Báo cáo khoa học: Structural studies of thymidine kinases from Bacillus anthracis and Bacillus cereus provide insights into quaternary structure and conformational changes upon substrate binding pot

Structural studies of thymidine kinases from
Bacillus anthracis and Bacillus cereus provide insights
into quaternary structure and conformational changes
upon substrate binding
Urszula Kosinska
1
, Cecilia Carnrot
2
, Michael P. B. Sandrini
3
, Anders R. Clausen
3
, Liya Wang
2
,
Jure Piskur
3
, Staffan Eriksson
2
and Hans Eklund
1
1 Department of Molecular Biology, Swedish University of Agricultural Sciences, Uppsala Biomedical Centre, Sweden
2 Molecular Biosciences, Swedish University of Agricultural Sciences, Uppsala Biomedical Centre, Sweden
3 Department of Cell and Organism Biology, Lund University, Sweden
Bacillus anthracis and Bacillus cereus are two closely
related species of the genus Bacillus. B. anthracis cau-
ses anthrax, a disease that in most cases has fatal
consequences [1]. B. cereus is a human pathogen asso-
ciated with food poisoning [2]. Both species produce
endospores under stressful conditions as a means
of survival through environmental stress. The major

genetic difference between these two species is associ-
ated with two toxin-encoding plasmids, pXO1 and
pXO2 [3,4], which are present in B. anthracis but not
in B. cereus.
Thymidine kinase (TK; EC 2.7.1.21) is a deoxyribo-
nucleoside kinase (dNK) that phosphorylates thymi-
dine to thymidine monophosphate. Mammals possess
Keywords
deoxythymidine triphosphate; dimer;
feedback inhibitor; phosphate donor;
tetramer
Correspondence
H. Eklund, Swedish University of
Agriculturla Sciences, Box 590, BMC,
Uppsala SE-75124, Sweden
E-mail:
(Received 28 August 2006, revised 17
November 2006, accepted 24 November
2006)
doi:10.1111/j.1742-4658.2006.05617.x
Thymidine kinase (TK) is the key enzyme in salvaging thymidine to pro-
duce thymidine monophosphate. Owing to its ability to phosphorylate
nucleoside analogue prodrugs, TK has gained attention as a rate-limiting
drug activator. We describe the structures of two bacterial TKs, one from
the pathogen Bacillus anthracis in complex with the substrate dT, and the
second from the food-poison-associated Bacillus cereus in complex with the
feedback inhibitor dTTP. Interestingly, in contrast with previous structures
of TK in complex with dTTP, in this study dTTP occupies the phosphate
donor site and not the phosphate acceptor site. This results in several con-
formational changes compared with TK structures described previously.

One of the differences is the way tetramers are formed. Unlike B. anthracis
TK, B. cereus TK shows a loose tetramer. Moreover, the lasso-domain is
in open conformation in B. cereus TK without any substrate in the active
site, whereas in B. anthracis TK the loop conformation is closed and
thymidine occupies the active site. Another conformational difference lies
within a region of 20 residues that we refer to as phosphate-binding b-hair-
pin. The phosphate-binding b-hairpin seems to be a flexible region of the
enzyme which becomes ordered upon formation of hydrogen bonds to the
a-phosphate of the phosphate donor, dTTP. In addition to descriptions of
the different conformations that TK may adopt during the course of reac-
tion, the oligomeric state of the enzyme is investigated.
Abbreviations
Ba-TK, Bacillus anthracis thymidine kinase; Bc-TK, Bacillus cereus thymidine kinase; Ca-TK, Clostridium acetobutylicum thymidine kinase;
dCK, deoxycytidine kinase; dGK, deoxyguanosine kinase; dNK, deoxyribonucleoside kinase; hTK1, human thymidine kinase 1; MPD,
2-methyl-2,4-pentadiol; P-b-hairpin, phosphate-binding b-hairpin; TK, thymidine kinase; Uu-TK, Ureaplasma urealyticum thymidine kinase.
FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS 727
four deoxyribonucleoside-specific dNKs: cytosolic
deoxycytidine kinase (dCK) and TK1 and mitochond-
rial deoxyguanosine kinase (dGK) and TK2. Bacteria,
however, have a smaller group of dNKs. Most
bacteria have TK that both sequence wise and struc-
turally resembles TK1 [5]. In addition, one or two
non-TK1-like dNKs can be found in most Gram-
positive bacteria. Besides TK, there are two
dCK ⁄ dGK-like dNKs in B. anthracis and B. cereus.
The amino-acid sequence identity between the TKs
from B. anthracis (Ba-TK) and B. cereus (Bc-TK) is
as high as 96%. The sequence identity with human
TK1 (hTK1) is 37–38%.
From amino-acid sequence analysis, it was suggested

that dCK, dGK and TK2 belong to one group, which
will be referred to as dNKs, whereas TK1-like enzymes
form a group of their own. This was confirmed by sub-
sequent structure determinations of a multisubstrate
Drosophila melanogaster dNK together with human
dGK [6], followed by human dCK [7], and later on
hTK1 [8,9] and TK from Ureaplasma urealyticum
(Uu-TK) [9]. However, Herpes simplex virus type 1
thymidine kinase shares structural and sequential simi-
larities with dNKs and does not belong to the TK1-
like group of enzymes. dNKs are biological dimers
with overlapping substrate specificity, which can be
attributed to differences of a few residues in the active
site [7,10]. TK1-like enzymes only accept thymidine
and deoxyuridine as substrates, and all interactions
between the substrate and the enzyme are by main-
chain hydrogen bonds to polar groups of the base.
The active site of TK1-like enzymes is smaller than
that found in dNKs and lined with hydrophobic resi-
dues. Whereas TK1-like enzymes have a lasso-domain
which covers the active site when the substrate is
bound, the active site of dNKs is covered by a helical
domain containing an arginine-rich lid. In both
enzyme families, the active site is situated at the C-ter-
minus of the central parallel b-sheet in the a ⁄ b-
domain, which contains a conserved P-loop
(GXXXGKS ⁄ T). Yet another difference between
dNKs and TK1-like enzymes is the presence of a struc-
tural Zn
2+

