Structural and Computational Analysis of Pseudomonas aeruginosa DNA Gyrase Reveals Molecular Characteristics That May Contribute to Ciprofloxacin Resistance

Abstract

Pseudomonas aeruginosa is considered a priority pathogen by the World Health Organization due to its resistance to antibiotics. Isolates resistant to ciprofloxacin (CPFX), a bactericide commonly used against P. aeruginosa, usually carry the mutations T83I or D87N in the GyrA subunit of the DNA gyrase. Yet, the molecular mechanisms by which these mutations confer CPFX-resistance to P. aeruginosa are unknown. Here we solved the crystal structure of the P. aeruginosa gyrase catalytic cleavage core and used it to carry out molecular dynamic (MD) simulations of CPFX-gyrase binding in the wild-type as well as the T83I and the D87N mutant systems. Our results show that DNA plays the most prominent stabilizing role once CPFX is bound, with relatively minor contributions from Thr83 or Asp87. Interestingly, we found a solvent cavity adjacent to these residues that may provide CPFX access to the active site. Interaction energy analysis using Umbrella Sampling indicates that Thr83 and Asp87 may influence CPFX trajectory during binding. In the mutant systems, the repulsive potential increases at the cavity site, which may hinder CPFX accessing the binding site. These results shed light on P. aeruginosa resistance to CPFX and may help provide a methodology to identify new therapeutic agents to target fluoroquinolone resistant bacteria.

1. Introduction

Pseudomonas aeruginosa, a ubiquitous gram-negative bacterium, is an important opportunistic pathogen causing high morbidity and mortality. It is a major species contributing to nosocomial infections in patients with compromised immune systems, and the leading cause of chronic lung infections in cystic fibrosis and chronic obstructive pulmonary disease patients [1,2,3]. Infections are difficult to treat because P. aeruginosa is naturally resistant to high concentrations of many common, first-line antibiotics and has a remarkable ability to acquire mutations conferring resistance to antimicrobial agents [4,5,6,7,8]. The World Health Organization (WHO) has listed it as one of the six pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, P. aeruginosa, and Enterobacter species or ESKAPE organisms) as top species prioritized for research due to their development of resistance to antibiotics [9].

Quinolones are one of the few antibiotic classes that are effective against P. aeruginosa infections, with the fluoroquinolone ciprofloxacin (CPFX) among the most used [10]. Their intense application, though, has led to a rapidly increasing bacterial resistance to this group of antibiotics and treatment failure [11,12,13,14]. Resistance is often determined by chromosomal mutations affecting drug permeability (increased efflux or reduced levels of drug accumulation in the cells) or type II topoisomerase (DNA gyrase, the quinolones’ target) sensitivity to the drug [11,12,13,14]. In CPFX-resistant isolates of P. aeruginosa identified in the clinic, the most frequently found mutations are mapped to the gene gyrA, which codes for the gyrase subunit GyrA (reviewed in [10]). Mutations in GyrA are located in a domain termed the “Quinolone Resistance-Determining Region” (QRDR), which encompass amino acids 67 to 106, with the most common being the specific mutations GyrA T83I and GyrA D87N [10,12,15].

DNA gyrase is an ATP-dependent bacterial type II DNA topoisomerase that performs the critical function of introducing negative supercoils into a double-stranded DNA (dsDNA) to prevent hyper-supercoiling. DNA gyrase is a heterotetrameric enzyme comprising two heterodimers (complex I and II), each containing one GyrA and one GyrB subunit (Figure 1A). Both subunits are involved in the binding and cleaving of DNA; GyrB also hydrolyzes ATP. To produce negative supercoils, gyrase binds two close segments of dsDNA: a G (or Gate) segment, and a T (or Transported) segment (Figure 1A). Then it cleaves both strands of the G-segment through nucleophilic attack by a tyrosine located in each of the GyrA subunits, leaving a four-base stagger. This results in the formation of a reversible covalent bond between the newly generated 5′-phophates and the catalytic tyrosine residues (Figure 1B). DNA cleavage is synchronized with ATP hydrolysis-facilitated passage of the T-segment through the G-segment and DNA-gate, followed by re-ligation of the G-segment via the reverse reaction and release of the DNA strands (reviewed in [16,17,18]).

Quinolones inhibit DNA gyrase through the formation of a stable ternary complex with the gyrase and covalently bound DNA whereby two quinolone molecules intercalate the DNA at both nick sites of the G-segment (Figure 1B), thus restraining the gyrase from re-ligating this segment (reviewed in [13,14,20,21,22]). These complexes can be reversed at low quinolone concentrations, but high drug concentrations produce dsDNA breaks that lead to chromosomal fragmentation and cell death (reviewed in [14,20]).

Analyses of wild-type and quinolone-resistant mutant gyrase cryo-EM and crystal structures alone or in combination with molecular docking and molecular dynamics (MD) simulations are useful means for understanding the drug–gyrase interaction and the drug mechanism of inhibition. Such structural studies provide the basis for the design of novel antibacterial agents or modification and implementation of existent ones (e.g., [19,21,23,24,25,26,27,28]). Partial or complete crystal or cryo-EM gyrase structures exist for Escherichia coli [29,30], S. aureus [19,31], A. baumanii [23], Thermus thermophilus [32], Streptococcus pneumoniae [33,34], and Mycobacterium tuberculosis [24], but not for P. aeruginosa. To date, dynamic models of P. aeruginosa gyrase are based on the gyrase structures of other bacteria (e.g., [10,15,35]), which may render an incomplete picture of how resistant-mutant versions of this enzyme interact with different drugs.

