New activators of eIF2α Kinase Heme-Regulated Inhibitor (HRI) with improved biophysical properties

Article abstract of DOI:10.1016/j.ejmech.2019.111973

Heme-regulated inhibitor (HRI), a eukaryotic translation initiation factor 2 alpha (eIF2α) kinase, is critically important for coupling protein synthesis to heme availability in reticulocytes and adaptation to various environmental stressors in all cells. HRI modifies the severity of several hemoglobin misfolding disorders including β-thalassemia. Small molecule activators of HRI are essential for studying normal- and patho-biology of this kinase as well as for the treatment of various human disorders for which activation of HRI or phosphorylation of eIF2α may be beneficial. We previously reported development of 1-((1,4-trans)-4-aryloxycyclohexyl)-3-arylureas (cHAUs) as specific HRI activators and demonstrated their potential as molecular probes for studying HRI biology and as lead compounds for treatment of various human disorders. To develop more druglike cHAUs for in vivo studies and drug development and to expand the chemical space, we undertook bioassay guided structure–activity relationship studies replacing cyclohexyl ring with various 4-6-membered rings and explored further substitutions on the N-phenyl ring. We tested all analogs in the surrogate eIF2α phosphorylation and cell proliferation assays, and a subset of analogs in secondary mechanistic assays that included endogenous eIF2α phosphorylation and expression of C/EBP homologous protein (CHOP), a downstream effector. Finally, we determined specificity of these compounds for HRI by testing their anti-proliferative activity in cells transfected with siRNA targeting HRI or mock. These compounds have significantly improved cLogPs with no loss of potencies, making them excellent candidates for lead optimization for development of investigational new drugs that potently and specifically activate HRI.

Full text of DOI:10.1016/j.ejmech.2019.111973

Journal Pre-proof
New activators of eIF2α Kinase Heme-Regulated Inhibitor (HRI) with improved
biophysical properties
Qingwen Zhang, Ronghui Du, Guilherme Rodrigo Reis Monteiro dos Santos, Revital
Yefidoff-Freedman, Andrew Bohm, Jose Halperin, Michael Chorev, Bertal H. Aktas
PII:
S0223-5234(19)31125-0
DOI:
https://doi.org/10.1016/j.ejmech.2019.111973
EJMECH 111973
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 31 May 2019
Revised Date: 13 December 2019
Accepted Date: 13 December 2019
Please cite this article as: Q. Zhang, R. Du, G.R. Reis Monteiro dos Santos, R. Yefidoff-Freedman, A.
Bohm, J. Halperin, M. Chorev, B.H. Aktas, New activators of eIF2α Kinase Heme-Regulated Inhibitor
(HRI) with improved biophysical properties, European Journal of Medicinal Chemistry (2020), doi: https://
doi.org/10.1016/j.ejmech.2019.111973.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
©
2019 Published by Elsevier Masson SAS.

R1
R4
H
H
R2
R3
N
N
X
CF3
m
O
O
X
n
str uctur e-activity relationship (SAR) studies
*
Concentration that inhibits growth of human melanoma CRL-2813 cells by 50%

New Activators of eIF2α Kinase Heme-Regulated Inhibitor (HRI) with Improved
Biophysical Properties
a,b #
b,c #
b #
Qingwen Zhang
, Ronghui Du
, Guilherme Rodrigo Reis Monteiro dos Santos
b
d
b
b
Revital Yefidoff-Freedman , Andrew Bohm , Jose Halperin , Michael Chorev *, and
b*
Bertal H. Aktas
a
Division of Medicinal and Process Chemistry, Shanghai Institute of Pharmaceutical
Industry, Pudong, Shanghai 201203, China
b
Hematology Laboratory for Translational Research, Department of Medicine, Brigham
and Women’s Hospital and Harvard Medical School, Boston, MA 02115, USA
c
Medicine School of Nanjing University, Nanjing, Jiangsu 210093, China
d
Tufts University Medical School, Boston, MA 02117, USA
#
These authors contributed equally.
*
Corresponding authors
Phone: (617) 525-3142. Fax: (617) 582-6069.
E-mail: huseyin_aktas@hms.harvard.edu
Keywords:
Heme-regulated inhibitor kinase (HRI)
eIF2α phosphorylation
Cap-dependent translation
HRI activator
Drug-likeness
Cancer
1
/ 47

Abstract
Heme-regulated inhibitor (HRI), a eukaryotic translation initiation factor 2 alpha
(eIF2α) kinase, is critically important for coupling protein synthesis to heme
availability in reticulocytes and adaptation to various environmental stressors in all
cells. HRI modifies the severity of several hemoglobin misfolding disorders including
β-thalassemia. Small molecule activators of HRI are essential for studying normal- and
patho-biology of this kinase as well as for the treatment of various human disorders for
which activation of HRI or phosphorylation of eIF2α may be beneficial. We previously
reported development of 1-((1,4-trans)-4-aryloxycyclohexyl)-3-arylureas (cHAUs) as
specific HRI activators and demonstrated their potential as molecular probes for
studying HRI biology and as lead compounds for treatment of various human
disorders. To develop more druglike cHAUs for in vivo studies and drug development
and to expand the chemical space, we undertook bioassay guided structure–activity
relationship studies replacing cyclohexyl ring with various 4-6-membered rings and
explored further substitutions on the N-phenyl ring. We tested all analogs in the
surrogate eIF2α phosphorylation and cell proliferation assays, and a subset of analogs
in secondary mechanistic assays that included endogenous eIF2α phosphorylation and
expression of C/EBP homologous protein (CHOP), a downstream effector. Finally, we
tested specificity of these compounds for HRI by testing their anti-proliferative
activity in cells transfected with siRNA targeting HRI or mock. These compounds
have significantly improved cLogPs with no loss of potencies, making them excellent
2
/ 47