ion in the lasso-domain of TKs. There are
no structural metals in members of the dNK family.
Furthermore, all TKs that have been structurally
determined form tetramers in the crystals. Enzymes
from the dNK and TK1 family can use different
NTPs, usually prefer ATP as phosphate donor, and
are feedback inhibited by the respective dNTP, such
that dTTP is a feedback inhibitor of TK1-like enzymes
[11]. It can be concluded that, despite structural differ-
ences, dNKs and TKs catalyze the phosphorylation of
deoxyribonucleosides in similar ways [9,12].
In this study, we describe the 3D structure of Ba-TK
in complex with thymidine and a phosphate ion, as well
as Bc-TK with an occupied phosphate donor site. As
these enzymes are essentially identical, these structures
represent the enzyme trapped in different conforma-
tional stages, which reflect structural conformations
that TK adopts along its reaction pathway.
Results
Overall structure
TKs from B. anthracis and B. cereus share 96%
amino-acid sequence identity. Bc-TK consists of 195
amino acids, and Ba-TK is one amino acid shorter.
The last five and the last four amino acids in Bc-TK
and Ba-TK, respectively, are different. Besides the dif-
ferences in the C-termini, there are only three addi-
tional amino acids that are not conserved. The lysine
at position 76 in Ba-TK is a glutamic acid in Bc-TK,
the methionine at position 82 in Ba-TK is a leucine in
Bc-TK, and the alanine at position 147 in Ba-TK is a

valine in Bc-TK. These minor differences do not affect
the overall structures of Ba-TK and Bc-TK, thus these
proteins may be considered structurally identical.
The overall structures of Ba-TK and Bc-TK closely
resemble previously described TK structures [8,9,
13,14]. The enzymes are tetramers with 222-fold sym-
metry and two types of subunit–subunit interaction.
One is formed between helices a1 of two neighboring
subunits, and the other is between the edges of the
b-sheets. The subunit comprises two domains, the
N-terminal a ⁄ b-domain and the C-terminal lasso-
domain (Fig. 1). The a ⁄ b-domain is formed from a
central, six-stranded, parallel b-sheet situated between
a long a-helix, a1, and a flexible loop on one side and
three shorter helices, a2–a4, on the other side. We
have chosen to name the flexible loop, which is about
20–25 residues in length (amino acids 46–68 for Ba-TK
and Bc-TK), the phosphate-binding b-hairpin
(P-b-hairpin). As previously described [14], the P-b-
hairpin is a flexible part of the TK, which has been
reported as missing or having a variety of different
conformations. There is also a phosphate-binding
motif, the P-loop (GXXXXGKS ⁄ T), in the junction
between b1 and a1 of the a ⁄ b-domain. The lasso-
domain, so called because of its ability to capture and
position the substrate [9], comprises two perpendicular
b-hairpins, where the longer hairpin opens up to form
a lasso-shaped loop. A Zn
2+
ion ligated by four cys-

teine residues (Cys145, 148, 183 and 186) stabilizes the
lasso-domain. The active site is situated between the
a ⁄ b-domain and the lasso-domain.
Bacillus thymidine kinase structures U. Kosinska et al.
728 FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS
Ba-TK
The structure of the Ba-TK–dT complex was refined at
2.7 A
˚
resolution to a final R-factor of 20.3% and R
free
of 24.4% (Table 1). There is one subunit in the asym-
metric unit of the space group I4
1
22. Application of
crystallographic symmetries generates the tetramer.
The crystal packing creates a mixed, four-stranded
b-sheet between the tetramers. The N-terminus and
C-terminus from two neighboring subunits of one tetra-
mer form a parallel b-sheet, which is connected in an
antiparallel manner with the N-terminus and C-termi-
nus of a neighboring tetramer. Because of the crystal
contacts, it was possible to trace the entire N-terminus
as well as two residues from the His-tag. At the C-ter-
minus, only the last residue, Arg194, is missing, and
Lys192 and Gln193 have flexible side chains which
lack electron density. Residues 46–62, which are situ-
ated on the P-b-hairpin, also lack electron density and
could not be traced. In previously described TK struc-
tures, this region was reported to have a variety of

conformations or to be missing because of flexibility
[14]. As will be described below, this part of the
enzyme becomes ordered when the phosphate donor
site is occupied.
The Ba-TK–dT complex is very similar to the
Uu-TK–dT complex [14]. The substrate is bound in a
hydrophobic pocket between the a ⁄ b-domain and the
lasso-domain, surrounded by Phe92, Leu116, Phe120,
Phe125 and Ile170 (Fig. 2A). All hydrogen bonds
between the thymine and the enzyme are to main-chain
atoms such that O2 and N3 form hydrogen bonds to
main-chain atoms of residues in the lasso-domain and
O4 to main-chain atoms in the a ⁄ b-domain. The
methyl group of thymine points towards Thr155. O3¢
of the deoxyribose makes a hydrogen bond with main-
chain nitrogen of Gly174 in the lasso-domain, and O5¢
is hydrogen-bonded to Glu89, which has been sugges-
ted to be the catalytic base (Fig. 2B) [9].
Besides strong electron density for dT, there is addi-
tional density close to the P-loop which has been inter-
preted as a phosphate ion originating from the
crystallization buffer. The position of the phosphate
corresponds to the c-phosphate of the dTTP molecule
bound as feedback inhibitor to hTK1 and Uu-TK
[8,9]. The phosphate ion is coordinated by the residues
in the P-loop: the side chain of Lys21 and main-chain
atoms of residues 18–20.
Bc-TK
Bc-TK crystallized in the same space group, I4
1

