Serine Racemase Expression by Striatal Neurons

Abstract

d-serine is synthesized by serine racemase (SR) and is a co-agonist at forebrain N-methyl-d-aspartate receptors (NMDARs). d-serine and SR are expressed primarily in neurons, but not in quiescent astrocytes. In this study, we examined the localization of d-serine and SR in the mouse striatum and the effects of genetically silencing SR expression in GABAergic interneurons (iSR−/−). iSR−/− mice had substantially reduced SR expression almost exclusively in striatum, but only exhibited marginal d-serine reduction. SR positive cells in the striatum showed strong co-localization with dopamine- and cyclic AMP-regulated neuronal phosphoprotein (DARPP32) in wild type mice. Transgenic fluorescent reporter mice for either the D1 or D2 dopamine receptors exhibited a 65:35 ratio for co-localization with D1and D2 receptor positive cells, respectively. These results indicate that GABAergic medium spiny neurons receiving dopaminergic inputs in striatum robustly and uniformly express SR. In behavioral tests, iSR−/− mice showed a blunted response to the hedonic and stimulant effects of cocaine, without affecting anxiety-related behaviors. Because the cocaine effects have been shown in the constitutive SR−/− mice, the restriction of the blunted response to cocaine to iSR−/− mice reinforces the conclusion that d-serine in striatal GABAergic neurons plays an important role in mediating dopaminergic stimulant effects. Results in this study suggest that SR in striatal GABAergic neurons is synthesizing d-serine, not as a glutamatergic co-transmitter, but rather as an autocrine whereby the GABAergic neurons control the excitability of their NMDARs by determining the availability of the co-agonist, d-serine.

Introduction

Since d-serine was discovered in the mammalian brain, it has emerged as an important N-methyl-d-aspartate receptor (NMDAR) co-agonist (Hashimoto et al. 1992). NMDARs consist of two GluN1 subunits that form an ion channel, usually blocked by a single magnesium (Mg²⁺) ion, and two GluN2 subunits that bind glutamate. Receptor activation requires concurrent membrane depolarization to release the Mg²⁺ ion from the channel, glutamate binding to the GluN2 subunits, and either glycine or d-serine binding at the glycine modulatory site (GMS) on the GluN1 subunits. d-serine has three times higher binding affinity at GluN1 than glycine; so, d-serine is the primary endogenous co-agonist for forebrain NMDARs (Balu and Coyle 2015; Matsui et al. 1995).

d-serine is synthesized from l-serine by the cytosolic enzyme serine racemase (SR), a pyridoxal 5′ phosphate (PLP) containing enzyme (Coyle et al. 2010; Nishikawa 2011). The reaction is reversible and requires Mg²⁺ and ATP as cofactors (Goto et al. 2009; Balu et al. 2014). SR-null mutant (Srr−/−) mice have brain d-serine levels 85–90% less than wild-type (WT) mice. The remaining d-serine most likely comes from diet or intestinal bacteria (Basu et al. 2009).

SR is enriched in cortico-limbic regions of the adult brain, whereas glycine appears to predominate in the cerebellum and brainstem. Because astrocytes produce the d-serine precursor, l-serine, and express its catabolic enzyme, d-amino acid oxidase (DAAO), SR was originally thought to be concentrated in glia (Mothet et al. 2005; Schell et al. 1995; Wolosker et al. 1999). Reactive astrocytes express SR (Balu et al. 2019; Perez et al. 2017), and cultured astrocytes begin expressing SR as they acquire the “reactive” or inflammatory phenotype (Li et al. 2018). But immunohistochemistry with Srr−/− mice as controls shows that under healthy physiological conditions, SR is expressed primarily by neurons (Miya et al. 2008; Ding et al. 2011; Ehmsen et al. 2013; Wolosker et al. 2016). These immunocytochemical findings were confirmed by conditionally silencing SR expression in a cell-specific fashion. By crossing Srrfl/fl mice with a Cre recombinase line driven by the promoter for glial fibrillary acidic protein (GFAP), SR expression in astrocytes would be expected to be suppressed; however, this astrocyte selective construct had no effect on the expression of SR in cortex. In contrast, crossing Srrfl/fl mice with a Cre recombinase line driven by the promoter for Ca²⁺/calmodulin-dependent kinase II (CaMKII) eliminated SR in cortico-limbic glutamatergic neurons, resulting in major reductions in SR expression (Balu et al. 2014; Balu and Coyle 2011). In the present study, we used mice with Cre recombinase driven by the promoter for glutamic acid decarboxylase-65 (GAD65; CreGAD2) to silence Srrfl/fl expression in GABAergic neurons. In this report, we have focused primarily on the impact of silencing SR expression in striatal GABAergic neurons.

