Type 1 cannabinoid receptor (CB1) is expressed in different neuronal populations in the mammalian brain. In particular, CB1 on GABAergic or glutamatergic neurons exerts different functions and display different pharmacological properties in vivo. This suggests the existence of neuron-type specific signalling pathways activated by different subpopulations of CB1. In this study, we analysed CB1 expression, binding and signalling in the hippocampus of conditional mutant mice, bearing CB1 deletion in GABAergic (GABA-CB1-KO mice) or cortical glutamatergic neurons (Glu-CB1-KO mice). Compared to their wild-type littermates, Glu-CB1-KO displayed a small decrease of CB1 mRNA amount, immunoreactivity and [³H]CP55,940 binding. Conversely, GABA-CB1-KO mice showed a drastic reduction of these parameters, confirming that CB1 is present at much higher density on hippocampal GABAergic interneurons than glutamatergic neurons. Surprisingly, however, saturation analysis of HU210-stimulated [35S]GTPγS binding demonstrated that ‘glutamatergic’ CB1 is more efficiently coupled to G protein signalling than ‘GABAergic’ CB1. Thus, the minority of CB1 on glutamatergic neurons is paradoxically several fold more strongly coupled to G protein signalling than ‘GABAergic’ CB1. This selective signalling mechanism raises the possibility of designing novel cannabinoid ligands that differentially activate only a subset of physiological effects of CB1 stimulation, thereby optimizing therapeutic action.
Type 1 cannabinoid receptor (CB1) is one of the most abundant G protein-coupled receptors in the mammalian brain with highest expression levels in the cerebellum, basal ganglia, cortex and limbic system (Marsicano and Kuner 2008). CB1 generally couples to Gi/o proteins to inhibit adenylyl cyclase (Howlett and Fleming 1984). Further intracellular effects include regulation of numerous kinases, transcription factors and ion channels (Demuth and Molleman 2006; Bosier et al. 2010). These intracellular effects collectively drive CB1′s cellular functions, of which the most prominent is the retrograde inhibition of transmitter release (Kano et al. 2009). Several studies demonstrated that CB1 activation inhibits the release of glutamate, GABA, glycine, acetylcholine, norepinephrine, dopamine, serotonin and cholecystokinin (Kano et al. 2009).
In the cortical regions of the mammalian brain, CB1 is mainly expressed in two major neuronal populations, GABAergic interneurons and glutamatergic principal neurons. There is a robust difference between the expression levels of CB1 in these two types of neurons. GABAergic interneurons express very high levels of CB1, whereas glutamatergic neurons express low to moderate levels (Marsicano and Lutz 1999). This implies that the most important inhibitory and stimulatory neurotransmitters in the cortex are, at least partially, under the control of the same regulatory system.
To dissect the physiological roles of these various receptor populations, we have generated several conditional mutant mice in which CB1 is specifically deleted in a particular neuronal population (Marsicano et al. 2003; Monory et al. 2006, 2007). Two of these mouse lines, Glu-CB1-KO and GABA-CB1-KO (Monory et al. 2006), are particularly interesting to study the differential impact of CB1-dependent control of cortical glutamatergic and GABAergic neurons.
Indeed, studies with these mice provided ample evidence about the involvement of these two populations of CB1 in mammalian brain physiology. Interestingly, contrary to general expectations, CB1 on glutamatergic cells, though much lower in numbers, produced in some cases more pronounced effects than their counterparts on GABAergic cells. Thus, behavioural analysis of Glu-CB1-KO and GABA-CB1-KO showed strong involvement of CB1 on glutamatergic cells in functions partly or fully dependent on hippocampus, such as protection against excitotoxic insults (Monory et al. 2006), stress response (Steiner et al. 2008), fear coping, stress and anxiety (Jacob et al. 2009; Kamprath et al. 2009; Dubreucq et al. 2012; Metna-Laurent et al. 2012). Similarly, CB1 on glutamatergic cells plays an important role in feeding behaviour (Lafenêtre et al. 2009; Bellocchio et al. 2010) and brain development (Mulder et al. 2008). Conversely, CB1 expressed on GABAergic neurons seems to be dispensable for several functions of endocannabinoid signalling or bears a completely opposite impact on these functions (Monory et al. 2006; Lafenêtre et al. 2009; Bellocchio et al. 2010; Dubreucq et al. 2012; Metna-Laurent et al. 2012). These observations might be simply because of the inhibition by CB1 activation of both excitatory and inhibitory neurotransmission, which logically will lead to different network effects. However, the administration of exogenous CB1 agonist to conditional mutant mice suggested that the CB1 pools expressed on the two neuronal populations likely possess different pharmacological properties. Thus, GABAergic neurons are not involved in the so-called tetrad effect induced by cannabinoids, which is mainly mediated by ‘glutamatergic’ CB1 (Monory et al. 2007). Interestingly, cannabinoids often exert biphasic effects on several behaviours, with low and high doses resulting in opposite effects (Sulcova et al. 1998; Chaperon and Thiébot 1999; Tzavara et al. 2003). We recently found that CB1 on glutamatergic cells is responsible for the effects of low doses of agonists, while CB1 on GABAergic cells responds only to higher doses (Bellocchio et al. 2010; Rey et al. 2012).
Altogether, these data suggest that CB1 on GABAergic or glutamatergic neurons might differ in its binding and/or signalling properties. Using hippocampal protein extracts from conditional mutant mice, we describe here neuronal type-specific G protein signalling that might provide mechanistic explanation for differential functions of CB1 and for the biphasic effects of cannabinoids.
