In silico study of febuxostat analogs as inhibitors of xanthine oxidoreductase: A combined 3D-QSAR and molecular docking study

Article abstract of DOI:10.1016/j.molstruc.2019.01.017

We explored molecular docking and a three-dimensional quantitative structure-activity relationship (3D-QSAR) model of 107 xanthine oxidoreductase (XOR) inhibitors containing a phenyl-substituted five-membered heterocycle. Molecular-docking results showed that Arg880, Phe914, and Phe1009 might be potential active residues targeted by the 107 XOR inhibitors evaluated in this study. Topomer comparative molecular field analysis (CoMFA) (q² = 0.571; r² = 0.833) was used for 3D-QSAR. The results indicated that benzene substituted with moderately bulky substituents, a cyano group, and a five-membered heterocycle with a carboxyl group might enhance XOR inhibitory activity. Four new compounds were designed based on these results, and each exhibited potential XOR inhibitory activity in vitro.

Full text of DOI:10.1016/j.molstruc.2019.01.017

Journal of Molecular Structure 1181 (2019) 428e435
Contents lists available at ScienceDirect
Journal of Molecular Structure
journal homepage: http://www.elsevier.com/locate/molstruc
In silico study of febuxostat analogs as inhibitors of xanthine
oxidoreductase: A combined 3D-QSAR and molecular docking study
Xiaolei Li 1, Haiyan Zhou 1, Xianwei Mo, Lei Zhang, Jing Li*
School of Bioscience and Bioengineering, South China University of Technology, Guangzhou, 510006, PR China
a r t i c l e i n f o
a b s t r a c t
Article history:
We explored molecular docking and a three-dimensional quantitative structure-activity relationship
(3D-QSAR) model of 107 xanthine oxidoreductase (XOR) inhibitors containing a phenyl-substituted five-
membered heterocycle. Molecular-docking results showed that Arg880, Phe914, and Phe1009 might be
potential active residues targeted by the 107 XOR inhibitors evaluated in this study. Topomer compar-
ative molecular field analysis (CoMFA) (q2 ¼ 0.571; r2 ¼ 0.833) was used for 3D-QSAR. The results indi-
cated that benzene substituted with moderately bulky substituents, a cyano group, and a five-membered
heterocycle with a carboxyl group might enhance XOR inhibitory activity. Four new compounds were
designed based on these results, and each exhibited potential XOR inhibitory activity in vitro.
© 2019 Published by Elsevier B.V.
Received 25 June 2018
Received in revised form
3 January 2019
Accepted 4 January 2019
Available online 7 January 2019
Keywords:
Xanthine oxidoreductase inhibitors
Molecular docking
3D-QSAR
Topomer CoMFA
1. Introduction
XOR inhibitors, it is necessary to evaluate febuxostat analogs with
reported half-maximal inhibitory concentration (IC50) values to
Xanthine oxidoreductase (XOR) is a target for treatment of gout
and conditions associated with hyperuricemia [1e3]. In 1966,
allopurinol (Fig. 1), a structural isomer of hypoxanthine, was the
first XOR inhibitor marketed for treatment of gout and other
hyperuricemia-associated conditions. However, allopurinol and its
analogs have severe life-threatening side-effects, collectively called
“allopurinol intolerance syndrome” [4,5]. Febuxostat is a thiazole-
carboxylic acid derivative that inhibits XOR activity, approved by
the US Food and Drug Administration (FDA) for treatment of gout in
2009 [6]. Owing to its excellent potency and better toxicological
profile, many febuxostat analogs have been designed, with a pri-
mary focus on replacement of the thiazole group with other five-
membered heterocycles, such as triazole [7], imidazole [8], sele-
nazoles [9], pyrazoles (e.g., Y-700) [10], and isoxazoles [11] (Fig. 1).
To better understand the mechanisms of XOR inhibition by
febuxostat and febuxostat analogs, and to guide the design of novel
determine generic structure-activity relationships (SARs) for XOR
inhibitors with a phenyl-substituted five-membered heterocycle
[7e11]. In the current study, a generic 3D-QSAR model of a series of
compounds was established by Topomer CoMFA [12e14]. In par-
allel, a molecular docking study was performed to gain insight into
binding interactions between XOR and febuxostat-related XOR
inhibitors.
2. Materials and methods
2.1. Computer simulations
The research procedures used in this study are shown in Fig. 2.
Data preparation, ligand-based study (3D-QSAR model), and
receptor-based study (molecular docking) were the three main
components of this study.
2.1.1. Data set
A series of febuxostat analogs that act as XOR inhibitors, and
their biologic activities, were obtained through a literature search
and used as a dataset for molecular modeling [7e11,15e18]. Among
Abbreviations: XOR, xanthine oxidoreductase; 3D-QSAR, three-dimensional
quantitative structure-activity relationship; CoMFA, comparative molecular field
analysis; IC50, half-maximal inhibitory concentration; SARs, structure-activity re-
lationships; PDB, Protein Data Bank.
these 107 inhibitors, the range of IC50 values was 0.0014e26.13
suggesting that the diversity in the dataset was sufficient to
construct stable 3D-QSAR models. The pIC50 [pIC50 ¼ ꢀlog (IC50
106)] values for these inhibitors ranged from 4.58 to 8.85. Based on
mM,
* Corresponding author.
E-mail address: lij@scut.edu.cn (J. Li).
These authors contributed equally to this work and should be considered as co-
1
/
first authors.
https://doi.org/10.1016/j.molstruc.2019.01.017
0022-2860/© 2019 Published by Elsevier B.V.

X. Li et al. / Journal of Molecular Structure 1181 (2019) 428e435
429
Fig. 1. Chemical structures of selected XOR inhibitors.
