A novel potent metal-binding NDM-1 inhibitor was identified by fragment virtual, SPR and NMR screening

Article abstract of DOI:10.1016/j.bmc.2020.115437

NDM-1 can hydrolyze nearly all available β-lactam antibiotics, including carbapenems. NDM-1 producing bacterial strains are worldwide threats. It is still very challenging to find a potent NDM-1 inhibitor for clinical use. In our study, we used a metal-binding pharmacophore (MBP) enriched virtual fragment library to screen NDM-1 hits. SPR screening helped to verify the MBP virtual hits and identified a new NDM-1 binder and weak inhibitor A1. A solution NMR study of ¹⁵N-labeled NDM-1 showed that A1 disturbed all three residues coordinating the second zinc ion (Zn2) in the active pocket of NDM-1. The perturbation only happened in the presence of zinc ion, indicating that A1 bound to Zn2. Based on the scaffold of A1, we designed and synthesized a series of NDM-1 inhibitors. Several compounds showed synergistic antibacterial activity with meropenem against NDM-1 producing K. pneumoniae.

Full text of DOI:10.1016/j.bmc.2020.115437

Journal Pre-proofs
A Novel Potent Metal-binding NDM-1 Inhibitor was Identified by Fragment
Virtual, SPR and NMR Screening
Huifang Guo, Kai Cheng, Yan Gao, Weiqi Bai, Cai Wu, Wei He, Conggang
Li, Zhuorong Li
PII:
S0968-0896(20)30247-9
https://doi.org/10.1016/j.bmc.2020.115437
BMC 115437
DOI:
Reference:
To appear in:
Bioorganic & Medicinal Chemistry
Received Date:
Revised Date:
Accepted Date:
26 January 2020
5 March 2020
10 March 2020
Please cite this article as: H. Guo, K. Cheng, Y. Gao, W. Bai, C. Wu, W. He, C. Li, Z. Li, A Novel Potent Metal-
binding NDM-1 Inhibitor was Identified by Fragment Virtual, SPR and NMR Screening, Bioorganic &
Medicinal Chemistry (2020), doi: https://doi.org/10.1016/j.bmc.2020.115437
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover
page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version
will undergo additional copyediting, typesetting and review before it is published in its final form, but we are
providing this version to give early visibility of the article. Please note that, during the production process, errors
may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2020 Elsevier Ltd. All rights reserved.

A Novel Potent Metal-binding NDM-1 Inhibitor was Identified
by Fragment Virtual, SPR and NMR Screening
Huifang Guo a, 1, Kai Cheng b, 1, Yan Gao a, Weiqi Bai a, Cai Wu c, Wei He c, Conggang Li, b, **
Zhuorong Li a, *
a
Institute of Medicinal Biotechnology, Chinese Academy of Medical Science and Peking Union Medical College,
Beijing 100050, China.
b Key Laboratory of Magnetic Resonance in Biological Systems, State Key Laboratory of Magnetic Resonance
and Atomic and Molecular Physics, National Center for Magnetic Resonance in Wuhan, Wuhan Institute of
Physics and Mathematics, Chinese Academy of Sciences, Wuhan 430071, China.
c School of Pharmaceutical Sciences, Tsinghua University, Beijing 100084, China.
1 These authors contributed equally to this work.
* Corresponding author.
** Corresponding author.
E-mail addresses: lizhuorong@imb.pumc.edu.cn (ZR Li), conggangli@wipm.ac.cn (CG Li).
Abbreviations used. NDM-1, New Delhi Metallo-1; SPR, surface plasmon resonance; NMR,
nuclear magnetic resonance; MBP, metal-binding pharmacophore; MBL, metallo-β-lactamase;
FBDD, fragment-based drug discovery; PDB, Protein Data Bank; CNCL PKU, Chinese National
Compound Library of Peking University; LE, ligand efficiency; Hepes, 2-(4-(2-hydroxyethyl)-1-
piperazinyl)-ethansulfonic acid; RU, response unit; DMSO, dimethyl sulfoxide.

Abstract
NDM-1 can hydrolyze nearly all available β-lactam antibiotics, including carbapenems.
NDM-1 producing bacterial strains are worldwide threats. It is still very challenging to find a
potent NDM-1 inhibitor for clinical use. In our study, we used a metal-binding pharmacophore
(MBP) enriched virtual fragment library to screen NDM-1 hits. SPR screening helped to verify
the MBP virtual hits and identified a new NDM-1 binder and weak inhibitor A1. A solution
NMR study of 15N-labeled NDM-1 showed that A1 disturbed all three residues coordinating
the second zinc ion (Zn2) in the active pocket of NDM-1. The perturbation only happened in
the presence of zinc ion, indicating that A1 bound to Zn2. Based on the scaffold of A1, we
designed and synthesized a series of NDM-1 inhibitors. Several compounds showed synergistic
antibacterial activity with meropenem against NDM-1 producing K. pneumoniae.
Keywords
NDM-1, metal-binding pharmacophore, fragment-based drug design, SPR, NMR
1. Introduction
The problem of antimicrobial resistance (AMR) becomes disturbing more than ever [1, 2].
According to WHO priority pathogens list for R&D of new antibiotics [3, 4], carbapenem-
resistant Acinetobacter baumannii, Pseudomonas aeruginosa and Enterobacteriaceae are the
highest ranked bacteria and regarded as a global priority for discovering new drugs.
Carbapenems, among the most used β-lactam antibiotics worldwide, are the most effective
antibiotics against both Gram-positive and Gram-negative bacteria. Since carbapenems are highly
effective against many bacterial species, less vulnerable to most β-lactam resistance
determinants and safer to use than last-line resort polymyxins, they are considered the most
reliable last-resort treatment for intractable infections [5-7]. Thus, the spread of carbapenem
resistance poses a substantial threat to global morbidity and mortality.
The most challenging carbapenem-resistance cause is metallo-β-lactamase (MBL) NDM-
1 [8]. NDM-1 can hydrolyze nearly all available β-lactam antibiotics, including carbapenems.
Since its characterization in 2009 [9], NDM-1 has become a world-wide problem conferring
bacterial resistance to all β-lactam antibiotics, with the only exception of monobactams (i.e.,
aztreonam) [10]. However, monobactam antibiotics can be inactivated by serine β-lactamase
(SBL) normally coproduced with NDM-1, thus expanding bacterial resistance. Several SBL
inhibitors are available in therapy [11, 12]. Valuable progresses have been made in the
structural, mechanistic information and even activity inhibition of NDM-1 (Scheme 1) [13-23].
It had been very challenging to find NDM-1 inhibitors for clinical use [10]. Recently, cyclic

boronate SBL and MBL inhibitor Taniborbactam (VNRX-5133, Scheme 1) is in phase 3
clinical development [24]. Drug combination of cefepime and VNRX-5133 showed potent in
vitro activity against Enterobacteriaceae and P. aeruginosa, including cefepime-, meropenem-,
or piperacillin-tazobactam-non-susceptible isolates. [25] VNRX-5133 also restored cefepime
antibacterial activity in both NDM-producing recombinant strains of E. coli and in clinical
isolates of Enterobacteriaceae. [26] It is promising to see that NDM-1 might not an intractable
drug target.
NH2
O
O
O
HO OH
HO
HO
H
O
SH
O
SH
NH2
OH
OH
B
N
H
N
O
O
HO
N
N
N
O
H
HO
O
O
O
OH
2
(S, S, S)-AMA
D-captopril
IC50 = 7.9 µM
L-captopril
IC50 = 202 µM
Cyclic boronate
IC50 = 7.0 µM
IC50 = 0.029 µM
H
H
N
S
O
O
S
N
OH
N
Se
N
H
SH
H2N
O
B
N
H
HO
O
O
HS
O
HO
COOH
VNRX-5133
L-CS319
Ebselen
Ki = 0.38 µM
Thiorphan
IC50 = 0.01 µM
IC50 = 23 µM
IC50 = 1.8 µM
Scheme 1. Structures of some reported NDM-1 inhibitors [10, 13, 14, 17-19, 21, 24].
There are two catalytic zinc ions (Zn1 and Zn2) in the active site of NDM-1 [13]. NDM-1
catalyzes the hydrolysis of β-lactam antibiotics by the mediation of the two zinc ions [15],
which coordinate a nucleophilic hydroxide to hydrolyze the β-lactam ring. Thus, the discovery
of zinc-binding scaffold probably can help us to find new NDM-1 inhibitors. In our study, by
the analysis of virtual screening hits of NDM-1, we developed a metal-binding pharmacophore
(MBP) enriched fragment library. Then we developed first amine coupling-based method to
immobilize tag-free NDM-1 onto CM7 chip, which helped us to verify NDM-1 hits with SPR
screen. Fragment A1 was found as a new NDM-1 binder and weak inhibitor. A solution NMR
study of 15N-labeled NDM-1 [16] found that A1 disturbed all three residues coordinating
catalytic Zn2 of NDM-1. A1’s perturbation to the residues only happened in the presence of
zinc ion. This is in accordance with molecular docking result that A1 binds to Zn2 in the active
pocket of NDM-1. Based on the scaffold of A1, we designed and synthesized a series of NDM-1
inhibitors. Several compounds showed some synergistic antibacterial activity with meropenem
against NDM-1 producing K. pneumoniae.
2. Results and discussion
2.1 Virtual screening and set up of metal-binding pharmacophore enriched fragment library

