Keywords

The neurotransmitter glutamate once released in the synaptic cleft binds and acts on various transporters, ionotropic and metabotropic membrane receptors located onto neuronal elements (Traynelis et al. 2010), but also onto glial cells including astrocytes and oligodendrocytes (Parpura and Verkhratsky 2013). The ionotropic glutamate-gated receptor (iGluR) family encodes 18 gene products or subunits subdivided into three main subtypes characterized by their preferential agonists: AMPA receptors (GluARs), kainate receptors (GluKRs), and NMDARs (GluNs) (Traynelis et al. 2010; Paoletti et al. 2013). AMPARs and KARs are cation-permeable ion channels which upon binding of glutamate mediate fast excitatory synaptic transmission, while NMDARs are calcium permeable and drive the transduction of specific temporal patterns of synaptic activity into long-term structural and functional changes. N-Methyl d-aspartic receptors (NMDARs) are ionotropic glutamate-gated receptors that are central to many physiological processes including learning and memory and are involved in neurotoxicity and psychiatric disorders (Traynelis et al. 2010; Paoletti et al. 2013).

1 The NMDA Receptors

1.1 Structure and Subunit Diversity of NMDARs in the CNS

NMDA receptors are heterotetramers associating two obligatory GluN1 subunits with two identical or different GluN2 or a mixture of GluN2 and GluN3 subunits (Traynelis et al. 2010; Paoletti et al. 2013) (Fig. 3.1). The GluN1 subunit exists as eight distinct splice variants of a single encoding gene (GRIN1). Four distinct GluN2 subunits (GluN2A-D) and two GluN3 subunits (GluN3A, B) have been identified which are encoded by four (GRIN2A-D) and two (GRIN3A, B) different genes, respectively (Traynelis et al. 2010; Paoletti et al. 2013). The large repertoire of diheteromeric and triheteromeric NMDARs display distinct trafficking, biophysical, and pharmacological properties and are differentially populating synapses depending on neuronal subpopulations and on development (Traynelis et al. 2010; Paoletti et al. 2013).

Fig. 3.1
figure 1

The NMDA receptors. (a) Linear organization of monomeric subunits. Seven subunits (GluN) have been identified: GluN1, GluN2A–GluN2D, and GluN3A and GluN3B. M1–M4 refer to membrane segments. (b) Modular organization of a GluN subunit. The N-terminal domain (NTD); the agonist-binding domain (ABD) that binds to the ligand (L), i.e., glycine or d-serine in GluN1 and GluN3 and glutamate in GluN2; the transmembrane domain (TMD) containing the ion channel; and an intracellular C-terminal domain (CTD). With permission from Mothet et al. (2015) J Neurochem. (c) Possible subunits compositions for diheteromeric and triheteromeric receptors

Full size image

Similar to all iGluR subunits, every GluN subunit has four domains (Fig. 3.1). The extracellular region comprises the amino-terminal domain (NTD) involved in subunit association and the agonist-binding domain (ABD) formed by two segments S1 and S2 which bind the co-agonists glycine and d-serine on GluN1 and GluN3 subunits or glutamate on GluN2 subunits. The transmembrane domain (TMD) is formed by three helices (M1, M3, M4) and a hairpin (M2) forming the channel pore and conferring the ion selectivity. The cytoplasmic carboxy-terminal domain (CTD) is involved in protein-protein interactions which determine trafficking (membrane targeting, lateral diffusion, stabilization, recycling), posttranslational regulation, and coupling to specific intracellular signaling cascades. All domains except the CTD are the targets of multiple endogenous or synthetic allosteric modulators (Traynelis et al. 2010; Paoletti et al. 2013).

1.2 NMDARs and the Glycine Modulatory Binding Site

NMDARs possess unique biophysical and pharmacological properties among the iGluR family (Traynelis et al. 2010; Paoletti et al. 2013). In contrast to GluKRs and GluARs, all NMDARs show much slower gating and deactivating kinetics, thus displaying a large single channel conductance (~40 pS) depending on the subunit composition. The NMDARs display high Ca2+ permeability (with subunit-specific variations) and behave as unique Hebbian-like coincidence detectors owing to their voltage-dependent Mg2+ block and the presence of many modulatory sites on the different domains of each subunit. But a major property of NMDAR activation of interest for us here is the need of a concomitant binding of glutamate on GluN2 subunits and of a co-agonist d-serine or glycine on GluN1 (Martineau et al. 2006; Traynelis et al. 2010; Paoletti et al. 2013).

The absolute necessity of a co-agonist together with glutamate for the activation of the NMDARs stemmed from the original observations that NMDA-evoked inward currents recorded in native NMDARs (Johnson and Ascher 1987) and from NMDARs expressed in oocytes (Kleckner and Dingledine 1988) require the addition of glycine at high nanomolar range. In their seminal work, Kleckner and Dingledine already showed that several glycine analogs such as d-serine or d-alanine could substitute in activating NMDARs (Kleckner and Dingledine 1988). This observation is consistent with early binding experiments and electrophysiological recordings showing that glycine and d-serine target NMDARs with relatively similar nanomolar affinity (Monaghan et al. 1988; Matsui et al. 1995). Cocrystal structures of the GluN1 S1-S2 ligand-binding core with glycine or d-serine reveal that agonist binding is critically dependent upon a series of hydrogen bonds to side-chain and main-chain atoms, as well as to water molecules in the binding pocket. Because it makes three additional hydrogen bonds and displaces a water molecule, d-serine binds more tightly to the receptor in comparison to glycine (Furukawa and Gouaux 2003) and would therefore virtually make it more efficient to activate the NMDAR. Nevertheless, the affinity of GluN1 for d-serine or glycine is very similar (EC50 ~ 0.1–1 μM, (Priestley and Kemp 1994; Chen et al. 2008). The affinity of glycine or d-serine for GluN1 is in fact dictated by the identity of GluN2 subunit. Indeed, the binding sites of the co-agonist and glutamate are allosterically coupled. Accordingly, heterodimeric GluN2B-containing NMDARs display 10× higher affinity for glycine than those containing the GluN2A subunit (EC50 ~ 0.1 μM versus 1 μM). Conversely, the affinity of d-serine for GluN1 seems to be higher when the NMDARs are containing the GluN2A subunits (Priestley and Kemp 1994; Matsui et al. 1995). While the binding of d-serine at NMDARs was described more than 25 years ago, only glycine was thought to be the right co-agonist since d-serine was not expected to exist in higher living organisms. The discovery that d-amino acids including d-serine are present in organisms throughout evolution including in humans has forced neuroscientists to reconsider the mode of activation of NMDARs. In different tissues including the brain, d-amino acids have emerged as important signaling molecules (Ohide et al. 2011). The field has grown very fast since the pioneer observations made in early 1990s and results from intensive research and from many groups around the world now support that d-serine is the preferred endogenous ligand for the glycine site of NMDARs and thus that this unexpected d-amino acid is central to many brain functions.

