Insights into the Mechanism of Inhibition of Novel Bacterial Topoisomerase Inhibitors from Characterization of Resistant Mutants of Staphylococcus aureus

Abstract

The type II topoisomerases DNA gyrase and topoisomerase IV are clinically validated bacterial targets that catalyze the modulation of DNA topology that is vital to DNA replication, repair, and decatenation. Increasing resistance to fluoroquinolones, which trap the topoisomerase-DNA complex, has led to significant efforts in the discovery of novel inhibitors of these targets. AZ6142 is a member of the class of novel bacterial topoisomerase inhibitors (NBTIs) that utilizes a distinct mechanism to trap the protein-DNA complex. AZ6142 has very potent activity against Gram-positive organisms, including Staphylococcus aureus, Streptococcus pneumoniae, and Streptococcus pyogenes. In this study, we determined the frequencies of resistance to AZ6142 and other representative NBTI compounds in S. aureus and S. pneumoniae. The frequencies of selection of resistant mutants at 4× the MIC were 1.7 × 10⁻⁸ for S. aureus and <5.5 × 10⁻¹⁰ for S. pneumoniae. To improve our understanding of the NBTI mechanism of inhibition, the resistant S. aureus mutants were characterized and 20 unique substitutions in the topoisomerase subunits were identified. Many of these substitutions were located outside the NBTI binding pocket and impact the susceptibility of AZ6142, resulting in a 4- to 32-fold elevation in the MIC over the wild-type parent strain. Data on cross-resistance with other NBTIs and fluoroquinolones enabled the differentiation of scaffold-specific changes from compound-specific variations. Our results suggest that AZ6142 inhibits both type II topoisomerases in S. aureus but that DNA gyrase is the primary target. Further, the genotype of the resistant mutants suggests that domain conformations and DNA interactions may uniquely impact NBTIs compared to fluoroquinolones.

INTRODUCTION

A recent report from the Centers for Disease Control and Prevention estimates that antibiotic resistance could result in 10 million deaths worldwide each year by 2050 (1). Also striking was the projected economic impact of this issue, which suggests that antibiotic resistance would cost the world up to US$100 trillion per annum. The number of potential drugs effective against multidrug-resistant pathogens is limited, and the increase in usage of antibiotics that are the drugs of last resort results in further resistance that renders them ineffective (2–5). One of the most notable examples is methicillin-resistant Staphylococcus aureus (MRSA). First isolated in 1961, MRSA has overcome almost all of the introduced antibiotics, including vancomycin and fluoroquinolones (FQ), which have long been relied upon as the last-resort antibiotics (6–8). These resistance challenges with preexisting antibiotics necessitate the discovery and development of novel scaffolds with well-differentiated modes of action to avoid the impact of preexisting resistance in the clinic (9–11). However, the route of exploiting novel antibacterial targets has met with significant failure over the last decade (12). A better strategy may be to identify novel scaffolds that function by binding to novel pockets on clinically well-validated targets that are conserved among bacteria, are differentiated from those of humans, and have a distinct mode of inhibition.

Bacterial DNA type II topoisomerases are clinically validated antibacterial drug targets. These two topoisomerases, DNA gyrase and topoisomerase IV (Topo IV), which are heterotetrameric A₂B₂ complexes comprised of two GyrA/GyrB and ParC/ParE subunits for DNA gyrase and Topo IV, respectively, regulate DNA topology during replication (13). DNA gyrase introduces negative supercoils, while Topo IV is mainly responsible for the decatenation of the sister chromatids at the end of the replication process. The fluoroquinolone class of antibacterial drugs has been used for decades to treat a variety of bacterial infections. They inhibit topoisomerase function by trapping the enzyme in its cleaved complex during DNA nicking, thus stabilizing double-strand breaks in the DNA, which leads to rapid bacterial killing. Due to the increased and constant use of fluoroquinolones, resistance to these drugs is extremely high among clinical isolates (14). Amino acid changes in the quinolone resistance-determining region (QRDR) of GyrA and/or ParC have become markers of fluoroquinolone resistance (15–17). The most common substitutions are found in key residues that coordinate a magnesium ion that is required for efficient binding of the fluoroquinolone molecule (18). However, there are other substitutions, including some in the B subunits, that have been shown to impact the potency of certain fluoroquinolones (19).

