Structural and mechanistic insight into covalent substrate binding by Escherichia coli dihydroxyacetone kinase

Article abstract of DOI:10.1073/pnas.1012596108

The Escherichia coli dihydroxyacetone (Dha) kinase is an unusual kinase because (i) it uses the phosphoenolpyruvate carbohydrate: phosphotransferase system (PTS) as the source of high-energy phosphate, (ii) the active site is formed by two subunits, and (iii) the substrate is covalently bound to His218^K* of the DhaK subunit. The PTS transfers phosphate to DhaM, which in turn phosphorylates the permanently bound ADP coenzyme of DhaL. This phosphoryl group is subsequently transferred to the Dha substrate bound to DhaK. Here we report the crystal structure of the E. coli Dha kinase complex, DhaK-DhaL. The structure of the complex reveals that DhaK undergoes significant conformational changes to accommodate binding of DhaL. Combined mutagenesis and enzymatic activity studies of kinase mutants allow us to propose a catalytic mechanism for covalent Dha binding, phosphorylation, and release of the Dha-phosphate product. Our results show that His56^K is involved in formation of the covalent hemiaminal bond with Dha. The structure of H56N ^K with noncovalently bound substrate reveals a somewhat different positioning of Dha in the binding pocket as compared to covalently bound Dha, showing that the covalent attachment to His218^K orients the substrate optimally for phosphoryl transfer. Asp109^K is critical for activity, likely acting as a general base activating the γ-OH of Dha. Our results provide a comprehensive picture of the roles of the highly conserved active site residues of dihydroxyacetone kinases.

Full text of DOI:10.1073/pnas.1012596108

Structural and mechanistic insight into covalent
substrate binding by Escherichia coli
dihydroxyacetone kinase
Rong Shia, Laura McDonaldb, Qizhi Cuic, Allan Mattec, Miroslaw Cyglera,c, and Irena Ekielb,c,1
aDepartment of Biochemistry, McGill University, Montréal, QC, Canada H3G 1Y6; bDepartment of Chemistry and Biochemistry, Concordia University,
c
Montréal, QC, Canada H4B 1R6; and Biotechnology Research Institute, National Research Council of Canada, 6100 Royalmount Avenue, Montréal, QC,
Canada H4P 2R2
Edited by Gregory A. Petsko, Brandeis University, Waltham, MA, and approved November 30, 2010 (received for review August 26, 2010)
The Escherichia coli dihydroxyacetone (Dha) kinase is an unusual
kinase because (i) it uses the phosphoenolpyruvate carbohydrate:
phosphotransferase system (PTS) as the source of high-energy
phosphate, (ii) the active site is formed by two subunits, and (iii)
the substrate is covalently bound to His218K* of the DhaK subunit.
The PTS transfers phosphate to DhaM, which in turn phosphory-
lates the permanently bound ADP coenzyme of DhaL. This phos-
phoryl group is subsequently transferred to the Dha substrate
bound to DhaK. Here we report the crystal structure of the E. coli
Dha kinase complex, DhaK–DhaL. The structure of the complex
reveals that DhaK undergoes significant conformational changes
to accommodate binding of DhaL. Combined mutagenesis and
enzymatic activity studies of kinase mutants allow us to propose
a catalytic mechanism for covalent Dha binding, phosphorylation,
and release of the Dha-phosphate product. Our results show that
His56K is involved in formation of the covalent hemiaminal bond
with Dha. The structure of H56NK with noncovalently bound
substrate reveals a somewhat different positioning of Dha in the
binding pocket as compared to covalently bound Dha, showing
that the covalent attachment to His218K orients the substrate op-
timally for phosphoryl transfer. Asp109K is critical for activity, likely
acting as a general base activating the γ-OH of Dha. Our results
provide a comprehensive picture of the roles of the highly con-
served active site residues of dihydroxyacetone kinases.
freundii (5), which has 30% sequence identity to the E. coli kinase.
Both ATP- and PTS-dependent Dha kinases display high struc-
tural similarity of their corresponding domains/subunits and are
structurally distinct from other kinases. The active site residues
are highly conserved between bacteria and eukaryotes reflecting
the unique phosphotransfer mechanism. The structure of the
homodimeric, ATP-dependent C. freundii Dha kinase shows swap-
ped K- and L-domains, with bound Dha and an ATP analogue
separated by approximately 14 Å (5). It is therefore unlikely that
this structure represents a phosphotransfer-competent state of
this enzyme, leaving the question of the mechanism unanswered.
In addition to the structures of DhaK and DhaL, the structure of
the Lactococcus lactis DhaM domain has been reported as well as
its complex with DhaL, providing insight into phosphotransfer
from phosphohistidine to ADP (6).
We have determined the crystal structure of the E. coli DhaK–
DhaL complex bound to ADP and Dha at 2.2-Å resolution,
providing a snapshot of the final step of phosphoryl transfer from
ATP to Dha. Site-directed mutagenesis, activity measurements,
and structural data indicate that His56K is important for forma-
tion of the covalent hemiaminal bond. We show further that the
covalent attachment of Dha to His218K optimally positions the
substrate for phosphoryl transfer but is likely dispensable for en-
zyme activity. Together, these data allow us to propose a general
catalytic mechanism for dihydroxyacetone kinases, consistent
with the available structural and biochemical data.