Materials and methods
Mutant mice, specifically lacking CB1 in cortical glutamatergic (Glu-CB1-KO; Monory et al. 2006) or GABAergic neurons (GABA-CB1-KO; Monory et al. 2006) and their wild-type littermates in a predominant C57BL/6N background (> 7 backcrosses into C57BL/6N) were used in this study. Glu-CB1-KO mice were obtained by crossing CB1f/f (Marsicano et al. 2003) with NEX-Cre mice (Kleppisch et al. 2003; Schwab et al. 2000). GABA-CB1-KO mice were generated by crossing CB1f/f (Marsicano et al. 2003) with Dlx5/6-Cre (Monory et al. 2006; Massa et al. 2010). Genotyping was performed by PCR as described for Glu-CB1-KO, GABA-CB1-KO (Monory et al. 2006) and for CB1f/f (Marsicano et al. 2003). The use of a LacZ reporter for Dlx5/6-Cre (Monory et al. 2006) or NEX-Cre (Monory et al. 2006; Goebbels et al. 2006) showed that Cre recombinase expression in these transgenic mice recapitulates GABAergic and cortical glutamatergic neuronal expression with high accuracy. The detailed anatomical characterization of CB1 expression in GABA-CB1-KO and Glu-CB1-KO was published in Monory et al. (2006) and Bellocchio et al. (2010). Animals were 3–8 months old, housed in a temperature- and humidity-controlled room with a 12 h light–dark cycle and had access to food and water ad libitum. The experimental protocols were carried out in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC) and approved by the Ethical Committee on animal care and use of Rhineland-Palatinate, Germany.
Mice were anaesthetized with isofluran, decapitated and brains were quickly removed. For radioligand binding and in situ hybridization, isolated brains were snap-frozen on dry ice and stored at −80°C until use. Brains were removed from −80°C, equilibrated in the cryostat to −18°C and mounted on Tissue Freezing Medium (Jung, Nussloch, Germany). Cryosections of 18 μm thickness were prepared on a Microm HM560 cryostat (Microm, Walldorf, Germany), transferred onto cold Super Frost Plus microscope slides (Menzel, Braunschweig, Germany), briefly dried at RT and stored at −20°C. For qPCR, brain regions of interest were quickly isolated on ice under a binocular, snap-frozen on dry-ice and stored at −80°C for RNA isolation. For [35S]GTPγS binding assays, hippocampi were dissected on ice and then homogenized in ice-cold homogenization buffer (10 mM Tris-HCl, 2 mM EDTA, 1 mM dithiothreitol, 50 μM phenylmethylsulfonyl fluoride; pH 8.0) using a glass homogenizer. Protein concentration was determined using Bio-Rad protein assay (Bio-Rad Laboratories, Munich, Germany). Homogenates were aliquoted and stored at −80°C until use.
In situ hybridization
To label CB1+-cells, a 1530 bp long [35S]-labelled CB1 receptor riboprobe was used while GABAergic cells were identified by a 1041-bp long digoxigenin (DIG) labelled glutamic acid decarboxylase (GAD65) riboprobe. Generation of probe sequences is described by Marsicano and Lutz (1999); synthesis of labelled probes is described by Hermann et al. (2002).
For linearization of the probe-containing plasmids and in vitro transcription of probes, restriction enzymes (New England Biolabs, Ipswich, MA, USA) and RNA polymerases (Roche Molecular Diagnostics, Mannheim, Germany) were as follows: CB1 sense, PstI, T7; CB1 antisense, BamHI, T3; GAD65 sense, EcoRI, T7; GAD65 antisense, BamHI, T3. In situ hybridization experiments using sense controls gave no detectable signal (data not shown), whereas antisense probes gave distribution patterns identical to previously published results (Marsicano and Lutz 1999; Monory et al. 2006).
In situ hybridization was carried out as described in Hermann et al. (2002). Briefly, frozen slides were warmed up and fixed in ice-cold 4% paraformaldehyde (PFA), rinsed then incubated in methanol containing 1% H2O2 to block endogenous peroxidase activity. After another rinsing step, incubations with 0.2 M HCl and 50 mM TE with 0.4 U/mL Proteinase K followed to permeabilize tissue samples. PFA fixation step was then repeated subsequently, slides were incubated in triethanolamine/acetic acid anhydride solution to reduce non-specific binding of probes. After another rinsing step, slide pre-treatment was finished with tissue dehydration in a series of graded ethanol solutions. Air-dried tissue sections were covered with hybridization buffer containing 50 000–70 000 cpm/μL [35S]-labelled CB1 receptor riboprobe and 300 ng/mL DIG-labelled GAD65 riboprobe and incubated overnight at 54°C. High stringency post-hybridization washes were carried out at 62°C – slides were washed in three different formamide-saline sodium citrate buffer (SSC) solutions of descending SSC concentrations. This washing series was followed by incubation with a 15 mM iodacetamide solution to deactivate endogenous alkaline phosphatases. After a rinsing step, slides were incubated in two different blocking solutions: first in 4% heat-inactivated sheep serum then in 0.5% Perkin-Elmer blocking reagent. This was followed by incubation with anti-DIG(Fab)-POD antibody (1 : 1200; Roche Molecular Diagnostics) for 1.5 h at 30°C. Slides were rinsed then the signal amplification step, incubation with a tyramide-biotin solution, followed. After another rinsing step slides were incubated with streptavidine-AP conjugates (1 : 1000; Roche Molecular Diagnostics) for 1 h at 30°C. After yet another rinsing step non-radioactive signal was visualized by the alkaline phosphatase substrate Vector Red (Vector Laboratories, Burlingame, CA, USA). Reaction was stopped by immersing the slides in phosphate-buffered saline (PBS) after appropriate signal strength was achieved. Slides were then fixed in 2.5% glutaraldehyde, rinsed in 1-1× SSC then dehydrated in a series of graded ethanol solutions. To estimate radioactive signal strength, air-dried slides were exposed to Biomax MR film (Kodak, Germany) overnight. After development of the film, slides were dipped in photo emulsion (NTB2, Kodak, Stuttgart, Germany), dried and kept at 4°C for 5–30 days before developing. Slides were then counter-stained with toluidine blue and mounted in Roti-Histol (Roth, Karlsruhe, Germany).
