An attempt to chemically state the cross-talk between monomers of COX homodimers by double/hybrid inhibitors mofezolac-spacer-mofezolac and mofezolac-spacer-arachidonic acid

Article abstract of DOI:10.1016/j.ejmech.2020.112919

Cardiovascular diseases (CVDs) account for over 17 million death globally each year, including arterial thrombosis. Platelets are key components in the pathogenesis of this disease and modulating their activity is an effective strategy to treat such throm

Full text of DOI:10.1016/j.ejmech.2020.112919

Journal Pre-proof
An attempt to chemically state the cross-talk between monomers of COX homodimers
by double/hybrid inhibitors mofezolac-spacer-mofezolac and mofezolac-spacer-
arachidonic acid
Maria Grazia Perrone, Morena Miciaccia, Paola Vitale, Savina Ferorelli, Cristina
da Costa Bernardes Araújo, Gabriella Silva de Almeida, Thaisa Francielle Souza
Domingos, Luiz Claudio Rodrigues Pereira da Silva, Marcelo de Pádula, Lucio
Mendes Cabral, Plínio Cunha Sathler, Carmela Bonaccorso, Cosimo G. Fortuna,
Antonio Scilimati
PII:
S0223-5234(20)30891-6
DOI:
https://doi.org/10.1016/j.ejmech.2020.112919
EJMECH 112919
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 3 July 2020
Revised Date: 18 September 2020
Accepted Date: 5 October 2020
Please cite this article as: M.G. Perrone, M. Miciaccia, P. Vitale, S. Ferorelli, C. da Costa Bernardes
Araújo, G. Silva de Almeida, T.F. Souza Domingos, L.C. Rodrigues Pereira da Silva, M. de Pádula,
L.M. Cabral, P.C. Sathler, C. Bonaccorso, C.G. Fortuna, A. Scilimati, An attempt to chemically state
the cross-talk between monomers of COX homodimers by double/hybrid inhibitors mofezolac-spacer-
mofezolac and mofezolac-spacer-arachidonic acid, European Journal of Medicinal Chemistry, https://
doi.org/10.1016/j.ejmech.2020.112919.
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.
© 2020 Published by Elsevier Masson SAS.

An attempt to chemically state the cross-talk between monomers of COX homodimers by
double/hybrid inhibitors mofezolac-spacer-mofezolac and mofezolac-spacer-arachidonic
acid
Maria Grazia Perronea, Morena Miciacciaa, Paola Vitalea, Savina Ferorellia, Cristina da Costa
Bernardes Araújob, Gabriella Silva de Almeidab, Thaisa Francielle Souza Domingosb, Luiz
Claudio Rodrigues Pereira da Silvab, Marcelo de Pádulab, Lucio Mendes Cabralb, Plínio Cunha
Sathlerb, Carmela Bonaccorsoc, Cosimo G. Fortunac, Antonio Scilimatia,*
aDepartment of Pharmacy - Pharmaceutical Sciences, University of Bari “Aldo Moro”, Via E.
Orabona 4, 70125, Bari, Italy. bFaculty of Pharmacy, Federal University of Rio de Janeiro, Center
of Health Sciences, Carlos Chagas Filho Avenue, 373, 21941599, Rio de Janeiro, Brazil.
cDepartment of Chemical Science, University of Catania, Viale Andrea Doria 6, 95125, Catania,
Italy.
Corresponding author: Antonio Scilimati phone, +39 080 5442753; fax, +39 080 5442724,
email: antonio.scilimati@uniba.it.
1

ABSTRACT
Cardiovascular diseases (CVDs) account for over 17 million death globally each year, including
arterial thrombosis. Platelets are key components in the pathogenesis of this disease and
modulating their activity is an effective strategy to treat such thrombotic events. Cyclooxygenase-
1 (COX-1) isoenzyme is involved in platelet activation and is the main target of non-steroidal anti-
inflammatory drugs (NSAIDs) and new selective inhibitor research. Such inhibitors of general
formula mofezolac-spacer-mofezolac (mof-spacer-mof) and mofezolac-spacer-arachidonic acid
(mof-spacer-AA) were projected to investigate the possible cross-talk between the two monomers
(Eallo and Ecat) forming the COX-1 homodimer. Mofezolac was chosen as either one or two
moieties of these molecules being the known most potent and selective COX-1 inhibitor and
administered to humans as Disopain™, then arachidonic acid (AA) was used to develop molecules
bearing, in the same compound, in addition to the inhibitor moiety (mofezolac) also the natural
COX substrate. Depending on the nature of the spacer, COX-1 and COX-2 activity was
differently inhibited by mof-spacer-mof set with a preferential COX-1 inhibition. The highest
COX-1 selectivity was exhibited by the compound in which the spacer was the benzidine [N,N'-
(biphenyl-4,4'-di-yl)bis(2-(3,4-bis(4-methoxyphenyl)isoxazol-5-yl)acetamide) (15): COX-1 IC50 =
0.08 µM, COX-2 IC50 > 50 µM, Selectivity Index (SI) > 625]. In the case of mof-spacer-AA set,
the COX inhibitory potency and also the isoform preference changed. (5Z, 8Z, 11Z, 14Z)-N-(4-{2-
[3,4-Bis(4-methoxyphenyl)isoxazol-5-yl]acetamido}butyl)icosa-5,8,11,14-tetraenamide (19) and
(5Z, 8Z, 11Z, 14Z)-N-(4'-{2-[3,4-bis(4-methoxyphenyl)isoxazol-5-yl]acetamido}-[1,1'-biphenyl]-
4-yl)icosa-5,8,11,14-tetraenamide (21), in which the spacer is the 1,2-diaminobutane or benzidine,
respectively, selectively inhibited the COX-2, whereas
when the spacer is the 1,4-
phenylendiamine [(5Z, 8Z, 11Z, 14Z)-N-(4-{2-[3,4-bis(4-methoxyphenyl)isoxazol-5-
yl]acetamido}phenyl)icosa-5,8,11,14-tetraenamide) (20) the COX preference is COX-1 (COX-1
2

