d-Serine differently modulates NMDA receptor function in rat CA1 hippocampal pyramidal cells and interneurons

Abstract

The organization of the neuronal hippocampal network depends on the tightly regulated interaction between pyramidal cells (PCs) and interneurons (Ints). NMDA receptor (NMDAR) activation requires the binding of glutamate and co-activation of the ‘glycine site’. It has been reported that d-serine is a more potent endogenous agonist than glycine for that site. While many studies have focused on NMDAR function in PCs, little is known regarding the modulation of NMDARs in Ints. We studied the modulatory effect of d-serine on NMDAR EPSCs in PCs and in stratum radiatum Ints using whole-cell patch-clamp recording in rat acute hippocampal slices. We found that d-serine enhances NMDAR function and differently modulates NMDAR currents in both cell types. The augmentation of NMDAR currents by d-serine was significantly larger in PCs compared with Ints. Moreover, we found differences in the kinetics of NMDAR currents in PCs and Ints. Our findings indicate that regulation of NMDAR through the ‘glycine site’ depends on the cell types. We speculate that the observed differences arise from assemblies of diverse NMDAR subunits. Overall, our data suggest that d-serine may be involved in regulation of the excitation-inhibition balance in the CA1 hippocampal region.

The N-methyl-d-aspartate receptor (NMDAR) plays a crucial role in a variety of physiological processes such as fast excitatory neurotransmission (MacDermott et al. 1986) and synaptic plasticity (Bliss & Collingridge, 1993; Malenka & Nicoll, 1999), as well as in a number of pathological conditions such as epilepsy (Czuczwar & Meldrum, 1982; Meldrum, 1985) and schizophrenia (Carlsson & Carlsson, 1990; Javitt & Zukin, 1991; Olney & Farber, 1995). The NMDAR is composed of different subunits of the NR1, NR2 (NR2A-D; Monyer et al. 1992; Meguro et al. 1992) and NR3 families (NR3A-B; Ciabarra et al. 1995; Sucher et al. 1995; Das et al. 1998; Chatterton et al. 2002). Different combinations of these subunits confer the pharmacological profile, gating properties and Mg²⁺ sensitivity to the NMDAR complex (Sucher et al. 1996).

NMDAR function is regulated by agents acting on a number of sites other than the glutamate binding site (Hollmann & Heinemann, 1994). One of these sites is the strychnine-insensitive binding site where glycine acts to allosterically facilitate the NMDAR function (Johnson & Ascher, 1987; Mayer et al. 1989; Thomson et al. 1989). d-Serine mimics the effect of glycine (Johnson & Ascher, 1987; Kemp & Leeson, 1993) and is up to three times more potent than glycine at the ‘glycine site’ (Matsui et al. 1995; Priestley et al. 1995). High levels of d-amino acids like d-serine and d-aspartate have been found in the mammalian brain, including that of humans (Hashimoto et al. 1992, 1993a,b). Immunohistochemical studies revealed the presence of d-serine in astrocytes and showed a high degree of co-localization of this amino acid with NMDARs in the forebrain (Hashimoto et al. 1993b; Schell et al. 1995, 1997; Hashimoto & Oka, 1997). The highest densities of d-serine binding sites in the brain are in the CA1 molecular layers (Schell et al. 1995). In the CA1 region of the hippocampus, where the NMDAR neurotransmission is prominent, d-serine-containing astrocytes are found in close proximity to the NR2A/B-enriched dendrites of pyramidal cells, which is consistent with a role for d-serine in regulating the ‘glycine site’ of these receptors (Schell et al. 1997). In the stratum radiatum of CA1, d-serine is most concentrated in the foot process of the astrocytes (Schell et al. 1997). Using biochemical and electrophysiological methods, Mothet et al. (2000) showed that selective degradation of endogenous d-serine with d-amino acid oxidase (DAAOX, present in astrocytes) greatly reduced NMDAR-mediated activity in brain slices and cell culture preparations. They concluded that d-serine is an endogenous modulator of the ‘glycine site’ of NMDARs and fully saturates this site at some functional synapses. However, there are still controversies regarding the saturation of the ‘glycine site’ (Danysz & Parsons, 1998) even though experiments from different laboratories, both in vivo (Salt, 1989; Wood et al. 1989; Thiels et al. 1992) and in vitro (Wilcox et al. 1996; Bergeron et al. 1998) have suggested that the ‘glycine site’ is not saturated.

The hippocampal formation is a complex network that consists of tightly regulated interaction between excitation (glutamatergic granular cells, CA1 and CA3 pyramidal cells) and inhibition (GABAergic interneurons; Woodson et al. 1989). Inhibitory interneurons play a crucial role in regulating the complex interactions between pyramidal cells, including population oscillations, plasticity, epileptic synchronization, hormonal effects and cortical development. Despite the important role of interneurons, little is known regarding their NMDAR-mediated responses to glutamatergic inputs. Multiple subtypes of interneurons have been described in the hippocampus (for review see Freund & Buzsáki, 1996). It is known that CA1 hippocampal interneurons receive two types of excitatory inputs: feedback and feedforward (Schwartzkroin & Mathers, 1978; Knowles & Schwartzkroin, 1981; Lacaille et al. 1987; Riback & Peterson, 1991; Kneisler & Dingledine, 1995a,b), depending on the subtypes and location of interneurons. In the stratum radiatum of the CA1 region, glutamatergic inputs to interneurons are predominantly from local collaterals of pyramidal cells or Schaffer collateral fibres (Freund & Buzsáki, 1996). We focused our work on interneurons of the stratum radiatum because of their well-described morphology, electrophysiological characteristics and connectivity (Freund & Buzsáki, 1996; Morin et al. 1996; Parra et al. 1998). It is noteworthy that this particular region of the hippocampus contains a high level of d-serine (Schell et al. 1997).

