Endocannabinoid signalling triggered by NMDA receptor-mediated calcium entry into rat hippocampal neurons

Abstract

Endocannabinoids are released from neurons in activity-dependent manners, act retrogradely on presynaptic CB₁ cannabinoid receptors, and induce short-term or long-term suppression of transmitter release. The endocannabinoid release is triggered by postsynaptic activation of voltage-gated Ca²⁺ channels and/or G_q-coupled receptors such as group I metabotropic glutamate receptors (I-mGluRs) and M₁/M₃ muscarinic receptors. However, the roles of NMDA receptors, which provide another pathway for Ca²⁺ entry into neurons, in endocannabinoid signalling have been poorly understood. In the present study, we investigated the possible contribution of NMDA receptors in endocannabinoid production by recording IPSCs in cultured hippocampal neurons. Under the conditions minimizing the activation of voltage-gated Ca²⁺ channels, local application of NMDA (200 μm) transiently suppressed cannabinoid-sensitive IPSCs, but not cannabinoid-insensitive IPSCs. This NMDA-induced suppression was abolished by blocking NMDA receptors, CB₁ receptors and diacylglycerol lipase, but not by inhibiting voltage-gated Ca²⁺ channels. When the postsynaptic neuron was dialysed with 30 mm BAPTA, the NMDA-induced suppression was reduced significantly. A lower dose of NMDA (20 μm) exerted little effect when applied alone, but markedly enhanced the cannabinoid-dependent suppression driven by muscarinic receptors or I-mGluRs. These data clearly indicate that the activation of NMDA receptors facilitates the endocannabinoid release either alone or in concert with the G_q-coupled receptors.

Activity-dependent change in synaptic strength is thought to be a fundamental process that underlies higher brain functions including learning and memory. NMDA-type glutamate receptors are known to play crucial roles in induction of synaptic plasticity in various regions of the CNS (Malenka & Nicoll, 1999). Recent studies have revealed that the endocannabinoid system also contributes to activity-dependent synaptic modulation in the CNS (Chevaleyre et al. 2006). Endocannabinoids are bioactive lipids and mediate a retrograde signal in neural tissues (Alger, 2002). They are released from postsynaptic neurons and activate presynaptic CB₁ cannabinoid receptors, thereby inducing short-term or long-term suppression of transmitter release. Endocannabinoid-mediated synaptic plasticity has been identified in various brain regions including the hippocampus, cerebellum, amygdala, basal ganglia and neocortex (Chevaleyre et al. 2006; Hashimotodani et al. 2007b).

As the mechanisms of endocannabinoid production, two main intracellular pathways have been reported (Ohno-Shosaku et al. 2005). One is a phospholipase Cβ (PLCβ)-dependent pathway, which is driven by G_q-coupled receptors including group I metabotropic glutamate receptors (I-mGluRs) (Maejima et al. 2001; Varma et al. 2001; Ohno-Shosaku et al. 2002) and M₁/M₃ muscarinic receptors (Ohno-Shosaku et al. 2003; Fukudome et al. 2004). These receptors stimulate PLCβ, which yields diacylglycerol (DAG) and inositol-1,4,5-trisphosphate. DAG is converted to the major endocannabinoid 2-arachidonolyglycerol (2-AG) by DAG lipase (Piomelli, 2003). 2-AG is then released from neurons and retrogradely activates presynaptic CB₁ receptors. The receptor-driven endocannabinoid release is enhanced by a small Ca²⁺ elevation to submicromolar levels (Maejima et al. 2005) that is attributable to Ca²⁺-dependent enhancement of PLCβ activity (Hashimotodani et al. 2005). The other pathway is independent of PLCβ (Hashimotodani et al. 2005; Maejima et al. 2005) and is driven by a large Ca²⁺ elevation to micromolar levels (Brenowitz & Regehr, 2003). This pathway is responsible for depolarization-induced suppression of inhibition (DSI) (Kreitzer & Regehr, 2001a; Ohno-Shosaku et al. 2001; Wilson & Nicoll, 2001) or excitation (DSE) (Kreitzer & Regehr, 2001b). Postsynaptic depolarization activates voltage-gated Ca²⁺ channels and causes a large Ca²⁺ elevation, thereby inducing endocannabinoid release. The released endocannabinoids then suppress the inhibitory or excitatory transmission (DSI/DSE). It has been suggested that DSI/DSE is also mediated by 2-AG (Kim & Alger, 2004; Straiker & Mackie, 2005; Szabo et al. 2006; Hashimotodani et al. 2007c; Uchigashima et al. 2007), although the molecular identities of enzymes involved in this pathway are still unclear.

