Glycine Induces Bidirectional Modifications in N-Methyl-d-aspartate Receptor-mediated Synaptic Responses in Hippocampal CA1 Neurons*

Background: NMDA receptor plays roles central to various brain functions. Results: Glycine, an amino acid widely presents in the brain, induces bidirectional modifications in NMDA receptor-mediated responses. Conclusion: Trafficking of NMDA receptors at postsynaptic sites may underlie bidirectional modifications in NMDA receptor responses. Significance: Revealing the molecular mechanism by which glycine dictates direction of persistent modifications in NMDA receptor function. Glycine can persistently potentiate or depress AMPA responses through differential actions on two binding sites: NMDA and glycine receptors. Whether glycine can induce long-lasting modifications in NMDA responses, however, remains unknown. Here, we report that glycine induces long-term potentiation (LTP) or long-term depression (LTD) of NMDA responses (Gly-LTPNMDA or Gly-LTDNMDA) in a dose-dependent manner in hippocampal CA1 neurons. These modifications of NMDA responses depend on NMDAR activation. In addition, the induction of Gly-LTPNMDA requires binding of glycine with NMDARs, whereas Gly-LTDNMDA requires that glycine bind with both sites on NMDARs and GlyRs. Moreover, activity-dependent exocytosis and endocytosis of postsynaptic NMDARs underlie glycine-induced bidirectional modification of NMDA excitatory postsynaptic currents. Thus, we conclude that glycine at different levels induces bidirectional plasticity of NMDA responses through differentially regulating NMDA receptor trafficking. Our present findings reveal important functions of the two glycine binding sites in gating the direction of synaptic plasticity in NMDA responses.

Glycine is a multifaceted bioactive molecule in the central nervous system (CNS) and exerts its effect through at least three targets (1,2). First, glycine is a strychnine-insensitive coagonist for classical NMDA receptors (NMDARs) and its bind-ing with NMDARs GluN1 subunit is essential for NMDAR activation (3)(4)(5). Second, glycine is a primary inhibitory neurotransmitter in certain brain regions such as the spinal cord and brain stem and displays its inhibitory effect through activation of glycine receptors (GlyRs). Enormous evidence over the last decade has suggested that functional GlyRs are present in all regions of the hippocampus and play an important role in regulating cell excitability and synaptic plasticity (6,7). These strychnine-sensitive GlyRs, if located postsynaptically, are mostly in extrasynaptic sites (8). Third, NMDAR GluN3A subunit has very high affinity for glycine with an apparent dissociation constant less than that of GluN1 (9).
Under most physiological conditions, the glycine concentration in cerebrospinal fluid is estimated to be in the low micromolar range (12). The affinity of glycine for NMDA receptors (EC 50 ϭ 0.5 M) is substantially higher than that of GlyRs (EC 50 ϭ 0.67 mM; see Refs. 10 and 11) but lower than that of GluN3A. It seems GluN3A has the priority to bind to glycine. However, compared with the GluN1 and GluN2 subunits of NMDAR, the expression level of GluN3A is much lower in hippocampus (13,14). Moreover, at CA1 hippocampal synapses GluN3A is only localized at extra-and perisynaptic sites but not postsynaptic sites of the plasma membrane (13,15). Therefore, in the hippocampus the endogenous glycine released from the presynaptic site is more likely to exert its excitatory effect through postsynaptic NMDAR co-agonist binding (16,17). Under some pathophysiological conditions, such as brain ischemia and epilepsy, glycine concentration in the synaptic cleft is elevated significantly (18,19). This raises the possibility that excessive glycine accumulation in the synaptic cleft of the hippocampus can spill over to extrasynaptic sites and reach the dosage threshold of activating the functional GlyRs inhibitory effect. This possible bidirectional effect by glycine is supported by our recent findings that suggest that glycine dictates the direction of synaptic plasticity in a dose-dependent manner (6). Through its binding with NMDARs, glycine at relatively low levels (0.6 mM) in Mg 2ϩ -free perfusion medium induces long-term potentiation (LTP) 3 of AMPA receptor-mediated excitatory postsynaptic currents (AMPA EPSCs) in CA1 neurons of hippocampal slices. In contrast, glycine at high levels (1.5 mM) activates both GlyRs and NMDARs and produces long-term depression (LTD).