Methods

Animals

WT, constitutive Srr−/− , and selective neuronal Srr−/− (nSR−/−) or astrocytic SR−/− (aSR−/−) mice were bred as previously described, by crossing SRfl/fl mice with Cre recombinase lines driven by either a CaMKII neuronal promoter or a glial fibrillary acidic protein (GFAP) astrocytic promoter (Basu et al. 2009; Benneyworth et al. 2012). We bred SR+/− parents to produce WT and SR−/− siblings and used SRfl/fl mice as controls against their iSR−/−, nSR−/− and aSR−/− littermates. All mice were backcrossed at least ten generations onto a C57BL/6J background.

Glutamic acid decarboxylase synthesizes the inhibitory neurotransmitter γ-aminobutyric acid (GABA) by decarboxylating glutamate. It has a 65-kDa isoform (GAD65; GAD2) and a 67-kDa isoform (GAD67; GAD1), produced by two distinct GABAergic populations. GAD2 mCherry mice which have a transgenic fluorescent protein driven by the promoters for the gene (Peron et al. 2015) were obtained from the Jackson Laboratory. A GAD2 Cre recombinase line [Gad2tm2(cre), Jackson Laboratories, Bar Harbor, ME] was bred with srrfl/fl mice to selectively remove SR from GAD65-expressing interneurons (iSR−/−).

Striatal inhibitory neurons can also be classified by their dopamine receptor expression. To examine SR expression in D1 direct projections and D2 indirect projections, we obtained D1- and D2-tdTomato mice, which have fluorescent transgenes driven by the promoters for either D1 or D2 dopamine receptors (Thibault et al. 2013); kindly provided by Michael Benneyworth, PhD, University of Minnesota). Three-month-old mice were used for experiments on adults. Animals were housed in a temperature- and humidity-controlled facility (22 °C) on 12 h light/dark cycles with food and water ad libitum. All animal procedures were approved by the McLean Hospital Institutional Animal Care and Use Committee.

Western Blotting

The striatum was dissected for immunoblotting, as previously described (Balu and Coyle 2011), with goat anti SR (1:500; Santa Cruz Biotechnology, Santa Cruz, CA), an antibody validated against Srr−/− mice (Wolosker et al. 2016; Takagi et al. 2015). β-Actin was a loading control. Antiserum (goat polyclonal) against d-amino acid oxidase (DAAO) was purchased from Abcam (ab34697).

High-Performance Liquid Chromatography (HPLC)

As previously described (Benneyworth et al. 2012), the striatum was dissected and homogenized, then the protein concentration was adjusted to 1 mg/mL and 4 mM/g l-homocysteic acid was added as an internal standard. Proteins were precipitated from the samples by centrifuging in 5% trichloroacetic acid for 30 min at 18,000 × g in 4 °C; and then washing three times with water-saturated diethyl ether.

Amino acids were derivatized with o-phthaldialdehyde (Alfa Aesar, Ward Hill, MA) and N-tert-butyloxycarbonyl-l-cysteine (Novabiochem, Gibbstown, NJ) as previously described (Benneyworth et al. 2012), and concentrations were calculated by normalizing each internal standard to a standard sample run that morning. The HPLC system used consists of an SCL-10A controller, two LC-10AT VP pumps, an SIL-10AD auto injector, a DGU-20A5 degasser and an RF-551 fluorescence monitor (all from Shimadzu Corporation, Kyoto, Japan).

