Targeting Bacillosamine Biosynthesis in Bacterial Pathogens: Development of Inhibitors to a Bacterial Amino-Sugar Acetyltransferase from Campylobacter jejuni

Article abstract of DOI:10.1021/acs.jmedchem.6b01869

The glycoproteins of selected microbial pathogens often include highly modified carbohydrates such as 2,4-diacetamidobacillosamine (diNAcBac). These glycoconjugates are involved in host-cell interactions and may be associated with the virulence of medically significant Gram-negative bacteria. In light of genetic studies demonstrating the attenuated virulence of bacterial strains in which modified carbohydrate biosynthesis enzymes have been knocked out, we are developing small molecule inhibitors of selected enzymes as tools to evaluate whether such compounds modulate virulence. We performed fragment-based and high-throughput screens against an amino-sugar acetyltransferase enzyme, PglD, involved in biosynthesis of UDP-diNAcBac in Campylobacter jejuni. Herein we report optimization of the hits into potent small molecule inhibitors (IC₅₀ < 300 nM). Biophysical characterization shows that the best inhibitors are competitive with acetyl coenzyme A and an X-ray cocrystal structure reveals that binding is biased toward occupation of the adenine subpocket of the AcCoA binding site by an aromatic ring.

Full text of DOI:10.1021/acs.jmedchem.6b01869

Subscriber access provided by University of Newcastle, Australia
Article
Targeting Bacillosamine Biosynthesis in Bacterial
Pathogens: Development of Inhibitors to a Bacterial
Amino-Sugar Acetyltransferase from Campylobacter jejuni
Joris W. De Schutter, James P Morrison, Michael James Morrison, Alessio Ciulli, and Barbara Imperiali
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.6b01869 • Publication Date (Web): 09 Feb 2017
Downloaded from http://pubs.acs.org on February 12, 2017
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 free 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 accessible to all
readers and 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 82
Journal of Medicinal Chemistry
1
2
Title
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
Targeting Bacillosamine Biosynthesis in Bacterial Pathogens: Development of Inhibitors to a
‡
Bacterial Amino-Sugar Acetyltransferase from Campylobacter jejuni
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
Authors
1
1
1
2
Joris W. De Schutter, James P. Morrison, Michael J. Morrison, Alessio Ciulli, Barbara
1
,3
Imperiali*
1
Department of Chemistry, Massachusetts Institute of Technology, 77 Massachusetts Ave.,
Cambridge, MA 02139, USA
2
Division of Biological Chemistry and Drug Discovery, School of Life Sciences, University of
Dundee, Dundee, Scotland
3
Department of Biology, Massachusetts Institute of Technology, 77 Massachusetts Ave.,
Cambridge, MA 02139, USA
‡
This paper is dedicated to the memory of Austin L. Travis
*
To whom correspondence should be addressed
Abstract
The glycoproteins of selected microbial pathogens often include highly modified
carbohydrates such as 2,4-diacetamidobacillosamine (diNAcBac). These glycoconjugates are
involved in host cell interactions and may be associated with the virulence of medically-
significant Gram-negative bacteria. In light of genetic studies demonstrating the attenuated
virulence of bacterial strains in which modified carbohydrate biosynthesis enzymes have been
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 82
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
knocked out, we are developing small molecule inhibitors of selected enzymes as tools to evaluate
whether such compounds modulate virulence.
We performed fragment-based and high-throughput screens against an amino-sugar
acetyltransferase enzyme, PglD, involved in biosynthesis of UDP-diNAcBac in C. jejuni. Herein
we report optimization of the hits into potent small molecule inhibitors (IC50 <300 nM).
Biophysical characterization shows that the best inhibitors are competitive with acetyl coenzyme
A and an X-ray co-crystal structure reveals that binding is biased towards occupation of the
adenine sub-pocket of the AcCoA binding site by an aromatic ring.
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
ACS Paragon Plus Environment