Because NMDAR plays a central role in many brain functions including synaptic plasticity, learning, and memory (20,21), upor down-regulation of NMDAR number and/or function may contribute to the bidirectional modifications of AMPA EPSCs indirectly by glycine. Therefore, it is tempting to examine the role of glycine in eliciting persistent modifications of NMDARmediated synaptic responses. In the present study, we demonstrate that similar to its effect on AMPA responses, glycine at different levels induces bidirectional modifications in NMDA receptor-mediated synaptic responses in hippocampal CA1 neurons. These effects are mainly attributable to the modulatory effects of glycine on NMDARs and GlyRs. Moreover, activity-dependent exocytosis and endocytosis of postsynaptic NMDARs underlie glycine-induced bidirectional modifications of NMDA responses. Our findings reveal important functions of the two glycine binding sites in gating the direction of synaptic plasticity of NMDA receptor-mediated synaptic responses.
Whole-cell voltage-clamp recordings were made with patch pipettes containing (in mM): 132.5 Cs-gluconate, 17.5 CsCl, 2 MgCl 2 , 0.5 EGTA, 10 HEPES, 4 ATP, and 5 QX-314, with pH adjusted to 7.2 by CsOH. Hippocampal slices were continuously superfused in 34°C ACSF bubbled with 95% O 2 , 5% CO 2 . Excitatory postsynaptic responses were evoked by stimulating Schaffer fibers through a constant current pulse delivered using a bipolar tungsten electrode. Synaptic responses were evoked at 0.05 Hz, except during glycine application. EPSCs were recorded at Ϫ65 mV. NMDAR-mediated EPSCs (NMDA EPSCs) were isolated with continuous low-Mg 2ϩ (0.25 mM) ACSF perfusion containing bicuculline methiodide (BMI, 10 M, Tocris) to block GABA A receptor-mediated inhibitory syn-aptic currents and NBQX (10 M, Sigma) to block AMPARmediated excitatory synaptic currents (AMPA EPSCs). ACSF was always kept in 0.25 mM Mg 2ϩ during the whole recordings. CA1 neurons were viewed under upright microscopy (ECLIPSE E600-FN, Nomasky, Nikon Corporation, Tokyo, Japan) and recorded using an Axopatch-200B amplifier (Molecular Devices, Palo Alto, CA). Changes in NMDA EPSCs amplitude were examined during the last 5 min of recording. Records were low-pass filtered at 2 kHz and acquired at 5-10 kHz. Recordings from each neuron lasted at least 60 to 80 min. The series resistance (R s ) in the recordings varied between 4 and 6 megaohm. For fear of re-sealing the ruptured membrane, changes in kinetics and amplitude of the EPSCs, caused by R s , were always monitored during recording. Cells in which the R s or capacitance deviated by Ͼ20% from initial values were excluded from the analysis. Cells with R s Ͼ20 megaohm at any time during the recording were also excluded from the analysis. Data were collected with pClamp 9.2 software and analyzed using Clampfit 9.2 (Molecular Devices, Palo Alto, CA).
Immunofluorescence Labeling and Analysis-Fluorescent immunostaining was applied to determine the colocalization of GluN1 and PSD-95 in rat hippocampal slices. 400-m slices were prepared and incubated for 10 min with glycine (0.6 mM or 1.5 mM). After a 15-min recovery time, slices were fixed in icecold 4% paraformaldehyde overnight, dehydrated in 30% sucrose for 24 h at 4°C, and sectioned at 30 -40 m on a freezing microtome (Leica CM1900). Sections were made permeable in 0.3% Triton X-100 for 60 min (37°C), then blocked with 10% cattle serum for 60 min. Sections were then incubated with goat anti-GluN1 antibody (1:200, Santa Cruz), mouse anti-PSD-95 antibody (1:100, Millipore) in PBS containing 10% cattle serum at 4°C for 48 h. After thoroughly washing with PBS, sections were probed with FITC-and TRITC-conjugated secondary antibody (1:500, Jackson ImmunoResearch Laboratories) overnight at 4°C. The sections were mounted with a mounting medium (Vectashield, Vector Laboratories) after rinsing with PBS. A confocal imaging system (Olympus FV1000) with a ϫ60 oil immersion lens was used for image acquisition. A series of optical sections were collected at 0.5-m steps and a resolution of 1024 ϫ 1024 pixels. Each image was collected by averaging five scans. Because of variability in brightness, it was necessary to use different gain and contrast settings for different cells. To control for this, all