(S)-1-(Pent-4′-enoyl)-4-(hydroxymethyl)-azetidin-2-one derivatives as inhibitors of human fatty acid amide hydrolase (hFAAH): synthesis, biological evaluation and molecular modelling

Abstract

A series of lipophilic ester derivatives (2a–g) of (S)-1-(pent-4′-enoyl)-4-(hydroxymethyl)-azetidin-2-one has been synthesised in three steps from (S)-4-(benzyloxycarbonyl)-azetidin-2-one and evaluated as novel, reversible, β-lactamic inhibitors of endocannabinoid-degrading enzymes (human fatty acid amide hydrolase (hFAAH) and monoacylglycerol lipase (hMAGL)). The compounds showed IC₅₀ values in the micromolar range and selectivity for hFAAH versus hMAGL. The unexpected 1000-fold decrease in activity of 2a comparatively to the known regioisomeric structure 1a (i.e. lipophilic chains placed on N₁ and C₃ positions of the β-lactam core) could be explained on the basis of docking studies into a revisited model of hFAAH active site, considering one or two water molecules in interaction with the catalytic triad.

• β-lactam
• docking
• fatty acid amide hydrolase
• rapid dilution
• reversible inhibition

Introduction

Nowadays, the endocannabinoid (EC) system is considered as one of the promising fields in pain and central nervous system disorder treatments1,2. The EC system is formed by the cannabinoid receptors (CB1, CB2), their endogenous ligands (anandamide (AEA), 2-arachidonoylglycerol (2-AG)) and enzymes/proteins involved in the EC biosynthesis, transport and regulation3,4. Amongst these enzymes, the EC-degrading enzymes, namely fatty acid amide hydrolase (FAAH) and monoacylglycerol lipase (MAGL), are of particular interest. Indeed, FAAH and MAGL hydrolyse, respectively, AEA and 2-AG, with the consequence of the termination of their action on CB1/CB2. Because the activation of the cannabinoid receptors results in multiple beneficial effects, a prolonged action of the cannabinoids is often desirable. The best way to achieve this goal consists in the inhibition of the EC-degrading enzymes5–7. The search of selective inhibitors of FAAH and MAGL has already produced very active compounds featuring various chemical structures, and is still a growing research domain in medicinal chemistry8–10.

FAAH is a serine amidase which possesses the non-traditional Ser-Ser-Lys catalytic triad. The processing of natural substrates involves the nucleophilic attack of Ser241 residue on the sessile amide carbonyl, leading to an unstable acyl-enzyme covalent intermediate which is rapidly hydrolysed for regenerating the active enzyme. The 3D-structure of the human FAAH (hFAAH) is not known. However, the rat isoforme (rFAAH) has been crystallised in the presence of MAFP (methylarachidonoyl fluorophosphonate) and the structure of the resulting enzyme-inhibitor complex has been resolved11–13. More recently, an engineered humanised rat FAAH (h/rFAAH) has been produced and successfully crystallised in the presence of three inhibitors, representative of the major classes of FAAH covalent inhibitors14–17. MAGL is also a serine hydrolase characterised by the classical Ser-His-Asp catalytic triad and two non-catalytic Cys residues located in the vicinity of the active site. The 3D-structure of the human MAGL (hMAGL) has been disclosed in 2010, independently by two research groups7,18.

The mechanism-based design of hFAAH inhibitors focuses on electrophilic entities susceptible to form stable tetrahedal- or acyl-enzyme covalent intermediates, as usually considered for the long-term (or irreversible) inhibition of serine enzymes. Accordingly, the main classes of hFAAH inhibitors are as follows: α-keto-heterocycles (OL-135), carbamates (URB-597), ureas (PF-622) and boronic acids (WO 2009 126691)8. Studies on the inhibition of hMAGL reveal the same trends with urea-(SAR-629) and carbamate-based compounds (JZL-184) (Figure 1)8–10. Surprisingly, the β-lactone and β-lactam templates are scarcely described in the field of EC system control19–21, although oxetan-2-ones22 and mainly azetidin-2-ones23–27 are well-known inhibitors of D,D-peptidases, β-lactamases, leukocyte elastase, thrombin, prostate specific antigen and human cytomegalovirus protease and among others.

Figure 1. Representative inhibitors of hFAAH and hMAGL.

A β-lactone (OMDM169)19) inspired of a serine hydrolase inhibitor, tetrahydrolipstatin, has been disclosed by Di Marzo’s group in 2008 for the inhibition of hMAGL (Figure 2). Since 2008, our group is interested in the development of β-lactams as selective inhibitors of hFAAH28. The lead compound 1a (IC₅₀ = 5.3 nM) displays two lipophilic chains anchored on the N₁ and C₃ positions of the azetidin-2-one ring (Figure 2). The origin of the β-lactam template is the chiral precursor of carbapenem antibiotics which is commercially available, i.e. (3 R,4 R)-4-acetoxy-3-((1′R)-(t-butyldimethylsilyloxy)-ethyl)-azetidin-2-one29, named here chiron A (Figure 3).

Figure 2. β-Lactone and β-lactam inhibitors of hMAGL and hFAAH, respectively.

Figure 3. General structures of studied hFAAH inhibitors and their respective precursors.

