Synthesis of Computationally Designed 2,5(6)-Benzimidazole Derivatives via Pd-Catalyzed Reactions for Potential E. coli DNA Gyrase B Inhibition

A pharmacophore model for inhibitors of Escherichia coli’s DNA Gyrase B was developed, using computer-aided drug design. Subsequently, docking studies showed that 2,5(6)-substituted benzimidazole derivatives are promising molecules, as they possess key hydrogen bond donor/acceptor groups for an efficient interaction with this bacterial target. Furthermore, 5(6)-bromo-2-(2-nitrophenyl)-1H-benzimidazole, selected as a core molecule, was prepared on a multi-gram scale through condensation of 4-bromo-1,2-diaminobenzene with 2-nitrobenzaldehyde using a sustainable approach. The challenging functionalization of the 5(6)-position was carried out via palladium-catalyzed Suzuki–Miyaura and Buchwald-Hartwig amination cross-coupling reactions between N-protected-5-bromo-2-nitrophenyl-benzimidazole and aryl boronic acids or sulfonylanilines, with yields up to 81%. The final designed molecules (2-(aminophen-2-yl)-5(6)-substituted-1H-benzimidazoles), which encompass the appropriate functional groups in the 5(6)-position according to the pharmacophore model, were obtained in yields up to 91% after acid-mediated N-boc deprotection followed by Pd-catalyzed hydrogenation. These groups are predicted to favor interactions with DNA gyrase B residues Asn46, Asp73, and Asp173, aiming to promote an inhibitory effect.


Introduction
Imidazole-based heterocycles are a common motif found in both natural and synthetic compounds, present in several clinically approved drugs [1][2][3]. In particular, the benzimidazole moiety is a relevant scaffold for multiple applications in clinic, showing antiulcer, antihypertensive, antiparasitic, anticancer, antifungal, and antibacterial activities [4,5]. For instance, ridinilazole, a recently developed antibacterial containing the benzimidazole scaffold, is currently in phase III clinical trials and has shown great promise for treating Clostridioides difficile infections [6].
Indeed, due to extensive and widespread bacterial resistance to current therapeutics [7] there is an urgent need to develop more efficient synthetic processes to obtain potential new antibiotics derived from a computer-aided rational design. Aiming for the development of inhibitors for the bacterial target Escherichia coli's DNA Gyrase B [3,[8][9][10], we have used a pharmacophore model created in the Molecular Operating Environment (MOE) molecular design software (Chemical Computing Group) [11] to provide insights into the ideal structure of potential antibacterial molecules. Following the analysis of the computational pharmacophore model herein described, we have planned the synthesis of families of potential antibacterial molecules derived from the 1H-benzimidazole scaffold ( Figure 1). This core was selected as a starting point for our studies, owing to its recently reported antibacterial activity [6] and its synthetic amenability [12]. We hypothesized that the synthesis of 5(6)-halogenated-2-substituted benzimidazole by classic methods [13][14][15] involving a condensation reaction of a halogenated orthophenylenediamine with 2-nitrocarbonyl/carboxylphenyl would allow the preparation of a core synthon for further derivatization to match the features predicted by the pharmacophoric model.
