Heteroaryl Phosphonates as Noncovalent Inhibitors of Both Serine- and Metallocarbapenemases

Article abstract of DOI:10.1021/acs.jmedchem.9b00728

Gram-negative pathogens expressing serine β-lactamases (SBLs) and metallo-β-lactamases (MBLs), especially those with carbapenemase activity, threaten the clinical utility of almost all β-lactam antibiotics. Here we describe the discovery of a heteroaryl phosphonate scaffold that exhibits noncovalent cross-class inhibition of representative carbapenemases, specifically the SBL KPC-2 and the MBLs NDM-1 and VIM-2. The most potent lead, compound 16, exhibited low nM to low μM inhibition of KPC-2, NDM-1, and VIM-2. Compound 16 potentiated imipenem efficacy against resistant clinical and laboratory bacterial strains expressing carbapenemases while showing some cytotoxicity toward human HEK293T cells only at concentrations above 100 μg/mL. Complex structures with KPC-2, NDM-1, and VIM-2 demonstrate how these inhibitors achieve high binding affinity to both enzyme classes. These findings provide a structurally and mechanistically new scaffold for drug discovery targeting multidrug resistant Gram-negative pathogens and more generally highlight the active site features of carbapenemases that can be leveraged for lead discovery.

Full text of DOI:10.1021/acs.jmedchem.9b00728

Subscriber access provided by Nottingham Trent University
Article
Heteroaryl phosphonates as non-covalent inhibitors
of both serine and metallo-carbapenemases
Orville Antonio Pemberton, Priyadarshini Jaishankar, Afroza Akhtar,
Jessie L Adams, Lindsey N. Shaw, Adam R. Renslo, and Yu Chen
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.9b00728 • Publication Date (Web): 04 Sep 2019
Downloaded from pubs.acs.org on September 5, 2019
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
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 53
Journal of Medicinal Chemistry
1
2
3
4
5
Heteroaryl phosphonates as non-covalent inhibitors of both serine-
and metallo-carbapenemases
6
7
8
9
Orville A. Pemberton1, Priyadarshini Jaishankar2, Afroza Akhtar1, Jessie L. Adams3, Lindsey N. Shaw3,
Adam R. Renslo2*, & Yu Chen1*
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
1Department of Molecular Medicine, University of South Florida Morsani College of Medicine, 12901
Bruce B. Downs Blvd, MDC 3522, Tampa, Florida 33612, United States
2Department of Pharmaceutical Chemistry and Small Molecule Discovery Center, University of California
San Francisco, 600 16th Street, Genentech Hall N574, San Francisco, California 94158, United States
3Department of Cell Biology, Microbiology & Molecular Biology, University of South Florida, 4202 E.
Fowler Ave., Tampa, Florida 33620, United States
*Address correspondence to Adam R. Renslo, adam.renslo@ucsf.edu; Yu Chen, ychen1@health.usf.edu
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 53
1
2
3
Abstract
4
5
6
7
8
9
Gram-negative pathogens expressing serine β-lactamases (SBLs) and metallo-β-lactamases (MBLs),
especially those with carbapenemase activity, threaten the clinical utility of almost all β-lactam antibiotics.
Here we describe the discovery of a heteroaryl phosphonate scaffold that exhibits non-covalent cross-
class inhibition of representative carbapenemases, specifically the SBL KPC-2 and the MBLs NDM-1 and
VIM-2. The most potent lead, compound 16, exhibited low nM to low µM inhibition of KPC-2, NDM-1, and
VIM-2. Compound 16 potentiated imipenem efficacy against resistant clinical and laboratory bacterial
strains expressing carbapenemases, while showing some cytotoxicity toward human HEK293T cells only
at concentrations above 100 µg/mL. Complex structures with KPC-2, NDM-1, and VIM-2 demonstrate how
these inhibitors achieve high binding affinity to both enzyme classes. These findings provide a structurally
and mechanistically new scaffold for drug discovery targeting multidrug resistant Gram-negative
pathogens, and more generally highlight the active site features of carbapenemases that can be leveraged
for lead discovery.
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
ACS Paragon Plus Environment

Page 3 of 53
Journal of Medicinal Chemistry
1
2
3
Introduction
4
5
6
7
8
9
The penicillins were among the first effective antibiotics, and current generation β-lactams such
as the cephalosporins and carbapenems remain crucial agents in the treatment of serious, life-
threatening bacterial infections1. Resistance to β-lactam antibiotics most commonly involves the
production of β-lactamase enzymes that hydrolyze the β-lactam ring2. There are four classes of β-
lactamase (A, B, C, and D), which differ from one another based on their amino acid sequence and
mechanism of action. Classes A, C, and D are serine β-lactamases (SBLs), whereas class B are metallo-β-
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
lactamases (MBLs) that use either one or two zinc ions to promote the hydrolysis reaction3, 4
.
Carbapenems (e.g., imipenem, meropenem, doripenem, and ertapenem) are often considered
to be “last-resort” antibiotics on account of their effectiveness in treating serious, multidrug resistant
infections5-7. Unfortunately, the recent emergence of carbapenem-resistant Enterobacteriaceae (CRE)
and multidrug resistant Pseudomonas aeruginosa threatens the utility of all currently available β-lactam
antibiotics. CRE and P. aeruginosa infections are associated with high rates of morbidity and mortality
worldwide due to limited treatment options8-10. Carbapenem resistance in CRE and P. aeruginosa can be
mediated by several factors, especially by the production of carbapenemases11-13. Carbapenemase
activity is exhibited by some class A and D SBLs and by all class B MBLs14, 15. Most serine carbapenemases
are versatile enzymes that can hydrolyze nearly all β-lactam substrates7, 14. While MBLs also have a
broad substrate spectrum, they are unable to catalyze the hydrolysis of monobactams4, 16, 17. Among
carbapenemases, of greatest concern currently are the class A SBL KPC-2 and the MBLs NDM-1 and VIM-
2, all of which are commonly isolated from CRE, P. aeruginosa, and other Gram-negative pathogens16.
A number of β-lactamase inhibitors have been developed to address the growing problem of β-
lactam antibiotic resistance, although currently marketed agents target only SBLs and not MBLs16, 18
Classical β-lactamase inhibitors like clavulanic acid are active against many class A SBLs but possess
limited activity against class C and D SBLs, and no activity against most serine or metallo-
.
carbapenemases16, 19. Newer, non-β-lactam derived compounds like avibactam20 cover an expanded
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 53
1
2
3
4
5
6
7
8
spectrum of class A, C, and D SBLs, including KPC-2, but are inactive against MBLs21, 22. Notwithstanding
some recent progress with boronic acid-based inhibitors23, 24, finding cross-class inhibitor leads remains a
significant and challenging problem25.
9
In this study, we used structure-based design, synthetic chemistry, and X-ray crystallography to
identify and explore common features of non-covalent small molecule binding by SBLs and MBLs. We
identified a heteroaryl phosphonate scaffold with low μM to low nM affinities against KPC-2, NDM-1,
and VIM-2, that also exhibits promising cell-based activity and good lead-like properties. Complex
crystal structures of several analogs have revealed underappreciated structural features that likely
underlie the broad substrate spectrum of carbapenemases. Taken together, these data offer a new
chemotype and actionable information to guide development of new agents to counter carbapenemase-
mediated resistance in pathogenic Gram-negative bacteria.
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
Results
A novel inhibitor chemotype for the KPC-2 serine carbapenemase
To develop novel inhibitor chemotypes, we sought to identify fragment- or lead-sized
compounds that bound with high ligand efficiency (L.E.) (defined as binding enthalpy per non-H atom)26
to the SBL KPC-2. The decision to start with KPC-2 was based on the expectation that ligand affinity for
KPC-2 would derive from non-covalent interactions with active-site residues, while screening against
MBLs could be dominated by chelation-driven interactions 27. We thus performed molecular docking of
KPC-2 using the ZINC database28 and identified several hits, including the coumarin-4-
methylphosphonate 1 that is the focus of the current study (Table 1).
Using a biochemical assay with the nitrocefin substrate, we observed 1 to be a moderately
potent and ligand-efficient inhibitor of KPC-2 (Ki = 32.9 µM; L.E. = 0.34 kcal mol-1 heavy atom-1). To probe
SAR around this hit, we identified five related commercial analogs (2-6) that differed in substitution at
positions C5-C7 of the coumarin ring. The potencies and ligand efficiencies of analogs 2-5 revealed that
ACS Paragon Plus Environment

