Introduction

Infection due to pathogenic bacteria is a leading cause of morbidity and mortality worldwide. The alarming rise of drug-resistant bacteria, such as Gram-positive Staphylococcus aureus [1, 2], emphasizes the development of new antibacterial agents for treatment. S. aureus is one of the leading causative organisms of infections acquired in the community and after surgery or hospital. It is a human pathogen causing significant morbidity and mortality globally. Apart from skin infections, S. aureus is now a major causative organism of bacteremia and infective endocarditis [3, 4] with a negative impact on patient outcomes. This highlights the necessity for developing new antibacterial agents. The intracellular adenosine triphosphate (ATP)-dependent Mur ligase enzymes (MurC-F) are validated as potential therapeutic targets in the antibacterial drug discovery.

The structural integrity of the bacterial cell wall is critically important for the survival of both Gram-positive and Gram-negative bacteria. By maintaining the cell shape and providing rigidity, the bacterial peptidoglycan plays a crucial role in cell survival. This peptidoglycan, comprising sugar molecules and peptides is synthesized and integrated outside the cytoplasmic membrane of the cell. For many years, disruption of the peptidoglycan layer is considered a prime target for the development of effective antibacterial agents. This rigid layer is incorporated by a multi-step synthesis and consists of many targets that do not exist in the human cell. Central to this pathway are four topologically similar ATP-dependent Mur ligases MurC, MurD, MurE, and MurF. These enzymes are involved in the intracellular assembly of peptidoglycan by catalyzing the consecutive addition of l-Ala, d-glutamic acid (d-Glu), meso-diaminopimelic acid (m-DMPA) (or l-Lys), and d-Ala-d-Ala to the starting precursor UDP-N-acetylmuramic acid (UDP-MurNAc) [5, 6]. The Mur ligases operate by a common reaction mechanism [7] and share similar three-dimensional structures with conserved amino acid regions [8,9,10]. They are functionally essential for the survival of bacteria. Further, the ATP-dependent catalytic pocket of the Mur ligases shows the highest level of sequence and structural similarity [8] but they are not closely related to any ATP-utilizing human enzyme [11]. MurD is second in the series of Mur ligases that catalyzes the formation of a peptide bond between d-Glu and cytoplasmic intermediate uridine-5′-diphosphate-N-acetylmueamoyl-l-alanine (UMA). Due to its high specificity for the d-amino acid substrate, involvement in the synthesis of peptidoglycan, and its absence in mammals, MurD enzyme is considered as an attractive target for the development of new antibacterial agents [12, 13]. Further, the cloned MurD from S. aureus showed high percentage similarity to MurD proteins from Streptococcus pyogenes (66%), Bacillus subtilis (65%), Escherichia coli (54%), and Haemophilus influenzae (55%) [14].

The catalytic pocket of MurD comprises an N-terminal domain, a central domain, and a C-terminal domain. The N-terminal domain comprises five-stranded parallel β-sheet surrounded by four helices [15, 16] and is mainly involved in the specific recognition of the nucleotide substrate. This domain binds the nucleotide substrate UDP moiety through a cleft constituted by the diphosphate-binding pocket and two hydrophobic loops. The central binding domain consists of six-stranded parallel β-sheet which is surrounded by seven α-helices and flanked on both sides by smaller antiparallel three-stranded β-sheet. These parallel β-sheets play an important role in the binding of adenosine triphosphate (ATP). The C-terminal domain consists of six-stranded β-sheet (five parallel and one antiparallel β-strands) and is surrounded on both sides by five α-helices. This domain also called as Rossmann fold is primarily responsible for the binding of substrates amino acid or dipeptide [6, 17]. The ATP-binding pocket of MurD enzyme is present in a cleft formed between the central and C-terminal domains. In addition to loop, the C- and N-terminal binding domains also undergo conformational changes when the substrate binds with the enzyme. The MurD bring together the UMA and ATP in the cleft and orient them properly in the catalytic pocket for the formation of an acylphosphate intermediate. At the binding site, MurD also orients the d-glutamine for the nucleophilic attack and thereby stabilizes the tetrahedral intermediate [18]. Further, site-directed mutagenesis indicated that among different bacterial Mur ligases, the GXXGKT/S ATP-binding consensus sequence and seven residues are common invariants [19]. However, differences exist in the overall topologies and amino acid sequences of MurD binding pockets from S. aureus, E. coli, Mycobacterium tuberculosis, Streptococcus pneumonia, and Borrelia burgdorferi. S. aureus MurD showed a high degree of sequence similarity with S. pneumonia MurD but low sequence similarity was observed with MurD from B. burgdorferi, E. coli, and M. Tuberculosis [20].

