Anti-MRSA drug discovery by ligand-based virtual screening and biological evaluation

Article abstract of DOI:10.1016/j.bioorg.2021.105042

S. aureus resistant to methicillin (MRSA) is one of the most-concerned multidrug resistant bacteria, due to its role in life-threatening infections. There is an urgent need to develop new antibiotics against MRSA. In this study, we firstly compiled a data set of 2,3-diaminoquinoxalines by chemical synthesis and antibacterial screening against S. aureus, and then performed cheminformatics modeling and virtual screening. The compound with the Specs ID of AG-205/33156020 was discovered as a new antibacterial agent, and was further identified as a Gyrase B (GyrB) inhibitor. In light of the common features, we hypothesized that the 6c as the representative of 2,3-diaminoquinoxalines also inhibited GyrB and eventually proved it. Via molecular docking and molecular dynamics simulations, we identified binding modes of AG-205/33156020 and 6c to the ATPase domain of GyrB. Importantly, these GyrB inhibitors inhibited the MRSA strains and showed selectivity to HepG2 and HUVEC. Taken together, this research work provides an effective ligand-based computational workflow for scaffold hopping in anti-MRSA drug discovery, and discovers two new GyrB inhibitors that are worthy of further development.

Full text of DOI:10.1016/j.bioorg.2021.105042

Bioorganic Chemistry 114 (2021) 105042
Contents lists available at ScienceDirect
Bioorganic Chemistry
journal homepage: www.elsevier.com/locate/bioorg
Anti-MRSA drug discovery by ligand-based virtual screening and
biological evaluation
,
,c
Xu Liana, Zhonghua Xiab, Xueyao Lia, Pavel Karpovb c, Hongwei Jind, Igor V. Tetkob
,
,
,*
Jie Xiaa *, Song Wua
a State Key Laboratory of Bioactive Substance and Function of Natural Medicines, Department of New Drug Research and Development, Institute of Materia Medica,
Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing 100050, China
b
¨
Institute of Structural Biology, Helmholtz Zentrum München-Research Center for Environmental Health (GmbH), Ingolstadter Landstraße 1, 85764 Neuherberg,
Germany
c BIGCHEM GmbH, Valerystr. 49, 85716 Unterschleißheim, Germany
d State Key Laboratory of Natural and Biomimetic Drugs, School of Pharmaceutical Sciences, Peking University, Beijing 100191, China
A R T I C L E I N F O
A B S T R A C T
Keywords:
S. aureus resistant to methicillin (MRSA) is one of the most-concerned multidrug resistant bacteria, due to its role
in life-threatening infections. There is an urgent need to develop new antibiotics against MRSA. In this study, we
firstly compiled a data set of 2,3-diaminoquinoxalines by chemical synthesis and antibacterial screening against
S. aureus, and then performed cheminformatics modeling and virtual screening. The compound with the Specs ID
of AG-205/33156020 was discovered as a new antibacterial agent, and was further identified as a Gyrase B
(GyrB) inhibitor. In light of the common features, we hypothesized that the 6c as the representative of 2,3-dia-
minoquinoxalines also inhibited GyrB and eventually proved it. Via molecular docking and molecular dynamics
simulations, we identified binding modes of AG-205/33156020 and 6c to the ATPase domain of GyrB.
Importantly, these GyrB inhibitors inhibited the MRSA strains and showed selectivity to HepG2 and HUVEC.
Taken together, this research work provides an effective ligand-based computational workflow for scaffold
hopping in anti-MRSA drug discovery, and discovers two new GyrB inhibitors that are worthy of further
development.
Antibiotic resistance
MRSA
Antibacterial agent
Virtual screening
DNA Gyrase inhibitors
1. Introduction
pneumonia, and has led to remarkable increases in morbidity, mortality
as well as overall healthcare costs [4]. At present, several drugs are
The increasing emergence of antibiotic resistance poses a serious
threat to public health worldwide [1]. Among the bacteria that devel-
oped antibiotic resistance, the ESKAPE pathogens, i.e., E. faecium, S.
aureus, K. pneumoniae, A. baumannii, P. aeruginosa and Enterobacter
species are the most important, as they are involved in many life-
threatening infections and difficult to treat [2]. The ESKAPE patho-
gens resistant to clinically used antibiotics were covered by the World
Health Organization (WHO) “priority pathogens” list for R&D of new
antibiotics [3]. According to this list, S. aureus resistant to methicillin
(MRSA) belongs to the class of “high-priority” pathogens.
available for the treatment of MRSA infections in clinic (e.g. vancomy-
cin, daptomycin and linezolid). However, the safety concerns have
limited their use in clinical practice [4]. Though new antibacterial
agents with favorable toxicity profiles are in late clinical development
phases, they are derived from previously approved drug classes and their
mechanisms are the same [5]. To avoid cross-resistance, anti-MRSA
agents with new chemotypes or new modes of action are urgently
needed.
Virtual screening is a widely used strategy for hit identification [6],
and has been proved fast and effective for discovery of new-chemotype
antibacterial agents [7,8]. As such, we applied this strategy to discover
novel anti-MRSA agents. Unlike other studies that completely used the
known chemical structures with known activity as references for
MRSA is a major cause of both community and hospital-acquired
infections such as complicated skin and skin structure infections,
bacteremia, diabetic foot infection and community acquired
* Corresponding authors at: Institute of Materia Medica, Chinese Academy of Medical Sciences & Peking Union Medical College, No. 2 Nanwei Road, Beijing
100050 China.
E-mail addresses: jie.william.xia@hotmail.com (J. Xia), ws@imm.ac.cn (S. Wu).
https://doi.org/10.1016/j.bioorg.2021.105042
Received 5 February 2021; Received in revised form 17 May 2021; Accepted 26 May 2021
Available online 3 June 2021
0045-2068/© 2021 Elsevier Inc. All rights reserved.

X. Lian et al.
Bioorganic Chemistry 114 (2021) 105042
building ligand-based models [9–11], we newly compiled a set of
compounds for cheminformatics modeling by chemical synthesis and
bacterial growth inhibition assay. The chemical series that we focused
on was the 2,3-diaminoquinoxaline (cf. Fig. 1). Recently, several de-
rivatives were reported as antibacterial agents against S. aureus
RCMB010010 [12]. Due to the rather limited chemical data and the lack
of knowledge on mode of action, however, no cheminformatics
modeling work based on this chemical series had been done. Based on
the data set, we built the common-feature pharmacophore models with
six active 2,3-diaminoquinoxalines in the modeling set, and determined
the optimal model by its performance evaluation with the actives and
inactive 2,3-diaminoquinoxalines in the test set. Apart from this anal-
ysis, we also used the most potent derivative to generate a shape-based
model and FCFP_6 fingerprints. By integrating the models into a
computational workflow, we performed in silico and in vitro screening to
identify diverse antibacterial agents against S. aureus with scaffolds
different from the 2,3-diaminoquinoxalines. We also studied mode of
action of the hit as well as the 2,3-diaminoquinoxalines, and proposed
plausible binding modes by molecular docking and molecular dynamics
simulations. In order to highlight the clinical significance of this
research work, we presented the antibacterial activity against clinically-
isolated MRSA and the cytotoxicity profile.
spectra as the signals from 7.20 to 8.32. As for compounds 6 g–l, the
unique methyl groups on the quinoxalines can be identified in both 1H
NMR and 13C NMR spectra. To be specific, their proton signals were
respectively shown at δ 2.45, 2.45, 2.47, 2.43, 2.43 and 2.43 ppm in 1H
NMR, and their carbon signals were observed at δ 21.41, 21.41, 21.42,
21.40, 19.23, 21.40 ppm in 13C NMR spectra. With regard to compounds
7a–c, the signals of two NH protons appeared at the positions different
from those of compounds 6a–f in the 1H NMR spectra. The two peaks
were located at δ 8.77 ppm and 7.18 ppm for both 7a and 7c, and at 8.84
ppm and 7.21 ppm for 7b. The other features of compounds 7a-c
included two signals of the ethyl groups in the 13C NMR spectra, i.e., δ
36.37 and 14.67 ppm for 7a, δ 36.38 and 14.67 ppm for 7b, and δ 36.38
and 14.66 ppm for 7c. The structures of compounds 8a–c were easy to
validate from the 1H NMR spectra, according to the peaks at δ 12.36,
12.37, 12.33 ppm (for the NH protons of sulfonamide) and the signals at
δ 9.10, 9.23, 8.93 ppm (for the NH protons of the phenylamino groups).
Their chemical structures and MICs are listed in Table 1. Although
the structures of 6a, 6b, 6c, 6e and 6f were previously reported, all of
them were never tested against S. aureus ATCC29213 [12,14,17]. As
shown in the table, 6a-l showed potent antibacterial activity against
S. aureus ATCC29213 (MICs: 0.1–3.13
S. aureus ATCC29213 at the concentrations between 50
g/mL. However, 8a-c were not active, with MICs greater than 100
μ
g/mL). 7a-c could inhibit
g/mL and 100
g/
μ
μ
μ
2. Results and discussions
mL. The above data demonstrated a preliminary relationship between
the chemical structures of 2,3-diaminoquinoxalines and the antibacte-
rial activity for S. aureus ATCC29213. To be specific, (i) symmetrical
substitutions of the 2,3-dichloroquinoxaline with the phenylamino
groups were optimal for antibacterial activity (6a-l vs. 7a-c/8a-c).