Page 5 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
substitution at C7, as well as a lack of substitution at C6, were important for activity. Thus, regioisomers
4 and 5 bearing methyl groups at C7 or C6, respectively, exhibited Ki (and L.E.) values of 23.3 µM (0.38
kcal mol-1 HA-1) and 154.4 µM (0.31 kcal mol-1 HA-1) against KPC-2. Analogs 2 and 6, both unsubstituted
at C6, were the most potent and ligand efficient of the commercial analogs, with Ki values in the low-µM
regime and L.E. values of 0.39 and 0.45, respectively.
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
Having established an apparent preference for substitution at both C5 and C7 (as in compound
6), we evaluated additional, similarly substituted synthetic analogs. First, the steric and electronic
Table 1. Biochemical activities of phosphonate compounds against KPC-2, NDM-1 and VIM-2.
Compound a
Ki
Ki
Ki
L.E. a
(KPC-2)
X
Y
Z
R1
R2
R3
(KPC-2)
µM
(NDM-1)
µM
(VIM-2)
µM
32.9 ± 3.3
1
O
C=O
C
Me
Me
H
H
0.34
44.3 ± 7.3
2.1 ± 0.1
8.4 ± 0.4
15.3 ± 2.5
23.3 ± 4.6
154.4
2
3
O
O
O
O
O
C=O
C=O
C=O
C=O
C=O
C
C
C
C
C
OMe
H
H
0.39
0.35
0.38
0.31
0.45
123.7 ± 3.6
N/A b
2.1 ± 0.2
N/A
OCH2O
4
5c
Me
H
H
Me
H
H
N/A
N/A
H
N/A
N/A
6
Me
Me
1.4 ± 0.2
33.8 ± 3.8
0.82 ± 0.1
7
8
O
O
C=O
C=O
C=O
C=O
C=O
C
C
C
C
C
C
Me
F
H
H
H
H
H
H
OMe
F
11.9 ± 1.1
9.3 ± 1.8
0.36
0.39
0.49
0.39
0.33
0.32
N/A
N/A
No inhib
3.3 ± 0.1
1.2 ± 0.16
3.3 ± 0.03
6.4 ± 0.04
7.5 ± 0.02
9
O
Me
Me
H
Br
F
0.246 ± 0.047
8.5 ± 0.3
34.2 ± 8.0
230.2 ± 3.2
190.2 ± 22.6
81.0 ± 12.8
10
11
12
O
NH
H
142.3 ± 22.3
101.5 ± 8.6
NMe C=O
H
H
13
14
15
NH
C=O
C
N
C
Me
H
H
H
H
Me
-
2.2 ± 0.8
0.43
0.27
0.30
56.2 ± 4.3
741.3 ± 53.1
No inhib
3.4 ± 0.03
22.3 ± 1.3
30.3 ± 3.2
C=C
N=C
730.4 ± 87.4
326.1 ± 56.8
H
H
16
NH
C=O
C
Br
H
Br
0.020 ± 0.007
0.59
31.4 ± 1.9
0.316 ± 0.015
a L.E. = Ligand Efficiency in units of kcal mol-1 number of heavy atom-1 or kcal mol-1 HA-1
b N/A, not tested.
c Compound 5 was tested only once due to limited availability.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 53
1
2
3
4
5
6
7
8
9
character of the C5 substituent was modified systematically in analogs sharing a C7-Me group
(compounds 7, 9, 10). Increasing the size of the C5 substituent from –F to –CH3 to –Br improved potency
and ligand efficiency, with the C5-Br analog 9 providing a new benchmark in potency and ligand
efficiency (Ki = 0.246 µM; L.E. = 0.49 kcal mol-1 HA-1). With respect to the electronic nature of the
coumarin ring, the dimethyl substitution as in compound 6 was preferred over the more electron rich
(compound 7) or electron deficient (compound 8) coumarin analogs.
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
Next, we explored replacement of O1 of the coumarin ring with nitrogen to produce the 1,2-
dihydroquinolin-2-one (DHQ) analogs 11-13. This change was well tolerated, with 5,7-dimethyl-DHQ
analog 13 (Ki = 2.2 µM; L.E. = 0.43 kcal mol-1 HA-1) essentially equipotent to its coumarin congener 6.
Substitution on the ring N atom in the DHQ scaffold was also tolerated, with analogs 11 and 12
exhibiting comparable potencies and ligand efficiencies. The favorable effect of C5/C7 substitution
noted in the coumarin series is true as well in DHQ analogs, with dimethyl analog 13 notably more
potent than unsubstituted DHQ comparators 11 and 12. Quinoline analogs 14 and 15 lacking the ring
carbonyl of the DHQ scaffold exhibited very weak potencies and ligand efficiencies 2-5 times inferior to
the relevant DHQ comparator 11, suggesting that the carbonyl group is involved in recognition and
binding by KPC-2. Finally, the introduction of bromine substituents at C5 and C7 of the DHQ scaffold
produced compound 16, with activity ~7,000-fold improved over the parent DHQ analog 11, and an
impressive Ki of 20 nM against KPC-2 that is to our knowledge the highest affinity non-covalent inhibitor
of any SBL reported to date.
To define key residues and interactions that contribute to inhibitor binding by KPC-2, we
determined complex crystal structures with compounds 1, 2, 3, 6, and 9 (Fig. 1 and Fig. S2) at resolutions
of 1.45 Å and better. All ligands were unambiguously identified in the composite simulated annealing
ACS Paragon Plus Environment

Page 7 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
a
b
wat
wat
wat
H274
R220
H274
C238
R220
T235
T237
T235
T237
C69
wat
7
C238
8
9
C69
Compound 1
K234
Compound 6
K234
S130
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
T216
T216
S70
K73
S70
S130
E166
K73
W105
L167
L167
N132
E166
W105
P104
P104
d
c
H274
H274
C238
R220
R220
C238
T235
T237
C69
T235
T237
C69
Sulfate
Compound 1
E166
Compound 1
Compound 6
K234
T216
K234
S130
T216
S70
K73
S70
K73
S130
L167
N132
E166
N132
L167
W105
W105
P104
P104
Figure 1. Crystal structures reveal binding mode of coumarin phosphonate analogs 1 and 6 to KPC-2.
The composite simulated annealing mFo-DFc omit maps are colored in gray and contoured at 3 σ.
Water molecules are shown as red spheres. Black dashed lines represent hydrogen bonds. (a)
Compound 1 bound to KPC-2 (PDB ID 6D15). (b) Compound 6 bound to KPC-2 (PDB ID 6D18). (c)
Compound 1 complex (white and green) superimposed onto KPC-2 apo structure (purple, PDB ID
5UL8). (d) Compound 1 complex (white and green) superimposed onto compound 2 complex (purple).
mFo-DFc omit maps contoured at 3 σ. Interestingly, for the complex with compound 1, there appears to
be additional density corresponding to a potential second copy of 1; however, the occupancy was too
low to confidently model in. In each complex structure, the phosphonate moiety establishes hydrogen
bonds (HB) with S70, S130, T235, T237, and forms a water-mediated HB with the backbone carbonyl of
T216. Notably, the carbonyl oxygen atom of the inhibitor forms a water-mediated HB with R220 and
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 53
1
2
3
4
5
H274, perhaps revealing why quinoline analogs 14 and 15 that lack such carbonyl functionality are
weaker inhibitors of KPC-2.
6
7
8
9
Also revealed in these structures were some important details concerning the KPC-2 active site.
The unfavorable effect of C6-Me substitution is explained by a potential steric clash between the C6
methyl group and N132, which restricts the binding mode of 1 (Fig. 1a). Moving this substituent to the
C5 position as in 6 and 9 avoids the steric occlusion and allows the coumarin ring to swing towards
W105 (Fig. 1b). Concomitantly, a conformational change in W105 brings the side chain closer to the face
of the coumarin ring, forming a π-π stacking interaction. At the same time, the C5-Me substituent in
compounds 6 and 9 itself contributes to binding affinity by forming non-polar interactions with W105.
Regarding KPC-2 enzymology, the large hydrophobic surface of W105, coupled with a conformational
flexibility revealed in our complex and apo structures (Fig. 1c, 1d), appears to endow KPC-2 with
significant versatility in interacting with small molecule substrates and inhibitors.
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 SAR analysis and complex crystal structures also highlight the contribution to ligand binding
of a non-polar surface consisting of P104 and especially L167, which is in contact with the C7-Me or C7-
Br substitutions in 6, 9 and 16 (Fig. 1 and S2). Replacing the C7-Me in 9 by –Br in 16 increases the activity
by ~12-fold, through increasing the interactions with this non-polar surface.
Cross-class inhibition of metallo-β-lactamases
Having established that heteroaryl phosphonates are recognized by the active site of KPC-2, we
next asked whether the clinically relevant MBLs NDM-1 and VIM-2 might also recognize these
compounds, as SBLs and MBLs clearly do with their shared β-lactam substrates 29. We first tested a
subset of the coumarin, DHQ, and quinoline analogs for biochemical activity against NDM-1 and VIM-2.
Notably, inhibition of NDM-1 and VIM-2 was generally strongest for analogs with substitution at both C5
and C7, mirroring our findings with KPC-2 (Table 1). Thus, compounds 6, 8, 9, 13, and 16 were better
ACS Paragon Plus Environment