Novel phosphinates, phosphonates, and N-acetylmuramic acid analogs have been developed as transition-state analog inhibitors of E. coli MurD; however, these compounds did not show promising results [21,22,23,24]. Phosphorylated hydroxyethylamines have been reported to have inhibitory activity against the whole MurC to MurF cascade [25] with IC50 values in the micromolar range. This paved the way for the development of multiple inhibitors of Mur ligases. In a further development, substituted naphthalene-N-sulfonyl derivatives of glutamic and substituted analogs of d-glutamic acid have been synthesized and evaluated as a potential inhibitor of the MurD from E. coli [26,27,28,29]. Further, it has been shown that l-Glu-containing compounds can also inhibit MurD ligase, despite the intrinsic preference for the d-Glu amino acid by MurD [26, 27]. 5-Benzylidenerhodanine and 5-benzylidenethiazolidine-2,4-dione-based E. coli MurD inhibitors were designed and synthesized. High-resolution 3D crystal structures of these d-Glu-based inhibitors [30, 31] provided insight into the molecular basis for binding within the catalytic pocket. In further development, naphthalene-N-sulfonyl derivatives possessing conformationally rigidified mimetics of d-Glu have been synthesized and evaluated as inhibitors of E. coli MurD [32, 33]. Some of these rigid mimetics displayed promising inhibitory activity against the whole cascade of Mur ligases (MurD-MurF). Structure-based virtual screenings and target-based design approach have been employed to identify novel MurD inhibitors [34,35,36]. But these efforts not resulted in inhibitors with potent antibacterial activity. Further, based on the virtual screening result and de novo structure-based design, E. coli MurD d-Glu-based inhibitors possessing 5-benzylidenethiazolidin-4-one scaffold are reported with improved potency (IC50 between 3 and 7 μM) [37]. However, only a few of these compounds exhibited weak activity against S. aureus and Enterococcus faecalis. Further, 2-thioxothiazolidin-4-one-based inhibitors with IC50 in the micromolar range against S. aureus MurD enzyme is reported [38]. One of these compounds exhibited significant inhibitory activity against methicillin-resistant S. aureus (MRSA) ATCC 43300 and S. aureus ATCC 29213 with MIC values of 8 μg/mL.

The aim of the present study was to identify potential virtual hits with inhibitory potency against S. aureus MurD ligase. The homolog 3D structure of S. aureus MurD prepared earlier [39] was used for the present study. Using structure-based in silico high-throughput virtual screening (HTVS) approach, potential hits were identified from library databases of approximately 1.6 million compounds. A filter based on docking energy identified ten potential hits with high negative binding free energy against the modeled S. aureus MurD ligase. The binding mode and stability of one of the hit and protein complex were analyzed by molecular dynamics (MD) simulation.