Here, to understand the fluoroquinolone resistance mechanism in P. aeruginosa at an atomic level, we examine how the GyrA mutations T83I and D87N modify P. aeruginosa gyrase interactions with CPFX (Figure 1C). We generated a P. aeruginosa GyrB/GyrA fusion construct that concatenates the DNA binding/cleavage domains and solved its structure using crystallography in the apo- and DNA/CPFX-bound forms. These structures were used as starting models for MD analysis of wild-type (WT) and mutant proteins to examine how the T83I and D87N mutations may contribute to CPFX resistance. The combined structural/computational analysis revealed that these residues lie along a potential solvent entry channel providing CPFX access to the active binding site. How these residues impact access and may affect resistance is discussed.

2. Materials and Methods

2.1. Cloning, Expression, and Purification of P. aeruginosa Gyrase Catalytic Cleavage Core

A construct (referred to as gyrB-GGS-gyrA525) composed of full-length P. aeruginosa gyrB (coding for residues 1 to 806)-linker (GGS)-truncated gyrA (coding for residues 2 to 525), optimized for expression in E. coli, was obtained through gene synthesis (Genewiz, South Plainfield, NJ, USA). A truncated gyrB (coding for residues 397 to 806)-linker (GGS)-truncated gyrA (coding for residues 2 to 525) construct (referred to as gyrB397-GGS-gyrA525) (Figure 1A) was amplified using the synthetized version as template and the following primers: forward primer, 5′-TCGgcggccgcAATGAAGGGTGCGCT-3′; reverse primer, 5′-TAGaagcttTTAGCTGGCCACAAT-3′. (Lowercase letters denote NotI and HindIII restriction sites, respectively). Both constructs were cloned into a pGEXM vector based on the pGEX4T3 backbone (Cityva, Marlborough, MA, USA) using the NotI/HindIII restriction sites, and sequenced [36]. pGEXM vectors carrying the constructs were transformed into Rosetta2(DE3) cells (Millipore Sigma, Burlington, MA, USA) for expression. Overnight (ON) cultures were grown at 37 °C in shaker flasks in Terrific Broth (TB) media containing 100 µg/mL ampicillin (Millipore Sigma, Burlington, MA, USA) and 35 µg/mL chloramphenicol (Millipore Sigma, Burlington, MA, USA). The next day, the ON cultures were added to Fernbach flasks containing 1 L of TB media and antibiotics and placed at 37 °C on shaker. When the OD₆₀₀ reached 0.9, the temperature was set to 18 °C. After 45 min, IPTG (isopropal-β-D-1-thiogalactopyranoside; GoldBio, St. Louis, MO, USA) was added to a concentration of 200 µM, and the cultures were shaken ON at 18 °C. Cells were then pelleted by centrifugation at 4000 g for 15 min and resuspended in sonication buffer consisting of 25 mM Tris pH 7.5 (Medical Inc., Columbia, MD, USA), 500 mM NaCl (Fisher Chemicals, Ottawa, ON, Canada), 1 mM EDTA (Millipore Sigma, Burlington, MA, USA), and 1 mM DTT (VWR Chemicals, Solon, OH, USA), followed by sonication. Cell debris were pelleted by centrifugation for 35 min at 47,900 g. The soluble fraction was equilibrated with Glutathione 4B Sepharose resin (Cytiva, Marlborough, MA, USA; 1 mL resin per liter of cell growth) in batch on a rocker at 4 °C for 1.25 h, and the resin was washed multiple times in batch with sonication buffer. Protein was removed from the resin-bound glutathione-S-transferase with TEV (Tobacco Etch Virus, expressed in house; 100 µL of 1.9 mg/mL TEV per ml of resin) in batch on rocker at 4 °C ON. Protein was concentrated, then further purified by gel filtration on a Superdex200 gel filtration column (Cytiva, Marlborough, MA, USA) equilibrated in 25 mM Tris pH 7.5, 75 mM NaCl, 1 mM DTT. Eluted fractions containing protein were pooled and concentrated (Figure S1). While the gyrB-GGS-gyrA525 construct produced little protein, the gyrB397-GGS-gyrA525 overexpressed well and could be concentrated to ~19 mg/mL for crystallization experiments.

2.2. Crystallization and Structure Determination of P. aeruginosa Gyrase Catalytic Cleavage Core

Crystals of the apo gyrB397-GGS-gyrA525 protein were grown at 4 °C using the vapor diffusion hanging drop method. This was originally an attempt to obtain a protein/DNA complex as the protein used consisted of 17.9 mg/mL protein mixed with an annealed 24mer duplex DNA (2:1 DNA to protein ratio; 24merF GGTCATGAATGACTATGCACGTAA, 24merR TTACGTGCATAGTCATTCATGACC) in 25 mM Tris pH 7.5, 150 mM NaCl, and 5 mM MgCl₂ (Millipore Sigma, Burlington, MA, USA), that was allowed to equilibrate for 4 h at 4 °C prior to setup. For crystallization, a solution consisting of 2 µL of protein/DNA complex was mixed with 2 µL of the reservoir consisting of 100 mM HEPES pH 7.5 (Millipore Sigma, Burlington, MA, USA), 5% PEG 8000 (Microlytic, Burlington, MA, USA), and 26% MPD (Hampton Research, Aliso Viejo, CA, USA) equilibrated for one day against the reservoir prior to streak seeding. The crystal used for data collection was harvested after six days, transferred to a cryo-solution consisting of 100 mM HEPES pH 7.5, 40% MPD, and 5% PEG 8000, then flash frozen in liquid nitrogen. Data were collected on the Berkeley Center for Structural Biology beamline 5.0.2 at The Advanced Light Source, and processed using HKL3000 (HKL Research Inc., Charlottesville, VA, USA) [37]. Protein Data Bank coordinates (PBD) 3NUH were used as a search model for molecular replacement in PHASER [38,39]. The model was refined with multiple cycles of refinement in PHENIX and manual model building in COOT [40,41,42]. No DNA was visible in the electron density. Several regions within the structure lack sufficient electron density to be modeled: GyrB residues 397–403, 453–461, 503–504, 570–575, 703–706, 717–720, and 798–806; GGS linker; GyrA residues 2–5, 175–176, 401–405, 415–420, and 427–430). Data statistics and model analysis are shown in

Table 1. Crystallographic data collection and refinement.

