Fragment-Based Drug Discovery of Inhibitors of Phosphopantetheine Adenylyltransferase from Gram-Negative Bacteria

Article abstract of DOI:10.1021/acs.jmedchem.7b01691

The discovery and development of new antibiotics capable of curing infections due to multidrug-resistant and pandrug-resistant Gram-negative bacteria are a major challenge with fundamental importance to our global healthcare system. Part of our broad program at Novartis to address this urgent, unmet need includes the search for new agents that inhibit novel bacterial targets. Here we report the discovery and hit-to-lead optimization of new inhibitors of phosphopantetheine adenylyltransferase (PPAT) from Gram-negative bacteria. Utilizing a fragment-based screening approach, we discovered a number of unique scaffolds capable of interacting with the pantetheine site of E. coli PPAT and inhibiting enzymatic activity, including triazolopyrimidinone 6. Structure-based optimization resulted in the identification of two lead compounds as selective, small molecule inhibitors of bacterial PPAT: triazolopyrimidinone 53 and azabenzimidazole 54 efficiently inhibited E. coli and P. aeruginosa PPAT and displayed modest cellular potency against the efflux-deficient E. coli ΔtolC mutant strain.

Full text of DOI:10.1021/acs.jmedchem.7b01691

Subscriber access provided by UNIV OF SCIENCES PHILADELPHIA
Article
Fragment-Based Drug Discovery of Inhibitors of Phosphopantetheine
Adenylyltransferase from Gram-Negative Bacteria
Robert J Moreau, Colin K Skepper, Brent A. Appleton, Anke Blechschmidt, Carl J. Balibar, Bret M.
Benton, Joseph E Drumm, Brian Y Feng, Mei Geng, Cindy Li, Mika K Lindvall, Andreas Lingel, Yipin Lu,
Mulugeta Mamo, Wosenu Mergo, Valery Polyakov, Thomas M Smith, Kenneth Takeoka, Kyoko Uehara,
Lisha Wang, Jun-Rong Wei, Andrew H. Weiss, Lili Xie, Wenjian Xu, Qiong Zhang, and Javier de Vicente
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.7b01691 • Publication Date (Web): 02 Mar 2018
Downloaded from http://pubs.acs.org on March 3, 2018
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Fragment-Based Drug Discovery of Inhibitors of
Phosphopantetheine Adenylyltransferase from
Gram-Negative Bacteria
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Robert J. Moreau,* Colin K. Skepper, Brent A. Appleton, Anke Blechschmidt, Carl J. Balibar,
Bret M. Benton, Joseph E. Drumm, III, Brian Y. Feng, Mei Geng, Cindy Li, Mika K. Lindvall,
Andreas Lingel, Yipin Lu, Mulugeta Mamo, Wosenu Mergo, Valery Polyakov, Thomas M. Smith,
Kenneth Takeoka, Kyoko Uehara, Lisha Wang, Jun-Rong Wei, Andrew H. Weiss, Lili Xie,
Wenjian Xu, Qiong Zhang, Javier de Vicente
Novartis Institutes for BioMedical Research, 5300 Chiron Way, Emeryville, California 94608,
United States
KEYWORDS. FBS, FBDD, PPAT, phosphopantetheine, adenylyltransferase, coaD, Gram-
negative, antibacterial, antibiotic.
ABSTRACT. The discovery and development of new antibiotics capable of curing infections
due to multidrug-resistant and pandrug-resistant Gram-negative bacteria is a major challenge
with fundamental importance to our global healthcare system. Part of our broad program at
Novartis to address this urgent, unmet need includes the search for new agents that inhibit novel
bacterial targets. Here we report the discovery and hit-to-lead optimization of new inhibitors of
phosphopantetheine adenylyltransferase (PPAT) from Gram-negative bacteria. Utilizing a
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
fragment-based screening approach, we discovered a number of unique scaffolds capable of
interacting with the pantetheine site of E. coli PPAT and inhibiting enzymatic activity, including
triazolopyrimidinone 6. Structure-based optimization resulted in the identification of two lead
compounds as selective, small molecule inhibitors of bacterial PPAT: triazolopyrimidinone 53
and azabenzimidazole 54 efficiently inhibited E. coli and P. aeruginosa PPAT and displayed
modest cellular potency against the efflux-deficient E. coli ΔtolC mutant strain.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
INTRODUCTION
The discovery of safe and effective broad-spectrum antibiotics stands as one of the landmark
th
scientific achievements of the 20 century. The golden age of antibiotics, beginning in the
1
940s, was characterized by the development of drugs such as penicillin and vancomycin that
transformed the practice of medicine, turning life-threatening infections into curable, short-lived
illnesses. Antibiotics also enabled the development of many medical procedures that we now
consider routine including complex surgeries, implantable devices and cancer chemotherapy.
Over the past few decades, however, it has become increasingly evident that continued use of
available antibiotics is threatened by the rise of resistance, a situation that has prompted the
1
World Health Organization and others to warn of an impending “post-antibiotic era”. The CDC
currently estimates that antibiotic-resistant infections are responsible for at least 2 million
illnesses and 23,000 deaths each year in the United States, with direct societal costs as high as
2
$
20 billion. Of particular concern is the spread of multidrug-resistant and pandrug-resistant
Gram-negative bacteria. The Gram-negative cell envelope presents a formidable barrier to entry
3
for small-molecule drugs and additional, specific resistance mechanisms have evolved that
dramatically decrease the efficacy of many known classes of antibiotics (e.g. β-lactams,
4
fluoroquinolones, aminoglycosides, etc.). As a result, the toxic, polycationic peptide colistin
ACS Paragon Plus Environment
2

