Serine Racemase Regulated by Binding to Stargazin and PSD-95

Background: d-Serine, generated by serine racemase (SR), is an endogenous co-agonist for NMDA receptors. Results: SR binds PSD-95 and stargazin, which inhibits SR enzymatic activity. This complex is disrupted by AMPA receptor activation, activating d-serine synthesis. Conclusion: SR/stargazin/PSD-95 interactions mediate NMDA/AMPA receptor cross-talk. Significance: d-Serine may link AMPA and NMDA neurotransmission. d-Serine, an endogenous co-agonist for the glycine site of the synaptic NMDA glutamate receptor, regulates synaptic plasticity and is implicated in schizophrenia. Serine racemase (SR) is the enzyme that converts l-serine to d-serine. In this study, we demonstrate that SR interacts with the synaptic proteins, postsynaptic density protein 95 (PSD-95) and stargazin, forming a ternary complex. SR binds to the PDZ3 domain of PSD-95 through the PDZ domain ligand at its C terminus. SR also binds to the C terminus of stargazin, which facilitates the cell membrane localization of SR and inhibits its activity. AMPA receptor activation internalizes SR and disrupts its interaction with stargazin, therefore derepressing SR activity, leading to more d-serine production and potentially facilitating NMDA receptor activation. These interactions regulate the enzymatic activity as well as the intracellular localization of SR, potentially coupling the activities of NMDA and AMPA receptors. This shuttling of a neurotransmitter synthesizing enzyme between two receptors appears to be a novel mode of synaptic regulation.

Glutamate, the principal excitatory neurotransmitter in mammalian brain, acts via metabotropic and ionotropic glutamate receptors, of which the two best characterized are the NMDA receptor (NMDAR) 2 and AMPA receptor (AMPAR). Typically, the sodium/calcium channels of NMDARs are opened only after neurons are depolarized by AMPA receptor activation, which relieves a magnesium-mediated block of NMDA channels.
NMDARs are unique in that they require co-activation by glutamate and another agent, first identified as glycine (1). D-Serine has only been recently appreciated as a major neurotransmitter/neuromodulator which, similar to glycine, coactivates NMDARs (2)(3)(4)(5). Evidence for its physiologic mediation of NMDA neurotransmission includes the much greater co-localization of D-serine than glycine with NMDARs (3) and the profound diminution of NMDA transmission following selective degradation of D-serine (3)(4)(5)(6). D-Serine is generated by serine racemase (SR), which converts L-to D-serine (7). The enzymatic activity of SR is enhanced by binding to glutamate receptor-interacting protein (8) and PICK1 (protein interacting with C kinase 1) (9, 10), which are proteins associated with AMPA receptors. By contrast, SR is inhibited by binding to the phospholipid phosphatidylinositol 4,5-bisphosphate (11) and S-nitrosylation that occurs following activation of NMDARs (12).
NMDA and AMPA receptors are regulated by a variety of accessory proteins. Stargazin is one of the best characterized AMPA receptor accessory proteins (13). Stargazin facilitates AMPA receptor cell surface expression, synaptic clustering, and recycling and modulates its desensitization and deactivation (14 -17). Postsynaptic density protein 95 (PSD-95) is one of the most abundant proteins of postsynaptic densities, binding to multiple proteins both in AMPA and NMDAR complexes (18). By binding directly to NMDARs and to neuronal nitric oxide synthase, PSD-95 serves as a signaling scaffold, mediating the activation of neuronal nitric oxide synthase by calcium-calmodulin concommitant with the entry of calcium via NMDAR channels (19). Also by anchoring AMPA receptors at postsynaptic densities, PSD-95 interfaces between AMPA and NMDA receptors (18).
In the present study, we report that SR binds to stargazin. We also identify binding of SR to PSD-95 and provide evidence for a ternary complex of SR with stargazin and PSD-95. We elucidate how these binding interactions influence SR function and, presumably, glutamate neurotransmission. These findings imply a role for SR in cross-talk between AMPA and NMDA neurotransmission.

EXPERIMENTAL PROCEDURES
Animal Husbandry-Mice containing targeted mutations of SR have been described previously (6). Both male and female mice were used, and all studies were conducted on matched littermates. Experiments were performed in accordance with protocols approved by the animal care and use committee at The Johns Hopkins University.
