The neuronal endocannabinoid system is known to depress synaptic inputs retrogradely in an activity-dependent manner. This mechanism has been generally described for excitatory glutamatergic and inhibitory GABAergic synapses. Here, we report that neurones in the auditory brainstem of the Mongolian gerbil (Meriones unguiculatus) retrogradely regulate the strength of their inputs via the endocannabinoid system. By means of whole-cell patch-clamp recordings, we found that retrograde endocannabinoid signalling attenuates both glycinergic and glutamatergic post-synaptic currents in the same types of neurones. Accordingly, we detected the cannabinoid receptor 1 in excitatory and inhibitory pre-synapses as well as the endocannabinoid-synthesising enzymes (diacylglycerol lipase α/β, DAGLα/β) post-synaptically through immunohistochemical stainings. Our study was performed with animals aged 10–15 days, that is, in the time window around the onset of hearing. Therefore, we suggest that retrograde endocannabinoid signalling has a role in adapting inputs during the functionally important switch from spontaneously generated to sound-related signals.
We report retrograde endocannabinoid modulation of synaptic strength in auditory brainstem nuclei of the Mongolian gerbil. Utilising electrophysiological recordings and immunohistochemistry, we found endocannabinoid-dependent suppression of excitatory and inhibitory glycinergic currents in the same neurone types. We propose that retrograde endocannabinoid signalling contributes to adapting inputs to sound environment in the time period around the onset of functional hearing.
Endocannabinoids are lipid-derived molecules produced on demand, which activate cannabinoid receptors. The endocannabinoid system operates as a retrograde negative feedback system in many brain areas (Herkenham et al. 1990; Katona and Freund 2012). Upon stimulation of a post-synaptic cell, endocannabinoid synthesis is initiated via a Ca2+- dependent activation of endocannabinoid synthesising enzymes, which leads to the activation of pre-synaptic cannabinoid receptor 1 (CB1) (Liu et al. 2006; Alger and Kim 2011; Di Marzo 2011; Ueda et al. 2011; Castillo et al. 2012). CB1 receptors are G-protein coupled and – upon activation – attenuate Ca2+ influx into the pre-synaptic terminal and thus transmitter release by blocking vesicle fusion (Guo and Ikeda 2004). This feedback mechanism is called depolarization-induced suppression of inhibition (DSI) (Pitler and Alger 1992, 1994; Ohno-Shosaku et al. 2001; Wilson and Nicoll 2001) or depolarization-induced suppression of excitation (DSE) (Kreitzer and Regehr 2001), depending on whether the released neurotransmitter exerts an inhibitory or an excitatory action on the post-synaptic neurone (Ohno-Shosaku et al. 2012). DSI and DSE have been observed throughout the brain for GABAergic and glutamatergic synapses to provide a means for a cell to down-regulate its inputs in an activity-dependent manner. Retrograde glycinergic DSI has so far only been reported once in hypoglossal motor neurones of mice and rats (Mukhtarov et al. 2005). Direct modulation of glycinergic currents by binding of endocannabinoids to post-synaptic glycine receptors has been thoroughly investigated, leading to depression (Lozovaya et al. 2005) or potentiation of glycine receptor-mediated currents (Hejazi et al. 2006).
We studied a potential retrograde modulation of glycinergic neurotransmission by endocannabinoids in two nuclei of the superior olivary complex in the auditory brainstem, namely the medial and the lateral superior olive (MSO and LSO, respectively). These nuclei are involved in localization of sound sources in an acoustic environment and their neurones receive excitatory glutamatergic as well as inhibitory glycinergic projections (Kandler and Friauf 1995; Brand et al. 2002; Grothe 2003; Pecka et al. 2008; Grothe et al. 2010) (Fig. 1). Especially the strength of the glycinergic projections is crucial for allowing the computation of sound related inputs in the physiological range (Brand et al. 2002; Grothe 2003). Therefore, the existence of a potential DSI mechanism in these auditory neurones would be important in the context of adaptation to changes in the acoustic environment. In the LSO of young rats, which have not yet developed functional hearing, expression of the CB1 receptor at glycinergic medial nucleus of the trapezoid body (MNTB) terminals has already been shown by immunohistochemistry and a modulatory involvement of the endocannabinoid system in stimulus processing has been suggested (Chi and Kandler 2012). Electrophysiological recordings in other nuclei of the auditory brainstem revealed a strong impact of endocannabinoid signalling on excitatory neurotransmission (DSE) for the MNTB (Kushmerick et al. 2004) and the dorsal cochlear nucleus (Zhao et al. 2009; Sedlacek et al. 2011; Zhao and Tzounopoulos 2011).
