Rational design of inhibitors of the bacterial cell wall synthetic enzyme GlmU using virtual screening and lead-hopping

Article abstract of DOI:10.1016/j.bmc.2014.08.017

An aminoquinazoline series targeting the essential bacterial enzyme GlmU (uridyltransferase) were previously reported (Biochem. J. 2012, 446, 405). In this study, we further explored SAR through a combination of traditional medicinal chemistry and structure-based drug design, resulting in a novel scaffold (benzamide) with selectivity against protein kinases. Virtual screening identified fragments that could be fused into the core scaffold, exploiting additional binding interactions and thus improving potency. These efforts resulted in a hybrid compound with target potency increased by a 1000-fold, while maintaining selectivity against selected protein kinases and an improved level of solubility and protein binding. Despite these significant improvements no significant antibacterial activity was yet observed within this class.

Full text of DOI:10.1016/j.bmc.2014.08.017

Bioorganic & Medicinal Chemistry 22 (2014) 6256–6269
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Rational design of inhibitors of the bacterial cell wall synthetic
enzyme GlmU using virtual screening and lead-hopping
a
b
a
c
b
Peter Doig a, , P. Ann Boriack-Sjodin , Jacques Dumas , Jun Hu , Kenji Itoh , Kenneth Johnson ,
Steven Kazmirski a, Tomohiko Kinoshita c, Satoru Kuroda c, Tomo-o Sato c, Kaori Sugimoto c,
Katsumi Tohyama c, Hiroshi Aoi c, Kazusa Wakamatsu c, Hongming Wang b
⇑
a Discovery Sciences, AstraZeneca R&D Boston, 35 Gatehouse Drive, Waltham, MA 02451, United States
b Infection Innovative Medicines, AstraZeneca R&D Boston, Waltham, MA 02451, United States
c Wakunaga Pharmaceutical Co. Ltd, Akitakata City, Hiroshima 739-1195, Japan
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 1 November 2013
Revised 6 August 2014
Accepted 18 August 2014
Available online 16 September 2014
An aminoquinazoline series targeting the essential bacterial enzyme GlmU (uridyltransferase) were pre-
viously reported (Biochem. J. 2012, 446, 405). In this study, we further explored SAR through a combina-
tion of traditional medicinal chemistry and structure-based drug design, resulting in a novel scaffold
(benzamide) with selectivity against protein kinases. Virtual screening identified fragments that could
be fused into the core scaffold, exploiting additional binding interactions and thus improving potency.
These efforts resulted in a hybrid compound with target potency increased by a 1000-fold, while main-
taining selectivity against selected protein kinases and an improved level of solubility and protein bind-
ing. Despite these significant improvements no significant antibacterial activity was yet observed within
this class.
Keywords:
GlmU
Aminoquinazoline
Cell wall biosynthesis
N-Acetylglucosamine-1-phosphate
Uridyltransferase
Virtual screening
Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction
transfers uridyl monophosphate from UTP to form UDP-GlcNAc and
a pyrophosphate ion (PPi) and takes place entirely within the
Bacterial resistance to marketed antibiotics is a growing threat
to their efficacy and continued use in the clinic.1 This is particularly
true for Gram-negative bacteria, for which effective options for the
treatment of new resistant forms (e.g., New Delhi metallo-beta-lac-
tamase 1) continue to diminish. In the search for the next genera-
tion of antibiotics, circumventing existing resistance that resides
within the clinic has become the focus. One such approach exploits
novel targets for which mechanism-based resistance is absent in
the bacterial pools found in hospitals and other care centers.
N-terminal domain. Inhibitors of this enzyme would be expected
to have a dual mode of action in Gram-negative bacteria, preventing
both peptidoglycan and lipopolysaccharide synthesis, making it an
excellent target for drug discovery. In addition, absence of an
equivalent enzyme in humans makes GlmU especially attractive
as a target for antibacterial drug design.3
Recent efforts have been fruitful in finding small molecule
inhibitors to both the C-terminal acetyltransferase domain4,5 and
the N-terminal uridyltransferase domain.6,7 The sulfonamide ace-
tyltransferase inhibitors described by Buurman et al.4 have been
used to validate GlmU as a productive antibacterial target. Their
studies demonstrated that a potent inhibitor of GlmU function
has antibacterial properties consistent with disruption of peptido-
glycan and fatty acid biosyntheses.4 Further, mutants resistant to
the action of the sulfonamides were generated and the mutations
responsible for the resistance phenotype shown to reside in the
GlmU acetyltransferase domain.
GlmU is a bifunctional enzyme that converts D-glucosamine-1-
phosphate (GlcN-1-P) into uridyldiphosphate-N-acetylglucosamine
(UDP-GlcNAc), an essential precursor in both lipopolysaccharide
and peptidoglycan synthesis in bacteria. The enzyme catalyzes
two separate reactions in two distinctly separate structural
domains. The first reaction transfers an acetyl group from
acetyl-CoA to glucosamine-1-phosphate producing N-acetylgluco-
samine-1-phosphate (GlcNAc-1-P) and CoA.2 This reaction occurs
in the carboxy-terminal end of the polypeptide. The second reaction
The 4-anilino-quinazolines, exemplified by compound 1, were
recently identified by high-throughput screening (HTS) and data-
base mining.7 These analogs provide a starting point for the design
of potent GlmU inhibitors. This series inhibits Gram-negative
⇑
Corresponding author. Tel.: +1 781 839 5434; fax: +1 781 839 4570.
E-mail address: Peter.Doig@astrazeneca.com (P. Doig).
http://dx.doi.org/10.1016/j.bmc.2014.08.017
0968-0896/Ó 2014 Elsevier Ltd. All rights reserved.