The limitations imposed by classic multi-step methods [16,17] on the functionalization of 5(6) position of benzimidazoles have driven our efforts towards the use of palladiumcatalyzed coupling reactions, namely Suzuki-Miyaura and Buchwald-Hartwig. Current literature on this topic shows multiple reports of palladium-catalyzed reactions involving the oxidative addition to activated 2-halogen substituted benzimidazoles [18][19][20][21][22][23][24][25][26], or 2-aminobenzimidazoles [27], however there are only a few examples describing their application for functionalization of the less reactive 5(6) position. These examples include Suzuki-Miyaura reactions [28,29] using vinyltrifluoroboronates or benzyltrifluoroborates as coupling agents, which, in general, do not include an appropriate hydrogen-bond acceptor according to our model, and/or require multi-step processes for their preparation. There are also a few examples regarding the catalytic amination at the 5(6) position [30][31][32][33][34], but some challenges remain to be overcome, particularly the enhancement of the low reaction yields, and/or the need for huge amounts of expensive catalysts. In this paper, we first describe the computer-aided design of 5(6)-substituted-2-(2-aminophenyl)benzimidazole derivatives aiming at the development of potential E. coli GyrB inhibitors. In addition, we report optimized synthetic processes for preparing these newly designed benzimidazole families, which encompass the appropriate substituents, via catalytic modulation of the less explored 5(6)-positions, using benchmark palladium-catalyzed reactions, namely Suzuki-Miyaura and Buchwald-Hartwig couplings with good yields.

Computer-Aided Design of Benzimidazole Derivatives with Potential E. coli DNA GyrB Inhibitory Activity
To generate the pharmacophore model, an alignment of the 18 training set molecules (see Supplementary Materials: Figure S2) through a stochastic conformer search was performed in MOE (Chemical Computing Group) [11] (Figure 2A).
The common structural features were identified, from which several pharmacophore queries were generated and further refined (by varying feature types, number of features and their radius). The selection and validation of the final pharmacophore model were grounded on its performance against a dataset (test set) composed of 90 compounds [9,10,[35][36][37][38][39][40][41] whose activity is well-known (61 active and 30 inactive compounds) (see Supplementary Materials: Figure S3). The best pharmacophore query was generated using MOE's Unified scheme, and contains five features: (i) a hydrogen bond acceptor region; (ii) an aromatic or hydrophobic region; (iii) one hydrophobic region; and (iv) two hydrogen-bond donor regions. This model ( Figure 2B) accurately predicted 90% of the active compounds (from the test set), with only 5% false positives. Figure 2B shows the optimized pharmacophore model superimposed with the selected benzimidazole scaffold bearing an -NH 2 (hydrogen bond donor) at 2-position and either (methylsulfonyl)phenyl, (methoxycarbonyl)phenyl and methoxyphenyl (hydrogen bond acceptors) at 5(6)-positions. Our next goal was to determine which type of functional groups are best suited to introduce in the 5(6)-position of the benzimidazole ring. To achieve this goal, we generated a virtual library of 2-(2-aminophenyl)-5(6)-substituted-benzimidazole derivatives (in total, 6681 compounds), using MOE tools. Initially, we screened the virtual library using the pharmacophore model, which we had previously selected and validated, in order to remove those derivatives whose features did not have hydrogen-bond acceptors. Next, docking studies were performed, using E. coli's DNA GyrB ATPase binding site (PDB entry 4KFG) [42], in order to identify the derivatives with the most potential for effective inhibition of this enzyme. Briefly, using GOLD (Cambridge Crystallographic Data Centre) [43] and the ChemScore scoring function, we ranked the compounds with the highest potential. ChemScore is a fitness (or scoring) function implemented in GOLD software to estimate the receptor-ligand binding affinity. This scoring function includes several terms, namely, a protein-ligand atom clash term and an internal energy term, taking account of hydrophobic-hydrophobic contact area, hydrogen bonding, ligand flexibility, and metal interaction. Then, we compared the results to a correlation established between ChemScore values and biological activies of known inhibitors (see Supplementary Materials: Figure S3). Based on this, we selected some of the highest ranking compounds based on their synthetic feasibility (Table 1). As an example, Figure 3 shows the docking pose obtained for the highest scoring compound 15 (Table 1, entry 1). From the analysis of the best scoring docking poses, we can observe three relevant hydrogen bond interactions: two between the -NH groups and Asn46 and Asp73; and another between the S=O group and Arg136. In addition, there are hydrophobic interactions between the aminophenyl ring and the surrounding non-polar protein side-chains. This corroborates the information obtained by the pharmacophore model as it states the importance of having hydrogen bond donors and acceptors in specific portions of the molecule, as well as aromatic/hydrophobic portions.