A random screening of our in-house library of β-lactams30, made from the same precursor A, showed that trisubstituted derivatives (at positions N₁, C₃ and C₄) exhibit only moderate activity against hFAAH (see Supplementary Material). Therefore, we envisaged the synthesis of a novel family of disubstituted β-lactams, featuring the lipophilic chains on the positions N₁ and C₄ (Figure 3, series of compounds 2). The (S)-configuration of the C₄ carbon has been selected by comparison of molecules 1a and 2a (n = 3, R = Ph; Figure 3) positioned into the homology model of hFAAH active site previously published28. The starting material of the series 2 of potential inhibitors is a known chiral β-lactam (named chiron B), used for the total synthesis of isocephem antibiotics, i.e. (S)-4-(benzyloxycarbonyl)-azetidin-2-one; B is readily accessible in large scale from (S)-aspartic acid31,32.

In this article, we describe the synthesis of representative members of the β-lactam family 2 and their evaluation for the inhibition of hFAAH and hMAGL. A comparison of inhibitors 1a and 2a has been performed by molecular modelling studies, guided by the recent crystallographic structures of h/rFAAH. Our previous model of active site has been completed, considering water molecules in interaction with the inhibitors and amino acid residues.

Experimental

Synthetic chemistry

Generalities

The solvents (technical and anhydrous) and reagents were purchased from Acros Organics (Thurnout, Belgium), Alfa Aesar (Bischeim, France), Cayman Chemical (Monluçon, France), Sigma-Aldrich-Fluka (Diegem, Belgium) or VWR (Leuven, Belgium), and used as received. The reactions in dry conditions were performed under argon atmosphere, in flame-dried vessel. Thin layer chromatography was realised on silica gel plate Merck 60F254, and revealed under UV light. Preparative chromatography was made with flash silica gel from Rocc.

The mass spectra (MS) were acquired on Thermo Finnigan LQC spectrometer, in electrospray ionisation (ESI), negative mode. ¹H and ¹³C nuclear magnetic resonance (NMR) spectra were recorded on Bruker Avance 300 (Eke, Belgium) (300 MHz, for proton and 75 MHz for carbon) and Bruker Avance 500 equipments (500 MHz for proton and 125 MHz for carbon), using CDCl₃ or CD₃OD as solvent and tetramethylsilane as internal standard. Chemical shifts are reported in ppm (residual non-deuterated solvent: CDCl₃ at 7.26 (H) and 77.16 (C) ppm; CD₃OD at 3.31 (H) and 40.00 (C) ppm). Coupling constants (J) are given in Hertz (Hz). The infra-red spectra were acquired on Shimadzu FTIR-8400S apparatus (Deurne Antwerpen, Belgium). Samples were deposited on the Se–Zn crystal by evaporation from dichloromethane (DCM) solution to give a film. High resolution MS (HRMS) were measured with a QExactive spectrometer (Thermo Scientific, Bruxelles, Belgium). Rotations were obtained with a polarimeter Atago Ap-100 at 25 °C. Samples were dissolved in CHCl₃ (concentrations given in percentages, i.e. g 100/mL). The purity of the tested compounds (≥95%) was controlled by HPLC.

Procedure for the reduction of (S)-4-(benzyloxycarbonyl)-2-azetidinone (B)

(S)-4-(Hydroxymethyl)-2-azetidinone (3)

To a solution of (S)-4-(benzyloxycarbonyl)-2-azetidinone (1.00 g, 4.87 mmol, 1 equiv.) in methanol (25 mL) was added NaBH₄ (556 mg, 14.6 mmol, 3 equiv.), in small portions, at 0 °C and under argon atmosphere. The mixture was stirred for 15 min at 0 °C and then for 2 h at 20 °C. After neutralisation by the addition of acetic acid (10% aqueous solution, 15 mL), and concentration under vacuum, the crude alcohol was purified by filtration over a silica gel pad with DCM/methanol (9:1, v/v) as the eluent. 3 was recovered as a colourless oil (495 mg, 100% yield). Rf = 0.244 (DCM/MeOH, 9:1, v/v). δ_H (300 MHz; CD₃OD) 2.67 (1 H, dd, J = 2.0, 14.9), 2.90 (1 H, dd, J = 4.6, 14.9), 3.54–3.61 (1 H, m) and 3.67–3.72 (2 H, m) ppm. δ_C (75 MHz; CD₃OD) 40.0, 49.7, 64.7 and 171.1 ppm.

General procedure of esterification according to Method (ii)

To a solution of 3 (1 equiv.), dicyclohexylcarbodiimide (DCC, 1.1 equiv.) and 4-dimethylaminopyridine (DMAP, 0.05 equiv.) in dry DCM (10 mL/mmol), was added the appropriate carboxylic acid (1.1 equiv.) in DCM (5 mL/mmol), drop-wise under argon atmosphere. The mixture was stirred overnight at room temperature, then cooled at 0 °C (for precipitation of dicyclohexylurea) and filtered. The filtrate was concentrated under vacuum and the residue was purified by flash chromatography (elution with DCM/MeOH, 9:1, v/v) to furnish the ester 4 (for 4d, 4e and 4g, see Supplementary Material).