candidates for lead optimization for development of investigational new drugs that
potently and specifically activate HRI.
Abbreviations:
ΦOcHΦUs, 1-phenyl-3-(4-phenoxy)cyclohexyl)ureas; ΦOcAlkΦUs,
1
1
1
1
1
1
-phenyl-3-(4-phenoxy)cycloalkyl)ureas; 4-CF -ΦOcAlkΦUs,
3
-phenyl-3-((1,3/4-trans)-4-(4-trifluoromethyl)phenoxy)cycloalkyl)ureas; 4-CF -ΦOcBΦUs,
3
-phenyl-3-((1,3-trans)-4-(4-trifluoromethyl)phenoxy)cyclobutyl)ureas; 4-CF -ΦOcPΦUs,
3
-phenyl-3-((1,3-trans)-4-(4-trifluoromethyl)phenoxy)cyclopentyl)ureas; 4-CF -ΦOcHΦUs,
3
-phenyl-3-((1,4-trans)-4-(4-trifluoromethyl)phenoxy)cyclohexyl)ureas; 4-CF -ΦOPyΦUs,
3
-phenyl-3-(5-(4-(trifluoromethyl)phenoxy)pyrimidin-2-yl)ureas; HRI, heme regulated inhibitor or
heme regulated eIF2α kinase; eIF2, eukaryotic translation initiation factor 2; eIF2α, eukaryotic
translation initiation factor 2 alpha; GTP, guanosine triphosphate; Met-tRNA , initiator methionyl
i
tRNA; mRNA, messenger RNA; ATF-4, activating transcription factor 4; CHOP, C/EBP
homology protein; PPAR-γ, peroxisome proliferator-activated receptor γ; SAR, structure-activity
relationship; LC-MS, liquid chromatography-mass spectrometry; NMR, nuclear magnetic
resonance; DIAD, diisopropyl azodicarboxylate; TFA, trifluoroacetic acid; BOC,
tert-butyloxycarbonyl; DMF, N,N-dimethylformamide; NMP, N-methyl-2-pyrrolidinone; DMSO,
dimethyl sulfoxide; DRL assay, surrogate dual-luciferase eIF2α phosphorylation assay; uORFs,
upstream open reading frames; 5’UTR, 5' untranslated region; Cycloalk. config., Cycloalkyl
configuration.
1. Introduction
Eukaryotic translation initiation factor 2 alpha (eIF2α) kinases play critical roles in
responding to various stress conditions and adapting cellular metabolism to extracellular
cues. Heme regulated eIF2α kinase, also known as heme regulated inhibitor (HRI), was
first of four eIF2α kinases to be discovered. HRI expression is highest in the red blood cell
(RBC) precursors where it critically contributes to differentiation and maturation of
myelogenic lineage into RBCs. In the RBC precursors, HRI is maintained in the inactive
state by heme [1]. Low levels of free heme lead to HRI activation through
3
/ 47

autophosphorylation [2]. Activated HRI phosphorylates eIF2α, which reduces the level of
the translation initiation complex formed by eIF2, GTP and Met-tRNA (the ternary
i
complex, eIF2·GTP·Met-tRNA ), which is critical for the formation of the 43S
i
pre-initiation complex [3]. Reducing the amount of the ternary complex inhibits
translation initiation thereby reducing global protein synthesis. By coupling globin
synthesis to heme availability, HRI plays a critical role in attenuating severity of
iron-deficiency anemia, β-thalassemia and most likely other anemic disorders [4].
HRI expression is not limited to myelogenic lineage; it is expressed in almost all tissues
examined. HRI is the only eIF2ɑ kinase activated by arsenate induced oxidative stress [5].
It is also activated by nitrous oxide [6], osmotic shock, and heat shock [7]. By
phosphorylating eIF2ɑ and reducing the amount of the ternary complex, HRI activates
downstream effectors of this pathway including activating transcription factor 4 (ATF-4)
and pro-apoptotic transcription factor C/EBP homology protein (CHOP). Activation of
HRI (and the resulting phosphorylation of eIF2ɑ) may initially be cytoprotective but its
sustained activation may cause cell death [8].
Recent discovery of small molecule activators of HRI [9-12] provided scientific
community with tools to better understand the role of this kinase as well as eIF2ɑ
phosphorylation in normal- and patho-biology. For example, these agents allowed us and
others to study HRI’s regulation of fibroblast growth factor 21 (FGF21) activity and its
role in diabetes and fatty liver disease and interaction of HRI/eIF2ɑ-P/ATF-4 pathway
with the PPAR-γ pathway [13,14], regulation of host/intracellular pathogen interactions
4
/ 47

[
15,16], and treatment of therapy resistant multiple myeloma cancers [8,17]. Importantly,
when employed in combination with other eIF2α kinase activators, HRI activators are
invaluable tools for dissecting contribution of eIF2α vs other substrates of eIF2α kinases
to normal- and patho-biology. This cannot be accomplished by using non-specific eIF2α
activators or eIF2α phosphatase inhibitors [18,19]. Understanding the molecular basis of
HRI activation by small molecules will be helpful for better understanding diverse roles
heme and heme regulated proteins in normal- and patho-physiology [20].
Demonstration by us others that modulation of eIF2α phosphorylation is a viable
approach for the treatment of animal models of proliferative and some neurodegenerative
disorders [18,19,21-24] prompted us to further screen libraries of small molecules to
identify inducers of eIF2α phosphorylation. Initially, we screened a diverse
N,N’-disubstituted urea library, discovered 1-phenyl-3-(4-phenoxy)cyclohexyl)ureas
(ΦOcHΦUs) as potent in vitro inducers of HRI-dependent phosphorylation and carried out
a limited structure-activity relationship (SAR) study [11]. Subsequently, design and
synthesis of focused library of ΦOcHΦUs led us to novel druglike analogs with improved
biophysical properties that potently induce eIF2α phosphorylation and expression of its
downstream effector CHOP and inhibit cancer cell proliferation at sub-micromolar
concentrations [12]. Moreover, one of the ΦOcHΦUs with more potent in vitro activity
(1-(3-cyano-5-trifluoromethyl)phenyl-3-(trans-4-(4-(trifluoromethyl)phenoxy)cyclohexyl)
urea, 1-VI in Table 1) displayed in vivo efficacy in xenograft mice model of human
melanoma with no apparent organ toxicity and a very encouraging target validation
5
/ 47