Page 3 of 82
Journal of Medicinal Chemistry
1
2
Introduction
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 clinically relevant antibiotics are targeted at essential bacterial survival functions
1
including replication, transcription, translation, and cell wall biosynthesis. This is an outcome of
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 vitro antibiotic screening and discovery, which has relied on observing cell death in culture in
laboratory settings. The emergence of resistance against all major mechanisms of antibiotic action
requires new paradigms for combating infectious disease. A potentially promising approach
includes the development of agents that target bacterial virulence in vivo in human hosts. Such
approaches may mitigate the effects of infectious disease, while potentially resulting in less
2
selective pressure for resistance development. Virulence factors are implicated in many bacterial
processes including host-cell adhesion, invasion, and colonization, as well as quorum sensing and
2-5
biofilm formation. In order to develop antivirulence agents, it is critical to identify validated
pathogen-specific processes that cause virulence in the targeted human hosts.
Protein glycosylation is widespread in nature and regulates a variety of cellular functions
6
including protein folding, cell-cell interactions, cell signaling, and the host immune response.
Glycans are attached to proteins via either serine/threonine (O-linked) or the amide nitrogen of
asparagine (N-linked). It is now recognized that selected bacteria possess the biosynthetic
machinery for O- and/or N-glycosylation and that this modification may play a role in
7
-11
pathogenicity.
N-glycosylation was first discovered in C. jejuni in 1999 and the protein
12,13
glycosylation (pgl) pathway has been characterized in detail for this organism (Figure 1).
In
C. jejuni, more than 60 proteins, including confirmed virulence factors, are modified with a
1
4
conserved heptasaccharide.
In a significant divergence between prokaryotes and eukaryotes, bacteria and archaea have
specialized enzymatic processes to modify the structures of selected carbohydrates for
incorporation into glycoconjugates. Furthermore, the discovery of unique prokaryote-specific
3
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 82
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
sugars is continuing with the pace of bacterial genome sequencing and bioanalytical methods
7
development. In contrast to the glycosyltransferase enzymes, which assemble complex glycans
and share common folds and mechanisms across domains of life, the specialized sugar-modifying
enzymes are attractive targets for developing targeted antivirulence agents because they tend not
to have mammalian homologs and because the associated glycoconjugates are linked with
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
1
5
16
bacterial pathogenicity. Of particular interest is di-N-acetylbacillosamine (diNAcBac), which
is derived from N-acetylglucosamine (GlcNAc). DiNAcBac is found, for example, at the reducing
end of O-linked glycans in selected strains of N. gonorrhoeae and A. baumannii, and N-linked
glycans in C. jejuni, highlighting the importance of diNAcBac biosynthesis pathways in these
1
7
Gram-negative pathogens. Intriguingly, while protein glycosylation pathways from these
pathogens glycosylate diverse proteins with different glycans, the reducing-end sugar is
1
8-20
diNAcBac.
The N-linked protein glycosylation (pgl) pathway of C. jejuni and steps leading to
diNAcBac biosynthesis are illustrated in Figure 1. The first two steps of UDP-diNAcBac
+
biosynthesis utilize an NAD -dependent dehydratase (PglF) followed by a pyridoxal phosphate-
dependent aminotransferase (PglE) to produce a UDP-4-amino-sugar, which is then acetylated by
1
8
PglD using acetyl coenzyme A (AcCoA) as a co-substrate (Figure 1 inset). Subsequent glycan
assembly onto an undecaprenyl-diphosphate carrier, is catalyzed by a series of glycosyl
transferases. After assembly, the completed heptasaccharide is translocated across the inner
membrane and the glycan is transferred to protein substrates in the bacterial periplasm by the
oligosaccharyl transferase PglB.
Studies have shown that disruption of genes responsible for diNAcBac biosynthesis (pglF,
21
pglE, pglD) perturb production of the native heptasaccharide glycosyl donor in C. jejuni. In
addition, ΔpglD and ΔpglE strains show greatly reduced colonization of the gastrointestinal tract
of 1-day-old chicks, thereby establishing a link between protein N-glycosylation and C. jejuni
4
ACS Paragon Plus Environment

Page 5 of 82
Journal of Medicinal Chemistry
1
2
2
2
pathogenicity in host cells. Further insight into these effects came from transposon mutagenesis
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
experiments, which identified pglF and pglE as essential genes for colonization. In mice,
23
mutation of pglE impaired invasion of intestinal epithelial cells and colonization of the gut. The
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
causative glycoconjugates underpinning these findings remain unknown, but several molecular
associations between N-glycosylation and C. jejuni virulence have been defined. For example,
VirB10, a structural component to the type IV secretion system (TFSS), needs to be glycosylated
2
4
at Asn97, otherwise a 10-fold decrease in natural competency results. Recently, 16 N-linked
glycoproteins were identified and found to be associated with C. jejuni outer membrane vesicles
2
5
(
OMVs) including the PEB3 adhesin. Pathogens deploy OMVs to deliver bacterial proteins into
host cells, making this an important finding in the relationship between periplasmic glycoproteins
26
and virulence. Protein O-glycosylation is also associated with virulence; for example, loss of
glycosylation of PilE, a constituent of the type IV pilin in N. gonorrhoeae, leads to a significant
2
7
decrease in epithelial cell adhesion.
Herein we report the development of inhibitors targeting the C. jejuni PglD, a UDP-amino-
sugar acetyltransferase, which catalyzes the third step in diNAcBac biosynthesis. PglD represents
an attractive target for inhibitor development as it is well understood from a structural and
2
8,29
mechanistic perspective.
Additionally, PglD is a soluble, well-expressed enzyme, which
makes it tractable for structure/activity-driven inhibitor discovery. Crystallographic analysis of
PglD reveals a homotrimeric structure with three equivalent active sites formed at the interfaces
2
8,29
between adjacent protomers.
Like many other bacterial UDP-amino-sugar acetyltransferases,
PglD is a member of the left-handed β-helix family comprising two domains (Figure 2a); a N-
terminal domain with a β−α−β−α−β−α Rossmann fold motif that binds the UDP-4-amino-sugar,
and a C-terminal hexapeptide repeat motif that defines the left-handed β-helix that contributes to
the AcCoA binding site. PglD has been co-crystallized in the presence of the AcCoA and UDP-4-
5
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 82
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
2
8
amino-sugar substrates (Figure 2b). The structural features of PglD and related homologs
differentiate these prokaryotic AcCoA-dependent acetyltransferases from their mammalian
counterparts, including for example HAT1 of the GCN5-related N-acetyltransferase (GNAT)
3
0
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
superfamily.
Currently, many questions remain as to why modified carbohydrates, such as diNAcBac,
are incorporated into bacterial glycoconjugates. Therefore, chemical tools such as small molecule
inhibitors that selectively block diNAcBac biosynthesis would be valuable agents for
understanding the significance of glycan diversification in selected bacteria. In addition, such
tools could help to validate the pgl pathway enzymes as potential antivirulence targets.
Consequently, we have established a program to develop inhibitors of diNAcBac biosynthesis
with a focus on the well-characterized C. jejuni pathway. This Gram-negative pathogen is a
common cause of gastroenteritis in humans and may also result in the development of Guillain-
Barré syndrome, an auto-immune disease in which the peripheral nerves are attacked resulting in
3
1,32
damage to the myelin insulation.
In recent years, C. jejuni has shown increased resistance
towards front-line antibiotics including the macrolides and fluoroquinolones, which inhibit
3
3
protein synthesis and DNA unwinding, respectively.
6
ACS Paragon Plus Environment

