YgiV promoter mutations cause resistance to cystobactamids and reduced virulence factor expression in Escherichia coli

Antimicrobial resistance (AMR) is a major threat for modern medicine, influencing not only primary therapeutic intervention but also essential treatments such as surgery, cancer chemotherapy and prevention of secondary infections. This silent pandemic, mainly driven by misuse of existing antimicrobials and lack of novel antibiotics, causes prolonged hospitalization, higher treatment costs and increased mortality¹. Recent studies show that AMR was associated with 4.95 million deaths in 2019, with increasing numbers potentially reaching 10 million deaths in 2050². Multidrug-resistant (MDR) Escherichia coli was found to be the leading cause of deadly infections associated with or attributed to antimicrobial resistance in 2019³. This pathogen shows high adaptability to a broad spectrum of known antimicrobial classes such as beta-lactams, quinolone antibiotics, sulfonamides and many more. The spread of AMR results in decreased treatment options, prolonged treatment time leading to treatment complications and finally increased mortality^4,5.

Cystobactamids (CYSs) have a high potential to be developed as a new antibiotic class due to their novel structural scaffold and mode of action. This natural compound class was first isolated from Cystobacter spp. and their biosynthetic pathway, as well as their total synthetic route, have been investigated and optimized^6,7. Natural CYS derivatives incorporate multiple para-aminobenzoic acid (PABA) units and an N-terminal para-nitrobenzoic acid linked through a central amino acid, similar to the structurally related natural compounds albicidin, which carries in contrast to CYSs an N-terminal para-methylcoumaric acid, and coralmycin^8,9. A number of structure-activity guided chemical optimization studies led to the discovery of new derivatives with favorable properties in terms of antibacterial spectrum coverage and potency¹⁰. CYSs show broad activity against multi-drug resistant (MDR) Gram-negative and Gram-positive pathogens such as Acinetobacter baumannii, E. coli, Staphylococcus aureus, and Enterococcus spp. mediated by inhibition of the bacterial gyrase and topoisomerase IV. This potent activity coupled with the lack of cytotoxic effects on eukaryotic cells, render CYSs as highly valuable for the development of a potential novel antibiotic class.

During the development of a novel antimicrobial compound, it is of utmost importance to understand the mechanism(s) of resistance in target pathogens. In contrast to the structurally related albicidins, where the tsx-encoded outer membrane channel mutation was found as a major cause for stable resistance in E. coli, the resistance mechanism for CYSs remained unknown¹¹. In this study, the CYS mode of resistance in E. coli as the target pathogen was investigated. To this end we generated resistant mutants and subsequently characterized the mutants using genomic and proteomic analyses. Previous studies already suggested that, in contrast to fluoroquinolones, CYS resistance (CYS^R) in E. coli is barely affected by efflux through the AcrAB-TolC pump or by typical target mutations in the quinolone resistance determining region (QRDR) of gyrase and topoisomerase IV^12,13. Here, we could confirm this finding and identified the YgiV protein as major cause of high-level resistance. YgiV is a soluble 17.8 kDa protein consisting of 160 amino acids, whose biological function is described as a repressor of the mcbR biofilm gene¹⁴. Furthermore, it incorporates a DNA binding domain and shows homology to GyrI-like proteins¹⁵. GyrI-like proteins are known for their protective effect on bacterial cells via numerous mechanisms, including gyrase protection or binding as well as modification of harmful xenobiotics¹⁶. With further characterization we have been able to conclude CYS neutralization is due to competitive binding to YgiV rather than an efflux-mediated mode of resistance. As opposed to recently published literature on albicidin resistance, YgiV overexpression in CYS^R is caused by point mutations in its promotor region, which in turn also affect bacterial virulence factor expression¹⁷.

Resistance development leads to stable high-level resistance against different CYS derivatives

Among many synthetic and natural CYS derivatives, the early-stage CYS derivative CN-861-2 [1] was used as the major reference compound in this study, because of its typical CYS structural composition, which is closely related to the natural products and its well-studied biological behavior^10,12. CNDM-861 [2] and Cysto-180 [3] (synthesis see Supplementary Figs. 1 to 10) were further chosen as control CYSs. The early stage derivative CNDM-861 [2] is closely related to CN-861-2 [1], missing a single methoxy-group in the linker region. Cys-180 [3] is a later stage CYS derivative with a highly modified B-ring and linker region, leading to superior physicochemical properties (Fig. 1). The well characterized E. coli K12 (BW25113) strain was used for generating CYS^R mutants because of its outstanding annotation and comparability with single gene knockout (KO) strains from the Keio library¹⁸. Single bacterial colonies were selected after overnight incubation on CYS CN-861-2-[1]-containing solid medium, with a moderate to low frequency of resistance (FoR) of 2.6 × 10⁻⁸ at 4-fold MIC and no resistance development at 8-, 16- and 32-fold MIC (FoR <10⁻⁹) (Supplementary Table 1). Eight individual mutant clones were selected and tested for susceptibility to three CYS derivatives, as well as to the clinically approved topoisomerase poisons ciprofloxacin (CIP) and levofloxacin (LEV). All isolated mutants showed high-level resistance against tested CYS derivatives with 20- to ≥ 640-fold shifts in the minimum inhibitory concentration (MIC). No shift in MIC was observed for these mutants when tested against quinolone antibiotics (Table 1), however, high-level cross-resistance to the structurally related albicidin [4] was detected (Supplementary Table 2). This indicated that the occurring resistance mechanism is highly specific for CYS and structurally related compounds, and does not rely on the specific linker or ring composition, nor the physicochemical properties of the derivatives. This specificity is supported by the lack of cross-resistance to the quinolone topoisomerase poisons, sharing the same targets. When MICs were re-evaluated after ten days of serial passaging without selective pressure, reversibility of resistance was not observed, suggesting a stable mutant genotype. To study whether the underlying resistance mechanism is accompanied by a fitness cost, we analyzed the in vitro growth and metabolic heat profiles of CYS^R mutants. There were no considerable differences between the mutants and the wild-type (WT) (Supplementary Fig. 11), indicating that there is no evolutionary disadvantage such as metabolic fitness loss resulting from the resistance-causing mutations, and explaining the stability of the occurring resistance.