(S)-(4-Oxoazetidin-2-yl)methyl 2-(biphenyl-4-yl)acetate (4c)

The title compound (234 mg, 87% yield) was obtained as a white gummy solid from 3 (0.92 mmol). Rf = 0.35 (DCM/EtOAc, 1:1, v/v). ν_max/cm 1732, 1714, 1487, 1413, 1242, 1222, 1149 and 1008. δ_H (500 MHz; CDCl₃) 2.73 (1 H, ddd, J = 1.5, 2.4, 14.9), 3.05 (1 H, ddd, J = 2.0, 5.2, 14.9), 3.70 (2 H, s), 3.84–3.88 (1 H, m), 4.07 (1 H, dd, J = 6.4, 11.7), 4.43 (1 H, dd, J = 3.7, 11.7), 6.05 (1 H, br s), 7.32–7.38 (3 H, m), 7.41–7.46 (2 H, m) and 7.55–7.60 (4 H, m). δ_C (125 MHz; CDCl₃) 40.85, 40.87, 46.1, 66.0, 127.2, 127.55, 127.49, 128.9, 129.7, 132.6, 140.4, 140.7, 167.0 and 171.5. m/z (ESI) 296.15 (M + H)⁺, 318.11(M + Na)⁺. HRMS m/z found 295.12041, calc. for C₁₈H₁₇NO₃ 295.12029.

(S)-(4-Oxoazetidin-2-yl)methyl 3-(1H-indol-3-yl)propanoate (4f)

The title compound (249 mg, 92% yield) was obtained as a yellow oil from 3 (0.99 mmol). Rf = 0.45 (DCM/MeOH, 9:1, v/v). [α]_D +5.97 (c 1.8 in CHCl₃). ν_max/cm 2922, 1730 (br), 1456, 1340 and 1155. δ_H (300 MHz; CDCl₃) 2.58 (1 H, ddd, J = 1.9, 5.2, 15), 2.77 (2 H, t, J = 7.3), 2.92 (1 H, ddd, J = 1.2, 2.1, 15), 3.12 (2 H, t, J = 7.3), 3.65 (1 H, m), 3.94 (1 H, dd, J = 6.8, 11.7), 4.28 (1 H, dd, J = 3.7, 11.7), 5.62 (1 H, br s), 6.96 (1 H, d, J = 2.1), 7.12 (1 H, td, J = 6.9, 0.9), 7.20 (1 H, td, J = 6.9, 1.1), 7.36 (1 H, d, J = 7.9), 7.59 (1 H, d, J = 7.8) and 8.06 (1 H, br s) ppm. δ_C (75 MHz; CDCl₃) 20.9, 34.9, 40.7, 46.0, 65.7, 111.4, 114.6, 118.8, 119.5, 121.7, 122.3, 127.2, 136.3, 167.0 and 173.3 ppm. m/z (ESI) 273.12 (M + H)⁺, 295.10 (M + Na)⁺. HRMS m/z found 273.12338, calc. for C₁₅H₁₇N₂O₃ 273.12337.

General procedure of esterification according to Method (iii)

To a solution of 3 (1 equiv.) in dry DCM solution (15 mL/mmol) at 20 °C, were added successively pyridine (1 equiv.) and the appropriate acid chloride (1 equiv.) under argon atmosphere. The mixture was stirred overnight, then diluted with DCM (15 mL/mmol) and washed with aqueous Na₂CO₃ (10%). The organic layer was washed with 1 N HCl and brine, dried over MgSO₄, filtered and concentrated under vacuum. The residue was purified by flash chromatography (elution with DCM/EtOAc, 7:3, v/v) to furnish the ester 4 (for 4h, see Supplementary Material).

(S)-(4-Oxoazetidin-2-yl)methyl 4-phenylbutanoate (4a)

The title compound (88 mg, 72% yield) was obtained as a yellow oil from 3 (0.49 mmol). Rf = 0.38 (DCM/MeOH, 95:5, v/v). [α]_D = +10.07 (c = 1.3). ν_max/cm 3329, 2943, 1730, 1688, 1616, 1496 and 1142. δ_H (300 MHz; CDCl₃) 1.96 (2 H, quint, J = 7.5), 2.36 (2 H, t, J = 7.5), 2.65 (2 H, t, J = 7.5), 2.75 (1 H, ddd, J = 1.4, 2.4, 14.9), 3.09 (1 H, ddd, J = 2.0, 5.2, 14.9), 3.85 (1 H, m), 4.01 (1 H, dd, J = 6.5, 11.7), 4.37 (1 H, dd, J = 3.8, 11.7), 6.06 (1 H, br s), and 7.13–7.33 (5 H, m) ppm. δ_C (75 MHz; CDCl₃) 26.4, 33.4, 35.1, 40.9, 46.2, 65.5, 126.2, 128.5, 128.6, 141.2, 167.0 and 173.3 ppm. m/z (ESI) 248.13 (M + H)⁺, 270.11 (M + Na)⁺. HRMS m/z found 248.12823, calc. for C₁₄H₁₈NO₃ 248.12812.

(S)-(4-Oxoazetidin-2-yl)methyl biphenyl-4-carboxylate (4b)

The title compound (122 mg, 78% yield) was obtained as a white amorphous powder from 3 (0.55 mmol). Rf = 0.27 (Cyclohexane/EtOAc, 2:8, v/v). ν_max/cm 2910, 2850, 1738, 1714, 1606, 1406, 1286, 1269 and 1123. δ_H (300 MHz; CDCl₃) 2.90 (1 H, ddd, J = 1.6, 2.2, 14.9), 3.17 (1 H, ddd, J = 1.9, 5.2, 14.9), 4.03 (1 H, m), 4.34 (1 H, dd, J = 6.3, 11.7), 4.64 (1 H, dd, J = 3.7, 11.7), 6.16 (1 H, br s), 7.36–7.53 (3 H, m), 7.57–7.74 (4 H, m) and 8.10 (2 H, d, J = 8.3) ppm. δ_C (75 MHz; CDCl₃) 41.0, 46.4, 66.1, 127.3 (2 C), 127.4, 128.4, 129.1, 130.3, 139.9, 146.3, 166.3 and 167.0 ppm. m/z (ESI) 304.09 (M + Na)⁺, 585.19 (2 M+ Na)⁺. HRMS m/z found 304.09475, calc. for C₁₇H₁₅NO₃Na 304.09441.

