Nybomycin Inhibits both Fluoroquinolone-Sensitive and Fluoroquinolone-Resistant Escherichia coli DNA Gyrase

Bacterial type II topoisomerases, DNA gyrase and topoisomerase IV, are targets of many antibiotics, including fluoroquinolones (FQs). Unfortunately, a number of bacterial species easily acquire resistance to FQs by mutations in either DNA gyrase or topoisomerase IV genes.

ABSTRACT

Bacterial type II topoisomerases, DNA gyrase and topoisomerase IV, are targets of many antibiotics, including fluoroquinolones (FQs). Unfortunately, a number of bacterial species easily acquire resistance to FQs by mutations in either DNA gyrase or topoisomerase IV genes. The emergence of resistant pathogenic strains is a global problem in health care; therefore, identifying alternative pathways to thwart their persistence is the current frontier in drug discovery. Nybomycins are an attractive class of compounds, reported to be “reverse antibiotics” that selectively inhibit growth of some Gram-positive FQ-resistant bacteria by targeting the mutant form of DNA gyrase while being inactive against wild-type strains with FQ-sensitive gyrases. The strong “reverse” effect was demonstrated only for a few Gram-positive organisms resistant to FQs due to the S83L/I mutation in the GyrA subunit of DNA gyrase. However, the activity of nybomycins has not been extensively explored among Gram-negative species. Here, we observed that in a ΔtolC strain of the Gram-negative Escherichia coli with enhanced permeability, wild-type gyrase and a GyrA S83L mutant, resistant to fluoroquinolones, are similarly sensitive to nybomycin.

INTRODUCTION

Bacterial cells employ an array of specialized enzymes, topoisomerases, in a bid to maintain their genomes under proper topological conditions during replication. Topoisomerases can be classified into two distinct groups, types I and II, which differ in the types of DNA breaks they introduce to operate with DNA topology (single-stranded and double-stranded breaks, respectively). Bacterial type II topoisomerases, represented by DNA gyrase and topoisomerase IV (topo IV), perform their function by binding DNA, producing a double-stranded break with a covalent intermediate between hydroxyls of catalytic tyrosines and 5′-phosphates of cleaved DNA, causing topological changes via passing the second DNA duplex through the break, religating cleaved DNA, and dissociating from DNA (1). DNA gyrase, or simply gyrase, and topo IV are heterotetramers that exist as GyrA₂GyrB₂ (gyrase) and ParC₂ParE₂ (topo IV) complexes. DNA gyrase is able to introduce negative supercoils into DNA, relax supercoiled DNA, and resolve catenated DNA circles, while topo IV relaxes DNA supercoils and possesses decatenation activity (2). The major functions of DNA gyrase in vivo are DNA supercoiling and relaxation of positive supercoils, whereas topo IV mediates the segregation of daughter bacterial chromosomes prior to cell division (3).

Since these enzymes play an important role during key cellular events, they are an attractive target for a wide range of antimicrobial agents. Fluoroquinolones (FQs) are a well-known family of antibiotics that inhibit topoisomerase activity (4). FQs target type II topoisomerases, DNA gyrase, and topo IV by stabilizing a covalent cleavage complex of the enzyme and DNA with a double-stranded break which is toxic for a cell (5). Despite their potent activity against many relevant Gram-negative and Gram-positive pathogens, the frequency of emergence of isolates resistant to FQs is relatively high (6). In most cases, mutations that confer resistance to FQs are located in quinolone resistance-determining regions (QRDRs) of GyrA/ParC or, more rarely, of GyrB/ParE subunits of DNA gyrase/topo IV (7–9). The extensive misuse of antibiotics, including FQs, has led to the emergence of several resistant pathogens that are of clinical relevance, for example, (i) Gram-positive bacteria such as methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant enterococci (VRE) strains and (ii) Gram-negative bacteria such as Pseudomonas aeruginosa, Acinetobacter baumannii, and Escherichia coli (7, 9, 10).

A possible way of combating fluoroquinolone-resistant (FQR) species is based on the use of nybomycin compounds (NYBs), so-called “reverse antibiotics” which selectively inhibit growth of certain types of bacteria resistant to FQs but possess little or no activity against corresponding wild-type FQ-sensitive (FQS) strains. The first compounds shown to have such properties are nybomycin (NYB) and deoxynybomycin (DNM) (11), which suppress growth of FQR strains of S. aureus having either mutant GyrA or both GyrA and ParC. The S83L (E. coli numbering) mutation in GyrA is most prevalent among susceptible strains. In contrast, the compounds demonstrated little or no activity against FQS S. aureus strains lacking such mutation in GyrA; moreover, sensitive FQR strains acquired resistance to the compounds via backward mutation of residue 83 to serine. Reverse antibiotics were able to inhibit FQR DNA gyrase from S. aureus with the S83L GyrA mutation, and no activity was observed against the wild-type enzyme. The presence of an equivalent mutation in ParC, a protein structurally similar to GyrA, had no effect on susceptibility in vivo. Finally, it was concluded that the mutant DNA gyrase, and not wild-type or mutant topo IV, is a key target of NYB and DNM in S. aureus. Researchers supposed that such a type of antimicrobials is an evolutionarily derived weapon that “cuts down” the otherwise endless process of acquiring resistance by returning multidrug resistant species to a basal level of susceptibility.

After the first report of NYB and DNM as reverse antibiotics, the family of NYB compounds (NYB, DNM, and their derivatives), or NYBs, was extended further and studied in detail as an alternative scaffold for antimicrobial agents. Synthetic derivatives of NYB and DNM generally have a tendency to act as potent inhibitors of a panel of Gram-positive FQR pathogens with the S83L/I GyrA mutation, while not affecting wild-type strains of these organisms (10, 12, 13). Unfortunately, little is still known about NYBs’ activity against other bacteria, especially Gram-negative species, which could potentially be due to the poor cellular accumulation of these compounds (12). However, the available data indicate that NYB compounds do not always have a strong “reversibility” against these organisms (e.g., E. coli) (12–14), which needs to be explained.