Immunohistochemistry and Immunofluorescence

SR-expressing neurons were counted on immune-histochemically stained 20-μm sections from SRfl/fl and iSR−/− mice after transcardially perfusing them with 0.1 M phosphate-buffered saline (PBS) and 4% paraformaldehyde. As previously described (Balu et al. 2014), the sections were incubated ~ 40 h at 4 °C with mouse anti SR (1:2000, BD Biosciences, San Jose, CA) diluted in PBS with 3.0% bovine serum albumin and 0.1% Triton X-100 followed by 90 min in a biotinylated horse anti-mouse secondary antibody (1:1000, Vector Laboratories; Burlingame, CA). The immunocomplexes were stained with 3,3-diaminobenzidine (DAB) using ABC Elite reagents (Vector Laboratories) according to manufacturers’ instructions. Experimental and control samples were processed in parallel.

Dual immunofluorescent staining was used to determine which subpopulations of striatal neurons expressed SR. Slices were incubated overnight at room temperature with SR anti-goat (1:100, Santa Cruz) along with either mouse anti-DARPP-32 (1:2000, Santa Cruz), mouse anti-CamKII (1:500, Abcam), mouse anti-GAD67 (1:100, Millipore, Billerica, MA), goat anti-choline acetyl transferase (ChAT) (1:500, Millipore) or mouse anti-parvalbumin (1:1000, Sigma-Aldrich, St. Louis, MO) followed by Alexa Fluor secondary antibodies. Contrast and brightness of immunofluorescent images in the figures were adjusted in PowerPoint 2011 (Microsoft, Redmond, WA).

SR co-localization with D1 and D2 dopamine receptors was calculated using the Optical Fractionator module of Stereo Investigator (MBF Bioscience, Williston, VT). The chosen grid sizes (500 µm² in striatum; 650 µm² in nucleus accumbens) and counting frame (120 µm²) estimated densities with high precision (coefficient of error < 0.09). Three sections per region spanning the rostral–caudal axis according to Paxinos and Franklin (2001) were counted from three mice and balanced by hemisphere.

Behavioral Testing

Elevated Plus Maze

The elevated plus maze consisted of four arms elevated 80 cm above the floor. Each arm (12 cm wide, 76 cm long) was joined to the others by a central square (12 cm × 12 cm) in a cross-like disposition. A wall (30 cm in height) enclosed two opposite arms, while the other two arms were open. Placing the mouse on the platform facing an open arm started the test. During the next 10 min the number of entries in open or closed-arms and the time spent in open or closed-arms was recorded. The mice were considered to have entered an arm when all four limbs were located in an arm of the maze. All sessions were video recorded by a camera mounted above the maze and analyzed in real-time using EthoVision XT software (version 7.1; Noldus Information Technology, Leesburg, VA). The percentage of time the mice spent in the open arms and the percentage of entries into open arms were calculated.

Open Field Test

Spontaneous locomotor activity and habituation to a novel open field environment were assessed using clear Plexiglas® chambers fitted with three 16-beam I/R arrays and placed inside sound-attenuating cubicles (ENV-510 System, Med Associates, Inc., St. Albans, VT, USA). Photobeam breaks were recorded and translated into locomotor activity using open field activity software (Activity Monitor 5.0, Med Associates, Inc.) The center zone was defined using the coordinates (4.5, 12; 12, 12; 4.5, 4.5; 12, 4.5) of the 16 × 16 cm chamber. The experiment time was 600 s. The distance moved by mice and the duration of time spent in the center zone were recorded.

Place Conditioning Test

Place conditioning was conducted in four identical acrylic chambers (30 cm × 15 cm × 15 cm), as previously described (Smith et al. 2014). Chamber floors consisted of interchangeable halves with distinct textures (hole or grid). Hole floors were made from perforated 16-gauge stainless steel mounted on acrylic rails with 6.4-mm round holes placed at 9.5-mm staggered centers. Grid floors were made from 2.3-mm stainless steel rods mounted on acrylic rails 6.4 mm apart. All place-conditioning sessions were video recorded and analyzed in real-time using EthoVision XT software.

