Shooting three inflammatory targets with a single bullet: Novel multi-targeting anti-inflammatory glitazones

Article abstract of DOI:10.1016/j.ejmech.2019.02.034

In search for effective multi-targeting drug ligands (MTDLs) to address low-grade inflammatory changes of metabolic disorders, we rationally designed some novel glitazones-like compounds. This was achieved by incorporating prominent pharmacophoric motifs

Full text of DOI:10.1016/j.ejmech.2019.02.034

Accepted Manuscript
Shooting three inflammatory targets with a single bullet: Novel multi-targeting anti-
inflammatory glitazones
Perihan A. Elzahhar, Rana Alaaeddine, Tamer M. Ibrahim, Rasha Nassra, Azza
Ismail, Benjamin S.K. Chua, Rebecca L. Frkic, John B. Bruning, Nadja Wallner, Tilo
Knape, Andreas von Knethen, Hala Labib, Ahmed F. El-Yazbi, Ahmed S.F. Belal
PII:
S0223-5234(19)30148-5
DOI:
https://doi.org/10.1016/j.ejmech.2019.02.034
EJMECH 11126
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 10 November 2018
Revised Date: 4 February 2019
Accepted Date: 10 February 2019
Please cite this article as: P.A. Elzahhar, R. Alaaeddine, T.M. Ibrahim, R. Nassra, A. Ismail, B.S.K.
Chua, R.L. Frkic, J.B. Bruning, N. Wallner, T. Knape, A. von Knethen, H. Labib, A.F. El-Yazbi,
A.S.F. Belal, Shooting three inflammatory targets with a single bullet: Novel multi-targeting anti-
inflammatory glitazones, European Journal of Medicinal Chemistry (2019), doi: https://doi.org/10.1016/
j.ejmech.2019.02.034.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Graphical Abstract:
Prototypical balanced modulation of 3 inflammatory targets offered by some novel
glitazone-like compounds.

ACCEPTED MANUSCRIPT
Shooting Three Inflammatory Targets with a Single Bullet: Novel
Multi-Targeting Anti-Inflammatory Glitazones
a
b
c
d
Perihan A. Elzahhar , Rana Alaaeddine , Tamer M. Ibrahim , Rasha Nassra , Azza
a
e
f
f
g
Ismail , Benjamin S.K. Chua , Rebecca L. Frkic , John B. Bruning , Nadja Wallner ,
Tilo Knape , Andreas von Knethen , Hala Labib , Ahmed F . El-Yazbi , Ahmed S.
g
g,h
a,i
b,j*
a*
F. Belal
a
Department of Pharmaceutical Chemistry, Faculty of Pharmacy, Alexandria
University, Alexandria 21521, Egypt.
b
Department of Pharmacology and Toxicology, Faculty of Medicine and Medical
Centre, American University of Beirut, Beirut, Lebanon.
c
Department of Pharmaceutical Chemistry, Faculty of Pharmacy, Kafr-Elsheikh
University, Kafr El-Sheikh 33516, Egypt.
d
Department of Medical Biochemistry, Faculty of Medicine, Alexandria University,
Egypt.
e
School of Biological Sciences, The University of Adelaide, Adelaide, South
Australia 5005, Australia.
f
Institute for Photonics and Advanced Sensing, The School of Biological Sciences, the
University of Adelaide, North Tce, Adelaide, South Australia 5005, Australia.
g
Fraunhofer Institute for Molecular Biology and Applied Ecology IME, Project Group
Translational Medicine & Pharmacology TMP, Theodor-Stern-Kai 7, 60596
Frankfurt, Germany.
h
Institute of Biochemistry I, Faculty of Medicine, Goethe-University Frankfurt,
Theodor-Stern-Kai 7, 60596 Frankfurt, Germany.
i
Department of Pharmaceutical Chemistry, College of Pharmacy, Arab Academy for
Science, Technology and Maritime Transport, Alexandria, Egypt.
j
Department of Pharmacology and Toxicology, Faculty of Pharmacy, Alexandria
University, Alexandria 21521, Egypt.
1

ACCEPTED MANUSCRIPT
ABSTRACT:
In search for effective multi-targeting drug ligands (MTDLs) to address low-grade
inflammatory changes of metabolic disorders, we rationally designed some novel
glitazones-like compounds. This was achieved by incorporating prominent
pharmacophoric motifs from previously reported COX-2, 15-LOX and PPARγ
ligands. Challenging our design with pre-synthetic docking experiments on PPARγ
showed encouraging results. In vitro tests have identified 4 compounds as
simultaneous partial PPARγ agonist, potent COX-2 antagonist (nanomolar IC50
values) and moderate 15-LOX inhibitor (micromolar IC50 values). We envisioned
such outcome as a prototypical balanced modulation of the 3 inflammatory targets. In
vitro glucose uptake assay defined six compounds as insulin-sensitive and the other
two as insulin-independent glucose uptake enhancers. Also, they were able to induce
PPARγ nuclear translocation in immunohistochemical analysis. Their anti-
inflammatory potential has been translated to effective inhibition of monocyte to
macrophage differentiation, suppression of LPS-induced inflammatory cytokine
production in macrophages, as well as significant in vivo anti-inflammatory activity.
Ligand co-crystallized PPARγ X-ray of one of MTDLs has identified new clues that
could serve as structural basis for its partial agonism. Docking of the most active
compounds into COX-2 and 15-LOX active sites, pinpointed favorable binding
patterns, similar to those of the co-crystallized ligands. Finally, in silico assessment of
pharmacokinetics, physicochemical properties, drug-likeness and ligand efficiency
indices was performed. Hence, we anticipate that the prominent biological profile of
such series will rationalize relevant anti-inflammatory drug development endeavors.
*
Corresponding Authors:
For AFE: Phone: +961-1-350 000, e-mail: ae88@aub.edu.lb,
ORCID ID: 0000-0003-3432-3038.
For ASFB: Phone: +2-03-4871317, e-mail: ahmed.belal@alexu.edu.eg,
ORCID ID: 0000-0001-9463-0844.
Keywords: Inflammation; Cyclooxygenase-2; 15-Lipoxygenase; Peroxisome
proliferator-activated receptor-γ; Multi-targeting.
2

