Structure-guided development of purine amide, hydroxamate, and amidoxime for the inhibition of non-small cell lung cancer

Article abstract of DOI:10.1016/j.ejmech.2019.07.054

An 8-oxopurine-6-carboxamide compound (1a) was previously identified as an inhibitor of non-small cell lung cancer (NSCLC). In this study, more than 30 purine-6-carboxamide derivatives with variations at the C2, N7, C8, and N9 positions were synthesized to investigate the structure?activity relationship as a basis for the construction of an advanced pharmacophore model. This model suggests that purine-6-hydroxamate and purine-6-amidoxime analogs could form more hydrogen bonds with a target protein to enhance the inhibitory activities against H1975 cells. Among the series of analogs, hydroxamate 17 and amidoxime 19a exhibited excellent potency against H1975 cells (IC₅₀ < 1.5 μM) and other lung cancer cells with either wild-type or mutated epidermal growth factor receptor (EGFR). Mouse experiments indicated that compounds 17 and 19a were efficient anticancer agents with no appreciable toxicity. The mechanisms of action for the induction of cell apoptosis were determined to involve microtubule fragmentation and p53-mediated signaling pathways.

Full text of DOI:10.1016/j.ejmech.2019.07.054

Accepted Manuscript
Structure-guided development of purine amide, hydroxamate, and amidoxime for the
inhibition of non-small cell lung cancer
Meng-Ruo Huang, Yuan-Lin Hsu, Tzu-Chen Lin, Ting-Jen Cheng, Ling-Wei Li, Yu-
Wei Tseng, Yu-shu Chou, Jyung-Hurng Liu, Szu-Hua Pan, Jim-Min Fang, Chi-Huey
Wong
PII:
S0223-5234(19)30675-0
DOI:
https://doi.org/10.1016/j.ejmech.2019.07.054
EJMECH 11551
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 21 May 2019
Revised Date: 7 July 2019
Accepted Date: 19 July 2019
Please cite this article as: M.-R. Huang, Y.-L. Hsu, T.-C. Lin, T.-J. Cheng, L.-W. Li, Y.-W. Tseng, Y.-
s. Chou, J.-H. Liu, S.-H. Pan, J.-M. Fang, C.-H. Wong, Structure-guided development of purine amide,
hydroxamate, and amidoxime for the inhibition of non-small cell lung cancer, European Journal of
Medicinal Chemistry (2019), doi: https://doi.org/10.1016/j.ejmech.2019.07.054.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Structure-guided development of purine amide, hydroxamate, and amidoxime for the
inhibition of non-small cell lung cancer
a
b
a
c
a
Meng-Ruo Huang , Yuan-Lin Hsu , Tzu-Chen Lin , Ting-Jen Cheng , Ling-Wei Li , Yu-Wei
a
a
d, e, f, g,
b, h, i,
a, c,
Tseng , Yu-shu Chou , Jyung-Hurng Liu
*, Szu-Hua Pan
*, Jim-Min Fang
*,
c
and Chi-Huey Wong
a
Department of Chemistry, National Taiwan University, Taipei 106, Taiwan
b
Graduate Institute of Medical Genomics and Proteomics, College of Medicine, National
Taiwan University, Taipei 100, Taiwan
c
The Genomics Research Center, Academia Sinica, Taipei 115, Taiwan
Institute of Genomics and Bioinformatics, National Chung Hsing University, Taichung 402,
Taiwan
d
e
f
Department of Life Sciences, National Chung Hsing University, Taichung 402, Taiwan
Agricultural Biotechnology Center, National Chung Hsing University, Taichung 402, Taiwan
Rong Hsing Research Center for Translational Medicine, National Chung Hsing University,
Taichung 402, Taiwan
g
h
Doctoral Degree Program of Translational Medicine, National Taiwan University, Taipei 100,
Taiwan
i
Genome and Systems Biology Degree Program, National Taiwan University and Academia
1

ACCEPTED MANUSCRIPT
Sinica, Taipei 100, Taiwan
*
Corresponding authors. Institute of Genomics and Bioinformatics, National Chung Hsing
University, Taichung 402, Taiwan. Graduate Institute of Medical Genomics and Proteomics,
College of Medicine, National Taiwan University, Taipei 100, Taiwan. Department of
Chemistry, National Taiwan University, Taipei 106, Taiwan.
E-mail addresses: jhliu@nchu.edu.tw (J.-H. Liu), shpan@ntu.edu.tw (S.-H. Pan),
jmfang@ntu.edu.tw (J.-M. Fang)
ABSTRACT
An 8-oxopurine-6-carboxamide compound (1a) was previously identified as an inhibitor
of non-small cell lung cancer (NSCLC). In this study, more than 30 purine-6-carboxamide
derivatives with variations at the C2, N7, C8, and N9 positions were synthesized to
investigate the structure−activity relationship as a basis for the construction of an advanced
pharmacophore model. This model suggests that purine-6-hydroxamate and
purine-6-amidoxime analogs could form more hydrogen bonds with a target protein to
enhance the inhibitory activities against H1975 cells. Among the series of analogs,
hydroxamate 17 and amidoxime 19a exhibited excellent potency against H1975 cells (IC <
5
0
2