Several well-conducted virtual screening of inhibitors of NDM-1 have been reported and
new NDM-1 inhibitors were found, including sulfonamides [27, 28], thiols [29], N-
arylsulphonyl hydrazones [29], succinic acids [29], mercaptocarboxylate acids [29], natural
products [30, 31] and so on [32, 33, 34]. In our study, we try to run the virtual screen to identify
NDM-1 binding fragments fragments with metal-binding pharmacophore.
We used two compound libraries. One library was purchased from Maybridge (Cambridge,
UK) and consisted of 2,500 fragments (Maybridge Ro3 diversity fragment library). The other
library was Chinese National Compound Library of Peking University (CNCL PKU) and
consisted of about 120,000 compounds. The whole Maybridge Ro3 diversity fragment library
was run virtual screening with NDM-1 (PDB code: 4hl2, chain A). For CNCL PKU, compounds
with molecular weight range out of 150 ~ 350 Da were all filtered out. Then the structure
diversity of fragments was maximized with Discovery Studio and the top 28, 000 compounds
were used to run virtual screening with NDM-1.
Virtual screening results first ranked the virtual fragment hits by total scores grading by
Surflex-Dock of SYBYL-X2.0. The total scores of D- and L-captopril with NDM-1 calculated
by Surflex-Dock were 5.9928 and 5.1273, respectively. Since in the docking by Surflex-Dock,
the binding site is calculated as a rigid and inflexible site. More assays are needed to tell whether
a virtual hit is a good hit. We then filtered the hits by whether they have metal-binding
pharmacophores (MBPs). Some chemotypes are thought to be malleable MBPs to the binding
of metalloenzymes, including picolinic acids, hydroxypyrones, hydroxylamines, salicylic acids
and so on [35-37]. From the virtual hits with higher total score, we searched for compounds
which had MBPs (Table 1, Scheme 2). Fingerprint module of Discovery Studio can help to
filter and find the hits with specified MBP. High frequency of carboxyl groups was observed
within high total score virtual hits from both CNCL PKU and Maybridge Ro3 (Table 1).
However, other reported MBPs were rare. There are two possible reasons. One is that both
compound libraries are not MBP rich. The other one is that many compounds with MBPs were
not on our good virtual hit list. However, to enrich MBPs in our new filtered library for NDM-1
binding test with SPR, we manually checked virtual hits from CNCL PKU (with total score >
6.66) and Maybridge Ro3 (with total score > 5.0) to find more possible chemotypes.
Empirically, we selected 31 virtual fragment hits with other MBPs (Scheme 2) from
CNCL PKU, which possibly could bind metal from the view of a chemist. We didn’t find other
MBPs than carboxyl group from virtual hits within Maybridge Ro3. At this point, we made the
decisions only based on the chemical feature of fragments and didn’t consider where the MBPs
bound in NDM-1 active pocket. During the process of manual filtration of the virtual fragment
hits, we kept some hits without any MBPs, since they have high total scores in virtual screening
and passed further Lipinski rule [38] and Verber rule [39] filtration. Thus, we integrated an
MBP enriched NDM-1 virtual fragment hits library. This library had 166 fragments from CNCL

PKU and 120 fragments from Maybridge Ro3 library (Table 1). In total, 286 fragments (50%
fragments had MBP) were all tested by SPR to check whether they can bind to NDM-1 in
solution.
Table 1. Metal-binding pharmacophore enriched NDM-1 virtual fragment hits library and
NDM-1 binders identified by SPR.
O
OH
O
O
HO
R
R
OH
O
Carboxyl
Salicylic acid
Hydroxypyrone
O
HO
O
O
O
O
R1
R2
O
O
R
O
O
N
R1
R2
R
R3
R4
O
O
HO
S
O
S
O
N
N
H
R
N
R
N
N
R
S
H
H
OH
O
Other metal-binding pharmaphocore examples from virtual NDM-1 hits
Scheme 2. Metal-binding pharmacophores identified from NDM-1 virtual fragment hits in
this study.
2.2 Identify NDM-1 binders by SPR screening
With the 286 fragments we got, we then investigated whether the hits that can bind to
NDM-1 in solution by SPR. Sensitive biophysical techniques like SPR are well suited to
identify protein fragment binders in fragment-based drug discovery, since functional assays are
often not sensitive enough to detect the weak interaction between fragment and protein target
[40].
Christopeit et al once described a SPR based biosensor assay for NDM-1 [41]. They
immobilized the enzyme to the surface of a SPR sensor chip by biotin-streptavidin capturing

method and the immobilization level of NDM-1 was around 2,400 response units (RU). Their
method required biotinylated NDM-1. We developed a new method to immobilize tag-free
NDM-1 to CM7 SPR chip by establishing a covalent bond between amine group of N-terminal
amino acid and the carboxyl group of CM7 chip. The immobilization level of NDM-1 reached
more than 8,000 RU with our method, which might give higher responses during the screening
of fragments. Real-time binding analysis of meropenem and L-captopril [10] to NDM-1 showed
that the immobilized protein was active on CM7 chip (Figure 1).
Advantages of SPR screening also include that this method can realize medium to high
through-put of screening and the initial amount of each virtual hit can be very low, as low as 5
µg each hit in our case. By the SPR method we developed, affinity screening was run to identify
NDM-1 binders from MBP enriched hits library which included 120 fragments of Maybridge
Ro3 library and 166 fragments purchased from CNCL PKU. The affinities of hits to NDM-1
were ranked according to the KDs of fragments. We defined fragments with higher affinities (0
< KD ≤ 1 mM) as priority I binders and hits with lower affinities (1 < KD ≤ 2 mM) as priority
II binders. In total, we got 41 priority I binders (hit rate = 25%) from CNCL PKU and 18 priority
I binders (hit rate = 15%) from Maybridge Ro3 library (Table 1). It was not surprising that we
got higher hit rate from CNCL PKU since the total scores of virtual hits we chose from the
virtual screening were higher.
Figure 1. Binding sensorgrams of L-Captopril, meropenem, running buffer and compound A1
dilutions. Tag-free NDM-1 coupled to CM7 chip by amine coupling can be used to sceening
fragment library. The affinities (KD) of meropenem and A1 were fit with the steady state affinity
binding model. The affinity of L-Captopril was fit kinetics with 1: 1 binding model. The figures
were generated by Biacore T200 evaluation software.
2.3 A solution NMR study of 15N-enriched NDM-1 observed A1’s binding to Zn2 in the active

pocket
Solution-state NMR spectroscopy has proved to be a powerful technique to obtain atomic
resolution structures of proteins and their complexes in solution [42, 43]. In contrast to X-ray
crystallography, NMR provides an invaluable tool for studying weak fragment-protein
complexes that are not amenable to crystallization. The x-ray structures of apo NDM-1 or
complexes were reported [13, 15, 44]. In the current study, we titrated some of the NDM-1
15
priority I binders identified by SPR affinity screening into N-labeled NDM-1 and utilized
NMR spectra to identify the detail binding sites of priority I binders in NDM-1, according to
our previous backbone chemical shift assignments [16].
A1 was an α-hydrazono carboxylic acid fragment (Figure 1). The affinity of A1 to NDM-1
was 219 µM measured by SPR. We observed chemical shift perturbations (CSPs) in the 1H-15N
HSQC spectra of NDM-1 when titrated with priority I binder A1 (Figure 2a). The chemical
shifts of NDM-1 changed continuously as A1 was added. Residues from Zn binding sites and
active sites showed larger CSPs (Figure 2b). Molecular docking result showed that A1 bound
to the active pocket of NDM-1 and its carboxyl group coordinated to catalytic Zn2 in the pocket
(Figure 3a). D124, C208 and H250 are the three residues coordinating Zn2 (Figure 3b). These
residues showed larger CSPs than the residues coordinating Zn1 (H120, H122, H189) (Figure
2b) in NMR titration experiment, which confirmed the coordination of A1 to Zn2. NMR
titration experiment showed the affinity of A1 to NDM-1 was 0.62 ± 0.02 mM based on the
shift of D124 (Supplementary material). Within the three residues coordinating Zn1 (H120,
H122, H189), only H122 was disturbed slightly. We also observed large CSPs of T119, V73,
Q123 and L65 in the active sites, indicating these residues are also involved in the A1 binding,
which is consistent with the results of molecular docking.
We also noted that there was no CSPs of NDM-1 when titrated with A1 after we removed
all zinc ions from NDM-1 solution by dialysis (Figure S1). When two equivalents of zinc ions
were added, the perturbation was restored (Figure S1). This observation further suggested that
the interaction between A1 and Zn2 was the major contribution of the binding and the carboxyl
group of A1 should be kept in further structure design and modification.

