Structure-Based Design of 6-Chloro-4-aminoquinazoline-2-carboxamide Derivatives as Potent and Selective p21-Activated Kinase 4 (PAK4) Inhibitors

Article abstract of DOI:10.1021/acs.jmedchem.7b01342

Herein, we report the discovery and characterization of a novel class of PAK4 inhibitors with a quinazoline scaffold. Based on the shape and chemical composition of the ATP-binding pocket of PAKs, we chose a 2,4-diaminoquinazoline series of inhibitors as a starting point. Guided by X-ray crystallography and a structure-based drug design (SBDD) approach, a series of novel 4-aminoquinazoline-2-carboxamide PAK4 inhibitors were designed and synthesized. The inhibitors' selectivity, therapeutic potency, and pharmaceutical properties were optimized. One of the best compounds, 31 (CZh226), showed remarkable PAK4 selectivity (346-fold vs PAK1) and favorable kinase selectivity profile. Moreover, this compound potently inhibited the migration and invasion of A549 tumor cells by regulating the PAK4-directed downstream signaling pathways in vitro. Taken together, these data support the further development of 31 as a lead compound for PAK4-targeted anticancer drug discovery and as a valuable research probe for the further biological investigation of group II PAKs.

Full text of DOI:10.1021/acs.jmedchem.7b01342

Subscriber access provided by Gothenburg University Library
Article
Structure-based design of 6-chloro-4-aminoquinazoline-2-carboxamide
derivatives as potent and selective p21-activated kinase 4 (PAK4) inhibitors
Chenzhou Hao, Fan Zhao, Hong Yan SONG, Jing Guo, Xiaodong Li, Xiaolin Jiang, Ran Huan,
Shuai Song, Qiaoling Zhang, Ruifeng Wang, Kai Wang, Yu Pang, Tongchao Liu, Tianqi Lu, Wanxu
Huang, Jian Wang, Bin Lin, Zhonggui He, Haitao Li, Feng Li, Dongmei Zhao, and Maosheng Cheng
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.7b01342 • Publication Date (Web): 30 Nov 2017
Downloaded from http://pubs.acs.org on November 30, 2017
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
He, Zhonggui; Shenyang Pharmaceutical University, School of Pharmacy
Li, Haitao; Tsinghua University, Tsinghua-Peking Joint Center for Life
Sciences, Beijing Advanced Innovation Center for Structural Biology,
Department of Basic Medical Sciences, School of Medicine
Li, Feng; China Medical University, Department of Cell Biology, Key
Laboratory of Cell Biology, Ministry of Public Health, and Key Laboratory of
Medical Cell Biology, Ministry of Education
Zhao, Dongmei; Shenyang Pharmaceutical University, Key Laboratory of
Structure-Based Drug Design & Discovery of Ministry of Education
Cheng, Maosheng; Shenyang Pharmaceutical University, Key Laboratory of
Structure-Based Drug Design & Discovery of Ministry of Education
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Structureꢀbased design of
6
ꢀchloroꢀ4ꢀaminoquinazolineꢀ2ꢀcarboxamide
derivatives as potent and selective
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
p21ꢀactivated kinase 4 (PAK4) inhibitors
†
,§
∥
,§
⊥
,§
†
‡
Chenzhou Hao , Fan Zhao , Hongyan Song , Jing Guo , Xiaodong Li , Xiaolin
†
‡
†
†
†
†
Jiang , Ran Huan , Shuai Song , Qiaoling Zhang , Ruifeng Wang , Kai Wang , Yu
†
†
‡
†
†
†
Pang , Tongchao Liu , Tianqi Lu , Wanxu Huang , Jian Wang , Bin Lin , Zhonggui
#
,
∥
,‡
,†
,†
He , Haitao Li* , Feng Li* , Dongmei Zhao* , Maosheng Cheng*
†
Key Laboratory of StructureꢀBased Drug Design & Discovery of Ministry of
Education, Shenyang Pharmaceutical University, Shenyang 110016, China
‡
Department of Cell Biology, Key Laboratory of Cell Biology, Ministry of Public
Health, and Key Laboratory of Medical Cell Biology, Ministry of Education, China
Medical University, Shenyang 110001, China
∥
TsinghuaꢀPeking Joint Center for Life Sciences, Beijing Advanced Innovation
Center for Structural Biology, Department of Basic Medical Sciences, School of
Medicine, Tsinghua University, Beijing 100084, China
⊥Institute of Materials Research and Engineering, A*STAR (Agency for Science,
Technology and Research), 2 Fusionopolis Way, Innovis, #08ꢀ03, Singapore 138634
#
School of Pharmacy, Shenyang Pharmaceutical University, Shenyang 110016, China
1
ACS Paragon Plus Environment

