Monlunabant

Differential neuromodulatory role of endocannabinoids in the rodent trigeminal sensory ganglion and cerebral cortex relevant to pain processing

Francesca Eroli a, Inge C.M. Loonen b, Arn M.J.M. van den Maagdenberg b, c, Else A. Tolner b, c, Andrea Nistri a, *

a b s t r a c t
Endocannabinoids are suggested to control pain, even though their clinical use is not fully validated and the underlying mechanisms are incompletely understood. To clarify the targets of endocannabinoid actions, we studied how activation of the endocannabinoid CB1 receptor (CB1R) affects neuronal re- sponses in two in vitro preparations of rodents, namely the trigeminal sensory ganglion (TG) in culture and a coronal slice of the cerebral cortex. On TG small-medium size neurons, we tested whether sub- micromolar concentrations of the endogenous CB1R agonist anandamide (AEA) modulated inhibitory GABAA receptors and excitatory ATP-gated P2X3 receptors. AEA reversibly depressed GABA-mediated membrane currents without altering P2X3 receptor responses.

The AEA antagonism was non- competitive, prevented by the CB1R antagonist AM-251, mimicked by the other cannabinoids 2- arachidonylglycerol and WIN 55,212-2, and insensitive to TRPV1 blocker capsazepine. AEA inhibited the potentiation of GABAergic responses by the cAMP activator forskolin, in line with the canonical in- hibition of cAMP synthesis by CB1Rs. In the cerebral cortex, AEA or WIN 55,212-2 did not affect electrically-evoked local field potentials or characteristics of cortical spreading depolarization (CSD) elicited by high potassium application. The GABAA receptor blocker gabazine, however, strongly enhanced field potentials without affecting CSD properties, suggesting that CSD was not dominantly controlled by GABAergic mechanisms. Our data propose that, despite the widespread expression of CB1Rs peripherally and centrally, the functional effects of AEA are region-specific and depend on CB1R coupling to downstream effectors. Future studies concerned with the mechanisms of AEA analgesia should perhaps be directed to discrete subcortical nuclei processing trigeminal inputs.

1.Introduction
Modulation of pain processing by endocannabinoids is a well- known phenomenon used clinically to treat acute or chronic pain
(Guindon and Hohmann, 2009; Hohmann and Suplita, 2006; Pacher et al., 2006; Zogopoulos et al., 2013). Early reports indi- cated that the pain-suppressing activity by endocannabinoids implied a major site of action at the level of the central nervous system likely via depression of excitatory neurotransmission (Hohmann and Suplita, 2006; Manzanares et al., 2006), although peripheral targets have later been identified (Agarwal et al., 2007). Membrane receptors termed cannabinoid receptor-1 (CB1R) transduce the action of endocannabinoids like anandamide (AEA) and 2-arachidonoylglycerol (2-AG), the main natural ligands for CB1Rs (Agarwal et al., 2007; Devane et al., 1992; Guindon and Hohmann, 2009; Hohmann et al., 2005; Piomelli, 2005) via depression of cAMP synthesis (Howlett et al., 2010) on neuronsprocessing nociceptive and affective components of pain perception.

We recently observed that AEA down-regulated locomotor network activity in the rat spinal cord by inhibiting GABAA re- ceptors (Veeraraghavan and Nistri, 2015), a novel target of endo- cannabinoid action, even though a direct causal link between these events was not demonstrated. Since CB1Rs are present on sensory ganglion neurons (Price et al., 2003; Veress et al., 2013) and may be coupled to GABAA receptors (Kano et al., 2009), we wondered if CB1R activation may also control GABAergic mechanisms of sensory neurons that offer a simplified experimental model without the complexity of central brain networks (Fabbro and Nistri, 2004; Kondo et al., 1994; Nistri and Constanti, 1979). In vitro (Ault and Hildebrand, 1994; Waite et al., 2000) and in vivo (Jang et al., 2017) experiments have indicated that activation of GABAA receptors of sensory neurons may produce excitation and a pro-nociceptive effect by neuronal membrane depolarization (Adams and Brown, 1975). Thus, we investigated whether CB1R activity could modu- late GABAA receptors on trigeminal ganglion (TG) sensory neurons in culture, as in ganglion tissue GABA is proposed as a diffusible neurotransmitter via neuron-glia crosstalk (Hayasaki et al., 2006).

Though widely expressed in the central nervous system (Mackie, 2005), CB1Rs are distributed in a discrete manner in cortical layers (Gutie´rrez-Rodríguez et al., 2017). Since activation of presynaptic CB1Rs can decrease GABAergic inhibition (Hoffman and Lupica, 2000; el Manira and Clarac, 1994), this would be ex- pected to facilitate cortical network excitation. Strong neuronal excitation can produce cortical spreading depolarization (CSD) (Somjen, 2001), a wave of depolarization that can arise from visual areas to engulf the cerebral cortex and is likely responsible for migraine aura (Hadjikhani et al., 2001; Lauritzen, 1994). Interest- ingly, nociceptive stimulation of TG largely facilitates the onset of CSD (Toriumi et al., 2016). In turn, CSD is reported, at least in experimental animal models, to activate the trigeminovascular system responsible for pain signaling in headache (Noseda and Burstein, 2013), thus closely linking central and peripheral noci- ceptive processes.

