Discovery of bis-aryl urea derivatives as potent and selective Limk inhibitors: Exploring Limk1 activity and Limk1/ROCK2 selectivity through a combined computational study

Article abstract of DOI:10.1016/j.bmc.2015.10.041

Lim kinase (Limk), a proline/serine-rich sequence, can regulate the polymerization of the actin filaments by phosphorylating, and it is found to be highly involved in various human diseases. In this paper, 47 reported Limk1 inhibitors with bis-aryl urea s

Full text of DOI:10.1016/j.bmc.2015.10.041

Bioorganic & Medicinal Chemistry 23 (2015) 7464–7477
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Discovery of bis-aryl urea derivatives as potent and selective Limk
inhibitors: Exploring Limk1 activity and Limk1/ROCK2 selectivity
through a combined computational study
a
a
a
Jiaxin Cui a, Mei Ding a, Wei Deng a, Yan Yin a,b, , Zhonghua Wang , Hong Zhou , Guofeng Sun ,
⇑
Yu Jiang a, Yangbo Feng c
a School of Chemical and Environmental Engineering, Shanghai Institute of Technology, 100 Hai Quan Rd, Shanghai 201418, PR China
b Key Laboratory of Synthetic Chemistry of Natural Substances, Shanghai Institute of Organic Chemistry, Chinese Academy of Sciences, 345 Ling Ling Rd, Shanghai 200032, PR China
c Medicinal Chemistry, Translational Research Institute, The Scripps Research Institute, Florida, Jupiter, FL 33458, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Lim kinase (Limk), a proline/serine-rich sequence, can regulate the polymerization of the actin filaments
by phosphorylating, and it is found to be highly involved in various human diseases. In this paper, 47
reported Limk1 inhibitors with bis-aryl urea scaffold were used to design potent and selective Limk inhi-
bitors by computational approaches. Firstly, the structure-Limk1 activity relationship models (3D-QSAR)
and structure-Limk1/ROCK2 selectivity relationship models (3D-QSSR) were developed and both 3D-
QSAR and 3D-QSSR models showed good correlative and predictive abilities. Then, the molecular docking
and molecular dynamics (MD) simulations were employed to validate the optimal docking conformation
and explore the binding affinities. Finally, five new compounds were designed and all of them exhibited
good Limk1 inhibition and Limk1/ROCK2 selectivity after synthesis and biological evaluation, which
demonstrated that the obtained information from computational studies were valuable to guide Limk
inhibitors’ design.
Received 30 September 2015
Revised 28 October 2015
Accepted 29 October 2015
Available online 30 October 2015
Keywords:
Limk1 inhibitors
Limk1/ROCK2 selectivity
Urea derivatives
Molecular modeling
Molecular docking
Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction
Although lots of studies about Limk were been done in recent
years,2,17,28–39 to the best of our knowledge few Limk inhibitors
LIM Kinase (Limk) is a serine-threonine protein and two iso-
forms are identified as Limk1 and Limk2.1,2 The structure of Limk
have been solved by solution NMR or X-ray diffraction,3 which con-
sists of two amino-terminal LIM domains, adjacent PDZ and pro-
line/serine-rich regions, following by a carboxyl-terminal protein
kinase domain. Limk is downstream of ROCK and PAK,4 and regu-
late the polymerization of actin filaments in the signaling path-
ways.5–9 Activated Limk can phosphorylate and inactivate cofilin,
thereby lead a dynamic regulation of actin cytoskeleton. Accumu-
lated evidence suggests that many disorders are correlated to the
regulator mechanism of Limk and the Limk inhibitors will be
potential drugs for a variety of human diseases including syn-
drome,10 Alzheimer’s disease,11 Parkinson’s disease,12 Williams’
disease,13 psoriatic epidermal lesions,14 preeclampsia’ disease,15
intracranial aneurysms,16 ocular hypertension/glaucoma,17 HIV
and other viral infections,18–20 and tumors.21–27
were described.2,17,28–36 Very recently our group reported aryl urea
derivatives as potent and selective Limk inhibitors.40 We also dis-
closed aryl ureas as ROCK inhibitors and explored their ROCK2/
PKA selectivity by 3D-QSAR, molecular docking and molecular
dynamic simulation.41 In this work, we will reveal the structural
and chemical properties that favor Limk1 inhibition and the
Limk1/ROCK2 selectivity for aryl urea derivatives through several
computational approaches. Comparative molecular field analyses
(CoMFA) and comparative molecular similarity indices analyses
(CoMSIA) are applied to obtain insights into key structural factors
that affect the inhibitory activity and selectivity of aryl urea based
Limk inhibitors. Molecular docking followed by molecular dynam-
ics (MD) simulations are then conducted to validate the models
and to further explore the origin of the selectivity at the amino acid
residue level. Finally, the residues which introduced ligand
potency and Limk1/ROCK2 selectivity are discovered and further
validated by chemical synthesis and biological evaluation. Herein,
the computer-aided drug design of highly potent and selective
Limk inhibitors will be described in detail, and the synthetic proce-
dures and structural characterization of the newly designed com-
⇑
Corresponding author. Tel.: +86 21 60877220; fax: +86 21 60877231.
E-mail address: yinyan@sit.edu.cn (Y. Yin).
http://dx.doi.org/10.1016/j.bmc.2015.10.041
0968-0896/Ó 2015 Elsevier Ltd. All rights reserved.

