Novel bacterial topoisomerase inhibitors: unique targeting activities of amide enzyme-binding motifs for tricyclic analogs

ABSTRACT

Antimicrobial resistance has made a sizeable impact on public health and continues to threaten the effectiveness of antibacterial therapies. Novel bacterial topoisomerase inhibitors (NBTIs) are a promising class of antibacterial agents with a unique binding mode and distinct pharmacology that enables them to evade existing resistance mechanisms. The clinical development of NBTIs has been plagued by several issues, including cardiovascular safety. Herein, we report a sub-series of tricyclic NBTIs bearing an amide linkage that displays promising antibacterial activity, potent dual-target inhibition of DNA gyrase and topoisomerase IV (TopoIV), as well as improved cardiovascular safety and metabolic profiles. These amide NBTIs induced both single- and double-strand breaks in pBR322 DNA mediated by Staphylococcus aureus DNA gyrase, in contrast to prototypical NBTIs that cause only single-strand breaks. Unexpectedly, amides 1a and 1b targeted human topoisomerase IIα (TOP2α) causing both single- and double-strand breaks in pBR322 DNA, and induced DNA strand breaks in intact human leukemia K562 cells. In addition, anticancer drug-resistant K/VP.5 cells containing decreased levels of TOP2α were cross-resistant to amides 1a and 1b. Together, these results demonstrate broad spectrum antibacterial properties of selected tricyclic NBTIs, desirable safety profiles, an unusual ability to induce DNA double-stranded breaks, and activity against human TOP2α. Future work will be directed toward optimization and development of tricyclic NBTIs with potent and selective activity against bacteria. Finally, the current results may provide an additional avenue for development of selective anticancer agents.

INTRODUCTION

Bacterial infections result in extensive mortality. In 2019, there were an estimated 7.7 million deaths globally from the 33 most significant bacterial pathogens (1). Additionally, the rise in antibiotic resistance threatens the effectiveness of current treatment options, resulting in 1.3 million directly attributable deaths in 2019 (2). Within the United States, the Centers for Disease Control and Prevention has prioritized antibiotic-resistant bacteria based on their impact on public health. Examples include the serious threats from methicillin-resistant Staphylococcus aureus (MRSA) and multidrug-resistant (MDR) Pseudomonas aeruginosa and the urgent threats from carbapenem-resistant Enterobacterales and Acinetobacter baumannii (3). Therefore, the development of novel therapies to address these and other bacterial pathogens is warranted.

Novel bacterial topoisomerase inhibitors (NBTIs) (4) target the bacterial type II topoisomerases DNA gyrase and topoisomerase IV (TopoIV), enzymes responsible for modulating the topology of DNA during replication, transcription, and chromosomal dysjunction during binary fission. DNA gyrase uniquely introduces negative supercoils into relaxed DNA and also functions ahead of the replication fork to remove positive supercoiling, whereas TopoIV serves to decatenate DNA (5). These enzymes carry out their respective functions via transient double-strand DNA cleavage reactions followed by religation (5, 6). Previous research has demonstrated that NBTIs primarily stabilize single-strand DNA breaks (SSBs) generated by DNA gyrase and TopoIV (7 – 9). The unique binding mode (7, 8) and distinct pharmacological properties of NBTIs make them an advantageous choice for targeting MDR pathogens.

The fluoroquinolone (FQ) class of antibacterial agents also targets DNA gyrase and TopoIV. In contrast to most NBTIs, FQs stabilize double-strand DNA breaks (DSBs), preventing religation of DNA and resulting in bacterial cell death (9 – 12). However, resistance conferred by mutations in the target enzymes has eroded the effectiveness of FQs (9 – 12). NBTIs, with their distinctive structures and mode of action, largely evade cross-resistance with FQs (7, 8), making them a promising class of antibacterial agents for clinical development. Indeed, gepotidacin is currently in Phase III clinical trials for the treatment of uncomplicated urinary tract infections and urogenital gonorrhea. Gepotidacin has proven efficacious in patients against many pathogenic bacteria including Escherichia coli, MRSA, and Neisseria gonorrhoeae (13 – 15), laying a powerful foundation for the development of future NBTIs.

The advancement of additional NBTI clinical candidates has been hindered by several issues, especially cardiovascular toxicity driven by inhibition of the human ether-à-go-go-related gene (hERG) K⁺ ion channel (16). Gepotidacin, by contrast, is a weak hERG inhibitor (IC₅₀ = 588 µg/mL) with a low risk of proarrhythmia at clinically relevant dosing/exposures (17). However, gepotidacin and other NBTIs typically target DNA gyrase more potently than TopoIV in S. aureus and have additional liabilities associated with emergence of resistance via mutations to gyrase (14, 18 – 20). We have sought structurally distinctive NBTIs with potent anti-MRSA activity [minimal inhibitory concentrations (MICs) of ≤0.25 µg/mL], potent inhibition of both supercoiling by S. aureus DNA gyrase, and decatenation by TopoIV, in vitro cardiac safety (hERG IC₅₀ ≥ 100 µM), and selectivity for targeting bacterial vs human type II topoisomerases. Owing to the need for new therapies directed at life-threatening Gram-negative bacteria such as Enterobacteriaceae (including E. coli) and A. baumannii, we also routinely test these NBTIs against representative Gram-negative pathogens.