IC50 = 0.05 µM, COX-2 IC50 > 50 µM, with a COX-1 selectivity > 1000). Molecular modelling by
using FLAP algorithm shows fundamental interactions of the novel compounds at the entry
channel of COX and inside its catalytic cavity. The effect of these mof-spacer-mof and mof-
spacer-AA in inhibiting in vitro free arachidonic acid-induced platelet aggregation was also
determined. A positive profile of hemocompatibility in relation to their influence on the blood
coagulation cascade and erythrocyte toxicity was observed. Cytotoxicity and genotoxicity safety
were also found for these two novel sets of compounds.
KEYWORDS: Mofezolac, Arachidonic acid, COX-1 and COX-2 selective inhibitors,
Fingerprints for Ligands and Proteins, Drug design, Antiplatelet agents, Thrombosis,
Cardiovascular diseases
3

1. Introduction
According to World Health Organization (WHO), cardiovascular diseases (CVDs) are
responsible to 17.7 million deaths worldwide per year, representing 31% of all deaths. This
number is expected to grow to 23.6 million until 20301,2. Atherosclerosis is the most common
clinical manifestation related to CVDs, which is characterized by an atherosclerotic plaque
accumulation causing arterial wall thickening and luminal narrowing1,3. Rupture of atherosclerotic
plaques in vessels may cause thrombotic occlusion leading to a myocardial infarction and
ischemic stroke4,5.
Platelets play a key role in pathogenesis of atherothrombosis due to their ability of aggregation
and adhesion in the content of atherosclerotic plaques exposed in arterial circulation after rupture,
forming with them a thrombus rich in fibrin6–8. Antiplatelet therapy, such as clopidogrel and
aspirin, is used for treatment and prevention of atherothrombosis9. Aspirin is a drug that act as an
irreversible inhibitor of both COX-1 and COX-2, and at low dose selectively inhibits COX-19,10
.
COX-1 is an enzyme present also in platelet surface, where it is responsible to oxidize the
arachidonic acid (AA), its natural substrate, leading to the thromboxane A2 (TXA2) formation,
which is in turn a platelet activator and increases the platelet aggregation11. Therefore, COX-1
inhibitors are used as antiplatelets, reducing the production of TXA2 and consequently platelet
aggregation12. However, use of antiplatelets demands knowledge of their characteristics, risks and
benefits, since they can cause severe adverse reaction as bleeding, neutropenia, thrombocytopenia
and drug resistance13,14
.
Diarylisoxazoles are well known COX-1 inhibitors15-18, abundantly studied also by X-ray in
complex with the enzyme to gain insight of COX-1 catalytic core19. The selectivity towards COX-
1 has advantages, such as inhibition of platelet TXA2 production and lack of gastrointestinal
toxicity20,21, whereas non-selective COXs inhibitors have adverse cardiovascular side effects due
4

to their action as reducer of prostacyclin (PGI2) biosynthesis, which has cardioprotective effect
being a vasodilator and a potent platelet aggregation inhibitor22,23. Herein, we report the design
and synthesis of two sets of novel isoxazoles having the general formula mof-spacer-mof and mof-
spacer-AA (Figure 1), respectively. They were evaluated as COXs and arachidonic acid-induced
platelet aggregation inhibitors. In vitro coagulation, hemocompatibility to health human
erythrocytes and cyto- and geno-toxicity tests were also performed. Structure Based Virtual
Screening (SBVS) was carried out and the binding poses of selected compounds in the catalytic
pockets of both COX-1 and COX-2 enzymes were analyzed24.
Figure 1. Chemical structures of mofezolac, arachidonic acid and two sets of novel isoxazoles
(mofezolac-spacer-mofezolac and mofezolac-spacer-AA).
2. Results and discussion
COXs are homodimers composed of ~72 kDa subunits that are tightly bonded to one another via
an interface spanning 2500 Å2. Each COX monomer contains an epidermal growth factor-like
domain, a membrane binding domain (MBD), and a large hydrophobic L-shaped catalytic core.
5

