Evybactin is a DNA gyrase inhibitor that selectively kills Mycobacterium tuberculosis

Abstract

The antimicrobial resistance crisis requires the introduction of novel antibiotics. The use of conventional broad-spectrum compounds selects for resistance in off-target pathogens and harms the microbiome. This is especially true for Mycobacterium tuberculosis, where treatment requires a 6-month course of antibiotics. Here we show that a novel antimicrobial from Photorhabdus noenieputensis, which we named evybactin, is a potent and selective antibiotic acting against M. tuberculosis. Evybactin targets DNA gyrase and binds to a site overlapping with synthetic thiophene poisons. Given the conserved nature of DNA gyrase, the observed selectivity against M. tuberculosis is puzzling. We found that evybactin is smuggled into the cell by a promiscuous transporter of hydrophilic compounds, BacA. Evybactin is the first, but likely not the only, antimicrobial compound found to employ this unusual mechanism of selectivity.

Since the discovery of streptomycin in 1943, many additional compounds to combat tuberculosis have been introduced^1–3. These agents include natural products such as rifampicin as well as synthetic antibiotics. Despite these advances, tuberculosis remains a major infectious disease, killing millions of people each year⁴. The need for a lengthy treatment leads to poor compliance and, with it, the rise of antibiotic resistance^5,6. The currently recommended regimen is a 2-month intensive phase comprised of rifampicin, isoniazid, pyrazinamide and ethambutol, followed by a 4–6-month continuation phase with rifampicin and isoniazid⁴. Of these, rifampicin is broadly active against Gram-positive bacteria, whereas isoniazid, pyrazinamide and ethambutol act selectively against M. tuberculosis. More recent additions to the tuberculosis antimicrobial arsenal include selective compounds, such as bedaquiline, an inhibitor of the mycobacterial ATP-synthase; moxifloxacin, a broad-spectrum fluoroquinolone poison of bacterial type II topoisomerases, such as DNA gyrase; and nitroimidazoles delamanid and pretomanid, which are mycolic acid biosynthesis inhibitors. Selectively acting compounds have a distinct advantage, as they do not harm the microbiome and do not contribute to general antibiotic resistance by acting against off-target bacteria. The microbiome affects many aspects of human health and disease⁷, and application of broad-spectrum antibiotics is especially problematic for a treatment that lasts 6 months.

The inclusion of broad-spectrum compounds in the current regimens of tuberculosis treatment prevents us from taking advantage of selectively acting anti–M. tuberculosis antibiotics. With this in mind, we have pursued a targeted selective approach, screening environmental microorganisms against M. tuberculosis and counter-screening against Staphylococcus aureus. Apart from providing for selective compounds, differential screening eliminates the very large background of toxic and known antimicrobials, markedly increasing the probability of finding target-specific antibiotics. Given that traditional cultivable actinomycetes have been largely overmined, we had focused on screening previously uncultured bacteria grown in situ. This screen led to the discovery of lassomycin, an inhibitor of the mycobacterial ClpP1P2C1 protease from Lentzea kentuckyensis sp., and amycobactin, an inhibitor of mycobacterial SecY protein exporter, produced by Amycolatopsis sp^8,9.

Compounds from uncultured soil bacteria did not evolve to function in animals. By contrast, Xenorhabdus and Photorhabdus, which are symbionts of nematode microbiome, produce compounds that must function systemically in an animal. These bacteria are released into insect larvae when nematodes invade them and produce antimicrobials. The compounds need to travel through the tissues of the larvae and be non-toxic to their nematode host. Odilorhabdins produced by Xenorhabdus are broad-spectrum inhibitors of translation and show efficacy in animal models of infection^10,11. We recently identified darobactins, a novel class of antibiotics produced by Photorhabdus that act selectively against Gram-negative bacteria both in vitro and in vivo^12,13. Darobactins inhibit BamA, a chaperone that inserts β-barrel proteins, such as porins, into the outer membrane.

Whether Photorhabdus harbor compounds with selective activity against mycobacteria is unknown. In the present study, we set out to explore this possibility. Here we report the discovery of evybactin, a novel cyclic depsipeptide DNA gyrase inhibitor produced by P. noenieputensis that acts selectively against M. tuberculosis. Evybactin is smuggled into M. tuberculosis by its multi-solute transporter BacA, which explains the mechanism of selectivity. Surprisingly, evybactin acts at a site known to be targeted by synthetic thiopene agents, an allosterically acting class of gyrase antagonists¹⁴, distinguishing its mode of action from widely used fluoroquinolone antibiotics.

Results

Identification of evybactin.

We screened culture extracts from 58 strains of Photorhabdus and Xenorhabdus nematode symbionts against M. tuberculosis H37Rv mc²6020 expressing mCherry as a growth indicator⁸. To avoid non-specifically acting compounds, we used S. aureus as a counter-screen. This screen yielded 33 hits, 12 of which were selective against M. tuberculosis. A supernatant from P. noenieputensis DSM 25462 showed potent activity against M. tuberculosis and was inactive against S. aureus. An extract from a culture grown in TNM-FH medium emulating insect hemolymph had a two- to fourfold-higher activity as compared to Luria–Bertani Broth (LBB) and Tryptic Soy Broth (TSB). To isolate the active compound, the supernatant was concentrated and fractionated using high-performance liquid chromatography (HPLC). Then, 101 fractions were collected, and activity was present in fractions 57–68. Active fractions were further subjected to HPLC for final purification. High-resolution electrospray ionization–mass spectrometry (MS) analysis revealed the molecular mass of the active compound ([M + H] + = 1,488.68), which did not match any known compounds in Antibase (Supplementary Fig. 1). The structure of the compound, which we named evybactin (1), was determined by nuclear magnetic resonance (NMR) and MS spectroscopic analyses (Fig. 1a, Extended Data Figs. 1 and 2, Supplementary Figs. 2, 3 and 4 and Supplementary Tables 1 and 2). The peptidic nature of evybactin was evident from the ¹H and ¹³C NMR spectra showing the presence of the α-protons and amide carbonyl signals. A detailed two-dimensional (2D) NMR spectroscopic analysis (COSY, HSQC, HMBC and ROESY) determined the composition and sequence of the amino acid residues in the peptide. The discrimination between α- and β-aspartic acids was particularly challenging. The 1,1-ADEQUATE in combination with ROESY and HMBC experiments revealed the presence of two β-aspartic acids in the molecule. The absolute configuration of the amino acids was determined by Marfey’s analysis. This structurally novel antibiotic is a cyclic depsipeptide composed of 12 L- and D-amino acids with an N-formylated branch.

Fig. 1 |. Evybactin is produced by P. noenieputensis.

a, The structure of evybactin. b, The BGC of evybactin. Gene alignment of the BGC of evybactin in the producer strain. A–e are NrPS genes, and T1 and T2 are transporter genes.

The biosynthetic gene cluster (BGC) of evybactin was determined using bioinformatic analysis of the genome. The genome was sequenced by a combination of Nanopore and Illumina reads (Microbial Genome Sequencing Center (MiGS)) and assembled into two contigs with a total size of 5.5 megabases. antiSMASH 5.0 (ref. ¹⁵) was used to analyze BGCs in the contigs. As amino acids are the main building blocks of evybactin, the non-ribosomal peptide synthetase (NRPS) with adjacent tailoring enzymes, such as formyltransferase and methyltransferase, was deduced to be responsible for the biosynthesis of evybactin. The BGC of evybactin was identified as NRPS with a core BGC spanning 49.6 kilobases (Fig. 1b, Supplementary Fig. 5 and Supplementary Table 3). The number of NRPS modules is in accordance with the number of amino acids in evybactin. The BGC has five core type I NRPS genes containing 12 linear modules. Adenylation domains were used to predict amino acid substrate specificity using antiSMASH and Prism default settings. A formyltransferase was identified from module 1, which is consistent with formylation of the tryptophan amide. Similarly, a methyltransferase was identified from module 5, consistent with the N-methylation of the histidine moiety. Both arginine loading modules (modules 3 and 10) contain an epimerization domain. Modules 7 and 11 were predicted to incorporate serine moieties, whereas an epimerization domain was found only in module 7. These results are consistent with Marfey’s analysis and suggest that the serine in module 7 has a D-configuration (Supplementary Table 2). The GC content of the evybactin BGC is 46%, and there is no closely related BGC in other bacterial species based on antiSMASH search. It appears that evybactin is unique to P. noenieputensis.

Spectrum of activity.

Notably, evybactin is highly potent against M. tuberculosis, with a minimum inhibitory concentration (MIC) of 0.25 μg ml⁻¹ (Table 1). Consistent with its isolation based on differential screening, it has lower activity against other pathogens. We tested evybactin against commensal bacteria, including Lactobacillus sp. and Bacteroides sp., and found no activity (Table 1). Additionally, evybactin showed no toxicity against HepG2, FaDu and HEK293 human cells (Table 1). Taken together, these results demonstrate that evybactin is highly selective against M. tuberculosis. This selectivity suggested action against a target specifically present in Mycobacteria and absent from human cells.

Table 1 |.

