Discovery of Bis-sydnone styryl ketone as a selective cyclooxygenase-2 inhibitor

Various literature sources have documented a wide spectrum of therapeutic properties of sydnones including anti-inflammatory, anticancer, antimicrobial activities. Phenyl styryl ketones and their derivatives as members of the chalcone family have also been reported as significant bioactive molecules. The current study was initiated to evaluate the anti-inflammatory activity of sydnone-based compounds including some novel bis-sydnone styryl ketone hybrids. Twenty-five sydnone-containing compounds were successfully synthesized. Compounds 46-48 and 56-58 were reported as new sydnone derivatives. Whereas, compounds 61-63 were synthesized as novel molecules containing two sydnone rings linked via α,β-unsaturated ketone. The structures of the synthesized compounds were confirmed by FTIR, 1H NMR, 13C NMR and ToF-MS analyses. The in vitro COX inhibition assay showed varied activity. Compounds 47, 51, 58 and 63 showed the most potent COX inhibitory effects at a concentration of 200 μM. The selectivity index revealed that 63 was the best selective COX-2 inhibitor. Acetylation of the sydnone ring at C-4 was fruitful for the COX inhibitory effects. Docking analysis showed that COX-2 selectivity was due to a favourable positive charged interaction occurring between the sydnone ring of 63 and Arg513 of COX-2. Compound 51 was hydrogen bonded to Arg513. On the other hand, the low inhibitory effect of 63 against COX-1 was due to an unfavourable polar interaction with His513 in the binding pocket of COX-1. The compounds were successfully synthesized and characterized. Compound 63 had a common architecture and pharmacophoric features with known selective COX-2 inhibitors (the coxib family) which make it a suitable candidate for the designing of selective and safe NSAID.


Background
Sydnone can be defined as a five-membered pseudoaromatic heterocyclic molecule. Oxadiazoles, classically 1,2,3-oxadiazole, form the main skeleton for sydnone (Fig. 1). The molecule has delocalized balanced positive and negative charges. The positive charge is shared between the five annular atoms, but mainly located at the nitrogen atom N3. An enolate-like exocyclic oxygen atom bears the negative charge. Studies on sydnone structure showed that there is no single polar or covalent structure able to represent the sydnone ring satisfactorily. The hydrogen atom at the position C4 was demonstrated to have acidic and nucleophilic functionalities making the sydnone ring reactive towards electrophilic reagents [1].
The history of sydnone dates back to the early nineteenth century when Earl and his colleagues reported the formation of the sydnone ring as the result of the effect of acetic anhydride on N-nitroso amino acids [1,2]. It was well documented that sydnone-containing compounds have a broad spectrum of biological activities including antitumor, antidiabetic, antimicrobial, antioxidant and anti-inflammatory [3]. The exploration of the bioactivity of sydnones appended to other molecules was explored. Structure-activity relationship (SAR) studies were mainly directed towards the adjunct structures like thiazole, benzothiazole, pyrimidine and sulphonamide overlooking the unique sydnone ring [4,5]. The anti-inflammatory activity of sydnones was initially published by Wagner and Hill who reported 3-(2arylthioethyl)sydnone and 3-(2-arylsulfoxyethyl)sydnone as efficacious anti-edema agents in rats [6,7].
Considering these issues, we undertook this research to cover the lack of consistent SAR data on activity variations resulted from structural modification on the sydnone ring itself. We hypothesize that joining two sydnone rings through a phenyl styrylketone linker will produce selective cyclooxygenase-2 (COX-2) inhibitors by giving the new molecules common architecture and pharmacophoric features with known selective COX-2 inhibitors (coxib family). We hypothesize that the nucleophilic hydrogen atom at C4 of the sydnone ring, the various substituents to the phenyl ring at N3, and connecting two sydnone rings via a phenyl styrylketone linker have a crucial impact on the COX-2 inhibitory activity of sydnone-based compounds.