Previous reports described a disparity in expression of NMDAR subunit subtypes in different cell types (Monyer et al. 1994). Consequently, the regulation of the NMDAR may differ in various cell populations. The aim of the present study was to compare the modulatory effect of d-serine on NMDAR-mediated responses in hippocampal pyramidal cells vs. interneurons. We found that d-serine differently modulates the NMDAR currents in the two cell types. We discuss this finding and speculate about functional significance of NMDAR properties in interneurons and how they may affect CA1 hippocampal circuit excitability.

METHODS

Preparation of hippocampal slices

Coronal brain slices (300 μm) containing the hippocampus were obtained from Sprague-Dawley rats (21–28 days old). Prior to decapitation, the animals were anaesthetized with isofluorane, in agreement with the guidelines of the Canadian Council of Animal Care. The brain was removed and placed in an oxygenated (95 % O₂-5 % CO₂) physiological solution, artificial cerebrospinal fluid (ACSF) at 4 °C, containing (mm): 126 NaCl, 2.5 KCl, 1 MgCl₂, 26 NaHCO₃, 1.25 NaH₂PO₄, 2 CaCl₂ and 10 glucose. The osmolarity of the ACSF was adjusted to 300 mosmol l⁻¹ and the pH to 7.2. A block containing the region of interest was prepared, and sections (300 μm) were obtained with a vibrating microtome (Leica VT 1000S). The slices were stored for 1 h in an oxygenated chamber at room temperature before they were used for the experiments.

Data recording and analysis

In current-clamp experiments, whole-cell recordings were obtained with borosilicate pipettes filled with a solution containing (mm): 130 K-gluconate, 10 N-2-hydroxy-ethylpiperazine-N′-2-ethanesulphonic acid (Hepes), 10 KCl, 2 MgCl₂, 2 ATP-Mg and 0.2 GTP-tris(hydroxy-methil)aminomethane. In voltage-clamp experiments, lidocaine N-ethyl bromide (QX-314, 5 mm; Andrade, 1991) and caesium-BAPTA (10 mm) were included in the intracellular solution. D-890 (3 mm; Knoll, Ludwigshafen, Germany) was also added to the intracellular solution in order to block the currents through voltage-gated Ca²⁺ channels and partially block the voltage-gated Na⁺ and K⁺ channels (Kovalchuk et al. 2000). This allows a voltage control that is comparable to that obtained when using intracellular Cs⁺. The pH was adjusted to 7.2 and osmolarity to 280–290 mosmol l⁻¹. With this solution, the liquid junction potential was measured (10 mV) and the membrane potential (V_m) was corrected accordingly. The pipettes had a resistance of 3–7 MΩ when filled with that solution. Recordings with series resistance higher than 20 MΩ were discarded. Bridge balance was monitored regularly during the recordings. During the first minutes of recordings it was possible to identify the firing pattern of the recorded cells. To allow the drugs added in the pipette to induce their pharmacological action, a delay of 10–15 min was systematically observed prior to recording. To further minimize current attenuation, we performed some experiments with a solution containing (mm): 130 Cs⁺-methane sulphonate, 10 Hepes, 10 KCl, 2 MgCl₂, 2 ATP-Mg and 0.2 GTP-tris(hydroxy-methil)aminomethane, 5 QX-314, 10 caesium-BAPTA, and 3 D-890. No differences were observed and the data were pooled together.

Current- and voltage-clamp recordings were obtained with a Multiclamp 700A amplifier (Axon Instruments, Foster City, CA, USA) under visual control using differential interference contrast and infrared video microscopy (IR-DIC; Leica DMLFSA, Germany). Electroresponsive properties of neurons were studied by applying 500 ms current pulses from rest. The amplitude of current pulses was varied in fixed increments of 10 pA. The input resistance (R_in) was estimated in the linear portion of current-voltage plots. Whole-cell currents were recorded from individual PCs and Ints voltage-clamped at room temperature. Cells were held at −70 mV. The PCs were recorded in the pyramidal layer and the Ints in the stratum radiatum of the CA1 region of the hippocampus.

The CA3 input to CA1 pyramidal cells, mediated by the Schaffer collaterals, is glutamatergic (Amaral & Witter, 1989). Postsynaptic responses were evoked by electrical stimulation of the Schaffer collaterals with a bipolar microelectrode positioned in the stratum radiatum. The stimulation intensity, consisting of 100 μs current pulses (0.1–1 mA; 0.3–0.01 Hz), was adjusted to evoke an EPSC amplitude in the range of 60–120 pA at V_m=−70 mV. The recordings were first obtained in normal ACSF. To isolate the NMDAR-mediated component of evoked responses, we used an ACSF containing a low concentration of MgCl₂ (0.1 mm), the α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor (AMPAR) antagonist 1,2,3,4-tetrahydro-6-nitro-2,3-dioxobenzo[f]quinoxaline-7-sulphonamide (NBQX), the GABA_A receptor antagonist picrotoxin, the GABA_B receptor antagonist 3-[[(3,4-dichlorophenyl)methyl]amino]propyl]diethoxymethyl)phosphinic acid (CGP 52432) and the glycine receptor antagonist strychnine. The composition of this ACSF (low-Mg²⁺ ACSF) was identical to that described previously. The difference in concentration of MgCl₂ was replaced with an equimolar amount of CaCl₂. The concentrations of drugs applied in the perfusate were (μm): 20 NBQX, 50 picrotoxin, 10 CGP 52432 and 0.5 strychnine, 50 dl-2-amino-5-phosphonovaleric acid (AP-5), 50 7-chlorokynurenic acid and 0.5 tetrodotoxin (TTX). NBQX is highly selective for AMPAR and does not act at the ‘glycine site’ of the NMDAR (Yu & Miller, 1995). All drugs were obtained from RBI (Natick, MA, USA), with the exception of CGP 52432 and 7-chlorokynurenic acid (Tocris, Bristol, UK). The D-890 was a generous gift from Knoll (Ludwingshafen, Germany).