Reagents-HA mouse monoclonal antibody was purchased from Covance. Myc mouse monoclonal antibody was from Roche Life Science. GFP rabbit antibody was from Abcam. Transferrin mouse monoclonal antibody was from Invitrogen. Rabbit polyclonal lactate dehydrogenase antibody was from Santa Cruz Biotechnology. Mouse SR antibody used for immunoprecipitation was from BD Biosciences. SR rabbit antiserum used for Western blotting was made in-house and has been described previously (2). Stargazin rabbit antibody used for Western blotting was from Millipore (AB9876), and the one used for immunoprecipitation was a generous gift from Dr. Richard Huganir of The Johns Hopkins University. Synapseassociated protein 102 (SAP102) rabbit antibody, GluR1 rabbit antibody, and the pRK5-GFP-GluN2B construct were also gifts from Dr. Huganir. PSD-95 rabbit monoclonal antibody (D27E11) was obtained from Cell Signaling. Sulfo-NHS-biotin and NMDA were purchased from Tocris Bioscience. Tubulin-HRP antibody was from Abcam. Anti-mouse IgG and anti-rabbit HRP-conjugated secondary antibodies were from GE Healthcare. Neutravidin beads were purchased from Thermo Scientific. Complete protease inhibitor mixture tablets were from Roche Life Science. The GFP-tagged truncated PSD-95 plasmid constructs except GFP-PSD-95-PDZ3 were kind gifts from Dr. Katherine Roche from National Institute of Neurological Disorders and Stroke, National Institutes of Health.
Human Proteome Microarray-We utilized a human proteome microarray constructed from a library of 16,368 unique full-length human ORFs, as described previously (20). They were expressed as N-terminal GST-RGS(His) 6 fusion proteins and purified from yeast. Each ORF is printed in duplicate on the microarrays. Two preparations of SR (independently purified in the laboratory of S. H. S. and H. W.) and D-serine deaminase (DsdA) protein were labeled with Alexa Fluor 647 C2-maleimide (Invitrogen) and used as bait proteins. The microarrays were blocked with SuperBlock Blocking Buffers (Thermo Scientific) supplemented with 3% (w/v) BSA, 100 mM NaCl, and 0.05% (v/v) Tween 20 for 1 h at room temperature and then independently hybridized to the bait proteins for 1 h at room temperature. The microarrays were subsequently washed three times in the TBST buffer and imaged using GenePix 4000B scanner at 635 nm. The amount of each protein immobilized to the microarray was determined by staining with an Alexa Fluor 555-labeled GST antibody and imaged at 532 nm. The intensity of each spot on the microarray was quantified and ranked. Positive candidates were defined as those ranked in the top 300 in both preparations of SR but not within the top 800 in the DsdA control microarray.
Immunoprecipitation and Western Blotting-Immunoprecipitation from cells was carried out 48 h after transfection with the constructs specified. The cells were harvested in lysis buffer (50 mM Tris⅐HCl, pH 7.8, 150 mM NaCl, 1% (v/v) Triton X-100, 1 mM EDTA, 1 mM PMSF, 10% (v/v) glycerol, and protease inhibitor tablet). Alternatively, mouse brains were quickly removed, and frontal cortex was isolated on ice-cold phosphate-buffered saline and subsequently homogenized using an overhead stirrer in the aforementioned lysis buffer. Lysates were allowed to rock at 4°C for 30 min. After centrifugation at 16,000 ϫ g at 4°C for 15 min, the supernatant was harvested. Some immunoprecipitations were replicated by centrifuging at 100,000 ϫ g at 4°C for 40 min with essentially the same results (data not shown). After the protein concentration was determined using the BCA assay, supernatant containing 50 g (overexpression) or 80 g (endogenous) of protein was saved as inputs (10%). Primary antibodies (ϳ2 g) were added to the supernatant containing 500 g (overexpression) or 800 g (endogenous) of protein and incubated at 4°C overnight. The next day, EZView red protein G or protein A affinity gel (Sigma-Aldrich) were added to the mixture for 2 h at 4°C and washed with washing buffer (lysis buffer with 250 -400 mM NaCl) three times. Bound proteins were analyzed by Western blotting. The optical density (O.D.) of protein bands on each digitized image was normalized to the O.D. of the loading control (␤-tubulin, 1:10,000). Densitometry was done using ImageJ software. Normalized values were used for analyses.