Page 7 of 82
Journal of Medicinal Chemistry
1
2
Figure 1: Schematic representation of the C. jejuni N-linked protein glycosylation pathway.
Enzymes are shown in italics with the oligosaccharyl transferase PglB shown as determined in
PDB 3RCE. Also shown is an N-linked glycosylation substrate PEB3 (PDB: 2HXW), which is a
virulence factor in periplasm that is modified by N-linked glycosylation. Inset highlights the three
sugar-modifying enzymes that convert UDP-GlcNAc to UDP-diNAcBac.
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
Results
Screening
Initial hits for the development of inhibitors for PglD were identified by employing
3
4
fragment-based screening and high-throughput molecular library approaches. The fragment-
based screen was performed with a fragment set that included a selection of the Maybridge Ro3
3
5
chemical library as described previously. The primary screen involved differential scanning
fluorimetry (DSF) analysis against PglD to identify fragments that bind to and stabilize the
natively-folded protein. Following this, hits were confirmed by NMR spectroscopy
(
WaterLOGSY and STD) and further validated using a UV-based biochemical assay, which
3
6
measures CoASH release following acetyl transfer to the UDP-4-amino-sugar. Using this
workflow, 10 compound hits with IC50 values in the 1-10 mM range and ligand efficiencies in the
0.29 – 0.32 range were validated. For example, compound 1 (Figure 2), which was analyzed by
X-ray in complex with PglD, binds to the central AcCoA binding groove occupying the
pantetheine site proximal to active-site residues involved in amino-sugar activation and thioester
attack (Figure 2c). Unfortunately, efforts to build upon this, and other validated fragments yielded
only modest improvements in affinity. Coincident with these studies a high-throughput screen
was performed in collaboration with the Broad Institute, using a Diversity-Orientated Synthesis
(DOS) compound collection (83,000) and the NIH Molecular Libraries Probe Production Centers
7
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 82
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
3
7
Network (MLPCN) library (276,000) of small molecules. Compounds were screened at 10 µM
using a PglD end-point assay monitoring the generation of CoASH from the acetyltransferase
reaction. Despite the low hit rate (defined as >20% inhibition) of 0.02% and 0.12% respectively,
the screen led to the discovery of thienopyrimidinone-6-carboxylate compound 2 as a promising
hit for further development. For direct comparison with the validated fragment hits, 2 was
deconstructed to the core structure 3, which demonstrated a validated IC50 value of 860 µM and
LE of 0.31. In light of these results and the tractability of the core towards synthetic manipulation
at four points of diversification (C2, C4, C5, and C6; Scheme 1) this compound class became the
focus of our continuing medicinal chemistry efforts.
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
Preliminary structure activity studies to vary substitution at C2 generated analog 4 as the
best inhibitor with an IC50 value of 1 µM (LE 0.33). However, although this compound
represented a good improvement in potency, the candidate suffered from some disadvantages.
Specifically, 4 features a planar structure with poor solubility properties and a metabolic liability
due to the 1,3-unsubsituted indole and conjugated alkene systems. Furthermore, 4 and related
analogs showed poor long-term stability in solution. A significant effort was also made to
crystallize PglD with several of the thienopyrimidinone hits, however, this proved unsuccessful.
Therefore, we performed a docking study on fragment 3 to generate a computational model as a
guide for future synthetic efforts aimed at eliminating the compound liabilities while improving
38
potency. The thienopyrimidone-6-carboxylate core was predicted to occupy a binding site
overlapping with the panthetheine moiety of AcCoA, as had been previously observed in the co-
crystal structure of fragment 1 (PDB 5TYH), with similar placement of the carboxylate moiety
and close agreement between the positions of the furan oxygen of 1 and exocyclic oxygen of 2
(compare Figure 2c and d). Given these observations, we exploited the computational model as a
guide for the design of the next generation of analogs, which recognized the opportunity to extend
8
ACS Paragon Plus Environment

