Báo cáo khoa học: Crystal structures of open and closed forms of D-serine deaminase from Salmonella typhimurium – implications on substrate specificity and catalysis pptx

Although holoenzyme was used for crystallization of both wild-type StDSD WtDSD and selenomethionine labelled StDSD SeMetDSD, significant electron density was not observed for the cofactor

deaminase from Salmonella typhimurium – implications on substrate specificity and catalysis

Sakshibeedu Rajegowda Bharath1, Shveta Bisht1, Handanhal Subbarao Savithri2and

Mattur Ramabhadrashastry Narasimha Murthy1

1 Molecular Biophysics Unit, Indian Institute of Science, Bangalore, India

2 Department of Biochemistry, Indian Institute of Science, Bangalore, India

Keywords

D -serine deaminase; open and closed

conformations; pyridoxal 5¢ phosphate

dependent Foldtype II enzyme; X-ray

diffraction; a, b elimination

Correspondence

M R N Murthy, Molecular Biophysics Unit,

Indian Institute of Science, Bangalore 560

012, India

Fax: +91 80 2360 0535

Tel: +91 80 2293 2458

E-mail: mrn@mbu.iisc.ernet.in

(Received 12 March 2011, revised 29 May

2011, accepted 7 June 2011)

doi:10.1111/j.1742-4658.2011.08210.x

Metabolism of D-amino acids is of considerable interest due to their key importance in cell structure and function Salmonella typhimurium D-serine deaminase (StDSD) is a pyridoxal 5¢ phosphate (PLP) dependent enzyme that catalyses degradation of D-Ser to pyruvate and ammonia The first crystal structure ofD-serine deaminase described here reveals a typical Foldtype II

or tryptophan synthase b subunit fold of PLP-dependent enzymes Although holoenzyme was used for crystallization of both wild-type StDSD (WtDSD) and selenomethionine labelled StDSD (SeMetDSD), significant electron density was not observed for the cofactor, indicating that the enzyme has a low affinity for the cofactor under crystallization conditions Interestingly, unexpected conformational differences were observed between the two struc-tures The WtDSD was in an open conformation while SeMetDSD, crystal-lized in the presence of isoserine, was in a closed conformation suggesting that the enzyme is likely to undergo conformational changes upon binding of substrate as observed in other Foldtype II PLP-dependent enzymes Electron density corresponding to a plausible sodium ion was found near the active site of the closed but not in the open state of the enzyme Examination of the active site and substrate modelling suggests that Thr166 may be involved in abstraction of proton from the Ca atom of the substrate Apart from the physiological reaction, StDSD catalyses a, b elimination ofD-Thr,D-Allothr and L-Ser to the corresponding a-keto acids and ammonia The structure of StDSD provides a molecular framework necessary for understanding differ-ences in the rate of reaction with these substrates

Introduction

Pyridoxal 5¢ phosphate (PLP) dependent enzymes

constitute a diverse family of proteins involved in the

metabolism of amino acids, amino sugars and amino

group containing lipids A majority of them are key

enzymes in the metabolism of amino acids The

reac-tions catalysed include the transfer of amino group,

decarboxylation, inter-conversion of l- and d-amino

acids and removal or replacement of chemical groups

at a, b or c positions [1] Functionally, PLP-dependent enzymes have been classified into three groups (a, b and c families) based on the carbon atom at which the net reaction takes place [2] Structurally, they have been classified into five groups [3–6]: Foldtype I enzymes that belong to the aspartate aminotransferase

Abbreviations

DNPH, 2,4-dinitrophenyl hydrazine; EcDSD, D -serine deaminase from Escherichia coli; LSD, L -serine dehydratase; PLP, pyridoxal

5¢-phosphate; SeMetDSD, selenomethionine D -serine deaminase; SpSR, Schizosaccharomyces pombe serine racemase; StDSD,

Salmonella typhimurium D -serine deaminase; TRPSb, tryptophan synthase b; WtDSD, wild-type D -serine deaminase.

family, Foldtype II that resemble tryptophan synthase

b, Foldtype III that are related to the alanine racemase

family, Foldtype IV enzymes related to the d-amino

acid aminotransferase family and Foldtype V or

glyco-gen phosphorylase family

Most of the amino acid dehydratases belong to the

tryptophan synthase b (TRPSb) family or Foldtype II

PLP-dependent enzymes [5] These enzymes catalyse

irreversible degradation of amino acids to the

respec-tive a-keto acids l-serine and l-threonine dehydratases

have been purified and characterized both structurally

and biochemically from several organisms [7–9]

Comparison of the amino acid sequence of d-serine

deaminase from Salmonella typhimurium (StDSD) with

those of other PLP-dependent enzymes suggests that it

belongs to the Foldtype II or TRPSb family, although

the sequence identities are low (15–23%) However,

there are d-serine deaminases (unrelated in sequence to

StDSD) which are annotated in sequence databases as

Foldtype III or as members of the alanine racemase

family

d-Serine deaminase from Escherichia coli (EcDSD;