Place-conditioning procedures were conducted, as previously described (Smith et al. 2014; Bechtholt et al. 2004; Cunningham 1995; Cunningham et al. 1999) in the apparatus described above. Specifically, animals were randomly assigned to GRID+ or GRID− conditioning groups, with genotype (SRfl/fl, iSrr−/−) counterbalanced. Conditioning consisted of five phases, as follows: habituation, conditioning, preference testing, and extinction (Fig. 2). Prior to each session, animals were transported from the colony room to the place-conditioning room where they were allowed to habituate for 1 h. Lighting was dimmed (∼10 lx). On day 1 (habituation), all animals received injections (i.p.; injection volume 10 ml/kg) of 0.9% bacteriostatic saline and were immediately placed in the conditioning chambers on a smooth Plexiglas floor with white matting beneath it for 30 min. On days 2 and 3 (conditioning), animals received injections (i.p.; injection volume 10 ml/kg) of saline in the morning and cocaine (20 mg/kg) in the afternoon (4 h later) and were immediately placed in the conditioning chambers for 30 min. GRID+ animals were placed on the hole side for saline injections and on the grid side for cocaine injections, whereas GRID− animals were placed on the grid side for saline injections and on the hole side for cocaine injections. On day 4 (preference testing), all animals received injections (i.p.; injection volume 10 ml/kg) of 0.9% bacteriostatic saline and were immediately placed in the conditioning chambers with access to both the grid and hole sides for 30 min. Days 5–7 (extinction) were identical to day 4 (preference testing).

Fig. 2.

SR reduction in striatum of nSR−/− mice is similar to iSR−/− . a and b SR positive cells are as sparse in nSR−/− as in iSR−/− mice. White bar in a indicates 200 μm and in b indicates 40 μm. c and d SR reduction in striatum (c) and nucleus accumbens (d) of nSR−/− mice is confirmed by cell counting and it is similar to iSR−/− . The number of SR positive cells was counted in the striatum and nucleus accumbens of SRfl/fl (n = 3; white bars) and nSR−/− (n = 3; black bars) mice. Values are expressed as the estimated cell population (cells/cm³). All values represent the mean ± SEM

Intracranial Self-stimulation (ICSS)

ICSS procedures were performed as previously described (Puhl et al. 2019). The mice (approximately 2 months old) were anesthetized (i.p., ketamine 80 mg/kg and xylazine 10 mg/kg), and monopolar-stimulating electrodes were stereotactically implanted in the medial forebrain bundle (in mm from bregma, anterior–posterior: − 1.9, mediolateral: − 1.0, dorsal–ventral: − 5.0 below dura, according to the atlas of Paxinos and Franklin 2001). Following ICSS electrode implantation surgery, the animals were individually housed and allowed a 1-week recovery period. Thereafter, all mice were trained to respond for brain stimulation by spinning an operant wheel during daily 1-h sessions over the course of 2–4 weeks (Muschamp et al. 2012). Stimulation frequency was fixed at 141 Hz, and stimulation current was adjusted to the lowest value that would support stable responding (i.e. one quarter wheel turn, which translated to one brain stimulation reward, per second). After identifying this minimum current for each mouse, the current was held constant during subsequent training sessions, where the mice were allowed to respond to 15 stimulation frequencies presented in descending order (0.05 log₁₀ unit steps) during fifteen 50-s trials. Each trial was preceded by a 5-s priming period, during which noncontingent stimulation was given at the frequency used for that trial. Also, each trial was followed by a 5-s timeout period, during which responding was not reinforced. Responding was recorded during each 50-s trial in each set of 15 trials (termed a “pass”). The range of frequencies was adjusted over the course of training for 4–6 weeks, so that the mice responded through the highest six to seven frequencies stably over six passes (90 min of training). The lowest frequency that supported responding (ICSS threshold or T0) was computed with least-squares line of best-fit analysis (Carlezon and Chartoff 2007). When the animals were observed to have stable mean ICSS thresholds (< 10% variability over three consecutive days), the effects of vehicle and different doses of cocaine (3, 6, 12 mg/kg) on ICSS threshold were determined. The mice responded through three passes immediately before drug treatment, and thresholds from the second and third passes were averaged to obtain the baseline (ICSS threshold and maximal response rate) parameters. Each mouse then received either vehicle or cocaine (3, 6, 12 mg/kg) and was tested for three passes. The mice were treated with vehicle or each dose of cocaine twice, in ascending then descending order, with a minimum of 2 days between treatments.