Sequential Immunoprecipitation-SH-SY5Y cells were lysed in lysis buffer (50 mM Tris⅐HCl, pH 7.4, 150 mM NaCl, 0.8% (v/v) Nonidet P-40, 10% (v/v) glycerol, and protease inhibitor tablet) 48 h post-transfection. Lysates were cleared and 30 l of a prewashed 1:1 slurry of anti-HA affinity gel (Sigma) was incubated with 600 g of whole-cell lysate overnight at 4°C. Next day, the anti-HA affinity gel complexes were washed three times with washing buffer (50 mM Tris⅐HCl, pH 7.4, 150 mM NaCl, 0.1% (v/v) Nonidet P-40, and protease inhibitor tablet). The last wash was done without the protease inhibitors. The affinity gel was eluted by HA peptide solution (final concentration of 200 g/ml) three times at 4°C 30 min each, and the eluents were pooled (ϳ90 l of total volume). The anti-stargazin antibody (1.2 g, gift from Dr. Richard Huganir) or anti-GFP antibody (1 g) and EZView Protein A affinity gel were incubated with the eluent overnight. The Protein A affinity gel complexes were washed three times with washing buffer and eluted in 2ϫ SDS loading buffer, followed by SDS-PAGE and immunoblotting.
Subcellular Fractionation-HEK-293 cells or primary cortical neuronal cultures were resuspended in 350 l/10 cm dish of fractionation buffer containing 250 mM sucrose, 20 mM HEPES, pH 7.4, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA, 1 mM DTT, and protease inhibitors. The lysates were passed through a 26G needle eight times and then left on ice for 20 min. The lysates were then centrifuged for 10 min at 1,000 ϫ g to pellet the nuclear fractions. The supernatant was recovered and centrifuged at 120,000 ϫ g for 40 min, and the supernatant was kept as the cytosolic fraction. The pellet was dissolved in the fractionation buffer and centrifuged again at 120,000 ϫ g for 40 min. The pellet was suspended in the fractionation buffer supplemented by 0.1% (v/v) SDS and stored at Ϫ80°C as the membrane fraction.
Cell Surface Protein Biotinylation Assay-Cells grown in 60 mm dishes were rinsed twice with ice cold PBS. Then 2.5 ml of freshly made sulfo-NHS-biotin solution was added to each dish and incubated on ice for 15 min in the dark. The cells were then gently washed with ice cold PBS, followed by a wash with PBS supplemented with 50 mM glycine to quench any unreacted biotin, and two washes with PBS.
The cells were then lysed in the lysis buffer and centrifuged at 16,000 ϫ g at 4°C for 15 min. Cleared lysates containing 500 g of protein were incubated with neutravidin beads overnight at 4°C. The next day, the beads were washed 5 times with the lysis buffer. Subsequently, proteins bound to the beads were eluted with elution buffer (1% (v/v) 2-Mercaptoethanol in PBS) at 37°C for 25 min rocking at 900 rpm. The eluents were then subjected to Western blotting.
SR Activity Assay-The SR activity assay was performed as previously described (22,23) with modifications. In short, 24 h after transfection, the cell culture medium was replaced with fresh medium containing 8 mM L-serine. After 24 h, the medium was harvested and spun down at 16,000 ϫ g for 10 min, and the supernatant was stored at Ϫ80°C. The level of D-serine was measured by HPLC as described previously (24). The amount of contaminating D-serine in the commercial L-serine was determined and subtracted. To determine the specific activity of SR, D-serine levels in the media were normalized by the amount of SR expressed (determined by the O.D. from Western blotting analysis).
Immunocytochemistry-For immunofluorescence, rat primary cortical neurons were cultured in 35-mm glass-bottomed culture dishes (MatTek). The cells were not treated, or treated with NMDA at 40 M for 4 min or at 100 M for 40 min or 100 M AMPA for 10 or 40 min. Subsequently, the cells were washed with PBS twice and fixed in 4% (w/v) paraformaldehyde in PBS for 20 min. After washing in PBS for four times, the cells were blocked with 7% (v/v) goat serum with 0.1% (v/v) Triton X-100 in PBS for 1 h at room temperature. The cells were incubated with primary antibodies (rabbit anti-SR serum, 1:1000; rabbit anti-PSD-95 1:250; rabbit anti-stargazin, 1:100) overnight at 4°C. The cells were then incubated with Alexa Fluor 488-or 568-labeled species-specific goat secondary antibodies (Invitrogen) diluted at 1:600 for 1.5 h at room temperature.
