Binding inhibitors of the bacterial sliding clamp by design

Article abstract of DOI:10.1021/jm2004333

The bacterial replisome is a target for the development of new antibiotics to combat drug resistant strains. The β₂ sliding clamp is an essential component of the replicative machinery, providing a platform for recruitment and function of other replisomal components and ensuring polymerase processivity during DNA replication and repair. A single binding region of the clamp is utilized by its binding partners, which all contain conserved binding motifs. The C-terminal Leu and Phe residues of these motifs are integral to the binding interaction. We acquired three-dimensional structural information on the binding site in β₂ by a study of the binding of modified peptides. Development of a three-dimensional pharmacophore based on the C-terminal dipeptide of the motif enabled identification of compounds that on further development inhibited α-β₂ interaction at low micromolar concentrations. We report the crystal structure of the complex containing one of these inhibitors, a biphenyl oxime, bound to β₂, as a starting point for further inhibitor design. Published 2011 by the American Chemical Society.

Full text of DOI:10.1021/jm2004333

ARTICLE
pubs.acs.org/jmc
Binding Inhibitors of the Bacterial Sliding Clamp by Design†
Gene Wijffels,*,‡ Wynona M. Johnson,§ Aaron J. Oakley,||,^ Kathleen Turner,§ V. Chandana Epa,#
Susan J. Briscoe,‡ Mitchell Polley,§ Andris J. Liepa,§ Albert Hofmann,§ Jens Buchardt,3,(
Caspar Christensen,3,( Pavel Prosselkov,||,z Brian P. Dalrymple,‡ Paul F. Alewood,3 Philip A. Jennings,‡,þ
Nicholas E. Dixon,||,^ and David A. Winkler§,O
‡CSIRO Livestock Industries, 306 Carmody Road, St. Lucia, Queensland 4067, Australia
§CSIRO Material Science and Engineering, Bayview Avenue, Clayton, Victoria 3168, Australia
Research School of Chemistry, Australian National University, Canberra, Australian Capital Territory 0200, Australia
^School of Chemistry, University of Wollongong, New South Wales 2522, Australia
#CSIRO Material Science and Engineering, 343 Royal Parade, Parkville, Victoria 3052, Australia
3Institute of Molecular Bioscience, 306 Carmody Road, University of Queensland, St. Lucia, Queensland 4067, Australia
OMonash Institute for Pharmaceutical Sciences, 381 Royal Parade, Parkville, Victoria 3052, Australia
S Supporting Information
b
ABSTRACT: The bacterial replisome is a target for the
development of new antibiotics to combat drug resistant strains.
The β2 sliding clamp is an essential component of the replicative
machinery, providing a platform for recruitment and function of
other replisomal components and ensuring polymerase proces-
sivity during DNA replication and repair. A single binding
region of the clamp is utilized by its binding partners, which
all contain conserved binding motifs. The C-terminal Leu and
Phe residues of these motifs are integral to the binding inter-
action. We acquired three-dimensional structural information
on the binding site in β2 by a study of the binding of modified
peptides. Development of a three-dimensional pharmacophore based on the C-terminal dipeptide of the motif enabled
identification of compounds that on further development inhibited Rꢀβ2 interaction at low micromolar concentrations. We
report the crystal structure of the complex containing one of these inhibitors, a biphenyl oxime, bound to β2, as a starting point for
further inhibitor design.
’ INTRODUCTION
replication and repair.14,15 The essential and central role of the
clamp, the conservation of its interactions with many partners,
and its limited cellular availability of 300ꢀ600 β2 dimers per
cell16,17 make it an attractive target for rational design of new
antimicrobial agents.1,2,5
Bioinformatic analysis identified a conserved clamp binding
site, a C-terminal or proximal C-terminal pentameric motif,
QL[S/D]LF, within the sequences of DNA polymerases of many
bacteria.18 Experimental evidence substantiated that peptides
bearing the consensus clamp binding motif (CBM) or related
hexapeptide sequences bind directly to the clamp and compete
successfully with clamp binding proteins.18,19 Mutation of the
first-identified (internal) CBM in the Escherichia coli Pol III R
polymerase subunit with various versions of the motif signifi-
cantly affected Pol III function in vitro and in vivo.20
The bacterial replisome is a clear but underexploited target for
the development of new classes of antibiotics to address the
problem of drug resistance.1,2 The complex multiprotein ma-
chinery of the bacterial DNA replication and repair systems offer
many target sites for the intervention of drugs to ameliorate or
arrest the progress of infectious disease.2,3 Indeed, the bacterial
replicative polymerases and even the complete replisome have
been the targets of recent drug screening and structureꢀfunction
studies in the search for new antimicrobial leads.2ꢀ6
A highly conserved component of the bacterial replisome is
the sliding clamp, a homodimer of two β subunits that form a
stable donut shaped ring around DNA,7,8 tethering the replica-
tive machinery while allowing the complex to slide along the
duplex.9 The clamp hosts each of the DNA polymerases (Pol
IꢀV) that function in chromosomal DNA replication and
repair,10,11 and also interacts directly with the δ subunit of clamp
loader complex,12 with Hda, a protein involved in initiation of
replication,13 and several other proteins that function in DNA
Received: April 12, 2011
Published: May 23, 2011
Published 2011 by the American Chemical Society
4831
dx.doi.org/10.1021/jm2004333 J. Med. Chem. 2011, 54, 4831–4838
|

Journal of Medicinal Chemistry
ARTICLE
An alternative region of the R subunit (of the E. coli Pol III
replisome), located very close to its C-terminus, had earlier been
proposed to be the more critical clamp-binding site.21 A peptide
bearing this sequence, QVELEFD, did indeed compete success-
fully in clamp binding and interfere in DNA replication assays.