Numerical evaluation of coexpression of CB1 and GAD65
In the hippocampus, there are two different CB1-expressing cell populations: low CB1-expressing glutamatergic pyramidal neurons and high CB1-expressing GABAergic interneurons (Marsicano and Lutz 1999). To validate the deletion of CB1 in Glu-CB1-KO mice from glutamatergic neurons, but not from GABAergic neurons, all red stained cells (GABAergic neurons) and all cells displaying both red staining and silver grains (GABAergic, high CB1-expressing neurons) were counted in hippocampi of Glu-CB1-KO mice and their wild-type littermates. The numerical evaluation of CB1-positive cells was performed in three distinct areas of the hippocampal formation, CA1, CA3 region and dentate gyrus. From both genotypes, three animals were taken and from all animals three sections were selected, ensuring that all sections are originating of the same hippocampal areas. Sections were analysed on a Leica DM RA/RXA Microscope (Leica Microsystems, Wetzlar, Germany) by a researcher blind to the genotype of the samples. Values are given as mean ± SEM.
Immunohistochemistry was performed according to Häring et al. (2007) with small modifications. Briefly, mice were deeply anaesthetized with pentobarbital, trans-cardially perfused with PBS containing 5 U/mL heparin and followed by perfusion with 4% PFA in PBS. The isolated brains were fixed for 24 h in 4% PFA in PBS solution, treated with 30% sucrose/PBS solution for 48 h and stored at −80 °C until use. Thirty micrometre thick coronal brain sections were prepared on a Microm HM560 cryostat (Microm) and stored at −20°C in cryoprotection solution (25% glycerol, 25% ethylene glycol and 50% PBS) until use. The free-floating brain sections were first rinsed in Tris-buffered saline (TBS) (25 mM Tris/HCl, 150 mM NaCl, pH 7.6) for 10 min, and then incubated in methanol containing 1.5% H2O2. After washing twice in PBS for 10 min, sections were incubated in blocking solution (5% normal donkey serum, 2.5% bovine serum albumin, 0.3% Triton X-100 in TBS) for 1 h. Afterwards, the sections were incubated overnight at 4°C with a polyclonal antibody against CB1 receptor (rabbit L15 antiserum, directed against the last 15 amino acids of CB1 receptor, diluted 1 : 5000; kind gift of Dr. Ken Mackie, Department of Psychological and Brain Sciences and Program in Neuroscience, Indiana University, Bloomington, USA) diluted in blocking solution. On the subsequent day, sections were washed five times in TBS-T (TBS, 0.1% Triton-X100), 5 min each, and then incubated for 2 h with a goat anti-rabbit antibody conjugated with horseradish peroxidase (Dianova, Hamburg, Germany) 1 : 100 diluted in blocking solution. The incubation was followed by five washing steps in TBS-T and subsequent incubation in Nova Red Solution (Vector Laboratories) for 5–10 min. Then, sections were washed twice for 2 min in distilled water, carefully transferred into TBS and mounted onto glass slides. After drying at RT for 2–4 h, slides were dipped for 2 s into distilled water, dried overnight and finally mounted in Roti-Histol (Roth).
RNA isolation and cDNA synthesis
Frozen tissue samples were transferred to 2 mL tubes from a Precellys ceramic kit (ceramic bead diameter 1.4 mm) containing homogenization buffer RLT from the RNeasy Mini-Kit (Qiagen, Hilden, Germany; ß-mercaptoethanol added, Carl Roth) and tissue was homogenized with a Precellys 24 (Peqlab, Erlangen, Germany) at 6000 rpm for 20 s. Total RNA was isolated using the RNeasy Mini-Kit (Qiagen) including the on-column DNA digestion step (RNase-Free DNase kit, Qiagen). Total RNA was reverse-transcribed using High Capacity cDNA Reverse Transcription Kit with random primer hexamers (Applied Biosystems, Carlsbad, CA, USA).
Quantification of cDNA was performed with an ABI 7300 real time PCR cycler (Applied Biosystems) utilizing the SYBR green method. Transcript-specific primer pairs for Gusb and CB1 were designed using VectorNTI software (Invitrogen, Darmstadt, Germany; Gusb: sense 5′ CTCTGGTGGCCTTACCTGAT, antisense 5′ CAGTTGTTGTCACCTTCACCTC; CB1: sense 5′ CTTCCACGTGTTCCACCGCA, antisense 5′ CCCACAGATGCTGTGAAGGAGG). PCR efficiencies were obtained from standard curves of serial dilutions of whole brain cDNA. PCR was performed in a volume of 20 μL containing 10 μL Power SYBR Green PCR Mastermix (2× concentrated; Applied Biosystems), 3 μL of each forward and reverse primer (2 μM) and 4 μL pre-diluted cDNA (pre-diluted to approximately 5 ng/μL after cDNA synthesis). Reactions were performed in duplicates. The thermal profile used for amplification was 2 min at 50°C, 15 min at 95°C, and 40 cycles of 15 s at 95°C, 30 s at 60°C and 30 s at 72°C. At the end of the amplification phase, a melting-curve analysis was carried out to confirm the formation of a single PCR product. Expression levels of the quantified transcripts were first normalized to that of the reference gene Gusb and then to that of control mice.
