Biochem. J. (2012) 446, 405–413 (Printed in Great Britain)
doi:10.1042/BJ20120596
405
An aminoquinazoline inhibitor of the essential bacterial cell wall synthetic
enzyme GlmU has a unique non-protein-kinase-like binding mode
Nicholas A. LARSEN*, Tory J. NASH†, Marshall MORNINGSTAR†, Adam B. SHAPIRO†, Camil JOUBRAN†, Carolyn J. BLACKETT‡,
Arthur D. PATTEN†, P. Ann BORIACK-SJODIN* and Peter DOIG*1
*Discovery Sciences Unit, AstraZeneca R&D Boston, Waltham, MA 02451, U.S.A., †Infection Innovative Medicines Unit, AstraZeneca R&D Boston, Waltham, MA 02451, U.S.A., and
‡Discovery Sciences Unit, AstraZeneca UK, Alderley Park, Macclesfield, Chesire SK10 4TG, U.K.
GlmU is a bifunctional enzyme with acetyltransferase and
uridyltransferase activities, and is essential for the biosynthesis
of the bacterial cell wall. Inhibition results in a loss of cell
viability. GlmU is therefore considered a potential target for
novel antibacterial agents. A HTS (high-throughput screen)
identified a series of aminoquinazolines with submicromolar
potency against the uridyltransferase reaction. Biochemical and
biophysical characterization showed competition with UTP
binding. We determined the crystal structure of a representative
aminoquinazoline bound to the Haemophilus influenzae
isoenzyme at a resolution of 2.0 Å. The inhibitor occupies part
of the UTP site, skirts the outer perimeter of the GlcNAc1-P (N-acetylglucosamine-1-phosphate) pocket and anchors a
hydrophobic moiety into a lipophilic pocket. Our SAR (structure–
activity relationship) analysis shows that all of these interactions
are essential for inhibitory activity in this series. The crystal
structure suggests that the compound would block binding of
UTP and lock GlmU in an apo-enzyme-like conformation, thus
interfering with its enzymatic activity. Our lead generation effort
provides ample scope for further optimization of these compounds
for antibacterial drug discovery.
INTRODUCTION
PPi (pyrophosphate ion). This reaction is contained entirely in the
N-terminal domain of each GlmU monomer [5] and is the ratelimiting step of the overall catalytic reaction. Structural studies
have shown that UTP and GlcNAc-1-P bind to separate regions
of the binding site [7–10]. The kinetic reaction mechanism is
ordered, with UTP binding first followed by GlcNAc-1-P, which
triggers a conformational change that is required for product
formation [10,11].
GlmU is an attractive target for antibacterial drug discovery
due to its importance in bacterial cell wall biosynthesis and the
absence of any comparable enzyme in humans [12]. Previous
publications have shown success in targeting both domains of
GlmU with small molecule inhibitors. An allosteric inhibitor of
the H. influenzae isoenzyme was identified through a HTS (highthroughput screen) targeting the uridyltransferase domain [13].
There are two structures in the PDB (codes 2W0V and 2W0W)
showing quinazoline compounds in the UTP-binding pocket,
which are presumably UTP competitive. Inhibitors have also
been identified for the acetyltransferase reaction in the C-terminal
domain that were either competitive or showed mixed competition
with acetyl-CoA [14]. More recently, sulfonamide containing
inhibitors against the C-terminal domain of H. influenzae GlmU
have been reported [15,16], and several biophysical methods
were used to confirm binding to the acetyl-CoA-binding site. The
sulfonamides with antimicrobial activity against an acrB efflux
pump mutant of H. influenzae were shown to act via GlmU using
strains that overexpress that target. Additionally, resistant mutants
were isolated that mapped to the acetyltransferase domain of
GlmU [15]. Together, these data provided validation that chemical
inhibition of the GlmU enzyme can halt bacterial growth in vitro.
Inhibition of cell wall biosynthesis has been a successful approach
for developing antibiotics. Both β-lactam-containing compounds
and glycopeptides exploit this mode of action and are well
established in the clinic. However, the emergence of mechanismbased resistance is an ongoing challenge [1]. Resistance can be
overcome in some cases by additional optimization of existing
scaffolds [2] and in other cases by combinations with other drugs
[3]. Inhibition of novel targets is another way of circumventing
existing clinical resistance, and some examples are now entering
clinical testing [1].
GlmU is a bifunctional enzyme that converts GlcN-1-P
(D-glucosamine-1-phosphate) into UDP-GlcNAc (uridyldiphosphate-N-acetylglucosamine), an essential precursor in
lipopolysaccharide and peptidoglycan synthesis in bacteria.
Inhibition of GlmU results in a decrease in UDP-GlcNAc
available for cell wall biosynthesis, resulting in a loss of
cell viability [4,5]. GlmU performs two distinct reactions [6]
(Figure 1). The first is the transfer of an acetyl group from
acetyl-CoA to glucosamine-1-phosphate producing GlcNAc1-P (N-acetylglucosamine-1-phosphate) and CoA [5]. This
reaction occurs in the C-terminal portion of the polypeptide.
Structural studies of Haemophilus influenzae, Escherichia coli
and Streptococcus pneumoniae isoenzymes bound to acetyl-CoA
have shown that the C-terminal domain assembles as a trimer
and the catalytic sites are composed of portions of all three of
the monomers at the interface of two subunits and the extended
C-terminal tail of the third [7,8]. The second reaction transfers
uridyl monophosphate from UTP to form UDP-GlcNAc and a
Key words: aminoquinazoline, cell wall biosynthesis,
GlmU,
high-throughput
screening,
inhibitor,
kinase,
N-acetylglucosamine-1-phosphate uridyltransferase.
Abbreviations used: DTT, dithiothreitol; GlcNAc-1-P, N -acetylglucosamine-1-phosphate; GlcN-1-P, D -glucosamine-1-phosphate; HTS, high-throughput
screen; ITC, isothermal titration calorimetry; MIC, minimum inhibitory concentration; PPi , pyrophosphate ion; RMSD, root mean square deviation; SAR,
structure–activity relationship; UDP-GlcNAc, uridyldiphosphate-N -acetylglucosamine.
The atomic co-ordinates and structures factors from the present study have been deposited in the PDB under accession code 4E1K.