	Apo	DNA + CPFX
PDB ID code	9NX5	9NX4
Data collection
Space group	P6322	P6322
a, b, c (Å)	218.159, 218.159, 108.803	179.033, 179.033, 163.239
α, β, γ (°)	90, 90, 120	90, 90, 120
Resolution Range (Å)	50.00–2.40 (2.44–2.40)	50.00–2.85 (2.90–2.85)
Rmerge (%)	9.9 (125.9)	16.8 (125.2)
Rmeas (%)	10.2 (130.7)	17.4 (129.6)
Rpim (%)	2.6 (34.5)	4.3 (32.7)
CC1/2	0.994/0.785	0.992/0.815
I/σI	1.5	1.3
Completeness (%)	100.0 (100.0)	99.2 (99.8)
Redundancy	15.7 (14.0)	15.5 (15.3)
No. reflections	939,593	564,175
Refinement
Rwork/Rfree (%)	24.18/25.67	20.50/23.44
No. atoms	6746	6204
Protein	6578	5696
Solvent	141	16
DNA	-	407
CPFX	-	48
Average B-factors (Å2)
Protein	71.43	57.58
Solvent	52.74	46.63
DNA	-	67.01
CPFX		61.75
Clashscore	0.85	0.66
R.m.s. deviations
Bond lengths (Å)	0.003	0.008
Bond angles (°)	0.563	0.954
Ramachandran Plot
Favored (%)	98.93	97.12
Allowed (%)	0.95	2.88
Outliers (%)	0.12	0.00

.

Crystals of the DNA\CPFX-bound gyrB397-GGS-gyrA525 protein were grown at 21 °C using the vapor diffusion hanging drop method. To form the complex, 16 mg/mL of protein was equilibrated with an annealed 20mer duplex DNA (3:1 duplex DNA to protein ratio; 20mer AGCCGTAGGGCCCTACGGCT) in 25 mM Tris pH 7.5, 75 mM NaCl, 1 mM MnCl2 (Millipore Sigma, Burlington, MA, USA), 1 mM DTT, and 1 mM CPFX, and allowed to equilibrate for ~5 h. For crystallization, 2 µL of protein/DNA/CPFX complex were mixed with 1 µL of reservoir solution consisting of 100 mM sodium acetate pH 6.0, 200 mM ammonium sulfate, and 14% PEG 4000 (Microlytic), and equilibrated for one day against the reservoir prior to streak seeding. Crystal was harvested after 11 days and flash frozen in liquid nitrogen in a cryo-solution consisting of 100 mM sodium acetate pH 6.0, 200 mM ammonium sulfate, 15% PEG 4000, 20% ethylene glycol, and 1 mM CPFX. Data were collected on the APS SER-CAT 22ID beamline at Argonne National Laboratory. The apo model of gyrB397-GGS-gyrA525 was used to solve the molecule replacement for the DNA\CPFX-bound structure using PHASER [38]. The DNA and CPFX were built into the map, and the model was refined with multiple cycles of refinement in PHENIX and manual model building in COOT [40,41,42]. The following residues lack sufficient density to be modeled: GyrB 397–404, 479–482, 562–735 (insertion domain), and 799–806; GGS-linker; GyrA 2–6. Quality of both the apo and DNA-bound structures were assessed using MolProbity [43]. The apo and DNA-bound structures have been deposited in the PDB with ID codes 9NX5 and 9NX4, respectively, with data statistics and model analysis results reported in Table 1.

2.3. Determination of P. aeruginosa Gyrase Catalytic Cleavage Core Construct Activity

Determination of gyrB397-GGS-gyrA525 activity and inhibition by CPFX was performed according to Nitiss et al., 2012 [44]. In brief, 20 µL reaction mixtures were prepared with 11 µL of H₂O, 2 µL of 10× buffer (ProFoldin, Hudson, MA, USA; P. aeruginosa DNA gyrase assay kit plus-100; final concentrations are: 20 mM Tris-HCl pH 8, 35 mM NH₄OAc, 4.6% glycerol, 1 mM DTT, 0.005% Brij35, 8 mM MgCl₂), 1 and 2 µL of purified gyrB397-GGS-gyrA525 (1 and 2 µM final concentrations), 1 µL of pUC18Ω-Tc (obtained from Dr. Joanna B. Goldberg, Emory University of Medicine, Atlanta, GA, USA; 30 nM final concentration), 2 µL of 10 mM ATP (ProFoldin; 1 mM final concentration), and 2 µL of either H₂O or CPFX solution (final concentrations 2, 4, and 8 µg/mL). Positive controls included two reactions with commercial E. coli gyrase (Millipore Sigma, Burlington, MA, USA; 1 and 2 µL of a 2 units/µL solution), and negative controls included all components except the gyrase. Reaction mixtures were incubated at 37 °C for 1 h and stopped by adding 5 µL of 5 × stop solution (Topogen, Buena Vista, CO, USA; 5% sarkosyl, 0.125% bromophenol blue, 25% glycerol). Proteinase K solution (Thermo Fisher, Waltham, MA, USA; 20 mg/mL) was then added to a final proteinase concentration of 0.4 mg/mL, and the reaction mixtures were incubated at 37 °C for 30 min. A sample (20 µL) of the mixtures was loaded in a 0.8% agarose gel. After electrophoresis, the gel was stained for 30 min in an ethidium bromide solution before visualization (Figure S2).

2.4. Molecular Dynamics (MD) Simulations

The initial structures of GyrB (residues 405–561 and 736–798) and GyrA (residues 7–525) of P. aeruginosa were derived from the X-ray crystal structure (PDB ID 9NX4). Missing residues were added using Modeller-10.2 [45]. DNA nucleotides spanning the nick sites (Figure 1B) were introduced based on the structure of M. tuberculosis (PDB ID 5BTC) [24]. CPFX molecules were realigned, and magnesium ions were incorporated as in the crystal structure mentioned above.