Recent efforts in the discovery of novel antibacterial agents have resulted in compounds with potent antimicrobial activity against multidrug-resistant pathogens, including resistance to fluoroquinolones. One of the promising classes of inhibitors among them is the novel bacterial topoisomerase inhibitors (NBTIs), reported by many groups (20–25), of which GSK2140944 is in phase 2 of clinical development for respiratory tract and acute bacterial skin and skin structure infections (26). NBTIs are a new class of compounds that inhibit both bacterial gyrase and Topo IV with minimal cross-reactivity to human topoisomerase. The absence of the formation of a DNA cleavage complex suggested a novel mode of action for these inhibitors, in contrast to the fluoroquinolones, which stabilize a cleaved-complex intermediate (22). The co-crystal structures of two representative NBTI molecules, GSK299423 (Fig. 1A) and AM8191, to S. aureus gyrase (20) confirmed that the binding mode of NBTIs was distinct from that of fluoroquinolones. They bind to the dimer interface of the DNA-protein complex, away from the cleavage site, and intercalate with the uncleaved DNA, which is trapped in the active site in a “stretched” A form. The protein interface forming the NBTI pocket is comprised of two juxtaposed winged-helix domains (WHDs), whose relative rotational configuration has been shown to correlate with the DNA cleavage status, which in turn determines the position of the C gate in either an open or closed configuration (20). The commonly observed substitutions of D83 and M121, which result in resistance to NBTIs (20, 22), are located in this binding site and have become the signature resistance markers for the scaffold. It has been proposed that the NBTI binds to the DNA prior to double-stranded cleavage and inhibits the cleavage and the subsequent step of extensive conformational changes that are the result of the G-gate opening for strand passage (20). This mechanism differs from that of the fluoroquinolones, which interfere with the phosphotransfer chemistry and religation of the DNA backbone upon cleavage. The common substitutions that result in FQ resistance in S. aureus, S84L and E88K, are located on the same helix of the WHD that contain the NBTI binding residues. In fact, the NBTI resistance residue, D83, is adjacent to the fluoroquinolone resistance residue, S84, in the primary sequence. However, their locations on the helix of the WHD subdomain place them in two distinct pockets located on opposite sides of the helix, with the fluoroquinolone residue facing the DNA cleavage site while the NBTI residue faces the dimer interface, resulting in lack of cross-resistance between the two classes of inhibitors (20).

FIG 1.

Chemical structures of the NBTI compounds. The two AZ compounds, AZ0217 and AZ6142, are called compounds 7a and R,S-7c by Reck et al. (27).

To further understand this intriguing conformation-driven inhibition mechanism and estimate the resistance risks associated with allosteric inhibition, we challenged a select number of Gram-positive pathogens with increasing concentrations of a representative NBTI, AZ6142 (27). In addition to determining the resistance frequencies, we also characterized representative resistant colonies to understand the alterations that would result in resistance to NBTIs and to obtain additional insight into the locations that are important for this mechanism of inhibition. Several additional previously described (27, 28) NBTI compounds (Fig. 1), as well as representative fluoroquinolones, were also tested against these resistant strains to confirm scaffold-specific changes and to understand cross-resistance. Numerous novel substitutions were identified in this study that support the evolving understanding of the function of this DNA-protein machinery and highlight locations of changes outside the binding pocket that can impact the effectiveness of the inhibition mechanism.

MATERIALS AND METHODS

Bacterial strains and antimicrobial susceptibility testing.

The clinical isolates used in this study are part of the AstraZeneca culture collection; they have been collected from multiple global sites and represent isolates with a range of resistance to other antibacterial agents. S. aureus ARC516 (a methicillin-susceptible clinical isolate from the United Kingdom) and Streptococcus pneumoniae NCTC7466 were used for the resistance studies described. The MIC against each isolate was determined using the broth microdilution method following Clinical and Laboratory Standards Institute (CLSI) guidelines. All compounds were tested in accordance with CLSI recommendations. The reference compounds used were obtained from the U.S. Pharmacopeial Convention (Rockville, MD). The quality control isolates (S. aureus ATCC 29213, S. pneumoniae ATCC 49619, Enterococcus faecalis ATCC 29212, Escherichia coli ATCC 25922, and Pseudomonas aeruginosa ATCC 27853) were obtained from the American Type Culture Collection

Killing kinetics.

Bacterial cultures were grown in cation-adjusted Mueller-Hinton broth (MHB2) (Sigma-Aldrich, St. Louis, MO) medium for S. aureus or MHB2 with the addition of 2.5% lysed horse blood (Hema Resource Inc., Aurora, OR) for S. pneumoniae to an optical density at 600 nm (OD₆₀₀) of 0.3 to 0.6. This culture was diluted in fresh medium back to an OD₆₀₀ of 0.001, 20-ml aliquots were dispensed into 125-ml flasks, and a dilution series was prepared in MHB2 and plated onto blood agar plates (Remel, Lenexa, KS) for CFU determination. The compound to be tested was added at 0.5×, 1×, 2×, 4×, 8×, and 16× the predetermined MIC. CFU determinations were made immediately after sampling by preparing a dilution series, plating aliquots onto blood agar plates, and incubating the plates overnight at 37°C in ambient air for S. aureus or 5% CO₂ for S. pneumoniae.

Frequency of spontaneous resistance.

Microorganisms were harvested from fresh blood agar plates (Remel, Lenexa, KS) and suspended in MHB2 (Sigma-Aldrich, St. Louis, MO) to an OD₆₀₀ of approximately 3.0, which corresponds to a density of approximately 10⁹ to 10¹⁰ CFU/ml, significantly higher than the 10⁴ CFU/spot that is employed using the CLSI-validated agar dilution MIC method. A dilution series of this suspension was plated onto blood agar plates to determine the number of CFU. In addition, 100-μl volumes of this suspension were spread evenly in triplicate onto agar plates containing the test compound at 2-fold-increasing concentrations. The plates were made with Mueller-Hinton 2 agar (BD, Franklin Lakes, NJ) for S. aureus and contained 5% sheep red blood cells (RBC) (Hema Resource Inc., Aurora, OR) for S. pneumoniae. Resistant variants were enumerated after 2 days of incubation at 37°C in ambient air for S. aureus or 5% CO₂ for S. pneumoniae. The colonies growing on plates at multiples of the concentration that prevented the confluent growth were counted, and resistance rates were calculated.

