Discovering COX-2 Inhibitors from Flavonoids and Diterpenoids

Article history: Received on: 21/11/2016 Accepted on: 06/01/2017 Available online: 30/07/2017 Cyclooxygenase is the enzyme that catalyzes the biosynthesis of prostaglandins from its substrate, arachidonic acid (AA). The reactions involve two steps which are (1) the oxidation of AA to the hydroperoxyendoperoxide PGG2, followed by (2) its subsequent reduction to the hydroxyl endoperoxide PGH2. Selective COX-2 inhibitors do not bind to Arg120, an amino acid residue used by AA and by the nonselective NSAIDs, all of which are carboxylic acids. In this work we studied the interaction of 54 compounds against COX enzymes for antiinflammatory discovery using molecular docking simulation. Docking simulation for each compound was repeated 100x using Linux script command for AutoDockVina embedded in MGLTools v.1.5.6. Discovery Studio v.2.5.5 was employed to predict the volume of both COX binding pockets.21 compounds were selected according to their best scoring values and were calculated their selectivity index (cSI). Selective COX-2 inhibitors, respectively, are (1) 3,19-O-diacetylandrographolide; (2) 2-((1R,4aS,5R,6R,8aS)-6-hydroxy-5(hydroxymethyl)-5,8a-dimethyl-2 methylene decahydronaphthalen-1-yl)-1-(2-oxo-2,5-dihydrofuran-3-yl) ethyl 4-methylbenzoate; and (3)12,13-dihydroandrographolide. Preferential COX-2 inhibitorsare (1) coronarin D; (2) 19-O-acetylhydroandrographolide; (3) p-methoxycinnamic acid; (4) kaempferide. The rest of the ligands are categorized as non-selective inhibitors.


INTRODUCTION
Cyclooxygenases are membrane -associated hemecontaining homodimers enzymes that catalyze the biosynthesis of prostaglandins from their substrate, arachidonic acid (AA).The reactions involve two steps which are (1) the oxidation of AA to the hydroperoxyendoperoxide PGG 2 , followed by (2) its subsequent reduction to the hydroxyl endoperoxide PGH 2 (Vane et al, 1998;Vecchio et al, 2010).Selective COX-2 inhibitors do not bind to Arg120, an amino acid residue used by AA and by the nonselective NSAIDs, all of which are carboxylic acids (Mancini et al, 1995).The mechanism of inhibition of COX activity by acetosal and NSAIDs was first described by Vane in 1971(Vane, 1971).Although AA is the preferred substrate, other fatty acids are oxygenated by these enzymes with varying efficiencies.The interactions identified between the enzyme and the fatty acids when bound to COX-1 are conserved in COX-2 structures, with the only difference is the lack of interaction of the carboxylate of AA and eicosapentaenoic acid (EPA) with Arg-120.Leu-531 exhibits a different side chain conformation when the nonproductive and productive binding modes of AA are compared.It was speculated that the mobility of the Leu-531 side chain increased the volume available at the opening of the cyclooxygenase channel and contributes to the observed ability of COX-2 to oxygenate a broad spectrum of fatty acid and fatty ester substrates (Vecchio et al, 2010).However, NSAIDs cause gastrointestinal adverse effects, mainly because of their inhibition of the constitutive isoform of COX.Since selective COX-2 inhibitors fail to inhibit constitutive COX-1 isoform, they have no gastrointestinal adverse effects (Dilber et al, 2010).
Selective COX-2 inhibitors, e.g.coxib drugs such as rofecoxib (Vioxx ® ) and valdecoxib (Bextra ® ),were withdrawn from the market in 2004 and 2005, respectively, because of their effects on increased risk of heart attacks and strokes in long term use (Mason et al, 2006).Bioactive compounds, such as flavonoids and diterpenoids, have been proven to show anti-inflammatory activity.While many studies on flavonoids have already been explored, in this work we studied the interaction of 54 diterpenoids against COX enzymes for anti-inflammatory discovery using molecular docking simulation.Plants selected were Andrographis paniculata, Kaempferia galanga L., Hibiscus sabdariffa L., Tripterygium wilfordii Hook, f. (TWHF), and Eunicellane-based diterpenoids (Levita et al., 2010;Lin et al., 2007;Gonzales et al, 2015).

Protein preparation
The X-ray crystallographic 3D structures of COX-1 (PDB code: 3N8X, resolution 2.75Åcomplexed with nimesulide, crystallized by Sidhu et al, 2010) and COX-2 (PDB code: 5IKR, 2.34 Å complexed with mefenamic acid, crystallized by Orlando and Malkowski, 2016) were downloaded from online Protein Data Bank (Fig. 1).Water molecules, ligands, and other hetero atoms were removed from the protein molecule along with the chain B, C and D, using Swiss-Pdb Viewer v4.0.4 (Swiss Institute of Bioinformatics, downloaded fromwww.expasy.org).Addition of hydrogen atoms and Kollman charges to the protein was performed using AutoDockVina (Molecular Graphics Laboratory-The Scripps Research Institute, downloaded from http://autodock.scripps.edu).The binding modes of COX-1 with nimesulide and COX-2 with mefenamic acid were studied using Ligand Explorer Viewer v.4.1.0(Research Col laboratory for Structural Bioinformatics, embedded onhttp://www.pdb.org/pdb/explore).The binding pockets were calculated their volumes by using bind module in PLANTS 1.2and were visualized using Discovery Studio v.2.5.5 (Fig. 2 and 3).