The bifunctional catalytic subunit houses both COX and peroxidase (POX) enzymatic activities.
Both COX isoforms exhibit half-of-sites COX activity when a ligand binds to one monomer of an
unoccupied dimer, that monomer becomes the allosteric subunit (Eallo), and the partner monomer
becomes the catalytic monomer (Ecat). The catalytic efficiency of the catalytic monomer is
determined by its interaction with its allosteric partner and this, in turn, is different for each
different ligand that binds the allosteric monomer. With the aim to study the relationship between
the activity of the two monomers, we have designed and prepared compounds bearing (i) two
molecules of inhibitor linked by a spacer “mofezolac-spacer-mofezolac” (mof-spacer-mof) (9-
15) or (ii) an inhibitor molecule linked to the arachidonic acid (COXs endogenous substrate)
"mofezolac-spacer-AA" (mof-spacer-AA) (19-21) (Figure 1). Both two sets of molecules were
projected with the idea that each compound eventually could simultaneously interact with the two
monomers Eallo and Ecat, respectively, being formed by either two close mofezolac moieties or
mofezolac and AA.
2.1
Rational behind the design of the novel compounds
The novel isoxazoles were though, as “double/hybrid molecules”, taking into consideration the
dual characteristic of the COX-1 homodimer (PDB code 6Y3C) and the eventual cross-talk
between monomers. The “double molecules” refers to compounds bearing two mofezolac units,
instead “hybrid molecules” is used to indicate compound constituted by oone mofezolac moiety
and an arachidonic acid unit. Specifically, the novel molecules are constituted by two parts linked
by a flexible or rigid moiety (spacer). One of the two parts is mofezolac (mof), a potent and
selective COX-1 inhibitor (COX-1 IC50 = 0.0079 µM and COX-2 IC50 >50 µM), mainly
recognized by Arg120 residue located close to the COX membrane binding domain and at the
6

beginning of the long hydrophobic channel containing the COX catalytic site19,22 and the second
molecular-part is another mofezolac molecule, affording mof-spacer-mof (double molecules 9-15).
The hypothesis underlined such an overall structure is that the two monomers forming the COX
homodimer play a different role, the first monomer when is occupied by a non-selective NSAIDs
behaves as an allosteric enzymatic portion (Eallo), determining a conformational change of the
partner monomer (Ecat) that exerts, then, the COX and POX catalysis. Hence, when one of the
two mofezolac present in mof-spacer-mof interacts with Eallo, the second mofezolac should
interact with Ecat, being close to it. This will definitively inhibit any COX catalytic activity.
The importance of understanding the cross-talk between the COX-allosteric and -catalytic
monomers was highlighted by a work showing that COX-2 inhibitors can bind tightly to one
monomer of COX-1 and attenuate the actions of aspirin on COX-1 in vitro and in vivo25. From a
pharmacological viewpoint, this condition resembles the case in which a patient treated with one
of non-selective NSAIDs (i.e., a profen, such as ibuprofen) or COX-2 selective inhibitors
(COXIBs: i.e., celeCOXib) to control an inflammatory status and contemporaneously takes the
anti-platelet low-dose aspirin. In this case, it should be expected that ibuprofen binding to Eallo
induces an Ecat conformational rearrangement to impair the aspirin access to Ecat catalytic site.
This will favor, instead, the access of the arachidonic acid to Ecat. In other words, a patient that is
treating inflammation with one of NSAIDS fails to prevent CVD events if simultaneously is
assuming low-dose aspirin (85-300 mg/die targets only COX-1), because COX-1-Ecat is
unoccupied and hence will still catalyze the conversion of AA in PGH2, in turn transformed in
thromboxane (TXA2). Therefore, the “inflamed” patient believes to be protected against adverse
CVD events but in practice he is not.
7

The second set of molecules “hybrid molecules”, were obtained by replacing one mofezolac unit
with the arachidonic acid (AA), endogenous COX substrate, obtaining mof-spacer-AA compounds
(19-21). In this case, the hypothesis is mof-moiety should interact with Eallo and AA moiety with
Ecat.
2.2
Chemistry
For the preparation of the two acetamide bonds of "mof-spacer-mof" series (9-15), mofezolac was
in situ activated in inert atmosphere with 1-(3-dimethylaminopropyl)-3-ethylcarbodimide
hydrochloride (EDC.HCl) and 1-hydroxybenzotriazole monohydrate (HOBt.H2O) at 0 °C in
anhydrous CH2Cl2 and then reacted with the opportune diamine (spacer) (2-8) in the presence of
N,N-diisopropylethylamine (DIEA), affording, after overnight stirring the final products in 25-83
% yields (Scheme 1).
Scheme 1. Synthesis of mof-spacer-mof set (9-15). Reagents and conditions: (a) HOBt.H2O,
EDC.HCl, CH2Cl2, 0 °C, 2h; (b) diamine, DIEA, CH2Cl2, r.t., 16h; (c) H2O.
8

With a similar approach the mofezolac/arachidonic bis-amides of mof-spacer-AA series were
obtained by reacting mofezolac with the opportune diamine (1 mmol) (3, 6 and 8), in the presence
of N,N-diisopropylethylamine (DIEA), 1-hydroxybenzotriazole monohydrate (HOBt.H2O) and 1-
(3-dimethylaminopropyl)-3-ethylcarbodimide hydrochloride (EDC.HCl) in anhydrous CH2Cl2 at 0
°C and under an inert atmosphere, affording the corresponding acetamides intermediates in 65-92
% yields (Scheme 2), that were finally reacted with arachidonic acid, activated by EDC.HCl and
HOBt.H2O, to afford the corresponding amides in 31-85 % yields.
9

