Turning pyridoxal-5′-phosphate-dependent enzymes into thermostable binding proteins: d -Serine dehydratase from baker’s yeast as a case study

Biochimie 94 (2012) 479e486

Contents lists available at SciVerse ScienceDirect

Biochimie journal homepage: www.elsevier.com/locate/biochi

Research paper

Turning pyridoxal-50 -phosphate-dependent enzymes into thermostable binding proteins: D-Serine dehydratase from baker’s yeast as a case study Maurizio Baldassarre1, Andrea Scirè1, Fabio Tanfani* Dipartimento di Scienze della Vita e dell’Ambiente, Università Politecnica delle Marche, Via Ranieri, 60131 Ancona, Italy

a r t i c l e i n f o

a b s t r a c t

Article history: Received 23 May 2011 Accepted 24 August 2011 Available online 30 August 2011

D-serine

Keywords: D-serine dehydratase Pyridoxal-50 -phosphate Infrared spectroscopy Saccharomyces cerevisiae Protein structure Schiff base

dehydratase from Saccharomyces cerevisae is a recently discovered dimeric enzyme catalyzing the b-elimination of D-serine to pyruvate and ammonia. The reaction is highly enantioselective and depends on cofactor pyridoxal-50 -phosphate (PLP) and Zn2þ. In our work, the aldimine linkage tethering PLP to recombinant, tagged D-serine dehydratase (Dsd) has been reduced by treatment with NaBH4 so as to yield an inactive form of the holoenzyme (DsdR), which was further treated with a protease in order to remove the amino-terminal purification tag. Fourier Transform infrared (FT-IR) spectroscopic analysis revealed that both the reduced form (DsdR) and the reduced/detagged form (DsdRD) maintain the overall secondary structure of Dsd, but featured a significant increased thermal stability. The observed Tm values for DsdR and for DsdRD shifted to 71.5  C and 73.3  C, respectively, resulting in nearly 11  C and 13  C higher than the one measured for Dsd. Furthermore, the analysis of the FT-IR spectra acquired in the presence of D-serine and L-serine indicates that, though catalytically inert, DsdRD retains the ability to enantioselectively bind its natural substrate. Sequence analysis of D-serine dehydratase and other PLPdependent enzymes also highlighted critical residues involved in PLP binding. In virtue of its intrinsic properties, DsdRD represents an ideal candidate for the design of novel platforms based on stable, nonconsuming binding proteins aimed at measuring D-serine levels in biological fluids. Ó 2011 Elsevier Masson SAS. All rights reserved.

1. Introduction Just like matter and antimatter, nature has mysteriously chosen acids to outnumber D-amino acids by several orders of magnitude [1,2]. However, D- and L-amino acids coexist to shape life as we know it [3]. D-amino acids have, indeed, been found in nearly all life forms, from single-celled organisms as simple as bacteria or baker’s yeast to complex, multicellular eukaryotes, such as plants and animals. Humans make no exception, relying on a number of different D-amino acids for cell communication [4,5] and regulation of hormone synthesis and release [6,7]. In the human brain, D-serine is produced by astrocytes surrounding specific classes of neurons, such as motor neurons [8], where it acts as a coactivator for N-methyl-D-aspartate glutamate receptors [9]. Research on this type of receptor suggests that robust L-amino

Abbreviations: FT-IR, Fourier transform infrared; Dsd, recombinant, unreduced, tagged D-serine dehydratase; DsdR, reduced, tagged D-serine dehydratase; DsdRD, reduced, detagged D-serine dehydratase; PLP, pyridoxal-50 -phosphate; IPTG, isopropyl-b-thiogalactoside. * Corresponding author. Tel.: þ39 071 2204687; fax: þ39 071 2204398. E-mail address: [email protected] (F. Tanfani). 1 These authors contributed equally to the work. 0300-9084/$ e see front matter Ó 2011 Elsevier Masson SAS. All rights reserved. doi:10.1016/j.biochi.2011.08.016

activation relies on the presence of D-serine at the synaptic cleft, thus making this simple, unusual molecule a key component in the transmission of nervous stimuli along motor neurons. Changes in the synaptic D-serine levels beyond physiological thresholds result in glutamatergic hypofunction, eventually leading to impairment of signal transmission. Not surprisingly, altered levels of D-serine are usually measured in the cerebrospinal fluid and blood of individuals affected by one of a number of neurodegenerative disorders [10], including Alzheimer’s disease [11], amyotrophic lateral sclerosis [12] and schizophrenia [13]. Measurement of D-serine levels in biological fluids might contribute significantly in determining the onset and evolution of these pathologies on one hand, and in assessing the efficacy of novel drugs on the other. The enzyme D-serine dehydratase from Saccharomyces cerevisiae catalyzes the b-elimination of D-serine to yield pyruvate and ammonia and has been shown to be highly enantioselective [14]. The reaction depends on a zinc ion and cofactor pyridoxal-50 phosphate (PLP). This is bound within the enzyme active site via formation of a Schiff base with a specific lysyl residue, supposedly Lys57. In our previous work, this interesting enzyme was expressed in a recombinant, tagged form (hereafter referred to as Dsd) and its structure was investigated by means of Fourier Transform infrared (FT-IR) spectroscopy [15]. Dsd showed a marked instability in most

480

M. Baldassarre et al. / Biochimie 94 (2012) 479e486

buffers and formation of a precipitate was observed over medium to long storage times even at 4  C. Unexpectedly, we found that deuterium oxide enhanced protein instability. This spurred us to use 1H2O for our infrared measurements, instead of the more common 2H2O. We hypothesized that loss of the bound cofactor might be the main cause for the observed instability, as reported in the literature for other PLP-dependent enzymes [16]. In this work, FT-IR spectroscopy has been used to investigate the structure and thermal stability of Dsd following aldimine reduction with NaBH4, resulting in an inactive, cofactor-blocked form of Dsd (DsdR). Infrared data revealed that no significant changes in the secondary structure occurred upon reduction; however the reduced form showed a markedly higher thermal stability. DsdR was also analysed following removal of the amino-terminal purification tag (DsdRD) resulting in a slightly more stable form than DsdR. In-depth spectroscopic analyses were carried out on DsdRD in order to evaluate the importance of Zn2þ on the secondary structure and thermal stability of the protein. In addition, this catalytically-inert sample was tested for the retained ability to bind D-serine preferentially over L-serine with the aim to evaluate its potential use as biological material for the development of D-serine sensitive biosensors.

