Identification of a Hydrophobic Cleft in the LytTR Domain of AgrA as a Locus for Small Molecule Interactions that Inhibit DNA Binding

Abstract

The AgrA transcription factor regulates the quorum-sensing response in Staphylococcus aureus, controlling the production of hemolysins and other virulence factors. AgrA binds to DNA via its C-terminal LytTR domain, a domain not found in humans but common in many pathogenic bacteria, making it a potential target for antimicrobial development. We have determined the crystal structure of the apo AgrA LytTR domain and screened a library of 500 fragment compounds to find inhibitors of AgrA DNA-binding activity. Using NMR, the binding site for five compounds has been mapped to a common locus at the C-terminal end of the LytTR domain, a site known to be important for DNA-binding activity. Three of these compounds inhibit AgrA DNA binding. These results provide the first evidence that LytTR domains can be targeted by small organic compounds.

S. aureus typically causes skin or soft tissue infections at a localized lesion.¹ Patients are at risk of more serious life threatening diseases such as pneumonia, osteomyelitis, bacteremia, endocarditis and toxic shock syndrome^2–5 if S. aureus is able to cross into the bloodstream. The treatment of S. aureus infections has become increasingly problematic due to the emergence of antibiotic resistant strains. Methicillin resistant S. aureus (MRSA) is now the most common antibiotic resistant pathogen identified in developed world hospitals⁶ and poses a significant threat to human health.

The pathogenicity of S. aureus requires the coordinated expression of a large number of virulence factors. The agr quorum-sensing system, which allows the pathogen to sense both the density of the local S. aureus population and its degree of confinement, plays a central role in the regulation of the S. aureus virulon. Activation of the agr system represses the production of cell surface adhesins and promotes the secretion of extracellular toxins.

The agr operon consists of four genes, designated agrBDCA. The propeptide, AgrD, is processed and exported by AgrB. The secreted cyclic peptide is the autoinducing peptide (AIP) that functions as the quorum-sensing signal. When the extracellular AIP concentration is high, the histidine kinase, AgrC, is activated, resulting in increased phosphorylation of the cytoplasmic response regulator protein, AgrA. Phosphorylated AgrA activates transcription of the agrBDCA operon and genes encoding phenol soluble modulins⁷ and the effector RNA molecule, RNAIII.⁸ In addition to its function as the messenger RNA for hemolysin δ,⁹ RNAIII uses antisense RNA mechanisms to down-regulate adhesins and activate the transcription of genes encoding hemolysins, Panton-Valentine leukocidin (PVL) and enterotoxins.^10–13

Deletion of the agr operon attenuates the S. aureus infection in mouse and rabbit animal models of infection demonstrating the importance of agr quorum sensing for pathogenesis.^14–19 Indeed, AIP analogues that inhibit the AgrC histidine kinase are effective in reducing the severity of S. aureus infection.^17,19 However, attempts to identify inhibitors of response regulator AgrA have not been reported, despite the attractiveness of AgrA as a target because of the absence of LytTR DNA-binding domain proteins in mammalian proteomes.²⁰

We have determined the high-resolution crystal structure of the AgrA C-terminal LytTR domain (AgrA_C) and used a fragment screening approach to search for small molecule binding sites on the DNA-binding surface of this domain. Fragment-screening approaches have been widely used to efficiently screen a broad area of potential chemical space, using a relatively small library of compounds. Although the affinity of small fragment compounds for a target protein is expected to be relatively low, due to the low molecular weight of the compounds used (<300 g·mol⁻¹), fragment screening can identify energetic focal points on a protein surface that can be targeted by small molecule compounds.²¹

EXPERIMENTAL PROCEDURES

Protein expression and purification

Unlabeled S. aureus AgrA_C protein (AgrA residues Asp137 to Ile238) samples were produced in E. coli grown in terrific broth (TB) media as previously described.²² For NMR studies, ¹⁵N and ¹⁵N/¹³C isotopically-enriched protein samples of AgrA_C were prepared by expressing the AgrA_C protein in E. coli BL21 (DE3) pLysS grown in M9 minimal media at 18 °C using 2.5 g L⁻¹ (¹⁵NH₄)₂SO₄ and 2 g L⁻¹ [¹³C]-glucose (Cambridge Isotope Laboratories) as appropriate. The M9 media was supplemented with 50 μg mL⁻¹ kanamycin, 50 μM FeCl₂, 2 μM CuCl₂, 2 μM Na₂MoO₄, 2 μM NiCl₂, 2 μM CoCl₂, 2 μM H₃BO₃, 10 μM MnCl₂, 10 μM ZnSO₄, 20 μM CaCl₂, and 1 μM of each of the following micronutrients: nicotininc acid, pyridoxine, thiamine, biotin, riboflavin, folic acid, D-pantothenic acid and myo-inositol. Recombinant protein expression was induced using 0.3 mM isopropyl-β-D-thiogalactopyranoside.