Materials and methods

In silico high-throughput virtual screening

We carried out structure-based high-throughput virtual screening using the Glide docking module incorporated in Schrödinger, LLC, New York, 2018-1. Earlier prepared homolog 3D structure of S. aureus MurD was employed as 3D search query using the available library databases (Enamine, lifeChemicals, TimTec, Asinex, Specs, ChemDiv, KeyOrganics, Maybridge) of approximately 1.60 million compounds. Molecules with high binding affinity within the catalytic pocket of S. aureus MurD were captured as hits and ranked according to their standard precision docking score. A total of top 3820 ligands ranked by the HTVS were clustered using normalized volume overlap with a grid spacing of 1.0 Å option into ten groups each inhabiting a different chemical space. Forty molecules from each cluster with the highest docking score were selected and subjected to Glide standard precision (SP) docking. Top 10% of molecules were analyzed to discard compounds with lower Glide score and not forming hydrogen bonding and hydrophobic interactions in the catalytic pocket. The selected 40 virtual hits were further docked into the catalytic pocket of modeled S. aureus MurD protein in an extra-precision mode using energetic-based Glide [40] without applying any constraints. Ten virtual hits H1–H10 were selected based on Glide score, Glide model energy, Glide energy, and hydrogen bond interactions (Table 1, Supplementary Fig. S1, and Table S1). Subsequently, their binding free energies were calculated using the Prime Molecular Mechanics Generalized Born Surface Area (MM-GBSA) method [41] (Table 2) with an OPLS3 force field and VSGB 2.0 energy model [42]. Molecular docking and binding free energy calculation were carried out using protocols described earlier [39]. Notably, the identified hits H1–H10 had scaffold topology different from earlier reported inhibitors of MurD. Additionally, Qikprop simulation was performed on all ten identified compounds to characterize the absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of all compounds (Supplementary Table S2). Further, the docked complexes of virtual hit H5 with modeled S. aureus MurD were used to run 30 ns molecular dynamics simulation.

Table 1 Extra-precision docking score of selected virtual hits H1–H10 in the active site of modeled S. aureus MurD ligase (kcal/mol)
Full size table
Table 2 Binding free energy contribution (MM-GBSA) of selected virtual hits H1–H10 in the active site of modeled S. aureus MurD ligase (kcal/mol)
Full size table

Molecular dynamics

In order to validate the stability and to get insight into the binding modes of virtual hit H5 within the catalytic pocket of modeled S. aureus MurD, 30 ns MD simulations were performed using the Desmond [43] software with OPLS3 force field [44]. The docked complex H5/S. aureus MurD protein was solvated with the TIP4P solvent model [45, 46] in an orthorhombic box keeping a 10 Å buffer region between protein atoms and box sides. Overlapping water molecules were deleted and then the system was neutralized by adding 10 sodium ions. The solvated system contains 47,571 atoms (including 13,478 water molecules) and 1,122,523 Å3 final box volume. Limited-memory Broyden Fletcher Goldfarb Shanno (LBFGS) minimization of the whole system was carried out with a minimum of 10 steepest descent steps, 3 vectors, and a gradient threshold of 25 kcal/mol/Å. During minimization, the maximum iterations were 2000 with a convergence threshold of 1.0 kcal/mol/Å. The system was then gradually heated to 300° K in the NPT ensemble with a time step of 2 fs. A multiple time step reversible reference system propagator algorithms (RESPA) integration algorithm was employed throughout the MD simulation with time steps of 2 fs for bonded, 2 fs for near nonbonded, and 6 fs for far nonbonded interactions. The particle mesh Ewald (PME) method [47] was employed for computing the long-range electrostatic interactions with a 9 Å short-range cutoff radius for van der Waals and Coulomb interactions. The system was then subjected to a 30 ns MD simulation in the NPT ensemble (T = 300 K, thermostat relaxation time = 200 ps; p = 1 atm; barostat relaxation time = 200 ps) using a Nose–Hoover thermostat [48] and Martyna–Tobias–Klein barostat [49]. The MD trajectory and 3D structures were visually inspected and analyzed using the Maestro graphical interface. The results of root mean square deviation (RMSD), root mean squared fluctuation (RMSF), and interactions of H5 with the catalytic pocket residues have been incorporated into the results.

