Computer-Aided Identification of Recognized Drugs as Pseudomonas aeruginosa Quorum-Sensing Inhibitors

Abstract

Attenuation of Pseudomonas aeruginosa virulence by the use of small-molecule quorum-sensing inhibitors (referred to as the antipathogenic drug principle) is likely to play a role in future treatment strategies for chronic infections. In this study, structure-based virtual screening was used in a search for putative quorum-sensing inhibitors from a database comprising approved drugs and natural compounds. The database was built from compounds which showed structural similarities to previously reported quorum-sensing inhibitors, the ligand of the P. aeruginosa quorum-sensing receptor LasR, and a quorum-sensing receptor agonist. Six top-ranking compounds, all recognized drugs, were identified and tested for quorum-sensing-inhibitory activity. Three compounds, salicylic acid, nifuroxazide, and chlorzoxazone, showed significant inhibition of quorum-sensing-regulated gene expression and related phenotypes in a dose-dependent manner. These results suggest that the identified compounds have the potential to be used as antipathogenic drugs. Furthermore, the results indicate that structure-based virtual screening is an efficient tool in the search for novel compounds to combat bacterial infections.

One of the major problems in the treatment of infectious diseases is the occurrence of antibiotic resistance, which is prevalent among many bacterial species, in particular, strains of the opportunistic pathogen Pseudomonas aeruginosa (39). This bacterium is a common gram-negative species found in nosocomial infections such as urinary tract infections, respiratory system infections, dermatitis, chronic wounds, soft-tissue infections, and a variety of systemic infections, particularly in patients with severe burns and in cancer and AIDS patients who are immunocompromised (43, 66). Furthermore, people who are suffering from the genetic disease cystic fibrosis are highly susceptible to chronic lung infection with this organism (16).

The resistance of P. aeruginosa to multiple antibiotics is the result of a variety of specific mechanisms that includes an inherent β-lactamase and a broad-spectrum efflux pump system. Furthermore, impermeability of the membrane, adaptive mutations, and horizontal transfer of resistance genes also contribute to its resistance (38). Besides these specific resistance mechanisms, the ability of P. aeruginosa to grow as structured communities of cells enclosed in a self-produced polymeric matrix, known as a biofilm (11), significantly adds to its tolerance to antimicrobial agents (12). Current antibiotics are susceptible to resistance development, as there will inevitably be selection pressure for bacteria able to grow in the presence of these growth-inhibiting compounds (39). Therefore, new approaches to combat microbes without selecting for resistance would hold great promise for the treatment of infectious diseases. An approach that does not target bacterial growth is the use of antipathogenic drugs that work by decreasing bacterial virulence and rendering bacteria incapable of establishing infection. An area where the use of antipathogenic drugs has received much attention is virulence attenuation by blocking bacterial intercellular communication, i.e., quorum sensing (QS) (recently reviewed in reference 52). In gram-negative bacteria, most QS systems are members of the LuxR-LuxI homologous system that use acyl homoserine lactone (AHL) signal molecules. These systems function by means of a LuxR homolog, the transcriptional activator, and a LuxI homolog, the AHL synthase. LuxI produces the required AHL molecule, which diffuses out into the local environment, and upon reaching the required concentration, the signal molecule binds to and activates LuxR, which in turn activates the transcription of the target genes (20, 59, 68). In P. aeruginosa, QS is mediated through the LuxRI-type systems LasRI and RhlRI, which sense 3-oxo-C12-homoserine lactone (HSL) and C4-HSL, respectively, and via the Pqs system, which senses Pqs (2-heptyl-3-hydroxy-4-quinolone). The systems are hierarchically arranged, with LasR regulating the Rhl and Pqs systems (35, 60). AHL-mediated QS coordinates the production of virulence factors and plays a role in biofilm formation (4, 15, 45). Furthermore, QS-controlled activities, as well as QS signals, evidently also affect the activities of host immune systems (34, 37, 53, 64). By using various QS inhibition assays and high-throughput screening, small antagonistic molecules displaying some structural similarities to AHL signals have been identified as QS inhibitors (QSIs) and shown to greatly reduce the virulence of P. aeruginosa both in vitro and in vivo (29, 42, 50).

Recently, computer-aided drug design, especially structure-based virtual screening (SB-VS), has emerged as a new tool in pharmaceutical chemistry (40, 54). The increasing availability of structural data and the affordability of high-performance computing platforms have broadened the applicability of this method (54). SB-VS has been adopted as an effective paradigm for lead discovery that fits in well alongside high-throughput screening programs. Screening has been successfully used to find inhibitors of various enzymes and proteins, including human carbonic anhydrase II, human protein tyrosine phosphatase 1B, and the omnipresent bacterial enzyme DNA gyrase, which in the latter case resulted in novel classes of inhibitors with potential for use as antibiotics (6, 17, 27). In addition to the discovery of novel antibiotics by virtual screening, the approach also holds great promise for the discovery of antipathogenic drugs and especially for the discovery of new QSIs. The structures of TraR from Agrobacterium tumefaciens and the ligand binding domain of LasR from P. aeruginosa bound to their natural ligands have become available recently and can be exploited in SB-VS (7, 67, 71). The availability of the LasR structure is particularly important, as it can be used to complement the traditional QSI discovery strategies previously mentioned.

In the present study, 147 recognized drugs and natural compounds were selected from the SuperNatural and SuperDrug databases (18, 26) on the basis of their two-dimensional (2D) structural similarity to the P. aeruginosa LasR natural ligand, the identified QSIs furanone C30 and patulin, or the QS agonist TP-1. The automated docking program Molegro Virtual Docker (MVD) was used to screen these selected compounds for QSI candidates. Six top-ranking drugs were acquired and tested for biological activity. Three compounds, salicylic acid, nifuroxazide, and chlorzoxazone, showed significant, dose-dependent inhibition of QS-regulated gene expression and related phenotypes.