Structural differences to cystobactamid CN-861-2 [1] (reference in this study) are marked in red.

Full size image

Genome analysis revealed SNPs in the potential promotor region of the ygiV gene

Whole-genome sequencing of CYS^R E. coli enabled the identification of two major mutation sites in close proximity to each other that are associated to resistance (Fig. 2). All eight mutants showed SNPs upstream of the ygiV gene. Mutants M3 and M6, showing the highest MIC shifts, differed in SNP location from the other CYS^R mutants. This is contrary to a recently published report on the related albicidin [4], where gene amplification events in the genetic locus of ygiV were identified to cause resistance¹⁷. Gene amplifications are intrinsically unstable due to higher fitness costs¹⁹, explaining the difference in stability of albicidin [4] and CYS resistance in E. coli, respectively. The previously investigated albicidin-[4]-resistant mutants lost their higher copy number of the respective genetic area in the absence of selective pressure¹⁷, whereas CYS^R mutants carrying SNPs maintained high-level resistance under similar experimental conditions. When testing the Keio ygiV KO strain, we did not detect a shift in MIC compared to the parent strain (Supplementary Table 3). However, when trying to evaluate the FoR for this strain, it was not possible to generate resistant mutants and a FoR of <10⁻¹⁰ was determined (Supplementary Table 4). Moreover, our attempts to induce resistance in the ygiV KO strain through prolonged antibiotic exposure over 5 days, utilizing parallel serial passaging in low to high CYS concentrations (0.25-, 0.5-, 1.0-, 2-, 4-fold MIC) remained unsuccessful, indicating an essential role of this gene for CYS resistance in E. coli which cannot be compensated by other resistance mechanisms such as target mutation.

Single nucleotide polymorphisms were found in the promotor region of ygiV of all cystobactamid-resistant mutants, with some carrying additional mutations in qseC (cp. Table 2) (created with BioRender.com).

Full size image

Furthermore, six out of eight CYS^R E. coli showed mutations in qseC, encoding for the sensor histidine kinase of the two-component system (TCS) QseBC²⁰, leading to a change in the amino acid sequence, a frameshift or an insertion of a premature stop codon (Table 2). These mutations were located in the direct genetic neighborhood to ygiV, probably mutually influencing their transcription due to a shared promotor site²⁰. However, the level of resistance of individual mutants did not correlate with the presence or absence of mutations in the qseC gene.

Full-proteome analysis revealed extensive overexpression of YgiV in CYS^R mutants compared to the WT

Three CYS^R mutants M3, M6 and M8 with varying mutation sites (Table 2) and resistance phenotype (Table 1) were selected to study their differential proteome expression. All selected mutants carry mutations in the promotor region of ygiV, with M8 representing the most common genotype, and additionally carrying qseC mutations. The analyzed mutants showed a high consensus of significantly upregulated proteins (p-value [-log₁₀] > 1.3, difference [log₂] > 1.5). Eighteen out of 32 proteins were overexpressed in all analyzed mutants, and two functional clusters were identified (Fig. 3a, b). One cluster represented phage-shock proteins (Psp), which are related to survival under nutrient- or energy-limited conditions and known for stabilizing the cell membrane^21,22. The second cluster comprised upregulated proteins known to play a role in resistance against cationic antimicrobial peptides (AMPs) by affecting bacterial membranes. This includes the transcriptional regulator protein QseB, the undecaprenyl-phosphate 4-deoxy-4-formamido-L-arabinose transferase ArnC, the phosphoethanolamine transferase EptC, the bifunctional polymyxin resistance protein ArnA, the transcriptional regulatory protein BasR and the lipopolysaccharide core heptose(II)-phosphate phosphatase Ais (Fig. 3b)^23,24,25. This upregulation of proteins, potentially causing cationic peptide resistance due to positively charged membrane modifications, can be explained as a downstream effect of detected QseB overexpression, which is known to influence the expression of ArnA, ArnC and EptC²³. This overexpression of QseB in CYS^R mutant bacteria was very likely caused by the identified mutations (Table 2), suggesting a shared transcriptional regulatory region between ygiVW and qseBC²⁰. However, no cross-resistance to colistin and polymyxin B was identified, which led us to conclude that, although being significantly upregulated, these membrane modifying proteins only play a minor role in contributing to the CYS resistance phenotype of E. coli, further highlighting a very specific mode of resistance that occurs also independent of qseC mutation (Supplementary Table 2). In case of the most resistant CYS^R M6, YgiV was the most highly overexpressed protein compared to the WT proteome (Fig. 3c). Additionally, relative YgiV expression levels differ significantly between mutants with different SNP positions (p-value < 0.0001, two-tailed student´s t-test, n = 4) and correlated with observed MIC shifts for tested CYS^R strains (M6 > M3 > M8) (Fig. 3d), indicating a crucial role of the protein in CYS^R. YgiV expression levels in WT bacteria were below the limit of detection (LoD), hence its abundance for relative quantification was based on imputed values. This suggests that found SNPs, occurring in the promotor region of the ygiV gene, potentially cause a change in affinity for a yet unknown transcriptional regulator, resulting in upregulation of the protein. This hypothesis is supported by the differences we identified in overexpression and the degree of CYS^R, seen for mutants with different SNP positions in the ygiV promotor region. In addition, we analogously generated resistant mutants from a set of three clinically relevant E. coli strains and performed a full proteome analysis of these mutants with the highest level of CYS resistance in order to determine the relevance of YgiV overexpression in clinical isolates²⁶. Firstly, multiple MDR clinical isolates were tested for their susceptibility to CYS, showing sensitivity with no pre-existing resistances (Supplementary Tables 5 and 6). Secondly, resistance to CYS was generated and the differential proteome expressions of the most resistant CYS^R mutants was analyzed, revealing highly consistent differential proteome profiles to those generated from BW25113, including YgiV overexpression (Supplementary Fig. 12).

a Comparison of upregulated proteins in mutants M3, M6 and M8 showed an intersection of 18 out of 32 proteins. (created using BioVenn). b Two main functional clusters were identified, including phage-shock proteins and membrane modifying proteins (created using STRING database). c Volcano plot shows that YgiV was the most upregulated protein in highly resistant strain M6 and major downregulated proteins in all tested mutants were found to be OmpF and FimA. d YgiV expression levels correlated to the level of resistance (M6 > M3 > M8). Data shown are mean ± SD of n = 4 (**** p-value < 0.0001). Significant up-/downregulation was analyzed by two-tailed student’s t-test with p-value [–log₁₀] > 1.3 and abundance difference [log₂] > 1.5 as cut-offs (n = 4).