J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
7465
pounds will also be reported along with the biological experimen-
tal protocols.
The correlations between the experimental and predicted val-
ues of CoMFA 3D-QSAR model, CoMSIA 3D-QSAR model, CoMFA
3D-QSSR model and CoMSIA 3D-QSSR model were exhibited in
Figure 1. Most points were well distributed along the line Y = X also
suggested that all chosen models for this work were correlative
and predictive.
2. Results and discussion
2.1. Data set
2.3. Counter maps analysis
Chemical structures and IC50 values against Limk1 and ROCK2
of 47 reported Limk inhibitors were listed in Table 1.40 Forty com-
pounds with accurate Limk1 inhibitory activity were chosen to
build quantitative structure-Limk1 activity relationship (3D-QSAR)
models and sorted into training set and test set randomly (30 com-
pounds into training set and 10 compounds into test set). Twenty
two compounds with accurate Limk1 and ROCK2 inhibition IC50
values were collected for quantitative structure-Limk1/ROCK2
selectivity relationship (3D-QSSR) models and also sorted ran-
domly (17 into training set and 5 into test set).42
For the convenience of counter maps analysis, the urea-based
Limk1 inhibitors were divided into five regions: R1 region, R2
region, R3 region, R4 region, and Ar region, and the structure–activ-
ity/selectivity relationships derived from 3D-QSAR and 3D-QSSR
studies were summarized in Figure 2.
2.3.1. Contour maps of steric field
The steric field contour maps were shown in Figure 3. As shown in
Figure 3A and B, there was a big green contour around the N atom
(which is attached to the terminal phenyl ring), and two big yellow
contours in the R1 and R3 regions, which indicated that bulky groups
in R4 region were favored for Limk1 inhibition activity
(IC50Limk1 = 43 nM for 27 vs 142 nM for 18), and bulky groups in R3
and R1 regions were disfavored for Limk1 inhibition activity
(IC50Limk1 = 1090 nM for 23, 9240 nM for 24, >10,000 nM for 25 vs
35 nM for 22, and IC50Limk1 = 710 nM for 17 vs 62 nM for 12). Accord-
ing to Figure 3C and D, there was a big yellow contour at the 6-mem-
ber ring side of the bottom pyrrolopyrimidine and the 2-position of
middle phenyl ring, and some green contours at the terminal phenyl
ring. These data suggested that bulky groups at the bottom pyrim-
idine side and bulky groups in the R1 region were disfavored for
2.2. Statistical results
Since compound 47 was the most active Limk1 inhibitor and 36
was one of the most Limk1/ROCK2 selective Limk1 inhibitors
among those in Table 1, they were used as the reference com-
pounds for 3D-QSAR models and 3D-QSSR models, respectively.
Models including CoMFA 3D-QSAR models, CoMSIA 3D-QSAR mod-
els, CoMFA 3D-QSSR models, and CoMSIA 3D-QSSR models were
developed and optimal models were chosen according to the sta-
tistical parameter after the partial least square (PLS) analysis. The
q2 (cross-validated correlation coefficient), r2 (correlation coeffi-
cient), see (standard error of estimate), and F-value (r2/1 À r2) were
0.618, 0.995, 0.06, and 501.848 for the chosen CoMFA 3D-QSAR
model, 0.572, 0.943, 0.185, and 80.136 for the chosen CoMSIA
3D-QSAR model, 0.561, 0.995, 0.058, and 273.753 for the chosen
CoMFA 3D-QSSR model, and 0.541, 0.994, 0.068, and 199.531 for
the chosen CoMSIA 3D-QSSR model, respectively, (Table 2). The
q2 values of all the chosen models were more than 0.5 which indi-
cated that these chosen models were reliable and predictive.
The contribution was 0.441 for steric field and 0.559 for electro-
static field in CoMFA 3D-QSAR model, and the contribution was
0.407 for steric field and 0.593 for electrostatic field in CoMFA
3D-QSSR model, so electrostatic field was the main contribution
in CoMFA models. The contributions of steric field, electrostatic
field, H-acceptor, H-donor, and hydrophobic field were 0.116,
0.369, 0.139, 0.186, 0.191 in CoMSIA 3D-QSAR model, and 0.137,
0.167, 0.139, 0.104, and 0.453 in CoMSIA 3D-QSSR model, respec-
tively, (Table 2).
Limk1/ROCK2 selectivity (
12), andbulkygroupsintheR2 regionwerefavoredforLimk1/ROCK2
selectivity pIC50selectivity = 2.205 for 36 with 4-CH3, 2.225
for 33 with 4-Cl, 2.225 for 32 with 3-Cl vs 2.184 for 27, and
pIC50selectivity = 1.414for12with3-CONHCH(CH3)2 vs1.221for18).
DpIC50selectivity = 0.999 for 17 vs 1.414 for
(D
D
2.3.2. Contour maps of electrostatic field
The electrostatic field contour maps were shown in Figure 4.
According to Figure 4A and B, there were both blue contours and
red contours in the R4 region, which suggested that groups with
an appropriate charge would benefit the Limk1 inhibition. For
example, CH2CH2OH and CH2CH2NH2 were both good for Limk1
inhibitory activity (pIC50Limk1 = 43 nM for 27 vs 142 nM for 18,
and 27 nM for 40 vs 35 nM for 22). In Figure 4A, there was a red
contour in the R1 region which suggested that an electronegative
group such as F was favored for Limk1 inhibition. For example,
compound 16 had a lower IC50 than inhibitor 12 (pIC50Limk1 = 18 -
nM for 16 vs 62 nM for 12). According to Figure 4C and D, there
were a red contour around the 4-position of the terminal phenyl
ring (R2 region), which indicated that electronegative groups such
as Cl were favored for selectivity, for example, the selectivity of 33
Limk1 inhibition activity (pIC50Limk1), residual of Limk1 inhibi-
tion activity (
activity were calculated and listed in 3D-QSAR models of Table 3.
Limk1/ROCK2 selectivity pIC50selectivity), residual of Limk1/
DpIC50activity), and residual rate of Limk1 inhibition
(
D
ROCK2 selectivity (DDpIC50selectivity), and residual rate of Limk1/
ROCK2 selectivity were calculated and listed in 3D-QSSR models
of Table 3. In 3D-QSAR models, residuals were between À0.403
and 0.304 and residual rates were between À6.1% and 4.1% for
the training set, and residuals were between À0.657 and 0.831
and residual rates were between À9.9% and 12% for the test set.
In 3D-QSSR models, residuals were between À0.061 and 0.132
and the residual rates were between À5% and 10.6% for the training
set, and residuals were between À0.519 and 0.433 and the residual
rates were between À29% and 35% for the test set. Although the
range of residual rates in the test set of 3D-QSSR model were a lit-
tle bit broader than others (À29% to 35% vs À6.1% to 4.1%, À5% to
10.6% and À9.9% to 12%), all models including 3D-QSAR model and
3D-QSSR model exhibited good correlative and predictive power.
was better than 27 (DpIC50selectivity = 2.241 for 33 vs 2.184 for 27).
2.3.3. Contour maps of hydrophobic and H-bond donor/acceptor
field
Contour maps of hydrophobic and H-bond donor/acceptor field
were shown in Figure 5. As shown in Figure 5A and B, there were
big white contours at the R4 region and big yellow contours at
the 4-position of the terminal phenyl ring, which demonstrated
that hydrophilic group such as –CH2CH2OH in the R4 region and
hydrophobic group such as CH3 in the R2 region were favored for
both the Limk1 inhibitory activity and the Limk1/ROCK2 selectivity
(IC50Limk1 = 43 nM and
DpIC50selectivity = 2.184 for 27 vs 142 nM and
1.221 for 18, and IC50Limk1 = 37 nM and
36 vs 43 nM and 2.184 for 27).
DpIC50selectivity = 2.205 for

7466
J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
Table 1
The structures and experimental IC50 values of urea-based Limk inhibitors
H CO
3
N
NH
NH
N
O
O
H CO
3
N
H
N
H
N
H
N
N
H
a, c
1
2
IC
Limk1(nM)
>10000
2
642
188
50
ROCK2 (nM)
NH
NH
R
Ar
1
S
O
O
O
N
N
N
N
N
R
N
R
N
R
N
H
2
N
N
N
N
H
4
3
H
H
b
a, c
3
4
5-47
IC
50
Limk1(nM)
4507
203
2290
ROCK2 (nM) >10000
Compd
R1
H
R2
R3
H
R4
H
Ar
Limk1
IC50 (nM)
ROCK2
IC50 (nM)
N
5a,c
3-CONHCH(CH3)2
201
1365
N
H
N
HN
N
6
7
H
H
3-CONHCH(CH3)2
3-CONHCH(CH3)2
H
H
H
H
>10,000
>10,000
45
90
N
N
N
8
H
H
H
3-CONHCH(CH3)2
3-CONHCH(CH3)2
3-CONHCH(CH3)2
H
H
H
H
H
H
>10,000
>10,000
>10,000
166
132
247
NH2
9
N
N
H
N
N
10
O
N
H
N
N
N
N
11a,c
12a,c
13b
H
H
H
3-CONHCH(CH3)2
3-CONHCH(CH3)2
3-CONHCH(CH3)2
H
H
H
H
H
H
1527
62
5570
N
H
N
N
1608
N
H
N
80
>10,000
N
H
N
N
N
14a
H
3-CONHCH(CH3)2
H
H
80
>10,000
N
H
N
15a,c
16a,c
17a,c
18a,d
19a,c
20b,c
21a,c
CF3
3-CONHCH(CH3)2
3-CONHCH(CH3)2
3-CONHCH(CH3)2
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
60
976
N
H
N
N
N
N
N
N
N
N
N
N
N
N
N
F
18
781
N
H
–OCH2CH2N(CH3)2
710
142
315
283
75
7083
2358
5421
6652
2572
N
H
H
H
H
H
N
H
3-F
N
H
2-OCH3
N
H
3-OCH3
N
H