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth media.

P. aeruginosa PAO1 (33), an isogenic P. aeruginosa lasR mutant (29), and an isogenic P. aeruginosa lasI rhlI mutant (29) were used in this study. PAO1 with a mini-Tn5 insert containing a translational fusion of the LasR-regulated lasB promoter and a gene coding for an unstable version of green fluorescent protein (GFP), gfp(ASV), and an extra copy of lasR under the influence of the lac promoter (constitutively expressed in P. aeruginosa) (28) was used for the LasR inhibition assay. For the RhlR inhibition assays, the plasmid pMHRA was introduced by electroporation into P. aeruginosa PAO1 and P. aeruginosa lasR. pMHRA contains an RlhR-regulated rhlA::gfp(ASV) translational fusion inserted into the vector pMH391 (28). The P. aeruginosa PAO1 strain and the lasR mutant used for the Pqs inhibition assay carried plasmid pAC37 containing a transcriptional fusion of the Pqs-regulated pqsA promoter and gfp(ASV) (70). Measurement of virulence factor production was carried out with the wild-type PAO1 strain and a QS-deficient PAO1 lasI rhlI mutant strain (29). GFP-tagged variants of these two strains were used for flow cell biofilm experiments. Tagging of the strains was done with mini-Tn7-enhanced GFP as previously described (36). Serratia liquefaciens MG1 (wild type) and an swrI (swarming-deficient) mutant (19) were used to assess the effect of the QSIs on the swarming motility of the bacterium.

For the QS inhibition and LasR specificity assays, ABT minimal medium (BT plus 10% A10) supplemented with 0.5% glucose (wt/vol) and 0.5% Casamino Acids (wt/vol) (28) was used to grow the cells. This medium supplemented with 0.6% Bacto agar (BD, NJ) was used in the S. liquefaciens swarming assay. Modified Luria-Bertani (3) medium containing 4 g NaCl liter⁻¹ was used as a basic medium to grow cells overnight (ON) prior to carrying out the swarming assay and the virulence factor assays and prior to inoculating the flow cell biofilm system. Virulence factor assays were carried out with cultures grown in modified BM2 medium (24) containing 0.5 mM Mg²⁺, 0.01 mM FeCl₃, 0.06 M glucose, and 0.2% (wt/vol) Casamino Acids. Flow cell biofilms were cultivated in FAB minimal medium (30) supplemented with 0.3 mM glucose. Selective media were supplemented with ampicillin (100 mg liter⁻¹), carbenicillin (500 mg liter⁻¹), gentamicin (60 mg liter⁻¹), or streptomycin (100 mg liter⁻¹) where appropriate.

Protein structure file and drug/natural ligand database.

The X-ray crystallographic structure of the P. aeruginosa LasR ligand binding domain bound to its natural ligand n-3-oxododecanoyl-l-homoserine lactone (OdDHL; Protein Data Bank [PDB] identification [ID]: 2UV0) (7) was used in the SB-VS of small molecules. Recognized drugs and natural compounds were selected from the SuperDrug and SuperNatural databases (18, 26) on the basis of their 2D structural similarity to the P. aeruginosa LasR ligand OdDHL (45), the identified QSIs furanone C30 and patulin (29, 51), or the LasR agonist TP-1 (42). The structures of the similarity templates are displayed in Fig. 1A. The 2D structures of the selected compounds were downloaded from Chembank (http://chembank.broad.harvard.edu/) (56) or drawn manually in MarvinSketch ver. 4.1.10 (ChemAxon Ltd., Hungary) and then saved as 2D coordinate structure-data file (SDF) files. The SDF files were merged into a single SDF file database by OpenBabel ver. 2.1.1 (OpenEye Scientific Software) and then converted to a three-dimensional (3D) structure database in CORINA ver. 3.4 (Molecular Networks GmbH, Germany).

FIG. 1.

2D structures of the compounds tested in this study. (A) Structures of the known LasR binding compounds used in the 2D similarity search against the SuperNatural and SuperDrug databases (OdDHL, TP-1, furanone C30, and patulin) and included in the SB-VS for reference (OdDHL) and comparison (all). (B) Structures of the six compounds tested for the ability to inhibit LasR in P. aeruginosa.

Molecular docking.

The automated docking software MVD ver. 2007.2.2.0 (65) (Molegro Aps., Denmark) was used in the SB-VS to dock the compounds from the created database. The LasR structure file contained four monomers of the ligand binding domain each in a complex with one OdDHL ligand, but only the E monomer in a complex with the ligand named OHN 1169 [E] was chosen for importation into the MVD workspace. Water was excluded from the workspace, and standard preparation of the molecules was used in the importation of the database.

Docking of the database was carried out in the following way. Initially, a template based on the present ligand was created to award poses of the docked compounds that showed similarity to the molecule. The template consisted of four individual criteria: steric, hydrogen donor, hydrogen acceptor, and ring. Subsequently, the actual docking procedure was set up by choosing template docking, the Ligand Evaluator scoring function, and the default search algorithm MolDock Optimizer. The binding site restraining the search space was sphere shaped and centered on the OdDHL ligand with a radius of 15 Å. Docking of each compound was repeated five times to ensure conformation to the lowest-energy state due to the iterative nature of the process. Five poses of each molecule in a complex with the protein were returned, one for each run, and these were finally ranked according to their rerank scores. For the specific settings for the docking process, see Fig. S1 and Table S1 in the supplemental material. In addition to the database of selected compounds, the docking process was supplemented with the LasR natural ligand structure for reference together with the structures of known LasR agonists and antagonists for comparison (Fig. 1A and Table 1). Scenes of molecules in complex with the LasR ligand binding domain were generated with PyMOL ver. 1.00 (DeLano Scientific LLC, San Carlos, CA; http://www.pymol.org) by importing the 3D poses of the docked molecules from MVD together with the structure of the LasR ligand binding domain (PDB ID: 2uV0).