profile. To better delineate the pharmacophore in the ΦOcHΦUs and further optimize the
biophysical properties to improve bioavailability and develop a promising lead candidate
for a comprehensive preclinical development, we undertook a new round of SAR
optimization. To this end, our study focuses on the cyclohexyl linker scaffolding the
phenoxy and the N-phenyl ureido moieties in the ΦOcHΦU chemotype. Herein we report
on novel more druglike ΦOcHΦUs that activate eIF2α phosphorylation and its
downstream effectors, and potently inhibit cancer cells proliferation. With this information
in hand we believe to have promising molecular probes and candidates for comprehensive
in vivo preclinical evaluation aimed at understanding the role of HRI and eIF2α
phosphorylation in normal- and patho-biology. These studies could help the advancement
of HRI activators as mechanistically unique agents for treatment of human disorders.
2. Results and discussion
2.1. Molecular design
Recently we reported progress in optimizing the ΦOcHΦU chemotype in which a
combination
of
N-(3-trifluoromethyl-5-cyano)phenyl
and
N-(1,4-trans)-(4-(4-trifluoromethyl)phenoxy)cyclohexyl) moieties contributes to enhanced
potency and more favorable cLogP values [12]. Herein we designed focused libraries of
N,N’-disubstituted ureas to examine the role of the N-(1,4-trans)-cyclohexyl moiety as a
locally constraining linker by replacing it with cyclopentyl and cyclobutyl that are smaller,
less hydrophobic and more constrained cycloalkyl moieties. In addition, we also sought to
test the hypothesis that global constraint introduced by replacing the
6
/ 47

(1,4-trans)-cyclohexyl with a 2,5-pyrimidyl may form a 6-membered hydrogen bond with
the distant ureido NH. Furthermore, we were eager to explore the impact of nature and
position of the N-phenyl substituents on HRI activity and drug-likeness of the desired
N,N’-disubstituted ureas.
2
.2. Synthesis
The common and principal intermediates for the synthesis of this series of
-phenyl-3-(4-phenoxy)cycloalkyl)ureas (ΦOcAlkΦUs) are the
/4-(4-(trifluoromethyl)phenoxy)cycloalkan-1-amines (B) (Scheme 1). Based on the
1
3
commercial availability of starting materials the synthesis of B was accomplished by 2
different approaches: 1. Mitsunobu coupling of variety of tert-butyl
3/4-hydroxycycloalkyl)carbamates with 4-(trifluoromethyl)phenol in the presence of
triphenylphosphine and diisopropyl azodicarboxylate (DIAD) in anhydrous THF delivered
the Mitsunobu products tert-butyl
3/4-(4-(trifluoromethyl)phenoxy)cycloalkyl)carbamates (A). Treatment of A with
a
(
(
trifluoroacetic acid (TFA) removed the protecting group BOC (tert-butyloxycarbonyl) and
yielded the amines B in good yields (Pathway AI) [25]. The Mitsunobu reaction provides a
highly reliable method to inverting the configuration of secondary alcohol chiral carbon
[
26]. A-V was prepared from optically pure tert-butyl (1S,3R)-3-hydroxycyclopentyl
carbamate (C, m=0, n=1). We have successfully harvested from methylene chloride single
crystals of A-V, whose absolute configuration was determined to be (1S,3S) by single
crystal X-ray diffraction (Cambridge Crystallographic Data Centre 1871144) (Fig. 1). As
anticipated, this result confirms the inversion of configuration of the chiral center
7
/ 47

undergoing
1,4-trans)-4-aminocyclohexan-1-ol by 1-fluoro-4-(trifluoromethyl)benzene in the
presence of NaH in DMF furnished the known intermediate B (m=1, n=1) (Pathway AII)
11,12].
the
Mitsunobu
reaction.
2.
A
direct
O-alkylation
of
(
[
With this common and principal intermediate B in hand, the desired
N,N’-disubstituted ureas were readily accessed by acylation with either commercially
available isocyanates D (Pathway A) or substituted phenyl carbamates E that were
conveniently prepared from the corresponding anilines F and phenyl chloroformate
(Pathway B) (Scheme 1).
8
/ 47

Scheme
1.
Synthetic
pathways
employed
for
the
synthesis
of
1
-phenyl-3-(3/4-(4-(trifluoromethyl)phenoxy)cycloalkyl)ureas (4-CF -ΦOcAlkΦUs). Pathway
A
3
generated the desired N,N’-disubstituted ureas utilizing isocyanates to acylate the substituted
cycloalkylamine B. The later was generated through either a Mitsunobu coupling of BOC protected
aminocycloalkanol C (m=0, n=1 for cyclopentyl series; m=0, n=0 for cyclobutyl series) and
4
-(trifluoromethyl)phenol followed by TFA deprotection (Pathway AI) or O-alkylation of
1,4-trans)-4-aminocyclohexan-1-ol (m=1, n=1 for cyclohexyl series) by
-fluoro-4-(trifluoromethyl)benzene (Pathway AII). Pathway B employed phenyl carbamates E as
(
1
acylating agents of the O-substituted 3/4-hydroxycycloalkan-1-amines B. The phenyl carbamates E were
obtained by acylating substituted anilines F with phenyl chloroformate.
Fig. 1. X-ray ORTEP diagram of A-V.
The
synthesis
of
this
series
of
1
-phenyl-3-(5-(4-(trifluoromethyl)phenoxy)pyrimidin-2-yl)ureas (4-CF -ΦOPyΦUs: 1-py –
3
4-py, 7-py and 10-py) utilized Pathways AII and B (Scheme 2). The initial attempted
condensation of 5-chloropyrimidin-2-amine and 4-(trifluoromethyl)phenol in the presence of
potassium hydroxide and potassium carbonate at elevated temperature in DMF or NMP
failed to deliver the key intermediate 5-(4-(trifluoromethyl)phenoxy)pyrimidin-2-amine (G).
Alternatively,
2-aminopyrimidin-5-ol
was
condensed
with
1-fluoro-4-(trifluoromethyl)benzene. Due to the greater acidity of 2-aminopyrimidin-5-ol vs
(1,4-trans)-4-aminocyclohexan-1-ol,
NaH/DMF used for the formation of
(1,4-trans)-4-(4-(trifluoromethyl)phenoxy)cyclohexan-1-amine (B, m=1, n=1) was replaced
9
/ 47

with Cs CO /DMSO for the formation of the key intermediate G albeit at lower yields
2
3
(Scheme 2).
Scheme
2.
Synthetic
pathways
employed
for
the
synthesis
of
1
-phenyl-3-(5-(4-(trifluoromethyl)phenoxy)pyrimidin-2-yl)urea series (4-CF -ΦOPyΦUs). Pathway A
3
delivered the desired N,N’-disubstituted ureas utilizing isocyanates to acylate the key intermediate G,
5
-(4-(trifluoromethyl)phenoxy)pyrimidin-2-amine. The later was generated through O-arylation of
2
-aminopyrimidin-5-ol by 1-fluoro-4-(trifluoromethyl)benzene in the presence of cesium carbonate in
DMSO (Pathway AII). Pathway B employed phenyl carbamates E as acylating agents of the key
intermediate G. The phenyl carbamates E were obtained by acylating substituted anilines F with phenyl
chloroformate.
It’s advisable to replace B with G in Scheme 2 to avoid confusion.
On the basis of analytical reverse-phase high-performance liquid chromatography
(RP-HPLC) analysis, the purity of all final N,N’-disubstituted ureas submitted to biological
characterization and reported herein equaled or exceeded 95%. Their structural identity and
1
13
19
integrity were confirmed by LC−MS as well as H-, C-, and F-NMR.
1
0 / 47