Chemical synthesis
All chemicals and solvents were purchased from commercial suppliers and used without purification. Reactions were monitored using thin-layer chromatography (TLC) performed on Silica Gel 60 F 254 plates (Merck) using a suitable mobile phase system. Column chromatography was carried out on silica gel (60-200 mesh) supplied from Merck. Melting points (Celsius,°C) were uncorrected and determined using open-end capillary tubes on Stuart Scientific apparatus. Infrared spectra (IR) were obtained via potassium bromide (Merck, IR grade) discs using Perkin-Elmer Spectrum Two FT-IR instrument. 1 H-NMR (700 MHz) and 13 C-NMR (76 MHz) spectra were recorded at 298 K on a Bruker Ascend spectrometer equipped with 5 mm TCI inverse detection cryoprobe. Chemical shifts were recorded in parts per million (ppm, δ) relative to tetramethylsilane (TMS) as an internal standard. High-resolution mass spectrometry (HRMS) was conducted using electrospray ionization techniques (ESI) by Bruker micrOTOF-Q II.

Synthesis of methyl 4-aminobenzoate 12
The general protocol of Fisher's esterification was followed to prepare methyl 4-aminobenzoate 12 (Fig. 2). In 50 mL (1.26 mol) hot methanol, 0.105 mol paraamino benzoic acid 11 was dissolved. Concentrated sulphuric acid (2.50 mol) was added dropwise with continuous cooling. The resulting white paste was heated at 90-95°C for 1 h. The reaction mixture was cooled down to ambient room temperature, then neutralised with an aqueous solution of sodium carbonate 5% w/v. The white solid was then collected and recrystallised from ethanol.

Synthesis of nitro-containing N-arylglycine 22-25
An aqueous solution of chloroacetic acid (0.2 mol) was neutralised with an equimolar of NaOH and then added to a boiling suspension of nitro-containing aniline derivatives 18-21 (0.1 mol) over a period of 6 h (Fig. 2). Reflux was continued for another 72 h with stirring. The black residue was then removed by vacuum filtration while hot, and the filtrate was kept in the fridge overnight. The resulting yellow solid was collected and gently refluxed with ammonia (20% v/v) for 30 min. After cooling to room temperature, the unreacted initiating material was removed via filtration. The filtrate was acidified with concentrated hydrochloric acid [9]. The yellow product was vacuum-filtered and recrystallised from ethanol to yield the corresponding amino acids 22-25.

Synthesis of N-nitroso-N-arylglycine 27-36
To a suspension of the prepared N-arylglycine (0.04 mol) in 100 mL cold water, one drop of concentrated hydrochloric acid was added. In 20 mL of cold water, 0.12 mol sodium nitrite was dissolved and added dropwise with vigorous stirring at 0-5°C over a period of 1 h (Fig. 2). Stirring at ambient room temperature continued for 0.5-4 h until a clear liquor solution was achieved. The concentrated hydrochloric acid was then added slowly, and the precipitated nitroso derivatives were vacuum-filtered [8]. The products 27-36 were used for the following steps without further purification.

Synthesis of 3-arylsydnones 37-46
In 15 mL dichloromethane, 0.015 mol of the prepared nitroso compounds 27-36 and 0.022 mol of trifluoroacetic anhydride (TFAA) were stirred at 0-5°C until complete dissolution as shown in Fig. 3 [11]. Next, the dichloromethane was evaporated under a nitrogen stream, and the resultant pink solid was washed with 5 mL sodium carbonate (10% w/v) and then recrystallised from ethanol 95% to produce the corresponding sydnones 37-46.

Synthesis of alcohol sydnone derivatives 47
A mixture of sydnone 46 (0.018 mol) and sodium borohydride (0.052 mol) was heated until reflux in 190 mL of tert-butanol (Fig. 3). Then, 20 mL methanol was added dropwise for 30 min with reflux continuing for a further 2 h. After cooling to room temperature, 200 mL water was added. Fractional distillation evaporated the remaining methanol and t-butanol as azeotrope mixture [10]. The remaining liquor aqueous solution was kept in the fridge overnight to precipitate the alcohol sydnone derivative as a dark red solid which was recrystallised

Synthesis of aldehyde sydnone derivatives 48
A suspension of 47 (0.02 mol) and MnO 2 (0.14 mol) in 360 mL dichloromethane was stirred at room temperature for 12 h (Fig. 3). The reaction mixture was then filtered to remove the remaining starting materials [10]. Dichloromethane was evaporated under reduced pressure, and the resulting crude product was recrystallised from ethyl acetate to afford 3-(4-formylphenyl)sydnone 48.