When the phenylamino group on one side was replaced with either
ethylamino- or sulfonamide- group, the antibacterial activity was
significantly reduced. (ii) The electron-withdrawing substituents at the
phenylamino group are better for antibacterial activity than the
electron-donating groups, e.g. 6a-d vs. 6e-f, 6 g-i vs. 6 k-l. This set of
compounds represents an ideal set of compounds for cheminformatics
modeling.
2.1. 2,3-diaminoquinoxalines for computational modeling
Since pharmacophore modeling required a certain number of active
and inactive compounds, we synthesized 18 derivatives with the 2,3-dia-
minoquinoxaline as the core scaffold and tested their antibacterial ac-
tivity in term of MIC against S. aureus ATCC29213, a wild-type and
methicillin-sensitive S. aureus (MSSA) strain, by the broth microdilution
method [13].
The synthetic routes of the derivatives were designed and performed
according to the previous publications [14–16], with minor modifica-
tion. As shown in Scheme 1, the synthesis of 6a-l started with the re-
action between o-phenylenediamines (1a-b) and oxalic acid for the
production of the 2,3-Quinoxalinediones (2a-b) [12]. By the chlorina-
tion of the 2,3-Quinoxalinediones and the substitution with various
anilines to the 2,3-dichloroquinolines (3a-b), the target compounds
were obtained. Notably, we used DMF as the solvent and aluminum
trichloride as the catalyst in the last step, which led to a high reaction
rate and a good yield (59%-80%). As for 7a-c, the un-substituted 2,3-
dichloroquinoline reacted with ethyl amine to produce the 3-chloro-N-
ethylquinoxalin-2-amine (4), which was further substituted by different
anilines at the position 3. The synthesis of 8a–c was based on the same
scheme as 7a–c, except that 4-chlorobenzenesulfonamide was used to
replace ethyl amine as the reagent in the reaction. All the target com-
pounds were structurally validated by melting points, 1H NMR, 13C NMR
and HRMS.
2.2. In silico and in vitro screening for antibacterial agents
2.2.1. Computational models
We used Catalyst/HipHop module of Discovery Studio (v16.1.0,
`
Dassault Systemes Biovia Corp) to generate 10 common-feature phar-
macophore models, based on six 2,3-diaminoquinoxalines with potent
antibacterial activity against S. aureus, i.e. 6a-f. The features of the
pharmacophore models are shown in Table 2. The ranks of the phar-
macophore models were very close, with 75.92 as the maximum and
73.26 as the minimum. The values of Direct Hit, Partial Hit and Max Hit
were the same for 10 models.
To facilitate model selection, the other 12 2,3-diaminoquinoxaline
derivatives, i.e. 6 g-l, 7a-c and 8a-c were used as a test set for the
evaluation of model performances in discriminating actives from in-
actives. According to Fig. 2a, it is easy to identify pharm_07 as the
optimal model as it assigned the maximal values (4.7–5.0) to the actives
whereas the minimal values (0.05–2.5) to the inactives. This model was
composed of one hydrophobic feature for aromatic rings and two gen-
eral hydrophobic features, one hydrogen bond donor, one hydrogen
bond acceptor (cf. Fig. 2b).
The NH protons of 6a-f revealed peaks in 1H NMR spectra at δ 9.15,
9.20, 9.37, 9.93, 9.81 and 8.96 ppm, respectively. The corresponding
1
–
aromatic C H protons of these compounds were displayed in H NMR
We used ROCS to build the shape-based model, with the lowest-
energy conformer generated by Discovery Studio as the query. The
model is shown in Fig. 2c.
2.2.2. Computational workflow
The computational workflow consisted of pharmacophore filtering,
shape matching and fingerprints-based similarity search (cf. Scheme 2).
We screened the Specs chemical library (~210,000 compounds) for
potential antibacterial agents with the computational workflow. A total
of 45,550 Specs compounds with their corresponding conformers/
Fig. 1. The 2,3-dianilinoquinoxaline scaffold in antibacterial agents
against S. aureus.
2

X. Lian et al.
Bioorganic Chemistry 114 (2021) 105042
Scheme 1. The general synthetic route of the 2,3-diaminoquinoxalines. (a) 4 N HCl, oxalic acid, reflux, 4 h; (b) SOCl2, DMF; (c) Ethyl amine, EtOH, reflux, 8 h; (d) 4-
chlorobenzenesulfonamide, DMF, K2CO3, reflux, 4 h; (e) R2-PhNH2, AlCl3, DMF.
stereoisomers that matched Pharm_07 passed the pharmacophore
filtering. We only selected 5,901 compounds with the pharmacophore
FitValues>3.2. By matching the conformers to the shape-based models,
we picked 2500 top-scoring compounds. Furthermore, we calculated 2D
similarity in term of Tanimoto coefficient (Tc) between these com-
pounds and compound 6c based on FCFP-6 fingerprints. To identify hits
with core scaffolds different from the 2,3-diaminoquinoxaline, we
excluded the most similar compounds with the Tc values greater than
0.3 (i.e. 2,3-diaminoquinoxalines). From the other compounds with Tc
values less than 0.3, we selected 503 compounds (Tc: 0.24–0.12). The
structural clustering into 25 clusters based on FCFP_6 fingerprints
facilitated the selection of structurally diverse compounds. By visual
inspection, we selected one or two chemical structures from each cluster
by giving priority to those with greater FitValues and shape Tc values,
less fingerprint similarity as well as good synthetic feasibility. We
selected 15 Specs compounds that were commercially available and had
never been previously reported as antibacterial agents. The FitValues of
these compounds ranged from 3.40 to 4.38, indicating of a relatively
high pharmacophore similarity (Table 3). Their shape similarity values
(Tc) were between 0.57 and 0.69, which demonstrated that they
generally fit the shape of the most active derivative 6c. Besides, the
minimum of the fingerprint similarity (Tc) was 0.13, while the
maximum was 0.22. It implied that the compounds somewhat shared
common fingerprints with the most active derivative 6c but also con-
tained unique scaffolds.
validated the effectiveness of our ligand-based computational workflow
for scaffold hopping.
2.3. Mode of action
2.3.1. Inhibition of S. aureus DNA Gyrase B
As mentioned above, AG-205/33156020 was never tested against
S. aureus. In order to understand its mode of action, we further analyzed
the fragments of its chemical structure. It was interesting to see that this
compound contained the fragment named arylaminotriazine, which
appeared in the Gyrase inhibitors from Astrazeneca [18]. A few publi-
cations pointed out that Astrazeneca arylaminotriazines inhibited DNA
supercoiling by targeting the ATP binding site of GyrB [19,20]. With this
evidence, we tested it for its inhibitory activity on GyrB. As a result, it
showed moderate inhibition, with the IC50 value of 1.1 ± 0.1 μM (cf.
Fig. 3a). Because its chemical structure well matched the optimal
common-feature pharmacophore model and the shape-based model
derived from the 2,3-diaminoquinoxalines, we also tested the repre-
sentative compound 6c against GyrB. As expected, 6c was a GyrB in-
hibitor, with an IC50 value of 5.2 ± 0.1 μM (cf. Fig. 3a).
2.3.2. Plausible binding modes
The data above demonstrated that AG-205/33156020 and 6c were
two new GyrB inhibitors with moderate antibacterial activity against
S. aureus. With the crystal structure of GyrB (PDB code: 4URO), we
performed molecular docking of the two compounds against GyrB with
OEDocking module in OpenEye and then molecular dynamics (MD)
simulations for 100 ns with AMBER 20 to further analyze the pro-
tein–ligand interactions. The RMSDs of the heavy atoms in GyrB and the
two ligands as a function of time are shown in Fig. 3b, where the protein
structure or ligand structure in every frame is superimposed on the
corresponding starting conformation by the CPPTRAJ program [21]. As
for the system of GyrB in complex with AG-205/33156020, it reached
equilibrium after about 40 ns. The system of GyrB in complex with 6c
reached a plateau after approximately 80 ns. The respective pro-
tein–ligand binding mode after equilibrium was extracted and is shown
2.2.3. Experimentally validated hits
We purchased and tested 15 potential hits for their antibacterial
activity against S. aureus ATCC29213. As shown in Table 3, the com-
pound with the Specs ID of AG-205/33156020 was experimentally
validated as an antibacterial agent, and its MIC value was 6.25 μg/mL.
Though not as potent as some of the 2,3-diaminoquinoxalines, it was
structurally different from the 2,3-diaminoquinoxalines, with the
fingerprint similarity (Tc) as low as 0.16. Fig. S3 shows AG-205/
33156020 was well mapped to the pharmacophore model Pharm_07
(FitValue: 3.72). The successful identification of the diverse hit
3

X. Lian et al.
Bioorganic Chemistry 114 (2021) 105042
in Fig. 3c.