Page 3 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
(
polymyxin E) has seen a resurgence in use and is now widely regarded as the antibiotic of last
resort. Alarmingly, strains of E. coli were recently identified that harbored a transferable,
plasmid-mediated mechanism of resistance to colistin (MCR-1), and the mcr-1 gene is
5
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
spreading.
Compounding the issue of antibiotic resistance is the fact that very few new drugs have been
approved to treat Gram-negative infections in recent decades, and the majority of these have
6
come from well-known classes of drugs discovered half a century ago. Indeed, since the
introduction of carbapenems in the 1980s, no compound from a new structural class has been
approved for the treatment of Gram-negative infections. This decline in antibiotic discovery is
due in part to the withdrawal from the field by many large pharmaceutical companies as a result
of the conjoined factors of unfavorable economics and the considerable scientific challenge
7
posed by the impermeability of Gram-negative pathogens.
One strategy to mitigate the risks associated with antibiotic development is to identify analogs
of an existing structural class with reduced susceptibility to resistance and/or improved intrinsic
6
activity. However, the inevitability of cross-resistance dictates that there is a clear, unmet need
for agents that inhibit novel bacterial targets. One such target that has received attention recently
is phosphopantetheine adenylyltransferase (PPAT, CoaD). PPAT is a hexameric enzyme that
catalyzes the penultimate step in CoA biosynthesis. CoA is an essential cofactor found in all
living organisms, functioning as a carrier for activated acyl groups in many vital metabolic
processes. For example, CoA plays a central role in lipid metabolism by providing
phosphopantetheine prosthetic groups that are post-translationally appended to acyl carrier
proteins upon which fatty acids are biosynthesized. CoA also has roles in various other cellular
8
processes, from the citric acid cycle to amino acid degradation. In bacteria, CoA is a critical
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
cofactor in the biogenesis of membrane lipids, peptidoglycan, teichoic acids in Gram-positive
9
,10
organisms, and Lipid A in Gram-negative organisms.
The biochemical pathway for the formation of CoA from pantothenic acid (vitamin B5)
consists of five enzymatic steps that are conserved across all taxa (Figure 1). Biosynthesis
begins with the phosphorylation of pantothenic acid to 4’-phosphopantothenate catalyzed by
pantothenate kinase (PanK). 4’-Phosphopantothenate is subsequently condensed with cysteine
by phosphopantothenoylcysteine synthetase (PPCS) to provide 4’-phosphopantothenoylcysteine,
which is in turn decarboxylated by phosphopantothenoylcysteine decarboxylase (PPCDC) to
form 4’-phosphopantetheine. An AMP unit is appended to 4’-phosphopantetheine by PPAT to
give dephospho-CoA, the substrate for phosphorylation of the ribose 3’-hydroxyl by dephospho-
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
11
CoA kinase (DPCK) to form CoA.
PanK
CoaA)
PPCS
O
O
O
O
HO C
O
O
2
-
-
(
(CoaB)
OPO 2
HS
OPO 2
OH
3
3
HO
N
HO
N
N
N
H
H
H
H
OH
OH
OH
acid
B5)
pantothenic
4'-phosphopantothenate
4'-phosphopantothenoylcysteine
(vitamin
PPCDC
(CoaC)
NH2
7
6
N
1
N
PPAT
O
O
O
O
O
O
N
(CoaD)
-
N
OPO 2
3
HS
O
P
O
P
O
+
HS
ATP
N
N
-
-
N
N
H
H
O
O
O
H
H
OH
OH
PPi
4'-phosphopantetheine
O
OH
=
R
R
H
,
R
dephospho-CoA
DPCK
CoaE)
(
=
PO32-, coenzyme A
Figure 1. Biosynthetic pathway for the conversion of pantothenic acid (vitamin B5) to CoA.
PPAT (CoaD), which catalyzes the reaction between ATP and 4’-phosphopantetheine, is
12
essential for bacterial growth and exhibits high sequence homology amongst bacterial isoforms.
In the human CoA biosynthesis pathway, the role of CoaD is performed by a bifunctional
PPAT/DPCK enzyme called CoA synthase (CoaSY), which does not share significant homology
ACS Paragon Plus Environment
4

Page 5 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
13
with bacterial PPAT, though it is similar to bacterial DPCK.
As such, bacterial PPAT
represents an attractive target for development of a new class of antibacterial agents. This is
supported by a report from AstraZeneca describing the discovery of a series of
1
4
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
cycloalkylpyrimidine PPAT inhibitors using high-throughput screening. Optimization of this
series provided compounds with potent inhibitory activity against Gram-positive species, but
much weaker activity against Gram-negative isolates. While small molecule inhibitors of
1
5
bacterial PPAT had been reported previously (1, Figure 2), the AstraZeneca compounds (2, 3)
were the first to demonstrate inhibition of bacterial growth at the cellular level and in an animal
infection model. However, several observed issues including toxicity, solubility, protein binding
and high in vivo clearance appear to limit the future potential of this class of compounds. Below
and in the accompanying manuscript, we describe our own work on the discovery and
optimization of potent inhibitors of PPAT from Gram-negative bacteria using an alternative
approach based on fragment-based screening.
O
N
N
NH2
O
N
O
CO H
HN
O
2
N
OH
N
NH
N
H
N
N
H
N
O
O
H
N
Cl
Cl
Cl
Cl
CO H
N
2
N
H H
H H
H
O
NHFmoc
1
2
3
Figure 2. Structures of known inhibitors of bacterial PPAT.
RESULTS AND DISCUSSION
The selection of a target enzyme for screening was guided by in silico analysis of PPAT from
four Gram-negative organisms: E. coli, P. aeruginosa, K. pneumoniae, and A. baumannii. Based
on the high degree of conservation observed for residues in and around the enzyme active site, E.
coli and P. aeruginosa PPAT were selected as representative enzymes for assay development. In
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
the event, the greater stability of E. coli PPAT led to a larger suite of more robust biochemical
and biophysical assays, including SPR, and this enzyme was chosen for assay miniaturization
and compound screening. Validated inhibitors would be profiled against P. aeruginosa PPAT,
and evaluated for selectivity against S. aureus PPAT and human CoaSY. Broad biochemical and
cellular profiling was planned for later in the program once sufficient gains in potency had been
achieved.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
In an effort to circumvent the well-described incompatibilities between HTS campaigns and
antibacterial drug targets, we placed an emphasis on fragment-based lead discovery to escape the
confines of our optimized compound library and discover smaller hits with lower
7
,16
lipophilicity.
Primary fragment screening against E. coli PPAT was conducted using
orthogonal biochemical and biophysical techniques. Inhibition of enzymatic activity was
evaluated in the context of a PPAT/pyrophosphatase coupled-enzyme assay that relied on
indirect measurement of pyrophosphate, the stoichiometric by-product of the PPAT-catalyzed
17
reaction between 4’-phosphopantetheine and ATP. During the course of the assay, inorganic
pyrophosphatase converted pyrophosphate into phosphate, which was subsequently detected
1
8
using a molybdate/malachite green reagent. This coupled-enzyme assay is amenable to both
1
9
absorbance- and fluorescence-based readouts. Two separate biochemical FBS campaigns were
conducted. With an absorbance-based readout in 384-well format at 500 µM compound
th
20
concentration, the Novartis 4 generation core fragment library of 1408 members, an expanded
library of their fragment analogs, and other compounds solubilized at 50 mM in DMSO-d were
6
screened in the coupled-enzyme assay. From a total of 25K compounds screened, 631 hits were
identified that achieved ≥30% inhibition. In a complementary screen with a fluorescence-based
readout in 1536-well format at 625 µM compound concentration, the core and expanded
ACS Paragon Plus Environment
6