Freshly isolated hippocampi were homogenized in 0.5 mL ice-cold membrane buffer (50 mM Tris-HCl, pH 7.4, 3 mM MgCl2, 0.2 mM EGTA) containing protease inhibitors (Complete, Roche Applied Science, Mannheim, Germany) using a glass homogenizer. Protein concentration was determined using Bio-Rad protein assay (Bio-Rad Laboratories). Fifty microgram of protein was mixed with 5× Laemmli reducing sample buffer, denatured for 5 min at 60°C, separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes (Protran, Whatman; GE Healthcare, Dassel, Germany). Membranes were blocked for 45 min in TBS-T containing 5% fat-free milk powder, and incubated in TBS-T containing 2% fat-free milk powder and anti-CB1 antibody (1 : 500, rabbit, polyclonal, generated against the terminal 31 amino acids of mouse CB1; Frontier Institute, Hokkaido, Japan) and anti-actin antibody (1 : 4000, rabbit, monoclonal, Millipore, Temecula, CA, USA) overnight at 4°C. Membranes were then washed three times 10 min in TBS-T and incubated 1 h in TBS-T containing 5% fat-free milk powder and goat anti-rabbit horseradish peroxidase-conjugated secondary antibody (Dianova, Hamburg, Germany) followed by ECL-detection (Amersham ECL Prime, GE healthcare, Buckinghamshire, UK). Visualization of chemiluminescence was performed using a Fusion SL instrument (Peqlab Biotechnologies), image acquisition was performed with Fusion software (Vilber Lourmat, Eberhardzell, Germany) and quantification of chemiluminescence signal was performed with Bio-ID software (Vilber Lourmat). Data were collected from six animals per group, signal intensity measured as arbitrary units and expressed as percentage of wild-type.
Slides with brain sections were incubated in 50 mM Tris-HCl buffer (pH 7.4) containing 5% fatty acid-free bovine serum albumin (BSA) and 5 nM [³H]CP55,940 (139.6 Ci/mmol; PerkinElmer, Rodgau, Germany) for 3 h at 30°C. Non-specific binding was determined in the presence of 10 μM CP55,940 (Sigma-Aldrich, St Louis, MO, USA). After incubation, slides were washed twice for 90 min at 4°C in 50 mM Tris-HCl buffer (pH 7.4) containing 1% BSA, then fixed in the same buffer containing 0.5% formalin and dipped in distilled water. Dried sections were exposed to Kodak Biomax MR film (Eastman Kodak, Rochester, NY, USA) together with tritium standards (American Radiolabeled Chemicals, St Louis, MO, USA) for 5 weeks. Films were scanned under equal light conditions using a digital camera (Roper Scientific, Ottobrunn/Munich, Germany) and grey values of hippocampi were quantified based on known tritium standard values using MCID Basic 7.0 software (Imaging Research Inc., St. Catharines, ON, Canada). nCi/g tissue values were then converted to pmol/mg tissue data based on the specific activity value of [³H]CP55,940.
Agonist-stimulated [35S]GTPγS binding
Agonist-stimulated [35S]GTPγS binding assays were performed as described in Massa et al. (2010) with small modifications. Briefly, hippocampal homogenates were prepared as described above, then thawed, re-homogenized and diluted in assay buffer. Samples were pre-incubated for 10 min at 30°C in the presence of 0.004 U/mL adenosine deaminase (Sigma-Aldrich GmbH, Taufkirchen, Germany). Hippocampus homogenates (5-8 μg of protein) were incubated with 0.05 nM [35S]GTPγS (1250 Ci/mmol, PerkinElmer). For dose–response experiments, 10−12 M–10−5 M of CB1 agonists in assay buffer (50 mM Tris-HCl, pH 7.4, 3 mM MgCl2, 0.2 mM EGTA; 100 mM NaCl), containing 30 μM GDP and 0.5% fatty acid-free BSA in a final volume of 0.5 mL, were incubated for 60 min at 30°C. For [35S]GTPγS saturation experiments, the maximally effective 100 nM HU-210 concentration was used to stimulate [35S]GTPγS binding in assay buffer containing 30 μM GDP, 0.5% BSA and different concentrations of [35S]GTPγS (0.05, 0.1, 0.2, 0.5, 1, 2 and 5 nM) in a final volume of 0.5 mL for 60 min at 30°C. Specific binding values were generated by subtracting non-specific binding from total binding determined in presence of 30 μM GTPγS for each [35S]GTPγS concentration. Bound [35S]GTPγS was harvested by vacuum filtration through Whatman GF/B filters with a Brandel Cell Harvester (Gaithersburg, MD, USA), followed by three washing steps with 3 mL of ice-cold 50 mM Tris-HCl buffer (pH 7.4). After overnight incubation of filters in 2.5 mL scintillation cocktail Aquasafe 300 plus from Zinsser Analytic (Frankfurt/Main, Germany), radioactivity was measured by liquid scintillation counting using Tri-Carb 2800 TR (Perkin-Elmer, Boston, MA, USA).
All graphs and statistics were generated by GraphPad Prism 4 software (La Jolla, CA, USA). Statistical analysis was performed by Student′s t-test. Data are expressed as mean ± SEM or as mean ± SE. p < 0.05 was considered statistically significant.
Differential distribution of hippocampal CB1 protein in GABAergic and glutamatergic neurons
To detect functional differences between CB1 expressed in different neuronal populations, we first performed an immunohistochemical analysis in the hippocampal formation of Glu-CB1-KO and GABA-CB1-KO mice and their wild-type littermates (Fig. 1a–c). Distribution of CB1 protein in the wild-type brain (Fig. 1a) agreed with previously published results (Katona et al. 1999; Tsou et al. 1999; Egertová and Elphick 2000; Monory et al. 2006; Bellocchio et al. 2010), showing strong signal in the pyramidal cell layer of the CA1-CA3 region, and moderate to strong signal in the stratum radiatum of the CA1-CA3 region and in the stratum molecularis of the dentate gyrus (DG). Specific deletion of CB1 from cortical glutamatergic neurons (Glu-CB1-KO mice; Fig. 1b) resulted in no apparent change in the CB1 immunoreactivity pattern in the hippocampal formation of mice. Specifically, deleting CB1 from GABAergic neurons (GABA-CB1-KO mice, Fig. 1c), however, led to major loss of CB1 protein in the hippocampus, drastically changing CB1 immunoreactivity pattern. In these mutants, the large majority of CB1 signal is lost in the stratum radiatum of the CA1-CA3 region, and in the outer two-thirds of the stratum molecularis of the DG. Faint but clear signal is visible however in the pyramidal cell layer of the CA1-CA3 region and in the inner third of the stratum molecularis of the DG (Monory et al. 2006; Bellocchio et al. 2010).