To examine a possible influence of endocannabinoid signalling in MSO and LSO neurones, we studied the distribution of CB1 and of endocannabinoid-synthesising enzymes diacylglycerol lipases α/β (DAGLα/β) by immunohistochemical stainings. In addition, we performed electrophysiological recordings to investigate, whether post-synaptic currents are influenced by retrograde endocannabinoid signalling. We conducted our study in the Mongolian gerbil (Meriones unguiculatus), which represents a suitable model organism for auditory neurosciences, as it possesses an audible spectrum and an organization of the auditory brainstem similar to humans (Heffner and Heffner 1988; Muller 1990).
Materials and methods
All experiments complied with institutional guidelines and national and regional laws. Ethical clearance for animal experiments was granted by the ‘Regierung von OberbayerN'. Mongolian gerbils (Meriones unguiculatus; both sexes) at post-natal day (P) 10 and P13–15 (majority P13) were used. All animals were bred in the animal house of the Department of Biology II, Ludwig Maximilians University Munich.
Animals were anaesthetized using 100 mg/kg body weight metamizol (Novalgin®; Sanofi-Aventis Deutschland GmbH, Frankfurt, Germany), p.o., followed by 200 mg/kg body weight pentobarbital (Narcoren®; Merial GmbH, Hallbergmoos, Germany), i.p. After the animals had reached a deep anaesthetic stage, they were transcardially perfused at a flow rate of 4 mL/min with Ringer solution supplemented with 0.1% heparin (Meditech Vertriebs GmbH, Parchim, Germany) for 10 min followed by 4% paraformaldehyde (PFA) for 20 min. The brains were post-fixed overnight in 4% PFA at 4°C. Using a Leica VT1200S vibratome (Leica Microsystems GmbH, Wetzlar, Germany), 50 μm sections of the auditory brainstem were collected and washed 4 times in 0.1 M phosphate buffered saline (PBS) for 5 min each. Then, unspecific binding sites were blocked using a blocking solution (0.3% Triton X-100, 0.1% saponine and 1% bovine serum albumin) for 1 hour at 22°C on a shaker. Sections were incubated in the primary antibody mix diluted in blocking solution over night at 4°C on a shaker. Primary antisera used were: chicken anti-microtubule associated protein 2 [MAP2, 1 : 1000; Neuromics (Edina, MN, USA), CH22103], rabbit anti-CB1 [1 : 300; Alomone Labs (Jerusalem, Israel), ACR-001], mouse anti-vesicular glutamate transporter 1 [VGluT1, 1 : 500; EMD Millipore Cooperation (Billerica, MA, USA), MAB5502], guinea pig anti-glycine transporter 2 (GlyT2, 1 : 1000, Millipore, AB1773) as well as rabbit anti-DGLα (1 : 500) and rabbit anti-DGLβ (1 : 500, both gift from Dr. Ken Mackie, Indiana University). Thereafter, sections were washed four times in 0.1 M PBS for 5 min each and incubated with secondary antibodies for 3–4 h at 22°C on a shaker. Secondary antisera used were: donkey anti-rabbit DyLight 649 [1 : 300; Dianova GmbH (Hamburg, Germany), 711-496-152], donkey anti-rabbit DyLight 488 (1 : 100, Dianova, 711-486-152), goat anti-mouse Cy5 (1 : 100, Dianova, 115-175-166), donkey anti-guinea pig Cy3 (1 : 300, Millipore, AP193C), goat anti-chick Alexa488 [1 : 400; Molecular probes (Life Technologies GmbH, Darmstadt, Germany), A11039] and donkey anti-chick Cy3 (1 : 300, Dianova, 703-166-155). Afterwards, the sections were washed four times in 0.1 M PBS for 10 min each. Finally, the slices were mounted with Vectashield supplemented with 4′,6-diamidino-2-phenylindole (DAPI) (H-1200; Vector Laboratories Inc., Burlingame, CA, USA).