S. aureus MurD activity inhibition assay

High-throughput virtual hits H5 (Enamine T1827917) and H10 (Enamine T1990672) were procured from Enamine Ltd., Kiev, Ukraine. The inhibitory activity of these two selected compounds was determined against S. aureus MurD (ProFoldin, USA) as described earlier [50, 51] by the malachite green assay with slight modifications. All of the experiments were run in duplicate. The assay mixture with final volume of 50 μL consisted of the following: 500 mM TrisHCl, pH 8.0, 400 mM KCl, 10 mM DTT, 0.05% Triton X-100, 10 mM MgCl2, 1 mM UDP-MurNAc-l-Ala, 5 mM d-Glu, 10 mM ATP, 5000 nM purified MurD from S. aureus, and 0.6 μL of different concentrations of each tested compound dissolved in DMSO. In all cases, the final concentration of DMSO was 5% (v/v). The reaction was initiated by the addition of MurD enzyme to the assay mixture and was incubated at 37 °C for 60 min and then quenched with Dye MPA3000. Incubated for 5 min and absorbance was measured at 650 nm and percentage inhibition of each compound was calculated with respect to a negative control (0% activity) and enzyme reaction (100% activity). IC50 values were determined using the GraphPad PRISM.

Antibacterial activity

The antibacterial activity of two selected virtual hits H5 (Enamine T1827917) and H10 (Enamine T1990672) was carried out against Gram-positive S. aureus (NCIM 5021 and NCIM 5022), methicillin-resistant S. aureus (MRSA strain 43,300), and B. subtilis (NCIM 2545), and Gram-negative strains E. coli (NCIM 2567), K. pneumonia (NCIM 2706), and P. aeruginosa (NCIM 2036) according to the guidelines of the Clinical Laboratories Standard Institute [52]. Triplicate analyses were performed for each of the selected strains. Microdilution broth technique was employed to determine MIC values by using the Mueller Hinton medium (Hi-media). Test compounds were dissolved in sterile dimethyl sulfoxide (DMSO) and sterile DMSO was used as a negative control. Ciprofloxacin and methicillin were used as control drugs (positive) in sterile DMSO. Plates were incubated at 37 °C for 24 h and were read for the MIC (Fig. 1).

Fig. 1
figure 1

Represents chemical structures of potential hits H1–H10 obtained by structure-based high-throughput virtual screening

