Serine Racemase Modulates Intracellular D-Serine Levels through an α,β-Elimination Activity*

Mammalian brain contains high levels of d-serine, an endogenous co-agonist of N-methyl d-aspartate type of glutamate receptors. d-Serine is synthesized by serine racemase, a brain enriched enzyme converting l- to d-serine. Degradation of d-serine is achieved by d-amino acid oxidase, but this enzyme is not present in forebrain areas that are highly enriched in d-serine. We now report that serine racemase catalyzes the degradation of cellular d-serine itself, through the α,β-elimination of water. The enzyme also catalyzes water α,β-elimination with l-serine and l-threonine. α,β-Elimination with these substrates is observed both in vitro and in vivo. To investigate further the role of α,β-elimination in regulating cellular d-serine, we generated a serine racemase mutant displaying selective impairment of α,β-elimination activity (Q155D). Levels of d-serine synthesized by the Q155D mutant are several-fold higher than the wild-type both in vitro and in vivo. This suggests that the α,β-elimination reaction limits the achievable d-serine concentration in vivo. Additional mutants in vicinal residues (H152S, P153S, and N154F) similarly altered the partition between the α,β-elimination and racemization reactions. α,β-Elimination also competes with the reverse serine racemase reaction in vivo. Although the formation of l- from d-serine is readily detected in Q155D mutant-expressing cells incubated with physiological d-serine concentrations, reversal with wild-type serine racemase-expressing cells required much higher d-serine concentration. We propose that α,β-elimination provides a novel mechanism for regulating intracellular d-serine levels, especially in brain areas that do not possess d-amino acid oxidase activity. Extracellular d-serine is more stable toward α,β-elimination, likely due to physical separation from serine racemase and its elimination activity.

D-Serine is a D-amino acid that occurs at high levels in the mammalian brain and is an endogenous ligand of the "glycine site" of N-methyl D-aspartate (NMDA) 1 receptors (1)(2)(3)(4). NMDA receptors play key roles in excitatory synaptic transmission, plasticity, and learning and memory (5). Overactivation of the NMDA receptor and the resultant influx of calcium into cells is a major culprit in the cell death that occurs following stroke and neurodegenerative diseases. Blockers of the "glycine site" of the receptor are neuroprotective in animal models of stroke (5).
Endogenous D-serine is required for NMDA receptor activation, and its removal markedly decreases NMDA receptor activity (3). In the vertebrate retina, endogenous D-serine may also mediate the light-dependent increase in neuronal activity by activating NMDA receptors (6). More recently, D-serine was suggested to play a role in the long term potentiation of synaptic transmission in the hippocampus, indicating a role of endogenous D-serine in long term synaptic plasticity (7). D-Serine is synthesized by serine racemase, a pyridoxal phosphate (PLP)-dependent enzyme enriched in the mammalian brain (8,9). Serine racemase has high sequence homology with the fold-type II group of PLP enzymes, such as serine/threonine dehydratase and D-serine dehydratase (10,11). In addition to converting L-to D-serine, serine racemase catalyzes the ␣,␤elimination of water from L-serine to form pyruvate and ammonia (12). The initial rates of racemization and ␣,␤-elimination of L-serine by serine racemase are strongly stimulated by magnesium and ATP, indicating that the complex Mg⅐ATP is a physiological ligand of the enzyme (12).
In accordance with accepted mechanisms of PLP-catalyzed reactions (13)(14)(15)(16), a mechanism for racemization and ␣,␤-elimination catalyzed by serine racemase is depicted in Scheme 1. PLP, bound to the enzyme through an internal aldimine with Lys 56 , reacts with L-serine to give an external aldimine intermediate. Subsequent ␣-proton abstraction forms a resonance stabilized carbanion. Reprotonation of this intermediate on the opposite face of the planar carbanion generates the D-serine external aldimine intermediate. D-Serine is released via transimination with Lys 56 , regenerating the free enzyme. The resonance stabilized carbanion is also an intermediate for the ␣,␤-elimination reaction. Elimination of the ␤-hydroxyl group from the carbanionic intermediate leads to the formation of the aminoacrylate-PLP intermediate. Subsequent transimination releases the initial aminoacrylate product and regenerates free enzyme. The aminoacrylate released undergoes rapid non-enzymatic hydrolysis to give pyruvate and ammonia.
The termination of signaling by a neurotransmitter in the brain normally requires its re-uptake and metabolism. D-Serine signaling is thought to involve its release from cells to stimulate NMDA receptors in neurons, but its re-uptake and degradation are not well understood. D-Amino acid oxidase is the only mammalian enzyme known to degrade D-serine. It is restricted to the cerebellum and brainstem, with negligible activity in the cerebral cortex and other forebrain areas where serine racemase is present (2). Thus, the pathways for D-serine removal and degradation are still unknown and are very important to delineate to have a full understanding of D-serine signaling.
