Cyclization mechanism catalyzed by an ATP-grasp enzyme essential for d-cycloserine biosynthesis

Article abstract of DOI:10.1111/febs.15163

In the biosynthetic pathway of an antitubercular antibiotic d-cycloserine (d-CS), O-ureido-d-serine (d-OUS) is converted to d-CS. We have previously demonstrated that DcsG, classified into the ATP-grasp superfamily enzyme, catalyzes the ring formation to generate d-CS, which is accompanied by the cleavage of a bond in the urea moiety of d-OUS to remove a carbamoyl group. Although the general ATP-grasp enzymes catalyze an ATP-dependent ligation reaction between two substrates, DcsG catalyzes specifically the generation of an intramolecular covalent bond. In the present study, cyanate was found in the reaction mixture, suggesting that carbamoyl group is eliminated as an isocyanic acid during the reaction. By the crystallographic and mutational investigations of DcsG, we anticipate the residues necessary for the binding of d-OUS. An acylphosphate intermediate must be bound at the narrow pocket of DcsG in a folded conformation, inducing the bond cleavage and the new bond formation to generate cyanate and d-CS, respectively. Database: Structural data are available in Protein Data Bank database under the accession number 6JIL.

Full text of DOI:10.1111/febs.15163

DR YASUYUKI MATOBA (Orcid ID : 0000-0002-0366-0851)
Received Date : 04-Oct-2019
Revised Date : 20-Nov-2019
Accepted Date : 02-Dec-2019
Color : Fig 1
Cyclization mechanism catalyzed by an ATP-grasp enzyme essential for D-cycloserine
biosynthesis
Yasuyuki Matoba1, *, Narutoshi Uda2, Mako Kudo2, and Masanori Sugiyama2,*
1Faculty of Pharmacy, Yasuda Women's University, Yasuhigashi 6-13-1, Asaminami-ku,
Hiroshima 731-0153, Japan
2Graduate School of Biomedical & Health Sciences, Hiroshima University, Kasumi 1-2-3,
Minami-ku, Hiroshima 734-8551, Japan
Running title: Cyclization mechanism by DcsG
Key words: ATP-grasp superfamily; Biosynthesis; Catalytic mechanism; Crystal structure; D-
cycloserine.
This article has been accepted for publication and undergone full peer review but has not been
through the copyediting, typesetting, pagination and proofreading process, which may lead to
differences between this version and the Version of Record. Please cite this article as doi:
10.1111/FEBS.15163
This article is protected by copyright. All rights reserved

Correspondence: Y. Matoba (email: matoba@yasuda-u.ac.jp) or M. Sugiyama (email:
sugi@hiroshima-u.ac.jp).
Abbreviations: D-CS, D-cycloserine; D-OUS, O-ureido-D-serine; ORF, open reading frame; rms,
root mean square.
Database: Structural data are available in Protein Data Bank database under the accession number
6JIL.
Abstract
In the biosynthetic pathway of an anti-tubercular antibiotic D-cycloserine (D-CS), O-ureido-D-
serine (D-OUS) is converted to D-CS. We have previously demonstrated that DcsG, classified into
the ATP-grasp superfamily enzyme, catalyzes the ring formation to generate D-CS, which is
accompanied by the cleavage of a bond in the urea moiety of D-OUS to remove a carbamoyl group.
Although the general ATP-grasp enzymes catalyze an ATP-dependent ligation reaction between
two substrates, DcsG catalyzes specifically the generation of an intramolecular covalent bond. In
the present study, cyanate was found in the reaction mixture, suggesting that carbamoyl group is
eliminated as an isocyanic acid during the reaction. By the crystallographic and mutational
investigations of DcsG, we anticipate the residues necessary for the binding of D-OUS. An
acylphosphate intermediate must be bound at the narrow pocket of DcsG in a folded conformation,
inducing the bond cleavage and the new bond formation to generate cyanate and D-CS,
respectively.
Introduction
An anti-tubercular antibiotic, D-cycloserine (D-CS), is produced by several Streptomyces
species [1]. D-CS, a cyclic analog of D-alanine, inhibits catalytic activities of alanine racemase [2–
6] and D-alanyl-D-alanine ligase [7–10], which are necessary for the peptidoglycan biosynthesis of
the bacterial cell wall. In addition, D-CS is known to inhibit the activity of branched-chain
aminotransferase [11], a pyridoxal-5’-phosphate-dependent enzyme, like alanine racemase. The
antibiotic is clinically used as a second-line defense against Mycobacterium tuberculosis, which
has acquired resistance to other anti-tubercular antibiotics [12,13]. On the other hand, D-CS has
This article is protected by copyright. All rights reserved

been recently demonstrated to be a partial agonist for the N-methyl-D-aspartate receptor; therefore,
the drug may be useful for the treatment of various psychological dysfunctions [14–16].
In the early 1980s, a putative D-CS biosynthetic pathway was biochemically proposed by
other research groups based on substrate-feeding experiments [17–19]. We have recently cloned a
D-CS biosynthetic gene cluster from the chromosomal DNA of D-CS-producing Streptomyces
lavendulae ATCC11924 [20]. The gene cluster is composed of 10 open reading frames (ORF),
designated dcsA to dcsJ. Our group has demonstrated that both dcsI and dcsJ genes play a role in
the self-resistance for the D-CS producer [21,22]. From the sequence information of the ORFs
located in the D-CS-biosynthetic gene cluster, almost all the reactions, except the hydroxylation of
L-arginine, were deduced to be catalyzed by the enzymes encoded in the cluster, as presented in
Figure 1A [20]. More recently, we have shown that DcsA contributes to the hydroxylation of L-
arginine in vivo, although the hydroxylation reaction has not been confirmed in vitro [23]. In
addition, the functional analyses of DcsB, DcsC, DcsD, DcsE, and DcsG were performed to
confirm the proposed biosynthetic pathway of D-CS [20,24–26], and the tertiary structure of DcsD
and DcsE were determined [27,28].
An early study in the 1980s has suggested that O-ureido-D-serine (D-OUS) is generated from
L-serine and L-arginine as a final intermediate for D-CS biosynthesis [19]. The study has also
suggested that the synthesis of D-CS from D-OUS is dependent on ATP and Mg(II). For the D-CS
generation, two reactions, i.e. the cleavage of a bond in the urea moiety of D-OUS to remove a
carbamoyl group and the generation of a new covalent bond for the ring formation, are required.
When the radioisotope-labeled D-OUS, whose ureido carbon is replaced by 14C, was used as a
substrate, 14CO2 was generated with a rate similar to that of the D-CS formation [19]. The coupled
generation of CO2 suggested that a cyanate or a carbamate, which is further divided into CO2 (or
bicarbonate) and ammonium in solution, was released by the cleavage in the urea moiety during
the generation of D-CS. However, since the crude enzyme prepared from S. garyphalus was used
for the analysis [19], it remained to be elucidated whether the ring formation and the cleavage of a
bond in the urea moiety are catalyzed by an enzyme or separate enzymes.
Our previous study using the purified recombinant DcsG has demonstrated that the enzyme
catalyzes both reactions [25]. The study has also revealed that DcsG catalyzes the cyclization of
other D-amino acids. For example, D-homocysteine and β-aminooxy-D-alanine were converted by
DcsG to D-homocysteine thiolactone and D-CS, respectively. However, it remains unknown
whether the cleavage of a bond in the urea moiety occurs by hydrolysis, phosphorolysis, or
This article is protected by copyright. All rights reserved

