Integrated Experimental and Computational Discovery of Repurposed and Novel Inhibitors Targeting Staphylococcus aureus Biofilms

Abstract

Background:

Bacterial biofilms are highly resistant to antibiotics, leading to chronic infections. New approaches targeting biofilm formation (antivirulence) are urgently needed.

Objectives:

To evaluate the antibiofilm efficacy of repurposed small-molecule drugs against Staphylococcus aureus and Pseudomonas aeruginosa, and to investigate their mechanisms of action via in vitro assays and in silico modeling.

Methods:

Twenty-nine compounds were screened for antibacterial activity (MIC) and for biofilm inhibition using microtiter plate assays. Seven lead compounds were identified and further tested for biofilm prevention and disruption. Computational studies included molecular docking of leads to key targets (e.g. P. aeruginosa LasR quorum-sensing receptor, S. aureus sortase A), 100 ns molecular dynamics simulations, and ADMET property prediction.

Results:

The lead compounds showed potent, dose-dependent inhibition of P. aeruginosa biofilm formation (approximately 49–87% at 250 µg/mL, and 71–100% at 500 µg/mL) without affecting planktonic growth (MIC >250 µg/mL). Two compounds (KAUS-31, KAUS-38) completely prevented P. aeruginosa biofilm formation at 500 µg/mL. Docking suggested high-affinity binding of the leads to LasR (GlideScores up to –12.6 vs –8.4 for native autoinducer), supporting a quorum-sensing inhibition mechanism. MD simulations indicated the lead–LasR complexes were less stable than the native ligand. All seven compounds showed drug-like profiles (no Lipinski rule violations), high predicted oral absorption, and no serious in silico toxicity flags.

Conclusion:

Repurposed small molecules can serve as effective antibiofilm agents against S. aureus and P. aeruginosa. These compounds likely antagonize bacterial quorum sensing to suppress biofilms, while exhibiting favorable pharmacokinetic properties, highlighting a promising antivirulence strategy against biofilm-associated infections.

INTRODUCTION

Bacterial biofilms are structured communities of microorganisms attached to surfaces and encased in a self-produced extracellular polymeric matrix, which provides a protected niche for the bacteria.[1,2] In this sessile mode, cells exhibit an extreme tolerance to antibiotics and host immune defenses, leading to persistent infections that are notoriously difficult to eradicate.[1,3] For instance, bacteria within biofilms can withstand antibiotic concentrations up to 10²–10³-fold higher than those lethal to planktonic cells.[2] Indeed, biofilms are implicated in approximately 65–80% of chronic and recurrent microbial infections in humans, underscoring their tremendous clinical significance.[2,4]

Conventional antibiotics, developed to eliminate free-floating (planktonic) bacteria, often fail to eradicate biofilm-based infections.[1,5,6] To date, there are no approved therapeutic agents that specifically target or disperse bacterial biofilms.[7,8] As a result, clinicians must frequently resort to aggressive measures—such as surgical debridement or removal of infected implants, alongside prolonged antibiotic therapy to manage biofilm-associated infections.[4,9] Even then, relapses are common due to the recalcitrance of biofilm infections.[4,9] This lack of effective interventions highlights an urgent need for novel antibiofilm treatment strategies.

One emerging approach to combat biofilm infections is to disarm pathogens by blocking their biofilm formation processes (an antivirulence strategy) rather than attempting to kill bacteria outright.[4,5,10] By interfering with biofilm-specific functions—such as initial surface adhesion, extracellular polymer production, or quorum sensing, these strategies aim to prevent biofilm development and thereby render bacteria more susceptible to treatment.[10,11,12] Quorum sensing (QS), a cell-density-dependent communication system, is a particularly attractive target, as it globally regulates biofilm maturation and the expression of bacterial virulence factors.[13,14] Accordingly, quorum-sensing inhibitors (QSIs) have been shown to significantly attenuate biofilm formation by silencing bacterial communication signals.[12,13,15] Notably, because QSIs and other antivirulence agents do not directly kill the microbes, they impose less selective pressure for resistance and can even synergize with conventional antibiotics.[12,15,16]

Another promising strategy for discovering antibiofilm agents is drug repurposing, screening existing approved drugs for new antibiofilm activities.[17,18] Repurposing offers a faster and more cost-effective route to therapy, since candidate compounds often have well-characterized pharmacological and safety profiles that can expedite development.[19] Moreover, numerous approved drugs have unrecognized antibiofilm or QS-inhibitory properties that can be leveraged. For example, the antifungal ciclopirox and the antiviral ribavirin (neither a traditional antibiotic) were recently found to inhibit Pseudomonas aeruginosa QS signaling and biofilm formation without affecting planktonic growth.[17,19] Building on this concept, the present study integrated in vitro biofilm assays with in silico molecular modeling to identify and characterize novel antibiofilm agents from a library of pharmacologically active compounds.[20,21] The overall goal was to discover repurposed molecules that effectively disrupt biofilms—potentially by targeting QS or other key pathways—thereby addressing the critical need for new strategies against biofilm-associated infections.