Page 3 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
ABSTRACT
Herein, we report the discovery and characterization of a novel class of PAK4
inhibitors with a quinazoline scaffold. Based on the shape and chemical composition
of the ATPꢀbinding pocket of PAKs, we chose a 2, 4ꢀdiaminoquinazoline series of
inhibitors as a starting point. Guided by Xꢀray crystallography and a structureꢀbased
drug design (SBDD) approach, a series of novel 4ꢀaminoquinazolineꢀ2ꢀcarboxamide
PAK4 inhibitors were designed and synthesized. The inhibitors’ selectivity,
therapeutic potency, and pharmaceutical properties were optimized. One of the best
compounds, 31 (CZh226), showed remarkable PAK4 selectivity (346ꢀfold vs PAK1)
and favorable kinase selectivity profile. Moreover, this compound potently inhibited
the migration and invasion of A549 tumor cells by regulating the PAK4ꢀdirected
downstream signaling pathways in vitro. Taken together, these data support the
further development of 31 as a lead compound for PAK4ꢀtargeted anticancer drug
discovery and as a valuable research probe for the further biological investigation of
group II PAKs.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
INTRODUCTION
The p21ꢀactivated kinases (PAKs) are serine/threonine (Ser/Thr) protein kinases that
1, 2
have been identified as downstream signaling effectors of Rhoꢀfamily GTPases.
The six mammalian PAK isoforms are categorized into two groups: PAK1ꢀ3 (group I)
and PAK4ꢀ6 (group II), based on their structural homologies and biochemical
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
3
features. As key components of the RasꢀRac/Cdc42ꢀPAK pathway, PAKs have
pivotal roles in many fundamental cellular processes, including cytoskeletal
reorganization, focal adhesion, cell motility, morphological changes, cellꢀcycle
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
progression, etc. Moreover, the overexpression, amplification and mutational
activation of PAK isoforms, in particular, PAK1 and PAK4, have been linked to many
human diseases, including breast cancer, lung cancer, prostate cancer, colon cancer,
4
and human head and neck squamous cell carcinoma. Consequently, PAKs have
emerged as attractive targets for new anticancer therapies and have been the subject of
5
extensive drug discovery efforts.
Although the two groups of PAK proteins are similar in overall sequence and
6
structure, they are differentiated by their tissue expression profiles, subcellular
7
7
8
localization, GTPase specificity, activation mechanism, and downstream substrate
2
, 7
specificity.
Studies using knockout mice lacking one or more specific PAK
isoforms revealed the role of each isoform in normal tissue development, with
2
, 9, 10
phenotypes that range from no apparent effect to early embryonic death.
Among
all of the PAKs, PAK4 is the most studied group II PAK member, and it has a place at
critical nodal points in multiple signaling pathways that are associated with cell
1
1
growth, cytoskeletal dynamics, cell polarity, survival, and development. PAK4 is
1
2
particularly highly expressed in prostate, testis, lung, heart, brain and liver. It has
attracted considerable interest because of its role in cancer invasion, metastasis and
13
proliferation of BRAFꢀ or KRASꢀdriven cancers. In addition to PAK4, there is
14
emerging evidence for the roles for PAK5 and PAK6 in cancer progression.
3
ACS Paragon Plus Environment

Page 5 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Moreover, a recent study revealed that PAK2 inhibition correlates with increased
1
5
acute cardiovascular toxicity, which may be enhanced by PAK1 inhibition. Thus,
the development of specific and potent PAK4 inhibitors is highly desirable for
minimizing the risk of the potential side effects associated with the inhibition of
normal function of group I PAKs and will also shed further light on its role in cancer
progression.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5,
Figure 1. Classification of representative PAK inhibitors by mechanism of function.
6
Over the past two decades, several PAK inhibitors have been developed that exhibit
5
, 6
different levels of selectivity across the PAK isoforms. On the basis of their binding
1
6
modes, PAK inhibitors can be categorized into several types (Figure 1). PFꢀ3758309
17
(
2), a “PanꢀPAK inhibitor”, was the first compound advanced to phase I trials
(NCT00932126) by Pfizer, targeting both group I and II PAKs. However, further
development of 2 was called to a stop because of unsatisfactory pharmacokinetic data
18
19
20
in phase I testing. Compound 4 and 5 (GNEꢀ2861) , a type I 1/2 kinase inhibitor,
show significant group I and group II PAK selectivity, respectively. Most recently, a
2
1
new dual PAK4/NAMPT modulator 8 (KPTꢀ9274) advanced to phase I human
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
clinical trials (NCT02702492) for the treatment of solid malignancies or
nonꢀHodgkin’s Lymphoma (NHL).
Our research group has focused on the discovery and optimization of PAK4
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
22ꢀ24
inhibitors as new therapeutic agents for over a decade,
and we identified
quinazolineꢀbased compound 9 (LCHꢀ7749944) as a moderate PAK4 inhibitor via
22
structureꢀbased virtual screening in 2012. Compound 9 was notable in that it
possessed higher potency against PAK4 than PAK1 and significantly inhibited the
migration and invasion of human gastric cancer cells in conjunction with the
concomitant
blockage
of
the
PAK4/LIMK1/cofilin
and
PAK4/MEKꢀ1/ERK1/2/MMP2 pathways. Herein, we describe our structureꢀbased
design efforts toward the discovery of potent group II PAK inhibitors specifically
targeting PAK4 and the structureꢀactivity relationship for this series. Our goal for
optimization was to improve the group II inhibitory activity of early lead 10a while
maintaining high selectivity over group I PAKs (PAK1 and PAK4 as representatives
for group I and group II PAKs, respectively). The binding mode revealed in the Xꢀray
crystallographic studies of 10a, combined with analysis of the residue differences
among PAK isoforms, shed light on some of the selectivity determinants within the
ATP binding site. The following optimization led to the discovery of compound 31, a
potent PAK4 inhibitor with substantially lower inhibition for group I PAKs (Figure
2).
5
ACS Paragon Plus Environment