Images were taken with a Zeiss LSM 510 confocal laser scanning microscope at The Johns Hopkins University Neuroscience Multiphoton/Electrophysiology Core Facility. Co-localization of two fluorescent channels was quantified by the Pearson correlation coefficient using the Velocity software (PerkinElmer Life Science).
Cell Viability Assay-HEK-293 cells were plated in 12-well plates for viability assays. The cells were transfected with 1.4 g of SR and/or 1.0 or 1.4 g of stargazin. Thirty-six hours after transfection, cell viability was determined via MTT assay as described previously (25).
Statistical Analysis-All results are expressed as the mean Ϯ S.E. and were analyzed by the Student's two-tailed paired t test. Data for each lane in the quantification were derived from at least three independent experiments. p values were calculated using the GraphPad Prism software (GraphPad Software, Inc.).

SR Binds PSD-95 and Associated
Proteins-Appreciation of D-serine, a product of SR, as an endogenous co-agonist of the NMDAR, places SR at a pivotal position in glutamate transmission. However, as it is a relatively recently identified enzyme, there have been few studies of its interactions with synaptic proteins. We wondered whether SR links with synaptic proteins other than the already identified PICK1 (9) and glutamate receptor-interacting protein (8). Accordingly, we screened a library of 16,368 GST-tagged human proteins purified from yeast (20). After screening for proteins that robustly bind to two SR preparations from two independent laboratories, but not to the DsdA control, we identified only two candidates, SAP102 and Golgi reassembly-stacking protein 2 (GORASP2) (Fig. 1A). Co-immunoprecipitation studies confirmed the binding to both mouse and human forms of SAP102, but the interaction with GORASP2 appears to be detectable only for its mouse orthologs (Fig. 1B). Considering the role of D-serine as a neurotransmitter, we decided to focus on SAP102. SAP102 belongs to the family of MAGUK (membrane-associated guanylate kinase) proteins, of which the best characterized is PSD-95, a protein that binds both NMDA and AMPA receptor complexes (18). We observed robust co-immunoprecipitation between SR and PSD-95 as well as SAP102, which is suggestive of binding, whereas no binding is evident for SAP-97 (Fig. 1C). We also detect binding of endogenous SR with SAP102 and PSD-95 by immunoprecipitation experiments with mouse brain extracts and primary cortical neurons (Fig. 1, D and E). We mapped binding sites for PSD-95 and SR. PSD-95 possesses 3 PDZ domains that mediate binding to diverse proteins (26). Our mapping experiments reveal that PSD-95 interacts with SR via its PDZ3 domain (Fig. 1, F and G). Accordingly, overexpression of the PDZ3 domain of PSD-95 disrupts the interaction between SR and PSD-95 (Fig. 1H). Consensus sequences enabling proteins to bind PDZ domains of other proteins involve the last several amino acids at the C terminus (27). Deletion of the C-terminal Thr-Val-Ser-Val (TVSV) of SR abolishes its binding to PSD-95, indicating that this typical "PDZ ligand" domain of SR is responsible for interactions with PSD-95 (Fig.  1I). We also demonstrate co-localization of PSD-95 and SR in the soma and dendrites of primary cortical neurons in culture preparations (Fig. 1J). We used Pearson correlation coefficient (PCC) to quantify co-localization because it is independent of signal intensity levels and signal offset (background) and therefore is relatively unbiased (28). SR/PSD-95 has a PCC of 0.68 Ϯ 0.01.