1
To whom correspondence should be addressed (email [email protected]).
c The Authors Journal compilation c 2012 Biochemical Society
406
Figure 1
N. A. Larsen and others
The biochemical reactions catalysed by GlmU
GlmU is a bifunctional enzyme performing two distinct reactions. The first step transfers an acetyl group from acetyl-CoA to GlcN-1-P, producing GlcNAc-1-P and CoA. The second reaction transfers
uridine 5 -monophosphate from UTP to form UDP-GlcNAc and PPi .
The compounds used in the present study were either described or
synthesized according to methods described previously [19–21].
For the high-throughput screening, the assay was adapted as
follows to run on the automation employed. The reaction was
run in a clear 384-well plate (Greiner) with a final reaction
volume of 30 μl. A 100 % inhibition in control wells was achieved
with 20 mM EDTA. Reactions were allowed to proceed at room
temperature for 40 min prior to being quenched with 30 μl of
Malachite Green reagent.
For HPLC detection, the reaction was quenched by 1:1 addition
of 50 mM NaH2 PO4 /Na2 HPO4 , with a pH of 2.8 adjusted with
acetic acid. A sample (10 μl) was injected on to a Vydac
3021IC4.6 ion-exchange column (4.6 mm×100 mm), running
at a flow rate of 2 ml/min at 22 ◦ C, equilibrated with 25 mM
NaH2 PO4 /Na2 HPO4 , with pH adjusted to 2.8 with acetic acid
(Buffer A). Elution was with a gradient of Buffer A and Buffer B
(125 mM NaH2 PO4 /Na2 HPO4 , with pH adjusted to 2.9 with acetic
acid) as follows: 2 min at 100 % Buffer A, followed by a linear
gradient from 0 % to 100 % Buffer B in 8 min and then held at
100 % Buffer B for 2 min. Quantitative UTP depletion and UDPGlcNAc production were measured by absorbance at 260 nm.
Nephelometry was used to estimate the solubility of compounds
in the assay buffer [50 mM Hepes-NaOH (pH 7.5) and 2 % (v/v)
DMSO]. Measurements were made using a Nephelostar (BMG
Labtech), and solubility was determined as the point of signal
change due to light scattering.
Susceptibility testing
NMR
MICs (minimum inhibitory concentrations) were determined
using broth microdilution according to the guidelines of the
Clinical and Laboratory Standards Institute [22].
E. coli GlmU (5.9 mg/ml) was equilibrated with NMR buffer
consisting of 50 mM deuterated Tris/HCl (Cambridge Isotope
Laboratories) adjusted to pH 7.6 with HCl, 5 % (v/v), 2 H2 O
and 5 mM deuterated DTT (Cambridge Isotope Laboratories)
using Micro Biospin-6 centrifugal gel-filtration columns (BioRad Laboratories). The concentration of the equilibrated sample
was measured using the method of Bradford [24]. Aliquots of the
protein were flash frozen on dry ice and stored at − 80 ◦ C.
The 1 H spectra were acquired at room temperature utilizing
a Bruker DRX-500MHz spectrometer with a 5 mm room
temperature broadband indirect probe. All of the samples were
dissolved in 10 % 2 H2 O and 90 % H2 O. Water suppression
was accomplished with the use of an excitation sculpting pulse
sequence [25,25].
The present study describes the identification of a series of
aminoquinazolines that are UTP competitive. Quinazolines are
known hinge binders in kinases [17] and we anticipated analogous
interactions in the UTP pocket. However, the crystal structure
revealed an unexpected binding mode, with ample scope for future
structure-based drug design.
EXPERIMENTAL
GlmU expression and purification
Genomic DNA and GlmU constructs were prepared
from H. influenzae and E. coli as described previously [15].
The H. influenzae glmU was subcloned into pET30a. The E. coli
glmU was cloned into pZT7#3.3 [18]. Both expression constructs
consist of full-length native sequence, with no additional affinity
tags. Both H. influenzae and E. coli GlmU enzymes were
expressed in E. coli BL21(DE3) cells and purified as described
previously [15].
Compounds
Assay conditions
The standard conditions for compound inhibition assays
were as follows: 50 mM Hepes/NaOH (pH 7.5), 5 mM
DTT (dithiothreitol), 0.01 % Brij-35, 0.3 units/ml inorganic
pyrophosphatase (Roche Applied Science), 5 mM MgCl2 and
30 μM UTP (Sigma). The H. influenzae isoenzyme assay
contained 22 μM GlcNac-1-P (Sigma) and 260 pM GlmU trimer.
The E. coli isoenzyme assay contained 32 μM GlcNac-1-P
(Sigma) and 150 pM GlmU trimer. Substrate concentrations were
set at approximately 2-fold of the K m value. Reactions were run
in a volume of 100 μl in a clear 96-well plate (Costar). A
100 % inhibition in control wells was achieved with 10 mM
EDTA. Reactions were allowed to proceed at room temperature
(22 ◦ C) for 30 min prior to being quenched. To quench the
reaction and detect phosphate release, Malachite Green-based
detection (150 μl) was used [23]. Reactions were incubated at
room temperature for 5 min and then the absorbance was read
at 650 nm.
c The Authors Journal compilation c 2012 Biochemical Society
ITC (isothermal titration calorimetry)
ITC experiments were conducted at 23 ◦ C using a Microcal
VP ITC. H. influenzae GlmU was formulated at 20 μM in 25 mM
Tris (pH 7.5), 50 mM NaCl and 2 % DMSO. Approximately
1.4 ml of protein was loaded into the sample cell. Compound
1 was solubilized at 10 mM in DMSO and diluted into the
same buffer to 200 μM (final DMSO 2 %). A total of 24
Competitive inhibitors of GlmU uridyltransferase reaction
injections of ligand were sequentially added to the sample cell,
titrating the protein and producing a final compound-to-protein
molar ratio of approximately 2.2:1. Thermodynamic parameters
N (stoichiometry), K a (association constant), H◦ (enthalpic
change) and TS (entropy change) were obtained using Origin
ITC software (v. 6.0, Microcal Software).