Nevertheless, Dohrmann and McHenry20 demonstrated that the
internal R subunit site, QADMF, is absolutely required by Pol III
holoenzyme for processive DNA synthesis. The precise roles of
the two CBMs in the Pol III R subunit remain to be fully
elucidated.
Crystal structures of complexes of CBM peptides from Pols II,
III (C-terminal site) and IV bound to β2 are now available.5,22,23
The peptides occupy the same surface pocket comprising two
subsites. In simple terms, subsite 1 is an 8.5 Å deep square pocket
that accommodates the C-terminal Leu-Phe (in pentapeptides)
or Leu-Xaa-Phe (in hexapeptides) portion of the motif. Subsite 2
is a narrow shallow slot (4.5 Å deep) that is occupied by the
N-terminal portion of the CBM.5,22,23
In a search for small molecule binders of the sliding clamp, the
O’Donnell group5 screened a library of 30,600 small synthetic
molecules and successfully identified a lead entity, 1 (RU75).
This compound inhibited in vitro DNA synthesis by Pol III and
also competed with a peptide containing the Streptococcus
pyogenes Pol C CBM for binding to S. pyogenes β2.5 A cocrystal
structure of 1 with the clamp revealed that it was bound only
within the deeper subsite 1 and with fewer contacts than the
larger synthetic peptides.5
Table 1. Inhibition of rꢀβ2 and β2ꢀConsensus Peptide
Assays by Peptide Variants
peptide sequence
Rꢀβ2 binding IC50 (nM)a
SPR IC50 (nM)a
DnaE Nonomer Peptides with Substitutions at Positions 4 and 5
of the DnaE CBM, QADMF
IG QLSLF GV
IG QADMF GV
IG QADLF GV
IG QADIF GV
IG QADFF GV
IG QADWF GV
IG QADMW GV
2,300 ( 1,100
96,000
550
3,000b
nd
56,000
315,000
174,000
345,000
282,000
nd
nd
nd
nd
N-Acetylation of the Pentameric CBM, QLDLF
QLDLF
2,600 ( 1,400
4,500 ( 700
70 ( 20
620
nd
pyroQLDLF
Ac-QLDLF
400
Most Competitive Peptides with Substitution in Positions
4 and 5 of Ac-QLDLF
Ac-QLDL[2ClF]
34 ( 8
11 ( 1
27 ( 18
21 ( 6
37 ( 1
117 ( 49
Ac-QLDL[3ClF]
Ac-QLDL[4ClF]
Ac-QLDL[3,4ClF]
Ac-QLDM[3,4ClF]
49 ( 8
20 ( 5
23 ( 11
33 ( 1
With limited structural information available at the time
we commenced this study, we adopted the alternate strategy
of rational molecular design. This required assembly of three-
dimensional (3D) spatial information about the target site
through interrogation with competitive binders (ligand based
design) and/or protein crystal structures (receptor structure
based design). We anticipated that a combination of these app-
roaches would lead to an informed design process.
To build up a 3D model through ligand interrogation, we
generated a series of synthetic pentapeptides incorporating
natural and non-natural substitutions of the C-terminal LF
component of the CBM. Assessment of these as inhibitors in
binding assays led to an optimized peptide. Since modifications
of LF gave improved binding, we developed a novel 3D infor-
matic technique to define a pharmacophore query based on the
likely conformations of the [S/D]LF moiety. Searches of data-
bases of small synthetic compounds with this 3D motif identified
a small number of hits that showed weak inhibitory activity. A
group of small molecules with related structures were then
selected for confirmation of binding, and those with improved
binding were used for cocrystallization with E. coli β2.
N-Terminal Extensions to QLDL[3,4ClF]
propionyl-
63 ( 11
62 ( 19
67 ( 17
66 ( 7
61 ( 17
69 ( 23
337 ( 18
537 ( 76
27 ( 2
pivaloyl-
benzyloxyacetyl-
3-indoleacetyl-
2,4-dimethoxybenzoyl-
benzoyl-
42 ( 15
69 ( 24
20 ( 15
86 ( 0.5
456 ( 80
41 ( 21
89 ( 10
isobutyloxycarbonyl-
methyloxycarbonyl-
a Assayed as described.19 IC50 values represent the concentration of
peptide required to inhibit interactions by 50%. Mean ( SEM of at least
2 independent measurements. nd, not done. b Data from Wijffels et al.19
competition assay.19 Where available, the dissociation constants
(KD) of peptideꢀβ2 complexes measured directly by surface
plasmon resonance (SPR) also concur with their inhibitory
activities.19 Since the Leu4-Phe5 portion of the consensus
pentameric CBM (QL[S/D]LF) is buried deeply in subsite
1,5,22,23 we anticipated that changes in these residues would have
significant potential for optimization.
A series of substitutions were made within the context of
pepDnaE-n (IGQADMFGV).19 This peptide with Q1A2D3M4F5
as its CBM exhibited a moderate KD (SPR) and modest IC50 in
the Rꢀβ2 plate binding assays relative to those of peptides
containing the consensus sequence.19 Consistent with previous
studies, substitution of Met4 with Leu improved inhibitory
activity in the Rꢀβ2 and SPR assays (Table 1). Peptides with
Ile, Phe, and Trp in position 4 or Trp in position 5 were weak
inhibitors in the plate assays. Substitutions with Cys, Val, or Tyr
at position 4 or Cys, Leu, Ile, Val, or Tyr at position 5 abolished
inhibitory activities (not shown).
We now report the structure of a new small chemical entity
thus identified, bound in the CBM binding site of β2. We
compare it to the binding of 1 in the same site and a model of
binding of the optimized pentapeptide. Together, the structural
information from these sources allows inference of ligand design
modifications that may be expected to provide compounds with
improved binding to β2.