2.3. Biological Results and Discussion
Structure-activity relationship studies. All newly synthesized N,N’-disubstituted
ureas were initially evaluated by testing them in the surrogate eIF2α phosphorylation
dual-luciferase reporter (DLR) assay [9,27]. Briefly, this assay takes advantage of the fact
that activated HRI phosphorylates eIF2α which inhibits eIF2 guanine nucleotide exchange
factor eIF2B, responsible for exchanging the GDP in the eIF2.GDP complex for GTP. The
reduction in the recycling of eIF2.GDP into eIF2.GTP interferes with the formation of
eIF2·GTP·Met-tRNA ternary complex. This results in the inhibition of overall translation,
i
but paradoxically, increases translation of a small subset of mRNAs that contain multiple
upstream open reading frames (uORFs) in their 5' untranslated region (5’UTR). These
include mRNA coding for activating transcription factor 4 (ATF-4) [18,19,28]. In our
assay, firefly (F) luciferase ORF is fused to the 5'UTR of ATF-4 mRNA that has multiple
uORFs while renilla (R) luciferase ORF is fused to a 5’UTR lacking any uORFs. Agents
that reduce the amount of the eIF2·GTP·Met-tRNA ternary complex, such as those
i
N,N’-disubstituted ureas that activate HRI, increase F luciferase expression while reducing
the R luciferase expression, resulting in an increased F/R luciferase ratio. The newly
synthesized ureas were tested at 10, 5, 2.5, and 1.25 µM concentrations in 96-well assay
plates [9,29]. We calculated the activity scores as the F/R ratios for every
compound-treated well and normalized these to F/R ratio of vehicle-treated (DMSO) wells
in the same plate, arbitrarily set at 1 (F/R = 1). Data obtained in the surrogate
dual-luciferase eIF2α phosphorylation (DLR) assay for the series of
11 / 47

1-phenyl-3/4-(4-trifluorometyl)phenoxy-cycloalkyl)ureas
(4-CF -ΦOcAlkΦUs)
and
3
1-phenyl-3-(5-(4-(trifluoromethyl)phenoxy)pyrimidin-2-yl)ureas
(4-CF -ΦOPyΦUs)
3
reported herein are shown in respective Tables 1 and 2.
Table 1 (Uploaded separately)
Table 2
Substitution pattern and cLogP of the 1-phenyl-3-(5-(4-(trifluoromethyl)phenoxy-pyrimidin-2-yl)ureas
(4-CF )ΦOPyrΦUs)*
(
3
R1
H
R2
CF3
CF3
CN
F
R3
H
H
H
H
H
F
R4
CN
H
cLogP
5.27
5.80
4.37
4.52
4.97
5.12
1
2
3
4
7
-py
-py
-py
-py
-py
H
H
H
H
CN
F
CH3
H
H
1
0-py
F
H
*No activity in the surrogate dual-luciferase eIF2α phosphorylation assay up to 20 µM.
In our previous reports we have identified two 1-phenyl-3-((1,
4
-trans)-4-(4-trifluorometyl)phenoxy)cyclohexyl)ureas (4-CF -ΦOcHΦUs) 1-VI and 2-VI
3
as promising leads [11,12]. Both have a (4-trifluoromethyl)phenoxy ring and while 2-VI
has a N-(3-trifluoromethyl)phenyl 1-VI carries the N-(3-trifluormethyl-5-cyano)phenyl
moiety leading to slightly different cLogP (6.41 and 6.20, respectively). Accordingly, both
activate HRI though 1-VI is more active (3.6- and 5.7-fold increase in the surrogate eIF2α
phosphorylation assay at 1.25µM for 2-VI and 1-VI, respectively). The subsequent
1
2 / 47

focused library of 4-CF -ΦOcHΦUs reported herein was prepared to reduce
3
hydrophobicity without compromising potency. All include a (4-trifluoromethyl)phenoxy
moiety and differ in nature and position of the substituents on the N-phenyl moiety. The
most potent analog in this library, 5-VI, is almost equipotent with 1-VI (5.5- and 5.7-fold
increase at 1.25 µM, respectively) but has significantly lower cLogP than the latter
(cLogP=5.38 vs 6.20, respectively). Moreover, the 2,3-difluorophenyl carrying
4
-CF -ΦOcHΦU 11-VI has potency that falls in between the two earlier leads, 1-VI and
3
2-VI, but it is somewhat less hydrophobic than 5-VI (cLogP=5.17 vs 5.38, respectively).
Evidently, replacement of the hydrophobic CF substituent in 1-VI with F, a smaller and
3
less hydrophobic substituent (cf. 1-VI with 5-VI) results in an insignificant loss in potency
but a significant reduction in hydrophobicity. Interestingly, the least hydrophobic
4
-CF -ΦOcHΦUs in this series 6-VI and 8-VI, presenting two polar and electron
3
withdrawing
substituents
as
either
N-(2-fluoro-5-cyano)phenyl
or
N-(2-fluoro-4-cyano)phenyl, respectively, are not among the most potent analogs. We
suspect that the nature of substituents and their positions on the N-phenyl ring are more
important for high potency than the overall hydrophobicity.
Interestingly, in the 4-CF -ΦOcBΦUs series we have 3 pairs of cis and trans isomers
3
1
-IV and 1-IVc, 2-IV and 2-IVc and 3-IV and 3-IVc listed in the decreasing order of
potency but almost equipotent within the pairs (e.g causing 4.4- and 4.3-fold increase in
the F/R ratio in the surrogate eIF2α phosphorylation assay at 1.25 µM for 1-IV and 1- IVc,
respectively, and IC of 0.9 µM for both vs increasing F/R ratio 1.6- and 1.1-fold at 1.25
5
0
1
3 / 47