TABLE 1.

Names and types of LasR binding compounds used as references and for comparison in the SB-VS

Compound	Type	Reference
3-Oxo-C12-HSL	Natural ligand	45
3-Oxo-C12-(2-aminophenol)	Inhibitor	62
4-Nitropyridine-N-oxide	Inhibitor	50
Furanone C30	Inhibitor	29
Patulin	Inhibitor	51
Penicillic acid	Inhibitor	51
TP-1	Activator	42
TP-5	Inhibitor	42

QSI candidates selected for assessment of QS-inhibitory activity.

The drugs selected on the basis of the docking results were tested for the ability to inhibit LasR-controlled lasB::gfp gene expression from P. aeruginosa. Salicylic acid, nifuroxazide, and chlorzoxazone were purchased from Sigma-Aldrich in their pure form, while indoramine (Wydora; Riemser Arzneimittel AG), tiaprofenic acid (Surgamyl; Patheon Ltd., United Kingdom), and donepezil (Aricept; Pfizer SA) were acquired as pills. Stock solutions or suspensions of the compounds were made by dissolving or suspending the compounds and pills in 96% ethanol. The pill solutions were filtered to remove the binding material, and sufficient content and purity of the active compounds were verified by high-performance liquid chromatography (HPLC)-mass spectrometry (MS).

QS inhibition assays.

The LasR inhibition assay was carried out by growing the P. aeruginosa lasB::gfp strain in a 96-well microtiter tray (black polystyrene; Nunc) together with serial twofold dilutions of the putative inhibitors as previously described with modifications (50). Briefly, growth medium and inhibitor stock solution were initially added to the first column of wells to give a final inhibitor concentration of 1,600 μg/ml in a volume of 300 μl. A volume of 150 μl of medium was added to each of the rest of the wells. Next, serial twofold dilutions of the inhibitors were made by transferring 150 μl of medium between the wells in a sequential manner. The final two columns of wells were kept without added inhibitor as a reference. Finally, 150 μl of cell culture with an optical density at 450 nm (OD₄₅₀) of 0.2 was added to the wells to give a final OD₄₅₀ of 0.1 and inhibitor concentrations ranging from 800 μg/ml to 1.5625 μg/ml in a volume of 300 μl. The added culture was prepared from an ON culture grown at 30°C by appropriate dilution with growth medium based on the measurement of OD₄₅₀. The OD of the ON culture was measured with a Thermo Scientific GENESYS 10 UV spectrophotometer. The microtiter tray was incubated at 34°C in a Wallac 1420 VICTOR2 plate reader (Perkin-Elmer, MA), and the instrument was set to measure GFP(ASV) expression by means of the protein's fluorescence at 535 nm upon excitation at 485 nm and the growth in the wells as the OD₄₅₀ every 15 min for at least 14 h. The Rhl and Pqs inhibition assays were carried out in essentially the same way as the LasR inhibition assay. However, instead of using serial twofold dilutions, salicylic acid, nifuroxazide, and chlorzoxazone were added to final concentrations of 200 μg/ml, 200 μg/ml, and 100 μg/ml, respectively.

LasR specificity assay.

The LasR specificity assay was carried out in the same way as the Rhl and Pqs inhibition assays. P. aeruginosa lasR harboring reporter plasmid pMHRA or pAC37 was grown ON with the required antibiotics to maintain the plasmids, while the assay was carried out without added antibiotics to avoid secondary effects from the possible interaction between the antibiotics and the added inhibitors.

Measurement of virulence factors.

P. aeruginosa PAO1 and the QS-deficient lasI rhlI strain were inoculated in modified BM2 medium to an OD₆₀₀ of 0.01 from ON cultures of the two strains. Inhibitors were added from stock solutions to a final concentration of 200 μg/ml for salicylic acid and nifuroxazide and to a final concentration of 100 μg/ml for chlorzoxazone. After 18 h of growth at 37°C, the OD₆₀₀ of the cultures was measured and the assays were carried out. The OD was measured with a Thermo Scientific GENESYS 10 UV spectrophotometer. Regular BM2 medium was used as a blank sample. Estimation of exogenous proteolytic activity was carried out as described by Hentzer et al. (28), with some modifications. Briefly, 150 μl of culture supernatant, centrifuged at 15,000 × g and 4°C for 15 min, was mixed with 250 μl of an azocasein solution (2% [wt/vol] in 2 mM CaCl₂ and 40 mM Tris-HCl [pH 7.8]; Sigma-Aldrich) and allowed to react for 45 min at 37°C. Undigested substrate was precipitated with 1.2 ml of trichloroacetic acid (10% [wt/vol]) for 15 min at room temperature and subsequently centrifuged at 15,000 × g for 10 min. Nine hundred microliters of the supernatant was transferred to new Eppendorf tubes and mixed with 750 μl of 1 M NaOH. Proteolytic activity was measured as the OD₄₄₀ against a blank sample run in parallel with the other samples and divided by the OD₆₀₀ of the culture to estimate the relative protease production. Pyoverdine production was measured according to a spectrophotometric method previously described by Höfte et al. (32), with modifications. Briefly, the concentration of pyoverdine in the culture supernatant was measured directly at 400 nm. Because the compound nifuroxazide also absorbed light at 400 nm, a sample containing medium and nifuroxazide was used as a control for measurement of the nifuroxazide-treated replicates. Relative pyoverdine production was estimated as the OD₄₀₀ divided by the OD₆₀₀ of the cultures. Rhamnolipid production was estimated by two modified biosurfactant production assays: a drop collapse assay (5) and an emulsification activity assay (10). For the drop collapse assay, 2-μl spots of mineral oil were added to the lid of a 96-well microtiter plate (Nunc). The lid was equilibrated for 1 h at room temperature, after which 5-μl drops of culture supernatant were added to the surface of the oil spots. The shape of the drops was inspected after 1 min. For the emulsification activity assay, 2 ml of n-hexadecane was added to 2 ml of culture supernatant. The mixture was vortexed at high speed for 2 min and allowed to stand for 6 h before inspection. Emulsification activity was defined as the height of the emulsion layer divided by the total height and expressed as a percentage.

