Synthesis, biological evaluation, and molecular modelling studies of potent human neutrophil elastase (HNE) inhibitors

Abstract We report the synthesis and biological evaluation of a new series of 3- or 4-(substituted)phenylisoxazolones as HNE inhibitors. Due to tautomerism of the isoxazolone nucleus, two isomers were obtained as final compounds (2-NCO and 5-OCO) and the 2-NCO derivatives were the most potent with IC50 values in the nanomolar range (20–70 nM). Kinetic experiments indicated that 2-NCO 7d and 5-OCO 8d are both competitive HNE inhibitors. Molecular modelling on 7d and 8d suggests for the latter a more crowded region about the site of the nucleophilic attack, which could explain its lowered activity. In addition molecular dynamics (MD) simulations showed that the isomer 8d appears more prone to form H-bond interactions which, however, keep the reactive sites quite distant for the attack by Ser195. By contrast the amide 7d appears more mobile within the active pocket, since it makes single H-bond interactions affording a favourable orientation for the nucleophilic attack.


Introduction
Proteases are enzymes implicated in cellular reactions involving the cleavage of protein substrates 1 . Serine proteases are characterised by the presence of a serine residue at the active site 2 . They are divided into four classes: chymotrypsin, subtilisin, carboxypeptidase Y, and caseinolytic protease 3 . Human neutrophil elastase (HNE), proteinase 3 (PR3), cathepsin G, and the recently discovered NSP4 4 are serine proteases belonging to the chymotrypsin family and represent neutrophil serine proteases (NSP) 5,6 . NSP are synthesised and expressed in neutrophil azurophilic granules. Neutrophils play a pivotal role in host defence, inflammation and tissue remodelling, and HNE is a key mediator of neutrophil-driven inflammation 7 .
HNE is a small, basic, and soluble glycoprotein of about 30 kDa 8 that performs many functions in our body. For example, HNE is involved in the maintenance of tissue homeostasis; it degrades a variety of structural proteins of the extracellular matrix, such as elastin, fibronectin, collagen, proteoglycan, and laminin; and it repairs damaged tissue 9 . Moreover, HNE plays an important and dual role in inflammation by degrading pro-inflammatory cytokines to reduce the intensity of the inflammation but also increasing the secretion of pro-inflammatory factors 10 . In the case of infection, HNE plays as an intracellular function in destroying phagocytosed pathogens, as well as an extracellular function through the formation of neutrophil extracellular traps, which can trap and kill microorganisms 11,12 . The powerful HNE activity is tightly controlled by the presence of extracellular neutralizing endogenous serine protease inhibitors, such as a-1 antitrypsin, secretory leucocyte protease inhibitor (SLPI), and elafin 13 . a1-Antitrypsin is a water-soluble glycoprotein classified as systemic HNE inhibitor that is synthesised in the liver and is particularly abundant in the lungs 14 . In contrast, SLPI and elafin are classified as alarm inhibitors because they are produced and released directly into the airway epithelium in response to the release of cytokines, regulating the immune response and inflammatory processes [15][16][17] . Under physiological conditions, the balance between protease and anti-protease supports the maintenance of tissue homeostasis. However, if this balance fails, excessive HNE activity can cause tissue the damage associated with some serious chronic diseases 4,6 . Among the pathologies associated with increased HNE activity are adult respiratory distress syndrome (ARDS) 18 , chronic obstructive pulmonary disease (COPD) 19,20 , cystic fibrosis (CF) 21,22 , and other disorders with an inflammatory component, such as rheumatoid arthritis 23 , atherosclerosis 24 , psoriasis, and dermatitis 25 . Recently, it was demonstrated that HNE is also implicated in the progression of non-small cell lung cancer 26 .
Although a large number of molecules have been reported as HNE inhibitors 27,28 , only two drugs are currently available for clinical use: Prolastin (purified a1-AT), a peptide drug synthesised by recombinant DNA techniques 29 and used for the treatment of a1antitripsin deficiency (AATD) 30 , and Sivelestat (Elaspol V R 100), a nonpeptide low molecular weight compound, belonging to the second generation of HNE inhibitors 31 . Sivelestat has an IC 50 ¼ 44 nM and is currently marketed only in Japan and South Korea 32,33 (Figure 1). AZD9668 (Alvelestat, AstraZeneca) 34 and BAY 85-8501 (Bayer HealthCare) 35 are two potent HNE inhibitors, belonging to the third and fifth generations of HNE inhibitors 31 , respectively, that have recently reached Phase II of clinical trials for COPD, CF, and BE ( Figure 1).