ACCEPTED MANUSCRIPT
1
.5 µM) and other lung cancer cells with either wild-type or mutated epidermal growth factor
receptor (EGFR). Mouse experiments indicated that compounds 17 and 19a were efficient
anticancer agents with no appreciable toxicity. The mechanisms of action for the induction of
cell apoptosis were determined to involve microtubule fragmentation and p53-mediated
signaling pathways.
Keywords:
Lung cancer; Purine Amide; Amidoxime; Hydroxamate; Microtubule.
3

ACCEPTED MANUSCRIPT
Table of Contents (Graphic)
4

ACCEPTED MANUSCRIPT
Research highlights
•
•
•
Purine-6-amides, hydroxamate and amidoximes are synthesized by fragment-based strategy.
Advanced pharmacophore model is constructed with strong predictive power.
Amidoxime 19a binds katanin to induce microtubule fragmentation and other signaling
pathways.
•
Compound 19a is active toward most EGFR wild-type and mutated lung adenocarcinoma
cells.
5

ACCEPTED MANUSCRIPT
Abbreviations:
ALT, alanine transaminase; AST, aspartate transaminase; BCA, bicinchoninic acid; Boc,
tert-butoxycarbonyl; BUN, blood urea nitrogen; CC, correlation coefficient; CDI,
1
,1′-carbonyldiimidazole; CREA, creatinine; DAPI, 4′,6-diamidino-2-phenylindole; DIEA,
N,N-diisopropylethylamine; DMF, N,N-dimethyl formamide; DMSO, dimethylsulfoxide; DS,
Discovery Studio; ECL, enhanced chemiluminescence; EGFR, epidermal growth factor
receptor; ELISA, enzyme-linked immunosorbent assay; FBS, fetal bovine serum; FITC,
fluorescein isothiocyanate; H&E, hematoxylin and eosin; HBA, hydrogen bond acceptor;
HBD, hydrogen bond donor; HBTU, (2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium
hexafluorophosphate; HPLC, high-performance liquid chromatography; HY, hydrophobic; IP,
immunoprecipitation; JNK, c-Jun N-terminal kinase; MD, molecular dynamics; MTA,
microtubule targeting agent; NSCLC, non-small cell lung cancer; PBS, phosphate-buffered
saline; QSAR, quantitative structure–activity relationship; RA, ring aromatic; RMSD,
root-mean-squared difference; SAR, structure−activity relationship; SD, standard deviation;
SRB, sulforhodamine B; TFAA, trifluoroacetic anhydride; THF, tetrahydrofuran; TKI,
tyrosine kinase inhibitor; TUNEL, terminal deoxynucleotidyl transferase dUTP nick end
labeling.
6

ACCEPTED MANUSCRIPT
1
. Introduction
Lung cancer is one of the most fatal diseases worldwide [1]. Epidermal growth factor
®
receptor (EGFR) tyrosine kinase inhibitors (TKI), such as gefitinib (iressa ) and erlotinib
®
(
tarceva ), are used in clinical therapy to treat non-small cell lung cancer (NSCLC) [2−4].
However, the T790M mutation of EGFR in NSCLC has led to resistance to these clinical
EGFR TKIs [5]. To overcome this problem, new NSCLC drugs have been developed, such as
®
®
afatinib (gilotrif ) [3, 6] and osimertinib (tagrisso ) [7].
Compounds with other mechanisms have also been considered to target
EGFR-independent pathways as another approach. For example, microtubule-targeting agents
(MTA) such as vinca alkaloids and taxanes have also been applied in cancer treatment [8].
Microtubules are tubulin-composing polymers that play significant roles in maintaining
cellular structure, transporting vesicles, and constructing mitosis process [9, 10]. During
mitosis, microtubules separate duplicated chromosomes into two different cells through the
formation of mitotic spindles [10]. This process is extremely important in cell division, which
makes the microtubules a desirable target for anticancer therapy.
Recently, using cell-based high-throughput screening, we identified a purine-type
compound GRC0321 (1a in Fig. 1) as a novel MTA that has good effects against NSCLC. It
can target katanin, which is an important microtubule-severing protein that is composed of
p60 and p80 subunits [11–13]. Furthermore, it can induce microtubule fragmentation, activate
7

ACCEPTED MANUSCRIPT
c-Jun N-terminal kinases (JNK)-mediated cell apoptosis, lead to cell cycle arrest at the G2/M
phase, and suppress tumor growth [14].
Fig. 1. Structure of purine-type compound 1a (GRC0321) and possible roles of the 8-oxo,
6
-carboxamide, C2-aryl, and ethoxy groups in binding with katanin as predicted by an initial
D-QSAR pharmacophore model [14]. HBA, hydrogen bond acceptor; HBD, hydrogen bond
3
donor; RA, ring aromatic; HY, hydrophobic.
We also generated a 3D-QSAR pharmacophore model based on the structures of
compound 1a, as well as nine analogs with variation of the aryl substituents at the C2 and N9
positions [14]. The initial pharmacophore model suggests a quantitative correlation between
the chemical structures of the purine-type compounds and their NSCLC inhibition activities.
For example, hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), ring aromatic
(RA), and hydrophobic (HY) interactions are mapped to the 8-oxo, 6-carboxamide, C2-aryl,
and ethoxy groups, respectively. The initial model used compounds of high structural
homology with a narrow activity range, which resulted in only small differences between the
total and null cost value (∆Cost = 54.763). According to the Discovery Studio (DS) manual,
∆Cost ≥ 60 is typical for a suitable pharmacophore model with 90% statistical significance.
Therefore, we aimed to build a better model with ∆Cost greater than 60 to guide studies on
the structure–activity relationship and to enable efficient lead optimization.
8