General procedure for N-acylation according to Method (iv)

To a solution of NH-azetidinone 4 (1 equiv.) in DCM (8 mL/mmol) were added successively pyridine (2 equiv.) and 4-pentenoyl chloride (2 equiv.) under argon atmosphere. The mixture was refluxed during 24 h, then diluted with DCM (20 mL/mmol). After washing with aqueous Na₂CO₃ (10%) (three times), the organic layer was washed with 3 N HCl and brine, dried over MgSO₄, filtered and concentrated under vacuum. The residue was purified by flash chromatography (elution with DCM/EtOAc, 98:2, v/v) to furnish the N-acyl azetidinone 2 (for 2d and 2e, see Supplementary Material).

(S)-(4-Oxo-1-pent-4-enoylazetidin-2-yl)methyl 4-phenylbutanoate (2a)

The title compound (26 mg, 43% yield) was obtained as a colourless oil from 4a (0.18 mmol). Rf = 0.5 (DCM/MeOH, 9:1, v/v). [α]_D = −10.63 (c = 3.1). ν_max/cm 2956, 2924, 1791, 1738, 1699, 1454, 1383, 1313 and 1144. δ_H (300 MHz; CDCl₃) 2.01 (2 H, quint, J = 7.4), 2.34–2.50 (4 H, m), 2.69 (2 H, t, J = 7.4), 2.84 (2 H, t, J = 7.4), 2.95 (1 H, dd, J = 3.4, 16.3), 3.17 (1 H, dd, J = 6.3, 16.3), 4.31 (1 H, m), 4.43 (1 H, dd, J = 3.1, 12.0), 4.57 (1 H, dd, J = 4.1, 12.0), 5.01–5.15 (2 H, m), 5.85 (1 H, m), 7.18–7.26 (3 H, m) and 7.29–7.36 (2 H, m) ppm. δ_C (75 MHz; CDCl₃) 26.4, 28.0, 33.4, 35.1, 35.8, 39.2, 48.5, 61.6, 115.9, 126.2, 128.5 (2 C), 141.2, 136.4, 164.2, 170.7 and 173.1 ppm. m/z (ESI) 330.17 (M + H)⁺, 352.15 (M + Na)⁺. HRMS m/z found 352.1519, calc. for C₁₉H₂₃NO₄Na 352.15193.

(S)-(4-Oxo-1-pent-4-enoylazetidin-2-yl)methyl biphenyl-4-carboxylate (2b)

The title compound (25 mg, 55% yield) was obtained as a colourless oil from 4b (0.12 mmol). Rf = 0.38 (DCM/EtOAc, 95:5, v/v). [α]_D = −27.31 (c = 0.8). ν_max/cm 2921, 1790, 1716, 1703, 1608, 1383, 1269 and 1113. δ_H (300 MHz; CDCl₃) 2.37–2.48 (2 H, m), 2.79–2.91 (2 H, m), 3.07 (1 H, dd, J = 3.4, 16.3), 3.23 (1 H, dd, J = 6.3, 16.3), 4.42 (1 H, m), 4.64 (1 H, dd, J = 3.1, 12.0), 4.82 (1 H, dd, J = 4.0, 12.0), 4.94–5.11 (2 H, m), 5.85 (1 H, m), 7.36–7.51 (3 H, m), 7.58–7.71 (4 H, m) and 8.06 (2 H, d, J = 8.5) ppm. δ_C (75 MHz; CDCl₃) 28.1, 35.8, 39.3, 48.7, 62.2, 116.0, 127.37, 127.42, 128.4, 128.2, 129.1, 130.3, 136.4, 139.9, 146.3, 164.3, 166.0 and 170.7 ppm. m/z (ESI) 364.15 (M + H)⁺, 386.14 (M + Na)⁺. HRMS m/z found 364.15462, calc. for C₂₂H₂₂NO₄ 364.15433.

(S)-(4-Oxo-1-pent-4-enoylazetidin-2-yl)methyl 3-(1H-indol-3-yl)propanoate (2f)

The title compound (82 mg, 63% yield) was obtained as a yellow oil from 4f (0.37 mmol). Rf = 0.60 (DCM/MeOH, 95:5, v/v). [α]_D = −12.42 (c = 0.9). ν_max/cm 2934, 1790, 1736, 1698, 1641, 1458, 1315 and 1155. δ_H (300 MHz; CDCl₃) 2.38 (2 H, m), 2.66–2.79 (5 H, m), 2.97 (1 H, dd, J = 6.4, 16.3), 3.11 (2 H, t, J = 7.4), 4.18 (1 H, m), 4.38 (1 H, dd, J = 3.1, 12, 4.50 (1 H, ddd, J = 2.1, 4.3, 12), 4.97–5.11 (2 H, m), 5.81 (1 H, m), 6.99 (1 H, d, J = 2.1), 7.09–7.26 (2 H, m), 7.35 (1 H, d, J = 8.0), 7.58 (1 H, d, J = 7.7) and 8.08 (1 H, br s) ppm. δ_C (75 MHz; CDCl₃) 20.7, 27.9, 34.7, 35.7, 38.9, 48.4, 61.5, 111.3, 114.5, 115.9, 118.7, 119.4, 121.6, 122.2, 127.1, 136.3, 136.4, 164.3, 170.7 and 173.0 ppm. m/z (ESI) 355.16 (M + H)⁺, 377.15 (M + Na)⁺. HRMS m/z found 355.16541, calc. for C₂₀H₂₃N₂O₄ 355.16523.

