In Silico Identification and Experimental Validation of Novel KPC-2 β-lactamase Inhibitors

Bacterial resistance has become a worldwide concern, particularly after the emergence of resistant strains overproducing carbapenemases. Among these, the KPC-2 carbapenemase represents a significant clinical challenge, being characterized by a broad substrate spectrum that includes aminothiazoleoxime and cephalosporins such as cefotaxime. Moreover, strains harboring KPC-type β-lactamases are often reported as resistant to available β-lactamase inhibitors (clavulanic acid, tazobactam and sulbactam). Therefore, the identification of novel non β-lactam KPC-2 inhibitors is strongly necessary to maintain treatment options. This study explored novel, non-covalent inhibitors active against KPC-2, as putative hit candidates. We performed a structure-based in silico screening of commercially available compounds for non-β-lactam KPC-2 inhibitors. Thirty-two commercially available high-scoring, fragment-like hits were selected for in vitro validation and their activity and mechanism of action vs the target was experimentally evaluated using recombinant KPC-2. N-(3-(1H-tetrazol-5-yl)phenyl)-3-fluorobenzamide (11a), in light of its ligand efficiency (LE = 0.28 kcal/mol/non-hydrogen atom) and chemistry, was selected as hit to be directed to chemical optimization to improve potency vs the enzyme and explore structural requirement for inhibition in KPC-2 binding site. Further, the compounds were evaluated against clinical strains overexpressing KPC-2 and the most promising compound reduced the MIC of the β-lactam antibiotic meropenem by four fold.


Introduction 39
The emergence of KPC-2 class-A Beta-Lactamase (BL) carbapenemase, which confers resistance to last 40 resort carbapenems, poses a serious health threat to the public. KPC-2, a class A BL, uses a catalytic 41 serine to hydrolyze the β-lactam ring. Specifically, the hydrolysis reaction proceeds through a series of 42 steps involving: (i) the formation of a precovalent complex, (ii) the conversion to a high-energy 43 tetrahedral acylation intermediate, (iii) followed by a low-energy acyl-enzyme complex, (iv) a high-44 energy tetrahedral de-acylation intermediate consequent to catalytic water attack, and (v) finally the 45 release of the hydrolyzed β-lactam ring product from the enzyme. [1][2][3][4][5][6]. 46 Notably to treat infections caused by bacteria that produce class A BLs, mechanism-based inhibitors 47 (i.e., clavulanic acid, sulbactam, and tazobactam) are administered in combination with β-lactam 48 antibiotics. However, strains harboring KPC-type β-lactamases are reported to be resistant to available 49 β-lactamase inhibitors. Moreover, because of KPC-2's broad spectrum of activity (which includes 50 penicillins, cephalosporins, and carbapenems) treatment options against KPC-2-producing bacteria are 51 scarce, and "last-resort" carbapenems are ineffective as well [7]. Therefore, studies directed to the 52 discovery of novel, non β-lactam KPC-2 inhibitors have multiplied in the last years. Recently, new drugs 53 able to restore susceptibility to β-lactams i.e. the novel inhibitor avibactam in combination with 54 ceftazidime (CAZ) and RPX7009 (vaborbactam) with meropenem have been approved ( As attention on KPC-2 rises, the number of crystal structures of its apo and complexed form disclosed 56 in the PDB has increased, making KPC-2 a druggable target for structure based drug design efforts and 57 for the study of novel, non β-lactam like inhibitors of this threatening carbapenemase [9-12] 58 Recently, two crystal structures of the hydrolyzed β-lactam antibiotics cefotaxime and faropenem in 59 complex with KPC-2 were determined (PDB codes 5UJ3, 5UJ4; Fig. 2).
[13] 60 Left: binding mode of hydrolyzed cefotaxime (PDB code 5UJ3). Right: binding mode of hydrolyzed faropenem (PDB code 5UJ4). The second rotamer of Trp105 adopted in the apo-enzyme is coloured in beige, protein side chains in blue and ligands in green. Hydrogen bonds are indicated as black dots.