Crystallography
Purified H. influenzae GlmU was crystallized as described
previously [13] with minor modification. The protein was
buffer exchanged into 10 mM Hepes (pH 7.5), 500 mM
NaCl and 0.2 mM TCEP [tris-(2-carboxyethyl)phosphine], and
concentrated to 16 mg/ml for crystallization. The compound was
solubilized in DMSO to a concentration of 100 mM and diluted
into protein to a final concentration of 2.5 mM giving ∼ 8:1
molar excess of ligand to protein. Crystals were grown by vapour
diffusion at room temperature. Optimal conditions were obtained
in 4 μl hanging drops containing equal parts of protein and
reservoir solution of 1.6–1.8 M ammonium sulfate and 0.1 M Mes
(pH 6.1). Crystals appeared overnight, grew over several days and
were frozen in cryosolution containing 8 μl of reservoir solution,
2 μl of ethylene glycol and 0.5 μl of the compound solubilized in
100 % DMSO at a concentration of 100 mM.
Diffraction data were obtained in-house from a Rigaku FRE + generator outfitted with a Saturn 944 + CCD (chargecoupled device) detector. Data were integrated and scaled using
autoPROC software [26]. The structure was determined by
molecular replacement using AMoRe [27] and refined using
Refmac [27] and Buster (http://www.globalphasing.com/buster/)
with rebuilding in Coot [29]. Water molecules were added
to the structure using Buster and removed after visual inspection
of the electron density. Final refinement statistics converged with
Rwork = 0.181 and Rfree = 0.206.
RESULTS
Hit evaluation summary
An HTS was performed against the H. influenzae GlmU
uridyltransferase reaction using an inorganic pyrophosphatasecoupled Malachite green assay to monitor the release of the byproduct phosphate. The screen identified an aminoquinazoline
inhibitor, and the series was further defined using closely
analogous compounds (Table 1). Compounds had IC50 values
against H. influenzae and E. coli GlmU ranging from 160 nM
to >200 μM. To exclude false positives, IC50 values against H.
influenzae GlmU uridyltransferase were verified with an HPLC
assay. Additionally, compounds were determined to be soluble
and free of insoluble, large aggregates at concentrations above
their IC50 value by nephelometry. IC50 values determined over
a range of H. influenzae enzyme concentrations showed no
significant shifts in potency, indicating specificity of binding.
Together, these results indicate that these aminoquinazolines are
specific inhibitors of GlmU and are not inhibiting the enzyme due
to aggregation or non-specific binding to the enzyme. None of the
compounds reported in the present study have significant MICs
against H. influenzae or E. coli, including the arcB − and tolC −
mutants of each species respectively.
SAR (structure–activity relationship) of the aminoquinazoline
series
A number of representative aminoquinazoline compounds
enabled investigation of SAR. One key observation was an
407
increased potency against both isoenzymes with the 7-hydroxy
over a 7-methoxy substituent with the 1,4 diaminophenyl
core (compounds 1 and 2). The 2,5-diaminopyrimidine core
(compounds 4 and 5) had greater potency than the 1,4diaminophenyl core (compound 1) against the E. coli isoenzyme,
but similar potency against that of H. influenzae. The 2,5diaminopyrimidine core had better solubility as determined
by nephelometry. Several substitution patterns appeared to be
tolerated in the anilino-linked side chain of the aminoquinazoline
core, since halogen substitutions on the benzyl ring (compounds 4
and 5) are active and substitution to pyridine (compound 3) caused
no loss of potency in matched pair analysis (compounds 2 and
3). Diamino fragments lacking the quinazoline core (compounds
6 and 7), a compound replacing the bicyclic quinazoline ring
system with a single ring (compound 8) and a compound lacking
the aromatic right-hand-side moiety (compound 9) were inactive.
These results indicate that several regions of the molecule are
important for binding.
Characterization of binding of aminoquinazoline series to GlmU
Several experiments were performed to characterize the
mechanism of inhibition and the nature of the binding mode. The
binding of compound 1 to H. influenzae GlmU was confirmed
in solution by ITC. The K d value was determined to be 0.5 μM
with H of − 12.9 +
− 0.4kcal/mol, TS of − 4.4 kcal/mol and
stoichiometry (N) of 1.17 +
− 0.02 (Figures 2A and 2B). The
measured K d value is consistent with the IC50 value of 0.27 μM.
The enthalpic contribution is generally indicative of hydrogen
bond, electrostatic and van der Waals interactions, whereas the
entropic contribution generally reflects hydrophobic interactions
[30]. In this case, the binding is mainly enthalpically driven with
an unfavourable entropic contribution.
In order to determine with which substrate(s) the
aminoquinazoline series competes, one-dimensional 1 H NMR
binding competition experiments were performed between
compound 4 and GlcNAc-1-P and UTP [31,32]. In the first
experiment (Figure 2C), 10 μM (monomer concentration) E. coli
GlmU was added to 50 μM GlcNAc-1-P. In the one-dimensional
1
H NMR spectrum, a reduction in intensity and broadening of the
acetyl methyl proton resonance at δ1.98 ppm was observed due
to binding with the protein, as expected. We then added 11 μM
of compound 4 to the NMR tube and observed no increase in
intensity/sharpening of the acetyl resonance, indicating that the
compound did not displace, and therefore does not compete with,
the substrate. In the second experiment (Figure 2D), we added
5 μM GlmU to 50 μM UTP and observed broadening of the
pyrimidine resonance at δ7.85 ppm, demonstrating binding of
UTP to GlmU. We then added 25 μM compound 4 and observed
sharpening of the UTP resonances, indicating that the compound
displaced UTP from the binding site. This result shows that
compound 4 is competitive with UTP.
Additional kinetic characterization of the series with E. coli
GlmU was undertaken to further confirm the nature of compound
binding with respect to the binding of the substrates UTP
and GlcNAc-1-P. Kinetic experiments were done at several
concentrations in triplicate using compound 4, and regression
analysis and global fitting of the data were performed with
GraFit 5.0 (http://www.erithacus.com/grafit/). Data were fitted
to models of inhibition using Lineweaver–Burk transformations.
When the concentration of GlcNAc-1-P was varied, compound
4 showed parallel lines in the Lineweaver–Burk plot indicative
of uncompetitive inhibition (Figure 2E). In the Lineweaver–
Burk plot in which UTP was varied, compound 4 showed lines
intersecting on the y-axis indicative of competitive inhibition
c The Authors Journal compilation c 2012 Biochemical Society
408
Table 1
N. A. Larsen and others
Structures and biochemical properties of aminoquinazoline inhibitors of GlmU uridyltransferase
NT, not tested.
