Dipicolinic Acid Derivatives as Inhibitors of New Delhi Metallo-β-lactamase-1

Article abstract of DOI:10.1021/acs.jmedchem.7b00407

The efficacy of β-lactam antibiotics is threatened by the emergence and global spread of metallo-β-lactamase (MBL) mediated resistance, specifically New Delhi metallo-β-lactamase-1 (NDM-1). By utilization of fragment-based drug discovery (FBDD), a new class of inhibitors for NDM-1 and two related β-lactamases, IMP-1 and VIM-2, was identified. On the basis of 2,6-dipicolinic acid (DPA), several libraries were synthesized for structure-activity relationship (SAR) analysis. Inhibitor 36 (IC₅₀ = 80 nM) was identified to be highly selective for MBLs when compared to other Zn(II) metalloenzymes. While DPA displayed a propensity to chelate metal ions from NDM-1, 36 formed a stable NDM-1:Zn(II):inhibitor ternary complex, as demonstrated by ¹H NMR, electron paramagnetic resonance (EPR) spectroscopy, equilibrium dialysis, intrinsic tryptophan fluorescence emission, and UV-vis spectroscopy. When coadministered with 36 (at concentrations nontoxic to mammalian cells), the minimum inhibitory concentrations (MICs) of imipenem against clinical isolates of Eschericia coli and Klebsiella pneumoniae harboring NDM-1 were reduced to susceptible levels.

Full text of DOI:10.1021/acs.jmedchem.7b00407

Subscriber access provided by Georgetown University | Lauinger and Blommer Libraries
Article
Dipicolinic Acid Derivatives as Inhibitors of New Delhi Metallo-#-lactamase-1
Allie Y. Chen, Pei W. Thomas, Alesha C. Stewart, Alexander Bergstrom, Zishuo Cheng, Callie
Miller, Christopher R. Bethel, Steven H. Marshall, Cy V. Credille, Christopher L Riley, Richard C.
Page, Robert A. Bonomo, Michael W. Crowder, David L. Tierney, Walter Fast, and Seth M. Cohen
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.7b00407 • Publication Date (Web): 15 Aug 2017
Downloaded from http://pubs.acs.org on August 15, 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 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Dipicolinic Acid Derivatives as Inhibitors of New Delhi
Metalloꢀβꢀlactamaseꢀ1
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
1
2
2
3
3
Allie Y. Chen , Pei W. Thomas , Alesha C. Stewart , Alexander Bergstrom , Zishuo Cheng ,
3
4
4
1
Callie Miller , Christopher R. Bethel , Steven H. Marshall , Cy V. Credille , Christopher L.
5
3
4,6*
3*
3*
Riley , Richard C. Page , Robert A. Bonomo , Michael W. Crowder , David L. Tierney ,
2
*
1*
Walter Fast , and Seth M. Cohen
1
Department of Chemistry and Biochemistry, University of California, San Diego, La Jolla, CA
92093, United States
2
Division of Chemical Biology & Medicinal Chemistry, College of Pharmacy, University of
Texas, Austin, TX 78712, United States
3
Department of Chemistry and Biochemistry, Miami University, Oxford, OH 45056, United
States
4
Research Services, Louis Stokes Cleveland Department of Veterans Affairs Medical Center,
Cleveland, OH 44106, United States
5
Department of Molecular Biosciences, University of Texas, Austin, TX 78712, United States
6
Departments of Medicine, Molecular Biology and Microbiology, Biochemistry, and
Pharmacology, Case Western Reserve University, Cleveland, OH 44106, United States
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 65
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
ABSTRACT. The efficacy of βꢀlactam antibiotics is threatened by the emergence and global
spread of metalloꢀβꢀlactamaseꢀ(MBL) mediated resistance, specifically New DelhiꢀMetalloꢀβꢀ
lactamaseꢀ1 (NDMꢀ1). Utilizing fragmentꢀbased drug discovery (FBDD), a new class of
inhibitors for NDMꢀ1 and two related βꢀlactamases, IMPꢀ1 and VIMꢀ2, was identified. Based on
2,6ꢀdipicolinic acid (DPA), several libraries were synthesized for structureꢀactivity relationship
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
(SAR) analysis. Inhibitor 36 (IC50 = 80 nM) was identified to be highly selective for MBLs
when compared to other Zn(II) metalloenzymes. While DPA displayed a propensity to chelate
metal ions from NDMꢀ1, 36 formed a stable NDMꢀ1:Zn(II):inhibitor ternary complex, as
1
demonstrated by H NMR, electron paramagnetic resonance (EPR) spectroscopy, equilibrium
dialysis, intrinsic tryptophan fluorescence emission, and UVꢀVis spectroscopy. When coꢀ
administered with 36 (at concentrations nonꢀtoxic to mammalian cells), the minimum inhibitory
concentration (MIC) of imipenem against clinical isolates of Eschericia coli and Klebsiella
pneumoniae harboring NDMꢀ1 were reduced to susceptible levels.
ACS Paragon Plus Environment