A relationship between the serine racemase ␣,␤-elimination reaction and the synthesis of D-serine has not been previously addressed. Moreover, ␣,␤-elimination was proposed to be specific for L-serine. D-serine was thought not to be a physiological ␣,␤-elimination substrate (17). Because ␣,␤-elimination is a degradative process, we considered the possibility that ␣,␤elimination may be a mechanism for down-regulation of D-serine levels. Thus, we have now explored further the physiological role and specificity of the serine racemase ␣,␤elimination activity.
It was found that, in addition to L-serine, serine racemase catalyzes the ␣,␤-elimination of water from L-threonine and D-serine in an ATP-modulated manner. ␣,␤-Elimination of Dserine occurs through an apparent futile reaction in which part of D-serine synthesized by serine racemase is converted to pyruvate both in vitro and in vivo. This suggests a novel mechanism to limit intracellular D-serine concentrations: serine racemase may degrade excess D-serine in areas lacking D-amino acid oxidase. Additionally, we propose that the export of D-serine to the extracellular milieu increases its stability by preventing ␣,␤-elimination by serine racemase.
Cell Culture and Transfection-HEK 293 cells were cultured in DMEM containing 10% fetal bovine serum and antibiotics. For transfection, cells were split into 6-well tissue culture plates (Nunc) at 70 -90% confluence. On the next day, cells were transfected with 0.4 g of mouse full-length serine racemase gene, a catalytically inactive mutant constructed by replacing Lys 56 with glycine (9) or green fluorescent protein (GFP) cloned into pRK5-KS using Polyfect reagent (Qiagen). Cells were used 24 -48 h after transfection.
Assay of Serine and Threonine in Cells-HEK 293-transfected cells were incubated in 6-well plates with 1 ml of DMEM medium. To measure L-and D-amino acid levels, a 0.1-ml aliquot of culture medium was removed, and the samples were processed for HPLC as described previously (18). Briefly, samples were first deproteinized by addition of trichloroacetic acid at 5% final concentration. The suspension was centrifuged at 20,000 ϫ g for 5 min, and the supernatant was analyzed by HPLC after removal of trichloroacetic acid by four extractions with water-saturated diethyl ether. To calculate specific synthesis of D-serine, the values of contaminating D-serine in DMEM (about 6 M contaminating D-serine) were subtracted from the values obtained after addition of 10 mM L-serine. For intracellular amino acid determination, cells were washed twice with cold PBS, followed by addition of 5% trichloroacetic acid to extract free amino acids. The suspension was centrifuged at 20,000 ϫ g for 5 min, and the supernatant was analyzed after removal of trichloroacetic acid by three extractions with watersaturated diethyl ether.
Assay of Keto Acids-To monitor pyruvate and 2-ketobutyrate, generated from L-serine and L-threonine ␣,␤-elimination, respectively, a 0.1-ml aliquot of culture medium was removed and boiled for 5 min to inactivate endogenous lactate dehydrogenase activity. The samples were centrifuged at 20,000 ϫ g for 10 min to remove precipitated protein, and the supernatant was analyzed by monitoring the decrease in NADH (0.2 mM) absorbance at 340 nm, because the keto acids are reduced by added lactate dehydrogenase (1 g/ml) (19).
Recombinant Serine Racemase-Mouse full-length serine racemase DNA was subcloned into pET 28cϩ, which encodes a 6ϫ-histidine tag and introduced into BL21 codon plus bacteria (Stratagene). Expression of serine racemase was induced by 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside (Sigma) for 12 h at 30°C. Cells were collected by centrifugation and disrupted by sonication in medium containing 20 mM Tris-HCl (pH 7.4), 15 M PLP, 10 mM imidazole, and 400 mM NaCl. After addition of 1% Triton X-100, the suspension was cleared by centrifugation (40,000 ϫ g for 15 min), and serine racemase was purified from the supernatant by binding to Talon resin (Clontech) or nickel-nitrilotriacetic acid-agarose (Qiagen) according to the manufacturers' instructions. About 1-2 mg of protein was obtained for every liter of bacterial culture. The activity could be easily checked by adding 20 l of enzyme bound to the resin to the assay medium. The enzyme bound to beads was active, and it displayed the same kinetic characteristics of the purified enzyme when eluted with imidazole and subjected to dialysis. In some experiments, recombinant serine racemase was purified from mammalian cells as described before (20), with identical results. Recombinant enzyme was of high purity as revealed by SDS-PAGE (data not shown).
Site-directed Mutagenesis-Point mutants of serine racemase were obtained by PCR with Pfu turbo polymerase (Stratagene) using complementary primers containing the desired base changes. After 22 cycles, the PCR product was digested with DPN I and transformed into XL1-Blue Escherichia coli. The mutants were verified by double-strand DNA sequence.