Figure 2. Compound A1 disturbs the amino acid residues in NDM-1 active pocket. a: 1H-15N
HSQC spectra of di-Zn-NDM-1 titrated with A1. b: The expanded regions of 1H-15N HSQC
spectra of residues that located in Zn1 binding sites (H120, H122, H189), Zn2 binding sites
(D124, C208, H250) and active sites (V73, Q123, G222).
Figure 3. Molecular docking calculations of the binding modes of α-hydrazono carboxylic
acids with NDM-1 (PDB code: 4hl2). The figures were prepared with Discovery Studio 2018.
a: A1 (yellow) was docked into the active pocket of NDM-1 and the A1 carbonyl oxygen was
found bind to catalytic zinc ion 2 (Zn2, yellow ball). b: Residues disturbed by A1 binding
(colored in blue) locate in NDM-1 active pocket. All of the three Zn2 coordinating residues
(D124, C208 and H250) were disturbed, pointing to the existence of the binding A1 to Zn2, not
Zn1. Only one Zn1 coordinating residue (H122) was disturbed, and the disturbance is probably
transmitted from the dramatic disturbance of D124.
2.4 Design and synthesis of NDM-1 inhibitors
We synthesized A1 by the route described in scheme 3 [45, 46] to carry on NMR and NDM-1
enzymatic inhibition tests. The NDM-1 enzymatic inhibition activity (IC50) of fragment A1 was
124.0 ± 8.3 µM (Table 2). The parallel tests of an NDM-1 inhibitor L-captopril showed the
enzymatic inhibitory activity of A1 was stronger than L-captopril. L-captopril is an approved drug
in the treatment of hypertension [47]. The thiol of L-captopril was found intercalating the two
zinc ions of NDM-1 in the X-ray structure [15]. To get more potent NDM-1 inhibitors, we
decided to design and synthesize new compounds based on the scaffold of A1. The carboxyl
group of A1 was the metal-binding pharmacophore coordinated to Zn2 of NDM-1 according to
the NMR and molecular docking results. So, it was kept intact in the following design and
syntheses. The newly designed compounds (Table 2) included changing the aryl rings and
substituents of the α-position of the carboxyl group. We evaluated NDM-1 enzymatic inhibitory
activities of the newly synthesized α-hydrazono carboxylic acids (Table 2). The compounds
showed weak to moderate strong NDM-1 inhibiting activities.

NHNH2
N
OH
O
N
N
N
COOH
a
b
N
N
H
N
H
N
H
OH
NH
Cl-
NH
Cl-
O
NH
Cl-
O
H
1
2
A1
(E)-isomer
(Z)-isomer
Scheme 3. Synthesis of the hydrochloride of A1. Conditions and reagents: a) 80% hydrazine
hydrate, reflux 10 h. 40% yield. b) CH3COCOOH, HCl, H2O, r.t., 2 h. 77% yield.
Table 2. Structures, IC50 Values, NDM-1 binding affinities, and ALogPs of newly designed α-
hydrazono carboxylic acids against NDM-1.
N
COOH
N
COOH
N
COOH
HN
Cl-
N
H
N
COOH
N
H
N
H
N
H
NH
Cl-
R
R
R
R
A1 A4
B1 B4
C1 C4
D1 D4
–
–
–
–
ID
A1
A2
A3
A4
B1
B2
B3
B4
C1
C2
C3
C4
D1
D2
D3
D4
R a
KD (µM) b
IC50 (µM) c
124.0 ± 8.3
68.2 ± 19.9
82.3 ± 13.6
> 500
ALogP d
CH3
Et
290.5 ± 177.4
275.1 ± 165.6
221.4 ± 63.0
435.0 ± 40.2
255.2 ± 21.8
449.4 ± 62.1
> 1000
-0.545
0.122
0.578
1.002
2.439
3.106
3.562
3.985
-1.744
-1.077
-0.621
-0.198
1.531
2.198
2.654
3.077
0.667
--
n-Pr
t-Bu
CH3
Et
388.9 ± 191.0
28.1 ± 41.1
> 500
n-Pr
t-Bu
CH3
Et
306.8 ± 93.3
> 1000
94.6 ± 44.0
n. t.
> 1000
n. t.
n-Pr
t-Bu
CH3
Et
702.9 ± 185.9
> 1000
n. t.
n. t.
167.8 ± 54.5
190.7 ± 153.4
187.7 ± 31.7
438.4 ± 187.6
14.1 ± 1.9
n. t.
n. t.
n-Pr
t-Bu
--
n. t.
n. t.
L-Captopril
EDTA-2Na
390.5 ± 87.6
17.2 ± 10.2
> 1000
--
a Et = ethyl, n-Pr = n-propyl, t-Bu = tert-butyl. b KD values are collected from three
c
separate tests tested by SPR. IC50 values are collected as the mean ± SD from
three separate tests. d ALogP values were calculated with Discovery Studio 2018.
2.4 The NDM-1 inhibition and binding activity of synthesized α-hydrazono carboxylic acids.

The ability of synthesized α-hydrazono carboxylic acids to bind and inhibit NDM-1 were
evaluated. From the binding affinity data, it was found that fragments with quinoline (A1 – A4)
naphthalene (B1, B2, B4) and benzene (D1 – D4) rings showed medium to weak binding
affinities to NDM-1 (Table 2). Compare the binding affinities of fragment A1 – A4 and D1 –
D4, it can be found that the R groups of methyl, ethyl and n-propyl contribute to the affinity
similarly, while bulk tert-butyl substitution can decrease the affinity.
In NDM-1 inhibition tests of synthesized α-hydrazono carboxylic acids, nitrocefin is not a
suitable chromogenic substrate, since inhibition effects were always observed at lower but not
higher fragment concentrations in our tests. With meropenem as the chromogenic substrate, we
observed evident NDM-1 inhibition by fragments A2, A3 and B2 (Table 2). Fragments with
pyridine (C1 – C4) and benzene (D1 – D4) have absorption around 300 nm and NDM-1
inhibition data were not tested.
L-captopril is a strong binder but weak NDM-1 inhibitor. EDTA-2Na showed evident
NDM-1 inhibiting activity by its well-known metal chelating property. However, EDTA-2Na
didn’t show obvious binding to NDM-1, possibly due to the lack of three-dimensional
complementary with NDM-1’s pocket. Compare the data of fragments A1 – A4 and B1 – B4,
most the affinities and IC50s were correlated except B2, possibly influenced by the different
tautomer in the tested condition.
2.5 The synergistic antibacterial activity of synthesized α-hydrazono carboxylic acids with
meropenem against NDM-1 producing K. pneumoniae.
The ability of synthesized α-hydrazono carboxylic acids to restore the antibacterial activity
of a clinically significant carbapenem antibiotic, meropenem, was assessed by performing MIC
assays with K. pneumoniae (ATCC BAA-2470) cells expressing NDM-1. At the concentrations
of 100 µg/mL, compounds A2, A4 and B3 restored the meropenem MIC to 64, 64 and 32 µg/mL.
We suppose that A2 restore the meropenem MIC by binding and inhibiting NDM-1, since
A2 showed evident NDM-1 binding and inhibiting activities in the SPR and enzyme inhibitory
tests (Table 1). In the NMR test, we observed the primary hit A1 can disturb all three residues
coordinating the Zn2 in the active pocket of NDM-1 by binding Zn2. The structure of A1 and
A2 are very close. So, we also suppose that A2 might inhibit NDM-1 by binding Zn2 in K.
pneumoniae (ATCC BAA-2470) cell. However, the synergistic antibacterial activity of A2 was
weak, mostly due to A2 is a fragment, the binding and inhibiting activity need to be improved
by more structure evolution. For the future optimization strategy, we suggest further designing
based on the scaffold of A2. The carboxyl group of A2 should be kept. If the imine group can
be replaced with a carbon-carbon double bond, tautomerization can be avoided. Fragment grow
at the quinoline ring may also can increase the affinity and potency of the scaffold. The outer