Page 3 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
Antibioticꢀresistant in pathogenic bacteria has become a critical public health threat. A
major mechanism of antibiotic resistance is microbial degradation of drugs by enzymes such as
2
βꢀlactamases. Three classes of βꢀlactamases (A, C, and D) utilize an activeꢀsite serine in
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
covalent mechanisms that can be targeted by βꢀlactamase inhibitors coꢀformulated with βꢀlactam
drugs. In contrast, class B consists of metalloꢀβꢀlactamases (MBLs) that utilize one or two active
3
site Zn(II) ion(s) to catalyze the hydrolysis of the βꢀlactam ring. MBLs have a broad substrate
scope that enables hydrolysis of bicyclic βꢀlactams, but due to MBLs employing a different
reaction mechanism, they evade the current βꢀlactamase inhibitors used in approved drug
combinations. Thus, MBLs can confer resistance to the entire category of βꢀlactam drugs, except
4
,5
monobactams.
MBLs with the greatest clinical impact are found in the B1 subfamily, with
New Delhi MetalloꢀβꢀLactamaseꢀ1 (NDMꢀ1), Verona Integrinꢀencoded MBL (VIMꢀ2), and
Imipenemase (IMPꢀ1) serving as examples. Of these, NDMꢀ1 represents the most troubling
development due to its carriage on easily transmissible plasmids and the emergence of
communityꢀacquired infections in addition to nosocomial infections typically found carrying
5ꢀ8
other MBLs.
Despite ongoing efforts to identify NDMꢀ1 inhibitors, approved drugs to counter its
activity are not yet available. Development of NDMꢀ1 inhibitors has been slow due to a number
9
of factors, including a shallow, relatively featureless active site that has is difficult to target,
concerns about a lack of inhibitor selectivity leading to in vivo metal stripping or offꢀtarget
10ꢀ12
inhibitory activity,
and a lack of novel scaffolds that selectively target the active site. Three
1
3
examples of NDMꢀ1 inhibitors are presented in Table 1. The hypertension drug
Lꢀcaptopril
1
4ꢀ16
represents the repurposing of this sulfhydrylꢀcontaining inhibitor.
Both Dꢀand Lꢀcaptopril
have been reported as inhibitors of NDMꢀ1, with IC50 values of ~8 ꢁM and ~200 ꢁM,
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 65
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
13,14
respectively.
a βꢀbridging fashion between the Zn(II) ions, displacing the catalytic hydroxide anion.
Although a crystal structure of the more potent ꢀcaptopril bound to NDMꢀ1 is not available,
analysis shows that the thiol of each isomer likely interacts with the dinuclear Zn(II) center in a
A crystal structure of Lꢀcaptopril bound to NDMꢀ1 reveals the thiolate bound in
1
3
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
1
3,17
similar manner,
and the carboxylate of the Dꢀisomer may mimic the carboxylate conserved in
most βꢀlactam substrates. However, adverse effects associated with thiolꢀcontaining compounds
1
8,19
may present a therapeutic liability for captopril and its derivatives.
A second example of a NDMꢀ1 inhibitor is aspergillomarasmine A (AMA), a fungal
12
natural product identified through cellꢀbased screening (IC50 value = 4.0 µM). Combination
therapy of AMA with meropenem was effective in microbial culture and in mice infected with a
lethal dose of NDMꢀ1ꢀpositive K. pneumoniae. However, the structural similarity of AMA to
metal chelators such as ethylenediaminetetraacetic acid (EDTA) raises the possibility of offꢀ
target effects. This is consistent with the finding that AMA inhibits NDMꢀ1 through a metalꢀ
1
2
chelation/stripping mechanism.
A third example is the cyclic boronate shown in Table 1. This and related compounds
exhibit excellent activity against NDMꢀ1 (IC = 4 nM) as well as VIMꢀ2, IMPꢀ1, and Bacillus
5
0
2
0
cereus MBL (BcII). Investigation of these cyclic boronates as inhibitors of B1 MBLs reveal a
binding mode that mimics substrate binding, including the formation of a tetrahedral complex.
The similarity of these inhibitors to a βꢀlactam tetrahedral intermediate transition state may
underlie the impressive activity of these compounds. More comprehensive reviews of MBL
2
1ꢀ23
inhibitors are provided elsewhere,
but these three examples illustrate some successful
strategies and pitfalls in developing effective inhibitors for this class of enzymes.
ACS Paragon Plus Environment

Page 5 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Table 1. Examples of Select NDMꢀ1 Inhibitors.
Structure
Name
IC50
1
4
D
ꢀCaptopril
7.9 µM
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
14
L
ꢀCaptopril
202.0 µM
Aspergillomarasmine A
(AMA)
1
2
4
.0 µM
Diaminoꢀsubstituted cyclic
boronate
2
0
4
nM
2
4
Herein, we report a fragmentꢀbased drug discovery (FBDD) strategy for the design,
synthesis, and evaluation of a novel class of NDMꢀ1 inhibitors. A focused library of metalꢀ
binding pharmacophores (MBPs) enabled a systematic search of compounds that target the
dinuclear Zn(II) site of B1 MBLs. SAR analysis and optimization of primary hit 2,6ꢀdipicolinic
acid (DPA) led to the discovery of 36 with an IC value of ~80 nM, putting it on par with the
50
2
0
most potent NDMꢀ1 inhibitors presently reported in the literature.
Compound 36 showed
increase inhibition against IMPꢀ1 and VIMꢀ2 when compared to DPA, and no significant
inhibition against several other Zn(II)ꢀdependent metalloenzymes. The interaction of 36 with the
1
active site of diꢀZn(II) or diꢀCo(II) metalloforms of NDMꢀ1 was characterized via H NMR
spectroscopy, electron paramagnetic resonance (EPR) spectroscopy, equilibrium dialysis,
intrinsic tryptophan fluorescence quenching, and UVꢀVis spectroscopy. As inhibition of the lead
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 65
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 was improved, a metal ion stripping mechanism was replaced in favor of a ternary
NDMꢀ1:Zn(II):inhibitor complex mechanism of action. Microdilution broth studies performed
with clinical E. coli and K. pneumoniae strains harboring blaNDMꢀ1 reveal that coꢀadministration
with 36 significantly reduced the minimum inhibitory concentration (MIC) values of imipenem.
Furthermore, no toxicity was observed for 36 against a mammalian cell culture.
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 and Discussion
Screening of MBP Library and Selection of DPA for Optimization
The extreme sequence diversity found within the family of MBL proteins complicates the
design of a single compound to serve as an effective broad spectrum MBL inhibitor. Even
within the B1 subclass of MBLs, which contains the most clinically relevant targets, sequence
identity is low. Pairwise protein sequence alignments between all combinations of the three
9
,25
MBLs tested here (NDMꢀ1, IMPꢀ1, and VIMꢀ2) have identities of approximately only 30%.
This diversity is also present at the active site where very few residues are essential for catalysis
9
due to a mechanism that relies mostly on substrate interactions with the diꢀZn(II) site. However,
this same property makes the diꢀZn(II) active site an attractive target for inhibition by MBPs.
MBPs are small molecular fragments that are known to have the capacity to bind metal ions in
the active site of metalloproteins. The Cohen laboratory has assembled a MBP library consisting
of ~300 fragments and this focused library has been show useful in the development of other
26ꢀ31
metalloprotein inhibitors, including antibacterial virulence factors.
Here, we search for MBL
inhibitors through a more systematic exploration of chemical space by screening the MBP library
against three representative B1 MBLs (NDMꢀ1, IMPꢀ1, and VIMꢀ2). Several MBPs were
identified as attractive leads for MBL inhibitor development.
ACS Paragon Plus Environment