’ RESULTS AND DISCUSSION
SAR of Modifications to the Peptide Motif. Synthetic
peptides with different CBMs generally exhibit correlated in-
hibitory activities (IC50) in a simple Rꢀβ2 plate binding assay
and a β2ꢀconsensus CBM surface plasmon resonance (SPR)
The observation that the β2 binding site tolerates some amino
acids with bulky but neutrally charged (lipophilic) side chains led
4832
dx.doi.org/10.1021/jm2004333 |J. Med. Chem. 2011, 54, 4831–4838

Journal of Medicinal Chemistry
ARTICLE
Figure 2. (A) Consensus superimposition of examples of LDLF motifs
from the PDB. (B) 3D pharmacophore query for the dominant DLF
conformation. The vertex points represent the terminal side chain
moieties of D (carboxyl carbon), L (dimethyl), and F (phenyl ring
centroid).
Figure 1. Computational model for the complex between β (in orange
surface representation) and the peptide Ac-QLDL[3,4ClF] in stick
representation.
of N-terminal extensions of the peptide that project from subsite
2 does not always impact peptide binding.
Modeling of Non-Natural Substitutions at Position 5 of
the Pentameric CBM. With the availability of experimentally
determined structures of β2 bound to various peptide and protein
ligands,5,22,23 we modeled the structure of its complex with
peptide Ac-QLDL[3,4ClF] (Figure 1). The model indicated
that β residues Arg152, Thr172, Gly174 to Leu177, Pro242,
Val247, Tyr323, Val344, Ser346, Val360, and Met362 to Arg365
interact with the peptide. Leu2, Leu4, and [3,4-dichloro-Phe]5 of
the peptide are all situated in the hydrophobic subsites in β; the
3,4-dichloro-Phe contacts residues Thr172, Leu177, Pro242, and
Val247 of subsite 1.
The model is fully consistent with the experimental data.
Chlorine substitution at the 3- and 4-positions of the Phe5 ring
produces better van der Waals’ contacts and stronger hydrophobic
interactions, consistent with the lower IC50 value (Table 1).
Subsite 1 is not large enough to accommodate benzyl, biphenyl,
or naphthalene moieties. Hence, peptides where Phe5 is sub-
stituted with these groups exhibited weaker interactions with β.
While a pyridine could be accommodated in the pocket as easily
as a phenyl ring, it would have a larger desolvation energy penalty
on transfer from water to the hydrophobic protein environment.
This would explain the higher IC50 values observed when the
phenyl ring is substituted by pyridine. The N-acetyl substituent
of Gln1 (labeled as ACE-0 in Figure 1) interacts favorably with β,
forming a hydrogen bond with the backbone of Arg365, account-
ing for the significant increase in binding affinity upon N-acetyl-
ation (Table 1). Because of the space around the N-acetyl group,
larger N-terminal amides can be accommodated without affect-
ing interaction with β (Table 1). However, replacement of the
N-acetyl group with bulky sulfonyl groups (Table S3, Supporting
Information) has an adverse effect on this favorable interaction
due to the change in geometry (tetrahedral sulfonyl cf. the planar
amide group) and steric clashes.
us to seek increased structural diversity by substitution of
positions 4 and 5 with large non-natural side chains and ring
structures. This study was conducted using the N-terminally
acetylated pentamer, Ac-Q1L2D3L4F5. The unprotected peptide
was found to be susceptible to spontaneous cyclization of
the N-terminal Gln to pyroglutamine, and this was a concern as
the pyroGln pentamer was a less effective inhibitor in the Rꢀβ2
plate assay (Table 1). However, in a serendipitous finding,
acetylation of the N-terminus by itself caused a 40-fold decrease
in the Rꢀβ2 IC50 from 2.6 μM to 70 nM. Thus, a series of
peptides based on Ac-QLDLF were prepared where Phe5 was
replaced by substituted phenylalanines, a benzylalanine,
naphthylalanines, or pyridylalanines. Assay of these peptides
revealed that those with chlorinated phenylalanines were super-
ior, with a 3ꢀ7-fold improvement in the Rꢀβ2 inhibitory activity
relative to that of the parent peptide Ac-QLDLF (Table 1 and
Table S1, Supporting Information).
In the next iteration, a series of peptides with substitutions of
non-natural amino acids at position 4 of Ac-QLDL[3,4ClF]
(3,4ClF = 3,4-dichloro-Phe) were studied. Only the methionine
substituted pentamer was competitive. The 2- and 3-chloro-Phe,
methyl-leucine, 3-allylglycine, and cyclohexylalanine derivatives
retained various levels of inhibitory activity (with Rꢀβ2 IC50
values of 0.1ꢀ0.6 μM; Table S2, Supporting Information). Thus,
in the context of a 3,4-dichloro-Phe5, there appeared to be limited
scope to enhance activity with modifications at position 4.
The improved inhibitory activity of the N-acetylated pentamer
also encouraged further investigation of substitution at the
N-terminus (Table S3, Supporting Information). A series of
aliphatic and aromatic amides, carbamates, and sulphonamides
were tested as N-terminal derivatives of QLDL[3,4ClF]
(Table 1). The best inhibitors contained the aliphatic amide
functionality, where linear and cyclic groups were well tolerated
in the Rꢀβ2 assay. Two aromatic amides, benzoyl and 2,4-
dimethyl benzoyl, were also good inhibitors, as were the simple
isobutyl- and methyl carbamates. None of the sulphonamides
were inhibitory (Table S3, Supporting Information).
Conserved Tertiary Structure in the Peptide Motif. The
peptide study indicated scope for improvement of binding
through modifications of the Leu4-Phe5 moiety of the CBM.