In this work, we studied the activity of NYB on different E. coli topoisomerases. Unlike S. aureus DNA gyrase, both E. coli enzymes, FQS and FQR enzymes with the S83L mutation, are inhibited by NYB. Moreover, NYB suppresses the activity of E. coli topo IV with higher potency, which may be an indication that this enzyme could be primarily targeted by the compound in E. coli. We further demonstrate that NYB inhibits human topoisomerase IIα, which could explain NYB cytotoxicity against several human cancer and normal cell lines (15).

RESULTS

NYB inhibits growth of both FQS and FQR E. coli.

Previously, we found NYB to be active against FQS E. coli ΔtolC, a strain with enhanced permeability due to the lack of the tolC gene, which mediates efflux of antibiotics and toxins (15). According to previously published data for S. aureus (11), FQR mutants are expected to exhibit high sensitivity to NYB. To verify this prediction for E. coli, we selected FQR strains obtained from E. coli ΔtolC grown at different FQ concentrations to further assay their susceptibility to the compound. Resistant clones harbored the S83L or D87Y mutation in GyrA, which was validated by sequencing the genome. The MIC values for NYB and the fluoroquinolone ciprofloxacin (CIP), as well as for other FQ and non-FQ antibiotics, were then assessed (Table 1). Interestingly, MIC values for NYB were not in agreement with the concept of reverse antibiotic. While CIP was much less active against resistant S83L and D87Y mutants (10- and 5-fold increases in MIC values, respectively), we did not observe any increase in NYB activity for the S83L mutant. Furthermore, NYB had a slightly lower potency against cells having the D87Y GyrA variant (2-fold increase in MIC). Whole-genome sequencing of wild-type and mutant strains further confirmed that the obtained results were from the respective GyrA substitutions, as only these mutations were scored upon analysis.

TABLE 1.

MIC values of nybomycin, fluoroquinolones ciprofloxacin and levofloxacin, and other antibiotics against E. coli strain JW5503 ΔtolC and its fluoroquinolone-resistant derivatives with mutations in GyrA^a

Compound	MIC (μg/ml) for indicated GyrA mutation
	None	S83L	D87Y
NYB	2.5	2.5	5
CIP	0.01	0.1	0.05
LVX	0.02	0.2	0.1
ERY	4	4	4
CHL	0.5	1	0.5
AMP	2	2	2

^aNYB, nybomycin; CIP, ciprofloxacin; LVX, levofloxacin; ERY, erythromycin; CHL, chloramphenicol; AMP, ampicillin.

NYB affects activity of the wild-type E. coli DNA gyrase and its mutant variant S83L GyrA.

In order to understand if the activity of NYB in vivo could be explained by its mode of action against E. coli DNA gyrase, we then studied how the compound influences DNA cleavage mediated by the wild-type and S83L GyrA mutant enzymes in vitro (Fig. 1A and C). In this type of assay, gyrase binds supercoiled DNA in the absence of ATP and may cause DNA relaxation. FQ antibiotics stabilize a cleavage complex with a double-stranded DNA break, which leads to formation of linear DNA, while the activity of novel bacterial type II topoisomerase inhibitors (NBTIs) and NYBs causes production of nicked, or open circular DNA, that has a single-stranded break. Values for 50% inhibitory concentration (IC₅₀) and minimal concentration at which the inhibitory effect was detected are listed in Table 2. For the wild-type enzyme (Fig. 1A), we found that the presence of NYB led to an increase in nicked DNA even at 2 μM (0.625 μg/ml), which is 4 times lower than the MIC, while addition of the same solution lacking NYB (NYB buffer) did not cause any detectable changes in the relaxation pattern. At higher concentrations of NYB, the appearance of the linear DNA form was detected (at 4 μM and higher) and complete inhibition of relaxation was observed (at 67 μM and higher). CIP, used as a control, induced double-stranded breaks at 3 and 0.6 μM. We compared this result with the inhibition of S83L GyrA E. coli enzyme caused by NYB (Fig. 1C). While this variant was highly resistant to CIP, it was only slightly more sensitive to NYB. At a concentration as low as 2 μM, NYB caused the appearance of both nicked and linear DNA forms, and their corresponding intensities were slightly higher than the inhibition observed for the wild-type enzyme. Gyrase-mediated relaxation was suppressed to a significant extent at 17 μM NYB. As an additional control, we assessed the activity of NBTI compound GSK126 against the wild-type and S83L GyrA enzymes (see Fig. S1 in the supplemental material). Although the compound was reported to inhibit E. coli gyrase by inducing single-stranded breaks in DNA (16), we did not observe such an effect at a concentration as high as 9.5 μM, whereas suppression of DNA relaxation began to be identifiable at only 47.5 μM GSK126. We also did not detect any difference in GSK126 activities against two gyrase variants.

FIG 1.

Effect of S83L GyrA mutation on susceptibility to NYB in DNA cleavage (A and C) and supercoiling (B and D) assays. In the DNA cleavage assay, as illustrated in panel E, the wild-type (WT) or S83L GyrA (S83L) DNA gyrase binds negatively supercoiled DNA (SC) and may relax DNA by producing a number of relaxed (R) topoisomers (gyrase lane). In the presence of DNA gyrase inhibitors which stabilize cleavage complexes, the reaction may result in formation of enzyme-DNA complexes with linear DNA (L) or open circular DNA (OC) that has a single-stranded break. Treatment of reactions with SDS and proteinase K removes the enzyme from mixtures and allows detection of different DNA forms by gel analysis. In the DNA supercoiling assay, as depicted in panel F, the wild-type or S83L GyrA DNA gyrase binds relaxed DNA and produces negatively supercoiled DNA in ATP-dependent manner (gyrase lane). When DNA gyrase inhibitors are added, the conversion may be disrupted, which can be observed by gel analysis. DNA, supercoiled/relaxed DNA not treated with DNA gyrase. NYB buf, NYB buf L, and NYB buf H reactions have the same amount of DMSO and salts as in NYB (2 to 134, 8 to 134, and 402 μM, respectively) but lack NYB.

TABLE 2.

