d-Amino acids in the brain: d-serine in neurotransmission and neurodegeneration


H. Wolosker, Department of Biochemistry, B. Rappaport Faculty of Medicine, Technion-Israel Institute of Technology, Haifa 31096, Israel
Fax: +972 4 8295384
Tel: +972 4 8295386
E-mail: hwolosker@tx.technion.ac.il


The mammalian brain contains unusually high levels of d-serine, a d-amino acid previously thought to be restricted to some bacteria and insects. In the last few years, studies from several groups have demonstrated that d-serine is a physiological co-agonist of the N-methyl d-aspartate (NMDA) type of glutamate receptor – a key excitatory neurotransmitter receptor in the brain. d-Serine binds with high affinity to a co-agonist site at the NMDA receptors and, along with glutamate, mediates several important physiological and pathological processes, including NMDA receptor transmission, synaptic plasticity and neurotoxicity. In recent years, biosynthetic, degradative and release pathways for d-serine have been identified, indicating that d-serine may function as a transmitter. At first, d-serine was described in astrocytes, a class of glial cells that ensheathes neurons and release several transmitters that modulate neurotransmission. This led to the notion that d-serine is a glia-derived transmitter (or gliotransmitter). However, recent data indicate that serine racemase, the d-serine biosynthetic enzyme, is widely expressed in neurons of the brain, suggesting that d-serine also has a neuronal origin. We now review these findings, focusing on recent questions regarding the roles of glia versus neurons in d-serine signaling.


amyotrophic lateral sclerosis


α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid


in situ hybridization


long-term potentiation


N-methyl d-aspartate


N-methyl d-aspartate receptor

N-methyl d-aspartate receptors (NMDARs) are key excitatory neurotransmitter receptors in the brain and are involved in many physiological processes, including memory formation, synaptic plasticity and development [1]. The NMDARs are composed of multiple subunits and their activity is regulated by numerous mechanisms, including different ligands and interacting proteins [2]. The NMDARs display high permeability to Ca2+, which is known to play a central role in synaptic plasticity and many signal transduction mechanisms [1]. NMDAR overstimulation promotes neurotoxicity and is implicated in several pathological conditions, such as stroke and neurodegenerative diseases [3].

The NMDARs are unique in their requirement for more than one agonist to operate. Glutamate, the main NMDAR agonist, does not activate the receptors unless a co-agonist binding site located at the NR1 subunit is occupied [4,5]. d-Serine, an unusual d-amino acid present in mammalian brain, is now recognized as a physiological ligand of the NMDAR co-agonist site, mediating several NMDAR-dependent processes [6–15].

At first, the NMDAR co-agonist site was thought to be occupied by glycine. Hence, the co-agonist site is also generally referred to as the ‘glycine site’. In addition, to be essential for NMDAR activity, the co-agonist site exerts neuromodulatory roles. Thus, co-agonist binding increases the receptor’s affinity for glutamate [16], decreases its desensitization [17] and promotes NMDAR turnover by internalization [18].

Since its discovery, the role of the co-agonist site in regulating the activity of the NMDAR has been controversial. The high extracellular concentration of glycine was assumed to saturate the co-agonist site in vivo, with some supporting studies demonstrating that this site is indeed saturated in tissue slices [19,20]. However, most studies now agree that the co-agonist site is not saturated under resting conditions, indicating that co-agonist binding exerts a dynamic regulation of the NMDARs [1,21].

As the dispute about the degree of saturation of the co-agonist site became abated, a new controversy emerged regarding the identity of the physiological NMDAR co-agonist. Like glycine, d-serine is a high-affinity ligand of the co-agonist site, displaying up to threefold higher affinity than glycine [22,23]. At first, d-amino acids like d-serine were not thought to exist at significant quantities in eukaryotes. Therefore, in contrast to glycine, d-serine was not viewed as a physiological ligand of NMDARs.

The serendipitous discovery of large amounts of endogenous d-serine in the brain, by Hashimoto et al., quickly changed this view [24]. Following this discovery, studies from several laboratories have shown that endogenous d-serine is a physiological regulator of NMDARs through binding to the co-agonist site. Endogenous d-serine has been implicated in several physiological and pathological NMDAR-dependent processes, including normal NMDAR transmission and synaptic plasticity [6,7,10–12,14,15], cell migration [9] and neurotoxicity [8,25–29].