As immunohistochemistry is an excellent tool to determine tissue distribution and cellular localization of proteins, but it is not well suited for quantitative analysis, we next performed Western blot analysis (Fig. 1e) to quantify the levels of CB1 protein in hippocampal preparations of Glu-CB1-KO and GABA-CB1-KO mice. We found only a slight and not significant difference between CB1 protein levels in the hippocampi of Glu-CB1-KO mice and their wild-type littermates (n = 6, wild-type: 100.0% ± 15.64%, Glu-CB1-KO: 70.60% ± 7.25%, p = 0.1189; Fig. 1f). In the hippocampi of GABA-CB1-KO mice, however, there was a substantial loss of CB1 protein level (n = 6, wild-type: 100.0% ± 11.76%, GABA-CB1-KO: 24.23% ± 5.34%, p = 0.0002; Fig. 1g). These data confirm that the large majority of CB1 protein in the hippocampal formation is present on GABAergic neurons (Kawamura et al. 2006; Bellocchio et al. 2010).
Cannabinoid binding in GABAergic and glutamatergic neurons of the hippocampus
The amount of a receptor protein might not fully relate to its ligand binding capacity as the latter is also influenced by receptor conformation, interacting partners, cofactors, signal transduction molecules (Maudsley et al. 2005; Nelson and Challiss 2007). We therefore performed radioligand binding experiments with [3H]CP55,940, a potent cannabinoid agonist, on hippocampal sections of Glu-CB1-KO and GABA-CB1-KO mice and their wild-type littermates (Fig. 2). The signal pattern of the [3H]CP55,940 ligand binding is similar to that seen with immunohistochemical analysis. However, the strong [3H]CP55,940 signal in the pyramidal cell layer and the stratum radiatum of the CA1-CA3 region and the stratum molecularis of the DG (Glu-CB1-WT mice; Fig. 2a, GABA-CB1-WT mice; Fig. 2c) seems only slightly changed after deletion of CB1 from hippocampal glutamatergic neurons (Glu-CB1-KO mice; Fig. 2b), but there is a major loss of [3H]CP55,940 ligand binding after deletion of CB1 from GABAergic neurons (GABA-CB1-KO mice; Fig. 2d). Quantification showed 9.35 ± 0.19 nCi/g bound [3H]CP55,940 in the hippocampi of Glu-CB1-WT mice and 9.38 ± 0.12 nCi/g in GABA-CB1-WT mice. In Glu-CB1-KO mice, we observed a slight but significant decrease of [3H]CP55,940 binding (Fig. 2e, 8.18 ± 0.13 nCi/g, 87% of wild-type; one-way anova, p < 0.01 Glu-CB1-KO vs. Glu-CB1-WT). In GABA-CB1-KO mice, however, a major loss of [3H]CP55,940 binding could be detected (Fig. 2e, 3.06 ± 0.10 nCi/g, 33% of wild-type) that was significantly different from values of wild-types as well as from Glu-CB1-KO (one-way anova, p < 0.001 GABA-CB1-KO vs. GABA-CB1-WT, p < 0.001 Glu-CB1-KO vs. GABA-CB1-KO).
Comparing the loss of [3H]CP55,940 binding to the apparent loss of CB1 protein measured by Western blot analysis in GABA-CB1-KO and Glu-CB1-KO mice (Figs 1f–g and 2e) reveals that no major difference in cannabinoid binding affinity exists between CB1 expressed on hippocampal GABAergic and glutamatergic neurons.
Specific loss of CB1 mRNA in glutamatergic neurons in Glu-CB1-KO mutant mice
Conditional mutants are excellent tools to dissect cell population-specific physiological or pathophysiological functions, but a specific limitation of this technology is the possible occurrence of unwanted ectopic Cre-mediated recombination, causing target gene deletion in other cell populations than originally planned. To check for such possible drawback, a double in situ hybridization study was performed to detect CB1 mRNA in combination with mRNA of GAD65, a specific marker of GABAergic neurons in the hippocampal formation of Glu-CB1-KO mice and their wild-type littermates. In cortical regions (such as neocortex and hippocampus), CB1 is expressed in two clearly distinct neuronal populations, with low levels in glutamatergic principal neurons and with high levels in cholecystokinin (CCK)-positive GABAergic interneurons (Marsicano and Lutz 1999). We have already shown that vGluT1-expressing (glutamatergic) cells do not express CB1 in Glu-CB1-KO mice (Monory et al. 2006; Bellocchio et al. 2010). However, the possible ectopic recombination in GABAergic neurons of these mutant mice, albeit not visually apparent, was not quantitatively tested so far. To this aim, we counted all neurons expressing GAD65, all neurons co-expressing GAD65 and CB1 and calculated the ratio of the two in CA1, CA3 and DG regions of the hippocampus (Table 1). Uniformly, in all tested regions, approximately 30% of all GABAergic neurons expressed CB1. We could not detect any significant difference between Glu-CB1-KO and wild-type littermates (p = 0.2392), indicating that no Cre-mediated recombination occurred in GABAergic neurons of Glu-CB1-KO mice. As a result of the dense localization of glutamatergic neurons in the hippocampus, the quantitative analysis of CB1 expression in these cells of GABA-CB1-KO was not feasible. However, qualitative imaging of double staining CB1/vGluT1 in the hippocampus and quantifications in the neocortex (Marinelli et al. 2009) strongly suggest that no ectopic unwanted recombination in glutamatergic neurons occurred in these mutant mice.