We next investigated structural conservation of [D/S]LF motifs,
mindful that in simple binding assays the linear tripeptides alone
were poor inhibitors.18 We extracted from the PDB all occur-
rences of the LDLF sequence in proteins whose structure
had been determined to better than 3 Å resolution. Of the
122 examples found, 85 LDLF segments (70%) adopted a
similar conformation (Figure 2A). Two populations of Phe5
Good inhibitory activity in the plate binding assay was not
consistently paralleled by competitive performance in the SPR
assay. The presence of large ring structures markedly increased
the SPR IC50, the one exception being the 2,4-dimethylbenzoyl
substituent. From a design viewpoint, it is clear that incorporation
4833
dx.doi.org/10.1021/jm2004333 |J. Med. Chem. 2011, 54, 4831–4838

Journal of Medicinal Chemistry
ARTICLE
Figure 4. Compound 4 bound in the CBM binding site of the E. coli β2
sliding clamp. The 2mFo ꢀ DFc electron density omit-map is drawn in
“wire basket” form, contoured at 0.33 e/Å3 (1σ). Carbon atoms are in
yellow, and water molecules are shown as red spheres.
Figure 3. Structures of small molecule mimics of the DLF component
of the CBM.
somewhat higher than expected for a structure at 1.9 Å resolution,
most likely due to twinning of the crystal. Nevertheless, the final
electron density for the protein and ligand is excellent (Figure 4).
Overall, the structures of ligand free25 and ligand bound forms
of the sliding clamp are very similar, superimposing with an rmsd
of 1.3 Å over all 732 CR atoms. There is a slight change in the
overall shape of the β2 ring. However, this may reflect the effects
of crystal packing rather than ligand binding. The X-ray data
revealed that 4 is bound in a hydrophobic cleft on the surface of
both β monomers (see Figure 4). This pocket is formed by
Arg152, Tyr154, Gly174, Pro242, Val247, Val360, and Met362
(nominally subsite 1). The biphenyl group of 4 participates in
hydrophobic interactions with the pocket, and the distal aromatic
ring in 4 is twisted out of the plane of the proximal ring. The
cyano group fits into a small hydrophobic pocket that would not
accommodate a larger ligand. This single hydrophobic interac-
tion at this end of the molecule may account for the improved
binding of this compound as compared to that of 2 and 3. The
conformations of the oxime link and propyl side chains are
different in the ligands bound to each monomer (Figure 5C).
Curiously, a single hydrogen bond is evident between one
carbonyl oxygen and Arg240 in one β monomer but not in the
other. Potentially Arg152 could twist around to form a hydrogen
bond with the other carbonyl, but it does not. These observations
are not surprising when one considers that the cyclohexanedione
portion of the ligand is predominantly exposed to the solvent and
that there may be little energetic advantage in the formation of a
hydrogen bond between the ligand and the protein as compared
to the solvent.
conformations was the dominant variation among these struc-
tures, while the Asp and Leu side chains occupied relatively
restricted regions of conformational space.
Virtual Screening for Small Molecule Tripeptide Motif
Mimetics. By choosing average structures representative of each
of the two dominant Phe5 conformations, 3D pharmacophore
queries could be generated for each; Figure 2B shows a pharmaco-
phore query for the DLF peptide motif. These pharmacophores
were used to conduct 3D flexible database searches in the
National Cancer Institute (NCI) and in-house chemical data-
bases. The NCI database24 search generated several hits that
matched the query that did not contain significant additional
structural features that may interfere with binding to β2. Two
compounds (of 25 tested) showed weak inhibitory activity in the
Rꢀβ2 assay: NCI 131123 and NCI 52310-Q (Figure 3) had
Rꢀβ2 IC50 values of 210 and 300 μM, respectively.
A similar virtual screen was conducted on the in-house
collection of 32,000 compounds, followed by an Rꢀβ2 plate
binding assay of 20 virtual hits. Two compounds, 2 (Rꢀβ2 IC50 =
270 μM) and 3 (350 μM), showed consistent, albeit weak,
inhibition. Significantly, both compounds (Figure 3) were of
related oximinodioxo-cyclohexane and -pyran chemical classes.
Structure of β2 with a Bound Small Molecule Peptide
Motif Mimic. Assay of 155 analogues of these oxime ether hits
indicated that Rꢀβ2 binding inhibition was characteristic of
these classes (details to be reported elsewhere).
Crystallization trials on a subset of oximinodioxo-cyclohexane
and -pyran peptide mimetics that exhibited superior inhibitory
activity resulted in the successful cocrystallization of biphenylox-
ime ether 4 with E. coli β2. This compound exhibited an Rꢀβ2
IC50 of 40 ( 17 μM (7 replicates) and an SPR IC50 of 8.3 μM
(2 replicates). The initial electron density maps calculated from
the X-ray diffraction data at 1.9 Å resolution showed positive
electron density in the CBM binding region of each β monomer
(peaks >6σ in the mFo ꢀ DFc electron density map). This was
easily interpreted as compound 4. After several rounds of model
building and refinement, the final model consisted of a β2 dimer
with one molecule of 4 in each CBM binding site. Statistics
pertaining to the X-ray structure and model are given in Table S4
(Supporting Information). The final R and Rfree values are
By superimposing the structure of ligand free β2 with the
above structure, it is possible to determine which residues have
moved upon the binding of 4. The most dramatic movement is
seen in Met362, which is partly disordered in the ligand free
structure. In the complex, the Met362 side chain lies above one of
the biphenyl rings. There are also compensatory movements in
Val247, Pro242, and Arg152, which have moved toward the
biphenyl system of 4.