IC₅₀ values and minimal effective concentrations of NYB and CIP in in vitro assays^a

Antibiotic	IC50 (μM)/minimal effective concn (μM)
	DNA cleavage		DNA supercoiling		kDNA decatenation
	WT gyrase	S83L GyrA gyrase	WT gyrase	S83L GyrA gyrase	Topo IV	Human topo IIα
Nybomycin	47 ± 17/4	41 ± 15/2	211 ± 40/268	75*/134	70 ± 7/67	15*/17
Ciprofloxacin	1.2 ± 0.1/0.6	R	0.3 ± 0.2/0.4	R	16*/8	NA

^aIC₅₀ values are shown as averages ± SD and were calculated from at least two experiments, except for the values labeled with an asterisk, which were obtained from a single experiment. For NYB in DNA cleavage assays, IC₅₀ indicates NYB concentration at which the fraction of supercoiled DNA equals 50% (with WT gyrase) or at which the fraction of relaxed DNA subtracted from the fraction of supercoiled DNA equals 50% (S83L mutant); see Materials and Methods for details. The minimal effective concentration is the concentration of the compound at which inhibitory effect (DNA linearization in cleavage assay, supercoiling suppression in supercoiling assay, and decatenation suppression in decatenation assay) starts to be observed and describes the gels in Fig. 1 and 2 and Fig. S1. R, inhibitory effect was not observed due to the resistance of the target; NA, inhibitory effect was not assessed.

To assess the difference in NYB activities against the wild-type and S83L GyrA E. coli DNA gyrases in more detail, we studied how NYB inhibits DNA supercoiling mediated by these enzymes (Fig. 1B and D). In the presence of ATP, gyrase is able to introduce negative supercoils into relaxed DNA, causing formation of supercoiled DNA, and gyrase inhibitors can halt the process. Calculated IC₅₀ values and minimal effective concentrations are listed in Table 2. We found that for the wild-type enzyme (Fig. 1B), 402 μM NYB completely suppressed the formation of the supercoiled form, while for the S83L mutant (Fig. 1D), a 3-times-lower concentration (134 μM) was enough to produce the same effect. The blurry bands of topoisomers and shifts in mobility of nicked and supercoiled forms observed at high NYB concentrations could be due to possible intercalative properties of the compound. We then examined whether NYB intercalates DNA by using a classical intercalator test (Fig. S2). Our results demonstrated that the compound exhibited properties of an intercalating agent, as the profile was reminiscent of ethidium bromide activity.

NYB inhibits E. coli topoisomerase IV and human topoisomerase IIα.

Topo IV shares a significant similarity to DNA gyrase, and inhibitors of type II topoisomerases (e.g., FQ and NBTIs) usually target both of them. To test if another E. coli type II topoisomerase could be more susceptible to NYB than DNA gyrase, which could explain the decreased sensitivity observed in vivo for the S83L GyrA mutant, we assayed the activity of the compound against topo IV (Fig. 2A) in a decatenation assay. Topo IV is able to decatenate a kinetoplast DNA (kDNA) network, producing separate DNA circles, which have higher mobility than the substrate, and topo IV inhibitors prevent the separation. IC₅₀ values and minimal effective concentrations are listed in Table 2. We found that NYB inhibited decatenation activity of topo IV at 67 μM and higher. At the two lowest concentrations of NYB tested, the enzyme successfully decatenated circles of kDNA, but the resulting set of DNA forms were represented mostly by open circular or completely relaxed DNA instead of the distribution of topoisomers observed for NYB buffer. The increase in this form is probably due to DNA intercalation by NYB and/or stabilization of cleavage complexes with nicked DNA that were released upon treatment with SDS/proteinase K. However, the same pattern was observed in the absence of treatment (Fig. S3), which could be explained by disruption of such complexes by addition of loading buffer containing SDS. At 34 μM NYB, minicircles in kDNA were still decatenated and released DNA had increased mobility compared to that obtained with the addition of NYB buffer instead of the antibiotic. The increase in NYB concentration gradually suppressed the decantation activity of the enzyme, while free DNA circles migrated faster, reaching maximum mobility at 134 μM NYB. At the highest concentration tested (268 μM), NYB almost completely inhibited catalytic function of topoisomerase IV, and only a minor fraction of released circles was observed. Taken together, these results indicate that similar to the inhibition of DNA gyrase, NYB may stabilize covalent cleavage complexes between topoisomerase IV and nicked DNA at relatively low concentrations (equal to or 2 times higher than MIC value) but starts to suppress topo IV catalytic activity at concentrations (67 μM) lower than required for DNA gyrase inhibition (268 μM for the wild-type gyrase-mediated DNA supercoiling and 134 μM for the S83L mutant).

FIG 2.

Inhibition of topo IV (A) and human topoisomerase IIα (B) by NYB in a decatenation assay with SDS/proteinase K treatment. In the DNA decatenation assay, E. coli topoisomerase IV or human topoisomerase IIα binds kinetoplast DNA (kDNA) formed by the network of DNA circles and produces nicked DNA (open circular) or completely relaxed circles and relaxed topoisomers (topo IV and topo IIα lanes). The presence of topoisomerase inhibitors may either stop the process, result in an incompletely resolved DNA network (K), or induce linear DNA formation. The presence of intercalating agents may lead to supercoiled DNA formation. Treatment of reactions with SDS and proteinase K removes the enzyme from mixtures and allows detection of linear and nicked DNA forms by gel analysis. DNA, kDNA not treated with the enzyme. NYB buf, NYB buf L, and NYB buf H have the same amount of DMSO and salts as in NYB (1 to 134, 8 to 134, and 268 μM, respectively) but lack NYB.

In order to explain the cytotoxic properties of NYB reported previously (15), we assessed its activity against human topoisomerase IIα, which is known to be a target of several NYBs (17). From the decatenation assay (Fig. 2B; IC₅₀ value and minimal effective concentration are listed in Table 2), we observed that at 17 μM, NYB started to inhibit the enzyme’s catalytic activity, and a further increase to 134 μM led to nearly complete suppression of kDNA decatenation.

NYB putative binding site.