22, as
Ba-TK but with different crystal packing and unit cell
parameters (Table 1). As in Ba-TK crystals, there is
Table 1. Data reduction and refinement statistics. Values in par-
entheses refer to outer resolution shell.
Ba-TK Bc-TK
Space group I4
1
22 I4
1
22
Unit cell parameters (A
˚
)a¼ b ¼ 73.2
c ¼ 223.7
a ¼ b ¼ 95.4
c ¼ 204.9
Resolution (A
˚
) 2.7 2.8
No. of unique reflections 8799 12035
Multiplicity 14.1 13.6
Completeness (%) 99.9 (99.9) 99.8 (99.8)
R
meas
10.6 (49.3) 10.0 (53.4)
<I⁄ rI > 22.6 (5.9) 26.8 (4.2)
Refinement
R (%) 20.3 19.6
R

free
(%) 24.4 23.9
R.m.s.d. bond length (A
˚
) 0.011 0.011
R.m.s.d. bond angle (°) 1.34 1.48
Average B factors (A
˚
2
)
a
42.8 57.8
a
Average B factor is calculated for residual B factors.
Fig. 1. Superposition of subunits of Ba-TK with dT (in green) and
Bc-TK with phosphate donor-mimicking dTTP and MPD bound in
the thymine-binding pocket (in yellow). The lasso-loop is in closed
conformation when dT is present and in open conformation when
the substrate is absent. The phosphate donor stabilizes the
P-b-hairpin. This part of the molecule is flexible and could not be
traced in Ba-TK.
U. Kosinska et al. Bacillus thymidine kinase structures
FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS 729
one subunit in the asymmetric unit, thus the tetramer
is formed after application of symmetry operators. The
N-termini of two subunits within the same tetramer
form an antiparallel b-sheet. The crystallographic
interactions between the tetramers involve only the
lasso-domains, which are packed such that the lasso of
one tetramer partly covers the lasso of a crystallo-

graphically related molecule in another tetramer (sup-
plementary Fig. S1). The electron density is continuous
from residue 1 through 191. In contrast with Ba-TK,
the entire region between residue 46 and 62 is fully
traceable, forming a b-hairpin.
The formation of the hairpin is mediated by a nucleo-
tide binding in the phosphate-binding site. Bc-TK was
cocrystallized with the feedback inhibitor dTTP, hence
we expected it to bind as thymine in the substrate-bind-
ing site between the lasso-domain and a ⁄ b-domain, and
the c-phosphate bound to the P-loop as described previ-
ously [8,9]. Interestingly, there is no electron density for
the inhibitor in the substrate-binding site. Instead, there
is strong positive electron density in the phosphate
donor site, which is situated opposite the substrate-
binding site. It was not possible to conclude from the
initial map whether the electron density represented an
ATP molecule originating from buffers used during
protein purification or a dTTP molecule mimicking a
phosphate donor. Consequently, during the early steps
of ligand fitting, refinement was carried out with both
ATP and dTTP. The electron density corresponding to
the ribose moiety was negative at the 2¢-OH position
when ATP was used in the refinement, and the size of
the electron density for the base was more compatible
with a pyrimidine. From this, we concluded that there
was a deoxyribonucleoside triphosphate, i.e. a dTTP
molecule, occupying the phosphate donor site (Fig. 2B).
dTTP can act as a phosphate donor for Ba-TK, but it
does so poorly compared with ATP: dTTP is only 3%

as efficient as ATP as phosphate donor when dT is used
as substrate [15].
An occupied phosphate donor site gives rise to a
3-A
˚
dissociation of subunits interacting by a1. The
base of dTTP is inserted between the a1-helix of two
subunits and is stacked between the rings of Phe18
and Phe34 from the adjacent subunit (Fig. 2B). These
two residues are conserved as hydrophobic residues in
all organisms but Gram-negative bacteria where Phe18
is replaced by asparagine and Phe34 is replaced by
glutamic acid (Fig. 3). The exchange of hydrophobic
residues for polar ones abolishes the hydrophobic
stacking interactions between the base and the enzyme.
The pattern of interaction of ATP with TKs from
Gram-negative bacteria remains to be evaluated. O4 of
the thymine makes a hydrogen bond with the main-
chain nitrogen of Val144. O3¢ of deoxyribose is hydro-
gen-bonded to Glu23, and O4¢ is hydrogen-bonded to
His58. The phosphates are stabilized by main-chain
nitrogens of P-loop residues as well as by side-chain
interactions with Lys21 and Ser22. In addition to the-
ses interactions, Ser57 and the main-chain nitrogen
from His58, both situated on the P-b-hairpin, also
make hydrogen bonds with the phosphates (Fig. 2B).
The b-phosphate of dTTP as phosphate donor is very
well aligned with the c-phosphate of dTTP bound as a
feedback inhibitor, as observed in Uu-TK and hTK1
[8,9]. dTTP not only provides binding partners for resi-

dues of the P- b-hairpin, but also affects the interac-
tions between subunits of the tetramer.
During the refinement and rebuilding process, posit-
ive density started to appear in the substrate-binding
pocket and was interpreted as a 2-methyl-2,4-pentadiol
(MPD) molecule originating from the crystallization
solution. The position of the MPD molecule
Fig. 2. (A) The active site of Ba-TK is occupied by dT and a phosphate ion. The active site is lined by hydrophobic residues. The map is a
Fo-Fc map contoured at 3r (0.1 e ⁄ A
˚
3
). (B) The phosphate donor site of Bc-TK with dTTP mimicking the phosphate donor. The base of the
phosphate donor is stacked between Phe18 and Phe34 each from adjacent subunits shown in yellow and orange, respectively. The phos-
phates are ligated by side-chain and main-chain atoms from the P-loop and P-b-hairpin. The map is a Fo-Fc map contoured at 3r (0.1 e ⁄ A
˚
3
).
Bacillus thymidine kinase structures U. Kosinska et al.
730 FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS
corresponds to the location of thymine of dT or dTTP
as observed in Ba-TK with dT, hTK1 with dTTP, and
Uu-TK with dT or dTTP (Fig. 1). The oxygens of the
MPD molecule form hydrogen bonds to main-chain
and side-chain atoms of the residues in the lasso-loop
(supplementary Fig. S2).
A dNTP molecule can generally bind as a phosphate
donor or a bisubstrate inhibitor. Whether it binds in
one or the other direction is primarily determined by
the affinity of the base of the dNTP for the substrate
site. Normally, the preferred bisubstrate inhibitor is