Page 7 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
A 96ꢀwell plate, continuous assay was optimized for the hydrolysis of the colorimetric βꢀ
lactam substrate chromacef by NDMꢀ1 (substrate concentration = 3×K ), IMPꢀ1 (substrate
M
concentration = K ), and VIMꢀ2 (substrate concentration = K ). The substrate concentrations
M
M
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
were kept close to K values to better detect competitive inhibitors, yet high enough to allow for
M
32
maintenance of linear initial rates. The assays showed good discrimination between positive
(enzyme added) and negative (enzyme omitted) samples, with Z' factors of 0.7, 0.7, and 0.8 for
3
3
NDMꢀ1, IMPꢀ1, and VIMꢀ2, respectively (Figure S1). Each fragment in the MBP library was
screened at a concentration of 50 µM against each of the three MBL enzymes and the resulting
percent inhibition was measured. The activity of MBPs ranged from 0 ꢀ 100% inhibition, with
an average of ~27% inhibition for the entire library screened against all three enzymes. Using a
cutoff of >60% inhibition, six compounds remained as hits (Figure 1). One primary hit, 1,10ꢀ
phenanthroline, is a known metal stripping agent and was omitted. Another hit, 2ꢀ(pyridineꢀ2ꢀ
yl)ꢀ4,5ꢀdihydrothiazole, was disregarded because mass spectrometry indicated the stock solution
of the MBP had degraded (data not shown). The remaining four primary hits consisted of: 3ꢀ
hydroxypyridineꢀ2(1H)ꢀthione, which shares some structural similarities to the known MBL
3
4
inhibitor thiomaltol; two 8ꢀhydroxyquinoline compounds, shown to inhibit a different diꢀZn(II)
3
5
enzyme, and 2,6ꢀdipicolinic acid (DPA). DPA was selected as the initial scaffold for
optimization because previous reports describe dicarboxylate compounds and 2ꢀ
9
,36
carboxypyridines as inhibitors of other MBLs.
Furthermore, DPA bears resemblance to the
core of a hydrolyzed cephalosporin, mimicking both the carboxylate formed upon βꢀlactam
hydrolysis and the leaving group nitrogen. The inhibition of DPA was first determined via
colorimetric assay to have an IC = 520 nM.
5
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 65
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 1. Primary screening hits observed to inhibit all three B1 MBLs tested: NDMꢀ1, IMPꢀ1,
and VIMꢀ2. See Table S1 for detailed percent inhibition values.
Design and Synthesis of DPA Derivatives
To augment the inherent affinity of DPA for the Zn(II) component of MBLs, we sought
to design DPAꢀbased inhibitors of NDMꢀ1 that focus on increasing the binding interactions with
protein residues surrounding the activeꢀsite Zn(II) ions. The active site of NDMꢀ1 (Figure 2)
includes a mobile βꢀhairpin loop, which moves over the Zn(II) ions and back to the original
3
7
position in the millisecond timescale. This βꢀhairpin loop likely plays a role in ligand binding
and/or catalysis, in part by presenting a flexible hydrophobic surface as one wall of the active
site. Flanking the diꢀZn(II) active site are two sides of a shallow, relatively featureless
hydrophobic channel. This channel has a few notable residues available as potential binding
partners, Asn220, Gln123, and Lys211. Based on the structure of the coordination complex of
3
8
DPA with free Zn(II), we assumed a similar coordination at the active site of NDMꢀ1 to guide
inhibitor optimization. The design and synthesis of subsequent sublibraries aimed to improve
interactions with amino acid residues surrounding the diꢀZn(II) site.
ACS Paragon Plus Environment

