be fine-tuned by anandamide signaling. One potential explanation for the lack of effects of FAAH inhibition on GABAergic synaptic strength could stem from the fact that anandamide, as mentioned above, is a partial agonist of CB 1 , and hence it may not elicit a response by itself at these specific GABAergic synapses. As a consequence, it may either facilitate or occlude the effects on tonic 2-AG signaling depending on how 2-AG saturates the functionally relevant CB 1 receptor pool on the axon terminals. Indeed, in paired recordings from CB 1 -positive perisomatic interneurons and CA1 pyramidal neurons (Fig. 7A), when MGL and FAAH were inhibited simultaneously by JZL184 (100 nM) and PF3845 (1 M), both the proportion of successful postsynaptic events ( Fig. 7B; JZL184, 17 Ϯ 3%, n ϭ 10 pairs; JZL184 ϩ PF3845, 41 Ϯ 8%, n ϭ 6 pairs; unpaired t test, p Ͻ 0.05) and the amplitude of euIPSCs ( Fig. 7C; JZL184, 8.9 Ϯ 2.1 pA, n ϭ 10 pairs; JZL184 ϩ PF3845, 37.9 Ϯ 11.0 pA, n ϭ 6 pairs; unpaired t test, p Ͻ 0.05) were significantly larger compared with the selective blockade of MGL with JZL184 alone. Thus, this observation demonstrates that, although enhanced anandamide signaling is not effective in itself, it can readily interfere with tonic 2-AG signaling at perisomatic GABAergic synapses.
These findings may result from an occlusion of 2-AG effects on presynaptic CB 1 activation by anandamide. However, the facts that FAAH is postsynaptic, whereas MGL is presynaptic (Gulyás Figure 6. Acute inhibition of endocannabinoid-degrading enzymes does not affect CB 1 numbers on the plasma membrane of perisomatic interneuron boutons. A, CB 1 -positive perisomatic interneurons were filled with biocytin in acute hippocampal slice preparations that were pretreated with vehicle (0.1% DMSO; 40 min), JZL184 (100 nM; 40 min), or PF3845 (1 M; 40 min) to determine whether increased 2-AG or anandamide levels affect the number of CB 1 receptors on perisomatic axon terminals. Deconvolved confocal images of identified boutons of perisomatic interneurons that were visualized by biocytin labeling (orange) are shown together with the corresponding STORM images of CB 1 immunostaining. LPs (green dots) in the raw STORM images represent the position of CB 1 receptors in the axon terminals. To analyze the nanoscale distribution of CB 1 receptors, individual LPs were first visualized in 3D and illustrated using Visual Molecular Dynamics. Based on the acquired 3D localization coordinates, a convex hull was then fitted onto the outermost CB 1 LPs to approximate the plasma membrane. To distinguish membrane-attached LPs (green) and those that are likely intracellular receptors located away from the surface of the convex hull (blue), an internalization index was calculated (for details, see Materials and Methods). B, C, Summary data of the STORM super-resolution imaging experiment reveals that neither the total number of CB 1 LPs nor the proportion of intracellular CB 1 LPs are changed significantly by pretreatment of MGL or FAAH inhibitors, indicating that any observed changes in synaptic functions during pharmacological treatments are not attributable to an altered number of available CB 1 receptors on the axon terminal surface. The number of individual experiments is shown in parentheses. Data were normalized to average control CB 1 number of localization points (NLP) values. Error bars represent SEM. ns, Not significant; Ctrl, Control. et al., 2004), and that brain 2-AG concentrations are 170 times greater than anandamide concentrations (Stella et al., 1997) do not support this scenario. Conversely, DGL-␣, the synthesizing enzyme of 2-AG, also occurs postsynaptically in the somatodendritic domain of pyramidal neurons, in which FAAH is located (Gulyás et al., 2004;Katona et al., 2006;Yoshida et al., 2006). This molecular architecture indicates that postsynaptic elevation of anandamide levels may interfere with the postsynaptic mobilization of 2-AG. Indeed, coapplication of JZL184 (100 nM) with PF3845 (1 M) significantly reduced the increase in 2-AG levels compared with MGL blockade alone ( Fig. 7D; paired t test, p Ͻ 0.05; n ϭ 5 slices). In striking contrast, increased anandamide levels induced by PF3845 pretreatment were not attenuated by the coapplication of PF3845 with JZL184 ( Fig. 7E; paired t test, p Ͼ 0.05; n ϭ 5 slices).