EC 4.3.1.18) exhibits b-elimination activity with d-Ser,

d-Thr, d-Allothr and l-Ser with a pH optimum of 8.0

[10] In most bacteria, DSD probably acts as a

detoxi-fying enzyme, carrying out degradation of d-Ser Most

organisms fail to survive and propagate on d-Ser

containing nutrient media This is attributed to the

formation of d-Ser activated aminoacyl tRNA leading

to toxicity and retardation of cell growth d-Ser is a

co-agonist of NMDA channel receptors and therefore

EcDSD has been routinely included in the purification

of NMDA receptors from organotypic hippocampal

slices [11] The Saccharomyces cerevisiae DSD has been

used in diagnostic laboratories for quantitative

deter-mination of d-Ser in human brain and urine [12] The

E coli dsdA gene has been found to be an excellent

marker for construction of strains for which the use of

antibiotic resistance genes as selective markers is not

allowed [13,14]

EcDSD (48 kDa) [15] and Klebsiella DSD (46 kDa)

[16] are known to be functional as monomers in

contrast to the majority of Foldtype II PLP-dependent

enzymes, which are dimers However, DSD is a dimer

of 118 kDa in S cerevisiae [17] and a heterodimer of

40 and 40.4 kDa subunits in chicken [18] EcDSD has

been found to be activated by NHþ4 and K+and to a

lesser extent by Na+ions [10,19] It was proposed that

K+is not involved directly in catalysis but is required

for stabilizing the active site geometry [20] Although

crystals of EcDSD suitable for X-ray diffraction

studies have been obtained [21], its structure has not

been reported in the literature Comparative studies on

monomeric StDSD and the more common dimeric forms of Foldtype II PLP-dependent enzymes will allow examination of the plausible role of oligomeric state in these enzymes [22]

In this paper, we report the first crystal structure of

a Foldtype II d-serine deaminase and describe features

of the active site essential for catalysis The crystal structures reported are wild-type StDSD (WtDSD) and selenomethionine incorporated StDSD (SeMetDSD) crystallized in the presence of isoserine Although nei-ther of the structures had density corresponding to the cofactor, WtDSD was in an open conformation while SeMetDSD was in the closed conformation, suggesting that StDSD exhibits a domain movement similar to those of rat liver l-serine deaminase (rat liver LSD) [9] and serine racemase from Schizosaccharomyces pombe (SpSR) [23] Examination of the monomeric structure

of StDSD suggests that a dimeric structure similar to those of other Foldtype II PLP-dependent enzymes would lead to unacceptable van der Waals contacts involving segments of StDSD that are insertions with respect to other Foldtype II enzymes Electron density for a putative Na+ion was located close to the active site of SeMetDSD but not of WtDSD The active site geometry allows identification of residues that may play a key role in catalysis

Results and Discussion

Biochemical studies on StDSD Recombinant StDSD was expressed in E coli as a hexa-histidine tagged protein and purified by nickel nitrilotriacetic acid affinity and size exclusion chroma-tography The purified protein was yellow in colour with an absorbance maximum at 415 nm, indicating the presence of PLP as an internal aldimine As in several other PLP-dependent enzymes, a small peak at

340 nm was also observed The A280⁄ A415 ratio was close to 10 The peak at 415 nm was independent of

pH in the range 6.0–9.0 However, a small increase in the peak at 330–340 nm was observed close to pH 6.0 Similar observations have been reported for the EcDSD by Dupourque et al [10] The peaks at 415 and 330 nm have been attributed to the ketoenamine and enolimine forms, respectively, of the internal aldi-mine [24]

StDSD was most active with d-Ser The activities

of WtDSD (2.09 lmolÆmg)1Æmin)1) and SeMetDSD (2.12 lmolÆmg)1Æmin)1) with d-Ser as the substrate were comparable It has been reported that the pres-ence of Na+ or K+ ions enhances the activity of EcDSD [20,24] The enzymatic properties of StDSD

were therefore examined in the presence of Na+ or

K+ ions and the results are tabulated (Table 1) Km

and Vmax for d-Ser were about twice as high in the

presence of Na+ than with K+ In the presence of

Na+, Km for d-Ser was higher than that of d-Thr

However, Vmax was an order of magnitude higher for

d-Ser In contrast, in the presence of K+ions, Kmwas

lower and Vmax was higher for d-Ser than for d-Thr

The enzyme was much less active with d-Allothr and

l-Ser With both d-Thr and d-Allothr, the enzyme was

more active in the presence of K+than Na+

Structure and model quality

Crystal structures of SeMetDSD (1.9 A˚) and WtDSD

(2.4 A˚) were solved using four-wavelength anomalous

dispersion (4W-MAD) and molecular replacement

methods respectively The data collection and structure

refinement statistics are given Tables 2 and 3

respec-tively In WtDSD, except for two short stretches

(68–71 and 234–239) electron density is of good

quality throughout the polypeptide main chain In

SeMetDSD, electron density is absent for only two

C-terminal residues (439–440) A total of 17 and seven

residues have been truncated according to the extent of

observed electron density in WtDSD and SeMetDSD,

respectively In both structures, the residues forming

the C-terminal hexa-histidine tag were not included in

the model due to absence of a well-defined electron

density In SeMetDSD, 94.4% and 4.7% of residues

were in favoured and additionally allowed regions,

respectively, of the Ramachandran plot [25,26] One

residue (Ile111) was in the disallowed region The

WtDSD structure had 93.1% and 6.3% of the residues

in the favoured and additionally allowed regions and

two residues (Ile111 and His319) in the disallowed

region Statistics of the Ramachandran plot obtained

from procheck [27] are given in Table 3 Interestingly,

a well-defined density for PLP was not observed in the

two structures

The polypeptide fold of StDSD is illustrated in Fig 1 The secondary structural elements have been assigned using dssp [28] As in other PLP-dependent enzymes of Foldtype II, the StDSD monomer consists

