Cyclooxygenase-2, prostaglandin E2 glycerol ester and nitric oxide are involved in muscarine-induced presynaptic enhancement at the vertebrate neuromuscular junction

Abstract

Previous work has demonstrated that activation of muscarinic acetylcholine receptors at the lizard neuromuscular junction (NMJ) induces a biphasic modulation of evoked neurotransmitter release: an initial depression followed by a delayed enhancement. The depression is mediated by the release of the endocannabinoid 2-arachidonylglycerol (2-AG) from the muscle and its binding to cannabinoid type 1 receptors on the motor nerve terminal. The work presented here suggests that the delayed enhancement of neurotransmitter release is mediated by cyclooxygenase-2 (COX-2) as it converts 2-AG to the glycerol ester of prostaglandin E2 (PGE₂-G). Using immunofluorescence, COX-2 was detected in the perisynaptic Schwann cells (PSCs) surrounding the NMJ. Pretreatment with either of the selective COX-2 inhibitors, nimesulide or DuP 697, prevents the delayed increase in endplate potential (EPP) amplitude normally produced by muscarine. In keeping with its putative role as a mediator of the delayed muscarinic effect, PGE₂-G enhances evoked neurotransmitter release. Specifically, PGE₂-G increases the amplitude of EPPs without altering that of spontaneous miniature EPPs. As shown previously for the muscarinic effect, the enhancement of evoked neurotransmitter release by PGE₂-G depends on nitric oxide (NO) as the response is abolished by application of either N^G-nitro-l-arginine methyl ester (l-NAME), an inhibitor of NO synthesis, or carboxy-PTIO, a chelator of NO. Intriguingly, the enhancement is not prevented by AH6809, a prostaglandin receptor antagonist, but is blocked by capsazepine, a TRPV1 and TRPM8 receptor antagonist. Taken together, these results suggest that the conversion of 2-AG to PGE₂-G by COX-2 underlies the muscarine-induced enhancement of neurotransmitter release at the vertebrate NMJ.

Key points

• The synapse between a nerve and muscle, called the neuromuscular junction (NMJ), undergoes a biphasic modulation, a decrease followed by an increase, when muscarinic acetylcholine receptors are continuously activated.

• The initial depression is caused by the endocannabinoid 2-arachidonylglycerol (2-AG), which is synthesized in and released from the muscle; 2-AG then activates cannabinoid receptors on the presynaptic nerve.

• In the work presented here, we explored the mechanism responsible for the delayed enhancement, uncovering a role for the enzyme cyclooxygenase-2 and locating it in the glial cells at the NMJ called perisynaptic Schwann cells (PSCs) where it converts 2-AG into the glycerol ester of prostaglandin E2.

• These results reveal a complex mechanism for regulating neurotransmitter release that involves the nerve, muscle and PSCs (i.e. the tripartite synapse) and may serve to ensure reliable neuromuscular transmission during periods of intense or long-term activity.

Introduction

Since the discovery of endocannabinoids (eCBs) much research has focused on the function of membrane-derived lipids in synaptic plasticity. At most synapses, eCBs are released from the postsynaptic cell in response to depolarization (Ohno-Shosaku et al. 2001; Wilson & Nicoll, 2001) and/or the activation of metabotropic receptors, such as muscarinic acetylcholine (ACh) receptors (Kim et al. 2002; Fukudome et al. 2004). Once released, eCBs bind to the cannabinoid type 1 (CB₁) receptor on the presynaptic terminal and inhibit neurotransmitter release (Maejima et al. 2001). Although eCBs were first shown to modulate synapses in the CNS, they have also been implicated in peripheral synapses (Newman et al. 2007; Sánchez-Pastor et al. 2007; Silveira et al. 2010). At the vertebrate neuromuscular junction (NMJ), the eCB 2-arachidonoylglycerol (2-AG) is responsible for the inhibition of neurotransmitter release initiated either by long-term, low-frequency stimulation or by activation of M₃ muscarinic receptors. In both cases, this inhibition requires the presence of nitric oxide (NO; Newman et al. 2007).

With continued activation of muscarinic receptors at the NMJ, specifically the M₁ receptor, the reduction of neurotransmitter release gives way, approximately 30 min later, to an enhancement of release (Graves et al. 2004). Other than also requiring NO (Graves et al. 2004), the mechanism of this delayed enhancement has remained a mystery. As Sang et al. (2006, 2007) found that several products derived from the cyclooxygenation of eCBs increase neurotransmitter release in the mouse hippocampus, the present study examined whether a similar process might underlie the delayed enhancement of neurotransmitter release at the NMJ. In particular, we asked whether the prostaglandin E2 glycerol ester (PGE₂-G), which is produced by the cyclooxygenation of 2-AG, mediates the delayed muscarine-induced enhancement. After first localizing cyclooxygenase-2 (COX-2) to the NMJ using immunofluorescence, we demonstrated its functional relevance by blocking the muscarine-induced enhancement with COX-2 inhibitors. We also demonstrated that application of PGE₂-G mimicked the enhancement, including its requirement for NO. Interestingly, as had been previously shown in the hippocampus (Sang et al. 2006), PGE₂-G does not act via known prostanoid receptors.

Methods

Ethical approval

All of the procedures used in the research reported here were approved by the Institutional Animal Use and Care Committee at Grinnell College.

Experimental preparation

To facilitate quick and accurate ablation of the forebrain and to minimize discomfort, small (5–8 cm) lizards (Anolis carolinensis; Carolina Biological Supply Co., Burlington, NC, USA) of either sex were placed at 7–10 °C for 8–10 min prior to decapitation. The ceratomandibularis muscle and its motor nerve, a small branch of the hypoglossal nerve, were isolated as described by Lindgren and Moore (1989) and pinned in a Sylgard-coated dish containing fresh Ringer solution (158 mm NaCl₂, 2 mm KCl, 2 mm MgCl₂, 5 mm Hepes, 2 mm CaCl₂, and 2 g l⁻¹ dextrose, pH 7.3). Ringer solution was made daily from stock solutions. The bathing solution was changed every 10 min to increase the longevity of the muscle. All chemicals, unless noted otherwise, were purchased from Sigma-Aldrich (St Louis, MO, USA).

Electrophysiology and data analysis

To prevent action potentials and muscle contraction, end-plate potentials (EPPs) were reduced below action potential threshold by applying 8 μm d-tubocurarine chloride (dTC) in the Ringer solution. EPPs were evoked by stimulating the motor nerve axon with square pulses at 0.2 Hz, 1.4 V, 0.2 ms using a Grass S88 stimulator. EPPs were recorded using a glass micropipette filled with 3 m KCl (resistance 15–20 MΩ). Membrane potentials were amplified with a Dagan 8700 Cell Explorer Amplifier, filtered with HumBug noise eliminator (Quest Scientific, North Vancouver, BC, Canada), and collected with a Maclab data acquisition system (ADInstruments, Colorado Springs, CO, USA). The amplitude of each EPP was measured after averaging 8–16 individual sweeps. Two protocols were used for monitoring changes in EPP amplitude during experiments. In one case, an EPP was recorded from a single end-plate for the duration of the experiment (see Figs 3A, 4B and 5B). In the second protocol used, EPPs were recorded from four or five randomly chosen synapses to determine a mean baseline EPP amplitude. After a treatment (e.g. drug application), EPPs were again recorded from four or five randomly chosen synapses. Treatment effects on EPP amplitudes were calculated as percentage change from baseline (see Figs 3B, 4A and 5A). Each treatment was repeated the number of times indicated in the text or figure legends, where n indicates the number of muscles examined. Unless stated otherwise, data are presented as mean ± SEM. A single-factor ANOVA was used to analyse the data, taking P < 0.05 as significant. Miniature end-plate potentials (MEPPs) were measured in the absence of stimulation and dTC. Only muscles with resting membrane potentials of at least −80 mV were included in this study.