In recent research focused on the development of new HNE inhibitors, we investigated various bicyclic scaffolds, such as indazole 36,37 , indole 38 , and cinnolinone 39 , compounds. The most interesting inhibitors were effective in the nanomolar range, with a potency comparable to Sivelestat. Subsequently, we focused our research on the design and synthesis of monocyclic derivatives with an isoxazol-5(2H)-one core, and the first series of 3/4-alkyl-(di)substituted isoxazolones was recently published 40 . We report here the synthesis of a new series of isoxazolone derivatives bearing a (substituted)phenyl at positions 3 and 4 and biological evaluation of their HNE inhibitory activity.

Material and methods
All melting points were determined on a B€ uchi apparatus (New Castle, DE) and are uncorrected. Extracts were dried over Na 2 SO 4 , and the solvents were removed under reduced pressure. Merck F-254 commercial plates (Merck, Durham, NC) were used for analytical TLC to follow the course of reactions. Silica gel 60 (Merck 70-230 mesh, Merck, Durham, NC) was used for column chromatography. 1 H NMR, 13 C NMR, HSQC, HMBC, and NOESY bidimensional spectra were recorded on an Avance 400 instrument (Bruker Biospin Version 002 with SGU, Bruker Inc., Billerica, MA). Chemical shifts (d) are reported in ppm to the nearest 0.01 ppm using the solvent as an internal standard. Coupling constants (J values) are given in Hz and were calculated using TopSpin 1.3 software (Nicolet Instrument Corp., Madison, WI) and are rounded to the nearest 0.1 vHz. Mass spectra (m/z) were recorded on an ESI-TOF mass spectrometer (Brucker Micro TOF, Bruker Inc., Billerica, MA), and reported mass values are within the error limits of ± 5 ppm mass units. Microanalyses indicated by the symbols of the elements or functions were performed with a Perkin-Elmer 260 elemental analyser (PerkinElmer, Inc., Waltham, MA) for C, H, and N, and the results were within ± 0.4% of the theoretical values, unless otherwise stated. Reagents and starting material were commercially available. (1.14 mmol), and 1-(chloromethyl)-3-methylbenzene (0.86 mmol) in 2 ml of anhydrous acetonitrile was stirred at reflux for 2 h. After cooling, the mixture was concentrated in vacuo, diluted with ice-cold water (10 ml), and extracted with ethyl acetate (3 Â 15 ml). The organic phase was dried over sodium sulphate and the solvent was evaporated in vacuo to afford the final compound 2, which was purified by column chromatography using cyclohexane/ethyl acetate 2:1 as eluent. Yield ¼57%; oil. 1    General procedure for compounds (3a-c) To a suspension of the appropriate 4-substituted benzensulfonyl chloride (0.16 mmol) in 3 ml of anhydrous pyridine, 0.79 mmol of intermediate 1a 41 was added. The mixture was stirred at room temperature for 4 h. The solvent was concentrated in vacuo to afford the final compounds 3a-c which were purified by column chromatography using cyclohexane/ethyl acetate in different ratio (2:1 for 3a, 4:1 for 3b) or toluene/ethyl acetate 9:1 for 3c as eluents.