Cultivation of biofilms in flow chambers.

Biofilms were grown for 4 days in flow chambers with individual channel dimensions of 1 by 4 by 40 mm. The flow system was assembled and prepared as described previously (63). Inoculation of the system was carried out by injecting 300 μl of an ON culture diluted to an OD₆₀₀ of 0.001 into each flow channel with a small syringe. After inoculation, the flow chambers were left upside down without flow for 1 h to allow bacterial attachment to the glass cover, after which medium flow was started with a Watson Marlow 205S peristaltic pump. The flow chambers were irrigated with medium with or without QSIs (salicylic acid at 20 μg ml⁻¹, nifuroxazide at 20 μg ml⁻¹, and chlorzoxazone at 10 μg ml⁻¹). The concentrations of the QSIs were reduced by a factor of 10 compared to those in the batch assays due to the continuous flowthrough of medium. The mean flow velocity in the flow chambers was 0.2 mm s⁻¹, corresponding to laminar flow with a Reynolds number of 0.02. The biofilms were grown at 30°C.

Microscopy and image acquisition.

All microscopy observations and image acquisitions were done with a Zeiss LSM510 confocal laser scanning microscope (CLSM) equipped with detectors and filter sets for monitoring of GFP. Images were obtained with a 40×/1.3 objective. Simulated 3D images and sections were generated with the Imaris software package (Bitplane AG).

COMSTAT image analysis.

CLSM images were analyzed by use of the computer program COMSTAT (30). Thresholds for the different image stacks were determined automatically, and connected volume filtration was used in the analysis.

S. liquefaciens swarming assay.

Swarming plates consisted of ABT minimal medium supplemented with glucose, Casamino Acids, and 0.6% Bacto agar. QSIs were mixed into the medium immediately before casting at concentrations of 200 μg/ml for salicylic acid and nifuroxazide and 100 μg/ml for chlorzoxazone. The plates were left for drying without lids in a fume hood for 1 h at room temperature. Five-microliter drops of 10-times-diluted ON cultures of S. liquefaciens MG1 and an swrI mutant were placed on the appropriate plates and incubated for 18 h at room temperature before pictures of each plate were taken. The areas of the swarming zones were calculated by use of the free ImageJ software. Measurements were done in triplicate. The viability of S. liquefaciens in the presence of the QSI compounds was assessed by monitoring growth at 30°C in a 96-well microtiter tray via OD₆₀₀ measurement. The concentrations of the QSI compounds were equal to those used in the swarming assay.

RESULTS

Virtual screening for LasR QSI candidates.

The initial 2D similarity search for drugs/natural compounds similar to the P. aeruginosa LasR natural ligand 3-oxo-C12-HSL, the proven QSIs furanone C30 and patulin, and the tri-phenyl LasR agonist TP-1 (Fig. 1A) in the publicly available (http://bioinformatics.charite.de/content/index.php) SuperNatural and SuperDrug databases (18, 26) resulted in 149 similar compounds. A 3D structural database of the compounds was subsequently docked against the ligand binding domain of LasR (PDB ID: 2UV0) in the docking program MVD. Furthermore, the screening included docking of the natural ligand itself and structures of known LasR antagonists and agonists (Fig. 1A and Table 1) for comparison. Six top-ranking drug/natural ligands (Fig. 1B) displaying docking scores better than that of the recognized QSI 4-nitropyridine-N-oxide (4-NPO), which was set as the cutoff for selection, were acquired to test their LasR-inhibitory potential. The docking scores of the six acquired compounds, together with those of the OdDHL reference and the known LasR binding compounds, are displayed in Table 2.

TABLE 2.

Rerank scores, similarity scores, and molecular weights of the acquired compounds, the reference ligand, and the known LasR binding compounds included for comparison^a

Compound	Mol wt	Similarity score	Rerank score
Acquired compounds
Indoramine	340.4	−491.29	−121.281
Nifuroxazide	274.2	−447.124	−86.3983
Tiaprofenic acid	258.3	−396.031	−71.4197
Donepezil	370.4	−499.284	−70.7016
Chlorzoxazone	169.6	−344.493	−69.6008
Salicylic acid	138.1	−336.595	−66.156
Reference, 3-oxo-C12-HSL	297.4		−129.739
Known LasR binders
3-Oxo-C12-(2-aminophenol)	305.4	−470.817	−93.8255
Patulin	154.1	−354.068	−70.1662
4-NPO	140.1	−305.698	−63.1366
Furanone C30	257.9	−265.816	−47.9107
Penicillic acid	170.2	−315.516	75.4073
TP-1	568.6	−497.462	220.864
TP-5	385.2	−398.532	431.434

^aThe compounds were ranked on the basis of their rerank scores. The scores of the individual compounds are the best of five scores derived from independent docking rounds.

Determination of LasR inhibition.