the dNTP where the base represents the best substrate.
Otherwise, it binds as a phosphate donor. A switch
from the bisubstrate situation to the phosphate donor
situation can be achieved by competing binding in the
substrate site. This was recently shown in a study of
deoxyadenosine kinase, where dCTP could be bound
as a bisubstrate inhibitor in the absence of substrate
but acted as a phosphate donor in the presence of sub-
strate [16]. The high concentration of MPD as a pre-
cipitant in the crystallization,  2.5 m, had some
unexpected consequences. Most surprisingly, it preven-
ted the dTTP molecule from binding in its natural site
as a bisubstrate inhibitor and instead promoted bind-
ing to the phosphate donor site. Although MPD binds
with much lower affinity than dTTP, at this high con-
centration it is able to compete with dTTP, which is
present at about 1000 times lower concentration.
The lasso-loop in Bc-TK has a different conforma-
tion from that observed in Ba-TK (Fig. 1). In Ba-TK,
where dT is occupying the active site, the lasso is
closed down over the active site and stabilized by
hydrogen bonds to thymidine. The absence of a nat-
ural substrate with hydrogen interaction partners, as is
the case in the Bc-TK structure, makes the lasso-loop
flexible. Because of crystallographic interactions, we
were able to trace the entire lasso-loop (supplementary
Fig. S1). An open conformation of the lasso-loop is
also present in the structure of TK from Clostrid-
ium acetobutylicum (Ca-TK) in complex with ADP
(PDB code 1XX6) [17]. In the Ca-TK structure, there

are neither substrates nor crystal contacts that can pro-
vide stabilizing partners. Therefore, parts of the lasso-
loop are missing. The presence of an MPD molecule in
the substrate site of Bc-TK may also add stabilizing
interaction partners for the lasso-loop, but crystal con-
tacts are probably more important for this stabiliza-
tion. Without such, the lasso-loop might have been as
flexible as in Ca-TK.
Subunit–subunit interactions
There are two types of subunit–subunit interaction in
the tetramer. One is between the long a-helices, a1,
between adjacent subunits (Fig. 4A). The helices make
an antiparallel helix pair with hydrophilic or basic
Fig. 3. Amino-acid sequence alignment of the TK1-like enzymes from B. anthracis (AAT57468), B. cereus (DQ384595), C. acetobutylicum
(NP_349490.1), U. urealyticum (NP_078433), human (P04183), mouse (NP_033413), Arabidosis thaliana (AAM63086.1), Escherichia coli
(NP_415754.1) and Yersinia pestis (NP_405720.1). The secondary-structure elements for Ba-TK and Bc-TK are shown above the alignment.
The P-loop and the zinc coordinating motifs are boxed. The catalytic Glu89 is marked in red, and the Phe18 and Phe34, which stack the base
of the phosphate donor, are marked in green. Whereas the catalytic base is conserved among TKs from different kingdoms, the stacking
phenylalanines are exchanged for hydrophilic residues in Gram-negative bacteria. The residues marked in blue take part in subunit–subunit
interactions.
U. Kosinska et al. Bacillus thymidine kinase structures
FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS 731
residues facing the interaction surface. In the case of
Ba-TK, several hydrogen bonds form between side-
chain atoms and main-chain atoms of the neighboring
helices: NH1 of Arg27 is hydrogen-bonded to main-
chain oxygen of Ser19, and NH1 of Arg31 forms a
hydrogen bond to the carboxyl oxygen of Phe18. In
Gram-negative bacteria, Arg27 and Arg31 are replaced
by glutamine and asparagine, respectively (Fig. 3).

Despite this change in amino acids, the possibility of
forming a hydrogen bond between the two subunits is
retained. In Ba-TK, the position of Arg31 is such that
ionic interactions between Arg30 and Arg31 with
Glu23 are formed. However, in Bc-TK, the distance
between the interfacing helices is 3 A
˚
longer than in
Ba-TK, and the only subunit–subunit interaction is by
hydrogen bonds between NH2 of Arg31 and main-
chain oxygen of Gln142 and Ala143 but not with
Phe18 as observed in Ba-TK.
In the second type of subunit–subunit interaction,
the b-sheet of one subunit is attached in an antiparallel
way to the b-sheet in the neighboring subunit
(Fig. 4B). This subunit interaction predominantly
involves residues from strands b6 and helix a4.
Val138, which is situated in the middle of the b6
Fig. 4. (A) The subunit–subunit interactions along a1 as observed in Ba-TK (green) and Bc-TK (yellow). The distance between neighboring
subunits is  3A
˚
wider in Bc-TK than in Ba-TK. (B) The b-sheet subunit–subunit interactions. In this interaction area, the distance between
adjacent subunits is the same for Ba-TK (green) and Bc-TK (yellow). (C) The helical interactions as observed in Bc-TK (yellow) and Ca-TK
(grey). Both structures are of the enzymes with occupied phosphate-donor sites, but without substrates. The distance between the adjacent
helices is the same in both structures, and the lasso-loops are in open conformation. (D) The superposition of Ba-TK (green), Uu-TK (grey)
and hTK1 (black) shows that the helices in adjacent subunits are closer together when the phosphate-donor is absent. In all three enzymes,
the lasso-loop is closed down over the active site.
Bacillus thymidine kinase structures U. Kosinska et al.
732 FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS
strand, is within van der Waals distance of Val138 on

the symmetry-related strand. This valine is in some
organisms replaced by other hydrophobic residues such
as isoleucine or leucine (Fig. 3). At each end of the
interaction area, the side chains of Lys140 make
hydrogen bonds to main-chain atoms of the neighbor-
ing subunits. This lysine is conserved among plants,
mammals and Gram-positive bacteria including Myco-
bacteria, but replaced with glutamic acid in Gram-
negative bacteria. There are also salt bridges between
Glu136 and Arg184 from each of the subunits. The
salt-bridge formation is conserved among other spe-
cies, such that, for hTK1, Glu144 and Arg186 are
involved, for Uu-TK, Asp144 and Arg192 are
involved, and for Ca-TK, Glu136 and Arg184 make
ionic interactions between the side chains. In some
organisms, arginine is replaced by lysine. In addition
to these interactions, a chain of water molecules is
bridged between main-chain atoms of the subunits.
The interaction area, calculated as the difference in
total accessible surface areas of isolated and interacting
structures divided by two [18], formed between
b-sheets is larger than that between the a-helices. The
b-sheet interactions and the a-helix interactions are
970 A
˚
2
and 790 A
˚
2
, respectively, for Ba-TK, and