Together, these results uncovered that tonic 2-AG signaling at perisomatic GABAergic synapses in the hippocampus are controlled by the anandamide-degrading enzyme FAAH via potential postsynaptic interaction mechanisms affecting 2-AG synthesis.

TRPV1 channels are located on intracellular membrane cisternae at perisomatic symmetrical synapses
To explore how postsynaptic anandamide signaling may interfere with tonic retrograde 2-AG signaling, in the final set of experiments, we tested the hypothesis whether TRPV1 receptors represent a link between the two lipid signaling pathways at hippocampal GABAergic synapses. TRPV1 appeared to be a potential candidate, because anandamide is a full agonist of TRPV1 (Zygmunt et al., 1999), and anatomical evidence at the light microscopic level shows that TRPV1 is present in the somatodendritic domain of CA3 pyramidal neurons together with FAAH (Cristino et al., 2008). In addition, emerging evidence indicates that anandamide signaling elicits synaptic plasticity via postsynaptic TRPV1 activation in other brain areas (Chávez et al., 2010(Chávez et al., , 2014Grueter et al., 2010;Puente et al., 2011), and, most importantly, anandamide inhibits metabotropic glutamate receptorinduced 2-AG synthesis and phasic endocannabinoid signaling in the striatum via TRPV1 activation (Maccarrone et al., 2008).
Although recent evidence showed that TRPV1 is found postsynaptically at excitatory synapses (Puente et al., 2011(Puente et al., , 2015, there was no direct evidence as to whether this ion channel is also located at hippocampal GABAergic synapses, which would be a prerequisite for it to serve as the regulator of tonic endocannabinoid signaling at perisomatic inhibitory synapses. Therefore, we used immunogold electron microscopy to study the localization of TRPV1 at symmetrical synapses in the stratum pyramidale and stratum radiatum of the CA1 subregion of the hippocampus (Fig.  8 A, B). In wild-type mice, immunogold particles representing the position of TRPV1 were found frequently in the cell body of CA1 pyramidal neurons (Fig. 8A). Moreover, the subcellular distribution analysis of TRPV1 immunolabeling revealed a significant population of this ion channel postsynaptically, adjacent to symmetrical, putative GABAergic synapses in the stratum pyramidale ( Fig. 8 A, C). In contrast, immunogold particles were found only occasionally in the latter postsynaptic compartment in TRPV1 knock-out mice (Fig. 8 B, C), validating the specificity of the postsynaptic distribution pattern of TRPV1 at inhibitory synapses (Pearson's 2 test, p Ͻ 0.001; n ϭ 429 and 445 synapses in two wild-type and two knock-out mice, respectively). Specific TRPV1 immunolabeling was not found in the presynaptic GABAergic axon terminals or on mitochondria. Analysis of single 60 nm ultrathin sections, including normalization of the data to the nonspecific labeling within the same compartment in TRPV1 knock-out mice, revealed that 18% of perisomatic GABAergic synapses were positive for TRPV1. Given the size of CB 1 -positive perisomatic symmetrical synapses, which span several consecutive ultrathin sections, and considering their proportion among all symmetrical synapses in the stratum pyramidale of the CA1 subfield (35-40%;Takács et al., 2015), our data are consistent with the possibility that most, if not all, CB 1 -positive perisomatic synapses contain postsynaptic TRPV1 channels. Conversely, no significant difference was detected in the ratio of immunogold-containing dendritic GABAergic synapses in the stratum radiatum between WT and TRPV1 knock-out mice (6.08 and 3.7%; n ϭ 115 and 108 dendritic synapses, respectively; p ϭ 0.41, 2 test).