Table 1
Chemical structures of the 2,3-diaminoquinoxalines and their antibacterial ac-
AG-205/33156020 binds to GyrB via forming (1) the hydrogen
bonds between the phenolic hydroxyl group and Asp81/Gly85 as well as
between the triazine fragment and His124, (2) the hydrophobic in-
teractions with Pro87, Ala98, His124, Ile51, Ile86 and Ile175. 6c binds
at the entrance of the ATP binding site of GyrB. As expected, its 2,3-dia-
minoquinoxaline scaffold plays the most essential role in protein–ligand
binding as it forms the electrostatic interaction with Arg84 and the
hydrogen bonds with Lys118 and Glu58. It should be noted that only one
of the secondary amine groups interacts with the protein. Two
dichlorophenyl groups interact with Thr173/Ile175 and Lys118/Ile102
via hydrophobic interactions, respectively. The details of the pro-
tein–ligand interactions from MD simulations may provide insights for
further structural optimization.
tivity against S. aureus ATCC29213 in term of MIC (
μg/mL).
Compound ID
R1
R2
S. aureus ATCC29213(
μ
g/mL)
Usage
Modeling set
6a
6b
6c
6d
6e
6f
H
4-Cl
0.2
H
4-Br
0.2
H
3,4-diCl
3,4-diF
4-Me
4-Et
0.1
H
0.2
H
3.13
3.13
0.39
0.2
2.4. Anti-MRSA activity and cytotoxicity
H
6 g
6 h
6i
CH3
CH3
CH3
CH3
CH3
CH3
4-Cl
Test set
4-Br
To explore whether the two GyrB inhibitors were promising for
further development, we tested their antibacterial activity against clin-
ical isolates of MRSA by the broth microdilution assay [13], and their
cytotoxicity to two mammalian cell lines by the sulforhodamine B (SRB)
assay [22]. Table 4 lists the MIC values for three clinical isolates of
MRSA. As shown in the table, both compounds can inhibit the growth of
3,4-diCl
4-Me
4-Et
0.2
6j
1.56
1.56
3.13
6 k
6 l
4-
Propyl
MRSA below the concentration of 10
last-resort antibiotics, i.e. vancomycin, AG-205/33156020 achieved
the MIC values of 4–8 g/mL (or 9.5–19.0 M). 6c was equivalent to
vancomycin in term of potency, with the MIC value of 0.5 g/mL (or 1.1
M). Regarding the cytotoxicity, the CC50 values of AG-205/33156020
for human hepatocellular carcinoma (HepG2) cells and human umbilical
vein endothelial cells (HUVEC) were 21 M and 34 M, while the values
of 6c were 4.9 M and 10 M, respectively. According to the above data,
μg/mL. Though not as potent as the
μ
μ
μ
μ
7a
7b
7c
H
H
H
4-F
100
50
4-Cl
4-Br
μ
μ
50
μ
μ
it is concluded that both hits showed anti-MRSA activity and were
somewhat selective to HepG2 and HUVEC, but remained to be
optimized.
3. Conclusion
8a
8b
8c
H
H
H
4-F
>100
>100
>100
MRSA has been defined by WHO as “high-priority” for new antibi-
otics discovery and development. This is somewhat attributed to cross-
resistance, i.e. bacterial resistance to all the drugs of the same chemo-
type or mode of action. Discovery of novel antibacterial chemotypes
may overcome this issue. In this study, we comprehensively employed
chemical synthesis, cheminformatics analysis, ligand-based virtual
screening and biological evaluation to identify novel anti-MRSA hits.
The 2,3-diaminoquinoxalines were previously reported as a new
class of antibacterial agents against S. aureus [12], but the amount of the
available derivatives was insufficient for cheminformatics modeling. To
close the gap in experimental data, we first compiled a data set
composed of 18 compounds with the 2,3-diaminoquinoxaline as the
3,4-F
4-Me
Table 2
10 common-feature pharmacophore models generated with the Catalyst/
HipHop module in Discovery Studio.
Pharmacophore
Model
Featuresa
Rank
Direct Hitb
Partial Hitc
Max Fit
pharm_01
pharm_02
pharm_03
pharm_04
pharm_05
pharm_06
pharm_07
pharm_08
pharm_09
pharm_10
XXZDH
XZZDH
XXZDA
RXZDH
XZZDA
XZZDH
XZZDH
XXZDH
XXZDH
RXZDA
75.92
74.75
74.72
74.46
73.55
73.43
73.35
73.34
73.31
73.26
111111
111111
111111
111111
111111
111111
111111
111111
111111
111111
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
5
5
5
5
5
5
5
5
5
5
scaffold and with the MIC values ranging from 0.1 μg/mL to > 100 μg/
mL, by chemical synthesis and antibacterial screening against S. aureus
ATCC29213. Based on the data set, we built 10 common-feature phar-
macophore models with the DS/HipHop module and selected pharm_07
out as the optimal model. We then integrated the pharmacophore model,
the shape-based model and the FCFP-6 fingerprints of 6c into the
computational workflow, screened the Specs chemical library and
identified a structurally different compound with moderate antibacterial
activity against S. aureus ATCC29213, namely AG-205/33156020.
Furthermore, we performed bioassay to understand the mode of
action and demonstrated that AG-205/33156020 was a S. aureus GyrB
inhibitor. Due to the same pharmacophore features and shape shared by
6c, we hypothesized and eventually validated GyrB as one protein target
of 6c. This is quite interesting as the protein targets of the 2,3-diamino-
quinoxalines had never been reported before this study. By molecular
docking and molecular dynamics simulations, we have proposed the
most plausible binding modes of the two GyrB inhibitors, which would
a
X, hydrophobic from aromatic rings; Z, general hydrophobic; D, hydrogen
bond donor; H, hydrogen bond acceptor lipid; A, hydrogen bond acceptor; R,
ring aromatic.
b
Direct Hit: 1, a ligand fully matches the pharmacophore; 111111, all the
ligands directly match the pharmacophore;
c
Partial Hit: 1, a ligand partially rather than fully matches the pharmaco-
phore; 000000, no ligand partially matches the pharmacophore.
4

X. Lian et al.
Bioorganic Chemistry 114 (2021) 105042
Fig. 2. The models of the computational workflow. (a) The 2,3-diaminoquinoxalines in the test set are mapped to every pharmacophore model. (b) The optimal
common-feature pharmacophore model. The model is composed of one hydrophobic feature from aromatic rings (blue) and two general hydrophobic features (light
blue), one hydrogen bond donor (purple), one hydrogen bond acceptor (green). The most active 2,3-diaminoquinoxaline 6c is mapped to the model. (c) The shape-
based model from the lowest-energy conformation of 6c. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of
this article.)
4. Materials and methods
4.1. Chemistry
4.1.1. General methods
All the reactions were monitored by thin layer chromatography
(TLC) on the silica gel plates GF254 (0.20 mm, Yantai Chemical Industry
Research Institute, China). High resolution mass spectrometry (HRMS)
was performed by the LC/MSD TOF mass spectrometer system (Agilent
Technologies Inc., USA). 1H NMR spectra were all recorded on Bruker
Avance III 500 spectrometer (Varian Mercury, USA) at the frequency of
500 MHz. 13C NMR (101 MHz or 126 MHz) spectra were recorded on
Bruker 400 or 500 spectrometer. For the spectrometry, DMSO‑d6 was
used as the solvent and tetramethylsilane was used as an internal stan-
dard. The column chromatography on silica gel (200–300 mm; Qingdao
Haiyang Chemical Co., Ltd, China) was used for the purification of
products. All the solvents and organic reagents were commercially
available and not further purified.
4.1.2. General procedure for the synthesis of intermediates 2a-b
The intermediates were prepared according to the synthetic routes
proposed by El-Atawy M.A. et al. [12]. The o-phenylenediamine or 4-
methylbenzene-1,2-diamine (1a/1b, 1 equiv) and anhydrous oxalic
acid (1 equiv) were dissolved in 4 N HCl (100 mL) in a round-bottom
flask in a size of 250 mL. The mixture was refluxed at 100 ◦C for 4 h
Scheme 2. The computational workflow for virtual screening.
while being stirred. The reaction mixture was concentrated under
reduced pressure. The residue was washed with ethanol (50 mL) and
dried, which afforded 1,4-dihydroquinoxaline-2,3-dione (2a/2b) as
white solid.
be helpful for hit-to-lead optimization. Further evaluation of anti-MRSA
activity of the two new GyrB inhibitors and cytotoxicity to HepG2 and
HUVEC has shown they effectively and somewhat selectively inhibit the
growth of the MRSA strains. As no GyrB inhibitor was approved for
clinical use after the withdrawing of novobiocin [23], both AG-205/
33156020 and 6c may serve as good starting structures for development
of new antibiotics.
4.1.2.1. 1,4-dihydroquinoxaline-2,3-dione (2a). Yield 83%, white and
powder-like solid. 1H NMR (500 MHz, DMSO‑d6) δ 11.93 (s, 2H), 7.12
(m, 4H). 13C NMR (126 MHz, DMSO‑d6) δ 155.79, 155.49, 132.73,
125.86, 124.20, 123.71, 115.60, 115.43, 20.98.