In sum, our aim to synthesize new families of 2-(2-aminophenyl)-5(6)-substitutedbenzimidazoles is explained by the need to insert hydrogen bond donor groups at 2position, while modulation of the 5(6)-position will allow the insertion of hydrogen bond acceptor groups. These groups will favor interactions with Asp73 and Arg136, respectively, and therefore increase their inhibition potential for E. coli's DNA gyrase B.

Synthetic Methods for the Preparation of 2-Aminophenyl-5(6)-Substituted Benzimidazole Derivatives
A simple synthetic pathway was developed for modulation of the desired 5(6)-position of the benzimidazole ring, starting with a set of reagents with an embedded functional group that could then be transformed to yield chemically diverse derivatives. Therefore, aiming toward the preparation of benzimidazoles [44] we started with 4-bromo-1,2diaminobenzene and 2-nitrobenzaldehyde (Scheme 1). Scheme 1. Synthesis of the core 5(6)-bromo-2-(2-nitrophenyl)-1H-benzimidazole substrate.
The optimization of reaction conditions was performed for the synthesis of the nitro derivative 1. The conditions explored and the results obtained are depicted in Table 2. Firstly, the reaction was carried out by mixing approximately equimolar quantities of 4-bromo-1,2-diaminobenzene and 2-nitrobenzaldehyde, using nitrobenzene both as solvent and oxidant, and the temperature was maintained at 180 • C for 8 h ( Table 2, entry 1).
Since the product could not be isolated by precipitation from the reaction mixture, the nitrobenzene was distilled at reduced pressure. The product was then purified by column chromatography, using silica gel as stationary phase, and a mixture of dichloromethane/ethyl acetate as eluent. Under these conditions, product 1 was obtained at 48% isolated yield. 13 C and 1 H nuclear magnetic resonance showed broad signals (118.1, 116.5, 114.7 ppm), typical of tautomerism associated with this class of compounds (see Supplementary Materials: Figures S5 and S6). Then, in order to avoid this troublesome distillation step, nitrobenzene was replaced by ethanol and the reaction was carried out both at reflux temperature (80 • C) and room temperature (25 • C) in air atmosphere, over 3 h (Table 2, entry 2-3). This methodology gave similar yields for product 1 (47-50%), but a significantly easier work-up. Indeed, oxidation/aromatization could proceed using the atmospheric oxygen [45], without the need of using nitrobenzene, usually described as the oxidant [44,46]. To improve the yield via activation of carbonyl group, we evaluated the effect of Montmorillonite K10 as a heterogeneous acid reusable catalyst, and both the reproducibility of the reaction and the final isolated yield increased to 62% when the reaction was performed at 25 • C (Table 2, entries 4). Using this synthetic methodology, compound 1 was obtained at a multi-gram scale (2.64 g, 8.3 mmol) and was then used as a starting material for further modifying the 5(6)-position of the benzimidazole ring.
Before starting the functionalization through Pd-catalyzed reactions, we proceeded with the benzimidazole -NH protection using benzyl or boc as protecting groups. In short, to a solution of the starting material 1, in dry tetrahydrofuran (THF) at 0 • C and under inert atmosphere, NaH was added, followed by benzyl bromide and a catalytic amount of tetra-n-butylammonium iodide (TBAI). The reaction was then heated to 70 • C for 2 h. After conventional work-up procedures, product 2 (2a + 2b) was isolated as a mixture of two regioisomers with 87% yield. As can be seen from the 1 H NMR spectra (see Supplementary Materials: Figure S11), the isomers were formed in a 1:1 ratio, and were used as such in further reactions. We proceeded with Pd-catalyzed reactions, using the mixture of N-benzyl-benzimidazole isomers, 2, as substrate. Scheme 2 shows all the optimized reaction conditions and corresponding yields.