Page 9 of 82
Journal of Medicinal Chemistry
1
2
the molecule both at the C2 position to interact with a hydrophobic region on the β-helix, and at
the C4 position, by transitioning to the thienopyrimidine core, to further occupy the AcCoA
binding groove (Scheme 1).
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 2: Structure analysis of C. jejuni PglD. (a) Crystal structure of the PglD homotrimer (PDB
3BSY) in cartoon representation; individual protomers are colored in green, blue and orange. (b)
Expanded of view of the PglD active site protein surface with the AcCoA substrate shown in stick
representation (magenta, PDB 3BSY) as well as the UDP-4-amino-sugar substrate (yellow, PDB
3BSS). (c) Crystal structure of fragment 1 (magenta) bound to PglD with a semi-transparent
9
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 82
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
surface and selected side-chain residues shown in stick representation and annotated (PDB
3
8
5
TYH). (d) Docking output pose generated for compound 3 (Glide, Maestro).
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
Scheme 1: General structure and numbering system of the thienopyrimidine core.
In order to reach cytoplasmic targets in Gram-negative pathogens, compounds may enter
cells through passive diffusion or pass through the bacterial porins to reach the periplasm and
3
9,40
then reach the cytoplasm via passive or active mechanisms.
With these parameters in mind,
our overall goal was to develop potent (IC50 <500 nM) inhibitors of C. jejuni PglD based on the
thienopyrimidine-6-carboxylate core with good ligand efficiency (LE >0.30) together with
4
1
balanced physicochemical properties including cLogP values between 2 – 4.
Chemistry
We developed modular synthetic routes to provide access to the target thienopyrimidine
carboxylic acids with readily-modifiable substitution at the C2 and C4 positions (Schemes 2 and
3
).
For modification at the C2-position (Scheme 2, Route I), the synthesis was initiated by
condensation of commercially available thiophene, 5, with nitriles bearing the desired R
2
substituent, under strongly acidic conditions. In most cases, the thienopyrimidone products of
general structure 6, proved highly insoluble, and therefore chromatography and NMR
characterization was carried out after the subsequent step. In cases where nitriles with the
appropriate R
2
moiety were inaccessible, an alternative approach (Route II) was employed. This
route is somewhat longer and includes a Michaelis-Arbuzov reaction with triethyl phosphite and
1
0
ACS Paragon Plus Environment

Page 11 of 82
Journal of Medicinal Chemistry
1
2
2 2
analog 6 (Core structure: R = -CH Cl) to furnish phosphonate 8. Subsequent Horner-Wadsworth-
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
Emmons (HWE) olefination and hydrogenation generated thienopyrimidinones of general
structure 9 with a variety of phenylethyl substituents at the C2-position.
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
Installation of a substituent at the C4-position of thienopyrimidines is typically
accomplished by transforming the C4 oxygen of the thienopyrimidone into chloride or bromide,
with POCl
3 3 N
and PBr respectively, followed by a S Ar reaction, both performed at high
temperatures. To avoid use of these harsh conditions and strong lachrymators, we instead
employed a one-step, room temperature reaction modified from chemistry developed by
4
2
researchers at Wyeth.
Utilization of
a
sufficiently strong base, such as 1,8-
diazabicyclo[5.4.0]undec-7-ene (DBU), in excess solubilized the thienopyrimidinones in
acetonitrile and replacing the toxic and expensive BOP reagent for the readily available PyBOP,
allowed for robust formation of the desired pyrimidines of general structure 10 in good yields
(
40-90%). These thienopyrimidines could be isolated by column chromatography in excellent
purity (>95% homogeneity as determined by LC-MS). The final saponification of
thioenopyrimidine esters represented by 7 and 10 was achieved in a THF/MeOH/H O mixture
2
containing NaOH; occasionally mild heating was required to achieve complete conversion. The
final inhibitors with core scaffold 11 were isolated by trituration as easy to handle solids.
1
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 82
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
Scheme 2: Synthesis of inhibitors with thienopyrimidine-6-carboxylate core 11
Conditions: (a) 1.2 equiv. R
.3 equiv. PyBOP, 3 equiv. DBU, MeCN, RT (40-90%); (c) 1.2 equiv. ClCH
dioxane, 90°C (50-90%) (d) 20 equiv. P(OEt) , 145°C (85%); (e) 1.5 equiv. NaH, 2 equiv.
ArCHO, THF, RT (70-90%); (f) 10% Pd(OH) /C, THF/MeOH 1:1, 60°C (quant.); (g) 1.5 equiv.
NH , 1.3 equiv. PyBOP, 3 equiv. DBU, MeCN, RT (40-90%) (h) 3 equiv. NaOH,
THF/MeOH/H O 2:1:1, RT (50-80%).
2
CN, 4 M HCl in 1,4-dioxane, 90°C (50-90%); (b) 1.5 equiv. R
4 2
NH ,
1
2
CN, 4 M HCl in 1,4-
3
2
R
4
2
2
A related thienopyrimidine isomer with general structure 16, was accessed as described in
Scheme 3. In this case, the thiophene starting material 14 is not commercially available but could
1
2
ACS Paragon Plus Environment

