ATP binding to human serine racemase is cooperative and modulated by glycine

Authors


Correspondence

A. Mozzarelli, Department of Pharmacy, University of Parma, Parco Area delle Scienze 23/a, 43124 Parma, Italy

Fax: +39 0521 905151

Tel: +39 0521 905138

E-mail: andrea.mozzarelli@unipr.it

Abstract

The N-methyl d-aspartate (NMDA) receptors play a key role in excitatory neurotransmission, and control learning, memory and synaptic plasticity. Their activity is modulated by the agonist glutamate and by the co-agonists d-serine and glycine. In the human brain, d-serine is synthesized from l-serine by the dimeric pyridoxal 5′-phosphate-dependent enzyme serine racemase, which also degrades l- and d-serine to pyruvate and ammonia. The dependence of l- and d-serine β-elimination and l-serine racemization activities on ATP concentration was characterized, and was found to be strongly cooperative, with Hill coefficients close to 2 and apparent ATP dissociation constants ranging from 0.22 to 0.41 mm. ATP binding to the holo-enzyme, monitored by the fluorescence changes of the coenzyme, was also determined to be cooperative, with an apparent dissociation constant of 0.24 mm. Glycine, an active-site ligand, increased the serine racemase affinity for ATP by ~ 22-fold, abolishing cooperativity. Conversely, ATP increased the non-cooperative glycine binding15-fold. These results indicate cross-talk between allosteric and active sites, leading to the stabilization of two alternative protein conformations with ATP affinities of ~ 10 μM and 1.8 mm, as evaluated within the Monod, Wyman and Changeux model. Therefore, intracellular ATP and glycine control d-serine homeostasis, and, indirectly, NMDA receptor activity. Because hyper- and hypo-activation of NMDA receptors are associated with neuropathologies, the development of allosteric drugs modulating serine racemase activity is a promising therapeutic strategy.

Abbreviations
DAAO

d-amino acid oxidase

NMDA

N-methyl d-aspartate

PLP

pyridoxal 5′-phosphate

SR

serine racemase

Introduction

The N-methyl d-aspartate (NMDA) receptors are a sub-type of ionotropic receptors for glutamate, the main excitatory neurotransmitter in the central nervous system of vertebrates. NMDA receptors play a key role in excitatory neurotransmission, and control learning, memory and synaptic plasticity [1, 2]. NMDA receptor activation requires binding of the agonist glutamate as well as one of the co-agonists d-serine or glycine [3-5]. Recent studies proposed that d-serine is the main agonist of synaptic NMDA receptors, whereas glycine is the agonist of extra-synaptic NMDA receptors [6-8]. NMDA receptors are also modulated by antagonists such as kynurenic acid [9], produced by kynurenine aminotransferase [10]. Hypo-activation of NMDA receptors is associated with schizophrenia [11, 12], whereas hyper-activation is associated with several neuropathologies, including Alzheimer's and Parkinson's diseases, ischemia [13], amyotrophic lateral sclerosis and Rett syndrome [14].

d-serine is produced by serine racemase (SR, EC 5.1.1.18) [15], a pyridoxal 5′-phosphate (PLP)-dependent, homodimeric enzyme that is localized both in neurons and astrocytes [16-18]. SR shows a dual activity: reversible racemization of l-serine to d-serine and irreversible β-elimination of water from both enantiomers of serine to generate pyruvate and ammonia [19-21]. β-elimination and racemization take place simultaneously, albeit with different catalytic efficiencies [19, 22], which dictate the relative amount of the products. Both reactions have been proposed to be physiologically relevant, with the former leading to d-serine production and the latter to d-serine breakdown, a route for control of its homeostasis in brain areas [20] that lack the main degradative enzyme for d-serine, d-amino acid oxidase [23, 24]. Additional complex mechanisms regulate d-serine concentration, including shuttle of d-serine from neurons to astrocytes, and its release in the synaptic space [25], and SR modulation by ATP, divalent cations, at least four proteins and post-translational modifications (Fig. 1) [14, 26, 27].

Figure 1.