Local drug injections were performed by applying air pressure pulses (3–10 ms) with a picospritzer (Parker Ins., Fairfield, NJ, USA) to a patch pipette containing 100 μmN-methyl-d-aspartic acid (NMDA). NMDA was dissolved in ACSF and applied every 30 s. The ejection pipette was positioned directly above the proximal dendrites. Recordings were first obtained in ACSF to identify the cells’ firing pattern. NMDA was then applied in the presence of TTX (0.5 μm) to avoid polysynaptic phenomena.

Kinetic analysis was performed on averaged EPSCs (usually 20–25 consecutive traces). The rise times of NMDA-mediated currents were measured at 10–90 % peak. Their decays were fitted with the exponential functions: y =A_f exp(−t/τ_f) +A_sexp(−t/τ_s) for double- and y =A₁exp(−t/τ) for single-exponential decay, where A is the amplitude, τ is the decay time constant, and the subscripts ‘f’ and ‘s’ denote fast and slow components.

Data were collected using software pCLAMP 8.2 (Axon Instruments). Analyses were performed off-line with the software, IGOR (WaveMetrics Inc., Lake Oswego, OR, USA), running on a PC microcomputer. Statistical significance of the results was determined with paired t tests (two-tailed). All values are expressed as means ±s.e.m.

Morphological identification of recorded cells

The recorded cells were dialysed with Lucifer Yellow (2 mm) in the following experiments: in all the experiments in current clamp aimed at recording cells in the stratum radiatum; in some experiments in voltage clamp where D-890 was added to the intracellular solution; and in all the experiments with Cs⁺-methane sulphonate. The slices were removed from the chamber and fixed for 1–3 days in 0.1 m PBS, pH 7.4, containing 4 % paraformaldahyde. Slices were then washed in dimethylsulphoxide (DMSO) for 1 h. The cells were then visualized with a confocal laser-scanning microscope LMS 510 (Zeiss, Germany), × 10 and × 40 water immersion objectives. Scanning and image acquisition were controlled with custom software. The computer generated a three-dimensional reconstruction of the neurons.

RESULTS

A total of 70 PCs and 57 Ints were recorded in the whole-cell configuration. Because of the great heterogeneity of interneuronal population in the hippocampus, we first describe the morphological and physiological properties of the recorded neurons. Then, we analyse the modulatory effect of d-serine on the NMDAR currents in PCs and Ints.

Morphological properties of pyramidal cells and interneurons

Cells were visually identified in slices with IR-DIC and selected for recordings on the basis of their morphology and localization in specific layers of the CA1 region. PCs and Ints were recorded in the pyramidal layer and stratum radiatum, respectively. A total of 30 Ints and 20 PCs were morphological identified using Lucifer Yellow. In the experiments designed to record Ints in the stratum radiatum, the pipettes were aimed towards the ovoid somatic profile. All cells chosen in this way (n = 30) were found to be Ints with aspiny, very sparsely spiny or varicose dendrites (Freund & Buszáki, 1996; Fig. 1B3). The soma measured 24.07 ± 1.18 μm in the longest axis and 15.47 ± 0.71 μm in the shortest axis (n = 30). The dendritic branches of the Ints radiated in stellate or bitufted configuration; the longest dendrite spanned 244.68 ± 21.44 μm (n = 30) and, generally, the dendritic trees were restricted to the stratum radiatum. In one cell the dendrites extended in the stratum pyramidale. In four cases, we chose cells with an ovoid vertical-oriented soma of small diameter (∼10 μm). These four cells had morphological profiles that are similar to interneurons that innervate other interneurons and have been classified as interneuron-selective (IS) cells (Freund & Buszaki, 1996). We did not include these cells in our database.

Figure 1. Contrasting morphological and physiological properties of pyramidal cells (PCs) and interneurons (Ints) in CA1 region of the hippocampus.

A1 and B1, confocal images of one PC and one Int of CA1 pyramidal layer and stratum radiatum, respectively. The cells were recorded at the positions indicated in A2 and B2 (asterisks; confocal and DIC image superimposed). A3 and B3, confocal images of the boxed dendritic regions shown in A1 and B1, respectively. Note the spiny dendritic segment of the PC (A3) in contrast with the aspiny (B3) one of the Int. The arrowheads indicate dendritic spines. A4 and B4 show the voltage responses of the same PC and Int in A1 and B1, to a series of intracellular current pulses. The current was applied at rest (−78 and −65 mV, respectively). Abbreviations: alv, alveus; so, stratum oriens; sp, stratum pyramidale; sr, stratum radiatum; slm, stratum lacunosum moleculare. Calibration bars in A2 and A4 are applied to B2 and B4, respectively.

All PC (n = 20) cell bodies were in the stratum pyramidale. The soma measured 27.3 ± 2.04 μm in the longest axis and 17.6 ± 2.18 μm in the shortest axis (n = 10). By contrast with Ints, these cells had spiny dendrites (Fig. 1A3). The apical dendrites (592 ± 32.59 μm; n = 10) crossed the entire stratum radiatum and ended in the stratum lacunosum moleculare.

