D-Serine, the endogenous coagonist of N-methyl-D-aspartate receptors (NMDARs), is considered to be an important gliotransmitter, and is essential for the induction of long-term potentiation. However, less is known about the role of D-serine in another form of synaptic plasticity, long-term depression (LTD). In this study, we found that exogenous D-serine regulated LTD in the hippocampal CA1 region in a “bell-shaped” concentration-dependent manner through regulating the function of NMDARs in the same manner, whereas endogenous D-serine was activity-dependently released and, in turn, contributed to the induction of LTD during low-frequency stimulation. Furthermore, impairing glial functions with sodium fluoroacetate (NaFAC) reduced the magnitude of LTD, which could be restored by exogenous D-serine, indicating that endogenous D-serine is mainly glia-derived during LTD induction. More interestingly, similar to the effects on LTD, exogenous D-serine enhanced spatial memory retrieval in the Morris water maze in a bell-shaped dose-dependent manner and rescued the NaFAC-induced impairment of memory retrieval, suggesting links between LTD and spatial memory retrieval. Our study thus provides direct evidence of the bell-shaped D-serine actions on hippocampal LTD and spatial memory retrieval, and underscores the importance of D-serine in synaptic plasticity, learning, and memory.
D-Serine is the most abundant D-amino acid in the mammalian brain. It is not incorporated into proteins, but rather works as a physiological coagonist of N-methyl-D-aspartate receptors (NMDARs) (Mothet et al. 2000). The conversion from L-serine to D-serine by serine racemase (SR) is the only source for endogenous D-serine in the brain (Wolosker et al. 1999). Thus, the restricted expression of SR to astrocytes leads to the major distribution of D-serine in astrocytes (Wolosker et al. 1999; Xia et al. 2004), although some SR expression and D-serine have also been found in neurons and microglia (Kartvelishvily et al. 2006; Williams et al. 2006). Furthermore, astrocytic D-serine can be released by glutamate receptor stimulation through a Ca2+- and soluble N-ethylmaleimide sensitive fusion protein attachment protein receptor-dependent exocytotic pathway (Mothet et al. 2005). This activity-dependent release of astrocytic D-serine suggests that it acts as a gliotransmitter to mediate neuron-glia cross-talk and modulates NMDAR-dependent synaptic activity. Indeed, the functions of D-serine parallel those of NMDARs (Martineau et al. 2006). For example, during early neuronal development, D-serine released by Bergmann glia seems to be essential in the NMDAR-dependent migration of granule cells (Kim et al. 2005). On the other hand, in models of neurotoxicity, D-serine is a key coagonist required for the cell death caused by NMDAR overactivation (Katsuki et al. 2004; Shleper et al. 2005). Furthermore, several studies on astrocytic D-serine focus on the astrocyte-dependent modulation of synaptic plasticity. In both neuronal cultures and slices, NMDAR-dependent long-term potentiation (LTP) is determined by the concentration of extracellular D-serine, which depends on the glial environment surrounding neurons (Yang et al. 2003; Mothet et al. 2006; Panatier et al. 2006). Several studies suggest that D-serine is also involved in the induction of another type of synaptic plasticity, long-term depression (LTD) (Panatier et al. 2006; Duffy et al. 2007). However, the mechanism underlying D-serine modulation of LTD remains unknown.
Synaptic plasticity has been proposed as a cellular mechanism of learning and memory in the brain, but most studies focus on the link between LTP and memory (Lynch 2004). Compared with LTP, the study of LTD has been relatively limited. One possible reason may be that the magnitude of LTD decreases with age and LTD is difficult to induce in adult hippocampal slices (Bear and Abraham 1996). More recently, hippocampal LTD has been shown to increase in stressful conditions (Xu et al. 1997). In addition, in the nucleus accumbens, LTD has been related to the expression of specific patterns of behavior, such as behavioral sensitization (Brebner et al. 2005). This suggests a more complex role for LTD in learning and memory (Braunewell and Manahan-Vaughan 2001). Furthermore, LTD changes synaptic strength which, in turn, regulates the induction of LTP, thereby contributing indirectly to information storage (Braunewell and Manahan-Vaughan 2001). Although LTP and LTD have many processes in common, LTD is not the mechanistic inverse of LTP. Thus, understanding the role of D-serine in LTD induction also has importance, perhaps equal to that of LTP. In this study, we found that exogenous D-serine regulates NMDAR-dependent LTD in the hippocampal CA1 region in a bell-shaped concentration-dependent manner, through regulating the function of NMDARs in the same manner, whereas endogenous astrocytic D-serine is released in an activity-dependent manner that modulates the induction of LTD. Behavioral experiments further showed that D-serine plays a crucial role in spatial memory retrieval. Our findings provide direct evidence of the bell-shaped D-serine actions on hippocampal LTD and spatial memory retrieval, and suggest the existence of links between LTD and spatial memory retrieval.