P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
6257
isozymes, but have poor activity against Gram-positive isozymes.
This finding and a subsequent analysis of the binding mode from
crystallographic studies suggest the potential for further optimiza-
tion as Gram-negative specific agents. Compound 1 exhibits an
unexpected binding mode (Fig. 1). Despite being competitive with
UTP, the quinazoline ring binds into the ribose pocket of GlmU
rather than directly into the uracil pocket, via a single hydrogen
bond between the 7-position hydroxyl and the backbone of
Ala13 (also formed by the 2-oxygen in the uracil ring of UTP).
The lead does not exploit key hydrogen bonding interactions of
the uracil pocket, notably with Gln76, nor does it extend into the
GlcNAc binding pocket (Fig. 1). In addition, significant Aurora
kinase B inhibitory activity is observed in biochemical assays. We
expected that by maintaining and optimizing the two ends of this
lead series, improved potency and selectivity could be achieved.
Virtual screening (VS) based on either pharmacophore models
or docking is well precedented as an approach to finding and
improving inhibitor series.8 The success of this approach is greatly
aided by an understanding of the binding mode and ligand binding
constraints. Using VS to grow existing inhibitors towards unfilled
binding pockets has been employed to improve potency, usually
by linking known fragments into a large, more potent molecule.
More rarely was it employed to identify fragments to further drive
potency.
In addition, it is well known in the literature that substitution at
this position is not tolerated for binding into protein kinases, due
to a steric clash with the backbone of the hinge region.9
The result of C2 exploration is outlined in Table 1. Results of
simple substitution at C2 such as methyl (5a), amino (10), and
chloro (11) were encouraging, as GlmU inhibitory activities were
maintained in the low micromolar range. This led to the belief that
key hydrogen bonds are largely maintained with the C2 substitu-
tion. Additionally, Aurora kinase B selectivity was greatly improved
as both 5a and 10 showed >30 lM IC50 (Table 2). Although early
prototypes such as 5b and 5c did not extend all the way to the
sugar pocket of the active site and made few if any interactions
with the enzyme (Fig. 2A), these compounds highlighted the
potential of C2 as a vector towards this region of the enzyme,
and maintained a high level of selectivity against kinases tested.
The most extended analogs 12a and 12b were designed to extend
deeper in the pocket. Potencies in the biochemical assays were
mostly unchanged against Escherichia coli, but sometimes
improved against Haemophilus influenzae. However, all analogs
failed to make productive hydrogen bond contacts with residues
in the GlcNAc pocket. Additionally, low aqueous solubilities and
high plasma protein binding remained, although the potential for
higher solubility was observed with 12b. With these results, fur-
ther exploration of the GlcNAc pocket was abandoned.
In this study, a combination of traditional medicinal chemistry
and virtual screening resulted in a hybrid compound with ꢀ1000-
fold increased potency against GlmU, with selectivity towards
selected protein kinases.
Modeling data suggested that the hydroxyl group of the quinaz-
oline at C7 could be used as a vector towards the uracil pocket, and
several analogs were prepared to test this hypothesis (7a, 7b, and
7c). Of these, 7b showed the most promise with an improved aque-
ous solubility and balanced E. coli/H. influenzae GlmU activity in the
micromolar range (Table 1). The crystal structure showed that the
potency lost due to the lack of H-bond with Ala13 is partially offset
by two new hydrogen bonds formed with Gln76 and Gln79 in the
uracil pocket (Fig. 2B).
2. Results
2.1. Structure–activity relationships: Progress towards GlcNAc
and uracil pockets and kinase selectivity
Based on the modest results obtained from the exploitation of
the C2 position of the quinazoline scaffold, scaffold hopping was
pursued as an alternative approach to decrease kinase activity.
As the nitrogen at position one of the scaffolds does not interact
with residues in the N-terminal domain of GlmU, but is key
binding element for protein kinases, the decision was made to
remove it from the core structure. The 2-aminobenzamide 16b
The first synthetic chemistry efforts were to explore the uracil
and GlcNAc (sugar) pockets of GlmU to improve the potency of
the initial hit, while addressing the kinase selectivity of the series.
A close examination of the compound 1 binding mode (Fig. 1) sug-
gested that substitution at the C2 position of the quinazoline ring
of compound 1 could result in interactions with the GlcNAc pocket.
Figure 1. Quinazoline lead structure compound 1 (light grey/green7; PDB code 4E1K) superimposed with the structure of the UDP-GlcNAc product in H. influenzae GlmU
(dark grey23; PDB code 2V0I). The uracil and GlcNAc binding areas and the vectors from the lead compound that target those areas are indicated.

6258
P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
Table 1
H
N
O
HN
MeO
N
R1
N
R2
M)
H. influenzae
Compound
R1
R2
IC50
(l
Solubility
pH 7.0
Serum binding (%)
E. coli
pH 3.5
pH 8.0
1
OH
OH
OH
OH
H
CH3
2-Pyridyl
–Bn
1.3 0.3
2.2 0.8
2.7 0.8
4.8
0.26 0.04
0.39 0.11
1.5 1.2
1.8
—
—
—
—
5a
5b
5c
20
0.2
0.5
46
1.5
0.6
169
6.3
2.5
90.3
>99
>99
HO
*
*
7a
H
35
19
—
—
—
—
O
O
HO
7b
H
6.7 0.3
3.2 1.3
14
169
178
>99
O
H2N
*
O
7c
H
>22
>22
—
—
—
—
O
10
11
OH
OH
NH2
Cl
10.5 5.1
0.37 0.11
1.7 0.01
0.65 0.15
12
2.2
14
4.9
39
75
97.7
97.9
12a
12b
OH
OH
0.65 0.05
0.38 .01
1.12 .24
63
1.1
13
4.7
41
>99
*
N
N
H
*
N
N
4.5 1.5
127
98.8
N
H
Table 2
Protein kinase activity (IC50 in lM) of key analogs
Compound
AurB
IRAK4
IRAK1
JAK1
JAK2
JAK3
CDK1
CDK2
CDK9
1
5a
7a
7b
7c
10
12a
14a
14b
16a
16b
17a
24a
24b
24c
27
0.275 0.03a
>30
0.0273 0.0004
0.689 0.018
0.155 0.011
>30
>30
>30
24.4 2.4
>30
>30
>30
0.579 0.024
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
5.91 7.12
23.1 2.67
>30
>30
>30
>30
7.51 1.11
>30
14.3 0.5
>30
18.1 0.2
33
37
a
N = 2.
was the first prototype prepared (Table 3). This analog retains the
conformation of the quinazoline series via an internal hydrogen
bond between the amino and keto groups. Additionally, these com-
pounds retain GlmU biochemical activity while increasing selectiv-
ity against kinases, as expected (Table 2). Further simplification of
the core structure led to the minimum pharmacophore repre-
sented by 16a, and its pyridine analog 14a. Both compounds are
as potent against H. influenzae GlmU as the quinazoline series,
but are slightly less potent against the E. coli enzyme. Interestingly,
replacing the methoxy group of 14a with a nitro group, as in 14b,
slightly improved potency, presumably via an additional H-bond
interaction through the nitro group. The crystal structure of 14b
confirmed that the benzamide core is aligned with the phenyl ring
of the quinazoline scaffold (Fig. 2C), and all expected hydrogen
bonds were maintained, including the hydrogen bond between
the linker nitrogen and Asp 105. Additionally, the position of the
terminal aromatic ring was relatively unchanged. Interestingly,
the change to the benzamidine core resulted in a dramatic chance
to the linker portion of the molecule. Most dramatically, the
orientation of the terminal amide group flipped, eliminating the
hydrogen bond seen between the nitrogen and the backbone
carbonyl of Val 223. However, the new position of the amide
carbonyl places it within hydrogen bond distance of the backbone
amide of Gly 225. In effect, one inhibitor-protein interaction was

P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
6259
Figure 2. Structures of (A) compound 5c and (B) compound 7b in H. influenzae GlmU with final 2Fo–Fc density (1
r
). Hydrogen bonds are shown as dotted lines. (C)
Superposition of benzamide compound 14b (green/light grey) with compound 1 (dark grey). The positions of the benzamide headgroup and the terminal aromatic ring match
well with that of the original series. The benzamide headgroup changes the preferred orientation of the linker portion of the molecule.