J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
7467
Table 1 (continued)
Compd
R1
R2
R3
R4
Ar
Limk1
ROCK2
IC50 (nM)
IC50 (nM)
N
N
22a
23a
H
H
4-OCH3
4-OCH3
H
H
H
35
>10,000
N
H
N
N
CH3
1090
NR
N
H
N
N
N
N
N
N
24a
H
H
H
H
H
4-OCH3
4-OCH3
H
H
H
9240
>10,000
368
NR
N
H
N
N
N
25
NR
N
H
N
26a
H
H
H
NR
N
H
N
N
27b,c
28b,d
H
CH2CH2OH
CH2CH2OH
43
6565
1605
N
H
2-F
132
N
H
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
29a,c
30b,c
31b,c
32b,d
33a,d
34a
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
3-F
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
CH2CH2OH
CH2CH2OH
CH2CH2OH
CH2CH2OH
CH2CH2OH
CH2CH2OH
CH2CH2OH
CH2CH2OH
CH2CH2OH
CH2CH2OH
CH2CH2OH
CH2CH2NH2
CH2CH2NH2
CH2CH2N(CH3)2
CH2CH2N(CH3)2
101
86
1898
3239
3339
11,270
4357
>10,000
8940
5932
>10,000
3219
>10,000
NR
N
H
4-F
N
H
2-Cl
58
N
H
3-Cl
67
N
H
4-Cl
25
N
H
2-CH3
3-CH3
4-CH3
2-OCH3
3-OCH3
4-OCH3
4-OCH3
4-Cl
350
151
37
N
H
35a,d
36b,c
37a
N
H
N
H
913
100
53
N
H
38a,c
39a
N
H
N
H
40a
27
N
H
41a
21
NR
N
H
42a
4-OCH3
4-Cl
47
NR
N
H
43a
20
NR
N
H
(continued on next page)

7468
J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
Table 1 (continued)
Compd
R1
R2
R3
R4
Ar
Limk1
ROCK2
IC50 (nM)
IC50 (nM)
N
N
N
N
44b
45a
46a
F
F
F
F
4-Cl
4-Cl
4-Cl
4-Cl
H
H
H
H
CH2CH2OH
21
21
19
8
NR
NR
NR
NR
N
H
N
N
N
N
CH2CH2NH2
CH2CH2N(CH3)2
CH2CH2OCH3
N
H
N
H
47a
N
H
a
Training set of 3D-QSAR models.
Test set of 3D-QSAR models.
Training set of 3D-QSSR models.
Test set of 3D-QSSR models.
b
c
d
through molecular docking in order to find more valuable informa-
tion for designing potent and selective Limk inhibitors.
Table 2
Statistical parameters of 3D-QSAR and 3D-QSSR models
The structures of compounds 12 and 5 were similar with 12
having one more methyl group at the 3-position of bottom
pyrrolopyrimidine ring. Inhibitor 12 exhibited a 3-fold higher
Limk1 potency. However, it had a similar ROCK2 inhibition
potency to compound 5 (IC50Limk1 = 62 nM for 12 and 201 nM for
5, IC50ROCK2 = 1608 nM for 12 and 1365 nM for 5). Docking 12 and
5 into the binding site of human Limk1 (PDB: 3S95) demonstrated
3D-QSAR model
3D-QSSR model
CoMFA
CoMSIA
CoMFA
CoMSIA
q2
0.618
8
0.572
5
0.561
7
0.541
7
NOCa
r2
SEE
F-value
0.995
0.06
501.848
0.943
0.185
80.136
0.995
0.058
273.753
0.994
0.068
199.531
that there were four H-bonds and one cation–p interaction in both
Contribution
Steric
Electrostatic
H-acceptor
H-donor
3S95-12 and 3S95-5 complexes (Fig. 6). However, there was an
alkyl hydrophobic interaction between the –CH3 group from the
bottom pyrrolopyrimidine ring of inhibitor 12 and residues
Leu467 and Ala353 of Limk1, which is likely the reason that inhi-
bitor 12 had a higher Limk1 inhibitory activity than inhibitor 5.
So 3-methyl pyrrolopyrimidine moiety should be kept as a key
structure unit in further deign of potent and selective Limk
inhibitors.
0.441
0.559
—
0.116
0.369
0.139
0.186
0.191
0.407
0.593
—
0.137
0.167
0.139
0.104
0.453
—
—
—
—
Hydrophobic
a
Optimal number of principal components.
In Figure 5C, the N atom of the urea linker attached to the ter-
Compound 33 had the highest selectivity (DpIC50selectivity = 2.241)
minal phenyl ring was surrounded by two big purple contours
from both the upper and the down sides, which revealed that a
unsubstituted NH was not as good as an alkylated N atom for the
Limk1 inhibitory activity. For example, the IC50Limk1 is 142 nM for
compound 18 while that for inhibitor 27 is 43 nM. In Figure 5D,
cyan contours occupied the R4 region, which showed that groups
containing H-bond donor moieties, such as CH2CH2OH and CH2-
CH2NH2, were favored for the Limk1/ROCK2 selectivity. For exam-
ple, compound 27 exhibited a better inhibitory activity and a
and compound 2 hadthelowest selectivity (
D
pIC50selectivity = À0.534).
They werebothdocked into the ATP-bindingpockets of human Limk1
(PDB: 3S95) and human ROCK2 (PDB: 4L6Q) (Fig. 7). In the 3S95-33
complex, there were three H-bonds between the pyrrolopyrimidine
ring and residues Ile416, Glu414 and Thr413. A fourth H-bond was
formed between the OH group and the carbonyl group of Lys347. In
addition, a cation–p interaction was formed between the terminal
phenyl ring and His464. Both H-bonds interactions and hydrophobic
interactions contributed to the high Limk1 affinity of compound 33
(Fig. 7A). On the other hand, only H-bonds were formed between
compound 2 and residues Ile416 and Glu414 in the 3S95-2 complex
(Fig. 7B). Therefore, the higher Limk1 inhibition activity of compound
33 comparedto2 (IC50Limk1 = 25 nMfor33vs IC50Limk1 = 642 nMfor 2)
islikelydue to the hydrophobicinteractionsfrom the terminal phenyl
ring and the H-bond interaction at the R4 region.
Previous studies demonstrated that both H-bonds and
hydrophobic interactions contributed to the high activity of urea-
based ROCK2 inhibitors and the residues for H-bonds were
Glu170, Met172, Lys121 and Asp176.43,44 As shown in Figure 7C,
only one H-bond (between NH (N1) and Met172) was formed in
the 4L6Q-33 complex, which is likely the reason for the low ROCK2
inhibition of 33 (IC50ROCK2 = 4357 nM). In the 4L6Q-2 complex, four
higher selectivity than compound 18 (DpIC50selectivity = 2.205 for
27 vs 1.221 for 18). In Figure 5E, there were magenta contours
around the R4 region, which demonstrated that groups with a H-
bond acceptor, such as –CH2CH2OCH3 and –CH2CH2N(CH3)2, were
favored for Limk1 inhibition. For example, compounds 46 and 47
were among the most potent Limk inhibitors. In Figure 5F, the R4
region was covered by a big red contour, which indicated that
groups with a H-bond acceptor were disfavored and groups
containing a H-bond donor were favored for the Limk1/ROCK2
selectivity (DpIC50selectivity = 2.205 for 27 vs 1.221 for 18).
2.4. Docking results
Docking protocols are widely used to explore the binding affini-
ties of ligands. We further explored the difference of Limk1 inhibi-
tory activity and Limk1/ROCK2 selectivity at the amino acid level
H-bonds and one cation–
for H-bonds were Glu170, Met172 and Lys121, and the residue
for cation– interaction was Phe384. The distance between the
p interaction were formed. The residues
p