A structural explanation for the selective effects of d-serine on NMDARs comes from inspection of the crystal structure of the binding core of the NR1 subunit of NMDARs. d-Serine binds more tightly to the receptor in comparison with glycine because it makes three additional hydrogen bonds and displaces a water molecule from the binding pocket [22]. There is also unique selectivity for the d-isomer of serine, as the hydroxyl group of l-serine interacts unfavorably in the binding pocket [22].

d-Serine – a physiological co-agonist of NMDARs

d-Serine is present at very high levels in the mammalian brain and at a much lower concentration in the peripheral tissues (Fig. 1). Brain d-serine accounts for one-third of the l-serine and its levels are higher than most essential amino acids [24,30]. In contrast to l-amino acids, d-serine is not incorporated into proteins or peptides, thus constituting a free amino acid pool. Experiments of brain microdialysis show that the extracellular concentration of endogenous d-serine is twice that of glycine in the striatum and comparable to the concentration of glycine in the cerebral cortex [31].

Figure 1.

 Localization of d-serine in the rat brain. The highest d-serine densities (white areas) are observed in the forebrain. AON, anterior olfactory nuclei; C, cerebral cortex; H, hippocampus; MOL, molecular layer of the cerebellum; OT, olfactory tubercule; S, striatum; T, thalamus. Reproduced from Schell et al. [34].

Hashimoto et al. initially observed that d-serine was enriched in rat forebrain areas, where NMDARs are abundant [32]. Subsequently, immunohistochemical studies carried out by the Snyder group (Fig. 1) demonstrated that the regional distribution of d-serine in rat brain co-localized almost perfectly with that of NMDARs [33,34]. The density of d-serine is much lower in the caudal part of the brain, including the adult cerebellum and brainstem (Fig. 1). This is because of the emergence of d-amino acid oxidase in adult animals, which degrades endogenous d-serine almost completely in these regions [34,35].

In contrast with d-serine, glycine immunoreactivity is higher in the caudal areas of the brain, where the density of NMDARs is lower [36]. The inverse localizations of d-serine and glycine led Schell et al. to propose that endogenous d-serine was physically closer to NMDARs than glycine.

d-Serine is enriched in protoplasmic astrocytes, raising the possibility that it is released from astrocytes ensheathing the synapse to activate neuronal NMDARs [33,34]. d-Serine was subsequently shown to be present also in neurons, where the d-serine biosynthetic enzyme, serine racemase, is robustly expressed, indicating that d-serine also has a neuronal origin [25,37–41].

A more direct demonstration that d-serine is a physiological NMDAR co-agonist arose from experiments that employed d-serine metabolic enzymes to remove, in a selective manner, endogenous d-serine from brain slices and cultures, leaving the levels of glycine unchanged. Using this strategy, endogenous d-serine was shown to mediate a variety of physiological NMDAR-dependent events.

In a pioneer study, Snyder, Mothet et al. depleted endogenous d-serine from neural cultures by applying d-amino acid oxidase, which specifically degrades d-amino acids, but not l-amino acids. Depletion of d-serine led to a 60% decrease in the spontaneous activity attributed to the postsynaptic NMDAR, whereas α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) responses were unaffected [10]. Subsequent studies demonstrated that endogenous d-serine is required for the NMDAR-mediated light-evoked responses in the retina [6,12]. Likewise, along with glutamate, endogenous d-serine mediates the long-term potentiation of synaptic activity in the hippocampus, a region associated with learning and memory [7,14,15,42].

One concern with the experiments employing d-amino acid oxidase to deplete d-serine is that the enzyme displays a very low affinity for d-serine – about 50 mm at physiological conditions [43]. As the affinity of NMDARs to d-serine is at least five orders of magnitude higher than that of d-amino acid oxidase, it is conceivable that a significant fraction of endogenous d-serine remains bound to the receptors. Furthermore, some commercial preparations of d-amino acid oxidase contain many impurities that may negatively affect the tissue preparation, including large quantities of d-aspartate oxidase, which quickly degrades N-methyl d-aspartate (NMDA) itself [28].