Synthesis of 4-acetyl-3-arylsydnones 49-58
In 20 mL of dry benzene 0.00375 mol of sydnones 37-48 and 0.0105 mol of phosphorous pentoxide were heated until reflux. Then, 0.600 mL glacial acetic acid was slowly added during the next 20 min with vigorous stirring (Fig. 3). The reaction mixture was then refluxed for a further 5 h under a nitrogen condition. The benzene layer was decanted, and the black residue was washed with 3 × 20 mL hot benzene. The combined benzene extracts were dried to produce the crude 4acetylated sydnone [12]. Compounds 49-53 were purified by recrystallisation from warm ethanol (60°C). Compounds 54-58 were eluted through a column of silica gel 60 (0.063-0.200 mm) using ethyl acetate and hexane (1:1) as a mobile phase to remove the unreacted initiating materials, followed by recrystallisation from boiling ethanol. Compounds 59 bearing -CH 2 OH at the para position of the benzene ring (corresponding to sydnone 47) and 60 bearing -CHO at the para position of the benzene ring (corresponding to sydnone 48) were not found.

Molecular docking analysis Ligand and protein preparation
The 2D structures of 37-58 and 61-63 were sketched using the freeware ACD/ChemSketch version 12.01 (ACD/Labs Release, Canada). These were then converted into a standard structure-data file (SDF) format using the freely available open source toolbox, Open Babel 2.3.2. Energy minimisation was conducted using optimised potential for liquid simulations (OPLS3) force field [15] using the Lig Prep Module provided by Maestro (Schrödinger, LCC, New York, 2017). The 3D structure of COX-1 (PDB code: 3N8Z, 2.9 Å) and COX-2 (PDB code: 3NT1, 1.73 Å) were retrieved from Protein Data Bank (PDB) and processed using the Protein Preparation Wizard from Maestro. Protein preparation was executed for chain A only, while the other chain and non-proteinic molecules like heme, water and arachidonic acid were removed. Glide Receptor Grid Generation platform as a part of Maestro, was employed to create a rigid grid around the interacting residues of the active site from the prepared protein structure [16].

Extra precision molecular docking and binding energy calculation
To predict the binding pose of ligands 37-58 and 61-63 with COX-1/COX-2, the extra precision (XP) docking procedure was applied using Glide workflow from Maestro [17]. The ligands were docked flexibly into the rigid docking box. Internal conformations like ring conformers, acyclic bond torsion and nitrogen inversion were generated during the docking process. For each ligand, one docking hit with the lowest glide docking score was generated and used to study the docking results. The docking poses generated from Glide XP docking were used as an input to calculate the binding energy. The binding free energy calculation was performed using Prime Molecular Mechanics/Generalised Born Surface Area MM-GBSA (Prime, Schrödinger, LCC, New York, 2017). All atoms of the receptor and ligands were optimised and minimised using VSGB 2.0 solvation model [18]. The binding energies were calculated by OPLS force field [15].    Supplementary data (Spectrum S117-S120).
Biological evaluation (in vitro cyclooxygenase inhibitory assay) The in vitro cyclooxygenase (COX-1/COX-2) inhibitory assay was performed for all sydnone  and bissydnone styryl ketone derivatives (61-63) and presented in Table 1. The two reference drugs were SC-560 as a selective COX-1 inhibitor [19] and DuP-697 as a selective COX-2 inhibitor [20]. Both 3-arylsydnone and 4acetyl-3-arylsydone showed moderate to low inhibitory activity against both COX isozymes. The calculated selectivity indices ranged from 0.29 to 1.27. Consequently, none of the compounds 37-56 can be considered a promising selective or non-selective COX inhibitor. Nevertheless, the findings of the study disclosed that acetylation of the sydnone ring at C-4 had increased the inhibitory activity of most of the tested compounds against COX enzymes. For example, 51, 54, 55, 57 and 58 inhibited more than 50% of at least one COX isoform while their non-acetylated analogues exhibited inhibitory activities in the range of 20-40%. Therefore, it can be speculated that the presence of an acetyl group at C-4 contributes in the interaction of the sydnone ring with COX active site. Noticeably, compound 63 at a concentration of 200 μM was the most potent COX-2 inhibitor suppressing 89.45% of the enzyme activity. On the other hand, it inhibited only 33% of COX-1 activity. Consequently, the calculated selectivity index of 63 was 2.70. On the contrary, 61 and 62 were almost inactive against COX-2 while they inhibited 44% and 73% of COX-1, respectively. The results thus revealed the importance of combining two sydnone molecules via a styryl ketone linker as a new selective COX-2 inhibitor, provided that the terminal phenyl ring bears an electron-withdrawing group like a chlorine atom.