elimination and whether the cleavage occurs prior to or following the formation of the
intramolecular covalent bond.
The amino acid sequence of DcsG has indicated that the enzyme is a member of the ATP-
grasp superfamily [29,30], which includes D-alanyl-D-alanine ligase and glutathione synthetase.
All the ATP-grasp enzymes catalyze an ATP- and Mg(II)-dependent ligation between a carboxyl
group of a substrate and an amino group or a thiol group of the other substrate (Fig. 1B) [29,30].
In the initial step of the reaction, a carboxylate-containing substrate is activated by ATP to form a
high-energy acylphosphate intermediate, together with the conversion of ATP to ADP.
Subsequently, an amino or thiol group of the second substrate nucleophilically targets the carbonyl
carbon of the intermediate, leading to the formation of a tetrahedral intermediate. Finally, an
amide or thioester bond is formed between the first and second substrates with the release of
inorganic phosphate. It should be noted that DcsG specifically catalyzes the bond formation within
one substrate in order to cyclize it.
Due to the sequence similarity of DcsG with the ATP-grasp superfamily enzymes, the ring-
formation reaction catalyzed by DcsG is suggested to proceed via the acylphosphate intermediate.
The D-OUS molecule may be directly phosphorylated or may be phosphorylated after the
conversion to β-aminooxy-D-alanine, in which carbamoyl group is removed. In fact, DcsG was
found to catalyze the generation of D-CS from β-aminooxy-D-alanine [25]. However, the kcat/Km
value of DcsG for D-OUS is 15-fold higher compared with that of β-aminooxy-D-alanine [25].
Especially, Km value for D-OUS is quite lower than that for β-aminooxy-D-alanine, indicating the
higher affinity of DcsG toward D-OUS. D-OUS is a more efficient substrate for DcsG compared
with β-aminooxy-D-alanine, suggesting that phosphorylated D-OUS is formed as an intermediate
prior to the cleavage of the bond in the urea moiety. If so, the phosphate and carbamoyl groups
must be removed from the phosphorylated D-OUS intermediate during the reaction of DcsG.
In the present study, we predicted the binding mode of D-OUS at the active site via
crystallographic and mutational investigations of DcsG. The substrate may be bound in a folded
conformation enhancing the cyclization reaction. In addition, we demonstrated that cyanate is
generated by the reaction. Cyanate is known to be mainly generated in vivo by the decomposition
of carbamoyl phosphate [31], which is generally used for the biosynthesis of L-arginine (or urea)
or pyrimidine nucleotide, in which the carbamoyl group is transferred to L-ornithine or L-aspartate,
respectively, by the enzymatic activity of the respective carbamoyltransferase [32–34]. However,
in the DcsG reaction to synthesize D-CS from D-OUS, the carbamoyl phosphate may not be
This article is protected by copyright. All rights reserved

generated. On the basis of the results obtained by the structural, mutational and enzymatic
analyses, we propose a catalytic mechanism underlying DcsG, which includes a step where the
intramolecular transfer of a proton from the carbamoyl group to the phosphate group initiates the
ligation reaction.
Results
Detection of cyanate
In the previous study, we have demonstrated that the stoichiometric ratio of synthesized D-CS
to ADP was approximately 1:1 when DcsG was incubated with D-OUS and ATP [25]. The present
study showed that cyanate is generated by the reaction. Kinetic parameters of DcsG for D-OUS
determined on the basis of the concentration of the generated cyanate (Km = 2.3 ± 0.4 mM, kcat
=
1.2 ± 0.1 s-1) are similar to those determined on the basis of the concentration of the generated
ADP (Km = 10 ± 1 mM, kcat = 2.2 ± 0.2 s-1) [25]. These results indicate that the two reactants (ATP
and D-OUS) were converted into ADP, phosphate, cyanate, and D-CS during the reaction of DcsG.
The cyanate was not observed in the absence of ATP (even in the presence of ADP and phosphate),
indicating that the carbamoyl group in D-OUS was removed after the formation of the
phosphorylated D-OUS intermediate. Cyanate is a toxic compound that can carbamoylate the
functional groups in the proteins [31]. In the D-CS-producing bacterium, the compound must be
immediately detoxified by the enzyme cyanase that can degrade it to CO2 and ammonia [35,36].
The early observation, where the generation of CO2 was coupled with the production of D-CS [19],
may be due to the cyanase activity in the crude enzyme solution. In addition, our previous study
indicated that an ORF encoding the putative cyanase is located near the D-CS biosynthetic gene
cluster [20].
When L-ornithine and L-ornithine carbamoyltransferase, which catalyzes the transfer of a
carbamoyl group from carbamoyl phosphate to L-ornithine, coexisted in the reaction mixture of
DcsG, L-citrulline was not generated. Considering that the half-life of carbamoyl phosphate is
about 5 min under the physiological conditions [37], carbamoyl phosphate is unlikely released to
the solution. In the case of the DcsG reaction, the cyanate may be generated not via carbamoyl
phosphate. Another possibility is that DcsG can catalyze the decomposition of carbamoyl
phosphate, which is intermediately formed during the reaction. In the present study, we found that
DcsG could not stimulate the formation of cyanate from carbamoyl phosphate at 25°C, where
This article is protected by copyright. All rights reserved

DcsG has a catalytic activity to synthesize D-CS, but natural decomposition of the carbamoyl
phosphate is inhibited, supporting that cyanate is generated not via carbamoyl phosphate.
Overall structure of DcsG
The crystals of DcsG in a complex with ADP and Mg(II) ions were grown when potassium
sodium L-tartrate was used as a precipitant. Since the crystals of DcsG did not grow under any
other conditions, we could not obtain the tertiary structure in a complex with the substrate D-OUS
or the product D-CS. In general, the amino acid sequence similarities among the ATP-grasp
enzymes are low [29,30]. Therefore, the structure of DcsG was determined by a single anomalous
dispersion method using the L-selenomethionine-labeled derivative (Table 1). An asymmetric unit
of the crystal contains four monomers. At a glance, DcsG seems to form a tetramer, where two
dimers, monomers of which is related by 4-fold axis, is related by 2-fold axis. However, there are
a few intermolecular interactions among the monomers. Buried surface area formed upon the
tetramer formation is 6,600 Å2, which is only 14% of the surface area of the tetramer. This result
agrees with our previous observation that DcsG exists as a monomer in solution [25]. The four
monomers have similar structures. The root mean square (rms) deviation value of the most
deviated pair was determined to be 0.38 Å for 297 Cα atoms.
The chain fold of DcsG, which is similar to other enzymes in the ATP-grasp superfamily
[29,30], consists of thirteen β-strands and ten α-helices that are arranged in three domains referred
to as N-terminal, central, and C-terminal domains (Fig. 2A). The N-terminal domain consists of
the residues of Gly2–Ile80 and has a central four-stranded parallel β-sheet (β2, β1, β3, and β4).
This β-sheet is flanked by helices α1–α3 in the domain. The central domain consists of residues
Pro107–Tyr177 and has a four-stranded antiparallel β-sheet (β5, β8, β6, and β7). This β-sheet is
flanked by helices α6 and α7 in the domain. The C-terminal domain consists of residues Val178–
Lys298 and has a five-stranded antiparallel β-sheet (β11, β10, β9, β12, and β13). This β-sheet is
flanked by helices α8 and α9 in the domain. The α4- and α5-helices are at the connecting region
between the N-terminal and central domains, but it is spatially close to the C-terminal domain. In
addition, the C-terminal α10-helix is spatially close to the N-terminal domain.
A structural similarity investigation of proteins by the DALI program [38] suggests that DcsG
is similar to ATP-grasp enzymes (Fig. 2). For example, DcsG can be superimposed on ArgX from
Thermus thermophilus (PDB ID 3VPB) [39] with an rms deviation of 2.8 Å for the 263 Cα atoms,
LysX from T. thermophilus (PDB ID 3VPD) [39] with rms deviation of 2.6 Å for the 261 Cα
This article is protected by copyright. All rights reserved

atoms, ribosomal protein S6 modification protein RimK from Escherichia coli (PDB ID 4IWX)
[40] with rms deviation of 3.1 Å for the 261 Cα atoms, glutathione synthetase from E. coli (PDB
ID 1GSA) [41] with rms deviation of 2.9 Å for the 263 Cα atoms (Fig. 2B), and D-alanyl-D-
alanine synthetase DdlB from E. coli (PDB ID 2DLN) [42] with rms deviation of 3.3 Å for the
263 Cα atoms (Fig. 2C). However, DcsG shares a low amino acid sequence identity with the
aforementioned five enzymes in the range of 13–17%.
The active site of the ATP-grasp enzymes contains pockets to accommodate ATP and the
substrates. The ATP-binding pocket has 2−3 binding sites for the Mg(II) ion, which aid binding of
ATP to the pocket. The ATP-grasp enzymes are known to adopt distinct conformations, including
open and closed. Binding of ATP (or ADP) and substrates into each binding site of the enzymes
induces the structural change from the open to the closed conformation, which is caused by
domain movements. As a result, the binding pockets for ATP (or ADP) and the substrates decrease
in size. When ATP (or ADP) and the substrates are bound to an enzyme at the same time, the
enzyme takes the most closed conformation. The crystals of DcsG were grown in the presence of
ADP, Mg(II), and L-tartrate. An active site of DcsG contains ADP and L-tartrate in addition to two
Mg(II) ions (Fig. 3). It is important to note that DcsG can be well superimposed on the closed
forms of ATP-grasp enzymes. It is likely that the binding of ADP and L-tartrate to the active site
of DcsG induced the conformational alteration to the closed form.
ATP-binding site
Like other ATP-grasp enzymes, the ADP molecule is positioned between the β-sheet in the
central domain and that in the C-terminal domain (Fig. 2A). The central domain acts as a flexible
lid in the absence of ATP (or ADP) that undergoes a conformational change and clamps down
over the active site upon nucleotide binding. The ADP molecule is also covered by a loop
connecting the β10- and β11-strands in the C-terminal domain. The loop in DcsG, which is longer
than the corresponding loop in the other ATP-grasp enzymes, locates near the central domain. In
particular, the carbonyl oxygen atoms of Leu206 and His213 in the loop form hydrogen bonds
with the side chains of Arg149 and Ser144 in the central domain, respectively.
The adenine ring of ADP is surrounded by the side chains of hydrophobic residues, including
Ile106, Val135, Tyr177, Val178, Val181, Leu206, Val258, and Leu268 (Fig. 4A). In addition,
hydrophilic interactions are formed with the adenine ring (Fig. 4A). The N1 nitrogen forms a
hydrogen bond with the main-chain nitrogen atom of Val178, the N6 amino group forms hydrogen
This article is protected by copyright. All rights reserved