The role of CB1R in cortical activity and CSD remains, however, unclear and apparently unrelated to GABAergic processes. In fact, earlier work with neocortical slices described suppressing effects of endocannabinoids on CSD, suggesting that CB1 activation may target glutamatergic signaling (Kazemi et al., 2012). Data from animal (Akerman et al., 2013; Greco et al., 2010a, 2010b; Nagy-Gro´cz et al., 2016) and clinical (Sarchielli et al., 2007) studies proposed that a deficit in CB1R activity may contribute to migraine attacks. Consequently, enhancing CB1R ac- tivity was put forward as potential migraine treatment and was shown to influence trigeminovascular nociception in rats (Akerman et al., 2013). Still, although many migraine patients may self- administer cannabis (Baron, 2015), which contain phytocannabi- noids that activate CB1 and CB2 receptors, there is little clinical evidence that this drug has a beneficial effect (Rhyne et al., 2016). The present investigation aimed at finding out if increased endo- cannabinoid (CB1R) activity may affect GABAAemediated effects on TG neurons, and synaptic network activity and CSD in the cerebral cortex to further explore potential targets for the action of endo- cannabinoids against pain.

2.Materials and methods
2.1.Animals
Primary cultures of TG were obtained from neonatal Wistar rats (P0-P2) and C57BL/6J mice (P12) mice. Because of the virtually complete homology (99.5% amino acid identity) between rat and mouse CB1R gene products (Abood et al., 1997; Matsuda et al., 1990), we used either species for the present experiments. For brain slice experiments, male C57BL/6J mice were used. All exper- imental procedures were performed following recommendations of the Italian and Dutch Animal Welfare Act and the European Communities Council Directive (2010/63/EU) and were approved by the SISSA (prot. 3599, May 2012) and LUMC (Dec 2014-13070) experimental animal ethical committees. All experiments were carried out in accordance with ARRIVE guidelines. All efforts were made to minimize suffering.

2.2.Electrophysiology of primary TG cultures
TG were isolated and dissociated as previously described (Hullugundi et al., 2013; Simonetti et al., 2006) and cultures were used 24 h after plating. From one animal we used both TG to pro- duce two or three cultures. As previously reported in detail (Hullugundi et al., 2013; Nair et al., 2010; Vilotti et al., 2013), tri- geminal neurons were superfused continuously (2 mL/min) with physiological solution containing (in mM): 152 NaCl, 5 KCl, 1 MgCl2, 2 CaCl2, 10 glucose, and 10 HEPES (pH adjusted to 7.4 with NaOH). Cells were patch-clamped in the whole-cell configuration using pipettes with a resistance of 3e4 MU when filled with the following solution (in mM): 140 KCl, 0.5 CaCl2, 2 MgCl2, 2 Mg2ATP3, 2 GTP, 10 HEPES, and 10 EGTA (pH adjusted to 7.2 with KOH). Recordings were performed on small- and medium-size neurons (capacitance < 22 pF; cell body diameter 15e25 mm) that all express functional GABAA receptors and mostly (about 80%) express func- tional P2X3 receptors (Hullugundi et al., 2013; Vilotti et al., 2013). Cells were held at —65 mV after correcting for the liquid junction potential; currents were filtered at 1 kHz using an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA, USA) and acquired by means of a DigiData 1200 Interface and pClamp 8.2 software (Molecular Devices, Sunnyvale, CA, USA). To induce P2X3 receptor currents, the synthetic agonist a,b-methylene-adenosine-5'- triphosphate (a,b-meATP; Sigma) was applied (2 s) at the concen- tration of 10 mM to evoke near maximal responses (Hullugundi et al., 2013; Franceschini et al., 2013; Nair et al., 2010), using a fast superfusion system (Rapid Solution Changer RSC-200; BioLogic Science Instruments, Claix, France). GABA was applied with 2-s pulses (Fabbro and Nistri, 2004; Vilotti et al., 2013) at various concentrations to obtain dose/response curves. For systematic testing the 10 mM concentration was employed. Various protocols were applied to investigate the effect of CB1R ligands on GABAA and P2X3 receptors. The endocannabinoid AEA was first tested at 5 and 10 mM concentrations as reported by previous studies (Price et al., 2004; Eljaschewitsch et al., 2006; Veeraraghavan and Nistri, 2015); then several concentrations were used to generate a full dose/response curve (Fig. 1D). For standard experiments the concentration used was 0.5 mM, which was the calculated IC50 value from dose/response experiments on GABA currents, as shown in Fig. 1D. The concentration of 2-AG (0.5 mM) was similar to the one of AEA and in line with previ- ously reported data (Lu et al., 2014; Sheinin et al., 2008) and confirmed with cAMP assay to be equieffective as 0.5 mM AEA (Laura Celotto and Sandra Vilotti, unpublished data). The synthetic cannabinoid agonist WIN 55, 212-2 mesylate was tested on TG cultured neurons at the concentration of 0.5, 1, 5 and 10 mM (Kazemi et al., 2012; Akerman et al., 2013; Sun et al., 2015) and routinely used for subsequent experiments at 5 mM as established with the concentration/effect plot (Fig. 4B). The 0.1 mM concen- tration of AM-251 was selected from preliminary experiments as adequate to inhibit the action by AEA on GABA currents without changes in control responses and it accords with previous reports (Baur et al., 2012; Lenkey et al., 2015). The TRPV1 antagonist cap- sazepine (CZ, 10 mM; Bevan et al., 1992) was used to verify theFig. 