Molecular docking analysis
Findings of the current study showed that the extra precision docking scores of 37-58 and 61-63 with COX-2 ranged from − 4.134 to − 1.384, and the binding energy MM-GBSA ranged from − 43.827 to − 5.336 kcal/mol (Table 2). Apparently, the acetylated sydnones 49-58 exhibited a lower binding energy than their non-acetylated counterparts revealing a more stable binding of 4-acetyl-3-arylsydnone with COX-2. Consequently, the existence of the acetyl group at the carbon C-4 of the sydnone ring is important for COX-2 inhibitory activity. Compound 63 had the lowest MM-GBSA value which came in accordance with its potent in vitro COX-2 inhibitory activity. In contrast, the inactive sydnone analogues against COX-2 like 37, 41 and 62 ranked high in terms of the binding energy with COX-2.
On the other hand, the docking scores of the studied compounds with the active site of COX-1 were − 5.524 to − 1.694 and the calculated MM-GBSA ranged from − 30.664 to − 1.831 kcal/mol as shown in Table 2. Except for compounds 49, the acetylated sydnones had a lower binding energy than that of the non-acetylated ones suggesting a high affinity of 4-acetyl-3-arylsydnone to COX-1. Our finding displayed a strong correlation (r 2 = 0.83) between the calculated MM-GBSA and the in vitro COX-2 inhibition by AChE ELISA, while docking scores were not well correlated to the inhibitory activity (Fig.  5a). Likewise, a moderate linear correlation (r 2 = 0.63) was observed between MM-GBSA and the in vitro activity against COX-1 (Fig. 5b).
Analysis of the binding poses disclosed that the sydnones of the present study oriented similarly in the COX-2 binding cavity; i.e. binding with residues from the MBD and the CD (Tables 3 and 4). Moreover, the binding modes were close to that of celecoxib, especially the interaction with Arg513 which is pivotal for selective COX-2 inhibition. However, celecoxib binds noncovalently with Arg513 through positive charge interaction between the sulphonamide and guanidinium. Sydnones interact with Arg513 via different types of noncovalent interactions like positive charged, π-π, πcation and salt-bridging. Access of celecoxib and sydnones to the adjunct pocket were through hydrophobic interaction with Tyr91. Other types of interactions like polar, hydrogen bonding and negative/positive charges with other residues in the binding site were also involved. Interestingly, the acetyl function at C-4 of the sydnone ring contributed as a hydrogen bond acceptor through its carbonyl oxygen atom with residues at the top of the COX-2 binding site.
Compound 63 (bis-sydnone styryl ketone) had the lowest MM-GBSA value (− 34.763 kcal/mol) with a considerable low docking score with COX-2 suggesting a good binding pose and a stable ligand-enzyme complex as displayed in Fig. 6a and supplementary data S121, Tables 3 and 4. It was found that one sydnone ring and the styryl ketone moiety bound at the MBD (residues 75-115) and the mouth of the catalytic region, while the other sydnone and the terminal benzene fit inside the catalytic domain. At the MBD and the mouth of the catalytic domain, the following interactions were also observed: Tyr91 had a hydrophobic interaction and faceto-face π-π interaction with the benzene ring, edge-toface π-π interaction with the sydnone ring, and π-cation interaction with sydnone N-3 atom. The carbonyl of the styryl ketone had polar interactions with both Asn87 and Thr88. In the catalytic domain, 63 had the following interactions: The benzene had π-cation interaction with Lys511 and hydrophobic interactions with Pro512 and Pro514, and the sydnone ring had a positively charged interaction with Arg513. The chlorine atom had a negatively charged interaction with Glu510 and Glu520 at the core of the catalytic site and a hydrophobic  interaction with the side pocket residue Tyr475 ( Fig. 6a and supplementary data S121). Likewise, 4-acetyl-3-(4methoxyphenyl)sydnone 51 and 4-acetyl-3-[4-(methoxycarbonyl) phenyl]sydnone 58 had remarkably low MM-GBSA values of − 30.737 and − 29.728 kcal/mol ( Table  2) with binding poses like that of compound 63. However, compound 51 binds essentially at the base of the catalytic domain with fewer interactions inside the active site ( Fig. 6b and supplementary data S121).
The binding pose of celecoxib as a selective COX-2 inhibitor was also investigated (Fig. 6c, Tables 3 and 4). The sulphonamide group of celecoxib had hydrogen bonds with the side chain of Asn87 and the carbonyl function of Lys511. The methyl group and benzene rings had hydrophobic interactions with Tyr91 of the MBD and Tyr475, Pro512 and Pro514 of the catalytic domain. Moreover, positively charged interactions with Lys511 and Arg513 were also observed. The residues Glu480, Glu510 and Glu520 had negatively charged interactions. Many residues at the MBD such as Asn87, His90, Thr94 and His95 were also involved in polar interactions with celecoxib atoms.
Analysis of the binding poses of sydnones with COX-1 revealed varied binding hits. Compounds 37-41 had a revealed that its orientation inside the binding site of COX-1 is an antipode to its binding with COX-2 (Fig. 7a). The chlorine atom oriented towards the MBD, whereas the styryl ketone moiety was located inside the catalytic domain. Of interest, 51 had a binding free energy value of − 12.93 kcal/mol with a relatively high docking score of − 1.694 which means it did not fit well inside the binding pocket of COX-1. As shown in Fig. 7b, the carbonyl of the acetyl group had a hydrogen bonding with Ser87. The residue Lys511 of the catalytic region had a π-cation interaction with the phenyl ring and a positively charged interaction with N-3 of the sydnone ring. Hydrophobic interactions (Cys512 and Pro514), polar interactions (Ser87 and His513) and positively charged interactions (Glu520) were also observed. The low docking scores revealed that these interactions are disfavoured.