Statistical Analyses

The results of western blotting, HPLC, elevated plus maze, place preference and open field test were compared using unpaired Student t-test. Mixed factorial and one-way analysis of variance (ANOVA) tests were conducted for ICSS. These analyses were done with the GraphPad PRISM software (version 7.03, GraphPad Software, Inc., La Jolla, CA) Values of p < 0.05 were considered statistically significant.

Results

iSrr−/− Mice Have Reduced SR Expression in Striatum Without a Corresponding Reduction in d-Serine

To examine the role of SR in striatal inhibitory neurons (Balu et al. 2014), we selectively silenced the Srr gene in GAD65-expressing interneurons and measured the resulting changes in protein expression and d-serine concentrations. We found significant SR protein decreases in striatum (p = 0.015) and modest effects on hippocampal and cortical SR in adult mice (Fig. 1a–c).

Fig. 1.

Western blot and cell counting showing SR reduction in striatum of iSR−/− mice. a–c Reduction of SR protein in iSR−/− mice. In iSR−/− mice, SR is reduced in striatum. There is no or very small reduction in hippocampus or frontal cortex. Protein levels of SR was measured in the striatum/hippocampus/frontal cortex of SRfl/fl mice (n = 8; white bars) and iSR−/− (n = 8; black bars) mice. Values are expressed as the optical density (OD) normalized to fl/fl mice. Asterisk indicates significant differences from the SRfl/fl mice (p < 0.05). d–g SR positive cells are sparse in iSR−/− mice (f, g) than SRfl/fl (d, e). White bars in d and f indicate 200 μm and in e and g indicate 40 μm. h and i SR reduction in striatum (h) and nucleus accumbens (i) of iSR−/− mice is confirmed by cell counting. Number of SR positive cells was counted in the striatum and nucleus accumbens of SRfl/fl (n = 3; white bars) and iSR−/− (n = 3; black bars) mice. Values are expressed as the estimated cell population (cells/cm³). All values represent the mean ± SEM

We found approximately 70% fewer SR-expressing cells in striatum and 65% fewer in the nucleus accumbens of iSR−/− mice than in the SRfl/fl controls (Fig. 1h–i). We found nearly the same reductions in nSR−/− mice in the nucleus accumbens (Fig. 2). It should be emphasized that nSR construct silences SR expression only in cells that express CamKII. This finding indicates that the GAD65 cells targeted in our iSR−/− mice account for nearly all CamKII expressing neurons in these regions, which we confirmed by co-staining SRlf/fl striatal slices for SR and CamKII. In both the immunohistochemical and immunofluorescence studies, no SR expression was observed in glia.

We used HPLC to measure the d-serine concentrations in the same samples and observed only a small and statistically insignificant reduction in d-serine in striatum (p = 0.140) and other areas (Fig. 3a–c). This finding is perplexing given the large reduction in SR, but a similar result was previously observed in nSR−/− mice, which showed consistent strong reductions in SR protein but a much smaller reduction in d-serine levels (Benneyworth et al. 2012). We measured striatal expression of DAAO, an astrocytic protein that catabolizes d-serine, to determine whether iSR−/− mice maintained d-serine homeostasis by decreasing DAAO expression. However, DAAO protein levels in iSR− /− mice were not affected (Fig. 3d).

Fig. 3.

d-serine concentrations in the iSR−/− mice show only a small and statistically insignificant reductions in striatum and other areas. a–c HPLC was used to measure d-serine in iSR−/− and SRfl/fl mice. d-serine was measured in the striatum, hippocampus and frontal cortex of SRfl/fl (n = 8; white bars), iSR−/− mice (n = 8; black bars). Values are expressed as the μmol/g tissue. All values represent the mean ± SEM. d DAAO protein is not reduced in striatum of iSR−/− mice. Protein levels of DAAO were measured in the striatum of SRfl/fl (n = 8; white bars) and iSR−/− (n = 8; black bars) mice. Values are expressed as the optical density (OD) normalized to SRfl/fl