1. Effect of AEA or AM-251 on GABAA- or P2X3-mediated currents. A, Histograms show average peak amplitude values of GABA-mediated currents induced by pulse application of GABA (10 mM, 2 s) in trigeminal ganglion (TG) neurons in control, after application of AEA (0.5 mM, 10 min) and after 5 min washout (n ¼ 13, 13, 9, respectively;*p < 0.0004). At the bottom, representative traces from the same cell of GABA responses in control, after AEA application and washout. In this and subsequent figures washout ofdrug treatment is abbreviated as ‘w/o’. B, Histograms show average peak amplitudes of currents evoked by GABA pulses (10 mM, 2 s) in control condition or after AM-251 application (0.1 mM, 10 min); n ¼ 12. At the bottom, examples of GABA-activated currents in control and after AM-251 application to the same cell. C, Histograms show no change in a,b-meATP-evoked currents after 10 min application of AEA (0.5 mM) or AM-251 (0.1 mM); n ¼ 10, 8, respectively. On the right, representative traces of P2X3-mediated currents induced bypulse application of a,b-meATP (10 mM, 2 s) in control and after application of AEA or AM-251. Note different time calibrations from A, B. D, dose/response curve shows reduction of GABA responses amplitude induced by AEA application at different concentrations (0.1, 0.5, 1 and 5 mM); n ≥ 8 for each concentration. E, Dose/response curves for GABA-mediated currents in control and after 10 min AEA application (0.5 mM); n ≥ 17, n ≥ 13 for each concentration, in control and AEA curves respectively; p < 0.0004 at 30 and 100 mM, p ¼ 0.02 at 10 mM.possible involvement of TRPV1 receptors. Finally, we explored changes in GABA responses after stimulation of cAMP synthesis induced by forskolin (FSK, 20 mM; D'Arco et al., 2009). 2.3.Brain slice preparation Experiments were performed with adult mouse coronal slices containing the primary visual cortex. To obtain slices, the brain was removed following decapitation of a deeply isoflurane- anesthetized mouse and placed in ice-cold carbogenated (95% O2/ 5% CO2) slice solution, containing (in mM): NaCl 118.3, KCl 3.5, CaCl2 1, NaH2PO4 1, MgSO4 0.8, MgCl2: 4 NaHCO3 23.8 and glucose 15.1. The brain was dissected and cut into 350-mm thick coronal slices with a vibrating microtome (Leica biosystems, Wetzlar, Germany). Slices containing the primary visual cortex were incubated in recording solution containing (in mM): NaCl 118.8, KCl 3, CaCl2 1.8, NaH2PO4 1, MgSO4 0.8, NaHCO3 23.8 and glucose 15.1 for 1 h at room temperature and subsequently for at least 1 h at 37 ◦C when slices were transferred to a submerged recording chamber main- tained at 35 ◦C and perfused with carbogenated recording solution (flow rate 3 mL/min) for electrophysiological recordings (Tolner et al., 2007). 2.4.Brain slice electrophysiology Extracellular local field potential (LFP) and DC-potential changes were recorded with a glass microelectrode filled with 154 mM NaCl placed in layer 2/3 of the primary visual cortex. Evoked LFP re- sponses were induced by stimulating cortical layer 4 by delivering brief monophasic current pulses every 30 s via a bipolar platinum/ iridium electrode (25 mm exposed tip; CE2C55 FHC, Bowdoin, ME, USA) placed at approximately 0.25 mm from the recording elec- trode located in layer 2/3 (see schematic in Fig. 6A; Heimel et al., 2010). Data were filtered (LFP: 0.03 Hz to 2 kHz; DC: 5 Hz low- pass filter) using an EXT 10-2F amplifier (NPI electronic GmbH, Tamm, Germany), sampled at 5 kHz and acquired by means of a DigiData 1322A Interface and Clampex software (Molecular De- vices). After initial recording of LFP responses to 20-mA stimulation, experimental sessions followed the timeline shown in Supplementary Fig. 1A. Thus, we first recorded baseline LFP and DC activity for 5 min, and constructed input-output (IO) LFP plots for stimulus intensities ranging from 3 mA to 300 mA. Stimuli were given at the frequency of 0.033 Hz (five responses for each in- tensity) with a 200-ms pulse duration. From the IO curve, current intensities required to evoke 10, 20, 50, 70 and 100% LFP were used for subsequent stimulation protocols referred to as 'short IO-para- digm'. This short IO-paradigm was run three times throughout the experiment: before (IO 1) and after (IO 2) a first CSD induction under control conditions (no drug) and before (IO 3) a second CSD induction in the presence or absence of AEA, WIN 55,212-2 or the GABAA receptor antagonist gabazine. The first short IO-paradigm (IO 1) was followed by the first CSD induction protocol consisting of 1 min baseline recording, induction of CSD (CSD 1) by infusion of 40 mM KCl in recording solution for 60 s, followed by 9 min re- covery in recording solution. After the first CSD induction protocol and the second short IO-paradigm (IO 2), recordings were continued in recording solution (control) or in the same solution containing AEA (0.5 mM), WIN 55,212-2 (5.0 mM), or gabazine (1 mM). After this drug application, the third short IO-paradigm (IO 3) was performed and a second CSD (CSD 2) was induced in the presence (or absence) of a drug (Supplementary Fig. 1A). Data were analyzed offline using Clampfit (Molecular Devices) and Spike2 (Cambridge Electronic Design, Cambridge, UK) software. For anal- ysis of evoked LFP responses, the amplitude of the first negative wave (N1) was measured (Fig. 6A and B; Heimel et al., 2010). For IO curves, values were averaged from 5 responses per each stimulus intensity. For CSD events, amplitude and duration were determined (see sample trace in Fig. 7A). 