3-Methyl-2-(3-methylbenzoyl)-4-phenylisoxazol-5(2H)-one (4a). Yield ¼ 52%; mp ¼ 85-88 C (EtOH). 1           General procedure for compounds (4i, 4l) To a suspension of 4-(pivaloyoxy)benzoic acid 45 or 4-(pivalamido)benzoic acid 46 (0.32 mmol) in 1 ml of anhydrous toluene, 0.64 mmol of SOCl 2 , and a catalytic amount of DMF (0.05 mmol) were added. The mixture was stirred at reflux for 2 h. The solvent was concentrated in vacuo and the crude compound was used without purification and added to a previously prepared solution composed of 0.29 mmol of intermediate 1a 41 and 0.64 mmol of sodium hydride in 5 ml of anhydrous THF. The mixture was stirred at room temperature overnight. After evaporation of the solvent, the product was purified by column chromatography using toluene/ethyl acetate 9:1 for 4i and hexane/acetone 4:1 for 4l as eluents.  General procedure for compounds (4m, 4u) To a solution of intermediate 1b 42 or 1f 47 (0.32 mmol) in 2 ml of t-BuOH, 0.35 mmol of K 2 CO 3 and 0.64 mmol of m-toluoyl chloride were added. The mixture was stirred at reflux for 3 h. The solvent was concentrated in vacuo, diluted with ice-cold water (10 ml), and extracted with DCM (3 Â 15 ml). The organic phase was dried over sodium sulphate and the solvent was evaporated in vacuo to afford the final compounds 4m,u, which were purified by column chromatography using cyclohexane/ethyl acetate in different ratio: 5:1 for 4m and 1:1 for 4u as eluent.  General procedure for compounds (5e, 5f) To a cooled (0 C) suspension of 5d 48 (0.68 mmol) in anhydrous CH 2 Cl 2 (2 ml), Et 3 N (1.36 mmol) and 2.04 mmol of appropriate acyl chloride were added. The mixture was stirred at 0 C for 2 h and then at room temperature for an additional 2 h. The solvent was evaporated, cold water was added and the mixture was neutralised with 0.5 N NaOH. The reaction mixture was extracted with CH 2 Cl 2 (3 Â 15 ml), the solvent was dried over sodium sulphate, evaporated in vacuo, and compounds 5e and 5f were purified by column chromatography using dichloromethane/methanol (9:1 for 5e; 99:1 for 5f) as eluents. General procedure for compounds (6a-e) A 3.00 mmol of appropriate intermediate 5a-c [48][49][50] and 5e,f was dissolved in 1.3 ml of water and heated at 80 C. To this solution, 3.3 mmol of hydroxylamine hydrochloride in 6.5 ml of methanol was added. The mixture was stirred at reflux for 5 h. After evaporation of the solvent, the residue was mixed with ice-cold water (20 ml). Compounds 6a,c were recovered by extraction with ethyl acetate (3 Â 15 ml), while compounds 6b,d,e were recovered by vacuum filtration. The final compounds 6b,d were purified by crystallisation with ethanol, while the compounds 6a,c,e were purified by column chromatography using dichoromethane/methanol 9:1 as eluent.  N-(4-(3-methyl-5-oxo-2,5-dihydroisoxazol-4-yl)phenyl)cyclopropanecarboxamide (6e). Yield ¼ 58%; mp ¼ 208-210 C (EtOH). 1  General procedure for compounds (7a-e, 8a,b,d,e) Compounds 7a-e and 8a,b,d,e were obtained following the same procedure performed for compounds 4a-h, 4n-t but starting from precursors 6a-e. The solvent was concentrated in vacuo to obtain the final compounds which were purified by column chromatography using petroleum ether/ethyl acetate 10:1 for 7a/8a, hexane/ ethyl acetate (5:1 for 7b/8b; 5:2 for 7c), dichloromethane/methanol (98:2 for 7d/8d; 99:1 for 7e/8e) as eluents.

General procedure for compounds (10a-c)
To suspension of the substrate 9 51 (0.37 mmol) in tert-Butanol (3 ml), K 2 CO 3 (0.41 mmol), and 0.74 mmol of the appropriate acyl chloride were added. The mixture was stirred at reflux for 3 h. After evaporation of the solvent, the residue was mixed with icecold water (20 ml) and extracted with ethyl acetate (3 Â 15 ml). The organic phase was dried over sodium sulphate, and the solvent was evaporated in vacuo to afford the final compounds 10a-c, which were purified by column chromatography using cyclohexane/ethyl acetate (5:1) as eluent.   (C, H, N).

HNE inhibition assay
Compounds were dissolved in 100% DMSO at 5 mM stock concentrations. The final concentration of DMSO in the reactions was 1%, and this level of DMSO had no effect on enzyme activity. The HNE inhibition assay was performed in black flat-bottom 96-well microtiter plates. Briefly, a buffer solution containing 200 mM Tris-HCl, pH 7.5, 0.01% bovine serum albumin, and 0.05% Tween-20 and 20 mU/mL of HNE (Calbiochem) was added to wells containing different concentrations of each compound. The reaction was initiated by addition of 25 mM elastase substrate (N-methylsuccinyl-Ala-Ala-Pro-Val-7-amino-4-methylcoumarin, Calbiochem) in a final reaction volume of 100 ml/well. Kinetic measurements were obtained every 30 s for 10 min at 25 C using a Fluoroskan Ascent FL fluorescence microplate reader (Thermo Electron, MA) with excitation and emission wavelengths set at 355 and 460 nm, respectively. For all compounds tested, the concentration of inhibitor that caused 50% inhibition of the enzymatic reaction (IC 50 ) was calculated by plotting % inhibition versus logarithm of inhibitor concentration (at least six points). The data are presented as the mean values of at least three independent experiments with relative standard deviations of <15%.