Page 9 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
inhibitors of the MBLs than comparators lacking C5 and C7 substituents (11, 12, 14, 15). Compounds 9
and 16, the most potent KPC-2 inhibitors, were also two of the most potent inhibitors of NDM-1 (Ki =
34.2 µM and 31.4 µM) and VIM-2 (Ki = 1.2 µM and 0.316 µM). While the compounds inhibited VIM-2
more potently than NDM-1, rank-order activity of the compound series was similar for the two enzymes.
Generally, and across all three enzymes, inhibitor potency improved with increasing hydrophobic
surface area. There is a particularly striking correlation between the activities for KPC-2 and VIM-2
(Table 1, e.g. comparing 15, 6 and 16). However, compared with KPC-2, the MBLs were less sensitive to
the positioning of ring substituents (e.g., C5 vs C6, comparing 1 and 6). Ring fluorination as in analogs 8
and 10 was unexpectedly detrimental to NDM-1 activity, an observation that will require further
investigation.
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
Complex structures were determined for NDM-1 with compounds 1-3, 6-9, and 11-14, and for
VIM-2 with compounds 8 and 14, all solved at resolutions of 1.70 Å and better (Fig. 2, Fig. S3-S4), initially
using soaking conditions at acidic pH (pH 3.8) to maintain stability of NDM-1 (vide infra). The binding
orientation of the inhibitors was unambiguously identified in the composite simulated annealing mFo-
DFc omit maps contoured at 3 σ (Fig. 2, Fig. S3-S4). In each structure, the phosphonate group bridges
the zinc ions in NDM-1 and VIM-2, displacing a hydroxide ion that is essential for catalysis4. For NDM-1,
the coumarin, DHQ, and quinoline analogs all form common interactions within the active site. These
interactions include HBs from the phosphonate group to N220 and D124, the latter of which forms a
similar interaction with the hydroxide ion in apo structures30, 31. The heteroaryl ring of all three scaffolds
form non-polar interactions with M67, W93 and V73. An edge-face interaction between the heteroaryl
ring and F70 is evident in many NDM-1 complex structures.
One interesting and unexpected aspect revealed in the NDM-1 structures is the ability of the
enzyme to bind similar ligands in one of two distinct orientations: one pointing the ring carbonyl group
away from V73 (pose 1, e.g., compound 1, Fig. 2a) and the other towards it (pose 2, e.g., 6, Fig. 2b).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 53
1
2
3
4
5
6
7
8
9
Pose 1 is observed in the complex structures of compounds lacking C5 substitution (e.g., 1, Fig. 2 and 2,
3, 11, 12 Fig. S3). In contrast, pose 2 is favored for compounds with C5 substitutions as it allows
additional interactions with H122 (e.g., pi-sigma interaction with -Me in 6 (distance 3.67 Å), Fig. 2b; pi-
alkyl interaction with -Br in 9 (distance 3.78 Å), Fig. S3d). This small pocket near H122 in NDM-1 is
analogous to the slot adjacent to W105 in KPC-2, which similarly accommodates the C5 substituent.
Also, of note, is that the side chain of F70 in NDM-1 exhibits significant conformational mobility,
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
a
b
H189
Zn1
H120
H189
Acetate
H120
H122
N220
Zn1
H122
N220
Acetate
C208
Zn2
C208
Zn2
H250
D124
D124
H250
Compound 1
Compound 6
M67
W93
V73
W93
M67
F70
F70-conf A
F70-conf B
V73
c
d
H189
H122
N220
H196
H118
Zn1
H120
Acetate
Zn1
Compound 8
F70-conf B
H116
C221
N233
Compound 8
C208
Zn2
D120
D124
Zn2
H263
W87
H250
F61
M67
W93
F70-conf A
R228
Y67
V73
Figure 2. X-ray crystal structures of NDM-1 and VIM-2 bound to coumarin phosphonate derivatives.
The composite simulated annealing mFo-DFc omit maps, determined at 1.15-1.50 Å resolutions, are
colored in gray and contoured at 3 σ, except panel d, which has been contoured at 2 σ. Black dashed
lines represent hydrogen bonds. Zinc coordination is shown as thin sticks. (a) NDM-1 with 1 (PDB ID
6D1A). (b) NDM-1 with 6 (PDB ID 6D1D). (c) NDM-1 with 8 (PDB ID 6D1F). (d) VIM-2 with 8 (PDB ID
6DD0). F70 of NDM-1 adopts two conformations (labeled F70-conf A and F70-conf B) in the complex
structures of 6 and 8.
ACS Paragon Plus Environment

Page 11 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
allowing the formation of favorable CH-π interactions with the C7-Me of 1 (Fig. 2a) and edge-face
interactions with compounds 11 and 12 (Fig. S3). Overall the relative openness of the NDM-1 active site
and the flexibility of the M67-F70 loop accommodates distinct binding orientations depending on ligand
substitution pattern32. To some extent this echoes the role of W105 in KPC-2 while affording more
plasticity in ligand binding33. However, as described earlier, KPC-2 is more sensitive than NDM-1 to the
positions of the substituents on the heteroaryl ring, as demonstrated by the different activities of
compounds 1 (C6/C7 substitution) and 6 (C5/C7 substitution) against KPC-2, in contrast to the similar Ki
values against NDM-1.
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
We solved structures of the 5,7-difluoro compound 8 (Fig. 2) and the smaller quinoline analog
14 (Fig. S3) bound to both NDM-1 and VIM-2, also at pH 3.8, to compare how these two MBLs interacted
with a common ligand. Although compound 8 did not show inhibition of NDM-1 under our standard
experimental conditions, pre-incubation of the compound with NDM-1 demonstrated weak inhibitory
activity (data not shown), suggesting the compound does interact with NDM-1. In both structures with
NDM-1 and VIM-2, compound 8 adopts pose 2, with the phosphonate group forming a HB with N233
and D120 in VIM-2 (Fig. 2d). The coumarin ring of compound 8 forms a π-π stacking interaction with Y67
of VIM-2, van der Waals interactions with W87, and an edge-face interaction with H263, analogous to
the contacts made by other analogs with F70/W93/H250 in NDM-1. Unique to VIM-2 is the presence of
R228 near the active site, which can establish a HB with the carbonyl oxygen of compound 8 and may
enhance binding and constrain the orientation of the compound, particularly in comparison to NDM-1.
In addition, the stacking interactions between the inhibitor and Y67 in VIM-2 may be more favorable
than contacts with similarly positioned residues M67 or F70 in NDM-1 and can also contribute to
stronger binding affinity for VIM-2.
The observed displacement of the active site hydroxide by phosphonate-bearing inhibitors in
the structures described above, together with the low pH of the soaks, suggested that the phosphonate
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 53
1
2
3
4
5
6
7
8
9
group was in the protonated monoanionic state, and therefore was able to form a HB with the
negatively charged D124, similar to the hydroxide. The use of these conditions was required as the WT
NDM-1 crystals were only stable at low pH. We hypothesized that this was due to the need for a
neutral/protonated D199 at the crystal packing interface. We subsequently constructed a D199N
mutant and were able to determine the complex structures of 7 and 16 at physiological pH (pH 7.5) (Fig.
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
H120
H189
Zn1
H189
a
b
H122
Zn1
H122
N220
C208
Zn2
wat
wat
N220
Zn2
Compound 7
F70
D124
H250
D124
W93
Compound 16
F70
W93
H250
M67
V73
M67
V73
H189
H189
H120
c
d
H120
Zn1
Zn1
H122
N220
H122
N220
C208
Zn2
C208
Zn2
wat
wat
H250
H250
D124
D124
Compound 9
F70
Compound 16
W93
Compound 7
F70
W93
M67
V73
M67
V73
Figure 3. X-ray crystal structures of NDM-1 bound to phosphonate derivatives at pH 7.5. The
composite simulated annealing mFo-DFc omit maps, determined at 1.30-1.40 Å resolutions, are colored
in gray and contoured at 3 σ. Black dashed lines represent hydrogen bonds. Zinc coordination is shown
as thin sticks. (a) NDM-1 with 16 (PDB ID 6NY7). (b) NDM-1 with 7 (PDB ID 6D1E). (c) NDM-1 with 16
(light blue and green), superimposed with the NDM-1 with 9 determined at pH 3.8 (yellow, PDB ID
6D1G). (d) NDM-1 with 7 (light blue and green), superimposed with the same complex determined at
pH 3.8 (yellow, PDB ID 6D1E).
ACS Paragon Plus Environment