2. Materials and methods 2.1. Materials The following chemicals were purchased from Aldrich (SigmaeAldrich S.r.l., Milan, Italy): deuterium oxide (99.9% 2H2O), 2HCl, NaO2H, pyridoxal-50 -phosphate (PLP), sodium borohydride, EDTA, LDH from rabbit muscle, D-serine, L-serine, NADH, thrombin and zinc chloride. All other chemicals were commercial samples of the purest quality. 2.2. Purification of recombinant Dsd Expression and purification of recombinant D-serine dehydratase from baker’s yeast was performed as previously described [15]. Briefly, Escherichia coli cells strain BL21(DE3) were transformed with the pET-15b expression vector harbouring the gene of Dsd cloned downstream of a nucleotide sequence encoding a 20mer purification tag, and grown aerobically in LB medium supplemented with 50 mg mL1 ampicillin. Expression of Dsd was triggered by addition of IPTG to a final concentration of 0.5 mM and let to continue for a further 3 h. The culture was harvested by centrifugation and the cells were lysed on a French pressure cell press apparatus. Recombinant Dsd was purified from the bacterial intracellular content by means of a nickel-chelating resin (Ni-NTA, Qiagen). Pure, homogeneous Dsd was dialysed against cold (4  C) storage buffer consisting of 50 mM potassium phosphate buffer (pH 7.9), 20 mM ZnCl2 and 20 mM PLP, and quantified spectrophotometrically using a molar extinction coefficient of 47,850 M1 cm1 at 280 nm [17]. The protein at this stage will be hereafter referred to as untreated or unreduced, tagged Dsd or more simply Dsd. All subsequent treatments, including dialyses and centrifugations, were carried out at 4  C unless specified. All samples were store in aluminium foil-protected tubes to minimize PLP decomposition. 2.3. Sodium borohydride treatment Fresh 1 M NaBH4 was prepared by dissolving an adequate amount of moisture-free powder in distilled, ice-cold water. Reduction of the aldimine linkage between PLP and Dsd was

achieved by addition of concentrated sodium borohydride (final concentration 5 mM) to samples of Dsd in storage buffer containing 20 mM PLP. During the reduction, the concentration of PLP was not increased further to avoid non specific interactions of PLP with surface exposed lysyl residues, as reported in literature [18]. Excess NaBH4 and unbound PLP were removed by extensive dialysis of samples against PLP-free storage buffer. These procedures were performed in the dark to minimize PLP decomposition. To assess the extent of holoenzyme reduction, UV/visible absorption spectra of samples before and after reduction were recorded between 250 and 500 nm and compared. This procedure yields the reduced, tagged form of Dsd (DsdR). 2.4. Removal of the N-terminal purification tag Removal of the purification tag endogenously encoded by the expression vector was performed by addition of thrombin from bovine plasma (final concentration 10 units/mg protein) to sample of DsdR, followed by dialysis of the mixture against 20 mM Tris/HCl (pH 7.9) containing 0.5 M NaCl and 20 mM ZnCl2. To overcome the low cleavage efficiency at low temperatures, the protease was allowed to act for no less than four days. Recovery of digested DsdR from the undigested fraction as well as free tags was achieved by subjecting the dialysate to a second round of nickel-affinity chromatography as described previously. The flow-through fractions, containing the tag-freed enzyme, were pooled and dialysed against PLP-free storage buffer. This procedure yields the reduced, detagged form of Dsd (DsdRD). 2.5. Activity assays The catalytic activity of Dsd, DsdR and DsdRD on D-serine was measured spectrophotometrically by coupling pyruvate production to the lactate dehydrogenase-mediated NADH oxidation [19]. All reactions were carried out on a Shimadzu UV-2401PC spectrophotometer in 1 mL of 50 mM Hepes/NaOH buffer (pH 8.0) containing 10 mM D-serine, 10 units of LDH and 0.3 mM NADH. PLP was either omitted or added to a final concentration of 20 mM. Prior to adding the enzyme (1.5 mg), the reaction mixture was let to equilibrate to 30  C for 5 min in the thermostated cuvette holder. Upon enzyme addition, the absorbance at 340 nm was recorded continuously for 5 min. Blanks lacking the different forms of the protein were recorded, so as to subtract the contribution of spontaneous NADH oxidation. 2.6. Preparation of samples for IR measurements The following buffers were prepared and used for IR measurements. Buffer (A): 50 mM Hepes/NaOH, 20 mM ZnCl2, 20 mM PLP, pH 7.9; buffer (B): 50 mM Hepes/NaOH, 20 mM ZnCl2, pH 7.9; buffer (C*): 50 mM Hepes/NaO2H, 20 mM ZnCl2, p2H 7.9; buffer (D*): 50 mM Hepes/NaO2H, 1 mM EDTA, p2H 7.9; buffer (E*): 50 mM Hepes/NaO2H, 20 mM ZnCl2, 10 mM D-serine, p2H 7.9; buffer (F*): 50 mM Hepes/NaO2H, 20 mM ZnCl2, 10 mM L-serine, p2H 7.9. The symbol (*) is used throughout this text to denote buffers prepared in 2H2O. For these buffers (C* through F*), the p2H corresponds to the pH-meter reading þ 0.4 [20]. Protein samples (1e1.5 mg) in storage buffer were concentrated to approximately 20e30 mL with Amicon Ultra 0.5-mL centrifugal filters (MWCO 30 kDa) for 1 h at 10,000  g on a refrigerated Allegra X-22R centrifuge (Beckman-Coulter). This was followed by addition of 250 mL of the desired buffer (A, B and C*eF*), and the protein solution was centrifuged again. This procedure was repeated several times so as to completely exchange the storage buffer for the desired buffer prepared in 1H2O or 2H2O. The