To overcome some of the limitations of the d-amino acid oxidase treatment, new enzyme preparations were developed to allow direct comparison between the effects of d-serine and glycine in stimulating NMDARs. One of these new enzyme preparations is the recombinant bacterial d-serine deaminase enzyme. This enzyme displays both high affinity and high specificity to d-serine, and efficiently degrades it in organotypic hippocampal slices [28], neuronal cultures [25] and retina preparations [6]. We found that depletion of endogenous d-serine by d-serine deaminase virtually abolished NMDA-elicited neurotoxicity in organotypic hippocampal slices (Fig. 2). This indicates that d-serine, and not glycine, is the dominant co-agonist required for the neuronal cell death elicited by NMDAR stimulation in the hippocampus (Fig. 2). Likewise, the essential role of d-serine for NMDAR neurotoxicity was also observed in cortical slices subjected to ischemic cell death induced by oxygen/glucose deprivation [8,26].

Figure 2.

 The role of endogenous d-serine in NMDAR-elicited neurotoxicity. The removal of d-serine by d-serine deaminase enzyme (DsdA) completely prevented NMDA-elicited cell death. (A) Control. (B) NMDA (500 μm) elicited robust cell death in all hippocampal areas, as measured by propidium iodide (PI) uptake. (C) Control treated with DsdA (10 μg·mL−1 for 90 min). (D) Destruction of d-serine by DsdA protected against NMDA-elicited cell death. (E) The NMDA effect was prevented by addition of the antagonist MK-801. (F) The NMDA effect was not prevented by addition of the AMPA receptor antagonist, DNQX. Densitometric analysis of PI uptake revealed almost complete neuroprotection by the removal of endogenous d-serine. Reproduced with slight modifications from Shleper et al. [28].

The dominant role of d-serine for NMDAR activity observed in neurotoxicity experiments is also supported by electrophysiological experiments. d-Serine is essential for the NMDAR-mediated light-evoked responses in the rat retina, shown by Miller et al. to contain endogenous d-serine [6,12,44]. A similar dominant role for endogenous d-serine in NMDAR transmission was observed in the supra-optic nucleus of the hypothalamus [11]. In this study, Panatier et al. demonstrated that a more efficient recombinant d-amino acid oxidase preparation destroyed d-serine from hypothalamic slices and blocked the NMDAR responses. By contrast, the degradation of endogenous glycine by a glycine oxidase enzyme had no effect, suggesting that d-serine, rather than glycine, is the dominant NMDAR co-agonist in the supra-optic nucleus [11].

The role of endogenous d-serine as the foremost co-agonist of NMDARs, as suggested by some studies, is at odds with the very high levels of extracellular glycine [31]. In hippocampal organotypic slice cultures, the removal of endogenous d-serine completely blocked the NMDAR-elicited neurotoxicity, even though extracellular glycine was present at a level was 10-fold higher than d-serine [28]. Glycine alone was very inefficient in promoting NMDAR neurotoxicity. When d-serine was completely removed from the slice cultures, the amount of glycine needed to cause maximal NMDAR neurotoxicity was two orders of magnitude higher than its dissociation constant from purified NMDARs [28]. Likewise, in hypoglossal neurons, exogenously applied d-serine was almost two orders of magnitude more effective than glycine in stimulating NMDAR responses [45].

Why is glycine much less efficient than d-serine in slice preparations? Glycine and d-serine display comparable affinities to NMDARs, and therefore the difference in their functional efficiency may be related to the availability of the two co-agonists at synaptic or extra-synaptic sites. The synaptic glycine concentration is efficiently regulated by a high-affinity glycine transport that limits glycine access to NMDAR sites [45,46]. Accordingly, the addition of a selective inhibitor of the GlyT1 transporter potentiates NMDAR responses [47] and elicits NMDAR neurotoxicity after the endogenous d-serine has been enzymatically destroyed [28].

In contrast to glycine, d-serine behaves as a poorly transported analogue. Specific d-serine transporters have not yet been identified, and neutral amino acid uptake systems capable of transporting d-serine display only low-to-moderate affinity [48–50]. It is therefore conceivable that d-serine can more easily reach synaptic or extrasynaptic NMDARs by evading the neutral amino acid re-uptake systems. Nevertheless, the relative contributions of d-serine versus glycine in mediating NMDAR transmission are still largely unknown. Further studies will be required to map the relative contributions of d-serine and glycine for NMDAR transmission in different brain regions.