Structure elucidation of compounds 61-63
Claisen-Schmidt's condensation is an aldol condensation where an aldehyde/ketone possessing an α hydrogen reacts with another carbonyl compound lacking the α hydrogen to produce an α,β-unsaturated product. The condensation usually requires the presence of a strong base as a catalyst, even though some reported a successful acid-catalysed condensation [21]. In the current research, we applied the base-catalysed aldol condensation approaches to join the 4-acetylated sydnones (49, 50 and 52) with the aldehyde derivative 48 in order to produce bis-sydnone styryl ketone molecules. Consequently, three compounds 61-63 were successfully synthesized, purified and characterized. Because 61-63 were considered to be novel molecules, a comprehensive structural elucidation was performed using FTIR, 1 H NMR, 13 C NMR and MS spectra. To make the peak-atom assignments, the obtained spectral data was compared to the most similar analogues found in the literature like phenyl styryl ketones [22,23] and mono-sydnone styryl ketones [13,24]. The FTIR data was summarized in Table 5 and supplementary data (S109, S113 and S117). Interestingly, the appearance of three strong bands in the carbonyl region delineated the existence of three carbonyl groups owing to the two sydnones and the styryl ketone. The first two bands (≈1780 and 1760-1740 cm −1 ) distinctive of the sydnone carbonyl were at high wavenumbers due to the sydnone ring's strain [1]. The styryl carbonyl was shifted to a lower frequency due to the conjugation with the π electron systems of the styryl and the sydnone [22,23]. Further structural confirmation was made on the basis of sp 2 C-H stretches of the sydnone and styryl moieties which occurred within their expected ranges. The sydnone C4-H appeared at high wavenumbers (3170-3100 cm −1 ) while the styryl C-H appeared about 3100 cm −1 . Because the out-of-plane bending of the sp 2 C-H is of great importance for determining the substitution pattern of the benzene ring and the styryl compounds, they were well examined. First, the strong band about 1030 cm −1 indicated the formation of trans substituted styryl double bonds [23]. Second, the appearance of one band in the range of 850-800 cm −1 revealed that all benzene rings of compounds 62 and 63 were para-disubstituted. Last, one strong band at 835 cm −1 and another two bands at 765 and 695 cm −1 confirmed the existence of para-disubstituted and mono-substituted benzene rings in the structure of 61. Analysis of the 1 H NMR spectra of compounds 61-63 (Table 6) revealed the existence of sydnone protons, styryl protons and benzene protons within the expected ranges when compared to the spectral data of the structurally-related analogues like styryl phenyl ketone [23], styryl naphthyl ketone [22] and mon-sydnone styryl ketones [13,25]. Interestingly, the styryl protons of 62 and 63 appeared as doublets with coupling constants ≈16 Hz denoting the formation of trans conformers [23]. The splitting pattern of the styryl protons of 61 was not clear because they were overlapped with the aryl protons. However, the two bulky substituents favour the formation of trans isomer. Therefore, we surmise that the aldol condensation of the acetylated sydnones and the aldehyde sydnone derivative resulted in bis-sydnone trans-styryl ketones.
Additional structural elucidation was performed using 13C NMR spectroscopy (supplementary data S111, S115 and S119). The obtained spectra displayed three peaks in the carbonyl region. The first peak had an almost constant chemical shift about 174 ppm in the three spectra and it was assigned to the carbonyl of the styryl ketone C14 [23]. The second peak at 168.92 ppm was also stable and it was assigned to the sydnone carbonyl C5 because the chemical environment of this carbon atom was nearly identical in the three compounds. The third carbonyl peak at ≈166 ppm was assigned to the sydnone carbonyl C5′ by a simple comparison with the acetylated sydnone counterparts (49, 50 and 52). The sydnone carbon atom C4 appeared at ≈95 ppm in the three spectra which was in accordance with the literature reported values for this sydnone characteristic peak. Similarly, the carbon atom C4′ appeared