J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
7469
Table 3
Experimental and predicted values
3D-QSAR models
CoMFA
a
Compd
Actual pIC50Limk1
CoMSIA
Residual
Predicted
pIC50Limk1
Residual
pIC50activity
Residual
ratec (%)
Predicted
pIC50Limk1
Residual
a
b
a
b
D
D
pIC50activity
ratec (%)
Training set
2
4
6.192
6.693
6.697
5.816
7.208
7.097
7.097
7.222
7.745
6.149
6.502
6.548
7.125
7.456
5.963
5.034
6.996
7.602
6.456
6.821
7.432
6.040
7.000
7.276
7.569
7.678
7.678
7.678
7.721
8.096
6.215
6.720
6.691
5.773
7.166
7.138
7.087
7.195
7.758
6.152
6.538
6.511
7.108
7.446
5.944
5.079
6.981
7.498
6.475
6.879
7.377
5.980
7.090
7.289
7.547
7.740
7.821
7.625
7.736
8.028
À0.023
À0.027
0.006
0.043
0.042
À0.37
À0.40
0.09
0.73
0.58
6.345
6.662
6.834
5.695
7.129
7.036
7.196
7.258
7.528
6.122
6.905
6.370
6.952
7.498
5.889
5.039
6.892
7.328
6.721
7.017
7.128
6.183
6.898
7.509
7.722
6.854
7.817
7.697
7.743
7.937
À0.153
0.031
À2.5
0.46
À2.0
2.1
5
À0.137
11
12
13
14
15
16
17
19
20
21
22
23
24
29
33
34
35
36
37
38
39
40
41
44
45
46
47
0.121
0.079
0.061
À0.099
À0.036
0.217
1.1
À0.041
À0.57
0.85
À1.4
À0.49
2.8
0.44
À6.1
2.7
0.010
0.14
0.027
0.37
À0.013
À0.003
À0.036
0.037
0.017
0.010
À0.16
À0.05
À0.55
0.56
0.23
0.13
0.027
À0.403
0.178
0.173
2.4
À0.042
À0.56
0.019
0.31
0.074
1.2
À0.045
À0.89
À0.005
0.104
0.274
À0.265
À0.196
0.304
À0.1
1.4
3.6
À4.1
À2.8
4.1
0.015
0.21
0.104
1.36
À0.019
À0.058
0.055
À0.29
À0.85
0.74
0.060
0.99
À0.143
À2.4
À0.090
À0.013
0.022
À0.038
À0.143
0.053
À1.2
À0.17
0.29
À0.49
À1.8
0.69
0.102
1.4
À0.233
À0.153
0.142
À0.139
À0.019
À0.022
0.159
À3.2
À2.0
1.8
À1.8
À0.24
À0.28
2.0
À0.015
À0.19
0.068
0.83
Test set
3
18
26
27
28
30
31
32
5.346
6.848
6.434
7.367
6.879
7.066
7.237
7.173
7.328
7.699
5.831
6.195
6.720
7.231
7.104
7.723
6.628
7.417
7.686
7.358
À0.485
0.653
À9.0
9.5
5.878
6.017
6.811
6.976
6.506
7.328
6.736
7.133
7.339
7.151
À0.532
0.831
À9.9
12
À0.286
À4.4
À0.377
À5.8
0.136
1.8
0.391
5.3
À0.225
À0.657
0.609
À0.244
À0.358
0.341
À3.2
À9.2
8.4
À3.4
À4.8
4.4
0.373
5.4
À0.262
À3.7
0.501
6.9
0.040
0.55
À0.15
7.1
42
43
À0.011
0.548
3D-QSSR models
CoMFA
d
Compd
Actual
D
pIC50selectivity
CoMSIA
Residual
Predicted
Residual
Residual
ratef (%)
Predicted
Residual
ratef (%)
d
e
d
e
D
pIC50selectivity
DDpIC50selectivity
D
pIC50selectivity
DDpIC50selectivity
Training set
2
4
5
À0.534
1.053
0.832
0.562
1.414
1.211
1.638
0.999
1.236
1.371
1.535
2.184
1.274
1.576
1.761
2.205
1.508
À0.524
1.053
0.832
0.549
1.403
1.226
1.646
0.999
1.164
1.363
1.592
2.078
1.33
À0.01
0
0
1.8
0
0
2.3
0.77
À1.2
À0.48
0
0.48
0.58
À3.7
4.8
À0.507
1.034
0.807
0.572
1.415
1.208
1.67
0.995
1.104
1.404
1.542
2.075
1.303
1.637
1.802
2.239
1.525
À0.027
0.019
0.025
5
1.8
3
11
12
15
16
17
19
20
21
27
29
30
31
36
38
0.013
0.011
À0.015
À0.008
0
0.072
0.008
À0.057
0.106
À0.056
À0.034
0.026
À0.067
0.012
À0.01
À1.7
À0.07
0.24
À1.9
0.4
10.6
À2.4
À0.45
4.9
À2.2
À3.8
À2.3
À1.5
À1.1
À0.001
0.003
À0.032
0.004
0.132
À0.033
À0.007
0.109
À0.029
À0.061
À0.041
À0.034
À0.017
À4.3
À2.1
1.5
1.61
1.735
2.272
1.496
À3
0.79
Test set
18
28
1.221
1.084
0.788
1.432
0.433
0.348
35
32
0.919
0.849
0.302
0.235
24
21
(continued on next page)

7470
J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
Table 3 (continued)
3D-QSSR models
CoMFA
d
Compd
Actual
D
pIC50selectivity
CoMSIA
Predicted
pIC50selectivity
Residual
DDpIC50selectivity
Residual
ratef (%)
Predicted
pIC50selectivity
Residual
DDpIC50selectivity
Residual
ratef (%)
d
e
d
e
D
D
32
33
35
2.225
2.241
1.772
2.190
2.193
2.158
0.035
0.048
À0.386
1.6
2.1
À21
1.824
1.992
2.291
0.401
0.249
À0.519
18
11
À29
a
b
c
d
e
f
pIC50Limk1 = Àlog IC50Limk1
Residual
pIC50activity = actual pIC50Limk1 À predicted pIC50Limk1
Residual rate = pIC50activity/actual pIC50Limk1
pIC50selectivity = pIC50Limk1 À pIC50ROCK2
Residual DDpIC50selectivity = actual
pIC50selectivity À predicted
Residual rate = DDpIC50selectivity/actual pIC50selectivity
.
D
.
D
.
D
.
D
D
pIC50selectivity.
D
.
Figure 1. Plots of actual versus predicted values. (A) CoMFA 3D-QSAR model. (B) CoMSIA 3D-QSAR model. (C) CoMFA 3D-QSSR model. (D) CoMSIA 3D-QSSR model.
urea carbonyl group and the terminal phenyl ring of 33 was shorter
than that of 2 (two and three backbone atoms for 33 and 2, respec-
tively), and the structural elements of compounds 33 could not fit
well into the hydrophobic pocket. Therefore, compound 33 exhib-
ited lower ROCK2 potency than 2 (IC50ROCK2 = 4357 nM for 33 vs
IC50ROCK2 = 188 nM for 2). So carbonyl group-aromatic ring system
(CONH-Ar) will be good structure moiety in next inhibitors’ design.
4249 150 KJ/mol from 3.5 ns to 5.0 ns, which showed that the
final structure of the 5.0 ns MDs was stable. Both initial and final
docked inhibitors were in the same binding pocket and their phar-
macophore structures were basically consistent, which demon-
strated that docking results were reliable.
2.6. Designs of urea-based Limk1 inhibitors
2.5. MD simulations results
Optimal computational results should have the following
advantages: the results fit the rule well and it can provide guidance
for inhibitor designs. Therefore, we test our computational results
through synthesis and biological evaluation of newly designed
Limk1 inhibitors. Molecular docking showed that 3-methyl
In order to validate the docking accuracy, we performed 5 ns
MD simulation for 3S95-33 complex (Fig. 8). The total-energy
reduced sharply at the beginning, and basically remained at

