Identification of N-Acyl Hydrazones as New Non-Zinc-Binding MMP-13 Inhibitors by Structure-Based Virtual Screening Studies and Chemical Optimization

Matrix metalloproteinase 13 plays a central role in osteoarthritis (OA), as its overexpression induces an excessive breakdown of collagen that results in an imbalance between collagen synthesis and degradation in the joint, leading to progressive articular cartilage degradation. Therefore, MMP-13 has been proposed as a key therapeutic target for OA. Here we have developed a virtual screening workflow aimed at identifying selective non-zinc-binding MMP-13 inhibitors by targeting the deep S1′ pocket of MMP-13. Three ligands were found to inhibit MMP-13 in the µM range, and one of these showed selectivity over other MMPs. A structure-based analysis guided the chemical optimization of the hit compound, leading to the obtaining of a new N-acyl hydrazone-based derivative with improved inhibitory activity and selectivity for the target enzyme.


Introduction
Osteoarthritis (OA) is the most common form of arthritis [1], affecting half of the elderly population (>65 years) [2]. It is characterized by the progressive degradation of articular collagen and can ultimately result in the prosthetic replacement of joints as they become completely dysfunctional. Matrix metalloproteinase 13 (MMP-13), also known as collagenase-3, plays a central role in the pathology as it is the main enzyme responsible for the cleavage of type II collagen in patients with OA [3]. MMP-13 is significantly overexpressed in the joints and articular cartilage in patients with OA and is not present in normal adult cartilage; therefore, it has been proposed as a key therapeutical target for the treatment of OA [4]. MMP-13 belongs to the MMP family, which consists of 23 zinc-dependent enzymes responsible for the degradation of different extracellular matrix (ECM) components [5]. In addition to tissue remodeling, MMPs are involved in the cleavage of many non-matrix targets, such as cell surface receptors, cytokines, chemokines, cell-cell adhesion molecules, clotting factors, and other proteinases [6]. Despite some broad-spectrum MMP inhibitors (MMPIs) having halted the destruction of cartilage in preclinical assays, they have failed clinical trials as patients developed musculoskeletal syndrome (MSS), possibly resulting from the alteration of the physiological functions of different members of the MMP family [7,8]. Therefore, selectivity is currently considered a priority in the development of MMP inhibitors, even if the high homology among MMP catalytic domains makes it a challenging task. So far, in MMPI design, the catalytic zinc ion chelation has been crucial, leading to inhibitors presenting mainly a hydroxamic acid in the development of MMP inhibitors, even if the high homology among MMP catalytic domains makes it a challenging task. So far, in MMPI design, the catalytic zinc ion chelation has been crucial, leading to inhibitors presenting mainly a hydroxamic acid or carboxylic acid as a zinc-binding group (ZBG) [9,10], with nanomolar activity but poor selectivity. In fact, the presence of a ZBG can negatively affect the selective targeting of a specific MMP, cross-interfering with other metalloproteases such as ADAMs (A Disintegrin and Metalloproteinase) [11,12] and ADAMTSs (A Disintegrin and Metalloproteinase with Thrombospondin Motifs) [13], but also with other zinc metalloenzymes such as carbonic anhydrases [14,15] or histone deacetylases [16,17]. In this regard, the characteristics of the catalytic binding site of MMP-13 are slightly different from those of other MMPs, thus providing an edge in the identification of selective inhibitors for this enzyme. More specifically, an adjacent region to the catalytic site, known as the S1′ pocket, is different in MMP-13 as the loop that delimits the pocket (Ω-loop) is longer and shows more flexibility in MMP-13 than in other MMPs [18]. In particular, the Ω-loop of MMP-13 encloses the socalled S1″ specificity pocket. This allows for an opportunity to identify inhibitors with a different binding mode that is not possible for other enzymes in the MMP family. Severa non-zinc chelating inhibitors ( Figure 1) exploited this difference in the MMP-13 binding site to achieve selectivity for this enzyme, as shown by their X-ray crystal structures [19][20][21][22]. In the present paper, we used these crystallographic data to design a virtual screening (VS) methodology able to identify MMP-13 inhibitors that can adopt a similar binding mode and therefore achieve selectivity towards MMP-13. Enzymatic assays of a limited selection of candidate compounds (20) allowed us to find three novel hits that were structurally unrelated to the known MMPIs. The most promising compound underwent a hit optimization study in order to improve its activity and selectivity profile, and a novel series of 12 derivatives was synthesized and tested in vitro.  [19], B [20], C [21] and D [22]) with reported X-ray crystal structures in complex with the MMP-13 catalytic domain used for the VS set-up.