In order to determine the QS-inhibitory potential of the six acquired drugs, they were initially screened by means of a LasR inhibition assay based on a lasB::gfp(ASV) translational fusion in P. aeruginosa PAO1. lasB codes for the virulence factor elastase and has been shown to be under the transcriptional control of LasR (22). As shown in Fig. 2, three of the six compounds, salicylic acid, nifuroxazide, and chlorzoxazone, showed significant reductions in the number of relative fluorescence units (RFU) over time in a dose-dependent manner. This suggests that the compounds possess inhibitory activity against LasR and function as QSIs. Furthermore, the compounds did not inhibit growth at the concentrations used. These results are in accordance with the finding that QS deficiency in P. aeruginosa exerts no effect on the growth of the bacterium (25).

FIG. 2.

Dose-response curves of the three drugs salicylic acid (A), nifuroxazide (B), and chlorzoxazone (C) when incubated together with PAO1 lasB::gfp(ASV). The three compounds reduce the RFU level (GFP fluorescence units divided by OD₄₅₀) at various concentrations (right), indicating inhibition of LasR, while displaying no effect on growth (left). Growth is displayed as the increase in OD₄₅₀ compared to the initial level. Salicylic acid and nifuroxazide were tested at concentrations of 400 (▪), 200 (Δ), 50 (•), 12.5 (□), 3.125 (▴), and 0 (○) μg/ml, while chlorzoxazone was tested at 200 (▪), 100 (▵), 20 (•), 6.25 (□), 1.5625 (▴), and 0 (○) μg/ml. Results are representative of three independent experiments.

Inhibition of the Rhl and Pqs QS systems.

The activity of the three compounds salicylic acid, nifuroxazide, and chlorzoxazone against LasR-controlled lasB::gfp expression prompted us to investigate their inhibitory effects on the other major parts of the P. aeruginosa QS circuit, the Rhl and Pqs systems, which are both under LasR regulation (35, 60). Inhibition of the systems was investigated by assays similar to the LasR assay, with Rhl system inhibition being assessed by an rhlA::gfp(ASV) translational fusion and the Pqs inhibition being assessed by a pqsA::gfp(ASV) transcriptional fusion as described in Materials and Methods. rhlA is the first gene of the rhlAB operon that codes for a rhamnosyltransferase essential for the production of rhamnolipid, the major product controlled by the Rhl QS system (44), and pqsA is the first gene of the pqsABCDE operon that is required for production of the Pqs signal (21). The two assays should thus give a relevant indication of the levels of Rhl and Pqs QS system inhibition by the three identified active compounds. The compounds were tested at concentrations of 200 μg/ml for salicylic acid and nifuroxazide and 100 μg/ml for chlorzoxazone. These concentrations displayed the largest degree of LasR inhibition without inhibiting growth (Fig. 2). As Fig. 3A shows, treatment of the two reporter strains with the three compounds resulted in a reduction of RFU levels in all cases. The reductions were statistically significant at the α = 0.01 level (single-factor analysis of variance [ANOVA]). This added further evidence to the finding that the three identified compounds induce conditions of reduced QS activity in P. aeruginosa.

FIG. 3.

Expression of rhlA::gfp(ASV) (A) and pqsA::gfp(ASV) (B) in wild-type and lasR mutant P. aeruginosa PAO1 treated with the three identified LasR inhibitors. Results are average RFU values taken from a single time point measurement corresponding to maximal induction of the reporters in the late log phase of growth. Inhibitors were added at concentrations of 200 μg/ml for salicylic acid and nifuroxazide and 100 μg/ml for chlorzoxazone. Averages and SDs of eight replicates are shown.

The results, however, were inconclusive regarding the specificity of the compounds. It could not be inferred if inhibition of the Rhl and Pqs systems occurred solely as a result of LasR inhibition or if the two systems were also directly inhibited by the compounds. To investigate this further, the GFP reporter plasmids for the two systems were introduced into a P. aeruginosa lasR mutant. The fluorescence levels from these reporter strains should only relate to the activity of the Rhl and Pqs systems abolishing fluorescence occurring as a result of system activation by LasR. As shown in Fig. 3B, expression of the rhlA::gfp(ASV) and pqsA::gfp(ASV) reporters was significantly lower in the P. aeruginosa lasR mutant than in the wild type, in accordance with LasR acting as a positive regulator of both the Rhl and Pqs systems. Interestingly, the experiment also showed that the three compounds reduced the expression of the rhlA::gfp(ASV) and pqsA::gfp(ASV) reporters in the lasR mutant compared to that in the untreated lasR mutant, indicating that the compounds, to some degree, exert an effect on the Rhl and Pqs systems independently of the effects seen from LasR inhibition.

Influence of QSIs on virulence factor production.

We subsequently tested the effects of our identified QSIs on the production of three QS-controlled virulence factors: exogenous proteases, pyoverdine, and rhamnolipid. The production of exogenous proteases was assessed indirectly by measuring the ability of culture supernatant to degrade the colored protein substrate azocasein. Pyoverdine production was estimated directly through spectrophotometric measurements of the pyoverdine concentrations in culture supernatants. Rhamnolipid production was estimated indirectly by an emulsification activity assay and a drop collapse assay that estimates the presence of biosurfactants.