2.5.Chemicals GABA, AEA, 2-AG, AM-251, SR144528 and gabazine (SR 95531 hydrobromide) were purchased from Tocris (Bristol, UK); a,b- meATP, CZ and FSK were purchased from Sigma (Milan, Italy); WIN 55, 212-2 Mesylate was purchased from MedChemExpress (Sol- lentuna, Sweden). The stock solution of AEA was prepared with ethanol (final concentration less than 0.05%). Stock solutions of 2- AG, WIN55,212-2, AM-251, SR144528, CZ and FSK were prepared by dissolving them in DMSO (final solvent 0.1%). GABA and gaba- zine stocks were prepared in distilled water. Fig. 2. Effect of AEA in the presence of CB1R or TRPV1 blockers. A, Histograms show average peak amplitudes of GABA-mediated responses in control, after 10 min co-application of AEA with AM-251 (0.5 and 0.1 mM, respectively) and after 5 min washout; n ¼ 21, 21, 16, respectively. On the right, representative traces of GABA-evoked currents in control, after AEA plus AM-251 co-administration and washout. B, Histograms summarize average peak amplitudes of GABA-mediated responses in control and after application of AEA alone or in the presence of CZ (10 mM, 10 min); n ¼ 13, 9 respectively; **p ¼ 0.00008, *p ¼ 0.005. On the right, representative traces of GABA-evoked currents in control and after 10 min application of AEA alone or co-applied with CZ. 2.6.Statistics Data are expressed as mean ± standard error of the mean (SEM), where n indicates the number of independent experiments or the number of investigated cells. Statistical analysis was performed using the Student's t-test or the ManneWhitney rank sum test, and the paired t-test or the Wilcoxon matched-pairs signed rank test after the software-directed choice of parametric or nonparametric data, and two-way repeated measures ANOVA (SigmaPlot and Systat Software Inc., San Jose, CA, USA and GraphPad Software, La Jolla, CA, USA). A p value of ≤0.05 was accepted as indicative of a statistically significant difference. 3.Results 3.1.Effect of AEA on GABAA- or P2X3-mediated currents in TG neurons Patch-clamp experiments were performed with TG neurons in culture to explore whether activation of CB1Rs by AEA could modify responses mediated by GABAA or P2X3 receptors. Fig. 1A shows that continuous superfusion with AEA (0.5 mM) significantly and selec- tively depressed GABA-mediated currents by an average of 52%: this effect was manifested within 10 min and was reversible with washout. AEA application did not involve any apparent change in the GABA current kinetics (see examples in Fig. 1A, bottom). Conversely, no effect of AEA was observed on currents elicited by a,b-meATP (10 mM, 2 s) (Fig. 1C). The application of the CB1R antagonist/inverse agonist AM-251 alone (0.1 mM) did not change the amplitude of GABAA or P2X3-mediated currents either (Fig. 1B and C). This suggests that, under these experimental conditions, there was no constitutive activity of CB1Rs modulating GABAA or P2X3 receptor activity. Furthermore, neither AEA nor AM-251 had an effect on baseline current (33 ± 7 pA in control versus 36 ± 5 pA after AEA, n ¼ 13; 30 ± 7 pA in control versus 36 ± 6 pA after AM- 251, n ¼ 14). Fig. 1D shows the dose/response curve for AEA-induced decrease in GABA response amplitude (AEA concentrations tested: 0.1, 0.5, 1 and 5 mM). Thus, AEA decreased GABA-mediated currents in a concentration-dependent fashion with a calculated IC50 value of 0.5 mM. The dose/response plot shown in Fig. 1E de- picts the depression induced by AEA on GABAA receptors activated by different GABA concentrations (1, 3, 10, 30 and 100 mM). Since the plot was shifted downwards in a non-parallel fashion, a non- competitive inhibition by AEA was apparent. To determine whether the decrease in GABA-activated currents induced by AEA was mediated by CB1Rs, the CB1R antagonist AM- 251 was applied as shown in Fig. 2A. When AEA plus AM-251 were continuously co-administered for 10 min, any depression of GABA- mediated currents was almost completely prevented, thus sup- porting the notion that GABAA receptors down-regulation was mediated by activation of CB1Rs. 3.2.TRPV1 channels were not involved in the effect of AEA on TG GABAA-currents A population of small- and medium-sized TG neurons express TRPV1 receptors (Bae et al., 2004; Guo et al., 1999), that are non- selective cation channels responsible for the transduction of certain noxious signals by sensory neurons (Caterina et al., 2000). Previous studies (Roberts et al., 2002; Smart et al., 2000; Zygmunt et al., 1999) have indicated the ability of high concentrations of AEA to activate TRPV1 receptors. To investigate whether the modulation by AEA of GABA receptors might involve TRPV1 activation, we tested the effect of AEA in the presence of CZ, a selective TRPV1 antagonist (Bevan et al., 1992; Zygmunt et al., 1999). Fig. 2B shows that continuous superfusion (10 min) of AEA plus CZ (10 mM) did not affect the depression induced by AEA on GABA-evoked cur- rents, thus indicating that the rapid down-regulation of GABAA receptor activity did not apparently require the intermediate step of TRPV1 channel activation. 