ACCEPTED MANUSCRIPT
1
. Introduction:
Inflammation is a complex and dynamic response to defend the host against potential
threats such as pathogens or tissue insults [1,2]. It is associated with diverse signaling
pathways involving enzymes, membrane and cellular receptors, transcription factors
and others, operating in an orchestrated manner to counteract the imminent threat
[
2,3]. Neutrophils and tissue macrophages represent the first line of defense, where
they stimulate the production of a plethora of inflammatory mediators as a result of
sensing an initial insult [4,5]. Although inflammation is intended to be protective,
exaggerated and uncontrollable neutrophil and/or macrophages activity, in addition to
failure to achieve resolution often leads to persistent tissue damage, a condition
underlying many chronic inflammatory disorders (such as psoriasis, multiple
sclerosis, and rheumatoid arthritis) [1].
On the other hand, diabetes is perceived as a multifaceted disease implicating
metabolic, cardiovascular, and immune component events; though an understanding
of the underlying relating mechanisms are not clear [6,7]. Recent results from our
group implicated low-grade adipose inflammation as potential underlying cause of
cardiovascular and cardiac autonomic dysfunction associated with metabolic
challenge [8]. Indeed, reversal of adipose inflammation was associated with an
improved vascular and cardiac autonomic function [8].
Among the different pathways involved in the inflammatory response, arachidonic
acid (AA) plays a major role as the biological precursor for many inflammatory
mediators. Two key metabolic enzymes are involved in generating AA derivatives,
namely: cyclooxygenase (COX) and lipoxygenase (LOX) [9]. COX-2 expression is
induced by pro-inflammatory stimuli leading to the generation of inflammatory
signaling prostaglandins (PGs) [9,10]. On the other hand, LOXs catalyze the
oxygenation of polyunsaturated fatty acids (PUFAs) such as arachidonic and linoleic
acids to give oxygenated lipid mediators that are involved in important cellular
3

ACCEPTED MANUSCRIPT
signaling mechanisms [11]. Based on the site of oxygenation of arachidonic acid,
several mammalian LOX isoforms were discovered and denoted 5-, 12- and 15-LOX
[
11].
respiratory disorders through the production of 5-hydroperoxy-eicosatetraenoic acid
5-HPETE) which is subsequently converted to leukotrienes (LTs) [11].
5-LOX is famous for playing an important role in the pathology of
(
Contrarily, 15-LOX converts polyunsaturated fatty acids into 15-S-HPETE which is
metabolized into either lipoxins or eoxins [11–13]. It is necessary to mention that 5-
LOX also contributes to the synthesis of lipoxins [11–13]. While lipoxins are pro-
resolving and anti-inflammatory [11–13], eoxins (also referred to as 14,15-LTs) have
shown to be pro-inflammatory mediators specially in the context of atherosclerosis,
adipocyte differentiation and development of obesity [12,14]. Also, eoxins were
shown to have similar actions as leukotrienes, and hence, the probability of their
implications in inflammatory and respiratory diseases could not be excluded [12,14].
Moreover, 15-LOX is hypothesized to initiate and/or promote atherosclerosis through
LDL oxidation [12,14]. It is not surprising that IL-4 and/or IL-13 induce the
expression of 15-LOX in cultured mast cells, monocytes, and epithelial cells [12,14].
Thus, 15-LOX can be regarded as a double-edged weapon that exert both pro- and
anti-inflammatory properties and the details of the interplay between them in
physiological and pathophysiological contexts are still lacking [12,14]. Up till this
moment, various reports indicated that 15-LOX-inhibitors (such as PD146176 among
others) could serve as effective therapeutics for a number of 15-LOX-related diseases,
which include diabetes, hypertension, obesity, atherosclerosis and other disorders
linked to chronic adipose tissue inflammation [14].
The notion of using selective COX-2 inhibitors as anti-inflammatories with minimal
gastrointestinal (GI) side effects was very resonating in the late 1990s and early 2000s
[
15,16]. However, it turned out that blocking only COX-2 would shunt the
inflammatory pathways towards more activity of LOX and hence blocking both COX-
2
1
and LOX would provide safer and more effective anti-inflammatory activity [17–
9].
Interestingly, in vitro studies demonstrated that LOX modulators can activate
peroxisome proliferator-activated receptors (PPARs) [20]. Indeed, recent studies in
4

ACCEPTED MANUSCRIPT
vivo implicated a role for interplay between LOX-derived AA metabolites and PPARγ
in mitigating post-ischemic injury in the brain [21]. PPARs are subclasses of the
intracellular receptors family, that when activated act as nuclear hormones modulating
the transcription of plenty of factors involved in many physiological processes [22–
2
4]. PPARγ activity was first thought to be limited to lipid metabolism and glucose
homeostasis [22–24]. However, there is ample evidence demonstrating the role of
activated PPARγ in reducing the expression of pro-inflammatory mediators such as
cytokines and adhesion molecules involved in neutrophils recruitment to
inflammatory foci [23,25]. Moreover, PPARγ agonists were reported to be capable of
inhibiting the mRNA expression of COX-2 and nitric oxide synthase (NOS), which
further emphasizes the essential role that PPARγ plays in inflammation [23,25,26].
Significantly, clinical trials examining the use of pioglitazone for managing psoriasis,
multiple sclerosis, and rheumatoid arthritis are currently ongoing [19].
The last decade has witnessed a paradigm shift in drug discovery programs from
designing highly selective single-target ligands to modulating a multiplicity of targets
within related signaling networks [27]. Hence, in seeking multi-targeting anti-
inflammatory leads with potential future applications in metabolic disorders, we
herein report on the development of the first-in-class MTDLs as potential
simultaneous COX-2/LOX inhibitors and PPARγ agonists. As a proof of concept, the
designed compounds were challenged with different biological assays and in silico
studies. We believe that the outcomes of this study might aid in expanding the
knowledge on the targeting of inflammatory reactions as a valid approach of
managing metabolic disorders.
2
2
. Results and discussion:
.1.Rationale for the design of multi-target directed ligands (MTDLs):
Thiazolidine-2,4-diones (TZDs)/rhodanines constitute diverse scaffolds that are
known to exert a vast array of pleiotropic benefits, including insulin-sensitizing, anti-
inflammatory and antiproliferative activities [28–30]. They were also reported to
prevent the progression of atherosclerotic lesions in patients with type 2 diabetes [31].
Over and above, telmisartan-rosiglitazone hybrid molecules were evaluated as dual
5