General procedure for N-acylation according to Method (v)

To a solution of NH-azetidinone 4 (1 equiv.) in dry tetrahydrofuran (THF) solution (10 mL/mmol) cooled at −78 °C, was added drop-wise, under argon atmosphere, a solution of LiHMDS (lithium hexamethyldisilazide) (1 M in THF, 1.1 equiv.) and then DMF (dimethylformamide) (0.1 mL/mmol). The mixture was stirred for 30 min. at −78 °C, then 4-pentenoyl chloride was added drop-wise (1.1 equiv.). After 1 h stirring at −78 °C, the solution was allowed to warm at room temperature and left 2 h at 20 °C. After concentration under vacuum, the residue was dissolved in DCM (20 mL/mmol) and washed three times with brine. The organic layer was dried over MgSO₄, concentrated and purified by flash chromatography (elution with DCM/EtOAc, 95:5, v/v) to furnish 2 (for 2g, see Supplementary Material).

(S)-(4-Oxo-1-pent-4-enoylazetidin-2-yl)methyl 2-(biphenyl-4-yl)acetate (2c)

The title compound (198 mg, 70% yield) was obtained as a yellow oil from 4c (0.75 mmol). Rf = 0.22 (DCM/EtOAc, 98:2, v/v). [α]_D = −23.1 (c = 1.7). ν_max/cm 3164, 1789, 1739, 1701, 1317 and 1153. δ_H (500 MHz; CDCl₃) 2.33–2.39 (2 H, m), 2.66–2.80 (2 H, m), 2.83 (1 H, dd, J = 3.4, 16.2), 3.06 (1 H, dd, J = 6.4, 16.2), 3.69 (2 H, s), 4.22–4.27 (1 H, m), 4.39 (1 H, dd, J = 2.9, 12.1), 4.61 (1 H, dd, J = 3.8, 12.1), 4.96–5.07 (2 H, m), 5.74–5.84 (1 H, m), 7.27–7.40 (3 H, m), 7.40–7.51 (2 H, m) and 7.51–7.67 (4 H, m) ppm. δ_C (125 MHz; CDCl₃) 27.9, 35.6, 38.9, 40.9, 48.4, 61.8,115.8, 127.1, 127.4, 127.5, 128.9, 129.7, 132.6, 136.4, 140.3, 140.7, 164.0, 170.6 and 171.0 ppm. m/z (ESI) 378.02 (M + H)⁺, 400.15 (M + Na)⁺. HRMS m/z found 400.1511, calc. for C₂₃H₂₃NO₄Na 400.1525.

Enzyme inhibition

In vitro assays for hFAAH

Tubes containing the enzyme (10 mM Tris-HCl, 1 mM ethylenediamine tetraacetic acid (EDTA), 0.1% (w/v) bovine serum albumin (BSA), pH 7.4, 165 µL), test compounds in DMSO (dimethylsulfoxide) or DMSO alone for controls (10 µL) and [³H]-AEA (50 000 dpm, 2 µM final concentration, 25 µL) were incubated at 37 °C for 10 min. Reactions were stopped by rapidly placing the tubes in ice and adding 400 µL of ice-cold chloroform/methanol (1:1 v/v) followed by vigorous mixing. Phases were separated by centrifugation at 850 g, and aliquots (200 µL) of the upper methanol/buffer phase were counted for radioactivity by liquid scintillation counting (LSC). In all experiments, tubes containing buffer only were used as control for chemical hydrolysis (blank) and this value was systematically subtracted. Using these conditions, URB-597 inhibits hFAAH with an IC₅₀ value of 40 nM.

In vitro assays for hMAGL

Tubes containing purified enzyme (10 mM Tris-HCl, 1 mM EDTA, 0.1% (w/v) BSA, pH 8.0, 165 µL), test compounds in DMSO or DMSO alone for controls (10 µL) and [³H]-2-OG (50 000 dpm, 2 µM final concentration, 25 µL) were preincubated at 20 °C for 30 min and incubated at 37 °C for 10 min. Reactions were stopped by rapidly placing the tubes in ice and adding 400 µL of ice-cold chloroform/methanol (1:1 v/v) followed by vigorous mixing. Phases were separated by centrifugation at 850 g, and aliquots (200 µL) of the lower chloroform phase were counted for radioactivity by LSC. In all experiments, tubes containing buffer only were used as control for chemical hydrolysis (blank) and this value was systematically subtracted.

Reversibility studies

Human recombinant FAAH (27.5 µg) was incubated in the presence of the inhibitors or DMSO during 1 h at room temperature, in a total of 15 µL assay buffer (Tris 10 mM, EDTA 1 mM, pH 7.4). Following this step, assay buffer was added to obtain a 100-fold dilution of the samples. A 175 µL aliquot was then taken after 0, 30 or 90 minutes and AEA, diluted in the same buffer supplemented with 0.1% BSA, was immediately added ([³H]-AEA; 25 µL; 50 000 dpm, 2 μM final AEA concentration). The hydrolysis reaction was allowed to proceed during 30 min at 37 °C and 400 μL of ice-cold chloroform/methanol (1:1 v/v) were added to stop the reaction, followed by vigorous mixing. Tubes were centrifuged at 850 g during 5 min and the radioactivity present in the upper aqueous phase was measured by LSC. Tubes lacking the enzyme were used as blanks, the value of which was systematically subtracted from other values. Results were expressed as per cent of control activity (three experiments performed in triplicate). Initial compound concentrations ([2a] = 100, 150 and 250 µM; [URB597] = 5 µM; [CAY10402] = 10 nM) were carefully selected in preliminary experiments omitting the dilution step, so that maximal and minimal inhibition should be observed at the concentrations obtained before and after the dilution, respectively. The concentrations of URB597 and CAY10402 correspond to about 100 times their respective IC₅₀ determined in our laboratory.