5

X. Lian et al.
Bioorganic Chemistry 114 (2021) 105042
Table 3
15 potential hits selected from the computational workflow. The chemical structures, the similarity based on pharmacophore features, shape features and 2D fin-
gerprints as well as their antibacterial activity (MIC) against S. aureus ATCC29213 are listed.
Specs ID
Structure
Similarity
S. aureus ATCC29213
(μ
g/mL)b
Pharmacophore
(FitValue)
Shape (Tc)a
0.68
Fingerprint
(Tc)
AT-057/43468612
3.40
0.14
>100
AO-476/41339614
AS-710/43364291
4.38
3.50
0.61
0.69
0.14
0.13
>100
>100
AG-205/40649626
AO-081/41480557
3.58
3.96
0.58
0.58
0.13
0.18
>100
>100
AK-968/41927008
AE-641/40793067
3.90
4.08
0.60
0.61
0.17
0.15
>100
>100
AG-205/33156020
AG-690/12245283
3.72
3.75
0.62
0.60
0.16
0.17
6.25
>100
AH-487/41801452
AO-476/42169377
4.00
4.05
0.65
0.64
0.13
0.18
>100
50
AG-690/11231009
AN-584/43492661
4.03
4.00
0.65
0.57
0.22
0.19
>100
>100
AP-064/41806449
AG-690/08506044
3.93
4.22
0.64
0.59
0.14
0.16
>100
>100
a
Tanimoto coefficient as a metric of structural similarity.
6

X. Lian et al.
Bioorganic Chemistry 114 (2021) 105042
b
The assay was performed in duplicate and vancomycin was the positive drug (MIC: 1.56 μg/mL).
Fig. 3. The molecular mechanism of compounds
AG-205/33156020
and
6c.
(a)
The
concentration-dependent ATPase inhibition of
S. aureus Gyrase B (GyrB). The calculated IC50
values of AG-205/33156020 and 6c are 1.1 ±
0.1 μM and 5.2 ± 0.1 μM, respectively. (b) Heavy-
atom RMSDs of S. aureus GyrB and the ligands as
a function of time during the 100-ns MD simula-
tions. Each RMSD was calculated with the start-
ing conformation/pose as the reference. (c)
Plausible binding modes of the two ligands to the
ATP binding site of S. aureus GyrB as derived
from MD simulations. Images were created with
Discovery Studio 2016. The interacting residues
and the ligands are shown in stick representa-
tions. Color codes: orange, residues; green, AG-
205/33156020; purple, 6c. (For interpretation of
the references to color in this figure legend, the
reader is referred to the web version of this
article.)
4.1.2.2. 6-methyl-1,4-dihydroquinoxaline-2,3-dione (2b). Yield 74%,
white and powder-like solid. 1H NMR (500 MHz, DMSO‑d6) δ11.85 (d, J
= 5.9 Hz, 2H), 7.01 (d, J = 8.2 Hz, 1H), 6.92 (s, 1H), 6.87 (d, J = 8.2 Hz,
1H), 2.26 (s, 3H). 13C NMR (126 MHz, DMSO‑d6) δ155.79, 155.49,
132.73, 125.86, 124.2, 123.71, 115.60, 115.43, 20.98.
Table 4
Antibacterial activity (MIC) and cytotoxicity (CC50) of two newly-identified
GyrB inhibitors.
Compound
ID
MIC (μg/mL)
CC50
(
μM, mean ±
SD)a
MRSA
MRSA
MRSA
HepG2
HUVEC
4.1.3. General procedure for the synthesis of intermediates 3a-b
The intermediates were prepared according to the reference [24].
The compounds 2a/2b (1 equiv.) was placed in a round-bottom flask in a
size of 250 mL and then SOCl2 (10 equiv.) was added. The reaction
mixture was stirred at 70 ◦C for 4 h, followed by the concentration under
reduced pressure. The residue was washed with n-hexane (50 mL) and
then dried, which afforded 2,3-dichloroquinoxaline (3a/3b) as a pale
yellow solid.
15–1
15–2
15–3
AG-205/
33156020
6c
4
8
4
21 ± 3
34 ± 0.5
10 ± 2
n.d.
0.5
0.5
0.5
1
0.5
1
4.9 ±
0.4
vancomycin
a
n.d.
values represent cytotoxicity after 72-hour treatment with the compounds.
mean, the average of duplicate; SD, standard deviation.
4.1.3.1. 2,3-dichloroquinoxaline (3a). Yield 90%, pale-yellow solid. 1H
NMR (500 MHz, DMSO‑d6) δ 8.13 (m, 2H), 8.00 (dd, J = 21.0, 10.7 Hz,
7

X. Lian et al.
Bioorganic Chemistry 114 (2021) 105042
2H). 13C NMR (126 MHz, DMSO‑d6) δ 145.08, 140.50, 132.25, 128.39,
28.14, 16.24. HRMS calcd for C24H24N4 [M + H]+, 369.2074; found,
123.44, 115.58.
369.2069
4.1.3.2. 2,3-dichloro-6-methylquinoxaline (3b). Yield 77%, pale-yellow
solid. 1H NMR (500 MHz, DMSO‑d6) δ 7.98 (t, J = 7.0 Hz, 1H), 7.87
(d, J = 6.1 Hz, 1H), 7.80 (d, J = 8.5 Hz, 1H), 2.59 (s,3H). 13C NMR (126
MHz, DMSO‑d6) δ 144.86, 143.97, 142.91, 140.58, 138.95, 134.30,
127.8, 127.12, 21.76.
4.1.4.7. N2,N3-bis(4-chlorophenyl)-6-methylquinoxaline-2,3-diamine
(6g). Yield 62%, yellow-green solid. m.p.: 228–231 ℃. 1H NMR (500
MHz, DMSO‑d6) δ 9.42 (s, 2H), 8.01 (t, J = 9.1 Hz, 4H), 7.50 (d, J = 8.2
Hz, 1H), 7.46 (dd, J = 9.1, 2.9 Hz, 4H), 7.43 – 7.40 (m, 1H), 7.23 (d, J =
8.2 Hz, 1H), 2.45 (s, 3H). 13C NMR (101 MHz, DMSO) δ 141.02, 139.91,
139.87, 136.32, 136.21, 135.42, 134.27, 134.19, 128.91, 127.60,
126.40, 126.27, 122.38, 122.27, 21.41.HRMS calcd for C21H6N4Cl2 [M
+ H]+, 395.0825; found, 395.0815
4.1.4. General procedure for the synthesis of target compounds 6a-l
The synthesis of 6a-l were performed according to the earlier pub-
lication [14]. A suspension of compounds 3a-b (1 equiv.), aromatic
amine (1 equiv.) and AlCl3 (1.1 equiv.) in DMF (50 mL) was stirred at
110◦ C for 8 h. The reaction was quenched by cold water (50 mL). The
product was extracted with ethyl acetate (3 × 75 mL). The organic phase
was washed with water (2 × 20 mL) and brine (30 mL), dried with
anhydrous Na2SO4. The solvent was evaporated under reduced pressure
and then the solid was purified by flash column chromatography with
petroleum ether/ethyl acetate (25:1) as the eluent.