of a small domain (residues 43–75, 109–238) and a large domain (residues 1–42, 76–108 and 239–440) The small domain folds as an open twisted a⁄ b structure consist-ing of a four-stranded (S4–S7) parallel b-sheet sand-wiched between one helix (H11) on the solvent facing side and two helices (H9 and H10) on the other side Four more helices (H4, H5, H7 and H8) occur in this domain The core of the large domain contains a seven-stranded mixed b-sheet surrounded by eight helices on the solvent facing side (H6, H17, H18, H19 and H20 on one side and H1, H2 and H3 on the other side) and six helices that occur between the two domains (H12, H13, H14, H15, H16 and H21) In the central b-sheet of the large domain, all except two short strands at the peri-phery (S1 and S2) are parallel The strands of the cen-tral b-sheet are strongly twisted The N-terminal helices H1 and H2 protrude away from the large domain StDSD structure was used as the template for identi-fying other proteins with similar folds in the Protein Data Bank (PDB) using the program dali [29] with the view of identifying shared and unique structural fea-tures of StDSD There were 245 hits with Z-scores higher than 20.0, of which 36 were unique structures The top hits corresponded to threonine deaminase, serine racemase and l-serine dehydratase Although the overall fold of StDSD is similar to those of the PLP-dependent Foldtype II enzymes, the helices H1, H2, H3 and H19 and the antiparallel b strand S1 in the large domain and the helices H5 and H8 in the small domain are significantly different and could be considered as additions to the fold of the TRPSb family These struc-tural segments are shown in orange in Fig 1

Gel filtration studies indicate that StDSD is a mono-mer in solution This is consistent with the earlier results obtained with EcDSD [10,15] However, it is in contrast to most of the enzymes belonging to the TRPSb family, which are dimers in solution (except for threonine synthase from E coli (PDB code 1VB3) and yeast (PDB code 1KL7), which are also mono-mers) Modelling of StDSD resembling the dimeric structures of other Foldtype II PLP-dependent enzymes such as O-acetylserine sulfhydrylase (OASS)

or cystathionine b-synthase suggests that H5, H6 and H21 prevent dimer formation by causing steric clashes

Comparison between WtDSD and SeMetDSD The rmsd upon superposition of corresponding

Ca atoms of WtDSD and SeMetDSD polypeptides is

Table 1 Kinetic parameters of StDSD with various substrates,

phosphate buffer pH 7.5 Vmaxis expressed as micromoles of

pyru-vate formed per milligram of protein per minute of reaction.

Substrate Buffer K m (m M ) V max

D -Ser Na + phosphate 0.87 ± 0.28 90.98 ± 13.12

K + phosphate 0.42 ± 0.08 54.56 ± 6.51

D -Thr Na+phosphate 0.45 ± 0.10 6.28 ± 1.31

K + phosphate 0.53 ± 0.09 13.83 ± 1.62

D -Allothr Na + phosphate 0.63 ± 0.08 0.40 ± 0.08

K+phosphate 0.88 ± 0.11 0.63 ± 0.09

L -Ser Na + phosphate 11.78 ± 2.53 0.32 ± 0.06

K + phosphate 10.23 ± 3.70 0.47 ± 0.11

1.32 A˚ Superposition of large and small domains of

StDSD with the corresponding domains of SeMetDSD

results in rmsd values of 0.53 and 1.64 A˚, respectively

A number of local conformational changes in the small domains of WtDSD and SeMetDSD are also observed These local changes lead to large rmsd in the compari-son of the small domains Superposition of the Ca atoms of the large domains (Fig 2A) in these struc-tures leaves the small domains with a residual rotation

of 15 Structures of WtDSD and SeMetDSD resemble the open and closed forms, respectively, of other Fold-type II PLP-dependent enzymes The solvent accessible surface areas of StDSD in the open and closed forms are 16 994 and 16 540 A˚2, respectively Ligand-induced movement of the small domain with respect to the large domain has been observed in SpSR (Fig 2B) [23], serine racemase from Rattus norvegicus and Homo sapiens [30], and OASS from S typhimurium [31] In tryptophan synthase b from E coli [32] and

l-serine dehydratase from Rattus norvegicus (rat liver LSD) [9], a similar domain movement is observed between apo and holo forms of the enzymes Consider-ing that SeMetDSD crystal was grown in the presence

of the inhibitor (isoserine), it is reasonable to assume that SeMetDSD represents the ligand bound closed form of the enzyme while the WtDSD represents the unliganded open form of the enzyme It is likely that, upon formation of external aldimine with isoserine, the enzyme undergoes conformational change to the closed

Table 2 Data collection statistics Values in parentheses refer to the highest resolution shell R merge = (R hkl R i |I i (hkl ) ) ÆI(hkl)æ|) ⁄ R hkl RIi(hkl ), where I i (hkl) is the intensity of the ith measurement of reflection (hkl) and ÆI(hkl )æ is its mean intensity R pim [46] = (R hkl [1 ⁄ N ) 1] 1 ⁄ 2 ) ÆI(hkl )æ|) ⁄ R hkl RIi(hkl ), where Ii(hkl ) is the intensity of the ith measurement of reflection (hkl ), ÆI(hkl )æ is its mean intensity and N is the number

of measurements (redundancy) I is the integrated intensity and r(I) is the estimated standard deviation of that intensity.