DhaK–DhaL complex ∣ kinase mechanism
ihydroxyacetone phosphate (Dha-P), an intermediate for the
Results and Discussion
D
synthesis of pyruvate, can be generated in bacteria through
Structure of the DhaK–DhaL Complex Reveals Conformational Rear-
rangements upon Complex Formation. Each asymmetric unit con-
tains one molecule each of DhaK and DhaL with all residues
well defined in the electron density map. The biological unit
of the complex is a heterotetramer containing a central DhaK
dimer and two molecules of DhaL (2∶2 stoichiometry). Each
DhaL subunit interacts with only one DhaK subunit of the dimer,
with the two DhaL subunits being approximately 37 Å apart
(Fig. 1A). An ADP molecule and two magnesium ions are bound
to DhaL, whereas DhaK contains a Dha molecule covalently
bound to His218K (Fig. 1B).
glycerol fermentation by the sequential actions of glycerol dehy-
drogenase forming dihydroxyacetone and Dha kinase generating
Dha-P. Dha kinases are divided into two classes, those using ATP
as the phosphoryl donor or those using the PTS-dependent
DhaM protein to provide the phosphoryl group (1). The ATP-
dependent Dha kinases are present in animals, plants, and some
bacteria and consist of a single chain containing two domains,
whereas the PTS-dependent Dha kinases exist only in bacteria
and are composed of two subunits. The most studied represen-
tative of the PTS-dependent family is the Escherichia coli Dha
kinase, which is composed of DhaK and DhaL subunits. DhaM,
the phosphotransferase subunit of Dha kinase, is phosphorylated
by the small phospho-carrier protein HPr of the PTS (1). Phos-
phorylated DhaM then forms a complex with the ADP-bound
DhaL subunit and transiently phosphorylates ADP to ATP (1).
The ATP-loaded DhaL subsequently associates with the DhaK
subunit containing the Dha substrate covalently bound to
His218K* through a hemiaminal bond (2). Finally, ATP bound to
DhaL, serving as a coenzyme (3), transfers phosphate to Dha
yielding Dha-P.
Author contributions: R.S., A.M., and I.E. designed research; R.S., L.M., Q.C., and A.M.
performed research; R.S., L.M., Q.C., A.M., and I.E. analyzed data; and R.S., L.M., Q.C.,
A.M., M.C., and I.E. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Data deposition: The atomic coordinates and structure factors have been deposited in the
Protein Data Bank, www.pdb.org [PDB ID codes 3PNK (DhaK), 3PNL (DhaK–DhaL complex),
3PNM (DhaK H56A), 3PNO (DhaK H56N), and 3PNQ ( DhaK H56N-Dha)].
K
*Superscript letter (or L) denotes kinase subunit.
Crystal structures for the two individual subunits of the E. coli
PTS-dependent Dha kinase—DhaK (2) and DhaL (4)—have
been previously determined, as well as the structure of an ATP-
dependent single chain, two-domain Dha kinase from Citrobacter
1To whom correspondence should be addressed. E-mail: irena.ekiel@nrc.ca.
This article contains supporting information online at www.pnas.org/lookup/suppl/
doi:10.1073/pnas.1012596108/-/DCSupplemental.
1302–1307 ∣ PNAS ∣ January 25, 2011 ∣ vol. 108 ∣ no. 4
www.pnas.org/cgi/doi/10.1073/pnas.1012596108

DhaL binds to DhaK through the narrower end of an eight-
helix barrel that harbors the ADP. This is consistent with the role
of DhaL-ATP as the phosphoryl donor to the DhaK-bound Dha.
The structure of DhaL in the complex is very similar to that of
free DhaL (4), with an rmsd of 0.42 Å for all Cα atoms. Binding to
DhaK involves residues in loops α1L∕α2L, α3L∕α4L, α7L∕α8L, and
portions of helices α1L, α2L, α4L, and α5L (Fig. 1C). In contrast,
DhaK undergoes significant conformational changes in several
regions as compared to the free DhaK previously reported
[PDB ID code 1OI2, (2)] (Fig. 2A). The rmsd for the Cα atoms
of free and complexed DhaK is 1.38 Å (for 336 aligned residues).
The loop β6K∕α4K undergoes a significant relocation to accom-
modate the narrow tip of DhaL, with the Cα atoms of Leu142K
and Tyr143K moving by approximately 13 Å. Removing this loop
from the comparison results in an rmsd of 0.5 Å for 328 Cα atoms.
The loop β8K∕β9K moves by approximately 3 Å at Pro221K to
accommodate Leu182L. A consequence of the adjustment of loop
β8K∕β9K is the ordering of the neighboring loop β7K∕β8K, which
is disordered in the free DhaK (2). This ordering effect also in-
cludes the N-terminal segment of DhaK (residues 2–9), which is
0
0
now sandwiched between the β7K∕β8K loop and the β9K ∕α6K
loop of the other DhaK subunit and forms part of the dimeriza-
tion interface. DhaK residues involved in the interface are found
in loops β4K∕α2K, β6K∕α4K, β7K∕β8K, β8K∕β9K and in helix α3K
(for more details see SI Text).