9
Reactions were monitored using a Beckmann DU640® spectrophotometer at 405nM for CENTA and 143 480 nM wavelength for nitrocefin [35]. The test compounds were synthesized as described below or 144 purchased from Enamine, TimTec, Vitas-M, ChemBridge, Otava, Life Chemicals or Apollo Scientific 145 and assayed without further purification. Compounds were dissolved in dimethyl sulfoxide (DMSO) to 146 a concentration of 25 mM and stored at -20°C. The highest concentration at which the compounds were 147 tested was up to 1 mM (depending on their solubility). All experiments were performed in duplicate and 148 the error never exceeded 5%. The reaction was typically initiated by adding KPC-2 to the reaction buffer 149 last. To control for incubation effects, protein was added to the reaction buffer first, and the reaction 150 was initiated by the addition of reporter substrate after 10 minutes of enzyme-compound incubation. 151 The results are reported in Tables 2 and Table 3. 152 Competitive inhibition mechanism and the Ki for compound 9a was determined by Lineweaver-Burk 153 (LB) and Dixon plots. For compound 11a, already reported as competitive inhibitor of the extended 154 General procedure for the synthesis of sulfonamides 1-6b 171 To a solution of 3-(1H-tetrazol-5-yl)aniline (1 eq.) in DCM dry (25 mL) at room temperature and under 172 nitrogen atmosphere, pyridine (3 eq.) and the appropriate sulfonyl-chloride (1.2 eq.) were added. The 173 mixture was reacted at room temperature for 2-12 h. The reaction was quenched with aqueous satured 174 solution of NH4Cl and acidified at pH 4 with aqueous 1N HCl. The aqueous phase was extracted with 175 AcOEt, and the organic phase washed with brine, dried over Na2SO4 and concentrated. The crude was 176 crystalized from MeOH or Et2O to give the desired product.

224
Only little diversity with respect to bound ligands was found in the KpKPC2 structures. To obtain a 225 more detailed picture on key interactions and to derive a pharmacophore hypothesis, PoSSuM -Search 226 K was used to search for similar binding sites containing non-covalent ligands. This resulted in thirteen 227 structures (Table 1), all having tetrazoles or carboxylates derivatives bound in the hydrophilic pocket 228 formed by the amino acids corresponding to Thr235, Thr237, Ser130 and Ser70 in KpKPC2 (Fig.4).

238
Based on the retrieved structures, a pharmacophore hypothesis was derived. All of the ligands in these 239 structures as well as the β-lactamase binding protein (PDB code 3E2L, 3E2K) and the covalent ligand 240 of the structure used as receptor, formed a hydrogen-bond with Thr235 or Thr237. Accordingly, a 241 hydrogen-bond acceptor at the corresponding ligand position was considered to be crucial for binding 242 (Fig. 3). Further, in most of the structures the ligands formed interactions with Trp105 (Ambler 243 numbering) [22]. Therefore, this interaction was also included in the pharmacophore hypothesis. 244 Hydrogen-bond interactions to Asn130 were found in four structures (PDB codes 3RXW, 3G32, 3G30, 245 4EUZ) and included as well. 246 247 A hierarchical approach was adopted for virtual screening. First, our in-house database of around five 248 million purchasable compounds was filtered for lead-like molecules [23]. In the second step, the 249 obtained hits were screened with the above-described pharmacophore resulting in 44658 compounds. 250 Out of these, 31122 compounds could be docked into the Kp KPC2 binding site. Filtering these binding 251 14 poses again with the pharmacophore resulted in 2894 compounds. These were divided into three 252 clusters, depending on the functional group placed in the hydrophilic pocket (tetrazoles, carboxylates, 253 sulfonamides) and inspected by eye. Finally, 31 compounds were selected for hit validation (Table 2). 254 Most of the selected chemotypes carried an anionic group, mainly a carboxylic group or its bioisostere,

263
The majority of the selected candidates were fragment-like as defined by the "rule of three" [14]. Thus, 264 potencies in the high micromolar to millimolar range were expected. Unfortunately, the required high 265 concentrations for ligand testing could not always be achieved due to solubility issues which might have 266 resulted false negatives after testing. However, some of the tested molecules inhibited the hydrolytic 267 activity of KPC-2 with millimolar potency. Among those, compounds 9a and 11a were the most 268 promising compounds with micromolar affinities (IC50 of 0.15 and 0.036 mM, translating to ligand 269 efficiencies (LE) of 0.38 and 0.28 kcal/mol/non-hydrogen atom, respectively; Table 2) and were thus 270 further investigated. 271 Compound 9a was predicted to place its carboxylate group in proximity of the catalytic Ser70, in the 272 carboxylic acid binding site mentioned above, forming hydrogen bond interactions with the side chains 273 corresponding to amino acids Ser130, Thr235 and Thr237 (Fig. 6). Thr 237 in KPC-2 is known to be 274 the presence of the sulphur atom of the benzothiazole system seems critical for affinity as the related 286 compound 19a, the benzimidazole analog, resulted 6-fold less active. Similar, the presence of the 287 carboxylic group appeared to be crucial as compound 32a, without such a functionality, was 8-fold less 288 active. 289