3.3.2-AG effect on TG GABAA-evoked responses We further sought to find out if a different endocannabinoid might share the depressant action of AEA. To this end, we studied Fig. 3. Effect of 2-AG on GABAA-evoked responses. A, Histograms show average current amplitude values of responses evoked by 10 mM GABA (2-s pulses) in control, after application of 2-AG (0.5 mM, 10 min) and 5 min washout; n ¼ 9, 9, 7, respectively; *p < 0.005. B, C, Histograms show average current amplitudes of GABA-mediated responses in control and after 10 min co-application of 2-AG with AM-251 or SR144528 (0.1 mM) followed by 5 min washout; n ¼ 9, 7, respectively; *p < 0.004. Note that the effect of 2-AG on GABA responses is prevented by AM-251 co-application, while the presence of SR144528 does not affect the reduction of GABA currents by 2-AG. Fig. 4. Effect of WIN 55,212-2 on GABAA- or P2X3-evoked currents. A, Histograms show average current amplitude values of GABA-mediated currents induced by pulse application of GABA (10 mM, 2 s) to TG neurons in control solution, after application of WIN 55,212-2 (5 mM, 10 min), and after 5 min washout (n ¼ 12, 12, 7, respectively; *p < 0.00005). At the bottom, representative traces of GABA responses in control, after WIN 55,212-2 application and washout (all from the same cell). B, dose/response curve shows depression of GABA response amplitude induced by WIN 55,212-2 at different concentrations (0.5, 1, 5 and 10 mM); n ≥ 10 for each concentration. C, Histograms show no change in P2X3-mediated responses after 10 min application of WIN 55,212-2 (5 mM); n ¼ 14. On the right, representative traces of P2X3-mediated currents induced by pulse application of a,b-meATP (10 mM, 2 s) in control and after application of WIN 55,212-2. Note different time calibrations from A, C. In this and subsequent figures WIN 55,212-2 is abbreviated as WIN the effect of 2-AG that, like AEA, is an endogenous cannabinoid binding CB1 and CB2 receptors (Guindon and Hohmann, 2009; Howlett, 2002). The histogram in Fig. 3A shows that 2-AG appli- cation (0.5 mM, 10 min) significantly decreased GABA-activated currents by an average of 46%, i.e. an effect comparable to that of AEA, and reversible on a washout. The down-regulation of GABA responses induced by 2-AG was abolished by co-application of 2- AG with AM-251 (Fig. 3B), confirming the role of CB1Rs in the GABA depression produced by AEA or 2-AG application. As further validation of CB1R involvement, we tested the effect of the CB2 selective antagonist SR144528 (Bouaboula et al., 1999; Fan et al., 2011), that did not block the depression of GABA-activated cur- rents by 2-AG (Fig. 3C), indicating lack of CB2 receptor contribution. 3.4.Effect of WIN 55,212-2 on TG GABAA- or P2X3-evoked currents Endocannabinoids like AEA and 2-AG are also known to act on GPR55 receptors (Akerman et al., 2013; Kress and Kuner, 2009; Ryberg et al., 2007). To rule out the possibility that the observed modulation of GABAA currents may be due to signaling via GPR55, we also investigated the effect of WIN 55,212-2, a synthetic cannabinoid receptor agonist acting on both CB1 and CB2 receptors (Akerman et al., 2013; Corcoran et al., 2015; Kazemi et al., 2012; Sun et al., 2015), with no reported activity at other receptor subtypes (Akerman et al., 2013; Johns et al., 2007). To this end, we tested WIN 55,212-2 on GABAA or P2X3 responses. Histograms and trace ex- amples in Fig. 4A show that continuous superfusion with WIN 55,212-2 (5 mM) significantly down-regulated GABA-evoked cur- rents by 51% average in analogy with the effect of AEA and 2-AG. This effect was manifested within 10 min and was partially reversible with 5 min washout. On the contrary, continuous application of WIN 55,212-2 with the same protocol (5 mM, 10 min) did not affect currents mediated by P2X3 receptors (Fig. 4C). The plot in Fig. 4B shows the dose/response curve for WIN 55,212-2-dependent depression of GABA response amplitude (WIN 55,212-2 concentrations tested: 0.5, 1, 5 and 10 mM), from which the calculated IC50 value was 5 mM. These results further support the idea that GABAA receptor depression was mediated by CB1R activity without involving GPR55 receptor signaling. 