6260
P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
Table 3
H
N
X
O
HN
R3
R1
O
R2
Compound
R1
R2
R3
X
IC50
(
l
M)
Solubility
pH 7.0
Serum binding (%)
E. coli
H. influenzae
pH 3.5
pH 8.0
14a
14b
16a
16b
OH
OH
OH
OH
H
H
H
NH2
OMe
NO2
OMe
OMe
N
N
C
8.2 3.3
3.1 0.1
18 3.3
5.9 1.1
0.51 0.29
0.05 0.007
1.4 0.2
10
8.3
18
8.5
18
23
43
19
44
98.4
>99
>99
98.9
0.7
8.9
54
C
6.1 0.3
HO
*
O
17a
17b
H
H
OMe
OMe
C
C
>67
>22
23 3.0
>22
—
—
—
—
—
—
—
—
O
H2N
*
O
O
substituted with another resulting in no gain or loss in activity.
Finally, attempts were made to extend the benzamide scaffold
towards the uracil pocket (compounds 17a and 17b), as previously
done with the quinazoline scaffold; unfortunately, this effort did
not lead to any breakthroughs in potency.
2.3. Structure–activity relationships: Hybrid analogs
The first hybrid prototype prepared, 20, added the uracil unit of
compound 38, the headgroup with superior ligand efficiency10
onto the minimum pharmacophore 16a, resulting in a significant
increase in biochemical potency (Table 4). We hypothesized that
the hydroxyl group of 20 would no longer contribute to potency,
and might even negatively affect the orientation of the uracil unit.
Therefore, we further optimized the series by removing this group,
and replacing the distal phenyl group with a pyridine. The result-
ing analog 24a shows potent inhibitory activity in the nanomolar
range against in both E. coli and H. influenzae GlmU biochemical
assays, yet maintains a molecular mass below 500 Da. Although
the biochemical potency towards the GlmU protein was dramati-
cally improved, solubility and plasma protein binding were nega-
tively affected.
In order to confirm the binding mode of the new series, the
structure of 24a was solved to 2.25 Å in H. influenzae GlmU. The
interactions seen with the fragment and the benzamide series
were maintained in the hybrid structure (Fig. 3C). Some flexibility
is seen in the uracyl binding pocket to accommodate the slight dif-
ference in the position of the uracyl group when attached to the
larger molecule.
Analysis of the phenyl ring (Table 4, compounds 24a, 24b, and
24c) yielded potent analogs, and confirmed the negative contribu-
tion of the substituent ortho to the uracil group. From this set, the
3-methoxy substitution also emerged as the most favorable, and
was used for further investigation of uracil replacements (30a,
30b, 33, and 36). Overall, while all uracil replacements showed a
fair level of activity, none matched the potency of the uracyl analog
24a. These analogs also share significant challenges associated
with solubility and protein binding.
The attempts to restore pharmaceutical properties while main-
taining fair levels of biochemical potency are summarized in
Table 5. The working hypothesis was that the N-benzoyl-1,4-phe-
nylene-diamine unit of the minimum pharmacophore 16a played
a major role in negatively impacting plasma protein binding and
aqueous solubility. With potent compounds in hand, modifications
were introduced that were predicted to improve solubility. In all
cases, the analogs prepared did have a drop in biochemical
potency, which appeared more pronounced in H. influenzae com-
pared to E. coli. Simpler amides containing sp3 hybridized carbons
2.2. Fragment identification and characterization
In order to address the moderate potency of the compounds, a
virtual screening approach was initiated in order to obtain novel
chemical matter as starting points for a new chemical series. Struc-
ture-based docking and ligand-based pharmacophore screening
utilized both the UTP and GlcNAc substrates as well as known
active compounds in order to expand the number of potential hits.
Additional areas of the binding site not engaged in inhibitor inter-
actions were also targeted. Fragments were selected for this virtual
screen in order to increase diversity of the potential hits. After
visual inspection of the highest scoring hits in the virtual screen,
320 fragments were selected for single point inhibition testing in
the H. influenzae GlmU assay. Active compounds (>33% activity at
10 mM concentration) were further characterized through dose
response and nephelometry experiments. An additional compound
quality check was performed by liquid chromatography/mass
spectrometry. Two fragments, compound 38 and 39, were identi-
fied, postulated to bind in the uracil pocket and selected for further
structural and biophysical studies (Fig. 3A).
NMR experiments were performed to confirm binding to the
UTP pocket. WaterLOGSY experiments with compound 38 and
compound 39 indicated that the compounds bound to H. influenzae
GlmU (data not shown). Subsequent addition of UTP resulted in
intensity reduction of both compounds’ WaterLOGSY signals,
which is indicative of competitive binding of the fragments with
UTP (data not shown). While a complex of compound 38 with H.
influenzae GlmU could not be obtained likely due to its poor solu-
bility, the structure of compound 39 was solved to a resolution of
2.1 Å (Fig. 3B). This fragment binds in the uracil binding site as pre-
dicted, with a tight network of contacts with Gln76 and the back-
bones of Ala13 and Gly81. An overlay of the fragment with the
benzamide series revealed that the phenyl ring of the fragment
overlapped with the hydroxy-methoxy phenyl headgroup of 14b,
suggesting that the uracil moiety could be merged onto the benz-
amide series (Fig. 3C).

P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
6261
Figure 3. (A) Structures, assay results and ligand efficiencies of fragment compounds found by virtual screening. (B) Structure of compound 39 in H. influenzae GlmU with
final 2Fo–Fc density (1s). Hydrogen bonds are indicated as dotted lines. (C) Superposition of compounds 39 (green) and 14b (cyan) onto the hybrid compound 24a (magenta).
such as 27 are well tolerated, yielding balanced activity in both
E. coli and H. influenzae GlmU biochemical assays, and much
improved solubility and plasma protein binding. Finally, a simple
ethyl urea, as in 37, also provides a promising avenue for future
investigation. While these analogs only represent a small set of
the possible variations, this study clearly demonstrates the
significant potential to modulate pharmaceutical properties with
this section of the molecule.