membrane of Gram-negative bacteria made most small molecules are harder to penetrate and
accumulate inside the cell. Michelle et al [48, 49] reported that small molecules are most likely
to accumulate with an amine group, are amphiphilic and rigid, and have low globularity. It is
important to increase the accumulation inside the cell of the compounds.
Since compounds A4 and B3 didn’t show evident affinity and inhibiting activities to NDM-
1 (Table 1), we guess that their weak synergistic antibacterial activities are not due to NDM-1
inhibition.
Table 3. MICs of meropenem alone or in combination with α-hydrazono carboxylic acids (100
µg/mL) against NDM-1 producing K. pneumoniae (ATCC BAA-2470).
Combination
MICs (µg/mL) of
Combination
MICs (µg/mL) of
compounds
meropenem
compounds
meropenem
>64
A1
>64
64
C1
C2
C3
C4
D1
D2
D3
D4
A2
>64
A3
>64
64
>64
A4
>64
B1
>64
>64
32
>64
B2
>64
B3
B4
>64
>64
>64
>64
No combination
3. Conclusions
NDM-1 is a tough AMR factor that can inactivate almost all β-lactam antibiotics, even
carbapenems. Discovery of potent NDM-1 inhibitors for clinical use is still very challenging.
The zinc ions in the active site are essential for NDM-1’s activity. The activity of NDM-1 can
be inhibited by blocking the zinc ions. Metal binding has been a successful drug design strategy
in the discovery of inhibitors of viral enzymes containing magnesium [35] and other
metalloenzymes [50]. It is also optimistic to find drugs inhibiting NDM-1 through metal-
binding. Fragment-based drug discovery strategy is very helpful to find NDM-1 inhibitors with
metal-binding pharmacophore.
Our virtual screening of fragment library found primary NDM-1 virtual hits. By analysis
and filtration of the virtual hits, we established a metal-binding pharmacophore enriched
fragment library. Biophysical methods SPR and NMR then helped us to identify NDM-1
binders and the zinc ion binder A1. Based on the scaffold of A1, we designed and synthesized
a series of potent NDM-1 inhibitors. Some showed synergistic antibacterial activity with
meropenem in NDM-1 producing K. pneumoniae. By means of fragment MBP searching, it is
promising to find more potent metal-binding NDM-1 inhibitors.

4. Experimental section
4.1 General information
All reagents and solvents were purchased and used without further purification unless
specified. Virtual sceening was performed with Surflex-Dock of SYBYL-X2.0. Molecular
docking was calculated with CDOCKER of Discovery Studio 2018. TECAN (SPARK 20M)
Plate Reader was used to measure absorbance in the NDM-1 enzymatic activity assay. Semi-
preparative liquid chromatography was performed with Biotage Isolera One with a Biotage
SNAP Cartridge KP-C18-HS 12 g column. Melting points of the compounds were observed
with Mettler Toledo MP90 melting point system or Beijing Tech Instrument X-4 digital
microscope melting point system. The 1H NMR and13C NMR spectra of small compounds were
recorded with a Bruker BioSpin GmbH spectrometer (400 MHz, 500 MHz or 600 MHz). The
internal solvent residual peaks were used as references. The high-resolution mass spectra
(HRMS) of small compounds were recorded on a Thermo Scientific LTQ Orbitrap XL with an
ESI mass selective detector. NMR experiments of 15N-labelled NDM-1 were carried out at 298
K and 293 K on a Bruker Avance 800 MHz spectrometer equipped with 5 mm triple-resonance
HCN-cryoprobe.
4.2 NDM-1 expression and purification
NDM-1 was expressed according to the reported method [13]. Briefly, the NDM-1 gene
was cloned into a pET-28a vector and the recombinant protein (residues I31 ~ K268) was
expressed in E. coli strain BL21 (DE3) cells. E. coli BL21 (DE3) was transformed with
recombinant plasmids. A single colony was inoculated into LB medium containing kanamycin
(100 mg/L). Cultures were incubated in 37 ºC at 200 rpm. IPTG (400 μM) was introduced for
induction. Cells were allowed to grow for another 18−20 h at 22 °C and harvested. NDM-1 was
obtained by three‐step purification using Ni column, thrombin cleavage, Ni column and
Superdex 200. Tag removal was performed to get the tag‐free protein. Fractions were pooled
and dialyzed followed by 0.22 μm filter sterilization. Proteins were analyzed by SDS‐PAGE by
standard protocols for molecular weight (27 kDa) and purity measurements (> 95%). Storage
buffer of NDM-1 was 50 mM hepes (pH 7.5), 150 mM NaCl, 100 µM ZnCl2, 1 mM DTT.
4.3 15N-enriched NDM-1 expression and purification
15N-enriched NDM-1 was expressed and purified according to the reported method [16].

Generally, the NDM-1 gene was cloned into a pET-21a vector and the recombinant protein
(residue 36-270) contains a C-terminal His6-tag (LEHHHHHH). The protein was expressed in
15
E. coli strain BL21 (DE3) cells. Cells were grown in M9 media containing NH4Cl at 37 C
and induced with 500 μM IPTG at 16 C for 24 h when OD600 reaching 0.8 to 0.9. Cells were
harvested by centrifugation and resuspended in Ni-column buffer A (50 mM Tris, 500 mM
NaCl, 10 mM imidazole, pH 7.5). Cells were lysed by sonication (JY92 IIN, 50% amplitude,
60 min, 37.5% duty cycle) on ice. The supernatant was collected after centrifugation at 16000
g at 4 C for 30 min. NDM-1 was first purified using a prepacked Ni-NTA column (GE
Healthcare), then followed by size exclusion chromatography (S100) with a buffer comprising
50 mM HEPES, 200 mM NaCl, pH 7.5. Proteins were assessed by SDS‐PAGE. The
concentration of NDM-1 was determined by using a molar extinction coefficient of 27960 M-1
cm-1 at 280 nm, which was calculated from the amino acid sequence using
http://web.expasy.org/protparam/. For nonfunctional NDM-1(apo-NDM-1), 20 equivalents of
EDTA were added to the purified protein and the EDTA was soon removed by buffer exchange.
For functional NDM-1(di-Zn-NDM-1), two equivalents of Zn2+ were added to apo-NDM-1.
The purified proteins were lyophilized and stored at -80 °C.
4.4 NMR spectroscopy
NMR titration experiments were carried out at 298 K and 293 K on a Bruker spectrometer
operating at 800 MHz. NDM-1 were dissolved in 50 mM HEPES, 200 mM NaCl at pH 7.5.
15
The concentration of N-enriched NDM-1 was 0.11 mM and 0.165 mM. All inhibitors were
dissolved in DMSO with concentration of 20 mM and 50 mM. For the titrations of di-Zn-NDM-
1
1 and apo-NDM-1 with A1, a series of H-15N HSQC spectra were acquired. The NDM-
1/inhibitor mole ratios varied from 1: 0 to 1: 0.5, 1: 1.5, 1: 2.5, 1: 4, 1: 6, 1: 8, 1: 10, 1: 12 and
1: 15 for A1. After apo-NDM-1 were titrated with inhibitors, two equivalents of Zn2+ were
added to the solutions.
4.5 Affinity screening by SPR
SPR experiments were carried out using a Biacore T200 (GE Healthcare) at 25 °C. NDM-1
was preconditioned to 50 μg/ml in pH 5.0 sodium acetate and then immobilized to a CM7 chip
by amine coupling. Running buffer was 100 mM Hepes (pH 7.5), 150 mM NaCl, 100 µM ZnCl2,
0.005% P20 v/v. When the affinity screening of fragments was run, 2% DMSO was added to
the buffer. Six concentrations of each fragment were used. Fragment concentration was diluted
by two-fold serial dilution. Compounds were injected to the surface of the protein-coupled chip