The levels of d-serine are high in the cerebellum of neonatal rats, decreasing to very low levels in the third week of life as a result of the emergence of the d-amino acid oxidase enzyme, which destroys endogenous d-serine [34,35]. The transient presence of d-serine in the cerebellum coincides with the postnatal cerebellar development, in which granule cells migrate from the external to the internal granule cell layer in an NMDAR-dependent manner [51]. Blockage of the NMDAR at granule cells decreases the rate of migration [51]. Bergman glia cells, which contain high levels of endogenous d-serine, serve as a scaffold for granule cell migration. Endogenous d-serine, presumably released by the Bergman glia, mediates the NMDAR-dependent neuronal migration in the cerebellum [9]. As migrating granule cells do not make conventional synaptic connections, the modulatory action of glial-released d-serine reflects a novel mechanism for neuromodulation [9].

Origin of brain d-serine

The role of d-serine as a possible regulator of NMDARs was initially viewed with scepticism, or even ignored, for d-amino acids were not thought to be synthesized in mammals. The discovery of the biosynthetic enzyme for d-serine paved the way for additional advances in the field. We found that endogenous d-serine is synthesized from l-serine by serine racemase, a brain-enriched enzyme [52–54]. Serine racemase requires pyridoxal 5′-phosphate as a cofactor and, in addition to racemization, it de-aminates l-serine into pyruvate and ammonia [52,55]. The enzyme is unique among the pyridoxal 5′-phosphate enzymes as a result of its requirement for divalent cations and the Mg.ATP complex for its activity [52,56–58].

The regional localization of serine racemase matches those of endogenous d-serine, indicating a physiological role in d-serine synthesis [53]. Preliminary reports indicate that serine racemase knockout mice display an 80–90% decrease in brain d-serine levels, confirming the role of serine racemase as the biosynthetic enzyme for d-serine [59–62]. Serine racemase knockout animals exhibit decreased NMDAR transmission, impaired long-term potentiation of synaptic activity in the hippocampus, and are more resistant to stroke damage upon middle-cerebral artery occlusion [60,61]. These preliminary reports support previous biochemical and electrophysiological data, indicating that d-serine is indeed a physiologically relevant endogenous co-agonist of NMDARs.

Is d-serine a transmitter?

By adopting a more liberal conceptualization of a neurotransmitter, Snyder and Ferris proposed that d-serine belongs to a new class of transmitters that only partially fulfill the criteria previously used to define classic neurotransmitters [63]. The existence of a biosynthetic pathway, a target receptor, an uptake system and a degradative enzyme for d-serine favors the notion that d-serine is indeed a neurotransmitter (Table 1). Unlike classical chemical transmitters, however, d-serine was originally thought to be specifically produced and released from astrocytes, suggesting that d-serine is a glial-transmitter (also known as a gliotransmitter) [33,34,64]. A boost to the notion that d-serine is a gliotransmitter was recently provided by Mothet et al. who demonstrated that cultured astrocytes are capable of the vesicular release of d-serine [65]. In this study, AMPA receptor stimulation was shown to promote the release of d-serine by exocytosis, an effect blocked by cocanamycin A, an inhibitor of the vesicular filling of neurotransmitters that blocks the generation of the electrochemical proton gradient across the vesicles [65]. In support for a possible vesicular localization of d-serine, Pow et al. observed vesicular-like structures containing d-serine in astrocytes in situ, which may correspond to synaptic-like vesicles [66]. The question remains, however, whether the vesicular pathway in cultured astrocytes surpasses nonvesicular forms of d-serine release and whether the vesicular release of d-serine occurs in more physiological preparations or in vivo.