ACCEPTED MANUSCRIPT
In this study, we applied a fragment-based approach to synthesize a series of new
purine-type derivatives and evaluated their inhibitory activities against the growth of H1975
lung cancer cells. H1975 is a cell line of human lung adenocarcinoma that harbors EGFR
L858R/ T790M double mutations and is resistant to most of the EGFR TKIs. Fragment-based
drug design is a powerful strategy for drug discovery [15]. Based on the chemical structures
of hit compounds, a series of derivatives can be prepared in a combinatorial or relatively
accessible manner to build up a chemical library for studies of the structure−activity
relationship (SAR). We also constructed a better pharmacophore model with strong predictive
power to look into the SAR for development of anticancer drugs.
2
. Results and discussion
2
.1. Synthesis of 8-oxopurine-6-carboxamides with structural variation at N9 and C2
positions
In the primitive pharmacophore model (Fig. 1), a hydrophobic region is mapped onto the
para-position of the C2-phenyl ring, but no information is provided about the structural
feature at N9. To understand the influence of N9 and C2 substituents on the proliferation of
1
2
cancer cells, we synthesized a series of 8-oxopurine-6-carboxamides with different R and R
groups (Scheme 1).
9

ACCEPTED MANUSCRIPT
Scheme 1. Synthesis of 8-oxopurine-6-carboxamides (1a−1j, 1k−1x and 1a′−1h′) with
1
2
variation of R and R substituents at the N9 and C2 positions.
Reagents and conditions: (a) THF, rt, 15−40 h; (b) Et N, MeOH, rt, 15−68 h. Overall yields
3
(%): 1a (32), 1b (38), 1c (28), 1d (36), 1e (38), 1f (39), 1g (52), 1h (37), 1i (26), 1j (46), 1k
(34), 1l (50), 1m (38), 1n (30), 1o (37), 1p (41), 1q (37), 1r (44), 1s (47), 1t (20), 1u (44), 1v
(37), 1w (42), 1x (44), 1a′ (37), 1b′ (39), 1c′ (40), 1d′ (36), 1e′ (42), 1f′ (33), 1g′ (12), 1h′
(3).
The condensation reactions of diaminomaleonitrile (2) with various aryl isocyanates (3)
afforded the urea compounds 4a−4j, which were subsequently treated with
4
-ethoxybenzaldehyde to obtain the corresponding purine-type compounds 1a−1j, which
1
carry the R substituent with either electron-donating or electron-withdrawing properties at
the N9 position [16]. Urea compound 4a was reacted with a variety of (substituted)
benzaldehydes to afford purine-type compounds 1k−1x, which have different phenyl groups
2
(
R ) at the C2 position. Furthermore, a similar procedure was used to synthesize the
1
0

ACCEPTED MANUSCRIPT
purine-type compounds 1a′−1h′, which have at least one aliphatic or heteroaromatic
substituent at the N9 or C2 position.
The growth-inhibiting activities of the synthesized purine-type compounds against H1975
2
cells were further examined (Table 1). Compounds 1a−1j all have the same substituent (R =
4
-ethoxyphenyl) at the C2 position. In this series, an IC value of 1.1 ± 0.3 µM was obtained
50
for compound 1e, which has an unsubstituted phenyl group at the N9 position. In comparison,
there tended to be higher inhibitory activity for compounds 1a, 1b, and 1d, in which the
N9-phenyl group is decorated with an electron-donating substituent (MeO or Et) at the ortho-
or meta- position. In contrast, the inhibitory activity deteriorated in compounds 1f−1i, which
have electron-withdrawing substituents (CO Me, NO or CF ) on the N9-phenyl group. After
2
2
3
replacing the phenyl group in 1e with a 1-naphthyl group, compound 1j also showed less
inhibitory activity (IC = 12.1 ± 1.6 µM).
5
0
An IC value of 31.6 ± 1.9 µM was obtained for compounds 1a and 1k–1x, which have a
5
0
2
-methoxyphenyl group at the N9 position, as well as compound 1s, which has a phenyl
group at the C2 position. The inhibitory activity increased in most of the analogous
compounds (1a, 1k, 1n−1r, and 1v−1x), which have a phenyl group bearing either
electron-donating or electron-withdrawing substituents. In the series of 1a, 1k, and 1l, the
inhibitory activity sequentially decreased when placing the ethoxy substituent at the para-,
meta-, and ortho-positions, respectively. A similar trend of decreasing activity was observed
1
1

ACCEPTED MANUSCRIPT
in the series of 1o, 1q, and 1r, which have p-, m-, and o-bromophenyl groups at the C2
position.
Compounds 1m−1p have different halide substituents at the para-position of the C2
phenyl ring. The most potent was compound 1p, which had an iodine substituent, while the
weakest activity toward H1975 cells occurred for compound 1m, which had a fluorine
substituent. Compound 1w had a 3-hydroxy-4-methoxyphenyl group at the C2 position and
exhibited excellent inhibitory activity (IC = 0.6 ± 0.1 µM), but inactivity was observed for
5
0
compounds 1t and 1u, which had p-hydroxyl and p-acetamido substituted phenyl groups (IC50
100 µM). This result is consistent with our previously proposed binding model (Fig. 1),
>
which suggested that a hydrophobic feature was favored at the para-position of the C2 phenyl
ring.
Our initial pharmacophore suggested the importance of aromaticity at the C2, which is
supported by the observation of the inactive compound 1a′, which has an aliphatic substituent
at the C2 position. Compounds 1f′−1h′, which have aliphatic groups at the C2 and N9
positions, showed inferior inhibitory activities to the analogous compounds with aryl groups
(
e.g., 1e). The results revealed moderate activity (IC = 35.3 ± 1.4 µM) for compound 1e′,
50
which bears a 2-thienyl group at the C2 position. The result was also similar to that of
compound 1s, which has a C2 phenyl group (IC = 31.6 ± 1.9 µM). However, the biological
5
0
activities were lost for compounds 1b′−1d′, which have furan or pyridine rings at the C2
1
2