M. Baldassarre et al. / Biochimie 94 (2012) 479e486

481

concentrated protein solutions typically had volumes of 20e30 mL and a concentration ranging from 3 to 7% (w/v), the latter for measurements in 1H2O medium. 2.7. Acquisition of IR spectra The concentrated protein samples were injected in a Graseby Specac 20500 cell (Graseby Specac, Orpington, Kent, U.K.) fitted with CaF2 windows and 6-mm tin or 25-mm Mylar spacers for samples in 1H2O and 2H2O, respectively. FT-IR spectra were recorded by means of a Perkin Elmer 1760-X spectrometer equipped with a deuterated triglycine sulphate (DTGS) detector and using a BeereNorton apodization function. At least 24 h before and during data acquisition, the spectrometer was continuously purged with dry air at a dew point of 70  C produced by a Parker-Balston purge gas apparatus. The temperature of the cell was controlled by means of an external water bath circulator (HAAKE F3) and measured with a thermocouple placed directly onto one of the CaF2 windows. Spectra of buffers and samples were recorded at a resolution of 2 cm1 under the same scanning and temperature conditions. In thermal denaturation experiments, the temperature was raised by 5  C steps in the 20e95  C range. At every fixed temperature, the IR spectra of the sample were averaged over 32 repeated accumulations. 2.8. Spectra analysis All spectra were processed using the SPECTRUM application from PerkineElmer. Correct subtraction of H2O was judged to yield an approximately flat baseline between 1900 and 1400 cm1, whilst subtraction of 2H2O was adjusted to the removal of the 2H2O bending absorption around 1220 cm1 [21]. Second-derivative spectra were calculated over a 9 data-point range (9 cm1). The deconvoluted parameters were set with a gamma value of 2.5 and a smoothing length of 60 points. To determine the Tm of the protein samples under investigation, the width at 3/4 height (W3/4H) of the Amide I or Amide I0 bands was plotted against the temperature and the plot was fitted with the Boltzmann sigmoidal function using OriginPro software (OriginLab, Northampton, MA) [22]. 2.9. Bioinformatics tools Amino acid sequences were retrieved from the UniProt database (http://www.uniprot.org). Multiple sequences alignments were performed using the ClustalW2 function embedded in UniProt. Crystallographic structures were retrieved from the Brookhaven Protein Databank (http://www.rcsb.org/pdb). 3. Results 3.1. Expression and purification of recombinant Dsd Recombinant tagged Dsd (Dsd) was produced in and purified from E. coli cultures as described previously. Purification from the bacterial lysate was performed in a single chromatographic step by virtue of the nickel-chelating, poly-histidyl tag fused to the protein N-terminus. The eluted protein showed a high degree of purity by SDS-PAGE (data not shown), with an apparent molecular weight matching the one expected on the basis of its amino acid sequence (w50 kDa). The UV/visible absorption spectrum of freshly purified Dsd revealed the presence of a band near 410 nm (Fig. 1, solid line) indicating the presence of PLP bound to the protein via an aldimine linkage.

Fig. 1. UVeVisible absorption spectra of tagged Dsd before (solid line) and after (dashed line) treatment with NaBH4. Spectra were acquired at 20  C in 50 mM potassium phosphate buffer (pH 7.9) containing 20 mM ZnCl2. Panels A and B represent the chemical structures of the Schiff base (lmax ¼ 410 nm) and secondary amine (lmax ¼ 320 nm) forms of PLP, respectively.

3.2. Sequence analysis A BLAST search of the UniProtKB database using the sequence of yeast’s Dsd (Uniprot ID: P53095) as a query returned a number of proteins from fungi and moulds. None of them appears to have been characterized, with most sequences deriving from the in silico translation of microbial genomes. Cross-searching the literature, however, a number of low-scoring hits from the BLAST search appeared to have been extensively characterized over the last decade. These include: D-threo-3-hydroxyaspartate dehydratase from Delftia sp. HT23 (Dthadh, B2DFG5), D-serine dehydratase from chicken kidney (DsdCh, A9CP13), D-threonine aldolase from Arthrobacter sp. (Dta, O82872), amino acid aldolase/racemase from Idiomarina lohiensis (Aaar, Q5QX87) and D-serine deaminase from Burkholderia xenovorans (Dsa, Q145Q0). All these proteins have been shown to belong to the fold-type III of PLP-dependent enzymes. Among these, the structures of Aaar (PDB ID: 3LLX), Dsa (PDB ID: 3GWQ) and DsdCh (PDB ID: 3R0Z) have been solved experimentally. A multiple sequence alignment (Fig. 2) of the above proteins revealed that a number of conserved regions can be detected in spite of the low overall sequence identity (w30%). In our previous work, we have identified the speculative sites of interaction between PLP and a homology model of Dsd based solely on the crystallographic structure of Aaar. Extending the sequence analysis to similar sequences of proteins known to belong to the same fold-type, it appears that some of the residues previously identified are conserved in all the sequences under examination (Table 1). These include His55, Lys57 (supposed to be the site of aldimine formation), Arg171 and Gly286. Other residues are highly conserved, such as His205 (corresponding to Tyr in Dsda and Dta), Tyr210 (corresponding to Lys in Dsda and Gln in Dta) and Thr255 (corresponding to Glu in Dsda). His284 appears to be the least conserved. A second hotspot of conserved residues can be found in the C-terminal domain, around the expected sites of zinc coordination (His398 and Cys400). The catalytic activity of some of the above proteins has been correlated to the availability of divalent cations. For instance, the dehydratase activity of Dthadh depends on Mn2þ, Co2þ and Ni2þ [23], while the aldolase activity of Dta depends on Mn2þ [24]. The activity of Dsd from chicken depends on Zn2þ and is inhibited by EDTA treatment [25]. In this structure His347 and Cys349 coordinate the metal ion. These results support our