6262
P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
Table 4
H
N
X
O
HN
R1
O
R2
R3
Compound
R1
R2
R3
X
C
IC50
(l
M)
Solubility
pH 7.0
Serum binding (%)
E. coli
H. influenzae
pH 3.5
4.7
pH 8.0
8.1
*
O
20
OMe
OH
0.26 0.03
0.02 0.01
3.9
0.3
98.1
HN
HN
NH
O
*
O
24a
OMe
H
H
N
N
0.027 0.005
0.0017 0.002
0.4
2.1
0.3
2.5
-
NH
O
*
O
24b
H
0.35 0.01
0.005 0.001
1.6
98.1
HN
NH
O
*
O
24c
30a
H
OMe
N
N
>22
0.95 0.45
0.55 0.13
6.2
5.2
8.4
>99
HN
NH
O
*
N
N
OMe
H
21 2.2
0.06
175
188
96.8
N
N
H
*
N
O
NH
O
30b
33
OMe
OMe
H
H
N
N
>22
1.7 0.3
6.4
4.3
2.8
3.1
4
98.4
90.3
*
N
O
1.7 0.27
0.071 0.011
4.7
NH
O
*
36
OMe
H
N
4.2 1.4
0.98 0.02
7.5
7.1
12
>99
HN
N
O
molecule, high plasma protein binding, low solubility, modest
ligand efficiency and activity against other targets, leading to a
selectivity issue.
3. Discussion
Over the past decade, the discovery of novel antibiotic classes
has proven elusive. Previously we reported on a novel class of
quinazolines that were specific UTP competitive inhibitors of the
bacterial target GlmU uridyltransferase binding site.7 This HTS-
based effort had yielded limited SAR, with compounds exhibiting
low hundreds of nanomolar activity in in vitro enzyme assays. In
this study, we undertook the optimization of this quinazoline lead
series in an attempt to improve potency against the target and
develop measurable antibacterial activity. This series, exemplified
by compound 1, highlighted the typical challenges often encoun-
tered with hits derived from HTS: a relatively flat and lipophilic
Since GlmU is an enzyme only found in bacteria, target based
selectivity was expected. However, since HTS compounds fre-
quently originate from other drug discovery efforts, off target
selectivity is often an issue. Indeed, similar quinazolines to com-
pound 1 have been shown to be potent inhibitors of protein
kinases previously,11,12 and this series showed similar activities.
Potent inhibition such protein kinases would represent a risk of
secondary pharmacology that would not be desirable in an antibi-
otic. Therefore, we decided to first address target potency and
selectivity against human protein kinases. Exploitation of X-ray

P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
6263
Table 5
H
N
R1
O
HN
O
O
O
HN
NH
O
Compound
R1
IC50
(
lM)
Solubility
pH 7.0
Serum binding (%)
E. coli
H. influenzae
pH 3.5
177
pH 8.0
156
27
37
0.12 0.04
0.05 0.01
160
26
60.5
84.1
N
H
*
H
N
0.0.97 0.03
0.077 0.01
34
67
*
crystallography data (both on GlmU and on quinazolines bound to
protein kinases) was instrumental in quickly defining rational
approaches to manage kinase selectivity. For this effort, we opted
to prevent hinge binding either by substitution at the 2-position
of the quinazoline ring, or with the elimination of the position
one nitrogen atom of the quinazoline via scaffold-hopping (the
benzamide series exemplified by 16a). Both approaches were suc-
cessful in mitigating the kinase inhibitory activity from the series,
while having relatively little effect on GlmU inhibitory activity.
The ability to increase selectivity of the series versus a select set
of protein kinases was encouraging. Both approaches taken to mit-
igate this activity did so by selectively disrupting interactions with
the protein kinase and did not directly address the issue of GlmU
target potency. Structure-based design of analogs that could enter
the GlcNAc pocket was undertaken in an attempt to further drive
potency. While the position 2 carbon of the quinazoline scaffold
provided an ideal vector to add substituents directed towards
and into the GlcNAc pocket, we were unable to establish specific
contacts with this region of the enzyme. While potency appeared
to be better with some analogs, although not to a level that would
not be a step change, neither solubility nor ligand efficiency were
improved. Compound size and solubility (ie maximum attainable
concentration) are important attributes that affect permeation of
Gram-negative outer membranes,13 a key requirement to access
a cytoplasmic target such as GlmU. The inability to significantly
alter these properties while retaining and/or improving potency
with our attempts to access the GlcNAc pocket indicated that there
was little hope that an inhibitor with antibacterial activity could be
identified from this subclass. Alternatively, the ‘deconstruction’ of
the quinazoline ring towards benzamides also addressed the
kinase selectivity issue, and provided a slightly more favorable
minimum pharmacophore 16a in terms of size, albeit with high
plasma protein binding and solubility.
could then be incorporated into existing scaffolds using this
approach was rapid compared to traditional library synthetic
approaches. As was the case for the benzamide series, the ability
to exploit X-ray crystallography data and having well characterized
binding modes greatly strengthened this approach, increasing its
likelihood of success. The resulting hybrid analogs were tight bind-
ing (low nanomolar range), representing
a breakthrough in
potency, and offered a range of new possibilities. Further efforts
to improve the physicochemical properties while retaining
potency yielded improved compounds (27 and 37), but still failed
to show significant MICs even in efflux pump deficient strains.
Based on these results, it is hypothesized that failure to obtain
MICs is not the result of bacterial efflux alone, but rather poor com-
pound flux across the outer membrane. In addition it is likely that
further potency gains would be needed, requiring inhibitors with
sub-nanomolar activity.
The GlmU binding pocket is very large, and a number of hydro-
gen bonds are needed to secure the biochemical potencies usually
associated with functional activity in bacteria. In addition, subtle
SAR differences between the H. influenzae and the E. coli enzymes
were noted, adding further complexity to the identification of
broad spectrum Gram-negative compounds targeting this binding
site. Nevertheless, the most advanced analogs of the set (27 and
37) lose 1–2 logs of potency compared to 24a, but exhibit balanced
activity, with more favorable aqueous solubility and plasma pro-
tein binding.
Overall, the design of potent inhibitors of the N-terminal
domain of GlmU remains
a significant challenge. While we
improved potency by 1000 fold in the biochemical assays, func-
tional activity (MICs) remained elusive throughout the program.
4. Material and methods
4.1. GlmU protein expression and purification
Faced with these challenges, we undertook a virtual screening
approach to indentify fragment sized compounds that could access
binding sites adjacent to that of compounds such as 16a. In this
way we hoped to identify small pharmacophores with good ligand
efficiency that could then be exploited in future designs. The vir-
tual screening approach provided two new, small molecule frag-
ments, which could be hybridized with 16a. The example
described here showcases the strength of the fragment screen-
ing/hybridization technique in generating new ideas and starting
points. The identification of small fragment-like molecules that
E. coli and H. influenzae GlmU were expressed and purified as
previously described.4
4.2. GlmU inhibition assays
Activity (IC50s) of compounds against E. coli and H. influenzae
GlmU was determined using the malachite green based assay sim-
ilar to that previously described.7 The assay mixture consisted of