Page 7 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 2. The design and modification strategies of novel and selective PAK4
inhibitors.
RESULTS AND DISCUSSION
Identification and Characterization of Early Lead 10a. Inspired by our previous
22
study of the PAK4 inhibitor 9, we identified compound 10a as a moderately potent
i
biochemical inhibitor of PAK4 (PAK4 K = 0.710 ꢀM) without measurable binding to
PAK1 (PAK1 K > 10 ꢀM) from a collection of quinazoline derivatives. For an
i
independent confirmation of activity and to address potential secondary effects in the
biochemical assay, such as PAK4 conformation change induced by peptide substrate
effect, or fluorescent false signal, two direct binding assays were used to identify the
binding between 10a and the truncated PAK4. Differential scanning fluorimetry
25
(
DSF) showed a melting temperature shift of 0.5 °C for the compound 10aꢀPAK4
complex compared to the truncated PAK4 protein (Supporting Information Table S4).
26
Surface plasmon resonance (SPR) assay resolved the dissociation constant of 10a
with PAK4 to be 0.156 ꢀM, by multiꢀconcentration kinetic analysis and fitted by 1:1
binding mode. The results demonstrate that compound 10a binds directly to the
truncated PAK4 kinase domain.
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
i i
To gain insight into the origin of the PAK4 selectivity (PAK1 K /PAK4 K )
displayed by compound 10a (> 14ꢀfold), the crystal structure of PAK4 kinase domain
in complex with compound 10a was obtained. The coꢀcrystal structure determined at
2.65 Å resolution (Supporting Information Table S1) clearly shows that PAK4 adopts
an active conformation (DFGꢀin) with a salt bridge (2.3 Å) between Lys350 and
Glu366 in helix αC, and 10a is located in the ATP binding cleft between the Nꢀ and
Cꢀlobes (Figure 3). The imidazole ring of 10a aligns in the hinge region, forming two
Hꢀbonds with the main chain of Leu398, and an additional nonꢀclassical Hꢀbond (3.0
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
7
Å) is formed between Glu396 and the C2ꢀH of the amino imidazole. The Nꢀ1
methyl group makes effective hydrophobic interactions with Met395, the gatekeeper
residue. A favorable electrostatic quinazoline N3 to imidazole C5 interaction was
2
proposed to orient these rings in a coplanar manner. The piperidine NH group makes
a hydrogen bond to the backbone carbonyl of Glu329 (glycine rich loop/Pꢀloop) of 2.9
Å. The coꢀcrystal structure confirmed that 10a is a typical type I inhibitor that binds to
the hinge region without perturbing the DFGꢀin state. The simplicity of the molecule
28
(
MW = 357.5) and its reasonable binding efficiency (LE = 0.31) made it an
attractive starting point for further optimization.
7
ACS Paragon Plus Environment

Page 9 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3. The crystal structure of PAK4 in complex with 10a (PDB code 5XVF).
Note: Asp458 is oriented towards the NH
counterpart Asp407 in PAK1 as in the structure 4O0R (cyan stick). The distance
between the piperidine NH of 10a and the carboxyl group (OD2 atom) of
is 4.7 Å and 6.0 Å, respectively (green line). The ∠
2
in the piperidine of 10a compared to its
2
PAK4
PAK1
Asp458
and Asp407
O(Asp458)−N(piperidine N)−N(piperidine NH2) angle is approximately 51.7° (cyan line).
To further improve the binding affinity of compound 10a, SAR at the hinge binding
motif was explored based on the structural insight provided by the crystal structure
above. As shown in Table 1, the role of the Nꢀ1 group in imidazole analogues is
critical, as cyclopropyl analog 10b and isopropyl analog 10c exhibited similar
inhibitory activity with methyl 10a, while the activity of bulky phenyl 10d was only
half of 10a. Pyrazole analogues 11a [PAK4 K
i
d
= 0.099 ꢀM, PAK4 K (SPR) = 0.043
ꢀ
M] and 11b [PAK4 K = 0.016 ꢀM, PAK4 K
i
d
(SPR) = 0.006 ꢀM] were found to
exhibit significantly improved biochemical and cellular potency (data not shown)
when the donor CꢀH hydrogen bond to the hinge in 10a is replaced by a NꢀH in 11.
Such results clearly showed that the nonꢀclassical CꢀH hydrogen bond is not as good
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
as the classical NꢀH hydrogen bond in this case and the change of the hydrogen bond
from the nonꢀclassical type to the classical type led to significant improvement. The
1
Hꢀpyrazolꢀ3ꢀamine as the hinge binding motif was maintained in the following
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
optimization. Although the pyrazole analogues showed dramatically improved
inhibitory activities against PAK4, they unfortunately did not lead to any
improvement in their PAK4/1 selectivity relative to 10a (11a, 18ꢀfold; 11b, 19ꢀfold).
To discover PAK4 inhibitors with better selectivity over PAK1, next we sought to
exploit the subtle differences between PAK4 and PAK1 crystallographic data
available in the PDB database.
Table 1. SAR of the initial optimization of the hinge binding motif.
X
N
Y
R
HN
N
Cl
N
N
NH2
a
a
b
X
PAK4 Ki
PAK1 Ki
PAK4
N
c
d
e
f
Compound
10a
Y
clogP
PSA
LE /LLE
R
(µM)
(µM)
select.index
0.710
0.652
0.884
1.440
0.099
0.016
>10
>14x
2.6
3.0
3.2
3.8
2.3
2.9
71.3
70.9
70.4
71.0
86.4
85.0
0.34/3.61
0.32/3.19
0.31/2.85
0.27/2.04
0.39/4.67
0.40/4.85
g
1
0b
ND
ND
ND
ND
19x
18x
10c
ND
ND
1
0d
11a
1.891
0.288
11b
11c
0.077
ND
ND
3.1
84.8
0.37/4.01
9
ACS Paragon Plus Environment