As shown in Fig. 4A, the three compounds significantly inhibited the production of exogenous proteases (α = 0.01; single-factor ANOVA). However, it was evident that the reduction in protease production observed for the three inhibitors was less than that observed in the QS-deficient lasI rhlI mutant strain, suggesting that some QS activity remains in P. aeruginosa when it is treated with the inhibitors at the concentrations used. Treatment of P. aeruginosa with the three identified QSIs resulted in a reduction in the supernatant levels of pyoverdine (Fig. 4B), suggesting that the three compounds reduce the production of the siderophore. The results were statistically significant at the α = 0.01 level in a single-factor ANOVA. Interestingly, inhibition of pyoverdine production occurred to a greater extent than the inhibition of protease production and the chlorzoxazone treatment resulted in pyoverdine levels comparable to that of the P. aeruginosa lasI rhlI mutant. This indicates that chlorzoxazone, at the concentration tested, is able to reduce pyoverdine production to a level resembling the level observed in QS deficiency. Finally, evidence was obtained that rhamnolipid production was inhibited by the three identified QSIs. An emulsification activity assay resulted in activities that were significantly reduced in cultures treated with the three compounds (α = 0.01; single-factor ANOVA) compared to the activity in untreated wild-type cultures (Fig. 4C). As for the pyoverdine results, the emulsification results indicate that chlorzoxazone has the ability to reduce rhamnolipid production to a level resembling that seen in a QS-deficient P. aeruginosa lasI rhlI mutant. The results were corroborated by a qualitative estimation of rhamnolipid production based on a drop collapse biosurfactant assay (5), which indicated that treatment with the three inhibitors affected the production of rhamnolipid by P. aeruginosa as follows: wild type, 100% production; lasI rhlI, 0%; salicylic acid treatment, 50%; nifuroxazide treatment, 25%; chlorzoxazone treatment, 0% (see Fig. S2 in the supplemental material).

FIG. 4.

Inhibition of exogenous protease production (A), pyoverdine production (B), and rhamnolipid production (C) in P. aeruginosa treated with the three identified LasR inhibitors. 1, untreated; 2, lasI rhlI mutant (QS deficient); 3, salicylic acid treated; 4, nifuroxazide treated; 5, chlorzoxazone treated. Results were taken after 18 h of growth at 37°C. Averages and SDs of five replicates are shown. Relat., relative; activ., activity; prod., production.

Effects of the QSIs on flow chamber biofilm development.

QS has been reported to play an important role in P. aeruginosa biofilm development and biofilm-related tolerance to antibiotics (4, 15, 45). Evidence has previously been presented that biofilms grown under conditions that repress QS are subject to sloughing (28), which may, at least in part, be caused by a lack of extracellular DNA matrix material (1). This prompted us to test the influence of the identified QSIs on P. aeruginosa biofilm formation. Biofilms were grown on glucose minimal medium in flow chambers in the presence or absence of non-growth-inhibitory concentrations of the QSI compounds, and as shown in Fig. 5, the characteristic mushroom-shaped structures normally observed in such P. aeruginosa biofilms (36) were not present in the inhibitor-treated biofilms. Instead, the inhibitor-treated biofilms resembled the QS-deficient P. aeruginosa lasI rhlI biofilm both in structure (flat and relatively unstructured) and in thickness. The finding that the inhibitor-treated biofilms were flatter and less structured than the wild-type biofilm was supported by objective image analysis. COMSTAT analysis was used for estimation of the total biomass present in biofilms formed by untreated, QSI treated, and QS-deficient cells. As shown in Fig. 6, inhibitor-treated biofilms contained less biomass than untreated biofilms, indicating impairment of the ability of P. aeruginosa to develop or maintain biofilms when treated with the selected QSIs. In support of these results, the QSI-treated biofilms had a total biomass comparable to that of biofilms formed by the QS-deficient lasI rhlI mutant strain.

FIG. 5.

CLSM pictures of 4-day-old P. aeruginosa biofilms. A, untreated; B, lasI rhlI mutant (QS deficient); C, salicylic acid treated; D, nifuroxazide treated; E, chlorzoxazone treated. The strains were GFP tagged for visualization. The main pictures are top-down 3D projections, while the flanking pictures are vertical sections. Bars, 20 μm.

FIG. 6.

Results of COMSTAT analysis for total biomass calculation of the individual biofilms. 1, untreated; 2, lasI rhlI mutant (QS deficient); 3, salicylic acid treated; 4, nifuroxazide treated; 5, chlorzoxazone treated. Averages and SDs from analysis of 12 images taken at random positions in three different biofilms are shown.

Assessment of the inhibitory activity of the QSI compounds in another microbial species.

The results of the Rhl inhibition assay and the LasR specificity assays indicated that the QSIs, in addition to the effect seen through inhibition of LasR, to some degree directly inhibit the activity of RhlR. It was therefore of interest to investigate whether the three compounds could inhibit AHL-based QS in other organisms. S. liquefaciens uses QS by means of C4-HSL and C6-HSL in the coordination of cellular functions such as swarming motility (19). In order to assess the activity of the QSIs on S. liquefaciens QS, we investigated the effects of the three compounds on S. liquefaciens swarming motility. A reduction of the swarming zone was observed when the bacteria were grown on agar with salicylic acid (0.347 cm²; n = 3, standard deviation [SD] = 0.03) or chlorzoxazone (0.150 cm²; n = 3, SD = 0.007), compared to that of bacteria grown on plates with no QSI (11.1 cm²; n = 3, SD = 0.9). The sizes of the zones were similar to that of the zone observed for an swrI mutant (0.143 cm²; n = 3, SD = 0.01) that is defective in swarming motility due to a lack of AHL synthesis (19). Nifuroxazide inhibited swarming at an intermediate, but significant, level (5.41 cm²; n = 3, SD = 0.5). Growth experiments confirmed that the QSIs do not affect the growth of S. liquefaciens at the concentrations used in the swarmer assays (data not shown). The results thus suggest that the three QSIs also have the potential of inhibiting QS in organisms using short-to-medium-chain AHLs.

Investigation of the structural features required for LasR-inhibitory activity.