Medium Spiny Neurons Express Striatal SR; Cholinergic and Parvalbumin Expressing Neurons Do Not

Approximately 95% of striatal neurons are medium spiny neurons (MSN), identifiable by dopamine- and cAMP-regulated neuronal phosphoprotein 32 (DARPP32) expression (Kemp and Powell 1971; Ouimet et al. 1984, 1998). We observed strong co-localization of SR with DARPP32 (Fig. 4a–c). The striatum has a small but densely branching population of cholinergic interneurons, identifiable by ChAT expression (Phelps and Vaughn 1986; Dautan et al. 2014), These striatal cholinergic neurons do not co-express SR (Fig. 4d). The striatum also contains a population of isolated fast-spiking interneurons that express parvalbumin. These interneurons also do not co-express SR (Fig. 4e).

Fig. 4.

Medium spiny neurons express striatal SR; cholinergic and parvalbumin expressing neurons do not. a–c Robust co-localization of SR with DARPP32 is observed. White bars indicate 40 μm. a DARPP32, b SR, c merged figure of a and b. DARPP32 was shown in green fluorescence. SR was shown in red fluorescence. White bars indicate 40 μm. d Striatal cholinergic neurons do not co-express SR. ChAT is shown in green fluorescence. SR is shown in red fluorescence. White bars indicate 40 μm. e Fast-spiking interneurons do not co-express SR. Parvalbumin is shown in green fluorescence. SR is shown in red fluorescence. White bars indicate 200 μm

Association of SR Expression with Dopamine Receptors

MSN’s have two distinct subtypes distinguished by the dopamine receptor subtype expressed and their projection patterns. D1-receptor-expressing neurons project directly to basal ganglia output nuclei, whereas D2-receptor-expressing neurons project indirectly by way of the globus pallidus and sub-thalamic nuclei (Soares-Cunha et al. 2016; Surmeier et al. 2007). We stained for SR in D1R and D2R-tdTomato expressing mice, and found that ~ 66% of striatal SR cells expressed D1R and that ~ 37% expressed D2R (Table 1).

Table 1.

SR co-localization with D1 and D2 dopamine receptors in striatum

	SR cell count	DR cell count	% DR cells in SR	% SR cells in D1/D2R
D1-tdTomato	892	314	100	37
D2-tdTomato	426	280	100	66

Reporter mice with tdTomato transgene expressed by either D1 or D2 dopamine receptors showed complete co-expression with SR in immunofluorescent-stained slices. Cells were counted using Stereo Investigator stereology software for unbiased random sampling throughout the striatum. The results the total number of cells ascertained from three sections from three mice

DR dopamine receptor

iSR−/− Mice Showed Impaired Sensitization to Cocaine

Constitutive SR−/− have shown a blunted response to the hedonic and stimulant effects of cocaine (Puhl et al. 2019), as well as elevated basal extracellular levels of glutamate and dopamine with blunted responses to cocaine in the ventral striatum. Consistent with these results, the iSR−/− mice showed significantly less cocaine locomotor sensitization than SRfl/fl mice (Fig. 5e; p = 0.010) and did not maintain a preference for the cocaine chamber on extinction days in place conditioning test, whereas the SRfl/fl controls did not extinguish (Fig. 5e–g).

Fig. 5.

