Synthesis, antitumour activities and molecular docking of thiocarboxylic acid ester-based NSAID scaffolds: COX-2 inhibition and mechanistic studies

Abstract A new series of NSAID thioesters were synthesized and evaluated for their in vitro antitumor effects against a panel of four human tumor cell lines, namely: HepG2, MCF-7, HCT-116 and Caco-2, using the MTT assay. Compared to the reference drugs 5-FU, afatinib and celecoxib, compounds 2b, 3b, 6a, 7a, 7b and 8a showed potent broad-spectrum antitumor activity against the selected tumour cell lines. Accordingly, these compounds were selected for mechanistic studies about COX inhibition and kinase assays. In vitro COX-1/COX-2 enzyme inhibition assay results indicated that compounds 2b, 3b, 6a, 7a, 7b, 8a and 8 b selectively inhibited the COX-2 enzyme (IC50 = ∼0.20–0.69 μM), with SI values of (>72.5–250) compared with celecoxib (IC50 = 0.16 μM, COX-2 SI: > 312.5); however, all the tested compounds did not inhibit the COX-1 enzyme (IC50 > 50 μM). On the other hand, EGFR, HER2, HER4 and cSrc kinase inhibition assays were evaluated at a 10 μM concentration. The selected candidates displayed limited activities against the various tested kinases; the compounds 2a, 3b, 6a, 7a, 7b and 8a showed no activity to weak activity (% inhibition = ∼0–10%). The molecular docking study revealed the importance of the thioester moiety for the interaction of the drugs with the amino acids in the active sites of COX-2. The aforementioned results indicated that thioester based on NSAID scaffolds derivatives may serve as new antitumor compounds.


Introduction
Malignancy is global health problem and is the leading cause of death in children until fifteen years of age 1 . Non-steroidal antiinflammatory drugs (NSAIDs) such as sulindac, indomethacin and celecoxib are commonly used for treating arthritis via inhibition of the cyclooxygenase enzyme (COX) 2,3 . COX-2 levels are overexpressed in human tumours, unlike in normal cells and could develop a tumorigenic potential 4 . Selective enzyme inhibition and restoration of normal apoptotic responses is known as COX-2dependent anticancer mechanism 4-6 . On the other hand, COX-2independent mechanisms function via apoptosis stimulation, angiogenesis arrest, or cancer cell growth inhibition by blocking signal transduction pathways for cell proliferation [7][8][9][10] .
Drug repositioning development is a more important process for saving money and time than the production of a new drug 11 . NSAIDs and coxib such as naproxen, ibuprofen, indomethacin, sulindac, celecoxib and their analogues ( Figure 1) have diverse scaffolds; modifying their basic structures is relatively safe, applicable for oral use, associated with multiple therapeutic features, such as analgesic, antipyretic, anti-inflammatory and anticancer activities [12][13][14][15][16][17][18] . For example, sulindac amides ( Figure 1) showed a good activity against a panel of lymphoblastic leukemia cell lines in nanomolar concentrations 18 . Additionally, celecoxib reduced the number and size of colorectal polyps in adenomatous polyposis ( Figure 1) [19][20][21] . Antiproliferative and apoptosis effects of celecoxib in colon, stomach, lung, prostate and pancreatic cancer cells have been observed by selective COX-2 inhibition [22][23][24][25] . On the other hand, a combination of drugs (NSAIDs) such as indomethacin, sulindac, tolmetin, acemetacin, zomepirac and mefenamic acid at non-toxic levels, and different chemotherapeutic drugs such as anthracyclines (doxorubicin, daunorubicin and epirubicin), in addition to VP-16, vincristine and teniposide, led to a significantly synergistic cytotoxicity of these chemotherapeutic drugs in the human COR L23R, DLKP, A549 and COR L23P lung cancer cell lines, and the human HL60/ADR leukaemia cell line 3 .
A molecular docking technique was used in order to predict the binding geometry requirements of the target molecules, which is important for the antitumour activity.

Experimental
Melting points were recorded on a Barnstead 9100 Electrothermal melting apparatus. IR spectra (KBr) were recorded on an FT-IR Perkin-Elmer spectrometer (m cm À1 ). 1 H and 13 C NMR spectra were recorded on Bruker 500 or 700 MHz spectrometers using DMSO-d 6 or CDCl 3 as the solvent. Microanalytical data (C, H and N) were obtained using a Perkin-Elmer 240 analyser and the proposed structures were within ±0.4% of the theoretical values. Mass spectra were recorded on a Varian TQ 320 GC/MS/MS mass spectrometer. NSAIDs thioester was obtained according to reported method 43 .