Table 1. Percentage of coexpression of CB1 mRNA with GAD65 mRNA in the hippocampal formation of Glu-CB1-KO and Glu-CB1-WT mice (n = 3) evaluated by double ISH. Values represent mean ± SEM
Number of cells expressing
% of coexpression
33.3 ± 3.8
117.3 ± 2.2
20.0 ± 1.5
80. 7 ± 4.1
10. 7 ± 0. 9
40. 7 ± 2.7
64.0 ± 4.0
238.7 ± 6.3
42.0 ± 1.0
151.7 ± 3.9
20.7 ± 1.9
78.3 ± 6.2
7.7 ± 0.9
36.0 ± 1.5
71.0 ± 3.1
265.7 ± 5.2
Quantification of CB1 mRNA in the hippocampi of Glu-CB1-KO and GABA-CB1-KO mice
For a quantitative analysis of CB1 mRNA amount in the hippocampi of Glu-CB1-KO and GABA-CB1-KO mice, we performed real time quantitative PCR. CB1 expression in Glu-CB1-KO (Fig. 3a) was significantly reduced to 66% (wild-type: 100.0% ± 6.3%, Glu-CB1-KO 65.8% ± 1.9%; p < 0.01), whereas in GABA-CB1-KO (Fig. 3b), the expression was significantly reduced to 27% (wild-type: 100.0% ± 7.5%, GABA-CB1-KO 26.9% ± 3.2%; p < 0.001) of the wild-type level. Essentially, these data are in good agreement with the ones quantifying CB1 protein amount (Glu-CB1-KO: 70% of WT, GABA-CB1-KO: 24% of WT; Fig. 1f–g) or binding (Glu-CB1-KO: 87% of WT, GABA-CB1-KO: 33% of WT; Fig. 2e).
Differential effects of cannabinoid-stimulated [35S]GTPγS binding
CB1 is a 7-transmembrane domain G protein-coupled receptor that upon agonist stimulation responds by activating heterotrimeric G proteins. Using the radioactively labelled non-hydrolysable GTP-analogue, [35S]GTPγS, we measured activation of hippocampal CB1 of Glu-CB1-KO and GABA-CB1-KO mice by two synthetic cannabinoids (HU-210 and CP55,940; Fig. 4) and two endocannabinoids (2-arachidonoylglycerol; 2-AG and anandamide; AEA; Fig. 5).
Synthetic cannabinoids strongly stimulated [35S]GTPγS binding in hippocampal homogenates of Glu-CB1-WT and GABA-CB1-WT mice with no apparent difference between the two lines (Fig. 4, Table 2). 2-AG and AEA stimulated [35S]GTPγS binding with a slightly lower Emax (Fig. 5). Similarly, however, we could not observe any difference between Glu-CB1-WT and GABA-CB1-WT mice (Table 2). Loss of CB1 in glutamatergic neurons (Glu-CB1-KO) led to a substantial decrease of Emax of agonist-stimulated [35S]GTPγS binding with all the agonists tested (Fig. 4 and 5; Table 2). Similarly, loss of CB1 in GABAergic neurons (GABA-CB1-KO) resulted in a significantly lower Emax of agonist-stimulated [35S]GTPγS binding in the hippocampus when HU-210, CP55,940 or 2-AG was tested (Fig. 4 and 5; Table 2). Comparing Glu-CB1-KO and GABA-CB1-KO mice demonstrated a significantly greater loss of agonist-stimulated [35S]GTPγS binding in Glu-CB1-KO mice (48.7% of the wild-type value in Glu-CB1-KO vs. 68.2% of the wild-type value in GABA-CB1-KO; Table 2). The potencies (EC50 values) of all cannabinoid agonists to stimulate [35S]GTPγS binding are comparable in all animals tested, except for 2-AG, which seems to be significantly more potent in Glu-CB1-KO mice than in their wild-type littermates. However, when compared with GABA-CB1-KO mice, no significant difference was observed.
Table 2. Comparison of the efficacy (Emax) and potency (EC50) of CB1 agonists to stimulate [35S]GTPγS binding in hippocampal homogenates
*p < 0.05, ***p < 0.001 KO versus WT. †p < 0.05, †††p < 0.001 Glu-CB1-KO versus GABA-CB1-KO (Student's t-test, n = 6, for AEA n = 5). Emax values are expressed as % specific [35S]GTPγS bound of basal.
Quantitative analysis of cannabinoid-activated G proteins in the hippocampi of Glu-CB1-KO and GABA-CB1-KO mice
Dose–response curves of different synthetic and endogenous cannabinoids showed a significant decrease of G protein activation in Glu-CB1-KO and, to a smaller extent, in GABA-CB1-KO compared with their wild-type littermates. We next performed saturation binding analysis of HU-210-stimulated [35S]GTPγS binding in hippocampal homogenates of Glu-CB1-KO and GABA-CB1-KO mice to quantify this effect (Fig. 6).
[35S]GTPγS saturation binding showed a significant loss of HU-210-activated G proteins in Glu-CB1-KO mice as revealed by 2-way anova (Fig. 6a; p < 0.0001), whereas no significant difference could be observed in GABA-CB1-KO mice (Fig. 6b; p = 0.2425). Bmax values show that Glu-CB1-KO mice have less than half of cannabinoid-activated G proteins than their wild-type littermates (Glu-CB1-WT: 66 ± 5 fmol/mg protein; Glu-CB1-KO: 27 ± 3 fmol/mg protein). At the same time, GABA-CB1-KO mice display a much smaller loss of cannabinoid-activated G proteins (GABA-CB1-WT: 64 ± 7 fmol/mg protein; GABA-CB1-KO: 62 ± 7 fmol/mg protein). KD values were comparable in all animals tested: Glu-CB1-WT: 1.00 ± 0.26 nM; Glu-CB1-KO: 1.70 ± 0.49 nM; GABA-CB1-WT: 0.60 ± 0.23 nM; GABA-CB1-KO: 1.32 ± 0.49 nM.