Spectrum of evybactin

Strain	BacA/SbmAg	MIC (μg ml−1)
Pathogenic bacteria (MIC)
Mycobacterium tuberculosis H37Rv mc26020 mCherryd	Present	0.0625
Mycobacterium tuberculosis H37Rv mc26020d	Present	0.25
Mycobacterium tuberculosis H37Rv mc26020 ΔbacAd	Present	16
Mycobacterium tuberculosis H37Rv mc26020 gyrA G88Sd	Present	64
Mycobacterium tuberculosis H37Rv mc26020 gyrA G88Cd	Present	128
Mycobacterium smegmatis mc2155d	Present	8
Mycobacterium abscessus ATCC 19977d	Present	16
Staphylococcus aureus HG003e	Absent	128
Escherichia coli WO153 (AB1157; recJ asmB1 ΔtolC::Kanr)e	Present	0.0625
Escherichia coli ΔtolC a,e	Present	0.25
Escherichia coli ATCC 25922e	Present	8
Escherichia coli BW25113e	Present	16
Escherichia coli MG1655e	Present	16
Escherichia coli ΔsbmA a,e	Absent	128
Klebsiella pneumoniae ATCC 700603e	Present	32
Clostridium perfringens KLE 2523b,c,f	Absent	64
Enterococcus faecalis KLE 2341b,c,f	Absent	>128
Acinetobacter baumannii ATCC 17978e	Present	128
Salmonella enterica KLE 2601b,c,f	Present	128
Pseudomonas aeruginosa PA01e	Present	>128
Symbiotic bacteria (MIC)
Lactobacillus reuteri LTH5448b,f	Absent	>128
Lactobacillus paracasei KLE 2504b,c,f	Absent	>128
Streptococcus parasanguinis KLE 2509b,c,f	Absent	>128
Bacteroides fragilis KLE 2244b,c,f	Absent	64
Bacteroides stercoris KLE 2537b,c,f	Absent	>128
Veillonella ratti KLE 2365b,c,f	Absent	>128
Human cell line (IC50)
HepG2	ND	>128
FaDu	ND	>128
HEK293	ND	>128

^aKeio collection mutants

^bCultivated under anaerobic conditions

^cHuman stool isolate, K.L. laboratory collection

^dDifco Middlebrook 7H9 medium

^eMHIIB medium

^fBHI-Ych medium

^gPresence of BacA (NP_216335.1)/SbmA (NP_414911.1) homologs (above 50% identity) were analyzed by Blastp search. ND, no data.

Animal efficacy.

We next tested evybactin in a simple model of septicemia infection with Escherichia coli to evaluate its potential for activity in vivo. First, to assess the toxicity of evybactin in a mouse, evybactin was administrated intraperitoneally at 100 mg kg⁻¹, and survival was observed for 24 hours. There were no indications of toxicity. Because evybactin showed a relatively low MIC against E. coli American Type Culture Collection (ATCC) 25922, we used this strain for a preliminary animal study. Mice were infected with E. coli intraperitoneally for 1 hour, followed by intraperitoneal administration of evybactin. A single dose of 25 mg kg⁻¹ of evybactin showed significant efficacy, and a 100 mg kg⁻¹ dose of evybactin completely protected mice from E. coli infection, whereas 83% of untreated control animals died within 24 hours (Extended Data Fig. 3).

Mechanism of selectivity.

To identify the target of evybactin, we isolated resistant mutants from M. tuberculosis H37Rv mc²6020. M. tuberculosis cells were seeded onto 7H9 nutrient agar medium containing 10× MIC of evybactin and gave rise to evybactin-resistant mutants with a frequency of 7.2 × 10⁻⁶ to 1.6 × 10⁻⁵ (Fig. 2a). We sequenced the whole genome of three spontaneous mutants and found that all of the strains carry mutations (L469P, L470P and S577R) in the membrane transporter bacA (Rv1819c) (Fig. 2b and Supplementary Table 4). All the spontaneous evybactin-resistant mutants, as well as a bacA deletion mutant, had an increased MIC (8–16 μg ml⁻¹ or 32–64 × MIC) (Table 1).

Fig. 2 |. Evybactin is transported into the cell via ABC transporter BacA and targets DNA gyrase.

a, Frequency of generating drug-resistant mutants in M. tuberculosis. b, evybactin-resistant mutations mapped in BacA. TMD, transmembrane domain; NBD, nucleotide-binding domain. c, evybactin-resistant mutations mapped onto the structure of the MtbGyrase DNA cleavage core (PDB ID: 5BS8 (ref. ³³)). Yellow: moxifloxacin binding site, the catalytic tyrosine responsible for DNA cleavage (GyrA_Y129) and moxifloxacin site are labeled in the inset. d, Scheme of the incorporation of evybactin and resistance mechanism. eVY, evybactin; IM, inner membrane; PG, peptidoglycan; OM, outer membrane.

Several antibiotics are taken up by specific transporters such as sideromycins and rifabutin^16,17. BacA is annotated as a vitamin B₁₂ transporter¹⁸; however, a recent study proposed that BacA serves as a multi-solute ABC-type transporter for hydrophilic molecules¹⁹. Evybactin is a highly hydrophilic compound whose solubility in water is more than 40 mg ml⁻¹. BacA was previously reported to transport bleomycin²⁰. We, therefore, evaluated the susceptibility of bacA evybactin-resistant mutants to bleomycin. All three mutants showed decreased susceptibility to bleomycin (MIC shift from 0.06 μg ml⁻¹ to 1 μg ml⁻¹). This result suggests that BacA is non-functional in the M. tuberculosis mutants resistant to evybactin. We also evaluated the effect of vitamin B₁₂ on evybactin susceptibility in M. tuberculosis but observed no change in the susceptibility to evybactin even in the presence of 1,000-fold molar excess of vitamin B₁₂ (Supplementary Table 5). Apparently, B₁₂ and evybactin bind to different sites of this multi-solute transporter.

BacA homologues are sparsely distributed among other bacteria and are found in E. coli (SbmA, which serves as a peptide antibiotic microcin transporter)^21,22 (Extended Data Fig. 4). We took advantage of this homology to test the possible role of SbmA in the susceptibility of E. coli to evybactin. The evybactin MIC for wild-type E. coli is 16 μg ml⁻¹, which is considerably higher than M. tuberculosis. Susceptibility to evybactin further decreases in an E. coli mutant with a knockout in sbmA (Table 1). Notably, we found that E. coli WO153 with a compromised penetration barrier²³ is highly sensitive to evybactin, with an MIC of 0.0625 μg ml⁻¹ (Table 1). E. coli WO153 expresses less lipopolysaccharide (LPS) and lacks the outer membrane porin TolC that serves as a docking port for multidrug resistance pumps. The MIC for an E. coli tolC deletion mutant was 0.25 μg ml⁻¹, which is four times higher than the E. coli WO153 strain, and 64 times lower than in the wild-type. This finding suggests that both the outer membrane barrier and efflux across it contribute to the high resistance of wild-type E. coli to evybactin. Taken together, these results suggest that, in E. coli, evybactin is transported into the cell by SbmA but efficiently effluxed by TolC-dependent MDRs (Fig. 2d). The Gram-positive M. tuberculosis lacks a similar restrictive penetration barrier and, as a result, is sensitive to evybactin, which is transported into the cell by BacA. This consideration explains the selectivity of evybactin against M. tuberculosis and also explains resistance of gut commensal bacteria that lack a BacA-type transporter.

The target of evybactin.

Considering that BacA is non-essential and likely only serves as the transporter for evybactin, we assumed that the true target is located in the cytoplasm. M. tuberculosis cells treated with evybactin were approximately two times longer than non-treated cells (Fig. 3a,b). This morphological change is typically observed in bacteria treated by DNA synthesis inhibitors, FtsZ inhibitors and β-lactams^24–26. In addition, a label incorporation assay revealed that evybactin inhibits DNA synthesis but has less of an effect on RNA, protein and fatty acid synthesis (Extended Data Fig. 5). These results suggest that evybactin inhibits DNA synthesis.

Fig. 3 |. Efficacy of evybactin.

a, Optical microscopy and analysis of M. tuberculosis exposed to 10× MIC antibiotics. None, no antibiotic; rif, rifampicin (0.6 μg ml⁻¹); Mox, moxifloxacin (1.25 μg ml⁻¹); Cip, ciprofloxacin (1.25 μg ml⁻¹); evy, evybactin (2.5 μg ml⁻¹). Scale bars, 5 μm. b, Cell elongation in the presence of antibiotics. c,d, The time-dependent killing of early exponential (c) and stationary (d) cells of M. tuberculosis by 10× MIC evybactin (2.5 μg ml⁻¹) and moxifloxacin (1.25 μg ml⁻¹). n = 3 biologically independent samples. Data are mean ± s.d. Significance was determined by ordinary one-way ANOVA with Tukey’s multiple comparisons test. None versus rif, P = 0.0256 (*), CI (−0.5441, −0.02235); None versus Mox, P < 0.0001 (****), CI (−0.7592, −0.2153); None versus Cip, P < 0.0001 (****), CI (−1.436, −0.8729); None versus evy, P < 0.0001 (****), CI (−2.061, −1.522). CI, confidence interval.