Although the overall structure of both DhaK and DhaL
subunits in the E. coli complex are similar to the corresponding
domains of the ATP-dependent C. freundii Dha kinase, with rmsd
values of 1.4 Å (307 Cα atoms) and 1.7 Å (162 Cα atoms), respec-
tively, our crystal structure reveals that the binding interface be-
tween these two subunits/domains is quite different as a result of
Fig. 1. Crystal structure of the E. coli DhaK–DhaL complex. (A) Cartoon repre-
sentation of the DhaK–DhaL heterotetramer. The DhaL molecules are shown
in cyan with the two DhaK subunits A and B colored yellow and mauve, respec-
tively. The Dha molecules covalently bound to His218K are colored green. The
ADP molecules and magnesium ions bound to DhaL are shown as red sticks and
magenta spheres, respectively. (B) Active site of the DhaK–DhaL complex. The
DhaK (yellow) and DhaL (cyan) molecules are shown in cartoon representa-
tion. The ADP molecule and residues described in the text are shown in stick
mode. The two magnesium ions are shown as magenta spheres. Interactions
involving the Dha molecule and the two metal ions are indicated by green
dashed lines. The distance between the γ-OH group of Dha and the oxygen
atom of the ADP β-phosphate is 4.8 Å, which is shown as an orange dashed
line. The difference (Fo − Fc) electron density map for Dha covalently linked
to His218K as well as ADP (omitted from refinement) contoured at 4σ is shown
in green. (C) Binding interface and interactions between DhaK and DhaL. The
color scheme is the same as B. Hydrogen bonding interactions are shown
as magenta dashed lines. The water molecule mediating the interaction
between Y181L and N105K is shown as a green sphere.
Fig. 2. Conformational changes in the DhaK subunit induced by complex
formation. (A) Uncomplexed forms of DhaK (PDB ID code 1OI2, shown in pale
green) and DhaL (PDB ID code 2BTD, shown in wheat) superimposed onto
their respective subunits from the DhaK–DhaL complex (yellow and cyan, re-
spectively). The four regions in DhaK that undergo significant conforma-
tional changes are colored in magenta. The second subunit of DhaK is
shown in white. (B) Model of the DhaK–DhaL(ATP) complex refined by
molecular dynamics. The ATP molecule, two magnesium ions, and active site
residues are shown. In this model, the γ-phosphoryl atom of ATP is located
3.9 Å away (indicated by the orange dashed line) from the γ-OH group of
Dha. The interactions between the Dha molecule and the residues His56K
and Asp109K, which we show are important in catalysis, are shown as green
dashed lines.
Shi et al.
PNAS
∣
January 25, 2011
∣
vol. 108
∣
no. 4
∣
1303

domain swapping in C. freundii (Fig. S1A). Importantly, and in
contrast to the C. freundii Dha kinase structure, in the E. coli
structure, the orientation of DhaL relative to DhaK is consistent
with a direct transfer of the γ-phosphoryl moiety from ATP to the
γ-OH group of Dha (Fig. 1B).
We have also crystallized E. coli DhaK in a different crystal
form (Table S1) than that previously reported (2). Interestingly,
in our free DhaK structure, the loop β6K∕α4K (residues 138–145)
displays a conformation intermediate between that of free DhaK
(PDB ID code 1OI2) and that of DhaL-bound DhaK (see above),
providing a snapshot of the movement of this loop (Fig. S2). The
flexibility of the loop β6K∕α4K is pivotal for the ability of DhaK to
bind DhaL, because conformations of this loop in both free DhaK
structures are not compatible with complex formation. The dif-
ference between the two structures of free DhaK is likely caused
by changes in packing environment.
Table 1. Kinetic constants of wild-type DhaK and His56K mutants
for Dha substrate
WT
H56AK
H56NK
kcat
Km
(min−1
)
507 25
<0.008*
96.1 2.0
0.302 0.005
318
223 4.3
2.20 0.15
101
app
app
(mM)
app
kcatapp∕Km
>6.3 × 104
(mM−1 min−1
)
Standard deviations were calculated from three independent experiments.
*Due to the limited sensitivity of the glycerol-3-phosphate dehydrogenase
app
coupled assay for WT at low substrate concentrations the Km
could
not be accurately determined.
or alanine (2) also abolishes activity, consistent with its role in
binding and orienting the substrate. In addition to the coupled en-
zyme assay, similar results were obtained by NMR spectroscopy
(Fig. S6).
The Dha Kinase Active Site Is Formed at the Interface of DhaL-DhaK
Subunits. In order to bring DhaL-bound ADP in proximity
to DhaK-bound Dha, the tip of the DhaL helical barrel contain-
ing loop α7L∕α8L (residues 175–185) that caps the nucleotide-
binding site, dips into the cavity formed between the β6K∕α4K
and β8K∕β9K loops, displacing both loops from their locations
observed in free DhaK (Fig. 2A). The distance between the
β-phosphate oxygen of ADP and the γ-OH group of Dha is
4.8 Å, sufficient to accommodate the missing ATP γ-phosphate
in our DhaK–DhaL(ADP) complex as shown by our model of
DhaK–DhaL(ATP) (Fig. 2B).
Structures of DhaK H56NK and H56AK Reveal the Ability of Dha to Bind
Noncovalently. To gain further insight into the reduced activity
exhibited by the H56NK and H56AK mutant enzymes we deter-
mined their crystal structures. The overall structure of either the
H56NK or H56AK enzyme is almost identical to that of the pre-
viously reported structure of wild-type DhaK (2), although they
crystallized in a different packing environment (Table S1) with
four subunits in the asymmetric unit. However, in contrast to
the structures of wild-type DhaK, no electron density correspond-
ing to the Dha substrate was observed in any of the subunits in the
structures of these mutant enzymes (Fig. S7). These structures
were immediately suggestive of a critical role for His56K in hemi-
aminal bond formation with Dha. The H56NK and H56AK struc-
tures reveal very small conformational changes in the active site
as compared to wild-type DhaK with covalently bound Dha.
These differences include an approximatley 35° rotation of the
imidazole ring of His218K, possibly resulting from the shorter
Asn56K and Ala56K side chains in the mutant enzyme structures.