ACCEPTED MANUSCRIPT
PPARγ agonists/angiotensin II antagonists for management of metabolic syndrome
[
32]. In particular, TZDs, known as high-affinity PPARγ agonists, have been
introduced into clinical practice for the treatment of type 2 diabetes [33].
Thus, in our search for multi-target anti-inflammatory leads with potential efficacy in
metabolic disorders, and based on the structural features of four pleiotropic lead
compounds; LYSO-7 (A) [34,35], pioglitazone (B), a reported LOX-inhibiting
thiazolidinedione (C) [36] and a dual COX-2 / 15-LOX inhibitor (D) [37] recently
reported by our lab, we adopted a pharmacophoric molecular hybridization approach
to design and synthesize some novel triazolyl-thiazolidinedione/rhodanine hybrids as
potential simultaneous COX-2/LOX inhibitors and PPARγ agonists. It is worth
mentioning that developing a small molecule modulating these 3 targets was not
encountered before.
Over and above, the designed assembly has its pharmacophoric parts rooted from
pioglitazone. It comprises the essential pharmacophore TZD or its rhodanine
bioisostere. This pharmacophore is decorated with a methoxybenzylidene group like
compounds A and C (Figure 1), which in turn is connected to the 1,2,3-triazole
moiety. This should retain the 3-carbon atom spacer between the oxygen of the
phenoxy- linker and nitrogen atom of the pyridine ring (as in pioglitazone) or N-1 of
the triazole nucleus (in our designed compounds). Substitution pattern at N-1 of the
triazole ring with various aryl and aralkyl groups is attempted to grant variable
electronic and lipophilic environments to investigate their effect on the anticipated
biological activity and with the hope of achieving better binding interactions with the
potential biological targets.
6

ACCEPTED MANUSCRIPT
Figure 1. Rationale for the design of the target compounds
In addition, our design agrees with the three-module framework proposed by Pirat et
al. to represent PPARγ agonist glitazones [38]. It is composed of an acidic polar head
group, TZD or rhodanine, joined to a hydrophobic tail through an aromatic linker [38].
We altered the tail group to afford chemical and biological uniqueness while
conserving the polar head and the linker regions [38]. Our tail moiety was carefully
chosen to be 1,4-disubstituted-1,2,3-triazole which is a pharmacophore we recently
reported to show dual COX-2 and LOX inhibitory activity (Structure D, Figure 1)
[37]. Furthermore, target compounds were mapped out with a benzylidene attachment
between the polar head and the linker instead of the usual benzyl, which is one
distinct difference from pioglitazone (Figure 1). We thought that this would be more
convenient since recent reports indicated benzylidene TZD derivatives are endowed
with partial PPARγ agonistic properties, which retain the desired pharmacological
actions while avoiding the untoward adverse effects [39,40].
7

ACCEPTED MANUSCRIPT
To the best of our knowledge, the click chemistry-derived 1,4-disubstituted-1,2,3-
triazoles have not been explored before for their synchronous PPARγ agonistic and
COX-2/LOX inhibitory activities. Together with their wide-ranging biological
activities, bioorthogonal nature, being non-classical bioisosteres of the amide group
and their propensity to be involved in hydrogen bond and π-π stacking interactions
with potential targets [41], it would be tempting to employ this easily accessible
structural motif into the aforementioned framework.
2
.2.Pre-synthetic PPARγ docking study:
To gain substantial theoretical evidence into the potential PPARγ agonistic activity at
the molecular level, molecular docking studies were pre-synthetically applied on 18
compounds that contained our proposed design. Our selection of compounds for
PPARγ crystallization and structure determination for in silico validation was based
on its upstream occurrence in inflammatory pathways (since it is a nuclear receptor),
plus its implications in lipid metabolism and glucose homeostasis.
Regarding the chemical structures of the docked compounds, the reader is advised to
consult Scheme 1 using the appropriate coding. Based on literature, the screening
performance of PPARγ DEKOIS 2.0 benchmark set employing various docking tools
(
e.g., GOLD [ChemPLP], Glide [SP] and AutoDock Vina) showed comparable and
nonrandom screening performance with all three docking tools [42]. This led us to
employ GOLD (ChemPLP) [43,44] as scoring function for pose prediction purposes.
Additionally, we employed two additional scoring functions, AutoDock Vina [45] and
AutoDock VinaXB [46], for extracting consensus ranking of our proposed
compounds.
The crystal structure of PPARγ co-crystallized with rosiglitazone (PDB ID: 2PRG)
was used as a model to gauge the binding affinities of the test compounds with the
enzyme. The scoring function, hydrogen bonds formed with key amino acids and the
relative orientation of the docked compounds with respect to the co-crystallized ligand
rosiglitazone were used to estimate the binding propensities to the ligand binding
domain (LBD) of PPARγ. Pose-retrieval experiment of the co-crystallized ligand
reproduced the key interaction pattern in the binding site of PPARγ with acceptable
8

ACCEPTED MANUSCRIPT
RMSD values (e.g., for GOLD docking RMSD = 1.25 Å, Supplementary
Information Figure SM1).
Examination of the predicted docking scores of the test compounds using GOLD, as
indicated in Supplementary Information Table SM1, revealed that four compounds
were superior to rosiglitazone while other compounds showed slightly lower and
comparable scores. According to AutoDock Vina and AutoDock VinaXB, the
predicted docking scores (Table SM1) almost showed no difference, which excludes
halogen-bonding as a possible type of intermolecular interaction between these poses
and the backbone of the protein. In addition, all docked compounds showed higher
predicted docking scores than the co-crystallized ligand.
The LBD of PPARγ consists of 13 alpha helices and a small beta sheet. Activation
function 2 (AF2) is the part of the LBD that is important for ligand dependent
coactivator binding and is formed by helices 3-5 and helix 12 [22]. Ser289, His323,
His449 and Tyr473 are the most important conserved amino acids that are involved in
target recognition of PPARγ ligands [22]. Of particular interest, the formation of a
hydrogen bond between a ligand and Tyr473 triggers conformational change in the
AF2 site, which enhances the recruitment of coregulatory factors to regulate gene
expression [22].
Most of our compounds reproduced the key hydrogen bonding interactions of the co-
crystallized ligand (rosiglitazone) via the TZD ring with the residues His449, Gln289,
Tyr473, His323 and Ser289, as shown in Figure 2A and exemplified by compounds
5
e and 5j. Moreover, some compounds showed extra binding interactions outside AF2
site as in compound 5j that formed a hydrogen bond with Ser342.
It is worth mentioning that compounds 5b and 5g displayed comparable interaction
pattern to rosiglitazone in the binding site of PPARγ via their TZD rings. As well,
they showed additional binding interactions with Arg288 and Met264 with the
bromophenyl group of 5b and middle phenoxy group of 5g, respectively (Figure 2B).
Regarding the relative orientation of the compounds in the active site of PPARγ with
respect to the co-crystallized ligand, it is clearly evident that the compounds were
perfectly accommodated by the LBD and occupied the same position and spatial area
9