4.1.4.8. N2,N3-bis(4-bromophenyl)-6-methylquinoxaline-2,3-diamine
(6h). Yield 67%, light yellow solid. m.p.: 234–236 ℃.1H NMR (500
MHz, DMSO‑d6) δ 9.49 (s, 2H), 7.97 (t, J = 8.8 Hz, 4H), 7.58 (m, J = 9.0,
3.0 Hz, 4H), 7.50 (d, J = 8.3 Hz, 1H), 7.42 (s, 1H), 7.24 (d, J = 8.3 Hz,
1H), 2.45 (s, 3H).13C NMR (101 MHz, DMSO‑d6) δ141.49, 140.98,
140.30, 136.09, 135.47, 134.17, 131.80, 127.63, 125.46, 125.28,
122.80, 122.69, 114.40, 114.26, 21.41. HRMS calcd for C21H16Br2N4
[M + H]+ , 482.9814; found: 482.9842
4.1.4.1. N2,N3-bis(4-chlorophenyl)quinoxaline-2,3-diamine (6a). Yield
71%, light yellow solid. m.p.: 236–239 ℃. 1H NMR (500 MHz,
DMSO‑d6) δ 9.15 (s, 2H), 7.96 (d, J = 8.6 Hz, 4H), 7.60 (dd, J = 6.1, 3.5
4.1.4.9. N2,N3-bis(3,4-dichlorophenyl)-6-methylquinoxaline-2,3-diamine
(6i). Yield 81%, light yellow solid. m.p.: 169–171 ℃. 1H NMR (500
MHz, DMSO‑d6) δ 9.22 (s, 1H), 9.19 (s, 1H), 8.32 (d, J = 2.5 Hz, 1H),
8.28 (d, J = 2.5 Hz, 1H), 7.86 (td, J = 8.3, 2.5 Hz, 2H), 7.65 (dd, J = 8.8,
3.2 Hz, 2H), 7.53 (d, J = 8.3 Hz, 1H), 7.47 – 7.43 (m, 1H), 7.29 (dd, J =
8.3, 2.0 Hz, 1H), 2.47 (s, 3H).13C NMR (101 MHz, DMSO‑d6) δ141.31,
141.07, 140.98, 140.75, 136.42, 136.16, 134.50, 131.26, 130.90,
128.26, 125.85, 125.63, 123.91, 123.77, 121.47, 121.35, 120.49,
120.36, 21.42. HRMS calcd for.C21H14Cl4N4 [M + Na]+, 484.9973;
found: 484.9858
Hz, 2H), 7.48 (d, J = 8.6 Hz, 4H), 7.40 (dd, J = 6.2, 3.5 Hz, 2H). 13
C
NMR (101 MHz, DMSO‑d6) δ 141.52, 139.67, 136.57, 128.97, 126.47,
125.98, 122.40. HRMS calcd for C20H14Cl2N4 [M + H]+, 381.0666;
found, 381.0668
4.1.4.2. N2,N3-bis(4-bromophenyl)quinoxaline-2,3-diamine (6b). Yield
74%, light yellow solid. m.p.: 253–255 ℃. 1H NMR (500 MHz,
DMSO‑d6) δ 9.20 (s, 2H), 7.92 (d, J = 8.2 Hz, 4H), 7.60 (d, J = 8.6 Hz,
6H), 7.43 – 7.37 (m, 2H).13C NMR (101 MHz, DMSO‑d6) δ141.50,
140.11, 136.55, 131.86, 126.01, 125.98, 122.79, 114.43. HRMS calcd
for C20H14Br2N4 [M + H]+, 468.9658; found, 468.9623
4.1.4.10. 6-methyl-N2,N3-di-p-tolylquinoxaline-2,3-diamine (6j). Yield
80%, light yellow solid. m.p.: 161–163 ℃. 1H NMR (500 MHz,
DMSO‑d6) δ 8.89 (s, 1H), 8.85 (s, 1H), 7.78 (dd, J = 10.4, 8.2 Hz, 4H),
7.43 (d, J = 8.2 Hz, 1H), 7.37 – 7.33 (m, 1H), 7.22 (dd, J = 8.4, 2.5 Hz,
4H), 7.20 – 7.14 (m, 1H), 2.43 (s, 3H), 2.34 (d, J = 2.9 Hz, 6H). 13C NMR
(101 MHz, DMSO) δ 141.73, 141.22, 138.24, 138.17, 136.55, 134.75,
134.60, 131.95, 131.83, 129.51, 127.03, 125.44, 125.35, 121.16,
121.08, 21.40, 20.99.HRMS calcd for C25H26N4 [M + H]+, 355.1917;
found: 355.1926
4.1.4.3. N2,N3-bis(3,4-dichlorophenyl)quinoxaline-2,3-diamine
(6c).
Yield 74%, yellow-green solid. m.p.: 151–153 ℃. 1H NMR (400 MHz,
DMSO‑d6) δ 9.37 (s, 2H), 8.31 (d, J = 2.5 Hz, 2H), 7.89 (dd, J = 8.8, 2.6
Hz, 2H), 7.68 – 7.58 (m, 4H), 7.44 (dd, J = 6.2, 3.5 Hz, 2H).13C NMR
(101 MHz, DMSO‑d6) δ141.35, 140.95, 136.48, 131.26, 130.94, 126.55,
126.16, 124.04, 121.62, 120.61. HRMS calcd for C20H12N4Cl4 [M + H]+,
448.9889; found, 448.9883
4.1.4.11. N2,N3-bis(4-ethylphenyl)-6-methylquinoxaline-2,3-diamine
(6k). Yield 45%, yellow-green solid. m.p.: 155–158 ℃. 1H NMR (500
MHz, DMSO‑d6) δ 8.91 (s, 1H), 8.86 (s, 1H), 7.80 (t, J = 8.3 Hz, 4H), 7.43
(d, J = 8.2 Hz, 1H), 7.35 (s, 1H), 7.28 – 7.23 (m, 4H), 7.17 (d, J = 8.2 Hz,
1H), 2.64 (qd, J = 7.5, 2.5 Hz, 4H), 2.43 (s, 3H), 1.24 (t, J = 7.7 Hz,
6H).13C NMR (101 MHz, DMSO) δ 141.09, 140.57, 138.20, 138.08,
137.01, 136.91, 135.53, 134.11, 133.59, 126.94, 125.72, 123.90,
123.79, 119.82, 27.16, 19.23, 14.18. HRMS calcd for C25H26N4 [M +
H]+, 383.2230; found, 383.2224.
4.1.4.4. N2,N3-bis(3,4-difluorophenyl)quinoxaline-2,3-diamine
(6d).
Yield 63%, yellow-green solid. m.p.: 216–218 ℃. 1H NMR (500 MHz,
DMSO‑d6) δ 9.93 (s, 2H), 8.32 (d, J = 7.6 Hz, 1H), 8.29 (d, J = 7.2 Hz,
1H), 7.84 – 7.78 (m, 2H), 7.63 (dt, J = 6.6, 3.4 Hz, 2H), 7.49 (d, J = 9.6
Hz, 1H), 7.46 (d, J = 9.6 Hz, 1H), 7.41 (dd, J = 6.3, 3.5 Hz, 2H).13C NMR
(126 MHz, DMSO)
δ 141.35, 137.92, 135.65, 126.11, 125.53,
123.47,117.67, 117.30, 115.60, 109.70.HRMS calcd for C20H12N4F4 [M
+ H]+, 385.1071; found, 385.1064
4.1.4.5. N2,N3-di-p-tolylquinoxaline-2,3-diamine (6e). Yield 73%, light
yellow solid. m.p.: 146–148 ℃. 1H NMR (500 MHz, DMSO‑d6) δ 9.81 (s,
2H), 7.77 (d, J = 7.9 Hz, 4H), 7.52 (dt, J = 6.2, 3.3 Hz, 2H), 7.29 (dd, J
= 6.3, 3.4 Hz, 2H), 7.20 (d, J = 7.8 Hz, 4H), 2.30 (s, 6H).13CNMR (101
MHz, DMSO‑d6) δ141.73, 137.48, 132.81, 129.67, 125.62, 121.75,
21.03. HRMS calcd for C22H20 N4 [Mꢀ H]-, 339.1615; found: 339.1605
4.1.4.12. 6-methyl-N2,N3-bis(4-propylphenyl)quinoxaline-2,3-diamine
(6l). Yield 72%, light yellow solid. m.p.: 144–146 ℃. 1H NMR (500
MHz, DMSO‑d6) δ 8.91 (s, 1H), 8.87 (s, 1H), 7.81 (t, J = 8.3 Hz, 4H), 7.44
(d, J = 8.1 Hz, 1H), 7.39 – 7.35 (m, 1H), 7.23 (dd, J = 8.4, 3.2 Hz, 4H),
7.17 (dd, J = 8.3, 1.8 Hz, 1H), 2.62 – 2.55 (m, 3H), 2.43 (s, 3H), 1.64 (h,
J = 7.5 Hz, 4H), 0.95 (t, J = 7.3 Hz, 6H).13C NMR (101 MHz, DMSO‑d6)
δ141.67, 141.15, 138.50, 138.41, 136.75, 136.56, 134.74, 134.62,
128.87, 127.03, 125.44, 125.36, 121.01, 120.89, 37.25, 24.69, 21.40,
14.15. HRMS calcd for C27H30N4 [M + H]+, 411.2543; found, 411.2537.
4.1.4.6. N2,N3-bis(4-ethylphenyl)quinoxaline-2,3-diamine
(6f). Yield
58%, light yellow solid. m.p.: 127–130 ℃. 1H NMR (500 MHz,
DMSO‑d6) δ 8.96 (s, 1H), 7.82 (d, J = 8.2 Hz, 2H), 7.54 (dd, J = 6.2, 3.5
Hz, 1H), 7.34 (dd, J = 6.1, 3.6 Hz, 1H), 7.27 (d, J = 8.1 Hz, 2H), 2.64 (q,
J = 7.6 Hz, 2H), 1.25 (t, J = 7.6 Hz, 3H). 13C NMR (101 MHz, DMSO‑d6)
δ 141.67, 138.53, 138.27, 136.66, 128.32, 125.73, 125.38, 121.26,
4.1.5. General procedure for the synthesis of target compounds 7a–c
According to the literature [15], a solution of 70% ethyl amine (2
equiv.) was dropwise added to a solution of compound 3a (1 equiv.)
8

X. Lian et al.
Bioorganic Chemistry 114 (2021) 105042
dissolved in absolute ethanol (50 mL). The reaction mixture was then
refluxed for 8 h. The solvent was evaporated under reduced pressure.
The residue was purified by flash column chromatography with petro-
leum ether/ethyl acetate (10:1) as the eluent to afford 3-chloro-N-eth-
ylquinoxalin-2-amine (4) as a white solid. The intermediate 4 was
directly used for the next reaction. It was dissolved in DMF (50 mL), to
which aromatic amine (1 equiv) and AlCl3 (1.1 equiv.) were added. The
mixture was stirred at 110◦ C for 8 h, then the reaction was quenched by
cold water (50 mL). The product was extracted with ethyl acetate (3 ×
75 mL), and the organic phase was washed with water (2 × 20 mL) and
brine (30 mL), dried with anhydrous Na2SO4. The solvent was evapo-
rated under reduced pressure and the solid was purified by flash column
chromatography with petroleum ether/ethyl acetate (10:1) as the
eluent.