In the first approach, through a Suzuki-Miyaura Pd-coupling reaction [47][48][49], the substrate 2 and the selected boronic acid were dissolved in a mixture of DME/EtOH 1:1, and the solvent was degassed. Then, Pd(OAc) 2 and PPh 3 were added and, after an incubation time of 15 min at room temperature, an aqueous solution of K 2 CO 3 was added, after which the reaction was heated and maintained at 90 • C for 72 h. After the usual work-up procedure followed by purification by column silica gel chromatography, the compound 4 was isolated, as a mixture of isomers, in 30% yield. Aiming to improve the yield of this reaction, the solvent was changed to a mixture of THF/H 2 O (4:1). After being degassed, Pd(OAc) 2 , Ph 3 P, and K 2 CO 3 were added, and the reaction was carried out at 70 • C for only 16 h. After a simple work-up, 4 was isolated in 66% yield. By applying these optimized reaction conditions, the formation of ester hydrolysis products was not observed and 5 was isolated in 80% yield.
Then, to prepare a family of molecules containing the sulfonylaniline group at 5(6)benzimidazol position (the best scoring family according to the model: Table 1, entry 1), the optimization of Pd-catalyzed Buchwald-Hartwig amination [50] reaction was carried out and the results are presented in Table 3. In a typical experiment, substrate 2 was mixed with 4-(methylsulfonyl)aniline, using Cs 2 CO 3 as base and Pd(OAc) 2 /phosphine, and the reaction was carried out at 100 • C for 16 h. First, the reaction was performed in toluene, using Pd/BINAP, Pd/DPEphos, and Pd/XPhos and a remarkable effect of the phosphine structure was observed (Table 3, entries 1-3). The reaction could only be carried out in the presence of palladium/XPhos, yielding 91% conversion after 16 h (Table 3, entry 3). To improve the solubility of the reaction components and evaluate the solvent effect, toluene was then replaced by dioxane and the reaction was performed under the same conditions. Again, BINAP and DPEphos did not originate an active catalytic system (Table 3, entry 4-5), the palladium/XPhos being slightly more active in this solvent (100% conversion, Table 3, entry 6). Then, we reduced the reaction time from 16 h to 8 h, and 79% conversion was obtained (Table 3, entry 7). Finally, we increased the substrate/catalyst ratio from 10 to 20 and, after 16 h, only 57% substrate conversion was obtained ( Table 3, entry 8). Overall, these results show that the catalytic activity is strongly dependent of the structure of palladium/phosphine catalyst, with the sterically hindered monodentate phosphine XPhos providing the most active catalytic system under these conditions, as previously reported [51].
Under our optimized conditions (Table 3, entry 6), and upon work-up procedure and purification by column chromatography in silica gel, the products 6 and 7 were isolated in yields of 78% and 81%, respectively (Scheme 2). Do, despite the lower nucleophilicity of the para derivative, when compared with the meta analogue, this factor did not translate into a noteworthy difference in reaction yield under the described conditions. To obtain the initially designed structures ( Table 1), deprotection of the benzyl group was performed via catalytic hydrogenation using H 2 and Pd/C [52], under mild conditions (50 • C, 3 bar H 2 ) for 8 h. Nevertheless, after this time no benzyl deprotection occurred, and only the reduction of -NO 2 was observed. Therefore, we used more vigorous reaction conditions (80 • C, 5 bar H 2 ), but a complex mixture of products was obtained.