Materials and Methods
The care and use of animals in these experiments followed the guidelines of, and the protocols were approved by, the Institutional Animals Care and Use Committee of the Institute of Neuroscience, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences.
Hippocampal Slice Preparation
Hippocampal slices were prepared from postnatal day 13–18 male Wistar rats. In brief, rats were decapitated and brains were quickly removed and placed in ice-cold, oxygenated (95% O2/5% CO2) artificial cerebrospinal fluid (ACSF). Coronal hippocampal slices (400 μm thick) were cut with a vibratome (Leica ATF-1000, Leica Instruments Ltd, Wetzlar, Germany) and transferred into a submersion-type holding chamber containing oxygenated ACSF at 35 °C for 1 h. Then a single slice was gently transferred into the recording chamber and held submerged between 2 nylon nets and maintained at room temperature (23–25 °C), where they were perfused throughout the experiment with oxygenated ACSF at a flow rate of 3–5 mL/min. The same ACSF was used in cutting, incubation and recording, and contained (in mM): NaCl, 120; KCl, 2.5; NaHCO3, 26; NaH2PO4, 1.25; CaCl2, 2.0; MgSO4, 2.0; and D-glucose, 10.
Extracellular recordings were made in the CA1 region of the hippocampus at room temperature (23–25 °C). A bipolar platinum-iridium stimulating electrode was placed among the Schaffer collateral axons to elicit field excitatory postsynaptic potentials (fEPSPs), which were recorded from the CA1 stratum radiatum using a glass microelectrode (1–3 MΩ) filled with 3 M NaCl. The recording electrodes were pulled from borosilicate glass tubing (1.5 mm outer diameter, 0.84 mm inner diameter; World Precision Instruments, Inc., Sarasota, FL) with a Brown-Flaming micropipette puller (P-97; Sutter Instruments Co., Novato, CA). Stimuli (0.1 ms in duration) were delivered every 30 s. Stable basal responses were obtained for at least 10 min. The stimulus intensity during baseline recording was adjusted to evoke approximately 50% of the maximum response. LTD was induced by low-frequency stimulation (LFS; 900 pulses, 1 Hz). Stimulus intensity during LFS was adjusted to evoke approximately 50% of the maximum response unless otherwise stated. In some experiments, the stimulus intensity during LFS was adjusted to evoke approximately 30% or 80% of the maximum response. To record NMDAR-mediated fEPSP, stimulus intensity was adjusted to evoke approximately 50% of the maximum response and then 10 μM 6-cyano-7-nitroquinoxaline-2,3-dione was added to block α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor-mediated component. Field potentials were amplified by an Axonpatch-200B amplifier (Axon Instruments, Foster City, CA) and filtered at 5K Hz. Data were acquired and analyzed using a Digidata1322A interface and Clampfit 9.0 software (Axon Instruments). The magnitude of LTD was determined 30 min after LFS and expressed as percentage reduction in fEPSP slope relative to baseline.
Enzyme-Linked Immunosorbent Assay
The sample solution of about 1 mL was collected from the recording chamber during baseline recording and again during LFS. The concentrations of D-serine in samples were assayed by competitive enzyme-linked immunosorbent assay (ELISA). To prepare antigen, 80 mg bovine serum albumin (BSA) and 4 mg D-serine were dissolved in 4 mL of phosphate-buffered saline (PBS) and then 2 mL of glutaraldehyde (0.02 M) was gently added. The mixed solution was dialyzed in 500 mL of PBS, which was replaced 4 times every 6–12 h, then used as stock solution for coating and stored at −60 °C. Stock solution (200 μL) was diluted 1:50 with 10 mL of PBS and then coated into 96-well plates (100 μL for each well) for 16 h at 4 °C. Then the coated solution was removed and each well was washed 3 times with PBS. The 96-well plates were dried and stored at 4 °C. Before use, each well in the plate underwent blocking by BSA for 1 h at room temperature and then was washed twice and dried. The samples or standard were mixed with anti-D-serine antibody (1:1500) (Sigma, St Louis, MO), and then shaken for 60 min at 37 °C. Then this mixed solution was added to the 96-well plate. After 90 min of shaking at room temperature, the plate was washed 5 times with PBS and dried. The plate was then incubated with secondary rabbit antibody conjugated to horseradish peroxidase (HRP) (1:1000) at room temperature for 90 min. After 5 washes with PBS, HRP substrate was added to produce a color reaction. The absorbance was read in a plate-reading spectrophotometer and the concentrations of D-serine in samples were determined by comparison with the standard curve (Fig. 3C). To avoid the system error, we reproduced the standard curve before each measurement.