To identify the true target of evybactin, we selected M. tuberculosis mutants in the presence of a high concentration of the compound, 25 μg ml⁻¹ (100× MIC), to avoid selection for bacA mutants (MIC 8 μg ml⁻¹, 32× MIC). This treatment resulted in the selection of mutants highly resistant to evybactin, at a frequency of 6.1 × 10⁻⁹ to 1.7 × 10⁻⁸ (Fig. 2a). We sequenced the whole genome of two high-level evybactin-resistant mutants and found that they harbor G88S or G88C mutations in gyrA (Rv0006), which codes for DNA gyrase subunit A (Fig. 2c and Supplementary Table 6). We confirmed that the high level of evybactin resistance is due to mutations in gyrA by constructing gyrA recombinant M. tuberculosis mutants in a clean background (Table 1).

To test whether there is cross-resistance between evybactin and moxifloxacin, a known DNA gyrase inhibitor, we isolated spontaneous moxifloxacin-resistant mutants of M. tuberculosis (GyrA D94N). We then evaluated the susceptibility of evybactin and moxifloxacin against evybactin-resistant mutants (GyrA G88C and G88S) and the moxifloxacin-resistant mutant (GyrA D94N). The GyrA G88C mutation is known to confer fluoroquinolone resistance to M. tuberculosis²⁷, and, in agreement with this finding, GyrA G88C mutant is resistant to moxifloxacin (Supplementary Table 7). However, contrary to our expectations, the GyrA G88S mutation makes M. tuberculosis more susceptible to moxifloxacin, whereas the GyrA D94N mutation did not have any effect on evybactin susceptibility (Supplementary Table 7). These results suggest that the two structurally distinct compounds bind differently to DNA gyrase.

DNA gyrase poisons are bactericidal antibiotics. We, therefore, evaluated the killing ability of evybactin. The compound was highly bactericidal against exponentially growing and stationary M. tuberculosis with activity similar to moxifloxacin, which is often used as a second-line antibiotic for extended multidrug-resistant mutants of M. tuberculosis (Fig. 3c,d and Table 1).

Evybactin is a non-quinolone gyrase poison.

To verify gyrase as the target of evybactin in M. tuberculosis, we conducted in vitro biochemical assays using the purified M. tuberculosis enzyme (MtbGyrase). Type II topoisomerases, including DNA gyrase, regulate DNA supercoiling and chromosome entanglements by creating a transient double-stranded break in one DNA duplex and passing a second double-stranded segment through the break^28,29. Bacterial type II topoisomerase poisons, such as moxifloxacin, corrupt this strand passage process, leading to persistent double-stranded DNA breaks and cell death^30,31.

We first tested the ability of evybactin to induce MtbGyrase-mediated DNA cleavage by titrating the enzyme against a fixed amount of the compound and plasmid, quenching with EDTA and SDS/proteinase K, and separating the reactants by native agarose gel electrophoresis. Evybactin not only inhibited the supercoiling activity of MtbGyrase but also induced DNA cleavage, as evidenced by the persistence of relaxed plasmid DNA at intermediate concentrations of enzyme and the appearance of linear cleavage products at higher enzyme concentrations (Fig. 4a, top panel). Because E. coli WO153 and tolC deletion strains were susceptible to evybactin, we further assayed the ability of the agent to induce cleavage by the two known E. coli type II topoisomerases, gyrase and topoisomerase IV (topo IV). Evybactin stimulated DNA cleavage by both enzymes to an extent similar to that of MtbGyrase (strong cleavage visible at 5 nM enzyme for both species) (Fig. 4a, middle and bottom panels); however, the compound was a more potent inhibitor of DNA supercoiling by gyrase than of DNA supercoil relaxation by topo IV. These data demonstrate that, similarly to the fluoroquinoline antibiotics, evybactin is a general poison of bacterial type IIA topoisomerases, with a preference for gyrase as compared to topo IV.

Fig. 4 |. Evybactin is a bacterial gyrase and topo IV poison.

a, Native agarose gel-based supercoiling assays for M. tuberculosis gyrase (top panel) and E. coli gyrase (middle panel) and supercoil relaxation assays for E. coli topo IV (bottom panel). enzyme amounts are indicated in nanomolar units (0–20 nM; plasmid DNA is 6 nM). Assays were conducted in the absence or presence of 100 μM evybactin. The migration positions for linear, nicked and supercoiled DNAs are labeled as indicated. All assays were repeated at least three times with similar results. b, Quantitation and IC₅₀ determination for evybactin-stimulated and moxifloxacin-stimulated DNA cleavage activity using 20 nM M. tuberculosis gyrase, 6 nM plasmid and 0–100 μM drug. Cleavage was assessed from reactions run on agarose gels containing ethidium bromide, which separates nicked and linear species from closed DNA circles (relaxed or supercoiled) to aid quantitation. Values are plotted as mean values ± s.d. n = 3 independent biochemical experiments. c, ATP-independent supercoil relaxation assay for M. tuberculosis gyrase. enzyme amounts are indicated in nanomolar (0–160 nM, with 6 nM plasmid). Assays were conducted in the absence or presence of 100 μM evybactin. d, ATPase rates for MtbGyrase (250 nM) containing no drug, 100 μM evybactin or 100 μM moxifloxacin. Data are reported in molecules of ATP per second per enzyme; values are plotted as mean values ± s.d. n = 3 independent biochemical experiments.

Because evybactin and moxifloxacin exhibited similar killing of M. tuberculosis, and both act as gyrase poisons, we sought to compare the effect of the two compounds on MtbGyrase in vitro. Using a fixed amount of MtbGyrase and titrating moxifloxacin in the presence of ATP, we observed robust levels of DNA cleavage (half maximal inhibitory concentration (IC₅₀) ≈ 1 μM) (Fig. 4b). Evybactin stimulated cleavage with similar efficiency to moxifloxacin in the presence of ATP, with an IC₅₀ also close to 1 μM. Interestingly, little to no cleavage was observed for evybactin in the absence of ATP, whereas moxifloxacin produced robust nucleotide-independent cleavage at a concentration of 5–10 μM. Given the strict ATP dependence for evybactin-induced cleavage, we were interested in assessing the effect of evybactin on the ATP-independent supercoil relaxation activity of MtbGyrase (in addition to supercoiling relaxed DNA in the presence of ATP, MtbGyrase can relax supercoiled substrates in the absence of ATP³²). We conducted supercoiling relaxation assays with MtbGyrase, analogous to the supercoil relaxation assays performed for E. coli topo IV, but without ATP (Fig. 4c). The resultant data showed that MtbGyrase exhibits similar efficiencies of supercoil relaxation in either the absence or presence of 100 μM evybactin (evybactin was able to induce DNA nicking at higher enzyme concentrations in the absence of ATP but produced no linear DNA products).

To further investigate the differences between the mechanisms of action for moxifloxacin and evybactin, we tested the effects of both compounds on the DNA-stimulated ATPase activity of MtbGyrase. Whereas evybactin had a negligible effect on the overall rate (V_max) of MtbGyrase’s ATPase activity (0.9 ± 0.2 molecules of ATP per second per enzyme with or without the compound), the addition of moxifloxacin resulted in a roughly twofold reduction of ATP consumption (0.4 ± 0.1 molecules of ATP per second per enzyme) (Fig. 4d). Taken together, the disparate effects of the D94N and G88S mutations on MIC values obtained for evybactin and moxifloxacin (Supplementary Table 7), the strict dependence on ATP by evybactin as well as its negligible effect on ATPase activity suggested that the mechanism of action for the molecule is distinct from that of fluoroquinolones.

Evybactin binds to an allosteric site targeted by thiophenes.

To more precisely determine how evybactin acts on MtbGyrase, we co-crystallized a portion of the enzyme bound to the compound and a singly nicked duplex DNA substrate (Methods). For this effort, we employed a fusion construct comprising the DNA binding and cleavage region of M. tuberculosis gyrase (MtbGyrBA_core) that we generated previously to study cleavage complexes of the enzyme bound to fluoroquinolone poisons³³ but harboring a Y129F mutation, which prevents the enzyme from cleaving DNA. The resultant structure revealed that evybactin binds at a site distal to the fluoroquinolone binding pocket (Fig. 5a and Supplementary Table 8). The macrocycle depsipeptide portion of evybactin engages a winged-helix domain in GyrA using an ‘edge-on’ conformation in which the protein is engaged by only one face of the compound (Fig. 5b). Edge-on binding poses are among the more common types of protein–macrocycle interactions reported in the literature and, along with ‘face-on’ and ‘compact’ binding modes where the macrocycle lays flat or within a binding pocket, account for most observed geometries of macrocycles bound to proteins³⁴.