iSR−/− mice showed impaired sensitization to cocaine. a and b Elevated Plus Maze test shows no difference in anxiety between SRfl/fl (n = 8, black dots) and iSR−/− (n = 8, black squares). a The percentage of time that mice spent in the open arms in the total experiment time is shown. SRfl/fl do differ from iSR−/−. b The percentage of the number of entries in open arms in the number of entries in both open and closed arms is shown. SRfl/fl do not differ from iSR−/−. c and d Open field test shows no difference in anxiety between SRfl/fl (n = 8, black dots) and iSR−/− mice (n = 8, black square). c The distance travelled by mice during the experiment time is not different between SRfl/fl (n = 8, black dots) and iSR−/− mice (n = 8, black square). d The time spent in the center zone by mice during the experiment time is not different between SRfl/fl (n = 8, black dots) and iSR−/− mice (n = 8, black squares). e–g The place conditioning task with cocaine shows less cocaine locomotor sensitization for iSR−/− mice, which do not maintain a preference for the cocaine chamber on extinction days. e On the conditioning days, iSR−/− mice (n = 9, black bars) travel significantly less distance on second conditioning day than SRfl/fl mice (n = 9, white bars, p = 0.010). f There is no difference in the distance travelled on test day and extension days between iSR−/− (black squares) and SRfl/fl mice (white dots). g At the test day, both iSR−/− mice (n = 9, black squares) and SRfl/fl (n = 9, white dots). show a significant preference for the cocaine chamber (0.05). However, iSR−/− mice do not show a significant preference for the cocaine chamber on the subsequent extinction days whereas SRfl/fl maintained significant preference for the cocaine chamber (p = 0.014, 0.001, 0.001, respectively). h and i iSR−/− mice react less in ICSS threshold following cocaine administration. h ICSS thresholds showed significant main effect of cocaine dose (F = 46.756, p < 2e−16) and close to significant main effects of genotype (F = 3.803, p = 0.0522). A t-test for the result at 12 mg/kg showed significant p value (p = 0.022). iSR−/− (n = 5, black bars), SRfl/fl (n = 7, white bars). i ICSS maximum response rates show significant main effects of cocaine dose (F = 5.429, p = 0.0012). Although a tendency for less dependent response to cocaine dose in iSR−/− mice (black bars) than in fl/fl mice (white bars) is confirmed, the main effect of genotype was not found to be significant

In the ICSS experiment, a consistent result was also obtained (Fig. 5h, i). Mice were implanted with ICSS electrodes and trained to respond to electrical stimulation (see Methods section). Upon reaching the stability criteria for ICSS responding, the mice received cocaine treatment. ICSS thresholds and maximum response rates were analyzed by using repeated measures ANOVAs varying genotype (SRfl/fl or iSR−/−) and cocaine dose (0, 3, 6 or 12 mg/kg). For ICSS thresholds, significant main effects of cocaine dose (F = 46.756, p < 2e−16) and genotype (F = 3.803, p = 0.05) were found. In the dose response study, an unpaired t-test showed a significant genotype difference at the 12 mg/kg dose (p = 0.0222), but not at the lower doses. This result suggests that iSR−/− mice react less in ICSS threshold following cocaine administration, consistent with previous results (Puhl et al. 2019). These results suggest that striatal GAD65 interneurons synthesize the d-serine involved in the hedonic response to cocaine.

Discussion

The striatum receives cholinergic innervation from interneurons and brainstem sources (e.g. latero-dorsal tegmental area, pedunculopontine nuclei) and GABAergic innervation from striatal MSNs (Kita 1993; Dautan et al. 2014). The MSNs in turn project directly to the substantia nigra or indirectly to the substantia nigra via the globus pallidus. Additionally, the striatum receives dopaminergic projections from the ventral tegmental area (VTA) and the substantia nigra (SNr) and glutamatergic inputs from several areas, including the cortex, hippocampus, amygdala, and thalamus (Swanson 1982; Phillipson and Griffiths 1985; Finch 1996; Groenewegen et al. 1999; Britt et al. 2012). These glutamatergic inputs synapse onto the heads of dendritic spines of the striatal GABAergic MSNs, whereas dopaminergic inputs make synapses en-passage onto the spine neck, allowing for an important and complex interaction between these two inputs in the modulation of MSN activity (Freund et al. 1984; Xu et al. 1989).

Miya et al. (2008) first described the expression of SR in GABAergic MSNs in the striatum. They reported that the SR immunoreactivity was primarily localized to the perikarya and the shafts of the dendrites. Our study confirms the universal expression of SR in the striatal MSNs including the 65% that express dopamine D2 receptors and project to the substantia nigra and the 35% that express dopamine D1 and project to the globus pallidus. In contrast, we found that a sub-population of fast-firing PV-expressing GABAergic interneurons in the striatum, which modulate the activity of MSNs (Lee et al. 2017), do not express SR. PV-expressing GABAergic neurons in the amygdala, an extension of the striatum, also did not express SR (Balu et al. 2018). In contrast, d-serine was co-localized with a substantial but variable portion of cortical and hippocampal PV-expressing interneurons (Balu et al. 2014), thus suggesting that co-localization of PV and SR is not uniform across brain regions. In addition, the striatal cholinergic interneurons do not express SR.