Analysis of compound stability
Spontaneous hydrolysis of selected derivatives was evaluated at 25 C in 0.05 M phosphate buffer, pH 7.3. Kinetics of hydrolysis were monitored by measuring changes in the absorbance spectra over time using a SpectraMax Plus microplate spectrophotometer (Molecular Devices, Sunnyvale, CA). Absorbance (A t ) at the characteristic absorption maxima of each compound was measured at the indicated times until no further absorbance decreases occurred (A 1 ) 52 . Using these measurements, we created semilogarithmic plots of log(A t -A 1 ) versus time, and k 0 values were determined from the slopes of these plots. Half-conversion times were calculated using t 1/2 ¼ 0.693/k 0 , as described previously 36,37 .

Molecular modelling procedures
The programs used for the energy minimisation, MD, and docking were the simulation protocols Minimisation, Standard Dynamics Cascade, Analyse Trajectory, and CDocker implemented in Accelrys Discovery Studio 2.1 53 . The Force Field used for all simulations was CHARMm 54 .
The following parameters were used for MD simulations, both in vacuum and in implicit solvent (the latter was simulated by using distance dependent dielectric constant set to 4r): time step ¼ 1 fs, equilibration time ¼ 100 ps, production time ¼ 1000 ps (5000 ps for the inhibitor-HNE assembly), T ¼ 300 and 600 K. Ten snapshot conformations with evenly spaced intervals were extracted from each MD trajectory and subsequently minimised (using the steepest descent and conjugate gradient algorithms) in order to obtain the starting geometries for the subsequent quantum chemical calculations (QC) and MD simulations with the receptor. GAUSSIAN09 (Rev. C01) 55 was used for quantum chemical calculations (QC) on 7d and 8d by using the B3LYP 56,57 and B97D 58 functionals. The basis set was 6-31 þ G(d,p) 59 , and the Berny algorithm was used 60 . Reliability of the stationary points was assessed by evaluation of the vibrational frequencies. For each inhibitor, different conformational isomers, chosen from amongst the low lying energy conformers (as found in MD simulations and subsequent geometry optimisation) were considered. Molecular plots were produced by the program Discovery Studio Visualiser (v 4.5) 61 .

Results and discussion
Chemistry It is well-known that the isoxazolone nucleus exhibits three tautomers 43,62,63 , as illustrated below. The NH form seems to be the most representative, especially in polar solvents 62,64-66 . However, it is possible to find examples of alkylation and acylation products in the literature, as shown by the NH and the OH forms 67 . Taking into account this information, we performed the synthesis of our final compounds, as shown in Schemes 1-3, and the structures were confirmed on the basis of analytical and spectral data.
Scheme 1 depicts the synthetic pathway followed to obtain the final 2-N-substituted isoxazolones of type 2, 3, and 4, which have different groups at positions 3 and 4. The previously described key intermediate isoxazol-5-(2H)-ones of type 1 were treated under various conditions to obtain the final compounds 2, 3, and 4. The alkylation of 1a 41 with 3-methylbenzyl chloride and K 2 CO 3 in anhydrous acetonitrile at reflux resulted in compound 2. On the other hand, treatment of compounds 1a-f (1a 41 , 1b 42 , 1c,d 43 , 1e 44 , and 1f 47 ) with the appropriate acyl/aroyl chloride and NaH in anhydrous THF at room temperature (compounds 4a-l,n-t) or with m-toluyl chloride, and K 2 CO 3 in t-ButOH at 80 C (compounds 4m,u) resulted in the corresponding 2-NCO derivatives of type 4. Likewise, we synthesised the sulfonamide derivatives 3a-c by treatment of intermediate 1a 41 with the appropriate commercially available phenyl sulfonyl chloride in pyridine at room temperature. All of these reactions led to a single derivative originating from the NH form of the isoxazolone nucleus, in agreement with previous data reported in literature. Scheme 2 shows the synthetic procedures used to obtain the final compounds of type 7 and 8, which have a 4-substituted phenyl at position 4. b-Ketoesters 5a-d (5a 49 , 5b,d 48 , and 5c 50 ) were synthesised as described previously [48][49][50] , while compounds 5e,f were obtained by acylation of 5d 48 with the appropriate acyl chloride and Et 3 N in dichloromethane. These compounds served as the starting material for synthesis of the key intermediates of type 6 with an isoxazolone nucleus. Cyclisation of 5a-c,e,f with hydroxylamine hydrochloride in a mixture MeOH/H 2 O 1:1 at reflux resulted in the intermediates 6a-e which, in turn, were treated with m-toluoyl chloride under the same conditions reported in Scheme 1. Unexpectedly this last step resulted the pair of isomers of type 7 and 8 (NCO/OCO ratio 3:1), with the only exception being the nitro derivative 7c, which was obtained only in the N-CO form.