Page 9 of 65
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 2. Activeꢀsite pocket of NDMꢀ1. A ribbon diagram of NDMꢀ1 covered by a transparent
surface depicts a shallow hydrophobic channel flanking the diꢀZn(II) site (gray spheres).
Selected residues are shown in stick form, including the residues ligating the Zn(II) ions (gray),
hydrophobic residues in the neighboring βꢀhairpin loop (green), residues with the potential to
interact with an inhibitor via hydrogenꢀbonding (purple and orange), and a conserved Lys211
(
yellow). All heteroatoms are colored by type (blue for nitrogen, red for oxygen, gold for
sulfur). The figure was prepared with coordinates from Protein Data Bank accession code
3Q6X, chain A, omitting the hydrolyzed ampicillin product.
The first of three sublibraries contained compounds with substituents designed to extend
into the hydrophobic channel, as well as determine the necessity of both carboxylate groups on
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 65
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
the DPA fragment for inhibition. These substitutions were introduced through the conversion of
one of the carboxylate groups of DPA to various amide substituents via standard peptide
coupling procedures (Compounds 3ꢀ20, Scheme 1). The carboxylic acids of DPA were first
esterified to give the dimethyl ester derivative 1, then hydrolyzed with KOH at 0 °C to obtain 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
monomethyl ester 2 product.
Coupling 2 with primary amines by 1ꢀethylꢀ3ꢀ(3ꢀ
dimethylaminopropyl)carbodiimide (EDC) and hydroxybenzotriazole (HOBT) afford the desired
amide derivatives. Finally, hydrolysis of the methyl ester in 1M NaOH/ THF produced the
monoꢀcarboxylic acid DPA derivatives that comprise sublibrary 1 (3-20).
Scheme 1. Synthetic scheme for sublibrary 1.
(
a) MeOH, H SO (cat.), 75 °C, 24 h, 100%; (b) MeOH, KOH, 0 °C, 4 h, 81%; (c) EDC, HOBT,
2 4
H NR, 18 h; then 4:1 1 M NaOH:THF, 30 min – 1 h, 4 M HCl, 34ꢀ87%.
2
The second sublibrary (Compounds 22ꢀ30) consisted of DPA derivatives with
substituents at the 4ꢀposition of the pyridine ring connected through an amine linker (Scheme 2).
This sublibrary was designed to determine what substituents could be tolerated in the active site
pocket below the βꢀhairpin loop. Compounds 22-30 were synthesized starting from dimethyl 4ꢀ
chloropyridineꢀ2,6ꢀdicarboxylate followed by saponification in a 1 M NaOH/THF solution to
ACS Paragon Plus Environment

Page 11 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
afford 21. Next, 21 was heated using a microwave reactor in the presence of the corresponding
amine to generate compounds 22-30.
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
Scheme 2. Synthetic scheme for sublibrary 2.
(a) 10:1 1 M NaOH:THF, 70 °C, 3 h, 80%; (b) amine, H O, microwave at 160 °C, 30 min, 10ꢀ
2
80%.
Lastly, the third sublibrary (Compounds 33ꢀ47) also contained DPA derivatives with
substituents at the 4ꢀposition, but incorporated a series of substituted aryl substituents to enhance
possible hydrophobic interactions at the base of the βꢀhairpin loop, and included substitutions to
make possible hydrogen bonding interactions with nearby residue side chains (e.g., Asn220,
Gln123). In this sublibrary, instead of an amine linker, the aryl substituents are attached through
a direct carbonꢀcarbon bond formed via Pdꢀcatalyzed Suzuki crossꢀcoupling (Scheme 3). 4ꢀ
Hydroxypyridineꢀ2,6ꢀdicarboxylic acid was esterified to 31 by heating to reflux in MeOH with
catalytic H SO .
Next, 31 was converted into the bromide derivative 32 using
2
4
tetrabutylammonium bromide (TBAB) and phosphorus pentoxide (P O ). Compounds 33ꢀ47
4
10
were then synthesized via Suzuki crossꢀcoupling procedures utilizing Pd(PPh3)4 and
K PO /CH CO K, with the final compounds obtained by saponification with 1 M NaOH/THF.
3
4
3
2
Scheme 3. Synthetic scheme for sublibrary 3.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 65
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(
a) MeOH, H SO (cat.), 75 °C, 24 h, 88%; (b)TBAB, P O , toluene, 100 °C, 3 h, 92%; (c)
2
4
4
10
Pd(PPh ) , K PO /CH CO K, 1,4ꢀdioxane, 85 °C, overnight; then 4:1 1 M NaOH:THF, H SO ,
3
4
3
4
3
2
2
4
22ꢀ84%.
ACS Paragon Plus Environment

Page 13 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Table 2A. IC50 Values of Sublibrary 1 Compounds (µM) Against NDMꢀ1. Values were
determined with chromacef at a substrate concentration of 3×KM.
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
Number Compound
IC (µM)
Number Compound
IC (µM)
5
0
50
3
> 50
> 50
12
24 ± 1
49 ± 1
4
5
13
14
> 50
> 50
22 ± 1
> 50
6
15
7
8
> 50
> 50
16
17
> 50
> 50
9
> 50
18
> 50
1
0
1
> 50
> 50
19
20
> 50
> 50
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 65
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
Table 2B. IC50 Values of Sublibrary 2 and Sublibrary 3 Compounds (µM) Against NDMꢀ1.
Values were determined with chromacef at a substrate concentration of 3×K . Values in bold
M
typeface were determined with fluorocillin at a substrate concentration <KM.
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
Number
Compound
IC (µM)
Number
Compound
IC (µM)
5
0
50
0
.52 ± 0.04
0.60 ± 0.03
DPA
35
0.41 ± 0.02
0.43 ± 0.01
0
.32 ± 0.01
2
1
2
5.0 ± 0.1
36
37
0.08 ± 0.002
0
.82 ± 0.01
2
13.6 ± 0.8
0.43 ± 0.01
0
.50 ± 0.02
2
3
4
3.4 ± 0.2
8.4 ± 1.1
38
39
0.50 ± 0.01
0.52 ± 0.02
2
0.13 ± 0.01
0
.43 ± 0.02
2
5
6
9.1 ± 0.2
4.5 ± 0.1
40
41
0.38 ± 0.02
0.66 ± 0.03
2
0.39 ± 0.02
ACS Paragon Plus Environment