The percentage of recovery of hFAAH activity after dilution was calculated according to the following formula: where %^act high []^{set up} = percentage of activity in high inhibitor concentration (before dilution); %^act low []^{set up} = percentage of activity in low inhibitor concentration (after dilution).

Computational chemistry

The previously determined modelled structure of hFAAH was used28. Docking of the inhibitors into the active site of FAAH was performed using the GOLD program33. It is based on a genetic algorithm (GA), performing docking of flexible ligands into proteins with partial flexibility in the neighbourhood of the active site. The binding region was defined as a 17 Å radius sphere centred on Phe192. Two explicit water molecules were allowed to switch on and off and to rotate around their three principal axes. For the 20 GA runs performed, a total of 100 000 genetic operations were carried out on 5 islands, each containing 100 individuals. The niche size was set to 2, and the value for the selection pressure was set to 1.1. Genetic operator weights for crossover, mutation and migration were set to 95, 95 and 10, respectively. The solutions were then ranked by GOLD score. The GOLD fitness function is made up of four components: protein-ligand hydrogen bond energy, protein-ligand van der Waals energy, ligand internal van der Waals energy and ligand torsional strain energy.

The compounds surrounded by the closer amino acids were optimised by a combined hybrid quantum mechanics/molecular mechanics method at the B3LYP/dzp/PM6 level using the GAUSSIAN 09 program package34. The figures were produced using PyMOL35.

Results and discussion

Synthesis

The chiron B is prepared in four steps from (S)-aspartic acid according to the literature31,32 (Scheme 1). Selective reduction of the benzyl ester function is obtained with sodium borohydride in quantitative yield36. The esterification reaction of the C₄ hydroxymethyl substituent is performed via two methods, namely the condensation of a carboxylic acid in the presence of DCC and 4-DMAP, or the acylation by an acid chloride in the presence of pyridine. The choice of the end-chain substituent R and the chain length n is inspired from the series 1 of inhibitors previously published37,38. Except for 4e, the yields of esters 4a–h (Table 1) are excellent after chromatographic purification. Most of the reagents are commercially available; 4-(pyridin-4-yl)-butanoic acid and 2-(pyridin-4-yl)-butanoic acid have been prepared by Michael addition of dimethyl malonate on vinylpyridines, followed by saponification and decarboxylation (see Supplementary Material)39,40.

Scheme 1. Synthesis of 1,4-disubstituted β-lactams: (i) NaBH₄, MeOH, 0–20 °C, 3 h; (ii) carboxilic acid, DCC, DMAP (catal.), DCM, 20 °C, 24 h; (iii) acid chloride, pyridine or DMAP, DCM, 20 °C, 24 h; (iv) ClO-(CH₂)₂-CH=CH₂, pyridine, DCM, reflux, 24 h; (v) ClCO-(CH₂)₂-CH=CH₂, LiHMDS, THF-DMF, −78–20 °C, 1 h. See Table 1 for yields of 4 and 2.

Table 1. Yields of compounds (%) of the Scheme 1.

n	R*	4	Method	Yield (%)	2	Method	Yield (%)	Overall yield from B
3	Ph	4a	iii	72	2a	iv	43†	31
0	biPh	4b	iii	78	2b	v	65‡	51
1	biPh	4c	ii	87	2c	v	70	61
3	2-Pyr	4d	ii	79	2d	iv	17	13
3	4-Pyr	4e	ii	44	2e	iv	19	8
2	3-Ind	4f	ii	92	2f	iv	63	58
1	O-Bn	4g	ii	87	2g	v	66	57
2	O-Ph	4h	iii	75	2h	v	<5	<4

*  †Method v gave 40% yield.

‡Method iv gave 55% yield.

The next step is the N-acylation with 4-pentenoyl chloride to furnish the series of β-lactams 2a–h equipped with the N₁ chain of the best inhibitors of the series 1 (Figure 3). Deprotonation of 4 with lithium hexamethyldisilazide at low temperature and quenching of the anion by the acid chloride41 give the expected products 2, except 2h (Table 1). We postulate that the intermediate anion of 4h is inert, because of the formation of a stable O,O,N-complex with the lithium cation (Figure 4). The regioisomeric structure, i.e. anion from 4g, should not form similar complex due to cyclic strain. Prolonged treatment of the precursors 4 with 4-pentenoyl chloride and pyridine in refluxing DCM leads also to the desired compounds 2 in few cases, but in low yields (Table 1). The β-lactams 2 have been purified by column chromatography and characterised by the usual spectroscopic methods. For instance, the ¹³C NMR spectra show three characteristic carbonyl signals, at 164 ppm (β-lactam), 170 ppm (imide) and 171–173 ppm (ester). In the ¹H NMR spectra, the β-lactamic protons give a typical ABX pattern, with H₃ at 2.9 ppm (dd, J_3,4trans = 3 Hz and J_3,3′ = 16 Hz), H_3′ at 3.2 ppm (dd, J_3′,4cis = 6 Hz and J_3,3′ = 16 Hz) and H₄ at 4.0 ppm (m). Another ABX pattern is visible, attributed to the CH₂O motif fixed at C₄: one doublet of doublet at 4.3 ppm (J = 6 and 12 Hz) and another at 4.6 ppm (J = 4 and 12 Hz).