Fig. 2. Flowchart for computational drug design.
pIC50 values, the inhibitors were divided into two sets at an
approximate ratio of 4:1, with 86 compounds assigned to the
training set and 21 compounds assigned to the test set, (Table S1)
[19].
for those specifically mentioned. Using Tripos force field, energy
minimizations were performed using the Powell method when the
numerical value met the energy convergence criterion of
0.005 kcal/mol A. The independent variables were Topomer CoMFA
descriptors and the dependent variables were pIC50 values [20].
2.1.2. Topomer CoMFA modeling
All 2D structures were sketched in ChemDraw Ultra v8.0.3
(PerkinElmer, Waltham, MA, USA). They were converted into 3D
structures using SYBYL-X 2.0 (Tripos Software, Saint Louis, MO,
USA) running on Windows™ 7.0 64-bit operating system. All pa-
rameters of Topomer CoMFA methods were default values except
2.1.3. Model validation
To assess the predictability and reliability of the 3D-QSAR
models, both internal validation and external validation were per-
formed. For internal validation, the partial least squares [21e24]
algorithm was used to determine the optimal number of

430
X. Li et al. / Journal of Molecular Structure 1181 (2019) 428e435
components by leave-one-out cross-validation [25], and defined
the highest cross-validation correlation coefficient (q2) to be:
hydrolysis, target compounds a-d were obtained from 3a-d.
2.2.2. Experimental protocols
ꢂ
ꢀ
ꢁ
X
X
2
ꢃ
ꢄ
2.2.2.1. General information. 1H NMR and 13C NMR spectra were
recorded in CDCl3 or DMSO-d6 on Bruker Avance III HD 400 MHz
spectrometers (Bruker, Germany) using TMS as the internal stan-
2
q2 ¼ 1 ꢀ
Ypred ꢀ Yexp
Yexp ꢀ Ymean
where Ypred, Yexp, and Ymean represented the predicted, experi-
mental, and mean biologic activities of training-set compounds,
respectively. For external validation, the predicted correlation co-
dard, respectively. Chemical shifts (d) were reported in ppm with
respect to TMS. Infrared spectra were recorded on a Bruker Vertex
33 instrument (Bruker, Germany). HRMS spectra were recorded on
an Agilent 1290 instrument in ESI mode (Agilent Technologies,
USA). All the reaction progress were checked on by thin layer
chromatography (TLC) using silica gel glass cards (Qingdao Ocean
Chemical, Qingdao, Shandong, P. R. China) with UV light indicator at
254 nm. Silica gel column chromatography was performed using
200e300 mesh of silica gel with the indicated solvents as the
eluent. Unless otherwise indicated, commercially available re-
agents were used without further purification.
efficient r2
[26e28] reflected the ability to predict 3D-QSAR
pred
models and defined r2pred as:
r2pred ¼ ðSD ꢀ PRESSÞ=SD
where SD expressed the sum of the squared deviation between the
molecular biological activity of the test set and the molecular mean
biological activity of the training set, and PRESS was the sum of
squares between the predictions and the experimental biological
activity of the test set molecules [29].
2.2.2.2. General procedures for the synthesis of compounds a-d
2.2.2.2.1. The preparation of 5-bromo-2-mercaptobenzonitrile (2).
A mixture of compound 1 (3.0 g, 15 mmol), Na2S (1.3 g, 16.5 mmol),
and DMF (10 mL) was stirred at room temperature for 5.0 h. The
resulting mixture was treated with NaOH (1 M, 300 mL) and stirred
for 0.5 h. Then the mixture was added EtOAc (200 mL). The organic
layer was separated, and the aqueous phase was extracted with
EtOAc (100 mL ꢁ 2). After that, the water phase was acidified with
6 M HCl solution to pH ¼ 2 and extracted with EtOAc (100 mL ꢁ 2).
The combined organic layer was washed with brine (300 mL ꢁ 2),
dried over anhydrous MgSO4, filtered and evaporated. The residue
was purified by flash column chromatography (0e10% EtOAc/pe-
troleum ether) to afford 2 as yellow solid (31.2%); 1H NMR (CDCl3)
2.1.4. Molecular docking
To further study binding interactions between the crucial
groups of febuxostat analogs and XOR, molecular docking was
performed using SYBYL-X 2.0 (Surflex-Dock method) [30]. The
protein structures of XOR complexed with Y-700 were taken from
the Research Collaboration for Structural Bioinformatics Protein
Data Bank (PDB ID: 1VDV) (www.rscb.org/pdb) and used in docking
experiments [31]. Before docking, the “Prepare Protein Structure”
procedure was used to prepare the inputted XOR structures for
docking, which comprised extracting the ligand Y-700, removing
all the watery molecules, and adding polar hydrogen atoms to the
XOR receptor [32]. Meanwhile, the “Ligand Structure Preparation”
procedure was used to prepare the inputted ligands for docking,
which comprised removing duplicates, adding polar hydrogen
atoms, and generating 3D conformations. Docking simulations
were carried out using a standard Surflex-Dock protocol with
default values for adjustable parameters [33].
d
7.78 (d, J ¼ 2.0 Hz, 1H, AreH), 7.71 (dd, J ¼ 2.0, 8.6 Hz, 1H, AreH),
7.63 (d, J ¼ 8.6 Hz, 1H, AreH).