Thus, it is well established that an intracellular Ca²⁺ signal is crucial for induction of endocannabinoid release through both the PLCβ-dependent and PLCβ-independent pathways. In most of the previous studies, Ca²⁺ elevation for triggering endocannabinoid release was caused by activation of voltage-gated Ca²⁺ channels. Under physiological conditions, however, Ca²⁺ elevation can be induced through multiple pathways. Although NMDA receptors constitute a major calcium entry pathway into central neurons, it is not well understood how NMDA receptors are involved in endocannabinoid signalling. In the present study, we used cannabinoid-sensitive IPSCs in cultured hippocampal neurons, and examined whether NMDA receptors can contribute to the endocannabinoid release through the PLCβ-dependent and PLCβ-independent pathways. Our data demonstrate that NMDA receptors provide a novel Ca²⁺ entry pathway for endocannabinoid production in hippocampal neurons.

Methods

Electrophysiology

All experiments were performed according to the guidelines laid down by the animal welfare committees of Kanazawa University and Osaka University. Cultured hippocampal neurons were prepared from newborn rats as previously described (Ohno-Shosaku et al. 2001). Briefly, rats were anaesthetized with ether and decapitated. Hippocampi were rapidly removed and cells were mechanically dissociated. The cultures were kept at 36°C in 5% CO₂ for 10–15 days before use. The neurons were whole-cell clamped with patch pipettes, and the current responses were recorded with a patch-clamp amplifier (EPC9/3 or EPC10/2, HEKA, Lambrecht/Pfalz, Germany). For recording IPSCs, the presynaptic neuron was stimulated by applying voltage pulses (from −80 mV to 0 mV, 2 ms) at 0.5 Hz, and the evoked IPSCs were recorded from the postsynaptic neuron at −80 mV, unless otherwise indicated. The magnitude of IPSC suppression was calculated from mean amplitudes of 10–12 consecutive IPSCs before treatment and seven consecutive IPSCs obtained between 4 s and 18 s after the end of local puff application. In some cases, electrical stimulation failed to excite action potentials in the presynaptic neurons for a short period of time after the end of puff application of a Na⁺-free solution. Such failure traces were not included in the calculation. Voltage-gated Ca²⁺ currents were evoked by applying depolarizing voltage pulses (from −80 mV to −10 mV, 50 ms), and the amplitude of inward currents was measured. The voltage-gated Ca²⁺ current was assumed to be blocked completely by 0.1 mm CdCl₂. All experiments were performed at room temperature (24–26°C).

Solutions

A standard external solution contained (mm): 140 NaCl, 2.5 KCl, 1 MgCl₂, 2 CaCl₂, 10 Hepes, 10 glucose and 0.01 CNQX (pH 7.3, adjusted with NaOH). The bath was continuously perfused with this solution at a flow rate of 1–3 ml min⁻¹. WIN55,212-2, AM281, 3-((R)-2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP), and voltage-gated Ca²⁺ channel blockers were bath-applied. For induction of endocannabinoid release, NMDA, oxotremorine M (oxo-M) and 3,5-dihydroxyphenylglycine (DHPG) were locally applied via a 250 μm wide puff pipette located near the neurons, by using a perfusion valve controller (VC-6M, Warner Instruments, Hamden, CT, USA). The solution used for this local application contained (mm): 145 N-methyl-d-glucamine (NMDG), 135 HCl, 2.5 KCl, 3 CaCl₂, 10 Hepes and 0.01 CNQX (pH 7.3, adjusted with HCl), unless otherwise indicated. In the experiments shown in Fig. 5, NMDA was locally applied by using a solution that contained (mm): 145 NMDG, 135 HCl, 2.5 KCl, 2 CaCl₂, 1 MgCl₂, 10 Hepes and 0.01 CNQX (pH 7.3, adjusted with HCl). For measurements of Ca²⁺ currents, the external solution contained (mm): 130 NaCl, 10 TEA-Cl, 2.5 KCl, 1 MgCl₂, 2 CaCl₂, 10 Hepes, 10 glucose and 1 4-aminopyridine (pH 7.3, adjusted with NaOH). Patch pipettes were filled with a solution containing (mm): 130 potassium gluconate, 15 KCl, 6 MgCl₂, 0.2 EGTA, 10 Hepes, 5 Na₂ATP and 0.2 Na₂GTP (pH 7.3 adjusted with KOH). In some experiments (Fig. 1), the pipettes were filled with a Cs⁺-based solution, which contained (mm): 120 CsCl, 5 MgCl₂, 1 CaCl₂, 10 EGTA, 10 Hepes, 30 CsOH, 5 Na₂ATP and 0.2 Na₂GTP (pH 7.3 adjusted with CsOH). In the experiments shown in Fig. 5, solutions containing 5 mm EGTA or 30 mm BAPTA were used. The EGTA solutions contained (mm): 120 potassium gluconate, 15 KCl, 6 MgCl₂, 5 EGTA, 10 Hepes, 5 Na₂ATP and 0.2 Na₂GTP (pH 7.3 adjusted with KOH). The BAPTA solution contained (mm): 72 potassium gluconate, 15 KCl, 6 MgCl₂, 3 CaCl₂, 30 BAPTA, 10 Hepes and 5 Na₂ATP (pH 7.3 adjusted with KOH). The electrode resistance ranged from 2 to 5 MΩ.