3.5.Effect of FSK on TG GABAA currents CB1Rs are Gi/o protein-coupled receptors whose activation re- sults in a cascade of events that decrease the intracellular cAMP level, thus inhibiting the downstream PKA system (Howlett et al., 2010). We, therefore, investigated whether GABAA receptor activ- ity was modulated by cAMP and whether interference by CB1R activation with the cAMP signaling pathway was the intermediate link between these two phenomena. To this end, we used FSK to evoke activation of adenylyl cyclase and cAMP synthesis and tested how this approach could change responses evoked by GABA. His- tograms and representative traces (Fig. 5A) show that stimulation of cAMP synthesis by FSK (20 mM, 10-min application) enhanced by 34% the amplitude of GABA-mediated currents. The effect produced by FSK was inhibited by AEA that almost halved GABA responses (40% reduction), as shown in Fig. 5B. 3.6.Effects of AEA, WIN 55,212-2, or gabazine on LFP and CSD in the visual cortex First, slice experiments were performed to determine the effects of CB1R activation by AEA (0.5 mM) and WIN 55,212-2 (5.0 mM) on LFP responses and CSD properties in the cerebral cortex in vitro. Evoked LFP responses were taken as an index of local network synaptic activity (schematic and sample traces in Fig. 6AeB). Supplementary Fig. 1A shows the timeline of evoked LFP and CSD recording in the slice experiments in which short IO-paradigms were run before (IO 1) and after (IO 2) the first CSD induction (CSD 1). Full recovery of synaptic transmission following CSD was observed with the exception of the response at 70% maximum which was slightly, yet significantly decreased (Supplementary Fig. 1B). On average, LFP amplitude was unaffected by AEA when compared to control (Fig. 6C). There was also no difference between IO 3 in the presence of AEA compared to IO 3 of control slices kept in recording solution (p ¼ 0.84, two-way repeated measurements ANOVA). LFP amplitude was also unaffected by CB1 and CB2 receptor agonist WIN 55,212-2 when compared to control. CSD 1 and CSD 2 in untreated slices had similar amplitude (Supplementary Fig. 1C) and duration (Supplementary Fig. 1D). These data, there- fore, validated the reproducibility of CSD in these preparations. With respect to the properties of CSD before (control) and after AEA application (sample traces in Fig. 7AeB), there was no change in average amplitude (Fig. 7C) or duration (Fig. 7D). Likewise, WIN 55,212-2 did not affect CSD average amplitude (Fig. 7G) or duration (Fig. 7H; sample traces are shown in Fig. 7EeF). These observations imply a lack of effect of CB1R modulation on the cortex in vitro and/ or lack of GABAA receptor contribution to CSD. These data led us to investigate whether the GABAA receptor blocker gabazine (1 mM) might affect evoked LFP responses as well as CSD. Gabazine altered LFP properties, whereby the initial fast negative peak was followed by an afterdischarge consisting of high- frequency activity superimposed on a prolonged negative shift Fig. 5. Stimulation of cAMP synthesis by FSK enhances the amplitude of GABAA currents. A, Histograms show average peak amplitude values of currents evoked by GABA pulses (10 mM, 2 s) in control and after 10 min application of FSK (20 mM) to the same cells; n ¼ 10; *p ¼ 0.00001. On the right, example of GABA-elicited currents in control and after FSK. B, Histograms show average peak amplitude values of GABA currents in control and after 10 min application of FSK, followed by AEA (0.5 mM) applied to the same cells; n ¼ 10; **p < 0.00007, *p ¼ 0.002. On the right, representative traces of GABA-mediated currents in control and after application of FSK and AEA. Fig. 6. AEA and WIN 55,212-2 do not affect evoked local field potential responses in cortical brain slices. A, Schematic representation of coronal slice recordings with stim- ulation in layer 4 of the primary visual cortex and local field potential (LFP) recording in layer 2/3. B, Sample traces show evoked LFPs in recording solution (control) and after AEA (0.5 mM) or WIN 55,212-2 (5 mM) application. The amplitude of the first negative wave N1 was used to compare peak responses. C,D, Input output (IO) curves before (control) and after AEA application (C; n ¼ 9 slices, 8 mice) or before and after WIN 55,212-2 application (D; n ¼ 9 slices, 5 mice) shown as mean ± SEM for 10, 20, 50, 70, and 100% of the maximal evoked LFP response amplitude. Differences when comparing short IO-paradigms before and after AEA or WIN 55,212-2 application were not significant (AEA: p ¼ 0.89, two-way repeated measurements ANOVA; maximum response control: 1.11 ± 0.14 mV versus AEA: 1.16 ± 0.16 mV; p ¼ 0.12, paired t-test. WIN 55,212-2: p ¼ 0.58, two-way repeated measurements ANOVA; maximum response control: 1.09 ± 0.10 mV versus WIN 55,212-2: 1.10 ± 0.11 mV; p ¼ 0.90, paired t-test).(Fig. 8A). On the same slices, gabazine potentiated the N1 ampli- tude of the evoked LFP response (IO 3) in comparison with the control LFP from IO 2 (Fig. 8B). Nevertheless, in the same slice preparations, there was no effect of gabazine on CSD (sample traces in Fig. 8CeD) in terms of average amplitude (Fig. 8E) and duration (Fig. 8F). 4.Discussion The principal finding of this study is the demonstration that AEA significantly and reversibly depressed GABAA receptor function in TG neurons, while exerting no apparent effect on synaptic trans- mission or CSD in the visual cortex. Thus, our in vitro models indicate that changes in CB1R activation seem not involved in altering cortical function in the context of CSD. The inference is that CB1R may, therefore, not play a crucial role in the migraine aura phase, a notion to be tested in future with in vivo experiments. On the other hand, our data seem to indicate that a possible beneficial action by AEA in relieving trigeminal pain potentially include modulation of the TG local circuitry (Hayasaki et al., 2006) where GABA is supposed to control afferent nociceptive signaling (Ault and Hildebrand, 1994; Jang et al., 2017; Waite et al., 2000). 