Scheme 2. Synthesis of mof-spacer-NH2 (16-18) and mof-spacer-AA (19-21). Reagents and
conditions: (a) HOBt.H2O, EDC.HCl, CH2Cl2, 0 °C, 2h; (b) H2N-spacer-NH2 (3, 6, 8; *1,4-
diaminobutane (3) was first converted into NH2(CH2)4NHBoc (3a) and allowed to react with
mofezolac), DIEA, CH2Cl2, r.t., 16h; (c) H2O; (d) 16-18, DIEA, CH2Cl2, r.t.; (e) 2h stirred
solution kept at 0 °C under argon of AA, HOBt.H2O, EDC.HCl, CH2Cl2, r.t., 16h; (f) H2O.
2.3
Cyclooxygenase catalytic activity inhibition evaluation
Cyclooxygenase activity inhibition was determined by a colorimetric COX Inhibitor Screening
Assay which measures the peroxidase component (POX) of the cyclooxygenases, monitoring the
appearance of oxidized N, N, N’, N’-tetramethyl-p-phenylenediamine (TMPD) at λ = 590 nm
(Table 1 and 2).
Mof-C2-mof (9) is constituted by two mofezolac molecules linked by an ethylene bridge (Table 1).
It inhibits both isoenzymes with a preference towards the COX-1 isoform (COX-1 IC50 = 0.013
µM and COX-2 IC50 = 0.12 µM), being its selectivity index SI equal to 9. A higher SI value is
obtained by elongating the linker of two carbon units, substituting the ethylene to a butylene
spacer, obtaining mof-C4-mof (10). COX-1 selectivity is retained when the linker is longer, as in
the case of mof-C6-mof (11) and mof-C12-mof (12), even if the inhibitory potency (IC50) of COX-
1 activity decreases (COX-1 IC50 = 9.0 and 7.1 µM, respectively). Mof-C6H4-mof (13) was
obtained by replacing the C2÷C12 carbon units as spacers by the conformationally rigid phenyl
ring. A good grade of inhibition was restored registering an 85% percentage of inhibition, an IC50
of 0.15 µM and a SI of 333. No inhibition at all was observed towards both COX isoforms when a
-CH2C6H4CH2- moiety was used as a spacer. The absence of any COXs activity inhibition by mof-
CH2C6H4CH2-mof (14) was likely due to its higher possible conformational states. The presence
of benzidine as a linker in mof-C6H4-mof (13) seemed to be responsible of its improved inhibitory
activity with respect to 10-12. This result prompted us to design the most selective compound of
10

the series (mof-C6H4-C6H4-mof) (15). Specifically, it has an IC50 = 0.08 µM and an SI = 625
(Table 1).
Table 1. COX inhibitory activity of mof-spacer-mof molecules (9-15) bearing different spacers
(2-8).
11

IC50 (µM)a
cLogPe
Compound
(% inhibition)b
SIc MW
COX-1
COX-2
0.0079 ± 0.15
>50
6392 339.11 2.51
0.013±0.0003 0.12±0.002
4.19
9
702.75
(75)
(45)
n = 1, spacer = -(CH2)2- (mof-C2-mof) (9)
n = 2, spacer = -(CH2)4- (mof-C4-mof) (10)
5.5±0.1
(74)
3.94
>50
>50
>50
> 9 730.30
5.5 758.86 5.00
842.43 8.17
9.0±0.2
(50)
n = 3, spacer = -(CH2)6- (mof-C6-mof) (11)
n = 6, spacer = -(CH2)12- (mof-C12-mof) (12)
7.1±0.3
(40)
7
0.15±0.04
(85)
>50
333 750.27 5.45
mof-C6H4-mof (13)
>50d
>50d
-
778.30 5.49
12

mof-CH2C6H4CH2-mof (14)
mof-C6H4-C6H4-mof (15)
0.08±0.001 (73)
>50
625 826.30 7.34
aIC50 values are the means of at least three independent measurements. bInhibition percentage (%)
c
determined at the highest final inhibitor concentration used (50 µM). Selectivity Index (SI) =
COX-2 IC50/COX-1 IC50. eMeasured by using ChemBio3D Ultra.
Among the set of compounds mof-C2÷C12-mof (9-12) (Table 1), the alkyl C4 spacer appeared to
be the best length to achieve the highest COX-1 selectivity and a good extent of COX-1 activity
inhibition. Then, the alkyl C4 spacer was used to conjugate mofezolac to arachidonic acid (AA)
(Scheme 2, Table 2), the natural endogenous COXs substrate. In this case, an inversion of
selectivity was observed. In fact, mof-C4-AA (19) is a preferential COX-2 inhibitor with a COX-2
IC50 of 0.8 µM, even though its precursor mof-C4-NH2 (16) is a COX-1 selective inhibitor (COX-1
IC50 = 0.95 µM, COX-2 IC50 = 3.5 µM). As from Table 1, COX-1 inhibition preference was
observed when a phenyl ring is the spacer in mof-C6H4-mof (13). The corresponding mof-C6H4-
AA (20) was found to be the most potent and selective COX-1 inhibitor of the series with a COX-
1 IC50 of 0.05 µM, COX-2 IC50 > 50 µM, and a selectivity index higher than 1,000. Its precursor
mof-C6H4-NH2 (17), instead, exhibited a COX-2 inhibition preference (COX-1 IC50 = 24.0 µM,
COX-2 IC50 = 0.5 µM). The COXs selectivity preference in favor of COX-2 (SI = 189) was also
observed when a benzidine was used as spacer to prepare mof-C6H4-C6H4-AA (21). Such a SI
value is higher than any of COXIBs selectivity, despite of the behavior of its precursor (18) that
showed a COX-1 inhibition preference (COX-1 IC50 = 0.075 µM, COX-2 IC50 = 0.3 µM).
Table 2. COX inhibitory activity of compounds composed by mofezolac and arachidonic acid
(19-21) and their precursors (mofezolac linked to the spacers) (16-18).
13