Page 11 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1d
0.256
.009/
ND
ND
3.5
3.7
85.1
69.8
0.31/3.09
0.32/4.35
0
0.003/
0.3x/
1
2
-
-
h
i
h
h
i
(
0.006)
0.026/
(0.0006)
(0.1x)
0.052/
2.0x/
4.3
98.3
0.30/3.29
i
(
0.015)
(0.036)
(2.4x)
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
a
PAK4 and PAK1 kinase inhibition was determined using a FRETꢀbased Z’ꢀLyte
assay according to the manufacturer’s instructions (Invitrogen, Carlsbad, USA). The
b
K
i
values are the average of at least two duplicate experiments. SEM < ±20 %. PAK1
c
K / PAK4 K , x = fold. The clogP values were calculated by the Qikprop software
i
i
d
with default settings (pH = 7.0). Van der Waals surface area of polar nitrogen and
e
oxygen atoms. Calculated by the Qikprop software. Ligand efficiency = (ꢀ1.4 log
2
8 f
29 g
i i
K )/(n heavy atoms). Ligandꢀlipophilicity efficiency = (ꢀlog K ) ꢀ (cLogP). ND =
h
19 i
19
not determined. Reported data. Reported data.
Active-site Sequences and Structures of Group I and II PAKs: Rational Design
of a New Series of PAK4 Inhibitors by Targeting the DFG Motif. Like all other
typical protein kinases, the kinase domain of PAKs consists of a Nꢀterminal domain
30
and a Cꢀterminal domain that are connected via a hinge. As shown in Figure 4A, the
ATP binding pocket is located between the two lobes. The selective inhibition of
PAK4 over group I PAKs has proven to be a challenging endeavor, in part because of
the high sequence homology within the ATP binding site of the PAKs (Supporting
Information, Figure S1). To address this challenge, we exploited the
threeꢀdimensional structures of PAKs and defined the ATP binding pocket as a set of
40 residues (P1 through P40, starting at the Nꢀterminus, in Figure 4A) able to interact
with ATPꢀcompetitive inhibitors (checked using KLIFS, a structural kinaseꢀligand
1
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
3
1, 32
interaction database).
The direct comparison of the residues set at the primary
sequence level revealed that more than 65 % (26/40) of residues are conserved in the
respective ATP pockets of PAK1 and PAK4 (Figure 4C). We believe that sequence
variations and unique active site features among PAKs will provide opportunities for
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
16
the selective targeting of these family members.
Figure 4. A comparison of the ATPꢀbinding pocket between group I and II PAKs. (A)
The structure of the PAK kinase domain (generated from the PDB structure 2X4Z).
Residues contributing to the binding cleft (P1ꢀP40) are depicted as spheres. (B)
Alignment (main chain atoms only) of the DFGꢀaspartate of PAK1 and PAK4 in the
active state. An updated survey of all PDB PAK1 structures in the active DFGꢀin state
indicated that 74 % (26/35) of the DFGꢀaspartate side chain is pointing away from the
ATPꢀcompetitive ligand. For the corresponding residue in PAK4, 71 % (22/31) of the
DFGꢀaspartate side chain is pointing toward the ligand. Figures were generated using
PyMOL (The PyMOL Molecular Graphics System, Version 1.7.2 Schrödinger,
11
ACS Paragon Plus Environment

Page 13 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
LLC.). The definition of the dihedral angles χ1(chi): NꢀCαꢀCβꢀCγ. (C) Sequence
alignment of the residues surrounding the ATP binding site in PAK1 and PAK4.
Differential residues between the two kinases are highlighted.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Most protein kinases share a strictly conserved DFG (AspꢀPheꢀGly, P38ꢀP40) motif
in the ATP site that can adopt two distinct conformations, the active DFGꢀin and the
16
inactive DFGꢀout states. Further alignment of the protein crystal structures available
in the Protein Data Bank (PDB) revealed that the orientation of the PAK4
PAK4
DFGꢀaspartate (Asp458
) side chain is different from the corresponding residue in
PAK1 19
PAK1 (Asp407
). As a proof of concept, Crawford et al. proved that taking
advantage of differing orientations of the DFG Asp motif serves as a viable means to
1
9
design Group I selective PAK inhibitors. An updated survey of all PDB PAK1
structures in the active DFGꢀin state indicated that 74 % (26/35) of the DFGꢀaspartate
side chain is pointing away from the ATPꢀcompetitive ligand, and for the
corresponding residue in PAK4, 71 % (22/31) of the DFGꢀaspartate side chain is
pointing toward the ligand (Figure 4B, see the Supporting Information for details).
Presumably, this can be explained by a key residue difference in the DFGꢀ1 residue
406
457
(P36, Thr
in PAK1 and Ser
in PAK4): there is potentially an unfavorable
interaction of the larger, branched Thr residue with the Asp residue in PAK1. In the
2+
active DFGꢀin conformation, the aspartate is required to chelate Mg and helps to
33
orient the γꢀphosphate for its transfer. The important biological function of the DFG
PAK4
PAK1
motif and orientation differences in the DFG Asp (Asp458
/Asp407
)
1
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
suggested to us that the modification of the early lead compounds by specifically
targeting the DFG motif would enhance PAK4 inhibitory activity and selectivity.
Indeed, the PAK4/10a crystal structure showed that the distance from piperidine
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
NH
2
to the side chain of Asp458 is 4.7 Å, and the ∠O(Asp458)−N(piperidine N)−N(piperidine
NH
2
)
angle is 51.7° (Figure 3), which could allow the installation of an appropriate
moiety to interact with the DFG motif. Furthermore, to generate electrostatic
PAK4
interactions or hydrogen binding with the negatively charged residue Asp458
through the quinazoline C2 position, the side chain should adopt an appropriate angle
to ensure that it could make favorable interactions with the DFG motif. In light of
this, we incorporated a putative DFG binding moiety to the central core of 10a by
tethering a cationic amino group or hydrogen bond donor (HBD) group through an
amide carbonyl linkage in consideration of its sp2 hybridization. In doing so, a novel
4ꢀ((1Hꢀpyrazolꢀ3ꢀyl)ꢀamino)quinazolineꢀ2ꢀcarboxamide scaffold was designed by
inserting a carbonyl group at the C2 side chain of our initial set of compounds (see
Table 2).
Structural Modification and SAR Interpretation. The inhibitory activities of the
designed compounds for PAK1 and PAK4 were first evaluated via a wellꢀestablished
fluorescence resonance energy transfer (FRET)ꢀbased Z’ꢀLyte assays (Invitrogen,
3
4
17
Carlsbad, USA). Two well characterized PAK inhibitors, 1 and 2 , were used as
positive controls.
Following the rational design approach above, the incorporation of various tail
amide group led to compounds 12-17 with diverse activities (Table 2). The
1
3
ACS Paragon Plus Environment