Full size image

Intriguingly, we found two proteins, namely the porin OmpF and the type-1 fimbrial protein A (FimA), to be substantially underexpressed in all three tested mutants (Fig. 3c). In order to study the possible involvement of OmpF in cell entry of CYS we tested an ompF knockout strain for susceptibility to the antibiotics. However, no MIC shift was observed compared to the WT (Supplementary Table 3). The FimA protein was abundant in all WT samples but was below or close to the LoD in mutant samples, showing a strongly reduced FimA expression level for CYS^R mutants. FimA is a key building protein of fimbriae or pili, which are responsible for cell adhesion and involved in pathogenicity. The influence of the observed FimA downregulation in CYS^R mutant proteomes, might result in reduced pathogenicity due to loss of ability to colonize epithelial cells of specific organs²⁷. As an additional virulence marker, we tested the bacterial cell motilities of CYS^R mutants and confirmed highly reduced swarming behavior of resistant mutants (Supplementary Fig. 13). The flagellin protein, the primary structural component of bacterial flagella responsible for motility, could not be quantified and compared in E. coli BL21 WT and CYS^R mutant strains, as its abundances were below the LoD. However, in all three tested CYS^R clinical isolates, flagellin expression was significantly reduced compared to their respective WT strain, which may account for the observed loss of motility (Supplementary Fig. 12). It is known that qseC defects and qseB upregulation cause reduction of flagellar motility and fimbrial hemagglutination^20,28. This observation supports the notion that development of resistance to CYSs, caused by SNPs located upstream of the ygiV gene, contribute to the upregulation of QseB due to formation of open promotor complexes for both gene loci²⁰, which, in turn, influences FimA expression and cell motility²⁸. This molecular cascade very likely results in a reduced expression of these specific virulence factors in E. coli. In the context of infection, reduction in FimA expression might result in less effective epithelial attachment of the pathogen. Accompanied by the observed decrease in cell motility and migration, this reduction could enhance the host´s ability to clear the pathogen. These expression changes of FimA and flagellin might be of particular advantage for the infected host in the context of urinary tract infections, where epithelial adhesion and bacterial motility are critical determinants of pathogenicity²⁹. However, this needs to be confirmed in vivo before firm conclusions can be drawn.

Overexpression of YgiV caused high-level CYS^R

The ygiV sequence was isolated from E. coli WT genomic DNA and cloned into a vector for protein overexpression via an inducible T7 promotor (pET-28b) (Supplementary Fig. 12). This inducible vector was transformed into E. coli Lemo21 (DE3) suitable for protein expression of a wide range of proteins. Following induction of YgiV overexpression with IPTG (isopropyl-β-D-thiogalactopyranoside), the MIC of CYSs increased 1600- to 6,400-fold, compared to the strain containing the empty vector (Table 3). These susceptibility shifts confirm that overexpression of YgiV in CYS^R mutants is the leading cause for their high-level resistance.

The biological function of YgiV is barely described in literature but is known as a transcriptional repressor of mcbR¹⁴. McbR is known to bind in the promotor region of mcbA and thereby inhibits its transcription, which prevents overproduction of colonic acid and mucoidy¹⁴. To test if this repression of mcbR causes CYS^R, the respective Keio knockout strain was tested and no shift in CYS antimicrobial activity was observed compared to the WT (Supplementary Table 3), excluding this pathway as a major mode of resistance.

CYS inhibitory activity on gyrase was lost in the presence of YgiV

To further characterize the resistance determinant we purified YgiV after expression in E. coli Lemo21 (DE3) by affinity and size exclusion chromatography (Supplementary Fig. 15). On-target activity of CYS in a gel-based assay in the presence and absence of YgiV showed a clear YgiV-concentration dependent shift in the IC₅₀ (half maximal inhibitory concentration) in the presence of 0.6 µM YgiV and completely abolished activity with 20 µM YgiV (Fig. 4a). When YgiV is present in a similar molarity as CYS, the inhibitory effect of CYS was neutralized. To rule out that YgiV degrades CYSs, we incubated YgiV and CN-861-2 [1] for 1 h and measured compound concentration. No reduction in the CYS amount was detected, excluding compound degradation as the YgiV-mediated mode of resistance (Supplementary Fig. 16). Thus, we view it as likely that CYS binding by YgiV prevents the antibiotics from interacting with its target, resulting in inactivity towards the gyrase. Native protein mass spectrometry (native MS) revealed seemingly stoichiometric binding of YgiV to all tested CYSs (Fig. 4b, Supplementary Fig. 17) after equimolar incubation. Subsequently, by using microscale thermophoresis (MST), a two-digit nanomolar K_D (36.41 ± 16.86 nM) (equilibrium dissociation constant) was determined (Fig. 4c, Supplementary Fig. 18 and 19), identifying the high-affinity binding of CYS to YgiV as a major cause for neutralization. This high-affinity binding is in line, with the previously reported binding of YgiV to the structurally related albicidin [4]¹⁷. In addition, due to multiple m/z shifts in the native protein MS with the mass of tested CYS derivatives, multiple binding sites per YgiV molecule seem to be possible. The exact binding mode will be further investigated by co-structure elucidation using cryo-EM and/or X-ray crystallography.

a Gyrase supercoiling assay showed neutralization of CN-861-2 [1] inhibitory activity (mean ± SD of n = 3) by addition of YgiV. b Native MS measurements showed binding of CN-861-2 [1] to YgiV with at least one but potentially multiple binding sites. c Microscale thermophoresis revealed low nanomolar binding affinity (K_D = 36.41 ± 16.86 nM, mean ± SD of n = 3) of CN-861-2 [1] to YgiV. d YgiV induced linearization of plasmid DNA in a concentration dependent manner. Intensity after background subtraction of the linear plasmid band at 20 µM YgiV was set as 100%. Negative control was set as 0%. (n = 3).