Cartoon representation of a model of human SR complexed with malonate, Mn2+ and ATP. The three-dimensional structure of holo human SR in the presence of malonate and Mn2+ was determined by Smith et al. [67] (PDB code 3L6B). The large and the small domains are shown in red and cyan, respectively; PLP bound to the active site Lys56 is shown in yellow in space-filling mode; the competitive inhibitor malonate is shown in blue in space-filling mode; the Mn2+ ion is shown in purple. By overlapping the structure of human SR with the structure of SR from Schizosaccharomyces pombe (PDB code 1V71) [33], which contains the ATP analog 5′-adenylyl methylenediphosphonate, shown in green space-filling mode, the site of the ATP binding site to human SR was modeled.

Despite the pharmacological relevance of SR as the source of the neuromodulator d-serine [4, 28], many aspects of its activity and regulation are still unclear. In particular, although the allosteric activation of murine SR by ATP was described more than ten years ago [19], the molecular details of this regulatory mechanism, which modulates the overall activity of SR and the relative rates of racemization and β-elimination [19, 20, 29], have never been thoroughly characterized. For instance, based on the reported apparent dissociation constant (KD) of ~ 10 μM for ATP from the murine enzyme [19, 30] and an intracellular concentration of ATP ranging from ~ 1 to 6 mm [31, 32], it was assumed that ATP constitutively saturates the two symmetric sites at the dimer interface (Fig. 1) [33], resulting in a permanently fully activated enzyme.

Here, we show that ATP binds to human SR with strong cooperativity, and increases the affinity of the active-site ligand glycine. Conversely, glycine stabilizes a protein conformation that binds ATP non-cooperatively and with high affinity, indicating a cross-talk between allosteric and active sites. Cooperativity evolutionarily developed as a molecular mechanism to maximize the dependence of protein function on the concentration of an effector in a narrow concentration range [34-38]. The cooperative behavior of human SR suggests the possibility of physiologically relevant regulation of d-serine synthesis by the intracellular levels of ATP and glycine. The pharmacological implications of these novel findings are discussed.

Results and Discussion

Catalytic parameters of human SR for the β-elimination and racemization reactions

We first measured the catalytic activity of human SR in the β-elimination reaction, using either l- or d-serine as substrates, in the absence and presence of ATP (Table 1 and Fig. S1A,B). The KM values for l-serine were determined to be 76 ± 10 and 12 ± 1 mm in the absence and presence of ATP, respectively, and those for d-serine were 56 ± 12 and 144 ± 27 mm, The kcat values for l-serine were 37 ± 4 and 147 ± 3 min−1 in the absence and presence of ATP, respectively, and those for d-serine were 0.9 ± 0.1 and 6.6 ± 0.6 min−1. Therefore, ATP decreases the KM for l-serine sixfold and increases the kcat approximately fourfold. The resulting catalytic efficiency (kcat/KM) values for l-serine elimination are 0.49 and 12.25 min−1·mm−1 in the absence and presence of ATP, respectively, indicating a 25-fold ATP-induced activation of β-elimination (Table 1). For d-serine, ATP increases both the KM and the kcat (Table 1). The resulting catalytic efficiency values are 0.016 and 0.046 min−1·mm−1 in the absence and presence of ATP, respectively, indicating an ATP-induced activation of only 2.9-fold for the d-serine β-elimination (Table 1). Because, in contrast to previously reported studies, our assay buffer solutions contained 150 mm sodium chloride, we evaluated its effect on the rate of d-serine elimination and found it to be negligible (Table 1). The observed ATP-induced activation of the l-serine elimination reaction is similar to that observed previously for the murine enzyme [19]. Overall, the net effect of ATP binding on the β-elimination reaction is a strong stimulation of l-serine degradation with a small effect on d-serine degradation.