bonds with the carbonyl oxygen atom of Pro176 and the Oε1 atom of Gln175, and the N7 nitrogen
forms a hydrogen bond with the amino group of Lys137. Regarding the ribose moiety of ADP, its
2’- and 3’-hydroxyl groups interact with the main-chain nitrogen atom of Met205 and the side-
chain carboxylate group of Glu186, respectively. With respect to the pyrophosphate moiety of
ADP, an α-phosphate oxygen forms a salt bridge with the side-chain amino group of Lys137, a β-
phosphate oxygen forms a salt bridge with the side-chain amino group of Lys92, and another β-
phosphate oxygen forms a hydrogen bond with the main-chain nitrogen atom of Ser144 (Fig. 4A).
In three out of four DcsG monomers determined in the present study, the positions of two
Mg(II) ions at the ATP-binding site are similar to those in the other ATP-grasp enzymes taking the
closed conformation. A magnesium ion (MgA) interacts with one of α-phosphate oxygens and one
of β-phosphate oxygens, whereas the other ion, MgB, interacts with another β-phosphate oxygen
(Figs. 4A & 4B). In addition, MgA interacts with one of two carboxylate oxygens of Glu269,
whereas MgB interacts with two carboxylate oxygens of Glu269 and one carboxylate oxygen of
Glu271. Furthermore, MgA binds with two water molecules, whereas MgB binds with one water
molecule. One of the MgA-ligating water molecules is hydrogen-bonded to the side chains of
Glu186 and Asp256, whereas the other is hydrogen-bonded to the 3’-hydroxyl group in ADP. On
the other hand, the MgB-linking water molecule is hydrogen-bonded to the backbone carbonyl
oxygen of Gly141.
In the other ATP-grasp enzymes, the residues corresponding to Asp256 and Glu269 of DcsG
form coordination bonds with MgA, whereas those corresponding to Glu269 and Glu271 form
coordination bonds with MgB (Figs. 4B, 4C & 4D). The Asp256 and Glu269 residues in DcsG are
conserved among the ATP-grasp enzymes, whereas the residue corresponding to the Glu271 of
DcsG is generally Asp or Asn. The distance between MgA and the carboxylate oxygen of Asp256
in DcsG is longer compared with the corresponding distance in the other ATP-grasp enzymes with
the closed conformation, and, as a result, Asp256 does not form a coordination bond with MgA
(Fig. 4B). The structural difference identified in DcsG may be due to the interaction between
Asp256 and Arg254 residues. In DcsG, the carboxylate group of Asp256 interacts with the
guanidino group of Arg254 with two hydrogen bonds. On the other hand, in the other ATP-grasp
enzymes, only one hydrogen bond is formed between the corresponding Arg and Asp residues,
and one carboxylate oxygen of the Asp residue, which does not interact with the Arg residue, can
be positioned near the MgA ion (Figs. 4C & 4D). During the catalytic reaction of DcsG, only one
This article is protected by copyright. All rights reserved

hydrogen bond may be formed between Asp256 and Arg254 residues, leading to the strong
interaction between Asp256 and MgA.
At the active site in another monomer, the distance between MgA and MgB is longer compared
with that in the aforementioned three monomers. Only one carboxylate oxygen of the Glu269
residue binds to MgB in this monomer, and distances between MgB and oxygen ligands are longer
compared with those in the other three monomers. This monomer appears to adopt the most
inactive structure among the four monomers determined in the present study.
Substrate-binding site
Although the chemical structure of L-tartrate is distinct from that of L-OUS, the L-tartrate may
enter the pocket as an analog of L-OUS or its phosphorylated form. However, since the L-tartrate
ion is not a true substrate for DcsG, the monomers of DcsG bound to L-tartrate adopt inactive
structures to a greater or lesser extent, as described above. The substrate-binding pocket is formed
between the N- and C-terminal domains, similar to the other ATP-grasp enzymes (Fig. 2). The
pocket is also covered by a loop connecting β6- and β7-strands in the central domain. In particular,
the orientation of the side chain of the Tyr143 residue in the loop region restricts the size of the
pocket (Fig. 4B).
Almost all the residues at the substrate-binding pocket of DcsG are not conserved among
ATP-grasp enzymes. The large structural variability of the substrate-binding pockets in the ATP-
grasp enzymes facilities the binding of the specific substrates and the suitable placements of the
nucleophile and acylphosphate intermediate for the ligation. The catalytic mechanism underlying
the ATP-grasp enzymes, which includes the binding modes of substrates or the acylphosphate
intermediate, has been anticipated from the crystal structures. For example, the catalytic
mechanism underlying the glutathione synthetase was anticipated from the crystal structure
complexed with a product (glutathione), ADP, and sulfate (Fig. 4C) [41], whereas the mechanism
underlying ArgX was anticipated from the crystal structure complexed with two substrates (LysW
and L-glutamate), ADP, and sulfate [39]. The positions of the sulfate ions are believed to be
similar to the position of the γ-phosphate group in ATP or the phosphate group in the
acylphosphate intermediate. In the case of DdlB, the crystal structure in a complex with ADP and
a tetrahedral intermediate analog, which imitates the compound formed immediately after the
binding of the second substrate to the acylphosphate intermediate, was determined (Fig. 4D) [42],
and the catalytic mechanism was predicted on the basis of this structure.
This article is protected by copyright. All rights reserved

Among three out of the four monomers of DcsG, the binding modes of L-tartrate to the
substrate-binding pocket are similar to each other. On the other hand, the binding mode in another
monomer, in which MgB is most distant from MgA among four monomers, varies from that in the
other three monomers, and appeared to be the most divergent from the binding pose of the true
substrate. Therefore, the binding mode of L-tartrate in the later monomer is excluded from further
discussions. When compared with the previously determined crystal structures, the position of the
1-carboxylate group in the L-tartrate bound to DcsG (Fig. 4B) is almost identical to that of the
phosphate group in the tetrahedral intermediate analog bound to DdlB (Fig. 4D) [42] and to that of
the sulfate ion bound to ArgX [39] or the glutathione synthetase (Fig. 4C) [41]. This indicates that
the γ-phosphate group in ATP or the phosphate group in the phosphorylated D-OUS intermediate
at the active site of DcsG may be bound to the binding site of the 1-carboxylate group of the L-
tartrate. In the crystal structure of DcsG, one of the two 1-carboxylate oxygens in the L-tartrate
binds to MgA, whereas the other binds to MgB (Figs. 4A & 4B). The later carboxylate oxygen
further forms a hydrogen bond with the main-chain nitrogen of Tyr143.
The putative substrate-binding pocket of DcsG is formed by Trp57, Cys142, Tyr143, Lys202,
Pro215, Arg220, Arg254, Glu271, and Ser276 (Fig. 4B). In these residues, the hydroxyl group of
Ser276 forms a hydrogen bond with a guanidino group of Arg220. Of note, the size of the
substrate-binding pocket in DcsG seems to be smaller compared with that in the other ATP-grasp
enzymes complexed with the substrate or its analog, suggesting that the pocket in DcsG
accommodates only one substrate.
The Glu271 residue is only one acidic residue at the substrate-binding pocket and one of the
two carboxylate oxygens binds to MgB (Fig. 4B). The other carboxylate oxygen, which forms a
hydrogen bond with the 3-hydroxyl group in the L-tartrate, may be important for the recognition of
the amino group in the substrate D-OUS. Among three basic residues (Lys202, Arg220, and
Arg254) at the substrate-binding pocket of DcsG, Arg254 is closest to Glu271 (Fig. 4B). DdlB has
an Arg255 residue at the corresponding position of Arg254 in DcsG (Fig. 4D) [42]. Furthermore,
the position of the guanidino group of the Arg254 residue in DcsG is occupied by the guanidino
group in glutathione synthetase (Arg210 in Fig. 4C) [41] and ArgX [39], although the positions of
the main-chain sections of the Arg residues vary. The Arg residues in DdlB, glutathione synthetase,
and ArgX are commonly thought to serve an important role in recognition of the carboxylate
group of the first substrate. Therefore, the Arg254 of DcsG may be used for the recognition of the
carboxylate group in D-OUS. The other basic residues, Lys202 and Arg220, interact with the 4-
This article is protected by copyright. All rights reserved