Crystal Se-Met Se-Met Se-Met Se-Met Se-Met Wt

Wavelength (A ˚ ) 0.97848

(peak)

0.97872 (inflection)

1.01876 (low remote)

0.97083 (high remote)

1.5418 (Cu Ka) 1.5418 (Cu Ka) Cell parameters (A ˚ )

a 56.26 56.37 56.41 56.38 56.46 100.02

b 187.65 187.83 187.94 187.88 188.39 46.79

c 46.48 46.53 46.54 46.57 46.59 100.04

a b c 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 90, 93.75, 90 Space group P21212 P21212 P21212 P21212 P21212 C2

Resolution range 45.1–2.2 (2.3–2.2) 37.6–2.0 (2.1–2.0) 41.9–2.1 (2.2–2.1) 35.3–1.8 (1.9–1.8) 48.4–1.9 (2.0–1.9) 49.9–2.4 (2.5–2.4)

R merge 0.065 (0.145) 0.055 (0.226) 0.049 (0.157) 0.069 (0.304) 0.053 (0.174) 0.109 (0.492)

Rpim 0.023 (0.045) 0.036 (0.146) 0.028 (0.096) 0.038 (0.153) 0.019 (0.058) 0.070 (0.300) Total measurements 367 430 (52 487) 143 786 (18 234) 129 720 (16 424) 262 803 (39 256) 329 613 (45 978) 72 766 (8757) Unique reflections 25 862 (3704) 34 178 (4867) 30 465 (4311) 46 887 (6704) 36 067 (4560) 17 806 (2301)

ÆI ⁄ r(I)æ 30.4 (18.4) 13.5 (4.4) 15.8 (6.0) 11.7 (3.6) 28.02 (12.0) 6.0 (2.3) Completeness 100 (100) 99.5 (98.7) 99.7 (99.0) 100 (100) 93.7 (82.5) 97.1 (87.8) Multiplicity 14.2 (14.2) 4.2 (3.7) 4.3 (3.8) 5.6 (5.9) 9.1 (10.1) 4.1 (3.8) Mosaicity 0.38 0.45 0.46 0.49 0.33 1.2

Wilson B-factor (A˚2 ) 21.9 22.8 23.9 21.6 14.7 37.4

Anomalous

completeness

100 (100) 90.9 (81.1) 92.7 (83.9) 99.6 (100) 91.0 (79.4) 92.9 (78.4) Anomalous

multiplicity

7.6 (7.4) 2.3 (2.2) 2.3 (2.2) 2.9 (3.0) 4.9 (5.3) 2.2 (2.1)

Table 3 Structure validation and refinement statistics.

Rwork= (Rhkl|Fo) F c |) ⁄ R hkl Fowhere Foand Fcare the observed and

calculated structure factors Rfree[47] is calculated as for Rworkbut

from a randomly selected subset of the data (5%), which were

excluded from the refinement.

SeMet Native

RMSD bond length (A ˚ ) 0.006 0.007

RMSD bond angle ( o ) 0.915 1.076

Ramachandran plot

Favoured region (%) 94.4 93.1

Additionally allowed region (%) 4.7 6.3

Generously allowed region (%) 0.6 0.0

Outliers (%) 0.3 0.5

Number of

Protein atoms 3307 3218

Water atoms 384 174

Non-water hetero-atoms 54 63

Average B-factor (A˚2 )

Protein atoms 14.3 27.5

Water atoms 26.9 29.9

Non-water hetero-atoms 22.4 50.6

form which is retained even after the removal of the

external aldimine under crystallization conditions or

from the crystals Thus, the WtDSD and SeMetDSD

structures can be viewed as open and closed forms of

the enzyme

In the WtDSD structure, two segments (68–71 and 234–239) were not built due to absence of significant electron density Of these, 68–71 stretch is involved in crystal contacts and is 27 A˚ away from the active site

In SeMetDSD, this stretch is not involved in crystal contacts and is ordered Therefore, the disorder observed in the WtDSD may be due to crystal pack-ing Residues 234–239 are close to the active site and are ordered in SeMetDSD suggesting that this segment may undergo disorder–order transition upon domain closure An aspartate residue (Asp236) occurring in this segment might have a role in catalysis (see Impli-cations for catalysis)

In the closed form of the structure (SeMetDSD), an isolated electron density close to Cys276 that could correspond to an Na+ ion was observed Although

Na+ions cannot be unambiguously distinguished from water molecules on the basis of electron density, bind-ing of an Na+ or K+ to EcDSD has been demon-strated through NMR chemical shift data [20] Also,

an equivalent site is known to bind divalent cations (Mg2+ or Mn2+) in SpSR [23], serine racemase from Rattus norvegicusand Homo sapiens [30] Therefore, an