Page 15 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
4
ꢀaminoꢀpiperidines 12 and 14 displayed a decrease in PAK4 affinity compared to
i
11a, with K values of 0.354 and 0.674 ꢀM, respectively. Meanwhile, alcohol
analogues 13 and 15 lost an order of magnitude in potency, suggesting that the amino
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
group at the tail position is preferable presumably to interact with the side chain of
PAK4
Asp458
NH
via forming a hydrogen bond or ionic interactions. The methylation of the
2
as in 16 resulted in a decrease by half. The overall trend of decreasing activity is
probably due to steric hindrance with the introduction of a slightly bulkier functional
group at the terminal position of the piperidine ring and more importantly, the
introduction of the carbonyl group between the quinazoline moiety and the piperidine
moiety (Supporting Information, Figure S2C and S2D). Such evidence prompted us to
try to contract the exo nitrogen atom into an endo one, which led to compound 17. As
expected, compound 17 showed dramatically improved binding affinities [PAK4 K
0.045 ꢀM, PAK4 K (SPR) = 0.066 ꢀM] over those of 12. Compared to the 2,
ꢀdiaminoquinazoline analogue 11a, the introduction of an amide sidechain in
compound 17 was clearly essential in improving its PAK4 selectivity (18 fold vs >
22ꢀfold, as shown in Table 3). A molecular docking simulation predicted that the
i
=
d
4
2
two compounds adopted very similar conformations, and their aromatic rings
superimposed very well. The most significant difference between these two inhibitors
is the additional interactions between the piperazine moiety and the DFG binding
3
5
motif of PAK4, which likely provided a chargeꢀassisted hydrogen bond (CAHB,
traditional hydrogen bonding is accompanied by Coulombic interactions) interactions
PAK4
with the deprotonated carboxylic acid side chain of Asp458
(2.3 Å, Figure 5A).
1
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
To obtain more insight into the interactions critical to the excellent activity of
compound 17, compounds 18-20 were designed, synthesized and evaluated.
Unsurprisingly, the importance of this electrostatic interaction is elucidated by the fact
that compound 18 [PAK4 K = 2.172 ꢀM, PAK4 K (SPR) = 1.940 ꢀM], in which the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
i
d
positively charged nitrogen atom was replaced with an oxygen atom, is a very poor
inhibitor (Figure S2E and S2F). Compound 19, an Nꢀmethylated derivative of 17, was
17ꢀfold less active, highlighting the importance of the amine proton acting as a
hydrogen bond donor. When the amine proton was replaced by an amide proton (20,
PAK4 K = 2.408 ꢀM), the activity was decreased by 54ꢀfold (Figure S2G and S2H).
i
The dramatic decrease in activity could be attributed to the fact that the amide group
PAK4
could not form salt bridge interactions with the side chain of Asp458
.
Furthermore, changing the piperazine ring to other heterocyclic secondary amines was
not tolerable, as indicated by the activities of compounds 21-24. This indicated that
the conformation of the amide side chain is critical to the activity of the compounds.
The cyclopropyl analog 25 [PAK4 K
PAK1 K = 2.750 ꢀM] showed improved activity, but its selectivity was not improved
172ꢀfold, shown in Table 3). In summary, an additional hydrogen bond conjoins with
i d
= 0.016 ꢀM, PAK4 K (SPR) = 0.016 ꢀM,
i
(
strong salt bridge interactions, and the correct orientation of the piperazine ring likely
explains the enhanced activity and selectivity of 17 and 25.
Table 2. The SAR of carboxamide substituents.
1
5
ACS Paragon Plus Environment

Page 17 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
a
a
a
PAK4 K
i
PAK4 K
i
PAK4 K
i
Compound
R
Compound
17
R
Compound
22
R
(
µM)
(µM)
(µM)
1
1
1
2
3
4
0.354
3.817
0.045
2.172
0.325
b
18
23
1.630
0.674
3.890
19
20
0.768
2.408
24
1
0.782
0.009
15
16
ꢀ
ꢀ
0.672
21
0.576
2
0.026
a
PAK4 kinase inhibition was determined using a FRETꢀbased Z’ꢀLyte assay according
i
to the manufacturer’s instructions (Invitrogen, Carlsbad, USA). The K values are the
b
average of at least two duplicate experiments. SEM <±20 %. Racemic mixture.
Figure 5. The predicted binding mode of 17. (A) The superimposition of structures of
1
7 (green) with 10a from the crystal structure (PDB code 5XVF). PDB entry 2X4Z
was used for molecular docking simulation with Autodock4. Note: Asp458 in PAK4
1
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
overlaid with a key residue different in PAK1 (cyan, Asp407 from structure 4O0R).
(B) The overlay of PAK4 and PAK1 with 17 (generated from the PDB structures
2X4Z and 4O0R, respectively).
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Beginning with the new scaffold of compounds 17 and 25, further structural
optimization was undertaken to enhance the binding affinity of the inhibitors by
increasing the spatial occupancy of the kinase ATP pocket. Meanwhile, the “floor
pocket” on the bottom of the ATP binding pocket near the DFG motif and the “upper
pocket” beneath the glycine rich loop (Pꢀloop) also captured our attention (shown in
Figure 5B). In contrast, PAK1 has a much wider and shallower “floor pocket”, and
PAK4 has a deeper and overall smaller subꢀpocket. This difference in size was
attributed to the distinct conformations of DFG Asp, as mentioned above, and key
PAK4
PAK1
PAK4
residues differences inside the subꢀpocket (Ser457
vs Thr406
; Ser445
vs
PAK1
Asn394
). Because of the subtle conformation differences in the Pꢀloop (shown in
Figure 5B), the specific features of the “upper pocket” in different PAK isoforms
were also observed. It was apparent from the modeling structures that these
hydrophobic pockets are located around the piperazine ring of 17 and could be
accessed via an axial substituent on the piperazine ring. We hypothesized that an
improvement in the binding affinity and selectivity might be achieved by targeting
these subꢀpockets. Table 3 shows our efforts at potency optimization through
piperazine substitution. The analogues with the methyl group positioned next to the
amide nitrogen (compounds 26-29) were proven unsuccessful at improving potency
and selectivity. Among them, the compounds with a (R)ꢀconfiguration at the methyl
1
7
ACS Paragon Plus Environment