Assignment of the isomer structures was first performed using the 1 H NMR chemical shift value of the methyl at position 3. Compounds 4a-l (Scheme 1) had shifts of 2.8 ppm, which is characteristic of NCO derivatives. Likewise, we attributed the NCOstructure to compounds with a 2.8 ppm shift for 3-CH 3 (7a-e) and the OCO-structure to compounds with a 2 .4 ppm shift (8a,b,d,e) (see Supporting Information). In order to verify our findings, we performed additional techniques, such as IR spectroscopy and 2 D NMR ( 1 H-13 C HSQC, 1 H-13 C HMBC, and 1 H-1 H NOESY).
In Scheme 3, we show the synthetic pathways used to obtain the benzoisoxazolone derivatives 10a-c, which are an elaboration of the previous isoxazolone scaffold. The acylation of benzoisoxazolone intermediate 9 51 with the appropriate acyl chloride and potassium carbonate in t-BuOH led to final compounds 10a-c.

Biological evaluation and structure-activity relationship (SAR) analysis
All compounds were evaluated for their ability to inhibit HNE in comparison with Sivelestat, a reference HNE inhibitor, and the results presented in Tables 1-3. Previously 40 , we examined the possibility of a two point of attack for Ser195, 2-NCO, and 5-CO, and docking studies confirmed that the endocyclic C ¼ O at position 5 was involved in the catalysis process, whereas the amidic C ¼ O group was important for anchoring to the sub-pocket of the binding site. We could expect a similar trend in this new series of isoxazolones.
Beginning analysis of the data with the 3-methyl-4-phenylisoxazol-5(2H)-one derivatives (compounds 4a-l), the results reported in Table 1 suggest that, similar to our previous series 40 , the best substituents at position N-2 are a m-methylbenzoyl or a cyclopropanecarbonyl fragment, which resulted in compounds 4a and 4b that were active in the nanomolar range (IC 50 ¼ 77 and 59 nM, respectively). Displacement of the methyl group from the meta to the para (4c) or ortho (4d) positions on the phenyl led to compounds with activity one order of magnitude lower than 4a, which is different than our previous series 40 Scheme 1. Reagents and conditions: (a) 3-methylbenzyl chloride, K 2 CO 3 , anhydrous CH 3 CN, 80 C, 2 h; (b) 4-X-Ph-SO 2 Cl, anhydrous pyridine, r.t., 4 h; (c) for 4a-l, n-t: R 2 -COCl, NaH, anhydrous THF, r.t., 24 h; for 4m,u: m-toluoyl chloride, K 2 CO 3 , t-BuOH, 80 C, 3 h. to the para position had no effect on HNE inhibitory activity. Substitution of m-methyl with other groups, such as trifluoromethyl (4e), cyano (4g,h), or methylsulfonyl (4f), which are found in other potent HNE inhibitors, was not favourable for activity, and only compound 4e exhibited activity in the submicromolar range (IC 50 ¼ 200 nM).
To evaluate the importance of the 2-CO amidic group in this series, we synthesised the alkyl derivative 2, which was completely inactive, suggesting that the carbonyl group is important for HNE inhibitory activity and also probably involved in catalysis. The insertion of the 3-methyl-4-phenylisoxazol-5(2H)-one scaffold of the sulfonamide fragment of the drug Sivelestat at position 2 led to compound 3b, which exhibited comparable activity to Sivelestat (IC 50 ¼ 59 nM, Table 1). In contrast, its pivalamide derivative (3c) was completely inactive, suggesting that the possible point of attack of Ser195 is the CO of the pivalate function, which is also present in Sivelestat. These data were also confirmed by the inactivity of the hydrolysed derivative 3a. Moreover, we synthesised compounds 4l and 4i by substituting the SO 2 group at position 2 (3b and 3c, respectively) with an amidic function. The amide 4l had activity (IC 50 ¼ 0.48 mM) that was one order of magnitude higher than the corresponding sulfonamide 3b, while the amide 4i had a higher potency (IC 50 ¼ 8.5 mM) than its corresponding sulfonamide 3c, which was completely inactive. These results suggest that the SO 2 group at position 2 of 3b could also be important for anchoring the ligand to the sub-pocket of the binding site.
