Exploring the fatty acid amide hydrolase and cyclooxygenase inhibitory properties of novel amide derivatives of ibuprofen

Abstract Inhibition of fatty acid amide hydrolase (FAAH) reduces the gastrointestinal damage produced by non-steroidal anti-inflammatory agents such as sulindac and indomethacin in experimental animals, suggesting that a dual-action FAAH-cyclooxygenase (COX) inhibitor could have useful therapeutic properties. Here, we have investigated 12 novel amide analogues of ibuprofen as potential dual-action FAAH/COX inhibitors. N-(3-Bromopyridin-2-yl)−2-(4-isobutylphenyl)propanamide (Ibu-AM68) was found to inhibit the hydrolysis of [3H]anandamide by rat brain homogenates by a reversible, mixed-type mechanism of inhibition with a Ki value of 0.26 µM and an α value of 4.9. At a concentration of 10 µM, the compound did not inhibit the cyclooxygenation of arachidonic acid by either ovine COX-1 or human recombinant COX-2. However, this concentration of Ibu-AM68 greatly reduced the ability of the COX-2 to catalyse the cyclooxygenation of the endocannabinoid 2-arachidonoylglycerol. It is concluded that Ibu-AM68 is a dual-acting FAAH/substrate-selective COX inhibitor.


Introduction
The non-steroidal anti-inflammatory agents (NSAIDs) such as ibuprofen, naproxen and diclofenac have widespread usage around the world, but their use is hampered by the incidence of severe gastrointestinal side effects. The elderly have a high consumption of NSAIDs, and this has resulted in a high incidence of NSAIDrelated hospitalisations and deaths 1 . There is thus much to be gained by the discovery and development of compounds that are as efficacious as the NSAIDs, but which lack these deleterious gastrointestinal effects.
Taken together, the studies above suggest that a compound with dual-action effects towards both cyclooxygenase (COX, the primary target of NSAIDs) and FAAH (or monoacylglycerol lipase) may be a potentially useful anti-inflammatory agent lacking the problematic gastrointestinal unwanted effects associated with NSAIDs. In 2015, the Piomelli group reported the synthesis and pharmacological properties of ARN2508 ((±)À2-[3-fluoro-4-[3-(hexylcarbamoyloxy)phenyl]phenyl]propanoic acid), a compound combining the structural elements of URB597 and the NSAID flurbiprofen 7,8 . The compound inhibited FAAH, COX-1 and COX-2 with IC 50 values of 31, 12 and 420 nM, respectively, and produced anti-inflammatory effects in vivo without causing gastric irritation 7 . The carbamate group in the molecule was required for (presumably irreversible) FAAH inhibition, but not for COX-inhibition 5 . Similar to the profens 9 , the compound shows substrate-selective inhibition of COX-2, being more potent when 2-AG is used as substrate than when arachidonic acid (AA) is used 10 .
An alternative approach has been to design compounds based on ibuprofen, which has modest FAAH-inhibitory activity 11 , and to optimise the FAAH-inhibitory properties while retaining the COXinhibitory properties of the parent compound. The first such compound, a heterocyclic amide ibuprofen analoge, Ibu-AM5 (2-(4-isobutylphenyl)-N-(3-methylpyridin-2-yl)propenamide, Figure  1) had been shown previously by one of us in 2003 to have analgesic activity with respect to acetic acid-induced visceral nociception in the mouse, without appreciable ulcerogenic potency 12 , and successively further described in 2007 for its FAAH inhibitory activity 13 . Further studies by us have shown that the compound inhibits FAAH in a mixed-type manner in sub-micromolar concentrations (i.e. 2-3 orders of magnitude more potent than ibuprofen itself) while retaining the substrate-selective inhibition of COX-2 seen with ibuprofen 14,15 .
Chemistry NMR spectra were recorded on an Inova 500 spectrometer (Varian, Palo Alto, CA). The chemical shifts (d) are reported in part per million downfield from tetramethylsilane (TMS), which was used as internal standard, and the spectra were recorded in hexadeuteriodimethylsulphoxide (DMSO-d 6 ). Infra-red spectra were recorded on a Vector 22 spectrometer (Bruker, Bremen, Germany) in Nujol mulls. The main bands are given in cm À1 . Positive-ion electrospray ionisation (ESI) mass spectra were recorded on a double-focusing MAT 95 instrument (Finnigan, Waltham, MA) with BE geometry. Melting points (mp) were determined on a SMP1 Melting Point apparatus (Stuart Scientific, Stone, UK) and are uncorrected. All products reported showed 1 H NMR spectra in agreement with the assigned structures. The purity of the tested compounds was determined by combustion elemental analyses conducted by the Microanalytical Laboratory of the Department of Chemical and Pharmaceutical Sciences of the University of Ferrara with a MT-5 CHN recorder elemental analyser (Yanagimoto, Kyoto, Japan) and the values found were within 0.4% of theoretical values. Ibuprofen amides Ibu-AM38-73 were synthesised according to Schemes 1 and 2.