Full size image

Results and discussion

Extra-precision docking and binding free energy by prime MM-GBSA

Analysis of the extra-precision docking poses of virtual hits H1–H10 exposed hydrogen bonding, π-π stacking, and salt bridge interactions mainly in the region of Lys19 to Gly147 and Lys328-Gln432 (Fig. 2 also Supplementary Fig. S1 and Table S1). A comparison of different docking poses of H1–H10 exhibited interactions with Lys19, Gly80, Glu101, Ser168, His196, Arg355, Trp425, Asp426, and Gln432 (Supplementary Table S1) residues belonging to the amino acid stretch that forms the catalytic pocket. Glide score of compounds H1–H10 ranges from − 10.23 to − 9.02 kcal/mol (Table 1); the highest glide score of − 10.23 kcal/mol was observed with H5. To illustrate the binding mode, compounds H1, H5, and H10 were analyzed and results were compared. As shown in the docked pose of H1 (Fig. 2a) in the active site of S. aureus MurD, the NH of one of the protonated piperidine ring formed hydrogen bonding interaction with the side chain carbonyl oxygen of Asp426 ((>N+(H)H⋯O=C<, 2.18 Å). The nitrogen of the same ring established a salt bridged interaction with the carboxylate oxygen of Asp426. The carbonyl oxygen and NH adjacent to this piperidine ring exhibited hydrogen bonding interactions respectively with the backbone NH of Trp425 (>C=O⋯NH, 2.50 Å) and backbone NH of Pro79 (NH⋯O=C<, 1.96 Å). The other carbonyl oxygen of this compound formed two hydrogen bonding interactions one each, respectively with the side chain protonated ring nitrogen of His196 (>C=O⋯rNH+, 1.82 Å) and the side chain NH of Arg355 (>C=O⋯NH, 2.05 Å). The docked pose of compound H5 exhibited seven hydrogen bonds (Fig. 2b), two more than compound H1. Precisely, OH group present on one of the terminal phenyl ring accepted a hydrogen bond from the side chain carboxylate oxygen of Glu23 (OH⋯−O–C=O, 1.84 Å). The two NH of –NHCONH– fragment adjacent to this hydroxyphenyl ring formed one hydrogen bonds each with the backbone NH of Gly147, while the carbonyl oxygen of the same fragment exhibited hydrogen bonding interaction with the side chain NH of protonated Lys19 (>C=O⋯HN(H2)+). Additionally, this hydroxyphenyl ring also exhibited two π-π stackings, one each with the pyrrole and benzo parts of the Trp425. This lucrative hydrophobic interactions further stabilized compound H5 in the catalytic pocket. The central phenyl ring showed a π-cation interaction with the protonated nitrogen of Lys19 while the OH present on another terminal hydroxyphenyl ring established two hydrogen bonding interactions one each with the backbone NH of Gly80 (HO⋯NH, 2.04 Å) and the side chain carboxylate oxygen of Glu101 (OH⋯−O–C=O, 1.71 Å). It is evident that all of the hydrogen bonding interactions for this compound is in the N-terminal and central domain of the catalytic pocket. On the other hand, docking pose of compound H10 (Fig. 2c) showed four hydrogen bonding interactions, three less compared to compound H5. Hydrogen bonding interaction for this compound was observed with Lys19, Ser168, Ala423, and Gln432, key binding residues of catalytic pocket. Our extra-precision docking result also exposed a π-π stacking interaction of one of the phenyl ring with the phenyl ring of Phe170. However, no π-cation interaction was observed for this compound. It is evident that compound H10 occupied all three domains of the catalytic pocket and established hydrogen bonding interactions with residues of all three domains. Further, generated binding cavity pocket (Supplementary Fig. S2) showed that the compound H5 is well placed within the catalytic pocket and both –NH–CO–NH– fragments are oriented towards hydrogen bond acceptor regions (red color).

Fig. 2
figure 2

Represent the interactions of virtual hits (a) H1, (b) H5, and (c) H10 with the catalytic pocket residues of modeled S. aureus MurD ligase

Full size image

Prime MM-GBSA-based binding free energy was calculated for rank ordering of the poses of virtual hits H1–H10. The calculated free energies of binding (ΔGbind) ranges from − 88.89 to 65.03 kcal/mol indicating the high binding affinity of compounds H1–H10 within the catalytic pocket of modeled S. aureus MurD enzyme. It is evident from Table 2 that the van der Waals (ΔGvdW, − 82.75 to − 31.29 kcal/mol) and Coulomb (ΔGCoul, 26.85 to 81.41 kcal/mol) energy terms are major favorable contributors for ligands binding. The electrostatic solvation energy (ΔGSolv, 14.85 to 67.31 kcal/mol) term strongly disfavors the binding, whereas non-polar contribution (ΔGLipo) interactions moderately favor the binding of all compounds. Covalent binding energy moderately favors the binding of compounds H1 and H7, whereas it disfavors the binding of all other compounds. Moreover, high negative values of ΔGCoul indicated that the catalytic pocket is lined with the polar residues and there is less contribution from the hydrophobic residues for ligands binding to the modeled S. aureus MurD enzyme. It is also evident from the high negative values of ΔGvdW that ligands are buried well within the polar region upon binding and this is in agreement with our docking result (Table 1) where van der Waals interaction energy also favors the ligands binding.