MATERIALS AND METHODS

Chemicals and compound library

All chemical reagents and culture media were of analytical grade and used as received. An in-house library of 29 small-molecule compounds (coded KAUS-1 to KAUS-55, CAB-7 to CAB-29, etc.) previously synthesized by one of the authors was utilized for screening. These compounds were originally designed for other biological targets and were repurposed here for antimicrobial and antibiofilm evaluation. Each compound’s identity and purity were confirmed by spectroscopic methods (NMR, MS) before use. Stock solutions (10–20 mM) were prepared in DMSO and stored at –20 °C, with working dilutions made fresh in growth medium for biological assays. Ciprofloxacin (a broad-spectrum antibiotic) was included as a positive control reference in antibacterial and biofilm assays, whereas other antibiotics or antivirulence agents were used as appropriate comparators.

Bacterial strains and culture conditions

The test organisms used in this study were Staphylococcus aureus (Gram-positive) and Pseudomonas aeruginosa (Gram-negative). S. aureus (MRSA clinical isolate, Microbiology Lab, Faculty of Pharmacy, KAU) and P. aeruginosa PAO1 strain (a standard laboratory biofilm-forming strain) were maintained on nutrient agar slants. For experiments, bacteria were grown in Mueller–Hinton broth (MHB) or tryptic soy broth (TSB) at 37 °C with aeration. For biofilm assays, TSB was supplemented with 1% glucose to enhance biofilm formation.[22]

Antimicrobial susceptibility testing (MIC determination)

The minimum inhibitory concentrations (MICs) of the compounds against S. aureus and P. aeruginosa were determined by the broth microdilution method following CLSI/EUCAST guidelines.[23] Briefly, sterile 96-well microtiter plates were used to prepare two-fold serial dilutions of each test compound in cation-adjusted MHB. Bacterial suspensions were added to each well at a final inoculum of ~5 × 10⁵ CFU/mL.[23] After incubation for 18–20 h at 37 °C, wells were examined for visible bacterial growth. The MIC was defined as the lowest compound concentration that completely inhibited visible growth of the organism. Ciprofloxacin was included as a positive control (reference MIC: 0.25 µg/mL for S. aureus, 0.5 µg/mL for P. aeruginosa), and a DMSO-only well served as a negative (growth) control. All MIC determinations were performed in triplicate independent experiments.

Biofilm inhibition assay

The antibiofilm activity of the compounds was evaluated by a microtiter crystal violet (CV) staining assay.[22] Overnight bacterial cultures were adjusted to ~1 × 10⁸ CFU/mL and diluted 1:100 in TSB + 1% glucose. Aliquots of 200 µL were added to 96-well flat-bottom polystyrene plates containing test compounds at various concentrations (ranging from 0.25 × MIC up to 2 × MIC). Controls included wells with bacteria but no compound (negative control for biofilm formation) and wells with sterile medium (background control). Plates were incubated under static conditions at 37 °C for 24 h to allow biofilm development in the presence of compounds. After incubation, planktonic supernatants were gently aspirated, and each well was washed twice with sterile PBS to remove loose cells. The remaining attached biofilms were fixed with methanol (15 min), then stained with 0.4% (w/v) crystal violet (200 µL per well) for 20 min.[22] Excess stains were rinsed off by washing with distilled water, and the plates were air-dried. The bound CV dye was then solubilized with 33% acetic acid (200 µL), and biofilm biomass was quantified by measuring the optical density at 570 nm using a microplate reader.[22] The percentage inhibition of biofilm formation was calculated by comparing the OD₅₇₀ of treated wells to that of the untreated control biofilm (taken as 0% inhibition). Compounds that inhibited ≥50% of biofilm formation were classified as active hits for antibiofilm activity.[24] All biofilm assays were performed in triplicate wells and repeated in three independent experiments. Results were expressed as mean percent biofilm inhibition ± standard deviation.

Biofilm disruption assay on pre-formed biofilms

To assess the ability of compounds to disrupt established biofilms, a biofilm disruption (eradication) assay was conducted.[22] Biofilms of S. aureus and P. aeruginosa were first allowed to form in 96-well plates by incubating bacterial suspensions (prepared as above in TSB + 1% glucose) for 24 h at 37 °C. After biofilm maturation, the medium and planktonic cells were carefully removed by aspiration. Wells were gently washed twice with PBS to remove non-adherent bacteria. Fresh TSB (200 µL) containing test compounds at concentrations equal to the MIC, 2 × MIC, and 4 × MIC was then added to the pre-formed biofilms.[22] The plates were incubated for an additional 24 h at 37 °C to allow the compounds to act on the established biofilm matrix. Thereafter, the medium was removed, and biofilm biomass was quantified by the CV staining procedure as described above (fixation, staining, and OD₅₇₀ measurement). Percent biofilm reduction (disruption) was calculated relative to untreated biofilm controls. Any compound achieving ≥50% reduction of pre-formed biofilm was noted as a potential biofilm-disrupting hit. All experiments were done in triplicate, and observations were confirmed in two separate runs.