7.34 (m, 2H).13C NMR (126 MHz, DMSO) δ 150.25, 150.15, 148.33,
148.22, 146.66, 144.64, 141.61, 140.57, 137.85, 136.53, 129.54,
128.91, 126.56, 126.33, 126.21, 117.62, 117.48, 110.02. HRMS calcd
for C20H13N4O2F2S [M + H]+, 447.0489; found:447.0487.
4.1.6.3. 4-chloro-N-(3-(p-tolylamino)quinoxalin-2-yl)benzenesulfonamide
(8c). Yield 70%, light yellow solid. m.p.: 226–228 ℃. 1H NMR (500
MHz, DMSO‑d6) δ 12.33 (s, 1H), 8.93 (s, 1H), 8.15 (d, J = 8.3 Hz, 2H),
7.96–7.88 (m, 1H), 7.85 (d, J = 8.0 Hz, 2H), 7.69 (d, J = 8.3 Hz, 2H),
7.56 (d, J = 7.8 Hz, 1H), 7.40 (t, J = 7.4 Hz, 1H), 7.36 (t, J = 7.5 Hz, 1H),
7.20 (d, J = 8.1 Hz, 2H), 2.32 (s, 3H).13C NMR (101 MHz, DMSO) δ
141.71, 140.65, 137.83, 136.76, 132.86, 129.56, 129.51, 128.96,
126.53, 126.01, 125.76, 121.19, 114.20, 109.80, 20.99. HRMS calcd for
C
21H17N4O2SCl [M + Na]+, 447.0653; found, 447.0656.
4.1.5.1. N2-ethyl-N3-(4-fluorophenyl)quinoxaline-2,3-diamine
(7a).
Yield 59%, yellow solid. m.p.: 138-140℃. 1H NMR (500 MHz, DMSO‑d6)
δ 8.77 (s, 1H), 7.94 (dd, J = 9.0, 5.0 Hz, 2H), 7.50 (d, J = 7.9 Hz, 2H),
7.30 (t, J = 7.5 Hz, 1H), 7.27 – 7.21 (m, 3H), 7.18 (d, J = 5.0 Hz, 1H),
3.58 (qd, J = 7.3, 4.7 Hz, 2H), 1.34 (t, J = 7.2 Hz, 3H). 13C NMR (101
MHz, DMSO‑d6) 13C NMR (101 MHz, DMSO) δ144.19, 141.04, 137.85,
137.14, 137.11, 135.54, 125.74, 125.29, 125.18, 124.02, 122.42,
122.34, 115.73, 115.51, 36.37, 14.67. HRMS calcd for C16H15N4F
[Mꢀ H]-, 281.1194; found, 281.1208
4.2. Computational modeling
4.2.1. Pharmacophore model generation
The Catalyst/HipHop implemented in Discovery Studio was used to
generate common feature pharmacophores based on the ligands in the
modeling set. The ligands were prepared by using the “prepare ligand”
module. This module added hydrogen atoms to the chemical structures
and generated the protonated state at the pH of 7.4. The “Principal” and
“MaxOmitFeat” properties were assigned to the ligands according to
their MIC values for S. aureus ATCC29213. To be specific, the values of 2
and 0 were respectively assigned to the “Principal” and the ‘Max-
OmitFeat’ attributes of 6a-d, which indicated that these compounds
were highly active and none of the pharmacophore features from them
were allowed to omit in model generation. For the other two moderately
active compounds, i.e. 6e and 6f, both the “Principal” and the “Max-
OmitFeat” attributes were set as 1. The “Fast” method was applied to
quickly generate a maximum of 255 diverse low-energy conformations
of each ligand, within the relative energy threshold of 20 kcal/mol. The
“Feature Mapping” module was applied to identify possible locations of
different pharmacophore features in the generated ligand conforma-
tions. From 272 features, hydrogen bond acceptor, hydrogen bond
donor, general hydrophobic features including those from aromatic
rings and from aliphatic chains, ring aromatic features were the most
frequent and important features. In the module of “Common Feature
Pharmacophore Generation”, the above-mentioned features were
selected and the allowed number of each feature in a pharmacophore
was set as 0 to 5. A maximum of 10 models were allowed to generate by
this module.
4.1.5.2. N2-ethyl-N3-(4-chlorophenyl)quinoxaline-2,3-diamine
(7b).
Yield 78%, light yellow solid. m.p.: 142–144 ℃. 1H NMR (500 MHz,
DMSO‑d6) δ 8.84 (s, 1H), 7.98 (d, J = 8.7 Hz, 2H), 7.52 (dd, J = 13.0, 8.0
Hz, 2H), 7.45 (d, J = 8.6 Hz, 2H), 7.32 (t, J = 7.4 Hz, 1H), 7.26 (t, J =
7.5 Hz,1H),7.21 (t, J = 4.9 Hz, 1H), 3.58 (qd, J = 7.4, 4.9 Hz 2H), 1.33
(t, J = 7.3 Hz, 3H). 13C NMR(101 MHz, DMSO‑d6) δ144.27, 140.77,
139.84, 138.0, 135.4, 128.9, 126.2, 125.87, 125.57, 125.24, 124.07,
121.98, 36.38, 14.67. HRMS calcd for C16H15N4Cl4 [Mꢀ H]-, 297.0912;
found, 297.0911
4.1.5.3. N2-ethyl-N3-(4-bromophenyl)quinoxaline-2,3-diamine
(7c).
Yield 77%, light yellow solid. m.p.: 150–153 ℃. 1H NMR (500 MHz,
DMSO‑d6) δ 8.84 (s, 1H), 7.98 (d, J = 8.6 Hz, 2H), 7.52 (dd, J = 13.0, 7.9
Hz, 2H), 7.45 (d, J = 8.6 Hz, 2H), 7.32 (t, J = 7.4 Hz, 1H), 7.26 (t, J =
7.4 Hz, 1H), 7.21 (t, J = 4.9 Hz, 1H), 3.58 (qd, J = 7.4, 5.0 Hz, 2H), 1.33
(t, J = 7.2 Hz, 3H).13C NMR(101 MHz, DMSO‑d6) δ144.29, 140.73,
140.26, 137.96, 135.37, 131.83, 125.89, 125.61, 125.24, 124.08,
122.39, 114.12, 36.38, 14.66. HRMS calcd for C16H15BrN4 [M + H]+
,
343.0553; found, 343.0541
The pharmacophore models were evaluated with
a data set
composed of 12 newly-synthesized active/inactive compounds. The
“Ligand Profiler” module in Discovery Studio was run to map all the
compounds to every generated pharmacophore model. In the module,
the parameter of “Maximum Omitted Features” was set to ꢀ 1, indicating
all feature subsets of the pharmacophore were considered in model
evaluation. Other parameters were set as default. The pharmacophore
model that can best discriminate actives from inactives by FitValue was
regarded as the optimal model.
4.1.6. General procedure for the synthesis of target compounds 8a–c
The synthesis of 8a-c was almost the same as that of 7a-c, except for a
solution of chlorobenzenesulfonamide was used to react with compound
3a.
4.1.6.1. 4-chloro-N-(3-((4-fluorophenyl)amino)quinoxalin-2-yl)benzene-
sulfonamide (8a). Yield 61%, light yellow solid. m.p.: 244–247 ℃. 1
H
NMR (500 MHz, DMSO‑d6) δ 12.36 (s, 1H), 9.10 (s, 1H), 8.16 (d, J = 8.3
Hz, 2H), 8.06 – 7.98 (m, 2H), 7.97–7.84 (m, 1H), 7.70 (d, J = 8.4 Hz,
2H), 7.56 (d, J = 7.7 Hz, 1H), 7.39 (p, J = 7.1 Hz, 2H), 7.23 (t, J = 8.8
Hz, 2H). 13C NMR (101 MHz, DMSO) δ 159.83, 157.41, 137.86, 135.80,
135.76, 135.72, 129.56, 128.83, 126.55, 126.04, 125.94, 123.32,
115.69, 115.47. HRMS calcd for C20H14ClFN4O2S [M + H]+ , 428.0499;
found, 428.0483
4.2.2. Shape generation
The lowest-energy conformation of the most active derivative 6c was
generated by the “Quick Minimization” tool in Discovery Studio. Based
on the conformation, ROCS (version 3.3.1.2, OpenEye Scientific Soft-
ware Inc., Santa Fe, NM, USA) was used to generate the shape-based
model. During the shape generation, no additional editing of the
model was performed.
4.1.6.2. 4-chloro-N-(3-((3,4-difluorophenyl)amino)quinoxalin-2 yl)ben-
zenesulfonamide (8b). Yield 73%, yellow-green solid. m.p.: 253–255 ℃.