A review of the available data shows that there are some species against which NYBs either act as potent reverse antibiotics (S. aureus and enterococci) (10, 11, 13) or seem not to have such an effect (E. coli, A. baumannii, Enterobacter cloacae, Klebsiella pneumoniae, and P. aeruginosa) (12–14). This difference is probably due to the structural features of their gyrases, which are resistant to the compounds when serine (S83 in GyrA) is present and show greatly increasing susceptibility if serine is changed to leucine/isoleucine (as in S. aureus) or are sensitive to the compounds in the wild-type state and show a small, if any, increase in sensitivity with serine-to-leucine/isoleucine substitution in GyrA (like in E. coli). In order to understand what structural factors may define different affinity to NYB, we decided to predict and analyze its putative binding site in DNA gyrase using molecular docking.

In contrast to FQs that stabilize gyrase-mediated double-stranded DNA breaks, NYB activity leads to the accumulation of nicked DNA (Fig. 1A and C), similarly to that obtained with NBTI compounds. This fact suggests that NYB and FQs likely inhibit different stages of DNA gyrase catalytic cycle. In conjunction with this, molecular docking of NYB was carried out with three different types of DNA gyrase-DNA complexes: with uncleaved DNA and with a single-stranded or double-stranded break in DNA. Molecular docking was performed with all three types of complexes for S. aureus gyrase and for a complex with a double-stranded DNA break for E. coli enzyme (no available structures with uncleaved DNA or a single-stranded break in DNA).

The best result was achieved for S. aureus DNA gyrase with a single-stranded DNA break, consistent with the observed stabilization of nicked DNA-gyrase complex by NYB. Docking results for complexes with uncleaved DNA for S. aureus and double-stranded breaks in DNA for both species were unacceptable because of the low docking score, suggesting that nicked DNA is essential for NYB binding. NYB may bind in a hydrophobic pocket between GyrA subunits, the site relatively distant from that of FQ moxifloxacin (Fig. 3). The planar system of conjugated rings of NYB partially intercalates into DNA, which could explain its intercalative properties. In addition, hydrogen bonds can be formed between the carbonyl oxygen of NYB and the backbone of M121 (M120 in E. coli) as well as between the hydroxyl group of NYB and the side chain of D83 (D82 in E. coli). Moreover, R122 (R121 in E. coli) can also play an important role in NYB-DNA gyrase binding, forming a cation–π interaction with the aromatic system of NYB.

FIG 3.

NYB putative binding site in DNA gyrase according to the docking results shown in different views (A and B). Molecular docking was performed with the structure of S. aureus DNA gyrase bound to DNA (PDB ID 5IWI), and then the E. coli DNA gyrase structure (PDB ID 6RKS) was aligned to the resulting complex. To compare the NYB binding site with that of fluoroquinolones, the structure of S. aureus DNA gyrase (not shown) bound to moxifloxacin (FQ; PDB ID 5CDQ) was aligned to the resulting complex. S. aureus DNA gyrase is represented in white, and E. coli DNA gyrase is shown in green. Yellow and light blue dash lines depict hydrogen bonds and π-π stacking between NYB and S. aureus DNA gyrase and DNA base pairs, respectively. Residues of E. coli and S. aureus DNA gyrases that could be responsible for NYB binding are shown as sticks and are labeled with slashes.

To explain the differences in susceptibility to NYB between the wild-type enzymes of E. coli and S. aureus, two DNA gyrase structures were aligned (Fig. 3). The residues that are probably involved in the interaction coincide between structures. The key difference from structural alignments lies in the hydrophobic pocket where NYB was predicted to bind. Residue I74 of E. coli DNA gyrase is more hydrophobic than M75 of the S. aureus enzyme, which could provide a basis for stronger binding of NYB in the E. coli gyrase pocket. However, there is no direct contact between this position and NYB, indicating that the presence of methionine may lead to conformational changes in the NYB binding site (e.g., through a shift of flexible R121), making NYB binding less favorable and more S83L dependent. To understand if this difference is conserved between species against which NYBs are “reverse” and “normal,” we aligned GyrA sequences from these organisms (Fig. 4A). We found that position 74 (E. coli numbering) is represented by methionine in the first group (“reverse”), while isoleucine is located in GyrA of the second group (“normal”) of bacteria. Moreover, residues at positions 89, 98, 116, and 125 differ in their physicochemical characteristics between these two groups of species. As with position 74, although these residues are distant from the predicted binding site, they could influence NYB reversibility via conformational changes. Such separation of selected organisms based on NYB reversibility appears to reflect their relationship, since phylogenetic trees built using GyrA (Fig. 4B) and 16S rRNA gene (Fig. 4C) sequences are almost identical, showing the same division of species into two clades.

FIG 4.

(A) Multiple alignment for GyrA. Numbering is relative to the E. coli sequence. Sequences from organisms against which nybomycins are reverse or seem not to be reverse are placed below or above the blue line, respectively. The red star indicates position 74, represented by isoleucine in the upper group and by methionine in the bottom group and which could determine difference in reversibility/reversibilities of nybomycins against these species. Blue stars point out residues that could potentially interact with NYB within its putative binding site in DNA gyrase. (B and C) Molecular phylogenetic analyses built using GyrA protein sequences (B) and 16S rRNA gene sequences (C) by the maximum likelihood method. The trees with the highest log likelihood (−7,786.71 for GyrA and −5,176.32 for 16S rDNA) are shown.