6264
P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
50 mM Hepes-NaOH buffer (pH 7.5) containing 5 mM DTT, 0.01%
Brij-35, 100 M EDTA, 0.3 U/mL pyrophosphatase, GlmU isozymes,
UTP, and N-acetyl-glucosamine-1-phosphate. The concentration of
UTP and N-acetyl-glucosamine-1-phosphate was 22 and 30 M for
H. influenzae, and 46 and 30 M for E. coli, respectively. The final
concentration of the GlmU isozymes was 0.26 and 0.11 nM for H.
influenzae and E. coli, respectively. Two L of a 3-fold serial DMSO
dilution of test compound ranging from 10 to 0.041 mM were
mixed with 70 L of the assay mixture per well in a 96-well plate.
After pre-incubation at room temperature for 15 min, 30 L of
5 mM MgCl2 were added to initiate the enzyme reaction. The reac-
tion was quenched by addition of 150 L of malachite green
reagent after 30 min incubation at room temperature, and the
absorbance of the solutions was measured at 665 nm using a
microplate reader (Envision, PerkinElmer Inc., US). As negative
and positive control, DMSO and 20 mM EDTA were used instead
of test compounds, respectively. The enzyme inhibitory activity
of test compounds was calculated by the following
4.4. Solubility and plasma protein binding measurements
l
The solubility of test compounds at three different pH values was
measured by a solution-precipitation method as follows. Solutions
containing 200 lM of test compound were prepared by diluting
10 mM compound, dissolved in 100% (v/v) DMSO, into each of three
buffer solutions (0.1 M citrate buffer at pH 3.5, 0.1 M phosphate buf-
fer at 7.0, and EPPS (3-[4-(2-hydroxyethyl)-1-piperazinyl]propane-
sulfonic acid) buffer at 8.0) directly into a 96-well filter plate device
l
l
l
l
l
(0.22 lm pore size filter, MultiScreen HTS, Millipore, US). These
solutions were incubated for 90 min shaking at 2000 rpm at 20 °C
(M-BR 022UP Taitec Bioshaker, Taitec, Japan). After centrifugation
of the filter plate to remove any precipitate, the compound concen-
tration in the filtrate was determined by high performance liquid
chromatography (HPLC) analysis. An HPLC system LC20A series
(Shimadzu Corporation, Japan), consisting of a model LC20AT pump,
a model CTO-20A autosampler, a model CTO-20A column oven, and
a model SPD-20A ultraviolet light detector, was used. The HPLC sep-
aration was performed using reverse phase C18 column (ODS80-
l
formula:(O.D.tested compound ꢁ O.D.negative control)/(O.D.positive
control
ꢁ O.D. negative control).
TM, 5
l
m, 4.6 ꢂ 150 mm, Tosoh Bioscience LLC., Japan) at a flow rate
IC50 values were calculated by using Kyplot (ver. 5.0, Keyence,
Japan). Hill slopes were generally within experimental error of 1
(range 0.7–1.6).
of 1 mL/min with 254 nm UV detection. The column temperature
was set at 40 °C. The mobile phase was consisted of acetonitrile
and water containing 20 mM sodium 1-decansulfonate (Tokyo
Chemical Industry, Japan), 40 mM phosphoric acid (Wako Pure
Chemical Industries, Japan), and 0.2% (v/v) triethylamine (Wako
Pure Chemical Industries, Japan) (40:60 to 50:50, v/v).
4.3. Kinase selectivity assays
A screening panel of protein kinases with representative mem-
bers from 4 branches of the kinome was assembled. Activity of
Aurora kinase B (AurB), IRAK1, IRAK4, JAK1, JAK2, JAK3, CDK1,
CDK2 and CDK9 was determined in vitro using a mobility shift
assay on a Caliper LC3000 reader (Caliper, MA), which measures
fluorescence of a phosphorylated and unphosphorylated fluores-
cent peptide substrate and calculates a ratiometric value to deter-
mine percent turnover. Phosphorylation of the peptide in the
presence and absence of the compound of interest was determined.
Enzyme/substrate/adenosine triphosphate (ATP) mix (5 mL) was
preincubated with 2 mL of compound for 20 min at 25 °C. Reac-
tions were initiated with 5 mL of 24 mM MgCl2 in 1.2ꢂ buffer
and incubated at 25 °C for 90 min and reactions were stopped by
addition of 5 mL of Stop mix consisting of 100 mM HEPES
(pH7.3), 121 mM EDTA, 0.8% Coatin Reagent 3 (Caliper, MA), and
0.01% Tween. Phosphorylated and unphosphorylated substrate
was detected by a Caliper LC3000 reader (Caliper, MA) in the pres-
ence of separation buffer consisting of 100 mM HEPES (pH7.3),
16 mM EDTA, 0.1% Coatin Reagent 3 (Caliper, MA), 0.015% Brij-
35, 5% DMSO, and 5.6 mM MgCl2.
Human serum protein binding was determined by an ultrafil-
tration method using a 96-well filter plate device (10 kDa cut-off
membrane filter, MultiScreen PPB, Millipore, US) and human
serum (Sigma-Aldrich, US). The protein binding was calculated
by following formula: Protein binding (%) = 100 ꢂ (Ct ꢁ Cu)/Ct,
where Cu and Ct are the bound-free and total concentrations of test
compounds in the serum solution, respectively. For the measure-
ment of Cu, 10 lM test compound solution was prepared by dilut-
ing a 1 mM solution of compound dissolved in 100% (v/v) DMSO
with human serum onto the well of the filter plate, and incubated
for 30 min at 37 °C. After centrifugation of the filter plate, the com-
pound concentration in the serum filtrate was determined by HPLC
analysis. For the measurement of Ct, 10 lM test compound solution
was prepared in 0.05 M phosphate buffer containing 0.09 M
sodium chloride (pH 7.4), and analyzed with HPLC.
4.5. Virtual screening methodology
Structure-based virtual screening was performed using a H.
influenzae GlmU crystal structure. The PPREP module of Maestro
Table 6
Crystallographic data collection and refinement statistics
5c
7b
14b
39
24a
Space group
H32
H32
H32
H32
H32
Cell constants a; b; c (Å)
109.3; 109.3; 328.0
108.7; 108.7; 326.2
108.2; 108.2; 325.9
108.8; 108.8; 327.5
108.2; 108.2; 327.9
Cell constants
Resolution limit (Å)
Resolution Range (Å)
a
; b;
c
(°)
90; 90; 120
2.10
42.0–2.10 [2.18–2.10]
90; 90; 120
1.90
40.8–1.90 [1.97–1.90]
90; 90; 120
2.21
31.04–2.21 [2.32–2.21]
90; 90; 120
2.09
48.7–2.09 [2.15–2.09]
90; 90; 120
2.31
30.6–2.31 [2.39–2.31]
Completeness overall (%)
Reflections, unique
Multiplicity
99.9 [100]
42262
6.9 [6.7]
11.2 [52.7]
21.6
99.8 [100]
55796
10.0 [9.7]
7.6 [52.1]
20.5
98.6 [100]
36717
10.9 [10.5]
8.1 [39.8]
18.5
99.5 [96.3]
41865
7.8 [7.1]
7.6 [47.6]
22.6
99.6 [99.6]
32767
9.9 [4.5]
6.5 [37.8]
19.4
1
RSym (%)
R-factoroverall (%)2
R-factorfree (%)
24.9
21.7
22.6
20.0
23.6
Non-hydrogen protein atoms
Non-hydgrogen ligand atoms
Solvent molecules
3442
88
427
3442
82
584
3437
86
615
3426
75
510
3442
92
366
PDB accession code
4KNR
4KNX
4KPX
4KPZ
4KQL