Virtual Screening Studies
Co-crystallized MMP-13 inhibitors that bind to the S1′ pocket without interacting with the catalytic zinc ion show a similar binding mode [19][20][21][22]. All of them present two common characteristics: (a) they contain two aromatic rings or ring systems at both ends of the molecule (with the exception of the co-crystallized inhibitor in the structure with  [19], B [20], C [21], and D [22]) with reported X-ray crystal structures in complex with the MMP-13 catalytic domain used for the VS set-up. First, the compounds were filtered by molecular weight (MW); then, a s alignment was performed to keep only the compounds that could adopt a co similar to that of the co-crystallized ligands; next, protein-ligand docking was on MMP-13; and, finally, compounds were selected based on the interactio binding site of MMP-13 that improved MMP-13 inhibitor activity in previou structure-activity relationship (SAR) studies. The compounds obtained from database were filtered by MW in order to discard compounds too small t posterior protein-ligand docking constraints and compounds too large comp reference ligands used for the subsequent shape-based alignment, therefore r computational time of the subsequent steps. The 300-700 Da range was sel filter, taking into account that the compounds used as references in the s alignment step have MWs between 392 and 491 Da. As a result of this first compounds were filtered out ( Figure 2). Next, a maximum of 10 conform generated for each compound that survived the MW filter. These conform compared to those of selective MMP-13 inhibitors co-crystallized with MMPto the S1′ pocket and do not contain a ZBG. Only conformations similar to crystallized ligands were kept in order to reduce the computational cost o compounds unable to adopt a similar shape to that of the co-crystallized lig likely not fit in the S1′ cavity during the protein-ligand docking step. As a r step, 30,693 compounds were filtered out ( Figure 2). The ligands resulting fro similarity filter were docked onto MMP-13 using the crystal structure with First, the compounds were filtered by molecular weight (MW); then, a shape-based alignment was performed to keep only the compounds that could adopt a conformation similar to that of the co-crystallized ligands; next, protein-ligand docking was performed on MMP-13; and, finally, compounds were selected based on the interactions with the binding site of MMP-13 that improved MMP-13 inhibitor activity in previously reported structure-activity relationship (SAR) studies. The compounds obtained from the Specs database were filtered by MW in order to discard compounds too small to fulfill the posterior protein-ligand docking constraints and compounds too large compared to the reference ligands used for the subsequent shape-based alignment, therefore reducing the computational time of the subsequent steps. The 300-700 Da range was selected as the filter, taking into account that the compounds used as references in the shape-based alignment step have MWs between 392 and 491 Da. As a result of this first step, 83,222 compounds were filtered out ( Figure 2). Next, a maximum of 10 conformations were generated for each compound that survived the MW filter. These conformations were compared to those of selective MMP-13 inhibitors co-crystallized with MMP-13 that bind to the S1 pocket and do not contain a ZBG. Only conformations similar to those of co-crystallized ligands were kept in order to reduce the computational cost of the VS, as compounds unable to adopt a similar shape to that of the co-crystallized ligands would likely not fit in the S1 cavity during the protein-ligand docking step. As a result of this step, 30,693 compounds were filtered out ( Figure 2). The ligands resulting from the shape similarity filter were docked onto MMP-13 using the crystal structure with PDB code 3WV1 [21], as it contains an inhibitor that binds to the S1 pocket. In order to discard ligands unable to adopt a similar binding mode in the S1 pocket to that of previously known selective MMP-13 inhibitors, two positional constraints were defined (one closer to the zinc-binding region and another one deep in the S1 pocket) to be fulfilled by aromatic atoms. Moreover, it was required that the ligand perform a hydrogen bond interaction with Thr245 or Thr247, as all the co-crystallized inhibitors performed at least one of these interactions with the core of the molecule. In order to select the compounds that performed the appropriate interactions in the S1 pocket of MMP-13, several SAR studies were analyzed to obtain information regarding which interactions are important to achieve high activity towards MMP-13. In particular, a potent MMP-13 inhibitor should: (1) have a negatively charged ring substituent that can establish a salt bridge interaction with Lys140, which is unique at the bottom of the S1 side pocket of MMP-13 [21,22,24]; (2) make a π-π interaction with Tyr246 and Phe252 [21]; (3) have a hydrophobic moiety occupying the S1 pocket [21,22]; (4) have an appropriate linker to join the S1 pocket with the S1 side pocket [24]; (5) have a hydrogen or halogen bond acceptor towards Met253 main chain nitrogen [21]; (6) make a hydrophobic interaction with Pro255 [20,24]; (7) have a hydrogen bond acceptor towards the side chain of Thr247 [21,22]; (8) establish hydrophobic contacts in the S1 pocket [24]; (9) have a hydrogen bond acceptor towards Thr245 main chain nitrogen [24]; (10) have a hydrogen bond donor towards Ala238 main chain carbonyl oxygen [24]; and (11) have an appropriate ring substituent in the region close to the zinc-binding group [20,22].
After docking, the top 10 docked poses for each compound were selected based on their docking scores. The interactions they performed with the protein were carefully inspected in order to select the compounds that accomplished the criteria obtained through the analysis of the above-mentioned SAR data. Ideally, in this step, we would like to obtain a compound that meets the 11 criteria. However, this was not the case, as the compound that accomplished more criteria was compound 1 (Figure S1), with a total of eight. Therefore, this compound and compounds that fulfilled most of the criteria were selected, obtaining 20 compounds for in vitro tests ( Figure S1). The structures of these 20 compounds were compared, using their molecular fingerprints, to those of previously reported MMP-13 inhibitors in the Reaxys [25] database. Except for compound 14, which showed a Tanimoto value of 0.71 with a described MMP-13 inhibitor, the Tanimoto similarity values of the selected compounds with any of the previously reported MMP-13 inhibitors analyzed were at most 0.6. Upon visual inspection, compound 14 was not considered structurally similar to the MMP-13 inhibitor obtained from Reaxys (see Figure S2). Moreover, hit compounds that were structurally similar to previously selected hit compounds were discarded to ensure that the final 20 compounds selected for in vitro tests were structurally different from each other (see Figure S3). Therefore, the hit compounds obtained by this virtual screening methodology proved not only to be different from previously reported MMP-13 inhibitors but also structurally diverse.
After the selection of the 20 hit compounds, they were purchased from Specs, and their activity for MMP-13 was analyzed in vitro. The activity data (% inhibition at 100 µM) were determined by a fluorometric assay on human recombinant enzyme, and four compounds were discarded due to very low solubility in DMSO or high fluorescence emission interfering with the assay. ARP100 [26], a hydroxamate-based MMP inhibitor previously developed by our research group, was used in the same assay conditions as the reference compound. To exclude any possible nonspecific inhibition of MMP-13 due to aggregate formation, we performed all the assays in the presence of 0.05% Brij-35, a nonionic detergent similar to Triton X-100, as indicated by Shoichet et al. [27]. Three ligands, out of the twenty tested, presented a % inhibition >40 at 100 µM and were further characterized. All other compounds were not further investigated (see Figure S1 for chemical structures). In Table 1, are displayed the structures, Specs codes, clogP, and MMP-13% inhibition of the selected ligands.