H. influenzae IC50 (μM)
E. coli IC50 (μM)
Nephelometry (μM)
1
0.26 +
− 0.04 (n = 2)
1.30 +
− 0.3 (n = 2)
100
2
2.2 +
− 0.9 (n = 2)
11 (n = 1)
50
3
1.8 +
− 0.9 (n = 3)
10.1 (n = 1)
>200
4
0.16 +
− 0.02 (n = 6)
0.21 (n = 1)
>200
5
0.46 +
− 0.1 (n = 2)
0.46 +
− 0.07 (n = 2)
200
6
>200 (n = 1)
NT
200
7
>200 (n = 1)
NT
200
8
>12.5 (n = 1)
NT
12.5
9
>200 (n = 1)
>200 (n = 1)
>200
10
9.8 (n = 1)
30 (n = 1)
>200
11
NT
NT
NT
Compound
Structure
c The Authors Journal compilation c 2012 Biochemical Society
Competitive inhibitors of GlmU uridyltransferase reaction
Table 1
Compound
409
Continued
Structure
12
H. influenzae IC50 (μM)
E. coli IC50 (μM)
Nephelometry (μM)
NT
NT
NT
(Figure 2F). The NMR and kinetic results are consistent with
the conclusion that these aminoquinazolines series inhibit the
uridyl transferase activity of GlmU by competition with the UTP
substrate.
Structure determination of compound 1 bound to GlmU
To define the molecular interactions, crystallographic studies
of compound 1 with H. influenzae GlmU were undertaken.
The structure of the inhibitor bound to GlmU was determined
by molecular replacement from in-house diffraction data at 2.0 Å
(1 Å = 0.1 nm) using a previously reported structure (PDB code
2VD4) as a search model. Clearly defined electron density was
visible in the initial difference maps for the expected inhibitor.
The inhibitor was readily modelled into the density and refined
(Figure 3). An additional strong difference density peak in the
active site was modelled as a sulfate ion. The data collection and
refinement statistics are found in Table 2.
GlmU has two distinct domains, each of which is
responsible for a separate biochemical reaction (Figure 4A). The
acetyltransferase reaction occurs in the C-terminal domain. This
C-terminal domain assembles as a trimer to form a pedestal-like
structure composed of an extended β-helix [34]. Catalysis occurs
at the interface of the subunits. The uridyltransferase reaction
occurs in the N-terminal domain (Figure 4A).
The uridyltransferase domain undergoes several changes in
conformation from the apo structure (PDB code 2V0H) to the
product-bound structure (PDB code 2V0I) [11,13] (Figure 4B).
The apo- and UTP-bound structures have an open activesite conformation, whereas in the product-bound structure,
the active site has contracted and become more closed. This
closure involves a rigid body movement of the N-terminal lobe
(residues 131–170) (Figure 4B). It was previously postulated
that GlcNAc-1-P binding induces conformational changes at this
subsite in order to create the proper stereochemical environment
needed for attack of the α-phosphate group of UTP [11]. This
movement drives the reaction by bringing the two substrates
into close proximity for catalysis. These movements are further
illustrated by comparison of the compound-1-bound structure
with previously reported GlmU structures (apo-, UDP- and UDPGlcNAc-bound forms) (Table 3). Determination of the RMSDs
(root mean square deviations) for each comparison indicated
that the uridyltransferase domain of the compound 1 inhibitorbound structure resembles more closely the apo- and UDPbound forms than the product-bound GlcNac-UDP structure. In
all of the cases, the acetyltransferase domains superimpose very
closely. The conformational changes are even more dramatic in
the S. pneumoniae isoenzyme (results not shown), with similar
movements of the upper N-terminal lobe and constriction of the
active site, but also additional rearrangement of the loop 184–196
to close over the active site.
Compound 1 skirts the perimeter of the GlcNAc-1-P-binding
pocket, with the quinazoline core occupying the binding pocket
for the ribose sugar of UTP (Figure 5A), consistent with the
kinetic and NMR data. The quinazoline core packs on top of
Leu11 and Tyr103 , where the ribose from UTP would otherwise
bind. The 7-hydroxyl forms a key hydrogen bond interaction
with the backbone amide of Ala13 . This hydrogen bond is also
seen with the 2-oxygen in the uridine ring of UTP. This is the
only hydrogen bond interaction shared between substrate and
inhibitor (Figure 5A). A bifurcated hydrogen bond interaction
is observed between the aniline-nitrogen and the side chain of
Asp105 (Figure 5A). An additional hydrogen bond is formed
between the benzamide nitrogen and the backbone carbonyl of
Val223 (Figure 5A). The number of key hydrogen bond interactions
observed is consistent with the strong binding enthalpy measured
by ITC.
With compound 1 bound, the structure of H. influenzae GlmU
more closely resembles the open, apo, form of the enzyme than
the closed, product-bound, form, with some notable exceptions.
Loop 75–81 forms the backside of the UTP pocket. In the
structure with compound 1, Gln79 is located in the pocket
to form hydrogen bond bridging interactions between Gln76
and Gly81 (Supplementary Figure S1 at http://www.BiochemJ.
org/bj/446/bj4460405add.htm). As a result, part of a pocket that
would normally have been occupied by UTP has been filled by
protein residue side chains. When UTP or product is bound, Gln79
moves out of the way, allowing Gln76 to form both donor and
acceptor hydrogen bond interactions with UTP (Supplementary
Figure S1). Efforts to optimize compound 1 to engage Gln76 could
be a possible rationale for improving the inhibitory potency of the
series.
Structural rationale for the observed SAR
Some of the key observations can be explained with regard
to the left-hand side aminoquinazoline ring, central aromatic
linker and right-hand side benzamide. With the left-hand side
aminoquinazoline, a methyl substituent at the 7-hydroxy position
(compounds 1 and 2) would be expected to decrease potency of the
compound since it would weaken the hydrogen bond interaction
with Ala13 (Figure 5A). This is reflected in a 13-fold loss in
potency of inhibition of H. influenzae GlmU. Complete removal
of the quinazoline moiety (compounds 6, 7 and 8) resulted in a
loss of inhibition, indicating its importance as an anchor in the
active site.