To visualize immunohistochemical stainings, confocal optical sections were acquired with a Leica 6000CS SP5 confocal laser-scanning microscope (Leica Microsystems, Mannheim, Germany) equipped with a Plan 63x/NA1.32 oil immersion objective. Fluorochromes were visualized by using a UV laser with an excitation wavelength of 405 nm (emission 420–475 nm for DAPI), an argon laser with an excitation wavelength of 488 nm (emission 494–555 nm for Alexa 488/DyLight 488), a DPSS laser with a laser line of 561 nm (emission 565–606 nm for Cy3) and a helium-neon laser with an excitation wavelength of 633 nm (emission 640–740 nm for DyLight 649). For each optical section the images were collected sequentially for the different fluorochromes. Stacks were obtained with axial distances of 300 nm – the image size was 512 × 512 pixels. To obtain an improved signal-to-noise ratio each section image was averaged from four successive line scans. The images shown in the figures of this publication are single images from the stacks and thus have an axial distance of 300 nm.
Animals were anaesthetized with isoflurane, followed by rapid decapitation. The brains were removed and placed in ice-cold dissecting solution containing (in mM, all from Sigma-Aldrich Chemie GmbH, München, Germany): 50 sucrose, 25 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 3 MgCl2, 0.1 CaCl2, 25 glucose, 0.4 ascorbic acid, 3 myo-inositol and 2 Na-pyruvate. The pH amounted to 7.4 when bubbled with 95% O2 and 5% CO2. Horizontal brainstem sections of 200 μm thickness were cut using a VT1000S vibratome (Leica). Slices were incubated in extracellular recording solution continuously bubbled with 95% O2 and 5% CO2 for 45 min at 36°C and then stored at 22°C until use. The extracellular recording solution contained (in mM, all from Sigma-Aldrich): 125 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 1 MgCl2, 25 glucose, 0.4 ascorbic acid, 3 myo-inositol and 2 Na-pyruvate.
For recording, slices were transferred into a chamber mounted on a microscope (Olympus BX51WI, Olympus Europa Holding GmbH, Hamburg, Germany) and continuously perfused with fresh extracellular recording solution at room temperature (≈ 22°C). Voltage-clamp whole-cell recordings were performed with an EPC9 amplifier (HEKA Elektronik Dr. Schulze GmbH, Lambrecht/Pfalz, Germany) on visually identified MSO neurones. Cells were clamped at −60 mV unless stated otherwise. Series resistance was compensated to a residual of 3–4 MΩ. Glass pipettes used for recording had a resistance between 2.0 and 4.5 MΩ. The intracellular solution contained in mM (all from Sigma-Aldrich): 105 Cs-gluconate, 26.7 CsCl, 10 HEPES, 20 TEA-Cl, 3.3 MgCl2, 2 Na2ATP, 0.2 NaGTP, 3 Na-phosphocreatine and 1 EGTA. The solution was adjusted to pH 7.2 with CsOH. To investigate whether retrograde endocannabinoid signalling is dependent on a rise in post-synaptic Ca2+ levels, we carried out some recordings in the presence of 30 mM 1,2-bis(o-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid (BAPTA) (Sigma-Aldrich) in the intracellular recording solution. Glycinergic post-synaptic currents for DSI recordings were pharmacologically isolated by adding 50 μM D-APV (BioTrend Chemikalien GmbH, Köln, Germany), 20 μM DNQX (BioTrend) and 10 μM gabazine (Sigma-Aldrich) to the extracellular recording solution. Excitatory post-synaptic currents for DSE recordings were pharmacologically isolated by supplementing the extracellular solution with 0.5 μM strychnine (Sigma-Aldrich) and 10 μM gabazine. To investigate the effects of the endocannabinoid system on evoked post-synaptic currents the following CB1 agonists and antagonists were used: rimonabant (2 μM, Cayman Chemical Company, Ann Arbor, MI, USA), WIN 55,212-2 mesylate (1 μM, BioTrend) and anandamide (1 μM, Cayman Chemical). Recordings with tetrahydrolipstatin (THL) to test an involvement of 2-arachidonylglycerol (2-AG) in endocannabinoid signalling, were carried out in brain slices that were incubated in 10 μM THL (Cayman Chemicals) in extracellular recording solution for 1.5 h prior to the recording, which was also carried out in the presence of 10 μM THL.