The structure of the β2 complex with 4 can be compared with
the complex with 1,5 a small-molecule compound that selectively
inhibits Pol III. Compound 1 features a dibrominated aromatic
ring that overlaps with the fluorobiphenyl group of 4 (Figure 6A).
Both make hydrophobic interactions with Leu177, Val247,
Val360, and Met362 in subsite 1. The aromatic groups of both
4834
dx.doi.org/10.1021/jm2004333 |J. Med. Chem. 2011, 54, 4831–4838

Journal of Medicinal Chemistry
ARTICLE
Figure 5. (A) Superimposition of the binding conformation and orientation of ligand 4 (in cyan) and the AMSLF peptide (in magenta; from the δꢀβ
structure, PDB ID: 1JQL26) from the X-ray structures of β complexes. Note the close proximity of the L and F side chains of the peptide with the
hydrophobic biphenyl moiety in 4. (B) Superimposition of the X-ray conformation of the AMSLF peptide (in yellow; PDB ID 1JQL26) with two DLF
conformers from the PDB (in gray) close to the consensus conformation for SLF used for the virtual screen 3D pharmacophore query. (C)
Superimposition of different X-ray conformations of 4 from the two binding sites in β2. The cyclohexanedione and biphenyl rings align well, but the
oxime and propyl chains adopt alternate conformations.
ligand that could be accessed by suitable substituents. Thr172 lies
approximately 2.5 Å from the 2 positions on both phenyl groups
and could interact with a small hydrogen bond acceptor like a
fluoro substituent, or form a new hydrogen bond following
conversion of one of the phenyl rings to a pyridine. The former
strategy may be more effective, as it retains the hydrophobic
character of the ligand, and a pyridyl substituent may cause the
biphenyl group to rotate by 180ꢀ so that the more hydrophilic
edge is exposed to the solvent.
Met362 and Ser346 both occupy positions above the distal
end of the hydrophobic pocket. A methyl substituent on the 2
position of the distal phenyl ring could make hydrophobic
contact with Met362. A small hydrophobic substituent would
be required in this position both to prevent twisting of the phenyl
rings and because of the short distance (3.5 Å) between the
S-methyl group of Met362 and the phenyl ring. Ser346 could
form a favorable interaction if a hydrogen bond acceptor was
placed in the 3 position of the distal phenyl ring or the fluorine
was replaced with a better hydrogen bond acceptor. The distance
between the 3 and 4 carbon atoms of 4 and the Ser hydroxyl
Figure 6. Orientation of ligands in the protein binding pocket of the β2
sliding clamp from X-ray structures. (A) Stereo view of the βꢀ4 and
βꢀ1 (PDB ID: 3D1G) complexes, with β shown in cartoon form and
the compounds in gray sticks. Side-chains of β that interact with 4 have
magenta carbon atoms, and those that interact with 1 have cyan carbon
atoms. (B) Biphenyloxime ether ligand 4. (C) AMSLF ligand (PDB ID:
1JQL). The protein is shown as a molecular surface encoded for
lipophilicity, with brown indicating hydrophobic and blue indicating
hydrophilic regions. The magenta arrows denote the positions of Met362.
oxygen is 5.3 and 5.0 Å, respectively.
Arg152 is 3.8 Å from the methylene group connecting the
oxime to the biphenyl moiety of 4 and 6ꢀ7 Å from the C1 atom
of the propyl side chain. Both of these methylene carbons could
be substituted by hydrogen bond acceptors or acidic groups to
create stronger hydrogen bond or salt bridge interactions with
this residue. Similarly, Tyr154 is 4ꢀ6 Å from this methylene
group and the relatively mobile propyl side chain carbons;
therefore, decoration of these with hydrogen bond acceptors
could lead to enhanced binding. Furthermore, extension of the
propyl chain culminating in a hydrogen bond acceptor could
access favorable interaction with Tyr153 at the end of this pocket.
In the X-ray structure, Arg246 is quite close to the end of the
cyclohexanedione ring, but faces away from the ligand. This
residue seems to be relatively mobile; therefore, a suitable
hydrogen bond acceptor attached to the equatorial methyl group
at the end of the cyclohexane ring may allow Arg246 to swing
around and make a hydrogen bond or salt bridge interaction with
the ligand.
compounds are approximately coplanar, but the bulky bromine
atoms cause the aromatic ring of 1 to be pushed out of the
binding site by about 1.4 Å compared with that of 4 (Figure 6A).
The remaining portions of each ligand have little in common.
The guanidinium group of Arg152 is in an adjusted position so
as to form a hydrogen bond with 1. Furthermore, the carboxy-
late moiety of 1 forms a hydrogen bond with the backbone
amide of Asp243. Overall, the biphenyl group in 4 makes more
hydrophobic contacts, whereas 1 has more hydrogen bonding
interactions.
Comparison of the structure of the bound AMSLF peptide
from the structure of the βꢀδ (clamp loader) complex26 with
that of 4 reveals that the two ligands only partially overlap
(Figures 5A and 6B,C), but the binding site retains a very similar
structure. Satisfyingly, the conformation of the SLF moiety of the
Insights for Improved Ligand Design. The β2ꢀ4 cocrystal
structure identifies important interactions between the fluorobi-
phenyl moiety and subsite 1 of β. However, inspection of the
structure shows that there are several other residues near the
4835
dx.doi.org/10.1021/jm2004333 |J. Med. Chem. 2011, 54, 4831–4838

Journal of Medicinal Chemistry
ARTICLE
CBM in the βꢀδ structure closely resembles the consensus
structure for one of its two conserved conformations from the
PDB 3D motif search (Figure 5B). The conserved LF residues
from the bioinformatics analysis and the biphenyl functional
group in 4 selected by the 3D pharmacophore search also bind in
the same region of the protein (Figures 5A and 6B,C). However,
the small section at the edge of the hydrophobic pocket into
which the LF and biphenyl moieties bind changes shape as the
ligand is changed. This primarily involves rotation of Met362
(arrows in Figure 6B,C). When the peptide binds, Met362
(arrow in Figure 6C) swings aside to open up a channel in the
side of the hydrophobic pocket through which the peptide
backbone can pass. When the biphenyl ligand binds, Met362
rotates to be near His175, closing off this channel (Figure 6B). A
deeper pocket on the side of the hydrophobic cleft (Figure 6) is
occupied by Phe but not by 4, and placement of a hydrophobic
group on the 3 position of the proximal biphenyl ring may
enhance binding at this end of the molecule.