No specific interactions between the protein and the central
aromatic linker were observed. This is consistent with the
similar inhibitory potencies of compounds containing either
the 1,4-diaminophenyl core (compounds 1–3) or the 2,5diaminopyrimidine core (compounds 4 and 5).
The right-hand side phenyl ring of the benzamide sits in a
lipophilic pocket and the crystal structure suggests that some
substitution there would be tolerated (Figure 5A). Substitution
of this phenyl group (compound 2) with pyridine (compound 3),
or addition of halogens (compounds 4 and 5), had little effect
on potency. Increasing the lipophilicity with Cl/F substitutions
(compound 4) slightly improved potency, with Cl being preferred
c The Authors Journal compilation c 2012 Biochemical Society
410
Figure 2
N. A. Larsen and others
Characterization of the binding and mode of inhibition of compound 1 compared with GlmU
(A) Binding of compound 1 to H. influenzae GlmU measured by ITC. Compound 1 was titrated into GlmU and the heats were measured. (B) Integrated enthalpies were fit to a one-site-binding model,
giving the enthalpy H , binding constant K a , stoichiometry (N ) and the entropy T S . (C) NMR spectra showing compound 4 is not competitive with GlcNAc-1P. The peak at δ1.98 ppm represents
the acetyl methyl protons of GlcNAc-1P. a, buffer; b, 50 μM GlcNAc-1P in buffer; c, addition of 10 μM GlmU; and d, addition of 11 μM compound 4. (D) NMR spectra showing compound 4 is
competitive with UTP. The doublet at δ7.85 ppm is the pyrimidine proton of UTP. A, 50 μM UTP; b, addition of 5 μM GlmU (monomer concentration); and c, addition of 25 μM compound 4. (E)
Lineweaver–Burk plots. Plotting the inverse of the rates at fixed concentrations of UTP as a function of GlcNAc-1P gives parallel lines indicating that compound 4 is uncompetitive with GlcNAc-1P.
(F) When plotted as a function of UTP with fixed concentrations of GlcNAc-1P, the lines intersect on the y -axis indicating that compound 4 is competitive with UTP.
at the meta (4) rather than para (5) position. Modification
of the benzamide linker from an amide to a more flexible
alkoxymethylene (compound 10) results in a 10-fold loss of
potency. This substitution would eliminate a hydrogen bond
interaction seen from the amide nitrogen and the backbone
carbonyl of Val223 (Figure 5A). Removal of the benzamide
(compound 9) eliminates both the lipophilic interaction and
the hydrogen bond to Val223 . Not surprisingly, compound 9 is
completely inactive.
DISCUSSION
The enzymes involved in biosynthesis of the bacterial cell
wall are important and attractive targets in drug discovery.
Inhibition of GlmU causes loss of cell viability [4,5], and
the lack of corresponding enzymes in humans makes this
c The Authors Journal compilation c 2012 Biochemical Society
target attractive from a safety perspective. An HTS to find
inhibitors of the enzymatic activity of the N-terminal domain
of GlmU identified a series of aminoquinazolines. Enzymatic
and biophysical characterization showed that the compounds are
competitive with UTP, but not GlcNAc. The enzyme–inhibitor
co-crystal structure showed that the compound occupies part of
the UTP-binding pocket, locking the catalytic site in an open
conformation and blocking UTP binding.
Upon binding of the second substrate (GlcNac-1-P), the enzyme
undergoes a conformational change to a closed form (Figure 4B).
This closure constricts the lipophilic pocket where the right-hand
side benzamide tail sits. By occupying this pocket, compound
1 locks GlmU in the open conformation, which is incompatible
with enzyme catalysis. Meanwhile, the left-hand quinazoline head
group prevents binding of UTP. It is expected that by maintaining
and optimizing the two sides of this lead series improved target
potency can be achieved.
Competitive inhibitors of GlmU uridyltransferase reaction
411
Table 2 Data collection and refinement statistics for GlmU bound to
compound 1
Values in parentheses
i || I i − I | /
shell. R merge = h kl
are for highest-resolution
h kl | | F obs | · R free is the cross-validation R
i I i · R work =
h kl || F obs | − || F calc | |/
factor computed for the test set of 5 % of unique reflections.
Parameter
Value
Data collection
Space group
Cell dimensions
a , b , c (Å)
α, β, γ (◦ )
Resolution (Å)
R merge
Completeness (%)
I /σ I
Multiplicity
Refinement
R work /R free
Number of atoms
Protein
Ligand
Solvent/ions
Average B values (Å2 )
Protein
Ligand
Solvent/ions
RMSDs
Bond lengths
Bond angles
Ramachandran (%)
Most favoured
Allowed
Generously allowed
Disallowed
Figure 3
R32
109, 109, 324
90.0, 90.0, 120
61–2.0 (2.07–2.0)
0.046 (0.23)
95.6 (79.0)
20.2 (3.6)
4.7 (2.2)
0.181/0.206
3476
29
424
31.2
24.8
45.2
0.010
1.06
92.9
6.9
0.0
0.3
Electron density on the quinazoline inhibitor
Data collected in-house yielded data that were processed to 2.0 Å resolution. The quinazoline
compound is highlighted in magenta. An additional strong difference density peak in the active
site was modelled as a sulfate ion (arrow). Protein side chains are shown in grey and waters as
red spheres. The 2F o − F c electron density map is displayed as a blue mesh and contoured at
1 s. The overall clarity of the map allowed for the unambiguous placement of the inhibitor in the
active site.
Figure 4
X-ray crystal structure of the GlmU monomer from H. influenzae
(A) GlmU is a bifunctional enzyme with the acetyltransferase reaction occurring in the C-terminal
domain and the uridyltransferase reaction occurring in the N-terminal domain. The bound
inhibitor, compound 1, is highlighted in magenta. (B) Grey shows the compound 1 bound form
of the enzyme, and green shows the product-bound form (PDB code 2V0I). The product is shown
with green carbon atoms. During catalysis, the active site constricts from an open to a closed
conformation. This involves movement of several loops, as indicated by the arrows. Compound
1 binds to the open form that is almost identical to the apo-enzyme.