Physiological properties of pyramidal cells and interneurons

Whole-cell recordings were obtained in current-clamp mode to examine the intrinsic membrane properties of PCs (n = 39) and Ints (n = 19). The PCs generated spike trains that exhibited frequency adaptation when depolarized (Fig. 1A4). The Ints could sustain high firing rates with various degrees of, or no accommodation (Fig. 1B4). In addition, they had significantly more pronounced fast afterhyperpolarization (fAHP; 13.65 ± 0.91 mV in Ints, n = 19, vs. 4.84 ± 0.22 mV in PCs, n = 39; P < 0.005). The physiological properties of PCs and Ints are listed in Table 1. In Ints, the resting potential was significantly more depolarized (−62 vs.−76 mV in PCs, P < 0.05), the action potential amplitude was significantly smaller (80 vs. 99 mV in PCs, P < 0.05), input resistance was significantly larger (346 vs. 99 MΩ in PCs, P < 0.05), while the duration of the action potential at half-amplitude was not significantly different (1.3 vs. 1.1 ms in PCs, P > 0.05).

Table 1.

Physiological properties of pyramidal cells and interneurons

	Physiological properties					Effect of D-serine on NMDARs
Cell type	Resting Vm	Rin	τ	Spike amplitude	Spike duration at half-amplitude	Electrical stimuli	Pressure application of NMDA
	(mV)	(Mω)	(ms)	(mV)	(ms)	(% increase)	(% increase)
Pyramidal cells	−76.25 ± 0.62 n = 39	99.26 ± 1.88 n = 39	25.49 ± 1.78 n = 39	99.26 ± 1.88 n = 39	1.17 ± 0.05 n = 39	51.66 ± 7.89 n = 17	49.20 ± 8.29 n = 10
Interneurons	−62.11 ± 1.54 n = 19	346.85 ± 21.63 n = 19	63.03 ± 5.0 n = 19	80.24 ± 1.89 n = 19	1.38 ± 0.10 n = 19	25.07 ± 5.65 n = 12	28.15 ± 3.84 n = 10

Values are means ± s.e.m. Electrical stimuli: electrically evoked NMDAR currents in a low-Mg²⁺ ACSF (0.1 idm) in the presence of NBQX (20 μm), picrotoxin (50 μm), CGP 52432 (10 μm) and strychnine (0.5 μm). Pressure application of NMDA: responses to local pressure application of NMDA (100 μm) in the presence of TTX (0.5 μm).

Effect of d-serine on evoked NMDAR-mediated currents

Previous studies of mRNA expression have shown that PCs and Ints have differences in their respective NMDAR subunit composition (Monyer et al. 1994). Since the different combinations in subunits confer the pharmacological profile, gating properties and Mg²⁺ sensitivity to the NMDAR complex (Sucher et al. 1996), we investigated the action of d-serine on the evoked NMDAR-mediated current in PCs and Ints. To evoke postsynaptic glutamatergic currents, the Schaffer collaterals were stimulated with a bipolar electrode. The evoked postsynaptic currents were recorded at V_m=−70 mV, both for PCs and Ints. The NMDAR-mediated component of the postsynaptic currents was pharmacologically isolated using the following drugs: NBQX (20 μm), picrotoxin (50 μm), CGP 52432 (10 μm) and strychnine (0.5 μm) to block AMPA-, GABA_A-, GABA_B- and glycine-mediated responses, respectively. These drugs were included in a low-Mg²⁺ ACSF (see Methods).

d-Serine had no effect on 5 out of 22 PCs and 3 out of 15 Ints. Application of d-serine (10 μm) induced a significant increase in the amplitude of NMDAR currents of 51.66 ± 7.89 % (n = 17; P < 0.005) in PCs (Fig. 2A1 and A3) and 25.07 ± 5.65 % (n = 12; P < 0.005) in Ints (Fig. 2B1 and B3) without affecting the τ_act (see Table 2). The change in the current amplitude during d-serine application was significantly larger in PCs than in Ints (P < 0.05). In control condition, the peak of NMDAR current was measured at 20.31 ± 2.62 ms and 14.83 ± 2.68 ms in PCs (n = 17) and Ints (n = 12), respectively. Due to the current amplitude increase in the presence of d-serine, the peak was significantly shifted to the right (3.48 ± 1.01 ms in PCs, n = 17, and 4.93 ± 2.12 ms in Ints, n = 12; P < 0.05). NMDAR currents were completely blocked when AP-5 (50 μm) was applied simultaneously with d-serine in PCs (n = 5) and Ints (n = 3; data not shown). Overall, these results suggest that the effect of d-serine in PCs and Ints is mediated via activation of NMDARs that have different properties.

Figure 2. Effect of d-serine on the NMDAR currents in PCs and Ints.