Fig. 5 |. Crystal structure of evybactin bound to M. tuberculosis gyrase.

a, Structures of gyrase bound to evybactin (left), moxifloxacin (middle, PDB ID: 5BS8 (ref. ³³)) or a thiophene (right, PDB ID: 5NPK (ref. ¹⁴)). GyrA/GyrB heterodimer subunits are colored in orange and blue, with DNA depicted in gray. The drug binding location is outlined, with the drugs themselves represented as transparent surfaces colored in magenta, yellow and green for evybactin, moxofloxacin and a thiophene, respectively. b, Close-up view of the evybactin binding site. evybactin is depicted in stick representation with a transparent sphere overlay. Hydrogen bonds are represented as dotted black lines, and residues forming the binding pocket of evybactin are represented as sticks and labeled. A subset of the amino acids that comprise the evybactin peptide macrocycle are labeled in circled, magenta, single-letter code. The thiophene binding site (and its overlap with that of evybactin) is indicated with a green outline. c, Intercalation of GyrB r482 at the site of DNA cleavage is shown for the evybactin-bound M. tuberculosis gyrase structure (blue) and for the intercalation of the equivalent r458 in the thiophene-bound S. aureus structure (light blue, PDB ID: 5NPK). The moxifloxacin binding site is illustrated as a yellow outline; the positions of the guanidinium groups of the arginines overlap with that of the fluoroquinolone.

Evybactin is constructed such that one end of its macrocycle is composed of a short, (D)Ser-(L)Phe-(L)Gly-(D)Arg stretch of residues (Fig. 1a). This portion of the compound orients the serine and arginine sidechains atop a largely hydrophilic surface on GyrA (Fig. 5b). The phenylalanine on the macrocycle, as well as nearby methylated histidine, do not appear to directly contact GyrA in the structure (Extended Data Fig. 6a). The opposite end of the compound is a branch terminating in an N-formylated tryptophan residue. Surprisingly, the indole ring of this tryptophan moiety occupies a pocket that has been previously shown to be exploited by the azaindole or chlorophenyl groups of thiophenes, a synthetic class of gyrase poisons that act by an allosteric mechanism (Fig. 5 and Extended Data Fig. 6b)¹⁴.

Given that resistance mutations obtained for evybactin map to the region of the enzyme where fluoroquinolones bind, the placement observed for the compound was unexpected. However, inspection of the structure revealed a conformational commonality that is shared with a thiopene-inhibited complex obtained with the cleavage core of the S. aureus enzyme but not with gyrases bound to fluoroquinolone poisons. In particular, the evybactin-bound and thiopene-bound gyrase structures both contain an unusual architecture in which an arginine (R482 in MtbGyrase and R458 in S. aureus gyrase) intercalates between the DNA bases present in one of the enzyme’s two cleavage centers, displacing one of the bases from that active site. This conformation is striking, given that the arginine sits >20 Å from the site of evybactin or thiophene binding. The position of the intercalating arginine also overlaps with the site where a fluoroquinolone would normally bind to the enzyme (Fig. 5c and Extended Data Fig. 6c).

To further probe the nature of the evybactin–gyrase interaction, we generated 13 MtbGyrase variants that contain single mutations in the binding surface for the compound, along with the MtbGyrA_G88C and MtbGyrA_G88S mutants isolated from in vivo evybactin resistance assays. In contrast to the evybactin binding pocket mutants, MtbGyrA_G88C and MtbGyrA_G88S are located close to the active site tyrosine in GyrA that is required for DNA cleavage. The purified constructs were then screened to look for changes in the ability of evybactin to promote DNA cleavage using an agarose gel-based cleavage assay. As expected from the in vivo resistance screen, MtbGyrA_G88C displayed strong resistance to evybactin-induced and moxifloxacin-induced DNA cleavage in vitro (100-fold versus 80-fold, respectively; Extended Data Fig. 7). Interestingly, MtbGyrA_G88S also led to substantially reduced evybactin-induced DNA cleavage (40-fold), whereas moxifloxacin-induced DNA cleavage was actually stimulated by the G88S mutation (fourfold). The reduction in cleavage activity for MtbGyrA_G88C in the presence of evybactin or moxifloxacin indicates that mutations at this site generally repress drug-induced cleavage. By contrast, the MtbGyrA_G88S mutation results in the introduction of a serine at a position that would be expected to help chelate a magnesium that is known to aid fluoroquinolone binding and stimulate cleavage complex formation^33,35,36. For the MtbGyrase mutants constructed within the evybactin binding pocket, many variants had no or little effect on evybactin-induced cleavage (Extended Data Fig. 8). This observation is consistent with the extended binding site of evybactin along the surface of MtbGyrBA_core, which is mostly composed of weak van der Waals interactions (Fig. 5b). However, MtbGyrA_M33A, MtbGyrA_P353L, MtbGyrA_A32V and MtbGyrA_I36F all displayed reduced cleavage (5–100-fold) in the presence of evybactin compared to wild-type MtbGyrase (Extended Data Fig. 7). These changes map to the hydrophobic binding pocket shared by the thiophenes and the tryptophan residue of evybactin.

Interestingly, along with reduced sensitivity to evybactin, both MtbGyrA_M33A and MtbGyrA_I36F also exhibited a ~fivefold to tenfold reduction in supercoiling activity by the mutant enzymes as compared to wild-type MtbGyrase (Extended Data Fig. 9). This general loss of function likely accounts for the reduced levels of cleavage seen for these constructs in the presence of evybactin. By comparison, the general supercoiling activity of MtbGyrA_P353L and MtbGyrA_A32V, as well as MtbGyrA_G88C and MtbGyrA_G88S, were only slightly reduced overall but showed relatively strong resistance against evybactin (Extended Data Figs. 7 and 9). In E. coli, the P353L and A32V substitutions also give rise to resistance to thiopenes¹⁴, consistent with our structural data showing that the two agents share a common binding pocket. Interestingly, in addition to reducing the cleavage-promoting efficiency of evybactin, MtbGyrA_P353L and MtbGyrA_A32V also promoted resistance to cleavage induced by moxifloxacin (Extended Data Fig. 7), even though these amino acid substitutions are far removed from the site of fluoroquinolone binding. This result indicates that there is allosteric coupling between the binding sites for the two classes of poisons.

Discussion

A differential screen of a small collection of Photorhabdus symbionts of the nematode microbiome resulted in the isolation of evybactin, a novel cyclic depsipeptide antibiotic acting potently and selectively against M. tuberculosis. The compound is highly polar and not well-suited to diffuse across a hydrophobic cytoplasmic membrane. The target is intracellular, the well-conserved bacterial DNA gyrase. All currently known compounds acting selectively against M. tuberculosis hit a unique target (or a unique site). This is not the case with evybactin; its activity depends on the BacA transporter, which explains both the penetration of this polar compound into the cell and the mechanism of selectivity. BacA is an unusual ABC-type ‘multi-solute transporter’ that apparently transports vitamin B₁₂ into the cell¹⁸. The same transporter was also found to translocate hydrophilic bacitracin into M. tuberculosis. Mycobacteria seem to be a rare group of Gram-positive species to harbor BacA; the only other example is Streptococcus pneumoniae, which has a microcin B17 transporter with 99.7% identity to E. coli SbmA and was probably acquired via horizontal gene transfer. Other members of this family of transporters are sparsely scattered among Gram-negative species³⁷. BacA-type transporters seem to be absent in most of human gut symbionts, explaining the low activity of evybactin against them. Notably, evybactin has low activity against wild-type E. coli carrying the SbmA homolog of BacA but is very potent against a mutant with a disrupted outer membrane permeability barrier. Our results suggest that, in E. coli, evybactin penetration is restricted by the outer membrane and efflux by multidrug pumps, and only some of the compound gets smuggled into the cell by SbmA. This efflux system masks the role of SbmA, and we expect that evybactin will be similarly inactive against most Gram-negative bacteria.

The mechanism of action for evybactin is distinct from that of fluoroquinolones. Evybactin binds at an allosteric site distal from the site of fluoroquinolone binding. A portion of this locus was first identified as a binding pocket for synthetic thiophenes¹⁴. The existence of natural products that target this allosteric site highlights the importance of the pocket as a critical node for gyrase activity and a point for small-molecule intervention. Interestingly, DNA cleavage induced by evybactin is highly ATP dependent, further distinguishing this molecule from fluoroquinolones. Comparative structural analyses of the evybactin binding pocket reveals that it is also the binding site for a Corynebacteriales-specific loop within the ATPase domains of DNA gyrase³⁸. In the absence of ATP and DNA, the ATPase domains of MtbGyrase fold down against the cleavage core of the enzyme³⁸, an action that we found occludes the evybactin binding pocket (Extended Data Fig. 10). This ‘open’ conformation of the ATPase domains within MtbGyrase may account for the strict ATP dependence for evybactin-induced cleavage.

Treating tuberculosis requires a constant introduction of novel compounds to combat emerging resistance. The rise of MDR and XDR-TB strains resistant to most currently available antibiotics underscores the need for new therapies. BacA null mutants resistant to evybactin occur with high frequency but have reduced virulence²⁰. This correlation suggests that only low-frequency gyrase mutants may pose a problem. In the case of M. tuberculosis, this concern is ameliorated, because all drugs are introduced as combinations. So far, each new compound has been added to an existing regimen, which provides only a temporary relief from emerging resistance. Ideally, such combinations should contain only novel compounds free of pre-existing resistance; this approach will effectively prevent resistance development. Drug combinations should also consist of M. tuberculosis–selective compounds to avoid harming the microbiome. The current pipeline of anti-tuberculosis drugs in development and efficient methods to discover novel selectively acting natural products makes this strategy realistic.