The identified virtual hits H1–H10 obeyed Lipinski’s rule of five (0 to 2), indicating the drug like property of these molecules (Supplementary Table S2). These molecules are devoid of CNS activity as indicated by the predicted central nervous system (CNS) activity (− 2 to 1) on a scale of − 2 as inactive and + 2 as active. The polar surface area (PSA) which is indicative of van der Waals surface area of polar nitrogen and oxygen atoms of molecules ranges between 76.98 to 172.52 Å2 and is well within the recommended range of 7–200 A2. QPlogPo/w which is related to the hydrophobicity of the molecule is in the range of 2.19 to 5.22 and is well within the recommended range of—2.0 to 6.5. The predicted apparent Caco-2 cell permeability which is a model for the gut-blood barrier is the range 60.79 to 1023 nm/s, indicating non-active transport of these molecules. Predicted human oral absorption of H1–H10 is in the range of 71.71 to 100% on a scale of 0 to 100%. For virtual hits H5 and H10, the predicted IC50 values for the blockage of human ether-a-go-go (HERG) K+ channels (QPlogHERG) are − 4.76 and − 4.58, respectively indicating the safety of these molecules. While QPlogHERG values for other eight compounds (H1-H4 and H6-H9) range between − 6.55 to − 9.02, indicating the blockage of HERG K+ channels and hence probable cardiac toxicity of these molecules. Further, lower solvent accessible surface area (SASA) of these molecules (688.42 to 973.40 Å2) on a recommended range of 300–1000 Å2 indicated the burial of these molecules within the catalytic pocket.

S. aureus MurD activity inhibition assay

Based on the extra-precision docking and MM-GBSA results, two best-ranked virtual hits H5 (Enamine T1827917) and H10 (Enamine T1990672) were tested against S. aureus MurD in the malachite green assay [50, 51]. The IC50 values were determined for both compounds, and in both cases, they were in the micromolar range. The most active compound, H5 (Enamine T1990672), has an IC50 of 7 μM whereas compound H10 showed an IC50 of 30 μM. The two active compounds H5 and H10 arising from the molecular docking analysis were among the first 10 top-ranked database compounds identified.

Antibacterial activity

The antibacterial activity of the hit compounds H5 (Enamine T1827917) and H10 (Enamine T1990672) was determined against Gram-positive S. aureus (NCIM 5021 and NCIM 5022), methicillin-resistant S. aureus (MRSA strain 43,300), B. subtilis (NCIM 2545), and Gram-negative strains E. coli (NCIM 2567), K. pneumonia (NCIM 2706), and P. aeruginosa (NCIM 2036) (Table 3). None of the tested hit compounds demonstrated antibacterial activity against E. coli (NCIM 2567), K. pneumonia (NCIM 2706), P. aeruginosa (NCIM 2036), and Bacillus subtilis (NCIM 2545). Compound H10 was also observed to be inactive against S. aureus (NCIM 5021 and NCIM 5022), methicillin-resistant S. aureus (MRSA strain 43,300), and B. subtilis (NCIM 2545) while compound H5 prevented growth of S. aureus NCIM 5021, S. aureus NCIM 5022, and methicillin-resistant S. aureus (MRSA strain 43,300) at higher concentrations (MIC 128, 128, and 256 μg/mL, respectively). The lack of antibacterial activity is probably due to the failure of these compounds to enter the bacterial cell.

Table 3 Antibacterial activity of synthesized compounds 4a-s and 9a-f against Gram-positive and Gram-negative bacteria
Full size table