J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
7471
Bulky groups were
disfavored for activity.
Bulky groups with appropriated charge were
favored for both activity and selectivity.
R3 R4
N
N
Bulky groups with electronegative
charge here, especially at 4-position
were favored for selectivity.
R2
O
Ar
R1
Bulky groups were disfavored for both
activity and selectivity and electronegative
groups were favored for activity.
Bulky groups at the 5, 6, or 7-position
of bottom pyrrolopyrimidine were
disfavored for selectivity.
Figure 2. Structure–activity/selectivity relationships derived from 3D-QSAR and 3D-QSSR studies.
Figure 3. Counter maps of steric field. (A) CoMFA 3D-QSAR model. (B) CoMSIA 3D-QSAR model. (C) CoMFA 3D-QSSR model. (D) CoMSIA 3D-QSSR model. Bulky groups were
favored in the green regions and disfavored in the yellow regions.
pyrrolopyrimidine moiety (Fig. 6) and carbonyl-aromatic ring sys-
tem (CONH-Ar, Fig. 7) were good structure units and should be
kept in further design of potent and selective Limk inhibitors. Con-
tour map of electrostatic field suggested that electronegative group
such as F, or –CH2CH2OH with appropriate charges in the R4 region
was favored for Limk1 inhibition (Fig. 4). So fragments including 3-
methyl pyrrolopyrimidine at the bottom aromatic region, carbonyl
group-aromatic ring system (CONH-Ar), F atom in the R1 region,
and –CH2CH2OH in the R4 region were introduced in the newly
designs of Limk inhibitors, which were favored for both Limk1
inhibition and the Limk1/ROCK2 selectivity. Compounds (D1–D5)
were therefore designed and synthesized, and the synthetic route
and biochemical data of them were shown in Figure 9 and Table 4,
respectively. Remarkably, the IC50 values obtained from computa-
tional models and biochemical assays were quite similar and
within the allowed error range (less than 10 times) except the
IC50 value of D1 against ROCK2 which was out of the test win-
dow.45 D1 exhibited a better Limk1 inhibitory activity compared
to 22 (IC50Limk1 = 10 nM for D1 vs 35 nM for 22). D5 also exhibited
a better Limk1 potency and an enhanced Limk1/ROCK2 selectivity
compared to inhibitor 18 (IC50Limk1 = 85 nM for D2 vs 142 nM for
18, and 53-fold Limk1/ROCK2 selectivity for D2 vs 16-fold for
18). Changing the tailor phenyl ring to pyridine or pyrimidine also
led to potent and selective Limk inhibitors (IC50Limk1 = 11 nM for
D2, 19 nM for D3, and 55 nM for D4, and IC50ROCK2 = 2114 nM for
D2, 2011 nM for D3, and 1021 nM for D4). So the computational
results obtained in this work were effective for designing new
Limk1 inhibitors.
3. Conclusion
From the data set collected from bis-aryl urea derivatives as
Limk inhibitors, CoMFA 3D-QSAR models, CoMSIA 3D-QSAR mod-
els, CoMFA 3D-QSSR models, and CoMSIA 3D-QSSR models were
built up, and the structure–activity/selectivity relationships were
obtained by analyzing the contour maps of CoMFA and CoMSIA
models. Molecular docking further demonstrated that hydrophobic
interaction with residues Leu467 and Ala353 was one of the key
elements for Limk1 inhibition. A 5 ns MD simulations certified
the reliability of docking results. Finally, new compounds were
designed, synthesized, and biologically evaluated, and these com-
pounds exhibited good Limk inhibitory activity and selectivity.
These results demonstrated that the computational results were
effective for designing highly potent and Limk1/ROCK2 selective

7472
J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
Figure 4. Contour maps of electrostatic field. (A) CoMFA 3D-QSAR model. (B) CoMSIA 3D-QSAR model. (C) CoMFA 3D-QSSR model. (D) CoMSIA 3D-QSSR model.
Electropositive groups were favored in the blue regions, and electronegative groups were favored in the red regions.
Figure 5. Contour maps of hydrophobic and H-bond donor/acceptor field. (A) Hydrophobic field of CoMSIA 3D-QSAR model. (B) Hydrophobic field of CoMSIA 3D-QSSR model.
(C) H-bond donor fields of CoMSIA 3D-QSAR model. (D) H-bond donor fields of CoMSIA 3D-QSSR model. (E) H-bond acceptor fields of CoMSIA 3D-QSAR model. (F) H-bond
acceptor fields of CoMSIA 3D-QSSR model. Hydrophobic groups were favored in the yellow regions and disfavored in the white regions. H-bond donor groups were favored
the cyan regions and disfavored in the purple regions. H-bond acceptor groups were favored in the magenta regions and disfavored in red regions.

J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
7473
Figure 6. Docking results. (A) Docking of 12 into the binding site of Limk1. (B) Docking of 5 into the binding site of Limk1. Hydrogen bonds were shown as green lines, alkyl
hydrophobic interactions were shown as pink lines, and cation– interactions were shown as red lines.
p
Limk inhibitors. The exploration of the F-substitution effects on the
Limk inhibitory activity and selectivity using computational tools
is underway in our labs and will be published in due course.
performed with Thermo Finnigan Orbitrap mass analyzer. Data
were acquired in the positive ion mode at resolving power of
100,000 at m/z 400. Calibration was performed with an external
calibration mixture immediately prior to analysis.
4. Experimental
4.1. Synthetic procedure and structural characterization of
D1–D5
Commercially available reagents and anhydrous solvents were
used without further purification unless otherwise specified. Thin
layer chromatography (TLC) analyses were performed with
precoated silica gel 60 F254. The mass spectra were recorded by
LC/MS with Finnigan LCQ Advantage MAX spectrometer of Thermo
ElectronÒ. Flash chromatography was performed on prepacked col-
umns of silica gel (230–400 Mesh, 40–63
with EtOAc/hexane or MeOH/DCM as eluent. The preparative HPLC
was performed on SunFire C18 OBD 10
CH3CN + 50% MeOH/H2O + 0.1% TFA as eluent to purify the targeted
compounds. Analytic HPLC was performed on Agilent technologies
1200 series with CH3CN (Solvent B)/H2O + 0.9% CH3CN + 0.1% TFA
(Solvent A) as eluent and the targeted products were detected by
UV in the detection range of 215–310 nm. All compounds were
determined to be >95% pure by this method. NMR spectra were
4-Methoxy-phenylamine,
6-methyl-pyridin-3-ylamine,
2-
methyl-pyridin-4-ylamine, pyrimidin-5-ylamine, or 2-pheny-
lamino-ethanol (0.2 mmol) was added to the solution of 4-
bromo-2-fluoro-1-isocyanato-benzene (0.2 mmol) in DCM (1 mL).
The mixture was stirred at room temperature for 2 h. Then, the sol-
vent was removed in vacuo to give the crude bromide A for next
step without further purification. Bis-(pinacolato)diboron
(0.24 mmol), A (0.2 mmol), and PdCl2(dppf) (0.02 mmol) were dis-
solved in degassed dioxane (5 mL). After refluxing for 2 h, the mix-
ture was diluted with water and extracted with ethyl acetate
(3 Â 5 mL). The organic layers were combined, dried over anhy-
drous Na2SO4 and concentrated in vacuo to give crude boronic acid
pinacol ester B. Finally, B (0.3 mmol) and 4-chloro-5-methyl-7H-
pyrrolo[2,3-d]pyrimidine (0.2 mmol) were dissolved in degassed
5:1 dioxane/H2O. Pd(PPh3)4 (0.02 mmol) and 2 M solution of
K2CO3 (0.6 mmol) were added sequentially under argon and the
mixture was heated at 95 °C for 2 h. After cooling to room temper-
ature, the mixture was diluted with water and extracted with ethyl
l
m) by CombiFlashÒ
l
m (30 Â 250 mm) with
recorded with
a
BrukerÒ 400 MHz spectrometer at ambient
temperature with the residual solvent peaks as internal standards.
The line positions of multiplets were given in ppm (d) and the
coupling constants (J) were given in Hertz. The high-resolution
mass spectra (HRMS, electrospray ionization) experiments were

