NMDA receptor-dependent presynaptic inhibition at the calyx of Held synapse of rat pups

N-Methyl-d-aspartate receptors (NMDARs) play diverse roles in synaptic transmission, synaptic plasticity, neuronal development and neurological diseases. In addition to their postsynaptic expression, NMDARs are also expressed in presynaptic terminals at some central synapses, and their activation modulates transmitter release. However, the regulatory mechanisms of NMDAR-dependent synaptic transmission remain largely unknown. In the present study, we demonstrated that activation of NMDARs in a nerve terminal at a central glutamatergic synapse inhibits presynaptic Ca2+ currents (ICa) in a GluN2C/2D subunit-dependent manner, thereby decreasing nerve-evoked excitatory postsynaptic currents. Neither presynaptically loaded fast Ca2+ chelator BAPTA nor non-hydrolysable GTP analogue GTPγS affected NMDAR-mediated ICa inhibition. In the presence of a glutamate uptake blocker, the decline in ICa amplitude evoked by repetitive depolarizing pulses at 20 Hz was attenuated by an NMDAR competitive antagonist, suggesting that endogenous glutamate has a potential to activate presynaptic NMDARs. Moreover, NMDA-induced inward currents at a negative holding potential (−80 mV) were abolished by intra-terminal loading of the NMDAR open channel blocker MK-801, indicating functional expression of presynaptic NMDARs. We conclude that presynaptic NMDARs can attenuate glutamate release by inhibiting voltage-gated Ca2+ channels at a relay synapse in the immature rat auditory brainstem.

Despite these findings, recent studies have challenged the existence of axonal/presynaptic NMDA receptors. In a previous study, focal iontophoretic application of the NMDAR agonist L-aspartate onto the axons of cerebellar stellate cells failed to elicit Ca 2þ transients in axonal varicosities. However, L-aspartate application onto the dendrites of these cells elicited Ca 2þ transients in axonal varicosities via opening of voltage-gated Ca 2þ channels (VGCCs) triggered by passive propagation of depolarization from somatodendritic sites down along axons [41]. Subsequent studies by the same research group revealed no evidence of functional NMDAR expression in the axons of L5 pyramidal cells in the visual cortex [42], or basket cells in the cerebellum [43]. Given such controversial findings, presynaptic recordings should be used to explore whether other types of cells in the CNS exhibit axonal/presynaptic NMDAR expression.
At the calyx of Held synapse in the rat auditory brainstem, whose presynaptic structure is large enough to enable direct whole-cell patch-clamp recordings, application of exogenous L-glutamate inhibits nerve-evoked release of the endogenous neurotransmitter glutamate [44]. This presynaptic inhibitory action is mediated by metabotropic glutamate receptors (mGluRs) [45] and also by a-amino-3-hydroxy-5-methyl-4isoxazolepropionic acid receptors (AMPARs) [44] expressed in the presynaptic terminal. Both types of glutamate receptors inhibit VGCCs via the activation of heterotrimeric G proteins. However, mGluR and AMPA/kainate receptor antagonists only partially impair the inhibitory effect of L-glutamate on presynaptic Ca 2þ currents (I Ca ), suggesting the involvement of additional mechanisms. In the present study, we show that activation of presynaptic NMDARs induces inward currents at a negative holding potential, inhibits I Ca and decreases action potential-dependent excitatory postsynaptic currents (EPSCs) at the calyx of Held synapse in the immature rat brainstem.