channels at the flow rate of 30 µL/min. The affinities (KD) were calibrated with the steady state
affinity 1:1 binding model with the evaluation software of Biacore T200. The affinities of the
fragments were tested once in SPR screen study.
In the binding study of synthesized α-hydrazono carboxylic acids (A1 ~ D4, in Table 2) to
NDM-1, the running buffer was 10 mM Hepes (pH 7.5), 150 mM NaCl, 100 µM ZnCl2 with 2%
DMSO. EDTA-2Na was tested without DMSO in the buffer. Eight two-fold serial diluted
concentrations of each compounds were tested. The affinities of compounds were tested at least
three times.
4.6 NDM-1 Enzyme Inhibition Assay
NDM-1 enzymatic activity inhibition tests were carried on according to the reported
method [21]. Briefly, the inhibitors were added to the NDM-1 (12 nM) and incubated for 20
min at room temperature. L-captopril was used as the parallel positive control. Then, the
substrate (meropenem, 100 μM) was added. The final buffer contained 2% DMSO in 50 mM
hepes (pH 7.24). Assays were performed in 96-well microplate format, and the absorbance was
recorded at 300 nm using a TECAN plate reader (SPARK 20M) at 30 °C. For each compound,
two parallel wells were used. And the tests were replicated for three times. IC50s were calculated
from a plot of percent loss of activity versus inhibitor concentration with GraphPad InStat 3.
4.7 Virtual screening and molecular docking calculations
X-ray crystal structure of NDM-1 (PDB code 4hl2, chain A) was used for virtual screening
and docking studies. The ligand was extracted from the catalytic active site of NDM-1 crystal
structure, which was defined as binding site for virtual screening and docking calculations.
Virtual screening was run with Surflex-Dock of SYBYL-X2.0. Chain A of 4hl2 (PDB code
4hl2) was used to prepare the protein. The ligand ampicillin in the co-crystal structure was
extracted to define the docking site. The H2O701 and two zinc ions were not kept when the
other substructures were removed. Prepare the protein by adding hydrogens. Before virtual
screening, all the fragments were energy minimized with max iterations of 2000, Tripos force
field and Gasteiger-Huckel charges loading. CDOCKER of Discovery Studio 2018 was used
to carry out the docking simulation of the compounds to NDM-1. Chain A of 4hl2 (PDB code
4hl2) was used to prepare the protein. Ampicillin-binding site was defined as the docking site
with site sphere radius of 9. Ligands were prepared with the small molecule tool of Discovery
Studio and fully minimized before docking. Ligands have two energy-minimized
conformations. All the carboxyl groups of the top docking models showed coordination to Zn2.
Ligands in syn conformation showed higher -CDOCKER energy score and was regarded as the

dominating pose.
4.8 Chemistry
General procedure for the syntheses of A1: as described in scheme 3, quinolin-8-ol (1, 2
g, 13.8 mmol) was added to 80% hydrazine hydrate (16 mL). The resulting mixture was heated
to reflux and reacted for 10 h. The reaction was cooled to room temperature. To the reaction
mixture, ~ 10 mL of 3M NaOH was added. Dichloromethane (20 mL) was then added to extract
targeting material. Organic layer was washed with water and brine, and then dried over
anhydrate MgSO4. Organic layer was collected by filtration and concentrated by rotary
evaporator. The crude was then purified by silica column chromatography (petroleum ether:
ethyl acetate = 10: 1 to 1: 1). 8-Hydrazinylquinoline (2) was then got as a brown solid (40%
1
yield). H NMR (400 MHz, CDCl3) δ (ppm): 8.74 – 8.73 (m, 1H), 8.09 (d, 1H, J = 8.4 Hz),
7.46 (t, 1H, J = 8.0 Hz), 7.41- 7.38 (m, 1H), 7.17 (d, 1H, J = 8.0 Hz), 7.10 (d, 1H, J = 7.6 Hz).
8-Hydrazinylquinoline (2, 360 mg, 2.26 mmol) was added to 3 mL of water. HCl (240 µL)
and a solution of 2-oxopropanoic acid (219 mg, 2.49 mmol) in water (0.4 mL) were added
sequentially. Then the mixture was stirred for 2 h at room temperature. Filtrate to collect the
yellow solid crude and wash the crude with water. Vacuum drying gave A1 as a yellow solid
1
(77% yield). Mp: 176.0 – 178.0 ℃. H NMR (600 MHz, d6-DMSO) δ (ppm) (major isomer:
(E)-isomer) 9.95 (s, 1H), 8.94 (dd, 1H, J = 4.2 Hz, 1.8 Hz), 8.54 (d, 1H, J = 8.4 Hz), 7.86 (dd,
1H, J = 7.8 Hz, 0.6 Hz), 7.71 (q, 1H, J = 4.2 Hz), 7.64 (t, 1H, J = 7.8 Hz), 7.57 (t, 1H, J = 7.2
Hz), 2.22 (s, 3H). 13C NMR (150 MHz, d6-DMSO) δ (ppm) (major isomer: (E)-isomer) 165.92,
147.66, 138.42, 137.83, 136.48, 136.15,128.39, 127.97, 122.17, 119.49, 108.25, 11.03. HRMS
+
(ESI) calculated for C12H12N3O2 [M + H] + 230.0924, found 230.0922.
All the other compounds were synthesized with a similar procedure of the synthesis of A1.
For targeting material need further purification, the purification was pursued with semi-
preparative liquid chromatography (Biotage Isolera One). A Biotage SNAP Cartridge KP-C18-
HS 12 g column with a gradient of 10 - 50% acetonitrile in water to was used. Lyophilization
of the eluent gave respective products. The purities of all the compounds were above 95%
(normalization method of peak area, see supporting information). The characterization data are
as followed:
1
A2 is obtained as yellow solid (53% yield). Mp: 123 – 126 ℃. H NMR (500 MHz, d6-
DMSO) δ (ppm) (major isomer) 13.25 (s, 1H), 8.84 (d, J = 2.5 Hz, 1H), 8.32 (d, J = 8.0 Hz,
1H), 7.65 – 7.50 (m, 4H), 7.43 (d, J = 8.0 Hz, 1H), 2.55 (q, J = 7.3 Hz, 2H), 1.18 – 1.13 (m,
3H). 13C NMR (101 MHz, d6-DMSO) (major isomer) δ 165.12, 148.95, 139.95, 138.72, 136.76,