Table 1.   Some transmitter-like properties of brain d-serine. ASCT, alanine, serine, cysteine and thrionine transporter.
 d-SerineSelected references
OccurrenceEnriched in the mammalian brain and in some insects[24,30]
Target receptorNMDAR[10,22,93]
ActionsModulates NMDAR transmission, long-term potentiation of synaptic activity, NMDAR-elicited neurotoxicity and NMDAR-dependent cell migration[6–12,14,15,25–29]
BiosynthesisSerine racemase[52–54,56–58,94,95]
Metabolismd-amino acid oxidase enzyme and β-elimination catalyzed by serine racemase[34,55,96]
TransportNeutral amino acid transporters; Asc-1 in neurons and ASCT-like in astrocytes[49,50,97,98]
ReleaseVesicular and nonvesicular release modes described in neurons and astrocytes[25,50,65,99,100]

In order to function as a gliotransmitter, d-serine actions should depend on an intimate relationship between astrocytes and neurons. In an elegant study, Oliet et al. found that NMDAR transmission in the supra-optic nucleus depends on the degree of astrocytic coverage of neurons [11]. The neuronal centers in the supra-optic nucleus undergo an extensive reduction of astrocytic ensheathing of its neurons and synapses under conditions such as lactation. Using this model, the authors showed that lactating rats display reduced NMDAR activity compared with virgin rats as a result of reduced levels of d-serine release. The data indicate that variations in the astrocytic environment of neurons and synapses play a prominent role in the postsynaptic control of excitatory neurotransmission by releasing d-serine [11].

Key to the hypothesis that d-serine is a gliotransmitter is the notion that the electrophysiological effects of d-serine should be attributable to astrocytic rather than neuronal release of d-serine. Most studies demonstrating a role for d-serine in mediating NMDAR activity attributed its effects solely to glial d-serine and overlooked a possible neuronal origin. Although glial d-serine is prominent, a number of recent studies have reported the presence of d-serine also in neurons. Thus, purified neuronal cultures were recently shown to synthesize large amounts of d-serine [25]. d-Serine was also identified in situ by immunohistochemistry in neurons of the nervous system (Fig. 3), including the cerebral cortex [25,39], some nuclei of the hindbrain [38,39,66] and in ganglion cells of the retina [67]. Recently, Puyal et al. showed that d-serine displays a developmental glia-to-neuron switch. In the vestibular nuclei of young rats, d-serine is predominantly glial, whereas in adult rats, d-serine is exclusively present in neurons in these regions [38].

Figure 3.

d-Serine localizes to neurons and astrocytes in the brain. Staining for d-serine was performed in pyramidal neurons of layer V of the cerebral cortex and in astrocytes in the corpus callosum of a P9 rat. The lower panels depict double-labeling immunofluorescence for d-serine (labeled for SR in the original publication) and a neuronal nucleus marker (NeuN) in layer VI of the cerebral cortex of a P9 rat. Reproduced with slight modifications from Kartvelishvily et al. [25].

Neuronal d-serine in NMDAR regulation

Is there a role for neurons in synthesizing and releasing d-serine? Although present at a lower level than in astrocytes, d-serine is detectable in pyramidal neurons of the cerebral cortex in situ (Fig. 3) [25,39]. Originally regarded as an elusive or nonimportant source of d-serine, the extent of the neuronal pool of d-serine became apparent when we re-investigated the expression of serine racemase using new antibodies [25]. We observed widespread and prominent neuronal serine racemase in situ, especially in the cerebral cortex and hippocampal formation, in which neuronal serine racemase predominates (Fig. 4A–C). Furthermore, recent studies indicate that cultured neurons contain both serine racemase mRNA and protein, and catalyze the synthesis of d-serine to levels comparable to that observed with astrocytes [25,40,41]. Neuronal staining for serine racemase was also recently observed in ganglion cells of the retina [67].

Figure 4.

 Localizations of serine racemase protein and mRNA in the brain. (A) Staining for serine racemase in the cerebral cortex (Ctx) of a P7 rat (layers IV–VI). (B) Staining of neurons in the stratum pyramidale (Pyr) of the CA1 region of the hippocampus. (C) Staining for serine racemase in the pyramidal cell layers and dentate gyrus of the hippocampus. (D) In situ hybridization (ISH) for serine racemase in the hippocampus of adult mice, showing the highest serine racemase mRNA levels in pyramidal cell and dentate gyrus layers (saggital image series 392945, Srr_110, Allen Brain Atlas). (E) ISH of adult mouse brain (saggital image series 392945, Srr_110, Allen Brain Atlas). (F) Dark-field ISH for serine racemase in the hippocampus of adult mice using a 33P-labelled RNA probe and silver grain emulsion (saggital image 38687, Gensat project). (G) Dark-field ISH for serine racemase in the hippocampus of adult mice (coronal image 36854, Gensat project). (A–C) Reproduced with slight modifications from Kartvelishvily et al. [25]. (D–E) ISH images from the Allen Institute of Brain Science [101,102], and the Gensat project [103]. bs, brainstem; cb, cerebellum; cc, corpus callosum; ctx, cerebral cortex; DG, dentate gyrus; H or Hipp, hippocampus; ob, olfactory bulb; st, striatum.