These observations of SR expression in subpopulations of GABAergic neurons raise the question about the function of d-serine in inhibitory neurons. Obviously, d-serine is not a “co-transmitter” with GABA as the post-synaptic GABAA receptors do not have a glycine modulatory site. A more parsimonious explanation is that d-serine is serving as a co-agonist at the NMDARs on the GABAergic dendrites post-synaptic to glutamatergic afferents. Further undermining the role of d-serine as a co-transmitter is the evidence that it is not released by Ca²⁺-dependent exocytotic mechanisms but rather facilitated transport, likely by AscT-1 (Sason et al. 2017). While the evidence supports an autocrine role for d-serine synthesized in the striatal MSNs, this finding does not preclude a potential role for facilitated release of d-serine from striatal glutamatergic afferents innervating the MSNs.

The recent tissue culture study of cerebral cortex by Lynch’s laboratory elegantly demonstrated the clumping of SR immunoreactivity around the post-synaptic density in dendrites under the VGut-expressing glutamatergic terminals on both glutamatergic and GABAergic neurons (Lin et al. 2016). This distribution of SR is consistent with the original description by Miya et al. (2008) of SR having a predominant dendritic localization in striatal MSNs. Our recent research has demonstrated that SR expression is dynamically regulated with increased expression in circumstances associated with increased glutamatergic neurotransmission (Balu et al. 2018). This regulation of the SR expression by glutamatergic neurotransmission is consistent with a post-synaptic localization of SR.

Our recent in vivo dialysis study revealed a striking and significant elevation of basal dopamine release in the ventral striatum and a blunted response to a cocaine challenge (Puhl et al. 2019). This finding is consistent with a disinhibition of dopaminergic neurons in the substantia nigra and VTA. Given the major GABAergic innervation of the dopaminergic perikarya and proximal dendrites (Bolam and Smith 1990), reduced GABAergic inhibitory drive due to impaired glutamatergic neurotransmission resulting from deficient function of NMDA receptors could account for the increased dopamine release in the striatum. This finding suggests that pharmacologically enhancing NMDA receptor function could counter-intuitively reduce dopaminergic disinhibition. Pharmacologic and gene knock-out studies support this hypothesis that enhanced NMDA receptor function is associated with increased locomotor sensitization with stimulants and ICSS (Benneyworth and Coyle 2012; Puhl et al. 2015; Areal et al. 2019).

In summary, SR is robustly expressed in striatal GABAergic neurons post-synaptic to a diffuse glutamatergic input from the cerebral cortex and other structures. The localization of SR to dendrites and, more specifically, within the post-synaptic density adjacent to glutamatergic terminal boutons suggests that SR is synthesizing d-serine not as a glutamatergic co-transmitter but rather as an autocrine, whereby the GABAergic neurons control the excitability of their own post-synaptic NMDA receptors by determining the availability of the co-agonist, d-serine. These GABAergic NMDA receptors appear to modulate the locomotor sensitization associated with repeated exposure to stimulants and ICSS.

Author Contributions

ST and JTC designed the study. DTB and JTC contributed to analysis and interpretation of data and assisted in the preparation of the manuscript. MDP and TA contributed to the histochemical studies and animal behavior testing. All authors had contributed to data collection and interpretation, and critically reviewed the manuscript. All authors approved the final version of the manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Compliance with Ethical Standards

Conflict of interest

JTC reports consulting with Concert Pharm and holding a Patent on d-serine for the Treatment of Serious Mental Illness, which is owned by Massachusetts General Hospital. MDP is currently employed by Alkermes Pharm. DTB served as a Consultant for LifeSci Capital and received research support from Takeda Pharmaceuticals. ST and TA report no conflicts of interest.

Ethical Approval

All animal procedures were conducted in accordance with the Guide for the Care and Use of Laboratory Animals (National Academy of Sciences, 2011) and were approved by the McLean Hospital Institutional Animal Care and Use Committee.