Methyl 2-(4-isobutylphenyl)acetate (2)
A solution of Ibufenac 1 (1.92 g, 10 mmol) in CH 3 OH (10 ml) was added at room temperature (r.t.) with 37% HCl (0.5 ml) and refluxed for 4 h. The solvent was removed under vacuum and the crude methyl ester was used without purification in the further step. Yield 85%. Oil. 1  General procedure for the synthesis of esters 3 and 4 Lithium bis-(trimethylsilyl)amide (4.00 g, 24 mmol) was added to a solution of ester 2 (2.00 g, 9.7 mmol) in dry THF (40 ml) under argon at À78 C, the mixture was stirred at this temperature for 45 min. Then methyl iodide (3.40 g, 24 mmol) or 1,2-dibromoethane (4.51 g, 12 mmol) was added dropwise to the stirred solution for an additional 1 h. The mixture was poured in water and the desired product was extracted with diethyl ether (2 Â 30 ml). The solvent was dried over Na 2 SO 4, then it was evaporated under reduced pressure. The crude residue was purified via silica gel (200-400 mesh silica gel Merk KGaA) chromatography using petroleum ether 40-60 C and AcOEt 20:1.
General procedure for the synthesis of acids 5 and 6 To a solution of ester 3 or 4 (1 mmol) in EtOH (10 ml) 5 N solution of NaOH (2 ml) and water (2 ml) were added. The resulting mixture was stirred at r.t. for 24 h. After removing EtOH under vacuum to the resulting solution ice was added and then acidified with aqueous 20% HCl solution until pH 3-4. The formed precipitate was filtrated, washed with water and re-crystallized from n-hexane.

FAAH assay
Frozen (À80 C) brains (minus cerebella) from adult Wistar or Sprague-Dawley rats were thawed and homogenised in 20 mM HEPES, 1 mM MgCl 2 , pH 7.0. The homogenates were then centrifuged at $35000 x g for 20 min at 4 C followed by washing (by recentrifugation and by resuspension in the buffer) twice before incubation at 37 C for 15 min in order to hydrolyse all endogenous FAAH substrates. After a further centrifugation, pellets were resuspended in 50 mM Tris-HCl buffer, pH 7.4, containing 1 mM EDTA and 3 mM MgCl 2 , and frozen at À80 C in aliquots until used for the assay. For the FAAH assay 20 , test compounds, homogenates and [ 3 H]AEA (diluted with non-radioactive AEA to give a substrate concentration of 0.5 lM) in 10 mM Tris-HCl, 1 mM EDTA, pH 7.4, containing 1% w/v fatty acid-free bovine serum albumin were incubated for 10 min at 37 C. Activated charcoal in 0.5 M HCl was added to adsorb the unmetabolized [ 3 H]AEA and the samples were mixed and left at r.t. for $30 min. Following centrifugation at 2000 g for 10 min, aliquots of the supernatants, containing the [ 3 H]ethanolamine produced by hydrolysis of [ 3 H]AEA, were analysed for tritium content by liquid scintillation spectroscopy with quench correction. Blank values were obtained by the use of buffer rather than homogenate.
Data were expressed as % of vehicle (ethanol) control and analysed using the algorithm log(inhibitor) vs. responsevariable slope (four parameters) built into the GraphPad Prism computer programme v8.3 for the Macintosh (GraphPad Software Inc., San Diego, CA). The programme reports 95% confidence limits (profile likelihood) for the IC 50 values and these presented in the results.