Neurons were recorded in a low-Mg²⁺ ACSF (0.1 mm) in the presence of NBQX (20 μm), picrotoxin (50 μm), CGP 52432 (10 μm) and strychnine (0.5 μm). Each trace is an average of 20 traces. A and B, responses evoked by bipolar electrical stimuli at V_m=−70 mV in a pyramidal cell (PC) and in an interneuron (Int). A1 and B1, the NMDAR currents recorded in low-Mg²⁺ ACSF (thin line), during application of d-serine (10 μm; thick line), and after recovery from the drug (dotted line), are superimposed. A2 and B2, decay time course of NMDAR currents in low-Mg²⁺ ACSF (thin line) and during application of d-serine (10 μm; thick line). The traces of the NMDAR currents in low-Mg²⁺ ACSF (thin line) are normalized to the amplitude of the traces after the application of d-serine (thick line). The decay of the NMDAR currents in Ints was best fitted with two exponentials (τ_f and τ_s, fast and slow decay components). Note that the decay of the NMDAR currents in the PC during the application of d-serine (10 μm) is slower than in low-Mg²⁺ ACSF. The same was observed for the τ_f in the Int. A3 and B3, histograms of the effect of d-serine (10 μm) on NMDAR currents (51.66 ± 7.89 % in PCs, n = 17 and 25.07 ± 5.65 % in Ints, n = 12). The asterisk indicates that the difference in percentage due to the application of d-serine on PCs compared with Ints is significant (P < 0.005). The values of the Ints’τ_f and τ_s are in the range of 30 to 52.4 ms and 43 to 387 ms for control condition, and 39 to 86.8 ms and 55 to 400 ms following the application of d-serine.

Table 2.

Rise and decay time constants of the NMDAR currents in CA1 pyramidal cells and interneurons

	Low-Mg2+ ACSF			D-serine
		Inactivation			Inactivation
Cell types	τact (ms)	τf (ms)	τs (ms)	τact (ms)	τf (ms)	τs (ms)
Pyramidal cells	13.18 ± 2.64 n = 17	56.62 ± 7.44 n = 17	—	13.28 ± 2.68 n = 17	71.52 ± 6.46 n = 17	—
Interneurons	11.46 ± 2.49 n = 12	36.83 ± 2.76 n = 12	242.4 ± 33.79 n = 12	12.94 ± 1.88 n = 12	55.78 ± 5.38 n = 12	244.0 ± 31.35 n = 12

Values are means ± s.e.m. Low-Mg²⁺ ACSF: electrically evoked NMDAR currents were recorded in a low-Mg²⁺ ACSF (0.1 mM) in the presence of NBQX (20 μm), picrotoxin (50 μm), CGP 52432 (10 μm) and strychnine (0.5 μM). The fast and the slow decay components are designated by τ_f, and τ_s, respectively. D-serine: the ACSF was the same as in Low-Mg²⁺ ACSF plus D-serine (10 μm). In PCs and Ints, the τ_f was significantly increased by the application of D-serine (P < 0.05).

The enhancing effect of d-serine on the amplitude of NMDAR currents isolated in ACSF-containing picrotoxin (50 μm), CGP 52432 (10 μm), strychnine (0.5 μm) and NBQX (20 μm) was also observed at +30 mV. Ints showed a significant increase of 18.97 ± 3.06 % (n = 19; P < 0.005) and PCs of 33.31 ± 2.62 % (n = 9, P < 0.005; Fig. 3A1 and B1) in the amplitude of NMDAR currents. Moreover, the voltage-dependent properties of the NMDAR-mediated component of the evoked EPSCs in PCs (Fig. 3A2) and Ints (Fig. 3B2) were not changed by d-serine application.

Figure 3. Effect of d-serine on the voltage-dependent properties of the NMDAR currents in PCs and Ints.

Neurons were recorded in normal ACSF in the presence of picrotoxin (50 μm), CGP 52432 (10 μm), strychnine (0.5 μm) and NBQX (20 μm). Each trace is an average of five traces. A1 and B1, the NMDAR currents of one PC and one Int shown before (thin line) and during the application of d-serine (10 μm; thick line), at the indicated membrane potentials, are superimposed. A2 and B2, peak current-voltage relations are shown for PCs and Ints before (▪) and during (•) the application of d-serine. The peak current-voltage relations are averages of three PCs and three Ints.

We then investigated the effect of d-serine (10 μm) over time (Fig. 4) in the two cell types. Electrical stimulation of the Schaffer collaterals was applied every 5 s. We observed that d-serine induced its maximal effect more rapidly in Ints (Fig. 4B1) than in PCs (Fig. 4A1). d-Serine produced its maximal effect in PCs (n = 8) and Ints (n = 6) after ∼2 and ∼1 min, respectively (Fig. 4). We recorded PCs (n = 2) and Ints (n = 2) for more than 40 min in the presence of d-serine. The effect of d-serine was still present after 40 min in PCs (n = 2; see one example in Fig. 4A2) and Ints (n = 2; see one example in Fig. 4B2).

Figure 4. Effect of d-serine on the NMDAR currents in PCs and Ints as a function of time.

Neurons were recorded in a low-Mg²⁺ ACSF (0.1 mm) in the presence of NBQX (20 μm), picrotoxin (50 μm), CGP 52432 (10 μm) and strychnine (0.5 μm). A1 and B1, graphs plotting the normalized NMDAR-mediated current (NMDAR-mc) amplitude (average of the control) as a function of time for PCs (n = 8) and Ints (n = 6). A2 and B2, one PC and one Int showing the effect of d-serine over a period of 40 min. d-Serine was applied at time zero (arrowhead).

To rule out the effect of d-serine on AMPAR, we pharmacologically isolated the AMPAR-mediated component of the evoked EPSCs using AP-5 (50 μm), picrotoxin (50 μm), CGP 52432 (10 μm) and strychnine (0.5 μm). d-Serine (10 μm) did not have any significant effect on the AMPAR-mediated component of the evoked EPSCs (n = 4 PCs and n = 2 Ints; data not shown).