µM and IC =4.9 and >10 µM for 3-IV and 3-IVc, respectively) (Table 1). In line with
5
0
this
1,4-trans)-4-aminocyclohexan-1-ol and its greater metabolic stability [30] we select the
trans isomer as preferential leads for further preclinical development.
observation,
the
greater
commercial
availability
of
(
Our working hypothesis is that the cycohexyl moiety in the ΦOcHΦU series serves
as a scaffold to link between the two parts of the pharmacophore, the N-phenyl substituted
urea moiety and the substituted phenoxy moiety. As such, reducing the size of the
cycloalkyl linker will reduce the overall hydrophobicity and at the same time will also
affect the overall rigidity of the N-aryl urea. With that in mind we generated two focused
libraries
of
1-phenyl-3-((1,3-trans)-4-(4-trifluorometyl)phenoxy)cyclopentyl)ureas
(
4-CF -ΦOcPΦUs)
and
3
1
-phenyl-3-((1,3-trans)-4-(4-trifluorometyl)phenoxy)cyclobutyl)ureas (4-CF -ΦOcBΦUs).
3
Indeed, reducing the ring size of the cycloalkyl moiety was accompanied by a consistent
lowering of cLogP (e.g. 6.41, 6.29, 6.16 for the 4-CF -ΦOcHΦU – 2-VI, 4-CF -ΦOcPΦU
3
3
–
2-V and 4-CF -ΦOcBΦU – 2-IV, respectively). Remarkably, none of the analogs
3
incorporating either the cyclobutyl or the cyclopentyl moieties were more potent or even
equipotent to the corresponding cyclohexyl containing analog. Possibly, the
conformational rigidification accompanying reduction in ring size and/or compromise of
critical hydrophobic interactions enabled by the cyclohexyl ring prevent both the
4
-CF -ΦOcPΦUs and the 4-CF -ΦOcBΦUs from achieving the same effective target
3
3
complementing interactions provided by the 4-CF -ΦOcHΦUs.
3
1
4 / 47

Replacement of the (1,4-trans)-4-(4-trifluoromethyl)phenoxycyclohexyl moiety in
-CF -ΦOcHΦUs with 5-(4-trifluoromethyl)phenoxy)pyrimidin-2-yl as in
-CF -ΦOPyΦUs aimed in addition to testing another mode of reducing hydrophobicity
4
4
3
3
this time by replacing the 1,4-disubstituted cyclohexyl with 2,5-disubstituted pyrimid-2-yl
cf. e.g. cLogP of 2-VI and 2-py 6.41 and 5.80, respectively) also testing for an interesting
(
mode of global molecular rigidification. This rigidification involves the replacement of the
flexible puckering cyclohexyl ring with the aromatic planar pyrimidyl ring and
enablement of the formation of a planar conjugated pseudo six-membered ring generated
by an intramolecular hydrogen bond within the pyrimidin-2-yl-urea motif (denoted in red
in Scheme 3), which was calculated to be more stable than the extended conformation
[
31,32]. Nevertheless, due to the pairing of one hydrogen bond donor and one hydrogen
bond acceptor to pseudo six-membered ring these H-donor and -acceptors are no longer
available for intermolecular interactions with water and therefore aqueous solubility was
lower than the one anticipated from compounds of the same cLogP that do not form
intramolecular hydrogen bonds [33]. All the compounds within the focused library of
4
-CF -ΦOPyΦUs were devoid of activity in the surrogate eIF2α phosphorylation assay
3
even up to 20 µM as measured in the DLR assay. We suspect that this mode of molecular
rigidification may lock compounds in an orientation that could not accommodate the HRI
activation and therefore did not lead to the activity in the surrogate eIF2α phosphorylation
assay.
1
5 / 47

Scheme 3. Potential molecular rigidification by a putative planar pseudo six-membered ring generated
by an intramolecular hydrogen bond within the pyrimidin-2-yl-urea motif
Biological activities in secondary mechanistic assays. To validate the newly
synthesized 4-CF -ΦOcAlkΦUs and 4-CF -ΦOPyΦUs as activators of HRI, and thereby
3
3
inducers of eIF2α phosphorylation, we selected most of the 4-CF -ΦOcHΦUs (4-VI,
3
5
-VI, 6-VI, 7-VI, 11-VI and 1-VI) and the most potent 4-CF -ΦOcBΦU (1-IV) (Table 1)
3
and tested them in the secondary mechanistic assays that included phosphorylation of
endogenous eIF2α and expression of its downstream effector CHOP.
Endogenous eIF2α is the best-known substrate of HRI and the upstream regulator of
the eIF2·GTP·Met-tRNA ternary complex abundance, while CHOP expression is a
i
downstream effector of eIF2α phosphorylation. For assessing eIF2α phosphorylation we
blotted cell lysates treated for two hours with vehicle or selected compounds using
antibodies specific for the total-eIF2α and the phosphorylated-eIF2α (T-eIF2α and
51
[
PhoS ]-eIF2α), respectively. [34] (Figure 2A). The least potent analog 6-VI in
stimulating eIF2α phosphorylation as measured by the surrogate eIF2α phosphorylation
assay was also the least active one in phosphorylating endogenous eIF2α in adherent
human melanoma CRL-2813 cells as determined by Western blot analysis. Moreover, in
CRL-2813 cells treated with the N,N’-disubstituted ureas for eight hours, expression of
CHOP protein as well as expression of cell cycle regulatory proteins, the oncogenic
Kip1
protein cyclin D1 and the cyclin dependent kinase (CDK) inhibitor p27
that prevents
activation of cyclin/CDK complexed at the G1 phase of cell cycle, was very revealing. As
anticipated, relative to DMSO, the control vehicle, active 4-CF -ΦOcHΦUs (1-VI, 4-VI,
3
5
-VI, 7-VI and 11-VI) and 4-CF -ΦOcBΦUs (1-IV) increased the expression of CHOP, a
3
1
6 / 47