measurements were expressed in terms of ratios. All measurements were performed using ImageJ (NIH) software. To quantify the staining, sections from 3 animals were used for quantitative analysis. Generally, 3-4 images of each slice were averaged to determine each value. A close-up view was obtained from one segment of an apical dendrite (about 50 -150 m away from the cell body layer). The data were analyzed with ANOVA LSD for statistical significance and expressed as mean Ϯ S.E. Subcellular Fractionation and Immunoblotting-Hippocampal slices were prepared as described in electrophysiological recordings. After a 1-h recovery, slices were incubated for 10 min in the presence of glycine (0.6 mM or 1.5 mM). After washing 3 times with fresh ACSF, slices were incubated for 10 -15 min. Then slices were stored in liquid nitrogen immediately in cold 0.32 M sucrose containing 1 mM HEPES, 1 mM MgCl 2 , 1 mM NaHCO 3 , 20 mM sodium pyrophosphate, 20 mM ␤-phosphoglycerol, 0.2 mM dithiothreitol, 1 mM EDTA, 1 mM EGTA, 50 mM NaF, 1 mM Na 3 VO 4 , 1 mM p-nitrophenyl phosphate (pH 7.4), in the presence of protease inhibitors and phosphatase inhibitors: 1 mM phenylmethylsulfonyl fluoride (PMSF), 5 g/ml of aprotinin, 5 g/ml of leupeptin, 5 g/ml of pepstatin A, and 16 g/ml of benzamidine. The homogenate was centrifuged at 1,000 ϫ g for 10 min. 200 l were removed from the total protein fractions (H) and the remaining supernatant was centrifuged at 3,000 ϫ g for 15 min to obtain the fraction of mitochondria and synaptosomes. The pellet was resuspended in 8 ml of hypotonic buffer with the presence of protease inhibitors and centrifuged at 100,000 ϫ g for 1 h and then suspended again in 8 ml of buffer containing 75 mM KCl and 1% Triton X-100 and centrifuged at 100,000 ϫ g for 1 h. The final pellet was homogenized 3 times in 20 mM HEPES. This fraction is regarded as a Triton-insoluble fraction. The Triton-insoluble fraction was used instead of the classical PSD (45).
Data Analysis-All populations are described as mean Ϯ S.E. Data were analyzed by analysis of variance (ANOVA) or Student's t test. Within-group comparisons were performed using paired-sample t tests, and differences between groups were compared using independent-sample t test and ANOVA post hoc comparisons. A one-way ANOVA LSD test was used when equal variances were assumed. Differences were considered significant when p Ͻ 0.05 (*, p Ͻ 0.05; **, p Ͻ 0.01).

Glycine-induced Bidirectional Modification in NMDA EPSCs-
To examine whether glycine induces changes in NMDAR-mediated synaptic responses, we performed whole-cell patch clamp recordings of evoked NMDAR-mediated excitatory postsynaptic currents (NMDA EPSCs) in CA1 pyramidal neurons in hippocampal slices. NMDA EPSCs were recorded at a Ϫ65 mV holding potential in a low-Mg 2ϩ (0.25 mM) ACSF to relieve blockade of NMDARs by Mg 2ϩ . The ACSF perfusion medium contains BMI (10 M) and NBQX (10 M) to block GABA A and AMPA receptor-mediated synaptic currents, respectively (Fig. 1A). Following 10 min of stable baseline EPSCs recording, 0.6 mM glycine was applied for 10 min and then washed out immediately. Changes in EPSCs amplitude were examined during the last 10 min of recording. We found that 0.6 mM glycine significantly increased NMDA EPSCs and induced LTP of NMDA EPSCs (Gly-LTP NMDA , normalized amplitudes 1.31 Ϯ 0.06, n ϭ 5, p Ͻ 0.01, paired-samples t test; Fig. 1, B and F). This persistent modification in NMDAR-mediated synaptic responses was reproducible and consistent (i.e. observed in more than 80% of recorded cells), except when exogenous glycine reached 1.0 mM, in which no persistent change in NMDA EPSCs was observed (n ϭ 6, p ϭ 0.07). Further increasing the glycine concentration to 1.5 mM produced LTD of NMDA EPSCs (Gly-LTD NMDA , 0.62 Ϯ 0.06, n ϭ 6; p Ͻ 0.01, paired-samples t test). Notably, Gly-LTD NMDA was not due to the rundown of NMDA EPSCs caused by the deterioration of the recorded neurons (data not shown) or to an adverse effect on the recording of the neurons during glycine treatment, because glycine at this concentration does not display toxic effects on nerve cells (22)(23)(24). The above results suggest that exogenously applied glycine at different levels induces persistent bidirectional modifications in NMDAR-mediated synaptic responses.