COX-1 and 2 assay
The assay was performed essentially according to the method of Meade et al 21

FAAH receptor and ligand preparation
The crystal structure of the rat fatty acid amide hydrolase (rFAAH) (PDB ID: 3QK5) was downloaded from the Protein Data Bank website. Both monomers A and B were treated with the Protein Preparation Wizard 22 tool implemented in Maestro ver. 11.12 23 , in order to add all the hydrogen atoms and assign the correct bond orders. Subsequently, both the co-crystallized ligands and water molecules were removed. Residue Lys142 was considered in its deprotonated form, according to the proposed catalytic mechanism of FAAH [24][25][26] . The 3 D structure of Ibu-AM68 was built using the Graphical User Graphical User Interface (GUI) of Maestro ver.
11.12 23 . The protonation state of Ibu-AM68 at pH 7.4 in water has been calculated using the Epik module 27 . Finally, Ibu-AM68 was then minimised using a protocol already adopted for Ibu-AM5: 17 OPLS 2005 force field using the Polak-Ribiere Conjugate Gradient (PRCG) 28 algorithm and 2500 iteration steps.

Docking of Ibu-AM68 in FAAH
The molecular docking of Ibu-AM68 was performed only on the monomer A of the rat FAAH (rFAAH) receptor. Docking procedure was carried out with the Glide software package 29 , using the Standard Precision (SP) algorithm of the GlideScore function 30,31 and the OPLS 2005 force field 32 . A grid box of 29 Â 29 Â 29 Å centred on the ligand binding cavity was created. A total amount of 200 poses was generated and the conformational sampling of the ligand was enhanced by two times, as reported by the default setting of Glide. Docking conformations of Ibu-AM68 were then clusterised based on their RMSD cut-off of 2 Å. Globally, ten clusters were obtained and, among them, only the conformation included in the most populated cluster owing both the Glide Emodel and GlideScore lowest-energy value was considered (Figure 4). Such conformation was, finally, submitted to a further minimisation protocol using the OPLS 2005 force field 32 , 20,000 minimisation steps and the Polak-Ribiere Conjugate Gradient (PRCG) algorithm 28 .

Results and discussion
The potency of Ibu-AM5 towards AEA hydrolysis has been measured by different groups with different assay methodologies, FAAH preparations (rat brain, mouse brain, recombinant human FAAH) and substrate concentrations (0.5-2 mM) [13][14][15]33,34 . The IC 50 value for Ibu-AM5 from different studies in our laboratory using the same assay as here (of importance given the mixed-type nature of its inhibition of FAAH) ranges from 0.52 À 1.2 mM 14,15 and we therefore have used the most potent value for comparative purposes, since the aim is to identify more potent compounds.