Figure 5. Effects of postsynaptic BAPTA injection on the NMDA-induced suppression of IPSCs.

A and B, representative experiment showing the effects of NMDA application (200 μm, 5 s) in the presence of Mg²⁺. IPSCs were recorded at the holding potential of −80 mV (left) and 0 mV (right) in the postsynaptic neuron dialysed with the pipette solution containing 5 mm EGTA (A) or 30 mm BAPTA (B). C, averaged data for the NMDA-induced suppression of IPSCs at these four recording conditions.

Figure 1. NMDA-induced whole-cell currents under three different conditions.

A, examples of NMDA-induced currents recorded at +40 mV (upper traces) and −80 mV (lower traces). NMDA (200 μm) was locally applied for 2 s. NMDA was dissolved in a standard extracellular solution (a), an Mg²⁺-free standard solution (b) or an Mg²⁺–Na⁺-free solution (c). The patch pipettes were filled with a Cs⁺-based solution. B, averaged data for relative amplitudes of NMDA-induced currents at −80 mV compared to those recorded at +40 mV in the three experimental conditions. The number of experiments is indicated in parentheses for this and subsequent figures.

Drugs

WIN55,212-2, AM281, CPP, DHPG and oxo-M, were purchased from Tocris Cookson (Bristol, UK). SNX-482, ω-agatoxin IVA (AgTX), and ω-conotoxin GVIA (CgTX) were from Peptide Institute (Osaka, Japan). CNQX, NMDA and tetrahydrolipstatin (THL) were from Sigma-Aldrich (St Louis, MO, USA). Nifedipine was from Nacalai Tesque (Kyoto, Japan).

Statistics

Averaged data from different experiments are presented as means ±s.e.m. Statistical significance was assessed by Student's t test for paired or unpaired data. One, two and three asterisks indicate P < 0.05, P < 0.01 and P < 0.001, respectively.

Results

Condition for NMDA receptor activation without apparent depolarization

We looked for suitable conditions in which NMDA receptors of recorded neurons were sufficiently activated without apparent depolarization. When NMDA (200 μm) was dissolved in a standard NaCl-based solution and locally applied for 2 s, a large outward current was induced at + 40 mV (Fig. 1Aa, upper trace). In the same neuron, NMDA application induced only a small inward current at −80 mV (Fig. 1Aa, lower trace), as expected from the ‘Mg²⁺ block’ of NMDA receptors. The relative amplitude of NMDA-induced currents at −80 mV compared to that recorded at + 40 mV was −0.02 ± 0.01 (n= 6) (Fig. 1B, left). When Mg²⁺ was omitted from the NMDA-containing solution, NMDA application induced a large inward current even at −80 mV (Fig. 1Ab, lower trace). The relative amplitude was −1.07 ± 0.22 (n= 7) (Fig. 1B, middle). Such a large inward current could induce local depolarization at distal dendrites, and thereby might activate voltage-gated Ca²⁺ channels. To minimize NMDA-induced local depolarization, Na⁺ in the Mg²⁺-free NMDA-containing solution was replaced with a membrane-impermeant cation, NMDG. With this solution, the NMDA-induced inward current at −80 mV was markedly reduced (−0.06 ± 0.05, n= 7, relative to the current at +40 mV) (Fig. 1A and B, right). Under this condition, sufficient NMDA receptor activation was expected to be achieved without activation of voltage-gated Ca²⁺ channels. Therefore, this NMDG-based solution was used for NMDA application in the following experiments.