(DIPEA). As derivatization reagents (Table S5, Supporting Information),
carboxylic acid chlorides, sulfonyl chlorides, and chloroformates were
used directly, with DIPEA added to the resin prior to the addition of the
reagent, while carboxylic acids were used together with O-(benzotriazol-
1-yl)-N,N,N0,N0-tetramethyluronium hexafluorophosphate and DIPEA
(4 equiv) as standard couplings. The peptides were purified using
standard reverse-phase HPLC methods. All peptides were >95% pure
by HPLC and mass spectrometric analyses were consistent with the
calculated values.
Inhibition Assays. The Rꢀβ2 plate binding assays were as
described.19 Stock 2.5 or 5 mM solutions of peptides were prepared in
methanol, and dilutions were freshly prepared in aqueous buffers. The
NCI and in-house library compounds were dissolved in DMSO at
10 mM and then diluted in aqueous buffers. Only soluble compounds
proceeded to assay. The SPR assay was as described.19
Molecular Modeling. Pharmacophore searching used the Tripos
software pages Sybyl 8.0 (Tripos Inc.) and Unity. The biopolymer
peptide sequence search facility was used to extract all examples of the
LDLF peptide sequence from structures in the PDB and to align them by
CR superimposition in 3D space. Since there were clearly only two
preferred conformations, the few outliers were carefully removed. The
terminal binding groups of the amino acid side chains were then used as
3D database queries, with suitable variations in the intergroup distances
being permitted. The database searches were carried out in Unity using
the 3D flexible search option that allows the main low energy con-
formations of any putative hit structure to be accessed during the search.
To understand the structural basis of observed trends in binding
affinities of peptides with β2 (Table 1), an atomic model of the peptide
Ac-QLDL[3,4ClF] bound to β was constructed using the crystal
structure5 of the Pol II peptide (GQLGLF) complex (PDB ID:
3D1E) as the template. After deleting the N-terminal Gly and N-acety-
lating the Gln, the central Gly was mutated to Asp, selecting an
energetically most favorable side-chain rotamer pointing toward the
external solvent. The C-terminal Phe was then modified by 3,4-dichloro
substitution on the phenyl ring. These modifications were done with
InsightII v. 2005 (Accelrys, San Diego, CA). This structure was then
energy minimized with Discover v. 2.98 (Accelrys), using the CVFF
force field and a distance-dependent dielectric function. During energy
minimization, backbone and all heavy atoms of amino acid residues of β
not interacting with the peptide were held fixed.
These alterations address improvement in binding to the
intermolecular binding site on β, but additional modifications
would also be required to optimize ADME properties for delivery
and bacterial uptake in vivo. In particular, the ClogP of 4 is
probably too high, and polar groups would need to be introduced
to bring it within the ideal range.
’ CONCLUSIONS
Studies by several groups have shown that the interaction site
of the β2 sliding clamp and its polymerase partners is amenable to
modulation by small chemical entities. We have chosen a rational
design process to identify such entities.
We used simple modifications of the pentapeptide clamp
binding motif (CBM) to derive structural information about
the binding requirements of the interaction site on β2 and then
prepared an optimized pentapeptide. A 120-fold improvement of
the Rꢀβ2 IC50 and 25-fold improvement in the SPR assay was
achieved by N-terminal acetylation and chlorine substitution on
the Phe ring in position 5 of the pentameric CBM.
A new method was developed for estimating the likely
conformation of the DLF motif within the CBM using a
consensus of conserved conformations from the PDB. This
allowed pharmacophore models to be defined, and these identi-
fied lead molecules in a virtual screen. The crystal structure of β2
with one of the inhibitors bound revealed details of interactions
that can be used to improve the activity of these lead compounds
by targeted design modifications. The medicinal chemistry and
QSAR of these leads will be described in subsequent papers.
Together with the structure of the β2ꢀ1 complex,5 the new
structural data form a significant basis for the design of new ligands
through lead-hopping, isostere replacement, and computer-aided
design. These examples should encourage further investment to
meet the urgent need for new classes of antibiotics to challenge the
rising tide of drug resistance among bacterial pathogens.
Chemistry. Detailed synthetic procedures and characterization for
all new compounds are in Supporting Information. In brief, the
oximinodioxo-cyclohexanes and -pyrans were prepared from the corre-
sponding N-alkoxyphthalimide by treatment with N,N-diethylethylene
diamine, followed by reaction with the appropriate 2-acyl-1,3-dicarbonyl
compound under acidic conditions.
Structural Biology. The β2 clamp was prepared as described.25
Crystals were grown by the hanging drop vapor diffusion method. The
reservoir contained 1 mL of a solution of 30% PEG400, 0.17 M calcium
acetate, and 0.1 M MES at pH 6.5. Hanging drops contained 2 μL of
protein (47 mg/mL), 2.5 μL of 2.3 mM 4 in ethanol, and 2 μL of
reservoir solution. X-ray data were collected on a MAR345 area detector
using CuꢀKR radiation from a Rigaku RU-200 rotating anode X-ray
generator and processed with DENZO and SCALEPACK from the
HKL package.27 The model was refined with REFMAC,28 with model
building in O.29 Figures were prepared with PyMOL (DeLano Scientific;
http://pymol.sourceforge.net/).