Quinazoline-containing compounds often exhibit inhibitory
activity against protein kinases by competition with ATP. The
quinazoline moiety binds where the adenine base of ATP engages
the enzyme. The 1-nitrogen of the quinazoline is an hydrogen
bond acceptor from the backbone nitrogen of the kinase hinge
motif (Figure 5B) [17]. This hinge interaction mimics that of
the nucleotide. Given our kinetic and NMR data establishing the
aminoquinazoline series as UTP competitive, it was anticipated
that the quinazoline core would bind in the base-binding region of
the UTP-binding site and mimic interactions seen by the uridine
ring. Surprisingly, our crystal structure revealed that none of these
interactions were made. Instead, the quinazoline ring binds in the
ribose binding site of GlmU. The only interaction made
in the UTP pocket with this series is the hydrogen bond between
the 7-position hydroxyl and Ala13 , also formed by the 2-oxygen
in the uridine ring of UTP (Figure 5A).
The presence of the quinazoline in this scaffold may raise
selectivity concerns due to the potential for inhibition of kinases
present in humans. However, the binding mode offers scope to
mitigate this concern by removal of the 1-nitrogen from the
quinazoline core. This would likely eliminate kinase binding with
little impact on GlmU binding.
Three inhibitor-bound structures with GlmU have been reported
(PDB codes 2W0V and 2W0W, and [13]). In two of these cocrystal structures, the ligand is a quinazoline (compound 11).
Interestingly, the central linker and lipophilic right-hand side of
c The Authors Journal compilation c 2012 Biochemical Society
412
N. A. Larsen and others
Table 3 Comparison of the compound 1-bound structure of H. influenzae
GlmU using RMSD analysis to the respective apo-, UDP- and UDP-GlcNAcbound structures
Uridyltransferase domain
Acetyltransferase domain
Overall
Apo
(PDB code 2V0H)
UDP
(PDB code 2V0K)
GlcNac-UDP
(PDB code 2V0I)
0.566
0.449
0.584
0.563
0.322
0.537
1.330
0.432
1.682
our series is replaced by a small acidic trifluoro sulfonamide or
a proton at the 4-quinazoline position. When we overlay these
quinazoline structures with others (Figure 5C), our ligand is
shifted and flipped in the binding site relative to the previously
reported compounds. The 2-position of the quinazoline ring
superimposes with our 4-position. It seems that the substitution
pattern at the 2- and 4-positions may dictate the binding mode of
the quinazoline core.
As a result of this alternative binding mode there are a number of
differences in their interactions. Although it would not be expected
to form a hydrogen bond interaction at physiological pH, Asp105 is
within hydrogen bond range of the 1-nitrogen of the quinazoline
core in the earlier reported structures. In contrast, the nitrogen
atoms in our core do not form any specific interactions (Figures 5A
and 5C). Another difference is the presence of a hydrogen bond
between the carbonyl of the tetrahydronapthyridinone substituent
(compound 11) and the backbone amide of Gly225 , whereas our
compound forms the hydrogen bond to the backbone carbonyl
of Val223 . Finally, there is an isobutyl group of compound 11
occupying the lipophilic pocket instead of an aromatic ring. The
Figure 5
inorganic sulfate in our crystal structure is in the same vicinity
as the trifluoro sulfonamide in the previously reported structure
(Figure 5C), suggesting a strategy to build further potency into
the series. Substitution at this 2-position could enhance potency
by accessing the GlcNAc-1-P-binding site.
An allosteric inhibitor of GlmU was reported (IC50 = 18 μM)
with a completely different binding mode than our quinazoline
series (Figure 5D, compound 12) [13]. This compound binds
outside both the GlcNAc and the UTP pockets. The only similarity
to our compound is that the phenyl substituent binds in the
lipophilic pocket with a very similar binding mode. By occupying
the lipophilic pocket, the allosteric inhibitor locks the enzyme
in the open conformation, thereby inhibiting the conformational
changes to the closed conformation, necessary for catalysis.
Compound 1 and other members of the aminoquinazoline
series inhibit isoenzymes of both Gram-negative species
tested, H. influenzae and E. coli. In contrast, there was no
measurable inhibition of isoenzymes of the Gram-positive species
S. pneumoniae or Staphyloccous aureus (results not shown).
The difference in potency between Gram-negative and Grampositive isoenzymes is not surprising given the comparatively
low sequence conservation in the N-terminal domains of
GlmU between Gram-positive and Gram-negative bacteria,
including some insertions and deletions that impact the tertiary
conformations of various loops (Supplementary Figures S2 and
S3 at http://www.BiochemJ.org/bj/446/bj4460405add.htm). The
overall amino acid sequence identity between H. influenzae
and S. pneumoniae GlmU is only 43 %, whereas that between
H. influenzae and E. coli is 65 %. Comparison of the S.
pneumoniae apo GlmU structure with GlmU from H. influenzae
reveals several amino acid sequence differences that would
be expected to impact binding affinity. Tyr103 forms a key
Details of the H. influenzae GlmU active site
(A) Compound 1 (magenta) superimposed on the product UDP-GlcNAc (green). Hydrogen bonds are indicated by broken lines. The quinazoline core superimposes on the ribose sugar. One hydrogen
bond is conserved between the ribose sugar and the 7-hydroxyl of the quinazoline. A sulfate ion was also modelled into difference density, but is not shown in the Figure for clarity (see C and D). (B)
Illustration of how quinazolines bind to the hinge region of kinases (PDB codes 2ITX and 2ITY). In this case, gefitinib (magenta) is shown bound to EGFR kinase, superimposed on where ATP (green)
binds. There is one hydrogen bond from the quinazoline to the backbone of the hinge, whereas ATP forms two. (C) Compound 1 superimposed on the quinazoline (pink, compound 11) found in
PDB code 2W0W. The quinazoline cores are flipped relative to each other. Arrow indicates position of the sulfate ion. (D) Compound 1 superimposed on an allosteric inhibitor (pink, compound 12)
of GlmU [13]. The only region that overlaps is the benzyl ring in the lipophilic pocket. Arrow indicates position of the sulfate ion.
c The Authors Journal compilation c 2012 Biochemical Society
Competitive inhibitors of GlmU uridyltransferase reaction
packing interaction in H. influenzae and this residue is an
alanine residue in the S. pneumoniae structure. In addition, the
lipophilic pocket in S. pneumoniae GlmU is smaller and is not
expected to accommodate the benzamide side chain without
further rearrangement. Therefore differences seen in the active
site and current SAR of the series suggest it would be difficult to
produce broad-spectrum inhibition with this scaffold, but it should
be feasible to obtain more potent inhibition of both Gram-negative
isoenzymes.