We then investigated whether such changes in NMDA EPSCs could also be reproduced by endogenous glycine. To test this possibility, we inhibited endogenous glycine re-uptake with a selective blocker of type 1 glycine transporters (24,25) to observe how the accumulation of extracellular glycine could affect NMDA EPSCs. A high-affinity glycine transporter type 1 (GlyT1) was reported in glial cells and glutamatergic neurons (26) (27). A 10-min application of a low concentration of NFPS (0.2 M), which should cause a relatively low level of endogenous glycine accumulation in extracellular milieu, induced LTP of NMDA EPSCs (1.32 Ϯ 0.06, n ϭ 8, p Ͻ 0.01; Fig. 1, C and F). In contrast, high concentrations of NFPS (2.0 M) in low Mg 2ϩ perfusion medium elicited a persistent depression of NMDA EPSCs (0.83 Ϯ 0.06, n ϭ 6, p Ͻ 0.01, paired samples t test; Fig. 1, D and F). These results are consistent with the above data on exogenous glycine and provide further support to the notion that glycine at different levels induces bidirectional modifications in NMDA EPSCs.

. N-[3-([1,1-Biphenyl]-4-yloxy)-3-(4-fluorophenyl)propyl]-N-methylglycine (NFPS) is a specific and potent GlyT1 blocker and displays dose-dependent GlyT1 inhibition
To further confirm that the observed changes in NMDA EPSCs upon NFPS treatment are caused by accumulation of extracellular glycine at different levels, we modulated the duration of NFPS treatment. The assumption is that if the duration of NFPS treatment is extended to 0.2 M, glycine accumulation will be higher, and, as a result, the direction of modification in NMDA EPSCs may change. To examine this assumption, we switch the 0.2 M NFPS perfusing time from 10 to 30 min. We indeed observed Gly-LTD NMDA rather than Gly-LTP NMDA (0.89 Ϯ 0.15, n ϭ 5, p Ͻ 0.05; Fig. 1, C and E). In contrast, shortening the duration of 2.0 M NFPS treatment from 10 to 4 min, which should lead to less extracellular glycine accumulation, induced Gly-LTP NMDA rather than Gly-LTD NMDA (1.47 Ϯ 0.14, n ϭ 6, p Ͻ 0.01; Fig. 1, D and E). These results re-emphasize that the amount of extracellular glycine is a critical factor that dictates the direction of glycine-induced synaptic plasticity in NMDA EPSCs.
Modification of NMDA EPSCs Depends on NMDAR Activation-To examine whether the induction of Gly-LTP NMDA and Gly-LTD NMDA requires the activation of NMDARs, we co-applied the competitive NMDAR antagonist DL-AP5 (50 M) with glycine. Under this treatment, we failed to detect any obvious changes in NMDA EPSCs (0.6 mM Gly ϩ AP5, 0.96 Ϯ 0.13, n ϭ 6, compared with baseline, p ϭ 0.09, paired samples t test; 1.5 mM Gly ϩ AP5, 1.02 Ϯ 0.18, n ϭ 6, compared with baseline, p ϭ 0.45, paired samples t test; Fig. 2, A and C). In addition, we performed another parallel set of experiments in ACSF with normal Mg 2ϩ concentration (1.0 mM) perfusion solution only when glycine was applied. Because of the Mg 2ϩ block of NMDARs, this treatment could block the selective activation of synaptic NMDARs. Similar to the above results using AP5, no Gly-LTP NMDA or Gly-LTD NMDA were detected (0.6 mM Gly, 0.99 Ϯ 0.09, n ϭ 6, compared with baseline, p ϭ 0.72, paired samples t test; 1.5 mM Gly, 0.98 Ϯ 0.03, n ϭ 6, compared with baseline, p ϭ 0.07, paired samples t test; Fig. 2, B and C). These data strongly suggest that both LTP NMDA and LTD NMDA induced by glycine at different concentrations depend on NMDAR activation.
NMDAR Trafficking Underlies Modification in NMDA EPSCs-An important mechanism underlying persistent modification in neurotransmission is to change the receptor number at postsynaptic sites. To examine whether the glycine-induced changes in NMDA EPSCs are caused by alteration of postsynaptic NMDAR number, we performed immunofluorescence assays in hippocampal slices to determine the amount of synaptic NMDARs by quantifying colocalization of NMDARs with PSD-95, a postsynaptic marker. We detected a significant increase and decrease in postsynaptic expression of NMDARs shortly after 0.6 and 1.5 mM glycine treatments, respectively (normalized GluN1 intensity, 1.37 Ϯ 0.03, n ϭ 14; 0.69 Ϯ 0.02, n ϭ 14, compared with control, p Ͻ 0.01, one-way ANOVA LSD test; Fig. 3, A and B). This observation is further supported with Western blot data. Using a Western blot assay of the Triton-insoluble fraction of hippocampal tissue, which roughly represents the subcellular fraction at postsynaptic sites (28,29), we found a dramatic elevation and decline of postsynaptic expression of NMDARs after 0.6 and 1.5 mM glycine treatments, respectively (1.24 Ϯ 0.08, p Ͻ 0.01, n ϭ 7, one-way ANOVA LSD test; 0.83 Ϯ 0.04, p Ͻ 0.01, n ϭ 7, one-way ANOVA LSD test; Fig. 3, E and F). Taken together, these results demonstrate that glycine at different levels induces bidirectional changes in NMDAR expression at postsynaptic sites.