To overcome this synthetic challenge, we decided to protect the benzimidazole 1 with boc, yielding 3a and 3b (Scheme 1). The reaction was carried out in dichloromethane (DCM) for 24 h, at room temperature, with the addition of tert-butyldicarbonate and 4dimethylaminopyridine (DMAP) as base. Following standard work-up, products 3a and 3b were obtained at 44.5% yield each (89% combined yield). It should be noted that, in this case, both isomers 3a and 3b could be easily separated by silica chromatography, being isolated and fully characterized. The assignment of each regioisomer was done using 2D-NOESY (see Supplementary Materials: Figures S18 and S23). Then, using the reaction conditions described above, 3a or 3b were coupled with 3,4,5-trimethoxyphenyl boronic acid or 3-fluoro-4-(methoxycarbonyl)phenyl boronic acid, giving products 8a or 8b and 9a or 9b with approximately the same NMR yields (75% and 70%, respectively). This result points to the fact that both regioisomers have similar reactivity towards Suzuki-Miyaura coupling. For further studies, isolation and full characterization, only the products 8a and 9a have been isolated upon column chromatography, in 72% and 66% isolated yields, respectively (Scheme 3). Full characterization is presented in the experimental section and SI. For 8a, deprotection was performed under the usual trifluoroacetic acid (TFA)/ dichloromethane (DCM) 1:1 for 2 h, resulting in quantitative yield of product 10. For 9a, due to the presence of the methyl ester group, we decided to use TFA/DCM 1:5 (5 h) and, following purification, 11 was obtained in 65% yield.
It should be noted that in the amination reaction of 3a with 3-(methylsulfonyl)aniline (Scheme 4), using the reaction conditions described above, without isolation of the protected product, we obtained product 12 in 48% isolated yield. This slightly lower yield may be attributed to partial deprotection of the boc group during the reaction, since carbamates are much easier to cleave under basic conditions with less basic amines such as imidazole or benzimidazole [53]. Therefore, the boc-protected product was not isolated, and we proceeded directly to the deprotection step. When the reaction was finished, the solid was filtered, the solvent was evaporated and the solid was re-dissolved in a mixture of TFA/DCM 1:1 for 2 h. After basic work-up to neutralize the acid, product 12 was purified by column chromatography, and isolated in 48% yield. Finally, to prepare our initially designed compounds (Table 1) with all the appropriate functional groups, according to the pharmacophore model developed, a catalytic hydrogenation of the nitrophenyl group was the final step of the synthetic route (Scheme 5). Briefly, the substrate, NH 2 NH 2 .H 2 O, and Pd/C were mixed in MeOH. The reaction was conducted at reflux temperature for 10-30 min, then filtered, and the solvent was evaporated, to give 13 in 91% yield, 14 in 56% yield (contaminated with dimeric amide in~25%, see Supplementary Materials: Figure S55), and 15 in 80% yield. These new chemical entities show a good match to our initially proposed computational model (ChemScore of 33.7, 28.7 and 27.5 for compounds 15, 14, and 13, respectively). Scheme 5. Pd/C catalyzed reduction of compounds 10-12 using hydrazine monohydrat.

Generation of the Pharmacophore Model
Through a comprehensive literature search [9,10,[35][36][37][38][39][40][41], several ligands of the E. coli DNA gyrase subunit B with different affinities were identified. From a total of 145 compounds, 61 were classified as actives (IC 50 ≤ 1.0 µM), 54 as intermediates (1.0 µM < IC 50 < 100 µM) and 30 as inactives (IC 50 ≥ 100 µM). From the actives dataset, the most structurally diverse 18 compounds were selected as the training set while the remaining 43 were included in the test set. In cases where multiple derivatives were present, the compound with the highest activity was chosen. For the purpose of generating the pharmacophore queries, Molecular Operating Environment (MOE) 2018.0802 software [11] was used, using the Unified annotation scheme. After the training set ligands' structural alignment, common features were identified using MOE's Unified annotation scheme. From a set of about 50 pharmacophoric queries generated through variation of feature type, number, and radii, the best model was selected based on their performance in discriminating between actives and inactives in the test sets.