The carboxylate group of Asp109K also rotates approximatley
35°, possibly due to the loss of its hydrogen bonding interaction
with the hydroxyl groups of Dha.
Both subunits contribute residues to form the active site. The
key residues in the Dha binding pocket, including His218K cova-
lently bound to Dha, Gly53K, His56K, Lys104K, and Asp109K, are
in almost identical positions as in the free DhaK structure. Dha is
further anchored by Arg178L, whose guanidinium group hydro-
gen bonds to the γ-OH of Dha and likely stabilizes the trigonal
bipyramidal transition state during phosphoryl transfer. All of
these residues are highly conserved in bacterial Dha kinases.
Mutations of His218K in E. coli DhaK or Arg161L in L. lactis
DhaL (corresponding to Arg178L in E. coli) result in inactive
enzymes (2, 6).
To better understand how the H56NK mutant enzyme could
retain some activity in the absence of covalent substrate binding,
we determined the crystal structure of H56NK in the presence of
Dha by supplementing the crystallization buffer with 10 mM Dha.
These crystals contained four molecules in the asymmetric unit
(Table S1). In one subunit there is clear electron density showing
the presence of Dha bound in the active site (Fig. 3A). Although
this non-covalently bound Dha molecule occupies the same
binding pocket as in the covalently bound structure, its C2 and
O2 atoms are displaced by approximately 2 Å and approximately
3 Å, respectively, as compared with covalently bound Dha. The
absence of the covalent bond results in somewhat different inter-
actions between Dha and the surrounding residues. The Dha car-
bonyl group is anchored via hydrogen bonds with the main chain
amide group of Gly53K; the hydroxyl group (α-OH) buried in the
binding pocket is within H-bonding distance of Lys104KðNZÞ and
His218KðNE2Þ, while the other exposed hydroxyl group (γ-OH)
forms a hydrogen bond to Asp109KðOD1Þ and Ser80KðOGÞ
(Fig. 3A). Although this γ-OH is still orientated toward the
β-phosphate group of ADP, it is shifted approximately 1.7 Å away
from its position in the covalently bound substrate (Fig. 3B). We
surmise that this rearrangement contributes to the observed re-
duced catalytic efficiency of the mutants (Table 1). Another con-
tribution to the lower activity of the H56NK and H56AK mutants
is their weaker affinity for the substrate, as reflected by a signifi-
Despite the lack of conformational changes in DhaL upon
binding to DhaK, the adenine ring of ADP undergoes a syn to anti
conversion (Fig. S3) when compared with the free DhaL structure
(4). Our molecular dynamics (MD) simulations indicate that in
the syn conformation, ADP has a potential steric clash of its ade-
nine ring with Thr107K (Fig. S4). Flipping of the adenine ring is
accompanied by formation of three additional water-mediated
H bonds, two between its purine N1 and the carbonyl group of
Val117L and the amide group of Ala123L and the third between
its purine N6 and the OH group of Thr107K. In addition, one of
the two Mg2þ ions (M2) in the ADP binding site of DhaL moves
by approximately 1.0 Å compared to free DhaL and becomes a
bridging element at the DhaK–DhaL interface (Fig. S3). Its coor-
dination number increases from three to five with the addition of
the carbonyl of Phe78K and a water molecule.
His56K and Asp109K Are Important for Dha Kinase Activity. In order to
evaluate the contributions of residues in the active site for activ-
ity, we performed site-directed mutagenesis and activity measure-
ments for several mutant enzymes, focusing on the conserved
residues His56K, Asp109K, His218K, and Arg178L. Gel filtration
chromatography revealed that all of the DhaK mutant enzymes
(H56AK, H56NK, D109AK, D109NK, and H218KK) were able to
form complexes with DhaL, while no complex formation between
DhaK and R178EL was observed (see SI Text). Two variants of
cant increase in the Kmapp (Table 1). Although our upper estimate
His56K, H56AK and H56NK, show a moderate decrease in kcat
app
app
app
of Km
for Dha in wild-type DhaK is 8 μM, the corresponding
but at least a 40- to 300-fold increase in Km
(Table 1). Both
D109AK and D109NK mutant enzymes showed no measurable ac-
tivity (Fig. S5). Mutation of His218K to either lysine (Fig. S5) (2)
Kmapp is 302 μM and 2.2 mM for H56AK and H56NK, respectively
(Table 1). Despite the different positioning of Dha in the H56NK
1304
∣
www.pnas.org/cgi/doi/10.1073/pnas.1012596108
Shi et al.

Fig. 4. Proposed catalytic mechanism for the DhaK–DhaL kinase, consisting
of (i) initial formation of the hemiaminal linkage between His218K and Dha;
(ii) phosphoryl transfer from ATP to Dha, yielding Dha-P; and (iii) breaking of
the hemiaminal linkage and release of Dha-P.
ing residues in the active site have not been previously implicated
as having a role in catalysis.
Fig. 3. Crystal structure of the DhaK H56NK mutant. (A) Active site of the
DhaK H56NK mutant bound noncovalently to Dha. The DhaK molecule is
shown in cartoon representation. Asn56K and the residues involved in sub-
strate binding are shown in stick mode. Hydrogen bonding interactions in-
volving Dha are indicated by magenta dashed lines. Red spheres represent
two water molecules. The difference (Fo − Fc) electron density map for
Dha (omitted from refinement) contoured at 2.5σ is shown in green. (B) Sub-
strate binding pocket of the DhaK H56NK–Dha complex superimposed onto
that of the DhaK–DhaL complex. The carbon atoms of the labeled residues
in the DhaK H56NK mutant and in the DhaK–DhaL complex are colored in
yellow and green, respectively. The phosphate moieties of ADP as well as
the two magnesium ions in the DhaK–DhaL complex are also shown. Hydro-
gen bonds involving the noncovalently and covalently bound Dha molecules
are represented by magenta and black dash lines, respectively.