Figure 3. PGE₂-G increases neurotransmitter release.

A, end-plate potentials (EPPs) measured in a single muscle cell with an intracellular microelectrode are plotted during the application of PGE₂-G via a pressure pulse from a pipette positioned directly over the NMJ. The PGE₂-G in the pipette was dissolved in Ringer solution at a concentration of 468 μm and applied with a 10 s, 20 p.s.i. pulse at the time indicated by the arrow. B, mean percentage change from baseline EPP amplitude is plotted during bath application of PGE₂-G (4.68 μm, n= 10); WASH (i.e. immediately following washout of PGE₂-G with normal saline, n= 10); PGD₂-G (4.69 μm, n= 4); PGE₂-G and AH6809 (10 μm, n= 4); PGE₂-G and capsazepine (2 μm, n= 5); and PGD₂-G and capsazepine (2 μm, n= 3). EPPs were recorded from 4–5 randomly chosen synapses to determine a mean baseline EPP amplitude. After a treatment (e.g. drug application), EPPs were again recorded from 4–5 randomly chosen synapses. Treatment effects on EPP amplitudes were calculated as percentage change from baseline. Each treatment was repeated the number of times indicated in the text or figure legends, where n indicates the number of muscles examined. Changes that are significantly different from baseline are indicated with an asterisk (P < 0.01; one-way ANOVA). C, sample MEPPs recorded before (top) and after (bottom) the application of PGE₂-G (4.68 μm). Calibration, 1 mV, 1 s. D, bars represent either the mean change from baseline of frequency (solid) or amplitude (open) of MEPPs recorded during the application of PGE₂-G (4.68 μm) in three preparations. All data are expressed as a percentage of the mean frequency or amplitude before application of PGE₂-G. Error bars represent ± SEM. The baseline MEPP amplitude and frequency were 0.506 ± 0.045 mV and 0.449 ± 0.056 Hz, respectively. Resting membrane potentials were at least −80 mV. The asterisks indicate the mean is significantly different from control (P < 0.05; one-way ANOVA).

Figure 4. The synaptic enhancement induced by PGE₂-G requires NO.

A, mean percentage change from baseline EPP amplitude is plotted during bath application of the following: PGE₂-G (4.68 μm; n= 10); PGE₂-G and l-NAME (0.3 mm; n= 3); PGE₂-G, l-NAME and DEA/NO (0.1 mm; n= 3); and PGE₂-G and carboxy-PTIO (cPTIO, 40 μm; n= 3). The percentage change from baseline EPP amplitude was determined as described in Fig. 3B. Changes that are significantly different from baseline are indicated with an asterisk (P < 0.01; one-way ANOVA). B, percentage change from baseline of end-plate potentials (EPPs) measured in a single muscle cell with an intracellular microelectrode is plotted before and during the application of PGE₂-G (4.68 μm), and following the addition of carboxy-PTIO (cPTIO, 40 μm) in the continued presence of PGE₂-G. Each data point represents the average of 16 AC-coupled sweeps. Sample EPP traces (averages of 8–16 AC-coupled sweeps) collected during each condition are displayed above the corresponding data point. Resting membrane potentials were approximately −90 mV. Calibration bar: 0.5 mV, 2 ms.

Figure 5. The muscarine-induced synaptic enhancement requires COX-2 and is blocked by capsezepine.

A, mean percentage change in EPP amplitudes measured before and 30 min after incubation with muscarine (5 μm throughout). The percentage change is plotted for muscles in muscarine alone (n= 4); muscarine with the COX inhibitor DuP 697 (1 μm; n= 8); muscarine with the COX inhibitor nimesulide (3 μm; n= 12), and muscarine with capsazepine (2 μm; n= 4). The percentage change from baseline EPP amplitude was determined as described in Fig. 2B. The mean percentage change with only muscarine in the saline is significantly different from the change with the addition of either DuP 697, nimesulide or capsazepine (*P < 0.01; one-way ANOVA). Furthermore, in the presence of nimesulide, the application of muscarine significantly reduced EPP amplitudes below baseline (‡P < 0.05, one-way ANOVA). B, percentage change from baseline of EPPs measured in a single muscle cell with an intracellular microelectrode is plotted before and during the application of muscarine (5 μm), and following the addition of capsazepine (2 μm) in the continued presence of muscarine. Each trace represents the average of 16 sweeps. Resting membrane potentials were approximately −90 mV. Calibration bars: 0.5 mV, 2 ms.

Drug application

Application of all drugs was conducted in the same manner: the preparation was bathed in the given concentration of the drug dissolved in fresh Ringer solution. Stock aliquots were prepared ahead of time and then diluted to the following concentration immediately before application: 5.0 μm muscarine, 4.7 μm PGE₂-G, 4.7 μm prostaglandin D2 glycerol ester (PGD₂-G), 10 μm AH6809 (6-isopropoxy-9-xanthione-2-carboxylic acid), 2 μm capsazepine, 0.3 mm N^G-nitro-l-arginine methyl ester (l-NAME), 0.1 mm Diethylamine NONOate (DEA-NO) and 40 μm 2-(4-carboxyphenyl)-4, 5-dihydro- 4, 4, 5, 5-tetramethyl-1H-imidazolyl-1-oxy-3-oxide (carboxy-PTIO). The physiological effects of solvents were considered to be negligible as applications of the solvents per se at comparable dilutions (1:1000) showed no effect.

Immunofluorescence

Muscles were pre-incubated at 24°C for approximately 1 h in Ringer solution containing muscarine (5 μm). They were then immediately fixed in 3% paraformaldehyde in glucose-free Ringer solution at 4°C for 1 h, rinsed for 1 h at 24°C in glucose-free Ringer solution (pH 8), permeabilized for 30 min at 37°C in 0.3% Triton X-100, and rinsed for 60 min at 24°C in blocking solution (BS; 0.01% Triton X-100, 1% IgG-free bovine serum albumin). After fixation, muscles were pre-incubated for 1 h at 37°C in BS, rinsed in BS at 24°C for 5 min, and incubated in primary antibody (2 μg ml⁻¹ rabbit anti-COX-2 polyclonal antibody #AB5118, Millipore Corporation, Billerica, MA, USA) for 12–24 h at 4°C. Muscles were then rinsed for 1 h in BS, incubated with Alexa Fluor 488-conjugated goat anti-rabbit IgG secondary antibody (5 μg ml⁻¹; American Qualex, San Clemente, CA, USA) or with Alexa Fluor 555-conjugated goat anti-rabbit IgG secondary antibody (Invitrogen, Carlsbad, CA, USA) for 2 h at 37°C, rinsed in BS for 60 min, and mounted on slides with ProLong Gold antifade reagent with DAPI (Invitrogen). Control experiments were performed by adding the secondary antibody without the primary antibody and by preabsorbing the primary antibody with recombinant human COX-2 (Invitrogen) for 5 h at 4°C prior to being added to the tissue.