The neuronal expression of serine racemase was confirmed by in situ hybridization (ISH), which revealed prominent serine racemase mRNA in neurons of the brain [41] and in neuronal ganglion cells of the retina [67]. Like the immunohistochemistry for serine racemase, the ISH of rat brain shows striking neuronal predominance [41]. The neuronal-like distribution of serine racemase mRNA in the hippocampus is evident, with little or no serine racemase message in astrocytes at the corpus callosum, as revealed by ISH from both the Allen Institute for Brain Science and the Gensat project (Fig. 4D–G).

Does neuronal d-serine activate NMDARs? We found that endogenous d-serine released by neuronal cultures lacking significant levels of astrocytes mediates a considerable fraction of NMDAR-elicited neurotoxicity [25]. Like astrocytes, cultured neurons release d-serine in a regulated manner, involving ionotropic glutamate receptor stimulation and depolarization by KCl [25]. In contrast to that previously reported with cultured astrocytes [65], however, the neuronal d-serine was not released through exocytosis of synaptic vesicles under our experimental conditions [25]. It remains to be established whether neuron-derived d-serine affects normal NMDAR transmission and if neurons indeed release d-serine in more physiological preparations.

In light of the widespread expression of serine racemase in forebrain neurons, which lack significant levels of d-amino acid oxidase, one would predict that d-serine should be present in all neurons. A few studies detected the presence of d-serine in some neuronal populations in situ, including the pyramidal neurons of the cerebral cortex [25,38,39]. Neuronal d-serine, however, is scarcely seen in most studies. We speculate that this may be attributed to technical difficulties or to low sensitivity of the immunohistochemical methods to detect d-serine. Being a small amino acid, d-serine may be poorly fixed by the commonly used fixatives, or even released from cells during the perfusion of the brain. In this framework, it is conceivable that the antibodies against d-serine miss many neuronal populations that contain significant levels of d-serine.

Many questions remain to be solved regarding the relative roles of glia versus neurons in the synthesis and release of d-serine. In the original model of d-serine signaling, d-serine was thought to be exclusively released from astrocytes. The predominance of serine racemase expression in neurons led us to propose an alternative model of d-serine signaling, in which d-serine may be released from both neurons and astrocytes (Fig. 5). This model assumes that the neuronal serine racemase enzyme is active towards d-serine synthesis and, like astrocytes, neurons release d-serine in a regulated manner (Fig. 5).

Figure 5.

 Proposed roles of glia and neurons in d-serine signaling. The scheme depicts two modes of d-serine release. A glia to neuron d-serine flux would be achieved through activation of glial AMPA receptors by glutamate (reaction 1) [34]. This leads to the release of astrocytic d-serine, possibly from a vesicular pool [65], to activate neuronal NMDARs (reaction 2). Because serine racemase (SR) occurs predominantly in neurons [25,41], astrocytes may obtain d-serine by re-uptake from the extracellular medium. Alternatively, the higher ability of astrocytes to synthesize l-serine from glucose [69] might also allow the synthesis of d-serine by some astrocytes containing serine racemase; the relative importance of each pathway leading to astrocytic accumulation of d-serine is unknown. A neuron to glia d-serine flux would be achieved by the release of d-serine from neurons, presumably by membrane depolarization (reaction 3) [25]. Released d-serine will activate NMDARs or be taken up by astrocytes (reaction 4). It is not clear whether neuronal d-serine synthesis and release occur at presynaptic or postsynaptic sites. Because neurons are mostly devoid of the ability to synthesize l-serine from glucose [69], they should rely on the export of l-serine from astrocytes (reaction 5) [70].