1030 A
˚
2
and 413 A
˚
2
for Bc-TK, when residues 8–45
and 65–187 are used. The a-helical interface area for
Bc-TK increases to 550 A
˚
2
when the P-b-hairpin is
included in surface calculations. As a result, binding of
the base of the phosphate donor between subunits
reduces the subunit–subunit interaction area formed by
the a-helices by one third. For comparison, the b-sheet
interaction area for hTK1, Uu-TK and Ca-TK is
860–970 A
˚
2
, and the helical interaction area is 790–
820 A
˚
2
for hTK1 and Uu-TK and 500 A
˚
2
for Ca-TK.
Thus, the interaction areas for Ba-TK are consistent
with the subunit interactions of Uu-TK and hTK1,

which, like Ba-TK, have unoccupied phosphate donor
sites, whereas contact areas in Bc-TK resemble the
contact areas in Ca-TK, which, like Bc-TK, has occu-
pied phosphate donor sites and empty acceptor sites.
Quaternary structure
Our findings regarding the quaternary structure of the
Bacilllus TKs called for further studies by other meth-
ods. We thus analyzed both enzymes by gel filtration,
and, in both cases, obtained only tetramers corres-
ponding to a molecular mass of  100 kDa (supple-
mentary Fig. S3). Furthermore, we analyzed the
enzyme at two different concentrations with dynamic
light scattering. In these experiments, Ba-TK at a
protein concentration of 3 mgÆmL
)1
occurred only as
tetramers, whereas the same enzyme at 1 mgÆmL
)1
in
both the presence and absence of ATP existed as
dimers. Bc-TK under all these conditions appeared as
dimers (Fig. 5).
Discussion
In this study, we describe the structures of two
homologues, TKs from B. anthracis and B. cereus,in
complex with substrate dT and phosphate donor-
mimicking dTTP. Because of the high sequence iden-
tity between these two enzymes, they can be regarded
as the same enzyme at different stages of its reaction.
0

2
4
6
8
10
12
14
16
18
20
22
0.1 1 10 100
Volume (%)
Size (r.nm)
Size Distribution by Volume
A
0
2
4
6
8
10
12
14
16
18
20
22
0.1 1 10 100
Size (r.nm)

Volume (%)
Size Distribution by Volume
B
Fig. 5. Size distribution by volume of (A) Ba-TK at a concentration of 3 mgÆmL
)1
(red), 1 mgÆmL
)1
(green) and 1 mgÆmL
)1
+2 mM ATP (blue).
The molecular masses calculated from the hydrodynamic radius (4.3 nm for Ba-TK at 3 mgÆmL
)1
and 3.2 nm for Ba- TK at 1 mgÆmL
)1
+2 mM
ATP) are 104 kDa (tetramer) and 51 kDa (dimer), respectively. (B) Bc-TK at a concentration of 3 mgÆmL
)1
(red), 1 mgÆmL
)1
(green) and
1mgÆmL
)1
+2 mM ATP (blue). The molecular mass calculated from the hydrodynamic radius (3.1–3.3 nm for all Bc-TK samples) is 46 kDa,
corresponding to a dimer.
U. Kosinska et al. Bacillus thymidine kinase structures
FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS 733
The lasso-domains of Ba-TK and Bc-TK show a
closed and an open form, respectively. In Ba-TK, the
lasso-loop is closed down over thymidine and stabil-
ized by hydrogen bond donors and acceptors of the

substrate, whereas lack of substrate in Bc-TK leaves
the lasso-loop flexible. The open conformation of the
lasso-domain in Bc-TK resembles the lasso-loop of
Ca-TK [17], where part of the loop is missing because
of flexibility. In contrast with Ca-TK where the active
site is empty and there are no crystallographic interac-
tions to stabilize the loop, the lasso of Bc-TK could be
fully traced because of stabilizing crystal-packing inter-
actions and MPD in the substrate site.
The conformation of the P-b-hairpin observed in the
Bc-TK structure is similar to that seen in Ca-TK in
complex with ADP, Uu-TK in complex with dTTP,
and one molecule of Uu-TK in complex with dT
[9,14,17]. Together with the lasso-domain, the P-b-hair-
pin shields the active site from the solvent. The confor-
mations of the P-b-hairpin region observed in
subunit A in hTK1 (1W4R) and subunit A in Uu-TK
(2B8T) differ significantly from the structures men-
tioned above [14]. They may have arisen because of
crystal contacts. Nevertheless, lack of electron density
or highly diverse conformations when the phosphate
donor is absent followed by formation of a b-hairpin
in the presence of a phosphate donor indicates that
this region of the molecule takes part in the creation
of a phosphate donor–acceptor complex. Residues at
the tip of the hairpin form hydrogen bonds to the
a-phosphate of dTTP bound as phosphate donor.
In the literature, TK1-like enzymes have been des-
cribed as both dimers and tetramers. The oligomeric
state of TK has been studied under various conditions

with divergent results [19–23]. In early studies of hTK1
derived from HeLa cells, it was concluded from sedi-
mentation in glycerol gradients and gel filtration that
hTK1 was a tetramer in solution [23]. Munch-Petersen
et al. [22] showed that tetramerization at low protein
concentration was an ATP-dependent process in which
incubation of hTK1 with ATP reversibly shifted the
oligomeric state from a dimer to a tetramer. A recent
study of the quaternary state of hTK1 demonstrated
that the oligomeric state of wild-type hTK1 is tetra-
meric, whereas the N-terminal and C-terminal trun-
cated mutant was a dimer irrespective of the
concentration and incubation with dT ⁄ ATP [19].
Furthermore, Uu-TK was found as dimers in solu-
tion [21]. To investigate if the enzyme could be found
as dimers and tetramers, we performed further experi-
ments with the Bacillus enzymes. Earlier gel-filtration
studies have shown that Ba-TK occurs as dimers [15],
whereas the oligomeric state for Bc-TK had not been
determined. Our gel-filtration studies of both enzymes
demonstrated that both were in the tetrameric state
only (supplementary Fig. S3). This is surprising as the
conditions in the present and earlier experiments were
similar. However, the salt and protein concentrations
differ somewhat.
We also investigated the enzyme with dynamic light
scattering methods. The results in this case were as
puzzling as the gel-filtration studies. In only one case,
Ba-TK at 3 mgÆmL
)1