Resistant variants were first passed on compound-containing plates and subsequently passed twice on compound-free blood agar plates, and the stability of the reduced susceptibility was verified using the broth microdilution method following guidelines of the Clinical and Laboratory Standards Institute.

Serial resistance.

Freshly grown S. aureus ARC516 was harvested from blood agar plates (Remel, Lenexa, KS) and suspended in 3 ml MHB2 (Sigma-Aldrich, St. Louis, MO), and a sample was frozen as passage zero. The suspension was diluted to an OD₆₀₀ of 0.001, representing approximately 5 × 10⁵ CFU/ml, in fresh medium. The medium and incubation conditions used throughout the experiment were MHB2 and 37°C in ambient air. A series of 2-fold-increasing concentrations of AZ6142 was made and added to 2 ml of the diluted culture that had been dispensed into the wells of 24-well culture plates (Costar, Corning, NY). The plates were incubated for 20 to 24 h, and the OD₆₀₀ was determined. Bacterial cells from the well containing the highest concentration of test compound that permitted ≥80% growth of the control well were frozen at −70°C and were also diluted back to an OD₆₀₀ of 0.001 in fresh medium. This diluted culture, representing 0.5× MIC, was then dispensed into a fresh 24-well culture plate containing the series of increasing concentrations of the test compound. This was repeated for 20 days. For the fixed-concentration passage experiment, S. aureus ARC516 was grown in AZ6142 at 0.5× MIC (0.03 μg/ml) overnight in MHB2 at 37°C in ambient air. Each day, the culture was sampled, diluted back 1:1,000 in fresh medium containing 0.03 μg/ml AZ6142, and reincubated.

The daily frozen samples from both experiments were then single-colony purified on blood agar plates in the absence of compound and tested for susceptibility using a standard broth microdilution assay. Organisms from passages with elevated MIC values, representing a decrease in susceptibility, were selected for further investigation by sequence analysis.

Genetic analysis.

Whole genomic DNA was prepared using a standard genomic DNA preparation kit (Promega, Madison, WI). Topoisomerase II subunit genes were amplified using a high-fidelity PCR mixture (Roche, Nutley, NJ). The PCR product was purified using a QIAquick PCR purification kit (Qiagen, Valencia, CA) and sequenced in an Applied Biosystems 3100 genetic analyzer (ABI, Foster City, CA). If changes were observed, a second, independent PCR product was amplified and sequenced to confirm that the variation was genuine and not the result of an incorporation error during PCR amplification.

Structural analysis.

Structural data from the public domain with NBTI, Protein Data Bank (PDB) ID 2XCS (20), and ciprofloxacin, PDB ID 2XCT, bound to the S. aureus gyrase cleavage complex were used for the analysis. The substitutions were mapped onto the protein using Pymol. The structural figures were generated using Pymol and Adobe Photoshop.

RESULTS AND DISCUSSION

In vitro susceptibility profiles of NBTI compounds.

The three representative NBTI compounds (Fig. 1) all displayed good activity against Gram-positive pathogens, although their activities against Gram-negative species were more variable (Table 1). The evolution of this chemical scaffold was from the C-linked series exemplified by the Morphochem compound MCHEM18 (28) (Fig. 1) to the N-linked aminopiperidine AZ0217 (Fig. 1) described previously (27). Further work determined that reducing the pK_a of the N-linked aminopiperidine inhibitors resulted in a lower human ether-à-go-go-related gene (hERG) 50% inhibitory concentration (IC₅₀) and an improved safety profile while maintaining excellent potency against key Gram-positive pathogens. The most promising compound, AZ6142 (Fig. 1C), was microbiologically profiled further in this study. AZ6142 demonstrated robust activity against S. aureus isolates, with MIC₉₀ values of 0.125 μg/ml and with no isolate exhibiting a MIC of >0.25 μg/ml. Further, the activity of AZ6142 was unaffected by preexisting resistance to methicillin or levofloxacin (Table 2). The activities of AZ6142 were equivalent in a panel of coagulase-negative staphylococci. The MIC₉₀ value of AZ6142 against S. pneumoniae was 0.5 μg/ml, although some of the levofloxacin-resistant isolates tested displayed reduced susceptibility to AZ6142 (Table 2). The activities against Streptococcus pyogenes and Enterococcus faecium were slightly reduced, with MIC₉₀ values of 0.5 and 1 μg/ml, respectively.

TABLE 1.

Antibacterial activities of NBTI compounds

Strain	MIC (μg/ml)
	AZ6142	AZ0217	MCHEM18	Levofloxacin
S. aureus (MSSAa) ARC516	0.03	0.06	0.06	0.125
S. aureus (MRSA) ARC517	0.125	0.125	≤0.06	>8
S. pneumoniae ATCC 10813	0.125	0.06	0.125	0.5
S. pyogenes ARC838	0.25	0.25	0.25	1
E. faecium ARC1182	0.5	1	0.5	>16
E. coli W3110	4	0.5	8	0.03
Haemophilus influenzae ATCC 51907	4	2	8	≤0.015
P. aeruginosa PAO1	>16	8	>16	0.25

^aMSSA, methicillin-susceptible S. aureus.