In Vitro Activity Assays-Reaction media contained 40 mM Tris-HCl (pH 7.4), 15 M PLP, and the different test substrates in a final reaction volume of 50 -70 l. The reaction was started by addition of recombinant SR (0.02-0.05 mg/ml final concentration) and stopped after 1-2 h SCHEME 1. Reaction mechanism for serine racemization and ␣,␤-elimination.
at 37°C by boiling for 5 min. Blanks were carried out with heatinactivated enzyme. When indicated, the free Mg 2ϩ concentration was varied by using a Mg-EDTA buffer. The concentration of EDTA was fixed at 2 mM, and the amounts of MgCl 2 needed to obtain the desired free Mg 2ϩ concentrations were calculated using Mg-EDTA and Mg⅐ATP association constants previously determined (21). Both L-and D-serine were monitored by HPLC analysis. Formation of keto acids was monitored with lactate dehydrogenase as described above. In some experiments, keto acids formed from L-serine, D-serine, or L-threonine were monitored by detection of hydrazone derivatives of 2,4-dinitrophenylhydrazine as described (22). Ammonia was monitored by the oxidation of NADPH in the presence of 2-oxoglutarate and glutamate dehydrogenase (ammonia kit, Sigma).
Immunocytochemistry for D-Serine-Intracellular D-serine levels were revealed with an antibody against glutaraldehyde-conjugated Dserine (Mobitec, Germany). Briefly, cells were fixed for 20 min with 4.5% glutaraldehyde and 0.2% sodium metabisulfite dissolved in PBS. After reduction of the free aldehyde groups with 0.5% sodium borohydride, cells were blocked and permeabilized with 5% normal goat serum, 0.2% Triton X-100 in PBS for 45 min. Overnight incubation with primary antibody (1:500 dilution) was carried out at 4°C in PBS, 5% normal goat serum, 0.1% Triton X-100, and 0.5 mM of L-serine glutaraldehyde conjugate to block immunoreactivity against L-serine. Antirabbit secondary antibody was used at 1:200 in 2% normal goat serum, 0.1% Triton X-100 in PBS. The staining was revealed by ABC Elite kit (Vector laboratories) using 3,3-diaminobenzidine tetrahydrochloride as peroxidase substrate.
Primary Astrocyte and Slice Cultures-Primary astrocyte cultures were prepared from cerebral cortex of P0 -P2 Sprague-Dawley rats as described (23). Cells were cultured in 6-well plates in BME plus 10% fetal bovine serum for 2 weeks before use. Synthesis of D-serine from the cultures was monitored by HPLC as described (12). Cortico-striatal organotypic slices were cut using a McIlwain tissue chopper and cultured in Millicel inserts (Millipore), as described (24). The slices were cultured in minimum essential medium-Hepes medium supplemented with 20% horse serum for 7-12 days prior to determination of D-serine.
Lentivirus Production-Mouse full-length serine racemase gene was sub-cloned into pTK208 lentiviral vector (a gift from Dr. T. Kafri, University of North Carolina, Chapel Hill, NC) containing CMV promoter. The virus was produced by calcium-mediated co-transfection of the lentiviral vector (5 g), packing vector pCMV-dR8.74 (3 g), and vesicular stomatitis virus coat envelope pMD2G (2 g) (a gift from Prof. D. Trono, University of Geneva) in HEK 293 cells grown in 10-mm culture dishes (25). Control virus consisted of green fluorescent protein (GFP) under the control of CMV promoter. Viral stocks were produced by concentrating the viral particles present in the culture medium by centrifugation at 120,000 ϫ g for 2 h. The pellet was suspended in a small volume of BME culture medium and stored at Ϫ70°C until use. To infect primary astrocyte cultures, viral stocks were added to cultured cells grown in 24-well plates, and the transfection efficiency was determined after 4 days with fluorescent microscopy (for GFP) or immunocytochemistry (for serine racemase). Preparations exhibiting virtually 100% transfection efficiency were employed for experiments of D-serine metabolism.

RESULTS
Serine racemase has been shown previously to possess two distinct catalytic activities: (i) the synthesis of D-serine and (ii) the ␣,␤-elimination of water from L-serine to give pyruvate and ammonia (12, 26) (Scheme 1). We now demonstrate that serine racemase utilizes other substrates for the ␣,␤elimination activity.
We examined cells transfected with serine racemase for utilization of L-threonine added to the culture medium. Transfection of serine racemase is associated with a large decrease in medium L-threonine when compared with control cells transfected with green fluorescent protein (GFP) or with a catalytically inactive mutant of the enzyme (K56G) (Fig. 1). The decrease in L-threonine levels is comparable to that observed with L-serine (Fig. 1, A and B). L-Threonine disappearance is accompanied by increases in keto acid concentration in the culture medium ( Fig. 1, C and D). In contrast to the production of D-serine by racemization of L-serine, D-threonine synthesis from L-threonine is not detectable, either in intact cells or with recombinant enzyme (Fig. 1, E and F, and data not shown).