Page 15 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
.03 ± 0.01
2
7
8
7.2 ± 0.4
5.8 ± 0.4
42
43
0.54 ± 0.02
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
0
.57 ± 0.02
2
0.19 ± 0.01
0.89 ± 0.03
29
5.2 ± 0.1
4.9 ± 0.1
0.99 ± 0.02
44
45
46
47
0.20 ± 0.08
0.99 ± 0.04
30
0.93 ± 0.02
0.86 ± 0.02
3
3
4
0.13 ± 0.01
0.80 ± 0.02
0.86 ± 0.02
3
1.07 ± 0.07
0.59 ± 0.06
In Vitro Activity of DPA Derivatives as NDM-1 Inhibitors
An initial ranking of NDMꢀ1 inhibition by DPA derivatives in sublibraries 1ꢀ3 was first
completed by determining the IC50 values using the colormetric substrate chromacef.
Chromacef, at a concentration of 3×K , allowed for linear hydrolysis rates over the assay
M
1
6,39
duration. A secondary assay, using the fluorescent substrate fluorocillin,
was then used to
verify ranking of the most potent compounds (sublibrary 3). By using a concentration less than
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 65
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
K (Supporting Information), the fluorescent substrate allowed for increased signal response, a
M
longer linear portion of the kinetic assay, and minimized substrate perturbation of IC values for
50
competitive inhibitors. Addition of a fluorescenceꢀbased assay also provides an orthogonal
method to verify inhibition. False positives in the screen arising from compounds with
absorbance at wavelengths similar to the product detected in the absorbanceꢀbased assay would
not interfere with the excitation or emission wavelengths used in the fluorescenceꢀbased assay.
IC values determined using either substrate give similar relative rankings of potency, but are 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
5
0
different absolute magnitudes due to differing assay conditions (most significantly substrate
32
concentration relative to K values). A summary of the inhibition data is provided in Table 2A
M
and 2B.
Utilizing a colorimetric assay, it was found that most of the compounds in sublibrary 1
(
3ꢀ20) showed ≥100ꢀfold loss in activity when compared to the unsubstituted DPA (IC = 520
50
nM). This indicates that both carboxylate groups on DPA are required for inhibition. Both large
8) and small (15) substitutions were detrimental; however, substituents that reintroduce a
potential metal coordinating ligand (12ꢀ14) retained some activity. DPA derivatized with
(
D
ꢀ
valine (14), in comparison with 13 and 16 suggests that a carboxylate at this position drives
activity, and that bulky hydrophobic substitutions can also improve inhibition, albeit to a lesser
degree. Based on this SAR, both carboxylates of the DPA scaffold were retained in the design of
subsequent sublibraries.
Compounds in sublibrary 2 (21-30) showed a modest loss in inhibition (7ꢀ to 26ꢀfold)
relative to DPA, with a small dimethylamine substituent (22) causing the most significant loss.
Addition of a hydrophobic aryl group to the nitrogen linker (as in compound 23) significantly
rescued activity and resulted in the most potent inhibitor in this series (IC50 = 3.4 µM).
ACS Paragon Plus Environment

Page 17 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Extending the length of the linker (23 vs. 24 or 25) reduced activity. Substitution of the aryl ring
(26ꢀ28) or replacement with different heterocycles (29ꢀ30) did not result in any significant
improvement. These results suggest that substitution at the 4ꢀposition by an amine linkage is not
favorable (even relative to unsubstituted DPA), but addition of hydrophobic substituents can
rescue activity. These results are consistent with the hypothesis that substitution at this position
can lead to interactions with the hydrophobic surface presented by the βꢀhairpin loop adjacent to
the active site (Figure 2). Therefore, the aryl substituent was kept in the design of the third
sublibrary, but a different linkage was utilized.
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
In contrast to the previous two sublibraries, compounds in sublibrary 3 (33ꢀ47) showed
inhibition of NDMꢀ1 that is on the same order of magnitude as DPA, and in some cases,
significantly better. Examination of the fluorescent assay results reveal that DPA with aryl
groups containing orthoꢀsubstituents (38ꢀ39) showed similar or improved inhibition relative to
DPA (IC = 410 nM). In general, those with paraꢀsubstituents (40ꢀ43) showed modest changes
50
in inhibition. The same general trend is seen with metaꢀsubstitutions (34ꢀ35, 37, 44ꢀ47), except
for compound 36. Compound 36 stood out in sublibrary 3, with an IC50 value of 80±2 nM
(Figure S2). This compound represents a 4ꢀfold improvement in inhibition compared to DPA,
and achieves one of the lowest IC values for NDMꢀ1 reported to date. Comparison of 36 with
5
0
42 suggests that the increase in inhibition is not likely due to the electron donating effects of 36
on metal binding, as a greater effect would be expected from the paraꢀsubstituted compound.
The substitution of a methoxy (34), alcohol (35), or dimethyl amine (37) group at the same
position shows a loss in activity, suggesting that specific interactions (possibly by hydrogen bond
donation) of 36 with the protein are responsible for the improved inhibition. It is proposed that
aryl substituents directly linked to the 4ꢀposition of DPA augmented by metaꢀ substituents are
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 65
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
capable of specific protein interactions, and represent an effective route for fragment growth.
Notably, addition of exogenous Zn(II) in the assay buffer has little impact on the IC value of
50
36, suggesting a specific, direct interaction of 36 with amino acid residues in the active site 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
NDMꢀ1. In contrast, exogenous Zn(II) in the assay buffer diminishes the inhibition activity of
DPA against NDMꢀ1, presumably due to sequestering of the excess Zn(II) by DPA (Figure S3).
In comparison with the most potent NDMꢀ1 inhibitor currently reported, a diaminoꢀ
2
0
substituted cyclic boronate (Table 1, IC = 4 nM), 36 is a smaller, achiral compound with a
5
0
4
0
higher ligand efficiency (0.5 versus 0.4 kcal/mol/nonꢀH atom). Furthermore, a review profiling
,200 antibacterial project compounds with whole cell activity found that active compounds
3
4
1
against Gramꢀnegative bacteria have an average of calculated log D (clogD pH 7.4) of <0. The
clogD7.4 of 36 (ꢀ5.19 ~ ꢀ4.48) falls into the typical hydrophobic region of such antibacterial
drugs. Structural determination of the ternary complex between 36 with NDMꢀ1 is in progress
and will be required to reveal the precise active site interactions, but molecular modeling
suggests that placement of the DPA scaffold at the dinuclear Zn(II) active site may position the
4ꢀaryl substituent near the hydrophobic base of the activeꢀsite βꢀhairpin loop, and situate the
metaꢀ substituent close to active site Asn220 (Figure S4). Molecular docking studies for
metalloproteins are rather challenging, hence we consider the model here as a rough guide until
more conclusive data on the binding of DPA and compound 36 are obtained.
Investigation of Inhibitor Selectivity
To determine the selectivity of 36 for MBLs, compound 36 was screened against two
other B1 MBLs and a panel of other clinically relevant Zn(II)ꢀdependent metalloenzymes.
ACS Paragon Plus Environment