ACCEPTED MANUSCRIPT
as rosiglitazone (as observed with 5e and 5j) (Figure 2A). Guided by the above-
mentioned data, it was encouraging to synthesize these proposed structures.
A
B
A
Figure 2. (A) Overlay of the best-scored pose of 5j (purple) and 5e (gold) on the co-
crystallized ligand (cyan) in the binding site of PPARγ crystal structure (PDB: 2PRG).
(B) Overlay of the best-scored pose of 5b (blue) on the pose of 5g (olive) in the
binding site of the PPARγ crystal structure. The yellow dashed-lines represent the
polar contacts (H-bonding interactions). Non-polar hydrogen atoms were omitted for
clarity.
2
.3. Chemistry:
The synthetic strategies for the synthesis of intermediate and target compounds are
outlined in Scheme 1. The key intermediate 4-propargyloxybenzaldeyhde (2) was
obtained by the alkylation of 4-hydroxy benzaldehyde with propargyl bromide in the
presence of anhydrous potassium carbonate in refluxing dry acetone. Azide-alkyne
cycloaddition of compound 2 and the appropriate aromatic azides in the presence of
catalytic amount of copper sulfate pentahydrate and sodium ascorbate in DMF/H O
2
afforded the desired regioselective 1,4-disubstituted 1,2,3-triazoles (3a-h) in good
yields comparable to reported ones. Knoevenagel condensation of compounds (3a-h)
with thiazolidene-2,4-dione or rhodanine in the presence of catalytic amount of
piperidine generated the corresponding arylidenes 5a-h and 6a-h, respectively. The
1
H NMR of 5,6 (a-h) showed triazole C -H aromatic singlet around 8.32-9.15 ppm,
5
1
3
along with the characteristic benzylidene CH around 7.58-7.77 ppm. Furthermore, C
NMR of thiazolidene-2,4-diones (5a-h) displayed 2 C=O characteristic peaks at
1
3
~
168.5 and 160.1 ppm, and benzylidene CH around 143-145 ppm. Additionally, C
10

ACCEPTED MANUSCRIPT
NMR of rhodanines (6a-h) displayed a C=S characteristic peak at ~191.8-196.1 ppm,
a C=O characteristic peak at 166.8-173.7 ppm, and benzylidene CH around 143-145
ppm.
Attempts to produce arylidene triazoles 5i-j and 6i-j (from phenacyl azides) using the
aforementioned method were fruitless. Hence, we decided to adopt an alternative
sequence; starting with Knoevenagel condensation to produce arylidenes 4a,b, which
were then subjected to CuAAC to produce the desired compounds 5i-j and 6i-j. It is
noteworthy to indicate that the latter approach enabled us to screen for the biological
1
activity of arylidenes 4a,b as well. The H NMR of 4 (a,b) showed a propargylic
characteristic triplet in the range of
δ 3.61-3.63 ppm, doublet in the range of δ 4.88-
4
.91 ppm, that corresponds to terminal CH and CH , respectively. Both protons
2
13
underwent a long range coupling of ~2.32 Hz. The C NMR of 4 (a,b) showed
propargylic characteristic peaks at 56.2, 79.1 and 79.2 ppm corresponding to CH2,
δ
CH and quaternary carbon, respectively. Other characteristic peaks appeared at their
expected chemical shifts such as C=O and C=S. The IR spectra of 4 (a,b) displayed
-1
characteristic sharp acetylenic-CH stretching in the range of 3275-3255 cm and C≡C
-1
1
stretching in the range of 2376 cm . The H NMR of 5,6 (i,j) showed triazole C
5
-H
aromatic singlet at 8.24 ppm, along with the disappearance of propargylic terminal
1
3
4 5
CH. Furthermore, C NMR of 5,6 (i,j) displayed the triazole C and C peaks, along
with the disappearance of previously mentioned propargylic CH and quaternary
carbon peaks. Moreover the IR spectra of 5,6 (i,j) were associated with the
disappearance of ethynyl CH and C≡C stretching bands, clearly confirming the
formation of the triazole products.
11

ACCEPTED MANUSCRIPT
Scheme 1. Synthesis of (Z)-5-(4-((1-(aryl or aralkyl)-1H-1,2,3-triazol-4-
yl)methoxy)benzylidene)thiazolidine-2,4-diones
thioxothiazolidin-4-one derivatives (compounds 6a-j). Reagents and conditions:
Propargyl bromide, K CO
CuSO .5H O (5 mole %), Sodium Ascorbate (20 mole %), DMF/ H
overnight. iii) Piperidine (10 mole%), EtOH, reflux for 12 h.
(compounds
5a-j)
or
2-
i)
2
3
, Acetone, reflux for 2.5 h. ii) Appropriate azide,
O, stirring
4
2
2
2
2
.4.Biological evaluation:
.4.1. In vitro COX-1 and COX-2 inhibitory assay:
All synthesized compounds were subjected to an in vitro COX-1/COX-2 inhibition
assay using an ovine COX-1/human recombinant COX-2 assay kit (Catalog no.
5
60131; Cayman Chemicals Inc. Ann Arbor, MI, USA).
The half maximal inhibitor concentrations (IC50 µM) were determined and the
selectivity index (SI) values were calculated as IC (COX-1)/ IC (COX-2).
5
0
50
As summarized in Table 1; all synthesized compounds, with the exception of
compound 5f, were more potent COX-2 inhibitors than the two reference drugs
diclofenac and indomethacin. They also showed higher SI than two reference drugs
12