Computational studies

Target selection and preparation

In order to elucidate possible mechanisms of action, in silico studies were performed on molecular targets relevant to bacterial virulence and biofilm formation. Four protein targets were selected: (1) Staphylococcus aureus Sortase A (SrtA), a membrane-anchored transpeptidase responsible for anchoring surface proteins essential for Gram-positive bacterial adhesion and biofilm development; (2) Pseudomonas aeruginosa LasR, a quorum-sensing receptor and transcriptional activator that regulates virulence factor production and biofilm formation in Gram-negative bacteria[25]; (3) LuxS (S-ribosylhomocysteine lyase), an enzyme involved in AI-2 autoinducer synthesis for interspecies quorum sensing, which has been implicated in modulating biofilm formation and virulence in various bacteria[26]; and (4) S. aureus ClpP protease, a caseinolytic protease whose activity influences stress responses and biofilm formation (via the Agr system) as well as overall virulence in staphylococci.[27] High-resolution crystal structures for each target were obtained from the Protein Data Bank: SrtA (PDB ID: 1T2P), LasR ligand-binding domain (PDB ID: 3IX3), LuxS (PDB ID: 1IE0), and S. aureus ClpP (PDB ID: 3V5E). Each protein structure was prepared using the Schrödinger Protein Preparation Wizard (Maestro 2024-1) to add hydrogen atoms, assign proper bond orders, and optimize the orientation of hydroxyl and amide groups. Protonation states of ionizable residues were adjusted to pH 7.0 using PropKa, and a restrained energy minimization was performed (OPLS4 force field) to relieve any steric clashes while maintaining the heavy atom positions within 0.3 Å of the crystal coordinates.

Ligand preparation

All 29 compounds were prepared for docking using Schrödinger’s LigPrep module (Maestro, 2024) to generate low-energy 3D conformations. Ionization states were generated at pH 7.0 ± 2.0 (using Epik) to account for possible protonation under physiological conditions. For each compound, tautomers and stereoisomers (if applicable) were considered, and the lowest-energy structures were retained. The compounds were energy-minimized with the OPLS4 force field to obtain geometrically optimized ligand structures for docking.

Molecular docking and induced-fit refinement

Initial molecular docking was carried out using the Glide tool (Schrödinger 2025-1) in Extra Precision (XP) mode.[28,29] The prepared protein structures were defined with a grid centered on the active site or known ligand-binding pocket (for LasR, the binding site of the autoinducer OH-LAS was used; for SrtA, the catalytic site; for LuxS, the active site cavity; and for ClpP, the substrate-binding chamber). Standard Glide XP docking was performed for all 29 compounds against each target, with flexible ligand sampling and the van der Waals radii of nonpolar receptor atoms scaled by 0.8 (to accommodate ligand fit). Top-scoring poses were recorded based on the GlideScore. To account for receptor flexibility upon ligand binding, an induced-fit docking (IFD) protocol was applied for the top hits on each target. The IFD methodology iteratively combines rigid receptor docking (Glide) with protein structure refinement in Prime.[30] In our protocol, the highest-ranked pose of each ligand from Glide XP was used to initiate Prime refinement of nearby protein residues (within 5 Å of the ligand). The ligand was allowed to flex and re-dock into the induced-fit protein conformations, and the resulting complexes were scored by the IFD score (combining Prime energy and GlideScore).[30] This procedure enables modeling of side-chain rearrangements and other conformational changes in the binding site induced by ligand binding. The most favorable binding poses for the ach ligand–target complex were selected based on the lowest docking energy and visual inspection of key interactions. All protein–ligand interactions (hydrogen bonds, hydrophobic contacts, π–π stacking, etc.) were analyzed in Maestro. Docking results were summarized as XP GlideScores for rank-ordering the compounds on each target.

Molecular Dynamics (MD) simulations

To further examine the stability of the ligand–target interactions, molecular dynamics simulations were performed on the top selected complexes. The system setup was done with Desmond (Schrödinger Release 2025-1), using the OPLS4 force field for both protein and ligand parameters. Each protein-ligand complex was embedded in an orthorhombic box of TIP3P water molecules extending at least 10 Å from the protein, and neutralizing counterions (Na⁺ or Cl^–) were added to achieve 0.15 M NaCl physiological ionic strength. After a brief minimization, each system was equilibrated using Desmond’s default relaxation protocol (Brownian dynamics at 10 K, then NVT and NPT ensemble heating to 300 K) followed by a production run. Production MD simulations were carried out for 100 ns for each complex under NPT conditions (300 K, 1 atm) with a 2-fs time step, using the Martyna-Tuckerman-Klein barostat and Nosé-Hoover thermostat. Periodic boundary conditions and the Particle Mesh Ewald method for long-range electrostatics were applied. Trajectory snapshots were saved every 50 ps. The stability of ligand binding was evaluated by monitoring root-mean-square deviations (RMSD) of the protein backbone and ligand heavy atoms, intermolecular hydrogen bonds, and the ligand root-mean-square fluctuation (RMSF) over time. The MD simulation results confirmed whether the docked poses remained stable and helped identify any significant conformational changes in the protein or ligand. All MD simulations were conducted using the Desmond MD engine (D. E. Shaw Research) integrated in Maestro.[31]

In Silico ADMET profiling

Drug-likeness and ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) properties of the compounds were predicted using QikProp (Schrödinger Release 2025-1). QikProp generates physically significant descriptors and pharmaceutically relevant properties for organic molecules and compares them to known ranges for 95% of approved drugs.[32] For each compound, key properties such as molecular weight, QPlog Po/w (predicted octanol/water partition coefficient), QPlog S (aqueous solubility), QPlog BB (brain/blood partition coefficient), percentage human oral absorption, number of likely metabolic sites, and rule-of-five compliance were computed.[32] The compounds were evaluated against drug-likeness filters: for example, molecular weight between 150 and 500, no more than 5 hydrogen bond donors and 10 acceptors, and rotatable bond count within an acceptable range. Toxicity risks were assessed using in silico predictors for mutagenicity, hERG inhibition, and oral rat LD_50 (where available in QikProp’s database). The QikProp results aided in prioritizing the hit compounds by identifying those with favorable ADMET profiles (e.g., good permeability, low toxicity) alongside potent biofilm inhibitory activity.