Page 13 of 82
Journal of Medicinal Chemistry
1
2
be readily prepared by the esterification of dicarboxylic acid 12 followed by nitration and
hydrogenation. Condensation with nitriles provided thienopyrimidinones of general structure 15,
which differ from the scaffold 11 series in the relative location of the sulfur atom and the absence
of a methyl group on the thiophene ring. Installation of substituents at C4 and saponification were
performed as described previously, to provide inhibitors with core scaffold 16.
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
Scheme 3: Synthesis of inhibitors with thienopyrimidine-6-carboxylate core 16
Conditions: (a) 4 equiv. SOCl
Pd(OH) /C, THF/MeOH 1:1, RT (93% over 2 steps); (d) 1.2 equiv. R
dioxane, 90°C (60-80%); (e) 1.5 equiv. R NH , 1.3 equiv. PyBOP, 3 equiv. DBU, MeCN, RT (40-
0%); (f) 3 equiv. NaOH in THF/MeOH/H O 2:1:1, RT (50-80%).
2
, MeOH, 50°C (99%); (b) 2 equiv. HNO
3
, H
2 4
SO , RT; (c) 10%
2
2
CN, 4 M HCl in 1,4-
4
2
9
2
Initial SAR analysis
All target thienopyrimidine carboxylic acids 11 and 16 (Figure 3) were assayed for
inhibition of recombinantly-expressed and purified C. jejuni NCTC 11168 PglD
1
8
acetyltransferase. The continuous assay monitors the release of CoASH using 5,5'-dithiobis-(2-
1
3
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 82
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
3
6
nitrobenzoic acid, and was carried out at 300 µM AcCoA (K
amino-sugar (K of 274 µM). IC50 values of selected compounds are summarized in Table 1. The
AcCoA concentration was aligned with reported physiologic concentrations in typical Gram-
m
of 295 µM), and 500 µM UDP-4-
m
4
3
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
negative bacteria.
Our initial strategy was to explore the C2- and C4-substituents individually and then
combine the optimal structural components from each series to capture advantageous synergistic
binding effects. We observed that short aliphatic chains and saturated carbocycles at C2 (with R
4
=
Me) provided weak inhibitors with IC50 values in the 10 µM to 1 mM range (data not shown).
Based on these studies we selected an unsubstituted phenylethyl moiety (Figure 3, substituent d)
as a favorable C2-substituent in terms of potency gain versus added mass (e.g. 17, IC50 = 2.2 µM,
LE 0.35, Table 1). However, when we explored C4-substituents (with R
proved to be poor inhibitors (IC50 > 500 µM, data not shown) and SAR trends could not be well
defined from the data. Accordingly, we elected to directly prepare disubstituted analogs, with R
phenylethyl (d), and iteratively build up the R substituent in terms of length, sterics and
2
= H), these analogs
2
=
4
heteroatom substitution (e.g. inhibitors 18-35; shorter and smaller substituents omitted for
brevity). In general, only modest gains in potency could be achieved, and all the gains were at the
cost of poorer ligand efficiencies. Ultimately, we observed that a phenylethyl moiety was also
well tolerated at the C4-position of the thienopyrimidine core (20, R
.4 µM).
In order to gain a better understanding of the binding behaviour of the inhibitors, we
examined the results of the phenyl scan on the R substituent, which positioned moieties with
2 4
= R = phenylethyl, IC50 =
1
4
different characteristics (sterics, polarity and hydrogen-bonding capabilities) ortho, meta or para
on the aryl ring (inhibitors 21-35). However, while we had predicted that these substitutions
would either complement or clash with the protein surface resulting in a measurable difference in
1
4
ACS Paragon Plus Environment

Page 15 of 82
Journal of Medicinal Chemistry
1
2
IC50 values, we instead observed a “flat" SAR with inhibitors exhibiting IC50 values around 0.5
µM (e.g. 21-35). Indeed, only substitution with an acetamido group (33-35) elicited clear
variances, with the ortho substitution being the most disfavored (35, IC50 = 9.0 µM). A potential
explanation for the flat SAR could be that if the aryl group occupies the AcCoA binding groove
then one edge would face the protein surface and the other would be solvent exposed, thus
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
precluding any interaction between the moieties on the R
4
-aryl ring and the protein surface. This
hypothesis was tested by preparing analogs bearing two moieties on the aryl ring (inhibitors 36-
3
8), however, potency remained at 0.5 µM.
In light of these data, we developed two hypotheses. One potential hypothesis was that the
compounds might have a different mode of action other than competitive inhibition with the
AcCoA substrate. We explored this by applying several biochemical and biophysical approaches,
which are presented in the next section (vide infra). The second possibility was that the inhibitor
binding mode differed significantly from that predicted by the computational model; for example,
either the ligands bind in a different location on the PglD protein surface or the enzyme itself
undergoes a significant conformational change. We also noted that the C2,C4-disubstituted
inhibitors were inherently quite symmetrical, with a pseudo C
thienopyrimidine core (Figure 4a). This feature might explain the flat SAR observed for the
phenyl scan of the R -substituent since the effect of any moiety on the aryl group could
potentially be masked if the ligand flipped 180° and positioned the R -substituent in the AcCoA
2
axis of rotation in the plane of the
4
2
binding groove instead. To test this hypothesis, we synthesized inhibitors 49 and 50, which
included the quasi-regioisomeric thienopyrimidine core 16 and, if the binding orientation remains
true, should position the sulfur atom towards the protein surface. Indeed, these inhibitors had
similar potency to their parent analogs (compare 49 IC50 = 3.9 µM vs. 17 IC50 = 2.2 µM, and 50
IC50 = 1.4 µM vs. 41 IC50 = 1.1 µM), suggesting the thienopyrimidine carboxylic core can bind in
1
5
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 82
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
either of two orientations into the AcCoA binding groove. To further validate this hypothesis, we
developed inhibitors that would be locked in a fixed conformation by introducing rigidifying
elements in the alkyl linkers between the thienopyrimidine core and substituent aryl groups.
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
Synthetically it was straightforward to add a methyl group on the second carbon of the R
2
-
substituent and a methoxy group on the second carbon of the R -substituent; the prerequisite
4
starting materials were commercially available as both stereoisomers and could be easily
transformed into the corresponding nitrile or amine for the modular syntheses (Figure 4b). With
the four rigidified analogs we would expect to see data consistent with a matched/mismatched
phenomenon; either substituent should have an enantiomer that complements the protein surface
better than the other and when combined these effects should result in one poor inhibitor, two
moderate ones and a single preferred analog. Indeed, this was observed with compounds 51-54
(Figure 4b), albeit in a modest IC50 range. These combined results strongly suggested that the
inhibitors were capable of binding to the PglD active site with either the R
the AcCoA groove.
2 4
-, or R -substituent in
Overall, the SAR data (Figure 4) shows that while binding may be somewhat promiscuous,
2
-substituent was far more important for binding potency. This experimental observation was
the R
therefore inconsistent with our initial computational model, which had oriented this substituent
over the β-helix (and quite solvent exposed) and placed the R -substituent in the AcCoA groove.
4
These findings prompted us to investigate the binding mode in more detail, to provide a
foundation for the next synthetic and biochemical efforts.
1
6
ACS Paragon Plus Environment