All AgrA_C protein samples were purified from E. coli lysates following the previously published procedure²² using HiTrap SP HP cation exchange, HiLoad Phenyl Sepharose HP hydrophobic interaction and HiLoad Superdex 75 gel filtration chromatograpy (GE Healthcare). Purified protein was transferred into appropriate buffers by dialysis.

Crystallization of AgrA_C

Initial AgrA_C crystals were prepared at 4 °C by hanging drop vapor diffusion by mixing 1 μL of 1 mM AgrA_C (dissolved in 20 mM Bis Tris, 100 mM NaCl and 10 mM DTT at pH 6.0) with 1 μL of reservoir solution (100 mM Tris, 150 mM LiSO₄ and 11% (w/v) PEG 4000 at pH 8.0). The hanging drop was suspended above a 1-mL reservoir. The crystals were improved by streak seeding after 24 h incubation at the same condition except the PEG 4000 concentration was lowered to 8% (w/v).

Data collection and structure refinement

AgrA_C crystals were soaked for 30 s in 50 mM Tris, 75 mM LiSO₄, 8% PEG 4000 and 20% glycerol at pH 8 before flash freezing in liquid nitrogen. A native data set was collected at 100 K using a Rigaku MicroMax-007 HF generator equipped with RAXIS-IV++ detector. Data were processed and scaled with DENZO and SCALEPACK.²³ The structure of apo AgrA_C was solved by molecular replacement using Phaser 2.0²⁴ with the DNA-bound state of AgrA_C as the search model (PDB 3BS1). The R factor after rigid body refinement of the best molecular replacement solution was 0.440. From the initial molecular replacement solution, the structure was rebuilt from scratch using RESOLVE²⁵ before iterative refinement using COOT 6.02²⁶ and Phenix.²⁷ The refined model contains two molecules of AgrA_C within the asymmetric unit; chains A and B contain residues Glu141 to Ile238 and Ser139 to Ile238, respectively. The model was refined to 1.52 Å with R factor and R_free values of 0.180 and 0.209 respectively. All residues lie within the allowed regions of the Ramachandran plot and exhibit favorable bond angles and bond lengths. An overview of the data collection and refinement statistics is provided in Table 1.

Table 1.

X-ray data collection and refinement statistics (molecular replacement)

	AgrACa
Data collection
Space group	P212121
Cell dimensions
a, b, c (Å)	41.4, 45.9, 112.1
α, β, γ (°)	90.0, 90.0, 90.0
Resolution range (Å)	25.00-1.52 (1.55-1.52)b
Rsymc	0.045 (0.563)
I/σ(I)	32.9 (2.0)
Completeness (%)	97.2 (89.8)
Redundancy	3.3 (2.7)
Refinement
Resolution (Å)	20.72-1.52
No. reflections	31964
Data cutoff	σ(F) > 0
Rworkd/Rfreee	0.180/0.209
No. atoms per asymmetric unit
Protein	1702
Glycerol	18
Water	254
Average B-factors (A2)
Protein	20.5
Glycerol	32.3
Water	32.6
Rms deviations from ideality
Bond lengths (Å)	0.006
Bond angles (°)	1.048

^aA single crystal was used for the data collection.

^bValues in parentheses are for the highest-resolution shell.

^cR_sym = Σ_h|I_h – <I>|/Σ_hI_h, where I_h and <I> represent the diffraction intensity values of individual measurements and the corresponding mean values, respectively.

^dR_work = (Σ_h||F_o| – |F_c||)/Σ_h|F_o|, where F_o and F_c are observed and calculated structure factor amplitudes, respectively.

^eR_free was calculated for 5% of the randomly selected reflections of the data set that were not used in the refinement.