4.1.CB1Rs and GABAA receptors of TG neurons While the depolarizing action of GABA on sensory neurons has been known for decades (Nistri and Constanti, 1979), earlier studies indicated that GABA-mediated inhibition was observed only when the neuronal membrane potential was positive with respect to the Cl— reversal potential (Adams and Brown, 1975). Subsequentstudies have, however, shown that GABA and chemically relatedagonists can actually excite afferent sensory neurons (Ault andHildebrand, 1994; Jang et al., 2017; Waite et al., 2000) to provide a pro-nociceptive effect. In this context it was interesting to observe that AEA, WIN 55,212-2 and 2-AG could selectively and reversible decrease GABAA receptor responses via CB1R activation. Data ob- tained with WIN 55,212-2 (that is more metabolically stable than AEA and has less effect on GPR55) validate the link between CB1 receptor activity and inhibition of GABAA receptors. This phenom- enon was not due to a direct action of the endocannabinoids on neuronal excitability because there was no change in baseline current. Furthermore, the endocannabinoids targeted selectively GABAA receptors because there was no alteration in the ATP-gated P2X3 receptors, and no influence of TRPV1 receptors which can be activated by high doses of AEA (Roberts et al., 2002; Ross, 2003). It is noteworthy that, as amply reviewed by Buczynski and Parsons (2010), the agonist concentrations of AEA or 2-AG used in the present study are clearly larger than those reported by assaying these endogenous compounds in brain tissue, although endoge- nous levels vary within a broad range presumably reflecting fluc- tuations in ongoing neuronal activity.The inhibitory action by AEA was observed as a non-parallelshift of the GABA dose-response curve, suggesting a non- competitive type of pharmacological antagonism manifested as a nonlinear relation between CB1R effects and GABA responses. Since CB1R activation is transduced into inhibition of adenylyl cyclase activity and depression of cAMP synthesis (Howlett, 2002; Howlett et al., 2010), the observation that AEA strongly suppressed the facilitatory effect by forskolin on GABAA mediated responses is in accordance with this notion. Previous studies (Angelotti et al., 1993; Connelly et al., 2013; Kapur and Macdonald, 1996; Vithlani et al., 2011) have shown intracellular cAMP (via activation of protein ki- nase A and GABAA receptor phosphorylation) to induce large enhancement of GABAA receptor function. Fig. 7. AEA and WIN 55,212-2 do not affect cortical spreading depolarization (CSD) in cortical brain slices. A,B, Sample traces of CSD induced by 40 mM KCl for 60 s before (A; control) and after AEA application (B). Arrows indicate how data were measured. C, D, CSD characteristics amplitude (C) and duration (D) before (control) and after AEA were not altered by AEA (amplitude: control: 2.18 ± 0.29 mV versus AEA: 2.29 ± 0.29 mV; p ¼ 0.49, paired t-test. Duration; control: 11.65 ± 3.67 s versus AEA: 10.18 ± 3.71 s; p ¼ 0.91, Wilcoxon matched-pairs signed rank test; n ¼ 9 slices, 8 mice). Data points refer to individual experiments. E,F, Sample traces of CSD before (E; control) and after (F) WIN 55,212-2 application. G,H, CSD characteristics amplitude (G) and duration (H) before (control) and after WIN 55,212-2 were not altered by WIN 55,212-2 (amplitude: control: 3.89 ± 0.50 mV versus WIN 55,212-2: 3.64 ± 0.57 mV; p ¼ 0.32, paired t-test; n ¼ 9 slices, 5 mice. Duration: control: 11.79 ± 3.36 s versus WIN 55,212-2: 13.85 ± 4.22 s; p ¼ 0.21, paired t-test; n ¼ 7 slices, 5 mice). 4.2.CB1 and GABAA receptor modulation in cortical slices CB1R activation by AEA, or CB1 and CB2 receptor activation by WIN 55,212-2, did not have an effect on neuronal network trans- mission and CSD in the primary visual cortex as assessed with our brain slice preparations. Hence, this observation did not apparently accord with the large potentiation of synaptic responses (which turned into repeated discharges) by the GABA blocker gabazine. In the cerebral cortex, CB1R activation can down-regulate both glutamatergic and GABAergic neurotransmission probably via distinct intracellular mechanisms (Fortin and Levine, 2007; Kano et al., 2009). It seems likely that, in the present experiments, the lack of AEA and WIN 55,212-2 effect on synaptic responses may originate from a relatively sparse coupling of CB1Rs, adenylyl cyclase and GABAA receptor function. Although the GABAA receptor blocker gabazine potentiated LFPs and evoked epileptiform re- sponses, as expected as a result of disinhibition (Kernig et al., 2012; Ko€hling et al., 2003), strong excitation following GABA receptor block originates mainly from the primary motor cortex (Castro- Alamancos and Rigas, 2002). Indeed, in the absence of focal stim- uli by high K+ solution, human visual cortex slices fail to generate sustained discharges after GABA receptor block (Dreier et al., 2012). AEA, WIN 55,212-2, or gabazine did not change the properties of CSD in our experiments. Since CSD was apparently