carboxylate group in the L-tartrate (Fig. 4B). In addition, the amino group of Lys202 is close to
one of the 1-carboxylate group oxygens in the L-tartrate that ligates to MgA.
Mutational analysis
The kcat/Km values of DcsG for D-OUS, β-aminooxy-D-alanine, and D-homocysteine at pH 8.0
were 210 ± 30, 14 ± 1, and 45 ± 9 M-1 s-1, respectively [25]. It is important to note that the
carbamoyl group is released only when DcsG reacts with the most suitable substrate D-OUS,
suggesting that the group plays a role in facilitating the reaction.
To elucidate the substrate-binding manner of DcsG, each of the residues at the substrate-
binding pocket (Trp57, Cys142, Tyr143, Lys202, Pro215, Arg220, Arg254, Glu271, or Ser276)
were replaced by Ala, and the activities of each mutant toward D-OUS, β-aminooxy-D-alanine,
and D-homocysteine were evaluated. The Glu271 residue was also replaced with Gln to evaluate
the significance of the negative charge in the residue. As a result, all the mutations were revealed
to reduce the activity of the three substrates investigated (Fig. 5), although the activity ratios
varied between 0 and 77%. This observation indicated that all the residues were important for the
binding of these compounds in suitable conformations for cyclization. In particular, Cys142 was
demonstrated to be important for the cyclization of β-aminooxy-D-alanine or D-homocysteine and
less important for D-OUS. On the other hand, although mutations at Arg220 and Ser276 markedly
reduced the activities for D-OUS (1.6 and 7.0% of the activity of the wild type, respectively), the
reducing ratio was decreased in the cases using β-aminooxy-D-alanine or D-homocysteine as a
substrate.
Discussion
In general, the ATP-grasp enzymes catalyze an ATP-dependent ligation reaction between two
substrates [29,30]. To the best of our knowledge, DcsG is the first example that ATP-grasp
enzyme catalyzes the formation of intra-molecular covalent bond. The conversion of D-OUS to D-
CS includes the reactions of the formation of a new amide bond and the cleavage of a bond present
within the urea moiety. In the previous study, although we have demonstrated that DcsG catalyzes
both reactions [25], we have not evaluated how the carbamoyl group is removed. In the present
study, we clearly showed that cyanate is generated by the reaction.
This article is protected by copyright. All rights reserved

Like the other ATP-grasp superfamily enzymes, the ligation reaction of DcsG may proceed
via the formation of the phosphorylated intermediate. Since D-OUS is a more efficient substrate
for DcsG compared with β-aminooxy-D-alanine [25], the phosphorylated D-OUS may be formed
as an intermediate prior to the cleavage of the bond in the urea moiety. The crystallographic
analysis of DcsG suggests that the phosphorylated D-OUS intermediate must be accommodated
into the narrow pocket of DcsG in a folded conformation, in which the carbamoyl and phosphate
groups are nearly positioned.
There are two possible mechanisms of DcsG for the cyanate generation from the
phosphorylated D-OUS (Fig. 6): the first one is that the phosphate group initially attacks the
carbamoyl carbon atom nucleophilically to generate a carbamoyl phosphate (Fig. 6A). A similar
step is included in the phosphorylase reaction catalyzed by L-ornithine or L-aspartate
carbamoyltransferasae to generate carbamoyl phosphate by using carbamoylated substrate and
inorganic phosphate [43,44]. Then, isocyanic acid, which is further converted into cyanate, is
generated by the decomposition of the carbamoyl phosphate under the neutral pH conditions [37].
However, we could not detect the presence of carbamoyl phosphate in the reaction mixture of
DcsG, suggesting that cyanate is formed not via carbamoyl phosphate.
With regard to the second possibility, like the decomposition of the carbamoyl phosphate, the
phosphate group in the acylphosphate intermediate removes initially the proton from the
carbamoyl group, leading to the direct generation of isocyanic acid (Fig. 6B). At this moment, this
is a more hopeful catalytic mechanism of DcsG. From our experiment incubating DcsG and
carbamoyl phosphate, it is suggested that the enzyme does not stimulate the decomposition of
carbamoyl phosphate. The positional relationship of the phosphate and carbamoyl groups in the
phosphorylated D-OUS bound to DcsG may be similar to that in the carbamoyl phosphate
decomposed in solution. However, DcsG may be unable to bind carbamoyl phosphate in a
conformation suitable for the elimination of the carbamoyl group.
We propose a reaction mechanism underlying DcsG (Fig. 7). An ordered reaction mechanism,
in which ATP is bound first and ADP is released last, is generally accepted for the ATP-grasp
enzymes owing to the shape of the pocket at the catalytic site (Fig. 7A). Following ATP binding to
DcsG, the D-OUS molecule is accommodated into the substrate-binding pocket, in which the
amino and carboxylate groups of D-OUS interact with Glu271 and Arg254, respectively (Fig. 7B).
This concept is consistent with the results that all the E271A, E271Q, and R254A mutations
This article is protected by copyright. All rights reserved

markedly reduced the activity of the three substrates analyzed (Fig. 5). Owing to the shape of the
substrate-binding pocket, L-amino acids are unlikely to be accommodated in the pocket of DcsG.
The crystal structure of DcsG indicates that the substrate-binding pocket accommodates only
one substrate (Fig. 4B). To accommodate D-OUS in the narrow pocket of DcsG, the substrate
must adopt a folded conformation (Fig. 7B), which may be unfavorable due to the van der Waals
violation. However, the narrow pocket provides an advantage for enhancing the formation of the
internal bond to generate D-CS. In particular, the Nδ atom approaches the carbonyl carbon to be
attacked.
Based on the active-site architecture of DcsG (Fig. 4B) and the results of the mutation
analysis (Fig. 5), the hydrophobic side chains of Trp57, Tyr143, and Pro215 may be important to
restrict the size of the pocket (Fig. 7B), although Tyr143 may be less important compared with the
others. The C142A mutation reduced the activity toward β-aminooxy-D-alanine or D-
homocysteine in the range of 30 and 40%, whereas the mutation only slightly reduced the activity
toward D-OUS (77%). The side chain of β-aminooxy-D-alanine or D-homocysteine, which is
smaller compared with that of D-OUS, may be occasionally accommodated into the vacant space
generated by the C142A mutation, leading to the reduced activity of the mutant toward β-
aminooxy-D-alanine or D-homocysteine.
Basic residue Arg220 may also play a role to reduce the pocket size (Fig. 4B). In addition, the
residue may form hydrophilic interactions with the side chain of D-OUS (Fig. 7B). We have
previously revealed that DcsG dislikes the substrate with a positive charge at the side chain [25].
The positive charge at the side chain of Arg220 may prevent the binding of the positively charged
substrates. The activity of R220A and S276A mutants toward D-OUS was significantly decreased
(Fig. 5). The orientation of the Arg220 residue may be adjusted by the hydrogen bond with Ser276
to achieve the binding of D-OUS in a conformation suitable for the cyclization reaction. On the
other hand, the activities toward β-aminooxy-D-alanine and D-homocysteine were not significantly
reduced by mutations at the Arg220 and Ser276 residues (Fig. 5). On the basis of the substrate-
binding model fitted manually, the guanidino group of Arg220 interacts with the Oζ atom of D-
OUS (Fig. 7B).
Subsequently, the γ-phosphate group of ATP is transferred to the carboxylate group of D-OUS,
resulting in the formations of ADP and the phosphorylated substrate (Figs. 6B & 7B). In the
proposed structure of the phosphorylated D-OUS, the carbamoyl and phosphate groups are close to
one another. Specifically, the phosphate group appears to be close to the Nδ and Nζ atoms in the
This article is protected by copyright. All rights reserved