P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
6265
(Schrodinger, LLC, Portland, OR) was used for protein preparation
at pH = 7. After addition of hydrogens, the protein was subjected
to molecular mechanics refinement using the OPLS2001 force field
incorporated in IMPREF and minimized until the RMSD reached
0.3 Å.2 The minimized structure was used for docking. Water mol-
ecules within the active site were removed before generating the
docking grid (Schrodinger, Oregon, USA). The fragment library
was assembled using compounds from the internal corporate
collection as well as commercially available molecules. 3D
conformations were generated using Schrodinger’s LigPrep. The
fragments were subjected to hydrogen additions, ionization (from
pH 5–9) and generation of low-energy ring conformations. The
Scheme 1. Reagents and conditions: (a) Fe, AcOH; (b) picolinoyl chloride hydrochloride, Et3N, CH2Cl2; (c) 2-phenylacetyl chloride, Et3N, CH2Cl2; (d) 2 mol/L LiOH aq, 1,4-
dioxane; (e) (1) POCl3; (2) 40-aminobenzanilide, DBU, pyridine; (3) TFA; (f) (1) POCl3; (2) 40-aminobenzanilide, n-BuOH; (3) TFA.
Scheme 2. Reagents and conditions: (a) (1) 40-Aminobenzanilide, 4 mol/L HCl in 1,4-dioxane, DMAC, (2) TFA; (b) (1) (2-bromoethoxy)-tert-butyldimethylsilane, K2CO3, DMF,
(2) concd HCl, MeOH; (c) (1) ethyl 2-bromoacetate, K2CO3, DMF, (2) 2 mol/L LiOH aq, MeOH; (d) 2-iodoacetamide, K2CO3, DMF; (e) 4-nitroaniline, n-BuOH; (f) (1) 2,4-
dimethoxybenzylamine, microwave, (2) Fe, AcOH; (3) benzoyl chloride, pyridine; (4) TFA; (g) (1) H2, Pd/alumina, THF, (2) 40-aminobenzanilide, Et3N, IPA, DMF; (h) 2-(2-
aminoethyl)pyridine; (e) 1-(3-aminopropyl)imidazole.

6266
P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
chiralities of the fragments were preserved. Rigid receptor docking
was performed using Schrodinger’s Glide software.14 The docking
grid was generated using a box size of (18 ꢂ 18 ꢂ 18) Å3 and the
ligand range was defined using an inner box of (12 ꢂ 12 ꢂ 12) Å.3
Hits were ranked using Gscore scoring function. Final compounds
for enzymatic assay were selected after manual visualization.
4.7. Crystallography
Crystallization and structure determination of complexes with
H. influenzae GlmU was performed as previously described.7 Data
were collected on either a Fre+ rotation anode using a Saturn
944+ detector (Rigaku) or at the LRL-CAT beamline at the Advanced
Photon Source. Internal structures (data not shown) were used as
molecular replacement probes using AmoRe.17 Refinement was
performed using Refmac.18 with model building and water place-
ment performed using Coot.19 Final data collection, refinement sta-
tistics and PDB accession codes are reported in Table 6.
4.6. NMR
NMR binding experiments were conducted at 298°K on a
600 MHz NMR instrument with a Bruker AVANCE III console and
a triple-resonance cryogenic probe. In the WaterLOGSY experi-
ment,15 the first water-selective 180° Sinc pulse was 6 ms long,
and a weak rectangular pulse field gradient was applied during
the mixing time (1.8 s). A gradient recovery time of 2 ms was intro-
duced after the mixing time. Water suppression was achieved by
the excitation sculpting scheme16 and the water-selective 180°
Sinc shape pulse was 3 ms long. The data were collected with a
sweep width of 9157 Hz, 0.45 s acquisition time and 1.8 s for the
relaxation delay. 64 scans were recorded for each experiment,
which resulted in 5 min per spectrum. The data were zero filled
to 32768 complex points and multiplied by an exponential func-
tion (line broadening 3 Hz) prior to Fourier transformation. In a
4.8. Susceptibility testing
The minimum inhibitory concentration (MIC) was determined
by agar dilution method according to standard methods described
by the Japanese Society of Chemotherapy.20
One loop of bacteria was incubated in 3 mL Muller Hinton Broth
(MHB, Becton Dickinson and Company, US) at 37 °C for 18–24 h,
and 100-fold diluted with saline to prepare the 106 cfu/mL of inoc-
ulating solution of bacteria. For Streptococcus pneumoniae, Strepto-
coccus pyogenes, H. influenzae, and Moraxella catarrhalis, one loop of
bacteria cultivated on Chocolate II Agar plate (Becton Dickinson
and Company, US) was suspended in 3 mL MHB, followed by
25- or 50-fold dilution with MHB to prepare the inoculating solu-
tion of the bacteria.
typical WaterLOGSY experiment, 200
to 20 M H. influenzae GlmU in 50 mM HEPES pH 7.5, 5 mM MgCl2,
5 mM DTT, 0.1 mM EDTA and 5% D2O to examine binding; 200
lM compound was added
l
lM
UTP was then added to the same sample to compare the change of
signal intensities.
Five
lL of bacterial solutions was inoculated onto Muller
Hinton Agar (MHA, Becton Dickinson and Company, USA) plates
Scheme 3. Reagents and conditions: (a) (1) N-(4-Aminophenyl)pyridine-2-carboxamide, HBTU, HOBtꢃH2O, DIEA, DMF, (2) TFA, anisole; (b) N-(4-aminophenyl)pyridine-2-
carboxamide, HATU, HOAt, DIEA, DMF; (c) (1) 40-aminobenzanilide, HBTU, HOBtꢃH2O, DIEA, DMF, (2) H2, Pd(OH)2/C, EtOH; (d) (1) tert-butyl bromoacetate, K2CO3, DMF, (2)
TFA, CH2Cl2; (e) 2-bromoacetamide, K2CO3, DMF; (f) 40-aminobenzanilide, HBTU, HOBtꢃH2O, DIEA, DMF; (g) (1) uracil-5-boronic acid, Pd(dppf)2Cl2, 2 mol/L Na2CO3 aq, DMF,
(2) TFA.