ACCEPTED MANUSCRIPT
position. This phenomenon inspired us to modify our primitive pharmacophore in view of the
aromaticity at C2.
1
3

ACCEPTED MANUSCRIPT
Table 1
Inhibitory activities of 8-oxopurine-6-carboxamides against the growth of H1975 lung cancer
cells.
O
NH2
N
H
N
O
N
R1
R2
N
1
2
a
Cmpd
R =
R =
IC (µM)
50
1
a
2-MeOC
3-MeOC
4-MeOC
6
6
6
H
H
H
4
4
4
4-EtOC
4-EtOC
4-EtOC
4-EtOC
4-EtOC
4-EtOC
4-EtOC
4-EtOC
4-EtOC
4-EtOC
3-EtOC
2-EtOC
6
6
6
6
6
6
6
6
6
6
6
6
H
H
H
H
H
H
H
H
H
H
H
H
4
4
4
4
4
4
4
4
4
4
4
4
0.6 ± 0.3
0.6 ± 0.2
5.7 ± 0.3
0.2 ± 0.0
1.1 ± 0.3
1
b
1
c
1
d
2-EtC H
6 4
1
e
C H
6 5
1
f
3-MeO
2
2
CC
CC
6
6
H
4
4
27.3 ± 9.0
>100
1
g
4-MeO
H
1
h
3-O
3,5-(F
1-naphthyl
2
NC
6
H
4
33.0 ± 20.1
16.8 ± 1.4
12.1 ± 1.6
7.2 ± 2.1
86.1 ± 11.4
57.9 ± 6.6
7.7 ± 0.9
7.2 ± 0.2
3.5 ± 0.7
13.2 ± 1.5
29.2 ± 5.6
31.6 ± 1.9
>100
1
i
3
C)
2
C
6
H
3
1
j
1
k
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
2-MeOC
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
H
4
1
l
1
m
4-FC H
6 4
1
n
4-ClC
H
6 4
1
o
p
q
4-BrC
6
H
4
1
4-IC H
6 4
1
3-BrC
2-BrC
6
6
H
4
4
1
r
s
H
1
C H
6 5
1
t
4-HOC H
6 4
1
u
4-AcHNC
6
H
4
>100
1
v
(3-F-4-Br)C
6
H
3
4.0 ± 0.5
0.6 ± 0.1
7.6 ± 1.1
>100
1
w
(3-HO-4-MeO)C
3,4-(OCH O)C H
6 3
6 3
H
1
x
2
1
a′
b′
c′
d′
e′
f′
g′
isopropyl
2-furanyl
3-furanyl
4-pyridyl
2-thienyl
1
>100
1
>100
1
>100
1
35.3 ± 1.4
>100
1
n-hexyl
cyclohexyl
4-EtOC
4-EtOC
6
6
H
4
4
1
H
30.7 ± 8.1
1
4

ACCEPTED MANUSCRIPT
1
h′
cyclohexyl
Et
68.0 ± 1.3
a
IC value is half the maximal inhibitory concentration determined by cell proliferation assay
5
0
in at least three independent experiments.
2
.2. Modification of 8-oxopurine core structure
To modify the core structure of 8-oxopurine, direct alkylation at N7 was attempted by
treating 1a with iodomethane under various conditions using Et N, i-Pr NEt, NaOH, KOH,
3
2
K CO , or t-BuONa as a base. However, all of these attempts failed, presumably because NH
2
3
had a strong intramolecular hydrogen bond with the C6 amide group, which prevented
alkylation. To solve this problem, the amide group in 1a was converted to a cyano group by
treatment with trifluoroacetic anhydride [17], which formed compound 6 (Scheme 2). This
was done to overcome the interference from the intramolecular hydrogen bond. The
N-alkylation of 6 with different alkyl halides was thus successfully carried out and gave
compounds 8a−8c.
No O-alkylation occurred when relatively soft halides (iodomethane, allyl bromide, or
benzyl bromide) were used as the alkylating agent. The N-alkylation compounds 8a, 8b, and
1
3
8
c were characterized by the N-CH and N-CH signals at δ 28.2, 43.6, and 45.1 in the C
3 2 C
NMR spectra, respectively. As shown in the transformation of 8a to amide 9, the cyano group
was reverted to an amide group by hydrolysis in alkaline conditions using hydrogen peroxide
as a promoter [18]. Alternatively, the O-acylation product 7 was obtained from the reaction of
compound 6 with ethyl chloroformate [19]. The signals of C8 and carbonate in compound 7
1
5