The Morris water maze consisted of a circular pool (250 cm diameter, 60 cm deep) filled with water at 24–26 °C to a depth of 20 cm. The water surface was covered with floating black resin beads. Yellow curtains were drawn around the pool (50 cm from the pool periphery) and contained distinctive visual marks that served as distal cues. Before training, a 180 s free swim trial without the platform was run. For training, a submerged (1.5 cm below the surface of the water, invisible to the animal) Perspex platform (13 × 13 cm) was fixed in the center of a quadrant so that the animal had to learn the location of the platform, which was the only escape from the water. Adult Wistar rats (150–200 g) were given 8 trials with starting positions that were equally distributed around the perimeter of the maze in an acquisition session, and for 3 days were allowed to learn the position of the platform. A trial was terminated when the rat had climbed onto the escape platform or when 120 s had elapsed. Each rat was allowed to stay on the platform for 30 s. The probe trial was performed on the fourth day. The platform was removed and the rat behavior was recorded for 120 s. Swimming paths for training session and probe trial were monitored using an automatic tracking system. This system was used to record the swimming trace and calculate the time spent in each quadrant. In the drug-treated groups, sodium fluoroacetate (NaFAC) was intraperitoneally injected 6 h before, whereas D-serine and 7-chlorokynurenic acid (7-ClKY) was injected 3 h before the memory retrieval task.
All drugs and chemicals in these experiments were purchased from Sigma. In the experiments on slices, for bath application, the stock solutions of drugs were diluted into the bathing medium and the given concentration could be reached in the recording chamber after 3–5 min; for “rapid” application, we directly replaced the ACSF flowing into the chamber by ACSF containing drugs with a given concentration at the entrance of the recording chamber, so that the drugs at a given concentration could reach the slices immediately. All the data are shown as the mean ± SEM. Comparisons of 2 groups were made by Student's t-test. All statistical analysis was performed using Origin 7.0 (OriginLab, Northampton, MA).
Exogenous D-Serine Regulates NMDAR-Dependent LTD in a Bell-Shaped Concentration-Dependent Manner
To define the role of D-serine in LTD, we first examined the effects of exogenous D-serine on LFS-induced LTD. LFS (900 pulses, 1 Hz) with the 50% stimulus intensity induced an obvious LTD in the hippocampal CA1 region (Fig. 1A,B). With bath application of exogenous D-serine, we found that 5 μM significantly enhanced the magnitude of LTD from 19.3 ± 3.7% (n = 5) to 58.3 ± 5.6% (n = 6; P < 0.01), whereas 100 μM had no significant effect—the LTD magnitude was 21.6 ± 4.1% (n = 6; P > 0.05) (Fig. 1A,B). To further determine the exact concentration–effect relationship of D-serine on LTD, we measured its effects at various concentrations (3, 5, 10, 50, and 100 μM). As shown in Figure 1D, exogenous D-serine regulated LTD in a bell-shaped concentration-dependent manner with the effect reaching maximum at 5 μM. The NMDAR antagonist D-AP5 (50 μM) blocked LTD both in the absence and presence of D-serine (data not shown). In addition, the specific NMDAR glycine site antagonist 7-ClKY (50 μM) (Kemp et al. 1988) blocked LTD both in the absence (0.1 ± 7.5%, n = 5) and presence of 5 μM D-serine (0.5 ± 5.8%, n = 5) and 100 μM D-serine (1.5 ± 5.8%, n = 5) (Fig. 1C), indicating that the glycine site of NMDAR is required for inducing LTD whether or not exogenous D-serine was present. To ascertain that a NMDAR-dependent mechanism is involved in D-serine action, we further examined the effect of glycine. Similar to that of D-serine, glycine modulated LTD in a bell-shaped concentration-dependent manner with the effect reaching maximum at 10 μM (Fig. 1E). These results indicate that D-serine exerts its action on LTD through regulating NMDARs.