Figure 4. Intermediates resulting from the deprotonation of 4g and 4h, respectively.

Enzyme inhibition

Human recombinant enzymes, namely, hFAAH42 and hMAGL43, are prepared according to the literature. The β-lactams 2 are tested in a competitive hydrolytic assay using [³H]-labelled substrates, i.e. [³H]-AEA in the case of hFAAH and [³H]-2-OG in the case of hMAGL. Briefly, tested compound, enzyme and labelled substrate are incubated at 37 °C during a few minutes and the extent of enzyme inhibition is then evaluated by LSC of [³H]-ethanolamine and [³H]-glycerol resulting from the hydrolysis of [³H]-AEA and [³H]-2-OG, respectively (Table 2).

Table 2. hFAAH and hMAGL inhibition.

			IC50/FAAH (µM)		IC50/MAGL (µM)
Compound	n	R*	Series 2	Series 1†	Series 2	Series 1†
a	3	Ph	5.03	0.005	108.10	4.06
b	0	biPh	0.91	nd	nd	nd
c	1	biPh	7.82	0.012	nd	1.84
d	3	2-Pyr	87.06	nd	nd	nd
e	3	4-Pyr	173.0	0.310	nd	nd
f	2	3-Ind	8.05	nd	166.10	nd
g	1	O-Bn	42.52	0.031	nd	nd

*See Table 1.

†Results of references 37 and 38.

All compounds 2 present a micromolar activity against hFAAH, with IC₅₀ values ranging from 0.9 to 173 µM. The less active inhibitors feature a pyridine motif in the C₄ side-chain. Comparatively to the lead compound 1a of the previous series, 2a appears 1000 times less potent for inhibiting hFAAH. Similarly, compounds 2c, 2e and 2g are, respectively, 650, 560 and 1370 times less active than the corresponding inhibitors 1c, 1e and 1g28,37,38. Compounds 2 behave as quite selective inhibitors of hFAAH versus hMAGL: the selectivity factor of 2a and 2f is about 20. However, the selectivity is more pronounced for the lead compound 1a (S = 800)28,37,38.

We have previously demonstrated that 1a is a reversible inhibitor of hFAAH: it is not processed by the enzyme according to HPLC/HRMS studies37, and the enzyme activity is recovered after a rapid and large dilution of enzyme/inhibitor mixtures28. The mode of action of 2a has been similarly controlled by wash-out experiments, using URB-597 (belonging to the carbamate family of FAAH inhibitors, Figure 1) and CAY-10402 (belonging to the α-keto-heterocycle family of FAAH inhibitors) as standards for the irreversible and reversible inhibitions, respectively38. Briefly, hFAAH and 2a (at 100, 150 and 200 µM) are incubated (1 h, 25 °C), then 100-fold diluted with buffer. Aliquots are taken (after 0, 30 and 90 min) for enzyme residual activity measurement. As shown in Figure 5, the enzyme activity is totally recovered after rapid dilution of the 2a/hFAAH mixtures, indicating a reversible mode of action, like CAY-10402. On the contrary, the hFAAH activity remains largely inhibited in the case of URB-597.

Figure 5. Percentage of recovery of hFAAH activity after dilution with 2a versus controls.

Thus, 2a and 1a exhibit very similar reversible inhibition profiles but the former β-lactam is 1000 times less potent, and 40 times less selective, than the latter. This unexpected and disappointing result could not be predicted from the docking of the novel compounds 2a in the simple model of hFAAH active site and the comparison with the previous inhibitor 1a.

Docking studies

The position and confirmation of the amino acids of the modelled hFAAH active site we used in 200928 is close to the crystallographic structure of h/rFAAH published in 201144, where the authors have been able to identify both non-covalent and covalent states of a α-keto-heterocycle reversible inhibitor trapped into the catalytic pocket. The hFAAH active site is formed by a hydrophobic tunnel (named ACB for acyl chain binding), the hydrophilic catalytic triad (Ser241, Ser217 and Lys142), a second tunnel exposed to the solvent (named CA for cytoplasmic access) and a third channel, close to the ACB channel, composed of three phenylalanine residues (Phe338, Phe381 and Phe192) and named as “phenyl pocket”. In this model, the same binding mode (named mode 1)28 is found for 1a and 2a docked as non-covalent complexes in the active site as they are reversible, competitive inhibitors (Figure 6). The molecules feature the syn-conformation of the imide motif (formed by the β-lactam carbonyl C₂=O and the exocyclic carbonyl N₁-C=O). Hydrogen bonds are especially observed between Ser241 and Ser217 and the exocyclic CO, and between Thr236 and the β-lactam CO. The phenyl chain interacts with the “phenyl pocket” and the alkene chain lies at the beginning of ACB channel. The complexes of 1a and 2a have been optimised and their interaction energies have been calculated at the B3LYP-PM6 level. In this binding mode 1, the 2a complex is found more stable than the 1a complex (ΔE = 6.7 kcal/mol). This result is in disagreement with our experimental values of inhibition and thereby prompts us to consider water molecules into the active site model, which will induce different binding modes, allowing possibly the discrimination between the series of inhibitors.

Figure 6. Binding mode 1 of 1a and 2a in the active site of hFAAH. Hydrogens were removed for clarity.