ACCEPTED MANUSCRIPT
diclofenac and indomethacin. Although they showed lower COX-2 inhibition in
comparison to celecoxib, they operated within the same range of submicromolar
activity.
All compounds inhibited COX-1 in concentrations that are at least one order of
magnitude higher than that experienced with COX-2, which highlights the selectivity
displayed towards COX-2. Although SI values were lower than that of celecoxib, yet
it could be viewed as an advantage by potentially avoiding the cardiovascular side
effects of highly selective COX-2 inhibitors [47].
For triazoles prepared from benzyl azides, we observed that the addition of 4-bromo
substitution enhanced both COX-2 inhibitory activity and SI values when compared to
the unsubstituted analogs. The thiazolidine-2,4-dione derivative 5b was slightly more
potent than the rhodanine 5a. It reached about 8 times the activity of diclofenac and 5
times that of indomethacin. Intriguingly, 5b showed the highest SI value (159) out of
all the synthesized compounds in the study, which accounts for almost one half that of
celecoxib activity and selectivity towards COX-2.
Within the triazole series prepared from aryl azides, it was noticed that 4-substitution
with lipophilic π-deficient or π-rich substituents as was seen with 4-bromo (5e and 6e)
and 4-methoxy (5g and 6g) remarkably enhanced both COX-2 inhibitory activity and
SI values over the unsubstituted congeners. 4-Bromophenyl triazolyl compounds 5e
and 6e showed about 4-5 times the inhibitory activity of indomethacin with about 950
times its SI towards COX-2. They reached about 6-8 times the inhibitory activity of
diclofenac with about 19 times its SI towards COX-2. 4-Methoxyphenyl triazolyl
compounds 5g and 6g showed about a 3.5-6 times the inhibitory activity of
indomethacin with about 700 and 1260 times its SI towards COX-2, respectively.
They reached about 5-8 times the inhibitory activity of diclofenac with about 14 and
2
5 times its SI towards COX-2, respectively.
On the other hand, 4-substitition with methyl (5f and 6f) and carboxylic
functionalities significantly reduced COX-2 inhibitory activity and SI values in
comparison to their unsubstituted counterparts. Surprisingly, the novel propargyl
rhodanine 4b showed higher COX-2 inhibitory activity and SI value than that of
13

ACCEPTED MANUSCRIPT
propargyl thiazolidine-2,4-dione 4a and was among the most active compounds in the
whole study.
Cycloaddition reaction of 4a and 4b with 4-substituted phenacyl azides generally
enhanced the activity with the highest difference observed with the 4-bromo
substituted derivatives 5j and 6j.
It is noteworthy that rhodanine derivatives demonstrated superiority in both COX-2
inhibitory activity and SI values in comparison to their thiazolidine-2,4-dione
counterparts, with the exception of 5a and 5b. A general trend was observed that
bromo-containing compounds were among the most active within the whole study.
2
.4.2. In vitro 15-LOX inhibitory assay:
Compounds that showed the highest in vitro COX-2 inhibitory activity (4b, 5b, 5e,
g, 5j, 6b, 6e, 6g, 6i and 6j) were further subjected to in vitro lipoxygenase inhibition
5
assay using soybean 15-LOX assay kit (Catalog no. 760700; Cayman Chemicals Inc.
Ann Arbor, MI, USA). Both soybean LOX and human 15-LOX are structurally
similar in terms of the presence of both C-terminal catalytic domain and N-terminal
membrane binding domain [14]. Also, their catalytic domains display a high level of
conservation. More specifically, within 10 Angstrom of the catalytic binding site they
show more than 50 % identity [14]. Hence, soybean LOX has been used by a lot of
research groups as a valid model to gauge the ability of many novel compounds to
inhibit human 15-LOX [14]. In vitro 15-LOX enzymatic inhibitory activities are
expressed as IC50 values of the tested compounds and shown in Table 1. The
universal LOX inhibitor, nordihydroguaiaretic acid (NDGA)[48,49] and selective
1
2/15-LOX inhibitor quercetin [50,51] were used as positive controls for comparison.
All ten compounds showed substantial LOX inhibitory activity with IC in the range
5
0
of 3.14-5.92 µM. They operated in the same order of magnitude as NDGA and
quercetin. All ten compounds showed superior LOX inhibitory activity to that of
NDGA while displaying varying activity when compared to quercetin. Interestingly,
the most and least potent derivatives 4b and 5b had almost thrice and twice the
activity of NDGA, respectively. Additionally, 4b was more potent than quercetin
while 5e and 5g were almost equipotent to quercetin. Other compounds appeared in
14

ACCEPTED MANUSCRIPT
the following descending order of activity; 6i, 5e, 6b, 6j, 5j, 6g and 5b. Consequently,
the above results showed that all ten compounds have promising LOX inhibitory
activity.
Table 1: In vitro COX-1, COX-2, 15-LOX inhibitory IC50 values and COX SI
values of synthesized compounds.
a
IC µM
SI
5
0
Code
Structure
COX-1/
COX-2
COX-1
COX-2 15-LOX
b
Celecoxib
Diclofenac Na
Indomethacin
14.8
3.9
0.039
0.05
0.8
0.49
-
-
-
296
4.9
0.1
6
5
a
a
8.4
9.4
0.33
0.27
0.11
0.10
-
25.5
34.8
121.8
159
-
6
5
b
b
13.4
15.9
4.32
5.92
6
5
c
c
6.4
5.9
0.26
0.38
-
-
24.6
15.5
15

ACCEPTED MANUSCRIPT
N
N
N
O
O
6
5
d
d
8.6
6.9
0.19
0.26
0.11
0.12
0.41
0.52
0.10
0.14
-
-
45.3
26.5
95.5
96.7
18.0
11.3
126
70
S
S
NH
Cl
N
N
N
O
O
6
5
e
e
10.5
11.6
7.4
3.74
4.23
-
S
S
NH
Br
6
5
f
f
5.9
-
6
5
g
g
12.6
9.8
5.24
3.22
16