Na+ ion was built into the observed density The refined B-factor of the Na+ (22.4 A˚2) is about twice that of the atoms in its close proximity Figure 3 shows the residues that interact with the proposed

Na+ion The charge on the Na+is neutralized by the carboxylate of Glu303 Apart from Ser307 hydroxyl and Cys309 sulfhydryl, the Na+ is surrounded by three main chain carbonyl groups A similar geometry has been observed around the bound K+ in rat liver LSD [9] In contrast, no density that could correspond

to a bound ion was present at the equivalent position

in WtDSD indicating that the ion binds only to the

Fig 1 Polypeptide fold of SeMetDSD showing secondary

struc-tural elements a-helices are shown as cylinders, b-strands are

shown as ribbons The two structural domains are coloured in

dif-ferent shades of teal and loops are shown in red PLP– D- Ser

com-plex at the active site is shown in ball and stick representation The

Na + ion is shown as a yellow sphere Secondary structural

ele-ments which are insertions with respect to most of the other

Fold-type II family of PLP-dependent enzymes are shown in orange All

secondary structural elements are labelled.

Fig 2 (A) Open and closed forms of

StDSD Large domains of WtDSD and

SeMetDSD were superposed to depict the

relative movement of the small domain

between the two conformational states.

Open conformation of WtDSD is shown in

dark grey while the closed conformation of

SeMetDSD is shown in light grey The small

domains are related by a residual rotation of

15 The view is selected to highlight

domain movement (B) Large domains of

open (dark grey; PDB code 1V71 ) and

closed (light grey; PDB code 2ZPU ) forms of

SpSR are superposed illustrating similarity

to the conformational change observed in

StDSD (A).

closed form of the enzyme Similar observation has

been made in the apo form of rat liver LSD [9] In

these enzymes, the metal ion does not appear to have

a catalytic role as it is not in direct contact with the

substrate

Active site

The active site of Foldtype II PLP-dependent enzymes

is situated in a large crevice between the two domains

Based on structural comparisons with other Foldtype II

PLP-dependent enzymes, PLP is expected to bind

StDSD as an internal aldimine covalently bonded to

the e-amino group of Lys116 situated at the beginning

of helix H7 However, significant density to fit an intact PLP was not observed in the electron density maps of either WtDSD (Fig 4A) or SeMetDSD (Fig 4B) suggesting that there is a tendency for PLP

to diffuse away from the active site It is known that EcDSD is readily converted to the apo form when incubated with l-Cys in the presence of EDTA [15] However, l-Cys and⁄ or EDTA were not present in purification or crystallization steps Also, the purified protein was yellow in colour and was catalytically active indicating that PLP is indeed bound However, the crystals were not yellow Enzymatic assay carried out with a dissolved SeMetDSD crystal showed low activity, which increased by a factor of 5 upon addition

Fig 3 (A) Cartoon diagram illustrating the residues of SeMetDSD that interact with Na+ion (B) Relative position of Na+ion with respect to the modelled PLP– D- Ser complex is shown Cys276 and Gly277, which are part of the glycine-rich loop (residues 276–282) anchoring phos-phate of the cofactor, are close to the Na + ion.

Fig 4 Electron density (2Fo) F c contoured

at 1r corresponding to 0.41 electrons A˚)3) observed at the active site of (A) WtDSD and (B) SeMetDSD The carboxyl group of modelled PLP– D -Ser is in the density corre-sponding to a bound ethylene glycol A blob

of density that represents a bound sulfate ion was observed at the site of phosphate

in both the structures although density was absent for the cofactor PLP in WtDSD and PLP– D -Ser external aldimine complex in SeMetDSD modelled by comparison with other Foldtype II enzymes are shown in ball and stick representation.

of PLP (Fig 5A) These results suggest that the

struc-tures described here most probably correspond to the

apo forms As the purified enzyme samples had

cova-lently bound PLP, it is likely that the cofactor was

lost during crystallization In order to examine this

possibility, 1 mgÆmL)1 WtDSD and SeMetDSD were

dialysed extensively against crystallization condition

Activities of these samples with d-Ser as substrate were

determined (Fig 5B) The samples had only 20% of

the activity (0.38 and 0.42 lmolÆmg)1Æmin)1in WtDSD

and SeMetDSD, respectively) of undialysed samples

suggesting that crystallization condition leads to loss

of PLP

SeMetDSD had a blob of significant density (Fig 4) at

a position corresponding to PLP of other Foldtype II

enzymes A sulfate ion was fitted to this density as it

was a component of crystallization The sulfate ion

could be refined to a B-factor (10 A˚2 in SeMetDSD

and 28 A˚2 in WtDSD) comparable with those of

the surrounding atoms (8–10 A˚2 in SeMetDSD and

17–26 A˚2 in WtDSD) It is worth noting that StDSD

is active in the presence of excess added sulfate and

hence absence of density for the PLP ring is unlikely

to be due to its displacement by sulfate

As adequate density was not observed for the cofactor, PLP (internal aldimine in WtDSD) and PLP-isoserine, PLP-d-Ser, PLP-d-Thr, PLP-d-Allothr and PLP-l-Ser (external aldimines in SeMetDSD) were modelled at the active site based on the structure of rat liver LSD [9] The close similarity in the active sites