Effect of d-serine on responses to local pressure application of NMDA

To study further the direct action of d-serine on the postsynaptically activated NMDARs, we measured the effects of this amino acid on the response to local application of NMDA (100 μm) on PCs and Ints (see Methods). NMDA (100 μm) was applied in a low-Mg²⁺ ACSF (0.1 mm) in the presence of TTX (0.5 μm). The amplitude of the currents in response to NMDA application was 41.23 ± 6.85 pA (n = 10) in PCs and 30.14 ± 4.78 pA (n = 10) in Ints. Application of d-serine (10 μm) induced a significant enhancement of the evoked NMDA responses in PCs and Ints (Fig. 5 and Table 1). d-Serine had no significant effect on 6 out of 16 PCs and 4 out of 14 Ints. Application of AP-5 (50 μm) completely abolished the responses in PCs (n = 3) and Ints (n = 3; data not shown). The change in the amplitude following the application of d-serine was significantly more pronounced in PCs than in Ints (P < 0.05).

Figure 5. Effect of d-serine on the responses of one PC and one Int to local pressure application of NMDA.

The NMDA (100 μm) was applied through a patch pipette positioned directly above the proximal dendrites. TTX (0.5 μm) was present throughout the experiments. The cells were voltage-clamped at V_m=−70 mV. A and B, NMDA-evoked responses observed in low-Mg²⁺ ACSF (thick line), during application of d-serine (10 μm; thin line), and after recovery from the drug (dotted line) are superimposed for one PC and one Int, respectively. Each trace is an average of four traces. Time base is 1 s for A and 500 ms for B. The inset represents the same traces in B with a slower time base.

Effect of d-serine on the time course of evoked NMDAR-mediated currents

Non-pyramidal cells in the hippocampus show mRNA expression for the NR2C and NR2D subunits that has a slower decay time than that conferred by the NR2A and NR2B subunits observed in pyramidal neurons (Monyer et al. 1994). We investigated the effect of d-serine on the time course of NMDAR currents. All the values of the rise and decay time constants are shown in Table 2. NMDAR current rise times in PCs and Ints were similar and not significantly changed in the presence of d-serine. The decay of NMDAR currents in Ints and in PCs was best fitted with a biexponential (see Methods) and monoexponential function, respectively (Table 2 and Fig. 2). Application of d-serine (10 μm) significantly increased the fast decay time course of NMDAR currents (Table 2; P < 0.005; Fig. 2A2 and B2). The difference in the kinetics of NMDAR currents in PCs and Ints suggests that the receptors may have a different molecular make-up.

Effect of d-serine on the ‘glycine site’

To rule out any d-serine-mediated changes in transmitter release, we undertook a series of experiments with the paired-pulse paradigm. Paired pulses (60 ms) were delivered to the Schaffer collaterals. The second response showed facilitation under control condition (Fig. 6A, thick line), as well as during application of d-serine (Fig. 6A, thin line). The stimulations induced two postsynaptic currents showing a ratio (peak₂/peak₁) of 2.03 ± 0.25 (n = 4; Fig. 6B). Addition of d-serine (10 μm) in normal ACSF increased the amplitude of the two pulses but did not significantly change the ratio of the two paired pulses (2.25 ± 0.32; n = 4; P > 0.5; Fig. 6B). These findings suggest that d-serine does not affect transmitter release by acting on presynaptic afferents.

Figure 6. Effect of d-serine on paired pulse stimulation.

A, responses were evoked by bipolar electrical stimuli at V_m = −70 mV in PCs. Two pulses of identical intensity were delivered with an interval of 60 ms. Glutamatergic-evoked responses observed in normal ACSF (thick line), during application of d-serine (10 μm; thin line), and after recovery from the drug (dotted line) are superimposed. Each trace is an average of 20 traces. B, histogram of the ratio between the amplitude of the second and the first evoked response (2.03 ± 0.25 and 2.25 ± 0.32, n = 4, in control and in d-serine, respectively) is shown. This difference was not statistically significant (P > 0.5).

d-Serine mimics the effect of glycine (Kemp & Leeson, 1993) and is up to three times more potent than glycine at the ‘glycine site’ (Matsui et al. 1995; Priestley et al. 1995). If indeed d-serine enhances the NMDAR function through the ‘glycine site’, this effect should be prevented by the addition of a ‘glycine site’ antagonist. 7-Chlorokynurenic acid (50 μm) blocked the amplitude of NMDAR currents by 86.67 ± 1.68 % and 87.36 ± 3.92 % in PCs (n = 5; Fig. 8A) and Ints (n = 3; Fig. 7B), respectively. Application of d-serine (10 μm) for 6 min in the presence of 7-chlorokynurenic acid had no effect on NMDAR currents in PCs and Ints (Fig. 7A and B). These data are in agreement with the previous report that d-serine exerts its effect only through the ‘glycine site’ of the NMDAR (Kemp & Leeson, 1993). The lack of effect of d-serine on the residual NMDAR currents suggests that the glycine sites of these receptors are saturated by ambient glycine. Overall, these results suggest the presence of two populations of receptors, one with high affinity and the other with low affinity for glycine and/or d-serine (Kutsuwada et al. 1992; Priestley et al. 1995; Kew et al. 1998).

Figure 7. d-Serine (10 μm) has no effect when the ‘glycine site’ of the NMDARs on PCs (A) and Ints (B) is blocked by 7-chlorokynurenic acid (50 μm).