Keeping m-methylbenzoyl and ciclopropanecarbonyl at position N-2, we modified positions 3 and 4 of the isoxazolone (Table 1). Generally, we observed a collapse of activity. In particular, the 3,4-diphenylisoxazol-5(2H)-one derivatives (compounds 4m and 4n), the 4-methyl-3-phenylisoxazol-5(2H)-one derivatives (compounds 4o and 4p), and the 3-phenylisoxazol-5(2H)-one derivatives (compounds 4q,r and 4u) all had decreased activity (IC 50 ¼ 10-50 mM). The insertion in the para position of the phenyl ring at position 3 of compounds 4q,r with substituents that in the previous series gave good results (such as a nitro group or acetamide function) led to completely inactive (compounds 4s,t). Based on the results in Table 1, we can conclude that  In Table 2, we report the activity of the pair of isomers of type 7 and 8, the amidic and ester derivatives, respectively, which were obtained by introducing a substituent to the para position of the phenyl group at position 4. All N-benzoyl derivatives (7a-e), with the exception of the nitro compound 7c, Table 1. HNE inhibitory activity of isoxazolone derivatives 2, 3a-c, and 4a-u.  50 values are presented as the mean ± SD of three independent experiments. NA: no inhibitory activity was found at the highest concentration of compound tested (50 mM). exhibited good HNE inhibitory activity in the nanomolar range (IC 50 ¼ 20-70 nM). In addition, the activity seems to be independent of the electrophilic properties of the substituents (CH 3 , NO 2 , CN, NHCOR). The best compound of this series was the p-methyl derivative 7a, which had an IC 50 of 20 nM. The O-benzoyl derivatives 8b,d,e, which were derived from tautomerism of the isoxazolone scaffold, surprisingly exhibited activity, although at one order of magnitude lower than the corresponding N-benzoyl type 7 derivatives (Table 2). Only the ester isomer of the potent 7a had a significant loss in activity (8a, IC 50 ¼ 10 mM). The unexpected data related to esters 8a,b,d,e could indicate a different interaction with the target, since the CO endocyclic implicated in the catalysis is missing, and the only carbonyl group presented in the molecule is an ester function at position 5. Further docking studies and kinetic experiments could help us to understand the interaction of these compounds with the HNE catalytic site and their mechanism of action. Table 3 shows the HNE inhibitory activity of benzoisoxazolone derivatives 10a-c, which are elaborations of the isoxazolone scaffold. In this case, only the m-methylbenzoyl derivative 10c had reasonable activity, with an IC 50 of 638 nM, while the other compounds were less active (10a) or inactive (10b).

Molecular modelling
Several attempts were made using different solvents in order to obtain crystals of 7d and 8d. However, it was not possible to obtain crystals suitable for X-ray diffraction (probably in part due to their quite low melting points). As a consequence, the starting 3D geometry of these molecules was obtained by using as building blocks the solid state structures of 3,5-dicyano-4-(4-methoxyphenyl) isoxazole (QAQPON refcode) 68 found in the Cambridge Structural Database (CSD; v 5.37) 69 and that of the isoxazolone derivative 2j (3-ethyl-2-(3-methylbenzoyl)isoxazol-5(2H)-one) 40 . The 3D arrangement of 7d and 8d was then roughly improved by an energy minimisation procedure, followed by molecular dynamics simulations (details in the Supplementary Material).
Preliminary molecular dynamics (MD) simulations at 300 and 600 K were performed on 7d and 8d to evaluate their flexibility, accessible conformational space, and preferred 3D arrangements (low-energy conformations). As expected, the overall shape of both ligands did not change significantly on changing the simulation medium, as roughly determined by comparison of the dihedral angles, which define the overall shape of the molecules (see Figure S1, Supplementary Material). For 7d, MD trajectories (300 and 600 K) showed that s1 and s2 adopted a trans conformation that is maintained throughout the simulations, while s3-s5 access different conformations ( Figures S2 and S3, Supplementary Material). Overall, the molecule adopted an elongated cylindrical shape ( Figure S4, Supplementary Material). In particular, the conformational behaviour of dihedrals s1 and s4 seems to be quite important (vide infra), given that the endocyclic C ¼ O at position 5 appears to be involved in the catalytic process, while the closest C ¼ O group could play an important role in anchoring to the subpocket of the binding site 40 . The superimposition (Figure 2, panel A) of the minimised conformers of 7d extracted from the MD trajectory (T ¼ 600 K, e ¼ 4r), which are comprised within about 2 kcal mol À1 , exemplifies the conformational space accessible to the molecule through rotations about the s3-s5 dihedrals.