1H NMR (500 MHz, DMSO‑d6) δ 12.37 (s, 1H), 9.23 (s, 1H), 8.31 – 8.21
(m, 1H), 8.16 – 8.10 (m, 2H), 7.94 – 7.89 (m, 1H), 7.85 – 7.77 (m, 1H),
7.71 – 7.64 (m, 2H), 7.63 – 7.55 (m, 1H), 7.47 – 7.39 (m, 1H), 7.42 –
4.2.3. Virtual screening
The Specs chemical library that included more than 210,000 com-
pounds (version Jun. 2019, accessed at http://www.specs.net) was used
for the virtual screening. The compounds in the library were prepared by
9

X. Lian et al.
Bioorganic Chemistry 114 (2021) 105042
the “Prepare Ligands” module in DS. The ligand preparation included
the generation of all protonated states at the pH of 7.4 and the
enumeration of all potential stereoisomers. A multi-conformer database
of the prepared structures was built by the “Build 3D Database” module
in Discovery Studio. In the database, each prepared structure is repre-
sented by a maximum of 100 conformers. As the first step, all the con-
formers in the 3D database were mapped to the pharmacophore model
(i.e. Pharm_07) by a rigid fit algorithm (“FAST” search) implemented in
the “Search 3D Database” module in DS. The similarity of pharmaco-
phore features was measured by the metric of FitValue. All the con-
formers with the FitValues greater than zero were put out. For the
conformers/stereoisomers that belong to the same Specs ID, only the
conformer/stereoisomer that best matched the pharmacophore model
was saved. Then, a certain number of compounds that passed the
pharmacophore filter were mapped to the shape-based model by ROCS.
ShapeTanimoto was the scoring function to measure shape similarity.
Subsequently, the “Find Similar Molecules by Fingerprints” module in
DS was used for 2D similarity search. In this module, the most active
derivative 6c in the modeling set was set as the reference. FCFP-6 was
the molecular fingerprinting algorithm and Tc was the metric to mea-
sure similarity. In order to identify diverse hits, only the compounds
with the Tc value less than 0.3 were retained. Structural clustering based
on FCFP-6 fingerprints was performed to generate 25 clusters, from
which the potential hits were selected by taking pharmacophore Fit-
Value, shape Tc and fingerprints Tc as well as synthetic feasibility,
commercial availability into consideration.
(kcal/mol)/Å2. Each MD simulation consisted of the relaxation phase,
the equilibrium phase, and the sampling phase. At the relaxation phase,
the system was heated gradually from 0 K to 300 K, with a 5-ps simu-
lation for each 50-K increase in temperature and a force constant of 2
(kcal/mol)/Å2. At the temperature of 300 K, the system was simulated
for 500 ps. At the equilibrium phase, the system was simulated at 1 bar
for 10 ns, including 5-ns simulation with restraints and 5-ns simulation
without any restraint. Lastly, a 100-ns MD simulation was performed for
each system with no restraint. A total of 10 000 snapshots were recorded
at this sampling phase. During the simulation, the temperature was
regulated by the weak-coupling algorithm and the pressure was regu-
lated by the isotropic position scaling algorithm with a pressure relax-
ation time of 1.0 ps. The integration of the equations of motion was
conducted at a time interval of 0.5 fs at the heating phase and 2 fs for the
other phases. All bonds were constrained using the SHAKE algorithm.
The particle-mesh Ewald (PME) method was used to calculate long-
range electrostatic interactions.
4.3. Biology
4.3.1. Bacterial growth inhibition assay
The broth microdilution assay recommended by the Clinical and
Laboratory Standards Institute [13] was performed to determine the
MICs of the compounds. The 2-fold dilutions of the compound dissolved
in MH broth medium/DMSO (100
μ
L) were added to 12 wells of the 96-
L)
to each well. The resulting solutions of the compound were at the con-
centrations from 100 to 0.05 g/mL (for S. aureus ATCC29213) or 64 to
well plate, followed by the addition of the bacterial suspension (100
μ
4.2.4. Molecular docking
μ
The X-ray structure of S. aureus GyrB in complex with novobiocin
(PDB code: 4URO) was downloaded from the Protein Data Bank
(https://www.rcsb.org). Then, the identical protein chains and the
cocrystallized water molecules were deleted, and the cognate ligand
(novobiocin) was stripped from the crystal structure and saved for future
use. The “Clean Protein” tool of Discovery Studio was used to solve
potential problems in the protein structure such as nonstandard names,
incomplete residues, nonstandard atom orders, alternate conformations,
as well as incorrect connectivity and bond orders, modify all hydrogen
atoms and terminal residues, and generate the protonation state at pH
7.0.
0.03 μg/mL (for MRSA), while the resulting concentration of the bac-
terial suspension was approximately 105 CFU/ml. After incubation at
37 ◦C for 18–24 h, the MIC value was determined by visual inspection.
The MIC was the lowest concentration at which the bacterial growth was
completely inhibited. The assay was performed in duplicate.
4.3.2. S. aureus GyrB inhibition assay
The reaction mixture of the assay (10 μL) includes 5 nM S. aureus
Gyrase (Inspiralis Ltd., Norwich, United Kingdom), the buffer (40 mM
HEPES-KOH (pH 7.6), 10 mM magnesium acetate, 10 mM dithiothreitol,
50 g/L BSA, 500 mM potassium glutamate), the compound at the test
concentration, 1% DMSO, 10 nM linear pBR322 DNA, 100 mM ATP.
Firstly, the solutions of the compound at 10 times the test concentration
were prepared by using the assay buffer and DMSO as solvents. Then, the
As the first step of molecular docking using OpenEye (OpenEye
Scientific Software, Inc., Santa Fe, NM, USA), a maximum of 200 con-
formers was generated by the module named OMEGA (version 2.5.1.4)
[25]. Secondly, the prepared protein structure was converted to a re-
ceptor by OEDocking (version 3.0.1) [26], with the cognate ligand to
define the binding site. Thirdly, all the conformers of the compound
were positioned in the binding site of the receptor and scored by the
Chemgauss4 scoring function in OEDocking. Lastly, the top-scoring pose
was retained and used as the initial binding mode between the com-
pound and GyrB.
compound solution (1
sequential adding of the buffer (7
and S. aureus Gyrase (0.5 L) as well as the ATP (1
mixture was incubated at 37 ◦C for 30 min.
To quantify the generated ADP, ADP-Glo kits were used. The ADP-
Glo reagent (40
L) was added to the mixture and incubated at 37 ◦C
for 40 min. Following that, the detection reagent (50 L) was added and
μ
L) was put into a PCR tube, followed by the
L), the linear pBR322 DNA (0.5 L)
L). The reaction
μ
μ
μ
μ
μ
μ
the mixture was incubated at another 5 min. The luminescence of the
mixture was recorded by the BioTek Synergy 2 microplate reader. The
enzymatic activity (%) after compound treatment at the test concen-
tration was calculated based on the luminescence of the mixtures with
the compound and without the compound. According to the activity
values (%) of the enzyme after the treatment with different concentra-
4.2.5. Molecular dynamics simulation
The all-atom MD simulation was performed for every GyrB-ligand
complex with AMBER 20 software [27] on GPUs. The initial structure
of the complex was generated by the above-mentioned molecular
docking. Each system was composed of a GyrB protein, a ligand, around
10,700 TIP3P water molecules, and 8 sodium ions that neutralize the
whole system. The partial atomic charges of the ligand were the AM1-
BCC charges calculated by the antechamber module of the AMBER soft-
ware package, while the other force field parameters of the ligand were
from GAFF2. The AMBER FF14SB force field was applied to the GyrB
protein. Antechamber and LEaP were used to generate the topology files
of ligands and GyrB, respectively.
tions of the compound (i.e. 0.01 μM ꢀ 100 μM), the IC50 value was
determined by using GraphPad Prism 5 software (GraphPad Software
Inc., La Jolla, CA). The compounds were tested in duplicate. Novobiocin
was the positive control of this assay.
4.3.3. Cytotoxicity assay
The SRB assay was used to determine the cytotoxicity. The cells
(HepG2 or HUVEC) seeded in 96-well plates were treated with the 3-fold
For each protein–ligand complex, the solvent of the system was
minimized by both the steepest descent method (5000 steps) and the
conjugated gradient algorithm (20,000 steps). All the solute were
restrained using a harmonic potential, with a force constant of 500
dilutions of the compounds (i.e. 0.01 μM-100 μM), and incubated at the
conditions of 37 ◦C and 5% CO2 for 72 h. Then, the cells were fixed with
10% trichloroacetic acid (w/v), kept at 4 ◦C for 1 h, and washed with
10

X. Lian et al.
Bioorganic Chemistry 114 (2021) 105042
Modeling and Experimental Validation of the Imidazolium and Pyridinium Based
Ionic Liquids as Potential Antibacterials of MDR Acinetobacter Baumannii and
Staphylococcus Aureus, Int. J. Mol. Sci. 22 (2) (2021).
distilled water for 5 times and air-dried. Next, the cells were stained with
0.4% (w/v) sulforhodamine B (SRB) at the room temperature for 20 min
and washed with 1% acetic acid for 5 times. Lastly, the bound SRB was
solubilized with 10 mM Tris and the absorbance in term of optical
density was measured at 540 nm by using a Tecan Infinite M1000
microplate reader. According to the cell viability (%) after the treatment
with different concentrations of the compounds, the CC50 value was
determined by using GraphPad Prism 5 software (GraphPad Software
Inc., La Jolla, CA). Paclitaxel was used as the positive drug for this assay.