Page 19 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
stereocenter (28, 29) exhibited better potency and selectivity than their corresponding
(S)ꢀanalogues (26, 27). A significant improvement in PAK4 selectivity was obtained
by incorporating a (R)ꢀmethyl group next to the amine nitrogen on the piperazine ring
(30, 31), which most likely contributes to their shape complementarity to maximize
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
van der Waals contacts. In the end, compound 31 had robust PAK4 affinity (PAK4 K
i
=
0.009 ꢀM) and excellent selectivity at 346ꢀfold over PAK1 (PAK1 K
i
= 3.112 ꢀM),
2
0
comparable to that of 5 (PAK1 K
i
= 2.9 ꢀM ). Interestingly, the (S)ꢀmethyl
substituent (32, 33) was not tolerated, suggesting that chirality is a crucial element
that influences van der Waals interactions. The bulky (R)ꢀethyl analogue, 36 (PAK4
i i
K = 0.006 ꢀM, PAK1 K = 0.368 ꢀM), displayed similar potency against PAK4 in the
single digital nanomolar range, but it had little selectivity against PAK4 (61ꢀfold).
Further substitution of the piperazine ring, however, resulted in a significant decrease
in activity (dimethyl analogues 34 and 35).
Table 3. The in vitro biological and physical properties of piperazine analogues.
a
a
PAK4 K
i
PAK1 K
i
PAK4
select.index
>212x
c
d
e
f
g
Compound
R
1
clogP
pKa
PSA
LE /LLE
b
(
µM)
(µM)
>10
1
7
Me
0.045
0.016
1.6
2.2
7.8
7.8
98.7
97.4
0.40/5.78
0.39/5.63
25
cꢁPr
2.750
172x
2
2
2
6
7
8
Me
cꢁPr
Me
0.218
0.026
0.151
>4.52
0.383
>4.52
>21x
15x
2.4
3.0
1.9
8.0
8.0
8.0
91.4
90.1
92.5
0.35/4.30
0.37/4.61
0.35/4.94
>30x
1
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
29
cꢁPr
0.017
2.080
122x
2.5
8.0
91.1
0.38/5.29
3
3
0
1
Me
cꢁPr
Me
0.051
0.009
0.306
0.028
>10
>196x
346x
>15x
25x
2.0
2.6
2.0
2.6
7.9
7.9
7.9
7.9
97.5
96.2
97.5
96.1
0.38/5.34
0.39/5.49
0.34/4.57
0.36/4.99
3.112
>4.52
0.690
32
3
3
cꢁPr
i
h
3
3
4
5
Me
0.119
0.114
ND
ND
2.2
2.8
8.0
8.0
95.5
94.1
0.35/4.72
0.32/4.11
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
i
cꢁPr
>4.52
0.368
>40x
3
6
cꢀPr
0.006
61x
2.9
8.0
94.0
0.38/5.29
1
ꢀ
ꢀ
ꢀ
ꢀ
0.009
0.026
0.003
0.052
0.3x
2.0x
3.7
4.3
9.6
8.1
69.8
98.3
0.34/4.82
0.30/3.30
2
a
PAK4 and PAK1 kinase inhibition was determined using a FRETꢀbased Z’ꢀLyte
assay according to the manufacturer’s instructions (Invitrogen, Carlsbad, USA). The
b
K
i
values are the average of at least two duplicate experiments. SEM <±20 %. PAK1
c
K / PAK4 K , x = fold. The clogP values were calculated by the Qikprop software
i
i
d
with default settings (pH = 7.0). The pKa values were calculated using ChemAxon’s
e
MarvinSketch (version 5.4.11). The van der Waals surface area of polar nitrogen and
f
oxygen atoms. Calculated by the Qikprop software. Ligand efficiency = (ꢀ1.4 log
2
8 g
29 h
K
i
i
)/(n heavy atoms). Ligandꢀlipophilicity efficiency = (ꢀlog K ) ꢀ (cLogP). ND =
i
not determined. Mixture of diastereomers.
To better understand the origin of the excellent PAK4 selectivity, the crystal
structures of the PAK4 kinase domain in complex with 31 (PDB code 5XVG) and its
highly similar methyl analogue 30 (PDB code 5XVA) were solved to 2.10 and 1.85
Å, respectively (Supporting Information Table S1). The binding mode observed for
31 was consistent with that observed with 30 (Figure 6), and the key interactions are
shown in Figure 7. The aminopyrazole of 31 forms classic donorꢀacceptorꢀdonor
1
9
ACS Paragon Plus Environment

Page 21 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
PAK4
Hꢀbond interactions with the hinge region (carbonyl of Glu396
; carbonyl and NH
PAK4
of Leu398
), which is similar to that observed in the crystal structure of PAK4 in
complex with 10a. The cyclopropyl moiety makes effective hydrophobic interactions
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
PAK4
PAK4
with Met395
, the gatekeeper residue, and Val335
. Consistent with our
hypothesis, the NH on the piperazine ring forms a chargeꢀassisted hydrogen bond
PAK4
with the negatively charged residue Asp458
through the amide carbonyl linkage,
PAK4
with a distance of 2.8 Å. It is noteworthy that the carboxylic side chain of Asp444
induces conformational changes to form electrostatic interactions with the tail of 31
(
3.4 Å). In comparison with the putative structure of PAK1/31, the side chain of
PAK1
Asp407
is pointing away from compound 31 and hence not able to form
electrostatic interactions with the piperazine ring (Figure 7B). Additionally, the
piperazine ring of 31 makes a lipophilic contact with the “floor pocket” of PAK4
mentioned above, and the (R)ꢀmethyl group of the piperazine ring fills a hydrophobic
dimple in the glycine rich loop above, which also explains why an additional
(
R)ꢀmethyl group increased the binding affinity and selectivity for PAK4. The
differences in the binding modes for 31 to PAK4 and PAK1 are also directly linked to
3
6
the differences in docking scores (Xꢀscore : −8.95 for PAK4 and −6.75 for PAK1)
and to the experimentally observed selectivity. The beneficial van der Waals
interactions resulting from shape complementarity as well as the anticipated CAHB
PAK4
with Asp458
contribute to the increased binding affinity of 31, demonstrating the
rationale of our optimization hypothesis, thereby providing a molecular explanation
for the subtype selectivity.
2
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 6. The crystal structure of PAK4 in complex with 30 (PDB code 5XVA).
Superimposition of the structures of 30 (green) with 10a (cyan) from the crystal
structure (PDB code 5XVF).
Figure 7. (A) The crystal structure of PAK4 in complex with 31 (PDB code 5XVG).
(B) Predicted binding mode of 31 in PAK1 (PDB code 4O0R). A schematic
representation of the contacts between 31 and the binding site residues of PAK4 (C)
2
1
ACS Paragon Plus Environment