General method for the preparation of NSAIDs thioester
Trifluoroacetic acid (0.5 mmol) was added dropwise to a mixture of NSAIDs (0.1 mmol) and thiol (0.5 mmol) in dry acetonitrile that was heated for 10-12 h at 60 C. The reaction mixture was cooled, quenched using ammonium chloride solution, extracted with ethylacetate, washed with brine and dried over anhydrous sodium sulphate; the solvent was then evaporated, and the product obtained was chromatographed with hexane and CHCl 3 .

In vitro cyclooxygenase (COX) inhibition assay
The colorimetric COX (ovine) inhibitor screening assay kit (kit catalogue number 560101, Cayman Chemical, Ann Arbor, MI) was utilized according to the manufacturer's instructions to examine the ability of the test compounds and the reference drugs to inhibit the COX-1/COX-2 isozymes 51,52 .

Kinase inhibition assay
The assay for Kinases was performed at BPS Bioscience Inc. 6044 Cornerstone Court West, Ste. E, San Diego, CA 92121, USA using Kinase-Glo Plus luminescence kinase assay kit (Promega).
Luminescence signal was measured using a BioTek Synergy 2 microplate reader 53 .

Docking methodology
All modelling experiments were conducted with MOE programs running on PC computer [MOE 2008.10 of Chemical Computing Group. Inc] 54 . The docking protocol is summarized in supporting information 51,52,[55][56][57] .

Chemistry
The new thioesters were synthesized by the reaction of the carboxylic acid group of NSAIDs with thiophenol and cycloxanethiol in the presence of trifluoroacetic acid (TFA) 43 . The newly synthesized thioesters (Scheme 1) were confirmed by the presence of the carbonyl group (C¼O) at 1700-1669 cm À1 and stretching of the (C-S) group at 859-683 cm À1 in the IR spectra. Additionally, the newly synthesised thioesters were confirmed by a characteristic peak at 201. 21-195.15 ppm attributable to the (S-C¼O) group in addition to the characteristic peaks of the cyclohexane moiety at 25.40-42.42 ppm or aromatic peaks of the thiophenol moiety in the aromatic region of the 13 C NMR spectra. The 1 H NMR spectra of the new thioesters showed a singlet peak because of the S-CH moiety of S-cyclohexane at 3.66-3.37 ppm, as well as the other 10 protons of the cyclohexane moiety in the aliphatic region or the aromatic peaks of the thiophenol moiety in the aromatic region.
In vitro kinase assay Accordance to the cytotoxicity activity of the newly synthesized compounds (Table 1), six compounds were selected for further mechanistic investigations about the kinases, EGFR, HER2, HER4 and cSrc. The results of kinase inhibition assays indicated that compounds 2a, 3b, 6a, 7a, 7b and 8a showed limited activities against the kinase enzymes. As shown in Table 3, all the compounds showed no or weak activities against HER2, HER4 and cSrc, as indicated by their % inhibition when used at a concentration of 10 lM (% inhibition ¼ $0-10%), comparable to the 81-100% inhibition of the reference drug staurosporine, used at a concentration of 1 lM (Table 3).