Consistent with our structural data, we propose that formation
of the initial covalent complex with Dha (Fig. 4i) involves nucleo-
philic attack of His218K on the electrophilic carbonyl Cβ atom of
Dha. For this, the NE2 atom of His218K needs to be deproto-
nated. The proximal protonated His56K acts subsequently as a
general acid, donating its proton to form the Dha β-OH moiety
(Fig. 4i). Although in general a water molecule could also play the
role of general acid, there is no suitably located water molecule in
any of the available crystal structures of wild-type DhaK. Forma-
tion of the covalent enzyme–substrate complex allows for precise
orientation of the substrate with respect to other active site
residues, including Gly53K, His56K, Lys104K, and Asp109K.
Following formation of the covalent DhaK–Dha complex, the
next step is transfer of the γ-phosphoryl group of ATP to the
γ-OH of Dha (Fig. 4ii). This reaction requires activation of the
γ-OH through removal of a proton, in order to generate an oxy-
anion suitable for attack at the ATP γ-phosphorous atom. This
activation requires a residue acting as a general base, which we
identify as Asp109K, within H-bonding distance of Dha γ-OH.
This role is supported by the loss of activity observed for the
Asp109K mutant enzymes (Fig. S5). In the DhaK–DhaL complex,
the leaving oxygen (O2B of ADP) and the entering oxygen (Oγ of
Dha) are separated by 4.8 Å (Fig. 1B). Based on molecular
dynamics simulations on the DhaK–DhaL–ATP model, the
Pγ-Oγ(Dha) distance in the Michaelis complex is estimated to
be 3.9 Å (Fig. 2B). The locations of negative charges developing
during phosphoryl transfer may be inferred from the location of
stabilizing, positively charged groups provided by the enzyme. An
invariant residue, Arg178L, is located at the active site of the
DhaK–DhaL complex and has been suggested to be essential
active site, the γ-OH remains within H-bonding distance (∼2.9 Å)
of Asp109KðODÞ therefore retaining the ability of Asp109K to act
as a catalytic base by deprotonating Dha.
Mechanistic Insight into Phosphotransfer. The chemical mechanisms
of phosphoryl transfer in general, and with respect to kinases in
particular, have been well studied (reviewed in ref. 7). However,
little has been proposed regarding how DhaK forms the covalent
adduct, catalyzes phosphoryl transfer, and ultimately releases its
product. We divide the overall reaction into three stages: (i) bind-
ing of Dha to DhaK and formation of the covalent hemiaminal
linkage; (ii) phosphoryl transfer from ATP to the covalently
bound Dha, generating Dha-P and ADP; and (iii) release of
Dha-P from DhaK, setting the stage for a new catalytic cycle.
We propose here a plausible chemical mechanism for the overall
reaction (Fig. 4). With the exception of the demonstrated roles
for His218K of DhaK (2) and Arg178L of DhaL (6), the remain-
Shi et al.
PNAS
∣
January 25, 2011
∣
vol. 108
∣
no. 4
∣
1305

for activity by mutagenesis (6). This residue has been postulated
to stabilize the γ-phosphoryl group during phosphoryl transfer. In
the crystal structure, the guanidinium group of Arg178L is within
H-bonding distance of the γ-OH group of Dha while at the same
time does not directly interact with the β-phosphate of ADP
(3.8 Å away). As previously stated, the second magnesium ion
(M2) in the complex is coordinated by five oxygens, one of which
is from a water molecule located between the β-phosphate of
ADP and the substrate γ-OH group and likely mimics one of the
oxygen atoms in the γ-phosphate group of ATP. Such an arrange-
ment indicates that both Arg178L and M2 are important for sta-
bilizing the negative charges accumulated on the γ-phosphoryl
group (γ-phosphate) during phosphoryl transfer. In the DhaK–
DhaL complex, no positively charged residue other than Arg178L
could be found at the interface.
(9, 10, 11); however, covalent substrate binding by dihydroxyace-
tone kinases is distinctively different. Previous studies have sug-
gested that the primary role of the hemiaminal bond formed
between DhaK and Dha or Dha-P is to provide the enzyme with
a way of specifically selecting for short chain ketoses and aldoses
vs. polyalcohols, because the enzyme displays remarkable selec-
tivity for Dha in comparison with the closely related molecule,
glycerol (8). This is in contrast to glycerol kinase, which is able
to effectively utilize both Dha and glycerol as substrates for
phosphorylation (12, 13). Quantum-mechanical calculations did
not provide any evidence for an influence of the hemiaminal bond
on activation of the Dha γ-O atom involved in nucleophilic
attack of the γ-phosphate of ATP (8). In our hands, addition of
glycerol (10 mM–1 M) to phosphotransferase reactions showed
no decrease in kinase activity for wild-type and both H56K mutant
enzymes, indicating that the specificity of DhaK for Dha is not a
consequence of hemiaminal bond formation per se. The apparent
inability of glycerol to compete with Dha for binding to the DhaK
active site may be a consequence of subtle structural differences
between the two molecules, due to the sp2 β-C in Dha vs. the sp3
β-C of glycerol, resulting in steric conflicts of the latter within the
quite rigid DhaK active site (based on comparison of all available
crystal structures).