The notion that neurons play a role in d-serine signaling does not exclude a role of glia in releasing d-serine, as d-serine is clearly enriched in protoplasmic astrocytes in the forebrain [33,34]. One possibility is that the higher level of d-serine in astrocytes reflects the glial uptake of d-serine synthesized and released by neurons. In this case, one would expect that d-serine uptake by astrocytes in vivo will be more efficient than in neurons or, alternatively, that the d-serine half-life would be longer in astrocytes. Nevertheless, experimental data demonstrating specific vesicular or nonvesicular release of d-serine from astrocytes in brain slices or in vivo are still lacking.

The levels of l-serine relative to serine racemase expression will also influence the distribution of d-serine. l-Serine can be synthesized from either glycine or glucose or be obtained by uptake from the extracellular medium [68,69]. It is known that, along with many growth factors, astrocytes release l-serine to neurons [70,71]. Astrocytes have a higher l-serine content and can synthesize it from the glycolytic intermediate 3-phosphoglycerate, an ability that neurons apparently lack [72]. The importance of the 3-phosphoglycerate pathway for the synthesis of d-serine arose from the observation that children exhibiting 3-phosphoglycerate dehydrogenase deficiency display severe neurodevelopmental problems associated with lower levels of d-serine in the cerebrospinal fluid [73]. Thus, the ability of astrocytes to synthesize l-serine by the 3-phosphoglycerate pathway might allow higher synthesis of d-serine, even if the expression of serine racemase in astrocytes is lower than in neurons. On the other hand, neuronal synthesis of d-serine will require conversion of glycine into l-serine by the serine hydroxymethyltransferase enzyme [74] or uptake of l-serine from the extracellular medium.

Serine racemase knockout mice will be a valuable tool to ascertain definitively the cellular origin of d-serine. In light of the new data indicating a neuronal source of d-serine, serine racemase knockout mice will be useful in defining the specificity of serine racemase and d-serine antibodies previously employed, hopefully providing a more definitive answer as to whether d-serine originates from neurons or astrocytes, or from both. Indeed, a preliminary study by Mori et al. demonstrated, using serine racemase knockout mice as controls, that serine racemase is present mainly in neurons [62].

d-Serine in disease

As well as being important for normal NMDAR transmission, NMDAR-dependent plasticity and developmental processes (Fig. 6), d-serine signaling dysregulation might also be involved in the NMDAR dysfunction that occurs in several pathologies, including neuro-psychiatric and neurodegenerative diseases (Fig. 6).

Figure 6.

 Multitude of d-serine functions. d-Serine has been implicated in several physiological NMDAR-dependent processes, including normal transmission, synaptic plasticity and cell migration in the developing cerebellum. d-Serine dysregulation may also play pathological roles in schizophrenia, ageing and acute and chronic neurodegeneration (see the text for references).

An important pathological aspect of d-serine signaling relates to NMDAR hypofunction thought to occur in schizophrenia [75]. NMDA antagonists, such as phencyclidine, induce schizophrenic-like symptoms in healthy volunteers, and precipitate thought disorder and delusions in schizophrenia patients [75,76]. In mice, d-serine antagonizes the stereotypical behavior and ataxia caused by NMDAR antagonists [77]. Mice expressing lower levels of the NMDAR1 (NR1) subunit display behavioral abnormalities, including increased motor activity and stereotypy, and deficits in social and sexual interactions, which are ameliorated by conventional antipsychotic treatment [78].

Based on the NMDA hypofunction hypothesis, several clinical trials were carried out to evaluate the efficacy of stimulation of NMDAR in schizophrenia. The administration of d-serine greatly ameliorated the positive, negative and cognitive symptoms of schizophrenia when associated with conventional neuroleptics [79–81]. Currently, five additional clinical trials are evaluating the effects of d-serine administration in schizophrenia in larger patient groups, which include both phase II and phase III studies.