Full size image

Intriguingly, addition of 20 µM YgiV to the gyrase supercoiling assay seemed to reduce the conversion of relaxed to supercoiled plasmid. However, without gyrase enzyme (negative controls in Supplementary Fig. 20) it became apparent that YgiV has an influence on the plasmid DNA by converting/cleaving the relaxed plasmid mainly to its linear form. Consequently, in the presence of YgiV, supercoiling activity of gyrase was generally less pronounced. Importantly, cystobactamids inhibit gyrase activity with an IC₅₀ of 0.448 µM. In the presence of YgiV, where we observed less efficient gyrase supercoiling activity, much higher concentrations of cystobactamid were needed to block this reaction (Supplementary Fig. 20). These findings led us to conclude that YgiV might play a dual role in the resistance mechanism of cystobactamids by on the one hand neutralizing the antibiotic through high-affinity binding and on the other hand interfering with the gyrase reactions. To further assess potential DNA cleavage properties of YgiV, we titrated the protein to relaxed and supercoiled plasmid DNA. The addition of YgiV resulted in a concentration dependent increase of OC/nicked and linearized plasmid DNA (Fig. 4d and Supplementary Fig. 21), thus confirming direct single- and double-stranded DNA-cleavage activity of YgiV. Furthermore, YgiV shows sequential and structural similarity to GyrI-like proteins such as SbmC, known for its protective effect against the topoisomerase poison microcin B17. Microcin B17 shows a similar molecular mechanism on DNA gyrase as CYSs and albicidin [4]. These compound classes require DNA strand passaging for binding to the DNA gyrase, a mechanism that differs from quinolone antibiotics^30,31. It cannot be excluded that the YgiV-DNA interaction contributes to its properties in mediating CYS resistance. The hypothesis is supported by the finding that full length YgiV conveyed higher resistance against the structurally related albicidin [4] than constructs lacking the DNA binding domain¹⁷. The underlying molecular mechanism needs to be investigated in future studies.

YgiV expression in mutants was inducible via treatment with CYS

To investigate how mutants react and adapt to CYS treatment, their differential proteome expression was analyzed under CYS stress. Full-proteome comparison of CYS^R and WT bacteria revealed a dependency of the YgiV expression to sub-MIC CYS treatment. When treating CYS^R mutants with the CYS derivative CN-861-2 [1] at 0.25-fold MIC, YgiV expression was induced even further in comparison to the already elevated expression in the mutant without CYS treatment (Fig. 5a, b). The elevated expression rates still correlated with the determined MIC shifts seen for tested CYS^R mutants (M6 > M3 > M8) (Fig. 5b) and hint towards an adaptation of mutant bacteria under sub-MIC treatment, which probably allows the mutants to resist higher CYS concentrations in comparison to their basal YgiV expression level. In addition, the trimethylamine-N-oxide reductases TorA and TorC, which are coupled to energy yielding reactions, as well as the acyl-coenzyme A dehydrogenase FadE, catalyzing the first step in fatty acid beta-oxidation, were found to be induced by treatment. Furthermore, various proteins involved in the SOS response such as DinD and RecA, and the DNA gyrase inhibitor protein SbmC³², were overexpressed in treated WT and mutant bacteria (Fig. 5a, c). The remaining SOS response in mutant bacteria demonstrates that CYSs still reach their antimicrobial targets, DNA gyrase and topoisomerase IV. In order to test if DNA damage and SOS response cause induction of YgiV, we treated M6, the mutant with the highest YgiV overexpression, with ciprofloxacin (CIP) as a representative topoisomerase poison. We could not detect any additional upregulation of YgiV caused by the resulting SOS response. In fact, the relative YgiV abundance in CYS^R M6 was slightly reduced following treatment with CIP (Fig. 5d). Therefore, it seems that the mechanism of ygiV induction is specific for the interaction with CYSs and potentially related compound classes. Due to the known repressor characteristics of YgiV itself¹⁴, it might be possible that binding of CYS to YgiV (Fig. 4c, d) directly influences its own transcription in the form of abolishing a potential negative transcriptional feedback loop. However, the detailed understanding of this mechanism is outside the scope of this paper and will be further investigated.

a Volcano plot of differerencial protein abundance observed for CYS^R M6 treated vs. untreated, and c string protein network analysis of overexpressed proteins after CYS treatment (created using STRING database), revealed mainly DNA repair proteins as part of SOS response commonly upregulated. b YgiV overexpression in cystobactamid-resistant mutants was induced via cystobactamid treatment, whereas only minor effects were observed in WT cells (*p-value = 0.0344, ****p-value < 0.0001). d YgiV upregulation could not be induced via CIP treatment (* p-value = 0.0142). Data shown are mean ± SD of n = 4. Significant up-/downregulation was analyzed by two-tailed unpaired student’s t-test (cut-offs: p-value [–log₁₀] > 1.3 and abundance difference [log₂] > 1.5) (n = 4).