ACCEPTED MANUSCRIPT
occurred at δ 147.8 and 162.0, which is in agreement with the literature [19]. The acylation
C
favored at the relatively hard oxygen atom (compared to a nitrogen atom) was attributable to
the relatively hard characteristic of ethyl chloroformate.
Scheme 2. Synthesis of compounds 6−15 by modification at the core structure of 8-oxopurine.
o
o
Reagents and conditions: (a) TFAA, pyridine, 0 C, 3 h; (b) ClCO Et, DIEA, CH Cl , 0 C, 3 h;
2
2
2
(
c) MeI or allyl bromide or benzyl bromide, NaH, DMF, rt, 2−4 h; (d) H O , NaOH, CH Cl ,
2 2 2 2
o
MeOH, rt, 50 h; (e) POCl , 100 C, 72 h; (f) 7 M NH in MeOH, rt, 24 h, then Boc O, NaH,
3
3
2
THF, rt, 19 h; (g) THF, rt, 48 h; (h) Et N, MeOH, rt, 21 h; (i) MeI, Et N, DMF, rt, 18 h. Yields:
3
3
6
(91%), 7 (67%), 8a (94%), 8b (43%), 8c (64%), 9 (72%), 10 (8%), 11 (45% for 2 steps), 14′
20% for 2 steps), 15 (78%). Boc = tert-butoxycarbonyl; DIEA = N,N-diisopropylethylamine;
TFAA = trifluoroacetic anhydride.
(
1
6

ACCEPTED MANUSCRIPT
In addition to N-alkylated and O-acylated purine derivatives, we also synthesized
8-aminopurine (Scheme 2B) and 8-mercaptopurine derivatives (Scheme 2C). The
8-oxopurine-6-carbonitrile 6 was first treated with phosphoryl chloride to provide
8-chloropurine-6-carbonitrile 10 [20]. Further reaction of 10 with ammonia afforded an
intermediate 8-aminopurine-6-carbonitrile, which was then elaborated to carbamate using
di-tert-butyl dicarbonate, resulting compound 11. Starting with the condensation reaction of
diaminomaleonitrile (2) with phenyl isothiocyanate, the intermediate product 13 was reacted
with 4-ethoxybenzaldehyde to provide 8-mercaptopurine-6-carboxamide 14′ as a tautomer of
8
-thionopurine-6-carboxamide 14. The reaction of 14′ with iodomethane occurred exclusively
at the relatively soft sulfur atom, giving the S-alkylation product 15.
Table 2
Inhibitory activities of compounds with modification of the 8-oxopurine core structure.
a
a
Compound IC (µM)
Compound IC (µM)
50
5
0
6
7
16.4 ± 3.1
>100
9
>100
11
14′
15
37.7 ± 1.8
20.6 ± 1.4
>100
8
a
>100
8
b
>100
b
8
c
>100
a
b
IC50 values against the growth of H1975 lung cancer cells were determined by cell
proliferation assays in at least three independent experiments.
IC = 500 ± 27.6 µM.
5
0
Table 2 shows that compounds 11 and 14′ with carbamate and mercapto groups at the C8
position were still active against H1975 cells, although they were somewhat less effective
1
7

ACCEPTED MANUSCRIPT
than 8-oxopurine 6. After incorporation of alkyl or acyl groups, compounds 7, 8a−8b, 9, and
1
5 lost their activities.
2
.3. Construction of an advanced pharmacophore model
We were able to construct new 3D-QSAR pharmacophore models based on the IC50
values (Tables 1, 2) of the synthesized compounds containing the modified derivatives of
phenyl- and 8-oxopurine subsets. In the initial trial (supplementary Table S1), we found that
most models were mainly composed of HBD and HY features. Therefore, we increased the
weightings of HBD and HY in the calculation and generated a second series of advanced
pharmacophore models. Table 3 shows the results for the top 10 pharmacophore hypotheses
and their statistical parameters.
According to the Discovery Studio (DS) manual, a cost difference (∆Cost) of more than
6
0 bits between the fixed and null cost indicates a high chance (90%) of achieving statistical
significance with good pharmacophore models. We found that the first pharmacophore
hypothesis (Hypo-1) is the best because this model has the lowest total cost value (186.937),
the largest cost difference (293.777), the lowest root-mean-squared difference (RMSD)
(1.722), and the highest correlation coefficient (0.920). We also performed Fischer
randomization calculations to validate the statistical robustness of Hypo-1 (supplementary Fig.
S1). At the 95% confidence level, 19 models were generated using randomly assigned IC50
1
8

ACCEPTED MANUSCRIPT
values and compared with Hypo-1. The total cost of Hypo-1 was lower than that of the 19
random models (supplementary Fig. S1A), which meant that Hypo-1 was valid and stable. In
addition, the correlation coefficient of Hypo-1 was better than those of the randomly
generated 19 models (supplementary Fig. S1B). This result strongly suggests that Hypo-1 was
not created randomly and indeed predicted the relationship between the structures and
bioactivity in the training.
Table 3
Statistical significance, predictive power, and features of the top 10 hypotheses derived from
a
3
D-QSAR pharmacophore generation.
e
b
c
d
Features
Hypothesis Correlation RMSD Total cost ∆Cost
Max. Fit
HBD
HY
Hypo-1
Hypo-2
Hypo-3
Hypo-4
Hypo-5
Hypo-6
Hypo-7
Hypo-8
Hypo-9
Hypo-10
0.920
0.868
0.844
0.824
0.807
0.807
0.791
0.790
0.781
0.775
1.722
2.179
2.357
2.488
2.596
2.597
2.687
2.691
2.744
2.775
186.937 293.777
224.846 255.868
241.355 239.359
254.550 226.164
265.190 215.524
265.447 215.267
275.259 205.455
278.946 201.768
284.942 195.772
286.312 194.402
2
3
9.462
7.149
7.133
6.334
6.853
6.566
6.458
4.725
4.609
5.641
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
a
b
c
The null cost, fixed cost, and configuration cost are 480.714, 127.177, and 30.231
respectively.
The correlation coefficient is calculated based on linear regression derived from the
geometric fit index.
Root mean square difference of log[estimated activity] from log[measured activity]
normalized to log[uncertainty].
d
e
∆
Cost = null cost – total cost of the hypothesis.
HBD, hydrogen-bond donor; HY, hydrophobic.
1
9