Binding of SR to Stargazin in a Complex with PSD-95-SR has been previously shown to bind PICK1 and glutamate receptorinteracting protein, well characterized members of the AMPA receptor-associated protein complex (8,9). Stargazin is a more recently identified and prominent accessory protein for AMPA receptors (14). Accordingly, we examined its possible interactions with SR. Utilizing overexpressed proteins in HEK-293 cells, we observe robust co-precipitation of stargazin with SR ( Fig. 2A). Endogenous stargazin also co-immunoprecipitates with SR in mouse brain extracts (Fig. 2B). Presumably SR participates in AMPA receptor complexes that include stargazin. We wondered whether, additionally, SR might bind directly to AMPA receptors. However, SR failed to co-precipitate with the AMPA receptor subunit GluR1 (Fig. 2C). We also mapped the segment that is critical for stargazin binding to the N-terminal first 66 amino acids of SR (Fig. 2, D and E). It is of note that FIGURE 1. SR interacts with PSD-95 and associated proteins. A, SAP102 and GORASP2 were identified as possible interacting partners of SR. SR prepared from two independent laboratories, and DsdA were labeled with Alexa Fluor 647 C2-maleimide (C2-M) and hybridized to a chip harboring a library of 16,368 GST-tagged human proteins. SAP102 and GORASP2 bind to both preparations of SR but not the DsdA control. Each protein in the library was printed in duplicate on the chip. B, co-immunoprecipitation of SAP102 and GORASP2 with SR in HEK-293 cells. Note that both the mouse and human forms of SAP102 co-precipitate with the mouse and human forms of SR, respectively. Mouse GORASP2 co-precipitates with mouse SR, but no co-precipitation is observed between their human orthologs. An immunoblot obtained with lower exposure (low exp.) is also shown here. mSR, mouse SR; hSR, human SR; mSAP102, mouse SAP102; hSAP102, human SAP102; mGORASP2, mouse GORASP2; hGORASP2, human GORASP2. C, co-immunoprecipitation of SR and MAGUK family proteins in HEK-293 cells. Mouse SAP102 and PSD-95, but not SAP97, co-precipitate with SR. D, immunoprecipitation of endogenous SR and SAP102 in mouse brain. Serine racemase knock-out mouse was utilized as a negative control. E, endogenous binding of SR and PSD-95 in primary cortical neuronal cultures (days in vitro 17). stargazin was not identified as one of the positive interacting partners in the initial protein microarray screening. Stargazin only gave modest specific signals in both preparations of SR and none in the DsdA control. This could be due to the relatively weak binding affinity between SR and stargazin, suboptimal binding conditions or impaired binding due to the bulky GST tag fused to stargazin.
It is well established that the binding between stargazin and PSD-95 is inhibited by phosphorylation of stargazin at Thr-321 (29). We wondered whether SR-stargazin binding is also influenced by this phosphorylation event. Stargazin-T321A binds to SR as well as does wild-type stargazin. By contrast, mutation of stargazin-Thr-321 to aspartate or glutamate, mimicking phosphorylation, abolishes binding to SR (Fig. 2F). Thus, similar to the stargazin-PSD-95 interaction, stargazin-SR binding appears likely to be inhibited by stargazin phosphorylation.
To determine whether SR, stargazin and PSD-95 could form a ternary complex, we performed sequential immunoprecipitation experiments (Fig. 3, A and B). Neuroblastoma cell line SH-SY5Y, which expresses endogenous PSD-95, was transfected with HA-tagged SR and Myc-tagged stargazin. We isolated HA-SR immune complexes and then eluted these complexes with the HA peptide, isolated stargazin immune com-plexes, and confirmed that endogenous PSD-95 was present in these complexes (Fig. 3B). Similarly, we detected endogenous PSD-95 when the cells were transfected with HA-tagged SR and GFP-tagged GluN2B and HA and GFP immune complexes were isolated in order (Fig. 3C) (23, 30 -37). These observations suggest that SR, stargazin, and PSD-95 are present in the same complex: the N-terminal portion of SR binding stargazin, whereas its C-terminal TVSV sequence mediates binding to PDZ3 domain of PSD-95 (Fig. 3D). In contrast, stargazin binds to PDZ1 and PDZ2 of PSD-95 (38). Moreover, SR, PSD-95 and the GluN2B subunit of the NMDAR could also form a ternary complex (Fig. 3D). In addition to the SR-PSD-95 interaction described here, it is known that the second PDZ domain in PSD-95 binds to the C-terminal domain of GluN2 subunits of the NMDAR (39).