Page 17 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3: Representative examples of thienopyrimidine-6-carboxylate inhibitors of C. jejuni
2 4 2 4
PglD; Core 11: R = a – z; R = a – z; Core 16: R = d, h; R = a.
1
7
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 82
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
Compound # Scaffold
R
2
R
4
IC50/(µM)
2.2 ± 0.4
1.6
LE ΔT
0.35
m
/(°C) cLogP
1
1
7
8
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
d
d
d
d
d
d
d
d
d
d
d
d
d
d
d
d
s
a
b
c
7.4
3.6
nd
4.44
1.48
5.89
6.54
6.46
6.46
6.46
6.68
6.68
6.68
6.99
6.99
6.99
5.04
5.04
5.04
3.46
3.46
3.46
6.42
6.55
6.50
0.32
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
19
4.5
0.26
2
0
d
g
h
i
1.4
0.27
0.27
0.28
0.28
0.29
0.29
0.29
0.28
0.29
0.30
0.30
0.30
0.30
0.29
0.28
0.26
0.27
0.26
0.26
13.2
nd
21
0.54
0.45
0.48
0.42
0.39
0.48
0.72
0.42
0.25
0.39
0.34
0.42
2.4
2
2
2
2
2
3
4
5
nd
13.7
nd
j
k
l
nd
26
13.2
nd
2
2
2
3
7
8
9
0
m
n
o
p
q
r
nd
14.3
12.0
12.0
12.5
nd
31
3
3
3
3
2
3
4
5
a
a
a
x
y
z
t
3.5
nd
u
d
d
d
9.0
nd
36
1.0
nd
3
7
0.37
0.59
nd
38
11.5
1
8
ACS Paragon Plus Environment

Page 19 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
39
11
11
11
11
11
11
11
11
11
11
16
16
11
11
11
11
11
e
f
a
a
a
a
a
b
r
4.4
4.1
0.31
0.30
0.33
0.33
7.9
7.4
8.4
8.9
8.4
5.3
11.5
9.4
7.4
7.4
4.8
5.9
9.4
10.4
nd
4.97
5.50
4.36
4.58
2.94
0.50
4.06
5.15
5.32
4.41
4.45
4.37
6.50
6.50
6.50
6.50
5.12
4
4
0
1
h
k
q
s
1.1 ± 0.2
2.3
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
42
4
3
0.46 ± 0.05 0.39
44
1.4
0.28
2.1
0.28
0.27
0.33
0.28
0.31
0.34
0.34
0.24
0.25
0.27
0.27
4
4
4
4
5
6
7
8
s
v
w
z
a
a
a
a
a
4.6
1.8
49
d
h
3.9
5
5
5
5
0
1
2
3
1.4
2PP(S) 2MPE(R)
2PP(S) 2MPE(S)
2PP(R) 2MPE(R)
2PP(R) 2MPE(S)
2.0
1.1
0.87
0.42
54
9.9
12.5
5
5
s
2MPE(S) 0.27 ± 0.09 0.26
Table 1: Experimental data obtained for selected inhibitors (Figures 3 and 4): IC50 values
determined with the continuous C. jejuni PglD inhibition assay, calculated ligand efficiency (LE),
thermal stabilization of the PglD/ligand complex with the DSF assay, and cLogP values
®
calculated with built-in functionality of the ChemDraw commercial software package.
Compounds 51-54 (Figure 4): 2MPE: 2-methoxy-2-phenylethyl and 2PP: 2-phenylpropyl.
1
9
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 82
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 4: (a) Illustration of the quasi C
inhibitors 51 – 54 with rigidified R and R
stereochemical preference of R and R substituents with respect to the PglD protein surface and
2
symmetry axis of the disubstituted inhibitors (b) SAR of
2
4
substituents; magenta symbols represent
2
4
highlight the mismatched/matched pattern.
Biophysical characterization and inhibitor optimization
With the knowledge gained from SAR studies up to this point, we generated a potent C.
jejuni PglD inhibitor, 55 (Figure 5), with an IC50 of 270 nM; however, this was at the cost of
decreased ligand efficiency (0.26) and a high cLogP (5.12). In order to further optimize the
potency and physicochemical properties of these compounds, we needed to better characterize
their binding characteristics. Towards this end, we employed several complementary approaches
2
0
ACS Paragon Plus Environment