Our structural and biochemical results show that covalent
bond formation does contribute, in part, to orientation of the sub-
strate for optimal phosphoryl transfer. An additional potential
role for hemiaminal bond formation to that in catalysis is for
it to serve as an on/off switch, thereby exerting a regulatory func-
tion. A covalent bond between Dha and DhaK would support the
enzyme existing in either a “Dha-loaded” or “Dha-unloaded”
form, as dictated by (i) the inherent affinity of the enzyme for
substrate and (ii) the effective concentration of Dha within the
cell. The hemiaminal bond, once formed, commits DhaK to the
“loaded” state, regardless of further changes in intracellular Dha
concentration. It also provides a form of DhaK already primed
for catalysis, once an ATP-loaded form of DhaL is available. In
this manner, Dha binding and phosphoryl transfer are effectively
uncoupled, allowing, for example, variations in the intracellular
PEP pool independent of the particular state of DhaK. In a
similar manner, the covalent bond would allow for an on/off
switch with regards to transcriptional activation or repression that
has been shown to occur via interactions with the transcription
regulator DhaR (14). In either case, one would predict that a
DhaK enzyme incapable of forming the hemiaminal bond in vivo
would exhibit an attenuated response as opposed to an on/off
behavior, given that the substrate could bind/unbind to the
enzyme active site. Importantly, His56K is as strongly sequence
conserved as is the catalytic Asp109K in Dha kinases across var-
ious species, including higher eukaryotes. This suggests a physio-
logical importance for covalent substrate binding for this class of
kinase in both PEP and ATP-dependent branches. The precise
nature of the physiological consequences on cellular Dha meta-
bolism resulting from loss of hemiaminal bond formation is the
subject of future studies.
The final step in catalysis involves dissociation of Dha-P from
DhaK, requiring breaking the hemiaminal linkage. Based on the
structure of the active site, His56K is well positioned to act as a
general base, abstracting the proton from the β-OH of Dha-P,
with concomitant formation of the carbonyl group at C2 and
breakage of the Dha-P-His218NE2 bond (Fig. 4iii). The only other
amino acid that could potentially act as a general base, Asp109K,
which interacts directly with the primary hydroxyl groups of Dha
via its γ-carboxyl oxygens, is located too far to interact with
the β-OH of Dha. Here, the oxygen–oxygen distance is 5.4 Å as
opposed to 2.6 Å between NE2 of His56K and the β-OH of Dha.
As suggested previously (8), the γ-carboxyl oxygen of Asp109K
loses the ability to form an H bond with the primary hydroxyl
group of Dha after the phosphoryl-ester is formed. Unfavorable
interactions between the phosphate group of Dha-P and the
carboxylate group of Asp109K may help release the product from
the enzyme. Together our data allow us to propose a chemical
mechanism for DhaK supported by several structural and bio-
chemical studies, consistent with involvement of His56K and
Asp109K in addition to the previously characterized His218K (2)
and Arg178L (4). These residues are highly sequence conserved
in both bacterial and eukaryotic Dha kinases suggesting their
common mechanism.
Our results indicating some activity for the H56AK and H56NK
mutants require explanation. The first possibility is that cova-
lently linked substrate is not essential for catalysis. Noncovalent
binding of Dha, as reflected by our crystal structure, leads to
app
reduced affinity as indicated by increased Km
as well as to a
suboptimal orientation of γ-OH, affecting kcatapp. If the covalent
linking is not essential for activity, why are H218KK and H218AK
mutants dead? We believe that because the lysine mutant intro-
duces a more bulkly side chain it does not allow for the binding
of Dha. Furthermore, the small Dha molecule could not be well
positioned in the absence of the histidine imidazole ring in
the H218AK mutant. The second possibility is that catalysis in
H56K mutants is achieved through the covalently bound Dha,
assuming that His56K determines the equilibrium between the
covalently and noncovalently bound states of Dha. The covalent
bond might be formed without participation of His56K; for exam-
ple, a water molecule could replace His56K as general acid in the
Materials and Methods
app
H56AK and H56NK mutants. The fact that Km
rather than
Protein Expression, Purification, and Characterization. Details for cloning,
mutagenesis, expression, and purification are available in SI Text. Briefly,
DhaK and DhaL N-terminal His8-fusion proteins were purified by nickel
affinity chromatography and the tags on DhaK removed by cleavage with
tobacco etch virus (TEV) protease. Enzyme I and DhaM GST-fusion proteins
were purified by glutathione affinity chromatography and cleaved from
the column by TEV. HPr was purified by solubilizing inclusion bodies in urea
and refolding the protein on nickel resin. For crystallization, the His8-tag was
cleaved from DhaL and the DhaK–DhaL complex was further purified on a
Superdex 200 size exclusion column equilibrated with 50 mM Tris pH 8,
10 mM NaCl, 1 mM magnesium acetate, 1 μM AMP-PNP, and 1 mM DTT.
Complex formation of kinase mutants was analyzed by size exclusion chro-
matography in the same manner except that 1 μM ADP was substituted for
AMP-PNP.
kcat
is more affected by the His56K mutation provides some
app
support for this idea. However, we could not identify appropriate
water molecule in any of the His56K mutant structures. More-
over, despite having analyzed 40 independent subunits of the
crystal structures of DhaK H56AK or H56NK mutants, we were
not capable of capturing the covalently bound substrate even at
40 mM Dha. Therefore, it is more likely that covalent binding of
Dha is not essential for activity.
Functional Implications of Covalent vs. Noncovalent Substrate Binding
by Dha Kinases. Formation of a transient, covalent bond between
substrate and enzyme usually directly contributes to catalysis
1306
∣
www.pnas.org/cgi/doi/10.1073/pnas.1012596108
Shi et al.