Page 13 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
3), where the phosphonate group may also exist in the dianionic state, especially when in close
proximity to the zinc ions. As posited, the complex structures at pH 7.5 retained the active site
hydroxide. The phosphonate group in this case interacts with the hydroxide and one of the zinc ions
(Zn2), rather than both zinc ions as in the low pH structures. As a result, there is a shift in the position of
the phosphonate group, together with the rest of the inhibitor. This movement was accompanied by
different conformations of ligand-interacting residues such as M67 and F70. Comparing 16 (Fig. 3a) and
the mono-bromo coumarin analog 9 (Fig. S3d), a flip of the heteroarene ring was also observed, allowing
16 to adopt a binding mode similar to pose 1 as described above (Fig. 2a). This is possibly due to the
ability of the C5/C7 bromo substituents to establish favorable interactions with H250 and F70. Overall
the complex structure at pH 7.5 provide detailed and valuable information for future lead optimzation.
But the lower pH structures, particularly with their well-defined electron densities, offer important
insights into how active site flexibility enables NDM-1 to respond to changes in ligand/substrate
structure including small changes in the ligand protonation states.
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
H196
Zn1
H196
H189
H116
a
b
H116
H120
H118
C221
R228
H118
H122
C221
C208
wat
wat
G232
R228
N233
Zn1
Zn2
Zn2
D120
N233
N220
H263
H250
H263
D124
Compound 16
Compound 16
W87
W93
W87
F61
F70
Y67
V73
Y67
F61
M67
Figure 4. X-ray crystal structures of VIM-2 bound to phosphonate derivative 16 at pH 7.5. The
composite simulated annealing mFo-DFc omit map, determined at 2.1 Å resolution, is colored in gray
and contoured at 2 σ. Black dashed lines represent hydrogen bonds. Zinc coordination is shown as
thin sticks. (a) VIM-2 with 16 (PDB ID 6O5T). (b) VIM-2 with 16 (orange and green) superimposed
with the complex of NDM-1 and 16 (blue, PDB ID 6NY7).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 53
1
2
3
4
5
6
7
8
9
We also determined the complex crystal structure of VIM-2 with compound 16 at pH 7.5 (Fig.
4a, Fig. S4b). Unlike the complex structure with 8, the carbonyl group is pointed away from R228
without forming a HB. One oxygen of the phosphonate group is positioned 3.7 Å from R228, suggesting
favorable electrostatic interactions. The DHQ ring with the di-bromo subsitutions establishes extensive
interactions with H263, R228, W87 and especiallly Y67 and F61. In addition, compared with NDM-1, the
binding pocket is slightly narrower in VIM-2 with the loop of G232-N233 lowered closer to the ligand,
and the planar peptide bond between G232 and N233 stacked on top of the DHQ ring (Fig. 4a). The
new contacts with R228, G232 and N233, together with the favorable stacking interactions with Y67 and
F61, may explain the ~100x higher affinity of 16 for VIM-2 than NDM-1. However, despite these
differences, it is important to note that the overall binding orientation of 16 is similar between VIM-2
and NDM-1 (Fig. 4b), echoing the observations of the complex structures of 8 (Fig. 2), and
desmonstrating the feasbility of lead optimization simutaneously improving activity against both
enzymes.
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
Cell-based activity of compounds 9, 11, 13 and 16.
To investigate the effects of these inhibitors on bacterial cells, coumarin 9 and DHQ analogs 11
Table 2. In vitro activity (MIC) of imipenem against bacterial strains expressing KPC-2, NDM-1, and
VIM-2 in the presence of 9, 11, 13.
Characterized
Bacterial strain
E. coli 5980
β-lactamase
NDM-1
KPC-2
8
64
>64
2
2
1
2
4
2
1
>64/4
1/4
K. pneumoniae 4683
K. pneumoniae 5979
E. cloacae 5981
NDM-1
NDM-1
VIM-2
32
2
32
2
64
2
>64/4
2/4
P. aeruginosa 4698
>64
>64
>64
>64
32/4
a
b
Compound 9, 11, and 13 were tested at a concentration of 128 µg/mL.
Ceftazidime and 4 µg/mL avibactam was included as a positive control.
ACS Paragon Plus Environment

Page 15 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
and 13 were tested (at fixed 128 µg/mL concentrations) in combination with the carbapenem antibiotic
imipenem against Gram-negative clinical isolates known to produce carbapenemases (Table 2).
Compound 9 restored susceptibility to imipenem in a K. pneumoniae strain producing KPC-2 and
reduced the minimum inhibitory concentration (MIC) by 64-fold. It also decreased the MIC for imipenem
by 4-fold in an NDM-1 producing E. coli strain and more than 2-fold in an NDM-1 producing K.
pneumoniae strain. Compound 9 did not have any activity against NDM-1 producing Enterobacter
cloacae. Compared to KPC-2, the relatively lower cell-based activity of 9 against NDM-1 expressing
strains correlates with the different in vitro activities observed in biochemical testing. The DHQ analog
13 was comparable to compound 9, reducing the MIC of imipenem by 64-fold in KPC-2 producing K.
pneumoniae and 4-fold in NDM-1 producing E. coli, but unable to rescue imipenem activity in NDM-1
producing K. pneumoniae/E. cloacae or in VIM-2 producing P. aeruginosa. The variable susceptibility of
the different bacterial strains likely reflects the influence of other resistance mechanisms such as active
efflux. Consistent with its weaker biochemical activity, DHQ analog 11 was less effective than
compounds 9 or 13 in reducing the MIC of imipenem against the KPC-2 producing K. pneumoniae and
exhibited little if any effect against NDM-1 producing E. coli, K. pneumoniae, and E. cloacae. In general,
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. In vitro activity (MIC) of imipenem against bacterial strains expressing KPC-2, NDM-1, and
VIM-2 in the presence of 16
Bacterial strain
Characterized β- Imipenem (µg/mL)
Imipenem+16 (µg/mL)
lactamase
imipenem
16
K. pneumoniae C2
KPC-2
512
16
4
2
8
32
2
4
-
-
16
128
128
128
128
128
128
-
P. aeruginosa 4135
P. aeruginosa 4129
P. aeruginosa(eflx+) 4177
P. aeruginosa 4882
P. aeruginosa(eflx+) 4402
P. aeruginosa 4173
P. aeruginosa(eflx+) 4175
KPC-2
NDM-1
NDM-1
VIM-2
VIM-2
vector
vector
64
128
128
32
32
2
2
-
a The P. aeruginosa strains (without eflx+) lack major efflux pumps MexAB-OprM and MexXY-
OprM. The eflx+ strain overexpresses MexAB-OprM.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 53
1
2
3
4
5
6
7
8
the microbiological studies revealed that heteroaryl phosphonate analogs can cross the outer
membrane to target both serine and metallo-carbapenemases in Gram-negative bacteria, at least at
high concentrations.
9
Following these initial experiments, we performed a more extensive microbiological assessment
of the most potent DHQ analog 16 (Table 3). A checkerboard assay using a clinical KPC-2 expressing K.
pneumoniae strain varied both imipenem and inhibitor concentrations and demonstrated that
compound 16, at 128 µg/mL, reduced the MIC of imipenem by 128-fold, and by 32-fold when applied at
16 µg/mL concentration. In comparison, at 16 µg/mL concentration, avibactam lowered the imipenem
MIC by 512-fold to 1 µg/mL. These results illustrated the dose-dependent effect of compound 16, while
highlighting its activity at more therapeutically relevant concentrations. We also studied this compound
(at a fixed concentration of 128 µg/mL) across a series of isogenic carbapenemase-expressing P.
aeruginosa strains with deletion of the efflux pumps MexAB-OprM and MexXY-OprM, or over-
expressing MexAB-OprM. For the strains lacking efflux pumps, the imipenem MIC was reduced by 32,
16, and 16-fold for KPC-2, NDM-1 and VIM-2, respectively. It should be noted that, for KPC-2 (strain
4135) and VIM-2 (strain 4882), the MIC of imipenem was 2 µg/mL in these experiments in the presence
of compound 16, the same as the MIC of imipenem alone in the control strain without any
carbapenemase (strain 4173, Table 3). Thus, an MIC reduction of 32 and 16-fold is the maximum that we
could observe for KPC-2 and VIM-2, respectively. For the strains overexpressing efflux pumps, the
imipenem MIC was reduced by 4 and 8-fold for NDM-1 and VIM-2, respectively. Overall the MIC data in
the presence of compound 16 mirror the inhibition results from the biochemical assays, consistent with
the different activities against KPC-2, NDM-1, and VIM-2. Although overexpressing efflux pumps did
decrease its potency by up to 4-fold against NDM-1, the inhibitor retained significant cell-based activity
in these bacterial strains.
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
ACS Paragon Plus Environment