In addition to being a promising pharmacological treatment for schizophrenia, a number of recent studies indicate that the level of endogenous d-serine may also be altered in the disease. Schizophrenic patients display a higher ratio of l-serine to d-serine in the blood and cerebrospinal fluid [82–84]. The possible involvement of d-serine in schizophrenia was also highlighted by genetic studies showing polymorphisms in the genes of serine racemase [85] and of the d-serine metabolic enzyme, d-amino acid oxidase [86]. Confirmation of the above studies in larger populations will be important to ascertain the role of endogenous d-serine in the pathophysiology of schizophrenia.

Is d-serine dysregulation linked to cognitive deficits? The long-term potentiation (LTP) of the synaptic activity in the hippocampus has been thought to play a role in memory formation [87]. The role of endogenous d-serine in LTP [7,14,15,42] raises the possibility that d-serine dysfunction might cause cognitive deficits. Although this possibility has not been directly investigated, aged rats display a sharp decrease in hippocampal d-serine and serine racemase expression [42]. This is associated with impaired LTP, which is reversed by the addition of exogenous d-serine in aged rats [7]. By contrast, in young rats, LTP is not enhanced by exogenous d-serine. Thus, it is possible that the LTP impairment observed in aged rats is caused by specific deficits in local d-serine synthesis.

The overproduction or excessive release of glutamate has been widely implicated in a large number of acute and chronic degenerative diseases. The harmful effects of excessive glutamate occur mainly through activation of the NMDARs and consequently by massive calcium influx into the cell [1]. NMDAR over-activation is the main culprit in the cell death that occurs following stroke and in neurodegenerative diseases [3].

Blockers of NMDARs are neuroprotective in animal models of stroke, but they were not well tolerated in clinical trials because of the side effects caused by NMDAR blockage, such as hallucinations [88,89]. Recently, low-affinity NMDAR inhibitors, like memantine, have been proposed as an alternative to high-affinity NMDAR blockers, and are indeed well tolerated by patients [88,89]. Similarly to low-affinity NMDAR antagonists, serine racemase inhibitors offer a more gentle approach to decrease NMDAR activation, and are likely to be better tolerated than high-affinity antagonists. In this framework, selective serine racemase inhibitors provide a new strategy to prevent stroke damage and cell death in neurodegenerative diseases.

Excessive production or release of d-serine may also be involved in chronic neurodegeneration. The levels of d-serine and its biosynthetic enzyme, serine racemase, are greatly increased in the spinal cord of patients with familiar and sporadic forms of amyotrophic lateral sclerosis (ALS) [27]. Although the motoneuronal cell death in ALS is widely attributed to excessive AMPA receptor stimulation [90], a recent study indicates that endogenous d-serine mediates motoneuron cell death by excessive stimulation of NMDAR in the spinal cord of ALS mice [27]. In ALS transgenic mice harboring the G93A mutation in superoxide dismutase 1, activated microglia seem to be the main source of spinal d-serine, constituting a potential therapeutic target for ALS [27]. Activation of microglia by inflammatory stimuli induces overexpression of serine racemase, an effect mediated by the c-Jun terminal kinase [27,91]. Additionally, overexpression of the G39A mutant, superoxide dismutase 1, promotes the upregulation of serine racemase in a c-Jun terminal kinase-independent manner [27]. Removal of endogenous d-serine from spinal cord cultures of ALS transgenic mice protects the motoneurons against NMDAR-mediated cell death, linking d-serine to motoneuron degeneration [27]. The overproduction of d-serine by glia in ALS fits the notion that glial activation/dysfunction plays a role in the disease [92]. In this context, inhibitors of serine racemase may provide a new neuroprotective strategy against ALS.


d-Serine is now widely recognized as an important player in NMDAR transmission and in pathologies linked to NMDAR dysfunction. Whether or not d-serine satisfies all the criteria for a transmitter, its role in regulating NMDARs indicates an important physiological role. There is still much to be learned regarding the regulation of d-serine signaling, including its biosynthesis regulation and mechanisms of release. While the experimental data so far favor a role of d-serine as a transmitter, many of the effects previously attributed to astrocytic d-serine release may also be caused by neuronal d-serine. Furthermore, it is unclear whether d-serine is physiologically released in a tonic manner or in a fast and activity-dependent manner. Further studies will be required to define the release pathways for d-serine from both neurons and astrocytes, and to clarify their relative contributions in d-serine-mediated NMDAR signaling.


HW is supported by a grant from Israel Science Foundation.