ACCEPTED MANUSCRIPT
Hypo-1 contains two hydrogen-bond donors and three hydrophobic features (Fig. 2). Fig.
2
A shows the 3D spatial arrangement and distance constraints of these features. According to
this model, as exemplified by compound 1a (Fig. 2B), hydrogen-bond donors are mapped
onto 6-amido and 7-NH groups, while hydrophobic regions are mapped onto C2 phenyl ring,
the alkoxy substituent at the para- position, and the methoxy substituent at the ortho-position
of the N9 phenyl ring, respectively. To demonstrate the predictive power of Hypo-1, a
regression analysis was performed to estimate the activity of each training-set compound. As
shown in supplementary Table S2, most of the training set compounds showed an error value
of less than 5. An error value below 10 indicates only one order of magnitude of difference
between the estimated and actual activity. This result indicates that Hypo-1 is a reliable model
that consists of reasonable chemical features of the training-set compounds and accurately
estimates their experimental activity.
Fig. 2. The best pharmacophore model (Hypo-1) established from the modified derivatives of
phenyl- and 8-oxopurine subsets. (A) The pharmacophore model Hypo-1 represented with
2
0

ACCEPTED MANUSCRIPT
distance constraints. (B) Hypo-1 mapping with compound 1a. Pharmacophore features are
color-coded as follows: magenta, hydrogen bond donor (HBD); cyan, hydrophobic (HY).
In a previous study [14], we used 10 purine-type compounds to create a pharmacophore
model with a correlation coefficient of 0.984 and a maximum fit value of 8.465. However, its
∆Cost value was 54.763, and it failed to achieve 90% statistical significance. In this study,
with more newly synthesized compounds, we built a new model based on 40 purine-type
compounds, Hypo-1. This new model has a better ∆Cost value (293.777) and maximum fit
value (9.462). Although its correlation coefficient (0.920) is weaker than that of the previous
model, we conclude that it is more statistically reliable because it passed the Fischer
randomization validation and had a higher ∆Cost value.
The hydrophobic mapping onto the C2-aryl ring in the new model was validated by
several examples. Compounds containing a more hydrophilic pyridine (1d′) or furan (1b′)
ring are less favorable in the model, and their inhibitory activities are inferior to those
containing a hydrophobic phenyl (1s) or thiophene (1e′) ring. Furthermore, the hydrophobic
feature of the ethoxy group at the para-position of the C2 phenyl ring (e.g., 1a) also exerts an
additional hydrophobic interaction. As a result, compound 1a has higher potency than the
analogous compound 1k with an ethoxy substituent at the meta-position. Compound 1l with
an ethoxy substituent at the ortho-position is the least active. Moreover, compounds 1t and 1u
containing hydrophilic OH or NHAc substituents at the para-position are inactive toward
H1975 cells. The present pharmacophore model also reveals a hydrophobic feature mapped
2
1

ACCEPTED MANUSCRIPT
onto the 2-methoxyphenyl moiety at N9. Thus, we anticipate that the analogous compound 1d
with a 2-ethylphenyl group will also have a potent effect.
It is predicted that 6-amide and 7-NH are crucial hydrogen-bond donors for the
inhibition of H1975 cells. For example, the inhibitory activity deteriorates when the HBD
characteristic of 6-amide in 1a is eliminated by transformation into a cyano group in 6.
Further alkylation of NH in 6 renders the derivatives 8a−8c inactive toward H1975 cells. A
similar phenomenon was observed when OH in 6′ was blocked by acylation and when SH in
thiol 14′ was converted to SCH by methylation.
3
2
.4. Elaboration of 6-amide to 6-hydroxamic acid and 6-amidoxime
The 6-amide group and HBD have important roles in the suppression of cancer cell
growth, as predicted by the pharmacophore model. Inspired by these roles, we synthesized a
series of purine-type compounds bearing the 6-hydroxamate or 6-amidoxime group. We
speculated that hydroxamate and amidoxime groups containing the NH−OH moiety would
enhance the HBD potency. As shown in Scheme 3A, the amide compound 1a was hydrolyzed
with an aqueous solution of NaOH and gave the carboxylic acid 16, which was activated by
1
,1′-carbonyldiimidazole and then treated with hydroxylamine to produce the corresponding
hydroxamic acid 17 [21]. The purine-6-carbonitrile 6 obtained from amide 1a was treated
with hydroxylamine to afford amidoxime 19a (Scheme 3B) [22]. The X-ray diffraction
2
2