Page 17 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
The cytotoxicity of the phosphonate compounds was also evaluated using compound 16 and
human HEK 293T cells. Compound 16 showed cytotoxicity in these experiments only when used at
concentrations of 100 μg/mL and above, with 88% cell recovery rate at 100 μg/mL and 70% recovery
rate at 200 μg/mL (Fig. S5). In addition, without imipenem, neither 9 nor 16 showed any antibacterial
effects by itself at 128 μg/mL. Although we used up to 128 μg/mL inhibitor concentrations in our MIC
testing, compound 16 displayed significant cell-based activity at lower concentrations (e.g., reducing
imipenem MIC by 32-fold at 16 μg/mL) that will be more relevant to clinical applications. Nevertheless
the potential cytotoxicity of these inhibitors should be further characterized in future development
given previous studies showing the adverse effects of the phosphono-alkylquinoline compounds34.
To assess the drug-like properties of this new scaffold, compound 16 was evaluated in a small
panel of standard in vitro ADME assays (Table 4). The compound exhibited excellent stability when
incubated with mouse liver microsomes and also reasonably high solubility in pH 7.4 phosphate buffered
saline (PBS). Perhaps on account of its anionic nature, 16 was poorly permeable across MDCK-MDR
monolayers, predicting for low oral absorption and blood-brain barrier penetrance. Given that most β-
lactam antibiotics used to treat serious infections are administered parenterally, the in vitro ADME
Table 4. In vitro ADME data for compound 16 and controls.
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
liver microsome stability (mouse)
compound
verapamil
16
t1/2 (min)
8.6 ± 0.1
>120
CLint (µL/min/mg protein)
161.9 ± 2.7
< 11.6
kinetic solubility in pH 7.4 PBS
compound
µM
193
241
µg/mL
51.6
95.9
13
16
MDCK-MDR cell permeability
compound
digoxin
propranolol
16
Papp, A-B (x10-6 cm/s) Papp, B-A (x10-6 cm/s) recovery (%)
0.4
16.5
34.9
< 0.2
105
97
57.5
< 0.7
91
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 53
1
2
3
profile of 16 is quite satisfactory as a starting point for future in vivo proof-of-concept studies and lead
4
5
optimization efforts.
6
7
8
9
Discussion
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
Antibiotic resistance conferred by SBLs and MBLs in Gram-negative pathogens is a serious
threat to human health. We have identified and characterized the first series of non-electrophilic, cross-
class inhibitors of SBLs and MBLs. The structures and SAR described here provide insights into inhibitor
discovery against these important antibiotic targets by providing new leads, and by elucidating the
structural features underlying the broad substrate spectrum of carbapenemases.
Non-covalent phosphonate compounds as cross-class inhibitors
Phosphonate-bearing fragments of modest affinity have been described previously for SBLs35,
36, while inhibitors bearing electrophilic phosphonate esters have been studied as traditional suicide
substrate inhibitors of SBLs by Pratt and colleagues37-40. Mercaptophosphonates were found to inhibit
MBLs in a non-covalent fashion; however, they possess no activity against SBLs41. In contrast, the
inhibitors detailed herein bind both SBLs and MBLs though a reversible, non-covalent mechanism, as
evidenced by their complex crystal structures and the absence of time-dependent inhibition kinetics.
The role of the phosphonate function in these new compounds resides in its ability to mimic the
carboxylate group of substrates, rather than as a reactive surrogate of the scissile bond in substrates, as
in the case of activated phosphonate ester inhibitors. Their expanded spectrum of inhibition, novel
reversible mechanism, and minimal cytotoxicity in mammalian cells, suggests (2-oxo-1,2-
dihydroquinolin-4-yl)methyl)phosphonates as a viable new scaffold for antibiotic development.
The emergence of MBLs in recent years has led to increasing interest in novel inhibitor
development against these proteins and resulted in many promising lead scaffolds including some also
containing phosphonate groups42-46. Most of these compounds are not active against SBLs. A recently
ACS Paragon Plus Environment

Page 19 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
identified series of MBL inhibitors contain both carboxylate and phosphonate moieties, and it is the
carboxylate rather than the phosphonate group that interacts with the zinc ions in the complex crystal
structures43. Bisphosphonate compounds have also been shown to inhibit both MBLs and SBLs, but the
mode of action is not entirely clear for these compounds partially due to the lack of crystal structures44.
In addition, pre-incubation appeared to significantly increase these compounds’ potency against SBLs,
suggesting they may function as covalent inhibitors.
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
Whereas most inhibitor development efforts have focused on either SBLs47-57 or MBLs18, 43, 45, 46,
58-62, there have been some early successes of cross-class inhibitors44, 63, most notably the cyclobutanone
analogs of β-lactam antibiotics64, 65. The recently developed cyclic-boronate esters represent arguably
the most promising broad-spectrum inhibitors targeting both SBLs and MBLs24. One such inhibitor,
named VNRX-5133, has demonstrated promising activity in clinical trials against SBL and MBL producing
Gram-negative bacteria66. Essentially, these cyclic-boronates function as covalent transition state
analogs in the case of SBLs, and paradoxically as non-covalent inhibitors of MBLs. By contrast, the
compounds described herein are to our knowledge the highest affinity cross-class inhibitors with a
a
b
H189
N220
T237
H122
T235
H120
Zn1
Cefotaxime
Product
C208
Zn2
Ampicillin
Product
H250
D124
S130
Compound 16
Compound 9
W105
M67
W93
51
52
53
54
55
56
57
58
59
60
Figure 5. Superimposition of carbapenemase inhibitor and product complexes. (a) Superimposition of
KPC-2/compound 9 (white and green, PDB ID 6D19) and KPC-2/cefotaxime product (purple, PDB ID 5UJ3).
(b) Superimposition of NDM-1/compound 16 (light blue and green, PDB ID 6NY7) and NDM-1/ampicillin
product (yellow, PDB ID 4HL2).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 53
1
2
3
4
5
6
7
8
9
shared, reversible mode of non-covalent inhibition for both SBLs and MBLs. Importantly, these inhibitors
capture functionally equivalent structural features of both SBLs and MBLs, with the phosphonate moiety
placed in the subpocket recognizing the carboxylate of substrate in both classes of enzymes (Fig. 5) and
highlighting the potential to further improve activity simultaneously against both serine- and metallo-
enzymes. In KPC-2, this sub-pocket consists of S130, T235, and T237 67. In NDM-1/VIM-2, one of the two
zinc ions and neighboring residues play a similar role in recognizing the same substrate C3’/4’
carboxylate group4, 16. The phosphonate group serves as an anchor that situates the compound similarly
in KPC-2 and NDM-1/VIM-2 relative to substrate binding. Accordingly, inhibitors positioned in this way
form similar interactions with other conserved active site features in both SBLs and MBLs that recognize
the same substrate functional groups, such as W105 in KPC2 and W93/M67 in NDM-1 that interact with
the conjugated 5/6-membered ring of the β-lactam (Fig. 1, 3, 5). Such similarities in ligand binding
features are underscored by the correlation between the inhibitors’ activities for KPC-2, NDM-1 and
VIM-2, as the hydrophobic surface area of the inhibitor increases (Table 1). Meanwhile, the potency of
the inhibitors was consistently >10-fold weaker against NDM-1 than for KPC-2 or VIM-2, suggesting
further subtleties of compound binding/inhibition that will need to be understood and exploited as
these compounds are optimized towards more advanced lead compounds, which is the focus of our
continuing efforts. Compared with β-lactam substrates, the current inhibitors engage only those active
site surfaces that bind the core β-lactam structure – the fused bicycle and the C3’/4’ carboxylate group –
leaving largely unexplored the residues interacting with distal side chains (e.g., the amino-benzyl group
of ampicillin, Fig. 5b). Targeting these additional ligand binding hot spots is expected to further improve
potencies against NDM-1 as well as against the other carbapenemases. It should also be mentioned that
our current inhibitors showed no activity against OXA-48 Class D carbapenemase (data not shown),
possibly due to a narrower active site of Class D β-lactamases that may require a smaller core
heterocycle for effective inhibition. Nonetheless the heteroaryl phosphonate scaffold can be engineered
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
ACS Paragon Plus Environment