Analysis of intracellular levels of L-threonine confirmed that the amino acid is a serine racemase substrate. A specific decrease in L-threonine is observed both at low (0.6 mM) and high (7 mM) concentrations of L-threonine in the medium (Fig. 1G). Even in the presence of a 10-fold higher L-serine concentration in the medium (7 mM L-serine), a decrease in cellular L-threonine is observed (Fig. 1G).
To verify the characteristics of L-threonine ␣,␤-elimination, we checked its properties in vitro. Unlike other PLP-dependent enzymes, serine racemase activity displays a unique dependence on ATP, which acts as an activator (12). In the absence of ATP, no L-threonine ␣,␤-elimination was observed ( Fig. 2A). By contrast, partial activation of L-serine ␣,␤-elimination was achieved by increasing free magnesium concentration alone ( Fig. 2A). Mg⅐ATP greatly stimulated L-threonine ␣,␤-elimination with generation of 2-oxobutyrate (Fig. 2B). The concentration of Mg⅐ATP required for half-maximal stimulation of Lthreonine utilization was somewhat higher than that required for L-serine (Fig. 2B). Hydrolysis of ATP was not required for enzyme activation and non-hydrolyzable analogs of ATP stimulated threonine ␣,␤-elimination as well (data not shown). L-Threonine utilization was optimal at alkaline pH (Fig. 2C). The Michaelis constant for L-threonine was in the millimolar range (K m ϭ 55 mM), higher than for L-serine (K m ϭ 10 mM), which suggests a lower affinity for the former (Table I). The efficiency of L-threonine ␣,␤-elimination calculated by the k cat /K m ratio was 3.3 lower than that with L-serine as substrate (Table I). Ammonia generated from the aminoacrylate intermediate, formed by the ␣,␤-elimination of water from L-threonine, was also detected (data not shown). D-Threonine (up to 100 mM) was not used as substrate by serine racemase, indicating a higher specificity for the L-form.
The utilization of L-threonine by serine racemase indicates that the ␣,␤-elimination activity displays a broader specificity than previously assumed. This prompted us to investigate ␣,␤elimination of additional substrates. Although considered a reaction product, we found that D-serine was strongly eliminated by serine racemase in an ATP-dependent manner, generating pyruvate and ammonia (Table I). The k cat for D-serine ␣,␤-elimination was about one-fourth of that observed with L-serine in the presence of ATP, compared with one-tenth of L-serine ␣,␤-elimination in the absence of ATP. The K m for D-serine in the ␣,␤-elimination reaction was similar to that of L-serine, and the relative efficiency of D-serine ␣,␤-elimination in the presence of ATP was 3.6 lower than that of L-serine ␣,␤-elimination as estimated by the k cat /K m ratio (Table I).
The ␣,␤-elimination of D-serine by serine racemase is in apparent contradiction with its role in D-serine biosynthesis. Therefore, we sought to investigate in more detail the kinetic characteristics of D-serine ␣,␤-elimination both with recombinant enzyme and in intact cells.
Previous measurements of the initial rate of D-serine synthesis in vitro were carried out using a large excess of L-serine (10 mM (8, 12, 27)). This may have masked D-serine ␣,␤-elimina-tion. When present at physiological-like concentration (1 mM), L-serine was rapidly consumed by the combined racemization and ␣,␤-elimination activities of serine racemase (Fig. 3A). Under this condition, D-serine was unstable; its levels fell as the concentration of L-serine decreased, reflecting its ␣,␤-elimination (Fig. 3B). Because EDTA impairs ␣,␤-elimination (12), we examined its effects on the stability of both D-and L-serine. Consistent with a previous report (12), the initial rate of Dserine synthesis in the presence of EDTA was much slower than with Mg 2ϩ and ATP. In the presence of EDTA, however, synthesized D-serine was stable because of the lower ␣,␤-elimination activity. As a result, D-serine reached higher values at longer reaction times when ␣,␤-elimination was impaired by EDTA (Fig. 3B). A similar stabilization promoted by EDTA was observed for L-serine levels (Fig. 3A). Production of pyruvate inversely correlated with the decrease in L-serine levels (Fig. 3C). Fig. 3B suggests that ␣,␤-elimination of D-serine may limit its accumulation. Thus, we next examined if serine racemase is able to maintain the same ratio of L-to D-serine in vitro as that found in vivo (3:1 (18)). When 1 mM L-serine and 0.3 mM Dserine were present from the start of the reaction, levels of both L-and D-serine decreased over time in the presence of the cofactors Mg 2ϩ and ATP (Fig. 3D). The degradation of D-serine observed in Fig. 3D by the racemase was not due to consumption of the precursor L-serine. Even when levels of L-serine were only reduced by 20 -30%, a decrease in D-serine was observed (Fig. 3D). Addition of 1 mM L-serine did not significantly slow down D-serine ␣,␤-elimination by serine racemase, whereas EDTA effectively blocked it (Fig. 3E). As expected, levels of pyruvate inversely correlated with the decrease in L-and Dserine levels (Fig. 3F).