ACCEPTED MANUSCRIPT
analysis of amidoxime 19a indicates that the 6-amidoxime preferably exists in the
Z-configuration, as shown in Scheme 3B and supplementary Fig. S4.
Scheme 3. Synthesis of purine hydroxamic acid (17) and amidoximes (19a-series and 21).
Reagents and conditions: (a) NaOH, EtOH, reflux, 3 h; (b) CDI, THF, rt, 2 h, then
o
NH OH∙HCl, 64 h; (c) TFAA, pyridine, 0−25 C, 2−26 h; (d) NH OH∙HCl, AcONa∙3H O,
2
2
2
o
THF, MeOH, rt−60 C, 1−3 h. Overall yields (%): 16 (59), 17 (33), 18f (71), 18j (71), 18m
54), 18n (48), 18o (77), 18q (53), 18r (76), 18s (65), 18c′ (63), 18e′ (80), 19a (84), 19f (97),
9j (94), 19m (71), 19n (62), 19o (78), 19q (67), 19r (95), 19s (91), 19c′ (95), 19e′ (62), 20
(
1
(84), 21 (90). CDI = 1,1′-carbonyldiimidazole; TFAA = trifluoroacetic anhydride.
Using a similar procedure, a series of 8-oxopurine-6-carboxamides bearing different aryl
substituents at the C2 and N9 positions were treated with trifluoroacetic anhydride to provide
the corresponding 6-cyano compounds, which underwent nucleophilic additions with
2
3

ACCEPTED MANUSCRIPT
hydroxylamine and gave the desired 8-oxopurine-6-amidoxime derivatives. The
8
-mercaptopurine-6-carboxamide (14′) was also converted to the corresponding
-mercaptopurine-6-amidoxime (21) via the 6-cyano compound 20 as an intermediate
8
(
Scheme 3C).
The IC50 values of the purine-type derivatives bearing either hydroxamic acid or an
amidoxime group at the 6-position are listed in Table 4. The hydroxamate 17 and amidoxime
9a both exhibited excellent potency against H1975 cells (IC < 1.5 µM). In the comparison
1
5
0
of the IC values in Table 1 and Table 4, the amidoxime derivatives (19a-series) consistently
5
0
showed 2−8 times higher inhibitory activity than their 6-amide precursors (1a-series). The
8
-mercaptopurine-6-amidoxime 21 also had 2.7 times better activity than its amide precursor
4′. These results confirm our pharmacophore model, which predicts that the 6-amidoxime
1
group would be an enhanced HBD to bind with the target protein.
Table 4
Inhibitory activities of purine hydroxamic acid and amidoxime against the growth of H1975
lung cancer cells.
a
a
Compound
IC (µM)
Compound IC (µM)
50
5
0
1
7
1.3 ± 0.2
1.1 ± 0.3
19q
19r
19s
19c′
19e′
21
5.9 ± 0.0
1
9a
16.0 ± 0.9
12.1 ± 0.3
32.9 ± 0.8
12.5 ± 0.0
7.5 ± 1.6
1
1
9f
9j
11.2 ± 0.1
3.0 ± 0.4
7.2 ± 0.1
3.8 ± 0.3
3.7 ± 0.8
1
9m
9n
9o
1
1
a
IC values were determined by cell proliferation assays in at least three independent
5
0
2
4

ACCEPTED MANUSCRIPT
experiments.
Table 5
The docking scores of purine-6-carboxamides (1a-series and 14′), purine-6-hydroxamate (17),
and purine-6-amidoxamides (19a-series and 21) using Autodock Vina.
b
Average
Predicted Binding
Energy (kcal/mol)
−10.7
Contribution of each interaction (%)
Cmpd
a
IC (µM)
Steric
75
72
68
83
71
81
71
78
66
76
65
75
63
78
68
78
68
77
66
79
68
77
67
79
68
Hydrophobic
H-bonding
20
5
0
1
1
a
7
0.6
5
5
5
5
4
6
6
6
5
5
5
5
5
5
5
5
5
6
5
3
2
3
3
5
5
1.3
1.1
−10.9
23
19a
−11.1
27
1
f
27.3
11.2
12.1
3.0
−10.6
12
19f
−11.1
25
1
j
−11.6
13
1
9j
m
−12.1
24
1
57.9
7.2
−10.8
17
19m
−11.4
28
1
n
7.7
−10.7
19
19n
3.8
−11.2
31
1
o
7.2
−10.5
20
19o
3.7
−10.9
32
1
q
13.2
5.9
−10.8
17
1
1
1
9q
−11.7
28
1
r
29.2
16.0
31.6
12.1
>100
32.9
35.3
12.5
20.6
7.5
−10.7
17
9r
−11.3
28
1
s
−10.5
17
9s
−11.1
29
1c′
−9.9
18
1
1
9c′
−10.5
30
1e′
−10.0
20
9e′
−10.6
30
14′
−10.5
16
2
1
−11.1
27
a
b
Data taken from Tables 1, 2, and 4.
The contribution of each interaction is expressed as the percentage of predicted binding
energy because the predicted free energy of binding is calculated from the summation of
weighted scoring function terms. The scoring function of AutoDock Vina consists of three
2
5