Exogenous D-Serine Regulates NMDAR–fEPSP in a Bell-Shaped Concentration-Dependent Manner
To establish the critical role of NMDARs underlying the bell-shaped D-serine action on LTD, we directly examined the effects of exogenous D-serine on NMDAR-mediated fEPSPs. As shown in Figure 2A–C,G, exogenous D-serine at 5 μM but not 100 μM significantly increased the amplitude of NMDAR–fEPSP (5 μM, 133.6 ± 7.6% normalized to control, n = 8, P < 0.01; 100 μM, 101.8 ± 2.5% normalized to control, n = 7, P > 0.05), which could be blocked by the NMDAR glycine site antagonist 7-ClKY (50 μM). The amplitudes of NMDAR–fEPSPs were measured 20 min after D-serine application, when the NMDAR–fEPSPs reached stable value. Interestingly, following the application of 100 μM D-serine, a transient increase of the NMDAR–fEPSP amplitude was observed (Fig. 2C). We attributed this phenomenon to the time lag for local D-serine to reach the maximal concentration. To test this hypothesis, we altered the method of drug application in order to speed up D-serine to reach a high level (see Methods). Under such conditions, however, we still observed first an increase, then a decrease of NMDAR–fEPSP. (Fig. 2D,G). To make sure that this modulation is dependent on the NMDAR glycine site, we examined glycine with the same rapid application method as we used for applying D-serine. Consistently, we found that glycine at low concentration (10 μM) caused a persistent increase of NMDA response, whereas glycine at high concentration also induced a first increase, and then a decrease of NMDAR–fEPSP (10 μM, 125.9 ± 8.5% normalized to control, n = 5, P < 0.01; 100 μM, 100.2 ± 10.4% normalized to control, n = 5, P > 0.05) (Fig. 2E–G). These results together suggest that exogenous D-serine also modulates NMDAR response in a bell-shaped concentration-dependent manner through a NMDAR glycine site-dependent mechanism.
Endogenous D-Serine is Activity-Dependently Released during LTD Induction
Because exogenous D-serine regulated LTD in a concentration-dependent manner, an important question was whether endogenous D-serine also participates in the induction and regulation of LTD, similar to its well-known function in LTP (Yang et al. 2003; Mothet et al. 2006; Panatier et al. 2006). To address this question, we first determined whether endogenous D-serine release occurred during the induction of LTD. Regarding the question of whether D-serine release was activity-dependent, we choose 3 different stimulus intensities to mimic different degrees of neuronal activity. As shown in Figure 3A, LFS at 3 different stimulus intensities, which were adjusted to evoke 30%, 50%, and 80% of the maximum response, induced LTD with different magnitudes (30% group, 11.5 ± 2.7%, n = 6; 50% group, 19.3 ± 3.7%, n = 5; 80% group, 43.8 ± 5.2%, n = 5). Furthermore, all these LTD induced by different stimulus intensities were blocked by the NMDAR glycine site antagonist 7-ClKY (Fig. 3B), indicating a NMDAR glycine site-dependent mechanism underlying the differences among the LTD induced by different intensities. By collecting the solutions during baseline and LFS, we assessed the concentration of extracellular D-serine in recording chamber by ELISA assay and the concentrations of D-serine in samples were determined by comparison with the standard curve (Fig. 3C). The specificity of the assay for D-serine was validated by testing 2 other amino acids γ-amino butyric acid (GABA) and glycine, and samples treated with a specific enzyme that degrades D-serine, D-amino acid oxidase (DAAO, 0.1 units/mL, from porcine kidney, Sigma) (Table 1). In all 3 groups with different stimulus intensities, the concentrations of D-serine during LFS were significantly higher than those during baseline activity (30% group, 5.69 ± 0.39 μM, baseline, 2.78 ± 0.19 μM, n = 8, P < 0.01; 50% group, 10.46 ± 0.67 μM, baseline, 2.70 ± 0.26 μM, n = 8, P < 0.01; 80% group, 14.80 ± 0.71 μM, baseline, 3.08 ± 0.27 μM, n = 8, P < 0.01) (Fig. 3D). Furthermore, the extracellular D-serine levels during LFS at higher stimulus intensities were significantly higher (50% compared with 30%, P < 0.01; 80% compared with 50%, P < 0.05) (Fig. 3D). Besides the stimulus intensity, the extracellular D-serine also depended on the stimulus frequency. High-frequency stimulation (HFS) with 100 Hz/1 s could induce a robust LTP (data not shown). During HFS with 50% intensity, the concentration of D-serine in recording chamber was significantly higher than those during baseline activity (HFS, 18.45 ± 1.68 μM, baseline, 3.15 ± 0.93 μM, n = 8, P < 0.01) and LFS with 50% intensity (HFS compared with 50% LFS, n = 8, P < 0.01). These results suggest that D-serine is activity-dependently released during the induction of LTD, and this may contribute to the observed differences in LTD magnitudes with various LFS intensities.