The instability of D-serine in vitro prompted us to verify the extent to which D-serine ␣,␤-elimination occurs in intact cells. Thus, intracellular levels of D-serine were monitored in cells cultured in media containing D-serine, but with reduced Lserine concentration to avoid D-serine synthesis. Transfection of wild-type serine racemase was associated with a drastic decrease in D-serine immunoreactivity when compared with the inactive mutant K56G (Fig. 4A). Any cross-reactivity with L-serine was blocked by preincubating the antibody with Lserine-glutaraldehyde conjugate. When the antibody was preabsorbed with excess D-serine-glutaraldehyde conjugate, no immunoreactivity was observed (Fig. 4A, inset). HPLC analysis confirmed the decrease in intracellular D-serine in wild-type serine racemase-transfected cells, reflecting intracellular ␣,␤elimination of D-serine (Fig. 4B).
To verify whether D-serine ␣,␤-elimination also occurs in glial cells, we infected primary astrocyte cultures with lentivirus harboring the gene of serine racemase or GFP control and checked the total D-serine in the culture. The rate of D-serine consumption was higher in serine racemase infected astrocytes when compared with GFP control, reflecting the ␣,␤-elimination of D-serine by the enzyme (Fig. 4C).
To verify how the ␣,␤-elimination affects D-serine synthesis by serine racemase, we generated a point mutant (Q155D) that displays impaired ␣,␤-elimination activity against L-serine, Dserine, and L-threonine (Fig. 5 and Table I). Production of pyruvate from L-serine in vitro by the mutant Q155D was several times lower than the wild-type, with little change in the K m (Table I and Fig. 5A). Analysis of D-serine and L-threonine ␣,␤-elimination also indicated an impairment of ␣,␤-elimination toward these substrates promoted by the Q155D mutation, as revealed by a large decrease in k cat (Table I). The k cat /K m ratio of ␣,␤-elimination of L-serine by the mutant Q155D in the presence of ATP was 4.6-fold lower than the wild-type (Table I).
Conversely, racemization activity of the mutant Q155D was increased several-fold when compared with wild-type enzyme, with no change in K m (Fig. 5B and Table II). As a result, the ratio of ␣,␤-elimination versus racemization with the Q155D mutant was 15 times lower than the wild-type enzyme (0.25 versus 3.7, Table III).
ATP was found to affect moderately the partition of ␣,␤elimination and racemization reactions of L-serine, favoring the former. Addition of ATP increased the k cat /K m of ␣,␤-elimination by 4.5, whereas the k cat /K m of racemization was increased by 2-fold (compare Tables I and II). The activity of mutant Q155D was also strongly stimulated by ATP (Tables I and II). The concentration of ATP required for half-maximal stimulation of Q155D activity was unchanged when compared with wild type enzyme (data not shown).
The pH profile of ␣,␤-elimination and racemization of Lserine were similar and not significantly altered by the mutation Q155D (Fig. 5, compare C and D). Both the wild-type enzyme and the Q155D mutant displayed optimal activities at alkaline pH. Stimulation of activity at alkaline pH values was more prominent in the absence of ATP, but similar for ␣,␤-  elimination and racemization reactions. One exception was the ␣,␤-elimination of L-serine with Q155D mutant in the absence of ATP, which was very low and did not increase at alkaline pH values. The results imply that the changes in ␣,␤-elimination and racemization reactions by the mutation Q155D are not due to changes in pH profile of the enzyme but to changes in partitioning of the resonance-stabilized carbanion intermediate between reprotonation on C␣ and elimination of the ␤-hydroxyl group (Scheme 1).