Crystallization. Crystals of the DhaK–DhaL complex were obtained using the
Anion Suite screen (Qiagen). Optimization of conditions led to the best crystals
from hanging drop vapor diffusion at 19 °C by equilibrating 2 μL of protein
(10 mg∕ml) with 1 μL reservoir solution (0.1 M HEPES pH 7.5, 3.5 M sodium
formate) over 1 mL of reservoir solution. Crystals belong to space group
P41212 with unit cell dimensions a ¼ b ¼ 74.6, c ¼ 268.8 Å. Crystals of wild-
type DhaK and mutant enzymes were also obtained by hanging drop vapor
diffusion by equilibrating 1 μL of protein (2.5 mg∕ml of DhaK and 1.7 mg∕ml
of DhaL-his) with 1 μL reservoir solution (0.1 M sodium citrate pH 5.6, 20%
[w∕v] PEG 8000) over 1 mL of reservoir solution. Crystals of the wild-type DhaK
belong to space group P21 with a ¼ 49.8, b ¼ 91.5, c ¼ 73.2 Å, β ¼ 89.9°.
Crystals of H56AK and H56NK belong to space group P1 with unit cell dimen-
sions a ¼ 59.7, b ¼ 82.6, c ¼ 92.9 Å, α ¼ 77.9, β ¼ 78.1, γ ¼ 71.1°, whereas that
of the DhaK-H56NK-Dha complex form in space group P21 with a ¼ 82.2,
b ¼ 101.1, c ¼ 99.3 Å, β ¼ 89.95°. For data collection, crystals were transferred
to reservoir solution supplemented with 12% (v∕v) ethylene glycol and flash
cooled in a nitrogen stream at 100 K (Oxford Cryosystems).
was further purified by size exclusion (Superdex 200) in buffer containing
20 mM Tris pH 8, 300 mM NaCl, 1 mM DTT, 1 mM MgCl2, and 0.01 mM
ADP. Phosphotransferase reactions for specific activity measurements con-
tained 1 μM Enzyme I, 1 μM HPr, 0.5 μM DhaM, 0.5 μM DhaL, 0.5 μM DhaK,
50 mM potassium phosphate pH 7.5, 16 mM Dha, 2.5 mM DTT, 2.5 mM MgCl2,
2 units glycerol-3-phosphate dehydrogenase (Sigma Chemical Co.), and 1 mM
NADH and were initiated by the addition of 2 mM PEP. The glycerol-3-phos-
phate dehydrogenase coupling enzyme uses NADH to convert Dha-P to gly-
cerol-3-phosphate. The production of Dha-P in the reaction was followed by
monitoring oxidation of NADH at 340 nm in a Spectramax 250 plate reader at
room temperature. The effect of glycerol on phosphotransferase activity for
wild-type, H56AK and H56NK was examined for reactions in the presence of
10 mM, 100 mM and 1 M glycerol. kcatapp was measured for wild-type enzyme
from coupled phosphotransferase reactions (0–1.6 mM Dha) containing
0.5 μM Enzyme I, 0.5 μM HPr, 0.25 μM DhaM, 0.25 μM DhaL, 0.25 μM DhaK,
50 mM potassium phosphate pH 7.5, 2.5 mM DTT, 2.5 mM MgCl2, 2 units
glycerol-3-phosphate dehydrogenase (Sigma Chemical Co.), and
1 mM
NADH, initiated by addition of 2 mM PEP. These conditions are similar to
those reported previously (8), and kinetic constants were derived by varying
concentration of Dha only, leading to apparent values of kinetic parameters.
X-ray Data Collection, Structure Solution, and Refinement. Diffraction data for
free DhaK and the DhaK–DhaL complex were collected to 2.2 Å at the 31-ID
beamline (LRL-CAT), Advanced Photon Source, Argonne National Laboratory.
Data integration and scaling were performed with the program HKL2000
(15). Structure determination was performed by molecular replacement
using the program Phaser (16) from the CCP4 suite with the previously
reported E. coli DhaK (PDB 1OI2) (2) and DhaL (PDB ID code 2BTD) (4) struc-
tures as the search models. Refinement was carried out with the program
Refmac5 (17) with final Rwork∕Rfree of 0.172∕0.210 and 0.189∕0.225 for
DhaK and DhaK–DhaL, respectively. Diffraction data for crystals of DhaK
mutant enzymes were collected to 2.55 Å (H56AK), 1.97 Å (H56NK) and
2.2 Å (H56NK-Dha) at the CMCF1 beamline at the Canadian Light Source.
Data processing and structure determination was done similarly as described
above. All models have good geometry as analyzed with PROCHECK (18).
Final data collection and refinement statistics are shown in Table S1.
Due to the limited sensitivity of the assay for wild-type enzyme seen at
app
low substrate concentrations an accurate value for Km
could not be deter-
mined. For H56AK and H56NK, kcat
and Km
were determined from
app
app
reactions containing 1 μM Enzyme I, 1 μM HPr, 0.5 μM DhaM, 0.5 μM DhaL,
and 0.5 μM DhaK at a range from 0–16 mM and 0–64 mM Dha, respectively.
Values were calculated by nonlinear regression fit to a hyperbola.
ACKNOWLEDGMENTS. We thank Dr. Shaunivan Labiuk (Canadian Light Source;
CLS) for collecting several datasets; John Wagner, Ming-Ni Hung, and Linhua
Zhang for cloning; and Drs. Traian Sulea, Enrico O. Purisima, and Stephan
Grosse for helpful discussions. This research was supported by Canadian
Institutes of Health Research (CIHR) Grant MOP-48370 (I.E., A.M., and M.C.).