Page 23 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
3
7
and PAK1 (D) was visualized by LigandScout. Note: the piperazine ring of 31
PAK4
provides chargeꢀassisted hydrogen bond (CAHB) interactions with Asp458
and
PAK4
Asp444
.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 8. Microenvironment of the 6ꢀsubstituent of the inhibitors. The Xꢀray structure
of 31 in complex with PAK4 (green, PDB 5XVG) overlaid with key residue
differences with PAK1 (cyan, Tyr346 and Ala348 from structure 4O0R).
Lastly, the role of the substituent at the Cꢀ6 position of the quinazoline core was
investigated, and the data are summarized in Table 4. As shown in Figure 8, the
6ꢀchloro group lies in the clifflike entrance around the hinge area and leads to van der
PAK4
PAK4
PAK4
Waals interactions with Phe397
, Glu399
and Ile327
(P24, P26, and P1 in
Figure 4B, respectively). To probe this distinctive entrance region, the removal of the
chloro group (compound 37, PAK4 K = 0.017 ꢀM) caused a slight decrease in the
i
potency for PAK4 and a significant reduction in PAK4/1 selectivity (37, PAK4
selectivity: 63ꢀfold). In addition, the position of the chloro group was also found to
affect the selectivity profile. Although the 7ꢀchloro derivative 38 (PAK4 K
i
= 0.006
ꢀ
M) displayed better potency than that of 31, it suffered from reduced PAK4
2
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
selectivity (38, PAK4 selectivity: 57ꢀfold). Further investigation suggested that the
6ꢀchloro group in 31 could be replaced by electronꢀwithdrawing substituents, such as
F (39) and Br (40), and still retain its strong PAK4 affinity but the PAK4 selectivity
was reduced by variable degrees (PAK4 selectivity: 31ꢀfold and 190ꢀfold,
respectively). The compound bearing large electronꢀdonating group, such as a
methoxyl group (41), suffered from reduced potency for both PAK1 and PAK4
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
i i
(PAK4 K = 0.036 ꢀM; PAK1 K > 4.52 ꢀM). These key findings highlighted that the
importance of both shape and electrostatic complementarity at the 6ꢀposition of the
quinazoline core because the residues around the 6ꢀposition between PAK4 and
PAK4
PAK1
PAK4
PAK1
PAK1 are not conserved (Phe397
vs Tyr346
; Glu399
vs Ala348
,
Figure 8). It may be possible to further exploit such sequence differences for the
future design of inhibitors that distinguish PAK4 and PAK1 to a larger extent.
Table 4. The SAR of the quinazoline core.
a
a
PAK4
PAK1
PAK4
select.index
346x
d
e
f
g
h
Compound
R
1
R
2
clogP
pKa
PSA
LE / LLE
b
Ki (µM)
0.009
0.017
0.007
0.016
0.011
0.036
Ki (µM)
3.112
3
3
3
1
7
8
Cl
H
H
H
Cl
H
H
H
2.6
2.1
2.6
2.3
2.6
2.2
7.9
7.9
7.9
7.9
7.9
7.9
96.2
95.6
95.6
96.2
95.6
104.6
0.39/5.49
0.39/5.69
0.39/5.60
0.38/5.48
0.38/5.34
0.35/5.26
1.072
0.401
0.503
2.085
>4.52
63x
H
57x
39
F
31x
4
4
0
1
Br
190x
OMe
>126x
a
PAK4 and PAK1 kinase inhibition was determined using a FRETꢀbased Z’ꢀLyte
assay according to the manufacturer’s instructions (Invitrogen, Carlsbad, USA). The
2
3
ACS Paragon Plus Environment

Page 25 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
b
K
i
i
values are the average of at least two duplicate experiments. SEM <±20 %. PAK1
c
d
K
i
/ PAK4 K , x = fold. Determined by differential scanning fluorimetry (DSF). The
clogP values were calculated using the Qikprop software with the default settings (pH
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
e
= 7.0). The pKa values were calculated using ChemAxon’s MarvinSketch (version
f
5
.4.11). The van der Waals surface area of polar nitrogen and oxygen atoms.
g
Calculated using the Qikprop software. Ligand efficiency = (ꢀ1.4 log K
i
)/(n heavy
28 h
29
i
atoms). Ligandꢀlipophilicity efficiency = (ꢀlog K ) ꢀ (cLogP).
After several rounds of lead optimization aided by Xꢀray crystallography and
molecular docking simulations, several compounds were identified with significant
i
inhibitory potency against PAK4 with K values higher than or compatible to the
reference compound 2. Over the course of the optimization, a high degree of ligand
efficiency was maintained, with LE ≥ 0.38 and LLE ≥ 5.3 for the majority of subꢀ20
nM K
potency, selectivity, and desirable physicochemical properties (pKa = 7.93, clogP =
.54, PSA = 96.1). Encouraged by the potency of 31 and its selectivity over PAK1, its
kinome selectivity profile was investigated by determining % inhibition at 0.1 ꢀM and
.0 ꢀM (concentration equal to 11ꢀfold and 110ꢀfold above PAK4 K , respectively)
against a panel of 54 kinases at the ATP K concentration, and the full data set is
i
compounds. Among them, compound 31 demonstrated the best balance of
2
1
i
m
available in the supporting information. Good kinase selectivity was observed for 31,
with only PAK4 producing greater than 80 % inhibition at the screening concentration
of 0.1 ꢀM (PAK4 IC50 = 0.0111 ꢀM, Figure 9). These data demonstrated that
compound 31 is a potent PAK4 inhibitor with good kinase spectrum selectivity.
2
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Although the DFG motif is highly conserved among protein kinases, the proper
spatial arrangement of 31 with the DFG motif of PAK4 in combination with other
residues differences within the ATP binding site would likely explains the selectivity
of 31.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 9. Selectivity profile of compound 31 (A) measured at a concentration of 0.1
®
ꢀ
M and (B) 1.0 ꢀM in a panel of 54 kinases generated with the SelectScreen Profiling
Service from Life Technologies (red columns denote >80 %, yellow columns between
40 and 80 %, and green columns <40 % inhibition).
Cellular effects of Compound 31. Previous studies have shown that PAK4 is
overexpressed in several nonꢀsmall cell lung cancer (NSCLC) cell lines and human
1
2, 38
NSCLC tissues.
Increased expression of PAK4 was correlated with metastasis,
shorter overall survival, advanced stage of NSCLC. Furthermore, selective
suppression of PAK4 with siRNA has been shown to substantially suppress the
migration and invasion of the PAK4ꢀdependent A549 cells (human lung
38
adenocarcinoma epithelial cell). Considering the potent PAK4 inhibition of
compound 31, we first investigated its antiꢀproliferative effects on A549 cells.
Meanwhile, the tumour cell line NCIꢀH460 (human large cell lung cancer cell), whose
2
5
ACS Paragon Plus Environment