Page 7 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
fragment libraries were screened. From a total of 16K compounds, 126 hits were identified
using a similar cutoff. While it was reassuring to find a high degree of overlap between
screening hits (>80%) despite the different assay formats and readouts, we corrected the low hit
rate in the latter screen by rescuing ~300 weakly-active hits and carrying them forward through
hit validation. We also obtained 26 fragment hits by screening an additional 4K fragments at 50
µM compound concentration as part of an HTS of the Novartis compound archive.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
0
In parallel, biophysical evaluation of the core fragment library from above in a set of 176 8-
1
fragment mixtures was performed by screening at 200 µM in three ligand-observed H NMR
binding experiments: T1ρ, water-ligand observed by gradient spectroscopy (WaterLOGSY), and
2
1–23
saturation transfer difference (STD).
These NMR experiments produced 201 hits, which
were combined with the biochemical hits from above. A limited hit expansion to include very
close analogs gave 1963 fragments that were advanced to biochemical and biophysical hit
validation, where fragments that were either ≥3-fold selective for PPAT over inorganic
pyrophosphatase or showed >1 °C ΔT by DSF were advanced. These criteria provided 488
m
validated fragment hits that displayed low overlap between validation methods, and 125
fragments were prioritized based on structural diversity, polarity, ligand efficiency and level of
hit validation. These fragments were examined using a robust E. coli PPAT X-ray
crystallography soaking system that typically produced crystals that diffracted at ≤2 Å, and 39
high-quality X-ray co-crystal structures were obtained. A diverse set of these co-crystallized
fragments is shown in Figure 3.
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
N
N
N
H
N
N
N
OH
NH
N
N
CF3
O
F
N
H
4
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
PPAT IC50 = 273 µM
PPAT DSF DT = 2.3 °C
LE = 0.31, cLogD7.4 = 2.0
PPAT IC50 = 230 µM
PPAT DSF DT = 8.1 °C
LE = 0.36, cLogD7.4 = 2.1
PPAT IC50 = 31 µM
PPAT DSF DT = 2.5 C
LE = 0.30, cLogD7.4 = 1.8
°
m
m
m
O
N
N
N
N
N
N
N
Cl
t-Bu
N
NH2
N
N
O
H
CF3
7
8
9
PPAT IC50 = 264 µM
PPAT DSF DT = 2.7 °C
LE = 0.27, cLogD7.4 = 1.9
PPAT IC50 = 37 µM
PPAT DSF DT = 5.1 °C
LE = 0.3, cLogD7.4 = 2.4
PPAT IC50 >2500 µM
PPAT DSF DT = 2.5 °C
m
m
m
cLogD7.4 = 2.0
O
O
O
t-Bu
N
S
NH
H
NH2
NH2
O
N
N
N
N
H
H N
2
10
11
12
PPAT IC50 = 74 µM
PPAT DSF DT = 8.7 °C
LE = 0.29, cLogD7.4 = 1.2
PPAT IC50 = 199 µM
PPAT DSF DT = 3.1 °C
LE = 0.26, cLogD7.4 = 2.7
PPAT IC50 = 1076 µM
PPAT DSF DT = N/A
LE = 0.32, cLogD7.4 = -0.38
m
m
m
O
O
O
H
N
N
N
O
H
NH2
N
N
F
HN
H
N
N
F
O
F
O
13
14
15
PPAT IC50 = 69 µM
PPAT DSF DT = 3.4 °C
LE = 0.27, cLogD7.4 = 0.27
PPAT IC50 = 237 µM
PPAT IC50 = 163 µM
°
PPAT DSF DT = 2.9 C
m
°
PPAT DSF DT = 3.8 C
m
m
LE = 0.24, cLogD7.4 = 1.6
LE = 0.25, cLogD7.4 = 1.0
Figure 3. Selected fragments with available E. coli X-ray co-structure and associated
biochemical, crystallography and DSF data are shown. Atoms colored in green highlight their
interaction with the backbone NH of M74. For ligand efficiency, LE = (1.37/HA) × pIC ,
5
0
where HA is the number of non-hydrogen (heavy) atoms.
ACS Paragon Plus Environment
8