deshielded at ≈108 ppm because of the substituted acetyl group [26][27][28][29][30]. Interestingly, the chemical shifts of C12 and C13 of the styryl double bond were also nearly similar in the three compounds. The carbon C12 was the most deshielded because of the π electron systems of the styryl and the benzene and it appeared at 142.5-141.5 ppm in the three spectra which came in agreement with the reported values for the carbon β of other styryl ketone derivatives; 148-141 ppm. Moreover, the carbon atom C13 appeared at ≈124 ppm which well matched the literature spectral chemical shifts of 125-121 ppm for the carbon α [13,23]. The carbon atoms of the two benzene rings C6-C11 and C6′-C11′ were assigned to the corresponding peaks by comparing the 13 C NMR spectra of 61, 62 and 63 with those of the corresponding acetylated analogues 49, 50 and 52. In vitro cyclooxygenase inhibitory activity The anti-inflammatory activity of sydnone-based derivatives dates back to the 70s when researchers reported 3-(2-arylthioethyl)sydnone and 3-(2-arylsulfoxyethyl)sydnone as potent anti-inflammatory drugs [6,7]. Later, more experiments were undertaken to prepare other anti-inflammatory agents containing sydnone scaffolds like 4-(4-thiazolyl)sydnone [31]. However, studies on the anti-inflammatory activity of sydnones were conducted in vivo, and they were reported as anti-inflammatory, analgesic, antipyretic and anti-arthritis agents without exploring the effect of sydnones on the main enzyme involved in the inflammation process; cyclooxygenase [32,33]. Because the pharmaceutical inhibition of COX enzymes can provide relief from inflammation and pain, we initiated the current study to evaluate the inhibitory activity and the selectivity of various sydnone derivatives against both COX isoforms.
The outcomes of the present study showed that sydnone ring is not expected to form a general structure type for designing a potent COX inhibitor. On the contrary, the literature sources reported many sydnone derivatives as efficacious anti-inflammatory agents. A displayed in Fig. 8, a deep insight into structures I-IV which were delineated in the literature as effective antiarthritis and anti-edema agents [6,31,33] concludes the fact that their bioactivity cannot be attributed to 3phenylsydnone.
Other studies reported antiinflammatory effects of triheterocyclic thiazoles [34] and sulfoxide-containing compounds [34,35]. Thereupon, thiazole, benzophenone oxime and sulfoxide are the main acting pharmacophores of I-IV. This discussion explains the reason for the moderate to low COX inhibitory activity of 37-58 which had the sydnone as the main scaffold.
Therefore, we attempted to improve the antiinflammatory properties of sydnones by joining two sydnone rings using an α,β-unsaturated ketone linker (styryl ketone). The new compounds 61-63 were designed to have a common architecture and pharmacophoric characters with some selective COX-2 inhibitors as highlighted in Fig. 9. As expected, the new molecules exhibited potent to moderate inhibitory activity at a concentration of 200 μM against at least one of the COX isozymes. Noteworthy, we found that compound 63 selectively inhibited COX-2 with minimal effect on COX-1. Some studies had already demonstrated the anti-inflammatory activity of phenyl styryl ketone (chalcone). These researches showed that 3,4-dihydroxychalcone derivatives had a potent antiinflammatory activity due to 5-lipoxygenase and cyclooxygenase inhibition with IC 50 values in the μM range [36]. Other researchers reported that some chalcone derivatives significantly reduced prostaglandin production in mouse peritoneal macrophages due to the inhibition of COX-2 synthesis. They demonstrated that chalcone derivatives did not affect cell viability indicating a low cellular toxicity [37]. More recently, Deshpande and his co-researchers reported the synthesis of sydnonyl styryl ketone as potential non-steroidal anti-inflammatory agents. They proved that these molecules had a topical and systematic anti-inflammatory activity with less ulcerogenic side effects. The most significant activity was associated with the existence of electron-withdrawing substituents at the para position of the benzene ring [13].