TABLE 2.

Antibacterial activities of AZ6142 and comparator compounds

Organism (no. of isolates)	Antimicrobial agent	MIC (μg/ml)
		Range	50%	90%
S. aureus (163)	AZ6142	0.016 to 0.25	0.06	0.125
	Levofloxacin	0.06 to >8	0.25	>8
	Linezolid	1 to 4	2	2
	Daptomycin	0.25 to 1	0.5	0.5
Methicillin-susceptible S. aureus (52)	AZ6142	0.03 to 0.25	0.06	0.125
	Levofloxacin	0.06 to 4	0.125	0.25
	Linezolid	1 to 4	2	4
	Daptomycin	0.25 to 1	0.5	0.5
Methicillin-resistant S. aureus (111)	AZ6142	0.016 to 0.125	0.06	0.06
	Levofloxacin	0.06 to >8	4	>8
	Linezolid	1 to 4	2	2
	Daptomycin	0.25 to 1	0.5	0.5
Quinolone-resistant S. aureus (61)	AZ6142	0.016 to 0.125	0.03	0.06
	Levofloxacin	4 to >8	8	>8
	Linezolid	1 to 4	2	2
	Daptomycin	0.25 to 1	0.5	1
CoNS (105)a	AZ6142	≤0.008 to 0.25	0.06	0.125
	Levofloxacin	0.125 to >16	0.25	16
	Linezolid	0.25 to 32	1	2
	Clindamycin	0.03 to >16	0.06	>16
	Tetracycline	0.06 to >16	0.5	>16
S. pneumoniae (119)	AZ6142	0.016 to 2	0.25	0.5
	Levofloxacin	0.5 to >16	1	16
	Linezolid	0.5 to 2	1	1
	Erythromycin	0.008 to >8	0.06	>8
	Chloramphenicol	1 to 32	4	16
Quinolone-resistant S. pneumoniae (17)	AZ6142	0.06 to 2	0.5	1
	Levofloxacin	8 to 32	16	32
	Linezolid	1 to 2	1	2
	Erythromycin	0.03 to >8	2	>8
	Chloramphenicol	1 to 16	4	16
S. pyogenes (97)	AZ6142	0.06 to 1	0.25	0.5
	Levofloxacin	0.125 to 2	0.5	1
	Linezolid	0.5 to 32	2	2
	Erythromycin	0.016 to >8	0.06	0.125
E. faecium (63)	AZ6142	0.125 to 16	0.5	2
	Levofloxacin	0.5 to >16	>16	>16
	Linezolid	0.5 to 64	2	16
	Vancomycin	0.125 to >64	64	>64

^aCoNS (coagulase-negative staphylococci) consist of 48 Staphylococcus epidermidis, 23 Staphylococcus haemolyticus, 14 Staphylococcus lugdunensis, 16 Staphylococcus saprophyticus, 2 Staphylococcus capitis, 1 Staphylococcus simulans, and 1 Staphylococcus sciuri isolates.

Time-kill and postantibiotic effect studies were also done with AZ6142 in representative S. aureus (n = 4), S. pneumoniae (n = 3), and S. pyogenes (n = 2) isolates. AZ6142 displayed bacterial killing most similar to that observed with levofloxacin in these isolates. The compound displayed rapid initial bacterial killing in these isolates up to 8 h, similar to that observed with levofloxacin, although partial rebound of growth was observed by 24 h in the S. aureus isolates. This rapid initial kill was also observed in the fluoroquinolone-resistant S. aureus and S. pneumoniae isolates tested (a 3-log₁₀-unit reduction in CFU/ml after 6 h at 4× MIC). Although AZ6142 failed to induce a significant postantibacterial effect in any of the S. aureus and S. pneumoniae isolates, a postantibiotic effect of 2.5 to 3 h at suprainhibitory concentrations (16× MIC) was observed in both S. pyogenes isolates.

Selection and characterization of resistant variants.