We previously reported a series of novel bicyclic 5-amino-1,3-dioxane-linked inhibitors with potent antibacterial activity against Gram-positive pathogens and improved cardiovascular safety profiles (20). This series, however, lacked potent Gram-negative activity and generally displayed rapid metabolism in microsomal assays. Optimization efforts of this series included increasing rigidity and polarity, resulting in analogs with tricyclic DNA-binding moieties (21), e.g., amines 2a and 2b (Fig. 1). Selected azatricyclic and diazatricyclic amine NBTIs displayed potent anti-MRSA activity, reduced hERG inhibition, and improved metabolic stability compared to the bicyclic inhibitors (20, 21).

Fig 1.

Design rationale for scaffold modification of tricyclic analogs.

Herein, we describe a subseries of tricyclic NBTIs bearing an amide linkage to the enzyme-binding moiety. This strategy further increases rigidity with the removal of a rotatable bond and replaces the weakly basic secondary amine with a non-basic amide. The potent antibacterial activity of these amides against both Gram-positive and Gram-negative pathogens prompted us to conduct biochemical and whole-cell experiments. These studies revealed an atypical mechanism of action vs DNA gyrase and unexpected targeting of human topoisomerase IIα (TOP2α).

RESULTS

Tricyclic amide NBTIs possess in vitro antibacterial activity

Antibacterial activity was measured as MICs against both Gram-positive and Gram-negative pathogens as previously described (20, 21). The azatricyclic amide-linked pyridooxazinone 1a was the most potent compound against a fluoroquinolone-resistant MRSA strain, with an MIC of 0.125 µg/mL (Table 1); slightly more potent than the clinical candidate gepotidacin. Amide 1a also exhibited promising Gram-negative activity against A. baumannii and E. coli, with an MIC of 2 µg/mL; the activity against A. baumannii was comparable to that of ciprofloxacin. We previously reported (21) that 2a, the amine counterpart to amide 1a, also had potent anti-MRSA activity, with an MIC of 0.25 µg/mL. In contrast, MICs for 2a were eightfold higher than 1a against the two Gram-negative bacteria (Table 1). The diazatricyclic amide-linked pyridooxazinone 1b was substantially weaker than 1a, with an eightfold loss in activity against MRSA, 16-fold loss against A. baumannii, and fourfold loss against E. coli (Table 1). These results suggest that the presence of the additional nitrogen in the diazatricyclic scaffold of 1b led to reduced activity. Amide 1b and its amine counterpart, 2b, displayed similar antibacterial activity, although 1b was fourfold more potent against E. coli (Table 1).

TABLE 1.

Minimal inhibitory concentrations (µg/mL) ^a of NBTIs

Compound	Structure	MRSA, USA 300	A. baumannii, ATCC 19606	E. coli, ATCC 25922
1a		0.125	2	2
1b		1	32	8
2a b		0.25	16	16
2b b		1	32	32
Gepotidacin b		0.25–1	32–64	1–2
Ciprofloxacin b		16	1–8	0.015–0.03

^a Determined in triplicate (at a minimum) according to Clinical & Laboratory Standards Institute (CLSI) guidelines (22). Observed ranges are reported where appropriate.

^b Values for amine analogs, gepotidacin, and ciprofloxacin were previously reported (21).

Tricyclic amide NBTI cardiovascular safety and metabolic profiles

Cardiovascular safety is a common concern for the development of NBTIs (16), and we have prioritized compounds with hERG IC₅₀ values ≥100 µM. We previously demonstrated that reducing the lipophilicity of dioxane-linked amine analogs (targeting cLogP < 1.5) generally resulted in low hERG inhibition (20). Notably, the pyridooxazinone enzyme-binding moiety common to 1a,b and 2a,b routinely afforded more potent hERG inhibition, even for highly polar compounds. Amides 1a and 1b replace the weakly basic amine, a strategy our laboratories (22) and others (23, 24) have previously successfully employed to reduce hERG inhibition. hERG inhibition was measured using an automated electrophysiology assay (IonWorks Barracuda system), as previously reported in reference (25) and shown in Table 2. Amides 1a and 1b both displayed favorable hERG IC₅₀ values, meeting our goal criteria of ≥100 µM. This finding is especially notable given the somewhat higher lipophilicity of 1a (cLogP = 1.82). Notably, both amides 1a and 1b were superior to their previously reported matched pairs, amines 2a and 2b, respectively, supporting the hERG design strategy.

TABLE 2.

In vitro hERG IC₅₀ and metabolic stability

Compound	hERG IC50 (µM) a	Half-life (min) b	cLogP
1a	>100	28	1.82
1b	>100	41	0.95
2a c	30	18.4	1.32
2b c	97	66.3	0.43

^a Charles River (Cleveland, OH, USA).

^b Charles River (Worcester, MA, USA).

^c Values for amine analogs were previously reported (21).

Metabolic clearance is also strongly correlated with lipophilicity (26), and dioxane-linked amine NBTIs generally required cLogP < 1.5 to achieve a half-life >30 min in mouse microsomes (20). The incorporation of the amide moiety blocks a potential site of metabolism (20), and azatricyclic amide 1a was less rapidly metabolized than amine 2a (Table 2). The diazatricyclic series showed the opposite trend; amide 1b had a shorter half-life than amine 2b. Thus, the incorporation of the amide linkage did not consistently improve the half-life of this subseries of compounds. The more polar diazatricyclic compounds 1b and 2b were more stable than their azatricyclic counterparts 1a and 2a, suggesting that reducing lipophilicity continues to be a productive strategy to enhance metabolic stability.