IC50 (µM)a
SIc
(1/SI)
cLogPd
MW
(% inhibition)b
Compound
COX-1
COX-2
0.95±0.01
(100)
3.5±0.12
(51)
3.7
(0.27)
409.20 1.49
mof-C4-NH2 (16)
mof-C4-AA (19)
0.05
16.0±0.2
(57)
0.8±0.02
(61)
(20) 695.43 8.66
24.0±0.9
(74)
0.5±0.0
(79)
0.02
429.17 2.57
(48)
mof-C6H4-NH2 (17)
0.05±0.002
(66)
> 1,000
>50
715.40 10.44
(> 0.001)
mof-C6H4-AA (20)
0.075±0.002
(85)
0.3±0.01
(71)
4
505.20 4.46
(0.25)
mof-C6H4-C6H4-NH2 (18)
17±0.7
(50)
0.09±0.003
(60)
0.005
791.43 12.29
(189)
14

mof-C6H4-C6H4-AA (21)
aIC50 values are the means of at least three independent measurements. bInhibition percentage (%)
c
determined at the highest final inhibitor concentration used (50 µM). Selectivity Index (SI) =
COX-2 IC50/COX-1 IC50. dMeasured by using ChemBio3D Ultra.
It is evident from the data listed in Table 1 and 2, that even if more and more work has been done
to understand the rationale behind the selectivity of COXs inhibitors no definitive rules are
available to design a highly selective COX-1 or COX-2 inhibitor. The two sets of the novel mof-
spacer-mof and mof-spacer-AA were projected to try to define the role of Eallo and Ecat, the two
monomers forming the COX homodimers, often considered conformationally heterodimers25. The
condition used to measure the inhibitory capability of the novel compounds requires an
arachidonic acid concentration kept at 100 µM. The highest concentration at which the new sets of
compounds are usually tested is 50 µM, this could mean that if AA has a higher affinity for Eallo
(0.25 µM for both COX isoforms) than for Ecat (0.5-2.0 and 5.0-10.0 µM for COX-1 and COX-2,
respectively) there could be both a AA mass effect along and compounds chemical structures
(size, polarity, lipo- and hydro-philia, the presence of specific functional groups, specific
interactions with the almost 50 amino acid interactions present in the long hydrophobic channel
constituting the COXs active site and involved in the substrate/inhibitors recognition) to direct the
selectivity of inhibitors either towards COX-1 or COX-2. In vivo, the situation is still more
complex for the presence of circulating saturated, mono- and poly-unsaturated fatty acids.
Saturated and mono-unsaturated free fatty acids bind Eallo, act as COXs non-substrates, and when
their ratio with respect AA is equal or higher than 20, they behave as inhibitors of COX-1 and
activator of COX-2. Poly-unsaturated fatty acids bind Ecat of COX-1 but Eallo of COX-2 and this
could be the reason of the cardioprotective role of ω-3 fish oil fatty acids, particularly when ω-3
fatty acids/AA ratios are equal or higher than 5.0.
15

2.4
Molecular Modeling
A Structure Based Virtual Screening (SBVS), based on the Fingerprints for Ligands and Proteins
(FLAP) algorithm, was carried out as an attempt to rationalize the inhibition capabilities of the
novel synthesized compounds and the binding poses of selected compounds in the catalytic
pockets of both COX-1 and COX-2 enzymes were analyzed19.
The SBVS was performed using the X-ray crystal structures of ovine COX-1 in complex with
mofezolac (PDB code: 5WBE)19 or AA (PDB code: 1DIY)26 and of mouse COX-2 with
celeCOXib (PDB code: 3LN1)27, upon removal of the co-crystallized ligands. The COX-1 binding
poses of the mof-spacer-mof compounds (9, 13 and 15) were generated within the internal cavities
computed for the 5WBE structure; both COX-1 structures (5WBE and 1DIY) were used to
investigate the mof-spacer-AA compounds (19, 20 and 21) and, interestingly, similar binding
poses were obtained. As in our previous studies28,29, the crystallographic poses of mofezolac19 and
celeCOXib27 in the three-dimensional structure of COX-1 and COX-2, respectively, as well as that
of AA in COX-126, were also reported for comparison purposes (Figure S1-S3).
According to the FLAP poses of the mof-spacer-mof compounds (9, 13 and 15) within COX-1
(Figure 2), one of the two mofezolac moieties is always located at the top of the catalytic pocket.
The insertion of different flexible or rigid spacers induced significant changes in the interactions
of both the inner and the outermost mofezolac units. Overall, these compounds strongly interact
with (i) the residues Arg120/Tyr355 at the catalytic site entry channel, an enzyme zone crucial for
substrate/inhibitor recognition, (ii) the residues in the central hydrophobic region and (iii) the top
of the cavity, where the catalytic inducer Tyr385 is located.
16