Full size image

Affinity-based protein profiling confirmed binding of CYSs to YgiV in the whole cell environment and revealed additional binding partners

The CYS photo-affinity probe Cysto-33 [5] (synthesis see Supplementary Figs. 22 to 26) was designed and used to check whether binding to YgiV plays a role in the complex environment of the whole cell. The incorporation of a diazirine group for photo-reactive coupling and the alkyne moiety for click-chemistry in the CYS photo probe [5] relied on current structure-activity relationship (SAR) studies of CYSs¹⁰. These functional groups were used during the affinity-based protein profiling (AfBPP) to couple the CYS derivative covalently to binding proteins after UV-irradiation with subsequent linking to biotin via click-chemistry. This complex of CYS binding proteins linked to biotin was used for enrichment on avidin beads, which were relatively quantified against the untreated vehicle control after tryptic digest using mass spectrometry (Fig. 6a). Using this approach we aimed at identifying potential additional target proteins and binding partners of CYSs. The probe was first tested for bioactivity on WT and CYS^R mutants (Supplementary Table 2) as well as for inhibition of gyrase supercoiling activity (Supplementary Fig. 27), thus showing activity in the same range as other CYSs. Affinity-based protein profiling using CYS^R E. coli mutant M6 showed significant enrichment (p-value [-log₁₀] > 1.3, difference [log₂] > 1.0) of YgiV, thereby confirming binding of CYS to the upregulated protein in the whole-cell environment (Fig. 6b). Co-incubation with CN-861-2 [1] during AfBPP was used to check for competition for the same binding site with Cysto-33 [5]. Thereby, YgiV enrichment was not competed by addition of sub-MIC CN-861-2 [1], indicating a high extent of YgiV CYS binding capacity in M6. Additionally, target proteins (gyrase subunit B, ParE (topoisomerase IV subunit B)) were confirmed by enrichment and competition as direct interaction partners of CYSs, confirming the remaining binding to these proteins by sub-MIC CYS treatment, which ultimately explains the identified SOS response in mutants (Fig. 4a, c). Moreover, the efflux pump protein AcrB, the probable protease SohB and the plasma membrane protein YdgA were identified as potential additional binding partners of CYS via competition assays (Fig. 6c). Knockout (KO) strains of these proteins were tested to check if CYS binding to these proteins influences its biological activity. For ΔsohB and ΔydgA, no activity shift was found. However, the KO-strain of the efflux pump AcrB, which is part of the AcrAB-TolC tripartite efflux pump, showed a slight increase in activity (4x). Using a KO-strain of the negative regulator of the multidrug efflux pump AcrAB-TolC (ΔmarR)³³, a fourfold decrease in activity compared to the WT could be shown. Interestingly, only the ΔacrB but not the tested ΔtolC strain showed a slight increase of CYS susceptibility (Supplementary Table 3). These findings are in line with previously published data where it could be shown that natural CYS derivatives are effluxed through E. coli AcrAB-TolC, whereas this liability could be overcome with some natural derivatives that inspired the total synthesis of advanced CYS compounds overcoming this resistance mechanism¹³.

a AfBPP cross-linked the photo-affinity probe Cysto-33 [5] to target proteins after UV-irradiation with following click-reaction to biotin used for avidin-bead enrichment of bound proteins. After tryptic digest, the samples were measured using nanoElute-timsTOF (Bruker) and analyzed using DIA-NN and Perseus (created with BioRender.com). b YgiV was significantly enriched (p-value [-log₁₀] > 1.3, difference [log₂] > 1.0), thereby confirming binding to the cystobactamid probe⁵ inside the whole cell environment (*** p-value = 0.0001, two-tailed unpaired t-test, mean ± SD of n = 4). c Known target proteins, gyrase and topoisomerase IV, could be identified, as well as AcrB, SohB and YdgA as potential cystobactamid binding proteins. However, of these only the efflux pump protein AcrB seemed to have an effect on cystobactamid´s bioactivity (Supplementary Table 3).

Full size image

However, the observed activity shifts by KO or overexpression of AcrB through ΔmarR disruption are rather low, compared to the large shift in MIC in the presence of elevated YgiV expression. Additionally, overexpression of AcrB or other components of the tripartite efflux pump was not observed in tested mutants. It cannot be excluded that efflux may play a role in (low-level) resistance to CYSs in clinical application against E. coli, since the AcrAB-TolC efflux pump is common³⁴. However, overexpression of YgiV is very likely the major cause of high-level resistance in E. coli. Furthermore, high-level resistance development in a YgiV homolog expressing pathogens like Salmonella, Vibrio and Pseudomonas is probable¹⁷. In the event that homologous modes of resistance emerge in related pathogens, it would be of scientific and clinical interest to investigate whether a demonstrated reduced expression of virulence factors associated with CYS^R, could effectively reduce the pathogenicity of these bacterial isolates in vivo.

Cystobactamid resistance in E. coli is mediated through mutations in the promotor region of ygiV, leading to significant overexpression of the protein. Mutants further adapt against cystobactamids during treatment via substantial amplification of YgiV overexpression. The efflux pump AcrB was identified as an off-target protein but only influences cystobactamid bioactivity slightly. Furthermore, we observed that cystobactamid-resistant mutants show a reduction of fimbrial protein expression and cell motility, associated with cystobactamid induced ygiV promotor mutations.

Bacterial strains and culture conditions

The parent strain E. coli K12 (BW25113) WT was used for mutant generation. KO mutants ΔsohB, ΔydgA, ΔacrB and ΔompF were provided from the in-house E. coli K12 Keio collection¹⁸. E. coli AB100 parent and E. coli AB100 ΔmarR were provided by Prof. Peter Heisig (University of Hamburg). Clinically isolated E. coli strains were provided by PD Dr. Stefano Mancini (University of Zürich). Routine culture was performed in lysogeny broth (LB-Lennox) media under ambient conditions (37 °C, 180 rpm). MALDI-biotyping (Ultraflex III, Bruker) was regularly used to ensure quality of used strains.

Frequency of resistance (FoR) and mutant generation

For FoR and mutant generation, an overnight culture (ONC) of E. coli K12 was inoculated and grown at 37 °C, 180 rpm. The next day, bacteria were centrifuged, washed and suspended in PBS with an OD₆₀₀ of 12 (5.15 × 10⁹ CFU/mL). 100 µL of this bacteria suspension was used to inoculate selective LB-agar plates containing CN-861-2 [1] (4x, 8x, 16x and 32x MIC) and cultured overnight (37 °C) to determine the FoR. Bacterial dilution series and unselective agar plates were used to determine the exact bacterial load. Single colonies were picked and isolated in CN-861-2 [1] (10× WT MIC) containing LB-media and plated again. Cryo-stocks were made by mixing bacteria with 1:1 LB-media and 50% glycerol.