Glycinergic currents were evoked by placing a bipolar tungsten electrode at the fibre bundle originating in the ipsilateral MNTB. Excitatory currents for DSE recordings were evoked by placing the stimulation electrode dorso-lateral to the MSO, to stimulate fibres originating in the ipsilateral cochlear nucleus. Fibres were stimulated with a rate of 0.33 Hz and pulse duration of 0.2 ms. The stimulus strength was set generally to 2–3 V for inhibitory currents and to 3–9 V for excitatory currents. The reason for this difference is that inhibitory fibres originating from the MNTB are easy to target as they run in a thick fibre bundle. Excitatory fibres cannot be located visually under the microscope and thus we sometimes had to apply higher stimulation strengths to recruit fibres that run deeper within the tissue. Before starting a DSI/DSE experiment, we ensured that the baseline current amplitude was stable for at least 21 stimulations (≥ 1 min). For conducting DSI/DSE experiments the cell was depolarized for 5 s to 0 mV, representing the standard protocol for DSI induction (Wilson and Nicoll 2001), after seven stimulations, which were used to calculate the baseline current for normalization. For DSI recordings also other depolarization protocols were tested to evoke endocannabinoid signalling: Either depolarization to 0 mV for 0.5 s or the StimAP stimulus. The StimAP stimulus is a voltage-clamp stimulus given instead of the 5 s constant depolarization to 0 mV that was generally used to elicit DSI. The stimulus consisted of a square wave pulse alternating between −40 and +40 mV at a frequency of 200 Hz for a duration of 5 s. This stimulus mimics the membrane response during a high frequent action potential train. For quantification all currents were normalized to the average baseline current before the depolarization pulse. During pharmacological experiments values for each time point were calculated by averaging the amplitudes of the three evoked post-synaptic currents (ePSCs) preceding depolarization (before depol.), the amplitudes of the first three ePSCs following depolarization (after depol.) or the amplitudes of the 12th to 14th ePSC after depolarization (recovery). For paired-pulse ratio (PPR) recordings, a pulse length of 0.2 ms and an inter-pulse interval of 9.8 ms were used.
Statistical analyses were performed using Prism 5.0 (Graphpad Software Inc., La Jolla, CA, USA). Average values are given as mean ± SEM. We compared PPR data using a paired t-test. All other data were tested for deviation from 1.0 with a one-sample t-test, as all data were normalized. We assigned significance for p < 0.1. The respective p-values are given in the Figure legends.
The endocannabinoid system is expressed in neurones of the MSO and LSO
Using immunohistochemistry, we detected a strong neuronal CB1 expression in the MSO and LSO at both P10 (not shown) and P13 (Figs 2 and 3). CB1 was partially co-localized with VGluT1 (Figs 2a and 3a; arrows mark sites of co-localization), which is a pre-synaptic marker for excitatory glutamatergic synapses, as well as with GlyT2 (Figs 2b and 3b; arrows mark sites of co-localization), a pre-synaptic marker for inhibitory glycinergic synapses. Considering endocannabinoid production, we investigated the expression of DAGLα/β with immunohistochemical labelling. We observed that DAGLα/β were expressed somatically by most neurones at both P10 (not shown) and P13 in the MSO (Fig. 2c) and LSO (Fig. 3c), as indicated by a co-localization with the neuronal marker MAP2. Lower levels of DAGLα/β were also detected in MAP2-negative cells, which are presumably glial cells (Fig. 2c).
We showed that inhibitory glycinergic currents, evoked by stimulation of axons originating from the MNTB, were reduced in amplitude upon depolarization of the post-synaptic cell, in both MSO (P10, Fig. 4a; P13 Fig. 4b) and LSO (P13, Fig. 8a; P10 not shown) for both age groups. The amplitude and time course of DSI was similar in MSO and LSO for both P10 and P13 (Fig. 4c). Generally, a depolarization of 5 s to 0 mV resulted in about 45% current amplitude reduction and the baseline amplitude was reached again within 20–30 s after depolarization. As the standard protocol with a depolarization pulse of 0 mV for 5 s does not represent a physiological input in these neurones, we tried to elicit DSI using more relevant stimuli in MSO neurones at P13 (Fig. 4d). A constant depolarization to 0 mV for 0.5 s did not result in any alteration of evoked current amplitude. However, a square wave function at 200 Hz with a maximum of +40 mV for 5 s (StimAP), which simulates a 200 Hz action potential train and thus represents a physiologically more relevant stimulus than the constant depolarization generally used in the literature (Wilson and Nicoll 2001), elicited DSI in a similar way. As the CB1 antagonist rimonabant completely blocked DSI of glycinergic currents (Fig. 6a), we conclude that DSI is retrogradely mediated by endocannabinoids. Endocannabinoid-dependent DSI has been previously shown to depend on raised post-synaptic Ca2+ levels (Ohno-Shosaku et al. 2001). Thus, we performed recordings in MSO neurones at P13 with 30 mM BAPTA in the intracellular solution. When intracellular Ca2+ was chelated by BAPTA, glycinergic DSI could not be elicited with 5 s depolarization to 0 mV (Fig. 7a).