Docking Studies on E. coli DNA GyrB
Docking studies were performed in GOLD 5.4 (Cambridge Crystallographic Data Centre, Cambridge, UK) [43], using E. coli's DNA Gyrase subunit B ATPase binding domain crystallographic structure (PDB entry 4KFG) [42]. For protein preparation, hydrogen atoms were added to the binding site residues and correct tautomers and protonation states were assigned. Water molecules and the ligand were deleted from the crystal structure before the docking studies. The binding site region was defined as the amino acid residues within a radius of 15 Å around the THR165 residue. To validate our protocol, the crystal structure ligand was docked into the defined binding site and the best scoring pose was able to reproduce the crystallographic pose with a root-mean-square deviation (RMSD) value of 0.57 Å. A set of 10 structurally diverse compounds with varying values of IC 50 was docked using the defined protocol. The following scoring functions were tested: ChemScore, GOLDscore, and ChemPLP. The scoring function that yielded a better docking score correlation with experimental IC 50 (ChemScore) was selected. From a set containing 6681 computationally generated derivatives of 2-aminophenyl-5(6)-substituted benzimidazoles, GOLD's ChemScore function was used to rank their predicted inhibitory activity. The best scoring poses were then graphically analyzed in MOE and the relevant protein side-chain interactions were determined.

5(6)-bromo-2-(2-nitrophenyl)-1H-benzimidazole (1):
In a round-bottom flask, 4-bromo-1,2-benzenediamine (2.5 g; 13.4 mmol), 2-nitrobenzaldehyde (2.23 g; 14.8 mmol), and 250 mg of Montmorillonite K10 were mixed in ethanol (30 mL). The reaction was stirred at room temperature for 4 h. The mixture was filtered, the solvent was evaporated, and the dark orange slurry obtained was dissolved in ethyl acetate, followed by the addition of silica powder and the evaporation of the solvent to dryness. The resulting solid was loaded in a silica column, and the crude was purified dichloromethane/ethyl acetate 10:1 (R f = 0.33). The product was obtained as a yellow solid in 62% yield (2.64 g). Characterization data in accordance to literature [45]. 1

Conclusions
The development of a pharmacophore model of E. coli DNA Gyrase B inhibitors, followed by docking studies using the crystallographic structure of this target, proved to be a relevant tool for the design and synthesis of the new chemical entities, based on 2-(aminophen-2-yl)-5(6)-substituted-1H-benzimidazoles.
The optimization of the condensation/oxidation reaction of 4-bromo-1,2-diaminobenzene with 2-nitrobenzaldehyde, yielding 5(6)-bromo-2-(nitrophen-2-yl)-1H-benzimidazole 1, led us to find a sustainable synthetic approach, using ethanol as solvent and Montmorillonite K10 as a reusable catalyst, which allowed a significant improvement of product yield and isolation process. Additionally, the N-boc revealed to be the ideal benzimidazole protecting group since its deprotection significantly does not lead to the formation of the side products obtained when using benzyl protecting group.
These studies pave the way for efficient functionalization of the more synthetically challenging 5(6) position, via cross-coupling Suzuki-Miyaura and Buchwald-Hartwig reactions, using multifunctionalized phenylboronic acids (products 8a and 9a were isolated with 72% and 66%, respectively) and amines (3 or 4-(methylsulfonyl)aniline), catalyzed by Pd/XPhos (79-81% isolated yields of 6 and 7). Finally, the use of NH 2 NH 2 .H 2 O and Pd/C, under air atmosphere, revealed to be a clean strategy to promote the reduction of the nitro group to the corresponding amine, giving 13, 14, and 15 with 91%, 56%, and 80% yields respectively.
In summary, we have developed efficient synthetic methods to prepare two different families of 2-(2-aminophenyl)-5(6)-substituted-1H-benzimidazoles, which encompass in their structures hydrogen bond donor groups at 2-position, and hydrogen bond acceptor groups at 5(6)-position. According to the predicted docking pose, these groups will favor interactions with Asp73 and Arg136, respectively, and therefore will potentially increase their inhibition potential for E.