Enzymatic Assays
After the preliminary screening, compounds 11, 12, and 13 (Table 1) displayed the highest MMP-13 inhibitory activity, and their IC50 values were calculated (91 µM, 105 µM, and 14.6 µM, respectively). Next, the inhibitory activities of these three compounds towards MMP-1, MMP-2, MMP-9, and MMP-14 were determined ( Table 2). The selectivity profile of the three ligands pointed out that compound 13, with an N-acyl hydrazone scaffold, presented low micromolar activity against MMP-13 and a promising selectivity over the other tested MMPs. For this reason, 13 was chosen as the hit compound and underwent an optimization study in order to improve its activity and selectivity profile. The Nacyl hydrazone group has been recently introduced in several approved drugs and molecules in clinical trials since it represents a peptide-mimetic subunit endowed with high metabolic stability [29,30].

Enzymatic Assays
After the preliminary screening, compounds 11, 12, and 13 (Table 1) displayed the highest MMP-13 inhibitory activity, and their IC50 values were calculated (91 µM, 105 µM, and 14.6 µM, respectively). Next, the inhibitory activities of these three compounds towards MMP-1, MMP-2, MMP-9, and MMP-14 were determined ( Table 2). The selectivity profile of the three ligands pointed out that compound 13, with an N-acyl hydrazone scaffold, presented low micromolar activity against MMP-13 and a promising selectivity over the other tested MMPs. For this reason, 13 was chosen as the hit compound and underwent an optimization study in order to improve its activity and selectivity profile. The Nacyl hydrazone group has been recently introduced in several approved drugs and molecules in clinical trials since it represents a peptide-mimetic subunit endowed with high metabolic stability [29,30].

Enzymatic Assays
After the preliminary screening, compounds 11, 12, and 13 (Table 1) displayed the highest MMP-13 inhibitory activity, and their IC50 values were calculated (91 µM, 105 µM, and 14.6 µM, respectively). Next, the inhibitory activities of these three compounds towards MMP-1, MMP-2, MMP-9, and MMP-14 were determined ( Table 2). The selectivity profile of the three ligands pointed out that compound 13, with an N-acyl hydrazone scaffold, presented low micromolar activity against MMP-13 and a promising selectivity over the other tested MMPs. For this reason, 13 was chosen as the hit compound and underwent an optimization study in order to improve its activity and selectivity profile. The Nacyl hydrazone group has been recently introduced in several approved drugs and molecules in clinical trials since it represents a peptide-mimetic subunit endowed with high metabolic stability [29,30].

Enzymatic Assays
After the preliminary screening, compounds 11, 12, and 13 (Table 1) displayed the highest MMP-13 inhibitory activity, and their IC 50 values were calculated (91 µM, 105 µM, and 14.6 µM, respectively). Next, the inhibitory activities of these three compounds towards MMP-1, MMP-2, MMP-9, and MMP-14 were determined ( Table 2). The selectivity profile of the three ligands pointed out that compound 13, with an N-acyl hydrazone scaffold, presented low micromolar activity against MMP-13 and a promising selectivity over the other tested MMPs. For this reason, 13 was chosen as the hit compound and underwent an optimization study in order to improve its activity and selectivity profile. The N-acyl hydrazone group has been recently introduced in several approved drugs and molecules in clinical trials since it represents a peptide-mimetic subunit endowed with high metabolic stability [29,30].