In addition to being labelled with anti-COX-2 antibody, as described above, each muscle was co-stained with a second fluorophore, as follows. To reveal the nicotinic ACh receptors at the muscle end-plate, α-bungarotoxin (α-BTX), conjugated to Alexa-Fluor 555, was applied (2 μg ml⁻¹) for 15 min at 24°C, just prior to mounting the tissue. To visualize nerve terminals, either: (1) preparations were incubated with 2 μg ml⁻¹ mouse anti-synaptotagmin monoclonal antibody (mAb 48, Developmental Studies Hybridoma Bank at the University of Iowa) and either goat anti-mouse secondary antibody conjugated to Alexa Fluor 555 or chicken anti-mouse secondary antibody conjugated to Alexa Fluor 647 (5 μg ml⁻¹; Invitrogen); or (2) the cut end of the motor axon was dipped into a small (1–2 μl) well containing 20 mm Texas Red conjugated to 10,000 molecular weight dextran (Molecular Probes, Carlsbad, CA, USA) in 10 mm Hepes buffer (pH 7.2) and incubated overnight at 9°C to allow the nerve terminals to fill with the Texas Red dextran. To visualize the perisynaptic Schwann cells (PSCs), preparations were either (1) incubated with YOYO-1 Iodide (125 nm, Y3601; Invitrogen) for 5 min at 24°C just prior to mounting or (2) incubated with 2 μg ml⁻¹ mouse anti-HNK-1 IgM monoclonal antibody (C6680; Sigma-Aldrich) and goat anti-mouse IgM secondary antibody conjugated to TRITC (5 μg ml⁻¹; American Qualex).

Microscopy

After being stained, NMJs were imaged with an Olympus IX81 microscope, 60× objective (numerical aperture 1.4), with a DSU confocal attachment (disc no. 2) and a Hamamatsu Orca EM camera. The following filter sets were used to image fluorophores: (1) a standard FITC filter set (Ex 470/90 nm; DM 495 nm; Em 525/50 nm) for Alexa 488, (2) a standard TRITC filter set (Ex 545/30 nm; DM 570 nm; Em 620/60 nm) for TRITC or Alexa Fluor 555, (3) a DAPI filter set (Ex 350/50 nm; DM 400 nm; Em 460/50 nm) for DAPI and (4) a Cy5 filter set (Ex 635/20 nm; DM 640 nm; Em 655 nm LP) for Alexa Fluor 647. All of the images were analysed using SlideBook (Intelligent Imaging Innovations, Inc., Denver, CO, USA). Some of the images were further processed for three-dimensional rendering using Metamorph (Molecular Devices, Inc., Sunnyvale, CA, USA). For all figures in which an image collected using differential interference contrast (DIC) optics was superimposed onto images collected using epifluorescence, the DIC image was shifted slightly (16 pixels) from the epifluorescence image to compensate for the offset created by a 45° mirror in the filter turret. This offset was calibrated previously using prepared slides containing structures that could be unambiguously identified using either DIC or epifluorescence.

Western blot analysis

Western clots were performed on ceratomandibularis muscle or whole brain tissue. The following procedure was modified from Inoue et al. (2006). After being rinsed twice with Ringer solution, the tissue was homogenized and lysed using an ice cold buffer (1% Triton X-100, 50 mm Tris pH 7.4, 150 mm NaCl, and protease inhibitor mixture (Roche, Indianapolis, IN, USA)). The lysate was cleared by centrifugation at 14,000 r.p.m. for 20 min at 4°C. Total protein concentration was measured using a BCA assay kit (Pierce, Rockford, IL, USA). Samples (30 μg protein) were denatured and separated using a Bis-Tris 11% SDS-PAGE gel (BioRad, Hercules, CA, USA) and transferred to PVDF membrane. The membranes were blocked with Tris-buffered saline and 0.1% Tween (TBST) with 5% non-fat milk for 1 h at 24°C. The membrane was then incubated in primary rabbit antibody (1:1000) overnight at 4°C. The membrane was washed for 1 h with TBST and then incubated in horseradish peroxidase-conjugated goat anti-rabbit secondary antibody (1:500; American Qualex) for 2 h at room temperature. Immunoreactive protein was detected using chemiluminescence (Perkin Elmer, Waltham, MA, USA), and images were captured with a digital photo-documentation system (Alpha Innotech, Santa Clara, CA, USA).

Results

As shown previously, the activation of muscarinic ACh receptors (mAChRs) at the lizard NMJ triggers a biphasic modulation of ACh release from the presynaptic terminal (Graves et al. 2004). This automodulation begins as a reduction and is followed by an enhancement of ACh release. Although there is variability in the timing of the switchover from reduction to enhancement, ranging from 15 to 35 min, the enhancement is always preceded by depression and is always maximal by at least 1 h of muscarine application (Fig. 1). The initial inhibition of ACh release has been shown to involve the synthesis and release of the endocannabinoid 2-AG, followed by activation of presynaptic CB₁ receptors (Newman et al. 2007). The mechanism for the delayed component of muscarinic action is the subject of this paper. Following the lead of Sang et al. (2006, 2007) we asked whether this delayed enhancement was due to the conversion of 2-AG to PGE₂-G by the enzyme COX.

Figure 1. Biphasic effect of muscarine.

The continuous application of muscarine (5 μm) yields two distinct effects at the lizard NMJ, an initial depression followed by a delayed enhancement. The percentage change of EPP amplitude from an initial baseline value is plotted as a function of the time after muscarine application began. Each data point (closed square) depicts the mean percentage change from baseline. The x error bars depict the range of times included in each mean. The y error bars depict the standard deviation of the percenatage changes (n= 11 for all the time ranges except for 7.5–22.5 and 30 min where n= 6 and 9, respectively). The shaded area depicts the range of percenatage changes observed. Note that the switch from depression to enhancement occurred within the range 15–35 min after the start of muscarine application; all NMJs were maximally enhanced by 60 min.

COX-2 is present at the vertebrate NMJ

Despite some pharmacological data suggesting a role for COX at the NMJ (Madden & Van der Kloot, 1982, 1985; Arkhipova et al. 2006; Pinard & Robitaille, 2008), there are no direct reports of COX localization at the vertebrate NMJ. Thus, we first attempted to detect COX using immunofluorescence. In our initial attempts, the binding of COX-2 antibodies was variable, with some NMJs/muscles immunoreactive and others not, or only minimally so. However, once we began pre-incubating muscles in muscarine (5 μm) for at least 1 h prior to fixation, we consistently observed high levels of immunoreactivity for COX-2, as illustrated in Fig. 2. One hour of incubation with muscarine was chosen because by this time synaptic transmission is enhanced at all the NMJs we have studied (see Fig. 1). As controls, two muscles were exposed to the secondary antibody without the primary anti-COX-2 antibody and another two muscles were exposed to COX-2 antibody after preabsorbing the antibody with recombinant COX-2 protein. No specific fluorescence could be detected at the NMJs in any of these preparations.