To prepare the starting configurations for MD trajectories, missing atoms and protons were added using the LEaP module of Amber.22 [46]. Counter ions were introduced, and the system was solvated in a box of TIP3P water, with the box boundary extending 20 Å from the nearest peptide atom. This setup resulted in a simulation box containing 279,472 atoms. Histidine residues 132, 195, 358, 379, and 465 in GyrB were δ-protonated, while the remaining histidine residues were ε-protonated. Charge neutralization and a 100 mM effective salt concentration, approximating physiological conditions, were achieved by adding 359 Na⁺ and 231 Cl⁻ ions.

Before equilibration, the solvated system underwent the following sequential steps:

(1)

Belly dynamics: 500 ps of simulation with fixed peptide. Constant volume constant temperature (NVT) simulation was performed at 100 K to move water molecules and ions with 1 fs time step.

(2)

Energy minimization: 5000 steps. (Constant volume; all atoms were allowed to move to remove hotspots with the steepest descent optimization).

(3)

Low-temperature dynamics: 1 ns of constant-pressure simulation at 200 K with a fixed protein, ensuring a reasonable starting density. (NPT; 1 fs time step at 1 atm).

(4)

Second energy minimization: 5000 steps. (Constant volume; all atoms were allowed to move with the steepest descent optimization).

(5)

Stepwise heating: Constant-volume MD simulations increasing the temperature from 0 K to 300 K over 3 ns. (NVT; 1 fs time step).

(6)

Constrained equilibration dynamics: 20 ns of constant-volume simulation with a 10 kcal/mol constraint force applied only to backbone heavy atoms. (NVT; 1 fs time step).

After a 30 ns equilibration period during which all constraints were removed, sampling was extended through three independent, constant-temperature, constant-volume MD simulations of 1 μs each. All MD trajectories were computed using the PMEMD (Particle Mesh Ewald Molecular Dynamics) module of Amber.22 with a 1 fs time step. Long-range Coulombic interactions were calculated using the PME method with a 10 Å cutoff for direct interactions. Amino acid parameters were derived from the FF14SB force field of Amber.22. CPFX atomic charges were calculated using ChelpG procedure of Gaussian.16 [47]. The Amber prepc file containing the atom types and partial atomic charges used in CPFX to construct the Amber forcefiled file is given in Table S1. The magnesium ion was treated as a divalent cation (+2e charge) with the Lennard Jones parameters taken directly from the standard Amber forcefield.

Configurations selected at each nanosecond of the production runs were used for all analyses utilizing the Cpptraj module of Amber.22 [48]. Interaction-free energies between gyrase residues and CPFX were calculated using the MM/PBSA (Molecular Mechanics/Poisson–Boltzmann Surface Area) protocol of Amber.22 [49]. These calculations were based on 1000 configurations extracted at each nanosecond interval from each trajectory, resulting in a total of 3000 configurations. An ionic strength of 0.1 M was selected for the MM/GBSA (Molecular Mechanics/Generalized Born Surface Area) calculations. Videos of the trajectories were generated using Chimera 1.18 [50].

Sets of MD runs were conducted for the WT as well as for the T83I and D87N mutant systems, with and without CPFX. The original WT simulation system was modified to construct these mutant systems. MD simulations were performed in triplicates, and the analysis methodology mirrored the approach described above for the WT system. Table S2 summarizes information on all systems studied in this work.

MD simulations in combination with Umbrella Sampling were carried out to evaluate the potential effects of specified mutations on CPFX binding. An area, identified from the original simulations, involving residues D87 and T83, referred to now as “the cavity site”, was deemed crucial as the entry point for CPFX along its transition pathway. CPFX was driven out from its binding site towards the cavity site by selecting 35 windows with 0.4 Å width along the direction from the bound position of CPFX towards the cavity site. For each window, a 1.2 ns of MD run was performed with the anchor strength of 100 kcal/mol. To apply the forces, a line along the center of mass of CPFX and the center of mass of four phosphates from the DNA that reside next to the cavity site was selected. Five hundred configurations chosen from the last 1 ns of each window were used in the MM/GBSA energy analysis. Two types of MM/GBSA calculations were performed: (1) selecting CPFX as the ligand to calculate the residue contribution for interaction energies, and (2) selecting CPFX and the magnesium ion bound to it to calculate the MM/GBSA binding energies for the CPFX/Mg complex.

3. Results

3.1. Structural Analysis of P. aeruginosa Gyrase Catalytic Cleavage Core

Previous structural studies on bacterial DNA gyrases have demonstrated the benefits of fusing the N-terminus of GyrA to the C-terminus of GyrB [22,24]. In this report, we demonstrate that a P. aeruginosa gyrase catalytic cleavage core construct composed of residues 397–806 of GyrB fused with residues 2–525 of GyrA via a GlyGlySer linker (gyrB397-GGS-gyrA525) (Figure 1A) is functional, i.e., can generate DNA double strand breaks (Figure S2). To better understand how the T83I and D87N GyrA mutations may contribute to CPFX resistance in P. aeruginosa, we solved the structure of the gyrB397-GGS-gyrA525 construct with and without DNA/CPFX bound (Figure 2). The apo form crystallized with one molecule in the asymmetric unit forming the expected dimer with a neighboring symmetry-related molecule in the crystal (Figure 2A). While these structures are the first reported for P. aeruginosa, ample gyrase structural data exists for E. coli, which shares with P. aeruginosa ~72% sequence identity within the gyrase catalytic cleavage core included in our construct [30,39,51]. Based on superposition, the apo catalytic cleavage core from both enzymes displays good agreement in the positions of the individual domains except for the domain unique to gram-negative bacteria, the insertion domain, that in P. aeruginosa pivots by ~23° from the connection to the TOPRIM domain, resulting in a shift of greater than 20 Å for the remote residues (Figure S3A,B) [39].