DISCUSSION

According to the concept of reverse antibiotics, NYBs should be inactive against FQS bacteria and potent against FQR species by selectively inhibiting FQR DNA gyrase with the S83L/I GyrA mutation (E. coli numbering). Despite an obvious need for the study of NYB activity against a panel of the wild-type organisms and isogenic FQR mutants to decipher a clear consensus in reversibility, some trends can be already revealed. In most Gram-positive species studied, S83L/I GyrA mutation indeed leads to an increase in susceptibility, whereas other FQR GyrA substitutions and mutations in ParC homological to the S83L in the FQR enzyme seem to have little or no effect in vivo (10, 11, 13). In fact, a significant difference in sensitivities between FQS wild-type and S83L GyrA mutant S. aureus is observed, with the half-inhibitory concentration of NYB in the cleavage assay being about 30 times higher for the wild-type gyrase than for the FQR mutant, which could explain the 30-fold decrease observed in MIC values (11). In Gram-negative organisms, including E. coli, in which gyrase is a primary target of FQs (18), frequent S83L GyrA and equivalent mutations in ParC generally seem either not to change sensitivity or to slightly increase it, but to a much lower extent than the effect of the same mutation in Gram-positive strains (12–14). However, in all these studies, FQS and FQR strains were not clearly isogenic, meaning that other factors apart from gyrase/topo IV mutations could be responsible for the observed effect. According to our results on isogenic E. coli strains, FQS E. coli with increased permeability to antibiotics is sensitive to NYB, and classical S83L mutation in GyrA does not increase sensitivity in vivo, making gyrase only slightly more susceptible in vitro. Equal sensitivities of the wild-type and GyrA S83L mutant E. coli (and probably for all Gram-negative strains in general) to NYB might be caused by preferential inhibition of topo IV rather than DNA gyrase. Thus, so far, reversibility of NYBs is naturally observed only against certain organisms and only against their gyrases, not topoisomerases IV. In this case, reverse activity of NYBs can be extended to Gram-negative species by chemical modifications, so that novel NYB compounds would be able to primarily inhibit DNA gyrase in these organisms and rely mostly on the S83L substitution or other FQR-related mutations in order to retain this unusual reverse effect and indeed act as a treatment of Gram-negative FQR infections.

While the observed absence of effect of the S83L mutation in vivo could be explained by different primary targets of NYB in Gram-positive and Gram-negative species, it is still interesting that we detected NYB activity against FQS E. coli DNA gyrase, which, according to the concept of reverse antibiotics, should be resistant to NYB. Even though S83L mutation indeed renders the enzyme more sensitive to the compound, the increase in sensitivity is relatively small. It is opposite to NYB activity against S. aureus enzyme, where the wild-type gyrase is initially highly resistant to NYB and FQR mutation seems to significantly improve binding of the antibiotic. In our attempt to find structural features responsible for such a difference in activities of NYB, we tried to predict and analyze its binding site in DNA gyrase. According to our results from molecular docking, nicked DNA seems to be critical for NYB affinity and the compound binds the enzyme at a site distinct from that of FQs, suggesting that the mechanism of action is different from that of FQs and similar to that of NBTI compounds. Analysis of the putative binding pocket of NYB in DNA gyrase revealed several residues (D82, M120, and R121 of GyrA) that could be involved in the interaction. Noticeably, these residues are highly conserved among type II topoisomerases of E. coli and S. aureus (Fig. S4), indicating that NYB binding site in topoisomerase IV may be identical to that in DNA gyrase. Interestingly, D82 is important for binding NBTIs, which inhibit gyrase in a way similar to that of NYB (19, 20), and the D82N substitution is observed in E. coli mutant resistant to the derivative of NYB, 6DNM-NH3 (17).

The most obvious difference that we observed near the binding site between S. aureus and E. coli gyrases is position 74 in GyrA, represented by methionine and isoleucine, respectively. Methionine is also present in enterococcal enzyme, in which NYBs are reverse, and isoleucine is located in gyrases of species where this feature is either weak or absent, which could indicate a role for position 74 in the reversibility status of NYB compounds. Moreover, the equivalent position in the ParC subunit of topoisomerase IV is represented by isoleucine and leucine in S. aureus and E. coli, respectively (Fig. S4), so NYBs are probably not reverse against these targets. This may explain why there was no observable effect of ParC mutations on sensitivity to NYB in S. aureus (11), tested on isogenic strains, or in E. coli, found for a nonisogenic clinical isolate (14). It was previously suggested that position 447 in GyrB, represented by arginine in S. aureus and lysine in E. coli, may define the reversibility of NYB derivative DNM-2 (13). However, this position is located relatively far away from NYB putative binding site suggested by our docking analysis, on the opposite side of the DNA helix (Fig. S5), so its effect on reversibility would probably be weaker than that of the I74M GyrA substitution. It is possible that position 447 may influence activity of DNM-2 since the compound is shown to be inactive against the wild-type E. coli gyrase (10), in contrast to NYB, and thus probably has a different mode of action. Unfortunately, we were not able to test the hypothesis proposed from the molecular docking data, as the search for NYB binding site lies outside the scope of this work. Further studies involving analysis of the residues potentially participating in interactions with NYB, or determining reversibility of the compound, would be crucial to investigate the mode of NYB binding and reveal new horizons to the development of novel nybomycins.

In addition to antimicrobial activity, NYB and some of its derivatives possess cytotoxic activity against different human cancer and normal cell lines (15, 17, 21, 22). This property of the compounds was proposed to be due to either targeting of eukaryotic topoisomerase I or dual inhibition of topoisomerases I and IIα (17, 21). Here, we have shown that the cytotoxicity of NYB we previously reported (15) is likely due to its inhibition of human topoisomerase IIα. Because of the structural homology among many type II topoisomerases, NYB might bind to human topoisomerase IIα at a site similar to the one predicted for bacterial gyrase (Fig. S4). Topoisomerase IIα has equivalent residues (E82 instead of D82 in E. coli and arginine at position 121) which are important for the proposed interaction. However, proline (P803) at the position 120 in the human enzyme (instead of methionine in bacteria) is not able to be a donor of a hydrogen bond, suggesting a lower affinity of NYB.

MATERIALS AND METHODS

Strains.

E. coli ΔtolC (E. coli JW5503), a kanamycin-resistant strain, was obtained from the Keio collection (23). Isogenic strains with S83L and D87Y mutations in GyrA were obtained previously in our laboratory by selection of clones resistant to FQs. The presence of GyrA mutations was confirmed by sequencing the gyrA gene.

Selection of ciprofloxacin- and nybomycin-resistant mutants.

E. coli ΔtolC was grown overnight in LB medium supplemented with 50 μg/ml of kanamycin, and then 10⁹ cells were plated on LB agar plates containing 50 μg/ml of kanamycin and different concentrations of CIP (from 0.01 to 0.5 μg/ml) and NYB (from 2 to 20 μg/ml). Several colonies appeared on plates with 0.05 and 0.1 μg/ml of CIP after 24 h of incubation; these were cultivated and the MICs of CIP and NYB were determined. No colonies were observed on plates with NYB. This procedure was repeated several times with different concentrations of the compound. CIP-resistant mutants were analyzed by whole-genome sequencing.