2 Astrocytes and Neurons Use Different Pathways to Release d-Serine

2.1 d-Serine: Glial and Neuronal

In mammals, d-serine is exclusively synthesized from l-serine by serine racemase (SR) (Wolosker and Mori 2012), an enzyme widely distributed in the CNS. SR is a highly regulated enzyme reviewed in (Campanini et al. 2013; Martineau et al. 2014) that binds to several receptor-interacting proteins that modulate the production of d-serine. In addition, d-serine degradation is controlled by the flavoenzyme DAAO, a peroxisomal enzyme that eliminates the amino acid retrieved from the extracellular space (Martineau et al. 2006). The expression and activity of DAAO is widespread in the human and rodent CNS being expressed in both astrocytes and neurons (Verrall et al. 2007; Sasabe et al. 2014). Despite many progresses, our knowledge on the biochemical cascades modulating DAAO activity is still very limited. DAAO interacts with pLG72, a primate-specific protein present in mitochondria (Sacchi et al. 2008; Otte et al. 2014). The interaction with DAAO would involve interaction of the cytosolic fraction of DAAO with pLG72 potentially exposed at the surface of mitochondria as suggested by Sacchi and colleagues (Sacchi et al. 2011). We do not know if an ortholog of pLG72 may exist in non-primate mammals. Proteomics analyses have revealed that DAAO interacts also with Bassoon and Piccolo, two proteins of the presynaptic active zone (Popiolek et al. 2011) supporting the neuronal localization for DAAO.

In which cell type d-serine is formed? Although a glial localization has been largely attributed to SR and d-serine (Schell et al. 1997; Wolosker et al. 1999; Panatier et al. 2006; Williams et al. 2006), allowing the amino acid to be considered as a gliotransmitter, several studies indicate that SR may be instead preferentially expressed in excitatory and inhibitory neurons in rodent and human brains rather than in glia (Miya et al. 2008; Ehmsen et al. 2013; Balu et al. 2014). Accordingly, neuronal-specific conditional deletion of SR induces a much larger decrease in cerebral and hippocampal d-serine contents compared to deletion limited to the astrocyte population (Benneyworth et al. 2012). However, variations in the cellular expression of SR and d-serine between species could not be totally ruled out. Accordingly, Suzuki et al. (2015), using SR-KO as negative controls, recently reported that astrocytes abundantly express SR in the human subiculum. The notion that d-serine could be formed in astroglia is further reinforced by the fact that SR interacts with DISC-1 in astrocytes (Ma et al. 2013). Furthermore, using multiplex single-cell RT-PCR, we observe that both neurons and astrocytes express SR at least at the levels of transcripts in the prefrontal cortex of adult rat (Dallerac, Turpin & Mothet, Personal communication). Thus, the cellular distribution of SR remains an open issue particularly since a recent transcriptome analysis of isolated purified brain cell groups shows that the enzyme is more expressed in glial cell lineage and notably astrocytes than in neurons in the mature cortex of mice (Zhang et al. 2014). As neurons are not capable of synthesizing much l-serine, neuronal SR activity would depend on the supply of the d-serine precursor which is primarily formed in astrocytes from glucose by 3-phosphoglycerate dehydrogenase (Yang et al. 2010; Ehmsen et al. 2013).