482

M. Baldassarre et al. / Biochimie 94 (2012) 479e486

Fig. 2. Multiple sequence alignment among Dsd and other members of the fold-type III of PLP-dependent enzymes. The amino acid sequence of Dsd from S. cerevisiae (Dsd) was aligned with those of D-threo-3-hydroxyaspartate dehydratase from Delftia sp. HT23 (Dthadh), D-serine dehydratase from chicken kidney (DsdCh), D-threonine aldolase from Arthrobacter sp. (Dta), amino acid aldolase/racemase from Idiomarina lohiensis (Aaar) and D-serine deaminase from Burkholderia xenovorans (Dsda). Residues common to at least half of the sequences are shaded black. Residues with similar properties are shaded grey. The symbols (;) and (B) refer to the site of PLP addition and zinc coordination, respectively.

previous conclusions that the residues mentioned above might be the sites of interaction with PLP and zinc in yeast’s Dsd. 3.3. Sodium borohydride treatment and activity assays Treatment with NaBH4 completely abolished the 410-nm absorption band displayed by Dsd. Reduction of Dsd was accompanied by the appearance of a new band around 320 nm (Fig. 1, dashed line). A blue shift of the cofactor absorption maximum of nearly 100 nm is typical of the Schiff base-form of PLP being reduced to its corresponding secondary amine. This results in irreversible inactivation of most PLP-dependent enzymes, since the aldimine carbon becomes inert towards attack by nucleophiles on the incoming substrate molecule. To assess whether this holds true for Dsd, the catalytic activities of Dsd, DsdR and DsdRD on D-serine were measured by the LDH-coupled spectrophotometric method [19]. In this assay, the rate of NADH oxidation is directly proportional to that of pyruvate formation. The activity of Dsd in the

Table 1 Aligned positions corresponding to the sites of PLP interactions. The sites of PLP interactions have either been observed in experimental structures (Dsda and Aaar) or are expected to bind PLP on the basis of sequence alignments (for the remaining sequences). The aligned positions are divided into identical and highly conserved residues. See Fig. 2 for the complete sequence alignment. Identical residues Dsd (S. cerevisiae) DsdCh (G. gallus) Dsda (B. xenovorans) Dthadh (Delftia) Aaar (I. lohiensis) Dta (Arthrobacter sp.)

H55 H43 H76 H41 H44 H57

K57 K45 K78 K43 K46 K59

R171 R143 R178 R141 R144 R157

Highly conserved residues G286 G241 G285 G236 G235 G251

H205 H176 Y209 H172 H171 Y187

Y210 Y181 K214 Y177 Y176 Q192

T255 T222 E259 T218 T217 T233

H284 H239 D283 R234 R233 Q249

presence of 20 mM PLP was considered as 100%. In the absence of exogenously added cofactor, a residual activity of 66% was measured indicating that about 30% of the enzyme lacks PLP as a consequence of the breakage of the aldimine linkage between enzyme and cofactor. When the enzyme is stored in the presence of PLP it is much more stable, but a long-term precipitation occurs, which may also be explained by the release of part of PLP causing protein instability [15]. When PLP was available in the reaction mixture, DsdR and DsdRD showed a residual activity of 20% and 32%, respectively. No activity could be detected with either samples when the medium was kept void of PLP. Taken together, these results suggest that the sodium borohydride reduction inactivates Dsd. On the other hand, the observation that the reduced protein still retains some activity in the presence of PLP, with no activity being detected in the absence of PLP, suggests that the reduced sample also contains the apo form of the enzyme. Loss of PLP could be due in part to decomposition of the cofactor during the storage, and more likely to a non complete reduction of the holoenzyme. In the latter case, loss of PLP could occur during the dialysis against PLPfree buffer.

3.4. Effect of aldimine reduction and tag removal on secondary structure and thermal stability In a previous work [15] we derived the secondary structure of Dsd by analysing the IR spectrum of the sample in H2O medium. Here we obtained the IR spectra of DsdR and DsdRD dissolved in H2O medium and compared them with the spectrum of Dsd. Dsd was analysed in the presence of PLP since it was previously shown that the cofactor is important for the enzyme stability [15]. On the

M. Baldassarre et al. / Biochimie 94 (2012) 479e486

other hand the DsdR and DsdRD samples do not require the presence of PLP since it is bound covalently to the protein. The IR absorption spectrum of Dsd in 50 mM Hepes/NaOH (pH 7.9) containing 20 mM ZnCl2 and in the presence of 20 mM PLP (buffer A) displays two composites centred around 1645 cm1 and 1550 cm1, corresponding to the Amide I and Amide II bands [26,27], respectively (Fig. 3, panel A). Deconvolution and secondderivative procedures allowed for the identification of the main bands making up the Amide I mode. Distinct signals were detected at 1684.9 cm1 (b-strands and/or turns), 1655.8 cm1 (a-helices þ unordered structures), 1641.2 cm1 and 1629.9 cm1 (b-strands), and a shoulder around 1674 cm1 (turns) (Fig. 3, panel C) [15]. Other significant signals in the observed spectral region arise from the Amide II mode (1550 cm1) [28] and the n(CC) ring vibration of tyrosyl residues (1515 cm1) [29]. The second-derivative spectra of DsdR and DsdRD in 50 mM Hepes/NaOH (pH 7.9) containing 20 mM ZnCl2 and in the absence of PLP (buffer B) feature a band pattern that is nearly identical to that of Dsd (Fig. 3, panels D and E, respectively). This indicates that neither the removal of the purification tag nor the reduction of the aldimine linkage between PLP and Dsd lead to detectable differences in the secondary structure composition of recombinant yeast’s D-serine dehydratase which was previously estimated to be 24% a-helix, 29% b-sheet and 47% unordered structures [15]. In addition, these data suggest that the