NBTIs inhibit supercoiling by DNA gyrase and decatenation by TopoIV

Isolated S. aureus DNA gyrase and TopoIV were used to assess inhibition of supercoiling and decatenation, respectively, by tricyclic NBTIs. IC₅₀ values are shown in Table 3. Amide 1a was the most potent inhibitor of supercoiling and decatenation, with IC₅₀ values of 150 nM and 653 nM, respectively. Notably, 1a was equipotent to gepotidacin against DNA gyrase and fourfold more potent against TopoIV. Amide 1a was also 10-fold more potent than previously reported amine 2a against DNA gyrase and twofold more potent against TopoIV. A similar trend was seen with the diazatricyclic compounds. Amide 1b was 18-fold more active than amine 2b against DNA gyrase and twofold improved for TopoIV. The azatricyclic amide 1a was superior to diazatricyclic 1b (Table 3), as we had previously reported for amines 2a and 2b (21). Taken together, these data support an important role for the amide moiety in disrupting enzyme function and suggest that lipophilicity may also contribute to activity.

TABLE 3.

Target inhibition by tricyclic NBTIs

Compound	S. aureus DNA gyrase IC50 (nM) a	S. aureus TopoIV IC50 (nM) b
1a	150	653
1b	229	755
2a c	1,600	1,300 (n = 5)
2b c	4,200	1,400
Gepotidacin (racemic) c	170	2,800
Ciprofloxacin c	28,000	8,600

^a DNA gyrase supercoiling inhibition assay; n = 2.

^b TopoIV decatenation inhibition assay; n = 2 unless otherwise noted in parentheses.

^c Values for amine analogs, gepotidacin, and ciprofloxacin were previously reported (21).

Tricyclic NBTI effects on DNA cleavage induced by S. aureus DNA gyrase

NBTIs primarily enhance gyrase-mediated SSBs, as has been well characterized with gepotidacin (8). In contrast, ciprofloxacin, a fluoroquinolone, primarily enhances gyrase-mediated DSBs (10, 11). We previously reported that bicyclic dioxane-linked amide NBTIs atypically stabilize both SSBs and DSBs (22). Therefore, we next evaluated the tricyclic amide 1a for its activity in cleavage assays using isolated S. aureus DNA gyrase. Incubation with 1a resulted in a concentration-dependent increase in both SSBs and DSBs at nanomolar concentrations (Fig. 2A), whereas the gepotidacin control at a concentration of 20 µM showed only SSBs. Interestingly, SSBs and DSBs induced by amide 1a were nearly equivalent (Fig. 2B and C). In contrast, amine 2a displayed the prototypical behavior for NBTIs, inducing a concentration-dependent increase in SSBs with no appreciable DSBs (Fig. 3). These results suggest that the amide motif plays a critical role in accumulation of DSBs, consistent with results from pyridooxazinone and pyridothiazinone amides in our bicyclic series (22).

Fig 2.

Compound 1a induction of DNA strand breaks in the presence of gyrase. (A) The ethidium-stained gel indicates positions of DNA after incubation with enzyme in the absence or presence of compound 1a (0.2–50 nM), a dioxane-linked azatricyclic NBTI containing an amide-linked enzyme-binding moiety. The various DNA forms are indicated as negatively supercoiled pBR322 DNA {[-(SC)]}, linearized DNA (Lin); nicked open-circular DNA (Nick). Ciprofloxacin and gepotidacin were included as controls at the indicated concentrations. (B) Quantitation of compound 1a induced DNA single-strand breaks and double-strand breaks was accomplished by first measuring fluorescence in each lane with corrections for differential fluorescent emission efficiency in (-) SC compared to linearized (Lin) and nicked (Nick) DNA bands (27) followed by the calculation of percent total fluorescence in the respective bands. Percent cleavage in enzyme controls was subtracted to yield final results. (C) Percent DNA cleavage induced by compound 1a (50 nM). Results shown are the mean ± standard deviation (SD) from six separate experiments performed on separate days.

Fig 3.

Compound 2a induction of DNA strand breaks in the presence of gyrase. (A) The ethidium-stained gel indicates positions of DNA after incubation with enzyme in the absence or presence of compound 2a (0.01–10 µM), a dioxane-linked azatricyclic NBTI containing an amine-linked enzyme-binding moiety. Ciprofloxacin and gepotidacin were included as controls at the indicated concentrations. (B) Quantitation of compound 2a induced DNA single-strand breaks and double-strand breaks was obtained by the same method as that explained in Fig. 2B. (C) Percent DNA cleavage induced by compound 2a (10 µM). Results shown are the mean ± SD from five separate experiments performed on separate days.