, did the enzyme occur as tetra-
mers. In all other cases, we obtained dimers. The pres-
ence of ATP did not influence the quaternary structure.
However, the available TK structures, including
truncated hTK1, are tetramers with comparable sub-
unit–subunit interactions. These results suggest that
TK exists as both dimers and tetramers. In solution,
the two forms may be in equilibrium, which in crystal-
lization conditions, with high protein concentrations
and small volumes, is shifted towards the tetrameric
form. The observation that, in Bc-TK, the base of
dTTP in the phosphate donor position is inserted
between two subunits implies that the enzyme was in
the dimeric form before the formation of the tetramer
and that it can be stable in both states.
As mentioned above, all tetramers found in the crys-
tal structures are comparable. This study reveals one
major difference. Empty phosphate donor sites make it
possible for the subunits to form tight tetramers,
whereas, when the phosphate donor sites are occupied,
as in Bc-TK, the subunits interacting with the long a1
helix are not able to come as close together to make
the tight tetramer. On the other hand, the base of the
phosphate donor, which is stuck between two subunits,
contributes to the subunit–subunit interaction and
helps to create the tetrameric state of the enzyme.
However, we are not able to determine in this study
whether the phosphate donor induces separation of a
tetramer formed earlier or if it prevents formation of a
tight tetramer. Is the observed position of the phos-

phate donor with the base inserted between the sub-
units the physiological one? If the enzyme acts as a
dimer, it seems likely. However, for a tetramer, separ-
ation into dimers to allow insertion of the base would
be unfavorable.
Moreover, we would like to establish whether it is
possible from structural data to determine which of
the subunit–subunit interactions is the most stable. In
contrast with Birringer et al. [8], we believe that dimers
with the b-sheet interaction area are the ones that are
observed in solution. There are a number of structural
indications that support this conclusion. The interac-
tion area formed between the b-sheets is larger than
the one formed between the long a-helices and includes
Bacillus thymidine kinase structures U. Kosinska et al.
734 FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS
conserved salt-bridge interactions, which together with
water molecules hydrogen-bonded between the two
sheets, make strong contacts. As described above,
binding of the phosphate donor does not alter the
b-sheet interaction, but significantly decreases the
a-helical interaction area. This has also been observed
in the structure of Ca-TK complexed with ADP [17],
but not in any other TK structure with unoccupied
phosphate-donor-binding sites. A longer distance
between the adjacent a1-helices is necessary for the
base of the phosphate donor to have space to fit
between Phe18 and Phe34. Yet another indication that
the b-sheet interaction is the most stable one comes
from the Uu-TK structure. In contrast with hTK1, the

structure of Uu-TK was determined for the full-length
enzyme where the C-terminus forms an a-helix, which
makes hydrophobic interactions with a2 and a3 of the
neighboring subunit in the b-sheet interaction dimer.
This increases the interaction area from 870 A
˚
2
to
2040 A
˚
2
. Formation of an a-helical dimer as suggested
by Birringer et al. [8] would leave the hydrophobic pat-
ches of the C-terminal helix unshielded.
To conclude, we have described the structures of
two almost identical TKs showing open and closed
conformations with respect to phosphate acceptor and
donor sites. This study shows that an empty TK
adopts an open form with two flexible parts. When the
substrate, feedback inhibitor, and phosphate donor are
all not bound to the enzyme, both the lasso-loop and
the P-b-hairpin are flexible. The lasso-loop acts as a lid
and traps the substrate or the inhibitor, whereas the
P-b-hairpin plays a part in the positioning of the phos-
phate donor. Moreover, we have illustrated that the
tetramer formed in the presence of the phosphate
donor is weaker than that observed in TKs with empty
phosphate-binding sites.
Experimental procedures
Protein expression and purification

The B. anthracis Sterne strain (34F2) tk gene was cloned
into the pCRÒ4-TOPO vector, subcloned into the pET-14b
expression vector and transformed into the chemically com-
petent Escherichia coli strain, BL21(DE3) pLysS (Novagen,
Madison, WI, USA). Recombinant Ba-TK, fused with an
N-terminal His
6
-tag, was expressed by induction with iso-
propyl b-d-thiogalactoside and purified by metal ion affin-
ity chromatography (HisBind; Novagen) as previously
described [15]. The buffers were changed to 20 mm
Tris ⁄ HCl (pH 7.6) ⁄ 150 mm NaCl ⁄ 2mm MgCl
2
⁄ 5mm di-
thiothreitol ⁄ 10% glycerol on a PD10 desalting column (GE
Healthcare, Uppsala, Sweden).
The tdk gene from B. cereus (GenBank accession No.
DQ384595) was cloned into the pGEX-2T vector (GE
Healthcare) and expressed in E. coli KY895 [24]. Recombin-
ant N-terminal glutathione S-transferase (GST)-tagged
Bc-TK fusion protein was expressed by induction with iso-
propyl b-d-thiogalactoside and subsequently purified on a
glutathione–Sepharose FF column (GE Healthcare). Cell-
free extract prepared by French Press, centrifugation (Sor-
vall RC5 centrifuge, SA-600 rotor, 30 min at 24 500 g) and
filtering was loaded on the column, and unbound proteins
eluted with buffer A (NaCl ⁄ P
i
pH 7.3, 10% glycerol,
0.1% Triton X-100). Then one column volume 10 mm