2.2 Cell-Type-Specific Mechanisms of Release: Exocytosis Versus Hetero-exchange

As shown so far, d-serine can be formed in both neurons and astrocytes (Martineau et al. 2014). Different elements of the literature support the notion that neurons and astroglia use different molecular machinery and signaling pathways to release d-serine into the extracellular space (Fig. 3.2). Early studies have shown that d-serine release is triggered by agonists of the ionotropic and metabotropic glutamate receptors (iGluR and mGluR, respectively; (Schell et al. 1995; Mothet et al. 2005; Martineau et al. 2008; Ishiwata et al. 2013). We have accumulated evidence that the release of d-serine by astrocytes is mainly dependent on an increase in cytosolic Ca2+. Indeed, disrupting Ca2+ signaling inside astrocytes reduces d-serine release from astrocytes in culture and in hippocampal slices (Mothet et al. 2005; Henneberger et al. 2010; Shigetomi et al. 2013) but also from astrocytes in vivo (Takata et al. 2011). Calcium certainly represents the most critical signaling hub for d-serine release from astrocytes, even though we are still largely ignoring the spatial and temporal signature of these Ca2+ signals: activation of astrocytic transforming growth factor (TGF)-β (TGFR; (Diniz et al. 2012)), bradykinin-type2 (B2R; (Martineau et al. 2008)), adenosine-type 2 (A2R; (Scianni et al. 2013)), ephrinB3 (Zhuang et al. 2010), and muscarinic acetylcholine (Takata et al. 2011; Lopez-Hidalgo et al. 2012) receptors. The activation of these G protein-coupled receptors (GPCRs) is associated with the recruitment of Ca2+ from the intracellular stores mainly via inositol-1,4,5-trisphosphate receptors (IP3R) located on the endoplasmic reticulum (ER) (Zorec et al. 2012). But Ca2+ driving d-serine release could also originate from the extracellular space through channel-mediated transmembrane Ca2+ fluxes in astrocytes. For example, astrocytic transient receptor potential A1 (TRPA1) channels contribute to basal Ca2+ signals which are required for d-serine release (Shigetomi et al. 2013). These studies clearly established that astrocytes express a plethora of functional receptors which activation is coupled to the release of d-serine. Then, what could be the mechanisms downstream Ca2+ triggering d-serine release? Although not exclusive of another release mechanism, we have found that d-serine release from astrocytes is dependent on soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins, indicating that glial Ca2+-regulated exocytosis is a main release mechanism used by astroglia to release d-serine. Indeed, astrocytes as most eukaryotic cells expressed synaptobrevins, also referred to as vesicle-associated proteins-VAMPs, or R (for arginine)-SNAREs) and the plasma membrane (syntaxins and synaptosome-associated protein of 23 kDa (SNAP23); Q (for glutamine)-SNAREs) (Jahn and Fasshauer 2012). The vesicle fusion is triggered by an increase in cytosolic Ca2+, which presumably binds to vesicular synaptotagmins (Montana et al. 2006). Astroglial vesicles also possess proteins necessary for vesicular filling, such as the vacuolar type H+-ATPase (V-ATPase) (Martineau et al. 2013) which provides the proton gradient necessary for intravesicular loading of gliotransmitters via appropriate transporter(s). Cleavage of Sb2 and cellubrevin by tetanus neurotoxin causes a strong inhibition of Ca2+-dependent d-serine release (Mothet et al. 2005; Henneberger et al. 2010). Additionally, the blockade of d-serine vesicular uptake using a V-ATPase blocker inhibits d-serine release from astrocytes (Mothet et al. 2005). Indeed, based on electron microscopy (EM), d-serine accumulates in clear vesicles with a diameter of 36 nm in the perisynaptic processes of hippocampal and cortical astrocytes (Bergersen et al. 2012; Martineau et al. 2013). In the adult hippocampus, clear vesicles are organized in small groups of 2–15 vesicles preferentially located within 100 nm from the astrocytic plasma membrane (Bergersen et al. 2012) and observed at sites adjacent to neuronal elements bearing NMDARs (Bezzi et al. 2004). Thus, astrocytes possess small vesicles resembling those found at synaptic terminals, albeit at lower density. It should be noted, however, that the vesicular size in astrocytes appears to be more diverse than that described for neurons. In addition to Ca2+- and SNARE-dependent exocytotic release of d-serine from astrocytes, this gliotransmitter can be released via alternative non-exocytotic conduits at the plasma membrane, including volume-regulated anion channels (VRAC; (Rosenberg et al. 2010)), alanine–serine–cysteine transporter (ASCT; (Ribeiro et al. 2002), but see (Maucler et al. 2013)) and likely through connexin 43 hemichannels (Stehberg et al. 2012). Although the vesicular transporter for d-serine has not been identified, the transport of d-serine inside astroglial vesicles was recently characterized (Martineau et al. 2013). While glutamate transport is observed in both synaptic and astroglial vesicles, the transport of d-serine is specific to astroglial vesicles (Martineau et al. 2013). Its apparent affinity is ~7 mM, consistent with the affinity of vesicular inhibitory amino acid transporter for γ-aminobutyric acid, another neutral amino acid (Chaudhry 2008; Martineau et al. 2013). Similar to glutamate, extravesicular chloride concentration modulates d-serine transport into astroglial vesicles, reaching the maximum activity at 4 mM. Because d-serine transport induces vesicular acidification and critically relies on chloride, the vesicular d-serine transporter is proposed to be a d-serine/chloride co-transporter (Martineau et al. 2013) (Fig. 3.2). A spatial association of SR activity and d-serine vesicular transport was observed, resulting in a functional coupling between d-serine synthesis and uptake (Martineau et al. 2013). Finally, d-serine and glutamate vesicular loading exert a mutual stimulation which indicates a functional crosstalk between the two transporters (Martineau et al. 2013). This synergy at vesicular level can only be explained if both transporters reside on the same vesicle, indicating the co-storage and thus the co-release of both gliotransmitters. However, the immunogold colabeling of d-serine and glutamate in the adult hippocampus did not reveal a population of vesicles containing both gliotransmitters (Bergersen et al. 2012); this seemingly disparate findings could be a result of the limited sensitivity of the immuno-EM. Nonetheless, the mechanism underlying the vesicular synergy between d-serine and glutamate uptake requires additional investigation. The possible co-storage of glutamate and d-serine, at least in a subpopulation of astroglial vesicles, points to the possibility of interdependence of their dynamics and modulatory functions at synaptic and extrasynaptic sites.

Fig. 3.2
figure 2

Modes of release of d-serine. (a) Preembedding peroxidase immunohistochemistry showing that d-serine immunoreactivity is mainly distributed in astrocytic processes (arrows) at the level of asymmetric synapses in the adult rat cerebral cortex. Representative low-magnification (top) and high-magnification (bottom) electron micrographs showing that no peroxidase staining is detected with a preabsorbed anti-d-serine antibody. Ast astrocyte, D dendrite, S dendritic spine, B axonal bouton. Scale bars, 500 nm. Modified with permission from The Society of Neuroscience (Martineau et al. 2013). (b) Immunogold labeling (small gold particles) of d-serine in small synaptic-like vesicles (indicated by red arrowheads) in astrocytic processes (Ast) positive for EAAT2 (large gold particles) contacting asymmetric synapses (stars) between nerve terminals (Term) and dendritic spines (Sp). Insets: higher magnification showing the similarities between SLMVs (red dotted arrows) and SVs (black dotted arrow). Note that the astrocytic SLMVs are often localized in small clusters close to the plasma membrane. Scale bars: 100 and 50 nm in insets. Right immunogold quantification of d-serine gold particles in astrocytic processes. The bar charts show the mean number of d-serine gold particles/μm2 ± SD in SLMVs and the cytoplasmic matrix of astrocytic processes (Acyt). Reprinted with permission from Oxford University Press (Bergersen et al 2012). (c) Graphical abstract showing that in astrocytes d-serine is mainly released by exocytosis of some vesicles containing also glutamate, while in neurons d-serine is released from the cytosol through asc-1. For explanation, see text