Compound 8d appears to be a little bit more flexible compared to 7d, as suggested by the larger variability of the distance separating the centroids of the phenyl rings. While the s1, s2, and s3 dihedrals were almost frozen irrespective of the simulation temperature and medium [they adopted cis, trans (with only few exceptions), and cis (with only few exceptions) conformations, respectively], increasing the temperature from 300 to 600 K made the s5 and s6 dihedrals freely rotate ( Figures S5 and S6, Supplementary Material). Dihedral s4 showed a preference for a gauche arrangement. Overall, 8d is V-shaped: in most cases both of the phenyl rings were rotated with respect to the mean plane defined by the heterocyclic ring ( Figure S7, Supplementary Material). The superimposition (Figure 2, panel B) of the minimised conformers of 8d extracted from MD simulations (T ¼ 600 K, e ¼ 4r) revealed two possible orientations for the carbonyl group of the ester function at position 5, which is supposed to be involved in  the catalytic process (energies are comprised within about 5 kcal mol À1 ).
In both inhibitors, accessibility of the carbonyl group involved in the catalytic process was roughly estimated by the dimension of a sphere centred on the C ¼ O carbon atom (r ¼ 2.5 Å, in red in Figure 3, panel A). Similarly a sphere (r ¼ 2.5 Å, in blue in Figure 3, panel B) centred on the oxygen atom, which could be involved in H-bond interactions with the receptor points, was used to assess the exposure to the environment of C ¼ O in 7d. Panel A of Figure  3 suggests a more crowded region about the site of the nucleophilic attack of 8d, which could account for its lower activity compared to 7d (in 8d, the carbonyl group is directly bound to the phenyl ring and faces the pending arm at position 4). In addition, the almost cylindrical shape of 7d caused, at least in principle, the carbonyl carbon atom to be attacked from both the side of the plane defined by the 5-membered ring ( Figure S8, Supplementary Material). In contrast, only one side of the ester group in 8d appeared to be easily accessible to the hydroxyl group of Ser195 ( Figure S9, Supplementary Material). Finally, in both cases the terminal amide group appeared to be well exposed and thus prone for anchoring to the sub-pocket of the binding site.
Two different approaches were used to gain an idea of the interaction of 7d and 8d with HNE. First, docking of the inhibitors (two different starting conformations were considered) into the active site of HNE using the CDocker protocol (in vacuum, target temperature 300 K, 20 poses retained) and secondly, MD simulations on each inhibitor-HNE complex (in vacuum, simulation time 5 ns, T ¼ 300K, with the atoms outside the binding sphere constrained to fixed points in space in order to save computational time). The structure of HNE complexed with a peptide chloromethyl ketone inhibitor was used for the docking study and MD simulations (1HNE 70 entry of the Protein Data Bank). The binding site of HNE was defined as a sphere with a 12 Å radius centred at the centroid of the five-membered ring of the peptide chloromethyl ketone inhibitor which covered all the active site amino acids of the HNE enzyme. All water molecules and bound inhibitor were removed from the macromolecule and hydrogen atoms were added.
Interaction energies from docking protocols on 7d and 8d with HNE did not significantly differ, and the two inhibitors showed comparable binding modes. In particular for 8d, most of the saved poses featured Gly193 as the anchor group, and the H-bond interaction with this amino acid pushed the site of the nucleophilic attack quite distant from Ser195. However, a few poses show Gly193 involved as an H-bond donor to the ester function of 8d and, in these cases, the distance HO(Ser195) … CO ranged from 3.3 to 4.8 Å (Figure 4, Panel A). The same amino acids are also involved in the interaction with 7d. Anchoring of the terminal NHCO group to Ser195 via H-bonds pushed the CO at position 5 distant from the nucleophilic -OH of Ser195, while H-bonds between 5-CO and Gly193 brought the two partners of the nucleophilic attack a little bit closer (distances range from 4.3 to 5.1 Å, Figure 4, Panel B). However, the mean distance between the inhibitor-HNE reacting sites, as determined from the saved poses, was very long for both complexes (ca 7 Å).