Electrophysiological recordings
Whole-cell patch-clamp recordings were made from presynaptic calyceal nerve terminals or postsynaptic MNTB principal neurons. For recording I Ca and I Ba , calyces were voltage-clamped at a holding potential of -80 mV, and depolarizing voltage steps (duration: 3 ms) were applied every 20 s. The I Ca amplitude was measured 2-3 ms after the onset of depolarizing pulses. For recording evoked EPSCs, MNTB neurons were voltage-clamped at a holding potential of 270 mV, and presynaptic axons were stimulated every 20 s by using a tungsten bipolar electrode positioned halfway between the midline and the MNTB. The EPSC amplitude was measured at their peaks. For presynaptic recordings, the electrode resistance was 4-8 MV, and the access resistance was 5-18 MV with its compensation by 80%. Leak currents in presynaptic recordings were subtracted by the scaled pulse (P/8) protocol. For postsynaptic recordings, the electrode resistance was 2.5-4 MV, and access resistance was 5-15 MV with its compensation by 70%. Voltage-clamp recordings were made using a patch-clamp amplifier (Axopatch-200B, Axon Instruments, Foster City, CA, USA). Current-clamp recordings of presynaptic action potentials were made using another patch-clamp amplifier (MultiClamp 700A, Axon Instruments) equipped with a high input impedance (10 11 V) voltage follower. Recorded signals were low-pass-filtered at 5 kHz and digitized at 20 -50 kHz by an analogue -digital converter (Digidata 1322A, Axon Instruments) with pCLAMP 9 software (Axon Instruments). Liquid -junction potentials between the pipette solutions and the aCSF were not corrected for. Drugs were bath-applied by switching superfusates using a peristaltic pump or a gravity-fed perfusion system ( perfusion rate, 2.0-6.0 ml min 21 ). Experiments were carried out at room temperature (21 -258C).

Statistical analysis
Data are presented as mean + s.e.m.. For comparison of paired data from one group, we first used Shapiro-Wilk normality test, then employed Student's paired t-test. For comparison of data from the control groups and groups with various kinds of manipulations such as extracellular/intracellular application of NMDAR agonists/antagonists, we first used Shapiro-Wilk normality test, then employed Student's unpaired t-test. Since all the data in this study passed the normality test, nonparametric statistical analysis was not necessary. Unless otherwise described, Student's unpaired t-test was employed. Statistical significance was considered when p-value was less than 0.05 (SIGMAPLOT 12.0, Systat Software Inc., San Jose, CA, USA), and significance level is denoted using asterisks (*p , 0.05, **p , 0.01 and ***p , 0.001).

Inhibitory effect of NMDA on presynaptic Ca 2þ currents (I Ca )
Previous studies have revealed that activation of mGluRs [45] or AMPARs [44] in the calyx of Held presynaptic terminal inhibits I Ca . First, we investigated whether NMDA also inhibits I Ca . As illustrated in figure 1a, bath-application of NMDA (50 mM) in Mg 2þ -free aCSF inhibited I Ca by 8.3 + 2.7% at 0 mV (n ¼ 5) without a clear shift in the current-voltage relationship (figure 1b). The inhibitory effect of NMDA on I Ca was concentration-dependent with an IC 50 of 135 mM (figure 1c). Based on this result, we used a relatively high concentration of NMDA (500 mM) for controls in order to securely analyse the property of NMDA-induced I Ca inhibition. As illustrated in figure 1d, bath-application of NMDA at this concentration in Mg 2þ -free aCSF more evidently inhibited I Ca by 26.0 + 3.7% at 0 mV (n ¼ 5), again without a clear shift in the current-voltage relationship (figure 1e). The NMDAR competitive antagonist D-AP5 (500 mM) abolished this NMDA effect (to 1.2 + 0.9%, n ¼ 5, ***p , 0.001; figure 1f,i).
These results indicate that NMDA-induced I Ca inhibition was indeed mediated by NMDARs. To exclude the possibility that NMDARs expressed in surrounding neurons and/or glia might mediate this NMDA effect, we loaded the NMDAR open channel blocker MK-801 (500 mM) directly into the calyceal nerve terminal through a whole-cell patch pipette.
We set the concentration of intra-terminal MK-801 to a range of sub-millimolar based on the procedure used in a previous study, in which MK-801 (1 mM) was applied into presynaptic neurons through the patch pipettes [38]. Under this condition, intra-terminal MK-801 nearly abolished NMDA-induced I Ca inhibition (to 7.3 + 2.4%, n ¼ 4, **p , 0.01; figure 1g,i), suggesting that functional NMDARs are expressed in calyceal nerve terminals, and that their activation inhibits presynaptic VGCCs.