Molecular docking analysis
Cyclooxygenase (COX) or prostaglandin endoperoxide H synthase has two isozymes, commonly referred to as COX-1 and COX-2, each possessing different physiological effects. Mainly, COX-2 is highly expressed in stimulated macrophages, fibroblasts, osteoblasts, synoviocytes, endothelial cells, smooth muscle cells, central nervous system, cells of the female reproductive system and mast cells as a response to inflammatory mediators like LPS, IL-1, TNFα and INFγ. On the other hand, COX-1 is regularly present in most body tissues and is involved in normal homeostasis [38]. There is a high degree of amino-acid sequence similarity between both isozymes. The tertiary and quaternary structures of both COX isoforms are identical with some minor differences that are considered essential for selectively inhibiting either one. The COX molecule is a homodimer with a molecular weight of 67-72 kDa. The monomer of both isoforms consists of three distinct regions: The Nterminal epidermal growth factor-like module (EGF) and the membrane binding domain (MBD) lying at the base of the third region which is the catalytic domain (CD). The CD contains peroxidase and COX active sites [39]. The CD is the largest region, and its structure is virtually similar between COX-1 and COX-2 (RMS deviation of 0.4 Å). The residues Arg120, Tyr355 and Glu524 form Table 5 FTIR spectral data of compounds 61-63 * Supplementary S109. ** Supplementary S113. *** Supplementary S117 a [2,13,22,23,25] Abdualkader et al. the entrance of the active site, while Tyr385 and Tyr348 lie at the apex of the active site and they are hydrogen bonded. The main differences between both active sites are at the positions 434, 513 and 523. In other words, Ile434 and Ile523 in COX-1 are replaced by valine residues in COX-2. The hydrophobic His513 in COX-1 was superseded by the positively charged arginine in COX-2. These differences resulted in altering the shape, size and the chemical environment of the active site of COX-2, and consequently, they are the key points of selective inhibitors [39,40]. Binding energy ΔG bind was widely used in the literature as a predictive tool to compare the potency of congeneric inhibitors against a specific enzyme. The free binding energy ΔG bind was calculated from the formula: where ΔE MM is the difference in energy between ligandenzyme complex and the sum of the energies of the ligand and free protein, ΔG solv is the difference in the GBSA solvation energy between the complex and the free ligand and protein and ΔG SA is the difference in the surface area energy between the complex and the sum of the surface area energies for the ligand and free protein [41]. The linear relationship between the virtually estimated ΔG bind and the experimental activity was well established for structurally similar data set. For example, Xie and his co-researchers reported a strong linear relationship (r 2 = 0.87) between the experimental IC 50 values of benzodiazepines and the calculated MM-GBSA of their complexes with gammaaminobutyric acid (GABA) receptor type A [42]. Moreover, MM-GBSA was 87-85% able to correctly rank the potency of checkpoint kinase 1 (CK1) inhibitors [16]. In the present study, it was found that the calculated binding energy values were able to explain 83% and 63% of the variation in the activity against COX-2 and COX-1, respectively. In other words, the in vitro COX-1/COX-2 inhibitory findings were well correlated with the in silico predicted binding poses of sydnones 37-63.
The residue 513 plays a paramount role in distinguishing the structural differences between the active sites of COX-1 and COX-2. Arginine forms a stable positive charge in the core of the COX-2 active site while histidine forms an exclusively hydrophobic pocket in COX-1 [43,44]. The reason mentioned above illustrates one way in which some sydnone analogues of the present study selectively inhibit COX-2. For example, compound 63 was found to be the best selective COX-2 inhibitor in vitro. It exhibited many polar, positive or negative charged interactions with fewer hydrophobic interactions inside the entirely hydrophobic pocket of the COX-1 resulting in an Table 6 1 H NMR spectral data of compounds 61-63 * Supplementary S110. ** Supplementary S114. *** Supplementary S118 a [13,22,23,[25][26][27]  unstable complex with this enzyme (Fig. 7a). On the contrary, the sydnone ring had a positive charge interaction with Arg513 of COX-2 forming a tight enzyme-ligand complex ( Fig. 6a and supplementary data S121). The same discussion is likewise with compound 51, which formed a positive charge interaction and a hydrogen bond with Arg513 of COX-2 ( Fig. 6b and supplementary data S121), albeit it had a polar interaction with the hydrophobic histidine residue of COX-1 (Fig. 7b).
Cyclooxygenase enzymes were also known as membrane-bound proteins. They bind to the endoplasmic reticulum (ER) membrane through the interaction of the hydrophobic residues of MBD with one sheet of the lipid bilayer where they are processed and folded to become mature enzymes [45]. Apart from its role in protein-membrane binding, the MBD forms the route from which arachidonic acid (AA) enters the COX active site. The residues from 379 to 385 are the apex of the  active site and have been known to form a proper positioning of AA [46]. Furthermore, Tyr355 lies on the opposite side of the mouth of the active site and controls the stereoselectivity of the inhibitors towards COX-2 [45]. The preceding discussion interprets the rationale for the high COX-1 inhibitory activity of compounds 37, 39 and 41 which had hydrophobic interactions with Tyr355, Trp387 and Tyr385 blocking AA from producing cyclooxygenase product PGG2 (Fig. 10). In contrast, the sydnone and benzene rings in these three molecules were involved in hydrophobic interactions with residues at the top of the COX-2 active site (Pro512, Pro514, etc.) rather than the apex which justifies their moderate COX-2 inhibitory activity.