downstream effector of eIF2α phosphorylation, and decreased the expression of cyclin D1
whose expression is dependent on the abundance of the ternary complex [28,29] (Fig. 2B).
Consistently, 4-CF -ΦOcHΦU 6-VI, which was inactive in the DLR assay, neither
3
increased expression of CHOP and nor inhibited the expression of cyclin D1 [28,35] (Fig.
2B). As anticipated, none of the analogs affected the expression of housekeeping proteins
Kip1
such as p27
and β-actin (Fig. 2B).
In summary, the results obtained with the secondary and tertiary mechanistic assays
reported herein are in full agreement with the data from the surrogate eIF2α
phosphorylation reporter assay and thus support our conclusion that our active
4
-CF -ΦOcAlkΦUs are targeting the ternary complex by activating HRI.
3
B
Fig. 2. HRI activators induce eIF2α phosphorylation and modify its downstream
effectors. A. CRL-2813 cells were treated with 5 µM of each compound for 2 hours, cell
1
7 / 47

lysates were probed with phosphorylated (P-eIF2α) and total eIF2α (T-eIF2α) and β-actin
specific antibodies. B. CRL-2813 cells were treated with 5 µM HRI activators for 8 hours
Kip1
and lysates were immunoblotted with antibodies against CHOP, p27 , cyclin D1 and
β-actin (Also uploaded separately as a TIFF Figure).
In vitro cell proliferation studies. All 4-CF -ΦOcAlkΦUs were tested in the
3
sulforhodamine B (SRB) cell proliferation assay to establish the concentrations that inhibit
human melanoma CRL-2813 cells growth by 50% (IC ) (Table 1). This activity reflects a
5
0
combined effect of membrane permeability and overall anti-proliferative effects that
include on target as well as off-target effects. Therefore, by correlating the IC s with the
5
0
activity of the compounds in the surrogate dual-luciferase eIF2α phosphorylation reporter
DLR) assay we may gain valuable insight into compounds’ specificity and mode of
action (Table 1). Evidently, the most active analogs that are part of the
-CF -ΦOcAlkylΦU series display IC s in the sub-micromolar range (e.g. 0.46 and 0.9
(
4
3
50
µM for 5-VI and 1-IV, respectively). Moreover, there is a very good correlation between
the activity of the compounds in the surrogate eIF2α phosphorylation reporter assay and
the IC values in the cell proliferation assay (e.g. 5.7-, 3.1- and 4.4-fold at 1.25 µM and
5
0
13.8-, 8.3- and 10-fold at 5 µM in the surrogate eIF2α-P assay vs IC s of 0.35, 1.2 and
50
0.9 µM, for respective 1-VI, 1-V and 1-IV) (Table 1). Together, these results strongly
support our hypothesis that anti-proliferative activity of 4-CF -ΦOcAlkylΦUs analogs is
3
primarily contributed through the activation of HRI. This is consistent with our previously
1
8 / 47

reported 4-CF -ΦOcHΦUs 1-VI, 2-VI and 3-VI that inhibit cell proliferation by activating
3
HRI and inducing eIF2α phosphorylation [9-12].
Specificity of HRI activators. To determine the specificity of our newly synthesized
lead compounds for HRI we knockdown expression of HRI in MCF-7 cells using a pool of
siRNA targeting HRI and used cell proliferation assay as a biologic response parameter
[
9,11]. We choose this assay because it compares specificity of the compounds for HRI
compared to all other cellular targets that can impinge on cell proliferation. We treated
MCF-7 human breast cancer cells transfected with siRNA targeting HRI or vehicle with
various concentrations of 4-CF -ΦOcHΦUs (1-VI, 4-VI, and 5-VI) and 4-CF -ΦOcBΦU
3
3
(1-IV). MCF-7 cells were chosen because knockdown efficiency in these cells is
significantly higher than in CRL-2813 cells [9]. As shown in Figure 3 knocking down HRI
expression caused a dramatic reduction in the activity of all four compounds tested. These
data demonstrate that 4-CF -ΦOcHΦUs specifically activate HRI.
3
1
-VI NT
1-VI HRI
1-IV NT
1-IV HRI
4-VI NT
4-VI HRI
5-VI NT
5-VI HRI
1
25
00
1
75
50
25
0
0
0.27
0.9
3
10
-
25
Compoundconcentration (µM)
Figure 3. Inhibition of cell proliferation by lead compounds is dependent on HRI.
MCF-7 cells transfected with (solid bars) or without (hatched bars) siRNA targeting HRI
were treated with the indicated concentrations of each compound for three days starting
1
9 / 47

one day after transfection and cell proliferation was measured by sulforhodamine B assay
as previously reported [9] (Also uploaded as a TIFF file)
Conclusions
The study reported herein is a continuation of our previously published studies that
established the 1-phenyl-3-((1,4-trans)-4-phenoxycyclohexyl)ureas (ΦOcHΦUs) as a
promising scaffold for activation of HRI and as such activators of eIF2α phosphorylation
and subsequent inhibitors of translation initiation. The major objective of this study was to
further optimize the leads 1-VI and 2-VI, identified previously, by lowering their
lipophilicity and if possible enhance their potency. Removal of the CF substituent from
3
the N-phenyl in 1-VI and replacing it with a F generating N-(3-F,5-CN)phenyl and
N-(3-F,4-CN)phenyl moieties as in 4-VI and 5-VI, respectively, that were almost as potent
as the previous leads but had significantly lower lipophilicity as estimated from their
cLogP, both 5.23 (cf. 6.20 for 1-VI). However, reducing lipophilicity by replacing the
(1,4-trans)-disubstituted cyclohexyl ring with smaller cycloalkyls such as cyclobutyl and
cyclopentyl resulted in lower potency in the DLR assay and therefore did not generate any
advantageous compounds. Interestingly, replacement of the (1,4-trans)-disubstituted
cyclohexyl ring with a 2,5-disubstituted pyrimidine ring led to significantly lower cLogPs
but reduced apparent solubility and abolished potency as measured in the surrogate eIF2α
phosphorylation assay. The planarity and aromatic nature of the pyrimidine ring and the
putative extended rigidification originating from the potential formation of an
intramolecular hydrogen bond within the pyrimidin-2-yl-urea motif do not accommodate
interactions with the solvent or productive interactions with the cognate macromolecular
2
0 / 47

targets of these ligands. Taken together, these results may underscore the unique role of
the (1,4-trans)-disubstituted cyclohexyl moiety as an advantageous linker connecting the
two pharmacophore fragments, the N-phenyl-ureido and the phenoxy motif, allowing them
acquire the correct spatial orientation and providing the right molecular flexibility, bulk
and lipophilicity. Figure 4 depicts a potential mechanism of HRI activation by
4
-CF -ΦOcHΦUs.
3
In addition, this study contributes two 4-CF -ΦOcHΦUs, 4-VI and 5-VI, as promising
3
leads to the previously reported 1-VI and 2-VI and provides us with the molecular tools to
explore their pharmacokinetic and pharmacodynamic properties in animals and as drug
candidates in animal model of diseases such as cancer, neurodegenerative disorders and
β-thalassemia. Finally, when employed in combination with other eIF2α kinase activators,
these compounds are invaluable tools for dissecting contribution of eIF2α vs other
substrates specific to each eIF2α kinase to normal- and patho-biology of those kinases.
2
1 / 47