Full size image

Release of d-serine as a neurotransmitter operates through a different mechanism than those described before. Neuronal d-serine is mainly released in response to depolarization, induced by veratridine, a voltage-dependent Na+ channel activator, or by increased extracellular potassium concentration, both in vitro and in vivo (Hashimoto et al. 2000; Rosenberg et al. 2010) (Fig. 3.2). Intriguingly, neuronal silencing by tetrodotoxin increased d-serine extracellular concentration in vivo (Maucler et al. 2013) although some studies show no effect (Hashimoto et al. 1995). Chelation of intracellular and extracellular Ca2+ does not affect depolarization-elicited d-serine release. In addition, inhibition of V-ATPase by bafilomycin A1 or cleavage of Sb2 by tetanus neurotoxin failed to inhibit d-serine release from neurons, excluding the vesicular release mechanism (Rosenberg et al. 2010) which is consistent with d-serine absence in the lumen of synaptic vesicles (Martineau et al. 2013), d-serine is released by neuronal presynaptic elements from a cytosolic pool through the Na+-independent antiporter alanine– serine–cysteine transporter 1 (asc-1; (Rosenberg et al. 2010; Rosenberg et al. 2013); Fig. 3.2). Asc-1 is restricted to neurons and catalyzes neutral amino acid hetero-exchange (Fukasawa et al. 2000; Helboe et al. 2003; Matsuo et al. 2004), between d-serine and another neutral amino acid. Neuronal d-serine release could be triggered by neutral amino acids, such as l-serine, d-alanine, or d-isoleucine, through this hetero-exchange (Rosenberg et al. 2010, 2013). Two independent studies published in 2013 (Ishiwata et al. 2013; Maucler et al. 2013) have explored in vivo the dynamics of d-serine release in rat cortex. Using a second-by-second time scale microelectrode biosensors, Maucler et al. (2013) found that the application of the asc-1 transporter inhibitor, S-methyl-l-cysteine, depressed the release of d-serine evoked by l-serine, while l-asparagine, a substrate of ASCT2, did not induce d-serine release in vivo. Conversely, intra-medial frontal cortex infusion of S-methyl-l-cysteine, caused a concentration-dependent increase in the microdialysate contents of d-serine (Ishiwata et al. 2013). The reasons of these opposite results are not known but could reflect differences in timescale of both analytical methods used in the two studies but clearly show that asc-1 could work in both directions. Which direction (uptake or release) in vivo prevails is not known so far.

3 Functions of d-Serine: From Development to Cognition

What could be the function of d-serine in the CNS? Early studies in the 1980–1990s evidenced that exogenous d-serine can potentiate NMDARs activity (Danysz et al. 1990). Two pioneer observations in the 1990s put d-serine in the spotlight by revealing that endogenous d-serine may be linked to NMDARs activation. First, its distribution in the rodent brain mirrors the expression of NMDARs, as one might expect for an NMDAR co-agonist, during development and in the adult animal (Hashimoto et al. 1993; Schell et al. 1995, 1997).

3.1 Role of d-Serine in the Adult Brain

Although d-serine could serve as a co-agonist of NMDAR, the experimental evidence for it awaited for the development of specific pharmacological and genetic tools allowing manipulating the levels of the endogenous amino acid.

A first approach we introduced consisted in the use of purified DAAO on top cultured neurons or brain slices to deplete endogenous d-serine while recording synaptic events (Mothet et al. 2000). We then discovered that depleting d-serine with DAAO consistently reduces synaptic events and NMDAR-mediated currents just demonstrating that d-serine rather than glycine as initially proposed would be the endogenous ligand for synaptic NMDARS in the hippocampus and juvenile cerebellum (Mothet et al. 2000). Therefore, it was predicted that the functions of d-serine and glycine could be dependent on the location of specific synapses within different brain area. But, subsequently acute depletion of d-serine with recombinant Rhodotorula gracilis DAAO (RgDAAO) or serine deaminase (Dsda), two enzymatic d-serine scavengers, consistently affects NMDAR activity on glutamatergic neurons in widespread area of the mature brain including the Schaeffer collateral (SC)-CA1 synapse in the hippocampus (Papouin et al. 2012; Le Bail et al. 2015), medial prefrontal cortex (Fossat et al. 2012), hypothalamic supraoptic nucleus (Panatier et al. 2006), nucleus accumbens (Curcio et al. 2013), the medial nucleus of the trapezoid body (MNTB) (Reyes-Haro et al. 2010), or even the retina (Stevens et al. 2003). These observations lead us to jump on the conclusion that d-serine and not glycine is the right endogenous co-agonist for synaptic NMDARs at most central synapses. This conclusion was reinforced by the observations that a concomitant removal of endogenous glycine with Bacillus subtilis GO (BsGO) has only weak or no effects at the SC-CA1 glutamatergic synapse in the hippocampus (Papouin et al. 2012; Le Bail et al. 2015) or in the hypothalamus (Panatier et al. 2006). Conversely, in some case like the nucleus tractus solitarii (NTS), a structure located in the brainstem, glycine acts as the primary co-agonist for synaptic receptors (Panatier et al. 2006). In the structures where d-serine action predominates, it is thought that extrasynaptic glia-derived glycine should be unable to reach synaptic sites due to efficient capture by GlyT1 (Fossat et al. 2012; Papouin et al. 2012; Le Bail et al. 2015). Nevertheless, this dichotomic view is challenged by the fact that a substantial fraction of the synaptic NMDAR responses remained systematically resistant to the enzymatic degradation of d-serine. In addition, neuronal-derived glycine released through Asc-1 physiologically modulates NMDAR activation (Rosenberg et al. 2013) at the SC-CA1 hippocampal synapses. Introduction of animal models for d-serine represented another important opening to define the function of d-serine. A first model was developed early by Konno and collaborators which is a mutant mouse strain (ddY/DAO−) lacking DAO activity (Konno and Yasumura 1983). The ddY/DAO− mice have a missense mutation (G181R) that causes a complete loss of enzyme activity (Sasaki et al. 1992). As expected, these mutant mice exhibit larger NMDA/AMPA ratio due to a higher occupancy of the glycine site of NMDARs in diverse areas like the hippocampus (Maekawa et al. 2005) and the retina (Romero et al. 2014) and also show enhanced hippocampal LTP (Maekawa et al. 2005). More recently, specific DAAO inhibitors have been developed by different companies. Duplantier and colleagues (2011) show that inhibition of DAAO increases NMDA receptor-mediated synaptic currents in primary neuronal cultures from rat hippocampus and resulted in a significant increase in evoked hippocampal theta rhythm, an in vivo electrophysiological model of hippocampal activity (Strick et al. 2011). Likewise, inhibition of DAAO was reported to increase NMDAR activity and LTP formation in the mature hippocampus (Hopkins et al. 2013a, b; Le Bail et al. 2015) and to normalize the activity of impaired NMDARs in the supraoptic nucleus of lactating rats (Hopkins et al. 2013a) by increasing d-serine levels. These results predict that DAAO inhibition would have a positive effect on cognition, most probably through its ability to augment NMDAR-mediated currents. In 2009, Coyle and colleagues introduced the first genetically invalidation of SR in mice (Basu et al. 2009). Despite some subtle behavioral and structural deficits (see Sects. 3.3.2 and 3.3.3), electrophysiological phenotyping of NMDARs activity and synaptic plasticity intriguingly revealed that a 90 % decrease in forebrain d-serine content (Basu et al. 2009), had only a moderate impact on NMDARs at SC-CA1 synapses. Indeed, the amplitude of NMDA-mediated synaptic currents was normal; only the decay kinetics of evoked NMDAR-mediated currents were slower in hippocampal slices in SR-KO mice (Basu et al. 2009) due certainly to a compensatory mechanism by glycine. The most striking effect in SR-KO is the absence of LTP induced by a pairing protocol (Basu et al. 2009; Rosenberg et al. 2013; Le Bail et al. 2015). It is worth noticing that the selective depletion of both d-serine and glycine reduce NMDAR synaptic responses in the lateral nucleus of the amygdala (LNA) (Li et al. 2013). Therefore, the contribution of the two co-agonists in activating NMDARs shows general overlap in the brain, but regional differences in the preference for one co-agonist over the other certainly occur.