Page 9 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Interestingly, all fragment hits that crystallized with E. coli PPAT were found to bind at the
pantetheine site of PPAT. This was somewhat unexpected considering the well-defined adenine
2
4a
site in the reported X-ray co-crystal structures of E. coli PPAT, though binding does not
appear to involve the typical N1/N6 hydrogen bond acceptor/donor interactions that are well
represented by fragments in our library. Instead, adenine enjoys hydrophobic interactions with
several surrounding residues, a direct hydrogen bond between N6 and the carbonyl of W124, and
a water-mediated hydrogen bond between N7 and the sidechain hydroxyl of T15. Consideration
of the full ATP binding site reveals several polar interactions with the triphosphate that ensure
proper orientation of the α-phosphate and stabilization of the pentavalent transition state during
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
4
catalysis.
Overall, this results in a more polar, shallow and solvent-exposed binding site
relative to the phosphopantetheine site, which is deeper and has a good balance of
hydrophobicity and polarity. Given that we screened at 10x the K for phosphopantetheine and
m
only 2x the Km for ATP, it is especially surprising that we enriched so strongly for
phosphopantetheine site binders. This suggests that the phosphopantetheine site is primed for
binding at lower concentrations, in line with each substrate’s K and their expected relative
m
intracellular concentrations. However, a recent report on P. aeruginosa PPAT nicely delineated
the complex regulation and binding dynamics of PPAT, and especially the ATP pocket thereof,
24b
which could lead to alternate proposals that help rationalize our observations.
A feature common to the binding of most of our fragments to E. coli PPAT is the ability to
accept a hydrogen bond from the NH of M74, which is the same residue at the center of the
pantetheine site that engages the cysteamide carbonyl of the natural substrate. Detailed analysis
of our fragment-bound X-ray co-crystal structures, especially those with overlapping fragments,
2
5
enabled us to map the phosphopantetheine binding site of PPAT and identify “hot spots” and
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
polar interactions to help drive our structure-based drug discovery effort. This information
2
6
would prove valuable in the subsequent optimization campaign.
To advance our lead generation efforts, we performed fragment hit prioritization according to
the following desirable properties: high ligand efficiency and polarity, attractive vectors for
expansion, and high-quality interactions with residues conserved across multiple bacterial
species. As a result of this analysis, fragments 4–6 were chosen as starting points for medicinal
chemistry. Fragment optimization efforts for compound 4 began with the installation of a methyl
group at C5 (16, Figure 4) in order to gain additional interactions with V135, M105 and L131 in
a small, hydrophobic site in the binding pocket into which several other ligands were observed to
project a small, lipophilic group (Figure 5A). To our delight, this modification yielded a 15-fold
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
27
improvement in biochemical activity. Incorporation of the halogen found in closely-related
fragment 7 produced m-chlorobenzyl analog 17, which possessed single-digit micromolar
inhibitory activity. Unfortunately, attempts to grow this fragment at the 2-position of the
azabenzimidazole for further optimization proved challenging due to the substituent at N1, so we
deprioritized this fragment and focused instead on exploring a more efficient inhibitor, 4-
hydroxybenzimidazole 5, which possessed a slightly different binding mode from that of the
azabenzimidazoles.
N
N
N
5
N
N
N
N
N
N
2
Cl
4
16
17
PPAT IC50 = 273 µM
PPAT IC50 = 19 µM
PPAT IC50 = 4.3 µM
LE = 0.31, cLogD7.4 = 2.0
LE = 0.38, cLogD7.4 = 2.4
LE = 0.41, cLogD7.4 = 3.1
Figure 4. Fragment hit optimization of E. coli PPAT inhibitors azabenzimidazoles 4, 16–17.
ACS Paragon Plus Environment
1
0

Page 11 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
A
B
Figure 5. Fragment-bound X-ray crystal structures of E. coli PPAT. (A) Azabenzimidazole 4.
B) Hydroxybenzimidazole 5. Protein carbon atoms are shown in gray, ligand carbon atoms in
(
orange, nitrogen in blue, oxygen in red, and fluorine in green. Water molecules are shown as red
spheres, hydrogen bonds as cyan dotted lines.
In contrast to the binding mode of fragment 4, where the azabenzimidazole core accepted
hydrogen bonds from both M74 and a conserved water molecule (Figure 5A), the hydroxyl
group of fragment 5 displaced the conserved water to form direct interactions with the sidechain
carbonyl of N106 and the NH of A75 (Figure 5B). This resulted in deeper binding of the
hydroxybenzimidazole relative to the azabenzimidazole and slightly altered the trajectory of the
ACS Paragon Plus Environment
1
1

Journal of Medicinal Chemistry
Page 12 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
lone pair at N3 without disrupting its ability to accept a hydrogen bond from M74; however, the
substituent vectors away from this pivot point were significantly affected. Despite the successful
improvement in biochemical activity observed following the installation of a methyl group at C5
of azabenzimidazole 4, the deeper binding and altered vector at C5 of hydroxybenzimidazole 5
suggested that the addition of a methyl group would be unfavorable due to a clash with the N106
sidechain, with which the hydroxyl made a key hydrogen bonding interaction, as well as the
sidechain of L109. We turned our attention to the opposite end of the molecule, where the now
repositioned and unsubstituted N1 of fragment 5 participated in a hydrogen bond between the
benzimidazole NH and the sidechain of E134. This glutamate sidechain extends from the short
α-helix that forms the upper limit of the pantetheine pocket and it engages the cysteamide NH of
the natural substrate. To maintain this interaction and explore the new vector at C2, we
conducted a survey of our compound archive for analogs differentiated at the 2-position of the
benzimidazole. We were fortunate to discover hydroxymethyl analog 18, which displayed
improved activity and efficiency (Table 1). An X-ray co-crystal structure with E. coli PPAT
confirmed the previously observed network of hydrogen bonds, and an overlay with
azabenzimidazole 17 (not shown) suggested that we could use the hydroxyl as a handle to
similarly position a phenyl group and improve the potency. This was indeed the case – m-
chlorophenyl ether 19 was 70-fold more potent than 18, and further gains were realized when we
replaced the ether oxygen with a methylene (20). Gratifyingly, compound 20 demonstrated
cellular activity against Gram-negative bacteria, albeit with an MIC of 32 µg/mL against the
efflux-deficient E. coli ΔtolC mutant strain. While our efforts to reinforce a favorable
arrangement of the two aromatic groups through the installation of branching methyl groups
were only modestly successful, cyclic linkers proved to be appropriate biasing elements that also
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment
1
2