Binding to Stargazin Inhibits SR Catalytic Activity and Facilitates Its Surface Expression-We investigated consequences of the binding between SR and stargazin. We explored the influence of stargazin on SR catalytic activity. Stargazin reduces the activity of co-expressed SR ϳ35%, indicated by the amount of D-serine produced (Fig. 4, A-C). In contrast, SAP102 and PSD-95 do not influence SR activity. Co-expressing SAP102 or PSD95 also does not significantly alter the influence of star-FIGURE 2. SR binds to stargazin. A, co-immunoprecipitation of SR and stargazin in HEK-293 cells. The ϳ41-kDa top band denoted by the arrowhead is the stargazin immunoreactive band. B, co-immunoprecipitation of endogenous SR and stargazin in mouse brain. SR Ϫ/Ϫ mice were used as negative controls. C, SR does not co-immunoprecipitate with GluR1 when overexpressed in HEK-293 cells. D, co-immunoprecipitation of truncated and point mutants of SR with stargazin to map the residues of SR that are critical for binding to stargazin. The top band denoted by the arrowhead is the specific stargazin band. stg, stargazin. SR V339G was employed here as this mutation in the PDZ domain ligand sequence reduces the binding between SR and PSD-95 (data not shown). E, schematic showing that the N-terminal 1-66 residues of SR are responsible for binding to stargazin. F, mutating Thr-321 of stargazin to Asp or Glu, which mimics the phosphorylated state, reduces the binding between SR and stargazin in HEK-293 cells. Immunoblots in A-F are representative of at least n ϭ 3 independent immunoprecipitations from n ϭ 3 independent transfections (where applicable). mSR, mouse SR; WB, Western blot. gazin upon SR activity (Fig. 4, A and C). These actions are not attributable to cytotoxicity, as cell viability is not influenced by the expression of stargazin (Fig. 4D).
One of the most prominent roles of stargazin is to facilitate the membrane expression of the AMPA receptor (40). Because stargazin binds SR, we wondered whether it could also shuttle SR to the proximity of the membrane. Indeed, overexpressing stargazin doubled surface levels of SR monitored both by cell surface biotinylation (Fig. 5A) and subcellular fractionation assays (Fig. 5B). Although SR is cytosolic, it is associated with the transmembrane protein stargazin, which is directly labeled by biotin. Accordingly, precipitation of stargazin leads to coprecipitation of serine racemase.
We wondered whether the SR-stargazin-PSD-95 complex is dynamically regulated and coupled to glutamate neurotransmission. We explored this possibility by monitoring the level of cell surface SR upon receptor activation (Fig. 5C). AMPA exposure markedly reduces surface levels of SR and diminishes GluR1 membrane levels, consistent with the well known internalization of GluR1 elicited by AMPA activation (41). By contrast, NMDA caused an increase in membrane levels of SR. The effect of AMPA on cell surface SR is abolished when co-treated with 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide, an AMPAR antagonist (Fig. 5D). We wondered whether dissociation of SR from membranes by AMPA activation reflects influences of AMPA signaling upon SR-stargazin binding. We demonstrate that AMPA exposure dissociates SR from stargazin (Fig. 5E). In contrast, dihydroxyphenylglycine, a metabotropic glutamate receptor agonist, fails to alter stargazin-SR binding. These findings imply that stargazin mediates SR association with membrane fractions. Consistent with these findings, AMPA treatment diminishes the co-localization of FIGURE 3. SR and PSD-95 form a ternary complex with stargazin and the NMDA receptor. Schematic (A) and results (B) of the sequential immunoprecipitation using neuroblastoma cell line SH-SY5Y expressing HA-tagged SR and Myc-tagged stargazin. SR immune complexes were isolated and then eluted with the HA peptide. Stargazin immune complexes were then purified, and endogenous PSD-95 was detected by immunoblotting. The top band denoted by the arrowhead is the specific stargazin band. C, SR, PSD-95, and the GluN2B subunit of the NMDA receptor also form a ternary complex in SH-SY5Y cells as determined by sequential co-immunoprecipitation experiments. SH-SY5Y cells were transfected with HA-tagged SR and GFP-tagged GluN2B. SR immune complexes were isolated and then eluted with the HA peptide. GFP-GluN2B immune complexes were then purified, and endogenous PSD-95 was detected by immunoblotting. Immunoblots in B and C are representative of at least n ϭ 3 independent immunoprecipitations from n ϭ 3 independent transfections. It is of note that only GFP-tagged GluN2B subunit of the NMDAR was transfected. It has been reported previously that undifferentiated SH-SY5Y cell lines express substantial amounts of endogenous GluN1 subunits (30,31) as well as functional NMDA receptors (23,(32)(33)(34)(35). Therefore, GFP-GluN2B will assemble with other