Growth inhibitory effects of tricyclic NBTIs in human leukemia cells

Given the ubiquity and essentiality of type II topoisomerases across kingdoms, achieving selective targeting and activity against bacterial compared to human type II topoisomerases (e.g., TOP2α) is an important consideration for antibacterial development. We have routinely assessed NBTI compounds for potential TOP2α targeting by measuring growth inhibition in a human chronic myelogenous leukemia cell line (K562) and in an acquired anticancer drug (etoposide)-resistant clonal subline (K/VP.5) (20 – 22, 25). TOP2α levels are reduced in K/VP.5 cells to 15% of that found in parental K562 cells (28). Thus, selective NBTIs would be expected to exhibit little or no difference in growth inhibitory activity in K562 compared to K/VP.5 cells as an indication that these compounds lack targeting against the human enzyme, as we have found most often (20 – 22, 25).

Forty-eight-hour growth inhibition experiments were conducted as previously reported (29). The anticancer agent etoposide yielded expected results with 26-fold resistance in K/VP.5 cells, coordinate with a reduction in the level of TOP2α (Fig. 4A; Table 4). Surprisingly, 1a inhibited the growth of K562 cells with an IC₅₀ value of 0.44 µM (Table 4). Importantly, 10-fold cross-resistance was observed in K/VP.5 cells (Fig. 4B; Table 4), yielding an IC₅₀ value of 4.57 µM (Table 4) and suggesting that the amide 1a targets TOP2α. In stark contrast, amine 2a exhibited much less potent growth inhibition in K562 with no difference in activity in K/VP.5 cells (Fig. 4C; Table 4), an indication that this amine-containing NBTI did not target human TOP2α. Diazatricyclic amide 1b and amine 2b exhibited a similar trend (Table 4), indicating a critical role for the amide moiety in targeting TOP2α.

Fig 4.

Growth inhibitory effects of etoposide and NBTIs in K562 and K/VP.5 cells. Log-phase cells were incubated for 48 h with various concentrations of (A) etoposide, (B) 1a, (C) 2a after which cells were counted on a model Z1 Coulter counter (Beckman Coulter, Danvers, MA, USA). The extent of growth beyond the starting concentration in drug treated vs dimethyl sulfoxide (DMSO) controls was expressed ultimately as percent inhibition. The concentration-response curves for each of the tested drugs are shown as scattergrams from multiple experiments. The 50% inhibitory concentrations (IC₅₀ values) are shown in Table 4, averaging results from multiple experiments performed on separate days.

TABLE 4.

Growth inhibitory activity of tricyclic NBTIs in K562 and K/VP.5 cells

Compound	K562, IC50 (µM) b	K/VP.5, IC50 (µM)	Fold resistance c
Etoposide	0.10 ± 0.02 d	2.56 ± 1.50	25.6
1a	0.44 ± 0.17 (6) d	4.57 ± 1.66 (6)	10.4
1b	2.26 ± 0.79 (3)	22.11 ± 11.02 (3)	9.8
2a a	100.66 ± 0.25	127.44 ± 25.02	1.26
2b a	>200	>200	NA

^a Values for amine analogs were previously reported (21).

^b IC₅₀ in a 48-h growth inhibition assay.

^c IC₅₀ of K/VP.5 divided by that of the parental K562 cell line.

^d Mean ± range or SD; number of independent experiments performed on different days; n = 2 unless otherwise noted in parentheses.

Tricyclic amide NBTI-induced DNA damage in human leukemia cells

The results from growth inhibition experiments prompted us to investigate the action of amide 1a in intact mammalian cells. We conducted single-cell gel electrophoresis (Comet assay) (30) experiments to assess DNA damage in K562 and K/VP.5 cells (29). K562 cells were incubated for 30 or 60 min with DMSO vehicle control and 10–50 µM 1a followed by quantitation of DNA damage (Fig. 5A) (29). Additionally, amine 2a (50 µM) was used as a comparator, given its low growth inhibitory activity in K562 cells and lack of cross-resistance in K/VP.5 cells (Fig. 4C; Table 4). A concentration-dependent increase in DNA damage was observed with 1a with no appreciable DNA damage seen with 2a compared to DMSO control (Fig. 5A). Compounds 1a and 2a were compared across multiple experiments after 30-min incubation at 50 µM validating the activity of 1a and the relative lack of activity of 2a (Fig. 5B). The direct comparison between 1a and 2a further highlights the distinct effects of the amide in causing the accumulation of DNA strand breaks. No further increase in 1a-induced DNA damage was observed after 60 min (Fig. 5B). DNA damage induced by 1a was dramatically attenuated in K/VP.5 compared to K562 cells consistent with the diminished cellular level of TOP2α in these anticancer drug-resistant cells (Fig. 5C). In addition, 2a induced little or no DNA damage in either cell line (Fig. 5C). Together, results are consistent with the amide 1a targeting human TOP2α.

Fig 5.

NBTI induction of DNA strand breaks in human leukemia K562 cells. (A) Effects of 1a and 2a on DNA damage in K562 cells. K562 cells were incubated with 1a (10, 25, and 50 µM), 2a (50 µM), or DMSO (control) for 30 min followed by alkaline (pH 13) Comet assays. Results shown are from all data points (cells/nuclei) within a representative experiment. (B) Effects of 1a and 2a (50 µM) on DNA damage (Comet assays) in K562 cells incubated for 30 min and 1a (50 µM) for 60 min followed by alkaline (pH 13) Comet assays. Results shown are the mean ± SD from 4–5 experiments run on separate days. (C) K562 and K/VP.5 cells were incubated for 30 min at 37°C with DMSO (control) and 1a (50 µM) or 2a (50 µM) followed by evaluation of DNA damage by Comet assays. For all experimental conditions, more than 100 cells were evaluated by OpenComet software (31). *P < 0.001 compared to respective DMSO controls.