Figure 2. COX-2 is present at the NMJ.

In all panels (A–E), ceratomandibularis muscles were dissected from Anolis lizards, pre-treated with muscarine (5 μm) for at least 1 h, fixed in 3% paraformaldehyde and then incubated with rabbit anti-COX-2 primary antibody (Ab) followed by goat anti-rabbit IgG, Alexa Fluor 488 (green) or Alexa Fluor 555 (red). In each panel, additional fluorophores were added as described below. A, Alexa Fluor 555 α-Bungarotoxin (α-BTX) was applied to stain the nAChRs located on the muscle end-plate. An image collected using DIC optics is superimposed on a single image plane from a confocal section through the NMJ, revealing the relative location of α-BTX (red) and COX-2 (green). The bottom two panels show enlargements (zooms) of the area indicated by the dashed white rectangle. The zoom on the right shows the positions of the PSC nuclei, which are stained by DAPI (blue). The arrows point to a typical raised surface that encircles the postsynaptic ridges containing the AChRs. COX-2 is found primarily in these raised surfaces, which appear to be the edges of surrounding PSCs. B, the motor nerve was back-loaded with Texas Red conjugated to 10,000 molecular weight dextran to reveal the nerve and nerve terminal branches (red). The image shown is a maximum projection of 10 confocal images collected at 0.5 μm intervals along the z-axis. Note that COX-2 is located adjacent to the nerve terminals; in some cases it is located very close to the Texas Red stain, but the two do not co-localize. C, a mouse monoclonal anti-synaptotagmin (SYT) antibody followed by goat anti-mouse secondary antibody conjugated to Alexa Fluor 555 (red) were applied to label the synaptic vesicles and thereby reveal the location of the nerve terminal boutons. A single confocal image plane is shown. Note that the majority of COX-2 staining is outside, although close to, the presynaptic boutons. The DAPI (blue) reveals nuclei, most of which are from PSCs. Note the COX-2 near the motor axon (see arrow). This probably indicates the presence of COX-2 in the myelinating Schwann cells, but other interpretations are possible. D, YOYO-1 (green) was applied to stain the nucleotides inside the PSCs, revealing the nucleus and cytoplasm. DAPI (blue) reveals the nuclei per se. The presynaptic nerve terminal was labelled with mouse monoclonal anti-SYT antibody followed by chicken anti-mouse secondary antibody conjugated to Alexa fluor 647 (white). A single confocal image plane is displayed. In the top panel, SYT is omitted to make it easier to see the overlap of the COX-2 (red) and the PSCs (blue and green). Note that COX-2 (red) is predominantly located in the fine PSC processes, stained exclusively by YOYO-1 (green). In the bottom panel, the SYT (white) is included, revealing the lack of overlap of COX-2 (red) and the nerve terminal boutons. E, a mouse monoclonal anti-HNK1 IgM antibody followed by goat anti-mouse IgM secondary antibody conjugated to TRITC (red) were applied to label the membranes of the PSCs. The image shown is a maximum projection of 18 confocal images collected at 0.5 μm intervals along the z-axis. COX-2 significantly overlaps with HNK-1 (yellow) indicating the close proximity of COX-2 and the PSC membrane. Scale bars = 10 μm (A–E).

The panels in Fig. 2A show a muscle stained for COX-2 and co-stained with α-BTX Alexa Fluor 555, which binds to nicotinic ACh receptors (nAChRs). The top panel displays the NMJ using DIC optics. In the middle panel, the DIC image is overlaid with a confocal image collected midway through the NMJ, showing the location of the nAChRs (red) on the ridges of large post-junctional folds which hold the nerve terminal boutons. This arrangement can be appreciated best in the enlarged zoom images at the bottom of Fig. 2A. In the lower panel, COX-2 immunostaining (green) is added to the overlay. Note that COX-2 is located immediately outside the postsynaptic ridges that contain the nAChRs. This spatial arrangement is maintained throughout the NMJ, as seen in the z-stack of confocal images collected throughout the full extent of the NMJ (see Supplemental Movie 1). The zoomed images reveal that COX-2 is restricted to narrow finger-like processes (arrows) that lie between the DAPI-labelled PSC nuclei (blue) and the nAChRs (red). These are most probably the cytoplasmic extensions of PSCs that can be seen in electron micrographs to tightly abut the nerve terminal membrane (Walrond & Reese, 1985).

Although the above images strongly suggest that COX-2 is in the PSC processes, the following experiments were performed to determine unambiguously whether this was indeed the case. First, the motor nerve was back-loaded with Texas Red dextran, revealing the motor nerve and its branched ending. As seen in Fig. 2B, the COX-2 antibodies (green) did not co-localize at all with the nerve terminal (red), but were instead found clustered in the gaps between the nerve terminal branches and boutons, the area occupied by the PSCs. Secondly, when synaptic vesicles, known to completely fill the presynaptic boutons in this preparation (see fig. 7A in Lindgren et al. 1997), are labelled with an anti-synaptotagmin monoclonal antibody, they are seen to occupy a different compartment from COX-2. As revealed in Fig. 2C, which is a single confocal image plane taken midway through an NMJ, the COX-2 antibodies (green) bind mostly outside the area stained by anti-synaptotagmin (red), although there are a few places where the two are very close, if not overlapping. The general absence of COX-2 in the nerve terminal can be best appreciated in the full z-stack of confocal images collected at this NMJ (see Supplemental Movie 2). COX-2 was also often observed near the motor axon as it approaches the muscle (see arrow in Fig. 2C). This COX-2 is most likely within the myelinating Schwann cells as it was never observed in the axons back-loaded with Texas Red dextran (see also Fig. 2D below).

To confirm the localization of COX-2 to the periphery of the PSCs as suggested by Fig. 2A, we used YOYO-1 (Invitrogen), a nucleotide stain that visualizes the nuclei and cytoplasm of the PSCs (see Walder et al. 2013). As seen in Fig. 2D (top), COX-2 immunofluorescence (red) overlays YOYO-1 (green) particularly where YOYO-1 reveals the fine processes of the PSCs. Furthermore, as also shown in Fig. 2C, COX-2 is close to but does not significantly overlap the anti-synaptotagmin antibody (white), which labels the presynaptic nerve terminal boutons. Thus, as suggested by the images shown in Fig. 2A, COX-2 is located in the periphery of the PSCs at positions that are in close proximity to the presynaptic nerve terminal. This location of COX-2 can be best appreciated in Supplemental Movie 3, which is a 360° rotation of a three-dimensional surface projection of an NMJ stained with DAPI, YOYO-1, anti-COX-2 and anti-synaptotagmin.