The assay was performed in duplicate.
[9] S. Hinsberger, K. Husecken, M. Groh, M. Negri, J. Haupenthal, R.W. Hartmann,
Discovery of novel bacterial RNA polymerase inhibitors: pharmacophore-based
virtual screening and hit optimization, J. Med. Chem. 56 (21) (2013) 8332–8338.
[10] E. Karhu, J. Isojarvi, P. Vuorela, L. Hanski, A. Fallarero, Identification of Privileged
Antichlamydial Natural Products by a Ligand-Based Strategy, J. Nat. Prod. 80 (10)
(2017) 2602–2608.
[11] K. Mitra, A. Chadha, M. Doble, Pharmacophore based approach to screen and
evaluate novel Mycobacterium cell division inhibitors targeting FtsZ - A modelling
and experimental study, Eur. J. Pharm. Sci. 135 (2019) 103–112.
[12] M.A. El-Atawy, E.A. Hamed, M. Alhadi, A.Z. Omar, Synthesis and Antimicrobial
Activity of Some New Substituted Quinoxalines, Molecules 24 (22) (2019).
[13] M.A. Wikler, Methods for dilution antimicrobial susceptibility tests for bacteria
that grow aerobically : approved standard, CLSI (NCCLS) 26 (2006) M7–A7.
[14] S.A.M. Khan, P. Manchanda, Synthesis and antimicrobial activity of 2,3-di-
substituted quinoxalines, Indian J. Heterocycl. Chem. 18 (2) (2008) 197–198.
[15] A. Keivanloo, M. Bakherad, A. Rahimi, Synthesis of Unexpected Pyrrolo[2,3-b]
quinoxaline-2-carbaldehydes via Sonogashira Coupling Reaction, Synthesis 2010
(10) (2010) 1599–1602.
Declaration of Competing Interest
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.
[16] S.V. Litvinenko, V.I. Savich, L.D. Bobrovnik, Synthesis, structure, and chemical
properties of some N-(3-chloro-2-quinoxalyl)arylsulfonamides, Chem. Heterocycl.
Compd. 30 (3) (1994) 340–344.
Acknowledgments
[17] K. Ingram-Sieber, N. Cowan, G. Panic, M. Vargas, N.R. Mansour, Q.D. Bickle, T.
N. Wells, T. Spangenberg, J. Keiser, Orally active antischistosomal early leads
identified from the open access malaria box, PLoS Negl Trop Dis 8 (1) (2014),
e2610.
This work was supported by the CAMS Innovation Fund for Medical
Sciences (Grant No. 2017-I2M-1-012) and PUMC Graduate Students
Innovation Fund (Grant No. 2019-1007-01) and by the China Scholar-
ship Council (CSC) for Z.X. (Grant No. 201706880010). We are grateful
to Sichuan Primed Bio-Tech Group Co., Ltd for the bioassays against
MRSA. We also thank Dr. Wenxuan Zhang and Dr. Tianlei Li for helpful
discussions on the chemical synthesis.
[18] B.T.J.P. Poyser, D. Timms, M.H. Block, N.J. Hales, Triazine derivatives and their
use as antibacterials, WO Pat (1999), 99/01442.
[19] Z. Jakopin, J. Ilas, M. Barancokova, M. Brvar, P. Tammela, M. Sollner Dolenc,
T. Tomasic, D. Kikelj, Discovery of substituted oxadiazoles as a novel scaffold for
DNA gyrase inhibitors, Eur. J. Med. Chem. 130 (2017) 171–184.
[20] N.S. Haiba, H.H. Khalil, M.A. Moniem, M.H. El-Wakil, A.A. Bekhit, S.N. Khattab,
Design, synthesis and molecular modeling studies of new series of s-triazine
derivatives as antimicrobial agents against multi-drug resistant clinical isolates,
Bioorg. Chem. 89 (2019), 103013.
Appendix A. Supplementary material
[21] D.R. Roe, T.E. Cheatham, PTRAJ and CPPTRAJ: Software for Processing and
Analysis of Molecular Dynamics Trajectory Data, J. Chem. Theory Comput. 9 (7)
(2013) 3084–3095.
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.bioorg.2021.105042.
[22] H. Quan, H. Liu, C. Li, L. Lou, 1,4-Diamino-2,3-dicyano-1,4-bis(methylthio)
butadiene (U0126) Enhances the Cytotoxicity of Combretastatin A4 Independently
of Mitogen-Activated Protein Kinase Kinase, J. Pharmacol. Exp. Ther. 330 (1)
(2009) 326.
References
[23] G.S. Bisacchi, J.I. Manchester, A New-Class Antibacterial-Almost Lessons in Drug
Discovery and Development: A Critical Analysis of More than 50 Years of Effort
toward ATPase Inhibitors of DNA Gyrase and Topoisomerase IV, ACS Infect. Dis. 1
(1) (2015) 4–41.
[1] J.M. Stokes, K. Yang, K. Swanson, W. Jin, A. Cubillos-Ruiz, N.M. Donghia, C.
R. MacNair, S. French, L.A. Carfrae, Z. Bloom-Ackermann, V.M. Tran,
A. Chiappino-Pepe, A.H. Badran, I.W. Andrews, E.J. Chory, G.M. Church, E.
D. Brown, T.S. Jaakkola, R. Barzilay, J.J. Collins, A Deep Learning Approach to
Antibiotic Discovery, Cell 180 (4) (2020) 688–702 e13.
[24] J. Araújo, F.G. Menezes, H.F.O. Silva, D.S. Vieira, S.R.B. Silva, A.J. Bortoluzzi,
C. Sant’Anna, M. Eugenio, J.M. Neri, L.H.S. Gasparotto, Functionalization of gold
nanoparticles with two aminoalcohol-based quinoxaline derivatives for targeting
[2] L.B. Rice, Federal funding for the study of antimicrobial resistance in nosocomial
pathogens: no ESKAPE, J. Infect. Dis. 197 (8) (2008) 1079–1081.
[3] 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, W.H.O.P.P.L.W. Group, Discovery, research, and
development of new antibiotics: the WHO priority list of antibiotic-resistant
bacteria and tuberculosis, Lancet Infect. Dis. 18 (3) (2018) 318–327.
[4] M. Bassetti, A. Russo, A. Carnelutti, M. Wilcox, Emerging drugs for treating
methicillin-resistant Staphylococcus aureus, Expert Opin. Emerging Drugs 24 (3)
(2019) 191–204.
phosphoinositide 3-kinases (PI3Kα), New J. Chem. 43 (4) (2019) 1803–1811.
[25] P.C. Hawkins, A.G. Skillman, G.L. Warren, B.A. Ellingson, M.T. Stahl, Conformer
generation with OMEGA: algorithm and validation using high quality structures
from the Protein Databank and Cambridge Structural Database, J. Chem. Inf.
Model. 50 (4) (2010) 572–584.
[26] M. McGann, FRED pose prediction and virtual screening accuracy, J. Chem. Inf.
Model. 51 (3) (2011) 578–596.
[27] K.B.D.A. Case, I.Y. Ben-Shalom, S.R. Brozell, D.S. Cerutti, T.E. Cheatham III, V.W.
D. Cruzeiro, T.A. Darden, R.E. Duke, G. Giambasu, M.K. Gilson, H. Gohlke, A.
W. Goetz, R. Harris, S. Izadi, S.A. Izmailov, K. Kasavajhala, A. Kovalenko,
R. Krasny, T. Kurtzman, T.S. Lee, S. LeGrand, P. Li, C. Lin, J. Liu, T. Luchko, R. Luo,
V. Man, K.M. Merz, Y. Miao, O. Mikhailovskii, G. Monard, H. Nguyen, A. Onufriev,
F. Pan, S. Pantano, R. Qi, D.R. Roe, A. Roitberg, C. Sagui, S. Schott-Verdugo,
J. Shen, C.L. Simmerling, N.R. Skrynnikov, J. Smith, J. Swails, R.C. Walker,
J. Wang, L. Wilson, R.M. Wolf, X. Wu, Y. Xiong, Y. Xue, D.M. York, P.A. Kollman,
AMBER 2020, University of California, San Francisco, 2020.
[5] P. Fernandes, E. Martens, Antibiotics in late clinical development, Biochem.
Pharmacol. 133 (2017) 152–163.
[6] W.P. Walters, R. Wang, New Trends in Virtual Screening, J. Chem. Inf. Model. 59
(9) (2019) 3603–3604.
[7] R.C. Diniz, L.W. Soares, L.C. Nascimento da Silva, Virtual Screening for the
Development of New Effective Compounds Against Staphylococcus aureus, Curr.
Med. Chem. 25 (42) (2018) 5975–5985.
[8] I.V. Semenyuta, M.M. Trush, V.V. Kovalishyn, S.P. Rogalsky, D.M. Hodyna,
P. Karpov, Z. Xia, I.V. Tetko, L.O. Metelytsia, Structure-Activity Relationship
11