Page 21 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
to target a wide range of carbapenemases due to the need of these enzymes to recognize the same
substrates.
6
7
8
9
Structural basis for carbapenemase broad-spectrum substrate profile
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 novel inhibitors and complex crystal structures described here also shed light on particular
aspects of carbapenemase substrate recognition that add to our understanding of these enzymes’
broad-spectrum activity and that will be useful for future antibiotic development. The new complex
structures reveal the contributions of several hydrophobic residues to ligand binding, notably
W105/L167 in KPC-2, Y67/W87 in VIM-2, and M67/F70/V73/W93 in NDM-1. The conformational
flexibility of specific residues like W105 in KPC-2, F70 in NDM-1, and Y67/R228 in VIM-2 moreover
provide important adaptability in the recognition of small molecules. Thus, in KPC-2 we uncovered
multiple W105 conformations in binding of compounds 1 and 6, a mobility rarely observed in the
analogous residue Y105 of other Class A SBLs33, 68-70. Likewise, residue F70 and the surrounding loop in
NDM-1 is observed in our structures in distinct conformations depending on ligand structure. In VIM-2,
R228 adopts different conformations upon binding to compounds 8 and 14, making important HBs in
both cases. The adaptability of MBLs is also revealed in the two different ligand conformations (poses 1
and 2) we observed in the complex structures of NDM-1 (e.g., 1 vs 6, Fig. 2a, b), resulting partially from
the relative openness of the MBL active sites. Together, these features endow carbapenemases with a
ligand binding versatility that may underlie their ability to hydrolyze a wide variety of β-lactam
antibiotics but paradoxically may also enable and facilitate new inhibitor discovery14, 67. The
susceptibility to small molecule inhibition is highlighted by the remarkable L.E. of compound 16 against
KPC-2 (0.59 kcal mol-1 HA-1), demonstrating the feasibility of using non-covalent inhibitors to achieve
potent activity against carbapenemases. In addition, the ligand binding versatility of carbapenemases is
also demonstrated by the observation of multiple copies of some inhibitors in the active site (e.g.,
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 53
1
2
3
4
5
compound 8 in VIM-2, Fig. S3), even though some of them were left unmodeled due to low occupancy
(e.g., compound 1 in KPC-2).
6
7
8
9
Conclusion
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 summary, we have presented herein multiple heteroaryl phosphonate scaffolds with cell-
based activity, low-nM to low-µM biochemical activities against both SBLs and MBLs, and with favorably
lead-like physiochemical properties (clogD, solubility) and in vitro ADME properties. As such, this new
scaffold appears quite suitable for further carbapenemase inhibitor discovery efforts. To support such
efforts more broadly, we have detailed new structural information that enhances current understanding
of carbapenemase activity and spectrum, reveals the structural basis for ligand/substrate binding, and
should help guide drug discovery efforts aimed at better addressing the most intractable of Gram-
negative bacterial infections.
Experimental Section
Compound Synthesis
Compounds 1-6 were obtained from commercial sources while compounds 7-15 were synthesized as
detailed below. All synthetic analogs submitted for biochemical or biological testing were judged to be
≥95% pure based on inspection of LC/MS and 1H NMR spectra. Scans of representative 1H NMR spectra
are provided in the Supporting Information. 1H and 13C NMR spectra were recorded on Varian INOVA-
400 400 MHz and Bruker 500 MHz spectrometers. Chemical shifts are reported in δ units (ppm). NMR
spectra were referenced relative to residual NMR solvent peaks. Coupling constants (J) are reported in
hertz (Hz). Air and/or moisture sensitive reactions were carried out under an argon atmosphere in oven-
dried glassware using anhydrous solvents from commercial suppliers. Air and/or moisture sensitive
reagents were transferred via syringe or cannula and were introduced into reaction vessels through
ACS Paragon Plus Environment

Page 23 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
rubber septa. Solvent removal was accomplished with a rotary evaporator at ca. 10-50 Torr. Column
chromatography was carried out using a Biotage SP1 flash chromatography system and silica gel
cartridges from Silicycle. Analytical TLC plates from EM Science (Silica Gel 60 F254) were employed for TLC
analyses. Mass analyses and compound purity were determined using Waters Micromass ZQTM,
equipped with Waters 2795 Separation Module, Waters 2996 Photodiode Array Detector, and Waters
2424 ELS detector Separations were carried out with an XTerra MS C18, 5µm, 4.6 x 50 mm column, at
ambient temperature (unregulated) using a mobile phase of water-methanol containing a constant 0.10
% formic acid. Fluorescence and absorbance data was collected on a Molecular Devices Flex Station 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
[(5-methoxy-7-methyl-2-oxo-2H-chromen-4-yl)methyl]phosphonic acid (7)
Step 1. A mixture of ethyl 4-bromoacetoacetate (0.66 mL, 4.8 mmol) and triethyl phosphite (0.818 mL,
4.8 mmol) were stirred at 100 oC for 25 min in a microwave reactor. Purification by flash column
chromatography (50% EA/hexanes followed by 100% EA) yielded 380 mg (30%) of ethyl 4-
(diethoxyphosphoryl)-3-oxobutanoate as a pale yellow oil. 1H NMR (400 MHz, CDCl3) δ 4.18-4.06 (m,
6H), 3.61 (s, 2H), 3.24 (s, 1H), 3.19 (s, 1H), 1.35-1.20 (m, 9H). Step 2. To a cooled (0 oC) mixture of 5-
methylresorcinol (0.12 g, 0.94 mmol) and ethyl 4-(diethoxyphosphoryl)-3-oxobutanoate (0.25 g, 0.94
mmol), was added concentrated sulfuric acid (0.25 mL). The reaction mixture was stirred at ambient
temperature for 18 h. The reaction mixture was then poured into cold water. The precipitate formed
was filtered and dried to obtain about 0.23 g of crude 4-diethyl [(5-hydroxy-7-methyl-2-oxo-2H-
chromen-4-yl)methyl]phosphonate as a pale pink solid that was used without further purification in the
1
next step. H NMR (400 MHz, CDCl3) δ 6.56 (s, 1H), 6.32 (s, 1H), 6.05 (s, 1H), 4.19-4.12 (m, 4H), 3.79 (s,
1H), 3.73 (s, 1H), 2.12 (s, 3H), 1.33 (br. t, J = 7.02 Hz, 6H). LC-MS: m/z = 327 [M+H]+. Step 3. To a
solution of diethyl [(5-hydroxy-7-methyl-2-oxo-2H-chromen-4-yl)methyl]phosphonate (0.036 g, 0.1
mmol) in methanol (1 mL), was added (trimethylsilyl)diazomethane (1.1 mL, 2.2 mmol). The mixture was
stirred at ambient temperature for 18 h, concentrated and isolated by flash column chromatography (12
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 53
1
2
3
g, 5% methanol/dichloromethane) to obtain about 36 mg (96%) of diethyl [(5-methoxy-7-methyl-2-oxo-
4
5
1
2H-chromen-4-yl)methyl]phosphonate as a yellow solid. H NMR (400 MHz, CDCl3) δ 6.76 (s, 1H), 6.57 (s,
6
7
8
1H), 6.14 (s, 1H), 4.21-3.96 (m, 2H), 3.93 (s, 3H), 3.89-3.67 (m, 4H), 2.40 (s, 3H), 1.44-1.17 (m, 6H)
9
13C NMR (100 MHz, CDCl3) δ 160.25, 157.40, 155.33, 147.66, 147.56, 143.58, 115.43, 115.35, 110.55,
107.71, 62.49, 62.42, 55.97, 52.74, 52.68, 34.38, 33.02, 22.02, 16.41, 16.35. LC-MS: m/z = 341 [M+H]+.
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
Step 4. A mixture of diethyl [(5-methoxy-7-methyl-2-oxo-2H-chromen-4-yl)methyl]phosphonate (0.036
g, 0.1 mmol) and 4 M solution of hydrochloric acid in dioxane (1 mL) were stirred at 100 oC for 48 h. The
reaction mixture was concentrated down to dryness azeotropically with toluene, washed with ethyl
1
acetate to obtain 23 mg of compound 7 (77%) as a tan solid. H NMR (400 MHz, CD3OD) δ 6.78 (s, 2H),
6.18 (d, J = 4.7 Hz, 1H), 3.96 (s, 3H), 3.78 (s, 1H), 3.72 (s, 1H), 2.43 (s, 3H). 13C NMR (100 MHz, CD3OD) δ
161.22, 158.01, 155.13, 150.34, 150.24, 144.02, 114.10, 114.01, 109.47, 107.98, 107.32, 55.26, 35.34,
34.04, 20.49. LC-MS: m/z = 285 [M+H]+.
[(5,7-difluoro-2-oxo-2H-chromen-4-yl)methyl]phosphonic acid (8)
Step 1. To a cooled (0 oC) mixture of ethyl 4-bromoacetoacetate (0.26 mL, 1.9 mmol) in concentrated
sulfuric acid (0.5 mL), was added 3, 5-difluorophenol (0.25 g, 1.9 mmol). The reaction mixture was
stirred at ambient temperature for 18 h. Reaction mixture was then diluted with cold water and
extracted with ethyl acetate. The organic extracts were washed with saturated aqueous sodium
bicarbonate solution, brine, dried over magnesium sulfate and purified by flash column chromatography
(35% ethyl acetate/hexanes) to obtain 150 mg (28%) of 4-(bromomethyl)-5,7-difluoro-2H-chromen-2-
one as a white solid. 1H NMR (400 MHz, CDCl3) δ 6.99-6.93 (m, 1H), 6.90-6.82 (m, 1H), 6.50 (s, 1H), 4.60
(d, J = 2.2 Hz, 2H). LC-MS: m/z = 275 [M+H]+. Step 2. A mixture of 4-(bromomethyl)-5,7-difluoro-2H-
chromen-2-one (0.038 g, 0.1 mmol) and triethyl phosphite (0.047 mL, 0.3 mmol) were heated to 100 oC
for 18 h. The reaction mixture was purified by flash column chromatography (50% ethyl acetate/hexanes
ACS Paragon Plus Environment