Tricyclic NBTI effects on DNA cleavage induced by human TOP2α

To provide additional evidence for human TOP2α targeting, DNA cleavage experiments using isolated human TOP2α enzyme were conducted. Enzyme alone controls revealed conversion of negatively supercoiled pBR322 substrate to relaxed DNA forms (RLX) when ethidium bromide was included in the agarose gels for electrophoresis as well as some SSBs (Nick) (Fig. 6A) (27, 32). Incubation with the anticancer agent etoposide (100 µM), a positive control, resulted in an accumulation of SSBs (Nick) and DSBs (Lin) (Fig. 6A). Compound 1a resulted in a concentration-dependent increase in both SSBs and DSBs, though SSBs predominated (Fig. 6A). Results from replicate experiments using 50 µM 1a are shown in Fig. 6B. In contrast, amine 2a did not induce DNA cleavage compared to enzyme controls (Fig. 6C). These results further confirm targeting of TOP2α by amide 1a.

Fig 6.

Compounds 1a, 2a induction of DNA strand breaks in the presence of TOP2α. (A) The ethidium-stained gel shows positions of DNA after incubation with enzyme in the absence or presence of 1a (0.1–100 µM) or etoposide (100 µM). The forms of DNA are indicated as relaxed (RLX), negatively supercoiled [(-) SC], nicked open-circular (Nick), and linearized (Lin). Formation of SSBs (Nick) and DSBs (Lin) was observed by fluorescence band intensity during conversion of (-) SC to Nick and Lin reaction products. EcoRI digested pBR322 was run on one lane as a control for identification of linearized DNA and for quantitation of %DNA cleavage. (B) Percent DNA cleavage (SSB and DSB) induced by compound 1a (50 µM) was quantified relative to intensity of the EcoRI band after subtraction of enzyme alone SSB or DSB controls. Results shown are the mean ± SD from six separate experiments. (C) The ethidium-stained gel shows positions of DNA forms after incubation with enzyme in the absence or presence of 2a (0.1–100 µM) or etoposide (100 µM).

DISCUSSION

The development of NBTIs for the treatment of MDR bacterial infections is challenging. Optimization efforts require balancing multiple parameters including antibacterial activity, inhibition of DNA gyrase and TopoIV, cardiovascular safety, and metabolic stability. The previously described amine 2a demonstrated potent antibacterial activity against Gram-positive pathogens but suffered from poor Gram-negative activity, relatively potent hERG inhibition, and poor metabolic stability (21). Here, we hypothesized that replacing the amine of 2a with an amide in 1a would reduce hERG inhibition (21, 24) and improve metabolic stability by blocking the potentially oxidatively labile amino methylene with a carbonyl group (20). We postulated that increased rigidity through the removal of one rotatable bond could improve downstream pharmacokinetic properties such as oral bioavailability (33) and enhance accumulation in Gram-negative bacteria (34). Indeed, amide 1a demonstrated several notable advantages over amine 2a, including a greater than threefold increase in hERG IC₅₀ and modest improvement in microsomal metabolic stability (Table 2). We were particularly encouraged by the substantial improvement in Gram-negative antibacterial activity for 1a (Table 1). The MIC of 2 µg/mL against E. coli was eightfold better than for 2a and on par with NBTI 5463, an earlier NBTI optimized for Gram-negative activity (35). Similar improvements were observed for A. baumannii. However, not all of these findings were generalizable. Whereas amide 1b also demonstrated reduced hERG inhibition vs amine 2b, improvements in metabolic stability and Gram-negative activity were not observed. Whereas amides 1a,1b both targeted S. aureus DNA gyrase and TopoIV more potently than amines 2a,2b, the increased inhibition in these biochemical assays did not translate to marked improvements in S. aureus MICs. This observation may reflect reduced penetration into S. aureus cells by the amides, but we have not yet carried out such measurements.

Amide 1a also exhibited intriguing pharmacological properties demonstrating the formation of DNA DSBs with S. aureus DNA gyrase, in contrast to the exclusive SSBs traditionally observed with NBTIs (7 – 9). We have previously reported DSBs with other amide NBTIs bearing oxazinone and thiazinone amide moieties, whereas benzamide NBTIs yielded only SSBs (22). The basis of this differentiation is not currently understood, although structural biology efforts are underway in an attempt to elucidate mechanism(s). Such research has provided powerful insight into the mechanism of other classes of inhibitors, including the very recent cryoelectron microscopy studies with the albicidins and E. coli DNA gyrase (36). Whether other amide-type NBTIs also induce type II topoisomerase-mediated DSBs is an open question; DNA cleavage results for such compounds have not been reported (24, 35).

