Chemicals

The polyphenols were provided by the The Department of Pharmaceutical Biology, Faculty of Pharmacy, University of Ljubljana, and originally sourced from various companies including Alfa–Aesar (19), APIN Chemicals Ltd. (2), Carl Roth (7, 9, 15, 16, 18), ChromaDex (23), Fluka (4, 6, 10, 11, 12, 20), Janssen Chimica (21), Serva (5, 17, 22), Sigma–Aldrich (1, 8, 13, 24), LGC Standards (Quercetin) and Nookhandeh Institut für Naturstoffchemie Homburg (3). Silymarin was extracted from the Silliver® tablets and all compounds were stored in the dark in a refrigerator to protect them from light and possible decomposition. Prior to the biological assay, all compounds were dissolved in dimethyl sulfoxide (DMSO and stored at − 20 °C. All other chemicals were of ultra-pure analytical grade and were purchased from Sigma–Aldrich or Acros Chemical Company.

Computational analysisProtein preparation

In E. coli, only the apo form of MurF crystal structures are available (PDB Access Code 1GG4). Therefore, a homology model of the enzyme was constructed using SWISS-MODEL [14], a fully automated protein structure homology-modelling server. The protein sequence of MurF from E. coli was obtained from the UniProtKB database [15] with the code P11880 and an automated search was conducted to identify the most similar templates. The resulting templates were subsequently prioritized according to the quality of the resultant models. Among these, the crystal structure of MurF from A. baumanii (PDB Access Code 4QF5) which contained bound ATP and was the highest-ranked crystal structure, was chosen as a template based on chain A.

Although the two structures share only 40.4% identity, the ATP-binding pocket is highly conserved, and 15 of the 18 residues are identical. The initial structure of the ternary complex was further prepared using the Protein Preparation Wizard in Maestro, which is described below. Finally, the accuracy of the model was validated using three online tools: ERRAT [16], PROSA [17], and RAMPAGE [18].

Protein preparation wizard

A protein structure was generated using the SWISS-MODEL [14] and prepared using Schrödinger’s Protein Preparation Wizard, which is implemented in the Schrödinger Suite [19]. Water and other co-crystallized molecules, except for magnesium ions, were removed. Bond orders were automatically assigned, and hydrogen atoms were added. Selenomethionines were converted to methionines, and missing side chains were added. Water beyond the 3 Å radius of heteroatoms and with fewer than 3 hydrogen bonds to non-waters was removed. Heteroatoms were protonated at pH 7.0. Finally, the impred script was used to perform a constrained minimization of the protein with a maximum root mean square deviation (RMSD) of 0.30 Å.

SiteMap analysis

SiteMap is a software tool used to identify and evaluate binding sites. It is implemented in Schrödinger’s Maestro software [19]. We chose SiteMap because it has been validated using a set of 538 crystal structures [20]. SiteMap uses interaction energies between the protein and grid probes to identify energetically favorable sites. The program generates a grid of points in a three-step process, and then scores the sites based on their energetic properties. Up to 10 potential binding sites were allowed, and only those containing at least 15 points per site were considered. A more restrictive definition of hydrophobicity was used, along with a standard grid. Site maps located 4 Å or more from the nearest site points were truncated. The developers proposed a SiteScore value of 0.80 as a cutoff value to distinguish between sites that can and cannot bind ligands. We identified a total of seven binding sites, and selected the first binding sites with a SiteScore value greater than 0.8 for further analysis and grid box preparation.

Preparation of the grid box

The SiteMap software was used to identify binding sites, which were then used as a reference to define two boxes enclosing the three identified binding sites: (a) site 1, which is an ATP binding site with xyz coordinates of 7 Å, − 5 Å, 17 Å and a diameter of 27 Å, and (b) sites 2 and 3. Site 2 is an UDP-MurNAc-tripeptide (UM3DAP) binding site. A grid box enclosing both binding sites was defined with xyz coordinates of 6 Å, 2 Å, − 11 Å and a diameter of 24 Å. No constraints were applied during the process.

Preparation of ligands

Molecules were drawn using ChemDraw18 (PerkinElmer, MA, USA), and OpenBabel [21] was used to covert the structures to SMILES format. These structures were then imported into the Maestro program [19] and further prepared using LigPrep [22] from the Schrödinger Suite. The molecular conformations were generated using the OPLS4 force field, and the protonation states were adjusted using Epik [23] at a target pH of 7 ± 2. The resulting conformers were then stored in Maestro format for subsequent docking. The conformers were finally stored in Maestro format for further docking.

Docking procedure

Glide software [24] from Schrödinger Suite was used for docking in standard precision mode (SP). Active flavonoids were docked using the two previously defined lattice boxes. To enhance the flexibility of the nonpolar regions within the ligand, we applied a scaling factor of 0.8 to the van der Waals radii of ligand atoms. Additionally, ligand atoms possessing partial atomic charges with an absolute value below the specified cutoff of 0.15 were subjected to this scaling process. This approach allows for a controlled reduction in the nonpolar volume of the ligand while maintaining interactions with other atoms intact. Flexible ligand sampling was used and ring conformations were sampled. Epik state penalties were added to the docking score, intramolecular hydrogen bonding was rewarded and conjugated π groups had enhanced planarity. No constraints were used and post-docking energy minimization was performed with strain correction terms.

Visualization of results

All analyses and visualizations of the docking poses were performed using Schrodinger’s Maestro. [19] The original ATP-binding coordinates were used to compare ATP and flavonoid binding positions. To visualize the UM3DAP binding region another PDB structure (PDB Access Code 4QDI) with bound UDP was aligned to our model and then UDP was extracted for comparison with flavonoid binding position.