483

possible presence of the enzyme apo form in the samples does not affect the secondary structure of the protein significantly. Plotting the width of the Amide I band at 3/4 height as a function of the temperature allows to obtain thermal denaturation curves (Fig. 4) from which a Tm of 60.8  0.2  C for Dsd was calculated [22]. The Tm values for DsdR and DsdRD were 71.5  0.8  C and 73.3  0.6  C, respectively, indicating that the reduction of aldimine linkage markedly increases the protein thermostability and that the removal of the purification tag further stabilizes the protein. The possible presence of the apo form might lead to an underestimation of the real Tm values, which might be even higher than those reported above. In addition, in must be mentioned that the reduced forms of the protein remain stable towards long-term storage at 4  C without formation of precipitates which, instead, occur with Dsd. This observation suggests that the long-term instability of Dsd may be related to partial release of PLP. DsdR and DsdRD show also an enhanced stability in deuterated buffers with respect to the unreduced enzyme. It was postulated that the enhanced instability of Dsd in 2H2O leading to massive precipitate is due to hydrophobic interaction allowed by the high flexibility of the protein [15]. The covalent linkage between enzyme and PLP most likely decreases the flexibility of DsdR and DsdRD which, consequently, become more stable. 3.5. IR spectra of DsdRD in 2H2O Because of the enhanced stability of DsdRD in deuterated buffers, this protein was further analysed in 2H2O medium in order to obtain information on the (i) importance of zinc ion on the secondary structure and on the protein’s stability, (ii) binding capacity of DsdRD towards D-serine and L-serine. Briefly, Fig. 5 (panel A) shows the IR absorption spectrum of DsdRD in 50 mM Hepes/NaO2H, 20 mM ZnCl2, p2H 7.9 (buffer C*) between 1750 and 1500 cm1. The deconvoluted and second-derivative spectra (Fig. 5, panels B and C, respectively) point out the presence of five distinct signals within the Amide I0 band. The bands at 1687.3 and 1630 cm1 may be attributed to b-strands. The band 1653.0 cm1 arises from a-helices, while the bands at 1642 and 1675.4 cm1 are due to unordered structures and turns, respectively [26]. The band

Fig. 3. Infrared absorption spectra of Dsd, DsdR and DsdRD at 20  C. Dsd was analysed in 50 mM Hepes/NaOH, 20 mM ZnCl2, 20 mM PLP, pH 7.9 (buffer A). DsdR and DsdRD were analysed in 50 mM Hepes/NaOH, 20 mM ZnCl2, pH 7.9 (buffer B). Panels A, B and C refer to the absorption, deconvoluted and second-derivative spectra of Dsd. Panels D and E (dotted lines) show the second-derivative spectra of DsdR and DsdRD, respectively, superimposed to that of Dsd (solid line).

Fig. 4. Denaturation of Dsd, DsdR and DsdRD. Dsd was analysed in 50 mM Hepes/ NaOH, 20 mM ZnCl2, 20 mM PLP, pH 7.9 (buffer A). DsdR and DsdRD were analysed in 50 mM Hepes/NaOH, 20 mM ZnCl2, pH 7.9 (buffer B). Plots were obtained by measuring the Amide I bandwidth at 3/4 height at all temperatures values under investigation. The experimental data were fitted with Boltzmann’s sigmoidal function as described in Materials and Methods. Fits are shown as grey dashed lines. The R2 values are 0.9973, 0.9898 and 0.9939 for Dsd, DsdR and DsdRD fits, respectively.

484

M. Baldassarre et al. / Biochimie 94 (2012) 479e486

induced denaturation, thus resulting in enhanced thermal stability of the proteineligand complex with respect to the protein alone [31]. The Amide I bandwidth, peak position or peak intensity, measured as a function of the applied temperature make useful tools to assess ligand binding. Fig. 6 (panel A) shows the bandwidth at 3/4 height measured experimentally for DsdRD alone in the absence of amino acid (squares) in buffer (C*) and in the presence of 10 mM L-serine (triangles) or 10 mM D-serine (circles) in buffers (F*) and (E*), respectively. Fitting the raw data with Boltzmann’s function allowed for calculation of the thermal denaturation midpoint (Tm). DsdRD alone (squares) shows a Tm of 72.1  0.2  C, which is 1.2  C lower with respect the Tm calculated from the spectrum recorded in H2O medium (see Fig. 4 and corresponding results). This may be due to the lower protein concentration used for the analyses in 2H2O (3% w/v) with respect to the analysis in H2O medium (7% w/v). Addition of 10 mM L-serine (triangles) increases this value by 1.1  C (Tm ¼ 73.2  0.1). However, a more prominent increase is caused by addition of 10 mM D-serine (circles), resulting in a Tm value as high as 74.5  0.1  C. A similar trend was detected by monitoring the Amide I0 peak intensity (Fig. 6, panel B), with Tm values for DsdRD, DsdRD þ L-serine, and DsdRD þ D-serine of 69.6  0.4  C, 70.8  0.6  C and 71.6  0.5  C.

Fig. 5. IR spectra of DsdRD in deuterated medium at 20  C. Spectra were recorded in 50 mM Hepes/NaO2H, 20 mM ZnCl2 p2H 7.9 (buffer C*). The bands making up the Amide I0 are labelled according to their frequencies and secondary structure assignment (a, b, turns or unordered). Panels A, B and C represent the absorption, deconvoluted and second-derivative spectra of DsdRD in deuterated medium. Panels D and E (dashed lines) represent the second-derivative spectra of DsdRD in the presence of 10 mM D-serine and 10 mM L-serine, respectively. Panel F (dashed line) shows the second-derivative spectrum of DsdRD in the presence of 1 mM EDTA. The secondderivative spectrum of Dsd is shown as a control in panels DeF (solid line).

at 1546 cm1 represents the residual Amide II band, which provides useful information on the amount of hydrogen atoms retained by the protein structure following incubation in buffers prepared in 2 H2O [26,28]. The remaining signals below 1620 cm1 arise from the amino acid sidechain absorptions [29].

3.6. Metal depletion and binding of D-serine and L-serine Removal of the zinc ions by treatment with EDTA (buffer D*) did not affect the thermal stability of DsdRD (not shown) but exerted little effect on the secondary structure (Fig. 5, panel F) consisting only in a small decrease in the intensity of the a-helix signal. FT-IR experiments as simple as thermal denaturation analyses can provide useful information on the binding of ligands or substrates to the proteins or enzymes under investigation [30]. Since new bonds are formed, binding usually opposes to heat-

Fig. 6. Effect of D- and L-serine on the thermal stability of DsdRD. Amide I0 bandwidth at 3/4 height (A) and amide I0 peak intensity (B) of DsdRD. The protein was analysed in the absence (,) and in the presence of 10 mM D-serine (B) or 10 mM 2 L-serine (D), in 50 mM Hepes/NaOH, 20 mM ZnCl2 (p H 7.9). Plots were fitted as described in the text. Fits are shown as grey, dashed lines. Panel (A), the R2 values are 0.9971, 0.9989 and 0.9984 for DsdRD alone, DsdRD þ D-Ser and DsdRD þ L-Ser fits, respectively. Panel (B), the R2 values are 0.9970, 0.9963 and 0.9952 for DsdRD alone (,), DsdRD þ D-Ser (B) and DsdRD þ L-Ser (D) fits, respectively.