ACCEPTED MANUSCRIPT
6
5
h
h
8.2
5.9
0.36
0.45
-
-
22.8
13.1
N
O
O
O
O
N
N
O
O
O
O
O
6
i
i
S
S
12.4
10.9
14.9
0.15
0.22
0.11
4.22
82.7
49.5
NH
Cl
Cl
Br
Br
S
N
N
N
N
N
O
5
-
NH
O
N
N
O
6
5
j
j
S
4.34
135.5
NH
S
N
N
O
S
13.6
11.6
0.10
0.12
4.97
3.14
136
NH
O
4
b
96.7
4
a
6.9
-
0.33
-
-
20.9
-
NDGA
10.25
17

ACCEPTED MANUSCRIPT
Quercetin
-
-
3.27
-
a
IC is the concentration µM) needed to cause 50% inhibition of COX-1, COX-2 and
5
0
(
1
5-LOX enzymatic activity. All values are expressed as mean of three replicates with
standard deviation less than 10% of the mean.
b
Selectivity index (SI) = IC (COX-1)/ IC (COX-2)
5
0
50
2
.4.3. In vitro glucose uptake using rat hemi-diaphragm model:
In order to assess the antidiabetic activity of compounds that showed the highest in
vitro COX and LOX inhibitory activities (4b, 5b, 5e, 5g, 5j, 6b, 6e, 6g, 6i and 6j),
their glucose uptake potential by rat hemi-diaphragm method was measured as
previously reported [31,52,53]. The glucose content of the working solution was
measured, and the glucose uptake was calculated as the difference between the initial
and final glucose content at 2 mg of optimized drug concentration. The glucose
uptake by rat hemi-diaphragm was measured in mg/dl/45 min. Data were expressed as
mean ± standard error of mean (SEM) and are shown in Figure 3.
Data were analyzed using IBM SPSS software package version 20.0. Statistical
analysis was done by ANOVA followed by Tukey post hoc test. Results of the in vitro
glucose uptake assay revealed that compounds 4b, 5e, 6g and 6j showed equal or
higher glucose uptake than pioglitazone without insulin and significantly higher
uptake with insulin. Compounds 5j and 6i showed lower glucose uptake without
insulin than pioglitazone but equal or higher uptake with insulin. Interestingly,
compounds 5b and 5g showed equal or higher glucose uptake than pioglitazone in an
insulin-independent fashion.
18

ACCEPTED MANUSCRIPT
Figure 3. In vitro glucose uptake activity of the most active compounds both in
absence and presence of insulin using rat hemi-diaphragm model. Data represented
are mean ± SD of three replicates. Statistical analysis was performed using Two-way
ANOVA followed by Sidak post hoc test. A P-value < 0.05 was considered
significant. * denotes significance vs. the corresponding effect of each compound in
absence of insulin, # denotes significance vs. the corresponding arm of tyrode
exposure, while $ denotes significance vs. the corresponding arm of pioglitazone
treatment.
2
.4.4. Immunohistochemical analysis of ligand-stimulated PPARγ nuclear
localization:
PPARγ agonists are known to affect the sub-cellular localization of the receptor
increasing its translocation to the nucleus with subsequent regulation of gene
expression [54]. The ability of the synthesized compounds to induce PPARγ nuclear
translocation was studied in rat primary adipocytes as described previously [21,55].
Four compounds were selected; 4b & 5e shown to have an insulin-sensitive glucose
uptake effect comparable to pioglitazone, and 5b & 5g shown to have a significant
insulin-independent glucose uptake. The effect of a 3-hour exposure to 10 µM
concentration of each of those compounds on the nuclear localization of PPARγ in rat
adipocytes was compared to that of an equivalent concentration of pioglitazone.
19

ACCEPTED MANUSCRIPT
Figure 4 shows representative micrographs of control and treated adipocytes.
Untreated controls (exposed to an equivalent volume of DMSO as the vehicle)
showed diffuse cytosolic PPARγ staining. Adipocytes treated with pioglitazone or the
selected test compounds showed redistribution of PPARγ staining that is
superimposed with nuclear staining as depicted in the fluorescence intensity profiles
along the line scans showed in Figure 4. This is suggestive of the ability of the
selected compounds to induce PPARγ nuclear translocation similar to pioglitazone.
The confirmation of PPARγ nuclear translocation together with insulin-independent
glucose uptake of compounds 5b & 5g raise the possibility of selective targeting of
signaling pathways. We viewed such behavior as an analogy to the phenomenon of
biased signaling that has been extensively described for GPCRs [56,57]. Theoretically
speaking, the concept of signaling selective agonism can be applied to any multi-
signaling receptor, yet and up to our knowledge, compounds 5b and 5g are the first
small molecule compounds reported to exhibit such behavior on PPARγ.
Figure 4. Immunofluorescent staining of PPARγ in untreated rat primary adipocytes
and those exposed to 10 µM of pioglitazone and different test compounds for 3 hours.
Phase contrast cell contours are delineated in white. Nuclei were stained with DAPI.
The corresponding graphs show staining intensity for PPARγ (green) and DAPI (blue)
along the scan line indicated on the corresponding micrographs.
2
.4.5. PPARγ functional reporter gene assay:
The effect of the selected compounds showing PPARγ nuclear translocation on its
transcriptional activity was assessed using a functional gene reporter assay[58].
Human embryonic kidney (HEK293) cells transiently transfected with two
components: 1) a hybrid receptor comprising the N-terminal Gal4 DNA binding
20

ACCEPTED MANUSCRIPT
domain fused to the ligand binding domain of human PPARγ and 2) reporter Renilla
luciferase gene functionally linked to the Gal4 upstream activation sequence. The
transfected HEK293 cells were exposed to different concentrations of pioglitazone,
4
b, 5b, 5e, and 5g for 24 hours. Luciferase activity was measured and compared to
basal activity in cells treated with the vehicle (DMSO). As shown in Figure 5, at the
highest concentration (10 µM), pioglitazone showed a 6-fold increase in luciferase
activity compared to vehicle. At the same concentration, the tested compounds
showed a 2-3 fold increase, with transcriptional activity of 25-45% that of
pioglitazone. Based on the transcriptional response, a partial PPARγ agonistic activity
can be assumed for the tested compounds.
Figure 5. Graphical representation of the increased PPARγ-mediated transcriptional
activity following treatment with different concentrations of pioglitazone and selected
synthetic compounds. Results are expressed as fold change compared to luciferase
activity in vehicle-treated cells and represented as Mean ± Standard deviation.
Luciferase activity values were compared to vehicle with One-way ANOVA followed
by Dunett Multiple Comparisons test. * denotes a P-value < 0.05 vs. DMSO treated
cells.
2
.4.6. Inhibition of monocyte to macrophage differentiation:
21