|Group||ACSF||Glycine (20 μM)||GABA (20 μM)||50% baseline||50% LFS||50% DAAO||50% LFS DAAO|
|OD value||0.299 ± 0.014||0.275 ± 0.005||0.285 ± 0.004||0.204 ± 0.001||0.195 ± 0.001||0.234 ± 0.005||0.241 ± 0.006|
|[D-Serine] (μM)||0.003 ± 0.003||0.003 ± 0.001||0.001 ± 0.000||6.508 ± 0.323||17.693 ± 0.623||0.367 ± 0.162||0.298 ± 0.231|
|Group||ACSF||Glycine (20 μM)||GABA (20 μM)||50% baseline||50% LFS||50% DAAO||50% LFS DAAO|
|OD value||0.299 ± 0.014||0.275 ± 0.005||0.285 ± 0.004||0.204 ± 0.001||0.195 ± 0.001||0.234 ± 0.005||0.241 ± 0.006|
|[D-Serine] (μM)||0.003 ± 0.003||0.003 ± 0.001||0.001 ± 0.000||6.508 ± 0.323||17.693 ± 0.623||0.367 ± 0.162||0.298 ± 0.231|
Endogenous D-Serine Contributes to the Modulation of LTD
The concentration of endogenous D-serine varied with LFS intensity. Thus, we supposed that exogenous D-serine may exert different effects on LTD induced by different stimulus intensities, due to variations in the levels of endogenously released D-serine during LFS. Indeed, we found that the concentration–effect relationship of D-serine on LTD differed with stimulus intensity. As shown in Figure 1A,B, 5 μM D-serine greatly increased the magnitude of 50% stimulus-induced LTD, whereas 100 μM had no significant effect. However, in 30% stimulus-induced LTD (11.5 ± 2.7%, n = 6), 5 μM exogenous D-serine had little effect on LTD (18.0 ± 5.1%, n = 5, P > 0.05 compared with 30% control) but 100 μM significantly enhanced the LTD magnitude (42.6 ± 4.6%, n = 7, P < 0.01 compared with 30% control) (Fig. 4A,C), whereas in 80% stimulus-induced LTD (43.8 ± 5.2%, n = 5), 5 μM D-serine had no significant effect (37.5 ± 5.0%, n = 5, P > 0.05 compared with 80% control) but 100 μM significantly reduced the LTD magnitude (25.0 ± 4.8%, n = 8, P < 0.05 compared with 80% control) (Fig. 4B,C). These results indicate that endogenous D-serine released during LFS participates in the induction of LTD. Thus stimulus intensity and exogenous D-serine concurrently modulated LTD, suggesting that endogenous D-serine also contributes to the bell-shaped modulation of LTD.
To directly examine the role of endogenous D-serine in LTD, we used DAAO, a specific enzyme that degrades D-serine. Similar to its effect on NMDAR-dependent LTP induction (Yang et al. 2003), bath application of DAAO for more than 45 min (0.1 units/mL) significantly reduced the magnitude of LTD induced by both 50% stimulus (control group, 19.3 ± 3.7%, n = 5; DAAO group, −0.2 ± 4.6%, n = 8) and 80% stimulus (control group, 43.8 ± 5.2%, n = 5; DAAO group, −1.9 ± 7.8%, n = 6) (Fig. 5A,B). This impairment of LTD was rescued by exogenous application of 100 μM D-serine (28.4 ± 4.2%, n = 8) (Fig. 5C). Similarly, an inhibitor of glycine transporter 1 (GlyT1), sarcosine (1 mM), was able to rescue the impairment of LTD by DAAO (22.1 ± 5.8%, n = 9) (Fig. 5C), indicating that increasing glycine overcomes the loss of D-serine and restores NMDAR activation. Therefore, the effect of DAAO was due to the specific depletion of D-serine but not glycine. Taken together, these results indicate that endogenous D-serine participates in the induction of LTD.