a)
b)
c)
d)
e)
f)
Figure 2. Proposed 3D binding mode inside the active site of COX-1, determined by FLAP
analysis, for 9 (a, b), 13 (c, d) and 15 (e, f). Some of the COX-1 key residues located in the cavity
are highlighted in stick-mode.
The key hydrogen bond with Tyr355 occurred for all compounds, even though the H-bond
acceptor is the isoxazole nitrogen atom of the outer mofezolac for 9, the outer amide (NH) and the
isoxazole ring are the donor and acceptor unit, respectively, for 13 and only the amide carbonyl
(CO) for 15. The methoxy substituted phenyl ring of the outer mofezolac moieties (Figure S4)
interacts with an external region defined by two parallel α-helixes (one of the two helixes
17

containing also the Arg120). Hydrophobic and H-bond interactions (with the Tyr385), of both the
mofezolac inner moiety and the spacer moieties, inside the cavity contributed to obtain a favorable
binding for all the compounds.
A good fit inside the catalytic cavity and H-bond acceptor interactions, quantified by the FLAP H
and O probes respectively, relate to the compounds ability to act as COX-1 inhibitors. The
analysis of these parameters, in agreement with experimental data (Table 1), confirms that
compound 9 is the most active of the series while compounds 13 and 15 are almost equivalent
(H*O values are 0.62, 0.56, and 0.58, respectively), albeit with lower values than the reference
inhibitor mofezolac, whose H*O value is 0.70.
A deeper inspection of the binding poses (2D depiction in Figure 3 and Figure S5), allowed to
ascertain that the compound 9, bearing the flexible and shorter C2 unit spacer, displays the most
efficient interactions with both Arg120 and Tyr355 (Figure 3a); the biphenyl unit of 15 ensures the
best interaction with the central hydrophobic region defined by the residues of the two parallel α-
helixes that define the upper and the lower limits of the catalytic cavity (Figure 3b). In addition,
the external mofezolac unit of 15 is able to form a dense network of CH-π interactions between
the methoxy-substituted phenyl rings and different Val, Leu and Ile residues of the two-parallel
outermost α-helixes (Figure S4c). On the contrary, in the FLAP binding pose of 13, the inner
mofezolac moiety shifts towards the center of the cavity (Figure S5); this shift allows the
formation of a new H-bond interaction between Tyr355 and amide NH, but the outer mofezolac is
involved in less efficient CH-π interactions. Additional π-π interaction with Phe518, or other
aromatic residues, and H-bond are still possible for all the compounds.
18

a)
b)
Figure 3. Comparison of the 2D FLAP binding poses for 9 (a) vs 15 (b). The most important
COX-1 residues are shown together with their respective numbers. Purple regions indicate strong
interactions.
The FLAP poses of the ‘hybrid molecules’ 19, 20 and 21 within the COX-1 show that they
interact with the catalytic cavity through the fatty acid moiety (Figure 4), while the mofezolac is
always located in the outer region of the COX-1 active site. The features of the spacer linking the
mofezolac and AA, also in this series, guide the interactions established within the catalytic site,
resulting, then, in the enzyme activity inhibition.
a)
b)
19

c)
d)
e)
f)
Figure 4. Proposed 3D binding mode inside the active site of COX-1, determined from FLAP
analysis, for 19 (a, b), 20 (c, d) and 21 (e, f). Some of the COX-1 key residues located in the cavity
are highlighted in stick-mode.
Only compound 20 is able to realize (i) a good fit of the long AA chain inside the catalytic cavity,
(ii) strong H-bond interactions with Arg120 and Tyr355, and (iii) efficient interactions both with
the central hydrophobic area of the cavity and with the aromatic residues in the deepest region, i.e.
nearby the catalytic Tyr385 (Figure 5; the poses of 19 and 21 are shown for comparison in Figure
S6). It is possible to take advantage of the CH-π interactions between the central phenyl ring and
the residues of the two parallel α-helixes and, at the same time, to establish strong H-bond with the
outer amide linkage and the mofezolac-isoxazole. Moreover, as previously reported for 9 and 15,
the methoxy-substituted phenyl rings of the mofezolac unit realize a dense network of CH-π
interactions with the Val, Leu and Ile residues of the two-parallel outermost α-helixes.
20

Figure 5. 2D FLAP binding pose of compound 20. The most important COX-1 residues are
shown together with their respective numbers. Purple regions indicate strong interactions.
A check of the FLAP H and O probes, which correlated with the in vitro inhibitor behavior,
allowed us to rationalize that 20 is the most active of the series: the H*O values obtained through
the Structure Based Virtual Screening are 0.52, 0.50, and 0.47 for 20, 19 and 21, respectively.
It is worth noting that, when the AA moiety is removed (i.e. compounds 16, 17 and 18) the trend
observed for the inhibitory activity reverses and resembles what previously observed for ‘gemini
systems’ 9, 13 and 15, when one of the mofezolac and the spacer occupy the enzyme cavity. This
result further supports the inclusion of the substrate/AA moiety and the lower inhibition potency
observed for 19 and 21 than 20.
Among the synthesized compounds, only 9, 19 and 21 inhibit COX-2; although the mofezolac is
not a COX-2 inhibitor, the FLAP binding poses clearly indicate that this moiety is always
included, at variable depths, into the wider COX-2 cavity (Figure 6).
21