GluNR3A is another potential target of glycine. Therefore, glycine might also induce changes in GluN3A expression at postsynaptic sites, which in turn contributes to the overall change in NMDAR expression. To examine this possibility, we first determined the expression level of GluN3A using whole hippocampal tissue. We found that the GluN3A expression level is very low and hard to detect when using a regular amount of hippocampal tissues for the protein level assay ( Fig. 3C; see Ref. 13). When we harvested 5-fold hippocampal tissues, however, we do detect GluN3A expression (Fig. 3D). Interestingly, we observed developmental changes in GluN3A expression from P2 to adult, which is consistent with previous findings (14,30). Moreover, when we determined the GluN3A level in the Triton X-100-insoluble fraction of hippocampal tissue (P20), which roughly represents the subcellular fraction at postsynaptic sites, we did not detect any obvious changes in GluN3A expression (1.02 Ϯ 0.02, p ϭ 0.38, n ϭ 7, one-way ANOVA LSD test; 1.01 Ϯ 0.03, p ϭ 0.75, n ϭ 7, one-way ANOVA LSD test; Fig. 3, E and F). These results are also consistent with previous findings that GluN3 predominantly locates in extra-and perisynaptic sites of the hippocampus (15). Taking into consideration the fact that the other GluN3 subunit GluN3B is mainly expressed in motor neurons (31), these results suggest that GluN3 does not contribute to glycine-induced changes in NMDAR expression level.
How could these changes occur in postsynaptic expression of NMDARs following glycine treatment? It has been reported that receptor trafficking through exocytosis or endocytosis is a major route to adding or decreasing the receptor number at the postsynaptic surface (32). We determined next whether bidirectional trafficking of NMDARs contributes to glycine-induced bidirectional modification of NMDA EPSCs. To test this possibility, we loaded cells with the specific SNARE-dependent exocytosis inhibitor tetanus toxin (TeTx, 0.1 mM) by including it in the pipette solution during whole cell recording and examined whether Gly-LTP NMDA could be induced under this condition (33,34). We found that TeTx completely blocked Gly-LTP NMDA induction (0.99 Ϯ 0.06, n ϭ 6; p Ͼ 0.05; Fig. 4, A and C). To further confirm our observation, we used a peptide corresponding to the C-terminal region of SNAP-25 to acutely disturb the association of endogenous SNAP-25 with its SNARE complex binding partners, syntaxin-1 and VAMP2. This interfering peptide (corresponding to residues 182-192 of SNAP-25) contains the nonphosphorylatable residue Ala in place of Ser at residue 187 flanked on either side by 5 amino acids (35). Loading interfering peptide (10 M) in the recording pipette totally abolished Gly-LTP NMDA (Fig. 4, A and C). In contrast, when we used a scrambled peptide with the same residue composition in a randomized order that renders it incapable of blocking excytosis, glycine could still induce potentiation of NMDA currents (1.34 Ϯ 0.08, n ϭ 6, p Ͻ 0.01). These results indicate that SNARE-dependent NMDAR exocytosis is responsible for increased surface expression of NMDARs.