2.2.2.2.2. General procedure for the preparation of compounds
3a-d. A mixture of K2CO3 (1.0 g, 7.0 mmol), 5-bromo-2-
mercaptobenzonitrile (2) (0.5 g, 2.3 mmol) in DMF (10 mL) was
added in a 100 mL reaction vial, which was sealed and stirred at
room temperature for 1.0 h. Then the reaction mixture was allowed
to add alkyl halide (7.0 mmol) and heated at 60 ꢂC for 6 h. The
mixture was cooled to room temperature and diluted with H2O
(30 mL), the resulting mixture was added EtOAc (30 mL). The
organic layer was separated, and the aqueous phase was extracted
with EtOAc (20 mL ꢁ 3). The combined organic layers were washed
with brine (100 mL ꢁ 2), dried over anhydrous Na2SO4 and evapo-
rated in a vacuum. Afterward, the residue was purified by flash
column chromatography (5% EtOAc/petroleum ether) to afford the
desired product 3a-d.
2.2. XOR inhibitors design and activity evaluation
To further verify the reliability of molecular docking and 3D-
QSAR models, new XOR inhibitors were designed, synthesized, XOR
inhibitory activities evaluated in vitro.
2.2.1. Chemistry
Based on our previous studies [15], synthesis of compounds a-
d is shown in Scheme 1. Commercially available 5-bromo-2-
fluorobenzonitrile (1) was treated with Na2S to provide 2. Com-
pound 2 was alkylated with C3eC6 alkyl halides in the presence of
K2CO3 to give 3a-d. Through a CeN coupling reaction and
5-bromo-2-(isopropylthio)-benzonitrile (3a). Colourless oil
(yield: 35.6%); 1H NMR (CDCl3)
d
7.76 (d, 1H, J ¼ 2.2 Hz, AreH), 7.62
(dd, 1H, J ¼ 2.2, 8.5 Hz, AreH), 7.37 (d, 1H, J ¼ 8.5 Hz, AreH),
Scheme 1. Reagents and conditions: (i) Na2S, DMF, NaOH, CH2Cl2; (ii) K2CO3, DMF, alkyl halide, 50 ꢂC; (iii) K2CO3, CuI, ethyl 1H-pyrazole-4-carboxylate, (E)eN, N0-dimethyl-1, 2-
cyclohexane-diamine, DMF, 110 ꢂC; (iv) NaOH aq. (1 M), THF/ethanol (1:1), 60 ꢂC, HCl aq. (1 M), rt.

X. Li et al. / Journal of Molecular Structure 1181 (2019) 428e435
431
3.60e3.46 (m, 1H, CH(CH3)2), 1.34 (d, 6H, J ¼ 6.7 Hz, CH(CH3)2).
5-bromo-2-(isobutylthio)-benzonitrile (3b). Colourless oil
d was completed. Then the reaction mixture was allowed to cool to
room temperature and acidified with 1 M HCl to pH ¼ 1e2 in ice
water. The reaction mixture was diluted with H2O, the suspension
was stirred in ice water for 20 min, obtained the crude product by
filtration. The crude product was purified by washing with water
and recrystallizing in methanol to afford the desired product a-d.
1-(3-cyano-4-(isopropylthio) phenyl)-1H-pyrazole-4-carboxylic
(yield: 54.7%); 1H NMR (CDCl3)
d
7.65 (d, 1H, J ¼ 2.2 Hz, AreH), 7.54
(dd,1H, J ¼ 2.2, 8.6 Hz, AreH), 7.20 (d,1H, J ¼ 8.6 Hz, AreH), 2.82 (m,
2H, J ¼ 5.3 Hz, CH2CH(CH3)2), 1.93e1.75 (m, 1H, CH2CH(CH3)2),
1.10e0.92 (d, 6H, J ¼ 6.7 Hz, CH2CH(CH3)2).
5-bromo-2-(sec-butylthio)-benzonitrile (3c). Colourless oil
(yield: 53.9%); 1H NMR (CDCl3)
d
7.74 (d, 1H, J ¼ 2.2 Hz, AreH), 7.61
acid (a). White solid (yield:100%); 1H NMR (DMSO-d6)
d 8.87 (s, 1H,
(dd, 1H, J ¼ 2.2, 8.5 Hz, AreH), 7.35 (d, 1H, J ¼ 8.5 Hz, AreH),
3.40e3.27 (m, 1H, CHCH3CH2CH3), 1.74e1.52 (m, 2H,
CHCH3CH2CH3), 1.31 (d, 3H, J ¼ 6.7 Hz, CHCH3CH2CH3),1.06e0.98 (t,
3H, J ¼ 6.4 Hz, CHCH3CH2CH3).
CH), 8.16 (s, 1H, CH), 7.96 (d, 1H, J ¼ 8.7 Hz, AreH), 7.88 (s, 1H,
AreH), 7.54 (d, 1H, J ¼ 8.7 Hz, AreH), 3.47 (m, 1H, CH(CH3)2), 1.07 (d,
J ¼ 6.5 Hz, 6H, CH(CH3)2); 13C NMR (DMSO-d6)
d 163.8, 143.0, 138.7,
137.8, 132.8, 132.1, 124.3, 124.1, 118.2, 116.9, 114.8, 38.4, 23.0; HRMS:
Calcd for C14H13N3O2S [MþNa]þ 310.0629, Found 310.0633.