Statistical analysis

All experimental data are presented as the mean ± standard deviation (SD). GraphPad Prism 9.0 (GraphPad Software, San Diego, CA) was used for data analysis. For MIC assays, reproducibility was confirmed across independent experiments without further statistical comparison. For biofilm inhibition and disruption assays, one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was employed to compare treated groups to untreated control biofilm. The value of P < 0.05 was considered statistically significant. All assays were performed in at least three independent experiments (biological replicates), each with triplicate technical replicates per treatment.

RESULTS

Antibacterial activity of lead compounds

The compounds exhibited no significant bactericidal activity against the tested planktonic bacteria. For the great majority of compounds, the MIC was above the highest concentration tested (250 µg/mL), indicating an absence of growth inhibition at these levels. Notably, a few compounds showed modest antimicrobial effects: for example, CAB-15 inhibited S. aureus, E. coli, and C. albicans at 125 µg/mL, and CAB-29 had an MIC of 125 µg/mL against S. aureus (while >250 µg/mL against P. aeruginosa). All other compounds, including the seven lead compounds identified for antibiofilm activity, did not exhibit significant direct antibacterial effects (MIC >250 µg/mL in planktonic culture). This indicates that the antibiofilm efficacy of the leads is not due to overt growth suppression at the concentrations tested. The minimum inhibitory concentrations (MICs) of KAUS-31, KAUS-38, KAUS-1, KAUS-5, NPK-208, SIN-OX15, and KAUS-17 were all >250 µg/mL (the highest concentration tested). This indicates that these compounds do not markedly inhibit P. aeruginosa growth at concentrations up to 250 µg/mL, suggesting their anti-biofilm efficacy is not due to bactericidal or bacteriostatic effects at those doses.

Biofilm inhibition assay

In contrast to their minimal planktonic effects, all seven lead compounds demonstrated potent, dose-dependent inhibition of P. aeruginosa PAO1 biofilm formation. The biofilm inhibition percentages at 125, 250, and 500 µg/mL for each lead compound. At the lowest tested concentration (125 µg/mL), the compounds inhibited biofilm formation by approximately 20–65%. Notably, KAUS-31 and KAUS-38 already showed high efficacy at 125 µg/mL (both around 62% inhibition), whereas SIN-OX15 was less active at this dose (only 20% inhibition). Increasing concentration to 250 µg/mL markedly improved the antibiofilm activity of all leads, with inhibition rates rising to between 49% and 87%. KAUS-31 and KAUS-38 remained the most potent (87% inhibition at 250 µg/mL), and even the least active lead at this concentration (KAUS-17) achieved nearly 50% biofilm inhibition. At the highest concentration of 500 µg/mL, all seven compounds exhibited strong antibiofilm effects, ranging from 71% to 100% inhibition. KAUS-31 and KAUS-38 completely prevented biofilm formation at 500 µg/mL (100% inhibition), while the other lead compounds also achieved high inhibition values (KAUS-1: 89%, KAUS-5: 83%, NPK-208: 79%, SIN-OX15: 77%, and KAUS-17: 71%). These results confirm that each of the identified lead compounds is capable of significantly reducing P. aeruginosa biofilm development in a concentration-dependent manner, with several compounds (especially KAUS-31 and KAUS-38) showing profound antibiofilm efficacy at higher doses.

Molecular docking results

All seven lead compounds exhibited favorable binding to the LasR ligand-binding domain in molecular docking simulations. NPK-208 stood out with the most negative GlideScore (–12.64), indicating the strongest predicted affinity (more favorable than the native autoinducer OdDHL at –8.41). The next top-scoring leads—KAUS-1 and SIN-OX15—had docking scores of approximately –9.6 and –9.1, respectively. KAUS-31 and KAUS-38 followed (around –8.8 to –8.5), slightly outperforming OdDHL in silico. KAUS-17 and KAUS-5 showed somewhat lower docking scores (both about –8.0), comparable to the OdDHL reference. Notably, the rankings by docking score did not perfectly correlate with binding free energy estimates. For example, SIN-OX15 yielded a relatively favorable MM-GBSA ∆G_bind (–58.45 kcal/mol, approaching that of OdDHL at –104.4 kcal/mol), whereas NPK-208’s ∆G_bind was much weaker (–5.28 kcal/mol) despite its excellent docking score. This suggests NPK-208’s top-scoring pose, while geometrically optimal, may be thermodynamically less stable than other complexes.