Whole-genome sequencing.

Genomic DNA was extracted from overnight cultures (PureLink genomic DNA minikit; Thermo Fisher Scientific). Approximately 3 μg of DNA was used for whole-genome Illumina sequencing. Library preparation was carried out by using the NEBNext Ultra II DNA library prep kit for Illumina. Indexing was performed with a set of indexing primers (NEBNext multiplex oligonucleotides for Illumina [96 unique dual index primer pairs]). Sequencing was carried out via the HiSeq 4000 platform (Illumina) in paired-ends reading mode with a read length of 150 bp (150 + 150).

Proteins.

E. coli topoisomerase IV and human topoisomerase IIα were purchased from TopoGEN (catalog no. TG2000EC-1 and TG2000H-1, respectively). Calf thymus topoisomerase I was purchased from Invitrogen (catalog no. 38042-024).

For overexpression of wild-type E. coli GyrA and GyrB, E. coli strain K-12 from the ASKA(-) (24) collection having a plasmid with a corresponding gene tagged by six histidine residues at its N terminus was used. For S83L GyrA, the plasmid for overexpression was derived from the vector containing the wild-type E. coli GyrA [from the ASKA(-) collection] variant by PCR mutagenesis (S83L forward primer, 5′-CTGGCGGTCTATGACACG-3′; S83L reverse primer, 5′-GTCACCATGGGGATGG-3′). After PCR, the resulting vector was transformed into E. coli JM109 chemically competent cells, the resulting mutation was confirmed by sequencing, and the vector was then transformed into E. coli BL21 chemically competent cells.

To obtain a protein, an overnight culture of a strain with a corresponding plasmid in LB broth with 34 μg/ml of chloramphenicol (CHL) was diluted 1:100 in 300 ml of LB broth with 34 μg/ml of chloramphenicol and grown at 37°C with shaking at 200 rpm until the mid-log phase, in which overexpression was induced by adding 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) and continued for 3 to 4 h at 37°C with shaking at 200 rpm. The cell pellet was washed with phosphate-buffered saline (PBS), frozen in liquid nitrogen, and stored at −80°C. Then cells were thawed on ice and resuspended in 5 ml of lysis buffer (50 mM HEPES-KOH [pH 7.6], 1 M NH₄Cl, 10 mM MgCl₂, 7 mM β-mercaptoethanol, 0.1% [vol/vol] Triton X-100, 0.2 mM phenylmethylsulfonyl fluoride [PMSF], 0.3 mg/ml lysozyme), followed by incubation for 15 min on ice and sonication. The debris was removed by centrifugation at 16,000 × g for 30 min at 4°C, and the clear yellowish lysate was incubated for 1 h at 4°C under constant agitation with 400 μl of nickel-nitrilotriacetic acid (Ni-NTA) resin suspension (Ni Sepharose 6 fast flow, catalog no. 17531804; GE Healthcare) and washed with lysis buffer. The resin was washed thrice with 10 ml of wash buffer (50 mM HEPES-KOH [pH 7.6], 1 M NH₄Cl, 10 mM MgCl₂, 7 mM β-mercaptoethanol, 30 mM imidazole) for 5 min at 4°C under agitation. The protein of interest was then eluted with elution buffer (300 μl to 1 ml), whose composition was the same as that of wash buffer, except that the imidazole concentration was increased up to 200 mM. The resulting elution was either mixed 1:1 (vol/vol) with glycerol and stored at −20°C or directly purified. For the wild-type GyrA protein, the elution was dialyzed against dialysis buffer (50 mM HEPES-KOH [pH 7.6], 100 mM KCl, 10 mM MgCl₂, 7 mM β-mercaptoethanol) and then mixed 1:1 with glycerol. For the S83L GyrA variant, the elution was purified with Amicon Ultra-15 centrifugal filter units (Merck; catalog no. UFC900308) using a dialysis buffer containing 50% (vol/vol) glycerol to purify the sample. For GyrB, the elution was dialyzed against dialysis buffer with 30% (vol/vol) glycerol. The purity and concentration of proteins were verified by spectrophotometry and SDS-PAGE; all proteins had a single major band of corresponding size. The purified samples were stored at −20°C.

Antibiotics.

A stock solution of ciprofloxacin (CIP; 10 mg/ml) was diluted in water to 100 μg/ml (302 μM), with pH adjusted to ∼7.5 by adding a small amount of 2 M Tris-HCl (pH 7.5). The next dilutions were performed in water. NYB powder obtained as described previously (15) was dissolved in concentrated hydrochloric acid to 15 mg/ml, and this solution was used to prepare the final NYB solution, containing 400 μg/ml (1.34 mM) of NYB, 5% (vol/vol) dimethyl sulfoxide (DMSO), 10 mM Tris-HCl (pH 7.5), and 30 mM NaCl, with a pH of 7.5 to 7.8. A light pellet was observed, and homogenized suspension was either directly used in assays or diluted further in NYB buffer which had the same conditions as 1.34 mM solution but lacked NYB. Levofloxacin (LVX) solution was prepared similarly to CIP, erythromycin (ERY) solution was prepared in DMSO, and CHL solution was prepared in ethanol. Ampicillin (AMP) solution was prepared in water. GSK126 (obtained from Cayman Chemical Company) stock solution (5 mg/ml; 9.5 mM) was prepared in 100% DMSO, and further dilutions were performed in water.

MIC determination.

Overnight cultures of tested strains were diluted 1:200 to 1:500 in LB medium, and 2-fold dilutions of studied antibiotics were prepared in these cultures. Ninety-six-well 2-ml deep-well plates containing cultures with and without antibiotic, and LB medium as a control, were incubated at 37°C with shaking at 200 rpm overnight. Absorbance at 590 nm was then measured for each well, and MIC was determined as the lowest concentration of an antibiotic suppressing bacterial growth. The procedure was repeated twice for different preparations of NYB solution and at least thrice for other antibiotics.