The Q155D mutation lies in a conserved region of the racemase with previously unassigned function (amino acids 152-165) that is predicted to be a flexible random coil loop by using secondary structure analysis software (Protean, DNAStar, Inc.). The closest homologue whose structure has been determined is the biosynthetic threonine dehydratase enzyme from E. coli (PDB entry 1TDJ), to which serine racemase has significant homology at the amino acid level and for predicted secondary structure. The corresponding loop in threonine dehydratase (residues 154 -159, Fig. 5E) connects the catalytic subdomains N1 and N2, lying within 10 Å from the pyridoxal 5Ј-phosphate (28). An auxotrophic mutant at this loop (Pro 156 to Ser) exhibited a decrease in threonine dehydratase activity (29). To investigate the role of the homologous region in serine racemase, we mutated a number of residues in the vicinity of Gln-155, including P153S, which is homologous to the auxotrophic mutant P156S of threonine dehydratase. The residues mutated are depicted in bold and were aligned to other fold type-II PLP-dependent enzymes, which exhibited overall amino acid identity to serine racemase ranging from 25 to 32% (Fig. 5E). Interestingly, similar to Q155D mutation, the mutants H152S, P153S, and N154F all elicited a change in the partition of ␣,␤-elimination and racemization activities as revealed by the ratios of k cat /K m (Table III). The mutations H152S, Q155D, and N154F promote increased racemization activity, although they render this region of serine racemase to be even more similar to rat serine dehydratase (Fig. 5E,  SDHL). None of the mutations changed the K m for either ␣,␤elimination or racemization, nor did they change the concentration of ATP required for half-maximal stimulation of serine racemase activity (data not shown).
To verify the extent to which ␣,␤-elimination affects D-serine synthesis in vivo, HEK 293 cells were transfected with the Q155D mutant. The mutation doubled the production of Dserine in cells (Fig. 6A). Like that observed in vitro, ␣,␤-elimination of L-serine and L-threonine in cells was decreased to a large extent with the Q155D mutant (Fig. 6, B and C). Western blot analysis demonstrated that the levels of expression of wild-type serine racemase and Q155D mutant were identical in the experiments (data not shown).
Because the mutant Q155D displayed approximately a 4-fold increase in the efficiency of conversion of D-into L-serine (k cat /K m of 0.73 compared with 0.21 mM Ϫ1 min Ϫ1 , Table II), we used it as a tool to evaluate the effects of ␣,␤-elimination in the reversibility of the serine racemase catalytic cycle in intact FIG. 4. D-Serine ␣,␤-elimination in HEK 293 and primary astrocyte cultures. A, HEK 293 cells were transfected either with serine racemase wild-type (SR-WT) or a catalytically inactive mutant (K56G). Twenty-four hours after transfection, DMEM culture medium was replaced by BME supplemented with 0.3 or 1 mM D-serine. Transfection of SR-WT was associated to a significant decrease in intracellular D-serine levels as revealed by immunocytochemistry (upper panels). Preabsorption of the antibody with D-serine glutaraldehyde-conjugate abolished immunoreactivity (inset). B, HPLC analysis reveals significant depletion of intracellular D-serine in HEK 293 cells transfected with wild-type serine racemase. C, consumption of D-serine in primary cortical astrocyte cultures infected with lentivirus-containing mouse serine racemase (•) or green fluorescent protein (GFP) gene (E). The astrocyte medium (BME) was supplemented with 1 mM D-serine in the absence of L-serine and D-serine content in medium and cells was analyzed after different times.
cells. Transfected cells were incubated with media containing 0.3 mM D-serine, but no added L-serine, and back synthesis of L-serine was monitored. Transfection of the wild-type enzyme promoted a decrease in L-serine to levels one-third of the control inactive enzyme (K56G) as well as a decrease in D-serine levels (Fig. 6, D and E). This indicated that in intact cells expressing wild-type serine racemase most of D-serine is elim-inated instead of being converted to L-serine. Conversely, the ␣,␤-elimination-deficient mutant Q155D displayed a robust increase in L-serine, both in the intracellular compartment (Fig.  6D) and in the cultured medium (Fig. 6E), caused by racemization of D-serine. Similar reversal with the wild-type enzyme was only observed when cells were incubated with supraphysiological values of D-serine (5 mM) (data not shown).
Because serine racemase continuously degrades D-serine, we wondered if there are cellular mechanisms in place to enhance the stability of D-serine and allow its accumulation in vivo. We found that a large fraction of D-serine produced by the cells was exported to the extracellular medium, in which D-serine may be more stable. In HEK293 cells transfected with serine racemase, about 96% of total synthesized D-serine was present in culture medium (Fig. 7A). Likewise, about 98% of synthesized D-serine was found in the medium of primary astrocyte cultures or in a more physiological preparation consisting of cultured cortico-striatal slices (Fig. 7A). This raised the possibility that compartmentalization of D-serine between the intra-and extra- with serine/threonine dehydratase homologues in rat (SDHL) and biosynthetic (THD1) and catabolic (THD2) serine/threonine dehydratases from Saccharomyces cerevisiae, Salmonella typhimurium, and Haemophilus influenzae. Residues in bold represent the amino acids chosen for sitedirected mutagenesis. Shaded areas correspond to homologous residues. The results are representative of replicate experiments carried-out with three different enzyme preparations.