P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
6267
Scheme 4. Reagents and conditions: (a) (1) benzylbromide, K2CO3, DMF, (2) 1 mol/L NaOH aq, EtOH; (b) (1) benzylbromide, K2CO3, DMF, (2) 1 mol/L NaOH aq, MeOH; (c) (1)
N-(4-aminophenyl)pyridine-2-carboxamide, HATU, HOAt, DIEA, DMF, (2) H2, Pd(OH)2/C, EtOH; (d) (1) N-(4-aminophenyl)pyridine-2-carboxamide, HBTU, HOBtꢃH2O, DIEA,
DMF, (2) H2, Pd/C, MeOH, DMF; (e) (1) N-(4-aminophenyl)pyridine-2-carboxamide, HATU, HOAt, DIEA, DMF, (2) H2, Pd/C, DMF; (f) (1) PhNTf2, K2CO3, THF, (2) uracil-5-boronic
acid, Pd(dppf)2Cl2, 2 mol/L Na2CO3 aq, DMF; (g) (1) N-Boc-p-phenylenediamine, HBTU, HOBtꢃH2O, DIEA, DMF, (2) TFA, CH2Cl2; (h) (1) Boc- -pipecolic acid, HBTU, HOBtꢃH2O,
L
DIEA, DMF, (2) H2, Pd/C, MeOH, DMF; (i) (1) PhNTf2, K2CO3, THF, (2) uracil-5-boronic acid, Pd(dppf)2Cl2, 2 mol/L Na2CO3 aq, DMF, (3) TFA, CH2Cl2; (j) (1) Tf2O, pyridine, CH2Cl2,
(2) ZnCN2, Pd(Ph3P)4, DMF, (3) LiOHꢃH2O, water, THF, (4) N-(4-aminophenyl)pyridine-2-carboxamide, HBTU, HOBtꢃH2O, DIEA, DMF; (k) NaN3, Et3NꢃHCl, toluene; (l) (1) K2CO3,
NH2OHꢃHCl, DMF, (2) CDI, THF.
containing serial 2-fold dilutions of each test compound ranging
from 128 to 0.002 g/mL. MHA supplemented with Sheep defibrin-
benzanilide under acidic conditions followed by debenzylation. A
small series of O-alkylations with appropriate halides led to quinaz-
oline derivatives 7a–c. Preparation of 2-aminoquinazoline 10 was
achieved by reaction of 8 with 4-nitroaniline, reaction with 2,4-
dimethoxybenzylamine under microwave irradiation, nitro group
reduction and acylation, followed by deprotection. N-substituted
12a and 12b were obtained by catalytic debenzylation of 8, leading
to 11, and sequential displacement of the two chlorine atoms with
40-aminobenzanilide and the appropriate primary amines.
l
ated whole blood (5%, v/v, Nippon Bio-Supply Center, Japan) and
MHA supplemented with Fildes extract (5%, v/v, Oxoid Ltd, UK)
were used instead of MHA for S. pneumoniae and S. pyogenes, and
H. influenzae and M. catarrhalis, respectively. After incubation at
37 °C for 18–24 h, the MIC was determined as the lowest
compound concentration, at which bacterial growth is invisible.
4.9. Chemical synthesis
Benzamides 14a and 14b can be prepared (Scheme 3) from
corresponding carboxylic acids 13a and 13b via amide bond
forming reaction with N-(4-aminophenyl)pyridine-2-carboxam-
ide and deprotection as needed. Benzamides 16a and 16b were
obtained from corresponding carboxylic acids 15a and 15b via
amide bond forming reaction with 40-aminobenzanilide followed
by catalytic hydrogenation. Further O-alkylation of compound
16a afforded the substituted acetate 17a and acetamide 17b.
Lastly, uracil-substituted benzamide 20 can be accessed from
the brominated carboxylic acid 18 via amide bond forming and
Suzuki coupling with uracil boronic acid, followed by
deprotection.
The preparation of uracil-substituted benzamides 24a–c
(Scheme 4) from the corresponding carboxylic acids 21a-c
involved an amide bond forming reaction with N-(4-aminophenyl)
pyridine-2-carboxamide as a key step. The Suzuki coupling with
uracil boronic acid was enabled by activations via triflate. Several
Detailed synthetic routes for the quinazoline and benzamide
derivatives prepared in this work are summarized in the Schema
1–5 and in Supplementary materials. For the quinazoline series,
starting materials such as 4-benzyloxy-5-methoxy-2-nitrobenza-
mide 2,21 7-benzyloxy-4-chloro-6-methoxyquinazoline 622 and 7-
benzyloxy-2,4-dichloro-6-methoxyquinazoline 821 were prepared
according to previously described methods.
Compound 3 was obtained by reduction of the nitro group of 2,
leaving the 2-methyl quinazoline 4a as a side product. Compounds
4b and 4c were prepared by reaction of 3 with appropriate acid chlo-
rides followed by alkaline treatment. Preparations of C-substituted
quinazolines 5a–c were completed by chlorination with phospho-
rus oxychloride and reaction with 40-aminobenzanilide followed
by treatment with trifluoroacetic acid (Scheme 1). Our lead com-
pound 1 (Scheme 2) was prepared by reaction of 6 with 40-amino-