In-depth analysis of the docking poses revealed that all lead candidates engage the LasR binding pocket through a combination of hydrogen bonding and hydrophobic interactions characteristic of LasR’s native ligand. Figure 1 illustrates the docked conformation of NPK-208, the highest-ranking ligand. In the 2D interaction diagram [Figure 1a], NPK-208 is shown forming multiple hydrogen bonds with key polar residues Asp73, Ser129, Tyr64, and Trp60 (purple H-bond arrows) and a π–π stacking interaction with Tyr47 (green highlight) in the LasR binding site. The corresponding 3D view [Figure 1b] confirms that NPK-208 (green sticks) is positioned deep in the pocket, establishing H-bonds (dashed cyan lines) to Ser129 and Tyr64 as well as to Arg61, and π–π stacking with Trp60 and Tyr47. This extensive interaction network suggests NPK-208 can span a large portion of the binding site. By comparison, the native autoinducer OdDHL binds in a similar region of the LBD with the expected interaction pattern. The OdDHL docked pose shows its homoserine lactone moiety forming strong hydrogen bonds with Asp73, Trp60, Ser129, and Tyr56 (mimicking the crystal structure binding mode), and its acyl tail nestled in a hydrophobic pocket lined by residues such as Ile52, Leu36, Ala105, and Phe101. Notably, a water-bridged interaction to Arg61 is also observed for OdDHL (consistent with crystallographic data), underscoring that our docking reproduced the native ligand’s binding motif. Another lead compound, KAUS-17, showed a docked pose that bridges both the polar and nonpolar subsites of the LasR pocket. The 2D diagram for KAUS-17 indicates hydrogen bonds to Asp73 and Thr75, including a bridging water molecule interaction with Tyr56, along with extensive hydrophobic contacts with residues Trp60, Trp88, and Tyr93. In the 3D view, KAUS-17 is seen forming a direct H-bond with Asp73 and a network of interactions via Thr75 (and a nearby water) and π–π stacking with Trp88, while being surrounded by hydrophobic side chains (Trp60, Tyr93, Phe101) that help lock it in place. These docking results suggest that the selected repurposed compounds bind in the LasR LBD in a fashion analogous to the native signal molecule, leveraging the Asp73-centered H-bonding hotspot and the adjoining hydrophobic cavity.

Figure 1.

Docked binding pose of NPK-208 in the LasR receptor. (a) 2D schematic of NPK-208 interactions with LasR, highlighting hydrogen bonds to Asp73, Ser129, Tyr64, Trp60 (purple arrows), and a π–π stacking interaction with Tyr47 (green). (b) 3D view of the NPK-208–LasR complex (green = NPK-208), showing key residues (gray sticks) forming H-bonds with NPK-208 (dashed cyan lines to Ser129, Tyr64, Arg61) and π–π stacking with Trp60 and Tyr47

Molecular dynamics simulation

To verify the stability of the docked poses, molecular dynamics (MD) simulations (100 ns) were performed for selected LasR–ligand complexes (NPK-208 and KAUS-38) in comparison to the LasR–OdDHL native complex. The Cα-backbone of LasR remained stable throughout all simulations, with an RMSD fluctuating around 1.3–1.5 Å, indicating that binding of these ligands did not perturb the overall protein fold. However, clear differences emerged in ligand stability. NPK-208, despite its high docking score, showed a progressive increase in ligand RMSD from ~1 Å in the first few nanoseconds to ~4–5 Å by 100 ns [Figure 2a], signaling a gradual loss of its tight initial binding conformation. In contrast, the native OdDHL ligand remained firmly bound with minimal movement (ligand RMSD ~1–1.5 Å) over the course of 100 ns [Figure 3a], reflecting the high stability of the natural complex. KAUS-38 displayed intermediate behavior: its RMSD initially stayed around 1–2 Å but later fluctuated in the 3–4 Å range, never settling into a single rigid pose in the binding site. These RMSD results suggest that NPK-208 and especially KAUS-38 have a tendency to partially dissociate or rearrange in the binding pocket over time, whereas OdDHL remains stably engaged. Analysis of MD-derived interaction frequencies provided further insight into each ligand’s binding mode. For NPK-208, the simulation revealed that two critical hydrogen bonds persisted with high occupancy: Thr75 (80% of simulation time) and Tyr56 (74%). NPK-208 also maintained a π–π stacking interaction with Trp88 for a significant duration (~63% occupancy). However, its contact with Ser129 was sporadic (H-bond occupancy ~38%), indicating that NPK-208 did not consistently engage this key residue on the β-sheet side of the pocket. This incomplete H-bond network correlates with the ligand’s increasing RMSD and suggests why NPK-208’s binding loosened during MD. In stark contrast, the native OdDHL maintained a nearly continuous hydrogen-bond network throughout the simulation: Asp73 (100% occupancy), Trp60 (97%), and Ser129 (75%) all stayed engaged, with a weaker H-bond from Tyr56 also present ~39% of the time [Figure 2b]. OdDHL’s ability to simultaneously satisfy multiple anchor points (including the water-bridged Asp73–Arg61 interaction) explains its unwavering stability in the pocket. Meanwhile, KAUS-38 proved to have the weakest interaction profile of the group. As shown in Figure 3b, KAUS-38 failed to form any persistent direct H-bond with Asp73 or Ser129 (0% stable occupancy); only a water-mediated H-bond to Thr75 was observed (~47% occupancy). Few hydrophobic contacts with surrounding residues were noted for KAUS-38, and those that did form were transient. This paucity of strong interactions is consistent with KAUS-38’s high ligand RMSD and indicates that it binds LasR only weakly and fleetingly during the simulation. Overall, the MD results demonstrate that the docking poses of the top leads vary in dynamic stability, with some compounds (like NPK-208 and especially KAUS-38) exhibiting partial unbinding, whereas the native ligand remains tightly bound, mirroring experimental expectations.