Table 1. Catalytic parameters of human SR. Measurements were performed at 37 °C as described in 'Experimental procedures'. The kinetic parameters were determined by non-linear regression fitting of initial reaction rates to the Michaelis–Menten equation. Experimental curves and fittings are shown in Fig. S1. For d-serine β-elimination in the presence of ATP, the catalytic parameters obtained in the absence of NaCl are shown in parentheses.
ReactionKM (mm)kcat (min−1)kcat/KM (min−1·mm−1)
−ATP+ATP−ATP+ATP−ATP+ATP
l-serine β-elimination76 ± 1012 ± 137 ± 4147 ± 30.49 ± 0.1212.25 ± 1.28
d-serine β-elimination56 ± 12144 ± 27 (204 ± 27)0.9 ± 0.16.6 ± 0.6 (8.7 ± 1)0.016 ± 0.0050.046 ± 0.013 (0.042 ± 0.013)
l-serine racemization48 ± 1640 ± 152.9 ± 0.310 ± 10.06 ± 0.030.25 ± 0.14

We then measured the effect of ATP binding on racemization, using l-serine as the substrate (Table 1 and Fig. S1C). In the absence and presence of ATP, the KM values for l-serine were 48 ± 16 and 40 ± 15 mm, and the kcat values were 2.9 ± 0.3 and 10 ± 1 min−1, respectively (Table 1). The kcat/KM values are 0.06 and 0.25 min−1·mm−1 (Table 1). Therefore, ATP leads to a fourfold activation of l-serine racemization. It is interesting to note that, in the absence of ATP, the ratio for l-serine β-elimination and racemization efficiency is 8.2, whereas, in the presence of ATP, it is 50, indicating that ATP binding preferentially stimulates l-serine β-elimination rather than racemization. Furthermore, d-serine elimination occurs in both the absence and presence of ATP with very low efficiency compared with l-serine β-elimination (29- and 271-fold in the absence and presence of ATP, respectively), casting some doubt on its proposed role in controlling d-serine homeostasis [20]. When d-serine is formed, the rate of its degradation by human SR is very low, and d-serine accumulates. Interestingly, in the absence of ATP, the efficiency of all measured reactions is lower, thus indicating that ATP plays a significant role in controlling d-serine concentration. These findings suggest that the development of compounds that target the ATP allosteric site may be a powerful strategy for SR modulation.

With respect to the only published characterization of the catalytic parameters of human SR in the presence of ATP [22], we found a twofold lower efficiency for l-serine β-elimination, an ~ 17-fold lower efficiency for d-serine β-elimination, and an ~ 40-fold lower efficiency for l-serine racemization. These differences may be due to differences in enzyme assays, or different purification and stabilization protocols. Significant differences had previously been observed between the catalytic parameters of murine SR reported by Hoffman et al. [22] and Foltyn et al. [20].

ATP binding affinity to human SR measured in the presence of saturating concentrations of either l- or d-serine

Because ATP increases human SR catalytic activities, we exploited this effect to determine ATP affinity by monitoring the β-elimination activity at increasing ATP concentrations in the presence of saturating l- or d-serine (Fig. 2A,B) or by monitoring the racemization activity in the presence of saturating l-serine (Fig. 2C). Unexpectedly, the ATP binding curve exhibited a marked sigmoidal shape, indicating positive cooperativity (Fig. 2). Fitting of the titrations to the Hill equation allowed us to determine cooperative coefficients of 1.7 ± 0.2 for l-serine β-elimination, 1.9 ± 0.2 for d-serine β-elimination, and 2.1 ± 0.7 for l-serine racemization (Table 2). These Hill coefficients are close or equal to the maximum theoretical value for a protein containing two binding sites, indicating a strong link between the two ATP sites, which are 24 Å apart (Fig. 1). The apparent ATP KD values, determined from the fitting (Fig. 2 and Table 2), were 0.22 ± 0.01 mm for l-serine β-elimination, 0.41 ± 0.02 mm for d-serine β-elimination and 0.22 ± 0.05 mm for l-serine racemization (Table 2). Cooperative ATP binding was retained at sub-saturating concentrations of l-serine (Fig. S2), and was not associated with modulation of the monomer/dimer equilibrium of SR by ATP, as the specific activity was not dependent on protein concentration in the range 0.5–5 μM (data not shown). The observed cooperativity in ATP binding to SR indicates that the enzyme exists as an equilibrium of at least two conformations with significantly different microscopic dissociation constants for ATP.