G proteins and Ca 2þ are dispensable for NMDARmediated I Ca inhibition
At the calyx of Held, a variety of presynaptic receptors are coupled to the heterotrimeric G proteins, and direct interaction of Gbg subunits with presynaptic VGCCs inhibits I Ca as shown for mGluRs [ [52] and AMPARs [44]. To investigate the possibility that this mechanism also underlies NMDA-induced I Ca inhibition, we loaded the non-hydrolysable GTP analogue GTPgS (0.2 mM) into the presynaptic terminal through whole-cell patch pipettes. As GTPgS diffused into a terminal from a presynaptic pipette, I Ca became smaller in amplitude and slower in rise time, consistent with previous studies [48,53]. After the I Ca amplitude had reached a steady level, bath-application of NMDA (500 mM) attenuated I Ca (figure 3a,e) by 27.5 + 4.2% (n ¼ 5). This magnitude of inhibition in the presence of intra-terminal GTPgS was similar to that observed in its absence ( p ¼ 0.784), suggesting that NMDAR-mediated I Ca inhibition does not require G proteins. We then sought to confirm that the lack of occlusive effect of intra-terminal GTPgS on NMDA-induced I Ca inhibition was not due to a failure of drug action. Following bath-application of the high affinity group III mGluR agonist L-AP4 (100 mM), significant differences in the magnitude of I Ca inhibition were observed between the absence (22.5 + 1.6%, n ¼ 5) and presence (2.2 + 1.3%, n ¼ 5, ***p , 0.001, data not shown) of intraterminal GTPgS (0.2 mM). Thus, intra-terminal GTPgS securely occluded mGluR-mediated I Ca inhibition. After intra-terminal GTPgS (0.2 mM) had fully activated the mGluR-and AMPAR-mediated I Ca inhibition pathways, we examined whether L-glutamate further inhibits I Ca via the activation of presynaptic NMDARs. As shown in figure 3b, bath-application of L-glutamate (500 mM) inhibited  rsob.royalsocietypublishing.org Open Biol. 7: 170032 I Ca by 19.4 + 0.6% (n ¼ 4), and this effect was lessened to 6.1 + 0.9% (n ¼ 4, *p , 0.05) by a mixture of D-AP5 (500 mM) and 7-ClK (100 mM). These results suggest that presynaptic NMDARs mainly mediate L-glutamate-induced additional I Ca inhibition after full activation of mGluRs and AMPARs [44]. We then examined whether intra-terminal Ca 2þ , which is elevated by presynaptic NMDAR activation, mediates NMDA-induced I Ca inhibition. The fast Ca 2þ chelator BAPTA (10 mM) loaded into the calyceal terminal had no effect on NMDA-induced I Ca inhibition (31.1 + 1.8%, n ¼ 4, p ¼ 0.290; figure 3c,e). Moreover, replacement of the VGCC charge carrier Ca 2þ with Ba 2þ (2 mM) had no significant effect on the NMDA-induced inhibition of presynaptic Ba 2þ currents (21.2 + 1.0%, n ¼ 5, p ¼ 0.254; figure 3d,e). These results suggest that intra-terminal Ca 2þ does not contribute to NMDA-induced I Ca inhibition.
We also examined a possible effect of Na þ on the I Ca inhibition. When extracellular Na þ was replaced with equimolar TEA þ , bath-application of NMDA (500 mM) no longer inhibited I Ca (3.5 + 1.0%, n ¼ 4; **p , 0.01; figure 3e), suggesting that Na þ influx through presynaptic NMDARs may somehow mediate the I Ca inhibition.