Attempts were made to crystalize gyrB397-GGS-gyrA525 with a duplex DNA substrate spanning the active site of the dimer utilizing 20mer and 24mer duplexes as had been done previously for other gyrase/topoisomerase II crystal structures [22,24]. Crystals were obtained in the presence of CPFX and a 20mer palindromic DNA sequence that resulted in one strand of DNA and one molecule of the catalytic cleavage core (GyrB/GyrA) in the asymmetric unit, forming the expected dimer with a crystallographic symmetry mate (Figure 2B). In this crystal structure, most of the insertion domain is disordered. Interestingly, rather than obtaining the expected structure of the duplex DNA oligo binding across the active site of the two molecules in the crystallographic dimer, each active site is binding to the blunt end of the duplex DNA (Figure 2B and Figure 3A). This DNA-bound form of P. aeruginosa’s catalytic cleavage core also shows good agreement with the CryoEM structure of the full length E. coli gyrase that includes the N-terminal ATPase domain of GyrB and the C-terminal pinwheel domain of GyrA in a closed conformation binding DNA (Figure S3C) [30]. While the blunt end superimposes well with the position of DNA in other gyrase structures such as in E. coli, S. aureus and M. tuberculosis (Mtb), it leaves a gap that could accommodate four nucleotide base pairs between the two ends, consistent with spacing in other structures (Figure 4A) [22,24,30]. For P. aeruginosa, CPFX is found stacking with the blunt end of the DNA on one side and a solvent citrate molecule positioned on the other (Figure 4A,B). While the location of CPFX binding to the complex is similar to what has been seen in the S. aureus and Mtb structures, the relative orientation of CPFX is different, possibly due to the lack of duplex DNA on both sides of the drug (Figure 4A) [22,24].

As seen for other gyrases such as in E. coli [30,39], structural alterations take place in P. aeruginosa gyrase upon binding DNA (Figure 3). Like in E. coli, the TOPRIM domain of GyrB undergoes a large change including the C-terminal helix moving away from the active site allowing DNA to bind (Figure 3A,B). DNA binding also results in shifts in the Tower and N-terminus of the WHD domain of GyrA (Figure 3B) as well as the insertion domain of GyrB becoming mostly disordered (Figure 3C). Upon this movement, the Arg790 from the GyrB C-terminal helix becomes ordered and interacts with the DNA (Figure 3A, red asterisk). Residues on the loop between Ile452 and Asp461 become ordered as well, and shift to the position of the C-terminal helix in the apo structure with previously disordered residues Asn454 and Lys457 forming interactions with the DNA. The GyrA loop containing residues Ala175 to Met178 shifts to accommodate the DNA with the backbone amide of Gly177, forming a hydrogen bond to a DNA backbone phosphate. These changes further help to position the catalytic Tyr122 from the other molecule in the dimer, closer to the active site. The position of Tyr122 superimposes well with that of the inactive mutant Y123F of S. aureus [22], supporting it in a catalytically relevant position (Figure 4A). Residues Thr83 and Asp87, mutations of which confer CPFX resistance, are found to be adjacent to the blunt end of the DNA and CPFX in this study (Figure 3A).

Although the DNA did not bind across the active site as expected, it did create a closed form of the enzyme that resembles other gyrases structures with regions of DNA and active site residues in positions consistent with catalytically relevant structures (Figure S3C and Figure 4), generating a good starting model (Figure S4) for MD studies of the effects of mutations T83I and D87N on CPFX resistance.

3.2. T83I and D87N Mutations Do Not Drastically Change Structure of the Gyrase/DNA Complex or CPFX Binding

To gain mechanistic insight into P. aeruginosa CPFX resistance caused by T83I and D87N mutations in GyrA, we employed MD simulations to address the effects of these mutations on CPFX binding. We used six different systems (Materials and Methods, Table S2). The WT system was generated to serve as a reference; both GyrA molecules in the heterotetrameric system were mutated to generate either the T83I or the D87N mutant. Each of the three systems were subjected to MD in triplicate with and without CPFX bound at both active sites. To validate that these systems were dynamically stable, the Root Mean Squared Deviations (RMSDs), a measure of the movement of atoms within a structure compared to those within a reference structure, were assessed. RMSDs of all backbone atoms stayed below 4 Å (Figure S5), except in the case of the CPFX-free T83I mutant (Figure S5B), for which the final values were slightly above 4 Å in one of three simulations. In most cases, GyrB has larger RMSDs than GyrA, mainly due to the contributions from large deviations culminating from the termini of two peptide fragments (residues 405–561 and 736–798) representing GyrB. Overall results indicate that the protein assemblies during simulations were stable on a global scale with or without bound CPFX during our simulation timescale (1000 ns). All CPFX-bound structures displayed slightly reduced RMSDs compared with their ligand-free counterparts (Figure S5).

CPFX, in its bound conformation, has its quinoline ring stacked between the four DNA bases at the nick site (Figure S4). Accordingly, all RMSD values of the heavy atoms of CPFX are found to be close to 0.5 Å at the beginning of each production run and remain low throughout, suggesting that CPFX is tightly bound (Figure S6). Owing to the free rotation of the piperazine group around the connecting N-C bond to the quinolone core, the RMSDs can increase to values around 1Å for some systems during molecular dynamic simulation time frames (Figure S6). On average, it displayed a deviation of about 0.75 Å in the center of mass position (Figure S6), even after this rotation of the piperazine group was considered. Still, this represents a rather low deviation, given the fact that even the smallest Root Mean Squared Fluctuations (RMSFs) for individual backbone atoms of the amino acid residues of GyrA and GyrB surrounding the active site fall within this range (Figure 5). RMSF values represent the local fluctuations of each amino acid residue, and our MD runs do not show obvious difference between the RMSFs of the CPFX-bound and -unbound forms for the WT system or mutant systems (Figure 5). This stability of CPFX in the bound form may be attributed to the initial placement and orientation of CPFX that was based on the X-ray crystal structure of gyrase from Mtb with the complete DNA segment present (PDB ID 5BTC) [24]. The RMSF analysis shows that WT and mutant systems behave in a very similar way, irrespective of the presence of CPFX (Figure 5). This suggests that these mutations have little effect on the local folding of CPFX-bound structures, and, therefore, change in stability of CPFX-bound mutant complexes would likely not be a factor contributing to acquisition of antibiotic resistance.