Minimum inhibitory concentration

Molecular dynamics

The MD simulation of the docked H5/modeled S. aureus MurD complex was performed to get insight into the mechanistic aspects of the protein-ligand interactions. The trajectory frames obtained from the simulation were inspected to analyze the predicted binding pose of H5. The RMSD values of protein all backbone, C-α, and heavy atoms in the inhibitor/modeled protein complex increased to 4.41, 4.34, and 4.59 Å, respectively during equilibration phase of 15 ns (Fig. 3a). After the equilibration phase, the mean RMSD values of complex backbone, C-α, and heavy atoms were observed to be 0.51, 0.475, and 0.65 Å, respectively during 15 to 30 ns MD trajectory. This indicated that there are no significant structural changes in protein during simulation. It is also evident from Fig. 3 a that the binding pocket residues RMSD cope well with the ligand movement. The radius of gyration (rGyr) of both backbone (blue circle) and C-α (red triangle) atoms throughout the MD trajectory is shown in Supplementary Fig. S3. The rGyr of backbone (blue circle) and C-α (red triangle) atoms gradually increased up to 15 ns and then stabilized respectively in the range 3.24 to 4.54 and 3.12 to 4.51 Å, during rest of the simulation time. This further indicated that the whole system is in relaxed conformations resulted from solvent effect. Further, protein all backbone and C-α atoms of catalytic pocket binding residues exhibited the RMSF values in the range 0.87–2.43 Å and 0.85–2.40, respectively (Fig. 3b) indicating low fluctuations of these residues. The high fluctuations of backbone and C-α atoms RMSF were observed between amino acid stretches Leu43-Gln45 (2.70–3.24 and 2.52–3.34 Å, respectively) present on loop (Asp39-Ala47) and Ser341-Gln344 (3.30–3.68 and 3.32–3.3.78 Å, respectively) present on β-sheet (Ile346-Gly351). These residues are away from the catalytic pocket.

Fig. 3
figure 3

Represents a RMSD (Å) plot of the simulated positions of C-α, backbone, and heavy atoms of modeled S. aureus MurD from those in the initial structure. b Protein RMSF profile of H5/modeled MurD complex during MD simulation. c Virtual hit H5 and modeled MurD protein interaction fraction plot during MD simulation. d 2D interaction diagram of H5 with the catalytic pocket binding residues of modeled MurD protein throughout 30-ns simulation trajectory

Full size image

Among eight hydrogen bonds predicted in XP-docking simulation, seven hydrogen bonds were preserved for compound H5/S. aureus MurD complex during 30 ns MD simulation. It is evident from Fig. 3 c and d (also Supplementary Fig. S4) that compound H5 occupied mainly N-terminal and central domains of modeled MurD catalytic pocket. The binding of H5 within the catalytic pocket was observed to be driven by both hydrogen bond and hydrophobic interactions with the conserved residues Lys19, Glu23, Gly80, Glu101, and Gly147. It is also evident that both hydroxyphenyl rings and –NH–CO–NH– fragments are responsible for the recognition of this compound within the active site. Both hydroxyphenyl rings are positioned at the UDP-MurNAc-l-Ala-d-Glu (UMAG) site, the binding site of a d-Glu moiety of other MurD inhibitors containing the d-Glu moiety [37, 53] while urea moieties occupied both UMAG and ATP binding sites.

Accurately, OH present on one of the phenyl ring formed a water-mediated high-frequency hydrogen bond (75% of the MD trajectory) with the side chain carboxylate oxygen of Glu23 (OH⋯−O–C=O). This high-frequency interaction is evolved due to the less conformational flexibility of Glu residue (ψ = 30°) (Supplementary Fig. S5a). This phenyl ring also exhibited a high-frequency π-cation interaction (74% of the MD trajectory) with the side chain NH of protonated Lys19 and a low-frequency π-π stacking (26% of the MD trajectory) with the side chain phenyl ring of Trp425. The high-frequency interaction is due to the less conformational flexibility of Lys 19 (ψ = 30°, Supplementary Fig. S5b) whereas the low-frequency π-π stacking interaction is evolved due to high conformational flexibility of Trp425 (ψ = 120°, Supplementary Fig. S5c). Both NH of –NH–CO–NH– fragment attached to this phenyl ring exhibited strong hydrogen bonding interactions (86% of the MD trajectory in both cases) with the backbone carboxylate oxygen of Gly147 (NH⋯O=C<) in a bidentate manner. It is evident from supplementary Fig. S5d that this strong hydrogen bonding interaction is probably due to the less conformational flexibility of Gly147 residue (ψ = 35°) during most of the simulation time. The carboxylate oxygen of the same fragment accepted a high-frequency hydrogen bond (56% of the MD trajectory) from the side chain NH of Lys19 (>C=O⋯HN(H2)+). The oxygen of OH group present on another phenyl ring exhibited a strong frequency hydrogen bonding interaction with the backbone NH of Gly80 (HO⋯NH, 65% of the MD trajectory), while the hydrogen of same OH group formed a water-mediated low-frequency (27% of the MD trajectory) hydrogen bond with the side chain carboxylate oxygen of Glu101 (OH⋯−O–C=O). This low-frequency interaction is probably due to the high conformational flexibility of Glu101 residue throughout the 30-ns simulation. The carbonyl oxygen of –NH–CO–NH– function attached to this another phenyl ring exhibited a high frequency (54% of the MD trajectory) direct hydrogen bonding interaction (>C=O⋯HN(H2)+) and a water-mediated moderate frequency (34% of the MD trajectory) hydrogen bonding interaction with NH of Lys19 (>C=O⋯H(H)O⋯HN(H2)+).