side chain of D-OUS. The disadvantage for this intermediate to take the folded conformation
would be partially offset by forming the intramolecular hydrogen bonds.
With regard to the Lys202 residue, the K202A mutation reduced the activities for three
substrates analyzed to the same level (Fig. 5). The amino group of the Lys202 residue may form a
hydrogen bond with one of the γ-phosphate oxygens in ATP or one of the phosphate group
oxygens in the phosphorylated substrates, which ligates to MgA (Fig. 7B). Therefore, the Lys202
residue would be important to generate the acylphosphate intermediate and help it achieve a
suitable orientation for the cyclization reaction.
The present study strongly suggests that the carbamoyl group of D-OUS is eliminated as an
isocyanic acid, which is further converted into cyanate (Fig. 6B). Under the neutral pH conditions,
elimination of cyanate from carbamoyl phosphate is provoked by the proton transfer from the
amide group to the phosphate one [37]. The reaction is progressed, since the other product is a
stable mono-hydrogen phosphate ion. If the same reaction occurred after the formation of the
phosphorylated D-OUS, the product would have a deprotonated amino group. Such intermediate
would not be favored, therefore, the elimination of cyanate may be coupled with the nucleophilic
attack of the Nδ atom to the carbonyl carbon, resulting in the formation of the tetrahedral
intermediate (Fig. 6B). The generated oxy anion may be stabilized with the positive charge from
the Arg254 residue.
The alternative substrates, including D-homocysteine and β-aminooxy-D-alanine, are also
likely accommodated in the narrow pocket of DcsG with a folded conformation. The catalytic
mechanism, in which the proton is transferred from the side chain of the substrate to the bound
phosphate, may be in common among D-OUS and the alternative substrates. However, the affinity
between DcsG and the alternative substrates may be decreased by the absence of the interaction
with the Arg220 residue, which seems to form a hydrogen bond with the Oζ atoms in the side
chain of D-OUS (Fig. 7B).
The tetrahedral intermediate would be finally separated into D-CS and inorganic phosphate
(Fig. 6B). At this stage, the active site of DcsG accommodates ADP, D-CS, cyanate, and
phosphate. D-CS is positioned closer to the entrance of the active-site pocket compared with the
others. Therefore, D-CS may be released first from the active site, followed by the releases of
cyanate and phosphate ions into the solvent in this order (Fig. 7A). Finally, ADP is replaced by
ATP to catalyze the next reaction.
This article is protected by copyright. All rights reserved

Conclusion
Novel enzymes catalyzing the unprecedented reactions have been often discovered from the
biosynthetic pathway of the natural products. The cyclization reaction contributes to restrict the
conformational freedom of the compound and thereby to confer the biological activities. In general,
the cyclization of typical natural products (non-ribosomal peptides or polyketides) is catalyzed by
the thioesterase domain within the large multidomain enzyme. However, there are some reports, in
which the cyclization is catalyzed by a different enzyme [45,46]. With respect to the ribosome-
derived bacteriocins, cyclic thioester cross-link is introduced by a zinc-binding enzyme (for
example, NisC for the nisin synthesis) [47]. To understand the reaction mechanisms of these
enzymes is useful to generate the artificial macromolecules containing the cyclic structures. On the
other hand, there are some enzymes which catalyze the generation of small circles in the natural
products (for example, isopenicillin N synthase) [48].
DcsG is the first example that the ATP-grasp enzyme catalyzes the formation of intra-
molecular covalent bond, being significant for the researchers studying the natural products or the
enzymes. Through the crystallographic and mutational analyses of DcsG, we anticipate the
residues necessary for the binding of D-OUS. In addition, we detected cyanate in the reaction
mixture, suggesting that carbamoyl group of D-OUS is eliminated as an isocyanic acid
immediately before the ring formation. These results make it possible to explain the dual functions
of DcsG (bond cleavage and ring formation). The catalytic mechanism of DcsG may include the
particular process where the intra-molecular transfer of a proton from the carbamoyl group to the
phosphate group initiates the ligation reaction. The result may be useful for the researcher which
aims to create a new enzymatic catalyst to generate cyclic amino acids.
Materials and Methods
Reagents. Almost all the reagents used in the present study were of analytical grade. D-OUS and
β-aminooxy-D-alanine were chemically synthesized from D-CS, according to methods described
previously [49,50].
Expression and purification of DcsG. The pET-21a(+) plasmid carrying dcsG [25] was used to
express DcsG with His6-tag at the C-terminus. To obtain the recombinant DcsG protein, E. coli
BL21(DE3) transformed with the plasmid was grown in an LB medium. To overproduce L-
selenomethionine-labeled DcsG, the plasmid was introduced into E. coli B834(DE3). The cells
This article is protected by copyright. All rights reserved

were grown at 28ºC in Overnight Express Autoinduction System 2 medium (Novagen, Tokyo,
Japan) supplemented with 100 nM vitamin B12 and 125 μg mL-1 L-selenomethionine. The
expression and purification of recombinant DcsG was performed as described previously [25].
Mutagenesis. The KOD -Plus- Mutagenesis Kit (Toyobo, Osaka, Japan) was used to generate
DcsG mutants, according to the manufacturer’s protocol. The mutagenic primers containing the
desired
mutations
(underlined)
were
as
(W57A,
(W57A,
(C142A,
follows:
sense),
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
5’-
GCGACCTGGGCGGAGCGGCAGGCAGAGTTC-3’
GGGGGAGCGGAAGACGACGGCTTCGAAACG-3’
CTCCAAGGACGTCCAGCGCTATCAGCGG-3’
TAGGCGCCGTCGGTCGGCTTGACGAC-3’
CTCCAAGGACGTCCAGCGCTATCAGCGG-3’
GCGCAGCCGTCGGTCGGCTTGACGAC-3’
GCGGGCGCGATGCTCATGCCGGAC-3’
antisense),
sense),
(C142A,
(Y143A,
(Y143A,
antisense),
sense),
antisense),
sense),
(K202A,
(K202A,
GTGGATGGCGTGGCTGTAGACGCCG-3’
GCCACCCTCGACTTCCGCCAGGCCCGTGAC-3’
GACGTGCACCGTGCCGTCCGGCATGAG-3’
GCCCAGGCCCGTGACGCCGACGAGGAC-3’
GAAGTCGAGGGTGGGGACGTGCACCGTGCC-3’
GCGGTCGACCTGGTCCGCGGCGCCGAC-3’
TCCGCAGACCAGGGGCAGGTCCAGGCC-3’
CTCTGCGAGCCCTCGCTCAACCTC-3’
antisense),
sense),
(P215A,
(P215A,
(R220A,
antisense),
sense),
(R220A,
antisense),
sense),
(R254A,
(R254A,
antisense),
sense),
(E271A,
CGCCATCTCCAGCACCATGGGGG-3’
CTCTGCGAGCCCTCGCTCAACCTC-3’
CTGCATCTCCAGCACCATGGGGG-3’
CTCAACCTCACCTTCAGCGAGGACGGG-3’
(E271A,
(E271Q,
(E271Q,
(S276A,
antisense),
sense),
antisense),
sense),
and
CGCGGGCTCGCAGAGCTCCATCTCC-3’ (S276A, antisense). To generate the expression
vector for the mutant DcsG, the vector for the wild-type DcsG was amplified using sense and
antisense primers. The mutation was confirmed by DNA-sequencing analysis. Expression and
purification of the mutant proteins was performed according to the same method used for the wild
type.
This article is protected by copyright. All rights reserved

Kinetic analysis. Substrate specificity of DcsG had previously been investigated [25]. The
substrates used in the present study were D-OUS, β-aminooxy-D-alanine, and D-homocysteine. It
was revealed that generations of the cyclized D-amino acid by DcsG were approximately coupled
with the conversion of ATP to ADP; therefore, the catalytic activity of DcsG or the mutants for the
three substrates were determined by the ADP release-coupled assay [51]. The reaction mixture,
consisting of 250 mM Tris-HCl (pH 8.0), 10 mM KCl, 5 mM MgCl2, 1 mM dithiothreitol, 5 mM
ATP, 2.5 mM potassium phosphoenolpyruvate, 0.3 mM NADH, 45 U mL-1 pyruvate kinase
(Wako Chemical, Osaka, Japan), 20 U mL-1 lactate dehydrogenase (Oriental Yeast, Tokyo, Japan),
and DcsG or its mutant with a concentration range of 25−230 μg mL-1, was incubated at 30ºC in
the presence of a fixed concentration (50 mM) of each substrate. At the first time, all the assays
were done with the protein concentration of 25 μg mL-1. If the activity was too low to determine,
the concentration was increased. The rate of decrease in absorbance at 340 nm of NADH (ε =
6,220 M-1cm-1) was evaluated using a V-550 spectrophotometer (Jusco), and the activity was
determined based on the rate of decrease in absorbance and the enzyme concentration.
To quantify the generated cyanate, colorimetric analysis was performed [52,53]. The reaction
mixture (100 μL), consisting of 20 mM Hepes-KOH (pH 8.0), 10 mM KCl, 5 mM MgCl2, 5 mM
ATP, 0.613−20 mM D-OUS, and 50 μg mL-1 DcsG was incubated at 30°C for 10 min. Then, 50
μL of 100 mM acetic acid and 150 μL of 20 mM m-aminobenzoic acid was added to the mixture.
Under the low pH condition (ca. 4.5), reactivity of DcsG was lost, whereas the reaction between
m-aminobenzoic acid and cyanate was stimulated. After the incubation at 40°C for 10 min, 300 μL
of 12 M HCl was added, followed by the measurement of absorbance at 310 nm. Concentration of
cyanate was quantified based on the calibration curve prepared by using the standard potassium
cyanate. To investigate whether DcsG stimulate the formation of cyanate from carbamoyl
phosphate, the reaction mixture, consisting of 20 mM Hepes-KOH (pH 8.0), 10 mM KCl, 5 mM
MgCl2, 5 mM ADP, 5 mM carbamoyl phosphate, and 50 μg mL-1 DcsG was incubated in the
presence or absence of 5 mM D-CS at 25°C for 10 min. The mixture was added with acetic acid
and m-aminobenzoic acid, followed by incubation at 25°C for 30 min. After the reaction was
stopped by the addition of HCl, the absorbance of the solution was measured.
HPLC analysis. To know whether carbamoyl phosphate was generated from D-OUS and ATP,
the reaction mixture, consisting of 10 mM KCl, 5 mM MgCl2, 1 mM tris(2-
carboxyethyl)phosphine, 10 mM D-OUS, 5 mM ATP, 10 mM L-ornithine, 10 U mL-1 L-ornithine
This article is protected by copyright. All rights reserved

