Discovery of Novel DNA Gyrase Inhibitors by High-Throughput Virtual Screening^▿

Database preparation.

The National Cancer Institute/Developmental Therapeutics Program (NCI/DTP) maintains a repository of 139,644 samples (plated compound set) (38). The three-dimensional coordinates for the NCI/DTP plated compound set in the MDL sd format were converted to the mol2 format by the program SDF2MOL2 (UCSF). Partial atomic charges for ligand atoms were calculated using SYBDB (UCSF) (13) and added to the mol2 file representing the NCI/DTP plated compound set (approximately 140,000 small molecules).

Molecular docking.

The procedure for molecular docking involves (i) the selection of structural pockets in DNA gyrase suitable for interactions with drug-like small molecules and (ii) molecular docking simulations where each one of approximately 140,000 small molecules (molecular weight, <500) is positioned in the selected structural pocket and scored based on predicted polar (e.g., H bond) and nonpolar (e.g., van der Waals) interactions. The 10 highest-scoring compounds for each selected structural pocket were obtained for use in DNA gyrase inhibition assays. Docking calculations were performed with the 15 October 2002 development version of DOCK v5.1.0 (13, 41).

The coordinates for the crystal structure of a 59-kDa fragment of gyrase A from E. coli, Protein Data Bank (PDB) code 1AB4, were used in the molecular docking calculations (39). The biologically relevant dimeric form of DNA gyrase A was generated by applying the crystallographic symmetry operation (−1.0, 0.0, 0.0, 119.6 Å, 0.0, −1.0, 0.0, 119.6 Å, 0.0, 0.0, 1.0, 0.0) to the coordinates in PDB 1AB4 using CCP4 (23). The molecular surface of the structure was explored using sets of spheres to describe potential binding pockets. The spheres literally fill in the available pocket spaces where a ligand might be able to form a complex. DOCK uses the spheres as a guide to search for orientations of each molecule that fit into the selected sites. The sites selected for molecular docking were defined using the SPHGEN program and filtered through the CLUSTER program (13). The SPHGEN program generates an unbiased grid of points that reflect the actual shape of the selected site. The CLUSTER program groups the selected spheres to define the points that are used by DOCK to match (superimpose) potential ligand atoms with spheres. Seventy-two spheres were used to define the gyrase A site for molecular docking. Each compound in the NCI/DTP database was positioned in the selected site in 100 different orientations. Intermolecular AMBER energy scoring (van der Waals + columbic), contact scoring, and bump filtering were implemented in DOCK v5.1.0 (13). PYMOL (9) was used to generate molecular graphic images.

The ATP binding site in DNA gyrase subunit B was targeted with small molecules by screening approximately 140,000 compounds, using a molecular docking protocol similar to that used for DNA gyrase subunit A, described above. For molecular docking into a structural pocket of DNA gyrase subunit B, the site selection criteria were based on the position of novobiocin in the ATP binding site (27). The coordinates were from the crystal complex of novobiocin and Thermus thermophilus, PDB code 1KIJ (26). PDB 2SPH (UCSF) was used to place spheres at the positions of novobiocin atoms. The novobiocin-bound structure of E. coli gyrase B is not available, but conserved residues in the ATP binding site were identified by aligning DNA gyrase subunit B sequences from E. coli and T. thermophilus by using ClustalX (6). Sequence variability was plotted on the molecular surface of T. thermophilus by using PYMOL to demonstrate the high degree of sequence and structural similarities. Forty-four spheres were used to define the site on gyrase B for molecular docking. Scoring was calculated in a 5-Å grid surrounding the spheres (GRID; UCSF). All molecular docking jobs were performed on SGI Octane workstations running DOCK (UCSF) in IRIX6.5. The 10 highest-scoring compounds for each selected structural pocket were obtained for use in DNA gyrase inhibition assays. Analysis of docked compounds was performed with HBPLUS (36) and plotted using LIGPLOT (45).

Chemicals and enzymes.

The small-molecule inhibitors used in this study were obtained from the Drug Synthesis and Chemistry Branch of the National Cancer Institute and are listed in Tables 1 and 3. Plasmids containing the A (pPH11) or B (pAG111) subunit of E. coli DNA gyrase (kindly provided by Tony Maxwell, Norwich Research Park, United Kingdom) were overexpressed in E. coli XL1-Blue (Stratagene), and the resulting polypeptides were purified as described previously by Maxwell and Howells (34). DNA gyrase activity was then reconstituted by mixing the two subunits together at a ratio of 1:1.4 (GyrA/GyrB).

Preparation of relaxed DNA substrate.

Relaxed plasmid DNA substrate was prepared by incubating supercoiled pRSET A DNA (Invitrogen) with wheat germ topoisomerase I (Promega) according to the manufacturer's specifications. The relaxed pRSET A DNA was then extracted with phenol-chloroform before ethanol precipitation and resuspension in 10 mM Tris-HCl, pH 7.5, 1 mM EDTA.

DNA supercoiling assay.