Ligand preparation
All ligands were generated by using ChemBioDraw Ultra14.0 free trial (downloaded from www.cambridgesoft.com),and were geometry optimized using HyperChem Professional 8.0 by employing MMFF94 forcefield.The ligands were calculated their log P and were saved in pdb.file format for further process.

Validation of the docking simulation
Validation was performed re-docking of the ligand into its origin location for 50x using Linux script command, continued by calculating the SD of the binding energy and Ki.The re-docked ligand was then superimposed with the co-crystallized ligand extracted from the enzyme.

Docking simulation
Docking simulation for all 54 ligands was repeated 100x using Linux script command for AutoDockVina embedded in MGL Tools v.1.5.6 at position x = 26.8092;y = 33.6399;z = 199.8130for COX-1 whereas for COX-2was at x = -16.0647;y = 41.6941;z = 25.6819(these coordinates are within 10 Å distance centered to the ligand position).These coordinates were calculated by employing bind module in PLANTS 1.2.The default parameters of the automatic settings were used to set the genetic algorithm parameters.The docked conformation which had the highest docking score was selected to analyze the binding mode (Mason et al, 2006).Nimesulide and celecoxib were used as standards.
Molecular docking simulations of selected compounds in COX-1which showed interaction with Arg120 and/or Tyr355and in COX-2 which showed interaction with Tyr385 and/or Ser530 were provided in Fig. 7 and Fig. 8, respectively.
We compared our work with that of D'Mello who studied the modeling of flavonoids for COX inhibitors.They mentioned that chrysin and apigenin's5-hydroxyl interacted with Tyr355 forming a H-bond, while 4'-OH on the B-ring of apigenin formed an additional H-bond with Tyr385.Morin and kaempferol indicated an interaction with Arg120 and Tyr355 (D'Mello et al, 2011).Previous work of Mancini et al. (1995) concluded that there was biochemical evidence of the importance of the Arg1 20 .residue in COX-1 for interaction with arachidonic acid and NSAIDs containing a free carboxylic acid moiety (Mancini et al, 1995).Furthermore, a comparison was done with the work of Dash et al (2015) who discovered that salviifosides A of Alangium salvifolium was found to having three hydrogen bondings with Tyr355, Gln192, and Val523 (Dash et al, 2015).
The Ki values of the ligands against COX-1 and COX-2 could be seen in Fig. 9.
According to Nunthanavanit and Samee (2011), the inhibition constant (Ki) could be used to estimate the calculated selectivity index (cSI, the ratio of COX-2 Ki to COX-1 Ki of each complex), which are classified as non-selective (cSI> 1), preferential COX-2 (0.1 <cSI< 1.0), and selective inhibitors (cSI<0.1),respectively (Nunthanavanit and Samee, 2011).The cSIof the ligands was further calculated.The result is showed in Table 1.Table 1 showed that flavonoids and diterpenoids might play important role in interacting with COX enzymes.The selected compounds (cinnamaldehyde, ethyl cinnamate, ethyl-p-methoxy cinnamate, quercetin, kaempferol, and p-methoxy cinnamic acid) show interaction with Arg120 and/or Tyr355 in COX-1, which is similar with that of nonselective NSAIDs and other flavonoids.In COX-2 the same compounds showed interaction with Tyr385 and/or Ser530.Based on the selectivity index calculation (cSI), only a few of the flavonoids and diterpenoids are categorized as selective COX-2 inhibitors.
According to Wong et al (1997), mutations of residues of COX-1: His513 to Arg513 and Ile523 to Val 523 in COX-2, could strongly increase sensitivity to selective COX-2 inhibition and restore time-dependent inhibition.They also suggested that the corresponding Arg499 and Val509 residues of COX-2 are essential determinants in differentiating between the interaction of nonselective NSAIDs and selective COX-2 inhibitors and their mechanism of action (Wong et al, 1997).
Rieke and colleagues (1995) stated that the positively charged guanido group of Arg-120 interfered with the binding COX-2 inhibitors including NS398, DuP-697, and SC58125.NS398 did not cause time-dependent inhibition of R120Q hCOX-2, whereas DuP-697 and SC58125 were time-dependent inhibitors.They concluded that Arg-120 was important for the timedependent inhibition of hCOX-2 by NS398 but not by DuP-697 or SC58125 (Rieke et al, 1995).

Fig. 4 :
Fig. 4:The distribution of energy values after geometry optimization (left) and log P (right) of the ligands (listed in Fig.3).
and Table 1.

Table 1 :
The cSI of the ligands calculated by using Ki values on COX-1 and COX-2.