In conclusion, the present study reports the discovery of
a potent quinazoline series that inhibits GlmU. Although the
compounds are UTP competitive, they possess a unique binding
mode compared with structurally similar inhibitors binding to
protein kinases and demonstrate that similar compounds can have
significantly different action against different targets.
AUTHOR CONTRIBUTION
All of the authors participated in the writing of this paper. Peter Doig and Tory Nash
carried out the biochemical studies and developed the enzyme assays. Adam Shapiro
and Camil Joubran performed the NMR studies. Carolyn Blackett was responsible for
performance of the HTS and analysis of screening results. Nicholas Larsen was responsible
for the crystallization experiments and, with P. Ann Boriack-Sjodin, was responsible for
the analysis and interpretation of these results. Marshal Morningstar and Arthur Patten
performed chemical evaluation and analysis of the HTS data as well as compound synthesis.
ACKNOWLEDGEMENTS
We thank Beth Andrews and Ning Gao for constructs and protein reagents, the AstraZeneca
R&D Boston clinical susceptibility group for MIC determinations and Andrew D. Ferguson
for his help with structural refinements used in the present study. We also acknowledge
Helen Ashton for assistance with HTS screening.
FUNDING
This research received no specific grant from any funding agency in the public, commercial
or not-for-profit sectors.
REFERENCES
1 Silver, L. L. (2011) Challenges of antibacterial discovery. Clin. Microbiol. Rev. 24,
71–109
2 Bazan, J. A. and Martin, S. I. (2010) Ceftaroline fosamil: a novel broad-spectrum
cephalosporin. Drugs Today 46, 743–755
3 Kalan, L. and Wright, G. D. (2011) Antibiotic adjuvants: multicomponent anti-infective
strategies. Expert Rev. Mol. Med. 13, e5
4 Mengin-Lecreulx, D. and van Heijenoort, J. (1993) Identification of the glmU gene
encoding N -acetylglucosamine-1-phosphate uridyltransferase in Escherichia coli .
J. Bacteriol. 175, 6150–6157
5 Gehring, A. M., Lees, W. J., Mindiola, D. J., Walsh, C. T. and Brown, E. D. (1996)
Acetyltransfer precedes uridylyltransfer in the formation of UDP-N -acetylglucosamine in
separable active sites of the bifunctional GlmU protein of Escherichia coli . Biochemistry
35, 579–585
6 Mengin-Lecreulx, D. and van Heijenoort, J. (1994) Copurification of
glucosamine-1-phosphate acetyltransferase and N -acetylglucosamine-1-phosphate
uridyltransferase activities of Escherichia coli : characterization of the glmU gene product
as a bifunctional enzyme catalyzing two subsequent steps in the pathway for
UDP-N -acetylglucosamine synthesis. J. Bacteriol. 176, 5788–5795
7 Sulzenbacher, G., Gal, L., Peneff, C., Fassy, F. and Bourne, Y. (2001) Crystal structure of
Streptococcus pneumoniae N -acetylglucosamine-1-phosphate uridyltransferase bound to
acetyl-coenzyme A reveals a novel active site architecture. J. Biol. Chem. 276,
11844–11851
8 Olsen, L. R. and Roderick, S. L. (2001) Structure of the Escherichia coli GlmU
pyrophosphorylase and acetyltransferase active sites. Biochemistry 40, 1913–1921
413
9 Brown, K., Pompeo, F., Dixon, S., Mengin-Lecreulx, D., Cambillau, C. and Bourne, Y.
(1999) Crystal structure of the bifunctional N -acetylglucosamine 1-phosphate
uridyltransferase from Escherichia coli : a paradigm for the related pyrophosphorylase
superfamily. EMBO J. 18, 4096–4107
10 Kostrewa, D., D’Arcy, A., Takacs, B. and Kamber, M. (2001) Crystal structures of
Streptococcus pneumoniae N -acetylglucosamine-1-phosphate uridyltransferase, GlmU,
in apo form at 2.33 Å resolution and in complex with UDP-N-acetylglucosamine and
Mg2 + at 1.96 Å resolution. J. Mol. Biol. 305, 279–289
11 Mochalkin, I., Lightle, S., Zhu, Y., Ohren, J. F., Spessard, C., Chirgadze, N. Y., Banotai, C.,
Melnick, M. and McDowell, L. (2007) Characterization of substrate binding and catalysis
in the potential antibacterial target N -acetylglucosamine-1-phosphate uridyltransferase
(GlmU). Protein Sci. 16, 2657–2666
12 Milewski, S., Gabriel, I. and Olchowy, J. (2006) Enzymes of UDP-GlcNAc biosynthesis in
yeast. Yeast 23, 1–14
13 Mochalkin, I., Lightle, S., Narasimhan, L., Bornemeier, D., Melnick, M., Vanderroest, S.
and McDowell, L. (2008) Structure of a small-molecule inhibitor complexed with GlmU
from Haemophilus influenzae reveals an allosteric binding site. Protein Sci. 17, 577–582
14 Pereira, M. P., Blanchard, J. E., Murphy, C., Roderick, S. L. and Brown, E. D. (2009)
High-throughput screening identifies novel inhibitors of the acetyltransferase activity of
Escherichia coli GlmU. Antimicrob. Agents Chemother. 53, 2306–2311
15 Buurman, E. T., Andrews, B., Gao, N., Hu, J., Keating, T. A., Lahiri, S., Otterbein, L. R.,
Patten, A. D., Stokes, S. S. and Shapiro, A. B. (2011) In vitro validation of the acetyl
transferase activity of GlmU as an antibacterial target in Haemophilus influenzae . J. Biol.