136.57, 128.70, 127.92, 122.49, 119.02, 108.72, 27.01, 12.38. HRMS (ESI) calculated for
+
C13H14N3O2 [M + H] + 244.1081, found 244.1072.
A3 is obtained as yellow solid (53% yield). Mp: 136.3 – 139.6 ℃. 1H NMR (400 MHz, d6-
DMSO) δ (ppm) (major isomer) 13.22 (s, 1H), 8.86 (d, J = 2.8 Hz, 1H), 8.34 (d, J = 8.0 Hz,
1H), 7.68 – 7.54 (m, 4H), 7.46 (d, J = 8.0 Hz, 1H), 1.68 – 1.69 (m, 2H), 1.02 – 0.95 (m, 3H).
13C NMR (101 MHz, d6-DMSO) (major isomer) δ 164.55, 148.57, 139.32, 136.15, 132.52,
128.23, 127.45, 122.07, 118.78, 110.23, 108.38, 35.18, 20.52, 13.69. HRMS (ESI) calculated
+
for C14H16N3O2 [M + H] + 258.1237, found 258.1241.
A4 is obtained as yellow solid (53% yield). Mp: 135.6 – 138.9 ℃. 1H NMR (500 MHz, d6-
DMSO), Z) δ (ppm) 12.51 (s, 1H), 8.84 (s, 1H), 8.33 (d, J = 8.2 Hz, 1H), 7.66 – 7.50 (m, 3H),
7.44 (d, J = 7.9 Hz, 1H), 1.33 (s, 9H). 13C NMR (101 MHz, DMSO) δ 164.95, 148.94, 140.50,
140.15, 136.78, 136.61, 128.65, 127.94, 122.50, 118.93, 108.91, 37.51, 29.24. HRMS (ESI)
+
calculated for C15H18N3O2 [M + H] + 272.1394, found 272.1382.
B1 is obtained as yellowgreen solid (10% yield). Mp: 146.6 – 149.3 ℃. 1H NMR (400 MHz,
d6-DMSO) δ (ppm) (major isomer) 12.20 (br s, 1H), 9.52 (s, 1H), 8.30 (dd, J = 5.9, 3.1 Hz, 1H),
7.89 (dd, J = 5.9, 3.0 Hz, 1H), 7.66 (d, J = 7.3 Hz, 1H), 7.59 – 7.39 (m, 4H), 2.23 (s, 3H).13C
NMR (101 MHz, d6-DMSO) δ (ppm) (major isomer)166.81, 139.86, 134.30, 128.50, 126.68,
126.44, 125.43, 122.65, 121.96, 121.65, 111.53, 11.96. HRMS (ESI) calculated for
+
C13H13N2O2 [M + H] + 229.0972, found 229.0949.
B2 is obtained as yellowgreen solid (21% yield). Mp: 132.0 – 134.7 ℃. 1H NMR (400 MHz,
d6-DMSO) δ (ppm) 13.78 (br s, 1H), 13.10 (s, 1H), 7.93 (d, J = 7.9 Hz, 1H), 7.81 (d, J = 8.2
Hz, 1H), 7.66 – 7.40 (m, 5H), 2.56 (q, J = 7.4 Hz, 2H), 1.18 (t, J = 7.3 Hz, 3H). 13C NMR (101
MHz, d6-DMSO) δ 166.12, 138.73, 134.30, 133.98, 129.09, 127.04, 126.60, 126.35, 121.70,
+
+
121.13, 119.67, 107.64, 26.78, 12.33. HRMS (ESI) calculated for C14H15N2O2 [M + H]
243.1128, found 243.1121.
B3 is obtained as brown solid (25% yield). Mp: 122.7 – 125.4 ℃. 1H NMR (400 MHz, d6-
DMSO) δ (ppm) 13.81 (br s, 1H), 13.20 (s, 2H), 7.92 (d, J = 8.1 Hz, 1H), 7.80 (d, J = 8.1 Hz,
1H), 7.66 – 7.42 (m, 5H), 1.64 (dd, J = 14.9, 7.4 Hz, 2H), 0.97 (t, J = 7.4 Hz, 3H). 13C NMR
(101 MHz, d6-DMSO) δ 166.27, 138.79, 134.31, 129.07, 127.07, 126.59, 126.31, 121.71,
+
121.02, 119.74, 107.57, 35.51, 20.97, 14.18. HRMS (ESI) calculated for C15H17N2O2 [M + H]

+ 257.1285, found 257.1280.
1
B4 is obtained as geen solid (85% yield). Mp: 127.2 – 130.5 ℃. H NMR (500 MHz, d6-
DMSO) δ (ppm) 13.96 (br s, 1H), 12.65 (s, 1H), 7.92 (s, 1H), 7.79 (s, 1H), 7.58 – 7.49 (m, 5H),
1.33 (s, 9H). 13C NMR (101 MHz, d6-DMSO) δ 165.53, 138.78, 138.62, 133.83, 128.59, 126.58,
126.10, 125.81, 121.38, 120.62, 119.31, 107.43, 37.12, 28.81. HRMS (ESI) calculated for
+
C16H19N2O2 [M + H] + 271.1441, found 271.1452.
C1 is obtained as white solid (56% yield). Mp: 177.1 – 180.0 ℃. 1H NMR (400 MHz, d6-
DMSO) δ (ppm) (Only (E)-isomer was detected) 12.12 (br s, 1H), 9.93 (s, 1H), 8.61 (s, 1H),
13
8.11 (s, 1H), 7.73 (d, J = 6.4 Hz, 1H), 7.30 (s, 1H), 2.06 (s, 3H). C NMR (101 MHz, d6-
DMSO) δ 166.11, 141.62, 140.82, 136.09, 123.88, 120.61, 11.63, 0.11. HRMS (ESI) calculated
+
for C8H10N3O2 [M + H] + 180.0768, found 180.0771.
C2 is obtained as yellow solid (24% yield). Mp: 142.2 – 145.6 ℃. 1H NMR (400 MHz, d6-
DMSO) (major isomer) δ (ppm) (major isomer) 12.36 (br s, 1H), 10.96 (s, 1H), 8.91 (s, 1H),
8.38 – 8.36 (m, 2H), 7.90 – 7.87 (m, 1H), 2.66 (q, J = 7.4 Hz, 2H), 1.04 (t, J = 7.6 Hz, 3H). 13C
NMR (101 MHz, d6-DMSO) δ 165.60, 143.90, 143.02, 134.74, 128.96, 127.87, 127.29, 18.91,
+
10.61. HRMS (ESI) calculated for C9H12N3O2 [M + H]+ 194.0924, found 194.0918.
C3 is obtained as white solid (10% yield). Mp: 149.4 – 152.1 ℃.1H NMR (400 MHz, d6-
DMSO) δ (ppm) (major isomer) 12.01 (br s, 1H), 10.08 (s, 1H), 8.62 (d, J = 2.1 Hz, 1H), 8.12
(m, 1H), 7.74 (d, J = 8.4 Hz, 1H), 7.32 – 7.29 (m, 1H), 2.59 – 2.41 (m, 2H), 1.51 – 1.45 (m,
13
2H), 0.96 – 0.92 (m, 3H). C NMR (101 MHz, d6-DMSO) (major isomer) δ 166.36, 141.96,
141.31, 138.57, 136.41, 124.36, 121.13, 35.45, 19.15, 14.33. HRMS (ESI) calculated for
+
C10H14N3O2 [M + H]+ 208.1081, found 208.1079.
C4 is obtained as white solid (32% yield). Mp: 130.9 – 136.2 ℃. 1H NMR (500 MHz, d6-
DMSO) δ (ppm) 11.03 (s, 1H), 8.49 (s, 1H), 8.11 (s, 1H), 7.68 (s, 1H), 7.38 (s, 1H), 1.25 (d, J
13
= 4.8 Hz, 9H). C NMR (101 MHz, d6-DMSO) δ 165.17, 142.85, 141.68, 139.51, 133.84,
+
124.56, 121.25, 36.86, 28.51. HRMS (ESI) calculated for C11H16N3O2 [M + H] + 222.1237,
found 222.1234.
D1 is obtained as yellow solid (78% yield). Mp: 174.0 – 175.3 ℃. 1H NMR (400 MHz, d6-
DMSO) δ (ppm) 11.95 (s, 1H), 9.75 (s, 1H), 7.35 (d, J = 7.8 Hz, 2H), 7.26 (t, J = 7.8 Hz, 2H),
6.88 (t, J = 7.2 Hz, 1H), 2.04 (s, 3H). 13C NMR (101 MHz, d6-DMSO) δ (ppm) 166.29, 144.36,