1-(3-cyano-4-(isobutylthio) phenyl)-1H-pyrazole-4-carboxylic
5-bromo-2-((2-ethylbutyl)thio)-benzonitrile (3d). Colourless oil
(yield: 48.3%); 1H NMR (CDCl3)
d 7.71 (s, 1H, AreH), 7.60 (d,
J ¼ 8.5 Hz, 1H, AreH), 7.26 (d, J ¼ 6.7 Hz, 1H, AreH), 2.97 (d,
J ¼ 5.7 Hz, 1H, CH2CH(CH2CH3)2), 1.60 (s, 2H, CH2CH(CH2CH3)2),
1.57e1.43 (m, 4H, CH2CH(CH2CH3)2), 0.90 (t, J ¼ 7.2 Hz, 6H,
CH2CH(CH2CH3)2).
acid (b). White solid (yield: 82.9%); 1H NMR (DMSO-d6)
d 12.74
(s, 1H, COOH), 9.13 (s, 1H, CH), 8.40 (d, 1H, J ¼ 2.5 Hz, CH), 8.22e8.18
(dd, 1H, J ¼ 2.0, 8.2 Hz, AreH), 8.12 (s, 1H, AreH), 7.73 (d, 1H,
J ¼ 8.9 Hz, AreH), 3.05 (d, 2H, J ¼ 6.8 Hz, CH2CH(CH3)2), 1.92e1.76
(m, 1H, CH2CH(CH3)2), 1.02 (d, 6H, J ¼ 6.6 Hz, CH2CH(CH3)2); 13C
2.2.2.2.3. General procedure for the preparation of compounds
4a-d. Under the nitrogen atmosphere, a mixture of compounds 3a-
d (0.24 g, 1.0 mmol), 1H-pyrazole-3-carboxylate (0.11 g, 0.8 mmol),
NMR (DMSO-d6) d 163.8,142.9,140.7,137.1,132.0,130.0,124.4,118.1,
116.8, 112.6, 41.4, 28.1, 22.0; HRMS: Calcd for C15H15N3O2S [MþNa]þ
K2CO3
(0.23 g,
1.7 mmol),
(E)eN,
N0-Dimethyl-1,
2-
324.0777, Found 324.0779.
cyclohexanediamine (82 mg, 0.96 mmol), CuI (15 mg, 0.08 mmol)
in DMF (5 mL) were added in 25 mL reaction vial, which was heated
to 110 ꢂC and stirred for 24 h. Then the reaction mixture was cooled
to room temperature and diluted with H2O (20 mL), the resulting
mixture was added EtOAc (20 mL). The organic layer was separated,
and the aqueous phase was extracted with EtOAc (15 mL ꢁ 3). The
combined organic extracts were washed with brine (50 mL), dried
over anhydrous Na2SO4. After that, the organic phase was
concentrated in a vacuum and was purified by flash column chro-
matography (6e12% EtOAc/petroleum ether) to afford the desired
products 4a-d.
1-(4-(sec-butylthio)-3-cyanophenyl)-1H-pyrazole-4-carboxylic
acid (c). White solid (yield: 88.5%); 1H NMR (DMSO-d6)
d 11.85 (s,
1H, COOH), 8.28 (s, 1H, CH), 7.55 (s, 1H, CH), 7.34 (dd, J ¼ 2.5, 8.8 Hz,
1H, AreH), 7.26 (s, 1H, AreH), 6.92 (d, J ¼ 8.8 Hz, 1H, AreH), 2.69 (q,
J ¼ 6.5 Hz, 1H, CHCH3CH2CH3), 0.75 (m, 2H, CHCH3CH2CH3), 0.42 (d,
J ¼ 6.7 Hz, 3H, CHCH3CH2CH3), 0.12 (t, J ¼ 7.4 Hz, 3H,
CHCH3CH2CH3); 13C NMR (DMSO-d6):
d 163.8, 143.0, 138.7, 137.8,
132.9, 132.1, 124.1, 118.2, 116.9, 114.9, 44.9, 29.3, 20.5, 11.5; HRMS:
Calcd for C15H15N3O2S [MþNa]þ 324.0777, Found 324.0786.
1-(3-cyano-4-((2-ethylbutyl) thio) phenyl)-1H-pyrazole-4-
carboxylic acid (d). White solid (yield: 87.7%); 1H NMR (DMSO-
Ethyl
1-(3-cyano-4-(isopropylthio)phenyl)-1H-pyrazole-4-
d6) d 12.75 (s, 1H, COOH), 9.14 (s, 1H,CH), 8.41 (s, 1H, CH), 8.20 (d,
carboxylate (4a). White solid (yield: 98.0%); 1H NMR (CDCl3)
d
J ¼ 8.8 Hz, 1H, AreH), 8.13 (s, 1H, AreH), 7.74 (d, J ¼ 8.8 Hz, 1H,
AreH), 3.12 (d, J ¼ 5.9 Hz, 2H, CH2CH(CH2CH3)2), 1.52 (dt, J ¼ 6.1,
11.9 Hz,1H, CH2CH(CH2CH3)2),1.48e1.38 (m, 4H, CH2CH(CH2CH3)2),
0.87 (t, J ¼ 7.2 Hz, 6H, CH2CH(CH2CH3)2; 13C NMR (DMSO-d6)
8.44 (s, 1H, CH), 8.13 (s, 1H, CH), 8.05 (d, 1H, J ¼ 2.5 Hz, AreH), 7.89
(dd, 1H, J ¼ 2.5, 8.6 Hz, AreH), 7.64 (d, 1H, J ¼ 8.7 Hz, AreH), 4.37 (q,
2H, J ¼ 7.1 Hz, CH2CH3), 3.62 (m, 1H, CH(CH3)2), 1.44e1.40 (overlap,
3H, CH2CH3), 1.41e1.35 (d, 6H, J ¼ 1.4 Hz, CH(CH3)2).
d
163.81, 142.92, 140.70, 137.11, 132.01, 130.06, 124.31, 123.96, 118.14,
Ethyl
1-(3-cyano-4-(isobutylthio)phenyl)-1H-pyrazole-4-
116.76, 112.67; HRMS: Calcd for C17H19N3O2S [MþH]þ 330.1276,
carboxylate (4b). White solid (yield: 58.7%); 1H NMR (CDCl3)
Found 330.1257.
d
8.39 (s, 1H, CH), 8.10 (s, 1H, CH), 7.98 (d, 1H, J ¼ 2.4 Hz, AreH), 7.84
(dd, 1H, J ¼ 2.5, 8.7 Hz, AreH), 7.49 (d, 1H, J ¼ 8.7 Hz, AreH), 4.34 (q,
2H, J ¼ 7.1 Hz, CH2CH3), 2.93 (d, 2H, J ¼ 6.8 Hz, CH2CH(CH3)2), 1.92
(m, 1H, CH2CH(CH3)2), 1.38 (t, 3H, J ¼ 7.1 Hz, CH2CH3), 1.08 (d, 6H,
J ¼ 6.7 Hz, CH2CH(CH3)2).