NMDA-induced endocannabinoid release

We first examined whether activation of NMDA receptors causes endocannabinoid release, by monitoring the amplitude of IPSCs. IPSCs in hippocampal neurons are classified into two categories, namely cannabinoid sensitive and cannabinoid insensitive (Ohno-Shosaku et al. 2001). Their cannabinoid sensitivities can be checked easily by applying a depolarizing voltage pulse to the postsynaptic neuron (0 mV, 3–5 s). Cannabinoid-sensitive IPSCs display a transient suppression following depolarization, a phenomenon known as depolarization-induced suppression of inhibition (DSI) (Fig. 2A) (Ohno-Shosaku et al. 2001). By contrast, cannabinoid-insensitive IPSCs do not exhibit DSI (Fig. 2E). We compared effects of NMDA application in these two distinct types of IPSCs. In cannabinoid-sensitive IPSCs, local application of NMDA (200 μm) for 5 s always induced transient suppression (Fig. 2A). On average, effects of depolarization (3 s) and NMDA application were comparable (Fig. 2B). The NMDA-induced suppression was blocked by an NMDA receptor antagonist, CPP (50 μm) (Fig. 2A and C). In contrast to cannabinoid-sensitive IPSCs, cannabinoid-insensitive IPSCs were not affected by NMDA application (Fig. 2D and E).

Figure 2. Effects of NMDA application on IPSCs.

A, representative data showing NMDA-induced suppression of cannabinoid-sensitive IPSCs. The amplitude of IPSC evoked by presynaptic stimulation was plotted as a function of time. At the time indicated by the open arrow, the postsynaptic neuron was depolarized from −80 mV to 0 mV for 5 s. At the times indicated by the filled arrows, 200 μm NMDA was locally applied for 5 s. CPP (50 μm) was added to both the bath solution and the NMDA-containing solution for the duration indicated by the horizontal bar. IPSC traces acquired at the indicated time points (a–f) are shown on the lower panel. Each trace is the average of 3–5 consecutive IPSCs. B, averaged data for the IPSC suppression induced by depolarization (0 mV, 3 s) and NMDA application (200 μm NMDA, 5 s) from six neuron pairs. C, averaged data for the effects of CPP (50 μm) on NMDA-induced suppression. D and E, representative experiment (D) and averaged data (E) showing the effects of depolarization (0 mV, 3 s) and NMDA application (200 μm NMDA, 5 s) on cannabinoid-insensitive IPSCs.

The involvement of endocannabinoids was further confirmed by using the CB₁ antagonist AM281. Effects of depolarization and NMDA application on cannabinoid-sensitive IPSCs were tested before and after the treatment with AM281 (Fig. 3A), the action of which is known to be almost irreversible. We usually applied the cannabinoid agonist WIN55,212-2 prior to AM281, in order to confirm the cannabinoid sensitivity of the IPSC and the antagonizing effect of AM281 on WIN55,212-2. After the treatment with AM281, both DSI and NMDA-induced suppression were completely abolished (Fig. 3A and B). Using the same preparation, we previously reported that the DAG lipase inhibitor THL blocked DSI without affecting the presynaptic cannabinoid sensitivity, suggesting that DSI is mediated by 2-AG (Hashimotodani et al. 2007c). Like DSI, the NMDA-induced suppression of IPSCs was found to be abolished in the THL-pretreated neurons (Fig. 3C and D). Taken together, these data indicate that NMDA receptor activation induces endocannabinoid release and causes a transient suppression of cannabinoid-sensitive IPSCs.

Figure 3. Cannabinoid dependence of NMDA-induced suppression of IPSCs.

A, representative data showing blockade of NMDA-induced IPSC suppression by the CB₁ antagonist AM281. Effects of depolarization (0 mV, 3 s) and NMDA application (200 μm, 5 s) on IPSCs were examined before and after the sequential treatment with 0.3 μm WIN55,212-2 and 0.3 μm AM281. Each trace (lower panel) is the average of 5 consecutive IPSCs acquired at the indicated time point (a–j). B, averaged data for IPSC suppression induced by depolarization and NMDA application before and after the treatment with AM281. C and D, representative experiment (C) and averaged data (D) for the effects of treatment with THL (5 μm, 3–5 min) on NMDA-induced IPSC suppression.