a)
b)
c)
d)
e)
f)
Figure 6. Proposed 3D binding mode inside the COX-2 active site determined by FLAP analysis
for 9 (a, b), 19 (c, d) and 21 (e, f). Some of the COX-1 key residues located in the cavity are
highlighted in stick-mode.
The compounds with the short and flexible spacer, namely 9 and 19, place the mofezolac unit at
the bottom of the cavity mainly due to a network of CH-π, S-π and π-π interactions (Figure 6 and
7) and the H-bond interaction of the isoxazole ring with Ser530. Some of these interactions can be
22

observed also in the crystallographic pose of celeCOXib (Figure S3). An additional network of H-
bonds, between Arg120/Tyr355 and both the amidic carbonyl (CO) and the outer isoxazole ring,
and CH-π interaction, of the external mofezolac, at the entry of the cavity reinforces the overall
interaction of compound 9 with the enzyme. For compound 19, only the amidic carbonyl (CO) is
hydrogen-bonded to Tyr355; moreover, the AA moiety (Figures 6c, 6d and S8) is too close to the
two-parallel outermost α-helixes that define the cavity entry causing steric hindrance.
The FLAP binding pose of 21 inside the COX-2 catalytic site (Figure 6e, 6f and 7) shows that
mofezolac has a different orientation but is able to achieve strong π-π and S-π interactions, with
Met 522 and Tyr 385; the latter residue and Ser 530 are involved in H-bond interactions with the
isoxazole ring and the inner amidic NH; the biphenyl unit is located in the central region of the
binding pocket and interact with Tyr355 and the residues that define the upper and lower cavity
walls. Unlike compound 9 and 19, the AA moiety of 21 establishes strong CH-π interaction with
the external parallel α-helixes and strong H-bond with Arg120: these interactions are key points
for the inhibitory activity of this compound (Figure S9-11).
a)
b)
Figure 7. Comparison of the 2D FLAP binding poses for compounds 9 (a) vs 21 (b). The most
important COX-2 residues are shown together with their respective numbers. Purple regions
indicate strong interactions.
23

Hydrophobic interaction and H-bond acceptor interactions, quantified by the FLAP probes Dry
and O respectively, are strictly related to the COX-2 inhibition activity; the analysis of these
parameters confirms that 19 is the less active of the series, whereas compounds 9 and 21 are
almost equivalent (Dry*O values are 0.70 and 0.77, respectively) (Figure 7).
The Structure Based Virtual Screening clearly indicates that small structural differences lead to
large variations in the inhibition activity, not only towards the single COX isoform but also when
comparing the activity of the same substrate towards the two COX isoforms. This is not
surprisingly as the different catalytic activity and substrate selectivity of COX-1 and COX-2 are
determined by a single amino acid difference in the active site (valine 509 to isoleucine) and a
series of differences at the entry channel of the active site28.
Steric effects, hydrophobic contacts (evaluated by the probe H, Dry and O) and H-bond acceptor
sites of the compounds with the enzyme key regions help us to clarify the inhibition activity
observed. For example, the biphenyl spacer in compound 15 shows good potency towards COX-
1, but is not tolerated in the hybrid inhibitor 21; the selectivity is reversed when considering the
inhibition activity towards COX-2. The inhibition activity of 15 towards COX-1 is mainly
determined by interaction within the smaller cavity and at the catalytic site entry channel; it is
not possible to define a comparable interaction towards the COX-2 isoform as the small
mofezolac unit inside does not realize a god fit with the key region of the catalytic cavity while
the outer mofezolac unit experiences large steric hindrance with the extended entry mouth.
Hybrid molecules 19-21 are able to inhibit both COX isoforms but the overall effect derives from
different interaction: the arachidonic acid moiety is always included in the cavity of the COX-1
while the mofezolac unit occupy the COX-2 catalytic site. The phenyl moiety in compound 20
ensures the right complementarity (shape and interaction), then the highest inhibition, towards
COX-1. Only the biphenyl spacer of compound 21 allows effective interactions with the catalytic
24

site and the mouth of the cavity fo the COX-2 isoform, at the same times the interaction of the
arachidonic acid with enzyme entry channel operates in a synergic way; taken together, these
effects determine the reversed selectivity.
Compounds 13, 15, 20-21 were further characterized by evaluating their contribution in platelet
aggregation and blood coagulation and their hemolytic and genotoxicity effects as a continuation
of preceding investigations, mofezolac and its analogues have already been described as an
attempt to target diseases such as multiple myeloma29-31, thrombosis32, and ovarian cancer33-35
.
2.5
Inhibition of AA-induced Platelet aggregation
The IC50 values (Table 3) show that all compounds were able to efficiently inhibit platelet
aggregation. In particular, mof-C6H4-C6H4-mof (15) shows an IC50 = 0.44 μM superimposed with
mofezolac (IC50 = 0.45 μM). All other compounds 13, 2 and 20 have a comparable micromolar
activity revealing the potency of these antiplatelet compounds to inhibit the arachidonic acid
pathway of platelet aggregation and confirm their ability to reduce the TXA2 production by COX-
1 inhibition as a mechanism to impair normal platelet aggregation as for aspirin (IC50 = 1.11
μM)31.
Table 3. Antiplatelet profile of compounds on in vitro platelet aggregation of human citrated
platelet rich plasma induced by arachidonic acid (AA).
Compound
mof-C6H4-mof (13)
IC50 (μM)a
0.91 ±0.28
0.44 ± 0.13
0.61 ± 0.24
0.76 ± 0.16
0.45 ± 0.04
1.11 ± 1.17
mof-C6H4-C6H4-mof (15)
mof-C6H4-AA (20)
mof-C6H4-C6H4-AA (21)
mofezolac (1)
Aspirin
25