M. Baldassarre et al. / Biochimie 94 (2012) 479e486

These results suggest that reduced, detagged yeast D-serine dehydratase retains the ability to bind D-serine more strongly than the L-enantiomer, and are consistent with the higher binding specificity towards D-serine reported in literature [14]. As one can expect, neither D-serine nor L-serine are able to induce significant changes to the secondary structure of DsdRD at 20  C, since at this temperature the second-derivative spectra of the protein in the presence of either molecule are virtually superimposable to those of DsdRD alone (Fig. 5, panels D and E). 4. Discussion Alterations in D-serine levels have been measured in a number of brain disorders, such as Alzheimer’s disease [11], Amyothrophic lateral sclerosis [12] and schizophrenia [13]. Hence, the development of stable, protein-based platforms for the in vivo or ex vivo measurement of D-serine levels, as an alternative to the classical mass spectrometry [32,33] and column-switching HPLC methods [34,35], might be of valuable interest in healthcare and drug research. This requires, however, that the biological recognition element be accurately characterized from a structural and functional standpoint, with special emphasis on its stability and binding properties. In the light of these observations, FT-IR spectroscopy has been used to shed light on the structure, stability and binding properties of yeast D-serine dehydratase following reduction of the aldimine linkage between the enzyme and cofactor pyridoxal-50 phosphate. Reduction of the internal aldimine to the corresponding secondary amine did not alter qualitatively or quantitatively the secondary structure of the enzyme, nor did the removal of the N-terminal purification tag from the reduced enzyme. Interestingly, we found that reduction of the internal aldimine increased the thermal stability of the protein by nearly 11  C. In one of our previous works, pyridoxal-50 -phosphate was modelled within the active site of a Dsd model exploiting the structural information derived from the crystal structure of a bacterial PLP-dependent amino acid racemase, and the hypothetical interactions between sidechains and backbone with PLP were assessed [15]. The prediction suggested that PLP is able to bridge different secondary structure elements in the upper part of the TIM-barrel, with the pyridyl ring and the phosphate group accounting for most of the interactions with the protein scaffold. This feature is reactionindependent and is shared by all members in the fold-type III family of PLP-dependent enzymes whose structure has been experimentally determined, such as alanine racemase from Geobacillus stearothermophilus (PDB ID: 1BD0) and arginine decarboxylase from the Chlorella virus (PDB ID: 2NVA) [36,37]. In addition, the proposed sites of interaction correspond to positions that are highly conserved across multiple sequence alignments among Dsd and other members of the above family (Table 1 and Fig. 2), suggesting that they play an important role in coenzyme binding. As a general rule, the pyridyl ring lies between two aromatic residues with which it forms p-stacking interactions, while the phosphate group interacts with at least three H-bond donors. The reduction of the Schiff base-form of PLP causes the former aldimine carbon to lose its electrophilicity, thus yielding a stable, irreversible CeN bond that is inert towards nucleophilic attack by water or amino groups on substrates [18]. In the case of DsdR and DsdRD, however, what is lost in catalysis is gained in stability. Being permanently blocked within the active site of the enzyme, PLP might form an interaction pattern which exerts a stabilizing effect, allowing the structure of the reduced protein to endure higher temperatures, as is indeed shown by the FT-IR data. An even higher stability was achieved by enzymatically cleaving the purification tag, yielding a Tm as high as 73.3  C for reduced,

485

detagged Dsd. This suggests that the 20-mer purification tag contributes detrimentally to the stability of recombinant Dsd. The observation that tags often decrease protein stability, however, is not new. In a work by Ausili et al., 2D-IR correlation analysis has been used to demonstrate that a purification tag consisting of as few as 4 residues caused a significant decrease in the thermal stability of the b-glycosidase from the hyperthermophilic archaeon Sulfolobus solfataricus [38]. In the light of its peculiar properties, DsdRD has been the object of a thorough structural and functional characterization. The contribution of the zinc ion to the thermal stability was assessed. In the active enzyme, the metal ion has a functional role, promoting the removal of the b-hydroxyl group of D-serine. However, circular dichroism analyses carried out by Yoshimura and co-workers on the metal-bound and EDTA-treated enzyme suggested a structural role for zinc, since the CD signal at 220 nm of the cation-depleted form started to decrease at lower temperatures [14]. The result of our FT-IR denaturation analysis suggests that cation stripping by EDTA does not alter the thermal stability of the reduced, detagged form of Dsd. The homology model of Dsd previously assessed and discussed by us [15] shows that in spite of being coordinated by two residues (His398 and Cys400) in the b-rich domain, the metal ion is kept in close proximity (less than 6 Å) of PLP, at the mouth of the amino-terminal domain, but does not interact with the TIM-barrel or with PLP itself. Hence, either zinc does not have a structural role, in contrast to what affirmed by the Yoshimura group, or the stabilizing effect of the blocked cofactor might compensate for the decrease in stability brought about by loss of the metal ion. Hence, the decreased stability measured using the CD technique might stem from loss of PLP as a consequence of extensive dialysis against PLP-free buffers, as described in the literature for many PLPdependent enzymes [39]. Following reduction with sodium borohydride, the catalytic activity of Dsd on D-serine was completely abolished, although a little recovery was measured in the presence of exogenous PLP. This allowed us to record infrared spectra of the reduced and detagged sample (DsdRD) in the presence of D-serine without the risk of substrate consumption over time. The data indicate that DsdRD is able to bind partially, if not totally, D-serine. This finding stems from the observation of a positive shift of the Tm in the presence of D-serine. A lesser, though detectable, effect was also measured in the presence of an equal concentration of L-serine. Hence, FT-IR data indicate that reduced, detagged Dsd is able to bind D-serine more strongly/efficiently than L-serine. In virtue of its binding properties, DsdRD might represent an ideal candidate for the development of systems aimed at measuring the levels of D-serine in biological fluids. To this purpose, problems concerning of the low Km (0.39 mM) for this molecule must be overcome. Although the residues involved in substrate binding remain unknown, the availability of crystallographic data or more accurate models, in the next future, will allow for the straightforward design, via site directed mutagenesis, of mutants with increased affinity and selectivity. 5. Conclusions In our work, it is shown for the first time that a PLP-dependent enzyme can be greatly stabilized by irreversibly blocking the cofactor within the active site. This blocking stabilizes DsdR and DsdRD not only against high temperatures but also against long storage terms. In addition, this PLP-blocked form may be used as a binding protein since it retains the ability to bind its original substrate(s) without consuming them. In general, the reduction of PLP-dependent enzymes might provide significant advances in the development of protein-based platforms for the in vivo or ex vivo measurements of analytes of interest in biological fluids. Moreover,