Structure-Based Hit Optimization
The docked position of compound 13 ( Figures 3A and S4A) in the MMP-13 binding site shows the characteristic U shape observed in MMP-13 inhibitors that bind to the S1 pocket. The compound contains two amide bonds, one on each side, that establish hydrogen bond interactions with Phe241 and Thr247. These hydrogen bond interactions and the planar character of amide bonds were deemed important for the binding mode of the compound, so the amide moieties were not altered during the optimization. At the S1 pocket, compound 13 establishes a π stacking interaction with Phe252 and a salt bridge interaction with Lys140. Given the strong nature of the electrostatic interaction with Lys140 and the fact that this residue is not present in other MMPs, this region of the molecule should be highly important for both activity and selectivity, and it was further explored in the optimization. At the zinc-binding region, compound 13 establishes a π stacking interaction with His222, but the bromine substituent may cause low solubility. Thus, to increase compound solubility, modifications were proposed in this region as well as to the methyl group present at the center of the molecule. planar character of amide bonds were deemed important for the binding mode of the compound, so the amide moieties were not altered during the optimization. At the S1′ pocket, compound 13 establishes a π stacking interaction with Phe252 and a salt bridge interaction with Lys140. Given the strong nature of the electrostatic interaction with Lys140 and the fact that this residue is not present in other MMPs, this region of the molecule should be highly important for both activity and selectivity, and it was further explored in the optimization. At the zinc-binding region, compound 13 establishes a π stacking interaction with His222, but the bromine substituent may cause low solubility. Thus, to increase compound solubility, modifications were proposed in this region as well as to the methyl group present at the center of the molecule.
On the basis of these considerations, we planned to synthesize new derivatives of 13 in order to explore: (i) the p-substitution on the benzyl ring (R); (ii) the replacement of the methyl on the N-acyl hydrazone moiety with an hydrogen atom (R1); (iii) the elongation of the linker between the central phenyl ring and amido group; and (iv) the substitution on the furan nucleus (R2) (compounds 13a-o, Table 3). The rigid parts of the molecule have been kept to maintain the degrees of freedom and the conformation of the compounds.  On the basis of these considerations, we planned to synthesize new derivatives of 13 in order to explore: (i) the p-substitution on the benzyl ring (R); (ii) the replacement of the methyl on the N-acyl hydrazone moiety with an hydrogen atom (R 1 ); (iii) the elongation of the linker between the central phenyl ring and amido group; and (iv) the substitution on the furan nucleus (R 2 ) (compounds 13a-o, Table 3). The rigid parts of the molecule have been kept to maintain the degrees of freedom and the conformation of the compounds.

Chemistry
New compounds 13a-o were obtained as described in Schemes 1-5. Hit compound 13 was synthesized to be used as a positive control in the enzymatic assays.
The synthesis of N-acyl hydrazones 13, 13a, and 13c is outlined in Scheme 1. The condensation of carboxylic acids 21 and 22 with aromatic amines 23 or 24 was carried out through the intermediate formation of acyl chlorides by treatment with SOCl2 or (COCl)2, to give the corresponding amides 25-27. Hydrazide 29 was obtained by direct treatment of ethyl ester 28 with hydrazine hydrate in equimolar proportion in EtOH. Finally, the target compounds 13, 13a, and 13c were obtained by conjugation of hydrazide 29 with the proper ketones 25 and 26 or aldehyde 27 in EtOH at reflux. The final compounds were easily isolated by crystallization in the reaction solvent.

Chemistry
New compounds 13a-o were obtained as described in Schemes 1-5. Hit compound 13 was synthesized to be used as a positive control in the enzymatic assays.
The synthesis of N-acyl hydrazones 13, 13a, and 13c is outlined in Scheme 1. The condensation of carboxylic acids 21 and 22 with aromatic amines 23 or 24 was carried out through the intermediate formation of acyl chlorides by treatment with SOCl2 or (COCl)2, to give the corresponding amides 25-27. Hydrazide 29 was obtained by direct treatment of ethyl ester 28 with hydrazine hydrate in equimolar proportion in EtOH. Finally, the target compounds 13, 13a, and 13c were obtained by conjugation of hydrazide 29 with the proper ketones 25 and 26 or aldehyde 27 in EtOH at reflux. The final compounds were easily isolated by crystallization in the reaction solvent.

Chemistry
New compounds 13a-o were obtained as described in Schemes 1-5. Hit compound 13 was synthesized to be used as a positive control in the enzymatic assays.
The synthesis of N-acyl hydrazones 13, 13a, and 13c is outlined in Scheme 1. The condensation of carboxylic acids 21 and 22 with aromatic amines 23 or 24 was carried out through the intermediate formation of acyl chlorides by treatment with SOCl2 or (COCl)2, to give the corresponding amides 25-27. Hydrazide 29 was obtained by direct treatment of ethyl ester 28 with hydrazine hydrate in equimolar proportion in EtOH. Finally, the target compounds 13, 13a, and 13c were obtained by conjugation of hydrazide 29 with the proper ketones 25 and 26 or aldehyde 27 in EtOH at reflux. The final compounds were easily isolated by crystallization in the reaction solvent.

Chemistry
New compounds 13a-o were obtained as described in Schemes 1-5. Hit compound 13 was synthesized to be used as a positive control in the enzymatic assays.
The synthesis of N-acyl hydrazones 13, 13a, and 13c is outlined in Scheme 1. The condensation of carboxylic acids 21 and 22 with aromatic amines 23 or 24 was carried out through the intermediate formation of acyl chlorides by treatment with SOCl2 or (COCl)2, to give the corresponding amides 25-27. Hydrazide 29 was obtained by direct treatment of ethyl ester 28 with hydrazine hydrate in equimolar proportion in EtOH. Finally, the target compounds 13, 13a, and 13c were obtained by conjugation of hydrazide 29 with the proper ketones 25 and 26 or aldehyde 27 in EtOH at reflux. The final compounds were easily isolated by crystallization in the reaction solvent.

Chemistry
New compounds 13a-o were obtained as described in Schemes 1-5. Hit compound 13 was synthesized to be used as a positive control in the enzymatic assays.
The synthesis of N-acyl hydrazones 13, 13a, and 13c is outlined in Scheme 1.

Chemistry
New compounds 13a-o were obtained as described in Schemes 1-5. Hit compound 13 was synthesized to be used as a positive control in the enzymatic assays.
The synthesis of N-acyl hydrazones 13, 13a, and 13c is outlined in Scheme 1.