The precise subcellular localization of the nonselective cation channel TRPV1 has also functional significance. Therefore, we first measured the distance of individual gold particles from the synapse. In total, 68% of gold particles were found adjacent to symmetrical synapses (within 200 nm), which was significantly different from the background distribution observed in knock-out animals ( Fig. 8D; Kolmogorov-Smirnov test, p Ͻ 0.001; n ϭ 128 and 33 gold particles from two wild-type and two knock-out mice, respectively). Surprisingly, the majority of gold particles representing the subcellular position of TRPV1 was found to be attached to intracellular membrane cisternae (Fig. 8A 1 ,A 2 ) instead of a plasma membrane localization (55 Ϯ 6 vs 18 Ϯ 4%). The subcellular compartment of the remaining 27 Ϯ 2% of gold particles could not be identified unequivocally in the absence of clearly outlined nearby membranes, but these particles were all away from the plasma membrane. Considering that TRPV1 is an integral membrane channel and that those gold particles that were found close to the plasma membrane may also have be- Figure 8. TRPV1 is present postsynaptically at perisomatic GABAergic synapses and regulates synaptic strength at CB 1expressing perisomatic GABAergic synapses. A, At the electron microscopic level, immunogold staining revealed that TRPV1 channels are located postsynaptically at perisomatic GABAergic synapses (labeled by asterisk) formed on the somata of CA1 pyramidal neurons in wild-type mice. A 1 , A 2 , Higher-magnification electron micrographs show that gold particles 4 representing the position of TRPV1 were often attached to putative intracellular membrane cisternae (highlighted with black line in A 2 ). B, In contrast, postsynaptic immunogold particles were usually not located at symmetrical GABAergic synapses obtained from TRPV1 knock-out mice. C, Summary bar graph illustrates that the proportion of randomly selected GABAergic synapses forming symmetrical synapses on the cell bodies of CA1 pyramidal neurons that contained immunogold particles in close proximity to the synapse was significantly higher on sections derived from wild-type mouse hippocampus compared with that obtained from TRPV1 knock-out mice. D, To assess the subsynaptic distribution of TRPV1 channels, immunogold particle distribution was further analyzed by measuring distance of each gold particle from the postsynaptic membrane. Notably, a sharply biased distribution of gold particles toward the postsynaptic membrane was observed. Altogether, 68% of all gold particles were found in the vicinity of the postsynaptic plasma membrane (Ͻ200 nm) and six times more gold particles were found in wild-type versus TRPV1 knock-out animals in the first two 100 nm bins. E, In accordance with these anatomical findings, coapplication of the selective TRPV1 antagonist AMG9810 (1 M; 40 min) could efficiently rescue the effect of the FAAH inhibitor on 2-AGmediated tonic inhibition of GABA release. F, G, Both the ratio of successful postsynaptic events and the amplitude of euIPSCs decreased significantly when AMG9810 was applied together with JZL184 and PF3845 compared with those when JZL184 and PF3845 were applied alone. H-J, MGL blockade with JZL184 (100 nM) strongly reduced success rate and euIPSCs compared with vehicle treatment (ctrl) in TRPV1 knock-out animals. However, the antagonistic effect of FAAH inhibition (seen in F and G in wild-type animals) was completely absent in TRPV1 knock-out mice (H). Importantly, when JZL184 (100 nM) was applied alone or together with PF3845 (1 M), the success rate of action potentials evoking a postsynaptic IPSC and the amplitude of euIPSCs did not differ between the two conditions. The number of individual experiments is shown in parentheses. Error bars represent SEM. *p Ͻ 0.05, ***p Ͻ 0.001; ns, not significant; Ctrl, control. longed to intracellular membrane segments that were not identifiable, these data suggest that a surprisingly high proportion of TRPV1 is present on intracellular membrane compartments positioned under the postsynaptic plasma membrane of GABAergic synapses.
Collectively, these findings provide direct evidence that highly compartmentalized TRPV1 channels are concentrated on postsynaptic intracellular membranes at hippocampal perisomatic, but not at dendritic, GABAergic synapses. Moreover, the data show that TRPV1 activation by endovanilloid signaling regulates the magnitude of tonic 2-AG signaling at perisomatic GABAergic synapses.

Discussion
The qualitative diversity and quantitative variability of synaptic transmission are critically determined by neuromodulatory signaling (Marder, 2012). Persistent CB 1 receptor activity represents one of the major modulatory mechanisms to establish synaptic strength (Losonczy et al., 2004). Our present findings indicate the following: (1) constitutively active CB 1 receptors set GABA release probability selectively at perisomatic synapses in the hippocampus; (2) 2-AG is the primary mediator of ligand-driven tonic endocannabinoid signaling, but it is tightly controlled by presynaptic MGL activity; (3) a major regulatory mechanism calibrating the homosynaptic 2-AG-mediated endocannabinoid tone is endovanilloid signaling, which requires the activation of TRPV1 channels predominantly located on postsynaptic intracellular stores at GABAergic synapses; and (4) endovanilloid regulation of the 2-AG-mediated tone is controlled by FAAH. Together, these findings demonstrate a striking spatial segregation and functional division of labor of distinct molecular components of the endocannabinoid system in the tonic control of neurotransmitter release.