486

M. Baldassarre et al. / Biochimie 94 (2012) 479e486

PLP-blocking might represent a useful trick to enhance the stability of challenging PLP-dependent enzymes, for example in growing quality crystals for X-ray analysis or more simply to speed up the crystallization process. Acknowledgements This work was supported by a grant from Università Politecnica delle Marche (FT). We wish to thank the laboratory for molecular sensing of Dr. Sabato D’Auria (IBP e CNR, Naples, Italy) for providing the fermentative resources and guidance necessary to produce and purify recombinant Dsd. References [1] N. Fujii, T. Saito, Homochirality and life, Chem. Rec. 4 (2004) 267e278. [2] L. Caglioti, O. Holczknecht, N. Fujii, C. Zucchi, G. Palyi, Astrobiology and biological chirality, Orig. Life Evol. Biosph. 36 (2006) 459e466. [3] M. Friedman, Origin, microbiology, nutrition, and pharmacology of D-amino acids, Chem. Biodivers. 7 (2010) 1491e1530. [4] A. D’Aniello, M.M. Di Fiore, G.H. Fisher, A. Milone, A. Seleni, S. D’Aniello, A.F. Perna, D. Ingrosso, Occurrence of D-aspartic acid and N-methyl-D-aspartic acid in rat neuroendocrine tissues and their role in the modulation of luteinizing hormone and growth hormone release, FASEB J. 14 (2000) 699e714. [5] H. Wolosker, A. D’Aniello, S.H. Snyder, D-Aspartate disposition in neuronal and endocrine tissues: ontogeny, biosynthesis and release, Neuroscience 100 (2000) 183e189. [6] S. Ishio, H. Yamada, M. Hayashi, S. Yatsushiro, T. Noumi, A. Yamaguchi, Y. Moriyama, D-Aspartate modulates melatonin synthesis in rat pinealocytes, Neurosci. Lett. 249 (1998) 143e146. [7] Y. Nagata, H. Homma, J.A. Lee, K. Imai, D-Aspartate stimulation of testosterone synthesis in rat Leydig cells, FEBS Lett. 444 (1999) 160e164. [8] H. Wolosker, R. Panizzutti, J. De Miranda, Neurobiology through the lookingglass: D-serine as a new glial-derived transmitter, Neurochem. Int. 41 (2002) 327e332. [9] C.R. Yang, K.A. Svensson, Allosteric modulation of NMDA receptor via elevation of brain glycine and D-serine: the therapeutic potentials for schizophrenia, Pharmacol. Ther. 120 (2008) 317e332. [10] S.A. Fuchs, R. Berger, L.W. Klomp, T.J. de Koning, D-Amino acids in the central nervous system in health and disease, Mol. Genet. Metab. 85 (2005) 168e180. [11] K. Hashimoto, T. Fukushima, E. Shimizu, S. Okada, N. Komatsu, N. Okamura, K. Koike, H. Koizumi, C. Kumakiri, K. Imai, M. Iyo, Possible role of D-serine in the pathophysiology of Alzheimer’s disease, Prog. Neuropsychopharmacol. Biol. Psychiatry 28 (2004) 385e388. [12] J. Sasabe, S. Aiso, Aberrant control of motoneuronal excitability in amyotrophic lateral sclerosis: excitatory glutamate/D-serine vs. inhibitory glycine/ gamma-aminobutanoic acid (GABA), Chem. Biodivers. 7 (2010) 1479e1490. [13] J.T. Kantrowitz, A.K. Malhotra, B. Cornblatt, G. Silipo, A. Balla, R.F. Suckow, C. D’Souza, J. Saksa, S.W. Woods, D.C. Javitt, High dose D-serine in the treatment of schizophrenia, Schizophr. Res. 121 (2010) 125e130. [14] T. Ito, H. Hemmi, K. Kataoka, Y. Mukai, T. Yoshimura, A novel zinc-dependent D-serine dehydratase from Saccharomyces cerevisiae, Biochem. J. 409 (2008) 399e406. [15] M. Baldassarre, A. Scire, I. Fiume, F. Tanfani, Insights into the structural properties of D-serine dehydratase from Saccharomyces cerevisiae: an FT-IR spectroscopic and in silico approach, Biochimie 93 (2011) 542e548. [16] N. Matsuda, H. Hayashi, S. Miyatake, T. Kuroiwa, H. Kagamiyama, Instability of the apo form of aromatic L-amino acid decarboxylase in vivo and in vitro: implications for the involvement of the flexible loop that covers the active site, J. Biochem. 135 (2004) 33e42. [17] M.R. Wilkins, E. Gasteiger, A. Bairoch, J.C. Sanchez, K.L. Williams, R.D. Appel, D.F. Hochstrasser, Protein identification and analysis tools in the ExPASy server, Methods Mol. Biol. 112 (1999) 531e552. [18] K.L. Williams, Y. Zhang, N. Shkriabai, R.G. Karki, M.C. Nicklaus, N. Kotrikadze, S. Hess, S.F. Le Grice, R. Craigie, V.K. Pathak, M. Kvaratskhelia, Mass spectrometric analysis of the HIV-1 integrase-pyridoxal 50 -phosphate complex reveals a new binding site for a nucleotide inhibitor, J. Biol. Chem. 280 (2005) 7949e7955.