To examine the possible contribution of NMDAR endocytosis in the induction of Gly-LTD NMDA , we loaded cells with an inhibitor of dynamin-dependent endocytosis termed D15 during recordings. D15 is a cell membrane-impermeable peptide that possesses the PPPQVPSRRNRAPPG sequence and can specifically interfere with the binding of dynamin to amphiphysin (36 -38). Unexpectedly, we found that D15 (2 mM) reversed the polarity of synaptic plasticity from Gly-LTD NMDA to Gly-LTP NMDA (1.35 Ϯ 0.08, n ϭ 6, p Ͻ 0.01; Fig. 4, B and C), with the resultant potentiation magnitude comparable with Gly-LTP NMDA induced by 0.6 mM glycine alone (p Ͼ 0.05). To further confirm this observation, we also made use of a dynamin inhibitory peptide, QVPSRPNRAP, which also competitively binds to the amphiphysin SH3 domain and prevents endocytosis when applied intracellularly (7,39,40). Tat-QVPSRPNRAP is a membrane-permeable form of this peptide. Tat protein (Tyr-Gly-Arg-Lys-Lys-Arg-Arg-Gln-Arg-Arg-Arg), which was obtained originally from the cell membrane transduction domain of the human immunodeficiency virus type 1 (HIV-1), was fused to the constructed peptides and resulted in the fusion peptide Tat-QVPSRPNRAP. This manipulation allowed the constructed peptides to easily cross the membrane and exert their effects intracellularly (41). Consistent with the results using D15, briefly perfusing slices with both 1.5 mM glycine and interfering peptide (50 M) resulted in Gly-LTP NMDA (1.34 Ϯ 0.12, n ϭ 7, p Ͻ 0.01; Fig. 4, B and C). As a negative control, we applied the non-membrane-permeable form of the peptide (without the TAT) to bath solution and found that it did not affect the internalization of NMDARs (0.69 Ϯ 0.08, n ϭ 7, p Ͻ 0.01). These results indicate that dynamin-dependent NMDAR endocytosis underlies Gly-LTD NMDA . The reversal of plasticity polarity may be caused by competition between the dichotomous effects (excitatory versus inhibitory or exocytosis versus endocytosis) mediated by the two glycine binding sites. When endocytosis or the suppressive effect was removed, the exocytosis or excitatory effect was uncovered. Taken together, our observation strongly suggests that bidirectional NMDAR trafficking, which results in an increase or decrease of the number of postsynaptic NMDARs, is responsible for glycine-induced bidirectional modifications of NMDA EPSCs. If these interfering peptides do disturb intracellular NMDAR trafficking, then we should observe correlated changes in NMDAR expression at postsynaptic sites. To examine this possibility, we determined the postsynaptic NMDAR expression under various treatments. We found a significant elevation of NMDAR expression after 0.6 mM glycine treatment for 20 min (1.33 Ϯ 0.11, n ϭ 5, p Ͻ 0.05; Fig. 4D). Pretreatment of the The magnitude of resultant Gly-LTP NMDA is comparable with that of Gly-LTP NMDA induced by 0.6 mM glycine. The plotting data for Gly-LTD NMDA was borrowed from Fig. 1B for comparison. C, summary of data displaying changes in NMDA EPSCs following treatments of various exocytosis or endocytosis blockers. **, p Ͻ 0.01, compared between the indicated groups. D, inhibiting exocytosis reversed the enhancement in postsynaptic NMDAR expression induced by glycine at 0.6 mM (n ϭ 5). As a control, scramble peptide failed to display a similar effect (n ϭ 5). Pep., SNAP-25 peptide; Scr., scramble peptide. E, inhibiting endocytosis reversed the decrease in postsynaptic NMDAR expression induced by glycine at 1.5 mM (n ϭ 5). The control cell impermeable membrane peptide failed to display a similar effect (n ϭ 5). T, Tat-QVPSRPNRAP peptide; Q, QVPSRPNRAP peptide.
The reversal of Gly-LTD NMDA to Gly-LTP NMDA by blocking endocytosis could be caused by a counter NMDAR exocytosis.
If this assumption is true, then we can make the two following predictions: first, inhibiting NMDAR exocytosis upon 1.5 mM glycine application would induce Gly-LTD NMDA with greater depression magnitude; second, the Gly-LTP NMDA that was reversed from Gly-LTD NMDA upon endocytosis blockade would be abolished by further treatment with the exocytosis blocker. To test these predictions, we applied either TeTx or SNAP25 in the recording pipette solution upon 1.5 mM glycine application (0.70 Ϯ 0.07, n ϭ 6, compared with baseline, p Ͻ 0.01; 0.50 Ϯ 0.12, n ϭ 5, p Ͻ 0.01; 0.50 Ϯ 0.10, n ϭ 6, p Ͻ 0.01; Fig. 5, A and B). Changes in EPSCs amplitude were examined during the last 10 min of recording. We indeed detected Gly- The plotting data for Gly-LTD NMDA was borrowed from Fig. 1B for comparison. B, statistical plots of data in A. Changes in EPSCs amplitude were examined during the last 10 min of recording. C, co-applied TeTx or SNAP25 peptide with D15 in the recording pipette solution abolished reversed Gly-LTP NMDA upon D15 treatment alone. Neither Gly-LTP NMDA nor Gly-LTD NMDA was induced by 1.5 mM glycine under these conditions, indicating that the reversal of Gly-LTD NMDA to Gly-LTP NMDA was caused by a counter NMDAR exocytosis uncovered after blocking endocytosis (n ϭ 6). D, statistical plots of data in A. **, p Ͻ 0.01, compared with control, one-way ANOVA LSD test. E, D15 alone failed to display any obvious effects on basal NMDA response (n ϭ 6). F, statistical plots of data in E, compared with control, one-way ANOVA LSD test.