7474
J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
Figure 7. Docking results. (A) Docking of compound 33 into the binding site of Limk1. (B) Docking of compound 2 into the binding site of Limk1. (C) Docking of compound 33
into the binding site of ROCK2. (D) Docking of compound 2 into the binding site of ROCK2. Hydrogen bonds were shown as green lines, alkyl hydrophobic interactions were
shown as pink lines, and cation–p interactions were shown as red lines.
Figure 8. MDs results. (A) Plot of total energy versus time. (B) 3S95-33 complex. The initial structure represent in magenta and the final structure represent in green.
NCO
F
F
Ar
N
NH
N
H
N
H
N
Ar
F
R4
O
F
b
c
a
O
N
N
R4
B
Ar
R4
HN
O
O
O
Br
N
Br
A
B
D1-D5
N
N
R4 = H or CH2CH2OH
Ar =
OCH3
N
N
Figure 9. Synthesis of newly designed Limk inhibitors. Reagents and conditions: (a) Ar-NHR4, DCM, 0 °C; (b) bis(pinacolato)diboron, PdCl2dppf, dioxane, reflux; (c) 4-chloro-
5-methyl-7H-pyrrolo[2,3-d]pyrimidine, Pd(PPh3)4, dioxane/H2O, 95 °C.

J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
7475
Table 4
Biochemical data of D1–D5
Compd
Experimental IC50 (nM)
Pred. IC50 (nM) by CoMFA
Pred. IC50 (nM) by CoMSIA
Limk1
ROCK2
Limk1
ROCK2
Limk1
ROCK2
NH
O
N
F
O
10
11
>20,000
65
2187
98
1840
5572
N
N
N
H
H
D1
NH
N
F
O
N
2114
17
2152
45
N
H
N
N
H
D2
NH
N
F
O
N
19
55
85
2011
1021
4508
53
57
933
276
82
479
181
3793
893
N
H
N
N
H
D3
NH
N
F
N
O
N
N
N
N
H
H
D4
NH
N
F
O
N
111
3140
2333
HN
D5
N
HO
acetate (3 Â 5 mL). The organic layers were combined, dried over
anhydrous Na2SO4 and concentrated in vacuo. The residue was
then purified by preparative HPLC to give the targeted product
D1–D5 as white solids.
(m, 3H), 8.01–8.00 (m, 2H), 7.70–7.68 (m, 1H), 7.48–7.46 (m, 1H),
7.39–7.37 (m, 1H), 2.34 (s, 3H), 2.09 (s, 3H); 13C NMR (DMSO-d6,
100 MHz)
d 158.40, 158.06, 154.37, 153.84, 152.08, 145.75,
143.02, 140.61, 136.21, 131.13, 130.43, 129.30, 128.47, 127.64,
127.09, 118.82, 114.50, 111.62, 17.31, 12.51; LC/MS (M+H+):
377.14; HRMS (ESI-Orbitrap) Calcd for C20H18FN6O: 377.1526 [M
+H+], Found 377.1509.
4.1.1. 1-[2-Fluoro-4-(5-methyl-7H-pyrrolo[2,3-d]pyrimidin-4-
yl)-phenyl]-3-(4-methoxy-phenyl)-urea (D1)
50% yield in 3 steps. Purity >99.9% (detected by UV at 215 nm,
230 nm, 254 nm, 280 nm, and 310 nm). 1H NMR (DMSO-d6,
400 MHz) d 12.01 (s, br, 1H), 8.99 (s, 1H), 8.75 (s, 1H), 8.69–8.68
(m, 1H), 8.39–8.37 (m, 1H), 7.60–7.56 (m, 1H), 7.54–7.50 (m,
1H), 7.49–7.46 (m, 1H), 7.40–7.38 (m, 3H), 6.91–6.89 (m, 2H),
3.73 (s, 3H), 2.09 (s, 3H); 13C NMR (DMSO-d6, 100 MHz) d 154.70,
152.39, 152.15, 149.78, 148.14, 132.25, 129.73, 129.63, 128.95,
127.00, 126.48, 119.95, 119.20, 116.14, 115.93, 114.81, 114.05,
109.98, 55.14, 12.73; LC/MS (M+H+): 392.10; HRMS (ESI-Orbitrap)
Calcd for C21H18FN5O2: 392.1523 [M+H+], Found 392.1546.
4.1.4. 1-[2-Fluoro-4-(5-methyl-7H-pyrrolo[2,3-d]pyrimidin-4-
yl)-phenyl]-3-pyrimidin-5-yl-urea (D4)
51% yield in 3 steps. Purity >99.9% (detected by UV at 215 nm,
230 nm, 254 nm, 280 nm, and 310 nm). 1H NMR (DMSO-d6,
400 MHz) d 12.56 (s, br, 1H), 9.18 (s, 1H), 8.94–8.19 (m, 2H),
7.78–7.48 (m, 6H), 7.32–7.28 (m, 2H), 7.01–6.98 (m, 1H), 2.10 (s,
3H); 13C NMR (DMSO-d6, 100 MHz) d 154.43, 152.14, 146.62,
142.62, 142.54, 130.30, 129.37, 129.09, 127.78, 127.66, 127.59,
126.52, 119.18, 118.67, 114.56, 111.08, 12.61; LC/MS (M+H+):
364.17; HRMS (ESI-Orbitrap) Calcd for C18H15FN7O: 364.1322 [M
+H+], Found 364.1341.
4.1.2. 1-[2-Fluoro-4-(5-methyl-7H-pyrrolo[2,3-d]pyrimidin-4-
yl)-phenyl]-3-(6-methyl-pyridin-3-yl)-urea; compound with
methane (D2)
4.1.5. 3-[2-Fluoro-4-(5-methyl-7H-pyrrolo[2,3-d]pyrimidin-4-
yl)-phenyl]-1-(2-hydroxy-ethyl)-1-phenyl-urea (D5)
48% yield in 3 steps. Purity >99.9% (detected by UV at 215 nm,
230 nm, 254 nm, 280 nm, and 310 nm). 1H NMR (DMSO-d6,
400 MHz) d 12.20 (s, br, 1H), 9.41 (s, 1H), 8.90–8.80 (m, 2H),
8.41–8.40 (m, 1H), 8.23–8.21 (m, 1H), 7.89 (s, 1H), 7.71–7.65 (m,
2H), 7.48–7.5 (m, 1H), 2.34 (s, 3H), 2.07 (s, 3H); 13C NMR
(DMSO-d6, 100 MHz) d 158.11, 154.38, 153.97, 152.09, 145.87,
142.96, 142.34, 138.77, 130.42, 129.16, 128.37, 128.20, 127.32,
125.31, 124.73, 118.69, 114.50, 111.55, 20.95, 12.53; LC/MS (M
+H+): 377.13; HRMS (ESI-Orbitrap) Calcd for C20H18FN6O:
377.1526 [M+H+], Found 377.1515.
45% yield in 3 steps. Purity >99.9% (detected by UV at 215 nm,
230 nm, 254 nm, 280 nm, and 310 nm). 1H NMR (DMSO-d6,
400 MHz) d 12.59 (s, br, 1H), 8.89 (s, 1H), 8.75 (s, 1H), 7.66–7.50
(m, 6H), 7.42–7.26 (m, 3H), 3.72 (t, J = 6.4 Hz, 2H), 3.57 (d,
J = 6.4 Hz, 2H), 2.07 (s, 3H); 13C NMR (DMSO-d6, 100 MHz) d
159.17, 128.37, 128.03, 156.71, 154.39, 153.62, 152.07, 145.55,
143.00, 130.39, 129.81, 128.90, 125.09, 119.00, 116.45, 114.51,
111.76, 59.33, 51.98, 12.47; LC/MS(M+H+): 416.17;HRMS (ESI-Orbi-
trap) Calcd for C22H21FN5O2: 416.1679 [M+H+], Found 416.1692.
4.1.3. 1-[2-Fluoro-4-(5-methyl-7H-pyrrolo[2,3-d]pyrimidin-4-
yl)-phenyl]-3-(2-methyl-pyridin-4-yl)-urea (D3)
4.2. Limk1 assay
47% yield in 3 steps. Purity >99.9% (detected by UV at 215 nm,
230 nm, 254 nm, 280 nm, and 310 nm). 1H NMR (DMSO-d6,
400 MHz) d 12.19 (s, br, 1H), 9.71 (s, 1H), 8.83 (s, 1H), 8.29–8.19
Limk1 biochemical assay was carried out in Reaction Biology
Corporation and followed the protocols described on its Web
site. Newly designed compounds (D1–D5) and control compound