Page 13 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
better filled the surrounding pocket. One such compound, racemic trans-cyclopentane 21,
displayed a several-fold improvement in biochemical activity and cellular potency with an MIC
of 8 µg/mL against E. coli ΔtolC. From here, we investigated several cyclopentane replacements
to correct our tendency toward more lipophilic analogs and to explore new interactions with the
enzyme. Exchange of the distal stereogenic methine of 21 for nitrogen gave racemic pyrrolidine
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
2, which was clearly less active. Lactams 23 and 24 were efficient inhibitors (LipE two units
more favorable than the corresponding cyclopentane) that showed promising biochemical
activity, but these analogs did not address the large shift in potency between E. coli and P.
aeruginosa PPAT that was generally observed for this scaffold. Further, our reliance on
structural information from E. coli PPAT caused us to produce compounds with improved
potency against E. coli that tended to lose P. aeruginosa activity. The two enzymes are 76%
identical within 8 Å of the binding pocket with 58% overall sequence identity, and the
divergence in activity between E. coli and P. aeruginosa was rationalized by the presence of
several residue differences in the vicinity of the inhibitor, namely M74L, V135I, and H138L (E.
coli numbering used throughout this discussion). The alterations at positions 74 and 138 may be
responsible for the shift in potency through steric clashing with the appended arene and the
linker chain of the inhibitor, respectively. By removing the linker substituents and modifying the
nature of the aromatic ring or removing it completely, we identified compounds with more
balanced (benzisoxazole 25) or completely inverted (benzimidazole 26 and acetamide 27)
selectivity. More detailed studies were performed early in the program in an effort to better
understand the binding of peptidic inhibitor 1 to E. coli PPAT and rationalize its high specificity
2
8
for E. coli PPAT over P. aeruginosa PPAT. Here, seven reciprocal residue swaps between E.
coli and P. aeruginosa PPAT identified that the residue at position 74 was one of three residues
ACS Paragon Plus Environment
1
3

Journal of Medicinal Chemistry
Page 14 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
most heavily involved in determining susceptibility to 1 – the single M74L point mutation
resulted in a 25-fold loss in potency against E. coli PPAT. Interestingly, single residue
alterations at positions 135 and 138 in E. coli PPAT conferred greater susceptibility to 1.
A complimentary effort to target interactions with conserved, polar residues in the nearby
phosphate binding region of the active site led to the preparation of m-carboxylate 28, which was
equipotent to m-chloro analog 21 against E. coli PPAT but suffered a 16-fold loss in P.
aeruginosa activity (Table 1). The E. coli X-ray co-crystal structure confirmed that an
interaction with the R88 guanidine was achieved and encouraged further exploration (Figure 6).
The phosphate binding region of the enzyme accommodates the terminal phosphate group of the
natural substrate, 4’-phosphopantetheine, as well as the phosphate ester linkage of the enzymatic
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
reaction product, dephospho-CoA.
Pantetheine substrate analogs such as N-
pentylpantothenamide have been reported as CoA pathway inhibitors with the ability to inhibit
2
9
the growth of certain bacterial strains. It has been proposed that N-pentylpantothenamide does
not inhibit the CoA biosynthetic enzymes, but instead acts as an alternative substrate, forming
the CoA analog ethyldethia-CoA, which may be responsible for the observed antibacterial
3
0
activity by inhibition of the CoA- and acetyl-CoA-utilizing enzymes. We confirmed that N-
pentylpantothenamide lacked activity against E. coli PPAT up to 500 µM but displayed cellular
potency with MICs of 16 µg/mL against E. coli ΔtolC and 64 µg/mL against wild-type E. coli.
In an effort to explore a similar potential substrate analog bearing our fragment core, we
synthesized hydroxybenzimidazole-modified pantetheine 29, which showed modest on-target
potency. The corresponding X-ray co-crystal structure with E. coli PPAT confirmed excellent
overlap with dephospho-CoA, including an interaction between the terminal hydroxyl and the
R88 guanidine (Figure 7). In contrast to N-pentylpantothenamide, compound 29 did not possess
ACS Paragon Plus Environment
1
4

Page 15 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
any cellular activity. Clearly, 29 is too weak of an inhibitor to achieve cellular potency by direct
inhibition of PPAT, and it is likely not accepted by the biosynthetic enzymes along the CoA
pathway that would convert it to a toxic CoA analog or phosphorylated intermediate. Ultimately,
the challenges of achieving broad spectrum activity and the relatively high lipophilicity of many
compounds in this series prompted us to focus on an alternative series.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Table 1. Biochemical data of 4-hydroxybenzimidazole compounds 5, 18–29.
OH
N
R1
N
H
1
a
Compound
R
E. coli PPAT P. aeruginosa PPAT LE
cLogD7.4
IC (µM)
IC (µM)
5
0
50
5
CF
3
230
116
>500
108
0.36
0.45
2.1
18
0.37
OH
O
19
1.7
14
0.42
0.48
0.45
3.5
4.1
5.0
Cl
Cl
2
2
0
1
0.25
0.051
4.6
1.6
H
H
Cl
Cl
2
2
0.91
34
0.38
3.7
N
ACS Paragon Plus Environment
1
5

Journal of Medicinal Chemistry
Page 16 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
O
H
2
3
1.0
11
0.36
1.5
N
H
2
2
2
4
5
6
0.32
1.0
23
25
0.37
0.41
0.30
1.9
2.7
2.2
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
O
N
2.7
0.13
N
O
N
N
N
H
27
28
106
1.9
25
0.30
0.42
0.69
0.76
O
H
0.051
H
O
HO
2
9
30
114
0.28
0.21
HN
O
HO
OH
a = LE for E. coli PPAT data.
ACS Paragon Plus Environment
1
6

Page 17 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 6. X-ray crystal structure of E. coli PPAT with hydroxybenzimidazole analog 28.
Figure 7. Overlapping X-ray crystal structures of compound 29 (orange) and dephospho-
coenzyme A (light purple) bound to E. coli PPAT. Phosphorus atoms are shown in mustard.
In parallel with the hydroxybenzimidazole efforts, fragment 6 was selected for further
optimization. An X-ray co-structure of closely-related fragment 30 (Figure 8) revealed that this
triazolopyrimidinone bound in a manner similar to azabenzimidazole 4. In addition to the
ACS Paragon Plus Environment
1
7