Table 2. Binding parameters of ATP to human SR determined by monitoring either l- or d-serine β-elimination and l-serine racemization, and by coenzyme fluorescence changes in the absence and presence of glycine. The dependence of SR catalytic activity on ATP was performed as described in Fig. 2 and Experimental procedures., in the presence of either 530 mm l-serine or 750 mm d-serine, and was fitted to the Hill equation. The dependence of coenzyme fluorescence on ATP was performed as described in Figs 3 and 4 and Experimental procedures, in the absence and presence of 50 mm glycine.
Method of binding detectionKD ATP (mm)Hill coefficient
Activity assay (l-serine β-elimination)0.22 ± 0.011.7 ± 0.2
Activity assay (d-serine β-elimination)0.41 ± 0.031.9 ± 0.2
Activity assay (l-serine racemization)0.22 ± 0.052.1 ± 0.7
Coenzyme fluorescence (without glycine)0.26 ± 0.021.8 ± 0.2
Coenzyme fluorescence (with glycine)0.0049 ± 0.00061.02 ± 0.09
Figure 2.

(A) Dependence of the l-serine β-elimination activity of human SR on ATP concentration in the presence of saturating substrate. Experiments were performed at 37 °C, as described in Experimental procedures, in the presence of 530 mm l-serine. The experimental data were fitted to the Hill equation, yielding a Hill coefficient of 1.7 ± 0.2 and an apparent KD for ATP of 0.22 ± 0.01 mm. (B) Dependence of the d-serine β-elimination activity of human SR on ATP concentration in the presence of saturating substrate. Experiments were performed at 37 °C, as described in Experimental procedures, in the presence of 750 mm d-serine. The experimental data were fitted to a Hill equation, yielding a Hill coefficient of 1.9 ± 0.2 and an apparent KD for ATP of 0.41 ± 0.03 mm. (C) Dependence of the racemization activity of human SR on ATP concentration in the presence of a saturating concentration of l-serine. Experiments were performed at 37 °C, as described in Experimental procedures, in the presence of 200 mm l-serine. The experimental data were fitted to the Hill equation, yielding a Hill coefficient of 2.1 ± 0.7 and an apparent KD for ATP of 0.22 ± 0.05 mm.

ATP binding to human SR detected by fluorescence measurements

To investigate ATP binding to human SR by directly monitoring changes in protein conformation, we exploited the spectroscopic properties of the bound PLP. The absorption spectrum of SR exhibited a band centered at 412 nm (Fig. 3A), attributed to the ketoenamine form of the PLP Schiff base with the active site Lys56. When the enzyme was excited at 412 nm, a fluorescence emission band centered at 500 nm was observed (Fig. 3B). PLP emission at 500 nm is a very sensitive indication of the active-site conformation in terms of both intensity and wavelength of maximum emission, as previously demonstrated in the investigation of ligand binding to O-acetylserine sulfhydrylase [39-41]. Binding of ATP to SR causes an increase in the emission at 500 nm with a small blue shift (Fig. 3B). The dependence of fluorescence intensity on ATP concentration was found to be sigmoidal, with a Hill coefficient of 1.8 ± 0.2 and an apparent KD of 0.26 ± 0.02 mm (Fig. 3B), mirroring the cooperative behavior observed when monitoring enzyme activity (Fig. 2). In addition, the effect of ATP binding on the PLP fluorescence spectrum indicates that ATP induces a conformational change in the active site, with a decrease in its polarity and/or solvent accessibility. As that the apparent dissociation constant for ATP determined from fluorimetric measurements in the absence of ligands is close to the values determined by activity assays at saturating concentrations of substrates, it may be concluded that reaction intermediates of neither l- nor d-serine allosterically affect the ATP site.

Figure 3.