not modulated by gabazine-sensitive GABAergic mechanisms, our brain slice preparations did perhaps not contain enough GABAergic network elements to down-regulate the powerful CSD phenomenon. While GABAA receptor agonists were shown to have inhibitory effects on CSD in vivo (Hoffmann et al., 2014; Holland et al., 2010; Toriumi et al., 2016), the lack of an effect by gabazine on CSD in our study seems in line with reports showing that barbiturates, which enhance GABAA receptor-mediated transmission, do not affect CSD amplitude or frequency in vivo (Van Harreveld and Stamm, 1953; Kitahara et al., 2001; Kudo et al., 2008). Because during hippo- campal recording in vitro (Aiba and Shuttleworth, 2014) and in vivo (Aiba et al., 2012), gabazine was able to enhance CSD propagation while having no effect on CSD duration, it is likely that there are substantial regional differences in the role of GABAergic Fig. 8. Gabazine induces epileptiform local field potentials but does not affect cortical spreading depolarization. A, Sample trace of local field potential (LFP) recorded in the visual cortex following stimulation during gabazine (1 mM) application. Note that gabazine induced epileptiform activity. Inset shows high gain record of early response component. B, IO curves before (control) and after gabazine application shown as mean ± SEM for 10, 20, 50, 70, and 100% of the maximal response amplitude (p ¼ 0.01, two-way repeated measures ANOVA; maximum response control: 0.78 ± 0.12 mV versus gabazine: 0.97 ± 0.17 mV; *p ¼ 0.048, paired t-test; n ¼ 5 slices, 5 mice). C,D, Sample traces of cortical spreading depolarization (CSD) induced by 40 mM KCl for 60 s before (C; control) and after gabazine application (D). E,F, CSD amplitude (E) and duration (F) compared before (control) and after gabazine application show no effect of gabazine (amplitude: control: 2.51 ± 0.40 mV versus gabazine: 2.31 ± 0.37 mV; p ¼ 0.19, Wilcoxon signed rank test. Duration: control: 16.99 ± 2.75 s versus gabazine: 20.52 ± 2.55 s; p ¼ 0.44, Wilcoxon signed rank test; n ¼ 5 slices, 5 mice). Data points refer to individual experiments mechanisms in CSD. Although a previous study showed that CB1R activation in the rat cortex reduced CSD susceptibility (Kazemi et al., 2012), the discrepancy from our data may be due to the different procedure to prepare brain slices. In the cortex CB1R- mediated suppression of GABA release is differentially expressed between superficial and deep layers (Fortin and Levine, 2007) and is dependent on the network firing activity (Galarreta et al., 2008). Furthermore, dissimilar GABAergic projections across cortical re- gions may account for the lack of CSD effects in the coronal slice preparations that contain only a small extent of cortical architec- ture. Even though coronal slices including those containing the visual cortex are standard preparations to study CSD (Sawant- Pokam et al., 2017; Tang et al., 2014; Wendt et al., 2016), experi- ments with more intact connections should be made in future to systematically examine any region selectivity related to the action of CB1R and GABAA receptors. In chronic migraine patients, levels of AEA in the cerebrospinal fluid were decreased compared to non-migraine controls (Sarchielli et al., 2007), a finding that might point to a deficiency of the endocannabinoid system in migraineurs. In the light of the current study it is of relevance that phytocannabinoids from cannabis can bind to and activate CB1 and CB2 receptors (Hill et al., 2012). Although medical marijuana use seems to decrease the monthly frequency of headache in chronic migraine patients (Rhyne et al., 2016), this notion has not formally been investigated in a clinical trial. 5.Conclusions A former study indicated that a large component of the anal- gesic action of CB1Rs may arise from their modulation of peripheral dorsal root ganglion nociceptors (Agarwal et al., 2007). Our data suggest that TG sensory neurons are new candidates for expressing a peripheral nociceptive role by CB1Rs via modulation of GABAergic effects within the TG minimal network (Hayasaki et al., 2006). Despite the fact that CSD was apparently independent from GABAergic mechanisms or CB1R regulation, the recent demon- stration that CSD can be facilitated by trigeminal nociceptive inputs comprising GABAA-receptor mechanisms (Toriumi et al., 2016) provide a novel process to link CB1Rs peripherally and centrally. Future studies concerned with mechanisms of AEA in analgesia, for example in the context of migraine pain, should perhaps be directed to discrete subcortical nuclei involved in processing tri- geminal inputs, and be ideally validated with in vivo animal studies. Acknowledgments This work was supported by the European Union's (EU) Seventh Framework programme “EUROHEADPAIN” (nr. 602633) and EU Marie Curie IAPP Program “BRAINPATH” (nr 612360). The authors thank A. Heimel for expert help with the cortical slice paradigm. All authors approved the final manuscript. Conception and design of the work FE, AN; acquisition and analysis Monlunabant of data, FE, IL; interpretation of data, FE, IL, AM, ET, AN; wrote/revised the manuscript, FE, IL, AM, ET, AN. The authors declare no competing interests.