Minimal inhibitory concentration (MIC)

Bacteria were cultivated on agar for 24 h at 37 °C. Bacteria were suspended in 0.9% (w/v) sterile NaCl to obtain 0.5 McFarland. Pure media as well as media including bacteria suspension (1:50 dilution) were prepared. A serial compound dilution was prepared in 96 well plates, after which bacterial suspension was added. The plates were incubated for 24 h at 37 °C. The MIC was reported as the concentration where no bacterial growth was observable.

Growth curves and metabolic heat flow

For the evaluation of the growth profile of E. coli BW25113 WT and CYS^R mutants, strains were incubated overnight. Overnight cultures of the respective strains were inoculated in fresh LB-media with an OD₆₀₀ of 0.001 and 4 × 120 µL per strain was pipetted in a 96 well plate. The plate was incubated at 37 °C for 24 h. OD₆₀₀ measurement was done using a plate reader (Tecan) in kinetics mode. Same procedure was followed for microcalorimetry measurements using calScreener (Symcel). (Supplementary Fig. 11).

Genomic DNA analysis

Genomic DNA (gDNA) was isolated by phenol-chloroform extraction. Briefly, ONC was washed with 10 mL MilliQ water and suspended in 1.8 mL Tris-buffer (10 mM, pH=8.0). 100 µL RNase A (20 mg/mL) was added, followed by 30 min incubation at 37 °C before 200 µL proteinase K (20 mg/mL) with 20 µL SDS (20%) was added. This mixture was incubated additional 1.5 h at 55 °C. Afterwards, samples were poured to 2 mL phenol:chloroform:isoamylalcohol mix, and the supernatant was extracted multiple times before the gDNA was precipitated by the addition of 200 µL sodium acetate (3 M, pH = 4.8) and 5 mL ethanol (100%), followed by incubation overnight at −20 °C. Subsequently, the resulting pellet was washed using 5 mL 70% ethanol and dried, followed by reconstitution in 100 µL MilliQ water. Isolated gDNA was send for Illumina sequencing (MiSeq PE 300, PE 2 × 300 bp reads). Reads were analyzed using Geneious Prime (version 2022) by assembling to the E. coli K12 reference sequence (NC_000913). Consensus regions were generated and mauve alignment was completed comparing WT and mutant sequences, followed by manual SNP calling.

Full-proteome analysis

ONC of each strain was inoculated in 4× fresh media and incubated until early stationary phase. The samples were washed with 1 mL cold PBS, and the pellet was stored at −80 °C until lysis. Lysis was done by adding 210 µL 0.4% SDS in PBS and four rounds of sonication with an amplitude of 70% for 30 s (Bandelin Sonoplus). Subsequently, samples were centrifuged and supernatant was used for BCA assay (Thermo). Proteome concentrations of the samples were adjusted to 100 µg/200 µL per sample. Afterwards, the proteome was precipitated by adding 1 mL ice-cold acetone and incubation overnight at -20 °C. Next, the samples were centrifuged (16900 × g, 15 min, 4 °C), and the supernatant was discarded. Samples were washed two times with 1 mL ice-cold methanol. Preparation for protein digest was done by resuspension in 200 µL X-buffer (7 M urea, 2 M thiourea, 20 mM HEPES, pH = 7.5) followed by reduction of cysteine with 0.8 µL DTT (250 mM) (dithiothreitol) for 45 min at 25 °C, capping with 2 µL IAA (550 mM) (iodoacetic acid) for 30 min at 25 °C and adding additional 3.2 µL DTT (250 mM) for 30 min at 25 °C. 600 µL ammonium bicarbonate buffer (pH = 8.5) and trypsin (1 µg) (MS grade, Promega) were added for digest overnight. Digestion was stopped by adding 8 µL formic acid. Peptide samples were desalted on SepPak C18 columns (Waters) using the following protocol. Columns were primed by adding 1 mL 100% and 80% acetonitrile (ACN) with 0.5% FA, followed by equilibration with 3 × 1 mL H₂O + 0.1% TFA, before samples were loaded and washed with 3 × 0.1% TFA. Subsequently, samples were eluted into 2 mL LoBind (Eppendorf) by adding 3 × 250 µL 80% ACN + 0.5% FA and dried using a speedVac (Eppendorf) before dissolving in 1% FA with a proteome concentration of 1 µg/µL. Samples were filtered with Ultrafree centrifugal filters (Merck Millipore, UFC30GV0S) and transferred to QuanRecovery autosampler vials (Waters). Peptide samples were analyzed using an UltiMate 3000 nano HPLC system (Dionex) with an Acclaim C18 PepMap100 (75 µm ID x 2 cm) (Thermo) trap column and an Aurora Ultimate (25 cm × 75 µm ID, 1.6 µm FSC C18) (Ionoptics) separation column coupled to an Orbitrap Fusion (Thermo Fisher) in EASY-spray mode. 1 µL sample was loaded on the trap column and washed for 10 min with 0.1% TFA at 5 µL/min. Afterwards, peptides were transferred on the separation column, and analysis was done using a 132 min gradient (Buffer A: H2O + 0.1% FA; B: ACN + 0.1% FA) with a flow rate of 0.300 µL/min: in 7 min from 0% to 5% B, in 105 min from 5% to 22%, in 10 min from 22 to 35% and in another 10 min to 90% B. MS settings were adjusted as follows: ion transfer tube temperature at 275 °C, RF lens amplitude 60%, scan range from 300-1500 m/z, automatic gain control (AGC) target of 4.0 × 10⁵, 3 s cycle time and 50 ms maximal injection time. Lower intensity cut-off was set to 5.0 × 10³ and charge states between 2 and 7 were selected for fragmentation at 30% collision energy, before subsequent analysis in the ion trap using rapid scan rate. The isolation window was set to 1.6 m/z. AGC target was set to standard with a maximal injection time of 35 ms. Raw data were processed using MaxQuant (version 2.2.0.0) with E. coli K12 as reference proteome (Proteome ID: UP000000625). Subsequent analyses was performed using Perseus (version 2.0.5.0). GraphPad Prism (Version 9.1.1) was used for visual representation and further statistical analysis.