Astrocytes Contribute to the Release of D-Serine during LTD Induction
Because endogenous D-serine can be released in an activity-dependent manner and, in turn, contributes to the induction of LTD, we attempted to identify the source of endogenous D-serine. It is well known that D-serine is a gliotransmitter and astrocyte-derived D-serine is essential for the induction of LTP (Yang et al. 2003; Mothet et al. 2006). Thus, we first determined whether astrocytes also contribute to endogenous D-serine release during LTD induction. We used NaFAC (3 mM for 1 h), the glia-specific metabolic inhibitor (Swanson and Graham 1994) to “interfere” with glial functions in slices and found that LTD was significantly inhibited (3.8 ± 5.2%, n = 4) (Fig. 6A). Exogenous application of D-serine reversed this impairment in a concentration-dependent manner (Fig. 6B,C). D-Serine at 100 μM had the maximal effect, restoring the LTD to 29.8 ± 4.2% (n = 6, P < 0.01 compared with NaFAC group), and this effect was blocked by 50 μM 7-ClKY (8.5 ± 4.5%, n = 4, Fig. 6B). These results suggest that astrocyte-derived D-serine contributes to LTD, as it does to LTP (Yang et al. 2003; Panatier et al. 2006). Furthermore, we found that increasing the stimulus intensity to 80% could also partially reverse the NaFAC-induced impairment of LTD (19.9 ± 3.5%, n = 5) (Fig. 6D). This rescue effect was blocked by the treatment of DAAO (0.1 units/mL) (-1.1 ± 5.3%, n = 5) (Fig. 6D), suggesting that either NaFAC only partially impairs glial function or there is also a neuron-derived source of D-serine.
D-Serine Contributes to Spatial Memory Retrieval
It is widely believed that long-term synaptic plasticity serves as a cellular mechanism underlying learning and memory (Bliss and Collingridge 1993; Bear and Malenka 1994). Thus, we further examined the effects of D-serine on spatial memory retrieval by using the Morris water maze. In the probe trail after 3 days training, rats showed significantly greater preference for the target quadrant (the quadrant with platform during training session) than the opposite quadrant (the quadrant opposite platform), represented by the time spent in the quadrant (Control group: target, 69.76 ± 4.0 s; opposite, 27.35 ± 3.54 s; n = 10, P < 0.05) (Fig. 7A). Intraperitoneal injection of D-serine at various concentrations did not change the preference (100 mg/kg group: target, 64.76 ± 4.80 s, opposite, 38.16 ± 3.33 s, n = 10, P < 0.05; 1000 mg/kg group: target, 90.02 ± 3.20 s, opposite, 18.60 ± 2.83 s, n = 10, P < 0.05; 3000 mg/kg group: target, 65.53 ± 6.27 s, opposite, 32.37 ± 7.10 s, n = 10, P < 0.05). However, D-serine improved memory retrieval only at an intermediate dose of 1000 mg/kg, when more time was spent in the target quadrant (1000 mg/kg group compared with control group without D-serine injection, n = 10, P < 0.05) (Fig. 7A,B), which was in parallel with the bell-shaped actions of D-serine on LTD induction (Fig. 1D). This effect was completely abolished by coinjection of 7-ClKY (15 mg/kg) (Fig. 7A), suggesting that D-serine exerts its effects through the glycine site of NMDARs. Interestingly, 7-ClKY also blocked the rat's preference for the target quadrant (Target, 35.70 ± 6.82 s; opposite, 29.98 ± 6.52 s; n = 8, P > 0.05) (Fig. 7A), suggesting that endogenous D-serine derived from glial cells may play a role in establishing the target preference, as it does in LTD induction (Fig. 1C). Consistent with this assumption, intraperitoneal injection of 3 mg/kg NaFAC impaired memory retrieval by abolishing the preference for the target quadrant (control group: target, 66.30 ± 10.54 s, opposite, 13.18 ± 6.52 s, n = 7, P < 0.05; NaFAC group: target, 41.81 ± 5.02 s, opposite, 32.77 ± 7.49 s, n = 7, P > 0.05), which could be restored by coinjection of 1000 mg/kg D-serine (target, 60.21 ± 3.62 s; opposite, 28.24 ± 5.27 s; n = 7; P < 0.05) (Fig. 7C,D). Again, coinjection of 7-ClKY (15 mg/kg) diminished the rescue effects of D-serine on the re-establishment of the target preference (target, 32.97 ± 6.53 s; opposite, 36.47 ± 5.42 s; n = 7; P > 0.05) (Fig. 7C). Together, these results suggest that both exogenous and endogenous D-serine, contributes to spatial memory retrieval in the same manner as it affects LTD.