6268
P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
Scheme 5. Reagents and conditions: (a) (1) acrylic acid, toluene, (2) urea, AcOH; (3) 1 mol/L NaOH, MeOH; (b) N-(4-aminophenyl)pyridine-2-carboxamide, HATU, HOAt,
DIEA, DMF; (c) (1) NaNO2, concd HCl, water, (2) KI, water; (d) (1) N-(4-aminophenyl)pyridine-2-carboxamide, HATU, HOAt, DIEA, DMF, (2) 2-chloropyrimidine-5-boronic acid,
Pd(dppf)2Cl2, 2 mol/L Na2CO3 aq, DMF; (3) AcONa, AcOH; (e) (1) N-Boc-p-phenylenediamine, HBTU, HOBtꢃH2O, DIEA, DMF, (2) TFA, CH2Cl2, (3) phenyl chloroformate, pyridine,
CH2Cl2, (4) EtNH2, Et3N, DMF, (5) uracil-5-boronic acid, Pd(dppf)2Cl2, 2 mol/L Na2CO3 aq, DMF.
protection/deprotection sequences were necessary to complete
these routes. Piperidine 27 followed a similar scheme, this time
from carboxylic acid 22a according to the similar procedure of
preparations of 24a–c. Lastly, the cyano-benzamide intermediate
29 allowed the preparation of tetrazole 30a and oxadiazolidinone
30b according to conventional methods.
Preparation of 33 (Scheme 5) was achieved via reaction of ani-
line 31 with acrylic acid and cyclization with urea, followed by
amide bond forming reaction with N-(4-aminophenyl) pyridine-
2-carboxamide. Standard conversion of a similar starting material
to iodide 35 followed by sequential amide bond formation and
Suzuki couplings, according to the similar procedure of prepara-
tions of 24a–c, led to analogs 36 and 37.
vapor. Reagents and solvents were used as obtained from commer-
cial supplier without further purification.
Acknowledgments
The authors wish to thank Vanitha Subramanian and Allan Wu
for their laboratory assistance and Dr. Steve Wasserman and the
staff at LRL-CAT for crystallographic data collection.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmc.2014.08.017.
4.10. General
References and notes
1H NMR spectra were recorded with a JEOL JNM-ECP500
(500 MHz) or a Varian NMR System PS500 (500 MHz). Chemical
Shifts are given in part per million (ppm) using tetramethylsilane
as the internal standard for spectra obtained in DMSO-d6 or CDCl3.
All J values are given in Hz. Mass (MS) were measured on a Finni-
gan LTQ mass spectrometer using ESI for ionization. Column chro-
matography was performed with indicated solvents using Merck
silica gel 60 (70–230 mesh) or Biotage KP-SIL (Silica) cartridge
(40–63 mm, 60A). Monitoring of reactions was carried out using
Merk 60 F254 silica gel, glass-supported TLC plates, followed by
visualization with UV light (254 nm) and staining with iodine
1. Silver, L. L. Clin. Microbiol. Rev. 2011, 24, 71.
2. Gehring, A. M.; Lees, W. J.; Mindiola, D. J.; Walsh, C. T.; Brown, E. D. Biochemistry
(N.Y.) 1996, 35, 579.
3. Milewski, S.; Gabriel, I.; Olchowy, J. Yeast 2006, 23, 1.
4. Buurman, E. T.; Andrews, B.; Gao, N.; Hu, J.; Keating, T. A.; Lahiri, S.; Otterbein,
L. R.; Patten, A. D.; Stokes, S. S.; Shapiro, A. B. J. Biol. Chem. 2011, 286, 40734.
5. Pereira, M. P.; Blanchard, J. E.; Murphy, C.; Roderick, S. L.; Brown, E. D.
Antimicrob. Agents Chemother. 2009, 53, 2306.
6. Mochalkin, I.; Lightle, S.; Narasimhan, L.; Bornemeier, D.; Melnick, M.;
Vanderroest, S.; McDowell, L. Protein Sci. 2008, 17, 577.
7. Larsen, N. A.; Nash, T. J.; Morningstar, M.; Shapiro, A. B.; Joubran, C.; Blackett, C.
J.; Patten, A. D.; Boriack-Sjodin, P. A.; Doig, P. Biochem. J. 2012, 446, 405.
8. Caporuscio, F.; Tafi, A. Curr. Med. Chem. 2011, 18, 2543.

P. Doig et al. / Bioorg. Med. Chem. 22 (2014) 6256–6269
6269
9. Bridges, A. J.; Zhou, H.; Cody, D. R.; Rewcastle, G. W.; McMichael, A.; Showalter,
H. D.; Fry, D. W.; Kraker, A. J.; Denny, W. A. J. Med. Chem. 1996, 39, 267.
10. Abad-Zapatero, C. Expert Opin. Drug Discov. 2007, 2, 469.
11. Marzaro, G.; Guiotto, A.; Chilin, A. Exp. Opin. Ther. Patents 2012, 22, 223.
12. Mortlock, A. A.; Keen, N. J.; Jung, F. H.; Heron, N. M.; Foote, K. M.; Wilkinson, R.
W.; Green, S. Curr. Top. Med. Chem. 2005, 5, 807.
13. Pages, J. M.; James, C. E.; Winterhalter, M. Nat. Rev. Microbiol. 2008, 6, 893.
14. Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.; Klicic, J. J.; Mainz, D. T.;
Repasky, M. P.; Knoll, E. H.; Shelley, M.; Perry, J. K.; Shaw, D. E.; Francis, P.;
Shenkin, P. S. J. Med. Chem. 2004, 47, 1739.
16. Stott, K.; Stonehouse, J.; Keeler, J.; Hwang, T.; Shaka, A. J. J. Am. Chem. Soc. 1995,
117, 4199.
17. Bailey, S. Acta Cryst. 1994, D50, 760.
18. Murshudov, G. N.; Vagin, A. A.; Dodson, E. J. Acta Crystallogr. Sect. D: Biol.
Crystallogr. 1997, 53, 240.
19. Emsley, P.; Lohkamp, B.; Scott, W. G.; Cowtan, K. Acta Cryst. 2010, D66, 486.
20. Japan Society of Chemotherapy Chemotherapy 1990, 38, 102–105.
21. Mortlock, A.; Keen, N.; Jung, F.; Brewster, A. Patent publication number WO
2001021596, 2001.
22. Althuis, T. H.; Hess, H. J. J. Med. Chem. 1977, 20, 146.
23. Mochalkin, I.; Lightle, S.; Zhu, Y.; Ohren, J. F.; Spessard, C.; Chirgadze, N. Y.;
Banotai, C.; Melnick, M.; McDowell, L. Protein Sci. 2007, 16, 2657.
15. Dalvit, C.; Fogliatto, G.; Stewart, A.; Veronesi, M.; Stockman, B. J. Biomol. NMR
2001, 21, 349.