Chemistry
New compounds 13a-o were obtained as described in Schemes 1-5. Hit compound 13 was synthesized to be used as a positive control in the enzymatic assays.
The synthesis of N-acyl hydrazones 13, 13a, and 13c is outlined in Scheme 1. The condensation of carboxylic acids 21 and 22 with aromatic amines 23 or 24 was carried out through the intermediate formation of acyl chlorides by treatment with SOCl 2 or (COCl) 2 , to give the corresponding amides 25-27. Hydrazide 29 was obtained by direct treatment of ethyl ester 28 with hydrazine hydrate in equimolar proportion in EtOH. Finally, the target compounds 13, 13a, and 13c were obtained by conjugation of hydrazide 29 with the proper ketones 25 and 26 or aldehyde 27 in EtOH at reflux. The final compounds were easily isolated by crystallization in the reaction solvent.
The synthesis of N-acyl hydrazones 13b and 13d is reported in Scheme 2. The commercially available furoic alcohol 30 was acetylated by treatment with Ac 2 O/Et 3 N to yield the furoic acetate 31. Analogously to the synthesis described in Scheme 1, the furoic carboxylic acid 31 protected on the alcoholic function was condensed with aromatic amines 23 or 24 by treatment with thionyl chloride, affording amides 32 and 33. The latter were deprotected by treatment with a methanolic ammonia solution to give the corresponding alcohols. The final compounds 13b and 13d, obtained by conjugation of hydrazide 29 with the proper ketone 34 or aldehyde 35 in EtOH, were isolated as pure crystals from the reaction solvent.
The synthesis of N-acyl hydrazones 13e-l is reported in Scheme 3. Dicarboxylic acid 36 was selectively protected as an ethyl ester by controlled treatment with SOCl 2 in EtOH, affording the monoester 37. 5-Methyl-1,2,4 oxadiazole derivative 39 was obtained by cyclodehydration reaction from the known amidoxime 38 [31] and ethyl ester 37. The resulting ethyl esters 37 and 39 were directly transformed into the corresponding hydrazides 40 and 41 by hydrazinolysis. Lastly, the condensation between the hydrazides 40 or 41 with the proper ketone (25 or 34) or aldehyde (27 or 35) in EtOH afforded the N-acyl hydrazones 13h-l and 13e-g purified by crystallization.
The N-acyl hydrazone 13n was prepared following the synthetic path described in Scheme 4. Ethyl ester 43 was obtained from carboxylic acid 42 by esterification with SOCl 2 . Subsequently, 1H-tetrazole derivative 44 was obtained by catalytic cyclization promoted by trimethylstannyl azide from aryl nitrile 43. The target compound 13n was afforded in high yield by condensation between hydrazide 45 and aldehyde 27.
The synthesis of the longest derivatives with n = 1, N-acyl hydrazones 13m and 13o, is shown in Scheme 5. The N-acyl hydrazones 47 and 48 were prepared by condensation between the commercially available aldehyde 46 and the proper hydrazides 40 or 45. The removal of Boc protection was conducted by acid hydrolysis using an excess of trifluoroacetic acid (TFA) to give the pure trifluoroacetate salts of benzylamines 49 and 50. The furoic carboxylic acid 21 was activated as an NHS-ester by reaction with N-hydroxysuccinimide and EDC as coupling agents. The resulting NHS-ester 51 was coupled with amines 49 or 50 to give the corresponding compounds 13m or 13o in high yields. The synthesis of N-acyl hydrazones 13e-l is reported in Scheme 3. Dicarboxylic acid 36 was selectively protected as an ethyl ester by controlled treatment with SOCl2 in EtOH, affording the monoester 37. 5-Methyl-1,2,4 oxadiazole derivative 39 was obtained by cyclodehydration reaction from the known amidoxime 38 [31] and ethyl ester 37.  Usually, the N-acyl hydrazone structure (-C(O)-N-N=C<) is a hybrid between amide and imine functional groups, exhibiting both geometric and conformational stereoisomerism [32]. Rotation along C=N linkage results in the E/Z stereoisomers, where the E conformation is reported to be usually the preferred geometry in solution due to the unfavorable steric restriction of the Z form [33,34]. Moreover, the rotation across the amide C(O)-NH bond gives geometrical isomerism as synperiplanar (sp) and antiperiplanar (ap) conformers [35]. For these reasons, all the N-acyl hydrazones here reported (13 and 13a-o) are described in their E form and as a mixture of synperiplanar (sp) and antiperiplanar (ap) conformers. As already reported by Munir et al. [35], in all the 1 H NMR spectra of the synthesized N-acyl hydrazones (13 and 13a-o; see Supporting Information and Experimental section), a particular pattern of two signal sets was detected, corresponding to the mixture of rotational syn-and antiperiplanar conformers. In fact, in the 1 H NMR spectra resolved in DMSO-d 6 , duplicated signals were observed for -CONHN-, -CH 2 -, and -N=C(CH 3 ) protons. In Figure 4, the 1 H NMR spectra of two representative N-acyl hydrazones, 13l and 13a, are reported. In both spectra, the chemical shift of the -CONHN-proton (in green) resonated around 10.5-12 ppm, resulting in a double singlet, indicating the syn and the ap forms. Similar behavior is shown by benzylic -CH 2 -(in yellow) and methyl imine (-CH 3 ) (in blue) protons, which resonate, respectively, at 4.2-3.8 ppm and 2.3-2.25 ppm. For -C(O)NHNdoublet-like peaks, the downfield signal was attributed to the antiperiplanar (ap) conformer, whereas the upfield signals were attributed to the synperiplanar (sp) conformer, as established from the literature [36][37][38]. On the contrary, regarding the methylene (-CH 2 ) proton, the upfield signal is the anti-form and the downfield signal is the syn-isomer, as evidently demonstrated by the integration of the peaks. The ratio between two conformers was calculated considering the integral intensities of the paired peaks, revealing that a DMSO-d 6 solution of these compounds contained E-synperiplanar and E-antiperiplanar conformers in an approximate 2:1 ratio with slight variation from compound to compound. The N-acyl hydrazone 13n was prepared following the synthetic path described in Scheme 4. Ethyl ester 43 was obtained from carboxylic acid 42 by esterification with SOCl2. Subsequently, 1H-tetrazole derivative 44 was obtained by catalytic cyclization promoted by trimethylstannyl azide from aryl nitrile 43. The target compound 13n was afforded in high yield by condensation between hydrazide 45 and aldehyde 27.  Usually, the N-acyl hydrazone structure (-C(O)-N-N=C<) is a hybrid between amide and imine functional groups, exhibiting both geometric and conformational stereoisomerism [32]. Rotation along C=N linkage results in the E/Z stereoisomers, where the E conformation is reported to be usually the preferred geometry in solution due to the unfavorable steric restriction of the Z form [33,34]. Moreover, the rotation across the amide C(O)-NH bond gives geometrical isomerism as synperiplanar (sp) and antiperiplanar (ap) con- nar (sp) conformer, as established from the literature [36][37][38]. On the contrary, regarding the methylene (-CH2) proton, the upfield signal is the anti-form and the downfield signal is the syn-isomer, as evidently demonstrated by the integration of the peaks. The ratio between two conformers was calculated considering the integral intensities of the paired peaks, revealing that a DMSO-d6 solution of these compounds contained E-synperiplanar and E-antiperiplanar conformers in an approximate 2:1 ratio with slight variation from compound to compound.