subunits and be present at the cell surface, rather than being retained in the endoplasmic reticulum as when expressed alone (36,37). D, schematic showing that SR, stargazin and PSD-95 form a ternary complex closely associated with AMPARs and NMDARs. It is known that the second PDZ domain in PSD-95 binds to the C-terminal domain of GluN2 subunits of the NMDAR (39). See text for details. mSR, mouse SR; WB, Western blot; WCL, whole cell lysate; N, N terminus; C, C terminus. Twenty-four hours after transfection, the cell culture medium was supplemented with 8 mM L-serine. After overnight incubation, the cell culture medium was harvested, and HPLC was performed to determine the level of D-serine. B, SR protein level in each sample was determined by Western blotting and the ratio of optical density of SR/␤-tubulin was used to correct for the expression of SR when calculating specific enzymatic activities of SR. Levels of apparent D-serine in non-transfected controls are 18.0 Ϯ 0.78% of the amount of D-serine in transfected preparations. These low levels of background D-serine reflect the ϳ1% known contamination of D-serine in commercial L-serine sources. C, quantification of the effect of stargazin, PSD-95, and SAP102 on the enzymatic activity of SR. Stargazin co-expression causes a 35% decrease in the activity of SR. K56G is a catalytically inactive form of SR. PSD-95 and SAP102 alone did not alter the activity of SR. They also did not significantly augment the effect of stargazin on SR when co-expressed. The number of independent experiments and HPLC measurements for each lane is 5, 5, 3, 3, 3, 3, and 3, respectively. AU, arbitrary unit. ***, p Ͻ 0.0001. NS, not significant. D, stargazin co-expression does not affect cell viability as determined by the MTT assay. In the SR ϩ stargazin lane, 1.0 g of stargazin DNA was transfected into HEK-293 cells in 60-mm dishes. One g and 1.4 g of stargazin DNA were transfected in the two stargazin lanes, respectively. Each lane in the quantification represents independent MTT measurements from n ϭ 3 independent transfections. mSR, mouse SR; WB, Western blot. OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 stargazin and SR in primary cortical cultures as the PCC decreases from 0.49 Ϯ 0.02 to 0.36 Ϯ 0.02 (Fig. 5F).

DISCUSSION
In the present study, we have demonstrated physiologic binding interactions of SR with PSD-95 as well as stargazin. There appears to be a quinary complex consisting of AMPA receptors, stargazin, SR, PSD-95, and the NMDA receptors. In this complex, the C-terminal Thr-Pro-Val of stargazin binds to PDZ domains 1 and 2 of PSD-95 (38). By contrast, SR binds to the PDZ3 domain of PSD-95. In this complex, stargazin inhibits SR catalytic activity. Activation of AMPA receptors appears to dissociate SR from its association in membranes with stargazin leading to enhanced SR catalytic activity (Fig. 6). This model explains, at least in part, the known cross-talk between AMPA and NMDA neurotransmission wherein AMPA receptor activation augments NMDA transmission. Under resting conditions, SR activity is inhibited as a consequence of its binding in a complex with stargazin, PSD-95, and AMPA receptors. AMPA receptor activation dissociates the complex freeing up SR to generate D-serine. Synaptic D-serine, together with gluta-mate, activates NMDARs. This model assumes proximity of AMPA and NMDA receptors, which has been well demonstrated (26).
Initial interest in NMDAR/AMPAR cross-talk derived from the discovery of "silent synapses" wherein a proportion of excitatory synapses in the hippocampus are functionally silent at resting membrane potentials because they contain NMDARs but not AMPARs (42,43). In this model, long term potentiation is dependent on the emergence of AMPARs at those synapses. The kinetics of AMPA receptor activation are much more rapid than those of NMDARs. One mechanism functionally linking the two receptors involves depolarization elicited by AMPA activation triggering opening of NMDA channels. Our findings provide an additional model wherein SR mediates cross-talk between the two receptors. This shuttling of a neurotransmitter synthesizing enzyme between two receptors appears to be a novel mode of synaptic regulation. AMPA activated D-serine synthesis proposed here is also consistent with findings that kainate stimulation of AMPA receptors potentiates NMDAR channel activities (44).
NMDA and AMPA receptors can be linked via synaptic scaling wherein rapid and large increases in AMPA transmission lead to augmented NMDA signaling (45). Association of these receptors via stargazin and SR provides a molecular mechanism to mediate such synaptic scaling. Synaptic scaling mediates long term potentiation wherein marked increases in AMPA transmission are followed by proportional potentiation of NMDA signaling (46), which may also involve the SR mechanisms reported here.