Figure 2.

MD simulation of the LasR–NPK-208 complex over 100 ns. (a) Root-mean-square deviation (RMSD) of the protein Cα atoms (blue) and the NPK-208 ligand (orange) vs. time. The protein remains stable (~1.5 Å RMSD), while NPK-208’s RMSD rises to ~4–5 Å by 100 ns, indicating the ligand’s binding conformation becomes less stable over time. (b) 2D interaction frequency diagram from the 100 ns trajectory, showing NPK-208’s key contacts: a persistent H-bond with Thr75 (80% occupancy) and Tyr56 (74%), a moderate H-bond with Ser129 (38%), and a π–π stacking interaction with Trp88 (63% occupancy). The lower consistency of the Ser129 interaction correlates with NPK-208’s tendency to drift in the binding site

Figure 3.

MD simulation of the LasR–OdDHL (native ligand) complex. (a) RMSD plots for protein (blue) and OdDHL (orange) over 100 ns, showing a stable trajectory (protein ~1.3 Å, ligand ~1.0–1.5 Å) with minimal ligand displacement. (b) Interaction occupancy diagram for OdDHL during MD, indicating a robust hydrogen-bond network: Asp73 (100% occupancy), Trp60 (97%), Ser129 (75%), and a water-mediated bond to Arg61, with a lesser H-bond from Tyr56 (~39%). These persistent interactions explain the native ligand’s tight, stable binding throughout the simulation

ADMET prediction

In silico ADMET profiling of the lead antibiofilm compounds was performed using QikProp. All seven compounds exhibit drug-like physicochemical properties within recommended ranges. Molecular weights range from ~226 to 429 (well within the 130–725 range), and all compounds have acceptable hydrogen-bond donor (0–2) and acceptor counts (3–9), with zero violations of Lipinski’s “Rule of 5” for drug-likeness (each compound had 0 rule-of-five violations). The predicted CNS activity scores were mostly –2 (indicating likely inactive in CNS), suggesting low propensity to cross the blood-brain barrier—consistent with their calculated logBB values (QPlogBB ~–1.3 to –0.2), which are in the range of poor brain penetration. Human oral absorption predictions were high for all leads: each compound received a Human Oral Absorption category of 3 (high), with estimated oral absorption percentages between ~65% and 100%. Notably, KAUS-31, KAUS-38, KAUS-55, and SIN-OX15 showed excellent predicted permeability (QPPCaco values >1000 nm/s), indicating they are likely to be well absorbed in the intestine. By contrast, KAUS-1 and KAUS-17 had lower but still moderate QPPCaco values (~35–42 nm/s), suggesting these might have comparatively lower intestinal permeability. All compounds are predicted to be non-HERG-liable or only borderline: the predicted log HERG inhibition values (QPlogHERG) for KAUS-1 and KAUS-17 are –3.14 and –3.96, respectively (well above the concern threshold of –5), while some others (KAUS-31, KAUS-38, KAUS-55, NPK-208, SIN-OX15) have more negative QPlogHERG between –5.1 and –6.5. The latter values slightly exceed the concern threshold, indicating a potential risk of hERG interaction that may require attention in future optimization. Importantly, all leads show favorable solubility predictions (log S in the range –2.7 to –5.8, within acceptable limits) and low predicted skin permeability (QPlogKp ~–3.9 to –0.8). They also have low numbers of likely metabolic reactions (#metabolite sites 1–3, except SIN-OX15 with 0), suggesting a reasonable metabolic stability profile. Overall, the ADMET predictions support the potential of these compounds as orally bioavailable, non-toxic drug candidates, with no major red flags aside from slightly elevated hERG scores for a few compounds.

DISCUSSION

Antibiofilm efficacy of repurposed lead compounds

All seven lead compounds identified in this study displayed potent, dose-dependent inhibition of P. aeruginosa PAO1 biofilm formation, despite having no significant antibacterial activity against planktonic cells at the tested concentrations. This dichotomy—strong antibiofilm efficacy coupled with minimal bactericidal effect—suggests that these small molecules interfere with biofilm-specific pathways (such as quorum sensing or extracellular matrix formation) rather than exerting a traditional antibiotic mechanism. Notably, at 500 µg/mL all leads reduced biofilm biomass by at least ~70%, with two compounds (KAUS-31 and KAUS-38) completely preventing biofilm formation (100% inhibition). Even at the lowest concentration tested (125 µg/mL), several compounds (e.g. KAUS-31, KAUS-38) achieved substantial biofilm suppression (~60% inhibition), whereas others were less active at that dose (e.g. SIN-OX15 ~20% inhibition). These results confirm that each lead compound can significantly impede P. aeruginosa biofilm development in a concentration-dependent manner, and importantly, that this antibiofilm effect is not attributable to growth inhibition or killing of the bacteria in their planktonic state.