ATP binding to human SR. (A) Absorption spectrum of a solution containing 7.4 μM human SR, 50 mm TEA, 150 mm NaCl, 1 mm EDTA, pH 8.0. (B) Fluorescence emission spectra, upon excitation at 412 nm, of a solution containing 2.7 μM human SR, 50 mm TEA, 150 mm NaCl, 5 mm TCEP, 1 mm MgCl2, pH 8.0, and increasing concentrations of ATP, at 20 °C. (C) Dependence on ATP concentration of the fluorescence emission intensity at 490 nm upon excitation at 412 nm. The experimental points were fitted to the Hill equation, yielding a KD of 0.26 ± 0.02 mm and a Hill coefficient of 1.8 ± 0.2.

To assess whether the allosteric communication between the two ATP sites occurs when the enzyme is involved in a stable complex with a ligand, ATP binding was monitored in the presence of saturating concentrations of glycine, an amino acid capable of forming a reversible Schiff base with SR. Glycine is present in the brain at a concentration that varies significantly depending on localization, from 8–12 μM in the extracellular space [42, 43] to ~ 0.6 mm within neuronal cells [44], and is released by the same cells responsible for the liberation of d-serine [8] . The ATP binding curve in the presence of glycine (Fig. 4A) was found to be hyperbolic and strongly left-shifted with respect to the binding curve for the glycine-free SR. The calculated KD was 0.0049 ± 0.0006 mm, indicating very tight binding. A similar value for ATP affinity was reported previously [19]. Therefore, the presence of glycine stabilizes an SR conformation that binds ATP with high affinity and non-cooperatively. Thus ATP binding to the glycine-free and bound forms exhibits striking differences, confirming that SR possesses distinct inter-converting conformations.

Figure 4.

(A) Binding of ATP to human SR in the absence and presence of glycine. Fluorescence emission spectra upon excitation at 412 nm of a solution containing 2.7 μM human SR in 50 mm TEA, 150 mm NaCl, 5 mm TCEP, 1 mm MgCl2, pH 8.0, at 20 °C, were collected at increasing ATP concentrations in the absence (open circles, data from Fig. 3C) and the presence (closed circles) of 50 mm glycine. The dependence of the normalized fluorescence intensity at 490 nm on ATP concentration was fitted to the Hill equation, resulting in a Hill coefficient of 1.8 ± 0.2 and a KD of 0.26 ± 0.02 mm in the absence of glycine. In the presence of glycine, data were fitted to the quadratic equation for tight ligand binding (see 'Experimental procedures'), yielding a KD of 0.0049 ± 0.0006 mm. (B) Binding of glycine to human SR in the absence and presence of ATP. Fluorescence emission spectra upon excitation at 412 nm were collected at increasing glycine concentrations. Experiments were performed at 20 °C on a solution containing 2.7 μM human SR in 50 mm TEA, 150 mm NaCl, 5 mm TCEP, 1 mm MgCl2, pH 8.0, in the absence (open circles) and presence (closed circles) of 2 mm ATP. The dependence of the fluorescence emission intensity at 482 nm on glycine concentration was fitted to a binding isotherm, yielding a KD for glycine of 0.47 ± 0.03 mm in the presence of ATP and a KD of 7.0 ± 0.3 mm in the absence of ATP.

To further evaluate the allosteric effect of ATP binding on the active site, glycine binding was investigated in both the absence and presence of ATP (Fig. 4B). We found that glycine binds non-cooperatively in both cases. Non-cooperative binding was also observed for l- and d-serine, as indicated by the pure Michaelis–Menten rate dependence (Fig. S1) [20]. However, it is noteworthy that ATP decreased the KD for glycine ~ 15-fold, from 7 to 0.47 mm. Therefore, ATP-bound SR exhibits a KD for glycine that is lower than the intracellular concentration of glycine (0.6 mm [44]), suggesting a physiological inter-play between ATP and glycine that may lead to a complex pattern of SR modulation, that requires further investigation.