We further aimed to identify the intracellular mechanism(s) that links NMDAR activation and Ca 2þ channel inhibition in the calyceal terminal. Since presynaptic NMDARs are relevant to nitric oxide synthesis in the cerebellum [54,55] as well as protein kinase C activation in the neocortex [56], we examined whether such chemicals as the nitric oxide synthase inhibitor L-NNA (1 mM) or the protein kinase C inhibitor staurosporine (2 mM) attenuate I Ca inhibition induced by NMDA (500 mM). However, neither agent exerted a significant effect (32.1 + 3.1%, n ¼ 4, p ¼ 0.259 for L-NNA; 23.1 + 2.3%, n ¼ 4, p ¼ 0.560 for staurosporine, figure 3e). Moreover, we examined whether endocannabinoid signalling triggered via the activation of NMDARs in the postsynaptic MNTB neuron is associated with NMDA-induced I Ca inhibition. Based on the protocol in a previous study [57], we performed these experiments using the cannabinoid receptor type 1 blocker AM 251 (5 mM). In the presence of AM 251 in aCSF, NMDA application still inhibited I Ca by 23.6 + 3.0% (n ¼ 4, p ¼ 0.473, figure 3e), suggesting that endocannabinoid-dependent retrograde signalling is not involved in NMDA-induced I Ca inhibition.

NMDA-induced currents in the calyceal nerve terminal
We examined whether NMDARs expressed in the calyceal nerve terminal exhibit ionotropic channel properties. In Mg 2þ -free aCSF containing TTX (1 mM), at a holding potential of -80 mV, bath-application of NMDA (500 mM) induced inward currents (35.5 + 9.9 pA, n ¼ 7, figure 4a), which were accompanied by an increase in membrane noise (figure 4a). Intra-terminal MK-801 (500 mM) significantly reduced these NMDA currents to 8.9 + 2.2 pA (n ¼ 4; figure 4b). After maximal blockade of presynaptic K þ channels by using a Cs þ -based pipette solution containing TEA (10 mM) and also replacement of extracellular Ca 2þ with equimolar Ba 2þ (2 mM), bath-application of NMDA (500 mM) still induced inward currents (17.3 + 6.9 pA, n ¼ 5; figure 4c), which were again accompanied by an increase in membrane noise (figure 4c). Intra-terminal MK-801 (500 mM) had no effect on the presynaptic resting membrane potential (260.6 + 1.2 mV for control, 261.8 + 1.5 mV for MK-801, n ¼ 5 for each, p ¼ 0.471), peak amplitude (95.6 + 2.0 mV for control, 96.8 + 2.1 mV for MK-801, p ¼ 0.655) or half-width (0.49 + 0.04 ms for control, 0.47 + 0.07 ms for MK-801, p ¼ 0.835) of presynaptic action potentials, suggesting that its blocking effect is specific for NMDARs. These results indicate that functional NMDARs are expressed in the calyceal nerve terminals. The reduction in the inward current amplitude following blockade of K þ channels and small currents remaining in the presence of intra-terminal MK-801 (figure 4b) imply that activation of NMDARs in surrounding cells might additionally contribute to NMDA-induced inward currents via an increase in extracellular K þ concentration.

NMDAR-mediated I Ca inhibition by endogenous glutamate
We then investigated whether endogenous glutamate inhibits I Ca via the activation of presynaptic NMDARs. When I Ca were evoked by a train of 30 depolarizing pulses (to 0 mV, duration: 1 ms) at 20 Hz, the I Ca amplitude displayed an activity-dependent decline, and reached a steady-state (ss) level lower than that of the first I Ca (I ss /I 1st : 92.9 + 4.0%, n ¼ 5). Previous research reported that elevated extracellular glutamate by repetitive depolarizing stimuli activates presynaptic mGluRs, thereby reducing glutamate release [58]. This mechanism may have partly contributed to the I Ca decline that we observed in this experiment. In the presence of D-AP5 (500 mM) in aCSF, no changes in this I Ca decline were observed (I ss /I 1st : 92.6 + 4.3%, n ¼ 5, rsob.royalsocietypublishing.org Open Biol. 7: 170032 figure 5a). In contrast, when I Ca were evoked by a train at a higher-frequency of 200 Hz, the I Ca amplitude displayed activity-dependent facilitation as previously reported for rats [59] and for mice [60]. Under this condition, D-AP5 again failed to alter the magnitude of facilitation (I ss /I 1st : 117.0 + 2.2% for control, 118.6 + 1.6% for D-AP5, n ¼ 5, p ¼ 0.536, Student's paired t-test, figure 