Resistance to NBTI compounds has previously implicated several residues responsible for decreased susceptibility, including some that were not specifically part of the binding pocket (20, 22). Given the close proximity of the binding pockets, resistance studies were done in fluoroquinolone-susceptible isolates so that any impact on fluoroquinolone susceptibility could be identified. The frequencies of spontaneous resistance to the three NBTI compounds, AZ6142, AZ0217, and MCHEM18, were comparable and similar to that observed with fluoroquinolones in our experiment (Table 3), as well as that reported in the literature (29), although variations in S. aureus strains exist (30). The frequencies of spontaneous resistance in S. aureus ARC516 ranged between 5.5 × 10⁻⁸ and 4.8 × 10⁻⁹ at 4× the MIC, and resistant variants could be isolated for all of the compounds. The selection of spontaneous resistant variants in S. pneumoniae was more difficult, with frequencies below the limit of detection at 4× MIC for all compounds (Table 3). Multiple first-step mutants for the three compounds in S. aureus ARC516 were selected for follow-up characterization, and 14 unique representatives are listed in Table 4. All of the mutants contained nonsynonymous substitutions in the DNA gyrase subunits; 12 of them were in GyrA, and two mutants carried a GyrB substitution. The D83G and M121K substitutions, which have been seen previously with other NBTIs (22, 24, 31), were observed with all three compounds in this study, with the exception of D83G mutants with AZ6142, despite several independent experiments and characterization of multiple variants. The mutants were tested for cross-resistance against all three NBTI compounds, as well as two fluoroquinolones, ciprofloxacin and levofloxacin (Table 4). Whereas most mutants showed similar decreases in susceptibility across the NBTI compounds, the D83G mutant was the most variable. The elevation in the MICs of AZ6142 and AZ0217 against this mutant was quite modest (2- to 4-fold), yet there was a significant impact (128-fold) on MCHEM18, a level similar to that observed with other NBTI molecules in the literature (20, 22, 31). The modest 2-fold increase in the MIC of AZ6142 likely explains the inability to obtain this mutant despite multiple selection experiments with AZ6142.

TABLE 3.

Frequencies of spontaneous resistance of NBTI compounds

Compound	Methicillin-susceptible S. aureusa			S. pneumoniaeb
	MIC (μg/ml)	FORc		MIC (μg/ml)	FOR
		4×MIC	8×MIC		4×MIC	8×MIC
AZ6142	0.03	1.7 × 10−8	3.3 × 10−9	0.125	<5.5 × 10−9	<5.5 × 10−9
AZ0217	0.06	5.5 × 10−8	<1.2 × 10−9	0.06	<1.9 × 10−8	<1.9 × 10−8
MCHEM18	0.06	4.8 × 10−9	8.0 × 10−9	0.125	<6.3 × 10−9	<6.3 × 10−9
Levofloxacin	0.125	2 × 10−10	<2 × 10−10	0.5	<1.5 × 10−9	<1.5 × 10−9

^aStrain ARC516.

^bStrain NCTC7466.

^cFOR, frequency of resistance.

TABLE 4.

In vitro susceptibilities of S. aureus ARC516-resistant variants obtained after NBTI selection

Substitution(s)	Selectionj	MIC (μg/ml)k
		CIP	LVX	AZ6142	AZ0217	MCHEM18
S. aureus ARC516 (parent strain)	NA	0.125	0.125	0.03	0.06	0.06
GyrA P36Sa	FOR	0.06	0.06	0.5	0.25	0.5
GyrA P36Tc	FOR	0.125	0.125	0.5	0.25	0.5
GyrA P44Tc	FOR/SP-F	0.125	0.06	0.25	0.25	0.25
GyrA V45Ib	FOR	0.06	0.06	1	0.5	1
GyrA V45Ac	FOR	0.125	0.125	0.5	0.25	0.5
GyrA D83Gf	FOR	0.125	0.125	0.06	0.25	8
GyrA A89Gc	FOR	0.125	0.125	0.5	0.25	0.25
GyrA M90Va	FOR	0.125	0.125	0.5	0.5	0.5
GyrA R92Ca	FOR	0.06	0.125	0.5	0.25	0.5
GyrA Q95Sc	FOR	0.06	0.06	0.25	0.125	0.125
GyrA M121Kd	FOR	0.125	0.125	4	4	8
GyrA G171Dc	FOR	0.125	0.06	0.25	0.25	0.25
GyrB P456Le	FOR	0.06	0.06	0.5	0.25	0.5
GyrB K417Ec	FOR	0.06	≤0.03	0.25	0.125	0.125
GyrA M121K, ParC A30Tc,g	FOR-2	0.25	0.125	32	32	32
GyrA M121K, ParC M117Kc,g	FOR-2	0.25	0.25	>32	>64	64
GyrA A32Vc	SP-F	0.125	0.125	0.25	0.125	0.25
GyrA S98Ic	SP-F	0.125	0.125	0.25	0.125	0.25
GyrA F164Sc	SP-F	0.125	0.125	0.125	≤0.06	≤0.06
GyrB D437Nc	SP	0.06	0.125	1	0.5	1
GyrB D437N, ParC K94Rc,h	SP	0.25	0.25	2	1	2
GyrB D437N, ParC K94R, GyrA R12Lc,i	SP	0.25	0.25	16	4	8

^aMutant obtained after selection with MCHEM18.

^bMutant obtained after selection with AZ0217.

^cMutant obtained after selection with AZ6142.

^dMutant obtained after selection with MCHEM18, AZ0217, and AZ6142.

^eMutant obtained after selection with MCHEM18 and AZ6142.

^fMutant obtained after selection with MCHEM18 and AZ0217.

^gDaughter strain of parent with GyrA M121K substitution.

^hDaughter strain of parent with GyrB D437N substitution.

ⁱDaughter strain of parent with GyrB D437N and ParC K94R substitutions.

^jFOR, frequency of resistance study; FOR-2, second-step frequency of resistance study; SP-F, serial passage with fixed concentration; SP, serial passage with increasing concentration; NA, not applicable.