Page 21 of 82
Journal of Medicinal Chemistry
1
2
including the PglD inhibition assay with varying substrate concentrations, DSF analysis, dynamic
light scattering (DLS) and X-ray crystallography.
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
We had first initiated inhibitor development based on modeling studies, which suggested
that the thienopyrimidine carboxylate core competed with the AcCoA substrate for binding to the
PglD active site. To test this experimentally, we monitored inhibition in the presence of varying
concentrations of the two substrates. We observed that significantly higher concentrations of the
UDP-4-amino-sugar substrate had no measurable effect on the inhibition by a representative
analog of 11 (compound 37, supporting information Table S2). In contrast, a 2-, and 5-fold
increase in AcCoA substrate concentration resulted in measured IC50 values of 0.63 µM and 1.4
µM, respectively. These values represent a 2- and 4-fold decrease in potency and strongly suggest
that the thienopyrimidine carboxylate inhibitors compete with AcCoA for binding to the PglD
active site.
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
All of the data thus far has been based on the enzymatic inhibition assay and, despite the
4
4
judicious use of detergent and BSA to prevent aggregation phenomena, may be prone to
artefacts that obscure binding trends. We therefore employed DSF as a qualitative secondary
assay to assess binding of inhibitors to the C. jejuni PglD. DSF employs a fluorescent dye, Sypro
4
5
Orange, to report on the thermal denaturation of a protein or protein/ligand complex. Potent
inhibitors bind to their target enzyme more strongly that weak ones, resulting in a more stable
protein/ligand complex and higher measured melting temperature (T
stabilization (ΔT ) of PglD can be used to provide an indirect measure of the relative binding
affinity of the related family of compounds. We selected a subset of inhibitors with a range of
m
). Therefore, the thermal
m
m
IC50 values and structural characteristics to analyze by DSF and plotted the ΔT vs IC50 values
2
(Figure 6). The graph shows a positive correlation (R = 0.72) between inhibition potency and
thermal stabilization over a range of IC50 values indicating that the compounds stabilize the native
2
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 82
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
PglD homotrimer and that the more potent compounds, based on IC50 values, indeed stabilize
PglD more effectively.
Next, we analyzed the enzyme quaternary structure in solution by dynamic light scattering
(DLS) and found the calculated hydrodynamic radius of apo PglD to be ~34 Å, consistent with the
crystal structures of trimeric PglD (supporting information Figure S1). In the presence of 10-fold
excess of inhibitor 55, the measured radius was marginally larger, ~36 Å, yet still consistent with
the known trimeric structure. We also noted that the polydispersity index (PDI) of the PglD/55
complex (24 % PDI) was lower than apo PglD (31% PDI), providing strong support that the
inhibitor does not destabilize quaternary structure and indeed binds to the homotrimeric enzyme.
Throughout these efforts we also continued to pursue crystallization attempts to gain
structural insight on representative thienopyrimidine analogs. X-ray analysis of the C. jejuni
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
8,29
NCTC 11168 PglD enzyme has been successfully carried out in our group and others.
Typically, the reported precipitant solutions were either high ionic strength (1.9 M ammonium
sulfate) or contained acidic buffer components that were frequently observed to bind in the
2
8,29
enzyme active site (sulfate, phosphate, citrate).
Despite numerous attempts, application of
previously published protocols did not yield co-crystal structures with the current
thienopyrimidine-6-carboxylate inhibitors and thus we screened for new conditions that would be
more compatible with the binding of small organic molecules. Ultimately, a crystallization hit
with PEG-3350 was successfully optimized to reliably produce high quality crystals. The crystals
develop as hollow, hexagonal rods that fill in from one end as they mature; producing either
“
vase-like” or solid crystals (Figure S3). The diffraction quality did not appear to differ between a
solid or hollow end of such crystals. The crystals were tolerant of DMSO in concentrations up to
10%, but tended to be rather fragile and could break upon looping. Therefore, we employed an
2
2
ACS Paragon Plus Environment

Page 23 of 82
Journal of Medicinal Chemistry
1
2
“
in-drop soaking” protocol, in which a solution of the inhibitor is directly added to the drop
containing the crystals.
Using these crystallization conditions, we were able to obtain high-quality diffraction data
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
and successfully solve the co-crystal structure of the PglD/17 complex (Figure 7 and S4). From
the structure, two significant observations were immediately evident. Firstly, as anticipated from
the SAR studies, the thienopyrimidine core was flipped 180° compared to the computational
model developed with the core fragment, with the R
Figure 7, panel b). This position of the thienopyrimidine core directs the R
from the AcCoA groove and into solvent where it makes no significant interactions with the PglD
2
-subsituent binding in the AcCoA groove
(
4
-subsituent away
4
6
protein surface (Figure S5, LigPlot ). This observation is consistent with our SAR findings that
R
4
-subsituent makes small contributions to inhibitor potency at the expense of ligand efficiency
(Table 1). Secondly, the most surprising observation was that the ligand was bound 3.9 Å further
along the AcCoA groove, relocating the carboxylate distal to the acetyltransferase active site
where the AcCoA thioester is normally accommodated. This observation was rather unexpected as
we assumed that the Asn118, H134 and S136 residues would be preferred hydrogen-bonding
partners for the inhibitor carboxylic acid moiety. Furthermore, this displacement precludes a T-
shaped π-stacking between the side chain of Phe152 and the thienopyrimidine core; instead the
binding mode positions Phe152 at an offset angle over the inhibitor carboxylate (Figure S4). To
compensate, inhibitor 17 makes several other key interactions with the PglD surface; the
carboxylic acid moiety is positioned at a favorable distance and orientation to form hydrogen
bonds with the Ser136 side chain and the backbone amide NH of Ile155. Similarly, the
thienopyrimidine N1 atom is well placed to interact with Gly173. From the structure it appears
that a definitive interaction may be the placement of the R
2
phenyl ring within the AcCoA adenine
binding sub-pocket, displacing solvent waters from a relatively hydrophobic protein cavity. The
2
3
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 82
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
ligand binding mode is also internally favorable (low torsion) as the bound conformation adopts
an anti-periplanar angle of approximately 160° along the benzylidene bond between the
thienopyridimine core and R
Armed with this new structural information we set out to further optimize the inhibitor
series. First we investigated the possibility of simultaneously placing the R phenyl ring in the
adenine sub-pocket and the carboxylate in the active site hot spot by increasing the linker length
between the inhibitor core and R -substituent. This did not prove successful as analogs 39 and 40
with a 3- and 4-carbon linker respectively) demonstrated lower in vitro potency; presumably due
to internal strain of unfavorable torsion angles of the alkyl linker. Secondly we noted that
although the R phenyl group fits well in the adenine sub-pocket, it does not fully occupy the
2
phenyl ring (Figure 7c).
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
(
2
volume of the [6,5]-fused ring cavity and there was unoccupied space at the meta position relative
to the linker (Figure 7c). We explored several moieties at this position (41 –OMe, 42 –F, 46 –Cl
and 47 –CF
3
, Table 1), but only gained a minor improvement in potency with compound 41 (IC50
=
1.1 µM). Finally upon closer inspection of the adenine sub-pocket we speculated that if the
aromatic carbon at a position ortho to the linker is replaced by a nitrogen and protonated, it would
be well-oriented to form a hydrogen bond with the backbone oxygen of Gly173, as it does with
2
8
the exocyclic amine of the adenine of the AcCoA substrate (PDB ID 3BSY). While the resultant
pyridine nitrogen is only weakly basic, it could potentially be protonated in the microenvironment
of the enzyme and lead to an order of magnitude improvement in potency. For example, the
pyridine derivative risedronic acid is a ~300-fold more potent inhibitor of the human farnesyl
4
7
pyrophosphate synthase than the corresponding phenyl analog. Although not as significant, the
4
.8-fold fold enhancement we observed between inhibitor 43 (IC50 = 0.46 µM) and 17 (IC50 = 2.2
µM) is very interesting considering the minor nature of the structural change. Although 43 is not
the most potent inhibitor, it exhibits the best ligand efficiency, 0.39, and has a cLogP of 2.94,
2
4
ACS Paragon Plus Environment