Docking studies
To highlight the inhibition selectivity of different core analogues towards the COX-2 enzyme, automated docking studies were carried out using the MOE 2008.10 program 54 . The scoring functions, hydrogen bonds and hydrophobic interactions formed with the surrounding amino acids are used to predict the binding modes, the energy of interaction and orientation of the docked compounds at the active sites of the COX-2 enzyme (Figures 2-3). The protein-ligand complex was constructed based on the X-ray structure of COX-2, with its bound inhibitor SC-558, which was available through the RCSB Protein Data Bank (PDB entry 1CX-2) 60 . The active site of the enzyme was defined to include residues within a 10.0-Å radius around any of the inhibitor atoms. This active pocket consisted of amino acid residues such as arginine (Arg 510 ), histidine (His 90 ), glutamine (Gln 192 ) or tyrosine (Tyr 355 ), arginine (Arg 120 ), valine (Val 523 ) and methionine (Met 535 ), which play fundamental roles by forming H-bonds and hydrophobic interactions (Figures 2-3). In order to verify the reproducibility of the docking calculations, the cocrystallised ligand SC-558 was extracted from the complex and submitted for one-ligand run calculation. This reproduced 20 top scoring conformations falling within a root-mean-square deviation (rmsd) value between 0.4 Å and 2.0 Å, from the bound X-ray conformation for the COX-2 enzyme, suggesting that this method is valid enough to be used for docking studies of other compounds.
The present work is based on a comparative study to define the selectivity of most active COX-2 inhibitors, such as the thioester derivatives 6a, 7a, 7b, 8a and 8b of well-known and wellestablished NSAIDs, namely diclofenac, indomethacin and sulindac, by exploring their docking and complementarity to the COX-2 binding site. Generally, the results of the docking study indicated that the thioesters based on indomethacin, sulindac and diclofenac scaffolds matched perfectly with the configuration of the T-shaped merged COX-2 binding site, which easily accommodated the wide bulk SC-558 inhibitor.
Compounds 8a (IC 50 ¼ 0.20 mM), 7a (IC 50 ¼ 0.22 mM) and 6a (IC 50 ¼ 0.25 mM) were the most active analogues; they showed the highest recognition at the COX-2 binding site, which is consistent with the experimental results of the selectivity index obtained from the COX-2 assay (Table 2). Compound 8a was shown to have a unique binding configuration (Figure 2, lower panel). The phenyl thioester of sulindac showed promising binding affinity and proper complementarity, because the E-conformer allows the crest-configuration to embed properly within the merged active site of the 5-flouroindenyl group, via proper hydrophobic interactions with the amino acids of the merged cleft active site. The thiocarbonyl function was impressively recognized with the polar amino acid Glu 524 . The two terminal phenyl groups are surrounded by hydrophobic amino acid analogs, where Leu 534 showed an aromatic-aromatic interaction with the 4-methylsulfinylphenyl group. The terminal 4-methylsulfinyl fragment enhanced the strategic function that showed proper complementarity with the groove wall residues via both Val 344 and Val 349 . According to the selectivity index and computational binding, the hydrogenbonded compound 8a was considered the most promising selective lead.
Moreover, compound 7a was held by one hydrogen bond with Tyr 355 , via its carbonyl thioester, apart from the electrostatic interaction between the chloro-function and the mercapto moiety of the corresponding Met 535 (Figure 2, middle panel). Aromatic recognition also was observed between the aromatic phenyl thioester and the side chain of Arg 513 .
Additionally, the docking studies of compound 6a revealed outstanding interactions with one of the essential active-site Arg 120 residues formed via proper hydrogen bonding (Figure 2, upper panel). The two aminophenyl and dichlorophenyl groups augment the aromatic-aromatic interaction with a series of seven hydrophobic amino acids, Leu 531 , Met 113 , Val 116 , Leu 352 , Val 349 and Ile 345 , arranged in a continued chain, lining the wall of the cleft. However, because of the NH-amino group being embedded inwards and away from the surrounding residues, it does not interact with the active-site amino acids, owing to the bulkiness of the two phenyl substituents. The thiophenyl function protruded towards Val 523 , showing notable improvement in the net lipophilic stabilisation (Figure 2, upper panel).
In comparison to the aforementioned derivatives, compounds 7b and 8b showed moderately selective inhibition towards COX-2. Compound 7b revealed distinct binding wherein the cyclohexyl group was merged with the side-pocket, and the thioester function was exposed to the surrounding binding residues for interaction with the conserved amino acids Val 523 via proper hydrogen bonding. Additionally, methoxy oxygen was recognized by a single conventional hydrogen bond with the conserved Arg 120 (Figure 3,  upper panel).
Similarly, the cyclohexyl group of compound 8b was merged with the side-pocket, and the ester function was exposed to the surrounding binding residues, to be oriented ahead of the polar amino acids, Ser 353 and Arg 120 (Figure 3, lower panel). Along the lining wall of the pocket, all the hydrophobic amino acids are oriented complementarily with the hydrophobic-facing groups indene, methylene, and the terminal phenyl. From another site, the following hydrophobic amino acids are stuffed properly and sandwiched between the cyclohexyl ring and the terminal phenyl group.