TABLE II Kinetic parameters of racemization
The reaction was carried out as in Table I. Standard errors of three to six experiments are given in parenthesis.  a k(␣,␤-elim) ϭ k cat /K m for ␣,␤-elimination of L-serine and k(racem) ϭ k cat /K m for racemization of L-serine. Other conditions were as described in Table I. Standard errors are given in parenthesis. cellular milieu might render it less susceptible to ␣,␤-elimination. To examine this possibility, we monitored the half-lives of L-and D-serine in culture media of transfected cells. Medium L-serine decreased rapidly in serine racemase-transfected cells with very little remaining after 24 h (Fig. 7B). This was accompanied by synthesis of D-serine (Fig. 7C). Notice that levels of D-serine remained stable, even 36 h after medium L-serine was almost totally consumed (Fig. 7C). As expected, intracellular levels of L-serine were also decreased in serine racemase-transfected cells (Fig. 7D). When L-serine (1 mM) was added together with D-serine (0.3 mM), L-serine consumption was unchanged, and levels of D-serine in culture media were also stable (Fig. 7, E and F). Even though medium D-serine was unaffected, intracellular levels of D-serine were significantly depleted in serine racemase-transfected cells (Fig. 7G). This suggests that the intracellular D-serine pool is more susceptible to ␣,␤-elimination. Additionally, when 0.3 mM D-serine was added in the absence of L-serine, as much as 80% of D-serine still remained stable in medium after 24 h (data not shown). DISCUSSION Our results obtained both in vitro and in vivo suggest a novel mechanism for limiting cellular D-serine through serine racemase catalyzed ␣,␤-elimination, which also uses L-threonine as substrate. In contrast, the racemization reaction is specific for serine, because we did not detect epimerization of L-threonine. The homology of serine racemase to serine/threonine dehydratase of E. coli provides a rationale for threonine ␣,␤-elimination. However, it differs from the bacterial enzyme by its absolute dependence on ATP. The physiological concentrations of L-serine, D-serine, and L-threonine in the brain are about 1, 0.3, and 0.6 mM (18, 30), much below the K m values for ␣,␤-elimination and racemization (Tables I and II). The 5-to 6-fold higher K m of serine racemase for L-threonine versus L-serine (Table I) raises the doubt that threonine may not be a relevant substrate at physiological concentrations. However, due to its higher k cat when compared with L-serine (Table I), the efficiency of L-threonine ␣,␤-elimination is in the same range as for D-serine ␣,␤-elimination, with k cat /K m about three times lower than for L-serine (Table I). This explains the significant Lthreonine ␣,␤-elimination seen by transfecting serine racemase in HEK293 cells (Fig. 1). The amount of keto acid (2-oxobutyrate) accumulated in the culture medium was lower than the observed decrease in L-threonine levels. This may be explained by the cellular oxidation of 2-oxobutyrate through either the mitochondrial branched-chain oxoacid dehydrogenase complex or pyruvate dehydrogenase (31).
Brain D-serine exhibits a half-life of about 16 h (27), but its degradative pathway is not clear. The only enzyme known to degrade D-serine in mammals is D-amino acid oxidase, which occurs in the cerebellum and brainstem. This enzyme, however, occurs only at very low levels in the forebrain areas such as the cerebral cortex and hippocampus, where high levels of D-serine are quite constant from 3-to 86-week-old rats (32). Moreover, mutant mice possessing inactive D-amino acid oxidase enzyme exhibit large increases in D-serine in the cerebellum and brainstem, but do not display increased D-serine in forebrain areas (33). This strongly suggests that D-serine degradation is not mediated by D-amino acid oxidase in the forebrain.
The results presented here imply that the ability of serine racemase to eliminate D-serine may be important to limiting D-serine levels in brain areas where D-amino acid oxidase is poorly expressed. Fig. 8 summarizes the proposed model for removing and further degrading extracellular D-serine in the forebrain. Upon its synthesis and release from the cells, Dserine stimulates NMDA receptors at the co-agonist site. This process could be terminated by D-serine re-uptake into the cells. Our results suggest that serine racemase can degrade D-serine taken up into the cells through its ␣,␤-elimination activity, generating pyruvate and ammonia. Accordingly, primary astrocyte cultures overexpressing serine racemase by lentivirus infection have increased D-serine consumption by ␣,␤-elimination (Fig. 4C). This might constitute a mechanism for regulating intracellular D-serine concentration upon D-serine re-uptake (Fig. 8). One caveat of such a model is that the uptake is key as a mode of synaptic inactivation. Glial cells take up D-serine in vivo (34), and several transporter candidates have been identified for D-serine, both in astrocytes and in neurons. These include Na ϩ -dependent (23,35) and Na ϩindependent (36, 37) neutral amino acid transporters. However, these transporters mediate only slow or non-selective uptake of D-serine. Alternatively, part of D-serine can leave the brain by efflux transport across the blood-brain barrier and later be excreted in the urine (Fig. 8). This could account for the high levels of D-serine in the urine of rodents and humans (38), but the existence of such pathway remains to be determined.