Figure 4. A hypothetical model of HRI activation by 4-CF -ΦOcHΦUs. HRI binding of
3
heme or cellular chaperones such as HSP-90 and/or other endogenous inhibitors cause the
N-terminal domain (NTD) to interact with kinase domain such that kinase domain is
inactive cannot bind the eIF2α. The 4-CF -ΦOcHΦUs displace such inhibitors releasing
3
the NTD from kinase domain which results a series of auto-phosphorylation events that
changes the relative orientation of N-lobe and C-lobe of kinase domain rendering the
substrate binding domain accessible to eIF2α binding and catalysis (also submitted
separately in TIFF format)
4. Experimental
4.1. Materials and methods
4.1.1. General methods
All reagents and solvents were purchased from commercial sources and used as is.
Analytical HPLC was run on a Waters Alliance 2695 using a reverse-phase column
XBridge BEH130 C18, 4.6 × 100 mm, particle size 5 µm) eluting with a linear gradient
(
of acetonitrile in water containing 0.1% trifluoracetic acid (TFA). The purity of all target
compounds was greater than 95% by analytical HPLC inspection. LC-MS analysis was
run on a Waters Alliance 2695 with UV detector (214 and 254 nm) and Micromass ZQ
2
2 / 47

+
quadrupole mass detector in electrospray positive (ESI ) mode using a reverse-phase
column (Waters Symmetry C18, 2.1 × 100 mm, particle size 3.5 µm) eluting with a linear
gradient of acetonitrile in water containing 0.1% formic acid. TLC analysis was run on
Merck silica gel 60 F254 aluminum sheets. Flash chromatography purifications were
performed on Biotage SP1 using silica gel prepacked normal phase columns (200−400
mesh) eluting with a linear gradient of ethyl acetate in n-heptane, and fractions were
collected at 254 nm and monitored at 280 nm. Melting points were determined on a
Mel-Temp electrothermal apparatus equipped with a Barnaand thermometer and were
uncorrected. Proton, carbon, and fluorine NMR experiments were performed on a Varian
Inova 400 MHz spectrometer using DMSO-d as solvent. Chemical shifts (δ) are reported
6
in ppm relative to TMS as the internal standard.
4.2. Chemistry
The synthesis, purification and characterization of 1-VI, 2-VI and 3-VI were previously
reported by us [11,12] and they are brought here for reference only.
4.2.1. tert-Butyl (1S,3S)-3-(4-(trifluoromethyl)phenoxy)cyclopentylcarbamate (A-V)
To a solution of tert-butyl (1S,3R)-3-hydroxycyclopentylcarbamate (0.50 g, 2.5
mmol) ,4-(trifluoromethyl)phenol (0.49 g, 3 mmol) and triphenylphosphine (0.98 g, 3.75
mmol) in anhydrous THF (8 mL) was dropwise added DIAD (0.98 mL, 4.68 mmol) in
anhydrous THF (7 mL). The reaction solution was stirred at room temperature overnight.
The resulting solution was concentrated in vacuo, and the residue was subjected to flash
column chromatography on silica gel (eluting with EtOAc-heptane by a gradient of EtOAc
from 4% to 32%) to deliver title compound as colorless crystals (0.63g, 73%); mp
1
1
8
2
1
29.3-132.1 ºC. H NMR (400 MHz, DMSO-d ) δ 7.60 (d, J = 8.7 Hz, 2H), 7.04 (d, J =
6
.7 Hz, 2H), 6.93 (d, J = 6.8 Hz, 1H), 4.96 – 4.86 (m, 1H), 3.95 (dq, J = 13.8, 7.0 Hz, 1H),
.15 (ddd, J = 20.2, 8.8, 6.2 Hz, 1H), 1.94 (td, J = 13.9, 7.3 Hz, 2H), 1.86 – 1.76 (m, 1H),
1
3
.64 (ddd, J = 10.2, 8.6, 2.4 Hz, 1H), 1.52 – 1.39 (m, 1H), 1.36 (s, 9H). C NMR (100
2
3 / 47