Chem. 286, 40734–40742
16 Green, O. M., McKenzie, A. R., Shapiro, A. B., Otterbein, L., Ni, H., Patten, A., Stokes, S.,
Albert, R., Kawatkar, S. and Breed, J. (2012) Inhibitors of acetyltransferase domain of
N -acetylglucosamine-1-phosphate-uridyltransferase/glucosamine-1-phosphateacetyltransferase (GlmU). Part 1: hit to lead evaluation of a novel arylsulfonamide series.
Bioorg. Med. Chem. Lett. 22, 1510–1519
17 Liu, Y. and Gray, N. S. (2006) Rational design of inhibitors that bind to inactive kinase
conformations. Nature Chem. Biol. 2, 358–364
18 Pioli, D., Hockney, R. C., Kara, B. C. and Bundell, K. R. (1999) T7 promotor-based
expression system. WIPO Pat. WO/1999/005297
19 Mortlock, A. (2004) Preparation of phosphonooxy quinazoline derivatives as therapeutic
agents. WIPO Pat. WO/ 2004/058782
20 Mortlock, A., Keen, N., Jung, F. and Brewster, A. (2001) Prepartaion of quinazolines as
aurora kinase 2 inhibitors. WIPO Pat. WO/2001/021596
21 Brown, D., Morris, J. and Thomas, A. (1996) Preparation of N -phenyl-4-quinazolinamides
for the treatment of proliferative diseases. WIPO Pat. WO/1996/15118
22 Wikler, M. A. (2009) Methods for dilution antimicrobial susceptibility tests for bacteria
that grow aerobically: approved standard (9th edn), Clinical and Laboratory Standards
Institute, Wayne, PA
23 Lanzetta, P. A., Alvarez, L. J., Reinach, P. S. and Candia, O. A. (1979) An improved assay
for nanomole amounts of inorganic phosphate. Anal. Biochem. 100, 95–97
24 Bradford, M. M. (1976) A rapid and sensitive method for the quantitation of microgram
quantities of protein utilizing the principle of protein-dye binding. Anal. Biochem. 72,
248–254
25 Hwang, T. L. and Shaka, A. J. (1995) Water suppression that works. Excitation sculpting
using arbitrary wave-forms and pulsed-field gradients. J. Magn. Reson. Ser. A. 112,
275–279
26 Vonrhein, C., Flensburg, C., Keller, P., Sharff, A., Smart, O., Paciorek, W., Womack, T. and
Bricogne, G. (2011) Data processing and analysis with the autoPROC toolbox. Acta
Crystalogr. Sect. D Biol. Crystalogr. 67, 293–302
27 Bailey, S. (1994) The CCP4 suite: programs for protein crystallography. Acta Crystallogr.
Sect. D Biol. Crystallogr. 50, 760–763
28 Reference deleted
29 Emsley, P., Lohkamp, B., Scott, W. G. and Cowtan, K. (2010) Features and development of
Coot. Acta Crystallogr. Sect. D Biol. Crystallogr. 66, 486–501
30 Ladbury, J. E. (2010) Calorimetry as a tool for understanding biomolecular interactions
and an aid to drug design. Biochem. Soc. Trans. 38, 888–893
31 Siriwardena, A. H., Tian, F., Noble, S. and Prestegard, J. H. (2002) A straightforward
NMR-spectroscopy-based method for rapid library screening. Angew. Chem. Int. Ed.
Engl. 41, 3454–3457
32 Jahnke, W., Floersheim, P., Ostermeier, C., Zhang, X., Hemmig, R., Hurth, K. and Uzunov,
D. P. (2002) NMR reporter screening for the detection of high-affinity ligands. Angew.
Chem. Int. Ed. Engl. 41, 3420–3423
33 Reference deleted
34 Raetz, C. R. and Roderick, S. L. (1995) A left-handed parallel β helix in the structure of
UDP-N -acetylglucosamine acyltransferase. Science 270, 997–1000
Received 12 April 2012/13 June 2012; accepted 22 June 2012
Published as BJ Immediate Publication 22 June 2012, doi:10.1042/BJ20120596
c The Authors Journal compilation c 2012 Biochemical Society
Biochem. J. (2012) 446, 405–413 (Printed in Great Britain)
doi:10.1042/BJ20120596
SUPPLEMENTARY ONLINE DATA
An aminoquinazoline inhibitor of the essential bacterial cell wall synthetic
enzyme GlmU has a unique non-protein-kinase-like binding mode
Nicholas A. LARSEN*, Tory J. NASH†, Marshall MORNINGSTAR†, Adam B. SHAPIRO†, Camil JOUBRAN†, Carolyn J. BLACKETT‡,
Arthur D. PATTEN†, P. Ann BORIACK-SJODIN* and Peter DOIG*1
*Discovery Sciences Unit, AstraZeneca R&D Boston, Waltham, MA 02451, U.S.A., †Infection Innovative Medicines Unit, AstraZeneca R&D Boston, Waltham, MA 02451, U.S.A., and
‡Discovery Sciences Unit, AstraZeneca UK, Alderley Park, Macclesfield, Chesire SK10 4TG, U.K.
Figure S1
Detailed structural rearrangements in the UTP pocket
In the compound 1-bound structure (magenta, grey = GlmU), GlmU resembles the open form
of the enzyme. Gln79 donates a hydrogen bond to Gln76 . In the UTP- and/or product-bound
structure (green), Gln79 moves out of the way allowing Gln76 to form donor and acceptor
hydrogen bond interactions with UTP. Arrows indicate movements, and broken lines indicate
hydrogen bond interactions.
The atomic co-ordinates and structures factors from the present study have been deposited in the PDB under accession code 4E1K.
1
To whom correspondence should be addressed (email [email protected]).
c The Authors Journal compilation c 2012 Biochemical Society
N. A. Larsen and others
Figure S2
ClustalW sequence alignment of H. influenzae and E. coli GlmU
Blue lettering is identical, green is similar and red is dissimilar. The N-terminal uridyltransferase domain is 65 % identical and 25 % similar.
Figure S3
ClustalW sequence alignment of H. influenzae and S. pneumoniae GlmU
Blue lettering is identical, green is similar and black is dissimilar. The N-terminal uridyltransferase domain is 43 % identical and 26 % similar.
Received 12 April 2012/13 June 2012; accepted 22 June 2012
Published as BJ Immediate Publication 22 June 2012, doi:10.1042/BJ20120596
c The Authors Journal compilation c 2012 Biochemical Society