NMR spectroscopy

All NMR data were acquired on Varian INOVA 600 MHz or Varian 500 MHz spectrometers equipped with triple resonance probes. The NMR spectra were referenced to an internal 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS) standard as described previously.²⁸ All spectra were processed using NMRpipe²⁹ and analyzed using CCPNMR analysis 2.1.³⁰

For the backbone NMR assignments, ¹H-¹⁵N HSQC³¹, HNCACB, HN(CO)CACB, HNCA, HN(CO)CA and HNCO³² spectra were acquired at 298 K from a sample containing 1 mM AgrA_C in 20 mM sodium phosphate, 100 mM NaCl and 5% D₂O at pH 5.8. The backbone resonance assignments for AgrA_C were established for all residues between Asn138 and Ile238 except for Tyr229.

¹⁵N heteronuclear NOE³³ spectra were recorded with 3.0 s ¹H saturation in the latter part of a 3.5-s preparation period delay, which was also used without radio frequency pulses for the reference 2D spectrum without NOE.

A library of 500 fragment compounds (Maybridge Ro3 500 fragment library) was screened for binding to AgrA_C using a WATERGATE W5 LOGSY NMR experiment.³⁴ Each sample contained a mixture of 10 compounds (400 μM of each compound) and 15 μM AgrA_C, dissolved in 20 mM sodium phosphate, 100 mM NaCl, 10 mM DTT and 5% D₂O at pH 7.0. As a negative control, spectra were acquired for a second sample without AgrA_C. Hit compounds were identified by the inversion of the compound spectra when AgrA_C was present and validated using single compound repeats of the WATERGATE W5 LOGSY experiment. The WATERGATE W5 LOGSY spectra were acquired at 20 °C using a mixing time of 1.2 s, 8192 complex data points, spectral width of 10000 Hz and number of scans set to 128. The acquisition time was 0.819 s with a relaxation delay of 3.0 s. The total experiment time was 11 min.

To identify the compound-binding sites, ¹H- ¹⁵N HSQC spectra were acquired for samples containing 300 μM AgrA_C dissolved in 20 mM sodium phosphate, 100 mM NaCl, 10 mM DTT and 5% D₂O at pH 5.8 after each addition of compound to the protein sample. 200 mM stock solutions of each compound, dissolved in DMSO-d₆ were used for the titrations. A control experiment where 3% DMSO-d₆ was added to the AgrA_C NMR sample in the absence of compound was used to establish which chemical shift perturbations were compound dependent.

The combined chemical shift was calculated for each amide resonance after the addition of compound, using the established method,³⁵ and mapped onto the apo AgrA_C crystal structure using Pymol v1.2r2. For each compound the highest spectra obtained with the highest concentration of compound tested in the NMR titrations was used to determine the combined chemical shift perturbation.

Compound docking

In-silico docking of the compounds binding to AgrA_C was performed using Autodock Vina.³⁶

Inhibition Assays

The activities of five compounds as inhibitors of AgrA_C DNA binding were tested using an electrophoretic mobility shift assay performed in the presence of increasing amounts of each compound (from 0 mM to 5 mM) to test whether the presence of compound inhibits the DNA-binding activity of AgrA_C. A 19-bp DNA duplex was prepared by annealing 5′-ATTTAACAGTTAAGTATTT-3′ with its complementary oligonucleotide in 21 mM sodium phosphate, 105 mM NaCl, 10.5 mM DTT at pH 5.8 and purified by size exclusion chromatography using a Superdex 75 HR column. Solutions were prepared containing 2 μM DNA duplex, 20 μM AgrA_C and various compound concentrations in 20 mM sodium phosphate, 100 mM NaCl, 10 mM DTT and 2.5% DMSO-d₆ at pH 5.8. After incubation at room temperature for 30 min, 12% (w/v) Ficoll 400 was added to each sample to give a 2% (w/v) final concentration before running the samples on an 8% acrylamide/bis-acrylamide 40 mM Tris·HCl, 20 mM acetic acid, 1 mM EDTA at pH 8.0 (TAE) gel at 70 V and 4 °C. The gels were stained in TAE supplemented with 0.1 mg mL⁻¹ ethidium bromide for 10 min prior to destaining in TAE for 10 min. DNA was visualized with UV light and images were captured using a CCD camera imager (Alpha Innotech).