Page 19 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Despite the sequence diversity among B1 MBLs, the elaboration of DPA to 36 also resulted in
enhancement of potency for inhibition of VIMꢀ2 and IMPꢀ1 (Table 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
Table 3. IC Values of Selected Compounds with B1 MBLs Using Fluorocillin as a Substrate.
5
0
IC (ꢁM)
50
Compound
NDMꢀ1
VIMꢀ2
IMPꢀ1
DPA
0.41 ± 0.02
1.66 ± 0.03
3.03 ± 0.04
36
0.080 ± 0.002
0.21 ± 0.01
0.24 ± 0.01
Inhibitor 36 was screened against a panel of Zn(II)ꢀdependent metalloenzymes, including
4
2
43
histone deacetylases (HDACs), matrix metalloproteinases (MMPs), and carbonic anhydrases
4
4
45,46
(
CAs) (Table 4). Screening protocols followed previously published procedures
and values
are reported as percent inhibition. At a concentration of 10 µM (significantly higher than the
IC value against NDMꢀ1) compound 36 showed some inhibition against MMPꢀ2 and MMPꢀ12,
5
0
but did not show inhibition against HDACꢀ1, HDACꢀ6, and human CAII. Testing 36 at higher
concentrations (50 – 100 µM), resulted in some inhibition of these metalloenzymes; however, the
inhibition values were <50% (data not shown). Based on the data obtained, compound 36 is
between 100ꢀ and 1000ꢀfold more selective for NDMꢀ1 than for MMPs, between 500ꢀ and 1000ꢀ
fold more selective for HDACs, and at least 500ꢀfold more selective than for human CAII.
Overall, 36 has strong activity against MBLs (NDMꢀ1, IMPꢀ1, VIMꢀ2) and no significant
inhibition of other Zn(II)ꢀdependent metalloenzymes (at concentrations ≥10 µM, demonstrating
the potential for 36 to be broad spectrum MBL inhibitor that does not exhibit offꢀtarget
inhibition.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 65
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
Table 4. Percent Inhibition of 36 (at 10 µM) Against a Panel of Zn(II) Metalloenzymes.
MMPꢀ2
<6%
MMPꢀ12
48±8%
HDACꢀ1
HDACꢀ6
hCAII
3
6
No inhibition
No inhibition
No inhibition
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
Mode of Inhibition
1
H NMR Spectroscopy. In previous work, we examined a series of M (II)ꢀsubstituted
1
,2
1
47,48
analogs of NDMꢀ1 (M1,2 = Co, Cd, Zn) by H NMR.
Comparing the CoCd and ZnCo
analogs of NDMꢀ1, we assigned several resonances in the rich spectrum displayed by the CoCo
enzyme (shown at the bottom of each panel in Figure 3). The solvent exchangeable resonances
at 78, 73, and 65 ppm are ascribed to the three Co(II)ꢀcoordinated histidines at the Zn site, while
1
the peak at 107 ppm is assigned to the NH proton from His250 coordinated to Co(II) in the Zn2
site. The sharp resonance at 47 ppm can be attributed to the βꢀCH pair of Asp124, coordinated
2
to Co(II) at the Zn site, while the poorly resolved doublet near 170 ppm arises from βꢀCH2
2
protons of the Co(II)ꢀbound cysteine at the same site. In addition, several poorly resolved
resonances attributed to secondary interactions between Co(II) in the Zn site and distant
2
residues in the substrate binding pocket, are seen in the range from +40 to ꢀ80 ppm.
With a paramagnet in both sites, and the ability to attribute many of the signals to either
the Zn or Zn site, we used these signatures to examine the metalꢀbinding properties of the
1
2
individual inhibitors. CoCoꢀNDMꢀ1 was titrated with increasing molar equivalents of DPA and
13
36, along with a known chelator (EDTA) and a known competitive inhibitor (
L
ꢀcaptopril) as
controls. Each titration was compared to spectra from an aqueous solution of the inhibitor and
Co(II), in the absence of protein (Figure S5). The NDMꢀ1 titrations with ꢀcaptopril and EDTA
L
gave results as expected, with
L
ꢀcaptopril forming a ternary complex (Figure S6A) and EDTA
ACS Paragon Plus Environment