+
132.34, 128.94, 120.77, 113.83, 11.41. HRMS (ESI) calculated for C9H10N2NaO2 [M + Na]+
201.0634, found 201.0641.
D2 is obtained as yellow solid (7% yield). Mp: 135.2 – 137.9 ℃. 1H NMR (400 MHz, d6-
DMSO) δ (ppm) (major isomer) 11.87 (s, 1H), 9.91 (s, 1H), 7.34 (d, J = 7.6 Hz, 2H), 7.25 (t, J
= 7.6 Hz, 2H), 6.86 (t, J = 7.2 Hz, 2H), 2.57 (q, J = 7.2 Hz, 2H), 1.00 (t, J = 7.6 Hz, 3H). 13C
NMR (101 MHz, d6-DMSO) δ (ppm) 166.43, 144.83, 137.36, 129.38, 121.16, 114.25, 18.01,
+
10.60. HRMS (ESI) calculated for C10H13N2O2 [M + H]+ 193.0972, found 193.0971.
1
D3 is obtained as yellow solid (50% yield). Mp: 75.9 – 78.5 ℃. H NMR (400 MHz, d6-
DMSO) δ (ppm) (major isomer: (E)-isomer) 11.85 (br s, 1H), 9.90 (s, 1H), 7.35 (d, J = 8.0 Hz,
2H), 7.27 – 7.34 (m, 2H), 6.88 (t, J = 7.2 Hz, 1H), 2.57 (t, J = 7.6 Hz, 2H), 1.51 – 1.42 (m, 2H),
13
0.94 (t, J = 7.2 Hz, 3H). C NMR (101 MHz, d6-DMSO) δ (ppm) (major isomer) 166.69,
144.82, 136.25, 129.39, 121.17, 114.27, 35.39, 26.35, 21.00, 19.18, 14.33. HRMS (ESI)
+
calculated for C11H15N2O2 [M + H]+ 207.1128, found 207.1121.
D4 is obtained as yellow solid (74% yield). Mp: 158.4 – 161.7 ℃. 1H NMR (500 MHz, d6-
DMSO) δ (ppm) 13.60 (s, 1H), 11.34 (s, 1H), 7.25 (t, J = 7.7 Hz, 2H), 7.17 (d, J = 7.8 Hz, 2H),
6.86 (t, J = 7.1 Hz, 1H), 1.26 (s, 9H). 13C NMR (101 MHz, d6-DMSO) δ (ppm) 165.45, 144.79,
-
137.97, 129.63, 121.11, 113.60, 37.26, 29.21. HRMS (ESI) calculated for C12H15N2O2 [M-H]-
219.1139, found 219.1131.
4.9 Antimicrobial susceptibility testing
Susceptibility testing was performed by broth microdilution assay according to Clinical and
Laboratory Standards Institute (CLSI) methods (CLSI M07-A9). Briefly, fresh bacterial
overnight colony growth was resuspended in sterile saline, adjusted to a cell density of about 1
× 108 colony-forming units (CFU)/mL. Then the liquid was diluted by Mueller-Hinton Broth
and transferred 100 µL to the wells, with the final target inoculum of about 5 × 105 CFU/mL.
Tested compounds were prepared in 100% DMSO with the stock concentration of 20 mg/mL.
The compounds were added to the wells with final compound concentration of 100 µg/mL. The
two-fold diluted final concentrations of meropenem were 64 to 0.0625 µg/mL. Assay microtiter
plates were incubated for 20 – 24 hours at 37°C. Following incubation, assay plates were
monitored for bacterial growth. The minimal inhibitory concentration (MIC) is defined as the

lowest concentration of antibiotic at which the visible growth of the organism is completely
inhibited.
Appendix A. Supplementary data
Chemical shift perturbations of NDM-1 observed by NMR spectroscopy. NMR spectra of
compounds in this study. Analytical HPLC trace for compounds in this study.
Acknowledgements
We gratefully acknowledge the grants from the National Natural Science Foundation of China
(81703418, 21575156) and the CAMS Innovation Fund for Medical Sciences (CIFMS, 2017-I2M-
1-012).
References
[1] L. S. J. Roope, R. D Smith, K. B. Pouwels, J. Buchanan, L. Abel, P. Eibich, C. C. Butler, P. S. Tan, A. S. Walker,
J. V. Robotham, S Wordsworth, The challenge of antimicrobial resistance: what economics can contribute. Science.
364(6435) (2019) pii: eaau4679.
[2] M. Hancock, “Antimicrobial resistance needs an urgent global response” (speech given at World Economic
Forum in Davos, 24 January 2019); www.gov.uk/government/speeches/antimicrobial-resistance-needs-an-urgent-
global-response.
[3] World Health Organization, Global priority list of antibiotic-resistant bacteria to guide research, discovery, and
development of new antibiotics. Feb 27th, 2017.
[4] E. Tacconelli, E. Carrara, A. Savoldi, S. Harbarth, M. Mendelson, D. L. Monnet, C. Pulcini, G. Kahlmeter, J.
Kluytmans, Y. Carmeli, M. Ouellette, K. Outterson, J. Patel, M. Cavaleri, E. M. Cox, C. R. Houchens, M. L. Grayson,
P. Hansen, N. Singh, U. Theuretzbacher, N. Magrini, and the WHO Pathogens Priority List Working Group,
Discovery, research, and development of new antibiotics: the WHO priority list of antibiotic-resistant bacteria and
tuberculosis. Lancet. Infect. Dis. 18 (2018) 318-327.
[5] F. S. Codjoe, E. S. Donkor, Carbapenem resistance: a review. Med. Sci. (Basel). 6 (2017) pii: E1.
[6] G. Meletis, Carbapenem resistance: overview of the problem and future perspectives. Ther. Adv. Infect. Dis. 3
(2016) 15-21.
[7] M. I. El-Gamal, I. Brahim, N. Hisham, R. Aladdin, H. Mohammed, A Bahaaeldin, Recent updates of carbapenem
antibiotics. Eur. J. Med. Chem. 131 (2017) 185-195.

[8] G. Cornaglia, H. Giamarellou, G. M. Rossolini, Metallo-β-lactamases: a last frontier for β-lactams? Lancet. Infect.
Dis. 11 (2011) 381-393.
[9] D. Yong, M. A. Toleman, C. G. Giske, H. S. Cho, K. Sundman, K. Lee, T. R. Walsh, Characterization of a new
metallo-beta-lactamase gene, bla(NDM-1), and a novel erythromycin esterase gene carried on a unique genetic
structure in Klebsiella pneumoniae sequence type 14 from India. Antimicrob. Agents. Chemother. 53 (2009) 5046-
5054.
[10] P. Linciano, L. Cendron, E. Gianquinto, F. Spyrakis, D. Tondi, Ten years with New Delhi Metallo-β-
lactamase‑1 (NDM-1): from structural insights to inhibitor design. ACS Infect. Dis. 5 (2019) 9-34.
[11] S. M. Drawz, R. A. Bonomo, Three decades of beta-lactamase inhibitors. Clin. Microbiol. Rev. 23 (2010) 160-
201.
[12] K. Bush, P. A. Bradford, β-Lactams and β-lactamase inhibitors: an overview. Cold. Spring. Harb. Perspect.
Med. 6 (2016) pii: a025247.
[13] Y. Guo, J. Wang, G. Niu, W. Shui, Y. Sun, H. Zhou, Y. Zhang, C. Yang, Z. Lou, Z. Rao, A structural view of
the antibiotic degradation enzyme NDM-1 from a superbug. Protein Cell. 2 (2011) 384-394.
[14] N. Li, Y. Xu, Q. Xia, C. Bai, T. Wang, L. Wang, D. He, N. Xie, L. Li, J. Wang, H. G. Zhou, F. Xu, C. Yang,
Q. Zhang, Z. Yin, Y. Guo, Y. Chen, Simplified captopril analogues as NDM-1 inhibitors. Bioorg. Med. Chem. Lett.
24 (2014) 386-389.
[15] D. T. King, L. J. Worrall, R. Gruninger, N. C. Strynadka, New Delhi metallo-β-lactamase: structural insights
into β-lactam recognition and inhibition. J. Am. Chem. Soc. 134 (2012) 11362-11365.
[16] C. Yao, Q. Wu, G. Xu, C. Li, NMR backbone resonance assignment of New Delhi metallo-beta-lactamase.
Biomol. NMR Assign. 11 (2017) 239-242.
[17] J. Brem, R. Cain, S. Cahill, M. A. McDonough, I. J. Clifton, J. C. Jiménez-Castellanos, M. B. Avison, J. Spencer,
C. W. Fishwick, C. J. Schofield, Structural basis of metallo-β-lactamase, serine-β-lactamase and penicillin-binding
protein inhibition by cyclic boronates. Nat. Commun. 7 (2016) 12406.
[18] M. M. González, M. Kosmopoulou, M. F. Mojica, V. Castillo, P. Hinchliffe, I. Pettinati, J. Brem, C. J. Schofield,
G. Mahler, R. A. Bonomo, L. I. Llarrull, J. Spencer, A. J. Vila, Bisthiazolidines: a substrate-mimicking scaffold as
an inhibitor of the NDM-1 carbapenemase. ACS Infect. Dis. 1 (2015) 544-554.
[19] J. Zhang, S. Wang, Q. Wei, Q. Guo, Y. Bai, S. Yang, F. Song, L. Zhang, X. Lei, Synthesis and biological
evaluation of aspergillomarasmine A derivatives as novel NDM-1 inhibitor to overcome antibiotics resistance.
Bioorg. Med. Chem. 25 (2017) 5133-5141.