LTD NMDA with a greater depression magnitude. In addition, we co-applied TeTx or SNAP25 peptide with D15 in the recording pipette solution to identify whether competing trafficking pathways were operating. We found that the reversed Gly-LTP NMDA upon D15 treatment alone was abolished and neither Gly-LTP NMDA nor Gly-LTD NMDA was induced by 1.5 mM glycine (0.94 Ϯ 0.10, n ϭ 6, compared with baseline, p Ͼ 0.05; Fig. 5, C and D), indicating that NMDAR exocytosis contributes to the reversed Gly-LTP NMDA . Moreover, D15 alone did not display any obvious effects on basal NMDA responses (1.01 Ϯ 0.07, n ϭ 6, p ϭ 0.25; Fig. 5, E and F). These observations highly support our predictions and provide new evidence to the notion that NMDAR exocytosis and endocytosis underlie Gly-LTP NMDA and Gly-LTD NMDA , respectively, and these two receptor trafficking processes may compete with each other under glycine application.
Role of Glycine Binding Sites in Modification of NMDA EPSCs-In the hippocampus, glycine can exert its action NMDAR co-agonist binding site (site B) and extrasynaptic GlyR site (site A), which mediate excitatory and inhibitory actions, respectively. Because NMDARs have a substantially higher affinity for glycine than GlyRs (10,11), the potentiation effect by glycine at a relatively low level is more likely via its action on NMDAR co-agonist binding site (site B). As the level of glycine increases, it may further activate GlyRs (site A) largely located at the extrasynaptic site and display an inhibitory effect in hippocampal neurons. We assumed that this dichotomous property of glycine may take a role in bidirectional modification of NMDA EPSCs. To examine this hypothesis, we used antagonists of the two glycine binding sites to determine whether and how the observed changes in NMDA EPSCs could be affected by these blockers. A specific glycine site B antagonist, L689560 (60 M), was applied together with glycine in low-Mg 2ϩ ACSF. As a result, Gly-LTP NMDA was completely abolished (0.89 Ϯ 0.13, n ϭ 6, compared with baseline, p Ͼ 0.05, paired samples t test; Fig. 6, A and C). However, in slices treated with the specific GlyR antagonist strychnine (1 M), no significant change in the magnitude of Gly-LTP NMDA was detected (1.32 Ϯ 0.10, n ϭ 6, compared with baseline, p Ͻ 0.01, paired-samples t test; Fig. 6, A and C). Thus, the induction of Gly-LTP NMDA was dependent on glycine binding with NMDARs.
Because glycine at high levels may exert its effect on both the NMDAR co-agonist binding site (site B) and GlyRs (site A), we further investigated the mechanisms underlying Gly-LTD NMDA and determined how Gly-LTD NMDA was affected by the antagonists of the two glycine binding sites. Interestingly, we found that Gly-LTD NMDA was totally abolished by L689560 (60 M) but was reversed to Gly-LTP NMDA upon treatment of strychnine (1 M; 1.02 Ϯ 0.11, n ϭ 6, paired samples t test, p Ͼ 0.05; 1.19 Ϯ 0.12, n ϭ 6, compared with baseline, p Ͻ 0.01; Fig. 6, B and C), indicating that GlyRs was also involved in Gly-LTD NMDA induction. In addition, co-application of the two antagonists displayed similar blocking effects on Gly-LTD NMDA without further reversing Gly-LTD NMDA to Gly-LTP NMDA (Fig. 6, B and C). This was not unexpected, because L689560 blocks glycine binding at NMDARs, which was also required for Gly-LTP NMDA induction.
To confirm the role of these two binding sites in glycineinduced modifications of NMDA EPSCs, we next examined how the antagonists for these binding sites affect Gly-LTP NMDA and Gly-LTD NMDA induced by endogenous glycine accumulation following GlyT1 blockade. LTP elicited by 10 min perfusion of NFPS (0.2 M) was abolished by L689560 (60 M; 0.96 Ϯ 0.17, n ϭ 5, compared with control, p Ͻ 0.01; Fig. 7, A and C), whereas Gly-LTD NMDA elicited by 10 min perfusion of NFPS (2.0 M) was reversed to Gly-LTP NMDA by strychnine (1 M; 1.64 Ϯ 0.19, n ϭ 5, p Ͻ 0.01; Fig. 7, B and C). Combined with the above results, these findings provide strong evidence that glycine induces bidirectional modifications of NMDA response by separately targeting the two glycine binding sites. Taking into consideration the accompanying alteration of postsynaptic NMDAR expression upon treatment of these antagonists, we propose glycine-induced changes in NMDAR trafficking and subsequent changes in postsynaptic NMDAR number may contribute to the modification of NMDA responses.