To provide further evidence that the observed DSI is caused by endocannabinoid signalling, a series of pharmacological experiments was conducted. First, the paired-pulse ratio was shown to increase upon the wash-in of the synthetic CB1 receptor agonist WIN 55,212-2, which indicates a pre-synaptic mechanism in the MSO (Fig. 4e) and LSO (data not shown, n = 3, p = 0.0199; paired t-test). In addition, we checked whether the paired-pulse ratio was also affected by post-synaptic depolarization. Depolarization to 0 mV for 5 s significantly increased the paired-pulse ratio (n = 8, p = 0.0008, paired t-test), supporting that our depolarization stimulus indeed is adequate to elicit endocannabinoid signalling (Fig. 4f). Wash-in of the endogenous CB1 receptor agonist anandamide decreased evoked inhibitory post-synaptic current (eIPSC) to about 60%, whereas subsequent wash-in of the CB1 receptor inverse agonist rimonabant rescued eIPSC to about 80% of baseline and eliminated the effect of depolarization on current amplitude (Figs 4g and 8b). Application of the CB1 agonist WIN 55,212-2 instead of anandamide led to a comparable suppression of inhibitory post-synaptic currents, which was, however, less in amplitude (72% of the original value) when compared to anandamide application. As with anandamide, rimonabant was unable to fully recover the baseline current (87% of the original value), but blocked DSI in favor of a CB-mediated endocannabinoid action (Figure S1). Rimonabant application alone completely abolished the effect of depolarization and inhibitory post-synaptic currents were no longer reduced by depolarising the cell for 5 s to 0 mV (Fig. 6a). This block of the effect by rimonabant shows that the effect observed is mediated by retrograde cannabinoid signalling. Furthermore, we were interested, whether 2-AG contributes to endocannabinoid signalling, as we found high levels of DAGLα/β immunohistochemically (Figs 2c and 3c). To this extent we conducted electrophysiological recordings in MSO P13 neurones in the presence of THL, which is an inhibitor of the DAGLα/β enzymes and should thus block DSI, if 2-AG was necessary for DSI induction. Our recordings show that DSI cannot be elicited in the presence of 10 μM THL (Fig. 7c), suggesting that 2-AG is the main endocannabinoid mediating DSI in the MSO at P13.
Excitatory glutamatergic currents, evoked by stimulation of axons originating from the ipsilateral cochlear nucleus, were reduced upon a depolarization to 0 mV for 5 s in both MSO (Fig. 5a and b) and LSO neurones (Fig. 8c). The current amplitude at P13 was reduced by around 30% (MSO) and 38% (LSO), respectively. The time course of DSE was similar in neurones of both nuclei (Fig. 5c). DSE could also be evoked – to a lesser extent – by the StimAP stimulus, which caused a reduction in current amplitude to about 78% of the original value in MSO P13 cells (Fig. 5d). DSE is also mediated by retrograde endocannabinoid signalling, as it cannot be evoked in the presence of the CB1 antagonist rimonabant (Fig. 6b). To test, whether DSE is also dependent on intracellular Ca2+ increase, we conducted experiments where the intracellular recording solution was supplemented with 30 mM BAPTA. When Ca2+ was chelated by BAPTA, DSE could not be elicited (Fig. 7b), which strongly indicates a Ca2+ dependency of our effect, as would be expected from retrograde endocannabinoid-mediated DSI.
The paired-pulse ratio of evoked glutamatergic currents was increased upon administration of the CB1 agonist WIN 55,212-2 (Fig. 5e) or by constant depolarization for 5 s to 0 mV (Fig. 5f), which indicates a pre-synaptic mode of action.