Page 27 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
growth was not dependent on PAK4, and HEKꢀ293 cells (human embryonic kidney
293 cell) were used to test the potential offꢀtarget effects of potent PAK4 inhibitors. It
was shown that compound 31 has moderate antiproliferative activities against A549
cells (Figure 10A), which is similar to the previous observation of 5 on
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
20
MDAꢀMBꢀ436 and MCF10A PIK3CA cells. The relatively low cell growth
inhibitory activities suggested that PAK4 might not be the “driving force” for
proliferation of these cancer cells. Next, the effects of compound 31 on tumor cell
migration and invasion were assessed using wound healing and transwell assays.
Microphotographs showed that untreated A549 cells filled most of the wounded area
2
days after scratching the cell monolayer, whereas treatment with indicated doses of
compound 31 significantly suppressed the wound healing (Figure 10B). Compound
1 treatment also significantly inhibited the A549 cells migration (Figure 10C, E) and
3
invasion (Figure 10D, F) by transwell assay. In contrast, the decrease in cell migration
and invasion was not pronounced in NCIꢀH460 and HEKꢀ293 cells (Figure 10GꢀJ),
suggesting that inhibition of cell motility by compound 31 is in fact related to PAK4
inhibition. To further illustrate the mechanism, the effects of compound 31 on the
PAK4/LIMK1/cofilin, PAK4/MMP2, or PAK4/GEFꢀH1 signaling pathways were
examined by Western blot analysis. As shown in Figure 10K, phosphorylation of
PAK4, LIMK1, and cofilin were inhibited by compound 31 treatment in
doseꢀdependent manner. Furthermore, phosphorylation of Ser810 on GEFꢀH1 and
expression of MMPꢀ2 were also inhibited by compound 31, which may contribute to
its antiꢀmetastasis effect of A549 lung cancer cells.
2
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 10. Compound 31 suppresses the migratory and invasive potential of A549
lung cancer cells via inhibition of PAK4 kinase activity and its downstream pathways.
(A) MTT assay showed the antiꢀproliferative effect of compound 31 on A549,
HEKꢀ293 and NCIꢀH460 cells. (B) Compound 31 inhibits migration of A549 lung
cancer cells in a wound healing assay. (CꢀF) Transwell assay [ꢀ(D)/+ (C) matrigel]
was performed to show the effect of indicated concentrations of 31 on A549 lung
cancer cells migration (C) and invasion (D). And migrated cells were counted
respectively (EꢀF). The results are presented as the mean ± standard deviation (*, p <
0.05; **, p < 0.01; ***, p < 0.001 vs control). (GꢀH) Effect of 31 on HEKꢀ293 cells
migration (G) and invasion (H). (IꢀJ) Effect of 31 on NCIꢀH460 cells migration (I)
2
7
ACS Paragon Plus Environment

Page 29 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
and invasion (J). Scale bar, 500 ꢀm. (K) Western blot assay showed the effect of
4
74
compound 31 on the phosphoꢀPAK4 Ser level and on the phosphorylation levels of
PAK4 downstream targets LIMK1 and cofilin. The expression of MMPꢀ2 and the
phosphorylation of GEFꢀH1 were also downregulated by compound 31 treatment in a
doseꢀdependent manner. A549 cells were cultured with vehicle or the indicated
concentrations of compound 31 for 24 h, and then proteins were extracted and
subjected to analysis.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
In Vitro Metabolic Stability, CYP450 Inhibition, hERG Inhibition and in Vivo
Pharmacokinetics Properties. A preliminary ADMET properties assessment of 31
was conducted to evaluate its potential for further development. The in vitro
metabolic stability of the compound was assessed in liver microsomes (rat and human)
and rat blood plasma. As shown in Table 5, compound 31 possessed moderate
microsomal stability and excellent blood plasma stability (for more information, see
Supporting Information). Further CYP450 inhibition determination showed that 31
possessed favorable metabolic properties, as the inhibition ratios for the five main
CYPs (CYP1A2, 2C9, 2C19, 2D6, and 3A4) were less than 20 % at 10 ꢀM. Previous
studies demonstrated that a basic nitrogen (or generally the positive charge feature)
would lead to greater activity against human etherꢀaꢀgoꢀgoꢀrelated gene (hERG)
39
channels (Kv11.1). Gratifyingly, the pKa of compound 31 (calculated pKa = 7.9)
was reduced by greater than two units than the early lead (calculated pKa = 10.0 for
10a) and was found to be safe (IC50 > 40 ꢀM) when tested in a hERG patch clamp
assay for assessing hERGꢀassociated cardiotoxicity. On the basis of these favorable in
2
8
ACS Paragon Plus Environment