Page 25 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
followed by 5% methanol/dichloromethane) to obtain 15 mg (33%) of the title compound as a pale
yellow solid..1H NMR (400 MHz, CDCl3) δ 6.93 (dt, J = 8.5, 1.9 Hz, 1H), 6.88-6.75 (m, 1H), 6.31 (d, J = 4.9
Hz, 1H), 4.14 (quin, J = 7.4 Hz, 4H), 3.51 (s, 1H), 3.45 (s, 1H), 1.30 (t, J = 7.1 Hz, 6H). LC-MS: m/z = 333
[M+H]+. Step 3. A mixture of diethyl [(5,7-difluoro-2-oxo-2H-chromen-4-yl)methyl]phosphonate (0.014 g,
0.0 mmol) and 4 M solution of hydrochloric acid in dioxane (1 mL) were stirred at 100 oC for 48 h. The
reaction mixture was concentrated down to dryness azeotropically with toluene, washed with 50% ethyl
acetate /hexanes to obtain 11 mg of compound 8 (95%) as a tan solid. 1H NMR (400 MHz, CD3OD) δ 7.15-
6.98 (m, 2H), 6.38 (d, J = 4.9 Hz, 1H), 3.57 (s, 1H), 3.51 (s, 1H), LC-MS: m/z = 277 [M+H]+.
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
Synthesis of [(5-bromo-7-methyl-2-oxo-2H-chromen-4-yl)methyl]phosphonic acid (9)
Step 1. To a cooled (0 oC) mixture of 3-bromo-5-methylphenol (0.400 g, 2.1 mmol) and concentrated
sulfuric acid (0.5 mL), was added ethyl 4-bromoacetoacetate (0.296 mL, 2.1 mmol) slowly. The reaction
mixture was stirred at ambient temperature for 18 h. Reaction mixture was then diluted with cold water
and extracted with ethyl acetate. The organic extracts were washed with brine, dried over magnesium
sulfate and purified by flash column chromatography (0-25% ethyl acetate/hexanes) to obtain 134 mg
(24%) of 5-bromo-4-(bromomethyl)-7-methyl-2H-chromen-2-one as a cream colored solid. 1H NMR (400
MHz, CDCl3) δ 7.46 (s, 1H), 7.17 (s, 1H), 6.61 (s, 1H), 5.00 (s, 2H), 2.42 (s, 3H). LC-MS: m/z = 332 [M+H,
79Br]+, 334 [M+H, 81Br]+.Step 2. A mixture of 5-bromo-4-(bromomethyl)-7-methyl-2H-chromen-2-one
(0.130 g, 0.4 mmol) and triethyl phosphite (0.134 mL, 0.8 mmol) were stirred in at 100 oC for 24 h. The
reaction mixture was diluted with ethyl acetate and washed with water and brine. The organic layer was
dried over magnesium sulfate and purified by flash column chromatography (0-50% ethyl
acetate/hexanes) to obtain 90 mg (59%) of diethyl [(5-bromo-7-methyl-2-oxo-2H-chromen-4-
1
yl)methyl]phosphonate as a yellow oil that solidifies on standing. H NMR (400 MHz, CDCl3) δ 7.44 (s,
1H), 7.15 (s, 1H), 6.44 (d, J = 4.8 Hz, 1H), 4.24-4.08 (m, 4H), 4.06 (s, 1H), 4.00 (s, 1H), 2.41 (s, 3H), 1.28 (t,
J = 7.0 Hz, 6H). 13C NMR (100 MHz, CDCl3) δ 158.92, 158.89, 155.46, 147.33, 143.12, 133.20, 119.44,
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 53
1
2
3
4
5
6
7
8
9
119.00, 118.90, 118.08, 115.70, 62.65, 62.59, 55.97, 33.66, 32.29, 16.39, 16.33. LC-MS: m/z = 388 [M+H,
79Br]+, 390 [M+H, 81Br]+. Step 3. A mixture of diethyl [(5-bromo-7-methyl-2-oxo-2H-chromen-4-
yl)methyl]phosphonate (0.090 g, 0.2 mmol) and 4 M solution of hydrochloric acid in dioxane (0.6 mL)
were stirred at 100 oC for 24 h. The reaction mixture was concentrated, washed with ethyl acetate and
dried to obtain 70 mg (91%) of compound 9 as a tan powder. 1H NMR (400 MHz, d6-DMSO) δ 7.53 (s,
1H), 7.29 (s, 1H), 6.44 (d, J = 4.5 Hz, 1H), 3.91 (s, 1H), 3.84 (s, 1H), 2.37 (s, 3H). 13C NMR (100 MHz, d6-
DMSO) δ 156.56, 155.31, 149.40, 149.31, 143.46, 133.13, 119.39, 118.90, 118.06, 117.97, 117.46,
115.75, 115.71, 34.17, 32.85, 19.37. LC-MS: m/z = 332 [M+H, 79Br]+, 334 [M+H, 81Br]+.
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
[(5-fluoro-7-methyl-2-oxo-2H-chromen-4-yl)methyl]phosphonic acid (10)
Step 1. To a cooled ( 0 oC) mixture of 3-fluoro-5-methylphenol (0.200 g, 1.6 mmol) and ethyl 4-
bromoacetoacetate (0.219 mL, 1.6 mmol), was added concentrated sulfuric acid (0.8 mL). The reaction
mixture was stirred at ambient temperature for 18 h. Reaction mixture was then diluted with cold water
and extracted with ethyl acetate. The organic extracts were washed with brine, dried over magnesium
sulfate and concentrated in vacuo to obtain 0.23 g of the crude 4-(bromomethyl)-5-fluoro-7-methyl-2H-
chromen-2-one as a pale yellow solid which was used without further purification. Step 2. A mixture of
4-(bromomethyl)-5-fluoro-7-methyl-2H-chromen-2-one (0.200 g, 0.7 mmol) and triethyl phosphite
(0.253 mL, 1.5 mmol) were stirred at 100 oC for 18 h. The reaction mixture purified by flash column
chromatography (50% ethyl acetate/hexanes followed by 100% ethyl acetate) to obtain 155 mg (64%) of
diethyl [(5-fluoro-7-methyl-2-oxo-2H-chromen-4-yl)methyl]phosphonate as a pale yellow oil. 1H NMR
(400 MHz, CDCl3) δ 6.98 (s, 1H), 6.84 (d, J = 12.9 Hz, 1H), 6.44 (d, J = 4.8 Hz, 1H), 4.20-4.03 (m, 4H), 3.52
(s, 1H), 3.46 (s, 1H), 2.43 (s, 3H), 1.36-1.22 (m, 6H). LC-MS: m/z = 329 [M+H]+. Step 3. A mixture of
diethyl [(5-fluoro-7-methyl-2-oxo-2H-chromen-4-yl)methyl]phosphonate (0.026 g, 0.1 mmol) and 4 M
solution of hydrochloric acid in dioxane (1 mL) were stirred at 100 oC for 18 h. The reaction mixture was
concentrated and purified by reverse phase HPLC to obtain 16 mg (74%) of compound 10 as a white
ACS Paragon Plus Environment