RESULTS

Backbone Resonance Assignments of AgrA_C

Backbone resonance assignments were determined for AgrA_C using a ¹⁵N,¹³C-labeled sample of AgrA LytTR domain (Asp137 to Ile238). The protein sample gave excellent quality NMR spectra with well-dispersed chemical shifts (Fig. 1.) allowing for assignment of the backbone resonances. Assignments could be made for all of the residues between Asn138 and Ile238 except for Tyr229 whose resonances could not be confidently assigned, probably due to overlapping crosspeaks with another amide crosspeak. The unusual downfield shift of the serine 202 backbone amide proton resonance is caused by the hydrogen bonding of this proton with the imidazole ring of His200, as observed in the crystal structure of AgrA_C. A similar downfield shift has been described for the hydrogen bonding of Arg7 backbone amide to His106 side chain in B. subtilis chorismate mutase.³⁷

Figure 1.

2D ¹H-¹⁵N HSQC spectrum of AgrA_C. Residue specific cross-peak assignments for the backbone amides are indicated.

AgrA_C in the Absence of DNA

The structure of AgrA_C bound to its target DNA duplex has been reported previously,²² but any structural changes that occur within this domain upon DNA binding were unknown. To establish if there are structural differences between the apo and DNA-bound states of AgrA_C both NMR and X-ray crystallographic analyses were used to investigate the structure of AgrA_C protein in the absence of DNA.

AgrA_C was crystallized in its apo state, forming P2₁2₁2₁ space group crystals that diffracted to 1.5-Å resolution. The structure of apo AgrA_C was solved by molecular replacement. To remove model bias from the molecular replacement solution, the AgrA_C protein was rebuilt from scratch using RESOLVE and then iteratively refined. The final structure contains two molecules of AgrA_C in the asymmetric unit of the crystals: chain A contains residues Glu141 to Ile238 and chain B contains Ser139 to Ile238).

For both AgrA_C protomers, the N-terminal residues are not observed in the electron density map, suggesting that this region is disordered, consistent with the negative {¹H}¹⁵N heteronuclear NOE values observed for these amino acids (Fig. 2A) under solution conditions. The apo AgrA_C protein adopts a β-β-β-sandwich fold as observed previously for the DNA-bound state. The fold is stabilized by a set of conserved salt bridges: Glu141-Arg195, Asp157-His208, Asp193-Arg195, Asp157-Arg195 and His174-Glu226. The domain is very stable with heteronuclear NOE values for the backbone amide bond vectors between 0.7 and 1.0 for the majority of the protein. However, the β1-β2 loop, the β3-β4 loop and the C terminus have ¹H¹⁵N-heteronuclear NOE values between 0.6 and 0.7, suggesting greater conformational flexibility for these regions relative to the rest of the LytTR domain (Fig. 2A). In addition, alternate rotamer conformations are observed in the electron density map for several amino acid sidechains within the AgrA_C crystal structure (Ser164, Ser215 and Glu217 from protomer chain A and Ile143, Ser149 and Asp176 from chain B), indicative of localized conformational flexibility within the domain. When the two protomer chains in the AgrA_C crystal are superposed, the only differences in the protein backbone are observed for the N and C termini as well as the β1-β2, and β3-β4 loops, in excellent agreement with the heteronuclear NOE data. The root mean square (RMS) deviation for the two protomer chains in the apo AgrA_C crystal is 0.19 Å for all backbone atoms, but a much greater difference is seen when the DNA-bound state is compared with the apo AgrA_C protein chains A or B (RMS deviations of 0.41 or 0.46 respectively), suggesting that a conformational change takes place upon DNA binding.

Figure 2.

Conformational differences between apo and DNA-bound AgrA_C. A, Steady-state heteronuclear {¹H}¹⁵N-NOE values for each residue of AgrA_C plotted versus the secondary structure topology of the protein. β strands and α helices are depicted as arrows and cylinders respectively. B, Superposition of the apo AgrA_C crystal structure (green) with AgrA_C in its DNA-bound state (orange).²² The sidechains of amino acid residues within the DNA-binding surface that adopt different conformations between the apo and DNA-bound states are shown in stick format.