As the most common coagonist of NMDARs, it is not surprising that exogenous D-serine can regulate hippocampal LTD through modulating the functions of NMDARs. However, we found that exogenous D-serine modulated LTD in a bell-shaped concentration-dependent manner. The maximal LTD was only achieved at an intermediate concentration of D-serine. Interestingly, in the Morris water maze, we found that D-serine enhanced spatial memory retrieval in a similar profile to its effects on LTD. The mechanisms underlying D-serine's bell-shaped actions are currently unclear. There are several possible explanations for this phenomenon. The most obvious one is that the dose-dependence may be accounted for by the Bienenstock–Cooper–Munro (BCM) theory (Bienenstock et al. 1982), which incorporates a bell-shaped curve for LTD induction. In the BCM model, different stimulus intensities reflect different Ca2+ levels in postsynaptic neurons (Nishiyama et al. 2000). The induction of LTD needs a suitable level of Ca2+ influx to activate the specific downstream pathways, a level lower than that for LTP (Nishiyama et al. 2000; Franks and Sejnowski 2002). Thus, only a certain degree of NMDAR activation is needed for LTD induction. In the case of LTD induced by a 50% stimulus, it is possible that exogenous D-serine at 5 μM increases the magnitude of LTD, whereas at 100 μM D-serine may decrease LTD during the transition from LTD to LTP according to the BCM theory. However, our results showed that exogenous D-serine modulated NMDAR-mediated fEPSPs also in a bell-shaped manner, indicating that D-serine at high concentrations does not cause more activation of NMDARs. Therefore, the bell-shaped D-serine action on LTD may not represent a high [Ca2+] shift to inducing LTP. We noted that glycine and D-serine at high concentrations can prime NMDAR internalization through their specific interactions with the glycine-binding site (Nong et al. 2003). Interestingly, D-serine and glycine at high concentrations modulated NMDAR response in a biphasic mode, with first an increase and then a decrease (Fig. 2). The result suggests that the activation of NMDARs is required for the secondary decline of NMDAR response, which is consistent with the notion that a prebinding of D-serine at high concentration is needed for the NMDAR internalization (Nong et al. 2003). The secondary reduction of NMDAR response might be counteracted by D-serine enhancement of the residual uninternalized NMDARs, leading to a sustained NMDAR response comparable to the value observed in control conditions without D-serine treatment. Accordingly, it is possible that D-serine–induced NMDAR internalization could account for its bell-shaped actions on LTD and memory retrieval. Alternatively, a recent study suggests that D-serine specifically regulated NMDA-NR2B receptor-dependent hippocampal LTD (Duffy et al. 2007), consistent with the newly proposed “NMDAR NR2A/2B subtype” theory underlying bidirectional synaptic plasticity (Liu et al. 2004; Fox et al. 2006). Regardless of its underlying mechanisms, the bell-shaped concentration–effect relationship between D-serine and LTD is likely to be important, by providing a novel mechanism for activity-dependent homeostatic regulation of synaptic plasticity.
Astrocytes sense changes in neuronal activity and, in turn, regulate synaptic function by releasing a variety of neuroactive substances known as gliotransmitters (Haydon 2001; Fields and Stevens-Graham 2002; Volterra and Meldolesi 2005). Similar to glutamate and adenosine triphosphate, D-serine has been suggested to be a gliotransmitter, and there is evidence that astrocytic D-serine is essential for the induction of LTP (Yang et al. 2003; Mothet et al. 2006; Panatier et al. 2006). Our data indicates that endogenous D-serine is released in an activity-dependent manner during the induction of LTD. The magnitudes of LTD induced by different stimulus intensities paralleled measurements of D-serine released in slices. Although there could be other mechanisms underlying differing magnitudes of LTD, such as different numbers of synapses recruited and depressed, our results raise the possibility that release of endogenous D-serine regulates the induction of LTD. First, with both 50% and 80% LFS stimulus intensities, degradation of endogenous D-serine by DAAO could block the induction of LTD (Fig. 5). Second, stimulus intensity and exogenous D-serine cooperatively modulated LTD (Fig. 4C). Third, exogenous D-serine restored LTD in slices treated with the astrocytic metabolic inhibitor NaFAC, and did so in a concentration-dependent manner. Finally, we found that increased stimulus intensity during LFS compromised NaFAC-induced impairment of LTD, and DAAO diminished this rescuing effect, consistent with an elevated D-serine release with increased stimulus intensity.