In summary, docking results do not help to explain the different activities of the two inhibitors towards HNE. The fact that the CDocker protocol keeps the receptor rigid, while the inhibitor is allowed to flex during the refinement, could however significantly affect the results, and this is the reason why the ligand-receptor binding affinity was also assessed by MD. On the other hand, MD simulations confirmed the leading role of Gly193 as an anchoring group for both inhibitors. However, in contrast with the docking results, MD shows the 5-CO grouping of 7d was significantly closer to the HNE reactive site (distance less than 4.5 Å in a large fraction of the snapshot conformations) with respect to the corresponding active site of 8d (d < 5 Å in few of the sampled conformations). In contrast, a comparison of the total number of the intermolecular DH … A contacts (distances less than 2.5 Å) shows a greater propensity for 8d to form H-bond interactions compared to 7d. In addition, while multiple intermolecular H-bonds were present for 8d, most of the snapshot conformations of 7d exhibited only one H-bond interaction. In other words, the net of intermolecular Hbonds that involves 8d keeps the molecule quite distant from Ser195. Finally, it is noteworthy that in the HNE complexes with 8d featuring closer distances between the active sites, the heteroatoms of the 5-membered ring acted as H-bond acceptors both towards the Ser195 -OH group and Gly193. As a result of these interactions, the relative 3 D arrangement of 8d in the HNE active site does not appear to be propitious for approach of the nucleophile of Ser195 (see for example Figure 5, Panel A). In contrast, the H-bond between 5-CO and Gly193, which in most cases held together the 7d/HNE adduct, results in a favourable 3 D arrangement of the reacting partners (see for example Figure 5, panel B) for nucleophilic attack.

Stability and kinetic features
The most potent isoxazolones with IC 50 < 100 nM and their ester analogs were further evaluated for chemical stability in aqueous buffer using spectrophotometry to detect compound hydrolysis. The compounds had t 1/2 values from 2.9 to 9.6 h for spontaneous hydrolysis, indicating that the amides were more stable than the esters in the corresponding pairs of 7a/8a, 7d/8d, and 7e/8e (Table 4). In general, the isoxazolones were more stable than our previously described HNE inhibitors with cinnolinone 39 , N-benzoylindazole 36,37 , and N-benzoylpyrazole scaffolds 71 .
The most potent isoxazolones were also selected for evaluation of the reversibility of HNE inhibition over time. As shown in Figure  6, HNE inhibition was rapidly ($30 min) reversed for compounds 7d and 7b. The inhibition was maximal for up to 60 min with compound 7e and >120 min for the other tested compounds (4a, 4b, and 7a). However, inhibition by the compounds was eventually reversed, and full recovery of HNE activity was observed by 4 h after treatment with 8 mM of the compound (e.g. see Figure 6).
To better understand the mechanism of action of these isoxazolone HNE inhibitors, we performed kinetic experiments. As shown in Figure 7, the representative double-reciprocal Lineweaver-Burk plot of fluorogenic substrate hydrolysis by HNE in the absence and presence of compounds 7d and 8d indicates that these compounds are competitive HNE inhibitors.

Conclusions
In the present study, we report a new series of isoxazolones as potent HNE inhibitors, confirming our previous results 40 , which indicated this nucleus as an appropriate scaffold for this target. The most potent compounds had a methyl group at position 3 and a (substituted)phenyl ring at position 4, and the higher HNE inhibitory activity was found for compound 7a (IC 50 ¼ 20 nM). In addition to the 2-NCO derivatives, a number of 5-NCO compounds were obtained, although with lower activity than the corresponding amide. Studies of chemical stability in aqueous buffer indicated that amides were generally more stable than esters, while kinetic experiments confirmed that both amides and esters were competitive HNE inhibitors. Docking and molecular dynamics studies are the different approaches used to understand the interactions of the two isomers 7d and 8d with HNE. Both studies highlight the fundamental role of Gly193 as anchoring group for both the inhibitors, but the MD simulations help us to explain the difference in inhibitory activity between the inhibitors 7d and 8d. The ester isomer 8d shows a greater propensity to form H-bond interactions with respect to 7d and as a result the molecule is quite distant from the Ser195, the amino acid responsible for the nucleophilic attack. While the amide isomer 7b appears more mobile within the active site of HNE, being held in place by single H-bond interactions (vs. multiple in 8d), which leads the C ¼ O at position 5 to obtain a favourable orientation for the nucleophilic attack by Ser195.