X-ray diffraction data for this study were measured at Lilly Research Labora-
tory Collaborative Access Team (LRL-CAT) at the Advanced Photon Source,
Argonne National Laboratory and at CMCF1 at CLS. Use of the Advanced
Photon Source at Argonne National Laboratory was supported by the US
Department of Energy, Office of Science, Office of Basic Energy Sciences,
under Contract DE-AC02-06CH11357. Use of the LRL-CAT beamline at Sector
31 of the Advanced Photon Source was provided by Eli Lilly & Company,
which operates the facility. The Canadian Macromolecular Crystallography
Facility is supported by the Canadian Foundation for Innovation, the Natural
Sciences and Engineering Research Council of Canada, and CIHR. This is NRCC
publication 50686.
Molecular Dynamics Simulations. The DhaK–DhaL–ATP, DhaK–DhaL–syn–ADP,
and DhaK–DhaL–anti–ADP complexes were sampled by 5-ns MD simulations
using the AMBER10 suite of programs (19) together with the AMBER ff03
force field for the proteins and a modified force field for ATP and ADP
(20). Details of the calculations are available in SI Text.
Phosphotransferase Assay. Phosphorylation of Dha was measured in
a
coupled assay similar to that described previously (1). To remove residual
endogenous DhaK that could copurify with recombinant DhaL, the latter
1. Gutknecht R, Beutler R, Garcia-Alles LF, Baumann U, Erni B (2001) The dihydroxyace-
tone kinase of Escherichia coli utilizes a phosphoprotein instead of ATP as phosphoryl
donor. EMBO J 20:2480–2486.
2. Siebold C, Garcia-Alles LF, Erni B, Baumann U (2003) A mechanism of covalent substrate
binding in the x-ray structure of subunit K of the Escherichia coli dihydroxyacetone
kinase. Proc Natl Acad Sci USA 100:8188–8192.
3. Bächler C, Flükiger-Brühwiler K, Schneider P, Bähler P, Erni B (2005) From ATP as
substrate to ADP as coenzyme: Functional evolution of the nucleotide binding subunit
of dihydroxyacetone kinases. J Biol Chem 280:18321–18325.
4. Oberholzer AE, Schneider P, Baumann U, Erni B (2006) Crystal structure of the nucleo-
tide-binding subunit DhaL of the Escherichia coli dihydroxyacetone kinase. J Mol Biol
359:539–545.
10. Verschueren KH, Seljée F, Rozeboom HJ, Kalk KH, Dijkstra BW (1993) Crystallographic
analysis of the catalytic mechanism of haloalkane dehalogenase. Nature 363:693–698.
11. Heine A, et al. (2001) Observation of covalent intermediates in an enzyme mechanism
at atomic resolution. Science 294:369–374.
12. Jin RZ (1982) Glycerol kinase as a substitute for dihydroxyacetone kinase in a mutant of
Klebsiella pneumoniae. J Bacteriol 152:1303–1307.
13. Johnson EA, Burke SK, Forage RG, Lin EC (1984) Purification and properties of
dihydroxyacetone kinase from Klebsiella pneumoniae. J Bacteriol 160:55–60.
14. Bächler C, Schneider P, Bähler P, Lustig A, Erni B (2005) Escherichia coli dihydroxyace-
tone kinase controls gene expression by binding to transcription factor DhaR. EMBO J
24:283–293.
5. Siebold C, Arnold I, Garcia-Alles LF, Baumann U, Erni B (2003) Crystal structure of the
Citrobacter freundii dihydroxyacetone kinase reveals an eight-stranded alpha-helical
barrel ATP-binding domain. J Biol Chem 278:48236–48244.
15. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in
oscillation mode. Methods Enzymol 276:307–326.
16. McCoy AJ (2007) Solving structures of protein complexes by molecular replacement
with Phaser. Acta Crystallogr D 63:32–41.
6. Zurbriggen A, et al. (2008) X-ray structures of the three Lactococcus lactis dihydrox-
yacetone kinase subunits and of
a transient intersubunit complex. J Biol Chem
17. Murshudov GN, Vagin AA, Lebedev A, Wilson KS, Dodson EJ (1999) Efficient anisotro-
pic refinement of macromolecular structures using FFT. Acta Crystallogr D 55:247–255.
18. Laskowski RA, MacArthur MW, Moss DS, Thornton JM (1993) PROCHECK: A program to
check the stereochemical quality of protein structures. J Appl Crystallogr 26:283–291.
19. Case DA, et al. (2005) The Amber biomolecular simulation programs. J Comput Chem
26:1668–1688.
283:35789–35796.
7. Allen KN, Dunaway-Mariano D (2004) Phosphoryl group transfer: Evolution of a
catalytic scaffold. Trends Biochem Sci 29:495–503.
8. Garcia-Alles LF, et al. (2004) Phosphoenolpyruvate- and ATP-dependent dihydroxyace-
tone kinases: Covalent substrate-binding and kinetic mechanism. Biochemistry
43:13037–13045.
20. Meagher KL, Redman LT, Carlson HA (2003) Development of polyphosphate
parameters for use with the AMBER force field. J Comput Chem 24:1016–1025.
9. Lahiri SD, Zhang G, Dunaway-Mariano D, Allen KN (2003) The pentacovalent
phosphoryl transfer reaction. Science 299:2067–2071.
Shi et al.
PNAS
∣
January 25, 2011
∣
vol. 108
∣
no. 4
∣
1307