Analyzing ATP binding according to the Monod, Wyman and Changeux model

Within the framework of the Monod–Wyman–Changeux (MWC) model [45] applied to ATP binding, glycine-bound SR coincides with a high-affinity, non-cooperative state (R) that is in equilibrium with a low-affinity state (T) in the glycine-free or serine-bound SR. Under this assumption, the experimentally determined KD for ATP in the presence of glycine is assumed to correspond to the dissociation constant of the R state (KDR). Direct determination of this dissociation constant at 37 °C by fluorimetric titrations was hampered by the limited stability of SR and ATP. To overcome these limitations, the temperature dependence of the KDR was measured by ATP titrations in the presence of glycine at 10, 15 and 20 °C, and analyzed using the van't Hoff equation (Fig. S3). The extrapolated value of KDR for ATP at 37 °C was 0.0115 mm. Global fitting of the activity data reported in Fig. 2 using the MWC equation and this fixed KDR value allowed the estimation of a KDT for ATP of 1.8 ± 0.5 mm. This value falls within the intracellular ATP concentration, which ranges from 1 mm [32] to ~ 6 mm [31], but locally may be even lower, suggesting that SR is not constantly ATP-saturated. More specifically, this finding indicates that small changes in ATP concentration, as well as glycine concentration, have a strong impact on the distribution between low- and high-efficiency SR sites.

Conclusions

ATP elicits conformational changes both at the active site that is 15 Å away, and at the symmetric site for ATP binding that is 24 Å away (Fig. 1). Neither l- nor d-serine allosterically affect the ATP-binding sites or the symmetric active site, whereas glycine binding completely shifts the conformational equilibrium towards the ATP high-affinity state. The structural changes responsible for this complex cross-talk are not known. The yeast homolog of SR crystallized in the absence and presence of the ATP analog 5′-adenylyl methylenediphosphonate shows different relative orientations of the monomers [33]. It has been suggested that a hydrogen bond network connecting the active site to the ATP-binding site may be responsible for the ATP-dependent modulation of an open/closed conformational transition of the active site [33]. Investigations towards elucidation of the molecular basis of the communication between allosteric and active sites, exploiting both in silico and spectroscopic signals, are currently underway in our laboratory. Furthermore, by exploiting the sol–gel method for protein encapsulation [46], we may be able to trap and fully characterize distinct human SR conformations, as previously performed for other proteins, including hemoglobin [47-54], aspartate transcarbamilase [55], tryptophan synthase [56] and green fluorescent protein [57, 58]. Finally, given that the intracellular ATP concentration in neurons and astrocytes, where human SR has been detected, is ~ 1 mm and may vary significantly during metabolic activity and depending on energy supply, our findings suggest that the intracellular ATP concentration, as well as the relative concentration of glycine and l-serine, may play a significant role in controlling d-serine homeostasis and therefore NMDA receptor activity. In turn, these results support the notion of human SR [4, 14, 59-61] as a target for development of allosteric drugs that, by fine-tuning enzyme activity, may be used for treatment of psychiatric disorders and neuropathologies caused by NMDA receptor dysfunctions.

Experimental procedures

Materials

The chemicals, NADH, lysozyme and horseradish peroxidase were of the best commercial quality available and were purchased from Sigma-Aldrich (St Louis, MO), with the exception of Tris (2-carboxyethyl) phosphine (TCEP), which was purchased from Apollo Scientific (Denton, Manchester, UK). Recombinant d-amino acid oxidase (DAAO) from Rhodotorula gracilis was a generous gift from Loredano Pollegioni (Department of Biotechnology and Molecular Sciences, University of Insubria, Varese, Italy).