Conclusion
In the present study, we described the synthesis of twenty-five sydnone-containing compounds including 3arylsydnones (37-48), 4-acetyl-3-arylsydnones (49-58) and trans bis-sydnone styryl ketones (61-63). The classical Aldol condensation was used to prepare bis-sydnone styryl ketones with a yield of 18-25%. Compounds 61-63 containing two sydnone rings linked via a styryl ketone are three novel molecules reported herewith. The in vitro findings showed that neither 3-arylsydnone nor 4-acetyl-3-arylsydnone were a promising COX antagonist. Hybridization of sydnone with styryl ketone (chalcone) led to an improved COX inhibition. Compound 63 with a chlorine atom at the para position of the benzene ring inhibited more than 80% of COX-2 with a selectivity index above 2.42. The docking analysis showed that the binding hits of 63 (Tyr91, Tyr475, Pro512, Pro514) and 51 (Lys511, Arg513, Tyr91, Pro512, Pro514) are almost essential for COX-2 inhibitory activity. Of paramount, they form positively charged interaction and hydrogen bonding, respectively, with Arg513 of COX-2 active site. The general binding poses of 63 mimics that of celecoxib. Accordingly, bis-sydnone styryl ketones could be a potential lead for selective cyclooxygenase-2 inhibitors and worth further chemical and biological consideration and investigations.