^kCIP, ciprofloxacin; LVX, levofloxacin.

The most significant and consistent MIC increases were observed with the GyrA M121K substitution, ranging from 64- to 128-fold for all of the NBTI compounds. This change, like D83G, is also located in the binding pocket and has been shown to affect all other NBTI molecules studied thus far in the literature (20, 22, 31), reflecting its scaffold-specific attributes. Additionally, there was no cross-resistance observed with either ciprofloxacin or levofloxacin, as previously reported (Table 4). The resistant isolate containing the M121K substitution obtained with AZ6142 exposure was used to determine whether additional substitutions could be accumulated that further decrease the susceptibility to AZ6142. The frequency of the spontaneous second-step resistance mutant was <4.1 × 10⁻¹⁰ at ≥4× MIC. However, four colonies from the 2× MIC plate (frequency, 4.1 × 10⁻⁹) were characterized, and all exhibited AZ6142 MIC values of ≥64 μg/ml and were fully cross-resistant to the other NBTI compounds (Table 4). Genetic characterization identified either A30T or M117K substitutions in the ParC protein in these mutants, which supports the dual-targeting nature of the molecule.

We also evaluated the types of substitutions selected during serial passage (SP) (Table 4) using either fixed or increasing concentrations of AZ6142. After 20 daily passages using increasing concentrations, there were 3 stepwise decreases in AZ6142 susceptibility in S. aureus ARC516. The first-step substitution, which was detected in passage 8, was D437N in GyrB, accompanied by a 32-fold decrease in susceptibility. The second change, which gave a further 2-fold loss in susceptibility, was a K94R change in ParC in passage 12, again supporting the dual-target mode of inhibition. The final 8-fold loss in susceptibility, representing a 512-fold decrease over that of the initial parent isolate, was accompanied by an R12L substitution in GyrA and appeared after passage 19. These changes gave equivalent stepwise decreases in susceptibility to AZ0217 and MCHEM18 (Table 4). A culture of S. aureus ARC516 was also passaged 20 times with a fixed concentration (SP-F) (Table 4) of AZ6142 at 0.5× MIC (0.03 μg/ml). As there was no increasing selective pressure in the experiment, it was unknown whether one mutant population would dominate. Purification and testing of representatives from the different passages confirmed that multiple lineages of resistant variants indeed coexisted over the 20 daily passages. Four unique variants, all with a single substitution in GyrA, were identified that gave between 2- and 8-fold elevations in the AZ6142 MIC value. Three of the four mutants also displayed cross-resistance to AZ0217 or MCHEM18 (Table 4). Taken together, these data suggest that different methods of selective pressure may result in different mechanism-specific resistance substitutions.

Localization of NBTI substitutions.

Crystal structures of S. aureus DNA gyrase solved in complex with different NBTI compounds (PDB ID 2XCS [20] and 4PLB [25]), and modeling with a related compound (32), confirmed that the binding mode was conserved among all compounds with this scaffold. The crystal structure of the S. aureus enzyme (PDB ID 2XCS) has been extensively analyzed, and therefore, all the substitutions described in this study were mapped to this structure (Fig. 2). Interestingly, the mapping showed that nearly all residues outside the compound binding site were spread over the DNA-binding pocket (Fig. 2A). This pocket represents the G-gate DNA-substrate binding cleft that functions to bind, bend, and cleave the DNA and subsequently allows T-strand passage and DNA religation (33) and, as such, performs a vital and complex function in the topoisomerization reaction cycle (34–36). The substitutions observed in this study are spread as far as 38 Å (to Arg12) and 30 Å (to Phe164) in different directions from the NBTI binding pocket. No substitutions were observed in other parts of the protein that are distant from the G-gate DNA-binding pocket (Fig. 2B and 3A), including the N terminus of GyrB or the C terminus of GyrA, that are not a part of this crystal system.

FIG 2.

Substitutions mapped to the NBTI-bound cleaved complex structure of S. aureus gyrase (PDB ID 2XCS). (A) Top view, showing the DNA cleft. (B) Side view, showing the G and C gates. The two monomers of the gyrase surface are colored in shades of yellow and blue. GyrA is depicted in the darker shades, with GyrB in the lighter shades. DNA bound to the NBTI GSK299423 (green sticks) is shown as a gray coil, while for comparison, the DNA bound to ciprofloxacin (from PDB ID 2XCT) is shown as a red coil. The substitutions observed in this study are colored magenta.

FIG 3.