When the apo and DNA-bound structures of AgrA_C are compared, changes in the conformation of the DNA-binding surface are observed for residues that are known to interact with the DNA backbone (Ser 168, Arg170, Tyr183, Lys187, Ser202, Arg218, Asn234) as well as for three residues that make direct, base-specific contacts (His169, Asn201 and Arg233)²² (Fig. 2B). The largest change between the DNA-bound and apo states of the AgrA_C protein is observed for the β3-β4 loop (Ser165 to Arg170). There are no crystal contacts for this region for the chain B protomer in the apo protein structure, so crystal lattice packing does not explain the different configuration of the β3-β4 loop relative to that in the DNA-bound state of AgrA_C. The conformational differences between the apo and DNA-bound structures suggest that DNA binding traps the protein into a conformation that represents only one conformation from the population of states adopted by the apo AgrA_C protein. The conformational freedom of the DNA-binding surface would allow some flexibility for AgrA binding to its DNA target sequences, albeit with an entropic penalty for binding.

Fragment Screening for Ligands of AgrA_C

LytTR domains are considered attractive drug targets for the development of new antibiotics because these domains are not present in humans and are often found to activate virulence pathways in bacterial pathogens.²⁰ NMR fragment screening methods were chosen because they allow the entire molecular surface of the target protein to be probed by a fragment compound library, in contrast to X-ray diffraction screening where some surfaces are occluded by lattice contacts. Fragment screening by NMR has proven to be a powerful tool for identifying regions of a protein that can be targeted by low complexity, small molecule ligands (fragment compounds), which subsequently can be used as starting points for drug development.^38,39 A library of 500 compounds was screened for binding to AgrA_C using a WATERGATE W5 LOGSY NMR experiment. Five compounds, 4-phenoxyphenol, 9H-xanthene-9-carboxylic acid, 2-(4-methylphenyl)-1,3-thiazole-4-carboxylic acid, [5-(2-thienyl)-3-isoxazolyl]methanol and 4-hydroxy-2,6-dimethylbenzonitrile, were identified as binding to AgrA_C as indicated by inversion of the compound spectrum when AgrA_C was present in the solution.

Mapping the Compound-Binding Site

Having identified five compounds that interacted with AgrA_C in the initial screen, increasing amounts of the each compound were titrated into a ¹⁵N-labeled sample of AgrA_C and chemical shift perturbations in the ¹⁵N-¹H-HSQC spectrum of the protein were monitored (Fig. 3.). Addition of each of the five compounds resulted in the perturbation of the same chemical shifts corresponding to the backbone amides for residues Ser231, Val232, Arg233, Asn234, Lys236, Lys237 and Ile238, at the C-terminal end of the protein (Fig. 4). A control experiment confirmed that the chemical shift perturbations were dependent on the presence of the compounds and not caused by the addition of DMSO-d₆ to the NMR sample (Fig. 4F). Mapping the chemical shift perturbations onto the surface of AgrA_C revealed a surface corresponding to a shallow groove formed by the β10-α2 loop and the α2 helix. Interestingly, the β10-α2 loop is known to be important for DNA binding by AgrA²² so ligands that occlude this site would be expected to reduce the DNA-binding activity of AgrA.

Figure 3.

Chemical shift perturbations observed upon binding of fragment compounds to AgrA_C. ¹H-¹⁵N HSQC spectra of AgrA_C recorded in the presence of the indicated concentrations of A, 4-phenoxyphenol, B, 9H-xanthene-9-carboxylic acid, C, 2-(4-methylphenyl)-1,3-thiazole-4-carboxylic acid, D, [5-(2-thienyl)-3-isoxazolyl]methanol and E, 4-hydroxy-2,6-dimethylbenzonitrile. The residue specific assignments for backbone amide cross-peaks that shift in a compound concentration dependent manner are highlighted. F, ¹H-¹⁵N HSQC spectra of ¹⁵N-labeled AgrA_C before and after the addition of 3% DMSO-d₆.

Figure 4.

Mapping the compound dependent chemical shift perturbations. Combined chemical shift perturbations of backbone amide resonances plotted versus the amino acid residue number for the addition of 4-phenoxyphenol A, 9H-xanthene-9-carboxylic acid, B, 2-(4-methylphenyl)-1,3-thiazole-4-carboxylic acid, C, [5-(2-thienyl)-3-isoxazolyl]methanol, D, and E, 4-hydroxy-2,6-dimethylbenzonitrile. The normalized combined chemical shift perturbations for F, 4-phenoxyphenol, G, 9H-xanthene-9-carboxylic acid, H, 2-(4-methylphenyl)-1,3-thiazole-4-carboxylic acid, I, [5-(2-thienyl)-3-isoxazolyl]methanol and J, 4-hydroxy-2,6-dimethylbenzonitrile are mapped onto the surface of the crystal structure of apo AgrA_C with the region showing the greatest combined chemical shift perturbation colored in yellow.