Neurons were recorded in a low-Mg²⁺ ACSF (0.1 mm) in the presence of NBQX (20 μm), picrotoxin (50 μm), CGP 52432 (10 μm) and strychnine (0.5 μm). A and B, graphs plotting the normalized NMDAR-mediated current (NMDAR-mc) amplitude as a function of time for PCs (n = 5) and Ints (n = 3). 7-Chlorokynurenic acid (50 μm) was applied at time zero and maintained throughout the experiment. d-Serine (10 μm) was applied 3 min after the application of 7-chlorokynurenic acid, for a period of 6 min. Note that in both PCs and Ints, in the presence of 7-chlorokynurenic acid, d-serine did not enhance the NMDAR currents. The insets show traces of NMDAR currents at V_m = −70 mV in low-Mg²⁺ ACSF and in the presence of 7-chlorokynurenic acid plus d-serine. Every trace is the average of 20 traces. The current bar applies to each trace.

DISCUSSION

In the present study, we found that d-serine enhanced NMDAR function in CA1 PCs and Ints. We observed that NMDAR EPSCs in the two cell types were modulated differently by d-serine and that the kinetics of NMDAR currents in PCs and Ints were different. Overall, our results suggest that PCs and Ints express NMDARs with different subunit composition.

It is generally believed that NMDARs are heteromultimeric channels composed of six different subunits (NR1, NR2A-D, NR3) as identified to date. The functional NMDARs are formed by combination of NR1 and NR2 subunits, which contain the glycine and glutamate recognition sites, respectively (for review see Dingledine et al. 1999). NR2A-D, when co-expressed with NR1, determine the pharmacological profile, the gating properties and the Mg²⁺ sensitivity of the NMDAR complex (Nakanishi et al. 1990; Nakanishi, 1992; Kutsuwada et al. 1992; Monyer et al. 1992; Hollmann & Heinemann, 1994; McBain & Mayer, 1994; Sucher et al. 1996; Danysz & Parsons, 1998). Recently, the NR3 subunit family has been described (Ciabarra et al. 1995; Sucher et al. 1995; Das et al. 1998; Chatterton et al. 2002). The molecular complexity of NMDAR is increased by post-translational modifications such as alternative splicing and RNA editing (Seeburg, 1996), which affect several functional properties of these channels (e.g. receptor deactivation and desensitization).

The glycine-binding site is located on the NR1 subunit but the affinity for glycine and d-serine is controlled by the type of NR2 subunit co-assembled with NR1 (Kutsuwada et al. 1992; Priestley et al. 1995; Kew et al. 1998). Glycine has approximately 10-fold higher affinity for NR2B-, NR2C- or NR2D- compared with NR2A-containing receptors (Buller et al. 1994; Laurie & Seeburg, 1994; Priestley et al. 1995). Similar differences in affinities have also been observed for d-serine (Matsui et al. 1995; Priestley et al. 1995). The glycine concentration in the extracellular and cerebrospinal fluids has been estimated to be in the low micromolar range (Westergren et al. 1994); however, glycine transporters (Smith et al. 1992) might reduce the glycine concentration to well below 1 μm in the local microenvironment of NMDARs (Supplisson & Bergman, 1997). Moreover, in different cells, populations of NMDARs with different affinity for glycine can co-exist (Kew et al. 1998). In the presence of hypothetical glycine concentrations (300 nm to 1 μm; Supplisson & Bergman, 1997), the NMDARs with a relatively low affinity for glycine (microscopic dissociation constant (mK_D) =≈800 nm; Kew et al. 1998) are only occupied in a range from ≈20 % to ≈65 %. With these glycine concentrations, almost every high affinity receptor (K_D= 100–500 nm; Buller et al. 1994; Laurie & Seeburg, 1994; Matsui et al. 1995; Priestley et al. 1995) would be saturated. The addition of exogenous d-serine could enhance the NMDAR responses by acting on NMDARs not occupied by the ambient glycine.

Another important factor that might be involved in the regulation of the ‘glycine site’ is the function of d-serine as a neuromodulator (Schell et al. 1997). It has been established that d-serine is released in response to glutamatergic stimulation from d-serine-containing astrocytes, located in close proximity to NMDAR-containing synapses on PC dendrites (Schell et al. 1997). This mechanism, most likely, has phasic features and is mediated by the AMPA/kainate type of glutamate receptors (Schell et al. 1995). It is not known whether the above model is applicable for d-serine release and uptake at the vicinity of NMDARs located in Ints. Other mechanisms exist and may be associated with rapid release from the cytoplasmic pool. One of these mechanisms is the reversal of the Na⁺ co-transport uptake system in response to high potassium levels (Semba et al. 1995). The time course of d-serine transients in Ints could be more tonic or more phasic in nature. Clearly, the mechanism responsible for the availability of d-serine at NMDAR-containing synapses on Ints should directly influence the magnitude of NMDAR activation. Rapid fluctuations of local d-serine concentration coupled with specific receptor sensitivity to d-serine should determine the degree of receptor saturation and the efficacy of d-serine modulation (Danysz & Parsons, 1998).

Several in vitro experiments using brain slices have claimed that the ‘glycine site’ is saturated (Fletcher & Lodge, 1988; Kemp et al. 1988; Crawford & Roberts, 1989; Ramson & Deschenes, 1989; Mothet et al. 2000; Ballard et al. 2002). In contrast, other groups using in vivo (Salt, 1989; Wood et al. 1989; Thiels et al. 1992; Danysz & Parsons, 1998) and in vitro (Wilcox et al. 1996; Bergeron et al. 1998) paradigms argued against this assumption. Indeed, it is believed that the slice preparation is traumatic for the cells and that the application of the NMDA agonists leads to the release of glutamate and glycine. This suggests that this preparation is not ideal to address the question of ‘glycine site’ saturation (Danysz & Parsons, 1998). However, even in such unfavourable conditions, it has been reported that the application of glycine and/or d-serine enhances NMDAR-mediated responses (Thomson et al. 1988; Minota et al. 1989; Wilcox et al. 1996; Bergeron et al. 1998). The reported lack of an effect of glycine and/or d-serine on NMDAR-mediated responses may be due to a limitation of the techniques used (intracellular vs. patch-clamp whole-cell technique).