SAR Analysis
All new derivatives 13a-o were tested by a fluorometric assay on recombinant MMP-13 in comparison with hit compound 13, and their inhibitory activity is reported in Table 3. The replacement of the bromine substituent on the furan ring (R 2 ) with a sulfonamido or an alcohol group caused a drop in activity, with derivatives 13a and b displaying an IC 50 > 100 µM.
In general, all derivatives bearing a hydrogen atom in R 1 showed improved activity with respect to their N-methyl analogues, as can be seen by comparing compound 13d with compound 13b and compound 13g with compound 13f.
The introduction of a 1,2,4-oxadiazole ring in R (as in compounds 13e and f) caused a decrease in inhibitory activity for MMP-13 relative to 13. This heterocycle, known as an ester isostere, is present in many biologically active compounds [39] and was chosen in an attempt to ameliorate the solubility of our original N-acyl hydrazone scaffold. Actually, the simultaneous introduction of the oxadiazole ring in R with an alcohol group in R 2 and a hydrogen in R 1 caused an improvement in solubility, as can be seen from the logP value calculated for compound 13g with respect to 13 (Table 4), but did not increase the inhibitory activity. On the contrary, the replacement of the nitro group in R with an acidic moiety, such as a carboxylic acid in 13l or a 1H-tetrazole in 13n, led to improved inhibitory activity relative to 13c. The incorporation of these acidic moieties introduced more negative electrostatic surfaces close to the residue Lys140, which resulted in better electrostatic complementarities between these compounds and the enzyme at the S1 pocket, thus increasing their activities ( Figures 3C and S4C).
On the basis of these results and in order to increase its inhibitory potency, we decided to further modify 13l by introducing a methylene spacer between the central phenyl ring and amido group (n = 1). In fact, this modification could allow our hit compound to better interact with Lys140 in the S1 pocket by establishing the salt bridge necessary for a proper fit with the MMP-13 binding site (Figures 3B and S4B). As expected, compound 13m resulted in the best of the series, showing an 8-fold increase in inhibitory potency against MMP-13 (IC 50 = 1.8 µM) relative to 13.
Of note, the same modification in the tetrazole analogue 13o caused a drop in activity, probably due to the excessive bulkiness of this substituent, negatively affecting the binding with the enzyme. 1H-tetrazole is an aromatic heterocyclic bioisoster of the carboxylic acid group, characterized by a higher lipophilicity and a similar acidity but a different volume [40].
In general, derivatives 13a-o show good predicted ADMET properties (see Table 4). Thus, all of them fulfill the Lipinski rule of five: they have good intestinal absorption and show week/moderate cardiotoxicity. Only high hepatotoxicity is predicted for all of them, and this is necessary to be experimentally tested in future studies and before the next rounds of optimization of 13m (or other) derivatives.
Finally, the selectivity profile of 13m was evaluated to verify if the optimization process has led to a loss of selectivity for MMP-13. The results reported in Table 5 show that, with the exception of MMP-2, the selectivity over the other tested enzymes has been maintained or improved with respect to our hit compound 13. In particular, 13m displayed a 100-fold selectivity for MMP-13 over MMP-1 and MMP-14.