ACCEPTED MANUSCRIPT
It is widely recognized that early stages of the inflammatory response involve
monocyte-to-macrophage differentiation [59]. This process was shown to be
particularly important in the development of atherosclerosis [60]. Interestingly,
pioglitazone was shown to modulate the M1/M2 monocyte differentiation imbalance
in obese diabetic patients increasing the putative anti-inflammatory M2 polarization,
and this effect was postulated to underlie its glucose-independent anti-atherogenic
activity [61,62]. Additionally, COX-2 expression is induced during monocytes
activation into macrophages with subsequent contribution to the inflammatory
response and oxidant/anti-oxidant cellular imbalance [63,64]. Of particular interest,
prostaglandin E2, via its action on EP2 and EP4 receptors, was shown to stimulate the
production and release of Interleukin-10 (IL-10), a cytokine with reprogramming
effects on monocytes and macrophages [65]. Moreover, macrophage activation and
adhesion was shown to be dependent on 15-LOX activation [66]. Our previous studies
showed that compounds with dual COX-2 and LOX inhibitory activity were able to
inhibit monocyte-to-macrophage differentiation[37]. As such, we anticipated that the
four compounds (4b, 5b, 5e, and 5g) that showed dual COX-2 and LOX inhibitory
activities together with PPARγ partial agonistic activities, would maintain a similar, if
not enhanced, activity against THP-1 monocyte differentiation into macrophages.
Towards this end, we used the PMA-induced THP-1 differentiation assay as an
accepted in vitro model of this process [67]. The effects of these compounds in the
assay were compared to diclofenac and pioglitazone as reference compounds. Results
and IC50 values are summarized in Figure 6. Surprisingly, pioglitazone did not
produce an appreciable inhibition of PMA-induced monocyte differentiation up to 300
µM concentration. 5g produced effects similar to pioglitazone yielding IC50 values in
the mM range. Diclofenac produced a concentration-dependent inhibition of the
differentiation process yielding an IC value close to that previously reported in a
5
0
similar assay[37]. 5b and 5e showed IC50 values close to that of diclofenac, while 4b
appeared to be more potent than diclofenac in inhibiting this process. Yet, it is worth
mentioning that this assay measures the differentiation of monocytes into adherent
macrophages with no indication of potential differences in differential polarization, if
22

ACCEPTED MANUSCRIPT
any. Thus, a negative result might not completely rule out the beneficial effect of
PPARγ agonistic activity represented by pioglitazone.
Figure 6. Inhibition of monocyte to macrophage differentiation by diclofenac,
pioglitazone, and the selected synthetic compounds. Results are represented as
percentage macrophage metabolic activity of the PMA control at different exposure
concentrations. IC50 values estimated by non-linear regression were compared using
One-way ANOVA followed by Tukey Multiple Comparisons post hoc test. * denotes
a P-value < 0.05 vs. diclofenac while # denotes a P-value < 0.05 vs. pioglitazone.
2
.4.7. IL-1β and TNF-α expression and 20-HETE production assay:
Previous literature consistently showed that 15-LOX activation is associated with
increased cytokine production in variety of cell types [68–70]. Specifically, both the
addition of 15-LOX arachidonate metabolites or their de novo synthesis in monocytes
triggered cytokine production in macrophages, including tumor necrosis factor-α
(
TNF-α), interleukin-1β (IL-1β), IL-6, and monocyte chemo-attractant protein-1 [71].
Moreover, the produced IL-1β contributes to further stimulation of signaling
pathways leading to the enhancement of phospholipase A2-mediated arachidonate
release and metabolism in a positive feedback loop [72]. We attempted to provide a
pathophysiological context for the 15-LOX inhibitory action of our hybrid
compounds. We examined the effect of the three compounds showing the highest
activity against soybean LOX in vitro (4b, 5e & 5g) on TNF-α and IL-1β production
in THP-1 monocytes differentiated into macrophages by PMA, and activated by a
23

ACCEPTED MANUSCRIPT
lipopolysaccharide (LPS) challenge. Compared to the negative control, LPS-treated
macrophages showed massive increases in both TNF-α and IL-1β production Figure
7
A. The increase in both cytokines’ expression was greatly attenuated in cells treated
with our compounds. As well, we tested the effect of these compounds on the 20-
hydroxyeicosatetraenoic acid (20-HETE) production in these cells as a general
indicator of arachidonate metabolism. The concentration of 20-HETE produced in the
conditioned medium of PMA-differentiated THP-1 monocytes that were challenged
with LPS was measured by enzyme-linked immunosorbent assay (ELISA). 20-HETE
production increased following LPS challenge compared to untreated controls. Cells
treated with our compounds showed reduced 20-HETE production indicating a
possible decrease in arachidonic acid release possibly as a result of the interruption of
the IL-1β positive feedback loop (Figure 7B).
2
.0
*
A
1.5
1
0
0
.0
.5
.0
#
#
#
IL-1β
0
.8
.6
*
TNF-⍺
0
GAPDH
0.4
#
#
0
.2
.0
#
LPS
-
-
-
-
+
-
-
+
+
-
+
-
+
+
-
-
5e
0
5
g
LPS
-
-
-
-
+
-
-
+
+
-
+
-
+
+
-
-
4
b
-
-
-
+
5e
5g
4b
-
-
-
+
B
*
1
00
#
7
5
50
25
#
0
LPS
-
-
-
-
+
-
-
+
+
-
+
-
+
+
-
-
5
5
e
g
4
b
-
-
-
+
Figure 7. The most potent 15-LOX inhibiting compounds reduce cytokine production
and arachidonate metabolism in LPS-challenged PMA-differentiated THP-1 cells. A,
representative western blotting of IL-1β, TNF-α, and GAPDH from THP-1 cells
24