7476
J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
Figure 10. The alignments of training set molecules. (A) The alignment of 3D-QSAR models with 47 as the template compound. (B) The alignment of 3D-QSSR models with 36
as the template compound. (C) Common substructures were marked in red.
staurosporine were tested in 10-dose IC50 mode with 3-fold series
dilution starting at 10 M with IC50 measurements. Reactions were
carried out at 10 M ATP, 1 M substrate cofilin, and 50 nM Limk1
atoms were added. Then, the end residues were repaired and pro-
tein was performed energy minimization. In the docking process,
the protein was fixed while the ligands were flexible, the ligands
were docked into the ATP-binding site of Limk1 or ROCK2 by a
patented module in Surflex-Dock and the docked complex was
evaluated by empirical scoring function. By default, 30 conforma-
tions were generated for each ligand and their scores were fore-
casted based on the strength of receptor–ligand interactions. The
docked conformation with best score was selected for further
studies.
l
l
l
(final concentration).
4.3. ROCK2 assay
Assays were performed using the STK2 kinase system from Cis-
bio. A 5 L mixture of a 1 M STK2 substrate and ATP (ROCK-II:
20 M) in STK-buffer was added to the wells of the 384-well plate
l
l
l
using a BioRAPTR FRD Workstation (Aurora Discovery, Carlsbad,
CA). Twenty nanoliters of test compound was then dispensed using
a 384-head offline Pintool system (GNF Systems, San Diego, CA).
4.7. MD simulations
The reaction was started by adding either 5
in STK-buffer. After 4 h at rt the reaction was stopped by adding
10
1 h incubation at rt, the plates were read on the Viewlux in HTRF
mode.
l
L of 0.5 nM ROCK-II
The MD simulations were carried out in Sybyl 2.0 software. The
conformation with best score was picked as initial conformation.
Constant temperature (300 K) and volume were maintained with
the time constant for a heat bath coupling of 100 fs. The time step
of 1 fs was used to integrate the equations of motion, and the snap-
shot time was 100 fs and MD target was limited in the range of 3 Å
with ligand as the core. The Boltzmann initial velocity was used to
start the simulation. Other parameters were set by default in Sybyl
2.0.
l
L of 1Â antibody and 62.5 nM Sa-XL in Detection Buffer. After
4.4. Ligand preparation
The ligand preparation was performed using SYBYL 2.0. Energy
minimization of all the compounds was processed using Tripos
molecular mechanics force field with energy gradient criterion of
0.005 k cal/(Å mol), and the charges of Tripos force field were cal-
culated by the Gasteiger method. Then the training set compounds
were aligned on the template molecules by the alignment com-
mand in SYBYL. The optimal alignment for 3D-QSAR models and
3D-QSSR models were shown in Figure 10 and the common sub-
structures were marked in red.
Acknowledgments
The work was supported by National Natural Science Founda-
tion of China (Grant No. 21502117). Support from Prof. Gang Zhao
and Prof. Guanjun Wang was also greatly appreciated.
References and notes
1. Manetti, F. Drug Discovery Today 2012, 17, 81.
4.5. CoMFA and CoMSIA models
2. Scott, R. W.; Olson, M. F. J. Mol. Med. 2007, 85, 555.
3. Goodwin, N. C.; Cianchetta, G.; Burgoon, H. A.; Healy, J.; Mabon, R.; Strobel, E.
D.; Allen, J.; Wang, S. L.; Hamman, B. D.; Rawlins, D. B. ACS Med. Chem. Lett.
2015, 6, 53.
4. Manetti, F. Med. Res. Rev. 2012, 32, 968.
5. Nadella, K. S.; Saji, M.; Jacob, N. K.; Pavel, E.; Ringel, M. D.; Kirschner, L. S. EMBO
Rep. 2009, 10, 599.
6. Goyal, P.; Pandey, D.; Siess, W. J. Biol. Chem. 2006, 281, 25223.
7. Arber, S.; Barbayannis, F. A.; Hanser, H.; Schneider, C.; Stanyon, C. A.; Bernard,
O.; Caroni, P. Nature 1998, 393, 805.
8. Misra, U. K.; Deedwania, R.; Pizzo, S. V. J. Biol. Chem. 2005, 280, 26278.
9. Efrat, M. F.; Roni, R.; Galit, E. S.; Roni, H.; Shmuel, C.; Yoel, K.; Wolfson, H. J.
Oncotarget 2012, 3, 629.
The optimal chemical conformations after ligand preparation
were aligned and used directly to build CoMFA and CoMSIA mod-
els. 3D contour maps of steric and electrostatic fields in CoMFA
models and five fields including steric, electrostatic, hydrophobic,
hydrogen bond donor, and hydrogen bond acceptor in CoMSIA
models were graphed using the Sybyl 2.0 program (Tripos, Inc.,
USA). The regression analysis was carried out using the PLS method
and the final models were selected according to the statistical
parameters.
10. Jiang, X. S.; Wassif, C. A.; Backlund, P. S.; Song, L.; Holtzclaw, L. A.; Li, Z.; Yergey,
A. L.; Porter, F. D. Hum. Mol. Gen. 2010, 19, 1347.
11. Sun, L.; Guo, C. N.; Wang, T.; Li, X.; Li, G. J.; Luo, Y. L.; Xiao, S. F. J. Alzheimers Dis.
2014, 42, 543.
4.6. Molecular docking
12. Saal, K. Ann.; Koch, J. C.; Tatenhorst, L.; Szego, E. M.; Ribas, V. T.; Michel, U.;
Baehr, M.; Toenges, L.; Lingor, P. Neurobiol. Dis. 2015, 73, 150.
13. Matsumoto, N.; Kitani, R.; Kalinec, F. Commun. Integr. Biol. 2011, 4, 208.
14. Honma, M.; Benitah, S. A.; Watt, F. M. Mol. Biol. Cell. 2006, 17, 1888.
15. Zhou, Y.; Yuge, A.; Rajah, A. M.; Unek, G.; Rinaudo, P. F.; Maltepe, E. Am. J.
Pathol. 2014, 184, 3321.
The crystal structures of human Limk1 and human ROCK2 for
molecular docking were downloaded from RCSB Protein Data Bank
(http:/www.rcsb.org/pdb/home/home.do, PDB: 3S95 for Limk1 and
4L6Q for ROCK2). All water molecules were removed and hydrogen