Motility assay

Motility agar (25 mL; 30 g/L tryptic soy broth, 3.4 g/L bacto agar, 9.0 g/L glucose) was poured in petri dishes and 5 µL of bacterial suspension with OD₆₀₀ = 1 was spotted in the middle of the plate. Plates were incubated overnight at 37 °C. Photos were taken with transmission light and a top view camera. (Supplementary Fig. 13)

Cloning

Forward primer_ygiV (5´ATATCATATGGAAAACCTGTATTTTCAGGGCATGACAAACCTGACACTGGA) and reverse primer_ygiV (5´TATAAAGCTTTCACGCCAACGGCACATAA) were designed based on sequenced WT DNA (Geneious Prime) followed by PCR to amplify the ygiV sequence from isolated gDNA. Subsequently, PCR product and pET-28b vector were restricted by NdeI and HindIII followed by ligation of both digests. Ligation mix was transformed into E. coli HS996 by electroporation to amplify the ygiV-carrying plasmid. Kanamycin was used as selectivity marker. The plasmid was isolated using midiprep as described by the manufacturer (Thermo). The plasmid was LGC-sequenced as quality control of the edit genetic code (Supplementary Fig. 14). The construct pET-28b-YgiV with a N-terminal His-Tag and TEV protease cleavage site (ENLYFGQ) was transformed into electrocompetent E. coli Lemo 21 and plated on agar plates containing both selection antibiotics kanamycin (50 µg/mL) and chloramphenicol (25 µg/mL). Preculture in LB medium supplemented with the appropriate antibiotics was inoculated with a single colony, and the flask was incubated overnight at 37 °C with shaking at 180 rpm. The following day, this culture was used to inoculate the main culture (1:100) in TB medium. The culture was incubated at 37 °C and 180 rpm until an OD₆₀₀ of 0.8 was reached, and protein expression was then induced with 1 mM IPTG and incubated at 18 °C and 180 rpm overnight. The following day, the cells were harvested by centrifugation at 12,800 × g at 4 °C for 15 min, and the cell pellet was flash frozen in liquid nitrogen and stored at −80 °C until further use.

Protein purification

YgiV and His-YgiV were purified using lysis buffer A (50 mM Tris, pH 8.0, 500 mM NaCl, 10% glycerol, 20 mM imidazole, and 2 mM BME) supplemented with DNAse (0.4 mg/g wet cell pellet, Sigma) and cOmplete EDTA-free protease inhibitor tablets (Roche). The bacterial cell pellet was resuspended in lysis buffer and lysed by passing it through a cell disruptor twice at 24.5 kPSI and cooling at 4 °C. Cell debris was removed by centrifugation at 35,000 × g at 4 °C for 30 min. The lysate was filtered through a filter with a pore size of 0.45 µm and loaded onto a HisTrap HP 5 mL column pre-equilibrated with buffer A using a Äkta pure at 4 °C. The column was then washed with five CV of buffer A, and protein elution was performed with 100% buffer B (50 mM Tris, pH 8.0, 500 mM NaCl, 10% glycerol, 300 mM imidazole, and 2 mM BME). The YgiV-containing fractions were collected, and an aliquot was concentrated using a 10 kDA cut-off Amicon concentrator. This aliquot was used to purify His-YgiV by loading it onto a Superdex S200 16/600 column pre-equilibrated with storage buffer C (30 mM HEPES, pH 8.0, 200 mM NaCl and 2 mM DTT). The protein fractions containing His-YgiV were pooled and concentrated to 1.5 mg/mL. The remaining YgiV protein sample after the HisTrap HP column was buffer-exchanged into buffer A without imidazole using a HiPrepTM 26/10 desalting column. Subsequently, digestion with TEV protease was performed at a ratio of 1:10 at 4 °C overnight to remove His-tag. At the following day the protein sample was loaded onto a HisTrap HP column pre-equilibrated with buffer A. The flow-through containing the YgiV protein was concentrated and applied to a Superdex S200 16/600 column pre-equilibrated with final storage buffer D (10 mM Tris pH 7.0, 150 mM NaCl, and 2 mM DTT). The protein fractions were pooled together, and YgiV was concentrated to 3.4 mg/mL (190 µM). Protein concentrations were determined by the specific absorption coefficient at A280 using a Nanodrop, and purity was checked by SDS-PAGE (Supplementary Fig. 15).

Gyrase supercoiling and cleavage assay

Inhibition of gyrase activity was examined using gyrase supercoiling activity kit (Inspriralis) as indicated by the manufacturer. Briefly, a compound dilution series was prepared in water. Water, dilution buffer, assay buffer, plasmid and enzyme were mixed with compound solution and incubated. YgiV was added in indicated amounts. A positive control was prepared without compound and a negative control was prepared without compound and enzyme. The reaction was stopped by adding chloroform/isoamylalcohol and STEB-buffer. The samples were vortexed and centrifuged before running gel-electrophoresis. The gel was stained in ethidium bromide and imaged using a Fusion Fx gel imager (Vilber Lourmat). Subsequent analysis was done using ImageJ and GraphPad Prism (Version 9.1.1). The same procedure was followed for the DNA cleavage assays, except that the samples did not contain gyrase but only YgiV and supercoiled or relaxed plasmid DNA, respectively (Inspiralis). Assays were independently repeated three times.

Degradation assay

Cystobactamid CN-861-2 [1] (20 µM) was incubated with YgiV (20 µM) in gyrase assay buffer (35 mM Tris·HCl (pH 7.5), 24 mM KCl, 4 mM MgCl₂, 2 mM DTT, 1.8 mM spermidine, 1 mM ATP, 6.5% (w/v) glycerol, and 0.1 mg/ml albumin) for 60 min at 37 °C. Afterwards, compound amount was measured using HPLC-UV/Vis-MS. (Supplementary Fig. 16).