Journal of Medicinal Chemistry
Page 18 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
common interactions with the conserved water molecule and M74, the exocyclic NH of 30
provided a new hydrogen bond with the backbone carbonyl of D72. Returning to fragment 6, we
hypothesized that the 6-methyl substituent of the bicyclic system and o-fluorine of the
benzylamine were not significantly contributing to the activity of this fragment; removal of both
of these groups while maintaining the key 5-methyl substituent on the triazolopyrimidinone led
to compound 31 and only a 2.5-fold drop in potency (Figure 9). As before, installation of a
chlorine atom at the meta-position of the phenyl ring (32) provided a 10-fold boost in activity.
Based on our observations from other fragments, substitution of the benzylic position of the
benzylamine was expected to be tolerated; however, the 30-fold improvement observed for R-
methyl analog 33 exceeded our expectations. We believe the stereogenic methyl serves to
facilitate de-solvation of the adjacent NH, provide lipophilic interactions, and favor the bound
conformation. The latter point was supported by a Cambridge Structural Database (CSD)
analysis of the dihedral angle of similar molecules, which showed enrichment of the 32° torsion
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
1
angle that is preferred for the binding of 33 (Figure 10). Further confirmation was obtained by
synthesis – the S-methyl analog of 31 was a weaker inhibitor of E. coli PPAT (IC = 120 μM)
5
0
than the parent methylene compound, while the R-methyl analog again led to improved activity
IC = 6.4 μM) (structures not shown).
(
5
0
H
N
N
N
NH
N
O
Br
30
PPAT IC50 = 5.6 µM
LE = 0.36, cLogD7.4 = 1.6
ACS Paragon Plus Environment
1
8

Page 19 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Figure 8. Inhibition data and X-ray crystal structure of compound 30 bound to E. coli PPAT.
H
N
H
N
H
N
N
N
N
N
N
N
NH
NH
NH
N
N
N
6
O
O
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
O
Cl
Cl
31
32
33
PPAT IC50 = 77 µM
PPAT IC50 = 7.6 µM
PPAT IC50 = 0.25 µM
LE = 0.30, cLogD7.4 = 0.62
LE = 0.35, cLogD7.4 = 1.4
LE = 0.43, cLogD7.4 = 2.2
Figure 9. Fragment hit optimization of E. coli PPAT for triazolopyrimidinones 6, 31–33.
Figure 10. CSD torsion angle population distribution for substituted α-methyl benzylamines.
With an optimized benzylamine moiety in place, we explored modifications to the core
heterocycle. The binding pocket surrounding the C5 methyl group was found to be restrictive, as
exemplified by the several-fold loss in potency for 34 and 35 (Table 2). Interestingly,
triazolopyrimidine 36 was almost 20-fold less potent than 33, even though 36 would be expected
to enjoy similar interactions with the enzyme. It is perhaps misleading to structurally represent
the parent inhibitors as triazolopyrimidinones when, in fact, they engage PPAT in the tautomeric
7
-hydroxytriazolopyrimidine form (or its ionized state) to satisfy the hydrogen bonding network.
In this form, N4 accepts a hydrogen bond from the nearby structural water, which is anchored by
interactions with the sidechain carbonyl of N106 and the NH of A75. As such, the hydrogen
ACS Paragon Plus Environment
1
9

Journal of Medicinal Chemistry
Page 20 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
bond acceptor strength of both N3 and N4 is expected to strongly influence overall binding
affinity. Calculation of the electrostatic potential (ESP) was performed on a set of prospective
core analogs, which were then ranked in order of increasingly negative electrostatic potential at
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
the N3 and N4 surface to evaluate potential hydrogen bonding interactions with the enzyme
32
(
Figure 11). From this set, triazolopyrimidine 36 was calculated to possess the least negative
electrostatic potential at the N3 and N4 surface, and so it was predicted to form weaker
interactions with PPAT than parent compound 33. This was borne out in the relatively weak
biochemical potency observed for 36. Our hypothesis was further validated when we found that
the introduction of an amine at C7 (37) recaptured most of the lost potency, which could be again
ablated by the incorporation of another ring nitrogen to provide the more electron-deficient
triazolotriazine 38. Among azolopyridines 39–41, the calculated electrostatic potential at the N3
and N4 surface increased along the series O < S < N, and 4-azabenzoxazole 39, 4-
azabenzothiazole 40 and 4-azabenzimidazole 41 were found to be progressively more active
against PPAT. From our limited analysis set, 41 was calculated to possess the most negative
electrostatic potential at N3 and N4 surface, and it also proved to be the most potent inhibitor.
One could reasonably assume that an amine at C7 would further enhance the electron density at
the N3 and N4 surface for these analogs; the results of our efforts along these lines can be found
26
in the accompanying manuscript on our lead optimization campaign.
Beyond the triazolopyrimidinone-inspired, evolutionary expansion of the core SAR, we also
evaluated several distinct core modifications. For example, installation of the core from
fragment 5 provided 4-hydroxybenzimidazole 42, which was 5–6-fold less active against E. coli
PPAT than both 33 and 41. Compound 42 also did not compare favorably to the most potent
hydroxybenzimidazoles (see Table 1), nor did it resolve the large shift in potency between E. coli
ACS Paragon Plus Environment
2
0

Page 21 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
and P. aeruginosa PPAT observed for many compounds from that series. Compared to the
parent hydroxybenzimidazole analogs, the additional hydrogen bond between the exocyclic NH
of 42 and the backbone carbonyl of D72 appears to result in a slight rotation of the inhibitor in
the binding pocket that may weaken the other interactions with the enzyme and lead to lower
potency. We also found that bolder core alterations were not well tolerated. This included
analogs such as pyridopyrazine 43 and “reverse” azabenzimidazole 44, both of which were
essentially inactive. These data support a concept that may be either underappreciated or masked
during multi-parameter optimization of cellularly-active inhibitors: stronger enzyme inhibition
can be achieved when the shape and the charge distribution of the protein is better matched by
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
3
the shape and the charge distribution of the ligand, presumably as a result of tighter binding.
Table 2. Biochemical data of triazolopyrimidinones and heterocyclic core replacements 33–44.
R2 NH
Cl
2
Compound
R
E. coli PPAT IC (µM) P. aeruginosa PPAT IC (µM)
50 50
H
N
33
0.25
0.90
1.1
0.97
N
N
N
O
H
N
3
3
4
5
2.8
N
N
N
O
N
H
N
1.7
N
N
O
ACS Paragon Plus Environment
2
1