Enzyme preparation

A pET28a-derived plasmid purchased from Novagen (EMD Millipore, Darmstadt, Germany) encoding hexahistidine-fused human serine racemase [62] (provided by Michael Toney, Department of Chemistry, University of California, Davis, CA, USA) was transformed into Escherichia coli BL21-CodonPlus®(DE3)-RIL (Agilent Technologies, Santa Clara, CA, USA). Cells were grown at 37 °C in Luria–Bertani medium supplemented with kanamycin (50 μg·mL−1) and chloramphenicol (50 μg·mL−1). When the cells reached an attenuance at 600 nm of ~ 0.5, benzyl alcohol was added to the medium to a final concentration of 10 mm [63] and the growth temperature was lowered to 20 °C. Benzyl alcohol was added because it is known to increase the expression of molecular chaperones in response to stress conditions [63]. After 20 min, isopropyl thio-β-d-galactoside was added to a final concentration of 10 μM, and the culture was further incubated at 20 °C for 20 h. Cells were harvested by centrifugation (3500 g, 15 min, 4°C), and the pellet was resuspended in lysis buffer (50 mm sodium phosphate, 300 mm NaCl, 5 mm TCEP, 50 μM PLP, 200 μM phenylmethanesulfonyl fluoride, 200 μM benzamidine, 1.5 μM pepstatine at pH 8.0). Cell walls were then disrupted by treatment with lysozyme (1 mg·mL−1), followed by sonication. The homogenate was clarified by centrifugation (16000 g, 30 min, 4°C), and the supernatant was incubated with TALON His-tag purification resin (Clontech, Mountain View, CA, USA). After washing the resin three times for 10 min at 4°C, using a buffer containing 50 mm sodium phosphate, 150 mm NaCl, 20 mm imidazole, 5 mm TCEP, pH 8.0, the protein was recovered by suspending the resin in a buffer containing 50 mm sodium phosphate, 150 mm NaCl, 250 mm imidazole at pH 8.0. The protein solution was diafiltered in an Amicon™ cell (Merck/Millipore) in 50 mm triethanolamine (TEA), 150 mm NaCl, 1 mm EDTA, pH 8.0. All steps were performed at 4 °C. The protein solution was concentrated to 64 μM (monomers concentration), supplemented with 5% glycerol and 5 mm TCEP, flash-frozen and stored at −80 °C in small aliquots. Protein purity was assessed as 98% by densitometry of Coomassie blue-stained bands of an SDS/PAGE gel using a Chemidoc gel imaging system (Bio-Rad, Hercules, CA, USA) (Fig. S4). The yield of pure SR was 2.5 mg per liter of E. coli culture.

Activity assays

β-elimination reactions

The initial rate of l- or d-serine β-elimination was monitored by coupling the reaction with pyruvate reduction by lactate dehydrogenase and following NADH disappearance at 340 nm [27], using a Varian CARY400 spectrophotometer (Palo Alto, CA, USA) with a thermostated cell holder. Within the time course of an assay (~ 10 min), it was estimated that < 0.03% of the substrate (either l-serine or d-serine) is converted to its enantiomer via racemization. The typical activity assay solution contained 50 mm TEA, 150 mm NaCl, 50 μM PLP, 5 mm dithiothreitol, 1 mm MgCl2, 60 U·mL−1 l-lactate dehydrogenase (one unit reduces 1.0 μmole of pyruvate to l-lactate per min at pH 7.5 at 37 °C) and 300 μM NADH. The dependence of the initial rate on either l- or d-serine concentration was determined in the absence and presence of 1–2 mm ATP.

Racemization reaction

The initial rate of d-serine formation as a function of l-serine concentration was determined via a discontinuous assay based on oxidation of d-serine by DAAO, which produces hydrogen peroxide [64]. Hydrogen peroxide-mediated oxidation of o-dianisidine dihydrochloride by horseradish peroxidase leads to formation of a chromophoric product that, once treated with sulfuric acid to increase solubility, absorbs at 530 nm. l-serine at various final concentrations was added to a solution containing 50 mm TEA, 150 mm NaCl, 50 μM PLP, 1 mm MgCl2 and 2.3 μM human SR, either in the absence or presence of ATP at concentrations ranging from 0.01 to 2 mm. The reaction mixture was incubated at 37 °C, and aliquots were periodically removed for determination of d-serine concentration. Because the commercially available l-serine contains a significant amount of d-serine, l-serine solutions were purified prior to use by incubation for 48 h with 400 U·mL−1 DAAO and 100 U·mL−1 catalase, which were then thermally inactivated. Absolute concentrations of d-serine in the assays were assessed using calibration curves obtained by reacting pure d-serine at known concentrations with DAAO and peroxidase.