carbamoyltransferase (Sigma-Aldrich, Tokyo, Japan), and 100 μg mL-1 DcsG, was incubated for
the given times at 30°C in the 50 mM Hepes-KOH buffer. The reaction was blocked by boiling at
1, 2, or 3 h after the initiation of the incubation. The product was analyzed by HPLC following
derivatization with o-phthaldialdehyde and 2-mercaptoethanol [54]. An aliquot was injected into
an ODS-A column (YMC, Kyoto, Japan, 100×4.6 mm) equilibrated with 92% solvent A (40 mM
sodium acetate at pH 4.0 supplemented with 20% methanol) and 8% solvent B (100% methanol).
The ratio of solvent B was linearly increased from 8 to 60% during 30 min and from 60 to 100%
during the subsequent 10 min. The flow rate was set to 1.0 mL min-1. Amino acid derivatives were
detected by evaluating the absorbance at 340 nm using an MD-2010 multiwavelength detector
(Jusco, Tokyo, Japan). In the standard assays using authentic samples, derivatives of D-OUS, L-
citrulline, and D-CS were detected at 6.5, 10.5 and 20 min, respectively.
Crystallography. To crystallize DcsG, the buffer of the protein was changed to 20 mM Hepes-
NaOH (pH 7.5) supplemented with 100 mM NaCl, 10 mM MgCl2, 0.5 mM tris(2-
carboxyethyl)phosphine, and 5 mM ADP and subsequently concentrated to 10 mg mL-1 using
Amicon Ultra (Millipore). The DcsG crystals were grown by the sitting-drop vapor-diffusion
method with a 1:1 (v/v) ratio of protein to precipitant solutions. Thin plate crystals were formed
within 7 days using a 0.1 M Tris-HCl buffer (pH 7.5) supplemented with 1.0 M potassium sodium
L-tartrate.
In general, sequence similarities among the ATP-grasp enzymes are low; therefore, the
molecular replacement method was not used as the structural determination method. The tertiary
structure of DcsG was determined by a single anomalous dispersion method using the L-
selenomethionine derivative. The crystals were flash-frozen prior to data collection using a
cryoprotectant containing 30% (v/v) glycerol. The diffraction intensities of the crystals were
collected by synchrotron radiation (0.97864 Å) from BL38B1 at SPring-8 (Hyogo, Japan). The X-
ray diffraction was evaluated using a CCD camera equipped at the station, and the intensities were
integrated and scaled using the HKL2000 program [55]. A total of 24 selenium sites were
identified using the SnB program [56] and the accuracy of the phases was improved by the
SHARP program [57]. The resulting electron density map, visualized using Xfit from the
XtalView software package [58], enabled us to build a partial model. This model was refined with
simulated annealing and conventional restrained refinement methods using the CNS program [59].
A subset of 5% of the reflections was used to monitor the free R factor (Rfree) [60]. Each
This article is protected by copyright. All rights reserved

refinement cycle included the refinement of the positional parameters, individual isotropic B-
factors, and revision of the model. The structure was completed by further rounds of manual
model fitting. Details of data collection and refinement statistics are presented in Table 1.
References
1.
Hidy PH, Hodge EB, Young VV, Harned RL, Brewer GA, Phillips WF, Runge WF, Stavely
HE, Pohland A, Boaz H & Sullivan, HR (1955) Structure and reaction of cycloserine. J Am
Chem Soc 77, 2345–2346.
2.
3
Lambert MP & Neuhaus FC (1972) Mechanism of D-cycloserine action: alanine racemase
from Escherichia coli W. J Bacteriol 110, 978–987.
Fenn TD, Stamper GF, Morollo AA & Ringe D (2003) A side reaction of alanine racemase:
transamination of cycloserine. Biochemistry 42, 5775–5783.
4
Fenn TD, Holyoak T, Stamper GF & Ringe D (2005) Effect of a Y265F mutant on the
transamination-based cycloserine inactivation of alanine racemase. Biochemistry 44, 5317–
5327.
5
Nakatani Y, Opel-Reading HK, Merker M, Machado D, Andres S, Kumar SS,
Moradigaravand D, Coll F, Perdigão J, Portugal I, Schön T, Nair D, Devi KRU, Kohl TA,
Beckert P, Clark TG, Maphalala G, Khumalo D, Diel R, Klaos K, Aung HL, Cook GM,
Parkhill J, Peacock SJ, Swaminathan S, Viveiros M, Niemann S, Krause KL & Köser CU
(2017) Role of alanine racemase mutations in Mycobacterium tuberculosis D-cycloserine
resistance. Antimicrob Agents Chemother 61, e01575-17.
6
7
8
9
Tassoni R, van der Aart LT, Ubbink M, van Wezel GP & Pannu NS (2017) Structural and
functional characterization of the alanine racemase from Streptomyces coelicolor A3(2).
Biochem Biophys Res Commun 483, 122–128.
Neuhaus FC & Lynch JL (1964) The enzymatic synthesis of D-alanyl-D-alanine. III. On the
inhibition of D-alanyl-D-alanine synthetase by the antibiotic D-cycloserine. Biochemistry 3,
471–480.
Bruning J, Murillo A, Chacon O, Barletta RG & Sacchettini JC (2011) Structure of the
Mycobacterium tuberculosis D-alanine:D-alanine ligase, a target of the antituberculosis drug
D-cycloserine. Antimicrob Agents Chemother 55, 291–301.
Prosser GA & de Carvalho LP (2013) Reinterpreting the mechanism of inhibition of
Mycobacterium tuberculosis D-alanine:D-alanine ligase by D-cycloserine. Biochemistry 52,
This article is protected by copyright. All rights reserved

7145–7149.
10 Batson S, de Chiara C, Majce V, Lloyd AJ, Gobec S, Rea D, Fülöp V, Thoroughgood CW,
Simmons KJ, Dowson CG, Fishwick CWG, de Carvalho LPS & Roper DI (2017) Inhibition
of D-Ala:D-Ala ligase through a phosphorylated form of the antibiotic D-cycloserine. Nat
Commun 8, 1939.
11 Amorim Franco TM, Favrot L, Vergnolle O & Blanchard JS (2017) Mechanism-based
inhibition of the Mycobacterium tuberculosis branched-chain aminotransferase by D- and L-
cycloserine. ACS Chem Biol 12, 1235–1244.
12 Pfyffer G, Bonato DA, Ebrahimzadeh A, Gross W, Hotaling J, Kornblum J, Laszlo A,
Roberts G, Salfinger M, Wittwer F & Siddiqi S (1999) Multicenter laboratory validation of
susceptibility testing of Mycobacterium tuberculosis against classical second-line and newer
antimicrobial drugs by using the radiometric BACTEC 460 technique and the proportion
method with solid media. J Clin Microbiol 37, 3179–3186.
13 Haydel S (2010) Extensively drug-resistant tuberculosis: a sign of the times and an impetus
for antimicrobial discovery. Pharmaceuticals 3, 2268–2290.
14 Herberg LJ & Rose IC (1990) Effects of D-cycloserine and cycloleucine, ligands for the
NMDA-associated strychnine-insensitive glycinesite, on brain-stimulation reward and
spontaneous locomotion. Pharmacol Biochem Behav 36, 735–738.
15 Leeson PD & Iversen LL (1994) The glycine site on the NMDA receptor: Structure-activity
relationships and therapeutic potential. J Med Chem 37, 4053–4067.
16 Goff DC, Tsai G, Levitt J, Amico E, Manoach D, Schoenfeld DA, Hayden DL, McCarley R
& Coyle JT (1999) Placebo-controlled trial of D-cycloserine added to conventional
neuroleptics in patients with schizophrenia. Arch Gen Psychiatry 56, 21–27.
17 Tanaka N & Sashikata K (1969) Biogenesis of D-4-amino-3-isoxazolidone and O-cabamyl-
D-serine. J Gen Appl Microbiol 9, 409–414.
18 Svensson ML, Valerie K & Gatenbeck S (1981) Hydroxyurea, a natural metabolite and an
intermediate in D-cycloserine biosynthesis in Streptomyces garyphalus. Arch Microbiol 129,
210–212.
19 Svensson ML & Gatenbeck S (1982) The pathway of D-cycloserine biosynthesis in
Streptomyces garyphalus. Arch Microbiol 131, 129–131.
20 Kumagai T, Koyama Y, Oda K, Noda M, Matoba Y & Sugiyama M (2010) Molecular
cloning and heterologous expression of a biosynthetic gene cluster for the antitubercular
This article is protected by copyright. All rights reserved