Spinal hyperexcitability and persistent pain have been associated with NMDA/AMPA transmission (47)(48)(49). Thus, blockade of NMDA and AMPA receptors is antinociceptive (50,51). Stargazin may play a role in these processes, as susceptibility to chronic pain following nerve injury is genetically impacted by stargazin (52). Moreover, stargazin polymorphisms are associated with chronic pain in various populations of cancer patients (52). The stargazin-SR dynamics described here might underlie the AMPA/NMDA receptor cross-talk that mediates spinal mechanisms for chronic pain processing. FIGURE 5. Stargazin facilitates SR cell surface expression and dissociates from SR upon AMPAR activation. A, co-expression of stargazin increases the amount of SR in close proximity to the cell membrane, as determined by the cell surface biotinylation assay. HA-mouse SR (mSR) with or without stargazin were transfected into HEK-293 cells. Lactate dehydrogenase (LDH), heat shock protein 90 (Hsp90), and ␤-tubulin are cytoplasmic proteins and serve as negative controls to assure that the NHS-sulfo-biotin reagent for surface labeling did not penetrate into the cell. The transferrin receptor (TfR), a membrane protein, is used as a positive control for surface labeling. Each lane in the quantification represents n ϭ 4 independent surface expression assays from n ϭ 4 independent transfections. The amount of co-precipitated SR in the SR-only lane was normalized to 1. Representative immunoblots are shown. WCL, whole cell lysate. B, subcellular fractionation shows that the level of SR increases in the membrane fraction when co-expressed with stargazin in HEK-293 cells. C, cytosol; M, membrane. Each lane in the quantification represents n ϭ 4 independent fractionation experiments from n ϭ 4 independent transfections. The amount of membrane SR in the SR-only lane was normalized to 1. Representative immunoblots are shown. C, SR in the membrane fraction decreases when the primary cortical neuronal culture is stimulated with AMPA. NMDA treatment increases the membrane fraction of SR. No significant change in the cytosolic pool of SR was observed upon AMPA or NMDA treatment. LDH and transferrin receptor are used as cytosol and membrane fraction markers, respectively. Each lane in the quantification represents n ϭ 4 independent treatments from n ϭ 4 independent cultures. Representative immunoblots are shown. D, treating the primary neuronal culture with 30 M 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (NBQX) abolishes the effect of AMPA on the membrane fraction of SR. Cells were pretreated with 30 M 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide for 4 min, and then AMPA was added to the final concentration of 100 M for 10 min before lysing. Each lane in the quantification represents n ϭ 3 independent treatment from n ϭ 3 independent cultures. The levels of membrane SR in the non-treated cultures were normalized to 1. Representative immunoblots are shown. E, SR dissociates from stargazin upon AMPA stimulation (100 M, 10 min). Neither NMDA (40 M, 4 min) nor dihydroxyphenylglycine (DHPG; 100 M, 10 min) has any effect on SR-stargazin binding. The top band denoted by the arrowhead is the specific SR band. No treatment (no trmt), NMDA, and dihydroxyphenylglycine-treated lanes in the quantification represent n ϭ 3 independent treatment from n ϭ 3 independent cultures. AMPA-treated lanes represents n ϭ 5 independent treatment from n ϭ 5 independent cultures. The amount of co-precipitated SR in the non-treated cultures was normalized to 1. Representative immunoblots are shown. F, immunocytochemical staining of SR (green) and stargazin (red) in the primary cortical neuronal cultures from rat. Top two panels, untreated cells; bottom two panels, cells treated with 100 M AMPA for 10 min. PCC was used to quantify the co-localization. PCC was calculated from 20 independent microscope fields from three independent batches of neuronal cultures for each group. The PCC values before and after AMPA treatment were 0.49 Ϯ 0.02 and 0.36 Ϯ 0.02, respectively. Scale bar, 5 m. *, p Ͻ 0.05; ***, p Ͻ 0.0001. AU, arbitrary unit; WB, Western blot.

Serine Racemase Regulated by Binding to Stargazin and PSD-95
OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 In summary, our study establishes that the enzyme SR mediates cross-talk between AMPA and NMDARs that involves the synaptic proteins stargazin and PSD95. Regulation of such synaptic cross-talk by translocation of an enzyme involved in neurotransmitter biosynthesis affords a novel mode of synaptic signaling. Such synaptic regulation might take place at other sites where synapses employ families of such proteins and wherein synaptic cross-talk operates on a similar time scale.