Compound Binding Predictions

In silico docking was used to predict the binding mode of the compounds using AutoDock Vina.³⁶ The search box was restricted to the region at the C-terminal end of AgrA_C, identified by NMR as the binding site for the compounds (Fig. 5). All five compounds docked into a shallow groove located between Val232 and Lys236. 4-Phenoxyphenol is predicted to form hydrogen bond interactions with Lys236 and the backbone carbonyl of Arg233 and bends around Val232 to make tight Van der Waals interactions with this amino acid sidechain. 9H-xanthene-9-carboxylic acid binds with the planar xanthene ring structure between Val232 and Lys236. The predicted interaction is further stabilized by a salt bridge between the carboxylic acid group of the compound and the sidechain amine of Lys236. 2-(4-Methylphenyl)-1,3-thiazole-4-carboxylic acid binding is predicted to be stabilized by a hydrogen bond between the thiazole nitrogen and the carbonyl group of Val232 and a salt bridge between Lys237 and the carboxylic acid group in the compound. [5-(2-thienyl)-3-isoxazolyl]-methanol is predicted to bind with the isoxazolyl ring oxygen forming a hydrogen bond with Lys236 amine. The putative compound binding mode is further stabilized by a hydrogen bond between Val232 backbone carbonyl group and the hydroxyl group of the compound. The 4-hydroxy-2,6-dimethylbenzonitrile compound is predicted to bind in the shallow groove formed between Val232, Arg233, Lys236 and Lys237, with the hydroxyl group of the compound making hydrogen bond interactions with the backbone of Val232 and Lys236. The calculated binding energies for the best docking poses for 4-phenoxyphenol, 9H-xanthene-9-carboxylic acid, 2-(4-methylphenyl)-1,3-thiazole-4-carboxylic acid, [5-(2-thienyl)-3-isoxazolyl]methanol and 4-hydroxy-2,6-dimethylbenzonitrile were −4.1 kcal mol⁻¹, −4.1 kcal mol⁻¹, −4.0 kcal mol⁻¹, −3.5 kcal mol⁻¹ and −3.6 kcal mol⁻¹, respectively, in agreement with the millimolar affinity observed by NMR.

Figure 5.

In silico compound docking predictions. Autodock Vina predicted binding modes are shown for A, 4-phenoxyphenol, B, 9H-xanthene-9-carboxylic acid, C, 2-(4-methylphenyl)-1,3-thiazole-4-carboxylic acid, D, [5-(2-thienyl)-3-isoxazolyl]methanol and E, 4-hydroxy-2,6-dimethylbenzonitrile. Orientations of the protein are slightly different in each panel to optimize viewing of the predicted hydrogen bonds between the compound and protein, indicated by dashed lines.

Three Fragment Compounds Inhibit the DNA-binding Activity of AgrA_C

The five compounds identified by fragment screening all target the C-terminal end of AgrA_C, a region known to be involved in DNA binding.²² It was therefore important to establish whether the compounds alter the DNA-binding activity of AgrA. The DNA-binding activity of AgrA_C was tested using electrophoretic mobility shift assays in the presence of various compound concentrations from 75 μM to 5000 μM (Fig. 6). For three compounds (4-phenoxyphenol, 9H-xanthene-9-carboxylic acid and 2-(4-methylphenyl)-1,3-thiazole-4-carboxylic acid), the intensity of the band corresponding to the AgrA_C:DNA complex decreases as the compound concentration is increased; no AgrA_C:DNA complex is observed when 5 mM compound is present. Two other compounds ([5-(2-thienyl)-3-isoxazolyl]methanol and 4-hydroxy-2,6-dimethylbenzonitrile) had no apparent effect on the DNA-binding activity of AgrA_C at concentrations up to 5 mM. The DMSO concentration in all binding reactions was constant so the disruption to the DNA-binding activity of AgrA_C is dependent only on the presence of the compound. In the control experiments where no AgrA_C was present, there is no effect of the compounds on the migration of DNA through the gel. Some of the compounds that were identified to bind at the C terminus of AgrA_C can disrupt AgrA DNA-binding activity.

Figure 6.