Assay for gyrase-mediated DNA cleavage.

A DNA cleavage assay for DNA gyrases was performed according to the method of Fisher and Pan (25), with the following modifications. Supercoiled pHOT-1 plasmid was used instead of pBR322. The final concentrations were as follows: GyrA, 36 ng/μl; GyrB, 137 ng/μl; and pHOT-1, 20 ng/μl. To prepare 10× enzyme mix, the required amounts of GyrA and GyrB were mixed and the volume was adjusted with dilution buffer (50 mM HEPES-KOH [pH 7.6], 100 mM KCl, 10 mM MgCl₂, 7 mM β-mercaptoethanol, 50% [vol/vol] glycerol). In the negative-control (DNA) reaction, the dilution buffer was added instead of 10× enzyme mix, and water was added instead of tested compounds. For the positive control (gyrase reaction), water was added instead of tested compounds. In order to verify that effects of NYB addition were not due to the salt/DMSO presence, the same solution without NYB was added as a control (NYB buffer). For CIP activity, IC₅₀ was calculated as the concentration at which the fraction of linear DNA is 50% (not assessed for the S83L mutant due to the resistance). At the IC₅₀ of NYB, the fraction of supercoiled DNA is 50% (for wild-type enzyme) or the difference between fractions of supercoiled DNA and relaxed topoisomers is 50% (for the S83L mutant), since such approach allowed a better approximation of the data. At minimal effective concentrations, an increase in linear DNA starts to be observed. Such a parameter was used for NYB since the appearance of double-stranded breaks correlated better with its concentration than the appearance of nicked DNA.

Gyrase-mediated DNA supercoiling.

A DNA supercoiling assay for DNA gyrases was performed according to the method of Fisher and Pan (25), with the following modifications. Relaxed pHOT-1 plasmid was used instead of pBR322. The final concentrations were as follows: GyrA, 5 ng/μl; GyrB, 6.3 ng/μl; and pHOT-1, 10 ng/μl. To prepare 10× enzyme mix, the required amounts of GyrA and GyrB were mixed and the volume was adjusted with dilution buffer (50 mM HEPES-KOH [pH 7.6], 100 mM KCl, 10 mM MgCl₂, 7 mM β-mercaptoethanol, 50% [vol/vol] glycerol). In the negative control (DNA reaction), the dilution buffer was added instead of 10× enzyme mix, and water was added instead of tested compounds. In the positive control (gyrase reaction), water was added instead of tested compounds. In order to verify that effects of NYB addition were not due to the salt/DMSO presence, the same solution without NYB was added as a control (NYB buffer). IC₅₀s were calculated as the concentrations at which the fraction of supercoiled DNA was 50% relative to that of the gyrase reaction. The value was not calculated for the S83L mutant due to the resistance (for CIP). At minimal effective concentrations, suppression of supercoiling starts to be observed.

Intercalation assay.

An intercalation assay was performed in 50 mM Tris HCl (pH 7.5), 50 mM KCl, 10 mM MgCl₂, 0.5 mM dithiothreitol (DTT), 0.1 mM EDTA, and 30 μg/ml of bovine serum albumin (BSA) with 12 ng/μl of supercoiled/relaxed pHOT-1 DNA, in the presence or absence of 0.5 U of topoisomerase I and tested compounds. To verify that effects of NYB addition were not due to the salt/DMSO presence, the same solution without NYB was added as a control (NYB buffer). Half of the reaction volume was incubated at 37°C for 30 min and then treated with 0.45% (vol/vol) SDS and 0.45 mg/ml of proteinase K at 37°C for 15 min. The second half of the reaction volume was treated with butanol and chloroform/isoamyl alcohol according to the manual for a DNA unwinding assay kit (Inspiralis; catalog no. DUKSR001) to remove a possible intercalator. After treatment, 1:6 (vol/vol) 5× stop solution (5% [vol/vol] SDS, 25% [vol/vol] glycerol, 0.25 mg/ml of bromophenol blue) was added. For reactions without butanol and chloroform/isoamyl alcohol treatment, 1:6 (vol/vol) 5× stop solution was added after incubation with SDS/proteinase K. Samples were then analyzed by electrophoresis in 1% (wt/vol) agarose in 1× TBE (90 mM Tris, 90 mM boric acid, 2.5 mM EDTA) containing 1 mg/ml of chloroquine for better resolution of topoisomers. Gels were stained with ethidium bromide after running and UV photographed.

Decatenation assay for E. coli topoisomerase IV.

A decatenation assay was performed using the E. coli topoisomerase IV drug screening kit (TopoGEN, catalog no. TG1007-1). Each reaction contained 1× complete topoisomerase IV reaction buffer with 13 ng/μl of kDNA (from the insect trypanosome Crithidia fasciculata; the average size of minicircles was 2.5 kb; supplied by TopoGEN), 2 U of topoisomerase IV (except negative control) diluted in topoisomerase IV dilution buffer (TopoGEN), and tested compounds at appropriate concentrations. Reaction mixtures were incubated at 37°C for 30 min and then stopped by adding 1:6 (vol/vol) 5× stop solution (5% [vol/vol] SDS, 25% [vol/vol] glycerol, 0.25 mg/ml of bromophenol blue) and treated with 0.24% (vol/vol) SDS and 120 μg/ml of proteinase K at 37°C for 10 min when necessary. Samples were analyzed by electrophoresis in 1% (wt/vol) agarose in 1× TBE (90 mM Tris, 90 mM boric acid, 2.5 mM EDTA), with subsequent staining with ethidium bromide after running, and UV photographed. In the negative control (DNA reaction), topo IV dilution buffer was added instead of enzyme and water was added instead of tested compounds. In the positive control (topo IV reaction), water was added instead of tested compounds. In order to verify that effects of NYB addition were not due to the salt/DMSO presence, the same solution without NYB was added as a control (NYB buffer). At the IC₅₀ of each inhibitor, the fraction of decatenated kDNA is 50% relative to the that of the topo IV reaction. At minimal effective concentration, suppression of decatenation starts to be observed.

Decatenation assay for human topoisomerase IIα.