2.2.3. Biological activity in vitro
Xanthine (0.5 mM) and 0.5 mL/100 mL XOR (Sigma, from bovine
milk) were prepared by diluting in PBS (PBS refers to 1ꢁ PBS unless
otherwise specified). The required concentrations of compounds a-
d were prepared in PBS and used for XOR activity assays. According
to the procedure reported by Fukunari et al. [31], XOR activity with
xanthine as substrate was measured spectrophotometrically as
Ethyl
1-(4-(sec-butylthio)-3-cyano-phenyl)-1H-pyrazole-4-
carboxylate (4c). White solid (yield: 60.7%); 1H NMR (CDCl3)
d
8.41 (s, 1H, CH), 8.11 (s, 1H, CH), 8.02 (d, J ¼ 2.4 Hz, 1H, AreH), 7.86
(dd, J ¼ 8.7, 2.5 Hz, 1H, AreH), 7.60 (d, 1H, J ¼ 8.7 Hz, AreH), 4.34 (q,
2H, J ¼ 7.1 Hz, CH2CH3), 3.39 (m, 1H, CHCH3CH2CH3), 1.79e1.56 (m,
2H, CHCH3CH2CH3), 1.38 (t, 3H, J ¼ 7.1 Hz, CH2CH3), 1.34 (d, 3H,
J ¼ 4.7 Hz, CHCH3CH2CH3), 1.05 (t, 3H, J ¼ 7.4 Hz, CHCH3CH2CH3).
Ethyl 1-(3-cyano-4-((2-ethylbutyl) thio) phenyl)-1H-pyrazole-
4-carboxylate (4d). White solid (yield: 58.2%); 1H NMR (CDCl3)
follows. A blank solution (PBS), enzyme solution (100 mL), and test
compounds were added to a 96 well plate, then incubated for
3 min at 37 ꢂC. Following incubation, substrate solution was
immediately added to the plate. Samples were monitored at
295 nm every 30 s for 5 min using an Ensipre-2300 microplate
reader (Perkin Elmer, USA). IC50 values were calculated using Excel
2007 and Prism 6.0 statistical software (GraphPad Software, Inc.,
San Diego, CA).
d
8.36 (s, 1H, CH), 8.07 (s, 1H, CH), 7.95 (d, J ¼ 2.4 Hz, 1H, AreH), 7.81
(dd, J ¼ 8.7, 2.5 Hz, 1H, AreH), 7.47 (d, J ¼ 8.7 Hz, 1H, AreH), 4.31 (q,
J ¼ 7.1 Hz, 2H, CH2CH3), 2.99 (d, J ¼ 6.0 Hz, 2H, CH2CH(CH2CH3)2),
1.54 (dd, J ¼ 10.9, 4.8 Hz,1H, CH2CH(CH2CH3)2),1.46 (tdd, J ¼ 8.9, 7.1,
3.0 Hz, 4H, CH2CH(CH2CH3)2), 1.34 (t, J ¼ 7.1 Hz, 3H, CH2CH3), 0.88
(t, J ¼ 7.4 Hz, 6H, CH2CH(CH2CH3)2).
3. Results and discussion
3.1. 3D-QSAR and docking analyses
2.2.2.2.4. General procedure for the preparation of compounds a-
d. 4a-d (0.12 g, 0.38 mmol) in the mixed solvent of THF (5 mL)
ethanol (5 mL) was added in a 50 mL reaction vial, which was
sealed and heated at 60 ꢂC until TLC revealed that conversion of 4a-
3.1.1. Data processing
As shown in Fig. 3, 107 compounds with “high,” “moderate,” and
“low” activity were selected in approximately equal proportions in

432
X. Li et al. / Journal of Molecular Structure 1181 (2019) 428e435
Table 1
Statistical analyses of the two methods of splitting molecules.
Method
q2
r2
N
Three pieces
Four pieces
0.537
0.571
0.789
0.833
5
5
q2 ¼ leave-one-out cross-validation correlation coefficient.
r2 ¼ Non-cross-validation correlation coefficient.
N ¼ optional number of components.
Fig. 3. Distribution of inhibitory activities for the training set and test set in Topomer
CoFMA studies.
both the training set and test set by considering the distribution of
biologic data and structural diversity to ensure that the datasets
used for 3D-QSAR modeling were appropriate.
3.1.2. Topomer CoMFA
Fig. 5. Scatterplot of experimental versus predicted pIC50 values of training and test
In general, the splitting mode of the R-group is important for
model generation in Topomer CoMFA. As such, comparison of two
different splitting modes was performed (Fig. 4). Inhibitor structure
(Table S1) showed the substituent group at R2 to be eCN, sug-
gesting that the molecules could be split into three pieces: R1, R3,
and R2-substituted benzene. However, by comparing the values of
q2 and r2 (Table 1), it was clear that splitting the compounds into
four pieces was more appropriate to our analyses.
sets in a 3D-QSAR model.