Online content

Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41589-022-01102-7.

Methods

Screening conditions.

Photorhabdus sp. and Xenorhabdus sp. were cultivated in LBB, TSB and Nutrient Broth (NB) for 8 days at 28 °C. Concentrated culture extract and S. aureus HG003 overlay plate for antibacterial assay were prepared as previously described¹². M. tuberculosis H37Rv mc²6020 (ΔlysA ΔpanCD) expressing mCherry (ΔlysA ΔpanCD, pBEN_mCherry kan^r) was cultured in supplemented Difco Middlebrook 7H9 medium containing kanamycin and incubated at 37 °C and 100 r.p.m.⁸. The culture was diluted into fresh medium to a final optical density at 600 nm (OD₆₀₀) of 0.003, and 148.5 μl of culture was added to the wells of the 96-well black with a flat, clear-bottom microtiter plate (Corning) containing the 1.5 μl of culture extract (1:100 dilution). The plate was incubated for 7 days at 37 °C and 100 r.p.m., at which point the OD₆₀₀ and fluorescence with excitation at 580 nm and emission at 610 nm were measured on a plate reader. The extract was deemed to have activity against M. tuberculosis H37Rv mc²6020 if it had ≥75% growth inhibition when compared to the growth control. The assay was repeated for confirmation of activity. The culture samples, which showed activity only against M. tuberculosis H37Rv mc²6020, were determined as anti-tuberculosis selective extract.

Purification of evybactin.

P. noenieputensis DSM 25462 was inoculated in a 500-ml Erlenmeyer flask with 200 ml of LBB and incubated at 28 °C with aeration at 200 r.p.m. for overnight. Ten milliliters of overnight culture was inoculated into a 2-L Erlenmeyer flask with 1 L of TNM-FH insect medium (Sigma-Aldrich) and incubated for 10–14 days. Cells were removed by centrifugation (8,000g, 5 minutes), and supernatant was incubated with XAD16N resin (20–60 mesh, Sigma-Aldrich) for overnight. After the supernatant was removed, evybactin was eluted from XAD16N resin by 100% methanol. Samples were dried and resuspended in 5 ml of MilliQ water. Then, 5 ml of concentrated culture extract was subjected to reverse-phase HPLC on a C18 column (Luna 5 μm C18(2) 100 Å, LC Column, 250 mm × 21.2 mm, Phenomenex). HPLC conditions were as follows: solvent A, Milli-Q water and 0.1% (v/v) formic acid; solvent B, acetonitrile and 0.1% (v/v) formic acid. The initial concentration of 10% solvent B was maintained for 5 minutes, followed by a linear gradient to 35% over 20 minutes and maintained at 100% for 10 minutes with a flow rate of 7 ml min⁻¹. Fractions were collected every 20 seconds, UV detection by diode-array detector from 210 nm to 400 nm, and active fraction was eluted for 19–24 minutes. Active fraction was subjected to reverse-phase HPLC on a C18 column (XBridge, BEH C18 OBD prep column, 100 Å, 5 μm, 250 mm × 10 mm, Waters). HPLC conditions were as follows: solvent A, Milli-Q water and 0.1% (v/v) formic acid; solvent B, acetonitrile and 0.1% (v/v) formic acid. The initial concentration of 10% solvent B was maintained for 2 minutes, followed by a linear gradient to 38% over 8 minutes with a flow rate of 5 ml min⁻¹, UV detection by diode-array detector from 210 nm. Evybactin was eluted at 8 minutes, with a purity of 92% by UV.

Structure elucidation.

LC–MS analysis was conducted on a 6530 Q-TOF-LC/MS (Agilent). The HPLC column was a reversed-phase ZORBAX RRHT Extend-C18, 2.1 mm × 50 mm, 1.8 μm (Agilent). The mobile phases were water with 0.1% formic acid (A) and acetonitrile with 0.1% formic acid (B). A linear gradient was initiated with 2% acetonitrile and linearly increased to 52% at 2–12 minutes. The flow rate was 0.2 ml min⁻¹, and the injection volume was 5 μl. Mass spectra in the m/z range 111–3,000 were obtained by positive ion (+ESI) modes. The MS conditions were as follows: gas temperature 300 °C, N₂ flow rate 7 L min⁻¹, nebulizer gas pressure 35 psig, capillary voltage 3,500 V, fragmentor potentials 175 V, Vcap 3,500 V, Skimmer 65 V and Octopole RFPeak 750 V. Data acquisition and analysis were conducted using Agilent LC-MS-QTOF MassHunter Data Acquisition software, version 10.1, and Agilent MassHunter Qualitative Analysis software, version 10.0, respectively (Agilent).

All NMR data were recorded on a Bruker AVANCE II 700-MHz spectrometer with a 5-mm TXI probehead and a 600-MHz spectrometer with a cryoprobe. All NMR experiments were performed with 10 mg of evybactin spiked with tetramethylsilane (TMS) in DMSO-d₆ (320 K) and 4% D₂O in H₂O (300 K). All carbon and proton chemical shifts were referenced by TMS. Complete assignments were obtained using 2D experiments, including COSY (cosygpmfqf), TOCSY (dipsi2etgpsi), ¹H-¹³C HSQC (hsqcedetgpsisp2.3), ¹H-¹⁵N HSQC (hsqcetf3gpsi2), ¹H-¹³C HMBC (hmbcgplpndqf), ROESY (roesyphpp.2) and 1,1-ADEQUATE (adeq11etgprdsp_bbhd). All NMR data were processed with TopSpin 4.1.3.

Structural analysis of evybactin.

High-resolution ESI-MS analysis of evybactin showed the following data: m/z 1,488.68 [M + H]⁺, calculated for C₆₄H₉₀N₂₁O₂₁⁺, 1,488.66 (Supplementary Fig. 1). The ¹H and HSQC NMR data recorded in DMSO-d₆ showed numerous amide proton signals (δ_H 7.22–10.09) and α-amino methines (δ_H 4.13–5.22), indicative of a peptidic structure (Supplementary Table 1). This was supported by the presence of carbonyl carbons in ¹³C NMR spectrum. Evybactin showed unusual signals that are not observed in standard amino acids. The singlet methyl signal at δ_H 3.44 indicated the presence of N-methylated residue. N-formyl modification was readily deduced by distinctive signals at δ_C/H 160.6/7.82. Further comprehensive analysis of 2D NMR data (COSY, HMBC and ROESY) enabled the identification of 11 standard amino acids, including two aspartic acids, two serines, two arginines, two threonines, one glycine, one phenylalanine, one histidine and one tryptophan (Extended Data Fig. 2). The strong HMBC correlations from the singlet methyl group H-38 (δ_H 3.44) to C-35 (δ_C 126.6) and C-37 (δ_C 137.4) positioned the N-methyl group C-38 (δ_C 30.6) at the π position of the histidine. Likewise, the formyl group C-64 (δ_C 160.6) was shown to be positioned at the amino group of tryptophan (54-NH) by the HMBC correlation from the methine proton H-54 (δ_H 4.55) to C-64. The connectivity of the identified residues was determined on the basis of HMBC correlations from α-protons to amide carbonyls of adjacent residues. It still remained ambiguous whether aspartic acid was connected via C-1 (backbone) or C-4 (sidechain) because the HMBC experiment cannot differentiate two-bond and three-bond correlations. The 1,1-ADEQUATE spectra were measured to address this issue. Correlations from H-30 (δ_H 4.40) to C-31 (δ_C 176.9) and from H-3 (δ_H 4.38) to C-4 (δ_C 168.7) were observed (Supplementary Fig. 3), indicating the connection via the sidechain of aspartic acid. This result strongly showed that two aspartic acids in the compound are β-aspartic acids and was further supported by ROESY and HMBC experiments. A ROESY correlation between H-29 (δ_H 2.17) and NH-26 (δ_H 6.77) suggested a β-aspartic acid linkage between a serine moiety and a methylated histidine moiety (Supplementary Fig. 4). Another β-aspartic acid linkage between the terminal aspartic acid and threonine through an ester bond was established by the long-range HMBC correlation from H-41 (δ_H 5.02) to C-2 (δ_C 38.3). This connection was further supported by the ROESY correlation between H-2 (δ_H 4.38) and H-41 observed in the spectrum acquired in 4% D₂O in H₂O. The planar structure of evybactin was, therefore, characterized as a cyclic depsipeptide with an N-formylated branch.

The stereochemistries of chiral centers present at α- and β-carbons were assigned by applying derivatization methods coupled with chromatographic analysis. The advanced Marfey’s method using l- and d-FDLA (1-fluoro-2,4-dinitrophenyl-5-leucinamide) established the absolute configurations of amino acids: l-Asp, l-Ser, d-Ser, d-Arg, l-Phe, l-methylhistidine, l-Thr and l-Trp (Supplementary Table 2). The remaining chiral centers at β-carbons were determined by the LC–MS analysis of the GITC (2,3,4,6-tetra-O-acetyl-β-d-glucopyranosyl isothiocyanate) derivatives. The regiochemistry of l- and d-Ser in evybactin relied on A domain analysis of the BGC.