A decatenation assay was performed using a human topoisomerase II assay kit (TopoGEN; catalog no. TG1001-1A). Each reaction mixture contained 200 ng of kDNA (TopoGEN), 4 U of the enzyme (except negative control), and tested compounds when needed. In the negative control, (DNA reaction), water was added instead of both enzyme and tested compounds. In the positive control (topo IIα reaction), water was added instead of tested compounds. Samples were analyzed by electrophoresis in 1% (wt/vol) agarose in 1× TBE (90 mM Tris, 90 mM boric acid, 2.5 mM EDTA), with subsequent staining by ethidium bromide after running, and UV photographed. To verify that effects of NYB addition were not due to the salt/DMSO presence, the same solution without NYB was added as a control (NYB buffer). At the IC₅₀ of NYB, the fraction of decatenated kDNA is 50% relative to that of the topo IIα reaction. At the minimal effective concentration, suppression of decatenation starts to be observed.

Molecular docking.

Molecular docking was carried out with three structures of S. aureus DNA gyrase complex, corresponding to the different stages of the catalytic cycle: complexes with uncleaved DNA (PDB identifier [ID] 5IWM) and with single-stranded (PDB ID 5IWI) or double-stranded (PDB ID 5CDQ) breaks in DNA. Additionally, molecular docking was performed with E. coli gyrase in complex with double-stranded breaks (PDB ID 6RKS) in DNA. Three-dimensional (3D) models for molecular docking were generated and optimized by using Protein Preparation Wizard available in the Maestro software package (release 2019-4; Schrödinger, LLC, New York, NY). The NYB molecule was converted to 3D structures using the LigPrep tool (release 2019-4; Schrödinger, LLC). Searching for tautomers and isomers as well as energy minimization for compounds was carried out by applying the OPLS (optimized potentials for liquid simulations) force field. The receptor grid was centered around residue S84 of GyrA (S. aureus numbering) as the most crucial position for NYB resistance (11) using the receptor grid generation tool. Molecular docking was performed with extra precision approaches of the Glide package (26). Results were visualized by using PyMOL (version 2.0; Schrödinger, LLC).

For structural alignment of the S. aureus and E. coli DNA gyrases, structures with PDB IDs 5IWI and 6RKS, respectively, were used, and the corresponding protein chains (GyrA and GyrB) were aligned. To compare binding sites of NYB and FQs, the structure of S. aureus DNA gyrase bound to moxifloxacin (PDB ID 5CDQ) was aligned to the resulting complex of S. aureus DNA gyrase and NYB.

Multiple-sequence alignment.

For GyrA alignment, sequences from different organisms obtained from the UniProt database (27) (accession numbers: E. coli K-12, P0AES4; A. baumannii ATCC 19606, D0CBH9; Enterobacter cloacae subsp. cloacae ATCC 13047, A0A0H3CRA9; Klebsiella pneumoniae subsp. pneumoniae DSM 30104, J2X497; P. aeruginosa ATCC 15692, P48372; S. aureus N315, Q99XG5; and Enterococcus gilvus ATCC BAA-350, R2V8F0) were aligned using the GenomeNet CLUSTALW tool (https://www.genome.jp/tools-bin/clustalw) with default settings (slow/accurate), and the alignment figure was generated using CLC Sequence Viewer (CLC bio) and GIMP 2.10.6 (https://www.gimp.org).

For 16S rRNA gene alignment, sequences (partial or complete) from different organisms obtained from the NCBI nucleotide (https://www.ncbi.nlm.nih.gov/nucleotide) database (accession numbers and positions where required: E. coli K-12, CP032667.1, 454613 to 456166; A. baumannii ATCC 19606, CP046654.1, 1353818 to 1355361; E. cloacae subsp. cloacae ATCC 13047, NR_102794.2; K. pneumoniae subsp. pneumoniae DSM 30104, AJJI01000003.1, 368476 to 370008; P. aeruginosa ATCC 15692, AF094715.1; S. aureus N315, NC_002745.2, 506162 to 507715; and E. gilvus ATCC BAA-350, NR_043793.1) were aligned using the GenomeNet CLUSTALW tool (https://www.genome.jp/tools-bin/clustalw) with default settings (slow/accurate).

Structural alignment of type II topoisomerases and corresponding sequence alignment were carried out with the FATCAT web tool (http://fatcat.godziklab.org) using structures of GyrA (PDB IDs 6RKS and 5IWI for E. coli K-12 and S. aureus N315, respectively), ParC (PDB IDs 1ZVU and 2INR for E. coli K-12 and S. aureus MSSA476, respectively), and human topoisomerase IIα (PDB ID 5GWK). The alignment figure was generated using CLC Sequence Viewer (CLC bio) and GIMP 2.10.6 (https://www.gimp.org).

Molecular phylogenetic analysis.

The evolutionary history was inferred by using the maximum likelihood method based on the Tamura-Nei model (28). Initial trees for the heuristic search were obtained automatically by applying neighbor-joining and BioNJ algorithms to a matrix of pairwise distances estimated using the maximum composite likelihood (MCL) approach and then selecting the topology with superior log likelihood value. The trees are drawn to scale, with branch lengths measured in the number of substitutions per site. All positions containing gaps and missing data were eliminated. There were totals of 827 (GyrA) and 1,403 (16S rRNA gene) positions in the final data set. Evolutionary analyses were conducted in MEGA7 (29).

Figure preparation and gel quantification.

For gel images, figures were generated and assembled using GIMP 2.10.6 (https://www.gimp.org). ImageJ software (30) was used to quantify the band intensity from gels. For each band, raw data (distance – gray value) were collected and the curve was approximated by Gaussian peaks. Areas were then calculated under peaks of each fraction (e.g., supercoiled DNA/relaxed topoisomers/nicked DNA or catenated kDNA/fully + partly decatenated kDNA) and the background noise was subtracted. Areas of interest were normalized to the total area of the DNA forms in the lane and then, if necessary, normalized to the areas in enzyme lane and areas of DNA lane were subtracted. The final DNA fractions were values from 0 to 1, dependent on the concentration of inhibitors. Curves for DNA fraction/inhibitor concentration were then fitted to sigmoid function, and IC₅₀ values were calculated.