of these enzymes ensured that the modelling is reliable

A significant density was present at the active site at a position corresponding to the carboxylate of the mod-elled substrate This density most probably corre-sponds to a bound ethylene glycol molecule The pyridine ring of PLP occupies a cavity between Leu338 and Ile115 These residues may limit the tilting of the pyridine ring between the internal and external aldi-mine forms [31] The side chain of Asn168 and the hydroxyl group of Thr422 are at hydrogen bonding distances from O3 and N1 of PLP (2.2 and 2.8 A˚), respectively Thr422 is found to be replaced by Ser (in threonine deaminase [8], O-acetylserine sulfhydrase [31] and serine racemase [23]) and Cys (in l-serine dehydra-tase [9]) in the other members of the TRPSb family of PLP-dependent enzymes It has been proposed that Ser

or Cys residue is important for maintaining the elec-tronic state of the PLP–Schiff base conjugate [33] Pre-sumably, Thr422 fulfils the same role in StDSD A glycine-rich loop is conserved in all the PLP-dependent enzymes and provides interactions for the binding of PLP (PLP-binding cup) [34] This loop located at the N-terminal end of helix H12 in StDSD consists of resi-dues Gly277, Val278, Gly279, Gly280, Gly281 and Pro282 Most of these residues are conserved in all known PLP-dependent enzymes The metal ion may stabilize the conformation of the PLP-binding loop by its interaction with carbonyl groups of Gly277 and Cys276

The carboxyl group of the substrate is held by hydrogen bonding to the Ser165 hydroxyl and amide group of Leu169 Comparison of modelled PLP-d-Ser and PLP-l-Ser complexes in SeMetDSD (Fig 6) shows that the Ca protons of the two external aldimine com-plexes are in opposite orientations The Ca proton in the case of PLP-d-Ser complex faces the hydroxyl group of Thr166 (2.5 A˚) whereas in PLP-l-Ser com-plex it faces Lys116 (4.5 A˚) and Asp236 (4.2 A˚) Based

on the modelled PLP–isoserine external aldimine com-plex, the hydroxyl group of Ser165 (2.7 A˚) and the amide groups of Thr166 (3.3 A˚) and Leu169 (2.6 A˚) appear to be important for stabilizing the external aldi-mine (Fig 7)

Earlier work on EcDSD has shown that modification

of a particular Cys residue leads to enzyme inactivation [15] This might correspond to Cys276 as it is conserved and occurs near the phosphate binding loop

Fig 5 (A) Activity of dissolved SeMetDSD crystals without (C1)

and with (C2) added PLP (50 l M ); 5 m M D -Ser was used as the

sub-strate (B) Activity of WtDSD and SeMetDSD under different

condi-tions as described in the text W1 and W2 correspond to WtDSD

while S1 and S2 correspond to SeMetDSD W1 and S1 correspond

to holo enzyme in 50 m M HEPES pH 7.5 W2 and S2 correspond to

proteins dialysed against crystallization condition.

Implications for catalysis

Degradation of d-Ser to pyruvate and ammonia by

StDSD involves two steps In the first step, d-Ser is

converted to aminoacrylate by Ca proton abstraction

and protonation of the hydroxyl group of the substrate

resulting in the release of a water molecule In a

subse-quent non-enzymatic step, aminoacrylate is converted

to ammonia and the a-keto acid, pyruvate

In rat liver LSD [9], it has been observed that the

N1 atom of PLP is unlikely to be protonated in view

of its hydrogen bonding to the side chain S–H of

Cys303 These authors also note that the PLP is likely

to be in its less polarized form (HPO4) as no cation is found in its vicinity and it is held in place only by backbone amide groups of the residues from the gly-cine-rich phosphate binding loop Based on these observations, they suggest that the Lys41 (which is linked to PLP in the internal aldimine form of the enzyme) may abstract the proton from the Ca atom of the substrate Elimination of the substrate hydroxyl may be facilitated by the phosphate group acting as a general acid The active site geometry of SpSR [23] is closely similar to that of rat liver LSD Here, N1 is

Fig 6 Stereodiagram of the superposition of active sites of SeMetDSD PLP- D -Ser (light grey) and PLP- L -Ser (dark grey) complexes The Ca proton of D -Ser points towards Thr166 in SeMetDSD and that of L -Ser points towards Lys116 and Asp236.