Mechanisms of NMDA-induced endocannabinoid release

The mechanisms of NMDA-induced endocannabinoid release were further studied. A simple explanation is that Ca²⁺ influx through NMDA receptors directly stimulates endocannabinoid production. However, the possibility remains that the activation of NMDA receptors might cause local depolarization, which in turn activates voltage-gated Ca²⁺ channels and induces an endocannabinoid release. To test this possibility, we applied NMDA under the blockade of voltage-gated Ca²⁺ channels. In the first series of experiments, we used pharmacological tools. It is reported that N-type Ca²⁺ channels exclusively contribute to the transmitter release at cannabinoid-sensitive GABAergic terminals in the hippocampus (Wilson et al. 2001). Therefore, we tried to block voltage-gated Ca²⁺ channels other than N-type by using Ca²⁺ channel blockers for P/Q-type (AgTX), R-type (SNX-482) and L-type (nifedipine). Application of a mixture of 0.1 μm AgTX, 0.1 μm SNX-482 and 20 μm nifedipine decreased the total Ca²⁺ current by 69.4 ± 3.6% (n= 5). The treatment with these Ca²⁺ channel blockers markedly reduced DSI, but not the NMDA-induced suppression (Fig. 4). In the second series of experiments, we electrically inactivated voltage-gated Ca²⁺ channels. In this series of experiments, we used the Mg²⁺-containing solution for NMDA application. When NMDA was applied to the neurons held at −80 mV, IPSCs were not evidently suppressed (Fig. 5A and C). When the holding potential of the postsynaptic neuron was changed from −80 mV to 0 mV, a large outward current was induced and IPSCs were suppressed due to activation of voltage-gated Ca²⁺ channels. When the membrane potential was held at 0 mV for several minutes, the outward current decreased and IPSCs recovered gradually. At this steady-state condition, voltage-gated Ca²⁺ channels are expected to be inactivated. In these neurons held at 0 mV, NMDA application induced a robust suppression of IPSCs (Fig. 5A and C). The NMDA-induced suppression of IPSCs at 0 mV was less prominent in the neurons dialysed with 30 mm BAPTA (Fig. 5B and C, right; P < 0.01, unpaired t test). These results collectively indicate that voltage-gated Ca²⁺ channels make little contribution to the NMDA-induced suppression, and that the elevation of intracellular Ca²⁺ concentration caused by Ca²⁺ influx through NMDA receptors induces endocannabinoid release from hippocampal neurons.

Figure 4. Effects of voltage-gated Ca²⁺ channel blockers on DSI and NMDA-induced suppression of IPSCs.

A, representative experiment. Effects of depolarization (0 mV, 3 s) and NMDA application (200 μm, 5 s) on IPSCs were examined before (left) and after (right) the treatment with 0.1 μm AgTX, 0.1 μm SNX-482 and 20 μm nifedipine for more than 3.5 min. B, averaged data for the effects of voltage-gated Ca²⁺ channel blockers on DSI and NMDA-induced suppression of IPSCs.

NMDA-induced enhancement of receptor-driven endocannabinoid release

Next, we determined whether the activation of NMDA receptors can facilitate the endocannabinoid release through the receptor-driven pathway. When a low dose of the muscarinic agonist oxo-M (0.3 μm) was applied locally to the postsynaptic neuron for 1–5 s, IPSCs were suppressed slightly (Fig. 6A–C). Application of a low dose of NMDA (20 μm) caused only minor effects on IPSCs (Fig. 6A–C). When oxo-M (0.3 μm) and NMDA (20 μm) were applied simultaneously, IPSCs were markedly suppressed (Fig. 6A–C). Effects of oxo-M alone, NMDA alone and oxo-M plus NMDA were independent of the order of application, and were repeatable. The marked suppression induced by the coapplication of oxo-M and NMDA was antagonized by AM281 (Fig. 6A and D), indicating that the suppression is mediated by endocannabinoids. The enhancement of endocannabinoid release by coapplication of oxo-M and NMDA can be explained by the enhancement of PLCβ activity by NMDA-induced Ca²⁺ elevation, similar to the enhancement by depolarization-induced Ca²⁺ elevation (Hashimotodani et al. 2005). However, it is also possible that NMDA receptors might be potentiated by muscarinic activation (Harvey et al. 1993). To test this possibility, we measured inward currents induced by application of NMDA (20 μm) alone or coapplication of NMDA (20 μm) and oxo-M (0.3 μm). We obtained no evidence that shows potentiation of NMDA-induced currents by oxo-M under these experimental conditions (Fig. 6E and F).

Figure 6. Enhancement of muscarinic receptor-driven endocannabinoid release by NMDA application.