FAAH inhibition
Three series of Ibu-AM5 analogues were synthesised according to Schemes 1 and 2 and tested towards rat brain FAAH-catalysed hydrolysis of AEA. The first series of two compounds was motivated by the finding that for Ibu-AM5 removal of the methyl group at the C-2 carbon atom ("Ibufenac-AM1") reduced the potency roughly 60-fold 15 . In order to explore whether or not the methyl group at that position was optimal, two compounds were synthesised, with a dimethyl (Ibu-AM72) and cyclopropyl (Ibu-AM73) groups instead of the methyl group at the C-2 carbon atom. The amides were obtained starting from Ibufenac (1) that was converted into its methyl ester 2 and then alkylated at C-2 position to give intermediates 3 and 4 that were hydrolysed to the corresponding acid 5 and 6. These last were coupled with 2amino-3-methylpyridine by EDC method to afford the target amides Ibu-AM72 and Ibu-AM73. The compounds inhibited FAAH with IC 50 values of 1.0 and 4.1 mM for Ibu-AM72 and Ibu-AM73, respectively ( Figure 2A, Table 1). Although this is a very limited series, it would suggest that there is little to be gained by adjusting the methyl group at the C-2 carbon atom, and so we moved on to the amido moiety of Ibu-AM5 3-substituent on the pyridine ring of Ibu-AM5.
In our initial study 12 , we reported that the potency of the pyridinamides of ibuprofen towards FAAH is related to the presence of both methyl and pyridine nitrogen in ortho positions to the amide nitrogen as in Ibu-AM5, the methyl absence or its moving in a position different from ortho to the amide nitrogen results in activity decrease. On this basis to evaluate the influence on the activity of the pyridine nitrogen, we changed pyridine ring with a phenyl. With this purpose, we prepared the 2-methylphenyl analogue of Ibu-AM5. The replacement of pyridine ring with a phenyl as in Ibu-AM38 produced activity decrease (IC 50 2.0 lM). After further evaluation a group able to establish hydrogen bond with the enzyme was inserted on the amide phenyl ring. With this purpose, the 2-hydroxy (Ibu-AM59) and the 4-hydroxy (Ibu-AM65) derivatives were designed. As indicated by their IC 50 values, the presence of the hydroxy group in 4-position caused an increase in activity (3.8 mM for Ibu-AM65 vs 8.5 mM for Ibu-AM59), although these compounds were among the least potent in the series. Next step was the integration of this activity enhancement with the presence of a 2-methyl group, with this purpose the compound Ibu-AM66 was prepared by condensation of ibuprofen with 2methyl-4-hydroxyaniline. Ibu-AM66 showed very good activity with an IC 50 value of 0.35 lM. With the aim to study, the influence on the inhibitory activity of hydrogen bond donor or acceptor group Ibu-AM66 was modified by replacement of the 4-hydroxy by a methoxy group to afford Ibu-AM67 (IC 50 4.6 lM). The result was decrease in activity, indicating that a hydrogen bond donor is better than an acceptor. Thereafter, we focussed on alternative substituents to the 3-methyl group such us halogens, trifluoromethyl and hydroxy groups. The data for these amides are shown in Figure 2(B) and summarised in Table 1. The observed potencies of the substituents were -I (Ibu-AM69) and -Br (Ibu-AM68) > -CF 3 (Ibu-AM60) > -Cl (Ibu-AM58) > -OH (Ibu-AM70). The 95% confidence intervals for the mean IC 50 values for Ibu-AM69 (0.078-0.19 mM) and Ibu-AM68 (0.038-0.15) overlap, so we regard the two compounds as equipotent but more potent than Ibu-AM5.
The inhibition of FAAH by Ibu-AM68 was investigated in more detail. Preincubation of Ibu-AM68 with the homogenates for up to 60 min prior to addition of substrate did not increase the observed inhibition, indicating that there is no time-dependence of the inhibition (Figure 2(D)). For a fully reversible inhibitor, preincubation for 60 min with a concentration "x" of compound followed by a 20-fold dilution prior to addition of substrate should  produce the same observed inhibition as seen with a concentration of "x/20" of the compound added together with the substrate, and this was found to be the case for Ibu-AM68 ( Figure  2(E)). Finally, kinetic experiments indicated a simple mixed-model inhibition of FAAH with a K i value of 0.26 mM and an a value (the ratio of the K i intercept : K i slope values; for pure competitive inhibition, a! 1) of 4.9 (Figure 2(F)). Thus, Ibu-AM68 is a reversible mixed inhibitor of rat brain FAAH with a greater potency than Ibu-AM5.