agent D-cycloserine produced by Streptomyces lavendulae. Antimicrob Agents Chemother
54, 1132–1139.
21 Matsuo H, Kumagai T, Mori K & Sugiyama M (2003) Molecular cloning of a D-cycloserine
resistance gene from D-cycloserine-producing Streptomyces garyphalus. J Antibiot 56, 762–
767.
22 Noda M, Kawahara Y, Ichikawa A, Matoba Y, Matsuo H, Lee DG, Kumagai T & Sugiyama
M (2004) Self-protection mechanism in D-cycloserine-producing Streptomyces lavendulae:
gene cloning, characterization, and kinetics of its alanine racemase and D-alanyl-D-alanine
ligase, which are target enzymes of D-cycloserine. J Biol Chem 279, 46143–46152.
23 Kumagai T, Takagi K, Koyama Y, Matoba Y, Oda K, Noda M & Sugiyama M (2012) Heme
protein and hydroxyarginase necessary for biosynthesis of D-cycloserine. Antimicrob Agents
Chemother 56, 3682–3689.
24 Dietrich D, van Belkum MJ & Vederas JC (2012) Characterization of DcsC, a PLP-
independent racemase involved in the biosynthesis of D-cycloserine. Org Biomol Chem 10,
2248–2254.
25 Uda N, Matoba Y, Kumagai T, Oda K, Noda M & Sugiyama M (2013) Establishment of an
in vitro D-cycloserine-synthesizing system by using O-ureido-L-serine synthase and D-
cycloserine synthetase found in the biosynthetic pathway. Antimicrob Agents Chemother 57,
2603–2612.
26 Ahn YC, Fischer C, van Belkum MJ & Vederas JC (2018) PLP-independent racemization:
mechanistic and mutational studies of O-ureidoserine racemase (DcsC). Org Biomol Chem.
16, 1126–1133.
27 Oda K, Matoba Y, Kumagai T, Noda M & Sugiyama M (2013) Crystallographic study to
determine the substrate specificity of an L-serine-acetylating enzyme found in the D-
cycloserine biosynthetic pathway. J Bacteriol 195, 1741–1749.
28 Uda N, Matoba Y, Oda K, Kumagai T & Sugiyama M (2015) The structural and mutational
analyses of O-ureido-L-serine synthase necessary for D-cycloserine biosynthesis. FEBS J
282, 3929–3944.
29 Galperin MY & Koonin EV (1997) A diverse superfamily of enzymes with ATP-dependent
carboxylate-amine/thiol ligase activity. Protein Sci 6, 2639–2643.
30 Fawaz MV, Topper ME & Firestine SM (2011) The ATP-grasp enzymes. Bioorg Chem 39,
185–191.
This article is protected by copyright. All rights reserved

31 Taussig A. (1960) The effect of carbamyl phosphate and cyanate on Escherichia coli. Can J
Microbiol 6, 619–629.
32 Shambaugh GE 3rd (1977) Urea biosynthesis I. The urea cycle and relationships to the citric
acid cycle. Am J Clin Nutr 30, 2083–2087.
33 Shambaugh GE 3rd (1979) Pyrimidine biosynthesis. Am J Clin Nutr 32, 1290–1297.
34 Jackson MJ, Beaudet AL & O'Brien WE (1986) Mammalian urea cycle enzymes. Annu Rev
Genet 20, 431–464.
35 Taussig A (1960) The synthesis of the induced enzyme, “cyanase”, in E. coli. Biochim
Biophys Acta 44, 510–519.
36 Anderson PM (1980) Purification and properties of the inducible enzyme cyanase.
Biochemistry 19, 2882–2888.
37 Allen CM Jr & Jones ME (1964) Decomposition of carbamylphosphate in aqueous solutions.
Biochemistry 3, 1238–1247.
38 Holm L & Sander CJ (1993) Protein structure comparison by alignment of distance matrices.
J Mol Biol 233, 123–138.
39 Ouchi T, Tomita T, Horie A, Yoshida A, Takahashi K, Nishida H, Lassak K, Taka H,
Mineki R, Fujimura T, Kosono S, Nishiyama C, Masui R, Kuramitsu S, Albers SV,
Kuzuyama T & Nishiyama M (2013) Lysine and arginine biosyntheses mediated by a
common carrier protein in Sulfolobus. Nat Chem Biol 9, 277–283.
40 Zhao G, Jin Z, Wang Y, Allewell NM, Tuchman M & Shi D (2013) Structure and function
of Escherichia coli RimK, an ATP-grasp fold, L-glutamyl ligase enzyme. Proteins 81, 1847–
1854.
41 Hara T, Kato H, Katsube Y & Oda J (1996) A pseudo-Michaelis quaternary complex in the
reverse reaction of a ligase: structure of Escherichia coli B glutathione synthetase
complexed with ADP, glutathione, and sulfate at 2.0 Å resolution. Biochemistry 35, 11967–
11074.
42 Fan C, Moews PC, Walsh CT & Knox JR (1994) Vancomycin resistance: structure of D-
alanine:D-alanine ligase at 2.3 Å resolution. Science 266, 439–443.
This article is protected by copyright. All rights reserved

43 Foote J & Lipscomb WN (1981) Kinetics of aspartate transcarbamylase from Escherichia
coli for the reverse direction of reaction. J Biol Chem 256, 11428–11433.
44 Zúñiga M, Pérez G & González-Candelas F (2002) Evolution of arginine deiminase (ADI)
pathway genes. Mol Phylogenet Evol 25, 429–444.
45 Kuranaga T, Matsuda K, Sano A, Kobayashi M, Ninomiya A, Takada K, Matsunaga S &
Wakimoto T (2018) Total synthesis of the nonribosomal peptide surugamideꢀB and
identification of a new offloading cyclase family. Angew Chem Int Ed Engl 57, 9447–9451.
46 Dorival J, Risser F, Jacob C, Collin S, Dräger G, Paris C, Chagot B, Kirschning A, Gruez A
& Weissman KJ (2018) Insights into a dual function amide oxidase/macrocyclase from
lankacidin biosynthesis. Nat Commun 9, 3998
47 Li B, Yu JP, Brunzelle JS, Moll GN, van der Donk WA & Nair SK (2006) Structure and
mechanism of the lantibiotic cyclase involved in nisin biosynthesis. Science 311, 1464–1467.
48 Roach PL, Clifton IJ, Fülöp V, Harlos K, Barton GJ, Hajdu J, Andersson I, Schofield CJ &
Baldwin JE (1995) Crystal structure of isopenicillin N synthase is the first from a new
structural family of enzymes. Nature 375, 700–704.
49 Stammer CH (1962) β-Aminoxy-D-alanine. J Org Chem 27, 2957–2958.
50 Weaver JD, Busch NF & Stammer CH (1974) Cycloserine carbamates. J Med Chem 17,
1033–1035.
51 Williamson JR & Corkey BE (1969) Assays of intermediates of the citric acid cycle and
related compounds by fluorometric enzyme methods. Methods Enzymol 13, 434–513.
52 Guilloton M & Karst F (1985) A spectrophotometric determination of cyanate using reaction
with 2-aminobenzoic acid. Anal Biochem 149, 291–295.
53 Winder B, Mulholland MR & Mopper K (2013) Chromatographic determination of
nanomolar cyanate concentrations in estuarine and sea waters by precolumn fluorescence
derivatization. Anal Chem 85, 6661–6666.
54 Canevari L, Vieira R, Aldegunde M & Dagani F (1992) High-performance liquid
chromatographic separation with electrochemical detection of amino acids focusing on
neurochemical application. Anal Biochem 205, 137–142.
55 Otwinowski Z & Minor W (1997) Processing of X-ray diffraction data collected in
oscillation mode. Methods Enzymol 276, 307–326.
56 Miller R, Gallo SM, Khalak HG & Weeks CM (1994) SnB: crystal structure determination
This article is protected by copyright. All rights reserved