ATP ⁄ MgCl
2
in buffer A was circulated over the column for
1 h at room temperature and washed through with buffer A.
Finally, one column volume 50 UÆmL
)1
thrombin in buf-
fer A was loaded on the column before it was incubated at
room temperature for 16 h. Pure Bc-TK without the GST-
tag was then eluted from the column with buffer A while
uncleaved fusion protein and GST-tag remained on the col-
umn for elution with glutathione. The protein was further
purified on a SuperdexÔ200 HiLoadÔ16 ⁄ 60 prepgrade
(GE Healthcare). The buffer used was 10 mm Tris ⁄ HCl
(pH 7.6) ⁄ 150 mm NaCl ⁄ 5mm MgCl
2
⁄ 5mm dithiothreitol.
Determination of quaternary structure with
dynamic light scattering
Three samples of different concentration with and without
ATP were prepared of each Bacillus TK. The concentra-
tions were 3 mgÆmL
)1
TK, 1 mgÆmL
)1
TK and 1 mgÆmL
)1
TK with 2 mm ATP. The buffer was 20 mm Tris ⁄ HCl
(pH 7.6) ⁄ 150 mm NaCl ⁄ 5mm MgCl
2

⁄ 5mm dithiothreitol.
The samples were incubated for 8 h at 4 °C before the
measurements. All measurements were performed at 4 °C
with a Malvern Instruments (Uppsala, Sweden) dynamic
light scattering instrument.
Crystallization, data collection and structure
determination
Ba-TK (10 mgÆmL
)1
) was cocrystallized with dT (5 mm).
The crystals grew in hanging drops at 14 °C in 5–24%
(depending on the type of crystallization plate) 1,2-pro-
panediol, 0.1 m sodium ⁄ potassium phosphate (pH 6) ⁄ 10%
(v ⁄ v) glycerol. The data were collected at the European
Synchrotron Radiation Facilities (ESRF), Grenoble,
France. No cryo-protectant was used before freezing in
liquid nitrogen.
Data were processed with Mosflm and Scala in CCP4 [25].
The search model for molrep [26] was a polyalanine chain
of one subunit of Uu-TK (1XMR or 2B8T) with residues
50–67, 84–88 and 199–217 excluded because of structural
flexibility and sequential dissimilarity. One Ba-TK subunit
was found in the asymmetric unit. Simulated annealing was
U. Kosinska et al. Bacillus thymidine kinase structures
FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS 735
performed in cns [27], with further refinement and model
building performed with refmac5 [28] and o [29], respect-
ively. During the end of the refinement, translation libera-
tion and screw rotation displacement (TLS) refinement was
applied with all residues forming one TLS group. Twenty

water molecules were included. Statistical analysis of data
processing and refinement are reported in Table 1. The
coordinates have been deposited with PDB code 2J9R.
Bc-TK (10–15 mgÆmL
)1
) was mixed with dTTP (5 mm)
before crystallization setup. Equal volumes of protein solu-
tion were mixed with crystallization solution consisting of
55–65% MPD and 0.1 m Hepes, pH 7.0. Bipyramidal crys-
tals grew using hanging drop vapor diffusion at 4 ° C. No
cryo-protecting agent was used, and the crystals were fro-
zen in liquid nitrogen directly from crystallization drops.
The data were collected at ESRF. Indexing and scaling
were performed with Mosflm and Scala in CCP4 [25].
Molecular replacement calculations were performed with
molrep [26] using one subunit of Ba-TK as search model.
As in the case of Ba-TK, there was only one subunit of
Bc-TK occupying the asymmetric unit. Cycles of model
building, restrained refinement, and TLS refinement were
performed in o [29] and refmac5 [28], respectively. Twenty
water molecules were included in the final structure. Data
collection and refinement statistics are presented in Table 1.
The structure has been deposited with PDB code 2JA1.
The sequence alignment was performed in clustalw [30],
and the figure was made in alscript [31]. The figures pre-
senting the protein structures were made with pymol [32].
Acknowledgements
This work was supported by grants from the Swedish
Research Council for Environment, Agricultural Sci-
ences and Spatial Planning (to LY and SE), the Swe-

dish Research Council (to HE, JP and SE) and the
Swedish Cancer Foundation (to HE).
References
1 Dixon TC, Meselson M, Guillemin J & Hanna PC
(1999) Anthrax. N Engl J Med 341, 815–826.
2 Kotiranta A, Lounatmaa K & Haapasalo M (2000) Epi-
demiology and pathogenesis of Bacillus cereus infec-
tions. Microbes Infect 2, 189–198.
3 Okinaka R, Cloud K, Hampton O, Hoffmaster A, Hill
K, Keim P, Koehler T, Lamke G, Kumano S, Manter
D, et al. (1999) Sequence, assembly and analysis of
pX01 and pX02. J Appl Microbiol 87, 261–262.
4 Okinaka RT, Cloud K, Hampton O, Hoffmaster AR,
Hill KK, Keim P, Koehler TM, Lamke G, Kumano S,
Mahillon J, et al. (1999) Sequence and organization of
pXO1, the large Bacillus anthracis plasmid harboring
the anthrax toxin genes. J Bacteriol 181, 6509–6515.
5 Sandrini MP, Clausen AR, Munch-Petersen B & Piskur
J (2006) Thymidine kinase diversity in bacteria. Nucleo-
sides Nucleotides Nucleic Acids 25, 1153–1158.
6 Johansson K, Ramaswamy S, Ljungcrantz C, Knecht
W, Piskur J, Munch-Petersen B, Eriksson S & Eklund
H (2001) Structural basis for substrate specificities of
cellular deoxyribonucleoside kinases. Nat Struct Biol 8,
616–620.
7 Sabini E, Ort S, Monnerjahn C, Konrad M & Lavie A
(2003) Structure of human dCK suggests strategies to
improve anticancer and antiviral therapy. Nat Struct
Biol 10, 513–519.
8 Birringer MS, Claus MT, Folkers G, Kloer DP, Schulz