3D coefficient contour maps (Fig. 6 and Fig. 7). Results of Topomer
CoMFA models were interpreted graphically using field contribu-
tion maps. Fig. 7 shows the calculated Topomer CoMFA steric and
electrostatic contour maps. Steric field analysis suggested that only
moderately bulky substituents would be favorable, as shown by the
green contours of the R1 group, while bulky substituents would not
contribute to inhibition, as shown by the yellow contours. These
characteristics may explain higher activity of compound 20
(pIC50 ¼ 8.52) with a butoxyl group as R1, and lower activity of
compounds 17 (pIC50 ¼ 6.26) and 26 (pIC50 ¼ 5.92) with -methoxyl
and -octyloxyl R1 groups, respectively (R2 and R3 were identical for
these compounds). In electrostatic field, the advantage of nega-
tively charged groups was indicated by red contours of R1 group,
with the advantage of positively charged groups indicated by blue
contours. This phenomenon occurred because all R1 substituents
contained hetero atoms such as oxygen, nitrogen, and sulfur.
As shown in Fig. 7c, red contours at the R2 group suggested that
negatively charged groups would be favorable. Among the 107
compounds, compound 51 was the only one in which R2 was a
3.1.3. Validation of 3D-QSAR models
The predicted pIC50 [pIC50 ¼ ꢀlog (IC50/106)] values, including
the training and test sets, their residual values, and contributions of
R1, R2, and R3 fragments, were calculated (Table S2). After obtaining
an r2pred value of 0.656 from the test set, the predictive ability of the
3D-QSAR model was evaluated, and the robustness of the model
was evaluated by external verification. The scatterplot in Fig. 5
showed that the predicted pIC50 values of those compounds were
coincident with experimental values. These results demonstrated
that the Topomer CoMFA model was reliable for predicting pIC50
values.
In addition, by comparing the fragment contribution values of
R1, R2, and R3, we hypothesized that substitution at R2 had the
greatest impact on XOR inhibitory activity, followed by R1 and R3.
3.1.4. 3D contour maps
To facilitate analyses, febuxostat was chosen as the reference in
Fig. 4. The method of splitting compounds into pieces: (A) four pieces and (B) three
pieces.
Fig. 6. Splitting of febuxostat.

X. Li et al. / Journal of Molecular Structure 1181 (2019) 428e435
433
Fig. 8. Structural super-positioning of the co-crystal structure (green) and re-docked
structure (red) for XOR.
between Y-700 and XOR had four main features (docking
score: ꢀ6.309): 1) the carboxyl group extended to the catalytic
center containing a molybdenum atom, and interacted with Arg880
(O—HeN: 2.33 Å, O—HeN: 1.97 Å) and Thr1010 (O—HeN: 2.22 Å)
via hydrogen bonding; 2) the pyrazole ring was sandwiched be-
tween Phe914 (3.77 Å, 13.3ꢂ) and Phe1009 (4.69 Å, 69.4ꢂ) by
p-p
stacking; 3) the hydrogen bond between the eCN group and
Asn768 (N—HeN: 2.44 Å); 4) R1 occupied the channel through
which the substrate enters the catalytic center of XOR, a hydro-
phobic pocket composed of phe 649, leu 648, lys 771, Val 1011, leu
1014, and phe 1013. Compound 61 (docking score: ꢀ7.343), another
ligand with high XOR inhibitory activity, exhibited a similar inter-
action with XOR to that of Y-700, including hydrogen bonds with
Arg880 (O—HeN: 2.19 Å, O—HeN: 1.74 Å), Thr1010 (O—HeN:
2.38 Å), and Asn768 (N—HeN: 2.64 Å) and
p-p stacking between
Phe914 (6.04 Å, 8.8ꢂ) and Phe1009 (4.83 Å, 70.0ꢂ).
Fig. 7. CoMFA StDev * Coeff contour maps for febuxostat. (a) and (b) Steric and elec-
trostatic field maps of the R1 fragment, respectively. (c) and (d) Steric and electrostatic
field maps of the R2 fragment, respectively. (e) and (f) Steric and electrostatic field
maps of the R3 fragment, respectively (green and yellow contours denote sterically
favorable and unfavorable sites, respectively; blue and red contours denote regions
that favor electropositive and electronegative groups, respectively).
3.2. XOR inhibitors design and activity evaluation
Based on the results from molecular docking and 3D-QSAR
models, four new XOR inhibitors a-d were designed by choosing
pyrazoles as the five-membered heterocycles. The experimental
IC50 values, experimental pIC50 values, and predicted pIC50 values
determined by Topomer CoMFA modeling are shown in Table 2.
These results showed that predicted pIC50 values were consistent
with experimental values. In addition, with increasing R1 volume,
changes in XOR inhibitory activity were in agreement with our
hypothesis that only moderately bulky substituents at R1 were
favorable for XOR inhibition. Experimental IC50 values, experi-
mental pIC50 values, and predicted pIC50 values determined by
Tomoper CoMFA modeling are summarized in Table 2.
hydrogen, and this compound exhibited poor XOR inhibitory ac-
tivity (pIC50 ¼ 5.33). Nearly all other compounds had a cyano group
or a nitro group at R2.
Yellow contours at the R3 group (between the benzene ring and
S atom on the R3 group) suggested that bulky substituents would be
unfavorable at this site. This could explain significant reduction in
XOR inhibitory activity for compound 21 (pIC50 ¼ 8.52) and com-
pound 28 (pIC50 ¼ 5.92). Similarly, only moderately bulky sub-
stituents between the N atom and carboxyl group allowed for
inhibitory activity. Finally, red contours at the R3 group demon-
strated that all R3 substituents contained a carboxyl group as an
essential group.
4. Conclusion
A combined computational and experimental approach was
employed in this study to identify the structural determinants and
specific binding modes between XOR and its inhibitors. Higher
statistical values (q2 ¼ 0.571; r2 ¼ 0.833) suggested strong reli-
ability and predictive capability of the Topomer CoMFA model.