via shake-and-bake. J Appl Cryst 27, 613–621.
57 de la Fottelle E & Bricogne G (1997) Maximum-likelihood heavy-atom parameters
refinement in the MIR and MAD methods. Methods Enzymol 276, 472–494.
58 McRee DE (1992) A visual protein crystallographic software system for X11/XView. J Mol
Graphics 10, 44–46.
59 Brünger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-Kunstleve RW, Jiang JS,
Kuszewski J, Nilges M, Pannu NS, Read RJ, Rice LM, Simonson T & Warren GL (1998)
Crystallography & NMR system: a new software suite for macromolecular structure
determination. Acta Crystallogr Sect D Biol Crystallogr 54, 905–921.
60 Brünger AT (1992) The free R value: a novel statistical quantity for assessing the accuracy
of crystal structures. Nature 355, 472–475.
61 Liu S, Chang JS, Herberg JT, Horng MM, Tomich PK, Lin AH & Marotti KR (2006)
Allosteric inhibition of Staphylococcus aureus D-alanine:D-alanine ligase revealed by
crystallographic studies. Proc Natl Acad Sci USA 103, 15178–15183.
62 Kraulis PJ (1991) MOLSCRIPT: a program to produce both detailed and schematic plots of
protein structures. J Appl Cryst 24, 946–950.
63 DeLano WL (2002) PyMOL. DeLano Scientific, San Carlos, CA, 700.
This article is protected by copyright. All rights reserved

Table 1. Data collection and refinement statistics
Data set
Se-labeled DcsG
Data collection
Space group
Cell dimensions
a (Å)
P21
56.45
120.73
b (Å)
c (Å)
102.81
β (°)
101.01
Wavelength (Å)
Resolution (Å)
Unique reflection
Redundancya
Completeness (%)a
Rmerge (%)a, b
I/a
0.97864
100−2.32
56,548
3.6 (2.9)
95.3 (82.4)
11.7 (44.7)
10.1 (1.8)
Refinement
Resolution (Å)
Used reflections
No. of atoms
Protein
30−2.32
55,347
9,184
217
Ligand
Water
756
R (%)
19.5
25.1
Rfree (%)
Rms deviationsc
Bond length (Å)
Bond angle (º)
Mean B-factor (Å2)
Protein
0.008
1.3
24.6
33.6
Ligand
This article is protected by copyright. All rights reserved

Solvent
29.4
Ramachandran plot (%)
Inner core
79.8
96.8
Core
aValues in parentheses are for the highest resolution bin.
bRmerge=Σ|I-<I>|/ΣI, where I is the observed intensity and <I> is the mean value of I.
cRms deviations are calculated by CNS [59].
This article is protected by copyright. All rights reserved

Acknowledgments
The DNA sequence determination was performed at the Analysis Center of Life Science,
Hiroshima University (Japan). The synchrotron radiation experiments were performed at the
BL38B1 in SPring-8 (Japan), with the approval of the Japan Synchrotron Radiation Research
Institute (2014A1077 and 2014B1081). We thank the beam-line staff at SPring-8 for the kind help
provided for the collection of X-ray data and Dr. K. Oda at Hiroshima University for the figure
preparation. This study was supported by a Grant-in-Aid for Scientific Research (15K07997 to Y.
Matoba) from the Japan Society for the Promotion of Science and by research funds provided by
the Institute for Fermentation, Osaka (IFO) and Novozymes Japan, awarded to Y. Matoba.
Author contributions
Y. M. designed research; Y. M., N. U. and M. K. performed experiments; All authors analyzed
data and discussed; Y. M. prepared the draft of manuscript; M. S. revised the manuscript.
Conflict of interest
The authors declare no conflict of interest.
Figure Legends
Fig. 1. Chemical reactions. A, Predicted biosynthetic pathway of D-CS in S. lavendulae
ATCC11924. B, Catalytic mechanism underlying amide bond formation by the general ATP-grasp
enzymes.
Fig. 2. Ribbon diagrams of the crystal structures of DcsG and its associated enzymes. A, Structure
of one of four monomers in the asymmetric unit of the DcsG crystal. The monomer is bound to
ADP and L-tartrate. B, Structure of glutathione synthetase from E. coli in a complex with ADP,
glutathione, and sulfate (PDB ID 1GSA) [41]. C, Structure of D-alanyl-D-alanine synthetase DdlB
from E. coli in a complex with ADP and an intermediate analog (PDB ID 2DLN) [42]. D,
Structure of the apo form of the D-alanyl-D-alanine synthetase from Staphylococcus aureus (PDB
ID 2I87) [61] as a representative structure taking the open form. The catalytic site of each enzyme
This article is protected by copyright. All rights reserved

in A−C contains two Mg(II) ions presented as green spheres. View directions of A−D were
adjusted to be the same using the rotation and translation matrices determined by a DALI program
[38]. N-terminal, central, and C-terminal domains in each enzyme are presented in magenta, cyan,
and gold, respectively. All the ligands are presented as stick models. The figures were prepared by
MOLSCRIPT [62].
Fig. 3. Simulated annealing composite omit map of the ligands bound to DcsG. The map, which
was computed by CNS [59] after removal of an ADP molecule, a L-tartrate, and two Mg(II) ions
bound to the active site in the monomer A, are contoured at 3 σ. The figure was prepared by
PyMOL [63].
Fig. 4. Active site structure. A, ATP-binding site of DcsG, where an ADP molecule is
accommodated. B, Substrate-binding site of DcsG, where an L-tartrate ion is accommodated. C,
Substrates-binding site of E. coli glutathione synthetase (PDB ID 1GSA) [41], where a product
and a sulfate ion are accommodated. D, Substrates-binding site of E. coli DdlB (PDB ID 2DLN)
[42], where an intermediate analogue is accommodated. Carbon atoms of the bound ligands (ADP
and L-tartrate in A and B, glutathione and ADP in C, and intermediate analog and ADP in D) are
presented in orange, whereas the carbon atoms from each enzyme are presented in light gray. The
figures were prepared by MOLSCRIPT [62]. Abbreviations for amino acids in one-letter code are
used: C, cysteine; D, aspartic acid; E, glutamic acid; I, isoleucine; K, lysine; L, leucine; N,
asparagine; P, proline; Q, glutamine; R, arginine; S, serine; V, valine; W, tryptophan; Y, tyrosine.
Fig. 5. Catalytic activities of DcsG and its mutants. The enzymatic activities, which were
normalized by the concentration, were evaluated in the presence of 50 mM substrate. Color of the
bars is altered by the substrates tested: bars for D-OUS, D-homocysteine, and β-aminooxy-D-
alanine are gray, cyan, and orange, respectively. All the evaluations were performed in duplicate
and the average values were used for the graph preparation. Ratios of the activity of mutants
toward that of wild type (WT) were determined for each substrate.
Fig. 6. Possible mechanisms of DcsG for the generations of cyanate and D-CS from the
phosphorylated D-OUS intermediate. A, A mechanism, in which the phosphate group initially
attacks the carbamoyl carbon atom nucleophilically, leading to the generation of a carbamoyl
This article is protected by copyright. All rights reserved

phosphate. A positive charge on the Arg220 residue neutralizes the negative charge generated in
the first tetrahedral intermediate. B, A mechanism, in which the phosphate group initially removes
the proton from the carbamoyl group, leading to the direct generation of isocyanic acid. A positive
charge on the Arg220 residue decreases the pKa value of the carbamoyl group in D-OUS.
Fig. 7. Proposed catalytic mechanism underlying DcsG during the reaction with D-OUS. A,
Putative reaction scheme. Pi and Cya mean inorganic phosphate and cyanate, respectively. B,
Schematic representation of the interactions formed among two Mg(II) ions, an ADP molecule, a
phosphorylated D-OUS molecule, and the residues in DcsG. This is one of the putative
intermediate states formed during the conversion from ATP and D-OUS (second step in A) to
ADP, phosphate, cyanate, and D-CS (third step in A). The intramolecular interaction of the
substrate via the Nζ atom is formed only in D-OUS, but not in the smaller substrates, including β-
aminooxy-D-alanine and D-homocysteine, due to the absence of the atom.
This article is protected by copyright. All rights reserved

febs_15163_f1.tif
This article is protected by copyright. All rights reserved