Journal of Medicinal Chemistry
Page 22 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
N
N
N
N
3
6
4.7
2.6
N
N
N
N
37
0.41
5.9
0.38
NH2
N
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
N
N
3
8
4.6
N
N
NH2
N
N
O
3
4
4
9
0
1
2.4
4.9
N
N
N
S
0.41
0.19
2.3
N
0.61
N
H
OH
4
2
1.2
15
N
N
H
N
43
348
401
271
N
N
N
4
4
>500
N
HN
ACS Paragon Plus Environment
2
2

Page 23 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Figure 11. Electrostatic potential (ESP) maps of bicyclic core replacements. The color codes of
the ESP maps represent more negative (red) to more positive (violet) potentials from –50 to +50
kcal/mol. For each compound, the minimum atomic ESP value (in kcal/mol) on the molecular
surface for N3 and N4 were labelled as a reference.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
From the preceding fragment hit optimization, compounds 33 and 41 emerged as promising
leads. We selected triazolopyrimidinone 33 for further optimization primarily due to its greater
polarity (LogD7.4 1.1 vs. 3.3), which we anticipated would facilitate the discovery of advanced
inhibitors with the physicochemical properties expected to be required for wild-type Gram-
34
negative permeability. If this hypothesis turned out to be flawed, there would likely be some
SAR learnings that could then be applied to the 4-azabenzimidazole scaffold. The X-ray co-
crystal structure between 33 and E. coli PPAT suggested that there was an opportunity to explore
additional polar interactions with amino acid residues surrounding the stereogenic methyl.
Branching off of this methyl group with a number of polar substituents provided a series of
optically pure analogs (45–48) that were well tolerated; in the racemic series, both β-amino acid
4
9 and amide 50 showed an improvement in biochemical potency relative to the parent racemate,
rac-33 (Table 3). In the X-ray co-crystal structures of 46 and 49, a hydrogen bond was observed
between the newly installed polar sidechains and the NH of S39 (not shown). Similar to the
alcohol vs. ether matched pair (46 vs. 47), terminal methylation of acid 49 and amide 50 led to a
>10-fold loss in potency (51–52), perhaps due to size constraints in the vicinity of S39.
Fortunately, a breakthrough was identified in the form of nitrile analog 53, which engaged in the
desired interaction with S39 without increasing the hydrogen bond donor count for the inhibitor
(Figure 12). Not only did 53 display strong, equipotent activity against both E. coli and P.
aeruginosa PPAT with excellent selectivity against S. aureus PPAT (IC >500 μM), but it was
50
ACS Paragon Plus Environment
2
3

Journal of Medicinal Chemistry
Page 24 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
the first analog in the triazolopyrimidinone series to demonstrate cellular activity with an MIC of
1
6 µg/mL against E. coli ΔtolC (Figure 13). In addition, 53 also exhibited significantly
improved solubility relative to descyano analog 33 (>300 μg/mL vs. 68 μg/mL). To test our
earlier hypothesis regarding the relative merits of the more polar triazolopyrimidinone core and
the potential for translation of new SAR discoveries to other scaffolds, we synthesized a 4-
azabenzimidazole analog bearing the cyanomethyl benzylamine (54). Consistent with the
observations from Table 2, the E. coli and P. aeruginosa PPAT biochemical potency for lead
compounds 53 and 54 were comparable (54 was not tested against S. aureus PPAT); however, 54
showed improved cellular activity with an MIC of 4 µg/mL against E. coli ΔtolC. Neither
compound was active against wild-type E. coli.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Table 3. Biochemical data of triazolopyrimidinones 45–53.
H
N
N
R1
N
N
O
1
Compound
R
E. coli PPAT IC (µM) P. aeruginosa PPAT IC (µM)
50 50
H
N
3
3
0.25
0.55
0.24
2.4
0.97
0.66
2.0
Ar
Ar
Ar
Ar
H
N
45
OH
H
N
4
4
6
7
OH
H
N
11
O
O
48
H
N
0.25
3.9
O
NH2
Ar
ACS Paragon Plus Environment
2
4

Page 25 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
H
N
rac-33
0.76
0.20
0.45
2.5
1.4
Ar
Ar
Ar
Ar
Ar
Ar
H
N
49
1.1
OH
NH2
O
O
O
O
O
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
H
N
5
5
5
0
1
2
0.83
0.49
10
H
N
H
N
H
N
15
H
N
53
0.037
0.049
N
Ar = m-chlorophenyl.
Figure 12. X-ray crystal structure of compound 53 bound to E. coli PPAT.
ACS Paragon Plus Environment
2
5

Journal of Medicinal Chemistry
Page 26 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
H
N
N
N
N
N
H
H
N
N
N
N
N
N
H
O
Cl
Cl
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
53
PPAT Ec IC = 0.037 µM
PPAT Pa IC = 0.049 µM
54
PPAT Ec IC = 0.056 µM
(
(
)
50
(
)
50
PPAT Pa IC = 0.064 µM
)
50
( )
50
LE = 0.44,
= 6.89, LogD7.4 = 1.2
LE = 0.45,
= 3.88, LogD7.4 = 3.3
LipE
LipE
E. coli DtolC MIC = 16 µ
E. coli DtolC MIC = 4 µ
g/mL
g/mL
Figure 13. Comparison of lead compounds 53 and 54, including data for E. coli (Ec) and P.
aeruginosa (Pa) PPAT, as well as lipophilic efficiency (LipE = pIC – cLogP).
50
The ability of the triazolopyrimidinones to interact with the pantetheine site of E. coli PPAT
and inhibit enzymatic activity prompted us to briefly explore the mechanism of inhibition. We
expected the inhibitory activity of these compounds to be sensitive to an increase in the
concentration of 4’-phosphopantetheine employed in the biochemical assay, and this was indeed
the case. With a 10-fold increase in 4’-phosphopantetheine, a shift in potency was observed for
all 15 triazolopyrimidinone analogs that were evaluated; the data for compounds disclosed here
are shown in Table 4. While not a rigorous kinetic analysis of enzyme inhibition, these data
support the characterization of the triazolopyrimidinones as inhibitors competitive with 4’-
phosphopantetheine.
Table 4. E. coli PPAT biochemical data with 1x and 10x 4’-phosphopantetheine.
Compound E. coli PPAT IC (µM) E. coli 10x PhP PPAT IC (µM) Fold shift
5
0
50
3
3
2
3
7.6
43
5.7
3.7
4.2
0.25
0.76
0.92
3.2
rac-33
ACS Paragon Plus Environment
2
6