Page 25 of 82
Journal of Medicinal Chemistry
1
2
within the -0.5 - 3 range that is thought to be ideal to reach cytoplasmic targets of Gram-negative
bacteria; with these characteristics 43 represents our best candidate to date with balanced
physicochemical properties.
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 5: Structures of key inhibitors discussed in this section: 17, 43 and 55.
m
Figure 6: Correlation between the in vitro potency against PglD and the T (calculated from the
first derivative of the thermal melting curve) of the PglD/inhibitor complexes with core 11;
thermal melt assays were performed with 10 µM PglD and 100 µM compound. Correlation
2
5
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 82
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
4
coefficient provided for all data indicated by red diamond. Inhibitors with R = b (ethylamine)
(
blue diamond, circled) appear as outliers and are omitted from the calculation of the correlation
coefficient.
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
6
ACS Paragon Plus Environment

Page 27 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
7
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 82
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
Figure 7: Co-crystal structure of the PglD/17 complex (PDB 5T2Y). (a) Inhibitor 17 shown in
sticks presentation (magenta) and overlayed with AcCoA substrate in lines (cyan, PDB 3BSY)
and sugar substrate in sticks (yellow, PDB 3BSS). (b) overlay with the computational docking
pose of compound 2 in line representation (orange). (c) Side-view highlighting the AcCoA
adenine sub-pocket and interaction between 17 and residues Gly173, Ile155 and Ser136. White
dashed lines indicate distances between 17 and PglD side-chain residues; yellow dashed line
shows distance between the carboxylate computational docking pose and crystal structure. (d)
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
Composite omit map 2F
o c
–F contoured at 1.0 s generated with the built in functionality of the
PHENIX software and image generated in PYMOL.
2
8
ACS Paragon Plus Environment

Page 29 of 82
Journal of Medicinal Chemistry
1
2
Discussion and Conclusions
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
In the development of inhibitors of the C. jejuni UDP-4-amino-sugar acetyltransferase,
PglD, we initially screened both fragment-based and HTS libraries. These studies led to the
identification of a thienopyrimidine-6-carboxylate core as a promising hit. Over the course of the
SAR studies we found substitution at the C2-position of the core to be most impactful and that the
binding mode appeared to be poorly predicted in silico. By combining data from substrate
competition assays, DSF, and DLS we confirmed that the inhibitors bind exclusively to the
AcCoA binding groove of the active site without disrupting the homo-trimeric structure of PglD.
Critically, crystallographic analysis reveals a 180° flip of the thienopyrimidine core to
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
place the R -substituent in the AcCoA groove, as was anticipated from the SAR data. However, in
contrast to previously characterized small fragments, we observed a 3.9 Å displacement of the
inhibitor further along the AcCoA groove away from the known thioester binding site. Notably, in
this new binding mode, the inhibitor makes several hydrogen-bonding interactions that
compensate for the ones that would have been formed with the key catalytic residues (Asn118 &
His134), but also precludes a stabilizing T-shaped π-interaction between the thienopyrimidine
2
core and Phe152. Ultimately, it appears that placement of the R -phenyl into the adenine sub-
pocket of AcCoA binding site may be the main driving force of ligand binding and indeed further
optimization of this interaction resulted in to our most promising lead. Specifically, inhibitor 43
(
IC50 = 0.46 µM) shows an excellent ligand efficiency of 0.39 and a favorable clogP of 2.94. To
the best of our knowledge this is the first competitive inhibitor with an IC50 value < 1 µM for a
bacterial amino-sugar-modifying enzyme. A selection of inhibitors from the series presented
herein was also tested against the structurally homologous amino-sugar acetyltransferase from N.
gonorrhoeae (PglC in that O-linked glycosylation pathway) and no significant inhibition was
2
9
ACS Paragon Plus Environment