Fig 7 Stereodiagram of the active site geometry in WtDSD (dark grey) and SeMetDSD (light grey) PLP- D -Ser modelled in SeMetDSD is also shown The carboxyl group of modelled PLP- D -Ser is held by hydrogen bonding with the hydroxyl group of Ser165 and the main chain amide of Leu169 Lys116 is in different orientations in these structures Ser165 and Thr166 are closer to the substrate in SeMetDSD com-pared with WtDSD.

hydrogen bonded to the side chain of Ser308 and the

phosphate is held by the amide groups from the

glycine-rich phosphate binding loop A two-base

mech-anism in which Ser82 and Lys57 are involved in the

abstraction of a proton from Ca of d-Ser and l-Ser,

respectively, has been proposed for the racemase

reac-tion SpSR also exhibits a low level of a, b elimination

of d-Ser, for which Ser82 has been proposed as the

base in abstraction of the Ca proton

Examination of the active site geometry in StDSD

suggests that the active site Lys116, unlike in rat liver

LSD, is not at a position suitable for proton

abstrac-tion (Fig 6) Two residues, Thr166 and Tyr214, are

close to Ca of the modelled external aldimine (Fig 7)

and hence might be suitable for proton abstraction

Thr166 is part of the conserved loop which holds the

carboxyl group of the substrate and moves by 3.0 A˚

when StDSD undergoes transition from the open to

the closed form The modelled SeMetDSD–PLP-d-Ser

complex (Fig 7) suggests that the Ca proton of the

substrate is close to Thr166 hydroxyl (2.5 A˚) Thr166

is structurally equivalent to Ser82 of SpSR Therefore,

in StDSD Thr166 may fulfil the same role The side

chain hydroxyl of Tyr214 is at a distance of 6.3 A˚ (in

SeMetDSD) from the Ca atom of the substrate and is

disordered in WtDSD It undergoes substantial

dis-placement between the open and closed forms and

hence may be involved in proton abstraction, although

it appears to be a less likely candidate than Thr166

Further mutagenesis experiments need to be carried

out to clarify the role of these residues in catalysis As

in rat liver LSD [9], PLP may be involved in the

pro-tonation of the substrate hydroxyl group leading

to the release of a water molecule and formation of

aminoacrylate

It has been noted earlier that the amino group of

the incoming amino acid should be deprotonated to

make a nucleophilic attack on the C4¢ of PLP [23]

Occurrence of Tyr214 and Asp236 near the active

site of StDSD suggests that these residues might be

important for the initial formation of external

aldimine

StDSD exhibits substantial activity with d-Thr

Mode-lling of d-Thr as an external aldimine in SeMetDSD

shows no unacceptable contacts between the substrate

and protein atoms The lower rate of degradation with

d-Thr may be because of a lower rate of protonation

by the phosphate group StDSD has a low level of

activity with l-Ser Modelling l-Ser at the active site

suggests that the Ca proton points towards Lys116

and Asp236 and not towards Thr166 or Tyr214

Therefore, Lys166 or Asp236 may be involved in

deg-radation of l-Ser by StDSD

Based on fluorescence energy transfer and CD stud-ies in EcDSD [24], it was suggested that Trp197 (equivalent to StDSD Trp195) undergoes large dis-placement during catalysis and hence could be a key residue in catalysis However, Trp195 is not close to the active site and is unlikely to be important for the catalytic function

Conclusions

StDSD is a monomeric PLP-dependent enzyme that catalyses a, b elimination of d-Ser, d-Thr, d-Allothr and l-Ser to the corresponding keto acid and ammo-nia Structural data presented here suggest that StDSD protomer has a fold similar to those of other Foldtype II PLP-dependent enzymes and undergoes conforma-tional change from an open unliganded state to a closed liganded state It has a low affinity for the co-factor PLP under the conditions of crystallization An ion bound near the active site (most probably Na+) may be essential to keep the PLP binding loop in a conformation appropriate for cofactor binding and hence for catalysis The ion is unlikely to be directly involved in the enzyme reaction Differences in the cat-alytic rates with respect to different substrates (Table 1) in the presence of Na+and K+suggest that these ions affect the conformation of the PLP binding loop in subtle ways The positioning of Thr166 in these structures with respect to the substrate suggests that it

is suitable for abstraction of the proton from the Ca atom of the substrate Further structural and kinetic studies with site mutants of residues at the active site and determination of structures of ligand and inhibitor complexes will provide a deeper understanding of the catalytic mechanism of this Foldtype II PLP-dependent enzyme

Materials and methods

Cloning, overexpression and purification of StDSD

The dsdA gene from S typhimurium was amplified by PCR using the following gene-specific primers: StDSD-sense pri-mer CATATGGCTAGC ATG GAA AAC ATA CAA AAG CTC ATC; StDSD-antisense primer GGATCC TTA CTCGAG GCGTCC TTT TGC CAG GTA TTG The underlined bases correspond to restriction sites The sense primer had NdeI and NheI restriction sites, whereas the antisense primer had BamHI and XhoI sites The dsdA gene was cloned into pET21b between the NheI and XhoI sites The cloning strategy was such that the expressed protein had eight extra amino acids at the C-terminus