Native protein mass spectrometry

YgiV (10 µM) was buffered in ammonium acetate (10 mM) and incubated overnight with CYS (CN-861-2 [1], CNDM-861 [2] or Cysto-180 [3]) (10 µM) at 4 °C. Afterwards, samples were measured using a solariX (Bruker) by direct infusion. Parameters for ionization were set as follows; flow rate: 5.0 µL/min, capillary voltage: 4500 V, nebulizer: 3 bar, dry gas: 5 L/min and dry temperature: 200 °C. The recorded mass window was set between 988.97 to 5000.00 m/z and 64 scans with an accumulation time of 1 s. Measurements were analyzed by DataAnalysis (Bruker, version 5.3) using maximum entropy deconvolution.

Microscale thermophoresis

A Monolith His-Tag Labeling Kit RED-tris-NTA 2^nd Generation (NanoTemper) was used for K_D determination as recommended by the manufacturer. Briefly, YgiV concentration (200 nM) was adjusted in assay buffer and labeled by RED-tris-NTA dye (100 nM) for 30 min. This mixture was added to CYS dilution (c_start = 10 µM) series and measured using Monolith NT.115 (NanoTemper) with standard glass capillaries (NanoTemper). The assay was independently repeated three times with CN-861-2 [1].

Affinity-based proteome profiling

Affinity enrichment in combination with competition by CN-861-2 [1] was done by inoculation of an ONC of M6 in 4 × 50 mL fresh LB-media and incubated until early stationary stage at 37 °C, 180 rpm. Bacteria were harvested (3000 × g, 15 min) and washed with PBS followed by resuspension at an OD₆₀₀ of 20. This suspension was treated with sub-MIC CN-861-2 [1] (25 µM) for competition or DMSO before adding Cysto-33 [5] photo-affinity probe (2.5 µM) or DMSO as control. Afterwards, samples were UV-irradiated for 10 min in 24 well plates, washed with 1 mL PBS and pellets were stored until lysis. Lysis and proteome adjustment was done as described above. Subsequently, the click reaction was performed by adding 3 µL biotin-azide (10 µM), 10 µL copper(II) sulfate (50 mM), 10 µL TCEP (Tris(2-carboxyethyl)phosphin –hydrochlorid) (15 µg/mL) and 3 µL THPTA (Tris((1-benzyl-4-triazolyl)methyl)amine) (10 mM) and incubating for 1 h at room temperature. Precipitation and washing were performed as described above. Samples were reconstituted in 0.2% SDS in PBS followed by avidin-bead enrichment (Merck) by coupling biotin labeled proteins to the avidin-agarose for 1 h under continuous mixing, followed by 3x washing steps with 1 mL 0.2% SDS in PBS, 2 × 1 mL urea (6 mM) and 3 × 1 mL PBS (centrifuge 3 min, 400 × g, RT). Subsequent protein digest, desalting and preparation for HPLC was done as described above. Sample analysis was done by using nanoElute nano flow liquid chromatography system (Bruker, Germany) coupled with a timsTOF Pro (Bruker, Germany). Samples were loaded to the trap column (Thermo Trap Cartridge 5 mm) and washed with 6 µL 0.1% FA with a flow rate of 10 µL/min. Peptides were then transferred to the analytical column (Aurora Ultimate CSI 25 cm×75 µm ID, 1.6 µm FSC C18, IonOpticks) and separated by a gradient elution (eluent A: H₂O + 0.1% FA, B: ACN + 0.1% FA; 0% to 3% in 1 min, 3% to 17% in 57 min, 17% to 25% in 21 min, 25% to 34% in 13 min, 34% to 85% in 1 min, 85% kept for 8 min) with a flow rate of 400 nL/min. A Captive Spray nanoESI source (Bruker, Germany) was used to ionize the peptides at 1.5 kV with 180 °C dry temperature at 3 L/min gas flow. The timsTOF Pro (Bruker, Germany) was operated in default dia-PASEF long gradient mode with TIMS set between 1/K0 0.6 Vs/cm2 and 1.6 Vs/cm2 with a ramp and accumulation time of 100 ms each and a ramp rate of 9.43 Hz. The mass range was set from 100.0 Da to 1700 Da with positive ion polarity. Dia-PASEF mass range was set to 400.0 Da to 1201.0 Da with a mobility range of 0.60 1/K0 to 1.43 1/K0 and a cycle time of 1.80 s. The collision energy for 0.60 1/K0 was set to 20.00 eV and for 1.6 1/K0 to 59.00 eV. Tuning MIX ESI-TOF (Agilent) was used for calibration of m/z and mobility. Data were processed using DIA-NN (version 1.8.1), and proteins were identified against Uniprot E. coli reference proteome (Proteome ID: UP000000625, downloaded 18/01/2023). Settings were used as default except precursor charge range was changed from 2 to 4. C carbamidomethylation was set as a fixed modification. “–relaxed-prot-inf” was added in additional options to allow further data processing with Perseus Software. In Perseus (version 2.0.5.0) the values were transformed to their log₂-value and the replicates were grouped. Missing values were imputed by the default settings and the differential protein abundance between different conditions were evaluated using two-tailed student’s t-test. The cut-off for –log₁₀ p-value was set to 1.3 (p-value = 0.05) and log₂ t-test difference >1.0. Proteins fitting these thresholds were seen as significantly over- or underexpressed compared to the wild type.

Data and statistics

Proteomics data were analyzed using Perseus software (version 2.0.5.0) with two-tailed student’s t-test to determine significant expression changes or enrichment as indicated. Dissociation constants (K_D) were identified using MO.Control (v1.6, NanoTemper) and MO.Affinity Analysis (v2.3, NanoTemper) software. GraphPad Prism (Version 9.1.1) was used for visual data representation and IC₅₀ determination. Statistical significance for direct comparisons was determined by GraphPad Prism (Version 9.1.1) using two-tailed, unpaired student´s t-test with statistical significance given to a p-value < 0.05.