The ability to suppress biofilm formation without affecting planktonic growth is a hallmark of quorum-sensing inhibitors (QSIs) and other antivirulence agents.[33,34] By targeting regulatory processes underlying biofilm development rather than bacterial viability, such compounds impose less selective pressure for resistance and can even synergize with conventional antibiotics. Indeed, previous studies have shown that certain non-antibiotic drugs can be repurposed as QSIs—for example, the antifungal ciclopirox and the antiviral ribavirin were found to inhibit P. aeruginosa quorum signaling and biofilm formation without affecting planktonic growth. Our findings align with this paradigm: the repurposed molecules KAUS-1, KAUS-5, KAUS-17, KAUS-31, KAUS-38, NPK-208, and SIN-OX15 represent novel antibiofilm agents that effectively disarm P. aeruginosa biofilms without killing the bacteria. To our knowledge, this is the first report of these particular compounds being applied to attenuate P. aeruginosa biofilms, highlighting the potential of drug repurposing to uncover unconventional anti-infective strategies. Their novel activity against P. aeruginosa biofilms expands the arsenal of QSIs and anti-biofilm compounds in the literature and underscores the advantage of an antivirulence approach in managing biofilm-associated infections.

Quorum sensing inhibition and mechanistic insights

To probe the mechanism of action, we employed molecular docking and dynamics simulations focusing on LasR, the key P. aeruginosa QS receptor that regulates biofilm maturation. Strikingly, all seven leads showed high predicted affinity for the LasR ligand-binding domain, with Glide docking scores comparable to or better than that of the native autoinducer N-(3-oxo-dodecanoyl) homoserine lactone (OdDHL). In particular, NPK-208 stood out with an exceptionally favorable docking score (~–12.64), substantially more negative (i.e., more binding-favorable) than OdDHL (–8.41). Several others (e.g. KAUS-1 and SIN-OX15) also docked tightly (scores –9.1 to –9.5), and even the lowest-scoring lead (KAUS-17) had a docking score (–8.0) on par with the native ligand. These in silico results suggest that the repurposed inhibitors can snugly occupy the LasR binding pocket, potentially competing with the natural signal and thereby silencing QS-controlled biofilm formation. The docking poses indicated that all candidates engage the critical LasR binding-site residues through hydrogen bonding and hydrophobic contacts in a manner analogous to the native signal molecule. This is consistent with a LasR-mediated antivirulence effect: by binding to LasR, the compounds likely function as QS antagonists, interrupting the quorum-sensing cascade that P. aeruginosa requires for robust biofilm development.

Despite the promising docking affinities, our simulations revealed notable differences in binding stability between the leads and the native ligand. NPK-208, for example, achieved the best docking score but showed a much less favorable binding free energy (MM-GBSA ∆G_bind ≈ –5.3 kcal/mol versus –104.4 kcal/mol for OdDHL), hinting that its docked pose, while geometrically optimal, might be thermodynamically less stable. Consistent with this, 100 ns molecular dynamics (MD) simulations indicated that NPK-208 does not remain as stably bound as OdDHL: the ligand’s RMSD within the LasR pocket increased to ~4–5 Å over the course of the simulation, whereas the native OdDHL stayed firmly in place with minimal displacement (~1 Å RMSD). Similarly, another top compound, KAUS-38, failed to settle into a single rigid conformation in the binding site (ligand RMSD fluctuating between ~1 Å and 4 Å during MD). In contrast, the LasR–OdDHL complex showed unwavering stability, with OdDHL remaining deeply and consistently engaged in the binding pocket throughout the simulation (ligand RMSD ~1–1.5 Å). These observations suggest that while our lead compounds can bind LasR, they do so with lower dynamic stability than the native autoinducer, possibly resulting in a tendency to partially unbind or reposition over time.

Closer examination of the MD trajectories provides a structural rationale for this behavior. OdDHL maintained a nearly continuous hydrogen-bond network with key LasR residues (Asp73, Trp60, Ser129, and a water-bridged Arg61) throughout the simulation, fully exploiting the known anchor points in the LasR binding pocket. By contrast, the repurposed inhibitors engaged fewer of these critical interactions. For instance, KAUS-38 was unable to form any persistent hydrogen bond to Asp73 or Ser129 (0% occupancy for each) and showed only a transient water-mediated H-bond with Thr75. NPK-208, while forming H-bonds with LasR during docking, did not consistently retain the contact with Ser129 in MD (only ~38% occupancy) and relied mainly on interactions like Thr75 and Tyr56 that were not maintained 100% of the time. This incomplete interaction footprint likely explains the reduced binding stability of the compounds: without anchoring the same robust H-bond network as OdDHL, the inhibitors are more prone to diffuse within or slip out of the pocket over time. Nevertheless, the ability of these molecules to bind in the correct region of LasR and initiate key contacts (even if intermittently) supports their proposed role as QS antagonists. The dynamic behavior observed for NPK-208 and KAUS-38 also suggests an opportunity for optimization. Medicinal chemistry efforts aimed at strengthening specific ligand–LasR interactions (for example, introducing groups to better engage Asp73 or Ser129) could yield derivatives with improved LasR affinity and enhanced stability, translating to more potent and durable QS inhibition in vivo.