Page 21 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
(
Figure S6B) indiscriminately removing metal. The titration with DPA (Figure 3A) shows
evidence of metal removal at 1 equivalent, with the appearance of four very sharp, intense
resonances at 102, 91, 37 and 30 ppm, consistent with the formation of the Co(DPA) adducts
n
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
seen in control spectra. However, the rest of the spectrum is largely unaffected, suggesting that
DPA indiscriminately removes both metal ions, and shows no changes that can be attributed to
an NDMꢀ1:Co(II):DPA ternary complex. Additional added DPA simply resulted in removal of
both metal ions in the NDMꢀ1 active site and further loss in signal intensity.
Titration of CoCoꢀNDMꢀ1 with the DPA derivative 36 (Figure 3B) shows some evidence
for metal ion removal; however, unlike DPA, 36 also shows evidence of a ternary complex
formation. Addition of 1 equivalent of 36 led to a set of sharp resonances that indicate Co(36)n
formation (Figure S5). Although it is not possible to directly compare the intensity of these
sharp lines to the broader lines of the Co(II) ligand protons, the relative intensity of these lines to
those from Co(DPA) at 1 equivalent of added DPA is significantly lower. Meanwhile, at least
n
two new resonances (red markers at ~125 and ~30 ppm) attributed with an NDMꢀ1:Co(II):36
ternary complex appears. At two equivalents of 36, substantial metal removal is apparent,
evidenced by loss of intensity particularly from the Zn site ligands, suggesting 36 shows some
2
preference for the formation of a ternary complex.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 65
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
1
Figure 3. 300 MHz H NMR titrations of CoCoꢀNDMꢀ1 with (A) DPA and (B) 36. Resonances
assigned to Zn site ligands are marked with black squares, while those assigned to Zn ligands
1
2
are marked with gray squares. Resonances indicative of a ternary complex are marked with red
squares.
EPR Spectroscopy. To gain a better understanding of the ternary complex formed, we
turned to EPR spectroscopy. As can be seen in Figure 4, the addition of 1 equivalent of DPA
and 36 (similar to
Lꢀcaptopril, data not shown) led to minimal perturbation of the spectra, aside
from minor modulation of the line shapes, when compared to the resting CoCoꢀNDMꢀ1 spectra.
The reason for small perturbations produced by
Lꢀcaptopril in the EPR spectra of NDMꢀ1 is
49
unknown, but consistent with similar previous findings. This suggests the electronic structure
of the Co(II) ions, and their physical structures, by extension, are largely unchanged. However,
ACS Paragon Plus Environment

Page 23 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
as shown in the inset, the addition of 36 led to a new feature near 600 G, which has been seen in
5
0
EPR studies of MBLs in the presence of substrates during turnover. This feature, indicative of
a more tightly coupled diꢀCo(II) site, is suggestive of an additional bridging species, arising from
weak ferromagnetic coupling of the Co(II) ions. This can be more readily seen in the parallel
mode spectra, which show a deep negative feature (~800 G) that appears on addition of 1
equivalent DPA and 36. This data indicates not only the formation of a ternary complex, but
bridging binding modes for both DPA and the synthesized inhibitor.
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 4. Xꢀband EPR spectroscopy of CoCoꢀNDMꢀ1 (black) and CoCoꢀNDMꢀ1 with 1
equivalent of added DPA (blue) and 36 (red). Standard perpendicular mode spectra are on top,
parallel mode spectra (scaled 3ꢀfold) are shown below. Inset: Expansion of the lowꢀfield region
of the perpendicular mode spectra.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 65
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
UV-Vis Spectroscopy. Previously, UVꢀvis spectroscopy has been used to probe Co(II)
binding to NDMꢀ1. The optical spectrum of CoCoꢀNDMꢀ1 presented here (Figure 5A) is
4
8
identical to that previously published, with ligand field transitions in the 500ꢀ700 nm region
from proteinꢀbound highꢀspin Co(II) (the small absorbance at 420 nm has previously been
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
5
1
attributed to the presence of Co(III)). Control spectra from buffered solutions of Co(II) in the
presence of 3 eq of EDTA, captopril, DPA, or 36 showed minimal absorbance in the ligand field
region (Figure S7), suggesting the presence or absence of the ligand field bands can serve as a
marker for proteinꢀbound Co(II). Titration of CoCoꢀNDMꢀ1 with 1 eq of EDTA results in a
small reduction of the intensities of all four ligand field transitions, suggesting removal of small
amounts of Co(II). A second equivalent of EDTA results in complete loss of the ligand field
transitions, indicating that EDTA fully removed Co(II) from the active site of NDMꢀ1 (Figure
5
A). Conversely, the titration of CoCoꢀNDMꢀ1 with competitive inhibitor Lꢀcaptopril (Figure
B) results in a change in the shape of the ligand field bands, but no loss in intensity. While the
5
inclusion of a second sulfur atom in the ligand set will surely complicate the interpretation of the
captopril spectra, the retention of dꢀd band intensity strongly suggests the formation of an NDMꢀ
1:Co(II):captopril ternary complex. Titration of CoCoꢀNDMꢀ1 with DPA was similar to that
with EDTA, with the addition of the first equivalent of DPA resulting in a reduction in the ligand
field intensity (Figure 5C), and a more dramatic reduction at 2 eq of DPA. While these data
indicate that Co(II) is being removed from the active site, the retention of some ligand field
52
intensity at 2 eq of DPA shows that DPA does not bind Co(II) as tightly as EDTA. Meanwhile,
titration of CoCoꢀNDMꢀ1 with 36 (Figure 5D) up to 2 eq shows minimal loss of dꢀd band
intensity, suggesting minimal removal of the active site metal ion. The only perturbation
ACS Paragon Plus Environment