Persistent CB 1 activity at hippocampal GABAergic synapses
Although the mechanistic details and functional importance of phasic endocannabinoid signaling have received widespread attention (Cachope, 2012;Castillo et al., 2012), the molecular mechanisms and significance of tonic endocannabinoid signaling are much less understood. Tonic signaling represents a major mode of operation for several other neuromodulatory systems, with tonic adenosine and ambient GABA signaling being two major examples (Brickley and Mody, 2012; Dias et al., 2013), suggesting that a deeper understanding of tonic cannabinoid control on neurotransmitter release may also lead to important insights. Therefore, our primary goal was to delineate how tonic endocannabinoid signaling is organized at hippocampal GABAergic synapses. Because various physiological or pathophysiological processes modulate endocannabinoid-mediated synaptic plasticity (Iremonger et al., 2013;Melis et al., 2014), our secondary goal was to identify new regulatory mechanisms that can fine-tune the extent of tonic endocannabinoid signaling at inhibitory synapses.
In the present study, the data revealed that AM251, an inverse agonist of CB 1 receptors, could readily increase GABAergic neurotransmission. In contrast, NESS0327, a neutral antagonist, did not have an effect on tonic cannabinoid signaling, although it could block exogenous and endogenous ligand-activated CB 1 signaling. This intriguing finding raises the possibility that a certain proportion of the highly abundant CB 1 receptors on perisomatic boutons can take up a constitutively active conformation and thereby persistently block nearby voltage-gated calcium channels and GABA release (Szabó et al., 2014a). Indeed, biophysical modeling and mutational analysis identified several active conformations of CB 1 receptors that trigger distinct downstream signaling mechanisms with different kinetics and determine subcellular receptor targeting (D'Antona et al., 2006;Gyombolai et al., 2015). In light of previous evidence that CB 1mediated inhibition of GABA release depends on the activity pattern of the presynaptic neuron (Losonczy et al., 2004; together with the emerging therapeutic significance of CB 1 inverse agonism (Pertwee, 2005;Tam et al., 2012;Meye et al., 2013), it will be pivotal to determine in the future the extent of constitutive CB 1 activity at other specific circuit locations and during distinct physiological or pathophysiological states.

Tonic 2-AG signaling is controlled by MGL activity at hippocampal GABAergic synapses
An alternative mechanism leading to persistent CB 1 activity relies on tonic endocannabinoid mobilization. Indeed, somatodendritic CB 1 receptors on hippocampal neurons are activated tonically by 2-AG (Leterrier et al., 2006;Thibault et al., 2013;Ladarre et al., 2014). Thus, although it is conceivable that the allosteric conformational change induced by NESS0327 docking prevents the extracellular 2-AG from reaching its intramembrane binding pocket on CB 1 receptors in the presynaptic membrane, the possibility that NESS0327 fails to displace prebound 2-AG also cannot be fully excluded. Our finding that 2-AG is generated tonically in acute slice preparations indicates that the latter scenario is possible and has also general implication for synaptic research performed in acute slice preparations because of the widespread distribution of presynaptic CB 1 receptors throughout the brain .
Our parallel lipid measurements and electrophysiological recordings revealed that MGL activity limits 2-AG levels in the hippocampus and controls how postsynaptically mobilized 2-AG affects GABA release. In contrast, anandamide and FAAH did not have direct effects on GABAergic synapses. Generally, these findings are consistent with the anatomical distribution of the molecular constituents of the endocannabinoid system. FAAH is found somatodendritically throughout the brain (Egertová et al., 1998;Gulyás et al., 2004); hence, it is segregated spatially from the presynaptic effector site of tonic endocannabinoid signaling. MGL is distributed ubiquitously within axon terminals, which is the ideal subcellular position to terminate 2-AG effects at presynaptic CB 1 receptors (Gulyás et al., 2004;Ludányi et al., 2008;Uchigashima et al., 2011, Horváth et al., 2014. Because downregulation of MGL levels is associated with various physiological conditions and pathological situations (Sumislawski et al., 2011;Kallendrusch et al., 2012;Serrano et al., 2012), our results also highlight the importance to identify those mechanisms by which the presynaptic neuron can adjust MGL activity.