Dynamic Cross-Correlation (DCC) analyses of alpha carbons, employed to search for correlated motion between residues revealing functional networks, clearly indicate that, in the WT system, the segment of the GyrA directly in contact with the segment of GyrB or DNA has highly positive correlations, while the non-interacting portions (residues 350–513 of GyrA) distinctly display their own dynamics. This is true for each of the GyrB/GyrA complexes (I and II) (Figure 6). Further analysis of the cross-correlation behavior between complexes I and II reveals similar but much weaker correlation (Figure 6, top and middle right panels). The coiled-coil region (residues 348–354) of GyrA shows the least correlation with the adjacent complex in each system except for the gate residues at the base of the structure (Figure 2B; ~residues 390–470), which strongly correlates with the adjacent complex’s gate residues due to the surface contacts. No appreciable differences between the CPFX-free and -bound forms of the WT system points to having similar dynamical behavior irrespective of the presence of the drug (Figure 6, bottom panels). The same tendencies in the pattern of interactions were observed for both mutant systems (Figure 7). These data suggest that dynamics of interactions between GyrA and GyrB in mutant complexes do not differ significantly from the WT complexes. Moreover, collinearity of the protein dynamics of these complexes with or without the bound ligand CPFX, either in mutant or WT systems, argues that such dynamics do not underlie antibiotic resistance of mutant complexes.

3.3. DNA Plays a Significant Role in Maintaining the CPFX-Bound Conformation

Molecular Mechanics with Generalized Born and Surface Area (MM/GBSA) solvation energy calculations were used to quantify the residue (or nucleotide) contribution for CPFX interaction energy. Only the attractive interaction energies (negative values) are displayed (Figure 8A–C). At most residue Thr83 of WT GyrA exhibits appreciable, attractive binding interaction to CPFX with solvation energies of approximately −3 kcal/mol (Figure 8A, Thr83, white circles). In the T83I system, the interaction strength decreases by about 2 kcal/mol (Figure 8A, compare Thr83 WT (white circles) and T83I (red circles)). The interaction energy with Asp87 was undetectable and not shown. Surprisingly, a decrease in the interaction energy of Thr83 with CPFX is seen to a lesser extent even in the D87N system, indicating subtle local changes due to this mutation (Figure 8A, Thr83, blue circles). In addition, we report here a small but detectable interaction with GyrA Arg121. In fact, this residue belongs to GyrA of the heterodimer complex that is not primarily involved in CPFX interactions (Figure 8A, Arg121). Several residues of GyrB contribute to CPFX interactions, with Lys449 standing out as particularly prominent across all systems due to its proximity to CPFX (Figure 3A and Figure 8B). Other interacting residues, including Leu405, Gly450, and Glu468, show reduced interaction strengths of one or more kcal/mol (Figure 8B) compared to Lys449. Based on the MM/GBSA energy values of DNA nucleotide interactions with CPFX (Figure 8C), we concluded that the DNA plays the most significant role in trapping CPFX between the nucleotide bases. As shown in Figure 8D,E, the most significant interaction energy contribution comes specifically from the nucleotide bases directly engaging with CPFX, while neighboring nucleotides exhibit only residual interactions. Interestingly, CPFX in the crystal structure (Figure 4) is found binding against the equivalent base to dA25 in the simulation (Figure 8D,E), which shows the greatest contribution to binding (Figure 8C). This suggests its presence is enough to energetically bind CPFX, but base dC24 of the four-nucleotide gap helps to properly position CFPX, as seen in other crystal structures and utilized for this simulation [22,24].

3.4. Reverse MD Suggests the Importance of a Solvent Cavity Adjacent to Thr83 and Asn87 for CPFX Access

Since we did not find compelling evidence that the mutations T83I or D87N have a significant effect on gyrase structure or CPFX-binding dynamics, we focused on further investigating the dynamic process of CPFX binding rather than on the bound conformation itself. Notably, all structural analyses reveal a water-filled cavity adjacent to the bound CPFX that sits directly adjacent to both Thr83 and Asp87 (Figure 8D). The cavity is formed around the residue segment Asp82–Asp87 and residues Val90, Arg91, Gln94, Phe96, Ser97, Asn116, Ala117, Met120, Asn268, Asn270, Ala272, and Arg273, along with the backbones of the residue segment Ser111–Asp115, and hereafter referred to as the “cavity site”. While CPFX does not directly interact with Asp87 (Figure 3A and Figure 8D) in the bound form, we believe that the cavity site might offer insights into how Asp87 influences CPFX’s positioning between nucleotides for its function.

We hypothesize that in its path to the inhibition binding site, CPFX first diffuses to the cavity site from its solvated state before it finds its destination between the nucleotide bases. There is an additional small cavity that could act as another channel to enter the inhibition site located right above the four nucleotides that sandwich CPFX. However, the net negatively charged CPFX would have to travel near the DNA backbone, if it were to use this as the entry point. The small size of this cavity and the negative charge likely work against utilizing this second site, and therefore we did not further investigate CPFX binding via this cavity.