DISCUSSION
NMDARs mainly serve as the molecular trigger for synaptic changes. Activity-dependent changes in NMDAR-mediated synaptic strength are of great importance because they involve potentiation or depression of the molecular trigger itself (42). In the present study, we demonstrate that glycine, an important amino acid released by both glial cells and glutamatergic neurons, elicits persistent bidirectional modifications in NMDARmediated synaptic responses at different concentrations. Further investigation elucidated the mechanisms by which glycine exerts its role in the synaptic plasticity of NMDAR-mediated synaptic responses.
Glycine has been reported to potentiate the NMDA response through an allosteric activation of NMDARs in cultured mouse brain neurons (4). This effect is observed when glycine was co-applied with NMDA. It is still unclear whether this potentiation effect is acute or longer in duration. Our present findings that glycine at a relative low level can induce long-lasting facil-itation of NMDA EPSCs is the first demonstration of persistent modification of NMDA EPSCs by glycine. These findings, however, are in clear contrast to a previous study (43). In that study, glycine at a similar dosage induced LTP of AMPA receptormediated synaptic responses but failed to display significant effects on miniature NMDAR-mediated synaptic responses (mEPSCs NMDA ). This inconsistency could be caused by difficulty in detecting changes in relatively smaller spontaneous EPSC amplitudes, as well as greater variability of NMDA versus AMPA responses (44). The small change (Ͻ30%) in mEPSCs NMDA is not easily detected. Compared with mEPSCs NMDA , evoked NMDA EPSCs recorded in neurons in slices are usually much larger with relatively smaller variability.
Glycine exerts its inhibitory effect mainly through the activation of GlyRs and the subsequent opening of GlyR-gated chloride channels. In the present study, we show that, in contrast to the induction of LTP NMDA by glycine at relatively low concentrations, glycine at high concentrations produces LTD NMDA . We further demonstrate that decreasing the NMDAR number through endocytosis underlies the LTD NMDA . Persistent suppression of NMDA EPSCs can be blocked by either L689560 or strychnine, suggesting GlyRs may also have a role in this effect. One study has revealed that stimulation of the glycine site can initiate signaling through the NMDAR complex and prime the receptors for clathrin-dependent endocytosis (7). Our present observation of glycine-induced NMDA EPSC suppression and NMDAR endocytosis supports this finding. Notably, the priming effects of glycine in that study were observed in the persistent presence of the GlyRs blocker strychnine, thus failing to monitor the possible involvement of GlyRs in the effect of glycine on the NMDAR-mediated function.
Here we demonstrate that the activation of GlyR by glycine has additional contributions to the glycine-induced NMDAR endocytosis as well as suppression of NMDAR function. Thus, it would be expected that antagonizing glycine binding to both sites upon a high level glycine treatment may uncover the potentiation and excitatory effects on NMDAR. However, because glycine-induced potentiation in NMDAR currents require glycine binding at the NMDAR site, the resultant NMDA currents fail to display any obvious change. Moreover, it is still unclear how the effect of GlyR activation is transduced to NMDAR internalization.
Our present results strongly suggest that the concentration is a very important determinant for the effect of glycine on NMDAR-mediated neurotransmission. This may be largely attributable to differences in the affinity of glycine for the NMDAR co-agonist binding site (site B) and GlyRs (site A). NMDA receptors have a substantially higher affinity for glycine than GlyRs (10,11). When the level of endogenous glycine is low, it typically prefers to act through the NMDAR co-agonist binding site. Under this condition, glycine induces potentiation of NMDAR function (Fig. 8A). When the level of glycine gets higher and crosses the "set point" of the NMDAR endocytosis mechanism, it may activate GlyRs (Fig. 8B). This increase of the glycine level could occur under some pathophysiological states, such as seizure and ischemia. As a result, the inhibitory effect may be stronger than the excitatory effect and the net effect mediated by glycine is a depression of the NMDA response.
Our findings reveal that glycine is more than a co-agonist for NMDAR and may also bind to both GlyRs and NMDARs and induce NMDAR endocytosis. Because NMDARs are involved in a variety of physiological and pathological processes throughout the central nervous system, the bidirectional modifications of NMDA function by glycine we observed here may have a general implication in understanding the regulation of nervous system function in normal and disease states. A, under physiological conditions, the level of endogenous glycine is usually low. Glycine prefers to act through the NMDAR co-agonist binding site to induce NMDAR exocytosis and resultant potentiation of NMDAR function. B, under some pathophysiological states, the dysfunction of the glycine transporter causes accumulation of glycine in the synaptic cleft. Excessive glycine can then be recruited to two additional sites. Glycine spills over to neighboring extrasynaptic sites and activates GlyRs located there. As a result, glycine induces NMDAR endocytosis and depression of NMDAR response.