It was of interest to investigate the proposed binding of salicylic acid, nifuroxazide, and chlorzoxazone to the LasR ligand binding domain as determined by the docking program MVD compared to the binding modes proposed by the program for known QSIs and the natural ligand itself. Finding a certain pattern in the proposed interactions between the active compounds and the binding site would be of interest in the identification of other QSIs, as well as for optimization of the already discovered ones. As shown in Fig. 7A to C, the three discovered QSIs were proposed to align with a ring structure in place of the homoserine lactone moiety of the natural ligand, a feature also proposed to occur for the LasR antagonists 4-NPO and furanone C30 (Fig. 7D and E). This, however, was the only pattern present since specific interactions between the compounds and LasR revealed no similarity and was therefore inadequate to improve the screening used in the present study.

FIG. 7.

Scenes displaying the conformations (magenta) of the identified LasR inhibitors and known LasR inhibitors as proposed by MVD after docking of the compounds. A, salicylic acid; B, nifuroxazide; C, chlorzoxazone; D, 4-NPO, the known LasR inhibitor used as cutoff in the docking; E, furanone C30, a potent LasR and QSI; F, OdDHL, the natural LasR ligand, as it is present in the LasR crystal structure with its hydrogen bonds (black lines) to the protein (green residues). In scenes A to E, hydrogen bonds proposed by MVD to occur upon binding are shown as black dotted lines together with the interaction residues (green). The conformation of OdDHL (white) has been included in scenes A to E for comparison, and the backbone of the LasR ligand binding domain is shown in blue-white with α-helical residues 65 to 72 omitted for clarity.

DISCUSSION

The increase in antibiotic resistance seen in P. aeruginosa clinical isolates (8, 9) and their ability to form persistent infections through the formation of biofilms (13, 11) have drawn attention to the improvement of current treatment strategies. A great effort has been made to develop antipathogenic drugs and strategies (29, 57, 61), especially by means of reducing bacterial virulence through intercellular communication (QS). Blocking of QS in P. aeruginosa by the use of QSIs has been shown to be a promising strategy for the treatment of infections (29, 49), and a series of QSIs has been identified by different groups through traditional methods (29, 42, 50, 62). The traditional methods, however, have limitations that can be complemented by novel computer-aided drug design (40, 54).

Use of X-ray crystallography-derived 3D structures for structure-aided drug design is a common activity in drug discovery today (14). In this process, the structures of macromolecular targets, often proteins or protein complexes complexed with their natural ligands or identified inhibitors, are solved and the structural information is used to design inhibitors or screen molecules for putative inhibitory activity. In the present study, we have used structure-based virtual screening to identify novel P. aeruginosa QSIs. The database used consisted of recognized drugs/natural ligands, i.e., compounds that have a significant potential for the clinical treatment of P. aeruginosa infections.

Our screening identified three compounds, salicylic acid, nifuroxazide, and chlorzoxazone, that displayed QS-inhibitory activity at concentrations that did not affect bacterial growth (Fig. 2 and 3).

The results indicated a direct inhibition of LasR by the compounds, and they were also shown to inhibit the two remaining QS systems in P. aeruginosa, Rhl and Pqs. As LasR controls these two systems (35, 60), the observed inhibition could be an effect resulting from LasR inhibition, as opposed to direct inhibition of the two systems. To elaborate on the specificity of the compounds, Rhl and Pqs inhibition was investigated in a P. aeruginosa lasR mutant. The results indicated that the two systems were affected by the compounds to some degree in the absence of LasR, which suggests that the three identified QSIs directly affect the Rhl and Pqs systems as well. Salicylic acid appears to be a fairly potent inhibiter of Pqs-dependent signaling (Fig. 3), suggesting that this compound might act as a Pqs signal antagonist. Further work could usefully include microarray experiments to gain more insight into the mode of action of the three identified QSIs.

In addition, treatment of P. aeruginosa with the compounds resulted in a decrease in the production of QS-controlled virulence factors (Fig. 4). Interestingly, treatment with chlorzoxazone was found to result in virulence factor inhibition resembling that of a lasI rhlI mutant. This result suggests that chlorzoxazone has the potential to function as an effective antipathogenic drug which can be used in the treatment of P. aeruginosa infections. Treatment with salicylic acid and nifuroxazide resulted in virulence factor inhibition that occurred to a lesser extent than that seen in the QS-deficient lasI rhlI mutant strain, indicating that QS regulation is not completely switched off by these two identified inhibitors at the concentrations tested. The facts that chlorzoxazone was less inhibitory of lasB expression than salicylic acid and nifuroxazide (Fig. 2) but displayed more potent virulence factor inhibition (Fig. 4) support the findings that the identified QSIs affect the Rhl and Pqs systems, in addition to the Las system. Compared to other identified inhibitors, such as furanone C30, patulin, and penicillic acid, the inhibitors discovered here are not as potent (29, 51), but the fact that the compounds are already approved drugs for human use is a significant benefit in the further application and development of antipathogenic drugs. Our previously identified potent inhibitors are experimental drugs and have been important for proof of concept (29, 51), but they are not pharmaceutically relevant due to toxicity and instability.

The three compounds salicylic acid, nifuroxazide, and chlorzoxazone were also found to alter biofilm formation by wild-type P. aeruginosa PAO1. Treated biofilms were thinner and less structured than untreated ones and resembled the biofilms formed by the QS-deficient mutant (Fig. 5). An analysis of the total biomass present in the biofilms showed that the inhibitor-treated biofilms and the QS-deficient biofilm contained roughly half the biomass of the wild-type biofilm (Fig. 6). These results agree with a study by Davies et al. (15), who presented evidence that QS is involved in the maturation of P. aeruginosa biofilms. Furthermore, alteration of P. aeruginosa biofilms through the action of QSIs has been shown to result in a decrease in the tolerance of antibiotics and an increase in the effectiveness of phagocytosis by polymorphonuclear leukocytes (4, 13, 29). These observations emphasize the importance of the ability to alter biofilm formation by blocking QS, as shown by the inhibitors discovered in this study. Interestingly, the prolonged continuous exposure of P. aeruginosa to salicylic acid and nifuroxazide occurring in the flow chambers at decreased concentrations, compared to the virulence factor assays, apparently improved the potency of the compounds. This indicates that the activity of the compounds, and QSIs in general, against virulence factor production should also be assessed in comparable treatment schemes to fully estimate the potency of the compounds in question.