Could the preference for a co-agonist be related to the molecular composition of synaptic NMDARs? We have recently discovered that in the mature hippocampus, d-serine is the main co-agonist at SC-CA1 hippocampal synapses whereas glycine is preferentially concerned at connections within the dentate gyrus (Le Bail et al. 2015). The role of d-serine prevails in brain regions where the GluN2A-containing subtype of NMDARs predominates such as hippocampal SC-CA1 and prefrontal cortex synapses (Fossat et al. 2012; Papouin et al. 2012; Le Bail et al. 2015), while glycine acts as the preferred co-agonist in regions where the contribution of the GluN2B subunit is favored, as seen in the dentate gyrus (Le Bail et al. 2015). This segregation would explain why SR-deficient mice display moderate synaptic dysfunctions at the medial perforant path-dentate gyrus synapses (Balu et al. 2013).

But in addition to be synapse specific, the function of d-serine is developmentally regulated (Le Bail et al. 2015) (Fig. 3.3). Indeed, we have found that glycine is the preferred co-agonist at immature SC-CA1 synapses and demonstrated that a switch from glycine to d-serine occurs at these synapses during postnatal development (Le Bail et al. 2015) that parallels a progressive replacement of GluN2B by GluN2A subunits in NMDAR assemblies (Rodenas-Ruano et al. 2012). Nevertheless, a strict association between the identity of the co-agonist and the NMDAR molecular composition is not a general rule since d-serine acts as the main co-agonist in the supraoptic nucleus (Panatier et al. 2006) where GluN2B-containing receptors predominate but also in the nucleus accumbens and the prefrontal cortex (Fossat et al. 2012; Curcio et al. 2013), where both GluN2 subunits are equally expressed. Additionally, one has to keep in mind that a substantial fraction of synaptic NMDARs are triheteromers containing both GluN2A and GluN2B subunits (reviewed in (Traynelis et al. 2010; Paoletti et al. 2013)) and are thus expected to be regulated by both d-serine and glycine.

Fig. 3.3
figure 3

Functions of d-serine. (a) Immature versus mature synapses. At immature synapse where only GluN1/2B receptors are present, glycine serves as the co-agonist. During postnatal development, the replacement of GluN2B- by GluN2A-containing NMDARs at synapses parallels a change in the identity of the co-agonist from glycine to d-serine. Conversely, glycine is the main co-agonist for extrasynaptic NMDA receptors. (b) d-serine and metaplasticity. d-serine levels govern the activity of NMDA receptor-dependent synaptic plasticity. Decreasing the levels of d-serine shifts the relationship to higher activity values, whereas increasing them has the opposite effect. Reprinted with permission from Cell Press (Panatier et al. 2006). (c) d-serine and cognition. d-serine levels drive cognitive and behavioral performances. Decreased levels as observed in normal aging and below a pathological threshold as observed in SR–KO or GrinD481N mice induce deficits. Conversely, increased levels of d-serine as observed under DAAO inactivation increases performances. Current therapy interventions aimed to normalize d-serine levels