Page 27 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
solid. 1H NMR (400 MHz, CD3OD) δ 7.06 (s, 1H), 6.98 (d, J = 13.1 Hz, 1H), 6.35 (d, J = 4.8 Hz, 1H), 3.56 (s,
1H), 3.51 (s, 1H), 2.45 (s, 3H). 13C NMR (125 MHz, d6-DMSO) δ 160.33, 159.37, 158.31, 154.38, 148.39,
148.31, 144.17, 144.08, 116.21, 113.61, 113.11, 112.93, 107.37, 107.28, 35.57, 34.66, 34.57, 21.39
9
LC-MS: m/z = 273 [M+H]+.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
[(2-oxo-1,2-dihydroquinolin-4-yl)methyl]phosphonic acid (11)
Step 1. A mixture of 4-bromomethyl-2(1H)-quinolinone (0.500 g, 2.1 mmol) and triethyl phosphite
(0.720 ml, 4.2 mmol) were heated in dioxane (2 mL) to 100 oC for 18 h. The reaction mixture was
concentrated and purified by flash column chromatography (0-5% methanol/dichloromethane) to
obtain 300 mg (48%) of diethyl [(2-oxo-1,2-dihydroquinolin-4-yl)methyl]phosphonate as cream colored
solid. 1H NMR (400 MHz, CDCl3) δ 12.82 (br.s, 1H), 7.82 (d, J = 8.2 Hz, 1H), 7.39-7.57 (m, 2H), 7.25-7.23
(m, 1H), 6.72 (d, J = 3.9 Hz, 1H), 4.07 (quin, J = 7.3 Hz, 4H), 3.46 (s, 1H), 3.40 (s, 1H), 1.25 (t, J = 7.0 Hz,
6H) LC-MS: m/z = 296 [M+H]+. Step 2. A mixture of diethyl [(2-oxo-1,2-dihydroquinolin-4-
yl)methyl]phosphonate (0.025 g, 0.085 mmol) and 4 M solution of hydrochloric acid in dioxane (1 mL)
were stirred at 100 oC for 24 h. The reaction mixture was concentrated down to dryness purified by
reverse phase HPLC to obtain 11 mg (54%) of compound 11 as a white solid. 1H NMR (400 MHz, d6-
DMSO) δ 11.64 (br.s, 1H), 7.84 (d, J = 7.9 Hz, 1H), 7.46 (t, J = 7.4 Hz, 1H), 7.28 (d, J = 8.1 Hz, 1H), 7.15 (t, J
= 7.4 Hz, 1H), 6.43 (d, J = 4.0 Hz, 1H), 3.26 (s, 1H), 3.19 (s, 1H). 13C NMR (100 MHz, d6-DMSO) δ 161.89,
145.71, 145.63, 139.29, 130.51, 126.49, 122.17, 121.76, 119.60, 115.72, 33.21, 31.94. LC-MS: m/z = 240
[M+H]+.
[(1-methyl-2-oxo-1,2-dihydroquinolin-4-yl)methyl]phosphonic acid (12)
Step 1. To a cooled (0 oC) solution of diethyl [(2-oxo-1,2-dihydroquinolin-4-yl)methyl]phosphonate
(0.050 g, 0.2 mmol) in DMF (0.5 mL), was added sodium hydride, 60% (0.004 g, 0.2 mmol) and
iodomethane (0.032 mL, 0.5 mmol). The mixture was stirred at ambient temperature for 2 h. The
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 53
1
2
3
4
5
6
7
8
9
reaction mixture was then quenched with aqueous saturated ammonium chloride solution, extracted
with ethyl acetate, washed with water and brine. The organic layer was dried over magnesium sulfate
and purified by flash column chromatography (0-10% methanol/dichloromethane) to obtain 24 mg
(46%) of diethyl [(1-methyl-2-oxo-1,2-dihydroquinolin-4-yl)methyl]phosphonate as a clear oil. 1H NMR
(400 MHz, CDCl3) δ 7.87 (d, J = 7.6 Hz, 1H), 7.59 (t, J = 7.2 Hz, 1H), 7.39 (d, J = 8.5 Hz, 1H), 7.33-7.26 (m,
1H), 6.72 (d, J = 4.4 Hz, 1H), 4.11-4.04 (m, 4H), 3.72 (s, 3H), 3.40 (s, 1H), 3.35 (s, 1H), 1.28-1.24 (m, 6H).
LC-MS: m/z = 310 [M+H]+. Step 2. A mixture of diethyl [(1-methyl-2-oxo-1,2-dihydroquinolin-4-
yl)methyl]phosphonate (0.023 g, 0.1 mmol) and 4 M solution of hydrochloric acid in dioxane (2 mL) were
stirred at 100 oC for 24 h The reaction mixture was concentrated down to dryness and isolated by
reverse phase HPLC to obtain 7 mg (37%) of the product as a white powder. 1H NMR (400 MHz, d6-
DMSO) δ 7.93 (d, J = 7.9 Hz, 1H), 7.63-7.59 (m, 1H), 7.51 (d, J = 8.4 Hz, 1H), 7.26 (t, J = 7.4 Hz, 1H), 6.54
(d, J = 3.5 Hz, 1H), 3.62 (s, 3H), 3.26 (s, 1H), 3.20 (s, 1H). 13C NMR (100 MHz, d6-DMSO) δ 161.03, 144.42,
144.33, 140.11, 130.97, 127.17, 122.14, 121.92, 120.54, 115.07, 33.17, 31.87, 29.33. LC-MS: m/z = 254
[M+H]+.
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
[(5,7-dimethyl-2-oxo-1,2-dihydroquinolin-4-yl)methyl]phosphonic acid (13)
Step 1. A mixture of 3,5-dimethylaniline (0.514 mL, 4.1 mmol), ethyl acetoacetate (1.043 mL, 8.2 mmol)
and potassium tert-butoxide (0.046 g, 0.4 mmol) were stirred at 120 oC for 18 h. The reaction mixture
was diluted with ethyl acetate, washed with aqueous saturated ammonium chloride, water and brine.
The organic layer was dried over magnesium sulfate, concentrated and purified by flash column
chromatography (35% ethyl acetate /hexanes) to obtain 0.33 g (39%) of N-(3,5-dimethylphenyl)-3-
oxobutanamide as a yellow solid. 1H NMR (400 MHz, CDCl3) δ 9.06 (br.s, 1H), 7.18 (s, 2H), 6.77 (s, 1H),
3.55 (s, 2H), 2.30-2.20 (m, 9H). LC-MS: m/z = 206 [M+H]+. Step 2. To a mixture of N-(3,5-
dimethylphenyl)-3-oxobutanamide (0.200 g, 1.0 mmol) in EtOAc (2 mL), was added pyridinium bromide
perbromide (0.156 g, 0.5 mmol). The mixture was stirred at ambient temperature for an hour. The
ACS Paragon Plus Environment

Page 29 of 53
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
reaction mixture was concentrated in vacuo and purified by flash column chromatography (0-25% ethyl
acetate /hexanes) to obtain 88 mg (32%) of 4-bromo-N-(3,5-dimethylphenyl)-3-oxobutanamide.
Concentrated sulfuric acid (0.2 mL) was added to a cooled (0 oC) 4-bromo-N-(3,5-dimethylphenyl)-3-
oxobutanamide (0.088 g, 0.31 mmol). The mixture was stirred at 40 oC for 18 h. Reaction mixture was
then diluted with cold water and extracted with 5% methanol in dichloromethane. The organic extracts
were washed with brine, dried over magnesium sulfate and concentrated in vacuo to obtain 68 mg of
the crude 4-(bromomethyl)-5,7-dimethyl-1,2-dihydroquinolin-2-one as a cream solid which was used
without further purification. 1H NMR (400 MHz, CDCl3) δ 12.48 (br.s, 1H), 7.17 (s, 1H), 6.92 (s, 1H), 6.74
(s, 1H), 4.77 (s, 2H), 2.91 (s, 3H), 2.43 (s, 3H). LC-MS: m/z = 265 [M+H, 79Br]+, 267 [M+H, 81Br]+. Step 3. A
mixture of 4-(bromomethyl)-5,7-dimethyl-1,2-dihydroquinolin-2-one (0.033 g, 0.12 mmol) and triethyl
phosphite (0.17 mL, 0.96 mmol) was stirred in dioxane (2.0 mL) at 100 oC for 22 h. The reaction mixture
was concentrated and purified by flash column chromatography (0-10% methanol/dichloromethane) to
obtain 23 mg (57%) of diethyl [(5,7-dimethyl-2-oxo-1,2-dihydroquinolin-4-yl)methyl]phosphonate as a
white solid. 1H NMR (400 MHz, CDCl3) δ 12.49 (br.s, 1H), 7.14 (s, 1H), 6.86 (s, 1H), 6.64 (d, J = 4.6 Hz, 1H),
4.15-4.00 (m, 4H), 3.71 (s, 1H), 3.65 (s, 1H), 2.87 (s, 3H), 2.39 (s, 3H), 1.32-1.18 (m, 6H). LC-MS: m/z =
324 [M+H]+. Step 4. A mixture of diethyl [(5,7-dimethyl-2-oxo-1,2-dihydroquinolin-4-
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
yl)methyl]phosphonate (0.023 g, 0.07 mmol) and 4 M solution of hydrochloric acid in dioxane (0.8 mL)
were stirred at 100 oC for 24 h. The reaction mixture was concentrated down to dryness, purified by
reverse phase HPLC to obtain 7 mg (37%) of compound 13 as a white solid. 1H NMR (400 MHz, d6-DMSO)
δ 11.54 (br.s, 1H), 6.92 (s, 1H), 6.76 (s, 1H), 6.28 (br.s, 1H), 3.42-3.36 (m, 2H), 2.78 (s, 3H), 2.25 (s, 3H)
13C NMR (100 MHz, d6-DMSO) δ 161.15, 141.06, 139.57, 136.60, 128.18, 123.65, 117.01, 116.98, 114.89,
36.58, 35.31, 25.02, 21.19. LC-MS: m/z = 268 [M+H]+.
[(quinolin-8-yl)methyl]phosphonic acid (14)
ACS Paragon Plus Environment