MIC and cytotoxicity.

M. tuberculosis strains including strain H37Rv mc²6020 (ΔlysA ΔpanCD), H37Rv mc²6020 expressing an mCherry plasmid and conferring kanamycin resistance (ΔlysA ΔpanCD, pBEN_mCherry kan^r) and evybactin-resistant mutants of H37Rv mc²6020 and Mycobacterium smegmatis mc²155 and Mycobacterium abscessus ATCC 19977 were used in this study. The MIC was determined for M. tuberculosis H37Rv mc²6020, H37Rv mc²6020 expressing mCherry, M. smegmatis and M. abscessus by broth microdilution. For all strains, a final OD₆₀₀ of 0.003 was obtained by diluting an exponentially growing culture of bacteria into supplemented Difco Middlebrook 7H9 medium (10% Middlebrook Oleic Albumin Dextrose Catalase (OADC) Growth Supplement (Millipore Sigma), 5% glycerol, 1% casamino acids, 0.05% tyloxapol, 80 μg ml⁻¹ of lysine, 24 μg ml⁻¹ of pantothenate and 50 μg ml⁻¹ of kanamycin where appropriate for M. tuberculosis strains, 10% ADC, 5% glycerol and 0.05% tyloxapol for non-tuberculosis strains). The plates were incubated at 37 °C and 100 r.p.m. for 7 days (M. tuberculosis) or 3 days (M. smegmatis and M. abscessus). The MIC was defined as the lowest concentration of antibiotic with no visible growth. For vitamin B₁₂ competition assay, M. tuberculosis H37Rv mc²6020 was incubated in the presence of evybactin and vitamin B₁₂ with molar ratio 1:0, 1:10, 1:100 and 1:1,000, and, after 7-day incubation at 37 °C with aeration at 100 r.p.m., MIC was evaluated. In the case of strains grown under aerobic conditions, overnight cultures were diluted 1:100 in Mueller Hinton II Broth (MHIIB) and incubated at 37 °C with aeration at 220 r.p.m. Exponential cultures (OD₆₀₀ 0.1–0.9) were diluted to OD₆₀₀ 0.001 (approximately 5 × 10⁵ colony-forming units (c.f.u.) ml⁻¹) in MHIIB, and 98-μl aliquots were transferred into round-bottom, 96-well plates containing 2 μl of evybactin solution diluted serially twofold. After overnight incubation at 37 °C, the plate was observed by eye, and the concentration where growth of bacteria markedly inhibited was determined as MIC. For the strains grown under anaerobic conditions (Coy Vinyl Anaerobic chamber, 37 °C, 5% H₂, 10% CO₂, 85% N₂), overnight cultures grown in Brain Heart Infusion (BHI) broth, supplemented with 0.5% yeast extract, 0.1% l-cysteine hydrochloride and 15 μg ml⁻¹ of Hemin (BHI-Ych), were diluted 1:100 in BHI-Ych. Then, 98-μl aliquots were transferred into round-bottom, 96-well assay plates and prepared by twofold dilution of evybactin. After 24-hour incubation, the MIC was determined. All MIC assays were performed at least in triplicate. Cytotoxicity assay was performed as previously described¹².

Time-dependent killing.

The exponential cultures of M. tuberculosis H37Rv mc²6020 were prepared by growing the strain to mid-exponential (OD₆₀₀ 1–1.5) and then backdiluting to OD₆₀₀ 0.003. For stationary phase cultures, M. tuberculosis H37Rv mc²6020 was grown for 2 weeks to an OD₆₀₀ > 1.5. Cultures were challenged with either 10× MIC of evybactin or moxifloxacin. Cultures were incubated at 37 °C and 100 r.p.m. At intervals, 100-μl aliquots were removed from each culture, serially diluted and plated onto supplemented 7H10 medium to determine c.f.u. ml⁻¹. The exponential phase plates were incubated for 3 weeks, and the stationary phase plates were incubated for 2 weeks before counting, both at 37 °C. Experiments were performed with biological and technical replicates.

Staining and fluorescent imaging of M. tuberculosis.

M. tuberculosis cultures were grown to exponential phase (OD₆₀₀ −0.5) and then treated with 10× MIC of the indicated antibiotic. Aliquots were taken after 48 hours of treatment, washed once in PBS + 0.05% Tween 80 (PBST) and fixed in 4% paraformaldehyde for 2 hours at room temperature. The cells were washed once, resuspended in PBST and stained with FM4–64FX (Thermo Fisher Scientific) at a final concentration of 1 μg ml⁻¹ for 30 minutes in the dark at room temperature. Once stained, 1 μl of cells was spotted onto a 1.5% low-melting agarose pad and observed with a Nikon Ti2-E inverted fluorescence microscope using a ×100 oil immersion objective lens. Images were acquired by NIS-Elements at a resolution of 2,048 × 2,048 and processed with Fiji software³⁹. Segmentation and calculation of cellular length from these images was done using the plug-in MicrobeJ⁴⁰.

Resistance study.

Mutants to evybactin in M. tuberculosis H37Rv mc²6020 were selected by plating on supplemented 7H10 medium containing 10× and 100× MIC of evybactin. Isogenic cultures of M. tuberculosis H37Rv mc²6020 were obtained by plating 100 μl of an exponentially growing culture onto supplemented 7H10 medium without antibiotics and incubating at 37 °C for 3 weeks. Three independent colonies were picked and inoculated into 10 ml of supplemented Difco Middlebrook 7H9 medium, grown for 2 weeks, subcultured 1:100 into 40 ml of supplemented Difco Middlebrook 7H9 medium and grown to OD₆₀₀ −1.0. Cells were plated at 10⁷, 10⁸ and 10⁹ concentrations. This was achieved by removing 100 μl of culture, serially diluted and plated in triplicate for c.f.u.; removing 4,400 μl of culture, plating 400 μl (100 μl per plate) onto plates containing either 10× MIC of evybactin or rifampicin, the remaining 4 ml pelleted by centrifugation at 8,000 r.p.m. for 5 minutes, resuspended in 400 μl of supplemented Difco Middlebrook 7H9 medium and plated as previously described; the remaining 36 ml was washed once, pelleted by centrifugation at 8,000 r.p.m. for 5 minutes and resuspended in 4 ml of supplemented Difco Middlebrook 7H9 medium; and 200 μl was plated onto each of the remaining plates, five for each antibiotic. This was repeated for each isogenic culture at 100× MIC. The plates were incubated at 37 °C for 3 weeks, at which point the number of colonies on each plate was counted. Mutation frequency was calculated by dividing the number of mutants obtained by the total bacteria plated. Evybactin mutants were picked and inoculated into 10 ml of supplemented Difco Middlebrook 7H9 medium without antibiotics, grown for 2 weeks and subcultured 1:100 into 10 ml of supplemented Difco Middlebrook 7H9 medium without antibiotics. The broth MIC for the mutants was determined to confirm maintenance of evybactin resistance. Genome sequencing and variant calling were conducted by the MiGS. Whole-genome sequencing was performed by paired-end reads (2× 150 bp) with Illumina NextSeq 550, and M. tuberculosis genome information data in the National Center of Biotechnology Information (NCBI) (NCBI reference sequence: NC_000962.3) were used for variant calling.

Sequencing and variant calling.

Sequencing and variant calling were conducted by the MiGS. Unicycler⁴¹ was used to assemble the genome with Illumina and Nanopore sequencing data from MiGS under default settings. Variant calls were made using the program breseq⁴².

Targeted mutations of bacA and gyrA.

Targeted deletion of bacA in M. tuberculosis H37Rv mc²6020 was accomplished as in a previous study⁴³. The plasmid pMSGzeo was used to construct the recombination substrate consisting of a zeocin resistance marker flanked by DNA fragments with homology to the upstream and downstream regions of bacA⁴⁴. The sequence of the primers used to amplify the flanking regions are bacAKOLFFor −5′-TTAAGATCTCGGGCCA CCGGCGCCACAAAC-3′, bacAKOLFRev −5′-GGGAAGCTTAAACAATTTCGGGCCCAAGG-3′, bacAKORFFor −5′-GGGTCTAGAACGCTGAATCCGTCGATCTC-3′ and bacAKORFRev −5′-TTTGGTACCCTCCGTTACCGATCAGTGG-3′. Null mutants were selected on 7H10 agar supplemented with 100 μg ml⁻¹ of zeocin. Mutation was confirmed via polymerase chain reaction (PCR) and sequencing. Point mutations in M. tuberculosis gyrA were constructed via single-stranded recombineering as in a previous study⁴⁵ with plating on 100× MIC evybactin. Sequence of oligonucloetides used to make targeted mutations were gyrA G88C −5′-ATGCGCACCAGGCTGTCGTAGATCGACGCGTCGCAGTGC GGGTGGTAGTTGCCCATGGTCTCGGCAACCG-3′ and gyrA G88S −5′-ATGCGCACCAGGCTGTCGTAGATCGACGCGTCGCTGTGCGGGTGGTAGTTGCCCATGGTCTCGGCAACCG-3′. Targeted mutations were confirmed via PCR and Sanger sequencing.

Macromolecule incorporation assay.