Localization of substitutions. (A) Substitutions mapped to the overall secondary structure of S. aureus gyrase (PDB ID 2XCS). The two GyrB TOPRIM domains are colored light yellow and light cyan, and the WHDs from each monomer are colored light and dark gray, while the remainder of the GyrA subunits in the crystal structure are colored orange and blue. The N-terminal helix and β-strand of the monomer visible in the structure are colored green. The DNA is highlighted by green flat-base stacks, while the NBTI molecule is a green stick in the dimer interface. The residues substituted in this experiment are highlighted in magenta. (B) Group 1 and group 3 changes on the WHD interface. The substituted residues are depicted in a stick diagram; the group 1 changes are colored in magenta and the group 2 changes in orange, while the WHD of each dimer is colored in two shades of gray. Here and in panels C and D, the NBTI is depicted as yellow sticks and the uncleaved DNA bound to the NBTI is depicted in blue, whereas the cleaved DNA observed in ciprofloxacin (for comparison) is light gray. (C) Group 2 and TOPRIM domain. The GyrB from each monomer is colored light yellow or cyan. The mutated residues are depicted as orange sticks, while the ciprofloxacin molecules in each dimer are depicted as red sticks to highlight the location of DNA cleavage. The position of the catalytic tyrosine is marked by a red-stick view of F123, as there is a Y123F change in the crystal construct. The residues that are commonly mutated in fluoroquinolone-resistant strains, S84 and E88 (see the introduction), are also indicated in green. (D) Group 4 and group 5 changes and the GyrA-GyrB interface. The group 4 and group 5 residues are depicted as brown and dark-pink sticks, respectively. The TOPRIM domain is colored yellow, while the N-terminal β-strand and α-helix of GyrA are colored green. The WHD is colored gray. The locations of group 1 and 3 residues on the WHD are shown in magenta and orange.

Based on their locations, the residues with substitutions that confer reduced susceptibility to NBTI molecules can be divided into five groups (Table 5). The substitutions in group 1, M121K and D83G (Fig. 3B), are located in the compound binding pocket in GyrA and make direct contact with the inhibitor. The M121 residue provides a hydrophobic stacking interaction, whereas D83 participates in hydrogen bonding with the linker nitrogen. The group 2 substitutions, K417E, D437N, and P456L, are located in the GyrB TOPRIM domain and interact with the DNA backbone of the nucleotides that form the NBTI binding pocket (Fig. 3B). Of these, D437 directly associates with the nucleotide backbone, and subtle changes with a D437N substitution have the potential to move the plane of the nucleotide bases, which in turn can significantly impact the conformation of the NBTI pocket. It should be noted that D437 is also close to the cleavage region (highlighted by ciprofloxacin and F123 in Fig. 3C), and a D437N change has also been suggested to have a minor effect on the susceptibility of some fluoroquinolone compounds (37), as well as of AZD0914, a novel cleavage complex-mediated inhibitor (38). The other two residues in group 2, K417 and P456, are in close proximity to D437 and are likely to impact the optimal positioning of D437 needed for NBTI binding by charge or steric clashes, respectively.

TABLE 5.

Locations and impacts of mutated residues

Group	Role/location	Subunit	Substitution	Fold change of MICa
				NBTI	Fluoroquinolone
1	Binding pocket; direct contact with NBTI	GyrA	D83G	2–128	1
		GyrA	M121K	64–128	1
		ParC	M117K	8 to >16	2
2	Binding pocket; contact with DNA	GyrB	D437N	8–32	0.5–1
		GyrB	K417E	2–8	≥0.25–0.5
		GyrB	P456L	4–16	0.5
3	WHD changes involved in DNA bending	GyrA	V45A/I	4–32	0.5–1
		GyrA	P44T	4–8	0.5–1
		ParC	K94Rb	2	2
		GyrA	A89G	4–16	1
		GyrA	M90V	8–16	1
		GyrA	F164Sc	0–4	1
		GyrA	S98I	2–8	1
		GyrA	R92C	4–16	0.5–1
		GyrA	Q95S	2–8	0.5
		GyrA	G171Dc	4–8	0.5–1
4	N-terminal HLH changes involved in DNA bending	GyrA	P36S/T	4–16	0.5–1
		ParC	A30T	4–8	1–2
		GyrA	A32V	2–8	1
5	Interaction with GyrB	GyrA	R12L	4–8	1

^aFold MIC change over the parent (or the previous step in cases of multiple-step mutants).

^bThe ParC K94 residue is equivalent to the GyrA S98 residue in a multiple-sequence alignment.

^cThe residue was put in group 3, as it is close to the WHD, although not in the subdomain.

The group 3 and 4 substitutions are also involved in interactions with the DNA strand, like the group 2 substitutions. However, they are located further away from the NBTI binding pocket and are in the region of the DNA-binding cleft, where the DNA is bent sharply by ∼150° (Fig. 3B and D). These two groups of substitutions can be differentiated by the substructural domains in which they are located. The group 3 substitutions, which represent the largest number of substitutions observed in this study, are located in the WHD of GyrA, which also carries the group 1 substitutions (Fig. 3B), as well as most common fluoroquinolone resistance determinants. Of the group 3 changes, R92C, V45A/I, M90V, and A89G directly interact with the DNA, whereas others form the second shell. The group 4 residues are located on the α1 helix immediately upstream of the WHD and on the loop that connects these two substructures (Fig. 3D). All group 4 substitutions, P36S/T, A30T, and A32V, interact with the DNA backbone. Both the group 3 and group 4 substitutions result in comparable shifts in the MIC for a given compound (Table 5), suggesting that the impacts of altering the protein-DNA interaction in these regions are similar.