To explore CPFX access to the cavity site, we performed a reverse experiment, using MD and the Umbrella Sampling technique to guide CPFX into the mentioned cavity from its binding site, a path that is illustrated in Figure 8D. Specifically, we steered CPFX away from its bound position and constrained it within a window of 0.4 Å thickness, applying forces to its atoms. Each window was sampled for 1.2 ns; and 500 samples, collected from the final nanosecond of each window, were subjected to MM/GBSA energy analysis to determine the interaction energies of CPFX with Thr83 or Ile83 and Asp87 or Asn87.

Figure 9 illustrates the changes in distance between the oxygen atoms in CPFX and the gamma-oxygen of Thr83 and Asp87 over time, as shown by a representative Umbrella Sampling trajectory. Initially, Thr83 was in close contact with one or more oxygen atoms (sampling windows 1–12). Later, Thr83 and CPFX separated over the course of the simulation. In contrast, the distal gamma-carbon of Asp87 moved closer to one or more CPFX oxygen atoms within the same trajectory, shifting to CG (gamma-carbon) distances around 5 Å away from its original bound state for a set of sampling windows (10 to 20).

MM/GBSA energy analysis shows that, in the WT system (Figure 10A), Thr83 presents a distinct attractive center for CPFX within a range of 0–5 Å from its inhibitory binding state, with an interaction strength between 2 and 5 kcal/mol. Conversely, Asp87 exhibits a clear attraction to CPFX at a distance from its inhibitory binding state, with interaction magnitudes ranging from approximately 1 to 5 kcal/mol. The combined effects of these interactions may facilitate CPFX’s entry into the cavity site before it becomes locked between the two sets of nucleotide bases in the DNA.

In the T83I mutant system (Figure 10B), there is a suppression of the attractive well observed for Thr83 in the WT system at a 4 Å distance, accompanied by modifications of the Asp87–CPFX interactions as well. On the other hand, in the D87N system (Figure 10C), the Asn87–CPFX interactions are somewhat stronger across a broader distance range. Interestingly, Thr83–CPFX interactions within this complex are also altered (compared to Figure 10A, left panels), showing a weaker attractive potential than in the WT complex.

Since the cavity site may provide access to the inhibitory bound state of CPFX, our findings argue that the T83I mutation may have a direct effect on the process of CPFX entering the inhibitory binding site, with its greatest contribution when the drug has moved 2 to 4 Å from its bound state. This favorable interaction diminishes when Thr83 is mutated to isoleucine (Figure 10B). For Asp87, the interaction energy is the greatest when CPFX has moved from the inhibitory position to the cavity site. As CPFX progressed further into the cavity and approached Asp87, the interaction energy in the WT system increased (Figure 10A).

Throughout the Umbrella Sampling simulations, we consistently observed that the magnesium ion closely followed CPFX across all systems. Notably, positional restraints were applied only to the center of mass of CPFX, leaving the orientation of the CPFX and the magnesium ion unrestrained. This behavior supports treating CPFX and magnesium as a unified ligand complex in MM/GBSA calculations to evaluate its interaction energy with the surrounding system.

Interaction energies between the CPFX/Mg complex and the rest of the system—specifically the protein and DNA segment—were computed and are presented in Figure 11. In the initial sampling windows, CPFX resides near its bound conformation, where the WT and mutant systems exhibit comparable energetics (starting Sample Window, Figure 11). These findings are corroborated by structural dynamics and residue-level interaction analyses.

On average, the first five windows yield interaction energies of −0.68 ± 0.50, −0.88 ± 0.55, and −2.08 ± 0.45 kcal/mol for the WT, D87N, and T83I systems, respectively. However, in the cavity site sampling windows, both mutant systems display significantly higher repulsive interaction energies. The WT system shows an average interaction energy of 3.50 ± 0.44 kcal/mol in the last five windows, while the D87N and the T83I mutants exhibit markedly higher values of 13.55 ± 0.93 and 14.28 ± 0.80 kcal/mol, respectively.

These results suggest that the cavity site is energetically unfavorable for the mutant systems, potentially hindering their ability to traverse this region and reach the drug’s ultimate binding site.

4. Discussion

P. aeruginosa represents a clinically relevant target for new antibiotics due to its prevalence, severity of infection outcomes, and intrinsic, acquired, and adaptive resistance to antibiotics [52,53]. Two mutations in the DNA gyrase that have been shown to render P. aeruginosa resistant to fluoroquinolones such as CPFX are T83I and D87N [10,12,15]. However, how these mutations contribute to P. aeruginosa resistance is not well understood. To address this clinically relevant question, we structurally characterized the catalytic core of the WT gyrase relevant to CPFX resistance and performed molecular dynamics (MD) to evaluate the effect that these mutations might have on CPFX binding.

The apo crystal structure showed good agreement with the previously determined structure of the E. coli gyrase except for a shift in the position of the insertion domain (Figure S3A). DNA binding results in conformational changes surrounding the active site to accommodate the DNA and catalytically relevant positions of Tyr122 residues for DNA cleavage. Surprisingly, in our study, the DNA binds with two blunt ends in the active site. This structure may mimic that of the complex post-melting of the four DNA base pairs between the two nicks in the G-segment by leaving a four-base pair gap through which the T-strand would pass to supercoil the DNA. The positions of the DNA and catalytic Tyr122 are in good agreement with the configuration of the equivalent residues reported for other gyrase structures [22,24,30]. CPFX is also found binding at the same position as in the S. aureus and Mtb structures, although the relative orientation differs, potentially due to the lack of the stacking base pair on the gap side (Figure 4A) [22,24].