[19] T. Ito, K. Takahashi, T. Naka, H. Hemmi, T. Yoshimura, Enzymatic assay of D-serine using D-serine dehydratase from Saccharomyces cerevisiae, Anal. Biochem. 371 (2007) 167e172. [20] P. Salomaa, L.L. Schaleger, F.A. Long, Solvent deuterium isotope effects on acidebase equilibria, J. Am. Chem. Soc. 86 (1964) 1e7. [21] F. Tanfani, T. Galeazzi, G. Curatola, E. Bertoli, G. Ferretti, Reduced beta-strand content in apoprotein B-100 in smaller and denser low-density lipoprotein subclasses as probed by Fourier-transform infrared spectroscopy, Biochem. J. 322 (1997) 765e769. [22] A. Scirè, E. Pedone, A. Ausili, M. Saviano, M. Baldassarre, E. Bertoli, S. Bartolucci, F. Tanfani, High hydrostatic pressure-induced conformational changes in protein disulfide oxidoreductase from the hyperthermophilic archaeon Pyrococcus furiosus. A Fourier-transform infrared spectroscopic study, Mol. Biosyst. 6 (2010) 2015e2022. [23] T. Maeda, Y. Takeda, T. Murakami, A. Yokota, M. Wada, Purification, characterization and amino acid sequence of a novel enzyme, D-threo-3hydroxyaspartate dehydratase, from Delftia sp. HT23, J. Biochem. 148 (2010) 705e712. [24] J.Q. Liu, T. Dairi, N. Itoh, M. Kataoka, S. Shimizu, H. Yamada, A novel metalactivated pyridoxal enzyme with a unique primary structure, low specificity D-threonine aldolase from Arthrobacter sp. Strain DK-38. Molecular cloning and cofactor characterization, J. Biol. Chem. 273 (1998) 16678e16685. [25] H. Tanaka, M. Senda, N. Venugopalan, A. Yamamoto, T. Senda, T. Ishida, K. Horiike, Crystal structure of a zinc-dependent D-serine dehydratase from chicken kidney, J. Biol. Chem. (2011). [26] J.L. Arrondo, A. Muga, J. Castresana, F.M. Goni, Quantitative studies of the structure of proteins in solution by Fourier-transform infrared spectroscopy, Prog. Biophys. Mol. Biol. 59 (1993) 23e56. [27] A. Barth, Infrared spectroscopy of proteins, Biochim. Biophys. Acta 1767 (2007) 1073e1101. [28] E. Pedone, M. Saviano, S. Bartolucci, M. Rossi, A. Ausili, A. Scire, E. Bertoli, F. Tanfani, Temperature-, SDS-, and pH-induced conformational changes in protein disulfide oxidoreductase from the archaeon Pyrococcus furiosus: a dynamic simulation and fourier transform infrared spectroscopic study, J. Proteome Res. 4 (2005) 1972e1980. [29] A. Barth, The infrared absorption of amino acid side chains, Prog. Biophys. Mol. Biol. 74 (2000) 141e173. [30] P. Herman, I. Barvik Jr., M. Staiano, A. Vitale, J. Vecer, M. Rossi, S. D’Auria, Temperature modulates binding specificity and affinity of the D-trehalose/Dmaltose-binding protein from the hyperthermophilic archaeon Thermococcus litoralis, Biochim. Biophys. Acta 1774 (2007) 540e544. [31] A. Scirè, A. Marabotti, M. Staiano, L. Iozzino, M.S. Luchansky, B.S. Der, J.D. Dattelbaum, F. Tanfani, S. D’Auria, Amino acid transport in thermophiles: characterization of an arginine-binding protein in Thermotoga maritima. 2. Molecular organization and structural stability, Mol. Biosyst. 6 (2010) 687e698. [32] H. Hasegawa, T. Matsukawa, Y. Shinohara, T. Hashimoto, Assessment of the metabolic chiral inversion of D-leucine in rat by gas chromatographyemass spectrometry combined with a stable isotope dilution analysis, Drug Metab. Dispos. 28 (2000) 920e924. [33] Y. Song, Y. Feng, X. Lu, S. Zhao, C.W. Liu, Y.M. Liu, D-Amino acids in rat brain measured by liquid chromatography/tandem mass spectrometry, Neurosci. Lett. 445 (2008) 53e57. [34] H. Bruckner, T. Westhauser, Chromatographic determination of L- and D-amino acids in plants, Amino Acids 24 (2003) 43e55. [35] A. Morikawa, K. Hamase, T. Inoue, R. Konno, K. Zaitsu, Alterations in D-amino acid levels in the brains of mice and rats after the administration of D-amino acids, Amino Acids 32 (2007) 13e20. [36] J.P. Shaw, G.A. Petsko, D. Ringe, Determination of the structure of alanine racemase from Bacillus stearothermophilus at 1.9-A resolution, Biochemistry 36 (1997) 1329e1342. [37] P. LeMagueres, H. Im, A. Dvorak, U. Strych, M. Benedik, K.L. Krause, Crystal structure at 1.45 A resolution of alanine racemase from a pathogenic bacterium, Pseudomonas aeruginosa, contains both internal and external aldimine forms, Biochemistry 42 (2003) 14752e14761. [38] A. Ausili, B. Cobucci-Ponzano, B. Di Lauro, R. D’Avino, A. Scire, M. Rossi, F. Tanfani, M. Moracci, Structural basis of the destabilization produced by an amino-terminal tag in the beta-glycosidase from the hyperthermophilic archeon Sulfolobus solfataricus, Biochimie 88 (2006) 807e817. [39] F. Malerba, A. Bellelli, A. Giorgi, F. Bossa, R. Contestabile, The mechanism of addition of pyridoxal 50 -phosphate to Escherichia coli apo-serine hydroxymethyltransferase, Biochem. J. 404 (2007) 477e485.