In one additional set of experiments designed to visualize the location of COX-2 relative to the PSCs, we applied an anti-HNK-1 antibody because it binds to Schwann cells (both myelinating and non-myelinating) in this preparation (see Supplemental Fig. 1). As seen in Fig. 2E, COX-2 (green) significantly overlaps with the HNK-1 antigen (red). As the anti-HNK-1 antibody is most probably binding to the extracellular carbohydrate moiety of a membrane-bound glycoprotein (see Discussion), these results further support a localization of COX-2 near the perimeter of the PSCs, just below or within the cell membrane.

As the above experiments were carried out using a primary antibody that was created in rabbit from a 17 amino acid peptide sequence near the C terminus of human/rat/mouse COX-2 (AB5118; Millipore), we checked the specificity of this antibody for lizard COX-2 by performing a Western blot analysis. As displayed in Supplemental Fig. 2, the antibody recognizes a protein in lizard of approximately 71–72 kDa, which corresponds to the expected molecular weight of COX-2 in lizards (http://www.ensembl.org/).

PGE₂-G enhances neurotransmitter release

Given that COX-2 is present at lizard NMJs, especially if pretreated with muscarine (Fig. 2), and given that 2-AG is a modulator at this synapse (Newman et al. 2007), we asked whether PGE₂-G, the product of 2-AG metabolism by COX-2 (Kozak et al. 2002), modifies synaptic transmission. While recording the EPP from a single neuromuscular junction with an intracellular recording electrode, PGE₂-G was locally applied to the junction via pressure ejection from a glass pipette. Application of PGE₂-G caused a large and persistent increase in EPP amplitude (Fig. 3A). To better control the concentration and duration of application, PGE₂-G was dissolved in Ringer solution. Application of PGE₂-G in this way produced a similar increase in synaptic transmission at multiple randomly selected NMJs (Fig. 3B). Bath-applied PGE₂-G more than doubled EPP amplitude (176 ± 14% change from baseline, P= 4.73 × 10⁻¹⁰, n= 10). Note that this increase in EPP amplitude was reversed within minutes of removing PGE₂-G from the bath (Fig. 3B). In contrast, PGE₂, which lacks the glycerol moiety and is not produced by the cyclooxygenation of 2-AG, was without effect (data not shown). On the other hand, PGD₂-G, which is another known product of 2-AG cyclooxygenation, also enhanced EPP amplitude (106 ± 4% change from baseline, P= 1.1 × 10⁻⁸, n= 4), albeit not as much as PGE₂-G (see Fig. 3B).

To determine whether PGE₂-G acts via known prostanoid receptors, we used AH6809, an antagonist at EP₁ and EP₂ receptors. When applied in the presence of AH6809, PGE₂-G still enhanced the EPP amplitude by 168 ± 13% (P= 1.26 × 10⁻⁵, n= 4), an increase that is not significantly different from that induced by PGE₂-G alone (P= 0.76; Fig. 3B). This inability of AH6809 to block the action of PGE₂-G has also been noted in mice hippocampal neurons (Sang et al. 2006, 2007). Following the lead of Silveira et al. (2010), who observed at the frog NMJ that the enhancement of neurotransmitter release by the eCB agonist arachidonyl-2′-chloroethylamide (ACEA) was blocked by the vanilloid receptor antagonist capsazepine, we asked whether capsazepine could similarly block the enhancement of neurotransmitter release by PGE₂-G at the lizard NMJ. As shown in Fig. 3B, capsazepine prevented PGE₂-G from increasing EPP amplitude (5 ± 9% change from baseline, P= 0.31, n= 5). Similarly, capsazepine abolished the effect of PGD₂-G (−1 ± 4% change from baseline, P= 0.62, n= 3). As a control experiment, capsazepine was tested by itself and found to have a small, but statistically insignificant inhibitory effect on EPP amplitude (−13.9 ± 5.4% change from baseline, P= 0.09, n= 4, paired t test).

Lastly, to examine whether the change in EPP amplitude by PGE₂-G was due to a presynaptic increase in ACh release or a postsynaptic change in the sensitivity of the nAChRs, we recorded spontaneous MEPPs (Fig. 3C). As summarized in Fig. 3D, the unitary quantal size (as measured by the MEPP amplitude) did not vary during the application of PGE₂-G (99 ± 6% of baseline, P= 0.90, n= 3; the baseline MEPP amplitude was 0.506 ± 0.045 mV); however, the frequency of MEPPs was significantly increased (198 ± 33% of baseline, P= 0.04, n= 3; the baseline MEPP frequency was 0.449 ± 0.056 Hz). These results demonstrate that PGE₂-G has a presynaptic effect, increasing the quantal content of evoked ACh but not the size of individual quantal units.

Enhancement of neurotransmitter release by PGE₂-G requires NO

Since previous work has shown that the modulation of neurotransmitter release at the lizard NMJ by muscarine depends on NO (Graves et al. 2004), we asked whether the effect of PGE₂-G had a similar requirement for NO. Indeed, application of the NO synthase inhibitor l-NAME prevented PGE₂-G from significantly altering the EPP amplitude (mean EPP amplitude was 94 ± 5% of baseline, P= 0.25, n= 3; Fig. 4A). To demonstrate that this effect of l-NAME was due specifically to the inhibition of NO synthesis, we applied the NO donor DEA-NO in the continued presence of l-NAME and PGE₂-G. In this case, the application of exogenous NO was followed immediately by an increase in EPP amplitude (206 ± 20% of baseline, P= 5.8 × 10⁻³, n= 3; Fig. 4A).

To investigate the role of NO further, we used the NO chelator carboxy-PTIO, which prevents the extracellular accumulation of NO. PGE₂-G had no effect on EPP amplitude in the presence of carboxy-PTIO (mean EPP amplitude was 97 ± 3% of baseline, P= 0.28, n= 3; Fig. 4A). Thus, the enhancement of neurotransmitter release by PGE₂-G requires both the synthesis and the extracellular diffusion of NO.

To determine whether NO was required only during initiation of the PGE₂-G-mediated enhancement or was required throughout, we applied carboxy-PTIO after the EPP amplitude had already been increased by PGE₂-G. An example is shown in Fig. 4B. Within 4 min of adding carboxy-PTIO, in the continued presence of PGE₂-G, the effect of PGE₂-G on EPP amplitude was significantly reduced (28.3 ± 4.6% change from baseline vs. 130.0 ± 10.5% for PGE₂-G alone, P= 0.015, n= 3), indicating that the synaptic enhancement mediated by PGE₂-G requires the continuous presence of NO.