Dependence on ATP of l- and d-serine β-elimination and l-serine racemization

The dependence of the β-elimination initial velocity on ATP concentration was determined either in the presence of saturating l-serine (530 mm) or d-serine (750 mm), corresponding to an approximately tenfold excess with respect to the KM determined in the absence of ATP (Table 1). The dependence of the racemization initial velocity on ATP concentration was determined in the presence of 200 mM l-serine. The reactants were pre-equilibrated at 37.0 ± 0.5 °C in the thermostated cell holder of the spectrophotometer before starting the reaction by adding the enzyme at a final concentration of typically 0.4–0.5 μM. To ensure that dimer dissociation was not affecting activity measurements, the final concentration of SR was increased to 5 μM without observing any change in the catalytic parameters. This demonstrates that the specific enzyme activity did not depend on protein concentration either in the absence or presence of ATP.

Absorbance and fluorescence measurements

Absorption spectra were performed using a Varian CARY400 spectrophotometer, with the cell holder thermostated at 20.0 ± 0.5 °C. Solutions contained 50 mm TEA, 150 mm NaCl, 1 mm EDTA and 7.4 μM human SR, pH 8.0.

Human SR fluorescence spectra in the absence and presence of ligands were collected using a FluoroMax-3 fluorometer (HORIBA Jobin Yvon, Kyoto, Japan), with the cell holder thermostated at 20.0 ± 0.5 °C. Emission spectra upon direct excitation of the co-factor at 412 nm were recorded using slits set for optimal signal-to-noise ratio on a solution containing 2.7 μM enzyme, 50 mm TEA, 150 mm NaCl, 5 mm TCEP, 1 mm MgCl2, pH 8.0. The binding affinity of ligands to SR was determined by monitoring the increase in fluorescence emission of the coenzyme upon excitation at 412 nm, as previously reported for other PLP-dependent enzymes [65].

Data analysis

To determine the Hill coefficient and the apparent KD, data points were fitted to the Hill equation:

display math

where y is the fluorescence emission intensity or the initial reaction velocity, [L] is the ligand concentration (ATP), and n is the Hill coefficient. y0 is an offset and a is the amplitude.

Given the very high affinity, determination of the KD of ATP in the presence of glycine was performed by fitting the experimental points to a quadratic equation that describes tight binding [66]:

display math

where y is the fluorescence emission intensity, [P] is the fixed protein concentration and [L] is the variable ligand concentration. y0 is a horizontal offset and a is the amplitude.

For analysis of the data points using the MWC model [45], the following equation for a two-site protein was applied:

display math

where y is the fraction of the effect, LMWC [T]0/[R]0 is the MWC allosteric constant, i.e. the ratio between the concentration of the two states in the absence of ligand, c = KDR/KDT is the ratio between the microscopic dissociation constants of the two forms and α = [L]/KDR is the normalized concentration of the ligand with respect to KDR. y0 is a horizontal offset and a is the amplitude. The KDR used in the analysis was independently estimated from binding curves of ATP in the presence of glycine.

The dependence of KDR as a function of temperature was determined at 10, 15 and 20 °C, and the value at 37 °C was extrapolated by linear regression using the van't Hoff plot:

display math

where T is the temperature in K, ΔH0 is the standard enthalpy, ΔS0 is the standard entropy, and R is the gas constant.

Acknowledgments

We gratefully acknowledge Michael Toney (Department of Chemistry, University of California at Davis, CA, USA) for the generous gift of the expression plasmid of human SR, and Gabriele Costantino (University of Parma, Italy) for attracting our attention to SR and for stimulating discussions. We thank Loredano Pollegioni (Department of Biotechnology and Molecular Sciences, University of Insubria, Varese, Italy) for the generous gift of recombinant DAAO. We are grateful to Emiliano Bolesani, Claudia Caminiti, Pierfrancesco Lanzilotti, Laura Tigli and Letizia Tongiani for contributing to optimization of the protein purification protocol and SR assays. This work was supported by funds from the Italian Ministery of University and Research.

Ancillary