GE & Scapozza L (2005) Structure of a type II thymidine
kinase with bound dTTP. FEBS Lett 579, 1376–1382.
9 Welin M, Kosinska U, Mikkelsen NE, Carnrot C, Zhu
C, Wang L, Eriksson S, Munch-Petersen B & Eklund H
(2004) Structures of thymidine kinase 1 of human and
mycoplasmic origin. Proc Natl Acad Sci USA A101,
17970–17975.
10 Knecht W, Sandrini MP, Johansson K, Eklund H,
Munch-Petersen B & Piskur J (2002) A few amino acid
substitutions can convert deoxyribonucleoside kinase
specificity from pyrimidines to purines. EMBO J 21,
1873–1880.
11 Eriksson S, Munch-Petersen B, Johansson K & Eklund
H (2002) Structure and function of cellular deoxyribo-
nucleoside kinases. Cell Mol Life Sci 59, 1327–1346.
12 Sandrini MP & Piskur J (2005) Deoxyribonucleoside
kinases: two enzyme families catalyze the same reaction.
Trends Biochem Sci 30, 225–228.
13 El Omari K, Solaroli N, Karlsson A, Balzarini J &
Stammers DK (2006) Structure of vaccinia virus thymi-
dine kinase in complex with dTTP: insights for drug
design. BMC Struct Biol 6, 22.
14 Kosinska U, Carnrot C, Eriksson S, Wang L & Eklund
H (2005) Structure of the substrate complex of thymi-
dine kinase from Ureaplasma urealyticum and investiga-
tions of possible drug targets for the enzyme. FEBS J
272, 6365–6372.
15 Carnrot C, Vogel SR, Byun Y, Wang L, Tjarks W,
Eriksson S & Phipps AJ (2006) Evaluation of Bacillus
anthracis thymidine kinase as a potential target for the

development of antibacterial nucleoside analogs. Biol
Chem 387
, 1575–1581.
16 Welin M, Wang J, Eriksson S & Eklund H. (2006)
Structure-function analysis of a bacterial deoxyadeno-
sine kinase reveals the basis for substrate specificity.
J Mol Biol, doi: 10.1016/j.jmb.2006.12.010.
17 Kuzin AP, Abashidze M, Forouhar F, Vorobiev SM,
Acton TB, Ma LC, Xiao R, Montelione GT, Tong L &
Hunt JF. (2006) X-ray structure of Clostridium acetobu-
tylicum thymidine kinase with ADP. Northeast Struc-
tural Genomics Target Car26, PDB code: 1XXX6.
Bacillus thymidine kinase structures U. Kosinska et al.
736 FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS
18 Krissinel E & Henrick K (2005) Detection of protein
assemblies in crystals. Lecture Notes in Computer
Science: Computational Life Sciences: Proceedings of the
First International Symposium, Complife2005, Konstanz,
Germany, September 25–27, 2005 3695, 163–174.
19 Birringer MS, Perozzo R, Kut E, Stillhart C, Surber W,
Scapozza L & Folkers G (2006) High-level expression
and purification of human thymidine kinase 1: quatern-
ary structure, stability, and kinetics. Protein Expr Purif
47, 506–515.
20 Black ME & Hruby DE (1990) Quaternary structure of
vaccinia virus thymidine kinase. Biochem Biophys Res
Commun 169, 1080–1086.
21 Carnrot C, Wehelie R, Eriksson S, Bolske G & Wang L
(2003) Molecular characterization of thymidine kinase
from Ureaplasma urealyticum: nucleoside analogues as

potent inhibitors of mycoplasma growth. Mol Microbiol
50, 771–780.
22 Munch-Petersen B, Tyrsted G & Cloos L (1993) Rever-
sible ATP-dependent transition between two forms of
human cytosolic thymidine kinase with different enzy-
matic properties. J Biol Chem 268, 15621–15625.
23 Sherley JL & Kelly TJ (1988) Human cytosolic thymi-
dine kinase. Purification and physical characterization
of the enzyme from HeLa cells. J Biol Chem 263, 375–
382.
24 Igarashi K, Hiraga S & Yura T (1967) A deoxythymi-
dine kinase deficient mutant of Escherichia coli. II.
Mapping and transduction studies with phage phi 80.
Genetics 57, 643–654.
25 Collaborative Computational Project, number 4 (1994)
The CCP4 suite: programs for protein crystallography.
Acta Crystallogr D Biol Crystallogr 50, 760–763.
26 Vagin A & Teplyakov A (1997) MOLREP: an auto-
mated program for molecular replacement. J Appl Crys-
tallogr 30, 1022–1025.
27 Bru
¨
nger AT, Adams PD, Clore GM, DeLano WL, Gros
P, Grosse-Kunstleve RW, Jiang JS, Kuszewski J, Nilges
M, Pannu NS, et al. (1998) Crystallography & NMR
system: a new software suite for macromolecular struc-
ture determination. Acta Crystallogr D Biol Crystallogr
54, 905–921.
28 Murshudov GN, Vagin AA & Dodson EJ (1997)
Refinement of macromolecular structures by the maxi-

mum-likelihood method. Acta Crystallogr D Biol Crys-
tallogr 53, 240–255.
29 Jones TA, Zou JY, Cowan SW & Kjeldgaard M (1991)
Improved methods for building protein models in elec-
tron density maps and the location of errors in these
models. Acta Crystallogr 47, 110–119.
30 Chenna R, Sugawara H, Koike T, Lopez R, Gibson TJ,
Higgins DG & Thompson JD (2003) Multiple sequence
alignment with the Clustal series of programs. Nucleic
Acids Res 31, 3497–3500.
31 Barton GJ (1993) ALSCRIPT: a tool to format multiple
sequence alignments. Protein Eng 6, 37–40.
32 DeLano WL (2002) The PyMOL Molecular Graphics
System. DeLano Scientific, San Carlos, CA.
Supplementary material
The following supplementary material is available
online:
Fig. S1. Crystal contacts stabilize the lasso-loop.
Fig. S2. MPD binding in the phosphate acceptor site
of Bc-TK.
Fig. S3. (A) Size-exclusion chromatography of Ba-TK
on Superdex 200 10 ⁄ 300. (B) Size-exclusion chro-
matography of Bc-TK on Superdex 200 16 ⁄ 60. (C) A
typical chromatogram of Bc-TK run on Superdex 200
16 ⁄ 60.
This material is available as part of the online article
from
Please note: Blackwell Publishing is not responsible
for the content or functionality of any supplementary
materials supplied by the authors. Any queries (other

than missing material) should be directed to the corres-
ponding author for the article.
U. Kosinska et al. Bacillus thymidine kinase structures
FEBS Journal 274 (2007) 727–737 ª 2006 The Authors Journal compilation ª 2006 FEBS 737