PGE₂-G mediates the muscarine-induced delayed enhancement of neurotransmitter release

Since both PGE₂-G and muscarine require NO to enhance neurotransmitter release (Fig. 4; Graves et al. 2004), and since both the precursor of PGE₂-G (2-AG) and its synthetic enzyme (COX-2) are present at the NMJ (Fig. 2; Newman et al. 2007), we looked for evidence that endogenously produced PGE₂-G is responsible for the delayed enhancement of neurotransmitter release induced by activation of mAChRs. If the muscarinic effect requires the activity of COX-2 to generate PGE₂-G from 2-AG, then inhibition of COX-2 should prevent it. This is exactly what we found. As depicted in Fig. 5A, EPPs were increased after 30 min of muscarine application (92 ± 6%, P= 5.27 × 10⁻⁶, n= 4); however, this increase was not observed in NMJs pretreated with the COX inhibitors DuP 697 (−12 ± 8%, n= 7) or nimesulide (−22 ± 6%, n= 12). The changes in EPP amplitude induced by muscarine in the presence of the COX inhibitors were significantly different from that produced by muscarine alone (P= 9.2 × 10⁻³ for DuP 697 and P= 1.98 × 10⁻⁵ for nimesulide). In fact, inhibition of COX not only prevented the delayed enhancement of EPP amplitude, it unmasked an underlying depression of neurotransmitter release. Although this was not statistically significant in the case of DuP 697, nimesulide and muscarine produced a statistically significant decrease in EPP amplitude compared to baseline (P= 0.019, n= 12). Recalling that muscarine normally produces a biphasic modulation of neurotransmitter release at this synapse (Graves et al. 2004), this residual depression is expected. Inhibition of COX-2 would not only prevent the synthesis of PGE₂-G, it would also prevent the normal metabolism of 2-AG. Thus, in the presence of a potent COX inhibitor such as nimesulide, we would expect that the extended presence of 2-AG would prolong the early CB₁-mediated depression (Newman et al. 2007).

The experiments described so far have shown the following: (1) COX-2 is present in the PSCs of muscles pre-exposed to muscarine, (2) PGE₂-G increases neurotransmitter release in an NO-dependent manner, (3) capsazepine blocks the effect of PGE₂-G and (4) COX inhibitors prevent the muscarine-induced enhancement of ACh release. Taken together, these results suggest that the COX-2-mediated synthesis of PGE₂-G is necessary for the delayed increase in synaptic transmission following muscarine activation at the lizard NMJ. To test this final prediction, we compared the application of muscarine per se to the application of muscarine along with capsazepine. As shown in Fig. 5A, in the presence of capsazepine, muscarine had no effect (mean EPP amplitude −5 ± 4% of baseline, P= 0.27, n= 4). A one-way ANOVA comparing the results with muscarine and muscarine along with capsazepine indicated the difference was highly significant (P= 1.21 × 10⁻⁵). Additionally, we repeated the experiment with a small modification. In this case, capsazepine was applied after the EPP amplitude had been enhanced by muscarine. An example is shown in Fig. 5B. Even with the continued presence of muscarine, capsazepine significantly decreased the EPP amplitude within 6 min of application (18.3 ± 13.9% change from baseline for the combination of muscarine and capsezapine vs. 85.7 ± 1.5% for muscarine alone, P= 0.016, n= 3). Thus, the delayed muscarine-induced enhancement of neurotransmitter release requires the continuous activation of the PGE₂-G receptor, which is consistent with the result presented in Fig. 3B showing that the PGE₂-G enhancement is quickly reversed by its washout from the bath.

Discussion

The results we report here and elsewhere (Graves et al. 2004; Newman et al. 2007) can be summarized as follows. The activation of mAChRs induces the synthesis of 2-AG, which is released from the muscle into the synaptic cleft via an eCB transporter (Newman et al. 2007). Initially, 2-AG inhibits the evoked release of neurotransmitter (ACh) via the activation of CB₁ receptors located on the presynaptic nerve terminal (Newman et al. 2007). Based on the current work, we propose that 2-AG is subsequently converted to PGE₂-G by the enzyme COX-2 and that PGE₂-G increases neurotransmitter release by activating a capsazepine-sensitive receptor. This latter process accounts for the previously observed delayed muscarine-induced enhancement of neurotransmitter release (Graves et al. 2004).

The identity, localization and regulation of COX-2 at the NMJ

Cyclooxygenase exists in at least two isoforms, COX-1 and COX-2, although a splice variant of COX-1, called COX-1b, has been detected (Chandrasekharan et al. 2002). These isoforms are similar in structure and catalytic activity, with COX-1 generally regarded as constitutively expressed and COX-2 as rapidly inducible (Reddy & Herschman, 1994; however, see Funk, 2001).

Prior to the work reported here, there have been only a few reports suggesting that cyclooxygenase or prostaglandins may be involved in the modulation of neurotransmitter release at the vertebrate NMJ. Cyclooxygenase activity and/or exogenous PGE₂ have been observed to decrease (Arkhipova et al. 2006), increase (Pinard & Robitaille, 2008), and both increase and decrease (Madden & Van der Kloot, 1982, 1985) neurotransmitter release. These results, along with those presented in this paper, are consistent with either isoform of COX being responsible for the modulation of ACh release at the NMJ as COX inhibitors, such as DuP 697 and nimesulide, inhibit both isoforms at the concentrations applied (Riendeau et al. 1997). Although our immunofluorescence experiments (Fig. 2) suggest that COX-2 is the active isoform, further work is necessary to confirm this.

In our proposed model, the cyclooxygenation of 2-AG occurs within the PSCs. We propose this location based on our immunofluorescence experiments, specifically: (1) the position of COX-2 immediately outside the rings of nAChRs that decorate the ridges formed by the large post-junctional folds (Fig. 2A), (2) the minimal overlap of COX-2 and markers of the nerve terminal (Fig. 2B–D), (3) the location of COX-2 relative to the PSC nuclei and peri-nuclear RNA (Fig. 2D) and (4) the extensive overlap of COX-2 and a marker of the PSCs (Fig. 2E). In the latter case, the marker used, anti-HNK-1 antibody, labels the extracellular surface of the PSCs, suggesting that COX-2 is located just beneath the cell membrane. If so, this distribution of COX-2 in glial cells at the NMJ is different from its more general localization to perinuclear membranes in most mammalian cells (Ueno et al. 2005). COX-2, however, has been localized to other parts of the cell, including the endoplasmic reticulum (Spencer et al. 1998), mitochondria (Liou et al. 2005) and the cell membrane (Liou et al. 2001; Perrone et al. 2007). Our data are most consistent with a location near the PSC plasma membrane at the NMJ. Its apparent location in the periphery of PSC processes that are closely opposed to the presynaptic nerve terminal would be an optimal site for the rapid metabolism of 2-AG and the release of reaction product, PGE₂-G, into the synaptic cleft where that effector could then act on the nerve terminal.

We speculate that COX-2 is regulated at the level of gene transcription, with the activation of M₁ receptors on the PSCs leading to the induction of the gene for COX-2. Although we do not have quantitative evidence that such regulation occurs at the NMJ, it is supported qualitatively by our observation that incubating the muscle in muscarine for 1 h greatly increased COX-2 immunofluorescence compared to controls. Furthermore, regulation of gene expression is the primary control mechanism for COX-2 in other cells and tissues studied (see Reddy & Herschman, 1994).

The muscarine-induced enhancement of ACh release at the NMJ is delayed by at least 30 min and persists for hours (Graves et al. 2004). It is noteworthy that COX-2 protein levels in human colon cells, as quantified by Western blot analysis, were noticeably increased as early as 30 min and were maximal by 3 h of treatment with carbachol (Yang & Frucht, 2000). Thus, there is precedence for the up-regulation of COX-2 by the activation of M₁ and/or M₃ mAChRs with a time course that approximates the switch from the initial reduction of evoked ACh release to its enhancement at the NMJ (Graves et al. 2004). Work is currently underway to directly test this hypothesis using quantitative real-time PCR.