ACCEPTED MANUSCRIPT
steric terms (gauss1, gauss2, and repulsion) for all atom pairs, one hydrophobic term for
hydrophobic atoms, and one H-bonding term for hydrogen bonding pairs.
We performed molecular docking and molecular dynamics (MD) simulations with the
previously reported p60 model structure to gain insights into the binding mechanisms of the
hydroxamate 17 and amidoxime derivatives (19a-series and 21) [14]. The docking program
predicts that hydroxamate 17 will have a higher binding affinity (more negative binding
energy) than its amide precursor 1a (Table 5), although they have comparable IC values (0.6
5
0
±
0.3 µM for 1a and 1.3 ± 0.2 µM for 17). Notably, all the amidoxime derivatives (19a-series
and 21) exhibited better putative affinity than their amide precursors (1a-series and 14′). This
prediction is consistent with the experimental results of inhibitory activities against H1975
cells. Based on the scoring function [23], we calculated the contribution of each interaction to
the binding energy and found appreciable increases in the hydrogen-bonding interaction for
amidoxime derivatives. This finding indicates that the substitution of 6-amide group with
6
-amidoxime provides more hydrogen bonding between the compound and p60.
We examined the binding poses predicted by molecular docking to understand the
potential hydrogen-bonding in the environment around the 6-amide group (Fig. 3D, 3E, 3F).
These poses were further refined using MD simulation to optimize the ligand−protein
interaction (Fig. 3G, 3H, 3I). The MD simulation indicated that the predicted binding poses of
compounds 1a, 17, and 19a were similar.
2
6

ACCEPTED MANUSCRIPT
In the best pharmacophore model, hydrogen-bond donors are mapped onto 6-amide and
7
-NH groups of the purine-type compounds (Fig. 2B). In the binding pocket, Asp308
consistently forms a hydrogen bond with the 7-NH group (Fig. 3D−3I). The backbone amines
of Glu309 and Asp311 are hydrogen bond donors that bind with the carbonyl moiety of
6
-amide group in compound 1a (Fig. 3D, 3G). The amine moiety of the 6-amide group forms
a hydrogen bond with the backbone oxygen of Ala358 or with the sidechain oxygen of
Asp311. Furthermore, it can form an additional hydrogen bond with the sidechain oxygen of
Asn360.
For binding with the 6-hydroxamate group of compound 17, the backbone amines of
Glu309 and Asp311 still serve as hydrogen bond donors to the carbonyl moiety (Fig. 3E, 3H).
The amine moiety of hydroxamate serves as a hydrogen bond donor that interacts with the
backbone oxygen of Ala358 or with the sidechain oxygen of Asp311. The hydroxyl moiety of
hydroxamate also acts as another hydrogen bond donor that binds the sidechain oxygen of
Asn360. The sidechain of Asp311 changes orientation after MD refinement in the binding
poses of compounds 1a and 17 (Fig. 3G, 3H), which disrupts the formation of hydrogen
bonds between the amine moiety and the sidechain oxygen of Asp311. We speculate that this
hydrogen bonding may not be stable during the binding of compound 1a (or 17) to p60.
2
7

ACCEPTED MANUSCRIPT
Fig. 3. Putative binding poses of compounds 1a, 17, and 19a. Left panel: chemical structures
of 1a (A), 17 (B), and 19a (C); middle panel: docking poses of 1a (D), 17 (E), and 19a (F)
predicted by AutoDock Vina; right panel: binding poses of 1a (G), 17 (H), and 19a (I) refined
by MD simulation. The carbon atoms of the ligand are colored in pink. The carbon atoms of
amino acid residues involved in hydrogen bond formation are colored in green. Nitrogen
atoms are shown in blue, and oxygen is red. Yellow dashed lines represent potential hydrogen
bonds.
The 6-amidoxime group of compound 19a has more hydrogen bonds than the 6-amide
group (Fig. 3F, 3I). The hydroxyl moiety of amidoxime forms a hydrogen bond with the
backbone oxygen of Ala358 and the sidechain oxygen of Asp311. The imine moiety of
2
8

ACCEPTED MANUSCRIPT
amidoxime acts as a hydrogen bond acceptor and interacts with the backbone amines of
Glu309 and Asp311. The amine moiety of amidoxime can form hydrogen bonds with the
sidechain oxygen atoms of Asp311 and Asn360. The putative binding poses reveal a possible
molecular mechanism by which p60 interacts with 6-amidoxime derivatives.
Table 6
a−d
Inhibitory activities of 17 and 19a against different NSCLC cell lines.
f
IC (µM)
5
0
e
Cell line
EGFR status
1
7
19a
a
CL1-0
WT
WT
WT
WT
WT
MT
MT
MT
0.37 ± 0.003
0.29 ± 0.05
>5
0.39 ± 0.01
0.46 ± 0.02
3.47 ± 0.27
0.37 ± 0.01
0.48 ± 0.01
0.44 ± 0.02
0.47 ± 0.12
>5
a
CL1-5
a
CL83
a
CL141
0.32 ± 0.01
0.33 ± 0.01
0.47 ± 0.002
0.86 ± 0.09
>5
b
CL152
c
CL25
c
CL97
d
CL100
a
CL1-0, CL1-5, CL83, and CL141 are human lung adenocarcinoma cell lines with wild-type
EGFR.
b
c
d
e
f
CL152 is a human lung squamous cell line with wild-type EGFR.
CL25 and CL97 are human lung adenocarcinoma cell lines with mutated EGFR.
CL100 is a human lung squamous cell line with mutated EGFR.
WT: wild-type EGFR, MT: mutated EGFR.
The IC values were determined by sulforhodamine B (SRB) assay.
5
0
To extend the scope of our new compounds, hydroxamate 17 and amidoxime 19a were
incubated with several clinically isolated NSCLC cell lines that bear different EGFR statuses
Table 6). The compounds 17 and 19a displayed high potency toward both EGFR wild-type
(
2
9