’ EXPERIMENTAL SECTION
Peptide Synthesis. The synthesis of all the peptides is described in
Supporting Information. Briefly, the standard Fmoc/tBu protection
strategy was used for solid phase peptide synthesis on a 2-chlorotrityl
chloride polystyrene resin; some were synthesized in a 96 well format.
Acetylation of the pentamer (with and without the chlorine substitutions
on the Phe5 ring) used acetic anhydride and diisopropylethylamine
’ ASSOCIATED CONTENT
S
Supporting Information. Details of the synthesis and
b
characterization of peptides, peptide derivatives, and com-
pounds, including a table showing derivatization reagents, data
for inhibition by synthetic peptide derivatives of interactions of
4836
dx.doi.org/10.1021/jm2004333 |J. Med. Chem. 2011, 54, 4831–4838

Journal of Medicinal Chemistry
ARTICLE
β2 with R and an immobilized peptide (three tables), and data
collection and refinement statistics for X-ray crystallography.
This material is available free of charge via the Internet at
http://pubs.acs.org.
(7) Kong, X.-P.; Onrust, R.; O’Donnell, M.; Kuriyan, J. Three-
dimensional structure of the β subunit of E. coli DNA polymerase III
holoenzyme: a sliding DNA clamp. Cell 1992, 69, 425–437.
(8) Georgescu, R. E.; Kim, S.-S.; Yurieva, O.; Kuriyan, J.; Kong, X.-P.;
O’Donnell, M. Structure of a sliding clamp on DNA. Cell 2008, 132,
43–54.
(9) Stukenberg, P. T.; Studwell-Vaughan, P. S.; O’Donnell, M.
Mechanism of the sliding β-clamp of DNA polymerase III holoenzyme.
J. Biol. Chem. 1991, 266, 11328–11334.
(10) Lovett, S. T. Polymerase switching in DNA replication. Mol.
Cell 2007, 27, 523–526.
(11) Langston, L. D.; Indiani, C.; O’Donnell, M. Whither the
replisome: emerging perspectives on the dynamic nature of the DNA
replication machinery. Cell Cycle 2009, 8, 2686–2691.
(12) Jeruzalmi, D.; O’Donnell, M.; Kuriyan, J. Clamp loaders and
sliding clamps. Curr. Opin. Struct. Biol. 2002, 12, 217–224.
(13) Kurz, M.; Dalrymple, B.; Wijffels, G.; Kongsuwan, K. Interac-
tion of the sliding clamp β-subunit and Hda, a DnaA-related protein.
J. Bacteriol. 2004, 186, 3508–3515.
(14) Lꢀopez de Saro, F. J.; O’Donnell, M. Interaction of the β sliding
clamp with MutS, ligase, and DNA polymerase I. Proc. Natl. Acad. Sci.
U.S.A. 2001, 98, 8376–8380.
(15) Kongsuwan, K.; Josh, P.; Picault, M. J.; Wijffels, G.; Dalrymple,
B. The plasmid RK2 replication initiator protein (TrfA) binds to the
sliding clamp β subunit of DNA polymerase III: Implication for the
toxicity of a peptide derived from the amino-terminal portion of 33-kDa
TrfA. J. Bacteriol. 2006, 188, 5501–5509.
Accession Codes
†The atomic coordinates and structure factors of compound 4 in
complex with the E. coli sliding clamp have been deposited in the
Protein Data Bank, www.rcsb.org (PDB code 3QSB).
’ AUTHOR INFORMATION
Corresponding Author
*CSIRO Livestock industries, Level 7 QBP, 306 Carmody Rd.,
St. Lucia, Queensland 4067, Australia. Phone: 61 7 3214 2510.
Fax: 61 7 3214 2900. E-mail: gene.wijffels@csiro.au.
Present Addresses
ꢁ
(Novo Nordisk A/S, Novo Nordisk Park, DK-2760 Maløv,
Denmark.
zLaboratory for Behavioral Genetics, RIKEN Brain Science
Institute, Wako, Saitama 351-0198, Japan.
þCephalon Australia, Level 2, 37 Epping Road, Macquarie Park,
New South Wales 2113, Australia.
(16) Leu, F. P.; Hingorani, M. M.; Turner, J.; O’Donnell, M. The δ
subunit of DNA polymerase III holoenzyme serves as a sliding clamp
unloader in Escherichia coli. J. Biol. Chem. 2000, 275, 34609–34618.
(17) Sutton, M. D.; Duzen, J. M.; Maul, R. W. Mutant forms of the
Escherichia coli β sliding clamp that distinguish between its roles in
replication and DNA polymerase V-dependent translesionDNAsynthesis
Mol. Microbiol. 2005, 55, 1751–1766.
’ ACKNOWLEDGMENT
This work was supported in part by grants from the Australian
Research Council. N.E.D and A.J.O hold Australian Professorial
and Future Fellowships, respectively. P.A.J. was a CSIRO Fellow.
(18) Dalrymple, B. P.; Kongsuwan, K.; Wijffels, G.; Dixon, N. E.;
Jennings, P. A. A universal protein-protein interaction motif in the
eubacterial DNA replication and repair systems. Proc. Natl. Acad. Sci.
U.S.A. 2001, 98, 11627–11632.