aIC50 = concentration at which 50% inhibition of platelet aggregation is observed.
2.6
In vitro coagulation assays
Neither the extrinsic nor intrinsic pathways of the coagulation cascade were affected by the
presence of the novel isoxazoles, suggesting that their antithrombotic profile relies on the direct
impairment of platelet aggregation, thus differing from dual action molecules (Table 4). This
result reveals a small bleeding risk in comparison to dual acting molecules whose mechanism
often leads to severe bleeding disorders36.
Table 4. In vitro coagulation of human plasma (n = 3) activated by prothrombin time (PT) and
thromboplastin time (APTT).a
Compound (100µM)
PT (seconds)
13.9 ± 0.8
300 ± 0.0
15.5 ± 0.0
15.1 ± 0.3
15.3 ± 1.0
14.6 ± 0.5
14.2 ± 0.3
14.6 ± 1.0
APTT (seconds)
44.8 ± 1.5
300 ± 0.0
C -
C +
mof-C6H4-mof (13)
mof-C6H4-C6H4-mof (15)
mof-C6H4-AA (20)
mof-C6H4-C6H4-AA (21)
mofezolac (1)
47.5 ± 0.4
46.5 ± 1.1
55.6 ± 1.2
48.0 ± 1.6
50.8 ± 0.23
50.3 ± 0.3
Aspirin
a(C-) = negative control: DMSO 1%. (C+) = positive control: 100 µM rivaroxaban.
2.7
Hemocompatibility to healthy human erythrocytes
The hemocompatibility profile of some representative novel isoxazoles was determined based on
the degree of erythrocyte lysis observed in the 3h incubation time at 37°C. The hemolysis values
26

obtained were zero (%) or close to it, confirming the acceptable toxicity profile of the compounds
tested (Table 5). Hemolysis values below 10% are considered non-hemolytic37.
Table 5. Hemolytic profile of some representative novel isoxazoles through hemolysis assay.a
Compound (100 µM)
Hemolysis (%)
0.00 ± 0.22
100 ± 2.01
0.00 ± 0.04
0.00 ± 0.08
0.00 ± 0.01
0.14 ± 0.08
0.00 ± 0.02
0.0 ± 0.04
C-
C+
mof-C6H4-mof (13)
mof-C6H4-C6H4-mof (15)
mof-C6H4-AA (20)
mof-C6H4-C6H4-AA (21)
mofezolac (1)
Aspirin
a(C-) = 1% DMSO is the negative control; 1% Triton X-100 is the positive control (C+). Values
below 10% are considered non hemolytic.
2.8
Cytotoxicity activity
The cytotoxicity test in Vero cell line was performed by MTT reduction assay at 10 and 100 μM
for 48 hours (Table 6). Among all the tested compounds, none of them was considered toxic to
Vero cells, even at higher concentrations. Significantly statistical reduction (p-value ≤ 0.05) of cell
viability values was not considered in this test, since the results above 70% are biologically
acceptable according to ISO 10993:5 (2009). This result represents a starting point for advanced
pre-clinical studies, given the correlation between toxicity in Vero cells, isolated renal tissue and
nephrotoxicity in rats38.
Table 6. Vero cells viability after 48 hours of exposition to the tested novel compounds at 10 and
100 µM.a
27

Compounds
[µM]
Cell viability (%)
± S.D. (%)
Control
-
100
97
2.6
2.8
0.5
2.8
4.8
3.2
2.8
3.1
4.2
4.6
3.7
2.0
2.8
100
10
mof-C6H4-mof (13)
mof-C6H4-C6H4-mof (15)
mof-C6H4-AA (20)
mof-C6H4-C6H4-AA (21)
mofezolac (1)
98
100
10
92
100
79
100
10
93
100
10
84
94
100
10
91
97
100
10
95
Aspirin
100
aData are presented as mean of cell viability values (n = 4) ± standard deviations.
2.9
Genotoxicity Tests
The safety of the novel isoxazoles is reinforced by the genotoxicity tests which revealed no
mutagenicity profile against Salmonella typhimurium auxotroph mutant strains through reverse
mutagenesis and histidine prototrophy (Ames test), compared to the positive control 4-nitro-
quinoline 1-oxide (4NQO), nor genotoxicity evaluated against Escherichia coli through the SOS
chromotest. This low risk profile is observed even at the highest concentrations tested (500 μM,
Table 7), revealing the safety profile of the series as promising “lead prototypes”.
Table 7. Mutagenic and genotoxic activity of molecules without metabolic activation.a
Ames Test
SOS chromotest
28