Supercoiling by DNA gyrase was determined using a standard agarose gel assay as described previously (18). Reaction mixtures (20 μl) containing 25 mM Tris-acetate, pH 7.9, 20 mM potassium acetate, 10 mM magnesium acetate, 2 mM dithiothreitol, 1.5 mM ATP, 5 mM spermidine-HCl, 50 μg/ml bovine serum albumin, 0.5 μg relaxed pRSET A DNA (Invitrogen), and various concentrations of drug were incubated on ice for 10 min. Reactions were then initiated by adding DNA gyrase (2.5 nM GyrA and 3.5 nM GyrB) and incubating the mixture at 30°C for 30 min. Reactions were terminated by adding 4 μl of stop solution (3% sodium dodecyl sulfate, 30% Ficoll, 0.6 mg/ml bromophenol blue, 60 mM EDTA), and the samples were loaded onto a 1% agarose gel in 40 mM Tris-acetate, pH 8.1, 2 mM EDTA. Following electrophoresis at 12 V for 20 h, gels were stained with ethidium bromide (1 μg/ml) and the DNA was visualized under UV light. The gel image was captured using a Bio-Rad Gel Doc apparatus, and the DNA bands were quantitated using Scion Image Beta software version 4.0.2 (Scion Corporation). The percentage of supercoiled DNA in each drug-treated sample was calculated relative to that in non-drug-treated gyrase controls.

ATPase assay.

The DNA-independent ATPase activity of E. coli GyrB was measured by a spectrophotometric method that couples the hydrolysis of ATP with the oxidation of NADH (31). The methodology and reaction conditions were the same as those previously reported using 40 nM of purified E. coli GyrB and 2 mM ATP (23). The apparent K_i [K_i(app)] value for Cancer Chemotherapy National Service Center (NSC) compound 20116 was determined as follows: 1/slope of a plot of inhibitor concentration versus (V₀/V_i) − 1, where V₀ and V_i represent the rates of ATP hydrolysis in the absence and presence of inhibitor, respectively (23). The K_i value was then calculated using the equation K_i = K_i(app)/(1 + [ATP]/K_m), using a K_m of 0.3 mM for E. coli gyrase ATPase (44).

DNA unwinding assay.

Drug-induced DNA unwinding was measured as described previously (28), using pGEM2 plasmid DNA (Promega, Madison, WI) and wheat germ DNA topoisomerase I (Promega, Madison, WI).

Bacterial growth studies.

To determine the effects of drugs on E. coli XL1-Blue cell growth, overnight cultures grown in LB media at 37°C were diluted to an optical density at 595 nm (OD₅₉₅) of 0.05 in 50 ml fresh LB media and incubated with shaking at 37°C until cells entered log-phase growth. Cells (10 ml) were then transferred into flasks containing various concentrations of filter-sterilized drug, and cell growth continued with shaking at 37°C. Cell growth was monitored at OD₅₉₅ using a Pharmacia LKB Ultrospec III spectrophotometer (Cambridge, England).

Discovery of a novel structural pocket on E. coli DNA gyrase subunit A for structure-based drug design.

The goal of this approach was to develop a strategy to inhibit DNA gyrase activity in E. coli by interfering with a functionally important structural pocket that is not affected by common mutations that cause resistance. Resistance to fluoroquinolones is associated primarily with mutations that occur in a small region (QRDR) of the A subunits of the DNA gyrase and topoisomerase IV genes, which encode a 50- to 60-amino-acid stretch in the N-terminal regions of the GyrA and ParC polypeptides, respectively (8, 14, 50; reviewed in references 12 and 20). In the case of E. coli GyrA, fluoroquinolone resistance is most often associated with amino acid substitutions at S83 and D87, which reside within the putative DNA binding surface of α-helix 4 (Fig. 2). Since more than 80% of the mutations that confer fluoroquinolone resistance occur in α-helix 4 of E. coli DNA GyrA, we targeted small molecules to a structural pocket that is not in contact with S83 or D87. The targeted region includes α-helix 3, which forms part of the GyrA dimer interface, and is in close proximity to Y122 (3.9 Å), the key catalytic residue involved in the DNA breakage-reunion reaction of DNA gyrase (Fig. 2) (39). This structural pocket is in close proximity to the catalytic residue Y122 but distant (18 Å) from S83 and D87, which are commonly altered in fluoroquinolone-resistant bacteria.