To test the dependence of DSE on retrograde endocannabinoid signalling, pharmacological experiments were carried out (Figs 5g and 8d). Wash-in of anandamide depressed evoked excitatory current amplitude by about 40% and subsequent treatment of the neurones with rimonabant recovered the full baseline current amplitude after 15 min. Application of rimonabant alone completely blocked depolarization-induced suppression of glutamatergic currents, strongly indicating that the effect observed is retrogradely mediated by endocannabinoids (Fig. 6b). Furthermore, we investigated, whether 2-AG also mediates DSE, by trying to evoke DSE in the presence of 10 μM of the DAGLα/β inhibitor THL. Our results show that we were unable to elicit DSE, when 2-AG was not produced (Fig. 7d), indicating that 2-AG is necessary for DSE induction.
The data presented here show that a retrograde endocannabinoid system is functionally active for both inhibitory glycinergic and excitatory glutamatergic synapses targeting the same neurones in the developing MSO and LSO (P10-15). We found the presence of pre-synaptic CB1 receptors as well as somatic expression of diacylglycerol lipase α/β, enzymes critically involved in the synthesis of one of the most important endocannabinoids (i.e. 2-arachidonylglycerol, 2-AG) (Best and Regehr 2010). The pre-synaptic CB1 receptors were partially co-localized with VGluT1 (Figs 2a and 3a; arrows mark co-localization sites), a marker for excitatory glutamatergic synapses, as well as with GlyT2 (Figs 2b and 3b; arrows mark co-localization sites), a marker for glycinergic terminals indicative of modulation of glycinergic and glutamatergic neurotransmission by retrograde endocannabinoid signalling. In accordance with our findings in the LSO, CB1 expression on glycinergic terminals originating from the MNTB has been reported in developing rats (Chi and Kandler 2012). However, our immunohistochemical experiments suggest an additional extrasynaptic localization of CB1 receptors (Figs 2 and 3) and their involvement in volume transmission by endocannabinoids (Fuxe et al. 2012).
To determine, whether the endocannabinoid system components detected by immunohistochemistry also play a functional role in neuronal signalling, we conducted a series of electrophysiological recordings. Firstly, we observed an endocannabinoid-dependent retrograde depression of evoked glycinergic currents in both MSO and LSO at P10 and P13 (Figs 4, 8a and b), which was blocked by the CB1 antagonist rimonabant (Fig. 6a) and required the rise of intracellular Ca2+ levels induced by depolarization (Fig. 7a), as it is generally described for DSI (Ohno-Shosaku et al. 2001). Until now DSI has mostly been described for GABAergic synapses (Ohno-Shosaku et al. 2001; Wilson and Nicoll 2001) with one exception; in rodent hypoglossal motoneurones glycinergic DSI has been reported as well (Mukhtarov et al. 2005). Both amplitude and time course of our glycinergic DSI are comparable to those reported for GABAergic DSI (Ohno-Shosaku et al. 2001; Wilson and Nicoll 2001). This is plausible as the processes of both DSI and DSE affect neurotransmitter release by interacting with the same players (CB1, G proteins, voltage-dependent Ca2+ channels, etc.). Neurotransmitter specific processes (e.g. at the post-synapse) are not affected and, thus, DSI and DSE (i.e. their amplitude, duration and time course) are supposed to be independent of the type of neurotransmitter (Katona and Freund 2012). However, when we compared glycinergic DSI in the MSO to glycinergic DSI previously reported in hypoglossal motor neurones of rat and mouse (Mukhtarov et al. 2005), glycinergic DSI in our study was smaller in amplitude and recovered faster from depression (Fig. 4b, c and d). Reasons for these observed discrepancies could be differences in species, animal age or stimulation procedure.