Chemistry
Melting points were determined on a Leica Galen III Microscope (Leica/Cambridge Instruments, Cambridge, UK) and are uncorrected. 1 H and 13 C NMR spectra were recorded on a Bruker Avance III HD 400 MHz spectrometer (Fällander, Switzerland). Chemical shifts (δ) are reported in parts per million, and coupling constants (J) are reported in hertz (Hz). 13 C NMR spectra were fully decoupled. The following abbreviations were used to explain multiplicities: singlet (s), doublet (d), triplet (t), double doublet (dd), broad (br), and multiplet (m). Chromatographic separations were performed on silica gel columns by flash column chromatography (Kieselgel 40, 0.040−0.063 mm, Merck, Darmstad, Germany) or using ISOLUTE Flash Si II cartridges (Biotage) or using an Isolera automatic system (Biotage, Uppsala, Sweden) and SFÄR HC Duo silica cartridges (Biotage, Uppsala, Sweden). Reactions were followed by thin-layer chromatography (TLC) on Merck aluminum silica gel (60 F254) sheets that were visualized under a UV lamp. Evaporation was performed in vacuo (rotating evaporator). Sodium sulfate was always used as the drying agent. Commercially available chemicals were purchased from Merck (Darmstad, Germany). Elemental analysis was used to determine the purity of the target compounds (Table S1). Analytical results are within ±0.4% of the theoretical values. High-resolution ESI-MS spectra were recorded by direct injection at 5 (positive) and 7 (negative) µL min −1 flow rates in an Orbitrap high-resolution mass spectrometer (Thermo, San Jose, CA, USA), equipped with a HESI source.

General Procedure to Synthesize Conjugates 25-27 and 32-33
To a solution of furoic acids 21, 22, and 31 (1 eq) in dioxane (1 mL/mmol) or a mixture of DCM and a few drops of DMF, SOCl 2 (4 eq) was added under a nitrogen atmosphere. The reaction mixture was stirred at 100 • C for 12 h and then evaporated under an inert atmosphere (N 2 ). The crude product was dissolved in dioxane, or DCM (1.10 mL/mmol), and pyridine (1 eq) and the commercial aryl amines 23-24 (1 eq) were added under a nitrogen atmosphere. The reaction was stirred at RT for 1 h and then diluted with EtOAc. The organic phase was washed with water, NaHCO 3 saturated solution, and HCl 1N. The organic layer was dried over Na 2 SO 4 , filtered, and evaporated under reduced pressure. The pure product (25-27 and 33) was afforded either without any further purification or by column chromatography (32).
The title compound was synthesized as previously reported in the general procedure, starting from commercial furoic acid 21 (500 mg, 2.62 mmol) dissolved in dioxane and using commercial 4-aminoacetophenone 23 (354 mg). After workup, compound 25 was obtained as a yellow/orange solid (565 mg) without any further purification. Yield: 73%. 1    Compound 37 (429 mg, 2.06 mmol, 1 eq) was dissolved in DCM (2.75 mL) and added to SOCl 2 (0.45 mL, 4.12 mmol, 3 eq) under a nitrogen atmosphere. The reaction mixture was stirred at RT for 4 h and concentrated under a nitrogen flux. Then the resulting crude was diluted with acetone and added dropwise to a solution of amidoxime 38 (167.7 mg, 2.27 mmol, 1.1 eq) and K 2 CO 3 (569 mg, 4.12 mmol, 2 eq) in acetone (3 mL). The resulting mixture was stirred at RT overnight under a nitrogen atmosphere, and then the solvents were removed under vacuum conditions. In order to remove any remaining salt, the crude was treated with H 2 O, and the solid was filtrated under vacuum conditions and subsequentially warmed up to 130 • C solvent-free for 3 h. Commercially available 2-(4-cyanophenyl)acetic acid (42) (100 mg, 0.62 mmol, 1 eq) was first dissolved in EtOH (0.6 mL) and then added to SOCl 2 (0.06 mL, 0.81 mmol, 1.3 eq). The reaction was stirred at 65 • C for 2 h and subsequentially evaporated. Purification of the crude was achieved by trituration with n-hexane, obtaining pure 43 as a white solid (113.3 mg). Yield: 97%. 1  To a solution of compound 43 (113 mg, 0.6 mmol, 1 eq) in THF dry (1 mL), azido trimethyltin (IV) (124.71 mg, 0.6 mmol, 1 eq) was added under a nitrogen atmosphere. The reaction mixture was maintained at 70 • C for 3 days. After that, the solution was diluted with HCl 1N (25 mL) and extracted with EtOAc (3 × 25 mL). The organic phases were then dried over Na 2 SO 4 , filtrated, and evaporated under vacuum conditions. The resulting crude was purified by flash chromatography using an ISOLUTE Si II 5 g cartridge (n-hexane/EtOAc in a gradient from 40:1 to 1:1 v/v) to give the desired product 44 as a white solid (96.1 mg). Yield: 69%. 1  Hydrazine hydrate (10 eq) was added to a solution of ethyl esters 28, 37, 39, and 44 (1 eq) in EtOH (3 mL/mmol). The resulting mixture was stirred at 80 • C for 4-6 h under a nitrogen atmosphere and evaporated. The crude purification was conducted in different ways, depending on the substrate.