J. Cui et al. / Bioorg. Med. Chem. 23 (2015) 7464–7477
7477
16. Low, S. K.; Zembutsu, H.; Takahashi, A.; Kamatani, N.; Cha, P. C.; Hosono, N.;
Kubo, M.; Matsuda, K.; Nakamura, Y. J. Hum. Genet. 2011, 56, 211.
17. Harrison, B. A.; Whitlock, N. A.; Voronkov, M. V.; Almstead, Z. Y.; Gu, K. J.;
Mabon, R.; Gardyan, M.; Hamman, B. D.; Allen, J.; Gopinathan, S.; McKnight, B.;
Crist, M.; Zhang, Y.; Liu, Y.; Courtney, L. F.; Key, B.; Zhou, J.; Patel, N.; Yates, P.
W.; Liu, Q.; Wilson, A. G.; Kimball, S. D.; Crosson, C. E.; Rice, D. S.; Rawlins, D. B.
J. Med. Chem. 2009, 52, 6515.
18. Xu, X. H.; Guo, J.; Vorster, P.; Wu, Y. Commun. Integr. Biol. 2012, 5, 381.
19. Vorster, P. J.; Guo, J.; Yoder, A.; Wang, W.; Zheng, Y.; Xu, X.; Yu, D.; Spear, M.;
Wu, Y. J. Biol. Chem. 2011, 286, 12554.
20. Wen, X.; Ding, L.; Wang, J. J.; Qi, M.; Hammonds, J.; Chu, H.; Chen, X.; Hunter,
E.; Spearman, P. J. Virol. 2014, 88, 6906.
21. Chen, P. X.; Zeng, M. J.; Zhao, Y.; Fang, X. L. Oncol. Rep. 2014, 32, 2070.
22. Xu, M.; Chen, G.; Wang, S.; Liao, M.; Frank, J. A.; Bower, K. A.; Zhang, Z.; Shi, X.;
Luo, J. PLoS One 2012, 7, e47721.
23. Xia, H.; Sun, S. J.; Wang, B.; Wang, T.; Liang, C. Y.; Li, G.; Huang, C. B.; Qi, D. L.;
Chu, X. Y. Int. J. Mol. Sci. 2014, 15, 11973.
24. Moretti, R. M.; Mai, S.; Marelli, M. M.; Rizzi, F.; Bettuzzi, S.; Limonta, P. Int. J.
Oncol. 2011, 39, 225.
25. Scott, R. W.; Hooper, S.; Crighton, D.; Li, A.; Konig, I.; Munro, J.; Trivier, E.;
Wickman, G.; Morin, P.; Croft, D. R.; Dawson, J.; Machesky, L.; Anderson, K. I.;
Sahai, E. A.; Olson, M. F. J. Cell Biol. 2010, 191, 169.
26. Morin, P.; Wickman, G.; Munro, J.; Inman, G. J.; Olson, M. F. Eur. J. Cell Biol.
2011, 90, 13.
32. Ross-Macdonald, P.; de Silva, H.; Guo, Q.; Xiao, H.; Hung, C. Y.; Penhallow, B.;
Markwalder, J.; He, L. Q.; Attar, R. M.; Lin, T. A.; Seitz, S.; Tilford, C.; Wardwell-
Swanson, J.; Jackson, D. Mol. Cancer Ther. 2008, 7, 3490.
33. Mashiach-Farkash, E.; Rak, R.; Elad-Sfadia, G.; Haklai, R.; Carmeli, S.; Kloog, Y.;
Wolfson, H. J. Oncotarget 2012, 3, 629.
34. He, L. Q.; Seitz, S. P.; Trainor, G. L.; Tortolani, D.; Vaccaro, W.; Poss, M.; Tarby, C.
M.; Tokarski, J. S.; Penhallow, B.; Hung, C. Y.; Attar, R.; Lin, T. A. Bioorg. Med.
Chem. Lett. 2012, 22, 5995.
35. Harrison, B. A.; Whitlock, N. A.; Voronkov, M. V.; Almstead, Z. Y.; Gu, K. J.;
Mabon, R.; Gardyan, M.; Hamman, B. D.; Allen, J.; Gopinathan, S.; McKnight, B.;
Crist, M.; Zhang, Y. L.; Liu, Y.; Courtney, L. F.; Key, B.; Zhou, J. L.; Patel, N.; Yates,
P. W.; Liu, Q. Y.; Wilson, A. G. F.; Kimball, S. D.; Crosson, C. E.; Rice, D. S.;
Rawlins, D. B. J. Med. Chem. 2009, 52, 6515.
36. Harrison, B. A.; Almstead, Z. Y.; Burgoon, H.; Gardyan, M.; Goodwin, N. C.;
Healy, J.; Liu, Y.; Mabon, R.; Marinelli, B.; Samala, L.; Zhang, Y. L.; Stouch, T. R.;
Whitlock, N. A.; Gopinathan, S.; McKnight, B.; Wang, S. L.; Patel, N.; Wilson, A.
G. E.; Hamman, B. D.; Rice, D. S.; Rawlins, D. B. ACS Med. Chem. Lett. 2015, 6, 84.
37. Sun, J. R.; Lan, P.; Sun, P. H.; Chen, W. M. Lett. Drug Des. Discovery 2011, 8, 229.
38. Shen, M. Y.; Zhou, S. Y.; Li, Y. Y.; Li, D.; Hou, T. J. Mol. BioSyst. 2013, 9, 2435.
39. Mezna, M.; Wong, A. C.; Ainger, M.; Scott, R. W.; Hammonds, T.; Olson, M. F. J.
Biomol. Screening 2012, 17, 460.
40. Yin, Y.; Zheng, K.; Eid, N.; Howard, S.; Jeong, J.; Yi, F.; Guo, J.; Park, C.; Bibian,
M.; Wu, W.; Hernandez, P.; Wu, Y.; Luo, J. L.; LoGrasso, P. V.; Feng, Y. J. Med.
Chem. 2015, 58, 1846.
27. Vlecken, D. H.; Bagowski, C. P. Zebrafish 2009, 6, 433.
41. Ding, M.; Yin, Y.; Wu, F. H.; Cui, J. X.; Zhou, H.; Sun, G. F.; Jiang, Y. Bioorg. Med.
Chem. 2015, 23, 2505.
42. Sun, N. Y.; Lu, T.; Chen, Y. D.; Hao, L. H.; Xu, Y.; Li, R. J. Acta Phys. Chim. Sin. 2009,
25, 645.
43. Yin, Y.; Cameron, M. D.; Lin, L.; Khan, S.; Schroer, T.; Grant, W.; Pocas, J.; Chen, Y. T.;
Schurer, S.; Pachori, A.; LoGrasso, P.; Feng, Y. B. ACS Med. Chem. Lett. 2010, 1, 175.
44. Yin, Y.; Lin, L.; Ruiz, C.; Khan, S.; Cameron, M. D.; Grant, W.; Pocas, J.; Eid, N.;
Park, H.; Schroter, T.; LoGrasso, P.; Feng, Y. B. J. Med. Chem. 2013, 56, 3568.
45. Patel, P. D.; Patel, M. R.; Kaushik-Basu, N.; Talele, T. T. J. Chem. Inf. Model. 2008,
48, 42.
28. Sleebs, B. E.; Levit, A.; Street, I. P.; Falk, H.; Hammonds, T.; Wong, A. C.; Charles,
M. D.; Olson, M. F.; Baell, J. B. Med. Chem. Commun. 2011, 2, 977.
29. Sleebs, B. E.; Ganame, D.; Levit, A.; Street, I. P.; Gregg, A.; Falk, H.; Baell, J. B.
Med. Chem. Commun. 2011, 2, 982.
30. Sleebs, B. E.; Nikolakopoulos, G.; Street, I. P.; Falk, H.; Baell, J. B. Bioorg. Med.
Chem. Lett. 2011, 21, 5992.
31. Boland, S.; Bourin, A.; Alen, J.; Geraets, J.; Schroeders, P.; Castermans, K.; Kindt,
N.; Boumans, N.; Panitti, L.; Vanormelingen, J.; Fransen, S.; Velde, S. V.; Defert,
O. Bioorg. Med. Chem. Lett. 2015, 25, 4005.