In our experimental condition, d-serine increased the NMDAR-mediated responses to synaptically released glutamate and to local application of NMDA in PCs and Ints, suggesting that the extracellular concentration of glycine and d-serine might not be saturating in a portion of NMDARs. Notably, the ‘glycine sites’ of the high affinity receptors could be saturated, but not those with lower affinity.

We examined the effect of d-serine on the amplitude of the NMDAR EPSCs in PCs and Ints. We observed an enhancing effect of d-serine on the NMDAR currents following electrical stimulation at −70 mV (low-Mg²⁺ ACSF) and at +30 mV (normal ACSF); the effect was stronger in PCs compared with Ints. Moreover, d-serine had similar effects on the responses evoked by pressure application of NMDA. However, d-serine did not have any effect on the isolated AMPAR currents. We ruled out any d-serine-mediated changes in transmitter release or action on synaptic sites other than the ‘glycine site’ of the NMDAR. The strong effect of d-serine on the NMDA currents that we found is most likely mediated by NMDAR subtypes that are not saturated by background glycine or d-serine concentrations. These currents could be mediated by a subpopulation of NMDARs expressing a high degree of NR2A subunit, conferring to the receptor a low affinity for glycine and d-serine. Interestingly, d-serine enhanced the NMDAR currents in a more pronounced way in PCs compared with Ints. It is known that in the adult rat hippocampus, NR2A and NR2B mRNA are prominent in CA1 pyramidal cells, but NR2C and NR2D occur in different subsets of interneurons (Monyer et al. 1994). Thus, the stronger effect observed on PCs probably arises not only from receptors with different NMDA subunit assemblies, but also from a different ratio of NR2A:NR2B:NR2C:NR2D subunits.

Our experiments in the presence of 7-chlorokynurenic acid showed no effect of d-serine on the residual NMDAR currents. It has been reported that for channel opening, two independent glycine sites must be occupied in the presence of saturating concentrations of NMDA (Benveniste & Mayer, 1991). Current relaxation after removal of glycine (or addition of 7-chlorokynurenic acid) reflects the rate of glycine dissociation from the lower affinity site (Kew et al. 1998). The rate of dissociation of glycine varies significantly depending on the receptor affinity (Priestley & Kemp, 1993). 7-Chlorokynurenic acid (50 μm) might fully occupy the glycine low affinity receptors (mainly NR2A-containing receptors) and interfere with the d-serine binding. However, the unblocked high affinity receptors would continue to be saturated by glycine.

We observed differences in the kinetics of NMDAR currents in PCs and Ints. In accordance with previous reports, the rise time of NMDAR currents was the same in PCs and Ints (Sah et al. 1990; Hestrin et al. 1990a; Perouansky & Yaari, 1993), but the kinetics of the decay time course were different. Non-pyramidal cells in the hippocampus show mRNA expression for the NR2C and NR2D subunits that confer decreased sensitivity to Mg²⁺ (Monyer et al. 1994). On the other hand, the decay times observed on PCs are conferred by the NR2 subunits (Monyer et al. 1994). Studies with both recombinant and native systems suggest that the expression of NR1/NR2A subunits produces channels with faster deactivation than NR1/NR2B or NR1/NR2C (Vicini et al. 1998), whereas the NMDARs containing the NR1/NR2D subunits display very slow kinetics (Vicini et al. 1998; Misra et al. 2000). Our finding, that the decay time course of the NMDAR-mediated responses in PCs is faster compared with that in Ints, strongly suggests different assemblies of NMDAR subunits whose kinetics could influence the duration of the response to glutamatergic inputs. The EPSC depends largely on the activation of the non-NMDARs while the late component of the EPSC is due to the NMDAR. However, with depolarization more NMDARs are open. The delayed opening of the NMDARs is responsible for the characteristic late phase of the EPSC (Hestrin et al. 1990b; Sah et al. 1990). Because the kinetics of NMDAR currents in Ints are slower than those in PCs, the duration of the EPSC will be longer in Ints. Consequently, this will increase the time interval during which temporal summation of synaptic potential could occur. The prolonged NMDAR time course of Ints provides an increased time window for coincidence detection, which allows them the control of PC synchronization and oscillation behaviour.

The composition and properties of the NMDAR component of EPSCs in a particular cell are most likely related to its specific role in the hippocampal network. Since enhancement of the NMDAR currents induced by d-serine is more pronounced in PCs compared with Ints, d-serine may have a stronger effect on the glutamatergic responses mediated by the PCs. Dysfunction in d-serine modulation of NMDAR function in Ints could affect the recurrent inhibition on CA1 PCs. In fact, there is already evidence that NMDAR-mediated alteration in collateral inhibition may lead to a selective and partial disinhibition of PCs (Grunze et al. 1996). Abnormalities in recurrent circuits might be involved in NMDAR antagonist-dependent psychosis related to schizophrenia (Greene, 2001).

Our results demonstrate that the modulation of the NMDAR function through the ‘glycine site’ differs according to cell type. d-Serine, the most likely endogenous agonist for that site, has different effects in PCs and Ints. This difference can be explained by the diverse NMDAR subunit expression in the two cell types. In the CA1 region of the hippocampus, abnormalities in the NMDAR-dependent regulation of excitation-inhibition might be associated with the pathophysiology of several neuropsychiatric disorders.