A, representative data showing enhanced endocannabinoid release by simultaneous activation of NMDA and muscarinic receptors. The amplitude of cannabinoid-sensitive IPSCs was plotted as a function of time. At the time points indicated by open, filled and grey arrows, 0.3 μm oxo-M, 0.3 μm oxo-M plus 20 μm NMDA, and 20 μm NMDA, respectively, were locally applied for 3 or 5 s. WIN55,212-2 (0.3 μm) and AM281 (0.3 μm) were bath-applied for the duration indicated by horizontal bars. Each IPSC trace (lower panel) is the average of 5–10 consecutive IPSCs acquired at the indicated time points a–j. B and C, individual (B) and averaged (C) data for the IPSC suppression by application of oxo-M, oxo-M + NMDA, and NMDA obtained from the experiments similar to A. The duration of agonist application ranged from 1 to 5 s. D, averaged data for the IPSC suppression induced by application of oxo-M + NMDA before and after the treatment with AM281. E and F, representative experiment (E) and averaged data (F) for inward currents induced by consecutive applications of NMDA (20 μm) alone, NMDA (20 μm) + oxo-M (0.3 μm), and NMDA (20 μm) alone.

Similar experiments were performed with the I-mGluR agonist DHPG. We found that a low dose of NMDA (20 μm) significantly enhanced the suppression of IPSCs induced by DHPG. Application of DHPG (5 μm) induced only a small suppression of IPSCs (Fig. 7A–C) but this DHPG-induced suppression was significantly enhanced by the simultaneous application of 20 μm NMDA (Fig. 7A–C). This enhanced suppression was blocked by AM281. Co-application of DHPG (5 μm) did not significantly enhance NMDA-induced inward currents (Fig. 7E and F). These results strongly suggest that mild activation of NMDA receptors, which by itself cannot induce endocannabinoid release, can enhance the endocannabinoid release driven by G_q-coupled receptor activation.

Figure 7. Enhancement of I-mGluR-driven endocannabinoid release by NMDA application.

A, representative data showing enhanced endocannabinoid release by simultaneous activation of NMDA receptors and I-mGluRs. The amplitude of cannabinoid-sensitive IPSCs was plotted as a function of time. At the time points indicated by open, closed and grey arrows, 5 μm DHPG, 5 μm DHPG + 20 μm NMDA, and 20 μm NMDA, respectively, were locally applied for 1 s. B and C, individual (B) and averaged (C) data for the IPSC suppression by application of DHPG, DHPG + NMDA, and NMDA obtained from the experiments similar to A. The duration of agonist application ranged from 1 to 10 s. Data connected by a line were obtained from the same neuron pair. D, averaged data for the IPSC suppression induced by application of DHPG + NMDA before and after the treatment with AM281. E and F, representative experiment (E) and averaged data (F) for inward currents induced by consecutive applications of NMDA (20 μm) alone, NMDA (20 μm) + DHPG (5 μm), and NMDA (20 μm) alone.

Discussion

The present study demonstrates that the activation of NMDA receptors can directly contribute to the generation of endocannabinoid signalling through elevating postsynaptic Ca²⁺ concentration. Strong activation of NMDA receptors alone induced endocannabinoid release, and suppressed the cannabinoid-sensitive synaptic transmission. Mild activation of NMDA receptors, which by itself caused little effect, enhanced the endocannabinoid release driven by muscarinic receptors or I-mGluRs. These functions of NMDA receptors are essentially the same as those of voltage-gated Ca²⁺ channels (Ohno-Shosaku et al. 2005; Chevaleyre et al. 2006). Because NMDA receptor activation may cause depolarization of neurons and thereby open voltage-gated Ca²⁺ channels, it is not easy to separate the effect resulting from NMDA receptor activation from secondary activation of voltage-gated Ca²⁺ channels. In the present study, we tried to separate these two, and demonstrate that the effect of NMDA is attributable to Ca²⁺ influx through NMDA receptors into hippocampal neurons. Under physiological conditions, however, NMDA receptors and voltage-gated Ca²⁺ channels may work in concert with each other. Cooperative activation of Ca²⁺ channels, NMDA receptors and G_q-coupled receptors might be physiologically important for the activity-dependent generation of endocannabinoid signal.