We used molecular docking to position each of approximately 140,000 compounds in 100 different orientations at the GyrA dimer interface. The 10 highest-scoring compounds based on van der Waals and electrostatic potential were then obtained from the NCI/DTP (38) and tested for inhibition of DNA gyrase supercoiling activity in vitro by using drug concentrations of up to 100 μg/ml (Table 1 and Fig. 3). Compounds (defined by NSC identification numbers) NSC 103003, NSC 130847, and NSC 20115 were found to significantly inhibit DNA gyrase supercoiling activity with 50% inhibitory concentrations (IC₅₀s) of 50, 72, and 737 μM, respectively. The most potent compound, NSC 103003, also had the highest overall energy score, −15.68 kcal per mol. Inhibition of DNA gyrase by NSC 103003 was further tested in the presence and absence of Triton X-100 to ascertain whether inhibition was an artifact due to sequestration of DNA gyrase by drug aggregates. McGovern et al. has shown that low concentrations of Triton X-100 (0.01%) can prevent or reverse this type of promiscuous enzyme inhibition by promoting the dissociation of enzyme from drug aggregates (37). Inhibition of DNA gyrase activity by NSC 103003 was not significantly affected by Triton X-100 at either 0.01% or 0.025% (Table 2), indicating that inhibition was not due to nonspecific drug aggregation effects. To determine whether the inhibition of DNA gyrase might involve drug intercalation into DNA, a DNA unwinding assay was done. NSC 130847 and NSC 20115 did not cause DNA unwinding at 100 and 1,600 μM, respectively (data not shown). However, NSC 103003 did cause significant DNA unwinding at concentrations of ≥50 μM, indicating that this compound may inhibit gyrase by altering DNA structure.

In silico docking and in vitro functional testing of compounds targeting the ATP binding pocket of DNA gyrase subunit B.

The ATP binding pocket of DNA gyrase subunit B is highly conserved between species, and crystallographic and biochemical evidence indicates that coumarin and cyclothialidine drugs act by competing with ATP for binding to the structural pocket defined by N46, E50, D73, R76, I78, K103, V118, and T165 in T. thermophilus and E. coli GyrB (Fig. 4) (25, 27, 30; reviewed in reference 35). There are currently no solved structures of wild-type E. coli DNA gyrase B bound to novobiocin. Therefore, we used the crystal structure of novobiocin bound to the highly conserved ATP binding pocket of T. thermophilus (Fig. 4) for our docking analysis. Similar to the strategy used for the A subunit of DNA gyrase, we screened approximately 140,000 molecules in the structural pocket defined by novobiocin bound to GyrB. The 10 highest-scoring compounds (Table 3 and Fig. 5) were then tested for inhibition of E. coli DNA gyrase supercoiling activity. One advantage of this approach is that it should selectively identify inhibitors that act through conserved binding motifs present in both T. thermophilus and E. coli enzymes. Such inhibitors would be expected to have a broader spectrum of antibacterial activity than that of inhibitors that target species-specific motifs.

NSC compounds 20116 and 7784 inhibited DNA gyrase activity, with IC₅₀ values of 338 and 814 μM, respectively (Table 3). The inclusion of Triton X-100 in the assay did not reverse inhibition by NSC 20116, suggesting that inhibition was not due to enzyme adsorption to drug aggregates (Table 2). Inhibition is unlikely to involve DNA intercalation since neither NSC 20116 nor NSC 7784 caused DNA unwinding at concentrations of up to 640 μM and 1.6 mM, respectively (data not shown).

Both NSC 20116 and NSC 7784 were also tested for inhibition of gyrase ATPase activity. Although NSC 7784 had no effect on ATPase activity at concentrations of up to 1.5 mM, NSC 20116 significantly inhibited ATPase activity, with a K_i of 149 μM (Fig. 4E), supporting the hypothesis that this latter compound targets the ATP binding pocket of GyrB.

The chemical structure of NSC 20116 differs significantly from those of known DNA gyrase inhibitors, suggesting a different mode of interaction with the ATP binding pocket (Fig. 4B to D). Cyclothialidine and novobiocin share a common binding motif involving an H bond donor to D73 and an H bond acceptor from a conserved water molecule (25, 27, 30, 48). These drugs also interact with residue R136, and an alteration in this residue to L, H, C, S, or A confers drug resistance (reviewed in reference 35). Although NSC 20116 is predicted to form a hydrogen bond with D73, it is not predicted to interact with R136 (Fig. 6). Therefore, mutations at R136 that confer resistance to coumarin and cyclothialidine drugs are not likely to cause cross-resistance to NSC 20116.

Inhibition of E. coli growth by NSC 103003 and NSC 20116.

Logarithmically growing E. coli XL1-Blue cells were incubated with various concentrations of either NSC 103003 or NSC 20116, and growth was monitored by following the absorbance of cell cultures at OD₅₉₅ (Fig. 7). NSC 103003 did not significantly inhibit growth in the micromolar range, although some inhibition was seen at 1.2 mM. This concentration was approximately 20-fold higher than concentrations required to inhibit purified DNA gyrase activity in vitro, suggesting that NSC 103003 does not effectively accumulate inside E. coli. Alternatively, the lack of growth inhibition by NSC 103003 could be due to drug instability or metabolic inactivation by bacterial enzymes. In contrast, NSC 20116 completely inhibited growth within 30 min at 192 μM, a concentration that inhibits purified DNA gyrase activity by >30%. There is an apparent dose-related delay in growth inhibition by NSC 20116 in the low micromolar range since growth inhibition was delayed for 30 or 60 min following the treatment of cells with 64 μM or 21 μM of drug, respectively (Fig. 7).