Inhibition of COX isoenzymes by Ibu-AM68
The ability of Ibu-AM68 to inhibit the cyclooxygenation of AA and 2-AG by COX-1 and COX-2 was investigated ( Figure 3). In our hands under the assay conditions used, 30 mM ibuprofen itself produces approximately 50% inhibition of the cyclooxygenation of AA by COX-1 with at best minor inhibition of COX-2 at this concentration. However, !10 and 30 mM ibuprofen produce a marked inhibition of 2-AG and AEA cyclooxygenation by COX-2 (neither endocannabinoid is a substrate for COX-1) 14,35 . Ibu-AM5 also shows substrate selective inhibition, reducing the rate of cyclooxygenation to about half at concentrations of 50 mM (COX-1, AA as substrate) and 3 mM (COX-2, AEA as substrate) whilst 100 mM Ibu-AM5 is without effect upon COX-2 with AA as substrate. 14 At a concentration of 10 mM, a modest inhibition of AA cyclooxygenation by Ibu-AM68 was seen with COX-1 whereas the cyclooxygenation of 2-AG by COX-2 was almost completely inhibited. Higher concentrations of Ibu-AM68 (50 and 100 mM) produced a complete inhibition of COX-1 but did not inhibit AA cyclooxygenation by COX-2. This suggests that the substrate-selective inhibition of COX-2 reported first for the R-profens by Marnett and colleagues 35 is also seen with Ibu-AM68. The mechanism of this inhibition has not been investigated, but Marnett et al. 9,36 have suggested that it may be related to COX-2 (which has a homodimeric structure) acting as functional heterodimers, whereby the binding of the Rprofen to one site acts allosterically to block 2-AG but not AA cyclooxygenation. It is possible that such a mechanism can explain the actions of Ibu-AM5 and Ibu-AM68.

FAAH docking on Ibu-AM68
Molecular docking calculations on (S)-Ibu-AM68 were performed with the Glide software 29-31 in the crystal structure of rat FAAH (PDB ID: 3QK5) 37 . The software Glide was chosen since it showed to be able to well reproduce the binding poses of (R)-and (S)-Ibu-AM5 resulting by molecular dynamics and free energy calculations 0.17 The results were clustered and successively ranked according to the Glide Emodel and the Glide Score. The best pose showed the isobutyl moiety pointing to the catalytic triad and the pyridine moiety entering the membrane access channel (MAC) channel ( Figure 4). In particular, the substituted pyridine ring established hydrophobic contacts with Leu404, Ile407, a T-shaped p-p interaction with Trp531 and a H-bond interaction with the hydroxyl group of Thr488. Moreover, polar contacts between the bromine atom and the backbone hydrogens of residues Asp403 and Leu404 were observed. An additional H-bond was established between the NH group of the ligand with the carbonyl of the Gly485. The Ibu-AM68 hydrophobic isobutyl-phenyl moiety resulted embedded in a hydrophobic region in the Acyl Chain Binding channel (ACB), and established hydrophobic contacts with residues Leu192 Phe244, Leu380, Thr488 and Ile491. The comparison with the binding mode of Ibu-AM5 showed high similarity in the isobutyl-phenyl moiety, but a different conformation of the pyridine ring with respect to the amide moiety. This different conformational behaviour maybe be likely due to different dipole alignment, being the slightly negative bromine atom better aligned with the NH group of the amide bond, while the methyl substituent preferred an orientation orthogonal to the carbonyl group.
In conclusion, the present study has characterised in vitro an Ibu-AM5 analogue that is slightly more potent than Ibu-AM5 itself as FAAH inhibitor and which retains its COX-2 substrateselectivity. Further studies are necessary to determine whether this compound behaves like the dual action FAAH-COX inhibitor ARN2508 in producing potentially beneficial effects in models of inflammatory pain without the ulcerogenic effects that are an issue with current NSAIDs.