Additionally, the tricyclic amide 1a but not the amine 2a inhibited the growth of human leukemia K562 cells in the low micromolar range (Table 4). The observed cross-resistance (Fig. 4B) in an acquired anticancer drug-resistant clonal cell line (K/VP.5), containing reduced TOP2α, suggests that this tricyclic amide NBTI has the capability to target TOP2α. A limitation of the present work is that both amides 1a,1b and amines 2a,2b were assayed as racemic mixtures. It is not certain whether the TOP2α-targeting activity resides in one or both enantiomers. Amide NBTIs, in general, appear to lack potent targeting of TOP2α (22, 24); we have only rarely observed such cross-resistance in our previous research, for example, with compound 4e in reference (21).

Assessment of amide 1a-induced DNA damage in anticancer drug-resistant K/VP.5 cells compared to K562 cells (Fig. 5C) is another indication that 1a is capable of targeting TOP2α. No appreciable DNA damage was manifest with 2a in intact cells, in accord with a virtual lack of growth inhibition in intact cells (Fig. 5A through C). In addition, biochemical analysis of the effects of 1a and 2a using isolated TOP2α enzyme provided evidence that amide 1a stabilized SSBs and DSBs generated by TOP2α (Fig. 6A and B), whereas amine 2a had no apparent DNA damaging effects (<1% due to nicking) on the human enzyme (Fig. 6C).

In summary, azatricyclic amide 1a delivered several key advances relative to our earlier NBTIs, most notably its low hERG inhibition, favorable dual-target inhibition of DNA gyrase and TopoIV, and potent activity against Gram-negative pathogens. The unexpected targeting of TOP2α precludes its further development as an antibacterial candidate. This discovery may, however, pave the way for the identification of potential anticancer therapeutics based on this chemically attractive scaffold with good drug-like properties. There have been previous reports of TOP2-targeting anticancer agents derived from classical antibacterial scaffolds, for example, the quinolone voreloxin (37). Finally, future work will be directed toward optimization and development of tricyclic NBTIs with both potent and selective activity against bacteria.

MATERIALS AND METHODS

Synthesis

General amide formation method: 3-oxo-3,4-dihydro-2H-pyrido3,2-b][1,4]oxazine-6-carboxylic acid (1–1.5 equiv), hydroxybenzotriazole (HOBT, 1.1 equiv), the requisite primary amine (1 equiv), 4-dimethylaminopyridine (DMAP, 0.1 equiv), and N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide (EDC, 1.1 equiv) were dissolved in tetrahydrofuran (THF, 0.1 M), and triethylamine (4 equiv) was added. The resulting mixture was heated at reflux overnight, then cooled to room temperature, diluted with dichloromethane, washed with saturated NaHCO₃ solution, dried over Na₂SO₄, and concentrated in vacuo. The residue was purified by chromatography on silica gel with dichloromethane/methanol (0%–25% gradient) to give the desired compound. Purity was determined by high-performance liquid chromatography (HPLC, “Method 2”) as previously reported (21).

N-(2-((9-fluoro-4-oxo-1,2-dihydro-4H-pyrrolo[3,2,1-ij]quinolin-1-yl)methyl)- trans -1,3-dioxan-5-yl)-3-oxo-3,4-dihydro-2H-pyrido[3,2-b][1,4]oxazine-6-carboxamide (1a)

The title compound as a white solid (0.033 g, 0.069 mmol, 57%) was prepared following the general amide formation method with the previously described (21) primary amine (1-((5-amino-trans-1,3-dioxan-2-yl)methyl)-9-fluoro-1,2-dihydro-4H-pyrrolo[3,2,1-ij]quinolin-4-one): ¹H NMR (400 MHz, DMSO-d ₆) δ: 11.31 (s, 1H), 7.93–7.88 (m, 2H), 7.63–7.58 (m, 2H), 7.47 (d, J = 8.2 Hz, 1H), 7.02 (dd, J = 9.8, 8.8 Hz, 1H), 6.51 (d, J = 9.4 Hz, 1H), 4.77 (t, J = 4.7 Hz, 1H), 4.74 (s, 2H), 4.46 (dd, J = 12.6, 9.5 Hz, 1H), 4.21–4.00 (m, 5H), 3.69–3.58 (m, 2H), 2.24 (dt, J = 13.8, 4.2 Hz, 1H), 2.02 (ddd, J = 14.0, 10.3, 4.6 Hz, 1H). ¹³C NMR (100 MHz, DMSO-d ₆) δ: 165.31, 163.47, 159.70, 158.73 (d, J = 248.4 Hz), 144.21 (d, J = 12.1 Hz), 141.83, 141.04, 140.41, 137.13, 127.11 (d, J = 9.3 Hz), 123.47, 122.10 (d, J = 2.6 Hz), 118.20, 118.00 (d, J = 20.6 Hz), 113.87 (d, J = 2.0 Hz), 111.41 (d, J = 23.5 Hz), 99.52, 68.41, 68.38, 66.76, 53.69, 42.24, 37.80, 33.79 (d, J = 1.2 Hz). HRMS (ESI) m/z calc’d for C₂₄H₂₂FN₄O₆ [M + H]⁺: 481.1523; found: 481.1529. HPLC: ELSD rt: 14.871 min, purity: 100%.