Macromolecular synthesis assay against DNA, RNA, protein, fatty acids and cell wall was analyzed by ImQuest BioSciences with E. coli WO153 cells.

Purification of recombinant MtbGyrase.

Full-length M. tuberculosis GyrA and GyrB were prepared as previously described³³. In brief, M. tuberculosis GyrA and GyrB were expressed separately from modified pET vectors containing an N-terminal hexahistadine SUMO tag, using BL21[DE3] CodonPlus E. coli cells (Agilent). Cells were grown at 37 °C to mid-log phase in 2× TY media, after which the temperature was reduced to 30 °C, and protein production was induced with 0.5 mM IPTG for 4 hours before harvesting by centrifugation. Cells were resuspended in A1000 (30 mM Tris-HCl (pH 7.8); 1 M NaCl; 10 mM imidazole, pH 8.0; 10% glycerol; 0.5 mM TCEP; 1 μg ml⁻¹ of leupeptin; 1 μg ml⁻¹ of pepstatin; 1 mM PMSF). GyrA and GyrB were purified separately, following an identical procedure. Cells were lysed by the addition of 1 mg ml⁻¹ of egg white lysozyme, followed by sonication. Cell lysate was then clarified by centrifugation, and the soluble lysate was applied to a 5-ml HisTrap-HP column (Cytiva). The column was washed with 200 ml of A1000, followed by elution with 30 ml of B1000 (30 mM Tris-HCl (pH 7.8); 1 M NaCl; 500 mM imidazole, pH 8.0; 10% glycerol; 0.5 mM TCEP; 1 μg ml⁻¹o f leupeptin; 1 μg ml⁻¹ of pepstatin; 1 mM PMSF). SUMO-tagged protein was then cleaved with SENP protease and dialyzed overnight against A500 (30 mM Tris-HCl (pH 7.8); 500 mM NaCl; 10 mM imidazole, pH 8.0; 10% glycerol; 0.5 mM TCEP; 1 μg ml⁻¹ of leupeptin; 1 μg ml⁻¹ of pepstatin; 1 mM PMSF). Cleaved protein was passed over a 5-ml HisTrap-HP column, and the flow-through was then collected and concentrated. GyrA and GyrB were each run separately over a Superdex 200 10/300 column (Cytiva) equilibrated in C500 (50 mM Tris-HCl (pH 7.8); 500 mM KCl; 10% glycerol; 0.5 mM TCEP). Peak fractions were collected and concentrated, and the final glycerol concentration increased to 30% before flash-freezing in liquid nitrogen for storage at −80 °C.

Supercoiling and supercoil relaxation assays.

Purified M. tuberculosis GyrA and GyrB were mixed on ice in an equimolar ratio to form gyrase heterotetramers at a concentration of 40 μM. MtbGyrase was then serially diluted in twofold steps using gyrase dilution buffer (50 mM Tris-HCl (pH 7.8); 150 mM monopotassium glutamate; 5 mM magnesium acetate; 10% glycerol) to 10× working concentrations for supercoiling assays. Supercoiling assays were assembled by mixing the following on ice: 2 μl of 10× relaxed pSG483 plasmid DNA (68.75 nM), 2 μl of 10× gyrase dilutions (3.12–200 nM), 7 μl of deionized water, 4 μl of 5× reaction buffer (120 mM Tris-HCl (pH 7.8); 38 mM magnesium acetate; 340 mM monopotassium glutamate; 36% glycerol; 0.4 mg ml⁻¹ of BSA; 4 mM TCEP) and 2 μl of deionized water or 10× evybactin (1 mM). Reactions were initiated by the addition of 2 μl of 10× ATP (20 mM) and then incubated at 37 °C for 30 minutes (final reaction volume, 20 μl) before quenching using 3 μl of reaction stop buffer (125 mM EDTA (pH 8.0); 5% SDS), followed by the addition of 2 μl of 3 mg ml⁻¹ proteinase K. Reactions were digested of protein by further incubation at 37 °C for 30 minutes. Loading dye (5 μl of 5× loading dye) was added to reactions, and products were resolved on a 1.5% native TAE agarose gel by running at 35 V for 16.5 hours. Gels were post-stained with ethidium bromide and visualized by UV transillumination.

ATPase assays.

ATPase assays were conducted using an NADH-coupled assay⁴⁶. Gyrase heterotetramer was formed as described above for the supercoil relaxation assays. Reactions were assembled in the following manner on ice: 5 μl of 10× gyrase (2.5 μM), 5 μl of BamHI linearized plasmid DNA (2.5 mg ml⁻¹), 5 μl of deionized water, 25 μl of 2× Buffer/NADH solution (100 mM Tris (pH 7.8); 170 mM monopotassium glutamate; 10% glycerol; 0.2 mg ml⁻¹ of BSA; 5 mM magnesium acetate; 7 mM phosphoenolpyruvate; 0.6 mM NADH; 10% pyruvate kinase/lactic dehydrogenase enzymes from rabbit muscle (Sigma-Aldrich)) and 5 μl of 10× evybactin or moxifloxacin (1 mM), final reaction volume 50 μl. Samples were incubated at 37 °C before the addition of 5 μl of 10× ATP (0.625–40 mM) to initiate the reactions. Absorbance changes at 340 nm were observed over 30 minutes, and absorbance change over time corresponding to ATP consumption was calculated using NADH standard curves. K_cat, K_M and V_max values were calculated by non-linear curve fitting in GraphPad Prism 8 using the Mechaelis–Menten equation.

Plasmid cleavage assays.

Plasmid cleavage assays were conducted similarly to the supercoiling and supercoil relaxation assays, with a few modifications. Gyrase stock concentrations for the cleavage assays were 200 nM (final concentration, 20 nM), and evybactin or moxifloxacin stock concentrations were 64 nM to 1 mM (final concentrations, 6.4 nM to 100 μM). Cleavage assays were resolved on a 1.5% TAE agarose gel containing 1 μg ml⁻¹ of ethidium bromide by running at 35 V for 16.5 hours. Gels were then post-stained with ethidium bromide and visualized by UV transillumination. ImageJ (version 1.52) was used for quantitation of cleaved products, and fraction plasmid cleaved was calculated by taking the cleaved band intensities and dividing by the sum of the cleaved band and supercoiled band intensities. IC₅₀ values were calculated by non-linear curve fitting using GraphPad Prism 8 using the following equation: Y = Min + X × (Max − Min) / (IC₅₀ + X).

MtbGyrBA_core(Y129F) purification and X-ray crystallography.

MtbGyrBA_core(Y129F) was purified as previously described³³. In brief, the protein was expressed from a modified pET vector containing an N-terminal hexahistadine SUMO tag, using BL21[DE3] CodonPlus E. coli cells (Agilent). Cells were grown at 37 °C to OD₆₀₀ of 0.4 in M9ZB media⁴⁷, after which the temperature was reduced to 18 °C, and protein production was induced with 0.25 mM IPTG for 18 hours before harvesting by centrifugation. Cells were resuspended in A1000 buffer, lysed, clarified and captured on a 5-ml HisTrap-HP column as described above. Captured protein was washed on the column with 50 ml of A100 (30 mM Tris-HCl (pH 7.8); 100 mM NaCl; 10 mM imidazole (pH 8.0); 10% glycerol; 0.5 mM TCEP; 1 μg ml⁻¹ of leupeptin; 1 μg ml⁻¹ of pepstatin; 1 mM PMSF) to reduce the salt before eluting directly onto a 5-ml HiTrapQ-HP column (Cytiva) using B100 (30 mM Tris-HCl (pH 7.8); 100 mM NaCl; 500 mM imidazole (pH 8.0); 10% glycerol; 0.5 mM TCEP; 1 μg ml⁻¹ of leupeptin; 1 μg ml⁻¹ of pepstatin; 1 mM PMSF). The HiTrapQ-HP column was washed with 5 column volumes of Q100 (30 mM Tris-HCl (pH 7.8); 100 mM NaCl; 10% glycerol; 0.5 mM TCEP; 1 μg ml⁻¹ of leupeptin; 1 μg ml⁻¹ of pepstatin; 1 mM PMSF), and a gradient of 0–100% Q1000 (30 mM Tris-HCl (pH 7.8); 1 M NaCl; 10% glycerol; 0.5 mM TCEP; 1 μg ml⁻¹ of leupeptin; 1 μg ml⁻¹ of pepstatin; 1 mM PMSF) over 10 column volumes was conducted to elute captured protein. Peak fractions were pooled, and the hexahistadine-SUMO tag was removed through overnight cleavage with SENP protease. Cleaved protein was then applied to a 5-ml HisTrap-HP column, and the flow-through was collected and concentrated. Subsequently, MtbGyrBA_core(Y129F) was then applied to a Superdex 200 16/60 column equilibrated in C500, after which peak fractions were pooled and concentrated, and the final glycerol concentration was increased to 30% before flash-freezing in liquid nitrogen for storage at −80 °C.