To indirectly test signalling efficiency of CB1 in GABAergic versus glutamatergic cells, we compared the amounts of CB1 and CB1 activated G proteins in the hippocampi of the mutant mice (Table 3). In GABA-CB1-KO mice, a substantial loss of receptor protein is associated with a minimal loss of CB1 activated G proteins while in Glu-CB1-KO a small loss of receptor protein accompanies a considerable loss of CB1 activated G proteins. Consequently, in GABAergic cells, CB1/G protein ratios are 2 : 1, but there are about 3 G proteins coupled to one CB1 molecule in glutamatergic cells, that is, there is a more than 6-fold difference between signalling efficiencies of CB1 in GABAergic versus glutamatergic cells.
Table 3. CB1 activated G protein amount in hippocampi of wild-type, Glu-CB1-KO and GABA-CB1-KO mice. CB1 protein amounts were measured by [3H]CP55,940 autoradiography; CB1 activated G protein amounts were measured by HU-210 stimulated [35S]GTPγS saturation binding
CB1 binding (pmol/mg)
G protein (pmol/mg)
G protein per CB1
CB1 is expressed in different neuronal populations in the mammalian brain (Marsicano and Lutz 1999; Tsou et al. 1999; Monory et al. 2006). To dissect the physiological roles of these various receptor populations, we have generated several conditional mutant mice in which CB1 is deleted in specific neuronal populations (Marsicano et al. 2003; Monory et al. 2006, 2007). As most glutamatergic principal neurons and many GABAergic interneurons express CB1 in cortical areas, two of these lines, Glu-CB1-KO and GABA-CB1-KO, were chosen for this project to study hippocampal CB1 expression, binding and G protein signalling in GABAergic versus glutamatergic cells.
In the hippocampal formation, mainly two distinct cell populations express CB1 mRNA: CCK+ GABAergic interneurons express high levels of CB1 mRNA (Marsicano and Lutz 1999), while glutamatergic principal cells (pyramidal neurons and mossy cells but not granular cells) express low amounts (Marsicano and Lutz 1999; Marsicano et al. 2003; Monory et al. 2006). CB1 is an axonal protein and for this reason there is a specific difference between mRNA and protein distribution, as the mRNA is located in the perinuclear soma. Consistently to this notion, intense CB1 protein staining is visible at sites where basket cell terminals surround pyramidal and granular cells. Somewhat weaker signal is visible in the strata molecularis > radiatum and > oriens, where the CB1 signal is predominantly originating from the interconnecting network of inhibitory interneurons innervating the dendritic arbour of pyramidal or granular cell bodies (Katona et al. 1999, 2001; Egertová and Elphick 2000). CB1 staining of glutamatergic terminals amounts to a much weaker and diffuse signal most prominently visible at the inner third of the stratum molecularis (Katona et al. 2006; Kawamura et al. 2006; Monory et al. 2006). Apart from these two major components of the CB1 population of the hippocampus, however, extrahippocampal neurons contribute to this population, too. Several regions send projecting neurons to the hippocampus, some of them are CB1+. Most important of these are the inputs from the entorhinal cortex which are glutamatergic pyramidal cells and are terminating in all parts of the hippocampus (Shepherd, 1990). Glutamatergic inputs arrive from subcortical regions too – hypothalamic projections terminate mostly in the stratum molecularis and CA3. In addition, neurons of several other neurotransmitter systems send their axons to the hippocampus. Notably, cholinergic neurons from the septum (Nyíri et al. 2005), serotonergic neurons from the raphe nuclei (Häring et al. 2007) and noradrenergic neurons from the locus coeruleus (Scavone et al. 2010) were already shown to express CB1. Though according to our estimation based on [3H]CP55,940 binding data, these represent only a few per cent of the entire hippocampal CB1 population, the existence of this CB1 expression, originating from extrahippocampal neurons, but located in the hippocampus, has to be considered when interpreting any functional analysis. In particular, basal forebrain cholinergic neurons arise from Dlx-expressing neurons (Bachy and Rétaux 2006), and in a lac-Z reporter mouse line using the same regulatory Dlx5/6 intergenic region (Stühmer et al. 2002) as present in the Dlx5/6-cre recombinase transgene (Monory et al. 2006) the septum is lacZ-positive. Consistently with this notion, septal area is devoid of CB1 mRNA expression in the GABA-CB1-KO mice (Monory et al. 2006).
As the overwhelming majority of hippocampal CB1 is on GABAergic neurons (Marsicano and Lutz 1999), immunohistochemistry and radioligand binding show only minimal, albeit significant, differences between Glu-CB1-KO and wild-type littermates. Nevertheless, previous studies using the same mouse models demonstrated the functional importance of ‘glutamatergic’ CB1 in several behavioural paradigms (Monory et al. 2006, 2007; Steiner et al. 2008; Lafenêtre et al. 2009; Jacob et al. 2009; Bellocchio et al. 2010; Häring et al. 2011; Metna-Laurent et al. 2012). In agreement with such in vivo data, this study shows that the small change in receptor expression induces considerable consequences on G protein signalling in the hippocampal formation.
In this study, we applied different techniques to estimate CB1 amount in hippocampi of Glu-CB1-KO and GABA-CB1-KO mice. Western blotting, radioligand binding and qPCR consistently showed that glutamatergic neurons contain about one quarter of all hippocampal CB1, while GABAergic neurons contain about three quarters of all hippocampal CB1.