Receptor-based molecular docking studies suggested that hydrogen
bonding, p-p stacking, and hydrophobic interactions were the most
important interactions between XOR and its inhibitors. Further-
more, four newly designed compounds showed potential XOR
3.1.5. Docking analyses
To validate docking accuracy, co-crystalized ligand (Y-700) was
first re-docked to the active binding site of XOR (PDB ID: 1VDV,
www.rscb.org/pdb). As shown in Fig. 8, the occupied space and
conformation of the re-docked moieties were closely associated
with those of the co-crystalized moieties, suggesting acceptable
reliability of the docking procedure. As shown in Fig. 9, binding

434
X. Li et al. / Journal of Molecular Structure 1181 (2019) 428e435
Fig. 9. Interactions between compounds and XOR protein residues. (a) Y-700 forms H-bonds with Arg880, Thr1010, and Asn768. (b) Compound 61 forms H-bonds with Arg880,
Thr1010, and Asn768. (The docking graphics were created by MOE and polished by photoshop CS6).
potent inhibitor of xanthine oxidoreductase crystal structure of the enzyme-
inhibitor complex and mechanism of inhibition, J. Biol. Chem. 278 (2003)
1848e1855.
Table 2
Values for IC50 and Predicted pIC50 on XOR for compounds a-d.
Compd.
R1
IC50 (nM)a
Exp. pIC50
Pred. pIC50
[5] G. Biagi, I. Giorgi, F. Pacchini, O. Livi, V. Scartoni, Cheminform abstract: 2-
alkyloxyalkylthiohypoxanthines as new potent inhibitors of xanthine oxi-
dase, IL Farmaco 56 (2001) 809e813.
[6] J.U. Song, S.P. Choi, T.H. Kim, C.K. Jung, J.Y. Lee, S.H. Jung, G.T. Kim, Design and
synthesis of novel 2-(indol-5-yl)thiazole derivatives as xanthine oxidase in-
hibitors, Bioorg. Med. Chem. Lett 25 (2015) 1254e1258.
a
b
c
eSCH(CH3)2
10.4
9.8
8.8
7.98
8.01
8.06
8.02
7.30
7.98
7.60
7.93
eSCH2CH(CH3)2
eSCHCH3CH2CH3
eSCH2CH(CH2CH3)2
d
9.6
[7] A. Shi, D. Wang, H. Wang, Y. Wu, H. Tian, Q. Guan, K. Bao, W. Zhang, Synthesis
and bioevaluation of 2-phenyl-5-methyl-2H-1, 2, 3-triazole-4-carboxylic acid/
carbohydrazide derivatives as potent xanthine oxidase inhibitors, RSC Adv. 6
(2016) 114879e114888.
a
Values are the mean of three experiments.
inhibitory activity in vitro, with predicted pIC50 values that were
consistent with experimental values. In conclusion, the QSAR and
docking models obtained in this study could provide better virtual
screening for design of febuxostat analogs, and indicated future
research directions, such as further exploration of hydrophobic
[8] S. Chen, T. Zhang, J. Wang, F. Wang, H. Niu, C. Wu, S. Wang, Synthesis and
evaluation
of
1-hydroxy/methoxy-4-methyl-2-phenyl-1H-imidazole-5-
carboxylic acid derivatives as non-purine xanthine oxidase inhibitors, Eur. J.
Med. Chem. 103 (2015) 343e353.
[9] G. Qi, Z. Cheng, X. Ma, L. Wang, D. Feng, Y. Cui, B. Kai, W. Lan, W. Zhang,
Synthesis and bioevaluation of 2-phenyl-4-methyl-1, 3-selenazole-5-
carboxylic acids as potent xanthine oxidase inhibitors, Eur. J. Med. Chem. 85
(2014) 508e516.
interactions and
p-p stacking.
[10] S. Ishibuchi, H. Morimoto, T. Oe, T. Ikebe, H. Inoue, A. Fukunari, M. Kamezawa,
I. Yamada, Y. Naka, Synthesis and structure-activity relationships of 1-
phenylpyrazoles as xanthine oxidase inhibitors, Bioorg. Med. Chem. Lett 11
(2001) 879e882.
Conflicts of interest
The authors declare no conflict of interest.
[11] S. Wang, J. Yan, J. Wang, J. Chen, T. Zhang, Y. Zhao, M. Xue, Synthesis of some
5-phenylisoxazole-3-carboxylic acid derivatives as potent xanthine oxidase
inhibitors, ChemInform 45 (2010) 2663e2670.
[12] A.K. Halder, S.A. Amin, T. Jha, S. Gayen, Insight into the structural re-
quirements of pyrimidine-based phosphodiesterase 10A (PDE10A) inhibitors
by multiple validated 3D QSAR approaches, SAR QSAR Environ. Res. 28 (2017)
253e273.
[13] R.D. Cramer, Topomer CoMFA: a design methodology for rapid lead optimi-
zation, J. Med. Chem. 46 (2003) 374e388.
[14] C. Xu, Y. Ren, Molecular modeling studies of [6, 6, 5] tricyclic fused oxazoli-
dinones as FXa inhibitors using 3D-QSAR, Topomer CoMFA, molecular docking
and molecular dynamics simulations, Bioorg. Med. Chem. Lett 25 (2015)
4522e4528.
Acknowledgments
This work was supported by the Science and Technology Project
of Guangdong Province (2015A020211005) and the Guangzhou
Science Technology and Innovation Commission (201704020036).