For co-crystallization, a DNA substrate was adapted from previous structural studies of S. aureus gyrase⁴⁸ and designed to contain a single nick that is offset 2 nucleotides (nt) from the center of the substrate as well the DNA ends joined by ‘GAA’ triloop linkers: 5′-GGCCCTACGGCTgaaAGCCGTAGGGCCCTACGGCTgaaAGCCGTAG-3′. The 2-nt offset positions the nick in one catalytic center of the enzyme to ensure a precise binding register with the protein. The oligo was ordered from Integrated DNA Technologies and annealed in 10 mM Tris (pH 7.8); 50 mM NaCl; 1 mM EDTA, using a thermocycler to generate the appropriate substrate for crystallography. MtbGyrBA_core(Y129F) and annealed DNA were mixed in a 1:1.7 protein:DNA ratio (150 μM MtbGyrBA_core(Y129F) dimer:255 μM DNA oligo). Protein:DNA complex was then dialyzed against 20 mM Tris-HCl (pH 7.8); 150 mM NaCl; 10 mM MgCl₂; 0.5 mM TCEP. Dialyzed protein:DNA complex was incubated with 1 mM evybactin at 37 °C for 3 hours before conducting crystallization trials. Long, rod-like crystals formed after several days in hanging drops containing 7–12% PEG10K; 100 mM MES (pH 6.0); 200 mM magnesium acetate. Crystals were cryopreserved in 12% PEG10K; 100 mM MES (pH 6.0); 200 mM magnesium acetate; 1 mM TCEP; 1 mM evybactin; 25% glycerol. Diffraction data were collected at NSLS-II beamline 17-ID-1 (AMX) and initially autoprocessed using Fast DP (version 2017_10_02)⁴⁹. Further data processing and data reduction were carried out using XDS (version Mar 15, 2019) and CCP4 (version 7.1). Molecular replacement was conducted using Phaser⁵⁰ and a single monomer subunit of a prior MtbGyrBA_core model (Protein Data Bank (PDB) ID: 5BTA), stripped of ligands, DNA and waters. Coot⁵¹ was used for model building and ligand placement, and refinement was conducted in Phenix (version 1.18.2–3874). Figures were generated using Pymol (version 2.4, Schrödinger)

Animal study.

Animal study was performed at Northeastern University, approved by Northeastern Institutional Animal Care and Use Committee, and was performed according to institutional animal care policies. Experiments were not randomized or blinded, as it was not deemed necessary. Female CD-1 mice (20–25 g, experimentally naive, 6 weeks old) from Charles River Laboratories were used for all studies. Evybactin was tested in a septicemia model against E. coli ATCC 25922. Mice were infected with 0.5 ml of E. coli ATCC 25922 suspension in BHI with 5% mucin (1 × 10⁶ to 5 × 10⁶) via intraperitoneal injection. This dose achieves >83% mortality within 24 hours after infection. At 1 hour after infection, mice were treated with evybactin from 10 mg kg⁻¹ to 100 mg kg⁻¹ by intraperitoneal injection. Infection control mice were treated with 50 mg kg⁻¹ of gentamicin as a positive control. Survival was monitored for 120 hours.

Extended Data

Extended Data Fig. 1 |. NMr structural determination of evybactin in DMSo-d₆.

a, ¹H. b, ¹³C. c, COSY. d, rOeSY. e, ¹H-¹³C HSQC. f, ¹H-¹³C HMBC.

Extended Data Fig. 2 |. 2D NMr key correlations for evybactin structural assignment.

All correlations were measured in DMSO-d₆ except for the HMBC correlation from H-41 to C-2 was recorded in D₂O.

Extended Data Fig. 3 |. Efficacy of evybactin in an animal model.

Mice were infected by E. coli ATCC 25922 through intraperitoneal injection, and antibiotics were administrated 1 h later. Survival was monitored over 5 days. The experiment was repeated three times (n=4 biologically independent mice); lines are the mean of experiments. Gentamicin (Gen) was used as a positive control. All treatment are in mg kg⁻¹.

Extended Data Fig. 4 |. BacA homologs are distributed among bacteria.

BacA phylogenic tree was generated by using the Maximum Likelihood method based on the JTT matrix-based model⁵². The tree with the highest log likelihood (−8349.22) is shown. Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model, and then selecting the topology with superior log likelihood value. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. The analysis involved 17 amino acid sequences. All positions containing gaps and missing data were eliminated. There were a total of 274 positions in the final dataset. evolutionary analyses were conducted in MeGA7⁵³. Protein sequences were obtained from a previous study⁵⁴.

Extended Data Fig. 5 |. Evybactin inhibits DNA synthesis.

Effect of evybactin on macromolecular biosyntheses in E. coli WO153. Incorporation of ¹⁴C-thymidine (DNA), ¹⁴C- uridine (rNA), ¹⁴C -L-amino acid mixture (protein), ¹⁴C-Acetic acid (fatty acid) and ¹⁴C-acetyl-glucosamine (peptidoglycan) was determined in cells treated with 8x MIC of evybactin (grey bars). Ciprofloxacin (8x MIC), rifampicin (8x MIC), chloramphenicol (8x MIC), triclosan (8xMIC) and fosfomycin (8x MIC) were used as controls (white bars). Values are plotted as mean values ±SD, n=3 independent biochemical experiments.

Extended Data Fig. 6 |. Comparison of evybactin and thiophene binding.

a, Electron density omit maps for evybactin contoured at 1σ. Gyrase is depicted as a blue cartoon and evybactin as magenta sticks. b, Comparison of the evybactin-binding pocket (top panels) with the thiophene-binding pocket (bottom panels – PDB ID: 5NPK¹⁴). Gyrase subunits are colored in dark blue (GyrA) and light blue (GyrB) with evybactin and the thiophene colored magenta and green, respectively. Hydrophobic residues forming the shared evybactin and thiophene binding pocket are labeled (left panels). A glutamate residue in GyrB that is critical for thiophene binding to S. aureus gyrase (e634) is a threonine (T664) in M. tuberculosis gyrase (right panels). c, electron density omit maps (1σ) for the evybactin-bound M. tuberculosis gyrase structure (left) and the thiophene-bound S. aureus gyrase structure (right, PDBID: 5NPK). Gyrase is colored in green and DNA is orange.

Extended Data Fig. 7 |. Mutations at the evybactin binding site effect evybactin and moxifloxacin induced cleavage.

a, Plots represent quantitation of evybactin- and moxifloxacin-induced cleavage in the presence of ATP. Fraction of linearized plasmid plotted at indicated concentrations of compound (0–500 μM). Cleavage was conducted with 20 nM wild-type MtbGyrase or MtbGyrase GyrA mutants and 6 nM DNA. Lines represent non-linear fits to the data, as in Fig. 4, values are plotted as mean values ±SD, n=3 independent biochemical experiments. b, representative cleavage assays used for quantitation. Samples were separated on agarose gels run in the presence of ethidium bromide. The positions of nicked, linear, and uncleaved plasmid are indicated.

Extended Data Fig. 8 |. Evybactin and moxifloxacin stimulated cleavage activity of M. tuberculosis gyrase mutants.

Native agarose gel-based cleavage assay conducted with 125 nM gyrase, 6 nM plasmid DNA, and indicated amounts of evybactin or moxifloxacin (0–20 μM). Mutations in GyrA or GyrB are indicated above each panel. The positions of nicked, linear, and supercoiled plasmid DNA are indicated. Note that high protein concentrations are used to see activity in the resistance mutants; as a result, the DNA in the WT MtbGyrase reactions becomes degraded and disappears as evybactin concentrations are increased due to the presence of multiple, randomly spaced cleavage complexes. All assays repeated at least 2 times with similar results.

Extended Data Fig. 9 |. Supercoiling activities of M. tuberculosis gyrase mutants.

Native agarose gel analysis of supercoiling activity using indicated amounts of M. tuberculosis gyrase and resistance mutants (0–20 nM). The migration positions of relaxed starting material and supercoiled products are indicated. All assays repeated at least 3 times with similar results.

Extended Data Fig. 10 |. The evybactin binding pocket is concealed in the M. tuberculosis gyrase ‘AtPase open’ state.

Structure of S. cerevisiae TOP2 bound to DNA and nonhydrolyzable ATP analog, illustrating the ‘ATPase closed’ conformation of type-II topoisomerases (left, PDBID: 4GFH⁵⁵). The ATPase and transducer domains are colored yellow and light green, the nucleolytic core is illustrated in orange and blue, and DNA is shown in grey. Structure of M. tuberculosis gyrase in an ‘ATPase open’ state (right, PDBID: 6GAV³⁸), in which the ATPase regions are folded down from the position shown at left. The binding site for evybactin is illustrated as a black outline and shown as a purple surface in the inset. The inset shows the loop within the GyrB ATPase domain that is specific to Corynebacteriales gyrases and how this loop occludes the evybactin binding site in the ‘ATPase open’ conformation of the enzyme.

Data availability

M. tuberculosis genome data (NC_000962.3) were used as a reference of genome sequencing. All data supporting the findings of this study are available within the paper and its Supplementary Information or have been deposited to the indicated databases. The genome of P. noenieputensis DSM 25462 has been deposited to GenBank with accession number RCWC00000000.1. Structural data have been deposited in the PDB with PDB identifier 7UGW. All other data are available from the corresponding author upon reasonable request. Source data are provided with this paper.