The goal of our study was to understand more about how the information processing during CB1-dependent G protein signalling is different between inhibitory (GABAergic) and excitatory (glutamatergic) neurons. One obvious idea was that the signalling machinery responds to a certain ligand concentration with different sensitivity in the various neurons. Cell type-specific signalling (same agonist showing different intracellular effects through the same receptor in different cells) was already shown for CB1 (Peters and Scott 2009). However, contrary to our expectations, we found no difference in the ED50 values of [35S]GTPγS binding stimulated by two synthetic and two endocannabinoids in Glu-CB1-KO versus GABA-CB1-KO mice.
The Emax values, however, were significantly different in Glu-CB1-KO versus GABA-CB1-KO mice, indicating a higher amount of CB1-activated G proteins in glutamatergic cells. Quantitative analysis of G proteins in the mutant mice revealed that the relatively minor amount of receptor in glutamatergic cells is responsible for over 50% of the total cannabinoid-activated G proteins. Contrarily to what we observed in Glu-CB1-KO mice, hippocampal CB1 in GABAergic cells is responsible for only about 20–30% of the total cannabinoid-activated G proteins. In other words, signal amplification at the CB1-G protein interface is much higher in the glutamatergic cells than in the GABAergic cells. Taking into account that CB1 density is much higher in GABAergic cells than in glutamatergic cells, our data are in good agreement with the observation of Breivogel et al. (1997) that CB1-G protein coupling differs by region in the rat brain and the higher the receptor density in a particular region, the lower the receptor-G protein amplification factor is. We have now extended this observation to neuronal types as well.
The more effective CB1 signalling on glutamatergic cells is in good agreement with the observation that in the lateral amygdala, cannabinoid actions on glutamatergic synaptic transmission override those on GABAergic synaptic transmission, leading to an overall decrease of excitability (Azad et al. 2003). Moreover, according to behavioural studies (Bellocchio et al. 2010; Rey et al. 2012), systemic administration of low doses of cannabinoids activates CB1 on glutamatergic terminals, while high doses activate CB1 on GABAergic terminals. Somewhat conflicting these in vivo data, dose–response curves of cannabinoid-induced suppression of excitatory or inhibitory post-synaptic currents in hippocampal cultures revealed that the ED50 at glutamatergic synapses was approximately 30 times bigger than at GABAergic ones (Ohno-Shosaku et al. 2002). Electrophysiological measurements revealed another kind of difference between CB1 signalling in glutamatergic and GABAergic cells as well. These measurements showed that in glutamatergic cells CB1 has only a minimal (5–7%) tonic/constitutive activity (Roberto et al. 2010), while in GABAergic cells this value is 30–40% (Slanina and Schweitzer 2005). Therefore, the tonic and phasic activities of the endocannabinoid system seem to be unevenly distributed, CB1 in glutamatergic cells being in greater part responsible for the phasic activity than the receptor on GABAergic cells.
THC administration was shown to disrupt hippocampus-dependent memory formation through CB1 on GABAergic neurons (Puighermanal et al. 2009). Recently, it was shown that this memory-disturbing effect of THC is exacerbated by chronic caffeine (Panlilio et al. 2012) and this effect is mediated by the adenosine A1 receptor (Sousa et al. 2011). Interestingly, however, A1 receptor, though present in GABAergic neurons, does not directly participate in regulation of GABA release. Nevertheless, the study showed influence of A1 receptor activation on CB1-mediated decrease of GABA and glutamate release as well as CB1-mediated G-protein activation. Chronic caffeine treatment leads to an up-regulation of A1 and a down-regulation of CB1. Yet, THC-induced amnesic effects are intensified in caffeine-treated mice. As the study uses wild-type mice, it is not possible to tell, whether CB1 down-regulation occured in all hippocampal neurons or only in selected populations. These surprising results probably originate from a complex net effect of activation of different memory-related neuronal circuits.
An important limitation of experiments using [35S]GTPγS binding assays is that this approach has a strong bias towards pertussis toxin (PTX)-sensitive (Gi/Go) G proteins partly because PTX sensitive G proteins are more abundant and partly because they have a much faster guanine nucleotide exchange rate (Milligan 2003). Consequently, our experiments likely measured mainly CB1-dependent activation of Gi/Go proteins. Nevertheless, although the assay is not well suited for its detection, we cannot exclude the participation of other G proteins in the cannabinoid effects. Cannabinoid-induced activation of Gs (Glass and Felder 1997; Hampson et al. 2000; Bash et al. 2003; Peters and Scott 2009) and Gq proteins (Lauckner et al. 2005; McIntosh et al. 2007) were previously reported in in vitro systems. It is therefore possible that PTX-insensitive G proteins are responsible for part of the biphasic behavioural effects of cannabinoids.
Taken together, we describe here a neuron-type specific CB1-dependent G protein signalling in the mouse hippocampus. A far more effective receptor-G protein coupling in glutamatergic cells explains why this receptor population responds already to low doses of cannabinoids and why it also seems to be engaged in more physiological processes.
This study was supported by the German Research Foundation (by the Graduate School GRK1044 to F.S.; by Research Unit FOR926 to B.L. and K.M.), by the Fondation pour la Recherche Medicale to G.M. and by a MAIFOR Research Grant of the Johannes Gutenberg University to K.M. The authors thank Martin Purrio and Ruth Jelinek for technical assistance. The authors declare that they have no conflict of interest in this study. Authorship contributions: K. Monory and B. Lutz participated in research design; F. Steindel, R. Lerner, M. Häring and S. Ruehle conducted experiments; G. Marsicano, K. Monory and B. Lutz contributed new reagents or analytic tools; F. Steindel, R. Lerner and K. Monory performed data analysis; K. Monory, G. Marsicano, B. Lutz and F. Steindel wrote or contributed to the writing of the manuscript.