290
For compound 9a binding affinity and mode of inhibition was determined by using gradient 291 concentrations of CENTA. Fitting of the obtained data showed that compound 9a behaves as a 292 competitive inhibitor with a determined Ki of 112.0 µM (Fig. 7). Its binding affinity was also determined 293 towards other class A β-lactamases (IC50 vs CTX-M9 160 µM). For this compound aggregating behavior 294 was also excluded by dynamic light scattering experiment (data not shown) [40]. Compound 9a with its 295 fragment-like characteristic (MW 208.21, determined Ki 112.0 µM, LE 0.38 kcal/mol/non-hydrogen 296 atom) exerts an interesting activity vs KPC2-2 and represents a very promising molecule to be directed 297 to hit to lead optimization. bioisostere of the carboxylic group, was predicted to lie in the hydrophilic pocket formed by Thr235, 303 Thr237, Ser130 and Ser70, driving the binding of the inhibitor in KPC-2 active site (Fig. 8). The phenyl 304 ring attached to the tetrazole was predicted to be sandwiched between the Trp105 side with a distance 305 compatible with weak hydrophobic interactions and the backbone of Thr237. The amide group of 11a 306 was oriented in the canonical site delimited by Asn132, Asn170 and in a further distance Glu166 where 307 the R1 amide side chain of β-lactams is known to bind. However, the amine linker and the second phenyl 308 ring in 11a were not predicted to form any specific interactions with the protein, except for the amide 309 nitrogen contacting the backbone of Thr237. The distal fluoro-benzene ring was oriented at the entrance 310 of the active site against two hydrophobic patches, one defined by Leu167, closer, and the other by the 311 backbone of Asn170, a residue critical for carbapenemase activity. 312

313
Based on the predicted binding mode, the tetrazole group of 11a seemed to be crucial for affinity. (Table  314 1). Moreover, while the proximal ring appeared to be involved in specific interactions, the amide group 315 and the distal ring did not contact efficaciously the protein. Based on predicted binding mode, chemical 316 size, synthetic accessibility for a rapid structural optimization and ligand efficacy compound 11a was 317 directed to chemical synthesis development to improve its affinity and to investigate target binding 318 requirements for optimal inhibitor-enzyme interaction. 319

Hit derivatization and evaluation 320
In order to improve the binding affinity of 11a, the compound was subjected to a hit optimization 321 program. Therefore, the phenyl-tetrazole moiety, that seemed to strongly drive the binding, was retained 322 unaltered, whereas structural modifications on the linker and on the distal aromatic ring were introduced 323 in order to explore and maximize the interactions with the pocket formed by Asn132, Asn170 and 324 Leu167 (Fig. 6). Because the amide linker does not contact efficaciously the protein we chose to replace 325 it with a sulfonamide (Fig. 9). We meant to target residues proximal to the opening of the active site 326 while investigating the potentiality for sulfonamide derivatives.

329
Sulfonamides are more stable towards hydrolysis than carboxyamides, possess an additional hydrogen 330 bonding oxygen atom and their NH is a strong hydrogen bond donor. In addition, the dihedral angle '' 331 OSNH measures around 90° compared with the 180° '' OCNH angle of amide. Sulfonamides, in 332 addition, have a non-planar configuration that could orient the distal ring towards Leu167 and Asn170 333 (Fig. 10). Therefore, the introduction of a sp 3 geometry could allow a more efficacious spanning of the  Substituents with different electronic and steric properties (i.e. halogens, nitro, sulfonamide, carboxylic 341 acid, methyl, acetamide, amino groups) were inserted in the different position of the aromatic ring. In 342 addition, the benzene ring was replaced by heterocyclic or extended benzofused systems such as 343 benzimidazole, quinazolinone, naphthalene, or quinolone ring. Based on the availability of compound 344 or building blocks, 6 compounds (1b-6b) were synthesized and 8 compounds (7b-14b) were purchased 345 to test our hypothesis (Table 3). The fourteen new compounds were tested in vivo vs clinical strains 346 overproducing KPC-2 to evaluate their ability to restore bacteria susceptibility to carbapenem 347 meropenem (Table 4).  Compounds 1b-6b were synthesized in high yield (75-95% yield) and purity (>95%) through direct 357 reaction of 3-(1H-tetrazol-5-yl) aniline and the appropriate sulfonyl chloride in dichloromethane at room 358 temperature for 3 hours (Fig. 11). 359 360 Figure 11. Reagents and conditions. a) aryl-sulfonyl chloride (1.2 eq.), pyridine (3 eq.), dry DCM, N2, r.t, 3 h, 75-95% yield.