Incomplete blockade of LTD by astrocytic poisoning could be due to incomplete depletion of glial D-serine, or indicate additional neuronal release of D-serine (Kartvelishvily et al. 2006). It is reasonable to speculate that a suitable level of endogenous D-serine is needed for LTD induction. Besides D-serine, glycine is also a ligand for the glycine-binding site (Thomson et al. 1989). Both D-serine (Yang et al. 2003; Mothet et al. 2006; Panatier et al. 2006) and glycine (Martina et al. 2004; Zhang et al. 2007) have been implicated in regulating LTP induction. In our study, we found that, after DAAO treatment, application of glycine transporter GlyT1 inhibitor, sarcosine, was effective in restoring DAAO-impaired LTD, indicating the involvement of glycine site in NMDAR in LTD. Together, our findings support the idea that endogenous D-serine is a key coagonist of NMDAR in the induction of LTD.
In parallel with the bell-shaped relationship between D-serine and LTD, we found that intraperitoneal injection of D-serine also affected spatial memory retrieval in a bell-shaped dose-dependent manner. Furthermore, this aspect of memory was also impaired by NaFAC and restored by systematic D-serine injection. Synaptic plasticity, including both LTP and LTD at excitatory synapses, has been proposed as a cellular substrate of both information processing and memory storage (Bliss and Collingridge 1993; Bear and Malenka 1994). However, the exact roles of LTP and LTD in memory are far from clarified. Although some previous studies suggest that the balance between LTP and LTD in different pathways may determine the expression of specific patterns of behavior (Brebner et al. 2005; Goto and Grace 2005), most studies mainly focus on the correlation between LTP and memory. The observed correlation between LTD and memory retrieval in the present study, although not implying a causal relation, suggests that LTD may also be critical for learning and memory retrieval. Interestingly, we noted that another study also showed a specific D-serine action on LTD, which was accompanied by the change of spatial reversal learning behavior (Duffy et al. 2007). Although we have no idea about the relationship between these 2 tasks (memory retrieval and reverse learning) due to the different experimental systems, these results together suggest that D-serine and LTD may be involved in some specific patterns of behavior. A previous study has shown that intraperitoneal injection of D-serine at a dose of 10 mmol/kg (≈1000 mg/kg) causes a substantial rise in hippocampal D-serine (from about 0.2 to 0.35 mM) (Hashimoto and Chiba 2004). This happens to be the dose that was effective in our water maze memory retrieval testing (Fig. 7A,B). We have no way of knowing what brain regions are influenced by actions of D-serine on memory retrieval following systemic injection. Our data are reminiscent of D-serine treatment of diseases related to the hypofunction of NMDARs such as schizophrenia (Hashimoto et al. 2003) and Alzheimer's disease (Hashimoto et al. 2004). The relationship between impaired LTD and these psychiatric disorders awaits further exploration.
National Natural Science Foundation of China (No. 30621062); the National Basic Research Program of China (No. 2006CB500803); and the Knowledge Innovation Project from the Chinese Academy of Sciences (KSCX2-YW-R-35).
Note Added in Proof
Note: Added in Proof—While this paper was under review, Duffy et al. (Duffy S, Labrie V and Roder JC. 2007. D-Serine augments NMDA-NR2B receptor-dependent hippocampal long-term depression and spatial reversal learning. Neuropsychopharmacology. Forthcoming) reported similar D-serine effects on NMDA receptor-dependent hippocampal LTD.
We thank Dr. Patric K. Stanton for helpful comments, Bin Zhao for assistance with ELISA, Yue-Xiong Yang for assistance with behavioral test, Xue-Hua Sun and Ming-Hu Cui for assistance with slice preparation. Conflict of Interest: None declared.