Page 27 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
4
6
0.24
a PhP = 4’-phosphopantetheine
As mentioned in the introduction, bacterial PPAT shares no significant sequence similarity to
0.44
1.8
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
the PPAT domain of human CoaSY. Nonetheless, we developed a complimentary biochemical
assay capable of measuring the PPAT activity of CoaSY to assess the selectivity of our
inhibitors. Of the hit-to-lead compounds presented here, only hydroxybenzimidazole 42
displayed measureable inhibition of the human enzyme (IC = 11 μM), though it was almost 9-
5
0
fold selective for the E. coli enzyme. All other analogs were either >200-fold selective or
inactive over the compound concentration range tested in our assay; advanced inhibitor 53 did
35
not show inhibition up to 500 μM.
Although we lacked a direct biophysical measure of
compound binding to CoaSY, the strong correlation observed between the E. coli biochemical
2
IC and the SPR K across a potency range of 0.00934-1.54 μM (R = 0.878, n = 64) suggested
5
0
D
that, by extension, any binding to CoaSY should be weak. The SPR data for compounds
disclosed here are shown below (Table 5). These data provided confidence that further
optimization would lead to analogs capable of achieving selective bacterial growth inhibition due
to specific activity on bacterial PPAT.
Table 5. E. coli PPAT SPR and biochemical data.
5
-1
-1
-1
Compound k (x 10 M sec ) k (s )
K (µM) E. coli PPAT IC (µM)
D 50
a
d
2
3
5
1
3
3
1.96
0.0071 0.038
0.0028 0.17
0.051
0.25
0.16
23
0.021
0.013
0.0090
0.010
0.037
0.056
54
13
ACS Paragon Plus Environment
2
7

Journal of Medicinal Chemistry
Page 28 of 52
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
CHEMISTRY
Synthesis of the N1-substituted azabenzimidazole analogs 16 and 17 began with nitropyridine
55 (Scheme 1). Displacement of fluoride with ammonia, followed by N-formylation and nitro
3
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
reduction provided 56, which underwent tandem reductive amination/cyclization with
benzaldehyde or 3-chlorobenzaldehyde to furnish 16 and 17, respectively.
Scheme 1. Synthesis of azabenzimidazoles 16 and 17.
N
N
H
N
F
N
N
H
a-c
d
N
O
NO2
NH2
R3
55
56
1
1
6 R = H
7 R = Cl
(a) ammonium hydroxide, MW, 130 °C, 30 min, 93%. (b) Ac O, HCOOH, 60 °C, overnight,
2
1
00%. (c) H , Raney-Nickel, THF, MeOH, rt, overnight, 86%. (d) benzaldehyde or 3-
2
chlorobenzaldehyde, BH ·Py, 2 h; 16: 28% and 17: 34%.
3
Compounds 19–27 from the 4-hydroxybenzimidazole series were uniformly prepared by
condensation of 2,3-diaminophenol 57 with the requisite carboxylic acid (Scheme 2). The
carboxylic acid required for a synthesis of 21 was prepared as shown in Scheme 3. Enolization
and triflate formation from β-keto ester 58 gave 59, which underwent smooth Suzuki-Miyaura
coupling with 3-chlorophenylboronic acid to give 60. Hydrogenation of the tetrasubstituted
olefin was achieved with Adams’ catalyst to give cis-1,2-disubstituted cyclopentane 61, along
with some reduction of the chloro group. cis-Cyclopentane 61 was then equilibrated under basic
conditions to trans-cyclopentane 62, which was directly condensed with 57 to give 21.
Scheme 2. Synthesis of 4-hydroxybenzimidazole analogs 19–20 and 22–27.
ACS Paragon Plus Environment
2
8

Page 29 of 52
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
O
OH
OH
HO
R1
NH2
N
R1
neat, 150 °C
N
NH2
H
-
5
7
19 20, 22-27
O
O
H
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
O
O
O
9
HO
HO
HO
HO
O
Cl
N
N
H
Cl
Cl
1
20
22
23
overnight, 3%
1 h, 35%
overnight, 7%
overnight, 13%
O
O
O
O
H
HO
HO
N
HO
HO
N
N
O
O
N
O
N
24
25
130 °C, 18 h, 5%
26
27
overnight, 7%
130 °C, 18 h, 3%
130 °C, 18 h, 14%
Scheme 3. Synthesis of trans-cyclopentane 21.
Cl
O
OTf
a
c
b
H CO C
H CO C
3
2
3
2
H CO C
3
2
58
59
60
OH
R3
R3
N
H
d
e
N
H
H CO C
H CO C
3
2
3
2
H
H
H
H
H
Cl
61
62
21
3
3
3
3
R = Cl, R = H
R = Cl, R = H
(3:1)
(3:1)
(
a) NaH, Tf O, Et O, 0 °C, 1 h, 97% (b) Pd(OAc) , PPh , Na CO , EtOH-benzene (1:3), 140
2 2 2 3 2 3
°C, 30 min, MW, 86% (c) PtO , H (1 atm), EtOH, 1.5 h, 72% (d) NaOMe, MeOH, reflux, 1 h.
e) 57, neat, 150 °C, overnight, 18%.
2
2
(
Compound 28 was prepared in a similar manner, with vinyl triflate 59 undergoing Suzuki-
Miyaura coupling with 3-trimethylsilylphenlyboronic acid to give 63. Hydrogenation with
Adams’ catalyst yielded cis-cyclopentane 64, which after iodo-desilylation with ICl provided
ACS Paragon Plus Environment
2
9