The modelling and analysis of crystallographic data featuring waters in recently solved h/rFAAH structures have confirmed their presence as well as the importance of two water molecules being close to the catalytic triad16. We, therefore, considered docking studies, including them in the active site by using the option of GOLD allowing water molecules to switch on and off and to rotate45. Such strategy was repeatedly shown to better model the interactions of inhibitors into enzyme cavity46,47. Two models can be built: the first with only the water molecule closest to the triad (1 W model), and the second with both water molecules (2 W model).

The binding mode 1 appears still compatible with the presence of one water molecule (mode 1-1 W) in interaction with the imide function of the inhibitor (1a or 2a) and the active serine residue; in that position both complexes have the same stabilisation energy (ΔE = 0.7 kcal/mol) (see Figure S1 in Supplementary Material).

Another binding mode, named mode 2, can be found where the water molecule makes hydrogen bonds with the Ser241 residue and the ester function of the inhibitor (1a or 2a). In this pose, the imide conformation is anti and the phenyl chain lies in the ACB channel (Figure 7), as observed in the crystallographic complex formed by h/rFAAH and OL-13544. Here, computations suggest the 1a complex as the most favourable one (ΔE = 8.8 kcal/mol). With two water molecules in the active site (in interaction with Ser241), another binding mode (named mode 3) is revealed for both inhibitors 1a and 2a (Figure 8). The β-lactam carbonyl is turned to Ser241, the phenyl chain is located in the “phenyl pocket” and the alkene chain occupies the CA channel. The binding mode 3 gives a more stable complex for 1a, compared with 2a (ΔE = 16.6 kcal/mol). This binding mode is not compatible with the anti-conformation of the imide motif. Binding modes 2 and 3 are therefore in agreement with the experimentally determined potency of the compounds.

Figure 7. Binding mode 2-1 W of 1a and 2a in the active site of hFAAH. Hydrogens were removed for clarity.

Figure 8. Binding mode 3-2 W of 1a and 2a in the active site of hFAAH. Hydrogens were removed for clarity.

In the novel modes 2 and 3, like in mode 1, the C₄ chain of β-lactams lies in a hydrophobic environment slightly compatible with a pyridinyl substituent. This may explain the drop of activity observed for compounds 2d and 2e. The reduced activity of 2g featuring an ether function in the phenyl chain (2g is an isostere of 2a) can be similarly justified.

Conclusion

The search for non-covalent, reversible, competitive inhibitors of hFAAH is an important topic in current drug discovery, already illustrated by various heterocyclic structures48, such as benzoxazoles49, benzothiazoles50, benzimidazoles51, tetrahydropyridopyridines52 and piperidines53.

We previously demonstrated that β-lactams, which are usually covalent inhibitors of serine-enzymes by acylation reaction, behave as non-covalent inhibitors of hFAAH as exemplified with the lead compound 1a28,37,38. Interestingly, OMDM169 (β-lactone) was also reported to be a reversible inhibitor of MAGL20.

In this paper, our idea was to develop further efficient and selective hFAAH β-lactamic inhibitors, acting similarly to our previous lead compound 1a, and rapidly produced (three steps) from a cheap chiral starting material, available on large scale (precursor B). We could indeed easily obtain the novel series of β-lactams 2, but their inhibition activities against hFAAH were disappointing, although the molecules fitted perfectly into our first enzyme active site model, with an interaction energy even more favourable than 1a. Therefore, we questioned about the hFAAH model relevance as a predictive tool for the design of novel active β-lactams.

Since hFAAH is not yet crystallised, we considered the recent crystallographic data of h/rFAAH suggesting to take into account the presence of linked water molecules into the active site. Two water molecules identified in the free enzyme are still present in the enzyme/inhibitor complex (acyl-enzyme intermediate)16, or replaced by the inhibitor (tetrahedral intermediate)14,15,44,54 covalently fixed on Ser241. New docking experiments of β-lactams 1a and 2a have been performed by introducing two water molecules near the catalytic triad. Then, we have left the choice to remove one or two water molecules. The results show that at least one water molecule is conserved and new binding modes of the inhibitors have been identified. In particular, the positioning of the C₄ phenyl chain in the “phenyl pocket” and the alkene chain in the CA channel, instead of the ACB tunnel, has revealed a significantly higher stabilisation of the 1a non-covalent complex compared with 2a. This binding mode 3 could possibly explain the micromolar activity of 2a regarding the nanomolar activity of 1a.

In the absence of available experimental data on co-crystals formed from hFAAH and β-lactamic inhibitors, h/rFAAH active site models and related docking experiments can be used for the design of new inhibitors, though their outputs are mainly qualitative. At the present time, the biochemical screening of oriented libraries of compounds remains the most suitable strategy of lead discovery55,56. This is particularly true in the field of β-lactam derivatives which involve a privileged structure (azetidin-2-one) commonly found in natural products and (semi)-synthetic drugs. In this context, our series of β-lactams could be advantageously tested against serine hydrolases, including diacylglycerol lipase also involved in the EC system19.

Declaration of interest

This work was supported by the UCL, F.R.S.-FNRS (Belgium) and partially by the PAI program PROFUSA P6/19. C. M., J. M.-B., R. R. and E. A. P. are senior research associates of the F.R.S.-FNRS (Belgium).

The authors have no conflicts of interest.

Supplementary material available online Preliminary screening of tri-substituted β-lactams, synthesis of 4- and 2-(pyridin-4-yl)butanoic acids and docking of 1a and 2a into hFAAH with binding mode 1-1w are given as Supplementary material.