MHz, DMSO-d ) δ 160.78 (d, J = 1.1 Hz), 155.51, 127.34 (q, J = 3.7 Hz), 125.03 (q, J =
6
1
9
2
70.8 Hz), 121.24 (q, J = 32.1 Hz), 116.10, 78.16, 50.48, 30.85, 30.64, 28.70. F NMR
(
376 MHz, DMSO-d ) δ -59.83.
6
4.2.2. tert-Butyl trans-3-(4-(trifluoromethyl)phenoxy)cyclobutylcarbamate (A-IV)
A-IV was prepared in essentially the same way as A-V. White solid (0.78 g, 94%); mp
1
1
37.8-141.8 ºC. H NMR (400 MHz, DMSO-d ) δ 7.61 (d, J = 6.8 Hz, 2H), 7.28 (d, J = 5.0
6
Hz, 1H), 6.96 (d, J = 6.9 Hz, 2H), 4.86 (d, J = 2.7 Hz, 1H), 4.08 (d, J = 4.6 Hz, 1H), 2.41
1
3
–
1
7
2.22 (m, 4H), 1.36 (d, J = 2.4 Hz, 9H). C NMR (100 MHz, DMSO-d ) δ 160.50,
6
55.21, 127.42 (q, J = 3.7 Hz), 124.99 (q, J = 271.0 Hz), 121.57 (q, J = 32.2 Hz), 115.71,
1
9
0.04, 42.05, 36.89, 28.66. F NMR (376 MHz, DMSO-d ) δ -59.03.
6
4.2.3. tert-Butyl cis-3-(4-(trifluoromethyl)phenoxy)cyclobutylcarbamate (A-IVc)
A-IVc was prepared in essentially the same way as A-V. White solid (0.45 g, 27%); mp
1
1
57.0-159.0 ºC (CH Cl ). H NMR (400 MHz, DMSO-d ) δ 7.60 (d, J = 7.6 Hz, 2H), 7.18
2
2
6
(
d, J = 7.2 Hz, 1H), 7.00 (d, J = 7.8 Hz, 2H), 4.52 – 4.30 (m, 1H), 3.69 (dd, J = 15.7, 8.0
1
3
Hz, 1H), 2.76 (d, J = 6.3 Hz, 2H), 1.97 (d, J = 8.2 Hz, 2H), 1.36 (s, 9H). C NMR (100
MHz, DMSO-d ) δ 160.38 (d, J = 1.3 Hz), 155.02, 127.42 (q, J = 3.7 Hz), 124.98 (q, J =
6
1
9
2
70.6 Hz), 121.59 (q, J = 32.1 Hz), 115.64, 65.68, 38.51, 37.65, 28.66. F NMR (376
MHz, DMSO-d ) δ -59.11.
6
4.2.4. (1S,3S)-3-(4-(Trifluoromethyl)phenoxy)cyclopentanamine (B-V)
A-V (518 mg, 1.5 mmol) was dissolved in anhydrous methylene chloride (10 mL), and a
solution of trifluoroacetic acid (5.31 g, 45 mmol) in anhydrous methylene chloride (5 mL)
was added. The reaction solution was stirred at room temperature for 1h. The resulting
solution was rotavapored, and the residue was diluted in methylene chloride (60 mL),
washed with saturated aqueous sodium bicarbonate (2×20 mL), dried over anhydrous
sodium sulfate, and evaporated in vacuo to deliver title compound as semi-solid (380 mg,
1
1
03%) that was used in the next step without further purification. H NMR (400 MHz,
DMSO-d ) δ 7.59 (d, J = 8.7 Hz, 2H), 7.02 (d, J = 8.6 Hz, 2H), 4.93 (m, 2H), 3.41 (m, 1H),
6
2
1
.22 (m, 1H), 2.01 – 1.81 (m, 2H), 1.76 – 1.66 (m, 1H), 1.66 – 1.55 (m, 1H), 1.29 (m, 1H),
1
9
.39 – 1.17 (m, 1H). F NMR (376 MHz, DMSO-d ) δ -59.77.
6
4.2.5. trans-3-(4-(Trifluoromethyl)phenoxy)cyclobutanamine (B-IV)
2
4 / 47

B-IV was prepared in essentially the same way as B-V. Yellow semi-solid (0.49 g, 100%).
1
H NMR (400 MHz, DMSO-d ) δ 7.60 (d, J = 8.7 Hz, 2H), 6.95 (d, J = 8.6 Hz, 2H), 4.99
6
–
1
1
4.77 (m, 2H), 3.70 – 3.45 (m, 1H), 2.25 (ddd, J = 11.9, 7.5, 4.1 Hz, 2H), 2.14 (dt, J =
1
3
2.3, 6.1 Hz, 2H). C NMR (100 MHz, DMSO-d ) δ 160.70, 127.39 (q, J = 3.8 Hz),
6
19
25.02 (q, J = 271.0 Hz), 121.37 (q, J = 31.7 Hz), 115.70, 70.57, 43.93, 39.61. F NMR
(
376 MHz, DMSO-d ) δ -59.84.
6
4.2.6. cis-3-(4-(Trifluoromethyl)phenoxy)cyclobutan-1-amine (B-IVc)
B-IVc was prepared in essentially the same way as B-V. Off white solid (0.28 g, 100%);
mp 71.9-74.3 ºC.
4
.2.7. 5-(4-(Trifluoromethyl)phenoxy)pyrimidin-2-amine (B-py)
mixture of 2-aminopyrimidin-5-ol (0.5
-fluoro-4-(trifluoromethyl)benzene (0.74 g, 4.5 mmol) and cesium carbonate (4.4 g, 13.5
A
g,
4.5
mmol),
1
mmol) in DMSO (8 mL) were stirred under argon at 80 ºC for 5h. The resulting reaction
was diluted with ethyl acetate (90 mL) and water (30 mL). Aqueous phase was
back-extracted with ethyl acetate (60 mL). Combined extract was washed with brine
(5×30 mL), dried over anhydrous sodium sulfate overnight, and evaporated in vacuo. The
residue was recrystallized from hot ethyl acetate to deliver title compound as white
1
crystalline solid (0.12 g, 10%); mp 132.5-134.5 ºC (EtOAc). H NMR (400 MHz,
DMSO-d ) δ 8.20 (s, 2H), 7.68 (d, J = 8.7 Hz, 2H), 7.10 (d, J = 8.6 Hz, 2H), 6.73 (s, 2H).
6
1
3
C NMR (100 MHz, DMSO-d ) δ 161.95 (d, J = 1.3 Hz), 161.85, 151.94, 141.27, 127.85
6
1
9
(q, J = 3.8 Hz), 124.73 (q, J = 271.2 Hz), 123.38 (q, J = 32.2 Hz), 116.78. F NMR (376
MHz, DMSO-d ) δ -60.15.
6
4.2.8. 1-(3-Cyano-5-fluorophenyl)-3-(trans-4-(4-(trifluoromethyl)phenoxy)cyclohexyl)urea
(4-VI)
3
-Amino-5-fluorobenzonitrile (0.41 g, 3 mmol) and anhydrous pyridine (0.36 g, 4.5 mmol)
were dissolved in anhydrous methylene chloride (45 mL). Phenyl chloroformate (0.70 g,
.5 mmol) was added dropwise at 0 ºC. The reaction was stirred under argon at room
4
temperature for 4 h. The resulting solution was washed with consecutively with 1 mol/L
aqueous hydrochloride, water and brine, dried over anhydrous sodium sulfate, and
evaporated in vacuo to deliver phenyl (3-cyano-5-fluorophenyl)carbamate (E-4) as white
solid in quantitative yield and used as it without further purification.
Above
obtained
E-4
(0.13
g,
0.5
mmol)
and
2
5 / 47