MD simulations have previously been employed to investigate delafloxacin (also a fluoroquinolone) resistance in S. aureus gyrase [28], revealing that the mutations studied (S84L/E88K, S84L/S85P, and single mutation S84L) do not induce significant structural perturbations upon drug binding. Notably, the WT gyrase exhibited the most negative binding-free energy, suggesting it offers the most favorable environment for delafloxacin interaction. The study further shows that the mutations increased the plasticity within the CPFX-binding pocket and elevated the atomic mobility of the mutated proteins, as demonstrated by principal component analysis. Interestingly, residues Ser84, Ser85, and Glu88, located within the fluoroquinolone-binding pocket (residues Thr83 and Asp87 in P. aeruginosa correspond to Ser84 and Glu88 in S. aureus) were found to neither directly interact with delafloxacin nor significantly contribute to its binding energy [28]. However, the absence of DNA in the MD simulations performed in the quoted study may give an inaccurate description of the inhibitory site, since CPFX is found sandwiched between DNA bases in the gyrase crystal structure with DNA bound [22,24].

Using the GyrA crystal structure of Colwellia psychrerythraea, Sada et al. [15] constructed a homology model of the P. aeruginosa GyrA along with the GyrA T83I and the GyrA D87N single mutants, and the double mutant GyrA T83I/D87N, and performed a docking study with CPFX. The affinities calculated directly from the docking poses using the GyrA systems resulted in very similar binding energies for the WT and mutant systems (−6.98 ± 0.42, −6.2 ± 0.19,−6.9 ± 0.46, and −6.5 ± 0.52 kcal/mol for the WT, T83I, D87N, and T83I/D87N systems, respectively), with only a slight favor for the WT. Our study, based on an X-ray crystal structure of P. aeruginosa GyrB/GyrA-GyrB/GyrA catalytic core with bound DNA and CPFX, suggests that GyrB residues such as Lys449 (Figure 3A) can also influence CPFX binding and thus should be included when evaluating quinolones binding. When GyrB was included in the analysis, the architecture of the quinolone binding cavity was redefined: the floor was assigned as residues 87–99 of GyrA and the roof as residues 447–449 of GyrB. Moreover, the residues 67–106 of GyrA —collectively known as the Quinolone Resistance-Determining Region (QRDR)— were confirmed as a critical mutational hotspot associated with resistance development [54].

In the present study, we have shown that the mutations T83I and D87N, located in the QRDR, do not appear to cause notable structural or dynamical perturbations within the P. aeruginosa gyrase complex. Global stability, as measured by Root Mean Squared Deviations (RMSDs), and local conformational flexibility, assessed through Root Mean Squared Fluctuations (RMSFs), remained largely unaffected by either drug binding or mutation (Figure 5, Figures S5 and S6). Cross-correlation analyses highlighted a pronounced association between GyrA and GyrB subunits, while inter-monomer correlations were markedly lower than the autocorrelations observed within each individual monomer. A notable exception was found in the exit gate residues of GyrA, which exhibited enhanced correlated motion across the two monomers, indicating a region of increased cooperative dynamics (Figure 6 and Figure 7).

MD simulations rendered insight into a solvent-filled cavity, preserved in all simulated systems, that may provide access to the inhibition site (Figure 8D), offering an opportunity to investigate the reverse pathway, namely CPFX entry into its inhibition site via this cavity site. By applying a sampling technique commonly used in MD (Umbrella Sampling) and guiding CPFX along a pre-defined path back towards the cavity site, we were able to generate data informative for studying the reverse process of CPFX binding.

In the bound state, residue Thr83 appears to be in proximity (<5 Å) with the drug, while Asp87 remains spatially distant (sampling window 0, Figure 10A). Energy profile analyses of the WT system (Figure 10A) suggest that Thr83 contributes favorably to the interaction energy in the initial stages of CPFX extraction from the binding site, with its greatest contribution when the drug has moved 2 to 4 Å from its bound state. This favorable interaction diminishes when Thr83 is mutated to isoleucine (Figure 10B). For Asp87, the interaction energy is the greatest when CPFX has moved from the inhibitory position to the cavity site. As CPFX progressed further into the cavity and approached Asp87, the interaction energy in the WT system increased.

While the equivalent residues to Thr83 and Asp87 in other gyrase/topoisomerases have been seen to form water-mediated magnesium-ion-bridges to fluoroquinolones (19, 24), our energy analysis suggests that the T83I and D87N mutations do not adversely affect the CPFX/Mg complex at the binding site (Figure 11). The MM/GBSA energy calculations using CPFX/Mg as a unified ligand indicate that this complex may be unable to access the cavity site in the mutant systems due to the high repulsive interaction energies. If CPFX were to approach the cavity site without its coordinated magnesium ion, it would encounter strong charge–charge repulsion from the negatively charged Asp283 and the anionic CPFX itself. The magnesium ion normally mitigates this by bridging the two negative charges during the transient passage. The cavity site contains two negatively charged residues (Asp282 and Asp287) and two positively charged residues (Arg91 and Arg273) among mostly hydrophobic residues, the latter of which can introduce mild repulsion, as observed even in energies of the WT system.

Additionally, replacing the polar Thr83 with a hydrophobic Ile83 may increase the local hydrophobicity, creating an environment that is even less accommodating for the charged or polar CPFX/Mg complex. Substitution of Asp87 with Asn removes the negative charge that previously contributed to a favorable interaction with the CPFX/Mg complex. Moreover, Asn87 may form a hydrogen bond with Thr83, orienting its partially positive NH₂ group toward the incoming CPFX/Mg and further destabilizing the interaction.

5. Conclusions

Our results suggest that while the GyrA mutant residues T83I and D87N do not render an increase in CPFX binding energy (decreased affinity) in the bound state, they may influence the actual access of CPFX to the gyrase inhibitory site by reducing the attractive forces associated with the process of binding. These findings underscore the complexity of molecular interactions, in which even subtle, non-destabilizing mutations may alter the trajectory of CPFX entrance to the active site. Understanding the importance of the path by which CPFX enters the gyrase active site could not have been elucidated solely by analysis of static (crystal or cryo-EM) structures, but became possible in combination with modeling of the interaction dynamics. We believe that combined structural analysis and MD simulations that study antibiotic access may be a useful consideration for screening compound libraries to identify new therapeutic agents to target fluoroquinolone resistant bacteria.