Biochemical analysisMurC-F inhibition assay

Inhibition of the Mur ligases was determined using the Malachite green assay, with slight modifications [25]. The mixtures for the respective Mur ligase assays had a final volume of 50 µL, which contained 100 µM of each tested compound dissolved in DMSO, added to: (1) MurC: 50 mM Hepes, pH 8.0, 5 mM MgCl2, 0.005% Triton X-114, 120 µM l-Ala, 120 µM UDP-N-acetylmuramic acid, 450 µM ATP, and 50 nM purified E. coli MurC [26]; (2) MurD: 50 mM Hepes, pH 8.0, 5 mM MgCl2, 0.005% Triton X-114, 100 µM d-Glu, 80 µM UDP-N-acetylmuramoyl-l-alanine (UMA), 400 µM ATP, and 150 nM purified E. coli MurD [27]; (3) MurE: 50 mM Hepes, pH 8.0, 15 mM MgCl2, 0.005% Triton X-114, 60 µM meso-diaminopimelic acid, 100 µM UDP-N-acetylmuramoyl-l-alanine-d-glutamate, 1000 µM ATP, and 20 nM purified E. coli MurE [28]; (4) MurF: 50 mM Hepes, pH 8.0, 50 mM MgCl2, 0.005% Triton X-114, 600 µM d-Ala- d-Ala, 100 µM UDP-N-acetylmuramoyl-l-alanine-d-glutamate-2,6-diaminopimelic acid (UM3DAP), 500 µM ATP, and 10 nM purified E. coli MurF [25, 29]. In all cases, the final concentration of DMSO was 5% (v/v).

After incubation for 15 min at 37 °C, the enzyme reaction was terminated by the addition of 100 µM Biomol green reagent®, and the absorbance was measured at 650 nm after 5 min. Experiments to determine the RA run in triplicate. Residual activities were calculated with respect to control assays without the tested compounds, but with the 5% DMSO carrier. The IC50 values were determined by measuring the residual activities at seven different compound concentrations, and they represent the concentration of the compound at which the residual activity was 50%. IC50 values were obtained by plotting the residual enzyme activities against the applied inhibitor concentrations, fitting the experimental data to the 4-parameter Hill equation: \( Y = } - \frac} - }}}}_} - }} \right) \times }\,}} \right)}} }} \), where X is the logarithm of the inhibitor concentration, and Y is the residual activity. The IC50 values were determined in three independent experiments. GraphPad Prism 8.2.0 (GraphPad Software, San Diego, CA, USA) was used for the fitting procedure.

Assay interference screen

To assess the potential interference of compounds using the malachite green-based assay, the following procedure was conducted:

Compounds, dissolved in DMSO, at 100 µM (5% DMSO, v/v), were combined with a substrate mixture (50 µL; containing 50 mM Hepes pH 8.0, 5 mM MgCl2, and 500 µM ATP) along with Biomol® reagent (100 µL). After incubating the mixture for 5 min at room temperature, the absorbance was measured at 650 nm. All experiments were performed in triplicate. A blank sample was prepared under identical conditions, utilizing only DMSO (5%, v/v). Compounds exhibiting a difference in absorbance greater than 0.1 (Acpd − Ablank ≥ 0.1) were categorized as causing interference with the assay.

Steady-state kinetic analysis of compound 21

For compound 21, Ki values were determined against MurF from E. coli. Ki determinations were performed under similar conditions to those described for the inhibition assay, where the different concentrations of one substrate and a fixed concentration of the other two were used. First, the concentration of UM3DAP was varied (25, 50, 100, 200 µM) at fixed ATP (500 µM) and d-Ala-d-Ala (600 µM)), then the concentration of d-Ala-d-Ala was varied (50, 100, 200, 400, 600 µM) at fixed ATP (500 µM) and UM3DAP (100 µM), and finally, the concentration of ATP was varied (50, 100, 350, 500 µM) at fixed UM3DAP (100 µM) and d-Ala-d-Ala (600 µM). The concentrations of 21 were 0, 25, 50, 75, 100, 200, 350 and 500 µM and the concentration of the MurF was 20 nM. After a 15 min incubation, 100 µM Biomol green reagent® was added, and the absorbance was read at 650 nm after 5 min. All experiments were run in triplicate.

The data were analysed using the SigmaPlot 12.0 software. The initial velocities were fitted to competitive, non-competitive, uncompetitive and mixed enzyme inhibition. The Ki and mode of inhibition from the best ranking model were used, as provided by the software.

Antibacterial assays

Antimicrobial testing was performed by the broth microdilution method in 96-well plates following the Clinical and Laboratory Standards Institute guidelines and European Committee on Antimicrobial Susceptibility Testing recommendations (Clinical and Laboratory Standards Institute). Bacterial suspensions equivalent to 0.5 McFarland turbidity standard were diluted with cation-adjusted Mueller–Hinton broth with TES buffering (ThermoFisher Scientific), for a final inoculum of 105 CFU/ mL. Compounds dissolved in DMSO and the inoculum were mixed and incubated for 20 h at 35 °C. After this incubation, the minimal inhibitory concentrations (MICs) were determined by visual inspection, as the lowest dilution of the compounds that showed no turbidity. The MICs were determined against two reference bacterial strains, S. aureus (ATCC 29,213) and E. coli (ATCC 25,922). Tetracycline was used as the positive control on every assay plate, with MICs of 0.5 and 1 µg/mL for S. aureus and E. coli, respectively. All MICs were determined in three independent experiments.