N-(2-((3-fluoro-7-oxo-4,5-dihydro-7H-pyrrolo[3,2,1-de][1,5]naphthyridin-4-yl)methyl)-1,3- trans -dioxan-5-yl)-3-oxo-3,4-dihydro-2H-pyrido[3,2-b][1,4]oxazine-6-carboxamide (1b)

The title compound as a white solid (0.036 g, 0.075 mmol, 50%) was prepared following the general amide formation method with the previously described (21) primary amine (4-((5-amino-1,3-trans-dioxan-2-yl)methyl)-3-fluoro-4,5-dihydro-7H-pyrrolo[3,2,1-de][1,5]naphthyridin-7-one). ¹H NMR (400 MHz, DMSO-d ₆) δ: 11.30 (s, 1 h), 8.44 (d, J = 1.4 Hz, 1 h), 7.97 (d, J = 9.7 Hz, 1 h), 7.91 (d, J = 8.0 Hz, 1 h), 7.60 (d, J = 8.1 Hz, 1 h), 7.47 (d, J = 8.2 Hz, 1 h), 6.76 (d, J = 9.7 Hz, 1 h), 4.78 (t, J = 4.7 Hz, 1 h), 4.74 (s, 2 h), 4.49 (dd, J = 12.0, 9.0 Hz, 1 h), 4.22–4.07 (m, 5 h), 3.69–3.59 (m, 2 h), 2.24 (dt, J = 14.0, 4.1 Hz, 1 h), 2.12 (ddd, J = 13.8, 10.0, 4.8 Hz, 1 h). ¹³C NMR (100 MHz, DMSO-d ₆) δ: 165.34, 163.49, 159.30, 155.21 (d, J = 258.3 Hz), 141.86, 141.05, 140.50 (d, J = 8.9 Hz), 140.43, 137.94, 135.22 (d, J = 26.2 Hz), 131.42 (d, J = 2.6 Hz), 126.94 (d, J = 17.7 Hz), 125.89 (d, J = 2.5 Hz), 123.50, 118.23, 99.25, 68.40, 68.37, 66.77, 53.95, 42.22, 36.88, 34.59. HRMS (ESI) m/z calc’d for C₂₃H₂₁FN₅O₆ [M + H]⁺: 482.1476; found: 482.1482. HPLC: ELSD rt: 13.177 min, purity: 99.4%.

The preparation of amines 2a and 2b has been previously described (21).

Minimal inhibitory concentration assay

Minimal inhibitory concentrations were assayed in triplicate (at a minimum) by microbroth dilution according to CLSI guidelines (38).

Microsomal stability

The stability in mouse microsomes was determined at Charles River (Worcester, MA, USA) using the previously described methodology (20).

hERG IC₅₀

IC₅₀ values were measured at Charles River (Cleveland, OH, USA) using the IonWorks Barracuda system (Molecular Devices Corporation, Union City, CA, USA) as previously reported (25).

Staphylococcus aureus gyrase supercoiling assay

Assays were performed using an Inspiralis Ltd (Norwich, UK) kit with modifications. DNA gyrase was added to initiate reactions and was present at a level that was just able to supercoil all relaxed pBR322 under the experimental conditions utilized. DNA gyrase was incubated with 250 ng of relaxed pBR322 DNA, DMSO vehicle control (1.2%–2.7% final), and varying concentrations of drugs/compounds diluted in DMSO (1.2%–2.7% final) in 30 µL total reaction volume at 37°C for 30 min under the following conditions: 40 mM HEPES-KOH (pH 7.6), 10 mM magnesium acetate, 10 mM DTT, 2 mM ATP, 500 mM potassium glutamate and 0.05 mg/mL BSA. Each reaction was stopped with 30 µL STEB buffer [40% sucrose, wt/vol; 100 mM Tris-HCl (pH 8); 10 mM EDTA (pH 8); 0.5 mg/mL bromophenol blue] followed by the addition of 30 µL chloroform/iso-amyl alcohol (24:1) and loaded on a 1% agarose gel run at 70 V for 2 h. Gels were stained with ethidium bromide (1 µg/mL) followed by UV visualization.

Staphylococcus aureus TopoIV decatenation assay

Assays were performed using an Inspiralis Ltd (Norwich, UK) kit according to protocol with slight modifications. Decatenation of 100 ng of kDNA was assessed after 20 min incubation at 37°C in a final 30 µL reaction volume with DMSO vehicle control (1.7% final) and varying concentrations of drugs/compounds diluted in DMSO in a buffer containing 50 mM Tris-HCl (pH 7.5), 5 mM magnesium chloride, 5 mM dithiothreitol (DTT), 1.5 mM ATP, 350 mM potassium glutamate, 0.05 mg/mL bovine serum albumin (BSA). TopoIV was added to initiate reactions and was present at a level that was just able to decatenate 80%–100% of the kDNA under the experimental conditions utilized. Reactions were stopped with 30 µL STEB buffer [40% sucrose, wt/vol; 100 mM Tris-HCl (pH 8); 10 mM EDTA (pH 8); 0.5 mg/mL bromophenol blue] followed by the addition of 30 µL chloroform/iso-amyl alcohol (24:1). After mixing and centrifugation, aqueous fractions (20 µL) were loaded on a 1% agarose gel and run at 70 V for 2 h. Gels were stained with ethidium bromide (1 µg/mL) for UV visualization and quantitation of the percent decatenation in the presence or absence of various concentrations of test compounds for the assessment of 50% inhibitory concentrations.