Synapse specificity of tonic cannabinoid signaling
Our previous findings uncovered tonic endocannabinoid signaling at perisomatic, but not dendritic, synaptic connections . Present observations now extend these findings by showing that neither tonic MGL nor tonic FAAH activity affects GABAergic transmission, despite the presence of functionally active CB 1 receptors on dendritically projecting interneuron boutons. One simple explanation is that cannabinoid signaling is quantitatively different between perisomatic and dendritic axon terminals because of the distinct nanoscale architecture of the two terminals (Dudok et al., 2015). Indeed, we found that a CB 1 receptor agonist more strongly inhibits GABA release from perisomatically projecting interneurons. An alternative explanation could be the cell-type-specific expression of cannabinoid receptor-interacting proteins, which bind to the C terminus of CB 1 and block constitutive receptor activity (Niehaus et al., 2007;Smith et al., 2015).
The synapse specificity also provides clues as to why tonic endocannabinoid signaling may sometimes appear absent, when electrical stimulation of a heterogeneous mixture of fiber types is used to evoke compound IPSCs (Pan et al., 2011). Such a paradigm was shown to require the genetic inactivation of MGL to uncover the presence of endocannabinoid tone (Pan et al., 2011). Similarly, the absence of effect after FAAH blockade does not exclude that anandamide conducts tonic endocannabinoid signaling under specific physiological or pathophysiological conditions, such as during chronic inhibition of neuronal firing (Kim and Alger, 2010). Considering that postsynaptic neurons exploit tonic endocannabinoid signaling in an activity-and cell-type-specific manner (Losonczy et al., 2004;Hentges et al., 2005;Oliet et al., 2007;Huang and Woolley, 2012), our findings highlight the importance of focusing anatomical and physiological analyses on identified circuit locations to elucidate how cannabinoid signaling regulates circuit dynamics.
Besides these independent synaptic functions of endocannabinoid and endovanilloid signaling, we found that these two pathways antagonistically interact to adjust synaptic strength at hippocampal inhibitory synapses. Endovanilloid signaling was reported to control 2-AG-mediated short-term plasticity evoked by acute stimulation of mGlu5 glutamate receptors or M 1 muscarinic acetylcholine receptors in the striatum (Maccarrone et al., 2008;Musella et al., 2010). The model proposed by Maccarrone et al. (2008) also suggested that increased anandamide levels activate TRPV1 channels, which then leads to an inhibition of the 2-AG-synthesizing enzyme DGL-␣. Consequently, mGlu5triggered increase in striatal 2-AG levels and phasic endocannabinoid signaling were impaired after FAAH inhibition (Maccarrone et al., 2008). Our findings are entirely consistent with this model by showing the following: (1) selective FAAH blockade elevates anandamide levels but attenuates increased hippocampal 2-AG levels induced by MGL inhibition; (2) the latter effect is associated with a weakened tonic 2-AG signaling; (3) reduced 2-AG signaling after FAAH blockade can be rescued by using a potent TRPV1 antagonist or in TRPV1 knock-out mice; and (4) TRPV1 is present postsynaptically at GABAergic synapses.
The mechanistic similarity of how endovanilloid signaling controls endocannabinoid signaling in different brain areas suggests that this functional interaction may be a more general phenomenon than hitherto thought. Emerging evidence indicates that TRPV1 activation has opposite behavioral function with respect to CB 1 signaling in several cognitive processes, including hippocampus-dependent learning and anxiety (Marsch et al., 2007;Li et al., 2008, Moreira et al., 2012You et al., 2012). Importantly, human FAAH variants that cause reduced enzyme activity are also associated with decreased anxiety-related behaviors and attenuated trauma in posttraumatic stress disorder patients (Pardini et al., 2012;Dincheva et al., 2015). Thus, uncovering the functional interaction of these two signaling systems in relation to specific behavioral processes will also help to understand the neurobiological mechanisms underlying the potential of FAAH inhibitors as novel anxiolytics (Gunduz-Cinar et al., 2013).