[20] T. Christopeit, A. Albert, H. S. Leiros, Discovery of a novel covalent non-β-lactam inhibitor of the metallo-β-
lactamase NDM-1. Bioorg. Med. Chem. 24 (2016) 2947-2953.
[21] F. M. Klingler, T. A. Wichelhaus, D. Frank, J. Cuesta-Bernal, J. El-Delik, H. F. Müller, H. Sjuts, S. Göttig, A.
Koenigs, K. M. Pos, D. Pogoryelov, E. Proschak, Approved drugs containing thiols as inhibitors of metallo-β-
lactamases: strategy to combat multidrug-resistant bacteria. J. Med. Chem. 58 (2015) 3626-3630.
[22] L. Zhai, Y. L. Zhang, J. S. Kang, P. Oelschlaeger, L. Xiao, S. S. Nie, K. W. Yang, Triazolylthioacetamide: a
valid scaffold for the development of new delhi metallo-β-lactmase-1 (NDM-1) inhibitors. ACS Med. Chem. Lett.
7 (2016) 413-417.
[23] V. Yarlagadda, P. Sarkar, S. Samaddar, G. B. Manjunath, S. D. Mitra, K. Paramanandham, B. R. Shome, J.
Haldar, Vancomycin analogue restores meropenem activity against NDM-1 gram-gegative pathogens. ACS Infect.
Dis. 4 (2018) 1093-1101.
[24] A. Krajnc, J. Brem, P. Hinchliffe, K. Calvopiña, T. D. Panduwawala, P. A. Lang, J. J. A. G. Kamps, J. M.
Tyrrell, E. Widlake, B. G. Saward, T. R. Walsh, J. Spencer, C. J. Schofield, Bicyclic Boronate VNRX-5133 Inhibits
Metallo- and Serine-β-Lactamases. J Med Chem. 62 (2019) 8544-8556.
[25] H. Meredith, D. S., Antimicrobial Activity of Cefepime in Combination with VNRX-5133 Against a Global
2018 Surveillance Collection of Clinical Isolates; European Congress of Clinical Microbiology and Infectious
Diseases (ECCMID) in Amsterdam: The Netherlands (2019).
[26] A. Piccirilli, B. Segatore, D. Daigle, G. Amicosante, M. Perilli, VNRX-5133 maintains potent inhibitory activity
in engineered NDM-1 variants with increased cefepime hydrolytic efficiency. European Congress of Clinical
Microbiology and Infectious Diseases (ECCMID) in Amsterdam: The Netherlands (2019).
[27] X. Wang, M. Lu, Y. Shi, Y. Ou, X. Cheng, Discovery of novel New Delhi Metallo-β-Lactamases-1 inhibitors
by multistep virtual screening. PLoS One. 10 (2015) e0118290.
[28] J. S. Kang, A. L. Zhang, M. Faheem, C. J. Zhang, N. Ai, J. D. Buynak, W. J. Welsh, P. Oelschlaeger, Virtual
screening and experimental testing of B1 Metallo-β-lactamase inhibitors. J. Chem. Inf. Model. 58 (2018) 1902-1914.
[29] F. Spyrakis, G. Celenza, F. Marcoccia, M. Santucci, S. Cross, P. Bellio, L. Cendron, M. Perilli, D. Tondi,
Structure-based virtual screening for the discovery of novel inhibitors of New Delhi Metallo-β-lactamase-1. ACS
Med. Chem. Lett. 9 (2018) 45-50.
[30] C. Shi, J. Bao, Y. Sun, X. Kang, X. Lao, H. Zheng, Discovery of baicalin as NDM-1 inhibitor: virtual screening,
biological evaluation and molecular simulation. Bioorg. Chem. 88 (2019) 102953.

[31] C. Shi, J. Chen, B. Xiao, X. Kang, X. Lao, H. Zheng, Discovery of NDM-1 inhibitors from natural products. J.
Glob. Antimicrob. Resist. 18 (2019) 80-87.
[32] P. K. Thakur, J. Kumar, D. Ray, F. Anjum, M. I. Hassan, Search of potential inhibitor against New Delhi
metallo-beta-lactamase 1 from a series of antibacterial natural compounds. J. Nat. Sci. Biol. Med. 4 (2013) 51-56.
[33] A. U. Khan, A. Ali, Danishuddin, G. Srivastava, A. Sharma, Potential inhibitors designed against NDM-1 type
metallo-β-lactamases: an attempt to enhance efficacies of antibiotics against multi-drug-resistant bacteria. Sci. Rep.
7 (2017) 9207.
[34] M. T. Rehman, M. F. AlAjmi, A. Hussain, G. M. Rather, M. A. Khan. High-throughput virtual screening,
molecular dynamics simulation, and enzyme kinetics identified ZINC84525623 as a potential inhibitor of NDM-1.
Int J Mol Sci. 20 (2019) pii: E819.
[35] D. Rogolino, M. Carcelli, M. Sechi, N. Neamati, Viral enzymes containing magnesium: metal binding as a
successful strategy in drug design. Coordin. Chem. Rev. 256 (2012) 3063–3086.
[36] D. P. Martin, P. G. Blachly, J. A. McCammon, S. M. Cohen, Exploring the influence of the protein environment
on metal-binding pharmacophores. J. Med. Chem. 57 (2014) 7126-7135.
[37] J. A. Jacobsen, J. L. Fullagar, M. T. Miller, S. M. Cohen, Identifying chelators for metalloprotein inhibitors
using a fragment-based approach. J. Med. Chem. 54 (2011) 591-602.
[38] C. A. Lipinski, F. Lombardo, B. W. Dominy, P. J. Feeney, Experimental and computational approaches to
estimate solubility and permeability in drug discovery and development settings. Adv. Drug. Deliv. Rev. 23 (1997)
3-25.
[39] D. F. Veber, S. R. Johnson, H. Y. Cheng, B. R. Smith, K. W. Ward, K. D Kopple, Molecular properties that
influence the oral bioavailability of drug candidates. J. Med. Chem. 45 (2002) 2615-2623.
[40] U. H. Danielson, Fragment library screening and lead characterization using SPR biosensors. Curr. Top. Med.
Chem. 9 (2009) 1725-1735.
[41] T. Christopeit, H. K. Leiros, Fragment-based discovery of inhibitor scaffolds targeting the metallo-β-lactamases
NDM-1 and VIM-2. Bioorg. Med. Chem. Lett. 26 (2016) 1973-1977.
[42] A. Bax, G. M. Clore, Protein NMR: Boundless opportunities. J. Magn. Reson. pii: S1090-7807 (2019) 30150-
30158.
[43] H. Arthanari, K. Takeuchi, A. Dubey, G. Wagner, Emerging solution NMR methods to illuminate the structural
and dynamic properties of proteins. Curr. Opin. Struct. Biol. 52 (2019) 1-11.

[44] H. Feng, J. Ding, D. Zhu, X. Liu, X. Xu, Y. Zhang, S. Zang, D. C. Wang, W. Liu, Structural and mechanistic
insights into NDM-1 catalyzed hydrolysis of cephalosporins. J. Am. Chem. Soc. 136 (2014) 14694-14697.
[45] R. A. Taylora, N. M. Bonanno, D. Mirza, A. J. Lough, M. T. Lemaire, A trivalent cobalt complex with the new
redox-active ligand 10-(8-quinolylazo)-9-phenanthrol (qapl). Polyhedron. 131 (2017) 34-39.
[46] J. W. Lyga, A Convenient synthesis of 1-aryl-Δ2-1,2,4-triazolin-5-ones from arylhydrazines. Synthetic.
Commun. 16 (1986) 163-167.
[47] No author. Captopril approved for hypertension. FDA Drug Bull. 11 (1981) 10-11.
[48] M. F. Richter, B. S. Drown, A. P. Riley, A. Garcia, T. Shirai, R. L. Svec, P. J. Hergenrother. Predictive
compound accumulation rules yield a broad-spectrum antibiotic. Nature. 545(2017) 299-304.
[49] R. O'Shea, H. E. Moser, Physicochemical properties of antibacterial compounds: implications for drug
discovery. J. Med. Chem. 51(2008) 2871-2878.
[50] A. Y. Chen, R. N. Adamek, B. L. Dick, C. V. Credille, C. N. Morrison, S. M. Cohen, Targeting metalloenzymes
for therapeutic intervention. Chem. Rev. 119 (2019) 1323-1455.

Declaration of interests
 The authors declare that they have no known competing financial interests or personal
relationships that could have appeared to influence the work reported in this paper.
☐The authors declare the following financial interests/personal relationships which may be
considered as potential competing interests:

A novel scaffold of NDM-1 inhibitors is described. The scaffold was discovered from metal-
binding pharmacophore enriched library with virtual and SPR screen. Solution NMR found that the
scaffold binds to the NDM-1 active pocket and the zinc ion in the pocket. Design and syntheses
found compounds with synergistic antibacterial activity.