In the literature, there are several studies suggesting NMDA-induced endocannabinoid release. In a study with gas chromatography/mass spectrometry analysis, Stella & Piomelli (2001) measured the contents of endocannabinoids in cultured cortical neurons under various conditions. Incubation of neurons with 300 μm NMDA for 2.5 min caused a marked accumulation of 2-AG, which was enhanced by coincubation with the cholinergic agonist carbachol (1 mm). Although their experimental conditions are quite different from ours (i.e. a longer exposure time and higher agonist doses), these data are qualitatively consistent with those in the present study. In a electrophysiological study on hypoglossal motoneurons, Mukhtarov et al. (2005) investigated effects of postsynaptic Ca²⁺ elevation on glycinergic transmission. Although this study aimed to characterize depolarization-induced suppression of glycinergic transmission and Ca²⁺-dependent potentiation of glycine receptors, it additionally included the experiments with NMDA. Application of 200 μm NMDA dissolved in a standard external solution (2 mm Ca²⁺ and 1.3 mm Mg²⁺) for 5–10 s induced a suppression of glycinergic IPSCs at −70 mV, which was abolished by a CB₁ antagonist. The details of this phenomenon were not described, and it is not clear how NMDA receptors were activated under these conditions (−70 mV, 1.3 mm Mg²⁺). It seems likely that NMDA application induced a small local depolarization, which further activated NMDA receptors and induced a further depolarization, resulting in regenerative activation of NMDA receptors and membrane depolarization. If this regenerative process was activated, the observed NMDA-induced suppression of glycinergic IPSCs might be caused, at least partly, by the DSI mechanism. In the present study, we used a Mg²⁺-free, NMDG-based solution for NMDA application, and minimized the local depolarization. Furthermore, we confirmed that the NMDA-induced suppression of IPSCs persisted under the pharmacological or electrical blockade of voltage-gated Ca²⁺ channels. Therefore, the present study is the first clear demonstration of the direct effect of NMDA receptor activation on generation of the endocannabinoid signal in neurons.

In the present study, we used cannabinoid-sensitive IPSCs to monitor the endocannabinoid release. In our culture conditions, we found that pharmacological activation of NMDA receptors induced the endocannabinoid release and suppressed IPSCs. In physiological conditions, however, synaptically driven endocannabinoid release through activation of NMDA receptors would be restricted around excitatory postsynaptic sites of dendrites, and the released endocannabinoids may hardly act on somatic inhibitory synapses. Therefore, the retrograde modulation of inhibitory transmission observed in the present study may be limited to, if any, the dendritic inhibitory synapses that are located close to excitatory synapses. Alternatively, synaptic activation of NMDA receptors might also occur at inhibitory synapses. At a selected population of inhibitory synapses on CA1 pyramidal cell soma, it has been demonstrated that aspartate, an agonist of NMDA receptors, is present on presynaptic terminals, whereas NMDA receptors were concentrated on postsynaptic membranes (Gundersen et al. 2004). Interestingly, it has been reported that vesicular glutamate transporter type 3 (vGluT3) is selectively expressed at cholecystokinin (CCK)-positive inhibitory terminals, which are known to be cannabinoid sensitive, in the hippocampus (Somogyi et al. 2004). It might be possible that aspartate is packed into synaptic vesicles by vGluT3, released together with GABA from cannabinoid-sensitive inhibitory terminals, and activates postsynaptic NMDA receptors. Further investigation using acute slice preparations is needed to evaluate the physiological relevance of the NMDA-receptor-mediated endocannabinoid release in the modulation of inhibitory transmission.

Recently, the contribution of NMDA receptors to synaptically driven endocannabinoid release has been reported at excitatory synapses in several brain regions (Beierlein & Regehr, 2006; Crozier et al. 2007; Tzounopoulos et al. 2007). In stellate cells (SCs) of the cerebellum, a brief parallel fibre burst induced a CB₁-dependent transient suppression, which was prevented by postsynaptic BAPTA injection. This Ca²⁺-dependent endocannabinoid release was not altered by blocking either mGluR1 alone or NMDA receptors alone, but was prevented by blocking both types of receptors, indicating that the synaptically driven endocannabinoid release from SCs requires activation of NMDA receptors under the blockade of mGluR1 (Beierlein & Regehr, 2006). In pyramidal neurons of visual cortical layer 2/3, pairing stimulation (layer 4 stimulation with postsynaptic depolarization from −65 to −45 mV) at 1 Hz for 5–10 min induced CB₁-dependent long-term depression (LTD). This LTD was blocked by intracellular loading of the NMDA receptor antagonist MK801, indicating that postsynaptic NMDA receptors are involved in pairing stimulation-induced endocannabinoid release (Crozier et al. 2007). In interneurons of dorsal cochlear nucleus, LTD was induced by pre-post timing protocol in which a presynaptic input was followed by a postsynaptic spike, which is usually used to induce LTP in the sensory cortex. This timing-dependent LTD was expressed presynaptically, and depended on postsynaptic Ca²⁺ elevation, postsynaptic NMDA receptors and presynaptic CB₁ receptors, but not on I-mGluRs (Tzounopoulos et al. 2007). These data strongly suggest that the pairing stimulation could induce the endocannabinoid release through activation of NMDA receptors. In these studies, however, it is unclear whether Ca²⁺ influx through NMDA receptors was sufficient to induce the NMDA receptor-dependent endocannabinoid release or whether voltage-gated Ca²⁺ channels were also involved. Nevertheless, these studies have clearly demonstrated that synaptically activated NMDA receptors can contribute to the generation of the endocannabinoid signal.