The effects exerted by the mutations in the amino acids 152-155 of serine racemase (Table III) are related to changes in the partition ratios for ␣,␤-elimination versus racemization (Scheme 1). The results implicate this region as a possible determinant of the reaction specificity of serine racemase. Conceivably, these mutations might alter the conformation of the enzyme and affect the microenvironment of the active site. We found that ATP also had moderate effects on the partitioning between ␣,␤-elimination and racemization of L-serine, favoring the former reaction. ATP also stimulates the efficiency of Dserine ␣,␤-elimination more than 10-fold (Table I). In vivo magnesium and ATP levels are in the millimolar range, FIG. 6. ␣,␤-Elimination limits cellular D-serine and reversal of the racemase. When indicated, HEK 293 cells were transfected with inactive mutant of serine racemase (K56G) wild-type (WT) or Q155D mutant. Amino acid levels were determined 48 h after addition of substrates to culture media, which consisted of BME plus 10% fetal bovine serum. A, stimulation of D-serine synthesis by Q155D mutation. D-Serine produced was monitored after addition of 7 mM L-serine to culture media. B, Q155D mutant display decreased L-serine ␣,␤-elimination calculated by subtracting the amount of D-serine synthesized from the total decrease in medium L-serine. C, Q155D mutant display decreased L-threonine ␣,␤-elimination calculated by the decrease in total L-threonine (7 mM) added to culture media. D and E, reversal of serine racemase reaction with Q155D mutant. Culture media of HEK 293 transfected cells were supplemented with 300 M D-serine and intracellular (D) or medium (E) L-serine was monitored by HPLC analysis. The results represent the average Ϯ S.E. of three independent transfection experiments.
whereas the K m of the serine racemase for Mg⅐ATP lies in the low micromolar range (12). This suggests that serine racemase activity in the presence of Mg⅐ATP better reflects the in vivo situation. Failure to include ATP in the enzyme assays explains previous failure or underestimation of the ␣,␤-elimination activity catalyzed by serine racemase (8,17).
As depicted in Fig. 8, reversal of serine racemase could play a role in degrading excess D-serine. The results with the ␣,␤elimination-deficient mutant, however, suggest that ␣,␤-elimination effectively competes with the reversal of serine racemase in vivo at physiological D-serine concentrations. This is supported by the experiments in which wild-type enzyme failed to accumulate L-serine from D-serine in cells incubated with 0.3 mM D-serine, whereas robust generation of L-serine was observed with the ␣,␤-elimination-deficient mutant (Fig. 6). Thus, degradation of D-serine does not occur by its racemization to L-serine. The reverse racemization reaction would be less effective than D-serine ␣,␤-elimination in decreasing cellular D-serine.
Extracellular levels of D-serine in the brain are higher than many common amino acids (39). It has been shown that release of D-serine from astrocytes in vitro is elicited by stimulation of AMPA/kainate receptors or by small neutral amino acids (2,23). Export of D-serine to the extracellular medium might be a mechanism to increase its stability by avoiding the intrinsic D-serine ␣,␤-elimination by serine racemase. This would explain the accumulation of D-serine in the extracellular milieu, despite the serine racemase ␣,␤-elimination reaction. Experiments measuring the L-and D-serine half-life using transfected HEK 293 cells confirmed the stability of extracellular versus intracellular D-serine (Fig. 7). This implies that changes in the degree of D-serine compartmentalization and in the gradient across the membrane (e.g. by increased or decreased cellular transport of D-serine) may affect D-serine metabolism. Thus, increased transport of D-serine into the cells should favor its degradation by the ␣,␤-elimination activity of serine racemase.
Our results have implications for the design of enzyme inhibitors and activators. Serine racemase inhibitors will be useful for conditions in which overstimulation of NMDA receptors takes place, such as stroke and neurodegenerative diseases. Conceivably, inhibitors of both racemization and ␣,␤-elimination activities of serine racemase will block D-serine synthesis, but removal of preformed D-serine will be slow due to the absence of D-amino acid oxidase activity in the forebrain. On the other hand, ligands that stimulate selectively D-serine ␣,␤elimination will be more effective in decreasing brain D-serine. D-Serine administration has been shown to ameliorate the negative symptoms of schizophrenia (40). Direct D-serine administration may be problematic due to its accumulation in several tissues and nephrotoxic effects (41). Our results point to the feasibility of designing serine racemase ligands to stimulate Lserine racemization while impairing ␣,␤-elimination. Such compounds will be useful in increasing brain D-serine specifically in the places where the racemase is normally expressed.