(19) Wijffels, G.; Dalrymple, B. P.; Prosselkov, P.; Kongsuwan, K.;
Epa, V. C.; Lilley, P. E.; Jergic, S.; Buchardt, J.; Brown, S. E.; Alewood,
P. F.; Jennings, P. A.; Dixon, N. E. Inhibition of protein interactions with
the β2 sliding clamp of Escherichia coli DNA polymerase III by peptides
from β2-binding proteins. Biochemistry 2004, 43, 5661–5671.
(20) Dohrmann, P. R.; McHenry, C. S. A bipartite polymerase-
processivity factor interaction: Only the internal β binding site of the R
subunit is required for processive replication by the DNA polymerase III
holoenzyme. J. Mol. Biol. 2005, 350, 228–239.
’ ABBREVIATIONS USED
CBM, clamp binding motif; SPR, surface plasmon resonance;
NCI, National Cancer Institute; PDB, Protein Data Bank; DIPEA,
N,N-diisopropylethylamine
’ REFERENCES
(1) Robinson, A.; Brzoska, A. J.; Turner, K. M.; Withers, R.; Harry,
E. J.; Lewis, P. J.; Dixon, N. E. Essential biological processes of an
emerging pathogen: DNA replication, transcription and cell division in
Acinetobacter spp. Microbiol. Mol. Biol. Rev. 2010, 74, 273–297.
(2) Robinson, A.; Causer, R. J.; Dixon, N. E. Architecture and
conservation of the bacterial DNA replication machinery, an under-
exploited drug target. Curr. Drug Targets 2011in press.
(21) Lꢀopez de Saro, F. J.; Georgescu, R. E.; O’Donnell, M. A peptide
switch regulates DNA polymerase processivity. Proc. Natl. Acad. Sci.
U.S.A. 2003, 100, 14689–14694.
(22) Bunting, K. A.; Roe, S. M.; Pearl, L. H. Structural basis for
recruitment of translesion DNA polymerase Pol IV/DinB to the β-clamp.
EMBO J. 2003, 22, 5883–5892.
(23) Burnouf, D. Y.; Olieric, V.; Wagner, J.; Fujii, S.; Reinbolt, J.;
Fuchs, R. P. P.; Dumas, P. Structural and biochemical analysis of sliding
clamp/ligand interactions suggest a competition between replicative and
translesion DNA polymerases. J. Mol. Biol. 2004, 335, 1187–1197.
(24) Ihlenfeldt, W.-D.; Voigt, J. H.; Bienfait, B.; Oellien, F.; Nicklaus,
M. C. Enhanced CACTVS browser of the open NCI database. J. Chem.
Inf. Comput. Sci. 2002, 42, 46–57.
(3) Dallmann, H. G.; Fackelmayer, O. J.; Tomer, G.; Chen, J.;
Wiktor-Becker, A.; Ferrara, T.; Pope, C.; Oliveira, M. T.; Burgers,
P. M. J.; Kaguni, L. S.; McHenry, C. S. Parallel multiplicative target
screening against divergent bacterial replicases: Identification of specific
inhibitors with broad spectrum potential. Biochemistry 2010, 49,
2551–2562.
(4) Daly, J. S.; Giehl, T. J.; Brown, N. C.; Zhi, C.; Wright, G. E.;
Ellison, R. T., III. In vitro antimicrobial activities of novel anilinouracils
which selectively inhibit DNA polymerase III of Gram-positive bacteria.
Antimicrob. Agents Chemother. 2000, 44, 2217–2221.
(25) Oakley, A. J.; Prosselkov, P.; Wijffels, G.; Beck, J. L.; Wilce,
M. C. J.; Dixon, N. E. Flexibility revealed by the 1.85 Å crystal structure
of the β sliding-clamp subunit of Escherichia coli DNA polymerase III.
Acta Crystallogr., Sect. D 2003, 59, 1192–1199.
(26) Jeruzalmi, D.; Yurieva, O.; Zhao, Y.; Young, M.; Stewart, J.;
Hingorani, M.; O’Donnell, M.; Kuriyan, J. Mechanism of processivity
clamp opening by the delta subunit wrench of the clamp loader complex
of E. coli DNA polymerase III. Cell 2001, 106, 417–428.
(5) Georgescu, R. E.; Yurieva, O.; Kim, S.-S.; Kuriyan, J.; Kong, X.-P.;
O’Donnell, M. Structure of a small-molecule inhibitor of a DNA poly-
merase sliding clamp. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 11116–11121.
(6) Guiles, J.; Sun, X.; Critchley, I. A.; Ochsner, U.; Tregay, M.;
Stone, K.; Bertino, J.; Green, L.; Sabin, R.; Dean, F.; Dallmann, H. G.;
McHenry, C. S.; Janjic, N. Quinazolin-2-ylamino-quinazolin-4-ols as
novel non-nucleoside inhibitors of bacterial DNA polymerase III. Bioorg.
Med. Chem. Lett. 2009, 19, 800–802.
4837
dx.doi.org/10.1021/jm2004333 |J. Med. Chem. 2011, 54, 4831–4838

Journal of Medicinal Chemistry
ARTICLE
(27) Otwinowski, Z.; Minor, W. Processing of X-ray diffraction
data collected in oscillation mode. Methods Enzymol. 1997, 276,
307–326.
(28) Murshudov, G. N.; Vagin, A. A.; Lebedev, A.; Wilson, K. S.;
Dodson, E. J. Efficient anisotropic refinement of macromolecular struc-
tures using FFT. Acta Crystallogr., Sect. D 1999, 55, 247–255.
(29) Jones, T. A.; Zou, J. Y.; Cowan, S. W.; Kjeldgaard, M. Improved
methods for building protein models in electron-density maps and the
location of errors in these models. Acta Crystallogr., Sect. A 1991, 47,
110–119.
4838
dx.doi.org/10.1021/jm2004333 |J. Med. Chem. 2011, 54, 4831–4838