(LEHHHHHH) that included a hexa-histidine tag The

clone obtained was confirmed by sequencing The protein

was overexpressed in E coli BL21 (DE3) Rosetta cells The

cells were grown in LB medium containing 100 lgÆmL)1of

ampicillin at 37C until A600 reached 0.5 and were then

induced with 1.0 mm isopropyl thio-b-d-galactoside (IPTG)

and grown at 25C for a further 6 h The cells were

pel-leted by centrifugation at 4810 g for 10 min and

resus-pended in buffer A containing 50 mm Tris pH 8.0, 400 mm

NaCl, 30% glycerol and 50 lm PLP After sonication and

centrifugation, 1 mL of Ni-nitrilotriacetic acid beads were

added to 30 mL of the soluble fraction and kept for

end-to-end rotation for 3 h The unbound proteins from the

column were washed using buffer B containing 50 mm Tris

pH 8.0, 200 mm NaCl, 20% glycerol and 50 lm PLP

Non-specifically bound proteins were removed by a wash with

buffer B containing 20 mm imidazole In the last step, the

protein was eluted using buffer B containing 200 mm

imid-azole The eluted protein was concentrated to 1 mL using

centricon tubes, loaded onto a Sephacryl S-200 preparative

column for a final round of purification and eluted using

50 mm Hepes buffer pH 7.5 containing 100 mm NaCl The

purified protein, free of excess PLP was concentrated to

10 mg.mL)1in Centricon tubes and used for crystallization

The purified protein corresponded to a size of 49 kDa on a

12% SDS⁄ PAGE The molecular mass was confirmed by

MALDI-TOF

Selenomethionine incorporation

The plasmid containing dsdA gene was transformed into

BL21 (DE3) pLysS strain of E coli The cells were grown

in minimal medium Methionine biosynthesis was inhibited

by the addition of 50 mgÆL)1of Leu, Ile, Val, Lys, Thr and

Phe half an hour before induction with 1.0 mm IPTG

Sele-nomethionine was added at the time of induction

Purifica-tion of the enzyme was carried out following the same

protocol as used for the native enzyme, except that all

buf-fers contained 5 mm b-mercaptoethanol The

selenomethio-nine incorporation was confirmed by accurate mass

determination using ESI-MS

Biochemical studies

a-keto acids released from d-Ser, d-Thr, d-Allothr and

l-Ser by the enzymatic action of StDSD was estimated

using the 2,4-dinitrophenyl hydrazine (DNPH) method [17]

The reaction mixture for the assay with d-Ser consisted of

50 mm sodium or potassium phosphate buffer (pH 7.5),

20 lm PLP, varying concentrations of d-Ser and 50 ng of

StDSD in a final volume of 50 lL The reaction was

started by the addition of d-Ser and carried out at 37C

for 10 min Then 50 lL of 0.1% DNPH in 2 m HCl was

added to stop the reaction and the mixture was incubated

at 37C for 2 min, followed by the addition of 150 lL of

0.4 m NaOH After 5-min incubation at room temperature,

A540of the resultant hydrazone was measured The result-ing absorbance units corrected for enzyme-blank were plot-ted against substrate concentration The assay was again carried out with d-Thr, d-Allothr and l-Ser following a similar protocol Activity measurements were also carried out in the crystallization condition The kinetic parameters (Km and Vmax) were determined under two different condi-tions (sodium phosphate buffer pH 7.5 and potassium phosphate buffer pH 7.5) The amount of enzyme used for determining Km and Vmax was different with each of the substrates: 50 ng for the assay with d-Ser, 100 ng with

d-Thr, and 500 ng with d-Allothr and l-Ser The concen-trations of the substrates were varied between 0.2 and 9.5 mm Biochemical data were analysed using graphpad prism 5 (GraphPad software Inc, La Jolla, CA, USA) The oligomeric state of the purified protein in solution was determined using an analytical gel filtration column

Crystallization and data collection Crystallization trials were carried out with Hampton Crys-tal screens 1 and 2, PEG-ion screen, and Index screens 1 and 2 using microbatch (under oil) as well as sitting-drop vapour diffusion methods The best crystals of StDSD were obtained from 100 mm trisodium citrate, pH 6.1, containing

5 mgÆmL)1StDSD, 0.8 m lithium sulfate and 0.4 m ammo-nium sulfate in the hanging-drop vapour diffusion method Prior to crystallization, SeMetDSD (10 mgÆmL)1) in 50 mm Hepes buffer pH 7.5, 100 mm NaCl, was incubated with

40 mm dl-isoserine for about an hour at 4C and n-octyl-b-glucopyranoside was added to 0.1% Using this sample, crystals were obtained under the same condition as that of the WtDSD crystals The quality of these crystals was bet-ter than the WtDSD crystals or WtDSD crystals obtained

in the presence of isoserine

The crystals were mounted on a cryo-loop and frozen in liquid nitrogen for X-ray diffraction data collection Data

on a WtDSD crystal were collected to 2.4 A˚ resolution using

a Bruker AXS Microstar rotating anode X-ray generator and a MARRESEARCH image plate detector system The data were processed using mosflm and scala of the CCP4 suite [35] These crystals belonged to the space group C2 with unit cell parameters a = 100.02 A˚, b = 46.80 A˚,

c= 100.04 A˚ and b = 93.75 The crystal asymmetric unit was compatible with a monomer (solvent content 47.9%) Data on SeMetDSD crystals were collected at four different wavelengths, namely 0.97848 A˚ (peak), 0.97872 A˚ (inflection point), 0.97083 A˚ (high energy remote) and 1.01876 A˚ (low energy remote) at beamline 14 of ESRF, Grenoble, France The best of these data sets extended to 1.8 A˚ resolution The data were processed using mosflm and scala of the CCP4 suite The crystals belonged to the orthorhombic space group P21212 with a = 56.4 A˚, b = 188.4 A˚ and

c= 46.59 A˚ The asymmetric unit was compatible