5b). Further, to clarify whether endogenous glutamate activates presynaptic NMDARs, I Ca were evoked by a train at 20 Hz in the presence of the glutamate uptake blocker TBOA (100 mM) in aCSF. As observed in the absence of TBOA, the I Ca amplitude displayed activity-dependent decline and reached a steady-state level lower than that of the first I Ca (I ss /I 1st : 86.4 + 2.8%, n ¼ 7). Previous studies already revealed that not only repetitive depolarizing stimuli [58] but also glutamate transporter blockade by TBOA [61] raises extracellular glutamate, thereby activating presynaptic mGluR-dependent autoinhibition of glutamate release. Both mechanisms may have partly contributed to the I Ca decline that we observed with this protocol. Under this condition, bath-application of D-AP5 (500 mM) significantly weakened this magnitude of I Ca decline (I ss /I 1st : 90.3 + 1.6%, p , 0.05, figure 5c). These results imply that endogenous glutamate released from the nerve terminal or surrounding cells [14,62] inhibits I Ca via the activation of presynaptic NMDARs. Furthermore, these results suggest that NMDAR-mediated presynaptic inhibition may not occur under physiological conditions, in which extracellular glutamate is promptly cleared by glial uptake systems. However, this inhibition may occur under pathological conditions, in which extracellular glutamate concentration rises to a high level.  6c). In the paired-pulse stimulation protocol with an inter-pulse interval of 20 ms, NMDA (500 mM) increased the paired-pulse ratio (PPR, the ratio of second amplitude to the first) of AMPA-EPSCs by 28.7 + 2.6% (n ¼ 7, ***p , 0.001, Student's paired t-test; figure 6d). Thus, this finding confirmed that NMDA application indeed decreases evoked AMPA-EPSCs by means of a presynaptic mechanism. Moreover, we examined whether tonic activation of presynaptic NMDAR by endogenous glutamate alters basic synaptic transmission. Bath-application of the NMDAR competitive blocker D-AP5 (500 mM) altered neither amplitude (100.5 + 1.8% of control, n ¼ 4, p ¼ 0.829, figure 6e) nor PPR (96.9 + 1.9% of control, n ¼ 4, p ¼ 0.252, Student's paired t-test, figure 6f ) of evoked AMPA-EPSCs, suggesting that presynaptic NMDARs are not tonically activated to reduce action potential-dependent release.

Discussion
In the present study, we demonstrated that activation of GluN2C/2D subunit-containing presynaptic NMDARs inhibits VGCCs, thereby attenuating action potential-driven release at a central glutamatergic synapse of young rats. Furthermore, we successfully recorded NMDA-induced currents using presynaptic voltage-clamp recordings, confirming functional expression of presynaptic NMDARs at this synapse.
Presynaptic inhibitory effects of NMDAR activation on nerve-evoked synaptic currents were reported at inhibitory synapses in the cerebellum [26,28] as well as excitatory synapses in the spinal cord primary afferents [36]. These studies reported a weak blocking effect of Mg 2þ on NMDAR-mediated presynaptic inhibition, indicating that GluN2C/2D ( preferentially GluN2D) subunits may be involved. In the present study, we also observed a weak blocking effect of Mg 2þ on NMDAR-mediated I Ca inhibition ( figure 1g,h). Further, we confirmed that the GluN2C/2D selective antagonist DQP 1105 weakened (figure 2c,d) but the GluN2C/2D selective potentiator CIQ strengthened NMDAR-mediated I Ca inhibition (figure 2d). Postsynaptic MNTB neurons before the onset of hearing employ GluN2A/2B subunit-containing NMDARs [65], whereas presynaptic calyceal terminals use GluN2C/2D subunitcontaining NMDARs ( figure 2c,d). Thus, NMDA induced the GluN2C/2D-dependent I Ca inhibition, which may decrease the action potential-driven glutamate release, resulting in the reduction of evoked EPSCs (figure 6a). Notably, presynaptic NMDAR activation still inhibited VGCCs (figure 1h,i) and glutamate release (figure 6a) in the presence of a physiological concentration of Mg 2þ (1 mM) in aCSF, where postsynaptic NMDARs are blocked at resting membrane potential. It is also noteworthy that postsynaptic MNTB neurons predominantly employ GluN2A/2C subunit-containing NMDARs after the onset of hearing [66].