Effects of compounds on AgrA DNA-binding activity. Electrophoretic mobility shift assays were performed with AgrA_C, its target DNA duplex and the indicated concentrations of A, 4-phenoxyphenol, B, 9H-xanthene-9-carboxylic acid, C, 2-(4-methylphenyl)-1,3-thiazole-4-carboxylic acid, D, [5-(2-thienyl)-3-isoxazolyl]methanol and E, 4-hydroxy-2,6-dimethylbenzonitrile. DNA incubated with or without AgrA_C in the absence of compound are shown as positive and negative controls, respectively.

DISCUSSION

The agr quorum sensing system has been suggested as a possible drug target because deletion of the agr system attenuates S. aureus virulence in animal models of infection.^14–16,18 Furthermore, a functional agr system has been found to be important for bacterial survival in the early stages of abscess formation in animal models.¹⁹ On the other hand, isolation of agr-deficient mutants from patients with chronic S. aureus infections⁴⁰ suggests that agr is not important for maintaining the infection once it has been established. Thus substantial controversy surrounds the validity of the agr system as a target for antimicrobial drug development. agr inhibitors would be especially useful tools in addressing the role of the agr system during the course of infection in animal models of infection. Furthermore, agr inhibitors might be effective for prophylactic treatment. Indeed, co-inoculation of S. aureus with an agr inhibitor was sufficient to reduce virulence in a mouse skin infection model.¹⁹

To date, research on inhibitors of S. aureus agr quorum sensing has focused on analogues of the autoinducing peptide as inhibitors of AgrC activation.^{17,19,41–43} Targeting agr quorum sensing using AIP analogues is complicated by the existence of several Staphylococcal AgrC types that are inhibited or activated by different AIP molecules.^17,44,45

The results presented here represent the first attempt to inhibit the agr quorum sensing pathway by targeting the DNA-binding surface of the AgrA LytTR domain. Fragment screening can be used to identify an energetic focal point on the surface of a protein that is favorable to compound binding.²¹ While small fragments are efficient in covering large areas of chemical space and have advantages as starting points for drug development, their binding affinities are typically very low. Indeed, only three of the five compounds that were identified to bind to AgrA_C by NMR analyses disrupted DNA binding in electrophoretic mobility shift assays. For these compounds, inhibition was detected only at millimolar concentrations. The low affinity of ligand binding provides a likely explanation for the lack of inhibition observed with the other two compounds.

The five compounds identified by fragment screening as AgrA_C inhibitors all bind to a highly conserved region that forms a short helix at the C terminus of the AgrA protein. To date, there are 211 AgrA protein sequences across all Staphylococcal strains in the Uniprot knowledgebase database. Conservation of 100% sequence identity is observed for Ser231, Val232, Arg233, Asn234, Lys236 and Lys237 that are identified as residues in the compound-binding site. Val235 is conserved in 97% of all Staphylococcal sequences, although isoleucine or leucine amino acids are observed at this position in S. siminae, S. pseudintermedius, S. intermedius and S. carnosus strains. The C-terminal isoleucine residue (Ile238) is conserved in 94% of the AgrA protein sequences but is replaced by a lysine in S. aureus Mu3 and Mu50 strains or by valine in S. lugdunensis or S. saprophyticus strains. Other modifications at the AgrA C terminus give rise to short extensions. However, these extensions result in a reduced or completely deficient agr phenotype.⁴⁰ Given the high degree of sequence conservation of the compound-binding site, it is reasonable to expect that any compound binding to this surface would be able to inhibit agr quorum sensing across all Staphylococcal strains.

The fragment compounds identified in this study bind with millimolar affinity to the AgrA_C protein. The low affinity reflects the small size of the compounds that populate the library used in the screen. Despite the low affinity of initial hits, fragment screening has previously been shown to provide a useful starting point for the development of higher affinity inhibitors of protein:DNA interactions.^21,46 It is commonly observed that features of the original low affinity fragment compound, identified by the initial screen, are maintained in the compounds that ultimately enter clinical development^47–49. Further research is required to expand the compounds that have been identified as AgrA inhibitors, to improve their affinity and specificity. The apo AgrA_C crystal structure and NMR assignments that have been established in this research will provide valuable information to guide future compound development.

ABBREVIATIONS

AIP

autoinducing peptide

DMSO

dimethyl sulfoxide

DTT

dithiothreitol

NMR

nuclear magnetic resonance

NOE

Nuclear Overhauser Effect

PEG

polyethylene glycols

rms

root mean square