We could show a retrograde action of the endocannabinoid system on the release of glycinergic neurotransmitter vesicles by the complete prevention of DSI in the presence of rimonabant (Fig. 6a). Furthermore, we were interested whether a direct modulation of the glycine receptor by endocannabinoids might also be involved (Lozovaya et al. 2005). Unlike in the case of anandamide-evoked depression of evoked excitatory post-synaptic current (eEPSCs) – rimonabant could not fully revert the anandamide-evoked depression of glycinergic currents in the MSO (Figs 4g and 5g). In contrast to anandamide, the synthetic CB1 agonist WIN 55,212-2 does not directly modulate glycine receptors (Lozovaya et al. 2005), but has the disadvantage of directly modulating voltage-gated Ca2+ currents (Nemeth et al. 2008). However, recordings with WIN 55,212-2 yielded similar results as with anandamide in the MSO (Figure S1), that is, subsequent rimonabant wash-in could not fully recover the baseline amplitude of the currents. Therefore, we conclude that the inability of rimonabant to recover the baseline current amplitude after CB1 agonist administration is probably not caused by a CB1-independent endocannabinoid modulation of glycine receptors. On the basis of the complete blockage of DSI and DSE in the presence of rimonabant, we conclude that endocannabinoid signalling operates mainly retrogradely at this synapse.
Secondly, we also detected a retrograde modulation of glutamatergic synaptic transmission with a time course of recovery similar to glycinergic DSI found in this study and previously reported data on glutamatergic DSE (Kreitzer and Regehr 2001) (Fig. 5b and c). DSE induction was dependent on a rise in intracellular Ca2+ levels (Fig. 7b) and could be blocked by rimonabant (Fig. 6b), which proves that it is mediated by retrograde endocannabinoid signalling via pre-synaptically located CB1 receptors. A depression of glutamatergic currents could further be achieved by wash in of the endogenous CB1 agonist anandamide (Fig. 5g). In addition, we could also show electrophysiologically that the activity of the DAGLα/β enzymes is necessary for the induction of DSI and DSE (Fig. 7c and d), as suggested by our immunohistochemical stainings (Figs 2c and 3c).
Hearing onset in Mongolian gerbils occurs at P12-13 (Woolf and Ryan 1984). Before hearing onset, neurones in the auditory brainstem are stimulated by spontaneous activity, which is generated by ATP-signalling in cochlear hair cells (Tritsch et al. 2007). This spontaneous activity is required for the correct development of auditory circuits as well as for the fine-tuning of tonotopic maps (Kandler 2004; Leake et al. 2006; Leao et al. 2006). As retrograde endocannabinoid signalling acts as a negative feedback loop, inhibiting the inputs received by the neuron, it is conceivable that the endocannabinoid system is involved in balancing the spontaneous activity and thereby contributes to the functional development of the auditory system before hearing onset (Chi and Kandler 2012). After hearing onset, the endocannabinoid system could help adjusting the stimulus strength of physiological inputs. The persisting endocannabinoid system might help to attenuate these new stimuli until further mechanisms for adaptation to the acoustic environment have been established and, thereby, protect MSO and LSO neurones from excitotoxicity. As endocannabinoid signalling acts because of the hydrophobicity of the compounds very locally (Brown et al. 2003), it is conceivable that a selective modulation of certain inputs enables accurate computation of sound source localization in ambient noise.
Interestingly, we found that endocannabinoids modulate glycinergic and glutamatergic neurotransmission to a similar extent. As the balance between these two inputs is crucial for the physiological function of sound source localization achieved by MSO and LSO neurones, retrograde endocannabinoid signalling might exert its adjusting function without affecting the balance between excitatory and inhibitory inputs. Like in other neuronal systems, depolarization by glutamatergic stimulation might – via Ca2+ influx through post-synaptic voltage-sensitive Ca2+ channels – stimulate endocannabinoid synthesis (Ueda et al. 2011) (Fig. 7a and b). In addition, stimulation of post-synaptic metabotropic receptors might also enhance endocannabinoid synthesis (Katona and Freund 2012). With both mechanisms, excitatory synaptic activity would be a necessary prerequisite not only for glutamatergic DSE but also for glycinergic DSI in these neurones. This represents a reasonable assumption, as both excitatory and inhibitory inputs have to be activated in parallel in order that MSO and LSO neurones can exert their main function in the context of localization of sound sources (Grothe et al. 2010). On the basis of the time window during which the endocannabinoid actions have been observed, we suggest a role in development of functional hearing.
We express our sincere thanks to Ken Mackie, Indiana University, who gave us DAGLα and DAGLβ antibodies as a present. We are grateful for inspiring discussions with Felix Felmy. The authors thank the Graduate School of Systemic Neurosciences (GSN-LMU) and the Graduate School 1373 (TU Munich, DFG) for providing a grant to BT. This work was largely funded by the DFG (Collaborative Research Center 870).
Conflict of interest
The authors declare no competing financial interests.