Drug-likeness and in silico ADMET profile

An important aspect of repurposing known compounds is evaluating their pharmacokinetic and toxicity profiles. All seven lead compounds exhibit favorable drug-like physicochemical properties according to in silico ADMET predictions. Their molecular weights (∼226–429 Da) and hydrogen bond counts fall within ranges typical for orally active drugs, and none of the compounds violated Lipinski’s Rule-of-Five criteria for drug-likeness. This suggests a low risk of development-limiting issues related to size or permeability. Consistently, the predicted human oral absorption was high for all leads: each was categorized as class 3 (high absorption), with estimated oral absorption percentages on the order of ~65% to essentially 100%. Several compounds (e.g. KAUS-31, KAUS-38, KAUS-55, SIN-OX15) had exceptionally strong calculated Caco-2 permeabilities (QPPCaco >1000 nm/s) indicative of excellent intestinal epithelial penetration. A couple of leads (KAUS-1, KAUS-17) showed relatively lower but still acceptable permeability predictions (tens of nm/s), suggesting moderate absorption that might be improvable. Notably, all compounds were predicted to have poor CNS penetration (CNS activity score of –2 and logBB ~–1), which is a desirable trait for anti-infectives that do not need to cross the blood–brain barrier. In practical terms, this means the compounds are likely to remain in peripheral circulation, focusing their activity on infections outside the central nervous system and possibly reducing CNS-related side effects.

Safety-related in silico parameters were also encouraging. Importantly, none of the lead candidates showed a strong propensity for hERG channel inhibition—a common predictor of cardiotoxicity. Two compounds (KAUS-1 and KAUS-17) had QPlogHERG values of –3.14 and –3.96, well above the concern threshold (around –5), indicating minimal risk of cardiotoxic effects. The other leads (KAUS-31, KAUS-38, KAUS-55, NPK-208, and SIN-OX15) had slightly more negative QPlogHERG scores (approximately –5.1 to –6.5) that marginally exceed the threshold. These values suggest a potential but minor risk of hERG interaction that merits attention in future optimization, but they are not alarmingly low; many existing drugs carry similar borderline hERG liabilities and can be managed with careful dose selection or structural tweaking. Additionally, all seven compounds showed acceptable predicted aqueous solubility (log S in the range –2.7 to –5.8) and low skin permeability (log Kp ~–3.9 to –0.8), indicating they are sufficiently soluble for formulation and unlikely to cause issues if administered systemically. They also had a low number of predicted metabolic routes (mostly 1–3 likely metabolic sites per molecule), implying a reasonable metabolic stability profile. Overall, the ADMET predictions support the candidacy of these leads as orally bioavailable and relatively safe drug prospects. Aside from the slight hERG-related flags (which can be addressed in follow-up studies), no major red flags were identified. This favorable ADMET outlook is a significant advantage of the repurposing approach—many of these molecules were originally designed for pharmacological activity, so it is perhaps unsurprising that they largely conform to drug-like standards. It bodes well for the feasibility of advancing one or more of these antibiofilm agents toward in vivo testing and further development.

CONCLUSION

This study demonstrates the feasibility of repurposing small molecules as potent inhibitors of S. aureus and P. aeruginosa biofilms. The seven lead compounds discovered are remarkable for suppressing biofilm formation in P. aeruginosa (up to 100% inhibition in vitro) despite having no significant bactericidal activity at effective concentrations. This novel profile, strong antibiofilm efficacy without killing planktonic cells, is characteristic of an antivirulence strategy, wherein pathogens are disarmed rather than destroyed. Mechanistically, the leads appear to interfere with bacterial quorum sensing, as supported by their high affinity for the LasR QS receptor in docking simulations. By occupying the LasR ligand-binding site with affinities comparable to or exceeding the native autoinducer, these compounds likely block QS signaling and thereby prevent biofilm maturation. Importantly, the repurposed inhibitors also exhibit drug-like physicochemical and ADMET properties, underscoring their pharmaceutical promise. Taken together, our findings introduce a set of novel antibiofilm agents with a unique mode of action that targets bacterial communication and biofilm formation. Such non-traditional anti-infectives have the potential to complement standard antibiotics, offering a new therapeutic angle for tackling recalcitrant biofilm-associated infections. The lead compounds identified here represent promising candidates for further development as antibiofilm drugs, highlighting the value of drug repurposing in addressing unmet needs in infection control. Moving forward, our next steps will include evaluating these lead compounds in vivo (e.g., in animal models of biofilm-related infection) to confirm their efficacy and safety in a complex host environment. We also plan to optimize the lead molecules through medicinal chemistry modifications of key functional groups to enhance antibiofilm potency or reduce any potential liabilities (such as the slight hERG binding signals noted in silico). Additionally, future studies will involve comprehensive cytotoxicity assessments on mammalian cell lines and combination therapy experiments (using the lead compounds alongside conventional antibiotics) to fully establish their therapeutic window and to exploit potential synergies. By pursuing these directions, we aim to translate our findings into viable antibiofilm treatment strategies and improve the clinical readiness of these repurposed compounds.

Conflicts of interest

There are no conflicts of interest.

Funding Statement

This research work was funded by the Institutional Fund Projects under grant no. (IFPIP: 363-166-1443). The authors gratefully acknowledge the technical and financial support provided by the Ministry of Education and King Abdulaziz University DSR, Jeddah, Saudi Arabia.