Full size image

Besides the impact of the co-agonist regional specificity, a segregated role for d-serine and glycine at the synaptic cleft has also been reported. Indeed, NMDARs are not only present synaptic sites but are also located at extra- and peri-synaptic sites (Petralia 2012) where they may have different roles in cell death or survival programs (Hardingham and Bading 2010) and may be involved in different forms of synaptic plasticity (Papouin et al 2012; Mothet et al 2015). Enzymatic depletion of d-serine with RgDAAO consistently reduces NMDAR synaptic events at mature SC-CA1 connections (Papouin et al. 2012; Le Bail et al. 2015), while the pharmacologically isolated extrasynaptic NMDAR activation or the NMDAR-driven tonic glutamate current remain mostly unaffected by changes in d-serine levels (Papouin et al. 2012). Opposite effects were observed when endogenous glycine was depleted using Bacillus subtilis GO (BsGO), indicating that at those hippocampal synapses, d-serine would gate preferentially the synaptic GluN2A-NMDARs while glycine would target extrasynaptic receptors (Papouin et al. 2012). The segregated role of the two co-agonists at gating synaptic and extrasynaptic NMDARs at SC-CA1 synapses delineates their respective role in synaptic plasticity, with d-serine specifically required for the expression of LTP and both d-serine and glycine acting for the induction of long-term depression (LTD) (Papouin et al. 2012). Our observation for a role of d-serine in the formation of hippocampal LTD in the mature brain expands earlier observations made by Xu and colleagues (2008) that the d-amino acid regulates LTD in the hippocampal CA1 region in a “bell-shaped” concentration-dependent manner (Zhang et al. 2008). The retina is offering another model to appreciate the relative functions of d-serine and glycine at synaptic and extrasynaptic sites. The group of Miller have nourished evidence that AMPA receptor-dependent and light-evoked NMDA receptor-mediated currents recorded at retinal ganglion cells are shaped by the release of d-serine from Muller cells (Stevens et al. 2003, 2010; Gustafson et al. 2007), while glycine would rather activate extrasynaptic NMDARs (Sullivan and Miller 2012). However, this appealing dichotomic synaptic versus extrasynaptic segregation of the functions of d-serine and glycine has limitations. Several studies demonstrate that GluN2B-NMDARs preferring glycine are also required to induce LTP (reviewed in (Paoletti et al. 2013)). Furthermore, the repetitive observation that LTP at SC-CA1 synapses is only partly affected in SR-KO mice displaying trace levels of d-serine (Basu et al. 2009; Rosenberg et al. 2013) supports the idea that glycine could compensate for any d-serine synaptic dysfunction. Indeed, a couple of very recent studies have shown that the two endogenous co-agonists are necessary for LTP since both treatments with RgDAAO and BsGO result in significantly reduced LTP expression in CA1 hippocampal area and dentate gyrus, as well as in the LNA (Li et al. 2013; Le Bail et al. 2015). Interestingly, a robust LTP may be induced in the LNA of SR-KO mice by increasing the strength of presynaptic activation. This LTP is then consistently reduced under glycine depletion by BsGO (Li et al. 2013). These results indicate that independent of their synaptic location in the brain, both d-serine and glycine may be recruited for activating NMDARs during enhanced synaptic activity. Accordingly, treatments with RgDAAO but not BsGO reduces the NMDAR component of spontaneous excitatory postsynaptic currents in the LNA while synaptic responses evoked by enhanced afferent stimulation are impacted by BsGO (Li et al. 2013). Ambient d-serine could therefore drive tonic activation of NMDARs under low-synaptic activity in this structure, while the involvement of glycine should rely on increased afferent activity. In the hippocampus, although d-serine and glycine preferentially drive low-frequency-evoked NMDAR synaptic responses at mature SC-CA1 and dentate gyrus synapses, respectively, they both contribute to LTP expression driven by high-frequency stimulation at both hippocampal synapses (Le Bail et al. 2015) suggesting that both co-agonists may be necessary for memory formation. Therefore, although the identity of the NMDAR co-agonist may be synapse specific under low basal conditions, it appears that the activation by the two co-agonists is concerned when a sustained recruitment of synaptic activity is promoted.

NMDARs are also present at presynaptic sites where they regulate the release of glutamate by nerve terminals (Sjostrom et al. 2003; Corlew et al. 2008; Larsen et al. 2014). The mode of activation by either glycine or d-serine remained to be defined. Jones and colleagues, by using RgDAAO and BsGO to deplete d-serine and glycine, respectively, have recently reported that in the mature entorhinal cortex, these presynaptic NMDA receptors are tonically activated by d-serine but not by glycine, suggesting that d-serine is the endogenous co-agonist of presynaptic NMDARs (Lench et al. 2014).

Finally, d-serine and glycine synaptic functions closely rely on a fine coupling between astrocytes and neurons. Indeed, the profile of the tripartite synapse could also impact the identity of the predominant co-agonist acting on NMDAR. As mentioned earlier, a progressive developmental switch from glycine to d-serine occurs at SC-CA1 synapses that could reflect to some extent postnatal changes in the intimate neuron–glial coupling (Le Bail et al. 2015). At immature synapses, pharmacological blockade of astrocytic GlyT1 does not increase NMDAR activation as it does at mature connections, confirming that ambient glycine is able to saturate NMDAR binding sites early after birth. These elevated levels of glycine at birth presumably rely on a weak expression of GlyT1 (Zafra et al. 1995) related to the lack of a prominent astrocyte network that only progressively develops during the first postnatal weeks (Yang et al. 2013). The fact that inactivation of astrocyte activity with the gliotoxin fluoroacetate (FAC) does not significantly alter NMDAR activation at immature synapses further supports the view that this functional aspect of neuron–glial coupling is weakly effective early after birth (Le Bail et al. 2015). On the other hand, the postnatal formation of the tripartite synapse does not only result in a higher clearance of glycine by astrocytic GlyT1 but also in a better supply of d-serine either from astrocyte or through activation of the serine shuttle (Wolosker 2011). Further lines of experimental evidence show that d-serine derived from glia regulates NMDARs. Rusakov and colleagues reported that in the mature hippocampus, astrocyte processes contacting SC-CA1 synapses retain the ability to control Hebbian LTP within or near their individual territories involving Ca2+-and SNARE protein-dependent d-serine release (Henneberger et al. 2010) confirming earlier evidence for a vesicular release of d-serine from astrocytes (Mothet et al. 2005; Martineau et al. 2008). Because astrocytes express a plethora of receptors, they could link different neurotransmitter systems through the release of d-serine. Accordingly, potentiating effect of nicotine on hippocampal synaptic transmission and long-term synaptic plasticity recorded at SC-CA1 synapses depends on d-serine furnished by astrocytes (Lopez-Hidalgo et al. 2012). Noteworthy, α7 nAChR gene deletion leads to impaired synaptic NMDAR functions through specific loss of d-serine in the neocortex (Lin et al. 2014). Likewise, astrocyte TRPA1 channels contribute to basal Ca2+ levels and are required for constitutive d-serine release into the extracellular space, which in turn contributes to NMDA receptor-dependent LTP (Shigetomi et al. 2013). In layer 2/3 of the neocortex, activation of either astroglial cannabinoid or α1-adrenoreceptors receptors induces Ca2+-signaling and synaptic plasticity through notably the release of d-serine (Rasooli-Nejad et al. 2014; Pankratov and Lalo 2015). Such a relationship between the degree of the astrocyte–neuron coupling and the prevalence of d-serine as a co-agonist agrees with the differential impact of the amino acid in activating NMDARs at hypothalamus synapses during lactation where d-serine efficiency is linearly related to the extent of astrocytic coverage of synapses (Panatier et al. 2006). Indeed in virgin rats, astrocytic coverage of synapses is maximal guaranteeing an optimal activity of NMDARs and enabling formation of LTP by d-serine. Conversely, lactation induces loss of synapses coverage by glia. This anatomical remodeling leads to a decrease of occupancy by d-serine of NMDARs (Panatier et al. 2006). As a consequence, the activity dependence of LTP and LTD whose induction depends on NMDAR activation is modified by this neuron–glial remodeling. The levels of d-serine shift the activity dependence of long-term changes toward higher activity (Fig. 3.3).