At some other synapses, activation of presynaptic NMDARs enhances nerve-evoked transmitter release [23,67]. Whereas tonic elevation of intra-terminal Ca 2þ facilitates transmitter release, it potentially inhibits nerve-evoked transmitter release via adaptation of Ca 2þ sensor for exocytosis [68]. At the calyx of Held, however, either intra-terminal BAPTA (10 mM) or replacement of the charge carrier Ca 2þ with Ba 2þ had no effect on NMDA-induced I Ca inhibition. This excludes the involvement of Ca 2þ -dependent intracellular mechanism(s). Sustained depolarization of the nerve terminal upon NMDA application may inhibit evoked transmitter release by reducing the amplitude of presynaptic action potential. However, this was not the case for NMDAR-mediated presynaptic inhibition in the present study. Expected presynaptic depolarization from the inward currents produced by NMDA (less than 50 pA) is less than 10 mV [44]. Such a mild depolarization facilitates rather than inhibits transmitter release [69,70] by causing tonic Ca 2þ entry into the terminal [69]. However, upon activation  rsob.royalsocietypublishing.org Open Biol. 7: 170032 of presynaptic NMDARs, this facilitatory effect was actually masked by stronger inhibitory effect of presynaptic NMDAR-dependent I Ca inhibition. This effect may have been associated with the smaller reduction in evoked EPSC amplitude (32.9% in 1 mM Mg 2þ in aCSF, figure 6), compared to the reduction in I Ca amplitude (18.1% in 1 mM Mg 2þ in aCSF, figure 1h,i) (cf. the EPSC amplitude is proportional to the fourth power of I Ca [71]). This discrepancy may also be explained by spillover of a millimolar range of MK-801 from the patch pipette during its approach onto the postsynaptic MNTB neuron, which may have partially attenuated presynaptic NMDARs upon the EPSC recording.
Blockade of NMDA-induced I Ca inhibition by loading of intra-terminal MK-801 (figure 1g) and by omission of extracellular Na þ (figure 3e) implies that Na þ influx through presynaptic NMDARs may be involved. Interestingly, Na þ suppressed the enhancement of spontaneous transmitter release by presynaptic NMDAR activation in the mouse primary visual cortex [56]. The Na þ -mediated regulation mechanism should be elucidated in future studies.
The VGCC in the calyceal terminal is a common target for presynaptic G protein-coupled receptors (GPCRs), including mGluRs [45], GABA B Rs [48,72], noradrenaline a 2 Rs [50], adenosine A 1 Rs [51] and 5-HT 1B Rs [52] as well as presynaptic AMPARs [44]. These receptors activate heterotrimeric G proteins, and direct interaction of Gbg subunits with presynaptic Ca 2þ channels inhibits I Ca [49]. These presynaptic inhibitory effects of GPCR ligands on VGCCs occlude with each other [51] and are blocked by the non-hydrolysable GTP analogue GTPgS loaded into the calyceal terminal [49], implying that they share the same pathway. However, the present study showed intra-terminal GTPgS had no effect on NMDAinduced I Ca inhibition. This suggests that the mechanism which links NMDARs to VGCCs is distinct from the common GTP-G protein pathway.
Then, we aimed to identify the intracellular mechanism(s) to connect NMDAR activation to VGCC inhibition in the calyceal nerve terminal using the nitric oxide synthase inhibitor L-NNA, the protein kinase C inhibitor staurosporine and the CBR1 inhibitor AM 251. However, none of these inhibitors weakened NMDA-induced I Ca inhibition. Thus, future studies to determine the candidate intracellular mechanism(s) are needed.