Appendix A. Supplementary data
[15] J. Li, F. Wu, X. Liu, Y. Zou, H. Chen, Z. Li, L. Zhang, Synthesis and bioevaluation
of 1-phenyl-pyrazole-4-carboxylic acid derivatives as potent xanthine
oxidoreductase inhibitors, Eur. J. Med. Chem. 140 (2017) 20e30.
[16] T.J. Zhang, Q.X. Wu, S.Y. Li, L. Wang, Q. Sun, Y. Zhang, F.H. Meng, H. Gao,
Synthesis and evaluation of 1-phenyl-1H-1, 2, 3-triazole-4-carboxylic acid
derivatives as xanthine oxidase inhibitors, Bioorg. Med. Chem. Lett 27 (2017)
3812e3816.
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.molstruc.2019.01.017.
References
ꢀ
ꢀ
ꢀ
ꢀ
[1] D.T. Lowenthal, The treatment of hyperuricemia, Am. Fam. Physician 14
(1976) 98e100.
[2] E. Roddy, M. Doherty, Treatment of hyperuricaemia and gout, Clin. Med. 13
(2013) 400e403.
[3] F. Borges, E. Fernandes, F. Roleira, Progress towards the discovery of xanthine
oxidase inhibitors, Curr. Med. Chem. 9 (2002) 195e217.
[4] K. Okamoto, B.T. Eger, T. Nishino, S. Kondo, E.F. Pai, T. Nishino, An extremely
ꢁ
ꢁ
ꢁ
ꢁ
ꢀ ꢀ
[17] A. Smelcerovic, K. Tomovic, S. Zaklina, Z. Petronijevic, G. Kocic, T. Tomasic,
ꢀ
Z. Jakopin, M. Anderluh, Xanthine oxidase inhibitors beyond allopurinol and
febuxostat; an overview and selection of potential leads based on in silico
calculated physico-chemical properties, predicted pharmacokinetics and
toxicity, Eur. J. Med. Chem. 135 (2017) 491e516.
ꢁ
[18] S.L.D. Silva, S. Marangoni, K.C. Weber, P. Homem-De-Mello, K.M. Honorio,

X. Li et al. / Journal of Molecular Structure 1181 (2019) 428e435
Chem. Soc. 110 (1988) 5959e5967.
435
A.B.F.D. Silva, A multivariate study on the inhibition of the xanthine oxidase
by phenylpyrazole compounds, Int. Electron. J. Mol. Des. 4 (2005) 515e526.
[19] J.T. Leonard, K. Roy, On selection of training and test sets for the development
of predictive QSAR models, Qsar Comb. Sci. 25 (2010) 235e251.
[20] Y. Xiang, Z. Hou, Z. Zhang, Pharmacophore and QSAR studies to design novel
histone deacetylase 2 inhibitors, Chem. Biol. Drug Des. 79 (2012) 760e770.
[21] P. Geladi, Notes on the history and nature of partial least squares (PLS)
modelling, J. Chemom. 2 (2010) 231e246.
[28] E. Vrontaki, G. Melagraki, T. Mavromoustakos, A. Afantitis, Searching for an-
thranilic acid-based thumb pocket 2 HCV NS5B polymerase inhibitors through
a combination of molecular docking, 3D-QSAR and virtual screening, J. Enzym.
Inhib. Med. Chem. 31 (2016) 38e52.
[29] C.Z. Hao, S.W. Xia, H. Wang, J. Xue, L. Yu, Using 3D-QSAR and molecular
docking insight into inhibitors binding with complex-associated kinases
CDK8, J. Mol. Struct. 1173 (2018) 498e511.
[22] P. Geladi, B.R. Kowalski, Partial least-squares regression: a tutorial, Anal. Chim.
Acta 185 (1985) 1e17.
[23] B.L. Bush, R.B. Nachbar, Sample-distance partial least squares: PLS optimized
for many variables, with application to CoMFA, J. Com Aid Mol. Des. 7 (1993)
587e619.
[24] M. Clark, R.D.C. Iii, N.V. Opdenbosch, Validation of the general purpose tripos
5.2 force field, J. Comput. Chem. 10 (1989) 982e1012.
[25] C.R. 3Rd, D.E. Patterson, J.D. Bunce, Recent advances in comparative molecular
field analysis (CoMFA), Prog. Clin. Biol. Res. 291 (1989) 161e165.
[26] A. Golbraikh, A. Tropsha, Beware of q2!, J. Mol. Graph. Model. 20 (2002)
269e276.
[30] P.S. Ambure, R.P. Gangwal, A.T. Sangamwar, 3D-QSAR and molecular docking
analysis of biphenyl amide derivatives as p38 mitogen-activated protein ki-
nase inhibitors, Mol. Divers. 16 (2012) 377e388.
[31] A. Fukunari, K. Okamoto, T. Nishino, B.T. Eger, E.F. Pai, M. Kamezawa,
I. Yamada, N. Kato, Y-,700 [1-[3-Cyano-4-(2, 2-dimethylpropoxy)phenyl]-1H-
pyrazole-4-carboxylic acid]: a potent xanthine oxidoreductase inhibitor with
hepatic excretion, J. Pharmacol. Exp. Therapeut. 311 (2004) 519e528.
[32] G.W. Van, J.P. Overington, A ligand's-eye view of protein similarity, Nat.
Methods. 10 (2013) 116e117.
[33] A. Ghaleb, A. Aouidate, M. Ghamali, A. Sbai, M. Bouachrine, T. Lakhlifi, 3D-
QSAR modeling and molecular docking studies on a series of 2, 5 disubstituted
1, 3, 4-oxadiazoles, J. Mol. Struct. 1145 (2017) 278e284.
[27] R.D. Cramer, D.E. Patterson, J.D. Bunce, Comparative molecular field analysis
(CoMFA). 1. Effect of shape on binding of steroids to carrier proteins, J. Am.