General Procedure to Synthesize Final Compounds 13m,o
To a solution of bromofuranosyl derivative 51 (1 eq) in DMF dry (4.2 mL/mmol), DIPEA (2 eq) was added. Lastly, a solution of trifluoroacetate salt 49 or 50 (1 eq) in DMF dry (8.4 mL/mmol) was added dropwise to the reacting mixture. The reaction was stirred for 2 h and then diluted with EtOAc.
The title compound was synthesized as reported in the general procedure, starting with 51 (48 mg, 0.167 mmol) and trifluoroacetate 50 (75 mg, 0.167 mmol). After dilution with EtOAc (50 mL), the organic phase was washed with water (2 × 50 mL). The water phase was acidified using HCl 37% and extracted with EtOAc (2 × 50 mL). The organic layers were dried over Na 2 SO 4 , filtered, and evaporated under vacuum conditions. The pure final compound 13o was obtained as a pinkish powder (35 mg

Shape-Based Similarity
Conformations were generated using Omega [43,44] with default parameters and requiring a maximum of 10 conformations. The co-crystallized inhibitors used as references in the shape comparison corresponded to the ligands of the crystal structures with the following PDB codes: 2OW9 [19], 2OZR [19], 3KEC [20], 3KEJ [20], 3KEK [20], 3WV1 [21], and 5BPA [22]. Shape similarity between the library compounds and the reference compounds was calculated with ROCS [45,46] using the ShapeTanimoto coefficient, a value between 0 and 1 calculated by the following equation: ShapeTanimoto f,g = O f,g /(I f + I g − O f,g ) in which the I terms are the self-volume overlaps of each molecule, while the O term is the overlap between the two functions [47].
Conformations with a ShapeTanimoto value lower than 0.5 for any of the reference compounds were discarded.

Ligand Setup for Docking
Before docking, ligand molecules were prepared with LigPrep [48] with default parameter values except for the following options: (a) respect chiralities from input geometry when generating stereoisomers; (b) use Epik [49] for ionization and tautomerization; (c) use 7.0 as an effective pH; and (d) use 2.0 as a pH tolerance for generated structures.

Protein Preparation
After verifying the fitting of the coordinates of the residues in the binding site relative to their corresponding electron density map with VHELIBS, the B chain of the crystal structure with PDB code 3WV1 was prepared by using Maestro's Protein Preparation Wizard [50] through the following procedure: (a) align to 1ROS, chain A; (b) remove original hydrogens; (c) cap termini; (d) generate ionization and tautomeric states of the ligand with Epik; (e) assign hydrogen bonds at pH 7 with PROPKA; (f) use force field OPLS_2005 to minimize the structure at 0.30 Å; and (g) remove all water molecules from the structure.

Grid Preparation
The grid for protein-ligand docking was generated with Maestro [51] by using default parameter values except for the following settings: (a) the grid center coordinates were (46.0, 80.0, and −1.0); (b) halogens were included as acceptors; (c) the inner box size was (10, 10, and 10); (d) the outer box size was (30, 30, and 30); (e) hydrogen bond constraints were defined on the backbone nitrogen of the residues Thr245 and Thr247 as well as the sidechain oxygen of the residue Thr245; and (f) two positional constraints with a radius of 2 Å were defined on the coordinates (46.3, 80.1, and −7.5) and (51.2, 80.5, and 4.6), respectively.

Molecular Docking
Protein-ligand docking was performed with Glide [52] by using default parameter values except for the following settings: (a) SP precision; (b) enhance planarity of conjugated π groups; (c) include halogens as acceptors; (d) write out at most 10 poses per ligand; (e) include 50 poses per ligand in post-docking minimization; (g) require the accomplishment of both positional constraints by aromatic atoms; and (f) require the accomplishment of one hydrogen bond constraint.
The first picture in each panel of Figure 3 was obtained with Maestro [51], and the second and third pictures were obtained with Flare [28]. The docked poses were predicted with GlideXP [52].

Conclusions
In order to obtain potent and selective MMP-13 inhibitors devoid of a ZBG, we developed a virtual screening workflow aimed at identifying compounds that target the S1 pocket of MMP-13, a region in the MMP binding site that has been shown to be different for MMP-13 with respect to other MMPs. For this purpose, we first applied a MW filter to discard compounds unlikely to survive subsequent filters. Next, we used a shape-based similarity analysis to restrict the initial library of compounds to those able to adopt the characteristic U shape adopted by co-crystallized non-zinc-binding selective MMP-13 inhibitors. Then, we performed protein-ligand docking simulations to predict the binding modes of these compounds. Finally, we analyzed previously reported SAR studies to identify MMP-13 inhibitor interactions with the protein that are important for activity, and we selected the docked poses obtained in the protein-ligand docking according to these criteria. The bioactivity assays on isolated enzymes identified three putative hit compounds capable of inhibiting MMP-13 in the µM range, one of which displayed at least 4-fold selectivity over MMP-1, MMP-2, MMP-9, and MMP-14. Then, a structure-based optimization of the N-acyl hydrazone hit compound 13 guided the synthesis of a series of 12 new derivatives. Among these, a carboxylate derivative (13m) was found to selectively inhibit MMP-13 with an IC 50 = 1.8 µM. A docking study showed that the presence of an acidic moiety in R introduced more negative electrostatic surfaces close to the residue Lys140, which resulted in better electrostatic interaction between this compound and the enzyme at the S1 pocket, thus increasing its activity. The next round of optimization from hit to lead will be necessary to further develop this interesting class of new compounds.