But, astrocyte-derived d-serine could play also a critical role in neurovascular coupling. Indeed astrocytes modulate hemodynamism of microcirculation by providing a physical linkage from synapses to arterioles and through notably the release of vasoactive gliotransmitters (Haydon and Carmignoto 2006). In two elegant studies, Anderson and colleagues have discovered that endogenous d-serine released by astrocytes contributes to the vasodilatory response of penetrating cortical arterioles produced by astrocyte activation (LeMaistre et al. 2012; Stobart et al. 2013). Still, we could also predict that the progressive setup of a close astrocyte–neuron interaction will favor the astrocyte-derived precursor l-serine to be provided to neurons to boost d-serine synthesis. According to this possibility, the FAC-induced decrease in NMDAR activation recorded at mature SC-CA1 synapses is rescued by increasing synaptic availability of l-serine (Le Bail et al. 2015).

Adult neurogenesis offers another example where the functions of d-serine have been recently explored. In the adult mammalian brain, neurogenesis occurs mainly in the subventricular zone (SVZ) and the subgranular zone (SGZ) of DG in the hippocampus (Aimone et al. 2014). NSCs are the progenitor cells in the brain, which are capable of continuous self-renewal and differentiation into neurons, astrocytes, and oligodendrocytes (Aimone et al. 2014). The local environment may dictate the fate choice of NSCs. Nevertheless, we are still largely ignoring the identity of the extrinsic mediators that regulate proliferation, survival, migration, and differentiation of NSCs. Hu and colleagues have shown that forebrain neural stem cells (NSCs) in vitro could synthesize and release d-serine and that d-serine supports the proliferation and neuronal differentiation of NSCs (Huang et al. 2012). Furthermore, they show that d-serine promotes neurogenesis through acting on NMDARs and subsequently regulating the Ca2+-related signaling pathways, including the phosphorylation of ERK1/2-CREB and GSK-3β. These in vitro observations were later expanded by Toni and colleagues (2013) who reported that d-serine increased adult hippocampal neurogenesis in vivo and in vitro and increased the density of neural stem cells and transit amplifying progenitors. Furthermore, d-serine increased the survival of newborn neurons (Sultan et al. 2013).

3.2 Role of d-Serine in the Developing Brain

Even that d-serine function may preferentially dominate at mature synapses, this co-agonist could already play an important role in the developing brain by controlling neuronal migration and synaptogenesis. Indeed, series of evidence indicate that d-serine and its enzymes are already present early during embryonic and postnatal development in different like the retina (Romero et al. 2014) and the vestibular nuclei (Puyal et al. 2006) where it could critically influence postnatal development. Therefore defects in neonatal d-serine could lead to severe brain defaults. Mice with a 3-phosphoglycerate dehydrogenase deficiency lack both l-serine and d-serine throughout their lives (Yang et al. 2010), but maternal d-serine supplementation can completely reverse the abnormal neurological phenotype (Fuchs et al. 2006).

During neocorticogenesis, migrating neurons can adopt different types of trajectories and a large proportion of neurons migrate radially, along radial glial cells, from the germinative zone to their final place (Yacubova and Komuro 2003). Glutamate acting on NMDARs exerts a crucial role by acting as a motility-promoting signal for immature neurons. Blocking NMDARs impairs neuronal migration during postnatal development (Yacubova and Komuro 2003). Radial migration of immature granule cells in the developing cerebellum, along the Bergmann glia, is one of the best studied models. How NMDARs of migrating immature neurons are activated remains unknown. Because d-serine levels peak at P14 in the cerebellum at the time of intense granule cell migration (Schell et al. 1997) and due to the absolute requirement of NMDARs activation for normal migration, one can imagine that d-serine is involved in this process. Snyder and colleagues (2005) have demonstrated that d-serine released by Bergmann glial cells promotes the migration of granule cells from the external to the internal granular layer through activation of NMDARs (Kim et al. 2005). The physiological influence of d-serine on neuronal migration involves the activation of SR by GRIP.