A word of explanation is in order regarding our use of anti-HNK-1 antibody to label Schwann cells at the NMJ since this is a novel use of the antibody. The HNK-1 monoclonal antibody was created against an immunogen in a homogenate of cat primary visual cortex (Arimatsu et al. 1987). Subsequent work has shown that the epitope specifically recognized by this antibody is an N-linked carbohydrate found in several glycoproteins, including members of the N-CAM adhesion molecule family and myelin-associated glycoprotein (Naegele & Barnstable, 1991). In our experience using it on the lizard NMJ, it acts as a very specific and sensitive marker of both myelinating and non-myelinating Schwann cells (see Supplemental Fig. 1).

NO is required for automodulation at the neuromuscular junction

We have shown previously that although chelating NO or inhibiting its synthesis does not alter synaptic transmission by itself, it does prevent muscarinic or CB₁ receptor agonists from modulating neurotransmitter release, and that exogenous NO restores this modulation even though exogenous NO by itself has no effect (Graves et al. 2004; Newman et al. 2007). Collectively, these results indicate that NO is necessary but not sufficient for modulating neurotransmitter release at the NMJ. A similar co-involvement of NO and eCBs in the depression of neurotransmitter release has been revealed in synapses on CA1 pyramidal cells in the mouse hippocampus (Makara et al. 2007). In both cases, NO acts via the standard guanylate cyclase/cGMP/protein kinase G pathway. The present work shows that the PGE₂-G enhancement of ACh release at the NMJ also has a co-requirement for NO.

Although we have demonstrated previously that NO synthase (NOS) is present in all three compartments of the NMJ – the nerve terminal, the PSCs and the muscle (Graves et al. 2004) – we still do not know which of these sources is critical for automodulation, nor do we know what activates NOS under these conditions. Recent studies at the NMJ of the frog (Pinard & Robitaille, 2008) and toad (Etherington & Everett, 2004) have implicated the NOS at the muscle end-plate in the modulation of neurotransmitter release. N-Methyl-d-aspartate (NMDA) receptors have been identified at the muscle end-plate (Berger et al. 1995; Mays et al. 2009; Walder et al. 2013) and these could provide a source of Ca²⁺ needed to activate NOS (Bredt & Snyder, 1990). Indeed, NOS has been shown to co-localize with NMDA receptors via the dystrophin–glycoprotein complex at the NMJs of rat and mouse skeletal muscle (Grozdanovic & Gossrau, 1998). Interestingly, levels of NOS-I are significantly reduced in the junctional sarcolemma of muscles from patients with Duchenne muscular dystrophy, in whom the protein dystrophin is mutated (Brenman et al. 1995).

Despite a potentially prominent role for NMDA receptors in activating NO synthesis at the NMJ, the source of the endogenous NMDA agonist is unknown. Glutamate is a likely candidate and has long been known to be present at the NMJ, in both the nerve terminals and PSCs (Waerhaug & Ottersen, 1993). However, the mechanism by which glutamate might be released into the synaptic cleft is unclear. Pinard and Robitaille (2008) make a strong argument that glutamate is released from the PSCs in a frequency-dependent manner, but they also concede that glutamate may be released from the nerve terminals.

The discovery of the dipeptide N-acetylasparty lglutamate (NAAG) along with its hydrolytic enzyme, glutamate carboxypeptidase-II (GCP-II), at the vertebrate NMJ (Berger et al. 1995; Walder et al. 2013) suggests a third possibility. We recently showed that NAAG is released from lizard motor nerve terminals during high-potassium depolarization or electrical stimulation of the motor nerve (Walder et al. 2013). GCP-II, which is present on the extracellular surface of the PSCs (Walder et al. 2013), would be expected to hydrolyse released NAAG to N-acetylaspartate and glutamate. Glutamate produced in this way could stimulate NO synthesis by activating the NMDA receptor at the muscle end-plate. More work is needed to explore this novel suggestion.

Is PGE₂-G an endogenous modulator at the NMJ?

Although the requirement for COX-2 in the muscarine-induced enhancement of neurotransmitter release is very clear, the evidence that PGE₂-G is the sole or primary product of COX-2 responsible for synaptic enhancement has less support. The evidence for this proposition comes from our observations that: 2-AG is present at the NMJ (Newman et al. 2007), PGE₂-G mimics the delayed enhancement (Fig. 3) and its inhibitor, capsazepine, blocks the muscarine-induced enhancement (Fig. 5). However, it is possible that COX-2 produces other signalling molecules that enhance neurotransmitter release in a capsazepine-dependent manner. In fact, there are several other known products of the cyclooxygenation of 2-AG, namely PGI₂-G, PGD₂-G, PGF_2α-G and TXA₂-G (Yang & Chen, 2008), that are also plausible candidates. Indeed, we have shown that PGD₂-G has similar effects to PGE₂-G, although not as large (Fig. 3B). Interestingly, in our experiments, PGE₂ was without effect, suggesting that the glycerol moiety is necessary. It is also possible that 2-AG is not the only substrate for COX-2 at the NMJ, opening up the range of possible candidates even further. The identity of the actual product(s) generated cannot be resolved with an electrophysiological/pharmacological approach, but will require chemical analysis (as in Hu et al. 2008).

Interestingly, if PGE₂-G is the sole signalling molecule responsible for the delayed muscarine-induced enhancement, this raises the question as to the source of 2-AG. Since COX-2 is located in the PSCs, the 2-AG must either be transported into the PSCs after being released into the synaptic cleft from the muscle or it must be synthesized separately in the PSC. The observation that the delayed muscarine-induced enhancement of neurotransmitter release is not prevented by blocking M₃ receptors (Graves et al. 2004), which are responsible for the synthesis and release of 2-AG from the muscle (Newman et al. 2007), supports the latter suggestion. However, it is also possible that blocking M₃ receptors reduces 2-AG to a level below that required to produce observable depression but sufficient to serve as a substrate for PGE₂-G production. Further experiments are needed to determine which pool of 2-AG is actually used for the synthesis of PGE₂-G.