Page 25 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
produced by 36 is a strong chargeꢀtransfer transition at ca. 320 nm (an internal πꢀπ* transition of
36) that alters the background associated with the ligand field region.
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 5. UVꢀVis spectroscopy of CoCoꢀNDMꢀ1 (black lines) and CoCoꢀNDMꢀ1 (300 µM)
with EDTA (A) and
D).
Lꢀcaptopril (B) as control, and 1ꢀ2 equivalents of added DPA (C), and 36
(
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 65
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
Equilibrium Dialysis. To be certain the mode of inhibition determined above is
representative of the diꢀZn(II) enzymes, equilibrium dialysis studies were conducted. Not
surprisingly, the Zn(II) content of NDMꢀ1 was significantly reduced when incubated with 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
metal chelator EDTA, while the Zn(II) content did not change upon incubation with
(Figure 6). Incubation of NDMꢀ1 with DPA resulted in a significant loss in Zn(II) content, albeit
not as much as EDTA at similar concentrations. Unlike DPA or EDTA, 36 behaves more like
Lꢀcaptopril
L
ꢀ
captopril. At low concentrations of 36, incubation with NDMꢀ1 resulted in no significant
removal of Zn(II); only upon exposure to higher concentrations of these compounds (~24 µM,
about 240ꢀfold greater than the IC50 value) was 10ꢀ20% of the Zn(II) removed from NDMꢀ1.
This result further supports that the mode of inhibition exhibited by 36 is through the formation
of a NDMꢀ1:Zn(II):inhibitor ternary complex, rather than metal stripping, even at micromolar
concentrations.
Figure 6. Metal content of NDMꢀ1 (8 µM) after dialysis with various concentrations of EDTA,
Lꢀcaptorpril, DPA, and 36 in 50 mM HEPES, pH 7.5 buffer.
ACS Paragon Plus Environment

Page 27 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Tryptophan Fluorescence. Previously, stoppedꢀflow fluorescence studies were used to
4
7,48
probe the kinetic mechanism of NDMꢀ1.
These studies revealed that the binding of substrate
results in a rapid decrease in fluorescence, and as substrate is hydrolyzed and product is released,
there is a rateꢀlimiting return of fluorescence intensity. We attributed these fluorescence
properties to Trp93, which is 5.7 Å from the Zn(II) binding site. As before, we utilized EDTA
and Lꢀcaptopril as controls to evaluate how a metal chelator versus a competitive inhibitor affects
the fluorescence properties of NDMꢀ1. Incubation of 2 µM NDMꢀ1 with up to 32 µM EDTA
resulted in <10% quenching of intrinsic tryptophan fluorescence (Figure 7A). Conversely,
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
incubation of NDMꢀ1 with up to 32 µM
Lꢀcaptopril resulted in a 10% increase in fluorescence,
indicating a different binding mode for this compound. The increase in fluorescence is likely
1
3
due to the interaction of C3 and C5 in Lꢀcaptopril with Trp93. Incubation of NDMꢀ1 with DPA
resulted in a 20% quenching of tryptophan fluorescence emission (at 32 µM DPA where 20% of
the Zn(II) was removed, Figure 6). In contrast, incubation of NDMꢀ1 with 36 resulted in a 50ꢀ
6
0% quenching of the intrinsic tryptophan fluorescence (Figure 7A). Due to the aromatic ring in
3
6, we attribute the quenching of fluorescence to be due to the binding of these compounds to the
metal center, resulting in a change in the fluorescence properties of Trp93. In support of this
hypothesis, a control titration of ꢀtryptophan with EDTA, Lꢀcaptopril, and 36 was performed.
The titration of ꢀtryptophan with EDTA and ꢀcaptopril results in no change in the fluorescence
L
L
L
emission of tryptophan, while titration with 36 results in a 50% reduction of fluorescence
emission (Figure 7B). The results from fluorescence titrations are consistent with those from
previous analysis, which indicate compound 36 is binding in the active site, forming a ternary
NDMꢀ1:Zn(II):inhibitor complex.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 65
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
(
A)
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
(B)
Figure 7. (A) Relative intrinsic tryptophan fluorescence emission versus concentration of
inhibitor; (B) Relative intrinsic tryptophan fluorescence emission versus concentration of 36,
L
ꢀ
captopril and EDTA. The concentration of NDMꢀ1 was 2 µM, Lꢀtryptophan was 2 µM and the
buffer for both studies was 50 mM HEPES, pH 7.5.
Microdilution Broth Minimum Inhibitory Concentrations (MICs)
Microdilution broth MICs were next performed on E. coli (Table 5) and K. pneumoniae
53
(
Table 6) clinical isolates expressing blaNDMꢀ1 using imipenem and compound 36. The MICs of
imipenem alone were 4 ꢀ 16 mg/L. Notably, the addition of compound 36 at 100 mg/L
ACS Paragon Plus Environment

Page 29 of 65
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
significantly lowered the imipenem MICs (0.5 ꢀ 1 mg/L). For all E. coli and K. pneumoniae
strains, no effect on growth was observed with compound 36 alone, indicating a lack of toxicity
at these concentrations. Even though standard CLSI guidelines have not yet been established for
this imipenem:inhibitor combination, the reduction in MICs from 8 ꢀ 16 mg/L to 0.5 ꢀ 1 mg/L for
the clinical strains would represent a reduction in MICs from the “resistant” to “susceptible”
range. These results serve as an important proof of concept that these exploratory scaffolds are
critical starting points for novel design strategies. Noteworthy, the βꢀlactam resistance profile in
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
the selected strains is mediated by both extendedꢀspectrum βꢀlactamases (ESBLs) (blaCTXꢀMꢀ15
)
and plasmidꢀmediated AmpCs (blaCMYꢀ2). A therapeutic combination of imipenem and 36 able
to overcome the complex background of a metalloꢀβꢀlactamase (MBL), ESBL, and plasmidꢀ
mediated AmpC is notable. Furthermore, growth and morphology of cultured human HEK293
cells were also not adversely effected by 36 at concentrations relevant to the MIC studies
(Figures S7, S8).
Table 5. Microdilution Broth MICs of Clinical E. coli Isolates Expressing blaNDMꢀ1
.
MICs (mg/L)
E. coli
Imipenem
Imipenem + 36
Isolates
Ch8.68
16
4
1
Ch8.69
0.5
Ch8.70
Ah8.71
Ch8.72
Ah8.73
Ah8.74
16
16
16
16
8
0.5
1
0.5
0.5
0.5
ACS Paragon Plus Environment