Some pieces of evidence in the present study demonstrate functional expression of NMDARs in calyceal nerve terminals rather than in surrounding cells. First, NMDA-induced inward currents were recorded in the calyceal terminal following blockade of potassium conductance (figure 4c). Second, loading of MK-801 into the nerve terminal blocked NMDAinduced inhibition of I Ca (figure 1g) and NMDA-induced inward currents (figure 4b), whereas NMDA-induced presynaptic inhibition was observed in the presence of MK-801 in postsynaptic MNTB cells (figure 6a). Third, after replacement of extracellular Ca 2þ with Ba 2þ , NMDA inhibited presynaptic Ba 2þ currents (figure 3d), despite the fact that synaptic transmission is nearly terminated after replacement of Ca 2þ with Ba 2þ [44]. Unfortunately, the extremely low amplitude of NMDA-induced presynaptic membrane currents (17.3 pA at 280 mV, figure 5c) prevented us from further dissecting the properties of NMDARs expressed in calyceal terminals. rsob.royalsocietypublishing.org Open Biol. 7: 170032 The presynaptic inhibitory effect of NMDA had an IC 50 of 135 mM for I Ca (figure 1c) and 112 mM for evoked EPSCs (figure 6c), respectively. The recombinant NMDAR currents in Xenopus oocytes have an EC 50 of 30 -60 mM for NMDA, depending on GluN2 subunits co-expressed with the GluN1 subunit [73]. Similarly, the EC 50 of native NMDAR currents in CA1 neurons in hippocampal slices is 38 mM [74]. The relatively low affinity of presynaptic NMDARs may reflect the involvement of GluN1 splice variants with low ligand affinity [75]. Ambient glutamate concentration is 55 nM at the calyx of Held synapse of immature rats [76]. In hippocampal slices, the glutamate concentration is 30 nM in the absence of the glutamate transporter inhibitor TBOA, but rises to 200 nM in its presence [74,77]. In the synaptic cleft, the glutamate concentration is estimated to rise above 300 mM [78] or up to 1 mM [79] during excitatory transmission. In experimental anoxia, the glutamate concentration in the cerebral cortex can rise to 400 mM in vivo [80]. Our finding that D-AP5 affected I Ca evoked by repetitive stimulation in the presence of TBOA, but not in its absence (figure 5), suggests that presynaptic NMDARs do not operate under physiological conditions. However, under pathological conditions in which extracellular glutamate concentration rises to a high level (e.g. brain anoxia), presynaptic NMDARs may act to reduce glutamate release, thereby playing a protective role against neuronal death due to glutamate excitotoxicity.
Previous research reported that synaptic transmission and glutamate transporter activity at the calyx of Held synapse differ between physiological and room temperature [81]. Thus, the temperature-dependency of NMDAR-mediated rsob.royalsocietypublishing.org Open Biol. 7: 170032 presynaptic inhibition remains to be examined in order to further investigate the physiological and pathophysiological relevance. Besides, it needs to be noted that all data in this study were obtained using immature rats (P7-9). It is beyond the scope of our current study to clarify whether the NMDAR-dependent presynaptic inhibition is developmentally regulated, similar to TBOA-induced presynaptic mGluR activation [61].
In conclusion, the present study identified a novel regulatory mechanism for NMDAR-dependent presynaptic inhibition at an excitatory synapse in the auditory brainstem of rat pups by direct presynaptic recordings. Moreover, it also revealed the presence of functional presynaptic NMDARs. These findings bring significant insights to the controversial research field on presynaptic NMDARs. Finally, this study not only provides applicable implications to presynaptic inhibition at other central synapses, but also potentially serves to develop drugs targeting presynaptic NMDARs in the CNS.
Ethics. All experiments complied with the guideline of the Physiological Society of Japan as well as the institutional guidelines. The experimental protocols were approved by the animal experimentation committees of University of Tokyo, Tokyo Medical and Dental University, and National Rehabilitation Center for Persons with Disabilities.
Data accessibility. The datasets supporting this article have been uploaded as part of the electronic supplementary material.