Coincidence detection of two distinct signals is important for various aspects of neural functions. It is generally accepted that NMDA receptors function as a coincidence detector of postsynaptic depolarization and presynaptic glutamate release, which play crucial roles in induction of long-term potentiation and depression in various synapses in the CNS (Cotman et al. 1988; Bliss & Collingridge, 1993; Malenka & Nicoll, 1999). Another example is PLCβ. This enzyme is activated effectively when activation of G_q-coupled receptors coincides with Ca²⁺ elevation (Hashimotodani et al. 2007a). Thus, it can detect the coincidence of presynaptic transmitter release and postsynaptic depolarization (Ca²⁺ elevation). It is noteworthy that these two coincidence detectors are involved in generation of the endocannabinoid signal. The involvement of endocannabinoids in induction of synaptic plasticity has been reported in various brain regions (Chevaleyre et al. 2006; Hashimotodani et al. 2007b). It is interesting to investigate roles of these coincidence detectors in the cannabinoid-dependent forms of synaptic plasticity.

In the hippocampus, it has been demonstrated that LTD of inhibitory transmission (LTD_i) triggered by high-frequency stimulation is mediated by endocannabinoids (Chevaleyre & Castillo, 2003, 2004). This cannabinoid-dependent form of LTD_i was shown to require the activation of I-mGluRs, but not NMDA receptors. By contrast, several examples of LTD_i that require activation of postsynaptic NMDA receptors for induction have been reported (Stelzer et al. 1987; McLean et al. 1996; Wang & Stelzer, 1996; Caillard et al. 1999; Lu et al. 2000). This NMDA-dependent form of LTD_i has been assumed to be caused by down-regulation of postsynaptic GABA_A receptor functions. The present study, however, suggests the possibility that the endocannabinoid signal triggered by postsynaptic NMDA receptor activation might also contribute to LTD_i.

In the present study, we did not determine the precise mechanisms of the effects of NMDA receptor activation on endocannabinoid release. However, it is reasonable to assume that the Ca²⁺ elevation caused by NMDA receptor activation and that by voltage-gated Ca²⁺ channel activation trigger the same processes. Namely, a large Ca²⁺ elevation by strong activation of NMDA receptors induces the endocannabinoid release through a PLCβ-independent pathway, whereas a small Ca²⁺ elevation by mild activation of NMDA receptors enhances the receptor-driven 2-AG release through Ca²⁺-dependent enhancement of PLCβ activation and the subsequent DAG lipase activity. Although fundamental intracellular mechanisms for endocannabinoid production appear to be similar, activation of voltage-gated Ca²⁺ channels and that of NMDA receptors may lead to different consequences in synaptic modulation. The Ca²⁺ elevation driven by voltage-gated Ca²⁺ channels is generally global, whereas that by NMDA receptors is restricted to around postsynaptic sites. Therefore, the endocannainoid signal driven by NMDA receptors can contribute to spatially limited synaptic modulation. At hippocampal excitatory synapses, NMDA receptors, mGluR5, and DAG lipase α, which are involved in 2-AG production, are all located on the postsynaptic membranes close to each other (Takumi et al. 1999; Racca et al. 2000; Yoshida et al. 2006). The CB₁ receptor is located at excitatory synaptic terminals facing postsynaptic dendritic spines (Katona et al. 2006; Kawamura et al. 2006). The synaptic localization of these signalling molecules suggests the possibility that endocannabinoids are released from restricted regions of postsynaptic neurons in activity-dependent manners, modulate transmitter release from a small number of presynaptic terminals, and cause a pathway-selective plasticity.