The PGE₂-G receptor

It was recently shown that application of either the vanilloid agonist arachidonyl-2′-chloroethylamide (ACEA) or capsaicin increases quantal content at the frog NMJ and this could be blocked by the transient receptor potential vanilloid 1 (TRPV1) antagonist capsazepine (Silveira et al. 2010). While our results add further evidence of a capsazepine-sensitive receptor at the NMJ, we are unwilling to conclude that this is a TRPV1 receptor (for a contrasting viewpoint, see Silveira et al. 2010). First, capsazepine blocks not only TRPV1 but also transient receptor potential melastatin 8 (TRPM8) channels in mammals (Behrendt et al. 2004; Weil et al. 2005; Xu et al. 2005) and both TRPV1 and TRPM8 mRNA have been detected in peripheral muscle in reptiles (Seebacher & Murray, 2007). Secondly, the sensitivity of neurotransmitter release at the NMJ to capsaicin, which was the main criterion used by Silveira et al. (2010), is of questionable utility in the lizard since the sensitivity of the TRPV1 channel to capsaicin is believed to be limited to mammalian herbivores (Jordt & Julius, 2002). Lastly, although PGE₂-G has been shown by others to act independently of known prostanoid receptors (Nirodi et al. 2004; Sang et al. 2006; Hu et al. 2008), there have been no studies to date identifying its endogenous receptor. It is noteworthy that PGE2-G has been shown to mobilize intracellular calcium in a murine macrophage-like cell line (Nirodi et al. 2004). If a similar signalling pathway exists in nerve terminals at the lizard NMJ, the increased free Ca²⁺ could account for the observed enhancement of neurotransmitter release. Considerably more work is needed to clarify the pharmacological and cell physiological effects of PGE₂-G at the lizard NMJ and elsewhere.

Is the vertebrate NMJ a tripartite synapse?

Glial cells have been known to function as active signalling elements at synapses in the CNS for over two decades, leading one group to coin the term ‘tripartite synapse’ to refer to the presynaptic terminal, the postsynaptic terminal and the glial cells surrounding the synapse (Araque et al. 1999). Early evidence suggesting that PSCs play a similar role at the NMJ came from the observation that, just like their counterparts in the CNS, activation of neurotransmitter release leads to an increase in intracellular free Ca²⁺ concentration in the PSCs. This has been reported for NMJs in frog (Jahromi et al. 1992; Reist & Smith, 1992), lizard (Lindgren & Haydon, 1999) and mouse (Rochon et al. 2001). Direct evidence that PSCs play a role in synaptic plasticity was provided by Robitaille (1998), who found that short-term synaptic depression depended on the activation of G proteins in the PSCs at frog NMJs. Work from the same lab also revealed that Ca²⁺ signals in PSCs influence synaptic plasticity at the mouse NMJ (Todd et al. 2010). In contrast to these results, Reddy et al. (2003) claimed that the ablation of PSCs at the frog NMJ by application of a monoclonal antibody specific for PSCs together with complement (in guinea pig serum) failed to alter short-term synaptic depression within 5 h of ablation.

By demonstrating a requirement for COX-2 in the delayed synaptic enhancement mediated by muscarinic receptors, along with the evidence that COX-2 is localized to the PSCs, the results presented in this paper support the suggestion that, like central synapses, the NMJ is a tripartite synapse.

A proposed physiological function for COX-2 at the NMJ

The purpose of neuromuscular transmission in vertebrate animals is to ensure reliable conversion of action potentials in the motor nerve to physical contraction of innervated muscle fibres. Thus, any mechanism that improves the fidelity of that conversion will benefit the organism. This fidelity is regularly challenged during prolonged muscle activity (e.g. during exercise) when it becomes difficult to sustain high levels of neurotransmitter (i.e. ACh) release. We hypothesize that under such conditions, the accumulation of ACh in the synaptic cleft, and possibly even its overflow out of the cleft, leads to the activation of mAChRs.

The data presented here, along with previous work (Graves et al. 2004; Newman et al. 2007) reveal a surprisingly complicated scheme by which the activation of mAChRs modulates the release of neurotransmitter at the NMJ. The exact physiological conditions under which these modulatory processes come into play is not known. However, there is evidence for long-term presynaptic modulation at the NMJ following 20 min of continuous 1 Hz stimulation (Etherington & Everett, 2004; Newman et al. 2007) and also following 5–7 days of intermittent periods of 10 Hz stimulation (Hinz & Wernig, 1988; Bélair et al. 2005). In the latter case, not only was baseline neurotransmitter release decreased (approximately 50%), but the NMJs were more resistant to high-frequency synaptic depression (Bélair et al. 2005).

The above observations along with those presented in this paper lead us to speculate as to the benefit of mAChR-mediated synaptic modulation at the NMJ during times of intense and/or long-term synaptic activity. Initially, the activation of M₃ mAChRs induces the synthesis and release of the eCB 2-AG, which reduces evoked ACh release. Since the NMJ normally releases 2–4 times the amount of ACh needed to successfully convert a motor nerve action potential to a muscle fibre twitch (known as ‘safety factor’, see Wood & Slater, 2001), the release of less ACh per action potential will enhance neuromuscular endurance as long as the reduction of ACh release does not exceed the safety factor. It is noteworthy in this regard that the application of maximal concentrations of either muscarinic or CB₁ agonists never reduces ACh release by more than 50%.

Following this initial ‘ACh conserving’ reduction in neurotransmitter release, we hypothesize that sustained (≥30 min) high levels of activity trigger the second phase of modulation mediated by M₁ mAChRs and the conversion of 2-AG to PGE₂-G by COX-2. Whilst we observed levels of neurotransmitter release that were more than twice normal levels following the application of PGE₂-G (Fig. 3), under the physiological conditions during which this mechanism would be invoked (i.e. at least 30 min of intense activity) it is likely that the motor nerve endings are being challenged to release sufficient ACh to activate contraction of the muscle fibres. The production of PGE₂-G under these extreme conditions may increase ACh release just enough to prevent catastrophic failure.

Further work is needed to test the above scenarios and verify the more speculative aspects of our model. However, even at the current stage of investigation, it is obvious that the modulation of synaptic transmission at the NMJ shares many similarities with synaptic modulation at synapses in the CNS, including the hippocampus. Thus, learning more about the role and mechanism of membrane-derived lipids in synaptic modulation at the relatively simple and highly accessible NMJ promises to provide insights relevant to synapses in the CNS.

Glossary

ACh

acetylcholine

2-AG

2-arachidonylglycerol

α-BTX

α-bungarotoxin

CB₁

cannabinoid type 1

COX

cyclooxygenase

DIC

differential interference contrast

dTC

d-tubocurarine chloride

eCB

endocannabinoid

EPP

end-plate potential

GCP

glutamate carboxypeptidase

l-NAME

N^G-nitro-l-arginine methyl ester

MEPP

miniature end-plate potential

mAChR

muscarinic acetylcholine receptor

NAAG

N-acetylaspartylglutamate

nAChR

nicotinic acetylcholine receptor

NMDA

N-methyl-d-aspartate

NMJ

neuromuscular junction

NO

nitric oxide

NOS

nitric oxide synthase

PSC

perisynaptic Schwann cell

PGD₂-G

prostaglandin D2 glycerol ester

PGE₂-G

prostaglandin E2 glycerol ester

Additional information

Competing interests

None.

Author contributions

C.A.L. conceived and designed the experiments, collected, analysed and interpreted data, and drafted and revised the article. Z.L.N. conceived and designed experiments, collected, analysed and interpreted data, and revised the article for important critical content. S.R. collected, analysed and interpreted data, and revised the article for important critical content. J.J.M., K.A.B. and Z.S. collected, analysed and interpreted data. All authors read and approved the final version of the manuscript.

Funding

This work was supported by NIH grant 1R15NS072735.

Supplementary material

Supplemental Fig. 1

Supplemental Fig. 2

Supplemental Movie 1

Supplemental Movie 2

Supplemental Movie 3