Staphylococcus aureus DNA gyrase cleavage assay

Assays were performed using modifications of a kit from Inspiralis Ltd (Norwich, UK). Reactions in 30 µL total volume were incubated with varying concentrations of drug/compound diluted in DMSO (3% final DMSO concentrations) at 37°C for 30 min in assay buffer [40 mM HEPES-KOH (pH 7.6), 10 mM magnesium acetate, 10 mM DTT, 100 mM potassium glutamate and 0.05 mg/mL albumin] containing pBR322 (250 ng/condition) and DNA gyrase (15 nM). Complexes were trapped with the addition of 2% (wt/vol) SDS (3 µL) and proteinase K (10 mg/mL) (1.5 µL) and incubated at 45°C for an additional 30 min to digest enzyme and reveal single-strand (Nick) and double-strand (Lin) breaks. STEB buffer (30 µL) [40% (wt/vol) sucrose, 100 mM Tris-HCl (pH 8), and 10 mM EDTA, 0.5 mg/mL bromophenol blue] and chloroform:isoamyl alcohol (24:1) (30 µL) were added followed by vigorous mixing and centrifugation for 2 min. The aqueous phase (20 µL) was loaded onto a 1% agarose gel containing 0.7 µg/mL ethidium bromide. Gel electrophoresis was conducted (60 V, 2 h) in 1× Tris acetate EDTA (TAE) buffer containing 0.7 µg/mL ethidium bromide. DNA bands were visualized using ImageLab software (Bio-Rad Laboratories, Hercules, CA, USA). Total fluorescence in each lane was determined as previously reported in reference (22) using corrections for differential emission in supercoiled DNA compared to linearized and nicked DNA bands according to reference (27) followed by calculation of percent total fluorescence in these bands as a measure of DNA cleavage. Percent cleavage in enzyme alone controls was subtracted to yield final results plotted using Sigmaplot 14.5 (Systat Software, Inc., San Jose, CA, USA).

Human topoisomerase II DNA cleavage assay

The formation of nicked (Nick; SSBs), linear (Lin; DSBs), and relaxed (RLX) DNA from negatively supercoiled pBR322 (substrate) was detected by separating reaction products using ethidium bromide gel electrophoresis as previously described (27, 32). The cleavage assay reaction mixture (20 µL total) contained 300 ng of isolated TOP2α protein, 160 ng of pBR322 DNA (NEB, Ipswich, MA, USA), and 1 mM ATP in an assay buffer [10 mM (pH 7.5), 50 mM KCL, 50 mM NaCl, 0.1 mM EDTA, 5 mM MgCl₂, 2.5% glycerol] along with DMSO solvent and 0.1–100 µM compound 1a or 2a in DMSO. The anticancer agent etoposide was also evaluated at the indicated concentrations. Assay buffer (17 µL) and compound/DMSO (1 µL) were mixed at room temperature for 30 min followed by addition of TOP2α (2 µL) to initiate the reaction which proceeded at 37°C for 30 min. Reactions were quenched with 10% SDS (2 µL) and 5 min later with 10 mM EDTA-200 mM NaCl mixture (2 µL) followed by 2 µL proteinase K (10 mg/mL). Tubes were incubated at 55°C for 1 h followed by addition of pre-heated 10× DNA loading solution (3.5 µL) (glycerol, H₂O, and bromophenol blue) and run on a 1.3% agarose gel containing ethidium bromide (0.7 µg/mL) in 1× TAE buffer also containing ethidium bromide (0.7 µg/mL) at 20V for 18 h. EcoRI digested pBR322 (160 ng) was run on one lane as a control for identification of linearized DNA and or quantitation of %DNA cleavage.

Mammalian cell growth inhibition

Growth inhibition was conducted in a human myelogenous leukemia cell line (K562) and an etoposide-resistant clonal subline (K/VP.5) using methods previously reported (29). Varying concentrations of drugs were incubated (48 h) in a 24-well plate containing log-phase parental K562 and clonal drug-resistant K/VP.5 cells adjusted to 1–1.5 × 10⁶ cells. Cells were counted using a Z1 Coulter counter. Growth inhibition (IC₅₀ value) was determined based on comparison with DMSO control growth using non-linear regression plots.

DNA damage (Comet) assay

Alkaline single-cell gel electrophoresis was utilized to assess DNA damage induced by compounds 1a and 2a as previously reported (29). K562 cells were washed and resuspended (1 × 10⁶ cell/mL) in warm buffer (25 mM HEPES, 10 mM glucose, 1 mM MgCl₂, 5 mM KCl, 130 mM NaCl, 5 mM NaH₂PO₄, pH 7.4) followed by incubation with 10, 25 or 50 µM of 1a and 50 µM of 2a or DMSO (solvent control) for 30 or 60 min at 37°C. The treated cells were washed with ice-cold buffer (1× PBS) and resuspended to 0.28 × 10⁶ cell/mL, and then further diluted in low melt agarose. Following alkaline electrophoresis (of ~2,000 cells) and staining with fluorescent DNA intercalating dye (SYBR Gold), DNA damage was visualized and quantified by fluorescence microscopy. Migrating fragments (comet tails) from the nucleoids (comet heads) were visualized and used to quantify the Olive tail moment (30) using the ImageJ processing program with the open-source software tool OpenComet (31). Olive tail moments from more than 100 cells per sample condition were determined (29).