During MD simulation, the rGyr of the ligand was in the range 5.07 to 5.57 Å, which indicates the ligand stability within the catalytic pocket (Supplementary Fig. S6). The lower RMSD value in the range of 0.89 to 1.96 Å further shows the less conformational flexibility and stability of ligand during MD simulation. The solvent accessible surface area (SASA) of ligand H5 was observed to be stable in the range 135.44 to 238.62 Å2 throughout the 30-ns simulation time indicating no significant change in the binding pocket volume. The polar surface area (PSA) of H5 was observed to in the range 189.80 to 215.32 Å2 during the last 25 ns of MD simulation, indicating the complete burial of compound within the hydrophilic pocket and this is in agreement with the docking study result (Supplementary Fig. S2). The generated simulation quality analysis result (Supplementary Fig. S7) of H5/S. aureus complex showed the total energy of the system between − 125,600 to − 125,200 kcal/mol, further indicating the stability of the modeled protein during 30-ns MD simulation. Further, superposition of the conformations of H5 after MD simulation and best XP-docked pose exhibited similar orientation with a low RMSD value 0.821 Å (Supplementary Fig. S8), indicating the conformational stability of binding pocket and ligand H5 during MD simulation. Above obtained results also indicated the rationality and validity of docking protocols used.

Conclusion

The murD enzyme is second in the series of Mur ligases and plays an important role in the formation of a peptide bond between d-Glu and cytoplasmic intermediate UMA. Its high specificity for d-amino acid substrate and its absence in mammals makes MurD a promising target for the development of new antibacterial agents. In the present study, we performed structure based in silico high-throughput virtual screening against S. aureus MurD enzyme using a compound library of ~ 1.6 million small molecular entities. Ten virtual hits H1–H10 was selected based on the extra-precision Glide score, interactions within the catalytic pocket, and high negative values of binding free energy. Binding free energy calculation by MM-GBSA approach showed van der Waals and Coulomb energy terms as major favorable contributors for ligand binding. While electrostatic solvation energy term, strongly disfavors the binding of ligands to the S. aureus MurD enzyme. Among the ten selected compounds, two compounds H5 and H10 showed favorable ADMET profile with low affinity for HERG K+ channels. The inhibitory activity of these two compounds against S. aureus MurD enzyme was performed using the malachite green assay. Compound H5 exhibited the highest activity with IC50 values of 7 μM while compound H10 displayed lower activity with an IC50 value of 30 μM. Compound H5 inhibited the growth of S. aureus NCIM 5021, S. aureus NCIM 5022, and methicillin-resistant S. aureus (MRSA strain 43,300) at high concentrations, while compound H10 was inactive against all the tested bacterial strains. The lower activity of compound H5 and inactivity of compound H10 against the tested bacterial strains is probably due to the failure of these compounds to enter the bacterial cell. Further stability of compound H5/S. aureus MurD complex was validated by 30-ns molecular dynamics simulation. Hydrogen bonding interactions with Lys19, Glu23, Gly80, and Gly147 as well as π-cation interaction with Lys19 played a crucial role in the stabilization of inhibitor within the catalytic pocket. Hydrophobic interaction was observed to be less significant for the inhibitor binding to S. aureus MurD enzyme.