The motility-promoting role of d-serine is probably not restricted to the postnatal development of the brain and may be involved earlier during fetal development. Indeed SR is present in the perireticular nucleus, a transient structure of the human fetal brain supposed to be engaged in the lamination of the forebrain (Martineau et al. 2006). Furthermore, d-serine and NMDARs are already present at embryonic day 14 (Paoletti et al. 2013; Martineau et al. 2006). Thus d-serine is well positioned both spatially and temporally to control NMDAR-mediated neuronal migration and synaptogenesis. Although SR knockout mice develop normally and show no obvious deficit in lamination or ectopic neurons, the mice have a significantly reduced cortical volume (Balu et al. 2012, 2013). Detailed structural analysis of the CNS of SR knockout mice reveals that SR deletion when achieved before weaning results in altered cortical dendritic morphology of pyramidal neurons in the prefrontal cortex and in the primary somatosensory cortex (DeVito et al. 2011) supporting the hypothesis that d-serine through the modulation of NMDARs could contribute to neuronal differentiation and therefore to synaptogenesis. Noteworthy, SR knockout mice have a reduction in total brain-derived neurotrophic factor (BDNF) protein levels (Balu et al. 2012), an activity-regulated neurotrophin that plays a key role in promoting synapse formation and maturation and in regulating the functional development of neuronal circuits (Park and Poo 2013). Accordingly, Gomes and colleagues (2012) have discovered that astrocytes induce the formation of glutamatergic synapses through transforming growth factor (TGF)-β1 pathway and the TGF-β1 synaptogenic property is dependent on d-serine signaling (Diniz et al. 2012). Indeed, the growth factor increases d-serine release from astrocytes, and blocking the synthesis of d-serine or its function prevents TGF-β1 to promote the development of functional glutamatergic synapses in the cerebral cortex. Another study demonstrates that differentiation of P19 cells induces SR expression and d-serine formation and that d-serine shapes synaptogenesis, potentially by preventing widespread untargeted synaptogenesis (Fuchs et al. 2012). These results contribute to expand evidence for a role of d-serine in synaptic shaping and wiring of neuronal circuitry during brain development.

3.3 Role of d-Serine in Cognition and Healthy Aging

In addition to their role in shaping neuronal circuitries and in synapse dynamics, NMDARs are essential for memory formation and for higher cognitive functions and social interactions (Paoletti et al. 2013). Genetically or pharmacologically driven hypofunction of NMDARs leads to profound deficits in cognition and social activity (Paoletti et al. 2013). For example, mice with a point mutation in Grin1(D481N) showed abnormally persistent latent inhibition (a measure of information-processing deficit), reduced social approach behaviors, and enhanced startle reactivity (measure of sensory gatting) and impairments in nonassociative spatial object recognition task (Labrie et al. 2008). These deficits are reversed by administration of agents that enhance NMDAR glycine site function like d-serine (Labrie et al. 2008). Various regimens of systemic administration of low-dose d-serine (50 mg/kg/day) increase performances on BALB/c mice performing object recognition, T-maze alternation, and open-field exploration tasks (Bado et al. 2011). Analysis of SR-KO mice revealed that the animals exhibited a significantly disrupted representation of the order of events in distinct experimental paradigms as shown by object recognition and odor sequence tests (DeVito et al. 2011); However, SR mutant mice showed normal detection of novel objects and in spatial displacement and showed intact relational memory in a test of transitive inference. SR mice exhibited normal sociability and preference for social novelty (DeVito et al. 2011). Analysis of mice with inactive DAAO further offers complementary evidence that d-serine is crucial for normal cognition and social behavior. First, Roder and colleagues have generated mice carrying point mutations in DAO (Dao1 G181R). These mice have no active DAAO and display elevated brain levels of d-serine. Introducing the Dao1 G181R mutations in Grin1(D481N) mice corrects all cognitive deficits and normalizes social behavior deficit normally observed in Grin1(D481N) mice (Labrie et al. 2010). Second, Dao−/− mice generated more recently by Peirson and colleagues (2015) demonstrate enhanced spatial recognition memory performance, improved odor recognition memory performance, and enhanced spontaneous alternation in the T-maze (Pritchett et al. 2015). Overall, these observations are fully consistent with, and extend, findings in the natural mutant ddY/Dao− line introduced and largely popularized by Konno and colleagues (Almond et al. 2006; Zhang et al. 2011). Thus, an increased level of d-serine resulting from decreased catalysis (i.e., DAAO inactivation) increases the cognitive performance and behavior of mice in striking contrast to SR deficient mice (Fig. 3.3). Building on this ground, inhibitors for DAAO show remarkable benefit in improving cognitive performance in animal models of NMDAR hypofunction (Smith et al. 2009; Strick et al. 2011; Hopkins et al. 2013a, b). Nicotine has long been known to improve long-term hippocampal-dependent memory in both laboratory animals and humans (Changeux 2010). García-Colunga and colleagues (2012) have found that the facilitating effect of nicotine on long-term memory evaluated with the one-trial step-through inhibitory-avoidance task depends on glial d-serine (Lopez-Hidalgo et al. 2012). Accordingly, nicotine activates nAChRs present on glial cells stimulating the release of d-serine and then the activation of NMDARs in the hippocampus. These data further expand the earlier observation that glial cells through the release of d-serine contribute to spatial memory retrieval (Zhang et al. 2008).

Non-pathological brain aging is certainly one of the best documented situation where accumulated experimental evidence have shown that deficits in synaptic plasticity and cognitive functions notably memory and learning result from reduced activation of NMDARs. In collaboration with Jean-Marie Billard, we have discovered that those deficits are caused by a reduction in the levels of d-serine produced by SR during aging (Mothet et al. 2006; Turpin et al. 2011) thus resulting in a hypofunction of NMDARs (Fig. 3.3). More detailed analysis revealed that oxidative stress as occurring during aging affects SR and then the production of d-serine and that treatment with the reducing agent N-acetyl cysteine prevents hippocampal synaptic plasticity deficits by protecting d-serine-dependent NMDA receptor activation (Haxaire et al. 2012). Recently the notion that d-serine contributes to age-related cognitive decline has been expanded by Saitoe and colleagues who demonstrated that age-related memory impairments in Drosophila melanogaster are a result of reduced d-serine production by glia and that d-serine feeding suppresses these deficits (Yamazaki et al. 2014) thus showing that d-serine metabolism and functions may be conserved throughout evolution (Schell 2004).

4 Conclusion

d-serine has overturned fundamental axioms of biology. Intensive research during the last two decades has shown that this d-amino acid has emerged as a very important brain messenger with implication for Human disease. Despite many progresses since the breakthrough of Nishikawa (1992) and Snyder (1995) groups, there are still some opened questions regarding the functions of d-serine in the CNS and in the peripheral nervous system. The discovery that NMDAR co-activation takes more than glycine illustrates perfectly there’s no simple rule governing the regulation of NMDA receptors. There is still a great need for efforts to discriminate the relative contribution of these two NMDAR neuromodulators in the healthy and diseased nervous system.