The finding that the three QSIs, to some degree, exerted inhibitory effects on the Rhl QS system independently of LasR prompted an investigation of QS inhibition in other bacterial species that use AHLs similar in length to the Rhl system autoinducer C4-HSL. The compounds were found to reduce swarming motility of S. liquefaciens, which uses C4- and C6-HSL for QS-regulated swarming (19). This result thus indicates that the three compounds have the potential of inhibiting QS in other bacterial species, which can have implications for the treatment of mixed-species infections with, e.g., P. aeruginosa and Burkholderia cenocepacia.

Salicylic acid is a small organic acid that is produced by plants and functions in the induction of defense responses against pathogenic attack (58). It is the major metabolite of aspirin and is responsible for the antiinflammatory properties of the drug in humans (31). In agreement with our findings, previous studies have implicated salicylic acid in the reduction of P. aeruginosa virulence (2, 48). Besides reducing the production of virulence factors and biofilms, the compound was shown to reduce the infectious potential of the bacterium against the plant Arabidopsis thaliana and the nematode Caenorhabditis elegans (48). Our results confirm these previous findings and directly establish salicylic acid as a QSI in P. aeruginosa.

Nifuroxazide is a synthetic antimicrobial agent used in the treatment of enteric infections (41). The antimicrobial effects of the compound should, at first glance, prevent it from being a useful antipathogenic drug. The results presented here (Fig. 2) showed, however, that the compound, at the concentrations used, did not affect bacterial growth. This discrepancy between the normal use of the compound and the observed effect is not unique to nifuroxazide. Recent studies showed that ceftazidime, tobramycin, and macrolide antibiotics, used at subinhibitory concentrations, displayed strong QS-inhibitory effects (23, 60, 69). Thus, nifuroxazide should still be a candidate in the development of antipathogenic drugs despite its antimicrobial activity at high concentrations.

The third drug found to be QS inhibitory, chlorzoxazone, is a centrally acting muscle relaxant when administered to humans (46). In addition to its medical effects, the compound has also been found to be hepatotoxic in some patients (47). The high QSI activity of the compound at the concentrations tested, however, makes the compound an interesting antipathogenic drug candidate, and therefore the exact hepatotoxic effects at the amounts required should be assessed before the compound is excluded from further development. Due to its small size, the compound may also be used as a lead compound from which other nontoxic QSIs for use as antipathogenic drugs can be synthesized.

Our discovery of three new QSIs prompted us to investigate the proposed binding of the three compounds to the LasR ligand binding domain in order to improve the screening procedure. Knowing specific interactions required for inhibition is of great value in the search for new inhibitors. A pattern in the interactions between the molecule and its binding site indicating inhibitory activity was, however, not present (Fig. 7). The identified feature of having a ring structure in place of the homoserine lactone moiety of the natural ligand is inadequate for screening improvement. This lack of features indicating inhibitory activity should also be seen in light of the mode of binding between LasR and its natural ligand. The structure of LasR complexed to OdDHL reveals that the ligand is deeply buried in a cavity inside the protein (7). Combined with results displaying irreversible binding of the ligand to LasR (55), this indicates that competitive binding of QSIs occurs with the protein in a more open conformation. Therefore, the available structure of the LasR ligand binding domain with bound OdDHL may not be representative of the structure to which the inhibitors bind, which is why structural features and interactions required for inhibition may be impossible, or at least difficult, to determine. The combination of SB-VS and a preliminary similarity search for molecules similar to already known LasR binding compounds presented here has, however, proven to be successful in the discovery of new QSIs.

Despite its insufficiencies for using the structure of LasR as a powerful tool in the future discovery of QSIs, the screening strategy used here has proven successful, with a hit rate of 50% (three active compounds out of six tested). Unlike the three indentified QSIs, the three compounds that did not show QSI activity were all from pills. Although HPLC-MS confirmed that the extraction procedure was adequate, we cannot exclude the possibility that some of the filler used in the pills binds the active compound under the conditions of the experiment and thereby limits its accessibility to the bacteria. (Oral delivery of the pills to humans has the benefit of extremes of pH to disrupt such binding.) The relatively small amount of compounds included for the SB-VS renders the possibility of false negatives, i.e., compounds with undiscovered activity that were discarded on the basis of the virtual screening process. Despite the desire to detect as many active compounds as possible, the occurrence of false negatives lies in the very nature of virtual screening. The methods used in the process are far from perfect, resulting in the possible exclusion of active compounds. In the screening strategy used here, the potent inhibitor furanone C30 would have been discarded on the basis of the cutoff values used. To avoid or reduce this phenomenon, an increased number of compounds should be tested in future searches.

In conclusion, we have used SB-VS for the identification of three P. aeruginosa QSIs. One of them, salicylic acid, was discovered previously to be an inhibitor of P. aeruginosa virulence, including virulence regulated by QS, which has been further established in this study. In addition, we have identified the two new QSIs nifuroxazide and chlorzoxazone. These two compounds display a higher level of inhibition than salicylic acid, with chlorzoxazone treatment resembling conditions of QS deficiency, which is of particular interest in the further development of effective antipathogenic drugs active against P. aeruginosa.