ACCEPTED MANUSCRIPT
exposed to different treatment conditions. The bar graphs summarize the
quantification of three separate experiments. B, 20-HETE concentration in the
conditioned medium of THP-1 cells exposed to the indicated treatment conditions.
Concentration is normalized to the values measured in untreated cells challenged with
LPS. Statistical significance was measured by one-way ANOVA followed by Tukey
Multiple Comparisons post hoc test. * denotes a P-value < 0.05 vs. untreated control
while # denotes a P-value < 0.05 vs. LPS-challenged cells.
2
.4.8. In vivo anti-inflammatory assay:
Formalin-induced rat paw edema protocol as an acute inflammation model was used
to test the in vivo anti-inflammatory activity of the four most active compounds (4b,
5
b, 5e and 5g). Formalin was injected subcutaneously to induce inflammation and an
oral dose of 5 mg/kg body weight of the test compounds was administered.
The potencies of the tested compounds relative to the negative control were identified
by measuring the percentage inhibition of edema after 4 h. Celecoxib and diclofenac
sodium were used as positive controls. Intriguingly, all tested compounds with the
exception of 4b showed equivalent inhibitory effect to celecoxib and diclofenac.
Additionally, 4b demonstrated a statistically significant higher potency compared to
celecoxib (Supplementary Information Table SM2 and Figure 8).
Figure 8. Graphical representation of in vivo anti-inflammatory activities of selected
compounds in formalin-induced rat paw edema bioassay (acute inflammation model).
25

ACCEPTED MANUSCRIPT
Results presented are Mean ± SD of four replicates. Statistical analysis was performed
using One-way ANOVA followed by Tukey post hoc test. A P-value < 0.05 was
considered significant. * denotes significance vs. celecoxib.
2
.5. Structure determination of compound 4b bound to PPARγ:
Recently, 5-ene rhodanine substructures have been stigmatized by some research
groups and drug discovery programs as promiscuous hitters or pan assay interference
(
PAINS) elements [30]. However, the concept of PAINS is still controversial and, in
many instances, contradicts the concept of privileged scaffolds [73,74]. Additionally,
recent reports indicated that PAINS is highly structural context dependent [73].
Indeed, one of the most active compounds in this study (compound 4b) carries the ene
rhodanine substructure. Although the biological activity of 4b was confirmed by
multiple lines of evidences (molecular then cellular and finally in vivo), we wanted to
rule out any skepticism about its proven biological activity and exclude the presence
of any assay artifacts. Hence, we determined the X-ray crystal structure of compound
4
b in complex with the PPARγ ligand binding domain (LBD) which will give us the
chance to probe its binding mode and detect specific binding interactions. Coordinates
and structure factors were deposited in the Protein Data Bank under accession code
6
E5A.
Experimental details can be found in the methods section, the crystallographic data
table describing the data processing and refinement statistics can be found in the
Supplementary Information Table SM3, and a stereo figure displaying the quality
electron density around the ligands can be found in the Supplementary Information
Figure SM2. Two subunits per asymmetric unit were found and both subunits were
found bound to 4b in a similar manner. The overall fold of PPARγ was similar to
most other observed structures, for example the 4b bound structure displays an RMSD
to the rivoglitazone bound structure (PDB:5U5L) of 0.48Å (over 512 Cα atoms),
signifying that the binding of compound 4b does not induce major conformational
changes as compared to other TZD based compounds. Compound 4b is located in the
LBD with the ring systems of the compound located between H3 and the beta-sheet
region and the carbon tail inserted into the hydrophobic patch below the AF2 region
(
Figure 9A). 4b makes only one hydrogen bond to PPARγ, via its non-protonated
26

ACCEPTED MANUSCRIPT
nitrogen atom to the protonated nitrogen backbone atom of Ser342 (3.0Å). All other
contacts are hydrophobic and Van der Waals interactions. The double ring system of
compound 4b sits between H3 and the beta sheet region making these contacts with
Ile281 (H3), Ile341 side chain (beta-sheet), the non-polar portion of Ser342 side chain
(
(
beta-sheet), the non-polar portion of Arg288 side chain (H3), Leu353 side chain
H6), and the side chain of Met364 (H7). These interactions can be seen in Figure
9
B. The carbon tail of 4b makes hydrophobic interactions with H6 and H7 which
include those to the side chain of Met364, the side chain of Phe363 (H7), the side
chain of Phe360 (H7), and the side chain of Leu353 (H6) as shown in Figure 9C.
Given that we also produced a docked structure of 4b bound to PPARγ that was in
excellent agreement with our X-ray crystal structure, we hypothesized that further
structural analysis of the other compounds could be carried out using docking
techniques.
C
C
27

ACCEPTED MANUSCRIPT
Figure 9. Binding mode of 4b to PPARγ as determined by X-ray crystallography
(
PDB accession code 6E5A). PPARγ is shown in green ribbons, compound 4b is
shown in yellow sticks, and residues of PPARγ within 3.5Å of the ligand are depicted
as green sticks. A) Overview of the entire PPARγ LBD bound to 4b. B) Interactions
of 4b with the beta-sheet and H3 region. C) Interactions of the carbon tail of 4b with
residues of the hydrophobic pocket formed by H7, and H3.
2
2
.6. Molecular Modeling:
.6.1. Docking of compounds 4b, 5b, 5g and 5e into COX-2 active site
The screening performance of COX-2 DEKOIS 2.0 benchmark set showed a good and
non-random result with three docking tools (e.g., GOLD [ChemPLP], Glide [SP] and
AutoDock Vina) [42]. Hence, we employed GOLD (ChemPLP) as scoring function
for pose prediction purposes using PDB code of 1CX2 crystal structure for COX-2
docking. It has been reported that variation in molecular weight can cause obvious
bias in docking performance especially when employing empirical scoring functions,
since heavier molecules tend to produce superior scores [75,76]. Minimizing such bias
would be of benefit in the context of correlating the docking scores of compounds
with their biological in vitro activity values.
Due to variability in molecular weights of the test compounds and in order to
eliminate this bias, score normalization strategy was applied by dividing the docking
score by the number of heavy atoms (N) using different arithmetic weights (e.g.,
1
/2
1/3
2/3
square root of N (N ), cube root of N (N ) and cube root of the squared N (N ))
77]. To select the most suitable normalization strategy, we evaluated the docking
[
screening performance employing such strategies using COX-2 DEKOIS 2.0
benchmark set, and visualized the results using “pROC-Chemotype” plots
(
Supplementary information Figures SM3-7). Besides, employing benchmarking
would enable the effective detection of active ligands in a pool of their false positive
decoys.
28