Finally, the single group 5 substitution, R12L, is located on the N-terminal β-strand of GyrA that forms a β-sheet, together with the β-strands from the adjacent GyrB TOPRIM domain, and represents a part of the GyrA-GyrB interface (Fig. 3D). The side chain of R12 makes a strong salt bridge to E20 located on the α1 helix, which also carries the group 4 changes. An R12L substitution is expected to disrupt this tight interaction and likely alter critical protein-protein connectivity between the GyrB TOPRIM domain and the GyrA DNA-binding elements.

Mechanistic insight from NBTI mutants.

The biochemical data and the resistance substitutions with NBTI compounds in previous studies (20, 22), together with the co-crystal structure, have helped to differentiate both the binding mode and the mechanism of inhibition of this class of inhibitors from those of the fluoroquinolones (16, 39, 40). While the general mechanistic differences have been well established, the specific pathways of conformational changes that result in this allosteric inhibition remain ill defined. Recently, structural advances in the field of bacterial (13, 20, 34, 39, 40) topoisomerases have significantly improved the understanding of this complex machinery, where association and dissociation of various subunits result in DNA cleavage/religation and DNA strand passage. However, with a limited number of inhibitors having unique inhibition mechanisms, resistance-based validation of these hypotheses is sparse.

In this study, we report a large number of unique resistant mutants raised against NBTI compounds that provide further insight into this mechanism and corroborate many of the conformational changes suggested to be important for topoisomerase function (34–36, 40). Excluding the two substitutions in the binding pocket, most of the variations are located in the DNA-binding cleft near the DNA cleavage region and participate, either directly or indirectly, in interactions with the DNA backbone. Further, the majority of these changes are located in the WHD (group 3) or are close to the WHD (groups 4 and 5). The WHD is a critical functional subdomain of type II topoisomerases, comprised of ∼100 amino acids that form the hydrophobic protein-protein dimer interface of GyrA/ParC, as well as a large section of the cleavable gate segment DNA (G-DNA) binding pocket. Indeed, this dimer interface is where the NBTI binding pocket is located. The WHD also carries the catalytic active-site tyrosine, as well as the isoleucine (I175) that intercalates with the minor groove of the DNA to promote the sharp 150° bending of the duplex DNA (41). Further, among the residues that coordinate the catalytic metal ion, GyrB D512 of the TOPRIM domain has been seen to alternate between its interaction with the metal ion and GyrA R33 of the α1 helix during the catalytic cycle, which is thought to act as a conformational switch for the remainder of the protein upon DNA cleavage (20). The close interaction between the WHD and the α1 helix shows the importance of these domains in communication, during the cleavage-religation reaction of the DNA, with the rest of the protein for gate opening and closing (39, 40). Thus, the WHD and the secondary structures associated with it are the fulcrum points of various conformational and functional links.

The location of the NBTI binding pocket between the two WHDs, which is formed during the catalytic reaction cycle, already points to the importance of the relative movements of these subdomains for DNA cleavage and strand separation. Binding of the inhibitor in this transitory pocket traps the DNA-protein complex in a precleavage state where the DNA is stretched between the two active sites but remains uncleaved. The protein is unable to undergo further conformational change to complete the cleavage reaction and strand passage, resulting in allosteric inhibition. The fact that the majority of the substitutions were located in the WHD subdomain or the α1 helix strongly suggests that the binding of these mechanism-based inhibitors can be disrupted in multiple ways. The impact of the substitutions in the binding pocket is straightforward, but the protein can also prevent complete inhibition by altering conformational dynamics that are needed to trap the inhibitor. The impacts of these changes on the natural topoisomerase function remain to be studied.

It was interesting to observe that none of the DNA-binding changes that affected the activity of the NBTI compounds resulted in significant changes in susceptibility to other DNA-protein topoisomerase inhibitors, such as fluoroquinolones (Tables 4 and 5) or AZD0914 (data not shown). The multiple structures of topoisomerase proteins bound to fluoroquinolones (16, 20, 40, 42) reveal minimal differences in the interaction of these residues with the DNA bound to fluoroquinolones versus the NBTI complex, although the conformation of the DNA strand is slightly different (Fig. 2A). This lack of cross-reactivity could be due to differences in conformational pathways, resulting in inhibition. Another reason could be the difference in targets in S. aureus, as fluoroquinolone substitutions have suggested that Topo IV is the primary target in this species. Further biochemical and biophysical studies on purified proteins are needed to better understand the exact impacts of these changes on the inhibition of both compound classes, as well as on the function of the protein.

Conclusion.

In this work, we have generated and characterized multiple mutants that are resistant to AZ6142 and other NBTI compounds in the series. Multiple unique substitutions were observed in both DNA gyrase and Topo IV subunits, confirming the dual-targeting mechanism, as well as the conformation-dependent mechanistic properties of these compounds. The novel binding mode of this series with no preexisting resistance, together with a frequency of spontaneous resistance comparable to that of fluoroquinolones, suggests that the development of resistance against AZ6142 in the clinic may be relatively slow but would need to be closely monitored. The locations of the resistance substitutions in this study also confirmed the dual-targeting mechanism of AZ6142. Furthermore, the lack of cross-resistance to fluoroquinolones further confirms that, although the molecular targets are similar, the subtle differences in the binding pocket and binding modes and in the primary target would provide an advantage in the treatment of infections caused by fluoroquinolone-resistant S. aureus.